【22.03 SP1】【nvme】需合入上游6.1补丁nvme: fix multipath crash caused by flush request when blktrace is enabled，避免crash

产品测试环境利用mkfs.xfs初始化远端nvme盘时触发crash，panic栈如下：

```
[  682.498381] Buffer I/O error on dev nvme0n2, logical block 51167, async page read
[  706.716776] BUG: kernel NULL pointer dereference, address: 000000000000001a
[  706.718079] #PF: supervisor read access in kernel mode
[  706.719379] #PF: error_code(0x0000) - not-present page
[  706.720649] PGD 21070ca067 P4D 21070b9067 PUD 2118627067 PMD 0
[  706.721921] Oops: 0000 [#1] SMP NOPTI
[  706.723160] CPU: 45 PID: 12817 Comm: dpe_io_3_1 Kdump: loaded Tainted: G S                5.10.0-136.12.0.86.4.nos1.x86_64 #6
[  706.725604] Hardware name: New H3C Technologies Co., Ltd. H3C UniServer R4900 G5/RS35M2C16SB, BIOS 5.57 02/17/2023
[  706.726852] RIP: 0010:blk_add_trace_bio_complete+0x6/0x30
[  706.728078] Code: 31 c0 31 d2 6a 00 31 f6 6a 00 6a 00 e8 03 d9 ff ff 48 83 c4 20 e9 aa cf f5 ff 66 2e 0f 1f 84 00 00 00 00 00 41 
f 00 48 89 ee 4c 89 e7 5d 89 c1 ba
[  706.730599] RSP: 0000:ff856c6d4d6a0e90 EFLAGS: 00010282
[  706.731863] RAX: ff4dcda862ccd3d0 RBX: 0000000002000000 RCX: 0000000000000001
[  706.733140] RDX: 0000000000000000 RSI: ff4dcd7cc276d2f8 RDI: 0000000000000000
[  706.734418] RBP: ff4dcd7c5f50f440 R08: 0000000000000000 R09: 0000000299bce580
[  706.735707] R10: 0000000000000000 R11: 0000000000000010 R12: ff4dcd7cc276d2f8
[  706.736996] R13: ff4dcd7cc2769270 R14: ff4dcd7c0dc15490 R15: ff4dcda77670b800
[  706.738288] FS:  00007f5beeb5c640(0000) GS:ff4dcd9b00080000(0000) knlGS:0000000000000000
[  706.739586] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[  706.740881] CR2: 000000000000001a CR3: 000000013060c004 CR4: 0000000000771ee0
[  706.742188] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[  706.743487] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
[  706.744750] PKRU: 55555554
[  706.745973] Call Trace:
[  706.747187]  <IRQ>
[  706.748394]  nvme_complete_rq+0x109/0x1f0 [nvme_core]
[  706.749581]  nvme_rdma_recv_done+0xa6/0xf0 [nvme_rdma]
[  706.750755]  __ib_process_cq+0x6d/0x100 [ib_core]
[  706.751927]  ib_poll_handler+0x2b/0xb0 [ib_core]
[  706.753080]  irq_poll_softirq+0x7c/0x100
[  706.754216]  __do_softirq+0xc1/0x280
[  706.755324]  asm_call_irq_on_stack+0xf/0x20
[  706.756396]  </IRQ>
[  706.757432]  do_softirq_own_stack+0x37/0x50
[  706.758448]  irq_exit_rcu+0xbe/0x100
[  706.759442]  common_interrupt+0x74/0x130
[  706.760413]  ? asm_common_interrupt+0x8/0x40
[  706.761387]  asm_common_interrupt+0x1e/0x40
[  706.762350] RIP: 0033:0x7f6008fd2ce6
[  706.763298] Code: 7d 02 00 00 49 8d 7c 24 48 49 8b 6c 24 18 48 89 f8 48 c1 e8 03 80 b8 00 80 ff 7f 00 0f 85 64 02 00 00 4d 8b 64 
5 01 0f 84 b8 00 00 00 48 89 ef 48
```

crash分析：

```
crash> request.rq_disk ff4dcd7c5f50f440
rq_disk = 0xff4dcd7c1292c000

crash> gendisk.disk_name 0xff4dcd7c1292c000
 disk_name = "nvme0c0n1\000\000\000\000\000\000\000\000p\222\347\300\204\372\002\001\000\000\000\000\000\000"

crash> request.bio ff4dcd7c5f50f440
  bio = 0x0

crash> request.end_io ff4dcd7c5f50f440
end_io = 0xffffffff8470fb50

crash> sym 0xffffffff8470fb50
ffffffff8470fb50 (t) flush_end_io /usr/src/debug/kernel-5.10.0-136.12.0.86.4.nos1.x86_64/linux-5.10.0-136.12.0.86.4.nos1.x86_64/block/blk-flush.c: 215 
```

经过定位需合入上游补丁3659fb5ac29a5e6102bebe494ac789fd47fb78f4 解决
nvme: fix multipath crash caused by flush request when blktrace is enabled

由blk_kick_flush初始化的flush请求具有NULL的bio。在io完成期间，可能会被nvme_end_req处理。当启用blktrace时，激活多路径的nvme_trace_bio_complete尝试从flush请求访问NULL指针bio数据，导致系统crash。

经过确认OLK-5.10和2203SP2分支已经合入，顾需要同步到2203SP1分支

GVP openEuler/kernel
关闭

内容风险标识

评论 (1)

GVPopenEuler/kernel关闭

内容风险标识