mm/slab: Fix drain freelist excessively
Subsystems (2)
PATCH 1/3 mm/slab: Fix drain freelist excessively
The drain_freelist is called to drain slabs_free lists for cache reap, cache shrink, memory hotplug callback etc. The tofree parameter is the number of slab objects to free instead of the number of slabs to free. The parameter transfered from callers is n->free_objects or n->freelimit
+ 5 * (searchp->num - 1) / (5 * searchp->num), and both of them mean the number of slabs objects. I add printk to dump drain information:
[ 122.864255] tofree is 2, actually free is 52, cache size is 26
The number of objects which caller prefer to drain is 2, however, actually 52 objects are drained, this destroy the cache locality.This patch fix it by compare the number of slabs objects which already drained instead of compare the number of slabs to the number of slab objects prefer to drain.
--- mm/slab.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
mm/slab.c +2 -2
@@ -2479,7 +2479,7 @@ static void drain_cpu_caches(struct kmem_cache *cachep)/* * Remove slabs from the list of free slabs. -* Specify the number of slabs to drain in tofree. +* Specify the number of slab objects to drain in tofree. * * Returns the actual number of slabs released. */
@@ -2491,7 +2491,7 @@ static int drain_freelist(struct kmem_cache *cache,struct slab *slabp; nr_freed =0 ; -while (nr_freed < tofree && !list_empty (&n->slabs_free)) { +while (nr_freed * cache->num < tofree && !list_empty (&n->slabs_free)) {spin_lock_irq (&n->list_lock); p = n->slabs_free.prev;
Signature · 7 lines
1.7.10.4 -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
PATCH 2/3 mm/slab: Sharing s_next and s_stop between slab and slub
This patch shares s_next and s_stop between slab and slub.
--- mm/slab.c | 10 ---------- mm/slab.h | 3 +++ mm/slab_common.c | 4 ++-- 3 files changed, 5 insertions(+), 12 deletions(-)
mm/slab.c +0 -10
@@ -4430,16 +4430,6 @@ static int leaks_show(struct seq_file *m, void *p)return 0 ; } -static void *s_next (struct seq_file *m,void *p,loff_t *pos) -{ -return seq_list_next (p, &slab_caches, pos); -} - -static void s_stop (struct seq_file *m,void *p) -{ -mutex_unlock (&slab_mutex); -} -static const struct seq_operations slabstats_op = { .start = leaks_start, .next = s_next,
mm/slab.h +3 -0
@@ -271,3 +271,6 @@ struct kmem_cache_node {#endif }; + +void *s_next (struct seq_file *m,void *p,loff_t *pos); +void s_stop (struct seq_file *m,void *p);
mm/slab_common.c +2 -2
@@ -531,12 +531,12 @@ static void *s_start(struct seq_file *m, loff_t *pos)return seq_list_start (&slab_caches, *pos); } -static void *s_next (struct seq_file *m,void *p,loff_t *pos) +void *s_next (struct seq_file *m,void *p,loff_t *pos) {return seq_list_next (p, &slab_caches, pos); } -static void s_stop (struct seq_file *m,void *p) +void s_stop (struct seq_file *m,void *p) {mutex_unlock (&slab_mutex); }
Signature · 1 line
1.7.10.4
David Rientjes
reviewerRe: [PATCH 2/3] mm/slab: Sharing s_next and s_stop between slab and slub
Quoted 2 lines · Wanpeng Li
This patch shares
s_nextands_stopbetween slab and slub.
Just about the entire kernel includes slab.h, so I think you'll need to give these slab-specific names instead of exporting "s_next" and "s_stop" to everybody.
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Re: [PATCH 2/3] mm/slab: Sharing s_next and s_stop between slab and slub
Quoted 8 lines · David Rientjes
Quoted 2 lines · Wanpeng Li
This patch shares
s_nextands_stopbetween slab and slub.Just about the entire kernel includes
slab.h, so I think you'll need to give these slab-specific names instead of exporting "s_next" and "s_stop" to everybody.
Ok, I will update them in next version, thanks for your review. ;-)
Regards,
Wanpeng Li
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Re: [PATCH 2/3] mm/slab: Sharing s_next and s_stop between slab and slub
In reply to David RientjesQuoted 8 lines · David Rientjes
Quoted 2 lines · Wanpeng Li
This patch shares
s_nextands_stopbetween slab and slub.Just about the entire kernel includes
slab.h, so I think you'll need to give these slab-specific names instead of exporting "s_next" and "s_stop" to everybody.
