f2fs: avoid freed stat information
Subsystems (2)
Jaegeuk Kim
maintainerPATCH 1/12 f2fs: avoid freed stat information
The write_checkpoint can update stat information, so we should destroy the stat structure after it.
--- fs/f2fs/super.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)
fs/f2fs/super.c +3 -1
@@ -498,7 +498,6 @@ static void f2fs_put_super(struct super_block *sb) }kobject_del (&sbi->s_kobj); -f2fs_destroy_stats (sbi);stop_gc_thread (sbi);/*
@@ -514,6 +513,9 @@ static void f2fs_put_super(struct super_block *sb)write_checkpoint (sbi, &cpc); } +/* write_checkpoint can update stat informaion */ +f2fs_destroy_stats (sbi); +/* * normally superblock is clean, so we need to release this. * In addition, EIO will skip do checkpoint, we need this as well.
Signature · 9 lines
2.1.1 ------------------------------------------------------------------------------ Don't Limit Your Business. Reach for the Cloud. GigeNET's Cloud Solutions provide you with the tools and support that you need to offload your IT needs and focus on growing your business. Configured For All Businesses. Start Your Cloud Today. https://www.gigenetcloud.com/
Jaegeuk Kim
maintainerPATCH 2/12 f2fs: avoid to use failed inode immediately
Before iput is called, the inode number used by a bad inode can be reassigned to other new inode, resulting in any abnormal behaviors on the new inode. This should not happen for the new inode.
--- fs/f2fs/f2fs.h | 1 + fs/f2fs/inode.c | 19 ++++++++++++------- fs/f2fs/namei.c | 4 ++-- 3 files changed, 15 insertions(+), 9 deletions(-)
fs/f2fs/f2fs.h +1 -0
@@ -1343,6 +1343,7 @@ enum {FI_INC_LINK ,/* need to increment i_nlink */ FI_ACL_MODE ,/* indicate acl mode */ FI_NO_ALLOC ,/* should not allocate any blocks */ +FI_FREE_NID ,/* free allocated nide */ FI_UPDATE_DIR ,/* should update inode block for consistency */ FI_DELAY_IPUT ,/* used for the recovery */ FI_NO_EXTENT, /* not to use the extent cache */
fs/f2fs/inode.c +12 -7
@@ -314,7 +314,8 @@ int f2fs_write_inode(struct inode *inode, struct writeback_control *wbc)void f2fs_evict_inode (struct inode *inode) {struct f2fs_sb_info *sbi =F2FS_I_SB (inode); -nid_t xnid =F2FS_I (inode)->i_xattr_nid; +struct f2fs_inode_info *fi =F2FS_I (inode); +nid_t xnid = fi->i_xattr_nid;/* some remained atomic pages should discarded */ if (f2fs_is_atomic_file (inode))
@@ -334,7 +335,7 @@ void f2fs_evict_inode(struct inode *inode) goto no_delete;sb_start_intwrite (inode->i_sb); -set_inode_flag (F2FS_I (inode),FI_NO_ALLOC ); +set_inode_flag (fi,FI_NO_ALLOC );i_size_write (inode,0 );if (F2FS_HAS_BLOCKS (inode))
@@ -357,14 +358,18 @@ no_delete:invalidate_mapping_pages (NODE_MAPPING (sbi), inode->i_ino, inode->i_ino);if (xnid)invalidate_mapping_pages (NODE_MAPPING (sbi), xnid, xnid); -if (is_inode_flag_set (F2FS_I (inode),FI_APPEND_WRITE )) +if (is_inode_flag_set (fi,FI_APPEND_WRITE ))add_dirty_inode (sbi, inode->i_ino,APPEND_INO ); -if (is_inode_flag_set (F2FS_I (inode),FI_UPDATE_WRITE )) +if (is_inode_flag_set (fi,FI_UPDATE_WRITE ))add_dirty_inode (sbi, inode->i_ino,UPDATE_INO ); +if (is_inode_flag_set (fi,FI_FREE_NID )) { +alloc_nid_failed (sbi, inode->i_ino); +clear_inode_flag (fi,FI_FREE_NID ); + } out_clear:#ifdef CONFIG_F2FS_FS_ENCRYPTION -if (F2FS_I (inode)->i_crypt_info) -f2fs_free_encryption_info (inode,F2FS_I (inode)->i_crypt_info); +if (fi->i_crypt_info) +f2fs_free_encryption_info (inode, fi->i_crypt_info);#endif clear_inode (inode); }
@@ -384,9 +389,9 @@ void handle_failed_inode(struct inode *inode)remove_inode_page (inode); +set_inode_flag (F2FS_I (inode),FI_FREE_NID );clear_inode_flag (F2FS_I (inode),FI_INLINE_DATA );clear_inode_flag (F2FS_I (inode),FI_INLINE_DENTRY ); -alloc_nid_failed (sbi, inode->i_ino);f2fs_unlock_op (sbi);/* iput will drop the inode object */
fs/f2fs/namei.c +2 -2
@@ -78,9 +78,9 @@ out: fail:trace_f2fs_new_inode (inode, err);make_bad_inode (inode); -iput (inode);if (nid_free) -alloc_nid_failed (sbi, ino); +set_inode_flag (F2FS_I (inode),FI_FREE_NID ); +iput (inode);return ERR_PTR (err); }
Signature · 9 lines
2.1.1 ------------------------------------------------------------------------------ Don't Limit Your Business. Reach for the Cloud. GigeNET's Cloud Solutions provide you with the tools and support that you need to offload your IT needs and focus on growing your business. Configured For All Businesses. Start Your Cloud Today. https://www.gigenetcloud.com/
RE: [f2fs-dev] [PATCH 02/12] f2fs: avoid to use failed inode immediately
Quoted 11 lines
-----Original Message-----
From: Jaegeuk Kim [mailto:[email protected]]
Sent: Tuesday, June 30, 2015 2:39 AM
To: [email protected]; [email protected]; [email protected]
Cc: Jaegeuk Kim
Subject: [f2fs-dev] [PATCH 02/12] f2fs: avoid to use failed inode immediatelyBefore iput is called, the inode number used by a bad inode can be reassigned to other new inode, resulting in any abnormal behaviors on the new inode. This should not happen for the new inode.
Nice catch!
Jaegeuk Kim
maintainerPATCH 3/12 f2fs: convert inline_data for various fallocate
In reply to Jaegeuk KimFor newly added fallocate types, it should convert inline_data before handling block swapping.
--- fs/f2fs/file.c | 14 ++++++++++++++ 1 file changed, 14 insertions(+)
fs/f2fs/file.c +14 -0
@@ -885,6 +885,14 @@ static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)if (offset & (F2FS_BLKSIZE -1 ) || len & (F2FS_BLKSIZE -1 ))return -EINVAL ; +f2fs_balance_fs (F2FS_I_SB (inode)); + +if (f2fs_has_inline_data (inode)) { + ret =f2fs_convert_inline_inode (inode); +if (ret) +return ret; + } + pg_start = offset >>PAGE_CACHE_SHIFT ; pg_end = (offset + len) >>PAGE_CACHE_SHIFT ;
@@ -1033,6 +1041,12 @@ static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)f2fs_balance_fs (sbi); +if (f2fs_has_inline_data (inode)) { + ret =f2fs_convert_inline_inode (inode); +if (ret) +return ret; + } + ret =truncate_blocks (inode,i_size_read (inode), true);if (ret)return ret;
Signature · 9 lines
2.1.1 ------------------------------------------------------------------------------ Don't Limit Your Business. Reach for the Cloud. GigeNET's Cloud Solutions provide you with the tools and support that you need to offload your IT needs and focus on growing your business. Configured For All Businesses. Start Your Cloud Today. https://www.gigenetcloud.com/
RE: [f2fs-dev] [PATCH 03/12] f2fs: convert inline_data for various fallocate
Quoted 8 lines
-----Original Message-----
From: Jaegeuk Kim [mailto:[email protected]]
Sent: Tuesday, June 30, 2015 2:39 AM
To: [email protected]; [email protected]; [email protected]
Cc: Jaegeuk Kim
Subject: [f2fs-dev] [PATCH 03/12] f2fs: convertinline_datafor various fallocate
Quoted 4 lines
For newly added fallocate types, it should convert
inline_databefore handling block swapping.Signed-off-by Jaegeuk Kim <[email protected]>
Jaegeuk Kim
maintainerPATCH 4/12 f2fs: remove wrong f2fs_bug_on when merging extents
In reply to Jaegeuk KimIn f2fs_update_extent_tree, if there is existing extent, f2fs tries to split it with two parts.
