On Mon, May 20, 2019 at 06:15:42AM +0000, bugzilla-daemon@xxxxxxxxxxxxxxxxxxx wrote: > https://bugzilla.kernel.org/show_bug.cgi?id=203653 > > Bug ID: 203653 > Summary: XFS: Internal error xlog_clear_stale_blocks(2) at line > 1794 of file ../fs/xfs/xfs_log_recover.c > Product: File System > Version: 2.5 > Kernel Version: 5.1.3 > Hardware: All > OS: Linux > Tree: Mainline > Status: NEW > Severity: normal > Priority: P1 > Component: XFS > Assignee: filesystem_xfs@xxxxxxxxxxxxxxxxxxxxxx > Reporter: midwinter1993@xxxxxxxxx > Regression: No > > Created attachment 282847 > --> https://bugzilla.kernel.org/attachment.cgi?id=282847&action=edit > Crafted image > > ### When mounting the image (in the attached file), XFS prints an internal > error. > ### Tested under kernel 5.1.3 and 4.4.0. > > ### Reproduce > > 1. download `image.tar.gz` > 2. uncompress it: > > tar -xzvf image.tar.gz > > 3. mount it: > > mkdir dd > > mount bingo.img dd > > 4. check result: > > dmesg > > --- Following is the core dump (under kernel 5.1.3) --- > ``` > [ 67.737771] XFS (loop0): Internal error xlog_clear_stale_blocks(2) at line > 1794 of file ../fs/xfs/xfs_log_recover.c. Caller xlog_find_tail+0x622/0x7b0 > [ 67.742886] CPU: 0 PID: 2114 Comm: mount Not tainted 5.1.3 #2 > [ 67.744338] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS > Ubuntu-1.8.2-1ubuntu1 04/01/2014 > [ 67.746410] Call Trace: > [ 67.746983] dump_stack+0x5b/0x8b > [ 67.747718] xlog_clear_stale_blocks+0x2d5/0x3d0 > [ 67.748389] xlog_find_tail+0x622/0x7b0 > [ 67.748934] ? _sched_setscheduler+0x107/0x180 > [ 67.749559] ? xlog_verify_head+0x4d0/0x4d0 > [ 67.750148] ? __sched_setscheduler+0x1c90/0x1d70 > [ 67.750810] ? check_preempt_wakeup+0x2c6/0x840 > [ 67.751440] ? ttwu_do_wakeup.isra.92+0x13/0x2b0 > [ 67.752079] xlog_recover+0x89/0x470 > [ 67.752551] ? xlog_find_tail+0x7b0/0x7b0 > [ 67.753084] ? kmem_alloc+0x81/0x130 > [ 67.753562] xfs_log_mount+0x291/0x660 > [ 67.754063] xfs_mountfs+0x1059/0x1bd0 > [ 67.754565] ? xfs_mount_reset_sbqflags+0x130/0x130 > [ 67.755216] ? kasan_unpoison_shadow+0x31/0x40 > [ 67.755810] ? __kasan_kmalloc+0xd5/0xf0 > [ 67.756332] ? kasan_unpoison_shadow+0x31/0x40 > [ 67.756918] ? __kasan_kmalloc+0xd5/0xf0 > [ 67.757438] ? kmem_alloc+0x81/0x130 > [ 67.757912] ? xfs_filestream_put_ag+0x30/0x30 > [ 67.758499] ? xfs_mru_cache_create+0x33b/0x530 > [ 67.759099] xfs_fs_fill_super+0xbca/0x11d0 > [ 67.759660] ? xfs_test_remount_options+0x70/0x70 > [ 67.760287] mount_bdev+0x25d/0x310 > [ 67.760751] ? xfs_finish_flags+0x390/0x390 > [ 67.761306] legacy_get_tree+0xe4/0x1c0 > [ 67.761817] vfs_get_tree+0x80/0x370 > [ 67.762295] do_mount+0xd8c/0x2320 > [ 67.762755] ? lockref_put_return+0x130/0x130 > [ 67.763333] ? __fsnotify_update_child_dentry_flags.part.3+0x2e0/0x2e0 > [ 67.764191] ? copy_mount_string+0x20/0x20 > [ 67.764732] ? kasan_unpoison_shadow+0x31/0x40 > [ 67.765315] ? __kasan_kmalloc+0xd5/0xf0 > [ 67.765830] ? strndup_user+0x42/0x90 > [ 67.766316] ? __kmalloc_track_caller+0xc7/0x1c0 > [ 67.766928] ? _copy_from_user+0x73/0xa0 > [ 67.767449] ? memdup_user+0x39/0x60 > [ 67.767931] ksys_mount+0x79/0xc0 > [ 67.768376] __x64_sys_mount+0xb5/0x150 > [ 67.768889] do_syscall_64+0x8c/0x280 > [ 67.769377] ? async_page_fault+0x8/0x30 > [ 67.769906] entry_SYSCALL_64_after_hwframe+0x44/0xa9 > [ 67.770577] RIP: 0033:0x7f0c5185d48a > [ 67.771053] Code: 48 8b 0d 11 fa 2a 00 f7 d8 64 89 01 48 83 c8 ff c3 66 2e > 0f 1f 84 00 00 00 00 00 0f 1f 44 00 00 49 89 ca b8 a5 00 00 00 > 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 8b 0d de f9 2a 00 f7 d8 64 89 01 48 > [ 67.773494] RSP: 002b:00007fff68c36b38 EFLAGS: 00000202 ORIG_RAX: > 00000000000000a5 > [ 67.774493] RAX: ffffffffffffffda RBX: 00005565e0f62080 RCX: > 00007f0c5185d48a > [ 67.775420] RDX: 00005565e0f68db0 RSI: 00005565e0f63f60 RDI: > 00005565e0f68d90 > [ 67.776358] RBP: 0000000000000000 R08: 0000000000000000 R09: > 0000000000000020 > [ 67.777292] R10: 00000000c0ed0000 R11: 0000000000000202 R12: > 00005565e0f68d90 > [ 67.778227] R13: 00005565e0f68db0 R14: 0000000000000000 R15: > 00000000ffffffff > [ 67.779226] XFS (loop0): failed to locate log tail > [ 67.779881] XFS (loop0): log mount/recovery failed: error -117 > [ 67.780969] XFS (loop0): log mount failed > ``` > > -- > You are receiving this mail because: > You are watching the assignee of the bug. How exactly was this image created? Was this corruption encountered after a crash or was the image manually crafted? xfs_logprint shows the following last few records in the log: ============================================================================ cycle: 1 version: 2 lsn: 1,36 tail_lsn: 1,36 length of Log Record: 512 prev offset: 34 num ops: 1 uuid: e703f609-b433-4e72-a7d5-4872d7e81ae5 format: little endian linux h_size: 32768 ---------------------------------------------------------------------------- Oper (0): tid: cfbab016 len: 8 clientid: LOG flags: UNMOUNT Unmount filesystem ============================================================================ cycle: 1 version: 2 lsn: 1,90 tail_lsn: 1,38 length of Log Record: 512 prev offset: 36 num ops: 1 uuid: e703f609-b433-4e72-a7d5-4872d7e81ae5 format: little endian linux h_size: 32768 ---------------------------------------------------------------------------- Oper (0): tid: 2bb1cec8 len: 8 clientid: LOG flags: UNMOUNT Unmount filesystem ============================================================================ cycle: 1 version: 2 lsn: 1,40 tail_lsn: 218,40 length of Log Record: 512 prev offset: 38 num ops: 1 uuid: e703f609-b433-4e72-a7d5-4872d7e81ae5 format: little endian linux h_size: 32768 ---------------------------------------------------------------------------- Oper (0): tid: 3467bf1d len: 8 clientid: LOG flags: UNMOUNT Unmount filesystem ... So the log appears clean and relatively sane until we get to the last couple of unmount records. The lsn of the second to last looks bogus given the jump from the previous and that it's beyond the last record. The tail_lsn of the final record is clearly bogus and is what triggers the corruption error on log recovery because we expect the head and tail of the log to respect constraints expected of a circular log. Also note that this isn't exactly a crash. The dmesg output is simply more verbose error output generated by the log recovery code. All in all, this seems like expected behavior for the associated image. The more interesting question is how this corruption was caused. Brian
