* [PATCH] temporarily disable swap token on memory pressure
@ 2005-11-25 22:37 Rik van Riel
2005-11-26 10:37 ` Zwane Mwaikambo
0 siblings, 1 reply; 4+ messages in thread
From: Rik van Riel @ 2005-11-25 22:37 UTC (permalink / raw)
To: linux-mm; +Cc: zwane
Some users (hi Zwane) have seen a problem when running a workload
that eats nearly all of physical memory - th system does an OOM
kill, even when there is still a lot of swap free.
I suspect the problem is that that big task is holding the swap
token, and the VM has a very hard time finding any other page in
the system that is swappable.
Instead of ignoring the swap token when sc->priority reaches 0,
we could simply take the swap token away from the memory hog and
make sure we don't give it back to the memory hog for a few seconds.
This patch is untested, since I have not reproduced Zwane's problem
on my system. I would like to see test results from anybody who is
running into this problem.
This patch is against today's git head.
Signed-off-by: Rik van Riel <riel@redhat.com>
Index: linux-2.6-token/mm/thrash.c
===================================================================
--- linux-2.6-token.orig/mm/thrash.c
+++ linux-2.6-token/mm/thrash.c
@@ -57,14 +57,17 @@ void grab_swap_token(void)
/* We have the token. Let others know we still need it. */
if (has_swap_token(current->mm)) {
current->mm->recent_pagein = 1;
+ if (unlikely(!swap_token_default_timeout))
+ disable_swap_token();
return;
}
if (time_after(jiffies, swap_token_check)) {
- /* Can't get swapout protection if we exceed our RSS limit. */
- // if (current->mm->rss > current->mm->rlimit_rss)
- // return;
+ if (!swap_token_default_timeout) {
+ swap_token_check = jiffies + SWAP_TOKEN_CHECK_INTERVAL;
+ return;
+ }
/* ... or if we recently held the token. */
if (time_before(jiffies, current->mm->swap_token_time))
@@ -95,6 +98,7 @@ void __put_swap_token(struct mm_struct *
{
spin_lock(&swap_token_lock);
if (likely(mm == swap_token_mm)) {
+ mm->swap_token_time = jiffies + SWAP_TOKEN_CHECK_INTERVAL;
swap_token_mm = &init_mm;
swap_token_check = jiffies;
}
Index: linux-2.6-token/include/linux/rmap.h
===================================================================
--- linux-2.6-token.orig/include/linux/rmap.h
+++ linux-2.6-token/include/linux/rmap.h
@@ -89,7 +89,7 @@ static inline void page_dup_rmap(struct
/*
* Called from mm/vmscan.c to handle paging out
*/
-int page_referenced(struct page *, int is_locked, int ignore_token);
+int page_referenced(struct page *, int is_locked);
int try_to_unmap(struct page *);
/*
@@ -109,7 +109,7 @@ unsigned long page_address_in_vma(struct
#define anon_vma_prepare(vma) (0)
#define anon_vma_link(vma) do {} while (0)
-#define page_referenced(page,l,i) TestClearPageReferenced(page)
+#define page_referenced(page,l) TestClearPageReferenced(page)
#define try_to_unmap(page) SWAP_FAIL
#endif /* CONFIG_MMU */
Index: linux-2.6-token/include/linux/swap.h
===================================================================
--- linux-2.6-token.orig/include/linux/swap.h
+++ linux-2.6-token/include/linux/swap.h
@@ -239,6 +239,11 @@ static inline void put_swap_token(struct
__put_swap_token(mm);
}
