This patch series is trivial cleanup and fix of page migration.
==========================================================
Christoph pointed out inc_zone_page_state(NR_ISOLATED) should be placed
in right after isolate_page().
This patch does it.
Cc: Christoph Lameter <[email protected]>
Signed-off-by: KOSAKI Motohiro <[email protected]>
---
mm/memory_hotplug.c | 4 ++++
mm/mempolicy.c | 3 +++
mm/migrate.c | 12 ++++--------
3 files changed, 11 insertions(+), 8 deletions(-)
diff --git a/mm/memory_hotplug.c b/mm/memory_hotplug.c
index 821dee5..653bf1e 100644
--- a/mm/memory_hotplug.c
+++ b/mm/memory_hotplug.c
@@ -26,6 +26,7 @@
#include <linux/migrate.h>
#include <linux/page-isolation.h>
#include <linux/pfn.h>
+#include <linux/mm_inline.h>
#include <asm/tlbflush.h>
@@ -663,6 +664,9 @@ do_migrate_range(unsigned long start_pfn, unsigned long end_pfn)
if (!ret) { /* Success */
list_add_tail(&page->lru, &source);
move_pages--;
+ inc_zone_page_state(page, NR_ISOLATED_ANON +
+ page_is_file_cache(page));
+
} else {
/* Becasue we don't have big zone->lock. we should
check this again here. */
diff --git a/mm/mempolicy.c b/mm/mempolicy.c
index 7dd9d9f..473f888 100644
--- a/mm/mempolicy.c
+++ b/mm/mempolicy.c
@@ -89,6 +89,7 @@
#include <linux/security.h>
#include <linux/syscalls.h>
#include <linux/ctype.h>
+#include <linux/mm_inline.h>
#include <asm/tlbflush.h>
#include <asm/uaccess.h>
@@ -809,6 +810,8 @@ static void migrate_page_add(struct page *page, struct list_head *pagelist,
if ((flags & MPOL_MF_MOVE_ALL) || page_mapcount(page) == 1) {
if (!isolate_lru_page(page)) {
list_add_tail(&page->lru, pagelist);
+ inc_zone_page_state(page, NR_ISOLATED_ANON +
+ page_is_file_cache(page));
}
}
}
diff --git a/mm/migrate.c b/mm/migrate.c
index 1a4bf48..0f66803 100644
--- a/mm/migrate.c
+++ b/mm/migrate.c
@@ -746,13 +746,6 @@ int migrate_pages(struct list_head *from,
struct page *page2;
int swapwrite = current->flags & PF_SWAPWRITE;
int rc;
- unsigned long flags;
-
- local_irq_save(flags);
- list_for_each_entry(page, from, lru)
- __inc_zone_page_state(page, NR_ISOLATED_ANON +
- page_is_file_cache(page));
- local_irq_restore(flags);
if (!swapwrite)
current->flags |= PF_SWAPWRITE;
@@ -878,8 +871,11 @@ static int do_move_page_to_node_array(struct mm_struct *mm,
goto put_and_set;
err = isolate_lru_page(page);
- if (!err)
+ if (!err) {
list_add_tail(&page->lru, &pagelist);
+ inc_zone_page_state(page, NR_ISOLATED_ANON +
+ page_is_file_cache(page));
+ }
put_and_set:
/*
* Either remove the duplicate refcount from
--
1.6.0.GIT
if mbind() receive invalid address, do_mbind makes leaked page.
following test program detect its leak.
This patch fixes it.