In each trial, __insert_extent_tree checks __is_front/back_mergeable, and then if it hits to go, there is f2fs_bug_on(!den), which triggers a kernel panic.
Actually, we don't need to check this. Instead, we can do __try_back_merge only when there exists a den pointer.
--- fs/f2fs/data.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-)
fs/f2fs/data.c +4 -4
@@ -519,19 +519,19 @@ static struct extent_node *__insert_extent_tree(struct f2fs_sb_info *sbi,if (ei->fofs < en->ei.fofs) {if (__is_front_mergeable (ei, &en->ei)) { -f2fs_bug_on (sbi, !den); en->ei.fofs = ei->fofs; en->ei.blk = ei->blk; en->ei.len += ei->len; - *den =__try_back_merge (sbi, et, en); +if (den) + *den =__try_back_merge (sbi, et, en);return en; } p = &(*p)->rb_left; }else if (ei->fofs >= en->ei.fofs + en->ei.len) {if (__is_back_mergeable (ei, &en->ei)) { -f2fs_bug_on (sbi, !den); en->ei.len += ei->len; - *den =__try_front_merge (sbi, et, en); +if (den) + *den =__try_front_merge (sbi, et, en);return en; } p = &(*p)->rb_right;
Signature · 9 lines
2.1.1 ------------------------------------------------------------------------------ Don't Limit Your Business. Reach for the Cloud. GigeNET's Cloud Solutions provide you with the tools and support that you need to offload your IT needs and focus on growing your business. Configured For All Businesses. Start Your Cloud Today. https://www.gigenetcloud.com/
RE: [f2fs-dev] [PATCH 04/12] f2fs: remove wrong f2fs_bug_on when merging extents
Hi Jaegeuk,
Quoted 26 lines
-----Original Message-----
From: Jaegeuk Kim [mailto:[email protected]]
Sent: Tuesday, June 30, 2015 2:39 AM
To: [email protected]; [email protected]; [email protected]
Cc: Jaegeuk Kim
Subject: [f2fs-dev] [PATCH 04/12] f2fs: remove wrongf2fs_bug_onwhen merging extentsIn
f2fs_update_extent_tree, if there is existing extent, f2fs tries to split it with two parts.
In each trial,__insert_extent_treechecks __is_front/back_mergeable, and then if it hits to go, there isf2fs_bug_on(!den), which triggers a kernel panic.Actually, we don't need to check this. Instead, we can do
__try_back_mergeonly when there exists a den pointer.Signed-off-by Jaegeuk Kim <[email protected]>--- fs/f2fs/data.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-)fs/f2fs/data.c +0
-0
fs/f2fs/data.c static struct extent_node *__insert_extent_tree(struct f2fs_sb_info *sbi, · line 522
Quoted 4 lines
fs/f2fs/data.c +0
-1@@ -519,19 +519,19 @@ static struct extent_node *__insert_extent_tree(struct f2fs_sb_info *sbi,if (ei->fofs < en->ei.fofs) {if (__is_front_mergeable (ei, &en->ei)) { -f2fs_bug_on (sbi, !den);
I add a BUG_ON here because we assume that in extent cache there is no such two extents whose mapping address is continuous but without being merged, since whenever we add a new extent(not splitted one) into cache, we tries to merge it frontward/backward, then all extent with continuous mapping should be merged. So when we split one extent to two parts, each part should not be able to merge with others. Otherwise it should be a bug.
Is there some special case to trigger this BUG_ON?
Thanks,
Quoted 23 lines
fs/f2fs/data.c +4
-4en->ei.fofs = ei->fofs; en->ei.blk = ei->blk; en->ei.len += ei->len; - *den =__try_back_merge (sbi, et, en); +if (den) + *den =__try_back_merge (sbi, et, en);return en; } p = &(*p)->rb_left; }else if (ei->fofs >= en->ei.fofs + en->ei.len) {if (__is_back_mergeable (ei, &en->ei)) { -f2fs_bug_on (sbi, !den); en->ei.len += ei->len; - *den =__try_front_merge (sbi, et, en); +if (den) + *den =__try_front_merge (sbi, et, en);return en; } p = &(*p)->rb_right; --2.1.1
Quoted 10 lines
------------------------------------------------------------------------------
Don't Limit Your Business. Reach for the Cloud.
GigeNET's Cloud Solutions provide you with the tools and support that you need to offload your IT needs and focus on growing your business. Configured For All Businesses. Start Your Cloud Today.
https://www.gigenetcloud.com/
_______________________________________________
Linux-f2fs-devel mailing list [email protected]
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel
------------------------------------------------------------------------------
Don't Limit Your Business. Reach for the Cloud.
GigeNET's Cloud Solutions provide you with the tools and support that you need to offload your IT needs and focus on growing your business. Configured For All Businesses. Start Your Cloud Today.
https://www.gigenetcloud.com/
Jaegeuk Kim
maintainerRe: [f2fs-dev] [PATCH 04/12] f2fs: remove wrong f2fs_bug_on when merging extents
Quoted 29 lines · Chao Yu
Hi Jaegeuk,
Quoted 27 lines
-----Original Message-----
From: Jaegeuk Kim [mailto:[email protected]]
Sent: Tuesday, June 30, 2015 2:39 AM
To: [email protected]; [email protected]; [email protected]
Cc: Jaegeuk Kim
Subject: [f2fs-dev] [PATCH 04/12] f2fs: remove wrongf2fs_bug_onwhen merging extentsIn
f2fs_update_extent_tree, if there is existing extent, f2fs tries to split it with two parts.
In each trial,__insert_extent_treechecks __is_front/back_mergeable, and then if it hits to go, there isf2fs_bug_on(!den), which triggers a kernel panic.Actually, we don't need to check this. Instead, we can do
__try_back_mergeonly when there exists a den pointer.Signed-off-by Jaegeuk Kim <[email protected]>--- fs/f2fs/data.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-)fs/f2fs/data.c +0
-0@@ -519,19 +519,19 @@ static struct extent_node *__insert_extent_tree(struct f2fs_sb_info *sbi,
Quoted 10 lines
fs/f2fs/data.c static struct extent_node *__insert_extent_tree(struct f2fs_sb_info *sbi, · line 522
Quoted 3 lines
fs/f2fs/data.c +0
-1if (ei->fofs < en->ei.fofs) {if (__is_front_mergeable (ei, &en->ei)) { -f2fs_bug_on (sbi, !den);I add a
BUG_ONhere because we assume that in extent cache there is no such two extents whose mapping address is continuous but without being merged, since whenever we add a new extent(not splitted one) into cache, we tries to merge it frontward/backward, then all extent with continuous mapping should be merged. So when we split one extent to two parts, each part should not be able to merge with others. Otherwise it should be a bug.
Oh, I see.
Let me take a look at this one more time.
I'm confusing whether this patch was written before fixing block address calculation and then change the order of the patches.
Thanks,
Jaegeuk Kim
maintainerPATCH 5/12 f2fs: fix wrong block address calculation for a split extent
In reply to Jaegeuk KimThis patch fixes wrong calculation on block address field when an extent is split.
--- fs/f2fs/data.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
fs/f2fs/data.c +1 -1
@@ -674,7 +674,7 @@ static void f2fs_update_extent_tree(struct inode *inode, pgoff_t fofs, endofs = dei.fofs + dei.len -1 ;if (endofs - fofs >=F2FS_MIN_EXTENT_LEN ) {set_extent_info (&ei, fofs +1 , - fofs - dei.fofs + dei.blk, endofs - fofs); + fofs - dei.fofs + dei.blk +1 , endofs - fofs); en2 =__insert_extent_tree (sbi, et, &ei,NULL ); } }
Signature · 9 lines
2.1.1 ------------------------------------------------------------------------------ Don't Limit Your Business. Reach for the Cloud. GigeNET's Cloud Solutions provide you with the tools and support that you need to offload your IT needs and focus on growing your business. Configured For All Businesses. Start Your Cloud Today. https://www.gigenetcloud.com/
RE: [f2fs-dev] [PATCH 05/12] f2fs: fix wrong block address calculation for a split extent
Hi Jaegeuk,
Quoted 10 lines
-----Original Message-----
From: Jaegeuk Kim [mailto:[email protected]]
Sent: Tuesday, June 30, 2015 2:39 AM
To: [email protected]; [email protected]; [email protected]
Cc: Jaegeuk Kim
Subject: [f2fs-dev] [PATCH 05/12] f2fs: fix wrong block address calculation for a split extentThis patch fixes wrong calculation on block address field when an extent is split.
