* [PATCH v3] mm: add replace_page_cache_page() function
@ 2011-01-11 14:07 Miklos Szeredi
2011-01-11 14:25 ` Minchan Kim
0 siblings, 1 reply; 9+ messages in thread
From: Miklos Szeredi @ 2011-01-11 14:07 UTC (permalink / raw)
To: akpm
Cc: minchan.kim, kamezawa.hiroyu, nishimura, linux-fsdevel, linux-mm,
linux-kernel
(resent with fixed CC list, sorry for the duplicate)
Thanks for the review.
Here's an updated patch. Modifications since the last post:
- don't pass gfp_mask (since it's only able to deal with GFP_KERNEL
anyway)
- use mem_cgroup_prepare_migration() and mem_cgroup_end_migration()
instead of intrdoucing mem_cgroup_replace_cache_page() helper.
This can be done later if the performance impact of using the
migration functions turns out to be excessive.
Applies on top of memcg-fix-memory-migration-of-shmem-swapcache.patch
Thanks,
Miklos
---
From: Miklos Szeredi <mszeredi@suse.cz>
Subject: mm: add replace_page_cache_page() function
This function basically does:
remove_from_page_cache(old);
page_cache_release(old);
add_to_page_cache_locked(new);
Except it does this atomically, so there's no possibility for the
"add" to fail because of a race.
This is used by fuse to move pages into the page cache.
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
---
fs/fuse/dev.c | 10 +++----
include/linux/pagemap.h | 1
mm/filemap.c | 64 ++++++++++++++++++++++++++++++++++++++++++++++++
3 files changed, 69 insertions(+), 6 deletions(-)
Index: linux-2.6/mm/filemap.c
===================================================================
--- linux-2.6.orig/mm/filemap.c 2011-01-11 14:38:12.000000000 +0100
+++ linux-2.6/mm/filemap.c 2011-01-11 14:38:17.000000000 +0100
@@ -387,6 +387,70 @@ int filemap_write_and_wait_range(struct
EXPORT_SYMBOL(filemap_write_and_wait_range);
/**
+ * replace_page_cache_page - replace a pagecache page with a new one
+ * @old: page to be replaced
+ * @new: page to replace with
+ *
+ * This function replaces a page in the pagecache with a new one. On
+ * success it acquires the pagecache reference for the new page and
+ * drops it for the old page. Both the old and new pages must be
+ * locked. This function does not add the new page to the LRU, the
+ * caller must do that.
+ *
+ * The remove + add is atomic. The only way this function can fail is
+ * memory allocation failure.
+ */
+int replace_page_cache_page(struct page *old, struct page *new)
+{
+ int error;
+ struct mem_cgroup *memcg = NULL;
+
+ VM_BUG_ON(!PageLocked(old));
+ VM_BUG_ON(!PageLocked(new));
+ VM_BUG_ON(new->mapping);
+
+ /*
+ * This is not page migration, but prepare_migration and
+ * end_migration does enough work for charge replacement.
+ *
+ * In the longer term we probably want a specialized function
+ * for moving the charge from old to new in a more efficient
+ * manner.
+ */
+ error = mem_cgroup_prepare_migration(old, new, &memcg);
+ if (error)
+ return error;
+
+ error = radix_tree_preload(GFP_KERNEL);
+ if (!error) {
+ struct address_space *mapping = old->mapping;
+ pgoff_t offset = old->index;
+
+ page_cache_get(new);
+ new->mapping = mapping;
+ new->index = offset;
+
+ spin_lock_irq(&mapping->tree_lock);
+ __remove_from_page_cache(old);
+ error = radix_tree_insert(&mapping->page_tree, offset, new);
+ BUG_ON(error);
+ mapping->nrpages++;
+ __inc_zone_page_state(new, NR_FILE_PAGES);
+ if (PageSwapBacked(new))
+ __inc_zone_page_state(new, NR_SHMEM);
+ spin_unlock_irq(&mapping->tree_lock);
+ radix_tree_preload_end();
+ page_cache_release(old);
+ mem_cgroup_end_migration(memcg, old, new, true);
+ } else {
+ mem_cgroup_end_migration(memcg, old, new, false);
+ }
+
+ return error;
+}
+EXPORT_SYMBOL_GPL(replace_page_cache_page);
+
+/**
* add_to_page_cache_locked - add a locked page to the pagecache
* @page: page to add
* @mapping: the page's address_space
Index: linux-2.6/include/linux/pagemap.h
===================================================================
--- linux-2.6.orig/include/linux/pagemap.h 2011-01-11 14:38:12.000000000 +0100
+++ linux-2.6/include/linux/pagemap.h 2011-01-11 14:38:17.000000000 +0100
@@ -457,6 +457,7 @@ int add_to_page_cache_lru(struct page *p
pgoff_t index, gfp_t gfp_mask);
extern void remove_from_page_cache(struct page *page);
extern void __remove_from_page_cache(struct page *page);
+int replace_page_cache_page(struct page *old, struct page *new);
/*
* Like add_to_page_cache_locked, but used to add newly allocated pages:
Index: linux-2.6/fs/fuse/dev.c
===================================================================
--- linux-2.6.orig/fs/fuse/dev.c 2011-01-11 14:38:12.000000000 +0100
+++ linux-2.6/fs/fuse/dev.c 2011-01-11 14:38:17.000000000 +0100
@@ -737,14 +737,12 @@ static int fuse_try_move_page(struct fus
if (WARN_ON(PageMlocked(oldpage)))
goto out_fallback_unlock;
- remove_from_page_cache(oldpage);
- page_cache_release(oldpage);
-
- err = add_to_page_cache_locked(newpage, mapping, index, GFP_KERNEL);
+ err = replace_page_cache_page(oldpage, newpage);
if (err) {
- printk(KERN_WARNING "fuse_try_move_page: failed to add page");
- goto out_fallback_unlock;
+ unlock_page(newpage);
+ return err;
}
+
page_cache_get(newpage);
if (!(buf->flags & PIPE_BUF_FLAG_LRU))
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [PATCH v3] mm: add replace_page_cache_page() function
2011-01-11 14:07 [PATCH v3] mm: add replace_page_cache_page() function Miklos Szeredi
@ 2011-01-11 14:25 ` Minchan Kim
2011-01-12 0:44 ` KAMEZAWA Hiroyuki
0 siblings, 1 reply; 9+ messages in thread
From: Minchan Kim @ 2011-01-11 14:25 UTC (permalink / raw)
To: Miklos Szeredi
Cc: akpm, kamezawa.hiroyu, nishimura, linux-fsdevel, linux-mm, linux-kernel
On Tue, Jan 11, 2011 at 03:07:54PM +0100, Miklos Szeredi wrote:
> (resent with fixed CC list, sorry for the duplicate)
>
> Thanks for the review.
>
> Here's an updated patch. Modifications since the last post:
>
> - don't pass gfp_mask (since it's only able to deal with GFP_KERNEL
> anyway)
>
I am not sure it's a good idea.
Now if we need just GFP_KERNEL, we can't make sure it in future.
Sometime we might need GFP_ATOMIC and friendd functions
(ex, add_to_page_cache_lru,add_to_page_cache_locked) already have gfp_mask.
It's a exported function so it's hard to modify it in future.
I want to keep it.
Instead of removing it, we can change mem_cgroup_prepare_migration as
getting gfp_mask.
--
Kind regards,
Minchan Kim
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [PATCH v3] mm: add replace_page_cache_page() function
2011-01-11 14:25 ` Minchan Kim
@ 2011-01-12 0:44 ` KAMEZAWA Hiroyuki
2011-01-12 14:30 ` Miklos Szeredi
0 siblings, 1 reply; 9+ messages in thread
From: KAMEZAWA Hiroyuki @ 2011-01-12 0:44 UTC (permalink / raw)
To: Minchan Kim
Cc: Miklos Szeredi, akpm, nishimura, linux-fsdevel, linux-mm, linux-kernel
On Tue, 11 Jan 2011 23:25:28 +0900
Minchan Kim <minchan.kim@gmail.com> wrote:
> On Tue, Jan 11, 2011 at 03:07:54PM +0100, Miklos Szeredi wrote:
> > (resent with fixed CC list, sorry for the duplicate)
> >
> > Thanks for the review.
> >
> > Here's an updated patch. Modifications since the last post:
> >
> > - don't pass gfp_mask (since it's only able to deal with GFP_KERNEL
> > anyway)
> >
>
> I am not sure it's a good idea.
> Now if we need just GFP_KERNEL, we can't make sure it in future.
> Sometime we might need GFP_ATOMIC and friendd functions
> (ex, add_to_page_cache_lru,add_to_page_cache_locked) already have gfp_mask.
> It's a exported function so it's hard to modify it in future.
>
> I want to keep it.
> Instead of removing it, we can change mem_cgroup_prepare_migration as
> getting gfp_mask.
>
you're right.
