Bad page state at unmount
Hi folks,
I hit this running check-parallel a moment ago:
[80180.074658] BUG: Bad page cache in process umount pfn:7655f4 [80180.077259] page: refcount:9 mapcount:1 mapping:00000000ecd1b54a index:0x0 pfn:0x7655f4 [80180.080573] head: order:2 mapcount:4 entire_mapcount:0 nr_pages_mapped:4 pincount:0 [80180.083615] memcg:ffff888104f36000 [80180.084977] aops:xfs_address_space_operations ino:84 [80180.087175] flags: 0x17ffffc000016d(locked|referenced|uptodate|lru|active|head|node=0|zone=2|lastcpupid=0x1fffff) [80180.091380] raw: 0017ffffc000016d ffffea001745c648 ffffea0012b1da08 ffff8891726dae98 [80180.094469] raw: 0000000000000000 0000000000000000 0000000900000000 ffff888104f36000 [80180.097740] head: 0017ffffc000016d ffffea001745c648 ffffea0012b1da08 ffff8891726dae98 [80180.100988] head: 0000000000000000 0000000000000000 0000000900000000 ffff888104f36000 [80180.104129] head: 0017ffffc0000202 ffffea001d957d01 ffffffff00000003 0000000000000004 [80180.107232] head: 0000000000000004 0000000000000000 0000000000000000 0000000000000000 [80180.110338] page dumped because: still mapped when deleted [80180.112755] CPU: 32 UID: 0 PID: 832271 Comm: umount Not tainted 6.14.0-rc2-dgc+ #302 [80180.112757] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.16.3-debian-1.16.3-2 04/01/2014 [80180.112760] Call Trace: [80180.112763] <TASK> [80180.112766] dump_stack_lvl+0x3d/0xa0 [80180.112773] dump_stack+0x10/0x17 [80180.112775] filemap_unaccount_folio+0x151/0x1e0 [80180.112779] delete_from_page_cache_batch+0x61/0x2f0 [80180.112787] truncate_inode_pages_range+0x122/0x3e0 [80180.112807] truncate_inode_pages_final+0x40/0x50 [80180.112809] evict+0x1af/0x310 [80180.112817] evict_inodes+0x66/0xc0 [80180.112818] generic_shutdown_super+0x3c/0x160 [80180.112821] kill_block_super+0x1b/0x40 [80180.112823] xfs_kill_sb+0x12/0x30 [80180.112824] deactivate_locked_super+0x38/0x100 [80180.112826] deactivate_super+0x41/0x50 [80180.112828] cleanup_mnt+0x9f/0x160 [80180.112830] __cleanup_mnt+0x12/0x20 [80180.112831] task_work_run+0x89/0xb0 [80180.112833] resume_user_mode_work+0x4f/0x60 [80180.112836] syscall_exit_to_user_mode+0x76/0xb0 [80180.112838] do_syscall_64+0x74/0x130 [80180.112840] ? exc_page_fault+0x62/0xc0 [80180.112841] entry_SYSCALL_64_after_hwframe+0x76/0x7e .... [80180.131293] BUG: Bad page cache in process umount pfn:4ac768 [80180.131296] page: refcount:9 mapcount:1 mapping:00000000ecd1b54a index:0x4 pfn:0x4ac768 [80180.131299] head: order:2 mapcount:4 entire_mapcount:0 nr_pages_mapped:4 pincount:0 [80180.131301] memcg:ffff888104f36000 [80180.131302] aops:xfs_address_space_operations ino:84 [80180.218440] flags: 0x17ffffc000016d(locked|referenced|uptodate|lru|active|head|node=0|zone=2|lastcpupid=0x1fffff) [80180.222779] raw: 0017ffffc000016d ffffea001d957d08 ffffea000d980b08 ffff8891726dae98 [80180.226376] raw: 0000000000000004 0000000000000000 0000000900000000 ffff888104f36000 [80180.229546] head: 0017ffffc000016d ffffea001d957d08 ffffea000d980b08 ffff8891726dae98 [80180.232954] head: 0000000000000004 0000000000000000 0000000900000000 ffff888104f36000 [80180.232956] head: 0017ffffc0000202 ffffea0012b1da01 ffffffff00000003 0000000000000004 [80180.232958] head: 0000000500000004 0000000000000000 0000000000000000 0000000000000000 [80180.232958] page dumped because: still mapped when