btrfs/071 is unhappy on 6.18-rc2
I just kicked off a baseline run with the xfstests volume group and a SCRATCH_DEV_POOL with 5 virtual nvme devices to test a VFS change that affects іt a little, and it does not seem too happy.
btrfs/071 gets into slab poisoning:
[ 279.241695] BTRFS info (device nvme1n1 state M): use zlib compression, level 3 [ 279.247651] Oops: general protection fault, probably for non-canonical address 0x6b6b6b6b6b6b6d73:I [ 279.250656] CPU: 1 UID: 0 PID: 82037 Comm: btrfs-cleaner Tainted: GN 6.18.0-rc2 [ 279.250656] Tainted: [N]=TEST [ 279.250656] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.16.3-debian-1.16.3-2 04/01/4 [ 279.250656] RIP: 0010:btrfs_kill_all_delayed_nodes+0x145/0x1e0 [ 279.250656] Code: 08 48 c1 e5 03 4b 8b 5c 3d 00 48 89 df e8 23 d0 ff ff 48 85 db 74 0f 4a 8d 54 3c0 [ 279.250656] RSP: 0018:ffffc9000138bdc0 EFLAGS: 00010246 [ 279.250656] RAX: 6b6b6b6b6b6b6b6b RBX: ffff88810dad3d58 RCX: 0000000000000000 [ 279.250656] RDX: 0000000000000001 RSI: 0000000000000286 RDI: 00000000ffffffff [ 279.250656] RBP: 0000000000000008 R08: ffff88810dad3f60 R09: ffff88810dad3f10 [ 279.250656] R10: 000000000000000d R11: ffff88810dad3d58 R12: ffff88811bdfbc18 [ 279.250656] R13: ffffc9000138bdc8 R14: ffff88811bdfb800 R15: 0000000000000000 [ 279.250656] FS: 0000000000000000(0000) GS:ffff8883ef66a000(0000) knlGS:0000000000000000 [ 279.250656] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 279.250656] CR2: 00007fcba10194c8 CR3: 00000001217a3002 CR4: 0000000000772ef0 [ 279.250656] PKRU: 55555554 [ 279.250656] Call Trace: [ 279.250656] <TASK> [ 279.250656] ? __schedule+0x52c/0xb60 [ 279.250656] btrfs_clean_one_deleted_snapshot+0x72/0x100 [ 279.250656] cleaner_kthread+0xd3/0x150 [ 279.250656] ? __pfx_cleaner_kthread+0x10/0x10 [ 279.250656] kthread+0x109/0x220 [ 279.250656] ? __pfx_kthread+0x10/0x10 [ 279.250656] ? __pfx_kthread+0x10/0x10 [ 279.250656] ret_from_fork+0x120/0x160 [ 279.250656] ? __pfx_kthread+0x10/0x10 [ 279.250656] ret_from_fork_asm+0x1a/0x30 [ 279.250656] </TASK> [ 279.250656] Modules linked in: kvm_intel kvm irqbypass [ 279.277534] ---[ end trace 0000000000000000 ]---
similar things repeat a few times, and then it loops basically forever doing device replacements, I waited for 30 minutes before killing it.
Quoted 4 lines · Christoph Hellwig
I just kicked off a baseline run with the xfstests volume group and a
SCRATCH_DEV_POOLwith 5 virtual nvme devices to test a VFS change that affects іt a little, and it does not seem too happy.
Quoted 8 lines
btrfs/071 gets into slab poisoning:
[ 279.241695] BTRFS info (device nvme1n1 state M): use zlib compression, level 3 [ 279.247651] Oops: general protection fault, probably for non-canonical address 0x6b6b6b6b6b6b6d73:I [ 279.250656] CPU: 1 UID: 0 PID: 82037 Comm: btrfs-cleaner Tainted: GN 6.18.0-rc2 [ 279.250656] Tainted: [N]=TEST [ 279.250656] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.16.3-debian-1.16.3-2 04/01/4 [ 279.250656] RIP: 0010:btrfs_kill_all_delayed_nodes+0x145/0x1e0
Any line number/context and reproducibility?
