lkml.org 
[lkml]   [2015]   [Sep]   [3]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
SubjectData race in ext4_writepages
From
Hi!

We are working on a dynamic data race detector for the Linux kernel,
KernelThreadSanitizer (ktsan):
https://github.com/google/ktsan/wiki

We got a report while running ktsan on 4.2:

==================================================================
ThreadSanitizer: data-race in ext4_writepages

Write of size 8 by thread T210 (K741):
[< inlined >] ext4_writepages+0x10a6/0x1480
mpage_map_and_submit_extent fs/ext4/inode.c:2223
[<ffffffff8133bee6>] ext4_writepages+0x10a6/0x1480 fs/ext4/inode.c:2508
[<ffffffff811dbd23>] do_writepages+0x53/0x80 mm/page-writeback.c:2332
[<ffffffff812a76bf>] __writeback_single_inode+0x7f/0x530 fs/fs-writeback.c:1259
[<ffffffff812a7fd4>] writeback_sb_inodes+0x464/0x690 fs/fs-writeback.c:1516
[<ffffffff812a82c1>] __writeback_inodes_wb+0xc1/0x100 fs/fs-writeback.c:1562
[<ffffffff812a86ae>] wb_writeback+0x3ae/0x450 fs/fs-writeback.c:1666
[< inlined >] wb_workfn+0x203/0x780 wb_do_writeback
fs/fs-writeback.c:1801
[<ffffffff812a91c3>] wb_workfn+0x203/0x780 fs/fs-writeback.c:1852
[<ffffffff810b06ce>] process_one_work+0x28e/0x710 kernel/workqueue.c:2036
[<ffffffff810b1299>] worker_thread+0xb9/0x750 kernel/workqueue.c:2170
[<ffffffff810b9c61>] kthread+0x161/0x180 kernel/kthread.c:209
[<ffffffff81eb0a1f>] ret_from_fork+0x3f/0x70 arch/x86/entry/entry_64.S:526

Previous read of size 8 by thread T480 (K7162):
[< inlined >] ext4_mark_iloc_dirty+0x5ec/0xe20
ext4_do_update_inode fs/ext4/inode.c:4441
[<ffffffff8133a1ac>] ext4_mark_iloc_dirty+0x5ec/0xe20 fs/ext4/inode.c:4937
[<ffffffff8133ab8b>] ext4_mark_inode_dirty+0xdb/0x390 fs/ext4/inode.c:5053
[<ffffffff8133f929>] ext4_dirty_inode+0x59/0x80 fs/ext4/inode.c:5085
[<ffffffff812a7319>] __mark_inode_dirty+0x2c9/0x5f0 fs/fs-writeback.c:2015
[<ffffffff8128a58e>] generic_update_time+0xbe/0x150 fs/inode.c:1566
[< inlined >] file_update_time+0x112/0x1b0 update_time fs/inode.c:1582
[<ffffffff812890f2>] file_update_time+0x112/0x1b0 fs/inode.c:1785
[<ffffffff811cb175>] __generic_file_write_iter+0x105/0x2e0 mm/filemap.c:2570
[<ffffffff8132c3a4>] ext4_file_write_iter+0x254/0x740 fs/ext4/file.c:170
[< inlined >] __vfs_write+0x19c/0x1e0 new_sync_write fs/read_write.c:478
[<ffffffff8125d48c>] __vfs_write+0x19c/0x1e0 fs/read_write.c:491
[<ffffffff8125dde6>] vfs_write+0xf6/0x2a0 fs/read_write.c:538
[< inlined >] SyS_write+0x6b/0xd0 SYSC_write fs/read_write.c:585
[<ffffffff8125f37b>] SyS_write+0x6b/0xd0 fs/read_write.c:577
[<ffffffff81eb062e>] entry_SYSCALL_64_fastpath+0x12/0x71
arch/x86/entry/entry_64.S:186
==================================================================

One thread reads 'ei->i_disksize' multiple times while it's being
updated by another.
ei->i_disksize is 64 bits long so reads and writes are not atomic.

Thanks!


\
 
 \ /
  Last update: 2015-09-03 18:01    [W:0.038 / U:5.944 seconds]
©2003-2017 Jasper Spaans. hosted at Digital OceanAdvertise on this site