Thanks,
-Kame
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [PATCH v3] mm: add replace_page_cache_page() function
2011-01-12 0:44 ` KAMEZAWA Hiroyuki
@ 2011-01-12 14:30 ` Miklos Szeredi
2011-01-13 2:27 ` KAMEZAWA Hiroyuki
0 siblings, 1 reply; 9+ messages in thread
From: Miklos Szeredi @ 2011-01-12 14:30 UTC (permalink / raw)
To: KAMEZAWA Hiroyuki
Cc: minchan.kim, miklos, akpm, nishimura, linux-fsdevel, linux-mm,
linux-kernel
On Wed, 12 Jan 2011, KAMEZAWA Hiroyuki wrote:
> On Tue, 11 Jan 2011 23:25:28 +0900
> Minchan Kim <minchan.kim@gmail.com> wrote:
>
> > On Tue, Jan 11, 2011 at 03:07:54PM +0100, Miklos Szeredi wrote:
> > > (resent with fixed CC list, sorry for the duplicate)
> > >
> > > Thanks for the review.
> > >
> > > Here's an updated patch. Modifications since the last post:
> > >
> > > - don't pass gfp_mask (since it's only able to deal with GFP_KERNEL
> > > anyway)
> > >
> >
> > I am not sure it's a good idea.
> > Now if we need just GFP_KERNEL, we can't make sure it in future.
> > Sometime we might need GFP_ATOMIC and friendd functions
> > (ex, add_to_page_cache_lru,add_to_page_cache_locked) already have gfp_mask.
> > It's a exported function so it's hard to modify it in future.
> >
> > I want to keep it.
> > Instead of removing it, we can change mem_cgroup_prepare_migration as
> > getting gfp_mask.
> >
> you're right.
Okay, makes sense.
Here's an updated patch.
Thanks,
Miklos
---
From: Miklos Szeredi <mszeredi@suse.cz>
Subject: mm: add replace_page_cache_page() function
This function basically does:
remove_from_page_cache(old);
page_cache_release(old);
add_to_page_cache_locked(new);
Except it does this atomically, so there's no possibility for the
"add" to fail because of a race.
This is used by fuse to move pages into the page cache.
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
---
fs/fuse/dev.c | 10 ++----
include/linux/memcontrol.h | 4 +-
include/linux/pagemap.h | 1
mm/filemap.c | 65 +++++++++++++++++++++++++++++++++++++++++++++
mm/memcontrol.c | 4 +-
5 files changed, 74 insertions(+), 10 deletions(-)
Index: linux-2.6/mm/filemap.c
===================================================================
--- linux-2.6.orig/mm/filemap.c 2011-01-12 14:50:18.000000000 +0100
+++ linux-2.6/mm/filemap.c 2011-01-12 14:51:25.000000000 +0100
@@ -387,6 +387,71 @@ int filemap_write_and_wait_range(struct
EXPORT_SYMBOL(filemap_write_and_wait_range);
/**
+ * replace_page_cache_page - replace a pagecache page with a new one
+ * @old: page to be replaced
+ * @new: page to replace with
+ * @gfp_mask: allocation mode
+ *
+ * This function replaces a page in the pagecache with a new one. On
+ * success it acquires the pagecache reference for the new page and
+ * drops it for the old page. Both the old and new pages must be
+ * locked. This function does not add the new page to the LRU, the
+ * caller must do that.
+ *
+ * The remove + add is atomic. The only way this function can fail is
+ * memory allocation failure.
+ */
+int replace_page_cache_page(struct page *old, struct page *new, gfp_t gfp_mask)
+{
+ int error;
+ struct mem_cgroup *memcg = NULL;
+
+ VM_BUG_ON(!PageLocked(old));
+ VM_BUG_ON(!PageLocked(new));
+ VM_BUG_ON(new->mapping);
+
+ /*
+ * This is not page migration, but prepare_migration and
+ * end_migration does enough work for charge replacement.
+ *
+ * In the longer term we probably want a specialized function
+ * for moving the charge from old to new in a more efficient
+ * manner.
+ */
+ error = mem_cgroup_prepare_migration(old, new, &memcg, gfp_mask);
+ if (error)
+ return error;
+
+ error = radix_tree_preload(gfp_mask & ~__GFP_HIGHMEM);
+ if (!error) {
+ struct address_space *mapping = old->mapping;
+ pgoff_t offset = old->index;
+
+ page_cache_get(new);
+ new->mapping = mapping;
+ new->index = offset;
+
+ spin_lock_irq(&mapping->tree_lock);
+ __remove_from_page_cache(old);
+ error = radix_tree_insert(&mapping->page_tree, offset, new);
+ BUG_ON(error);
+ mapping->nrpages++;
+ __inc_zone_page_state(new, NR_FILE_PAGES);
+ if (PageSwapBacked(new))
+ __inc_zone_page_state(new, NR_SHMEM);
+ spin_unlock_irq(&mapping->tree_lock);
+ radix_tree_preload_end();
+ page_cache_release(old);
+ mem_cgroup_end_migration(memcg, old, new, true);
+ } else {
+ mem_cgroup_end_migration(memcg, old, new, false);
+ }
+
+ return error;
+}
+EXPORT_SYMBOL_GPL(replace_page_cache_page);
+
+/**
* add_to_page_cache_locked - add a locked page to the pagecache
* @page: page to add
* @mapping: the page's address_space
Index: linux-2.6/include/linux/pagemap.h
===================================================================
--- linux-2.6.orig/include/linux/pagemap.h 2011-01-12 14:50:18.000000000 +0100
+++ linux-2.6/include/linux/pagemap.h 2011-01-12 14:50:31.000000000 +0100
@@ -457,6 +457,7 @@ int add_to_page_cache_lru(struct page *p
pgoff_t index, gfp_t gfp_mask);
extern void remove_from_page_cache(struct page *page);
extern void __remove_from_page_cache(struct page *page);
+int replace_page_cache_page(struct page *old, struct page *new, gfp_t gfp_mask);
/*
* Like add_to_page_cache_locked, but used to add newly allocated pages:
Index: linux-2.6/fs/fuse/dev.c
===================================================================
--- linux-2.6.orig/fs/fuse/dev.c 2011-01-12 14:50:18.000000000 +0100
+++ linux-2.6/fs/fuse/dev.c 2011-01-12 14:50:31.000000000 +0100
@@ -737,14 +737,12 @@ static int fuse_try_move_page(struct fus
if (WARN_ON(PageMlocked(oldpage)))
goto out_fallback_unlock;
- remove_from_page_cache(oldpage);
- page_cache_release(oldpage);
-
- err = add_to_page_cache_locked(newpage, mapping, index, GFP_KERNEL);
+ err = replace_page_cache_page(oldpage, newpage, GFP_KERNEL);
if (err) {
- printk(KERN_WARNING "fuse_try_move_page: failed to add page");
- goto out_fallback_unlock;
+ unlock_page(newpage);
+ return err;
}
+
page_cache_get(newpage);
if (!(buf->flags & PIPE_BUF_FLAG_LRU))
Index: linux-2.6/include/linux/memcontrol.h
===================================================================
--- linux-2.6.orig/include/linux/memcontrol.h 2011-01-12 14:50:19.000000000 +0100
+++ linux-2.6/include/linux/memcontrol.h 2011-01-12 14:53:13.000000000 +0100
@@ -91,7 +91,7 @@ extern struct cgroup_subsys_state *mem_c
extern int
mem_cgroup_prepare_migration(struct page *page,
- struct page *newpage, struct mem_cgroup **ptr);
+ struct page *newpage, struct mem_cgroup **ptr, gfp_t gfp_mask);
extern void mem_cgroup_end_migration(struct mem_cgroup *mem,
struct page *oldpage, struct page *newpage, bool migration_ok);
@@ -225,7 +225,7 @@ static inline struct cgroup_subsys_state
static inline int
mem_cgroup_prepare_migration(struct page *page, struct page *newpage,
- struct mem_cgroup **ptr)
+ struct mem_cgroup **ptr, gfp_t gfp_mask)
{
return 0;
}
Index: linux-2.6/mm/memcontrol.c
===================================================================
--- linux-2.6.orig/mm/memcontrol.c 2011-01-12 14:50:19.000000000 +0100
+++ linux-2.6/mm/memcontrol.c 2011-01-12 14:50:39.000000000 +0100
@@ -2767,7 +2767,7 @@ static inline int mem_cgroup_move_swap_a
* page belongs to.
