diff options
Diffstat (limited to 'sys/contrib/openzfs/module')
124 files changed, 3619 insertions, 1715 deletions
diff --git a/sys/contrib/openzfs/module/Kbuild.in b/sys/contrib/openzfs/module/Kbuild.in index 667f061c6e16..3d6f288fa5da 100644 --- a/sys/contrib/openzfs/module/Kbuild.in +++ b/sys/contrib/openzfs/module/Kbuild.in @@ -406,6 +406,7 @@ ZFS_OBJS := \ zfs_byteswap.o \ zfs_chksum.o \ zfs_debug_common.o \ + zfs_crrd.o \ zfs_fm.o \ zfs_fuid.o \ zfs_impl.o \ @@ -494,3 +495,34 @@ UBSAN_SANITIZE_zfs/sa.o := n ifeq ($(CONFIG_ALTIVEC),y) $(obj)/zfs/vdev_raidz_math_powerpc_altivec.o : c_flags += -maltivec endif + +# The following recipes attempt to fix out of src-tree builds, where $(src) != $(obj), so that the +# subdir %.c/%.S -> %.o targets will work as expected. The in-kernel pattern targets do not seem to +# be working on subdirs since about ~6.10 +zobjdirs = $(dir $(zfs-objs)) $(dir $(spl-objs)) \ + $(dir $(zfs-$(CONFIG_X86))) $(dir $(zfs-$(CONFIG_UML_X86))) $(dir $(zfs-$(CONFIG_ARM64))) \ + $(dir $(zfs-$(CONFIG_PPC64))) $(dir $(zfs-$(CONFIG_PPC))) + +z_cdirs = $(sort $(filter-out lua/setjmp/ $(addprefix icp/asm-aarch64/, aes/ blake3/ modes/ sha2/) \ + $(addprefix icp/asm-x86_64/, aes/ blake3/ modes/ sha2/) \ + $(addprefix icp/asm-ppc/, aes/ blake3/ modes/ sha2/) \ + $(addprefix icp/asm-ppc64/, aes/ blake3/ modes/ sha2/), $(zobjdirs))) +z_sdirs = $(sort $(filter lua/setjmp/ $(addprefix icp/asm-aarch64/, aes/ blake3/ modes/ sha2/) \ + $(addprefix icp/asm-x86_64/, aes/ blake3/ modes/ sha2/) \ + $(addprefix icp/asm-ppc/, aes/ blake3/ modes/ sha2/) \ + $(addprefix icp/asm-ppc64/, aes/ blake3/ modes/ sha2/), $(zobjdirs))) + +define ZKMOD_C_O_MAKE_TARGET +$1%.o: $(src)/$1%.c FORCE + $$(call if_changed_rule,cc_o_c) + $$(call cmd,force_checksrc) +endef + +define ZKMOD_S_O_MAKE_TARGET +$1%.o: $(src)/$1%.S FORCE + $$(call if_changed_rule,as_o_S) + $$(call cmd,force_checksrc) +endef + +$(foreach target,$(z_cdirs), $(eval $(call ZKMOD_C_O_MAKE_TARGET,$(target)))) +$(foreach target,$(z_sdirs), $(eval $(call ZKMOD_S_O_MAKE_TARGET,$(target)))) diff --git a/sys/contrib/openzfs/module/Makefile.bsd b/sys/contrib/openzfs/module/Makefile.bsd index 7e7c3db73a43..3ba38c43f25b 100644 --- a/sys/contrib/openzfs/module/Makefile.bsd +++ b/sys/contrib/openzfs/module/Makefile.bsd @@ -217,6 +217,7 @@ SRCS+= abd_os.c \ vdev_label_os.c \ zfs_acl.c \ zfs_ctldir.c \ + zfs_crrd.c \ zfs_debug.c \ zfs_dir.c \ zfs_file_os.c \ diff --git a/sys/contrib/openzfs/module/Makefile.in b/sys/contrib/openzfs/module/Makefile.in index e9a268121762..859ba8649dd7 100644 --- a/sys/contrib/openzfs/module/Makefile.in +++ b/sys/contrib/openzfs/module/Makefile.in @@ -57,7 +57,7 @@ modules-Linux: $(if @KERNEL_LD@,LD=@KERNEL_LD@) $(if @KERNEL_LLVM@,LLVM=@KERNEL_LLVM@) \ $(if @KERNEL_CROSS_COMPILE@,CROSS_COMPILE=@KERNEL_CROSS_COMPILE@) \ $(if @KERNEL_ARCH@,ARCH=@KERNEL_ARCH@) \ - $(if @OBJTOOL_DISABLE_WERROR@,objtool=@top_builddir@/scripts/objtool-wrapper) \ + $(if @OBJTOOL_DISABLE_WERROR@,objtool=@abs_top_builddir@/scripts/objtool-wrapper) \ M="$$PWD" @KERNEL_MAKE@ CONFIG_ZFS=m modules modules-FreeBSD: diff --git a/sys/contrib/openzfs/module/avl/avl.c b/sys/contrib/openzfs/module/avl/avl.c index b6c1c02bc3f2..67cbcd3adeec 100644 --- a/sys/contrib/openzfs/module/avl/avl.c +++ b/sys/contrib/openzfs/module/avl/avl.c @@ -225,7 +225,7 @@ avl_nearest(avl_tree_t *tree, avl_index_t where, int direction) size_t off = tree->avl_offset; if (node == NULL) { - ASSERT(tree->avl_root == NULL); + ASSERT0P(tree->avl_root); return (NULL); } data = AVL_NODE2DATA(node, off); @@ -478,7 +478,7 @@ avl_insert(avl_tree_t *tree, void *new_data, avl_index_t where) size_t off = tree->avl_offset; #ifdef _LP64 - ASSERT(((uintptr_t)new_data & 0x7) == 0); + ASSERT0(((uintptr_t)new_data & 0x7)); #endif node = AVL_DATA2NODE(new_data, off); @@ -495,10 +495,10 @@ avl_insert(avl_tree_t *tree, void *new_data, avl_index_t where) AVL_SETBALANCE(node, 0); AVL_SETPARENT(node, parent); if (parent != NULL) { - ASSERT(parent->avl_child[which_child] == NULL); + ASSERT0P(parent->avl_child[which_child]); parent->avl_child[which_child] = node; } else { - ASSERT(tree->avl_root == NULL); + ASSERT0P(tree->avl_root); tree->avl_root = node; } /* @@ -608,7 +608,7 @@ avl_insert_here( ASSERT(diff > 0 ? child == 1 : child == 0); #endif } - ASSERT(node->avl_child[child] == NULL); + ASSERT0P(node->avl_child[child]); avl_insert(tree, new_data, AVL_MKINDEX(node, child)); } @@ -881,7 +881,7 @@ avl_create(avl_tree_t *tree, int (*compar) (const void *, const void *), ASSERT(size > 0); ASSERT(size >= offset + sizeof (avl_node_t)); #ifdef _LP64 - ASSERT((offset & 0x7) == 0); + ASSERT0((offset & 0x7)); #endif tree->avl_compar = compar; @@ -897,8 +897,8 @@ void avl_destroy(avl_tree_t *tree) { ASSERT(tree); - ASSERT(tree->avl_numnodes == 0); - ASSERT(tree->avl_root == NULL); + ASSERT0(tree->avl_numnodes); + ASSERT0P(tree->avl_root); } diff --git a/sys/contrib/openzfs/module/icp/core/kcf_sched.c b/sys/contrib/openzfs/module/icp/core/kcf_sched.c index 759f0d81d521..75e1052a4ed4 100644 --- a/sys/contrib/openzfs/module/icp/core/kcf_sched.c +++ b/sys/contrib/openzfs/module/icp/core/kcf_sched.c @@ -124,7 +124,7 @@ kcf_context_cache_destructor(void *buf, void *cdrarg) (void) cdrarg; kcf_context_t *kctx = (kcf_context_t *)buf; - ASSERT(kctx->kc_refcnt == 0); + ASSERT0(kctx->kc_refcnt); } void diff --git a/sys/contrib/openzfs/module/icp/io/aes.c b/sys/contrib/openzfs/module/icp/io/aes.c index ba703efa71fc..ca586eaf97ef 100644 --- a/sys/contrib/openzfs/module/icp/io/aes.c +++ b/sys/contrib/openzfs/module/icp/io/aes.c @@ -236,16 +236,16 @@ aes_encrypt_atomic(crypto_mechanism_t *mechanism, aes_xor_block); if (ret != CRYPTO_SUCCESS) goto out; - ASSERT(aes_ctx.ac_remainder_len == 0); + ASSERT0(aes_ctx.ac_remainder_len); } else if (mechanism->cm_type == AES_GCM_MECH_INFO_TYPE) { ret = gcm_encrypt_final((gcm_ctx_t *)&aes_ctx, ciphertext, AES_BLOCK_LEN, aes_encrypt_block, aes_copy_block, aes_xor_block); if (ret != CRYPTO_SUCCESS) goto out; - ASSERT(aes_ctx.ac_remainder_len == 0); + ASSERT0(aes_ctx.ac_remainder_len); } else { - ASSERT(aes_ctx.ac_remainder_len == 0); + ASSERT0(aes_ctx.ac_remainder_len); } if (plaintext != ciphertext) { @@ -337,7 +337,7 @@ aes_decrypt_atomic(crypto_mechanism_t *mechanism, ret = ccm_decrypt_final((ccm_ctx_t *)&aes_ctx, plaintext, AES_BLOCK_LEN, aes_encrypt_block, aes_copy_block, aes_xor_block); - ASSERT(aes_ctx.ac_remainder_len == 0); + ASSERT0(aes_ctx.ac_remainder_len); if ((ret == CRYPTO_SUCCESS) && (ciphertext != plaintext)) { plaintext->cd_length = @@ -349,7 +349,7 @@ aes_decrypt_atomic(crypto_mechanism_t *mechanism, ret = gcm_decrypt_final((gcm_ctx_t *)&aes_ctx, plaintext, AES_BLOCK_LEN, aes_encrypt_block, aes_xor_block); - ASSERT(aes_ctx.ac_remainder_len == 0); + ASSERT0(aes_ctx.ac_remainder_len); if ((ret == CRYPTO_SUCCESS) && (ciphertext != plaintext)) { plaintext->cd_length = diff --git a/sys/contrib/openzfs/module/nvpair/nvpair.c b/sys/contrib/openzfs/module/nvpair/nvpair.c index 811cfc87d7a4..eb8c14b4a783 100644 --- a/sys/contrib/openzfs/module/nvpair/nvpair.c +++ b/sys/contrib/openzfs/module/nvpair/nvpair.c @@ -265,7 +265,7 @@ nv_priv_alloc_embedded(nvpriv_t *priv) static int nvt_tab_alloc(nvpriv_t *priv, uint64_t buckets) { - ASSERT3P(priv->nvp_hashtable, ==, NULL); + ASSERT0P(priv->nvp_hashtable); ASSERT0(priv->nvp_nbuckets); ASSERT0(priv->nvp_nentries); @@ -334,7 +334,7 @@ nvt_lookup_name_type(const nvlist_t *nvl, const char *name, data_type_t type) i_nvp_t **tab = priv->nvp_hashtable; if (tab == NULL) { - ASSERT3P(priv->nvp_list, ==, NULL); + ASSERT0P(priv->nvp_list); ASSERT0(priv->nvp_nbuckets); ASSERT0(priv->nvp_nentries); return (NULL); @@ -540,7 +540,7 @@ nvt_add_nvpair(nvlist_t *nvl, nvpair_t *nvp) /* insert link at the beginning of the bucket */ i_nvp_t *new_entry = NVPAIR2I_NVP(nvp); - ASSERT3P(new_entry->nvi_hashtable_next, ==, NULL); + ASSERT0P(new_entry->nvi_hashtable_next); new_entry->nvi_hashtable_next = bucket; // cppcheck-suppress nullPointerRedundantCheck tab[index] = new_entry; diff --git a/sys/contrib/openzfs/module/os/freebsd/spl/spl_kmem.c b/sys/contrib/openzfs/module/os/freebsd/spl/spl_kmem.c index 6d198fad5203..ae6e36d988c2 100644 --- a/sys/contrib/openzfs/module/os/freebsd/spl/spl_kmem.c +++ b/sys/contrib/openzfs/module/os/freebsd/spl/spl_kmem.c @@ -160,7 +160,7 @@ kmem_cache_create(const char *name, size_t bufsize, size_t align, { kmem_cache_t *cache; - ASSERT3P(vmp, ==, NULL); + ASSERT0P(vmp); cache = kmem_alloc(sizeof (*cache), KM_SLEEP); strlcpy(cache->kc_name, name, sizeof (cache->kc_name)); diff --git a/sys/contrib/openzfs/module/os/freebsd/spl/spl_misc.c b/sys/contrib/openzfs/module/os/freebsd/spl/spl_misc.c index f9125a067cd1..3f360d167b17 100644 --- a/sys/contrib/openzfs/module/os/freebsd/spl/spl_misc.c +++ b/sys/contrib/openzfs/module/os/freebsd/spl/spl_misc.c @@ -101,6 +101,15 @@ spl_panic(const char *file, const char *func, int line, const char *fmt, ...) va_end(ap); } +/* + * Check if the current thread is a memory reclaim thread. + * Returns true if curproc is pageproc (FreeBSD's page daemon). + */ +int +current_is_reclaim_thread(void) +{ + return (curproc == pageproc); +} SYSINIT(opensolaris_utsname_init, SI_SUB_TUNABLES, SI_ORDER_ANY, opensolaris_utsname_init, NULL); diff --git a/sys/contrib/openzfs/module/os/freebsd/spl/spl_sysevent.c b/sys/contrib/openzfs/module/os/freebsd/spl/spl_sysevent.c index 9da633c2b1be..3c2d39b20c09 100644 --- a/sys/contrib/openzfs/module/os/freebsd/spl/spl_sysevent.c +++ b/sys/contrib/openzfs/module/os/freebsd/spl/spl_sysevent.c @@ -256,7 +256,7 @@ sysevent_worker(void *arg __unused) * free `ze`, so just inline the free() here -- events have already * been drained. */ - VERIFY3P(ze->ze_zevent, ==, NULL); + VERIFY0P(ze->ze_zevent); kmem_free(ze, sizeof (zfs_zevent_t)); kthread_exit(); diff --git a/sys/contrib/openzfs/module/os/freebsd/spl/spl_vm.c b/sys/contrib/openzfs/module/os/freebsd/spl/spl_vm.c index 733c2bd07ebb..9d5f025423a1 100644 --- a/sys/contrib/openzfs/module/os/freebsd/spl/spl_vm.c +++ b/sys/contrib/openzfs/module/os/freebsd/spl/spl_vm.c @@ -43,6 +43,7 @@ const int zfs_vm_pagerret_bad = VM_PAGER_BAD; const int zfs_vm_pagerret_error = VM_PAGER_ERROR; const int zfs_vm_pagerret_ok = VM_PAGER_OK; +const int zfs_vm_pagerret_pend = VM_PAGER_PEND; const int zfs_vm_pagerput_sync = VM_PAGER_PUT_SYNC; const int zfs_vm_pagerput_inval = VM_PAGER_PUT_INVAL; diff --git a/sys/contrib/openzfs/module/os/freebsd/zfs/abd_os.c b/sys/contrib/openzfs/module/os/freebsd/zfs/abd_os.c index fbf67f6a14a8..4bf487cdc469 100644 --- a/sys/contrib/openzfs/module/os/freebsd/zfs/abd_os.c +++ b/sys/contrib/openzfs/module/os/freebsd/zfs/abd_os.c @@ -507,7 +507,7 @@ abd_iter_at_end(struct abd_iter *aiter) void abd_iter_advance(struct abd_iter *aiter, size_t amount) { - ASSERT3P(aiter->iter_mapaddr, ==, NULL); + ASSERT0P(aiter->iter_mapaddr); ASSERT0(aiter->iter_mapsize); /* There's nothing left to advance to, so do nothing */ @@ -526,7 +526,7 @@ abd_iter_map(struct abd_iter *aiter) { void *paddr; - ASSERT3P(aiter->iter_mapaddr, ==, NULL); + ASSERT0P(aiter->iter_mapaddr); ASSERT0(aiter->iter_mapsize); /* There's nothing left to iterate over, so do nothing */ diff --git a/sys/contrib/openzfs/module/os/freebsd/zfs/dmu_os.c b/sys/contrib/openzfs/module/os/freebsd/zfs/dmu_os.c index 364bbfc60abd..26cc7981bfcd 100644 --- a/sys/contrib/openzfs/module/os/freebsd/zfs/dmu_os.c +++ b/sys/contrib/openzfs/module/os/freebsd/zfs/dmu_os.c @@ -156,7 +156,7 @@ dmu_read_pages(objset_t *os, uint64_t object, vm_page_t *ma, int count, if (dbp[0]->db_offset != 0 || numbufs > 1) { for (i = 0; i < numbufs; i++) { ASSERT(ISP2(dbp[i]->db_size)); - ASSERT3U((dbp[i]->db_offset % dbp[i]->db_size), ==, 0); + ASSERT0((dbp[i]->db_offset % dbp[i]->db_size)); ASSERT3U(dbp[i]->db_size, ==, dbp[0]->db_size); } } @@ -175,7 +175,7 @@ dmu_read_pages(objset_t *os, uint64_t object, vm_page_t *ma, int count, vm_page_sunbusy(m); break; } - ASSERT3U(m->dirty, ==, 0); + ASSERT0(m->dirty); ASSERT(!pmap_page_is_write_mapped(m)); ASSERT3U(db->db_size, >, PAGE_SIZE); @@ -201,7 +201,7 @@ dmu_read_pages(objset_t *os, uint64_t object, vm_page_t *ma, int count, if (m != bogus_page) { vm_page_assert_xbusied(m); ASSERT(vm_page_none_valid(m)); - ASSERT3U(m->dirty, ==, 0); + ASSERT0(m->dirty); ASSERT(!pmap_page_is_write_mapped(m)); va = zfs_map_page(m, &sf); } @@ -295,7 +295,7 @@ dmu_read_pages(objset_t *os, uint64_t object, vm_page_t *ma, int count, vm_page_sunbusy(m); break; } - ASSERT3U(m->dirty, ==, 0); + ASSERT0(m->dirty); ASSERT(!pmap_page_is_write_mapped(m)); ASSERT3U(db->db_size, >, PAGE_SIZE); diff --git a/sys/contrib/openzfs/module/os/freebsd/zfs/vdev_geom.c b/sys/contrib/openzfs/module/os/freebsd/zfs/vdev_geom.c index c8ab7cc7cf8e..bbd1dafc69be 100644 --- a/sys/contrib/openzfs/module/os/freebsd/zfs/vdev_geom.c +++ b/sys/contrib/openzfs/module/os/freebsd/zfs/vdev_geom.c @@ -1236,7 +1236,7 @@ vdev_geom_io_done(zio_t *zio) struct bio *bp = zio->io_bio; if (zio->io_type != ZIO_TYPE_READ && zio->io_type != ZIO_TYPE_WRITE) { - ASSERT3P(bp, ==, NULL); + ASSERT0P(bp); return; } diff --git a/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_acl.c b/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_acl.c index 334264f6da2f..b15a3e6e38c0 100644 --- a/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_acl.c +++ b/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_acl.c @@ -1632,7 +1632,7 @@ zfs_acl_ids_create(znode_t *dzp, int flag, vattr_t *vap, cred_t *cr, if (zfsvfs->z_replay == B_FALSE) ASSERT_VOP_ELOCKED(ZTOV(dzp), __func__); } else - ASSERT3P(dzp->z_vnode, ==, NULL); + ASSERT0P(dzp->z_vnode); memset(acl_ids, 0, sizeof (zfs_acl_ids_t)); acl_ids->z_mode = MAKEIMODE(vap->va_type, vap->va_mode); @@ -2014,7 +2014,7 @@ top: error = zfs_aclset_common(zp, aclp, cr, tx); ASSERT0(error); - ASSERT3P(zp->z_acl_cached, ==, NULL); + ASSERT0P(zp->z_acl_cached); zp->z_acl_cached = aclp; if (fuid_dirtied) @@ -2357,10 +2357,42 @@ zfs_zaccess(znode_t *zp, int mode, int flags, boolean_t skipaclchk, cred_t *cr, * In FreeBSD, we don't care about permissions of individual ADS. * Note that not checking them is not just an optimization - without * this shortcut, EA operations may bogusly fail with EACCES. + * + * If this is a named attribute lookup, do the checks. */ +#if __FreeBSD_version >= 1500040 + if ((zp->z_pflags & ZFS_XATTR) && (flags & V_NAMEDATTR) == 0) +#else if (zp->z_pflags & ZFS_XATTR) +#endif return (0); + /* + * If a named attribute directory then validate against base file + */ + if (is_attr) { + if ((error = zfs_zget(ZTOZSB(zp), + zp->z_xattr_parent, &xzp)) != 0) { + return (error); + } + + check_zp = xzp; + + /* + * fixup mode to map to xattr perms + */ + + if (mode & (ACE_WRITE_DATA|ACE_APPEND_DATA)) { + mode &= ~(ACE_WRITE_DATA|ACE_APPEND_DATA); + mode |= ACE_WRITE_NAMED_ATTRS; + } + + if (mode & (ACE_READ_DATA|ACE_EXECUTE)) { + mode &= ~(ACE_READ_DATA|ACE_EXECUTE); + mode |= ACE_READ_NAMED_ATTRS; + } + } + owner = zfs_fuid_map_id(zp->z_zfsvfs, zp->z_uid, cr, ZFS_OWNER); /* diff --git a/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_ctldir.c b/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_ctldir.c index 8d0ff9b25e30..61d0bb26d1e5 100644 --- a/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_ctldir.c +++ b/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_ctldir.c @@ -357,7 +357,7 @@ zfsctl_create(zfsvfs_t *zfsvfs) vnode_t *rvp; uint64_t crtime[2]; - ASSERT3P(zfsvfs->z_ctldir, ==, NULL); + ASSERT0P(zfsvfs->z_ctldir); snapdir = sfs_alloc_node(sizeof (*snapdir), "snapshot", ZFSCTL_INO_ROOT, ZFSCTL_INO_SNAPDIR); @@ -1367,7 +1367,7 @@ zfsctl_snapshot_unmount(const char *snapname, int flags __unused) int err = getzfsvfs(snapname, &zfsvfs); if (err != 0) { - ASSERT3P(zfsvfs, ==, NULL); + ASSERT0P(zfsvfs); return (0); } vfsp = zfsvfs->z_vfs; diff --git a/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_dir.c b/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_dir.c index 191df832d726..75ba2ea0cb9e 100644 --- a/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_dir.c +++ b/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_dir.c @@ -273,7 +273,7 @@ zfs_unlinked_add(znode_t *zp, dmu_tx_t *tx) zfsvfs_t *zfsvfs = zp->z_zfsvfs; ASSERT(zp->z_unlinked); - ASSERT3U(zp->z_links, ==, 0); + ASSERT0(zp->z_links); VERIFY0(zap_add_int(zfsvfs->z_os, zfsvfs->z_unlinkedobj, zp->z_id, tx)); @@ -437,7 +437,7 @@ zfs_rmnode(znode_t *zp) uint64_t count; int error; - ASSERT3U(zp->z_links, ==, 0); + ASSERT0(zp->z_links); if (zfsvfs->z_replay == B_FALSE) ASSERT_VOP_ELOCKED(ZTOV(zp), __func__); diff --git a/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_vfsops.c b/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_vfsops.c index 493ac9f69ad4..79b784288911 100644 --- a/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_vfsops.c +++ b/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_vfsops.c @@ -455,8 +455,13 @@ zfs_sync(vfs_t *vfsp, int waitfor) return (0); } - if (zfsvfs->z_log != NULL) - zil_commit(zfsvfs->z_log, 0); + if (zfsvfs->z_log != NULL) { + error = zil_commit(zfsvfs->z_log, 0); + if (error != 0) { + zfs_exit(zfsvfs, FTAG); + return (error); + } + } zfs_exit(zfsvfs, FTAG); } else { @@ -1091,7 +1096,7 @@ zfsvfs_setup(zfsvfs_t *zfsvfs, boolean_t mounting) if (mounting) { boolean_t readonly; - ASSERT3P(zfsvfs->z_kstat.dk_kstats, ==, NULL); + ASSERT0P(zfsvfs->z_kstat.dk_kstats); error = dataset_kstats_create(&zfsvfs->z_kstat, zfsvfs->z_os); if (error) return (error); @@ -1209,6 +1214,8 @@ zfs_set_fuid_feature(zfsvfs_t *zfsvfs) zfsvfs->z_use_sa = USE_SA(zfsvfs->z_version, zfsvfs->z_os); } +extern int zfs_xattr_compat; + static int zfs_domount(vfs_t *vfsp, char *osname) { @@ -1289,6 +1296,16 @@ zfs_domount(vfs_t *vfsp, char *osname) goto out; } +#if __FreeBSD_version >= 1500040 + /* + * Named attributes can only work if the xattr property is set to + * on/dir and not sa. Also, zfs_xattr_compat must be set. + */ + if ((zfsvfs->z_flags & ZSB_XATTR) != 0 && !zfsvfs->z_xattr_sa && + zfs_xattr_compat) + vfsp->mnt_flag |= MNT_NAMEDATTR; +#endif + vfs_mountedfrom(vfsp, osname); if (!zfsvfs->z_issnap) @@ -1812,6 +1829,14 @@ zfs_vget(vfs_t *vfsp, ino_t ino, int flags, vnode_t **vpp) err = vn_lock(*vpp, flags); if (err != 0) vrele(*vpp); +#if __FreeBSD_version >= 1500040 + else if ((zp->z_pflags & ZFS_XATTR) != 0) { + if ((*vpp)->v_type == VDIR) + vn_irflag_set_cond(*vpp, VIRF_NAMEDDIR); + else + vn_irflag_set_cond(*vpp, VIRF_NAMEDATTR); + } +#endif } if (err != 0) *vpp = NULL; @@ -1964,9 +1989,17 @@ zfs_fhtovp(vfs_t *vfsp, fid_t *fidp, int flags, vnode_t **vpp) *vpp = ZTOV(zp); zfs_exit(zfsvfs, FTAG); err = vn_lock(*vpp, flags); - if (err == 0) + if (err == 0) { vnode_create_vobject(*vpp, zp->z_size, curthread); - else +#if __FreeBSD_version >= 1500040 + if ((zp->z_pflags & ZFS_XATTR) != 0) { + if ((*vpp)->v_type == VDIR) + vn_irflag_set_cond(*vpp, VIRF_NAMEDDIR); + else + vn_irflag_set_cond(*vpp, VIRF_NAMEDATTR); + } +#endif + } else *vpp = NULL; return (err); } diff --git a/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_vnops_os.c b/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_vnops_os.c index 8a5006c488f3..1813c411b013 100644 --- a/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_vnops_os.c +++ b/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_vnops_os.c @@ -25,6 +25,7 @@ * Copyright (c) 2012, 2015 by Delphix. All rights reserved. * Copyright (c) 2014 Integros [integros.com] * Copyright 2017 Nexenta Systems, Inc. + * Copyright (c) 2025, Klara, Inc. */ /* Portions Copyright 2007 Jeremy Teo */ @@ -114,6 +115,8 @@ typedef uint64_t cookie_t; typedef ulong_t cookie_t; #endif +static int zfs_check_attrname(const char *name); + /* * Programming rules. * @@ -813,7 +816,12 @@ zfs_lookup(vnode_t *dvp, const char *nm, vnode_t **vpp, /* * Do we have permission to get into attribute directory? */ - error = zfs_zaccess(zp, ACE_EXECUTE, 0, B_FALSE, cr, NULL); + if (flags & LOOKUP_NAMED_ATTR) + error = zfs_zaccess(zp, ACE_EXECUTE, V_NAMEDATTR, + B_FALSE, cr, NULL); + else + error = zfs_zaccess(zp, ACE_EXECUTE, 0, B_FALSE, cr, + NULL); if (error) { vrele(ZTOV(zp)); } @@ -1093,7 +1101,7 @@ zfs_create(znode_t *dzp, const char *name, vattr_t *vap, int excl, int mode, zfs_exit(zfsvfs, FTAG); return (error); } - ASSERT3P(zp, ==, NULL); + ASSERT0P(zp); /* * Create a new file object and update the directory @@ -1185,8 +1193,8 @@ out: *zpp = zp; } - if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) - zil_commit(zilog, 0); + if (error == 0 && zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) + error = zil_commit(zilog, 0); zfs_exit(zfsvfs, FTAG); return (error); @@ -1315,9 +1323,8 @@ out: if (xzp) vrele(ZTOV(xzp)); - if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) - zil_commit(zilog, 0); - + if (error == 0 && zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) + error = zil_commit(zilog, 0); zfs_exit(zfsvfs, FTAG); return (error); @@ -1474,7 +1481,7 @@ zfs_mkdir(znode_t *dzp, const char *dirname, vattr_t *vap, znode_t **zpp, zfs_exit(zfsvfs, FTAG); return (error); } - ASSERT3P(zp, ==, NULL); + ASSERT0P(zp); if ((error = zfs_zaccess(dzp, ACE_ADD_SUBDIRECTORY, 0, B_FALSE, cr, mnt_ns))) { @@ -1548,8 +1555,8 @@ out: getnewvnode_drop_reserve(); - if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) - zil_commit(zilog, 0); + if (error == 0 && zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) + error = zil_commit(zilog, 0); zfs_exit(zfsvfs, FTAG); return (error); @@ -1629,8 +1636,8 @@ zfs_rmdir_(vnode_t *dvp, vnode_t *vp, const char *name, cred_t *cr) if (zfsvfs->z_use_namecache) cache_vop_rmdir(dvp, vp); out: - if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) - zil_commit(zilog, 0); + if (error == 0 && zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) + error = zil_commit(zilog, 0); zfs_exit(zfsvfs, FTAG); return (error); @@ -3001,8 +3008,8 @@ out: } out2: - if (os->os_sync == ZFS_SYNC_ALWAYS) - zil_commit(zilog, 0); + if (err == 0 && os->os_sync == ZFS_SYNC_ALWAYS) + err = zil_commit(zilog, 0); zfs_exit(zfsvfs, FTAG); return (err); @@ -3531,7 +3538,7 @@ out_seq: out: if (error == 0 && zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) - zil_commit(zilog, 0); + error = zil_commit(zilog, 0); zfs_exit(zfsvfs, FTAG); return (error); @@ -3723,7 +3730,7 @@ zfs_symlink(znode_t *dzp, const char *name, vattr_t *vap, *zpp = zp; if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) - zil_commit(zilog, 0); + error = zil_commit(zilog, 0); } zfs_exit(zfsvfs, FTAG); @@ -3913,8 +3920,8 @@ zfs_link(znode_t *tdzp, znode_t *szp, const char *name, cred_t *cr, vnevent_link(ZTOV(szp), ct); } - if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) - zil_commit(zilog, 0); + if (error == 0 && zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) + error = zil_commit(zilog, 0); zfs_exit(zfsvfs, FTAG); return (error); @@ -4299,6 +4306,43 @@ zfs_freebsd_getpages(struct vop_getpages_args *ap) ap->a_rahead)); } +typedef struct { + uint_t pca_npages; + vm_page_t pca_pages[]; +} putpage_commit_arg_t; + +static void +zfs_putpage_commit_cb(void *arg, int err) +{ + putpage_commit_arg_t *pca = arg; + vm_object_t object = pca->pca_pages[0]->object; + + zfs_vmobject_wlock(object); + + for (uint_t i = 0; i < pca->pca_npages; i++) { + vm_page_t pp = pca->pca_pages[i]; + + if (err == 0) { + /* + * Writeback succeeded, so undirty the page. If it + * fails, we leave it in the same state it was. That's + * most likely dirty, so it will get tried again some + * other time. + */ + vm_page_undirty(pp); + } + + vm_page_sunbusy(pp); + } + + vm_object_pip_wakeupn(object, pca->pca_npages); + + zfs_vmobject_wunlock(object); + + kmem_free(pca, + offsetof(putpage_commit_arg_t, pca_pages[pca->pca_npages])); +} + static int zfs_putpages(struct vnode *vp, vm_page_t *ma, size_t len, int flags, int *rtvals) @@ -4400,10 +4444,12 @@ zfs_putpages(struct vnode *vp, vm_page_t *ma, size_t len, int flags, } if (zp->z_blksz < PAGE_SIZE) { - for (i = 0; len > 0; off += tocopy, len -= tocopy, i++) { - tocopy = len > PAGE_SIZE ? PAGE_SIZE : len; + vm_ooffset_t woff = off; + size_t wlen = len; + for (i = 0; wlen > 0; woff += tocopy, wlen -= tocopy, i++) { + tocopy = MIN(PAGE_SIZE, wlen); va = zfs_map_page(ma[i], &sf); - dmu_write(zfsvfs->z_os, zp->z_id, off, tocopy, va, tx); + dmu_write(zfsvfs->z_os, zp->z_id, woff, tocopy, va, tx); zfs_unmap_page(sf); } } else { @@ -4424,19 +4470,48 @@ zfs_putpages(struct vnode *vp, vm_page_t *ma, size_t len, int flags, zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime); err = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); ASSERT0(err); - /* - * XXX we should be passing a callback to undirty - * but that would make the locking messier - */ - zfs_log_write(zfsvfs->z_log, tx, TX_WRITE, zp, off, - len, commit, B_FALSE, NULL, NULL); - zfs_vmobject_wlock(object); - for (i = 0; i < ncount; i++) { - rtvals[i] = zfs_vm_pagerret_ok; - vm_page_undirty(ma[i]); + if (commit) { + /* + * Caller requested that we commit immediately. We set + * a callback on the log entry, to be called once its + * on disk after the call to zil_commit() below. The + * pages will be undirtied and unbusied there. + */ + putpage_commit_arg_t *pca = kmem_alloc( + offsetof(putpage_commit_arg_t, pca_pages[ncount]), + KM_SLEEP); + pca->pca_npages = ncount; + memcpy(pca->pca_pages, ma, sizeof (vm_page_t) * ncount); + + zfs_log_write(zfsvfs->z_log, tx, TX_WRITE, zp, off, len, + B_TRUE, B_FALSE, zfs_putpage_commit_cb, pca); + + for (i = 0; i < ncount; i++) + rtvals[i] = zfs_vm_pagerret_pend; + } else { + /* + * Caller just wants the page written back somewhere, + * but doesn't need it committed yet. We've already + * written it back to the DMU, so we just need to put + * it on the async log, then undirty the page and + * return. + * + * We cannot use a callback here, because it would keep + * the page busy (locked) until it is eventually + * written down at txg sync. + */ + zfs_log_write(zfsvfs->z_log, tx, TX_WRITE, zp, off, len, + B_FALSE, B_FALSE, NULL, NULL); + + zfs_vmobject_wlock(object); + for (i = 0; i < ncount; i++) { + rtvals[i] = zfs_vm_pagerret_ok; + vm_page_undirty(ma[i]); + } + zfs_vmobject_wunlock(object); } - zfs_vmobject_wunlock(object); + VM_CNT_INC(v_vnodeout); VM_CNT_ADD(v_vnodepgsout, ncount); } @@ -4444,8 +4519,13 @@ zfs_putpages(struct vnode *vp, vm_page_t *ma, size_t len, int flags, out: zfs_rangelock_exit(lr); - if (commit) - zil_commit(zfsvfs->z_log, zp->z_id); + if (commit) { + err = zil_commit(zfsvfs->z_log, zp->z_id); + if (err != 0) { + zfs_exit(zfsvfs, FTAG); + return (err); + } + } dataset_kstats_update_write_kstats(&zfsvfs->z_kstat, len); @@ -4707,8 +4787,16 @@ zfs_freebsd_access(struct vop_access_args *ap) * ZFS itself only knowns about VREAD, VWRITE, VEXEC and VAPPEND, */ accmode = ap->a_accmode & (VREAD|VWRITE|VEXEC|VAPPEND); - if (accmode != 0) - error = zfs_access(zp, accmode, 0, ap->a_cred); + if (accmode != 0) { +#if __FreeBSD_version >= 1500040 + /* For named attributes, do the checks. */ + if ((vn_irflag_read(vp) & VIRF_NAMEDATTR) != 0) + error = zfs_access(zp, accmode, V_NAMEDATTR, + ap->a_cred); + else +#endif + error = zfs_access(zp, accmode, 0, ap->a_cred); + } /* * VADMIN has to be handled by vaccess(). @@ -4741,6 +4829,190 @@ struct vop_lookup_args { }; #endif +#if __FreeBSD_version >= 1500040 +static int +zfs_lookup_nameddir(struct vnode *dvp, struct componentname *cnp, + struct vnode **vpp) +{ + struct vnode *xvp; + int error, flags; + + *vpp = NULL; + flags = LOOKUP_XATTR | LOOKUP_NAMED_ATTR; + if ((cnp->cn_flags & CREATENAMED) != 0) + flags |= CREATE_XATTR_DIR; + error = zfs_lookup(dvp, NULL, &xvp, NULL, 0, cnp->cn_cred, flags, + B_FALSE); + if (error == 0) { + if ((cnp->cn_flags & LOCKLEAF) != 0) + error = vn_lock(xvp, cnp->cn_lkflags); + if (error == 0) { + vn_irflag_set_cond(xvp, VIRF_NAMEDDIR); + *vpp = xvp; + } else { + vrele(xvp); + } + } + return (error); +} + +static ssize_t +zfs_readdir_named(struct vnode *vp, char *buf, ssize_t blen, off_t *offp, + int *eofflagp, struct ucred *cred, struct thread *td) +{ + struct uio io; + struct iovec iv; + zfs_uio_t uio; + int error; + + io.uio_offset = *offp; + io.uio_segflg = UIO_SYSSPACE; + io.uio_rw = UIO_READ; + io.uio_td = td; + iv.iov_base = buf; + iv.iov_len = blen; + io.uio_iov = &iv; + io.uio_iovcnt = 1; + io.uio_resid = blen; + zfs_uio_init(&uio, &io); + error = zfs_readdir(vp, &uio, cred, eofflagp, NULL, NULL); + if (error != 0) + return (-1); + *offp = io.uio_offset; + return (blen - io.uio_resid); +} + +static bool +zfs_has_namedattr(struct vnode *vp, struct ucred *cred) +{ + struct componentname cn; + struct vnode *xvp; + struct dirent *dp; + off_t offs; + ssize_t rsize; + char *buf, *cp, *endcp; + int eofflag, error; + bool ret; + + MNT_ILOCK(vp->v_mount); + if ((vp->v_mount->mnt_flag & MNT_NAMEDATTR) == 0) { + MNT_IUNLOCK(vp->v_mount); + return (false); + } + MNT_IUNLOCK(vp->v_mount); + + /* Now see if a named attribute directory exists. */ + cn.cn_flags = LOCKLEAF; + cn.cn_lkflags = LK_SHARED; + cn.cn_cred = cred; + error = zfs_lookup_nameddir(vp, &cn, &xvp); + if (error != 0) + return (false); + + /* It exists, so see if there is any entry other than "." and "..". */ + buf = malloc(DEV_BSIZE, M_TEMP, M_WAITOK); + ret = false; + offs = 0; + do { + rsize = zfs_readdir_named(xvp, buf, DEV_BSIZE, &offs, &eofflag, + cred, curthread); + if (rsize <= 0) + break; + cp = buf; + endcp = &buf[rsize]; + while (cp < endcp) { + dp = (struct dirent *)cp; + if (dp->d_fileno != 0 && (dp->d_type == DT_REG || + dp->d_type == DT_UNKNOWN) && + !ZFS_XA_NS_PREFIX_FORBIDDEN(dp->d_name) && + ((dp->d_namlen == 1 && dp->d_name[0] != '.') || + (dp->d_namlen == 2 && (dp->d_name[0] != '.' || + dp->d_name[1] != '.')) || dp->d_namlen > 2)) { + ret = true; + break; + } + cp += dp->d_reclen; + } + } while (!ret && rsize > 0 && eofflag == 0); + vput(xvp); + free(buf, M_TEMP); + return (ret); +} + +static int +zfs_freebsd_lookup(struct vop_lookup_args *ap, boolean_t cached) +{ + struct componentname *cnp = ap->a_cnp; + char nm[NAME_MAX + 1]; + int error; + struct vnode **vpp = ap->a_vpp, *dvp = ap->a_dvp, *xvp; + bool is_nameddir, needs_nameddir, opennamed = false; + + /* + * These variables are used to handle the named attribute cases: + * opennamed - Is true when this is a call from open with O_NAMEDATTR + * specified and it is the last component. + * is_nameddir - Is true when the directory is a named attribute dir. + * needs_nameddir - Is set when the lookup needs to look for/create + * a named attribute directory. It is only set when is_nameddir + * is_nameddir is false and opennamed is true. + * xvp - Is the directory that the lookup needs to be done in. + * Usually dvp, unless needs_nameddir is true where it is the + * result of the first non-named directory lookup. + * Note that name caching must be disabled for named attribute + * handling. + */ + needs_nameddir = false; + xvp = dvp; + opennamed = (cnp->cn_flags & (OPENNAMED | ISLASTCN)) == + (OPENNAMED | ISLASTCN); + is_nameddir = (vn_irflag_read(dvp) & VIRF_NAMEDDIR) != 0; + if (is_nameddir && (cnp->cn_flags & ISLASTCN) == 0) + return (ENOATTR); + if (opennamed && !is_nameddir && (cnp->cn_flags & ISDOTDOT) != 0) + return (ENOATTR); + if (opennamed || is_nameddir) + cnp->cn_flags &= ~MAKEENTRY; + if (opennamed && !is_nameddir) + needs_nameddir = true; + ASSERT3U(cnp->cn_namelen, <, sizeof (nm)); + error = 0; + *vpp = NULL; + if (needs_nameddir) { + if (VOP_ISLOCKED(dvp) != LK_EXCLUSIVE) + vn_lock(dvp, LK_UPGRADE | LK_RETRY); + error = zfs_lookup_nameddir(dvp, cnp, &xvp); + if (error == 0) + is_nameddir = true; + } + if (error == 0) { + if (!needs_nameddir || cnp->cn_namelen != 1 || + *cnp->cn_nameptr != '.') { + strlcpy(nm, cnp->cn_nameptr, MIN(cnp->cn_namelen + 1, + sizeof (nm))); + error = zfs_lookup(xvp, nm, vpp, cnp, cnp->cn_nameiop, + cnp->cn_cred, 0, cached); + if (is_nameddir && error == 0 && + (cnp->cn_namelen != 1 || *cnp->cn_nameptr != '.') && + (cnp->cn_flags & ISDOTDOT) == 0) { + if ((*vpp)->v_type == VDIR) + vn_irflag_set_cond(*vpp, VIRF_NAMEDDIR); + else + vn_irflag_set_cond(*vpp, + VIRF_NAMEDATTR); + } + if (needs_nameddir && xvp != *vpp) + vput(xvp); + } else { + /* + * Lookup of "." when a named attribute dir is needed. + */ + *vpp = xvp; + } + } + return (error); +} +#else static int zfs_freebsd_lookup(struct vop_lookup_args *ap, boolean_t cached) { @@ -4753,6 +5025,7 @@ zfs_freebsd_lookup(struct vop_lookup_args *ap, boolean_t cached) return (zfs_lookup(ap->a_dvp, nm, ap->a_vpp, cnp, cnp->cn_nameiop, cnp->cn_cred, 0, cached)); } +#endif static int zfs_freebsd_cachedlookup(struct vop_cachedlookup_args *ap) @@ -4775,7 +5048,11 @@ zfs_cache_lookup(struct vop_lookup_args *ap) zfsvfs_t *zfsvfs; zfsvfs = ap->a_dvp->v_mount->mnt_data; +#if __FreeBSD_version >= 1500040 + if (zfsvfs->z_use_namecache && (ap->a_cnp->cn_flags & OPENNAMED) == 0) +#else if (zfsvfs->z_use_namecache) +#endif return (vfs_cache_lookup(ap)); else return (zfs_freebsd_lookup(ap, B_FALSE)); @@ -4798,6 +5075,11 @@ zfs_freebsd_create(struct vop_create_args *ap) vattr_t *vap = ap->a_vap; znode_t *zp = NULL; int rc, mode; + struct vnode *dvp = ap->a_dvp; +#if __FreeBSD_version >= 1500040 + struct vnode *xvp; + bool is_nameddir; +#endif #if __FreeBSD_version < 1400068 ASSERT(cnp->cn_flags & SAVENAME); @@ -4808,10 +5090,36 @@ zfs_freebsd_create(struct vop_create_args *ap) zfsvfs = ap->a_dvp->v_mount->mnt_data; *ap->a_vpp = NULL; - rc = zfs_create(VTOZ(ap->a_dvp), cnp->cn_nameptr, vap, 0, mode, - &zp, cnp->cn_cred, 0 /* flag */, NULL /* vsecattr */, NULL); + rc = 0; +#if __FreeBSD_version >= 1500040 + xvp = NULL; + is_nameddir = (vn_irflag_read(dvp) & VIRF_NAMEDDIR) != 0; + if (!is_nameddir && (cnp->cn_flags & OPENNAMED) != 0) { + /* Needs a named attribute directory. */ + rc = zfs_lookup_nameddir(dvp, cnp, &xvp); + if (rc == 0) { + dvp = xvp; + is_nameddir = true; + } + } + if (is_nameddir && rc == 0) + rc = zfs_check_attrname(cnp->cn_nameptr); +#endif + if (rc == 0) + rc = zfs_create(VTOZ(dvp), cnp->cn_nameptr, vap, 0, mode, + &zp, cnp->cn_cred, 0 /* flag */, NULL /* vsecattr */, NULL); +#if __FreeBSD_version >= 1500040 + if (xvp != NULL) + vput(xvp); +#endif + if (rc == 0) { *ap->a_vpp = ZTOV(zp); +#if __FreeBSD_version >= 1500040 + if (is_nameddir) + vn_irflag_set_cond(*ap->a_vpp, VIRF_NAMEDATTR); +#endif + } if (zfsvfs->z_use_namecache && rc == 0 && (cnp->cn_flags & MAKEENTRY) != 0) cache_enter(ap->a_dvp, *ap->a_vpp, cnp); @@ -4830,13 +5138,21 @@ struct vop_remove_args { static int zfs_freebsd_remove(struct vop_remove_args *ap) { + int error = 0; #if __FreeBSD_version < 1400068 ASSERT(ap->a_cnp->cn_flags & SAVENAME); #endif - return (zfs_remove_(ap->a_dvp, ap->a_vp, ap->a_cnp->cn_nameptr, - ap->a_cnp->cn_cred)); +#if __FreeBSD_version >= 1500040 + if ((vn_irflag_read(ap->a_dvp) & VIRF_NAMEDDIR) != 0) + error = zfs_check_attrname(ap->a_cnp->cn_nameptr); +#endif + + if (error == 0) + error = zfs_remove_(ap->a_dvp, ap->a_vp, ap->a_cnp->cn_nameptr, + ap->a_cnp->cn_cred); + return (error); } #ifndef _SYS_SYSPROTO_H_ @@ -4921,8 +5237,32 @@ struct vop_fsync_args { static int zfs_freebsd_fsync(struct vop_fsync_args *ap) { + vnode_t *vp = ap->a_vp; + int err = 0; - return (zfs_fsync(VTOZ(ap->a_vp), 0, ap->a_td->td_ucred)); + /* + * Push any dirty mmap()'d data out to the DMU and ZIL, ready for + * zil_commit() to be called in zfs_fsync(). + */ + if (vm_object_mightbedirty(vp->v_object)) { + zfs_vmobject_wlock(vp->v_object); + if (!vm_object_page_clean(vp->v_object, 0, 0, 0)) + err = SET_ERROR(EIO); + zfs_vmobject_wunlock(vp->v_object); + if (err) { + /* + * Unclear what state things are in. zfs_putpages() + * will ensure the pages remain dirty if they haven't + * been written down to the DMU, but because there may + * be nothing logged, we can't assume that zfs_sync() + * -> zil_commit() will give us a useful error. It's + * safest if we just error out here. + */ + return (err); + } + } + + return (zfs_fsync(VTOZ(vp), 0, ap->a_td->td_ucred)); } #ifndef _SYS_SYSPROTO_H_ @@ -4994,6 +5334,11 @@ zfs_freebsd_getattr(struct vop_getattr_args *ap) #undef FLAG_CHECK *vap = xvap.xva_vattr; vap->va_flags = fflags; + +#if __FreeBSD_version >= 1500040 + if ((vn_irflag_read(ap->a_vp) & (VIRF_NAMEDDIR | VIRF_NAMEDATTR)) != 0) + vap->va_bsdflags |= SFBSD_NAMEDATTR; +#endif return (0); } @@ -5136,15 +5481,24 @@ zfs_freebsd_rename(struct vop_rename_args *ap) vnode_t *fvp = ap->a_fvp; vnode_t *tdvp = ap->a_tdvp; vnode_t *tvp = ap->a_tvp; - int error; + int error = 0; #if __FreeBSD_version < 1400068 ASSERT(ap->a_fcnp->cn_flags & (SAVENAME|SAVESTART)); ASSERT(ap->a_tcnp->cn_flags & (SAVENAME|SAVESTART)); #endif - error = zfs_do_rename(fdvp, &fvp, ap->a_fcnp, tdvp, &tvp, - ap->a_tcnp, ap->a_fcnp->cn_cred); +#if __FreeBSD_version >= 1500040 + if ((vn_irflag_read(fdvp) & VIRF_NAMEDDIR) != 0) { + error = zfs_check_attrname(ap->a_fcnp->cn_nameptr); + if (error == 0) + error = zfs_check_attrname(ap->a_tcnp->cn_nameptr); + } +#endif + + if (error == 0) + error = zfs_do_rename(fdvp, &fvp, ap->a_fcnp, tdvp, &tvp, + ap->a_tcnp, ap->a_fcnp->cn_cred); vrele(fdvp); vrele(fvp); @@ -5398,12 +5752,33 @@ zfs_freebsd_pathconf(struct vop_pathconf_args *ap) return (0); } return (EINVAL); +#if __FreeBSD_version >= 1500040 + case _PC_NAMEDATTR_ENABLED: + MNT_ILOCK(ap->a_vp->v_mount); + if ((ap->a_vp->v_mount->mnt_flag & MNT_NAMEDATTR) != 0) + *ap->a_retval = 1; + else + *ap->a_retval = 0; + MNT_IUNLOCK(ap->a_vp->v_mount); + return (0); + case _PC_HAS_NAMEDATTR: + if (zfs_has_namedattr(ap->a_vp, curthread->td_ucred)) + *ap->a_retval = 1; + else + *ap->a_retval = 0; + return (0); +#endif +#ifdef _PC_HAS_HIDDENSYSTEM + case _PC_HAS_HIDDENSYSTEM: + *ap->a_retval = 1; + return (0); +#endif default: return (vop_stdpathconf(ap)); } } -static int zfs_xattr_compat = 1; +int zfs_xattr_compat = 1; static int zfs_check_attrname(const char *name) @@ -6436,9 +6811,11 @@ zfs_deallocate(struct vop_deallocate_args *ap) if (error == 0) { if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS || (ap->a_ioflag & IO_SYNC) != 0) - zil_commit(zilog, zp->z_id); - *ap->a_offset = off + len; - *ap->a_len = 0; + error = zil_commit(zilog, zp->z_id); + if (error == 0) { + *ap->a_offset = off + len; + *ap->a_len = 0; + } } zfs_exit(zfsvfs, FTAG); diff --git a/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_znode_os.c b/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_znode_os.c index 9bad1e13d7cc..7cd0a153577c 100644 --- a/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_znode_os.c +++ b/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_znode_os.c @@ -150,8 +150,6 @@ zfs_znode_cache_constructor(void *buf, void *arg, int kmflags) zp->z_xattr_cached = NULL; zp->z_xattr_parent = 0; zp->z_vnode = NULL; - zp->z_sync_writes_cnt = 0; - zp->z_async_writes_cnt = 0; return (0); } @@ -163,18 +161,15 @@ zfs_znode_cache_destructor(void *buf, void *arg) znode_t *zp = buf; ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs)); - ASSERT3P(zp->z_vnode, ==, NULL); + ASSERT0P(zp->z_vnode); ASSERT(!list_link_active(&zp->z_link_node)); mutex_destroy(&zp->z_lock); mutex_destroy(&zp->z_acl_lock); rw_destroy(&zp->z_xattr_lock); zfs_rangelock_fini(&zp->z_rangelock); - ASSERT3P(zp->z_acl_cached, ==, NULL); - ASSERT3P(zp->z_xattr_cached, ==, NULL); - - ASSERT0(atomic_load_32(&zp->z_sync_writes_cnt)); - ASSERT0(atomic_load_32(&zp->z_async_writes_cnt)); + ASSERT0P(zp->z_acl_cached); + ASSERT0P(zp->z_xattr_cached); } @@ -200,7 +195,7 @@ zfs_znode_init(void) /* * Initialize zcache */ - ASSERT3P(znode_uma_zone, ==, NULL); + ASSERT0P(znode_uma_zone); znode_uma_zone = uma_zcreate("zfs_znode_cache", sizeof (znode_t), zfs_znode_cache_constructor_smr, zfs_znode_cache_destructor_smr, NULL, NULL, 0, 0); @@ -229,7 +224,7 @@ zfs_znode_init(void) /* * Initialize zcache */ - ASSERT3P(znode_cache, ==, NULL); + ASSERT0P(znode_cache); znode_cache = kmem_cache_create("zfs_znode_cache", sizeof (znode_t), 0, zfs_znode_cache_constructor, zfs_znode_cache_destructor, NULL, NULL, NULL, KMC_RECLAIMABLE); @@ -358,8 +353,8 @@ zfs_znode_sa_init(zfsvfs_t *zfsvfs, znode_t *zp, ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs) || (zfsvfs == zp->z_zfsvfs)); ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zfsvfs, zp->z_id))); - ASSERT3P(zp->z_sa_hdl, ==, NULL); - ASSERT3P(zp->z_acl_cached, ==, NULL); + ASSERT0P(zp->z_sa_hdl); + ASSERT0P(zp->z_acl_cached); if (sa_hdl == NULL) { VERIFY0(sa_handle_get_from_db(zfsvfs->z_os, db, zp, SA_HDL_SHARED, &zp->z_sa_hdl)); @@ -456,8 +451,6 @@ zfs_znode_alloc(zfsvfs_t *zfsvfs, dmu_buf_t *db, int blksz, zp->z_blksz = blksz; zp->z_seq = 0x7A4653; zp->z_sync_cnt = 0; - zp->z_sync_writes_cnt = 0; - zp->z_async_writes_cnt = 0; atomic_store_ptr(&zp->z_cached_symlink, NULL); zfs_znode_sa_init(zfsvfs, zp, db, obj_type, hdl); @@ -1134,7 +1127,7 @@ zfs_rezget(znode_t *zp) } rw_exit(&zp->z_xattr_lock); - ASSERT3P(zp->z_sa_hdl, ==, NULL); + ASSERT0P(zp->z_sa_hdl); err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db); if (err) { ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); @@ -1305,7 +1298,7 @@ zfs_znode_free(znode_t *zp) zfsvfs_t *zfsvfs = zp->z_zfsvfs; char *symlink; - ASSERT3P(zp->z_sa_hdl, ==, NULL); + ASSERT0P(zp->z_sa_hdl); zp->z_vnode = NULL; mutex_enter(&zfsvfs->z_znodes_lock); POINTER_INVALIDATE(&zp->z_zfsvfs); diff --git a/sys/contrib/openzfs/module/os/freebsd/zfs/zvol_os.c b/sys/contrib/openzfs/module/os/freebsd/zfs/zvol_os.c index 212ef560db07..265dfd55fc4d 100644 --- a/sys/contrib/openzfs/module/os/freebsd/zfs/zvol_os.c +++ b/sys/contrib/openzfs/module/os/freebsd/zfs/zvol_os.c @@ -727,9 +727,9 @@ unlock: break; } - if (commit) { + if (error == 0 && commit) { commit: - zil_commit(zv->zv_zilog, ZVOL_OBJ); + error = zil_commit(zv->zv_zilog, ZVOL_OBJ); } resume: rw_exit(&zv->zv_suspend_lock); @@ -906,8 +906,8 @@ zvol_cdev_write(struct cdev *dev, struct uio *uio_s, int ioflag) zfs_rangelock_exit(lr); int64_t nwritten = start_resid - zfs_uio_resid(&uio); dataset_kstats_update_write_kstats(&zv->zv_kstat, nwritten); - if (commit) - zil_commit(zv->zv_zilog, ZVOL_OBJ); + if (error == 0 && commit) + error = zil_commit(zv->zv_zilog, ZVOL_OBJ); rw_exit(&zv->zv_suspend_lock); return (error); @@ -1117,7 +1117,7 @@ zvol_cdev_ioctl(struct cdev *dev, ulong_t cmd, caddr_t data, case DIOCGFLUSH: rw_enter(&zv->zv_suspend_lock, ZVOL_RW_READER); if (zv->zv_zilog != NULL) - zil_commit(zv->zv_zilog, ZVOL_OBJ); + error = zil_commit(zv->zv_zilog, ZVOL_OBJ); rw_exit(&zv->zv_suspend_lock); break; case DIOCGDELETE: @@ -1152,7 +1152,7 @@ zvol_cdev_ioctl(struct cdev *dev, ulong_t cmd, caddr_t data, } zfs_rangelock_exit(lr); if (sync) - zil_commit(zv->zv_zilog, ZVOL_OBJ); + error = zil_commit(zv->zv_zilog, ZVOL_OBJ); rw_exit(&zv->zv_suspend_lock); break; case DIOCGSTRIPESIZE: @@ -1248,9 +1248,11 @@ zvol_os_is_zvol(const char *device) return (device && strncmp(device, ZVOL_DIR, strlen(ZVOL_DIR)) == 0); } -void +int zvol_os_rename_minor(zvol_state_t *zv, const char *newname) { + int error = 0; + ASSERT(RW_LOCK_HELD(&zvol_state_lock)); ASSERT(MUTEX_HELD(&zv->zv_state_lock)); @@ -1304,14 +1306,94 @@ zvol_os_rename_minor(zvol_state_t *zv, const char *newname) args.mda_gid = GID_OPERATOR; args.mda_mode = 0640; args.mda_si_drv2 = zv; - if (make_dev_s(&args, &dev, "%s/%s", ZVOL_DRIVER, newname) - == 0) { + error = make_dev_s(&args, &dev, "%s/%s", ZVOL_DRIVER, newname); + if (error == 0) { dev->si_iosize_max = maxphys; zsd->zsd_cdev = dev; } } strlcpy(zv->zv_name, newname, sizeof (zv->zv_name)); dataset_kstats_rename(&zv->zv_kstat, newname); + + return (error); +} + +/* + * Allocate memory for a new zvol_state_t and setup the required + * request queue and generic disk structures for the block device. + */ +static int +zvol_alloc(const char *name, uint64_t volsize, uint64_t volblocksize, + zvol_state_t **zvp) +{ + zvol_state_t *zv; + uint64_t volmode; + int error; + + error = dsl_prop_get_integer(name, zfs_prop_to_name(ZFS_PROP_VOLMODE), + &volmode, NULL); + if (error) + return (error); + + if (volmode == ZFS_VOLMODE_DEFAULT) + volmode = zvol_volmode; + + if (volmode == ZFS_VOLMODE_NONE) + return (0); + + zv = kmem_zalloc(sizeof (*zv), KM_SLEEP); + mutex_init(&zv->zv_state_lock, NULL, MUTEX_DEFAULT, NULL); + cv_init(&zv->zv_removing_cv, NULL, CV_DEFAULT, NULL); + zv->zv_zso = kmem_zalloc(sizeof (struct zvol_state_os), KM_SLEEP); + zv->zv_volmode = volmode; + zv->zv_volsize = volsize; + zv->zv_volblocksize = volblocksize; + if (zv->zv_volmode == ZFS_VOLMODE_GEOM) { + struct zvol_state_geom *zsg = &zv->zv_zso->zso_geom; + struct g_provider *pp; + struct g_geom *gp; + + g_topology_lock(); + gp = g_new_geomf(&zfs_zvol_class, "zfs::zvol::%s", name); + gp->start = zvol_geom_bio_start; + gp->access = zvol_geom_access; + pp = g_new_providerf(gp, "%s/%s", ZVOL_DRIVER, name); + pp->flags |= G_PF_DIRECT_RECEIVE | G_PF_DIRECT_SEND; + pp->sectorsize = DEV_BSIZE; + pp->mediasize = 0; + pp->private = zv; + + zsg->zsg_provider = pp; + } else if (zv->zv_volmode == ZFS_VOLMODE_DEV) { + struct zvol_state_dev *zsd = &zv->zv_zso->zso_dev; + struct cdev *dev; + struct make_dev_args args; + + make_dev_args_init(&args); + args.mda_flags = MAKEDEV_CHECKNAME | MAKEDEV_WAITOK; + args.mda_devsw = &zvol_cdevsw; + args.mda_cr = NULL; + args.mda_uid = UID_ROOT; + args.mda_gid = GID_OPERATOR; + args.mda_mode = 0640; + args.mda_si_drv2 = zv; + error = make_dev_s(&args, &dev, "%s/%s", ZVOL_DRIVER, name); + if (error) { + kmem_free(zv->zv_zso, sizeof (struct zvol_state_os)); + kmem_free(zv, sizeof (zvol_state_t)); + return (error); + } + + dev->si_iosize_max = maxphys; + zsd->zsd_cdev = dev; + knlist_init_sx(&zsd->zsd_selinfo.si_note, &zv->zv_state_lock); + } + (void) strlcpy(zv->zv_name, name, MAXPATHLEN); + rw_init(&zv->zv_suspend_lock, NULL, RW_DEFAULT, NULL); + zfs_rangelock_init(&zv->zv_rangelock, NULL, NULL); + + *zvp = zv; + return (error); } /* @@ -1333,7 +1415,7 @@ zvol_os_free(zvol_state_t *zv) struct zvol_state_geom *zsg = &zv->zv_zso->zso_geom; struct g_provider *pp __maybe_unused = zsg->zsg_provider; - ASSERT3P(pp->private, ==, NULL); + ASSERT0P(pp->private); g_topology_lock(); zvol_geom_destroy(zv); @@ -1343,7 +1425,7 @@ zvol_os_free(zvol_state_t *zv) struct cdev *dev = zsd->zsd_cdev; if (dev != NULL) { - ASSERT3P(dev->si_drv2, ==, NULL); + ASSERT0P(dev->si_drv2); destroy_dev(dev); knlist_clear(&zsd->zsd_selinfo.si_note, 0); knlist_destroy(&zsd->zsd_selinfo.si_note); @@ -1364,11 +1446,11 @@ zvol_os_free(zvol_state_t *zv) int zvol_os_create_minor(const char *name) { - zvol_state_t *zv; + zvol_state_t *zv = NULL; objset_t *os; dmu_object_info_t *doi; uint64_t volsize; - uint64_t volmode, hash, len; + uint64_t hash, len; int error; bool replayed_zil = B_FALSE; @@ -1400,74 +1482,22 @@ zvol_os_create_minor(const char *name) if (error) goto out_dmu_objset_disown; - error = dsl_prop_get_integer(name, - zfs_prop_to_name(ZFS_PROP_VOLMODE), &volmode, NULL); - if (error || volmode == ZFS_VOLMODE_DEFAULT) - volmode = zvol_volmode; - error = 0; + error = zvol_alloc(name, volsize, doi->doi_data_block_size, &zv); + if (error || zv == NULL) + goto out_dmu_objset_disown; - /* - * zvol_alloc equivalent ... - */ - zv = kmem_zalloc(sizeof (*zv), KM_SLEEP); zv->zv_hash = hash; - mutex_init(&zv->zv_state_lock, NULL, MUTEX_DEFAULT, NULL); - cv_init(&zv->zv_removing_cv, NULL, CV_DEFAULT, NULL); - zv->zv_zso = kmem_zalloc(sizeof (struct zvol_state_os), KM_SLEEP); - zv->zv_volmode = volmode; - if (zv->zv_volmode == ZFS_VOLMODE_GEOM) { - struct zvol_state_geom *zsg = &zv->zv_zso->zso_geom; - struct g_provider *pp; - struct g_geom *gp; - - g_topology_lock(); - gp = g_new_geomf(&zfs_zvol_class, "zfs::zvol::%s", name); - gp->start = zvol_geom_bio_start; - gp->access = zvol_geom_access; - pp = g_new_providerf(gp, "%s/%s", ZVOL_DRIVER, name); - pp->flags |= G_PF_DIRECT_RECEIVE | G_PF_DIRECT_SEND; - pp->sectorsize = DEV_BSIZE; - pp->mediasize = 0; - pp->private = zv; - - zsg->zsg_provider = pp; - } else if (zv->zv_volmode == ZFS_VOLMODE_DEV) { - struct zvol_state_dev *zsd = &zv->zv_zso->zso_dev; - struct cdev *dev; - struct make_dev_args args; - - make_dev_args_init(&args); - args.mda_flags = MAKEDEV_CHECKNAME | MAKEDEV_WAITOK; - args.mda_devsw = &zvol_cdevsw; - args.mda_cr = NULL; - args.mda_uid = UID_ROOT; - args.mda_gid = GID_OPERATOR; - args.mda_mode = 0640; - args.mda_si_drv2 = zv; - if (make_dev_s(&args, &dev, "%s/%s", ZVOL_DRIVER, name) - == 0) { - dev->si_iosize_max = maxphys; - zsd->zsd_cdev = dev; - knlist_init_sx(&zsd->zsd_selinfo.si_note, - &zv->zv_state_lock); - } - } - (void) strlcpy(zv->zv_name, name, MAXPATHLEN); - rw_init(&zv->zv_suspend_lock, NULL, RW_DEFAULT, NULL); - zfs_rangelock_init(&zv->zv_rangelock, NULL, NULL); if (dmu_objset_is_snapshot(os) || !spa_writeable(dmu_objset_spa(os))) zv->zv_flags |= ZVOL_RDONLY; - zv->zv_volblocksize = doi->doi_data_block_size; - zv->zv_volsize = volsize; zv->zv_objset = os; - ASSERT3P(zv->zv_kstat.dk_kstats, ==, NULL); + ASSERT0P(zv->zv_kstat.dk_kstats); error = dataset_kstats_create(&zv->zv_kstat, zv->zv_objset); if (error) goto out_dmu_objset_disown; - ASSERT3P(zv->zv_zilog, ==, NULL); + ASSERT0P(zv->zv_zilog); zv->zv_zilog = zil_open(os, zvol_get_data, &zv->zv_kstat.dk_zil_sums); if (spa_writeable(dmu_objset_spa(os))) { if (zil_replay_disable) @@ -1490,13 +1520,14 @@ zvol_os_create_minor(const char *name) out_dmu_objset_disown: dmu_objset_disown(os, B_TRUE, FTAG); - if (error == 0 && volmode == ZFS_VOLMODE_GEOM) { + if (error == 0 && zv && zv->zv_volmode == ZFS_VOLMODE_GEOM) { g_error_provider(zv->zv_zso->zso_geom.zsg_provider, 0); + /* geom was locked inside zvol_alloc() function */ g_topology_unlock(); } out_doi: kmem_free(doi, sizeof (dmu_object_info_t)); - if (error == 0) { + if (error == 0 && zv) { rw_enter(&zvol_state_lock, RW_WRITER); zvol_insert(zv); zvol_minors++; diff --git a/sys/contrib/openzfs/module/os/linux/spl/spl-condvar.c b/sys/contrib/openzfs/module/os/linux/spl/spl-condvar.c index ce9c9e39e60c..aac5f2ebbfd2 100644 --- a/sys/contrib/openzfs/module/os/linux/spl/spl-condvar.c +++ b/sys/contrib/openzfs/module/os/linux/spl/spl-condvar.c @@ -66,9 +66,9 @@ void __cv_init(kcondvar_t *cvp, char *name, kcv_type_t type, void *arg) { ASSERT(cvp); - ASSERT(name == NULL); + ASSERT0P(name); ASSERT(type == CV_DEFAULT); - ASSERT(arg == NULL); + ASSERT0P(arg); cvp->cv_magic = CV_MAGIC; init_waitqueue_head(&cvp->cv_event); @@ -83,7 +83,7 @@ static int cv_destroy_wakeup(kcondvar_t *cvp) { if (!atomic_read(&cvp->cv_waiters) && !atomic_read(&cvp->cv_refs)) { - ASSERT(cvp->cv_mutex == NULL); + ASSERT0P(cvp->cv_mutex); ASSERT(!waitqueue_active(&cvp->cv_event)); return (1); } @@ -104,7 +104,7 @@ __cv_destroy(kcondvar_t *cvp) while (cv_destroy_wakeup(cvp) == 0) wait_event_timeout(cvp->cv_destroy, cv_destroy_wakeup(cvp), 1); - ASSERT3P(cvp->cv_mutex, ==, NULL); + ASSERT0P(cvp->cv_mutex); ASSERT3S(atomic_read(&cvp->cv_refs), ==, 0); ASSERT3S(atomic_read(&cvp->cv_waiters), ==, 0); ASSERT3S(waitqueue_active(&cvp->cv_event), ==, 0); diff --git a/sys/contrib/openzfs/module/os/linux/spl/spl-generic.c b/sys/contrib/openzfs/module/os/linux/spl/spl-generic.c index f37699b4347e..89ca4a648b2f 100644 --- a/sys/contrib/openzfs/module/os/linux/spl/spl-generic.c +++ b/sys/contrib/openzfs/module/os/linux/spl/spl-generic.c @@ -709,7 +709,7 @@ zone_get_hostid(void *zone) { uint32_t hostid; - ASSERT3P(zone, ==, NULL); + ASSERT0P(zone); if (spl_hostid != 0) return ((uint32_t)(spl_hostid & HW_HOSTID_MASK)); diff --git a/sys/contrib/openzfs/module/os/linux/spl/spl-kmem-cache.c b/sys/contrib/openzfs/module/os/linux/spl/spl-kmem-cache.c index fab80289b278..22e4ed169d03 100644 --- a/sys/contrib/openzfs/module/os/linux/spl/spl-kmem-cache.c +++ b/sys/contrib/openzfs/module/os/linux/spl/spl-kmem-cache.c @@ -296,7 +296,7 @@ spl_slab_free(spl_kmem_slab_t *sks, spl_kmem_cache_t *skc; ASSERT(sks->sks_magic == SKS_MAGIC); - ASSERT(sks->sks_ref == 0); + ASSERT0(sks->sks_ref); skc = sks->sks_cache; ASSERT(skc->skc_magic == SKC_MAGIC); @@ -598,7 +598,7 @@ static void spl_magazine_free(spl_kmem_magazine_t *skm) { ASSERT(skm->skm_magic == SKM_MAGIC); - ASSERT(skm->skm_avail == 0); + ASSERT0(skm->skm_avail); kfree(skm); } @@ -610,7 +610,7 @@ spl_magazine_create(spl_kmem_cache_t *skc) { int i = 0; - ASSERT((skc->skc_flags & KMC_SLAB) == 0); + ASSERT0((skc->skc_flags & KMC_SLAB)); skc->skc_mag = kzalloc(sizeof (spl_kmem_magazine_t *) * num_possible_cpus(), kmem_flags_convert(KM_SLEEP)); @@ -640,7 +640,7 @@ spl_magazine_destroy(spl_kmem_cache_t *skc) spl_kmem_magazine_t *skm; int i = 0; - ASSERT((skc->skc_flags & KMC_SLAB) == 0); + ASSERT0((skc->skc_flags & KMC_SLAB)); for_each_possible_cpu(i) { skm = skc->skc_mag[i]; @@ -679,8 +679,8 @@ spl_kmem_cache_create(const char *name, size_t size, size_t align, /* * Unsupported flags */ - ASSERT(vmp == NULL); - ASSERT(reclaim == NULL); + ASSERT0P(vmp); + ASSERT0P(reclaim); might_sleep(); @@ -863,11 +863,11 @@ spl_kmem_cache_destroy(spl_kmem_cache_t *skc) * Validate there are no objects in use and free all the * spl_kmem_slab_t, spl_kmem_obj_t, and object buffers. */ - ASSERT3U(skc->skc_slab_alloc, ==, 0); - ASSERT3U(skc->skc_obj_alloc, ==, 0); - ASSERT3U(skc->skc_slab_total, ==, 0); - ASSERT3U(skc->skc_obj_total, ==, 0); - ASSERT3U(skc->skc_obj_emergency, ==, 0); + ASSERT0(skc->skc_slab_alloc); + ASSERT0(skc->skc_obj_alloc); + ASSERT0(skc->skc_slab_total); + ASSERT0(skc->skc_obj_total); + ASSERT0(skc->skc_obj_emergency); ASSERT(list_empty(&skc->skc_complete_list)); ASSERT3U(percpu_counter_sum(&skc->skc_linux_alloc), ==, 0); @@ -986,7 +986,7 @@ spl_cache_grow(spl_kmem_cache_t *skc, int flags, void **obj) ASSERT0(flags & ~KM_PUBLIC_MASK); ASSERT(skc->skc_magic == SKC_MAGIC); - ASSERT((skc->skc_flags & KMC_SLAB) == 0); + ASSERT0((skc->skc_flags & KMC_SLAB)); *obj = NULL; diff --git a/sys/contrib/openzfs/module/os/linux/spl/spl-kmem.c b/sys/contrib/openzfs/module/os/linux/spl/spl-kmem.c index 337a4bcf76a0..9fe008cef868 100644 --- a/sys/contrib/openzfs/module/os/linux/spl/spl-kmem.c +++ b/sys/contrib/openzfs/module/os/linux/spl/spl-kmem.c @@ -302,13 +302,8 @@ spl_kmem_free_impl(const void *buf, size_t size) #ifdef DEBUG_KMEM /* Shim layer memory accounting */ -#ifdef HAVE_ATOMIC64_T atomic64_t kmem_alloc_used = ATOMIC64_INIT(0); -unsigned long long kmem_alloc_max = 0; -#else /* HAVE_ATOMIC64_T */ -atomic_t kmem_alloc_used = ATOMIC_INIT(0); -unsigned long long kmem_alloc_max = 0; -#endif /* HAVE_ATOMIC64_T */ +uint64_t kmem_alloc_max = 0; EXPORT_SYMBOL(kmem_alloc_used); EXPORT_SYMBOL(kmem_alloc_max); @@ -320,9 +315,9 @@ spl_kmem_alloc_debug(size_t size, int flags, int node) ptr = spl_kmem_alloc_impl(size, flags, node); if (ptr) { - kmem_alloc_used_add(size); - if (unlikely(kmem_alloc_used_read() > kmem_alloc_max)) - kmem_alloc_max = kmem_alloc_used_read(); + atomic64_add(size, &kmem_alloc_used); + if (unlikely(atomic64_read(&kmem_alloc_used) > kmem_alloc_max)) + kmem_alloc_max = atomic64_read(&kmem_alloc_used); } return (ptr); @@ -331,7 +326,7 @@ spl_kmem_alloc_debug(size_t size, int flags, int node) inline void spl_kmem_free_debug(const void *ptr, size_t size) { - kmem_alloc_used_sub(size); + atomic64_sub(size, &kmem_alloc_used); spl_kmem_free_impl(ptr, size); } @@ -595,7 +590,7 @@ spl_kmem_init(void) { #ifdef DEBUG_KMEM - kmem_alloc_used_set(0); + atomic64_set(&kmem_alloc_used, 0); @@ -617,9 +612,10 @@ spl_kmem_fini(void) * at that address to aid in debugging. Performance is not * a serious concern here since it is module unload time. */ - if (kmem_alloc_used_read() != 0) + if (atomic64_read(&kmem_alloc_used) != 0) printk(KERN_WARNING "kmem leaked %ld/%llu bytes\n", - (unsigned long)kmem_alloc_used_read(), kmem_alloc_max); + (unsigned long)atomic64_read(&kmem_alloc_used), + kmem_alloc_max); #ifdef DEBUG_KMEM_TRACKING spl_kmem_fini_tracking(&kmem_list, &kmem_lock); diff --git a/sys/contrib/openzfs/module/os/linux/spl/spl-kstat.c b/sys/contrib/openzfs/module/os/linux/spl/spl-kstat.c index 48f70b00c96b..02c5b42bc4a0 100644 --- a/sys/contrib/openzfs/module/os/linux/spl/spl-kstat.c +++ b/sys/contrib/openzfs/module/os/linux/spl/spl-kstat.c @@ -541,7 +541,7 @@ __kstat_create(const char *ks_module, int ks_instance, const char *ks_name, kstat_t *ksp; ASSERT(ks_module); - ASSERT(ks_instance == 0); + ASSERT0(ks_instance); ASSERT(ks_name); if ((ks_type == KSTAT_TYPE_INTR) || (ks_type == KSTAT_TYPE_IO)) diff --git a/sys/contrib/openzfs/module/os/linux/spl/spl-proc.c b/sys/contrib/openzfs/module/os/linux/spl/spl-proc.c index 4ed0deedd5b9..8cdd5fc5cfe5 100644 --- a/sys/contrib/openzfs/module/os/linux/spl/spl-proc.c +++ b/sys/contrib/openzfs/module/os/linux/spl/spl-proc.c @@ -82,11 +82,7 @@ proc_domemused(CONST_CTL_TABLE *table, int write, if (write) { *ppos += *lenp; } else { -#ifdef HAVE_ATOMIC64_T val = atomic64_read((atomic64_t *)table->data); -#else - val = atomic_read((atomic_t *)table->data); -#endif /* HAVE_ATOMIC64_T */ rc = proc_doulongvec_minmax(&dummy, write, buffer, lenp, ppos); } @@ -315,18 +311,14 @@ static struct ctl_table spl_kmem_table[] = { { .procname = "kmem_used", .data = &kmem_alloc_used, -#ifdef HAVE_ATOMIC64_T .maxlen = sizeof (atomic64_t), -#else - .maxlen = sizeof (atomic_t), -#endif /* HAVE_ATOMIC64_T */ .mode = 0444, .proc_handler = &proc_domemused, }, { .procname = "kmem_max", .data = &kmem_alloc_max, - .maxlen = sizeof (unsigned long), + .maxlen = sizeof (uint64_t), .extra1 = &table_min, .extra2 = &table_max, .mode = 0444, diff --git a/sys/contrib/openzfs/module/os/linux/spl/spl-thread.c b/sys/contrib/openzfs/module/os/linux/spl/spl-thread.c index 1398483a3ac8..8f5c73b13df5 100644 --- a/sys/contrib/openzfs/module/os/linux/spl/spl-thread.c +++ b/sys/contrib/openzfs/module/os/linux/spl/spl-thread.c @@ -28,6 +28,7 @@ #include <sys/kmem.h> #include <sys/tsd.h> #include <sys/string.h> +#include <sys/misc.h> /* * Thread interfaces @@ -79,7 +80,7 @@ __thread_create(caddr_t stk, size_t stksize, thread_func_t func, /* Option pp is simply ignored */ /* Variable stack size unsupported */ - ASSERT(stk == NULL); + ASSERT0P(stk); tp = kmem_alloc(sizeof (thread_priv_t), KM_PUSHPAGE); if (tp == NULL) @@ -197,3 +198,14 @@ issig(void) } EXPORT_SYMBOL(issig); + +/* + * Check if the current thread is a memory reclaim thread. + * Returns true if current thread is kswapd. + */ +int +current_is_reclaim_thread(void) +{ + return (current_is_kswapd()); +} +EXPORT_SYMBOL(current_is_reclaim_thread); diff --git a/sys/contrib/openzfs/module/os/linux/spl/spl-tsd.c b/sys/contrib/openzfs/module/os/linux/spl/spl-tsd.c index 34a61bef7d4f..2e8cedf0dc87 100644 --- a/sys/contrib/openzfs/module/os/linux/spl/spl-tsd.c +++ b/sys/contrib/openzfs/module/os/linux/spl/spl-tsd.c @@ -161,7 +161,7 @@ tsd_hash_add(tsd_hash_table_t *table, uint_t key, pid_t pid, void *value) ulong_t hash; int rc = 0; - ASSERT3P(tsd_hash_search(table, key, pid), ==, NULL); + ASSERT0P(tsd_hash_search(table, key, pid)); /* New entry allocate structure, set value, and add to hash */ entry = kmem_alloc(sizeof (tsd_hash_entry_t), KM_PUSHPAGE); diff --git a/sys/contrib/openzfs/module/os/linux/zfs/abd_os.c b/sys/contrib/openzfs/module/os/linux/zfs/abd_os.c index e1140b31a97a..8a8316f63c48 100644 --- a/sys/contrib/openzfs/module/os/linux/zfs/abd_os.c +++ b/sys/contrib/openzfs/module/os/linux/zfs/abd_os.c @@ -256,10 +256,6 @@ abd_unmark_zfs_page(struct page *page) #ifndef CONFIG_HIGHMEM -#ifndef __GFP_RECLAIM -#define __GFP_RECLAIM __GFP_WAIT -#endif - /* * The goal is to minimize fragmentation by preferentially populating ABDs * with higher order compound pages from a single zone. Allocation size is @@ -867,9 +863,9 @@ abd_iter_advance(struct abd_iter *aiter, size_t amount) * Ensure that last chunk is not in use. abd_iterate_*() must clear * this state (directly or abd_iter_unmap()) before advancing. */ - ASSERT3P(aiter->iter_mapaddr, ==, NULL); + ASSERT0P(aiter->iter_mapaddr); ASSERT0(aiter->iter_mapsize); - ASSERT3P(aiter->iter_page, ==, NULL); + ASSERT0P(aiter->iter_page); ASSERT0(aiter->iter_page_doff); ASSERT0(aiter->iter_page_dsize); @@ -901,7 +897,7 @@ abd_iter_map(struct abd_iter *aiter) void *paddr; size_t offset = 0; - ASSERT3P(aiter->iter_mapaddr, ==, NULL); + ASSERT0P(aiter->iter_mapaddr); ASSERT0(aiter->iter_mapsize); /* There's nothing left to iterate over, so do nothing */ diff --git a/sys/contrib/openzfs/module/os/linux/zfs/vdev_disk.c b/sys/contrib/openzfs/module/os/linux/zfs/vdev_disk.c index 154ca22d9513..830fad7fe793 100644 --- a/sys/contrib/openzfs/module/os/linux/zfs/vdev_disk.c +++ b/sys/contrib/openzfs/module/os/linux/zfs/vdev_disk.c @@ -552,7 +552,7 @@ vdev_bio_associate_blkg(struct bio *bio) #endif ASSERT3P(q, !