linux-kernel - Re: [RFC PATCH 0/2] Introduce per-task io utilization boost

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: <de1eba3c-2453-4c5c-bd80-dd7d7b33f60d@arm.com>
Date: Mon, 25 Mar 2024 12:06:16 +0000
From: Christian Loehle <christian.loehle@....com>
To: Bart Van Assche <bvanassche@....org>, Qais Yousef <qyousef@...alina.io>
Cc: linux-kernel@...r.kernel.org, peterz@...radead.org,
 juri.lelli@...hat.com, mingo@...hat.com, rafael@...nel.org,
 dietmar.eggemann@....com, vschneid@...hat.com, vincent.guittot@...aro.org,
 Johannes.Thumshirn@....com, adrian.hunter@...el.com, ulf.hansson@...aro.org,
 andres@...razel.de, asml.silence@...il.com, linux-pm@...r.kernel.org,
 linux-block@...r.kernel.org, io-uring@...r.kernel.org,
 linux-mmc@...r.kernel.org
Subject: Re: [RFC PATCH 0/2] Introduce per-task io utilization boost

On 21/03/2024 19:52, Bart Van Assche wrote:
> On 3/21/24 10:57, Christian Loehle wrote:
>> In the long-term it looks like for UFS the problem will disappear as we are
>> expected to get one queue/hardirq per CPU (as Bart mentioned), on NVMe that
>> is already the case.
> 
> Why the focus on storage controllers with a single completion interrupt?
> It probably won't take long (one year?) until all new high-end
> smartphones may have support for multiple completion interrupts.
> 
> Thanks,
> 
> Bart.
> 

Apart from going to "This patch shows significant performance improvements on
hardware that runs mainline today" to "This patch will have significant
performance improvements on devices running mainline in a couple years"
nothing in particular.
I'm fine with leaving it with having acknowledged the problem.
Maybe I would just gate the task placement on the task having been in
UFS (with multiple completion interrupts) or NVMe submission recently to
avoid regressions to current behavior in future versions. I did have that
already at some point, although it was a bit hacky.
Anyway, thank you for your input on that, it is what I wanted to hear!

Kind Regards,
Christian