thermal throttling on xps13: unchecked MSR access error

thermal throttling on xps13: unchecked MSR access error

[Linus Torvalds]

Hmm.

I don't think I've seen this before on my trusty old x86 laptop (XPS
13 9380 - it's a few years old)

    unchecked MSR access error: WRMSR to 0x1b1
      (tried to write 0x0000000004000aa8)
      at rIP: 0xffffffff8b8559fe (throttle_active_work+0xbe/0x1b0)

I'm blaming one of

  930d06bf071a ("thermal: intel: Protect clearing of thermal status bits")
  6fe1e64b6026 ("thermal: intel: Prevent accidental clearing of HFI status")

with no real reason apart from being the last commit to touch that
function, but also when it started happening.

The first kernel I see this for is 6.1.0-03225-g764822972d64, but
honestly, it's possible that it has happened before too, and the real
issue is that the machine just happened to be hot and throttling at
bootup and/or I just didn't notice.

The CPU in this thing is a

  Intel(R) Core(TM) i7-8565U CPU @ 1.80GHz

which hopefully makes somebody go "Ahh, yes, I missed that case".

I don't *think* the MSR access checking has changed, but maybe it did,
and I'm barking up the wrong tree.

Anybody?

                 Linus

Re: thermal throttling on xps13: unchecked MSR access error

[srinivas pandruvada]

Hi Linus,

Sorry about the issue.

On Tue, 2022-12-13 at 16:35 -0800, Linus Torvalds wrote:
> Hmm.
> 
> I don't think I've seen this before on my trusty old x86 laptop (XPS
> 13 9380 - it's a few years old)
> 
>     unchecked MSR access error: WRMSR to 0x1b1
>       (tried to write 0x0000000004000aa8)
>       at rIP: 0xffffffff8b8559fe (throttle_active_work+0xbe/0x1b0)
> 
You got a (PROCHOT#) throttling event.
 
> I'm blaming one of
> 
>   930d06bf071a ("thermal: intel: Protect clearing of thermal status
> bits")
>   6fe1e64b6026 ("thermal: intel: Prevent accidental clearing of HFI
> status")
> 
This is to blame. I am able to reproduce on an old system.

I sent a patch " thermal: intel: Don't set HFI status bit to 1"

Please check.

Thanks,
Srinivas

> with no real reason apart from being the last commit to touch that
> function, but also when it started happening.
> 
> The first kernel I see this for is 6.1.0-03225-g764822972d64, but
> honestly, it's possible that it has happened before too, and the real
> issue is that the machine just happened to be hot and throttling at
> bootup and/or I just didn't notice.
> 
> The CPU in this thing is a
> 
>   Intel(R) Core(TM) i7-8565U CPU @ 1.80GHz
> 
> which hopefully makes somebody go "Ahh, yes, I missed that case".
> 
> I don't *think* the MSR access checking has changed, but maybe it
> did,
> and I'm barking up the wrong tree.
> 
> Anybody?
> 
>                  Linus

Re: thermal throttling on xps13: unchecked MSR access error

[Linus Torvalds]

On Tue, Dec 13, 2022 at 6:12 PM srinivas pandruvada
<srinivas.pandruvada@linux.intel.com> wrote:
>
> I sent a patch " thermal: intel: Don't set HFI status bit to 1"
>
> Please check.

Yup, just replied to it.

Again - I can't guarantee that I actually triggered the situation that
caused the problem originally, but it does look fine now.

               Linus