Re: [PATCH] perf/x86/amd: Do not WARN on every IRQ

[Peter Zijlstra]

On Thu, Sep 14, 2023 at 04:52:13PM +0530, Sandipan Das wrote:
> On 9/14/2023 4:48 PM, Peter Zijlstra wrote:
> > On Thu, Sep 14, 2023 at 02:30:43AM -0700, Breno Leitao wrote:
> >> On Thu, Sep 14, 2023 at 11:12:34AM +0200, Peter Zijlstra wrote:
> >>> On Thu, Sep 14, 2023 at 02:25:40PM +0530, Sandipan Das wrote:
> >>
> >>>> I agree with using WARN_ON_ONCE() to make this less intrusive.
> >>>
> >>> Could you send a patch that AMD is happy with?
> >>
> >> Why the current patch is not good enough?
> > 
> > Sandipan, can you answer this? I don't tihnk I'm qualified to speak for
> > the AMD pmu and certainly I don't have insight into their design future.
> 
> Hi Breno,
> 
> Functionally, the patch looks good to me and I will be reusing it
> without any change to the authorship. However, as Peter suggested, I
> wanted to add a message to prompt users to update the microcode and
> also call out the required patch levels in the commit message since
> different Zen 4 variants and steppings use different microcode.
> 
> Here's what I plan to send.
> 
> diff --git a/arch/x86/events/amd/core.c b/arch/x86/events/amd/core.c
> index abadd5f23425..186a124bb3c0 100644
> --- a/arch/x86/events/amd/core.c
> +++ b/arch/x86/events/amd/core.c
> @@ -909,6 +909,13 @@ static int amd_pmu_v2_handle_irq(struct pt_regs *regs)
>                 status &= ~GLOBAL_STATUS_LBRS_FROZEN;
>         }
> 
> +       if (status & ~amd_pmu_global_cntr_mask)
> +               pr_warn_once("Unknown status bits are set (0x%llx), please consider updating microcode\n",
> +                            status);
> +
> +       /* Clear any reserved bits set by buggy microcode */
> +       status &= amd_pmu_global_cntr_mask;
> +
>         for (idx = 0; idx < x86_pmu.num_counters; idx++) {
>                 if (!test_bit(idx, cpuc->active_mask))
>                         continue;
> 
> --
> 
> Hi Peter,
> 
> There is another case where users will see warnings but the patch
> to fix it (link below) is yet to be reviewed. May I rebase and
> resend it along with the above?
> 
> https://lore.kernel.org/all/20230613105809.524535-1-sandipan.das@xxxxxxx/
> 

Sure, sorry I seem to have missed that :-(