deleted [80180.232961] CPU: 32 UID: 0 PID: 832271 Comm: umount Tainted: G B 6.14.0-rc2-dgc+ #302 [80180.232965] Tainted: [B]=BAD_PAGE ..... [80180.233052] BUG: Bad page cache in process umount pfn:36602c [80180.241951] page: refcount:9 mapcount:1 mapping:00000000ecd1b54a index:0x8 pfn:0x36602c [80180.241955] head: order:2 mapcount:4 entire_mapcount:0 nr_pages_mapped:4 pincount:0 [80180.241957] memcg:ffff888104f36000 [80180.241958] aops:xfs_address_space_operations ino:84 [80180.241961] flags: 0x17ffffc000016d(locked|referenced|uptodate|lru|active|head|node=0|zone=2|lastcpupid=0x1fffff) [80180.241965] raw: 0017ffffc000016d ffffea0012b1da08 ffffea000d585508 ffff8891726dae98 [80180.241966] raw: 0000000000000008 0000000000000000 0000000900000000 ffff888104f36000 [80180.241967] head: 0017ffffc000016d ffffea0012b1da08 ffffea000d585508 ffff8891726dae98 [80180.241969] head: 0000000000000008 0000000000000000 0000000900000000 ffff888104f36000 [80180.241970] head: 0017ffffc0000202 ffffea000d980b01 ffffffff00000003 0000000000000004 [80180.241971] head: 0000000500000004 0000000000000000 0000000000000000 0000000000000000 [80180.241972] page dumped because: still mapped when deleted [80180.241974] CPU: 32 UID: 0 PID: 832271 Comm: umount Tainted: G B 6.14.0-rc2-dgc+ #302 [80180.241976] Tainted: [B]=BAD_PAGE
I don't know which fstest triggered it, but this is a new failure that I haven't seen before. It looks like 3 consecutive order-2 folios on the same mapping all have the same problem....
The kernel was a post 6.14-rc2 kernel with linux-xfs/for-next merged into it. I'm going to update the kernel to TOT to see if this reproduces again, but I've only seen this once in dozens of tests runs on this kernel, so....
Has anyone seen something similar or have any ideas where to look?
-Dave.
Signature · 2 lines
Dave Chinner [email protected]
Quoted 79 lines · Dave Chinner
Hi folks,
I hit this running check-parallel a moment ago:
[80180.074658] BUG: Bad page cache in process umount pfn:7655f4 [80180.077259] page: refcount:9 mapcount:1 mapping:00000000ecd1b54a index:0x0 pfn:0x7655f4 [80180.080573] head: order:2 mapcount:4 entire_mapcount:0 nr_pages_mapped:4 pincount:0 [80180.083615] memcg:ffff888104f36000 [80180.084977] aops:xfs_address_space_operations ino:84 [80180.087175] flags: 0x17ffffc000016d(locked|referenced|uptodate|lru|active|head|node=0|zone=2|lastcpupid=0x1fffff) [80180.091380] raw: 0017ffffc000016d ffffea001745c648 ffffea0012b1da08 ffff8891726dae98 [80180.094469] raw: 0000000000000000 0000000000000000 0000000900000000 ffff888104f36000 [80180.097740] head: 0017ffffc000016d ffffea001745c648 ffffea0012b1da08 ffff8891726dae98 [80180.100988] head: 0000000000000000 0000000000000000 0000000900000000 ffff888104f36000 [80180.104129] head: 0017ffffc0000202 ffffea001d957d01 ffffffff00000003 0000000000000004 [80180.107232] head: 0000000000000004 0000000000000000 0000000000000000 0000000000000000 [80180.110338] page dumped because: still mapped when deleted [80180.112755] CPU: 32 UID: 0 PID: 832271 Comm: umount Not tainted 6.14.0-rc2-dgc+ #302 [80180.112757] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.16.3-debian-1.16.3-2 04/01/2014 [80180.112760] Call Trace: [80180.112763] <TASK> [80180.112766] dump_stack_lvl+0x3d/0xa0 [80180.112773] dump_stack+0x10/0x17 [80180.112775] filemap_unaccount_folio+0x151/0x1e0 [80180.112779] delete_from_page_cache_batch+0x61/0x2f0 [80180.112787] truncate_inode_pages_range+0x122/0x3e0 [80180.112807] truncate_inode_pages_final+0x40/0x50 [80180.112809] evict+0x1af/0x310 [80180.112817] evict_inodes+0x66/0xc0 [80180.112818] generic_shutdown_super+0x3c/0x160 [80180.112821] kill_block_super+0x1b/0x40 [80180.112823] xfs_kill_sb+0x12/0x30 [80180.112824] deactivate_locked_super+0x38/0x100 [80180.112826] deactivate_super+0x41/0x50 [80180.112828] cleanup_mnt+0x9f/0x160 [80180.112830] __cleanup_mnt+0x12/0x20 [80180.112831] task_work_run+0x89/0xb0 [80180.112833] resume_user_mode_work+0x4f/0x60 [80180.112836] syscall_exit_to_user_mode+0x76/0xb0 [80180.112838] do_syscall_64+0x74/0x130 [80180.112840] ? exc_page_fault+0x62/0xc0 [80180.112841] entry_SYSCALL_64_after_hwframe+0x76/0x7e .... [80180.131293] BUG: Bad page cache in process umount pfn:4ac768 [80180.131296] page: refcount:9 mapcount:1 mapping:00000000ecd1b54a index:0x4 pfn:0x4ac768 [80180.131299] head: order:2 mapcount:4 entire_mapcount:0 nr_pages_mapped:4 pincount:0 [80180.131301] memcg:ffff888104f36000 [80180.131302] aops:xfs_address_space_operations ino:84 [80180.218440] flags: 0x17ffffc000016d(locked|referenced|uptodate|lru|active|head|node=0|zone=2|lastcpupid=0x1fffff) [80180.222779] raw: 0017ffffc000016d ffffea001d957d08 ffffea000d980b08 ffff8891726dae98 [80180.226376] raw: 0000000000000004 0000000000000000 0000000900000000 ffff888104f36000 [80180.229546] head: 0017ffffc000016d ffffea001d957d08 ffffea000d980b08 ffff8891726dae98 [80180.232954] head: 0000000000000004 0000000000000000 0000000900000000 ffff888104f36000 [80180.232956] head: 0017ffffc0000202 ffffea0012b1da01 ffffffff00000003 0000000000000004 [80180.232958] head: 0000000500000004 0000000000000000 0000000000000000 0000000000000000 [80180.232958] page dumped because: still mapped when deleted [80180.232961] CPU: 32 UID: 0 PID: 832271 Comm: umount Tainted: G B 6.14.0-rc2-dgc+ #302 [80180.232965] Tainted: [B]=BAD_PAGE ..... [80180.233052] BUG: Bad page cache in process umount pfn:36602c [80180.241951] page: refcount:9 mapcount:1 mapping:00000000ecd1b54a index:0x8 pfn:0x36602c [80180.241955] head: order:2 mapcount:4 entire_mapcount:0 nr_pages_mapped:4 pincount:0 [80180.241957] memcg:ffff888104f36000 [80180.241958] aops:xfs_address_space_operations ino:84 [80180.241961] flags: 0x17ffffc000016d(locked|referenced|uptodate|lru|active|head|node=0|zone=2|lastcpupid=0x1fffff) [80180.241965] raw: 0017ffffc000016d ffffea0012b1da08 ffffea000d585508 ffff8891726dae98 [80180.241966] raw: 0000000000000008 0000000000000000 0000000900000000 ffff888104f36000 [80180.241967] head: 0017ffffc000016d ffffea0012b1da08 ffffea000d585508 ffff8891726dae98 [80180.241969] head: 0000000000000008 0000000000000000 0000000900000000 ffff888104f36000 [80180.241970] head: 0017ffffc0000202 ffffea000d980b01 ffffffff00000003 0000000000000004 [80180.241971] head: 0000000500000004 0000000000000000 0000000000000000 0000000000000000 [80180.241972] page dumped because: still mapped when deleted [80180.241974] CPU: 32 UID: 0 PID: 832271 Comm: umount Tainted: G B 6.14.0-rc2-dgc+ #302 [80180.241976] Tainted: [B]=BAD_PAGEI don't know which fstest triggered it, but this is a new failure that I haven't seen before. It looks like 3 consecutive order-2 folios on the same mapping all have the same problem....