Oh, that's a terrible mistake, sorry about it.
Thank you for fixing this.
Quoted 2 lines
Signed-off-by Jaegeuk Kim <[email protected]>
------------------------------------------------------------------------------
Don't Limit Your Business. Reach for the Cloud.
GigeNET's Cloud Solutions provide you with the tools and support that you need to offload your IT needs and focus on growing your business. Configured For All Businesses. Start Your Cloud Today.
https://www.gigenetcloud.com/
Jaegeuk Kim
maintainerPATCH 6/12 f2fs: update on-disk extents even under extent_cache
In reply to Jaegeuk KimPreviously, f2fs_update_extent_cache() updates in-memory extent_cache all the time, and then finally preserves its up-to-date extent into on-disk one during f2fs_evict_inode.
But, in the following scenario:
1. mount
2. open & write an extent X
3. f2fs_evict_inode; on-disk extent is X
4. open & update the extent X with Y
5. sync; trigger checkpoint
6. power-cut
after power-on, f2fs should serve extent Y, but we have an on-disk extent X.
This causes a failure on xfstests/311.
--- fs/f2fs/data.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
fs/f2fs/data.c +2 -2
@@ -899,9 +899,9 @@ void f2fs_update_extent_cache(struct dnode_of_data *dn) fofs =start_bidx_of_node (ofs_of_node (dn->node_page), fi) + dn->ofs_in_node; +/* we should call update_extent_info() to update on-disk extent */ if (test_opt (F2FS_I_SB (dn->inode),EXTENT_CACHE )) -return f2fs_update_extent_tree (dn->inode, fofs, - dn->data_blkaddr); +f2fs_update_extent_tree (dn->inode, fofs, dn->data_blkaddr);if (update_extent_info (dn->inode, fofs, dn->data_blkaddr))sync_inode_page (dn);
Signature · 9 lines
2.1.1 ------------------------------------------------------------------------------ Don't Limit Your Business. Reach for the Cloud. GigeNET's Cloud Solutions provide you with the tools and support that you need to offload your IT needs and focus on growing your business. Configured For All Businesses. Start Your Cloud Today. https://www.gigenetcloud.com/
RE: [f2fs-dev] [PATCH 06/12] f2fs: update on-disk extents even under extent_cache
Quoted 21 lines
-----Original Message-----
From: Jaegeuk Kim [mailto:[email protected]]
Sent: Tuesday, June 30, 2015 2:40 AM
To: [email protected]; [email protected]; [email protected]
Cc: Jaegeuk Kim
Subject: [f2fs-dev] [PATCH 06/12] f2fs: update on-disk extents even underextent_cachePreviously,
f2fs_update_extent_cache()updates in-memoryextent_cacheall the time, and then finally preserves its up-to-date extent into on-disk one duringf2fs_evict_inode.But, in the following scenario:
1. mount
2. open & write an extent X
3.f2fs_evict_inode; on-disk extent is X
4. open & update the extent X with Y
5. sync; trigger checkpoint
6. power-cut
Quoted 5 lines
after power-on, f2fs should serve extent Y, but we have an on-disk extent X.
This causes a failure on xfstests/311.
Signed-off-by Jaegeuk Kim <[email protected]>
Jaegeuk Kim
maintainerPATCH 7/12 f2fs: set cached_en after checking finally
In reply to Jaegeuk KimThis patch relocates cached_en not only to be covered by spin_lock, but also to set once after checking out completely.
--- fs/f2fs/data.c | 9 ++++----- 1 file changed, 4 insertions(+), 5 deletions(-)
fs/f2fs/data.c +4 -5
@@ -453,14 +453,12 @@ static struct extent_node *__lookup_extent_tree(struct extent_tree *et,while (node) { en = rb_entry(node, struct extent_node, rb_node); -if (fofs < en->ei.fofs) { +if (fofs < en->ei.fofs) node = node->rb_left; - }else if (fofs >= en->ei.fofs + en->ei.len) { +else if (fofs >= en->ei.fofs + en->ei.len) node = node->rb_right; - }else { - et->cached_en = en; +else return en; - } }return NULL ; }
@@ -625,6 +623,7 @@ static bool f2fs_lookup_extent_tree(struct inode *inode, pgoff_t pgofs,spin_lock (&sbi->extent_lock);if (!list_empty (&en->list))list_move_tail (&en->list, &sbi->extent_list); + et->cached_en = en;spin_unlock (&sbi->extent_lock);stat_inc_read_hit (sbi->sb); }
Signature · 1 line
2.1.1
RE: [f2fs-dev] [PATCH 07/12] f2fs: set cached_en after checking finally
Quoted 8 lines
-----Original Message-----
From: Jaegeuk Kim [mailto:[email protected]]
Sent: Tuesday, June 30, 2015 2:40 AM
To: [email protected]; [email protected]; [email protected]
Cc: Jaegeuk Kim
Subject: [f2fs-dev] [PATCH 07/12] f2fs: setcached_enafter checking finally
Quoted 4 lines
This patch relocates
cached_ennot only to be covered byspin_lock, but also to set once after checking out completely.Signed-off-by Jaegeuk Kim <[email protected]>
------------------------------------------------------------------------------
Don't Limit Your Business. Reach for the Cloud.
GigeNET's Cloud Solutions provide you with the tools and support that you need to offload your IT needs and focus on growing your business. Configured For All Businesses. Start Your Cloud Today.
https://www.gigenetcloud.com/
Jaegeuk Kim
maintainerPATCH 8/12 f2fs: introduce a shrinker for mounted fs
In reply to Jaegeuk KimThis patch introduces a shrinker targeting to reduce memory footprint consumed by a number of in-memory f2fs data structures.
