Re: [PATCH] PCI: vmd: Fix spinlock usage on config access for RT kernel

[Ryo Takakura <ryotkkr98@gmail.com>]

Hi Luis,

On Mon, 16 Dec 2024 11:33:56 -0300, Luis Claudio R. Goncalves wrote:
>On Sun, Dec 15, 2024 at 11:13:21PM +0900, Ryo Takakura wrote:
>> PCI config access is locked with pci_lock which is raw_spinlock.
>> Convert cfg_lock to raw_spinlock so that the lock usage is consistent
>> for RT kernel.
>> 
>> Reported as following:
>> [   18.756807] BUG: sleeping function called from invalid context at kernel/locking/spinlock_rt.c:48
>> [   18.756810] in_atomic(): 1, irqs_disabled(): 1, non_block: 0, pid: 1617, name: nodedev-init
>> [   18.756810] preempt_count: 1, expected: 0
>> [   18.756811] RCU nest depth: 0, expected: 0
>> [   18.756811] INFO: lockdep is turned off.
>> [   18.756812] irq event stamp: 0
>
>This problem has been discussed in the past at:
>
>    [PATCH v6.4-rc5-rt4] rt: vmd: make cfg_lock a raw spinlock
>    https://lore.kernel.org/linux-rt-users/20230608210232.056e731f@gandalf.local.home/T/#t
>
>And there was a suggestion to remove the lock altogether as (at the time) the
>only two call sites of vmd_pci_read() were already protected by pci_lock.
>
>I wrote the patch removing the lock and it worked as expected, but never
>sent it to the list. If you think that for the current code that makes
>sense, please have fun removing the lock. :)

I see, thanks for the link!
I took a look at current the code and I think that cfg_lock can still
safely be removed as explained by Sebastian [0].
I'll send another patch shortly!

>Best regards,
>Luis

Sincerely,
Ryo Takakura

[0] https://lore.kernel.org/linux-rt-users/20230620154434.MosrzRUh@linutronix.de/