From: Rik van Riel <riel@surriel.com>
To: Michel Lespinasse <walken@google.com>
Cc: Peter Zijlstra <peterz@infradead.org>,
Sasha Levin <sasha.levin@oracle.com>,
torvalds@linux-foundation.org, davidlohr.bueso@hp.com,
linux-kernel@vger.kernel.org, akpm@linux-foundation.org,
hhuang@redhat.com, jason.low2@hp.com, lwoodman@redhat.com,
chegu_vinod@hp.com, Dave Jones <davej@redhat.com>,
benisty.e@gmail.com, Ingo Molnar <mingo@redhat.com>
Subject: Re: [PATCH v2 -mm -next] ipc,sem: fix lockdep false positive
Date: Fri, 29 Mar 2013 08:07:35 -0400 [thread overview]
Message-ID: <51558407.5040005@surriel.com> (raw)
In-Reply-To: <CANN689GGwztb=beCNtUU8SJp6sfjn_mNTaHQxezt=ehtxvYBAA@mail.gmail.com>
On 03/28/2013 10:50 PM, Michel Lespinasse wrote:
> On Thu, Mar 28, 2013 at 1:23 PM, Rik van Riel <riel@surriel.com> wrote:
>> Subject: [PATCH -mm -next] ipc,sem: change locking scheme to make lockdep happy
>>
>> Unfortunately the locking scheme originally proposed has false positives
>> with lockdep. This can be fixed by changing the code to only ever take
>> one lock, and making sure that other relevant locks are not locked, before
>> entering a critical section.
>>
>> For the "global lock" case, this is done by taking the sem_array lock,
>> and then (potentially) waiting for all the semaphore's spinlocks to be
>> unlocked.
>>
>> For the "local lock" case, we wait on the sem_array's lock to be free,
>> before taking the semaphore local lock. To prevent races, we need to
>> check again after we have taken the local lock.
>>
>> Suggested-by: Peter Zijlstra <peterz@infradead.org>
>> Reported-by: Sasha Levin <sasha.levin@oracle.com>
>> Signed-off-by: Rik van Riel <riel@redhat.com>
>
> TL;DR: The locking algorithm is not familiar for me, but it seems
> sound. There are some implementation details I don't like. More
> below...
>
>> ---
>> ipc/sem.c | 55 ++++++++++++++++++++++++++++++++++++++++---------------
>> 1 files changed, 40 insertions(+), 15 deletions(-)
>>
>> diff --git a/ipc/sem.c b/ipc/sem.c
>> index 36500a6..87b74d5 100644
>> --- a/ipc/sem.c
>> +++ b/ipc/sem.c
>> @@ -320,24 +320,39 @@ void __init sem_init (void)
>> }
>>
>> /*
>> - * If the sem_array contains just one semaphore, or if multiple
>> - * semops are performed in one syscall, or if there are complex
>> - * operations pending, the whole sem_array is locked.
>> - * If one semop is performed on an array with multiple semaphores,
>> - * get a shared lock on the array, and lock the individual semaphore.
>> + * If the request contains only one semaphore operation, and there are
>> + * no complex transactions pending, lock only the semaphore involved.
>> + * Otherwise, lock the entire semaphore array, since we either have
>> + * multiple semaphores in our own semops, or we need to look at
>> + * semaphores from other pending complex operations.
>> *
>> * Carefully guard against sma->complex_count changing between zero
>> * and non-zero while we are spinning for the lock. The value of
>> * sma->complex_count cannot change while we are holding the lock,
>> * so sem_unlock should be fine.
>> + *
>> + * The global lock path checks that all the local locks have been released,
>> + * checking each local lock once. This means that the local lock paths
>> + * cannot start their critical sections while the global lock is held.
>> */
>> static inline int sem_lock(struct sem_array *sma, struct sembuf *sops,
>> int nsops)
>> {
>> int locknum;
>> + again:
>> if (nsops == 1 && !sma->complex_count) {
>> struct sem *sem = sma->sem_base + sops->sem_num;
>>
>> + /*
>> + * Another process is holding the global lock on the
>> + * sem_array. Wait for that process to release the lock,
>> + * before acquiring our lock.