=, NULL); - ASSERT3P(bio->bi_blkg, ==, NULL); + ASSERT0P(bio->bi_blkg); if (q->root_blkg && vdev_blkg_tryget(q->root_blkg)) bio->bi_blkg = q->root_blkg; @@ -574,7 +574,7 @@ vdev_bio_set_dev(struct bio *bio, struct block_device *bdev) bio->bi_bdev = bdev; ASSERT3P(q, !=, NULL); - ASSERT3P(bio->bi_blkg, ==, NULL); + ASSERT0P(bio->bi_blkg); if (q->root_blkg && vdev_blkg_tryget(q->root_blkg)) bio->bi_blkg = q->root_blkg; @@ -806,7 +806,7 @@ vbio_completion(struct bio *bio) * here; instead we stash vbio on the zio and take care of it in the * done callback. */ - ASSERT3P(zio->io_bio, ==, NULL); + ASSERT0P(zio->io_bio); zio->io_bio = vbio; zio_delay_interrupt(zio); diff --git a/sys/contrib/openzfs/module/os/linux/zfs/zfs_acl.c b/sys/contrib/openzfs/module/os/linux/zfs/zfs_acl.c index 1b169122f25b..daa4b5776837 100644 --- a/sys/contrib/openzfs/module/os/linux/zfs/zfs_acl.c +++ b/sys/contrib/openzfs/module/os/linux/zfs/zfs_acl.c @@ -1900,7 +1900,7 @@ zfs_acl_ids_create(znode_t *dzp, int flag, vattr_t *vap, cred_t *cr, if (!(flag & IS_ROOT_NODE) && (dzp->z_pflags & ZFS_INHERIT_ACE) && !(dzp->z_pflags & ZFS_XATTR)) { - VERIFY(0 == zfs_acl_node_read(dzp, B_TRUE, + VERIFY0(zfs_acl_node_read(dzp, B_TRUE, &paclp, B_FALSE)); acl_ids->z_aclp = zfs_acl_inherit(zfsvfs, vap->va_mode, paclp, acl_ids->z_mode, &need_chmod); @@ -2204,8 +2204,8 @@ top: } error = zfs_aclset_common(zp, aclp, cr, tx); - ASSERT(error == 0); - ASSERT(zp->z_acl_cached == NULL); + ASSERT0(error); + ASSERT0P(zp->z_acl_cached); zp->z_acl_cached = aclp; if (fuid_dirtied) diff --git a/sys/contrib/openzfs/module/os/linux/zfs/zfs_ctldir.c b/sys/contrib/openzfs/module/os/linux/zfs/zfs_ctldir.c index 84b25cb2c5ac..fb4de50480a3 100644 --- a/sys/contrib/openzfs/module/os/linux/zfs/zfs_ctldir.c +++ b/sys/contrib/openzfs/module/os/linux/zfs/zfs_ctldir.c @@ -494,9 +494,9 @@ zfsctl_inode_alloc(zfsvfs_t *zfsvfs, uint64_t id, if (!creation) now = current_time(ip); zp = ITOZ(ip); - ASSERT3P(zp->z_dirlocks, ==, NULL); - ASSERT3P(zp->z_acl_cached, ==, NULL); - ASSERT3P(zp->z_xattr_cached, ==, NULL); + ASSERT0P(zp->z_dirlocks); + ASSERT0P(zp->z_acl_cached); + ASSERT0P(zp->z_xattr_cached); zp->z_id = id; zp->z_unlinked = B_FALSE; zp->z_atime_dirty = B_FALSE; @@ -511,8 +511,6 @@ zfsctl_inode_alloc(zfsvfs_t *zfsvfs, uint64_t id, zp->z_pflags = 0; zp->z_mode = 0; zp->z_sync_cnt = 0; - zp->z_sync_writes_cnt = 0; - zp->z_async_writes_cnt = 0; ip->i_generation = 0; ip->i_ino = id; ip->i_mode = (S_IFDIR | S_IRWXUGO); @@ -592,7 +590,7 @@ zfsctl_inode_lookup(zfsvfs_t *zfsvfs, uint64_t id, int zfsctl_create(zfsvfs_t *zfsvfs) { - ASSERT(zfsvfs->z_ctldir == NULL); + ASSERT0P(zfsvfs->z_ctldir); zfsvfs->z_ctldir = zfsctl_inode_alloc(zfsvfs, ZFSCTL_INO_ROOT, &zpl_fops_root, &zpl_ops_root, 0); diff --git a/sys/contrib/openzfs/module/os/linux/zfs/zfs_dir.c b/sys/contrib/openzfs/module/os/linux/zfs/zfs_dir.c index 2f935bb3fc8c..e8de536606e2 100644 --- a/sys/contrib/openzfs/module/os/linux/zfs/zfs_dir.c +++ b/sys/contrib/openzfs/module/os/linux/zfs/zfs_dir.c @@ -463,7 +463,7 @@ zfs_unlinked_add(znode_t *zp, dmu_tx_t *tx) zfsvfs_t *zfsvfs = ZTOZSB(zp); ASSERT(zp->z_unlinked); - ASSERT(ZTOI(zp)->i_nlink == 0); + ASSERT0(ZTOI(zp)->i_nlink); VERIFY3U(0, ==, zap_add_int(zfsvfs->z_os, zfsvfs->z_unlinkedobj, zp->z_id, tx)); @@ -662,8 +662,8 @@ zfs_rmnode(znode_t *zp) uint64_t links; int error; - ASSERT(ZTOI(zp)->i_nlink == 0); - ASSERT(atomic_read(&ZTOI(zp)->i_count) == 0); + ASSERT0(ZTOI(zp)->i_nlink); + ASSERT0(atomic_read(&ZTOI(zp)->i_count)); /* * If this is an attribute directory, purge its contents. @@ -710,7 +710,7 @@ zfs_rmnode(znode_t *zp) &xattr_obj, sizeof (xattr_obj)); if (error == 0 && xattr_obj) { error = zfs_zget(zfsvfs, xattr_obj, &xzp); - ASSERT(error == 0); + ASSERT0(error); } acl_obj = zfs_external_acl(zp); @@ -744,12 +744,12 @@ zfs_rmnode(znode_t *zp) } if (xzp) { - ASSERT(error == 0); + ASSERT0(error); mutex_enter(&xzp->z_lock); xzp->z_unlinked = B_TRUE; /* mark xzp for deletion */ clear_nlink(ZTOI(xzp)); /* no more links to it */ links = 0; - VERIFY(0 == sa_update(xzp->z_sa_hdl, SA_ZPL_LINKS(zfsvfs), + VERIFY0(sa_update(xzp->z_sa_hdl, SA_ZPL_LINKS(zfsvfs), &links, sizeof (links), tx)); mutex_exit(&xzp->z_lock); zfs_unlinked_add(xzp, tx); @@ -872,7 +872,7 @@ zfs_link_create(zfs_dirlock_t *dl, znode_t *zp, dmu_tx_t *tx, int flag) ctime); } error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); - ASSERT(error == 0); + ASSERT0(error); mutex_exit(&zp->z_lock); @@ -894,7 +894,7 @@ zfs_link_create(zfs_dirlock_t *dl, znode_t *zp, dmu_tx_t *tx, int flag) &dzp->z_pflags, sizeof (dzp->z_pflags)); zfs_tstamp_update_setup(dzp, CONTENT_MODIFIED, mtime, ctime); error = sa_bulk_update(dzp->z_sa_hdl, bulk, count, tx); - ASSERT(error == 0); + ASSERT0(error); mutex_exit(&dzp->z_lock); return (0); @@ -986,7 +986,7 @@ zfs_drop_nlink_locked(znode_t *zp, dmu_tx_t *tx, boolean_t *unlinkedp) SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL, &links, sizeof (links)); error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); - ASSERT3U(error, ==, 0); + ASSERT0(error); if (unlinkedp != NULL) *unlinkedp = unlinked; @@ -1058,7 +1058,7 @@ zfs_link_destroy(zfs_dirlock_t *dl, znode_t *zp, dmu_tx_t *tx, int flag, /* The only error is !zfs_dirempty() and we checked earlier. */ error = zfs_drop_nlink_locked(zp, tx, &unlinked); - ASSERT3U(error, ==, 0); + ASSERT0(error); mutex_exit(&zp->z_lock); } else { error = zfs_dropname(dl, zp, dzp, tx, flag); @@ -1083,7 +1083,7 @@ zfs_link_destroy(zfs_dirlock_t *dl, znode_t *zp, dmu_tx_t *tx, int flag, NULL, &dzp->z_pflags, sizeof (dzp->z_pflags)); zfs_tstamp_update_setup(dzp, CONTENT_MODIFIED, mtime, ctime); error = sa_bulk_update(dzp->z_sa_hdl, bulk, count, tx); - ASSERT(error == 0); + ASSERT0(error); mutex_exit(&dzp->z_lock); if (unlinkedp != NULL) @@ -1167,7 +1167,7 @@ zfs_make_xattrdir(znode_t *zp, vattr_t *vap, znode_t **xzpp, cred_t *cr) ASSERT(error == 0 && parent == zp->z_id); #endif - VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), &xzp->z_id, + VERIFY0(sa_update(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), &xzp->z_id, sizeof (xzp->z_id), tx)); if (!zp->z_unlinked) diff --git a/sys/contrib/openzfs/module/os/linux/zfs/zfs_file_os.c b/sys/contrib/openzfs/module/os/linux/zfs/zfs_file_os.c index d193eb80dca2..c729947369c2 100644 --- a/sys/contrib/openzfs/module/os/linux/zfs/zfs_file_os.c +++ b/sys/contrib/openzfs/module/os/linux/zfs/zfs_file_os.c @@ -260,24 +260,12 @@ zfs_file_fsync(zfs_file_t *filp, int flags) { int datasync = 0; int error; - int fstrans; if (flags & O_DSYNC) datasync = 1; - /* - * May enter XFS which generates a warning when PF_FSTRANS is set. - * To avoid this the flag is cleared over vfs_sync() and then reset. - */ - fstrans = __spl_pf_fstrans_check(); - if (fstrans) - current->flags &= ~(__SPL_PF_FSTRANS); - error = -vfs_fsync(filp, datasync); - if (fstrans) - current->flags |= __SPL_PF_FSTRANS; - return (error); } @@ -292,14 +280,6 @@ int zfs_file_deallocate(zfs_file_t *fp, loff_t offset, loff_t len) { /* - * May enter XFS which generates a warning when PF_FSTRANS is set. - * To avoid this the flag is cleared over vfs_sync() and then reset. - */ - int fstrans = __spl_pf_fstrans_check(); - if (fstrans) - current->flags &= ~(__SPL_PF_FSTRANS); - - /* * When supported by the underlying file system preferentially * use the fallocate() callback to preallocate the space. */ @@ -308,9 +288,6 @@ zfs_file_deallocate(zfs_file_t *fp, loff_t offset, loff_t len) error = -fp->f_op->fallocate(fp, FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE, offset, len); - if (fstrans) - current->flags |= __SPL_PF_FSTRANS; - if (error) return (SET_ERROR(error)); diff --git a/sys/contrib/openzfs/module/os/linux/zfs/zfs_sysfs.c b/sys/contrib/openzfs/module/os/linux/zfs/zfs_sysfs.c index 1c187d7b9cab..895d80b2d79e 100644 --- a/sys/contrib/openzfs/module/os/linux/zfs/zfs_sysfs.c +++ b/sys/contrib/openzfs/module/os/linux/zfs/zfs_sysfs.c @@ -223,7 +223,7 @@ zfs_kobj_add(zfs_mod_kobj_t *zkobj, struct kobject *parent, const char *name) { /* zko_default_group.attrs must be NULL terminated */ ASSERT(zkobj->zko_default_group.attrs != NULL); - ASSERT(zkobj->zko_default_group.attrs[zkobj->zko_attr_count] == NULL); + ASSERT0P(zkobj->zko_default_group.attrs[zkobj->zko_attr_count]); kobject_init(&zkobj->zko_kobj, &zkobj->zko_kobj_type); return (kobject_add(&zkobj->zko_kobj, parent, name)); diff --git a/sys/contrib/openzfs/module/os/linux/zfs/zfs_vfsops.c b/sys/contrib/openzfs/module/os/linux/zfs/zfs_vfsops.c index a3837f784668..cd606e667bff 100644 --- a/sys/contrib/openzfs/module/os/linux/zfs/zfs_vfsops.c +++ b/sys/contrib/openzfs/module/os/linux/zfs/zfs_vfsops.c @@ -279,19 +279,14 @@ zfs_sync(struct super_block *sb, int wait, cred_t *cr) return (err); /* - * If the pool is suspended, just return an error. This is to help - * with shutting down with pools suspended, as we don't want to block - * in that case. + * Sync any pending writes, but do not block if the pool is suspended. + * This is to help with shutting down with pools suspended, as we don't + * want to block in that case. */ - if (spa_suspended(zfsvfs->z_os->os_spa)) { - zfs_exit(zfsvfs, FTAG); - return (SET_ERROR(EIO)); - } - - zil_commit(zfsvfs->z_log, 0); + err = zil_commit_flags(zfsvfs->z_log, 0, ZIL_COMMIT_NOW); zfs_exit(zfsvfs, FTAG); - return (0); + return (err); } static void @@ -883,7 +878,7 @@ zfsvfs_setup(zfsvfs_t *zfsvfs, boolean_t mounting) * operations out since we closed the ZIL. */ if (mounting) { - ASSERT3P(zfsvfs->z_kstat.dk_kstats, ==, NULL); + ASSERT0P(zfsvfs->z_kstat.dk_kstats); error = dataset_kstats_create(&zfsvfs->z_kstat, zfsvfs->z_os); if (error) return (error); @@ -1217,6 +1212,63 @@ zfs_root(zfsvfs_t *zfsvfs, struct inode **ipp) } /* + * Dentry and inode caches referenced by a task in non-root memcg are + * not going to be scanned by the kernel-provided shrinker. So, if + * kernel prunes nothing, fall back to this manual walk to free dnodes. + * To avoid scanning the same znodes multiple times they are always rotated + * to the end of the z_all_znodes list. New znodes are inserted at the + * end of the list so we're always scanning the oldest znodes first. + */ +static int +zfs_prune_aliases(zfsvfs_t *zfsvfs, unsigned long nr_to_scan) +{ + znode_t **zp_array, *zp; + int max_array = MIN(nr_to_scan, PAGE_SIZE * 8 / sizeof (znode_t *)); + int objects = 0; + int i = 0, j = 0; + + zp_array = vmem_zalloc(max_array * sizeof (znode_t *), KM_SLEEP); + + mutex_enter(&zfsvfs->z_znodes_lock); + while ((zp = list_head(&zfsvfs->z_all_znodes)) != NULL) { + + if ((i++ > nr_to_scan) || (j >= max_array)) + break; + + ASSERT(list_link_active(&zp->z_link_node)); + list_remove(&zfsvfs->z_all_znodes, zp); + list_insert_tail(&zfsvfs->z_all_znodes, zp); + + /* Skip active znodes and .zfs entries */ + if (MUTEX_HELD(&zp->z_lock) || zp->z_is_ctldir) + continue; + + if (igrab(ZTOI(zp)) == NULL) + continue; + + zp_array[j] = zp; + j++; + } + mutex_exit(&zfsvfs->z_znodes_lock); + + for (i = 0; i < j; i++) { + zp = zp_array[i]; + + ASSERT3P(zp, !=, NULL); + d_prune_aliases(ZTOI(zp)); + + if (atomic_read(&ZTOI(zp)->i_count) == 1) + objects++; + + zrele(zp); + } + + vmem_free(zp_array, max_array * sizeof (znode_t *)); + + return (objects); +} + +/* * The ARC has requested that the filesystem drop entries from the dentry * and inode caches. This can occur when the ARC needs to free meta data * blocks but can't because they are all pinned by entries in these caches. @@ -1267,6 +1319,14 @@ zfs_prune(struct super_block *sb, unsigned long nr_to_scan, int *objects) *objects = (*shrinker->scan_objects)(shrinker, &sc); #endif + /* + * Fall back to zfs_prune_aliases if kernel's shrinker did nothing + * due to dentry and inode caches being referenced by a task running + * in non-root memcg. + */ + if (*objects == 0) + *objects = zfs_prune_aliases(zfsvfs, nr_to_scan); + zfs_exit(zfsvfs, FTAG); dprintf_ds(zfsvfs->z_os->os_dsl_dataset, @@ -1611,7 +1671,7 @@ zfs_umount(struct super_block *sb) if (zfsvfs->z_arc_prune != NULL) arc_remove_prune_callback(zfsvfs->z_arc_prune); - VERIFY(zfsvfs_teardown(zfsvfs, B_TRUE) == 0); + VERIFY0(zfsvfs_teardown(zfsvfs, B_TRUE)); os = zfsvfs->z_os; /* @@ -1737,8 +1797,8 @@ zfs_vget(struct super_block *sb, struct inode **ipp, fid_t *fidp) ASSERT(*ipp != NULL); if (object == ZFSCTL_INO_SNAPDIR) { - VERIFY(zfsctl_root_lookup(*ipp, "snapshot", ipp, - 0, kcred, NULL, NULL) == 0); + VERIFY0(zfsctl_root_lookup(*ipp, "snapshot", ipp, + 0, kcred, NULL, NULL)); } else { /* * Must have an existing ref, so igrab() @@ -1840,7 +1900,7 @@ zfs_resume_fs(zfsvfs_t *zfsvfs, dsl_dataset_t *ds) goto bail; ds->ds_dir->dd_activity_cancelled = B_FALSE; - VERIFY(zfsvfs_setup(zfsvfs, B_FALSE) == 0); + VERIFY0(zfsvfs_setup(zfsvfs, B_FALSE)); zfs_set_fuid_feature(zfsvfs); zfsvfs->z_rollback_time = jiffies; @@ -2013,7 +2073,7 @@ zfs_set_version(zfsvfs_t *zfsvfs, uint64_t newvers) ZFS_SA_ATTRS, 8, 1, &sa_obj, tx); ASSERT0(error); - VERIFY(0 == sa_set_sa_object(os, sa_obj)); + VERIFY0(sa_set_sa_object(os, sa_obj)); sa_register_update_callback(os, zfs_sa_upgrade); } diff --git a/sys/contrib/openzfs/module/os/linux/zfs/zfs_vnops_os.c b/sys/contrib/openzfs/module/os/linux/zfs/zfs_vnops_os.c index ed9721dade76..6106726651a3 100644 --- a/sys/contrib/openzfs/module/os/linux/zfs/zfs_vnops_os.c +++ b/sys/contrib/openzfs/module/os/linux/zfs/zfs_vnops_os.c @@ -25,6 +25,7 @@ * Copyright (c) 2012, 2018 by Delphix. All rights reserved. * Copyright (c) 2015 by Chunwei Chen. All rights reserved. * Copyright 2017 Nexenta Systems, Inc. + * Copyright (c) 2025, Klara, Inc. */ /* Portions Copyright 2007 Jeremy Teo */ @@ -840,8 +841,8 @@ out: *zpp = zp; } - if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) - zil_commit(zilog, 0); + if (error == 0 && zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) + error = zil_commit(zilog, 0); zfs_exit(zfsvfs, FTAG); return (error); @@ -1202,8 +1203,8 @@ out: zfs_zrele_async(xzp); } - if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) - zil_commit(zilog, 0); + if (error == 0 && zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) + error = zil_commit(zilog, 0); zfs_exit(zfsvfs, FTAG); return (error); @@ -1391,14 +1392,15 @@ out: zfs_dirent_unlock(dl); - if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) - zil_commit(zilog, 0); - if (error != 0) { zrele(zp); } else { zfs_znode_update_vfs(dzp); zfs_znode_update_vfs(zp); + + if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) + error = zil_commit(zilog, 0); + } zfs_exit(zfsvfs, FTAG); return (error); @@ -1527,8 +1529,8 @@ out: zfs_znode_update_vfs(zp); zrele(zp); - if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) - zil_commit(zilog, 0); + if (error == 0 && zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) + error = zil_commit(zilog, 0); zfs_exit(zfsvfs, FTAG); return (error); @@ -2482,10 +2484,10 @@ top: new_mode = zp->z_mode; } err = zfs_acl_chown_setattr(zp); - ASSERT(err == 0); + ASSERT0(err); if (attrzp) { err = zfs_acl_chown_setattr(attrzp); - ASSERT(err == 0); + ASSERT0(err); } } @@ -2599,7 +2601,7 @@ out: if (err == 0 && xattr_count > 0) { err2 = sa_bulk_update(attrzp->z_sa_hdl, xattr_bulk, xattr_count, tx); - ASSERT(err2 == 0); + ASSERT0(err2); } if (aclp) @@ -2629,8 +2631,8 @@ out: } out2: - if (os->os_sync == ZFS_SYNC_ALWAYS) - zil_commit(zilog, 0); + if (err == 0 && os->os_sync == ZFS_SYNC_ALWAYS) + err = zil_commit(zilog, 0); out3: kmem_free(xattr_bulk, sizeof (sa_bulk_attr_t) * bulks); @@ -3156,7 +3158,7 @@ top: * zfs_link_create() to add back the same entry, but with a new * dnode (szp), should not fail. */ - ASSERT3P(tzp, ==, NULL); + ASSERT0P(tzp); goto commit_link_tzp; } @@ -3234,8 +3236,8 @@ out: zfs_dirent_unlock(sdl); zfs_dirent_unlock(tdl); - if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) - zil_commit(zilog, 0); + if (error == 0 && zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) + error = zil_commit(zilog, 0); zfs_exit(zfsvfs, FTAG); return (error); @@ -3435,7 +3437,7 @@ top: *zpp = zp; if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) - zil_commit(zilog, 0); + error = zil_commit(zilog, 0); } else { zrele(zp); } @@ -3653,8 +3655,8 @@ top: * operation are sync safe. */ if (is_tmpfile) { - VERIFY(zap_remove_int(zfsvfs->z_os, - zfsvfs->z_unlinkedobj, szp->z_id, tx) == 0); + VERIFY0(zap_remove_int(zfsvfs->z_os, + zfsvfs->z_unlinkedobj, szp->z_id, tx)); } else { if (flags & FIGNORECASE) txtype |= TX_CI; @@ -3669,18 +3671,20 @@ top: zfs_dirent_unlock(dl); - if (!is_tmpfile && zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) - zil_commit(zilog, 0); - - if (is_tmpfile && zfsvfs->z_os->os_sync != ZFS_SYNC_DISABLED) { - txg_wait_flag_t wait_flags = - spa_get_failmode(dmu_objset_spa(zfsvfs->z_os)) == - ZIO_FAILURE_MODE_CONTINUE ? TXG_WAIT_SUSPEND : 0; - error = txg_wait_synced_flags(dmu_objset_pool(zfsvfs->z_os), - txg, wait_flags); - if (error != 0) { - ASSERT3U(error, ==, ESHUTDOWN); - error = SET_ERROR(EIO); + if (error == 0) { + if (!is_tmpfile && zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) + error = zil_commit(zilog, 0); + + if (is_tmpfile && zfsvfs->z_os->os_sync != ZFS_SYNC_DISABLED) { + txg_wait_flag_t wait_flags = + spa_get_failmode(dmu_objset_spa(zfsvfs->z_os)) == + ZIO_FAILURE_MODE_CONTINUE ? TXG_WAIT_SUSPEND : 0; + error = txg_wait_synced_flags( + dmu_objset_pool(zfsvfs->z_os), txg, wait_flags); + if (error != 0) { + ASSERT3U(error, ==, ESHUTDOWN); + error = SET_ERROR(EIO); + } } } @@ -3690,24 +3694,39 @@ top: return (error); } -static void -zfs_putpage_sync_commit_cb(void *arg) +/* Finish page writeback. */ +static inline void +zfs_page_writeback_done(struct page *pp, int err) { - struct page *pp = arg; + if (err != 0) { + /* + * Writeback failed. Re-dirty the page. It was undirtied before + * the IO was issued (in zfs_putpage() or write_cache_pages()). + * The kernel only considers writeback for dirty pages; if we + * don't do this, it is eligible for eviction without being + * written out, which we definitely don't want. + */ +#ifdef HAVE_VFS_FILEMAP_DIRTY_FOLIO + filemap_dirty_folio(page_mapping(pp), page_folio(pp)); +#else + __set_page_dirty_nobuffers(pp); +#endif + } ClearPageError(pp); end_page_writeback(pp); } +/* + * ZIL callback for page writeback. Passes to zfs_log_write() in zfs_putpage() + * for syncing writes. Called when the ZIL itx has been written to the log or + * the whole txg syncs, or if the ZIL crashes or the pool suspends. Any failure + * is passed as `err`. + */ static void -zfs_putpage_async_commit_cb(void *arg) +zfs_putpage_commit_cb(void *arg, int err) { - struct page *pp = arg; - znode_t *zp = ITOZ(pp->mapping->host); - - ClearPageError(pp); - end_page_writeback(pp); - atomic_dec_32(&zp->z_async_writes_cnt); + zfs_page_writeback_done(arg, err); } /* @@ -3827,15 +3846,6 @@ zfs_putpage(struct inode *ip, struct page *pp, struct writeback_control *wbc, zfs_rangelock_exit(lr); if (wbc->sync_mode != WB_SYNC_NONE) { - /* - * Speed up any non-sync page writebacks since - * they may take several seconds to complete. - * Refer to the comment in zpl_fsync() for details. - */ - if (atomic_load_32(&zp->z_async_writes_cnt) > 0) { - zil_commit(zfsvfs->z_log, zp->z_id); - } - if (PageWriteback(pp)) #ifdef HAVE_PAGEMAP_FOLIO_WAIT_BIT folio_wait_bit(page_folio(pp), PG_writeback); @@ -3861,8 +3871,6 @@ zfs_putpage(struct inode *ip, struct page *pp, struct writeback_control *wbc, * was in fact not skipped and should not be counted as if it were. */ wbc->pages_skipped--; - if (!for_sync) - atomic_inc_32(&zp->z_async_writes_cnt); set_page_writeback(pp); unlock_page(pp); @@ -3874,18 +3882,15 @@ zfs_putpage(struct inode *ip, struct page *pp, struct writeback_control *wbc, err = dmu_tx_assign(tx, DMU_TX_WAIT); if (err != 0) { dmu_tx_abort(tx); -#ifdef HAVE_VFS_FILEMAP_DIRTY_FOLIO - filemap_dirty_folio(page_mapping(pp), page_folio(pp)); -#else - __set_page_dirty_nobuffers(pp); -#endif - ClearPageError(pp); - end_page_writeback(pp); - if (!for_sync) - atomic_dec_32(&zp->z_async_writes_cnt); + zfs_page_writeback_done(pp, err); zfs_rangelock_exit(lr); zfs_exit(zfsvfs, FTAG); - return (err); + + /* + * Don't return error for an async writeback; we've re-dirtied + * the page so it will be tried again some other time. + */ + return (for_sync ? err : 0); } va = kmap(pp); @@ -3908,36 +3913,70 @@ zfs_putpage(struct inode *ip, struct page *pp, struct writeback_control *wbc, err = sa_bulk_update(zp->z_sa_hdl, bulk, cnt, tx); - boolean_t commit = B_FALSE; - if (wbc->sync_mode != WB_SYNC_NONE) { - /* - * Note that this is rarely called under writepages(), because - * writepages() normally handles the entire commit for - * performance reasons. - */ - commit = B_TRUE; - } else if (!for_sync && atomic_load_32(&zp->z_sync_writes_cnt) > 0) { + /* + * A note about for_sync vs wbc->sync_mode. + * + * for_sync indicates that this is a syncing writeback, that is, kernel + * caller expects the data to be durably stored before being notified. + * Often, but not always, the call was triggered by a userspace syncing + * op (eg fsync(), msync(MS_SYNC)). For our purposes, for_sync==TRUE + * means that that page should remain "locked" (in the writeback state) + * until it is definitely on disk (ie zil_commit() or spa_sync()). + * Otherwise, we can unlock and return as soon as it is on the + * in-memory ZIL. + * + * wbc->sync_mode has similar meaning. wbc is passed from the kernel to + * zpl_writepages()/zpl_writepage(); wbc->sync_mode==WB_SYNC_NONE + * indicates this a regular async writeback (eg a cache eviction) and + * so does not need a durability guarantee, while WB_SYNC_ALL indicates + * a syncing op that must be waited on (by convention, we test for + * !WB_SYNC_NONE rather than WB_SYNC_ALL, to prefer durability over + * performance should there ever be a new mode that we have not yet + * added support for). + * + * So, why a separate for_sync field? This is because zpl_writepages() + * calls zfs_putpage() multiple times for a single "logical" operation. + * It wants all the individual pages to be for_sync==TRUE ie only + * unlocked once durably stored, but it only wants one call to + * zil_commit() at the very end, once all the pages are synced. So, + * it repurposes sync_mode slightly to indicate who issue and wait for + * the IO: for NONE, the caller to zfs_putpage() will do it, while for + * ALL, zfs_putpage should do it. + * + * Summary: + * for_sync: 0=unlock immediately; 1=unlock once on disk + * sync_mode: NONE=caller will commit; ALL=we will commit + */ + boolean_t need_commit = (wbc->sync_mode != WB_SYNC_NONE); + + /* + * We use for_sync as the "commit" arg to zfs_log_write() (arg 7) + * because it is a policy flag that indicates "someone will call + * zil_commit() soon". for_sync=TRUE means exactly that; the only + * question is whether it will be us, or zpl_writepages(). + */ + zfs_log_write(zfsvfs->z_log, tx, TX_WRITE, zp, pgoff, pglen, for_sync, + B_FALSE, for_sync ? zfs_putpage_commit_cb : NULL, pp); + + if (!for_sync) { /* - * If the caller does not intend to wait synchronously - * for this page writeback to complete and there are active - * synchronous calls on this file, do a commit so that - * the latter don't accidentally end up waiting for - * our writeback to complete. Refer to the comment in - * zpl_fsync() (when HAVE_FSYNC_RANGE is defined) for details. + * Async writeback is logged and written to the DMU, so page + * can now be unlocked. */ - commit = B_TRUE; + zfs_page_writeback_done(pp, 0); } - zfs_log_write(zfsvfs->z_log, tx, TX_WRITE, zp, pgoff, pglen, commit, - B_FALSE, for_sync ? zfs_putpage_sync_commit_cb : - zfs_putpage_async_commit_cb, pp); - dmu_tx_commit(tx); zfs_rangelock_exit(lr); - if (commit) - zil_commit(zfsvfs->z_log, zp->z_id); + if (need_commit) { + err = zil_commit_flags(zfsvfs->z_log, zp->z_id, ZIL_COMMIT_NOW); + if (err != 0) { + zfs_exit(zfsvfs, FTAG); + return (err); + } + } dataset_kstats_update_write_kstats(&zfsvfs->z_kstat, pglen); diff --git a/sys/contrib/openzfs/module/os/linux/zfs/zfs_znode_os.c b/sys/contrib/openzfs/module/os/linux/zfs/zfs_znode_os.c index 54e60b4820f6..bcaabeb32b8a 100644 --- a/sys/contrib/openzfs/module/os/linux/zfs/zfs_znode_os.c +++ b/sys/contrib/openzfs/module/os/linux/zfs/zfs_znode_os.c @@ -126,8 +126,6 @@ zfs_znode_cache_constructor(void *buf, void *arg, int kmflags) zp->z_acl_cached = NULL; zp->z_xattr_cached = NULL; zp->z_xattr_parent = 0; - zp->z_sync_writes_cnt = 0; - zp->z_async_writes_cnt = 0; return (0); } @@ -146,12 +144,9 @@ zfs_znode_cache_destructor(void *buf, void *arg) rw_destroy(&zp->z_xattr_lock); zfs_rangelock_fini(&zp->z_rangelock); - ASSERT3P(zp->z_dirlocks, ==, NULL); - ASSERT3P(zp->z_acl_cached, ==, NULL); - ASSERT3P(zp->z_xattr_cached, ==, NULL); - - ASSERT0(atomic_load_32(&zp->z_sync_writes_cnt)); - ASSERT0(atomic_load_32(&zp->z_async_writes_cnt)); + ASSERT0P(zp->z_dirlocks); + ASSERT0P(zp->z_acl_cached); + ASSERT0P(zp->z_xattr_cached); } static int @@ -183,13 +178,13 @@ zfs_znode_init(void) * backed by kmalloc() when on the Linux slab in order that any * wait_on_bit() operations on the related inode operate properly. */ - ASSERT(znode_cache == NULL); + ASSERT0P(znode_cache); znode_cache = kmem_cache_create("zfs_znode_cache", sizeof (znode_t), 0, zfs_znode_cache_constructor, zfs_znode_cache_destructor, NULL, NULL, NULL, KMC_SLAB | KMC_RECLAIMABLE); - ASSERT(znode_hold_cache == NULL); + ASSERT0P(znode_hold_cache); znode_hold_cache = kmem_cache_create("zfs_znode_hold_cache", sizeof (znode_hold_t), 0, zfs_znode_hold_cache_constructor, zfs_znode_hold_cache_destructor, NULL, NULL, NULL, 0); @@ -332,10 +327,10 @@ zfs_znode_sa_init(zfsvfs_t *zfsvfs, znode_t *zp, mutex_enter(&zp->z_lock); - ASSERT(zp->z_sa_hdl == NULL); - ASSERT(zp->z_acl_cached == NULL); + ASSERT0P(zp->z_sa_hdl); + ASSERT0P(zp->z_acl_cached); if (sa_hdl == NULL) { - VERIFY(0 == sa_handle_get_from_db(zfsvfs->z_os, db, zp, + VERIFY0(sa_handle_get_from_db(zfsvfs->z_os, db, zp, SA_HDL_SHARED, &zp->z_sa_hdl)); } else { zp->z_sa_hdl = sa_hdl; @@ -371,6 +366,12 @@ zfs_inode_alloc(struct super_block *sb, struct inode **ip) return (0); } +void +zfs_inode_free(struct inode *ip) +{ + kmem_cache_free(znode_cache, ITOZ(ip)); +} + /* * Called in multiple places when an inode should be destroyed. */ @@ -395,8 +396,15 @@ zfs_inode_destroy(struct inode *ip) nvlist_free(zp->z_xattr_cached); zp->z_xattr_cached = NULL; } - - kmem_cache_free(znode_cache, zp); +#ifndef HAVE_SOPS_FREE_INODE + /* + * inode needs to be freed in RCU callback. If we have + * super_operations->free_inode, Linux kernel will do call_rcu + * for us. But if we don't have it, since call_rcu is GPL-only + * symbol, we can only free synchronously and accept the risk. + */ + zfs_inode_free(ip); +#endif } static void @@ -522,9 +530,9 @@ zfs_znode_alloc(zfsvfs_t *zfsvfs, dmu_buf_t *db, int blksz, return (NULL); zp = ITOZ(ip); - ASSERT(zp->z_dirlocks == NULL); - ASSERT3P(zp->z_acl_cached, ==, NULL); - ASSERT3P(zp->z_xattr_cached, ==, NULL); + ASSERT0P(zp->z_dirlocks); + ASSERT0P(zp->z_acl_cached); + ASSERT0P(zp->z_xattr_cached); zp->z_unlinked = B_FALSE; zp->z_atime_dirty = B_FALSE; zp->z_is_ctldir = B_FALSE; @@ -535,8 +543,6 @@ zfs_znode_alloc(zfsvfs_t *zfsvfs, dmu_buf_t *db, int blksz, zp->z_blksz = blksz; zp->z_seq = 0x7A4653; zp->z_sync_cnt = 0; - zp->z_sync_writes_cnt = 0; - zp->z_async_writes_cnt = 0; zfs_znode_sa_init(zfsvfs, zp, db, obj_type, hdl); @@ -605,7 +611,7 @@ zfs_znode_alloc(zfsvfs_t *zfsvfs, dmu_buf_t *db, int blksz, * processing so do not hash unlinked znodes. */ if (links > 0) - VERIFY3S(insert_inode_locked(ip), ==, 0); + VERIFY0(insert_inode_locked(ip)); mutex_enter(&zfsvfs->z_znodes_lock); list_insert_tail(&zfsvfs->z_all_znodes, zp); @@ -805,7 +811,7 @@ zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr, } /* Now add in all of the "SA" attributes */ - VERIFY(0 == sa_handle_get_from_db(zfsvfs->z_os, db, NULL, SA_HDL_SHARED, + VERIFY0(sa_handle_get_from_db(zfsvfs->z_os, db, NULL, SA_HDL_SHARED, &sa_hdl)); /* @@ -895,7 +901,7 @@ zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr, acl_ids->z_fuid, acl_ids->z_fgid); } - VERIFY(sa_replace_all_by_template(sa_hdl, sa_attrs, cnt, tx) == 0); + VERIFY0(sa_replace_all_by_template(sa_hdl, sa_attrs, cnt, tx)); if (!(flag & IS_ROOT_NODE)) { /* @@ -1194,7 +1200,7 @@ zfs_rezget(znode_t *zp) } rw_exit(&zp->z_xattr_lock); - ASSERT(zp->z_sa_hdl == NULL); + ASSERT0P(zp->z_sa_hdl); err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db); if (err) { zfs_znode_hold_exit(zfsvfs, zh); @@ -1308,9 +1314,9 @@ zfs_znode_delete(znode_t *zp, dmu_tx_t *tx) zh = zfs_znode_hold_enter(zfsvfs, obj); if (acl_obj) { VERIFY(!zp->z_is_sa); - VERIFY(0 == dmu_object_free(os, acl_obj, tx)); + VERIFY0(dmu_object_free(os, acl_obj, tx)); } - VERIFY(0 == dmu_object_free(os, obj, tx)); + VERIFY0(dmu_object_free(os, obj, tx)); zfs_znode_dmu_fini(zp); zfs_znode_hold_exit(zfsvfs, zh); } @@ -1530,7 +1536,7 @@ zfs_extend(znode_t *zp, uint64_t end) zp->z_size = end; - VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(ZTOZSB(zp)), + VERIFY0(sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(ZTOZSB(zp)), &zp->z_size, sizeof (zp->z_size), tx)); zfs_rangelock_exit(lr); @@ -1720,7 +1726,7 @@ zfs_trunc(znode_t *zp, uint64_t end) SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, &zp->z_pflags, 8); } - VERIFY(sa_bulk_update(zp->z_sa_hdl, bulk, count, tx) == 0); + VERIFY0(sa_bulk_update(zp->z_sa_hdl, bulk, count, tx)); dmu_tx_commit(tx); zfs_rangelock_exit(lr); @@ -1787,7 +1793,7 @@ log: NULL, &zp->z_pflags, 8); zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime); error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); - ASSERT(error == 0); + ASSERT0(error); zfs_log_truncate(zilog, tx, TX_TRUNCATE, zp, off, len); @@ -1834,7 +1840,7 @@ zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx) moid = MASTER_NODE_OBJ; error = zap_create_claim(os, moid, DMU_OT_MASTER_NODE, DMU_OT_NONE, 0, tx); - ASSERT(error == 0); + ASSERT0(error); /* * Set starting attributes. @@ -1847,7 +1853,7 @@ zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx) const char *name; ASSERT(nvpair_type(elem) == DATA_TYPE_UINT64); - VERIFY(nvpair_value_uint64(elem, &val) == 0); + VERIFY0(nvpair_value_uint64(elem, &val)); name = nvpair_name(elem); if (strcmp(name, zfs_prop_to_name(ZFS_PROP_VERSION)) == 0) { if (val < version) @@ -1855,7 +1861,7 @@ zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx) } else { error = zap_update(os, moid, name, 8, 1, &val, tx); } - ASSERT(error == 0); + ASSERT0(error); if (strcmp(name, zfs_prop_to_name(ZFS_PROP_NORMALIZE)) == 0) norm = val; else if (strcmp(name, zfs_prop_to_name(ZFS_PROP_CASE)) == 0) @@ -1863,7 +1869,7 @@ zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx) } ASSERT(version != 0); error = zap_update(os, moid, ZPL_VERSION_STR, 8, 1, &version, tx); - ASSERT(error == 0); + ASSERT0(error); /* * Create zap object used for SA attribute registration @@ -1873,7 +1879,7 @@ zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx) sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE, DMU_OT_NONE, 0, tx); error = zap_add(os, moid, ZFS_SA_ATTRS, 8, 1, &sa_obj, tx); - ASSERT(error == 0); + ASSERT0(error); } else { sa_obj = 0; } @@ -1883,7 +1889,7 @@ zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx) obj = zap_create(os, DMU_OT_UNLINKED_SET, DMU_OT_NONE, 0, tx); error = zap_add(os, moid, ZFS_UNLINKED_SET, 8, 1, &obj, tx); - ASSERT(error == 0); + ASSERT0(error); /* * Create root znode. Create minimal znode/inode/zfsvfs/sb @@ -1916,7 +1922,7 @@ zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx) error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END, &zfsvfs->z_attr_table); - ASSERT(error == 0); + ASSERT0(error); /* * Fold case on file systems that are always or sometimes case @@ -1940,12 +1946,12 @@ zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx) mutex_init(&zfsvfs->z_hold_locks[i], NULL, MUTEX_DEFAULT, NULL); } - VERIFY(0 == zfs_acl_ids_create(rootzp, IS_ROOT_NODE, &vattr, + VERIFY0(zfs_acl_ids_create(rootzp, IS_ROOT_NODE, &vattr, cr, NULL, &acl_ids, zfs_init_idmap)); zfs_mknode(rootzp, &vattr, tx, cr, IS_ROOT_NODE, &zp, &acl_ids); ASSERT3P(zp, ==, rootzp); error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &rootzp->z_id, tx); - ASSERT(error == 0); + ASSERT0(error); zfs_acl_ids_free(&acl_ids); atomic_set(&ZTOI(rootzp)->i_count, 0); diff --git a/sys/contrib/openzfs/module/os/linux/zfs/zpl_file.c b/sys/contrib/openzfs/module/os/linux/zfs/zpl_file.c index 1a82c13e1523..d07317b0d910 100644 --- a/sys/contrib/openzfs/module/os/linux/zfs/zpl_file.c +++ b/sys/contrib/openzfs/module/os/linux/zfs/zpl_file.c @@ -22,6 +22,7 @@ /* * Copyright (c) 2011, Lawrence Livermore National Security, LLC. * Copyright (c) 2015 by Chunwei Chen. All rights reserved. + * Copyright (c) 2025, Klara, Inc. */ @@ -106,60 +107,52 @@ zpl_iterate(struct file *filp, struct dir_context *ctx) return (error); } +static inline int +zpl_write_cache_pages(struct address_space *mapping, + struct writeback_control *wbc, void *data); + static int zpl_fsync(struct file *filp, loff_t start, loff_t end, int datasync) { struct inode *inode = filp->f_mapping->host; znode_t *zp = ITOZ(inode); - zfsvfs_t *zfsvfs = ITOZSB(inode); cred_t *cr = CRED(); int error; fstrans_cookie_t cookie; /* - * The variables z_sync_writes_cnt and z_async_writes_cnt work in - * tandem so that sync writes can detect if there are any non-sync - * writes going on and vice-versa. The "vice-versa" part to this logic - * is located in zfs_putpage() where non-sync writes check if there are - * any ongoing sync writes. If any sync and non-sync writes overlap, - * we do a commit to complete the non-sync writes since the latter can - * potentially take several seconds to complete and thus block sync - * writes in the upcoming call to filemap_write_and_wait_range(). - */ - atomic_inc_32(&zp->z_sync_writes_cnt); - /* - * If the following check does not detect an overlapping non-sync write - * (say because it's just about to start), then it is guaranteed that - * the non-sync write will detect this sync write. This is because we - * always increment z_sync_writes_cnt / z_async_writes_cnt before doing - * the check on z_async_writes_cnt / z_sync_writes_cnt here and in - * zfs_putpage() respectively. + * Force dirty pages in the range out to the DMU and the log, ready + * for zil_commit() to write down. + * + * We call write_cache_pages() directly to ensure that zpl_putpage() is + * called with the flags we need. We need WB_SYNC_NONE to avoid a call + * to zil_commit() (since we're doing this as a kind of pre-sync); but + * we do need for_sync so that the pages remain in writeback until + * they're on disk, and so that we get an error if the DMU write fails. */ - if (atomic_load_32(&zp->z_async_writes_cnt) > 0) { - if ((error = zpl_enter(zfsvfs, FTAG)) != 0) { - atomic_dec_32(&zp->z_sync_writes_cnt); + if (filemap_range_has_page(inode->i_mapping, start, end)) { + int for_sync = 1; + struct writeback_control wbc = { + .sync_mode = WB_SYNC_NONE, + .nr_to_write = LONG_MAX, + .range_start = start, + .range_end = end, + }; + error = + zpl_write_cache_pages(inode->i_mapping, &wbc, &for_sync); + if (error != 0) { + /* + * Unclear what state things are in. zfs_putpage() will + * ensure the pages remain dirty if they haven't been + * written down to the DMU, but because there may be + * nothing logged, we can't assume that zfs_sync() -> + * zil_commit() will give us a useful error. It's + * safest if we just error out here. + */ return (error); } - zil_commit(zfsvfs->z_log, zp->z_id); - zpl_exit(zfsvfs, FTAG); } - error = filemap_write_and_wait_range(inode->i_mapping, start, end); - - /* - * The sync write is not complete yet but we decrement - * z_sync_writes_cnt since zfs_fsync() increments and decrements - * it internally. If a non-sync write starts just after the decrement - * operation but before we call zfs_fsync(), it may not detect this - * overlapping sync write but it does not matter since we have already - * gone past filemap_write_and_wait_range() and we won't block due to - * the non-sync write. - */ - atomic_dec_32(&zp->z_sync_writes_cnt); - - if (error) - return (error); - crhold(cr); cookie = spl_fstrans_mark(); error = -zfs_fsync(zp, datasync, cr); @@ -535,11 +528,30 @@ zpl_writepages(struct address_space *mapping, struct writeback_control *wbc) if (sync_mode != wbc->sync_mode) { if ((result = zpl_enter_verify_zp(zfsvfs, zp, FTAG)) != 0) return (result); - if (zfsvfs->z_log != NULL) - zil_commit(zfsvfs->z_log, zp->z_id); + + if (zfsvfs->z_log != NULL) { + /* + * We don't want to block here if the pool suspends, + * because this is not a syncing op by itself, but + * might be part of one that the caller will + * coordinate. + */ + result = -zil_commit_flags(zfsvfs->z_log, zp->z_id, + ZIL_COMMIT_NOW); + } + zpl_exit(zfsvfs, FTAG); /* + * If zil_commit_flags() failed, it's unclear what state things + * are currently in. putpage() has written back out what it can + * to the DMU, but it may not be on disk. We have little choice + * but to escape. + */ + if (result != 0) + return (result); + + /* * We need to call write_cache_pages() again (we can't just * return after the commit) because the previous call in * non-SYNC mode does not guarantee that we got all the dirty diff --git a/sys/contrib/openzfs/module/os/linux/zfs/zpl_inode.c b/sys/contrib/openzfs/module/os/linux/zfs/zpl_inode.c index f9f6406f8b47..f97662d052c7 100644 --- a/sys/contrib/openzfs/module/os/linux/zfs/zpl_inode.c +++ b/sys/contrib/openzfs/module/os/linux/zfs/zpl_inode.c @@ -247,7 +247,7 @@ zpl_mknod(struct inode *dir, struct dentry *dentry, umode_t mode, * and fifos, but we want to know if this behavior ever changes. */ if (S_ISSOCK(mode) || S_ISFIFO(mode)) - ASSERT(rdev == 0); + ASSERT0(rdev); crhold(cr); vap = kmem_zalloc(sizeof (vattr_t), KM_SLEEP); diff --git a/sys/contrib/openzfs/module/os/linux/zfs/zpl_super.c b/sys/contrib/openzfs/module/os/linux/zfs/zpl_super.c index a682bfd33c38..53819628627d 100644 --- a/sys/contrib/openzfs/module/os/linux/zfs/zpl_super.c +++ b/sys/contrib/openzfs/module/os/linux/zfs/zpl_super.c @@ -45,10 +45,19 @@ zpl_inode_alloc(struct super_block *sb) return (ip); } +#ifdef HAVE_SOPS_FREE_INODE +static void +zpl_inode_free(struct inode *ip) +{ + ASSERT0(atomic_read(&ip->i_count)); + zfs_inode_free(ip); +} +#endif + static void zpl_inode_destroy(struct inode *ip) { - ASSERT(atomic_read(&ip->i_count) == 0); + ASSERT0(atomic_read(&ip->i_count)); zfs_inode_destroy(ip); } @@ -455,6 +464,9 @@ zpl_prune_sb(uint64_t nr_to_scan, void *arg) const struct super_operations zpl_super_operations = { .alloc_inode = zpl_inode_alloc, +#ifdef HAVE_SOPS_FREE_INODE + .free_inode = zpl_inode_free, +#endif .destroy_inode = zpl_inode_destroy, .dirty_inode = zpl_dirty_inode, .write_inode = NULL, diff --git a/sys/contrib/openzfs/module/os/linux/zfs/zpl_xattr.c b/sys/contrib/openzfs/module/os/linux/zfs/zpl_xattr.c index a098197e7448..d93282db815a 100644 --- a/sys/contrib/openzfs/module/os/linux/zfs/zpl_xattr.c +++ b/sys/contrib/openzfs/module/os/linux/zfs/zpl_xattr.c @@ -1494,7 +1494,7 @@ zpl_posix_acl_free(void *arg) acl_rel_head = NULL; if (cmpxchg(&acl_rel_tail, &a->next, &acl_rel_head) == &a->next) { - ASSERT3P(a->next, ==, NULL); + ASSERT0P(a->next); a->next = freelist; freelist = a; break; @@ -1544,7 +1544,7 @@ zpl_posix_acl_release_impl(struct posix_acl *acl) a->time = ddi_get_lbolt(); /* atomically points tail to us and get the previous tail */ prev = xchg(&acl_rel_tail, &a->next); - ASSERT3P(*prev, ==, NULL); + ASSERT0P(*prev); *prev = a; /* if it was empty before, schedule the free task */ if (prev == &acl_rel_head) diff --git a/sys/contrib/openzfs/module/os/linux/zfs/zvol_os.c b/sys/contrib/openzfs/module/os/linux/zfs/zvol_os.c index 57a9711e9027..a73acdad34ae 100644 --- a/sys/contrib/openzfs/module/os/linux/zfs/zvol_os.c +++ b/sys/contrib/openzfs/module/os/linux/zfs/zvol_os.c @@ -84,8 +84,9 @@ static unsigned int zvol_blk_mq_blocks_per_thread = 8; static inline void zvol_end_io(struct bio *bio, struct request *rq, int error) { + ASSERT3U(error, >=, 0); if (bio) { - bio->bi_status = errno_to_bi_status(-error); + bio->bi_status = errno_to_bi_status(error); bio_endio(bio); } else { blk_mq_end_request(rq, errno_to_bi_status(error)); @@ -208,8 +209,14 @@ zvol_write(zv_request_t *zvr) disk = zv->zv_zso->zvo_disk; /* bio marked as FLUSH need to flush before write */ - if (io_is_flush(bio, rq)) - zil_commit(zv->zv_zilog, ZVOL_OBJ); + if (io_is_flush(bio, rq)) { + error = zil_commit(zv->zv_zilog, ZVOL_OBJ); + if (error != 0) { + rw_exit(&zv->zv_suspend_lock); + zvol_end_io(bio, rq, -error); + return; + } + } /* Some requests are just for flush and nothing else. */ if (io_size(bio, rq) == 0) { @@ -273,8 +280,8 @@ zvol_write(zv_request_t *zvr) dataset_kstats_update_write_kstats(&zv->zv_kstat, nwritten); task_io_account_write(nwritten); - if (sync) - zil_commit(zv->zv_zilog, ZVOL_OBJ); + if (error == 0 && sync) + error = zil_commit(zv->zv_zilog, ZVOL_OBJ); rw_exit(&zv->zv_suspend_lock); @@ -282,7 +289,7 @@ zvol_write(zv_request_t *zvr) blk_generic_end_io_acct(q, disk, WRITE, bio, start_time); } - zvol_end_io(bio, rq, -error); + zvol_end_io(bio, rq, error); } static void @@ -361,7 +368,7 @@ zvol_discard(zv_request_t *zvr) zfs_rangelock_exit(lr); if (error == 0 && sync) - zil_commit(zv->zv_zilog, ZVOL_OBJ); + error = zil_commit(zv->zv_zilog, ZVOL_OBJ); unlock: rw_exit(&zv->zv_suspend_lock); @@ -371,7 +378,7 @@ unlock: start_time); } - zvol_end_io(bio, rq, -error); + zvol_end_io(bio, rq, error); } static void @@ -449,7 +456,7 @@ zvol_read(zv_request_t *zvr) blk_generic_end_io_acct(q, disk, READ, bio, start_time); } - zvol_end_io(bio, rq, -error); + zvol_end_io(bio, rq, error); } static void @@ -480,7 +487,7 @@ zvol_request_impl(zvol_state_t *zv, struct bio *bio, struct request *rq, int rw = io_data_dir(bio, rq); if (unlikely(zv->zv_flags & ZVOL_REMOVING)) { - zvol_end_io(bio, rq, -SET_ERROR(ENXIO)); + zvol_end_io(bio, rq, SET_ERROR(ENXIO)); goto out; } @@ -499,7 +506,7 @@ zvol_request_impl(zvol_state_t *zv, struct bio *bio, struct request *rq, (long long unsigned)offset, (long unsigned)size); - zvol_end_io(bio, rq, -SET_ERROR(EIO)); + zvol_end_io(bio, rq, SET_ERROR(EIO)); goto out; } @@ -512,8 +519,8 @@ zvol_request_impl(zvol_state_t *zv, struct bio *bio, struct request *rq, #ifdef HAVE_BLK_MQ_RQ_HCTX blk_mq_hw_queue = rq->mq_hctx->queue_num; #else - blk_mq_hw_queue = - rq->q->queue_hw_ctx[rq->q->mq_map[rq->cpu]]->queue_num; + blk_mq_hw_queue = rq->q->queue_hw_ctx[ + rq->q->mq_map[raw_smp_processor_id()]]->queue_num; #endif taskq_hash = cityhash3((uintptr_t)zv, offset >> ZVOL_TASKQ_OFFSET_SHIFT, blk_mq_hw_queue); @@ -521,7 +528,7 @@ zvol_request_impl(zvol_state_t *zv, struct bio *bio, struct request *rq, if (rw == WRITE) { if (unlikely(zv->zv_flags & ZVOL_RDONLY)) { - zvol_end_io(bio, rq, -SET_ERROR(EROFS)); + zvol_end_io(bio, rq, SET_ERROR(EROFS)); goto out; } @@ -886,16 +893,18 @@ zvol_ioctl(struct block_device *bdev, fmode_t mode, case BLKZNAME: mutex_enter(&zv->zv_state_lock); - error = copy_to_user((void *)arg, zv->zv_name, MAXNAMELEN); + error = -copy_to_user((void *)arg, zv->zv_name, MAXNAMELEN); mutex_exit(&zv->zv_state_lock); + if (error) + error = SET_ERROR(error); break; default: - error = -ENOTTY; + error = SET_ERROR(ENOTTY); break; } - return (SET_ERROR(error)); + return (-error); } #ifdef CONFIG_COMPAT @@ -1302,27 +1311,30 @@ zvol_alloc_blk_mq(zvol_state_t *zv, zvol_queue_limits_t *limits) * Allocate memory for a new zvol_state_t and setup the required * request queue and generic disk structures for the block device. */ -static zvol_state_t * -zvol_alloc(dev_t dev, const char *name, uint64_t volblocksize) +static int +zvol_alloc(dev_t dev, const char *name, uint64_t volsize, uint64_t volblocksize, + zvol_state_t **zvp) { zvol_state_t *zv; struct zvol_state_os *zso; uint64_t volmode; int ret; - if (dsl_prop_get_integer(name, "volmode", &volmode, NULL) != 0) - return (NULL); + ret = dsl_prop_get_integer(name, "volmode", &volmode, NULL); + if (ret) + return (ret); if (volmode == ZFS_VOLMODE_DEFAULT) volmode = zvol_volmode; if (volmode == ZFS_VOLMODE_NONE) - return (NULL); + return (0); zv = kmem_zalloc(sizeof (zvol_state_t), KM_SLEEP); zso = kmem_zalloc(sizeof (struct zvol_state_os), KM_SLEEP); zv->zv_zso = zso; zv->zv_volmode = volmode; + zv->zv_volsize = volsize; zv->zv_volblocksize = volblocksize; list_link_init(&zv->zv_next); @@ -1396,12 +1408,13 @@ zvol_alloc(dev_t dev, const char *name, uint64_t volblocksize) snprintf(zso->zvo_disk->disk_name, DISK_NAME_LEN, "%s%d", ZVOL_DEV_NAME, (dev & MINORMASK)); - return (zv); + *zvp = zv; + return (ret); out_kmem: kmem_free(zso, sizeof (struct zvol_state_os)); kmem_free(zv, sizeof (zvol_state_t)); - return (NULL); + return (ret); } /* @@ -1422,7 +1435,7 @@ zvol_os_free(zvol_state_t *zv) ASSERT(!RW_LOCK_HELD(&zv->zv_suspend_lock)); ASSERT(!MUTEX_HELD(&zv->zv_state_lock)); ASSERT0(zv->zv_open_count); - ASSERT3P(zv->zv_zso->zvo_disk->private_data, ==, NULL); + ASSERT0P(zv->zv_zso->zvo_disk->private_data); rw_destroy(&zv->zv_suspend_lock); zfs_rangelock_fini(&zv->zv_rangelock); @@ -1470,7 +1483,9 @@ __zvol_os_add_disk(struct gendisk *disk) { int error = 0; #ifdef HAVE_ADD_DISK_RET - error = add_disk(disk); + error = -add_disk(disk); + if (error) + error = SET_ERROR(error); #else add_disk(disk); #endif @@ -1562,7 +1577,7 @@ zvol_os_add_disk(struct gendisk *disk) int zvol_os_create_minor(const char *name) { - zvol_state_t *zv; + zvol_state_t *zv = NULL; objset_t *os; dmu_object_info_t *doi; uint64_t volsize; @@ -1611,18 +1626,16 @@ zvol_os_create_minor(const char *name) if (error) goto out_dmu_objset_disown; - zv = zvol_alloc(MKDEV(zvol_major, minor), name, - doi->doi_data_block_size); - if (zv == NULL) { - error = SET_ERROR(EAGAIN); + error = zvol_alloc(MKDEV(zvol_major, minor), name, + volsize, doi->doi_data_block_size, &zv); + if (error || zv == NULL) goto out_dmu_objset_disown; - } + zv->zv_hash = hash; if (dmu_objset_is_snapshot(os)) zv->zv_flags |= ZVOL_RDONLY; - zv->zv_volsize = volsize; zv->zv_objset = os; /* Default */ @@ -1647,11 +1660,11 @@ zvol_os_create_minor(const char *name) blk_queue_flag_set(QUEUE_FLAG_SCSI_PASSTHROUGH, zv->zv_zso->zvo_queue); #endif - ASSERT3P(zv->zv_kstat.dk_kstats, ==, NULL); + ASSERT0P(zv->zv_kstat.dk_kstats); error = dataset_kstats_create(&zv->zv_kstat, zv->zv_objset); if (error) goto out_dmu_objset_disown; - ASSERT3P(zv->zv_zilog, ==, NULL); + ASSERT0P(zv->zv_zilog); zv->zv_zilog = zil_open(os, zvol_get_data, &zv->zv_kstat.dk_zil_sums); if (spa_writeable(dmu_objset_spa(os))) { if (zil_replay_disable) @@ -1689,7 +1702,7 @@ out_doi: * zvol_open()->zvol_first_open() and zvol_release()->zvol_last_close() * directly as well. */ - if (error == 0) { + if (error == 0 && zv) { rw_enter(&zvol_state_lock, RW_WRITER); zvol_insert(zv); rw_exit(&zvol_state_lock); @@ -1701,7 +1714,7 @@ out_doi: return (error); } -void +int zvol_os_rename_minor(zvol_state_t *zv, const char *newname) { int readonly = get_disk_ro(zv->zv_zso->zvo_disk); @@ -1728,6 +1741,8 @@ zvol_os_rename_minor(zvol_state_t *zv, const char *newname) set_disk_ro(zv->zv_zso->zvo_disk, readonly); dataset_kstats_rename(&zv->zv_kstat, newname); + + return (0); } void @@ -1755,10 +1770,10 @@ zvol_init(void) return (error); } - error = register_blkdev(zvol_major, ZVOL_DRIVER); + error = -register_blkdev(zvol_major, ZVOL_DRIVER); if (error) { printk(KERN_INFO "ZFS: register_blkdev() failed %d\n", error); - return (error); + return (SET_ERROR(error)); } if (zvol_blk_mq_queue_depth == 0) { diff --git a/sys/contrib/openzfs/module/zcommon/zfeature_common.c b/sys/contrib/openzfs/module/zcommon/zfeature_common.c index 0362d82efbc1..6ba9892eeb64 100644 --- a/sys/contrib/openzfs/module/zcommon/zfeature_common.c +++ b/sys/contrib/openzfs/module/zcommon/zfeature_common.c @@ -732,6 +732,12 @@ zpool_feature_init(void) ZFEATURE_FLAG_READONLY_COMPAT, ZFEATURE_TYPE_BOOLEAN, NULL, sfeatures); + zfeature_register(SPA_FEATURE_BLOCK_CLONING_ENDIAN, + "com.truenas:block_cloning_endian", "block_cloning_endian", + "Fixes BRT ZAP endianness on new pools.", + ZFEATURE_FLAG_READONLY_COMPAT, ZFEATURE_TYPE_BOOLEAN, NULL, + sfeatures); + zfeature_register(SPA_FEATURE_AVZ_V2, "com.klarasystems:vdev_zaps_v2", "vdev_zaps_v2", "Support for root vdev ZAP.", @@ -786,6 +792,24 @@ zpool_feature_init(void) ZFEATURE_TYPE_BOOLEAN, large_microzap_deps, sfeatures); } + zfeature_register(SPA_FEATURE_DYNAMIC_GANG_HEADER, + "com.klarasystems:dynamic_gang_header", "dynamic_gang_header", + "Support for dynamically sized gang headers", + ZFEATURE_FLAG_MOS | ZFEATURE_FLAG_NO_UPGRADE, + ZFEATURE_TYPE_BOOLEAN, NULL, sfeatures); + + { + static const spa_feature_t physical_rewrite_deps[] = { + SPA_FEATURE_EXTENSIBLE_DATASET, + SPA_FEATURE_NONE + }; + zfeature_register(SPA_FEATURE_PHYSICAL_REWRITE, + "com.truenas:physical_rewrite", "physical_rewrite", + "Support for preserving logical birth time during rewrite.", + ZFEATURE_FLAG_READONLY_COMPAT | ZFEATURE_FLAG_PER_DATASET, + ZFEATURE_TYPE_BOOLEAN, physical_rewrite_deps, sfeatures); + } + zfs_mod_list_supported_free(sfeatures); } diff --git a/sys/contrib/openzfs/module/zcommon/zfs_prop.c b/sys/contrib/openzfs/module/zcommon/zfs_prop.c index 8b4c4251703d..864e3898b365 100644 --- a/sys/contrib/openzfs/module/zcommon/zfs_prop.c +++ b/sys/contrib/openzfs/module/zcommon/zfs_prop.c @@ -640,7 +640,7 @@ zfs_prop_init(void) "<1.00x or higher if compressed>", "REFRATIO", B_FALSE, sfeatures); zprop_register_number(ZFS_PROP_VOLBLOCKSIZE, "volblocksize", ZVOL_DEFAULT_BLOCKSIZE, PROP_ONETIME, - ZFS_TYPE_VOLUME, "512 to 128k, power of 2", "VOLBLOCK", B_FALSE, + ZFS_TYPE_VOLUME, "512 to 16M, power of 2", "VOLBLOCK", B_FALSE, sfeatures); zprop_register_index(ZFS_PROP_VOLTHREADING, "volthreading", 1, PROP_DEFAULT, ZFS_TYPE_VOLUME, "on | off", "zvol threading", @@ -734,13 +734,12 @@ zfs_prop_init(void) /* inherit number properties */ zprop_register_number(ZFS_PROP_RECORDSIZE, "recordsize", SPA_OLD_MAXBLOCKSIZE, PROP_INHERIT, - ZFS_TYPE_FILESYSTEM, "512 to 1M, power of 2", "RECSIZE", B_FALSE, - sfeatures); + ZFS_TYPE_FILESYSTEM, "512 to 16M, power of 2", + "RECSIZE", B_FALSE, sfeatures); zprop_register_number(ZFS_PROP_SPECIAL_SMALL_BLOCKS, "special_small_blocks", 0, PROP_INHERIT, - ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME, - "zero or 512 to 1M, power of 2", "SPECIAL_SMALL_BLOCKS", B_FALSE, - sfeatures); + ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME, "0 to 16M", + "SPECIAL_SMALL_BLOCKS", B_FALSE, sfeatures); /* hidden properties */ zprop_register_hidden(ZFS_PROP_NUMCLONES, "numclones", PROP_TYPE_NUMBER, diff --git a/sys/contrib/openzfs/module/zcommon/zfs_valstr.c b/sys/contrib/openzfs/module/zcommon/zfs_valstr.c index c39ac62f654f..0cb9f584acc6 100644 --- a/sys/contrib/openzfs/module/zcommon/zfs_valstr.c +++ b/sys/contrib/openzfs/module/zcommon/zfs_valstr.c @@ -203,7 +203,7 @@ _VALSTR_BITFIELD_IMPL(zio_flag, { '?', "??", "[UNUSED 11]" }, { '.', "ND", "NODATA" }, { '.', "ID", "INDUCE_DAMAGE" }, - { '.', "AL", "IO_ALLOCATING" }, + { '.', "AT", "ALLOC_THROTTLED" }, { '.', "RE", "IO_RETRY" }, { '.', "PR", "PROBE" }, { '.', "TH", "TRYHARD" }, @@ -221,7 +221,6 @@ _VALSTR_BITFIELD_IMPL(zio_flag, { '.', "NP", "NOPWRITE" }, { '.', "EX", "REEXECUTED" }, { '.', "DG", "DELEGATED" }, - { '.', "DC", "DIO_CHKSUM_ERR" }, { '.', "PA", "PREALLOCATED" }, ) diff --git a/sys/contrib/openzfs/module/zfs/abd.c b/sys/contrib/openzfs/module/zfs/abd.c index 826928e67350..bf9b13c30509 100644 --- a/sys/contrib/openzfs/module/zfs/abd.c +++ b/sys/contrib/openzfs/module/zfs/abd.c @@ -563,7 +563,7 @@ abd_get_offset_impl(abd_t *abd, abd_t *sabd, size_t off, size_t size) left -= csize; off = 0; } - ASSERT3U(left, ==, 0); + ASSERT0(left); } else { abd = abd_get_offset_scatter(abd, sabd, off, size); } diff --git a/sys/contrib/openzfs/module/zfs/arc.c b/sys/contrib/openzfs/module/zfs/arc.c index 04ca32356a6d..df41e3b49204 100644 --- a/sys/contrib/openzfs/module/zfs/arc.c +++ b/sys/contrib/openzfs/module/zfs/arc.c @@ -1052,7 +1052,7 @@ static arc_buf_hdr_t * buf_hash_find(uint64_t spa, const blkptr_t *bp, kmutex_t **lockp) { const dva_t *dva = BP_IDENTITY(bp); - uint64_t birth = BP_GET_BIRTH(bp); + uint64_t birth = BP_GET_PHYSICAL_BIRTH(bp); uint64_t idx = BUF_HASH_INDEX(spa, dva, birth); kmutex_t *hash_lock = BUF_HASH_LOCK(idx); arc_buf_hdr_t *hdr; @@ -2239,8 +2239,8 @@ arc_evictable_space_increment(arc_buf_hdr_t *hdr, arc_state_t *state) ASSERT(HDR_HAS_L1HDR(hdr)); if (GHOST_STATE(state)) { - ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL); - ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL); + ASSERT0P(hdr->b_l1hdr.b_buf); + ASSERT0P(hdr->b_l1hdr.b_pabd); ASSERT(!HDR_HAS_RABD(hdr)); (void) zfs_refcount_add_many(&state->arcs_esize[type], HDR_GET_LSIZE(hdr), hdr); @@ -2278,8 +2278,8 @@ arc_evictable_space_decrement(arc_buf_hdr_t *hdr, arc_state_t *state) ASSERT(HDR_HAS_L1HDR(hdr)); if (GHOST_STATE(state)) { - ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL); - ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL); + ASSERT0P(hdr->b_l1hdr.b_buf); + ASSERT0P(hdr->b_l1hdr.b_pabd); ASSERT(!HDR_HAS_RABD(hdr)); (void) zfs_refcount_remove_many(&state->arcs_esize[type], HDR_GET_LSIZE(hdr), hdr); @@ -2319,7 +2319,7 @@ add_reference(arc_buf_hdr_t *hdr, const void *tag) if (!HDR_EMPTY(hdr) && !MUTEX_HELD(HDR_LOCK(hdr))) { ASSERT(state == arc_anon); ASSERT(zfs_refcount_is_zero(&hdr->b_l1hdr.b_refcnt)); - ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL); + ASSERT0P(hdr->b_l1hdr.b_buf); } if ((zfs_refcount_add(&hdr->b_l1hdr.b_refcnt, tag) == 1) && @@ -2503,7 +2503,7 @@ arc_change_state(arc_state_t *new_state, arc_buf_hdr_t *hdr) (void) zfs_refcount_add_many( &new_state->arcs_size[type], HDR_GET_LSIZE(hdr), hdr); - ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL); + ASSERT0P(hdr->b_l1hdr.b_pabd); ASSERT(!HDR_HAS_RABD(hdr)); } else { @@ -2547,7 +2547,7 @@ arc_change_state(arc_state_t *new_state, arc_buf_hdr_t *hdr) if (update_old && old_state != arc_l2c_only) { ASSERT(HDR_HAS_L1HDR(hdr)); if (GHOST_STATE(old_state)) { - ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL); + ASSERT0P(hdr->b_l1hdr.b_pabd); ASSERT(!HDR_HAS_RABD(hdr)); /* @@ -2631,7 +2631,7 @@ arc_space_consume(uint64_t space, arc_space_type_t type) ARCSTAT_INCR(arcstat_bonus_size, space); break; case ARC_SPACE_DNODE: - ARCSTAT_INCR(arcstat_dnode_size, space); + aggsum_add(&arc_sums.arcstat_dnode_size, space); break; case ARC_SPACE_DBUF: ARCSTAT_INCR(arcstat_dbuf_size, space); @@ -2677,7 +2677,7 @@ arc_space_return(uint64_t space, arc_space_type_t type) ARCSTAT_INCR(arcstat_bonus_size, -space); break; case ARC_SPACE_DNODE: - ARCSTAT_INCR(arcstat_dnode_size, -space); + aggsum_add(&arc_sums.arcstat_dnode_size, -space); break; case ARC_SPACE_DBUF: ARCSTAT_INCR(arcstat_dbuf_size, -space); @@ -2758,7 +2758,7 @@ arc_buf_alloc_impl(arc_buf_hdr_t *hdr, spa_t *spa, const zbookmark_phys_t *zb, VERIFY(hdr->b_type == ARC_BUFC_DATA || hdr->b_type == ARC_BUFC_METADATA); ASSERT3P(ret, !=, NULL); - ASSERT3P(*ret, ==, NULL); + ASSERT0P(*ret); IMPLY(encrypted, compressed); buf = *ret = kmem_cache_alloc(buf_cache, KM_PUSHPAGE); @@ -2982,7 +2982,7 @@ static void arc_share_buf(arc_buf_hdr_t *hdr, arc_buf_t *buf) { ASSERT(arc_can_share(hdr, buf)); - ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL); + ASSERT0P(hdr->b_l1hdr.b_pabd); ASSERT(!ARC_BUF_ENCRYPTED(buf)); ASSERT(HDR_EMPTY_OR_LOCKED(hdr)); @@ -3201,14 +3201,14 @@ arc_hdr_alloc_abd(arc_buf_hdr_t *hdr, int alloc_flags) if (alloc_rdata) { size = HDR_GET_PSIZE(hdr); - ASSERT3P(hdr->b_crypt_hdr.b_rabd, ==, NULL); + ASSERT0P(hdr->b_crypt_hdr.b_rabd); hdr->b_crypt_hdr.b_rabd = arc_get_data_abd(hdr, size, hdr, alloc_flags); ASSERT3P(hdr->b_crypt_hdr.b_rabd, !=, NULL); ARCSTAT_INCR(arcstat_raw_size, size); } else { size = arc_hdr_size(hdr); - ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL); + ASSERT0P(hdr->b_l1hdr.b_pabd); hdr->b_l1hdr.b_pabd = arc_get_data_abd(hdr, size, hdr, alloc_flags); ASSERT3P(hdr->b_l1hdr.b_pabd, !=, NULL); @@ -3290,7 +3290,7 @@ arc_hdr_alloc(uint64_t spa, int32_t psize, int32_t lsize, ASSERT(HDR_EMPTY(hdr)); #ifdef ZFS_DEBUG - ASSERT3P(hdr->b_l1hdr.b_freeze_cksum, ==, NULL); + ASSERT0P(hdr->b_l1hdr.b_freeze_cksum); #endif HDR_SET_PSIZE(hdr, psize); HDR_SET_LSIZE(hdr, lsize); @@ -3351,12 +3351,12 @@ arc_hdr_realloc(arc_buf_hdr_t *hdr, kmem_cache_t *old, kmem_cache_t *new) nhdr->b_l1hdr.b_state = arc_l2c_only; /* Verify previous threads set to NULL before freeing */ - ASSERT3P(nhdr->b_l1hdr.b_pabd, ==, NULL); + ASSERT0P(nhdr->b_l1hdr.b_pabd); ASSERT(!HDR_HAS_RABD(hdr)); } else { - ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL); + ASSERT0P(hdr->b_l1hdr.b_buf); #ifdef ZFS_DEBUG - ASSERT3P(hdr->b_l1hdr.b_freeze_cksum, ==, NULL); + ASSERT0P(hdr->b_l1hdr.b_freeze_cksum); #endif /* @@ -3375,7 +3375,7 @@ arc_hdr_realloc(arc_buf_hdr_t *hdr, kmem_cache_t *old, kmem_cache_t *new) * might try to be accessed, even though it was removed. */ VERIFY(!HDR_L2_WRITING(hdr)); - VERIFY3P(hdr->b_l1hdr.b_pabd, ==, NULL); + VERIFY0P(hdr->b_l1hdr.b_pabd); ASSERT(!HDR_HAS_RABD(hdr)); arc_hdr_clear_flags(nhdr, ARC_FLAG_HAS_L1HDR); @@ -3698,12 +3698,12 @@ arc_hdr_destroy(arc_buf_hdr_t *hdr) arc_hdr_free_abd(hdr, B_TRUE); } - ASSERT3P(hdr->b_hash_next, ==, NULL); + ASSERT0P(hdr->b_hash_next); if (HDR_HAS_L1HDR(hdr)) { ASSERT(!multilist_link_active(&hdr->b_l1hdr.b_arc_node)); - ASSERT3P(hdr->b_l1hdr.b_acb, ==, NULL); + ASSERT0P(hdr->b_l1hdr.b_acb); #ifdef ZFS_DEBUG - ASSERT3P(hdr->b_l1hdr.b_freeze_cksum, ==, NULL); + ASSERT0P(hdr->b_l1hdr.b_freeze_cksum); #endif kmem_cache_free(hdr_full_cache, hdr); } else { @@ -3771,7 +3771,7 @@ arc_evict_hdr(arc_buf_hdr_t *hdr, uint64_t *real_evicted) ASSERT(MUTEX_HELD(HDR_LOCK(hdr))); ASSERT(HDR_HAS_L1HDR(hdr)); ASSERT(!HDR_IO_IN_PROGRESS(hdr)); - ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL); + ASSERT0P(hdr->b_l1hdr.b_buf); ASSERT0(zfs_refcount_count(&hdr->b_l1hdr.b_refcnt)); *real_evicted = 0; @@ -3796,7 +3796,7 @@ arc_evict_hdr(arc_buf_hdr_t *hdr, uint64_t *real_evicted) DTRACE_PROBE1(arc__delete, arc_buf_hdr_t *, hdr); if (HDR_HAS_L2HDR(hdr)) { - ASSERT(hdr->b_l1hdr.b_pabd == NULL); + ASSERT0P(hdr->b_l1hdr.b_pabd); ASSERT(!HDR_HAS_RABD(hdr)); /* * This buffer is cached on the 2nd Level ARC; @@ -4490,7 +4490,7 @@ arc_evict(void) * target is not evictable or if they go over arc_dnode_limit. */ int64_t prune = 0; - int64_t dn = wmsum_value(&arc_sums.arcstat_dnode_size); + int64_t dn = aggsum_value(&arc_sums.arcstat_dnode_size); int64_t nem = zfs_refcount_count(&arc_mru->arcs_size[ARC_BUFC_METADATA]) + zfs_refcount_count(&arc_mfu->arcs_size[ARC_BUFC_METADATA]) - zfs_refcount_count(&arc_mru->arcs_esize[ARC_BUFC_METADATA]) @@ -5082,11 +5082,13 @@ arc_is_overflowing(boolean_t lax, boolean_t use_reserve) * in the ARC. In practice, that's in the tens of MB, which is low * enough to be safe. */ - int64_t over = aggsum_lower_bound(&arc_sums.arcstat_size) - arc_c - + int64_t arc_over = aggsum_lower_bound(&arc_sums.arcstat_size) - arc_c - zfs_max_recordsize; + int64_t dn_over = aggsum_lower_bound(&arc_sums.arcstat_dnode_size) - + arc_dnode_limit; /* Always allow at least one block of overflow. */ - if (over < 0) + if (arc_over < 0 && dn_over <= 0) return (ARC_OVF_NONE); /* If we are under memory pressure, report severe overflow. */ @@ -5097,7 +5099,7 @@ arc_is_overflowing(boolean_t lax, boolean_t use_reserve) int64_t overflow = (arc_c >> zfs_arc_overflow_shift) / 2; if (use_reserve) overflow *= 3; - return (over < overflow ? ARC_OVF_SOME : ARC_OVF_SEVERE); + return (arc_over < overflow ? ARC_OVF_SOME : ARC_OVF_SEVERE); } static abd_t * @@ -5552,7 +5554,7 @@ static void arc_hdr_verify(arc_buf_hdr_t *hdr, blkptr_t *bp) { if (BP_IS_HOLE(bp) || BP_IS_EMBEDDED(bp)) { - ASSERT3U(HDR_GET_PSIZE(hdr), ==, 0); + ASSERT0(HDR_GET_PSIZE(hdr)); ASSERT3U(arc_hdr_get_compress(hdr), ==, ZIO_COMPRESS_OFF); } else { if (HDR_COMPRESSION_ENABLED(hdr)) { @@ -5585,7 +5587,7 @@ arc_read_done(zio_t *zio) if (HDR_IN_HASH_TABLE(hdr)) { arc_buf_hdr_t *found; - ASSERT3U(hdr->b_birth, ==, BP_GET_BIRTH(zio->io_bp)); + ASSERT3U(hdr->b_birth, ==, BP_GET_PHYSICAL_BIRTH(zio->io_bp)); ASSERT3U(hdr->b_dva.dva_word[0], ==, BP_IDENTITY(zio->io_bp)->dva_word[0]); ASSERT3U(hdr->b_dva.dva_word[1], ==, @@ -5688,7 +5690,7 @@ arc_read_done(zio_t *zio) error = SET_ERROR(EIO); if ((zio->io_flags & ZIO_FLAG_SPECULATIVE) == 0) { spa_log_error(zio->io_spa, &acb->acb_zb, - BP_GET_LOGICAL_BIRTH(zio->io_bp)); + BP_GET_PHYSICAL_BIRTH(zio->io_bp)); (void) zfs_ereport_post( FM_EREPORT_ZFS_AUTHENTICATION, zio->io_spa, NULL, &acb->acb_zb, zio, 0); @@ -6107,7 +6109,7 @@ top: if (!embedded_bp) { hdr->b_dva = *BP_IDENTITY(bp); - hdr->b_birth = BP_GET_BIRTH(bp); + hdr->b_birth = BP_GET_PHYSICAL_BIRTH(bp); exists = buf_hash_insert(hdr, &hash_lock); } if (exists != NULL) { @@ -6130,14 +6132,14 @@ top: } if (GHOST_STATE(hdr->b_l1hdr.b_state)) { - ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL); + ASSERT0P(hdr->b_l1hdr.b_pabd); ASSERT(!HDR_HAS_RABD(hdr)); ASSERT(!HDR_IO_IN_PROGRESS(hdr)); ASSERT0(zfs_refcount_count( &hdr->b_l1hdr.b_refcnt)); - ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL); + ASSERT0P(hdr->b_l1hdr.b_buf); #ifdef ZFS_DEBUG - ASSERT3P(hdr->b_l1hdr.b_freeze_cksum, ==, NULL); + ASSERT0P(hdr->b_l1hdr.b_freeze_cksum); #endif } else if (HDR_IO_IN_PROGRESS(hdr)) { /* @@ -6231,7 +6233,7 @@ top: acb->acb_nobuf = no_buf; acb->acb_zb = *zb; - ASSERT3P(hdr->b_l1hdr.b_acb, ==, NULL); + ASSERT0P(hdr->b_l1hdr.b_acb); hdr->b_l1hdr.b_acb = acb; if (HDR_HAS_L2HDR(hdr) && @@ -6715,7 +6717,7 @@ arc_release(arc_buf_t *buf, const void *tag) nhdr = arc_hdr_alloc(spa, psize, lsize, protected, compress, hdr->b_complevel, type); - ASSERT3P(nhdr->b_l1hdr.b_buf, ==, NULL); + ASSERT0P(nhdr->b_l1hdr.b_buf); ASSERT0(zfs_refcount_count(&nhdr->b_l1hdr.b_refcnt)); VERIFY3U(nhdr->b_type, ==, type); ASSERT(!HDR_SHARED_DATA(nhdr)); @@ -6802,7 +6804,7 @@ arc_write_ready(zio_t *zio) if (HDR_HAS_RABD(hdr)) arc_hdr_free_abd(hdr, B_TRUE); } - ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL); + ASSERT0P(hdr->b_l1hdr.b_pabd); ASSERT(!HDR_HAS_RABD(hdr)); ASSERT(!HDR_SHARED_DATA(hdr)); ASSERT(!arc_buf_is_shared(buf)); @@ -6946,7 +6948,7 @@ arc_write_done(zio_t *zio) arc_buf_t *buf = callback->awcb_buf; arc_buf_hdr_t *hdr = buf->b_hdr; - ASSERT3P(hdr->b_l1hdr.b_acb, ==, NULL); + ASSERT0P(hdr->b_l1hdr.b_acb); if (zio->io_error == 0) { arc_hdr_verify(hdr, zio->io_bp); @@ -6955,7 +6957,7 @@ arc_write_done(zio_t *zio) buf_discard_identity(hdr); } else { hdr->b_dva = *BP_IDENTITY(zio->io_bp); - hdr->b_birth = BP_GET_BIRTH(zio->io_bp); + hdr->b_birth = BP_GET_PHYSICAL_BIRTH(zio->io_bp); } } else { ASSERT(HDR_EMPTY(hdr)); @@ -6971,7 +6973,7 @@ arc_write_done(zio_t *zio) arc_buf_hdr_t *exists; kmutex_t *hash_lock; - ASSERT3U(zio->io_error, ==, 0); + ASSERT0(zio->io_error); arc_cksum_verify(buf); @@ -6992,7 +6994,7 @@ arc_write_done(zio_t *zio) arc_hdr_destroy(exists); mutex_exit(hash_lock); exists = buf_hash_insert(hdr, &hash_lock); - ASSERT3P(exists, ==, NULL); + ASSERT0P(exists); } else if (zio->io_flags & ZIO_FLAG_NOPWRITE) { /* nopwrite */ ASSERT(zio->io_prop.zp_nopwrite); @@ -7005,7 +7007,7 @@ arc_write_done(zio_t *zio) ASSERT(ARC_BUF_LAST(hdr->b_l1hdr.b_buf)); ASSERT(hdr->b_l1hdr.b_state == arc_anon); ASSERT(BP_GET_DEDUP(zio->io_bp)); - ASSERT(BP_GET_LEVEL(zio->io_bp) == 0); + ASSERT0(BP_GET_LEVEL(zio->io_bp)); } } arc_hdr_clear_flags(hdr, ARC_FLAG_IO_IN_PROGRESS); @@ -7042,7 +7044,7 @@ arc_write(zio_t *pio, spa_t *spa, uint64_t txg, ASSERT3P(done, !=, NULL); ASSERT(!HDR_IO_ERROR(hdr)); ASSERT(!HDR_IO_IN_PROGRESS(hdr)); - ASSERT3P(hdr->b_l1hdr.b_acb, ==, NULL); + ASSERT0P(hdr->b_l1hdr.b_acb); ASSERT3P(hdr->b_l1hdr.b_buf, !=, NULL); if (uncached) arc_hdr_set_flags(hdr, ARC_FLAG_UNCACHED); @@ -7111,7 +7113,7 @@ arc_write(zio_t *pio, spa_t *spa, uint64_t txg, arc_hdr_set_compress(hdr, ZIO_COMPRESS_OFF); ASSERT(!arc_buf_is_shared(buf)); - ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL); + ASSERT0P(hdr->b_l1hdr.b_pabd); zio = zio_write(pio, spa, txg, bp, abd_get_from_buf(buf->b_data, HDR_GET_LSIZE(hdr)), @@ -7326,7 +7328,7 @@ arc_kstat_update(kstat_t *ksp, int rw) #if defined(COMPAT_FREEBSD11) as->arcstat_other_size.value.ui64 = wmsum_value(&arc_sums.arcstat_bonus_size) + - wmsum_value(&arc_sums.arcstat_dnode_size) + + aggsum_value(&arc_sums.arcstat_dnode_size) + wmsum_value(&arc_sums.arcstat_dbuf_size); #endif @@ -7368,7 +7370,7 @@ arc_kstat_update(kstat_t *ksp, int rw) &as->arcstat_uncached_evictable_metadata); as->arcstat_dnode_size.value.ui64 = - wmsum_value(&arc_sums.arcstat_dnode_size); + aggsum_value(&arc_sums.arcstat_dnode_size); as->arcstat_bonus_size.value.ui64 = wmsum_value(&arc_sums.arcstat_bonus_size); as->arcstat_l2_hits.value.ui64 = @@ -7738,7 +7740,7 @@ arc_state_init(void) wmsum_init(&arc_sums.arcstat_data_size, 0); wmsum_init(&arc_sums.arcstat_metadata_size, 0); wmsum_init(&arc_sums.arcstat_dbuf_size, 0); - wmsum_init(&arc_sums.arcstat_dnode_size, 0); + aggsum_init(&arc_sums.arcstat_dnode_size, 0); wmsum_init(&arc_sums.arcstat_bonus_size, 0); wmsum_init(&arc_sums.arcstat_l2_hits, 0); wmsum_init(&arc_sums.arcstat_l2_misses, 0); @@ -7897,7 +7899,7 @@ arc_state_fini(void) wmsum_fini(&arc_sums.arcstat_data_size); wmsum_fini(&arc_sums.arcstat_metadata_size); wmsum_fini(&arc_sums.arcstat_dbuf_size); - wmsum_fini(&arc_sums.arcstat_dnode_size); + aggsum_fini(&arc_sums.arcstat_dnode_size); wmsum_fini(&arc_sums.arcstat_bonus_size); wmsum_fini(&arc_sums.arcstat_l2_hits); wmsum_fini(&arc_sums.arcstat_l2_misses); diff --git a/sys/contrib/openzfs/module/zfs/bpobj.c b/sys/contrib/openzfs/module/zfs/bpobj.c index 8c19de93f12f..ea9fbd036c6e 100644 --- a/sys/contrib/openzfs/module/zfs/bpobj.c +++ b/sys/contrib/openzfs/module/zfs/bpobj.c @@ -160,8 +160,8 @@ bpobj_open(bpobj_t *bpo, objset_t *os, uint64_t object) memset(bpo, 0, sizeof (*bpo)); mutex_init(&bpo->bpo_lock, NULL, MUTEX_DEFAULT, NULL); - ASSERT(bpo->bpo_dbuf == NULL); - ASSERT(bpo->bpo_phys == NULL); + ASSERT0P(bpo->bpo_dbuf); + ASSERT0P(bpo->bpo_phys); ASSERT(object != 0); ASSERT3U(doi.doi_type, ==, DMU_OT_BPOBJ); ASSERT3U(doi.doi_bonus_type, ==, DMU_OT_BPOBJ_HDR); @@ -478,7 +478,7 @@ bpobj_iterate_impl(bpobj_t *initial_bpo, bpobj_itor_t func, void *arg, * We have unprocessed subobjs. Process the next one. */ ASSERT(bpo->bpo_havecomp); - ASSERT3P(bpobj_size, ==, NULL); + ASSERT0P(bpobj_size); /* Add the last subobj to stack. */ int64_t i = bpi->bpi_unprocessed_subobjs - 1; @@ -954,8 +954,8 @@ space_range_cb(void *arg, const blkptr_t *bp, boolean_t bp_freed, dmu_tx_t *tx) (void) bp_freed, (void) tx; struct space_range_arg *sra = arg; - if (BP_GET_LOGICAL_BIRTH(bp) > sra->mintxg && - BP_GET_LOGICAL_BIRTH(bp) <= sra->maxtxg) { + if (BP_GET_BIRTH(bp) > sra->mintxg && + BP_GET_BIRTH(bp) <= sra->maxtxg) { if (dsl_pool_sync_context(spa_get_dsl(sra->spa))) sra->used += bp_get_dsize_sync(sra->spa, bp); else diff --git a/sys/contrib/openzfs/module/zfs/brt.c b/sys/contrib/openzfs/module/zfs/brt.c index 27d9ed7ea2b0..40664354aa73 100644 --- a/sys/contrib/openzfs/module/zfs/brt.c +++ b/sys/contrib/openzfs/module/zfs/brt.c @@ -478,6 +478,18 @@ brt_vdev_create(spa_t *spa, brt_vdev_t *brtvd, dmu_tx_t *tx) sizeof (uint64_t), 1, &brtvd->bv_mos_brtvdev, tx)); BRT_DEBUG("Pool directory object created, object=%s", name); + /* + * Activate the endian-fixed feature if this is the first BRT ZAP + * (i.e., BLOCK_CLONING is not yet active) and the feature is enabled. + */ + if (spa_feature_is_enabled(spa, SPA_FEATURE_BLOCK_CLONING_ENDIAN) && + !spa_feature_is_active(spa, SPA_FEATURE_BLOCK_CLONING)) { + spa_feature_incr(spa, SPA_FEATURE_BLOCK_CLONING_ENDIAN, tx); + } else if (spa_feature_is_active(spa, + SPA_FEATURE_BLOCK_CLONING_ENDIAN)) { + spa_feature_incr(spa, SPA_FEATURE_BLOCK_CLONING_ENDIAN, tx); + } + spa_feature_incr(spa, SPA_FEATURE_BLOCK_CLONING, tx); } @@ -658,6 +670,8 @@ brt_vdev_destroy(spa_t *spa, brt_vdev_t *brtvd, dmu_tx_t *tx) rw_exit(&brtvd->bv_lock); spa_feature_decr(spa, SPA_FEATURE_BLOCK_CLONING, tx); + if (spa_feature_is_active(spa, SPA_FEATURE_BLOCK_CLONING_ENDIAN)) + spa_feature_decr(spa, SPA_FEATURE_BLOCK_CLONING_ENDIAN, tx); } static void @@ -855,16 +869,29 @@ brt_entry_fill(const blkptr_t *bp, brt_entry_t *bre, uint64_t *vdevidp) *vdevidp = DVA_GET_VDEV(&bp->blk_dva[0]); } +static boolean_t +brt_has_endian_fixed(spa_t *spa) +{ + return (spa_feature_is_active(spa, SPA_FEATURE_BLOCK_CLONING_ENDIAN)); +} + static int -brt_entry_lookup(brt_vdev_t *brtvd, brt_entry_t *bre) +brt_entry_lookup(spa_t *spa, brt_vdev_t *brtvd, brt_entry_t *bre) { uint64_t off = BRE_OFFSET(bre); if (brtvd->bv_mos_entries == 0) return (SET_ERROR(ENOENT)); - return (zap_lookup_uint64_by_dnode(brtvd->bv_mos_entries_dnode, - &off, BRT_KEY_WORDS, 1, sizeof (bre->bre_count), &bre->bre_count)); + if (brt_has_endian_fixed(spa)) { + return (zap_lookup_uint64_by_dnode(brtvd->bv_mos_entries_dnode, + &off, BRT_KEY_WORDS, sizeof (bre->bre_count), 1, + &bre->bre_count)); + } else { + return (zap_lookup_uint64_by_dnode(brtvd->bv_mos_entries_dnode, + &off, BRT_KEY_WORDS, 1, sizeof (bre->bre_count), + &bre->bre_count)); + } } /* @@ -1056,7 +1083,7 @@ brt_entry_decref(spa_t *spa, const blkptr_t *bp) } rw_exit(&brtvd->bv_lock); - error = brt_entry_lookup(brtvd, &bre_search); + error = brt_entry_lookup(spa, brtvd, &bre_search); /* bre_search now contains correct bre_count */ if (error == ENOENT) { BRTSTAT_BUMP(brt_decref_no_entry); @@ -1118,7 +1145,7 @@ brt_entry_get_refcount(spa_t *spa, const blkptr_t *bp) bre = avl_find(&brtvd->bv_tree, &bre_search, NULL); if (bre == NULL) { rw_exit(&brtvd->bv_lock); - error = brt_entry_lookup(brtvd, &bre_search); + error = brt_entry_lookup(spa, brtvd, &bre_search); if (error == ENOENT) { refcnt = 0; } else { @@ -1270,10 +1297,18 @@ brt_pending_apply_vdev(spa_t *spa, brt_vdev_t *brtvd, uint64_t txg) uint64_t off = BRE_OFFSET(bre); if (brtvd->bv_mos_entries != 0 && brt_vdev_lookup(spa, brtvd, off)) { - int error = zap_lookup_uint64_by_dnode( - brtvd->bv_mos_entries_dnode, &off, - BRT_KEY_WORDS, 1, sizeof (bre->bre_count), - &bre->bre_count); + int error; + if (brt_has_endian_fixed(spa)) { + error = zap_lookup_uint64_by_dnode( + brtvd->bv_mos_entries_dnode, &off, + BRT_KEY_WORDS, sizeof (bre->bre_count), 1, + &bre->bre_count); + } else { + error = zap_lookup_uint64_by_dnode( + brtvd->bv_mos_entries_dnode, &off, + BRT_KEY_WORDS, 1, sizeof (bre->bre_count), + &bre->bre_count); + } if (error == 0) { BRTSTAT_BUMP(brt_addref_entry_on_disk); } else { @@ -1326,7 +1361,7 @@ brt_pending_apply(spa_t *spa, uint64_t txg) } static void -brt_sync_entry(dnode_t *dn, brt_entry_t *bre, dmu_tx_t *tx) +brt_sync_entry(spa_t *spa, dnode_t *dn, brt_entry_t *bre, dmu_tx_t *tx) { uint64_t off = BRE_OFFSET(bre); @@ -1337,9 +1372,15 @@ brt_sync_entry(dnode_t *dn, brt_entry_t *bre, dmu_tx_t *tx) BRT_KEY_WORDS, tx); VERIFY(error == 0 || error == ENOENT); } else { - VERIFY0(zap_update_uint64_by_dnode(dn, &off, - BRT_KEY_WORDS, 1, sizeof (bre->bre_count), - &bre->bre_count, tx)); + if (brt_has_endian_fixed(spa)) { + VERIFY0(zap_update_uint64_by_dnode(dn, &off, + BRT_KEY_WORDS, sizeof (bre->bre_count), 1, + &bre->bre_count, tx)); + } else { + VERIFY0(zap_update_uint64_by_dnode(dn, &off, + BRT_KEY_WORDS, 1, sizeof (bre->bre_count), + &bre->bre_count, tx)); + } } } @@ -1368,7 +1409,8 @@ brt_sync_table(spa_t *spa, dmu_tx_t *tx) void *c = NULL; while ((bre = avl_destroy_nodes(&brtvd->bv_tree, &c)) != NULL) { - brt_sync_entry(brtvd->bv_mos_entries_dnode, bre, tx); + brt_sync_entry(spa, brtvd->bv_mos_entries_dnode, bre, + tx); kmem_cache_free(brt_entry_cache, bre); } diff --git a/sys/contrib/openzfs/module/zfs/btree.c b/sys/contrib/openzfs/module/zfs/btree.c index aa282f711bc3..725b96a3b2c7 100644 --- a/sys/contrib/openzfs/module/zfs/btree.c +++ b/sys/contrib/openzfs/module/zfs/btree.c @@ -1110,7 +1110,7 @@ zfs_btree_add_idx(zfs_btree_t *tree, const void *value, if (where->bti_node == NULL) { ASSERT3U(tree->bt_num_elems, ==, 1); ASSERT3S(tree->bt_height, ==, -1); - ASSERT3P(tree->bt_root, ==, NULL); + ASSERT0P(tree->bt_root); ASSERT0(where->bti_offset); tree->bt_num_nodes++; @@ -1947,7 +1947,7 @@ void zfs_btree_destroy(zfs_btree_t *tree) { ASSERT0(tree->bt_num_elems); - ASSERT3P(tree->bt_root, ==, NULL); + ASSERT0P(tree->bt_root); } /* Verify that every child of this node has the correct parent pointer. */ @@ -1969,10 +1969,10 @@ static void zfs_btree_verify_pointers(zfs_btree_t *tree) { if (tree->bt_height == -1) { - VERIFY3P(tree->bt_root, ==, NULL); + VERIFY0P(tree->bt_root); return; } - VERIFY3P(tree->bt_root->bth_parent, ==, NULL); + VERIFY0P(tree->bt_root->bth_parent); zfs_btree_verify_pointers_helper(tree, tree->bt_root); } diff --git a/sys/contrib/openzfs/module/zfs/dataset_kstats.c b/sys/contrib/openzfs/module/zfs/dataset_kstats.c index d3baabd6169f..e5abcd2044cf 100644 --- a/sys/contrib/openzfs/module/zfs/dataset_kstats.c +++ b/sys/contrib/openzfs/module/zfs/dataset_kstats.c @@ -44,6 +44,7 @@ static dataset_kstat_values_t empty_dataset_kstats = { { "zil_commit_error_count", KSTAT_DATA_UINT64 }, { "zil_commit_stall_count", KSTAT_DATA_UINT64 }, { "zil_commit_suspend_count", KSTAT_DATA_UINT64 }, + { "zil_commit_crash_count", KSTAT_DATA_UINT64 }, { "zil_itx_count", KSTAT_DATA_UINT64 }, { "zil_itx_indirect_count", KSTAT_DATA_UINT64 }, { "zil_itx_indirect_bytes", KSTAT_DATA_UINT64 }, diff --git a/sys/contrib/openzfs/module/zfs/dbuf.c b/sys/contrib/openzfs/module/zfs/dbuf.c index f1b5a17f337e..3d0f88b36336 100644 --- a/sys/contrib/openzfs/module/zfs/dbuf.c +++ b/sys/contrib/openzfs/module/zfs/dbuf.c @@ -523,7 +523,7 @@ dbuf_verify_user(dmu_buf_impl_t *db, dbvu_verify_type_t verify_type) return; /* Only data blocks support the attachment of user data. */ - ASSERT(db->db_level == 0); + ASSERT0(db->db_level); /* Clients must resolve a dbuf before attaching user data. */ ASSERT(db->db.db_data != NULL); @@ -866,8 +866,16 @@ dbuf_evict_notify(uint64_t size) * and grabbing the lock results in massive lock contention. */ if (size > dbuf_cache_target_bytes()) { - if (size > dbuf_cache_hiwater_bytes()) + /* + * Avoid calling dbuf_evict_one() from memory reclaim context + * (e.g. Linux kswapd, FreeBSD pagedaemon) to prevent deadlocks. + * Memory reclaim threads can get stuck waiting for the dbuf + * hash lock. + */ + if (size > dbuf_cache_hiwater_bytes() && + !current_is_reclaim_thread()) { dbuf_evict_one(); + } cv_signal(&dbuf_evict_cv); } } @@ -1120,8 +1128,8 @@ dbuf_verify(dmu_buf_impl_t *db) DB_DNODE_ENTER(db); dn = DB_DNODE(db); if (dn == NULL) { - ASSERT(db->db_parent == NULL); - ASSERT(db->db_blkptr == NULL); + ASSERT0P(db->db_parent); + ASSERT0P(db->db_blkptr); } else { ASSERT3U(db->db.db_object, ==, dn->dn_object); ASSERT3P(db->db_objset, ==, dn->dn_objset); @@ -1172,7 +1180,7 @@ dbuf_verify(dmu_buf_impl_t *db) /* db is pointed to by the dnode */ /* ASSERT3U(db->db_blkid, <, dn->dn_nblkptr); */ if (DMU_OBJECT_IS_SPECIAL(db->db.db_object)) - ASSERT(db->db_parent == NULL); + ASSERT0P(db->db_parent); else ASSERT(db->db_parent != NULL); if (db->db_blkid != DMU_SPILL_BLKID) @@ -1211,7 +1219,7 @@ dbuf_verify(dmu_buf_impl_t *db) int i; for (i = 0; i < db->db.db_size >> 3; i++) { - ASSERT(buf[i] == 0); + ASSERT0(buf[i]); } } else { blkptr_t *bps = db->db.db_data; @@ -1235,11 +1243,9 @@ dbuf_verify(dmu_buf_impl_t *db) DVA_IS_EMPTY(&bp->blk_dva[1]) && DVA_IS_EMPTY(&bp->blk_dva[2])); ASSERT0(bp->blk_fill); - ASSERT0(bp->blk_pad[0]); - ASSERT0(bp->blk_pad[1]); ASSERT(!BP_IS_EMBEDDED(bp)); ASSERT(BP_IS_HOLE(bp)); - ASSERT0(BP_GET_PHYSICAL_BIRTH(bp)); + ASSERT0(BP_GET_RAW_PHYSICAL_BIRTH(bp)); } } } @@ -1253,7 +1259,7 @@ dbuf_clear_data(dmu_buf_impl_t *db) { ASSERT(MUTEX_HELD(&db->db_mtx)); dbuf_evict_user(db); - ASSERT3P(db->db_buf, ==, NULL); + ASSERT0P(db->db_buf); db->db.db_data = NULL; if (db->db_state != DB_NOFILL) { db->db_state = DB_UNCACHED; @@ -1378,13 +1384,13 @@ dbuf_read_done(zio_t *zio, const zbookmark_phys_t *zb, const blkptr_t *bp, * All reads are synchronous, so we must have a hold on the dbuf */ ASSERT(zfs_refcount_count(&db->db_holds) > 0); - ASSERT(db->db_buf == NULL); - ASSERT(db->db.db_data == NULL); + ASSERT0P(db->db_buf); + ASSERT0P(db->db.db_data); if (buf == NULL) { /* i/o error */ ASSERT(zio == NULL || zio->io_error != 0); ASSERT(db->db_blkid != DMU_BONUS_BLKID); - ASSERT3P(db->db_buf, ==, NULL); + ASSERT0P(db->db_buf); db->db_state = DB_UNCACHED; DTRACE_SET_STATE(db, "i/o error"); } else if (db->db_level == 0 && db->db_freed_in_flight) { @@ -1578,7 +1584,7 @@ dbuf_read_impl(dmu_buf_impl_t *db, dnode_t *dn, zio_t *zio, dmu_flags_t flags, ASSERT(!zfs_refcount_is_zero(&db->db_holds)); ASSERT(MUTEX_HELD(&db->db_mtx)); ASSERT(db->db_state == DB_UNCACHED || db->db_state == DB_NOFILL); - ASSERT(db->db_buf == NULL); + ASSERT0P(db->db_buf); ASSERT(db->db_parent == NULL || RW_LOCK_HELD(&db->db_parent->db_rwlock)); @@ -1615,7 +1621,7 @@ dbuf_read_impl(dmu_buf_impl_t *db, dnode_t *dn, zio_t *zio, dmu_flags_t flags, */ if (db->db_objset->os_encrypted && !BP_USES_CRYPT(bp)) { spa_log_error(db->db_objset->os_spa, &zb, - BP_GET_LOGICAL_BIRTH(bp)); + BP_GET_PHYSICAL_BIRTH(bp)); err = SET_ERROR(EIO); goto early_unlock; } @@ -1676,7 +1682,7 @@ dbuf_fix_old_data(dmu_buf_impl_t *db, uint64_t txg) ASSERT(MUTEX_HELD(&db->db_mtx)); ASSERT(db->db.db_data != NULL); - ASSERT(db->db_level == 0); + ASSERT0(db->db_level); ASSERT(db->db.db_object != DMU_META_DNODE_OBJECT); if (dr == NULL || @@ -1895,8 +1901,8 @@ dbuf_noread(dmu_buf_impl_t *db, dmu_flags_t flags) while (db->db_state == DB_READ || db->db_state == DB_FILL) cv_wait(&db->db_changed, &db->db_mtx); if (db->db_state == DB_UNCACHED) { - ASSERT(db->db_buf == NULL); - ASSERT(db->db.db_data == NULL); + ASSERT0P(db->db_buf); + ASSERT0P(db->db.db_data); dbuf_set_data(db, dbuf_alloc_arcbuf(db)); db->db_state = DB_FILL; DTRACE_SET_STATE(db, "assigning filled buffer"); @@ -1923,7 +1929,7 @@ dbuf_unoverride(dbuf_dirty_record_t *dr) * comes from dbuf_dirty() callers who must also hold a range lock. */ ASSERT(dr->dt.dl.dr_override_state != DR_IN_DMU_SYNC); - ASSERT(db->db_level == 0); + ASSERT0(db->db_level); if (db->db_blkid == DMU_BONUS_BLKID || dr->dt.dl.dr_override_state == DR_NOT_OVERRIDDEN) @@ -1988,7 +1994,7 @@ dbuf_free_range(dnode_t *dn, uint64_t start_blkid, uint64_t end_blkid, mutex_enter(&dn->dn_dbufs_mtx); db = avl_find(&dn->dn_dbufs, db_search, &where); - ASSERT3P(db, ==, NULL); + ASSERT0P(db); db = avl_nearest(&dn->dn_dbufs, where, AVL_AFTER); @@ -2011,7 +2017,7 @@ dbuf_free_range(dnode_t *dn, uint64_t start_blkid, uint64_t end_blkid, if (db->db_state == DB_UNCACHED || db->db_state == DB_NOFILL || db->db_state == DB_EVICTING) { - ASSERT(db->db.db_data == NULL); + ASSERT0P(db->db.db_data); mutex_exit(&db->db_mtx); continue; } @@ -2154,6 +2160,12 @@ dbuf_redirty(dbuf_dirty_record_t *dr) ASSERT(arc_released(db->db_buf)); arc_buf_thaw(db->db_buf); } + + /* + * Clear the rewrite flag since this is now a logical + * modification. + */ + dr->dt.dl.dr_rewrite = B_FALSE; } } @@ -2701,6 +2713,38 @@ dmu_buf_will_dirty(dmu_buf_t *db_fake, dmu_tx_t *tx) dmu_buf_will_dirty_flags(db_fake, tx, DMU_READ_NO_PREFETCH); } +void +dmu_buf_will_rewrite(dmu_buf_t *db_fake, dmu_tx_t *tx) +{ + dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake; + + ASSERT(tx->tx_txg != 0); + ASSERT(!zfs_refcount_is_zero(&db->db_holds)); + + /* + * If the dbuf is already dirty in this txg, it will be written + * anyway, so there's nothing to do. + */ + mutex_enter(&db->db_mtx); + if (dbuf_find_dirty_eq(db, tx->tx_txg) != NULL) { + mutex_exit(&db->db_mtx); + return; + } + mutex_exit(&db->db_mtx); + + /* + * The dbuf is not dirty, so we need to make it dirty and + * mark it for rewrite (preserve logical birth time). + */ + dmu_buf_will_dirty_flags(db_fake, tx, DMU_READ_NO_PREFETCH); + + mutex_enter(&db->db_mtx); + dbuf_dirty_record_t *dr = dbuf_find_dirty_eq(db, tx->tx_txg); + if (dr != NULL && db->db_level == 0) + dr->dt.dl.dr_rewrite = B_TRUE; + mutex_exit(&db->db_mtx); +} + boolean_t dmu_buf_is_dirty(dmu_buf_t *db_fake, dmu_tx_t *tx) { @@ -2852,8 +2896,8 @@ dmu_buf_will_clone_or_dio(dmu_buf_t *db_fake, dmu_tx_t *tx) dbuf_clear_data(db); } - ASSERT3P(db->db_buf, ==, NULL); - ASSERT3P(db->db.db_data, ==, NULL); + ASSERT0P(db->db_buf); + ASSERT0P(db->db.db_data); db->db_state = DB_NOFILL; DTRACE_SET_STATE(db, @@ -2888,7 +2932,7 @@ dmu_buf_will_fill_flags(dmu_buf_t *db_fake, dmu_tx_t *tx, boolean_t canfail, ASSERT(db->db_blkid != DMU_BONUS_BLKID); ASSERT(tx->tx_txg != 0); - ASSERT(db->db_level == 0); + ASSERT0(db->db_level); ASSERT(!zfs_refcount_is_zero(&db->db_holds)); ASSERT(db->db.db_object != DMU_META_DNODE_OBJECT || @@ -3100,7 +3144,7 @@ dbuf_assign_arcbuf(dmu_buf_impl_t *db, arc_buf_t *buf, dmu_tx_t *tx, { ASSERT(!zfs_refcount_is_zero(&db->db_holds)); ASSERT(db->db_blkid != DMU_BONUS_BLKID); - ASSERT(db->db_level == 0); + ASSERT0(db->db_level); ASSERT3U(dbuf_is_metadata(db), ==, arc_is_metadata(buf)); ASSERT(buf != NULL); ASSERT3U(arc_buf_lsize(buf), ==, db->db.db_size); @@ -3165,7 +3209,7 @@ dbuf_assign_arcbuf(dmu_buf_impl_t *db, arc_buf_t *buf, dmu_tx_t *tx, VERIFY(!dbuf_undirty(db, tx)); db->db_state = DB_UNCACHED; } - ASSERT(db->db_buf == NULL); + ASSERT0P(db->db_buf); dbuf_set_data(db, buf); db->db_state = DB_FILL; DTRACE_SET_STATE(db, "filling assigned arcbuf"); @@ -3225,7 +3269,7 @@ dbuf_destroy(dmu_buf_impl_t *db) } ASSERT(db->db_state == DB_UNCACHED || db->db_state == DB_NOFILL); - ASSERT(db->db_data_pending == NULL); + ASSERT0P(db->db_data_pending); ASSERT(list_is_empty(&db->db_dirty_records)); db->db_state = DB_EVICTING; @@ -3277,11 +3321,11 @@ dbuf_destroy(dmu_buf_impl_t *db) db->db_parent = NULL; - ASSERT(db->db_buf == NULL); - ASSERT(db->db.db_data == NULL); - ASSERT(db->db_hash_next == NULL); - ASSERT(db->db_blkptr == NULL); - ASSERT(db->db_data_pending == NULL); + ASSERT0P(db->db_buf); + ASSERT0P(db->db.db_data); + ASSERT0P(db->db_hash_next); + ASSERT0P(db->db_blkptr); + ASSERT0P(db->db_data_pending); ASSERT3U(db->db_caching_status, ==, DB_NO_CACHE); ASSERT(!multilist_link_active(&db->db_cache_link)); @@ -3916,7 +3960,7 @@ dbuf_hold_impl(dnode_t *dn, uint8_t level, uint64_t blkid, if (fail_uncached) return (SET_ERROR(ENOENT)); - ASSERT3P(parent, ==, NULL); + ASSERT0P(parent); err = dbuf_findbp(dn, level, blkid, fail_sparse, &parent, &bp); if (fail_sparse) { if (err == 0 && bp && BP_IS_HOLE(bp)) @@ -4020,7 +4064,7 @@ dbuf_create_bonus(dnode_t *dn) { ASSERT(RW_WRITE_HELD(&dn->dn_struct_rwlock)); - ASSERT(dn->dn_bonus == NULL); + ASSERT0P(dn->dn_bonus); dn->dn_bonus = dbuf_create(dn, 0, DMU_BONUS_BLKID, dn->dn_dbuf, NULL, dbuf_hash(dn->dn_objset, dn->dn_object, 0, DMU_BONUS_BLKID)); dn->dn_bonus->db_pending_evict = FALSE; @@ -4372,7 +4416,7 @@ dbuf_check_blkptr(dnode_t *dn, dmu_buf_impl_t *db) * inappropriate to hook it in (i.e., nlevels mismatch). */ ASSERT(db->db_blkid < dn->dn_phys->dn_nblkptr); - ASSERT(db->db_parent == NULL); + ASSERT0P(db->db_parent); db->db_parent = dn->dn_dbuf; db->db_blkptr = &dn->dn_phys->dn_blkptr[db->db_blkid]; DBUF_VERIFY(db); @@ -4433,7 +4477,7 @@ dbuf_prepare_encrypted_dnode_leaf(dbuf_dirty_record_t *dr) ASSERT(MUTEX_HELD(&db->db_mtx)); ASSERT3U(db->db.db_object, ==, DMU_META_DNODE_OBJECT); - ASSERT3U(db->db_level, ==, 0); + ASSERT0(db->db_level); if (!db->db_objset->os_raw_receive && arc_is_encrypted(db->db_buf)) { zbookmark_phys_t zb; @@ -4544,7 +4588,7 @@ dbuf_sync_leaf_verify_bonus_dnode(dbuf_dirty_record_t *dr) /* ensure that everything is zero after our data */ for (; datap_end < datap_max; datap_end++) - ASSERT(*datap_end == 0); + ASSERT0(*datap_end); #endif } @@ -4552,7 +4596,7 @@ static blkptr_t * dbuf_lightweight_bp(dbuf_dirty_record_t *dr) { /* This must be a lightweight dirty record. */ - ASSERT3P(dr->dr_dbuf, ==, NULL); + ASSERT0P(dr->dr_dbuf); dnode_t *dn = dr->dr_dnode; if (dn->dn_phys->dn_nlevels == 1) { @@ -4695,7 +4739,7 @@ dbuf_sync_leaf(dbuf_dirty_record_t *dr, dmu_tx_t *tx) */ if (db->db_state == DB_UNCACHED) { /* This buffer has been freed since it was dirtied */ - ASSERT3P(db->db.db_data, ==, NULL); + ASSERT0P(db->db.db_data); } else if (db->db_state == DB_FILL) { /* This buffer was freed and is now being re-filled */ ASSERT(db->db.db_data != dr->dt.dl.dr_data); @@ -4712,9 +4756,9 @@ dbuf_sync_leaf(dbuf_dirty_record_t *dr, dmu_tx_t *tx) */ dbuf_dirty_record_t *dr_head = list_head(&db->db_dirty_records); - ASSERT3P(db->db_buf, ==, NULL); - ASSERT3P(db->db.db_data, ==, NULL); - ASSERT3P(dr_head->dt.dl.dr_data, ==, NULL); + ASSERT0P(db->db_buf); + ASSERT0P(db->db.db_data); + ASSERT0P(dr_head->dt.dl.dr_data); ASSERT3U(dr_head->dt.dl.dr_override_state, ==, DR_OVERRIDDEN); } else { ASSERT(db->db_state == DB_CACHED || db->db_state == DB_NOFILL); @@ -4899,7 +4943,7 @@ dbuf_write_ready(zio_t *zio, arc_buf_t *buf, void *vdb) dnode_diduse_space(dn, delta - zio->io_prev_space_delta); zio->io_prev_space_delta = delta; - if (BP_GET_LOGICAL_BIRTH(bp) != 0) { + if (BP_GET_BIRTH(bp) != 0) { ASSERT((db->db_blkid != DMU_SPILL_BLKID && BP_GET_TYPE(bp) == dn->dn_type) || (db->db_blkid == DMU_SPILL_BLKID && @@ -5186,7 +5230,7 @@ dbuf_remap_impl(dnode_t *dn, blkptr_t *bp, krwlock_t *rw, dmu_tx_t *tx) ASSERT(dsl_pool_sync_context(spa_get_dsl(spa))); drica.drica_os = dn->dn_objset; - drica.drica_blk_birth = BP_GET_LOGICAL_BIRTH(bp); + drica.drica_blk_birth = BP_GET_BIRTH(bp); drica.drica_tx = tx; if (spa_remap_blkptr(spa, &bp_copy, dbuf_remap_impl_callback, &drica)) { @@ -5201,8 +5245,7 @@ dbuf_remap_impl(dnode_t *dn, blkptr_t *bp, krwlock_t *rw, dmu_tx_t *tx) if (dn->dn_objset != spa_meta_objset(spa)) { dsl_dataset_t *ds = dmu_objset_ds(dn->dn_objset); if (dsl_deadlist_is_open(&ds->ds_dir->dd_livelist) && - BP_GET_LOGICAL_BIRTH(bp) > - ds->ds_dir->dd_origin_txg) { + BP_GET_BIRTH(bp) > ds->ds_dir->dd_origin_txg) { ASSERT(!BP_IS_EMBEDDED(bp)); ASSERT(dsl_dir_is_clone(ds->ds_dir)); ASSERT(spa_feature_is_enabled(spa, @@ -5320,7 +5363,7 @@ dbuf_write(dbuf_dirty_record_t *dr, arc_buf_t *data, dmu_tx_t *tx) } ASSERT(db->db_level == 0 || data == db->db_buf); - ASSERT3U(BP_GET_LOGICAL_BIRTH(db->db_blkptr), <=, txg); + ASSERT3U(BP_GET_BIRTH(db->db_blkptr), <=, txg); ASSERT(pio); SET_BOOKMARK(&zb, os->os_dsl_dataset ? @@ -5334,6 +5377,24 @@ dbuf_write(dbuf_dirty_record_t *dr, arc_buf_t *data, dmu_tx_t *tx) dmu_write_policy(os, dn, db->db_level, wp_flag, &zp); /* + * Set rewrite properties for zfs_rewrite() operations. + */ + if (db->db_level == 0 && dr->dt.dl.dr_rewrite) { + zp.zp_rewrite = B_TRUE; + + /* + * Mark physical rewrite feature for activation. + * This will be activated automatically during dataset sync. + */ + dsl_dataset_t *ds = os->os_dsl_dataset; + if (!dsl_dataset_feature_is_active(ds, + SPA_FEATURE_PHYSICAL_REWRITE)) { + ds->ds_feature_activation[ + SPA_FEATURE_PHYSICAL_REWRITE] = (void *)B_TRUE; + } + } + + /* * We copy the blkptr now (rather than when we instantiate the dirty * record), because its value can change between open context and * syncing context. We do not need to hold dn_struct_rwlock to read @@ -5403,6 +5464,7 @@ EXPORT_SYMBOL(dbuf_release_bp); EXPORT_SYMBOL(dbuf_dirty); EXPORT_SYMBOL(dmu_buf_set_crypt_params); EXPORT_SYMBOL(dmu_buf_will_dirty); +EXPORT_SYMBOL(dmu_buf_will_rewrite); EXPORT_SYMBOL(dmu_buf_is_dirty); EXPORT_SYMBOL(dmu_buf_will_clone_or_dio); EXPORT_SYMBOL(dmu_buf_will_not_fill); diff --git a/sys/contrib/openzfs/module/zfs/ddt.c b/sys/contrib/openzfs/module/zfs/ddt.c index 60cbb7755a7e..d6658375f810 100644 --- a/sys/contrib/openzfs/module/zfs/ddt.c +++ b/sys/contrib/openzfs/module/zfs/ddt.c @@ -397,7 +397,7 @@ ddt_object_create(ddt_t *ddt, ddt_type_t type, ddt_class_t class, ddt_object_name(ddt, type, class, name); - ASSERT3U(*objectp, ==, 0); + ASSERT0(*objectp); VERIFY0(ddt_ops[type]->ddt_op_create(os, objectp, tx, prehash)); ASSERT3U(*objectp, !=, 0); @@ -724,10 +724,13 @@ ddt_phys_extend(ddt_univ_phys_t *ddp, ddt_phys_variant_t v, const blkptr_t *bp) dvas[2] = bp->blk_dva[2]; if (ddt_phys_birth(ddp, v) == 0) { - if (v == DDT_PHYS_FLAT) - ddp->ddp_flat.ddp_phys_birth = BP_GET_BIRTH(bp); - else - ddp->ddp_trad[v].ddp_phys_birth = BP_GET_BIRTH(bp); + if (v == DDT_PHYS_FLAT) { + ddp->ddp_flat.ddp_phys_birth = + BP_GET_PHYSICAL_BIRTH(bp); + } else { + ddp->ddp_trad[v].ddp_phys_birth = + BP_GET_PHYSICAL_BIRTH(bp); + } } } @@ -891,14 +894,14 @@ ddt_phys_select(const ddt_t *ddt, const ddt_entry_t *dde, const blkptr_t *bp) if (ddt->ddt_flags & DDT_FLAG_FLAT) { if (DVA_EQUAL(BP_IDENTITY(bp), &ddp->ddp_flat.ddp_dva[0]) && - BP_GET_BIRTH(bp) == ddp->ddp_flat.ddp_phys_birth) { + BP_GET_PHYSICAL_BIRTH(bp) == ddp->ddp_flat.ddp_phys_birth) { return (DDT_PHYS_FLAT); } } else /* traditional phys */ { for (int p = 0; p < DDT_PHYS_MAX; p++) { if (DVA_EQUAL(BP_IDENTITY(bp), &ddp->ddp_trad[p].ddp_dva[0]) && - BP_GET_BIRTH(bp) == + BP_GET_PHYSICAL_BIRTH(bp) == ddp->ddp_trad[p].ddp_phys_birth) { return (p); } @@ -1008,7 +1011,7 @@ ddt_free(const ddt_t *ddt, ddt_entry_t *dde) { if (dde->dde_io != NULL) { for (int p = 0; p < DDT_NPHYS(ddt); p++) - ASSERT3P(dde->dde_io->dde_lead_zio[p], ==, NULL); + ASSERT0P(dde->dde_io->dde_lead_zio[p]); if (dde->dde_io->dde_repair_abd != NULL) abd_free(dde->dde_io->dde_repair_abd); @@ -1418,7 +1421,7 @@ ddt_key_compare(const void *x1, const void *x2) static void ddt_create_dir(ddt_t *ddt, dmu_tx_t *tx) { - ASSERT3U(ddt->ddt_dir_object, ==, 0); + ASSERT0(ddt->ddt_dir_object); ASSERT3U(ddt->ddt_version, ==, DDT_VERSION_FDT); char name[DDT_NAMELEN]; @@ -2392,7 +2395,7 @@ ddt_sync(spa_t *spa, uint64_t txg) * scan's root zio here so that we can wait for any scan IOs in * addition to the regular ddt IOs. */ - ASSERT3P(scn->scn_zio_root, ==, NULL); + ASSERT0P(scn->scn_zio_root); scn->scn_zio_root = rio; for (enum zio_checksum c = 0; c < ZIO_CHECKSUM_FUNCTIONS; c++) { diff --git a/sys/contrib/openzfs/module/zfs/ddt_log.c b/sys/contrib/openzfs/module/zfs/ddt_log.c index dbd381aa9609..3d30e244c1f7 100644 --- a/sys/contrib/openzfs/module/zfs/ddt_log.c +++ b/sys/contrib/openzfs/module/zfs/ddt_log.c @@ -116,7 +116,7 @@ static void ddt_log_create_one(ddt_t *ddt, ddt_log_t *ddl, uint_t n, dmu_tx_t *tx) { ASSERT3U(ddt->ddt_dir_object, >, 0); - ASSERT3U(ddl->ddl_object, ==, 0); + ASSERT0(ddl->ddl_object); char name[DDT_NAMELEN]; ddt_log_name(ddt, name, n); @@ -194,7 +194,7 @@ void ddt_log_begin(ddt_t *ddt, size_t nentries, dmu_tx_t *tx, ddt_log_update_t *dlu) { ASSERT3U(nentries, >, 0); - ASSERT3P(dlu->dlu_dbp, ==, NULL); + ASSERT0P(dlu->dlu_dbp); if (ddt->ddt_log_active->ddl_object == 0) ddt_log_create(ddt, tx); @@ -748,8 +748,8 @@ ddt_log_load(ddt_t *ddt) void ddt_log_alloc(ddt_t *ddt) { - ASSERT3P(ddt->ddt_log_active, ==, NULL); - ASSERT3P(ddt->ddt_log_flushing, ==, NULL); + ASSERT0P(ddt->ddt_log_active); + ASSERT0P(ddt->ddt_log_flushing); avl_create(&ddt->ddt_log[0].ddl_tree, ddt_key_compare, sizeof (ddt_log_entry_t), offsetof(ddt_log_entry_t, ddle_node)); diff --git a/sys/contrib/openzfs/module/zfs/dmu.c b/sys/contrib/openzfs/module/zfs/dmu.c index 21c465328134..f7f808d5b8f7 100644 --- a/sys/contrib/openzfs/module/zfs/dmu.c +++ b/sys/contrib/openzfs/module/zfs/dmu.c @@ -1343,7 +1343,7 @@ dmu_prealloc(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, if (size == 0) return; - VERIFY(0 == dmu_buf_hold_array(os, object, offset, size, + VERIFY0(dmu_buf_hold_array(os, object, offset, size, FALSE, FTAG, &numbufs, &dbp)); for (i = 0; i < numbufs; i++) { @@ -1872,7 +1872,7 @@ dmu_sync_ready(zio_t *zio, arc_buf_t *buf, void *varg) */ BP_SET_LSIZE(bp, db->db_size); } else if (!BP_IS_EMBEDDED(bp)) { - ASSERT(BP_GET_LEVEL(bp) == 0); + ASSERT0(BP_GET_LEVEL(bp)); BP_SET_FILL(bp, 1); } } @@ -1966,7 +1966,7 @@ dmu_sync_late_arrival_done(zio_t *zio) blkptr_t *bp_orig __maybe_unused = &zio->io_bp_orig; ASSERT(!(zio->io_flags & ZIO_FLAG_NOPWRITE)); ASSERT(BP_IS_HOLE(bp_orig) || !BP_EQUAL(bp, bp_orig)); - ASSERT(BP_GET_LOGICAL_BIRTH(zio->io_bp) == zio->io_txg); + ASSERT(BP_GET_BIRTH(zio->io_bp) == zio->io_txg); ASSERT(zio->io_txg > spa_syncing_txg(zio->io_spa)); zio_free(zio->io_spa, zio->io_txg, zio->io_bp); } @@ -2405,7 +2405,7 @@ dmu_write_policy(objset_t *os, dnode_t *dn, int level, int wp, zio_prop_t *zp) } } } else if (wp & WP_NOFILL) { - ASSERT(level == 0); + ASSERT0(level); /* * If we're writing preallocated blocks, we aren't actually @@ -2508,6 +2508,7 @@ dmu_write_policy(objset_t *os, dnode_t *dn, int level, int wp, zio_prop_t *zp) zp->zp_encrypt = encrypt; zp->zp_byteorder = ZFS_HOST_BYTEORDER; zp->zp_direct_write = (wp & WP_DIRECT_WR) ? B_TRUE : B_FALSE; + zp->zp_rewrite = B_FALSE; memset(zp->zp_salt, 0, ZIO_DATA_SALT_LEN); memset(zp->zp_iv, 0, ZIO_DATA_IV_LEN); memset(zp->zp_mac, 0, ZIO_DATA_MAC_LEN); @@ -2655,11 +2656,12 @@ dmu_read_l0_bps(objset_t *os, uint64_t object, uint64_t offset, uint64_t length, * operation into ZIL, or it may be impossible to replay, since * the block may appear not yet allocated at that point. */ - if (BP_GET_BIRTH(bp) > spa_freeze_txg(os->os_spa)) { + if (BP_GET_PHYSICAL_BIRTH(bp) > spa_freeze_txg(os->os_spa)) { error = SET_ERROR(EINVAL); goto out; } - if (BP_GET_BIRTH(bp) > spa_last_synced_txg(os->os_spa)) { + if (BP_GET_PHYSICAL_BIRTH(bp) > + spa_last_synced_txg(os->os_spa)) { error = SET_ERROR(EAGAIN); goto out; } @@ -2731,7 +2733,8 @@ dmu_brt_clone(objset_t *os, uint64_t object, uint64_t offset, uint64_t length, if (!BP_IS_HOLE(bp) || BP_GET_LOGICAL_BIRTH(bp) != 0) { if (!BP_IS_EMBEDDED(bp)) { BP_SET_BIRTH(&dl->dr_overridden_by, dr->dr_txg, - BP_GET_BIRTH(bp)); + BP_GET_PHYSICAL_BIRTH(bp)); + BP_SET_REWRITE(&dl->dr_overridden_by, 0); } else { BP_SET_LOGICAL_BIRTH(&dl->dr_overridden_by, dr->dr_txg); @@ -2862,7 +2865,7 @@ byteswap_uint64_array(void *vbuf, size_t size) size_t count = size >> 3; int i; - ASSERT((size & 7) == 0); + ASSERT0((size & 7)); for (i = 0; i < count; i++) buf[i] = BSWAP_64(buf[i]); @@ -2875,7 +2878,7 @@ byteswap_uint32_array(void *vbuf, size_t size) size_t count = size >> 2; int i; - ASSERT((size & 3) == 0); + ASSERT0((size & 3)); for (i = 0; i < count; i++) buf[i] = BSWAP_32(buf[i]); @@ -2888,7 +2891,7 @@ byteswap_uint16_array(void *vbuf, size_t size) size_t count = size >> 1; int i; - ASSERT((size & 1) == 0); + ASSERT0((size & 1)); for (i = 0; i < count; i++) buf[i] = BSWAP_16(buf[i]); diff --git a/sys/contrib/openzfs/module/zfs/dmu_diff.c b/sys/contrib/openzfs/module/zfs/dmu_diff.c index 86f751e886c9..fb13b2f87f57 100644 --- a/sys/contrib/openzfs/module/zfs/dmu_diff.c +++ b/sys/contrib/openzfs/module/zfs/dmu_diff.c @@ -224,8 +224,8 @@ dmu_diff(const char *tosnap_name, const char *fromsnap_name, * call the ZFS_IOC_OBJ_TO_STATS ioctl. */ error = traverse_dataset(tosnap, fromtxg, - TRAVERSE_PRE | TRAVERSE_PREFETCH_METADATA | TRAVERSE_NO_DECRYPT, - diff_cb, &da); + TRAVERSE_PRE | TRAVERSE_PREFETCH_METADATA | TRAVERSE_NO_DECRYPT | + TRAVERSE_LOGICAL, diff_cb, &da); if (error != 0) { da.da_err = error; diff --git a/sys/contrib/openzfs/module/zfs/dmu_direct.c b/sys/contrib/openzfs/module/zfs/dmu_direct.c index 12b0ffa2c99b..d44c686088fc 100644 --- a/sys/contrib/openzfs/module/zfs/dmu_direct.c +++ b/sys/contrib/openzfs/module/zfs/dmu_direct.c @@ -95,16 +95,16 @@ dmu_write_direct_done(zio_t *zio) abd_free(zio->io_abd); mutex_enter(&db->db_mtx); - ASSERT3P(db->db_buf, ==, NULL); - ASSERT3P(dr->dt.dl.dr_data, ==, NULL); - ASSERT3P(db->db.db_data, ==, NULL); + ASSERT0P(db->db_buf); + ASSERT0P(dr->dt.dl.dr_data); + ASSERT0P(db->db.db_data); db->db_state = DB_UNCACHED; mutex_exit(&db->db_mtx); dmu_sync_done(zio, NULL, zio->io_private); if (zio->io_error != 0) { - if (zio->io_flags & ZIO_FLAG_DIO_CHKSUM_ERR) + if (zio->io_post & ZIO_POST_DIO_CHKSUM_ERR) ASSERT3U(zio->io_error, ==, EIO); /* diff --git a/sys/contrib/openzfs/module/zfs/dmu_object.c b/sys/contrib/openzfs/module/zfs/dmu_object.c index b4ff7d224cc9..207cc6d0e713 100644 --- a/sys/contrib/openzfs/module/zfs/dmu_object.c +++ b/sys/contrib/openzfs/module/zfs/dmu_object.c @@ -90,7 +90,7 @@ dmu_object_alloc_impl(objset_t *os, dmu_object_type_t ot, int blocksize, if (allocated_dnode != NULL) { ASSERT3P(tag, !