*/
int mem_cgroup_prepare_migration(struct page *page,
- struct page *newpage, struct mem_cgroup **ptr)
+ struct page *newpage, struct mem_cgroup **ptr, gfp_t gfp_mask)
{
struct page_cgroup *pc;
struct mem_cgroup *mem = NULL;
@@ -2823,7 +2823,7 @@ int mem_cgroup_prepare_migration(struct
return 0;
*ptr = mem;
- ret = __mem_cgroup_try_charge(NULL, GFP_KERNEL, ptr, false);
+ ret = __mem_cgroup_try_charge(NULL, gfp_mask, ptr, false);
css_put(&mem->css);/* drop extra refcnt */
if (ret || *ptr == NULL) {
if (PageAnon(page)) {
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [PATCH v3] mm: add replace_page_cache_page() function
2011-01-12 14:30 ` Miklos Szeredi
@ 2011-01-13 2:27 ` KAMEZAWA Hiroyuki
2011-01-13 13:35 ` Miklos Szeredi
0 siblings, 1 reply; 9+ messages in thread
From: KAMEZAWA Hiroyuki @ 2011-01-13 2:27 UTC (permalink / raw)
To: Miklos Szeredi
Cc: minchan.kim, akpm, nishimura, linux-fsdevel, linux-mm, linux-kernel
On Wed, 12 Jan 2011 15:30:11 +0100
Miklos Szeredi <miklos@szeredi.hu> wrote:
> On Wed, 12 Jan 2011, KAMEZAWA Hiroyuki wrote:
> > On Tue, 11 Jan 2011 23:25:28 +0900
> > Minchan Kim <minchan.kim@gmail.com> wrote:
> >
> > > On Tue, Jan 11, 2011 at 03:07:54PM +0100, Miklos Szeredi wrote:
> > > > (resent with fixed CC list, sorry for the duplicate)
> > > >
> > > > Thanks for the review.
> > > >
> > > > Here's an updated patch. Modifications since the last post:
> > > >
> > > > - don't pass gfp_mask (since it's only able to deal with GFP_KERNEL
> > > > anyway)
> > > >
> > >
> > > I am not sure it's a good idea.
> > > Now if we need just GFP_KERNEL, we can't make sure it in future.
> > > Sometime we might need GFP_ATOMIC and friendd functions
> > > (ex, add_to_page_cache_lru,add_to_page_cache_locked) already have gfp_mask.
> > > It's a exported function so it's hard to modify it in future.
> > >
> > > I want to keep it.
> > > Instead of removing it, we can change mem_cgroup_prepare_migration as
> > > getting gfp_mask.
> > >
> > you're right.
>
> Okay, makes sense.
>
> Here's an updated patch.
>
Seems nicer. But you didn't fixed the caller of prepare_migraton()
in mm/migrate.c (I think this is the only caller..)
please modify it, too.
To check your patch, CONFIG_MIGRATE should be set.
Thanks,
-Kame
> Thanks,
> Miklos
> ---
>
> From: Miklos Szeredi <mszeredi@suse.cz>
> Subject: mm: add replace_page_cache_page() function
>
> This function basically does:
>
> remove_from_page_cache(old);
> page_cache_release(old);
> add_to_page_cache_locked(new);
>
> Except it does this atomically, so there's no possibility for the
> "add" to fail because of a race.
>
> This is used by fuse to move pages into the page cache.
>
> Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
> ---
> fs/fuse/dev.c | 10 ++----
> include/linux/memcontrol.h | 4 +-
> include/linux/pagemap.h | 1
> mm/filemap.c | 65 +++++++++++++++++++++++++++++++++++++++++++++
> mm/memcontrol.c | 4 +-
> 5 files changed, 74 insertions(+), 10 deletions(-)
>
> Index: linux-2.6/mm/filemap.c
> ===================================================================
> --- linux-2.6.orig/mm/filemap.c 2011-01-12 14:50:18.000000000 +0100
> +++ linux-2.6/mm/filemap.c 2011-01-12 14:51:25.000000000 +0100
> @@ -387,6 +387,71 @@ int filemap_write_and_wait_range(struct
> EXPORT_SYMBOL(filemap_write_and_wait_range);
>
> /**
> + * replace_page_cache_page - replace a pagecache page with a new one
> + * @old: page to be replaced
> + * @new: page to replace with
> + * @gfp_mask: allocation mode
> + *
> + * This function replaces a page in the pagecache with a new one. On
> + * success it acquires the pagecache reference for the new page and
> + * drops it for the old page. Both the old and new pages must be
> + * locked. This function does not add the new page to the LRU, the
> + * caller must do that.
> + *
> + * The remove + add is atomic. The only way this function can fail is
> + * memory allocation failure.
> + */
> +int replace_page_cache_page(struct page *old, struct page *new, gfp_t gfp_mask)
> +{
> + int error;
> + struct mem_cgroup *memcg = NULL;
> +
> + VM_BUG_ON(!PageLocked(old));
> + VM_BUG_ON(!PageLocked(new));
> + VM_BUG_ON(new->mapping);
> +
> + /*
> + * This is not page migration, but prepare_migration and
> + * end_migration does enough work for charge replacement.
> + *
> + * In the longer term we probably want a specialized function
> + * for moving the charge from old to new in a more efficient
> + * manner.
> + */
> + error = mem_cgroup_prepare_migration(old, new, &memcg, gfp_mask);
> + if (error)
> + return error;
> +
> + error = radix_tree_preload(gfp_mask & ~__GFP_HIGHMEM);
> + if (!error) {
> + struct address_space *mapping = old->mapping;
> + pgoff_t offset = old->index;
> +
> + page_cache_get(new);
> + new->mapping = mapping;
> + new->index = offset;
> +
> + spin_lock_irq(&mapping->tree_lock);
> + __remove_from_page_cache(old);
> + error = radix_tree_insert(&mapping->page_tree, offset, new);
> + BUG_ON(error);
> + mapping->nrpages++;
> + __inc_zone_page_state(new, NR_FILE_PAGES);
> + if (PageSwapBacked(new))
> + __inc_zone_page_state(new, NR_SHMEM);
> + spin_unlock_irq(&mapping->tree_lock);
> + radix_tree_preload_end();
> + page_cache_release(old);
> + mem_cgroup_end_migration(memcg, old, new, true);
> + } else {
> + mem_cgroup_end_migration(memcg, old, new, false);
> + }
> +
> + return error;
> +}
> +EXPORT_SYMBOL_GPL(replace_page_cache_page);
> +
> +/**
> * add_to_page_cache_locked - add a locked page to the pagecache
> * @page: page to add
> * @mapping: the page's address_space
> Index: linux-2.6/include/linux/pagemap.h
> ===================================================================
> --- linux-2.6.orig/include/linux/pagemap.h 2011-01-12 14:50:18.000000000 +0100
> +++ linux-2.6/include/linux/pagemap.h 2011-01-12 14:50:31.000000000 +0100
> @@ -457,6 +457,7 @@ int add_to_page_cache_lru(struct page *p
> pgoff_t index, gfp_t gfp_mask);
> extern void remove_from_page_cache(struct page *page);
> extern void __remove_from_page_cache(struct page *page);
> +int replace_page_cache_page(struct page *old, struct page *new, gfp_t gfp_mask);
>
> /*
> * Like add_to_page_cache_locked, but used to add newly allocated pages:
> Index: linux-2.6/fs/fuse/dev.c
> ===================================================================
> --- linux-2.6.orig/fs/fuse/dev.c 2011-01-12 14:50:18.000000000 +0100
> +++ linux-2.6/fs/fuse/dev.c 2011-01-12 14:50:31.000000000 +0100
> @@ -737,14 +737,12 @@ static int fuse_try_move_page(struct fus
> if (WARN_ON(PageMlocked(oldpage)))
> goto out_fallback_unlock;
>
> - remove_from_page_cache(oldpage);
> - page_cache_release(oldpage);
> -
> - err = add_to_page_cache_locked(newpage, mapping, index, GFP_KERNEL);
> + err = replace_page_cache_page(oldpage, newpage, GFP_KERNEL);
> if (err) {
> - printk(KERN_WARNING "fuse_try_move_page: failed to add page");
> - goto out_fallback_unlock;
> + unlock_page(newpage);
> + return err;
> }
> +
> page_cache_get(newpage);
>
> if (!(buf->flags & PIPE_BUF_FLAG_LRU))
> Index: linux-2.6/include/linux/memcontrol.h
> ===================================================================
> --- linux-2.6.orig/include/linux/memcontrol.h 2011-01-12 14:50:19.000000000 +0100
> +++ linux-2.6/include/linux/memcontrol.h 2011-01-12 14:53:13.000000000 +0100
> @@ -91,7 +91,7 @@ extern struct cgroup_subsys_state *mem_c
>
> extern int
> mem_cgroup_prepare_migration(struct page *page,
> - struct page *newpage, struct mem_cgroup **ptr);
> + struct page *newpage, struct mem_cgroup **ptr, gfp_t gfp_mask);
> extern void mem_cgroup_end_migration(struct mem_cgroup *mem,
> struct page *oldpage, struct page *newpage, bool migration_ok);
>
> @@ -225,7 +225,7 @@ static inline struct cgroup_subsys_state
>
> static inline int
> mem_cgroup_prepare_migration(struct page *page, struct page *newpage,
> - struct mem_cgroup **ptr)
> + struct mem_cgroup **ptr, gfp_t gfp_mask)
> {
> return 0;
> }
> Index: linux-2.6/mm/memcontrol.c
> ===================================================================
> --- linux-2.6.orig/mm/memcontrol.c 2011-01-12 14:50:19.000000000 +0100
> +++ linux-2.6/mm/memcontrol.c 2011-01-12 14:50:39.000000000 +0100
> @@ -2767,7 +2767,7 @@ static inline int mem_cgroup_move_swap_a
> * page belongs to.
