Message ID | 1377872041-390-1-git-send-email-jbacik@fusionio.com (mailing list archive) |
---|---|
State | New, archived |
Headers | show |
On Fri, Aug 30, 2013 at 10:14:01AM -0400, Josef Bacik wrote: > Btrfs uses an rwsem to control access to its extent tree. Threads will hold a > read lock on this rwsem while they scan the extent tree, and if need_resched() > they will drop the lock and schedule. The transaction commit needs to take a > write lock for this rwsem for a very short period to switch out the commit > roots. If there are a lot of threads doing this caching operation we can starve > out the committers which slows everybody out. To address this we want to add > this functionality to see if our rwsem has anybody waiting to take a write lock > so we can drop it and schedule for a bit to allow the commit to continue. > +/* > + * check to see if the rwsem we're holding has anybody waiting to acquire it. > + */ > +int rwsem_is_contended(struct rw_semaphore *sem) > +{ > + int ret = 0; > + unsigned long flags; > + > + if (!raw_spin_trylock_irqsave(&sem->wait_lock, flags)) > + return 1; > + if (!list_empty(&sem->wait_list)) > + ret = 1; > + raw_spin_unlock_irqrestore(&sem->wait_lock, flags); > + return ret; > +} > + > +EXPORT_SYMBOL(rwsem_is_contended); Modeled after spin_is_contended(), so no problem with that. One thing I was wondering about is if it wants to be called rwsem_is_write_contended() or similar, since it explicitly only tests for pending writers. -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Hi Josef, On Fri, Aug 30, 2013 at 7:14 AM, Josef Bacik <jbacik@fusionio.com> wrote: > Btrfs uses an rwsem to control access to its extent tree. Threads will hold a > read lock on this rwsem while they scan the extent tree, and if need_resched() > they will drop the lock and schedule. The transaction commit needs to take a > write lock for this rwsem for a very short period to switch out the commit > roots. If there are a lot of threads doing this caching operation we can starve > out the committers which slows everybody out. To address this we want to add > this functionality to see if our rwsem has anybody waiting to take a write lock > so we can drop it and schedule for a bit to allow the commit to continue. > Thanks, > > Signed-off-by: Josef Bacik <jbacik@fusionio.com> FYI, I once tried to introduce something like this before, but my use case was pretty weak so it was not accepted at the time. I don't think there were any objections to the API itself though, and I think it's potentially a good idea if you use case justifies it. Two comments: - Note that there are two rwsem implementations - if you are going to add functionality to rwsem.h you want to add the same functionality in rwsem-spinlock.h as well. - I would prefer if you could avoid taking the wait_lock in your rwsem.h implementation. In your use case (read lock is known to be held), checking for sem->count < 0 would be sufficient to indicate a writer is queued (or getting onto the queue). In the general case, some architectures have the various values set up so that RWSEM_WAITING_BIAS != RWSEM_ACTIVE_WRITE_BIAS - for these architectures at least, you can check for waiters by looking if the lowest bit of RWSEM_WAITING_BIAS is set in sem->count.
On 09/01/2013 04:32 AM, Michel Lespinasse wrote: > Hi Josef, > > On Fri, Aug 30, 2013 at 7:14 AM, Josef Bacik <jbacik@fusionio.com> wrote: >> Btrfs uses an rwsem to control access to its extent tree. Threads will hold a >> read lock on this rwsem while they scan the extent tree, and if need_resched() >> they will drop the lock and schedule. The transaction commit needs to take a >> write lock for this rwsem for a very short period to switch out the commit >> roots. If there are a lot of threads doing this caching operation we can starve >> out the committers which slows everybody out. To address this we want to add >> this functionality to see if our rwsem has anybody waiting to take a write lock >> so we can drop it and schedule for a bit to allow the commit to continue. >> Thanks, >> >> Signed-off-by: Josef Bacik <jbacik@fusionio.com> > > FYI, I once tried to introduce something like this before, but my use > case was pretty weak so it was not accepted at the time. I don't think > there were any objections to the API itself though, and I think it's > potentially a good idea if you use case justifies it. Exactly, I'm concerned about the use case: readers can't starve writers. Of course, lots of existing readers can temporarily prevent a writer from acquiring, but those readers would already have the lock. Any new readers wouldn't be able to prevent a waiting writer from obtaining the lock. Josef, Could you be more explicit, maybe with some detailed numbers about the condition you report? I say that because a subtle bug that could mistakenly wait a reader existed in the rwsem implementation until relatively recently. Is there some other lurking problem? > Two comments: > > - Note that there are two rwsem implementations - if you are going to > add functionality to rwsem.h you want to add the same functionality in > rwsem-spinlock.h as well. > > - I would prefer if you could avoid taking the wait_lock in your > rwsem.h implementation. In your use case (read lock is known to be > held), checking for sem->count < 0 would be sufficient to indicate a > writer is queued (or getting onto the queue). In the general case, > some architectures have the various values set up so that > RWSEM_WAITING_BIAS != RWSEM_ACTIVE_WRITE_BIAS - for these > architectures at least, you can check for waiters by looking if the > lowest bit of RWSEM_WAITING_BIAS is set in sem->count. Michel, I'm glad you point out a much better approach --- but why are we considering open-coding down_read_trylock()/down_write_trylock? Regards, Peter Hurley -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Mon, Sep 02, 2013 at 01:18:08PM -0400, Peter Hurley wrote: > On 09/01/2013 04:32 AM, Michel Lespinasse wrote: > >Hi Josef, > > > >On Fri, Aug 30, 2013 at 7:14 AM, Josef Bacik <jbacik@fusionio.com> wrote: > >>Btrfs uses an rwsem to control access to its extent tree. Threads will hold a > >>read lock on this rwsem while they scan the extent tree, and if need_resched() > >>they will drop the lock and schedule. The transaction commit needs to take a > >>write lock for this rwsem for a very short period to switch out the commit > >>roots. If there are a lot of threads doing this caching operation we can starve > >>out the committers which slows everybody out. To address this we want to add > >>this functionality to see if our rwsem has anybody waiting to take a write lock > >>so we can drop it and schedule for a bit to allow the commit to continue. > >>Thanks, > >> > >>Signed-off-by: Josef Bacik <jbacik@fusionio.com> > > > >FYI, I once tried to introduce something like this before, but my use > >case was pretty weak so it was not accepted at the time. I don't think > >there were any objections to the API itself though, and I think it's > >potentially a good idea if you use case justifies it. > > Exactly, I'm concerned about the use case: readers can't starve writers. > Of course, lots of existing readers can temporarily prevent a writer from > acquiring, but those readers would already have the lock. Any new readers > wouldn't be able to prevent a waiting writer from obtaining the lock. > > Josef, > Could you be more explicit, maybe with some detailed numbers about the > condition you report? > Sure, this came from a community member http://article.gmane.org/gmane.comp.file-systems.btrfs/28081 With the old approach we could block between 1-2 seconds waiting for this rwsem, and with the new approach where we allow many more of these caching threads we were staving out the writer for 80 seconds. So what happens is these threads will scan our extent tree to put together the free space cache, and they'll hold this lock while they are doing the scanning. The only way they will drop this lock is if we hit need_resched(), but because these threads are going to do quite a bit of IO I imagine we're not ever being flagged with need_resched() because we schedule while waiting for IO. So these threads will hold onto this lock for bloody ever without giving it up so the committer can take the write lock. His patch to "fix" the problem was to have an atomic that let us know somebody was waiting for a write lock and then we'd drop the reader lock and schedule. So really we're just using a rwsem in a really mean way for writers. I'm open to other suggestions but I think this probably the cleanest way. Thanks, Josef -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Sun, Sep 01, 2013 at 01:32:36AM -0700, Michel Lespinasse wrote: > Hi Josef, > > On Fri, Aug 30, 2013 at 7:14 AM, Josef Bacik <jbacik@fusionio.com> wrote: > > Btrfs uses an rwsem to control access to its extent tree. Threads will hold a > > read lock on this rwsem while they scan the extent tree, and if need_resched() > > they will drop the lock and schedule. The transaction commit needs to take a > > write lock for this rwsem for a very short period to switch out the commit > > roots. If there are a lot of threads doing this caching operation we can starve > > out the committers which slows everybody out. To address this we want to add > > this functionality to see if our rwsem has anybody waiting to take a write lock > > so we can drop it and schedule for a bit to allow the commit to continue. > > Thanks, > > > > Signed-off-by: Josef Bacik <jbacik@fusionio.com> > > FYI, I once tried to introduce something like this before, but my use > case was pretty weak so it was not accepted at the time. I don't think > there were any objections to the API itself though, and I think it's > potentially a good idea if you use case justifies it. > > Two comments: > > - Note that there are two rwsem implementations - if you are going to > add functionality to rwsem.h you want to add the same functionality in > rwsem-spinlock.h as well. > Sure thing. > - I would prefer if you could avoid taking the wait_lock in your > rwsem.h implementation. In your use case (read lock is known to be > held), checking for sem->count < 0 would be sufficient to indicate a > writer is queued (or getting onto the queue). In the general case, > some architectures have the various values set up so that > RWSEM_WAITING_BIAS != RWSEM_ACTIVE_WRITE_BIAS - for these > architectures at least, you can check for waiters by looking if the > lowest bit of RWSEM_WAITING_BIAS is set in sem->count. Question about this one, I can't just do if (sem->count < 0) since each arch has their own atomic way of looking at count, so I'd have to add something to do just a normal read of count for each arch and call that wouldn't I? If that's what you want me to do then I'm fine with that (though I'll need a really thorough review), just want to double check before I make a bunch of extra work for myself. Thanks, Josef -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Sat, Aug 