=, NULL); } else { - ASSERT3P(tag, ==, NULL); + ASSERT0P(tag); tag = FTAG; } diff --git a/sys/contrib/openzfs/module/zfs/dmu_objset.c b/sys/contrib/openzfs/module/zfs/dmu_objset.c index b3f792e4ae6b..a77f338bdfd3 100644 --- a/sys/contrib/openzfs/module/zfs/dmu_objset.c +++ b/sys/contrib/openzfs/module/zfs/dmu_objset.c @@ -345,12 +345,6 @@ smallblk_changed_cb(void *arg, uint64_t newval) { objset_t *os = arg; - /* - * Inheritance and range checking should have been done by now. - */ - ASSERT(newval <= SPA_MAXBLOCKSIZE); - ASSERT(ISP2(newval)); - os->os_zpl_special_smallblock = newval; } @@ -730,7 +724,7 @@ dmu_objset_from_ds(dsl_dataset_t *ds, objset_t **osp) if (err == 0) { mutex_enter(&ds->ds_lock); - ASSERT(ds->ds_objset == NULL); + ASSERT0P(ds->ds_objset); ds->ds_objset = os; mutex_exit(&ds->ds_lock); } @@ -1376,7 +1370,7 @@ dmu_objset_create(const char *name, dmu_objset_type_t type, uint64_t flags, 6, ZFS_SPACE_CHECK_NORMAL); if (rv == 0) - zvol_create_minor(name); + zvol_create_minors(name); crfree(cr); @@ -2232,7 +2226,7 @@ dmu_objset_userquota_get_ids(dnode_t *dn, boolean_t before, dmu_tx_t *tx) rf |= DB_RF_HAVESTRUCT; error = dmu_spill_hold_by_dnode(dn, rf, FTAG, (dmu_buf_t **)&db); - ASSERT(error == 0); + ASSERT0(error); mutex_enter(&db->db_mtx); data = (before) ? db->db.db_data : dmu_objset_userquota_find_data(db, tx); diff --git a/sys/contrib/openzfs/module/zfs/dmu_recv.c b/sys/contrib/openzfs/module/zfs/dmu_recv.c index 3a4bd7a1cea9..45c7af2bdcd2 100644 --- a/sys/contrib/openzfs/module/zfs/dmu_recv.c +++ b/sys/contrib/openzfs/module/zfs/dmu_recv.c @@ -866,7 +866,7 @@ dmu_recv_begin_sync(void *arg, dmu_tx_t *tx) */ if (dcp == NULL && drrb->drr_fromguid == 0 && drba->drba_origin == NULL) { - ASSERT3P(dcp, ==, NULL); + ASSERT0P(dcp); dcp = &dummy_dcp; if (featureflags & DMU_BACKUP_FEATURE_RAW) @@ -881,7 +881,7 @@ dmu_recv_begin_sync(void *arg, dmu_tx_t *tx) if (drba->drba_cookie->drc_fromsnapobj != 0) { VERIFY0(dsl_dataset_hold_obj(dp, drba->drba_cookie->drc_fromsnapobj, FTAG, &snap)); - ASSERT3P(dcp, ==, NULL); + ASSERT0P(dcp); } if (drc->drc_heal) { /* When healing we want to use the provided snapshot */ @@ -905,7 +905,7 @@ dmu_recv_begin_sync(void *arg, dmu_tx_t *tx) if (drba->drba_origin != NULL) { VERIFY0(dsl_dataset_hold(dp, drba->drba_origin, FTAG, &origin)); - ASSERT3P(dcp, ==, NULL); + ASSERT0P(dcp); } /* Create new dataset. */ @@ -1403,7 +1403,7 @@ corrective_read_done(zio_t *zio) /* Corruption corrected; update error log if needed */ if (zio->io_error == 0) { spa_remove_error(data->spa, &data->zb, - BP_GET_LOGICAL_BIRTH(zio->io_bp)); + BP_GET_PHYSICAL_BIRTH(zio->io_bp)); } kmem_free(data, sizeof (cr_cb_data_t)); abd_free(zio->io_abd); @@ -1530,7 +1530,7 @@ do_corrective_recv(struct receive_writer_arg *rwa, struct drr_write *drrw, } rrd->abd = abd; - io = zio_rewrite(NULL, rwa->os->os_spa, BP_GET_LOGICAL_BIRTH(bp), bp, + io = zio_rewrite(NULL, rwa->os->os_spa, BP_GET_BIRTH(bp), bp, abd, BP_GET_PSIZE(bp), NULL, NULL, ZIO_PRIORITY_SYNC_WRITE, flags, &zb); @@ -2792,7 +2792,7 @@ receive_read_payload_and_next_header(dmu_recv_cookie_t *drc, int len, void *buf) drc->drc_rrd->bytes_read = drc->drc_bytes_read; } } else { - ASSERT3P(buf, ==, NULL); + ASSERT0P(buf); } drc->drc_prev_cksum = drc->drc_cksum; @@ -3450,7 +3450,7 @@ dmu_recv_stream(dmu_recv_cookie_t *drc, offset_t *voffp) break; } - ASSERT3P(drc->drc_rrd, ==, NULL); + ASSERT0P(drc->drc_rrd); drc->drc_rrd = drc->drc_next_rrd; drc->drc_next_rrd = NULL; /* Allocates and loads header into drc->drc_next_rrd */ @@ -3468,7 +3468,7 @@ dmu_recv_stream(dmu_recv_cookie_t *drc, offset_t *voffp) drc->drc_rrd = NULL; } - ASSERT3P(drc->drc_rrd, ==, NULL); + ASSERT0P(drc->drc_rrd); drc->drc_rrd = kmem_zalloc(sizeof (*drc->drc_rrd), KM_SLEEP); drc->drc_rrd->eos_marker = B_TRUE; bqueue_enqueue_flush(&rwa->q, drc->drc_rrd, 1); @@ -3831,11 +3831,11 @@ dmu_recv_end(dmu_recv_cookie_t *drc, void *owner) nvlist_free(drc->drc_keynvl); } else if (!drc->drc_heal) { if (drc->drc_newfs) { - zvol_create_minor(drc->drc_tofs); + zvol_create_minors(drc->drc_tofs); } char *snapname = kmem_asprintf("%s@%s", drc->drc_tofs, drc->drc_tosnap); - zvol_create_minor(snapname); + zvol_create_minors(snapname); kmem_strfree(snapname); } diff --git a/sys/contrib/openzfs/module/zfs/dmu_redact.c b/sys/contrib/openzfs/module/zfs/dmu_redact.c index 65443d112f27..5a22ed71a5fe 100644 --- a/sys/contrib/openzfs/module/zfs/dmu_redact.c +++ b/sys/contrib/openzfs/module/zfs/dmu_redact.c @@ -370,8 +370,8 @@ redact_traverse_thread(void *arg) #endif err = traverse_dataset_resume(rt_arg->ds, rt_arg->txg, - &rt_arg->resume, TRAVERSE_PRE | TRAVERSE_PREFETCH_METADATA, - redact_cb, rt_arg); + &rt_arg->resume, TRAVERSE_PRE | TRAVERSE_PREFETCH_METADATA | + TRAVERSE_LOGICAL, redact_cb, rt_arg); if (err != EINTR) rt_arg->error_code = err; @@ -1067,7 +1067,7 @@ dmu_redact_snap(const char *snapname, nvlist_t *redactnvl, } if (err != 0) goto out; - VERIFY3P(nvlist_next_nvpair(redactnvl, pair), ==, NULL); + VERIFY0P(nvlist_next_nvpair(redactnvl, pair)); boolean_t resuming = B_FALSE; zfs_bookmark_phys_t bookmark; diff --git a/sys/contrib/openzfs/module/zfs/dmu_send.c b/sys/contrib/openzfs/module/zfs/dmu_send.c index 4f27f3df0e55..8ecb99d5f57c 100644 --- a/sys/contrib/openzfs/module/zfs/dmu_send.c +++ b/sys/contrib/openzfs/module/zfs/dmu_send.c @@ -962,7 +962,7 @@ do_dump(dmu_send_cookie_t *dscp, struct send_range *range) char *data = NULL; if (srdp->abd != NULL) { data = abd_to_buf(srdp->abd); - ASSERT3P(srdp->abuf, ==, NULL); + ASSERT0P(srdp->abuf); } else if (srdp->abuf != NULL) { data = srdp->abuf->b_data; } @@ -1084,7 +1084,7 @@ send_cb(spa_t *spa, zilog_t *zilog, const blkptr_t *bp, */ if (sta->os->os_encrypted && !BP_IS_HOLE(bp) && !BP_USES_CRYPT(bp)) { - spa_log_error(spa, zb, BP_GET_LOGICAL_BIRTH(bp)); + spa_log_error(spa, zb, BP_GET_PHYSICAL_BIRTH(bp)); return (SET_ERROR(EIO)); } @@ -1210,7 +1210,7 @@ send_traverse_thread(void *arg) err = traverse_dataset_resume(st_arg->os->os_dsl_dataset, st_arg->fromtxg, &st_arg->resume, - st_arg->flags, send_cb, st_arg); + st_arg->flags | TRAVERSE_LOGICAL, send_cb, st_arg); if (err != EINTR) st_arg->error_code = err; @@ -2514,7 +2514,7 @@ dmu_send_impl(struct dmu_send_params *dspp) * list in the stream. */ if (dspp->numfromredactsnaps != NUM_SNAPS_NOT_REDACTED) { - ASSERT3P(from_rl, ==, NULL); + ASSERT0P(from_rl); fnvlist_add_uint64_array(nvl, BEGINNV_REDACT_FROM_SNAPS, dspp->fromredactsnaps, (uint_t)dspp->numfromredactsnaps); if (dspp->numfromredactsnaps > 0) { @@ -2891,7 +2891,7 @@ dmu_send(const char *tosnap, const char *fromsnap, boolean_t embedok, &fromds); if (err != 0) { - ASSERT3P(fromds, ==, NULL); + ASSERT0P(fromds); } else { /* * We need to make a deep copy of the redact diff --git a/sys/contrib/openzfs/module/zfs/dmu_traverse.c b/sys/contrib/openzfs/module/zfs/dmu_traverse.c index f534a7dd64e3..dd1df1705040 100644 --- a/sys/contrib/openzfs/module/zfs/dmu_traverse.c +++ b/sys/contrib/openzfs/module/zfs/dmu_traverse.c @@ -74,6 +74,15 @@ static int traverse_dnode(traverse_data_t *td, const blkptr_t *bp, static void prefetch_dnode_metadata(traverse_data_t *td, const dnode_phys_t *, uint64_t objset, uint64_t object); +static inline uint64_t +get_birth_time(traverse_data_t *td, const blkptr_t *bp) +{ + if (td->td_flags & TRAVERSE_LOGICAL) + return (BP_GET_LOGICAL_BIRTH(bp)); + else + return (BP_GET_BIRTH(bp)); +} + static int traverse_zil_block(zilog_t *zilog, const blkptr_t *bp, void *arg, uint64_t claim_txg) @@ -85,7 +94,7 @@ traverse_zil_block(zilog_t *zilog, const blkptr_t *bp, void *arg, return (0); if (claim_txg == 0 && - BP_GET_LOGICAL_BIRTH(bp) >= spa_min_claim_txg(td->td_spa)) + get_birth_time(td, bp) >= spa_min_claim_txg(td->td_spa)) return (-1); SET_BOOKMARK(&zb, td->td_objset, ZB_ZIL_OBJECT, ZB_ZIL_LEVEL, @@ -110,7 +119,7 @@ traverse_zil_record(zilog_t *zilog, const lr_t *lrc, void *arg, if (BP_IS_HOLE(bp)) return (0); - if (claim_txg == 0 || BP_GET_LOGICAL_BIRTH(bp) < claim_txg) + if (claim_txg == 0 || get_birth_time(td, bp) < claim_txg) return (0); ASSERT3U(BP_GET_LSIZE(bp), !=, 0); @@ -194,7 +203,7 @@ traverse_prefetch_metadata(traverse_data_t *td, const dnode_phys_t *dnp, */ if (resume_skip_check(td, dnp, zb) != RESUME_SKIP_NONE) return (B_FALSE); - if (BP_IS_HOLE(bp) || BP_GET_LOGICAL_BIRTH(bp) <= td->td_min_txg) + if (BP_IS_HOLE(bp) || get_birth_time(td, bp) <= td->td_min_txg) return (B_FALSE); if (BP_GET_LEVEL(bp) == 0 && BP_GET_TYPE(bp) != DMU_OT_DNODE) return (B_FALSE); @@ -265,7 +274,7 @@ traverse_visitbp(traverse_data_t *td, const dnode_phys_t *dnp, zb->zb_object == DMU_META_DNODE_OBJECT) && td->td_hole_birth_enabled_txg <= td->td_min_txg) return (0); - } else if (BP_GET_LOGICAL_BIRTH(bp) <= td->td_min_txg) { + } else if (get_birth_time(td, bp) <= td->td_min_txg) { return (0); } diff --git a/sys/contrib/openzfs/module/zfs/dmu_tx.c b/sys/contrib/openzfs/module/zfs/dmu_tx.c index d85d8b89423e..40c0b3402a05 100644 --- a/sys/contrib/openzfs/module/zfs/dmu_tx.c +++ b/sys/contrib/openzfs/module/zfs/dmu_tx.c @@ -126,7 +126,7 @@ dmu_tx_hold_dnode_impl(dmu_tx_t *tx, dnode_t *dn, enum dmu_tx_hold_type type, * problem, but there's no way for it to happen (for * now, at least). */ - ASSERT(dn->dn_assigned_txg == 0); + ASSERT0(dn->dn_assigned_txg); dn->dn_assigned_txg = tx->tx_txg; (void) zfs_refcount_add(&dn->dn_tx_holds, tx); mutex_exit(&dn->dn_mtx); @@ -443,7 +443,7 @@ dmu_tx_count_free(dmu_tx_hold_t *txh, uint64_t off, uint64_t len) dnode_t *dn = txh->txh_dnode; int err; - ASSERT(tx->tx_txg == 0); + ASSERT0(tx->tx_txg); if (off >= (dn->dn_maxblkid + 1) * dn->dn_datablksz) return; @@ -607,7 +607,7 @@ dmu_tx_hold_zap_impl(dmu_tx_hold_t *txh, const char *name) dnode_t *dn = txh->txh_dnode; int err; - ASSERT(tx->tx_txg == 0); + ASSERT0(tx->tx_txg); dmu_tx_count_dnode(txh); @@ -681,7 +681,7 @@ dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object) { dmu_tx_hold_t *txh; - ASSERT(tx->tx_txg == 0); + ASSERT0(tx->tx_txg); txh = dmu_tx_hold_object_impl(tx, tx->tx_objset, object, THT_BONUS, 0, 0); @@ -706,7 +706,7 @@ dmu_tx_hold_space(dmu_tx_t *tx, uint64_t space) { dmu_tx_hold_t *txh; - ASSERT(tx->tx_txg == 0); + ASSERT0(tx->tx_txg); txh = dmu_tx_hold_object_impl(tx, tx->tx_objset, DMU_NEW_OBJECT, THT_SPACE, space, 0); @@ -1232,7 +1232,7 @@ dmu_tx_assign(dmu_tx_t *tx, dmu_tx_flag_t flags) { int err; - ASSERT(tx->tx_txg == 0); + ASSERT0(tx->tx_txg); ASSERT0(flags & ~(DMU_TX_WAIT | DMU_TX_NOTHROTTLE | DMU_TX_SUSPEND)); IMPLY(flags & DMU_TX_SUSPEND, flags & DMU_TX_WAIT); ASSERT(!dsl_pool_sync_context(tx->tx_pool)); @@ -1328,7 +1328,7 @@ dmu_tx_wait(dmu_tx_t *tx) dsl_pool_t *dp = tx->tx_pool; hrtime_t before; - ASSERT(tx->tx_txg == 0); + ASSERT0(tx->tx_txg); ASSERT(!dsl_pool_config_held(tx->tx_pool)); /* @@ -1644,12 +1644,12 @@ dmu_tx_hold_sa(dmu_tx_t *tx, sa_handle_t *hdl, boolean_t may_grow) dmu_tx_hold_zap(tx, sa->sa_layout_attr_obj, B_TRUE, NULL); if (sa->sa_force_spill || may_grow || hdl->sa_spill) { - ASSERT(tx->tx_txg == 0); + ASSERT0(tx->tx_txg); dmu_tx_hold_spill(tx, object); } else { DB_DNODE_ENTER(db); if (DB_DNODE(db)->dn_have_spill) { - ASSERT(tx->tx_txg == 0); + ASSERT0(tx->tx_txg); dmu_tx_hold_spill(tx, object); } DB_DNODE_EXIT(db); diff --git a/sys/contrib/openzfs/module/zfs/dnode.c b/sys/contrib/openzfs/module/zfs/dnode.c index 904a039edf95..963ff41232a3 100644 --- a/sys/contrib/openzfs/module/zfs/dnode.c +++ b/sys/contrib/openzfs/module/zfs/dnode.c @@ -86,6 +86,19 @@ int zfs_default_ibs = DN_MAX_INDBLKSHIFT; static kmem_cbrc_t dnode_move(void *, void *, size_t, void *); #endif /* _KERNEL */ +static char * +rt_name(dnode_t *dn, const char *name) +{ + struct objset *os = dn->dn_objset; + + return (kmem_asprintf("{spa=%s objset=%llu obj=%llu %s}", + spa_name(os->os_spa), + (u_longlong_t)(os->os_dsl_dataset ? + os->os_dsl_dataset->ds_object : DMU_META_OBJSET), + (u_longlong_t)dn->dn_object, + name)); +} + static int dbuf_compare(const void *x1, const void *x2) { @@ -201,7 +214,7 @@ dnode_dest(void *arg, void *unused) for (int i = 0; i < TXG_SIZE; i++) { ASSERT(!multilist_link_active(&dn->dn_dirty_link[i])); - ASSERT3P(dn->dn_free_ranges[i], ==, NULL); + ASSERT0P(dn->dn_free_ranges[i]); list_destroy(&dn->dn_dirty_records[i]); ASSERT0(dn->dn_next_nblkptr[i]); ASSERT0(dn->dn_next_nlevels[i]); @@ -218,10 +231,10 @@ dnode_dest(void *arg, void *unused) ASSERT0(dn->dn_assigned_txg); ASSERT0(dn->dn_dirty_txg); ASSERT0(dn->dn_dirtyctx); - ASSERT3P(dn->dn_dirtyctx_firstset, ==, NULL); - ASSERT3P(dn->dn_bonus, ==, NULL); + ASSERT0P(dn->dn_dirtyctx_firstset); + ASSERT0P(dn->dn_bonus); ASSERT(!dn->dn_have_spill); - ASSERT3P(dn->dn_zio, ==, NULL); + ASSERT0P(dn->dn_zio); ASSERT0(dn->dn_oldused); ASSERT0(dn->dn_oldflags); ASSERT0(dn->dn_olduid); @@ -305,7 +318,7 @@ dnode_kstats_update(kstat_t *ksp, int rw) void dnode_init(void) { - ASSERT(dnode_cache == NULL); + ASSERT0P(dnode_cache); dnode_cache = kmem_cache_create("dnode_t", sizeof (dnode_t), 0, dnode_cons, dnode_dest, NULL, NULL, NULL, KMC_RECLAIMABLE); kmem_cache_set_move(dnode_cache, dnode_move); @@ -496,7 +509,7 @@ dnode_buf_byteswap(void *vbuf, size_t size) int i = 0; ASSERT3U(sizeof (dnode_phys_t), ==, (1<<DNODE_SHIFT)); - ASSERT((size & (sizeof (dnode_phys_t)-1)) == 0); + ASSERT0((size & (sizeof (dnode_phys_t)-1))); while (i < size) { dnode_phys_t *dnp = (void *)(((char *)vbuf) + i); @@ -660,7 +673,7 @@ dnode_destroy(dnode_t *dn) objset_t *os = dn->dn_objset; boolean_t complete_os_eviction = B_FALSE; - ASSERT((dn->dn_id_flags & DN_ID_NEW_EXIST) == 0); + ASSERT0((dn->dn_id_flags & DN_ID_NEW_EXIST)); mutex_enter(&os->os_lock); POINTER_INVALIDATE(&dn->dn_objset); @@ -767,7 +780,7 @@ dnode_allocate(dnode_t *dn, dmu_object_type_t ot, int blocksize, int ibs, ASSERT0(dn->dn_next_maxblkid[i]); ASSERT(!multilist_link_active(&dn->dn_dirty_link[i])); ASSERT3P(list_head(&dn->dn_dirty_records[i]), ==, NULL); - ASSERT3P(dn->dn_free_ranges[i], ==, NULL); + ASSERT0P(dn->dn_free_ranges[i]); } dn->dn_type = ot; @@ -945,7 +958,7 @@ dnode_move_impl(dnode_t *odn, dnode_t *ndn) ndn->dn_dirty_txg = odn->dn_dirty_txg; ndn->dn_dirtyctx = odn->dn_dirtyctx; ndn->dn_dirtyctx_firstset = odn->dn_dirtyctx_firstset; - ASSERT(zfs_refcount_count(&odn->dn_tx_holds) == 0); + ASSERT0(zfs_refcount_count(&odn->dn_tx_holds)); zfs_refcount_transfer(&ndn->dn_holds, &odn->dn_holds); ASSERT(avl_is_empty(&ndn->dn_dbufs)); avl_swap(&ndn->dn_dbufs, &odn->dn_dbufs); @@ -2291,7 +2304,7 @@ dnode_free_range(dnode_t *dn, uint64_t off, uint64_t len, dmu_tx_t *tx) if ((off >> blkshift) > dn->dn_maxblkid) return; } else { - ASSERT(dn->dn_maxblkid == 0); + ASSERT0(dn->dn_maxblkid); if (off == 0 && len >= blksz) { /* * Freeing the whole block; fast-track this request. @@ -2436,8 +2449,10 @@ done: { int txgoff = tx->tx_txg & TXG_MASK; if (dn->dn_free_ranges[txgoff] == NULL) { - dn->dn_free_ranges[txgoff] = zfs_range_tree_create(NULL, - ZFS_RANGE_SEG64, NULL, 0, 0); + dn->dn_free_ranges[txgoff] = + zfs_range_tree_create_flags( + NULL, ZFS_RANGE_SEG64, NULL, 0, 0, + ZFS_RT_F_DYN_NAME, rt_name(dn, "dn_free_ranges")); } zfs_range_tree_clear(dn->dn_free_ranges[txgoff], blkid, nblks); zfs_range_tree_add(dn->dn_free_ranges[txgoff], blkid, nblks); @@ -2509,7 +2524,7 @@ dnode_diduse_space(dnode_t *dn, int64_t delta) } space += delta; if (spa_version(dn->dn_objset->os_spa) < SPA_VERSION_DNODE_BYTES) { - ASSERT((dn->dn_phys->dn_flags & DNODE_FLAG_USED_BYTES) == 0); + ASSERT0((dn->dn_phys->dn_flags & DNODE_FLAG_USED_BYTES)); ASSERT0(P2PHASE(space, 1<<DEV_BSHIFT)); dn->dn_phys->dn_used = space >> DEV_BSHIFT; } else { diff --git a/sys/contrib/openzfs/module/zfs/dnode_sync.c b/sys/contrib/openzfs/module/zfs/dnode_sync.c index 4067f221f1bf..046ceddb3609 100644 --- a/sys/contrib/openzfs/module/zfs/dnode_sync.c +++ b/sys/contrib/openzfs/module/zfs/dnode_sync.c @@ -209,8 +209,8 @@ free_verify(dmu_buf_impl_t *db, uint64_t start, uint64_t end, dmu_tx_t *tx) rw_exit(&dn->dn_struct_rwlock); if (err == ENOENT) continue; - ASSERT(err == 0); - ASSERT(child->db_level == 0); + ASSERT0(err); + ASSERT0(child->db_level); dr = dbuf_find_dirty_eq(child, txg); /* data_old better be zeroed */ @@ -868,7 +868,7 @@ dnode_sync(dnode_t *dn, dmu_tx_t *tx) dbuf_sync_list(list, dn->dn_phys->dn_nlevels - 1, tx); if (!DMU_OBJECT_IS_SPECIAL(dn->dn_object)) { - ASSERT3P(list_head(list), ==, NULL); + ASSERT0P(list_head(list)); dnode_rele(dn, (void *)(uintptr_t)tx->tx_txg); } diff --git a/sys/contrib/openzfs/module/zfs/dsl_bookmark.c b/sys/contrib/openzfs/module/zfs/dsl_bookmark.c index e301fe19f645..ee574c499f9f 100644 --- a/sys/contrib/openzfs/module/zfs/dsl_bookmark.c +++ b/sys/contrib/openzfs/module/zfs/dsl_bookmark.c @@ -243,7 +243,7 @@ dsl_bookmark_create_check_impl(dsl_pool_t *dp, /* error is retval of the following if-cascade */ if (strchr(source, '@') != NULL) { dsl_dataset_t *source_snap_ds; - ASSERT3S(snapshot_namecheck(source, NULL, NULL), ==, 0); + ASSERT0(snapshot_namecheck(source, NULL, NULL)); error = dsl_dataset_hold(dp, source, FTAG, &source_snap_ds); if (error == 0) { VERIFY(source_snap_ds->ds_is_snapshot); @@ -258,7 +258,7 @@ dsl_bookmark_create_check_impl(dsl_pool_t *dp, } } else if (strchr(source, '#') != NULL) { zfs_bookmark_phys_t source_phys; - ASSERT3S(bookmark_namecheck(source, NULL, NULL), ==, 0); + ASSERT0(bookmark_namecheck(source, NULL, NULL)); /* * Source must exists and be an earlier point in newbm_ds's * timeline (newbm_ds's origin may be a snap of source's ds) @@ -501,7 +501,7 @@ dsl_bookmark_create_sync_impl_snap(const char *bookmark, const char *snapshot, sizeof (uint64_t) * num_redact_snaps); local_rl->rl_phys->rlp_num_snaps = num_redact_snaps; if (bookmark_redacted) { - ASSERT3P(redaction_list, ==, NULL); + ASSERT0P(redaction_list); local_rl->rl_phys->rlp_last_blkid = UINT64_MAX; local_rl->rl_phys->rlp_last_object = UINT64_MAX; dsl_redaction_list_long_rele(local_rl, tag); @@ -1523,7 +1523,7 @@ dsl_bookmark_block_killed(dsl_dataset_t *ds, const blkptr_t *bp, dmu_tx_t *tx) * If the block was live (referenced) at the time of this * bookmark, add its space to the bookmark's FBN. */ - if (BP_GET_LOGICAL_BIRTH(bp) <= + if (BP_GET_BIRTH(bp) <= dbn->dbn_phys.zbm_creation_txg && (dbn->dbn_phys.zbm_flags & ZBM_FLAG_HAS_FBN)) { mutex_enter(&dbn->dbn_lock); diff --git a/sys/contrib/openzfs/module/zfs/dsl_crypt.c b/sys/contrib/openzfs/module/zfs/dsl_crypt.c index db568f42d24e..f519b937edc0 100644 --- a/sys/contrib/openzfs/module/zfs/dsl_crypt.c +++ b/sys/contrib/openzfs/module/zfs/dsl_crypt.c @@ -534,7 +534,7 @@ out: static void dsl_crypto_key_free(dsl_crypto_key_t *dck) { - ASSERT(zfs_refcount_count(&dck->dck_holds) == 0); + ASSERT0(zfs_refcount_count(&dck->dck_holds)); /* destroy the zio_crypt_key_t */ zio_crypt_key_destroy(&dck->dck_key); @@ -866,7 +866,7 @@ spa_keystore_load_wkey(const char *dsname, dsl_crypto_params_t *dcp, dsl_pool_rele(dp, FTAG); /* create any zvols under this ds */ - zvol_create_minors_recursive(dsname); + zvol_create_minors(dsname); return (0); @@ -1912,7 +1912,7 @@ dsl_dataset_create_crypt_sync(uint64_t dsobj, dsl_dir_t *dd, /* clones always use their origin's wrapping key */ if (dsl_dir_is_clone(dd)) { - ASSERT3P(dcp, ==, NULL); + ASSERT0P(dcp); /* * If this is an encrypted clone we just need to clone the diff --git a/sys/contrib/openzfs/module/zfs/dsl_dataset.c b/sys/contrib/openzfs/module/zfs/dsl_dataset.c index c0a7872c40ad..420687480a76 100644 --- a/sys/contrib/openzfs/module/zfs/dsl_dataset.c +++ b/sys/contrib/openzfs/module/zfs/dsl_dataset.c @@ -159,7 +159,7 @@ dsl_dataset_block_born(dsl_dataset_t *ds, const blkptr_t *bp, dmu_tx_t *tx) return; } - ASSERT3U(BP_GET_LOGICAL_BIRTH(bp), >, + ASSERT3U(BP_GET_BIRTH(bp), >, dsl_dataset_phys(ds)->ds_prev_snap_txg); dmu_buf_will_dirty(ds->ds_dbuf, tx); mutex_enter(&ds->ds_lock); @@ -194,7 +194,7 @@ dsl_dataset_block_born(dsl_dataset_t *ds, const blkptr_t *bp, dmu_tx_t *tx) * they do not need to be freed. */ if (dsl_deadlist_is_open(&ds->ds_dir->dd_livelist) && - BP_GET_LOGICAL_BIRTH(bp) > ds->ds_dir->dd_origin_txg && + BP_GET_BIRTH(bp) > ds->ds_dir->dd_origin_txg && !(BP_IS_EMBEDDED(bp))) { ASSERT(dsl_dir_is_clone(ds->ds_dir)); ASSERT(spa_feature_is_enabled(spa, @@ -263,7 +263,7 @@ dsl_dataset_block_kill(dsl_dataset_t *ds, const blkptr_t *bp, dmu_tx_t *tx, return (0); ASSERT(dmu_tx_is_syncing(tx)); - ASSERT(BP_GET_LOGICAL_BIRTH(bp) <= tx->tx_txg); + ASSERT(BP_GET_BIRTH(bp) <= tx->tx_txg); if (ds == NULL) { dsl_free(tx->tx_pool, tx->tx_txg, bp); @@ -281,7 +281,7 @@ dsl_dataset_block_kill(dsl_dataset_t *ds, const blkptr_t *bp, dmu_tx_t *tx, * they do not need to be freed. */ if (dsl_deadlist_is_open(&ds->ds_dir->dd_livelist) && - BP_GET_LOGICAL_BIRTH(bp) > ds->ds_dir->dd_origin_txg && + BP_GET_BIRTH(bp) > ds->ds_dir->dd_origin_txg && !(BP_IS_EMBEDDED(bp))) { ASSERT(dsl_dir_is_clone(ds->ds_dir)); ASSERT(spa_feature_is_enabled(spa, @@ -289,7 +289,7 @@ dsl_dataset_block_kill(dsl_dataset_t *ds, const blkptr_t *bp, dmu_tx_t *tx, bplist_append(&ds->ds_dir->dd_pending_frees, bp); } - if (BP_GET_LOGICAL_BIRTH(bp) > dsl_dataset_phys(ds)->ds_prev_snap_txg) { + if (BP_GET_BIRTH(bp) > dsl_dataset_phys(ds)->ds_prev_snap_txg) { int64_t delta; /* @@ -346,14 +346,14 @@ dsl_dataset_block_kill(dsl_dataset_t *ds, const blkptr_t *bp, dmu_tx_t *tx, ASSERT(dsl_dataset_phys(ds->ds_prev)->ds_num_children > 0); /* if (logical birth > prev prev snap txg) prev unique += bs */ if (dsl_dataset_phys(ds->ds_prev)->ds_next_snap_obj == - ds->ds_object && BP_GET_LOGICAL_BIRTH(bp) > + ds->ds_object && BP_GET_BIRTH(bp) > dsl_dataset_phys(ds->ds_prev)->ds_prev_snap_txg) { dmu_buf_will_dirty(ds->ds_prev->ds_dbuf, tx); mutex_enter(&ds->ds_prev->ds_lock); dsl_dataset_phys(ds->ds_prev)->ds_unique_bytes += used; mutex_exit(&ds->ds_prev->ds_lock); } - if (BP_GET_LOGICAL_BIRTH(bp) > ds->ds_dir->dd_origin_txg) { + if (BP_GET_BIRTH(bp) > ds->ds_dir->dd_origin_txg) { dsl_dir_transfer_space(ds->ds_dir, used, DD_USED_HEAD, DD_USED_SNAP, tx); } @@ -450,7 +450,7 @@ dsl_dataset_evict_sync(void *dbu) { dsl_dataset_t *ds = dbu; - ASSERT(ds->ds_owner == NULL); + ASSERT0P(ds->ds_owner); unique_remove(ds->ds_fsid_guid); } @@ -460,7 +460,7 @@ dsl_dataset_evict_async(void *dbu) { dsl_dataset_t *ds = dbu; - ASSERT(ds->ds_owner == NULL); + ASSERT0P(ds->ds_owner); ds->ds_dbuf = NULL; @@ -1187,7 +1187,7 @@ dsl_dataset_create_sync_dd(dsl_dir_t *dd, dsl_dataset_t *origin, ASSERT(origin == NULL || origin->ds_dir->dd_pool == dp); ASSERT(origin == NULL || dsl_dataset_phys(origin)->ds_num_children > 0); ASSERT(dmu_tx_is_syncing(tx)); - ASSERT(dsl_dir_phys(dd)->dd_head_dataset_obj == 0); + ASSERT0(dsl_dir_phys(dd)->dd_head_dataset_obj); dsobj = dmu_object_alloc(mos, DMU_OT_DSL_DATASET, 0, DMU_OT_DSL_DATASET, sizeof (dsl_dataset_phys_t), tx); @@ -2005,7 +2005,7 @@ dsl_dataset_snapshot(nvlist_t *snaps, nvlist_t *props, nvlist_t *errors) if (error == 0) { for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL; pair = nvlist_next_nvpair(snaps, pair)) { - zvol_create_minor(nvpair_name(pair)); + zvol_create_minors(nvpair_name(pair)); } } @@ -2112,7 +2112,7 @@ dsl_dataset_sync(dsl_dataset_t *ds, zio_t *rio, dmu_tx_t *tx) { ASSERT(dmu_tx_is_syncing(tx)); ASSERT(ds->ds_objset != NULL); - ASSERT(dsl_dataset_phys(ds)->ds_next_snap_obj == 0); + ASSERT0(dsl_dataset_phys(ds)->ds_next_snap_obj); /* * in case we had to change ds_fsid_guid when we opened it, @@ -2944,7 +2944,7 @@ dsl_dataset_modified_since_snap(dsl_dataset_t *ds, dsl_dataset_t *snap) if (snap == NULL) return (B_FALSE); rrw_enter(&ds->ds_bp_rwlock, RW_READER, FTAG); - birth = BP_GET_LOGICAL_BIRTH(dsl_dataset_get_blkptr(ds)); + birth = BP_GET_BIRTH(dsl_dataset_get_blkptr(ds)); rrw_exit(&ds->ds_bp_rwlock, FTAG); if (birth > dsl_dataset_phys(snap)->ds_creation_txg) { objset_t *os, *os_snap; @@ -3413,7 +3413,7 @@ dsl_dataset_clone(const char *clone, const char *origin) 6, ZFS_SPACE_CHECK_NORMAL); if (rv == 0) - zvol_create_minor(clone); + zvol_create_minors(clone); crfree(cr); @@ -4180,7 +4180,7 @@ dsl_dataset_clone_swap_sync_impl(dsl_dataset_t *clone, dsl_pool_t *dp = dmu_tx_pool(tx); int64_t unused_refres_delta; - ASSERT(clone->ds_reserved == 0); + ASSERT0(clone->ds_reserved); /* * NOTE: On DEBUG kernels there could be a race between this and * the check function if spa_asize_inflation is adjusted... diff --git a/sys/contrib/openzfs/module/zfs/dsl_deadlist.c b/sys/contrib/openzfs/module/zfs/dsl_deadlist.c index 3113d932fb68..475db3c89508 100644 --- a/sys/contrib/openzfs/module/zfs/dsl_deadlist.c +++ b/sys/contrib/openzfs/module/zfs/dsl_deadlist.c @@ -484,7 +484,7 @@ dsl_deadlist_insert(dsl_deadlist_t *dl, const blkptr_t *bp, boolean_t bp_freed, dl->dl_phys->dl_comp += sign * BP_GET_PSIZE(bp); dl->dl_phys->dl_uncomp += sign * BP_GET_UCSIZE(bp); - dle_tofind.dle_mintxg = BP_GET_LOGICAL_BIRTH(bp); + dle_tofind.dle_mintxg = BP_GET_BIRTH(bp); dle = avl_find(&dl->dl_tree, &dle_tofind, &where); if (dle == NULL) dle = avl_nearest(&dl->dl_tree, where, AVL_BEFORE); @@ -493,7 +493,7 @@ dsl_deadlist_insert(dsl_deadlist_t *dl, const blkptr_t *bp, boolean_t bp_freed, if (dle == NULL) { zfs_panic_recover("blkptr at %p has invalid BLK_BIRTH %llu", - bp, (longlong_t)BP_GET_LOGICAL_BIRTH(bp)); + bp, (longlong_t)BP_GET_BIRTH(bp)); dle = avl_first(&dl->dl_tree); } @@ -1037,7 +1037,7 @@ dsl_livelist_iterate(void *arg, const blkptr_t *bp, boolean_t bp_freed, avl_tree_t *avl = lia->avl; bplist_t *to_free = lia->to_free; zthr_t *t = lia->t; - ASSERT(tx == NULL); + ASSERT0P(tx); if ((t != NULL) && (zthr_has_waiters(t) || zthr_iscancelled(t))) return (SET_ERROR(EINTR)); diff --git a/sys/contrib/openzfs/module/zfs/dsl_deleg.c b/sys/contrib/openzfs/module/zfs/dsl_deleg.c index c01a06e98340..fdd37b36e280 100644 --- a/sys/contrib/openzfs/module/zfs/dsl_deleg.c +++ b/sys/contrib/openzfs/module/zfs/dsl_deleg.c @@ -102,7 +102,7 @@ dsl_deleg_can_allow(char *ddname, nvlist_t *nvp, cred_t *cr) nvlist_t *perms; nvpair_t *permpair = NULL; - VERIFY(nvpair_value_nvlist(whopair, &perms) == 0); + VERIFY0(nvpair_value_nvlist(whopair, &perms)); while ((permpair = nvlist_next_nvpair(perms, permpair))) { const char *perm = nvpair_name(permpair); @@ -189,8 +189,7 @@ dsl_deleg_set_sync(void *arg, dmu_tx_t *tx) const char *perm = nvpair_name(permpair); uint64_t n = 0; - VERIFY(zap_update(mos, jumpobj, - perm, 8, 1, &n, tx) == 0); + VERIFY0(zap_update(mos, jumpobj, perm, 8, 1, &n, tx)); spa_history_log_internal_dd(dd, "permission update", tx, "%s %s", whokey, perm); } @@ -225,7 +224,7 @@ dsl_deleg_unset_sync(void *arg, dmu_tx_t *tx) if (zap_lookup(mos, zapobj, whokey, 8, 1, &jumpobj) == 0) { (void) zap_remove(mos, zapobj, whokey, tx); - VERIFY(0 == zap_destroy(mos, jumpobj, tx)); + VERIFY0(zap_destroy(mos, jumpobj, tx)); } spa_history_log_internal_dd(dd, "permission who remove", tx, "%s", whokey); @@ -243,7 +242,7 @@ dsl_deleg_unset_sync(void *arg, dmu_tx_t *tx) if (zap_count(mos, jumpobj, &n) == 0 && n == 0) { (void) zap_remove(mos, zapobj, whokey, tx); - VERIFY(0 == zap_destroy(mos, + VERIFY0(zap_destroy(mos, jumpobj, tx)); } spa_history_log_internal_dd(dd, "permission remove", tx, @@ -332,7 +331,7 @@ dsl_deleg_get(const char *ddname, nvlist_t **nvp) basezc = kmem_alloc(sizeof (zap_cursor_t), KM_SLEEP); baseza = zap_attribute_alloc(); source = kmem_alloc(ZFS_MAX_DATASET_NAME_LEN, KM_SLEEP); - VERIFY(nvlist_alloc(nvp, NV_UNIQUE_NAME, KM_SLEEP) == 0); + VERIFY0(nvlist_alloc(nvp, NV_UNIQUE_NAME, KM_SLEEP)); for (dd = startdd; dd != NULL; dd = dd->dd_parent) { nvlist_t *sp_nvp; @@ -706,7 +705,7 @@ copy_create_perms(dsl_dir_t *dd, uint64_t pzapobj, ZFS_DELEG_LOCAL, &uid); if (zap_lookup(mos, zapobj, whokey, 8, 1, &jumpobj) == ENOENT) { jumpobj = zap_create(mos, DMU_OT_DSL_PERMS, DMU_OT_NONE, 0, tx); - VERIFY(zap_add(mos, zapobj, whokey, 8, 1, &jumpobj, tx) == 0); + VERIFY0(zap_add(mos, zapobj, whokey, 8, 1, &jumpobj, tx)); } za = zap_attribute_alloc(); @@ -716,8 +715,7 @@ copy_create_perms(dsl_dir_t *dd, uint64_t pzapobj, uint64_t zero = 0; ASSERT(za->za_integer_length == 8 && za->za_num_integers == 1); - VERIFY(zap_update(mos, jumpobj, za->za_name, - 8, 1, &zero, tx) == 0); + VERIFY0(zap_update(mos, jumpobj, za->za_name, 8, 1, &zero, tx)); } zap_cursor_fini(&zc); zap_attribute_free(za); @@ -761,10 +759,10 @@ dsl_deleg_destroy(objset_t *mos, uint64_t zapobj, dmu_tx_t *tx) zap_cursor_retrieve(&zc, za) == 0; zap_cursor_advance(&zc)) { ASSERT(za->za_integer_length == 8 && za->za_num_integers == 1); - VERIFY(0 == zap_destroy(mos, za->za_first_integer, tx)); + VERIFY0(zap_destroy(mos, za->za_first_integer, tx)); } zap_cursor_fini(&zc); - VERIFY(0 == zap_destroy(mos, zapobj, tx)); + VERIFY0(zap_destroy(mos, zapobj, tx)); zap_attribute_free(za); return (0); } diff --git a/sys/contrib/openzfs/module/zfs/dsl_destroy.c b/sys/contrib/openzfs/module/zfs/dsl_destroy.c index f5ec93b2dc5c..ea01ee586f8b 100644 --- a/sys/contrib/openzfs/module/zfs/dsl_destroy.c +++ b/sys/contrib/openzfs/module/zfs/dsl_destroy.c @@ -133,11 +133,11 @@ process_old_cb(void *arg, const blkptr_t *bp, boolean_t bp_freed, dmu_tx_t *tx) ASSERT(!BP_IS_HOLE(bp)); - if (BP_GET_LOGICAL_BIRTH(bp) <= + if (BP_GET_BIRTH(bp) <= dsl_dataset_phys(poa->ds)->ds_prev_snap_txg) { dsl_deadlist_insert(&poa->ds->ds_deadlist, bp, bp_freed, tx); if (poa->ds_prev && !poa->after_branch_point && - BP_GET_LOGICAL_BIRTH(bp) > + BP_GET_BIRTH(bp) > dsl_dataset_phys(poa->ds_prev)->ds_prev_snap_txg) { dsl_dataset_phys(poa->ds_prev)->ds_unique_bytes += bp_get_dsize_sync(dp->dp_spa, bp); @@ -315,8 +315,7 @@ dsl_destroy_snapshot_sync_impl(dsl_dataset_t *ds, boolean_t defer, dmu_tx_t *tx) ASSERT(RRW_WRITE_HELD(&dp->dp_config_rwlock)); rrw_enter(&ds->ds_bp_rwlock, RW_READER, FTAG); - ASSERT3U(BP_GET_LOGICAL_BIRTH(&dsl_dataset_phys(ds)->ds_bp), <=, - tx->tx_txg); + ASSERT3U(BP_GET_BIRTH(&dsl_dataset_phys(ds)->ds_bp), <=, tx->tx_txg); rrw_exit(&ds->ds_bp_rwlock, FTAG); ASSERT(zfs_refcount_is_zero(&ds->ds_longholds)); @@ -351,7 +350,7 @@ dsl_destroy_snapshot_sync_impl(dsl_dataset_t *ds, boolean_t defer, dmu_tx_t *tx) dsl_dataset_deactivate_feature(ds, f, tx); } if (dsl_dataset_phys(ds)->ds_prev_snap_obj != 0) { - ASSERT3P(ds->ds_prev, ==, NULL); + ASSERT0P(ds->ds_prev); VERIFY0(dsl_dataset_hold_obj(dp, dsl_dataset_phys(ds)->ds_prev_snap_obj, FTAG, &ds_prev)); after_branch_point = @@ -466,7 +465,7 @@ dsl_destroy_snapshot_sync_impl(dsl_dataset_t *ds, boolean_t defer, dmu_tx_t *tx) &used, &comp, &uncomp); dsl_dataset_phys(ds_next)->ds_unique_bytes += used; dsl_dataset_rele(ds_nextnext, FTAG); - ASSERT3P(ds_next->ds_prev, ==, NULL); + ASSERT0P(ds_next->ds_prev); /* Collapse range in this head. */ dsl_dataset_t *hds; @@ -526,7 +525,7 @@ dsl_destroy_snapshot_sync_impl(dsl_dataset_t *ds, boolean_t defer, dmu_tx_t *tx) /* remove from snapshot namespace */ dsl_dataset_t *ds_head; - ASSERT(dsl_dataset_phys(ds)->ds_snapnames_zapobj == 0); + ASSERT0(dsl_dataset_phys(ds)->ds_snapnames_zapobj); VERIFY0(dsl_dataset_hold_obj(dp, dsl_dir_phys(ds->ds_dir)->dd_head_dataset_obj, FTAG, &ds_head)); VERIFY0(dsl_dataset_get_snapname(ds)); @@ -729,8 +728,8 @@ kill_blkptr(spa_t *spa, zilog_t *zilog, const blkptr_t *bp, */ dsl_free(ka->tx->tx_pool, ka->tx->tx_txg, bp); } else { - ASSERT(zilog == NULL); - ASSERT3U(BP_GET_LOGICAL_BIRTH(bp), >, + ASSERT0P(zilog); + ASSERT3U(BP_GET_BIRTH(bp), >, dsl_dataset_phys(ka->ds)->ds_prev_snap_txg); (void) dsl_dataset_block_kill(ka->ds, bp, tx, B_FALSE); } @@ -1020,8 +1019,7 @@ dsl_destroy_head_sync_impl(dsl_dataset_t *ds, dmu_tx_t *tx) ASSERT(ds->ds_prev == NULL || dsl_dataset_phys(ds->ds_prev)->ds_next_snap_obj != ds->ds_object); rrw_enter(&ds->ds_bp_rwlock, RW_READER, FTAG); - ASSERT3U(BP_GET_LOGICAL_BIRTH(&dsl_dataset_phys(ds)->ds_bp), <=, - tx->tx_txg); + ASSERT3U(BP_GET_BIRTH(&dsl_dataset_phys(ds)->ds_bp), <=, tx->tx_txg); rrw_exit(&ds->ds_bp_rwlock, FTAG); ASSERT(RRW_WRITE_HELD(&dp->dp_config_rwlock)); diff --git a/sys/contrib/openzfs/module/zfs/dsl_dir.c b/sys/contrib/openzfs/module/zfs/dsl_dir.c index f24cd2049533..6ce1890cfea1 100644 --- a/sys/contrib/openzfs/module/zfs/dsl_dir.c +++ b/sys/contrib/openzfs/module/zfs/dsl_dir.c @@ -151,8 +151,8 @@ dsl_dir_evict_async(void *dbu) for (t = 0; t < TXG_SIZE; t++) { ASSERT(!txg_list_member(&dp->dp_dirty_dirs, dd, t)); - ASSERT(dd->dd_tempreserved[t] == 0); - ASSERT(dd->dd_space_towrite[t] == 0); + ASSERT0(dd->dd_tempreserved[t]); + ASSERT0(dd->dd_space_towrite[t]); } if (dd->dd_parent) diff --git a/sys/contrib/openzfs/module/zfs/dsl_pool.c b/sys/contrib/openzfs/module/zfs/dsl_pool.c index f1088d87208b..f47822df8b53 100644 --- a/sys/contrib/openzfs/module/zfs/dsl_pool.c +++ b/sys/contrib/openzfs/module/zfs/dsl_pool.c @@ -522,8 +522,8 @@ dsl_pool_create(spa_t *spa, nvlist_t *zplprops __attribute__((unused)), /* create and open the free_bplist */ obj = bpobj_alloc(dp->dp_meta_objset, SPA_OLD_MAXBLOCKSIZE, tx); - VERIFY(zap_add(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT, - DMU_POOL_FREE_BPOBJ, sizeof (uint64_t), 1, &obj, tx) == 0); + VERIFY0(zap_add(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT, + DMU_POOL_FREE_BPOBJ, sizeof (uint64_t), 1, &obj, tx)); VERIFY0(bpobj_open(&dp->dp_free_bpobj, dp->dp_meta_objset, obj)); } @@ -1056,7 +1056,7 @@ upgrade_clones_cb(dsl_pool_t *dp, dsl_dataset_t *hds, void *arg) * will be wrong. */ rrw_enter(&ds->ds_bp_rwlock, RW_READER, FTAG); - ASSERT0(BP_GET_LOGICAL_BIRTH(&dsl_dataset_phys(prev)->ds_bp)); + ASSERT0(BP_GET_BIRTH(&dsl_dataset_phys(prev)->ds_bp)); rrw_exit(&ds->ds_bp_rwlock, FTAG); /* The origin doesn't get attached to itself */ @@ -1077,7 +1077,7 @@ upgrade_clones_cb(dsl_pool_t *dp, dsl_dataset_t *hds, void *arg) dsl_dataset_phys(prev)->ds_num_children++; if (dsl_dataset_phys(ds)->ds_next_snap_obj == 0) { - ASSERT(ds->ds_prev == NULL); + ASSERT0P(ds->ds_prev); VERIFY0(dsl_dataset_hold_obj(dp, dsl_dataset_phys(ds)->ds_prev_snap_obj, ds, &ds->ds_prev)); @@ -1173,7 +1173,7 @@ dsl_pool_create_origin(dsl_pool_t *dp, dmu_tx_t *tx) dsl_dataset_t *ds; ASSERT(dmu_tx_is_syncing(tx)); - ASSERT(dp->dp_origin_snap == NULL); + ASSERT0P(dp->dp_origin_snap); ASSERT(rrw_held(&dp->dp_config_rwlock, RW_WRITER)); /* create the origin dir, ds, & snap-ds */ @@ -1250,7 +1250,7 @@ dsl_pool_user_hold_create_obj(dsl_pool_t *dp, dmu_tx_t *tx) { objset_t *mos = dp->dp_meta_objset; - ASSERT(dp->dp_tmp_userrefs_obj == 0); + ASSERT0(dp->dp_tmp_userrefs_obj); ASSERT(dmu_tx_is_syncing(tx)); dp->dp_tmp_userrefs_obj = zap_create_link(mos, DMU_OT_USERREFS, diff --git a/sys/contrib/openzfs/module/zfs/dsl_prop.c b/sys/contrib/openzfs/module/zfs/dsl_prop.c index b76f22df61e2..51f624da5689 100644 --- a/sys/contrib/openzfs/module/zfs/dsl_prop.c +++ b/sys/contrib/openzfs/module/zfs/dsl_prop.c @@ -815,7 +815,7 @@ dsl_prop_set_sync_impl(dsl_dataset_t *ds, const char *propname, */ err = zap_update(mos, zapobj, recvdstr, intsz, numints, value, tx); - ASSERT(err == 0); + ASSERT0(err); break; case (ZPROP_SRC_NONE | ZPROP_SRC_LOCAL | ZPROP_SRC_RECEIVED): /* @@ -1166,7 +1166,7 @@ dsl_prop_get_all_impl(objset_t *mos, uint64_t propobj, if (nvlist_exists(nv, propname)) continue; - VERIFY(nvlist_alloc(&propval, NV_UNIQUE_NAME, KM_SLEEP) == 0); + VERIFY0(nvlist_alloc(&propval, NV_UNIQUE_NAME, KM_SLEEP)); if (za->za_integer_length == 1) { /* * String property @@ -1179,8 +1179,7 @@ dsl_prop_get_all_impl(objset_t *mos, uint64_t propobj, kmem_free(tmp, za->za_num_integers); break; } - VERIFY(nvlist_add_string(propval, ZPROP_VALUE, - tmp) == 0); + VERIFY0(nvlist_add_string(propval, ZPROP_VALUE, tmp)); kmem_free(tmp, za->za_num_integers); } else { /* @@ -1191,8 +1190,8 @@ dsl_prop_get_all_impl(objset_t *mos, uint64_t propobj, za->za_first_integer); } - VERIFY(nvlist_add_string(propval, ZPROP_SOURCE, source) == 0); - VERIFY(nvlist_add_nvlist(nv, propname, propval) == 0); + VERIFY0(nvlist_add_string(propval, ZPROP_SOURCE, source)); + VERIFY0(nvlist_add_nvlist(nv, propname, propval)); nvlist_free(propval); } zap_cursor_fini(&zc); @@ -1215,7 +1214,7 @@ dsl_prop_get_all_ds(dsl_dataset_t *ds, nvlist_t **nvp, int err = 0; char setpoint[ZFS_MAX_DATASET_NAME_LEN]; - VERIFY(nvlist_alloc(nvp, NV_UNIQUE_NAME, KM_SLEEP) == 0); + VERIFY0(nvlist_alloc(nvp, NV_UNIQUE_NAME, KM_SLEEP)); if (ds->ds_is_snapshot) flags |= DSL_PROP_GET_SNAPSHOT; @@ -1333,18 +1332,18 @@ dsl_prop_nvlist_add_uint64(nvlist_t *nv, zfs_prop_t prop, uint64_t value) uint64_t default_value; if (nvlist_lookup_nvlist(nv, propname, &propval) == 0) { - VERIFY(nvlist_add_uint64(propval, ZPROP_VALUE, value) == 0); + VERIFY0(nvlist_add_uint64(propval, ZPROP_VALUE, value)); return; } - VERIFY(nvlist_alloc(&propval, NV_UNIQUE_NAME, KM_SLEEP) == 0); - VERIFY(nvlist_add_uint64(propval, ZPROP_VALUE, value) == 0); + VERIFY0(nvlist_alloc(&propval, NV_UNIQUE_NAME, KM_SLEEP)); + VERIFY0(nvlist_add_uint64(propval, ZPROP_VALUE, value)); /* Indicate the default source if we can. */ if (dodefault(prop, 8, 1, &default_value) == 0 && value == default_value) { - VERIFY(nvlist_add_string(propval, ZPROP_SOURCE, "") == 0); + VERIFY0(nvlist_add_string(propval, ZPROP_SOURCE, "")); } - VERIFY(nvlist_add_nvlist(nv, propname, propval) == 0); + VERIFY0(nvlist_add_nvlist(nv, propname, propval)); nvlist_free(propval); } @@ -1355,13 +1354,13 @@ dsl_prop_nvlist_add_string(nvlist_t *nv, zfs_prop_t prop, const char *value) const char *propname = zfs_prop_to_name(prop); if (nvlist_lookup_nvlist(nv, propname, &propval) == 0) { - VERIFY(nvlist_add_string(propval, ZPROP_VALUE, value) == 0); + VERIFY0(nvlist_add_string(propval, ZPROP_VALUE, value)); return; } - VERIFY(nvlist_alloc(&propval, NV_UNIQUE_NAME, KM_SLEEP) == 0); - VERIFY(nvlist_add_string(propval, ZPROP_VALUE, value) == 0); - VERIFY(nvlist_add_nvlist(nv, propname, propval) == 0); + VERIFY0(nvlist_alloc(&propval, NV_UNIQUE_NAME, KM_SLEEP)); + VERIFY0(nvlist_add_string(propval, ZPROP_VALUE, value)); + VERIFY0(nvlist_add_nvlist(nv, propname, propval)); nvlist_free(propval); } diff --git a/sys/contrib/openzfs/module/zfs/dsl_scan.c b/sys/contrib/openzfs/module/zfs/dsl_scan.c index 1b2cd3e361d1..fcd50c459d07 100644 --- a/sys/contrib/openzfs/module/zfs/dsl_scan.c +++ b/sys/contrib/openzfs/module/zfs/dsl_scan.c @@ -454,7 +454,7 @@ static inline void bp2sio(const blkptr_t *bp, scan_io_t *sio, int dva_i) { sio->sio_blk_prop = bp->blk_prop; - sio->sio_phys_birth = BP_GET_PHYSICAL_BIRTH(bp); + sio->sio_phys_birth = BP_GET_RAW_PHYSICAL_BIRTH(bp); sio->sio_birth = BP_GET_LOGICAL_BIRTH(bp); sio->sio_cksum = bp->blk_cksum; sio->sio_nr_dvas = BP_GET_NDVAS(bp); @@ -1768,7 +1768,7 @@ dsl_scan_zil_block(zilog_t *zilog, const blkptr_t *bp, void *arg, ASSERT(!BP_IS_REDACTED(bp)); if (BP_IS_HOLE(bp) || - BP_GET_LOGICAL_BIRTH(bp) <= scn->scn_phys.scn_cur_min_txg) + BP_GET_BIRTH(bp) <= scn->scn_phys.scn_cur_min_txg) return (0); /* @@ -1778,13 +1778,13 @@ dsl_scan_zil_block(zilog_t *zilog, const blkptr_t *bp, void *arg, * scrub there's nothing to do to it). */ if (claim_txg == 0 && - BP_GET_LOGICAL_BIRTH(bp) >= spa_min_claim_txg(dp->dp_spa)) + BP_GET_BIRTH(bp) >= spa_min_claim_txg(dp->dp_spa)) return (0); SET_BOOKMARK(&zb, zh->zh_log.blk_cksum.zc_word[ZIL_ZC_OBJSET], ZB_ZIL_OBJECT, ZB_ZIL_LEVEL, bp->blk_cksum.zc_word[ZIL_ZC_SEQ]); - VERIFY(0 == scan_funcs[scn->scn_phys.scn_func](dp, bp, &zb)); + VERIFY0(scan_funcs[scn->scn_phys.scn_func](dp, bp, &zb)); return (0); } @@ -1804,7 +1804,7 @@ dsl_scan_zil_record(zilog_t *zilog, const lr_t *lrc, void *arg, ASSERT(!BP_IS_REDACTED(bp)); if (BP_IS_HOLE(bp) || - BP_GET_LOGICAL_BIRTH(bp) <= scn->scn_phys.scn_cur_min_txg) + BP_GET_BIRTH(bp) <= scn->scn_phys.scn_cur_min_txg) return (0); /* @@ -1812,7 +1812,7 @@ dsl_scan_zil_record(zilog_t *zilog, const lr_t *lrc, void *arg, * already txg sync'ed (but this log block contains * other records that are not synced) */ - if (claim_txg == 0 || BP_GET_LOGICAL_BIRTH(bp) < claim_txg) + if (claim_txg == 0 || BP_GET_BIRTH(bp) < claim_txg) return (0); ASSERT3U(BP_GET_LSIZE(bp), !=, 0); @@ -1820,7 +1820,7 @@ dsl_scan_zil_record(zilog_t *zilog, const lr_t *lrc, void *arg, lr->lr_foid, ZB_ZIL_LEVEL, lr->lr_offset / BP_GET_LSIZE(bp)); - VERIFY(0 == scan_funcs[scn->scn_phys.scn_func](dp, bp, &zb)); + VERIFY0(scan_funcs[scn->scn_phys.scn_func](dp, bp, &zb)); } return (0); } @@ -1952,7 +1952,7 @@ dsl_scan_prefetch(scan_prefetch_ctx_t *spc, blkptr_t *bp, zbookmark_phys_t *zb) return; if (BP_IS_HOLE(bp) || - BP_GET_LOGICAL_BIRTH(bp) <= scn->scn_phys.scn_cur_min_txg || + BP_GET_BIRTH(bp) <= scn->scn_phys.scn_cur_min_txg || (BP_GET_LEVEL(bp) == 0 && BP_GET_TYPE(bp) != DMU_OT_DNODE && BP_GET_TYPE(bp) != DMU_OT_OBJSET)) return; @@ -2223,7 +2223,7 @@ dsl_scan_recurse(dsl_scan_t *scn, dsl_dataset_t *ds, dmu_objset_type_t ostype, if (dnp != NULL && dnp->dn_bonuslen > DN_MAX_BONUS_LEN(dnp)) { scn->scn_phys.scn_errors++; - spa_log_error(spa, zb, BP_GET_LOGICAL_BIRTH(bp)); + spa_log_error(spa, zb, BP_GET_PHYSICAL_BIRTH(bp)); return (SET_ERROR(EINVAL)); } @@ -2319,7 +2319,7 @@ dsl_scan_recurse(dsl_scan_t *scn, dsl_dataset_t *ds, dmu_objset_type_t ostype, * by arc_read() for the cases above. */ scn->scn_phys.scn_errors++; - spa_log_error(spa, zb, BP_GET_LOGICAL_BIRTH(bp)); + spa_log_error(spa, zb, BP_GET_PHYSICAL_BIRTH(bp)); return (SET_ERROR(EINVAL)); } @@ -2396,7 +2396,12 @@ dsl_scan_visitbp(const blkptr_t *bp, const zbookmark_phys_t *zb, if (f != SPA_FEATURE_NONE) ASSERT(dsl_dataset_feature_is_active(ds, f)); - if (BP_GET_LOGICAL_BIRTH(bp) <= scn->scn_phys.scn_cur_min_txg) { + /* + * Recurse any blocks that were written either logically or physically + * at or after cur_min_txg. About logical birth we care for traversal, + * looking for any changes, while about physical for the actual scan. + */ + if (BP_GET_BIRTH(bp) <= scn->scn_phys.scn_cur_min_txg) { scn->scn_lt_min_this_txg++; return; } @@ -2422,7 +2427,7 @@ dsl_scan_visitbp(const blkptr_t *bp, const zbookmark_phys_t *zb, * Don't scan it now unless we need to because something * under it was modified. */ - if (BP_GET_BIRTH(bp) > scn->scn_phys.scn_cur_max_txg) { + if (BP_GET_PHYSICAL_BIRTH(bp) > scn->scn_phys.scn_cur_max_txg) { scn->scn_gt_max_this_txg++; return; } @@ -4806,7 +4811,7 @@ dsl_scan_scrub_cb(dsl_pool_t *dp, { dsl_scan_t *scn = dp->dp_scan; spa_t *spa = dp->dp_spa; - uint64_t phys_birth = BP_GET_BIRTH(bp); + uint64_t phys_birth = BP_GET_PHYSICAL_BIRTH(bp); size_t psize = BP_GET_PSIZE(bp); boolean_t needs_io = B_FALSE; int zio_flags = ZIO_FLAG_SCAN_THREAD | ZIO_FLAG_RAW | ZIO_FLAG_CANFAIL; @@ -5136,7 +5141,7 @@ dsl_scan_io_queue_vdev_xfer(vdev_t *svd, vdev_t *tvd) mutex_enter(&svd->vdev_scan_io_queue_lock); mutex_enter(&tvd->vdev_scan_io_queue_lock); - VERIFY3P(tvd->vdev_scan_io_queue, ==, NULL); + VERIFY0P(tvd->vdev_scan_io_queue); tvd->vdev_scan_io_queue = svd->vdev_scan_io_queue; svd->vdev_scan_io_queue = NULL; if (tvd->vdev_scan_io_queue != NULL) diff --git a/sys/contrib/openzfs/module/zfs/dsl_userhold.c b/sys/contrib/openzfs/module/zfs/dsl_userhold.c index 57c70e4ce3d2..f91b7a1eb69a 100644 --- a/sys/contrib/openzfs/module/zfs/dsl_userhold.c +++ b/sys/contrib/openzfs/module/zfs/dsl_userhold.c @@ -335,7 +335,7 @@ dsl_dataset_user_hold(nvlist_t *holds, minor_t cleanup_minor, nvlist_t *errlist) dduha.dduha_holds = holds; /* chkholds can have non-unique name */ - VERIFY(0 == nvlist_alloc(&dduha.dduha_chkholds, 0, KM_SLEEP)); + VERIFY0(nvlist_alloc(&dduha.dduha_chkholds, 0, KM_SLEEP)); dduha.dduha_errlist = errlist; dduha.dduha_minor = cleanup_minor; diff --git a/sys/contrib/openzfs/module/zfs/fm.c b/sys/contrib/openzfs/module/zfs/fm.c index a092817efedd..ae788b2310d8 100644 --- a/sys/contrib/openzfs/module/zfs/fm.c +++ b/sys/contrib/openzfs/module/zfs/fm.c @@ -337,7 +337,7 @@ zfs_zevent_next(zfs_zevent_t *ze, nvlist_t **event, uint64_t *event_size, } } - VERIFY(nvlist_size(ev->ev_nvl, &size, NV_ENCODE_NATIVE) == 0); + VERIFY0(nvlist_size(ev->ev_nvl, &size, NV_ENCODE_NATIVE)); if (size > *event_size) { *event_size = size; error = ENOMEM; diff --git a/sys/contrib/openzfs/module/zfs/metaslab.c b/sys/contrib/openzfs/module/zfs/metaslab.c index 43b94eba2d58..9f4399af56bd 100644 --- a/sys/contrib/openzfs/module/zfs/metaslab.c +++ b/sys/contrib/openzfs/module/zfs/metaslab.c @@ -375,13 +375,23 @@ static metaslab_stats_t metaslab_stats = { #define METASLABSTAT_BUMP(stat) \ atomic_inc_64(&metaslab_stats.stat.value.ui64); +char * +metaslab_rt_name(metaslab_group_t *mg, metaslab_t *ms, const char *name) +{ + return (kmem_asprintf("{spa=%s vdev_guid=%llu ms_id=%llu %s}", + spa_name(mg->mg_vd->vdev_spa), + (u_longlong_t)mg->mg_vd->vdev_guid, + (u_longlong_t)ms->ms_id, + name)); +} + static kstat_t *metaslab_ksp; void metaslab_stat_init(void) { - ASSERT(metaslab_alloc_trace_cache == NULL); + ASSERT0P(metaslab_alloc_trace_cache); metaslab_alloc_trace_cache = kmem_cache_create( "metaslab_alloc_trace_cache", sizeof (metaslab_alloc_trace_t), 0, NULL, NULL, NULL, NULL, NULL, 0); @@ -446,16 +456,16 @@ metaslab_class_destroy(metaslab_class_t *mc) { spa_t *spa = mc->mc_spa; - ASSERT(mc->mc_alloc == 0); - ASSERT(mc->mc_deferred == 0); - ASSERT(mc->mc_space == 0); - ASSERT(mc->mc_dspace == 0); + ASSERT0(mc->mc_alloc); + ASSERT0(mc->mc_deferred); + ASSERT0(mc->mc_space); + ASSERT0(mc->mc_dspace); for (int i = 0; i < spa->spa_alloc_count; i++) { metaslab_class_allocator_t *mca = &mc->mc_allocator[i]; avl_destroy(&mca->mca_tree); mutex_destroy(&mca->mca_lock); - ASSERT(mca->mca_rotor == NULL); + ASSERT0P(mca->mca_rotor); ASSERT0(mca->mca_reserved); } mutex_destroy(&mc->mc_lock); @@ -750,7 +760,8 @@ metaslab_class_histogram_verify(metaslab_class_t *mc) } IMPLY(mg == mg->mg_vd->vdev_log_mg, - mc == spa_embedded_log_class(mg->mg_vd->vdev_spa)); + mc == spa_embedded_log_class(mg->mg_vd->vdev_spa) || + mc == spa_special_embedded_log_class(mg->mg_vd->vdev_spa)); for (i = 0; i < ZFS_RANGE_TREE_HISTOGRAM_SIZE; i++) mc_hist[i] += mg->mg_histogram[i]; @@ -1076,8 +1087,8 @@ metaslab_group_destroy(metaslab_group_t *mg) { spa_t *spa = mg->mg_class->mc_spa; - ASSERT(mg->mg_prev == NULL); - ASSERT(mg->mg_next == NULL); + ASSERT0P(mg->mg_prev); + ASSERT0P(mg->mg_next); /* * We may have gone below zero with the activation count * either because we never activated in the first place or @@ -1107,8 +1118,8 @@ metaslab_group_activate(metaslab_group_t *mg) ASSERT3U(spa_config_held(spa, SCL_ALLOC, RW_WRITER), !=, 0); - ASSERT(mg->mg_prev == NULL); - ASSERT(mg->mg_next == NULL); + ASSERT0P(mg->mg_prev); + ASSERT0P(mg->mg_next); ASSERT(mg->mg_activation_count <= 0); if (++mg->mg_activation_count <= 0) @@ -1153,8 +1164,8 @@ metaslab_group_passivate(metaslab_group_t *mg) if (--mg->mg_activation_count != 0) { for (int i = 0; i < spa->spa_alloc_count; i++) ASSERT(mc->mc_allocator[i].mca_rotor != mg); - ASSERT(mg->mg_prev == NULL); - ASSERT(mg->mg_next == NULL); + ASSERT0P(mg->mg_prev); + ASSERT0P(mg->mg_next); ASSERT(mg->mg_activation_count < 0); return; } @@ -1183,14 +1194,16 @@ metaslab_group_passivate(metaslab_group_t *mg) if (msp != NULL) { mutex_enter(&msp->ms_lock); metaslab_passivate(msp, - metaslab_weight_from_range_tree(msp)); + metaslab_weight(msp, B_TRUE) & + ~METASLAB_ACTIVE_MASK); mutex_exit(&msp->ms_lock); } msp = mga->mga_secondary; if (msp != NULL) { mutex_enter(&msp->ms_lock); metaslab_passivate(msp, - metaslab_weight_from_range_tree(msp)); + metaslab_weight(msp, B_TRUE) & + ~METASLAB_ACTIVE_MASK); mutex_exit(&msp->ms_lock); } } @@ -1288,7 +1301,8 @@ metaslab_group_histogram_add(metaslab_group_t *mg, metaslab_t *msp) mutex_enter(&mc->mc_lock); for (int i = 0; i < SPACE_MAP_HISTOGRAM_SIZE; i++) { IMPLY(mg == mg->mg_vd->vdev_log_mg, - mc == spa_embedded_log_class(mg->mg_vd->vdev_spa)); + mc == spa_embedded_log_class(mg->mg_vd->vdev_spa) || + mc == spa_special_embedded_log_class(mg->mg_vd->vdev_spa)); mg->mg_histogram[i + ashift] += msp->ms_sm->sm_phys->smp_histogram[i]; mc->mc_histogram[i + ashift] += @@ -1316,7 +1330,8 @@ metaslab_group_histogram_remove(metaslab_group_t *mg, metaslab_t *msp) ASSERT3U(mc->mc_histogram[i + ashift], >=, msp->ms_sm->sm_phys->smp_histogram[i]); IMPLY(mg == mg->mg_vd->vdev_log_mg, - mc == spa_embedded_log_class(mg->mg_vd->vdev_spa)); + mc == spa_embedded_log_class(mg->mg_vd->vdev_spa) || + mc == spa_special_embedded_log_class(mg->mg_vd->vdev_spa)); mg->mg_histogram[i + ashift] -= msp->ms_sm->sm_phys->smp_histogram[i]; @@ -1330,7 +1345,7 @@ metaslab_group_histogram_remove(metaslab_group_t *mg, metaslab_t *msp) static void metaslab_group_add(metaslab_group_t *mg, metaslab_t *msp) { - ASSERT(msp->ms_group == NULL); + ASSERT0P(msp->ms_group); mutex_enter(&mg->mg_lock); msp->ms_group = mg; msp->ms_weight = 0; @@ -2895,30 +2910,43 @@ metaslab_init(metaslab_group_t *mg, uint64_t id, uint64_t object, zfs_range_seg_type_t type = metaslab_calculate_range_tree_type(vd, ms, &start, &shift); - ms->ms_allocatable = zfs_range_tree_create(NULL, type, NULL, start, - shift); + ms->ms_allocatable = zfs_range_tree_create_flags( + NULL, type, NULL, start, shift, + ZFS_RT_F_DYN_NAME, metaslab_rt_name(mg, ms, "ms_allocatable")); for (int t = 0; t < TXG_SIZE; t++) { - ms->ms_allocating[t] = zfs_range_tree_create(NULL, type, - NULL, start, shift); - } - ms->ms_freeing = zfs_range_tree_create(NULL, type, NULL, start, shift); - ms->ms_freed = zfs_range_tree_create(NULL, type, NULL, start, shift); + ms->ms_allocating[t] = zfs_range_tree_create_flags( + NULL, type, NULL, start, shift, + ZFS_RT_F_DYN_NAME, + metaslab_rt_name(mg, ms, "ms_allocating")); + } + ms->ms_freeing = zfs_range_tree_create_flags( + NULL, type, NULL, start, shift, + ZFS_RT_F_DYN_NAME, metaslab_rt_name(mg, ms, "ms_freeing")); + ms->ms_freed = zfs_range_tree_create_flags( + NULL, type, NULL, start, shift, + ZFS_RT_F_DYN_NAME, metaslab_rt_name(mg, ms, "ms_freed")); for (int t = 0; t < TXG_DEFER_SIZE; t++) { - ms->ms_defer[t] = zfs_range_tree_create(NULL, type, NULL, - start, shift); + ms->ms_defer[t] = zfs_range_tree_create_flags( + NULL, type, NULL, start, shift, + ZFS_RT_F_DYN_NAME, metaslab_rt_name(mg, ms, "ms_defer")); } - ms->ms_checkpointing = - zfs_range_tree_create(NULL, type, NULL, start, shift); - ms->ms_unflushed_allocs = - zfs_range_tree_create(NULL, type, NULL, start, shift); + ms->ms_checkpointing = zfs_range_tree_create_flags( + NULL, type, NULL, start, shift, + ZFS_RT_F_DYN_NAME, metaslab_rt_name(mg, ms, "ms_checkpointing")); + ms->ms_unflushed_allocs = zfs_range_tree_create_flags( + NULL, type, NULL, start, shift, + ZFS_RT_F_DYN_NAME, metaslab_rt_name(mg, ms, "ms_unflushed_allocs")); metaslab_rt_arg_t *mrap = kmem_zalloc(sizeof (*mrap), KM_SLEEP); mrap->mra_bt = &ms->ms_unflushed_frees_by_size; mrap->mra_floor_shift = metaslab_by_size_min_shift; - ms->ms_unflushed_frees = zfs_range_tree_create(&metaslab_rt_ops, - type, mrap, start, shift); + ms->ms_unflushed_frees = zfs_range_tree_create_flags( + &metaslab_rt_ops, type, mrap, start, shift, + ZFS_RT_F_DYN_NAME, metaslab_rt_name(mg, ms, "ms_unflushed_frees")); - ms->ms_trim = zfs_range_tree_create(NULL, type, NULL, start, shift); + ms->ms_trim = zfs_range_tree_create_flags( + NULL, type, NULL, start, shift, + ZFS_RT_F_DYN_NAME, metaslab_rt_name(mg, ms, "ms_trim")); metaslab_group_add(mg, ms); metaslab_set_fragmentation(ms, B_FALSE); @@ -2989,7 +3017,7 @@ metaslab_fini(metaslab_t *msp) metaslab_group_remove(mg, msp); mutex_enter(&msp->ms_lock); - VERIFY(msp->ms_group == NULL); + VERIFY0P(msp->ms_group); /* * If this metaslab hasn't been through metaslab_sync_done() yet its @@ -3892,7 +3920,10 @@ metaslab_condense(metaslab_t *msp, dmu_tx_t *tx) type = metaslab_calculate_range_tree_type(msp->ms_group->mg_vd, msp, &start, &shift); - condense_tree = zfs_range_tree_create(NULL, type, NULL, start, shift); + condense_tree = zfs_range_tree_create_flags( + NULL, type, NULL, start, shift, + ZFS_RT_F_DYN_NAME, + metaslab_rt_name(msp->ms_group, msp, "condense_tree")); for (int t = 0; t < TXG_DEFER_SIZE; t++) { zfs_range_tree_walk(msp->ms_defer[t], @@ -3949,8 +3980,10 @@ metaslab_condense(metaslab_t *msp, dmu_tx_t *tx) * followed by FREES (due to space_map_write() in metaslab_sync()) for * sync pass 1. */ - zfs_range_tree_t *tmp_tree = zfs_range_tree_create(NULL, type, NULL, - start, shift); + zfs_range_tree_t *tmp_tree = zfs_range_tree_create_flags( + NULL, type, NULL, start, shift, + ZFS_RT_F_DYN_NAME, + metaslab_rt_name(msp->ms_group, msp, "tmp_tree")); zfs_range_tree_add(tmp_tree, msp->ms_start, msp->ms_size); space_map_write(sm, tmp_tree, SM_ALLOC, SM_NO_VDEVID, tx); space_map_write(sm, msp->ms_allocatable, SM_FREE, SM_NO_VDEVID, tx); @@ -5199,29 +5232,16 @@ next: /* * We were unable to allocate from this metaslab so determine - * a new weight for this metaslab. Now that we have loaded - * the metaslab we can provide a better hint to the metaslab - * selector. - * - * For space-based metaslabs, we use the maximum block size. - * This information is only available when the metaslab - * is loaded and is more accurate than the generic free - * space weight that was calculated by metaslab_weight(). - * This information allows us to quickly compare the maximum - * available allocation in the metaslab to the allocation - * size being requested. - * - * For segment-based metaslabs, determine the new weight - * based on the highest bucket in the range tree. We - * explicitly use the loaded segment weight (i.e. the range - * tree histogram) since it contains the space that is - * currently available for allocation and is accurate - * even within a sync pass. + * a new weight for this metaslab. The weight was last + * recalculated either when we loaded it (if this is the first + * TXG it's been loaded in), or the last time a txg was synced + * out. */ uint64_t weight; if (WEIGHT_IS_SPACEBASED(msp->ms_weight)) { - weight = metaslab_largest_allocatable(msp); - WEIGHT_SET_SPACEBASED(weight); + metaslab_set_fragmentation(msp, B_TRUE); + weight = metaslab_space_weight(msp) & + ~METASLAB_ACTIVE_MASK; } else { weight = metaslab_weight_from_range_tree(msp); } @@ -5233,13 +5253,6 @@ next: * For the case where we use the metaslab that is * active for another allocator we want to make * sure that we retain the activation mask. - * - * Note that we could attempt to use something like - * metaslab_recalculate_weight_and_sort() that - * retains the activation mask here. That function - * uses metaslab_weight() to set the weight though - * which is not as accurate as the calculations - * above. */ weight |= msp->ms_weight & METASLAB_ACTIVE_MASK; metaslab_group_sort(mg, msp, weight); @@ -5590,7 +5603,21 @@ remap_blkptr_cb(uint64_t inner_offset, vdev_t *vd, uint64_t offset, vdev_indirect_births_t *vib = oldvd->vdev_indirect_births; uint64_t physical_birth = vdev_indirect_births_physbirth(vib, DVA_GET_OFFSET(&bp->blk_dva[0]), DVA_GET_ASIZE(&bp->blk_dva[0])); - BP_SET_PHYSICAL_BIRTH(bp, physical_birth); + + /* + * For rewritten blocks, use the old physical birth as the new logical + * birth (representing when the space was allocated) and the removal + * time as the new physical birth (representing when it was actually + * written). + */ + if (BP_GET_REWRITE(bp)) { + uint64_t old_physical_birth = BP_GET_PHYSICAL_BIRTH(bp); + ASSERT3U(old_physical_birth, <, physical_birth); + BP_SET_BIRTH(bp, old_physical_birth, physical_birth); + BP_SET_REWRITE(bp, 0); + } else { + BP_SET_PHYSICAL_BIRTH(bp, physical_birth); + } DVA_SET_VDEV(&bp->blk_dva[0], vd->vdev_id); DVA_SET_OFFSET(&bp->blk_dva[0], offset); @@ -5712,7 +5739,7 @@ metaslab_unalloc_dva(spa_t *spa, const dva_t *dva, uint64_t txg) ASSERT(!vd->vdev_removing); ASSERT(vdev_is_concrete(vd)); ASSERT0(vd->vdev_indirect_config.vic_mapping_object); - ASSERT3P(vd->vdev_indirect_mapping, ==, NULL); + ASSERT0P(vd->vdev_indirect_mapping); if (DVA_GET_GANG(dva)) size = vdev_gang_header_asize(vd); @@ -5757,21 +5784,21 @@ metaslab_free_dva(spa_t *spa, const dva_t *dva, boolean_t checkpoint) } /* - * Reserve some allocation slots. The reservation system must be called - * before we call into the allocator. If there aren't any available slots - * then the I/O will be throttled until an I/O completes and its slots are - * freed up. The function returns true if it was successful in placing - * the reservation. + * Reserve some space for a future allocation. The reservation system must be + * called before we call into the allocator. If there aren't enough space + * available, the calling I/O will be throttled until another I/O completes and + * its reservation is released. The function returns true if it was successful + * in placing the reservation. */ boolean_t -metaslab_class_throttle_reserve(metaslab_class_t *mc, int slots, zio_t *zio, - boolean_t must, boolean_t *more) +metaslab_class_throttle_reserve(metaslab_class_t *mc, int allocator, + int copies, uint64_t io_size, boolean_t must, boolean_t *more) { - metaslab_class_allocator_t *mca = &mc->mc_allocator[zio->io_allocator]; + metaslab_class_allocator_t *mca = &mc->mc_allocator[allocator]; ASSERT(mc->mc_alloc_throttle_enabled); - if (mc->mc_alloc_io_size < zio->io_size) { - mc->mc_alloc_io_size = zio->io_size; + if (mc->mc_alloc_io_size < io_size) { + mc->mc_alloc_io_size = io_size; metaslab_class_balance(mc, B_FALSE); } if (must || mca->mca_reserved <= mc->mc_alloc_max) { @@ -5782,10 +5809,9 @@ metaslab_class_throttle_reserve(metaslab_class_t *mc, int slots, zio_t *zio, * worst that can happen is few more I/Os get to allocation * earlier, that is not a problem. */ - int64_t delta = slots * zio->io_size; + int64_t delta = copies * io_size; *more = (atomic_add_64_nv(&mca->mca_reserved, delta) <= mc->mc_alloc_max); - zio->io_flags |= ZIO_FLAG_IO_ALLOCATING; return (B_TRUE); } *more = B_FALSE; @@ -5793,13 +5819,13 @@ metaslab_class_throttle_reserve(metaslab_class_t *mc, int slots, zio_t *zio, } boolean_t -metaslab_class_throttle_unreserve(metaslab_class_t *mc, int slots, - zio_t *zio) +metaslab_class_throttle_unreserve(metaslab_class_t *mc, int allocator, + int copies, uint64_t io_size) { - metaslab_class_allocator_t *mca = &mc->mc_allocator[zio->io_allocator]; + metaslab_class_allocator_t *mca = &mc->mc_allocator[allocator]; ASSERT(mc->mc_alloc_throttle_enabled); - int64_t delta = slots * zio->io_size; + int64_t delta = copies * io_size; return (atomic_add_64_nv(&mca->mca_reserved, -delta) <= mc->mc_alloc_max); } @@ -5960,7 +5986,7 @@ metaslab_alloc_range(spa_t *spa, metaslab_class_t *mc, uint64_t psize, int error = 0; ASSERT0(BP_GET_LOGICAL_BIRTH(bp)); - ASSERT0(BP_GET_PHYSICAL_BIRTH(bp)); + ASSERT0(BP_GET_RAW_PHYSICAL_BIRTH(bp)); spa_config_enter(spa, SCL_ALLOC, FTAG, RW_READER); @@ -5971,16 +5997,16 @@ metaslab_alloc_range(spa_t *spa, metaslab_class_t *mc, uint64_t psize, } ASSERT(ndvas > 0 && ndvas <= spa_max_replication(spa)); - ASSERT(BP_GET_NDVAS(bp) == 0); + ASSERT0(BP_GET_NDVAS(bp)); ASSERT(hintbp == NULL || ndvas <= BP_GET_NDVAS(hintbp)); ASSERT3P(zal, !