Quoted 6 lines
The kernel was a post 6.14-rc2 kernel with linux-xfs/for-next merged into it. I'm going to update the kernel to TOT to see if this reproduces again, but I've only seen this once in dozens of tests runs on this kernel, so....
Has anyone seen something similar or have any ideas where to look?
I didn't see anything like that on -rc2, and -rc3 doesn't seem to be showing that either.
--D
Quoted 5 lines
-Dave.
Signature · 2 lines
Dave Chinner [email protected]
Quoted 4 lines · Dave Chinner
Hi folks,
I hit this running check-parallel a moment ago:
Quoted 13 lines
[80180.074658] BUG: Bad page cache in process umount pfn:7655f4 [80180.077259] page: refcount:9 mapcount:1 mapping:00000000ecd1b54a index:0x0 pfn:0x7655f4 [80180.080573] head: order:2 mapcount:4 entire_mapcount:0 nr_pages_mapped:4 pincount:0 [80180.083615] memcg:ffff888104f36000 [80180.084977] aops:xfs_address_space_operations ino:84 [80180.087175] flags: 0x17ffffc000016d(locked|referenced|uptodate|lru|active|head|node=0|zone=2|lastcpupid=0x1fffff) [80180.091380] raw: 0017ffffc000016d ffffea001745c648 ffffea0012b1da08 ffff8891726dae98 [80180.094469] raw: 0000000000000000 0000000000000000 0000000900000000 ffff888104f36000 [80180.097740] head: 0017ffffc000016d ffffea001745c648 ffffea0012b1da08 ffff8891726dae98 [80180.100988] head: 0000000000000000 0000000000000000 0000000900000000 ffff888104f36000 [80180.104129] head: 0017ffffc0000202 ffffea001d957d01 ffffffff00000003 0000000000000004 [80180.107232] head: 0000000000000004 0000000000000000 0000000000000000 0000000000000000 [80180.110338] page dumped because: still mapped when deleted
Do you have CONFIG_PT_RECLAIM enabled? If so, it's buggy (see linux-mm for a fix if you don't want to disable it).
Quoted 4 lines · Matthew Wilcox
Quoted 3 lines · Dave Chinner
Hi folks,
I hit this running check-parallel a moment ago:
FWIW, I just hit the bug again...
Quoted 14 lines
Quoted 5 lines
[80180.074658] BUG: Bad page cache in process umount pfn:7655f4 [80180.077259] page: refcount:9 mapcount:1 mapping:00000000ecd1b54a index:0x0 pfn:0x7655f4 [80180.080573] head: order:2 mapcount:4 entire_mapcount:0 nr_pages_mapped:4 pincount:0 [80180.083615] memcg:ffff888104f36000 [80180.084977] aops:xfs_address_space_operations ino:84Quoted 8 lines
[80180.087175] flags: 0x17ffffc000016d(locked|referenced|uptodate|lru|active|head|node=0|zone=2|lastcpupid=0x1fffff) [80180.091380] raw: 0017ffffc000016d ffffea001745c648 ffffea0012b1da08 ffff8891726dae98 [80180.094469] raw: 0000000000000000 0000000000000000 0000000900000000 ffff888104f36000 [80180.097740] head: 0017ffffc000016d ffffea001745c648 ffffea0012b1da08 ffff8891726dae98 [80180.100988] head: 0000000000000000 0000000000000000 0000000900000000 ffff888104f36000 [80180.104129] head: 0017ffffc0000202 ffffea001d957d01 ffffffff00000003 0000000000000004 [80180.107232] head: 0000000000000004 0000000000000000 0000000000000000 0000000000000000 [80180.110338] page dumped because: still mapped when deleted
Quoted 1 line
Do you have
CONFIG_PT_RECLAIMenabled?
Never heard of it - what's that do, and how long before the fix gets into the mainline tree?
Quoted 2 lines
If so, it's buggy (see linux-mm
for a fix if you don't want to disable it).
I'll disable it for now, don't feel like going on a fishing trip.
-Dave.
Signature · 2 lines
Dave Chinner [email protected]