31, 2013 at 04:51:36PM +0200, Peter Zijlstra wrote: > On Fri, Aug 30, 2013 at 10:14:01AM -0400, Josef Bacik wrote: > > Btrfs uses an rwsem to control access to its extent tree. Threads will hold a > > read lock on this rwsem while they scan the extent tree, and if need_resched() > > they will drop the lock and schedule. The transaction commit needs to take a > > write lock for this rwsem for a very short period to switch out the commit > > roots. If there are a lot of threads doing this caching operation we can starve > > out the committers which slows everybody out. To address this we want to add > > this functionality to see if our rwsem has anybody waiting to take a write lock > > so we can drop it and schedule for a bit to allow the commit to continue. > > > > +/* > > + * check to see if the rwsem we're holding has anybody waiting to acquire it. > > + */ > > +int rwsem_is_contended(struct rw_semaphore *sem) > > +{ > > + int ret = 0; > > + unsigned long flags; > > + > > + if (!raw_spin_trylock_irqsave(&sem->wait_lock, flags)) > > + return 1; > > + if (!list_empty(&sem->wait_list)) > > + ret = 1; > > + raw_spin_unlock_irqrestore(&sem->wait_lock, flags); > > + return ret; > > +} > > + > > +EXPORT_SYMBOL(rwsem_is_contended); > > Modeled after spin_is_contended(), so no problem with that. One thing I > was wondering about is if it wants to be called > rwsem_is_write_contended() or similar, since it explicitly only tests > for pending writers. > Well it checks all pending waiters, the waiters list isn't split between readers waiters and write waiters, so people holding the write lock could call this to see if readers are waiting on the lock. Thanks, Josef -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On 09/03/2013 09:18 AM, Josef Bacik wrote: > On Mon, Sep 02, 2013 at 01:18:08PM -0400, Peter Hurley wrote: >> On 09/01/2013 04:32 AM, Michel Lespinasse wrote: >>> Hi Josef, >>> >>> On Fri, Aug 30, 2013 at 7:14 AM, Josef Bacik <jbacik@fusionio.com> wrote: >>>> Btrfs uses an rwsem to control access to its extent tree. Threads will hold a >>>> read lock on this rwsem while they scan the extent tree, and if need_resched() >>>> they will drop the lock and schedule. The transaction commit needs to take a >>>> write lock for this rwsem for a very short period to switch out the commit >>>> roots. If there are a lot of threads doing this caching operation we can starve >>>> out the committers which slows everybody out. To address this we want to add >>>> this functionality to see if our rwsem has anybody waiting to take a write lock >>>> so we can drop it and schedule for a bit to allow the commit to continue. >>>> Thanks, >>>> >>>> Signed-off-by: Josef Bacik <jbacik@fusionio.com> >>> >>> FYI, I once tried to introduce something like this before, but my use >>> case was pretty weak so it was not accepted at the time. I don't think >>> there were any objections to the API itself though, and I think it's >>> potentially a good idea if you use case justifies it. >> >> Exactly, I'm concerned about the use case: readers can't starve writers. >> Of course, lots of existing readers can temporarily prevent a writer from >> acquiring, but those readers would already have the lock. Any new readers >> wouldn't be able to prevent a waiting writer from obtaining the lock. >> >> Josef, >> Could you be more explicit, maybe with some detailed numbers about the >> condition you report? >> > > Sure, this came from a community member > > http://article.gmane.org/gmane.comp.file-systems.btrfs/28081 > > With the old approach we could block between 1-2 seconds waiting for this rwsem, > and with the new approach where we allow many more of these caching threads we > were staving out the writer for 80 seconds. > > So what happens is these threads will scan our extent tree to put together the > free space cache, and they'll hold this lock while they are doing the scanning. > The only way they will drop this lock is if we hit need_resched(), but because > these threads are going to do quite a bit of IO I imagine we're not ever being > flagged with need_resched() because we schedule while waiting for IO. So these > threads will hold onto this lock for bloody ever without giving it up so the > committer can take the write lock. His patch to "fix" the problem was to have > an atomic that let us know somebody was waiting for a write lock and then we'd > drop the reader lock and schedule. Thanks for the additional clarification. > So really we're just using a rwsem in a really mean way for writers. I'm open > to other suggestions but I think this probably the cleanest way. Is there substantial saved state at the point where the caching thread is checking need_resched() that precludes dropping and reacquiring the extent_commit_sem (or before find_next_key())? Not that it's a cleaner solution; just want to understand better the situation. Regards, Peter Hurley -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On 09/03/2013 11:47 AM, Josef Bacik wrote: > On Sun, Sep 01, 2013 at 01:32:36AM -0700, Michel Lespinasse wrote: >> Hi Josef, >> >> On Fri, Aug 30, 2013 at 7:14 AM, Josef Bacik <jbacik@fusionio.com> wrote: >>> Btrfs uses an rwsem to control access to its extent tree. Threads will hold a >>> read lock on this rwsem while they scan the extent tree, and if need_resched() >>> they will drop the lock and schedule. The transaction commit needs to take a >>> write lock for this rwsem for a very short period to