aboutsummaryrefslogtreecommitdiff
diff options
context:
space:
mode:
-rw-r--r--ObsoleteFiles.inc1
-rw-r--r--UPDATING7
-rw-r--r--cddl/lib/libicp/Makefile2
-rw-r--r--cddl/lib/libicp_rescue/Makefile2
-rw-r--r--contrib/tcpdump/print-pfsync.c12
-rw-r--r--lib/libifconfig/Symbol.map2
-rw-r--r--lib/libifconfig/libifconfig.h1
-rw-r--r--sbin/pfctl/pfctl.824
-rw-r--r--share/man/man4/iflib.414
-rw-r--r--share/man/man4/iwlwifi.44
-rw-r--r--share/man/man5/src.conf.59
-rw-r--r--share/mk/src.opts.mk2
-rw-r--r--share/mk/src.sys.mk2
-rw-r--r--stand/libsa/zfs/zfsimpl.c227
-rw-r--r--sys/arm64/arm64/elf32_machdep.c2
-rw-r--r--sys/cddl/boot/zfs/zfsimpl.h4
-rw-r--r--sys/conf/files.amd643
-rw-r--r--sys/conf/kern.pre.mk3
-rwxr-xr-xsys/contrib/openzfs/.github/workflows/scripts/qemu-2-start.sh2
-rwxr-xr-xsys/contrib/openzfs/.github/workflows/scripts/qemu-4-build-vm.sh23
-rw-r--r--sys/contrib/openzfs/.github/workflows/zfs-qemu-packages.yml13
-rw-r--r--sys/contrib/openzfs/.mailmap4
-rw-r--r--sys/contrib/openzfs/AUTHORS13
-rw-r--r--sys/contrib/openzfs/META2
-rw-r--r--sys/contrib/openzfs/cmd/zdb/zdb.c61
-rw-r--r--sys/contrib/openzfs/config/kernel-mkdir.m42
-rw-r--r--sys/contrib/openzfs/config/toolchain-simd.m444
-rw-r--r--sys/contrib/openzfs/contrib/debian/control8
-rw-r--r--sys/contrib/openzfs/contrib/icp/gcm-simd/boringssl/LICENSE253
-rw-r--r--sys/contrib/openzfs/contrib/icp/gcm-simd/boringssl/README11
-rw-r--r--sys/contrib/openzfs/contrib/icp/gcm-simd/boringssl/aes-gcm-avx2-x86_64-linux.S1328
-rw-r--r--sys/contrib/openzfs/contrib/initramfs/scripts/zfs3
-rw-r--r--sys/contrib/openzfs/include/os/linux/kernel/linux/simd_x86.h26
-rw-r--r--sys/contrib/openzfs/include/os/linux/zfs/sys/trace_zil.h6
-rw-r--r--sys/contrib/openzfs/include/sys/spa.h2
-rw-r--r--sys/contrib/openzfs/include/sys/zil_impl.h18
-rw-r--r--sys/contrib/openzfs/include/sys/zio.h29
-rw-r--r--sys/contrib/openzfs/include/sys/zvol_impl.h5
-rw-r--r--sys/contrib/openzfs/lib/libicp/Makefile.am1
-rw-r--r--sys/contrib/openzfs/lib/libspl/include/sys/simd.h28
-rw-r--r--sys/contrib/openzfs/lib/libzpool/kernel.c75
-rw-r--r--sys/contrib/openzfs/lib/libzutil/zutil_import.c39
-rw-r--r--sys/contrib/openzfs/man/man4/zfs.44
-rw-r--r--sys/contrib/openzfs/man/man8/zdb.814
-rw-r--r--sys/contrib/openzfs/man/man8/zfs-send.824
-rw-r--r--sys/contrib/openzfs/module/Kbuild.in1
-rw-r--r--sys/contrib/openzfs/module/icp/algs/modes/gcm.c371
-rw-r--r--sys/contrib/openzfs/module/icp/algs/modes/modes.c2
-rw-r--r--sys/contrib/openzfs/module/icp/asm-x86_64/modes/THIRDPARTYLICENSE.boringssl253
-rw-r--r--sys/contrib/openzfs/module/icp/asm-x86_64/modes/THIRDPARTYLICENSE.boringssl.descrip1
-rw-r--r--sys/contrib/openzfs/module/icp/asm-x86_64/modes/aesni-gcm-avx2-vaes.S1323
-rw-r--r--sys/contrib/openzfs/module/icp/include/modes/modes.h13
-rw-r--r--sys/contrib/openzfs/module/os/freebsd/zfs/kmod_core.c7
-rw-r--r--sys/contrib/openzfs/module/os/freebsd/zfs/zvol_os.c210
-rw-r--r--sys/contrib/openzfs/module/os/linux/zfs/zvol_os.c153
-rw-r--r--sys/contrib/openzfs/module/zcommon/simd_stat.c4
-rw-r--r--sys/contrib/openzfs/module/zfs/dbuf.c7
-rw-r--r--sys/contrib/openzfs/module/zfs/multilist.c4
-rw-r--r--sys/contrib/openzfs/module/zfs/spa_config.c115
-rw-r--r--sys/contrib/openzfs/module/zfs/spa_misc.c8
-rw-r--r--sys/contrib/openzfs/module/zfs/zil.c349
-rw-r--r--sys/contrib/openzfs/module/zfs/zio.c51
-rw-r--r--sys/contrib/openzfs/module/zfs/zvol.c281
-rw-r--r--sys/contrib/openzfs/module/zstd/zfs_zstd.c8
-rwxr-xr-xsys/contrib/openzfs/scripts/spdxcheck.pl1
-rw-r--r--sys/contrib/openzfs/tests/runfiles/common.run2
-rw-r--r--sys/contrib/openzfs/tests/zfs-tests/cmd/crypto_test.c2
-rw-r--r--sys/contrib/openzfs/tests/zfs-tests/tests/Makefile.am1
-rwxr-xr-xsys/contrib/openzfs/tests/zfs-tests/tests/functional/zvol/zvol_stress/zvol_stress_destroy.ksh66
-rw-r--r--sys/dev/gpio/acpi_gpiobus.c165
-rw-r--r--sys/dev/gpio/acpi_gpiobusvar.h6
-rw-r--r--sys/dev/gpio/gpioaei.c204
-rw-r--r--sys/dev/gpio/gpiobus.c19
-rw-r--r--sys/dev/gpio/gpiobus_internal.h2
-rw-r--r--sys/dev/gpio/ofw_gpiobus.c14
-rw-r--r--sys/dev/mlx5/mlx5_accel/mlx5_ipsec_offload.c2
-rw-r--r--sys/modules/zfs/Makefile7
-rw-r--r--sys/modules/zfs/zfs_config.h16
-rw-r--r--sys/modules/zfs/zfs_gitrev.h2
-rw-r--r--sys/net/if_pfsync.h4
-rw-r--r--sys/net/iflib.c92
-rw-r--r--sys/net/pfvar.h69
-rw-r--r--sys/netpfil/pf/if_pfsync.c12
-rw-r--r--sys/netpfil/pf/pf.c38
-rw-r--r--sys/netpfil/pf/pf_nl.c2
-rwxr-xr-xtools/build/depend-cleanup.sh64
-rw-r--r--tools/build/options/WITH_CLEAN4
87 files changed, 5186 insertions, 1128 deletions
diff --git a/ObsoleteFiles.inc b/ObsoleteFiles.inc
index aaec7ace84fc..95d4c06fe07e 100644
--- a/ObsoleteFiles.inc
+++ b/ObsoleteFiles.inc
@@ -76,7 +76,6 @@ OLD_FILES+=usr/lib/libopencsd.so
# 20250801: Move compile_et to /usr/sbin
OLD_FILES+=usr/share/et/compile_et
-OLD_DIRS+=usr/share/et
# 20250728: zfsboot (MBR) removed
OLD_FILES+=boot/zfsboot
diff --git a/UPDATING b/UPDATING
index f90036c1a3ad..e00b0c21f5c6 100644
--- a/UPDATING
+++ b/UPDATING
@@ -27,6 +27,13 @@ NOTE TO PEOPLE WHO THINK THAT FreeBSD 15.x IS SLOW:
world, or to merely disable the most expensive debugging functionality
at runtime, run "ln -s 'abort:false,junk:false' /etc/malloc.conf".)
+20250819:
+ The CLEAN option has been switched back from default-on to default-off.
+ This reverts the 20250808 change below, which had reverted the 20240729
+ change before it. Note that some src.conf(5) options are known to break
+ ABI or compatibility in ways that may require a clean build initially
+ when switched.
+
20250816:
Sendmail's libmilter has been moved to its own package. If you want
to compile applications that use libmilter, you should install the
diff --git a/cddl/lib/libicp/Makefile b/cddl/lib/libicp/Makefile
index f097e7e6ff58..8e801246215f 100644
--- a/cddl/lib/libicp/Makefile
+++ b/cddl/lib/libicp/Makefile
@@ -11,6 +11,7 @@ ASM_SOURCES_AS = \
asm-x86_64/aes/aes_aesni.S \
asm-x86_64/modes/gcm_pclmulqdq.S \
asm-x86_64/modes/aesni-gcm-x86_64.S \
+ asm-x86_64/modes/aesni-gcm-avx2-vaes.S \
asm-x86_64/modes/ghash-x86_64.S \
asm-x86_64/sha2/sha256-x86_64.S \
asm-x86_64/sha2/sha512-x86_64.S \
@@ -112,6 +113,7 @@ CFLAGS.aes_amd64.S+= -DLOCORE
CFLAGS.aes_aesni.S+= -DLOCORE
CFLAGS.gcm_pclmulqdq.S+= -DLOCORE
CFLAGS.aesni-gcm-x86_64.S+= -DLOCORE
+CFLAGS.aesni-gcm-avx2-vaes.S+= -DLOCORE
CFLAGS.ghash-x86_64.S+= -DLOCORE
CFLAGS.sha256-x86_64.S+= -DLOCORE
CFLAGS.sha512-x86_64.S+= -DLOCORE
diff --git a/cddl/lib/libicp_rescue/Makefile b/cddl/lib/libicp_rescue/Makefile
index 3a8b6746fe61..0a5a81f4ab7f 100644
--- a/cddl/lib/libicp_rescue/Makefile
+++ b/cddl/lib/libicp_rescue/Makefile
@@ -11,6 +11,7 @@ ASM_SOURCES_AS = \
asm-x86_64/aes/aes_aesni.S \
asm-x86_64/modes/gcm_pclmulqdq.S \
asm-x86_64/modes/aesni-gcm-x86_64.S \
+ asm-x86_64/modes/aesni-gcm-avx2-vaes.S \
asm-x86_64/sha2/sha256-x86_64.S \
asm-x86_64/sha2/sha512-x86_64.S \
asm-x86_64/blake3/blake3_avx2.S \
@@ -109,6 +110,7 @@ CFLAGS.aes_amd64.S+= -DLOCORE
CFLAGS.aes_aesni.S+= -DLOCORE
CFLAGS.gcm_pclmulqdq.S+= -DLOCORE
CFLAGS.aesni-gcm-x86_64.S+= -DLOCORE
+CFLAGS.aesni-gcm-avx2-vaes.S+= -DLOCORE
CFLAGS.ghash-x86_64.S+= -DLOCORE
CFLAGS.sha256-x86_64.S+= -DLOCORE
CFLAGS.sha512-x86_64.S+= -DLOCORE
diff --git a/contrib/tcpdump/print-pfsync.c b/contrib/tcpdump/print-pfsync.c
index 6bf9abaf3903..e4f11930816c 100644
--- a/contrib/tcpdump/print-pfsync.c
+++ b/contrib/tcpdump/print-pfsync.c
@@ -53,8 +53,8 @@
static void pfsync_print(netdissect_options *, struct pfsync_header *,
const u_char *, u_int);
static void print_src_dst(netdissect_options *,
- const struct pfsync_state_peer *,
- const struct pfsync_state_peer *, uint8_t);
+ const struct pf_state_peer_export *,
+ const struct pf_state_peer_export *, uint8_t);
static void print_state(netdissect_options *, union pfsync_state_union *, int);
void
@@ -330,7 +330,7 @@ print_host(netdissect_options *ndo, struct pf_addr *addr, uint16_t port,
}
static void
-print_seq(netdissect_options *ndo, const struct pfsync_state_peer *p)
+print_seq(netdissect_options *ndo, const struct pf_state_peer_export *p)
{
if (p->seqdiff)
ND_PRINT("[%u + %u](+%u)", ntohl(p->seqlo),
@@ -341,8 +341,8 @@ print_seq(netdissect_options *ndo, const struct pfsync_state_peer *p)
}
static void
-print_src_dst(netdissect_options *ndo, const struct pfsync_state_peer *src,
- const struct pfsync_state_peer *dst, uint8_t proto)
+print_src_dst(netdissect_options *ndo, const struct pf_state_peer_export *src,
+ const struct pf_state_peer_export *dst, uint8_t proto)
{
if (proto == IPPROTO_TCP) {
@@ -390,7 +390,7 @@ print_src_dst(netdissect_options *ndo, const struct pfsync_state_peer *src,
static void
print_state(netdissect_options *ndo, union pfsync_state_union *s, int version)
{
- struct pfsync_state_peer *src, *dst;
+ struct pf_state_peer_export *src, *dst;
struct pfsync_state_key *sk, *nk;
int min, sec;
diff --git a/lib/libifconfig/Symbol.map b/lib/libifconfig/Symbol.map
index 2d80fb31652a..2e11ff963909 100644
--- a/lib/libifconfig/Symbol.map
+++ b/lib/libifconfig/Symbol.map
@@ -40,7 +40,6 @@ FBSD_1.6 {
ifconfig_open;
ifconfig_set_capability;
ifconfig_set_description;
- ifconfig_set_fib;
ifconfig_set_metric;
ifconfig_set_mtu;
ifconfig_set_name;
@@ -81,7 +80,6 @@ FBSD_1.6 {
ifconfig_sfp_fc_speed_symbol;
ifconfig_sfp_id_description;
ifconfig_sfp_id_display;
- ifconfig_sfp_id_is_qsfp;
ifconfig_sfp_id_symbol;
ifconfig_sfp_rev_description;
ifconfig_sfp_rev_symbol;
diff --git a/lib/libifconfig/libifconfig.h b/lib/libifconfig/libifconfig.h
index b2f0cf9744ea..a5ce7b375830 100644
--- a/lib/libifconfig/libifconfig.h
+++ b/lib/libifconfig/libifconfig.h
@@ -171,7 +171,6 @@ int ifconfig_set_name(ifconfig_handle_t *h, const char *name,
const char *newname);
int ifconfig_get_orig_name(ifconfig_handle_t *h, const char *ifname,
char **orig_name);
-int ifconfig_set_fib(ifconfig_handle_t *h, const char *name, int fib);
int ifconfig_get_fib(ifconfig_handle_t *h, const char *name, int *fib);
int ifconfig_set_mtu(ifconfig_handle_t *h, const char *name, const int mtu);
int ifconfig_get_mtu(ifconfig_handle_t *h, const char *name, int *mtu);
diff --git a/sbin/pfctl/pfctl.8 b/sbin/pfctl/pfctl.8
index f582c6301124..5a74a8fd3444 100644
--- a/sbin/pfctl/pfctl.8
+++ b/sbin/pfctl/pfctl.8
@@ -24,7 +24,7 @@
.\" (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
.\" THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.\"
-.Dd July 7, 2025
+.Dd August 5, 2025
.Dt PFCTL 8
.Os
.Sh NAME
@@ -410,6 +410,7 @@ This is the default behaviour.
.It Fl o Cm profile
Enable basic ruleset optimizations with profiling.
.El
+.Pp
For further information on the ruleset optimizer, see
.Xr pf.conf 5 .
.It Fl P
@@ -431,7 +432,7 @@ Perform reverse DNS lookups on states and tables when displaying them.
and
.Fl r
are mutually exclusive.
-.It Fl s Ar modifier
+.It Fl s Ar modifier Op Fl R Ar id
Show the filter parameters specified by
.Ar modifier
(may be abbreviated):
@@ -563,19 +564,16 @@ no free ports in translation port range
.It Fl S
Do not perform domain name resolution.
If a name cannot be resolved without DNS, an error will be reported.
-.It Fl T Ar command Op Ar address ...
+.It Fl t Ar table Fl T Ar command Op Ar address ...
Specify the
.Ar command
-(may be abbreviated) to apply to the table.
+(may be abbreviated) to apply to
+.Ar table .
Commands include:
.Pp
-.Bl -tag -width xxxxxxxxxxxx -compact
-.It Fl T Cm kill
-Kill a table.
-.It Fl T Cm flush
-Flush all addresses of a table.
+.Bl -tag -width "-T expire number" -compact
.It Fl T Cm add
-Add one or more addresses in a table.
+Add one or more addresses to a table.
Automatically create a persistent table if it does not exist.
.It Fl T Cm delete
Delete one or more addresses from a table.
@@ -586,6 +584,10 @@ seconds ago.
For entries which have never had their statistics cleared,
.Ar number
refers to the time they were added to the table.
+.It Fl T Cm flush
+Flush all addresses in a table.
+.It Fl T Cm kill
+Kill a table.
.It Fl T Cm replace
Replace the addresses of the table.
Automatically create a persistent table if it does not exist.
@@ -765,8 +767,6 @@ tables of the same name from anchors attached below it.
.It C
This flag is set when per-address counters are enabled on the table.
.El
-.It Fl t Ar table
-Specify the name of the table.
.It Fl v
Produce more verbose output.
A second use of
diff --git a/share/man/man4/iflib.4 b/share/man/man4/iflib.4
index 0114263e6ca2..2040698f0087 100644
--- a/share/man/man4/iflib.4
+++ b/share/man/man4/iflib.4
@@ -1,4 +1,4 @@
-.Dd September 27, 2018
+.Dd August 20, 2025
.Dt IFLIB 4
.Os
.Sh NAME
@@ -64,6 +64,18 @@ If this is zero or not set, an RX and TX queue pair will be assigned to each
core.
When set to a non-zero value, TX queues are assigned to cores following the
last RX queue.
+.It Va simple_tx
+When set to one, iflib uses a simple transmit routine with no queuing at all.
+By default, iflib uses a highly optimized, lockless, transmit queue called
+mp_ring.
+This performs well when there are more CPU cores than NIC
+queues and prevents lock contention for transmit resources.
+Unfortunately, mp_ring incurs unneeded overheads on workloads where
+resource contention is not a problem (well behaved applications on
+systems where there are as many NIC queues as CPU cores).
+Note that when this is enabled, the tx_abdicate sysctl is no longer
+applicable and is ignored.
+Defaults to zero.
.El
.Pp
These
diff --git a/share/man/man4/iwlwifi.4 b/share/man/man4/iwlwifi.4
index 4a251f239a55..660f6a9bf57c 100644
--- a/share/man/man4/iwlwifi.4
+++ b/share/man/man4/iwlwifi.4
@@ -27,7 +27,7 @@
.\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
.\" SUCH DAMAGE.
.\"
-.Dd June 13, 2025
+.Dd August 19, 2025
.Dt IWLWIFI 4
.Os
.Sh NAME
@@ -331,7 +331,7 @@ driver first appeared in
802.11n and 802.11ac support for the 22000 and later chipsets first appeared in
.Fx 14.3 .
.Sh BUGS
-Certainly.
+.Lk https://bugs.freebsd.org/bugzilla/showdependencytree.cgi?id=iwlwifi "iwlwifi known bugs"
.Pp
While
.Nm
diff --git a/share/man/man5/src.conf.5 b/share/man/man5/src.conf.5
index 2e694bfe3293..f79d160255bf 100644
--- a/share/man/man5/src.conf.5
+++ b/share/man/man5/src.conf.5
@@ -1,5 +1,5 @@
.\" DO NOT EDIT-- this file is @generated by tools/build/options/makeman.
-.Dd August 8, 2025
+.Dd August 19, 2025
.Dt SRC.CONF 5
.Os
.Sh NAME
@@ -411,8 +411,11 @@ Build clang-format.
.It Va WITHOUT_CLANG_FULL
Avoid building the ARCMigrate, Rewriter and StaticAnalyzer components of
the Clang C/C++ compiler.
-.It Va WITHOUT_CLEAN
-Do not clean before building world and/or kernel.
+.It Va WITH_CLEAN
+Clean before building world and/or kernel.
+Note that recording a new epoch in
+.Pa .clean_build_epoch
+in the root of the source tree will also force a clean world build.
.It Va WITHOUT_CPP
Do not build
.Xr cpp 1 .
diff --git a/share/mk/src.opts.mk b/share/mk/src.opts.mk
index f146a4b24424..8a52fe03221e 100644
--- a/share/mk/src.opts.mk
+++ b/share/mk/src.opts.mk
@@ -80,7 +80,6 @@ __DEFAULT_YES_OPTIONS = \
CDDL \
CLANG \
CLANG_BOOTSTRAP \
- CLEAN \
CPP \
CROSS_COMPILER \
CRYPT \
@@ -201,6 +200,7 @@ __DEFAULT_NO_OPTIONS = \
BHYVE_SNAPSHOT \
CLANG_EXTRAS \
CLANG_FORMAT \
+ CLEAN \
DIALOG \
DETECT_TZ_CHANGES \
DISK_IMAGE_TOOLS_BOOTSTRAP \
diff --git a/share/mk/src.sys.mk b/share/mk/src.sys.mk
index d5c2af0c559d..2b9fc255a26d 100644
--- a/share/mk/src.sys.mk
+++ b/share/mk/src.sys.mk
@@ -42,7 +42,7 @@ CFLAGS+= ${CFCOMMONFLAG}
CFLAGS+= -fmacro-prefix-map=${SRCTOP}=/usr/src -fdebug-prefix-map=${SRCTOP}=/usr/src
.endif
-DEFAULTWARNS= 6
+DEFAULTWARNS?= 6
# tempting, but bsd.compiler.mk causes problems this early
# probably need to remove dependence on bsd.own.mk
diff --git a/stand/libsa/zfs/zfsimpl.c b/stand/libsa/zfs/zfsimpl.c
index 971d71d098d3..f15d9b016068 100644
--- a/stand/libsa/zfs/zfsimpl.c
+++ b/stand/libsa/zfs/zfsimpl.c
@@ -107,11 +107,6 @@ typedef struct indirect_vsd {
} indirect_vsd_t;
/*
- * List of all vdevs, chained through v_alllink.
- */
-static vdev_list_t zfs_vdevs;
-
-/*
* List of supported read-incompatible ZFS features. Do not add here features
* marked as ZFEATURE_FLAG_READONLY_COMPAT, they are irrelevant for read-only!
*/
@@ -167,7 +162,6 @@ vdev_indirect_mapping_entry_phys_t *
static void
zfs_init(void)
{
- STAILQ_INIT(&zfs_vdevs);
STAILQ_INIT(&zfs_pools);
dnode_cache_buf = malloc(SPA_MAXBLOCKSIZE);
@@ -839,16 +833,27 @@ vdev_replacing_read(vdev_t *vdev, const blkptr_t *bp, void *buf,
return (kid->v_read(kid, bp, buf, offset, bytes));
}
+/*
+ * List of vdevs that were fully initialized from their own label, but later a
+ * newer label was found that obsoleted the stale label, freeing its
+ * configuration tree. We keep those vdevs around, since a new configuration
+ * may include them.
+ */
+static vdev_list_t orphans = STAILQ_HEAD_INITIALIZER(orphans);
+
static vdev_t *
-vdev_find(uint64_t guid)
+vdev_find(vdev_list_t *list, uint64_t guid)
{
- vdev_t *vdev;
+ vdev_t *vdev, *safe;
- STAILQ_FOREACH(vdev, &zfs_vdevs, v_alllink)
+ STAILQ_FOREACH_SAFE(vdev, list, v_childlink, safe) {
if (vdev->v_guid == guid)
return (vdev);
+ if ((vdev = vdev_find(&vdev->v_children, guid)) != NULL)
+ return (vdev);
+ }
- return (0);
+ return (NULL);
}
static vdev_t *
@@ -857,6 +862,11 @@ vdev_create(uint64_t guid, vdev_read_t *_read)
vdev_t *vdev;
vdev_indirect_config_t *vic;
+ if ((vdev = vdev_find(&orphans, guid))) {
+ STAILQ_REMOVE(&orphans, vdev, vdev, v_childlink);
+ return (vdev);
+ }
+
vdev = calloc(1, sizeof(vdev_t));
if (vdev != NULL) {
STAILQ_INIT(&vdev->v_children);
@@ -871,7 +881,6 @@ vdev_create(uint64_t guid, vdev_read_t *_read)
if (_read != NULL) {
vic = &vdev->vdev_indirect_config;
vic->vic_prev_indirect_vdev = UINT64_MAX;
- STAILQ_INSERT_TAIL(&zfs_vdevs, vdev, v_alllink);
}
}
@@ -1035,22 +1044,19 @@ vdev_init(uint64_t guid, const nvlist_t *nvlist, vdev_t **vdevp)
* STAILQ_INSERT_AFTER.
*/
static vdev_t *
-vdev_find_previous(vdev_t *top_vdev, vdev_t *vdev)
+vdev_find_previous(vdev_t *top_vdev, uint64_t id)
{
vdev_t *v, *previous;
- if (STAILQ_EMPTY(&top_vdev->v_children))
- return (NULL);
-
previous = NULL;
STAILQ_FOREACH(v, &top_vdev->v_children, v_childlink) {
- if (v->v_id > vdev->v_id)
+ if (v->v_id > id)
return (previous);
- if (v->v_id == vdev->v_id)
+ if (v->v_id == id)
return (v);
- if (v->v_id < vdev->v_id)
+ if (v->v_id < id)
previous = v;
}
return (previous);
@@ -1072,7 +1078,7 @@ vdev_child_count(vdev_t *vdev)
/*
* Insert vdev into top_vdev children list. List is ordered by v_id.
*/
-static void
+static vdev_t *
vdev_insert(vdev_t *top_vdev, vdev_t *vdev)
{
vdev_t *previous;
@@ -1085,7 +1091,7 @@ vdev_insert(vdev_t *top_vdev, vdev_t *vdev)
* so we can use either STAILQ_INSERT_HEAD or STAILQ_INSERT_AFTER
* as STAILQ does not have insert before.
*/
- previous = vdev_find_previous(top_vdev, vdev);
+ previous = vdev_find_previous(top_vdev, vdev->v_id);
if (previous == NULL) {
STAILQ_INSERT_HEAD(&top_vdev->v_children, vdev, v_childlink);
@@ -1094,7 +1100,8 @@ vdev_insert(vdev_t *top_vdev, vdev_t *vdev)
* This vdev was configured from label config,
* do not insert duplicate.
*/
- return;
+ free(vdev);
+ return (previous);
} else {
STAILQ_INSERT_AFTER(&top_vdev->v_children, previous, vdev,
v_childlink);
@@ -1103,26 +1110,28 @@ vdev_insert(vdev_t *top_vdev, vdev_t *vdev)
count = vdev_child_count(top_vdev);
if (top_vdev->v_nchildren < count)
top_vdev->v_nchildren = count;
+ return (vdev);
}
static int
-vdev_from_nvlist(spa_t *spa, uint64_t top_guid, uint64_t txg,
- const nvlist_t *nvlist)
+vdev_from_nvlist(spa_t *spa, uint64_t top_guid, uint64_t label_guid,
+ uint64_t txg, const nvlist_t *nvlist)
{
vdev_t *top_vdev, *vdev;
nvlist_t **kids = NULL;
int rc, nkids;
/* Get top vdev. */
- top_vdev = vdev_find(top_guid);
+ top_vdev = vdev_find(&spa->spa_root_vdev->v_children, top_guid);
if (top_vdev == NULL) {
rc = vdev_init(top_guid, nvlist, &top_vdev);
if (rc != 0)
return (rc);
top_vdev->v_spa = spa;
top_vdev->v_top = top_vdev;
+ top_vdev->v_label = label_guid;
top_vdev->v_txg = txg;
- vdev_insert(spa->spa_root_vdev, top_vdev);
+ (void )vdev_insert(spa->spa_root_vdev, top_vdev);
}
/* Add children if there are any. */
@@ -1143,7 +1152,7 @@ vdev_from_nvlist(spa_t *spa, uint64_t top_guid, uint64_t txg,
vdev->v_spa = spa;
vdev->v_top = top_vdev;
- vdev_insert(top_vdev, vdev);
+ vdev = vdev_insert(top_vdev, vdev);
}
} else {
/*
@@ -1162,30 +1171,6 @@ done:
return (rc);
}
-static int
-vdev_init_from_label(spa_t *spa, const nvlist_t *nvlist)
-{
- uint64_t pool_guid, top_guid, txg;
- nvlist_t *vdevs;
- int rc;
-
- if (nvlist_find(nvlist, ZPOOL_CONFIG_POOL_GUID, DATA_TYPE_UINT64,
- NULL, &pool_guid, NULL) ||
- nvlist_find(nvlist, ZPOOL_CONFIG_TOP_GUID, DATA_TYPE_UINT64,
- NULL, &top_guid, NULL) ||
- nvlist_find(nvlist, ZPOOL_CONFIG_POOL_TXG, DATA_TYPE_UINT64,
- NULL, &txg, NULL) != 0 ||
- nvlist_find(nvlist, ZPOOL_CONFIG_VDEV_TREE, DATA_TYPE_NVLIST,
- NULL, &vdevs, NULL)) {
- printf("ZFS: can't find vdev details\n");
- return (ENOENT);
- }
-
- rc = vdev_from_nvlist(spa, top_guid, txg, vdevs);
- nvlist_destroy(vdevs);
- return (rc);
-}
-
static void
vdev_set_state(vdev_t *vdev)
{
@@ -1232,14 +1217,14 @@ vdev_set_state(vdev_t *vdev)
}
static int
-vdev_update_from_nvlist(uint64_t top_guid, const nvlist_t *nvlist)
+vdev_update_from_nvlist(vdev_t *root, uint64_t top_guid, const nvlist_t *nvlist)
{
vdev_t *vdev;
nvlist_t **kids = NULL;
int rc, nkids;
/* Update top vdev. */
- vdev = vdev_find(top_guid);
+ vdev = vdev_find(&root->v_children, top_guid);
if (vdev != NULL)
vdev_set_initial_state(vdev, nvlist);
@@ -1255,7 +1240,7 @@ vdev_update_from_nvlist(uint64_t top_guid, const nvlist_t *nvlist)
if (rc != 0)
break;
- vdev = vdev_find(guid);
+ vdev = vdev_find(&root->v_children, guid);
if (vdev != NULL)
vdev_set_initial_state(vdev, kids[i]);
}
@@ -1271,10 +1256,6 @@ vdev_update_from_nvlist(uint64_t top_guid, const nvlist_t *nvlist)
return (rc);
}
-/*
- * Shall not be called on root vdev, that is not linked into zfs_vdevs.
- * See comment in vdev_create().
- */
static void
vdev_free(struct vdev *vdev)
{
@@ -1282,8 +1263,10 @@ vdev_free(struct vdev *vdev)
STAILQ_FOREACH_SAFE(kid, &vdev->v_children, v_childlink, safe)
vdev_free(kid);
- STAILQ_REMOVE(&zfs_vdevs, vdev, vdev, v_alllink);
- free(vdev);
+ if (vdev->v_phys_read != NULL)
+ STAILQ_INSERT_HEAD(&orphans, vdev, v_childlink);
+ else
+ free(vdev);
}
static int
@@ -1329,15 +1312,16 @@ vdev_init_from_nvlist(spa_t *spa, const nvlist_t *nvlist)
NULL, &guid, NULL);
if (rc != 0)
break;
- vdev = vdev_find(guid);
+ vdev = vdev_find(&spa->spa_root_vdev->v_children, guid);
/*
* Top level vdev is missing, create it.
* XXXGL: how can this happen?
*/
if (vdev == NULL)
- rc = vdev_from_nvlist(spa, guid, 0, kids[i]);
+ rc = vdev_from_nvlist(spa, guid, 0, 0, kids[i]);
else
- rc = vdev_update_from_nvlist(guid, kids[i]);
+ rc = vdev_update_from_nvlist(spa->spa_root_vdev, guid,
+ kids[i]);
if (rc != 0)
break;
}
@@ -1355,6 +1339,53 @@ vdev_init_from_nvlist(spa_t *spa, const nvlist_t *nvlist)
return (rc);
}
+static bool
+nvlist_find_child_guid(const nvlist_t *nvlist, uint64_t guid)
+{
+ nvlist_t **kids = NULL;
+ int nkids, i;
+ bool rv = false;
+
+ if (nvlist_find(nvlist, ZPOOL_CONFIG_CHILDREN, DATA_TYPE_NVLIST_ARRAY,
+ &nkids, &kids, NULL) != 0)
+ nkids = 0;
+
+ for (i = 0; i < nkids; i++) {
+ uint64_t kid_guid;
+
+ if (nvlist_find(kids[i], ZPOOL_CONFIG_GUID, DATA_TYPE_UINT64,
+ NULL, &kid_guid, NULL) != 0)
+ break;
+ if (kid_guid == guid)
+ rv = true;
+ else
+ rv = nvlist_find_child_guid(kids[i], guid);
+ if (rv)
+ break;
+ }
+
+ for (i = 0; i < nkids; i++)
+ nvlist_destroy(kids[i]);
+ free(kids);
+
+ return (rv);
+}
+
+static bool
+nvlist_find_vdev_guid(const nvlist_t *nvlist, uint64_t guid)
+{
+ nvlist_t *vdevs;
+ bool rv;
+
+ if (nvlist_find(nvlist, ZPOOL_CONFIG_VDEV_TREE, DATA_TYPE_NVLIST, NULL,
+ &vdevs, NULL) != 0)
+ return (false);
+ rv = nvlist_find_child_guid(vdevs, guid);
+ nvlist_destroy(vdevs);
+
+ return (rv);
+}
+
static spa_t *
spa_find_by_guid(uint64_t guid)
{
@@ -2023,8 +2054,8 @@ vdev_probe(vdev_phys_read_t *_read, vdev_phys_write_t *_write, void *priv,
{
vdev_t vtmp;
spa_t *spa;
- vdev_t *vdev;
- nvlist_t *nvl;
+ vdev_t *vdev, *top;
+ nvlist_t *nvl, *vdevs;
uint64_t val;
uint64_t guid, pool_guid, top_guid, txg;
const char *pool_name;
@@ -2083,6 +2114,7 @@ vdev_probe(vdev_phys_read_t *_read, vdev_phys_write_t *_write, void *priv,
if (nvlist_find(nvl, ZPOOL_CONFIG_POOL_TXG, DATA_TYPE_UINT64,
NULL, &txg, NULL) != 0 ||
+ txg == 0 ||
nvlist_find(nvl, ZPOOL_CONFIG_TOP_GUID, DATA_TYPE_UINT64,
NULL, &top_guid, NULL) != 0 ||
nvlist_find(nvl, ZPOOL_CONFIG_POOL_GUID, DATA_TYPE_UINT64,
@@ -2092,7 +2124,7 @@ vdev_probe(vdev_phys_read_t *_read, vdev_phys_write_t *_write, void *priv,
nvlist_find(nvl, ZPOOL_CONFIG_GUID, DATA_TYPE_UINT64,
NULL, &guid, NULL) != 0) {
/*
- * Cache and spare devices end up here - just ignore
+ * Cache, spare and replaced devices end up here - just ignore
* them.
*/
nvlist_destroy(nvl);
@@ -2119,22 +2151,47 @@ vdev_probe(vdev_phys_read_t *_read, vdev_phys_write_t *_write, void *priv,
nvlist_destroy(nvl);
return (ENOMEM);
}
- } else {
- struct vdev *kid;
-
- STAILQ_FOREACH(kid, &spa->spa_root_vdev->v_children,
- v_childlink)
- if (kid->v_guid == top_guid && kid->v_txg < txg) {
- printf("ZFS: pool %s vdev %s ignoring stale "
- "label from txg 0x%jx, using 0x%jx@0x%jx\n",
- spa->spa_name, kid->v_name,
- kid->v_txg, guid, txg);
+ }
+
+ /*
+ * Check if configuration is already known. If configuration is known
+ * and txg numbers don't match, we got 2x2 scenarios here. First, is
+ * the label being read right now _newer_ than the one read before.
+ * Second, is the vdev that provided the stale label _present_ in the
+ * newer configuration. If neither is true, we completely ignore the
+ * label.
+ */
+ STAILQ_FOREACH(top, &spa->spa_root_vdev->v_children, v_childlink)
+ if (top->v_guid == top_guid) {
+ bool newer, present;
+
+ if (top->v_txg == txg)
+ break;
+ newer = (top->v_txg < txg);
+ present = newer ?
+ nvlist_find_vdev_guid(nvl, top->v_label) :
+ (vdev_find(&top->v_children, guid) != NULL);
+ printf("ZFS: pool %s vdev %s %s stale label from "
+ "0x%jx@0x%jx, %s 0x%jx@0x%jx\n",
+ spa->spa_name, top->v_name,
+ present ? "using" : "ignoring",
+ newer ? top->v_label : guid,
+ newer ? top->v_txg : txg,
+ present ? "referred by" : "using",
+ newer ? guid : top->v_label,
+ newer ? txg : top->v_txg);
+ if (newer) {
STAILQ_REMOVE(&spa->spa_root_vdev->v_children,
- kid, vdev, v_childlink);
- vdev_free(kid);
+ top, vdev, v_childlink);
+ vdev_free(top);
+ break;
+ } else if (present) {
break;
+ } else {
+ nvlist_destroy(nvl);
+ return (EIO);
}
- }
+ }
/*
* Get the vdev tree and create our in-core copy of it.
@@ -2142,14 +2199,22 @@ vdev_probe(vdev_phys_read_t *_read, vdev_phys_write_t *_write, void *priv,
* be some kind of alias (overlapping slices, dangerously dedicated
* disks etc).
*/
- vdev = vdev_find(guid);
+ vdev = vdev_find(&spa->spa_root_vdev->v_children, guid);
/* Has this vdev already been inited? */
if (vdev && vdev->v_phys_read) {
nvlist_destroy(nvl);
return (EIO);
}
- rc = vdev_init_from_label(spa, nvl);
+ if (nvlist_find(nvl, ZPOOL_CONFIG_VDEV_TREE, DATA_TYPE_NVLIST, NULL,
+ &vdevs, NULL)) {
+ printf("ZFS: can't find vdev details\n");
+ nvlist_destroy(nvl);
+ return (ENOENT);
+ }
+
+ rc = vdev_from_nvlist(spa, top_guid, guid, txg, vdevs);
+ nvlist_destroy(vdevs);
nvlist_destroy(nvl);
if (rc != 0)
return (rc);
@@ -2158,7 +2223,7 @@ vdev_probe(vdev_phys_read_t *_read, vdev_phys_write_t *_write, void *priv,
* We should already have created an incomplete vdev for this
* vdev. Find it and initialise it with our read proc.
*/
- vdev = vdev_find(guid);
+ vdev = vdev_find(&spa->spa_root_vdev->v_children, guid);
if (vdev != NULL) {
vdev->v_phys_read = _read;
vdev->v_phys_write = _write;
diff --git a/sys/arm64/arm64/elf32_machdep.c b/sys/arm64/arm64/elf32_machdep.c
index 7cd5327b9f1b..5c81c6cdce3d 100644
--- a/sys/arm64/arm64/elf32_machdep.c
+++ b/sys/arm64/arm64/elf32_machdep.c
@@ -195,7 +195,7 @@ freebsd32_fetch_syscall_args(struct thread *td)
register_t *ap;
struct syscall_args *sa;
int error, i, nap, narg;
- unsigned int args[4];
+ unsigned int args[6];
nap = 4;
p = td->td_proc;
diff --git a/sys/cddl/boot/zfs/zfsimpl.h b/sys/cddl/boot/zfs/zfsimpl.h
index 915aeeda3c9e..c9de1fe4c391 100644
--- a/sys/cddl/boot/zfs/zfsimpl.h
+++ b/sys/cddl/boot/zfs/zfsimpl.h
@@ -2021,11 +2021,11 @@ typedef struct vdev_indirect_config {
typedef struct vdev {
STAILQ_ENTRY(vdev) v_childlink; /* link in parent's child list */
- STAILQ_ENTRY(vdev) v_alllink; /* link in global vdev list */
vdev_list_t v_children; /* children of this vdev */
const char *v_name; /* vdev name */
uint64_t v_guid; /* vdev guid */
- uint64_t v_txg; /* most recent transaction */
+ uint64_t v_label; /* label instantiated from (top vdev) */
+ uint64_t v_txg; /* most recent transaction (top vdev) */
uint64_t v_id; /* index in parent */
uint64_t v_psize; /* physical device capacity */
int v_ashift; /* offset to block shift */
diff --git a/sys/conf/files.amd64 b/sys/conf/files.amd64
index 80548320c3fc..9b6ba03b78df 100644
--- a/sys/conf/files.amd64
+++ b/sys/conf/files.amd64
@@ -419,6 +419,9 @@ contrib/openzfs/module/icp/asm-x86_64/blake3/blake3_avx512.S optional zfs com
contrib/openzfs/module/icp/asm-x86_64/blake3/blake3_sse2.S optional zfs compile-with "${ZFS_S}"
contrib/openzfs/module/icp/asm-x86_64/blake3/blake3_sse41.S optional zfs compile-with "${ZFS_S}"
+# zfs AVX2 implementation of aes-gcm from BoringSSL
+contrib/openzfs/module/icp/asm-x86_64/modes/aesni-gcm-avx2-vaes.S optional zfs compile-with "${ZFS_S}"
+
# zfs sha2 hash support
zfs-sha256-x86_64.o optional zfs \
dependency "$S/contrib/openzfs/module/icp/asm-x86_64/sha2/sha256-x86_64.S" \
diff --git a/sys/conf/kern.pre.mk b/sys/conf/kern.pre.mk
index 78178065e15b..1fcfd6467e7f 100644
--- a/sys/conf/kern.pre.mk
+++ b/sys/conf/kern.pre.mk
@@ -214,7 +214,8 @@ ZFS_CFLAGS+= -I$S/contrib/openzfs/module/icp/include \
.if ${MACHINE_ARCH} == "amd64"
ZFS_CFLAGS+= -D__x86_64 -DHAVE_SSE2 -DHAVE_SSSE3 -DHAVE_SSE4_1 -DHAVE_SSE4_2 \
- -DHAVE_AVX -DHAVE_AVX2 -DHAVE_AVX512F -DHAVE_AVX512VL -DHAVE_AVX512BW
+ -DHAVE_AVX -DHAVE_AVX2 -DHAVE_AVX512F -DHAVE_AVX512VL -DHAVE_AVX512BW \
+ -DHAVE_VAES -DHAVE_VPCLMULQDQ
.endif
.if ${MACHINE_ARCH} == "i386" || ${MACHINE_ARCH} == "powerpc" || \
diff --git a/sys/contrib/openzfs/.github/workflows/scripts/qemu-2-start.sh b/sys/contrib/openzfs/.github/workflows/scripts/qemu-2-start.sh
index 885a64037f89..70a2364f1fc6 100755
--- a/sys/contrib/openzfs/.github/workflows/scripts/qemu-2-start.sh
+++ b/sys/contrib/openzfs/.github/workflows/scripts/qemu-2-start.sh
@@ -109,7 +109,7 @@ case "$OS" in
KSRC="$FREEBSD_SNAP/../amd64/$FreeBSD/src.txz"
;;
freebsd15-0c)
- FreeBSD="15.0-CURRENT"
+ FreeBSD="15.0-PRERELEASE"
OSNAME="FreeBSD $FreeBSD"
OSv="freebsd14.0"
URLxz="$FREEBSD_SNAP/$FreeBSD/amd64/Latest/FreeBSD-$FreeBSD-amd64-BASIC-CI-ufs.raw.xz"
diff --git a/sys/contrib/openzfs/.github/workflows/scripts/qemu-4-build-vm.sh b/sys/contrib/openzfs/.github/workflows/scripts/qemu-4-build-vm.sh
index 17e976ebcc39..2807d9e77127 100755
--- a/sys/contrib/openzfs/.github/workflows/scripts/qemu-4-build-vm.sh
+++ b/sys/contrib/openzfs/.github/workflows/scripts/qemu-4-build-vm.sh
@@ -5,12 +5,13 @@
#
# Usage:
#
-# qemu-4-build-vm.sh OS [--enable-debug][--dkms][--poweroff]
-# [--release][--repo][--tarball]
+# qemu-4-build-vm.sh OS [--enable-debug][--dkms][--patch-level NUM]
+# [--poweroff][--release][--repo][--tarball]
#
# OS: OS name like 'fedora41'
# --enable-debug: Build RPMs with '--enable-debug' (for testing)
# --dkms: Build DKMS RPMs as well
+# --patch-level NUM: Use a custom patch level number for packages.
