From: Ingo Molnar <mingo@elte.hu>
To: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Steven Rostedt <rostedt@goodmis.org>,
LKML <linux-kernel@vger.kernel.org>,
Andrew Morton <akpm@linux-foundation.org>,
Thomas Gleixner <tglx@linutronix.de>,
Christoph Hellwig <hch@infradead.org>,
john stultz <johnstul@us.ibm.com>,
Oleg Nesterov <oleg@tv-sign.ru>,
"Paul E. McKenney" <paulmck@linux.vnet.ibm.com>,
Dipankar Sarma <dipankar@in.ibm.com>,
"David S. Miller" <davem@davemloft.net>,
matthew.wilcox@hp.com, kuznet@ms2.inr.ac.ru
Subject: Re: [RFC PATCH 0/6] Convert all tasklets to workqueues
Date: Fri, 22 Jun 2007 22:40:58 +0200 [thread overview]
Message-ID: <20070622204058.GA11777@elte.hu> (raw)
In-Reply-To: <alpine.LFD.0.98.0706221010490.3593@woody.linux-foundation.org>
* Linus Torvalds <torvalds@linux-foundation.org> wrote:
> Whether we actually then want to do 6 is another matter. I think we'd
> need some measuring and discussion about that.
basically tasklets have a number of limitations:
- tasklets have certain latency limitations over real tasks. (for
example they are not guaranteed to be re-executed when they are
triggered while they are running, so artificial latencies can be
introduced into the kernel workflow)
- tasklets have certain execution limitations. (only atomic functions
can be executed in them)
- tasklets have certain fairness limitations. (they are executed in
softirq context and thus preempt everything, even if there is some
potentially more important, high-priority task waiting to be
executed.)
- the 'priority levels' approach of softirqs is not really
self-documenting - unlike real locks. As a result we've got some
vague coupling between network softirq processing and timer softirq
processing, which spilled over into tasklets as well. The 'hi' and
'low' concept of tasklets isnt really used either. We should reduce
the amount of such opaque 'coupling' between workflows - it should be
spelled out explicitly via some synchronization construct.
- tasklets are duplicated infrastructure (over existing workqueues)
that, if it's possible to do it compatibly, would be a good idea to
eliminate.
when it comes to 'deferred processing', we've basically got two 'prime'
choices for deferred processing:
- if it's high-performance then it goes into a softirq.
- if performance is not important, or robustness and flexibility is
more important than performance, then workqueues are used.
basically tasklets do _neither_ really well. They are too 'global' to
scale really well on SMP (even the RCU tasklet wasnt a real tasklet: it
was a _per CPU tasklet_, which almost by definition is equivalent to a
softirq, some some extra glue overhead ...), and tasklets are also too
much tied to softirqs to be used as a generic processing context.
that's why i'd like them to be gently but firmly phased out =B-)
Ingo
next prev parent reply other threads:[~2007-06-22 20:43 UTC|newest]
Thread overview: 129+ messages / expand[flat|nested] mbox.gz Atom feed top
2007-06-22 4:00 [RFC PATCH 0/6] Convert all tasklets to workqueues Steven Rostedt
2007-06-22 4:00 ` [RFC PATCH 1/6] Convert the RCU tasklet into a softirq Steven Rostedt
2007-06-22 7:10 ` Christoph Hellwig
2007-06-22 7:43 ` Ingo Molnar
2007-06-22 12:35 ` Steven Rostedt
2007-06-22 12:55 ` Ingo Molnar
2007-06-22 4:00 ` [RFC PATCH 2/6] Split out tasklets from softirq.c Steven Rostedt
2007-06-22 7:11 ` Christoph Hellwig
2007-06-22 12:40 ` Steven Rostedt
2007-06-22 13:45 ` Akinobu Mita
2007-06-22 13:58 ` Steven Rostedt
2007-06-22 4:00 ` [RFC PATCH 3/6] Add a tasklet is-scheduled API Steven Rostedt
2007-06-22 4:00 ` [RFC PATCH 4/6] Make DRM use the tasklet is-sched API Steven Rostedt
2007-06-22 6:36 ` Daniel Walker
2007-06-22 6:49 ` Thomas Gleixner
2007-06-22 7:08 ` Daniel Walker
2007-06-22 12:15 ` Steven Rostedt
2007-06-22 15:36 ` Daniel Walker
2007-06-22 22:38 ` Ingo Molnar
2007-06-22 23:28 ` Daniel Walker
2007-06-22 16:10 ` Arnd Bergmann
2007-06-22 16:56 ` Steven Rostedt
2007-06-22 18:24 ` Christoph Hellwig
2007-06-22 23:38 ` Dave Airlie
2007-06-22 4:00 ` [RFC PATCH 5/6] Move tasklet.h to tasklet_softirq.h Steven Rostedt