switch out the commit >>> roots. If there are a lot of threads doing this caching operation we can starve >>> out the committers which slows everybody out. To address this we want to add >>> this functionality to see if our rwsem has anybody waiting to take a write lock >>> so we can drop it and schedule for a bit to allow the commit to continue. >>> Thanks, >>> >>> Signed-off-by: Josef Bacik <jbacik@fusionio.com> >> >> FYI, I once tried to introduce something like this before, but my use >> case was pretty weak so it was not accepted at the time. I don't think >> there were any objections to the API itself though, and I think it's >> potentially a good idea if you use case justifies it. >> >> Two comments: >> >> - Note that there are two rwsem implementations - if you are going to >> add functionality to rwsem.h you want to add the same functionality in >> rwsem-spinlock.h as well. >> > > Sure thing. > >> - I would prefer if you could avoid taking the wait_lock in your >> rwsem.h implementation. In your use case (read lock is known to be >> held), checking for sem->count < 0 would be sufficient to indicate a >> writer is queued (or getting onto the queue). In the general case, >> some architectures have the various values set up so that >> RWSEM_WAITING_BIAS != RWSEM_ACTIVE_WRITE_BIAS - for these >> architectures at least, you can check for waiters by looking if the >> lowest bit of RWSEM_WAITING_BIAS is set in sem->count. > > Question about this one, I can't just do > > if (sem->count < 0) > > since each arch has their own atomic way of looking at count, so I'd have to add > something to do just a normal read of count for each arch and call that wouldn't > I? Reading sem->count is atomic. For that matter, in your particular use case (which is more heuristic), you could perform the list_empty() check without acquiring the wait_lock. Regards, Peter Hurley -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Wed, Sep 04, 2013 at 07:46:56AM -0400, Peter Hurley wrote: > On 09/03/2013 09:18 AM, Josef Bacik wrote: > >On Mon, Sep 02, 2013 at 01:18:08PM -0400, Peter Hurley wrote: > >>On 09/01/2013 04:32 AM, Michel Lespinasse wrote: > >>>Hi Josef, > >>> > >>>On Fri, Aug 30, 2013 at 7:14 AM, Josef Bacik <jbacik@fusionio.com> wrote: > >>>>Btrfs uses an rwsem to control access to its extent tree. Threads will hold a > >>>>read lock on this rwsem while they scan the extent tree, and if need_resched() > >>>>they will drop the lock and schedule. The transaction commit needs to take a > >>>>write lock for this rwsem for a very short period to switch out the commit > >>>>roots. If there are a lot of threads doing this caching operation we can starve > >>>>out the committers which slows everybody out. To address this we want to add > >>>>this functionality to see if our rwsem has anybody waiting to take a write lock > >>>>so we can drop it and schedule for a bit to allow the commit to continue. > >>>>Thanks, > >>>> > >>>>Signed-off-by: Josef Bacik <jbacik@fusionio.com> > >>> > >>>FYI, I once tried to introduce something like this before, but my use > >>>case was pretty weak so it was not accepted at the time. I don't think > >>>there were any objections to the API itself though, and I think it's > >>>potentially a good idea if you use case justifies it. > >> > >>Exactly, I'm concerned about the use case: readers can't starve writers. > >>Of course, lots of existing readers can temporarily prevent a writer from > >>acquiring, but those readers would already have the lock. Any new readers > >>wouldn't be able to prevent a waiting writer from obtaining the lock. > >> > >>Josef, > >>Could you be more explicit, maybe with some detailed numbers about the > >>condition you report? > >> > > > >Sure, this came from a community member > > > >http://article.gmane.org/gmane.comp.file-systems.btrfs/28081 > > > >With the old approach we could block between 1-2 seconds waiting for this rwsem, > >and with the new approach where we allow many more of these caching threads we > >were staving out the writer for 80 seconds. > > > >So what happens is these threads will scan our extent tree to put together the > >free space cache, and they'll hold this lock while they are doing the scanning. > >The only way they will drop this lock is if we hit need_resched(), but because > >these threads are going to do quite a bit of IO I imagine we're not ever being > >flagged with need_resched() because we schedule while waiting for IO. So these > >threads will hold onto this lock for bloody ever without giving it up so the > >committer can take the write lock. His patch to "fix" the problem was to have > >an atomic that let us know somebody was waiting for a write lock and then we'd > >drop the reader lock and schedule. > > Thanks for the additional clarification. > > >So really we're just using a rwsem in a really mean way for writers. I'm open > >to other suggestions but I think this probably the cleanest way. > > Is there substantial saved state at the point where the caching thread is > checking need_resched() that precludes dropping and reacquiring the > extent_commit_sem (or before find_next_key())? Not that it's a cleaner > solution; just want to understand better the situation. > Yes I had thought of just dropping our locks everytime we had to do find_next_key() but that isn't going to work. We do have to save state but that's not the hard part, it's the fact that we could race with the committing transaction and lose space. So what would happen is