From: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
To: Julius Bairaktaris <julius@bairaktaris.de>,
Jeff Johnson <jjohnson@kernel.org>
Cc: ath10k@lists.infradead.org, linux-wireless@vger.kernel.org,
linux-kernel@vger.kernel.org,
Baochen Qiang <baochen.qiang@oss.qualcomm.com>
Subject: Re: [PATCH ath-next] wifi: ath10k: spread the PCI copy engine interrupts over the CPUs
Date: Wed, 18 Nov 2026 11:57:53 -0800 [thread overview]
Message-ID: <9b13a6c0-1ff9-4c63-b1fc-9ab6def252bd@oss.qualcomm.com> (raw)
In-Reply-To: <20261101164247.258200-1-julius@bairaktaris.de>
On 11/1/2026 8:42 AM, Julius Bairaktaris wrote:
> The PCI ath10k devices are capable of multiple MSI. The QCA9889 of a
> Xiaomi AX3600 advertises 8 vectors in its MSI capability (Multiple
> Message Capable 3) and, when granted all of them, raises the firmware
> interrupt and each copy engine on its own vector, the layout ath10k
> used until commit cfe9011a05a8 ("ath10k: remove MSI range support").
> Since then ath10k asks for one vector, so all its copy engine
> interrupts and the NAPI they schedule run on one CPU.
>
> Ask for as many MSI vectors as the device offers and spread them over
> the CPUs, starting each further device on the next CPU. Every vector
> keeps the shared handler, which schedules the NAPI that services all
> copy engines, so the copy engine to vector mapping does not matter for
> correctness. With one vector nothing changes.
>
> On the AX3600, with a station uploading 87-88 Mbit/s over the QCA9889,
> the HTT copy engine raised 36891-37590 interrupts in 14 s on vector 6
> and the WMI and HTC ones 165-313 each on vectors 2-4. The receive
> softirq moved from CPU0 (21-25%) to CPU2 (21-35%), with CPU0 at 1-3%.
> Measured with idle-priority spinners, the CPU time of the whole system
> stayed within the run-to-run spread: 25-42% of four CPUs against 22-34%
> uploading, 14-35% against 10-37% downloading (3 runs each). A firmware
> crash forced through debugfs recovered as without the change.
>
> Tested-on: QCA9887 hw1.0 PCI 10.2.4-1.0-00047
>
> Signed-off-by: Julius Bairaktaris <julius@bairaktaris.de>
I'm concerned that you are taking an approach to benefit routers which can
actually harm mobile clients (which want to minimize the number of CPU cores
in use to reduce power consumption). I'm very hesitant to consider a series
for upstream that could be detrimental to mobile devices.
This looks like something that should stay within OpenWrt, which is focused on
the AP, and not come upstream where we have to treat mobile devices nicely.
/jeff
prev parent reply other threads:[~2026-11-18 19:57 UTC|newest]
Thread overview: 5+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-11-01 16:42 Julius Bairaktaris
2026-11-02 12:13 ` Loic Poulain
2026-11-14 21:05 ` Julius Bairaktaris
2026-11-16 15:25 ` Loic Poulain
2026-11-18 19:57 ` Jeff Johnson [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=9b13a6c0-1ff9-4c63-b1fc-9ab6def252bd@oss.qualcomm.com \
--to=jeff.johnson@oss.qualcomm.com \
--cc=ath10k@lists.infradead.org \
--cc=baochen.qiang@oss.qualcomm.com \
--cc=jjohnson@kernel.org \
--cc=julius@bairaktaris.de \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-wireless@vger.kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®