Ok, I will update them in next version, thanks for your review. ;-)
Regards,
Wanpeng Li
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Re: [PATCH 2/3] mm/slab: Sharing s_next and s_stop between slab and slub
In reply to David RientjesQuoted 8 lines · David Rientjes
Quoted 2 lines · Wanpeng Li
This patch shares
s_nextands_stopbetween slab and slub.Just about the entire kernel includes
slab.h, so I think you'll need to give these slab-specific names instead of exporting "s_next" and "s_stop" to everybody.
He put the export into mm/slab.h. The headerfile is only included by mm/sl?b.c .
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Re: [PATCH 2/3] mm/slab: Sharing s_next and s_stop between slab and slub
Quoted 7 lines · Christoph Lameter
Quoted 5 lines · David Rientjes
Quoted 2 lines · Wanpeng Li
This patch shares
s_nextands_stopbetween slab and slub.
Quoted 6 lines
Quoted 3 lines
Just about the entire kernel includes
slab.h, so I think you'll need to give these slab-specific names instead of exporting "s_next" and "s_stop" to everybody.He put the export into
mm/slab.h. The headerfile is only included by mm/sl?b.c .
So I think this patch is ok, David?
Regards,
Wanpeng Li
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Re: [PATCH 2/3] mm/slab: Sharing s_next and s_stop between slab and slub
In reply to Christoph LameterQuoted 7 lines · Christoph Lameter
Quoted 5 lines · David Rientjes
Quoted 2 lines · Wanpeng Li
This patch shares
s_nextands_stopbetween slab and slub.
Quoted 6 lines
Quoted 3 lines
Just about the entire kernel includes
slab.h, so I think you'll need to give these slab-specific names instead of exporting "s_next" and "s_stop" to everybody.He put the export into
mm/slab.h. The headerfile is only included by mm/sl?b.c .
So I think this patch is ok, David?
Regards,
Wanpeng Li
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Re: [PATCH 2/3] mm/slab: Sharing s_next and s_stop between slab and slub
In reply to Christoph LameterQuoted 7 lines · Christoph Lameter
Quoted 5 lines · David Rientjes
Quoted 2 lines · Wanpeng Li
This patch shares
s_nextands_stopbetween slab and slub.
Quoted 6 lines
Quoted 3 lines
Just about the entire kernel includes
slab.h, so I think you'll need to give these slab-specific names instead of exporting "s_next" and "s_stop" to everybody.He put the export into
mm/slab.h. The headerfile is only included by mm/sl?b.c .
But he then went on to add globally visible symbols "s_next" and "s_stop" which is bad...
Please send me an incremental patch on top of slab/next to fix this up. Otherwise I'll revert it before sending a pull request to Linus.
Pekka
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Re: [PATCH 2/3] mm/slab: Sharing s_next and s_stop between slab and slub
Quoted 15 lines · Pekka Enberg
Quoted 7 lines · Christoph Lameter
Quoted 5 lines · David Rientjes
Quoted 2 lines · Wanpeng Li
This patch shares
s_nextands_stopbetween slab and slub.Quoted 6 lines
Quoted 3 lines
Just about the entire kernel includes
slab.h, so I think you'll need to give these slab-specific names instead of exporting "s_next" and "s_stop" to everybody.He put the export into
mm/slab.h. The headerfile is only included by mm/sl?b.c .
Quoted 7 lines
But he then went on to add globally visible symbols "
s_next" and "s_stop" which is bad...Please send me an incremental patch on top of slab/next to fix this up. Otherwise I'll revert it before sending a pull request to Linus.
Pekka
Hi Pekka,
I attach the incremental patch in attachment. ;-)
Regards,
Wanpeng Li
Pekka Enberg
AppliedRe: [PATCH 2/3] mm/slab: Sharing s_next and s_stop between slab and slub
Quoted 16 lines · Wanpeng Li
Quoted 15 lines · Pekka Enberg
Quoted 13 lines · Christoph Lameter
Quoted 8 lines · David Rientjes
Quoted 2 lines · Wanpeng Li
This patch shares
s_nextands_stopbetween slab and slub.Just about the entire kernel includes
slab.h, so I think you'll need to give these slab-specific names instead of exporting "s_next" and "s_stop" to everybody.He put the export into
mm/slab.h. The headerfile is only included by mm/sl?b.c .
Quoted 7 lines
Quoted 5 lines
But he then went on to add globally visible symbols "
s_next" and "s_stop" which is bad...Please send me an incremental patch on top of slab/next to fix this up. Otherwise I'll revert it before sending a pull request to Linus.
I attach the incremental patch in attachment. ;-)
Applied, thanks!