In addition, it newly adds:
- sbi->umount_mutex to avoid data races on shrinker and put_super
- sbi->shruinker_run_no to not revisit objects
Noteh that the basic implementation was copied from fs/btrfs/shrinker.c
--- fs/f2fs/Makefile | 1 + fs/f2fs/f2fs.h | 13 +++++++ fs/f2fs/shrinker.c | 104 +++++++++++++++++++++++++++++++++++++++++++++++++++++ fs/f2fs/super.c | 24 +++++++++++++ 4 files changed, 142 insertions(+) create mode 100644 fs/f2fs/shrinker.c
fs/f2fs/Makefile +1 -0
@@ -2,6 +2,7 @@ obj-$(CONFIG_F2FS_FS) += f2fs.o f2fs-y := dir.o file.o inode.o namei.o hash.o super.o inline.o f2fs-y += checkpoint.o gc.o data.o node.o segment.o recovery.o +f2fs-y += shrinker.o f2fs-$(CONFIG_F2FS_STAT_FS) += debug.o f2fs-$(CONFIG_F2FS_FS_XATTR) += xattr.o f2fs-$(CONFIG_F2FS_FS_POSIX_ACL) += acl.o
fs/f2fs/f2fs.h +13 -0
@@ -791,6 +791,11 @@ struct f2fs_sb_info {/* For sysfs suppport */ struct kobject s_kobj;struct completion s_kobj_unregister; + +/* For shrinker support */ +struct list_head s_list; +struct mutex umount_mutex; +unsigned int shrinker_run_no; };/*
@@ -1952,6 +1957,14 @@ int f2fs_read_inline_dir(struct file *, struct dir_context *, struct f2fs_str *);/* +* shrinker.c +*/ +unsigned long f2fs_shrink_count (struct shrinker *,struct shrink_control *); +unsigned long f2fs_shrink_scan (struct shrinker *,struct shrink_control *); +void f2fs_join_shrinker (struct f2fs_sb_info *); +void f2fs_leave_shrinker (struct f2fs_sb_info *); + +/* * crypto support */ static inline int f2fs_encrypted_inode (struct inode *inode)
fs/f2fs/shrinker.c new file +104 -0
104 lines not on this page · raw message
fs/f2fs/super.c +24 -0
@@ -39,6 +39,13 @@ static struct proc_dir_entry *f2fs_proc_root;static struct kmem_cache *f2fs_inode_cachep;static struct kset *f2fs_kset; +/* f2fs-wide shrinker description */ +static struct shrinker f2fs_shrinker_info = { + .scan_objects = f2fs_shrink_scan, + .count_objects = f2fs_shrink_count, + .seeks =DEFAULT_SEEKS , +}; +enum { Opt_gc_background, Opt_disable_roll_forward,
@@ -500,6 +507,9 @@ static void f2fs_put_super(struct super_block *sb)stop_gc_thread (sbi); +/* prevent remaining shrinker jobs */ +mutex_lock (&sbi->umount_mutex); +/* * We don't need to do checkpoint when superblock is clean. * But, the previous checkpoint was not done by umount, it needs to do
@@ -523,6 +533,9 @@ static void f2fs_put_super(struct super_block *sb)release_dirty_inode (sbi);release_discard_addrs (sbi); +f2fs_leave_shrinker (sbi); +mutex_unlock (&sbi->umount_mutex); +iput (sbi->node_inode);iput (sbi->meta_inode);
@@ -972,6 +985,9 @@ static void init_sb_info(struct f2fs_sb_info *sbi) sbi->dir_level =DEF_DIR_LEVEL ;clear_sbi_flag (sbi,SBI_NEED_FSCK ); + +INIT_LIST_HEAD (&sbi->s_list); +mutex_init (&sbi->umount_mutex); }/*
@@ -1214,6 +1230,8 @@ try_onemore: goto free_nm; } +f2fs_join_shrinker (sbi); +/* if there are nt orphan nodes free them */ recover_orphan_inodes (sbi);
@@ -1310,6 +1328,7 @@ free_root_inode:dput (sb->s_root); sb->s_root =NULL ; free_node_inode: +f2fs_leave_shrinker (sbi);iput (sbi->node_inode); free_nm:destroy_node_manager (sbi);
@@ -1406,6 +1425,9 @@ static int __init init_f2fs_fs(void) err =f2fs_init_crypto ();if (err) goto free_kset; + +register_shrinker (&f2fs_shrinker_info); + err =register_filesystem (&f2fs_fs_type);if (err) goto free_crypto;
@@ -1414,6 +1436,7 @@ static int __init init_f2fs_fs(void)return 0 ; free_crypto: +unregister_shrinker (&f2fs_shrinker_info);f2fs_exit_crypto (); free_kset:kset_unregister (f2fs_kset);
@@ -1435,6 +1458,7 @@ static void __exit exit_f2fs_fs(void) {remove_proc_entry ("fs/f2fs" ,NULL );f2fs_destroy_root_stats (); +unregister_shrinker (&f2fs_shrinker_info);unregister_filesystem (&f2fs_fs_type);f2fs_exit_crypto ();destroy_extent_cache ();
Signature · 9 lines
2.1.1 ------------------------------------------------------------------------------ Don't Limit Your Business. Reach for the Cloud. GigeNET's Cloud Solutions provide you with the tools and support that you need to offload your IT needs and focus on growing your business. Configured For All Businesses. Start Your Cloud Today. https://www.gigenetcloud.com/
RE: [f2fs-dev] [PATCH 08/12] f2fs: introduce a shrinker for mounted fs
Quoted 8 lines
-----Original Message-----
From: Jaegeuk Kim [mailto:[email protected]]
Sent: Tuesday, June 30, 2015 2:40 AM
To: [email protected]; [email protected]; [email protected]
Cc: Jaegeuk Kim
Subject: [f2fs-dev] [PATCH 08/12] f2fs: introduce a shrinker for mounted fs
Quoted 8 lines
This patch introduces a shrinker targeting to reduce memory footprint consumed by a number of in-memory f2fs data structures.
In addition, it newly adds:
-sbi->umount_mutexto avoid data races on shrinker andput_super
-sbi->shruinker_run_noto not revisit objectsNoteh that the basic implementation was copied from
fs/btrfs/shrinker.c
Great! Good to see it's being implemented in f2fs.
Quoted 2 lines
Signed-off-by Jaegeuk Kim <[email protected]>
[snip]
fs/f2fs/super.c static int __init init_f2fs_fs(void) · line 1429
Quoted 6 lines
fs/f2fs/super.c +2
-0@@ -1406,6 +1425,9 @@ static int __init init_f2fs_fs(void) err =f2fs_init_crypto ();if (err) goto free_kset; + +register_shrinker (&f2fs_shrinker_info);
This function can fail due to no memory, please check the return value here.
Thanks,
------------------------------------------------------------------------------
Don't Limit Your Business. Reach for the Cloud.
GigeNET's Cloud Solutions provide you with the tools and support that you need to offload your IT needs and focus on growing your business. Configured For All Businesses. Start Your Cloud Today.
https://www.gigenetcloud.com/
Jaegeuk Kim
maintainerRe: [f2fs-dev] [PATCH 08/12] f2fs: introduce a shrinker for mounted fs
Quoted 26 lines · Chao Yu
Quoted 8 lines
-----Original Message-----
From: Jaegeuk Kim [mailto:[email protected]]
Sent: Tuesday, June 30, 2015 2:40 AM
To: [email protected]; [email protected]; [email protected]
Cc: Jaegeuk Kim
Subject: [f2fs-dev] [PATCH 08/12] f2fs: introduce a shrinker for mounted fsQuoted 8 lines
This patch introduces a shrinker targeting to reduce memory footprint consumed by a number of in-memory f2fs data structures.
In addition, it newly adds:
-sbi->umount_mutexto avoid data races on shrinker andput_super
-sbi->shruinker_run_noto not revisit objectsNoteh that the basic implementation was copied from
fs/btrfs/shrinker.cGreat! Good to see it's being implemented in f2fs.
Quoted 2 lines
Signed-off-by Jaegeuk Kim <[email protected]>Reviewed-by Chao Yu <[email protected]>[snip]
Quoted 8 lines
fs/f2fs/super.c static int __init init_f2fs_fs(void) · line 1429
Quoted 6 lines
fs/f2fs/super.c +2
-0@@ -1406,6 +1425,9 @@ static int __init init_f2fs_fs(void) err =f2fs_init_crypto ();if (err) goto free_kset; + +register_shrinker (&f2fs_shrinker_info);This function can fail due to no memory, please check the return value here.
Agreed, done.
Thanks,
Quoted 2 lines
Thanks,
RE: [f2fs-dev] [PATCH 08/12] f2fs: introduce a shrinker for mounted fs
In reply to Jaegeuk KimQuoted 8 lines
-----Original Message-----
From: Jaegeuk Kim [mailto:[email protected]]
Sent: Tuesday, June 30, 2015 2:40 AM
To: [email protected]; [email protected]; [email protected]
Cc: Jaegeuk Kim
Subject: [f2fs-dev] [PATCH 08/12] f2fs: introduce a shrinker for mounted fs
Quoted 8 lines
This patch introduces a shrinker targeting to reduce memory footprint consumed by a number of in-memory f2fs data structures.
In addition, it newly adds:
-sbi->umount_mutexto avoid data races on shrinker andput_super
-sbi->shruinker_run_noto not revisit objectsNoteh that the basic implementation was copied from
fs/btrfs/shrinker.c
This file seems not exist...
fs/f2fs/super.c free_root_inode: · line 1331
Quoted 5 lines
fs/f2fs/super.c +1
-0@@ -1310,6 +1328,7 @@ free_root_inode:dput (sb->s_root); sb->s_root =NULL ; free_node_inode: +f2fs_leave_shrinker (sbi);
We should detach shrinker under sbi->umount_mutex. Otherwise we will access freed memory in following call path:
mount shrinker
->fill_super
Failed after f2fs_join_shrinker
->f2fs_leave_shrinker
->f2fs_shrink_scan
spin_lock
get sbi pointer
spin_unlock
spin_lock
list_del sbi->s_list
spin_unlock
free sbi
use-after-free for sbiThanks,
------------------------------------------------------------------------------
Don't Limit Your Business. Reach for the Cloud.