> */
> int mem_cgroup_prepare_migration(struct page *page,
> - struct page *newpage, struct mem_cgroup **ptr)
> + struct page *newpage, struct mem_cgroup **ptr, gfp_t gfp_mask)
> {
> struct page_cgroup *pc;
> struct mem_cgroup *mem = NULL;
> @@ -2823,7 +2823,7 @@ int mem_cgroup_prepare_migration(struct
> return 0;
>
> *ptr = mem;
> - ret = __mem_cgroup_try_charge(NULL, GFP_KERNEL, ptr, false);
> + ret = __mem_cgroup_try_charge(NULL, gfp_mask, ptr, false);
> css_put(&mem->css);/* drop extra refcnt */
> if (ret || *ptr == NULL) {
> if (PageAnon(page)) {
>
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [PATCH v3] mm: add replace_page_cache_page() function
2011-01-13 2:27 ` KAMEZAWA Hiroyuki
@ 2011-01-13 13:35 ` Miklos Szeredi
2011-01-14 1:00 ` KAMEZAWA Hiroyuki
2011-01-14 1:00 ` Daisuke Nishimura
0 siblings, 2 replies; 9+ messages in thread
From: Miklos Szeredi @ 2011-01-13 13:35 UTC (permalink / raw)
To: KAMEZAWA Hiroyuki
Cc: miklos, minchan.kim, akpm, nishimura, linux-fsdevel, linux-mm,
linux-kernel
On Thu, 13 Jan 2011, KAMEZAWA Hiroyuki wrote:
> On Wed, 12 Jan 2011 15:30:11 +0100
> Miklos Szeredi <miklos@szeredi.hu> wrote:
>
> > On Wed, 12 Jan 2011, KAMEZAWA Hiroyuki wrote:
> > > On Tue, 11 Jan 2011 23:25:28 +0900
> > > Minchan Kim <minchan.kim@gmail.com> wrote:
> > >
> > > > On Tue, Jan 11, 2011 at 03:07:54PM +0100, Miklos Szeredi wrote:
> > > > > (resent with fixed CC list, sorry for the duplicate)
> > > > >
> > > > > Thanks for the review.
> > > > >
> > > > > Here's an updated patch. Modifications since the last post:
> > > > >
> > > > > - don't pass gfp_mask (since it's only able to deal with GFP_KERNEL
> > > > > anyway)
> > > > >
> > > >
> > > > I am not sure it's a good idea.
> > > > Now if we need just GFP_KERNEL, we can't make sure it in future.
> > > > Sometime we might need GFP_ATOMIC and friendd functions
> > > > (ex, add_to_page_cache_lru,add_to_page_cache_locked) already have gfp_mask.
> > > > It's a exported function so it's hard to modify it in future.
> > > >
> > > > I want to keep it.
> > > > Instead of removing it, we can change mem_cgroup_prepare_migration as
> > > > getting gfp_mask.
> > > >
> > > you're right.
> >
> > Okay, makes sense.
> >
> > Here's an updated patch.
> >
>
> Seems nicer. But you didn't fixed the caller of prepare_migraton()
> in mm/migrate.c (I think this is the only caller..)
>
> please modify it, too.
Right you are. Thanks for the review.
Updated patch attached.
Thanks,
Miklos
---
From: Miklos Szeredi <mszeredi@suse.cz>
Subject: mm: add replace_page_cache_page() function
This function basically does:
remove_from_page_cache(old);
page_cache_release(old);
add_to_page_cache_locked(new);
Except it does this atomically, so there's no possibility for the
"add" to fail because of a race.
This is used by fuse to move pages into the page cache.
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
---
fs/fuse/dev.c | 10 ++----
include/linux/memcontrol.h | 4 +-
include/linux/pagemap.h | 1
mm/filemap.c | 65 +++++++++++++++++++++++++++++++++++++++++++++
mm/memcontrol.c | 4 +-
mm/migrate.c | 2 -
6 files changed, 75 insertions(+), 11 deletions(-)
Index: linux-2.6/mm/filemap.c
===================================================================
--- linux-2.6.orig/mm/filemap.c 2011-01-12 15:28:56.000000000 +0100
+++ linux-2.6/mm/filemap.c 2011-01-13 13:37:41.000000000 +0100
@@ -387,6 +387,71 @@ int filemap_write_and_wait_range(struct
EXPORT_SYMBOL(filemap_write_and_wait_range);
/**
+ * replace_page_cache_page - replace a pagecache page with a new one
+ * @old: page to be replaced
+ * @new: page to replace with
+ * @gfp_mask: allocation mode
+ *
+ * This function replaces a page in the pagecache with a new one. On
+ * success it acquires the pagecache reference for the new page and
+ * drops it for the old page. Both the old and new pages must be
+ * locked. This function does not add the new page to the LRU, the
+ * caller must do that.
+ *
+ * The remove + add is atomic. The only way this function can fail is
+ * memory allocation failure.
+ */
+int replace_page_cache_page(struct page *old, struct page *new, gfp_t gfp_mask)
+{
+ int error;
+ struct mem_cgroup *memcg = NULL;
+
+ VM_BUG_ON(!PageLocked(old));
+ VM_BUG_ON(!PageLocked(new));
+ VM_BUG_ON(new->mapping);
+
+ /*
+ * This is not page migration, but prepare_migration and
+ * end_migration does enough work for charge replacement.
+ *
+ * In the longer term we probably want a specialized function
+ * for moving the charge from old to new in a more efficient
+ * manner.
+ */
+ error = mem_cgroup_prepare_migration(old, new, &memcg, gfp_mask);
+ if (error)
+ return error;
+
+ error = radix_tree_preload(gfp_mask & ~__GFP_HIGHMEM);
+ if (!error) {
+ struct address_space *mapping = old->mapping;
+ pgoff_t offset = old->index;
+
+ page_cache_get(new);
+ new->mapping = mapping;
+ new->index = offset;
+
+ spin_lock_irq(&mapping->tree_lock);
+ __remove_from_page_cache(old);
+ error = radix_tree_insert(&mapping->page_tree, offset, new);
+ BUG_ON(error);
+ mapping->nrpages++;
+ __inc_zone_page_state(new, NR_FILE_PAGES);
+ if (PageSwapBacked(new))
+ __inc_zone_page_state(new, NR_SHMEM);
+ spin_unlock_irq(&mapping->tree_lock);
+ radix_tree_preload_end();
+ page_cache_release(old);
+ mem_cgroup_end_migration(memcg, old, new, true);
+ } else {
+ mem_cgroup_end_migration(memcg, old, new, false);
+ }
+
+ return error;
+}
+EXPORT_SYMBOL_GPL(replace_page_cache_page);
+
+/**
* add_to_page_cache_locked - add a locked page to the pagecache
* @page: page to add
* @mapping: the page's address_space
Index: linux-2.6/include/linux/pagemap.h
===================================================================
--- linux-2.6.orig/include/linux/pagemap.h 2011-01-12 15:28:56.000000000 +0100
+++ linux-2.6/include/linux/pagemap.h 2011-01-13 13:37:41.000000000 +0100
@@ -457,6 +457,7 @@ int add_to_page_cache_lru(struct page *p
pgoff_t index, gfp_t gfp_mask);
extern void remove_from_page_cache(struct page *page);
extern void __remove_from_page_cache(struct page *page);
+int replace_page_cache_page(struct page *old, struct page *new, gfp_t gfp_mask);
/*
* Like add_to_page_cache_locked, but used to add newly allocated pages:
Index: linux-2.6/fs/fuse/dev.c
===================================================================
--- linux-2.6.orig/fs/fuse/dev.c 2011-01-12 15:28:56.000000000 +0100
+++ linux-2.6/fs/fuse/dev.c 2011-01-13 13:37:41.000000000 +0100
@@ -737,14 +737,12 @@ static int fuse_try_move_page(struct fus
if (WARN_ON(PageMlocked(oldpage)))
goto out_fallback_unlock;
- remove_from_page_cache(oldpage);
- page_cache_release(oldpage);
-
- err = add_to_page_cache_locked(newpage, mapping, index, GFP_KERNEL);
+ err = replace_page_cache_page(oldpage, newpage, GFP_KERNEL);
if (err) {
- printk(KERN_WARNING "fuse_try_move_page: failed to add page");
- goto out_fallback_unlock;
+ unlock_page(newpage);
+ return err;
}
+
page_cache_get(newpage);
if (!(buf->flags & PIPE_BUF_FLAG_LRU))
Index: linux-2.6/include/linux/memcontrol.h
===================================================================
--- linux-2.6.orig/include/linux/memcontrol.h 2011-01-13 13:37:40.000000000 +0100
+++ linux-2.6/include/linux/memcontrol.h 2011-01-13 13:37:41.000000000 +0100
@@ -91,7 +91,7 @@ extern struct cgroup_subsys_state *mem_c
extern int
mem_cgroup_prepare_migration(struct page *page,
- struct page *newpage, struct mem_cgroup **ptr);
+ struct page *newpage, struct mem_cgroup **ptr, gfp_t gfp_mask);
extern void mem_cgroup_end_migration(struct mem_cgroup *mem,
struct page *oldpage, struct page *newpage, bool migration_ok);
@@ -225,7 +225,7 @@ static inline struct cgroup_subsys_state
static inline int
mem_cgroup_prepare_migration(struct page *page, struct page *newpage,
- struct mem_cgroup **ptr)
+ struct mem_cgroup **ptr, gfp_t gfp_mask)
{
return 0;
}
Index: linux-2.6/mm/memcontrol.c
===================================================================
--- linux-2.6.orig/mm/memcontrol.c 2011-01-13 13:37:40.000000000 +0100
+++ linux-2.6/mm/memcontrol.c 2011-01-13 13:37:41.000000000 +0100
@@ -2767,7 +2767,7 @@ static inline int mem_cgroup_move_swap_a
* page belongs to.