+static inline void disable_swap_token(void)
+{
+ put_swap_token(swap_token_mm);
+}
+
#else /* CONFIG_SWAP */
#define total_swap_pages 0
Index: linux-2.6-token/mm/vmscan.c
===================================================================
--- linux-2.6-token.orig/mm/vmscan.c
+++ linux-2.6-token/mm/vmscan.c
@@ -407,7 +407,7 @@ static int shrink_list(struct list_head
if (PageWriteback(page))
goto keep_locked;
- referenced = page_referenced(page, 1, sc->priority <= 0);
+ referenced = page_referenced(page, 1);
/* In active use or really unfreeable? Activate it. */
if (referenced && page_mapping_inuse(page))
goto activate_locked;
@@ -756,7 +756,7 @@ refill_inactive_zone(struct zone *zone,
if (page_mapped(page)) {
if (!reclaim_mapped ||
(total_swap_pages == 0 && PageAnon(page)) ||
- page_referenced(page, 0, sc->priority <= 0)) {
+ page_referenced(page, 0)) {
list_add(&page->lru, &l_active);
continue;
}
@@ -960,6 +960,8 @@ int try_to_free_pages(struct zone **zone
sc.nr_reclaimed = 0;
sc.priority = priority;
sc.swap_cluster_max = SWAP_CLUSTER_MAX;
+ if (!priority)
+ disable_swap_token();
shrink_caches(zones, &sc);
shrink_slab(sc.nr_scanned, gfp_mask, lru_pages);
if (reclaim_state) {
@@ -1056,6 +1058,10 @@ loop_again:
int end_zone = 0; /* Inclusive. 0 = ZONE_DMA */
unsigned long lru_pages = 0;
+ /* The swap token gets in the way of swapout... */
+ if (!priority)
+ disable_swap_token();
+
all_zones_ok = 1;
if (nr_pages == 0) {
@@ -1360,6 +1366,7 @@ int zone_reclaim(struct zone *zone, gfp_
sc.nr_reclaimed = 0;
/* scan at the highest priority */
sc.priority = 0;
+ disable_swap_token();
if (nr_pages > SWAP_CLUSTER_MAX)
sc.swap_cluster_max = nr_pages;
Index: linux-2.6-token/mm/rmap.c
===================================================================
--- linux-2.6-token.orig/mm/rmap.c
+++ linux-2.6-token/mm/rmap.c
@@ -292,7 +292,7 @@ pte_t *page_check_address(struct page *p
* repeatedly from either page_referenced_anon or page_referenced_file.
*/
static int page_referenced_one(struct page *page,
- struct vm_area_struct *vma, unsigned int *mapcount, int ignore_token)
+ struct vm_area_struct *vma, unsigned int *mapcount)
{
struct mm_struct *mm = vma->vm_mm;
unsigned long address;
@@ -313,7 +313,7 @@ static int page_referenced_one(struct pa
/* Pretend the page is referenced if the task has the
swap token and is in the middle of a page fault. */
- if (mm != current->mm && !ignore_token && has_swap_token(mm) &&
+ if (mm != current->mm && has_swap_token(mm) &&
rwsem_is_locked(&mm->mmap_sem))
referenced++;
@@ -323,7 +323,7 @@ out:
return referenced;
}
-static int page_referenced_anon(struct page *page, int ignore_token)
+static int page_referenced_anon(struct page *page)
{
unsigned int mapcount;
struct anon_vma *anon_vma;
@@ -336,8 +336,7 @@ static int page_referenced_anon(struct p
mapcount = page_mapcount(page);
list_for_each_entry(vma, &anon_vma->head, anon_vma_node) {
- referenced += page_referenced_one(page, vma, &mapcount,
- ignore_token);
+ referenced += page_referenced_one(page, vma, &mapcount);
if (!mapcount)
break;
}
@@ -356,7 +355,7 @@ static int page_referenced_anon(struct p
*
* This function is only called from page_referenced for object-based pages.
