linux-kernel - Re: [RFC PATCH] perf/x86/intel: Add a check for dynamic constraints

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] [day] [month] [year] [list]

Message-ID: <20250513151632.GB25639@noisy.programming.kicks-ass.net>
Date: Tue, 13 May 2025 17:16:32 +0200
From: Peter Zijlstra <peterz@...radead.org>
To: kan.liang@...ux.intel.com
Cc: mingo@...hat.com, acme@...nel.org, namhyung@...nel.org,
	irogers@...gle.com, linux-kernel@...r.kernel.org
Subject: Re: [RFC PATCH] perf/x86/intel: Add a check for dynamic constraints

On Mon, May 12, 2025 at 10:55:42AM -0700, kan.liang@...ux.intel.com wrote:
> From: Kan Liang <kan.liang@...ux.intel.com>
> 
> The current event scheduler has a limit. If the counter constraint of an
> event is not a subset of any other counter constraint with an equal or
> higher weight. The counters may not be fully utilized.
> 
> To workaround it, the commit bc1738f6ee83 ("perf, x86: Fix event
> scheduler for constraints with overlapping counters") introduced an
> overlap flag, which is hardcoded to the event constraint that may
> trigger the limit. It only works for static constraints.
> 
> Many features on and after Intel PMON v6 require dynamic constraints. An
> event constraint is decided by both static and dynamic constraints at
> runtime. See commit 4dfe3232cc04 ("perf/x86: Add dynamic constraint").
> The dynamic constraints are from CPUID enumeration. It's impossible to
> hardcode it in advance. It's not practical to set the overlap flag to all
> events. It's harmful to the scheduler.
> 
> For the existing Intel platforms, the dynamic constraints don't trigger
> the limit. A real fix is not required.
> 
> However, for virtualization, VMM may give a weird CPUID enumeration to a
> guest. It's impossible to indicate what the weird enumeration is. A
> check is introduced, which can list the possible breaks if a weird
> enumeration is used.
> 
> Check the dynamic constraints enumerated for normal, branch counters
> logging, and auto-counter reload.
> Check both PEBS and non-PEBS constratins.
> 
> Closes: https://lore.kernel.org/lkml/20250416195610.GC38216@noisy.programming.kicks-ass.net/
> Signed-off-by: Kan Liang <kan.liang@...ux.intel.com>

Yeah, that looks about right. Thanks!