something like this caching_thread: save last cached offset drop locks find_next_key get a ref on the current commit root search down to the next leaf re-take locks process leaf transaction committer: acquire locks swap commit root write transaction unpin all extents up to the last saved cached offset So if the caching thread grabs a ref on the commit root before the transaction committer swaps out the commit root we are dealing with too old of a tree. So say the leaf we're going to process next has data that was free'ed (and therefore would have been unpinned) during that transaction, because it is after our last cached offset we don't unpin it because we know that the caching thread is going to find the leaf where that extent is not there and add free space for that extent. However we got the leaf from two transactions ago which will still show that extent in use, so we won't add free space for it, which will cause us to leak the extent. We need to make sure we are always consistently on the previous extent root which is why we hold this lock. I hope that makes sense. Thanks, Josef -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Fri, 30 Aug 2013 10:14:01 -0400 Josef Bacik <jbacik@fusionio.com> wrote: > Btrfs uses an rwsem to control access to its extent tree. Threads will hold a > read lock on this rwsem while they scan the extent tree, and if need_resched() > they will drop the lock and schedule. The transaction commit needs to take a > write lock for this rwsem for a very short period to switch out the commit > roots. If there are a lot of threads doing this caching operation we can starve > out the committers which slows everybody out. To address this we want to add > this functionality to see if our rwsem has anybody waiting to take a write lock > so we can drop it and schedule for a bit to allow the commit to continue. > Thanks, > This sounds rather nasty and hacky. Rather then working around a locking shortcoming in a caller it would be better to fix/enhance the core locking code. What would such a change need to do? Presently rwsem waiters are fifo-queued, are they not? So the commit thread will eventually get that lock. Apparently that's not working adequately for you but I don't fully understand what it is about these dynamics which is causing observable problems. > I've cc'ed people who seemed like they may be in charge/familiar with this code, > hopefully I got the right people. > > include/linux/rwsem.h | 1 + > lib/rwsem.c | 17 +++++++++++++++++ This will break CONFIG_RWSEM_GENERIC_SPINLOCK=n? -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Mon, Sep 16, 2013 at 04:05:47PM -0700, Andrew Morton wrote: > On Fri, 30 Aug 2013 10:14:01 -0400 Josef Bacik <jbacik@fusionio.com> wrote: > > > Btrfs uses an rwsem to control access to its extent tree. Threads will hold a > > read lock on this rwsem while they scan the extent tree, and if need_resched() > > they will drop the lock and schedule. The transaction commit needs to take a > > write lock for this rwsem for a very short period to switch out the commit > > roots. If there are a lot of threads doing this caching operation we can starve > > out the committers which slows everybody out. To address this we want to add > > this functionality to see if our rwsem has anybody waiting to take a write lock > > so we can drop it and schedule for a bit to allow the commit to continue. > > Thanks, > > > > This sounds rather nasty and hacky. Rather then working around a > locking shortcoming in a caller it would be better to fix/enhance the > core locking code. What would such a change need to do? > > Presently rwsem waiters are fifo-queued, are they not? So the commit > thread will eventually get that lock. Apparently that's not working > adequately for you but I don't fully understand what it is about these > dynamics which is causing observable problems. > So the problem is not that its normal lock starvation, it's more our particular use case that is causing the starvation. We can have lots of people holding readers and simply never give them up for long periods of time, which is why we need this is_contended helper so we know to drop things and let the committer through. Thanks, Josef -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On 09/16/2013 05:05 PM, Josef Bacik wrote: > On Mon, Sep 16, 2013 at 04:05:47PM -0700, Andrew Morton wrote: >> On Fri, 30 Aug 2013 10:14:01 -0400 Josef Bacik <jbacik@fusionio.com> wrote: >> >>> Btrfs uses an rwsem to control access to its extent tree. Threads will hold a >>> read lock on this rwsem while they scan the extent tree, and if need_resched() >>> they will drop the lock and schedule. The transaction commit needs to take a >>> write lock for this rwsem for a very short period to switch out the commit >>> roots. If there are a lot of threads doing this caching operation we can starve >>> out the committers which slows everybody out. To address this we want to add >>> this functionality to see if our rwsem has anybody waiting to take a write lock >>> so we can drop it and schedule for a bit to allow the commit to continue. >>> Thanks, >>> >> >> This sounds rather nasty and hacky. Rather then working around a >> locking shortcoming in a caller it would be better to fix/enhance the >> core locking code. What would such a change need to do? >> >> Presently rwsem waiters are fifo-queued, are they not? So the commit >> thread will eventually get that lock. Apparently that's not working >> adequately for you