*/
-static int page_referenced_file(struct page *page, int ignore_token)
+static int page_referenced_file(struct page *page)
{
unsigned int mapcount;
struct address_space *mapping = page->mapping;
@@ -394,8 +393,7 @@ static int page_referenced_file(struct p
referenced++;
break;
}
- referenced += page_referenced_one(page, vma, &mapcount,
- ignore_token);
+ referenced += page_referenced_one(page, vma, &mapcount);
if (!mapcount)
break;
}
@@ -412,13 +410,10 @@ static int page_referenced_file(struct p
* Quick test_and_clear_referenced for all mappings to a page,
* returns the number of ptes which referenced the page.
*/
-int page_referenced(struct page *page, int is_locked, int ignore_token)
+int page_referenced(struct page *page, int is_locked)
{
int referenced = 0;
- if (!swap_token_default_timeout)
- ignore_token = 1;
-
if (page_test_and_clear_young(page))
referenced++;
@@ -427,15 +422,14 @@ int page_referenced(struct page *page, i
if (page_mapped(page) && page->mapping) {
if (PageAnon(page))
- referenced += page_referenced_anon(page, ignore_token);
+ referenced += page_referenced_anon(page);
else if (is_locked)
- referenced += page_referenced_file(page, ignore_token);
+ referenced += page_referenced_file(page);
else if (TestSetPageLocked(page))
referenced++;
else {
if (page->mapping)
- referenced += page_referenced_file(page,
- ignore_token);
+ referenced += page_referenced_file(page);
unlock_page(page);
}
}
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [PATCH] temporarily disable swap token on memory pressure
2005-11-25 22:37 [PATCH] temporarily disable swap token on memory pressure Rik van Riel
@ 2005-11-26 10:37 ` Zwane Mwaikambo
2005-11-26 10:41 ` Zwane Mwaikambo
0 siblings, 1 reply; 4+ messages in thread
From: Zwane Mwaikambo @ 2005-11-26 10:37 UTC (permalink / raw)
To: Rik van Riel; +Cc: linux-mm
On Fri, 25 Nov 2005, Rik van Riel wrote:
> Some users (hi Zwane) have seen a problem when running a workload
> that eats nearly all of physical memory - th system does an OOM
> kill, even when there is still a lot of swap free.
>
> I suspect the problem is that that big task is holding the swap
> token, and the VM has a very hard time finding any other page in
> the system that is swappable.
>
> Instead of ignoring the swap token when sc->priority reaches 0,
> we could simply take the swap token away from the memory hog and
> make sure we don't give it back to the memory hog for a few seconds.
>
> This patch is untested, since I have not reproduced Zwane's problem
> on my system. I would like to see test results from anybody who is
> running into this problem.
>
> This patch is against today's git head.
Very nice! With this patch my job actually completed.
MemTotal: 2049180 kB
MemFree: 15592 kB
Buffers: 1516 kB
Cached: 62972 kB
SwapCached: 190512 kB
Active: 1626228 kB
Inactive: 361288 kB
HighTotal: 0 kB
HighFree: 0 kB
LowTotal: 2049180 kB
LowFree: 15592 kB
SwapTotal: 3228760 kB
SwapFree: 2555688 kB
Dirty: 0 kB
Writeback: 0 kB
Mapped: 1824108 kB
Slab: 27392 kB
CommitLimit: 4253348 kB
Committed_AS: 2472188 kB
PageTables: 6688 kB
VmallocTotal: 34359738367 kB
VmallocUsed: 263796 kB
VmallocChunk: 34359474119 kB
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [PATCH] temporarily disable swap token on memory pressure
2005-11-26 10:37 ` Zwane Mwaikambo
@ 2005-11-26 10:41 ` Zwane Mwaikambo
0 siblings, 0 replies; 4+ messages in thread
From: Zwane Mwaikambo @ 2005-11-26 10:41 UTC (permalink / raw)
To: Rik van Riel; +Cc: linux-mm
On Sat, 26 Nov 2005, Zwane Mwaikambo wrote:
> On Fri, 25 Nov 2005, Rik van Riel wrote:
>
> > Some users (hi Zwane) have seen a problem when running a workload
> > that eats nearly all of physical memory - th system does an OOM
> > kill, even when there is still a lot of swap free.