Quoted 3 lines
[ 279.250656] Code: 08 48 c1 e5 03 4b 8b 5c 3d 00 48 89 df e8 23 d0 ff ff 48 85 db 74 0f 4a 8d 54 3c0 [ 279.250656] RSP: 0018:ffffc9000138bdc0 EFLAGS: 00010246 [ 279.250656] RAX: 6b6b6b6b6b6b6b6b
This looks like POISON_FREE, so some use-after-free bug?
If you're able to reproduce, mind to try KASAN?
As I just checked my logs, no failures on btrfs/071 recorded yet (but not on upstream rc2 yet)
Thanks,
Qu
Quoted 28 lines
RBX:
ffff88810dad3d58RCX: 0000000000000000[ 279.250656] RDX: 0000000000000001 RSI: 0000000000000286 RDI: 00000000ffffffff [ 279.250656] RBP: 0000000000000008 R08: ffff88810dad3f60 R09: ffff88810dad3f10 [ 279.250656] R10: 000000000000000d R11: ffff88810dad3d58 R12: ffff88811bdfbc18 [ 279.250656] R13: ffffc9000138bdc8 R14: ffff88811bdfb800 R15: 0000000000000000 [ 279.250656] FS: 0000000000000000(0000) GS:ffff8883ef66a000(0000) knlGS:0000000000000000 [ 279.250656] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 279.250656] CR2: 00007fcba10194c8 CR3: 00000001217a3002 CR4: 0000000000772ef0 [ 279.250656] PKRU: 55555554 [ 279.250656] Call Trace: [ 279.250656] <TASK> [ 279.250656] ? __schedule+0x52c/0xb60 [ 279.250656] btrfs_clean_one_deleted_snapshot+0x72/0x100 [ 279.250656] cleaner_kthread+0xd3/0x150 [ 279.250656] ? __pfx_cleaner_kthread+0x10/0x10 [ 279.250656] kthread+0x109/0x220 [ 279.250656] ? __pfx_kthread+0x10/0x10 [ 279.250656] ? __pfx_kthread+0x10/0x10 [ 279.250656] ret_from_fork+0x120/0x160 [ 279.250656] ? __pfx_kthread+0x10/0x10 [ 279.250656] ret_from_fork_asm+0x1a/0x30 [ 279.250656] </TASK> [ 279.250656] Modules linked in: kvm_intel kvm irqbypass [ 279.277534] ---[ end trace 0000000000000000 ]---similar things repeat a few times, and then it loops basically forever doing device replacements, I waited for 30 minutes before killing it.
Quoted 7 lines · Qu Wenruo
Quoted 5 lines
[ 279.247651] Oops: general protection fault, probably for non-canonical address 0x6b6b6b6b6b6b6d73:I [ 279.250656] CPU: 1 UID: 0 PID: 82037 Comm: btrfs-cleaner Tainted: GN 6.18.0-rc2 [ 279.250656] Tainted: [N]=TEST [ 279.250656] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.16.3-debian-1.16.3-2 04/01/4 [ 279.250656] RIP: 0010:btrfs_kill_all_delayed_nodes+0x145/0x1e0Any line number/context
Nope, that's it. Last lines before are:
[ 62.492209] BTRFS info (device nvme1n1): first mount of filesystem 975f6fd4-b50f-4f3d-8112-319c [ 62.492520] BTRFS info (device nvme1n1): using crc32c (crc32c-lib) checksum algorithm [ 62.510951] BTRFS info (device nvme1n1): checking UUID tree [ 62.511230] BTRFS info (device nvme1n1): enabling ssd optimizations [ 62.511452] BTRFS info (device nvme1n1): turning on async discard [ 62.511728] BTRFS info (device nvme1n1): enabling free space tree [ 62.642011] BTRFS info (device nvme1n1 state M): use zlib compression, level 3
Quoted 1 line
and reproducibility?
100% over a few runs.
Quoted 3 lines
If you're able to reproduce, mind to try KASAN?