migrate_efault.c
============================================
#include <numaif.h>
#include <numa.h>
#include <sys/mman.h>
#include <stdio.h>
#include <unistd.h>
#include <stdlib.h>
#include <string.h>
static unsigned long pagesize;
static void* make_hole_mapping(void)
{
void* addr;
addr = mmap(NULL, pagesize*3, PROT_READ|PROT_WRITE,
MAP_ANON|MAP_PRIVATE, 0, 0);
if (addr == MAP_FAILED)
return NULL;
/* make page populate */
memset(addr, 0, pagesize*3);
/* make memory hole */
munmap(addr+pagesize, pagesize);
return addr;
}
int main(int argc, char** argv)
{
void* addr;
int ch;
int node;
struct bitmask *nmask = numa_allocate_nodemask();
int err;
int node_set = 0;
while ((ch = getopt(argc, argv, "n:")) != -1){
switch (ch){
case 'n':
node = strtol(optarg, NULL, 0);
numa_bitmask_setbit(nmask, node);
node_set = 1;
break;
default:
;
}
}
argc -= optind;
argv += optind;
if (!node_set)
numa_bitmask_setbit(nmask, 0);
pagesize = getpagesize();
addr = make_hole_mapping();
err = mbind(addr, pagesize*3, MPOL_BIND, nmask->maskp, nmask->size, MPOL_MF_MOVE_ALL);
if (err)
perror("mbind ");
return 0;
}
============================================
Signed-off-by: KOSAKI Motohiro <[email protected]>
---
mm/mempolicy.c | 2 ++
1 files changed, 2 insertions(+), 0 deletions(-)
diff --git a/mm/mempolicy.c b/mm/mempolicy.c
index 473f888..824abf3 100644
--- a/mm/mempolicy.c
+++ b/mm/mempolicy.c
@@ -1061,6 +1061,8 @@ static long do_mbind(unsigned long start, unsigned long len,
if (!err && nr_failed && (flags & MPOL_MF_STRICT))
err = -EIO;
+ } else {
+ putback_lru_pages(&pagelist);
}
up_write(&mm->mmap_sem);
--
1.6.0.GIT
If migrate_prep is failed, new variable is leaked.
This patch fixes it.
Signed-off-by: KOSAKI Motohiro <[email protected]>
---
mm/mempolicy.c | 10 +++++-----
1 files changed, 5 insertions(+), 5 deletions(-)
diff --git a/mm/mempolicy.c b/mm/mempolicy.c
index 824abf3..38ce2a7 100644
--- a/mm/mempolicy.c
+++ b/mm/mempolicy.c
@@ -1027,7 +1027,7 @@ static long do_mbind(unsigned long start, unsigned long len,
err = migrate_prep();
if (err)
- return err;
+ goto mpol_out;
}
{
NODEMASK_SCRATCH(scratch);
@@ -1042,10 +1042,9 @@ static long do_mbind(unsigned long start, unsigned long len,
err = -ENOMEM;
NODEMASK_SCRATCH_FREE(scratch);
}
- if (err) {
- mpol_put(new);
- return err;
- }
+ if (err)
+ goto mpol_out;
+
vma = check_range(mm, start, end, nmask,
flags | MPOL_MF_INVERT, &pagelist);
@@ -1066,6 +1065,7 @@ static long do_mbind(unsigned long start, unsigned long len,
}
up_write(&mm->mmap_sem);
+ mpol_out:
mpol_put(new);
return err;
}
--
1.6.0.GIT
> Signed-off-by: KOSAKI Motohiro <[email protected]>
> ---
> mm/mempolicy.c | 2 ++
> 1 files changed, 2 insertions(+), 0 deletions(-)
>
> diff --git a/mm/mempolicy.c b/mm/mempolicy.c
> index 473f888..824abf3 100644
> --- a/mm/mempolicy.c
> +++ b/mm/mempolicy.c
> @@ -1061,6 +1061,8 @@ static long do_mbind(unsigned long start, unsigned long len,
>
> if (!err && nr_failed && (flags & MPOL_MF_STRICT))
> err = -EIO;
> + } else {
> + putback_lru_pages(&pagelist);
> }
>
> up_write(&mm->mmap_sem);
Oops, I forgot to remove unnecessary brace.
updated patch is here.
================================================================
Subject: [PATCH] Fix memory leak of never putback pages in mbind()
if mbind() receive invalid address, do_mbind makes leaked page.
following test program detect its leak.