> >
> > I suspect the problem is that that big task is holding the swap
> > token, and the VM has a very hard time finding any other page in
> > the system that is swappable.
> >
> > Instead of ignoring the swap token when sc->priority reaches 0,
> > we could simply take the swap token away from the memory hog and
> > make sure we don't give it back to the memory hog for a few seconds.
> >
> > This patch is untested, since I have not reproduced Zwane's problem
> > on my system. I would like to see test results from anybody who is
> > running into this problem.
> >
> > This patch is against today's git head.
>
> Very nice! With this patch my job actually completed.
>
> MemTotal: 2049180 kB
> MemFree: 15592 kB
> Buffers: 1516 kB
> Cached: 62972 kB
> SwapCached: 190512 kB
> Active: 1626228 kB
> Inactive: 361288 kB
> HighTotal: 0 kB
> HighFree: 0 kB
> LowTotal: 2049180 kB
> LowFree: 15592 kB
> SwapTotal: 3228760 kB
> SwapFree: 2555688 kB
> Dirty: 0 kB
> Writeback: 0 kB
> Mapped: 1824108 kB
> Slab: 27392 kB
> CommitLimit: 4253348 kB
> Committed_AS: 2472188 kB
> PageTables: 6688 kB
> VmallocTotal: 34359738367 kB
> VmallocUsed: 263796 kB
> VmallocChunk: 34359474119 kB
Some information about the application;
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
3974 root 25 0 2459m 1.7g 8588 R 99.9 88.2 118:38.43 simnow
root@morocco ~ {0:0} cat /proc/3974/maps | wc -l
73870
Lots of PAGE_SIZE mappings like so;
...
401a6c0000-401a6c1000 rwxp 401a6c0000 00:00 0
401a6c2000-401a6c3000 rwxp 401a6c2000 00:00 0
401a6c4000-401a6c5000 rwxp 401a6c4000 00:00 0
401a6c6000-401a6c7000 rwxp 401a6c6000 00:00 0
401a6c8000-401a6c9000 rwxp 401a6c8000 00:00 0
...
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>
^ permalink raw reply [flat|nested] 4+ messages in thread
* [PATCH] temporarily disable swap token on memory pressure
@ 2005-11-26 17:43 Rik van Riel
0 siblings, 0 replies; 4+ messages in thread
From: Rik van Riel @ 2005-11-26 17:43 UTC (permalink / raw)
To: Andrew Morton; +Cc: linux-kernel, linux-mm, zwane
Some users (hi Zwane) have seen a problem when running a workload
that eats nearly all of physical memory - th system does an OOM
kill, even when there is still a lot of swap free.
The problem appears to be a very big task that is holding the swap
token, and the VM has a very hard time finding any other page in
the system that is swappable.
Instead of ignoring the swap token when sc->priority reaches 0,
we could simply take the swap token away from the memory hog and
make sure we don't give it back to the memory hog for a few seconds.
This patch resolves the problem Zwane ran into.
This patch is against yesterday's git head.