2007-06-22 4:00 ` [RFC PATCH 6/6] Convert tasklets to work queues Steven Rostedt
2007-06-22 7:06 ` Daniel Walker
2007-06-22 13:29 ` Steven Rostedt
2007-06-22 15:52 ` Oleg Nesterov
2007-06-22 16:35 ` Steven Rostedt
2007-06-23 11:15 ` Arnd Bergmann
2007-06-22 7:09 ` [RFC PATCH 0/6] Convert all tasklets to workqueues Christoph Hellwig
2007-06-22 7:51 ` Ingo Molnar
2007-06-22 7:53 ` Christoph Hellwig
2007-06-22 11:23 ` Ingo Molnar
2007-06-22 12:32 ` Steven Rostedt
2007-06-22 12:38 ` Ingo Molnar
2007-06-22 12:58 ` Steven Rostedt
2007-06-22 13:12 ` Ingo Molnar
2007-06-22 14:27 ` Steven Rostedt
2007-06-22 13:13 ` Andrew Morton
2007-06-22 13:26 ` Ingo Molnar
2007-06-22 13:41 ` Andrew Morton
2007-06-22 14:00 ` Ingo Molnar
2007-06-22 13:35 ` Steven Rostedt
2007-06-22 14:25 ` Arjan van de Ven
2007-06-22 14:42 ` Steven Rostedt
2007-06-22 14:43 ` Arjan van de Ven
2007-06-22 17:16 ` Linus Torvalds
2007-06-22 17:31 ` Steven Rostedt
2007-06-22 18:32 ` Christoph Hellwig
2007-06-22 20:40 ` Ingo Molnar [this message]
2007-06-22 21:00 ` Christoph Hellwig
2007-06-22 21:10 ` Ingo Molnar
2007-06-22 21:13 ` Thomas Gleixner
2007-06-22 21:37 ` Linus Torvalds
2007-06-22 21:59 ` Ingo Molnar
2007-06-22 22:09 ` Ingo Molnar
2007-06-22 22:43 ` Roland Dreier
2007-06-22 22:57 ` Alan Cox
2007-06-22 22:58 ` Steven Rostedt
2007-06-23 6:23 ` Dave Airlie
2007-06-24 15:16 ` Jonathan Corbet
2007-06-24 15:52 ` Steven Rostedt
2007-06-25 16:50 ` Tilman Schmidt
2007-06-25 17:06 ` Steven Rostedt
2007-06-25 20:50 ` Tilman Schmidt
2007-06-25 21:03 ` Steven Rostedt
2007-06-25 19:52 ` Stephen Hemminger
2007-06-26 0:00 ` Jonathan Corbet
2007-06-26 0:52 ` Steven Rostedt
2007-06-25 18:48 ` Kristian Høgsberg
2007-06-25 19:11 ` Steven Rostedt
2007-06-25 20:07 ` Kristian Høgsberg
2007-06-25 20:31 ` Steven Rostedt
2007-06-25 21:08 ` Kristian Høgsberg
2007-06-25 21:15 ` Ingo Molnar
2007-06-25 23:36 ` Stefan Richter
2007-06-26 0:46 ` Steven Rostedt
2007-06-26 1:46 ` Dan Williams
2007-06-26 2:01 ` Steven Rostedt
2007-06-26 2:12 ` Dan Williams
2007-06-28 12:37 ` Steven Rostedt
2007-06-28 16:37 ` Oleg Nesterov
2007-06-28 18:02 ` Dan Williams
2007-06-28 20:46 ` Steven Rostedt
2007-06-28 21:23 ` Dan Williams
2007-06-28 21:40 ` Dan Williams
2007-06-28 22:01 ` Steven Rostedt
2007-06-28 22:00 ` Steven Rostedt
2007-06-28 5:48 ` Jeff Garzik
2007-06-28 9:23 ` Ingo Molnar
2007-06-28 14:38 ` Alexey Kuznetsov
2007-06-28 15:23 ` Jeff Garzik
2007-06-28 15:54 ` Steven Rostedt
2007-06-28 16:00 ` Ingo Molnar
2007-06-28 17:26 ` Jeff Garzik
2007-06-28 17:44 ` Jeff Garzik
2007-06-28 18:19 ` Andrew Morton
2007-06-28 20:07 ` Ingo Molnar
2007-06-29 11:34 ` Alexey Kuznetsov
2007-06-29 11:48 ` Duncan Sands
2007-06-29 13:36 ` Alexey Kuznetsov
2007-06-29 14:01 ` Duncan Sands
2007-06-29 16:34 ` Alexey Kuznetsov
2007-06-29 12:29 ` Ingo Molnar
2007-06-29 13:25 ` Alexey Kuznetsov
2007-06-29 13:43 ` Ingo Molnar
2007-06-29 15:23 ` Alexey Kuznetsov
2007-06-29 13:41 ` Steven Rostedt
2007-06-29 14:24 ` Jeff Garzik
2007-06-29 14:26 ` Oleg Nesterov
2007-06-29 19:04 ` Alexey Kuznetsov
2007-06-29 14:27 ` Alexey Kuznetsov
2007-06-29 15:51 ` Oleg Nesterov
2007-06-29 16:21 ` Alexey Kuznetsov
2007-06-29 16:52 ` Oleg Nesterov
2007-06-29 17:09 ` Oleg Nesterov
2007-06-30 11:25 ` Oleg Nesterov
2007-06-28 15:17 ` Jeff Garzik
2007-06-22 21:53 ` Daniel Walker
2007-06-22 22:09 ` david
2007-06-22 22:15 ` Daniel Walker
2007-06-22 22:44 ` Ingo Molnar
2007-06-22 23:28 ` Daniel Walker
2007-06-22 22:15 ` Ingo Molnar
2007-06-23 5:14 ` Stephen Hemminger
2007-06-26 13:03 Clemens Ladisch
2007-06-26 13:15 ` Takashi Iwai
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=20070622204058.GA11777@elte.hu \
--to=mingo@elte.hu \
--cc=akpm@linux-foundation.org \
--cc=davem@davemloft.net \
--cc=dipankar@in.ibm.com \
--cc=hch@infradead.org \
--cc=johnstul@us.ibm.com \
--cc=kuznet@ms2.inr.ac.ru \
--cc=linux-kernel@vger.kernel.org \
--cc=matthew.wilcox@hp.com \
--cc=oleg@tv-sign.ru \
--cc=paulmck@linux.vnet.ibm.com \
--cc=rostedt@goodmis.org \
--cc=tglx@linutronix.de \
--cc=torvalds@linux-foundation.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;
as well as URLs for NNTP newsgroup(s).