# --poweroff: Power-off the VM after building
# --release Build zfs-release*.rpm as well
# --repo After building everything, copy RPMs into /tmp/repo
@@ -21,6 +22,7 @@
ENABLE_DEBUG=""
DKMS=""
+PATCH_LEVEL=""
POWEROFF=""
RELEASE=""
REPO=""
@@ -35,6 +37,11 @@ while [[ $# -gt 0 ]]; do
DKMS=1
shift
;;
+ --patch-level)
+ PATCH_LEVEL=$2
+ shift
+ shift
+ ;;
--poweroff)
POWEROFF=1
shift
@@ -215,6 +222,10 @@ function rpm_build_and_install() {
run ./autogen.sh
echo "##[endgroup]"
+ if [ -n "$PATCH_LEVEL" ] ; then
+ sed -i -E 's/(Release:\s+)1/\1'$PATCH_LEVEL'/g' META
+ fi
+
echo "##[group]Configure"
run ./configure --enable-debuginfo $extra
echo "##[endgroup]"
@@ -328,7 +339,13 @@ fi
# almalinux9.5
# fedora42
source /etc/os-release
-sudo hostname "$ID$VERSION_ID"
+ if which hostnamectl &> /dev/null ; then
+ # Fedora 42+ use hostnamectl
+ sudo hostnamectl set-hostname "$ID$VERSION_ID"
+ sudo hostnamectl set-hostname --pretty "$ID$VERSION_ID"
+else
+ sudo hostname "$ID$VERSION_ID"
+fi
# save some sysinfo
uname -a > /var/tmp/uname.txt
diff --git a/sys/contrib/openzfs/.github/workflows/zfs-qemu-packages.yml b/sys/contrib/openzfs/.github/workflows/zfs-qemu-packages.yml
index 5b5afe746859..d8a95954fe1a 100644
--- a/sys/contrib/openzfs/.github/workflows/zfs-qemu-packages.yml
+++ b/sys/contrib/openzfs/.github/workflows/zfs-qemu-packages.yml
@@ -32,6 +32,11 @@ on:
options:
- "Build RPMs"
- "Test repo"
+ patch_level:
+ type: string
+ required: false
+ default: ""
+ description: "(optional) patch level number"
repo_url:
type: string
required: false
@@ -78,7 +83,13 @@ jobs:
mkdir -p /tmp/repo
ssh zfs@vm0 '$HOME/zfs/.github/workflows/scripts/qemu-test-repo-vm.sh' ${{ github.event.inputs.repo_url }}
else
- .github/workflows/scripts/qemu-4-build.sh --repo --release --dkms --tarball ${{ matrix.os }}
+ EXTRA=""
+ if [ -n "${{ github.event.inputs.patch_level }}" ] ; then
+ EXTRA="--patch-level ${{ github.event.inputs.patch_level }}"
+ fi
+
+ .github/workflows/scripts/qemu-4-build.sh $EXTRA \
+ --repo --release --dkms --tarball ${{ matrix.os }}
fi
- name: Prepare artifacts
diff --git a/sys/contrib/openzfs/.mailmap b/sys/contrib/openzfs/.mailmap
index b6d942c000b8..e6f09c6c9d43 100644
--- a/sys/contrib/openzfs/.mailmap
+++ b/sys/contrib/openzfs/.mailmap
@@ -23,6 +23,7 @@
# These maps are making names consistent where they have varied but the email
# address has never changed. In most cases, the full name is in the
# Signed-off-by of a commit with a matching author.
+Achill Gilgenast <achill@achill.org>
Ahelenia Ziemiańska <nabijaczleweli@gmail.com>
Ahelenia Ziemiańska <nabijaczleweli@nabijaczleweli.xyz>
Alex John <alex@stty.io>
@@ -37,6 +38,7 @@ Crag Wang <crag0715@gmail.com>
Damian Szuberski <szuberskidamian@gmail.com>
Daniel Kolesa <daniel@octaforge.org>
Debabrata Banerjee <dbavatar@gmail.com>
+Diwakar Kristappagari <diwakar-k@hpe.com>
Finix Yan <yanchongwen@hotmail.com>
Gaurav Kumar <gauravk.18@gmail.com>
Gionatan Danti <g.danti@assyoma.it>
@@ -145,6 +147,7 @@ Gaurav Kumar <gauravk.18@gmail.com> <gaurkuma@users.noreply.github.com>
George Gaydarov <git@gg7.io> <gg7@users.noreply.github.com>
Georgy Yakovlev <gyakovlev@gentoo.org> <168902+gyakovlev@users.noreply.github.com>
Gerardwx <gerardw@alum.mit.edu> <Gerardwx@users.noreply.github.com>
+Germano Massullo <germano.massullo@gmail.com> <Germano0@users.noreply.github.com>
Gian-Carlo DeFazio <defazio1@llnl.gov> <defaziogiancarlo@users.noreply.github.com>
Giuseppe Di Natale <dinatale2@llnl.gov> <dinatale2@users.noreply.github.com>
Hajo Möller <dasjoe@gmail.com> <dasjoe@users.noreply.github.com>
@@ -164,6 +167,7 @@ John Ramsden <johnramsden@riseup.net> <johnramsden@users.noreply.github.com>
Jonathon Fernyhough <jonathon@m2x.dev> <559369+jonathonf@users.noreply.github.com>
Jose Luis Duran <jlduran@gmail.com> <jlduran@users.noreply.github.com>
Justin Hibbits <chmeeedalf@gmail.com> <chmeeedalf@users.noreply.github.com>
+Kaitlin Hoang <kthoang@amazon.com> <khoang98@users.noreply.github.com>
Kevin Greene <kevin.greene@delphix.com> <104801862+kxgreene@users.noreply.github.com>
Kevin Jin <lostking2008@hotmail.com> <33590050+jxdking@users.noreply.github.com>
Kevin P. Fleming <kevin@km6g.us> <kpfleming@users.noreply.github.com>
diff --git a/sys/contrib/openzfs/AUTHORS b/sys/contrib/openzfs/AUTHORS
index a9d249a66f1e..6c34c07f39ef 100644
--- a/sys/contrib/openzfs/AUTHORS
+++ b/sys/contrib/openzfs/AUTHORS
@@ -10,6 +10,7 @@ PAST MAINTAINERS:
CONTRIBUTORS:
Aaron Fineman <abyxcos@gmail.com>
+ Achill Gilgenast <achill@achill.org>
Adam D. Moss <c@yotes.com>
Adam Leventhal <ahl@delphix.com>
Adam Stevko <adam.stevko@gmail.com>
@@ -59,6 +60,7 @@ CONTRIBUTORS:
Andreas Buschmann <andreas.buschmann@tech.net.de>
Andreas Dilger <adilger@intel.com>
Andreas Vögele <andreas@andreasvoegele.com>
+ Andres <a-d-j-i@users.noreply.github.com>
Andrew Barnes <barnes333@gmail.com>
Andrew Hamilton <ahamilto@tjhsst.edu>
Andrew Innes <andrew.c12@gmail.com>
@@ -72,6 +74,7 @@ CONTRIBUTORS:
Andrey Prokopenko <job@terem.fr>
Andrey Vesnovaty <andrey.vesnovaty@gmail.com>
Andriy Gapon <avg@freebsd.org>
+ Andriy Tkachuk <andriy.tkachuk@seagate.com>
Andy Bakun <github@thwartedefforts.org>
Andy Fiddaman <omnios@citrus-it.co.uk>
Aniruddha Shankar <k@191a.net>
@@ -120,6 +123,7 @@ CONTRIBUTORS:
Caleb James DeLisle <calebdelisle@lavabit.com>
Cameron Harr <harr1@llnl.gov>
Cao Xuewen <cao.xuewen@zte.com.cn>
+ Carl George <carlwgeorge@gmail.com>
Carlo Landmeter <clandmeter@gmail.com>
Carlos Alberto Lopez Perez <clopez@igalia.com>
Cedric Maunoury <cedric.maunoury@gmail.com>
@@ -200,6 +204,7 @@ CONTRIBUTORS:
Dimitri John Ledkov <xnox@ubuntu.com>
Dimitry Andric <dimitry@andric.com>
Dirkjan Bussink <d.bussink@gmail.com>
+ Diwakar Kristappagari <diwakar-k@hpe.com>
Dmitry Khasanov <pik4ez@gmail.com>
Dominic Pearson <dsp@technoanimal.net>
Dominik Hassler <hadfl@omniosce.org>
@@ -250,6 +255,7 @@ CONTRIBUTORS:
George Wilson <gwilson@delphix.com>
Georgy Yakovlev <ya@sysdump.net>
Gerardwx <gerardw@alum.mit.edu>
+ Germano Massullo <germano.massullo@gmail.com>
Gian-Carlo DeFazio <defazio1@llnl.gov>
Gionatan Danti <g.danti@assyoma.it>
Giuseppe Di Natale <guss80@gmail.com>
@@ -287,6 +293,7 @@ CONTRIBUTORS:
Igor K <igor@dilos.org>
Igor Kozhukhov <ikozhukhov@gmail.com>
Igor Lvovsky <ilvovsky@gmail.com>
+ Igor Ostapenko <pm@igoro.pro>
ilbsmart <wgqimut@gmail.com>
Ilkka Sovanto <github@ilkka.kapsi.fi>
illiliti <illiliti@protonmail.com>
@@ -326,6 +333,7 @@ CONTRIBUTORS:
Jinshan Xiong <jinshan.xiong@intel.com>
Jitendra Patidar <jitendra.patidar@nutanix.com>
JK Dingwall <james@dingwall.me.uk>
+ Joel Low <joel@joelsplace.sg>
Joe Stein <joe.stein@delphix.com>
John-Mark Gurney <jmg@funkthat.com>
John Albietz <inthecloud247@gmail.com>
@@ -374,6 +382,7 @@ CONTRIBUTORS:
Kevin Jin <lostking2008@hotmail.com>
Kevin P. Fleming <kevin@km6g.us>
Kevin Tanguy <kevin.tanguy@ovh.net>
+ khoang98 <khoang98@users.noreply.github.com>
KireinaHoro <i@jsteward.moe>
Kjeld Schouten-Lebbing <kjeld@schouten-lebbing.nl>
Kleber Tarcísio <klebertarcisio@yahoo.com.br>
@@ -447,6 +456,7 @@ CONTRIBUTORS:
Max Zettlmeißl <max@zettlmeissl.de>
Md Islam <mdnahian@outlook.com>
megari <megari@iki.fi>
+ Meriel Luna Mittelbach <lunarlambda@gmail.com>
Michael D Labriola <michael.d.labriola@gmail.com>
Michael Franzl <michael@franzl.name>
Michael Gebetsroither <michael@mgeb.org>
@@ -494,6 +504,7 @@ CONTRIBUTORS:
Orivej Desh <orivej@gmx.fr>
Pablo Correa Gómez <ablocorrea@hotmail.com>
Palash Gandhi <pbg4930@rit.edu>
+ Patrick Fasano <patrick@patrickfasano.com>
Patrick Mooney <pmooney@pfmooney.com>
Patrik Greco <sikevux@sikevux.se>
Paul B. Henson <henson@acm.org>
@@ -535,6 +546,7 @@ CONTRIBUTORS:
Remy Blank <remy.blank@pobox.com>
renelson <bnelson@nelsonbe.com>
Reno Reckling <e-github@wthack.de>
+ René Wirnata <rene.wirnata@pandascience.net>
Ricardo M. Correia <ricardo.correia@oracle.com>
Riccardo Schirone <rschirone91@gmail.com>
Richard Allen <belperite@gmail.com>
@@ -640,6 +652,7 @@ CONTRIBUTORS:
tleydxdy <shironeko.github@tesaguri.club>
Tobin Harding <me@tobin.cc>
Todd Seidelmann <seidelma@users.noreply.github.com>
+ Todd Zullinger <tmz@pobox.com>
Tom Caputi <tcaputi@datto.com>
Tom Matthews <tom@axiom-partners.com>
Tomohiro Kusumi <kusumi.tomohiro@gmail.com>
diff --git a/sys/contrib/openzfs/META b/sys/contrib/openzfs/META
index 47f0795bfa11..1a9c671feac6 100644
--- a/sys/contrib/openzfs/META
+++ b/sys/contrib/openzfs/META
@@ -6,5 +6,5 @@ Release: 1
Release-Tags: relext
License: CDDL
Author: OpenZFS
-Linux-Maximum: 6.15
+Linux-Maximum: 6.16
Linux-Minimum: 4.18
diff --git a/sys/contrib/openzfs/cmd/zdb/zdb.c b/sys/contrib/openzfs/cmd/zdb/zdb.c
index 66d5fbd6adbe..957e6c67dd12 100644
--- a/sys/contrib/openzfs/cmd/zdb/zdb.c
+++ b/sys/contrib/openzfs/cmd/zdb/zdb.c
@@ -127,6 +127,7 @@ static zfs_range_tree_t *mos_refd_objs;
static spa_t *spa;
static objset_t *os;
static boolean_t kernel_init_done;
+static boolean_t corruption_found = B_FALSE;
static void snprintf_blkptr_compact(char *, size_t, const blkptr_t *,
boolean_t);
@@ -250,6 +251,7 @@ sublivelist_verify_func(void *args, dsl_deadlist_entry_t *dle)
&e->svbr_blk, B_TRUE);
(void) printf("\tERROR: %d unmatched FREE(s): %s\n",
e->svbr_refcnt, blkbuf);
+ corruption_found = B_TRUE;
}
zfs_btree_destroy(&sv->sv_pair);
@@ -405,6 +407,7 @@ verify_livelist_allocs(metaslab_verify_t *mv, uint64_t txg,
(u_longlong_t)DVA_GET_ASIZE(&found->svb_dva),
(u_longlong_t)found->svb_allocated_txg,
(u_longlong_t)txg);
+ corruption_found = B_TRUE;
}
}
}
@@ -426,6 +429,7 @@ metaslab_spacemap_validation_cb(space_map_entry_t *sme, void *arg)
(u_longlong_t)txg, (u_longlong_t)offset,
(u_longlong_t)size, (u_longlong_t)mv->mv_vdid,
(u_longlong_t)mv->mv_msid);
+ corruption_found = B_TRUE;
} else {
zfs_range_tree_add(mv->mv_allocated,
offset, size);
@@ -439,6 +443,7 @@ metaslab_spacemap_validation_cb(space_map_entry_t *sme, void *arg)
(u_longlong_t)txg, (u_longlong_t)offset,
(u_longlong_t)size, (u_longlong_t)mv->mv_vdid,
(u_longlong_t)mv->mv_msid);
+ corruption_found = B_TRUE;
} else {
zfs_range_tree_remove(mv->mv_allocated,
offset, size);
@@ -526,6 +531,7 @@ mv_populate_livelist_allocs(metaslab_verify_t *mv, sublivelist_verify_t *sv)
(u_longlong_t)DVA_GET_VDEV(&svb->svb_dva),
(u_longlong_t)DVA_GET_OFFSET(&svb->svb_dva),
(u_longlong_t)DVA_GET_ASIZE(&svb->svb_dva));
+ corruption_found = B_TRUE;
continue;
}
@@ -542,6 +548,7 @@ mv_populate_livelist_allocs(metaslab_verify_t *mv, sublivelist_verify_t *sv)
(u_longlong_t)DVA_GET_VDEV(&svb->svb_dva),
(u_longlong_t)DVA_GET_OFFSET(&svb->svb_dva),
(u_longlong_t)DVA_GET_ASIZE(&svb->svb_dva));
+ corruption_found = B_TRUE;
continue;
}
@@ -655,6 +662,7 @@ livelist_metaslab_validate(spa_t *spa)
}
(void) printf("ERROR: Found livelist blocks marked as allocated "
"for indirect vdevs:\n");
+ corruption_found = B_TRUE;
zfs_btree_index_t *where = NULL;
sublivelist_verify_block_t *svb;
@@ -827,7 +835,7 @@ usage(void)
(void) fprintf(stderr, "Specify an option more than once (e.g. -bb) "
"to make only that option verbose\n");
(void) fprintf(stderr, "Default is to dump everything non-verbosely\n");
- zdb_exit(1);
+ zdb_exit(2);
}
static void
@@ -2583,19 +2591,17 @@ snprintf_blkptr_compact(char *blkbuf, size_t buflen, const blkptr_t *bp,
}
}
-static void
+static u_longlong_t
print_indirect(spa_t *spa, blkptr_t *bp, const zbookmark_phys_t *zb,
const dnode_phys_t *dnp)
{
char blkbuf[BP_SPRINTF_LEN];
+ u_longlong_t offset;
int l;
- if (!BP_IS_EMBEDDED(bp)) {
- ASSERT3U(BP_GET_TYPE(bp), ==, dnp->dn_type);
- ASSERT3U(BP_GET_LEVEL(bp), ==, zb->zb_level);
- }
+ offset = (u_longlong_t)blkid2offset(dnp, bp, zb);
- (void) printf("%16llx ", (u_longlong_t)blkid2offset(dnp, bp, zb));
+ (void) printf("%16llx ", offset);
ASSERT(zb->zb_level >= 0);
@@ -2610,19 +2616,38 @@ print_indirect(spa_t *spa, blkptr_t *bp, const zbookmark_phys_t *zb,
snprintf_blkptr_compact(blkbuf, sizeof (blkbuf), bp, B_FALSE);
if (dump_opt['Z'] && BP_GET_COMPRESS(bp) == ZIO_COMPRESS_ZSTD)
snprintf_zstd_header(spa, blkbuf, sizeof (blkbuf), bp);
- (void) printf("%s\n", blkbuf);
+ (void) printf("%s", blkbuf);
+
+ if (!BP_IS_EMBEDDED(bp)) {
+ if (BP_GET_TYPE(bp) != dnp->dn_type) {
+ (void) printf(" (ERROR: Block pointer type "
+ "(%llu) does not match dnode type (%hhu))",
+ BP_GET_TYPE(bp), dnp->dn_type);
+ corruption_found = B_TRUE;
+ }
+ if (BP_GET_LEVEL(bp) != zb->zb_level) {
+ (void) printf(" (ERROR: Block pointer level "
+ "(%llu) does not match bookmark level (%ld))",
+ BP_GET_LEVEL(bp), zb->zb_level);
+ corruption_found = B_TRUE;
+ }
+ }
+ (void) printf("\n");
+
+ return (offset);
}
static int
visit_indirect(spa_t *spa, const dnode_phys_t *dnp,
blkptr_t *bp, const zbookmark_phys_t *zb)
{
+ u_longlong_t offset;
int err = 0;
if (BP_GET_BIRTH(bp) == 0)
return (0);
- print_indirect(spa, bp, zb, dnp);
+ offset = print_indirect(spa, bp, zb, dnp);
if (BP_GET_LEVEL(bp) > 0 && !BP_IS_HOLE(bp)) {
arc_flags_t flags = ARC_FLAG_WAIT;
@@ -2652,8 +2677,15 @@ visit_indirect(spa_t *spa, const dnode_phys_t *dnp,
break;
fill += BP_GET_FILL(cbp);
}
- if (!err)
- ASSERT3U(fill, ==, BP_GET_FILL(bp));
+ if (!err) {
+ if (fill != BP_GET_FILL(bp)) {
+ (void) printf("%16llx: Block pointer "
+ "fill (%llu) does not match calculated "
+ "value (%lu)\n", offset, BP_GET_FILL(bp),
+ fill);
+ corruption_found = B_TRUE;
+ }
+ }
arc_buf_destroy(buf, &buf);
}
@@ -2909,6 +2941,7 @@ dump_full_bpobj(bpobj_t *bpo, const char *name, int indent)
(void) printf("ERROR %u while trying to open "
"subobj id %llu\n",
error, (u_longlong_t)subobj);
+ corruption_found = B_TRUE;
continue;
}
dump_full_bpobj(&subbpo, "subobj", indent + 1);
@@ -3088,6 +3121,7 @@ bpobj_count_refd(bpobj_t *bpo)
(void) printf("ERROR %u while trying to open "
"subobj id %llu\n",
error, (u_longlong_t)subobj);
+ corruption_found = B_TRUE;
continue;
}
bpobj_count_refd(&subbpo);
@@ -9634,7 +9668,7 @@ main(int argc, char **argv)
} else if (objset_str && !zdb_numeric(objset_str + 1) &&
dump_opt['N']) {
printf("Supply a numeric objset ID with -N\n");
- error = 1;
+ error = 2;
goto fini;
}
} else {
@@ -9936,5 +9970,8 @@ fini:
if (kernel_init_done)
kernel_fini();
+ if (corruption_found && error == 0)
+ error = 3;
+
return (error);
}
diff --git a/sys/contrib/openzfs/config/kernel-mkdir.m4 b/sys/contrib/openzfs/config/kernel-mkdir.m4
index c1aebc387abe..78b32447c593 100644
--- a/sys/contrib/openzfs/config/kernel-mkdir.m4
+++ b/sys/contrib/openzfs/config/kernel-mkdir.m4
@@ -84,6 +84,8 @@ AC_DEFUN([ZFS_AC_KERNEL_MKDIR], [
AC_DEFINE(HAVE_IOPS_MKDIR_DENTRY, 1,
[iops->mkdir() returns struct dentry*])
],[
+ AC_MSG_RESULT(no)
+
dnl #
dnl # 6.3 API change
dnl # mkdir() takes struct mnt_idmap * as the first arg
diff --git a/sys/contrib/openzfs/config/toolchain-simd.m4 b/sys/contrib/openzfs/config/toolchain-simd.m4
index 344807fc830c..f18c91007cde 100644
--- a/sys/contrib/openzfs/config/toolchain-simd.m4
+++ b/sys/contrib/openzfs/config/toolchain-simd.m4
@@ -24,6 +24,8 @@ AC_DEFUN([ZFS_AC_CONFIG_ALWAYS_TOOLCHAIN_SIMD], [
ZFS_AC_CONFIG_TOOLCHAIN_CAN_BUILD_AES
ZFS_AC_CONFIG_TOOLCHAIN_CAN_BUILD_PCLMULQDQ
ZFS_AC_CONFIG_TOOLCHAIN_CAN_BUILD_MOVBE
+ ZFS_AC_CONFIG_TOOLCHAIN_CAN_BUILD_VAES
+ ZFS_AC_CONFIG_TOOLCHAIN_CAN_BUILD_VPCLMULQDQ
ZFS_AC_CONFIG_TOOLCHAIN_CAN_BUILD_XSAVE
ZFS_AC_CONFIG_TOOLCHAIN_CAN_BUILD_XSAVEOPT
ZFS_AC_CONFIG_TOOLCHAIN_CAN_BUILD_XSAVES
@@ -447,6 +449,48 @@ AC_DEFUN([ZFS_AC_CONFIG_TOOLCHAIN_CAN_BUILD_MOVBE], [
])
dnl #
+dnl # ZFS_AC_CONFIG_TOOLCHAIN_CAN_BUILD_VAES
+dnl #
+AC_DEFUN([ZFS_AC_CONFIG_TOOLCHAIN_CAN_BUILD_VAES], [
+ AC_MSG_CHECKING([whether host toolchain supports VAES])
+
+ AC_LINK_IFELSE([AC_LANG_SOURCE([
+ [
+ int main()
+ {
+ __asm__ __volatile__("vaesenc %ymm0, %ymm1, %ymm0");
+ return (0);
+ }
+ ]])], [
+ AC_MSG_RESULT([yes])
+ AC_DEFINE([HAVE_VAES], 1, [Define if host toolchain supports VAES])
+ ], [
+ AC_MSG_RESULT([no])
+ ])
+])
+
+dnl #
+dnl # ZFS_AC_CONFIG_TOOLCHAIN_CAN_BUILD_VPCLMULQDQ
+dnl #
+AC_DEFUN([ZFS_AC_CONFIG_TOOLCHAIN_CAN_BUILD_VPCLMULQDQ], [
+ AC_MSG_CHECKING([whether host toolchain supports VPCLMULQDQ])
+
+ AC_LINK_IFELSE([AC_LANG_SOURCE([
+ [
+ int main()
+ {
+ __asm__ __volatile__("vpclmulqdq %0, %%ymm4, %%ymm3, %%ymm5" :: "i"(0));
+ return (0);
+ }
+ ]])], [
+ AC_MSG_RESULT([yes])
+ AC_DEFINE([HAVE_VPCLMULQDQ], 1, [Define if host toolchain supports VPCLMULQDQ])
+ ], [
+ AC_MSG_RESULT([no])
+ ])
+])
+
+dnl #
dnl # ZFS_AC_CONFIG_TOOLCHAIN_CAN_BUILD_XSAVE
dnl #
AC_DEFUN([ZFS_AC_CONFIG_TOOLCHAIN_CAN_BUILD_XSAVE], [
diff --git a/sys/contrib/openzfs/contrib/debian/control b/sys/contrib/openzfs/contrib/debian/control
index 96a2bdd88665..c5358dedc0fd 100644
--- a/sys/contrib/openzfs/contrib/debian/control
+++ b/sys/contrib/openzfs/contrib/debian/control
@@ -100,8 +100,8 @@ Depends: ${misc:Depends}, ${shlibs:Depends}
# The libcurl4 is loaded through dlopen("libcurl.so.4").
# https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=988521
Recommends: libcurl4
-Breaks: libzfs2, libzfs4, libzfs4linux, libzfs6linux
-Replaces: libzfs2, libzfs4, libzfs4linux, libzfs6linux
+Breaks: libzfs2, libzfs4, libzfs4linux, libzfs6linux, openzfs-libzfs4
+Replaces: libzfs2, libzfs4, libzfs4linux, libzfs6linux, openzfs-libzfs4
Conflicts: libzfs6linux
Description: OpenZFS filesystem library for Linux - general support
OpenZFS is a storage platform that encompasses the functionality of
@@ -128,8 +128,8 @@ Package: openzfs-libzpool6
Section: contrib/libs
Architecture: linux-any
Depends: ${misc:Depends}, ${shlibs:Depends}
-Breaks: libzpool2, libzpool5, libzpool5linux, libzpool6linux
-Replaces: libzpool2, libzpool5, libzpool5linux, libzpool6linux
+Breaks: libzpool2, libzpool5, libzpool6linux
+Replaces: libzpool2, libzpool5, libzpool6linux
Conflicts: libzpool6linux
Description: OpenZFS pool library for Linux
OpenZFS is a storage platform that encompasses the functionality of
diff --git a/sys/contrib/openzfs/contrib/icp/gcm-simd/boringssl/LICENSE b/sys/contrib/openzfs/contrib/icp/gcm-simd/boringssl/LICENSE
new file mode 100644
index 000000000000..04c03a37e0cb
--- /dev/null
+++ b/sys/contrib/openzfs/contrib/icp/gcm-simd/boringssl/LICENSE
@@ -0,0 +1,253 @@
+BoringSSL is a fork of OpenSSL. As such, large parts of it fall under OpenSSL
+licensing. Files that are completely new have a Google copyright and an ISC
+license. This license is reproduced at the bottom of this file.
+
+Contributors to BoringSSL are required to follow the CLA rules for Chromium:
+https://cla.developers.google.com/clas
+
+Files in third_party/ have their own licenses, as described therein. The MIT
+license, for third_party/fiat, which, unlike other third_party directories, is
+compiled into non-test libraries, is included below.
+
+The OpenSSL toolkit stays under a dual license, i.e. both the conditions of the
+OpenSSL License and the original SSLeay license apply to the toolkit. See below
+for the actual license texts. Actually both licenses are BSD-style Open Source
+licenses. In case of any license issues related to OpenSSL please contact
+openssl-core@openssl.org.
+
+The following are Google-internal bug numbers where explicit permission from
+some authors is recorded for use of their work. (This is purely for our own
+record keeping.)
+ 27287199
+ 27287880
+ 27287883
+ 263291445
+
+
+ OpenSSL License
+ ---------------
+
+/* ====================================================================
+ * Copyright (c) 1998-2011 The OpenSSL Project. All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ *
+ * 2. Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in
+ * the documentation and/or other materials provided with the
+ * distribution.
+ *
+ * 3. All advertising materials mentioning features or use of this
+ * software must display the following acknowledgment:
+ * "This product includes software developed by the OpenSSL Project
+ * for use in the OpenSSL Toolkit. (http://www.openssl.org/)"
+ *
+ * 4. The names "OpenSSL Toolkit" and "OpenSSL Project" must not be used to
+ * endorse or promote products derived from this software without
+ * prior written permission. For written permission, please contact
+ * openssl-core@openssl.org.
+ *
+ * 5. Products derived from this software may not be called "OpenSSL"
+ * nor may "OpenSSL" appear in their names without prior written
+ * permission of the OpenSSL Project.
+ *
+ * 6. Redistributions of any form whatsoever must retain the following
+ * acknowledgment:
+ * "This product includes software developed by the OpenSSL Project
+ * for use in the OpenSSL Toolkit (http://www.openssl.org/)"
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE OpenSSL PROJECT ``AS IS'' AND ANY
+ * EXPRESSED OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+ * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE OpenSSL PROJECT OR
+ * ITS CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
+ * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
+ * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
+ * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED
+ * OF THE POSSIBILITY OF SUCH DAMAGE.
+ * ====================================================================
+ *
+ * This product includes cryptographic software written by Eric Young
+ * (eay@cryptsoft.com). This product includes software written by Tim
+ * Hudson (tjh@cryptsoft.com).
+ *
+ */
+
+ Original SSLeay License
+ -----------------------
+
+/* Copyright (C) 1995-1998 Eric Young (eay@cryptsoft.com)
+ * All rights reserved.
+ *
+ * This package is an SSL implementation written
+ * by Eric Young (eay@cryptsoft.com).
+ * The implementation was written so as to conform with Netscapes SSL.
+ *
+ * This library is free for commercial and non-commercial use as long as
+ * the following conditions are aheared to. The following conditions
+ * apply to all code found in this distribution, be it the RC4, RSA,
+ * lhash, DES, etc., code; not just the SSL code. The SSL documentation
+ * included with this distribution is covered by the same copyright terms
+ * except that the holder is Tim Hudson (tjh@cryptsoft.com).
+ *
+ * Copyright remains Eric Young's, and as such any Copyright notices in
+ * the code are not to be removed.
+ * If this package is used in a product, Eric Young should be given attribution
+ * as the author of the parts of the library used.
+ * This can be in the form of a textual message at program startup or
+ * in documentation (online or textual) provided with the package.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the copyright
+ * notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in the
+ * documentation and/or other materials provided with the distribution.
+ * 3. All advertising materials mentioning features or use of this software
+ * must display the following acknowledgement:
+ * "This product includes cryptographic software written by
+ * Eric Young (eay@cryptsoft.com)"
+ * The word 'cryptographic' can be left out if the rouines from the library
+ * being used are not cryptographic related :-).
+ * 4. If you include any Windows specific code (or a derivative thereof) from
+ * the apps directory (application code) you must include an acknowledgement:
+ * "This product includes software written by Tim Hudson (tjh@cryptsoft.com)"
+ *
+ * THIS SOFTWARE IS PROVIDED BY ERIC YOUNG ``AS IS'' AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ *
+ * The licence and distribution terms for any publically available version or
+ * derivative of this code cannot be changed. i.e. this code cannot simply be
+ * copied and put under another distribution licence
+ * [including the GNU Public Licence.]
+ */
+
+
+ISC license used for completely new code in BoringSSL:
+
+/* Copyright 2015 The BoringSSL Authors
+ *
+ * Permission to use, copy, modify, and/or distribute this software for any
+ * purpose with or without fee is hereby granted, provided that the above
+ * copyright notice and this permission notice appear in all copies.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+ * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
+ * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY
+ * SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+ * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION
+ * OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN
+ * CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. */
+
+
+The code in third_party/fiat carries the MIT license:
+
+Copyright (c) 2015-2016 the fiat-crypto authors (see
+https://github.com/mit-plv/fiat-crypto/blob/master/AUTHORS).
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.
+
+
+Licenses for support code
+-------------------------
+
+Parts of the TLS test suite are under the Go license. This code is not included
+in BoringSSL (i.e. libcrypto and libssl) when compiled, however, so
+distributing code linked against BoringSSL does not trigger this license:
+
+Copyright (c) 2009 The Go Authors. All rights reserved.
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+
+ * Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+ * Redistributions in binary form must reproduce the above
+copyright notice, this list of conditions and the following disclaimer
+in the documentation and/or other materials provided with the
+distribution.
+ * Neither the name of Google Inc. nor the names of its
+contributors may be used to endorse or promote products derived from
+this software without specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+
+BoringSSL uses the Chromium test infrastructure to run a continuous build,
+trybots etc. The scripts which manage this, and the script for generating build
+metadata, are under the Chromium license. Distributing code linked against
+BoringSSL does not trigger this license.
+
+Copyright 2015 The Chromium Authors. All rights reserved.
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+
+ * Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+ * Redistributions in binary form must reproduce the above
+copyright notice, this list of conditions and the following disclaimer
+in the documentation and/or other materials provided with the
+distribution.
+ * Neither the name of Google Inc. nor the names of its
+contributors may be used to endorse or promote products derived from
+this software without specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
diff --git a/sys/contrib/openzfs/contrib/icp/gcm-simd/boringssl/README b/sys/contrib/openzfs/contrib/icp/gcm-simd/boringssl/README
new file mode 100644
index 000000000000..aa6fb6d477fa
--- /dev/null
+++ b/sys/contrib/openzfs/contrib/icp/gcm-simd/boringssl/README
@@ -0,0 +1,11 @@
+This directory contains the original BoringSSL [1] GCM x86-64 assembly
+files [2].
+
+The assembler files where then further modified to fit the ICP conventions.
+
+The main purpose to include these files (and the original ones) here, is to
+serve as a reference if upstream changes need to be applied to the files
+included and modified in the ICP.
+
+[1] https://github.com/google/boringssl
+[2] https://github.com/google/boringssl/blob/d5440dd2c2c500ac2d3bba4afec47a054b4d99ae/gen/bcm/aes-gcm-avx2-x86_64-linux.S
diff --git a/sys/contrib/openzfs/contrib/icp/gcm-simd/boringssl/aes-gcm-avx2-x86_64-linux.S b/sys/contrib/openzfs/contrib/icp/gcm-simd/boringssl/aes-gcm-avx2-x86_64-linux.S
new file mode 100644
index 000000000000..e7327c9de872
--- /dev/null
+++ b/sys/contrib/openzfs/contrib/icp/gcm-simd/boringssl/aes-gcm-avx2-x86_64-linux.S
@@ -0,0 +1,1328 @@
+// SPDX-License-Identifier: Apache-2.0
+// This file is generated from a similarly-named Perl script in the BoringSSL
+// source tree. Do not edit by hand.
+
+#include <openssl/asm_base.h>
+
+#if !defined(OPENSSL_NO_ASM) && defined(OPENSSL_X86_64) && defined(__ELF__)
+.section .rodata
+.align 16
+
+
+.Lbswap_mask:
+.quad 0x08090a0b0c0d0e0f, 0x0001020304050607
+
+
+
+
+
+
+
+
+.Lgfpoly:
+.quad 1, 0xc200000000000000
+
+
+.Lgfpoly_and_internal_carrybit:
+.quad 1, 0xc200000000000001
+
+.align 32
+
+.Lctr_pattern:
+.quad 0, 0
+.quad 1, 0
+.Linc_2blocks:
+.quad 2, 0
+.quad 2, 0
+
+.text
+.globl gcm_init_vpclmulqdq_avx2
+.hidden gcm_init_vpclmulqdq_avx2
+.type gcm_init_vpclmulqdq_avx2,@function
+.align 32
+gcm_init_vpclmulqdq_avx2:
+.cfi_startproc
+
+_CET_ENDBR
+
+
+
+
+
+ vpshufd $0x4e,(%rsi),%xmm3
+
+
+
+
+
+ vpshufd $0xd3,%xmm3,%xmm0
+ vpsrad $31,%xmm0,%xmm0
+ vpaddq %xmm3,%xmm3,%xmm3
+ vpand .Lgfpoly_and_internal_carrybit(%rip),%xmm0,%xmm0
+ vpxor %xmm0,%xmm3,%xmm3
+
+ vbroadcasti128 .Lgfpoly(%rip),%ymm6
+
+
+ vpclmulqdq $0x00,%xmm3,%xmm3,%xmm0
+ vpclmulqdq $0x01,%xmm3,%xmm3,%xmm1
+ vpclmulqdq $0x10,%xmm3,%xmm3,%xmm2
+ vpxor %xmm2,%xmm1,%xmm1
+ vpclmulqdq $0x01,%xmm0,%xmm6,%xmm2
+ vpshufd $0x4e,%xmm0,%xmm0
+ vpxor %xmm0,%xmm1,%xmm1
+ vpxor %xmm2,%xmm1,%xmm1
+ vpclmulqdq $0x11,%xmm3,%xmm3,%xmm5
+ vpclmulqdq $0x01,%xmm1,%xmm6,%xmm0
+ vpshufd $0x4e,%xmm1,%xmm1
+ vpxor %xmm1,%xmm5,%xmm5
+ vpxor %xmm0,%xmm5,%xmm5
+
+
+
+ vinserti128 $1,%xmm3,%ymm5,%ymm3
+ vinserti128 $1,%xmm5,%ymm5,%ymm5
+
+
+ vpclmulqdq $0x00,%ymm5,%ymm3,%ymm0
+ vpclmulqdq $0x01,%ymm5,%ymm3,%ymm1
+ vpclmulqdq $0x10,%ymm5,%ymm3,%ymm2
+ vpxor %ymm2,%ymm1,%ymm1
+ vpclmulqdq $0x01,%ymm0,%ymm6,%ymm2
+ vpshufd $0x4e,%ymm0,%ymm0
+ vpxor %ymm0,%ymm1,%ymm1
+ vpxor %ymm2,%ymm1,%ymm1
+ vpclmulqdq $0x11,%ymm5,%ymm3,%ymm4
+ vpclmulqdq $0x01,%ymm1,%ymm6,%ymm0
+ vpshufd $0x4e,%ymm1,%ymm1
+ vpxor %ymm1,%ymm4,%ymm4
+ vpxor %ymm0,%ymm4,%ymm4
+
+
+
+ vmovdqu %ymm3,96(%rdi)
+ vmovdqu %ymm4,64(%rdi)
+
+
+
+ vpunpcklqdq %ymm3,%ymm4,%ymm0
+ vpunpckhqdq %ymm3,%ymm4,%ymm1
+ vpxor %ymm1,%ymm0,%ymm0
+ vmovdqu %ymm0,128+32(%rdi)
+
+
+ vpclmulqdq $0x00,%ymm5,%ymm4,%ymm0
+ vpclmulqdq $0x01,%ymm5,%ymm4,%ymm1
+ vpclmulqdq $0x10,%ymm5,%ymm4,%ymm2
+ vpxor %ymm2,%ymm1,%ymm1
+ vpclmulqdq $0x01,%ymm0,%ymm6,%ymm2
+ vpshufd $0x4e,%ymm0,%ymm0
+ vpxor %ymm0,%ymm1,%ymm1
+ vpxor %ymm2,%ymm1,%ymm1
+ vpclmulqdq $0x11,%ymm5,%ymm4,%ymm3
+ vpclmulqdq $0x01,%ymm1,%ymm6,%ymm0
+ vpshufd $0x4e,%ymm1,%ymm1
+ vpxor %ymm1,%ymm3,%ymm3
+ vpxor %ymm0,%ymm3,%ymm3
+
+ vpclmulqdq $0x00,%ymm5,%ymm3,%ymm0
+ vpclmulqdq $0x01,%ymm5,%ymm3,%ymm1
+ vpclmulqdq $0x10,%ymm5,%ymm3,%ymm2
+ vpxor %ymm2,%ymm1,%ymm1
+ vpclmulqdq $0x01,%ymm0,%ymm6,%ymm2
+ vpshufd $0x4e,%ymm0,%ymm0
+ vpxor %ymm0,%ymm1,%ymm1
+ vpxor %ymm2,%ymm1,%ymm1
+ vpclmulqdq $0x11,%ymm5,%ymm3,%ymm4
+ vpclmulqdq $0x01,%ymm1,%ymm6,%ymm0
+ vpshufd $0x4e,%ymm1,%ymm1
+ vpxor %ymm1,%ymm4,%ymm4
+ vpxor %ymm0,%ymm4,%ymm4
+
+ vmovdqu %ymm3,32(%rdi)
+ vmovdqu %ymm4,0(%rdi)
+
+
+
+ vpunpcklqdq %ymm3,%ymm4,%ymm0
+ vpunpckhqdq %ymm3,%ymm4,%ymm1
+ vpxor %ymm1,%ymm0,%ymm0
+ vmovdqu %ymm0,128(%rdi)
+
+ vzeroupper
+ ret
+
+.cfi_endproc
+.size gcm_init_vpclmulqdq_avx2, . - gcm_init_vpclmulqdq_avx2
+.globl gcm_gmult_vpclmulqdq_avx2
+.hidden gcm_gmult_vpclmulqdq_avx2
+.type gcm_gmult_vpclmulqdq_avx2,@function
+.align 32
+gcm_gmult_vpclmulqdq_avx2:
+.cfi_startproc
+
+_CET_ENDBR
+
+
+
+ vmovdqu (%rdi),%xmm0
+ vmovdqu .Lbswap_mask(%rip),%xmm1
+ vmovdqu 128-16(%rsi),%xmm2
+ vmovdqu .Lgfpoly(%rip),%xmm3
+ vpshufb %xmm1,%xmm0,%xmm0
+
+ vpclmulqdq $0x00,%xmm2,%xmm0,%xmm4
+ vpclmulqdq $0x01,%xmm2,%xmm0,%xmm5
+ vpclmulqdq $0x10,%xmm2,%xmm0,%xmm6
+ vpxor %xmm6,%xmm5,%xmm5
+ vpclmulqdq $0x01,%xmm4,%xmm3,%xmm6
+ vpshufd $0x4e,%xmm4,%xmm4
+ vpxor %xmm4,%xmm5,%xmm5
+ vpxor %xmm6,%xmm5,%xmm5
+ vpclmulqdq $0x11,%xmm2,%xmm0,%xmm0
+ vpclmulqdq $0x01,%xmm5,%xmm3,%xmm4
+ vpshufd $0x4e,%xmm5,%xmm5
+ vpxor %xmm5,%xmm0,%xmm0
+ vpxor %xmm4,%xmm0,%xmm0
+
+
+ vpshufb %xmm1,%xmm0,%xmm0
+ vmovdqu %xmm0,(%rdi)
+ ret
+
+.cfi_endproc
+.size gcm_gmult_vpclmulqdq_avx2, . - gcm_gmult_vpclmulqdq_avx2
+.globl gcm_ghash_vpclmulqdq_avx2
+.hidden gcm_ghash_vpclmulqdq_avx2
+.type gcm_ghash_vpclmulqdq_avx2,@function
+.align 32
+gcm_ghash_vpclmulqdq_avx2:
+.cfi_startproc
+
+_CET_ENDBR
+
+
+
+
+
+
+ vmovdqu .Lbswap_mask(%rip),%xmm6
+ vmovdqu .Lgfpoly(%rip),%xmm7
+
+
+ vmovdqu (%rdi),%xmm5
+ vpshufb %xmm6,%xmm5,%xmm5
+
+
+ cmpq $32,%rcx
+ jb .Lghash_lastblock
+
+
+
+ vinserti128 $1,%xmm6,%ymm6,%ymm6
+ vinserti128 $1,%xmm7,%ymm7,%ymm7
+
+ cmpq $127,%rcx
+ jbe .Lghash_loop_1x
+
+
+ vmovdqu 128(%rsi),%ymm8
+ vmovdqu 128+32(%rsi),%ymm9
+.Lghash_loop_4x:
+
+ vmovdqu 0(%rdx),%ymm1
+ vpshufb %ymm6,%ymm1,%ymm1
+ vmovdqu 0(%rsi),%ymm2
+ vpxor %ymm5,%ymm1,%ymm1
+ vpclmulqdq $0x00,%ymm2,%ymm1,%ymm3
+ vpclmulqdq $0x11,%ymm2,%ymm1,%ymm5
+ vpunpckhqdq %ymm1,%ymm1,%ymm0
+ vpxor %ymm1,%ymm0,%ymm0
+ vpclmulqdq $0x00,%ymm8,%ymm0,%ymm4
+
+ vmovdqu 32(%rdx),%ymm1
+ vpshufb %ymm6,%ymm1,%ymm1
+ vmovdqu 32(%rsi),%ymm2
+ vpclmulqdq $0x00,%ymm2,%ymm1,%ymm0
+ vpxor %ymm0,%ymm3,%ymm3
+ vpclmulqdq $0x11,%ymm2,%ymm1,%ymm0
+ vpxor %ymm0,%ymm5,%ymm5
+ vpunpckhqdq %ymm1,%ymm1,%ymm0
+ vpxor %ymm1,%ymm0,%ymm0
+ vpclmulqdq $0x10,%ymm8,%ymm0,%ymm0
+ vpxor %ymm0,%ymm4,%ymm4
+
+ vmovdqu 64(%rdx),%ymm1
+ vpshufb %ymm6,%ymm1,%ymm1
+ vmovdqu 64(%rsi),%ymm2
+ vpclmulqdq $0x00,%ymm2,%ymm1,%ymm0
+ vpxor %ymm0,%ymm3,%ymm3
+ vpclmulqdq $0x11,%ymm2,%ymm1,%ymm0
+ vpxor %ymm0,%ymm5,%ymm5
+ vpunpckhqdq %ymm1,%ymm1,%ymm0
+ vpxor %ymm1,%ymm0,%ymm0
+ vpclmulqdq $0x00,%ymm9,%ymm0,%ymm0
+ vpxor %ymm0,%ymm4,%ymm4
+
+
+ vmovdqu 96(%rdx),%ymm1
+ vpshufb %ymm6,%ymm1,%ymm1
+ vmovdqu 96(%rsi),%ymm2
+ vpclmulqdq $0x00,%ymm2,%ymm1,%ymm0
+ vpxor %ymm0,%ymm3,%ymm3
+ vpclmulqdq $0x11,%ymm2,%ymm1,%ymm0
+ vpxor %ymm0,%ymm5,%ymm5
+ vpunpckhqdq %ymm1,%ymm1,%ymm0
+ vpxor %ymm1,%ymm0,%ymm0
+ vpclmulqdq $0x10,%ymm9,%ymm0,%ymm0
+ vpxor %ymm0,%ymm4,%ymm4
+
+ vpxor %ymm3,%ymm4,%ymm4
+ vpxor %ymm5,%ymm4,%ymm4
+
+
+ vbroadcasti128 .Lgfpoly(%rip),%ymm2
+ vpclmulqdq $0x01,%ymm3,%ymm2,%ymm0
+ vpshufd $0x4e,%ymm3,%ymm3
+ vpxor %ymm3,%ymm4,%ymm4
+ vpxor %ymm0,%ymm4,%ymm4
+
+ vpclmulqdq $0x01,%ymm4,%ymm2,%ymm0
+ vpshufd $0x4e,%ymm4,%ymm4
+ vpxor %ymm4,%ymm5,%ymm5
+ vpxor %ymm0,%ymm5,%ymm5
+ vextracti128 $1,%ymm5,%xmm0
+ vpxor %xmm0,%xmm5,%xmm5
+
+ subq $-128,%rdx
+ addq $-128,%rcx
+ cmpq $127,%rcx
+ ja .Lghash_loop_4x
+
+
+ cmpq $32,%rcx
+ jb .Lghash_loop_1x_done
+.Lghash_loop_1x:
+ vmovdqu (%rdx),%ymm0
+ vpshufb %ymm6,%ymm0,%ymm0
+ vpxor %ymm0,%ymm5,%ymm5
+ vmovdqu 128-32(%rsi),%ymm0
+ vpclmulqdq $0x00,%ymm0,%ymm5,%ymm1
+ vpclmulqdq $0x01,%ymm0,%ymm5,%ymm2
+ vpclmulqdq $0x10,%ymm0,%ymm5,%ymm3
+ vpxor %ymm3,%ymm2,%ymm2
+ vpclmulqdq $0x01,%ymm1,%ymm7,%ymm3
+ vpshufd $0x4e,%ymm1,%ymm1
+ vpxor %ymm1,%ymm2,%ymm2
+ vpxor %ymm3,%ymm2,%ymm2
+ vpclmulqdq $0x11,%ymm0,%ymm5,%ymm5
+ vpclmulqdq $0x01,%ymm2,%ymm7,%ymm1
+ vpshufd $0x4e,%ymm2,%ymm2
+ vpxor %ymm2,%ymm5,%ymm5
+ vpxor %ymm1,%ymm5,%ymm5
+
+ vextracti128 $1,%ymm5,%xmm0
+ vpxor %xmm0,%xmm5,%xmm5
+ addq $32,%rdx
+ subq $32,%rcx
+ cmpq $32,%rcx
+ jae .Lghash_loop_1x
+.Lghash_loop_1x_done:
+
+
+.Lghash_lastblock:
+ testq %rcx,%rcx
+ jz .Lghash_done
+ vmovdqu (%rdx),%xmm0
+ vpshufb %xmm6,%xmm0,%xmm0
+ vpxor %xmm0,%xmm5,%xmm5
+ vmovdqu 128-16(%rsi),%xmm0
+ vpclmulqdq $0x00,%xmm0,%xmm5,%xmm1
+ vpclmulqdq $0x01,%xmm0,%xmm5,%xmm2
+ vpclmulqdq $0x10,%xmm0,%xmm5,%xmm3
+ vpxor %xmm3,%xmm2,%xmm2
+ vpclmulqdq $0x01,%xmm1,%xmm7,%xmm3
+ vpshufd $0x4e,%xmm1,%xmm1
+ vpxor %xmm1,%xmm2,%xmm2
+ vpxor %xmm3,%xmm2,%xmm2
+ vpclmulqdq $0x11,%xmm0,%xmm5,%xmm5
+ vpclmulqdq $0x01,%xmm2,%xmm7,%xmm1
+ vpshufd $0x4e,%xmm2,%xmm2
+ vpxor %xmm2,%xmm5,%xmm5
+ vpxor %xmm1,%xmm5,%xmm5
+
+
+.Lghash_done:
+
+ vpshufb %xmm6,%xmm5,%xmm5
+ vmovdqu %xmm5,(%rdi)
+
+ vzeroupper
+ ret
+
+.cfi_endproc
+.size gcm_ghash_vpclmulqdq_avx2, . - gcm_ghash_vpclmulqdq_avx2
+.globl aes_gcm_enc_update_vaes_avx2
