From: Peter Zijlstra <peterz@infradead.org>
To: Andi Kleen <andi@firstfloor.org>
Cc: Huang Ying <ying.huang@intel.com>, Ingo Molnar <mingo@elte.hu>,
"H. Peter Anvin" <hpa@zytor.com>,
linux-kernel@vger.kernel.org
Subject: Re: [RFC][PATCH] irq_work
Date: Fri, 25 Jun 2010 12:38:51 +0200 [thread overview]
Message-ID: <1277462331.32034.221.camel@twins> (raw)
In-Reply-To: <20100624123804.GK578@basil.fritz.box>
On Thu, 2010-06-24 at 14:38 +0200, Andi Kleen wrote:
> The sleepable
> soft irq would have avoided that (that's not a show stopper)
I'm still not convinced sleepable softirq is a workable thing.
Softirqs:
A) are non-preemptible
B) are per-cpu because of A
C) can be ran from ksoftirqd context
D) generic kernel infrastructure with identical semantics on all archs
If you were to make something like a sleepable softirq, you'd loose A
(per definition), B (sleepable implies migratable to cpus_allowed) and
possibly D (unless you want to touch all architectures).
Now from your 'requirements':
> I have one case that needs to sleep (but only when interrupting user code)
> TIF works for user space, but it's a bit ugly because it requires adding
> more data to the task_struct because CPUs can change.
Which I read as:
1) needs to run in the task context of the task that got 'interrupted'
2) needs to stay on the cpu it got interrupted on.
So C is out of the window too, at which point there's nothing resembling
softirqs left.
To boot, x86_64 runs softirqs from the hardirq stack:
/* Call softirq on interrupt stack. Interrupts are off. */
ENTRY(call_softirq)
CFI_STARTPROC
push %rbp
CFI_ADJUST_CFA_OFFSET 8
CFI_REL_OFFSET rbp,0
mov %rsp,%rbp
CFI_DEF_CFA_REGISTER rbp
incl PER_CPU_VAR(irq_count)
cmove PER_CPU_VAR(irq_stack_ptr),%rsp
push %rbp # backlink for old unwinder
call __do_softirq
leaveq
CFI_DEF_CFA_REGISTER rsp
CFI_ADJUST_CFA_OFFSET -8
decl PER_CPU_VAR(irq_count)
ret
CFI_ENDPROC
END(call_softirq)
Also, -rt has something that could be considered sleepable softirqs,
although we call them preemptible softirqs. It runs all softirqs from
cpu bound kthreads, which again doesn't match your requirements.
So no, I don't think your idea of sleepable softirqs is sound.
next prev parent reply other threads:[~2010-06-25 10:39 UTC|newest]
Thread overview: 66+ messages / expand[flat|nested] mbox.gz Atom feed top
2010-06-24 3:04 [RFC 1/5] Make soft_irq NMI safe Huang Ying
2010-06-24 3:04 ` [RFC 2/5] NMI return notifier Huang Ying
2010-06-24 3:04 ` [RFC 3/5] x86, trigger NMI return notifier soft_irq earlier Huang Ying
2010-06-24 6:03 ` Peter Zijlstra
2010-06-24 3:04 ` [RFC 4/5] x86, Use NMI return notifier in MCE Huang Ying
2010-06-24 10:00 ` Andi Kleen
2010-06-24 3:04 ` [RFC 5/5] Use NMI return notifier in perf pending Huang Ying
2010-06-24 6:00 ` Peter Zijlstra
2010-06-24 6:09 ` [RFC 1/5] Make soft_irq NMI safe Peter Zijlstra
2010-06-24 6:45 ` Huang Ying
2010-06-24 6:35 ` [RFC][PATCH] irq_work Peter Zijlstra
2010-06-24 6:43 ` Huang Ying
2010-06-24 6:47 ` Peter Zijlstra
2010-06-24 6:50 ` Huang Ying
2010-06-24 6:58 ` Peter Zijlstra
2010-06-24 7:04 ` Huang Ying
2010-06-24 7:19 ` Peter Zijlstra
2010-06-24 7:27 ` Huang Ying
2010-06-24 7:32 ` Peter Zijlstra
2010-06-24 10:27 ` Andi Kleen
2010-06-24 10:30 ` Peter Zijlstra
2010-06-24 10:52 ` Andi Kleen
2010-06-24 10:58 ` Peter Zijlstra
2010-06-24 11:08 ` Andi Kleen
2010-06-24 11:10 ` Peter Zijlstra
2010-06-24 11:20 ` Andi Kleen
2010-06-24 11:33 ` Peter Zijlstra
2010-06-24 11:55 ` Andi Kleen
2010-06-24 11:57 ` Peter Zijlstra
2010-06-24 12:02 ` Andi Kleen
2010-06-24 12:18 ` Peter Zijlstra
2010-06-24 12:38 ` Andi Kleen
2010-06-25 10:38 ` Peter Zijlstra [this message]
2010-06-24 11:42 ` Peter Zijlstra
2010-06-24 11:58 ` Andi Kleen
2010-06-24 12:02 ` Peter Zijlstra
2010-06-24 11:23 ` Ingo Molnar
2010-06-24 11:34 ` Peter Zijlstra
2010-06-24 12:35 ` Ingo Molnar
2010-06-24 13:02 ` Andi Kleen
2010-06-24 13:20 ` Borislav Petkov
2010-06-24 13:33 ` Andi Kleen
2010-06-24 13:42 ` Ingo Molnar
2010-06-24 13:46 ` Ingo Molnar
2010-06-24 14:01 ` Andi Kleen
2010-06-24 15:41 ` Borislav Petkov
2010-06-24 16:09 ` Andi Kleen
2010-06-25 2:12 ` Huang Ying
2010-06-25 7:48 ` Peter Zijlstra
2010-06-25 9:17 ` Huang Ying
2010-06-25 9:23 ` Frederic Weisbecker
2010-06-25 9:30 ` Huang Ying
2010-06-25 9:44 ` Frederic Weisbecker
2010-06-25 9:30 ` Peter Zijlstra
2010-06-25 11:58 ` huang ying
2010-06-25 9:08 ` Andi Kleen
2010-06-25 18:30 ` [RFC][PATCH] irq_work -v2 Peter Zijlstra
2010-06-25 19:30 ` Andi Kleen
2010-06-25 19:39 ` Peter Zijlstra
2010-06-25 19:49 ` Peter Zijlstra
2010-06-25 22:29 ` Andi Kleen
2010-06-26 8:36 ` Peter Zijlstra
2010-06-26 10:08 ` Andi Kleen
2010-06-26 10:32 ` Peter Zijlstra
2010-06-25 19:47 ` Peter Zijlstra
2010-06-26 1:26 ` huang ying
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=1277462331.32034.221.camel@twins \
--to=peterz@infradead.org \
--cc=andi@firstfloor.org \
--cc=hpa@zytor.com \
--cc=linux-kernel@vger.kernel.org \
--cc=mingo@elte.hu \
--cc=ying.huang@intel.com \
/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;
as well as URLs for NNTP newsgroup(s).