Signed-off-by: Rik van Riel <riel@redhat.com>
Index: linux-2.6-token/mm/thrash.c
===================================================================
--- linux-2.6-token.orig/mm/thrash.c
+++ linux-2.6-token/mm/thrash.c
@@ -57,14 +57,17 @@ void grab_swap_token(void)
/* We have the token. Let others know we still need it. */
if (has_swap_token(current->mm)) {
current->mm->recent_pagein = 1;
+ if (unlikely(!swap_token_default_timeout))
+ disable_swap_token();
return;
}
if (time_after(jiffies, swap_token_check)) {
- /* Can't get swapout protection if we exceed our RSS limit. */
- // if (current->mm->rss > current->mm->rlimit_rss)
- // return;
+ if (!swap_token_default_timeout) {
+ swap_token_check = jiffies + SWAP_TOKEN_CHECK_INTERVAL;
+ return;
+ }
/* ... or if we recently held the token. */
if (time_before(jiffies, current->mm->swap_token_time))
@@ -95,6 +98,7 @@ void __put_swap_token(struct mm_struct *
{
spin_lock(&swap_token_lock);
if (likely(mm == swap_token_mm)) {
+ mm->swap_token_time = jiffies + SWAP_TOKEN_CHECK_INTERVAL;
swap_token_mm = &init_mm;
swap_token_check = jiffies;
}
Index: linux-2.6-token/include/linux/rmap.h
===================================================================
--- linux-2.6-token.orig/include/linux/rmap.h
+++ linux-2.6-token/include/linux/rmap.h
@@ -89,7 +89,7 @@ static inline void page_dup_rmap(struct
/*
* Called from mm/vmscan.c to handle paging out
*/
-int page_referenced(struct page *, int is_locked, int ignore_token);
+int page_referenced(struct page *, int is_locked);
int try_to_unmap(struct page *);
/*
@@ -109,7 +109,7 @@ unsigned long page_address_in_vma(struct
#define anon_vma_prepare(vma) (0)
#define anon_vma_link(vma) do {} while (0)
-#define page_referenced(page,l,i) TestClearPageReferenced(page)
+#define page_referenced(page,l) TestClearPageReferenced(page)
#define try_to_unmap(page) SWAP_FAIL
#endif /* CONFIG_MMU */
Index: linux-2.6-token/include/linux/swap.h
===================================================================
--- linux-2.6-token.orig/include/linux/swap.h
+++ linux-2.6-token/include/linux/swap.h
@@ -239,6 +239,11 @@ static inline void put_swap_token(struct
__put_swap_token(mm);
}
+static inline void disable_swap_token(void)
+{
+ put_swap_token(swap_token_mm);
+}
+
#else /* CONFIG_SWAP */
#define total_swap_pages 0
Index: linux-2.6-token/mm/vmscan.c
===================================================================
--- linux-2.6-token.orig/mm/vmscan.c
+++ linux-2.6-token/mm/vmscan.c
@@ -407,7 +407,7 @@ static int shrink_list(struct list_head
if (PageWriteback(page))
goto keep_locked;
- referenced = page_referenced(page, 1, sc->priority <= 0);
+ referenced = page_referenced(page, 1);
/* In active use or really unfreeable? Activate it. */
if (referenced && page_mapping_inuse(page))
goto activate_locked;
@@ -756,7 +756,7 @@ refill_inactive_zone(struct zone *zone,
if (page_mapped(page)) {
if (!reclaim_mapped ||
(total_swap_pages == 0 && PageAnon(page)) ||
- page_referenced(page, 0, sc->priority <= 0)) {
+ page_referenced(page, 0)) {
list_add(&page->lru, &l_active);
continue;
}
@@ -960,6 +960,8 @@ int try_to_free_pages(struct zone **zone
sc.nr_reclaimed = 0;
sc.priority = priority;
sc.swap_cluster_max = SWAP_CLUSTER_MAX;
+ if (!priority)
+ disable_swap_token();
shrink_caches(zones, &sc);
shrink_slab(sc.nr_scanned, gfp_mask, lru_pages);
if (reclaim_state) {
@@ -1056,6 +1058,10 @@ loop_again:
int end_zone = 0; /* Inclusive. 0 = ZONE_DMA */
unsigned long lru_pages = 0;
+ /* The swap token gets in the way of swapout... */
+ if (!priority)
+ disable_swap_token();
+
all_zones_ok = 1;
if (nr_pages == 0) {
@@ -1360,6 +1366,7 @@ int zone_reclaim(struct zone *zone, gfp_
sc.nr_reclaimed = 0;
/* scan at the highest priority */
sc.priority = 0;
+ disable_swap_token();
if (nr_pages > SWAP_CLUSTER_MAX)
sc.swap_cluster_max = nr_pages;
Index: linux-2.6-token/mm/rmap.c
===================================================================
--- linux-2.6-token.orig/mm/rmap.c
+++ linux-2.6-token/mm/rmap.c
@@ -292,7 +292,7 @@ pte_t *page_check_address(struct page *p
* repeatedly from either page_referenced_anon or page_referenced_file.