=, NULL); - uint64_t cur_psize = 0; - + uint64_t smallest_psize = UINT64_MAX; for (int d = 0; d < ndvas; d++) { - error = metaslab_alloc_dva_range(spa, mc, psize, max_psize, - dva, d, hintdva, txg, flags, zal, allocator, - actual_psize ? &cur_psize : NULL); + uint64_t cur_psize = 0; + error = metaslab_alloc_dva_range(spa, mc, psize, + MIN(smallest_psize, max_psize), dva, d, hintdva, txg, + flags, zal, allocator, actual_psize ? &cur_psize : NULL); if (error != 0) { for (d--; d >= 0; d--) { metaslab_unalloc_dva(spa, &dva[d], txg); @@ -6000,13 +6026,13 @@ metaslab_alloc_range(spa_t *spa, metaslab_class_t *mc, uint64_t psize, DVA_GET_VDEV(&dva[d]), allocator, flags, psize, tag); if (actual_psize) - max_psize = MIN(cur_psize, max_psize); + smallest_psize = MIN(cur_psize, smallest_psize); } } - ASSERT(error == 0); + ASSERT0(error); ASSERT(BP_GET_NDVAS(bp) == ndvas); if (actual_psize) - *actual_psize = max_psize; + *actual_psize = smallest_psize; spa_config_exit(spa, SCL_ALLOC, FTAG); @@ -6022,7 +6048,7 @@ metaslab_free(spa_t *spa, const blkptr_t *bp, uint64_t txg, boolean_t now) int ndvas = BP_GET_NDVAS(bp); ASSERT(!BP_IS_HOLE(bp)); - ASSERT(!now || BP_GET_LOGICAL_BIRTH(bp) >= spa_syncing_txg(spa)); + ASSERT(!now || BP_GET_BIRTH(bp) >= spa_syncing_txg(spa)); /* * If we have a checkpoint for the pool we need to make sure that @@ -6040,7 +6066,7 @@ metaslab_free(spa_t *spa, const blkptr_t *bp, uint64_t txg, boolean_t now) * normally as they will be referenced by the checkpointed uberblock. */ boolean_t checkpoint = B_FALSE; - if (BP_GET_LOGICAL_BIRTH(bp) <= spa->spa_checkpoint_txg && + if (BP_GET_BIRTH(bp) <= spa->spa_checkpoint_txg && spa_syncing_txg(spa) > spa->spa_checkpoint_txg) { /* * At this point, if the block is part of the checkpoint diff --git a/sys/contrib/openzfs/module/zfs/mmp.c b/sys/contrib/openzfs/module/zfs/mmp.c index f3665d29b8b4..7db72b9b04b0 100644 --- a/sys/contrib/openzfs/module/zfs/mmp.c +++ b/sys/contrib/openzfs/module/zfs/mmp.c @@ -260,7 +260,7 @@ mmp_thread_stop(spa_t *spa) zfs_dbgmsg("MMP thread stopped pool '%s' gethrtime %llu", spa_name(spa), gethrtime()); - ASSERT(mmp->mmp_thread == NULL); + ASSERT0P(mmp->mmp_thread); mmp->mmp_thread_exiting = 0; } diff --git a/sys/contrib/openzfs/module/zfs/range_tree.c b/sys/contrib/openzfs/module/zfs/range_tree.c index 373636c69254..ea2d2c7227c8 100644 --- a/sys/contrib/openzfs/module/zfs/range_tree.c +++ b/sys/contrib/openzfs/module/zfs/range_tree.c @@ -201,10 +201,10 @@ ZFS_BTREE_FIND_IN_BUF_FUNC(zfs_range_tree_seg64_find_in_buf, zfs_range_seg64_t, ZFS_BTREE_FIND_IN_BUF_FUNC(zfs_range_tree_seg_gap_find_in_buf, zfs_range_seg_gap_t, zfs_range_tree_seg_gap_compare) -zfs_range_tree_t * -zfs_range_tree_create_gap(const zfs_range_tree_ops_t *ops, +static zfs_range_tree_t * +zfs_range_tree_create_impl(const zfs_range_tree_ops_t *ops, zfs_range_seg_type_t type, void *arg, uint64_t start, uint64_t shift, - uint64_t gap) + uint64_t gap, uint64_t flags, const char *name) { zfs_range_tree_t *rt = kmem_zalloc(sizeof (zfs_range_tree_t), KM_SLEEP); @@ -236,6 +236,8 @@ zfs_range_tree_create_gap(const zfs_range_tree_ops_t *ops, rt->rt_ops = ops; rt->rt_gap = gap; + rt->rt_flags = flags; + rt->rt_name = name; rt->rt_arg = arg; rt->rt_type = type; rt->rt_start = start; @@ -248,10 +250,29 @@ zfs_range_tree_create_gap(const zfs_range_tree_ops_t *ops, } zfs_range_tree_t * +zfs_range_tree_create_gap(const zfs_range_tree_ops_t *ops, + zfs_range_seg_type_t type, void *arg, uint64_t start, uint64_t shift, + uint64_t gap) +{ + return (zfs_range_tree_create_impl(ops, type, arg, start, shift, gap, + 0, NULL)); +} + +zfs_range_tree_t * zfs_range_tree_create(const zfs_range_tree_ops_t *ops, zfs_range_seg_type_t type, void *arg, uint64_t start, uint64_t shift) { - return (zfs_range_tree_create_gap(ops, type, arg, start, shift, 0)); + return (zfs_range_tree_create_impl(ops, type, arg, start, shift, 0, + 0, NULL)); +} + +zfs_range_tree_t * +zfs_range_tree_create_flags(const zfs_range_tree_ops_t *ops, + zfs_range_seg_type_t type, void *arg, uint64_t start, uint64_t shift, + uint64_t flags, const char *name) +{ + return (zfs_range_tree_create_impl(ops, type, arg, start, shift, 0, + flags, name)); } void @@ -262,6 +283,9 @@ zfs_range_tree_destroy(zfs_range_tree_t *rt) if (rt->rt_ops != NULL && rt->rt_ops->rtop_destroy != NULL) rt->rt_ops->rtop_destroy(rt, rt->rt_arg); + if (rt->rt_name != NULL && (rt->rt_flags & ZFS_RT_F_DYN_NAME)) + kmem_strfree((char *)(uintptr_t)rt->rt_name); + zfs_btree_destroy(&rt->rt_root); kmem_free(rt, sizeof (*rt)); } @@ -271,15 +295,17 @@ zfs_range_tree_adjust_fill(zfs_range_tree_t *rt, zfs_range_seg_t *rs, int64_t delta) { if (delta < 0 && delta * -1 >= zfs_rs_get_fill(rs, rt)) { - zfs_panic_recover("zfs: attempting to decrease fill to or " - "below 0; probable double remove in segment [%llx:%llx]", + zfs_panic_recover("zfs: rt=%s: attempting to decrease fill to " + "or below 0; probable double remove in segment [%llx:%llx]", + ZFS_RT_NAME(rt), (longlong_t)zfs_rs_get_start(rs, rt), (longlong_t)zfs_rs_get_end(rs, rt)); } if (zfs_rs_get_fill(rs, rt) + delta > zfs_rs_get_end(rs, rt) - zfs_rs_get_start(rs, rt)) { - zfs_panic_recover("zfs: attempting to increase fill beyond " - "max; probable double add in segment [%llx:%llx]", + zfs_panic_recover("zfs: rt=%s: attempting to increase fill " + "beyond max; probable double add in segment [%llx:%llx]", + ZFS_RT_NAME(rt), (longlong_t)zfs_rs_get_start(rs, rt), (longlong_t)zfs_rs_get_end(rs, rt)); } @@ -319,14 +345,17 @@ zfs_range_tree_add_impl(void *arg, uint64_t start, uint64_t size, uint64_t fill) * the normal code paths. */ if (rs != NULL) { + uint64_t rstart = zfs_rs_get_start(rs, rt); + uint64_t rend = zfs_rs_get_end(rs, rt); if (gap == 0) { - zfs_panic_recover("zfs: adding existent segment to " - "range tree (offset=%llx size=%llx)", - (longlong_t)start, (longlong_t)size); + zfs_panic_recover("zfs: rt=%s: adding segment " + "(offset=%llx size=%llx) overlapping with existing " + "one (offset=%llx size=%llx)", + ZFS_RT_NAME(rt), + (longlong_t)start, (longlong_t)size, + (longlong_t)rstart, (longlong_t)(rend - rstart)); return; } - uint64_t rstart = zfs_rs_get_start(rs, rt); - uint64_t rend = zfs_rs_get_end(rs, rt); if (rstart <= start && rend >= end) { zfs_range_tree_adjust_fill(rt, rs, fill); return; @@ -348,7 +377,7 @@ zfs_range_tree_add_impl(void *arg, uint64_t start, uint64_t size, uint64_t fill) return; } - ASSERT3P(rs, ==, NULL); + ASSERT0P(rs); /* * Determine whether or not we will have to merge with our neighbors. @@ -451,6 +480,7 @@ zfs_range_tree_remove_impl(zfs_range_tree_t *rt, uint64_t start, uint64_t size, zfs_range_seg_t *rs; zfs_range_seg_max_t rsearch, rs_tmp; uint64_t end = start + size; + uint64_t rstart, rend; boolean_t left_over, right_over; VERIFY3U(size, !=, 0); @@ -464,12 +494,15 @@ zfs_range_tree_remove_impl(zfs_range_tree_t *rt, uint64_t start, uint64_t size, /* Make sure we completely overlap with someone */ if (rs == NULL) { - zfs_panic_recover("zfs: removing nonexistent segment from " - "range tree (offset=%llx size=%llx)", - (longlong_t)start, (longlong_t)size); + zfs_panic_recover("zfs: rt=%s: removing nonexistent segment " + "from range tree (offset=%llx size=%llx)", + ZFS_RT_NAME(rt), (longlong_t)start, (longlong_t)size); return; } + rstart = zfs_rs_get_start(rs, rt); + rend = zfs_rs_get_end(rs, rt); + /* * Range trees with gap support must only remove complete segments * from the tree. This allows us to maintain accurate fill accounting @@ -479,31 +512,36 @@ zfs_range_tree_remove_impl(zfs_range_tree_t *rt, uint64_t start, uint64_t size, if (rt->rt_gap != 0) { if (do_fill) { if (zfs_rs_get_fill(rs, rt) == size) { - start = zfs_rs_get_start(rs, rt); - end = zfs_rs_get_end(rs, rt); + start = rstart; + end = rend; size = end - start; } else { zfs_range_tree_adjust_fill(rt, rs, -size); return; } - } else if (zfs_rs_get_start(rs, rt) != start || - zfs_rs_get_end(rs, rt) != end) { - zfs_panic_recover("zfs: freeing partial segment of " - "gap tree (offset=%llx size=%llx) of " + } else if (rstart != start || rend != end) { + zfs_panic_recover("zfs: rt=%s: freeing partial segment " + "of gap tree (offset=%llx size=%llx) of " "(offset=%llx size=%llx)", + ZFS_RT_NAME(rt), (longlong_t)start, (longlong_t)size, - (longlong_t)zfs_rs_get_start(rs, rt), - (longlong_t)zfs_rs_get_end(rs, rt) - - zfs_rs_get_start(rs, rt)); + (longlong_t)rstart, (longlong_t)(rend - rstart)); return; } } - VERIFY3U(zfs_rs_get_start(rs, rt), <=, start); - VERIFY3U(zfs_rs_get_end(rs, rt), >=, end); + if (!(rstart <= start && rend >= end)) { + panic("zfs: rt=%s: removing segment " + "(offset=%llx size=%llx) not completely overlapped by " + "existing one (offset=%llx size=%llx)", + ZFS_RT_NAME(rt), + (longlong_t)start, (longlong_t)size, + (longlong_t)rstart, (longlong_t)(rend - rstart)); + return; + } - left_over = (zfs_rs_get_start(rs, rt) != start); - right_over = (zfs_rs_get_end(rs, rt) != end); + left_over = (rstart != start); + right_over = (rend != end); zfs_range_tree_stat_decr(rt, rs); @@ -829,7 +867,7 @@ zfs_range_tree_remove_xor_add_segment(uint64_t start, uint64_t end, next = zfs_btree_next(&removefrom->rt_root, &where, &where); } - VERIFY3P(curr, ==, NULL); + VERIFY0P(curr); if (start != end) { VERIFY3U(start, <, end); diff --git a/sys/contrib/openzfs/module/zfs/rrwlock.c b/sys/contrib/openzfs/module/zfs/rrwlock.c index 8ee784619839..d0df39b93560 100644 --- a/sys/contrib/openzfs/module/zfs/rrwlock.c +++ b/sys/contrib/openzfs/module/zfs/rrwlock.c @@ -108,7 +108,7 @@ rrn_add(rrwlock_t *rrl, const void *tag) rn->rn_rrl = rrl; rn->rn_next = tsd_get(rrw_tsd_key); rn->rn_tag = tag; - VERIFY(tsd_set(rrw_tsd_key, rn) == 0); + VERIFY0(tsd_set(rrw_tsd_key, rn)); } /* @@ -129,7 +129,7 @@ rrn_find_and_remove(rrwlock_t *rrl, const void *tag) if (prev) prev->rn_next = rn->rn_next; else - VERIFY(tsd_set(rrw_tsd_key, rn->rn_next) == 0); + VERIFY0(tsd_set(rrw_tsd_key, rn->rn_next)); kmem_free(rn, sizeof (*rn)); return (B_TRUE); } @@ -155,7 +155,7 @@ rrw_destroy(rrwlock_t *rrl) { mutex_destroy(&rrl->rr_lock); cv_destroy(&rrl->rr_cv); - ASSERT(rrl->rr_writer == NULL); + ASSERT0P(rrl->rr_writer); zfs_refcount_destroy(&rrl->rr_anon_rcount); zfs_refcount_destroy(&rrl->rr_linked_rcount); } @@ -188,7 +188,7 @@ rrw_enter_read_impl(rrwlock_t *rrl, boolean_t prio, const void *tag) } else { (void) zfs_refcount_add(&rrl->rr_anon_rcount, tag); } - ASSERT(rrl->rr_writer == NULL); + ASSERT0P(rrl->rr_writer); mutex_exit(&rrl->rr_lock); } diff --git a/sys/contrib/openzfs/module/zfs/sa.c b/sys/contrib/openzfs/module/zfs/sa.c index 5db470ce6242..7ad25d4d85ba 100644 --- a/sys/contrib/openzfs/module/zfs/sa.c +++ b/sys/contrib/openzfs/module/zfs/sa.c @@ -304,7 +304,7 @@ sa_get_spill(sa_handle_t *hdl) if (hdl->sa_spill == NULL) { if ((rc = dmu_spill_hold_existing(hdl->sa_bonus, NULL, &hdl->sa_spill)) == 0) - VERIFY(0 == sa_build_index(hdl, SA_SPILL)); + VERIFY0(sa_build_index(hdl, SA_SPILL)); } else { rc = 0; } @@ -432,7 +432,7 @@ sa_add_layout_entry(objset_t *os, const sa_attr_type_t *attrs, int attr_count, (void) snprintf(attr_name, sizeof (attr_name), "%d", (int)lot_num); - VERIFY(0 == zap_update(os, os->os_sa->sa_layout_attr_obj, + VERIFY0(zap_update(os, os->os_sa->sa_layout_attr_obj, attr_name, 2, attr_count, attrs, tx)); } @@ -505,7 +505,7 @@ sa_resize_spill(sa_handle_t *hdl, uint32_t size, dmu_tx_t *tx) } error = dbuf_spill_set_blksz(hdl->sa_spill, blocksize, tx); - ASSERT(error == 0); + ASSERT0(error); return (error); } @@ -717,7 +717,7 @@ sa_build_layouts(sa_handle_t *hdl, sa_bulk_attr_t *attr_desc, int attr_count, if (BUF_SPACE_NEEDED(spill_used, spillhdrsize) > hdl->sa_spill->db_size) - VERIFY(0 == sa_resize_spill(hdl, + VERIFY0(sa_resize_spill(hdl, BUF_SPACE_NEEDED(spill_used, spillhdrsize), tx)); } @@ -791,7 +791,7 @@ sa_build_layouts(sa_handle_t *hdl, sa_bulk_attr_t *attr_desc, int attr_count, hdl->sa_bonus_tab = NULL; } if (!sa->sa_force_spill) - VERIFY(0 == sa_build_index(hdl, SA_BONUS)); + VERIFY0(sa_build_index(hdl, SA_BONUS)); if (hdl->sa_spill) { sa_idx_tab_rele(hdl->sa_os, hdl->sa_spill_tab); if (!spilling) { @@ -801,10 +801,10 @@ sa_build_layouts(sa_handle_t *hdl, sa_bulk_attr_t *attr_desc, int attr_count, dmu_buf_rele(hdl->sa_spill, NULL); hdl->sa_spill = NULL; hdl->sa_spill_tab = NULL; - VERIFY(0 == dmu_rm_spill(hdl->sa_os, + VERIFY0(dmu_rm_spill(hdl->sa_os, sa_handle_object(hdl), tx)); } else { - VERIFY(0 == sa_build_index(hdl, SA_SPILL)); + VERIFY0(sa_build_index(hdl, SA_SPILL)); } } @@ -1733,10 +1733,10 @@ sa_add_projid(sa_handle_t *hdl, dmu_tx_t *tx, uint64_t projid) NULL, dxattr_obj, dxattr_size); } - VERIFY(dmu_set_bonustype(db, DMU_OT_SA, tx) == 0); - VERIFY(sa_replace_all_by_template_locked(hdl, attrs, count, tx) == 0); + VERIFY0(dmu_set_bonustype(db, DMU_OT_SA, tx)); + VERIFY0(sa_replace_all_by_template_locked(hdl, attrs, count, tx)); if (znode_acl.z_acl_extern_obj) { - VERIFY(0 == dmu_object_free(zfsvfs->z_os, + VERIFY0(dmu_object_free(zfsvfs->z_os, znode_acl.z_acl_extern_obj, tx)); } @@ -1858,7 +1858,7 @@ sa_attr_register_sync(sa_handle_t *hdl, dmu_tx_t *tx) continue; ATTR_ENCODE(attr_value, tb[i].sa_attr, tb[i].sa_length, tb[i].sa_byteswap); - VERIFY(0 == zap_update(hdl->sa_os, sa->sa_reg_attr_obj, + VERIFY0(zap_update(hdl->sa_os, sa->sa_reg_attr_obj, tb[i].sa_name, 8, 1, &attr_value, tx)); tb[i].sa_registered = B_TRUE; } @@ -2013,7 +2013,7 @@ sa_modify_attrs(sa_handle_t *hdl, sa_attr_type_t newattr, * Only a variable-sized attribute can be * replaced here, and its size must be changing. */ - ASSERT3U(reg_length, ==, 0); + ASSERT0(reg_length); ASSERT3U(length, !=, buflen); SA_ADD_BULK_ATTR(attr_desc, j, attr, locator, datastart, buflen); diff --git a/sys/contrib/openzfs/module/zfs/spa.c b/sys/contrib/openzfs/module/zfs/spa.c index 6b52c6cb1f9e..b3bb46da263b 100644 --- a/sys/contrib/openzfs/module/zfs/spa.c +++ b/sys/contrib/openzfs/module/zfs/spa.c @@ -100,6 +100,7 @@ #include <sys/vmsystm.h> #endif /* _KERNEL */ +#include "zfs_crrd.h" #include "zfs_prop.h" #include "zfs_comutil.h" #include <cityhash.h> @@ -311,6 +312,41 @@ static int zfs_livelist_condense_zthr_cancel = 0; static int zfs_livelist_condense_new_alloc = 0; /* + * Time variable to decide how often the txg should be added into the + * database (in seconds). + * The smallest available resolution is in minutes, which means an update occurs + * each time we reach `spa_note_txg_time` and the txg has changed. We provide + * a 256-slot ring buffer for minute-level resolution. The number is limited by + * the size of the structure we use and the maximum amount of bytes we can write + * into ZAP. Setting `spa_note_txg_time` to 10 minutes results in approximately + * 144 records per day. Given the 256 slots, this provides roughly 1.5 days of + * high-resolution data. + * + * The user can decrease `spa_note_txg_time` to increase resolution within + * a day, at the cost of retaining fewer days of data. Alternatively, increasing + * the interval allows storing data over a longer period, but with lower + * frequency. + * + * This parameter does not affect the daily or monthly databases, as those only + * store one record per day and per month, respectively. + */ +static uint_t spa_note_txg_time = 10 * 60; + +/* + * How often flush txg database to a disk (in seconds). + * We flush data every time we write to it, making it the most reliable option. + * Since this happens every 10 minutes, it shouldn't introduce any noticeable + * overhead for the system. In case of failure, we will always have an + * up-to-date version of the database. + * + * The user can adjust the flush interval to a lower value, but it probably + * doesn't make sense to flush more often than the database is updated. + * The user can also increase the interval if they're concerned about the + * performance of writing the entire database to disk. + */ +static uint_t spa_flush_txg_time = 10 * 60; + +/* * ========================================================================== * SPA properties routines * ========================================================================== @@ -390,10 +426,10 @@ spa_prop_add_user(nvlist_t *nvl, const char *propname, char *strval, { nvlist_t *propval; - VERIFY(nvlist_alloc(&propval, NV_UNIQUE_NAME, KM_SLEEP) == 0); - VERIFY(nvlist_add_uint64(propval, ZPROP_SOURCE, src) == 0); - VERIFY(nvlist_add_string(propval, ZPROP_VALUE, strval) == 0); - VERIFY(nvlist_add_nvlist(nvl, propname, propval) == 0); + VERIFY0(nvlist_alloc(&propval, NV_UNIQUE_NAME, KM_SLEEP)); + VERIFY0(nvlist_add_uint64(propval, ZPROP_SOURCE, src)); + VERIFY0(nvlist_add_string(propval, ZPROP_VALUE, strval)); + VERIFY0(nvlist_add_nvlist(nvl, propname, propval)); nvlist_free(propval); } @@ -417,11 +453,15 @@ spa_prop_get_config(spa_t *spa, nvlist_t *nv) alloc += metaslab_class_get_alloc(spa_special_class(spa)); alloc += metaslab_class_get_alloc(spa_dedup_class(spa)); alloc += metaslab_class_get_alloc(spa_embedded_log_class(spa)); + alloc += metaslab_class_get_alloc( + spa_special_embedded_log_class(spa)); size = metaslab_class_get_space(mc); size += metaslab_class_get_space(spa_special_class(spa)); size += metaslab_class_get_space(spa_dedup_class(spa)); size += metaslab_class_get_space(spa_embedded_log_class(spa)); + size += metaslab_class_get_space( + spa_special_embedded_log_class(spa)); spa_prop_add_list(nv, ZPOOL_PROP_NAME, spa_name(spa), 0, src); spa_prop_add_list(nv, ZPOOL_PROP_SIZE, NULL, size, src); @@ -925,7 +965,7 @@ spa_prop_set(spa_t *spa, nvlist_t *nvp) uint64_t ver = 0; if (prop == ZPOOL_PROP_VERSION) { - VERIFY(nvpair_value_uint64(elem, &ver) == 0); + VERIFY0(nvpair_value_uint64(elem, &ver)); } else { ASSERT(zpool_prop_feature(nvpair_name(elem))); ver = SPA_VERSION_FEATURES; @@ -1255,7 +1295,7 @@ spa_taskqs_fini(spa_t *spa, zio_type_t t, zio_taskq_type_t q) spa_taskqs_t *tqs = &spa->spa_zio_taskq[t][q]; if (tqs->stqs_taskq == NULL) { - ASSERT3U(tqs->stqs_count, ==, 0); + ASSERT0(tqs->stqs_count); return; } @@ -1679,6 +1719,8 @@ spa_activate(spa_t *spa, spa_mode_t mode) "embedded_log", msp, B_TRUE); spa->spa_special_class = metaslab_class_create(spa, "special", msp, B_FALSE); + spa->spa_special_embedded_log_class = metaslab_class_create(spa, + "special_embedded_log", msp, B_TRUE); spa->spa_dedup_class = metaslab_class_create(spa, "dedup", msp, B_FALSE); @@ -1794,9 +1836,9 @@ static void spa_deactivate(spa_t *spa) { ASSERT(spa->spa_sync_on == B_FALSE); - ASSERT(spa->spa_dsl_pool == NULL); - ASSERT(spa->spa_root_vdev == NULL); - ASSERT(spa->spa_async_zio_root == NULL); + ASSERT0P(spa->spa_dsl_pool); + ASSERT0P(spa->spa_root_vdev); + ASSERT0P(spa->spa_async_zio_root); ASSERT(spa->spa_state != POOL_STATE_UNINITIALIZED); spa_evicting_os_wait(spa); @@ -1853,6 +1895,9 @@ spa_deactivate(spa_t *spa) metaslab_class_destroy(spa->spa_special_class); spa->spa_special_class = NULL; + metaslab_class_destroy(spa->spa_special_embedded_log_class); + spa->spa_special_embedded_log_class = NULL; + metaslab_class_destroy(spa->spa_dedup_class); spa->spa_dedup_class = NULL; @@ -1976,7 +2021,7 @@ spa_unload_log_sm_flush_all(spa_t *spa) dmu_tx_t *tx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir); VERIFY0(dmu_tx_assign(tx, DMU_TX_WAIT | DMU_TX_SUSPEND)); - ASSERT3U(spa->spa_log_flushall_txg, ==, 0); + ASSERT0(spa->spa_log_flushall_txg); spa->spa_log_flushall_txg = dmu_tx_get_txg(tx); dmu_tx_commit(tx); @@ -2031,6 +2076,111 @@ spa_destroy_aux_threads(spa_t *spa) } } +static void +spa_sync_time_logger(spa_t *spa, uint64_t txg) +{ + uint64_t curtime; + dmu_tx_t *tx; + + if (!spa_writeable(spa)) { + return; + } + curtime = gethrestime_sec(); + if (curtime < spa->spa_last_noted_txg_time + spa_note_txg_time) { + return; + } + + if (txg > spa->spa_last_noted_txg) { + spa->spa_last_noted_txg_time = curtime; + spa->spa_last_noted_txg = txg; + + mutex_enter(&spa->spa_txg_log_time_lock); + dbrrd_add(&spa->spa_txg_log_time, curtime, txg); + mutex_exit(&spa->spa_txg_log_time_lock); + } + + if (curtime < spa->spa_last_flush_txg_time + spa_flush_txg_time) { + return; + } + spa->spa_last_flush_txg_time = curtime; + + tx = dmu_tx_create_assigned(spa_get_dsl(spa), txg); + + VERIFY0(zap_update(spa_meta_objset(spa), DMU_POOL_DIRECTORY_OBJECT, + DMU_POOL_TXG_LOG_TIME_MINUTES, RRD_ENTRY_SIZE, RRD_STRUCT_ELEM, + &spa->spa_txg_log_time.dbr_minutes, tx)); + VERIFY0(zap_update(spa_meta_objset(spa), DMU_POOL_DIRECTORY_OBJECT, + DMU_POOL_TXG_LOG_TIME_DAYS, RRD_ENTRY_SIZE, RRD_STRUCT_ELEM, + &spa->spa_txg_log_time.dbr_days, tx)); + VERIFY0(zap_update(spa_meta_objset(spa), DMU_POOL_DIRECTORY_OBJECT, + DMU_POOL_TXG_LOG_TIME_MONTHS, RRD_ENTRY_SIZE, RRD_STRUCT_ELEM, + &spa->spa_txg_log_time.dbr_months, tx)); + dmu_tx_commit(tx); +} + +static void +spa_unload_sync_time_logger(spa_t *spa) +{ + uint64_t txg; + dmu_tx_t *tx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir); + VERIFY0(dmu_tx_assign(tx, DMU_TX_WAIT)); + + txg = dmu_tx_get_txg(tx); + spa->spa_last_noted_txg_time = 0; + spa->spa_last_flush_txg_time = 0; + spa_sync_time_logger(spa, txg); + + dmu_tx_commit(tx); +} + +static void +spa_load_txg_log_time(spa_t *spa) +{ + int error; + + error = zap_lookup(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT, + DMU_POOL_TXG_LOG_TIME_MINUTES, RRD_ENTRY_SIZE, RRD_STRUCT_ELEM, + &spa->spa_txg_log_time.dbr_minutes); + if (error != 0 && error != ENOENT) { + spa_load_note(spa, "unable to load a txg time database with " + "minute resolution [error=%d]", error); + } + error = zap_lookup(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT, + DMU_POOL_TXG_LOG_TIME_DAYS, RRD_ENTRY_SIZE, RRD_STRUCT_ELEM, + &spa->spa_txg_log_time.dbr_days); + if (error != 0 && error != ENOENT) { + spa_load_note(spa, "unable to load a txg time database with " + "day resolution [error=%d]", error); + } + error = zap_lookup(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT, + DMU_POOL_TXG_LOG_TIME_MONTHS, RRD_ENTRY_SIZE, RRD_STRUCT_ELEM, + &spa->spa_txg_log_time.dbr_months); + if (error != 0 && error != ENOENT) { + spa_load_note(spa, "unable to load a txg time database with " + "month resolution [error=%d]", error); + } +} + +static boolean_t +spa_should_sync_time_logger_on_unload(spa_t *spa) +{ + + if (!spa_writeable(spa)) + return (B_FALSE); + + if (!spa->spa_sync_on) + return (B_FALSE); + + if (spa_state(spa) != POOL_STATE_EXPORTED) + return (B_FALSE); + + if (spa->spa_last_noted_txg == 0) + return (B_FALSE); + + return (B_TRUE); +} + + /* * Opposite of spa_load(). */ @@ -2052,6 +2202,9 @@ spa_unload(spa_t *spa) * we delay the final TXGs beyond what spa_final_txg is set at. */ if (spa->spa_final_txg == UINT64_MAX) { + if (spa_should_sync_time_logger_on_unload(spa)) + spa_unload_sync_time_logger(spa); + /* * If the log space map feature is enabled and the pool is * getting exported (but not destroyed), we want to spend some @@ -2127,7 +2280,7 @@ spa_unload(spa_t *spa) */ if (spa->spa_root_vdev) vdev_free(spa->spa_root_vdev); - ASSERT(spa->spa_root_vdev == NULL); + ASSERT0P(spa->spa_root_vdev); /* * Close the dsl pool. @@ -2265,8 +2418,8 @@ spa_load_spares(spa_t *spa) spa->spa_spares.sav_vdevs = kmem_zalloc(nspares * sizeof (void *), KM_SLEEP); for (i = 0; i < spa->spa_spares.sav_count; i++) { - VERIFY(spa_config_parse(spa, &vd, spares[i], NULL, 0, - VDEV_ALLOC_SPARE) == 0); + VERIFY0(spa_config_parse(spa, &vd, spares[i], NULL, 0, + VDEV_ALLOC_SPARE)); ASSERT(vd != NULL); spa->spa_spares.sav_vdevs[i] = vd; @@ -2393,8 +2546,8 @@ spa_load_l2cache(spa_t *spa) /* * Create new vdev */ - VERIFY(spa_config_parse(spa, &vd, l2cache[i], NULL, 0, - VDEV_ALLOC_L2CACHE) == 0); + VERIFY0(spa_config_parse(spa, &vd, l2cache[i], NULL, 0, + VDEV_ALLOC_L2CACHE)); ASSERT(vd != NULL); newvdevs[i] = vd; @@ -2646,7 +2799,7 @@ spa_passivate_log(spa_t *spa) vdev_t *tvd = rvd->vdev_child[c]; if (tvd->vdev_islog) { - ASSERT3P(tvd->vdev_log_mg, ==, NULL); + ASSERT0P(tvd->vdev_log_mg); metaslab_group_passivate(tvd->vdev_mg); slog_found = B_TRUE; } @@ -2669,7 +2822,7 @@ spa_activate_log(spa_t *spa) vdev_t *tvd = rvd->vdev_child[c]; if (tvd->vdev_islog) { - ASSERT3P(tvd->vdev_log_mg, ==, NULL); + ASSERT0P(tvd->vdev_log_mg); metaslab_group_activate(tvd->vdev_mg); } } @@ -2709,8 +2862,8 @@ spa_claim_notify(zio_t *zio) return; mutex_enter(&spa->spa_props_lock); /* any mutex will do */ - if (spa->spa_claim_max_txg < BP_GET_LOGICAL_BIRTH(zio->io_bp)) - spa->spa_claim_max_txg = BP_GET_LOGICAL_BIRTH(zio->io_bp); + if (spa->spa_claim_max_txg < BP_GET_BIRTH(zio->io_bp)) + spa->spa_claim_max_txg = BP_GET_BIRTH(zio->io_bp); mutex_exit(&spa->spa_props_lock); } @@ -3106,7 +3259,7 @@ spa_livelist_delete_cb(void *arg, zthr_t *z) static void spa_start_livelist_destroy_thread(spa_t *spa) { - ASSERT3P(spa->spa_livelist_delete_zthr, ==, NULL); + ASSERT0P(spa->spa_livelist_delete_zthr); spa->spa_livelist_delete_zthr = zthr_create("z_livelist_destroy", spa_livelist_delete_cb_check, spa_livelist_delete_cb, spa, @@ -3122,7 +3275,7 @@ static int livelist_track_new_cb(void *arg, const blkptr_t *bp, boolean_t bp_freed, dmu_tx_t *tx) { - ASSERT(tx == NULL); + ASSERT0P(tx); livelist_new_arg_t *lna = arg; if (bp_freed) { bplist_append(lna->frees, bp); @@ -3316,7 +3469,7 @@ spa_start_livelist_condensing_thread(spa_t *spa) spa->spa_to_condense.syncing = B_FALSE; spa->spa_to_condense.cancelled = B_FALSE; - ASSERT3P(spa->spa_livelist_condense_zthr, ==, NULL); + ASSERT0P(spa->spa_livelist_condense_zthr); spa->spa_livelist_condense_zthr = zthr_create("z_livelist_condense", spa_livelist_condense_cb_check, @@ -3333,7 +3486,7 @@ spa_spawn_aux_threads(spa_t *spa) spa_start_livelist_destroy_thread(spa); spa_start_livelist_condensing_thread(spa); - ASSERT3P(spa->spa_checkpoint_discard_zthr, ==, NULL); + ASSERT0P(spa->spa_checkpoint_discard_zthr); spa->spa_checkpoint_discard_zthr = zthr_create("z_checkpoint_discard", spa_checkpoint_discard_thread_check, @@ -3768,20 +3921,17 @@ out: * ZPOOL_CONFIG_MMP_HOSTID - hostid from the active pool */ if (error == EREMOTEIO) { - const char *hostname = "<unknown>"; - uint64_t hostid = 0; - if (mmp_label) { if (nvlist_exists(mmp_label, ZPOOL_CONFIG_HOSTNAME)) { - hostname = fnvlist_lookup_string(mmp_label, - ZPOOL_CONFIG_HOSTNAME); + const char *hostname = fnvlist_lookup_string( + mmp_label, ZPOOL_CONFIG_HOSTNAME); fnvlist_add_string(spa->spa_load_info, ZPOOL_CONFIG_MMP_HOSTNAME, hostname); } if (nvlist_exists(mmp_label, ZPOOL_CONFIG_HOSTID)) { - hostid = fnvlist_lookup_uint64(mmp_label, - ZPOOL_CONFIG_HOSTID); + uint64_t hostid = fnvlist_lookup_uint64( + mmp_label, ZPOOL_CONFIG_HOSTID); fnvlist_add_uint64(spa->spa_load_info, ZPOOL_CONFIG_MMP_HOSTID, hostid); } @@ -3941,11 +4091,11 @@ spa_ld_parse_config(spa_t *spa, spa_import_type_t type) nvlist_free(spa->spa_load_info); spa->spa_load_info = fnvlist_alloc(); - ASSERT(spa->spa_comment == NULL); + ASSERT0P(spa->spa_comment); if (nvlist_lookup_string(config, ZPOOL_CONFIG_COMMENT, &comment) == 0) spa->spa_comment = spa_strdup(comment); - ASSERT(spa->spa_compatibility == NULL); + ASSERT0P(spa->spa_compatibility); if (nvlist_lookup_string(config, ZPOOL_CONFIG_COMPATIBILITY, &compatibility) == 0) spa->spa_compatibility = spa_strdup(compatibility); @@ -4711,6 +4861,9 @@ spa_ld_get_props(spa_t *spa) if (error != 0 && error != ENOENT) return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO)); + /* Load time log */ + spa_load_txg_log_time(spa); + /* * Load the persistent error log. If we have an older pool, this will * not be present. @@ -5760,7 +5913,7 @@ spa_load_best(spa_t *spa, spa_load_state_t state, uint64_t max_request, nvlist_free(config); if (state == SPA_LOAD_RECOVER) { - ASSERT3P(loadinfo, ==, NULL); + ASSERT0P(loadinfo); spa_import_progress_remove(spa_guid(spa)); return (rewind_error); } else { @@ -5899,7 +6052,7 @@ spa_open_common(const char *pool, spa_t **spapp, const void *tag, } if (firstopen) - zvol_create_minors_recursive(spa_name(spa)); + zvol_create_minors(spa_name(spa)); *spapp = spa; @@ -6877,7 +7030,7 @@ spa_import(char *pool, nvlist_t *config, nvlist_t *props, uint64_t flags) mutex_exit(&spa_namespace_lock); - zvol_create_minors_recursive(pool); + zvol_create_minors(pool); spa_import_os(spa); @@ -7134,6 +7287,9 @@ spa_export_common(const char *pool, int new_state, nvlist_t **oldconfig, spa_config_exit(spa, SCL_ALL, FTAG); } + if (spa_should_sync_time_logger_on_unload(spa)) + spa_unload_sync_time_logger(spa); + /* * If the log space map feature is enabled and the pool is * getting exported (but not destroyed), we want to spend some @@ -8935,7 +9091,7 @@ spa_vdev_setfru(spa_t *spa, uint64_t guid, const char *newfru) int spa_scrub_pause_resume(spa_t *spa, pool_scrub_cmd_t cmd) { - ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == 0); + ASSERT0(spa_config_held(spa, SCL_ALL, RW_WRITER)); if (dsl_scan_resilvering(spa->spa_dsl_pool)) return (SET_ERROR(EBUSY)); @@ -8946,7 +9102,7 @@ spa_scrub_pause_resume(spa_t *spa, pool_scrub_cmd_t cmd) int spa_scan_stop(spa_t *spa) { - ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == 0); + ASSERT0(spa_config_held(spa, SCL_ALL, RW_WRITER)); if (dsl_scan_resilvering(spa->spa_dsl_pool)) return (SET_ERROR(EBUSY)); @@ -8963,7 +9119,7 @@ int spa_scan_range(spa_t *spa, pool_scan_func_t func, uint64_t txgstart, uint64_t txgend) { - ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == 0); + ASSERT0(spa_config_held(spa, SCL_ALL, RW_WRITER)); if (func >= POOL_SCAN_FUNCS || func == POOL_SCAN_NONE) return (SET_ERROR(ENOTSUP)); @@ -9092,6 +9248,8 @@ spa_async_thread(void *arg) old_space += metaslab_class_get_space(spa_dedup_class(spa)); old_space += metaslab_class_get_space( spa_embedded_log_class(spa)); + old_space += metaslab_class_get_space( + spa_special_embedded_log_class(spa)); spa_config_update(spa, SPA_CONFIG_UPDATE_POOL); @@ -9100,6 +9258,8 @@ spa_async_thread(void *arg) new_space += metaslab_class_get_space(spa_dedup_class(spa)); new_space += metaslab_class_get_space( spa_embedded_log_class(spa)); + new_space += metaslab_class_get_space( + spa_special_embedded_log_class(spa)); mutex_exit(&spa_namespace_lock); /* @@ -9388,7 +9548,7 @@ spa_sync_frees(spa_t *spa, bplist_t *bpl, dmu_tx_t *tx) { zio_t *zio = zio_root(spa, NULL, NULL, 0); bplist_iterate(bpl, spa_free_sync_cb, zio, tx); - VERIFY(zio_wait(zio) == 0); + VERIFY0(zio_wait(zio)); } /* @@ -9427,7 +9587,7 @@ spa_sync_nvlist(spa_t *spa, uint64_t obj, nvlist_t *nv, dmu_tx_t *tx) size_t nvsize = 0; dmu_buf_t *db; - VERIFY(nvlist_size(nv, &nvsize, NV_ENCODE_XDR) == 0); + VERIFY0(nvlist_size(nv, &nvsize, NV_ENCODE_XDR)); /* * Write full (SPA_CONFIG_BLOCKSIZE) blocks of configuration @@ -9437,15 +9597,15 @@ spa_sync_nvlist(spa_t *spa, uint64_t obj, nvlist_t *nv, dmu_tx_t *tx) bufsize = P2ROUNDUP((uint64_t)nvsize, SPA_CONFIG_BLOCKSIZE); packed = vmem_alloc(bufsize, KM_SLEEP); - VERIFY(nvlist_pack(nv, &packed, &nvsize, NV_ENCODE_XDR, - KM_SLEEP) == 0); + VERIFY0(nvlist_pack(nv, &packed, &nvsize, NV_ENCODE_XDR, + KM_SLEEP)); memset(packed + nvsize, 0, bufsize - nvsize); dmu_write(spa->spa_meta_objset, obj, 0, bufsize, packed, tx); vmem_free(packed, bufsize); - VERIFY(0 == dmu_bonus_hold(spa->spa_meta_objset, obj, FTAG, &db)); + VERIFY0(dmu_bonus_hold(spa->spa_meta_objset, obj, FTAG, &db)); dmu_buf_will_dirty(db, tx); *(uint64_t *)db->db_data = nvsize; dmu_buf_rele(db, FTAG); @@ -10180,6 +10340,8 @@ spa_sync(spa_t *spa, uint64_t txg) */ brt_pending_apply(spa, txg); + spa_sync_time_logger(spa, txg); + /* * Lock out configuration changes. */ @@ -10222,6 +10384,7 @@ spa_sync(spa_t *spa, uint64_t txg) dmu_tx_t *tx = dmu_tx_create_assigned(dp, txg); spa->spa_sync_starttime = gethrtime(); + taskq_cancel_id(system_delay_taskq, spa->spa_deadman_tqid); spa->spa_deadman_tqid = taskq_dispatch_delay(system_delay_taskq, spa_deadman, spa, TQ_SLEEP, ddi_get_lbolt() + @@ -10309,7 +10472,7 @@ spa_sync(spa_t *spa, uint64_t txg) metaslab_class_evict_old(spa->spa_normal_class, txg); metaslab_class_evict_old(spa->spa_log_class, txg); - /* spa_embedded_log_class has only one metaslab per vdev. */ + /* Embedded log classes have only one metaslab per vdev. */ metaslab_class_evict_old(spa->spa_special_class, txg); metaslab_class_evict_old(spa->spa_dedup_class, txg); @@ -10378,7 +10541,7 @@ spa_sync_tq_create(spa_t *spa, const char *name) { kthread_t **kthreads; - ASSERT(spa->spa_sync_tq == NULL); + ASSERT0P(spa->spa_sync_tq); ASSERT3S(spa->spa_alloc_count, <=, boot_ncpus); /* @@ -11095,6 +11258,13 @@ ZFS_MODULE_PARAM(zfs_livelist_condense, zfs_livelist_condense_, new_alloc, INT, "Whether extra ALLOC blkptrs were added to a livelist entry while it " "was being condensed"); +ZFS_MODULE_PARAM(zfs_spa, spa_, note_txg_time, UINT, ZMOD_RW, + "How frequently TXG timestamps are stored internally (in seconds)"); + +ZFS_MODULE_PARAM(zfs_spa, spa_, flush_txg_time, UINT, ZMOD_RW, + "How frequently the TXG timestamps database should be flushed " + "to disk (in seconds)"); + #ifdef _KERNEL ZFS_MODULE_VIRTUAL_PARAM_CALL(zfs_zio, zio_, taskq_read, spa_taskq_read_param_set, spa_taskq_read_param_get, ZMOD_RW, diff --git a/sys/contrib/openzfs/module/zfs/spa_errlog.c b/sys/contrib/openzfs/module/zfs/spa_errlog.c index 3e08f261fda1..7252fd534bdf 100644 --- a/sys/contrib/openzfs/module/zfs/spa_errlog.c +++ b/sys/contrib/openzfs/module/zfs/spa_errlog.c @@ -253,7 +253,7 @@ find_birth_txg(dsl_dataset_t *ds, zbookmark_err_phys_t *zep, if (error == 0 && BP_IS_HOLE(&bp)) error = SET_ERROR(ENOENT); - *birth_txg = BP_GET_LOGICAL_BIRTH(&bp); + *birth_txg = BP_GET_PHYSICAL_BIRTH(&bp); rw_exit(&dn->dn_struct_rwlock); dnode_rele(dn, FTAG); return (error); @@ -885,7 +885,7 @@ sync_upgrade_errlog(spa_t *spa, uint64_t spa_err_obj, uint64_t *newobj, if (error == EACCES) error = 0; else if (!error) - zep.zb_birth = BP_GET_LOGICAL_BIRTH(&bp); + zep.zb_birth = BP_GET_PHYSICAL_BIRTH(&bp); rw_exit(&dn->dn_struct_rwlock); dnode_rele(dn, FTAG); diff --git a/sys/contrib/openzfs/module/zfs/spa_misc.c b/sys/contrib/openzfs/module/zfs/spa_misc.c index f054e4290bbf..cce772eae598 100644 --- a/sys/contrib/openzfs/module/zfs/spa_misc.c +++ b/sys/contrib/openzfs/module/zfs/spa_misc.c @@ -471,9 +471,9 @@ spa_config_lock_destroy(spa_t *spa) spa_config_lock_t *scl = &spa->spa_config_lock[i]; mutex_destroy(&scl->scl_lock); cv_destroy(&scl->scl_cv); - ASSERT(scl->scl_writer == NULL); - ASSERT(scl->scl_write_wanted == 0); - ASSERT(scl->scl_count == 0); + ASSERT0P(scl->scl_writer); + ASSERT0(scl->scl_write_wanted); + ASSERT0(scl->scl_count); } } @@ -715,6 +715,7 @@ spa_add(const char *name, nvlist_t *config, const char *altroot) mutex_init(&spa->spa_feat_stats_lock, NULL, MUTEX_DEFAULT, NULL); mutex_init(&spa->spa_flushed_ms_lock, NULL, MUTEX_DEFAULT, NULL); mutex_init(&spa->spa_activities_lock, NULL, MUTEX_DEFAULT, NULL); + mutex_init(&spa->spa_txg_log_time_lock, NULL, MUTEX_DEFAULT, NULL); cv_init(&spa->spa_async_cv, NULL, CV_DEFAULT, NULL); cv_init(&spa->spa_evicting_os_cv, NULL, CV_DEFAULT, NULL); @@ -783,24 +784,23 @@ spa_add(const char *name, nvlist_t *config, const char *altroot) dp->scd_path = altroot ? NULL : spa_strdup(spa_config_path); list_insert_head(&spa->spa_config_list, dp); - VERIFY(nvlist_alloc(&spa->spa_load_info, NV_UNIQUE_NAME, - KM_SLEEP) == 0); + VERIFY0(nvlist_alloc(&spa->spa_load_info, NV_UNIQUE_NAME, KM_SLEEP)); if (config != NULL) { nvlist_t *features; if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_FEATURES_FOR_READ, &features) == 0) { - VERIFY(nvlist_dup(features, &spa->spa_label_features, - 0) == 0); + VERIFY0(nvlist_dup(features, + &spa->spa_label_features, 0)); } - VERIFY(nvlist_dup(config, &spa->spa_config, 0) == 0); + VERIFY0(nvlist_dup(config, &spa->spa_config, 0)); } if (spa->spa_label_features == NULL) { - VERIFY(nvlist_alloc(&spa->spa_label_features, NV_UNIQUE_NAME, - KM_SLEEP) == 0); + VERIFY0(nvlist_alloc(&spa->spa_label_features, NV_UNIQUE_NAME, + KM_SLEEP)); } spa->spa_min_ashift = INT_MAX; @@ -903,6 +903,7 @@ spa_remove(spa_t *spa) mutex_destroy(&spa->spa_vdev_top_lock); mutex_destroy(&spa->spa_feat_stats_lock); mutex_destroy(&spa->spa_activities_lock); + mutex_destroy(&spa->spa_txg_log_time_lock); kmem_free(spa, sizeof (spa_t)); } @@ -1308,6 +1309,7 @@ spa_vdev_config_exit(spa_t *spa, vdev_t *vd, uint64_t txg, int error, metaslab_class_validate(spa_log_class(spa)); metaslab_class_validate(spa_embedded_log_class(spa)); metaslab_class_validate(spa_special_class(spa)); + metaslab_class_validate(spa_special_embedded_log_class(spa)); metaslab_class_validate(spa_dedup_class(spa)); spa_config_exit(spa, SCL_ALL, spa); @@ -1896,6 +1898,8 @@ spa_get_slop_space(spa_t *spa) */ uint64_t embedded_log = metaslab_class_get_dspace(spa_embedded_log_class(spa)); + embedded_log += metaslab_class_get_dspace( + spa_special_embedded_log_class(spa)); slop -= MIN(embedded_log, slop >> 1); /* @@ -2001,6 +2005,12 @@ spa_special_class(spa_t *spa) } metaslab_class_t * +spa_special_embedded_log_class(spa_t *spa) +{ + return (spa->spa_special_embedded_log_class); +} + +metaslab_class_t * spa_dedup_class(spa_t *spa) { return (spa->spa_dedup_class); diff --git a/sys/contrib/openzfs/module/zfs/spa_stats.c b/sys/contrib/openzfs/module/zfs/spa_stats.c index 6d7cabcf766d..2c87122a0aa9 100644 --- a/sys/contrib/openzfs/module/zfs/spa_stats.c +++ b/sys/contrib/openzfs/module/zfs/spa_stats.c @@ -718,7 +718,7 @@ spa_mmp_history_set(spa_t *spa, uint64_t mmp_node_id, int io_error, for (smh = list_tail(&shl->procfs_list.pl_list); smh != NULL; smh = list_prev(&shl->procfs_list.pl_list, smh)) { if (smh->mmp_node_id == mmp_node_id) { - ASSERT(smh->io_error == 0); + ASSERT0(smh->io_error); smh->io_error = io_error; smh->duration = duration; error = 0; diff --git a/sys/contrib/openzfs/module/zfs/space_map.c b/sys/contrib/openzfs/module/zfs/space_map.c index c429e0edd168..5f24963f2291 100644 --- a/sys/contrib/openzfs/module/zfs/space_map.c +++ b/sys/contrib/openzfs/module/zfs/space_map.c @@ -817,7 +817,7 @@ space_map_open(space_map_t **smp, objset_t *os, uint64_t object, space_map_t *sm; int error; - ASSERT(*smp == NULL); + ASSERT0P(*smp); ASSERT(os != NULL); ASSERT(object != 0); diff --git a/sys/contrib/openzfs/module/zfs/space_reftree.c b/sys/contrib/openzfs/module/zfs/space_reftree.c index 9b2d5ed31dc9..889980e08c06 100644 --- a/sys/contrib/openzfs/module/zfs/space_reftree.c +++ b/sys/contrib/openzfs/module/zfs/space_reftree.c @@ -149,6 +149,6 @@ space_reftree_generate_map(avl_tree_t *t, zfs_range_tree_t *rt, int64_t minref) } } } - ASSERT(refcnt == 0); + ASSERT0(refcnt); ASSERT(start == -1ULL); } diff --git a/sys/contrib/openzfs/module/zfs/vdev.c b/sys/contrib/openzfs/module/zfs/vdev.c index 01758b0c54c0..9cf35e379000 100644 --- a/sys/contrib/openzfs/module/zfs/vdev.c +++ b/sys/contrib/openzfs/module/zfs/vdev.c @@ -243,6 +243,25 @@ vdev_dbgmsg_print_tree(vdev_t *vd, int indent) vdev_dbgmsg_print_tree(vd->vdev_child[i], indent + 2); } +char * +vdev_rt_name(vdev_t *vd, const char *name) +{ + return (kmem_asprintf("{spa=%s vdev_guid=%llu %s}", + spa_name(vd->vdev_spa), + (u_longlong_t)vd->vdev_guid, + name)); +} + +static char * +vdev_rt_name_dtl(vdev_t *vd, const char *name, vdev_dtl_type_t dtl_type) +{ + return (kmem_asprintf("{spa=%s vdev_guid=%llu %s[%d]}", + spa_name(vd->vdev_spa), + (u_longlong_t)vd->vdev_guid, + name, + dtl_type)); +} + /* * Virtual device management. */ @@ -282,12 +301,15 @@ vdev_getops(const char *type) * Given a vdev and a metaslab class, find which metaslab group we're * interested in. All vdevs may belong to two different metaslab classes. * Dedicated slog devices use only the primary metaslab group, rather than a - * separate log group. For embedded slogs, the vdev_log_mg will be non-NULL. + * separate log group. For embedded slogs, vdev_log_mg will be non-NULL and + * will point to a metaslab group of either embedded_log_class (for normal + * vdevs) or special_embedded_log_class (for special vdevs). */ metaslab_group_t * vdev_get_mg(vdev_t *vd, metaslab_class_t *mc) { - if (mc == spa_embedded_log_class(vd->vdev_spa) && + if ((mc == spa_embedded_log_class(vd->vdev_spa) || + mc == spa_special_embedded_log_class(vd->vdev_spa)) && vd->vdev_log_mg != NULL) return (vd->vdev_log_mg); else @@ -532,7 +554,7 @@ vdev_add_child(vdev_t *pvd, vdev_t *cvd) vdev_t **newchild; ASSERT(spa_config_held(cvd->vdev_spa, SCL_ALL, RW_WRITER) == SCL_ALL); - ASSERT(cvd->vdev_parent == NULL); + ASSERT0P(cvd->vdev_parent); cvd->vdev_parent = pvd; @@ -556,7 +578,7 @@ vdev_add_child(vdev_t *pvd, vdev_t *cvd) pvd->vdev_nonrot &= cvd->vdev_nonrot; cvd->vdev_top = (pvd->vdev_top ? pvd->vdev_top: cvd); - ASSERT(cvd->vdev_top->vdev_parent->vdev_parent == NULL); + ASSERT0P(cvd->vdev_top->vdev_parent->vdev_parent); /* * Walk up all ancestors to update guid sum. @@ -692,8 +714,9 @@ vdev_alloc_common(spa_t *spa, uint_t id, uint64_t guid, vdev_ops_t *ops) rw_init(&vd->vdev_indirect_rwlock, NULL, RW_DEFAULT, NULL); mutex_init(&vd->vdev_obsolete_lock, NULL, MUTEX_DEFAULT, NULL); - vd->vdev_obsolete_segments = zfs_range_tree_create(NULL, - ZFS_RANGE_SEG64, NULL, 0, 0); + vd->vdev_obsolete_segments = zfs_range_tree_create_flags( + NULL, ZFS_RANGE_SEG64, NULL, 0, 0, + ZFS_RT_F_DYN_NAME, vdev_rt_name(vd, "vdev_obsolete_segments")); /* * Initialize rate limit structs for events. We rate limit ZIO delay @@ -747,8 +770,9 @@ vdev_alloc_common(spa_t *spa, uint_t id, uint64_t guid, vdev_ops_t *ops) cv_init(&vd->vdev_rebuild_cv, NULL, CV_DEFAULT, NULL); for (int t = 0; t < DTL_TYPES; t++) { - vd->vdev_dtl[t] = zfs_range_tree_create(NULL, ZFS_RANGE_SEG64, - NULL, 0, 0); + vd->vdev_dtl[t] = zfs_range_tree_create_flags( + NULL, ZFS_RANGE_SEG64, NULL, 0, 0, + ZFS_RT_F_DYN_NAME, vdev_rt_name_dtl(vd, "vdev_dtl", t)); } txg_list_create(&vd->vdev_ms_list, spa, @@ -1077,10 +1101,10 @@ vdev_free(vdev_t *vd) { spa_t *spa = vd->vdev_spa; - ASSERT3P(vd->vdev_initialize_thread, ==, NULL); - ASSERT3P(vd->vdev_trim_thread, ==, NULL); - ASSERT3P(vd->vdev_autotrim_thread, ==, NULL); - ASSERT3P(vd->vdev_rebuild_thread, ==, NULL); + ASSERT0P(vd->vdev_initialize_thread); + ASSERT0P(vd->vdev_trim_thread); + ASSERT0P(vd->vdev_autotrim_thread); + ASSERT0P(vd->vdev_rebuild_thread); /* * Scan queues are normally destroyed at the end of a scan. If the @@ -1109,7 +1133,7 @@ vdev_free(vdev_t *vd) for (int c = 0; c < vd->vdev_children; c++) vdev_free(vd->vdev_child[c]); - ASSERT(vd->vdev_child == NULL); + ASSERT0P(vd->vdev_child); ASSERT(vd->vdev_guid_sum == vd->vdev_guid); if (vd->vdev_ops->vdev_op_fini != NULL) @@ -1138,7 +1162,7 @@ vdev_free(vdev_t *vd) */ vdev_remove_child(vd->vdev_parent, vd); - ASSERT(vd->vdev_parent == NULL); + ASSERT0P(vd->vdev_parent); ASSERT(!list_link_active(&vd->vdev_leaf_node)); /* @@ -1285,9 +1309,9 @@ vdev_top_transfer(vdev_t *svd, vdev_t *tvd) ASSERT0(tvd->vdev_indirect_config.vic_births_object); ASSERT0(tvd->vdev_indirect_config.vic_mapping_object); ASSERT3U(tvd->vdev_indirect_config.vic_prev_indirect_vdev, ==, -1ULL); - ASSERT3P(tvd->vdev_indirect_mapping, ==, NULL); - ASSERT3P(tvd->vdev_indirect_births, ==, NULL); - ASSERT3P(tvd->vdev_obsolete_sm, ==, NULL); + ASSERT0P(tvd->vdev_indirect_mapping); + ASSERT0P(tvd->vdev_indirect_births); + ASSERT0P(tvd->vdev_obsolete_sm); ASSERT0(tvd->vdev_noalloc); ASSERT0(tvd->vdev_removing); ASSERT0(tvd->vdev_rebuilding); @@ -1440,7 +1464,7 @@ vdev_remove_parent(vdev_t *cvd) if (cvd == cvd->vdev_top) vdev_top_transfer(mvd, cvd); - ASSERT(mvd->vdev_children == 0); + ASSERT0(mvd->vdev_children); vdev_free(mvd); } @@ -1508,8 +1532,13 @@ vdev_metaslab_group_create(vdev_t *vd) vd->vdev_mg = metaslab_group_create(mc, vd); if (!vd->vdev_islog) { - vd->vdev_log_mg = metaslab_group_create( - spa_embedded_log_class(spa), vd); + if (mc == spa_special_class(spa)) { + vd->vdev_log_mg = metaslab_group_create( + spa_special_embedded_log_class(spa), vd); + } else { + vd->vdev_log_mg = metaslab_group_create( + spa_embedded_log_class(spa), vd); + } } /* @@ -1624,9 +1653,10 @@ vdev_metaslab_init(vdev_t *vd, uint64_t txg) /* * Find the emptiest metaslab on the vdev and mark it for use for * embedded slog by moving it from the regular to the log metaslab - * group. + * group. This works for normal and special vdevs. */ - if (vd->vdev_mg->mg_class == spa_normal_class(spa) && + if ((vd->vdev_mg->mg_class == spa_normal_class(spa) || + vd->vdev_mg->mg_class == spa_special_class(spa)) && vd->vdev_ms_count > zfs_embedded_slog_min_ms && avl_is_empty(&vd->vdev_log_mg->mg_metaslab_tree)) { uint64_t slog_msid = 0; @@ -2104,14 +2134,14 @@ vdev_open(vdev_t *vd) * faulted, bail out of the open. */ if (!vd->vdev_removed && vd->vdev_faulted) { - ASSERT(vd->vdev_children == 0); + ASSERT0(vd->vdev_children); ASSERT(vd->vdev_label_aux == VDEV_AUX_ERR_EXCEEDED || vd->vdev_label_aux == VDEV_AUX_EXTERNAL); vdev_set_state(vd, B_TRUE, VDEV_STATE_FAULTED, vd->vdev_label_aux); return (SET_ERROR(ENXIO)); } else if (vd->vdev_offline) { - ASSERT(vd->vdev_children == 0); + ASSERT0(vd->vdev_children); vdev_set_state(vd, B_TRUE, VDEV_STATE_OFFLINE, VDEV_AUX_NONE); return (SET_ERROR(ENXIO)); } @@ -2167,7 +2197,7 @@ vdev_open(vdev_t *vd) * the vdev is accessible. If we're faulted, bail. */ if (vd->vdev_faulted) { - ASSERT(vd->vdev_children == 0); + ASSERT0(vd->vdev_children); ASSERT(vd->vdev_label_aux == VDEV_AUX_ERR_EXCEEDED || vd->vdev_label_aux == VDEV_AUX_EXTERNAL); vdev_set_state(vd, B_TRUE, VDEV_STATE_FAULTED, @@ -2176,7 +2206,7 @@ vdev_open(vdev_t *vd) } if (vd->vdev_degraded) { - ASSERT(vd->vdev_children == 0); + ASSERT0(vd->vdev_children); vdev_set_state(vd, B_TRUE, VDEV_STATE_DEGRADED, VDEV_AUX_ERR_EXCEEDED); } else { @@ -3449,7 +3479,9 @@ vdev_dtl_load(vdev_t *vd) return (error); ASSERT(vd->vdev_dtl_sm != NULL); - rt = zfs_range_tree_create(NULL, ZFS_RANGE_SEG64, NULL, 0, 0); + rt = zfs_range_tree_create_flags( + NULL, ZFS_RANGE_SEG64, NULL, 0, 0, + ZFS_RT_F_DYN_NAME, vdev_rt_name(vd, "vdev_dtl_load:rt")); error = space_map_load(vd->vdev_dtl_sm, rt, SM_ALLOC); if (error == 0) { mutex_enter(&vd->vdev_dtl_lock); @@ -3597,7 +3629,8 @@ vdev_dtl_sync(vdev_t *vd, uint64_t txg) ASSERT(vd->vdev_dtl_sm != NULL); } - rtsync = zfs_range_tree_create(NULL, ZFS_RANGE_SEG64, NULL, 0, 0); + rtsync = zfs_range_tree_create_flags(NULL, ZFS_RANGE_SEG64, NULL, 0, 0, + ZFS_RT_F_DYN_NAME, vdev_rt_name(vd, "rtsync")); mutex_enter(&vd->vdev_dtl_lock); zfs_range_tree_walk(rt, zfs_range_tree_add, rtsync); @@ -3912,7 +3945,7 @@ vdev_load(vdev_t *vd) if (error == 0 && checkpoint_sm_obj != 0) { objset_t *mos = spa_meta_objset(vd->vdev_spa); ASSERT(vd->vdev_asize != 0); - ASSERT3P(vd->vdev_checkpoint_sm, ==, NULL); + ASSERT0P(vd->vdev_checkpoint_sm); error = space_map_open(&vd->vdev_checkpoint_sm, mos, checkpoint_sm_obj, 0, vd->vdev_asize, @@ -3960,7 +3993,7 @@ vdev_load(vdev_t *vd) if (error == 0 && obsolete_sm_object != 0) { objset_t *mos = vd->vdev_spa->spa_meta_objset; ASSERT(vd->vdev_asize != 0); - ASSERT3P(vd->vdev_obsolete_sm, ==, NULL); + ASSERT0P(vd->vdev_obsolete_sm); if ((error = space_map_open(&vd->vdev_obsolete_sm, mos, obsolete_sm_object, 0, vd->vdev_asize, 0))) { @@ -4488,7 +4521,7 @@ top: /* * Prevent any future allocations. */ - ASSERT3P(tvd->vdev_log_mg, ==, NULL); + ASSERT0P(tvd->vdev_log_mg); metaslab_group_passivate(mg); (void) spa_vdev_state_exit(spa, vd, 0); @@ -5161,7 +5194,7 @@ vdev_stat_update(zio_t *zio, uint64_t psize) int64_t vdev_deflated_space(vdev_t *vd, int64_t space) { - ASSERT((space & (SPA_MINBLOCKSIZE-1)) == 0); + ASSERT0((space & (SPA_MINBLOCKSIZE-1))); ASSERT(vd->vdev_deflate_ratio != 0 || vd->vdev_isl2cache); return ((space >> SPA_MINBLOCKSHIFT) * vd->vdev_deflate_ratio); @@ -5253,8 +5286,8 @@ vdev_config_dirty(vdev_t *vd) if (nvlist_lookup_nvlist_array(sav->sav_config, ZPOOL_CONFIG_L2CACHE, &aux, &naux) != 0) { - VERIFY(nvlist_lookup_nvlist_array(sav->sav_config, - ZPOOL_CONFIG_SPARES, &aux, &naux) == 0); + VERIFY0(nvlist_lookup_nvlist_array(sav->sav_config, + ZPOOL_CONFIG_SPARES, &aux, &naux)); } ASSERT(c < naux); @@ -5642,7 +5675,7 @@ vdev_expand(vdev_t *vd, uint64_t txg) (vd->vdev_asize >> vd->vdev_ms_shift) > vd->vdev_ms_count && vdev_is_concrete(vd)) { vdev_metaslab_group_create(vd); - VERIFY(vdev_metaslab_init(vd, txg) == 0); + VERIFY0(vdev_metaslab_init(vd, txg)); vdev_config_dirty(vd); } } diff --git a/sys/contrib/openzfs/module/zfs/vdev_draid.c b/sys/contrib/openzfs/module/zfs/vdev_draid.c index feec5fd3ce17..a05289102af2 100644 --- a/sys/contrib/openzfs/module/zfs/vdev_draid.c +++ b/sys/contrib/openzfs/module/zfs/vdev_draid.c @@ -477,7 +477,7 @@ vdev_draid_generate_perms(const draid_map_t *map, uint8_t **permsp) VERIFY3U(map->dm_children, <=, VDEV_DRAID_MAX_CHILDREN); VERIFY3U(map->dm_seed, !=, 0); VERIFY3U(map->dm_nperms, !=, 0); - VERIFY3P(map->dm_perms, ==, NULL); + VERIFY0P(map->dm_perms); #ifdef _KERNEL /* @@ -590,7 +590,7 @@ vdev_draid_psize_to_asize(vdev_t *vd, uint64_t psize, uint64_t txg) uint64_t asize = (rows * vdc->vdc_groupwidth) << ashift; ASSERT3U(asize, !=, 0); - ASSERT3U(asize % (vdc->vdc_groupwidth), ==, 0); + ASSERT0(asize % (vdc->vdc_groupwidth)); return (asize); } @@ -704,7 +704,7 @@ vdev_draid_map_alloc_scrub(zio_t *zio, uint64_t abd_offset, raidz_row_t *rr) uint64_t skip_off = 0; ASSERT3U(zio->io_type, ==, ZIO_TYPE_READ); - ASSERT3P(rr->rr_abd_empty, ==, NULL); + ASSERT0P(rr->rr_abd_empty); if (rr->rr_nempty > 0) { rr->rr_abd_empty = abd_alloc_linear(rr->rr_nempty * skip_size, @@ -793,7 +793,7 @@ vdev_draid_map_alloc_empty(zio_t *zio, raidz_row_t *rr) uint64_t skip_off = 0; ASSERT3U(zio->io_type, ==, ZIO_TYPE_READ); - ASSERT3P(rr->rr_abd_empty, ==, NULL); + ASSERT0P(rr->rr_abd_empty); if (rr->rr_nempty > 0) { rr->rr_abd_empty = abd_alloc_linear(rr->rr_nempty * skip_size, @@ -807,7 +807,7 @@ vdev_draid_map_alloc_empty(zio_t *zio, raidz_row_t *rr) /* empty data column (small read), add a skip sector */ ASSERT3U(skip_size, ==, parity_size); ASSERT3U(rr->rr_nempty, !