As I just checked my logs, no failures on btrfs/071 recorded yet (but not on upstream rc2 yet)
A bit busy right now, but I'll try to do a KASAN run later.
Quoted 10 lines · Christoph Hellwig
Quoted 7 lines · Qu Wenruo
Quoted 5 lines
[ 279.247651] Oops: general protection fault, probably for non-canonical address 0x6b6b6b6b6b6b6d73:I [ 279.250656] CPU: 1 UID: 0 PID: 82037 Comm: btrfs-cleaner Tainted: GN 6.18.0-rc2 [ 279.250656] Tainted: [N]=TEST [ 279.250656] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.16.3-debian-1.16.3-2 04/01/4 [ 279.250656] RIP: 0010:btrfs_kill_all_delayed_nodes+0x145/0x1e0Any line number/context
Nope, that's it. Last lines before are:
I mean the code line number extracted from that RIP.
But I'll try to reproduce it after the recent direct IO problem solved.
Thanks,
Qu
Quoted 20 lines
[ 62.492209] BTRFS info (device nvme1n1): first mount of filesystem 975f6fd4-b50f-4f3d-8112-319c [ 62.492520] BTRFS info (device nvme1n1): using crc32c (crc32c-lib) checksum algorithm [ 62.510951] BTRFS info (device nvme1n1): checking UUID tree [ 62.511230] BTRFS info (device nvme1n1): enabling ssd optimizations [ 62.511452] BTRFS info (device nvme1n1): turning on async discard [ 62.511728] BTRFS info (device nvme1n1): enabling free space tree [ 62.642011] BTRFS info (device nvme1n1 state M): use zlib compression, level 3Quoted 1 line
and reproducibility?
100% over a few runs.
Quoted 3 lines
If you're able to reproduce, mind to try KASAN?
As I just checked my logs, no failures on btrfs/071 recorded yet (but not on upstream rc2 yet)A bit busy right now, but I'll try to do a KASAN run later.
KASAN output:
[ 75.341543] ================================================================== [ 75.341824] BUG: KASAN: slab-use-after-free in btrfs_kill_all_delayed_nodes+0x46f/0x4c0 [ 75.342082] Read of size 8 at addr ffff88812389f380 by task btrfs-cleaner/4493 [ 75.342310] [ 75.342369] CPU: 1 UID: 0 PID: 4493 Comm: btrfs-cleaner Tainted: G N 6.18.0-rc2+ #4115 PREEMPT(f [ 75.342372] Tainted: [N]=TEST [ 75.342373] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.16.3-debian-1.16.3-2 04/01/2014 [ 75.342374] Call Trace: [ 75.342375] <TASK> [ 75.342376] dump_stack_lvl+0x4b/0x70 [ 75.342379] print_report+0x174/0x4e7 [ 75.342382] ? __virt_addr_valid+0x1bb/0x2f0 [ 75.342384] ? btrfs_kill_all_delayed_nodes+0x46f/0x4c0 [ 75.342385] kasan_report+0xd2/0x100 [ 75.342387] ? btrfs_kill_all_delayed_nodes+0x46f/0x4c0 [ 75.342388] btrfs_kill_all_delayed_nodes+0x46f/0x4c0 [ 75.342389] ? _raw_spin_unlock+0x13/0x30 [ 75.342392] ? __pfx_btrfs_kill_all_delayed_nodes+0x10/0x10 [ 75.342393] ? do_raw_spin_lock+0x128/0x260 [ 75.342395] ? __pfx_do_raw_spin_lock+0x10/0x10 [ 75.342397] ? list_lru_add_obj+0xfb/0x1a0 [ 75.342399] ? do_raw_spin_lock+0x128/0x260 [ 75.342401] ? __pfx_do_raw_spin_lock+0x10/0x10 [ 75.342402] btrfs_clean_one_deleted_snapshot+0x143/0x370 [ 75.342405] cleaner_kthread+0x1ee/0x300 [ 75.342406] ? __pfx_cleaner_kthread+0x10/0x10 [ 75.342407] kthread+0x37f/0x6f0 [ 75.342409] ? __pfx_kthread+0x10/0x10 [ 75.342411] ? __pfx_kthread+0x10/0x10 [ 75.342412] ? __pfx_kthread+0x10/0x10 [ 75.342413] ret_from_fork+0x17d/0x240 [ 75.342415] ? __pfx_kthread+0x10/0x10 [ 75.342416] ret_from_fork_asm+0x1a/0x30 [ 75.342419] </TASK> [ 75.342419] [ 75.345517] Allocated by task 4527: [ 75.345517] kasan_save_stack+0x22/0x40 [ 75.345517] kasan_save_track+0x14/0x30 [ 75.345517] __kasan_slab_alloc+0x6e/0x70 [ 75.345517] kmem_cache_alloc_noprof+0x14c/0x400 [ 75.345517] btrfs_get_or_create_delayed_node+0x9e/0x9e0 [ 75.345517] btrfs_insert_delayed_dir_index+0xe4/0x8a0 [ 75.345517] btrfs_insert_dir_item+0x4c1/0x720 [ 75.345517] btrfs_add_link+0x173/0xa30 [ 75.345517] btrfs_create_new_inode+0x1551/0x2650 [ 75.345517] btrfs_create_common+0x17b/0x200 [ 75.345517] vfs_mknod+0x3a7/0x600 [ 75.345517] do_mknodat+0x34e/0x520 [ 75.345517] __x64_sys_mknodat+0xaa/0xe0 [ 75.345517] do_syscall_64+0x50/0xfa0 [ 75.345517] entry_SYSCALL_64_after_hwframe+0x76/0x7e [ 75.345517] [ 75.345517] Freed by task 4493: [ 75.345517] kasan_save_stack+0x22/0x40 [ 75.345517] kasan_save_track+0x14/0x30 [ 75.345517] __kasan_save_free_info+0x3b/0x70 [ 75.345517] __kasan_slab_free+0x43/0x70 [ 75.345517] kmem_cache_free+0x172/0x610 [ 75.345517] btrfs_kill_all_delayed_nodes+0x2db/0x4c0 [ 75.345517] btrfs_clean_one_deleted_snapshot+0x143/0x370 [ 75.345517] cleaner_kthread+0x1ee/0x300 [ 75.345517] kthread+0x37f/0x6f0 [ 75.345517] ret_from_fork+0x17d/0x240 [ 75.345517] ret_from_fork_asm+0x1a/0x30 [ 75.345517] [ 75.345517] The buggy address belongs to the object at ffff88812389f370 [ 75.345517] which belongs to the cache btrfs_delayed_node of size 440 [ 75.345517] The buggy address is located 16 bytes inside of [ 75.345517] freed 440-byte region [ffff88812389f370, ffff88812389f528) [ 75.345517] [ 75.345517] The buggy address belongs to the physical page: [ 75.345517] page: refcount:0 mapcount:0 mapping:0000000000000000 index:0x0 pfn:0x12389e [ 75.345517] head: order:1 mapcount:0 entire_mapcount:0 nr_pages_mapped:0 pincount:0 [ 75.345517] flags: 0x4000000000000040(head|zone=2) [ 75.345517] page_type: f5(slab) [ 75.345517] raw: 4000000000000040 ffff88810bcaadc0 ffffea0004487a10 ffff88810c6e6d80 [ 75.345517] raw: 0000000000000000 00000000000e000e 00000000f5000000 0000000000000000 [ 75.345517] head: 4000000000000040 ffff88810bcaadc0 ffffea0004487a10 ffff88810c6e6d80 [ 75.345517] head: 0000000000000000 00000000000e000e 00000000f5000000 0000000000000000 [ 75.345517] head: 4000000000000001 ffffea00048e2781 00000000ffffffff 00000000ffffffff [ 75.345517] head: ffffffffffffffff 0000000000000000 00000000ffffffff 0000000000000002 [ 75.345517] page dumped because: kasan: bad access detected [ 75.345517] [ 75.345517] Memory state around the buggy address: [ 75.345517] ffff88812389f280: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 [ 75.345517] ffff88812389f300: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fa fb [ 75.345517] >ffff88812389f380: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb [ 75.345517] ^ [ 75.345517] ffff88812389f400: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb [ 75.345517] ffff88812389f480: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb [ 75.345517] ================================================================== [ 75.501545] Disabling lock debugging due to kernel taint
gdb) l *(btrfs_kill_all_delayed_nodes+0x46f)0xffffffff82f2422f is in btrfs_kill_all_delayed_nodes (fs/btrfs/delayed-inode.h:219).