+.hidden aes_gcm_enc_update_vaes_avx2
+.type aes_gcm_enc_update_vaes_avx2,@function
+.align 32
+aes_gcm_enc_update_vaes_avx2:
+.cfi_startproc
+
+_CET_ENDBR
+ pushq %r12
+.cfi_adjust_cfa_offset 8
+.cfi_offset %r12,-16
+
+ movq 16(%rsp),%r12
+#ifdef BORINGSSL_DISPATCH_TEST
+.extern BORINGSSL_function_hit
+.hidden BORINGSSL_function_hit
+ movb $1,BORINGSSL_function_hit+8(%rip)
+#endif
+ vbroadcasti128 .Lbswap_mask(%rip),%ymm0
+
+
+
+ vmovdqu (%r12),%xmm1
+ vpshufb %xmm0,%xmm1,%xmm1
+ vbroadcasti128 (%r8),%ymm11
+ vpshufb %ymm0,%ymm11,%ymm11
+
+
+
+ movl 240(%rcx),%r10d
+ leal -20(,%r10,4),%r10d
+
+
+
+
+ leaq 96(%rcx,%r10,4),%r11
+ vbroadcasti128 (%rcx),%ymm9
+ vbroadcasti128 (%r11),%ymm10
+
+
+ vpaddd .Lctr_pattern(%rip),%ymm11,%ymm11
+
+
+
+ cmpq $127,%rdx
+ jbe .Lcrypt_loop_4x_done__func1
+
+ vmovdqu 128(%r9),%ymm7
+ vmovdqu 128+32(%r9),%ymm8
+
+
+
+ vmovdqu .Linc_2blocks(%rip),%ymm2
+ vpshufb %ymm0,%ymm11,%ymm12
+ vpaddd %ymm2,%ymm11,%ymm11
+ vpshufb %ymm0,%ymm11,%ymm13
+ vpaddd %ymm2,%ymm11,%ymm11
+ vpshufb %ymm0,%ymm11,%ymm14
+ vpaddd %ymm2,%ymm11,%ymm11
+ vpshufb %ymm0,%ymm11,%ymm15
+ vpaddd %ymm2,%ymm11,%ymm11
+
+
+ vpxor %ymm9,%ymm12,%ymm12
+ vpxor %ymm9,%ymm13,%ymm13
+ vpxor %ymm9,%ymm14,%ymm14
+ vpxor %ymm9,%ymm15,%ymm15
+
+ leaq 16(%rcx),%rax
+.Lvaesenc_loop_first_4_vecs__func1:
+ vbroadcasti128 (%rax),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+ addq $16,%rax
+ cmpq %rax,%r11
+ jne .Lvaesenc_loop_first_4_vecs__func1
+ vpxor 0(%rdi),%ymm10,%ymm2
+ vpxor 32(%rdi),%ymm10,%ymm3
+ vpxor 64(%rdi),%ymm10,%ymm5
+ vpxor 96(%rdi),%ymm10,%ymm6
+ vaesenclast %ymm2,%ymm12,%ymm12
+ vaesenclast %ymm3,%ymm13,%ymm13
+ vaesenclast %ymm5,%ymm14,%ymm14
+ vaesenclast %ymm6,%ymm15,%ymm15
+ vmovdqu %ymm12,0(%rsi)
+ vmovdqu %ymm13,32(%rsi)
+ vmovdqu %ymm14,64(%rsi)
+ vmovdqu %ymm15,96(%rsi)
+
+ subq $-128,%rdi
+ addq $-128,%rdx
+ cmpq $127,%rdx
+ jbe .Lghash_last_ciphertext_4x__func1
+.align 16
+.Lcrypt_loop_4x__func1:
+
+
+
+
+ vmovdqu .Linc_2blocks(%rip),%ymm2
+ vpshufb %ymm0,%ymm11,%ymm12
+ vpaddd %ymm2,%ymm11,%ymm11
+ vpshufb %ymm0,%ymm11,%ymm13
+ vpaddd %ymm2,%ymm11,%ymm11
+ vpshufb %ymm0,%ymm11,%ymm14
+ vpaddd %ymm2,%ymm11,%ymm11
+ vpshufb %ymm0,%ymm11,%ymm15
+ vpaddd %ymm2,%ymm11,%ymm11
+
+
+ vpxor %ymm9,%ymm12,%ymm12
+ vpxor %ymm9,%ymm13,%ymm13
+ vpxor %ymm9,%ymm14,%ymm14
+ vpxor %ymm9,%ymm15,%ymm15
+
+ cmpl $24,%r10d
+ jl .Laes128__func1
+ je .Laes192__func1
+
+ vbroadcasti128 -208(%r11),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+ vbroadcasti128 -192(%r11),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+.Laes192__func1:
+ vbroadcasti128 -176(%r11),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+ vbroadcasti128 -160(%r11),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+.Laes128__func1:
+ prefetcht0 512(%rdi)
+ prefetcht0 512+64(%rdi)
+
+ vmovdqu 0(%rsi),%ymm3
+ vpshufb %ymm0,%ymm3,%ymm3
+ vmovdqu 0(%r9),%ymm4
+ vpxor %ymm1,%ymm3,%ymm3
+ vpclmulqdq $0x00,%ymm4,%ymm3,%ymm5
+ vpclmulqdq $0x11,%ymm4,%ymm3,%ymm1
+ vpunpckhqdq %ymm3,%ymm3,%ymm2
+ vpxor %ymm3,%ymm2,%ymm2
+ vpclmulqdq $0x00,%ymm7,%ymm2,%ymm6
+
+ vbroadcasti128 -144(%r11),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+
+ vbroadcasti128 -128(%r11),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+
+ vmovdqu 32(%rsi),%ymm3
+ vpshufb %ymm0,%ymm3,%ymm3
+ vmovdqu 32(%r9),%ymm4
+ vpclmulqdq $0x00,%ymm4,%ymm3,%ymm2
+ vpxor %ymm2,%ymm5,%ymm5
+ vpclmulqdq $0x11,%ymm4,%ymm3,%ymm2
+ vpxor %ymm2,%ymm1,%ymm1
+ vpunpckhqdq %ymm3,%ymm3,%ymm2
+ vpxor %ymm3,%ymm2,%ymm2
+ vpclmulqdq $0x10,%ymm7,%ymm2,%ymm2
+ vpxor %ymm2,%ymm6,%ymm6
+
+ vbroadcasti128 -112(%r11),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+
+ vmovdqu 64(%rsi),%ymm3
+ vpshufb %ymm0,%ymm3,%ymm3
+ vmovdqu 64(%r9),%ymm4
+
+ vbroadcasti128 -96(%r11),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+ vpclmulqdq $0x00,%ymm4,%ymm3,%ymm2
+ vpxor %ymm2,%ymm5,%ymm5
+ vpclmulqdq $0x11,%ymm4,%ymm3,%ymm2
+ vpxor %ymm2,%ymm1,%ymm1
+
+ vbroadcasti128 -80(%r11),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+ vpunpckhqdq %ymm3,%ymm3,%ymm2
+ vpxor %ymm3,%ymm2,%ymm2
+ vpclmulqdq $0x00,%ymm8,%ymm2,%ymm2
+ vpxor %ymm2,%ymm6,%ymm6
+
+
+ vmovdqu 96(%rsi),%ymm3
+ vpshufb %ymm0,%ymm3,%ymm3
+
+ vbroadcasti128 -64(%r11),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+ vmovdqu 96(%r9),%ymm4
+ vpclmulqdq $0x00,%ymm4,%ymm3,%ymm2
+ vpxor %ymm2,%ymm5,%ymm5
+ vpclmulqdq $0x11,%ymm4,%ymm3,%ymm2
+ vpxor %ymm2,%ymm1,%ymm1
+ vpunpckhqdq %ymm3,%ymm3,%ymm2
+ vpxor %ymm3,%ymm2,%ymm2
+ vpclmulqdq $0x10,%ymm8,%ymm2,%ymm2
+ vpxor %ymm2,%ymm6,%ymm6
+
+ vbroadcasti128 -48(%r11),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+
+ vpxor %ymm5,%ymm6,%ymm6
+ vpxor %ymm1,%ymm6,%ymm6
+
+
+ vbroadcasti128 .Lgfpoly(%rip),%ymm4
+ vpclmulqdq $0x01,%ymm5,%ymm4,%ymm2
+ vpshufd $0x4e,%ymm5,%ymm5
+ vpxor %ymm5,%ymm6,%ymm6
+ vpxor %ymm2,%ymm6,%ymm6
+
+ vbroadcasti128 -32(%r11),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+
+ vpclmulqdq $0x01,%ymm6,%ymm4,%ymm2
+ vpshufd $0x4e,%ymm6,%ymm6
+ vpxor %ymm6,%ymm1,%ymm1
+ vpxor %ymm2,%ymm1,%ymm1
+
+ vbroadcasti128 -16(%r11),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+ vextracti128 $1,%ymm1,%xmm2
+ vpxor %xmm2,%xmm1,%xmm1
+
+
+ subq $-128,%rsi
+ vpxor 0(%rdi),%ymm10,%ymm2
+ vpxor 32(%rdi),%ymm10,%ymm3
+ vpxor 64(%rdi),%ymm10,%ymm5
+ vpxor 96(%rdi),%ymm10,%ymm6
+ vaesenclast %ymm2,%ymm12,%ymm12
+ vaesenclast %ymm3,%ymm13,%ymm13
+ vaesenclast %ymm5,%ymm14,%ymm14
+ vaesenclast %ymm6,%ymm15,%ymm15
+ vmovdqu %ymm12,0(%rsi)
+ vmovdqu %ymm13,32(%rsi)
+ vmovdqu %ymm14,64(%rsi)
+ vmovdqu %ymm15,96(%rsi)
+
+ subq $-128,%rdi
+
+ addq $-128,%rdx
+ cmpq $127,%rdx
+ ja .Lcrypt_loop_4x__func1
+.Lghash_last_ciphertext_4x__func1:
+
+ vmovdqu 0(%rsi),%ymm3
+ vpshufb %ymm0,%ymm3,%ymm3
+ vmovdqu 0(%r9),%ymm4
+ vpxor %ymm1,%ymm3,%ymm3
+ vpclmulqdq $0x00,%ymm4,%ymm3,%ymm5
+ vpclmulqdq $0x11,%ymm4,%ymm3,%ymm1
+ vpunpckhqdq %ymm3,%ymm3,%ymm2
+ vpxor %ymm3,%ymm2,%ymm2
+ vpclmulqdq $0x00,%ymm7,%ymm2,%ymm6
+
+ vmovdqu 32(%rsi),%ymm3
+ vpshufb %ymm0,%ymm3,%ymm3
+ vmovdqu 32(%r9),%ymm4
+ vpclmulqdq $0x00,%ymm4,%ymm3,%ymm2
+ vpxor %ymm2,%ymm5,%ymm5
+ vpclmulqdq $0x11,%ymm4,%ymm3,%ymm2
+ vpxor %ymm2,%ymm1,%ymm1
+ vpunpckhqdq %ymm3,%ymm3,%ymm2
+ vpxor %ymm3,%ymm2,%ymm2
+ vpclmulqdq $0x10,%ymm7,%ymm2,%ymm2
+ vpxor %ymm2,%ymm6,%ymm6
+
+ vmovdqu 64(%rsi),%ymm3
+ vpshufb %ymm0,%ymm3,%ymm3
+ vmovdqu 64(%r9),%ymm4
+ vpclmulqdq $0x00,%ymm4,%ymm3,%ymm2
+ vpxor %ymm2,%ymm5,%ymm5
+ vpclmulqdq $0x11,%ymm4,%ymm3,%ymm2
+ vpxor %ymm2,%ymm1,%ymm1
+ vpunpckhqdq %ymm3,%ymm3,%ymm2
+ vpxor %ymm3,%ymm2,%ymm2
+ vpclmulqdq $0x00,%ymm8,%ymm2,%ymm2
+ vpxor %ymm2,%ymm6,%ymm6
+
+
+ vmovdqu 96(%rsi),%ymm3
+ vpshufb %ymm0,%ymm3,%ymm3
+ vmovdqu 96(%r9),%ymm4
+ vpclmulqdq $0x00,%ymm4,%ymm3,%ymm2
+ vpxor %ymm2,%ymm5,%ymm5
+ vpclmulqdq $0x11,%ymm4,%ymm3,%ymm2
+ vpxor %ymm2,%ymm1,%ymm1
+ vpunpckhqdq %ymm3,%ymm3,%ymm2
+ vpxor %ymm3,%ymm2,%ymm2
+ vpclmulqdq $0x10,%ymm8,%ymm2,%ymm2
+ vpxor %ymm2,%ymm6,%ymm6
+
+ vpxor %ymm5,%ymm6,%ymm6
+ vpxor %ymm1,%ymm6,%ymm6
+
+
+ vbroadcasti128 .Lgfpoly(%rip),%ymm4
+ vpclmulqdq $0x01,%ymm5,%ymm4,%ymm2
+ vpshufd $0x4e,%ymm5,%ymm5
+ vpxor %ymm5,%ymm6,%ymm6
+ vpxor %ymm2,%ymm6,%ymm6
+
+ vpclmulqdq $0x01,%ymm6,%ymm4,%ymm2
+ vpshufd $0x4e,%ymm6,%ymm6
+ vpxor %ymm6,%ymm1,%ymm1
+ vpxor %ymm2,%ymm1,%ymm1
+ vextracti128 $1,%ymm1,%xmm2
+ vpxor %xmm2,%xmm1,%xmm1
+
+ subq $-128,%rsi
+.Lcrypt_loop_4x_done__func1:
+
+ testq %rdx,%rdx
+ jz .Ldone__func1
+
+
+
+
+
+ leaq 128(%r9),%r8
+ subq %rdx,%r8
+
+
+ vpxor %xmm5,%xmm5,%xmm5
+ vpxor %xmm6,%xmm6,%xmm6
+ vpxor %xmm7,%xmm7,%xmm7
+
+ cmpq $64,%rdx
+ jb .Llessthan64bytes__func1
+
+
+ vpshufb %ymm0,%ymm11,%ymm12
+ vpaddd .Linc_2blocks(%rip),%ymm11,%ymm11
+ vpshufb %ymm0,%ymm11,%ymm13
+ vpaddd .Linc_2blocks(%rip),%ymm11,%ymm11
+ vpxor %ymm9,%ymm12,%ymm12
+ vpxor %ymm9,%ymm13,%ymm13
+ leaq 16(%rcx),%rax
+.Lvaesenc_loop_tail_1__func1:
+ vbroadcasti128 (%rax),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ addq $16,%rax
+ cmpq %rax,%r11
+ jne .Lvaesenc_loop_tail_1__func1
+ vaesenclast %ymm10,%ymm12,%ymm12
+ vaesenclast %ymm10,%ymm13,%ymm13
+
+
+ vmovdqu 0(%rdi),%ymm2
+ vmovdqu 32(%rdi),%ymm3
+ vpxor %ymm2,%ymm12,%ymm12
+ vpxor %ymm3,%ymm13,%ymm13
+ vmovdqu %ymm12,0(%rsi)
+ vmovdqu %ymm13,32(%rsi)
+
+
+ vpshufb %ymm0,%ymm12,%ymm12
+ vpshufb %ymm0,%ymm13,%ymm13
+ vpxor %ymm1,%ymm12,%ymm12
+ vmovdqu (%r8),%ymm2
+ vmovdqu 32(%r8),%ymm3
+ vpclmulqdq $0x00,%ymm2,%ymm12,%ymm5
+ vpclmulqdq $0x01,%ymm2,%ymm12,%ymm6
+ vpclmulqdq $0x10,%ymm2,%ymm12,%ymm4
+ vpxor %ymm4,%ymm6,%ymm6
+ vpclmulqdq $0x11,%ymm2,%ymm12,%ymm7
+ vpclmulqdq $0x00,%ymm3,%ymm13,%ymm4
+ vpxor %ymm4,%ymm5,%ymm5
+ vpclmulqdq $0x01,%ymm3,%ymm13,%ymm4
+ vpxor %ymm4,%ymm6,%ymm6
+ vpclmulqdq $0x10,%ymm3,%ymm13,%ymm4
+ vpxor %ymm4,%ymm6,%ymm6
+ vpclmulqdq $0x11,%ymm3,%ymm13,%ymm4
+ vpxor %ymm4,%ymm7,%ymm7
+
+ addq $64,%r8
+ addq $64,%rdi
+ addq $64,%rsi
+ subq $64,%rdx
+ jz .Lreduce__func1
+
+ vpxor %xmm1,%xmm1,%xmm1
+
+
+.Llessthan64bytes__func1:
+ vpshufb %ymm0,%ymm11,%ymm12
+ vpaddd .Linc_2blocks(%rip),%ymm11,%ymm11
+ vpshufb %ymm0,%ymm11,%ymm13
+ vpxor %ymm9,%ymm12,%ymm12
+ vpxor %ymm9,%ymm13,%ymm13
+ leaq 16(%rcx),%rax
+.Lvaesenc_loop_tail_2__func1:
+ vbroadcasti128 (%rax),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ addq $16,%rax
+ cmpq %rax,%r11
+ jne .Lvaesenc_loop_tail_2__func1
+ vaesenclast %ymm10,%ymm12,%ymm12
+ vaesenclast %ymm10,%ymm13,%ymm13
+
+
+
+
+ cmpq $32,%rdx
+ jb .Lxor_one_block__func1
+ je .Lxor_two_blocks__func1
+
+.Lxor_three_blocks__func1:
+ vmovdqu 0(%rdi),%ymm2
+ vmovdqu 32(%rdi),%xmm3
+ vpxor %ymm2,%ymm12,%ymm12
+ vpxor %xmm3,%xmm13,%xmm13
+ vmovdqu %ymm12,0(%rsi)
+ vmovdqu %xmm13,32(%rsi)
+
+ vpshufb %ymm0,%ymm12,%ymm12
+ vpshufb %xmm0,%xmm13,%xmm13
+ vpxor %ymm1,%ymm12,%ymm12
+ vmovdqu (%r8),%ymm2
+ vmovdqu 32(%r8),%xmm3
+ vpclmulqdq $0x00,%xmm3,%xmm13,%xmm4
+ vpxor %ymm4,%ymm5,%ymm5
+ vpclmulqdq $0x01,%xmm3,%xmm13,%xmm4
+ vpxor %ymm4,%ymm6,%ymm6
+ vpclmulqdq $0x10,%xmm3,%xmm13,%xmm4
+ vpxor %ymm4,%ymm6,%ymm6
+ vpclmulqdq $0x11,%xmm3,%xmm13,%xmm4
+ vpxor %ymm4,%ymm7,%ymm7
+ jmp .Lghash_mul_one_vec_unreduced__func1
+
+.Lxor_two_blocks__func1:
+ vmovdqu (%rdi),%ymm2
+ vpxor %ymm2,%ymm12,%ymm12
+ vmovdqu %ymm12,(%rsi)
+ vpshufb %ymm0,%ymm12,%ymm12
+ vpxor %ymm1,%ymm12,%ymm12
+ vmovdqu (%r8),%ymm2
+ jmp .Lghash_mul_one_vec_unreduced__func1
+
+.Lxor_one_block__func1:
+ vmovdqu (%rdi),%xmm2
+ vpxor %xmm2,%xmm12,%xmm12
+ vmovdqu %xmm12,(%rsi)
+ vpshufb %xmm0,%xmm12,%xmm12
+ vpxor %xmm1,%xmm12,%xmm12
+ vmovdqu (%r8),%xmm2
+
+.Lghash_mul_one_vec_unreduced__func1:
+ vpclmulqdq $0x00,%ymm2,%ymm12,%ymm4
+ vpxor %ymm4,%ymm5,%ymm5
+ vpclmulqdq $0x01,%ymm2,%ymm12,%ymm4
+ vpxor %ymm4,%ymm6,%ymm6
+ vpclmulqdq $0x10,%ymm2,%ymm12,%ymm4
+ vpxor %ymm4,%ymm6,%ymm6
+ vpclmulqdq $0x11,%ymm2,%ymm12,%ymm4
+ vpxor %ymm4,%ymm7,%ymm7
+
+.Lreduce__func1:
+
+ vbroadcasti128 .Lgfpoly(%rip),%ymm2
+ vpclmulqdq $0x01,%ymm5,%ymm2,%ymm3
+ vpshufd $0x4e,%ymm5,%ymm5
+ vpxor %ymm5,%ymm6,%ymm6
+ vpxor %ymm3,%ymm6,%ymm6
+ vpclmulqdq $0x01,%ymm6,%ymm2,%ymm3
+ vpshufd $0x4e,%ymm6,%ymm6
+ vpxor %ymm6,%ymm7,%ymm7
+ vpxor %ymm3,%ymm7,%ymm7
+ vextracti128 $1,%ymm7,%xmm1
+ vpxor %xmm7,%xmm1,%xmm1
+
+.Ldone__func1:
+
+ vpshufb %xmm0,%xmm1,%xmm1
+ vmovdqu %xmm1,(%r12)
+
+ vzeroupper
+ popq %r12
+.cfi_adjust_cfa_offset -8
+.cfi_restore %r12
+ ret
+
+.cfi_endproc
+.size aes_gcm_enc_update_vaes_avx2, . - aes_gcm_enc_update_vaes_avx2
+.globl aes_gcm_dec_update_vaes_avx2
+.hidden aes_gcm_dec_update_vaes_avx2
+.type aes_gcm_dec_update_vaes_avx2,@function
+.align 32
+aes_gcm_dec_update_vaes_avx2:
+.cfi_startproc
+
+_CET_ENDBR
+ pushq %r12
+.cfi_adjust_cfa_offset 8
+.cfi_offset %r12,-16
+
+ movq 16(%rsp),%r12
+ vbroadcasti128 .Lbswap_mask(%rip),%ymm0
+
+
+
+ vmovdqu (%r12),%xmm1
+ vpshufb %xmm0,%xmm1,%xmm1
+ vbroadcasti128 (%r8),%ymm11
+ vpshufb %ymm0,%ymm11,%ymm11
+
+
+
+ movl 240(%rcx),%r10d
+ leal -20(,%r10,4),%r10d
+
+
+
+
+ leaq 96(%rcx,%r10,4),%r11
+ vbroadcasti128 (%rcx),%ymm9
+ vbroadcasti128 (%r11),%ymm10
+
+
+ vpaddd .Lctr_pattern(%rip),%ymm11,%ymm11
+
+
+
+ cmpq $127,%rdx
+ jbe .Lcrypt_loop_4x_done__func2
+
+ vmovdqu 128(%r9),%ymm7
+ vmovdqu 128+32(%r9),%ymm8
+.align 16
+.Lcrypt_loop_4x__func2:
+
+
+
+
+ vmovdqu .Linc_2blocks(%rip),%ymm2
+ vpshufb %ymm0,%ymm11,%ymm12
+ vpaddd %ymm2,%ymm11,%ymm11
+ vpshufb %ymm0,%ymm11,%ymm13
+ vpaddd %ymm2,%ymm11,%ymm11
+ vpshufb %ymm0,%ymm11,%ymm14
+ vpaddd %ymm2,%ymm11,%ymm11
+ vpshufb %ymm0,%ymm11,%ymm15
+ vpaddd %ymm2,%ymm11,%ymm11
+
+
+ vpxor %ymm9,%ymm12,%ymm12
+ vpxor %ymm9,%ymm13,%ymm13
+ vpxor %ymm9,%ymm14,%ymm14
+ vpxor %ymm9,%ymm15,%ymm15
+
+ cmpl $24,%r10d
+ jl .Laes128__func2
+ je .Laes192__func2
+
+ vbroadcasti128 -208(%r11),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+ vbroadcasti128 -192(%r11),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+.Laes192__func2:
+ vbroadcasti128 -176(%r11),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+ vbroadcasti128 -160(%r11),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+.Laes128__func2:
+ prefetcht0 512(%rdi)
+ prefetcht0 512+64(%rdi)
+
+ vmovdqu 0(%rdi),%ymm3
+ vpshufb %ymm0,%ymm3,%ymm3
+ vmovdqu 0(%r9),%ymm4
+ vpxor %ymm1,%ymm3,%ymm3
+ vpclmulqdq $0x00,%ymm4,%ymm3,%ymm5
+ vpclmulqdq $0x11,%ymm4,%ymm3,%ymm1
+ vpunpckhqdq %ymm3,%ymm3,%ymm2
+ vpxor %ymm3,%ymm2,%ymm2
+ vpclmulqdq $0x00,%ymm7,%ymm2,%ymm6
+
+ vbroadcasti128 -144(%r11),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+
+ vbroadcasti128 -128(%r11),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+
+ vmovdqu 32(%rdi),%ymm3
+ vpshufb %ymm0,%ymm3,%ymm3
+ vmovdqu 32(%r9),%ymm4
+ vpclmulqdq $0x00,%ymm4,%ymm3,%ymm2
+ vpxor %ymm2,%ymm5,%ymm5
+ vpclmulqdq $0x11,%ymm4,%ymm3,%ymm2
+ vpxor %ymm2,%ymm1,%ymm1
+ vpunpckhqdq %ymm3,%ymm3,%ymm2
+ vpxor %ymm3,%ymm2,%ymm2
+ vpclmulqdq $0x10,%ymm7,%ymm2,%ymm2
+ vpxor %ymm2,%ymm6,%ymm6
+
+ vbroadcasti128 -112(%r11),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+
+ vmovdqu 64(%rdi),%ymm3
+ vpshufb %ymm0,%ymm3,%ymm3
+ vmovdqu 64(%r9),%ymm4
+
+ vbroadcasti128 -96(%r11),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+ vpclmulqdq $0x00,%ymm4,%ymm3,%ymm2
+ vpxor %ymm2,%ymm5,%ymm5
+ vpclmulqdq $0x11,%ymm4,%ymm3,%ymm2
+ vpxor %ymm2,%ymm1,%ymm1
+
+ vbroadcasti128 -80(%r11),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+ vpunpckhqdq %ymm3,%ymm3,%ymm2
+ vpxor %ymm3,%ymm2,%ymm2
+ vpclmulqdq $0x00,%ymm8,%ymm2,%ymm2
+ vpxor %ymm2,%ymm6,%ymm6
+
+
+ vmovdqu 96(%rdi),%ymm3
+ vpshufb %ymm0,%ymm3,%ymm3
+
+ vbroadcasti128 -64(%r11),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+ vmovdqu 96(%r9),%ymm4
+ vpclmulqdq $0x00,%ymm4,%ymm3,%ymm2
+ vpxor %ymm2,%ymm5,%ymm5
+ vpclmulqdq $0x11,%ymm4,%ymm3,%ymm2
+ vpxor %ymm2,%ymm1,%ymm1
+ vpunpckhqdq %ymm3,%ymm3,%ymm2
+ vpxor %ymm3,%ymm2,%ymm2
+ vpclmulqdq $0x10,%ymm8,%ymm2,%ymm2
+ vpxor %ymm2,%ymm6,%ymm6
+
+ vbroadcasti128 -48(%r11),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+
+ vpxor %ymm5,%ymm6,%ymm6
+ vpxor %ymm1,%ymm6,%ymm6
+
+
+ vbroadcasti128 .Lgfpoly(%rip),%ymm4
+ vpclmulqdq $0x01,%ymm5,%ymm4,%ymm2
+ vpshufd $0x4e,%ymm5,%ymm5
+ vpxor %ymm5,%ymm6,%ymm6
+ vpxor %ymm2,%ymm6,%ymm6
+
+ vbroadcasti128 -32(%r11),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+
+ vpclmulqdq $0x01,%ymm6,%ymm4,%ymm2
+ vpshufd $0x4e,%ymm6,%ymm6
+ vpxor %ymm6,%ymm1,%ymm1
+ vpxor %ymm2,%ymm1,%ymm1
+
+ vbroadcasti128 -16(%r11),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+ vextracti128 $1,%ymm1,%xmm2
+ vpxor %xmm2,%xmm1,%xmm1
+
+
+
+ vpxor 0(%rdi),%ymm10,%ymm2
+ vpxor 32(%rdi),%ymm10,%ymm3
+ vpxor 64(%rdi),%ymm10,%ymm5
+ vpxor 96(%rdi),%ymm10,%ymm6
+ vaesenclast %ymm2,%ymm12,%ymm12
+ vaesenclast %ymm3,%ymm13,%ymm13
+ vaesenclast %ymm5,%ymm14,%ymm14
+ vaesenclast %ymm6,%ymm15,%ymm15
+ vmovdqu %ymm12,0(%rsi)
+ vmovdqu %ymm13,32(%rsi)
+ vmovdqu %ymm14,64(%rsi)
+ vmovdqu %ymm15,96(%rsi)
+
+ subq $-128,%rdi
+ subq $-128,%rsi
+ addq $-128,%rdx
+ cmpq $127,%rdx
+ ja .Lcrypt_loop_4x__func2
+.Lcrypt_loop_4x_done__func2:
+
+ testq %rdx,%rdx
+ jz .Ldone__func2
+
+
+
+
+
+ leaq 128(%r9),%r8
+ subq %rdx,%r8
+
+
+ vpxor %xmm5,%xmm5,%xmm5
+ vpxor %xmm6,%xmm6,%xmm6
+ vpxor %xmm7,%xmm7,%xmm7
+
+ cmpq $64,%rdx
+ jb .Llessthan64bytes__func2
+
+
+ vpshufb %ymm0,%ymm11,%ymm12
+ vpaddd .Linc_2blocks(%rip),%ymm11,%ymm11
+ vpshufb %ymm0,%ymm11,%ymm13
+ vpaddd .Linc_2blocks(%rip),%ymm11,%ymm11
+ vpxor %ymm9,%ymm12,%ymm12
+ vpxor %ymm9,%ymm13,%ymm13
+ leaq 16(%rcx),%rax
+.Lvaesenc_loop_tail_1__func2:
+ vbroadcasti128 (%rax),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ addq $16,%rax
+ cmpq %rax,%r11
+ jne .Lvaesenc_loop_tail_1__func2
+ vaesenclast %ymm10,%ymm12,%ymm12
+ vaesenclast %ymm10,%ymm13,%ymm13
+
+
+ vmovdqu 0(%rdi),%ymm2
+ vmovdqu 32(%rdi),%ymm3
+ vpxor %ymm2,%ymm12,%ymm12
+ vpxor %ymm3,%ymm13,%ymm13
+ vmovdqu %ymm12,0(%rsi)
+ vmovdqu %ymm13,32(%rsi)
+
+
+ vpshufb %ymm0,%ymm2,%ymm12
+ vpshufb %ymm0,%ymm3,%ymm13
+ vpxor %ymm1,%ymm12,%ymm12
+ vmovdqu (%r8),%ymm2
+ vmovdqu 32(%r8),%ymm3
+ vpclmulqdq $0x00,%ymm2,%ymm12,%ymm5
+ vpclmulqdq $0x01,%ymm2,%ymm12,%ymm6
+ vpclmulqdq $0x10,%ymm2,%ymm12,%ymm4
+ vpxor %ymm4,%ymm6,%ymm6
+ vpclmulqdq $0x11,%ymm2,%ymm12,%ymm7
+ vpclmulqdq $0x00,%ymm3,%ymm13,%ymm4
+ vpxor %ymm4,%ymm5,%ymm5
+ vpclmulqdq $0x01,%ymm3,%ymm13,%ymm4
+ vpxor %ymm4,%ymm6,%ymm6
+ vpclmulqdq $0x10,%ymm3,%ymm13,%ymm4
+ vpxor %ymm4,%ymm6,%ymm6
+ vpclmulqdq $0x11,%ymm3,%ymm13,%ymm4
+ vpxor %ymm4,%ymm7,%ymm7
+
+ addq $64,%r8
+ addq $64,%rdi
+ addq $64,%rsi
+ subq $64,%rdx
+ jz .Lreduce__func2
+
+ vpxor %xmm1,%xmm1,%xmm1
+
+
+.Llessthan64bytes__func2:
+ vpshufb %ymm0,%ymm11,%ymm12
+ vpaddd .Linc_2blocks(%rip),%ymm11,%ymm11
+ vpshufb %ymm0,%ymm11,%ymm13
+ vpxor %ymm9,%ymm12,%ymm12
+ vpxor %ymm9,%ymm13,%ymm13
+ leaq 16(%rcx),%rax
+.Lvaesenc_loop_tail_2__func2:
+ vbroadcasti128 (%rax),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ addq $16,%rax
+ cmpq %rax,%r11
+ jne .Lvaesenc_loop_tail_2__func2
+ vaesenclast %ymm10,%ymm12,%ymm12
+ vaesenclast %ymm10,%ymm13,%ymm13
+
+
+
+
+ cmpq $32,%rdx
+ jb .Lxor_one_block__func2
+ je .Lxor_two_blocks__func2
+
+.Lxor_three_blocks__func2:
+ vmovdqu 0(%rdi),%ymm2
+ vmovdqu 32(%rdi),%xmm3
+ vpxor %ymm2,%ymm12,%ymm12
+ vpxor %xmm3,%xmm13,%xmm13
+ vmovdqu %ymm12,0(%rsi)
+ vmovdqu %xmm13,32(%rsi)
+
+ vpshufb %ymm0,%ymm2,%ymm12
+ vpshufb %xmm0,%xmm3,%xmm13
+ vpxor %ymm1,%ymm12,%ymm12
+ vmovdqu (%r8),%ymm2
+ vmovdqu 32(%r8),%xmm3
+ vpclmulqdq $0x00,%xmm3,%xmm13,%xmm4
+ vpxor %ymm4,%ymm5,%ymm5
+ vpclmulqdq $0x01,%xmm3,%xmm13,%xmm4
+ vpxor %ymm4,%ymm6,%ymm6
+ vpclmulqdq $0x10,%xmm3,%xmm13,%xmm4
+ vpxor %ymm4,%ymm6,%ymm6
+ vpclmulqdq $0x11,%xmm3,%xmm13,%xmm4
+ vpxor %ymm4,%ymm7,%ymm7
+ jmp .Lghash_mul_one_vec_unreduced__func2
+
+.Lxor_two_blocks__func2:
+ vmovdqu (%rdi),%ymm2
+ vpxor %ymm2,%ymm12,%ymm12
+ vmovdqu %ymm12,(%rsi)
+ vpshufb %ymm0,%ymm2,%ymm12
+ vpxor %ymm1,%ymm12,%ymm12
+ vmovdqu (%r8),%ymm2
+ jmp .Lghash_mul_one_vec_unreduced__func2
+
+.Lxor_one_block__func2:
+ vmovdqu (%rdi),%xmm2
+ vpxor %xmm2,%xmm12,%xmm12
+ vmovdqu %xmm12,(%rsi)
+ vpshufb %xmm0,%xmm2,%xmm12
+ vpxor %xmm1,%xmm12,%xmm12
+ vmovdqu (%r8),%xmm2
+
+.Lghash_mul_one_vec_unreduced__func2:
+ vpclmulqdq $0x00,%ymm2,%ymm12,%ymm4
+ vpxor %ymm4,%ymm5,%ymm5
+ vpclmulqdq $0x01,%ymm2,%ymm12,%ymm4
+ vpxor %ymm4,%ymm6,%ymm6
+ vpclmulqdq $0x10,%ymm2,%ymm12,%ymm4
+ vpxor %ymm4,%ymm6,%ymm6
+ vpclmulqdq $0x11,%ymm2,%ymm12,%ymm4
+ vpxor %ymm4,%ymm7,%ymm7
+
+.Lreduce__func2:
+
+ vbroadcasti128 .Lgfpoly(%rip),%ymm2
+ vpclmulqdq $0x01,%ymm5,%ymm2,%ymm3
+ vpshufd $0x4e,%ymm5,%ymm5
+ vpxor %ymm5,%ymm6,%ymm6
+ vpxor %ymm3,%ymm6,%ymm6
+ vpclmulqdq $0x01,%ymm6,%ymm2,%ymm3
+ vpshufd $0x4e,%ymm6,%ymm6
+ vpxor %ymm6,%ymm7,%ymm7
+ vpxor %ymm3,%ymm7,%ymm7
+ vextracti128 $1,%ymm7,%xmm1
+ vpxor %xmm7,%xmm1,%xmm1
+
+.Ldone__func2:
+
+ vpshufb %xmm0,%xmm1,%xmm1
+ vmovdqu %xmm1,(%r12)
+
+ vzeroupper
+ popq %r12
+.cfi_adjust_cfa_offset -8
+.cfi_restore %r12
+ ret
+
+.cfi_endproc
+.size aes_gcm_dec_update_vaes_avx2, . - aes_gcm_dec_update_vaes_avx2
+#endif
diff --git a/sys/contrib/openzfs/contrib/initramfs/scripts/zfs b/sys/contrib/openzfs/contrib/initramfs/scripts/zfs
index c569b2528368..67707e9d80f4 100644
--- a/sys/contrib/openzfs/contrib/initramfs/scripts/zfs
+++ b/sys/contrib/openzfs/contrib/initramfs/scripts/zfs
@@ -979,7 +979,8 @@ mountroot()
touch /run/zfs_unlock_complete
if [ -e /run/zfs_unlock_complete_notify ]; then
- read -r < /run/zfs_unlock_complete_notify
+ # shellcheck disable=SC2034
+ read -r zfs_unlock_complete_notify < /run/zfs_unlock_complete_notify
fi
# ------------
diff --git a/sys/contrib/openzfs/include/os/linux/kernel/linux/simd_x86.h b/sys/contrib/openzfs/include/os/linux/kernel/linux/simd_x86.h
index e8004e18c4a4..326f471d7c9b 100644
--- a/sys/contrib/openzfs/include/os/linux/kernel/linux/simd_x86.h
+++ b/sys/contrib/openzfs/include/os/linux/kernel/linux/simd_x86.h
@@ -598,6 +598,32 @@ zfs_movbe_available(void)
}
/*
+ * Check if VAES instruction set is available
+ */
+static inline boolean_t
+zfs_vaes_available(void)
+{
+#if defined(X86_FEATURE_VAES)
+ return (!!boot_cpu_has(X86_FEATURE_VAES));
+#else
+ return (B_FALSE);
+#endif
+}
+
+/*
+ * Check if VPCLMULQDQ instruction set is available
+ */
+static inline boolean_t
+zfs_vpclmulqdq_available(void)
+{
+#if defined(X86_FEATURE_VPCLMULQDQ)
+ return (!!boot_cpu_has(X86_FEATURE_VPCLMULQDQ));
+#else
+ return (B_FALSE);
+#endif
+}
+
+/*
* Check if SHA_NI instruction set is available
*/
static inline boolean_t
diff --git a/sys/contrib/openzfs/include/os/linux/zfs/sys/trace_zil.h b/sys/contrib/openzfs/include/os/linux/zfs/sys/trace_zil.h
index 955462c85d10..e34ea46b3fe8 100644
--- a/sys/contrib/openzfs/include/os/linux/zfs/sys/trace_zil.h
+++ b/sys/contrib/openzfs/include/os/linux/zfs/sys/trace_zil.h
@@ -139,18 +139,18 @@
#define ZCW_TP_STRUCT_ENTRY \
__field(lwb_t *, zcw_lwb) \
__field(boolean_t, zcw_done) \
- __field(int, zcw_zio_error) \
+ __field(int, zcw_error) \
#define ZCW_TP_FAST_ASSIGN \
__entry->zcw_lwb = zcw->zcw_lwb; \
__entry->zcw_done = zcw->zcw_done; \
- __entry->zcw_zio_error = zcw->zcw_zio_error;
+ __entry->zcw_error = zcw->zcw_error;
#define ZCW_TP_PRINTK_FMT \
"zcw { lwb %p done %u error %u }"
#define ZCW_TP_PRINTK_ARGS \
- __entry->zcw_lwb, __entry->zcw_done, __entry->zcw_zio_error
+ __entry->zcw_lwb, __entry->zcw_done, __entry->zcw_error
/*
* Generic support for two argument tracepoints of the form:
diff --git a/sys/contrib/openzfs/include/sys/spa.h b/sys/contrib/openzfs/include/sys/spa.h
index db6de332ae67..66db16b33c51 100644
--- a/sys/contrib/openzfs/include/sys/spa.h
+++ b/sys/contrib/openzfs/include/sys/spa.h
@@ -880,7 +880,6 @@ extern kcondvar_t spa_namespace_cv;
#define SPA_CONFIG_UPDATE_VDEVS 1
extern void spa_write_cachefile(spa_t *, boolean_t, boolean_t, boolean_t);
-extern void spa_config_load(void);
extern int spa_all_configs(uint64_t *generation, nvlist_t **pools);
extern void spa_config_set(spa_t *spa, nvlist_t *config);
extern nvlist_t *spa_config_generate(spa_t *spa, vdev_t *vd, uint64_t txg,
@@ -1244,7 +1243,6 @@ extern void vdev_mirror_stat_fini(void);
/* Initialization and termination */
extern void spa_init(spa_mode_t mode);
extern void spa_fini(void);
-extern void spa_boot_init(void *);
/* properties */
extern int spa_prop_set(spa_t *spa, nvlist_t *nvp);
diff --git a/sys/contrib/openzfs/include/sys/zil_impl.h b/sys/contrib/openzfs/include/sys/zil_impl.h
index 44b776e16b52..ea1364a7e35a 100644
--- a/sys/contrib/openzfs/include/sys/zil_impl.h
+++ b/sys/contrib/openzfs/include/sys/zil_impl.h
@@ -41,8 +41,8 @@ extern "C" {
*
* An lwb will start out in the "new" state, and transition to the "opened"
* state via a call to zil_lwb_write_open() on first itx assignment. When
- * transitioning from "new" to "opened" the zilog's "zl_issuer_lock" must be
- * held.
+ * transitioning from "new" to "opened" the zilog's "zl_issuer_lock" and
+ * LWB's "lwb_lock" must be held.
*
* After the lwb is "opened", it can be assigned number of itxs and transition
* into the "closed" state via zil_lwb_write_close() when full or on timeout.
@@ -100,16 +100,22 @@ typedef enum {
* holding the "zl_issuer_lock". After the lwb is issued, the zilog's
* "zl_lock" is used to protect the lwb against concurrent access.
*/
+typedef enum {
+ LWB_FLAG_SLIM = (1<<0), /* log block has slim format */
+ LWB_FLAG_SLOG = (1<<1), /* lwb_blk is on SLOG device */
+ LWB_FLAG_CRASHED = (1<<2), /* lwb is on the crash list */
+} lwb_flag_t;
+
typedef struct lwb {
zilog_t *lwb_zilog; /* back pointer to log struct */
blkptr_t lwb_blk; /* on disk address of this log blk */
- boolean_t lwb_slim; /* log block has slim format */
- boolean_t lwb_slog; /* lwb_blk is on SLOG device */
+ lwb_flag_t lwb_flags; /* extra info about this lwb */
int lwb_error; /* log block allocation error */
int lwb_nmax; /* max bytes in the buffer */
int lwb_nused; /* # used bytes in buffer */
int lwb_nfilled; /* # filled bytes in buffer */
int lwb_sz; /* size of block and buffer */
+ int lwb_min_sz; /* min size for range allocation */
lwb_state_t lwb_state; /* the state of this lwb */
char *lwb_buf; /* log write buffer */
zio_t *lwb_child_zio; /* parent zio for children */
@@ -124,7 +130,7 @@ typedef struct lwb {
list_t lwb_itxs; /* list of itx's */
list_t lwb_waiters; /* list of zil_commit_waiter's */
avl_tree_t lwb_vdev_tree; /* vdevs to flush after lwb write */
- kmutex_t lwb_vdev_lock; /* protects lwb_vdev_tree */
+ kmutex_t lwb_lock; /* protects lwb_vdev_tree and size */
} lwb_t;
/*
@@ -149,7 +155,7 @@ typedef struct zil_commit_waiter {
list_node_t zcw_node; /* linkage in lwb_t:lwb_waiter list */
lwb_t *zcw_lwb; /* back pointer to lwb when linked */
boolean_t zcw_done; /* B_TRUE when "done", else B_FALSE */
- int zcw_zio_error; /* contains the zio io_error value */
+ int zcw_error; /* result to return from zil_commit() */
} zil_commit_waiter_t;
/*
diff --git a/sys/contrib/openzfs/include/sys/zio.h b/sys/contrib/openzfs/include/sys/zio.h
index 4f46eab3db89..353805fcb969 100644
--- a/sys/contrib/openzfs/include/sys/zio.h
+++ b/sys/contrib/openzfs/include/sys/zio.h
@@ -360,26 +360,26 @@ struct zbookmark_err_phys {
(zb)->zb_blkid == ZB_ROOT_BLKID)
typedef struct zio_prop {
- enum zio_checksum zp_checksum;
- enum zio_compress zp_compress;
+ enum zio_checksum zp_checksum:8;
+ enum zio_compress zp_compress:8;
uint8_t zp_complevel;
uint8_t zp_level;
uint8_t zp_copies;
uint8_t zp_gang_copies;
- dmu_object_type_t zp_type;
- boolean_t zp_dedup;
- boolean_t zp_dedup_verify;
- boolean_t zp_nopwrite;
- boolean_t zp_brtwrite;
- boolean_t zp_encrypt;
- boolean_t zp_byteorder;
- boolean_t zp_direct_write;
- boolean_t zp_rewrite;
+ dmu_object_type_t zp_type:8;
+ dmu_object_type_t zp_storage_type:8;
+ boolean_t zp_dedup:1;
+ boolean_t zp_dedup_verify:1;
+ boolean_t zp_nopwrite:1;
+ boolean_t zp_brtwrite:1;
+ boolean_t zp_encrypt:1;
+ boolean_t zp_byteorder:1;
+ boolean_t zp_direct_write:1;
+ boolean_t zp_rewrite:1;
+ uint32_t zp_zpl_smallblk;
uint8_t zp_salt[ZIO_DATA_SALT_LEN];
uint8_t zp_iv[ZIO_DATA_IV_LEN];
uint8_t zp_mac[ZIO_DATA_MAC_LEN];
- uint32_t zp_zpl_smallblk;
- dmu_object_type_t zp_storage_type;
} zio_prop_t;
typedef struct zio_cksum_report zio_cksum_report_t;
@@ -622,7 +622,8 @@ extern zio_t *zio_free_sync(zio_t *pio, spa_t *spa, uint64_t txg,
const blkptr_t *bp, zio_flag_t flags);
extern int zio_alloc_zil(spa_t *spa, objset_t *os, uint64_t txg,
- blkptr_t *new_bp, uint64_t size, boolean_t *slog);
+ blkptr_t *new_bp, uint64_t min_size, uint64_t max_size, boolean_t *slog,
+ boolean_t allow_larger);
extern void zio_flush(zio_t *zio, vdev_t *vd);
extern void zio_shrink(zio_t *zio, uint64_t size);
diff --git a/sys/contrib/openzfs/include/sys/zvol_impl.h b/sys/contrib/openzfs/include/sys/zvol_impl.h
index f3dd9f26f23c..5422e66832c0 100644
--- a/sys/contrib/openzfs/include/sys/zvol_impl.h
+++ b/sys/contrib/openzfs/include/sys/zvol_impl.h
@@ -20,7 +20,7 @@
* CDDL HEADER END
*/
/*
- * Copyright (c) 2024, Klara, Inc.
+ * Copyright (c) 2024, 2025, Klara, Inc.
*/
#ifndef _SYS_ZVOL_IMPL_H
@@ -56,6 +56,7 @@ typedef struct zvol_state {
atomic_t zv_suspend_ref; /* refcount for suspend */
krwlock_t zv_suspend_lock; /* suspend lock */
kcondvar_t zv_removing_cv; /* ready to remove minor */
+ list_node_t zv_remove_node; /* node on removal list */
struct zvol_state_os *zv_zso; /* private platform state */
boolean_t zv_threading; /* volthreading property */
} zvol_state_t;
@@ -135,7 +136,7 @@ int zvol_os_rename_minor(zvol_state_t *zv, const char *newname);
int zvol_os_create_minor(const char *name);
int zvol_os_update_volsize(zvol_state_t *zv, uint64_t volsize);
boolean_t zvol_os_is_zvol(const char *path);
-void zvol_os_clear_private(zvol_state_t *zv);
+void zvol_os_remove_minor(zvol_state_t *zv);
void zvol_os_set_disk_ro(zvol_state_t *zv, int flags);
void zvol_os_set_capacity(zvol_state_t *zv, uint64_t capacity);
diff --git a/sys/contrib/openzfs/lib/libicp/Makefile.am b/sys/contrib/openzfs/lib/libicp/Makefile.am
index ce24d13a760f..23adba10bc44 100644
--- a/sys/contrib/openzfs/lib/libicp/Makefile.am
+++ b/sys/contrib/openzfs/lib/libicp/Makefile.am
@@ -69,6 +69,7 @@ nodist_libicp_la_SOURCES += \
module/icp/asm-x86_64/aes/aes_aesni.S \
module/icp/asm-x86_64/modes/gcm_pclmulqdq.S \
module/icp/asm-x86_64/modes/aesni-gcm-x86_64.S \
+ module/icp/asm-x86_64/modes/aesni-gcm-avx2-vaes.S \
module/icp/asm-x86_64/modes/ghash-x86_64.S \
module/icp/asm-x86_64/sha2/sha256-x86_64.S \
module/icp/asm-x86_64/sha2/sha512-x86_64.S \
diff --git a/sys/contrib/openzfs/lib/libspl/include/sys/simd.h b/sys/contrib/openzfs/lib/libspl/include/sys/simd.h
index 1ef24f5a7d39..4772a5416b2e 100644
--- a/sys/contrib/openzfs/lib/libspl/include/sys/simd.h
+++ b/sys/contrib/openzfs/lib/libspl/include/sys/simd.h
@@ -102,7 +102,9 @@ typedef enum cpuid_inst_sets {
AES,
PCLMULQDQ,
MOVBE,
- SHA_NI
+ SHA_NI,
+ VAES,
+ VPCLMULQDQ
} cpuid_inst_sets_t;
/*
@@ -127,6 +129,8 @@ typedef struct cpuid_feature_desc {
#define _AES_BIT (1U << 25)
#define _PCLMULQDQ_BIT (1U << 1)
#define _MOVBE_BIT (1U << 22)
+#define _VAES_BIT (1U << 9)
+#define _VPCLMULQDQ_BIT (1U << 10)
#define _SHA_NI_BIT (1U << 29)
/*
@@ -157,6 +161,8 @@ static const cpuid_feature_desc_t cpuid_features[] = {
[PCLMULQDQ] = {1U, 0U, _PCLMULQDQ_BIT, ECX },
[MOVBE] = {1U, 0U, _MOVBE_BIT, ECX },
[SHA_NI] = {7U, 0U, _SHA_NI_BIT, EBX },
+ [VAES] = {7U, 0U, _VAES_BIT, ECX },
+ [VPCLMULQDQ] = {7U, 0U, _VPCLMULQDQ_BIT, ECX },
};
/*
@@ -231,6 +237,8 @@ CPUID_FEATURE_CHECK(aes, AES);
CPUID_FEATURE_CHECK(pclmulqdq, PCLMULQDQ);
CPUID_FEATURE_CHECK(movbe, MOVBE);
CPUID_FEATURE_CHECK(shani, SHA_NI);
+CPUID_FEATURE_CHECK(vaes, VAES);
+CPUID_FEATURE_CHECK(vpclmulqdq, VPCLMULQDQ);
/*
* Detect register set support
@@ -382,6 +390,24 @@ zfs_shani_available(void)
}
/*
+ * Check if VAES instruction is available
+ */
+static inline boolean_t
+zfs_vaes_available(void)
+{
+ return (__cpuid_has_vaes());
+}
+
+/*
+ * Check if VPCLMULQDQ instruction is available
+ */
+static inline boolean_t
+zfs_vpclmulqdq_available(void)
+{
+ return (__cpuid_has_vpclmulqdq());
+}
+
+/*
* AVX-512 family of instruction sets:
*
* AVX512F Foundation
diff --git a/sys/contrib/openzfs/lib/libzpool/kernel.c b/sys/contrib/openzfs/lib/libzpool/kernel.c
index e63153a03370..fea2f81458f9 100644
--- a/sys/contrib/openzfs/lib/libzpool/kernel.c
+++ b/sys/contrib/openzfs/lib/libzpool/kernel.c
@@ -38,6 +38,7 @@
#include <sys/processor.h>
#include <sys/rrwlock.h>
#include <sys/spa.h>
+#include <sys/spa_impl.h>
#include <sys/stat.h>
#include <sys/systeminfo.h>
#include <sys/time.h>
@@ -811,6 +812,79 @@ umem_out_of_memory(void)
return (0);
}
+static void
+spa_config_load(void)
+{
+ void *buf = NULL;
+ nvlist_t *nvlist, *child;
+ nvpair_t *nvpair;
+ char *pathname;
+ zfs_file_t *fp;
+ zfs_file_attr_t zfa;
+ uint64_t fsize;
+ int err;
+
+ /*
+ * Open the configuration file.
+ */
+ pathname = kmem_alloc(MAXPATHLEN, KM_SLEEP);
+
+ (void) snprintf(pathname, MAXPATHLEN, "%s", spa_config_path);
+
+ err = zfs_file_open(pathname, O_RDONLY, 0, &fp);
+ if (err)
+ err = zfs_file_open(ZPOOL_CACHE_BOOT, O_RDONLY, 0, &fp);
+
+ kmem_free(pathname, MAXPATHLEN);
+
+ if (err)
+ return;
+
+ if (zfs_file_getattr(fp, &zfa))
+ goto out;
+
+ fsize = zfa.zfa_size;
+ buf = kmem_alloc(fsize, KM_SLEEP);
+
+ /*
+ * Read the nvlist from the file.
+ */
+ if (zfs_file_read(fp, buf, fsize, NULL) < 0)
+ goto out;
+
+ /*
+ * Unpack the nvlist.
+ */
+ if (nvlist_unpack(buf, fsize, &nvlist, KM_SLEEP) != 0)
+ goto out;
+
+ /*
+ * Iterate over all elements in the nvlist, creating a new spa_t for
+ * each one with the specified configuration.
+ */
+ mutex_enter(&spa_namespace_lock);
+ nvpair = NULL;
+ while ((nvpair = nvlist_next_nvpair(nvlist, nvpair)) != NULL) {
+ if (nvpair_type(nvpair) != DATA_TYPE_NVLIST)
+ continue;
+
+ child = fnvpair_value_nvlist(nvpair);
+
+ if (spa_lookup(nvpair_name(nvpair)) != NULL)
+ continue;
+ (void) spa_add(nvpair_name(nvpair), child, NULL);
+ }
+ mutex_exit(&spa_namespace_lock);
+
+ nvlist_free(nvlist);
+
+out:
+ if (buf != NULL)
+ kmem_free(buf, fsize);
+
+ zfs_file_close(fp);
+}
+
void
kernel_init(int mode)
{
@@ -835,6 +909,7 @@ kernel_init(int mode)
zstd_init();
spa_init((spa_mode_t)mode);
+ spa_config_load();
fletcher_4_init();
diff --git a/sys/contrib/openzfs/lib/libzutil/zutil_import.c b/sys/contrib/openzfs/lib/libzutil/zutil_import.c
index a4a6e76a1d09..08367f4c064d 100644
--- a/sys/contrib/openzfs/lib/libzutil/zutil_import.c
+++ b/sys/contrib/openzfs/lib/libzutil/zutil_import.c
@@ -1903,30 +1903,43 @@ zpool_find_config(libpc_handle_t *hdl, const char *target, nvlist_t **configp,
*sepp = '\0';
pools = zpool_search_import(hdl, args);
+ if (pools == NULL) {
+ zutil_error_aux(hdl, dgettext(TEXT_DOMAIN, "no pools found"));
+ (void) zutil_error_fmt(hdl, LPC_UNKNOWN, dgettext(TEXT_DOMAIN,
+ "failed to find config for pool '%s'"), targetdup);
+ free(targetdup);
+ return (ENOENT);
+ }
- if (pools != NULL) {
- nvpair_t *elem = NULL;
- while ((elem = nvlist_next_nvpair(pools, elem)) != NULL) {
- VERIFY0(nvpair_value_nvlist(elem, &config));
- if (pool_match(config, targetdup)) {
- count++;
- if (match != NULL) {
- /* multiple matches found */
- continue;
- } else {
- match = fnvlist_dup(config);
- }
+ nvpair_t *elem = NULL;
+ while ((elem = nvlist_next_nvpair(pools, elem)) != NULL) {
+ VERIFY0(nvpair_value_nvlist(elem, &config));
+ if (pool_match(config, targetdup)) {
+ count++;
+ if (match != NULL) {
+ /* multiple matches found */
+ continue;
+ } else {
+ match = fnvlist_dup(config);
}
}
- fnvlist_free(pools);
}
+ fnvlist_free(pools);
if (count == 0) {
+ zutil_error_aux(hdl, dgettext(TEXT_DOMAIN,
+ "no matching pools"));
+ (void) zutil_error_fmt(hdl, LPC_UNKNOWN, dgettext(TEXT_DOMAIN,
+ "failed to find config for pool '%s'"), targetdup);
free(targetdup);
return (ENOENT);
}
if (count > 1) {
+ zutil_error_aux(hdl, dgettext(TEXT_DOMAIN,
+ "more than one matching pool"));
+ (void) zutil_error_fmt(hdl, LPC_UNKNOWN, dgettext(TEXT_DOMAIN,
+ "failed to find config for pool '%s'"), targetdup);
free(targetdup);
fnvlist_free(match);
return (EINVAL);
diff --git a/sys/contrib/openzfs/man/man4/zfs.4 b/sys/contrib/openzfs/man/man4/zfs.4
index 4a5f9fd93f4f..5c7958667f92 100644
--- a/sys/contrib/openzfs/man/man4/zfs.4
+++ b/sys/contrib/openzfs/man/man4/zfs.4
@@ -941,10 +941,6 @@ The target number of bytes the ARC should leave as free memory on the system.