*/
int mem_cgroup_prepare_migration(struct page *page,
- struct page *newpage, struct mem_cgroup **ptr)
+ struct page *newpage, struct mem_cgroup **ptr, gfp_t gfp_mask)
{
struct page_cgroup *pc;
struct mem_cgroup *mem = NULL;
@@ -2823,7 +2823,7 @@ int mem_cgroup_prepare_migration(struct
return 0;
*ptr = mem;
- ret = __mem_cgroup_try_charge(NULL, GFP_KERNEL, ptr, false);
+ ret = __mem_cgroup_try_charge(NULL, gfp_mask, ptr, false);
css_put(&mem->css);/* drop extra refcnt */
if (ret || *ptr == NULL) {
if (PageAnon(page)) {
Index: linux-2.6/mm/migrate.c
===================================================================
--- linux-2.6.orig/mm/migrate.c 2011-01-13 13:37:40.000000000 +0100
+++ linux-2.6/mm/migrate.c 2011-01-13 13:38:55.000000000 +0100
@@ -657,7 +657,7 @@ static int unmap_and_move(new_page_t get
}
/* charge against new page */
- charge = mem_cgroup_prepare_migration(page, newpage, &mem);
+ charge = mem_cgroup_prepare_migration(page, newpage, &mem, GFP_KERNEL);
if (charge == -ENOMEM) {
rc = -ENOMEM;
goto unlock;
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [PATCH v3] mm: add replace_page_cache_page() function
2011-01-13 13:35 ` Miklos Szeredi
@ 2011-01-14 1:00 ` KAMEZAWA Hiroyuki
2011-01-14 1:00 ` Daisuke Nishimura
1 sibling, 0 replies; 9+ messages in thread
From: KAMEZAWA Hiroyuki @ 2011-01-14 1:00 UTC (permalink / raw)
To: Miklos Szeredi
Cc: minchan.kim, akpm, nishimura, linux-fsdevel, linux-mm, linux-kernel
On Thu, 13 Jan 2011 14:35:09 +0100
Miklos Szeredi <miklos@szeredi.hu> wrote:
> On Thu, 13 Jan 2011, KAMEZAWA Hiroyuki wrote:
> > On Wed, 12 Jan 2011 15:30:11 +0100
> > Miklos Szeredi <miklos@szeredi.hu> wrote:
> >
> > > On Wed, 12 Jan 2011, KAMEZAWA Hiroyuki wrote:
> > > > On Tue, 11 Jan 2011 23:25:28 +0900
> > > > Minchan Kim <minchan.kim@gmail.com> wrote:
> > > >
> > > > > On Tue, Jan 11, 2011 at 03:07:54PM +0100, Miklos Szeredi wrote:
> > > > > > (resent with fixed CC list, sorry for the duplicate)
> > > > > >
> > > > > > Thanks for the review.
> > > > > >
> > > > > > Here's an updated patch. Modifications since the last post:
> > > > > >
> > > > > > - don't pass gfp_mask (since it's only able to deal with GFP_KERNEL
> > > > > > anyway)
> > > > > >
> > > > >
> > > > > I am not sure it's a good idea.
> > > > > Now if we need just GFP_KERNEL, we can't make sure it in future.
> > > > > Sometime we might need GFP_ATOMIC and friendd functions
> > > > > (ex, add_to_page_cache_lru,add_to_page_cache_locked) already have gfp_mask.
> > > > > It's a exported function so it's hard to modify it in future.
> > > > >
> > > > > I want to keep it.
> > > > > Instead of removing it, we can change mem_cgroup_prepare_migration as
> > > > > getting gfp_mask.
> > > > >
> > > > you're right.
> > >
> > > Okay, makes sense.
> > >
> > > Here's an updated patch.
> > >
> >
> > Seems nicer. But you didn't fixed the caller of prepare_migraton()
> > in mm/migrate.c (I think this is the only caller..)
> >
> > please modify it, too.
>
> Right you are. Thanks for the review.
>
> Updated patch attached.
>
> Thanks,
> Miklos
> ---
>
>
> From: Miklos Szeredi <mszeredi@suse.cz>
> Subject: mm: add replace_page_cache_page() function
>
> This function basically does:
>
> remove_from_page_cache(old);
> page_cache_release(old);
> add_to_page_cache_locked(new);
>
> Except it does this atomically, so there's no possibility for the
> "add" to fail because of a race.
>
> This is used by fuse to move pages into the page cache.
>
> Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
Thank you.
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
But my last concern is that GFP_KERNEL is ok or not for compaction...
(It was implicitly used.)
Maybe migrate_page() should have gfp_mask but it should be in another work.
Thanks,
-Kame
> ---
> fs/fuse/dev.c | 10 ++----
> include/linux/memcontrol.h | 4 +-
> include/linux/pagemap.h | 1
> mm/filemap.c | 65 +++++++++++++++++++++++++++++++++++++++++++++
> mm/memcontrol.c | 4 +-
> mm/migrate.c | 2 -
> 6 files changed, 75 insertions(+), 11 deletions(-)
>
> Index: linux-2.6/mm/filemap.c
> ===================================================================
> --- linux-2.6.orig/mm/filemap.c 2011-01-12 15:28:56.000000000 +0100
> +++ linux-2.6/mm/filemap.c 2011-01-13 13:37:41.000000000 +0100
> @@ -387,6 +387,71 @@ int filemap_write_and_wait_range(struct
> EXPORT_SYMBOL(filemap_write_and_wait_range);
>
> /**
> + * replace_page_cache_page - replace a pagecache page with a new one
> + * @old: page to be replaced
> + * @new: page to replace with
> + * @gfp_mask: allocation mode
> + *
> + * This function replaces a page in the pagecache with a new one. On
> + * success it acquires the pagecache reference for the new page and
> + * drops it for the old page. Both the old and new pages must be
> + * locked. This function does not add the new page to the LRU, the
> + * caller must do that.
> + *
> + * The remove + add is atomic. The only way this function can fail is
> + * memory allocation failure.
> + */
> +int replace_page_cache_page(struct page *old, struct page *new, gfp_t gfp_mask)
> +{
> + int error;
> + struct mem_cgroup *memcg = NULL;
> +
> + VM_BUG_ON(!PageLocked(old));
> + VM_BUG_ON(!PageLocked(new));
> + VM_BUG_ON(new->mapping);
> +
> + /*
> + * This is not page migration, but prepare_migration and
> + * end_migration does enough work for charge replacement.
> + *
> + * In the longer term we probably want a specialized function
> + * for moving the charge from old to new in a more efficient
> + * manner.