but I don't fully understand what it is about these >> dynamics which is causing observable problems. >> > > So the problem is not that its normal lock starvation, it's more our particular > use case that is causing the starvation. We can have lots of people holding > readers and simply never give them up for long periods of time, which is why we > need this is_contended helper so we know to drop things and let the committer > through. Thanks, You could easily achieve the same thing by putting an "is_contending" flag in parallel with the rwsem and testing that: DECLARE_RWSEM(foo); atomic_t is_contended = ATOMIC_INIT(0); . . . /* writing context */ atomic_inc(&is_contended); down_write(&foo); do_writing_action(); up_write(&foo); atomic_dec(&is_contended); /* reading context */ down_read(&foo); while (!atomic_read(&is_contended)) do_reading_actions(); up_read(&foo); David Daney -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On 09/16/2013 08:29 PM, David Daney wrote: > On 09/16/2013 05:05 PM, Josef Bacik wrote: >> On Mon, Sep 16, 2013 at 04:05:47PM -0700, Andrew Morton wrote: >>> On Fri, 30 Aug 2013 10:14:01 -0400 Josef Bacik <jbacik@fusionio.com> wrote: >>> >>>> Btrfs uses an rwsem to control access to its extent tree. Threads will hold a >>>> read lock on this rwsem while they scan the extent tree, and if need_resched() >>>> they will drop the lock and schedule. The transaction commit needs to take a >>>> write lock for this rwsem for a very short period to switch out the commit >>>> roots. If there are a lot of threads doing this caching operation we can starve >>>> out the committers which slows everybody out. To address this we want to add >>>> this functionality to see if our rwsem has anybody waiting to take a write lock >>>> so we can drop it and schedule for a bit to allow the commit to continue. >>>> Thanks, >>>> >>> >>> This sounds rather nasty and hacky. Rather then working around a >>> locking shortcoming in a caller it would be better to fix/enhance the >>> core locking code. What would such a change need to do? >>> >>> Presently rwsem waiters are fifo-queued, are they not? So the commit >>> thread will eventually get that lock. Apparently that's not working >>> adequately for you but I don't fully understand what it is about these >>> dynamics which is causing observable problems. >>> >> >> So the problem is not that its normal lock starvation, it's more our particular >> use case that is causing the starvation. We can have lots of people holding >> readers and simply never give them up for long periods of time, which is why we >> need this is_contended helper so we know to drop things and let the committer >> through. Thanks, > > You could easily achieve the same thing by putting an "is_contending" flag in parallel with the rwsem and testing that: Which adds a bunch more bus-locked operations to contended over, when a unlocked if (list_empty()) is sufficient. Regards, Peter Hurley -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On 09/16/2013 05:37 PM, Peter Hurley wrote: > On 09/16/2013 08:29 PM, David Daney wrote: >> On 09/16/2013 05:05 PM, Josef Bacik wrote: >>> On Mon, Sep 16, 2013 at 04:05:47PM -0700, Andrew Morton wrote: >>>> On Fri, 30 Aug 2013 10:14:01 -0400 Josef Bacik <jbacik@fusionio.com> >>>> wrote: >>>> >>>>> Btrfs uses an rwsem to control access to its extent tree. Threads >>>>> will hold a >>>>> read lock on this rwsem while they scan the extent tree, and if >>>>> need_resched() >>>>> they will drop the lock and schedule. The transaction commit needs >>>>> to take a >>>>> write lock for this rwsem for a very short period to switch out the >>>>> commit >>>>> roots. If there are a lot of threads doing this caching operation >>>>> we can starve >>>>> out the committers which slows everybody out. To address this we >>>>> want to add >>>>> this functionality to see if our rwsem has anybody waiting to take >>>>> a write lock >>>>> so we can drop it and schedule for a bit to allow the commit to >>>>> continue. >>>>> Thanks, >>>>> >>>> >>>> This sounds rather nasty and hacky. Rather then working around a >>>> locking shortcoming in a caller it would be better to fix/enhance the >>>> core locking code. What would such a change need to do? >>>> >>>> Presently rwsem waiters are fifo-queued, are they not? So the commit >>>> thread will eventually get that lock. Apparently that's not working >>>> adequately for you but I don't fully understand what it is about these >>>> dynamics which is causing observable problems. >>>> >>> >>> So the problem is not that its normal lock starvation, it's more our >>> particular >>> use case that is causing the starvation. We can have lots of people >>> holding >>> readers and simply never give them up for long periods of time, which >>> is why we >>> need this is_contended helper so we know to drop things and let the >>> committer >>> through. Thanks, >> >> You could easily achieve the same thing by putting an "is_contending" >> flag in parallel with the rwsem and testing that: > > Which adds a bunch more bus-locked operations to contended over Would that be a problem in this particular case? Has it been measured? > , when > a unlocked if (list_empty()) is sufficient. I don't object to adding rwsem_is_contended() *if* it is required. I was just pointing out that there may be other options. The patch adds a bunch of new semantics to rwsem. There is a trade off between increased complexity of core code, and generalizing subsystem specific