If zero, equivalent to the bigger of
.Sy 512 KiB No and Sy all_system_memory/64 .
.
-.It Sy zfs_autoimport_disable Ns = Ns Sy 1 Ns | Ns 0 Pq int
-Disable pool import at module load by ignoring the cache file
-.Pq Sy spa_config_path .
-.
.It Sy zfs_checksum_events_per_second Ns = Ns Sy 20 Ns /s Pq uint
Rate limit checksum events to this many per second.
Note that this should not be set below the ZED thresholds
diff --git a/sys/contrib/openzfs/man/man8/zdb.8 b/sys/contrib/openzfs/man/man8/zdb.8
index 3984aaac5866..0a5b6af73fdb 100644
--- a/sys/contrib/openzfs/man/man8/zdb.8
+++ b/sys/contrib/openzfs/man/man8/zdb.8
@@ -15,7 +15,7 @@
.\" Copyright (c) 2017 Lawrence Livermore National Security, LLC.
.\" Copyright (c) 2017 Intel Corporation.
.\"
-.Dd October 27, 2024
+.Dd April 23, 2025
.Dt ZDB 8
.Os
.
@@ -531,6 +531,18 @@ option, with more occurrences enabling more verbosity.
If no options are specified, all information about the named pool will be
displayed at default verbosity.
.
+.Sh EXIT STATUS
+The
+.Nm
+utility exits
+.Sy 0
+on success,
+.Sy 1
+if a fatal error occurs,
+.Sy 2
+if invalid command line options were specified, or
+.Sy 3
+if on-disk corruption was detected, but was not fatal.
.Sh EXAMPLES
.Ss Example 1 : No Display the configuration of imported pool Ar rpool
.Bd -literal
diff --git a/sys/contrib/openzfs/man/man8/zfs-send.8 b/sys/contrib/openzfs/man/man8/zfs-send.8
index c920a5a48798..f7c6b840303c 100644
--- a/sys/contrib/openzfs/man/man8/zfs-send.8
+++ b/sys/contrib/openzfs/man/man8/zfs-send.8
@@ -173,8 +173,10 @@ The receiving system must have the
feature enabled.
If the
.Sy lz4_compress
-feature is active on the sending system, then the receiving system must have
-that feature enabled as well.
+or
+.Sy zstd_compress
+features are active on the sending system, then the receiving system must have
+the corresponding features enabled as well.
Datasets that are sent with this flag may not be
received as an encrypted dataset, since encrypted datasets cannot use the
.Sy embedded_data
@@ -201,8 +203,10 @@ property for details
.Pc .
If the
.Sy lz4_compress
-feature is active on the sending system, then the receiving system must have
-that feature enabled as well.
+or
+.Sy zstd_compress
+features are active on the sending system, then the receiving system must have
+the corresponding features enabled as well.
If the
.Sy large_blocks
feature is enabled on the sending system but the
@@ -357,8 +361,10 @@ property for details
.Pc .
If the
.Sy lz4_compress
-feature is active on the sending system, then the receiving system must have
-that feature enabled as well.
+or
+.Sy zstd_compress
+features are active on the sending system, then the receiving system must have
+the corresponding features enabled as well.
If the
.Sy large_blocks
feature is enabled on the sending system but the
@@ -400,8 +406,10 @@ The receiving system must have the
feature enabled.
If the
.Sy lz4_compress
-feature is active on the sending system, then the receiving system must have
-that feature enabled as well.
+or
+.Sy zstd_compress
+features are active on the sending system, then the receiving system must have
+the corresponding features enabled as well.
Datasets that are sent with this flag may not be received as an encrypted
dataset,
since encrypted datasets cannot use the
diff --git a/sys/contrib/openzfs/module/Kbuild.in b/sys/contrib/openzfs/module/Kbuild.in
index 3d6f288fa5da..362d2295e091 100644
--- a/sys/contrib/openzfs/module/Kbuild.in
+++ b/sys/contrib/openzfs/module/Kbuild.in
@@ -135,6 +135,7 @@ ICP_OBJS_X86_64 := \
asm-x86_64/sha2/sha256-x86_64.o \
asm-x86_64/sha2/sha512-x86_64.o \
asm-x86_64/modes/aesni-gcm-x86_64.o \
+ asm-x86_64/modes/aesni-gcm-avx2-vaes.o \
asm-x86_64/modes/gcm_pclmulqdq.o \
asm-x86_64/modes/ghash-x86_64.o
diff --git a/sys/contrib/openzfs/module/icp/algs/modes/gcm.c b/sys/contrib/openzfs/module/icp/algs/modes/gcm.c
index c2a982b5a376..3cfa5b8165ce 100644
--- a/sys/contrib/openzfs/module/icp/algs/modes/gcm.c
+++ b/sys/contrib/openzfs/module/icp/algs/modes/gcm.c
@@ -46,6 +46,9 @@
#define IMPL_CYCLE (UINT32_MAX-1)
#ifdef CAN_USE_GCM_ASM
#define IMPL_AVX (UINT32_MAX-2)
+#if CAN_USE_GCM_ASM >= 2
+#define IMPL_AVX2 (UINT32_MAX-3)
+#endif
#endif
#define GCM_IMPL_READ(i) (*(volatile uint32_t *) &(i))
static uint32_t icp_gcm_impl = IMPL_FASTEST;
@@ -56,17 +59,16 @@ static uint32_t user_sel_impl = IMPL_FASTEST;
boolean_t gcm_avx_can_use_movbe = B_FALSE;
/*
* Whether to use the optimized openssl gcm and ghash implementations.
- * Set to true if module parameter icp_gcm_impl == "avx".
*/
-static boolean_t gcm_use_avx = B_FALSE;
-#define GCM_IMPL_USE_AVX (*(volatile boolean_t *)&gcm_use_avx)
+static gcm_impl gcm_impl_used = GCM_IMPL_GENERIC;
+#define GCM_IMPL_USED (*(volatile gcm_impl *)&gcm_impl_used)
extern boolean_t ASMABI atomic_toggle_boolean_nv(volatile boolean_t *);
static inline boolean_t gcm_avx_will_work(void);
-static inline void gcm_set_avx(boolean_t);
-static inline boolean_t gcm_toggle_avx(void);
-static inline size_t gcm_simd_get_htab_size(boolean_t);
+static inline boolean_t gcm_avx2_will_work(void);
+static inline void gcm_use_impl(gcm_impl impl);
+static inline gcm_impl gcm_toggle_impl(void);
static int gcm_mode_encrypt_contiguous_blocks_avx(gcm_ctx_t *, char *, size_t,
crypto_data_t *, size_t);
@@ -89,7 +91,7 @@ gcm_mode_encrypt_contiguous_blocks(gcm_ctx_t *ctx, char *data, size_t length,
void (*xor_block)(uint8_t *, uint8_t *))
{
#ifdef CAN_USE_GCM_ASM
- if (ctx->gcm_use_avx == B_TRUE)
+ if (ctx->impl != GCM_IMPL_GENERIC)
return (gcm_mode_encrypt_contiguous_blocks_avx(
ctx, data, length, out, block_size));
#endif
@@ -208,7 +210,7 @@ gcm_encrypt_final(gcm_ctx_t *ctx, crypto_data_t *out, size_t block_size,
{
(void) copy_block;
#ifdef CAN_USE_GCM_ASM
- if (ctx->gcm_use_avx == B_TRUE)
+ if (ctx->impl != GCM_IMPL_GENERIC)
return (gcm_encrypt_final_avx(ctx, out, block_size));
#endif
@@ -374,7 +376,7 @@ gcm_decrypt_final(gcm_ctx_t *ctx, crypto_data_t *out, size_t block_size,
void (*xor_block)(uint8_t *, uint8_t *))
{
#ifdef CAN_USE_GCM_ASM
- if (ctx->gcm_use_avx == B_TRUE)
+ if (ctx->impl != GCM_IMPL_GENERIC)
return (gcm_decrypt_final_avx(ctx, out, block_size));
#endif
@@ -631,23 +633,23 @@ gcm_init_ctx(gcm_ctx_t *gcm_ctx, char *param,
((aes_key_t *)gcm_ctx->gcm_keysched)->ops->needs_byteswap;
if (GCM_IMPL_READ(icp_gcm_impl) != IMPL_CYCLE) {
- gcm_ctx->gcm_use_avx = GCM_IMPL_USE_AVX;
+ gcm_ctx->impl = GCM_IMPL_USED;
} else {
/*
- * Handle the "cycle" implementation by creating avx and
- * non-avx contexts alternately.
+ * Handle the "cycle" implementation by creating different
+ * contexts, one per implementation.
*/
- gcm_ctx->gcm_use_avx = gcm_toggle_avx();
+ gcm_ctx->impl = gcm_toggle_impl();
- /* The avx impl. doesn't handle byte swapped key schedules. */
- if (gcm_ctx->gcm_use_avx == B_TRUE && needs_bswap == B_TRUE) {
- gcm_ctx->gcm_use_avx = B_FALSE;
+ /* The AVX impl. doesn't handle byte swapped key schedules. */
+ if (needs_bswap == B_TRUE) {
+ gcm_ctx->impl = GCM_IMPL_GENERIC;
}
/*
- * If this is a GCM context, use the MOVBE and the BSWAP
+ * If this is an AVX context, use the MOVBE and the BSWAP
* variants alternately.
*/
- if (gcm_ctx->gcm_use_avx == B_TRUE &&
+ if (gcm_ctx->impl == GCM_IMPL_AVX &&
zfs_movbe_available() == B_TRUE) {
(void) atomic_toggle_boolean_nv(
(volatile boolean_t *)&gcm_avx_can_use_movbe);
@@ -658,12 +660,13 @@ gcm_init_ctx(gcm_ctx_t *gcm_ctx, char *param,
* still they could be created by the aes generic implementation.
* Make sure not to use them since we'll corrupt data if we do.
*/
- if (gcm_ctx->gcm_use_avx == B_TRUE && needs_bswap == B_TRUE) {
- gcm_ctx->gcm_use_avx = B_FALSE;
+ if (gcm_ctx->impl != GCM_IMPL_GENERIC && needs_bswap == B_TRUE) {
+ gcm_ctx->impl = GCM_IMPL_GENERIC;
cmn_err_once(CE_WARN,
"ICP: Can't use the aes generic or cycle implementations "
- "in combination with the gcm avx implementation!");
+ "in combination with the gcm avx or avx2-vaes "
+ "implementation!");
cmn_err_once(CE_WARN,
"ICP: Falling back to a compatible implementation, "
"aes-gcm performance will likely be degraded.");
@@ -672,36 +675,20 @@ gcm_init_ctx(gcm_ctx_t *gcm_ctx, char *param,
"restore performance.");
}
- /* Allocate Htab memory as needed. */
- if (gcm_ctx->gcm_use_avx == B_TRUE) {
- size_t htab_len = gcm_simd_get_htab_size(gcm_ctx->gcm_use_avx);
-
- if (htab_len == 0) {
- return (CRYPTO_MECHANISM_PARAM_INVALID);
- }
- gcm_ctx->gcm_htab_len = htab_len;
- gcm_ctx->gcm_Htable =
- kmem_alloc(htab_len, KM_SLEEP);
-
- if (gcm_ctx->gcm_Htable == NULL) {
- return (CRYPTO_HOST_MEMORY);
- }
+ /*
+ * AVX implementations use Htable with sizes depending on
+ * implementation.
+ */
+ if (gcm_ctx->impl != GCM_IMPL_GENERIC) {
+ rv = gcm_init_avx(gcm_ctx, iv, iv_len, aad, aad_len,
+ block_size);
}
- /* Avx and non avx context initialization differs from here on. */
- if (gcm_ctx->gcm_use_avx == B_FALSE) {
+ else
#endif /* ifdef CAN_USE_GCM_ASM */
- if (gcm_init(gcm_ctx, iv, iv_len, aad, aad_len, block_size,
- encrypt_block, copy_block, xor_block) != CRYPTO_SUCCESS) {
- rv = CRYPTO_MECHANISM_PARAM_INVALID;
- }
-#ifdef CAN_USE_GCM_ASM
- } else {
- if (gcm_init_avx(gcm_ctx, iv, iv_len, aad, aad_len,
- block_size) != CRYPTO_SUCCESS) {
- rv = CRYPTO_MECHANISM_PARAM_INVALID;
- }
+ if (gcm_init(gcm_ctx, iv, iv_len, aad, aad_len, block_size,
+ encrypt_block, copy_block, xor_block) != CRYPTO_SUCCESS) {
+ rv = CRYPTO_MECHANISM_PARAM_INVALID;
}
-#endif /* ifdef CAN_USE_GCM_ASM */
return (rv);
}
@@ -767,6 +754,9 @@ gcm_impl_get_ops(void)
break;
#ifdef CAN_USE_GCM_ASM
case IMPL_AVX:
+#if CAN_USE_GCM_ASM >= 2
+ case IMPL_AVX2:
+#endif
/*
* Make sure that we return a valid implementation while
* switching to the avx implementation since there still
@@ -828,6 +818,13 @@ gcm_impl_init(void)
* Use the avx implementation if it's available and the implementation
* hasn't changed from its default value of fastest on module load.
*/
+#if CAN_USE_GCM_ASM >= 2
+ if (gcm_avx2_will_work()) {
+ if (GCM_IMPL_READ(user_sel_impl) == IMPL_FASTEST) {
+ gcm_use_impl(GCM_IMPL_AVX2);
+ }
+ } else
+#endif
if (gcm_avx_will_work()) {
#ifdef HAVE_MOVBE
if (zfs_movbe_available() == B_TRUE) {
@@ -835,7 +832,7 @@ gcm_impl_init(void)
}
#endif
if (GCM_IMPL_READ(user_sel_impl) == IMPL_FASTEST) {
- gcm_set_avx(B_TRUE);
+ gcm_use_impl(GCM_IMPL_AVX);
}
}
#endif
@@ -852,6 +849,7 @@ static const struct {
{ "fastest", IMPL_FASTEST },
#ifdef CAN_USE_GCM_ASM
{ "avx", IMPL_AVX },
+ { "avx2-vaes", IMPL_AVX2 },
#endif
};
@@ -887,7 +885,13 @@ gcm_impl_set(const char *val)
/* Check mandatory options */
for (i = 0; i < ARRAY_SIZE(gcm_impl_opts); i++) {
#ifdef CAN_USE_GCM_ASM
+#if CAN_USE_GCM_ASM >= 2
/* Ignore avx implementation if it won't work. */
+ if (gcm_impl_opts[i].sel == IMPL_AVX2 &&
+ !gcm_avx2_will_work()) {
+ continue;
+ }
+#endif
if (gcm_impl_opts[i].sel == IMPL_AVX && !gcm_avx_will_work()) {
continue;
}
@@ -915,11 +919,17 @@ gcm_impl_set(const char *val)
* Use the avx implementation if available and the requested one is
* avx or fastest.
*/
+#if CAN_USE_GCM_ASM >= 2
+ if (gcm_avx2_will_work() == B_TRUE &&
+ (impl == IMPL_AVX2 || impl == IMPL_FASTEST)) {
+ gcm_use_impl(GCM_IMPL_AVX2);
+ } else
+#endif
if (gcm_avx_will_work() == B_TRUE &&
(impl == IMPL_AVX || impl == IMPL_FASTEST)) {
- gcm_set_avx(B_TRUE);
+ gcm_use_impl(GCM_IMPL_AVX);
} else {
- gcm_set_avx(B_FALSE);
+ gcm_use_impl(GCM_IMPL_GENERIC);
}
#endif
@@ -952,6 +962,12 @@ icp_gcm_impl_get(char *buffer, zfs_kernel_param_t *kp)
for (i = 0; i < ARRAY_SIZE(gcm_impl_opts); i++) {
#ifdef CAN_USE_GCM_ASM
/* Ignore avx implementation if it won't work. */
+#if CAN_USE_GCM_ASM >= 2
+ if (gcm_impl_opts[i].sel == IMPL_AVX2 &&
+ !gcm_avx2_will_work()) {
+ continue;
+ }
+#endif
if (gcm_impl_opts[i].sel == IMPL_AVX && !gcm_avx_will_work()) {
continue;
}
@@ -993,9 +1009,6 @@ MODULE_PARM_DESC(icp_gcm_impl, "Select gcm implementation.");
/* Clear the FPU registers since they hold sensitive internal state. */
#define clear_fpu_regs() clear_fpu_regs_avx()
-#define GHASH_AVX(ctx, in, len) \
- gcm_ghash_avx((ctx)->gcm_ghash, (const uint64_t *)(ctx)->gcm_Htable, \
- in, len)
#define gcm_incr_counter_block(ctx) gcm_incr_counter_block_by(ctx, 1)
@@ -1010,20 +1023,77 @@ MODULE_PARM_DESC(icp_gcm_impl, "Select gcm implementation.");
static uint32_t gcm_avx_chunk_size =
((32 * 1024) / GCM_AVX_MIN_DECRYPT_BYTES) * GCM_AVX_MIN_DECRYPT_BYTES;
+/*
+ * GCM definitions: uint128_t is copied from include/crypto/modes.h
+ * Avoiding u128 because it is already defined in kernel sources.
+ */
+typedef struct {
+ uint64_t hi, lo;
+} uint128_t;
+
extern void ASMABI clear_fpu_regs_avx(void);
extern void ASMABI gcm_xor_avx(const uint8_t *src, uint8_t *dst);
extern void ASMABI aes_encrypt_intel(const uint32_t rk[], int nr,
const uint32_t pt[4], uint32_t ct[4]);
extern void ASMABI gcm_init_htab_avx(uint64_t *Htable, const uint64_t H[2]);
+#if CAN_USE_GCM_ASM >= 2
+extern void ASMABI gcm_init_vpclmulqdq_avx2(uint128_t Htable[16],
+ const uint64_t H[2]);
+#endif
extern void ASMABI gcm_ghash_avx(uint64_t ghash[2], const uint64_t *Htable,
const uint8_t *in, size_t len);
+#if CAN_USE_GCM_ASM >= 2
+extern void ASMABI gcm_ghash_vpclmulqdq_avx2(uint64_t ghash[2],
+ const uint64_t *Htable, const uint8_t *in, size_t len);
+#endif
+static inline void GHASH_AVX(gcm_ctx_t *ctx, const uint8_t *in, size_t len)
+{
+ switch (ctx->impl) {
+#if CAN_USE_GCM_ASM >= 2
+ case GCM_IMPL_AVX2:
+ gcm_ghash_vpclmulqdq_avx2(ctx->gcm_ghash,
+ (const uint64_t *)ctx->gcm_Htable, in, len);
+ break;
+#endif
+
+ case GCM_IMPL_AVX:
+ gcm_ghash_avx(ctx->gcm_ghash,
+ (const uint64_t *)ctx->gcm_Htable, in, len);
+ break;
+
+ default:
+ VERIFY(B_FALSE);
+ }
+}
+typedef size_t ASMABI aesni_gcm_encrypt_impl(const uint8_t *, uint8_t *,
+ size_t, const void *, uint64_t *, const uint64_t *Htable, uint64_t *);
extern size_t ASMABI aesni_gcm_encrypt(const uint8_t *, uint8_t *, size_t,
const void *, uint64_t *, uint64_t *);
+#if CAN_USE_GCM_ASM >= 2
+extern void ASMABI aes_gcm_enc_update_vaes_avx2(const uint8_t *in,
+ uint8_t *out, size_t len, const void *key, const uint8_t ivec[16],
+ const uint128_t Htable[16], uint8_t Xi[16]);
+#endif
+typedef size_t ASMABI aesni_gcm_decrypt_impl(const uint8_t *, uint8_t *,
+ size_t, const void *, uint64_t *, const uint64_t *Htable, uint64_t *);
extern size_t ASMABI aesni_gcm_decrypt(const uint8_t *, uint8_t *, size_t,
const void *, uint64_t *, uint64_t *);
+#if CAN_USE_GCM_ASM >= 2
+extern void ASMABI aes_gcm_dec_update_vaes_avx2(const uint8_t *in,
+ uint8_t *out, size_t len, const void *key, const uint8_t ivec[16],
+ const uint128_t Htable[16], uint8_t Xi[16]);
+#endif
+
+static inline boolean_t
+gcm_avx2_will_work(void)
+{
+ return (kfpu_allowed() &&
+ zfs_avx2_available() && zfs_vaes_available() &&
+ zfs_vpclmulqdq_available());
+}
static inline boolean_t
gcm_avx_will_work(void)
@@ -1035,33 +1105,67 @@ gcm_avx_will_work(void)
}
static inline void
-gcm_set_avx(boolean_t val)
+gcm_use_impl(gcm_impl impl)
{
- if (gcm_avx_will_work() == B_TRUE) {
- atomic_swap_32(&gcm_use_avx, val);
+ switch (impl) {
+#if CAN_USE_GCM_ASM >= 2
+ case GCM_IMPL_AVX2:
+ if (gcm_avx2_will_work() == B_TRUE) {
+ atomic_swap_32(&gcm_impl_used, impl);
+ return;
+ }
+
+ zfs_fallthrough;
+#endif
+
+ case GCM_IMPL_AVX:
+ if (gcm_avx_will_work() == B_TRUE) {
+ atomic_swap_32(&gcm_impl_used, impl);
+ return;
+ }
+
+ zfs_fallthrough;
+
+ default:
+ atomic_swap_32(&gcm_impl_used, GCM_IMPL_GENERIC);
}
}
static inline boolean_t
-gcm_toggle_avx(void)
+gcm_impl_will_work(gcm_impl impl)
{
- if (gcm_avx_will_work() == B_TRUE) {
- return (atomic_toggle_boolean_nv(&GCM_IMPL_USE_AVX));
- } else {
- return (B_FALSE);
+ switch (impl) {
+#if CAN_USE_GCM_ASM >= 2
+ case GCM_IMPL_AVX2:
+ return (gcm_avx2_will_work());
+#endif
+
+ case GCM_IMPL_AVX:
+ return (gcm_avx_will_work());
+
+ default:
+ return (B_TRUE);
}
}
-static inline size_t
-gcm_simd_get_htab_size(boolean_t simd_mode)
+static inline gcm_impl
+gcm_toggle_impl(void)
{
- switch (simd_mode) {
- case B_TRUE:
- return (2 * 6 * 2 * sizeof (uint64_t));
+ gcm_impl current_impl, new_impl;
+ do { /* handle races */
+ current_impl = atomic_load_32(&gcm_impl_used);
+ new_impl = current_impl;
+ while (B_TRUE) { /* handle incompatble implementations */
+ new_impl = (new_impl + 1) % GCM_IMPL_MAX;
+ if (gcm_impl_will_work(new_impl)) {
+ break;
+ }
+ }
- default:
- return (0);
- }
+ } while (atomic_cas_32(&gcm_impl_used, current_impl, new_impl) !=
+ current_impl);
+
+ return (new_impl);
}
@@ -1077,6 +1181,50 @@ gcm_incr_counter_block_by(gcm_ctx_t *ctx, int n)
ctx->gcm_cb[1] = (ctx->gcm_cb[1] & ~counter_mask) | counter;
}
+static size_t aesni_gcm_encrypt_avx(const uint8_t *in, uint8_t *out,
+ size_t len, const void *key, uint64_t *iv, const uint64_t *Htable,
+ uint64_t *Xip)
+{
+ (void) Htable;
+ return (aesni_gcm_encrypt(in, out, len, key, iv, Xip));
+}
+
+#if CAN_USE_GCM_ASM >= 2
+// kSizeTWithoutLower4Bits is a mask that can be used to zero the lower four
+// bits of a |size_t|.
+// This is from boringssl/crypto/fipsmodule/aes/gcm.cc.inc
+static const size_t kSizeTWithoutLower4Bits = (size_t)-16;
+
+/* The following CRYPTO methods are from boringssl/crypto/internal.h */
+static inline uint32_t CRYPTO_bswap4(uint32_t x) {
+ return (__builtin_bswap32(x));
+}
+
+static inline uint32_t CRYPTO_load_u32_be(const void *in) {
+ uint32_t v;
+ memcpy(&v, in, sizeof (v));
+ return (CRYPTO_bswap4(v));
+}
+
+static inline void CRYPTO_store_u32_be(void *out, uint32_t v) {
+ v = CRYPTO_bswap4(v);
+ memcpy(out, &v, sizeof (v));
+}
+
+static size_t aesni_gcm_encrypt_avx2(const uint8_t *in, uint8_t *out,
+ size_t len, const void *key, uint64_t *iv, const uint64_t *Htable,
+ uint64_t *Xip)
+{
+ uint8_t *ivec = (uint8_t *)iv;
+ len &= kSizeTWithoutLower4Bits;
+ aes_gcm_enc_update_vaes_avx2(in, out, len, key, ivec,
+ (const uint128_t *)Htable, (uint8_t *)Xip);
+ CRYPTO_store_u32_be(&ivec[12],
+ CRYPTO_load_u32_be(&ivec[12]) + len / 16);
+ return (len);
+}
+#endif /* if CAN_USE_GCM_ASM >= 2 */
+
/*
* Encrypt multiple blocks of data in GCM mode.
* This is done in gcm_avx_chunk_size chunks, utilizing AVX assembler routines
@@ -1091,8 +1239,15 @@ gcm_mode_encrypt_contiguous_blocks_avx(gcm_ctx_t *ctx, char *data,
size_t done = 0;
uint8_t *datap = (uint8_t *)data;
size_t chunk_size = (size_t)GCM_CHUNK_SIZE_READ;
+ aesni_gcm_encrypt_impl *encrypt_blocks =
+#if CAN_USE_GCM_ASM >= 2
+ ctx->impl == GCM_IMPL_AVX2 ?
+ aesni_gcm_encrypt_avx2 :
+#endif
+ aesni_gcm_encrypt_avx;
const aes_key_t *key = ((aes_key_t *)ctx->gcm_keysched);
uint64_t *ghash = ctx->gcm_ghash;
+ uint64_t *htable = ctx->gcm_Htable;
uint64_t *cb = ctx->gcm_cb;
uint8_t *ct_buf = NULL;
uint8_t *tmp = (uint8_t *)ctx->gcm_tmp;
@@ -1156,8 +1311,8 @@ gcm_mode_encrypt_contiguous_blocks_avx(gcm_ctx_t *ctx, char *data,
/* Do the bulk encryption in chunk_size blocks. */
for (; bleft >= chunk_size; bleft -= chunk_size) {
kfpu_begin();
- done = aesni_gcm_encrypt(
- datap, ct_buf, chunk_size, key, cb, ghash);
+ done = encrypt_blocks(
+ datap, ct_buf, chunk_size, key, cb, htable, ghash);
clear_fpu_regs();
kfpu_end();
@@ -1180,7 +1335,8 @@ gcm_mode_encrypt_contiguous_blocks_avx(gcm_ctx_t *ctx, char *data,
/* Bulk encrypt the remaining data. */
kfpu_begin();
if (bleft >= GCM_AVX_MIN_ENCRYPT_BYTES) {
- done = aesni_gcm_encrypt(datap, ct_buf, bleft, key, cb, ghash);
+ done = encrypt_blocks(datap, ct_buf, bleft, key, cb, htable,
+ ghash);
if (done == 0) {
rv = CRYPTO_FAILED;
goto out;
@@ -1293,6 +1449,29 @@ gcm_encrypt_final_avx(gcm_ctx_t *ctx, crypto_data_t *out, size_t block_size)
return (CRYPTO_SUCCESS);
}
+static size_t aesni_gcm_decrypt_avx(const uint8_t *in, uint8_t *out,
+ size_t len, const void *key, uint64_t *iv, const uint64_t *Htable,
+ uint64_t *Xip)
+{
+ (void) Htable;
+ return (aesni_gcm_decrypt(in, out, len, key, iv, Xip));
+}
+
+#if CAN_USE_GCM_ASM >= 2
+static size_t aesni_gcm_decrypt_avx2(const uint8_t *in, uint8_t *out,
+ size_t len, const void *key, uint64_t *iv, const uint64_t *Htable,
+ uint64_t *Xip)
+{
+ uint8_t *ivec = (uint8_t *)iv;
+ len &= kSizeTWithoutLower4Bits;
+ aes_gcm_dec_update_vaes_avx2(in, out, len, key, ivec,
+ (const uint128_t *)Htable, (uint8_t *)Xip);
+ CRYPTO_store_u32_be(&ivec[12],
+ CRYPTO_load_u32_be(&ivec[12]) + len / 16);
+ return (len);
+}
+#endif /* if CAN_USE_GCM_ASM >= 2 */
+
/*
* Finalize decryption: We just have accumulated crypto text, so now we
* decrypt it here inplace.
@@ -1306,10 +1485,17 @@ gcm_decrypt_final_avx(gcm_ctx_t *ctx, crypto_data_t *out, size_t block_size)
B_FALSE);
size_t chunk_size = (size_t)GCM_CHUNK_SIZE_READ;
+ aesni_gcm_decrypt_impl *decrypt_blocks =
+#if CAN_USE_GCM_ASM >= 2
+ ctx->impl == GCM_IMPL_AVX2 ?
+ aesni_gcm_decrypt_avx2 :
+#endif
+ aesni_gcm_decrypt_avx;
size_t pt_len = ctx->gcm_processed_data_len - ctx->gcm_tag_len;
uint8_t *datap = ctx->gcm_pt_buf;
const aes_key_t *key = ((aes_key_t *)ctx->gcm_keysched);
uint32_t *cb = (uint32_t *)ctx->gcm_cb;
+ uint64_t *htable = ctx->gcm_Htable;
uint64_t *ghash = ctx->gcm_ghash;
uint32_t *tmp = (uint32_t *)ctx->gcm_tmp;
int rv = CRYPTO_SUCCESS;
@@ -1322,8 +1508,8 @@ gcm_decrypt_final_avx(gcm_ctx_t *ctx, crypto_data_t *out, size_t block_size)
*/
for (bleft = pt_len; bleft >= chunk_size; bleft -= chunk_size) {
kfpu_begin();
- done = aesni_gcm_decrypt(datap, datap, chunk_size,
- (const void *)key, ctx->gcm_cb, ghash);
+ done = decrypt_blocks(datap, datap, chunk_size,
+ (const void *)key, ctx->gcm_cb, htable, ghash);
clear_fpu_regs();
kfpu_end();
if (done != chunk_size) {
@@ -1334,8 +1520,8 @@ gcm_decrypt_final_avx(gcm_ctx_t *ctx, crypto_data_t *out, size_t block_size)
/* Decrypt remainder, which is less than chunk size, in one go. */
kfpu_begin();
if (bleft >= GCM_AVX_MIN_DECRYPT_BYTES) {
- done = aesni_gcm_decrypt(datap, datap, bleft,
- (const void *)key, ctx->gcm_cb, ghash);
+ done = decrypt_blocks(datap, datap, bleft,
+ (const void *)key, ctx->gcm_cb, htable, ghash);
if (done == 0) {
clear_fpu_regs();
kfpu_end();
@@ -1424,13 +1610,42 @@ gcm_init_avx(gcm_ctx_t *ctx, const uint8_t *iv, size_t iv_len,
ASSERT3S(((aes_key_t *)ctx->gcm_keysched)->ops->needs_byteswap, ==,
B_FALSE);
+ size_t htab_len = 0;
+#if CAN_USE_GCM_ASM >= 2
+ if (ctx->impl == GCM_IMPL_AVX2) {
+ /*
+ * BoringSSL's API specifies uint128_t[16] for htab; but only
+ * uint128_t[12] are used.
+ * See https://github.com/google/boringssl/blob/
+ * 813840dd094f9e9c1b00a7368aa25e656554221f1/crypto/fipsmodule/
+ * modes/asm/aes-gcm-avx2-x86_64.pl#L198-L200
+ */
+ htab_len = (2 * 8 * sizeof (uint128_t));
+ } else
+#endif /* CAN_USE_GCM_ASM >= 2 */
+ {
+ htab_len = (2 * 6 * sizeof (uint128_t));
+ }
+
+ ctx->gcm_Htable = kmem_alloc(htab_len, KM_SLEEP);
+ if (ctx->gcm_Htable == NULL) {
+ return (CRYPTO_HOST_MEMORY);
+ }
+
/* Init H (encrypt zero block) and create the initial counter block. */
memset(H, 0, sizeof (ctx->gcm_H));
kfpu_begin();
aes_encrypt_intel(keysched, aes_rounds,
(const uint32_t *)H, (uint32_t *)H);
- gcm_init_htab_avx(ctx->gcm_Htable, H);
+#if CAN_USE_GCM_ASM >= 2
+ if (ctx->impl == GCM_IMPL_AVX2) {
+ gcm_init_vpclmulqdq_avx2((uint128_t *)ctx->gcm_Htable, H);
+ } else
+#endif /* if CAN_USE_GCM_ASM >= 2 */
+ {
+ gcm_init_htab_avx(ctx->gcm_Htable, H);
+ }
if (iv_len == 12) {
memcpy(cb, iv, 12);
diff --git a/sys/contrib/openzfs/module/icp/algs/modes/modes.c b/sys/contrib/openzfs/module/icp/algs/modes/modes.c
index 343591cd9691..ef3c1806e4b6 100644
--- a/sys/contrib/openzfs/module/icp/algs/modes/modes.c
+++ b/sys/contrib/openzfs/module/icp/algs/modes/modes.c
@@ -171,7 +171,7 @@ gcm_clear_ctx(gcm_ctx_t *ctx)
explicit_memset(ctx->gcm_remainder, 0, sizeof (ctx->gcm_remainder));
explicit_memset(ctx->gcm_H, 0, sizeof (ctx->gcm_H));
#if defined(CAN_USE_GCM_ASM)
- if (ctx->gcm_use_avx == B_TRUE) {
+ if (ctx->impl != GCM_IMPL_GENERIC) {
ASSERT3P(ctx->gcm_Htable, !=, NULL);
explicit_memset(ctx->gcm_Htable, 0, ctx->gcm_htab_len);
kmem_free(ctx->gcm_Htable, ctx->gcm_htab_len);
diff --git a/sys/contrib/openzfs/module/icp/asm-x86_64/modes/THIRDPARTYLICENSE.boringssl b/sys/contrib/openzfs/module/icp/asm-x86_64/modes/THIRDPARTYLICENSE.boringssl
new file mode 100644
index 000000000000..04c03a37e0cb
--- /dev/null
+++ b/sys/contrib/openzfs/module/icp/asm-x86_64/modes/THIRDPARTYLICENSE.boringssl
@@ -0,0 +1,253 @@
+BoringSSL is a fork of OpenSSL. As such, large parts of it fall under OpenSSL
+licensing. Files that are completely new have a Google copyright and an ISC
+license. This license is reproduced at the bottom of this file.
+
+Contributors to BoringSSL are required to follow the CLA rules for Chromium:
+https://cla.developers.google.com/clas
+
+Files in third_party/ have their own licenses, as described therein. The MIT
+license, for third_party/fiat, which, unlike other third_party directories, is
+compiled into non-test libraries, is included below.
+
+The OpenSSL toolkit stays under a dual license, i.e. both the conditions of the
+OpenSSL License and the original SSLeay license apply to the toolkit. See below
+for the actual license texts. Actually both licenses are BSD-style Open Source
+licenses. In case of any license issues related to OpenSSL please contact
+openssl-core@openssl.org.
+
+The following are Google-internal bug numbers where explicit permission from
+some authors is recorded for use of their work. (This is purely for our own
+record keeping.)
+ 27287199
+ 27287880
+ 27287883
+ 263291445
+
+
+ OpenSSL License
+ ---------------
+
+/* ====================================================================
+ * Copyright (c) 1998-2011 The OpenSSL Project. All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ *
+ * 2. Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in
+ * the documentation and/or other materials provided with the
+ * distribution.
+ *
+ * 3. All advertising materials mentioning features or use of this
+ * software must display the following acknowledgment:
+ * "This product includes software developed by the OpenSSL Project
+ * for use in the OpenSSL Toolkit. (http://www.openssl.org/)"
+ *
+ * 4. The names "OpenSSL Toolkit" and "OpenSSL Project" must not be used to
+ * endorse or promote products derived from this software without
+ * prior written permission. For written permission, please contact
+ * openssl-core@openssl.org.
+ *
+ * 5. Products derived from this software may not be called "OpenSSL"
+ * nor may "OpenSSL" appear in their names without prior written
+ * permission of the OpenSSL Project.
+ *
+ * 6. Redistributions of any form whatsoever must retain the following
+ * acknowledgment:
+ * "This product includes software developed by the OpenSSL Project
+ * for use in the OpenSSL Toolkit (http://www.openssl.org/)"
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE OpenSSL PROJECT ``AS IS'' AND ANY
+ * EXPRESSED OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+ * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE OpenSSL PROJECT OR
+ * ITS CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
+ * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
+ * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
+ * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED
+ * OF THE POSSIBILITY OF SUCH DAMAGE.
+ * ====================================================================
+ *
+ * This product includes cryptographic software written by Eric Young
+ * (eay@cryptsoft.com). This product includes software written by Tim
+ * Hudson (tjh@cryptsoft.com).
+ *
+ */
+
+ Original SSLeay License
+ -----------------------
+
+/* Copyright (C) 1995-1998 Eric Young (eay@cryptsoft.com)
+ * All rights reserved.
+ *
+ * This package is an SSL implementation written
+ * by Eric Young (eay@cryptsoft.com).
+ * The implementation was written so as to conform with Netscapes SSL.
+ *
+ * This library is free for commercial and non-commercial use as long as
+ * the following conditions are aheared to. The following conditions
+ * apply to all code found in this distribution, be it the RC4, RSA,
+ * lhash, DES, etc., code; not just the SSL code. The SSL documentation
+ * included with this distribution is covered by the same copyright terms
+ * except that the holder is Tim Hudson (tjh@cryptsoft.com).
+ *
+ * Copyright remains Eric Young's, and as such any Copyright notices in
+ * the code are not to be removed.
+ * If this package is used in a product, Eric Young should be given attribution
+ * as the author of the parts of the library used.
+ * This can be in the form of a textual message at program startup or
+ * in documentation (online or textual) provided with the package.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the copyright
+ * notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in the
+ * documentation and/or other materials provided with the distribution.
+ * 3. All advertising materials mentioning features or use of this software
+ * must display the following acknowledgement:
+ * "This product includes cryptographic software written by
+ * Eric Young (eay@cryptsoft.com)"
+ * The word 'cryptographic' can be left out if the rouines from the library
+ * being used are not cryptographic related :-).
+ * 4. If you include any Windows specific code (or a derivative thereof) from
+ * the apps directory (application code) you must include an acknowledgement:
+ * "This product includes software written by Tim Hudson (tjh@cryptsoft.com)"
+ *
+ * THIS SOFTWARE IS PROVIDED BY ERIC YOUNG ``AS IS'' AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ *
+ * The licence and distribution terms for any publically available version or
+ * derivative of this code cannot be changed. i.e. this code cannot simply be
+ * copied and put under another distribution licence
+ * [including the GNU Public Licence.]
+ */
+
+
+ISC license used for completely new code in BoringSSL:
+
+/* Copyright 2015 The BoringSSL Authors
+ *
+ * Permission to use, copy, modify, and/or distribute this software for any
+ * purpose with or without fee is hereby granted, provided that the above
+ * copyright notice and this permission notice appear in all copies.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+ * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
+ * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY
+ * SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+ * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION
+ * OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN
+ * CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. */
+
+
+The code in third_party/fiat carries the MIT license:
+
+Copyright (c) 2015-2016 the fiat-crypto authors (see
+https://github.com/mit-plv/fiat-crypto/blob/master/AUTHORS).
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.
+
+
+Licenses for support code
+-------------------------
+
+Parts of the TLS test suite are under the Go license. This code is not included
+in BoringSSL (i.e. libcrypto and libssl) when compiled, however, so
+distributing code linked against BoringSSL does not trigger this license:
+
+Copyright (c) 2009 The Go Authors. All rights reserved.
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+
+ * Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+ * Redistributions in binary form must reproduce the above
+copyright notice, this list of conditions and the following disclaimer
+in the documentation and/or other materials provided with the
+distribution.
+ * Neither the name of Google Inc. nor the names of its
+contributors may be used to endorse or promote products derived from
+this software without specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+
+BoringSSL uses the Chromium test infrastructure to run a continuous build,
+trybots etc. The scripts which manage this, and the script for generating build
+metadata, are under the Chromium license. Distributing code linked against
+BoringSSL does not trigger this license.
+
+Copyright 2015 The Chromium Authors. All rights reserved.
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+
+ * Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+ * Redistributions in binary form must reproduce the above
+copyright notice, this list of conditions and the following disclaimer
+in the documentation and/or other materials provided with the
+distribution.
+ * Neither the name of Google Inc. nor the names of its
+contributors may be used to endorse or promote products derived from
+this software without specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
diff --git a/sys/contrib/openzfs/module/icp/asm-x86_64/modes/THIRDPARTYLICENSE.boringssl.descrip b/sys/contrib/openzfs/module/icp/asm-x86_64/modes/THIRDPARTYLICENSE.boringssl.descrip
new file mode 100644
index 000000000000..f63a67a4d2ae
--- /dev/null
+++ b/sys/contrib/openzfs/module/icp/asm-x86_64/modes/THIRDPARTYLICENSE.boringssl.descrip
@@ -0,0 +1 @@
+PORTIONS OF AES GCM and GHASH FUNCTIONALITY
diff --git a/sys/contrib/openzfs/module/icp/asm-x86_64/modes/aesni-gcm-avx2-vaes.S b/sys/contrib/openzfs/module/icp/asm-x86_64/modes/aesni-gcm-avx2-vaes.S
new file mode 100644
index 000000000000..3d1b045127e2
--- /dev/null
+++ b/sys/contrib/openzfs/module/icp/asm-x86_64/modes/aesni-gcm-avx2-vaes.S
@@ -0,0 +1,1323 @@
+// SPDX-License-Identifier: Apache-2.0
+// This file is generated from a similarly-named Perl script in the BoringSSL
+// source tree. Do not edit by hand.
+
+#if defined(__x86_64__) && defined(HAVE_AVX) && \
+ defined(HAVE_VAES) && defined(HAVE_VPCLMULQDQ)
+
+#define _ASM
+#include <sys/asm_linkage.h>
+
+/* Windows userland links with OpenSSL */
+#if !defined (_WIN32) || defined (_KERNEL)
+
+.section .rodata
+.balign 16
+
+
+.Lbswap_mask:
+.quad 0x08090a0b0c0d0e0f, 0x0001020304050607
+
+
+
+
+
+
+
+
+.Lgfpoly:
+.quad 1, 0xc200000000000000
+
+
+.Lgfpoly_and_internal_carrybit:
+.quad 1, 0xc200000000000001
+
+.balign 32
+
+.Lctr_pattern:
+.quad 0, 0
+.quad 1, 0
+.Linc_2blocks:
+.quad 2, 0
+.quad 2, 0
+
+ENTRY_ALIGN(gcm_init_vpclmulqdq_avx2, 32)
+.cfi_startproc
+
+ENDBR
+
+
+
+
+
+ vmovdqu (%rsi),%xmm3
+ // KCF/ICP stores H in network byte order with the hi qword first
+ // so we need to swap all bytes, not the 2 qwords.
