From: Mikulas Patocka <mpatocka@redhat.com>
To: Jens Axboe <jens.axboe@oracle.com>
Cc: linux-kernel@vger.kernel.org, Neil Brown <neilb@suse.de>
Subject: Re: [PATCH 1/2] Avoid bio_endio recursion
Date: Thu, 3 Jul 2008 17:08:10 -0400 (EDT) [thread overview]
Message-ID: <Pine.LNX.4.64.0807031704120.20244@engineering.redhat.com> (raw)
In-Reply-To: <20080702082511.GF20055@kernel.dk>
On Wed, 2 Jul 2008, Jens Axboe wrote:
> On Wed, Jul 02 2008, Mikulas Patocka wrote:
>>> Right, that wont work of course. Completions are typically done through
>>> a softirq, so it is not currently done with hard interrupts disabled.
>>
>> I thought, from hardirq - that's what IDE is doing. And they are called
>> with interrupts disabled (maybe unless you specify unmaskirq, which is not
>> default). What block driver does completions with softirq? ... and why?
>
> The key word is 'typically', the old IDE driver really isn't used very
> much. The SCSI layer and eg cciss uses the block layer softirq
> completions, so that is what 99% of the uses will be.
I use the old IDE driver, I don't see a reason why driver should create
SCSI requests and lower layer translate them to ATA commands.
Just to look, this is the first version of the patch without disabling
interrupts. If you are concerned about disabling interrupts, you can go
this way.
Mikulas
Index: linux-2.6.26-rc5-devel/fs/bio.c
===================================================================
--- linux-2.6.26-rc5-devel.orig/fs/bio.c 2008-06-18 16:05:45.000000000 +0200
+++ linux-2.6.26-rc5-devel/fs/bio.c 2008-06-18 23:27:28.000000000 +0200
@@ -1168,6 +1168,25 @@
**/
void bio_endio(struct bio *bio, int error)
{
+ static DEFINE_PER_CPU(local_t, bio_end_queue) = LOCAL_INIT(0);
+ local_t *bio_end_queue_ptr;
+ unsigned long l;
+
+ bio->bi_next = NULL;
+
+ bio_end_queue_ptr = &get_cpu_var(bio_end_queue);
+ if (local_read(bio_end_queue_ptr)) {
+insert_race:
+ bio->bi_next = (void *)local_read(bio_end_queue_ptr);
+ if (unlikely(local_cmpxchg(bio_end_queue_ptr, (unsigned long)bio->bi_next, (unsigned long)bio) != (unsigned long)bio->bi_next))
+ goto insert_race;
+ put_cpu_var(bio_end_queue);
+ return;
+ }
+
+ local_set(bio_end_queue_ptr, 1);
+
+process_bio:
if (error)
clear_bit(BIO_UPTODATE, &bio->bi_flags);
else if (!test_bit(BIO_UPTODATE, &bio->bi_flags))
@@ -1175,6 +1194,19 @@
if (bio->bi_end_io)
bio->bi_end_io(bio, error);
+
+remove_race:
+ l = local_read(bio_end_queue_ptr);
+ if (l != 1) {
+ bio = (void *)l;
+ if (unlikely(local_cmpxchg(bio_end_queue_ptr, (unsigned long)bio, (unsigned long)bio->bi_next) != (unsigned long)bio))
+ goto remove_race;
+ goto process_bio;
+ }
+ if (unlikely(local_cmpxchg(bio_end_queue_ptr, 1, 0) != 1))
+ goto remove_race;
+
+ put_cpu_var(bio_end_queue);
}
void bio_pair_release(struct bio_pair *bp)
next prev parent reply other threads:[~2008-07-03 21:09 UTC|newest]
Thread overview: 20+ messages / expand[flat|nested] mbox.gz Atom feed top
2008-06-24 5:22 Mikulas Patocka
2008-06-24 6:08 ` Neil Brown
2008-06-24 14:36 ` Mikulas Patocka
2008-06-24 8:07 ` Jens Axboe
2008-06-24 14:27 ` Mikulas Patocka
2008-06-25 8:24 ` Jens Axboe
2008-06-26 0:13 ` Mikulas Patocka
2008-06-26 7:07 ` Jens Axboe
2008-07-02 4:09 ` Mikulas Patocka
2008-07-02 8:00 ` Alan Cox
2008-07-03 21:03 ` Mikulas Patocka
2008-07-02 8:25 ` Jens Axboe
2008-07-03 21:08 ` Mikulas Patocka [this message]
2008-07-03 21:04 ` Alan Cox
2008-07-03 22:54 ` Mikulas Patocka
2008-07-03 23:00 ` Alan Cox
2008-07-03 23:51 ` Mikulas Patocka
2008-07-03 23:44 ` Alan Cox
2008-07-04 3:26 ` Mikulas Patocka
2008-07-04 8:11 ` Alan Cox
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=Pine.LNX.4.64.0807031704120.20244@engineering.redhat.com \
--to=mpatocka@redhat.com \
--cc=jens.axboe@oracle.com \
--cc=linux-kernel@vger.kernel.org \
--cc=neilb@suse.de \
/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
Powered by JetHome