=, 0); - ASSERT3P(rc->rc_abd, ==, NULL); + ASSERT0P(rc->rc_abd); rc->rc_abd = abd_get_offset_size(rr->rr_abd_empty, skip_off, skip_size); skip_off += skip_size; @@ -1623,7 +1623,7 @@ vdev_draid_rebuild_asize(vdev_t *vd, uint64_t start, uint64_t asize, SPA_MAXBLOCKSIZE); ASSERT3U(vdev_draid_get_astart(vd, start), ==, start); - ASSERT3U(asize % (vdc->vdc_groupwidth << ashift), ==, 0); + ASSERT0(asize % (vdc->vdc_groupwidth << ashift)); /* Chunks must evenly span all data columns in the group. */ psize = (((psize >> ashift) / ndata) * ndata) << ashift; @@ -1634,7 +1634,7 @@ vdev_draid_rebuild_asize(vdev_t *vd, uint64_t start, uint64_t asize, uint64_t left = vdev_draid_group_to_offset(vd, group + 1) - start; chunk_size = MIN(chunk_size, left); - ASSERT3U(chunk_size % (vdc->vdc_groupwidth << ashift), ==, 0); + ASSERT0(chunk_size % (vdc->vdc_groupwidth << ashift)); ASSERT3U(vdev_draid_offset_to_group(vd, start), ==, vdev_draid_offset_to_group(vd, start + chunk_size - 1)); @@ -2272,7 +2272,7 @@ vdev_draid_init(spa_t *spa, nvlist_t *nv, void **tsd) ASSERT3U(vdc->vdc_groupwidth, <=, vdc->vdc_ndisks); ASSERT3U(vdc->vdc_groupsz, >=, 2 * VDEV_DRAID_ROWHEIGHT); ASSERT3U(vdc->vdc_devslicesz, >=, VDEV_DRAID_ROWHEIGHT); - ASSERT3U(vdc->vdc_devslicesz % VDEV_DRAID_ROWHEIGHT, ==, 0); + ASSERT0(vdc->vdc_devslicesz % VDEV_DRAID_ROWHEIGHT); ASSERT3U((vdc->vdc_groupwidth * vdc->vdc_ngroups) % vdc->vdc_ndisks, ==, 0); diff --git a/sys/contrib/openzfs/module/zfs/vdev_indirect.c b/sys/contrib/openzfs/module/zfs/vdev_indirect.c index fac2c3a5f154..7538f471e63c 100644 --- a/sys/contrib/openzfs/module/zfs/vdev_indirect.c +++ b/sys/contrib/openzfs/module/zfs/vdev_indirect.c @@ -792,7 +792,7 @@ spa_condense_indirect_start_sync(vdev_t *vd, dmu_tx_t *tx) DMU_POOL_CONDENSING_INDIRECT, sizeof (uint64_t), sizeof (*scip) / sizeof (uint64_t), scip, tx)); - ASSERT3P(spa->spa_condensing_indirect, ==, NULL); + ASSERT0P(spa->spa_condensing_indirect); spa->spa_condensing_indirect = spa_condensing_indirect_create(spa); zfs_dbgmsg("starting condense of vdev %llu in txg %llu: " @@ -882,7 +882,7 @@ spa_condense_fini(spa_t *spa) void spa_start_indirect_condensing_thread(spa_t *spa) { - ASSERT3P(spa->spa_condense_zthr, ==, NULL); + ASSERT0P(spa->spa_condense_zthr); spa->spa_condense_zthr = zthr_create("z_indirect_condense", spa_condense_indirect_thread_check, spa_condense_indirect_thread, spa, minclsyspri); @@ -1504,7 +1504,7 @@ vdev_indirect_splits_checksum_validate(indirect_vsd_t *iv, zio_t *zio) is != NULL; is = list_next(&iv->iv_splits, is)) { ASSERT3P(is->is_good_child->ic_data, !=, NULL); - ASSERT3P(is->is_good_child->ic_duplicate, ==, NULL); + ASSERT0P(is->is_good_child->ic_duplicate); abd_copy_off(zio->io_abd, is->is_good_child->ic_data, is->is_split_offset, 0, is->is_size); @@ -1842,7 +1842,7 @@ vdev_indirect_io_done(zio_t *zio) */ if (zio->io_flags & ZIO_FLAG_DIO_READ && ret == ECKSUM) { zio->io_error = ret; - zio->io_flags |= ZIO_FLAG_DIO_CHKSUM_ERR; + zio->io_post |= ZIO_POST_DIO_CHKSUM_ERR; zio_dio_chksum_verify_error_report(zio); ret = 0; } diff --git a/sys/contrib/openzfs/module/zfs/vdev_initialize.c b/sys/contrib/openzfs/module/zfs/vdev_initialize.c index 4274728578ad..27188c46e561 100644 --- a/sys/contrib/openzfs/module/zfs/vdev_initialize.c +++ b/sys/contrib/openzfs/module/zfs/vdev_initialize.c @@ -541,8 +541,9 @@ vdev_initialize_thread(void *arg) abd_t *deadbeef = vdev_initialize_block_alloc(); - vd->vdev_initialize_tree = zfs_range_tree_create(NULL, ZFS_RANGE_SEG64, - NULL, 0, 0); + vd->vdev_initialize_tree = zfs_range_tree_create_flags( + NULL, ZFS_RANGE_SEG64, NULL, 0, 0, + ZFS_RT_F_DYN_NAME, vdev_rt_name(vd, "vdev_initialize_tree")); for (uint64_t i = 0; !vd->vdev_detached && i < vd->vdev_top->vdev_ms_count; i++) { @@ -631,7 +632,7 @@ vdev_initialize(vdev_t *vd) ASSERT(MUTEX_HELD(&vd->vdev_initialize_lock)); ASSERT(vd->vdev_ops->vdev_op_leaf); ASSERT(vdev_is_concrete(vd)); - ASSERT3P(vd->vdev_initialize_thread, ==, NULL); + ASSERT0P(vd->vdev_initialize_thread); ASSERT(!vd->vdev_detached); ASSERT(!vd->vdev_initialize_exit_wanted); ASSERT(!vd->vdev_top->vdev_removing); @@ -652,7 +653,7 @@ vdev_uninitialize(vdev_t *vd) ASSERT(MUTEX_HELD(&vd->vdev_initialize_lock)); ASSERT(vd->vdev_ops->vdev_op_leaf); ASSERT(vdev_is_concrete(vd)); - ASSERT3P(vd->vdev_initialize_thread, ==, NULL); + ASSERT0P(vd->vdev_initialize_thread); ASSERT(!vd->vdev_detached); ASSERT(!vd->vdev_initialize_exit_wanted); ASSERT(!vd->vdev_top->vdev_removing); @@ -671,7 +672,7 @@ vdev_initialize_stop_wait_impl(vdev_t *vd) while (vd->vdev_initialize_thread != NULL) cv_wait(&vd->vdev_initialize_cv, &vd->vdev_initialize_lock); - ASSERT3P(vd->vdev_initialize_thread, ==, NULL); + ASSERT0P(vd->vdev_initialize_thread); vd->vdev_initialize_exit_wanted = B_FALSE; } diff --git a/sys/contrib/openzfs/module/zfs/vdev_label.c b/sys/contrib/openzfs/module/zfs/vdev_label.c index 6baa6236aac2..c44f654b0261 100644 --- a/sys/contrib/openzfs/module/zfs/vdev_label.c +++ b/sys/contrib/openzfs/module/zfs/vdev_label.c @@ -163,7 +163,7 @@ uint64_t vdev_label_offset(uint64_t psize, int l, uint64_t offset) { ASSERT(offset < sizeof (vdev_label_t)); - ASSERT(P2PHASE_TYPED(psize, sizeof (vdev_label_t), uint64_t) == 0); + ASSERT0(P2PHASE_TYPED(psize, sizeof (vdev_label_t), uint64_t)); return (offset + l * sizeof (vdev_label_t) + (l < VDEV_LABELS / 2 ? 0 : psize - VDEV_LABELS * sizeof (vdev_label_t))); @@ -768,12 +768,12 @@ vdev_top_config_generate(spa_t *spa, nvlist_t *config) } if (idx) { - VERIFY(nvlist_add_uint64_array(config, ZPOOL_CONFIG_HOLE_ARRAY, - array, idx) == 0); + VERIFY0(nvlist_add_uint64_array(config, + ZPOOL_CONFIG_HOLE_ARRAY, array, idx)); } - VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_VDEV_CHILDREN, - rvd->vdev_children) == 0); + VERIFY0(nvlist_add_uint64(config, ZPOOL_CONFIG_VDEV_CHILDREN, + rvd->vdev_children)); kmem_free(array, rvd->vdev_children * sizeof (uint64_t)); } @@ -1189,8 +1189,8 @@ vdev_label_init(vdev_t *vd, uint64_t crtxg, vdev_labeltype_t reason) * vdev uses as described above, and automatically expires if we * fail. */ - VERIFY(nvlist_add_uint64(label, ZPOOL_CONFIG_CREATE_TXG, - crtxg) == 0); + VERIFY0(nvlist_add_uint64(label, ZPOOL_CONFIG_CREATE_TXG, + crtxg)); } buf = vp->vp_nvlist; diff --git a/sys/contrib/openzfs/module/zfs/vdev_mirror.c b/sys/contrib/openzfs/module/zfs/vdev_mirror.c index a6aee9437066..18efdaac006f 100644 --- a/sys/contrib/openzfs/module/zfs/vdev_mirror.c +++ b/sys/contrib/openzfs/module/zfs/vdev_mirror.c @@ -532,7 +532,7 @@ vdev_mirror_child_select(zio_t *zio) uint64_t txg = zio->io_txg; int c, lowest_load; - ASSERT(zio->io_bp == NULL || BP_GET_BIRTH(zio->io_bp) == txg); + ASSERT(zio->io_bp == NULL || BP_GET_PHYSICAL_BIRTH(zio->io_bp) == txg); lowest_load = INT_MAX; mm->mm_preferred_cnt = 0; @@ -779,7 +779,7 @@ vdev_mirror_io_done(zio_t *zio) * being written out during self healing. */ if ((zio->io_flags & ZIO_FLAG_DIO_READ) && - (zio->io_flags & ZIO_FLAG_DIO_CHKSUM_ERR)) { + (zio->io_post & ZIO_POST_DIO_CHKSUM_ERR)) { zio_dio_chksum_verify_error_report(zio); zio->io_error = vdev_mirror_worst_error(mm); ASSERT3U(zio->io_error, ==, ECKSUM); diff --git a/sys/contrib/openzfs/module/zfs/vdev_queue.c b/sys/contrib/openzfs/module/zfs/vdev_queue.c index aa41f7066036..c12713b107bf 100644 --- a/sys/contrib/openzfs/module/zfs/vdev_queue.c +++ b/sys/contrib/openzfs/module/zfs/vdev_queue.c @@ -780,7 +780,7 @@ vdev_queue_aggregate(vdev_queue_t *vq, zio_t *zio) if (dio->io_flags & ZIO_FLAG_NODATA) { /* allocate a buffer for a write gap */ ASSERT3U(dio->io_type, ==, ZIO_TYPE_WRITE); - ASSERT3P(dio->io_abd, ==, NULL); + ASSERT0P(dio->io_abd); abd_gang_add(aio->io_abd, abd_get_zeros(dio->io_size), B_TRUE); } else { diff --git a/sys/contrib/openzfs/module/zfs/vdev_raidz.c b/sys/contrib/openzfs/module/zfs/vdev_raidz.c index 71c4bfbdaf00..b597d6daefde 100644 --- a/sys/contrib/openzfs/module/zfs/vdev_raidz.c +++ b/sys/contrib/openzfs/module/zfs/vdev_raidz.c @@ -412,7 +412,7 @@ vdev_raidz_map_free(raidz_map_t *rm) rm->rm_nphys_cols); } - ASSERT3P(rm->rm_lr, ==, NULL); + ASSERT0P(rm->rm_lr); kmem_free(rm, offsetof(raidz_map_t, rm_row[rm->rm_nrows])); } @@ -2206,11 +2206,7 @@ vdev_raidz_close(vdev_t *vd) /* * Return the logical width to use, given the txg in which the allocation - * happened. Note that BP_GET_BIRTH() is usually the txg in which the - * BP was allocated. Remapped BP's (that were relocated due to device - * removal, see remap_blkptr_cb()), will have a more recent physical birth - * which reflects when the BP was relocated, but we can ignore these because - * they can't be on RAIDZ (device removal doesn't support RAIDZ). + * happened. */ static uint64_t vdev_raidz_get_logical_width(vdev_raidz_t *vdrz, uint64_t txg) @@ -2249,10 +2245,9 @@ vdev_raidz_asize_to_psize(vdev_t *vd, uint64_t asize, uint64_t txg) vdev_raidz_t *vdrz = vd->vdev_tsd; uint64_t psize; uint64_t ashift = vd->vdev_top->vdev_ashift; - uint64_t cols = vdrz->vd_original_width; uint64_t nparity = vdrz->vd_nparity; - cols = vdev_raidz_get_logical_width(vdrz, txg); + uint64_t cols = vdev_raidz_get_logical_width(vdrz, txg); ASSERT0(asize % (1 << ashift)); @@ -2285,10 +2280,9 @@ vdev_raidz_psize_to_asize(vdev_t *vd, uint64_t psize, uint64_t txg) vdev_raidz_t *vdrz = vd->vdev_tsd; uint64_t asize; uint64_t ashift = vd->vdev_top->vdev_ashift; - uint64_t cols = vdrz->vd_original_width; uint64_t nparity = vdrz->vd_nparity; - cols = vdev_raidz_get_logical_width(vdrz, txg); + uint64_t cols = vdev_raidz_get_logical_width(vdrz, txg); asize = ((psize - 1) >> ashift) + 1; asize += nparity * ((asize + cols - nparity - 1) / (cols - nparity)); @@ -2345,7 +2339,7 @@ vdev_raidz_io_verify(zio_t *zio, raidz_map_t *rm, raidz_row_t *rr, int col) logical_rs.rs_start = rr->rr_offset; logical_rs.rs_end = logical_rs.rs_start + vdev_raidz_psize_to_asize(zio->io_vd, rr->rr_size, - BP_GET_BIRTH(zio->io_bp)); + BP_GET_PHYSICAL_BIRTH(zio->io_bp)); raidz_col_t *rc = &rr->rr_col[col]; vdev_t *cvd = zio->io_vd->vdev_child[rc->rc_devidx]; @@ -2437,7 +2431,7 @@ raidz_start_skip_writes(zio_t *zio) vdev_t *cvd = vd->vdev_child[rc->rc_devidx]; if (rc->rc_size != 0) continue; - ASSERT3P(rc->rc_abd, ==, NULL); + ASSERT0P(rc->rc_abd); ASSERT3U(rc->rc_offset, <, cvd->vdev_psize - VDEV_LABEL_END_SIZE); @@ -2568,7 +2562,7 @@ vdev_raidz_io_start(zio_t *zio) raidz_map_t *rm; uint64_t logical_width = vdev_raidz_get_logical_width(vdrz, - BP_GET_BIRTH(zio->io_bp)); + BP_GET_PHYSICAL_BIRTH(zio->io_bp)); if (logical_width != vdrz->vd_physical_width) { zfs_locked_range_t *lr = NULL; uint64_t synced_offset = UINT64_MAX; @@ -2691,7 +2685,7 @@ raidz_checksum_verify(zio_t *zio) */ if (zio->io_flags & ZIO_FLAG_DIO_READ && ret == ECKSUM) { zio->io_error = ret; - zio->io_flags |= ZIO_FLAG_DIO_CHKSUM_ERR; + zio->io_post |= ZIO_POST_DIO_CHKSUM_ERR; zio_dio_chksum_verify_error_report(zio); zio_checksum_verified(zio); return (0); @@ -3048,7 +3042,7 @@ raidz_reconstruct(zio_t *zio, int *ltgts, int ntgts, int nparity) /* Check for success */ if (raidz_checksum_verify(zio) == 0) { - if (zio->io_flags & ZIO_FLAG_DIO_CHKSUM_ERR) + if (zio->io_post & ZIO_POST_DIO_CHKSUM_ERR) return (0); /* Reconstruction succeeded - report errors */ @@ -3369,7 +3363,7 @@ vdev_raidz_io_done_reconstruct_known_missing(zio_t *zio, raidz_map_t *rm, * also have been fewer parity errors than parity * columns or, again, we wouldn't be in this code path. */ - ASSERT(parity_untried == 0); + ASSERT0(parity_untried); ASSERT(parity_errors < rr->rr_firstdatacol); /* @@ -3514,7 +3508,7 @@ vdev_raidz_io_done(zio_t *zio) } if (raidz_checksum_verify(zio) == 0) { - if (zio->io_flags & ZIO_FLAG_DIO_CHKSUM_ERR) + if (zio->io_post & ZIO_POST_DIO_CHKSUM_ERR) goto done; for (int i = 0; i < rm->rm_nrows; i++) { @@ -4591,8 +4585,10 @@ spa_raidz_expand_thread(void *arg, zthr_t *zthr) uint64_t shift, start; zfs_range_seg_type_t type = metaslab_calculate_range_tree_type( raidvd, msp, &start, &shift); - zfs_range_tree_t *rt = zfs_range_tree_create(NULL, type, NULL, - start, shift); + zfs_range_tree_t *rt = zfs_range_tree_create_flags( + NULL, type, NULL, start, shift, ZFS_RT_F_DYN_NAME, + metaslab_rt_name(msp->ms_group, msp, + "spa_raidz_expand_thread:rt")); zfs_range_tree_add(rt, msp->ms_start, msp->ms_size); zfs_range_tree_walk(msp->ms_allocatable, zfs_range_tree_remove, rt); @@ -4747,7 +4743,7 @@ spa_raidz_expand_thread(void *arg, zthr_t *zthr) void spa_start_raidz_expansion_thread(spa_t *spa) { - ASSERT3P(spa->spa_raidz_expand_zthr, ==, NULL); + ASSERT0P(spa->spa_raidz_expand_zthr); spa->spa_raidz_expand_zthr = zthr_create("raidz_expand", spa_raidz_expand_thread_check, spa_raidz_expand_thread, spa, defclsyspri); diff --git a/sys/contrib/openzfs/module/zfs/vdev_rebuild.c b/sys/contrib/openzfs/module/zfs/vdev_rebuild.c index 0e296606d037..47b3b9921abe 100644 --- a/sys/contrib/openzfs/module/zfs/vdev_rebuild.c +++ b/sys/contrib/openzfs/module/zfs/vdev_rebuild.c @@ -256,7 +256,7 @@ vdev_rebuild_initiate_sync(void *arg, dmu_tx_t *tx) "vdev_id=%llu vdev_guid=%llu started", (u_longlong_t)vd->vdev_id, (u_longlong_t)vd->vdev_guid); - ASSERT3P(vd->vdev_rebuild_thread, ==, NULL); + ASSERT0P(vd->vdev_rebuild_thread); vd->vdev_rebuild_thread = thread_create(NULL, 0, vdev_rebuild_thread, vd, 0, &p0, TS_RUN, maxclsyspri); @@ -413,7 +413,7 @@ vdev_rebuild_reset_sync(void *arg, dmu_tx_t *tx) mutex_enter(&vd->vdev_rebuild_lock); ASSERT(vrp->vrp_rebuild_state == VDEV_REBUILD_ACTIVE); - ASSERT3P(vd->vdev_rebuild_thread, ==, NULL); + ASSERT0P(vd->vdev_rebuild_thread); vrp->vrp_last_offset = 0; vrp->vrp_min_txg = 0; @@ -787,8 +787,9 @@ vdev_rebuild_thread(void *arg) vdev_rebuild_phys_t *vrp = &vr->vr_rebuild_phys; vr->vr_top_vdev = vd; vr->vr_scan_msp = NULL; - vr->vr_scan_tree = zfs_range_tree_create(NULL, ZFS_RANGE_SEG64, NULL, - 0, 0); + vr->vr_scan_tree = zfs_range_tree_create_flags( + NULL, ZFS_RANGE_SEG64, NULL, 0, 0, + ZFS_RT_F_DYN_NAME, vdev_rt_name(vd, "vr_scan_tree")); mutex_init(&vr->vr_io_lock, NULL, MUTEX_DEFAULT, NULL); cv_init(&vr->vr_io_cv, NULL, CV_DEFAULT, NULL); diff --git a/sys/contrib/openzfs/module/zfs/vdev_removal.c b/sys/contrib/openzfs/module/zfs/vdev_removal.c index db79ded6dce4..2f7a739da241 100644 --- a/sys/contrib/openzfs/module/zfs/vdev_removal.c +++ b/sys/contrib/openzfs/module/zfs/vdev_removal.c @@ -344,10 +344,10 @@ spa_vdev_remove_aux(nvlist_t *config, const char *name, nvlist_t **dev, for (int i = 0, j = 0; i < count; i++) { if (dev[i] == dev_to_remove) continue; - VERIFY(nvlist_dup(dev[i], &newdev[j++], KM_SLEEP) == 0); + VERIFY0(nvlist_dup(dev[i], &newdev[j++], KM_SLEEP)); } - VERIFY(nvlist_remove(config, name, DATA_TYPE_NVLIST_ARRAY) == 0); + VERIFY0(nvlist_remove(config, name, DATA_TYPE_NVLIST_ARRAY)); fnvlist_add_nvlist_array(config, name, (const nvlist_t * const *)newdev, count - 1); @@ -364,13 +364,15 @@ spa_vdev_removal_create(vdev_t *vd) spa_vdev_removal_t *svr = kmem_zalloc(sizeof (*svr), KM_SLEEP); mutex_init(&svr->svr_lock, NULL, MUTEX_DEFAULT, NULL); cv_init(&svr->svr_cv, NULL, CV_DEFAULT, NULL); - svr->svr_allocd_segs = zfs_range_tree_create(NULL, ZFS_RANGE_SEG64, - NULL, 0, 0); + svr->svr_allocd_segs = zfs_range_tree_create_flags( + NULL, ZFS_RANGE_SEG64, NULL, 0, 0, + ZFS_RT_F_DYN_NAME, vdev_rt_name(vd, "svr_allocd_segs")); svr->svr_vdev_id = vd->vdev_id; for (int i = 0; i < TXG_SIZE; i++) { - svr->svr_frees[i] = zfs_range_tree_create(NULL, ZFS_RANGE_SEG64, - NULL, 0, 0); + svr->svr_frees[i] = zfs_range_tree_create_flags( + NULL, ZFS_RANGE_SEG64, NULL, 0, 0, + ZFS_RT_F_DYN_NAME, vdev_rt_name(vd, "svr_frees")); list_create(&svr->svr_new_segments[i], sizeof (vdev_indirect_mapping_entry_t), offsetof(vdev_indirect_mapping_entry_t, vime_node)); @@ -421,7 +423,7 @@ vdev_remove_initiate_sync(void *arg, dmu_tx_t *tx) svr = spa_vdev_removal_create(vd); ASSERT(vd->vdev_removing); - ASSERT3P(vd->vdev_indirect_mapping, ==, NULL); + ASSERT0P(vd->vdev_indirect_mapping); spa_feature_incr(spa, SPA_FEATURE_DEVICE_REMOVAL, tx); if (spa_feature_is_enabled(spa, SPA_FEATURE_OBSOLETE_COUNTS)) { @@ -527,7 +529,7 @@ vdev_remove_initiate_sync(void *arg, dmu_tx_t *tx) * but in any case only when there are outstanding free i/os, which * there are not). */ - ASSERT3P(spa->spa_vdev_removal, ==, NULL); + ASSERT0P(spa->spa_vdev_removal); spa->spa_vdev_removal = svr; svr->svr_thread = thread_create(NULL, 0, spa_vdev_remove_thread, spa, 0, &p0, TS_RUN, minclsyspri); @@ -1179,8 +1181,9 @@ spa_vdev_copy_segment(vdev_t *vd, zfs_range_tree_t *segs, * relative to the start of the range to be copied (i.e. relative to the * local variable "start"). */ - zfs_range_tree_t *obsolete_segs = zfs_range_tree_create(NULL, - ZFS_RANGE_SEG64, NULL, 0, 0); + zfs_range_tree_t *obsolete_segs = zfs_range_tree_create_flags( + NULL, ZFS_RANGE_SEG64, NULL, 0, 0, + ZFS_RT_F_DYN_NAME, vdev_rt_name(vd, "obsolete_segs")); zfs_btree_index_t where; zfs_range_seg_t *rs = zfs_btree_first(&segs->rt_root, &where); @@ -1359,11 +1362,11 @@ vdev_remove_complete(spa_t *spa) txg_wait_synced(spa->spa_dsl_pool, 0); txg = spa_vdev_enter(spa); vdev_t *vd = vdev_lookup_top(spa, spa->spa_vdev_removal->svr_vdev_id); - ASSERT3P(vd->vdev_initialize_thread, ==, NULL); - ASSERT3P(vd->vdev_trim_thread, ==, NULL); - ASSERT3P(vd->vdev_autotrim_thread, ==, NULL); + ASSERT0P(vd->vdev_initialize_thread); + ASSERT0P(vd->vdev_trim_thread); + ASSERT0P(vd->vdev_autotrim_thread); vdev_rebuild_stop_wait(vd); - ASSERT3P(vd->vdev_rebuild_thread, ==, NULL); + ASSERT0P(vd->vdev_rebuild_thread); sysevent_t *ev = spa_event_create(spa, vd, NULL, ESC_ZFS_VDEV_REMOVE_DEV); @@ -1448,8 +1451,9 @@ spa_vdev_copy_impl(vdev_t *vd, spa_vdev_removal_t *svr, vdev_copy_arg_t *vca, * allocated segments that we are copying. We may also be copying * free segments (of up to vdev_removal_max_span bytes). */ - zfs_range_tree_t *segs = zfs_range_tree_create(NULL, ZFS_RANGE_SEG64, - NULL, 0, 0); + zfs_range_tree_t *segs = zfs_range_tree_create_flags( + NULL, ZFS_RANGE_SEG64, NULL, 0, 0, + ZFS_RT_F_DYN_NAME, vdev_rt_name(vd, "spa_vdev_copy_impl:segs")); for (;;) { zfs_range_tree_t *rt = svr->svr_allocd_segs; zfs_range_seg_t *rs = zfs_range_tree_first(rt); @@ -1610,8 +1614,9 @@ spa_vdev_remove_thread(void *arg) vca.vca_read_error_bytes = 0; vca.vca_write_error_bytes = 0; - zfs_range_tree_t *segs = zfs_range_tree_create(NULL, ZFS_RANGE_SEG64, - NULL, 0, 0); + zfs_range_tree_t *segs = zfs_range_tree_create_flags( + NULL, ZFS_RANGE_SEG64, NULL, 0, 0, + ZFS_RT_F_DYN_NAME, vdev_rt_name(vd, "spa_vdev_remove_thread:segs")); mutex_enter(&svr->svr_lock); @@ -1863,7 +1868,7 @@ spa_vdev_remove_cancel_sync(void *arg, dmu_tx_t *tx) vdev_indirect_mapping_t *vim = vd->vdev_indirect_mapping; objset_t *mos = spa->spa_meta_objset; - ASSERT3P(svr->svr_thread, ==, NULL); + ASSERT0P(svr->svr_thread); spa_feature_decr(spa, SPA_FEATURE_DEVICE_REMOVAL, tx); @@ -1895,8 +1900,9 @@ spa_vdev_remove_cancel_sync(void *arg, dmu_tx_t *tx) vdev_indirect_mapping_max_offset(vim)); } - zfs_range_tree_t *segs = zfs_range_tree_create(NULL, ZFS_RANGE_SEG64, - NULL, 0, 0); + zfs_range_tree_t *segs = zfs_range_tree_create_flags( + NULL, ZFS_RANGE_SEG64, NULL, 0, 0, ZFS_RT_F_DYN_NAME, + vdev_rt_name(vd, "spa_vdev_remove_cancel_sync:segs")); for (uint64_t msi = 0; msi < vd->vdev_ms_count; msi++) { metaslab_t *msp = vd->vdev_ms[msi]; @@ -2070,7 +2076,7 @@ spa_vdev_remove_log(vdev_t *vd, uint64_t *txg) ASSERT(vd->vdev_islog); ASSERT(vd == vd->vdev_top); - ASSERT3P(vd->vdev_log_mg, ==, NULL); + ASSERT0P(vd->vdev_log_mg); ASSERT(MUTEX_HELD(&spa_namespace_lock)); /* @@ -2106,7 +2112,7 @@ spa_vdev_remove_log(vdev_t *vd, uint64_t *txg) if (error != 0) { metaslab_group_activate(mg); - ASSERT3P(vd->vdev_log_mg, ==, NULL); + ASSERT0P(vd->vdev_log_mg); return (error); } ASSERT0(vd->vdev_stat.vs_alloc); diff --git a/sys/contrib/openzfs/module/zfs/vdev_trim.c b/sys/contrib/openzfs/module/zfs/vdev_trim.c index 842bb3e690d4..eee18b367909 100644 --- a/sys/contrib/openzfs/module/zfs/vdev_trim.c +++ b/sys/contrib/openzfs/module/zfs/vdev_trim.c @@ -902,7 +902,9 @@ vdev_trim_thread(void *arg) ta.trim_vdev = vd; ta.trim_extent_bytes_max = zfs_trim_extent_bytes_max; ta.trim_extent_bytes_min = zfs_trim_extent_bytes_min; - ta.trim_tree = zfs_range_tree_create(NULL, ZFS_RANGE_SEG64, NULL, 0, 0); + ta.trim_tree = zfs_range_tree_create_flags( + NULL, ZFS_RANGE_SEG64, NULL, 0, 0, + ZFS_RT_F_DYN_NAME, vdev_rt_name(vd, "trim_tree")); ta.trim_type = TRIM_TYPE_MANUAL; ta.trim_flags = 0; @@ -1008,7 +1010,7 @@ vdev_trim(vdev_t *vd, uint64_t rate, boolean_t partial, boolean_t secure) ASSERT(MUTEX_HELD(&vd->vdev_trim_lock)); ASSERT(vd->vdev_ops->vdev_op_leaf); ASSERT(vdev_is_concrete(vd)); - ASSERT3P(vd->vdev_trim_thread, ==, NULL); + ASSERT0P(vd->vdev_trim_thread); ASSERT(!vd->vdev_detached); ASSERT(!vd->vdev_trim_exit_wanted); ASSERT(!vd->vdev_top->vdev_removing); @@ -1030,7 +1032,7 @@ vdev_trim_stop_wait_impl(vdev_t *vd) while (vd->vdev_trim_thread != NULL) cv_wait(&vd->vdev_trim_cv, &vd->vdev_trim_lock); - ASSERT3P(vd->vdev_trim_thread, ==, NULL); + ASSERT0P(vd->vdev_trim_thread); vd->vdev_trim_exit_wanted = B_FALSE; } @@ -1305,8 +1307,10 @@ vdev_autotrim_thread(void *arg) * Allocate an empty range tree which is swapped in * for the existing ms_trim tree while it is processed. */ - trim_tree = zfs_range_tree_create(NULL, ZFS_RANGE_SEG64, - NULL, 0, 0); + trim_tree = zfs_range_tree_create_flags( + NULL, ZFS_RANGE_SEG64, NULL, 0, 0, + ZFS_RT_F_DYN_NAME, + vdev_rt_name(vd, "autotrim_tree")); zfs_range_tree_swap(&msp->ms_trim, &trim_tree); ASSERT(zfs_range_tree_is_empty(msp->ms_trim)); @@ -1360,8 +1364,10 @@ vdev_autotrim_thread(void *arg) if (!cvd->vdev_ops->vdev_op_leaf) continue; - ta->trim_tree = zfs_range_tree_create(NULL, - ZFS_RANGE_SEG64, NULL, 0, 0); + ta->trim_tree = zfs_range_tree_create_flags( + NULL, ZFS_RANGE_SEG64, NULL, 0, 0, + ZFS_RT_F_DYN_NAME, + vdev_rt_name(vd, "autotrim_tree")); zfs_range_tree_walk(trim_tree, vdev_trim_range_add, ta); } @@ -1533,7 +1539,7 @@ vdev_autotrim_stop_wait(vdev_t *tvd) cv_wait(&tvd->vdev_autotrim_cv, &tvd->vdev_autotrim_lock); - ASSERT3P(tvd->vdev_autotrim_thread, ==, NULL); + ASSERT0P(tvd->vdev_autotrim_thread); tvd->vdev_autotrim_exit_wanted = B_FALSE; } mutex_exit(&tvd->vdev_autotrim_lock); @@ -1600,7 +1606,9 @@ vdev_trim_l2arc_thread(void *arg) vd->vdev_trim_secure = 0; ta.trim_vdev = vd; - ta.trim_tree = zfs_range_tree_create(NULL, ZFS_RANGE_SEG64, NULL, 0, 0); + ta.trim_tree = zfs_range_tree_create_flags( + NULL, ZFS_RANGE_SEG64, NULL, 0, 0, + ZFS_RT_F_DYN_NAME, vdev_rt_name(vd, "trim_tree")); ta.trim_type = TRIM_TYPE_MANUAL; ta.trim_extent_bytes_max = zfs_trim_extent_bytes_max; ta.trim_extent_bytes_min = SPA_MINBLOCKSIZE; @@ -1704,7 +1712,7 @@ vdev_trim_l2arc(spa_t *spa) mutex_enter(&vd->vdev_trim_lock); ASSERT(vd->vdev_ops->vdev_op_leaf); ASSERT(vdev_is_concrete(vd)); - ASSERT3P(vd->vdev_trim_thread, ==, NULL); + ASSERT0P(vd->vdev_trim_thread); ASSERT(!vd->vdev_detached); ASSERT(!vd->vdev_trim_exit_wanted); ASSERT(!vd->vdev_top->vdev_removing); @@ -1735,7 +1743,9 @@ vdev_trim_simple(vdev_t *vd, uint64_t start, uint64_t size) ASSERT(!vd->vdev_top->vdev_rz_expanding); ta.trim_vdev = vd; - ta.trim_tree = zfs_range_tree_create(NULL, ZFS_RANGE_SEG64, NULL, 0, 0); + ta.trim_tree = zfs_range_tree_create_flags( + NULL, ZFS_RANGE_SEG64, NULL, 0, 0, + ZFS_RT_F_DYN_NAME, vdev_rt_name(vd, "trim_tree")); ta.trim_type = TRIM_TYPE_SIMPLE; ta.trim_extent_bytes_max = zfs_trim_extent_bytes_max; ta.trim_extent_bytes_min = SPA_MINBLOCKSIZE; diff --git a/sys/contrib/openzfs/module/zfs/zap.c b/sys/contrib/openzfs/module/zfs/zap.c index 9711c91d7e4e..3e4e997798a3 100644 --- a/sys/contrib/openzfs/module/zfs/zap.c +++ b/sys/contrib/openzfs/module/zfs/zap.c @@ -921,7 +921,7 @@ fzap_add_cd(zap_name_t *zn, ASSERT(RW_LOCK_HELD(&zap->zap_rwlock)); ASSERT(!zap->zap_ismicro); - ASSERT(fzap_check(zn, integer_size, num_integers) == 0); + ASSERT0(fzap_check(zn, integer_size, num_integers)); err = zap_deref_leaf(zap, zn->zn_hash, tx, RW_WRITER, &l); if (err != 0) @@ -1304,7 +1304,7 @@ zap_increment_int(objset_t *os, uint64_t obj, uint64_t key, int64_t delta, int fzap_cursor_retrieve(zap_t *zap, zap_cursor_t *zc, zap_attribute_t *za) { - int err = ENOENT; + int err; zap_entry_handle_t zeh; zap_leaf_t *l; @@ -1386,7 +1386,7 @@ again: } err = zap_entry_read_name(zap, &zeh, za->za_name_len, za->za_name); - ASSERT(err == 0); + ASSERT0(err); za->za_normalization_conflict = zap_entry_normalization_conflict(&zeh, @@ -1546,7 +1546,7 @@ zap_shrink(zap_name_t *zn, zap_leaf_t *l, dmu_tx_t *tx) boolean_t trunc = B_FALSE; int err = 0; - ASSERT3U(zap_leaf_phys(l)->l_hdr.lh_nentries, ==, 0); + ASSERT0(zap_leaf_phys(l)->l_hdr.lh_nentries); ASSERT3U(prefix_len, <=, zap_f_phys(zap)->zap_ptrtbl.zt_shift); ASSERT(RW_LOCK_HELD(&zap->zap_rwlock)); ASSERT3U(ZAP_HASH_IDX(hash, prefix_len), ==, prefix); @@ -1564,7 +1564,7 @@ zap_shrink(zap_name_t *zn, zap_leaf_t *l, dmu_tx_t *tx) uint64_t sl_hash = ZAP_PREFIX_HASH(sl_prefix, prefix_len); int slbit = prefix & 1; - ASSERT3U(zap_leaf_phys(l)->l_hdr.lh_nentries, ==, 0); + ASSERT0(zap_leaf_phys(l)->l_hdr.lh_nentries); /* * Check if there is a sibling by reading ptrtbl ptrs. diff --git a/sys/contrib/openzfs/module/zfs/zap_micro.c b/sys/contrib/openzfs/module/zfs/zap_micro.c index 411b1a9db5ab..ea4e3117a8b9 100644 --- a/sys/contrib/openzfs/module/zfs/zap_micro.c +++ b/sys/contrib/openzfs/module/zfs/zap_micro.c @@ -346,7 +346,7 @@ zap_name_alloc_uint64(zap_t *zap, const uint64_t *key, int numints) { zap_name_t *zn = kmem_cache_alloc(zap_name_cache, KM_SLEEP); - ASSERT(zap->zap_normflags == 0); + ASSERT0(zap->zap_normflags); zn->zn_zap = zap; zn->zn_key_intlen = sizeof (*key); zn->zn_key_orig = zn->zn_key_norm = key; @@ -1876,7 +1876,7 @@ zap_cursor_serialize(zap_cursor_t *zc) return (-1ULL); if (zc->zc_zap == NULL) return (zc->zc_serialized); - ASSERT((zc->zc_hash & zap_maxcd(zc->zc_zap)) == 0); + ASSERT0((zc->zc_hash & zap_maxcd(zc->zc_zap))); ASSERT(zc->zc_cd < zap_maxcd(zc->zc_zap)); /* @@ -1911,7 +1911,7 @@ zap_cursor_retrieve(zap_cursor_t *zc, zap_attribute_t *za) * we must add to the existing zc_cd, which may already * be 1 due to the zap_cursor_advance. */ - ASSERT(zc->zc_hash == 0); + ASSERT0(zc->zc_hash); hb = zap_hashbits(zc->zc_zap); zc->zc_hash = zc->zc_serialized << (64 - hb); zc->zc_cd += zc->zc_serialized >> hb; diff --git a/sys/contrib/openzfs/module/zfs/zcp.c b/sys/contrib/openzfs/module/zfs/zcp.c index 6960ea360b15..c6684f453e95 100644 --- a/sys/contrib/openzfs/module/zfs/zcp.c +++ b/sys/contrib/openzfs/module/zfs/zcp.c @@ -765,7 +765,7 @@ zcp_lua_alloc(void *ud, void *ptr, size_t osize, size_t nsize) return (NULL); } (void) memcpy(luabuf, ptr, osize); - VERIFY3P(zcp_lua_alloc(ud, ptr, osize, 0), ==, NULL); + VERIFY0P(zcp_lua_alloc(ud, ptr, osize, 0)); return (luabuf); } } @@ -1175,7 +1175,7 @@ zcp_eval(const char *poolname, const char *program, boolean_t sync, for (nvpair_t *pair = nvlist_next_nvpair(runinfo.zri_new_zvols, NULL); pair != NULL; pair = nvlist_next_nvpair(runinfo.zri_new_zvols, pair)) { - zvol_create_minor(nvpair_name(pair)); + zvol_create_minors(nvpair_name(pair)); } fnvlist_free(runinfo.zri_new_zvols); diff --git a/sys/contrib/openzfs/module/zfs/zfeature.c b/sys/contrib/openzfs/module/zfs/zfeature.c index 7dfe00d42a08..0816ea134bf3 100644 --- a/sys/contrib/openzfs/module/zfs/zfeature.c +++ b/sys/contrib/openzfs/module/zfs/zfeature.c @@ -210,8 +210,8 @@ spa_features_check(spa_t *spa, boolean_t for_write, za->za_name, 1, MAXPATHLEN, buf) == 0) desc = buf; - VERIFY(nvlist_add_string(unsup_feat, - za->za_name, desc) == 0); + VERIFY0(nvlist_add_string(unsup_feat, + za->za_name, desc)); } } } diff --git a/sys/contrib/openzfs/module/zfs/zfs_chksum.c b/sys/contrib/openzfs/module/zfs/zfs_chksum.c index 5c92be21c0c8..21852bf3d865 100644 --- a/sys/contrib/openzfs/module/zfs/zfs_chksum.c +++ b/sys/contrib/openzfs/module/zfs/zfs_chksum.c @@ -32,9 +32,6 @@ #include <sys/blake3.h> #include <sys/sha2.h> -/* limit benchmarking to max 256KiB, when EdonR is slower then this: */ -#define LIMIT_PERF_MBS 300 - typedef struct { const char *name; const char *impl; @@ -52,9 +49,15 @@ typedef struct { zio_checksum_tmpl_free_t *(free); } chksum_stat_t; +#define AT_STARTUP 0 +#define AT_BENCHMARK 1 +#define AT_DONE 2 + static chksum_stat_t *chksum_stat_data = 0; -static int chksum_stat_cnt = 0; static kstat_t *chksum_kstat = NULL; +static int chksum_stat_limit = AT_STARTUP; +static int chksum_stat_cnt = 0; +static void chksum_benchmark(void); /* * Sample output on i3-1005G1 System: @@ -129,6 +132,9 @@ chksum_kstat_data(char *buf, size_t size, void *data) static void * chksum_kstat_addr(kstat_t *ksp, loff_t n) { + /* full benchmark */ + chksum_benchmark(); + if (n < chksum_stat_cnt) ksp->ks_private = (void *)(chksum_stat_data + n); else @@ -176,47 +182,36 @@ chksum_run(chksum_stat_t *cs, abd_t *abd, void *ctx, int round, kpreempt_enable(); run_bw = size * run_count * NANOSEC; - run_bw /= run_time_ns; /* B/s */ + run_bw /= run_time_ns; /* B/s */ *result = run_bw/1024/1024; /* MiB/s */ } -#define LIMIT_INIT 0 -#define LIMIT_NEEDED 1 -#define LIMIT_NOLIMIT 2 - static void chksum_benchit(chksum_stat_t *cs) { abd_t *abd; void *ctx = 0; void *salt = &cs->salt.zcs_bytes; - static int chksum_stat_limit = LIMIT_INIT; memset(salt, 0, sizeof (cs->salt.zcs_bytes)); if (cs->init) ctx = cs->init(&cs->salt); + /* benchmarks in startup mode */ + if (chksum_stat_limit == AT_STARTUP) { + abd = abd_alloc_linear(1<<18, B_FALSE); + chksum_run(cs, abd, ctx, 5, &cs->bs256k); + goto done; + } + /* allocate test memory via abd linear interface */ abd = abd_alloc_linear(1<<20, B_FALSE); + + /* benchmarks when requested */ chksum_run(cs, abd, ctx, 1, &cs->bs1k); chksum_run(cs, abd, ctx, 2, &cs->bs4k); chksum_run(cs, abd, ctx, 3, &cs->bs16k); chksum_run(cs, abd, ctx, 4, &cs->bs64k); - chksum_run(cs, abd, ctx, 5, &cs->bs256k); - - /* check if we ran on a slow cpu */ - if (chksum_stat_limit == LIMIT_INIT) { - if (cs->bs1k < LIMIT_PERF_MBS) { - chksum_stat_limit = LIMIT_NEEDED; - } else { - chksum_stat_limit = LIMIT_NOLIMIT; - } - } - - /* skip benchmarks >= 1MiB when the CPU is to slow */ - if (chksum_stat_limit == LIMIT_NEEDED) - goto abort; - chksum_run(cs, abd, ctx, 6, &cs->bs1m); abd_free(abd); @@ -225,7 +220,7 @@ chksum_benchit(chksum_stat_t *cs) chksum_run(cs, abd, ctx, 7, &cs->bs4m); chksum_run(cs, abd, ctx, 8, &cs->bs16m); -abort: +done: abd_free(abd); /* free up temp memory */ @@ -243,7 +238,6 @@ chksum_benchmark(void) /* we need the benchmark only for the kernel module */ return; #endif - chksum_stat_t *cs; uint64_t max; uint32_t id, cbid = 0, id_save; @@ -251,8 +245,14 @@ chksum_benchmark(void) const zfs_impl_t *sha256 = zfs_impl_get_ops("sha256"); const zfs_impl_t *sha512 = zfs_impl_get_ops("sha512"); + /* benchmarks are done */ + if (chksum_stat_limit == AT_DONE) + return; + + /* count implementations */ - chksum_stat_cnt = 2; + chksum_stat_cnt = 1; /* edonr */ + chksum_stat_cnt += 1; /* skein */ chksum_stat_cnt += sha256->getcnt(); chksum_stat_cnt += sha512->getcnt(); chksum_stat_cnt += blake3->getcnt(); @@ -332,6 +332,17 @@ chksum_benchmark(void) } } blake3->setid(id_save); + + switch (chksum_stat_limit) { + case AT_STARTUP: + /* next time we want a full benchmark */ + chksum_stat_limit = AT_BENCHMARK; + break; + case AT_BENCHMARK: + /* no further benchmarks */ + chksum_stat_limit = AT_DONE; + break; + } } void @@ -341,7 +352,7 @@ chksum_init(void) blake3_per_cpu_ctx_init(); #endif - /* Benchmark supported implementations */ + /* 256KiB benchmark */ chksum_benchmark(); /* Install kstats for all implementations */ diff --git a/sys/contrib/openzfs/module/zfs/zfs_crrd.c b/sys/contrib/openzfs/module/zfs/zfs_crrd.c new file mode 100644 index 000000000000..f9267ed41d71 --- /dev/null +++ b/sys/contrib/openzfs/module/zfs/zfs_crrd.c @@ -0,0 +1,227 @@ +// SPDX-License-Identifier: CDDL-1.0 +/* + * CDDL HEADER START + * + * The contents of this file are subject to the terms of the + * Common Development and Distribution License (the "License"). + * You may not use this file except in compliance with the License. + * + * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE + * or https://opensource.org/licenses/CDDL-1.0. + * See the License for the specific language governing permissions + * and limitations under the License. + * + * When distributing Covered Code, include this CDDL HEADER in each + * file and include the License file at usr/src/OPENSOLARIS.LICENSE. + * If applicable, add the following below this CDDL HEADER, with the + * fields enclosed by brackets "[]" replaced with your own identifying + * information: Portions Copyright [yyyy] [name of copyright owner] + * + * CDDL HEADER END + */ +/* + * Copyright (c) 2024 Klara Inc. + * + * This software was developed by + * Mariusz Zaborski <mariusz.zaborski@klarasystems.com> + * Fred Weigel <fred.weigel@klarasystems.com> + * under sponsorship from Wasabi Technology, Inc. and Klara Inc. + */ +/* + * This file implements a round-robin database that stores timestamps and txg + * numbers. Due to limited space, we use a round-robin approach, where + * the oldest records are overwritten when there is no longer enough room. + * This is a best-effort mechanism, and the database should be treated as + * an approximation. Consider this before consuming it. + * + * The database is linear, meaning we assume each new entry is newer than the + * ones already stored. Because of this, if time is manipulated, the database + * will only accept records that are newer than the existing ones. + * (For example, jumping 10 years into the future and then back can lead to + * situation when for 10 years we wont write anything to database) + * + * All times stored in the database use UTC, which makes it easy to convert to + * and from local time. + * + * Each database holds 256 records (as defined in the `RRD_MAX_ENTRIES` macro). + * This limit comes from the maximum size of a ZAP object, where we store the + * binary blob. + * + * We've split the database into three smaller ones. + * The `minute database` provides high resolution (default: every 10 minutes), + * but only covers approximately 1.5 days. This gives a detailed view of recent + * activity, useful, for example, when performing a scrub of the last hour. + * The `daily database` records one txg per day. With 256 entries, it retains + * roughly 8 months of data. This allows users to scrub or analyze txgs across + * a range of days. + * The `monthly database` stores one record per month, giving approximately + * 21 years of history. + * All these calculations assume the worst-case scenario: the pool is always + * online and actively written to. + * + * A potential source of confusion is that the database does not store data + * while the pool is offline, leading to potential gaps in timeline. Also, + * the database contains no records from before this feature was enabled. + * Both, upon reflection, are expected. + */ +#include <sys/zfs_context.h> + +#include "zfs_crrd.h" + +rrd_data_t * +rrd_tail_entry(rrd_t *rrd) +{ + size_t n; + + if (rrd_len(rrd) == 0) + return (NULL); + + if (rrd->rrd_tail == 0) + n = RRD_MAX_ENTRIES - 1; + else + n = rrd->rrd_tail - 1; + + return (&rrd->rrd_entries[n]); +} + +uint64_t +rrd_tail(rrd_t *rrd) +{ + const rrd_data_t *tail; + + tail = rrd_tail_entry(rrd); + + return (tail == NULL ? 0 : tail->rrdd_time); +} + +/* + * Return length of data in the rrd. + * rrd_get works from 0..rrd_len()-1. + */ +size_t +rrd_len(rrd_t *rrd) +{ + + return (rrd->rrd_length); +} + +const rrd_data_t * +rrd_entry(rrd_t *rrd, size_t i) +{ + size_t n; + + if (i >= rrd_len(rrd)) { + return (0); + } + + n = (rrd->rrd_head + i) % RRD_MAX_ENTRIES; + return (&rrd->rrd_entries[n]); +} + +uint64_t +rrd_get(rrd_t *rrd, size_t i) +{ + const rrd_data_t *data = rrd_entry(rrd, i); + + return (data == NULL ? 0 : data->rrdd_txg); +} + +/* Add value to database. */ +void +rrd_add(rrd_t *rrd, hrtime_t time, uint64_t txg) +{ + rrd_data_t *tail; + + tail = rrd_tail_entry(rrd); + if (tail != NULL && tail->rrdd_time == time) { + if (tail->rrdd_txg < txg) { + tail->rrdd_txg = txg; + } else { + return; + } + } + + rrd->rrd_entries[rrd->rrd_tail].rrdd_time = time; + rrd->rrd_entries[rrd->rrd_tail].rrdd_txg = txg; + + rrd->rrd_tail = (rrd->rrd_tail + 1) % RRD_MAX_ENTRIES; + + if (rrd->rrd_length < RRD_MAX_ENTRIES) { + rrd->rrd_length++; + } else { + rrd->rrd_head = (rrd->rrd_head + 1) % RRD_MAX_ENTRIES; + } +} + +void +dbrrd_add(dbrrd_t *db, hrtime_t time, uint64_t txg) +{ + hrtime_t daydiff, monthdiff, minutedif; + + minutedif = time - rrd_tail(&db->dbr_minutes); + daydiff = time - rrd_tail(&db->dbr_days); + monthdiff = time - rrd_tail(&db->dbr_months); + + if (monthdiff >= 0 && monthdiff >= SEC2NSEC(30 * 24 * 60 * 60)) + rrd_add(&db->dbr_months, time, txg); + else if (daydiff >= 0 && daydiff >= SEC2NSEC(24 * 60 * 60)) + rrd_add(&db->dbr_days, time, txg); + else if (minutedif >= 0) + rrd_add(&db->dbr_minutes, time, txg); +} + +/* + * We could do a binary search here, but the routine isn't frequently + * called and the data is small so we stick to a simple loop. + */ +static const rrd_data_t * +rrd_query(rrd_t *rrd, hrtime_t tv, dbrrd_rounding_t rounding) +{ + const rrd_data_t *data = NULL; + + for (size_t i = 0; i < rrd_len(rrd); i++) { + const rrd_data_t *cur = rrd_entry(rrd, i); + + if (rounding == DBRRD_FLOOR) { + if (tv < cur->rrdd_time) { + break; + } + data = cur; + } else { + /* DBRRD_CEILING */ + if (tv <= cur->rrdd_time) { + data = cur; + break; + } + } + } + + return (data); +} + +static const rrd_data_t * +dbrrd_closest(hrtime_t tv, const rrd_data_t *r1, const rrd_data_t *r2) +{ + + if (r1 == NULL) + return (r2); + if (r2 == NULL) + return (r1); + + return (ABS(tv - r1->rrdd_time) < ABS(tv - r2->rrdd_time) ? r1 : r2); +} + +uint64_t +dbrrd_query(dbrrd_t *r, hrtime_t tv, dbrrd_rounding_t rounding) +{ + const rrd_data_t *data, *dm, *dd, *dy; + + data = NULL; + dm = rrd_query(&r->dbr_minutes, tv, rounding); + dd = rrd_query(&r->dbr_days, tv, rounding); + dy = rrd_query(&r->dbr_months, tv, rounding); + + data = dbrrd_closest(tv, dbrrd_closest(tv, dd, dm), dy); + + return (data == NULL ? 0 : data->rrdd_txg); +} diff --git a/sys/contrib/openzfs/module/zfs/zfs_fuid.c b/sys/contrib/openzfs/module/zfs/zfs_fuid.c index 10a6d289fbf8..2af1efe82e62 100644 --- a/sys/contrib/openzfs/module/zfs/zfs_fuid.c +++ b/sys/contrib/openzfs/module/zfs/zfs_fuid.c @@ -112,8 +112,7 @@ zfs_fuid_table_load(objset_t *os, uint64_t fuid_obj, avl_tree_t *idx_tree, uint64_t fuid_size; ASSERT(fuid_obj != 0); - VERIFY(0 == dmu_bonus_hold(os, fuid_obj, - FTAG, &db)); + VERIFY0(dmu_bonus_hold(os, fuid_obj, FTAG, &db)); fuid_size = *(uint64_t *)db->db_data; dmu_buf_rele(db, FTAG); @@ -125,22 +124,21 @@ zfs_fuid_table_load(objset_t *os, uint64_t fuid_obj, avl_tree_t *idx_tree, int i; packed = kmem_alloc(fuid_size, KM_SLEEP); - VERIFY(dmu_read(os, fuid_obj, 0, - fuid_size, packed, DMU_READ_PREFETCH) == 0); - VERIFY(nvlist_unpack(packed, fuid_size, - &nvp, 0) == 0); - VERIFY(nvlist_lookup_nvlist_array(nvp, FUID_NVP_ARRAY, - &fuidnvp, &count) == 0); + VERIFY0(dmu_read(os, fuid_obj, 0, + fuid_size, packed, DMU_READ_PREFETCH)); + VERIFY0(nvlist_unpack(packed, fuid_size, &nvp, 0)); + VERIFY0(nvlist_lookup_nvlist_array(nvp, FUID_NVP_ARRAY, + &fuidnvp, &count)); for (i = 0; i != count; i++) { fuid_domain_t *domnode; const char *domain; uint64_t idx; - VERIFY(nvlist_lookup_string(fuidnvp[i], FUID_DOMAIN, - &domain) == 0); - VERIFY(nvlist_lookup_uint64(fuidnvp[i], FUID_IDX, - &idx) == 0); + VERIFY0(nvlist_lookup_string(fuidnvp[i], FUID_DOMAIN, + &domain)); + VERIFY0(nvlist_lookup_uint64(fuidnvp[i], FUID_IDX, + &idx)); domnode = kmem_alloc(sizeof (fuid_domain_t), KM_SLEEP); @@ -246,35 +244,33 @@ zfs_fuid_sync(zfsvfs_t *zfsvfs, dmu_tx_t *tx) &zfsvfs->z_fuid_obj, tx) == 0); } - VERIFY(nvlist_alloc(&nvp, NV_UNIQUE_NAME, KM_SLEEP) == 0); + VERIFY0(nvlist_alloc(&nvp, NV_UNIQUE_NAME, KM_SLEEP)); numnodes = avl_numnodes(&zfsvfs->z_fuid_idx); fuids = kmem_alloc(numnodes * sizeof (void *), KM_SLEEP); for (i = 0, domnode = avl_first(&zfsvfs->z_fuid_domain); domnode; i++, domnode = AVL_NEXT(&zfsvfs->z_fuid_domain, domnode)) { - VERIFY(nvlist_alloc(&fuids[i], NV_UNIQUE_NAME, KM_SLEEP) == 0); - VERIFY(nvlist_add_uint64(fuids[i], FUID_IDX, - domnode->f_idx) == 0); - VERIFY(nvlist_add_uint64(fuids[i], FUID_OFFSET, 0) == 0); - VERIFY(nvlist_add_string(fuids[i], FUID_DOMAIN, - domnode->f_ksid->kd_name) == 0); + VERIFY0(nvlist_alloc(&fuids[i], NV_UNIQUE_NAME, KM_SLEEP)); + VERIFY0(nvlist_add_uint64(fuids[i], FUID_IDX, + domnode->f_idx)); + VERIFY0(nvlist_add_uint64(fuids[i], FUID_OFFSET, 0)); + VERIFY0(nvlist_add_string(fuids[i], FUID_DOMAIN, + domnode->f_ksid->kd_name)); } fnvlist_add_nvlist_array(nvp, FUID_NVP_ARRAY, (const nvlist_t * const *)fuids, numnodes); for (i = 0; i != numnodes; i++) nvlist_free(fuids[i]); kmem_free(fuids, numnodes * sizeof (void *)); - VERIFY(nvlist_size(nvp, &nvsize, NV_ENCODE_XDR) == 0); + VERIFY0(nvlist_size(nvp, &nvsize, NV_ENCODE_XDR)); packed = kmem_alloc(nvsize, KM_SLEEP); - VERIFY(nvlist_pack(nvp, &packed, &nvsize, - NV_ENCODE_XDR, KM_SLEEP) == 0); + VERIFY0(nvlist_pack(nvp, &packed, &nvsize, NV_ENCODE_XDR, KM_SLEEP)); nvlist_free(nvp); zfsvfs->z_fuid_size = nvsize; dmu_write(zfsvfs->z_os, zfsvfs->z_fuid_obj, 0, zfsvfs->z_fuid_size, packed, tx); kmem_free(packed, zfsvfs->z_fuid_size); - VERIFY(0 == dmu_bonus_hold(zfsvfs->z_os, zfsvfs->z_fuid_obj, - FTAG, &db)); + VERIFY0(dmu_bonus_hold(zfsvfs->z_os, zfsvfs->z_fuid_obj, FTAG, &db)); dmu_buf_will_dirty(db, tx); *(uint64_t *)db->db_data = zfsvfs->z_fuid_size; dmu_buf_rele(db, FTAG); diff --git a/sys/contrib/openzfs/module/zfs/zfs_ioctl.c b/sys/contrib/openzfs/module/zfs/zfs_ioctl.c index ebb1cfd07125..121b966b9864 100644 --- a/sys/contrib/openzfs/module/zfs/zfs_ioctl.c +++ b/sys/contrib/openzfs/module/zfs/zfs_ioctl.c @@ -1493,7 +1493,7 @@ zfs_ioc_pool_create(zfs_cmd_t *zc) goto pool_props_bad; (void) nvlist_remove_all(props, ZPOOL_HIDDEN_ARGS); - VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0); + VERIFY0(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP)); error = zfs_fill_zplprops_root(version, rootprops, zplprops, NULL); if (error != 0) @@ -1704,6 +1704,8 @@ zfs_ioc_pool_scan(zfs_cmd_t *zc) static const zfs_ioc_key_t zfs_keys_pool_scrub[] = { {"scan_type", DATA_TYPE_UINT64, 0}, {"scan_command", DATA_TYPE_UINT64, 0}, + {"scan_date_start", DATA_TYPE_UINT64, ZK_OPTIONAL}, + {"scan_date_end", DATA_TYPE_UINT64, ZK_OPTIONAL}, }; static int @@ -1712,6 +1714,7 @@ zfs_ioc_pool_scrub(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) spa_t *spa; int error; uint64_t scan_type, scan_cmd; + uint64_t date_start, date_end; if (nvlist_lookup_uint64(innvl, "scan_type", &scan_type) != 0) return (SET_ERROR(EINVAL)); @@ -1721,6 +1724,11 @@ zfs_ioc_pool_scrub(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) if (scan_cmd >= POOL_SCRUB_FLAGS_END) return (SET_ERROR(EINVAL)); + if (nvlist_lookup_uint64(innvl, "scan_date_start", &date_start) != 0) + date_start = 0; + if (nvlist_lookup_uint64(innvl, "scan_date_end", &date_end) != 0) + date_end = 0; + if ((error = spa_open(poolname, &spa, FTAG)) != 0) return (error); @@ -1732,7 +1740,24 @@ zfs_ioc_pool_scrub(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl) error = spa_scan_range(spa, scan_type, spa_get_last_scrubbed_txg(spa), 0); } else { - error = spa_scan(spa, scan_type); + uint64_t txg_start, txg_end; + + txg_start = txg_end = 0; + if (date_start != 0 || date_end != 0) { + mutex_enter(&spa->spa_txg_log_time_lock); + if (date_start != 0) { + txg_start = dbrrd_query(&spa->spa_txg_log_time, + date_start, DBRRD_FLOOR); + } + + if (date_end != 0) { + txg_end = dbrrd_query(&spa->spa_txg_log_time, + date_end, DBRRD_CEILING); + } + mutex_exit(&spa->spa_txg_log_time_lock); + } + + error = spa_scan_range(spa, scan_type, txg_start, txg_end); } spa_close(spa, FTAG); @@ -2220,7 +2245,7 @@ nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop) */ if ((error = zfs_get_zplprop(os, prop, &value)) != 0) return (error); - VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0); + VERIFY0(nvlist_add_uint64(props, zfs_prop_to_name(prop), value)); return (0); } @@ -2255,7 +2280,7 @@ zfs_ioc_objset_zplprops(zfs_cmd_t *zc) dmu_objset_type(os) == DMU_OST_ZFS) { nvlist_t *nv; - VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0); + VERIFY0(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP)); if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 && (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 && (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 && @@ -2458,7 +2483,7 @@ zfs_prop_set_userquota(const char *dsname, nvpair_t *pair) if (nvpair_type(pair) == DATA_TYPE_NVLIST) { nvlist_t *attrs; - VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); + VERIFY0(nvpair_value_nvlist(pair, &attrs)); if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE, &pair) != 0) return (SET_ERROR(EINVAL)); @@ -2513,9 +2538,8 @@ zfs_prop_set_special(const char *dsname, zprop_source_t source, if (nvpair_type(pair) == DATA_TYPE_NVLIST) { nvlist_t *attrs; - VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); - VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, - &pair) == 0); + VERIFY0(nvpair_value_nvlist(pair, &attrs)); + VERIFY0(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, &pair)); } /* all special properties are numeric except for keylocation */ @@ -2907,14 +2931,14 @@ props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops) { nvpair_t *pair; - VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0); + VERIFY0(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP)); pair = NULL; while ((pair = nvlist_next_nvpair(props, pair)) != NULL) { if (nvlist_exists(skipped, nvpair_name(pair))) continue; - VERIFY(nvlist_add_nvpair(*newprops, pair) == 0); + VERIFY0(nvlist_add_nvpair(*newprops, pair)); } } @@ -3039,11 +3063,11 @@ zfs_ioc_inherit_prop(zfs_cmd_t *zc) switch (type) { case PROP_TYPE_STRING: - VERIFY(0 == nvlist_add_string(dummy, propname, "")); + VERIFY0(nvlist_add_string(dummy, propname, "")); break; case PROP_TYPE_NUMBER: case PROP_TYPE_INDEX: - VERIFY(0 == nvlist_add_uint64(dummy, propname, 0)); + VERIFY0(nvlist_add_uint64(dummy, propname, 0)); break; default: err = SET_ERROR(EINVAL); @@ -3429,14 +3453,14 @@ zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver, /* * Put the version in the zplprops */ - VERIFY(nvlist_add_uint64(zplprops, - zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0); + VERIFY0(nvlist_add_uint64(zplprops, + zfs_prop_to_name(ZFS_PROP_VERSION), zplver)); if (norm == ZFS_PROP_UNDEFINED && (error = zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm)) != 0) return (error); - VERIFY(nvlist_add_uint64(zplprops, - zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0); + VERIFY0(nvlist_add_uint64(zplprops, + zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm)); /* * If we're normalizing, names must always be valid UTF-8 strings. @@ -3446,55 +3470,55 @@ zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver, if (u8 == ZFS_PROP_UNDEFINED && (error = zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8)) != 0) return (error); - VERIFY(nvlist_add_uint64(zplprops, - zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0); + VERIFY0(nvlist_add_uint64(zplprops, + zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8)); if (sense == ZFS_PROP_UNDEFINED && (error = zfs_get_zplprop(os, ZFS_PROP_CASE, &sense)) != 0) return (error); - VERIFY(nvlist_add_uint64(zplprops, - zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0); + VERIFY0(nvlist_add_uint64(zplprops, + zfs_prop_to_name(ZFS_PROP_CASE), sense)); if (duq == ZFS_PROP_UNDEFINED && (error = zfs_get_zplprop(os, ZFS_PROP_DEFAULTUSERQUOTA, &duq)) != 0) return (error); - VERIFY(nvlist_add_uint64(zplprops, - zfs_prop_to_name(ZFS_PROP_DEFAULTUSERQUOTA), duq) == 0); + VERIFY0(nvlist_add_uint64(zplprops, + zfs_prop_to_name(ZFS_PROP_DEFAULTUSERQUOTA), duq)); if (dgq == ZFS_PROP_UNDEFINED && (error = zfs_get_zplprop(os, ZFS_PROP_DEFAULTGROUPQUOTA, &dgq)) != 0) return (error); - VERIFY(nvlist_add_uint64(zplprops, - zfs_prop_to_name(ZFS_PROP_DEFAULTGROUPQUOTA), dgq) == 0); + VERIFY0(nvlist_add_uint64(zplprops, + zfs_prop_to_name(ZFS_PROP_DEFAULTGROUPQUOTA), dgq)); if (dpq == ZFS_PROP_UNDEFINED && (error = zfs_get_zplprop(os, ZFS_PROP_DEFAULTPROJECTQUOTA, &dpq)) != 0) return (error); - VERIFY(nvlist_add_uint64(zplprops, - zfs_prop_to_name(ZFS_PROP_DEFAULTPROJECTQUOTA), dpq) == 0); + VERIFY0(nvlist_add_uint64(zplprops, + zfs_prop_to_name(ZFS_PROP_DEFAULTPROJECTQUOTA), dpq)); if (duoq == ZFS_PROP_UNDEFINED && (error = zfs_get_zplprop(os, ZFS_PROP_DEFAULTUSEROBJQUOTA, &duoq)) != 0) return (error); - VERIFY(nvlist_add_uint64(zplprops, - zfs_prop_to_name(ZFS_PROP_DEFAULTUSEROBJQUOTA), duoq) == 0); + VERIFY0(nvlist_add_uint64(zplprops, + zfs_prop_to_name(ZFS_PROP_DEFAULTUSEROBJQUOTA), duoq)); if (dgoq == ZFS_PROP_UNDEFINED && (error = zfs_get_zplprop(os, ZFS_PROP_DEFAULTGROUPOBJQUOTA, &dgoq)) != 0) return (error); - VERIFY(nvlist_add_uint64(zplprops, - zfs_prop_to_name(ZFS_PROP_DEFAULTGROUPOBJQUOTA), dgoq) == 0); + VERIFY0(nvlist_add_uint64(zplprops, + zfs_prop_to_name(ZFS_PROP_DEFAULTGROUPOBJQUOTA), dgoq)); if (dpoq == ZFS_PROP_UNDEFINED && (error = zfs_get_zplprop(os, ZFS_PROP_DEFAULTPROJECTOBJQUOTA, &dpoq)) != 0) return (error); - VERIFY(nvlist_add_uint64(zplprops, - zfs_prop_to_name(ZFS_PROP_DEFAULTPROJECTOBJQUOTA), dpoq) == 0); + VERIFY0(nvlist_add_uint64(zplprops, + zfs_prop_to_name(ZFS_PROP_DEFAULTPROJECTOBJQUOTA), dpoq)); if (is_ci) *is_ci = (sense == ZFS_CASE_INSENSITIVE); @@ -3643,8 +3667,8 @@ zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) * file system creation, so go figure them out * now. */ - VERIFY(nvlist_alloc(&zct.zct_zplprops, - NV_UNIQUE_NAME, KM_SLEEP) == 0); + VERIFY0(nvlist_alloc(&zct.zct_zplprops, + NV_UNIQUE_NAME, KM_SLEEP)); error = zfs_fill_zplprops(fsname, nvprops, zct.zct_zplprops, &is_insensitive); if (error != 0) { @@ -4891,9 +4915,8 @@ zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr) * format. */ nvlist_t *attrs; - VERIFY(nvpair_value_nvlist(pair, &attrs) == 0); - VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, - &pair) == 0); + VERIFY0(nvpair_value_nvlist(pair, &attrs)); + VERIFY0(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, &pair)); } /* @@ -5000,15 +5023,6 @@ zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr) } break; - case ZFS_PROP_SPECIAL_SMALL_BLOCKS: - /* - * This property could require the allocation classes - * feature to be active for setting, however we allow - * it so that tests of settable properties succeed. - * The CLI will issue a warning in this case. - */ - break; - case ZFS_PROP_SHARESMB: if (zpl_earlier_version(dsname, ZPL_VERSION_FUID)) return (SET_ERROR(ENOTSUP)); @@ -5087,7 +5101,7 @@ zfs_check_clearable(const char *dataset, nvlist_t *props, nvlist_t **errlist) if (props == NULL) return (0); - VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0); + VERIFY0(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP)); zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP); (void) strlcpy(zc->zc_name, dataset, sizeof (zc->zc_name)); @@ -5099,9 +5113,8 @@ zfs_check_clearable(const char *dataset, nvlist_t *props, nvlist_t **errlist) sizeof (zc->zc_value)); if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 || (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) { - VERIFY(nvlist_remove_nvpair(props, pair) == 0); - VERIFY(nvlist_add_int32(errors, - zc->zc_value, err) == 0); + VERIFY0(nvlist_remove_nvpair(props, pair)); + VERIFY0(nvlist_add_int32(errors, zc->zc_value, err)); } pair = next_pair; } @@ -5111,7 +5124,7 @@ zfs_check_clearable(const char *dataset, nvlist_t *props, nvlist_t **errlist) nvlist_free(errors); errors = NULL; } else { - VERIFY(nvpair_value_int32(pair, &rv) == 0); + VERIFY0(nvpair_value_int32(pair, &rv)); } if (errlist == NULL) @@ -5128,16 +5141,14 @@ propval_equals(nvpair_t *p1, nvpair_t *p2) if (nvpair_type(p1) == DATA_TYPE_NVLIST) { /* dsl_prop_get_all_impl() format */ nvlist_t *attrs; - VERIFY(nvpair_value_nvlist(p1, &attrs) == 0); - VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, - &p1) == 0); + VERIFY0(nvpair_value_nvlist(p1, &attrs)); + VERIFY0(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, &p1)); } if (nvpair_type(p2) == DATA_TYPE_NVLIST) { nvlist_t *attrs; - VERIFY(nvpair_value_nvlist(p2, &attrs) == 0); - VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, - &p2) == 0); + VERIFY0(nvpair_value_nvlist(p2, &attrs)); + VERIFY0(nvlist_lookup_nvpair(attrs, ZPROP_VALUE, &p2)); } if (nvpair_type(p1) != nvpair_type(p2)) @@ -5146,14 +5157,14 @@ propval_equals(nvpair_t *p1, nvpair_t *p2) if (nvpair_type(p1) == DATA_TYPE_STRING) { const char *valstr1, *valstr2; - VERIFY(nvpair_value_string(p1, &valstr1) == 0); - VERIFY(nvpair_value_string(p2, &valstr2) == 0); + VERIFY0(nvpair_value_string(p1, &valstr1)); + VERIFY0(nvpair_value_string(p2, &valstr2)); return (strcmp(valstr1, valstr2) == 0); } else { uint64_t intval1, intval2; - VERIFY(nvpair_value_uint64(p1, &intval1) == 0); - VERIFY(nvpair_value_uint64(p2, &intval2) == 0); + VERIFY0(nvpair_value_uint64(p1, &intval1)); + VERIFY0(nvpair_value_uint64(p2, &intval2)); return (intval1 == intval2); } } @@ -5221,7 +5232,7 @@ extract_delay_props(nvlist_t *props) }; int i; - VERIFY(nvlist_alloc(&delayprops, NV_UNIQUE_NAME, KM_SLEEP) == 0); + VERIFY0(nvlist_alloc(&delayprops, NV_UNIQUE_NAME, KM_SLEEP)); for (nvp = nvlist_next_nvpair(props, NULL); nvp != NULL; nvp = nvlist_next_nvpair(props, nvp)) { @@ -5237,8 +5248,8 @@ extract_delay_props(nvlist_t *props) } if (delayable[i] != 0) { tmp = nvlist_prev_nvpair(props, nvp); - VERIFY(nvlist_add_nvpair(delayprops, nvp) == 0); - VERIFY(nvlist_remove_nvpair(props, nvp) == 0); + VERIFY0(nvlist_add_nvpair(delayprops, nvp)); + VERIFY0(nvlist_remove_nvpair(props, nvp)); nvp = tmp; } } @@ -5469,15 +5480,15 @@ zfs_ioc_recv_impl(char *tofs, char *tosnap, const char *origin, * using ASSERT() will be just like a VERIFY. */ if (recv_delayprops != NULL) { - ASSERT(nvlist_merge(recvprops, recv_delayprops, 0) == 0); + ASSERT0(nvlist_merge(recvprops, recv_delayprops, 0)); nvlist_free(recv_delayprops); } if (local_delayprops != NULL) { - ASSERT(nvlist_merge(localprops, local_delayprops, 0) == 0); + ASSERT0(nvlist_merge(localprops, local_delayprops, 0)); nvlist_free(local_delayprops); } if (inherited_delayprops != NULL) { - ASSERT(nvlist_merge(localprops, inherited_delayprops, 0) == 0); + ASSERT0(nvlist_merge(localprops, inherited_delayprops, 0)); nvlist_free(inherited_delayprops); } *read_bytes = off - noff; @@ -7326,8 +7337,8 @@ zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func, ASSERT3U(ioc, >=, ZFS_IOC_FIRST); ASSERT3U(ioc, <, ZFS_IOC_LAST); - ASSERT3P(vec->zvec_legacy_func, ==, NULL); - ASSERT3P(vec->zvec_func, ==, NULL); + ASSERT0P(vec->zvec_legacy_func); + ASSERT0P(vec->zvec_func); vec->zvec_legacy_func = func; vec->zvec_secpolicy = secpolicy; @@ -7350,8 +7361,8 @@ zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func, ASSERT3U(ioc, >=, ZFS_IOC_FIRST); ASSERT3U(ioc, <, ZFS_IOC_LAST); - ASSERT3P(vec->zvec_legacy_func, ==, NULL); - ASSERT3P(vec->zvec_func, ==, NULL); + ASSERT0P(vec->zvec_legacy_func); + ASSERT0P(vec->zvec_func); /* if we are logging, the name must be valid */ ASSERT(!allow_log || namecheck != NO_NAME); @@ -8132,7 +8143,7 @@ zfsdev_ioctl_common(uint_t vecnum, zfs_cmd_t *zc, int flag) spa_t *spa; nvlist_t *lognv = NULL; - ASSERT(vec->zvec_legacy_func == NULL); + ASSERT0P(vec->zvec_legacy_func); /* * Add the innvl to the lognv before calling the func, diff --git a/sys/contrib/openzfs/module/zfs/zfs_log.c b/sys/contrib/openzfs/module/zfs/zfs_log.c index 2ce25b72b288..ea17e049279f 100644 --- a/sys/contrib/openzfs/module/zfs/zfs_log.c +++ b/sys/contrib/openzfs/module/zfs/zfs_log.c @@ -607,8 +607,6 @@ zfs_log_rename_whiteout(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype, * called as soon as the write is on stable storage (be it via a DMU sync or a * ZIL commit). */ -static uint_t zfs_immediate_write_sz = 32768; - void zfs_log_write(zilog_t *zilog, dmu_tx_t *tx, int txtype, znode_t *zp, offset_t off, ssize_t resid, boolean_t commit, @@ -622,19 +620,12 @@ zfs_log_write(zilog_t *zilog, dmu_tx_t *tx, int txtype, if (zil_replaying(zilog, tx) || zp->z_unlinked || zfs_xattr_owner_unlinked(zp)) { if (callback != NULL) - callback(callback_data); + callback(callback_data, 0); return; } - if (zilog->zl_logbias == ZFS_LOGBIAS_THROUGHPUT || o_direct) - write_state = WR_INDIRECT; - else if (!spa_has_slogs(zilog->zl_spa) && - resid >= zfs_immediate_write_sz) - write_state = WR_INDIRECT; - else if (commit) - write_state = WR_COPIED; - else - write_state = WR_NEED_COPY; + write_state = zil_write_state(zilog, resid, blocksize, o_direct, + commit); (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(ZTOZSB(zp)), &gen, sizeof (gen)); @@ -672,7 +663,7 @@ zfs_log_write(zilog_t *zilog, dmu_tx_t *tx, int txtype, DMU_KEEP_CACHING); DB_DNODE_EXIT(db); if (err != 0) { - zil_itx_destroy(itx); + zil_itx_destroy(itx, 0); itx = zil_itx_create(txtype, sizeof (*lr)); lr = (lr_write_t *)&itx->itx_lr; wr_state = WR_NEED_COPY; @@ -938,6 +929,3 @@ zfs_log_clone_range(zilog_t *zilog, dmu_tx_t *tx, int txtype, znode_t *zp, len -= partlen; } } - -ZFS_MODULE_PARAM(zfs, zfs_, immediate_write_sz, UINT, ZMOD_RW, - "Largest data block to write to zil"); diff --git a/sys/contrib/openzfs/module/zfs/zfs_quota.c b/sys/contrib/openzfs/module/zfs/zfs_quota.c index b8fe512d4f09..2e91ccc27d6d 100644 --- a/sys/contrib/openzfs/module/zfs/zfs_quota.c +++ b/sys/contrib/openzfs/module/zfs/zfs_quota.c @@ -374,7 +374,7 @@ zfs_set_userquota(zfsvfs_t *zfsvfs, zfs_userquota_prop_t type, if (*objp == 0) { *objp = zap_create(zfsvfs->z_os, DMU_OT_USERGROUP_QUOTA, DMU_OT_NONE, 0, tx); - VERIFY(0 == zap_add(zfsvfs->z_os, MASTER_NODE_OBJ, + VERIFY0(zap_add(zfsvfs->z_os, MASTER_NODE_OBJ, zfs_userquota_prop_prefixes[type], 8, 1, objp, tx)); } mutex_exit(&zfsvfs->z_lock); @@ -386,7 +386,7 @@ zfs_set_userquota(zfsvfs_t *zfsvfs, zfs_userquota_prop_t type, } else { err = zap_update(zfsvfs->z_os, *objp, buf, 8, 1, "a, tx); } - ASSERT(err == 0); + ASSERT0(err); if (fuid_dirtied) zfs_fuid_sync(zfsvfs, tx); dmu_tx_commit(tx); diff --git a/sys/contrib/openzfs/module/zfs/zfs_rlock.c b/sys/contrib/openzfs/module/zfs/zfs_rlock.c index 53eb3ef1b66e..4035baff77d6 100644 --- a/sys/contrib/openzfs/module/zfs/zfs_rlock.c +++ b/sys/contrib/openzfs/module/zfs/zfs_rlock.c @@ -666,7 +666,7 @@ zfs_rangelock_reduce(zfs_locked_range_t *lr, uint64_t off, uint64_t len) /* Ensure there are no other locks */ ASSERT3U(avl_numnodes(&rl->rl_tree), ==, 1); - ASSERT3U(lr->lr_offset, ==, 0); + ASSERT0(lr->lr_offset); ASSERT3U(lr->lr_type, ==, RL_WRITER); ASSERT(!lr->lr_proxy); ASSERT3U(lr->lr_length, ==, UINT64_MAX); diff --git a/sys/contrib/openzfs/module/zfs/zfs_sa.c b/sys/contrib/openzfs/module/zfs/zfs_sa.c index 59b6ae4e4203..8b4fc6fd7fbd 100644 --- a/sys/contrib/openzfs/module/zfs/zfs_sa.c +++ b/sys/contrib/openzfs/module/zfs/zfs_sa.c @@ -169,7 +169,7 @@ zfs_sa_set_scanstamp(znode_t *zp, xvattr_t *xvap, dmu_tx_t *tx) ASSERT(MUTEX_HELD(&zp->z_lock)); VERIFY((xoap = xva_getxoptattr(xvap)) != NULL); if (zp->z_is_sa) - VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_SCANSTAMP(zfsvfs), + VERIFY0(sa_update(zp->z_sa_hdl, SA_ZPL_SCANSTAMP(zfsvfs), &xoap->xoa_av_scanstamp, sizeof (xoap->xoa_av_scanstamp), tx)); else { @@ -181,12 +181,12 @@ zfs_sa_set_scanstamp(znode_t *zp, xvattr_t *xvap, dmu_tx_t *tx) len = sizeof (xoap->xoa_av_scanstamp) + ZFS_OLD_ZNODE_PHYS_SIZE; if (len > doi.doi_bonus_size) - VERIFY(dmu_set_bonus(db, len, tx) == 0); + VERIFY0(dmu_set_bonus(db, len, tx)); (void) memcpy((caddr_t)db->db_data + ZFS_OLD_ZNODE_PHYS_SIZE, xoap->xoa_av_scanstamp, sizeof (xoap->xoa_av_scanstamp)); zp->z_pflags |= ZFS_BONUS_SCANSTAMP; - VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_FLAGS(zfsvfs), + VERIFY0(sa_update(zp->z_sa_hdl, SA_ZPL_FLAGS(zfsvfs), &zp->z_pflags, sizeof (uint64_t), tx)); } } @@ -286,7 +286,7 @@ zfs_sa_set_xattr(znode_t *zp, const char *name, const void *value, size_t vsize) dmu_tx_commit(tx); if (logsaxattr && zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) - zil_commit(zilog, 0); + error = zil_commit(zilog, 0); } out_free: vmem_free(obj, size); @@ -427,11 +427,10 @@ zfs_sa_upgrade(sa_handle_t *hdl, dmu_tx_t *tx) zp->z_pflags &= ~ZFS_BONUS_SCANSTAMP; } - VERIFY(dmu_set_bonustype(db, DMU_OT_SA, tx) == 0); - VERIFY(sa_replace_all_by_template_locked(hdl, sa_attrs, - count, tx) == 0); + VERIFY0(dmu_set_bonustype(db, DMU_OT_SA, tx)); + VERIFY0(sa_replace_all_by_template_locked(hdl, sa_attrs, count, tx)); if (znode_acl.z_acl_extern_obj) - VERIFY(0 == dmu_object_free(zfsvfs->z_os, + VERIFY0(dmu_object_free(zfsvfs->z_os, znode_acl.z_acl_extern_obj, tx)); zp->z_is_sa = B_TRUE; diff --git a/sys/contrib/openzfs/module/zfs/zfs_vnops.c b/sys/contrib/openzfs/module/zfs/zfs_vnops.c index 656ca4dc22ff..7bb9ba57c69e 100644 --- a/sys/contrib/openzfs/module/zfs/zfs_vnops.c +++ b/sys/contrib/openzfs/module/zfs/zfs_vnops.c @@ -27,6 +27,7 @@ * Copyright 2017 Nexenta Systems, Inc. * Copyright (c) 2021, 2022 by Pawel Jakub Dawidek * Copyright (c) 2025, Rob Norris <robn@despairlabs.com> + * Copyright (c) 2025, Klara, Inc. */ /* Portions Copyright 2007 Jeremy Teo */ @@ -49,6 +50,7 @@ #include <sys/dmu.h> #include <sys/dmu_objset.h> #include <sys/dsl_crypt.h> +#include <sys/dsl_dataset.h> #include <sys/spa.h> #include <sys/txg.h> #include <sys/dbuf.h> @@ -67,13 +69,14 @@ int zfs_bclone_enabled = 1; /* - * When set zfs_clone_range() waits for dirty data to be written to disk. - * This allows the clone operation to reliably succeed when a file is modified - * and then immediately cloned. For small files this may be slower than making - * a copy of the file and is therefore not the default. However, in certain - * scenarios this behavior may be desirable so a tunable is provided. + * When set to 1 the FICLONE and FICLONERANGE ioctls will wait for any dirty + * data to be written to disk before proceeding. This ensures that the clone + * operation reliably succeeds, even if a file is modified and then immediately + * cloned. Note that for small files this may be slower than simply copying + * the file. When set to 0 the clone operation will immediately fail if it + * encounters any dirty blocks. By default waiting is enabled. */ -int zfs_bclone_wait_dirty = 0; +int zfs_bclone_wait_dirty = 1; /* * Enable Direct I/O. If this setting is 0, then all I/O requests will be @@ -114,9 +117,7 @@ zfs_fsync(znode_t *zp, int syncflag, cred_t *cr) if (zfsvfs->z_os->os_sync != ZFS_SYNC_DISABLED) { if ((error = zfs_enter_verify_zp(zfsvfs, zp, FTAG)) != 0) return (error); - atomic_inc_32(&zp->z_sync_writes_cnt); - zil_commit(zfsvfs->z_log, zp->z_id); - atomic_dec_32(&zp->z_sync_writes_cnt); + error = zil_commit(zfsvfs->z_log, zp->z_id); zfs_exit(zfsvfs, FTAG); } return (error); @@ -375,8 +376,13 @@ zfs_read(struct znode *zp, zfs_uio_t *uio, int ioflag, cred_t *cr) frsync = !!(ioflag & FRSYNC); #endif if (zfsvfs->z_log && - (frsync || zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)) - zil_commit(zfsvfs->z_log, zp->z_id); + (frsync || zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)) { + error = zil_commit(zfsvfs->z_log, zp->z_id); + if (error != 0) { + zfs_exit(zfsvfs, FTAG); + return (error); + } + } /* * Lock the range against changes. @@ -1074,8 +1080,13 @@ zfs_write(znode_t *zp, zfs_uio_t *uio, int ioflag, cred_t *cr) return (error); } - if (commit) - zil_commit(zilog, zp->z_id); + if (commit) { + error = zil_commit(zilog, zp->z_id); + if (error != 0) { + zfs_exit(zfsvfs, FTAG); + return (error); + } + } int64_t nwritten = start_resid - zfs_uio_resid(uio); dataset_kstats_update_write_kstats(&zfsvfs->z_kstat, nwritten); @@ -1102,13 +1113,21 @@ zfs_rewrite(znode_t *zp, uint64_t off, uint64_t len, uint64_t flags, { int error; - if (flags != 0 || arg != 0) + if ((flags & ~ZFS_REWRITE_PHYSICAL) != 0 || arg != 0) return (SET_ERROR(EINVAL)); zfsvfs_t *zfsvfs = ZTOZSB(zp); if ((error = zfs_enter_verify_zp(zfsvfs, zp, FTAG)) != 0) return (error); + /* Check if physical rewrite is allowed */ + spa_t *spa = zfsvfs->z_os->os_spa; + if ((flags & ZFS_REWRITE_PHYSICAL) && + !spa_feature_is_enabled(spa, SPA_FEATURE_PHYSICAL_REWRITE)) { + zfs_exit(zfsvfs, FTAG); + return (SET_ERROR(ENOTSUP)); + } + if (zfs_is_readonly(zfsvfs)) { zfs_exit(zfsvfs, FTAG); return (SET_ERROR(EROFS)); @@ -1196,7 +1215,10 @@ zfs_rewrite(znode_t *zp, uint64_t off, uint64_t len, uint64_t flags, if (dmu_buf_is_dirty(dbp[i], tx)) continue; nw += dbp[i]->db_size; - dmu_buf_will_dirty(dbp[i], tx); + if (flags & ZFS_REWRITE_PHYSICAL) + dmu_buf_will_rewrite(dbp[i], tx); + else + dmu_buf_will_dirty(dbp[i], tx); } dmu_buf_rele_array(dbp, numbufs, FTAG); @@ -1249,8 +1271,8 @@ zfs_setsecattr(znode_t *zp, vsecattr_t *vsecp, int flag, cred_t *cr) zilog = zfsvfs->z_log; error = zfs_setacl(zp, vsecp, skipaclchk, cr); - if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) - zil_commit(zilog, 0); + if (error == 0 && zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) + error = zil_commit(zilog, 0); zfs_exit(zfsvfs, FTAG); return (error); @@ -1935,7 +1957,7 @@ unlock: ZFS_ACCESSTIME_STAMP(inzfsvfs, inzp); if (outos->os_sync == ZFS_SYNC_ALWAYS) { - zil_commit(zilog, outzp->z_id); + error = zil_commit(zilog, outzp->z_id); } *inoffp += done; diff --git a/sys/contrib/openzfs/module/zfs/zil.c b/sys/contrib/openzfs/module/zfs/zil.c index 00059b2c6de0..31b59c55f17b 100644 --- a/sys/contrib/openzfs/module/zfs/zil.c +++ b/sys/contrib/openzfs/module/zfs/zil.c @@ -24,6 +24,7 @@ * Copyright (c) 2011, 2018 by Delphix. All rights reserved. * Copyright (c) 2014 Integros [integros.com] * Copyright (c) 2018 Datto Inc. + * Copyright (c) 2025, Klara, Inc. */ /* Portions Copyright 2010 Robert Milkowski */ @@ -103,6 +104,7 @@ static zil_kstat_values_t zil_stats = { { "zil_commit_error_count", KSTAT_DATA_UINT64 }, { "zil_commit_stall_count", KSTAT_DATA_UINT64 }, { "zil_commit_suspend_count", KSTAT_DATA_UINT64 }, + { "zil_commit_crash_count", KSTAT_DATA_UINT64 }, { "zil_itx_count", KSTAT_DATA_UINT64 }, { "zil_itx_indirect_count", KSTAT_DATA_UINT64 }, { "zil_itx_indirect_bytes", KSTAT_DATA_UINT64 }, @@ -145,7 +147,7 @@ static uint64_t zil_slog_bulk = 64 * 1024 * 1024; static kmem_cache_t *zil_lwb_cache; static kmem_cache_t *zil_zcw_cache; -static void zil_lwb_commit(zilog_t *zilog, lwb_t *lwb, itx_t *itx); +static int zil_lwb_commit(zilog_t *zilog, lwb_t *lwb, itx_t *itx); static itx_t *zil_itx_clone(itx_t *oitx); static uint64_t zil_max_waste_space(zilog_t *zilog); @@ -367,6 +369,7 @@ zil_sums_init(zil_sums_t *zs) wmsum_init(&zs->zil_commit_error_count, 0); wmsum_init(&zs->zil_commit_stall_count, 0); wmsum_init(&zs->zil_commit_suspend_count, 0); + wmsum_init(&zs->zil_commit_crash_count, 0); wmsum_init(&zs->zil_itx_count, 0); wmsum_init(&zs->zil_itx_indirect_count, 0); wmsum_init(&zs->zil_itx_indirect_bytes, 0); @@ -392,6 +395,7 @@ zil_sums_fini(zil_sums_t *zs) wmsum_fini(&zs->zil_commit_error_count); wmsum_fini(&zs->zil_commit_stall_count); wmsum_fini(&zs->zil_commit_suspend_count); + wmsum_fini(&zs->zil_commit_crash_count); wmsum_fini(&zs->zil_itx_count); wmsum_fini(&zs->zil_itx_indirect_count); wmsum_fini(&zs->zil_itx_indirect_bytes); @@ -422,6 +426,8 @@ zil_kstat_values_update(zil_kstat_values_t *zs, zil_sums_t *zil_sums) wmsum_value(&zil_sums->zil_commit_stall_count); zs->zil_commit_suspend_count.value.ui64 = wmsum_value(&zil_sums->zil_commit_suspend_count); + zs->zil_commit_crash_count.value.ui64 = + wmsum_value(&zil_sums->zil_commit_crash_count); zs->zil_itx_count.value.ui64 = wmsum_value(&zil_sums->zil_itx_count); zs->zil_itx_indirect_count.value.ui64 = @@ -589,7 +595,7 @@ zil_clear_log_block(zilog_t *zilog, const blkptr_t *bp, void *tx, * that we rewind to is invalid. Thus, we return -1 so * zil_parse() doesn't attempt to read it. */ - if (BP_GET_LOGICAL_BIRTH(bp) >= first_txg) + if (BP_GET_BIRTH(bp) >= first_txg) return (-1); if (zil_bp_tree_add(zilog, bp) != 0) @@ -615,7 +621,7 @@ zil_claim_log_block(zilog_t *zilog, const blkptr_t *bp, void *tx, * Claim log block if not already committed and not already claimed. * If tx == NULL, just verify that the block is claimable. */ - if (BP_IS_HOLE(bp) || BP_GET_LOGICAL_BIRTH(bp) < first_txg || + if (BP_IS_HOLE(bp) || BP_GET_BIRTH(bp) < first_txg || zil_bp_tree_add(zilog, bp) != 0) return (0); @@ -640,7 +646,7 @@ zil_claim_write(zilog_t *zilog, const lr_t *lrc, void *tx, uint64_t first_txg) * waited for all writes to be stable first), so it is semantically * correct to declare this the end of the log. */ - if (BP_GET_LOGICAL_BIRTH(&lr->lr_blkptr) >= first_txg) { + if (BP_GET_BIRTH(&lr->lr_blkptr) >= first_txg) { error = zil_read_log_data(zilog, lr, NULL); if (error != 0) return (error); @@ -687,7 +693,7 @@ zil_claim_clone_range(zilog_t *zilog, const lr_t *lrc, void *tx, * just in case lets be safe and just stop here now instead of * corrupting the pool. */ - if (BP_GET_BIRTH(bp) >= first_txg) + if (BP_GET_PHYSICAL_BIRTH(bp) >= first_txg) return (SET_ERROR(ENOENT)); /* @@ -742,7 +748,7 @@ zil_free_write(zilog_t *zilog, const lr_t *lrc, void *tx, uint64_t claim_txg) /* * If we previously claimed it, we need to free it. */ - if (BP_GET_LOGICAL_BIRTH(bp) >= claim_txg && + if (BP_GET_BIRTH(bp) >= claim_txg && zil_bp_tree_add(zilog, bp) == 0 && !BP_IS_HOLE(bp)) { zio_free(zilog->zl_spa, dmu_tx_get_txg(tx), bp); } @@ -864,9 +870,9 @@ zil_free_lwb(zilog_t *zilog, lwb_t *lwb) ASSERT(MUTEX_HELD(&zilog->zl_lock)); ASSERT(lwb->lwb_state == LWB_STATE_NEW || lwb->lwb_state == LWB_STATE_FLUSH_DONE); - ASSERT3P(lwb->lwb_child_zio, ==, NULL); - ASSERT3P(lwb->lwb_write_zio, ==, NULL); - ASSERT3P(lwb->lwb_root_zio, ==, NULL); + ASSERT0P(lwb->lwb_child_zio); + ASSERT0P(lwb->lwb_write_zio); + ASSERT0P(lwb->lwb_root_zio); ASSERT3U(lwb->lwb_alloc_txg, <=, spa_syncing_txg(zilog->zl_spa)); ASSERT3U(lwb->lwb_max_txg, <=, spa_syncing_txg(zilog->zl_spa)); VERIFY(list_is_empty(&lwb->lwb_itxs)); @@ -991,8 +997,8 @@ zil_create(zilog_t *zilog) */ txg_wait_synced(zilog->zl_dmu_pool, zilog->zl_destroy_txg); - ASSERT(zh->zh_claim_txg == 0); - ASSERT(zh->zh_replay_seq == 0); + ASSERT0(zh->zh_claim_txg); + ASSERT0(zh->zh_replay_seq); blk = zh->zh_log; @@ -1104,7 +1110,7 @@ zil_destroy(zilog_t *zilog, boolean_t keep_first) zilog->zl_keep_first = keep_first; if (!list_is_empty(&zilog->zl_lwb_list)) { - ASSERT(zh->zh_claim_txg == 0); + ASSERT0(zh->zh_claim_txg); VERIFY(!keep_first); while ((lwb = list_remove_head(&zilog->zl_lwb_list)) != NULL) { if (lwb->lwb_buf != NULL) @@ -1250,7 +1256,7 @@ zil_check_log_chain(dsl_pool_t *dp, dsl_dataset_t *ds, void *tx) blkptr_t *bp; int error; - ASSERT(tx == NULL); + ASSERT0P(tx); error = dmu_objset_from_ds(ds, &os); if (error != 0) { @@ -1351,7 +1357,7 @@ zil_commit_waiter_link_lwb(zil_commit_waiter_t *zcw, lwb_t *lwb) ASSERT(!list_link_active(&zcw->zcw_node)); list_insert_tail(&lwb->lwb_waiters, zcw); - ASSERT3P(zcw->zcw_lwb, ==, NULL); + ASSERT0P(zcw->zcw_lwb); zcw->zcw_lwb = lwb; } @@ -1365,7 +1371,7 @@ zil_commit_waiter_link_nolwb(zil_commit_waiter_t *zcw, list_t *nolwb) { ASSERT(!list_link_active(&zcw->zcw_node)); list_insert_tail(nolwb, zcw); - ASSERT3P(zcw->zcw_lwb, ==, NULL); + ASSERT0P(zcw->zcw_lwb); } void @@ -1482,7 +1488,7 @@ zil_lwb_flush_vdevs_done(zio_t *zio) } while ((itx = list_remove_head(&lwb->lwb_itxs)) != NULL) - zil_itx_destroy(itx); + zil_itx_destroy(itx, 0); while ((zcw = list_remove_head(&lwb->lwb_waiters)) != NULL) { mutex_enter(&zcw->zcw_lock); @@ -1895,7 +1901,7 @@ zil_lwb_write_close(zilog_t *zilog, lwb_t *lwb, lwb_state_t state) /* * Finalize previously closed block and issue the write zio. */ -static void +static int zil_lwb_write_issue(zilog_t *zilog, lwb_t *lwb) { spa_t *spa = zilog->zl_spa; @@ -1909,8 +1915,13 @@ zil_lwb_write_issue(zilog_t *zilog, lwb_t *lwb) /* Actually fill the lwb with the data. */ for (itx_t *itx = list_head(&lwb->lwb_itxs); itx; - itx = list_next(&lwb->lwb_itxs, itx)) - zil_lwb_commit(zilog, lwb, itx); + itx = list_next(&lwb->lwb_itxs, itx)) { + error = zil_lwb_commit(zilog, lwb, itx); + if (error != 0) { + ASSERT3U(error, ==, ESHUTDOWN); + return (error); + } + } lwb->lwb_nused = lwb->lwb_nfilled; ASSERT3U(lwb->lwb_nused, <=, lwb->lwb_nmax); @@ -1928,7 +1939,7 @@ zil_lwb_write_issue(zilog_t *zilog, lwb_t *lwb) lwb->lwb_state = LWB_STATE_READY; if (BP_IS_HOLE(&lwb->lwb_blk) && lwb->lwb_error == 0) { mutex_exit(&zilog->zl_lock); - return; + return (0); } mutex_exit(&zilog->zl_lock); @@ -1997,7 +2008,7 @@ next_lwb: &slog); } if (error == 0) { - ASSERT3U(BP_GET_LOGICAL_BIRTH(bp), ==, txg); + ASSERT3U(BP_GET_BIRTH(bp), ==, txg); BP_SET_CHECKSUM(bp, nlwb->lwb_slim ? ZIO_CHECKSUM_ZILOG2 : ZIO_CHECKSUM_ZILOG); bp->blk_cksum = lwb->lwb_blk.blk_cksum; @@ -2065,6 +2076,8 @@ next_lwb: lwb = nlwb; if (lwb) goto next_lwb; + + return (0); } /* @@ -2095,6 +2108,19 @@ zil_max_waste_space(zilog_t *zilog) */ static uint_t zil_maxcopied = 7680; +/* + * Largest write size to store the data directly into ZIL. + */ +uint_t zfs_immediate_write_sz = 32768; + +/* + * When enabled and blocks go to normal vdev, treat special vdevs as SLOG, + * writing data to ZIL (WR_COPIED/WR_NEED_COPY). Disabling this forces the + * indirect writes (WR_INDIRECT) to preserve special vdev throughput and + * endurance, likely at the cost of normal vdev latency. + */ +int zil_special_is_slog = 1; + uint64_t zil_max_copied_data(zilog_t *zilog) { @@ -2102,6 +2128,46 @@ zil_max_copied_data(zilog_t *zilog) return (MIN(max_data, zil_maxcopied)); } +/* + * Determine the appropriate write state for ZIL transactions based on + * pool configuration, data placement, write size, and logbias settings. + */ +itx_wr_state_t +zil_write_state(zilog_t *zilog, uint64_t size, uint32_t blocksize, + boolean_t o_direct, boolean_t commit) +{ + if (zilog->zl_logbias == ZFS_LOGBIAS_THROUGHPUT || o_direct) + return (WR_INDIRECT); + + /* + * Don't use indirect for too small writes to reduce overhead. + * Don't use indirect if written less than a half of a block if + * we are going to commit it immediately, since next write might + * rewrite the same block again, causing inflation. If commit + * is not planned, then next writes might coalesce, and so the + * indirect may be perfect. + */ + boolean_t indirect = (size >= zfs_immediate_write_sz && + (size >= blocksize / 2 || !commit)); + + if (spa_has_slogs(zilog->zl_spa)) { + /* Dedicated slogs: never use indirect */ + indirect = B_FALSE; + } else if (spa_has_special(zilog->zl_spa)) { + /* Special vdevs: only when beneficial */ + boolean_t on_special = (blocksize <= + zilog->zl_os->os_zpl_special_smallblock); + indirect &= (on_special || !zil_special_is_slog); + } + + if (indirect) + return (WR_INDIRECT); + else if (commit) + return (WR_COPIED); + else + return (WR_NEED_COPY); +} + static uint64_t zil_itx_record_size(itx_t *itx) { @@ -2255,11 +2321,13 @@ cont: return (lwb); } +static void zil_crash(zilog_t *zilog); + /* * Fill the actual transaction data into the lwb, following zil_lwb_assign(). * Does not require locking. */ -static void +static int zil_lwb_commit(zilog_t *zilog, lwb_t *lwb, itx_t *itx) { lr_t *lr, *lrb; @@ -2271,7 +2339,7 @@ zil_lwb_commit(zilog_t *zilog, lwb_t *lwb, itx_t *itx) lrw = (lr_write_t *)lr; if (lr->lrc_txtype == TX_COMMIT) - return; + return (0); reclen = lr->lrc_reclen; dlen = zil_itx_data_size(itx); @@ -2357,16 +2425,35 @@ zil_lwb_commit(zilog_t *zilog, lwb_t *lwb, itx_t *itx) ". Falling back to txg_wait_synced().", error); zfs_fallthrough; - case EIO: - txg_wait_synced(zilog->zl_dmu_pool, - lr->lrc_txg); + case EIO: { + int error = txg_wait_synced_flags( + zilog->zl_dmu_pool, + lr->lrc_txg, TXG_WAIT_SUSPEND); + if (error != 0) { + ASSERT3U(error, ==, ESHUTDOWN); + /* + * zil_lwb_commit() is called from a + * loop over a list of itxs at the + * top of zil_lwb_write_issue(), which + * itself is called from a loop over a + * list of lwbs in various places. + * zil_crash() will free those itxs + * and sometimes the lwbs, so they + * are invalid when zil_crash() returns. + * Callers must pretty much abort + * immediately. + */ + zil_crash(zilog); + return (error); + } zfs_fallthrough; + } case ENOENT: zfs_fallthrough; case EEXIST: zfs_fallthrough; case EALREADY: - return; + return (0); } } } @@ -2374,6 +2461,8 @@ zil_lwb_commit(zilog_t *zilog, lwb_t *lwb, itx_t *itx) lwb->lwb_nfilled += reclen + dlen; ASSERT3S(lwb->lwb_nfilled, <=, lwb->lwb_nused); ASSERT0(P2PHASE(lwb->lwb_nfilled, sizeof (uint64_t))); + + return (0); } itx_t * @@ -2415,7 +2504,7 @@ zil_itx_clone(itx_t *oitx) } void -zil_itx_destroy(itx_t *itx) +zil_itx_destroy(itx_t *itx, int err) { ASSERT3U(itx->itx_size, >=, sizeof (itx_t)); ASSERT3U(itx->itx_lr.lrc_reclen, ==, @@ -2424,7 +2513,7 @@ zil_itx_destroy(itx_t *itx) IMPLY(itx->itx_callback != NULL, itx->itx_lr.lrc_txtype != TX_COMMIT); if (itx->itx_callback != NULL) - itx->itx_callback(itx->itx_callback_data); + itx->itx_callback(itx->itx_callback_data, err); zio_data_buf_free(itx, itx->itx_size); } @@ -2467,7 +2556,7 @@ zil_itxg_clean(void *arg) if (itx->itx_lr.lrc_txtype == TX_COMMIT) zil_commit_waiter_skip(itx->itx_private); - zil_itx_destroy(itx); + zil_itx_destroy(itx, 0); } cookie = NULL; @@ -2477,7 +2566,7 @@ zil_itxg_clean(void *arg) while ((itx = list_remove_head(list)) != NULL) { /* commit itxs should never be on the async lists. */ ASSERT3U(itx->itx_lr.lrc_txtype, !=, TX_COMMIT); - zil_itx_destroy(itx); + zil_itx_destroy(itx, 0); } list_destroy(list); kmem_free(ian, sizeof (itx_async_node_t)); @@ -2539,7 +2628,7 @@ zil_remove_async(zilog_t *zilog, uint64_t oid) while ((itx = list_remove_head(&clean_list)) != NULL) { /* commit itxs should never be on the async lists. */ ASSERT3U(itx->itx_lr.lrc_txtype, !=, TX_COMMIT); - zil_itx_destroy(itx); + zil_itx_destroy(itx, 0); } list_destroy(&clean_list); } @@ -2624,6 +2713,67 @@ zil_itx_assign(zilog_t *zilog, itx_t *itx, dmu_tx_t *tx) } /* + * Post-crash cleanup. This is called from zil_clean() because it needs to + * do cleanup after every txg until the ZIL is restarted, and zilog_dirty() + * can arrange that easily, unlike zil_sync() which is more complicated to + * get a call to without actual dirty data. + */ +static void +zil_crash_clean(zilog_t *zilog, uint64_t synced_txg) +{ + ASSERT(MUTEX_HELD(&zilog->zl_lock)); + ASSERT3U(zilog->zl_restart_txg, >, 0); + + /* Clean up anything on the crash list from earlier txgs */ + lwb_t *lwb; + while ((lwb = list_head(&zilog->zl_lwb_crash_list)) != NULL) { + if (lwb->lwb_alloc_txg >= synced_txg || + lwb->lwb_max_txg >= synced_txg) { + /* + * This lwb was allocated or updated on this txg, or + * in the future. We stop processing here, to avoid + * the strange situation of freeing a ZIL block on + * on the same or earlier txg than what it was + * allocated for. + * + * We'll take care of it on the next txg. + */ + break; + } + + /* This LWB is from the past, so we can clean it up now. */ + list_remove(&zilog->zl_lwb_crash_list, lwb); + if (lwb->lwb_buf != NULL) + zio_buf_free(lwb->lwb_buf, lwb->lwb_sz); + if (!BP_IS_HOLE(&lwb->lwb_blk)) + /* + * Free on the next txg, since zil_clean() is called + * once synced_txg has already been completed. + */ + zio_free(zilog->zl_spa, synced_txg+1, &lwb->lwb_blk); + zil_free_lwb(zilog, lwb); + } + + if (zilog->zl_restart_txg > synced_txg) { + /* + * Not reached the restart txg yet, so mark the ZIL dirty for + * the next txg and we'll consider it all again then. + */ + zilog_dirty(zilog, synced_txg+1); + return; + } + + /* + * Reached the restart txg, so we can allow new calls to zil_commit(). + * All ZIL txgs have long past so there should be no IO waiting. + */ + ASSERT(list_is_empty(&zilog->zl_lwb_list)); + ASSERT(list_is_empty(&zilog->zl_lwb_crash_list)); + + zilog->zl_restart_txg = 0; +} + +/* * If there are any in-memory intent log transactions which have now been * synced then start up a taskq to free them. We should only do this after we * have written out the uberblocks (i.e. txg has been committed) so that @@ -2638,6 +2788,15 @@ zil_clean(zilog_t *zilog, uint64_t synced_txg) ASSERT3U(synced_txg, <, ZILTEST_TXG); + /* Do cleanup and restart after crash. */ + if (zilog->zl_restart_txg > 0) { + mutex_enter(&zilog->zl_lock); + /* Make sure we didn't lose a race. */ + if (zilog->zl_restart_txg > 0) + zil_crash_clean(zilog, synced_txg); + mutex_exit(&zilog->zl_lock); + } + mutex_enter(&itxg->itxg_lock); if (itxg->itxg_itxs == NULL || itxg->itxg_txg == ZILTEST_TXG) { mutex_exit(&itxg->itxg_lock); @@ -2830,13 +2989,13 @@ zil_prune_commit_list(zilog_t *zilog) mutex_exit(&zilog->zl_lock); list_remove(&zilog->zl_itx_commit_list, itx); - zil_itx_destroy(itx); + zil_itx_destroy(itx, 0); } IMPLY(itx != NULL, itx->itx_lr.lrc_txtype != TX_COMMIT); } -static void +static int zil_commit_writer_stall(zilog_t *zilog) { /* @@ -2861,8 +3020,22 @@ zil_commit_writer_stall(zilog_t *zilog) */ ASSERT(MUTEX_HELD(&zilog->zl_issuer_lock)); ZIL_STAT_BUMP(zilog, zil_commit_stall_count); - txg_wait_synced(zilog->zl_dmu_pool, 0); + + int err = txg_wait_synced_flags(zilog->zl_dmu_pool, 0, + TXG_WAIT_SUSPEND); + if (err != 0) { + ASSERT3U(err, ==, ESHUTDOWN); + zil_crash(zilog); + } + + /* + * Either zil_sync() has been called to wait for and clean up any + * in-flight LWBs, or zil_crash() has emptied out the list and arranged + * for them to be cleaned up later. + */ ASSERT(list_is_empty(&zilog->zl_lwb_list)); + + return (err); } static void @@ -2902,19 +3075,14 @@ zil_process_commit_list(zilog_t *zilog, zil_commit_waiter_t *zcw, list_t *ilwbs) ASSERT(MUTEX_HELD(&zilog->zl_issuer_lock)); - /* - * Return if there's nothing to commit before we dirty the fs by - * calling zil_create(). - */ - if (list_is_empty(&zilog->zl_itx_commit_list)) - return; - - list_create(&nolwb_itxs, sizeof (itx_t), offsetof(itx_t, itx_node)); - list_create(&nolwb_waiters, sizeof (zil_commit_waiter_t), - offsetof(zil_commit_waiter_t, zcw_node)); - lwb = list_tail(&zilog->zl_lwb_list); if (lwb == NULL) { + /* + * Return if there's nothing to commit before we dirty the fs. + */ + if (list_is_empty(&zilog->zl_itx_commit_list)) + return; + lwb = zil_create(zilog); } else { /* @@ -2942,6 +3110,10 @@ zil_process_commit_list(zilog_t *zilog, zil_commit_waiter_t *zcw, list_t *ilwbs) } } + list_create(&nolwb_itxs, sizeof (itx_t), offsetof(itx_t, itx_node)); + list_create(&nolwb_waiters, sizeof (zil_commit_waiter_t), + offsetof(zil_commit_waiter_t, zcw_node)); + while ((itx = list_remove_head(&zilog->zl_itx_commit_list)) != NULL) { lr_t *lrc = &itx->itx_lr; uint64_t txg = lrc->lrc_txg; @@ -3030,7 +3202,7 @@ zil_process_commit_list(zilog_t *zilog, zil_commit_waiter_t *zcw, list_t *ilwbs) } else { ASSERT3S(lrc->lrc_txtype, !=, TX_COMMIT); zilog->zl_cur_left -= zil_itx_full_size(itx); - zil_itx_destroy(itx); + zil_itx_destroy(itx, 0); } } @@ -3041,9 +3213,14 @@ zil_process_commit_list(zilog_t *zilog, zil_commit_waiter_t *zcw, list_t *ilwbs) * the ZIL write pipeline; see the comment within * zil_commit_writer_stall() for more details. */ - while ((lwb = list_remove_head(ilwbs)) != NULL) - zil_lwb_write_issue(zilog, lwb); - zil_commit_writer_stall(zilog); + int err = 0; + while ((lwb = list_remove_head(ilwbs)) != NULL) { + err = zil_lwb_write_issue(zilog, lwb); + if (err != 0) + break; + } + if (err == 0) + err = zil_commit_writer_stall(zilog); /* * Additionally, we have to signal and mark the "nolwb" @@ -3061,7 +3238,7 @@ zil_process_commit_list(zilog_t *zilog, zil_commit_waiter_t *zcw, list_t *ilwbs) * the itx's callback if one exists for the itx. */ while ((itx = list_remove_head(&nolwb_itxs)) != NULL) - zil_itx_destroy(itx); + zil_itx_destroy(itx, 0); } else { ASSERT(list_is_empty(&nolwb_waiters)); ASSERT3P(lwb, !=, NULL); @@ -3111,14 +3288,21 @@ zil_process_commit_list(zilog_t *zilog, zil_commit_waiter_t *zcw, list_t *ilwbs) * possible, without significantly impacting the latency * of each individual itx. */ - if (lwb->lwb_state == LWB_STATE_OPENED && !zilog->zl_parallel) { + if (lwb->lwb_state == LWB_STATE_OPENED && + (!zilog->zl_parallel || zilog->zl_suspend > 0)) { zil_burst_done(zilog); list_insert_tail(ilwbs, lwb); lwb = zil_lwb_write_close(zilog, lwb, LWB_STATE_NEW); if (lwb == NULL) { - while ((lwb = list_remove_head(ilwbs)) != NULL) - zil_lwb_write_issue(zilog, lwb); - zil_commit_writer_stall(zilog); + int err = 0; + while ((lwb = + list_remove_head(ilwbs)) != NULL) { + err = zil_lwb_write_issue(zilog, lwb); + if (err != 0) + break; + } + if (err == 0) + zil_commit_writer_stall(zilog); } } } @@ -3177,10 +3361,23 @@ zil_commit_writer(zilog_t *zilog, zil_commit_waiter_t *zcw) zil_prune_commit_list(zilog); zil_process_commit_list(zilog, zcw, &ilwbs); + /* + * If the ZIL failed somewhere inside zil_process_commit_list(), it's + * will be because a fallback to txg_wait_sync_flags() happened at some + * point (eg zil_commit_writer_stall()). All cases should issue and + * empty ilwbs, so there will be nothing to in the issue loop below. + * That's why we don't have to plumb the error value back from + * zil_process_commit_list(), and don't have to skip it. + */ + IMPLY(zilog->zl_restart_txg > 0, list_is_empty(&ilwbs)); + out: mutex_exit(&zilog->zl_issuer_lock); - while ((lwb = list_remove_head(&ilwbs)) != NULL) - zil_lwb_write_issue(zilog, lwb); + int err = 0; + while ((lwb = list_remove_head(&ilwbs)) != NULL) { + if (err == 0) + err = zil_lwb_write_issue(zilog, lwb); + } list_destroy(&ilwbs); return (wtxg); } @@ -3436,7 +3633,7 @@ static void zil_free_commit_waiter(zil_commit_waiter_t *zcw) { ASSERT(!list_link_active(&zcw->zcw_node)); - ASSERT3P(zcw->zcw_lwb, ==, NULL); + ASSERT0P(zcw->zcw_lwb); ASSERT3B(zcw->zcw_done, ==, B_TRUE); mutex_destroy(&zcw->zcw_lock); cv_destroy(&zcw->zcw_cv); @@ -3473,6 +3670,96 @@ zil_commit_itx_assign(zilog_t *zilog, zil_commit_waiter_t *zcw) } /* + * Crash the ZIL. This is something like suspending, but abandons the ZIL + * without further IO until the wanted txg completes. No effort is made to + * close the on-disk chain or do any other on-disk work, as the pool may + * have suspended. zil_sync() will handle cleanup as normal and restart the + * ZIL once enough txgs have passed. + */ +static void +zil_crash(zilog_t *zilog) +{ + mutex_enter(&zilog->zl_lock); + + uint64_t txg = spa_syncing_txg(zilog->zl_spa); + uint64_t restart_txg = + spa_syncing_txg(zilog->zl_spa) + TXG_CONCURRENT_STATES; + + if (zilog->zl_restart_txg > 0) { + /* + * If the ZIL is already crashed, it's almost certainly because + * we lost a race involving multiple callers from + * zil_commit_impl(). + */ + + /* + * This sanity check is to support my understanding that in the + * event of multiple callers to zil_crash(), only one of them + * can possibly be in the codepath to issue lwbs; the rest + * should be calling from zil_commit_impl() after their waiters + * have completed. As I understand it, a second thread trying + * to issue will eventually wait on zl_issuer_lock, and then + * have no work to do and leave. + * + * If more lwbs had been created an issued between zil_crash() + * calls, then we probably just need to take those too, add + * them to the crash list and clean them up, but it complicates + * this function and I don't think it can happend. + */ + ASSERT(list_is_empty(&zilog->zl_lwb_list)); + + mutex_exit(&zilog->zl_lock); + return; + } + + zilog->zl_restart_txg = restart_txg; + + /* + * Capture any live LWBs. Depending on the state of the pool they may + * represent in-flight IO that won't return for some time, and we want + * to make sure they don't get in the way of normal ZIL operation. + */ + ASSERT(list_is_empty(&zilog->zl_lwb_crash_list)); + list_move_tail(&zilog->zl_lwb_crash_list, &zilog->zl_lwb_list); + + /* + * Run through the LWB list; erroring all itxes and signalling error + * to all waiters. + */ + for (lwb_t *lwb = list_head(&zilog->zl_lwb_crash_list); lwb != NULL; + lwb = list_next(&zilog->zl_lwb_crash_list, lwb)) { + itx_t *itx; + while ((itx = list_remove_head(&lwb->lwb_itxs)) != NULL) + zil_itx_destroy(itx, EIO); + + zil_commit_waiter_t *zcw; + while ((zcw = list_remove_head(&lwb->lwb_waiters)) != NULL) { + mutex_enter(&zcw->zcw_lock); + zcw->zcw_lwb = NULL; + zcw->zcw_zio_error = EIO; + zcw->zcw_done = B_TRUE; + cv_broadcast(&zcw->zcw_cv); + mutex_exit(&zcw->zcw_lock); + } + } + + /* + * Zero the ZIL header bp after the ZIL restarts. We'll free it in + * zil_clean() when we clean up the lwbs. + */ + zil_header_t *zh = zil_header_in_syncing_context(zilog); + BP_ZERO(&zh->zh_log); + + /* + * Mark this ZIL dirty on the next txg, so that zil_clean() will be + * called for cleanup. + */ + zilog_dirty(zilog, txg+1); + + mutex_exit(&zilog->zl_lock); +} + +/* * Commit ZFS Intent Log transactions (itxs) to stable storage. * * When writing ZIL transactions to the on-disk representation of the @@ -3587,9 +3874,17 @@ zil_commit_itx_assign(zilog_t *zilog, zil_commit_waiter_t *zcw) * but the order in which they complete will be the same order in * which they were created. */ -void +static int zil_commit_impl(zilog_t *zilog, uint64_t foid); + +int zil_commit(zilog_t *zilog, uint64_t foid) { + return (zil_commit_flags(zilog, foid, ZIL_COMMIT_FAILMODE)); +} + +int +zil_commit_flags(zilog_t *zilog, uint64_t foid, zil_commit_flag_t flags) +{ /* * We should never attempt to call zil_commit on a snapshot for * a couple of reasons: @@ -3606,7 +3901,7 @@ zil_commit(zilog_t *zilog, uint64_t foid) ASSERT3B(dmu_objset_is_snapshot(zilog->zl_os), ==, B_FALSE); if (zilog->zl_sync == ZFS_SYNC_DISABLED) - return; + return (0); if (!spa_writeable(zilog->zl_spa)) { /* @@ -3617,10 +3912,23 @@ zil_commit(zilog_t *zilog, uint64_t foid) * verifying that truth before we return to the caller. */ ASSERT(list_is_empty(&zilog->zl_lwb_list)); - ASSERT3P(zilog->zl_last_lwb_opened, ==, NULL); + ASSERT0P(zilog->zl_last_lwb_opened); for (int i = 0; i < TXG_SIZE; i++) - ASSERT3P(zilog->zl_itxg[i].itxg_itxs, ==, NULL); - return; + ASSERT0P(zilog->zl_itxg[i].itxg_itxs); + return (0); + } + + int err = 0; + + /* + * If the ZIL crashed, bypass it entirely, and rely on txg_wait_sync() + * to get the data out to disk. + */ + if (zilog->zl_restart_txg > 0) { + ZIL_STAT_BUMP(zilog, zil_commit_crash_count); + err = txg_wait_synced_flags(zilog->zl_dmu_pool, 0, + TXG_WAIT_SUSPEND); + goto out; } /* @@ -3632,14 +3940,43 @@ zil_commit(zilog_t *zilog, uint64_t foid) */ if (zilog->zl_suspend > 0) { ZIL_STAT_BUMP(zilog, zil_commit_suspend_count); - txg_wait_synced(zilog->zl_dmu_pool, 0); - return; + err = txg_wait_synced_flags(zilog->zl_dmu_pool, 0, + TXG_WAIT_SUSPEND); + if (err != 0) { + ASSERT3U(err, ==, ESHUTDOWN); + zil_crash(zilog); + } + goto out; } - zil_commit_impl(zilog, foid); + err = zil_commit_impl(zilog, foid); + +out: + if (err == 0) + return (0); + + /* + * The ZIL write failed and the pool is suspended. There's nothing else + * we can do except return or block. + */ + ASSERT3U(err, ==, ESHUTDOWN); + + /* + * Return error if failmode=continue or caller will handle directly. + */ + if (!(flags & ZIL_COMMIT_FAILMODE) || + spa_get_failmode(zilog->zl_spa) == ZIO_FAILURE_MODE_CONTINUE) + return (SET_ERROR(EIO)); + + /* + * Block until the pool returns. We assume that the data will make + * it out to disk in the end, and so return success. + */ + txg_wait_synced(zilog->zl_dmu_pool, 0); + return (0); } -void +static int zil_commit_impl(zilog_t *zilog, uint64_t foid) { ZIL_STAT_BUMP(zilog, zil_commit_count); @@ -3676,6 +4013,7 @@ zil_commit_impl(zilog_t *zilog, uint64_t foid) uint64_t wtxg = zil_commit_writer(zilog, zcw); zil_commit_waiter(zilog, zcw); + int err = 0; if (zcw->zcw_zio_error != 0) { /* * If there was an error writing out the ZIL blocks that @@ -3688,13 +4026,29 @@ zil_commit_impl(zilog_t *zilog, uint64_t foid) ZIL_STAT_BUMP(zilog, zil_commit_error_count); DTRACE_PROBE2(zil__commit__io__error, zilog_t *, zilog, zil_commit_waiter_t *, zcw); - txg_wait_synced(zilog->zl_dmu_pool, 0); + err = txg_wait_synced_flags(zilog->zl_dmu_pool, 0, + TXG_WAIT_SUSPEND); } else if (wtxg != 0) { ZIL_STAT_BUMP(zilog, zil_commit_suspend_count); - txg_wait_synced(zilog->zl_dmu_pool, wtxg); + err = txg_wait_synced_flags(zilog->zl_dmu_pool, wtxg, + TXG_WAIT_SUSPEND); } zil_free_commit_waiter(zcw); + + if (err == 0) + return (0); + + /* + * ZIL write failed and pool failed in the fallback to + * txg_wait_synced_flags(). Right now we have no idea if the data is on + * disk and the pool is probably suspended so we have no idea when it's + * coming back. All we can do is shut down and return error to the + * caller. + */ + ASSERT3U(err, ==, ESHUTDOWN); + zil_crash(zilog); + return (err); } /* @@ -3720,7 +4074,7 @@ zil_sync(zilog_t *zilog, dmu_tx_t *tx) mutex_enter(&zilog->zl_lock); - ASSERT(zilog->zl_stop_sync == 0); + ASSERT0(zilog->zl_stop_sync); if (*replayed_seq != 0) { ASSERT(zh->zh_replay_seq < *replayed_seq); @@ -3890,6 +4244,8 @@ zil_alloc(objset_t *os, zil_header_t *zh_phys) list_create(&zilog->zl_lwb_list, sizeof (lwb_t), offsetof(lwb_t, lwb_node)); + list_create(&zilog->zl_lwb_crash_list, sizeof (lwb_t), + offsetof(lwb_t, lwb_node)); list_create(&zilog->zl_itx_commit_list, sizeof (itx_t), offsetof(itx_t, itx_node)); @@ -3914,9 +4270,12 @@ zil_free(zilog_t *zilog) ASSERT0(zilog->zl_suspend); ASSERT0(zilog->zl_suspending); + ASSERT0(zilog->zl_restart_txg); ASSERT(list_is_empty(&zilog->zl_lwb_list)); list_destroy(&zilog->zl_lwb_list); + ASSERT(list_is_empty(&zilog->zl_lwb_crash_list)); + list_destroy(&zilog->zl_lwb_crash_list); ASSERT(list_is_empty(&zilog->zl_itx_commit_list)); list_destroy(&zilog->zl_itx_commit_list); @@ -3952,8 +4311,8 @@ zil_open(objset_t *os, zil_get_data_t *get_data, zil_sums_t *zil_sums) { zilog_t *zilog = dmu_objset_zil(os); - ASSERT3P(zilog->zl_get_data, ==, NULL); - ASSERT3P(zilog->zl_last_lwb_opened, ==, NULL); + ASSERT0P(zilog->zl_get_data); + ASSERT0P(zilog->zl_last_lwb_opened); ASSERT(list_is_empty(&zilog->zl_lwb_list)); zilog->zl_get_data = get_data; @@ -3972,7 +4331,8 @@ zil_close(zilog_t *zilog) uint64_t txg; if (!dmu_objset_is_snapshot(zilog->zl_os)) { - zil_commit(zilog, 0); + if (zil_commit_flags(zilog, 0, ZIL_COMMIT_NOW) != 0) + txg_wait_synced(zilog->zl_dmu_pool, 0); } else { ASSERT(list_is_empty(&zilog->zl_lwb_list)); ASSERT0(zilog->zl_dirty_max_txg); @@ -4073,6 +4433,17 @@ zil_suspend(const char *osname, void **cookiep) return (SET_ERROR(EBUSY)); } + if (zilog->zl_restart_txg > 0) { + /* + * ZIL crashed. It effectively _is_ suspended, but callers + * are usually trying to make sure it's empty on-disk, which + * we can't guarantee right now. + */ + mutex_exit(&zilog->zl_lock); + dmu_objset_rele(os, suspend_tag); + return (SET_ERROR(EBUSY)); + } + /* * Don't put a long hold in the cases where we can avoid it. This * is when there is no cookie so we are doing a suspend & resume @@ -4105,6 +4476,11 @@ zil_suspend(const char *osname, void **cookiep) zil_resume(os); else *cookiep = os; + + if (zilog->zl_restart_txg > 0) + /* ZIL crashed while we were waiting. */ + return (SET_ERROR(EBUSY)); + return (0); } @@ -4146,17 +4522,34 @@ zil_suspend(const char *osname, void **cookiep) * would just call txg_wait_synced(), because zl_suspend is set. * txg_wait_synced() doesn't wait for these lwb's to be * LWB_STATE_FLUSH_DONE before returning. + * + * However, zil_commit_impl() itself can return an error if any of the + * lwbs fail, or the pool suspends in the fallback + * txg_wait_sync_flushed(), which affects what we do next, so we + * capture that error. */ - zil_commit_impl(zilog, 0); + error = zil_commit_impl(zilog, 0); + if (error == ESHUTDOWN) + /* zil_commit_impl() has called zil_crash() already */ + error = SET_ERROR(EBUSY); /* * Now that we've ensured all lwb's are LWB_STATE_FLUSH_DONE, we * use txg_wait_synced() to ensure the data from the zilog has * migrated to the main pool before calling zil_destroy(). */ - txg_wait_synced(zilog->zl_dmu_pool, 0); + if (error == 0) { + error = txg_wait_synced_flags(zilog->zl_dmu_pool, 0, + TXG_WAIT_SUSPEND); + if (error != 0) { + ASSERT3U(error, ==, ESHUTDOWN); + zil_crash(zilog); + error = SET_ERROR(EBUSY); + } + } - zil_destroy(zilog, B_FALSE); + if (error == 0) + zil_destroy(zilog, B_FALSE); mutex_enter(&zilog->zl_lock); zilog->zl_suspending = B_FALSE; @@ -4170,7 +4563,8 @@ zil_suspend(const char *osname, void **cookiep) zil_resume(os); else *cookiep = os; - return (0); + + return (error); } void @@ -4333,7 +4727,7 @@ zil_replay(objset_t *os, void *arg, zilog->zl_replay = B_TRUE; zilog->zl_replay_time = ddi_get_lbolt(); - ASSERT(zilog->zl_replay_blks == 0); + ASSERT0(zilog->zl_replay_blks); (void) zil_parse(zilog, zil_incr_blks, zil_replay_log_record, &zr, zh->zh_claim_txg, B_TRUE); vmem_free(zr.zr_lr, 2 * SPA_MAXBLOCKSIZE); @@ -4418,3 +4812,9 @@ ZFS_MODULE_PARAM(zfs_zil, zil_, maxblocksize, UINT, ZMOD_RW, ZFS_MODULE_PARAM(zfs_zil, zil_, maxcopied, UINT, ZMOD_RW, "Limit in bytes WR_COPIED size"); + +ZFS_MODULE_PARAM(zfs, zfs_, immediate_write_sz, UINT, ZMOD_RW, + "Largest write size to store data into ZIL"); + +ZFS_MODULE_PARAM(zfs_zil, zil_, special_is_slog, INT, ZMOD_RW, + "Treat special vdevs as SLOG"); diff --git a/sys/contrib/openzfs/module/zfs/zio.c b/sys/contrib/openzfs/module/zfs/zio.c index 6d7bce8b0e10..3f0ddb63249d 100644 --- a/sys/contrib/openzfs/module/zfs/zio.c +++ b/sys/contrib/openzfs/module/zfs/zio.c @@ -339,8 +339,8 @@ zio_fini(void) } for (size_t i = 0; i < n; i++) { - VERIFY3P(zio_buf_cache[i], ==, NULL); - VERIFY3P(zio_data_buf_cache[i], ==, NULL); + VERIFY0P(zio_buf_cache[i]); + VERIFY0P(zio_data_buf_cache[i]); } if (zio_ksp != NULL) { @@ -692,7 +692,7 @@ error: zio->io_error = SET_ERROR(EIO); if ((zio->io_flags & ZIO_FLAG_SPECULATIVE) == 0) { spa_log_error(spa, &zio->io_bookmark, - BP_GET_LOGICAL_BIRTH(zio->io_bp)); + BP_GET_PHYSICAL_BIRTH(zio->io_bp)); (void) zfs_ereport_post(FM_EREPORT_ZFS_AUTHENTICATION, spa, NULL, &zio->io_bookmark, zio, 0); } @@ -771,7 +771,7 @@ zio_add_child_impl(zio_t *pio, zio_t *cio, boolean_t first) else mutex_enter(&cio->io_lock); - ASSERT(pio->io_state[ZIO_WAIT_DONE] == 0); + ASSERT0(pio->io_state[ZIO_WAIT_DONE]); uint64_t *countp = pio->io_children[cio->io_child_type]; for (int w = 0; w < ZIO_WAIT_TYPES; w++) @@ -821,7 +821,7 @@ zio_wait_for_children(zio_t *zio, uint8_t childbits, enum zio_wait_type wait) boolean_t waiting = B_FALSE; mutex_enter(&zio->io_lock); - ASSERT(zio->io_stall == NULL); + ASSERT0P(zio->io_stall); for (int c = 0; c < ZIO_CHILD_TYPES; c++) { if (!(ZIO_CHILD_BIT_IS_SET(childbits, c))) continue; @@ -850,15 +850,9 @@ zio_notify_parent(zio_t *pio, zio_t *zio, enum zio_wait_type wait, mutex_enter(&pio->io_lock); if (zio->io_error && !(zio->io_flags & ZIO_FLAG_DONT_PROPAGATE)) *errorp = zio_worst_error(*errorp, zio->io_error); - pio->io_reexecute |= zio->io_reexecute; + pio->io_post |= zio->io_post; ASSERT3U(*countp, >, 0); - /* - * Propogate the Direct I/O checksum verify failure to the parent. - */ - if (zio->io_flags & ZIO_FLAG_DIO_CHKSUM_ERR) - pio->io_flags |= ZIO_FLAG_DIO_CHKSUM_ERR; - (*countp)--; if (*countp == 0 && pio->io_stall == countp) { @@ -961,8 +955,8 @@ zio_create(zio_t *pio, spa_t *spa, uint64_t txg, const blkptr_t *bp, zio_t *zio; IMPLY(type != ZIO_TYPE_TRIM, psize <= SPA_MAXBLOCKSIZE); - ASSERT(P2PHASE(psize, SPA_MINBLOCKSIZE) == 0); - ASSERT(P2PHASE(offset, SPA_MINBLOCKSIZE) == 0); + ASSERT0(P2PHASE(psize, SPA_MINBLOCKSIZE)); + ASSERT0(P2PHASE(offset, SPA_MINBLOCKSIZE)); ASSERT(!vd || spa_config_held(spa, SCL_STATE_ALL, RW_READER)); ASSERT(!bp || !(flags & ZIO_FLAG_CONFIG_WRITER)); @@ -1110,7 +1104,8 @@ zfs_blkptr_verify_log(spa_t *spa, const blkptr_t *bp, "DVA[1]=%#llx/%#llx " "DVA[2]=%#llx/%#llx " "prop=%#llx " - "pad=%#llx,%#llx " + "prop2=%#llx " + "pad=%#llx " "phys_birth=%#llx " "birth=%#llx " "fill=%#llx " @@ -1123,9 +1118,9 @@ zfs_blkptr_verify_log(spa_t *spa, const blkptr_t *bp, (long long)bp->blk_dva[2].dva_word[0], (long long)bp->blk_dva[2].dva_word[1], (long long)bp->blk_prop, - (long long)bp->blk_pad[0], - (long long)bp->blk_pad[1], - (long long)BP_GET_PHYSICAL_BIRTH(bp), + (long long)bp->blk_prop2, + (long long)bp->blk_pad, + (long long)BP_GET_RAW_PHYSICAL_BIRTH(bp), (long long)BP_GET_LOGICAL_BIRTH(bp), (long long)bp->blk_fill, (long long)bp->blk_cksum.zc_word[0], @@ -1340,7 +1335,7 @@ zio_read(zio_t *pio, spa_t *spa, const blkptr_t *bp, { zio_t *zio; - zio = zio_create(pio, spa, BP_GET_BIRTH(bp), bp, + zio = zio_create(pio, spa, BP_GET_PHYSICAL_BIRTH(bp), bp, data, size, size, done, private, ZIO_TYPE_READ, priority, flags, NULL, 0, zb, ZIO_STAGE_OPEN, (flags & ZIO_FLAG_DDT_CHILD) ? @@ -1456,7 +1451,7 @@ zio_free(spa_t *spa, uint64_t txg, const blkptr_t *bp) metaslab_check_free(spa, bp); bplist_append(&spa->spa_free_bplist[txg & TXG_MASK], bp); } else { - VERIFY3P(zio_free_sync(NULL, spa, txg, bp, 0), ==, NULL); + VERIFY0P(zio_free_sync(NULL, spa, txg, bp, 0)); } } @@ -1564,7 +1559,7 @@ zio_read_phys(zio_t *pio, vdev_t *vd, uint64_t offset, uint64_t size, { zio_t *zio; - ASSERT(vd->vdev_children == 0); + ASSERT0(vd->vdev_children); ASSERT(!labels || offset + size <= VDEV_LABEL_START_SIZE || offset >= vd->vdev_psize - VDEV_LABEL_END_SIZE); ASSERT3U(offset + size, <=, vd->vdev_psize); @@ -1585,7 +1580,7 @@ zio_write_phys(zio_t *pio, vdev_t *vd, uint64_t offset, uint64_t size, { zio_t *zio; - ASSERT(vd->vdev_children == 0); + ASSERT0(vd->vdev_children); ASSERT(!labels || offset + size <= VDEV_LABEL_START_SIZE || offset >= vd->vdev_psize - VDEV_LABEL_END_SIZE); ASSERT3U(offset + size, <=, vd->vdev_psize); @@ -1649,7 +1644,7 @@ zio_vdev_child_io(zio_t *pio, blkptr_t *bp, vdev_t *vd, uint64_t offset, * through the mirror during self healing. See comment in * vdev_mirror_io_done() for more details. */ - ASSERT0(pio->io_flags & ZIO_FLAG_DIO_CHKSUM_ERR); + ASSERT0(pio->io_post & ZIO_POST_DIO_CHKSUM_ERR); } else if (type == ZIO_TYPE_WRITE && pio->io_prop.zp_direct_write == B_TRUE) { /* @@ -1685,7 +1680,7 @@ zio_vdev_child_io(zio_t *pio, blkptr_t *bp, vdev_t *vd, uint64_t offset, * If this is a retried I/O then we ignore it since we will * have already processed the original allocating I/O. */ - if (flags & ZIO_FLAG_IO_ALLOCATING && + if (flags & ZIO_FLAG_ALLOC_THROTTLED && (vd != vd->vdev_top || (flags & ZIO_FLAG_IO_RETRY))) { ASSERT(pio->io_metaslab_class != NULL); ASSERT(pio->io_metaslab_class->mc_alloc_throttle_enabled); @@ -1695,7 +1690,7 @@ zio_vdev_child_io(zio_t *pio, blkptr_t *bp, vdev_t *vd, uint64_t offset, ASSERT(!(pio->io_flags & ZIO_FLAG_IO_REWRITE) || pio->io_child_type == ZIO_CHILD_GANG); - flags &= ~ZIO_FLAG_IO_ALLOCATING; + flags &= ~ZIO_FLAG_ALLOC_THROTTLED; } zio = zio_create(pio, pio->io_spa, pio->io_txg, bp, data, size, size, @@ -1752,7 +1747,7 @@ zio_flush(zio_t *pio, vdev_t *vd) void zio_shrink(zio_t *zio, uint64_t size) { - ASSERT3P(zio->io_executor, ==, NULL); + ASSERT0P(zio->io_executor); ASSERT3U(zio->io_orig_size, ==, zio->io_size); ASSERT3U(size, <=, zio->io_size); @@ -1860,7 +1855,7 @@ zio_write_bp_init(zio_t *zio) blkptr_t *bp = zio->io_bp; zio_prop_t *zp = &zio->io_prop; - ASSERT(BP_GET_LOGICAL_BIRTH(bp) != zio->io_txg); + ASSERT(BP_GET_BIRTH(bp) != zio->io_txg); *bp = *zio->io_bp_override; zio->io_pipeline = ZIO_INTERLOCK_PIPELINE; @@ -1946,9 +1941,9 @@ zio_write_compress(zio_t *zio) } ASSERT(zio->io_child_type != ZIO_CHILD_DDT); - ASSERT(zio->io_bp_override == NULL); + ASSERT0P(zio->io_bp_override); - if (!BP_IS_HOLE(bp) && BP_GET_LOGICAL_BIRTH(bp) == zio->io_txg) { + if (!BP_IS_HOLE(bp) && BP_GET_BIRTH(bp) == zio->io_txg) { /* * We're rewriting an existing block, which means we're * working on behalf of spa_sync(). For spa_sync() to @@ -2085,7 +2080,7 @@ zio_write_compress(zio_t *zio) * spa_sync() to allocate new blocks, but force rewrites after that. * There should only be a handful of blocks after pass 1 in any case. */ - if (!BP_IS_HOLE(bp) && BP_GET_LOGICAL_BIRTH(bp) == zio->io_txg && + if (!BP_IS_HOLE(bp) && BP_GET_BIRTH(bp) == zio->io_txg && BP_GET_PSIZE(bp) == psize && pass >= zfs_sync_pass_rewrite) { VERIFY3U(psize, !=, 0); @@ -2441,7 +2436,7 @@ __zio_execute(zio_t *zio) ASSERT(!MUTEX_HELD(&zio->io_lock)); ASSERT(ISP2(stage)); - ASSERT(zio->io_stall == NULL); + ASSERT0P(zio->io_stall); do { stage <<= 1; @@ -2514,7 +2509,7 @@ zio_wait(zio_t *zio) int error; ASSERT3S(zio->io_stage, ==, ZIO_STAGE_OPEN); - ASSERT3P(zio->io_executor, ==, NULL); + ASSERT0P(zio->io_executor); zio->io_waiter = curthread; ASSERT0(zio->io_queued_timestamp); @@ -2556,7 +2551,7 @@ zio_nowait(zio_t *zio) if (zio == NULL) return; - ASSERT3P(zio->io_executor, ==, NULL); + ASSERT0P(zio->io_executor); if (zio->io_child_type == ZIO_CHILD_LOGICAL && list_is_empty(&zio->io_parent_list)) { @@ -2595,14 +2590,14 @@ zio_reexecute(void *arg) ASSERT(pio->io_child_type == ZIO_CHILD_LOGICAL); ASSERT(pio->io_orig_stage == ZIO_STAGE_OPEN); - ASSERT(pio->io_gang_leader == NULL); - ASSERT(pio->io_gang_tree == NULL); + ASSERT0P(pio->io_gang_leader); + ASSERT0P(pio->io_gang_tree); mutex_enter(&pio->io_lock); pio->io_flags = pio->io_orig_flags; pio->io_stage = pio->io_orig_stage; pio->io_pipeline = pio->io_orig_pipeline; - pio->io_reexecute = 0; + pio->io_post = 0; pio->io_flags |= ZIO_FLAG_REEXECUTED; pio->io_pipeline_trace = 0; pio->io_error = 0; @@ -2694,7 +2689,7 @@ zio_suspend(spa_t *spa, zio_t *zio, zio_suspend_reason_t reason) ASSERT(!