> + */
> + error = mem_cgroup_prepare_migration(old, new, &memcg, gfp_mask);
> + if (error)
> + return error;
> +
> + error = radix_tree_preload(gfp_mask & ~__GFP_HIGHMEM);
> + if (!error) {
> + struct address_space *mapping = old->mapping;
> + pgoff_t offset = old->index;
> +
> + page_cache_get(new);
> + new->mapping = mapping;
> + new->index = offset;
> +
> + spin_lock_irq(&mapping->tree_lock);
> + __remove_from_page_cache(old);
> + error = radix_tree_insert(&mapping->page_tree, offset, new);
> + BUG_ON(error);
> + mapping->nrpages++;
> + __inc_zone_page_state(new, NR_FILE_PAGES);
> + if (PageSwapBacked(new))
> + __inc_zone_page_state(new, NR_SHMEM);
> + spin_unlock_irq(&mapping->tree_lock);
> + radix_tree_preload_end();
> + page_cache_release(old);
> + mem_cgroup_end_migration(memcg, old, new, true);
> + } else {
> + mem_cgroup_end_migration(memcg, old, new, false);
> + }
> +
> + return error;
> +}
> +EXPORT_SYMBOL_GPL(replace_page_cache_page);
> +
> +/**
> * add_to_page_cache_locked - add a locked page to the pagecache
> * @page: page to add
> * @mapping: the page's address_space
> Index: linux-2.6/include/linux/pagemap.h
> ===================================================================
> --- linux-2.6.orig/include/linux/pagemap.h 2011-01-12 15:28:56.000000000 +0100
> +++ linux-2.6/include/linux/pagemap.h 2011-01-13 13:37:41.000000000 +0100
> @@ -457,6 +457,7 @@ int add_to_page_cache_lru(struct page *p
> pgoff_t index, gfp_t gfp_mask);
> extern void remove_from_page_cache(struct page *page);
> extern void __remove_from_page_cache(struct page *page);
> +int replace_page_cache_page(struct page *old, struct page *new, gfp_t gfp_mask);
>
> /*
> * Like add_to_page_cache_locked, but used to add newly allocated pages:
> Index: linux-2.6/fs/fuse/dev.c
> ===================================================================
> --- linux-2.6.orig/fs/fuse/dev.c 2011-01-12 15:28:56.000000000 +0100
> +++ linux-2.6/fs/fuse/dev.c 2011-01-13 13:37:41.000000000 +0100
> @@ -737,14 +737,12 @@ static int fuse_try_move_page(struct fus
> if (WARN_ON(PageMlocked(oldpage)))
> goto out_fallback_unlock;
>
> - remove_from_page_cache(oldpage);
> - page_cache_release(oldpage);
> -
> - err = add_to_page_cache_locked(newpage, mapping, index, GFP_KERNEL);
> + err = replace_page_cache_page(oldpage, newpage, GFP_KERNEL);
> if (err) {
> - printk(KERN_WARNING "fuse_try_move_page: failed to add page");
> - goto out_fallback_unlock;
> + unlock_page(newpage);
> + return err;
> }
> +
> page_cache_get(newpage);
>
> if (!(buf->flags & PIPE_BUF_FLAG_LRU))
> Index: linux-2.6/include/linux/memcontrol.h
> ===================================================================
> --- linux-2.6.orig/include/linux/memcontrol.h 2011-01-13 13:37:40.000000000 +0100
> +++ linux-2.6/include/linux/memcontrol.h 2011-01-13 13:37:41.000000000 +0100
> @@ -91,7 +91,7 @@ extern struct cgroup_subsys_state *mem_c
>
> extern int
> mem_cgroup_prepare_migration(struct page *page,
> - struct page *newpage, struct mem_cgroup **ptr);
> + struct page *newpage, struct mem_cgroup **ptr, gfp_t gfp_mask);
> extern void mem_cgroup_end_migration(struct mem_cgroup *mem,
> struct page *oldpage, struct page *newpage, bool migration_ok);
>
> @@ -225,7 +225,7 @@ static inline struct cgroup_subsys_state
>
> static inline int
> mem_cgroup_prepare_migration(struct page *page, struct page *newpage,
> - struct mem_cgroup **ptr)
> + struct mem_cgroup **ptr, gfp_t gfp_mask)
> {
> return 0;
> }
> Index: linux-2.6/mm/memcontrol.c
> ===================================================================
> --- linux-2.6.orig/mm/memcontrol.c 2011-01-13 13:37:40.000000000 +0100
> +++ linux-2.6/mm/memcontrol.c 2011-01-13 13:37:41.000000000 +0100
> @@ -2767,7 +2767,7 @@ static inline int mem_cgroup_move_swap_a
> * page belongs to.
> */
> int mem_cgroup_prepare_migration(struct page *page,
> - struct page *newpage, struct mem_cgroup **ptr)
> + struct page *newpage, struct mem_cgroup **ptr, gfp_t gfp_mask)
> {
> struct page_cgroup *pc;
> struct mem_cgroup *mem = NULL;
> @@ -2823,7 +2823,7 @@ int mem_cgroup_prepare_migration(struct
> return 0;
>
> *ptr = mem;
> - ret = __mem_cgroup_try_charge(NULL, GFP_KERNEL, ptr, false);
> + ret = __mem_cgroup_try_charge(NULL, gfp_mask, ptr, false);
> css_put(&mem->css);/* drop extra refcnt */
> if (ret || *ptr == NULL) {
> if (PageAnon(page)) {
> Index: linux-2.6/mm/migrate.c
> ===================================================================
> --- linux-2.6.orig/mm/migrate.c 2011-01-13 13:37:40.000000000 +0100
> +++ linux-2.6/mm/migrate.c 2011-01-13 13:38:55.000000000 +0100
> @@ -657,7 +657,7 @@ static int unmap_and_move(new_page_t get
> }
>
> /* charge against new page */
> - charge = mem_cgroup_prepare_migration(page, newpage, &mem);
> + charge = mem_cgroup_prepare_migration(page, newpage, &mem, GFP_KERNEL);
> if (charge == -ENOMEM) {
> rc = -ENOMEM;
> goto unlock;
> --
> 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/
>
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [PATCH v3] mm: add replace_page_cache_page() function
2011-01-13 13:35 ` Miklos Szeredi
2011-01-14 1:00 ` KAMEZAWA Hiroyuki
@ 2011-01-14 1:00 ` Daisuke Nishimura
2011-01-17 13:59 ` Minchan Kim
1 sibling, 1 reply; 9+ messages in thread
From: Daisuke Nishimura @ 2011-01-14 1:00 UTC (permalink / raw)
To: Miklos Szeredi
Cc: KAMEZAWA Hiroyuki, minchan.kim, akpm, linux-fsdevel, linux-mm,
linux-kernel, Daisuke Nishimura
On Thu, 13 Jan 2011 14:35:09 +0100
Miklos Szeredi <miklos@szeredi.hu> wrote:
> > Seems nicer. But you didn't fixed the caller of prepare_migraton()
> > in mm/migrate.c (I think this is the only caller..)
> >
> > please modify it, too.
>
> Right you are. Thanks for the review.
>
> Updated patch attached.
>
> Thanks,
> Miklos
> ---
>
>
> From: Miklos Szeredi <mszeredi@suse.cz>
> Subject: mm: add replace_page_cache_page() function
>
> This function basically does:
>
> remove_from_page_cache(old);
> page_cache_release(old);
> add_to_page_cache_locked(new);
>
> Except it does this atomically, so there's no possibility for the
> "add" to fail because of a race.
>
> This is used by fuse to move pages into the page cache.
>
> Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
> ---
> fs/fuse/dev.c | 10 ++----
> include/linux/memcontrol.h | 4 +-
> include/linux/pagemap.h | 1
> mm/filemap.c | 65 +++++++++++++++++++++++++++++++++++++++++++++
> mm/memcontrol.c | 4 +-
> mm/migrate.c | 2 -
> 6 files changed, 75 insertions(+), 11 deletions(-)
>
> Index: linux-2.6/mm/filemap.c
> ===================================================================
> --- linux-2.6.orig/mm/filemap.c 2011-01-12 15:28:56.000000000 +0100
> +++ linux-2.6/mm/filemap.c 2011-01-13 13:37:41.000000000 +0100
> @@ -387,6 +387,71 @@ int filemap_write_and_wait_range(struct
> EXPORT_SYMBOL(filemap_write_and_wait_range);
>
> /**
> + * replace_page_cache_page - replace a pagecache page with a new one
> + * @old: page to be replaced
> + * @new: page to replace with
> + * @gfp_mask: allocation mode
> + *
> + * This function replaces a page in the pagecache with a new one. On
> + * success it acquires the pagecache reference for the new page and
> + * drops it for the old page. Both the old and new pages must be
> + * locked. This function does not add the new page to the LRU, the
> + * caller must do that.
> + *
> + * The remove + add is atomic. The only way this function can fail is
> + * memory allocation failure.
> + */
> +int replace_page_cache_page(struct page *old, struct page *new, gfp_t gfp_mask)
> +{
> + int error;
> + struct mem_cgroup *memcg = NULL;
> +
> + VM_BUG_ON(!PageLocked(old));
> + VM_BUG_ON(!PageLocked(new));
> + VM_BUG_ON(new->mapping);
> +
> + /*
> + * This is not page migration, but prepare_migration and
> + * end_migration does enough work for charge replacement.
> + *
> + * In the longer term we probably want a specialized function
> + * for moving the charge from old to new in a more efficient
> + * manner.
