[PATCH] add rb_next in block throttle

[PATCH] add rb_next in block throttle

[Hillf Danton]

When selecting group for dispatching, the first entry of the service
tree is sequentially dequeued, then the first entry is recomputed by
calling rb_first().

When the first is removed from service tree, new first could also be
computed with rb_next(), since it could be faster than rb_first in
this special environment.

Signed-off-by: Hillf Danton <dhillf@gmail.com>
---

--- a/block/blk-throttle.c	2010-11-01 19:54:12.000000000 +0800
+++ b/block/blk-throttle.c	2010-11-26 21:27:40.000000000 +0800
@@ -251,7 +251,7 @@ static void rb_erase_init(struct rb_node
 static void throtl_rb_erase(struct rb_node *n, struct throtl_rb_root *root)
 {
 	if (root->left == n)
-		root->left = NULL;
+		root->left = rb_next(n);
 	rb_erase_init(n, &root->rb);
 	--root->count;
 }

Re: [PATCH] add rb_next in block throttle

[Vivek Goyal]

On Fri, Nov 26, 2010 at 10:06:27PM +0800, Hillf Danton wrote:
> When selecting group for dispatching, the first entry of the service
> tree is sequentially dequeued, then the first entry is recomputed by
> calling rb_first().
> 
> When the first is removed from service tree, new first could also be
> computed with rb_next(), since it could be faster than rb_first in
> this special environment.
> 
> Signed-off-by: Hillf Danton <dhillf@gmail.com>
> ---

Ok. So set the leftmost element with the help of rb_next() at the time of
deletion of existing leftmost element. Not sure whether it is cheaper or
not but does not harm doing it this way.

Have you tested it to make sure nothing is broken?

Thanks
Vivek

> 
> --- a/block/blk-throttle.c	2010-11-01 19:54:12.000000000 +0800
> +++ b/block/blk-throttle.c	2010-11-26 21:27:40.000000000 +0800
> @@ -251,7 +251,7 @@ static void rb_erase_init(struct rb_node
>  static void throtl_rb_erase(struct rb_node *n, struct throtl_rb_root *root)
>  {
>  	if (root->left == n)
> -		root->left = NULL;
> +		root->left = rb_next(n);
>  	rb_erase_init(n, &root->rb);
>  	--root->count;
>  }
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/

Re: [PATCH] add rb_next in block throttle

[Hillf Danton]

On Tue, Nov 30, 2010 at 10:36 PM, Vivek Goyal <vgoyal@redhat.com> wrote:
> On Fri, Nov 26, 2010 at 10:06:27PM +0800, Hillf Danton wrote:
>> When selecting group for dispatching, the first entry of the service
>> tree is sequentially dequeued, then the first entry is recomputed by
>> calling rb_first().
>>
>> When the first is removed from service tree, new first could also be
>> computed with rb_next(), since it could be faster than rb_first in
>> this special environment.
>>
>> Signed-off-by: Hillf Danton <dhillf@gmail.com>
>> ---
>
> Ok. So set the leftmost element with the help of rb_next() at the time of
> deletion of existing leftmost element. Not sure whether it is cheaper or
> not but does not harm doing it this way.
>
> Have you tested it to make sure nothing is broken?

Hm..if broken, we rush out and gnarl at Andrea Arcangeli as loud as we could.

Thanks
Hillf
>
> Thanks
> Vivek
>
>>
>> --- a/block/blk-throttle.c    2010-11-01 19:54:12.000000000 +0800
>> +++ b/block/blk-throttle.c    2010-11-26 21:27:40.000000000 +0800
>> @@ -251,7 +251,7 @@ static void rb_erase_init(struct rb_node
>>  static void throtl_rb_erase(struct rb_node *n, struct throtl_rb_root *root)
>>  {
>>       if (root->left == n)
>> -             root->left = NULL;
>> +             root->left = rb_next(n);
>>       rb_erase_init(n, &root->rb);
>>       --root->count;
>>  }
>> --
>> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
>> the body of a message to majordomo@vger.kernel.org
>> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>> Please read the FAQ at  http://www.tux.org/lkml/
>

Re: [PATCH] add rb_next in block throttle

[Vivek Goyal]

On Wed, Dec 01, 2010 at 09:11:01PM +0800, Hillf Danton wrote:
> On Tue, Nov 30, 2010 at 10:36 PM, Vivek Goyal <vgoyal@redhat.com> wrote:
> > On Fri, Nov 26, 2010 at 10:06:27PM +0800, Hillf Danton wrote:
> >> When selecting group for dispatching, the first entry of the service
> >> tree is sequentially dequeued, then the first entry is recomputed by
> >> calling rb_first().
> >>
> >> When the first is removed from service tree, new first could also be
> >> computed with rb_next(), since it could be faster than rb_first in
> >> this special environment.
> >>
> >> Signed-off-by: Hillf Danton <dhillf@gmail.com>
> >> ---
> >
> > Ok. So set the leftmost element with the help of rb_next() at the time of
> > deletion of existing leftmost element. Not sure whether it is cheaper or
> > not but does not harm doing it this way.
> >
> > Have you tested it to make sure nothing is broken?
> 
> Hm..if broken, we rush out and gnarl at Andrea Arcangeli as loud as we could.
> 

Well, some testing is always good. This looks simple enough though.

Acked-by: Vivek Goyal <vgoyal@redhat.com>

Vivek

> Thanks
> Hillf
> >
> > Thanks
> > Vivek
> >
> >>
> >> --- a/block/blk-throttle.c    2010-11-01 19:54:12.000000000 +0800
> >> +++ b/block/blk-throttle.c    2010-11-26 21:27:40.000000000 +0800
> >> @@ -251,7 +251,7 @@ static void rb_erase_init(struct rb_node
> >>  static void throtl_rb_erase(struct rb_node *n, struct throtl_rb_root *root)
> >>  {
> >>       if (root->left == n)
> >> -             root->left = NULL;
> >> +             root->left = rb_next(n);
> >>       rb_erase_init(n, &root->rb);
> >>       --root->count;
> >>  }
> >> --
> >> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> >> the body of a message to majordomo@vger.kernel.org
> >> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> >> Please read the FAQ at  http://www.tux.org/lkml/
> >