From: Al Viro <viro@ZenIV.linux.org.uk>
To: Mike Marshall <hubcap@omnibond.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>,
linux-fsdevel <linux-fsdevel@vger.kernel.org>,
Stephen Rothwell <sfr@canb.auug.org.au>
Subject: [RFC] bufmap-related wait logics (Re: Orangefs ABI documentation)
Date: Sun, 7 Feb 2016 20:01:00 +0000 [thread overview]
Message-ID: <20160207200100.GB17997@ZenIV.linux.org.uk> (raw)
In-Reply-To: <20160207035331.GZ17997@ZenIV.linux.org.uk>
On Sun, Feb 07, 2016 at 03:53:31AM +0000, Al Viro wrote:
> BTW, could you try to reproduce that WARN_ON with these two patches added
> and with bufmap debugging turned on? Both double-free and lack of rewinding
> are real; I can see scenarios where they would trigger, and I'm pretty sure
> that the latter is triggering in your reproducer. Moreover, I'm absolutely
> sure that spurious dropping of bufmap references is happening there; what I'm
> not sure is whether it was on this double-free or on something else...
AFAICS, with bufmap we have 6 kinds of events -
1) daemon installs a bufmap
2) daemon shuts down
3) wait_for_direct_io() requests a read/write slot
4) orangefs_readdir() requests a readdir slot
5) wait_for_direct_io() releases a slot
6) orangefs_readdir() releases a slot
and the whole thing can be described via two counters and two waitqueues.
Rules:
Initially C1 = C2 = -1
(1) if C1 >= 0
sod off, we'd already installed that thing
else
C1 = number of read/write slots
wake up that many of those who wait on Q1
C2 = number of readdir slots
wake up that many of those who wait on Q2
(2) C1 -= number of read/write slots + 1
C2 -= number of readdir slots + 1
wait on Q1 for C1 == -1
wait on Q2 for C2 == -1
(3) if C1 <= 0
end = now + 15 minutes
while true
if C1 < 0
interruptibly wait on Q1 for (C1 > 0)
up to min(end - now, 30s)
if C1 < 0
return -ETIMEDOUT
else
interruptibly wait on Q1 for (C1 > 0)
up to end - now, exclusive
if C1 > 0
break
if signal arrived
return -EINTR
if now after end
return -ETIMEDOUT
C1--, and grab a slot in read/write slots bitmap
(5) release a slot in bitmap; C1++; wake up Q1
(4,6) same as (3,5) with s/C1/C2/, s/Q1/Q2/
I'd probably use Q1.lock for serializing C1 and Q2.lock for C2; the only
obstacle is the lack of timeout versions of
wait_event_interruptible{,_exclusive}locked()
(and obscene identifier length of such beasts, of course).
The really annoying thing is that it's very similar to a couple of counting
semaphores; home-grown wait primitive is almost always a Bad Idea(tm) and
if somebody sees a sane way to cobble that out of higher-level ones, I'd
very much prefer that. Suggestions?
next prev parent reply other threads:[~2016-02-07 20:01 UTC|newest]
Thread overview: 111+ messages / expand[flat|nested] mbox.gz Atom feed top
2016-01-15 21:46 Orangefs ABI documentation Mike Marshall
2016-01-22 7:11 ` Al Viro
2016-01-22 11:09 ` Mike Marshall
2016-01-22 16:59 ` Mike Marshall
2016-01-22 17:08 ` Al Viro
2016-01-22 17:40 ` Mike Marshall
2016-01-22 17:43 ` Al Viro
2016-01-22 18:17 ` Mike Marshall
2016-01-22 18:37 ` Al Viro
2016-01-22 19:07 ` Mike Marshall
2016-01-22 19:21 ` Mike Marshall
2016-01-22 20:04 ` Al Viro
2016-01-22 20:30 ` Mike Marshall
2016-01-23 0:12 ` Al Viro
2016-01-23 1:28 ` Al Viro
2016-01-23 2:54 ` Mike Marshall
2016-01-23 19:10 ` Al Viro
2016-01-23 19:24 ` Mike Marshall