+ vmovdqu .Lbswap_mask(%rip),%xmm4
+ vpshufb %xmm4,%xmm3,%xmm3
+
+
+
+
+
+ vpshufd $0xd3,%xmm3,%xmm0
+ vpsrad $31,%xmm0,%xmm0
+ vpaddq %xmm3,%xmm3,%xmm3
+ vpand .Lgfpoly_and_internal_carrybit(%rip),%xmm0,%xmm0
+ vpxor %xmm0,%xmm3,%xmm3
+
+ vbroadcasti128 .Lgfpoly(%rip),%ymm6
+
+
+ vpclmulqdq $0x00,%xmm3,%xmm3,%xmm0
+ vpclmulqdq $0x11,%xmm3,%xmm3,%xmm5
+ vpclmulqdq $0x01,%xmm0,%xmm6,%xmm1
+ vpshufd $0x4e,%xmm0,%xmm0
+ vpxor %xmm0,%xmm1,%xmm1
+ vpclmulqdq $0x01,%xmm1,%xmm6,%xmm0
+ vpshufd $0x4e,%xmm1,%xmm1
+ vpxor %xmm1,%xmm5,%xmm5
+ vpxor %xmm0,%xmm5,%xmm5
+
+
+
+ vinserti128 $1,%xmm3,%ymm5,%ymm3
+ vinserti128 $1,%xmm5,%ymm5,%ymm5
+
+
+ vpclmulqdq $0x00,%ymm5,%ymm3,%ymm0
+ vpclmulqdq $0x01,%ymm5,%ymm3,%ymm1
+ vpclmulqdq $0x10,%ymm5,%ymm3,%ymm2
+ vpxor %ymm2,%ymm1,%ymm1
+ vpclmulqdq $0x01,%ymm0,%ymm6,%ymm2
+ vpshufd $0x4e,%ymm0,%ymm0
+ vpxor %ymm0,%ymm1,%ymm1
+ vpxor %ymm2,%ymm1,%ymm1
+ vpclmulqdq $0x11,%ymm5,%ymm3,%ymm4
+ vpclmulqdq $0x01,%ymm1,%ymm6,%ymm0
+ vpshufd $0x4e,%ymm1,%ymm1
+ vpxor %ymm1,%ymm4,%ymm4
+ vpxor %ymm0,%ymm4,%ymm4
+
+
+
+ vmovdqu %ymm3,96(%rdi)
+ vmovdqu %ymm4,64(%rdi)
+
+
+
+ vpunpcklqdq %ymm3,%ymm4,%ymm0
+ vpunpckhqdq %ymm3,%ymm4,%ymm1
+ vpxor %ymm1,%ymm0,%ymm0
+ vmovdqu %ymm0,128+32(%rdi)
+
+
+ vpclmulqdq $0x00,%ymm5,%ymm4,%ymm0
+ vpclmulqdq $0x01,%ymm5,%ymm4,%ymm1
+ vpclmulqdq $0x10,%ymm5,%ymm4,%ymm2
+ vpxor %ymm2,%ymm1,%ymm1
+ vpclmulqdq $0x01,%ymm0,%ymm6,%ymm2
+ vpshufd $0x4e,%ymm0,%ymm0
+ vpxor %ymm0,%ymm1,%ymm1
+ vpxor %ymm2,%ymm1,%ymm1
+ vpclmulqdq $0x11,%ymm5,%ymm4,%ymm3
+ vpclmulqdq $0x01,%ymm1,%ymm6,%ymm0
+ vpshufd $0x4e,%ymm1,%ymm1
+ vpxor %ymm1,%ymm3,%ymm3
+ vpxor %ymm0,%ymm3,%ymm3
+
+ vpclmulqdq $0x00,%ymm5,%ymm3,%ymm0
+ vpclmulqdq $0x01,%ymm5,%ymm3,%ymm1
+ vpclmulqdq $0x10,%ymm5,%ymm3,%ymm2
+ vpxor %ymm2,%ymm1,%ymm1
+ vpclmulqdq $0x01,%ymm0,%ymm6,%ymm2
+ vpshufd $0x4e,%ymm0,%ymm0
+ vpxor %ymm0,%ymm1,%ymm1
+ vpxor %ymm2,%ymm1,%ymm1
+ vpclmulqdq $0x11,%ymm5,%ymm3,%ymm4
+ vpclmulqdq $0x01,%ymm1,%ymm6,%ymm0
+ vpshufd $0x4e,%ymm1,%ymm1
+ vpxor %ymm1,%ymm4,%ymm4
+ vpxor %ymm0,%ymm4,%ymm4
+
+ vmovdqu %ymm3,32(%rdi)
+ vmovdqu %ymm4,0(%rdi)
+
+
+
+ vpunpcklqdq %ymm3,%ymm4,%ymm0
+ vpunpckhqdq %ymm3,%ymm4,%ymm1
+ vpxor %ymm1,%ymm0,%ymm0
+ vmovdqu %ymm0,128(%rdi)
+
+ vzeroupper
+ RET
+
+.cfi_endproc
+SET_SIZE(gcm_init_vpclmulqdq_avx2)
+ENTRY_ALIGN(gcm_gmult_vpclmulqdq_avx2, 32)
+.cfi_startproc
+
+ENDBR
+
+
+
+ vmovdqu (%rdi),%xmm0
+ vmovdqu .Lbswap_mask(%rip),%xmm1
+ vmovdqu 128-16(%rsi),%xmm2
+ vmovdqu .Lgfpoly(%rip),%xmm3
+ vpshufb %xmm1,%xmm0,%xmm0
+
+ vpclmulqdq $0x00,%xmm2,%xmm0,%xmm4
+ vpclmulqdq $0x01,%xmm2,%xmm0,%xmm5
+ vpclmulqdq $0x10,%xmm2,%xmm0,%xmm6
+ vpxor %xmm6,%xmm5,%xmm5
+ vpclmulqdq $0x01,%xmm4,%xmm3,%xmm6
+ vpshufd $0x4e,%xmm4,%xmm4
+ vpxor %xmm4,%xmm5,%xmm5
+ vpxor %xmm6,%xmm5,%xmm5
+ vpclmulqdq $0x11,%xmm2,%xmm0,%xmm0
+ vpclmulqdq $0x01,%xmm5,%xmm3,%xmm4
+ vpshufd $0x4e,%xmm5,%xmm5
+ vpxor %xmm5,%xmm0,%xmm0
+ vpxor %xmm4,%xmm0,%xmm0
+
+
+ vpshufb %xmm1,%xmm0,%xmm0
+ vmovdqu %xmm0,(%rdi)
+
+
+ RET
+
+.cfi_endproc
+SET_SIZE(gcm_gmult_vpclmulqdq_avx2)
+ENTRY_ALIGN(gcm_ghash_vpclmulqdq_avx2, 32)
+.cfi_startproc
+
+ENDBR
+
+
+
+
+
+
+ vmovdqu .Lbswap_mask(%rip),%xmm6
+ vmovdqu .Lgfpoly(%rip),%xmm7
+
+
+ vmovdqu (%rdi),%xmm5
+ vpshufb %xmm6,%xmm5,%xmm5
+
+
+ cmpq $32,%rcx
+ jb .Lghash_lastblock
+
+
+
+ vinserti128 $1,%xmm6,%ymm6,%ymm6
+ vinserti128 $1,%xmm7,%ymm7,%ymm7
+
+ cmpq $127,%rcx
+ jbe .Lghash_loop_1x
+
+
+ vmovdqu 128(%rsi),%ymm8
+ vmovdqu 128+32(%rsi),%ymm9
+.Lghash_loop_4x:
+
+ vmovdqu 0(%rdx),%ymm1
+ vpshufb %ymm6,%ymm1,%ymm1
+ vmovdqu 0(%rsi),%ymm2
+ vpxor %ymm5,%ymm1,%ymm1
+ vpclmulqdq $0x00,%ymm2,%ymm1,%ymm3
+ vpclmulqdq $0x11,%ymm2,%ymm1,%ymm5
+ vpunpckhqdq %ymm1,%ymm1,%ymm0
+ vpxor %ymm1,%ymm0,%ymm0
+ vpclmulqdq $0x00,%ymm8,%ymm0,%ymm4
+
+ vmovdqu 32(%rdx),%ymm1
+ vpshufb %ymm6,%ymm1,%ymm1
+ vmovdqu 32(%rsi),%ymm2
+ vpclmulqdq $0x00,%ymm2,%ymm1,%ymm0
+ vpxor %ymm0,%ymm3,%ymm3
+ vpclmulqdq $0x11,%ymm2,%ymm1,%ymm0
+ vpxor %ymm0,%ymm5,%ymm5
+ vpunpckhqdq %ymm1,%ymm1,%ymm0
+ vpxor %ymm1,%ymm0,%ymm0
+ vpclmulqdq $0x10,%ymm8,%ymm0,%ymm0
+ vpxor %ymm0,%ymm4,%ymm4
+
+ vmovdqu 64(%rdx),%ymm1
+ vpshufb %ymm6,%ymm1,%ymm1
+ vmovdqu 64(%rsi),%ymm2
+ vpclmulqdq $0x00,%ymm2,%ymm1,%ymm0
+ vpxor %ymm0,%ymm3,%ymm3
+ vpclmulqdq $0x11,%ymm2,%ymm1,%ymm0
+ vpxor %ymm0,%ymm5,%ymm5
+ vpunpckhqdq %ymm1,%ymm1,%ymm0
+ vpxor %ymm1,%ymm0,%ymm0
+ vpclmulqdq $0x00,%ymm9,%ymm0,%ymm0
+ vpxor %ymm0,%ymm4,%ymm4
+
+
+ vmovdqu 96(%rdx),%ymm1
+ vpshufb %ymm6,%ymm1,%ymm1
+ vmovdqu 96(%rsi),%ymm2
+ vpclmulqdq $0x00,%ymm2,%ymm1,%ymm0
+ vpxor %ymm0,%ymm3,%ymm3
+ vpclmulqdq $0x11,%ymm2,%ymm1,%ymm0
+ vpxor %ymm0,%ymm5,%ymm5
+ vpunpckhqdq %ymm1,%ymm1,%ymm0
+ vpxor %ymm1,%ymm0,%ymm0
+ vpclmulqdq $0x10,%ymm9,%ymm0,%ymm0
+ vpxor %ymm0,%ymm4,%ymm4
+
+ vpxor %ymm3,%ymm4,%ymm4
+ vpxor %ymm5,%ymm4,%ymm4
+
+
+ vbroadcasti128 .Lgfpoly(%rip),%ymm2
+ vpclmulqdq $0x01,%ymm3,%ymm2,%ymm0
+ vpshufd $0x4e,%ymm3,%ymm3
+ vpxor %ymm3,%ymm4,%ymm4
+ vpxor %ymm0,%ymm4,%ymm4
+
+ vpclmulqdq $0x01,%ymm4,%ymm2,%ymm0
+ vpshufd $0x4e,%ymm4,%ymm4
+ vpxor %ymm4,%ymm5,%ymm5
+ vpxor %ymm0,%ymm5,%ymm5
+ vextracti128 $1,%ymm5,%xmm0
+ vpxor %xmm0,%xmm5,%xmm5
+
+ subq $-128,%rdx
+ addq $-128,%rcx
+ cmpq $127,%rcx
+ ja .Lghash_loop_4x
+
+
+ cmpq $32,%rcx
+ jb .Lghash_loop_1x_done
+.Lghash_loop_1x:
+ vmovdqu (%rdx),%ymm0
+ vpshufb %ymm6,%ymm0,%ymm0
+ vpxor %ymm0,%ymm5,%ymm5
+ vmovdqu 128-32(%rsi),%ymm0
+ vpclmulqdq $0x00,%ymm0,%ymm5,%ymm1
+ vpclmulqdq $0x01,%ymm0,%ymm5,%ymm2
+ vpclmulqdq $0x10,%ymm0,%ymm5,%ymm3
+ vpxor %ymm3,%ymm2,%ymm2
+ vpclmulqdq $0x01,%ymm1,%ymm7,%ymm3
+ vpshufd $0x4e,%ymm1,%ymm1
+ vpxor %ymm1,%ymm2,%ymm2
+ vpxor %ymm3,%ymm2,%ymm2
+ vpclmulqdq $0x11,%ymm0,%ymm5,%ymm5
+ vpclmulqdq $0x01,%ymm2,%ymm7,%ymm1
+ vpshufd $0x4e,%ymm2,%ymm2
+ vpxor %ymm2,%ymm5,%ymm5
+ vpxor %ymm1,%ymm5,%ymm5
+
+ vextracti128 $1,%ymm5,%xmm0
+ vpxor %xmm0,%xmm5,%xmm5
+ addq $32,%rdx
+ subq $32,%rcx
+ cmpq $32,%rcx
+ jae .Lghash_loop_1x
+.Lghash_loop_1x_done:
+
+
+.Lghash_lastblock:
+ testq %rcx,%rcx
+ jz .Lghash_done
+ vmovdqu (%rdx),%xmm0
+ vpshufb %xmm6,%xmm0,%xmm0
+ vpxor %xmm0,%xmm5,%xmm5
+ vmovdqu 128-16(%rsi),%xmm0
+ vpclmulqdq $0x00,%xmm0,%xmm5,%xmm1
+ vpclmulqdq $0x01,%xmm0,%xmm5,%xmm2
+ vpclmulqdq $0x10,%xmm0,%xmm5,%xmm3
+ vpxor %xmm3,%xmm2,%xmm2
+ vpclmulqdq $0x01,%xmm1,%xmm7,%xmm3
+ vpshufd $0x4e,%xmm1,%xmm1
+ vpxor %xmm1,%xmm2,%xmm2
+ vpxor %xmm3,%xmm2,%xmm2
+ vpclmulqdq $0x11,%xmm0,%xmm5,%xmm5
+ vpclmulqdq $0x01,%xmm2,%xmm7,%xmm1
+ vpshufd $0x4e,%xmm2,%xmm2
+ vpxor %xmm2,%xmm5,%xmm5
+ vpxor %xmm1,%xmm5,%xmm5
+
+
+.Lghash_done:
+
+ vpshufb %xmm6,%xmm5,%xmm5
+ vmovdqu %xmm5,(%rdi)
+
+ vzeroupper
+ RET
+
+.cfi_endproc
+SET_SIZE(gcm_ghash_vpclmulqdq_avx2)
+ENTRY_ALIGN(aes_gcm_enc_update_vaes_avx2, 32)
+.cfi_startproc
+
+ENDBR
+ pushq %r12
+.cfi_adjust_cfa_offset 8
+.cfi_offset %r12,-16
+
+ movq 16(%rsp),%r12
+#ifdef BORINGSSL_DISPATCH_TEST
+.extern BORINGSSL_function_hit
+.hidden BORINGSSL_function_hit
+ movb $1,BORINGSSL_function_hit+6(%rip)
+#endif
+ vbroadcasti128 .Lbswap_mask(%rip),%ymm0
+
+
+
+ vmovdqu (%r12),%xmm1
+ vpshufb %xmm0,%xmm1,%xmm1
+ vbroadcasti128 (%r8),%ymm11
+ vpshufb %ymm0,%ymm11,%ymm11
+
+
+
+ movl 504(%rcx),%r10d // ICP has a larger offset for rounds.
+ leal -24(,%r10,4),%r10d // ICP uses 10,12,14 not 9,11,13 for rounds.
+
+
+
+
+ leaq 96(%rcx,%r10,4),%r11
+ vbroadcasti128 (%rcx),%ymm9
+ vbroadcasti128 (%r11),%ymm10
+
+
+ vpaddd .Lctr_pattern(%rip),%ymm11,%ymm11
+
+
+
+ cmpq $127,%rdx
+ jbe .Lcrypt_loop_4x_done__func1
+
+ vmovdqu 128(%r9),%ymm7
+ vmovdqu 128+32(%r9),%ymm8
+
+
+
+ vmovdqu .Linc_2blocks(%rip),%ymm2
+ vpshufb %ymm0,%ymm11,%ymm12
+ vpaddd %ymm2,%ymm11,%ymm11
+ vpshufb %ymm0,%ymm11,%ymm13
+ vpaddd %ymm2,%ymm11,%ymm11
+ vpshufb %ymm0,%ymm11,%ymm14
+ vpaddd %ymm2,%ymm11,%ymm11
+ vpshufb %ymm0,%ymm11,%ymm15
+ vpaddd %ymm2,%ymm11,%ymm11
+
+
+ vpxor %ymm9,%ymm12,%ymm12
+ vpxor %ymm9,%ymm13,%ymm13
+ vpxor %ymm9,%ymm14,%ymm14
+ vpxor %ymm9,%ymm15,%ymm15
+
+ leaq 16(%rcx),%rax
+.Lvaesenc_loop_first_4_vecs__func1:
+ vbroadcasti128 (%rax),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+ addq $16,%rax
+ cmpq %rax,%r11
+ jne .Lvaesenc_loop_first_4_vecs__func1
+ vpxor 0(%rdi),%ymm10,%ymm2
+ vpxor 32(%rdi),%ymm10,%ymm3
+ vpxor 64(%rdi),%ymm10,%ymm5
+ vpxor 96(%rdi),%ymm10,%ymm6
+ vaesenclast %ymm2,%ymm12,%ymm12
+ vaesenclast %ymm3,%ymm13,%ymm13
+ vaesenclast %ymm5,%ymm14,%ymm14
+ vaesenclast %ymm6,%ymm15,%ymm15
+ vmovdqu %ymm12,0(%rsi)
+ vmovdqu %ymm13,32(%rsi)
+ vmovdqu %ymm14,64(%rsi)
+ vmovdqu %ymm15,96(%rsi)
+
+ subq $-128,%rdi
+ addq $-128,%rdx
+ cmpq $127,%rdx
+ jbe .Lghash_last_ciphertext_4x__func1
+.balign 16
+.Lcrypt_loop_4x__func1:
+
+
+
+
+ vmovdqu .Linc_2blocks(%rip),%ymm2
+ vpshufb %ymm0,%ymm11,%ymm12
+ vpaddd %ymm2,%ymm11,%ymm11
+ vpshufb %ymm0,%ymm11,%ymm13
+ vpaddd %ymm2,%ymm11,%ymm11
+ vpshufb %ymm0,%ymm11,%ymm14
+ vpaddd %ymm2,%ymm11,%ymm11
+ vpshufb %ymm0,%ymm11,%ymm15
+ vpaddd %ymm2,%ymm11,%ymm11
+
+
+ vpxor %ymm9,%ymm12,%ymm12
+ vpxor %ymm9,%ymm13,%ymm13
+ vpxor %ymm9,%ymm14,%ymm14
+ vpxor %ymm9,%ymm15,%ymm15
+
+ cmpl $24,%r10d
+ jl .Laes128__func1
+ je .Laes192__func1
+
+ vbroadcasti128 -208(%r11),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+ vbroadcasti128 -192(%r11),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+.Laes192__func1:
+ vbroadcasti128 -176(%r11),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+ vbroadcasti128 -160(%r11),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+.Laes128__func1:
+ prefetcht0 512(%rdi)
+ prefetcht0 512+64(%rdi)
+
+ vmovdqu 0(%rsi),%ymm3
+ vpshufb %ymm0,%ymm3,%ymm3
+ vmovdqu 0(%r9),%ymm4
+ vpxor %ymm1,%ymm3,%ymm3
+ vpclmulqdq $0x00,%ymm4,%ymm3,%ymm5
+ vpclmulqdq $0x11,%ymm4,%ymm3,%ymm1
+ vpunpckhqdq %ymm3,%ymm3,%ymm2
+ vpxor %ymm3,%ymm2,%ymm2
+ vpclmulqdq $0x00,%ymm7,%ymm2,%ymm6
+
+ vbroadcasti128 -144(%r11),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+
+ vbroadcasti128 -128(%r11),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+
+ vmovdqu 32(%rsi),%ymm3
+ vpshufb %ymm0,%ymm3,%ymm3
+ vmovdqu 32(%r9),%ymm4
+ vpclmulqdq $0x00,%ymm4,%ymm3,%ymm2
+ vpxor %ymm2,%ymm5,%ymm5
+ vpclmulqdq $0x11,%ymm4,%ymm3,%ymm2
+ vpxor %ymm2,%ymm1,%ymm1
+ vpunpckhqdq %ymm3,%ymm3,%ymm2
+ vpxor %ymm3,%ymm2,%ymm2
+ vpclmulqdq $0x10,%ymm7,%ymm2,%ymm2
+ vpxor %ymm2,%ymm6,%ymm6
+
+ vbroadcasti128 -112(%r11),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+
+ vmovdqu 64(%rsi),%ymm3
+ vpshufb %ymm0,%ymm3,%ymm3
+ vmovdqu 64(%r9),%ymm4
+
+ vbroadcasti128 -96(%r11),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+ vpclmulqdq $0x00,%ymm4,%ymm3,%ymm2
+ vpxor %ymm2,%ymm5,%ymm5
+ vpclmulqdq $0x11,%ymm4,%ymm3,%ymm2
+ vpxor %ymm2,%ymm1,%ymm1
+
+ vbroadcasti128 -80(%r11),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+ vpunpckhqdq %ymm3,%ymm3,%ymm2
+ vpxor %ymm3,%ymm2,%ymm2
+ vpclmulqdq $0x00,%ymm8,%ymm2,%ymm2
+ vpxor %ymm2,%ymm6,%ymm6
+
+
+ vmovdqu 96(%rsi),%ymm3
+ vpshufb %ymm0,%ymm3,%ymm3
+
+ vbroadcasti128 -64(%r11),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+ vmovdqu 96(%r9),%ymm4
+ vpclmulqdq $0x00,%ymm4,%ymm3,%ymm2
+ vpxor %ymm2,%ymm5,%ymm5
+ vpclmulqdq $0x11,%ymm4,%ymm3,%ymm2
+ vpxor %ymm2,%ymm1,%ymm1
+ vpunpckhqdq %ymm3,%ymm3,%ymm2
+ vpxor %ymm3,%ymm2,%ymm2
+ vpclmulqdq $0x10,%ymm8,%ymm2,%ymm2
+ vpxor %ymm2,%ymm6,%ymm6
+
+ vbroadcasti128 -48(%r11),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+
+ vpxor %ymm5,%ymm6,%ymm6
+ vpxor %ymm1,%ymm6,%ymm6
+
+
+ vbroadcasti128 .Lgfpoly(%rip),%ymm4
+ vpclmulqdq $0x01,%ymm5,%ymm4,%ymm2
+ vpshufd $0x4e,%ymm5,%ymm5
+ vpxor %ymm5,%ymm6,%ymm6
+ vpxor %ymm2,%ymm6,%ymm6
+
+ vbroadcasti128 -32(%r11),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+
+ vpclmulqdq $0x01,%ymm6,%ymm4,%ymm2
+ vpshufd $0x4e,%ymm6,%ymm6
+ vpxor %ymm6,%ymm1,%ymm1
+ vpxor %ymm2,%ymm1,%ymm1
+
+ vbroadcasti128 -16(%r11),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+ vextracti128 $1,%ymm1,%xmm2
+ vpxor %xmm2,%xmm1,%xmm1
+
+
+ subq $-128,%rsi
+ vpxor 0(%rdi),%ymm10,%ymm2
+ vpxor 32(%rdi),%ymm10,%ymm3
+ vpxor 64(%rdi),%ymm10,%ymm5
+ vpxor 96(%rdi),%ymm10,%ymm6
+ vaesenclast %ymm2,%ymm12,%ymm12
+ vaesenclast %ymm3,%ymm13,%ymm13
+ vaesenclast %ymm5,%ymm14,%ymm14
+ vaesenclast %ymm6,%ymm15,%ymm15
+ vmovdqu %ymm12,0(%rsi)
+ vmovdqu %ymm13,32(%rsi)
+ vmovdqu %ymm14,64(%rsi)
+ vmovdqu %ymm15,96(%rsi)
+
+ subq $-128,%rdi
+
+ addq $-128,%rdx
+ cmpq $127,%rdx
+ ja .Lcrypt_loop_4x__func1
+.Lghash_last_ciphertext_4x__func1:
+
+ vmovdqu 0(%rsi),%ymm3
+ vpshufb %ymm0,%ymm3,%ymm3
+ vmovdqu 0(%r9),%ymm4
+ vpxor %ymm1,%ymm3,%ymm3
+ vpclmulqdq $0x00,%ymm4,%ymm3,%ymm5
+ vpclmulqdq $0x11,%ymm4,%ymm3,%ymm1
+ vpunpckhqdq %ymm3,%ymm3,%ymm2
+ vpxor %ymm3,%ymm2,%ymm2
+ vpclmulqdq $0x00,%ymm7,%ymm2,%ymm6
+
+ vmovdqu 32(%rsi),%ymm3
+ vpshufb %ymm0,%ymm3,%ymm3
+ vmovdqu 32(%r9),%ymm4
+ vpclmulqdq $0x00,%ymm4,%ymm3,%ymm2
+ vpxor %ymm2,%ymm5,%ymm5
+ vpclmulqdq $0x11,%ymm4,%ymm3,%ymm2
+ vpxor %ymm2,%ymm1,%ymm1
+ vpunpckhqdq %ymm3,%ymm3,%ymm2
+ vpxor %ymm3,%ymm2,%ymm2
+ vpclmulqdq $0x10,%ymm7,%ymm2,%ymm2
+ vpxor %ymm2,%ymm6,%ymm6
+
+ vmovdqu 64(%rsi),%ymm3
+ vpshufb %ymm0,%ymm3,%ymm3
+ vmovdqu 64(%r9),%ymm4
+ vpclmulqdq $0x00,%ymm4,%ymm3,%ymm2
+ vpxor %ymm2,%ymm5,%ymm5
+ vpclmulqdq $0x11,%ymm4,%ymm3,%ymm2
+ vpxor %ymm2,%ymm1,%ymm1
+ vpunpckhqdq %ymm3,%ymm3,%ymm2
+ vpxor %ymm3,%ymm2,%ymm2
+ vpclmulqdq $0x00,%ymm8,%ymm2,%ymm2
+ vpxor %ymm2,%ymm6,%ymm6
+
+
+ vmovdqu 96(%rsi),%ymm3
+ vpshufb %ymm0,%ymm3,%ymm3
+ vmovdqu 96(%r9),%ymm4
+ vpclmulqdq $0x00,%ymm4,%ymm3,%ymm2
+ vpxor %ymm2,%ymm5,%ymm5
+ vpclmulqdq $0x11,%ymm4,%ymm3,%ymm2
+ vpxor %ymm2,%ymm1,%ymm1
+ vpunpckhqdq %ymm3,%ymm3,%ymm2
+ vpxor %ymm3,%ymm2,%ymm2
+ vpclmulqdq $0x10,%ymm8,%ymm2,%ymm2
+ vpxor %ymm2,%ymm6,%ymm6
+
+ vpxor %ymm5,%ymm6,%ymm6
+ vpxor %ymm1,%ymm6,%ymm6
+
+
+ vbroadcasti128 .Lgfpoly(%rip),%ymm4
+ vpclmulqdq $0x01,%ymm5,%ymm4,%ymm2
+ vpshufd $0x4e,%ymm5,%ymm5
+ vpxor %ymm5,%ymm6,%ymm6
+ vpxor %ymm2,%ymm6,%ymm6
+
+ vpclmulqdq $0x01,%ymm6,%ymm4,%ymm2
+ vpshufd $0x4e,%ymm6,%ymm6
+ vpxor %ymm6,%ymm1,%ymm1
+ vpxor %ymm2,%ymm1,%ymm1
+ vextracti128 $1,%ymm1,%xmm2
+ vpxor %xmm2,%xmm1,%xmm1
+
+ subq $-128,%rsi
+.Lcrypt_loop_4x_done__func1:
+
+ testq %rdx,%rdx
+ jz .Ldone__func1
+
+
+
+
+
+ leaq 128(%r9),%r8
+ subq %rdx,%r8
+
+
+ vpxor %xmm5,%xmm5,%xmm5
+ vpxor %xmm6,%xmm6,%xmm6
+ vpxor %xmm7,%xmm7,%xmm7
+
+ cmpq $64,%rdx
+ jb .Llessthan64bytes__func1
+
+
+ vpshufb %ymm0,%ymm11,%ymm12
+ vpaddd .Linc_2blocks(%rip),%ymm11,%ymm11
+ vpshufb %ymm0,%ymm11,%ymm13
+ vpaddd .Linc_2blocks(%rip),%ymm11,%ymm11
+ vpxor %ymm9,%ymm12,%ymm12
+ vpxor %ymm9,%ymm13,%ymm13
+ leaq 16(%rcx),%rax
+.Lvaesenc_loop_tail_1__func1:
+ vbroadcasti128 (%rax),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ addq $16,%rax
+ cmpq %rax,%r11
+ jne .Lvaesenc_loop_tail_1__func1
+ vaesenclast %ymm10,%ymm12,%ymm12
+ vaesenclast %ymm10,%ymm13,%ymm13
+
+
+ vmovdqu 0(%rdi),%ymm2
+ vmovdqu 32(%rdi),%ymm3
+ vpxor %ymm2,%ymm12,%ymm12
+ vpxor %ymm3,%ymm13,%ymm13
+ vmovdqu %ymm12,0(%rsi)
+ vmovdqu %ymm13,32(%rsi)
+
+
+ vpshufb %ymm0,%ymm12,%ymm12
+ vpshufb %ymm0,%ymm13,%ymm13
+ vpxor %ymm1,%ymm12,%ymm12
+ vmovdqu (%r8),%ymm2
+ vmovdqu 32(%r8),%ymm3
+ vpclmulqdq $0x00,%ymm2,%ymm12,%ymm5
+ vpclmulqdq $0x01,%ymm2,%ymm12,%ymm6
+ vpclmulqdq $0x10,%ymm2,%ymm12,%ymm4
+ vpxor %ymm4,%ymm6,%ymm6
+ vpclmulqdq $0x11,%ymm2,%ymm12,%ymm7
+ vpclmulqdq $0x00,%ymm3,%ymm13,%ymm4
+ vpxor %ymm4,%ymm5,%ymm5
+ vpclmulqdq $0x01,%ymm3,%ymm13,%ymm4
+ vpxor %ymm4,%ymm6,%ymm6
+ vpclmulqdq $0x10,%ymm3,%ymm13,%ymm4
+ vpxor %ymm4,%ymm6,%ymm6
+ vpclmulqdq $0x11,%ymm3,%ymm13,%ymm4
+ vpxor %ymm4,%ymm7,%ymm7
+
+ addq $64,%r8
+ addq $64,%rdi
+ addq $64,%rsi
+ subq $64,%rdx
+ jz .Lreduce__func1
+
+ vpxor %xmm1,%xmm1,%xmm1
+
+
+.Llessthan64bytes__func1:
+ vpshufb %ymm0,%ymm11,%ymm12
+ vpaddd .Linc_2blocks(%rip),%ymm11,%ymm11
+ vpshufb %ymm0,%ymm11,%ymm13
+ vpxor %ymm9,%ymm12,%ymm12
+ vpxor %ymm9,%ymm13,%ymm13
+ leaq 16(%rcx),%rax
+.Lvaesenc_loop_tail_2__func1:
+ vbroadcasti128 (%rax),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ addq $16,%rax
+ cmpq %rax,%r11
+ jne .Lvaesenc_loop_tail_2__func1
+ vaesenclast %ymm10,%ymm12,%ymm12
+ vaesenclast %ymm10,%ymm13,%ymm13
+
+
+
+
+ cmpq $32,%rdx
+ jb .Lxor_one_block__func1
+ je .Lxor_two_blocks__func1
+
+.Lxor_three_blocks__func1:
+ vmovdqu 0(%rdi),%ymm2
+ vmovdqu 32(%rdi),%xmm3
+ vpxor %ymm2,%ymm12,%ymm12
+ vpxor %xmm3,%xmm13,%xmm13
+ vmovdqu %ymm12,0(%rsi)
+ vmovdqu %xmm13,32(%rsi)
+
+ vpshufb %ymm0,%ymm12,%ymm12
+ vpshufb %xmm0,%xmm13,%xmm13
+ vpxor %ymm1,%ymm12,%ymm12
+ vmovdqu (%r8),%ymm2
+ vmovdqu 32(%r8),%xmm3
+ vpclmulqdq $0x00,%xmm3,%xmm13,%xmm4
+ vpxor %ymm4,%ymm5,%ymm5
+ vpclmulqdq $0x01,%xmm3,%xmm13,%xmm4
+ vpxor %ymm4,%ymm6,%ymm6
+ vpclmulqdq $0x10,%xmm3,%xmm13,%xmm4
+ vpxor %ymm4,%ymm6,%ymm6
+ vpclmulqdq $0x11,%xmm3,%xmm13,%xmm4
+ vpxor %ymm4,%ymm7,%ymm7
+ jmp .Lghash_mul_one_vec_unreduced__func1
+
+.Lxor_two_blocks__func1:
+ vmovdqu (%rdi),%ymm2
+ vpxor %ymm2,%ymm12,%ymm12
+ vmovdqu %ymm12,(%rsi)
+ vpshufb %ymm0,%ymm12,%ymm12
+ vpxor %ymm1,%ymm12,%ymm12
+ vmovdqu (%r8),%ymm2
+ jmp .Lghash_mul_one_vec_unreduced__func1
+
+.Lxor_one_block__func1:
+ vmovdqu (%rdi),%xmm2
+ vpxor %xmm2,%xmm12,%xmm12
+ vmovdqu %xmm12,(%rsi)
+ vpshufb %xmm0,%xmm12,%xmm12
+ vpxor %xmm1,%xmm12,%xmm12
+ vmovdqu (%r8),%xmm2
+
+.Lghash_mul_one_vec_unreduced__func1:
+ vpclmulqdq $0x00,%ymm2,%ymm12,%ymm4
+ vpxor %ymm4,%ymm5,%ymm5
+ vpclmulqdq $0x01,%ymm2,%ymm12,%ymm4
+ vpxor %ymm4,%ymm6,%ymm6
+ vpclmulqdq $0x10,%ymm2,%ymm12,%ymm4
+ vpxor %ymm4,%ymm6,%ymm6
+ vpclmulqdq $0x11,%ymm2,%ymm12,%ymm4
+ vpxor %ymm4,%ymm7,%ymm7
+
+.Lreduce__func1:
+
+ vbroadcasti128 .Lgfpoly(%rip),%ymm2
+ vpclmulqdq $0x01,%ymm5,%ymm2,%ymm3
+ vpshufd $0x4e,%ymm5,%ymm5
+ vpxor %ymm5,%ymm6,%ymm6
+ vpxor %ymm3,%ymm6,%ymm6
+ vpclmulqdq $0x01,%ymm6,%ymm2,%ymm3
+ vpshufd $0x4e,%ymm6,%ymm6
+ vpxor %ymm6,%ymm7,%ymm7
+ vpxor %ymm3,%ymm7,%ymm7
+ vextracti128 $1,%ymm7,%xmm1
+ vpxor %xmm7,%xmm1,%xmm1
+
+.Ldone__func1:
+
+ vpshufb %xmm0,%xmm1,%xmm1
+ vmovdqu %xmm1,(%r12)
+
+ vzeroupper
+ popq %r12
+.cfi_adjust_cfa_offset -8
+.cfi_restore %r12
+ RET
+
+.cfi_endproc
+SET_SIZE(aes_gcm_enc_update_vaes_avx2)
+ENTRY_ALIGN(aes_gcm_dec_update_vaes_avx2, 32)
+.cfi_startproc
+
+ENDBR
+ pushq %r12
+.cfi_adjust_cfa_offset 8
+.cfi_offset %r12,-16
+
+ movq 16(%rsp),%r12
+ vbroadcasti128 .Lbswap_mask(%rip),%ymm0
+
+
+
+ vmovdqu (%r12),%xmm1
+ vpshufb %xmm0,%xmm1,%xmm1
+ vbroadcasti128 (%r8),%ymm11
+ vpshufb %ymm0,%ymm11,%ymm11
+
+
+
+ movl 504(%rcx),%r10d // ICP has a larger offset for rounds.
+ leal -24(,%r10,4),%r10d // ICP uses 10,12,14 not 9,11,13 for rounds.
+
+
+
+
+ leaq 96(%rcx,%r10,4),%r11
+ vbroadcasti128 (%rcx),%ymm9
+ vbroadcasti128 (%r11),%ymm10
+
+
+ vpaddd .Lctr_pattern(%rip),%ymm11,%ymm11
+
+
+
+ cmpq $127,%rdx
+ jbe .Lcrypt_loop_4x_done__func2
+
+ vmovdqu 128(%r9),%ymm7
+ vmovdqu 128+32(%r9),%ymm8
+.balign 16
+.Lcrypt_loop_4x__func2:
+
+
+
+
+ vmovdqu .Linc_2blocks(%rip),%ymm2
+ vpshufb %ymm0,%ymm11,%ymm12
+ vpaddd %ymm2,%ymm11,%ymm11
+ vpshufb %ymm0,%ymm11,%ymm13
+ vpaddd %ymm2,%ymm11,%ymm11
+ vpshufb %ymm0,%ymm11,%ymm14
+ vpaddd %ymm2,%ymm11,%ymm11
+ vpshufb %ymm0,%ymm11,%ymm15
+ vpaddd %ymm2,%ymm11,%ymm11
+
+
+ vpxor %ymm9,%ymm12,%ymm12
+ vpxor %ymm9,%ymm13,%ymm13
+ vpxor %ymm9,%ymm14,%ymm14
+ vpxor %ymm9,%ymm15,%ymm15
+
+ cmpl $24,%r10d
+ jl .Laes128__func2
+ je .Laes192__func2
+
+ vbroadcasti128 -208(%r11),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+ vbroadcasti128 -192(%r11),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+.Laes192__func2:
+ vbroadcasti128 -176(%r11),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+ vbroadcasti128 -160(%r11),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+.Laes128__func2:
+ prefetcht0 512(%rdi)
+ prefetcht0 512+64(%rdi)
+
+ vmovdqu 0(%rdi),%ymm3
+ vpshufb %ymm0,%ymm3,%ymm3
+ vmovdqu 0(%r9),%ymm4
+ vpxor %ymm1,%ymm3,%ymm3
+ vpclmulqdq $0x00,%ymm4,%ymm3,%ymm5
+ vpclmulqdq $0x11,%ymm4,%ymm3,%ymm1
+ vpunpckhqdq %ymm3,%ymm3,%ymm2
+ vpxor %ymm3,%ymm2,%ymm2
+ vpclmulqdq $0x00,%ymm7,%ymm2,%ymm6
+
+ vbroadcasti128 -144(%r11),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+
+ vbroadcasti128 -128(%r11),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+
+ vmovdqu 32(%rdi),%ymm3
+ vpshufb %ymm0,%ymm3,%ymm3
+ vmovdqu 32(%r9),%ymm4
+ vpclmulqdq $0x00,%ymm4,%ymm3,%ymm2
+ vpxor %ymm2,%ymm5,%ymm5
+ vpclmulqdq $0x11,%ymm4,%ymm3,%ymm2
+ vpxor %ymm2,%ymm1,%ymm1
+ vpunpckhqdq %ymm3,%ymm3,%ymm2
+ vpxor %ymm3,%ymm2,%ymm2
+ vpclmulqdq $0x10,%ymm7,%ymm2,%ymm2
+ vpxor %ymm2,%ymm6,%ymm6
+
+ vbroadcasti128 -112(%r11),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+
+ vmovdqu 64(%rdi),%ymm3
+ vpshufb %ymm0,%ymm3,%ymm3
+ vmovdqu 64(%r9),%ymm4
+
+ vbroadcasti128 -96(%r11),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+ vpclmulqdq $0x00,%ymm4,%ymm3,%ymm2
+ vpxor %ymm2,%ymm5,%ymm5
+ vpclmulqdq $0x11,%ymm4,%ymm3,%ymm2
+ vpxor %ymm2,%ymm1,%ymm1
+
+ vbroadcasti128 -80(%r11),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+ vpunpckhqdq %ymm3,%ymm3,%ymm2
+ vpxor %ymm3,%ymm2,%ymm2
+ vpclmulqdq $0x00,%ymm8,%ymm2,%ymm2
+ vpxor %ymm2,%ymm6,%ymm6
+
+
+ vmovdqu 96(%rdi),%ymm3
+ vpshufb %ymm0,%ymm3,%ymm3
+
+ vbroadcasti128 -64(%r11),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+ vmovdqu 96(%r9),%ymm4
+ vpclmulqdq $0x00,%ymm4,%ymm3,%ymm2
+ vpxor %ymm2,%ymm5,%ymm5
+ vpclmulqdq $0x11,%ymm4,%ymm3,%ymm2
+ vpxor %ymm2,%ymm1,%ymm1
+ vpunpckhqdq %ymm3,%ymm3,%ymm2
+ vpxor %ymm3,%ymm2,%ymm2
+ vpclmulqdq $0x10,%ymm8,%ymm2,%ymm2
+ vpxor %ymm2,%ymm6,%ymm6
+
+ vbroadcasti128 -48(%r11),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+
+ vpxor %ymm5,%ymm6,%ymm6
+ vpxor %ymm1,%ymm6,%ymm6
+
+
+ vbroadcasti128 .Lgfpoly(%rip),%ymm4
+ vpclmulqdq $0x01,%ymm5,%ymm4,%ymm2
+ vpshufd $0x4e,%ymm5,%ymm5
+ vpxor %ymm5,%ymm6,%ymm6
+ vpxor %ymm2,%ymm6,%ymm6
+
+ vbroadcasti128 -32(%r11),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+
+ vpclmulqdq $0x01,%ymm6,%ymm4,%ymm2
+ vpshufd $0x4e,%ymm6,%ymm6
+ vpxor %ymm6,%ymm1,%ymm1
+ vpxor %ymm2,%ymm1,%ymm1
+
+ vbroadcasti128 -16(%r11),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ vaesenc %ymm2,%ymm14,%ymm14
+ vaesenc %ymm2,%ymm15,%ymm15
+
+ vextracti128 $1,%ymm1,%xmm2
+ vpxor %xmm2,%xmm1,%xmm1
+
+
+
+ vpxor 0(%rdi),%ymm10,%ymm2
+ vpxor 32(%rdi),%ymm10,%ymm3
+ vpxor 64(%rdi),%ymm10,%ymm5
+ vpxor 96(%rdi),%ymm10,%ymm6
+ vaesenclast %ymm2,%ymm12,%ymm12
+ vaesenclast %ymm3,%ymm13,%ymm13
+ vaesenclast %ymm5,%ymm14,%ymm14
+ vaesenclast %ymm6,%ymm15,%ymm15
+ vmovdqu %ymm12,0(%rsi)
+ vmovdqu %ymm13,32(%rsi)
+ vmovdqu %ymm14,64(%rsi)
+ vmovdqu %ymm15,96(%rsi)
+
+ subq $-128,%rdi
+ subq $-128,%rsi
+ addq $-128,%rdx
+ cmpq $127,%rdx
+ ja .Lcrypt_loop_4x__func2
+.Lcrypt_loop_4x_done__func2:
+
+ testq %rdx,%rdx
+ jz .Ldone__func2
+
+
+
+
+
+ leaq 128(%r9),%r8
+ subq %rdx,%r8
+
+
+ vpxor %xmm5,%xmm5,%xmm5
+ vpxor %xmm6,%xmm6,%xmm6
+ vpxor %xmm7,%xmm7,%xmm7
+
+ cmpq $64,%rdx
+ jb .Llessthan64bytes__func2
+
+
+ vpshufb %ymm0,%ymm11,%ymm12
+ vpaddd .Linc_2blocks(%rip),%ymm11,%ymm11
+ vpshufb %ymm0,%ymm11,%ymm13
+ vpaddd .Linc_2blocks(%rip),%ymm11,%ymm11
+ vpxor %ymm9,%ymm12,%ymm12
+ vpxor %ymm9,%ymm13,%ymm13
+ leaq 16(%rcx),%rax
+.Lvaesenc_loop_tail_1__func2:
+ vbroadcasti128 (%rax),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ addq $16,%rax
+ cmpq %rax,%r11
+ jne .Lvaesenc_loop_tail_1__func2
+ vaesenclast %ymm10,%ymm12,%ymm12
+ vaesenclast %ymm10,%ymm13,%ymm13
+
+
+ vmovdqu 0(%rdi),%ymm2
+ vmovdqu 32(%rdi),%ymm3
+ vpxor %ymm2,%ymm12,%ymm12
+ vpxor %ymm3,%ymm13,%ymm13
+ vmovdqu %ymm12,0(%rsi)
+ vmovdqu %ymm13,32(%rsi)
+
+
+ vpshufb %ymm0,%ymm2,%ymm12
+ vpshufb %ymm0,%ymm3,%ymm13
+ vpxor %ymm1,%ymm12,%ymm12
+ vmovdqu (%r8),%ymm2
+ vmovdqu 32(%r8),%ymm3
+ vpclmulqdq $0x00,%ymm2,%ymm12,%ymm5
+ vpclmulqdq $0x01,%ymm2,%ymm12,%ymm6
+ vpclmulqdq $0x10,%ymm2,%ymm12,%ymm4
+ vpxor %ymm4,%ymm6,%ymm6
+ vpclmulqdq $0x11,%ymm2,%ymm12,%ymm7
+ vpclmulqdq $0x00,%ymm3,%ymm13,%ymm4
+ vpxor %ymm4,%ymm5,%ymm5
+ vpclmulqdq $0x01,%ymm3,%ymm13,%ymm4
+ vpxor %ymm4,%ymm6,%ymm6
+ vpclmulqdq $0x10,%ymm3,%ymm13,%ymm4
+ vpxor %ymm4,%ymm6,%ymm6
+ vpclmulqdq $0x11,%ymm3,%ymm13,%ymm4
+ vpxor %ymm4,%ymm7,%ymm7
+
+ addq $64,%r8
+ addq $64,%rdi
+ addq $64,%rsi
+ subq $64,%rdx
+ jz .Lreduce__func2
+
+ vpxor %xmm1,%xmm1,%xmm1
+
+
+.Llessthan64bytes__func2:
+ vpshufb %ymm0,%ymm11,%ymm12
+ vpaddd .Linc_2blocks(%rip),%ymm11,%ymm11
+ vpshufb %ymm0,%ymm11,%ymm13
+ vpxor %ymm9,%ymm12,%ymm12
+ vpxor %ymm9,%ymm13,%ymm13
+ leaq 16(%rcx),%rax
+.Lvaesenc_loop_tail_2__func2:
+ vbroadcasti128 (%rax),%ymm2
+ vaesenc %ymm2,%ymm12,%ymm12
+ vaesenc %ymm2,%ymm13,%ymm13
+ addq $16,%rax
+ cmpq %rax,%r11
+ jne .Lvaesenc_loop_tail_2__func2
+ vaesenclast %ymm10,%ymm12,%ymm12
+ vaesenclast %ymm10,%ymm13,%ymm13
+
+
+
+
+ cmpq $32,%rdx
+ jb .Lxor_one_block__func2
+ je .Lxor_two_blocks__func2
+
+.Lxor_three_blocks__func2:
+ vmovdqu 0(%rdi),%ymm2
+ vmovdqu 32(%rdi),%xmm3
+ vpxor %ymm2,%ymm12,%ymm12
+ vpxor %xmm3,%xmm13,%xmm13
+ vmovdqu %ymm12,0(%rsi)
+ vmovdqu %xmm13,32(%rsi)
+
+ vpshufb %ymm0,%ymm2,%ymm12
+ vpshufb %xmm0,%xmm3,%xmm13
+ vpxor %ymm1,%ymm12,%ymm12
+ vmovdqu (%r8),%ymm2
+ vmovdqu 32(%r8),%xmm3
+ vpclmulqdq $0x00,%xmm3,%xmm13,%xmm4
+ vpxor %ymm4,%ymm5,%ymm5
+ vpclmulqdq $0x01,%xmm3,%xmm13,%xmm4
+ vpxor %ymm4,%ymm6,%ymm6
+ vpclmulqdq $0x10,%xmm3,%xmm13,%xmm4
+ vpxor %ymm4,%ymm6,%ymm6
+ vpclmulqdq $0x11,%xmm3,%xmm13,%xmm4
+ vpxor %ymm4,%ymm7,%ymm7
+ jmp .Lghash_mul_one_vec_unreduced__func2
+
+.Lxor_two_blocks__func2:
+ vmovdqu (%rdi),%ymm2
+ vpxor %ymm2,%ymm12,%ymm12
+ vmovdqu %ymm12,(%rsi)
+ vpshufb %ymm0,%ymm2,%ymm12
+ vpxor %ymm1,%ymm12,%ymm12
+ vmovdqu (%r8),%ymm2
+ jmp .Lghash_mul_one_vec_unreduced__func2
+
+.Lxor_one_block__func2:
+ vmovdqu (%rdi),%xmm2
+ vpxor %xmm2,%xmm12,%xmm12
+ vmovdqu %xmm12,(%rsi)
+ vpshufb %xmm0,%xmm2,%xmm12
+ vpxor %xmm1,%xmm12,%xmm12
+ vmovdqu (%r8),%xmm2
+
+.Lghash_mul_one_vec_unreduced__func2:
+ vpclmulqdq $0x00,%ymm2,%ymm12,%ymm4
+ vpxor %ymm4,%ymm5,%ymm5
+ vpclmulqdq $0x01,%ymm2,%ymm12,%ymm4
+ vpxor %ymm4,%ymm6,%ymm6
+ vpclmulqdq $0x10,%ymm2,%ymm12,%ymm4
+ vpxor %ymm4,%ymm6,%ymm6
+ vpclmulqdq $0x11,%ymm2,%ymm12,%ymm4
+ vpxor %ymm4,%ymm7,%ymm7
+
+.Lreduce__func2:
+
+ vbroadcasti128 .Lgfpoly(%rip),%ymm2
+ vpclmulqdq $0x01,%ymm5,%ymm2,%ymm3
+ vpshufd $0x4e,%ymm5,%ymm5
+ vpxor %ymm5,%ymm6,%ymm6
+ vpxor %ymm3,%ymm6,%ymm6
+ vpclmulqdq $0x01,%ymm6,%ymm2,%ymm3
+ vpshufd $0x4e,%ymm6,%ymm6
+ vpxor %ymm6,%ymm7,%ymm7
+ vpxor %ymm3,%ymm7,%ymm7
+ vextracti128 $1,%ymm7,%xmm1
+ vpxor %xmm7,%xmm1,%xmm1
+
+.Ldone__func2:
+
+ vpshufb %xmm0,%xmm1,%xmm1
+ vmovdqu %xmm1,(%r12)
+
+ vzeroupper
+ popq %r12
+.cfi_adjust_cfa_offset -8
+.cfi_restore %r12
+ RET
+
+.cfi_endproc
+SET_SIZE(aes_gcm_dec_update_vaes_avx2)
+
+#endif /* !_WIN32 || _KERNEL */
+
+/* Mark the stack non-executable. */
+#if defined(__linux__) && defined(__ELF__)
+.section .note.GNU-stack,"",%progbits
+#endif
+
+#endif /* defined(__x86_64__) && defined(HAVE_AVX) && defined(HAVE_AES) ... */
diff --git a/sys/contrib/openzfs/module/icp/include/modes/modes.h b/sys/contrib/openzfs/module/icp/include/modes/modes.h
index ca734cf4f045..de11d9eafafb 100644
--- a/sys/contrib/openzfs/module/icp/include/modes/modes.h
+++ b/sys/contrib/openzfs/module/icp/include/modes/modes.h
@@ -42,7 +42,7 @@ extern "C" {
*/
#if defined(__x86_64__) && defined(HAVE_AVX) && \
defined(HAVE_AES) && defined(HAVE_PCLMULQDQ)
-#define CAN_USE_GCM_ASM
+#define CAN_USE_GCM_ASM (HAVE_VAES && HAVE_VPCLMULQDQ ? 2 : 1)
extern boolean_t gcm_avx_can_use_movbe;
#endif
@@ -129,6 +129,15 @@ typedef struct ccm_ctx {
#define ccm_copy_to ccm_common.cc_copy_to
#define ccm_flags ccm_common.cc_flags
+#ifdef CAN_USE_GCM_ASM
+typedef enum gcm_impl {
+ GCM_IMPL_GENERIC = 0,
+ GCM_IMPL_AVX,
+ GCM_IMPL_AVX2,
+ GCM_IMPL_MAX,
+} gcm_impl;
+#endif
+
/*
* gcm_tag_len: Length of authentication tag.
*
@@ -174,7 +183,7 @@ typedef struct gcm_ctx {
uint64_t gcm_len_a_len_c[2];
uint8_t *gcm_pt_buf;
#ifdef CAN_USE_GCM_ASM
- boolean_t gcm_use_avx;
+ enum gcm_impl impl;
#endif
} gcm_ctx_t;
diff --git a/sys/contrib/openzfs/module/os/freebsd/zfs/kmod_core.c b/sys/contrib/openzfs/module/os/freebsd/zfs/kmod_core.c
index c114db14a916..b218c0da8125 100644
--- a/sys/contrib/openzfs/module/os/freebsd/zfs/kmod_core.c
+++ b/sys/contrib/openzfs/module/os/freebsd/zfs/kmod_core.c
@@ -112,7 +112,6 @@ static int zfs__fini(void);
static void zfs_shutdown(void *, int);
static eventhandler_tag zfs_shutdown_event_tag;
-static eventhandler_tag zfs_mountroot_event_tag;
#define ZFS_MIN_KSTACK_PAGES 4
@@ -311,9 +310,6 @@ zfs_modevent(module_t mod, int type, void *unused __unused)
zfs_shutdown_event_tag = EVENTHANDLER_REGISTER(
shutdown_post_sync, zfs_shutdown, NULL,
SHUTDOWN_PRI_FIRST);
- zfs_mountroot_event_tag = EVENTHANDLER_REGISTER(
- mountroot, spa_boot_init, NULL,
- SI_ORDER_ANY);
}
return (err);
case MOD_UNLOAD:
@@ -322,9 +318,6 @@ zfs_modevent(module_t mod, int type, void *unused __unused)
if (zfs_shutdown_event_tag != NULL)
EVENTHANDLER_DEREGISTER(shutdown_post_sync,
zfs_shutdown_event_tag);
- if (zfs_mountroot_event_tag != NULL)
- EVENTHANDLER_DEREGISTER(mountroot,
- zfs_mountroot_event_tag);
}
return (err);
case MOD_SHUTDOWN:
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 265dfd55fc4d..0dd2ecd7fd8d 100644
--- a/sys/contrib/openzfs/module/os/freebsd/zfs/zvol_os.c
+++ b/sys/contrib/openzfs/module/os/freebsd/zfs/zvol_os.c
@@ -31,7 +31,7 @@
* Copyright (c) 2012, 2017 by Delphix. All rights reserved.
* Copyright (c) 2013, Joyent, Inc. All rights reserved.
* Copyright (c) 2014 Integros [integros.com]
- * Copyright (c) 2024, Klara, Inc.
+ * Copyright (c) 2024, 2025, Klara, Inc.
*/
/* Portions Copyright 2011 Martin Matuska <mm@FreeBSD.org> */
@@ -196,7 +196,6 @@ DECLARE_GEOM_CLASS(zfs_zvol_class, zfs_zvol);
static int zvol_geom_open(struct g_provider *pp, int flag, int count);
static int zvol_geom_close(struct g_provider *pp, int flag, int count);
-static void zvol_geom_destroy(zvol_state_t *zv);
static int zvol_geom_access(struct g_provider *pp, int acr, int acw, int ace);
static void zvol_geom_bio_start(struct bio *bp);
static int zvol_geom_bio_getattr(struct bio *bp);
@@ -226,25 +225,14 @@ zvol_geom_open(struct g_provider *pp, int flag, int count)
}
retry:
- rw_enter(&zvol_state_lock, ZVOL_RW_READER);
- /*
- * Obtain a copy of private under zvol_state_lock to make sure either
- * the result of zvol free code setting private to NULL is observed,
- * or the zv is protected from being freed because of the positive
- * zv_open_count.
- */
- zv = pp->private;
- if (zv == NULL) {
- rw_exit(&zvol_state_lock);
- err = SET_ERROR(ENXIO);
- goto out_locked;
- }
+ zv = atomic_load_ptr(&pp->private);
+ if (zv == NULL)
+ return (SET_ERROR(ENXIO));
mutex_enter(&zv->zv_state_lock);
if (zv->zv_zso->zso_dying || zv->zv_flags & ZVOL_REMOVING) {
- rw_exit(&zvol_state_lock);
err = SET_ERROR(ENXIO);
- goto out_zv_locked;
+ goto out_locked;
}
ASSERT3S(zv->zv_volmode, ==, ZFS_VOLMODE_GEOM);
@@ -257,8 +245,24 @@ retry:
drop_suspend = B_TRUE;
if (!rw_tryenter(&zv->zv_suspend_lock, ZVOL_RW_READER)) {
mutex_exit(&zv->zv_state_lock);
+
+ /*
+ * Removal may happen while the locks are down, so
+ * we can't trust zv any longer; we have to start over.