GigeNET's Cloud Solutions provide you with the tools and support that you need to offload your IT needs and focus on growing your business. Configured For All Businesses. Start Your Cloud Today.
https://www.gigenetcloud.com/
Jaegeuk Kim
maintainerRe: [f2fs-dev] [PATCH 08/12] f2fs: introduce a shrinker for mounted fs
Quoted 32 lines · Chao Yu
Quoted 8 lines
-----Original Message-----
From: Jaegeuk Kim [mailto:[email protected]]
Sent: Tuesday, June 30, 2015 2:40 AM
To: [email protected]; [email protected]; [email protected]
Cc: Jaegeuk Kim
Subject: [f2fs-dev] [PATCH 08/12] f2fs: introduce a shrinker for mounted fsQuoted 8 lines
This patch introduces a shrinker targeting to reduce memory footprint consumed by a number of in-memory f2fs data structures.
In addition, it newly adds:
-sbi->umount_mutexto avoid data races on shrinker andput_super
-sbi->shruinker_run_noto not revisit objectsNoteh that the basic implementation was copied from
fs/btrfs/shrinker.cThis file seems not exist...
fs/f2fs/super.c free_root_inode: · line 1331
Quoted 5 lines
fs/f2fs/super.c +1
-0@@ -1310,6 +1328,7 @@ free_root_inode:dput (sb->s_root); sb->s_root =NULL ; free_node_inode: +f2fs_leave_shrinker (sbi);We should detach shrinker under
sbi->umount_mutex. Otherwise we will access freed memory in following call path:mount shrinker
->fill_super
Failed afterf2fs_join_shrinker
->f2fs_leave_shrinker
Quoted 9 lines
->f2fs_shrink_scan
spin_lock
get sbi pointer
spin_unlock
spin_lock
list_del sbi->s_list
spin_unlock
free sbi
use-after-free for sbiRight, confirmed this.
Thanks,
Quoted 13 lines
Thanks,
------------------------------------------------------------------------------
Don't Limit Your Business. Reach for the Cloud.
GigeNET's Cloud Solutions provide you with the tools and support that you need to offload your IT needs and focus on growing your business. Configured For All Businesses. Start Your Cloud Today.
https://www.gigenetcloud.com/
_______________________________________________
Linux-f2fs-devel mailing list [email protected]
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel
------------------------------------------------------------------------------
Don't Limit Your Business. Reach for the Cloud.
GigeNET's Cloud Solutions provide you with the tools and support that you need to offload your IT needs and focus on growing your business. Configured For All Businesses. Start Your Cloud Today.
https://www.gigenetcloud.com/
Jaegeuk Kim
maintainerPATCH 9/12 f2fs: shrink nat_cache entries
In reply to Jaegeuk KimThis patch registers shrinking nat_cache entries.
--- fs/f2fs/node.c | 6 +++--- fs/f2fs/segment.c | 8 ++++++-- fs/f2fs/shrinker.c | 11 +++++++++-- 3 files changed, 18 insertions(+), 7 deletions(-)
fs/f2fs/node.c +3 -3
@@ -328,11 +328,11 @@ static void set_node_addr(struct f2fs_sb_info *sbi, struct node_info *ni,int try_to_free_nats (struct f2fs_sb_info *sbi,int nr_shrink) {struct f2fs_nm_info *nm_i =NM_I (sbi); +int nr = nr_shrink; -if (available_free_memory (sbi,NAT_ENTRIES )) +if (!down_write_trylock (&nm_i->nat_tree_lock))return 0 ; -down_write (&nm_i->nat_tree_lock);while (nr_shrink && !list_empty (&nm_i->nat_entries)) {struct nat_entry *ne; ne = list_first_entry(&nm_i->nat_entries,
@@ -341,7 +341,7 @@ int try_to_free_nats(struct f2fs_sb_info *sbi, int nr_shrink) nr_shrink--; }up_write (&nm_i->nat_tree_lock); -return nr_shrink; +return nr - nr_shrink; }/*
fs/f2fs/segment.c +6 -2
@@ -305,8 +305,12 @@ void f2fs_balance_fs_bg(struct f2fs_sb_info *sbi)/* try to shrink extent cache when there is no enough memory */ f2fs_shrink_extent_tree (sbi,EXTENT_CACHE_SHRINK_NUMBER ); -/* check the # of cached NAT entries and prefree segments */ -if (try_to_free_nats (sbi,NAT_ENTRY_PER_BLOCK ) || +/* check the # of cached NAT entries */ +if (!available_free_memory (sbi,NAT_ENTRIES )) +try_to_free_nats (sbi,NAT_ENTRY_PER_BLOCK ); + +/* checkpoint is the only way to shrink partial cached entries */ +if (!available_free_memory (sbi,NAT_ENTRIES ) ||excess_prefree_segs (sbi) || !available_free_memory (sbi,INO_ENTRIES ))f2fs_sync_fs (sbi->sb, true);
fs/f2fs/shrinker.c +9 -2
@@ -18,6 +18,11 @@ static LIST_HEAD(f2fs_list);static DEFINE_SPINLOCK (f2fs_list_lock);static unsigned int shrinker_run_no; +static unsigned long __count_nat_entries (struct f2fs_sb_info *sbi) +{ +return NM_I (sbi)->nat_cnt -NM_I (sbi)->dirty_nat_cnt; +} +unsigned long f2fs_shrink_count (struct shrinker *shrink,struct shrink_control *sc) {
@@ -37,7 +42,8 @@ unsigned long f2fs_shrink_count(struct shrinker *shrink, }spin_unlock (&f2fs_list_lock); -/* TODO: count # of objects */ +/* shrink clean nat cache entries */ + count +=__count_nat_entries (sbi);spin_lock (&f2fs_list_lock); p = p->next;
@@ -76,7 +82,8 @@ unsigned long f2fs_shrink_scan(struct shrinker *shrink, sbi->shrinker_run_no = run_no; -/* TODO: shrink caches */ +/* shrink clean nat cache entries */ + freed +=try_to_free_nats (sbi, nr);spin_lock (&f2fs_list_lock); p = p->next;
Signature · 9 lines
2.1.1 ------------------------------------------------------------------------------ Don't Limit Your Business. Reach for the Cloud. GigeNET's Cloud Solutions provide you with the tools and support that you need to offload your IT needs and focus on growing your business. Configured For All Businesses. Start Your Cloud Today. https://www.gigenetcloud.com/
RE: [f2fs-dev] [PATCH 09/12] f2fs: shrink nat_cache entries
Quoted 11 lines
-----Original Message-----
From: Jaegeuk Kim [mailto:[email protected]]
Sent: Tuesday, June 30, 2015 2:40 AM
To: [email protected]; [email protected]; [email protected]
Cc: Jaegeuk Kim
Subject: [f2fs-dev] [PATCH 09/12] f2fs: shrinknat_cacheentriesThis patch registers shrinking
nat_cacheentries.Signed-off-by Jaegeuk Kim <[email protected]>
------------------------------------------------------------------------------
Don't Limit Your Business. Reach for the Cloud.
GigeNET's Cloud Solutions provide you with the tools and support that you need to offload your IT needs and focus on growing your business. Configured For All Businesses. Start Your Cloud Today.
https://www.gigenetcloud.com/
Jaegeuk Kim
maintainerPATCH 10/12 f2fs: shrink extent_cache entries
In reply to Jaegeuk KimThis patch registers shrinking extent_caches.
