LKML Archive on lore.kernel.org
 help / color / Atom feed
From: NeilBrown <neilb@suse.com>
To: Jeff Layton <jlayton@kernel.org>,
	Alexander Viro <viro@zeniv.linux.org.uk>
Cc: "J. Bruce Fields" <bfields@fieldses.org>,
	Martin Wilck <mwilck@suse.de>,
	linux-fsdevel@vger.kernel.org,
	Frank Filz <ffilzlnx@mindspring.com>,
	linux-kernel@vger.kernel.org
Subject: Re: [PATCH 5/5] fs/locks: create a tree of dependent requests.
Date: Fri, 10 Aug 2018 09:25:04 +1000
Message-ID: <87ftznrulr.fsf@notabene.neil.brown.name> (raw)
In-Reply-To: <20411e8edbc29aa45599b408075548faa9a5b904.camel@kernel.org>

[-- Attachment #1: Type: text/plain, Size: 7806 bytes --]

On Thu, Aug 09 2018, Jeff Layton wrote:

> On Thu, 2018-08-09 at 12:04 +1000, NeilBrown wrote:
>> When we find an existing lock which conflicts with a request,
>> and the request wants to wait, we currently add the request
>> to a list.  When the lock is removed, the whole list is woken.
>> This can cause the thundering-herd problem.
>> To reduce the problem, we make use of the (new) fact that
>> a pending request can itself have a list of blocked requests.
>> When we find a conflict, we look through the existing blocked requests.
>> If any one of them blocks the new request, the new request is attached
>> below that request.
>> This way, when the lock is released, only a set of non-conflicting
>> locks will be woken.  The rest of the herd can stay asleep.
>> 
>> Reported-and-tested-by: Martin Wilck <mwilck@suse.de>
>> Signed-off-by: NeilBrown <neilb@suse.com>
>> ---
>>  fs/locks.c |   69 +++++++++++++++++++++++++++++++++++++++++++++++++++++++-----
>>  1 file changed, 63 insertions(+), 6 deletions(-)
>> 
>> diff --git a/fs/locks.c b/fs/locks.c
>> index fc64016d01ee..17843feb6f5b 100644
>> --- a/fs/locks.c
>> +++ b/fs/locks.c
>> @@ -738,6 +738,39 @@ static void locks_delete_block(struct file_lock *waiter)
>>  	spin_unlock(&blocked_lock_lock);
>>  }
>>  
>> +static void wake_non_conflicts(struct file_lock *waiter, struct file_lock *blocker,
>> +			       enum conflict conflict(struct file_lock *,
>> +						      struct file_lock *))
>> +{
>> +	struct file_lock *parent = waiter;
>> +	struct file_lock *fl;
>> +	struct file_lock  *t;
>> +
>> +	fl = list_entry(&parent->fl_blocked, struct file_lock, fl_block);
>> +restart:
>> +	list_for_each_entry_safe_continue(fl, t, &parent->fl_blocked, fl_block) {
>> +		switch (conflict(fl, blocker)) {
>> +		default:
>
> BUG or WARN here too please.

Maybe .... I'd rather not have the default case at all.
I can remove this one, but if I remove the next one, gcc complains
../fs/locks.c: In function ‘posix_locks_conflict’:
../fs/locks.c:912:1: warning: control reaches end of non-void function [-Wreturn-type]

event though control cannot reach the end of the function.
Maybe:
	switch (locks_conflict(caller_fl, sys_fl)) {
	default:
		WARN(1, "locks_conflict returned impossible value");
		/* fallthrough */
	case FL_NO_CONFLICT:


