Re: [PATCH] refscale: Fix use of uninitalized wait_queue_head_t

[Waiman Long <longman@redhat.com>]

On 7/7/23 12:16, Paul E. McKenney wrote:
> On Fri, Jul 07, 2023 at 10:56:51AM -0400, Waiman Long wrote:
>> On 7/7/23 10:07, Davidlohr Bueso wrote:
>>> On Thu, 06 Jul 2023, Waiman Long wrote:
>>>
>>>> It was found that running the refscale test might sometimes crash the
>>>> kernel with the following error:
>>>>
>>>> [ 8569.952896] BUG: unable to handle page fault for address:
>>>> ffffffffffffffe8
>>>> [ 8569.952900] #PF: supervisor read access in kernel mode
>>>> [ 8569.952902] #PF: error_code(0x0000) - not-present page
>>>> [ 8569.952904] PGD c4b048067 P4D c4b049067 PUD c4b04b067 PMD 0
>>>> [ 8569.952910] Oops: 0000 [#1] PREEMPT_RT SMP NOPTI
>>>> [ 8569.952916] Hardware name: Dell Inc. PowerEdge R750/0WMWCR, BIOS
>>>> 1.2.4 05/28/2021
>>>> [ 8569.952917] RIP: 0010:prepare_to_wait_event+0x101/0x190
>>>>   :
>>>> [ 8569.952940] Call Trace:
>>>> [ 8569.952941]  <TASK>
>>>> [ 8569.952944]  ref_scale_reader+0x380/0x4a0 [refscale]
>>>> [ 8569.952959]  kthread+0x10e/0x130
>>>> [ 8569.952966]  ret_from_fork+0x1f/0x30
>>>> [ 8569.952973]  </TASK>
>>>>
>>>> This is likely caused by the fact that init_waitqueue_head() is called
>>>> after the ref_scale_reader kthread is created. So the kthread may try
>>>> to use the waitqueue head before it is properly initialized. Fix this
>>>> by initializing the waitqueue head first before kthread creation.
>>>>
>>>> Fixes: 653ed64b01dc ("refperf: Add a test to measure performance of
>>>> read-side synchronization")
>>>> Signed-off-by: Waiman Long <longman@redhat.com>
>>> Strange this wasn't reported sooner.
>> Red Hat does have a pretty large QE organization that run all sort of tests
>> include this one pretty frequently. The race window is pretty small, but
>> they did hit this once in a while.
> I do run this fairly frequently, but haven't managed to hit it.
>
> Good show on making it happen, and looking forward to the updated patch!

v2 sent with a bit more explanation text in the commit log.

Cheers,
Longman