214 ref_tracker_dir_exit(&node->ref_dir.dir);
215 }
216
217 static inline void btrfs_delayed_node_ref_tracker_dir_print(struct btrfs_delayed_node *node)
218 {
219 if (!btrfs_test_opt(node->root->fs_info, REF_TRACKER))
220 return;
221
222 ref_tracker_dir_print(&node->ref_dir.dir,
223 BTRFS_DELAYED_NODE_REF_TRACKER_DISPLAY_LIMIT);
Quoted 96 lines · On Mon, 20 Oct 2025 07:19:39 -0700 Christoph Hellwig <[email protected]> wrote:
KASAN output:
[ 75.341543] ================================================================== [ 75.341824] BUG: KASAN: slab-use-after-free in btrfs_kill_all_delayed_nodes+0x46f/0x4c0 [ 75.342082] Read of size 8 at addr ffff88812389f380 by task btrfs-cleaner/4493 [ 75.342310] [ 75.342369] CPU: 1 UID: 0 PID: 4493 Comm: btrfs-cleaner Tainted: G N 6.18.0-rc2+ #4115 PREEMPT(f [ 75.342372] Tainted: [N]=TEST [ 75.342373] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.16.3-debian-1.16.3-2 04/01/2014 [ 75.342374] Call Trace: [ 75.342375] <TASK> [ 75.342376] dump_stack_lvl+0x4b/0x70 [ 75.342379] print_report+0x174/0x4e7 [ 75.342382] ? __virt_addr_valid+0x1bb/0x2f0 [ 75.342384] ? btrfs_kill_all_delayed_nodes+0x46f/0x4c0 [ 75.342385] kasan_report+0xd2/0x100 [ 75.342387] ? btrfs_kill_all_delayed_nodes+0x46f/0x4c0 [ 75.342388] btrfs_kill_all_delayed_nodes+0x46f/0x4c0 [ 75.342389] ? _raw_spin_unlock+0x13/0x30 [ 75.342392] ? __pfx_btrfs_kill_all_delayed_nodes+0x10/0x10 [ 75.342393] ? do_raw_spin_lock+0x128/0x260 [ 75.342395] ? __pfx_do_raw_spin_lock+0x10/0x10 [ 75.342397] ? list_lru_add_obj+0xfb/0x1a0 [ 75.342399] ? do_raw_spin_lock+0x128/0x260 [ 75.342401] ? __pfx_do_raw_spin_lock+0x10/0x10 [ 75.342402] btrfs_clean_one_deleted_snapshot+0x143/0x370 [ 75.342405] cleaner_kthread+0x1ee/0x300 [ 75.342406] ? __pfx_cleaner_kthread+0x10/0x10 [ 75.342407] kthread+0x37f/0x6f0 [ 75.342409] ? __pfx_kthread+0x10/0x10 [ 75.342411] ? __pfx_kthread+0x10/0x10 [ 75.342412] ? __pfx_kthread+0x10/0x10 [ 75.342413] ret_from_fork+0x17d/0x240 [ 75.342415] ? __pfx_kthread+0x10/0x10 [ 75.342416] ret_from_fork_asm+0x1a/0x30 [ 75.342419] </TASK> [ 75.342419] [ 75.345517] Allocated by task 4527: [ 75.345517] kasan_save_stack+0x22/0x40 [ 75.345517] kasan_save_track+0x14/0x30 [ 75.345517] __kasan_slab_alloc+0x6e/0x70 [ 75.345517] kmem_cache_alloc_noprof+0x14c/0x400 [ 75.345517] btrfs_get_or_create_delayed_node+0x9e/0x9e0 [ 75.345517] btrfs_insert_delayed_dir_index+0xe4/0x8a0 [ 75.345517] btrfs_insert_dir_item+0x4c1/0x720 [ 75.345517] btrfs_add_link+0x173/0xa30 [ 75.345517] btrfs_create_new_inode+0x1551/0x2650 [ 75.345517] btrfs_create_common+0x17b/0x200 [ 75.345517] vfs_mknod+0x3a7/0x600 [ 75.345517] do_mknodat+0x34e/0x520 [ 75.345517] __x64_sys_mknodat+0xaa/0xe0 [ 75.345517] do_syscall_64+0x50/0xfa0 [ 75.345517] entry_SYSCALL_64_after_hwframe+0x76/0x7e [ 75.345517] [ 75.345517] Freed by task 4493: [ 75.345517] kasan_save_stack+0x22/0x40 [ 75.345517] kasan_save_track+0x14/0x30 [ 75.345517] __kasan_save_free_info+0x3b/0x70 [ 75.345517] __kasan_slab_free+0x43/0x70 [ 75.345517] kmem_cache_free+0x172/0x610 [ 75.345517] btrfs_kill_all_delayed_nodes+0x2db/0x4c0 [ 75.345517] btrfs_clean_one_deleted_snapshot+0x143/0x370 [ 75.345517] cleaner_kthread+0x1ee/0x300 [ 75.345517] kthread+0x37f/0x6f0 [ 75.345517] ret_from_fork+0x17d/0x240 [ 75.345517] ret_from_fork_asm+0x1a/0x30 [ 75.345517] [ 75.345517] The buggy address belongs to the object at ffff88812389f370 [ 75.345517] which belongs to the cache btrfs_delayed_node of size 440 [ 75.345517] The buggy address is located 16 bytes inside of [ 75.345517] freed 440-byte region [ffff88812389f370, ffff88812389f528) [ 75.345517] [ 75.345517] The buggy address belongs to the physical page: [ 75.345517] page: refcount:0 mapcount:0 mapping:0000000000000000 index:0x0 pfn:0x12389e [ 75.345517] head: order:1 mapcount:0 entire_mapcount:0 nr_pages_mapped:0 pincount:0 [ 75.345517] flags: 0x4000000000000040(head|zone=2) [ 75.345517] page_type: f5(slab) [ 75.345517] raw: 4000000000000040 ffff88810bcaadc0 ffffea0004487a10 ffff88810c6e6d80 [ 75.345517] raw: 0000000000000000 00000000000e000e 00000000f5000000 0000000000000000 [ 75.345517] head: 4000000000000040 ffff88810bcaadc0 ffffea0004487a10 ffff88810c6e6d80 [ 75.345517] head: 0000000000000000 00000000000e000e 00000000f5000000 0000000000000000 [ 75.345517] head: 4000000000000001 ffffea00048e2781 00000000ffffffff 00000000ffffffff [ 75.345517] head: ffffffffffffffff 0000000000000000 00000000ffffffff 0000000000000002 [ 75.345517] page dumped because: kasan: bad access detected [ 75.345517] [ 75.345517] Memory state around the buggy address: [ 75.345517] ffff88812389f280: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 [ 75.345517] ffff88812389f300: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fa fb [ 75.345517] >ffff88812389f380: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb [ 75.345517] ^ [ 75.345517] ffff88812389f400: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb [ 75.345517] ffff88812389f480: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb [ 75.345517] ================================================================== [ 75.501545] Disabling lock debugging due to kernel taint
Quoted 12 lines
gdb) l *(btrfs_kill_all_delayed_nodes+0x46f)
0xffffffff82f2422fis inbtrfs_kill_all_delayed_nodes(fs/btrfs/delayed-inode.h:219).
214ref_tracker_dir_exit(&node->ref_dir.dir);
215 }
216
217 static inline voidbtrfs_delayed_node_ref_tracker_dir_print(struct btrfs_delayed_node *node)
218 {
219 if (!btrfs_test_opt(node->root->fs_info,REF_TRACKER))
220 return;
221
222 ref_tracker_dir_print(&node->ref_dir.dir,
223BTRFS_DELAYED_NODE_REF_TRACKER_DISPLAY_LIMIT);
This is a use after free bug with my ref_tracker patch, it's trying to print delayed_node ref_tracker stats after the delayed node has been freed. Will send a fix in a second.
Quoted 111 lines · On Mon, 20 Oct 2025 09:55:10 -0700 Leo Martins <[email protected]> wrote:
Quoted 96 lines · On Mon, 20 Oct 2025 07:19:39 -0700 Christoph Hellwig <[email protected]> wrote:
KASAN output:
[ 75.341543] ================================================================== [ 75.341824] BUG: KASAN: slab-use-after-free in btrfs_kill_all_delayed_nodes+0x46f/0x4c0 [ 75.342082] Read of size 8 at addr ffff88812389f380 by task btrfs-cleaner/4493 [ 75.342310] [ 75.342369] CPU: 1 UID: 0 PID: 4493 Comm: btrfs-cleaner Tainted: G N 6.18.0-rc2+ #4115 PREEMPT(f [ 75.342372] Tainted: [N]=TEST [ 75.342373] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.16.3-debian-1.16.3-2 04/01/2014 [ 75.342374] Call Trace: [ 75.342375] <TASK> [ 75.342376] dump_stack_lvl+0x4b/0x70 [ 75.342379] print_report+0x174/0x4e7 [ 75.342382] ? __virt_addr_valid+0x1bb/0x2f0 [ 75.342384] ? btrfs_kill_all_delayed_nodes+0x46f/0x4c0 [ 75.342385] kasan_report+0xd2/0x100 [ 75.342387] ? btrfs_kill_all_delayed_nodes+0x46f/0x4c0 [ 75.342388] btrfs_kill_all_delayed_nodes+0x46f/0x4c0 [ 75.342389] ? _raw_spin_unlock+0x13/0x30 [ 75.342392] ? __pfx_btrfs_kill_all_delayed_nodes+0x10/0x10 [ 75.342393] ? do_raw_spin_lock+0x128/0x260 [ 75.342395] ? __pfx_do_raw_spin_lock+0x10/0x10 [ 75.342397] ? list_lru_add_obj+0xfb/0x1a0 [ 75.342399] ? do_raw_spin_lock+0x128/0x260 [ 75.342401] ? __pfx_do_raw_spin_lock+0x10/0x10 [ 75.342402] btrfs_clean_one_deleted_snapshot+0x143/0x370 [ 75.342405] cleaner_kthread+0x1ee/0x300 [ 75.342406] ? __pfx_cleaner_kthread+0x10/0x10 [ 75.342407] kthread+0x37f/0x6f0 [ 75.342409] ? __pfx_kthread+0x10/0x10 [ 75.342411] ? __pfx_kthread+0x10/0x10 [ 75.342412] ? __pfx_kthread+0x10/0x10 [ 75.342413] ret_from_fork+0x17d/0x240 [ 75.342415] ? __pfx_kthread+0x10/0x10 [ 75.342416] ret_from_fork_asm+0x1a/0x30 [ 75.342419] </TASK> [ 75.342419] [ 75.345517] Allocated by task 4527: [ 75.345517] kasan_save_stack+0x22/0x40 [ 75.345517] kasan_save_track+0x14/0x30 [ 75.345517] __kasan_slab_alloc+0x6e/0x70 [ 75.345517] kmem_cache_alloc_noprof+0x14c/0x400 [ 75.345517] btrfs_get_or_create_delayed_node+0x9e/0x9e0 [ 75.345517] btrfs_insert_delayed_dir_index+0xe4/0x8a0 [ 75.345517] btrfs_insert_dir_item+0x4c1/0x720 [ 75.345517] btrfs_add_link+0x173/0xa30 [ 75.345517] btrfs_create_new_inode+0x1551/0x2650 [ 75.345517] btrfs_create_common+0x17b/0x200 [ 75.345517] vfs_mknod+0x3a7/0x600 [ 75.345517] do_mknodat+0x34e/0x520 [ 75.345517] __x64_sys_mknodat+0xaa/0xe0 [ 75.345517] do_syscall_64+0x50/0xfa0 [ 75.345517] entry_SYSCALL_64_after_hwframe+0x76/0x7e [ 75.345517] [ 75.345517] Freed by task 4493: [ 75.345517] kasan_save_stack+0x22/0x40 [ 75.345517] kasan_save_track+0x14/0x30 [ 75.345517] __kasan_save_free_info+0x3b/0x70 [ 75.345517] __kasan_slab_free+0x43/0x70 [ 75.345517] kmem_cache_free+0x172/0x610 [ 75.345517] btrfs_kill_all_delayed_nodes+0x2db/0x4c0 [ 75.345517] btrfs_clean_one_deleted_snapshot+0x143/0x370 [ 75.345517] cleaner_kthread+0x1ee/0x300 [ 75.345517] kthread+0x37f/0x6f0 [ 75.345517] ret_from_fork+0x17d/0x240 [ 75.345517] ret_from_fork_asm+0x1a/0x30 [ 75.345517] [ 75.345517] The buggy address belongs to the object at ffff88812389f370 [ 75.345517] which belongs to the cache btrfs_delayed_node of size 440 [ 75.345517] The buggy address is located 16 bytes inside of [ 75.345517] freed 440-byte region [ffff88812389f370, ffff88812389f528) [ 75.345517] [ 75.345517] The buggy address belongs to the physical page: [ 75.345517] page: refcount:0 mapcount:0 mapping:0000000000000000 index:0x0 pfn:0x12389e [ 75.345517] head: order:1 mapcount:0 entire_mapcount:0 nr_pages_mapped:0 pincount:0 [ 75.345517] flags: 0x4000000000000040(head|zone=2) [ 75.345517] page_type: f5(slab) [ 75.345517] raw: 4000000000000040 ffff88810bcaadc0 ffffea0004487a10 ffff88810c6e6d80 [ 75.345517] raw: 0000000000000000 00000000000e000e 00000000f5000000 0000000000000000 [ 75.345517] head: 4000000000000040 ffff88810bcaadc0 ffffea0004487a10 ffff88810c6e6d80 [ 75.345517] head: 0000000000000000 00000000000e000e 00000000f5000000 0000000000000000 [ 75.345517] head: 4000000000000001 ffffea00048e2781 00000000ffffffff 00000000ffffffff [ 75.345517] head: ffffffffffffffff 0000000000000000 00000000ffffffff 0000000000000002 [ 75.345517] page dumped because: kasan: bad access detected [ 75.345517] [ 75.345517] Memory state around the buggy address: [ 75.345517] ffff88812389f280: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 [ 75.345517] ffff88812389f300: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fa fb [ 75.345517] >ffff88812389f380: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb [ 75.345517] ^ [ 75.345517] ffff88812389f400: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb [ 75.345517] ffff88812389f480: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb [ 75.345517] ================================================================== [ 75.501545] Disabling lock debugging due to kernel taintQuoted 12 lines
gdb) l *(btrfs_kill_all_delayed_nodes+0x46f)
0xffffffff82f2422fis inbtrfs_kill_all_delayed_nodes(fs/btrfs/delayed-inode.h:219).
214ref_tracker_dir_exit(&node->ref_dir.dir);
215 }
216
217 static inline voidbtrfs_delayed_node_ref_tracker_dir_print(struct btrfs_delayed_node *node)
218 {
219 if (!btrfs_test_opt(node->root->fs_info,REF_TRACKER))
220 return;
221
222 ref_tracker_dir_print(&node->ref_dir.dir,
223BTRFS_DELAYED_NODE_REF_TRACKER_DISPLAY_LIMIT);
Quoted 2 lines
This is a use after free bug with my
ref_trackerpatch, it's trying to printdelayed_noderef_trackerstats after the delayed node has been freed. Will send a fix in a second.
I wasn't able to reproduce the crash by running btrfs/071. I sent out a fix, if you have time it would be great if you could check it against your reproducer.