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

From: Sandipan Das (AMD)
Date: Fri Jun 16 2023 - 10:43:39 EST


On 16/06/23 7:33 pm, Breno Leitao wrote:
> On Fri, Jun 16, 2023 at 03:29:54PM +0200, Peter Zijlstra wrote:
>> On Fri, Jun 16, 2023 at 04:53:15AM -0700, Breno Leitao wrote:
>>> On some systems, the Performance Counter Global Status Register is
>>> coming with reserved bits set, which causes the system to be unusable
>>> if a simple `perf top` runs. The system hits the WARN() thousands times
>>> while perf runs.
>>>
>>> WARNING: CPU: 18 PID: 20608 at arch/x86/events/amd/core.c:944 amd_pmu_v2_handle_irq+0x1be/0x2b0
>>>
>>> This happens because the "Performance Counter Global Status Register"
>>> (PerfCntGlobalStatus) MSR has bit 7 set. Bit 7 should be reserved according
>>> to the documentation (Figure 13-12 from "AMD64 Architecture Programmer’s
>>> Manual, Volume 2: System Programming, 24593"[1]
>>
>> Would it then not make more sense to mask out bit7 before:
>
> It is more than bit 7. This is the register structure according to the document
> above:
>
> Bits Mnemonic Description Access type
> 63:60 Reserved RO
> 59 PMCF Performance Counter Freeze RO
> 58 LBRSF Last Branch Record Stack Freeze RO
> 57:6 Reserved RO
> 5:0 CNT_OF Counter overflow for PerfCnt[5:0] RO
>
> In the code, bit GLOBAL_STATUS_LBRS_FROZEN is handled and cleared before
> we reach my changes
>
> That said, your approach is almost similar to what I did, and I will be happy
> to change in order to make the code clearer.
>
>> + status &= ~AMD_PMU_V2_GLOBAL_STATUS_RESERVED;
>> if (!status)
>> goto done;
>>
>> ?
>>
>> Aside from being reserved, why are these bits magically set all of a
>> sudden?
>
> That is probably a question to AMD.
>

The reserved bits should never be set. The purpose of the WARN_ON() is to catch such anomalies.
I am working out the details with Breno and will report back with a resolution.

- Sandipan

[sending from my personal email as I currently don't have access to my work laptop]