+ */
+ zv = atomic_load_ptr(&pp->private);
+ if (zv == NULL)
+ return (SET_ERROR(ENXIO));
+
rw_enter(&zv->zv_suspend_lock, ZVOL_RW_READER);
mutex_enter(&zv->zv_state_lock);
+
+ if (zv->zv_zso->zso_dying ||
+ zv->zv_flags & ZVOL_REMOVING) {
+ err = SET_ERROR(ENXIO);
+ goto out_locked;
+ }
+
/* Check to see if zv_suspend_lock is needed. */
if (zv->zv_open_count != 0) {
rw_exit(&zv->zv_suspend_lock);
@@ -266,7 +270,6 @@ retry:
}
}
}
- rw_exit(&zvol_state_lock);
ASSERT(MUTEX_HELD(&zv->zv_state_lock));
@@ -294,7 +297,7 @@ retry:
if (drop_namespace)
mutex_exit(&spa_namespace_lock);
if (err)
- goto out_zv_locked;
+ goto out_locked;
pp->mediasize = zv->zv_volsize;
pp->stripeoffset = 0;
pp->stripesize = zv->zv_volblocksize;
@@ -329,9 +332,8 @@ out_opened:
zvol_last_close(zv);
wakeup(zv);
}
-out_zv_locked:
- mutex_exit(&zv->zv_state_lock);
out_locked:
+ mutex_exit(&zv->zv_state_lock);
if (drop_suspend)
rw_exit(&zv->zv_suspend_lock);
return (err);
@@ -345,12 +347,9 @@ zvol_geom_close(struct g_provider *pp, int flag, int count)
boolean_t drop_suspend = B_TRUE;
int new_open_count;
- rw_enter(&zvol_state_lock, ZVOL_RW_READER);
- zv = pp->private;
- if (zv == NULL) {
- rw_exit(&zvol_state_lock);
+ zv = atomic_load_ptr(&pp->private);
+ if (zv == NULL)
return (SET_ERROR(ENXIO));
- }
mutex_enter(&zv->zv_state_lock);
if (zv->zv_flags & ZVOL_EXCL) {
@@ -377,6 +376,15 @@ zvol_geom_close(struct g_provider *pp, int flag, int count)
mutex_exit(&zv->zv_state_lock);
rw_enter(&zv->zv_suspend_lock, ZVOL_RW_READER);
mutex_enter(&zv->zv_state_lock);
+
+ /*
+ * Unlike in zvol_geom_open(), we don't check if
+ * removal started here, because we might be one of the
+ * openers that needs to be thrown out! If we're the
+ * last, we need to call zvol_last_close() below to
+ * finish cleanup. So, no special treatment for us.
+ */
+
/* Check to see if zv_suspend_lock is needed. */
new_open_count = zv->zv_open_count - count;
if (new_open_count != 0) {
@@ -387,7 +395,6 @@ zvol_geom_close(struct g_provider *pp, int flag, int count)
} else {
drop_suspend = B_FALSE;
}
- rw_exit(&zvol_state_lock);
ASSERT(MUTEX_HELD(&zv->zv_state_lock));
@@ -408,20 +415,6 @@ zvol_geom_close(struct g_provider *pp, int flag, int count)
return (0);
}
-static void
-zvol_geom_destroy(zvol_state_t *zv)
-{
- struct zvol_state_geom *zsg = &zv->zv_zso->zso_geom;
- struct g_provider *pp = zsg->zsg_provider;
-
- ASSERT3S(zv->zv_volmode, ==, ZFS_VOLMODE_GEOM);
-
- g_topology_assert();
-
- zsg->zsg_provider = NULL;
- g_wither_geom(pp->geom, ENXIO);
-}
-
void
zvol_wait_close(zvol_state_t *zv)
{
@@ -454,7 +447,7 @@ zvol_geom_access(struct g_provider *pp, int acr, int acw, int ace)
("Unsupported access request to %s (acr=%d, acw=%d, ace=%d).",
pp->name, acr, acw, ace));
- if (pp->private == NULL) {
+ if (atomic_load_ptr(&pp->private) == NULL) {
if (acr <= 0 && acw <= 0 && ace <= 0)
return (0);
return (pp->error);
@@ -921,25 +914,14 @@ zvol_cdev_open(struct cdev *dev, int flags, int fmt, struct thread *td)
boolean_t drop_suspend = B_FALSE;
retry:
- rw_enter(&zvol_state_lock, ZVOL_RW_READER);
- /*
- * Obtain a copy of si_drv2 under zvol_state_lock to make sure either
- * the result of zvol free code setting si_drv2 to NULL is observed,
- * or the zv is protected from being freed because of the positive
- * zv_open_count.
- */
- zv = dev->si_drv2;
- if (zv == NULL) {
- rw_exit(&zvol_state_lock);
- err = SET_ERROR(ENXIO);
- goto out_locked;
- }
+ zv = atomic_load_ptr(&dev->si_drv2);
+ if (zv == NULL)
+ return (SET_ERROR(ENXIO));
mutex_enter(&zv->zv_state_lock);
- if (zv->zv_zso->zso_dying) {
- rw_exit(&zvol_state_lock);
+ if (zv->zv_zso->zso_dying || zv->zv_flags & ZVOL_REMOVING) {
err = SET_ERROR(ENXIO);
- goto out_zv_locked;
+ goto out_locked;
}
ASSERT3S(zv->zv_volmode, ==, ZFS_VOLMODE_DEV);
@@ -954,6 +936,13 @@ retry:
mutex_exit(&zv->zv_state_lock);
rw_enter(&zv->zv_suspend_lock, ZVOL_RW_READER);
mutex_enter(&zv->zv_state_lock);
+
+ if (unlikely(zv->zv_flags & ZVOL_REMOVING)) {
+ /* Removal started while locks were down. */
+ err = SET_ERROR(ENXIO);
+ goto out_locked;
+ }
+
/* Check to see if zv_suspend_lock is needed. */
if (zv->zv_open_count != 0) {
rw_exit(&zv->zv_suspend_lock);
@@ -961,7 +950,6 @@ retry:
}
}
}
- rw_exit(&zvol_state_lock);
ASSERT(MUTEX_HELD(&zv->zv_state_lock));
@@ -989,7 +977,7 @@ retry:
if (drop_namespace)
mutex_exit(&spa_namespace_lock);
if (err)
- goto out_zv_locked;
+ goto out_locked;
}
ASSERT(MUTEX_HELD(&zv->zv_state_lock));
@@ -1016,9 +1004,8 @@ out_opened:
zvol_last_close(zv);
wakeup(zv);
}
-out_zv_locked:
- mutex_exit(&zv->zv_state_lock);
out_locked:
+ mutex_exit(&zv->zv_state_lock);
if (drop_suspend)
rw_exit(&zv->zv_suspend_lock);
return (err);
@@ -1030,12 +1017,9 @@ zvol_cdev_close(struct cdev *dev, int flags, int fmt, struct thread *td)
zvol_state_t *zv;
boolean_t drop_suspend = B_TRUE;
- rw_enter(&zvol_state_lock, ZVOL_RW_READER);
- zv = dev->si_drv2;
- if (zv == NULL) {
- rw_exit(&zvol_state_lock);
+ zv = atomic_load_ptr(&dev->si_drv2);
+ if (zv == NULL)
return (SET_ERROR(ENXIO));
- }
mutex_enter(&zv->zv_state_lock);
if (zv->zv_flags & ZVOL_EXCL) {
@@ -1060,6 +1044,15 @@ zvol_cdev_close(struct cdev *dev, int flags, int fmt, struct thread *td)
mutex_exit(&zv->zv_state_lock);
rw_enter(&zv->zv_suspend_lock, ZVOL_RW_READER);
mutex_enter(&zv->zv_state_lock);
+
+ /*
+ * Unlike in zvol_cdev_open(), we don't check if
+ * removal started here, because we might be one of the
+ * openers that needs to be thrown out! If we're the
+ * last, we need to call zvol_last_close() below to
+ * finish cleanup. So, no special treatment for us.
+ */
+
/* Check to see if zv_suspend_lock is needed. */
if (zv->zv_open_count != 1) {
rw_exit(&zv->zv_suspend_lock);
@@ -1069,7 +1062,6 @@ zvol_cdev_close(struct cdev *dev, int flags, int fmt, struct thread *td)
} else {
drop_suspend = B_FALSE;
}
- rw_exit(&zvol_state_lock);
ASSERT(MUTEX_HELD(&zv->zv_state_lock));
@@ -1101,7 +1093,8 @@ zvol_cdev_ioctl(struct cdev *dev, ulong_t cmd, caddr_t data,
int error;
boolean_t sync;
- zv = dev->si_drv2;
+ zv = atomic_load_ptr(&dev->si_drv2);
+ ASSERT3P(zv, !=, NULL);
error = 0;
KASSERT(zv->zv_open_count > 0,
@@ -1162,6 +1155,7 @@ zvol_cdev_ioctl(struct cdev *dev, ulong_t cmd, caddr_t data,
*(off_t *)data = 0;
break;
case DIOCGATTR: {
+ rw_enter(&zv->zv_suspend_lock, ZVOL_RW_READER);
spa_t *spa = dmu_objset_spa(zv->zv_objset);
struct diocgattr_arg *arg = (struct diocgattr_arg *)data;
uint64_t refd, avail, usedobjs, availobjs;
@@ -1186,6 +1180,7 @@ zvol_cdev_ioctl(struct cdev *dev, ulong_t cmd, caddr_t data,
arg->value.off = refd / DEV_BSIZE;
} else
error = SET_ERROR(ENOIOCTL);
+ rw_exit(&zv->zv_suspend_lock);
break;
}
case FIOSEEKHOLE:
@@ -1196,10 +1191,12 @@ zvol_cdev_ioctl(struct cdev *dev, ulong_t cmd, caddr_t data,
hole = (cmd == FIOSEEKHOLE);
noff = *off;
+ rw_enter(&zv->zv_suspend_lock, ZVOL_RW_READER);
lr = zfs_rangelock_enter(&zv->zv_rangelock, 0, UINT64_MAX,
RL_READER);
error = dmu_offset_next(zv->zv_objset, ZVOL_OBJ, hole, &noff);
zfs_rangelock_exit(lr);
+ rw_exit(&zv->zv_suspend_lock);
*off = noff;
break;
}
@@ -1400,42 +1397,65 @@ zvol_alloc(const char *name, uint64_t volsize, uint64_t volblocksize,
* Remove minor node for the specified volume.
*/
void
-zvol_os_free(zvol_state_t *zv)
+zvol_os_remove_minor(zvol_state_t *zv)
{
- ASSERT(!RW_LOCK_HELD(&zv->zv_suspend_lock));
- ASSERT(!MUTEX_HELD(&zv->zv_state_lock));
+ ASSERT(MUTEX_HELD(&zv->zv_state_lock));
ASSERT0(zv->zv_open_count);
+ ASSERT0(atomic_read(&zv->zv_suspend_ref));
+ ASSERT(zv->zv_flags & ZVOL_REMOVING);
- ZFS_LOG(1, "ZVOL %s destroyed.", zv->zv_name);
-
- rw_destroy(&zv->zv_suspend_lock);
- zfs_rangelock_fini(&zv->zv_rangelock);
+ struct zvol_state_os *zso = zv->zv_zso;
+ zv->zv_zso = NULL;
if (zv->zv_volmode == ZFS_VOLMODE_GEOM) {
- struct zvol_state_geom *zsg = &zv->zv_zso->zso_geom;
- struct g_provider *pp __maybe_unused = zsg->zsg_provider;
-
- ASSERT0P(pp->private);
+ struct zvol_state_geom *zsg = &zso->zso_geom;
+ struct g_provider *pp = zsg->zsg_provider;
+ atomic_store_ptr(&pp->private, NULL);
+ mutex_exit(&zv->zv_state_lock);
g_topology_lock();
- zvol_geom_destroy(zv);
+ g_wither_geom(pp->geom, ENXIO);
g_topology_unlock();
} else if (zv->zv_volmode == ZFS_VOLMODE_DEV) {
- struct zvol_state_dev *zsd = &zv->zv_zso->zso_dev;
+ struct zvol_state_dev *zsd = &zso->zso_dev;
struct cdev *dev = zsd->zsd_cdev;
+ if (dev != NULL)
+ atomic_store_ptr(&dev->si_drv2, NULL);
+ mutex_exit(&zv->zv_state_lock);
+
if (dev != NULL) {
- ASSERT0P(dev->si_drv2);
destroy_dev(dev);
knlist_clear(&zsd->zsd_selinfo.si_note, 0);
knlist_destroy(&zsd->zsd_selinfo.si_note);
}
}
+ kmem_free(zso, sizeof (struct zvol_state_os));
+
+ mutex_enter(&zv->zv_state_lock);
+}
+
+void
+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);
+ ASSERT0P(zv->zv_zso);
+
+ ASSERT0P(zv->zv_objset);
+ ASSERT0P(zv->zv_zilog);
+ ASSERT0P(zv->zv_dn);
+
+ ZFS_LOG(1, "ZVOL %s destroyed.", zv->zv_name);
+
+ rw_destroy(&zv->zv_suspend_lock);
+ zfs_rangelock_fini(&zv->zv_rangelock);
+
mutex_destroy(&zv->zv_state_lock);
cv_destroy(&zv->zv_removing_cv);
dataset_kstats_destroy(&zv->zv_kstat);
- kmem_free(zv->zv_zso, sizeof (struct zvol_state_os));
kmem_free(zv, sizeof (zvol_state_t));
zvol_minors--;
}
@@ -1538,28 +1558,6 @@ out_doi:
return (error);
}
-void
-zvol_os_clear_private(zvol_state_t *zv)
-{
- ASSERT(RW_LOCK_HELD(&zvol_state_lock));
- if (zv->zv_volmode == ZFS_VOLMODE_GEOM) {
- struct zvol_state_geom *zsg = &zv->zv_zso->zso_geom;
- struct g_provider *pp = zsg->zsg_provider;
-
- if (pp->private == NULL) /* already cleared */
- return;
-
- pp->private = NULL;
- ASSERT(!RW_LOCK_HELD(&zv->zv_suspend_lock));
- } else if (zv->zv_volmode == ZFS_VOLMODE_DEV) {
- struct zvol_state_dev *zsd = &zv->zv_zso->zso_dev;
- struct cdev *dev = zsd->zsd_cdev;
-
- if (dev != NULL)
- dev->si_drv2 = NULL;
- }
-}
-
int
zvol_os_update_volsize(zvol_state_t *zv, uint64_t volsize)
{
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 a73acdad34ae..bac166fcd89e 100644
--- a/sys/contrib/openzfs/module/os/linux/zfs/zvol_os.c
+++ b/sys/contrib/openzfs/module/os/linux/zfs/zvol_os.c
@@ -22,7 +22,7 @@
/*
* Copyright (c) 2012, 2020 by Delphix. All rights reserved.
* Copyright (c) 2024, Rob Norris <robn@despairlabs.com>
- * Copyright (c) 2024, Klara, Inc.
+ * Copyright (c) 2024, 2025, Klara, Inc.
*/
#include <sys/dataset_kstats.h>
@@ -679,28 +679,19 @@ zvol_open(struct block_device *bdev, fmode_t flag)
retry:
#endif
- rw_enter(&zvol_state_lock, RW_READER);
- /*
- * Obtain a copy of private_data under the zvol_state_lock to make
- * sure that either the result of zvol free code path setting
- * disk->private_data to NULL is observed, or zvol_os_free()
- * is not called on this zv because of the positive zv_open_count.
- */
+
#ifdef HAVE_BLK_MODE_T
- zv = disk->private_data;
+ zv = atomic_load_ptr(&disk->private_data);
#else
- zv = bdev->bd_disk->private_data;
+ zv = atomic_load_ptr(&bdev->bd_disk->private_data);
#endif
if (zv == NULL) {
- rw_exit(&zvol_state_lock);
return (-SET_ERROR(ENXIO));
}
mutex_enter(&zv->zv_state_lock);
-
if (unlikely(zv->zv_flags & ZVOL_REMOVING)) {
mutex_exit(&zv->zv_state_lock);
- rw_exit(&zvol_state_lock);
return (-SET_ERROR(ENXIO));
}
@@ -712,8 +703,28 @@ retry:
if (zv->zv_open_count == 0) {
if (!rw_tryenter(&zv->zv_suspend_lock, RW_READER)) {
mutex_exit(&zv->zv_state_lock);
+
+ /*
+ * Removal may happen while the locks are down, so
+ * we can't trust zv any longer; we have to start over.
+ */
+#ifdef HAVE_BLK_MODE_T
+ zv = atomic_load_ptr(&disk->private_data);
+#else
+ zv = atomic_load_ptr(&bdev->bd_disk->private_data);
+#endif
+ if (zv == NULL)
+ return (-SET_ERROR(ENXIO));
+
rw_enter(&zv->zv_suspend_lock, RW_READER);
mutex_enter(&zv->zv_state_lock);
+
+ if (unlikely(zv->zv_flags & ZVOL_REMOVING)) {
+ mutex_exit(&zv->zv_state_lock);
+ rw_exit(&zv->zv_suspend_lock);
+ return (-SET_ERROR(ENXIO));
+ }
+
/* check to see if zv_suspend_lock is needed */
if (zv->zv_open_count != 0) {
rw_exit(&zv->zv_suspend_lock);
@@ -724,7 +735,6 @@ retry:
drop_suspend = B_TRUE;
}
}
- rw_exit(&zvol_state_lock);
ASSERT(MUTEX_HELD(&zv->zv_state_lock));
@@ -821,11 +831,11 @@ zvol_release(struct gendisk *disk, fmode_t unused)
#if !defined(HAVE_BLOCK_DEVICE_OPERATIONS_RELEASE_1ARG)
(void) unused;
#endif
- zvol_state_t *zv;
boolean_t drop_suspend = B_TRUE;
- rw_enter(&zvol_state_lock, RW_READER);
- zv = disk->private_data;
+ zvol_state_t *zv = atomic_load_ptr(&disk->private_data);
+ if (zv == NULL)
+ return;
mutex_enter(&zv->zv_state_lock);
ASSERT3U(zv->zv_open_count, >, 0);
@@ -839,6 +849,15 @@ zvol_release(struct gendisk *disk, fmode_t unused)
mutex_exit(&zv->zv_state_lock);
rw_enter(&zv->zv_suspend_lock, RW_READER);
mutex_enter(&zv->zv_state_lock);
+
+ /*
+ * Unlike in zvol_open(), we don't check if removal
+ * started here, because we might be one of the openers
+ * that needs to be thrown out! If we're the last, we
+ * need to call zvol_last_close() below to finish
+ * cleanup. So, no special treatment for us.
+ */
+
/* check to see if zv_suspend_lock is needed */
if (zv->zv_open_count != 1) {
rw_exit(&zv->zv_suspend_lock);
@@ -848,7 +867,6 @@ zvol_release(struct gendisk *disk, fmode_t unused)
} else {
drop_suspend = B_FALSE;
}
- rw_exit(&zvol_state_lock);
ASSERT(MUTEX_HELD(&zv->zv_state_lock));
@@ -868,9 +886,10 @@ static int
zvol_ioctl(struct block_device *bdev, fmode_t mode,
unsigned int cmd, unsigned long arg)
{
- zvol_state_t *zv = bdev->bd_disk->private_data;
int error = 0;
+ zvol_state_t *zv = atomic_load_ptr(&bdev->bd_disk->private_data);
+ ASSERT3P(zv, !=, NULL);
ASSERT3U(zv->zv_open_count, >, 0);
switch (cmd) {
@@ -923,9 +942,8 @@ zvol_check_events(struct gendisk *disk, unsigned int clearing)
{
unsigned int mask = 0;
- rw_enter(&zvol_state_lock, RW_READER);
+ zvol_state_t *zv = atomic_load_ptr(&disk->private_data);
- zvol_state_t *zv = disk->private_data;
if (zv != NULL) {
mutex_enter(&zv->zv_state_lock);
mask = zv->zv_changed ? DISK_EVENT_MEDIA_CHANGE : 0;
@@ -933,17 +951,14 @@ zvol_check_events(struct gendisk *disk, unsigned int clearing)
mutex_exit(&zv->zv_state_lock);
}
- rw_exit(&zvol_state_lock);
-
return (mask);
}
static int
zvol_revalidate_disk(struct gendisk *disk)
{
- rw_enter(&zvol_state_lock, RW_READER);
+ zvol_state_t *zv = atomic_load_ptr(&disk->private_data);
- zvol_state_t *zv = disk->private_data;
if (zv != NULL) {
mutex_enter(&zv->zv_state_lock);
set_capacity(zv->zv_zso->zvo_disk,
@@ -951,8 +966,6 @@ zvol_revalidate_disk(struct gendisk *disk)
mutex_exit(&zv->zv_state_lock);
}
- rw_exit(&zvol_state_lock);
-
return (0);
}
@@ -971,16 +984,6 @@ zvol_os_update_volsize(zvol_state_t *zv, uint64_t volsize)
return (0);
}
-void
-zvol_os_clear_private(zvol_state_t *zv)
-{
- /*
- * Cleared while holding zvol_state_lock as a writer
- * which will prevent zvol_open() from opening it.
- */
- zv->zv_zso->zvo_disk->private_data = NULL;
-}
-
/*
* Provide a simple virtual geometry for legacy compatibility. For devices
* smaller than 1 MiB a small head and sector count is used to allow very
@@ -990,9 +993,10 @@ zvol_os_clear_private(zvol_state_t *zv)
static int
zvol_getgeo(struct block_device *bdev, struct hd_geometry *geo)
{
- zvol_state_t *zv = bdev->bd_disk->private_data;
sector_t sectors;
+ zvol_state_t *zv = atomic_load_ptr(&bdev->bd_disk->private_data);
+ ASSERT3P(zv, !=, NULL);
ASSERT3U(zv->zv_open_count, >, 0);
sectors = get_capacity(zv->zv_zso->zvo_disk);
@@ -1417,53 +1421,70 @@ out_kmem:
return (ret);
}
-/*
- * Cleanup then free a zvol_state_t which was created by zvol_alloc().
- * At this time, the structure is not opened by anyone, is taken off
- * the zvol_state_list, and has its private data set to NULL.
- * The zvol_state_lock is dropped.
- *
- * This function may take many milliseconds to complete (e.g. we've seen
- * it take over 256ms), due to the calls to "blk_cleanup_queue" and
- * "del_gendisk". Thus, consumers need to be careful to account for this
- * latency when calling this function.
- */
void
-zvol_os_free(zvol_state_t *zv)
+zvol_os_remove_minor(zvol_state_t *zv)
{
-
- ASSERT(!RW_LOCK_HELD(&zv->zv_suspend_lock));
- ASSERT(!MUTEX_HELD(&zv->zv_state_lock));
+ ASSERT(MUTEX_HELD(&zv->zv_state_lock));
ASSERT0(zv->zv_open_count);
- ASSERT0P(zv->zv_zso->zvo_disk->private_data);
+ ASSERT0(atomic_read(&zv->zv_suspend_ref));
+ ASSERT(zv->zv_flags & ZVOL_REMOVING);
- rw_destroy(&zv->zv_suspend_lock);
- zfs_rangelock_fini(&zv->zv_rangelock);
+ struct zvol_state_os *zso = zv->zv_zso;
+ zv->zv_zso = NULL;
+
+ /* Clearing private_data will make new callers return immediately. */
+ atomic_store_ptr(&zso->zvo_disk->private_data, NULL);
+
+ /*
+ * Drop the state lock before calling del_gendisk(). There may be
+ * callers waiting to acquire it, but del_gendisk() will block until
+ * they exit, which would deadlock.
+ */
+ mutex_exit(&zv->zv_state_lock);
- del_gendisk(zv->zv_zso->zvo_disk);
+ del_gendisk(zso->zvo_disk);
#if defined(HAVE_SUBMIT_BIO_IN_BLOCK_DEVICE_OPERATIONS) && \
(defined(HAVE_BLK_ALLOC_DISK) || defined(HAVE_BLK_ALLOC_DISK_2ARG))
#if defined(HAVE_BLK_CLEANUP_DISK)
- blk_cleanup_disk(zv->zv_zso->zvo_disk);
+ blk_cleanup_disk(zso->zvo_disk);
#else
- put_disk(zv->zv_zso->zvo_disk);
+ put_disk(zso->zvo_disk);
#endif
#else
- blk_cleanup_queue(zv->zv_zso->zvo_queue);
- put_disk(zv->zv_zso->zvo_disk);
+ blk_cleanup_queue(zso->zvo_queue);
+ put_disk(zso->zvo_disk);
#endif
- if (zv->zv_zso->use_blk_mq)
- blk_mq_free_tag_set(&zv->zv_zso->tag_set);
+ if (zso->use_blk_mq)
+ blk_mq_free_tag_set(&zso->tag_set);
+
+ ida_simple_remove(&zvol_ida, MINOR(zso->zvo_dev) >> ZVOL_MINOR_BITS);
- ida_simple_remove(&zvol_ida,
- MINOR(zv->zv_zso->zvo_dev) >> ZVOL_MINOR_BITS);
+ kmem_free(zso, sizeof (struct zvol_state_os));
+
+ mutex_enter(&zv->zv_state_lock);
+}
+
+void
+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);
+ ASSERT0P(zv->zv_zso);
+
+ ASSERT0P(zv->zv_objset);
+ ASSERT0P(zv->zv_zilog);
+ ASSERT0P(zv->zv_dn);
+
+ rw_destroy(&zv->zv_suspend_lock);
+ zfs_rangelock_fini(&zv->zv_rangelock);
cv_destroy(&zv->zv_removing_cv);
mutex_destroy(&zv->zv_state_lock);
dataset_kstats_destroy(&zv->zv_kstat);
- kmem_free(zv->zv_zso, sizeof (struct zvol_state_os));
kmem_free(zv, sizeof (zvol_state_t));
}
diff --git a/sys/contrib/openzfs/module/zcommon/simd_stat.c b/sys/contrib/openzfs/module/zcommon/simd_stat.c
index 11e2080ff9f2..007ae9e4fbbc 100644
--- a/sys/contrib/openzfs/module/zcommon/simd_stat.c
+++ b/sys/contrib/openzfs/module/zcommon/simd_stat.c
@@ -118,6 +118,10 @@ simd_stat_kstat_data(char *buf, size_t size, void *data)
"pclmulqdq", zfs_pclmulqdq_available());
off += SIMD_STAT_PRINT(simd_stat_kstat_payload,
"movbe", zfs_movbe_available());
+ off += SIMD_STAT_PRINT(simd_stat_kstat_payload,
+ "vaes", zfs_vaes_available());
+ off += SIMD_STAT_PRINT(simd_stat_kstat_payload,
+ "vpclmulqdq", zfs_vpclmulqdq_available());
off += SIMD_STAT_PRINT(simd_stat_kstat_payload,
"osxsave", boot_cpu_has(X86_FEATURE_OSXSAVE));
diff --git a/sys/contrib/openzfs/module/zfs/dbuf.c b/sys/contrib/openzfs/module/zfs/dbuf.c
index 3d0f88b36336..7403f10d91b7 100644
--- a/sys/contrib/openzfs/module/zfs/dbuf.c
+++ b/sys/contrib/openzfs/module/zfs/dbuf.c
@@ -2557,12 +2557,13 @@ dbuf_undirty(dmu_buf_impl_t *db, dmu_tx_t *tx)
/*
* Due to our use of dn_nlevels below, this can only be called
- * in open context, unless we are operating on the MOS.
- * From syncing context, dn_nlevels may be different from the
- * dn_nlevels used when dbuf was dirtied.
+ * in open context, unless we are operating on the MOS or it's
+ * a special object. From syncing context, dn_nlevels may be
+ * different from the dn_nlevels used when dbuf was dirtied.
*/
ASSERT(db->db_objset ==
dmu_objset_pool(db->db_objset)->dp_meta_objset ||
+ DMU_OBJECT_IS_SPECIAL(db->db.db_object) ||
txg != spa_syncing_txg(dmu_objset_spa(db->db_objset)));
ASSERT(db->db_blkid != DMU_BONUS_BLKID);
ASSERT0(db->db_level);
diff --git a/sys/contrib/openzfs/module/zfs/multilist.c b/sys/contrib/openzfs/module/zfs/multilist.c
index 7b85d19e19ee..46fb79269310 100644
--- a/sys/contrib/openzfs/module/zfs/multilist.c
+++ b/sys/contrib/openzfs/module/zfs/multilist.c
@@ -81,7 +81,7 @@ multilist_create_impl(multilist_t *ml, size_t size, size_t offset,
ml->ml_num_sublists = num;
ml->ml_index_func = index_func;
- ml->ml_sublists = kmem_zalloc(sizeof (multilist_sublist_t) *
+ ml->ml_sublists = vmem_zalloc(sizeof (multilist_sublist_t) *
ml->ml_num_sublists, KM_SLEEP);
ASSERT3P(ml->ml_sublists, !=, NULL);
@@ -134,7 +134,7 @@ multilist_destroy(multilist_t *ml)
}
ASSERT3P(ml->ml_sublists, !=, NULL);
- kmem_free(ml->ml_sublists,
+ vmem_free(ml->ml_sublists,
sizeof (multilist_sublist_t) * ml->ml_num_sublists);
ml->ml_num_sublists = 0;
diff --git a/sys/contrib/openzfs/module/zfs/spa_config.c b/sys/contrib/openzfs/module/zfs/spa_config.c
index 7d4d06659146..cf28955b0c50 100644
--- a/sys/contrib/openzfs/module/zfs/spa_config.c
+++ b/sys/contrib/openzfs/module/zfs/spa_config.c
@@ -48,18 +48,17 @@
/*
* Pool configuration repository.
*
- * Pool configuration is stored as a packed nvlist on the filesystem. By
- * default, all pools are stored in /etc/zfs/zpool.cache and loaded on boot
- * (when the ZFS module is loaded). Pools can also have the 'cachefile'
- * property set that allows them to be stored in an alternate location until
- * the control of external software.
+ * Pool configuration is stored as a packed nvlist on the filesystem. When
+ * pools are imported they are added to the /etc/zfs/zpool.cache file and
+ * removed from it when exported. For each cache file, we have a single nvlist
+ * which holds all the configuration information. Pools can also have the
+ * 'cachefile' property set which allows this config to be stored in an
+ * alternate location under the control of external software.
*
- * For each cache file, we have a single nvlist which holds all the
- * configuration information. When the module loads, we read this information
- * from /etc/zfs/zpool.cache and populate the SPA namespace. This namespace is
- * maintained independently in spa.c. Whenever the namespace is modified, or
- * the configuration of a pool is changed, we call spa_write_cachefile(), which
- * walks through all the active pools and writes the configuration to disk.
+ * The kernel independantly maintains an AVL tree of imported pools. See the
+ * "SPA locking" comment in spa.c. Whenever a pool configuration is modified
+ * we call spa_write_cachefile() which walks through all the active pools and
+ * writes the updated configuration to to /etc/zfs/zpool.cache file.
*/
static uint64_t spa_config_generation = 1;
@@ -69,94 +68,6 @@ static uint64_t spa_config_generation = 1;
* userland pools when doing testing.
*/
char *spa_config_path = (char *)ZPOOL_CACHE;
-#ifdef _KERNEL
-static int zfs_autoimport_disable = B_TRUE;
-#endif
-
-/*
- * Called when the module is first loaded, this routine loads the configuration
- * file into the SPA namespace. It does not actually open or load the pools; it
- * only populates the namespace.
- */
-void
-spa_config_load(void)
-{
- void *buf = NULL;
- nvlist_t *nvlist, *child;
- nvpair_t *nvpair;
- char *pathname;
- zfs_file_t *fp;
- zfs_file_attr_t zfa;
- uint64_t fsize;
- int err;
-
-#ifdef _KERNEL
- if (zfs_autoimport_disable)
- return;
-#endif
-
- /*
- * Open the configuration file.
- */
- pathname = kmem_alloc(MAXPATHLEN, KM_SLEEP);
-
- (void) snprintf(pathname, MAXPATHLEN, "%s", spa_config_path);
-
- err = zfs_file_open(pathname, O_RDONLY, 0, &fp);
-
-#ifdef __FreeBSD__
- if (err)
- err = zfs_file_open(ZPOOL_CACHE_BOOT, O_RDONLY, 0, &fp);
-#endif
- kmem_free(pathname, MAXPATHLEN);
-
- if (err)
- return;
-
- if (zfs_file_getattr(fp, &zfa))
- goto out;
-
- fsize = zfa.zfa_size;
- buf = kmem_alloc(fsize, KM_SLEEP);
-
- /*
- * Read the nvlist from the file.
- */
- if (zfs_file_read(fp, buf, fsize, NULL) < 0)
- goto out;
-
- /*
- * Unpack the nvlist.
- */
- if (nvlist_unpack(buf, fsize, &nvlist, KM_SLEEP) != 0)
- goto out;
-
- /*
- * Iterate over all elements in the nvlist, creating a new spa_t for
- * each one with the specified configuration.
- */
- mutex_enter(&spa_namespace_lock);
- nvpair = NULL;
- while ((nvpair = nvlist_next_nvpair(nvlist, nvpair)) != NULL) {
- if (nvpair_type(nvpair) != DATA_TYPE_NVLIST)
- continue;
-
- child = fnvpair_value_nvlist(nvpair);
-
- if (spa_lookup(nvpair_name(nvpair)) != NULL)
- continue;
- (void) spa_add(nvpair_name(nvpair), child, NULL);
- }
- mutex_exit(&spa_namespace_lock);
-
- nvlist_free(nvlist);
-
-out:
- if (buf != NULL)
- kmem_free(buf, fsize);
-
- zfs_file_close(fp);
-}
static int
spa_config_remove(spa_config_dirent_t *dp)
@@ -623,7 +534,6 @@ spa_config_update(spa_t *spa, int what)
spa_config_update(spa, SPA_CONFIG_UPDATE_VDEVS);
}
-EXPORT_SYMBOL(spa_config_load);
EXPORT_SYMBOL(spa_all_configs);
EXPORT_SYMBOL(spa_config_set);
EXPORT_SYMBOL(spa_config_generate);
@@ -634,8 +544,3 @@ EXPORT_SYMBOL(spa_config_update);
ZFS_MODULE_PARAM(zfs_spa, spa_, config_path, STRING, ZMOD_RD,
"SPA config file (/etc/zfs/zpool.cache)");
#endif
-
-#ifdef _KERNEL
-ZFS_MODULE_PARAM(zfs, zfs_, autoimport_disable, INT, ZMOD_RW,
- "Disable pool import at module load");
-#endif
diff --git a/sys/contrib/openzfs/module/zfs/spa_misc.c b/sys/contrib/openzfs/module/zfs/spa_misc.c
index cce772eae598..dceafbc27556 100644
--- a/sys/contrib/openzfs/module/zfs/spa_misc.c
+++ b/sys/contrib/openzfs/module/zfs/spa_misc.c
@@ -2548,13 +2548,6 @@ spa_name_compare(const void *a1, const void *a2)
}
void
-spa_boot_init(void *unused)
-{
- (void) unused;
- spa_config_load();
-}
-
-void
spa_init(spa_mode_t mode)
{
mutex_init(&spa_namespace_lock, NULL, MUTEX_DEFAULT, NULL);
@@ -2607,7 +2600,6 @@ spa_init(spa_mode_t mode)
chksum_init();
zpool_prop_init();
zpool_feature_init();
- spa_config_load();
vdev_prop_init();
l2arc_start();
scan_init();
diff --git a/sys/contrib/openzfs/module/zfs/zil.c b/sys/contrib/openzfs/module/zfs/zil.c
index 31b59c55f17b..0307df55aa21 100644
--- a/sys/contrib/openzfs/module/zfs/zil.c
+++ b/sys/contrib/openzfs/module/zfs/zil.c
@@ -819,34 +819,37 @@ zil_lwb_vdev_compare(const void *x1, const void *x2)
* we choose them here and later make the block allocation match.
*/
static lwb_t *
-zil_alloc_lwb(zilog_t *zilog, int sz, blkptr_t *bp, boolean_t slog,
- uint64_t txg, lwb_state_t state)
+zil_alloc_lwb(zilog_t *zilog, blkptr_t *bp, int min_sz, int sz,
+ boolean_t slog, uint64_t txg)
{
lwb_t *lwb;
lwb = kmem_cache_alloc(zil_lwb_cache, KM_SLEEP);
+ lwb->lwb_flags = 0;
lwb->lwb_zilog = zilog;
if (bp) {
lwb->lwb_blk = *bp;
- lwb->lwb_slim = (BP_GET_CHECKSUM(bp) == ZIO_CHECKSUM_ZILOG2);
+ if (BP_GET_CHECKSUM(bp) == ZIO_CHECKSUM_ZILOG2)
+ lwb->lwb_flags |= LWB_FLAG_SLIM;
sz = BP_GET_LSIZE(bp);
+ lwb->lwb_min_sz = sz;
} else {
BP_ZERO(&lwb->lwb_blk);
- lwb->lwb_slim = (spa_version(zilog->zl_spa) >=
- SPA_VERSION_SLIM_ZIL);
+ if (spa_version(zilog->zl_spa) >= SPA_VERSION_SLIM_ZIL)
+ lwb->lwb_flags |= LWB_FLAG_SLIM;
+ lwb->lwb_min_sz = min_sz;
}
- lwb->lwb_slog = slog;
+ if (slog)
+ lwb->lwb_flags |= LWB_FLAG_SLOG;
lwb->lwb_error = 0;
- if (lwb->lwb_slim) {
- lwb->lwb_nmax = sz;
- lwb->lwb_nused = lwb->lwb_nfilled = sizeof (zil_chain_t);
- } else {
- lwb->lwb_nmax = sz - sizeof (zil_chain_t);
- lwb->lwb_nused = lwb->lwb_nfilled = 0;
- }
+ /*
+ * Buffer allocation and capacity setup will be done in
+ * zil_lwb_write_open() when the LWB is opened for ITX assignment.
+ */
+ lwb->lwb_nmax = lwb->lwb_nused = lwb->lwb_nfilled = 0;
lwb->lwb_sz = sz;
- lwb->lwb_state = state;
- lwb->lwb_buf = zio_buf_alloc(sz);
+ lwb->lwb_buf = NULL;
+ lwb->lwb_state = LWB_STATE_NEW;
lwb->lwb_child_zio = NULL;
lwb->lwb_write_zio = NULL;
lwb->lwb_root_zio = NULL;
@@ -857,8 +860,6 @@ zil_alloc_lwb(zilog_t *zilog, int sz, blkptr_t *bp, boolean_t slog,
mutex_enter(&zilog->zl_lock);
list_insert_tail(&zilog->zl_lwb_list, lwb);
- if (state != LWB_STATE_NEW)
- zilog->zl_last_lwb_opened = lwb;
mutex_exit(&zilog->zl_lock);
return (lwb);
@@ -878,7 +879,7 @@ zil_free_lwb(zilog_t *zilog, lwb_t *lwb)
VERIFY(list_is_empty(&lwb->lwb_itxs));
VERIFY(list_is_empty(&lwb->lwb_waiters));
ASSERT(avl_is_empty(&lwb->lwb_vdev_tree));
- ASSERT(!MUTEX_HELD(&lwb->lwb_vdev_lock));
+ ASSERT(!MUTEX_HELD(&lwb->lwb_lock));
/*
* Clear the zilog's field to indicate this lwb is no longer
@@ -1019,7 +1020,7 @@ zil_create(zilog_t *zilog)
}
error = zio_alloc_zil(zilog->zl_spa, zilog->zl_os, txg, &blk,
- ZIL_MIN_BLKSZ, &slog);
+ ZIL_MIN_BLKSZ, ZIL_MIN_BLKSZ, &slog, B_TRUE);
if (error == 0)
zil_init_log_chain(zilog, &blk);
}
@@ -1028,7 +1029,7 @@ zil_create(zilog_t *zilog)
* Allocate a log write block (lwb) for the first log block.
*/
if (error == 0)
- lwb = zil_alloc_lwb(zilog, 0, &blk, slog, txg, LWB_STATE_NEW);
+ lwb = zil_alloc_lwb(zilog, &blk, 0, 0, slog, txg);
/*
* If we just allocated the first log block, commit our transaction
@@ -1324,10 +1325,12 @@ zil_check_log_chain(dsl_pool_t *dp, dsl_dataset_t *ds, void *tx)
* zil_commit() is racing with spa_sync().
*/
static void
-zil_commit_waiter_skip(zil_commit_waiter_t *zcw)
+zil_commit_waiter_done(zil_commit_waiter_t *zcw, int err)
{
mutex_enter(&zcw->zcw_lock);
ASSERT3B(zcw->zcw_done, ==, B_FALSE);
+ zcw->zcw_lwb = NULL;
+ zcw->zcw_error = err;
zcw->zcw_done = B_TRUE;
cv_broadcast(&zcw->zcw_cv);
mutex_exit(&zcw->zcw_lock);
@@ -1389,7 +1392,7 @@ zil_lwb_add_block(lwb_t *lwb, const blkptr_t *bp)
if (zil_nocacheflush)
return;
- mutex_enter(&lwb->lwb_vdev_lock);
+ mutex_enter(&lwb->lwb_lock);
for (i = 0; i < ndvas; i++) {
zvsearch.zv_vdev = DVA_GET_VDEV(&bp->blk_dva[i]);
if (avl_find(t, &zvsearch, &where) == NULL) {
@@ -1398,7 +1401,7 @@ zil_lwb_add_block(lwb_t *lwb, const blkptr_t *bp)
avl_insert(t, zv, where);
}
}
- mutex_exit(&lwb->lwb_vdev_lock);
+ mutex_exit(&lwb->lwb_lock);
}
static void
@@ -1415,12 +1418,12 @@ zil_lwb_flush_defer(lwb_t *lwb, lwb_t *nlwb)
/*
* While 'lwb' is at a point in its lifetime where lwb_vdev_tree does
- * not need the protection of lwb_vdev_lock (it will only be modified
+ * not need the protection of lwb_lock (it will only be modified
* while holding zilog->zl_lock) as its writes and those of its
* children have all completed. The younger 'nlwb' may be waiting on
* future writes to additional vdevs.
*/
- mutex_enter(&nlwb->lwb_vdev_lock);
+ mutex_enter(&nlwb->lwb_lock);
/*
* Tear down the 'lwb' vdev tree, ensuring that entries which do not
* exist in 'nlwb' are moved to it, freeing any would-be duplicates.
@@ -1434,7 +1437,7 @@ zil_lwb_flush_defer(lwb_t *lwb, lwb_t *nlwb)
kmem_free(zv, sizeof (*zv));
}
}
- mutex_exit(&nlwb->lwb_vdev_lock);
+ mutex_exit(&nlwb->lwb_lock);
}
void
@@ -1491,10 +1494,6 @@ zil_lwb_flush_vdevs_done(zio_t *zio)
zil_itx_destroy(itx, 0);
while ((zcw = list_remove_head(&lwb->lwb_waiters)) != NULL) {
- mutex_enter(&zcw->zcw_lock);
-
- ASSERT3P(zcw->zcw_lwb, ==, lwb);
- zcw->zcw_lwb = NULL;
/*
* We expect any ZIO errors from child ZIOs to have been
* propagated "up" to this specific LWB's root ZIO, in
@@ -1509,14 +1508,7 @@ zil_lwb_flush_vdevs_done(zio_t *zio)
* errors not being handled correctly here. See the
* comment above the call to "zio_flush" for details.
*/
-
- zcw->zcw_zio_error = zio->io_error;
-
- ASSERT3B(zcw->zcw_done, ==, B_FALSE);
- zcw->zcw_done = B_TRUE;
- cv_broadcast(&zcw->zcw_cv);
-
- mutex_exit(&zcw->zcw_lock);
+ zil_commit_waiter_done(zcw, zio->io_error);
}
uint64_t txg = lwb->lwb_issued_txg;
@@ -1588,7 +1580,7 @@ zil_lwb_write_done(zio_t *zio)
avl_tree_t *t = &lwb->lwb_vdev_tree;
void *cookie = NULL;
zil_vdev_node_t *zv;
- lwb_t *nlwb;
+ lwb_t *nlwb = NULL;
ASSERT3S(spa_config_held(spa, SCL_STATE, RW_READER), !=, 0);
@@ -1608,9 +1600,11 @@ zil_lwb_write_done(zio_t *zio)
* its write ZIO a parent this ZIO. In such case we can not defer
* our flushes or below may be a race between the done callbacks.
*/
- nlwb = list_next(&zilog->zl_lwb_list, lwb);
- if (nlwb && nlwb->lwb_state != LWB_STATE_ISSUED)
- nlwb = NULL;
+ if (!(lwb->lwb_flags & LWB_FLAG_CRASHED)) {
+ nlwb = list_next(&zilog->zl_lwb_list, lwb);
+ if (nlwb && nlwb->lwb_state != LWB_STATE_ISSUED)
+ nlwb = NULL;
+ }
mutex_exit(&zilog->zl_lock);
if (avl_numnodes(t) == 0)
@@ -1624,12 +1618,17 @@ zil_lwb_write_done(zio_t *zio)
* written out.
*
* Additionally, we don't perform any further error handling at
- * this point (e.g. setting "zcw_zio_error" appropriately), as
- * we expect that to occur in "zil_lwb_flush_vdevs_done" (thus,
- * we expect any error seen here, to have been propagated to
- * that function).
+ * this point (e.g. setting "zcw_error" appropriately), as we
+ * expect that to occur in "zil_lwb_flush_vdevs_done" (thus, we
+ * expect any error seen here, to have been propagated to that
+ * function).
+ *
+ * Note that we treat a "crashed" LWB as though it was in error,
+ * even if it did appear to succeed, because we've already
+ * signaled error and cleaned up waiters and committers in
+ * zil_crash(); we just want to clean up and get out of here.
*/
- if (zio->io_error != 0) {
+ if (zio->io_error != 0 || (lwb->lwb_flags & LWB_FLAG_CRASHED)) {
while ((zv = avl_destroy_nodes(t, &cookie)) != NULL)
kmem_free(zv, sizeof (*zv));
return;
@@ -1742,10 +1741,26 @@ zil_lwb_write_open(zilog_t *zilog, lwb_t *lwb)
return;
}
+ mutex_enter(&lwb->lwb_lock);
mutex_enter(&zilog->zl_lock);
lwb->lwb_state = LWB_STATE_OPENED;
zilog->zl_last_lwb_opened = lwb;
mutex_exit(&zilog->zl_lock);
+ mutex_exit(&lwb->lwb_lock);
+
+ /*
+ * Allocate buffer and set up LWB capacities.
+ */
+ ASSERT0P(lwb->lwb_buf);
+ ASSERT3U(lwb->lwb_sz, >, 0);
+ lwb->lwb_buf = zio_buf_alloc(lwb->lwb_sz);
+ if (lwb->lwb_flags & LWB_FLAG_SLIM) {
+ lwb->lwb_nmax = lwb->lwb_sz;
+ lwb->lwb_nused = lwb->lwb_nfilled = sizeof (zil_chain_t);
+ } else {
+ lwb->lwb_nmax = lwb->lwb_sz - sizeof (zil_chain_t);
+ lwb->lwb_nused = lwb->lwb_nfilled = 0;
+ }
}
/*
@@ -1762,6 +1777,8 @@ static uint_t
zil_lwb_plan(zilog_t *zilog, uint64_t size, uint_t *minsize)
{
uint_t md = zilog->zl_max_block_size - sizeof (zil_chain_t);
+ uint_t waste = zil_max_waste_space(zilog);
+ waste = MAX(waste, zilog->zl_cur_max);
if (size <= md) {
/*
@@ -1772,9 +1789,10 @@ zil_lwb_plan(zilog_t *zilog, uint64_t size, uint_t *minsize)
} else if (size > 8 * md) {
/*
* Big bursts use maximum blocks. The first block size
- * is hard to predict, but it does not really matter.
+ * is hard to predict, but we need at least enough space
+ * to make reasonable progress.
*/
- *minsize = 0;
+ *minsize = waste;
return (md);
}
@@ -1787,57 +1805,52 @@ zil_lwb_plan(zilog_t *zilog, uint64_t size, uint_t *minsize)
uint_t s = size;
uint_t n = DIV_ROUND_UP(s, md - sizeof (lr_write_t));
uint_t chunk = DIV_ROUND_UP(s, n);
- uint_t waste = zil_max_waste_space(zilog);
- waste = MAX(waste, zilog->zl_cur_max);
if (chunk <= md - waste) {
*minsize = MAX(s - (md - waste) * (n - 1), waste);
return (chunk);
} else {
- *minsize = 0;
+ *minsize = waste;
return (md);
}
}
/*
* Try to predict next block size based on previous history. Make prediction
- * sufficient for 7 of 8 previous bursts. Don't try to save if the saving is
- * less then 50%, extra writes may cost more, but we don't want single spike
- * to badly affect our predictions.
+ * sufficient for 7 of 8 previous bursts, but don't try to save if the saving
+ * is less then 50%. Extra writes may cost more, but we don't want single
+ * spike to badly affect our predictions.
*/
-static uint_t
-zil_lwb_predict(zilog_t *zilog)
+static void
+zil_lwb_predict(zilog_t *zilog, uint64_t *min_predict, uint64_t *max_predict)
{
- uint_t m, o;
+ uint_t m1 = 0, m2 = 0, o;
- /* If we are in the middle of a burst, take it into account also. */
- if (zilog->zl_cur_size > 0) {
- o = zil_lwb_plan(zilog, zilog->zl_cur_size, &m);
- } else {
+ /* If we are in the middle of a burst, take it as another data point. */
+ if (zilog->zl_cur_size > 0)
+ o = zil_lwb_plan(zilog, zilog->zl_cur_size, &m1);
+ else
o = UINT_MAX;
- m = 0;
- }
- /* Find minimum optimal size. We don't need to go below that. */
- for (int i = 0; i < ZIL_BURSTS; i++)
- o = MIN(o, zilog->zl_prev_opt[i]);
-
- /* Find two biggest minimal first block sizes above the optimal. */
- uint_t m1 = MAX(m, o), m2 = o;
+ /* Find two largest minimal first block sizes. */
for (int i = 0; i < ZIL_BURSTS; i++) {
- m = zilog->zl_prev_min[i];
- if (m >= m1) {
+ uint_t cur = zilog->zl_prev_min[i];
+ if (cur >= m1) {
m2 = m1;
- m1 = m;
- } else if (m > m2) {
- m2 = m;
+ m1 = cur;
+ } else if (cur > m2) {
+ m2 = cur;
}
}
- /*
- * If second minimum size gives 50% saving -- use it. It may cost us
- * one additional write later, but the space saving is just too big.
