Re: PTP vclock: BUG: scheduling while atomic

On Fri, Feb 3, 2023 at 1:10 AM Richard Cochran [off-list ref] wrote:

On Thu, Feb 02, 2023 at 05:33:15PM +0100, Miroslav Lichvar wrote:

quoted

On Thu, Feb 02, 2023 at 05:02:07PM +0100, Íñigo Huguet wrote:

quoted

Our QA team was testing PTP vclocks, and they've found this error with sfc NIC/driver:
  BUG: scheduling while atomic: ptp5/25223/0x00000002

The reason seems to be that vclocks disable interrupts with `spin_lock_irqsave` in
`ptp_vclock_gettime`, and then read the timecounter, which in turns ends calling to
the driver's `gettime64` callback.

The same issue was observed with the ice driver:
https://lists.osuosl.org/pipermail/intel-wired-lan/Week-of-Mon-20221107/030633.html

I tried to fix it generally in the vclock support, but was not
successful. There was a hint it would be fixed in the driver. I'm not
sure what is the best approach here.

Can ptp_vclock_gettime use a mutex instead?

I don't see any place where these vclock functions are called in
atomic context, so it might be possible, but there are many callback
indirections and I'm not sure if I might have missed any.

Thanks,
Richard

-- 
Íñigo Huguet