In the Linux kernel, the following vulnerability has been resolved:<\/p>\n
btrfs: always wait for ordered extents to avoid OE races<\/p>\n
[BUG]
\nSyzbot reported a bug that there can be conflicting OEs for the same
\nrange:<\/p>\n
BTRFS critical (device loop4): panic in insert_ordered_extent:264: overlapping ordered extents, existing oe file_offset 16384 num_bytes 430080 flags 0x1089, new oe file_offset 16384 num_bytes 430080 flags 0x80 (errno=-17 Object alrea[ 179.162726][ T6897] BTRFS critical (device loop4): panic in insert_ordered_extent:264: overlapping ordered extents, existing oe file_offset 16384 num_bytes 430080 flags 0x1089, new oe file_offset 16384 num_bytes 430080 flags 0x80 (errno=-17 Object already exists)
\n ————[ cut here ]————
\n kernel BUG at fs\/btrfs\/ordered-data.c:264!
\n Hardware name: Google Google Compute Engine\/Google Compute Engine, BIOS Google 05\/09\/2026
\n RIP: 0010:btrfs_alloc_ordered_extent+0x943\/0xad0
\n Call Trace:<\/p>\n
cow_file_range+0x744\/0x12a0
\n fallback_to_cow+0x5ea\/0xa00
\n run_delalloc_nocow+0x110c\/0x17a0
\n btrfs_run_delalloc_range+0xbe4\/0x1c20
\n writepage_delalloc+0x104d\/0x1ba0
\n btrfs_writepages+0x1667\/0x28b0
\n do_writepages+0x338\/0x560
\n filemap_fdatawrite_range+0x1f2\/0x300
\n btrfs_fdatawrite_range+0x54\/0xf0
\n btrfs_direct_write+0x6a0\/0xc30
\n btrfs_do_write_iter+0x329\/0x790
\n do_iter_readv_writev+0x624\/0x8d0
\n vfs_writev+0x34c\/0x990
\n __se_sys_pwritev2+0x17a\/0x2a0
\n do_syscall_64+0x174\/0x580
\n entry_SYSCALL_64_after_hwframe+0x77\/0x7f<\/p>\n
—[ end trace 0000000000000000 ]—<\/p>\n
[CAUSE]
\nSince commit ff66fe666233 (“btrfs: fix incorrect buffered IO fallback
\nfor append direct writes”), if the direct IO finished short, we will
\nrevert the isize back to the original one, so that append writes can be
\nrespected during the buffered fallback.<\/p>\n
Normally we rely on lock_and_cleanup_extent_if_need() function during
\nbuffered writeback to wait for any existing ordered extents.<\/p>\n
But that ordered extent waiting only happens if the start_pos is inside
\nthe isize.
\nSince we have reverted the isize during failed direct IO, we will not
\nwait for any ordered extents.<\/p>\n
This means we can have a race where the direct IO OE is still in the
\ntree, finished but not yet removed, then we’re inserting the OE for the
\nbuffered write, causing the above crash.<\/p>\n
[FIX]
\nMake the OE wait to be unconditional, to handle the reverted isize
\nsituation.<\/p>\n
And since lock_and_cleanup_extent_if_need() now either lock the
\nextents or return -EAGAIN, also remove the branches that handles
\nno-extent-locked cases, and rename it to remove the “_if_need” suffix.<\/p>\n
The following micro benchmark shows the runtime difference for
\nbtrfs_buffered_write(), doing `xfs_io -f -c “pwrite 0 1m”` workload,
\nall values are the average runtime in nano seconds.<\/p>\n
function runtime | before | after
\n ———————————–+————-+—————
\n lock_and_cleanup_extent_if_need() | 58.2 | 183.0
\n btrfs_buffered_write() | 2115.6 | 2973.3<\/p>\n
The overall runtime of btrfs_buffered_write() is still pretty
\ntiny (still less than 3 micro seconds), I’d say the extra cost is still
\nacceptable.<\/p>\n
An alternative to fix this problem is to wait ordered extents during
\niomap_end() where the isize revert is done.<\/p>\n
But that solution will break nowait requirement, as if a nowait direct
\nIO finished short, we have to wait for the OEs unconditionally or the
\nnext append buffered IO can still hit the same problem.<\/p>\n
So here we have to move the wait cost to buffered write, but at least
\nthe code is slightly more streamline.<\/p>\n","protected":false},"excerpt":{"rendered":"
In the Linux kernel, the following vulnerability has been resolved:<\/p>\n
btrfs: always wait for ordered extents to avoid OE races<\/p>\n
[BUG]
\nSyzbot reported a bug that there can be conflicting OEs for the sam<\/p>\n","protected":false},"template":"","meta":[],"categories":[],"tags":[],"class_list":["post-25673","vuln","type-vuln","status-publish","hentry"],"_links":{"self":[{"href":"https:\/\/www.54iter.cn\/index.php?rest_route=\/wp\/v2\/vuln\/25673","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.54iter.cn\/index.php?rest_route=\/wp\/v2\/vuln"}],"about":[{"href":"https:\/\/www.54iter.cn\/index.php?rest_route=\/wp\/v2\/types\/vuln"}],"wp:attachment":[{"href":"https:\/\/www.54iter.cn\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=25673"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.54iter.cn\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=25673"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.54iter.cn\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=25673"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}