--- fs/f2fs/data.c | 19 +++++++++++-------- fs/f2fs/f2fs.h | 2 +- fs/f2fs/segment.c | 3 ++- fs/f2fs/shrinker.c | 14 +++++++++++++- 4 files changed, 27 insertions(+), 11 deletions(-)
fs/f2fs/data.c +11 -8
@@ -767,7 +767,7 @@ out:update_inode_page (inode); } -void f2fs_shrink_extent_tree (struct f2fs_sb_info *sbi,int nr_shrink) +unsigned int f2fs_shrink_extent_tree (struct f2fs_sb_info *sbi,int nr_shrink) {struct extent_tree *treevec[EXT_TREE_VEC_SIZE ];struct extent_node *en, *tmp;
@@ -778,10 +778,7 @@ void f2fs_shrink_extent_tree(struct f2fs_sb_info *sbi, int nr_shrink)unsigned int node_cnt =0 , tree_cnt =0 ;if (!test_opt (sbi,EXTENT_CACHE )) -return ; - -if (available_free_memory (sbi,EXTENT_CACHE )) -return ; +return 0 ;spin_lock (&sbi->extent_lock);list_for_each_entry_safe (en, tmp, &sbi->extent_list, list) {
@@ -791,7 +788,9 @@ void f2fs_shrink_extent_tree(struct f2fs_sb_info *sbi, int nr_shrink) }spin_unlock (&sbi->extent_lock); -down_read (&sbi->extent_tree_lock); +if (!down_read_trylock (&sbi->extent_tree_lock)) + goto out; +while ((found =radix_tree_gang_lookup (&sbi->extent_tree_root, (void **)treevec, ino,EXT_TREE_VEC_SIZE ))) {unsigned i;
@@ -809,7 +808,9 @@ void f2fs_shrink_extent_tree(struct f2fs_sb_info *sbi, int nr_shrink) }up_read (&sbi->extent_tree_lock); -down_write (&sbi->extent_tree_lock); +if (!down_write_trylock (&sbi->extent_tree_lock)) + goto out; +radix_tree_for_each_slot (slot, &sbi->extent_tree_root, &iter,F2FS_ROOT_INO (sbi)) {struct extent_tree *et = (struct extent_tree *)*slot;
@@ -822,8 +823,10 @@ void f2fs_shrink_extent_tree(struct f2fs_sb_info *sbi, int nr_shrink) } }up_write (&sbi->extent_tree_lock); - +out:trace_f2fs_shrink_extent_tree (sbi, node_cnt, tree_cnt); + +return node_cnt + tree_cnt; }void f2fs_destroy_extent_tree (struct inode *inode)
fs/f2fs/f2fs.h +1 -1
@@ -1754,7 +1754,7 @@ void f2fs_submit_page_mbio(struct f2fs_io_info *);void set_data_blkaddr (struct dnode_of_data *);int reserve_new_block (struct dnode_of_data *);int f2fs_reserve_block (struct dnode_of_data *,pgoff_t ); -void f2fs_shrink_extent_tree (struct f2fs_sb_info *,int ); +unsigned int f2fs_shrink_extent_tree (struct f2fs_sb_info *,int );void f2fs_destroy_extent_tree (struct inode *);void f2fs_init_extent_cache (struct inode *,struct f2fs_extent *);void f2fs_update_extent_cache (struct dnode_of_data *);
fs/f2fs/segment.c +2 -1
@@ -303,7 +303,8 @@ void f2fs_balance_fs(struct f2fs_sb_info *sbi)void f2fs_balance_fs_bg (struct f2fs_sb_info *sbi) {/* try to shrink extent cache when there is no enough memory */ -f2fs_shrink_extent_tree (sbi,EXTENT_CACHE_SHRINK_NUMBER ); +if (!available_free_memory (sbi,EXTENT_CACHE )) +f2fs_shrink_extent_tree (sbi,EXTENT_CACHE_SHRINK_NUMBER );/* check the # of cached NAT entries */ if (!available_free_memory (sbi,NAT_ENTRIES ))
fs/f2fs/shrinker.c +13 -1
@@ -23,6 +23,11 @@ static unsigned long __count_nat_entries(struct f2fs_sb_info *sbi)return NM_I (sbi)->nat_cnt -NM_I (sbi)->dirty_nat_cnt; } +static unsigned long __count_extent_cache (struct f2fs_sb_info *sbi) +{ +return sbi->total_ext_tree +atomic_read (&sbi->total_ext_node); +} +unsigned long f2fs_shrink_count (struct shrinker *shrink,struct shrink_control *sc) {
@@ -42,6 +47,9 @@ unsigned long f2fs_shrink_count(struct shrinker *shrink, }spin_unlock (&f2fs_list_lock); +/* count extent cache entries */ + count +=__count_extent_cache (sbi); +/* shrink clean nat cache entries */ count +=__count_nat_entries (sbi);
@@ -82,8 +90,12 @@ unsigned long f2fs_shrink_scan(struct shrinker *shrink, sbi->shrinker_run_no = run_no; +/* shrink extent cache entries */ + freed +=f2fs_shrink_extent_tree (sbi, nr >>1 ); +/* shrink clean nat cache entries */ - freed +=try_to_free_nats (sbi, nr); +if (freed < nr) + freed +=try_to_free_nats (sbi, nr - freed);spin_lock (&f2fs_list_lock); p = p->next;
Signature · 9 lines
2.1.1 ------------------------------------------------------------------------------ Don't Limit Your Business. Reach for the Cloud. GigeNET's Cloud Solutions provide you with the tools and support that you need to offload your IT needs and focus on growing your business. Configured For All Businesses. Start Your Cloud Today. https://www.gigenetcloud.com/
RE: [f2fs-dev] [PATCH 10/12] f2fs: shrink extent_cache entries
Quoted 11 lines
-----Original Message-----
From: Jaegeuk Kim [mailto:[email protected]]
Sent: Tuesday, June 30, 2015 2:40 AM
To: [email protected]; [email protected]; [email protected]
Cc: Jaegeuk Kim
Subject: [f2fs-dev] [PATCH 10/12] f2fs: shrinkextent_cacheentriesThis patch registers shrinking
extent_caches.Signed-off-by Jaegeuk Kim <[email protected]>
------------------------------------------------------------------------------
Don't Limit Your Business. Reach for the Cloud.
GigeNET's Cloud Solutions provide you with the tools and support that you need to offload your IT needs and focus on growing your business. Configured For All Businesses. Start Your Cloud Today.
https://www.gigenetcloud.com/
Jaegeuk Kim
maintainerPATCH 11/12 f2fs: add noextent_cache mount option
In reply to Jaegeuk KimThis patch adds noextent_cache mount option.
--- fs/f2fs/super.c | 5 +++++ 1 file changed, 5 insertions(+)
fs/f2fs/super.c +5 -0
@@ -65,6 +65,7 @@ enum { Opt_nobarrier, Opt_fastboot, Opt_extent_cache, + Opt_noextent_cache, Opt_noinline_data, Opt_err, };
@@ -88,6 +89,7 @@ static match_table_t f2fs_tokens = { {Opt_nobarrier, "nobarrier"}, {Opt_fastboot, "fastboot"}, {Opt_extent_cache, "extent_cache"}, + {Opt_noextent_cache, "noextent_cache"}, {Opt_noinline_data, "noinline_data"}, {Opt_err, NULL}, };
@@ -389,6 +391,9 @@ static int parse_options(struct super_block *sb, char *options)case Opt_extent_cache:set_opt (sbi,EXTENT_CACHE );break ; +case Opt_noextent_cache: +clear_opt (sbi,EXTENT_CACHE ); +break ;case Opt_noinline_data:clear_opt (sbi,INLINE_DATA );break ;
Signature · 9 lines
2.1.1 ------------------------------------------------------------------------------ Don't Limit Your Business. Reach for the Cloud. GigeNET's Cloud Solutions provide you with the tools and support that you need to offload your IT needs and focus on growing your business. Configured For All Businesses. Start Your Cloud Today. https://www.gigenetcloud.com/
RE: [f2fs-dev] [PATCH 11/12] f2fs: add noextent_cache mount option
Quoted 9 lines
-----Original Message-----
From: Jaegeuk Kim [mailto:[email protected]]
Sent: Tuesday, June 30, 2015 2:40 AM
To: [email protected]; [email protected]; [email protected]
Cc: Jaegeuk Kim
Subject: [f2fs-dev] [PATCH 11/12] f2fs: addnoextent_cachemount optionThis patch adds
noextent_cachemount option.
How about adding some descriptions in Documentation to help our user?
The other thing is we'd better to show 'noextent_cache' option when user lookups the mount options.