This patch fixes it.
migrate_efault.c
=======================================
#include <numaif.h>
#include <numa.h>
#include <sys/mman.h>
#include <stdio.h>
#include <unistd.h>
#include <stdlib.h>
#include <string.h>
static unsigned long pagesize;
static void* make_hole_mapping(void)
{
void* addr;
addr = mmap(NULL, pagesize*3, PROT_READ|PROT_WRITE,
MAP_ANON|MAP_PRIVATE, 0, 0);
if (addr == MAP_FAILED)
return NULL;
/* make page populate */
memset(addr, 0, pagesize*3);
/* make memory hole */
munmap(addr+pagesize, pagesize);
return addr;
}
int main(int argc, char** argv)
{
void* addr;
int ch;
int node;
struct bitmask *nmask = numa_allocate_nodemask();
int err;
int node_set = 0;
while ((ch = getopt(argc, argv, "n:")) != -1){
switch (ch){
case 'n':
node = strtol(optarg, NULL, 0);
numa_bitmask_setbit(nmask, node);
node_set = 1;
break;
default:
;
}
}
argc -= optind;
argv += optind;
if (!node_set)
numa_bitmask_setbit(nmask, 0);
pagesize = getpagesize();
addr = make_hole_mapping();
err = mbind(addr, pagesize*3, MPOL_BIND, nmask->maskp, nmask->size, MPOL_MF_MOVE_ALL);
if (err)
perror("mbind ");
return 0;
}
=======================================
Signed-off-by: KOSAKI Motohiro <[email protected]>
---
mm/mempolicy.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
Index: b/mm/mempolicy.c
===================================================================
--- a/mm/mempolicy.c
+++ b/mm/mempolicy.c
@@ -1061,7 +1061,8 @@ static long do_mbind(unsigned long start
if (!err && nr_failed && (flags & MPOL_MF_STRICT))
err = -EIO;
- }
+ } else
+ putback_lru_pages(&pagelist);
up_write(&mm->mmap_sem);
mpol_put(new);
Reviewed-by: Christoph Lameter <[email protected]>
On Fri, 9 Oct 2009, KOSAKI Motohiro wrote:
> if mbind() receive invalid address, do_mbind makes leaked page.
> following test program detect its leak.
Acked-by: Christoph Lameter <[email protected]>
On Fri, 9 Oct 2009, KOSAKI Motohiro wrote:
> If migrate_prep is failed, new variable is leaked.
> This patch fixes it.
Acked-by: Christoph Lameter <[email protected]>
On Fri, 9 Oct 2009, KOSAKI Motohiro wrote:
> Oops, I forgot to remove unnecessary brace.
> updated patch is here.
Thats a style issue. There are other weird things in do_mbind as well
like starting a new block in the middle of another.
Having
}
{
in a program is a bit confusing. So could you do a cleanup patch for
mpol_bind? Preferably it should make it easy to read to and bring some
order to the confusing error handling.
>> Oops, I forgot to remove unnecessary brace.
>> updated patch is here.
>
> Thats a style issue. There are other weird things in do_mbind as well
> like starting a new block in the middle of another.
>
> Having
>
> }
> {
>
> in a program is a bit confusing. So could you do a cleanup patch for
> mpol_bind? Preferably it should make it easy to read to and bring some
> order to the confusing error handling.
Yes, I'll do.
On Fri, 9 Oct 2009 10:06:58 +0900 (JST)
KOSAKI Motohiro <[email protected]> wrote:
> This patch series is trivial cleanup and fix of page migration.
>
>
> ==========================================================
>
> Christoph pointed out inc_zone_page_state(NR_ISOLATED) should be placed
> in right after isolate_page().
The bugfixes are appropriate for 2.6.32 and should be backported into
-stable too, I think. I haven't checked to see how long those bugs
have been present.
The cleanup is more appropriate for 2.6.33 so I had to switch the order
of these patches. Hopefully the bugfixes were not dependent on the
cleanup.
> On Fri, 9 Oct 2009 10:06:58 +0900 (JST)
> KOSAKI Motohiro <[email protected]> wrote:
>
> > This patch series is trivial cleanup and fix of page migration.
> >
> >
> > ==========================================================
> >
> > Christoph pointed out inc_zone_page_state(NR_ISOLATED) should be placed
> > in right after isolate_page().
>
> The bugfixes are appropriate for 2.6.32 and should be backported into
> -stable too, I think. I haven't checked to see how long those bugs
> have been present.
>
> The cleanup is more appropriate for 2.6.33 so I had to switch the order
> of these patches. Hopefully the bugfixes were not dependent on the
> cleanup.
Yes, each patches are independent.
[1/3] is cleanup.
[2/3] and [3/3] are bugfixes.
I'm sorry for lack of prudence of patch order.