[LU-11658] The cl_cache may be uninitialized while osc is activating

• Type: Bug
• Resolution: Fixed
• Priority: Minor
• Fix Version/s: Lustre 2.13.0, Lustre 2.10.7, Lustre 2.12.1
• Severity: 3
• Assignee: Yang Sheng
• Reporter: Yang Sheng

Description

The obd_set_info_async will be invoked with KEY_CACHE_SET to initialize cl_cache while normal osc setup. We don't call it when osc has been disabled since export has not been setup. But it still be missed while osc active and causes the client to crash.

[3968842.116112] Lustre: MGC10.53.18.101@tcp: Connection restored to MGS (at 10.53.18.101@tcp)
[4195817.824439] BUG: unable to handle kernel NULL pointer dereference at           (null)
[4195817.824813] IP: [<ffffffffa0b934a5>] osc_lru_reserve+0x25/0x130 [osc]
[4195817.825016] PGD ec67e1067 PUD 112e776067 PMD 0
[4195817.825209] Oops: 0000 [#1] SMP

.....

[4195817.833867] Call Trace:
[4195817.834063]  [<ffffffffa0b98b36>] osc_io_write_iter_init+0x86/0x1e0 [osc]
[4195817.834280]  [<ffffffffa077448c>] cl_io_iter_init+0x5c/0x120 [obdclass]
[4195817.834474]  [<ffffffffa063037c>] lov_io_iter_init+0x1fc/0x5a0 [lov]
[4195817.834670]  [<ffffffffa0630898>] lov_io_rw_iter_init+0x178/0x2c0 [lov]
[4195817.834873]  [<ffffffffa077448c>] cl_io_iter_init+0x5c/0x120 [obdclass]
[4195817.835080]  [<ffffffffa0777b72>] cl_io_loop+0x42/0x190 [obdclass]
[4195817.835281]  [<ffffffffa0a9fc91>] ll_file_io_generic+0x601/0xad0 [lustre]
[4195817.835479]  [<ffffffffa0aa050d>] ll_file_aio_write+0x20d/0x870 [lustre]
[4195817.835679]  [<ffffffffa0aa0c3b>] ll_file_write+0xcb/0x1e0 [lustre]
[4195817.835873]  [<ffffffff811fe12d>] vfs_write+0xbd/0x1e0
[4195817.836063]  [<ffffffff811fee02>] SyS_pwrite64+0x92/0xc0
[4195817.836253]  [<ffffffff816965c9>] system_call_fastpath+0x16/0x1b
[4195817.836448] Code: 66 0f 1f 44 00 00 0f 1f 44 00 00 55 48 89 e5 41 54 49 89 fc 53 44 8b 87 4c 01 00 00 44 0f af 87 48 01 00 00 48 8b 97 98 07 00 00 <48> 8b 0a 4c 39 c6 49 0f 47 f0 48 39 f1 48 89 f3 72 40 48 89 ce
[4195817.837229] RIP  [<ffffffffa0b934a5>] osc_lru_reserve+0x25/0x130 [osc]
[4195817.837434]  RSP <ffff881190fffbb8>
[4195817.837612] CR2: 0000000000000000

Issue Links

• is related to: LU-17069 Crash when writing to a deactivated OSC