How about adding codes as below?
if (test_opt(sbi, EXTENT_CACHE))
seq_puts(seq, ",extent_cache");
+else
+ seq_puts(seq, ",noextent_cache");
Quoted 2 lines
Signed-off-by Jaegeuk Kim <[email protected]>
The other looks good to me, so please add:
------------------------------------------------------------------------------
Don't Limit Your Business. Reach for the Cloud.
GigeNET's Cloud Solutions provide you with the tools and support that you need to offload your IT needs and focus on growing your business. Configured For All Businesses. Start Your Cloud Today.
https://www.gigenetcloud.com/
Jaegeuk Kim
maintainerRe: [f2fs-dev] [PATCH 11/12] f2fs: add noextent_cache mount option
Quoted 19 lines · Chao Yu
Quoted 9 lines
-----Original Message-----
From: Jaegeuk Kim [mailto:[email protected]]
Sent: Tuesday, June 30, 2015 2:40 AM
To: [email protected]; [email protected]; [email protected]
Cc: Jaegeuk Kim
Subject: [f2fs-dev] [PATCH 11/12] f2fs: addnoextent_cachemount optionThis patch adds
noextent_cachemount option.How about adding some descriptions in Documentation to help our user?
The other thing is we'd better to show '
noextent_cache' option when user lookups the mount options.How about adding codes as below?
if (test_opt(sbi, EXTENT_CACHE))seq_puts(seq, ",extent_cache");
Review
Quoted 2 lines
(quoted diff) +2
-0+else + seq_puts(seq, ",noextent_cache");
Agreed.
Thanks,
Quoted 7 lines
Quoted 2 lines
Signed-off-by Jaegeuk Kim <[email protected]>The other looks good to me, so please add:
Reviewed-by Chao Yu <[email protected]>
Quoted 10 lines
------------------------------------------------------------------------------
Don't Limit Your Business. Reach for the Cloud.
GigeNET's Cloud Solutions provide you with the tools and support that you need to offload your IT needs and focus on growing your business. Configured For All Businesses. Start Your Cloud Today.
https://www.gigenetcloud.com/
_______________________________________________
Linux-f2fs-devel mailing list [email protected]
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel
------------------------------------------------------------------------------
Don't Limit Your Business. Reach for the Cloud.
GigeNET's Cloud Solutions provide you with the tools and support that you need to offload your IT needs and focus on growing your business. Configured For All Businesses. Start Your Cloud Today.
https://www.gigenetcloud.com/
Jaegeuk Kim
maintainerPATCH 12/12 f2fs: use extent_cache by default
In reply to Jaegeuk KimWe don't need to handle the duplicate extent infot showrmation.
The integrated rule is:
- update on-disk extent with largest one tracked by in-memory extent_cache
- destroy extent_tree for the truncation case
- drop per-inode extent_cache by shrinker
--- fs/f2fs/data.c | 349 ++++++++++++++++------------------------------------- fs/f2fs/f2fs.h | 20 ++- fs/f2fs/inode.c | 18 ++- fs/f2fs/namei.c | 2 + fs/f2fs/shrinker.c | 2 + fs/f2fs/super.c | 8 +- 6 files changed, 136 insertions(+), 263 deletions(-)
fs/f2fs/data.c +101 -248
@@ -266,103 +266,6 @@ int f2fs_reserve_block(struct dnode_of_data *dn, pgoff_t index)return err; } -static bool lookup_extent_info (struct inode *inode,pgoff_t pgofs, -struct extent_info *ei) -{ -struct f2fs_inode_info *fi =F2FS_I (inode); -pgoff_t start_fofs, end_fofs; -block_t start_blkaddr; - -read_lock (&fi->ext_lock); -if (fi->ext.len ==0 ) { -read_unlock (&fi->ext_lock); -return false; - } - -stat_inc_total_hit (inode->i_sb); - - start_fofs = fi->ext.fofs; - end_fofs = fi->ext.fofs + fi->ext.len -1 ; - start_blkaddr = fi->ext.blk; - -if (pgofs >= start_fofs && pgofs <= end_fofs) { - *ei = fi->ext; -stat_inc_read_hit (inode->i_sb); -read_unlock (&fi->ext_lock); -return true; - } -read_unlock (&fi->ext_lock); -return false; -} - -static bool update_extent_info (struct inode *inode,pgoff_t fofs, -block_t blkaddr) -{ -struct f2fs_inode_info *fi =F2FS_I (inode); -pgoff_t start_fofs, end_fofs; -block_t start_blkaddr, end_blkaddr; -int need_update = true; - -write_lock (&fi->ext_lock); - - start_fofs = fi->ext.fofs; - end_fofs = fi->ext.fofs + fi->ext.len -1 ; - start_blkaddr = fi->ext.blk; - end_blkaddr = fi->ext.blk + fi->ext.len -1 ; - -/* Drop and initialize the matched extent */ -if (fi->ext.len ==1 && fofs == start_fofs) - fi->ext.len =0 ; - -/* Initial extent */ -if (fi->ext.len ==0 ) { -if (blkaddr !=NULL_ADDR ) { - fi->ext.fofs = fofs; - fi->ext.blk = blkaddr; - fi->ext.len =1 ; - } - goto end_update; - } - -/* Front merge */ -if (fofs == start_fofs -1 && blkaddr == start_blkaddr -1 ) { - fi->ext.fofs--; - fi->ext.blk--; - fi->ext.len++; - goto end_update; - } - -/* Back merge */ -if (fofs == end_fofs +1 && blkaddr == end_blkaddr +1 ) { - fi->ext.len++; - goto end_update; - } - -/* Split the existing extent */ -if (fi->ext.len >1 && - fofs >= start_fofs && fofs <= end_fofs) { -if ((end_fofs - fofs) < (fi->ext.len >>1 )) { - fi->ext.len = fofs - start_fofs; - }else { - fi->ext.fofs = fofs +1 ; - fi->ext.blk = start_blkaddr + fofs - start_fofs +1 ; - fi->ext.len -= fofs - start_fofs +1 ; - } - }else { - need_update = false; - } - -/* Finally, if the extent is very fragmented, let's drop the cache. */ -if (fi->ext.len <F2FS_MIN_EXTENT_LEN ) { - fi->ext.len =0 ; -set_inode_flag (fi,FI_NO_EXTENT ); - need_update = true; - } -end_update: -write_unlock (&fi->ext_lock); -return need_update; -} -static struct extent_node *__attach_extent_node (struct f2fs_sb_info *sbi,struct extent_tree *et,struct extent_info *ei,struct rb_node *parent,struct rb_node **p)
@@ -394,23 +297,6 @@ static void __detach_extent_node(struct f2fs_sb_info *sbi, et->cached_en =NULL ; } -static struct extent_tree *__find_extent_tree (struct f2fs_sb_info *sbi, -nid_t ino) -{ -struct extent_tree *et; - -down_read (&sbi->extent_tree_lock); - et =radix_tree_lookup (&sbi->extent_tree_root, ino); -if (!et) { -up_read (&sbi->extent_tree_lock); -return NULL ; - } -atomic_inc (&et->refcount); -up_read (&sbi->extent_tree_lock); - -return et; -} -static struct extent_tree *__grab_extent_tree (struct inode *inode) {struct f2fs_sb_info *sbi =F2FS_I_SB (inode);
@@ -434,6 +320,9 @@ static struct extent_tree *__grab_extent_tree(struct inode *inode)atomic_inc (&et->refcount);up_write (&sbi->extent_tree_lock); +/* never died untill evict_inode */ +F2FS_I (inode)->extent_tree = et; +return et; }
@@ -522,7 +411,7 @@ static struct extent_node *__insert_extent_tree(struct f2fs_sb_info *sbi, en->ei.len += ei->len;if (den) *den =__try_back_merge (sbi, et, en); -return en; + goto update_out; } p = &(*p)->rb_left; }else if (ei->fofs >= en->ei.fofs + en->ei.len) {