*/
static int page_referenced_one(struct page *page,
- struct vm_area_struct *vma, unsigned int *mapcount, int ignore_token)
+ struct vm_area_struct *vma, unsigned int *mapcount)
{
struct mm_struct *mm = vma->vm_mm;
unsigned long address;
@@ -313,7 +313,7 @@ static int page_referenced_one(struct pa
/* Pretend the page is referenced if the task has the
swap token and is in the middle of a page fault. */
- if (mm != current->mm && !ignore_token && has_swap_token(mm) &&
+ if (mm != current->mm && has_swap_token(mm) &&
rwsem_is_locked(&mm->mmap_sem))
referenced++;
@@ -323,7 +323,7 @@ out:
return referenced;
}
-static int page_referenced_anon(struct page *page, int ignore_token)
+static int page_referenced_anon(struct page *page)
{
unsigned int mapcount;
struct anon_vma *anon_vma;
@@ -336,8 +336,7 @@ static int page_referenced_anon(struct p
mapcount = page_mapcount(page);
list_for_each_entry(vma, &anon_vma->head, anon_vma_node) {
- referenced += page_referenced_one(page, vma, &mapcount,
- ignore_token);
+ referenced += page_referenced_one(page, vma, &mapcount);
if (!mapcount)
break;
}
@@ -356,7 +355,7 @@ static int page_referenced_anon(struct p
*
* This function is only called from page_referenced for object-based pages.
*/
-static int page_referenced_file(struct page *page, int ignore_token)
+static int page_referenced_file(struct page *page)
{
unsigned int mapcount;
struct address_space *mapping = page->mapping;
@@ -394,8 +393,7 @@ static int page_referenced_file(struct p
referenced++;
break;
}
- referenced += page_referenced_one(page, vma, &mapcount,
- ignore_token);
+ referenced += page_referenced_one(page, vma, &mapcount);
if (!mapcount)
break;
}
@@ -412,13 +410,10 @@ static int page_referenced_file(struct p
* Quick test_and_clear_referenced for all mappings to a page,
* returns the number of ptes which referenced the page.
*/
-int page_referenced(struct page *page, int is_locked, int ignore_token)
+int page_referenced(struct page *page, int is_locked)
{
int referenced = 0;
- if (!swap_token_default_timeout)
- ignore_token = 1;
-
if (page_test_and_clear_young(page))
referenced++;
@@ -427,15 +422,14 @@ int page_referenced(struct page *page, i
if (page_mapped(page) && page->mapping) {
if (PageAnon(page))
- referenced += page_referenced_anon(page, ignore_token);
+ referenced += page_referenced_anon(page);
else if (is_locked)
- referenced += page_referenced_file(page, ignore_token);
+ referenced += page_referenced_file(page);
else if (TestSetPageLocked(page))
referenced++;
else {
if (page->mapping)
- referenced += page_referenced_file(page,
- ignore_token);
+ referenced += page_referenced_file(page);
unlock_page(page);
}
}
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>
^ permalink raw reply [flat|nested] 4+ messages in thread
end of thread, other threads:[~2005-11-26 17:43 UTC | newest]
Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2005-11-25 22:37 [PATCH] temporarily disable swap token on memory pressure Rik van Riel
2005-11-26 10:37 ` Zwane Mwaikambo
2005-11-26 10:41 ` Zwane Mwaikambo
2005-11-26 17:43 Rik van Riel
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox