writeback: expired dirty inodes can lead to a NULL dereference kernel panic issue in 'move_expired_inodes' function
Hello, Colin Ian King.
I am Paran Lee.
While tracing the null dereference kernel panic issue during the stress-ng(stress-ng-proc) test, I found the inode code block that could cause a null dereference kernel panic.
If this issue is confirmed as a bug that may have occurred in the past, wouldn't it be added to the link below?
@ Bugs found with stress-ng - https://github.com/ColinIanKing/stress-ng#bugs-found-with-stress-ng
* kernel log
[21881.096120] ICMPv6: process `stress-ng-procf' is using deprecated sysctl (syscall) net.ipv6.neigh.default.base_reachable_time - use net.ipv6.neigh.default.base_reachable_time_ms instead [22009.051158] BUG: unable to handle kernel NULL pointer dereference at 0000000000000008 [22009.051167] IP: move_expired_inodes+0x59/0x1a0 [22009.051168] PGD 0 P4D 0 [22009.051170] Oops: 0002 [#1] SMP NOPTI ... [22009.051221] RIP: 0010:move_expired_inodes+0x59/0x1a0 ... [22009.051257] Call Trace: [22009.051260] queue_io+0x66/0x110 [22009.051262] wb_writeback+0x253/0x300 [22009.051264] wb_workfn+0xc0/0x400 [22009.051265] ? wb_workfn+0xc0/0x400 [22009.051268] ? __switch_to_asm+0x35/0x70 [22009.051272] process_one_work+0x1de/0x420 [22009.051274] worker_thread+0x32/0x410 [22009.051276] kthread+0x121/0x140 [22009.051277] ? process_one_work+0x420/0x420 [22009.051279] ? kthread_create_worker_on_cpu+0x70/0x70 [22009.051280] ret_from_fork+0x1f/0x40 [22009.051282] Code: 04 25 28 00 00 00 48 89 45 d0 31 c0 4c 89 75 c8 c7 45 b0 00 00 00 00 c7 45 b4 00 00 00 00 eb 76 48 8b 3b 48 8b 43 08 83 45 b0 01 <48> 89 47 08 48 89 38 48 8b 45 c0 48 89 58 08 48 89 03 48 8d 43 [22009.051293] RIP: move_expired_inodes+0x59/0x1a0 RSP: ffffa2f68413bcb0
* trace log on crash utility
KERNEL: vmlinux-4.15.0-166-generic DUMPFILE: 202204072123-wb_inode-delaying_queue-list-null-check/dump.202204072123 [PARTIAL DUMP] CPUS: 4 DATE: Fri Apr 8 06:23:30 2022 UPTIME: 01:35:27 LOAD AVERAGE: 8.30, 8.37, 9.17 TASKS: 668 NODENAME: ubuntu1804 RELEASE: 4.15.0-166-generic VERSION: #174-Ubuntu SMP Wed Dec 8 19:07:44 UTC 2021 MACHINE: x86_64 (2394 Mhz) MEMORY: 16 GB PANIC: "BUG: unable to handle kernel NULL pointer dereference at 0000000000000008" PID: 22864 COMMAND: "kworker/u8:0" TASK: ffff91490b55d880 [THREAD_INFO: ffff91490b55d880] CPU: 0 STATE: TASK_RUNNING (PANIC)
crash> bt ... #9 [ffffa2f68413bc00] page_fault at ffffffffae401615 [exception RIP: move_expired_inodes+89] RIP: ffffffffadcb11f9 RSP: ffffa2f68413bcb0 RFLAGS: 00010202 RAX: 0000000000000000 RBX: ffff9148083af9b8 RCX: ffff91496659a090 RDX: 000000010052d046 RSI: ffff91496659a080 RDI: 0000000000000000 RBP: ffffa2f68413bd08 R8: ffff91496676727f R9: ffff914951111f76 R10: ffffa2f68413bdf8 R11: 0000000000000334 R12: 0000000000000000 R13: ffff91496659a070 R14: ffffa2f68413bcc8 R15: ffff91496659a080 ORIG_RAX: ffffffffffffffff CS: 0010 SS: 0018 #10 [ffffa2f68413bd10] queue_io at ffffffffadcb32d6 #11 [ffffa2f68413bd48] wb_writeback at ffffffffadcb77a3 #12 [ffffa2f68413bde8] wb_workfn at ffffffffadcb7eb0 #13 [ffffa2f68413be80] process_one_work at ffffffffadaaa68e #14 [ffffa2f68413bec8] worker_thread at ffffffffadaaa902 #15 [ffffa2f68413bf08] kthread at ffffffffadab1361 #16 [ffffa2f68413bf50] ret_from_fork at ffffffffae4001ef
crash> dis ffffffffadcb32c6 20 ... 0xffffffffadcb32c6 <queue_io+86>: jg 0xffffffffadcb32e0 <queue_io+112> 0xffffffffadcb32c8 <queue_io+88>: mov %rsi,%r14 0xffffffffadcb32cb <queue_io+91>: mov %r12,%rsi 0xffffffffadcb32ce <queue_io+94>: mov %rdx,%r13 0xffffffffadcb32d1 <queue_io+97>: callq 0xffffffffadcb11a0 <move_expired_inodes>
crash> dis 0xffffffffadcb11a0 0xffffffffadcb11a0 <move_expired_inodes>: nopl 0x0(%rax,%rax,1) [FTRACE NOP] ... 0xffffffffadcb11ec <move_expired_inodes+76>: jmp 0xffffffffadcb1264 <move_expired_inodes+196> 0xffffffffadcb11ee <move_expired_inodes+78>: mov (%rbx),%rdi 0xffffffffadcb11f1 <move_expired_inodes+81>: mov 0x8(%rbx),%rax 0xffffffffadcb11f5 <move_expired_inodes+85>: addl $0x1,-0x50(%rbp) 0xffffffffadcb11f9 <move_expired_inodes+89>: mov %rax,0x8(%rdi) <<<<<< Don't you think this is the NULL dereference RIP point of it? 0xffffffffadcb11fd <move_expired_inodes+93>: mov %rdi,(%rax) ... 0xffffffffadcb121e <move_expired_inodes+126>: callq 0xffffffffae3cd460 <_raw_spin_lock>
Have a good day. Paran Lee.
Signed-off-by: Paran Lee p4ranlee@gmail.com --- fs/fs-writeback.c | 11 +++++++++-- 1 file changed, 9 insertions(+), 2 deletions(-)
diff --git a/fs/fs-writeback.c b/fs/fs-writeback.c index 591fe9cf1659..23a7a567e443 100644 --- a/fs/fs-writeback.c +++ b/fs/fs-writeback.c @@ -1357,12 +1357,14 @@ static int move_expired_inodes(struct list_head *delaying_queue, LIST_HEAD(tmp); struct list_head *pos, *node; struct super_block *sb = NULL; - struct inode *inode; + struct inode *inode = NULL; int do_sb_sort = 0; int moved = 0;
while (!list_empty(delaying_queue)) { inode = wb_inode(delaying_queue->prev); + if(!inode) + continue; if (inode_dirtied_after(inode, dirtied_before)) break; list_move(&inode->i_io_list, &tmp); @@ -1385,7 +1387,12 @@ static int move_expired_inodes(struct list_head *delaying_queue,
/* Move inodes from one superblock together */ while (!list_empty(&tmp)) { - sb = wb_inode(tmp.prev)->i_sb; + inode = wb_inode(tmp.prev); + if(!inode) + continue; + sb = inode->i_sb; + if(!sb) + continue; list_for_each_prev_safe(pos, node, &tmp) { inode = wb_inode(pos); if (inode->i_sb == sb)