> + */
> + error = mem_cgroup_prepare_migration(old, new, &memcg, gfp_mask);
> + if (error)
> + return error;
> +
> + error = radix_tree_preload(gfp_mask & ~__GFP_HIGHMEM);
> + if (!error) {
> + struct address_space *mapping = old->mapping;
> + pgoff_t offset = old->index;
> +
> + page_cache_get(new);
> + new->mapping = mapping;
> + new->index = offset;
> +
> + spin_lock_irq(&mapping->tree_lock);
> + __remove_from_page_cache(old);
> + error = radix_tree_insert(&mapping->page_tree, offset, new);
> + BUG_ON(error);
> + mapping->nrpages++;
> + __inc_zone_page_state(new, NR_FILE_PAGES);
> + if (PageSwapBacked(new))
> + __inc_zone_page_state(new, NR_SHMEM);
> + spin_unlock_irq(&mapping->tree_lock);
> + radix_tree_preload_end();
> + page_cache_release(old);
> + mem_cgroup_end_migration(memcg, old, new, true);
> + } else {
> + mem_cgroup_end_migration(memcg, old, new, false);
> + }
> +
> + return error;
> +}
> +EXPORT_SYMBOL_GPL(replace_page_cache_page);
> +
> +/**
> * add_to_page_cache_locked - add a locked page to the pagecache
> * @page: page to add
> * @mapping: the page's address_space
> Index: linux-2.6/include/linux/pagemap.h
> ===================================================================
> --- linux-2.6.orig/include/linux/pagemap.h 2011-01-12 15:28:56.000000000 +0100
> +++ linux-2.6/include/linux/pagemap.h 2011-01-13 13:37:41.000000000 +0100
> @@ -457,6 +457,7 @@ int add_to_page_cache_lru(struct page *p
> pgoff_t index, gfp_t gfp_mask);
> extern void remove_from_page_cache(struct page *page);
> extern void __remove_from_page_cache(struct page *page);
> +int replace_page_cache_page(struct page *old, struct page *new, gfp_t gfp_mask);
>
> /*
> * Like add_to_page_cache_locked, but used to add newly allocated pages:
> Index: linux-2.6/fs/fuse/dev.c
> ===================================================================
> --- linux-2.6.orig/fs/fuse/dev.c 2011-01-12 15:28:56.000000000 +0100
> +++ linux-2.6/fs/fuse/dev.c 2011-01-13 13:37:41.000000000 +0100
> @@ -737,14 +737,12 @@ static int fuse_try_move_page(struct fus
> if (WARN_ON(PageMlocked(oldpage)))
> goto out_fallback_unlock;
>
> - remove_from_page_cache(oldpage);
> - page_cache_release(oldpage);
> -
> - err = add_to_page_cache_locked(newpage, mapping, index, GFP_KERNEL);
> + err = replace_page_cache_page(oldpage, newpage, GFP_KERNEL);
> if (err) {
> - printk(KERN_WARNING "fuse_try_move_page: failed to add page");
> - goto out_fallback_unlock;
> + unlock_page(newpage);
> + return err;
> }
> +
> page_cache_get(newpage);
>
> if (!(buf->flags & PIPE_BUF_FLAG_LRU))
> Index: linux-2.6/include/linux/memcontrol.h
> ===================================================================
> --- linux-2.6.orig/include/linux/memcontrol.h 2011-01-13 13:37:40.000000000 +0100
> +++ linux-2.6/include/linux/memcontrol.h 2011-01-13 13:37:41.000000000 +0100
> @@ -91,7 +91,7 @@ extern struct cgroup_subsys_state *mem_c
>
> extern int
> mem_cgroup_prepare_migration(struct page *page,
> - struct page *newpage, struct mem_cgroup **ptr);
> + struct page *newpage, struct mem_cgroup **ptr, gfp_t gfp_mask);
> extern void mem_cgroup_end_migration(struct mem_cgroup *mem,
> struct page *oldpage, struct page *newpage, bool migration_ok);
>
> @@ -225,7 +225,7 @@ static inline struct cgroup_subsys_state
>
> static inline int
> mem_cgroup_prepare_migration(struct page *page, struct page *newpage,
> - struct mem_cgroup **ptr)
> + struct mem_cgroup **ptr, gfp_t gfp_mask)
> {
> return 0;
> }
> Index: linux-2.6/mm/memcontrol.c
> ===================================================================
> --- linux-2.6.orig/mm/memcontrol.c 2011-01-13 13:37:40.000000000 +0100
> +++ linux-2.6/mm/memcontrol.c 2011-01-13 13:37:41.000000000 +0100
> @@ -2767,7 +2767,7 @@ static inline int mem_cgroup_move_swap_a
> * page belongs to.
> */
> int mem_cgroup_prepare_migration(struct page *page,
> - struct page *newpage, struct mem_cgroup **ptr)
> + struct page *newpage, struct mem_cgroup **ptr, gfp_t gfp_mask)
> {
> struct page_cgroup *pc;
> struct mem_cgroup *mem = NULL;
> @@ -2823,7 +2823,7 @@ int mem_cgroup_prepare_migration(struct
> return 0;
>
> *ptr = mem;
> - ret = __mem_cgroup_try_charge(NULL, GFP_KERNEL, ptr, false);
> + ret = __mem_cgroup_try_charge(NULL, gfp_mask, ptr, false);
> css_put(&mem->css);/* drop extra refcnt */
> if (ret || *ptr == NULL) {
> if (PageAnon(page)) {
> Index: linux-2.6/mm/migrate.c
> ===================================================================
> --- linux-2.6.orig/mm/migrate.c 2011-01-13 13:37:40.000000000 +0100
> +++ linux-2.6/mm/migrate.c 2011-01-13 13:38:55.000000000 +0100
> @@ -657,7 +657,7 @@ static int unmap_and_move(new_page_t get
> }
>
> /* charge against new page */
> - charge = mem_cgroup_prepare_migration(page, newpage, &mem);
> + charge = mem_cgroup_prepare_migration(page, newpage, &mem, GFP_KERNEL);
> if (charge == -ENOMEM) {
> rc = -ENOMEM;
> goto unlock;
This patch looks good and I want to ack, but this patch can conflict with patches
in https://lkml.org/lkml/2011/1/11/10 and https://lkml.org/lkml/2011/1/11/65.
So I think you'd better coordinate with Minchan.
Thanks,
Daisuke Nishimura.
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [PATCH v3] mm: add replace_page_cache_page() function
2011-01-14 1:00 ` Daisuke Nishimura
@ 2011-01-17 13:59 ` Minchan Kim
0 siblings, 0 replies; 9+ messages in thread
From: Minchan Kim @ 2011-01-17 13:59 UTC (permalink / raw)
To: Daisuke Nishimura
Cc: Miklos Szeredi, KAMEZAWA Hiroyuki, akpm, linux-fsdevel, linux-mm,
linux-kernel
On Fri, Jan 14, 2011 at 10:00:37AM +0900, Daisuke Nishimura wrote:
> On Thu, 13 Jan 2011 14:35:09 +0100
> Miklos Szeredi <miklos@szeredi.hu> wrote:
>
> > > Seems nicer. But you didn't fixed the caller of prepare_migraton()
> > > in mm/migrate.c (I think this is the only caller..)
> > >
> > > please modify it, too.
> >
> > Right you are. Thanks for the review.
> >
> > Updated patch attached.
> >
> > Thanks,
> > Miklos
> > ---
> >
> >
> > From: Miklos Szeredi <mszeredi@suse.cz>
> > Subject: mm: add replace_page_cache_page() function
> >
> > This function basically does:
> >
> > remove_from_page_cache(old);
> > page_cache_release(old);
> > add_to_page_cache_locked(new);
> >
> > Except it does this atomically, so there's no possibility for the
> > "add" to fail because of a race.
> >
> > This is used by fuse to move pages into the page cache.
> >
> > Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
> > ---
> > fs/fuse/dev.c | 10 ++----
> > include/linux/memcontrol.h | 4 +-
> > include/linux/pagemap.h | 1
> > mm/filemap.c | 65 +++++++++++++++++++++++++++++++++++++++++++++
> > mm/memcontrol.c | 4 +-
> > mm/migrate.c | 2 -
> > 6 files changed, 75 insertions(+), 11 deletions(-)
> >
> > Index: linux-2.6/mm/filemap.c
> > ===================================================================
> > --- linux-2.6.orig/mm/filemap.c 2011-01-12 15:28:56.000000000 +0100
> > +++ linux-2.6/mm/filemap.c 2011-01-13 13:37:41.000000000 +0100
> > @@ -387,6 +387,71 @@ int filemap_write_and_wait_range(struct
> > EXPORT_SYMBOL(filemap_write_and_wait_range);
> >
> > /**
> > + * replace_page_cache_page - replace a pagecache page with a new one
> > + * @old: page to be replaced
> > + * @new: page to replace with
> > + * @gfp_mask: allocation mode
> > + *
> > + * This function replaces a page in the pagecache with a new one. On
> > + * success it acquires the pagecache reference for the new page and
> > + * drops it for the old page. Both the old and new pages must be
> > + * locked. This function does not add the new page to the LRU, the
> > + * caller must do that.