Pekka
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Re: [PATCH 2/3] mm/slab: Sharing s_next and s_stop between slab and slub
In reply to Pekka EnbergQuoted 15 lines · Pekka Enberg
Quoted 7 lines · Christoph Lameter
Quoted 5 lines · David Rientjes
Quoted 2 lines · Wanpeng Li
This patch shares
s_nextands_stopbetween slab and slub.Quoted 6 lines
Quoted 3 lines
Just about the entire kernel includes
slab.h, so I think you'll need to give these slab-specific names instead of exporting "s_next" and "s_stop" to everybody.He put the export into
mm/slab.h. The headerfile is only included by mm/sl?b.c .
Quoted 7 lines
But he then went on to add globally visible symbols "
s_next" and "s_stop" which is bad...Please send me an incremental patch on top of slab/next to fix this up. Otherwise I'll revert it before sending a pull request to Linus.
Pekka
Hi Pekka,
I attach the incremental patch in attachment. ;-)
Regards,
Wanpeng Li
Quoted 1 line
From
02f008d7321a4a10babb4f7398b9eb8ebe13af24Mon Sep 17 00:00:00 2001
From: Wanpeng Li <[email protected]>
Date: Mon, 8 Jul 2013 08:08:28 +0800
Subject: [PATCH] mm/slab: Give s_next and s_stop slab-specific names
Give s_next and s_stop slab-specific names instead of exporting "s_next" and "s_stop".
--- mm/slab.c | 4 ++-- mm/slab.h | 4 ++-- mm/slab_common.c | 8 ++++---- 3 files changed, 8 insertions(+), 8 deletions(-)
mm/slab.c +2 -2
@@ -4440,8 +4440,8 @@ static int leaks_show(struct seq_file *m, void *p)static const struct seq_operations slabstats_op = { .start = leaks_start, - .next = s_next, - .stop = s_stop, + .next = slab_next, + .stop = slab_stop, .show = leaks_show, };
mm/slab.h +2 -2
@@ -272,5 +272,5 @@ struct kmem_cache_node { }; -void *s_next (struct seq_file *m,void *p,loff_t *pos); -void s_stop (struct seq_file *m,void *p); +void *slab_next (struct seq_file *m,void *p,loff_t *pos); +void slab_stop (struct seq_file *m,void *p);
mm/slab_common.c +4 -4
@@ -536,12 +536,12 @@ static void *s_start(struct seq_file *m, loff_t *pos)return seq_list_start (&slab_caches, *pos); } -void *s_next (struct seq_file *m,void *p,loff_t *pos) +void *slab_next (struct seq_file *m,void *p,loff_t *pos) {return seq_list_next (p, &slab_caches, pos); } -void s_stop (struct seq_file *m,void *p) +void slab_stop (struct seq_file *m,void *p) {mutex_unlock (&slab_mutex); }
@@ -618,8 +618,8 @@ static int s_show(struct seq_file *m, void *p)*/ static const struct seq_operations slabinfo_op = { .start = s_start, - .next = s_next, - .stop = s_stop, + .next = slab_next, + .stop = slab_stop, .show = s_show, };
Signature · 1 line
1.8.1.2
PATCH 3/3 mm/slab: Fix /proc/slabinfo unwriteable for slab
In reply to Wanpeng LiSlab have some tunables like limit, batchcount, and sharedfactor can be tuned through function slabinfo_write. Commit (b7454ad3: mm/sl[au]b: Move slabinfo processing to slab_common.c) uncorrectly change /proc/slabinfo unwriteable for slab, this patch fix it by revert to original mode.
--- mm/slab_common.c | 10 ++++++++++ 1 file changed, 10 insertions(+)
mm/slab_common.c +10 -0
@@ -631,10 +631,20 @@ static const struct file_operations proc_slabinfo_operations = { .release = seq_release, }; +#ifdef CONFIG_SLAB +static int __init slab_proc_init (void ) +{ +proc_create ("slabinfo" ,S_IWUSR |S_IRUSR ,NULL , &proc_slabinfo_operations); +return 0 ; +} +#endif +#ifdef CONFIG_SLUB static int __init slab_proc_init (void ) {proc_create ("slabinfo" ,S_IRUSR ,NULL , &proc_slabinfo_operations);return 0 ; } +#endif +module_init (slab_proc_init);#endif /* CONFIG_SLABINFO */
Signature · 7 lines
1.7.10.4 -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Re: [PATCH 3/3] mm/slab: Fix /proc/slabinfo unwriteable for slab
Quoted 13 lines · Wanpeng Li
1 file changed, 10 insertions(+)mm/slab_common.c +3
-0@@ -631,10 +631,20 @@ static const struct file_operations proc_slabinfo_operations = { .release = seq_release, }; +#ifdef CONFIG_SLAB +static int __init slab_proc_init (void ) +{
mm/slab_common.c static const struct file_operations proc_slabinfo_operations = { · line 646
Quoted 10 lines
mm/slab_common.c +5
-0+proc_create ("slabinfo" ,S_IWUSR |S_IRUSR ,NULL , &proc_slabinfo_operations); + return 0; +} +#endif +#ifdef CONFIG_SLUB static int __init slab_proc_init (void ) {proc_create ("slabinfo" ,S_IRUSR ,NULL , &proc_slabinfo_operations);return 0 ; }
It may be easier to define a macro SLABINFO_RIGHTS and use #ifdefs to assign the correct one. That way we have only one slab_proc_init().