(zio->io_flags & ZIO_FLAG_GODFATHER)); ASSERT(zio != spa->spa_suspend_zio_root); ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL); - ASSERT(zio_unique_parent(zio) == NULL); + ASSERT0P(zio_unique_parent(zio)); ASSERT(zio->io_stage == ZIO_STAGE_DONE); zio_add_child(spa->spa_suspend_zio_root, zio); } @@ -2749,11 +2744,14 @@ zio_resume_wait(spa_t *spa) * being nearly full, it calls zio_write_gang_block() to construct the * block from smaller fragments. * - * A gang block consists of a gang header (zio_gbh_phys_t) and up to - * three (SPA_GBH_NBLKPTRS) gang members. The gang header is just like - * an indirect block: it's an array of block pointers. It consumes - * only one sector and hence is allocatable regardless of fragmentation. - * The gang header's bps point to its gang members, which hold the data. + * A gang block consists of a a gang header and up to gbh_nblkptrs(size) + * gang members. The gang header is like an indirect block: it's an array + * of block pointers, though the header has a small tail (a zio_eck_t) + * that stores an embedded checksum. It is allocated using only a single + * sector as the requested size, and hence is allocatable regardless of + * fragmentation. Its size is determined by the smallest allocatable + * asize of the vdevs it was allocated on. The gang header's bps point + * to its gang members, which hold the data. * * Gang blocks are self-checksumming, using the bp's <vdev, offset, txg> * as the verifier to ensure uniqueness of the SHA256 checksum. @@ -2832,10 +2830,10 @@ zio_rewrite_gang(zio_t *pio, blkptr_t *bp, zio_gang_node_t *gn, abd_t *data, if (gn != NULL) { abd_t *gbh_abd = - abd_get_from_buf(gn->gn_gbh, SPA_GANGBLOCKSIZE); + abd_get_from_buf(gn->gn_gbh, gn->gn_gangblocksize); zio = zio_rewrite(pio, pio->io_spa, pio->io_txg, bp, - gbh_abd, SPA_GANGBLOCKSIZE, zio_gang_issue_func_done, NULL, - pio->io_priority, ZIO_GANG_CHILD_FLAGS(pio), + gbh_abd, gn->gn_gangblocksize, zio_gang_issue_func_done, + NULL, pio->io_priority, ZIO_GANG_CHILD_FLAGS(pio), &pio->io_bookmark); /* * As we rewrite each gang header, the pipeline will compute @@ -2906,14 +2904,16 @@ static zio_gang_issue_func_t *zio_gang_issue_func[ZIO_TYPES] = { static void zio_gang_tree_assemble_done(zio_t *zio); static zio_gang_node_t * -zio_gang_node_alloc(zio_gang_node_t **gnpp) +zio_gang_node_alloc(zio_gang_node_t **gnpp, uint64_t gangblocksize) { zio_gang_node_t *gn; - ASSERT(*gnpp == NULL); + ASSERT0P(*gnpp); - gn = kmem_zalloc(sizeof (*gn), KM_SLEEP); - gn->gn_gbh = zio_buf_alloc(SPA_GANGBLOCKSIZE); + gn = kmem_zalloc(sizeof (*gn) + + (gbh_nblkptrs(gangblocksize) * sizeof (gn)), KM_SLEEP); + gn->gn_gangblocksize = gn->gn_allocsize = gangblocksize; + gn->gn_gbh = zio_buf_alloc(gangblocksize); *gnpp = gn; return (gn); @@ -2924,11 +2924,12 @@ zio_gang_node_free(zio_gang_node_t **gnpp) { zio_gang_node_t *gn = *gnpp; - for (int g = 0; g < SPA_GBH_NBLKPTRS; g++) - ASSERT(gn->gn_child[g] == NULL); + for (int g = 0; g < gbh_nblkptrs(gn->gn_allocsize); g++) + ASSERT0P(gn->gn_child[g]); - zio_buf_free(gn->gn_gbh, SPA_GANGBLOCKSIZE); - kmem_free(gn, sizeof (*gn)); + zio_buf_free(gn->gn_gbh, gn->gn_allocsize); + kmem_free(gn, sizeof (*gn) + + (gbh_nblkptrs(gn->gn_allocsize) * sizeof (gn))); *gnpp = NULL; } @@ -2940,7 +2941,7 @@ zio_gang_tree_free(zio_gang_node_t **gnpp) if (gn == NULL) return; - for (int g = 0; g < SPA_GBH_NBLKPTRS; g++) + for (int g = 0; g < gbh_nblkptrs(gn->gn_allocsize); g++) zio_gang_tree_free(&gn->gn_child[g]); zio_gang_node_free(gnpp); @@ -2949,13 +2950,28 @@ zio_gang_tree_free(zio_gang_node_t **gnpp) static void zio_gang_tree_assemble(zio_t *gio, blkptr_t *bp, zio_gang_node_t **gnpp) { - zio_gang_node_t *gn = zio_gang_node_alloc(gnpp); - abd_t *gbh_abd = abd_get_from_buf(gn->gn_gbh, SPA_GANGBLOCKSIZE); + uint64_t gangblocksize = UINT64_MAX; + if (spa_feature_is_active(gio->io_spa, + SPA_FEATURE_DYNAMIC_GANG_HEADER)) { + spa_config_enter(gio->io_spa, SCL_VDEV, FTAG, RW_READER); + for (int dva = 0; dva < BP_GET_NDVAS(bp); dva++) { + vdev_t *vd = vdev_lookup_top(gio->io_spa, + DVA_GET_VDEV(&bp->blk_dva[dva])); + uint64_t psize = vdev_gang_header_psize(vd); + gangblocksize = MIN(gangblocksize, psize); + } + spa_config_exit(gio->io_spa, SCL_VDEV, FTAG); + } else { + gangblocksize = SPA_OLD_GANGBLOCKSIZE; + } + ASSERT3U(gangblocksize, !=, UINT64_MAX); + zio_gang_node_t *gn = zio_gang_node_alloc(gnpp, gangblocksize); + abd_t *gbh_abd = abd_get_from_buf(gn->gn_gbh, gangblocksize); ASSERT(gio->io_gang_leader == gio); ASSERT(BP_IS_GANG(bp)); - zio_nowait(zio_read(gio, gio->io_spa, bp, gbh_abd, SPA_GANGBLOCKSIZE, + zio_nowait(zio_read(gio, gio->io_spa, bp, gbh_abd, gangblocksize, zio_gang_tree_assemble_done, gn, gio->io_priority, ZIO_GANG_CHILD_FLAGS(gio), &gio->io_bookmark)); } @@ -2978,13 +2994,17 @@ zio_gang_tree_assemble_done(zio_t *zio) byteswap_uint64_array(abd_to_buf(zio->io_abd), zio->io_size); ASSERT3P(abd_to_buf(zio->io_abd), ==, gn->gn_gbh); - ASSERT(zio->io_size == SPA_GANGBLOCKSIZE); - ASSERT(gn->gn_gbh->zg_tail.zec_magic == ZEC_MAGIC); + /* + * If this was an old-style gangblock, the gangblocksize should have + * been updated in zio_checksum_error to reflect that. + */ + ASSERT3U(gbh_eck(gn->gn_gbh, gn->gn_gangblocksize)->zec_magic, + ==, ZEC_MAGIC); abd_free(zio->io_abd); - for (int g = 0; g < SPA_GBH_NBLKPTRS; g++) { - blkptr_t *gbp = &gn->gn_gbh->zg_blkptr[g]; + for (int g = 0; g < gbh_nblkptrs(gn->gn_gangblocksize); g++) { + blkptr_t *gbp = gbh_bp(gn->gn_gbh, g); if (!BP_IS_GANG(gbp)) continue; zio_gang_tree_assemble(gio, gbp, &gn->gn_child[g]); @@ -3009,10 +3029,11 @@ zio_gang_tree_issue(zio_t *pio, zio_gang_node_t *gn, blkptr_t *bp, abd_t *data, zio = zio_gang_issue_func[gio->io_type](pio, bp, gn, data, offset); if (gn != NULL) { - ASSERT(gn->gn_gbh->zg_tail.zec_magic == ZEC_MAGIC); + ASSERT3U(gbh_eck(gn->gn_gbh, + gn->gn_gangblocksize)->zec_magic, ==, ZEC_MAGIC); - for (int g = 0; g < SPA_GBH_NBLKPTRS; g++) { - blkptr_t *gbp = &gn->gn_gbh->zg_blkptr[g]; + for (int g = 0; g < gbh_nblkptrs(gn->gn_gangblocksize); g++) { + blkptr_t *gbp = gbh_bp(gn->gn_gbh, g); if (BP_IS_HOLE(gbp)) continue; zio_gang_tree_issue(zio, gn->gn_child[g], gbp, data, @@ -3119,6 +3140,13 @@ zio_write_gang_done(zio_t *zio) abd_free(zio->io_abd); } +static void +zio_update_feature(void *arg, dmu_tx_t *tx) +{ + spa_t *spa = dmu_tx_pool(tx)->dp_spa; + spa_feature_incr(spa, (spa_feature_t)(uintptr_t)arg, tx); +} + static zio_t * zio_write_gang_block(zio_t *pio, metaslab_class_t *mc) { @@ -3157,20 +3185,24 @@ zio_write_gang_block(zio_t *pio, metaslab_class_t *mc) ASSERT(ZIO_HAS_ALLOCATOR(pio)); int flags = METASLAB_GANG_HEADER; - if (pio->io_flags & ZIO_FLAG_IO_ALLOCATING) { + if (pio->io_flags & ZIO_FLAG_ALLOC_THROTTLED) { ASSERT(pio->io_priority == ZIO_PRIORITY_ASYNC_WRITE); ASSERT(has_data); flags |= METASLAB_ASYNC_ALLOC; } - error = metaslab_alloc(spa, mc, SPA_GANGBLOCKSIZE, + uint64_t gangblocksize = SPA_OLD_GANGBLOCKSIZE; + uint64_t candidate = gangblocksize; + error = metaslab_alloc_range(spa, mc, gangblocksize, gangblocksize, bp, gbh_copies, txg, pio == gio ? NULL : gio->io_bp, flags, - &pio->io_alloc_list, pio->io_allocator, pio); + &pio->io_alloc_list, pio->io_allocator, pio, &candidate); if (error) { pio->io_error = error; return (pio); } + if (spa_feature_is_active(spa, SPA_FEATURE_DYNAMIC_GANG_HEADER)) + gangblocksize = candidate; if (pio == gio) { gnpp = &gio->io_gang_tree; @@ -3179,23 +3211,24 @@ zio_write_gang_block(zio_t *pio, metaslab_class_t *mc) ASSERT(pio->io_ready == zio_write_gang_member_ready); } - gn = zio_gang_node_alloc(gnpp); + gn = zio_gang_node_alloc(gnpp, gangblocksize); gbh = gn->gn_gbh; - memset(gbh, 0, SPA_GANGBLOCKSIZE); - gbh_abd = abd_get_from_buf(gbh, SPA_GANGBLOCKSIZE); + memset(gbh, 0, gangblocksize); + gbh_abd = abd_get_from_buf(gbh, gangblocksize); /* * Create the gang header. */ - zio = zio_rewrite(pio, spa, txg, bp, gbh_abd, SPA_GANGBLOCKSIZE, + zio = zio_rewrite(pio, spa, txg, bp, gbh_abd, gangblocksize, zio_write_gang_done, NULL, pio->io_priority, ZIO_GANG_CHILD_FLAGS(pio), &pio->io_bookmark); zio_gang_inherit_allocator(pio, zio); - if (pio->io_flags & ZIO_FLAG_IO_ALLOCATING) { + if (pio->io_flags & ZIO_FLAG_ALLOC_THROTTLED) { boolean_t more; - VERIFY(metaslab_class_throttle_reserve(mc, gbh_copies, - zio, B_TRUE, &more)); + VERIFY(metaslab_class_throttle_reserve(mc, zio->io_allocator, + gbh_copies, zio->io_size, B_TRUE, &more)); + zio->io_flags |= ZIO_FLAG_ALLOC_THROTTLED; } /* @@ -3203,7 +3236,9 @@ zio_write_gang_block(zio_t *pio, metaslab_class_t *mc) * opportunistic allocations. If that fails to generate enough * space, we fall back to normal zio_write calls for nested gang. */ - for (int g = 0; resid != 0; g++) { + int g; + boolean_t any_failed = B_FALSE; + for (g = 0; resid != 0; g++) { flags &= METASLAB_ASYNC_ALLOC; flags |= METASLAB_GANG_CHILD; zp.zp_checksum = gio->io_prop.zp_checksum; @@ -3224,9 +3259,9 @@ zio_write_gang_block(zio_t *pio, metaslab_class_t *mc) memset(zp.zp_mac, 0, ZIO_DATA_MAC_LEN); uint64_t min_size = zio_roundup_alloc_size(spa, - resid / (SPA_GBH_NBLKPTRS - g)); + resid / (gbh_nblkptrs(gangblocksize) - g)); min_size = MIN(min_size, resid); - bp = &gbh->zg_blkptr[g]; + bp = &((blkptr_t *)gbh)[g]; zio_alloc_list_t cio_list; metaslab_trace_init(&cio_list); @@ -3236,6 +3271,7 @@ zio_write_gang_block(zio_t *pio, metaslab_class_t *mc) flags, &cio_list, zio->io_allocator, NULL, &allocated_size); boolean_t allocated = error == 0; + any_failed |= !allocated; uint64_t psize = allocated ? MIN(resid, allocated_size) : min_size; @@ -3268,6 +3304,29 @@ zio_write_gang_block(zio_t *pio, metaslab_class_t *mc) } /* + * If we used more gang children than the old limit, we must already be + * using the new headers. No need to update anything, just move on. + * + * Otherwise, we might be in a case where we need to turn on the new + * feature, so we check that. We enable the new feature if we didn't + * manage to fit everything into 3 gang children and we could have + * written more than that. + */ + if (g > gbh_nblkptrs(SPA_OLD_GANGBLOCKSIZE)) { + ASSERT(spa_feature_is_active(spa, + SPA_FEATURE_DYNAMIC_GANG_HEADER)); + } else if (any_failed && candidate > SPA_OLD_GANGBLOCKSIZE && + spa_feature_is_enabled(spa, SPA_FEATURE_DYNAMIC_GANG_HEADER) && + !spa_feature_is_active(spa, SPA_FEATURE_DYNAMIC_GANG_HEADER)) { + dmu_tx_t *tx = + dmu_tx_create_assigned(spa->spa_dsl_pool, txg + 1); + dsl_sync_task_nowait(spa->spa_dsl_pool, + zio_update_feature, + (void *)SPA_FEATURE_DYNAMIC_GANG_HEADER, tx); + dmu_tx_commit(tx); + } + + /* * Set pio's pipeline to just wait for zio to finish. */ pio->io_pipeline = ZIO_INTERLOCK_PIPELINE; @@ -3303,11 +3362,11 @@ zio_nop_write(zio_t *zio) zio_prop_t *zp = &zio->io_prop; ASSERT(BP_IS_HOLE(bp)); - ASSERT(BP_GET_LEVEL(bp) == 0); + ASSERT0(BP_GET_LEVEL(bp)); ASSERT(!(zio->io_flags & ZIO_FLAG_IO_REWRITE)); ASSERT(zp->zp_nopwrite); ASSERT(!zp->zp_dedup); - ASSERT(zio->io_bp_override == NULL); + ASSERT0P(zio->io_bp_override); ASSERT(IO_IS_ALLOCATING(zio)); /* @@ -3436,7 +3495,7 @@ zio_ddt_read_start(zio_t *zio) ddt_univ_phys_t *ddp = dde->dde_phys; blkptr_t blk; - ASSERT(zio->io_vsd == NULL); + ASSERT0P(zio->io_vsd); zio->io_vsd = dde; if (v_self == DDT_PHYS_NONE) @@ -3501,7 +3560,7 @@ zio_ddt_read_done(zio_t *zio) zio->io_vsd = NULL; } - ASSERT(zio->io_vsd == NULL); + ASSERT0P(zio->io_vsd); return (zio); } @@ -3836,7 +3895,7 @@ zio_ddt_write(zio_t *zio) * block and leave. */ if (have_dvas == 0) { - ASSERT(BP_GET_LOGICAL_BIRTH(bp) == txg); + ASSERT(BP_GET_BIRTH(bp) == txg); ASSERT(BP_EQUAL(bp, zio->io_bp_override)); ddt_phys_extend(ddp, v, bp); ddt_phys_addref(ddp, v); @@ -3864,6 +3923,23 @@ zio_ddt_write(zio_t *zio) * then we can just use them as-is. */ if (have_dvas >= need_dvas) { + /* + * For rewrite operations, try preserving the original + * logical birth time. If the result matches the + * original BP, this becomes a NOP. + */ + if (zp->zp_rewrite) { + uint64_t orig_logical_birth = + BP_GET_LOGICAL_BIRTH(&zio->io_bp_orig); + ddt_bp_fill(ddp, v, bp, orig_logical_birth); + if (BP_EQUAL(bp, &zio->io_bp_orig)) { + /* We can skip accounting. */ + zio->io_flags |= ZIO_FLAG_NOPWRITE; + ddt_exit(ddt); + return (zio); + } + } + ddt_bp_fill(ddp, v, bp, txg); ddt_phys_addref(ddp, v); ddt_exit(ddt); @@ -4078,9 +4154,11 @@ zio_io_to_allocate(metaslab_class_allocator_t *mca, boolean_t *more) * reserve then we throttle. */ if (!metaslab_class_throttle_reserve(zio->io_metaslab_class, - zio->io_prop.zp_copies, zio, B_FALSE, more)) { + zio->io_allocator, zio->io_prop.zp_copies, zio->io_size, + B_FALSE, more)) { return (NULL); } + zio->io_flags |= ZIO_FLAG_ALLOC_THROTTLED; avl_remove(&mca->mca_tree, zio); ASSERT3U(zio->io_stage, <, ZIO_STAGE_DVA_ALLOCATE); @@ -4164,8 +4242,10 @@ zio_dva_allocate(zio_t *zio) ASSERT3U(zio->io_child_type, ==, ZIO_CHILD_GANG); memcpy(zio->io_bp->blk_dva, zio->io_bp_orig.blk_dva, 3 * sizeof (dva_t)); - BP_SET_BIRTH(zio->io_bp, BP_GET_LOGICAL_BIRTH(&zio->io_bp_orig), - BP_GET_PHYSICAL_BIRTH(&zio->io_bp_orig)); + BP_SET_LOGICAL_BIRTH(zio->io_bp, + BP_GET_LOGICAL_BIRTH(&zio->io_bp_orig)); + BP_SET_PHYSICAL_BIRTH(zio->io_bp, + BP_GET_RAW_PHYSICAL_BIRTH(&zio->io_bp_orig)); return (zio); } @@ -4236,13 +4316,14 @@ again: * If we are holding old class reservation, drop it. * Dispatch the next ZIO(s) there if some are waiting. */ - if (zio->io_flags & ZIO_FLAG_IO_ALLOCATING) { + if (zio->io_flags & ZIO_FLAG_ALLOC_THROTTLED) { if (metaslab_class_throttle_unreserve(mc, - zio->io_prop.zp_copies, zio)) { + zio->io_allocator, zio->io_prop.zp_copies, + zio->io_size)) { zio_allocate_dispatch(zio->io_metaslab_class, zio->io_allocator); } - zio->io_flags &= ~ZIO_FLAG_IO_ALLOCATING; + zio->io_flags &= ~ZIO_FLAG_ALLOC_THROTTLED; } if (zfs_flags & ZFS_DEBUG_METASLAB_ALLOC) { @@ -4291,6 +4372,15 @@ again: error); } zio->io_error = error; + } else if (zio->io_prop.zp_rewrite) { + /* + * For rewrite operations, preserve the logical birth time + * but set the physical birth time to the current txg. + */ + uint64_t logical_birth = BP_GET_LOGICAL_BIRTH(&zio->io_bp_orig); + ASSERT3U(logical_birth, <=, zio->io_txg); + BP_SET_BIRTH(zio->io_bp, logical_birth, zio->io_txg); + BP_SET_REWRITE(zio->io_bp, 1); } return (zio); @@ -4324,18 +4414,17 @@ zio_dva_claim(zio_t *zio) static void zio_dva_unallocate(zio_t *zio, zio_gang_node_t *gn, blkptr_t *bp) { - ASSERT(BP_GET_LOGICAL_BIRTH(bp) == zio->io_txg || BP_IS_HOLE(bp)); - ASSERT(zio->io_bp_override == NULL); + ASSERT(BP_GET_BIRTH(bp) == zio->io_txg || BP_IS_HOLE(bp)); + ASSERT0P(zio->io_bp_override); if (!BP_IS_HOLE(bp)) { - metaslab_free(zio->io_spa, bp, BP_GET_LOGICAL_BIRTH(bp), - B_TRUE); + metaslab_free(zio->io_spa, bp, BP_GET_BIRTH(bp), B_TRUE); } if (gn != NULL) { - for (int g = 0; g < SPA_GBH_NBLKPTRS; g++) { + for (int g = 0; g < gbh_nblkptrs(gn->gn_gangblocksize); g++) { zio_dva_unallocate(zio, gn->gn_child[g], - &gn->gn_gbh->zg_blkptr[g]); + gbh_bp(gn->gn_gbh, g)); } } } @@ -4347,7 +4436,7 @@ int zio_alloc_zil(spa_t *spa, objset_t *os, uint64_t txg, blkptr_t *new_bp, uint64_t size, boolean_t *slog) { - int error = 1; + int error; zio_alloc_list_t io_alloc_list; ASSERT(txg > spa_syncing_txg(spa)); @@ -4372,14 +4461,34 @@ zio_alloc_zil(spa_t *spa, objset_t *os, uint64_t txg, blkptr_t *new_bp, int allocator = (uint_t)cityhash1(os->os_dsl_dataset->ds_object) % spa->spa_alloc_count; ZIOSTAT_BUMP(ziostat_total_allocations); + + /* Try log class (dedicated slog devices) first */ error = metaslab_alloc(spa, spa_log_class(spa), size, new_bp, 1, txg, NULL, flags, &io_alloc_list, allocator, NULL); *slog = (error == 0); + + /* Try special_embedded_log class (reserved on special vdevs) */ + if (error != 0) { + error = metaslab_alloc(spa, spa_special_embedded_log_class(spa), + size, new_bp, 1, txg, NULL, flags, &io_alloc_list, + allocator, NULL); + } + + /* Try special class (general special vdev allocation) */ + if (error != 0) { + error = metaslab_alloc(spa, spa_special_class(spa), size, + new_bp, 1, txg, NULL, flags, &io_alloc_list, allocator, + NULL); + } + + /* Try embedded_log class (reserved on normal vdevs) */ if (error != 0) { error = metaslab_alloc(spa, spa_embedded_log_class(spa), size, new_bp, 1, txg, NULL, flags, &io_alloc_list, allocator, NULL); } + + /* Finally fall back to normal class */ if (error != 0) { ZIOSTAT_BUMP(ziostat_alloc_class_fallbacks); error = metaslab_alloc(spa, spa_normal_class(spa), size, @@ -4450,8 +4559,8 @@ zio_vdev_io_start(zio_t *zio) zio->io_delay = 0; - ASSERT(zio->io_error == 0); - ASSERT(zio->io_child_error[ZIO_CHILD_VDEV] == 0); + ASSERT0(zio->io_error); + ASSERT0(zio->io_child_error[ZIO_CHILD_VDEV]); if (vd == NULL) { if (!(zio->io_flags & ZIO_FLAG_CONFIG_WRITER)) @@ -4642,7 +4751,7 @@ zio_vdev_io_done(zio_t *zio) ops->vdev_op_io_done(zio); if (unexpected_error && vd->vdev_remove_wanted == B_FALSE) - VERIFY(vdev_probe(vd, zio) == NULL); + VERIFY0P(vdev_probe(vd, zio)); return (zio); } @@ -4722,7 +4831,7 @@ zio_vdev_io_assess(zio_t *zio) * If a Direct I/O operation has a checksum verify error then this I/O * should not attempt to be issued again. */ - if (zio->io_flags & ZIO_FLAG_DIO_CHKSUM_ERR) { + if (zio->io_post & ZIO_POST_DIO_CHKSUM_ERR) { if (zio->io_type == ZIO_TYPE_WRITE) { ASSERT3U(zio->io_child_type, ==, ZIO_CHILD_LOGICAL); ASSERT3U(zio->io_error, ==, EIO); @@ -4794,7 +4903,7 @@ void zio_vdev_io_reissue(zio_t *zio) { ASSERT(zio->io_stage == ZIO_STAGE_VDEV_IO_START); - ASSERT(zio->io_error == 0); + ASSERT0(zio->io_error); zio->io_stage >>= 1; } @@ -4811,7 +4920,7 @@ void zio_vdev_io_bypass(zio_t *zio) { ASSERT(zio->io_stage == ZIO_STAGE_VDEV_IO_START); - ASSERT(zio->io_error == 0); + ASSERT0(zio->io_error); zio->io_flags |= ZIO_FLAG_IO_BYPASS; zio->io_stage = ZIO_STAGE_VDEV_IO_ASSESS >> 1; @@ -5031,7 +5140,7 @@ zio_checksum_verify(zio_t *zio) ASSERT3U(zio->io_prop.zp_checksum, ==, ZIO_CHECKSUM_LABEL); } - ASSERT0(zio->io_flags & ZIO_FLAG_DIO_CHKSUM_ERR); + ASSERT0(zio->io_post & ZIO_POST_DIO_CHKSUM_ERR); IMPLY(zio->io_flags & ZIO_FLAG_DIO_READ, !(zio->io_flags & ZIO_FLAG_SPECULATIVE)); @@ -5040,7 +5149,7 @@ zio_checksum_verify(zio_t *zio) if (error == ECKSUM && !(zio->io_flags & ZIO_FLAG_SPECULATIVE)) { if (zio->io_flags & ZIO_FLAG_DIO_READ) { - zio->io_flags |= ZIO_FLAG_DIO_CHKSUM_ERR; + zio->io_post |= ZIO_POST_DIO_CHKSUM_ERR; zio_t *pio = zio_unique_parent(zio); /* * Any Direct I/O read that has a checksum @@ -5090,7 +5199,7 @@ zio_dio_checksum_verify(zio_t *zio) if ((error = zio_checksum_error(zio, NULL)) != 0) { zio->io_error = error; if (error == ECKSUM) { - zio->io_flags |= ZIO_FLAG_DIO_CHKSUM_ERR; + zio->io_post |= ZIO_POST_DIO_CHKSUM_ERR; zio_dio_chksum_verify_error_report(zio); } } @@ -5115,7 +5224,7 @@ zio_checksum_verified(zio_t *zio) void zio_dio_chksum_verify_error_report(zio_t *zio) { - ASSERT(zio->io_flags & ZIO_FLAG_DIO_CHKSUM_ERR); + ASSERT(zio->io_post & ZIO_POST_DIO_CHKSUM_ERR); if (zio->io_child_type == ZIO_CHILD_LOGICAL) return; @@ -5187,9 +5296,9 @@ zio_ready(zio_t *zio) if (zio->io_ready) { ASSERT(IO_IS_ALLOCATING(zio)); - ASSERT(BP_GET_LOGICAL_BIRTH(bp) == zio->io_txg || + ASSERT(BP_GET_BIRTH(bp) == zio->io_txg || BP_IS_HOLE(bp) || (zio->io_flags & ZIO_FLAG_NOPWRITE)); - ASSERT(zio->io_children[ZIO_CHILD_GANG][ZIO_WAIT_READY] == 0); + ASSERT0(zio->io_children[ZIO_CHILD_GANG][ZIO_WAIT_READY]); zio->io_ready(zio); } @@ -5202,7 +5311,7 @@ zio_ready(zio_t *zio) if (zio->io_error != 0) { zio->io_pipeline = ZIO_INTERLOCK_PIPELINE; - if (zio->io_flags & ZIO_FLAG_IO_ALLOCATING) { + if (zio->io_flags & ZIO_FLAG_ALLOC_THROTTLED) { ASSERT(IO_IS_ALLOCATING(zio)); ASSERT(zio->io_priority == ZIO_PRIORITY_ASYNC_WRITE); ASSERT(zio->io_metaslab_class != NULL); @@ -5213,8 +5322,8 @@ zio_ready(zio_t *zio) * issue the next I/O to allocate. */ if (metaslab_class_throttle_unreserve( - zio->io_metaslab_class, zio->io_prop.zp_copies, - zio)) { + zio->io_metaslab_class, zio->io_allocator, + zio->io_prop.zp_copies, zio->io_size)) { zio_allocate_dispatch(zio->io_metaslab_class, zio->io_allocator); } @@ -5264,6 +5373,7 @@ zio_dva_throttle_done(zio_t *zio) vdev_t *vd = zio->io_vd; int flags = METASLAB_ASYNC_ALLOC; const void *tag = pio; + uint64_t size = pio->io_size; ASSERT3P(zio->io_bp, !=, NULL); ASSERT3U(zio->io_type, ==, ZIO_TYPE_WRITE); @@ -5273,16 +5383,19 @@ zio_dva_throttle_done(zio_t *zio) ASSERT3P(vd, ==, vd->vdev_top); ASSERT(zio_injection_enabled || !(zio->io_flags & ZIO_FLAG_IO_RETRY)); ASSERT(!(zio->io_flags & ZIO_FLAG_IO_REPAIR)); - ASSERT(zio->io_flags & ZIO_FLAG_IO_ALLOCATING); + ASSERT(zio->io_flags & ZIO_FLAG_ALLOC_THROTTLED); /* * Parents of gang children can have two flavors -- ones that allocated * the gang header (will have ZIO_FLAG_IO_REWRITE set) and ones that * allocated the constituent blocks. The first use their parent as tag. + * We set the size to match the original allocation call for that case. */ if (pio->io_child_type == ZIO_CHILD_GANG && - (pio->io_flags & ZIO_FLAG_IO_REWRITE)) + (pio->io_flags & ZIO_FLAG_IO_REWRITE)) { tag = zio_unique_parent(pio); + size = SPA_OLD_GANGBLOCKSIZE; + } ASSERT(IO_IS_ALLOCATING(pio) || (pio->io_child_type == ZIO_CHILD_GANG && (pio->io_flags & ZIO_FLAG_IO_REWRITE))); @@ -5295,9 +5408,10 @@ zio_dva_throttle_done(zio_t *zio) ASSERT(zio->io_metaslab_class->mc_alloc_throttle_enabled); metaslab_group_alloc_decrement(zio->io_spa, vd->vdev_id, - pio->io_allocator, flags, pio->io_size, tag); + pio->io_allocator, flags, size, tag); - if (metaslab_class_throttle_unreserve(zio->io_metaslab_class, 1, pio)) { + if (metaslab_class_throttle_unreserve(pio->io_metaslab_class, + pio->io_allocator, 1, pio->io_size)) { zio_allocate_dispatch(zio->io_metaslab_class, pio->io_allocator); } @@ -5328,17 +5442,15 @@ zio_done(zio_t *zio) * write. We must do this since the allocation is performed * by the logical I/O but the actual write is done by child I/Os. */ - if (zio->io_flags & ZIO_FLAG_IO_ALLOCATING && + if (zio->io_flags & ZIO_FLAG_ALLOC_THROTTLED && zio->io_child_type == ZIO_CHILD_VDEV) zio_dva_throttle_done(zio); for (int c = 0; c < ZIO_CHILD_TYPES; c++) for (int w = 0; w < ZIO_WAIT_TYPES; w++) - ASSERT(zio->io_children[c][w] == 0); + ASSERT0(zio->io_children[c][w]); if (zio->io_bp != NULL && !BP_IS_EMBEDDED(zio->io_bp)) { - ASSERT(zio->io_bp->blk_pad[0] == 0); - ASSERT(zio->io_bp->blk_pad[1] == 0); ASSERT(memcmp(zio->io_bp, &zio->io_bp_copy, sizeof (blkptr_t)) == 0 || (zio->io_bp == zio_unique_parent(zio)->io_bp)); @@ -5431,7 +5543,7 @@ zio_done(zio_t *zio) */ if (zio->io_error != ECKSUM && zio->io_vd != NULL && !vdev_is_dead(zio->io_vd) && - !(zio->io_flags & ZIO_FLAG_DIO_CHKSUM_ERR)) { + !(zio->io_post & ZIO_POST_DIO_CHKSUM_ERR)) { int ret = zfs_ereport_post(FM_EREPORT_ZFS_IO, zio->io_spa, zio->io_vd, &zio->io_bookmark, zio, 0); if (ret != EALREADY) { @@ -5446,14 +5558,14 @@ zio_done(zio_t *zio) if ((zio->io_error == EIO || !(zio->io_flags & (ZIO_FLAG_SPECULATIVE | ZIO_FLAG_DONT_PROPAGATE))) && - !(zio->io_flags & ZIO_FLAG_DIO_CHKSUM_ERR) && + !(zio->io_post & ZIO_POST_DIO_CHKSUM_ERR) && zio == zio->io_logical) { /* * For logical I/O requests, tell the SPA to log the * error and generate a logical data ereport. */ spa_log_error(zio->io_spa, &zio->io_bookmark, - BP_GET_LOGICAL_BIRTH(zio->io_bp)); + BP_GET_PHYSICAL_BIRTH(zio->io_bp)); (void) zfs_ereport_post(FM_EREPORT_ZFS_DATA, zio->io_spa, NULL, &zio->io_bookmark, zio, 0); } @@ -5467,7 +5579,7 @@ zio_done(zio_t *zio) */ if (zio->io_error == EAGAIN && IO_IS_ALLOCATING(zio) && zio->io_prop.zp_dedup) { - zio->io_reexecute |= ZIO_REEXECUTE_NOW; + zio->io_post |= ZIO_POST_REEXECUTE; zio->io_prop.zp_dedup = B_FALSE; } /* @@ -5479,11 +5591,11 @@ zio_done(zio_t *zio) if (IO_IS_ALLOCATING(zio) && !(zio->io_flags & ZIO_FLAG_CANFAIL) && - !(zio->io_flags & ZIO_FLAG_DIO_CHKSUM_ERR)) { + !(zio->io_post & ZIO_POST_DIO_CHKSUM_ERR)) { if (zio->io_error != ENOSPC) - zio->io_reexecute |= ZIO_REEXECUTE_NOW; + zio->io_post |= ZIO_POST_REEXECUTE; else - zio->io_reexecute |= ZIO_REEXECUTE_SUSPEND; + zio->io_post |= ZIO_POST_SUSPEND; } if ((zio->io_type == ZIO_TYPE_READ || @@ -5492,10 +5604,11 @@ zio_done(zio_t *zio) zio->io_error == ENXIO && spa_load_state(zio->io_spa) == SPA_LOAD_NONE && spa_get_failmode(zio->io_spa) != ZIO_FAILURE_MODE_CONTINUE) - zio->io_reexecute |= ZIO_REEXECUTE_SUSPEND; + zio->io_post |= ZIO_POST_SUSPEND; - if (!(zio->io_flags & ZIO_FLAG_CANFAIL) && !zio->io_reexecute) - zio->io_reexecute |= ZIO_REEXECUTE_SUSPEND; + if (!(zio->io_flags & ZIO_FLAG_CANFAIL) && + !(zio->io_post & (ZIO_POST_REEXECUTE|ZIO_POST_SUSPEND))) + zio->io_post |= ZIO_POST_SUSPEND; /* * Here is a possibly good place to attempt to do @@ -5514,7 +5627,8 @@ zio_done(zio_t *zio) */ zio_inherit_child_errors(zio, ZIO_CHILD_LOGICAL); - if ((zio->io_error || zio->io_reexecute) && + if ((zio->io_error || + (zio->io_post & (ZIO_POST_REEXECUTE|ZIO_POST_SUSPEND))) && IO_IS_ALLOCATING(zio) && zio->io_gang_leader == zio && !(zio->io_flags & (ZIO_FLAG_IO_REWRITE | ZIO_FLAG_NOPWRITE))) zio_dva_unallocate(zio, zio->io_gang_tree, zio->io_bp); @@ -5525,16 +5639,16 @@ zio_done(zio_t *zio) * Godfather I/Os should never suspend. */ if ((zio->io_flags & ZIO_FLAG_GODFATHER) && - (zio->io_reexecute & ZIO_REEXECUTE_SUSPEND)) - zio->io_reexecute &= ~ZIO_REEXECUTE_SUSPEND; + (zio->io_post & ZIO_POST_SUSPEND)) + zio->io_post &= ~ZIO_POST_SUSPEND; - if (zio->io_reexecute) { + if (zio->io_post & (ZIO_POST_REEXECUTE|ZIO_POST_SUSPEND)) { /* * A Direct I/O operation that has a checksum verify error * should not attempt to reexecute. Instead, the error should * just be propagated back. */ - ASSERT(!(zio->io_flags & ZIO_FLAG_DIO_CHKSUM_ERR)); + ASSERT0(zio->io_post & ZIO_POST_DIO_CHKSUM_ERR); /* * This is a logical I/O that wants to reexecute. @@ -5571,7 +5685,7 @@ zio_done(zio_t *zio) pio_next = zio_walk_parents(zio, &zl); if ((pio->io_flags & ZIO_FLAG_GODFATHER) && - (zio->io_reexecute & ZIO_REEXECUTE_SUSPEND)) { + (zio->io_post & ZIO_POST_SUSPEND)) { zio_remove_child(pio, zio, remove_zl); /* * This is a rare code path, so we don't @@ -5595,13 +5709,14 @@ zio_done(zio_t *zio) * "next_to_execute". */ zio_notify_parent(pio, zio, ZIO_WAIT_DONE, NULL); - } else if (zio->io_reexecute & ZIO_REEXECUTE_SUSPEND) { + } else if (zio->io_post & ZIO_POST_SUSPEND) { /* * We'd fail again if we reexecuted now, so suspend * until conditions improve (e.g. device comes online). */ zio_suspend(zio->io_spa, zio, ZIO_SUSPEND_IOERR); } else { + ASSERT(zio->io_post & ZIO_POST_REEXECUTE); /* * Reexecution is potentially a huge amount of work. * Hand it off to the otherwise-unused claim taskq. @@ -5614,7 +5729,8 @@ zio_done(zio_t *zio) } ASSERT(list_is_empty(&zio->io_child_list)); - ASSERT(zio->io_reexecute == 0); + ASSERT0(zio->io_post & ZIO_POST_REEXECUTE); + ASSERT0(zio->io_post & ZIO_POST_SUSPEND); ASSERT(zio->io_error == 0 || (zio->io_flags & ZIO_FLAG_CANFAIL)); /* diff --git a/sys/contrib/openzfs/module/zfs/zio_checksum.c b/sys/contrib/openzfs/module/zfs/zio_checksum.c index a91775b04af2..1d0646a61185 100644 --- a/sys/contrib/openzfs/module/zfs/zio_checksum.c +++ b/sys/contrib/openzfs/module/zfs/zio_checksum.c @@ -215,7 +215,7 @@ zio_checksum_info_t zio_checksum_table[ZIO_CHECKSUM_FUNCTIONS] = { spa_feature_t zio_checksum_to_feature(enum zio_checksum cksum) { - VERIFY((cksum & ~ZIO_CHECKSUM_MASK) == 0); + VERIFY0((cksum & ~ZIO_CHECKSUM_MASK)); switch (cksum) { case ZIO_CHECKSUM_BLAKE3: @@ -279,7 +279,7 @@ static void zio_checksum_gang_verifier(zio_cksum_t *zcp, const blkptr_t *bp) { const dva_t *dva = BP_IDENTITY(bp); - uint64_t txg = BP_GET_BIRTH(bp); + uint64_t txg = BP_GET_PHYSICAL_BIRTH(bp); ASSERT(BP_IS_GANG(bp)); @@ -545,14 +545,39 @@ zio_checksum_error(zio_t *zio, zio_bad_cksum_t *info) uint_t checksum = (bp == NULL ? zio->io_prop.zp_checksum : (BP_IS_GANG(bp) ? ZIO_CHECKSUM_GANG_HEADER : BP_GET_CHECKSUM(bp))); int error; - uint64_t size = (bp == NULL ? zio->io_size : - (BP_IS_GANG(bp) ? SPA_GANGBLOCKSIZE : BP_GET_PSIZE(bp))); + uint64_t size = bp ? BP_GET_PSIZE(bp) : zio->io_size; uint64_t offset = zio->io_offset; abd_t *data = zio->io_abd; spa_t *spa = zio->io_spa; + if (bp && BP_IS_GANG(bp)) { + if (spa_feature_is_active(spa, SPA_FEATURE_DYNAMIC_GANG_HEADER)) + size = zio->io_size; + else + size = SPA_OLD_GANGBLOCKSIZE; + } + error = zio_checksum_error_impl(spa, bp, checksum, data, size, offset, info); + if (error && bp && BP_IS_GANG(bp) && size > SPA_OLD_GANGBLOCKSIZE) { + /* + * It's possible that this is an old gang block. Rerun + * the checksum with the old size; if that passes, then + * update the gangblocksize appropriately. + */ + error = zio_checksum_error_impl(spa, bp, checksum, data, + SPA_OLD_GANGBLOCKSIZE, offset, info); + if (error == 0) { + ASSERT3U(zio->io_child_type, ==, ZIO_CHILD_VDEV); + zio_t *pio; + for (pio = zio_unique_parent(zio); + pio->io_child_type != ZIO_CHILD_GANG; + pio = zio_unique_parent(pio)) + ; + zio_gang_node_t *gn = pio->io_private; + gn->gn_gangblocksize = SPA_OLD_GANGBLOCKSIZE; + } + } if (zio_injection_enabled && error == 0 && zio->io_error == 0) { error = zio_handle_fault_injection(zio, ECKSUM); diff --git a/sys/contrib/openzfs/module/zfs/zio_compress.c b/sys/contrib/openzfs/module/zfs/zio_compress.c index 9f0ac1b63146..89ceeb58ad91 100644 --- a/sys/contrib/openzfs/module/zfs/zio_compress.c +++ b/sys/contrib/openzfs/module/zfs/zio_compress.c @@ -38,12 +38,6 @@ #include <sys/zstd/zstd.h> /* - * If nonzero, every 1/X decompression attempts will fail, simulating - * an undetected memory error. - */ -static unsigned long zio_decompress_fail_fraction = 0; - -/* * Compression vectors. */ zio_compress_info_t zio_compress_table[ZIO_COMPRESS_FUNCTIONS] = { @@ -171,15 +165,6 @@ zio_decompress_data(enum zio_compress c, abd_t *src, abd_t *dst, else err = ci->ci_decompress(src, dst, s_len, d_len, ci->ci_level); - /* - * Decompression shouldn't fail, because we've already verified - * the checksum. However, for extra protection (e.g. against bitflips - * in non-ECC RAM), we handle this error (and test it). - */ - if (zio_decompress_fail_fraction != 0 && - random_in_range(zio_decompress_fail_fraction) == 0) - err = SET_ERROR(EINVAL); - return (err); } diff --git a/sys/contrib/openzfs/module/zfs/zio_inject.c b/sys/contrib/openzfs/module/zfs/zio_inject.c index df7b01ba879e..981a1be4847c 100644 --- a/sys/contrib/openzfs/module/zfs/zio_inject.c +++ b/sys/contrib/openzfs/module/zfs/zio_inject.c @@ -1119,7 +1119,7 @@ zio_clear_fault(int id) kmem_free(handler->zi_lanes, sizeof (*handler->zi_lanes) * handler->zi_record.zi_nlanes); } else { - ASSERT3P(handler->zi_lanes, ==, NULL); + ASSERT0P(handler->zi_lanes); } if (handler->zi_spa_name != NULL) diff --git a/sys/contrib/openzfs/module/zfs/zrlock.c b/sys/contrib/openzfs/module/zfs/zrlock.c index 3c0f1b7bbbc1..09c110945c97 100644 --- a/sys/contrib/openzfs/module/zfs/zrlock.c +++ b/sys/contrib/openzfs/module/zfs/zrlock.c @@ -129,7 +129,7 @@ zrl_tryenter(zrlock_t *zrl) (uint32_t *)&zrl->zr_refcount, 0, ZRL_LOCKED); if (cas == 0) { #ifdef ZFS_DEBUG - ASSERT3P(zrl->zr_owner, ==, NULL); + ASSERT0P(zrl->zr_owner); zrl->zr_owner = curthread; #endif return (1); diff --git a/sys/contrib/openzfs/module/zfs/zthr.c b/sys/contrib/openzfs/module/zfs/zthr.c index 597a510528ea..d245ce4946e0 100644 --- a/sys/contrib/openzfs/module/zfs/zthr.c +++ b/sys/contrib/openzfs/module/zfs/zthr.c @@ -316,7 +316,7 @@ zthr_destroy(zthr_t *t) { ASSERT(!MUTEX_HELD(&t->zthr_state_lock)); ASSERT(!MUTEX_HELD(&t->zthr_request_lock)); - VERIFY3P(t->zthr_thread, ==, NULL); + VERIFY0P(t->zthr_thread); mutex_destroy(&t->zthr_request_lock); mutex_destroy(&t->zthr_state_lock); cv_destroy(&t->zthr_cv); diff --git a/sys/contrib/openzfs/module/zfs/zvol.c b/sys/contrib/openzfs/module/zfs/zvol.c index 3568d4f43fcb..29f51e230a37 100644 --- a/sys/contrib/openzfs/module/zfs/zvol.c +++ b/sys/contrib/openzfs/module/zfs/zvol.c @@ -102,6 +102,7 @@ extern int zfs_bclone_wait_dirty; zv_taskq_t zvol_taskqs; typedef enum { + ZVOL_ASYNC_CREATE_MINORS, ZVOL_ASYNC_REMOVE_MINORS, ZVOL_ASYNC_RENAME_MINORS, ZVOL_ASYNC_SET_SNAPDEV, @@ -110,10 +111,14 @@ typedef enum { } zvol_async_op_t; typedef struct { - zvol_async_op_t op; - char name1[MAXNAMELEN]; - char name2[MAXNAMELEN]; - uint64_t value; + zvol_async_op_t zt_op; + char zt_name1[MAXNAMELEN]; + char zt_name2[MAXNAMELEN]; + uint64_t zt_value; + uint32_t zt_total; + uint32_t zt_done; + int32_t zt_status; + int zt_error; } zvol_task_t; zv_request_task_t * @@ -210,8 +215,8 @@ zvol_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx) int error; uint64_t volblocksize, volsize; - VERIFY(nvlist_lookup_uint64(nvprops, - zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) == 0); + VERIFY0(nvlist_lookup_uint64(nvprops, + zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize)); if (nvlist_lookup_uint64(nvprops, zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE), &volblocksize) != 0) volblocksize = zfs_prop_default_numeric(ZFS_PROP_VOLBLOCKSIZE); @@ -220,21 +225,20 @@ zvol_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx) * These properties must be removed from the list so the generic * property setting step won't apply to them. */ - VERIFY(nvlist_remove_all(nvprops, - zfs_prop_to_name(ZFS_PROP_VOLSIZE)) == 0); + VERIFY0(nvlist_remove_all(nvprops, zfs_prop_to_name(ZFS_PROP_VOLSIZE))); (void) nvlist_remove_all(nvprops, zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE)); error = dmu_object_claim(os, ZVOL_OBJ, DMU_OT_ZVOL, volblocksize, DMU_OT_NONE, 0, tx); - ASSERT(error == 0); + ASSERT0(error); error = zap_create_claim(os, ZVOL_ZAP_OBJ, DMU_OT_ZVOL_PROP, DMU_OT_NONE, 0, tx); - ASSERT(error == 0); + ASSERT0(error); error = zap_update(os, ZVOL_ZAP_OBJ, "size", 8, 1, &volsize, tx); - ASSERT(error == 0); + ASSERT0(error); } /* @@ -249,7 +253,7 @@ zvol_get_stats(objset_t *os, nvlist_t *nv) error = zap_lookup(os, ZVOL_ZAP_OBJ, "size", 8, 1, &val); if (error) - return (SET_ERROR(error)); + return (error); dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_VOLSIZE, val); doi = kmem_alloc(sizeof (dmu_object_info_t), KM_SLEEP); @@ -262,7 +266,7 @@ zvol_get_stats(objset_t *os, nvlist_t *nv) kmem_free(doi, sizeof (dmu_object_info_t)); - return (SET_ERROR(error)); + return (error); } /* @@ -300,7 +304,7 @@ zvol_update_volsize(uint64_t volsize, objset_t *os) error = dmu_tx_assign(tx, DMU_TX_WAIT); if (error) { dmu_tx_abort(tx); - return (SET_ERROR(error)); + return (error); } txg = dmu_tx_get_txg(tx); @@ -332,7 +336,7 @@ zvol_set_volsize(const char *name, uint64_t volsize) error = dsl_prop_get_integer(name, zfs_prop_to_name(ZFS_PROP_READONLY), &readonly, NULL); if (error != 0) - return (SET_ERROR(error)); + return (error); if (readonly) return (SET_ERROR(EROFS)); @@ -348,7 +352,7 @@ zvol_set_volsize(const char *name, uint64_t volsize) FTAG, &os)) != 0) { if (zv != NULL) mutex_exit(&zv->zv_state_lock); - return (SET_ERROR(error)); + return (error); } owned = B_TRUE; if (zv != NULL) @@ -385,7 +389,7 @@ out: if (error == 0 && zv != NULL) zvol_os_update_volsize(zv, volsize); - return (SET_ERROR(error)); + return (error); } /* @@ -396,7 +400,7 @@ zvol_set_volthreading(const char *name, boolean_t value) { zvol_state_t *zv = zvol_find_by_name(name, RW_NONE); if (zv == NULL) - return (ENOENT); + return (SET_ERROR(ENOENT)); zv->zv_threading = value; mutex_exit(&zv->zv_state_lock); return (0); @@ -445,8 +449,10 @@ zvol_check_volblocksize(const char *name, uint64_t volblocksize) * We don't allow setting the property above 1MB, * unless the tunable has been changed. */ - if (volblocksize > zfs_max_recordsize) + if (volblocksize > zfs_max_recordsize) { + spa_close(spa, FTAG); return (SET_ERROR(EDOM)); + } spa_close(spa, FTAG); } @@ -613,7 +619,7 @@ zvol_clone_range(zvol_state_t *zv_src, uint64_t inoff, zvol_state_t *zv_dst, dmu_tx_t *tx; blkptr_t *bps; size_t maxblocks; - int error = EINVAL; + int error = 0; rw_enter(&zv_dst->zv_suspend_lock, RW_READER); if (zv_dst->zv_zilog == NULL) { @@ -639,23 +645,22 @@ zvol_clone_range(zvol_state_t *zv_src, uint64_t inoff, zvol_state_t *zv_dst, */ if (!spa_feature_is_enabled(dmu_objset_spa(outos), SPA_FEATURE_BLOCK_CLONING)) { - error = EOPNOTSUPP; + error = SET_ERROR(EOPNOTSUPP); goto out; } if (dmu_objset_spa(inos) != dmu_objset_spa(outos)) { - error = EXDEV; + error = SET_ERROR(EXDEV); goto out; } if (inos->os_encrypted != outos->os_encrypted) { - error = EXDEV; + error = SET_ERROR(EXDEV); goto out; } if (zv_src->zv_volblocksize != zv_dst->zv_volblocksize) { - error = EINVAL; + error = SET_ERROR(EINVAL); goto out; } if (inoff >= zv_src->zv_volsize || outoff >= zv_dst->zv_volsize) { - error = 0; goto out; } @@ -666,17 +671,15 @@ zvol_clone_range(zvol_state_t *zv_src, uint64_t inoff, zvol_state_t *zv_dst, len = zv_src->zv_volsize - inoff; if (len > zv_dst->zv_volsize - outoff) len = zv_dst->zv_volsize - outoff; - if (len == 0) { - error = 0; + if (len == 0) goto out; - } /* * No overlapping if we are cloning within the same file */ if (zv_src == zv_dst) { if (inoff < outoff + len && outoff < inoff + len) { - error = EINVAL; + error = SET_ERROR(EINVAL); goto out; } } @@ -686,7 +689,7 @@ zvol_clone_range(zvol_state_t *zv_src, uint64_t inoff, zvol_state_t *zv_dst, */ if ((inoff % zv_src->zv_volblocksize) != 0 || (outoff % zv_dst->zv_volblocksize) != 0) { - error = EINVAL; + error = SET_ERROR(EINVAL); goto out; } @@ -694,7 +697,7 @@ zvol_clone_range(zvol_state_t *zv_src, uint64_t inoff, zvol_state_t *zv_dst, * Length must be multiple of block size */ if ((len % zv_src->zv_volblocksize) != 0) { - error = EINVAL; + error = SET_ERROR(EINVAL); goto out; } @@ -766,13 +769,13 @@ zvol_clone_range(zvol_state_t *zv_src, uint64_t inoff, zvol_state_t *zv_dst, zfs_rangelock_exit(outlr); zfs_rangelock_exit(inlr); if (error == 0 && zv_dst->zv_objset->os_sync == ZFS_SYNC_ALWAYS) { - zil_commit(zilog_dst, ZVOL_OBJ); + error = zil_commit(zilog_dst, ZVOL_OBJ); } out: if (zv_src != zv_dst) rw_exit(&zv_src->zv_suspend_lock); rw_exit(&zv_dst->zv_suspend_lock); - return (SET_ERROR(error)); + return (error); } /* @@ -859,13 +862,8 @@ zil_replay_func_t *const zvol_replay_vector[TX_MAX_TYPE] = { }; /* - * zvol_log_write() handles synchronous writes using TX_WRITE ZIL transactions. - * - * We store data in the log buffers if it's small enough. - * Otherwise we will later flush the data out via dmu_sync(). + * zvol_log_write() handles TX_WRITE transactions. */ -static const ssize_t zvol_immediate_write_sz = 32768; - void zvol_log_write(zvol_state_t *zv, dmu_tx_t *tx, uint64_t offset, uint64_t size, boolean_t commit) @@ -878,15 +876,7 @@ zvol_log_write(zvol_state_t *zv, dmu_tx_t *tx, uint64_t offset, if (zil_replaying(zilog, tx)) return; - if (zilog->zl_logbias == ZFS_LOGBIAS_THROUGHPUT) - write_state = WR_INDIRECT; - else if (!spa_has_slogs(zilog->zl_spa) && - size >= blocksize && blocksize > zvol_immediate_write_sz) - write_state = WR_INDIRECT; - else if (commit) - write_state = WR_COPIED; - else - write_state = WR_NEED_COPY; + write_state = zil_write_state(zilog, size, blocksize, B_FALSE, commit); while (size) { itx_t *itx; @@ -905,7 +895,7 @@ zvol_log_write(zvol_state_t *zv, dmu_tx_t *tx, uint64_t offset, if (wr_state == WR_COPIED && dmu_read_by_dnode(zv->zv_dn, offset, len, lr + 1, DMU_READ_NO_PREFETCH | DMU_KEEP_CACHING) != 0) { - zil_itx_destroy(itx); + zil_itx_destroy(itx, 0); itx = zil_itx_create(TX_WRITE, sizeof (*lr)); lr = (lr_write_t *)&itx->itx_lr; wr_state = WR_NEED_COPY; @@ -924,7 +914,7 @@ zvol_log_write(zvol_state_t *zv, dmu_tx_t *tx, uint64_t offset, itx->itx_private = zv; - (void) zil_itx_assign(zilog, itx, tx); + zil_itx_assign(zilog, itx, tx); offset += len; size -= len; @@ -1034,7 +1024,7 @@ zvol_get_data(void *arg, uint64_t arg2, lr_write_t *lr, char *buf, zvol_get_done(zgd, error); - return (SET_ERROR(error)); + return (error); } /* @@ -1079,15 +1069,15 @@ zvol_setup_zv(zvol_state_t *zv) error = dsl_prop_get_integer(zv->zv_name, "readonly", &ro, NULL); if (error) - return (SET_ERROR(error)); + return (error); error = zap_lookup(os, ZVOL_ZAP_OBJ, "size", 8, 1, &volsize); if (error) - return (SET_ERROR(error)); + return (error); error = dnode_hold(os, ZVOL_OBJ, zv, &zv->zv_dn); if (error) - return (SET_ERROR(error)); + return (error); zvol_os_set_capacity(zv, volsize >> 9); zv->zv_volsize = volsize; @@ -1129,7 +1119,7 @@ zvol_shutdown_zv(zvol_state_t *zv) */ if (zv->zv_flags & ZVOL_WRITTEN_TO) txg_wait_synced(dmu_objset_pool(zv->zv_objset), 0); - (void) dmu_objset_evict_dbufs(zv->zv_objset); + dmu_objset_evict_dbufs(zv->zv_objset); } /* @@ -1206,7 +1196,7 @@ zvol_resume(zvol_state_t *zv) if (zv->zv_flags & ZVOL_REMOVING) cv_broadcast(&zv->zv_removing_cv); - return (SET_ERROR(error)); + return (error); } int @@ -1222,7 +1212,7 @@ zvol_first_open(zvol_state_t *zv, boolean_t readonly) boolean_t ro = (readonly || (strchr(zv->zv_name, '@') != NULL)); error = dmu_objset_own(zv->zv_name, DMU_OST_ZVOL, ro, B_TRUE, zv, &os); if (error) - return (SET_ERROR(error)); + return (error); zv->zv_objset = os; @@ -1434,6 +1424,48 @@ zvol_create_minors_cb(const char *dsname, void *arg) return (0); } +static void +zvol_task_update_status(zvol_task_t *task, uint64_t total, uint64_t done, + int error) +{ + + task->zt_total += total; + task->zt_done += done; + if (task->zt_total != task->zt_done) { + task->zt_status = -1; + if (error) + task->zt_error = error; + } +} + +static void +zvol_task_report_status(zvol_task_t *task) +{ +#ifdef ZFS_DEBUG + static const char *const msg[] = { + "create", + "remove", + "rename", + "set snapdev", + "set volmode", + "unknown", + }; + + if (task->zt_status == 0) + return; + + zvol_async_op_t op = MIN(task->zt_op, ZVOL_ASYNC_MAX); + if (task->zt_error) { + dprintf("The %s minors zvol task was not ok, last error %d\n", + msg[op], task->zt_error); + } else { + dprintf("The %s minors zvol task was not ok\n", msg[op]); + } +#else + (void) task; +#endif +} + /* * Create minors for the specified dataset, including children and snapshots. * Pay attention to the 'snapdev' property and iterate over the snapshots @@ -1451,14 +1483,27 @@ zvol_create_minors_cb(const char *dsname, void *arg) * 'visible' (which also verifies that the parent is a zvol), and if so, * a minor node for that snapshot is created. */ -void -zvol_create_minors_recursive(const char *name) +static void +zvol_create_minors_impl(zvol_task_t *task) { + const char *name = task->zt_name1; list_t minors_list; minors_job_t *job; + uint64_t snapdev; + int total = 0, done = 0, last_error, error; - if (zvol_inhibit_dev) + /* + * Note: the dsl_pool_config_lock must not be held. + * Minor node creation needs to obtain the zvol_state_lock. + * zvol_open() obtains the zvol_state_lock and then the dsl pool + * config lock. Therefore, we can't have the config lock now if + * we are going to wait for the zvol_state_lock, because it + * would be a lock order inversion which could lead to deadlock. + */ + + if (zvol_inhibit_dev) { return; + } /* * This is the list for prefetch jobs. Whenever we found a match @@ -1474,13 +1519,16 @@ zvol_create_minors_recursive(const char *name) if (strchr(name, '@') != NULL) { - uint64_t snapdev; - - int error = dsl_prop_get_integer(name, "snapdev", - &snapdev, NULL); - - if (error == 0 && snapdev == ZFS_SNAPDEV_VISIBLE) - (void) zvol_os_create_minor(name); + error = dsl_prop_get_integer(name, "snapdev", &snapdev, NULL); + if (error == 0 && snapdev == ZFS_SNAPDEV_VISIBLE) { + error = zvol_os_create_minor(name); + if (error == 0) { + done++; + } else { + last_error = error; + } + total++; + } } else { fstrans_cookie_t cookie = spl_fstrans_mark(); (void) dmu_objset_find(name, zvol_create_minors_cb, @@ -1495,41 +1543,30 @@ zvol_create_minors_recursive(const char *name) * sequentially. */ while ((job = list_remove_head(&minors_list)) != NULL) { - if (!job->error) - (void) zvol_os_create_minor(job->name); + if (!job->error) { + error = zvol_os_create_minor(job->name); + if (error == 0) { + done++; + } else { + last_error = error; + } + } else if (job->error == EINVAL) { + /* + * The objset, with the name requested by current job + * exist, but have the type different from zvol. + * Just ignore this sort of errors. + */ + done++; + } else { + last_error = job->error; + } + total++; kmem_strfree(job->name); kmem_free(job, sizeof (minors_job_t)); } list_destroy(&minors_list); -} - -void -zvol_create_minor(const char *name) -{ - /* - * Note: the dsl_pool_config_lock must not be held. - * Minor node creation needs to obtain the zvol_state_lock. - * zvol_open() obtains the zvol_state_lock and then the dsl pool - * config lock. Therefore, we can't have the config lock now if - * we are going to wait for the zvol_state_lock, because it - * would be a lock order inversion which could lead to deadlock. - */ - - if (zvol_inhibit_dev) - return; - - if (strchr(name, '@') != NULL) { - uint64_t snapdev; - - int error = dsl_prop_get_integer(name, - "snapdev", &snapdev, NULL); - - if (error == 0 && snapdev == ZFS_SNAPDEV_VISIBLE) - (void) zvol_os_create_minor(name); - } else { - (void) zvol_os_create_minor(name); - } + zvol_task_update_status(task, total, done, last_error); } /* @@ -1577,10 +1614,11 @@ zvol_free_task(void *arg) zvol_os_free(arg); } -void -zvol_remove_minors_impl(const char *name) +static void +zvol_remove_minors_impl(zvol_task_t *task) { zvol_state_t *zv, *zv_next; + const char *name = task ? task->zt_name1 : NULL; int namelen = ((name) ? strlen(name) : 0); taskqid_t t; list_t delay_list, free_list; @@ -1662,13 +1700,13 @@ zvol_remove_minors_impl(const char *name) } /* Remove minor for this specific volume only */ -static void +static int zvol_remove_minor_impl(const char *name) { zvol_state_t *zv = NULL, *zv_next; if (zvol_inhibit_dev) - return; + return (0); rw_enter(&zvol_state_lock, RW_WRITER); @@ -1684,7 +1722,7 @@ zvol_remove_minor_impl(const char *name) if (zv == NULL) { rw_exit(&zvol_state_lock); - return; + return (SET_ERROR(ENOENT)); } ASSERT(MUTEX_HELD(&zv->zv_state_lock)); @@ -1698,7 +1736,7 @@ zvol_remove_minor_impl(const char *name) mutex_exit(&zv->zv_state_lock); rw_exit(&zvol_state_lock); zvol_remove_minor_task(zv); - return; + return (0); } zvol_remove(zv); @@ -1708,16 +1746,20 @@ zvol_remove_minor_impl(const char *name) rw_exit(&zvol_state_lock); zvol_os_free(zv); + + return (0); } /* * Rename minors for specified dataset including children and snapshots. */ static void -zvol_rename_minors_impl(const char *oldname, const char *newname) +zvol_rename_minors_impl(zvol_task_t *task) { zvol_state_t *zv, *zv_next; - int oldnamelen; + const char *oldname = task->zt_name1; + const char *newname = task->zt_name2; + int total = 0, done = 0, last_error, error, oldnamelen; if (zvol_inhibit_dev) return; @@ -1732,24 +1774,31 @@ zvol_rename_minors_impl(const char *oldname, const char *newname) mutex_enter(&zv->zv_state_lock); if (strcmp(zv->zv_name, oldname) == 0) { - zvol_os_rename_minor(zv, newname); + error = zvol_os_rename_minor(zv, newname); } else if (strncmp(zv->zv_name, oldname, oldnamelen) == 0 && (zv->zv_name[oldnamelen] == '/' || zv->zv_name[oldnamelen] == '@')) { char *name = kmem_asprintf("%s%c%s", newname, zv->zv_name[oldnamelen], zv->zv_name + oldnamelen + 1); - zvol_os_rename_minor(zv, name); + error = zvol_os_rename_minor(zv, name); kmem_strfree(name); } - + if (error) { + last_error = error; + } else { + done++; + } + total++; mutex_exit(&zv->zv_state_lock); } rw_exit(&zvol_state_lock); + zvol_task_update_status(task, total, done, last_error); } typedef struct zvol_snapdev_cb_arg { + zvol_task_t *task; uint64_t snapdev; } zvol_snapdev_cb_arg_t; @@ -1757,26 +1806,31 @@ static int zvol_set_snapdev_cb(const char *dsname, void *param) { zvol_snapdev_cb_arg_t *arg = param; + int error = 0; if (strchr(dsname, '@') == NULL) return (0); switch (arg->snapdev) { case ZFS_SNAPDEV_VISIBLE: - (void) zvol_os_create_minor(dsname); + error = zvol_os_create_minor(dsname); break; case ZFS_SNAPDEV_HIDDEN: - (void) zvol_remove_minor_impl(dsname); + error = zvol_remove_minor_impl(dsname); break; } + zvol_task_update_status(arg->task, 1, error == 0, error); return (0); } static void -zvol_set_snapdev_impl(char *name, uint64_t snapdev) +zvol_set_snapdev_impl(zvol_task_t *task) { - zvol_snapdev_cb_arg_t arg = {snapdev}; + const char *name = task->zt_name1; + uint64_t snapdev = task->zt_value; + + zvol_snapdev_cb_arg_t arg = {task, snapdev}; fstrans_cookie_t cookie = spl_fstrans_mark(); /* * The zvol_set_snapdev_sync() sets snapdev appropriately @@ -1787,11 +1841,14 @@ zvol_set_snapdev_impl(char *name, uint64_t snapdev) } static void -zvol_set_volmode_impl(char *name, uint64_t volmode) +zvol_set_volmode_impl(zvol_task_t *task) { + const char *name = task->zt_name1; + uint64_t volmode = task->zt_value; fstrans_cookie_t cookie; uint64_t old_volmode; zvol_state_t *zv; + int error; if (strchr(name, '@') != NULL) return; @@ -1804,7 +1861,7 @@ zvol_set_volmode_impl(char *name, uint64_t volmode) */ zv = zvol_find_by_name(name, RW_NONE); if (zv == NULL && volmode == ZFS_VOLMODE_NONE) - return; + return; if (zv != NULL) { old_volmode = zv->zv_volmode; mutex_exit(&zv->zv_state_lock); @@ -1815,51 +1872,34 @@ zvol_set_volmode_impl(char *name, uint64_t volmode) cookie = spl_fstrans_mark(); switch (volmode) { case ZFS_VOLMODE_NONE: - (void) zvol_remove_minor_impl(name); + error = zvol_remove_minor_impl(name); break; case ZFS_VOLMODE_GEOM: case ZFS_VOLMODE_DEV: - (void) zvol_remove_minor_impl(name); - (void) zvol_os_create_minor(name); + error = zvol_remove_minor_impl(name); + /* + * The remove minor function call above, might be not + * needed, if volmode was switched from 'none' value. + * Ignore error in this case. + */ + if (error == ENOENT) + error = 0; + else if (error) + break; + error = zvol_os_create_minor(name); break; case ZFS_VOLMODE_DEFAULT: - (void) zvol_remove_minor_impl(name); + error = zvol_remove_minor_impl(name); if (zvol_volmode == ZFS_VOLMODE_NONE) break; else /* if zvol_volmode is invalid defaults to "geom" */ - (void) zvol_os_create_minor(name); + error = zvol_os_create_minor(name); break; } + zvol_task_update_status(task, 1, error == 0, error); spl_fstrans_unmark(cookie); } -static zvol_task_t * -zvol_task_alloc(zvol_async_op_t op, const char *name1, const char *name2, - uint64_t value) -{ - zvol_task_t *task; - - /* Never allow tasks on hidden names. */ - if (name1[0] == '$') - return (NULL); - - task = kmem_zalloc(sizeof (zvol_task_t), KM_SLEEP); - task->op = op; - task->value = value; - - strlcpy(task->name1, name1, sizeof (task->name1)); - if (name2 != NULL) - strlcpy(task->name2, name2, sizeof (task->name2)); - - return (task); -} - -static void -zvol_task_free(zvol_task_t *task) -{ - kmem_free(task, sizeof (zvol_task_t)); -} - /* * The worker thread function performed asynchronously. */ @@ -1868,25 +1908,29 @@ zvol_task_cb(void *arg) { zvol_task_t *task = arg; - switch (task->op) { + switch (task->zt_op) { + case ZVOL_ASYNC_CREATE_MINORS: + zvol_create_minors_impl(task); + break; case ZVOL_ASYNC_REMOVE_MINORS: - zvol_remove_minors_impl(task->name1); + zvol_remove_minors_impl(task); break; case ZVOL_ASYNC_RENAME_MINORS: - zvol_rename_minors_impl(task->name1, task->name2); + zvol_rename_minors_impl(task); break; case ZVOL_ASYNC_SET_SNAPDEV: - zvol_set_snapdev_impl(task->name1, task->value); + zvol_set_snapdev_impl(task); break; case ZVOL_ASYNC_SET_VOLMODE: - zvol_set_volmode_impl(task->name1, task->value); + zvol_set_volmode_impl(task); break; default: VERIFY(0); break; } - zvol_task_free(task); + zvol_task_report_status(task); + kmem_free(task, sizeof (zvol_task_t)); } typedef struct zvol_set_prop_int_arg { @@ -1931,23 +1975,17 @@ zvol_set_common_sync_cb(dsl_pool_t *dp, dsl_dataset_t *ds, void *arg) if (dsl_prop_get_int_ds(ds, prop_name, &prop) != 0) return (0); - switch (zsda->zsda_prop) { - case ZFS_PROP_VOLMODE: - task = zvol_task_alloc(ZVOL_ASYNC_SET_VOLMODE, dsname, - NULL, prop); - break; - case ZFS_PROP_SNAPDEV: - task = zvol_task_alloc(ZVOL_ASYNC_SET_SNAPDEV, dsname, - NULL, prop); - break; - default: - task = NULL; - break; - } - - if (task == NULL) + task = kmem_zalloc(sizeof (zvol_task_t), KM_SLEEP); + if (zsda->zsda_prop == ZFS_PROP_VOLMODE) { + task->zt_op = ZVOL_ASYNC_SET_VOLMODE; + } else if (zsda->zsda_prop == ZFS_PROP_SNAPDEV) { + task->zt_op = ZVOL_ASYNC_SET_SNAPDEV; + } else { + kmem_free(task, sizeof (zvol_task_t)); return (0); - + } + task->zt_value = prop; + strlcpy(task->zt_name1, dsname, sizeof (task->zt_name1)); (void) taskq_dispatch(dp->dp_spa->spa_zvol_taskq, zvol_task_cb, task, TQ_SLEEP); return (0); @@ -2001,15 +2039,34 @@ zvol_set_common(const char *ddname, zfs_prop_t prop, zprop_source_t source, } void -zvol_remove_minors(spa_t *spa, const char *name, boolean_t async) +zvol_create_minors(const char *name) { + spa_t *spa; zvol_task_t *task; taskqid_t id; - task = zvol_task_alloc(ZVOL_ASYNC_REMOVE_MINORS, name, NULL, ~0ULL); - if (task == NULL) + if (spa_open(name, &spa, FTAG) != 0) return; + task = kmem_zalloc(sizeof (zvol_task_t), KM_SLEEP); + task->zt_op = ZVOL_ASYNC_CREATE_MINORS; + strlcpy(task->zt_name1, name, sizeof (task->zt_name1)); + id = taskq_dispatch(spa->spa_zvol_taskq, zvol_task_cb, task, TQ_SLEEP); + if (id != TASKQID_INVALID) + taskq_wait_id(spa->spa_zvol_taskq, id); + + spa_close(spa, FTAG); +} + +void +zvol_remove_minors(spa_t *spa, const char *name, boolean_t async) +{ + zvol_task_t *task; + taskqid_t id; + + task = kmem_zalloc(sizeof (zvol_task_t), KM_SLEEP); + task->zt_op = ZVOL_ASYNC_REMOVE_MINORS; + strlcpy(task->zt_name1, name, sizeof (task->zt_name1)); id = taskq_dispatch(spa->spa_zvol_taskq, zvol_task_cb, task, TQ_SLEEP); if ((async == B_FALSE) && (id != TASKQID_INVALID)) taskq_wait_id(spa->spa_zvol_taskq, id); @@ -2022,10 +2079,10 @@ zvol_rename_minors(spa_t *spa, const char *name1, const char *name2, zvol_task_t *task; taskqid_t id; - task = zvol_task_alloc(ZVOL_ASYNC_RENAME_MINORS, name1, name2, ~0ULL); - if (task == NULL) - return; - + task = kmem_zalloc(sizeof (zvol_task_t), KM_SLEEP); + task->zt_op = ZVOL_ASYNC_RENAME_MINORS; + strlcpy(task->zt_name1, name1, sizeof (task->zt_name1)); + strlcpy(task->zt_name2, name2, sizeof (task->zt_name2)); id = taskq_dispatch(spa->spa_zvol_taskq, zvol_task_cb, task, TQ_SLEEP); if ((async == B_FALSE) && (id != TASKQID_INVALID)) taskq_wait_id(spa->spa_zvol_taskq, id); @@ -2144,7 +2201,7 @@ zvol_fini_impl(void) rw_destroy(&zvol_state_lock); if (ztqs->tqs_taskq == NULL) { - ASSERT3U(ztqs->tqs_cnt, ==, 0); + ASSERT0(ztqs->tqs_cnt); } else { for (uint_t i = 0; i < ztqs->tqs_cnt; i++) { ASSERT3P(ztqs->tqs_taskq[i], !=, NULL); diff --git a/sys/contrib/openzfs/module/zstd/zfs_zstd.c b/sys/contrib/openzfs/module/zstd/zfs_zstd.c index 4f7ac56a6273..391216d6e263 100644 --- a/sys/contrib/openzfs/module/zstd/zfs_zstd.c +++ b/sys/contrib/openzfs/module/zstd/zfs_zstd.c @@ -637,11 +637,10 @@ zfs_zstd_compress_buf(void *s_start, void *d_start, size_t s_len, size_t d_len, size_t actual_abort_size = zstd_abort_size; if (zstd_earlyabort_pass > 0 && zstd_level >= zstd_cutoff_level && s_len >= actual_abort_size) { - int pass_len = 1; abd_t sabd, dabd; abd_get_from_buf_struct(&sabd, s_start, s_len); abd_get_from_buf_struct(&dabd, d_start, d_len); - pass_len = zfs_lz4_compress(&sabd, &dabd, s_len, d_len, 0); + int pass_len = zfs_lz4_compress(&sabd, &dabd, s_len, d_len, 0); abd_free(&dabd); abd_free(&sabd); if (pass_len < d_len) { |