> > + *
> > + * The remove + add is atomic. The only way this function can fail is
> > + * memory allocation failure.
> > + */
> > +int replace_page_cache_page(struct page *old, struct page *new, gfp_t gfp_mask)
> > +{
> > + int error;
> > + struct mem_cgroup *memcg = NULL;
> > +
> > + VM_BUG_ON(!PageLocked(old));
> > + VM_BUG_ON(!PageLocked(new));
> > + VM_BUG_ON(new->mapping);
> > +
> > + /*
> > + * This is not page migration, but prepare_migration and
> > + * end_migration does enough work for charge replacement.
> > + *
> > + * In the longer term we probably want a specialized function
> > + * for moving the charge from old to new in a more efficient
> > + * manner.
> > + */
> > + error = mem_cgroup_prepare_migration(old, new, &memcg, gfp_mask);
> > + if (error)
> > + return error;
> > +
> > + error = radix_tree_preload(gfp_mask & ~__GFP_HIGHMEM);
> > + if (!error) {
> > + struct address_space *mapping = old->mapping;
> > + pgoff_t offset = old->index;
> > +
> > + page_cache_get(new);
> > + new->mapping = mapping;
> > + new->index = offset;
> > +
> > + spin_lock_irq(&mapping->tree_lock);
> > + __remove_from_page_cache(old);
> > + error = radix_tree_insert(&mapping->page_tree, offset, new);
> > + BUG_ON(error);
> > + mapping->nrpages++;
> > + __inc_zone_page_state(new, NR_FILE_PAGES);
> > + if (PageSwapBacked(new))
> > + __inc_zone_page_state(new, NR_SHMEM);
> > + spin_unlock_irq(&mapping->tree_lock);
> > + radix_tree_preload_end();
> > + page_cache_release(old);
> > + mem_cgroup_end_migration(memcg, old, new, true);
> > + } else {
> > + mem_cgroup_end_migration(memcg, old, new, false);
> > + }
> > +
> > + return error;
> > +}
> > +EXPORT_SYMBOL_GPL(replace_page_cache_page);
> > +
> > +/**
> > * add_to_page_cache_locked - add a locked page to the pagecache
> > * @page: page to add
> > * @mapping: the page's address_space
> > Index: linux-2.6/include/linux/pagemap.h
> > ===================================================================
> > --- linux-2.6.orig/include/linux/pagemap.h 2011-01-12 15:28:56.000000000 +0100
> > +++ linux-2.6/include/linux/pagemap.h 2011-01-13 13:37:41.000000000 +0100
> > @@ -457,6 +457,7 @@ int add_to_page_cache_lru(struct page *p
> > pgoff_t index, gfp_t gfp_mask);
> > extern void remove_from_page_cache(struct page *page);
> > extern void __remove_from_page_cache(struct page *page);
> > +int replace_page_cache_page(struct page *old, struct page *new, gfp_t gfp_mask);
> >
> > /*
> > * Like add_to_page_cache_locked, but used to add newly allocated pages:
> > Index: linux-2.6/fs/fuse/dev.c
> > ===================================================================
> > --- linux-2.6.orig/fs/fuse/dev.c 2011-01-12 15:28:56.000000000 +0100
> > +++ linux-2.6/fs/fuse/dev.c 2011-01-13 13:37:41.000000000 +0100
> > @@ -737,14 +737,12 @@ static int fuse_try_move_page(struct fus
> > if (WARN_ON(PageMlocked(oldpage)))
> > goto out_fallback_unlock;
> >
> > - remove_from_page_cache(oldpage);
> > - page_cache_release(oldpage);
> > -
> > - err = add_to_page_cache_locked(newpage, mapping, index, GFP_KERNEL);
> > + err = replace_page_cache_page(oldpage, newpage, GFP_KERNEL);
> > if (err) {
> > - printk(KERN_WARNING "fuse_try_move_page: failed to add page");
> > - goto out_fallback_unlock;
> > + unlock_page(newpage);
> > + return err;
> > }
> > +
> > page_cache_get(newpage);
> >
> > if (!(buf->flags & PIPE_BUF_FLAG_LRU))
> > Index: linux-2.6/include/linux/memcontrol.h
> > ===================================================================
> > --- linux-2.6.orig/include/linux/memcontrol.h 2011-01-13 13:37:40.000000000 +0100
> > +++ linux-2.6/include/linux/memcontrol.h 2011-01-13 13:37:41.000000000 +0100
> > @@ -91,7 +91,7 @@ extern struct cgroup_subsys_state *mem_c
> >
> > extern int
> > mem_cgroup_prepare_migration(struct page *page,
> > - struct page *newpage, struct mem_cgroup **ptr);
> > + struct page *newpage, struct mem_cgroup **ptr, gfp_t gfp_mask);
> > extern void mem_cgroup_end_migration(struct mem_cgroup *mem,
> > struct page *oldpage, struct page *newpage, bool migration_ok);
> >
> > @@ -225,7 +225,7 @@ static inline struct cgroup_subsys_state
> >
> > static inline int
> > mem_cgroup_prepare_migration(struct page *page, struct page *newpage,
> > - struct mem_cgroup **ptr)
> > + struct mem_cgroup **ptr, gfp_t gfp_mask)
> > {
> > return 0;
> > }
> > Index: linux-2.6/mm/memcontrol.c
> > ===================================================================
> > --- linux-2.6.orig/mm/memcontrol.c 2011-01-13 13:37:40.000000000 +0100
> > +++ linux-2.6/mm/memcontrol.c 2011-01-13 13:37:41.000000000 +0100
> > @@ -2767,7 +2767,7 @@ static inline int mem_cgroup_move_swap_a
> > * page belongs to.
> > */
> > int mem_cgroup_prepare_migration(struct page *page,
> > - struct page *newpage, struct mem_cgroup **ptr)
> > + struct page *newpage, struct mem_cgroup **ptr, gfp_t gfp_mask)
> > {
> > struct page_cgroup *pc;
> > struct mem_cgroup *mem = NULL;
> > @@ -2823,7 +2823,7 @@ int mem_cgroup_prepare_migration(struct
> > return 0;
> >
> > *ptr = mem;
> > - ret = __mem_cgroup_try_charge(NULL, GFP_KERNEL, ptr, false);
> > + ret = __mem_cgroup_try_charge(NULL, gfp_mask, ptr, false);
> > css_put(&mem->css);/* drop extra refcnt */
> > if (ret || *ptr == NULL) {
> > if (PageAnon(page)) {
> > Index: linux-2.6/mm/migrate.c
> > ===================================================================
> > --- linux-2.6.orig/mm/migrate.c 2011-01-13 13:37:40.000000000 +0100
> > +++ linux-2.6/mm/migrate.c 2011-01-13 13:38:55.000000000 +0100
> > @@ -657,7 +657,7 @@ static int unmap_and_move(new_page_t get
> > }
> >
> > /* charge against new page */
> > - charge = mem_cgroup_prepare_migration(page, newpage, &mem);
> > + charge = mem_cgroup_prepare_migration(page, newpage, &mem, GFP_KERNEL);
> > if (charge == -ENOMEM) {
> > rc = -ENOMEM;
> > goto unlock;
> This patch looks good and I want to ack, but this patch can conflict with patches
> in https://lkml.org/lkml/2011/1/11/10 and https://lkml.org/lkml/2011/1/11/65.
> So I think you'd better coordinate with Minchan.
And it can conflict with https://patchwork.kernel.org/patch/470311/
I am not sure Andrew accept it easily although many mm guys already acked and reviewed it.
So I hope patch ordering is following as after mmotm release
1. https://lkml.org/lkml/2011/1/11/10 and https://lkml.org/lkml/2011/1/11/65
2. your patch
3. https://patchwork.kernel.org/patch/470311/
If you have a problem, please let me know it.
>
> Thanks,
> Daisuke Nishimura.
--
Kind regards,
Minchan Kim
^ permalink raw reply [flat|nested] 9+ messages in thread
end of thread, other threads:[~2011-01-17 13:59 UTC | newest]
Thread overview: 9+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2011-01-11 14:07 [PATCH v3] mm: add replace_page_cache_page() function Miklos Szeredi
2011-01-11 14:25 ` Minchan Kim
2011-01-12 0:44 ` KAMEZAWA Hiroyuki
2011-01-12 14:30 ` Miklos Szeredi
2011-01-13 2:27 ` KAMEZAWA Hiroyuki
2011-01-13 13:35 ` Miklos Szeredi
2011-01-14 1:00 ` KAMEZAWA Hiroyuki
2011-01-14 1:00 ` Daisuke Nishimura
2011-01-17 13:59 ` Minchan Kim
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).