- */
- return ((m1 < m2 * 2) ? m1 : m2);
+ /* Minimum should guarantee progress in most cases. */
+ *min_predict = (m1 < m2 * 2) ? m1 : m2;
+
+ /* Maximum doesn't need to go below the minimum optimal size. */
+ for (int i = 0; i < ZIL_BURSTS; i++)
+ o = MIN(o, zilog->zl_prev_opt[i]);
+ m1 = MAX(m1, o);
+ m2 = MAX(m2, o);
+ *max_predict = (m1 < m2 * 2) ? m1 : m2;
}
/*
@@ -1845,12 +1858,13 @@ zil_lwb_predict(zilog_t *zilog)
* Has to be called under zl_issuer_lock to chain more lwbs.
*/
static lwb_t *
-zil_lwb_write_close(zilog_t *zilog, lwb_t *lwb, lwb_state_t state)
+zil_lwb_write_close(zilog_t *zilog, lwb_t *lwb)
{
- uint64_t blksz, plan, plan2;
+ uint64_t minbs, maxbs;
ASSERT(MUTEX_HELD(&zilog->zl_issuer_lock));
ASSERT3S(lwb->lwb_state, ==, LWB_STATE_OPENED);
+ membar_producer();
lwb->lwb_state = LWB_STATE_CLOSED;
/*
@@ -1875,27 +1889,34 @@ zil_lwb_write_close(zilog_t *zilog, lwb_t *lwb, lwb_state_t state)
* Try to predict what can it be and plan for the worst case.
*/
uint_t m;
- plan = zil_lwb_plan(zilog, zilog->zl_cur_left, &m);
+ maxbs = zil_lwb_plan(zilog, zilog->zl_cur_left, &m);
+ minbs = m;
if (zilog->zl_parallel) {
- plan2 = zil_lwb_plan(zilog, zilog->zl_cur_left +
- zil_lwb_predict(zilog), &m);
- if (plan < plan2)
- plan = plan2;
+ uint64_t minp, maxp;
+ zil_lwb_predict(zilog, &minp, &maxp);
+ maxp = zil_lwb_plan(zilog, zilog->zl_cur_left + maxp,
+ &m);
+ if (maxbs < maxp)
+ maxbs = maxp;
}
} else {
/*
* The previous burst is done and we can only predict what
* will come next.
*/
- plan = zil_lwb_predict(zilog);
+ zil_lwb_predict(zilog, &minbs, &maxbs);
}
- blksz = plan + sizeof (zil_chain_t);
- blksz = P2ROUNDUP_TYPED(blksz, ZIL_MIN_BLKSZ, uint64_t);
- blksz = MIN(blksz, zilog->zl_max_block_size);
- DTRACE_PROBE3(zil__block__size, zilog_t *, zilog, uint64_t, blksz,
- uint64_t, plan);
- return (zil_alloc_lwb(zilog, blksz, NULL, 0, 0, state));
+ minbs += sizeof (zil_chain_t);
+ maxbs += sizeof (zil_chain_t);
+ minbs = P2ROUNDUP_TYPED(minbs, ZIL_MIN_BLKSZ, uint64_t);
+ maxbs = P2ROUNDUP_TYPED(maxbs, ZIL_MIN_BLKSZ, uint64_t);
+ maxbs = MIN(maxbs, zilog->zl_max_block_size);
+ minbs = MIN(minbs, maxbs);
+ DTRACE_PROBE3(zil__block__size, zilog_t *, zilog, uint64_t, minbs,
+ uint64_t, maxbs);
+
+ return (zil_alloc_lwb(zilog, NULL, minbs, maxbs, 0, 0));
}
/*
@@ -1944,14 +1965,16 @@ zil_lwb_write_issue(zilog_t *zilog, lwb_t *lwb)
mutex_exit(&zilog->zl_lock);
next_lwb:
- if (lwb->lwb_slim)
+ if (lwb->lwb_flags & LWB_FLAG_SLIM)
zilc = (zil_chain_t *)lwb->lwb_buf;
else
zilc = (zil_chain_t *)(lwb->lwb_buf + lwb->lwb_nmax);
- int wsz = lwb->lwb_sz;
+ uint64_t alloc_size = BP_GET_LSIZE(&lwb->lwb_blk);
+ int wsz = alloc_size;
if (lwb->lwb_error == 0) {
abd_t *lwb_abd = abd_get_from_buf(lwb->lwb_buf, lwb->lwb_sz);
- if (!lwb->lwb_slog || zilog->zl_cur_size <= zil_slog_bulk)
+ if (!(lwb->lwb_flags & LWB_FLAG_SLOG) ||
+ zilog->zl_cur_size <= zil_slog_bulk)
prio = ZIO_PRIORITY_SYNC_WRITE;
else
prio = ZIO_PRIORITY_ASYNC_WRITE;
@@ -1959,16 +1982,17 @@ next_lwb:
ZB_ZIL_OBJECT, ZB_ZIL_LEVEL,
lwb->lwb_blk.blk_cksum.zc_word[ZIL_ZC_SEQ]);
lwb->lwb_write_zio = zio_rewrite(lwb->lwb_root_zio, spa, 0,
- &lwb->lwb_blk, lwb_abd, lwb->lwb_sz, zil_lwb_write_done,
+ &lwb->lwb_blk, lwb_abd, alloc_size, zil_lwb_write_done,
lwb, prio, ZIO_FLAG_CANFAIL, &zb);
zil_lwb_add_block(lwb, &lwb->lwb_blk);
- if (lwb->lwb_slim) {
+ if (lwb->lwb_flags & LWB_FLAG_SLIM) {
/* For Slim ZIL only write what is used. */
wsz = P2ROUNDUP_TYPED(lwb->lwb_nused, ZIL_MIN_BLKSZ,
int);
- ASSERT3S(wsz, <=, lwb->lwb_sz);
- zio_shrink(lwb->lwb_write_zio, wsz);
+ ASSERT3S(wsz, <=, alloc_size);
+ if (wsz < alloc_size)
+ zio_shrink(lwb->lwb_write_zio, wsz);
wsz = lwb->lwb_write_zio->io_size;
}
memset(lwb->lwb_buf + lwb->lwb_nused, 0, wsz - lwb->lwb_nused);
@@ -2004,13 +2028,53 @@ next_lwb:
BP_ZERO(bp);
error = lwb->lwb_error;
if (error == 0) {
- error = zio_alloc_zil(spa, zilog->zl_os, txg, bp, nlwb->lwb_sz,
- &slog);
+ /*
+ * Allocation flexibility depends on LWB state:
+ * if NEW: allow range allocation and larger sizes;
+ * if OPENED: use fixed predetermined allocation size;
+ * if CLOSED + Slim: allocate precisely for actual usage.
+ */
+ boolean_t flexible = (nlwb->lwb_state == LWB_STATE_NEW);
+ if (flexible) {
+ /* We need to prevent opening till we update lwb_sz. */
+ mutex_enter(&nlwb->lwb_lock);
+ flexible = (nlwb->lwb_state == LWB_STATE_NEW);
+ if (!flexible)
+ mutex_exit(&nlwb->lwb_lock); /* We lost. */
+ }
+ boolean_t closed_slim = (nlwb->lwb_state == LWB_STATE_CLOSED &&
+ (lwb->lwb_flags & LWB_FLAG_SLIM));
+
+ uint64_t min_size, max_size;
+ if (closed_slim) {
+ /* This transition is racy, but only one way. */
+ membar_consumer();
+ min_size = max_size = P2ROUNDUP_TYPED(nlwb->lwb_nused,
+ ZIL_MIN_BLKSZ, uint64_t);
+ } else if (flexible) {
+ min_size = nlwb->lwb_min_sz;
+ max_size = nlwb->lwb_sz;
+ } else {
+ min_size = max_size = nlwb->lwb_sz;
+ }
+
+ error = zio_alloc_zil(spa, zilog->zl_os, txg, bp,
+ min_size, max_size, &slog, flexible);
+ if (error == 0) {
+ if (closed_slim)
+ ASSERT3U(BP_GET_LSIZE(bp), ==, max_size);
+ else if (flexible)
+ nlwb->lwb_sz = BP_GET_LSIZE(bp);
+ else
+ ASSERT3U(BP_GET_LSIZE(bp), ==, nlwb->lwb_sz);
+ }
+ if (flexible)
+ mutex_exit(&nlwb->lwb_lock);
}
if (error == 0) {
ASSERT3U(BP_GET_BIRTH(bp), ==, txg);
- BP_SET_CHECKSUM(bp, nlwb->lwb_slim ? ZIO_CHECKSUM_ZILOG2 :
- ZIO_CHECKSUM_ZILOG);
+ BP_SET_CHECKSUM(bp, (nlwb->lwb_flags & LWB_FLAG_SLIM) ?
+ ZIO_CHECKSUM_ZILOG2 : ZIO_CHECKSUM_ZILOG);
bp->blk_cksum = lwb->lwb_blk.blk_cksum;
bp->blk_cksum.zc_word[ZIL_ZC_SEQ]++;
}
@@ -2039,14 +2103,15 @@ next_lwb:
if (nlwb) {
nlwb->lwb_blk = *bp;
nlwb->lwb_error = error;
- nlwb->lwb_slog = slog;
+ if (slog)
+ nlwb->lwb_flags |= LWB_FLAG_SLOG;
nlwb->lwb_alloc_txg = txg;
if (nlwb->lwb_state != LWB_STATE_READY)
nlwb = NULL;
}
mutex_exit(&zilog->zl_lock);
- if (lwb->lwb_slog) {
+ if (lwb->lwb_flags & LWB_FLAG_SLOG) {
ZIL_STAT_BUMP(zilog, zil_itx_metaslab_slog_count);
ZIL_STAT_INCR(zilog, zil_itx_metaslab_slog_bytes,
lwb->lwb_nused);
@@ -2220,7 +2285,6 @@ zil_lwb_assign(zilog_t *zilog, lwb_t *lwb, itx_t *itx, list_t *ilwbs)
ASSERT(MUTEX_HELD(&zilog->zl_issuer_lock));
ASSERT3P(lwb, !=, NULL);
- ASSERT3P(lwb->lwb_buf, !=, NULL);
zil_lwb_write_open(zilog, lwb);
@@ -2262,9 +2326,10 @@ cont:
(dlen % max_log_data == 0 ||
lwb_sp < reclen + dlen % max_log_data))) {
list_insert_tail(ilwbs, lwb);
- lwb = zil_lwb_write_close(zilog, lwb, LWB_STATE_OPENED);
+ lwb = zil_lwb_write_close(zilog, lwb);
if (lwb == NULL)
return (NULL);
+ zil_lwb_write_open(zilog, lwb);
lwb_sp = lwb->lwb_nmax - lwb->lwb_nused;
}
@@ -2554,7 +2619,7 @@ zil_itxg_clean(void *arg)
* called) we will hit this case.
*/
if (itx->itx_lr.lrc_txtype == TX_COMMIT)
- zil_commit_waiter_skip(itx->itx_private);
+ zil_commit_waiter_done(itx->itx_private, 0);
zil_itx_destroy(itx, 0);
}
@@ -2742,6 +2807,7 @@ zil_crash_clean(zilog_t *zilog, uint64_t synced_txg)
}
/* This LWB is from the past, so we can clean it up now. */
+ ASSERT(lwb->lwb_flags & LWB_FLAG_CRASHED);
list_remove(&zilog->zl_lwb_crash_list, lwb);
if (lwb->lwb_buf != NULL)
zio_buf_free(lwb->lwb_buf, lwb->lwb_sz);
@@ -2981,7 +3047,7 @@ zil_prune_commit_list(zilog_t *zilog)
* never any itx's for it to wait on), so it's
* safe to skip this waiter and mark it done.
*/
- zil_commit_waiter_skip(itx->itx_private);
+ zil_commit_waiter_done(itx->itx_private, 0);
} else {
zil_commit_waiter_link_lwb(itx->itx_private, last_lwb);
}
@@ -3212,15 +3278,21 @@ zil_process_commit_list(zilog_t *zilog, zil_commit_waiter_t *zcw, list_t *ilwbs)
* "next" lwb on-disk. When this happens, we must stall
* the ZIL write pipeline; see the comment within
* zil_commit_writer_stall() for more details.
+ *
+ * ESHUTDOWN has to be handled carefully here. If we get it,
+ * then the pool suspended and zil_crash() was called, so we
+ * need to stop trying and just get an error back to the
+ * callers.
*/
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_lwb_write_issue(zilog, lwb);
}
- if (err == 0)
+ if (err != ESHUTDOWN)
err = zil_commit_writer_stall(zilog);
+ if (err == ESHUTDOWN)
+ err = SET_ERROR(EIO);
/*
* Additionally, we have to signal and mark the "nolwb"
@@ -3230,7 +3302,7 @@ zil_process_commit_list(zilog_t *zilog, zil_commit_waiter_t *zcw, list_t *ilwbs)
*/
zil_commit_waiter_t *zcw;
while ((zcw = list_remove_head(&nolwb_waiters)) != NULL)
- zil_commit_waiter_skip(zcw);
+ zil_commit_waiter_done(zcw, err);
/*
* And finally, we have to destroy the itx's that
@@ -3238,7 +3310,7 @@ 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, 0);
+ zil_itx_destroy(itx, err);
} else {
ASSERT(list_is_empty(&nolwb_waiters));
ASSERT3P(lwb, !=, NULL);
@@ -3292,17 +3364,17 @@ zil_process_commit_list(zilog_t *zilog, zil_commit_waiter_t *zcw, list_t *ilwbs)
(!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);
+ lwb = zil_lwb_write_close(zilog, lwb);
if (lwb == NULL) {
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_lwb_write_issue(
+ zilog, lwb);
}
- if (err == 0)
- zil_commit_writer_stall(zilog);
+ if (err != ESHUTDOWN)
+ (void) zil_commit_writer_stall(zilog);
}
}
}
@@ -3470,7 +3542,7 @@ zil_commit_waiter_timeout(zilog_t *zilog, zil_commit_waiter_t *zcw)
* hasn't been issued.
*/
zil_burst_done(zilog);
- lwb_t *nlwb = zil_lwb_write_close(zilog, lwb, LWB_STATE_NEW);
+ lwb_t *nlwb = zil_lwb_write_close(zilog, lwb);
ASSERT3S(lwb->lwb_state, ==, LWB_STATE_CLOSED);
@@ -3546,7 +3618,7 @@ zil_commit_waiter(zilog_t *zilog, zil_commit_waiter_t *zcw)
* commit itxs. When this occurs, the commit waiters linked
* off of these commit itxs will not be committed to an
* lwb. Additionally, these commit waiters will not be
- * marked done until zil_commit_waiter_skip() is called via
+ * marked done until zil_commit_waiter_done() is called via
* zil_itxg_clean().
*
* Thus, it's possible for this commit waiter (i.e. the
@@ -3624,7 +3696,7 @@ zil_alloc_commit_waiter(void)
list_link_init(&zcw->zcw_node);
zcw->zcw_lwb = NULL;
zcw->zcw_done = B_FALSE;
- zcw->zcw_zio_error = 0;
+ zcw->zcw_error = 0;
return (zcw);
}
@@ -3728,6 +3800,9 @@ zil_crash(zilog_t *zilog)
*/
for (lwb_t *lwb = list_head(&zilog->zl_lwb_crash_list); lwb != NULL;
lwb = list_next(&zilog->zl_lwb_crash_list, lwb)) {
+ ASSERT(!(lwb->lwb_flags & LWB_FLAG_CRASHED));
+ lwb->lwb_flags |= LWB_FLAG_CRASHED;
+
itx_t *itx;
while ((itx = list_remove_head(&lwb->lwb_itxs)) != NULL)
zil_itx_destroy(itx, EIO);
@@ -3736,7 +3811,7 @@ zil_crash(zilog_t *zilog)
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_error = EIO;
zcw->zcw_done = B_TRUE;
cv_broadcast(&zcw->zcw_cv);
mutex_exit(&zcw->zcw_lock);
@@ -4014,7 +4089,7 @@ zil_commit_impl(zilog_t *zilog, uint64_t foid)
zil_commit_waiter(zilog, zcw);
int err = 0;
- if (zcw->zcw_zio_error != 0) {
+ if (zcw->zcw_error != 0) {
/*
* If there was an error writing out the ZIL blocks that
* this thread is waiting on, then we fallback to
@@ -4149,7 +4224,7 @@ zil_lwb_cons(void *vbuf, void *unused, int kmflag)
offsetof(zil_commit_waiter_t, zcw_node));
avl_create(&lwb->lwb_vdev_tree, zil_lwb_vdev_compare,
sizeof (zil_vdev_node_t), offsetof(zil_vdev_node_t, zv_node));
- mutex_init(&lwb->lwb_vdev_lock, NULL, MUTEX_DEFAULT, NULL);
+ mutex_init(&lwb->lwb_lock, NULL, MUTEX_DEFAULT, NULL);
return (0);
}
@@ -4158,7 +4233,7 @@ zil_lwb_dest(void *vbuf, void *unused)
{
(void) unused;
lwb_t *lwb = vbuf;
- mutex_destroy(&lwb->lwb_vdev_lock);
+ mutex_destroy(&lwb->lwb_lock);
avl_destroy(&lwb->lwb_vdev_tree);
list_destroy(&lwb->lwb_waiters);
list_destroy(&lwb->lwb_itxs);
@@ -4381,7 +4456,7 @@ zil_close(zilog_t *zilog)
if (lwb != NULL) {
ASSERT(list_is_empty(&zilog->zl_lwb_list));
ASSERT3S(lwb->lwb_state, ==, LWB_STATE_NEW);
- zio_buf_free(lwb->lwb_buf, lwb->lwb_sz);
+ ASSERT0P(lwb->lwb_buf);
zil_free_lwb(zilog, lwb);
}
mutex_exit(&zilog->zl_lock);
@@ -4472,16 +4547,16 @@ zil_suspend(const char *osname, void **cookiep)
cv_wait(&zilog->zl_cv_suspend, &zilog->zl_lock);
mutex_exit(&zilog->zl_lock);
- if (cookiep == NULL)
+ if (zilog->zl_restart_txg > 0) {
+ /* ZIL crashed while we were waiting. */
+ zil_resume(os);
+ error = SET_ERROR(EBUSY);
+ } else if (cookiep == NULL)
zil_resume(os);
else
*cookiep = os;
- if (zilog->zl_restart_txg > 0)
- /* ZIL crashed while we were waiting. */
- return (SET_ERROR(EBUSY));
-
- return (0);
+ return (error);
}
/*
diff --git a/sys/contrib/openzfs/module/zfs/zio.c b/sys/contrib/openzfs/module/zfs/zio.c
index 3f0ddb63249d..4cf8912d4269 100644
--- a/sys/contrib/openzfs/module/zfs/zio.c
+++ b/sys/contrib/openzfs/module/zfs/zio.c
@@ -4434,12 +4434,15 @@ zio_dva_unallocate(zio_t *zio, zio_gang_node_t *gn, blkptr_t *bp)
*/
int
zio_alloc_zil(spa_t *spa, objset_t *os, uint64_t txg, blkptr_t *new_bp,
- uint64_t size, boolean_t *slog)
+ uint64_t min_size, uint64_t max_size, boolean_t *slog,
+ boolean_t allow_larger)
{
int error;
zio_alloc_list_t io_alloc_list;
+ uint64_t alloc_size = 0;
ASSERT(txg > spa_syncing_txg(spa));
+ ASSERT3U(min_size, <=, max_size);
metaslab_trace_init(&io_alloc_list);
@@ -4448,7 +4451,7 @@ zio_alloc_zil(spa_t *spa, objset_t *os, uint64_t txg, blkptr_t *new_bp,
* Fill in the obvious ones before calling into metaslab_alloc().
*/
BP_SET_TYPE(new_bp, DMU_OT_INTENT_LOG);
- BP_SET_PSIZE(new_bp, size);
+ BP_SET_PSIZE(new_bp, max_size);
BP_SET_LEVEL(new_bp, 0);
/*
@@ -4463,43 +4466,51 @@ zio_alloc_zil(spa_t *spa, objset_t *os, uint64_t txg, blkptr_t *new_bp,
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);
+ error = metaslab_alloc_range(spa, spa_log_class(spa), min_size,
+ max_size, new_bp, 1, txg, NULL, flags, &io_alloc_list, allocator,
+ NULL, &alloc_size);
*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);
+ error = metaslab_alloc_range(spa,
+ spa_special_embedded_log_class(spa), min_size, max_size,
+ new_bp, 1, txg, NULL, flags, &io_alloc_list, allocator,
+ NULL, &alloc_size);
}
/* 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);
+ error = metaslab_alloc_range(spa, spa_special_class(spa),
+ min_size, max_size, new_bp, 1, txg, NULL, flags,
+ &io_alloc_list, allocator, NULL, &alloc_size);
}
/* 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);
+ error = metaslab_alloc_range(spa, spa_embedded_log_class(spa),
+ min_size, max_size, new_bp, 1, txg, NULL, flags,
+ &io_alloc_list, allocator, NULL, &alloc_size);
}
/* Finally fall back to normal class */
if (error != 0) {
ZIOSTAT_BUMP(ziostat_alloc_class_fallbacks);
- error = metaslab_alloc(spa, spa_normal_class(spa), size,
- new_bp, 1, txg, NULL, flags, &io_alloc_list, allocator,
- NULL);
+ error = metaslab_alloc_range(spa, spa_normal_class(spa),
+ min_size, max_size, new_bp, 1, txg, NULL, flags,
+ &io_alloc_list, allocator, NULL, &alloc_size);
}
metaslab_trace_fini(&io_alloc_list);
if (error == 0) {
- BP_SET_LSIZE(new_bp, size);
- BP_SET_PSIZE(new_bp, size);
+ if (!allow_larger)
+ alloc_size = MIN(alloc_size, max_size);
+ else if (max_size <= SPA_OLD_MAXBLOCKSIZE)
+ alloc_size = MIN(alloc_size, SPA_OLD_MAXBLOCKSIZE);
+ alloc_size = P2ALIGN_TYPED(alloc_size, ZIL_MIN_BLKSZ, uint64_t);
+
+ BP_SET_LSIZE(new_bp, alloc_size);
+ BP_SET_PSIZE(new_bp, alloc_size);
BP_SET_COMPRESS(new_bp, ZIO_COMPRESS_OFF);
BP_SET_CHECKSUM(new_bp,
spa_version(spa) >= SPA_VERSION_SLIM_ZIL
@@ -4527,8 +4538,8 @@ zio_alloc_zil(spa_t *spa, objset_t *os, uint64_t txg, blkptr_t *new_bp,
}
} else {
zfs_dbgmsg("%s: zil block allocation failure: "
- "size %llu, error %d", spa_name(spa), (u_longlong_t)size,
- error);
+ "min_size %llu, max_size %llu, error %d", spa_name(spa),
+ (u_longlong_t)min_size, (u_longlong_t)max_size, error);
}
return (error);
diff --git a/sys/contrib/openzfs/module/zfs/zvol.c b/sys/contrib/openzfs/module/zfs/zvol.c
index 29f51e230a37..2fd3e1c37045 100644
--- a/sys/contrib/openzfs/module/zfs/zvol.c
+++ b/sys/contrib/openzfs/module/zfs/zvol.c
@@ -38,25 +38,36 @@
* Copyright 2014 Nexenta Systems, Inc. All rights reserved.
* Copyright (c) 2016 Actifio, Inc. All rights reserved.
* Copyright (c) 2012, 2019 by Delphix. All rights reserved.
- * Copyright (c) 2024, Klara, Inc.
+ * Copyright (c) 2024, 2025, Klara, Inc.
*/
/*
* Note on locking of zvol state structures.
*
- * These structures are used to maintain internal state used to emulate block
- * devices on top of zvols. In particular, management of device minor number
- * operations - create, remove, rename, and set_snapdev - involves access to
- * these structures. The zvol_state_lock is primarily used to protect the
- * zvol_state_list. The zv->zv_state_lock is used to protect the contents
- * of the zvol_state_t structures, as well as to make sure that when the
- * time comes to remove the structure from the list, it is not in use, and
- * therefore, it can be taken off zvol_state_list and freed.
+ * zvol_state_t represents the connection between a single dataset
+ * (DMU_OST_ZVOL) and the device "minor" (some OS-specific representation of a
+ * "disk" or "device" or "volume", eg, a /dev/zdXX node, a GEOM object, etc).
*
- * The zv_suspend_lock was introduced to allow for suspending I/O to a zvol,
- * e.g. for the duration of receive and rollback operations. This lock can be
- * held for significant periods of time. Given that it is undesirable to hold
- * mutexes for long periods of time, the following lock ordering applies:
+ * The global zvol_state_lock is used to protect access to zvol_state_list and
+ * zvol_htable, which are the primary way to obtain a zvol_state_t from a name.
+ * It should not be used for anything not name-relateds, and you should avoid
+ * sleeping or waiting while its held. See zvol_find_by_name(), zvol_insert(),
+ * zvol_remove().
+ *
+ * The zv_state_lock is used to protect the contents of the associated
+ * zvol_state_t. Most of the zvol_state_t is dedicated to control and
+ * configuration; almost none of it is needed for data operations (that is,
+ * read, write, flush) so this lock is rarely taken during general IO. It
+ * should be released quickly; you should avoid sleeping or waiting while its
+ * held.
+ *
+ * zv_suspend_lock is used to suspend IO/data operations to a zvol. The read
+ * half should held for the duration of an IO operation. The write half should
+ * be taken when something to wait for IO to complete and the block further IO,
+ * eg for the duration of receive and rollback operations. This lock can be
+ * held for long periods of time.
+ *
+ * Thus, the following lock ordering appies.
* - take zvol_state_lock if necessary, to protect zvol_state_list
* - take zv_suspend_lock if necessary, by the code path in question
* - take zv_state_lock to protect zvol_state_t
@@ -67,9 +78,8 @@
* these operations are serialized per pool. Consequently, we can be certain
* that for a given zvol, there is only one operation at a time in progress.
* That is why one can be sure that first, zvol_state_t for a given zvol is
- * allocated and placed on zvol_state_list, and then other minor operations
- * for this zvol are going to proceed in the order of issue.
- *
+ * allocated and placed on zvol_state_list, and then other minor operations for
+ * this zvol are going to proceed in the order of issue.
*/
#include <sys/dataset_kstats.h>
@@ -1570,184 +1580,156 @@ zvol_create_minors_impl(zvol_task_t *task)
}
/*
- * Remove minors for specified dataset including children and snapshots.
- */
-
-/*
- * Remove the minor for a given zvol. This will do it all:
- * - flag the zvol for removal, so new requests are rejected
- * - wait until outstanding requests are completed
- * - remove it from lists
- * - free it
- * It's also usable as a taskq task, and smells nice too.
+ * Remove minors for specified dataset and, optionally, its children and
+ * snapshots.
*/
static void
-zvol_remove_minor_task(void *arg)
-{
- zvol_state_t *zv = (zvol_state_t *)arg;
-
- ASSERT(!RW_LOCK_HELD(&zvol_state_lock));
- ASSERT(!MUTEX_HELD(&zv->zv_state_lock));
-
- mutex_enter(&zv->zv_state_lock);
- while (zv->zv_open_count > 0 || atomic_read(&zv->zv_suspend_ref)) {
- zv->zv_flags |= ZVOL_REMOVING;
- cv_wait(&zv->zv_removing_cv, &zv->zv_state_lock);
- }
- mutex_exit(&zv->zv_state_lock);
-
- rw_enter(&zvol_state_lock, RW_WRITER);
- mutex_enter(&zv->zv_state_lock);
-
- zvol_remove(zv);
- zvol_os_clear_private(zv);
-
- mutex_exit(&zv->zv_state_lock);
- rw_exit(&zvol_state_lock);
-
- zvol_os_free(zv);
-}
-
-static void
-zvol_free_task(void *arg)
-{
- zvol_os_free(arg);
-}
-
-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;
+ boolean_t children = task ? !!task->zt_value : B_TRUE;
if (zvol_inhibit_dev)
return;
- list_create(&delay_list, sizeof (zvol_state_t),
- offsetof(zvol_state_t, zv_next));
- list_create(&free_list, sizeof (zvol_state_t),
- offsetof(zvol_state_t, zv_next));
+ /*
+ * We collect up zvols that we want to remove on a separate list, so
+ * that we don't have to hold zvol_state_lock for the whole time.
+ *
+ * We can't remove them from the global lists until we're completely
+ * done with them, because that would make them appear to ZFS-side ops
+ * that they don't exist, and the name might be reused, which can't be
+ * good.
+ */
+ list_t remove_list;
+ list_create(&remove_list, sizeof (zvol_state_t),
+ offsetof(zvol_state_t, zv_remove_node));
- rw_enter(&zvol_state_lock, RW_WRITER);
+ rw_enter(&zvol_state_lock, RW_READER);
for (zv = list_head(&zvol_state_list); zv != NULL; zv = zv_next) {
zv_next = list_next(&zvol_state_list, zv);
mutex_enter(&zv->zv_state_lock);
+ if (zv->zv_flags & ZVOL_REMOVING) {
+ /* Another thread is handling shutdown, skip it. */
+ mutex_exit(&zv->zv_state_lock);
+ continue;
+ }
+
+ /*
+ * This zvol should be removed if:
+ * - no name was offered (ie removing all at shutdown); or
+ * - name matches exactly; or
+ * - we were asked to remove children, and
+ * - the start of the name matches, and
+ * - there is a '/' immediately after the matched name; or
+ * - there is a '@' immediately after the matched name
+ */
if (name == NULL || strcmp(zv->zv_name, name) == 0 ||
- (strncmp(zv->zv_name, name, namelen) == 0 &&
+ (children && strncmp(zv->zv_name, name, namelen) == 0 &&
(zv->zv_name[namelen] == '/' ||
zv->zv_name[namelen] == '@'))) {
- /*
- * By holding zv_state_lock here, we guarantee that no
- * one is currently using this zv
- */
/*
- * If in use, try to throw everyone off and try again
- * later.
+ * Matched, so mark it removal. We want to take the
+ * write half of the suspend lock to make sure that
+ * the zvol is not suspended, and give any data ops
+ * chance to finish.
*/
- if (zv->zv_open_count > 0 ||
- atomic_read(&zv->zv_suspend_ref)) {
- zv->zv_flags |= ZVOL_REMOVING;
- t = taskq_dispatch(
- zv->zv_objset->os_spa->spa_zvol_taskq,
- zvol_remove_minor_task, zv, TQ_SLEEP);
- if (t == TASKQID_INVALID) {
- /*
- * Couldn't create the task, so we'll
- * do it in place once the loop is
- * finished.
- */
- list_insert_head(&delay_list, zv);
- }
+ mutex_exit(&zv->zv_state_lock);
+ rw_enter(&zv->zv_suspend_lock, RW_WRITER);
+ mutex_enter(&zv->zv_state_lock);
+
+ if (zv->zv_flags & ZVOL_REMOVING) {
+ /* Another thread has taken it, let them. */
mutex_exit(&zv->zv_state_lock);
+ rw_exit(&zv->zv_suspend_lock);
continue;
}
- zvol_remove(zv);
-
/*
- * Cleared while holding zvol_state_lock as a writer
- * which will prevent zvol_open() from opening it.
+ * Mark it and unlock. New entries will see the flag
+ * and return ENXIO.
*/
- zvol_os_clear_private(zv);
-
- /* Drop zv_state_lock before zvol_free() */
+ zv->zv_flags |= ZVOL_REMOVING;
mutex_exit(&zv->zv_state_lock);
+ rw_exit(&zv->zv_suspend_lock);
- /* Try parallel zv_free, if failed do it in place */
- t = taskq_dispatch(system_taskq, zvol_free_task, zv,
- TQ_SLEEP);
- if (t == TASKQID_INVALID)
- list_insert_head(&free_list, zv);
- } else {
+ /* Put it on the list for the next stage. */
+ list_insert_head(&remove_list, zv);
+ } else
mutex_exit(&zv->zv_state_lock);
- }
}
- rw_exit(&zvol_state_lock);
- /* Wait for zvols that we couldn't create a remove task for */
- while ((zv = list_remove_head(&delay_list)) != NULL)
- zvol_remove_minor_task(zv);
-
- /* Free any that we couldn't free in parallel earlier */
- while ((zv = list_remove_head(&free_list)) != NULL)
- zvol_os_free(zv);
-}
-
-/* Remove minor for this specific volume only */
-static int
-zvol_remove_minor_impl(const char *name)
-{
- zvol_state_t *zv = NULL, *zv_next;
-
- if (zvol_inhibit_dev)
- return (0);
+ rw_exit(&zvol_state_lock);
- rw_enter(&zvol_state_lock, RW_WRITER);
+ /* Didn't match any, nothing to do! */
+ if (list_is_empty(&remove_list)) {
+ if (task)
+ task->zt_error = SET_ERROR(ENOENT);
+ return;
+ }
- for (zv = list_head(&zvol_state_list); zv != NULL; zv = zv_next) {
- zv_next = list_next(&zvol_state_list, zv);
+ /* Actually shut them all down. */
+ for (zv = list_head(&remove_list); zv != NULL; zv = zv_next) {
+ zv_next = list_next(&remove_list, zv);
mutex_enter(&zv->zv_state_lock);
- if (strcmp(zv->zv_name, name) == 0)
- /* Found, leave the the loop with zv_lock held */
- break;
- mutex_exit(&zv->zv_state_lock);
- }
-
- if (zv == NULL) {
- rw_exit(&zvol_state_lock);
- return (SET_ERROR(ENOENT));
- }
- ASSERT(MUTEX_HELD(&zv->zv_state_lock));
+ /*
+ * Still open or suspended, just wait. This can happen if, for
+ * example, we managed to acquire zv_state_lock in the moments
+ * where zvol_open() or zvol_release() are trading locks to
+ * call zvol_first_open() or zvol_last_close().
+ */
+ while (zv->zv_open_count > 0 ||
+ atomic_read(&zv->zv_suspend_ref))
+ cv_wait(&zv->zv_removing_cv, &zv->zv_state_lock);
- if (zv->zv_open_count > 0 || atomic_read(&zv->zv_suspend_ref)) {
/*
- * In use, so try to throw everyone off, then wait
- * until finished.
+ * No users, shut down the OS side. This may not remove the
+ * minor from view immediately, depending on the kernel
+ * specifics, but it will ensure that it is unusable and that
+ * this zvol_state_t can never again be reached from an OS-side
+ * operation.
*/
- zv->zv_flags |= ZVOL_REMOVING;
+ zvol_os_remove_minor(zv);
mutex_exit(&zv->zv_state_lock);
+
+ /* Remove it from the name lookup lists */
+ rw_enter(&zvol_state_lock, RW_WRITER);
+ zvol_remove(zv);
rw_exit(&zvol_state_lock);
- zvol_remove_minor_task(zv);
- return (0);
}
- zvol_remove(zv);
- zvol_os_clear_private(zv);
+ /*
+ * Our own references on remove_list is the last one, free them and
+ * we're done.
+ */
+ while ((zv = list_remove_head(&remove_list)) != NULL)
+ zvol_os_free(zv);
- mutex_exit(&zv->zv_state_lock);
- rw_exit(&zvol_state_lock);
+ list_destroy(&remove_list);
+}
- zvol_os_free(zv);
+/* Remove minor for this specific volume only */
+static int
+zvol_remove_minor_impl(const char *name)
+{
+ if (zvol_inhibit_dev)
+ return (0);
- return (0);
+ zvol_task_t task;
+ memset(&task, 0, sizeof (zvol_task_t));
+ strlcpy(task.zt_name1, name, sizeof (task.zt_name1));
+ task.zt_value = B_FALSE;
+
+ zvol_remove_minors_impl(&task);
+
+ return (task.zt_error);
}
/*
@@ -2067,6 +2049,7 @@ zvol_remove_minors(spa_t *spa, const char *name, boolean_t async)
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));
+ task->zt_value = B_TRUE;
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);
@@ -2188,14 +2171,6 @@ zvol_fini_impl(void)
zvol_remove_minors_impl(NULL);
- /*
- * The call to "zvol_remove_minors_impl" may dispatch entries to
- * the system_taskq, but it doesn't wait for those entries to
- * complete before it returns. Thus, we must wait for all of the
- * removals to finish, before we can continue.
- */
- taskq_wait_outstanding(system_taskq, 0);
-
kmem_free(zvol_htable, ZVOL_HT_SIZE * sizeof (struct hlist_head));
list_destroy(&zvol_state_list);
rw_destroy(&zvol_state_lock);
diff --git a/sys/contrib/openzfs/module/zstd/zfs_zstd.c b/sys/contrib/openzfs/module/zstd/zfs_zstd.c
index 391216d6e263..3db196953f74 100644
--- a/sys/contrib/openzfs/module/zstd/zfs_zstd.c
+++ b/sys/contrib/openzfs/module/zstd/zfs_zstd.c
@@ -876,9 +876,9 @@ static void __init
zstd_mempool_init(void)
{
zstd_mempool_cctx =
- kmem_zalloc(ZSTD_POOL_MAX * sizeof (struct zstd_pool), KM_SLEEP);
+ vmem_zalloc(ZSTD_POOL_MAX * sizeof (struct zstd_pool), KM_SLEEP);
zstd_mempool_dctx =
- kmem_zalloc(ZSTD_POOL_MAX * sizeof (struct zstd_pool), KM_SLEEP);
+ vmem_zalloc(ZSTD_POOL_MAX * sizeof (struct zstd_pool), KM_SLEEP);
for (int i = 0; i < ZSTD_POOL_MAX; i++) {
mutex_init(&zstd_mempool_cctx[i].barrier, NULL,
@@ -924,8 +924,8 @@ zstd_mempool_deinit(void)
release_pool(&zstd_mempool_dctx[i]);
}
- kmem_free(zstd_mempool_dctx, ZSTD_POOL_MAX * sizeof (struct zstd_pool));
- kmem_free(zstd_mempool_cctx, ZSTD_POOL_MAX * sizeof (struct zstd_pool));
+ vmem_free(zstd_mempool_dctx, ZSTD_POOL_MAX * sizeof (struct zstd_pool));
+ vmem_free(zstd_mempool_cctx, ZSTD_POOL_MAX * sizeof (struct zstd_pool));
zstd_mempool_dctx = NULL;
zstd_mempool_cctx = NULL;
}
diff --git a/sys/contrib/openzfs/scripts/spdxcheck.pl b/sys/contrib/openzfs/scripts/spdxcheck.pl
index 88f5a235d70c..cdab5368f19c 100755
--- a/sys/contrib/openzfs/scripts/spdxcheck.pl
+++ b/sys/contrib/openzfs/scripts/spdxcheck.pl
@@ -190,6 +190,7 @@ my @path_license_tags = (
['BSD-2-Clause OR GPL-2.0-only', 'CDDL-1.0'],
'module/icp' => ['Apache-2.0', 'CDDL-1.0'],
+ 'contrib/icp' => ['Apache-2.0', 'CDDL-1.0'],
# Python bindings are always Apache-2.0
'contrib/pyzfs' => ['Apache-2.0'],
diff --git a/sys/contrib/openzfs/tests/runfiles/common.run b/sys/contrib/openzfs/tests/runfiles/common.run
index 131845f5ed40..2da46458289a 100644
--- a/sys/contrib/openzfs/tests/runfiles/common.run
+++ b/sys/contrib/openzfs/tests/runfiles/common.run
@@ -1093,7 +1093,7 @@ tests = ['zvol_misc_002_pos', 'zvol_misc_hierarchy', 'zvol_misc_rename_inuse',
tags = ['functional', 'zvol', 'zvol_misc']
[tests/functional/zvol/zvol_stress]
-tests = ['zvol_stress']
+tests = ['zvol_stress', 'zvol_stress_destroy']
tags = ['functional', 'zvol', 'zvol_stress']
[tests/functional/zvol/zvol_swap]
diff --git a/sys/contrib/openzfs/tests/zfs-tests/cmd/crypto_test.c b/sys/contrib/openzfs/tests/zfs-tests/cmd/crypto_test.c
index e08003f80464..cbebd33e0bf6 100644
--- a/sys/contrib/openzfs/tests/zfs-tests/cmd/crypto_test.c
+++ b/sys/contrib/openzfs/tests/zfs-tests/cmd/crypto_test.c
@@ -529,6 +529,8 @@ static const char *aes_gcm_impl[][2] = {
{ "aesni", "pclmulqdq" },
{ "x86_64", "avx" },
{ "aesni", "avx" },
+ { "x86_64", "avx2" },
+ { "aesni", "avx2" },
};
/* signature of function to call after setting implementation params */
diff --git a/sys/contrib/openzfs/tests/zfs-tests/tests/Makefile.am b/sys/contrib/openzfs/tests/zfs-tests/tests/Makefile.am
index b8b8bbe45a42..41e7b45ef4ec 100644
--- a/sys/contrib/openzfs/tests/zfs-tests/tests/Makefile.am
+++ b/sys/contrib/openzfs/tests/zfs-tests/tests/Makefile.am
@@ -2244,6 +2244,7 @@ nobase_dist_datadir_zfs_tests_tests_SCRIPTS += \
functional/zvol/zvol_stress/cleanup.ksh \
functional/zvol/zvol_stress/setup.ksh \
functional/zvol/zvol_stress/zvol_stress.ksh \
+ functional/zvol/zvol_stress/zvol_stress_destroy.ksh \
functional/zvol/zvol_swap/cleanup.ksh \
functional/zvol/zvol_swap/setup.ksh \
functional/zvol/zvol_swap/zvol_swap_001_pos.ksh \
diff --git a/sys/contrib/openzfs/tests/zfs-tests/tests/functional/zvol/zvol_stress/zvol_stress_destroy.ksh b/sys/contrib/openzfs/tests/zfs-tests/tests/functional/zvol/zvol_stress/zvol_stress_destroy.ksh
new file mode 100755
index 000000000000..669b59fac01f
--- /dev/null
+++ b/sys/contrib/openzfs/tests/zfs-tests/tests/functional/zvol/zvol_stress/zvol_stress_destroy.ksh
@@ -0,0 +1,66 @@
+#!/bin/ksh -p
+# 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) 2025, Klara, Inc.
+#
+
+. $STF_SUITE/include/libtest.shlib
+
+verify_runnable "global"
+
+typeset -i nzvols=1000
+typeset -i parallel=$(( $(get_num_cpus) * 2 ))
+
+function cleanup {
+ for zvol in $(zfs list -Ho name -t vol) ; do
+ log_must_busy zfs destroy $zvol
+ done
+}
+
+log_onexit cleanup
+
+log_assert "stress test concurrent zvol create/destroy"
+
+function destroy_zvols_until {
+ typeset cond=$1
+ while true ; do
+ IFS='' zfs list -Ho name -t vol | read -r -d '' zvols
+ if [[ -n $zvols ]] ; then
+ echo $zvols | xargs -n 1 -P $parallel zfs destroy
+ fi
+ if ! $cond ; then
+ break
+ fi
+ done
+}
+
+( seq $nzvols | \
+ xargs -P $parallel -I % zfs create -s -V 1G $TESTPOOL/testvol% ) &
+cpid=$!
+sleep 1
+
+destroy_zvols_until "kill -0 $cpid"
+destroy_zvols_until "false"
+
+log_pass "stress test done"
diff --git a/sys/dev/gpio/acpi_gpiobus.c b/sys/dev/gpio/acpi_gpiobus.c
index 170f23615416..0d2455cab399 100644
--- a/sys/dev/gpio/acpi_gpiobus.c
+++ b/sys/dev/gpio/acpi_gpiobus.c
@@ -37,6 +37,7 @@
#include <dev/gpio/gpiobusvar.h>
#include <dev/gpio/acpi_gpiobusvar.h>
#include <dev/gpio/gpiobus_internal.h>
+#include <sys/sbuf.h>
#include "gpiobus_if.h"
@@ -52,12 +53,11 @@ struct acpi_gpiobus_ctx {
struct acpi_gpiobus_ivar
{
- struct gpiobus_ivar gpiobus; /* Must come first */
- ACPI_HANDLE dev_handle; /* ACPI handle for bus */
- uint32_t flags;
+ struct gpiobus_ivar gpiobus;
+ ACPI_HANDLE handle;
};
-static uint32_t
+uint32_t
acpi_gpiobus_convflags(ACPI_RESOURCE_GPIO *gpio_res)
{
uint32_t flags = 0;
@@ -150,70 +150,24 @@ acpi_gpiobus_enumerate_res(ACPI_RESOURCE *res, void *context)
return (AE_OK);
}
-static struct acpi_gpiobus_ivar *
-acpi_gpiobus_setup_devinfo(device_t bus, device_t child,
- ACPI_RESOURCE_GPIO *gpio_res)
-{
- struct acpi_gpiobus_ivar *devi;
-
- devi = malloc(sizeof(*devi), M_DEVBUF, M_NOWAIT | M_ZERO);
- if (devi == NULL)
- return (NULL);
- resource_list_init(&devi->gpiobus.rl);
-
- devi->flags = acpi_gpiobus_convflags(gpio_res);
- if (acpi_quirks & ACPI_Q_AEI_NOPULL)
- devi->flags &= ~GPIO_PIN_PULLUP;
-
- devi->gpiobus.npins = 1;
- if (gpiobus_alloc_ivars(&devi->gpiobus) != 0) {
- free(devi, M_DEVBUF);
- return (NULL);
- }
-
- for (int i = 0; i < devi->gpiobus.npins; i++)
- devi->gpiobus.pins[i] = gpio_res->PinTable[i];
-
- return (devi);
-}
-
static ACPI_STATUS
acpi_gpiobus_enumerate_aei(ACPI_RESOURCE *res, void *context)
{
ACPI_RESOURCE_GPIO *gpio_res = &res->Data.Gpio;
- struct acpi_gpiobus_ctx *ctx = context;
- device_t bus = ctx->sc->sc_busdev;
- device_t child;
- struct acpi_gpiobus_ivar *devi;
+ uint32_t *npins = context, *pins = npins + 1;
- /* Check that we have a GpioInt object. */
+ /*
+ * Check that we have a GpioInt object.
+ * Note that according to the spec this
+ * should always be the case.
+ */
if (res->Type != ACPI_RESOURCE_TYPE_GPIO)
return (AE_OK);
if (gpio_res->ConnectionType != ACPI_RESOURCE_GPIO_TYPE_INT)
return (AE_OK);
- /* Add a child. */
- child = device_add_child_ordered(bus, 0, "gpio_aei", DEVICE_UNIT_ANY);
- if (child == NULL)
- return (AE_OK);
- devi = acpi_gpiobus_setup_devinfo(bus, child, gpio_res);
- if (devi == NULL) {
- device_delete_child(bus, child);
- return (AE_OK);
- }
- device_set_ivars(child, devi);
-
- for (int i = 0; i < devi->gpiobus.npins; i++) {
- if (GPIOBUS_PIN_SETFLAGS(bus, child, 0, devi->flags &
- ~GPIO_INTR_MASK)) {
- device_delete_child(bus, child);
- return (AE_OK);
- }
- }
-
- /* Pass ACPI information to children. */
- devi->dev_handle = ctx->dev_handle;
-
+ for (int i = 0; i < gpio_res->PinTableLength; i++)
+ pins[(*npins)++] = gpio_res->PinTable[i];
return (AE_OK);
}
@@ -296,6 +250,63 @@ err:
return (AE_BAD_PARAMETER);
}
+static void
+acpi_gpiobus_attach_aei(struct acpi_gpiobus_softc *sc, ACPI_HANDLE handle)
+{
+ struct acpi_gpiobus_ivar *devi;
+ ACPI_HANDLE aei_handle;
+ device_t child;
+ uint32_t *pins;
+ ACPI_STATUS status;
+ int err;
+
+ status = AcpiGetHandle(handle, "_AEI", &aei_handle);
+ if (ACPI_FAILURE(status))
+ return;
+
+ /* pins[0] specifies the length of the array. */
+ pins = mallocarray(sc->super_sc.sc_npins + 1,
+ sizeof(uint32_t), M_DEVBUF, M_WAITOK);
+ pins[0] = 0;
+
+ status = AcpiWalkResources(handle, "_AEI",
+ acpi_gpiobus_enumerate_aei, pins);
+ if (ACPI_FAILURE(status)) {
+ device_printf(sc->super_sc.sc_busdev,
+ "Failed to enumerate AEI resources\n");
+ free(pins, M_DEVBUF);
+ return;
+ }
+
+ child = BUS_ADD_CHILD(sc->super_sc.sc_busdev, 0, "gpio_aei",
+ DEVICE_UNIT_ANY);
+ if (child == NULL) {
+ device_printf(sc->super_sc.sc_busdev,
+ "Failed to add gpio_aei child\n");
+ free(pins, M_DEVBUF);
+ return;
+ }
+
+ devi = device_get_ivars(child);
+ devi->gpiobus.npins = pins[0];
+ devi->handle = aei_handle;
+
+ err = gpiobus_alloc_ivars(&devi->gpiobus);
+ if (err != 0) {
+ device_printf(sc->super_sc.sc_busdev,
+ "Failed to allocate gpio_aei ivars\n");
+ device_delete_child(sc->super_sc.sc_busdev, child);
+ free(pins, M_DEVBUF);
+ return;
+ }
+
+ for (int i = 0; i < pins[0]; i++)
+ devi->gpiobus.pins[i] = pins[i + 1];
+ free(pins, M_DEVBUF);
+
+ bus_attach_children(sc->super_sc.sc_busdev);
+}
+
static int
acpi_gpiobus_probe(device_t dev)
{
@@ -353,13 +364,8 @@ acpi_gpiobus_attach(device_t dev)
if (ACPI_FAILURE(status))
device_printf(dev, "Failed to enumerate GPIO resources\n");
- /* Look for AEI children */
- status = AcpiWalkResources(handle, "_AEI", acpi_gpiobus_enumerate_aei,
- &ctx);
-
- if (ACPI_FAILURE(status) && status != AE_NOT_FOUND)
- device_printf(dev, "Failed to enumerate AEI resources\n");
-
+ /* Look for AEI child */
+ acpi_gpiobus_attach_aei(sc, handle);
return (0);
}
@@ -390,10 +396,7 @@ acpi_gpiobus_read_ivar(device_t dev, device_t child, int which,
switch (which) {
case ACPI_GPIOBUS_IVAR_HANDLE:
- *result = (uintptr_t)devi->dev_handle;
- break;
- case ACPI_GPIOBUS_IVAR_FLAGS:
- *result = (uintptr_t)devi->flags;
+ *result = (uintptr_t)devi->handle;
break;
default:
return (gpiobus_read_ivar(dev, child, which, result));
@@ -402,6 +405,28 @@ acpi_gpiobus_read_ivar(device_t dev, device_t child, int which,
return (0);
}
+static device_t
+acpi_gpiobus_add_child(device_t dev, u_int order, const char *name, int unit)
+{
+ return (gpiobus_add_child_common(dev, order, name, unit,
+ sizeof(struct acpi_gpiobus_ivar)));
+}
+
+static int
+acpi_gpiobus_child_location(device_t bus, device_t child, struct sbuf *sb)
+{
+ struct acpi_gpiobus_ivar *devi;
+ int err;
+
+ err = gpiobus_child_location(bus, child, sb);
+ if (err != 0)
+ return (err);
+
+ devi = device_get_ivars(child);
+ sbuf_printf(sb, " handle=%s", acpi_name(devi->handle));
+ return (0);
+}
+
static device_method_t acpi_gpiobus_methods[] = {
/* Device interface */
DEVMETHOD(device_probe, acpi_gpiobus_probe),
@@ -410,6 +435,8 @@ static device_method_t acpi_gpiobus_methods[] = {
/* Bus interface */
DEVMETHOD(bus_read_ivar, acpi_gpiobus_read_ivar),
+ DEVMETHOD(bus_add_child, acpi_gpiobus_add_child),
+ DEVMETHOD(bus_child_location, acpi_gpiobus_child_location),
DEVMETHOD_END
};
diff --git a/sys/dev/gpio/acpi_gpiobusvar.h b/sys/dev/gpio/acpi_gpiobusvar.h
index f8d502eab9d1..288e8bd0f2af 100644
--- a/sys/dev/gpio/acpi_gpiobusvar.h
+++ b/sys/dev/gpio/acpi_gpiobusvar.h
@@ -33,16 +33,16 @@
#include <contrib/dev/acpica/include/acpi.h>
enum acpi_gpiobus_ivars {
- ACPI_GPIOBUS_IVAR_HANDLE = 10600,
- ACPI_GPIOBUS_IVAR_FLAGS,
+ ACPI_GPIOBUS_IVAR_HANDLE = 10600
};
#define ACPI_GPIOBUS_ACCESSOR(var, ivar, type) \
__BUS_ACCESSOR(acpi_gpiobus, var, ACPI_GPIOBUS, ivar, type)
ACPI_GPIOBUS_ACCESSOR(handle, HANDLE, ACPI_HANDLE)
-ACPI_GPIOBUS_ACCESSOR(flags, FLAGS, uint32_t)
#undef ACPI_GPIOBUS_ACCESSOR
+uint32_t acpi_gpiobus_convflags(ACPI_RESOURCE_GPIO *);
+
#endif /* __ACPI_GPIOBUS_H__ */
diff --git a/sys/dev/gpio/gpioaei.c b/sys/dev/gpio/gpioaei.c
index ecae8ccaf2fa..7b97277aaf61 100644
--- a/sys/dev/gpio/gpioaei.c
+++ b/sys/dev/gpio/gpioaei.c
@@ -45,13 +45,21 @@ enum gpio_aei_type {
ACPI_AEI_TYPE_EVT
};
-struct gpio_aei_softc {
- ACPI_HANDLE handle;
- enum gpio_aei_type type;
- int pin;
+struct gpio_aei_ctx {
+ SLIST_ENTRY(gpio_aei_ctx) next;
struct resource * intr_res;
- int intr_rid;
void * intr_cookie;
+ ACPI_HANDLE handle;
+ gpio_pin_t gpio;
+ uint32_t pin;
+ int intr_rid;
+ enum gpio_aei_type type;
+};
+
+struct gpio_aei_softc {
+ SLIST_HEAD(, gpio_aei_ctx) aei_ctx;
+ ACPI_HANDLE dev_handle;
+ device_t dev;
};
static int
@@ -65,69 +73,157 @@ gpio_aei_probe(device_t dev)
static void
gpio_aei_intr(void * arg)
{
- struct gpio_aei_softc * sc = arg;
+ struct gpio_aei_ctx * ctx = arg;
/* Ask ACPI to run the appropriate _EVT, _Exx or _Lxx method. */
- if (sc->type == ACPI_AEI_TYPE_EVT)
- acpi_SetInteger(sc->handle, NULL, sc->pin);
+ if (ctx->type == ACPI_AEI_TYPE_EVT)
+ acpi_SetInteger(ctx->handle, NULL, ctx->pin);
else
- AcpiEvaluateObject(sc->handle, NULL, NULL, NULL);
+ AcpiEvaluateObject(ctx->handle, NULL, NULL, NULL);
+}
+
+static ACPI_STATUS
+gpio_aei_enumerate(ACPI_RESOURCE * res, void * context)
+{
+ ACPI_RESOURCE_GPIO * gpio_res = &res->Data.Gpio;
+ struct gpio_aei_softc * sc = context;
+ uint32_t flags, maxpin;
+ device_t busdev;
+ int err;
+
+ /*
+ * Check that we have a GpioInt object.
