From: Philippe Gerum <rpm@xenomai.org>
To: Florian Bezdeka <florian.bezdeka@siemens.com>
Cc: xenomai@lists.linux.dev
Subject: Re: [PATCH RFC Dovetail 6.16 5/5] kernel/irq/chip: Do not call low level irq chip hooks directly
Date: Sun, 28 Sep 2025 10:21:27 +0200 [thread overview]
Message-ID: <875xd3rnqw.fsf@xenomai.org> (raw)
In-Reply-To: <87bjmvro65.fsf@xenomai.org> (Philippe Gerum's message of "Sun, 28 Sep 2025 10:12:18 +0200")
Philippe Gerum <rpm@xenomai.org> writes:
> Florian Bezdeka <florian.bezdeka@siemens.com> writes:
>
>> irq_mask() and irq_unmask() are tracking a software IRQ state that
>> might run out of sync when bypassing them.
>>
>> No functional change.
>>
>
> Actually, there is. Percpu IRQs are not serialized by the desc->lock, so
> calling mask_irq/unmask_irq in these handlers is unsafe since we may end
> up flipping bits from the irqd state concurrently on multiple CPUs for
> the same descriptor. This is the reason why masking/unmasking was
> open-coded there. This patch typically breaks my kvm/arm64 fixture at
> boot, not observed on kvm/x86 so far though.
With respect to the IRQ state consistency, the logic assumes that the
percpu IRQ is unmasked in the descriptor while its handler runs, so not
mirroring the hardware state into the irqd state word is correct.
--
Philippe.
next prev parent reply other threads:[~2025-09-28 8:21 UTC|newest]
Thread overview: 15+ messages / expand[flat|nested] mbox.gz Atom feed top
2025-09-25 13:05 [PATCH RFC Dovetail 6.16 0/5] Dovetail: Minor cleanups found while debugging the pipeline entry code Florian Bezdeka
2025-09-25 13:05 ` [PATCH RFC Dovetail 6.16 1/5] kernel/entry: irq_pipeline: Do not leave .noinstr.text section Florian Bezdeka
2025-09-26 8:49 ` Philippe Gerum
2025-09-26 10:26 ` Florian Bezdeka
2025-09-26 10:37 ` Philippe Gerum
2025-09-25 13:05 ` [PATCH RFC Dovetail 6.16 2/5] kernel/irq/chip: Change return value of get_flow_step() to int Florian Bezdeka
2025-09-25 13:05 ` [PATCH RFC Dovetail 6.16 3/5] kernel/irq/chip: Simplify may_start_flow() Florian Bezdeka
2025-09-26 8:50 ` Philippe Gerum
2025-09-25 13:05 ` [PATCH RFC Dovetail 6.16 4/5] kernel/irq/chip: Simplify should_feed_pipeline() Florian Bezdeka
2025-09-25 13:05 ` [PATCH RFC Dovetail 6.16 5/5] kernel/irq/chip: Do not call low level irq chip hooks directly Florian Bezdeka
2025-09-28 8:12 ` Philippe Gerum
2025-09-28 8:21 ` Philippe Gerum [this message]
2025-09-29 21:15 ` Florian Bezdeka
2025-09-30 7:52 ` Philippe Gerum
2025-09-28 12:25 ` [PATCH RFC Dovetail 6.16 0/5] Dovetail: Minor cleanups found while debugging the pipeline entry code Philippe Gerum
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=875xd3rnqw.fsf@xenomai.org \
--to=rpm@xenomai.org \
--cc=florian.bezdeka@siemens.com \
--cc=xenomai@lists.linux.dev \
/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