>> + */
>> + if (unlikely(spin_is_locked(&sma->sem_perm.lock))) {
>> + spin_unlock_wait(&sma->sem_perm.lock);
>> + goto again;
>> + }
>> +
>
> So, there are a few things I don't like about spin_unlock_wait():
>
> 1- From a lock ordering point of view, it is strictly equivalent to
> taking the lock and then releasing it - and yet, lockdep won't catch
> any deadlocks that involve spin_unlock_wait. (Not your fault here,
> this should be fixed as a separate change in lockdep. I manually
> looked at the lock ordering here and found it safe).
>
> 2- With the current ticket lock implementation, a stream of lockers
> can starve spin_unlock_wait() forever. Once again, not your fault and
> I suspect this could be fixed - I expect spin_unlock_wait() callers
> actually only want to know that the lock has been passed on, not that
> it actually got to an unlocked state.
>
> 3- Regarding your actual use here - I find it confusing to call
> spin_unlock_wait() before holding any other lock. The pattern I expect
> to see is that people take one lock, then see that the other lock they
> want is already taken, so they release the first lock and wait on the
> second. So, I'd suggest we remove the sem_perm.lock checks here and
> deal with this in a retry path later down.
>
>> /* Lock just the semaphore we are interested in. */
>> spin_lock(&sem->lock);
>>
>> @@ -347,17 +362,33 @@ static inline int sem_lock(struct sem_array *sma, struct sembuf *sops,
>> */
>> if (unlikely(sma->complex_count)) {
>> spin_unlock(&sem->lock);
>> - goto lock_all;
>> + goto lock_array;
>> + }
>> +
>> + /*
>> + * Another process is holding the global lock on the
>> + * sem_array; we cannot enter our critical section,
>> + * but have to wait for the global lock to be released.
>> + */
>> + if (unlikely(spin_is_locked(&sma->sem_perm.lock))) {
>> + spin_unlock(&sem->lock);
>> + goto again;
>
> This is IMO where the spin_unlock_wait(&sma->sem_perm.lock) would
> belong - right before the goto again.
That is where I had it initially. I may have gotten too clever
and worked on keeping more accesses read-only. If you want, I
can move it back here and re-submit the patch :)
> Also - I think there is a risk that an endless stream of complex
> semops could starve a simple semop here, as it would always find the
> sem_perm.lock to be locked ??? One easy way to guarantee progress
> would be to goto lock_array instead; however there is then the issue
> that a complex semop could force an endless stream of following simple
> semops to take the lock_array path. I'm not sure which of these
> problems is preferable to have...
If starvation turns out to be an issue, there is an even
simpler solution:
if (unlikely(spin_is_locked(&sma->sem_perm.lock))) {
spin_unlock(&sem->lock);
spin_lock(&sma->sem_perm.lock);
spin_lock(&sem->lock);
spin_unlock(&sma->sem_perm.lock);
}
Followed by unconditionally doing the critical section for
holding a single semaphore's lock, because we know that a
subsequent taker of sma->sem_perm.lock will either grab a
different semaphore's spinlock, or wait on the same semaphore's
spinlock, or wait for us to unlock our spinlock.
--
All rights reversed.
