linux-kernel - Re: [PATCH] x86/split_lock: add split lock counter

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]

Date: Tue, 2 Jan 2024 07:14:10 -0800
From: Dave Hansen <dave.hansen@...el.com>
To: Maksim Davydov <davydov-max@...dex-team.ru>
Cc: den-plotnikov@...dex-team.ru, tony.luck@...el.com,
 linux-kernel@...r.kernel.org, tglx@...utronix.de, mingo@...hat.com,
 bp@...en8.de, dave.hansen@...ux.intel.com, hpa@...or.com,
 peterz@...radead.org, juri.lelli@...hat.com, vincent.guittot@...aro.org,
 x86@...nel.org
Subject: Re: [PATCH] x86/split_lock: add split lock counter

On 12/16/23 02:56, Maksim Davydov wrote:
> On 12/15/23 21:16, Dave Hansen wrote:
>> On 12/15/23 06:01, Maksim Davydov wrote:
>>> Provides per task split locks counter to monitor split locks rate
>>> in the system. It can be helpful in split locks monitoring to get a
>>> clear
>>> sense of which process causing split locks and how many of them have
>>> happened by the moment. For instance, it might be used by cloud
>>> providers
>>> who can't control guest executable code and want to make decisions based
>>> on the rate value like ratelimiting or notifing the split lock origins.
>> Have you considered doing this with tracing instead?
>>
>> It seems a _little_ silly for everyone to pay the cost of having that
>> counter around.
> No, it just seemed like a good idea to make a simple machine-readable
> interface based on the existing flag (reported_split_lock).
> What if this interface is made together with the appropriate CONFIG flag
> (default false)? Thus it won't affect those who don't need it.

Distributions (effectively) run a single kernel config.  They turn on
almost everything.  Using a config option would subject all distro users
to the overhead.