>
>> +		case FL_NO_CONFLICT:
>> +			__locks_wake_one(fl);
>> +			break;
>> +		case FL_CONFLICT:
>> +			/* Need to check children */
>> +			parent = fl;
>> +			fl = list_entry(&parent->fl_blocked, struct file_lock, fl_block);
>> +			goto restart;
>> +		case FL_TRANSITIVE_CONFLICT:
>> +			/* all children must also conflict, no need to check */
>> +			continue;
>> +		}
>> +	}
>> +	if (parent != waiter) {
>> +		parent = parent->fl_blocker;
>> +		fl = parent;
>> +		goto restart;
>> +	}
>> +}
>> +
>>  /* Insert waiter into blocker's block list.
>>   * We use a circular list so that processes can be easily woken up in
>>   * the order they blocked. The documentation doesn't require this but
>> @@ -747,11 +780,32 @@ static void locks_delete_block(struct file_lock *waiter)
>>   * fl_blocked list itself is protected by the blocked_lock_lock, but by ensuring
>>   * that the flc_lock is also held on insertions we can avoid taking the
>>   * blocked_lock_lock in some cases when we see that the fl_blocked list is empty.
>> + *
>> + * Rather than just adding to the list, we check for conflicts with any existing
>> + * waiter, and add to that waiter instead.
>> + * Thus wakeups don't happen until needed.
>>   */
>>  static void __locks_insert_block(struct file_lock *blocker,
>> -					struct file_lock *waiter)
>> +				 struct file_lock *waiter,
>> +				 enum conflict conflict(struct file_lock *,
>> +							struct file_lock *))
>>  {
>> +	struct file_lock *fl;
>>  	BUG_ON(!list_empty(&waiter->fl_block));
>> +
>> +	/* Any request in waiter->fl_blocked is know to conflict with
>
> "known"
>
>> +	 * waiter, but it might not conflict with blocker.
>> +	 * If it doesn't, it needs to be woken now so it can find
>> +	 * somewhere else to wait, or possible it can get granted.
>
> "possibly it can be"

Both fixed, thanks.

>
>> +	 */
>> +	if (conflict(waiter, blocker) != FL_TRANSITIVE_CONFLICT)
>> +		wake_non_conflicts(waiter, blocker, conflict);
>> +new_blocker:
>> +	list_for_each_entry(fl, &blocker->fl_blocked, fl_block)
>> +		if (conflict(fl, waiter)) {
>> +			blocker =  fl;
>> +			goto new_blocker;
>> +		}
>>  
>> > 	waiter->fl_blocker = blocker;
>>  	list_add_tail(&waiter->fl_block, &blocker->fl_blocked);
>>  	if (IS_POSIX(blocker) && !IS_OFDLCK(blocker))
>
> I wonder if it might be better to insert the blocker first before waking
> up other waiters? Consider that anything awoken will end up contending
> for the flc_lock that is held by "current" at this point. Doing most of
> what you need to get done before waking them might mean less spinning in
> other tasks.
>

Maybe.
I think you are suggesting we move the call to wake_non_conflicts() to
the end of the function.
The main reason I put it at the top is to use the original value of
"blocker" before it gets changed.
Even if we move it to the end, there is still quite a few other little
tasks to be performed before the lock is dropped.

Will all this get done before some other processor has a chance to wake
up a process, and for that process to get a to spinlock ???

Maybe - though the first spinlock would be blocked_lock_lock in
locks_delete_block(), and that is dropped almost immediately.

I don't know ... it seems much of a muchness.
If the process will be woken that quickly, then some other processor
must be idle, and does it matter much if it spends a microsecond
spinning on a lock rather than being idle a bit longer?

Thanks.
I won't to do a least a little testing before I repost with any of these
changes.

NeilBrown


>> @@ -760,10 +814,12 @@ static void __locks_insert_block(struct file_lock *blocker,
>>  
>>  /* Must be called with flc_lock held. */
>>  static void locks_insert_block(struct file_lock *blocker,
>> -					struct file_lock *waiter)
>> +			       struct file_lock *waiter,
>> +			       enum conflict conflict(struct file_lock *,
>> +						      struct file_lock *))
>>  {
>>  	spin_lock(&blocked_lock_lock);
>> -	__locks_insert_block(blocker, waiter);
>> +	__locks_insert_block(blocker, waiter, conflict);
>>  	spin_unlock(&blocked_lock_lock);
>>  }
>>  
>> @@ -1033,7 +1089,7 @@ static int flock_lock_inode(struct inode *inode, struct file_lock *request)
>>  		if (!(request->fl_flags & FL_SLEEP))
>>  			goto out;
>>  		error = FILE_LOCK_DEFERRED;
>> -		locks_insert_block(fl, request);
>> +		locks_insert_block(fl, request, flock_locks_conflict);
>>  		goto out;
>>  	}
>>  	if (request->fl_flags & FL_ACCESS)
>> @@ -1107,7 +1163,8 @@ static int posix_lock_inode(struct inode *inode, struct file_lock *request,
>>  			spin_lock(&blocked_lock_lock);
>>  			if (likely(!posix_locks_deadlock(request, fl))) {
>>  				error = FILE_LOCK_DEFERRED;
>> -				__locks_insert_block(fl, request);
>> +				__locks_insert_block(fl, request,
>> +						     posix_locks_conflict);
>>  			}
>>  			spin_unlock(&blocked_lock_lock);
>>  			goto out;
>> @@ -1581,7 +1638,7 @@ int __break_lease(struct inode *inode, unsigned int mode, unsigned int type)
>>  		break_time -= jiffies;
>>  	if (break_time == 0)
>>  		break_time++;
>> -	locks_insert_block(fl, new_fl);
>> +	locks_insert_block(fl, new_fl, leases_conflict);
>>  	trace_break_lease_block(inode, new_fl);
>>  	spin_unlock(&ctx->flc_lock);
>>  	percpu_up_read_preempt_enable(&file_rwsem);
>> 
>> 
>
> -- 
> Jeff Layton <jlayton@kernel.org>