optimizations that may not be globally useful. Is it worth it in this case? I do not know. <insert quote relating to occam's razor> David Daney -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Mon, Sep 16, 2013 at 06:08:42PM -0700, David Daney wrote: > On 09/16/2013 05:37 PM, Peter Hurley wrote: > >On 09/16/2013 08:29 PM, David Daney wrote: > >>On 09/16/2013 05:05 PM, Josef Bacik wrote: > >>>On Mon, Sep 16, 2013 at 04:05:47PM -0700, Andrew Morton wrote: > >>>>On Fri, 30 Aug 2013 10:14:01 -0400 Josef Bacik <jbacik@fusionio.com> > >>>>wrote: > >>>> > >>>>>Btrfs uses an rwsem to control access to its extent tree. Threads > >>>>>will hold a > >>>>>read lock on this rwsem while they scan the extent tree, and if > >>>>>need_resched() > >>>>>they will drop the lock and schedule. The transaction commit needs > >>>>>to take a > >>>>>write lock for this rwsem for a very short period to switch out the > >>>>>commit > >>>>>roots. If there are a lot of threads doing this caching operation > >>>>>we can starve > >>>>>out the committers which slows everybody out. To address this we > >>>>>want to add > >>>>>this functionality to see if our rwsem has anybody waiting to take > >>>>>a write lock > >>>>>so we can drop it and schedule for a bit to allow the commit to > >>>>>continue. > >>>>>Thanks, > >>>>> > >>>> > >>>>This sounds rather nasty and hacky. Rather then working around a > >>>>locking shortcoming in a caller it would be better to fix/enhance the > >>>>core locking code. What would such a change need to do? > >>>> > >>>>Presently rwsem waiters are fifo-queued, are they not? So the commit > >>>>thread will eventually get that lock. Apparently that's not working > >>>>adequately for you but I don't fully understand what it is about these > >>>>dynamics which is causing observable problems. > >>>> > >>> > >>>So the problem is not that its normal lock starvation, it's more our > >>>particular > >>>use case that is causing the starvation. We can have lots of people > >>>holding > >>>readers and simply never give them up for long periods of time, which > >>>is why we > >>>need this is_contended helper so we know to drop things and let the > >>>committer > >>>through. Thanks, > >> > >>You could easily achieve the same thing by putting an "is_contending" > >>flag in parallel with the rwsem and testing that: > > > >Which adds a bunch more bus-locked operations to contended over > > Would that be a problem in this particular case? Has it been measured? > > >, when > >a unlocked if (list_empty()) is sufficient. > > I don't object to adding rwsem_is_contended() *if* it is required. I was > just pointing out that there may be other options. > > The patch adds a bunch of new semantics to rwsem. There is a trade off > between increased complexity of core code, and generalizing subsystem > specific optimizations that may not be globally useful. > > Is it worth it in this case? I do not know. > So what you suggested is actually what we did in order to prove that this was what the problem was. I'm ok with continuing to do that, I just figured adding something like rwsem_is_contended() would be nice in case anybody else runs into the issue in the future, plus it would save me an atomic_t in an already large structure. Thanks, Josef -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On 09/16/2013 09:11 PM, Josef Bacik wrote: > On Mon, Sep 16, 2013 at 06:08:42PM -0700, David Daney wrote: >> On 09/16/2013 05:37 PM, Peter Hurley wrote: >>> On 09/16/2013 08:29 PM, David Daney wrote: >>>> On 09/16/2013 05:05 PM, Josef Bacik wrote: >>>>> On Mon, Sep 16, 2013 at 04:05:47PM -0700, Andrew Morton wrote: >>>>>> On Fri, 30 Aug 2013 10:14:01 -0400 Josef Bacik <jbacik@fusionio.com> >>>>>> wrote: >>>>>> >>>>>>> Btrfs uses an rwsem to control access to its extent tree. Threads >>>>>>> will hold a >>>>>>> read lock on this rwsem while they scan the extent tree, and if >>>>>>> need_resched() >>>>>>> they will drop the lock and schedule. The transaction commit needs >>>>>>> to take a >>>>>>> write lock for this rwsem for a very short period to switch out the >>>>>>> commit >>>>>>> roots. If there are a lot of threads doing this caching operation >>>>>>> we can starve >>>>>>> out the committers which slows everybody out. To address this we >>>>>>> want to add >>>>>>> this functionality to see if our rwsem has anybody waiting to take >>>>>>> a write lock >>>>>>> so we can drop it and schedule for a bit to allow the commit to >>>>>>> continue. >>>>>>> Thanks, >>>>>>> >>>>>> >>>>>> This sounds rather nasty and hacky. Rather then working around a >>>>>> locking shortcoming in a caller it would be better to fix/enhance the >>>>>> core locking code. What would such a change need to do? >>>>>> >>>>>> Presently rwsem waiters are fifo-queued, are they not? So the commit >>>>>> thread will eventually get that lock. Apparently that's not working >>>>>> adequately for you but I don't fully understand what it is about these >>>>>> dynamics which is causing observable problems. >>>>>> >>>>> >>>>> So the problem is not that its normal lock starvation, it's more our >>>>> particular >>>>> use case that is causing the starvation. We can have lots of people >>>>> holding >>>>> readers and simply never give them up for long periods of time, which >>>>> is why we >>>>> need this is_contended helper so we know to drop things and let the >>>>> committer >>>>> through. Thanks, >>>> >>>> You could easily achieve the same thing by putting an "is_contending" >>>> flag in parallel