+ * Note that according to the spec this
+ * should always be the case.
+ */
+ if (res->Type != ACPI_RESOURCE_TYPE_GPIO)
+ return (AE_OK);
+ if (gpio_res->ConnectionType != ACPI_RESOURCE_GPIO_TYPE_INT)
+ return (AE_OK);
+
+ flags = acpi_gpiobus_convflags(gpio_res);
+ if (acpi_quirks & ACPI_Q_AEI_NOPULL)
+ flags &= ~GPIO_PIN_PULLUP;
+
+ err = GPIO_PIN_MAX(acpi_get_device(sc->dev_handle), &maxpin);
+ if (err != 0)
+ return (AE_ERROR);
+
+ busdev = GPIO_GET_BUS(acpi_get_device(sc->dev_handle));
+ for (int i = 0; i < gpio_res->PinTableLength; i++) {
+ struct gpio_aei_ctx * ctx;
+ uint32_t pin = gpio_res->PinTable[i];
+
+ if (__predict_false(pin > maxpin)) {
+ device_printf(sc->dev,
+ "Invalid pin 0x%x, max: 0x%x (bad ACPI tables?)\n",
+ pin, maxpin);
+ continue;
+ }
+
+ ctx = malloc(sizeof(struct gpio_aei_ctx), M_DEVBUF, M_WAITOK);
+ ctx->type = ACPI_AEI_TYPE_UNKNOWN;
+ if (pin <= 255) {
+ char objname[5]; /* "_EXX" or "_LXX" */
+ sprintf(objname, "_%c%02X",
+ (flags & GPIO_INTR_EDGE_MASK) ? 'E' : 'L', pin);
+ if (ACPI_SUCCESS(AcpiGetHandle(sc->dev_handle, objname,
+ &ctx->handle)))
+ ctx->type = ACPI_AEI_TYPE_ELX;
+ }
+
+ if (ctx->type == ACPI_AEI_TYPE_UNKNOWN) {
+ if (ACPI_SUCCESS(AcpiGetHandle(sc->dev_handle, "_EVT",
+ &ctx->handle)))
+ ctx->type = ACPI_AEI_TYPE_EVT;
+ else {
+ device_printf(sc->dev,
+ "AEI Device type is unknown for pin 0x%x\n",
+ pin);
+
+ free(ctx, M_DEVBUF);
+ continue;
+ }
+ }
+
+ err = gpio_pin_get_by_bus_pinnum(busdev, pin, &ctx->gpio);
+ if (err != 0) {
+ device_printf(sc->dev, "Cannot acquire pin 0x%x\n",
+ pin);
+
+ free(ctx, M_DEVBUF);
+ continue;
+ }
+
+ err = gpio_pin_setflags(ctx->gpio, flags & ~GPIO_INTR_MASK);
+ if (err != 0) {
+ device_printf(sc->dev,
+ "Cannot set pin flags for pin 0x%x\n", pin);
+
+ gpio_pin_release(ctx->gpio);
+ free(ctx, M_DEVBUF);
+ continue;
+ }
+
+ ctx->intr_rid = 0;
+ ctx->intr_res = gpio_alloc_intr_resource(sc->dev,
+ &ctx->intr_rid, RF_ACTIVE, ctx->gpio,
+ flags & GPIO_INTR_MASK);
+ if (ctx->intr_res == NULL) {
+ device_printf(sc->dev,
+ "Cannot allocate an IRQ for pin 0x%x\n", pin);
+
+ gpio_pin_release(ctx->gpio);
+ free(ctx, M_DEVBUF);
+ continue;
+ }
+
+ err = bus_setup_intr(sc->dev, ctx->intr_res, INTR_TYPE_MISC |
+ INTR_MPSAFE | INTR_EXCL | INTR_SLEEPABLE, NULL,
+ gpio_aei_intr, ctx, &ctx->intr_cookie);
+ if (err != 0) {
+ device_printf(sc->dev,
+ "Cannot set up an IRQ for pin 0x%x\n", pin);
+
+ bus_release_resource(sc->dev, ctx->intr_res);
+ gpio_pin_release(ctx->gpio);
+ free(ctx, M_DEVBUF);
+ continue;
+ }
+
+ ctx->pin = pin;
+ SLIST_INSERT_HEAD(&sc->aei_ctx, ctx, next);
+ }
+
+ return (AE_OK);
}
static int
gpio_aei_attach(device_t dev)
{
struct gpio_aei_softc * sc = device_get_softc(dev);
- gpio_pin_t pin;
- uint32_t flags;
ACPI_HANDLE handle;
- int err;
+ ACPI_STATUS status;
/* This is us. */
device_set_desc(dev, "ACPI Event Information Device");
- /* Store parameters needed by gpio_aei_intr. */
handle = acpi_gpiobus_get_handle(dev);
- if (gpio_pin_get_by_child_index(dev, 0, &pin) != 0) {
- device_printf(dev, "Unable to get the input pin\n");
+ status = AcpiGetParent(handle, &sc->dev_handle);
+ if (ACPI_FAILURE(status)) {
+ device_printf(dev, "Cannot get parent of %s\n",
+ acpi_name(handle));
return (ENXIO);
}
- sc->type = ACPI_AEI_TYPE_UNKNOWN;
- sc->pin = pin->pin;
-
- flags = acpi_gpiobus_get_flags(dev);
- if (pin->pin <= 255) {
- char objname[5]; /* "_EXX" or "_LXX" */
- sprintf(objname, "_%c%02X",
- (flags & GPIO_INTR_EDGE_MASK) ? 'E' : 'L', pin->pin);
- if (ACPI_SUCCESS(AcpiGetHandle(handle, objname, &sc->handle)))
- sc->type = ACPI_AEI_TYPE_ELX;
- }
- if (sc->type == ACPI_AEI_TYPE_UNKNOWN) {
- if (ACPI_SUCCESS(AcpiGetHandle(handle, "_EVT", &sc->handle)))
- sc->type = ACPI_AEI_TYPE_EVT;
- }
-
- if (sc->type == ACPI_AEI_TYPE_UNKNOWN) {
- device_printf(dev, "ACPI Event Information Device type is unknown");
- return (ENOTSUP);
- }
+ SLIST_INIT(&sc->aei_ctx);
+ sc->dev = dev;
- /* Set up the interrupt. */
- if ((sc->intr_res = gpio_alloc_intr_resource(dev, &sc->intr_rid,
- RF_ACTIVE, pin, flags & GPIO_INTR_MASK)) == NULL) {
- device_printf(dev, "Cannot allocate an IRQ\n");
- return (ENOTSUP);
- }
- err = bus_setup_intr(dev, sc->intr_res, INTR_TYPE_MISC | INTR_MPSAFE |
- INTR_EXCL | INTR_SLEEPABLE, NULL, gpio_aei_intr, sc,
- &sc->intr_cookie);
- if (err != 0) {
- device_printf(dev, "Cannot set up IRQ\n");
- bus_release_resource(dev, SYS_RES_IRQ, sc->intr_rid,
- sc->intr_res);
- return (err);
+ status = AcpiWalkResources(sc->dev_handle, "_AEI",
+ gpio_aei_enumerate, sc);
+ if (ACPI_FAILURE(status)) {
+ device_printf(dev, "Failed to enumerate AEI resources\n");
+ return (ENXIO);
}
return (0);
@@ -137,9 +233,15 @@ static int
gpio_aei_detach(device_t dev)
{
struct gpio_aei_softc * sc = device_get_softc(dev);
+ struct gpio_aei_ctx * ctx, * tctx;
+
+ SLIST_FOREACH_SAFE(ctx, &sc->aei_ctx, next, tctx) {
+ bus_teardown_intr(dev, ctx->intr_res, ctx->intr_cookie);
+ bus_release_resource(dev, ctx->intr_res);
+ gpio_pin_release(ctx->gpio);
+ free(ctx, M_DEVBUF);
+ }
- bus_teardown_intr(dev, sc->intr_res, sc->intr_cookie);
- bus_release_resource(dev, SYS_RES_IRQ, sc->intr_rid, sc->intr_res);
return (0);
}
diff --git a/sys/dev/gpio/gpiobus.c b/sys/dev/gpio/gpiobus.c
index c25c41f43042..5f1f6532a79b 100644
--- a/sys/dev/gpio/gpiobus.c
+++ b/sys/dev/gpio/gpiobus.c
@@ -57,7 +57,6 @@ static int gpiobus_suspend(device_t);
static int gpiobus_resume(device_t);
static void gpiobus_probe_nomatch(device_t, device_t);
static int gpiobus_print_child(device_t, device_t);
-static int gpiobus_child_location(device_t, device_t, struct sbuf *);
static device_t gpiobus_add_child(device_t, u_int, const char *, int);
static void gpiobus_hinted_child(device_t, const char *, int);
@@ -662,7 +661,7 @@ gpiobus_print_child(device_t dev, device_t child)
return (retval);
}
-static int
+int
gpiobus_child_location(device_t bus, device_t child, struct sbuf *sb)
{
struct gpiobus_ivar *devi;
@@ -674,16 +673,19 @@ gpiobus_child_location(device_t bus, device_t child, struct sbuf *sb)
return (0);
}
-static device_t
-gpiobus_add_child(device_t dev, u_int order, const char *name, int unit)
+device_t
+gpiobus_add_child_common(device_t dev, u_int order, const char *name, int unit,
+ size_t ivars_size)
{
device_t child;
struct gpiobus_ivar *devi;
+ KASSERT(ivars_size >= sizeof(struct gpiobus_ivar),
+ ("child ivars must include gpiobus_ivar as their first member"));
child = device_add_child_ordered(dev, order, name, unit);
if (child == NULL)
return (child);
- devi = malloc(sizeof(struct gpiobus_ivar), M_DEVBUF, M_NOWAIT | M_ZERO);
+ devi = malloc(ivars_size, M_DEVBUF, M_NOWAIT | M_ZERO);
if (devi == NULL) {
device_delete_child(dev, child);
return (NULL);
@@ -694,6 +696,13 @@ gpiobus_add_child(device_t dev, u_int order, const char *name, int unit)
return (child);
}
+static device_t
+gpiobus_add_child(device_t dev, u_int order, const char *name, int unit)
+{
+ return (gpiobus_add_child_common(dev, order, name, unit,
+ sizeof(struct gpiobus_ivar)));
+}
+
static void
gpiobus_child_deleted(device_t dev, device_t child)
{
diff --git a/sys/dev/gpio/gpiobus_internal.h b/sys/dev/gpio/gpiobus_internal.h
index de3f57663132..c198e5f79989 100644
--- a/sys/dev/gpio/gpiobus_internal.h
+++ b/sys/dev/gpio/gpiobus_internal.h
@@ -42,6 +42,8 @@ void gpiobus_free_ivars(struct gpiobus_ivar *);
int gpiobus_read_ivar(device_t, device_t, int, uintptr_t *);
int gpiobus_acquire_pin(device_t, uint32_t);
void gpiobus_release_pin(device_t, uint32_t);
+int gpiobus_child_location(device_t, device_t, struct sbuf *);
+device_t gpiobus_add_child_common(device_t, u_int, const char *, int, size_t);
extern driver_t gpiobus_driver;
#endif
diff --git a/sys/dev/gpio/ofw_gpiobus.c b/sys/dev/gpio/ofw_gpiobus.c
index fc5fb03d6824..b12b78fac18c 100644
--- a/sys/dev/gpio/ofw_gpiobus.c
+++ b/sys/dev/gpio/ofw_gpiobus.c
@@ -451,28 +451,22 @@ ofw_gpiobus_add_child(device_t dev, u_int order, const char *name, int unit)
device_t child;
struct ofw_gpiobus_devinfo *devi;
- child = device_add_child_ordered(dev, order, name, unit);
+ child = gpiobus_add_child_common(dev, order, name, unit,
+ sizeof(struct ofw_gpiobus_devinfo));
if (child == NULL)
- return (child);
- devi = malloc(sizeof(struct ofw_gpiobus_devinfo), M_DEVBUF,
- M_NOWAIT | M_ZERO);
- if (devi == NULL) {
- device_delete_child(dev, child);
- return (0);
- }
+ return (NULL);
/*
* NULL all the OFW-related parts of the ivars for non-OFW
* children.
*/
+ devi = device_get_ivars(child);
devi->opd_obdinfo.obd_node = -1;
devi->opd_obdinfo.obd_name = NULL;
devi->opd_obdinfo.obd_compat = NULL;
devi->opd_obdinfo.obd_type = NULL;
devi->opd_obdinfo.obd_model = NULL;
- device_set_ivars(child, devi);
-
return (child);
}
diff --git a/sys/dev/mlx5/mlx5_accel/mlx5_ipsec_offload.c b/sys/dev/mlx5/mlx5_accel/mlx5_ipsec_offload.c
index 978e5f25ceaf..cc0bc1f3fcd2 100644
--- a/sys/dev/mlx5/mlx5_accel/mlx5_ipsec_offload.c
+++ b/sys/dev/mlx5/mlx5_accel/mlx5_ipsec_offload.c
@@ -120,7 +120,7 @@ static void mlx5e_ipsec_packet_setup(void *obj, u32 pdn,
switch (attrs->dir) {
case IPSEC_DIR_OUTBOUND:
- if (attrs->replay_esn.replay_window != 0)
+ if (attrs->replay_esn.trigger)
MLX5_SET(ipsec_aso, aso_ctx, mode, MLX5_IPSEC_ASO_INC_SN);
else
MLX5_SET(ipsec_aso, aso_ctx, mode, MLX5_IPSEC_ASO_MODE);
diff --git a/sys/modules/zfs/Makefile b/sys/modules/zfs/Makefile
index 2dd9e2be3f56..ec531ed646a7 100644
--- a/sys/modules/zfs/Makefile
+++ b/sys/modules/zfs/Makefile
@@ -15,6 +15,7 @@ KMOD= zfs
${SRCDIR}/icp/asm-ppc64/sha2 \
${SRCDIR}/icp/asm-ppc64/blake3 \
${SRCDIR}/icp/asm-x86_64/blake3 \
+ ${SRCDIR}/icp/asm-x86_64/modes \
${SRCDIR}/icp/asm-x86_64/sha2 \
${SRCDIR}/os/freebsd/spl \
${SRCDIR}/os/freebsd/zfs \
@@ -40,7 +41,8 @@ CFLAGS+= -D__KERNEL__ -DFREEBSD_NAMECACHE -DBUILDING_ZFS \
.if ${MACHINE_ARCH} == "amd64"
CFLAGS+= -D__x86_64 -DHAVE_SSE2 -DHAVE_SSSE3 -DHAVE_SSE4_1 -DHAVE_SSE4_2 \
- -DHAVE_AVX -DHAVE_AVX2 -DHAVE_AVX512F -DHAVE_AVX512VL -DHAVE_AVX512BW
+ -DHAVE_AVX -DHAVE_AVX2 -DHAVE_AVX512F -DHAVE_AVX512VL -DHAVE_AVX512BW \
+ -DHAVE_VAES -DHAVE_VPCLMULQDQ
.endif
.if ${MACHINE_ARCH} == "i386" || ${MACHINE_ARCH} == "powerpc" || \
@@ -82,6 +84,9 @@ SRCS+= blake3_avx2.S \
blake3_avx512.S \
blake3_sse2.S \
blake3_sse41.S
+
+#icp/asm-x86_64/modes
+SRCS+= aesni-gcm-avx2-vaes.S
.endif
#icp/algs/sha2
diff --git a/sys/modules/zfs/zfs_config.h b/sys/modules/zfs/zfs_config.h
index c79c9eaa1a5f..12274bcceea1 100644
--- a/sys/modules/zfs/zfs_config.h
+++ b/sys/modules/zfs/zfs_config.h
@@ -704,6 +704,11 @@
/* iops->setattr() takes struct user_namespace* */
/* #undef HAVE_USERNS_IOPS_SETATTR */
+#ifdef __amd64__
+/* Define if host toolchain supports VAES */
+#define HAVE_VAES 1
+#endif
+
/* fops->clone_file_range() is available */
/* #undef HAVE_VFS_CLONE_FILE_RANGE */
@@ -743,6 +748,11 @@
/* __vmalloc page flags exists */
/* #undef HAVE_VMALLOC_PAGE_KERNEL */
+#ifdef __amd64__
+/* Define if host toolchain supports VPCLMULQDQ */
+#define HAVE_VPCLMULQDQ 1
+#endif
+
/* int (*writepage_t)() takes struct folio* */
/* #undef HAVE_WRITEPAGE_T_FOLIO */
@@ -830,7 +840,7 @@
/* #undef ZFS_DEVICE_MINOR */
/* Define the project alias string. */
-#define ZFS_META_ALIAS "zfs-2.3.99-539-FreeBSD_g1d0b94c4e"
+#define ZFS_META_ALIAS "zfs-2.3.99-571-FreeBSD_ga9410ccbd"
/* Define the project author. */
#define ZFS_META_AUTHOR "OpenZFS"
@@ -839,7 +849,7 @@
/* #undef ZFS_META_DATA */
/* Define the maximum compatible kernel version. */
-#define ZFS_META_KVER_MAX "6.15"
+#define ZFS_META_KVER_MAX "6.16"
/* Define the minimum compatible kernel version. */
#define ZFS_META_KVER_MIN "4.18"
@@ -860,7 +870,7 @@
#define ZFS_META_NAME "zfs"
/* Define the project release. */
-#define ZFS_META_RELEASE "539-FreeBSD_g1d0b94c4e"
+#define ZFS_META_RELEASE "571-FreeBSD_ga9410ccbd"
/* Define the project version. */
#define ZFS_META_VERSION "2.3.99"
diff --git a/sys/modules/zfs/zfs_gitrev.h b/sys/modules/zfs/zfs_gitrev.h
index 20fd58c620b5..5c265cf5b08e 100644
--- a/sys/modules/zfs/zfs_gitrev.h
+++ b/sys/modules/zfs/zfs_gitrev.h
@@ -1 +1 @@
-#define ZFS_META_GITREV "zfs-2.3.99-539-g1d0b94c4e"
+#define ZFS_META_GITREV "zfs-2.3.99-571-ga9410ccbd"
diff --git a/sys/net/if_pfsync.h b/sys/net/if_pfsync.h
index 1efc220aa8e1..e99df0b85ccf 100644
--- a/sys/net/if_pfsync.h
+++ b/sys/net/if_pfsync.h
@@ -160,8 +160,8 @@ struct pfsync_ins_ack {
struct pfsync_upd_c {
u_int64_t id;
- struct pfsync_state_peer src;
- struct pfsync_state_peer dst;
+ struct pf_state_peer_export src;
+ struct pf_state_peer_export dst;
u_int32_t creatorid;
u_int32_t expire;
u_int8_t timeout;
diff --git a/sys/net/iflib.c b/sys/net/iflib.c
index 2b43f6f19051..98c59e5de988 100644
--- a/sys/net/iflib.c
+++ b/sys/net/iflib.c
@@ -142,6 +142,7 @@ struct iflib_ctx;
static void iru_init(if_rxd_update_t iru, iflib_rxq_t rxq, uint8_t flid);
static void iflib_timer(void *arg);
static void iflib_tqg_detach(if_ctx_t ctx);
+static int iflib_simple_transmit(if_t ifp, struct mbuf *m);
typedef struct iflib_filter_info {
driver_filter_t *ifi_filter;
@@ -198,6 +199,7 @@ struct iflib_ctx {
uint8_t ifc_sysctl_use_logical_cores;
uint16_t ifc_sysctl_extra_msix_vectors;
bool ifc_cpus_are_physical_cores;
+ bool ifc_sysctl_simple_tx;
qidx_t ifc_sysctl_ntxds[8];
qidx_t ifc_sysctl_nrxds[8];
@@ -725,6 +727,7 @@ static void iflib_free_intr_mem(if_ctx_t ctx);
#ifndef __NO_STRICT_ALIGNMENT
static struct mbuf *iflib_fixup_rx(struct mbuf *m);
#endif
+static __inline int iflib_completed_tx_reclaim(iflib_txq_t txq, int thresh);
static SLIST_HEAD(cpu_offset_list, cpu_offset) cpu_offsets =
SLIST_HEAD_INITIALIZER(cpu_offsets);
@@ -2624,8 +2627,10 @@ iflib_stop(if_ctx_t ctx)
#endif /* DEV_NETMAP */
CALLOUT_UNLOCK(txq);
- /* clean any enqueued buffers */
- iflib_ifmp_purge(txq);
+ if (!ctx->ifc_sysctl_simple_tx) {
+ /* clean any enqueued buffers */
+ iflib_ifmp_purge(txq);
+ }
/* Free any existing tx buffers. */
for (j = 0; j < txq->ift_size; j++) {
iflib_txsd_free(ctx, txq, j);
@@ -3635,13 +3640,16 @@ defrag:
* cxgb
*/
if (__predict_false(nsegs + 2 > TXQ_AVAIL(txq))) {
- txq->ift_no_desc_avail++;
- bus_dmamap_unload(buf_tag, map);
- DBG_COUNTER_INC(encap_txq_avail_fail);
- DBG_COUNTER_INC(encap_txd_encap_fail);
- if ((txq->ift_task.gt_task.ta_flags & TASK_ENQUEUED) == 0)
- GROUPTASK_ENQUEUE(&txq->ift_task);
- return (ENOBUFS);
+ (void)iflib_completed_tx_reclaim(txq, RECLAIM_THRESH(ctx));
+ if (__predict_false(nsegs + 2 > TXQ_AVAIL(txq))) {
+ txq->ift_no_desc_avail++;
+ bus_dmamap_unload(buf_tag, map);
+ DBG_COUNTER_INC(encap_txq_avail_fail);
+ DBG_COUNTER_INC(encap_txd_encap_fail);
+ if ((txq->ift_task.gt_task.ta_flags & TASK_ENQUEUED) == 0)
+ GROUPTASK_ENQUEUE(&txq->ift_task);
+ return (ENOBUFS);
+ }
}
/*
* On Intel cards we can greatly reduce the number of TX interrupts
@@ -4014,6 +4022,12 @@ _task_fn_tx(void *context)
netmap_tx_irq(ifp, txq->ift_id))
goto skip_ifmp;
#endif
+ if (ctx->ifc_sysctl_simple_tx) {
+ mtx_lock(&txq->ift_mtx);
+ (void)iflib_completed_tx_reclaim(txq, RECLAIM_THRESH(ctx));
+ mtx_unlock(&txq->ift_mtx);
+ goto skip_ifmp;
+ }
#ifdef ALTQ
if (if_altq_is_enabled(ifp))
iflib_altq_if_start(ifp);
@@ -4027,9 +4041,8 @@ _task_fn_tx(void *context)
*/
if (abdicate)
ifmp_ring_check_drainage(txq->ift_br, TX_BATCH_SIZE);
-#ifdef DEV_NETMAP
+
skip_ifmp:
-#endif
if (ctx->ifc_flags & IFC_LEGACY)
IFDI_INTR_ENABLE(ctx);
else
@@ -5131,7 +5144,14 @@ iflib_device_register(device_t dev, void *sc, if_shared_ctx_t sctx, if_ctx_t *ct
scctx = &ctx->ifc_softc_ctx;
ifp = ctx->ifc_ifp;
-
+ if (ctx->ifc_sysctl_simple_tx) {
+#ifndef ALTQ
+ if_settransmitfn(ifp, iflib_simple_transmit);
+ device_printf(dev, "using simple if_transmit\n");
+#else
+ device_printf(dev, "ALTQ prevents using simple if_transmit\n");
+#endif
+ }
iflib_reset_qvalues(ctx);
IFNET_WLOCK();
CTX_LOCK(ctx);
@@ -6766,6 +6786,9 @@ iflib_add_device_sysctl_pre(if_ctx_t ctx)
SYSCTL_ADD_CONST_STRING(ctx_list, oid_list, OID_AUTO, "driver_version",
CTLFLAG_RD, ctx->ifc_sctx->isc_driver_version, "driver version");
+ SYSCTL_ADD_BOOL(ctx_list, oid_list, OID_AUTO, "simple_tx",
+ CTLFLAG_RDTUN, &ctx->ifc_sysctl_simple_tx, 0,
+ "use simple tx ring");
SYSCTL_ADD_U16(ctx_list, oid_list, OID_AUTO, "override_ntxqs",
CTLFLAG_RWTUN, &ctx->ifc_sysctl_ntxqs, 0,
"# of txqs to use, 0 => use default #");
@@ -7088,3 +7111,48 @@ iflib_debugnet_poll(if_t ifp, int count)
return (0);
}
#endif /* DEBUGNET */
+
+
+static inline iflib_txq_t
+iflib_simple_select_queue(if_ctx_t ctx, struct mbuf *m)
+{
+ int qidx;
+
+ if ((NTXQSETS(ctx) > 1) && M_HASHTYPE_GET(m))
+ qidx = QIDX(ctx, m);
+ else
+ qidx = NTXQSETS(ctx) + FIRST_QSET(ctx) - 1;
+ return (&ctx->ifc_txqs[qidx]);
+}
+
+static int
+iflib_simple_transmit(if_t ifp, struct mbuf *m)
+{
+ if_ctx_t ctx;
+ iflib_txq_t txq;
+ int error;
+ int bytes_sent = 0, pkt_sent = 0, mcast_sent = 0;
+
+
+ ctx = if_getsoftc(ifp);
+ if ((if_getdrvflags(ifp) & (IFF_DRV_RUNNING | IFF_DRV_OACTIVE)) !=
+ IFF_DRV_RUNNING)
+ return (EBUSY);
+ txq = iflib_simple_select_queue(ctx, m);
+ mtx_lock(&txq->ift_mtx);
+ error = iflib_encap(txq, &m);
+ if (error == 0) {
+ pkt_sent++;
+ bytes_sent += m->m_pkthdr.len;
+ mcast_sent += !!(m->m_flags & M_MCAST);
+ (void)iflib_txd_db_check(txq, true);
+ }
+ (void)iflib_completed_tx_reclaim(txq, RECLAIM_THRESH(ctx));
+ mtx_unlock(&txq->ift_mtx);
+ if_inc_counter(ifp, IFCOUNTER_OBYTES, bytes_sent);
+ if_inc_counter(ifp, IFCOUNTER_OPACKETS, pkt_sent);
+ if (mcast_sent)
+ if_inc_counter(ifp, IFCOUNTER_OMCASTS, mcast_sent);
+
+ return (error);
+}
diff --git a/sys/net/pfvar.h b/sys/net/pfvar.h
index c397f0b67896..d6c13470f2eb 100644
--- a/sys/net/pfvar.h
+++ b/sys/net/pfvar.h
@@ -1020,7 +1020,7 @@ struct pf_state_scrub_export {
#define PF_SCRUB_FLAG_VALID 0x01
uint8_t scrub_flag;
uint32_t pfss_ts_mod; /* timestamp modulation */
-};
+} __packed;
struct pf_state_key_export {
struct pf_addr addr[2];
@@ -1037,7 +1037,7 @@ struct pf_state_peer_export {
uint8_t state; /* active state level */
uint8_t wscale; /* window scaling factor */
uint8_t dummy[6];
-};
+} __packed;
_Static_assert(sizeof(struct pf_state_peer_export) == 32, "size incorrect");
struct pf_state_export {
@@ -1179,26 +1179,6 @@ struct pf_test_ctx {
* Unified state structures for pulling states out of the kernel
* used by pfsync(4) and the pf(4) ioctl.
*/
-struct pfsync_state_scrub {
- u_int16_t pfss_flags;
- u_int8_t pfss_ttl; /* stashed TTL */
-#define PFSYNC_SCRUB_FLAG_VALID 0x01
- u_int8_t scrub_flag;
- u_int32_t pfss_ts_mod; /* timestamp modulation */
-} __packed;
-
-struct pfsync_state_peer {
- struct pfsync_state_scrub scrub; /* state is scrubbed */
- u_int32_t seqlo; /* Max sequence number sent */
- u_int32_t seqhi; /* Max the other end ACKd + win */
- u_int32_t seqdiff; /* Sequence number modulator */
- u_int16_t max_win; /* largest window (pre scaling) */
- u_int16_t mss; /* Maximum segment size option */
- u_int8_t state; /* active state level */
- u_int8_t wscale; /* window scaling factor */
- u_int8_t pad[6];
-} __packed;
-
struct pfsync_state_key {
struct pf_addr addr[2];
u_int16_t port[2];
@@ -1208,8 +1188,8 @@ struct pfsync_state_1301 {
u_int64_t id;
char ifname[IFNAMSIZ];
struct pfsync_state_key key[2];
- struct pfsync_state_peer src;
- struct pfsync_state_peer dst;
+ struct pf_state_peer_export src;
+ struct pf_state_peer_export dst;
struct pf_addr rt_addr;
u_int32_t rule;
u_int32_t anchor;
@@ -1235,8 +1215,8 @@ struct pfsync_state_1400 {
u_int64_t id;
char ifname[IFNAMSIZ];
struct pfsync_state_key key[2];
- struct pfsync_state_peer src;
- struct pfsync_state_peer dst;
+ struct pf_state_peer_export src;
+ struct pf_state_peer_export dst;
struct pf_addr rt_addr;
u_int32_t rule;
u_int32_t anchor;
@@ -1323,39 +1303,10 @@ extern pflog_packet_t *pflog_packet_ptr;
/* for copies to/from network byte order */
/* ioctl interface also uses network byte order */
-#define pf_state_peer_hton(s,d) do { \
- (d)->seqlo = htonl((s)->seqlo); \
- (d)->seqhi = htonl((s)->seqhi); \
- (d)->seqdiff = htonl((s)->seqdiff); \
- (d)->max_win = htons((s)->max_win); \
- (d)->mss = htons((s)->mss); \
- (d)->state = (s)->state; \
- (d)->wscale = (s)->wscale; \
- if ((s)->scrub) { \
- (d)->scrub.pfss_flags = \
- htons((s)->scrub->pfss_flags & PFSS_TIMESTAMP); \
- (d)->scrub.pfss_ttl = (s)->scrub->pfss_ttl; \
- (d)->scrub.pfss_ts_mod = htonl((s)->scrub->pfss_ts_mod);\
- (d)->scrub.scrub_flag = PFSYNC_SCRUB_FLAG_VALID; \
- } \
-} while (0)
-
-#define pf_state_peer_ntoh(s,d) do { \
- (d)->seqlo = ntohl((s)->seqlo); \
- (d)->seqhi = ntohl((s)->seqhi); \
- (d)->seqdiff = ntohl((s)->seqdiff); \
- (d)->max_win = ntohs((s)->max_win); \
- (d)->mss = ntohs((s)->mss); \
- (d)->state = (s)->state; \
- (d)->wscale = (s)->wscale; \
- if ((s)->scrub.scrub_flag == PFSYNC_SCRUB_FLAG_VALID && \
- (d)->scrub != NULL) { \
- (d)->scrub->pfss_flags = \
- ntohs((s)->scrub.pfss_flags) & PFSS_TIMESTAMP; \
- (d)->scrub->pfss_ttl = (s)->scrub.pfss_ttl; \
- (d)->scrub->pfss_ts_mod = ntohl((s)->scrub.pfss_ts_mod);\
- } \
-} while (0)
+void pf_state_peer_hton(const struct pf_state_peer *,
+ struct pf_state_peer_export *);
+void pf_state_peer_ntoh(const struct pf_state_peer_export *,
+ struct pf_state_peer *);
#define pf_state_counter_hton(s,d) do { \
d[0] = htonl((s>>32)&0xffffffff); \
diff --git a/sys/netpfil/pf/if_pfsync.c b/sys/netpfil/pf/if_pfsync.c
index e34c08c8c4db..585c196391c0 100644
--- a/sys/netpfil/pf/if_pfsync.c
+++ b/sys/netpfil/pf/if_pfsync.c
@@ -123,8 +123,8 @@ union inet_template {
sizeof(struct pfsync_header) + \
sizeof(struct pfsync_subheader) )
-static int pfsync_upd_tcp(struct pf_kstate *, struct pfsync_state_peer *,
- struct pfsync_state_peer *);
+static int pfsync_upd_tcp(struct pf_kstate *, struct pf_state_peer_export *,
+ struct pf_state_peer_export *);
static int pfsync_in_clr(struct mbuf *, int, int, int, int);
static int pfsync_in_ins(struct mbuf *, int, int, int, int);
static int pfsync_in_iack(struct mbuf *, int, int, int, int);
@@ -330,7 +330,7 @@ SYSCTL_UINT(_net_pfsync, OID_AUTO, defer_delay, CTLFLAG_VNET | CTLFLAG_RW,
static int pfsync_clone_create(struct if_clone *, int, caddr_t);
static void pfsync_clone_destroy(struct ifnet *);
-static int pfsync_alloc_scrub_memory(struct pfsync_state_peer *,
+static int pfsync_alloc_scrub_memory(struct pf_state_peer_export *,
struct pf_state_peer *);
static int pfsyncoutput(struct ifnet *, struct mbuf *,
const struct sockaddr *, struct route *);
@@ -502,7 +502,7 @@ pfsync_clone_destroy(struct ifnet *ifp)
}
static int
-pfsync_alloc_scrub_memory(struct pfsync_state_peer *s,
+pfsync_alloc_scrub_memory(struct pf_state_peer_export *s,
struct pf_state_peer *d)
{
if (s->scrub.scrub_flag && d->scrub == NULL) {
@@ -1172,8 +1172,8 @@ pfsync_in_iack(struct mbuf *m, int offset, int count, int flags, int action)
}
static int
-pfsync_upd_tcp(struct pf_kstate *st, struct pfsync_state_peer *src,
- struct pfsync_state_peer *dst)
+pfsync_upd_tcp(struct pf_kstate *st, struct pf_state_peer_export *src,
+ struct pf_state_peer_export *dst)
{
int sync = 0;
diff --git a/sys/netpfil/pf/pf.c b/sys/netpfil/pf/pf.c
index 9d83e7b82e6f..8cd4fff95b15 100644
--- a/sys/netpfil/pf/pf.c
+++ b/sys/netpfil/pf/pf.c
@@ -2069,6 +2069,44 @@ pf_find_state_all_exists(const struct pf_state_key_cmp *key, u_int dir)
return (false);
}
+void
+pf_state_peer_hton(const struct pf_state_peer *s, struct pf_state_peer_export *d)
+{
+ d->seqlo = htonl(s->seqlo);
+ d->seqhi = htonl(s->seqhi);
+ d->seqdiff = htonl(s->seqdiff);
+ d->max_win = htons(s->max_win);
+ d->mss = htons(s->mss);
+ d->state = s->state;
+ d->wscale = s->wscale;
+ if (s->scrub) {
+ d->scrub.pfss_flags = htons(
+ s->scrub->pfss_flags & PFSS_TIMESTAMP);
+ d->scrub.pfss_ttl = (s)->scrub->pfss_ttl;
+ d->scrub.pfss_ts_mod = htonl((s)->scrub->pfss_ts_mod);
+ d->scrub.scrub_flag = PF_SCRUB_FLAG_VALID;
+ }
+}
+
+void
+pf_state_peer_ntoh(const struct pf_state_peer_export *s, struct pf_state_peer *d)
+{
+ d->seqlo = ntohl(s->seqlo);
+ d->seqhi = ntohl(s->seqhi);
+ d->seqdiff = ntohl(s->seqdiff);
+ d->max_win = ntohs(s->max_win);
+ d->mss = ntohs(s->mss);
+ d->state = s->state;
+ d->wscale = s->wscale;
+ if (s->scrub.scrub_flag == PF_SCRUB_FLAG_VALID &&
+ d->scrub != NULL) {
+ d->scrub->pfss_flags = ntohs(s->scrub.pfss_flags) &
+ PFSS_TIMESTAMP;
+ d->scrub->pfss_ttl = s->scrub.pfss_ttl;
+ d->scrub->pfss_ts_mod = ntohl(s->scrub.pfss_ts_mod);
+ }
+}
+
struct pf_udp_mapping *
pf_udp_mapping_create(sa_family_t af, struct pf_addr *src_addr, uint16_t src_port,
struct pf_addr *nat_addr, uint16_t nat_port)
diff --git a/sys/netpfil/pf/pf_nl.c b/sys/netpfil/pf/pf_nl.c
index 09754359ec2d..45b5b8dd5fef 100644
--- a/sys/netpfil/pf/pf_nl.c
+++ b/sys/netpfil/pf/pf_nl.c
@@ -118,7 +118,7 @@ dump_state_peer(struct nl_writer *nw, int attr, const struct pf_state_peer *peer
nlattr_add_u16(nw, PF_STP_PFSS_FLAGS, pfss_flags);
nlattr_add_u32(nw, PF_STP_PFSS_TS_MOD, sc->pfss_ts_mod);
nlattr_add_u8(nw, PF_STP_PFSS_TTL, sc->pfss_ttl);
- nlattr_add_u8(nw, PF_STP_SCRUB_FLAG, PFSYNC_SCRUB_FLAG_VALID);
+ nlattr_add_u8(nw, PF_STP_SCRUB_FLAG, PF_SCRUB_FLAG_VALID);
}
nlattr_set_len(nw, off);
diff --git a/tools/build/depend-cleanup.sh b/tools/build/depend-cleanup.sh
index cd51c59ff0e1..22bf34439758 100755
--- a/tools/build/depend-cleanup.sh
+++ b/tools/build/depend-cleanup.sh
@@ -50,12 +50,12 @@
# - Replacing generated files with files committed to the tree. This is special
# case of moving from one directory to another. The stale generated file also
# needs to be deleted, so that it isn't found in make's .PATH. Note the
-# unconditional `rm -f`: there's no need for an extra call to first check for
+# unconditional `rm -fv`: there's no need for an extra call to first check for
# the file's existence.
#
# # 20250110 3863fec1ce2d add strlen SIMD implementation
# clean_dep lib/libc strlen S arm-optimized-routines
-# run rm -f "$OBJTOP"/lib/libc/strlen.S
+# run rm -fv "$OBJTOP"/lib/libc/strlen.S
#
# A rule may be required for only one architecture:
#
@@ -152,6 +152,11 @@ run()
fi
}
+# Clean the depend and object files for a given source file if the
+# depend file matches a regex (which defaults to the source file
+# name). This is typically used if a file was renamed, especially if
+# only its extension was changed (e.g. from .c to .cc).
+#
# $1 directory
# $2 source filename w/o extension
# $3 source extension
@@ -162,13 +167,34 @@ clean_dep()
dirprfx=${libcompat:+obj-lib${libcompat}/}
if egrep -qw "${4:-$2\.$3}" "$OBJTOP"/$dirprfx$1/.depend.$2.*o 2>/dev/null; then
echo "Removing stale ${libcompat:+lib${libcompat} }dependencies and objects for $2.$3"
- run rm -f \
+ run rm -fv \
"$OBJTOP"/$dirprfx$1/.depend.$2.* \
"$OBJTOP"/$dirprfx$1/$2.*o
fi
done
}
+# Clean the object file for a given source file if it exists and
+# matches a regex. This is typically used if a a change in CFLAGS or
+# similar caused a change in the generated code without a change in
+# the sources.
+#
+# $1 directory
+# $2 source filename w/o extension
+# $3 source extension
+# $4 regex for egrep -w
+clean_obj()
+{
+ for libcompat in "" $ALL_libcompats; do
+ dirprfx=${libcompat:+obj-lib${libcompat}/}
+ if strings "$OBJTOP"/$dirprfx$1/$2.*o 2>/dev/null | egrep -qw "${4}"; then
+ echo "Removing stale ${libcompat:+lib${libcompat} }objects for $2.$3"
+ run rm -fv \
+ "$OBJTOP"/$dirprfx$1/$2.*o
+ fi
+ done
+}
+
extract_epoch()
{
[ -s "$1" ] || return 0
@@ -243,7 +269,7 @@ fi
if stat "$OBJTOP"/tests/sys/kqueue/libkqueue/*kqtest* \
"$OBJTOP"/tests/sys/kqueue/libkqueue/.depend.kqtest* >/dev/null 2>&1; then
echo "Removing old kqtest"
- run rm -f "$OBJTOP"/tests/sys/kqueue/libkqueue/.depend.* \
+ run rm -fv "$OBJTOP"/tests/sys/kqueue/libkqueue/.depend.* \
"$OBJTOP"/tests/sys/kqueue/libkqueue/*
fi
@@ -317,7 +343,7 @@ fi
if [ -f "$OBJTOP"/rescue/rescue/rescue.mk ] && \
! grep -q 'nvme_util.o' "$OBJTOP"/rescue/rescue/rescue.mk; then
echo "removing rescue.mk without nvme_util.o"
- run rm -f "$OBJTOP"/rescue/rescue/rescue.mk
+ run rm -fv "$OBJTOP"/rescue/rescue/rescue.mk
fi
# 20240910 e2df9bb44109
@@ -337,7 +363,7 @@ if [ ${MACHINE} = riscv ]; then
fi
if ! grep -q 'lib/libc/csu/riscv/reloc\.c' "$f"; then
echo "Removing stale dependencies and objects for libc_start1.c"
- run rm -f \
+ run rm -fv \
"$OBJTOP"/lib/libc/.depend.libc_start1.* \
"$OBJTOP"/lib/libc/libc_start1.*o
break
@@ -351,28 +377,28 @@ f="$p"/arm_mve_builtin_sema.inc
if [ -e "$f" ]; then
if grep -q SemaBuiltinConstantArgRange "$f"; then
echo "Removing pre-llvm19 clang-tblgen output"
- run rm -f "$p"/*.inc
+ run rm -fv "$p"/*.inc
fi
fi
# 20241025 cb5e41b16083 Unbundle hash functions fom lib/libcrypt
-clean_dep lib/libcrypt crypt-md5 c
-clean_dep lib/libcrypt crypt-nthash c
-clean_dep lib/libcrypt crypt-sha256 c
-clean_dep lib/libcrypt crypt-sha512 c
+clean_obj lib/libcrypt crypt-md5 c __MD5Init
+clean_obj lib/libcrypt crypt-nthash c __MD4Init
+clean_obj lib/libcrypt crypt-sha256 c __SHA256Init
+clean_obj lib/libcrypt crypt-sha512 c __SHA512Init
# 20241213 b55f5e1c4ae3 jemalloc: Move generated jemalloc.3 into lib/libc tree
if [ -h "$OBJTOP"/lib/libc/jemalloc.3 ]; then
# Have to cleanup the jemalloc.3 in the obj tree since make gets
# confused and won't use the one in lib/libc/malloc/jemalloc/jemalloc.3
echo "Removing stale jemalloc.3 object"
- run rm -f "$OBJTOP"/lib/libc/jemalloc.3
+ run rm -fv "$OBJTOP"/lib/libc/jemalloc.3
fi
if [ $MACHINE_ARCH = aarch64 ]; then
# 20250110 5e7d93a60440 add strcmp SIMD implementation
ALL_libcompats= clean_dep lib/libc strcmp S arm-optimized-routines
- run rm -f "$OBJTOP"/lib/libc/strcmp.S
+ run rm -fv "$OBJTOP"/lib/libc/strcmp.S
# 20250110 b91003acffe7 add strspn optimized implementation
ALL_libcompats= clean_dep lib/libc strspn c
@@ -391,7 +417,7 @@ if [ $MACHINE_ARCH = aarch64 ]; then
# 20250110 25c485e14769 add strncmp SIMD implementation
ALL_libcompats= clean_dep lib/libc strncmp S arm-optimized-routines
- run rm -f "$OBJTOP"/lib/libc/strncmp.S
+ run rm -fv "$OBJTOP"/lib/libc/strncmp.S
# 20250110 bad17991c06d add memccpy SIMD implementation
ALL_libcompats= clean_dep lib/libc memccpy c
@@ -402,11 +428,11 @@ if [ $MACHINE_ARCH = aarch64 ]; then
# 20250110 bea89d038ac5 add strlcat SIMD implementation, and move memchr
ALL_libcompats= clean_dep lib/libc strlcat c "libc.string.strlcat.c"
ALL_libcompats= clean_dep lib/libc memchr S "[[:space:]]memchr.S"
- run rm -f "$OBJTOP"/lib/libc/memchr.S
+ run rm -fv "$OBJTOP"/lib/libc/memchr.S
# 20250110 3863fec1ce2d add strlen SIMD implementation
ALL_libcompats= clean_dep lib/libc strlen S arm-optimized-routines
- run rm -f "$OBJTOP"/lib/libc/strlen.S
+ run rm -fv "$OBJTOP"/lib/libc/strlen.S
# 20250110 79e01e7e643c add bcopy & bzero wrapper
ALL_libcompats= clean_dep lib/libc bcopy c "libc.string.bcopy.c"
@@ -431,15 +457,15 @@ clean_dep usr.sbin/ctld uclparse c
# 20250425 2e47f35be5dc libllvm, libclang and liblldb became shared libraries
if [ -f "$OBJTOP"/lib/clang/libllvm/libllvm.a ]; then
echo "Removing old static libllvm library"
- run rm -f "$OBJTOP"/lib/clang/libllvm/libllvm.a
+ run rm -fv "$OBJTOP"/lib/clang/libllvm/libllvm.a
fi
if [ -f "$OBJTOP"/lib/clang/libclang/libclang.a ]; then
echo "Removing old static libclang library"
- run rm -f "$OBJTOP"/lib/clang/libclang/libclang.a
+ run rm -fv "$OBJTOP"/lib/clang/libclang/libclang.a
fi
if [ -f "$OBJTOP"/lib/clang/liblldb/liblldb.a ]; then
echo "Removing old static liblldb library"
- run rm -f "$OBJTOP"/lib/clang/liblldb/liblldb.a
+ run rm -fv "$OBJTOP"/lib/clang/liblldb/liblldb.a
fi
# 20250813 4f766afc1ca0 tcopy converted to C++
diff --git a/tools/build/options/WITH_CLEAN b/tools/build/options/WITH_CLEAN
new file mode 100644
index 000000000000..0bb05e33371b
--- /dev/null
+++ b/tools/build/options/WITH_CLEAN
@@ -0,0 +1,4 @@
+Clean before building world and/or kernel.
+Note that recording a new epoch in
+.Pa .clean_build_epoch
+in the root of the source tree will also force a clean world build.