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 832 bytes --]

  reply index

Thread overview: 28+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2018-08-09  2:04 [PATCH 0/5 - V2] locks: avoid thundering-herd wake-ups NeilBrown
2018-08-09  2:04 ` [PATCH 1/5] fs/locks: rename some lists and pointers NeilBrown
2018-08-09  2:04 ` [PATCH 2/5] fs/locks: allow a lock request to block other requests NeilBrown
2018-08-09  2:04 ` [PATCH 3/5] fs/locks: change all *_conflict() functions to return a new enum NeilBrown
2018-08-09 11:09   ` Jeff Layton
2018-08-09 13:09   ` J. Bruce Fields
2018-08-09 23:40     ` NeilBrown
2018-08-10  0:56       ` J. Bruce Fields
2018-08-09  2:04 ` [PATCH 4/5] fs/locks: split out __locks_wake_one() NeilBrown
2018-08-09  2:04 ` [PATCH 5/5] fs/locks: create a tree of dependent requests NeilBrown
2018-08-09 11:17   ` Jeff Layton
2018-08-09 23:25     ` NeilBrown [this message]
2018-08-09 14:13   ` J. Bruce Fields
2018-08-09 22:19     ` NeilBrown
2018-08-10  0:36       ` J. Bruce Fields
2018-08-09 17:32 ` [PATCH 0/5 - V2] locks: avoid thundering-herd wake-ups J. Bruce Fields
2018-08-09 22:12   ` NeilBrown
2018-08-10  0:29     ` J. Bruce Fields
2018-08-10  1:50       ` NeilBrown
2018-08-10  2:52         ` J. Bruce Fields
2018-08-10  3:17           ` NeilBrown
2018-08-10 15:47             ` J. Bruce Fields
2018-08-11 11:56               ` Jeff Layton
2018-08-11 12:35                 ` J. Bruce Fields
2018-08-11 11:51       ` Jeff Layton
2018-08-11 12:21         ` J. Bruce Fields
2018-08-11 13:15           ` Jeff Layton
2018-08-14  3:56 [PATCH 0/5 v2] " NeilBrown
2018-08-14  3:56 ` [PATCH 5/5] fs/locks: create a tree of dependent requests NeilBrown

Reply instructions:

You may reply publically 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=87ftznrulr.fsf@notabene.neil.brown.name \
    --to=neilb@suse.com \
    --cc=bfields@fieldses.org \
    --cc=ffilzlnx@mindspring.com \
    --cc=jlayton@kernel.org \
    --cc=linux-fsdevel@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=mwilck@suse.de \
    --cc=viro@zeniv.linux.org.uk \
    /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

LKML Archive on lore.kernel.org

Archives are clonable:
	git clone --mirror https://lore.kernel.org/lkml/0 lkml/git/0.git
	git clone --mirror https://lore.kernel.org/lkml/1 lkml/git/1.git
	git clone --mirror https://lore.kernel.org/lkml/2 lkml/git/2.git
	git clone --mirror https://lore.kernel.org/lkml/3 lkml/git/3.git
	git clone --mirror https://lore.kernel.org/lkml/4 lkml/git/4.git
	git clone --mirror https://lore.kernel.org/lkml/5 lkml/git/5.git
	git clone --mirror https://lore.kernel.org/lkml/6 lkml/git/6.git
	git clone --mirror https://lore.kernel.org/lkml/7 lkml/git/7.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 lkml lkml/ https://lore.kernel.org/lkml \
		linux-kernel@vger.kernel.org linux-kernel@archiver.kernel.org
	public-inbox-index lkml


Newsgroup available over NNTP:
	nntp://nntp.lore.kernel.org/org.kernel.vger.linux-kernel


AGPL code for this site: git clone https://public-inbox.org/ public-inbox