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Re: [PATCH 3/3] mm/slab: Fix /proc/slabinfo unwriteable for slab
Quoted 12 lines · Christoph Lameter
Quoted 10 lines · Wanpeng Li
1 file changed, 10 insertions(+)mm/slab_common.c +0
-0@@ -631,10 +631,20 @@ static const struct file_operations proc_slabinfo_operations = { .release = seq_release, };
Quoted 17 lines
mm/slab_common.c static const struct file_operations proc_slabinfo_operations = { · line 646
Quoted 13 lines
mm/slab_common.c +8
-0+#ifdef CONFIG_SLAB +static int __init slab_proc_init (void ) +{ +proc_create ("slabinfo" ,S_IWUSR |S_IRUSR ,NULL , &proc_slabinfo_operations); +return 0 ; +} +#endif +#ifdef CONFIG_SLUB static int __init slab_proc_init (void ) {proc_create ("slabinfo" ,S_IRUSR ,NULL , &proc_slabinfo_operations);return 0 ; }It may be easier to define a macro
SLABINFO_RIGHTSand use #ifdefs to assign the correct one. That way we have only oneslab_proc_init().
Greate point! I will do it in next version. ;-)
Regards,
Wanpeng Li
Quoted 5 lines
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Re: [PATCH 3/3] mm/slab: Fix /proc/slabinfo unwriteable for slab
In reply to Christoph LameterQuoted 12 lines · Christoph Lameter
Quoted 10 lines · Wanpeng Li
1 file changed, 10 insertions(+)mm/slab_common.c +0
-0@@ -631,10 +631,20 @@ static const struct file_operations proc_slabinfo_operations = { .release = seq_release, };
Quoted 17 lines
mm/slab_common.c static const struct file_operations proc_slabinfo_operations = { · line 646
Quoted 13 lines
mm/slab_common.c +8
-0+#ifdef CONFIG_SLAB +static int __init slab_proc_init (void ) +{ +proc_create ("slabinfo" ,S_IWUSR |S_IRUSR ,NULL , &proc_slabinfo_operations); +return 0 ; +} +#endif +#ifdef CONFIG_SLUB static int __init slab_proc_init (void ) {proc_create ("slabinfo" ,S_IRUSR ,NULL , &proc_slabinfo_operations);return 0 ; }It may be easier to define a macro
SLABINFO_RIGHTSand use #ifdefs to assign the correct one. That way we have only oneslab_proc_init().
Greate point! I will do it in next version. ;-)
Regards,
Wanpeng Li
Quoted 5 lines
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Quoted 3 lines · Wanpeng Li
The
drain_freelistis called to drainslabs_freelists for cache reap, cache shrink, memory hotplug callback etc. The tofree parameter is the number of slab objects to free instead of the number of slabs to free.
Well its intended to be the number of slabs to free. The patch does not fix the callers that pass the number of slabs.
I think the best approach would be to fix the callers that pass # of objects. Make sure they pass # of slabs.
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Quoted 6 lines · Christoph Lameter
Quoted 3 lines · Wanpeng Li
The
drain_freelistis called to drainslabs_freelists for cache reap, cache shrink, memory hotplug callback etc. The tofree parameter is the number of slab objects to free instead of the number of slabs to free.
Quoted 6 lines
Well its intended to be the number of slabs to free. The patch does not fix the callers that pass the number of slabs.
I think the best approach would be to fix the callers that pass # of objects. Make sure they pass # of slabs.
Good point, I will fix it in next version.
Regards,
Wanpeng Li
Quoted 5 lines
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Quoted 6 lines · Christoph Lameter
Quoted 3 lines · Wanpeng Li
The
drain_freelistis called to drainslabs_freelists for cache reap, cache shrink, memory hotplug callback etc. The tofree parameter is the number of slab objects to free instead of the number of slabs to free.
Quoted 6 lines
Well its intended to be the number of slabs to free. The patch does not fix the callers that pass the number of slabs.
I think the best approach would be to fix the callers that pass # of objects. Make sure they pass # of slabs.
Good point, I will fix it in next version.
Regards,
Wanpeng Li
Quoted 5 lines
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>