linux-kernel - Re: [PATCH v2 2/5] perf: arm_cspmu: Support shared interrupts

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <4bc59e32-a3cc-4e8c-0264-e509aadeb534@arm.com>
Date:   Fri, 2 Jun 2023 12:25:53 +0100
From:   Robin Murphy <robin.murphy@....com>
To:     Ilkka Koskinen <ilkka@...amperecomputing.com>
Cc:     Jonathan Corbet <corbet@....net>, Will Deacon <will@...nel.org>,
        Mark Rutland <mark.rutland@....com>,
        Besar Wicaksono <bwicaksono@...dia.com>,
        Suzuki K Poulose <suzuki.poulose@....com>,
        linux-doc@...r.kernel.org, linux-kernel@...r.kernel.org,
        linux-arm-kernel@...ts.infradead.org
Subject: Re: [PATCH v2 2/5] perf: arm_cspmu: Support shared interrupts

On 2023-06-02 08:04, Ilkka Koskinen wrote:
> 
> Hi Robin,
> 
> On Thu, 1 Jun 2023, Robin Murphy wrote:
>> On 2023-06-01 04:01, Ilkka Koskinen wrote:
>>> Some of the PMUs may share the interrupt. Support them by
>>> setting IRQF_SHARED
>>
>> This has the usual problem of allowing any PMU instance to move the 
>> IRQ affinity to a different CPU without also migrating all the other 
>> PMU contexts, and thus breaking perf core's assumptions of mutual 
>> exclusion.
> 
> I see, I wasn't aware of such an assumption. Sounds like there isn't 
> necessarily an easy and clean solution for the shared interrupt case. I 
> drop the patch and get back on the issue if we come up with something 
> reasonable later.

What comes to mind is factoring out the explicit interrupt-sharing 
machinery that we wrote to solve this problem in arm_dmc620_pmu, or 
possibly trying to do something with IRQ affinity notifiers (however, I 
recall looking into that a while ago and it didn't seem like they 
actually interact with CPU hotplug in the way we'd want).

Thanks,
Robin.