From: Jens Axboe <axboe@suse.de>
To: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: Andrew Morton <akpm@osdl.org>,
riel@redhat.com, marcelo.tosatti@cyclades.com, andrea@suse.de,
linux-kernel@vger.kernel.org
Subject: Re: memory barrier in ll_rw_blk.c (was Re: [PATCH][5/?] count writeback pages in nr_scanned)
Date: Thu, 6 Jan 2005 09:32:52 +0100 [thread overview]
Message-ID: <20050106083251.GH17821@suse.de> (raw)
In-Reply-To: <41DCF3EC.3090506@yahoo.com.au>
On Thu, Jan 06 2005, Nick Piggin wrote:
> Jens Axboe wrote:
> >On Thu, Jan 06 2005, Nick Piggin wrote:
>
> >>
> >>This memory barrier is not needed because the waitqueue will only get
> >>waiters on it in the following situations:
> >>
> >>rq->count has exceeded the threshold - however all manipulations of
> >>->count
> >>are performed under the runqueue lock, and so we will correctly pick up
> >>any
> >>waiter.
> >>
> >>Memory allocation for the request fails. In this case, there is no
> >>additional
> >>help provided by the memory barrier. We are guaranteed to eventually wake
> >>up waiters because the request allocation mempool guarantees that if the
> >>mem
> >>allocation for a request fails, there must be some requests in flight.
> >>They
> >>will wake up waiters when they are retired.
> >
> >
> >Not sure I agree completely. Yes it will work, but only because it tests
> ><= q->nr_requests and I don't think that 'eventually' is good enough :-)
> >
> >The actual waitqueue manipulation doesn't happen under the queue lock,
> >so the memory barrier is needed to pickup the change on SMP. So I'd like
> >to keep the barrier.
> >
>
> No that's right... but between the prepare_to_wait and the io_schedule,
> get_request takes the lock and checks nr_requests. I think we are safe?
It looks like it, yes you are right. But it looks to be needed a few
lines further down instead, though :-)
===== drivers/block/ll_rw_blk.c 1.281 vs edited =====
--- 1.281/drivers/block/ll_rw_blk.c 2004-12-01 09:13:57 +01:00
+++ edited/drivers/block/ll_rw_blk.c 2005-01-06 09:32:19 +01:00
@@ -1630,11 +1630,11 @@
if (rl->count[rw] < queue_congestion_off_threshold(q))
clear_queue_congested(q, rw);
if (rl->count[rw]+1 <= q->nr_requests) {
- smp_mb();
if (waitqueue_active(&rl->wait[rw]))
wake_up(&rl->wait[rw]);
blk_clear_queue_full(q, rw);
}
+ smp_mb();
if (unlikely(waitqueue_active(&rl->drain)) &&
!rl->count[READ] && !rl->count[WRITE])
wake_up(&rl->drain);
> >I'd prefer to add smp_mb() to waitqueue_active() actually!
> >
>
> That may be a good idea (I haven't really taken much notice of how other
> code uses it).
>
> I'm not worried about any possible performance advantages of removing it,
> rather just having a memory barrier without comments can be perplexing.
I fully agree, subtle things like that should always be commented.
--
Jens Axboe
next prev parent reply other threads:[~2005-01-06 8:33 UTC|newest]
Thread overview: 43+ messages / expand[flat|nested] mbox.gz Atom feed top
2005-01-03 17:25 [PATCH][5/?] count writeback pages in nr_scanned Rik van Riel
2005-01-05 10:08 ` Andrew Morton
2005-01-05 18:06 ` Andrea Arcangeli
2005-01-05 18:50 ` Rik van Riel
2005-01-05 17:49 ` Marcelo Tosatti
2005-01-05 21:44 ` Andrew Morton
2005-01-05 20:32 ` Marcelo Tosatti
2005-01-05 23:51 ` Nick Piggin
2005-01-06 1:27 ` Rik van Riel
2005-01-06 1:33 ` Nick Piggin
2005-01-06 1:37 ` Andrew Morton
2005-01-06 1:40 ` Nick Piggin
2005-01-06 1:52 ` Andrea Arcangeli
2005-01-06 1:36 ` Andrew Morton
2005-01-06 3:42 ` Rik van Riel
2005-01-06 3:50 ` Nick Piggin
2005-01-06 4:26 ` Andrew Morton
2005-01-06 4:35 ` Nick Piggin
2005-01-06 4:47 ` Andrew Morton
2005-01-06 4:55 ` Nick Piggin
2005-01-06 5:03 ` Andrea Arcangeli
2005-01-06 8:06 ` Jens Axboe
2005-01-06 8:16 ` memory barrier in ll_rw_blk.c (was Re: [PATCH][5/?] count writeback pages in nr_scanned) Nick Piggin
2005-01-06 8:32 ` Jens Axboe [this message]
2005-01-06 8:53 ` Nick Piggin
2005-01-06 12:00 ` Jens Axboe
2005-01-06 4:59 ` [PATCH][5/?] count writeback pages in nr_scanned Andrea Arcangeli
2005-01-06 5:05 ` Andrew Morton
2005-01-06 5:17 ` Andrea Arcangeli
2005-01-06 5:19 ` Nick Piggin
2005-01-06 5:25 ` Andrea Arcangeli
2005-01-06 5:36 ` Nick Piggin
2005-01-06 5:44 ` Nick Piggin
2005-01-06 5:37 ` Andrew Morton
2005-01-06 5:59 ` Andrea Arcangeli
2005-01-06 13:28 ` Rik van Riel
2005-01-06 5:32 ` Andrew Morton
2005-01-06 5:46 ` Andrea Arcangeli
2005-01-06 5:59 ` Andrew Morton
2005-01-06 6:16 ` Andrea Arcangeli
2005-01-06 5:06 ` Nick Piggin
2005-01-06 5:21 ` Andrea Arcangeli
2005-01-05 23:26 ` Andrew Morton
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=20050106083251.GH17821@suse.de \
--to=axboe@suse.de \
--cc=akpm@osdl.org \
--cc=andrea@suse.de \
--cc=linux-kernel@vger.kernel.org \
--cc=marcelo.tosatti@cyclades.com \
--cc=nickpiggin@yahoo.com.au \
--cc=riel@redhat.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).