linux-kernel - Re: [RFC PATCH 3/3] sched/Documentation: Point out use of preempt_schedule

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]

Date:   Fri, 1 Feb 2019 10:27:54 +0000
From:   Valentin Schneider <valentin.schneider@....com>
To:     Julien Thierry <julien.thierry@....com>,
        linux-kernel@...r.kernel.org
Cc:     Jonathan Corbet <corbet@....net>, Ingo Molnar <mingo@...hat.com>,
        Peter Zijlstra <peterz@...radead.org>,
        Thomas Gleixner <tglx@...utronix.de>, linux-doc@...r.kernel.org
Subject: Re: [RFC PATCH 3/3] sched/Documentation: Point out use of
 preempt_schedule_irq()

On 01/02/2019 08:45, Julien Thierry wrote:
[...]
>> +Kernel preemption
>> +=================
>> +When returning from interrupt context, you should call either of
>> +preempt_schedule() or preempt_schedule_irq() if preemption is enabled
>> +and need_resched() is true.
>> +
> 
> I don't think preempt_schedule() is really an option for a return from
> interrupt. First thing preempt_schedule() does is:
> 
>         if (likely(!preemptible()))
> 		return;
> 
> And preemptible() is:
> 
> preempt_count() == 0 && !irqs_disabled()
> 
> Generally on return from interrupt context interrupts are disabled, so
> we would never be preemptible() and preempt_schedule() would just do
> nothing.
> 
> Unless I'm missing something.
> 

No, I think you're right. The main reason I still mentioned it here is to
be conservative, although I must admit I've started confusing what we
have vs what we used to have since my dive into the history.

If I look at some archs that don't use preempt_schedule_irq() (e.g. alpha,
unicore32), they seem to be calling schedule() directly - but I don't see
any (*current*) user of preempt_schedule() on interrupt return.

preempt_schedule() still has this comment attached to it:

 * this is the entry point to schedule() from in-kernel preemption
 * off of preempt_enable. Kernel preemptions off return from interrupt
 * occur there and call schedule directly.

So I might just remove the mention to preempt_schedule() in the doc and
also change the comment.

Thanks,
Valentin