with the rwsem and testing that: >>> >>> Which adds a bunch more bus-locked operations to contended over >> >> Would that be a problem in this particular case? Has it been measured? >> >>> , when >>> a unlocked if (list_empty()) is sufficient. >> >> I don't object to adding rwsem_is_contended() *if* it is required. I was >> just pointing out that there may be other options. >> >> The patch adds a bunch of new semantics to rwsem. There is a trade off >> between increased complexity of core code, and generalizing subsystem >> specific optimizations that may not be globally useful. >> >> Is it worth it in this case? I do not know. >> > > So what you suggested is actually what we did in order to prove that this was > what the problem was. I'm ok with continuing to do that, I just figured adding > something like rwsem_is_contended() would be nice in case anybody else runs into > the issue in the future, plus it would save me an atomic_t in an already large > structure. I saw the original patch you linked to earlier in the discussion, and I agree that for your use case adding a contention test is cleaner and clearer than other options. That said, I think this extension is only useful for readers: writers should be getting their business done and releasing the sem. Also, I think the comment above the function should be clearer that the lock must already be held by the caller; IOW, this is not a trylock replacement. Regards, Peter Hurley -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
* Andrew Morton <akpm@linux-foundation.org> wrote: > On Fri, 30 Aug 2013 10:14:01 -0400 Josef Bacik <jbacik@fusionio.com> wrote: > > > Btrfs uses an rwsem to control access to its extent tree. Threads > > will hold a read lock on this rwsem while they scan the extent tree, > > and if need_resched() they will drop the lock and schedule. The > > transaction commit needs to take a write lock for this rwsem for a > > very short period to switch out the commit roots. If there are a lot > > of threads doing this caching operation we can starve out the > > committers which slows everybody out. To address this we want to add > > this functionality to see if our rwsem has anybody waiting to take a > > write lock so we can drop it and schedule for a bit to allow the > > commit to continue. Thanks, > > This sounds rather nasty and hacky. Rather then working around a > locking shortcoming in a caller it would be better to fix/enhance the > core locking code. What would such a change need to do? > > Presently rwsem waiters are fifo-queued, are they not? So the commit > thread will eventually get that lock. Apparently that's not working > adequately for you but I don't fully understand what it is about these > dynamics which is causing observable problems. It would be nice to see the whole solution, together with the btrfs patch. The problem I have is that this new primitive is only superficially like spin_is_contended(): in the spinlock case dropping the lock will guarantee some sort of progress, because another CPU will almost certainly pick up the lock if we cpu_relax(). In the rwsem case there's no such guarantee of progress, especially if a read-lock is dropped. So I'd like to see how it's implemented in practice. Thanks, Ingo -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/include/linux/rwsem.h b/include/linux/rwsem.h index 0616ffe..61a8802 100644 --- a/include/linux/rwsem.h +++ b/include/linux/rwsem.h @@ -35,6 +35,7 @@ extern struct rw_semaphore *rwsem_down_read_failed(struct rw_semaphore *sem); extern struct rw_semaphore *rwsem_down_write_failed(struct rw_semaphore *sem); extern struct rw_semaphore *rwsem_wake(struct rw_semaphore *); extern struct rw_semaphore *rwsem_downgrade_wake(struct rw_semaphore *sem); +extern int rwsem_is_contended(struct rw_semaphore *sem); /* Include the arch specific part */ #include <asm/rwsem.h> diff --git a/lib/rwsem.c b/lib/rwsem.c index 19c5fa9..20858cd 100644 --- a/lib/rwsem.c +++ b/lib/rwsem.c @@ -287,6 +287,23 @@ struct rw_semaphore *rwsem_downgrade_wake(struct rw_semaphore *sem) return sem; } +/* + * check to see if the rwsem we're holding has anybody waiting to acquire it. + */ +int rwsem_is_contended(struct rw_semaphore *sem) +{ + int ret = 0; + unsigned long flags; + + if (!raw_spin_trylock_irqsave(&sem->wait_lock, flags)) + return 1; + if (!list_empty(&sem->wait_list)) + ret = 1; + raw_spin_unlock_irqrestore(&sem->wait_lock, flags); + return ret; +} + +EXPORT_SYMBOL(rwsem_is_contended); EXPORT_SYMBOL(rwsem_down_read_failed); EXPORT_SYMBOL(rwsem_down_write_failed); EXPORT_SYMBOL(rwsem_wake);
Btrfs uses an rwsem to control access to its extent tree. Threads will hold a read lock on this rwsem while they scan the extent tree, and if need_resched() they will drop the lock and schedule. The transaction commit needs to take a write lock for this rwsem for a very short period to switch out the commit roots. If there are a lot of threads doing this caching operation we can starve out the committers which slows everybody out. To address this we want to add this functionality to see if our rwsem has anybody waiting to take a write lock so we can drop it and schedule for a bit to allow the commit to continue. Thanks, Signed-off-by: Josef Bacik <jbacik@fusionio.com> --- I've cc'ed people who seemed like they may be in charge/familiar with this code, hopefully I got the right people. include/linux/rwsem.h | 1 + lib/rwsem.c | 17 +++++++++++++++++ 2 files changed, 18 insertions(+), 0 deletions(-)