[LU-18845] NULL pointer exception with_imp_locked_nested

• Type: Bug
• Resolution: Fixed
• Priority: Minor
• Fix Version/s: Lustre 2.17.0
• Affects Version/s: Upstream

Epic/Theme

lproc

• Severity: 3
• Assignee: Vandana Rungta
• Reporter: Vandana Rungta

Description

Unable to handle kernel NULL pointer dereference at virtual address 0000000000000588

Call stack:

 324.423743] Unable to handle kernel NULL pointer dereference at virtual address 0000000000000588 
...

{{ Call trace:
[  324.459145]  down_read+0x28/0x100
[  324.459684]  active_show+0x2c/0x90 [osp]
[  324.460339]  lustre_attr_show+0x1c/0x2c [obdclass]
[  324.461095]  sysfs_kf_seq_show+0xb0/0x140
[  324.461729]  kernfs_seq_show+0x30/0x40
[  324.462377]  seq_read_iter+0x1cc/0x4c8
[  324.462973]  kernfs_fop_read_iter+0x38/0x40
[  324.463635]  new_sync_read+0xf0/0x190
[  324.464219]  vfs_read+0x150/0x1e8
[  324.464748]  ksys_read+0x70/0xfc
[  324.465265]  __arm64_sys_read+0x20/0x28
[  324.465880]  el0_svc_common.constprop.0+0x84/0x240
[  324.466635]  do_el0_svc+0x28/0x80
[  324.467165]  el0_svc+0x1c/0x28
[  324.467657]  el0_sync_handler+0x94/0x120}}

Observed on Lustre 2_15.
Analysis:

active_show calls macro with_imp_locked with an obd_device, which in
turn gets a lock on u.cli.cl_sem in obd_device.
The exception is at offset 0x588, and u.cli.cl_sem is at 0x588
in struct obd_device - as a result of obd being NULL.