next prev parent reply other threads:[~2013-03-29 12:07 UTC|newest]
Thread overview: 129+ messages / expand[flat|nested] mbox.gz Atom feed top
2013-03-20 19:55 ipc,sem: sysv semaphore scalability Rik van Riel
2013-03-20 19:55 ` [PATCH 1/7] ipc: remove bogus lock comment for ipc_checkid Rik van Riel
2013-03-20 19:55 ` [PATCH 2/7] ipc: introduce obtaining a lockless ipc object Rik van Riel
2013-03-20 19:55 ` [PATCH 3/7] ipc: introduce lockless pre_down ipcctl Rik van Riel
2013-03-20 19:55 ` [PATCH 4/7] ipc,sem: do not hold ipc lock more than necessary Rik van Riel
2013-03-20 19:55 ` [PATCH 5/7] ipc,sem: open code and rename sem_lock Rik van Riel
2013-03-22 1:14 ` Davidlohr Bueso
2013-03-20 19:55 ` [PATCH 6/7] ipc,sem: have only one list in struct sem_queue Rik van Riel
2013-03-22 1:14 ` Davidlohr Bueso
2013-03-20 19:55 ` [PATCH 7/7] ipc,sem: fine grained locking for semtimedop Rik van Riel
2013-03-22 1:14 ` Davidlohr Bueso
2013-03-22 23:01 ` Michel Lespinasse
2013-03-22 23:38 ` Rik van Riel
2013-03-22 23:42 ` [PATCH 7/7 part3] fix for sem_lock Rik van Riel
2013-03-20 20:49 ` ipc,sem: sysv semaphore scalability Linus Torvalds
2013-03-20 20:56 ` Linus Torvalds
2013-03-20 20:57 ` Davidlohr Bueso
2013-03-21 21:10 ` Andrew Morton
2013-03-21 21:47 ` Peter Hurley
2013-03-21 21:50 ` Peter Hurley
2013-03-21 22:01 ` Andrew Morton
2013-03-22 3:38 ` Rik van Riel
2013-03-26 19:28 ` Dave Jones
2013-03-26 19:43 ` Andrew Morton
2013-03-29 16:17 ` Dave Jones
2013-03-29 18:00 ` Linus Torvalds
2013-03-29 18:04 ` Dave Jones
2013-03-29 18:10 ` Linus Torvalds
2013-03-29 18:43 ` Linus Torvalds
2013-03-29 19:06 ` Dave Jones
2013-03-29 19:13 ` Linus Torvalds
2013-03-29 19:26 ` Linus Torvalds
2013-03-29 19:36 ` Peter Hurley
2013-04-02 16:08 ` Sasha Levin
2013-04-02 17:24 ` Linus Torvalds
2013-04-02 17:52 ` Linus Torvalds
2013-04-02 19:53 ` Sasha Levin
2013-04-02 20:00 ` Dave Jones
2013-03-29 19:33 ` Peter Hurley
2013-03-29 19:54 ` Linus Torvalds
2013-04-01 7:40 ` Stanislav Kinsbursky
2013-03-29 20:41 ` Linus Torvalds
2013-03-29 21:12 ` Linus Torvalds
2013-03-29 23:16 ` Linus Torvalds
2013-03-30 1:36 ` Emmanuel Benisty
2013-03-30 2:08 ` Davidlohr Bueso
2013-03-30 3:02 ` Emmanuel Benisty
2013-03-30 3:46 ` Linus Torvalds
2013-03-30 4:33 ` Emmanuel Benisty
2013-03-30 5:10 ` Linus Torvalds
2013-03-30 5:57 ` Emmanuel Benisty
2013-03-30 17:22 ` Linus Torvalds
2013-03-31 2:38 ` Emmanuel Benisty
2013-03-31 5:01 ` Davidlohr Bueso
2013-03-31 13:45 ` Rik van Riel
2013-03-31 17:10 ` Linus Torvalds
2013-03-31 17:02 ` Emmanuel Benisty
2013-03-30 2:09 ` Linus Torvalds
2013-03-30 2:55 ` Davidlohr Bueso
2013-03-29 19:01 ` Dave Jones
2013-05-03 15:03 ` Peter Hurley
2013-03-22 1:12 ` Davidlohr Bueso
2013-03-22 1:23 ` Linus Torvalds
2013-03-22 3:40 ` Rik van Riel
2013-03-22 7:30 ` Mike Galbraith
2013-03-22 11:04 ` Emmanuel Benisty
2013-03-22 15:37 ` Linus Torvalds
2013-03-23 3:19 ` Emmanuel Benisty
2013-03-23 19:45 ` Linus Torvalds
2013-03-24 13:46 ` Emmanuel Benisty
2013-03-24 17:10 ` Linus Torvalds
2013-03-25 13:47 ` Emmanuel Benisty
2013-03-25 14:00 ` Rik van Riel
2013-03-25 14:03 ` Rik van Riel
2013-03-25 15:20 ` Emmanuel Benisty
2013-03-25 15:53 ` Rik van Riel
2013-03-25 17:09 ` Emmanuel Benisty
2013-03-25 14:01 ` Rik van Riel
2013-03-25 14:21 ` Emmanuel Benisty
2013-03-26 17:59 ` Davidlohr Bueso
2013-03-26 18:14 ` Rik van Riel
2013-03-26 18:35 ` Andrew Morton
2013-04-16 23:30 ` Andrew Morton
2013-05-04 15:55 ` Jörn Engel
2013-05-04 18:12 ` Borislav Petkov
2013-05-06 14:47 ` Jörn Engel
2013-03-22 17:51 ` Davidlohr Bueso
2013-03-25 20:21 ` Sasha Levin
2013-03-25 20:38 ` [PATCH -mm -next] ipc,sem: fix lockdep false positive Rik van Riel
2013-03-25 21:42 ` Michel Lespinasse
2013-03-25 21:51 ` Michel Lespinasse
2013-03-25 21:56 ` Sasha Levin
2013-03-25 21:52 ` Sasha Levin
2013-03-26 13:19 ` Peter Zijlstra
2013-03-26 13:40 ` Michel Lespinasse
2013-03-26 14:27 ` Peter Zijlstra
2013-03-26 15:19 ` Rik van Riel
2013-03-27 8:40 ` Peter Zijlstra
2013-03-27 8:42 ` Peter Zijlstra
2013-03-27 11:22 ` Michel Lespinasse
2013-03-27 12:02 ` Peter Zijlstra
2013-03-27 20:00 ` Rik van Riel
2013-03-28 20:23 ` [PATCH v2 " Rik van Riel
2013-03-29 2:50 ` Michel Lespinasse
2013-03-29 9:57 ` Peter Zijlstra
2013-03-29 13:21 ` Michel Lespinasse
2013-03-29 12:07 ` Rik van Riel [this message]
2013-03-29 13:08 ` Michel Lespinasse
2013-03-29 13:24 ` Rik van Riel
2013-03-29 13:55 ` [PATCH v3 " Rik van Riel
2013-03-29 13:59 ` Michel Lespinasse
2013-03-26 14:25 ` [PATCH " Rik van Riel
2013-03-26 17:33 ` ipc,sem: sysv semaphore scalability Sasha Levin
2013-03-26 17:51 ` Davidlohr Bueso
2013-03-26 18:07 ` Sasha Levin
2013-03-26 18:17 ` Rik van Riel
2013-03-26 20:00 ` [PATCH -mm -next] ipc,sem: untangle RCU locking with find_alloc_undo Rik van Riel
2013-04-05 4:38 ` Mike Galbraith
2013-04-05 13:21 ` Rik van Riel
2013-04-05 16:26 ` Mike Galbraith
2013-04-16 12:37 ` Mike Galbraith
2013-03-26 17:55 ` ipc,sem: sysv semaphore scalability Paul E. McKenney
2013-03-28 15:32 ` [PATCH -mm -next] ipc,sem: untangle RCU locking with find_alloc_undo Rik van Riel
2013-03-28 21:05 ` Davidlohr Bueso
2013-03-29 1:00 ` Michel Lespinasse
2013-03-29 1:14 ` Sasha Levin
2013-03-30 13:35 ` Sasha Levin
2013-03-31 1:30 ` Rik van Riel
2013-03-31 4:09 ` Davidlohr Bueso
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=51558407.5040005@surriel.com \
--to=riel@surriel.com \
--cc=akpm@linux-foundation.org \
--cc=benisty.e@gmail.com \
--cc=chegu_vinod@hp.com \
--cc=davej@redhat.com \
--cc=davidlohr.bueso@hp.com \
--cc=hhuang@redhat.com \
--cc=jason.low2@hp.com \
--cc=linux-kernel@vger.kernel.org \
--cc=lwoodman@redhat.com \
--cc=mingo@redhat.com \
--cc=peterz@infradead.org \
--cc=sasha.levin@oracle.com \
--cc=torvalds@linux-foundation.org \
--cc=walken@google.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
Powered by JetHome