diff options
| author | Konstantin Belousov <kib@FreeBSD.org> | 2024-03-02 04:58:57 +0000 |
|---|---|---|
| committer | Konstantin Belousov <kib@FreeBSD.org> | 2024-03-02 05:02:55 +0000 |
| commit | 7e4ac11b6076e6a9bf7341ddeae22784284ed733 (patch) | |
| tree | 48cb90318da10db9c1c803b8d4f02922ac5e1f8f | |
| parent | beadbca67bd811ae6e6f4f4aff683026af90fd3c (diff) | |
| download | src-7e4ac11b6076e6a9bf7341ddeae22784284ed733.tar.gz src-7e4ac11b6076e6a9bf7341ddeae22784284ed733.zip | |
getblkx(9): be more tolerant but also strict with the buffer size checks
It is possible that on-disk filesystem format causes allocation of
buffers of size larger than maxbcachebuf. Currently, getblkx() and
indirectly bufkva_alloc() panic in that situation.
It is more useful to return an error instead, allowing the system to
continue running.
PR: 277414
Reported by: Robert Morris <rtm@lcs.mit.edu>
MFC after: 1 week
Sponsored by: The FreeBSD Foundation
| -rw-r--r-- | sys/kern/vfs_bio.c | 12 |
1 files changed, 10 insertions, 2 deletions
diff --git a/sys/kern/vfs_bio.c b/sys/kern/vfs_bio.c index 4e0832475c3e..b5466fb2cd53 100644 --- a/sys/kern/vfs_bio.c +++ b/sys/kern/vfs_bio.c @@ -3985,9 +3985,11 @@ getblkx(struct vnode *vp, daddr_t blkno, daddr_t dblkno, int size, int slpflag, ("GB_KVAALLOC only makes sense with GB_UNMAPPED")); if (vp->v_type != VCHR) ASSERT_VOP_LOCKED(vp, "getblk"); - if (size > maxbcachebuf) - panic("getblk: size(%d) > maxbcachebuf(%d)\n", size, + if (size > maxbcachebuf) { + printf("getblkx: size(%d) > maxbcachebuf(%d)\n", size, maxbcachebuf); + return (EIO); + } if (!unmapped_buf_allowed) flags &= ~(GB_UNMAPPED | GB_KVAALLOC); @@ -4161,6 +4163,12 @@ newbuf_unlocked: vmio = vp->v_object != NULL; if (vmio) { maxsize = size + (offset & PAGE_MASK); + if (maxsize > maxbcachebuf) { + printf( + "getblkx: maxsize(%d) > maxbcachebuf(%d)\n", + maxsize, maxbcachebuf); + return (EIO); + } } else { maxsize = size; /* Do not allow non-VMIO notmapped buffers. */ |