2016-01-23 21:35 ` Mike Marshall
2016-01-23 22:05 ` Al Viro
2016-01-23 21:40 ` Al Viro
2016-01-23 22:36 ` Mike Marshall
2016-01-24 0:16 ` Al Viro
2016-01-24 4:05 ` Al Viro
2016-01-24 22:12 ` Mike Marshall
2016-01-30 17:22 ` Al Viro
2016-01-26 19:52 ` Martin Brandenburg
2016-01-30 17:34 ` Al Viro
2016-01-30 18:27 ` Al Viro
2016-02-04 23:30 ` Mike Marshall
2016-02-06 19:42 ` Al Viro
2016-02-07 1:38 ` Al Viro
2016-02-07 3:53 ` Al Viro
2016-02-07 20:01 ` Al Viro [this message]
2016-02-08 22:26 ` Mike Marshall
2016-02-08 23:35 ` Al Viro
2016-02-09 3:32 ` Al Viro
2016-02-09 14:34 ` Mike Marshall
2016-02-09 17:40 ` Al Viro
2016-02-09 21:06 ` Al Viro
2016-02-09 22:25 ` Mike Marshall
2016-02-11 23:36 ` Mike Marshall
2016-02-09 22:02 ` Mike Marshall
2016-02-09 22:16 ` Al Viro
2016-02-09 22:40 ` Al Viro
2016-02-09 23:13 ` Al Viro
2016-02-10 16:44 ` Al Viro
2016-02-10 21:26 ` Al Viro
2016-02-11 23:54 ` Mike Marshall
2016-02-12 0:55 ` Al Viro
2016-02-12 12:13 ` Mike Marshall
2016-02-11 0:44 ` Al Viro
2016-02-11 3:22 ` Mike Marshall
2016-02-12 4:27 ` Al Viro
2016-02-12 12:26 ` Mike Marshall
2016-02-12 18:00 ` Martin Brandenburg
2016-02-13 17:18 ` Mike Marshall
2016-02-13 17:47 ` Al Viro
2016-02-14 2:56 ` Al Viro
2016-02-14 3:46 ` [RFC] slot allocator - waitqueue use review needed (Re: Orangefs ABI documentation) Al Viro
2016-02-14 4:06 ` Al Viro
2016-02-16 2:12 ` Al Viro
2016-02-16 19:28 ` Al Viro
2016-02-14 22:31 ` Orangefs ABI documentation Mike Marshall
2016-02-14 23:43 ` Al Viro
2016-02-15 17:46 ` Mike Marshall
2016-02-15 18:45 ` Al Viro
2016-02-15 22:32 ` Martin Brandenburg
2016-02-15 23:04 ` Al Viro
2016-02-16 23:15 ` Mike Marshall
2016-02-16 23:36 ` Al Viro
2016-02-16 23:54 ` Al Viro
2016-02-17 19:24 ` Mike Marshall
2016-02-17 20:11 ` Al Viro
2016-02-17 21:17 ` Al Viro
2016-02-17 22:24 ` Mike Marshall
2016-02-17 22:40 ` Martin Brandenburg
2016-02-17 23:09 ` Al Viro
2016-02-17 23:15 ` Al Viro
2016-02-18 0:04 ` Al Viro
2016-02-18 11:11 ` Al Viro
2016-02-18 18:58 ` Mike Marshall
2016-02-18 19:20 ` Al Viro
2016-02-18 19:49 ` Martin Brandenburg
2016-02-18 20:08 ` Mike Marshall
2016-02-18 20:22 ` Mike Marshall
2016-02-18 20:38 ` Mike Marshall
2016-02-18 20:52 ` Al Viro
2016-02-18 21:50 ` Mike Marshall
2016-02-19 0:25 ` Al Viro
2016-02-19 22:11 ` Mike Marshall
2016-02-19 22:22 ` Al Viro
2016-02-20 12:14 ` Mike Marshall
2016-02-20 13:36 ` Al Viro
2016-02-22 16:20 ` Mike Marshall
2016-02-22 21:22 ` Mike Marshall
2016-02-23 21:58 ` Mike Marshall
2016-02-26 20:21 ` Mike Marshall
2016-02-19 22:32 ` Al Viro
2016-02-19 22:45 ` Martin Brandenburg
2016-02-19 22:50 ` Martin Brandenburg
2016-02-18 20:49 ` Al Viro
2016-02-15 22:47 ` Mike Marshall
2016-01-23 22:46 ` write() semantics (Re: Orangefs ABI documentation) Al Viro
2016-01-23 23:35 ` Linus Torvalds
2016-03-03 22:25 ` Mike Marshall
2016-03-04 20:55 ` Mike Marshall
2016-01-22 20:51 ` Orangefs ABI documentation Mike Marshall
2016-01-22 23:53 ` Mike Marshall
2016-01-22 19:54 ` Al Viro
2016-01-22 19:50 ` Al Viro
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=20160207200100.GB17997@ZenIV.linux.org.uk \
--to=viro@zeniv.linux.org.uk \
--cc=hubcap@omnibond.com \
--cc=linux-fsdevel@vger.kernel.org \
--cc=sfr@canb.auug.org.au \
--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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.