@@ -530,7 +419,7 @@ static struct extent_node *__insert_extent_tree(struct f2fs_sb_info *sbi, en->ei.len += ei->len;if (den) *den =__try_front_merge (sbi, et, en); -return en; + goto update_out; } p = &(*p)->rb_right; }else {
@@ -538,7 +427,11 @@ static struct extent_node *__insert_extent_tree(struct f2fs_sb_info *sbi, } } -return __attach_extent_node (sbi, et, ei, parent, p); + en =__attach_extent_node (sbi, et, ei, parent, p); +update_out: +if (en && en->ei.len > et->largest.len) + et->largest = en->ei; +return en; } static unsigned int __free_extent_tree(struct f2fs_sb_info *sbi,
@@ -570,26 +463,36 @@ static unsigned int __free_extent_tree(struct f2fs_sb_info *sbi,return count - et->count; } -static void f2fs_init_extent_tree (struct inode *inode, -struct f2fs_extent *i_ext) +static void __drop_largest_extent (struct inode *inode,pgoff_t fofs) +{ +struct extent_info *largest = &F2FS_I (inode)->extent_tree->largest; + +if (largest->fofs <= fofs && largest->fofs + largest->len > fofs) + largest->len =0 ; +} + +void f2fs_init_extent_tree (struct inode *inode,struct f2fs_extent *i_ext) {struct f2fs_sb_info *sbi =F2FS_I_SB (inode);struct extent_tree *et;struct extent_node *en;struct extent_info ei; -if (le32_to_cpu (i_ext->len) <F2FS_MIN_EXTENT_LEN ) +if (!f2fs_may_extent_tree (inode))return ; et =__grab_extent_tree (inode); -write_lock (&et->lock); -if (et->count) - goto out; +if (!i_ext ||le32_to_cpu (i_ext->len) <F2FS_MIN_EXTENT_LEN ) +return ;set_extent_info (&ei,le32_to_cpu (i_ext->fofs),le32_to_cpu (i_ext->blk),le32_to_cpu (i_ext->len)); +write_lock (&et->lock); +if (et->count) + goto out; + en =__insert_extent_tree (sbi, et, &ei,NULL );if (en) { et->cached_en = en;
@@ -600,21 +503,18 @@ static void f2fs_init_extent_tree(struct inode *inode, } out:write_unlock (&et->lock); -atomic_dec (&et->refcount); }static bool f2fs_lookup_extent_tree (struct inode *inode,pgoff_t pgofs,struct extent_info *ei) {struct f2fs_sb_info *sbi =F2FS_I_SB (inode); -struct extent_tree *et; +struct extent_tree *et =F2FS_I (inode)->extent_tree;struct extent_node *en; -trace_f2fs_lookup_extent_tree_start (inode, pgofs); +f2fs_bug_on (sbi, !et); - et =__find_extent_tree (sbi, inode->i_ino); -if (!et) -return false; +trace_f2fs_lookup_extent_tree_start (inode, pgofs);read_lock (&et->lock); en =__lookup_extent_tree (et, pgofs);
@@ -631,27 +531,35 @@ static bool f2fs_lookup_extent_tree(struct inode *inode, pgoff_t pgofs,read_unlock (&et->lock);trace_f2fs_lookup_extent_tree_end (inode, pgofs, en); - -atomic_dec (&et->refcount);return en ? true : false; } -static void f2fs_update_extent_tree (struct inode *inode,pgoff_t fofs, +/* return true, if on-disk extent should be updated */ +static bool f2fs_update_extent_tree (struct inode *inode,pgoff_t fofs,block_t blkaddr) {struct f2fs_sb_info *sbi =F2FS_I_SB (inode); -struct extent_tree *et; +struct extent_tree *et =F2FS_I (inode)->extent_tree;struct extent_node *en =NULL , *en1 =NULL , *en2 =NULL , *en3 =NULL ;struct extent_node *den =NULL ; -struct extent_info ei, dei; +struct extent_info ei, dei, prev;unsigned int endofs; -trace_f2fs_update_extent_tree (inode, fofs, blkaddr); +if (!et) +return false; - et =__grab_extent_tree (inode); +trace_f2fs_update_extent_tree (inode, fofs, blkaddr);write_lock (&et->lock); +if (is_inode_flag_set (F2FS_I (inode),FI_NO_EXTENT )) { +write_unlock (&et->lock); +return false; + } + + prev = et->largest; + dei.len =0 ; +/* 1. lookup and remove existing extent info in cache */ en =__lookup_extent_tree (et, fofs);if (!en)
@@ -660,6 +568,8 @@ static void f2fs_update_extent_tree(struct inode *inode, pgoff_t fofs, dei = en->ei;__detach_extent_node (sbi, et, en); +__drop_largest_extent (inode, fofs); +/* 2. if extent can be split more, split and insert the left part */ if (dei.len >1 ) {/* insert left part of split extent into cache */
@@ -683,6 +593,14 @@ update_extent:if (blkaddr) {set_extent_info (&ei, fofs, blkaddr,1 ); en3 =__insert_extent_tree (sbi, et, &ei, &den); + +/* give up extent_cache, if split and small updates happen */ +if (dei.len >=1 && + prev.len <F2FS_MIN_EXTENT_LEN && + et->largest.len <F2FS_MIN_EXTENT_LEN ) { + et->largest.len =0 ; +set_inode_flag (F2FS_I (inode),FI_NO_EXTENT ); + } }/* 4. update in global extent list */
@@ -714,57 +632,12 @@ update_extent:if (den)kmem_cache_free (extent_node_slab, den); -write_unlock (&et->lock); -atomic_dec (&et->refcount); -} - -void f2fs_preserve_extent_tree (struct inode *inode) -{ -struct extent_tree *et; -struct extent_info *ext = &F2FS_I (inode)->ext; -bool sync = false; - -if (!test_opt (F2FS_I_SB (inode),EXTENT_CACHE )) -return ; - - et =__find_extent_tree (F2FS_I_SB (inode), inode->i_ino); -if (!et) { -if (ext->len) { - ext->len =0 ; -update_inode_page (inode); - } -return ; - } - -read_lock (&et->lock); -if (et->count) { -struct extent_node *en; - -if (et->cached_en) { - en = et->cached_en; - }else { -struct rb_node *node =rb_first (&et->root); - -if (!node) - node =rb_last (&et->root); - en = rb_entry(node, struct extent_node, rb_node); - } - -if (__is_extent_same (ext, &en->ei)) - goto out; +if (is_inode_flag_set (F2FS_I (inode),FI_NO_EXTENT )) +__free_extent_tree (sbi, et, true); - *ext = en->ei; - sync = true; - }else if (ext->len) { - ext->len =0 ; - sync = true; - } -out: -read_unlock (&et->lock); -atomic_dec (&et->refcount); +write_unlock (&et->lock); -if (sync) -update_inode_page (inode); +return !__is_extent_same (&prev, &et->largest); }unsigned int f2fs_shrink_extent_tree (struct f2fs_sb_info *sbi,int nr_shrink)
@@ -772,8 +645,7 @@ unsigned int f2fs_shrink_extent_tree(struct f2fs_sb_info *sbi, int nr_shrink)struct extent_tree *treevec[EXT_TREE_VEC_SIZE ];struct extent_node *en, *tmp;unsigned long ino =F2FS_ROOT_INO (sbi); -struct radix_tree_iter iter; -void **slot; +struct radix_tree_root *root = &sbi->extent_tree_root;unsigned int found;unsigned int node_cnt =0 , tree_cnt =0 ;
@@ -788,10 +660,10 @@ unsigned int f2fs_shrink_extent_tree(struct f2fs_sb_info *sbi, int nr_shrink) }spin_unlock (&sbi->extent_lock); -if (!down_read_trylock (&sbi->extent_tree_lock)) +if (!down_write_trylock (&sbi->extent_tree_lock)) goto out; -while ((found =radix_tree_gang_lookup (&sbi->extent_tree_root, +while ((found =radix_tree_gang_lookup (root, (void **)treevec, ino,EXT_TREE_VEC_SIZE ))) {unsigned i;
@@ -799,27 +671,15 @@ unsigned int f2fs_shrink_extent_tree(struct f2fs_sb_info *sbi, int nr_shrink)for (i =0 ; i < found; i++) {struct extent_tree *et = treevec[i]; -atomic_inc (&et->refcount);write_lock (&et->lock); node_cnt +=__free_extent_tree (sbi, et, false);write_unlock (&et->lock); -atomic_dec (&et->refcount); - } - } -up_read (&sbi->extent_tree_lock); - -if (!down_write_trylock (&sbi->extent_tree_lock)) - goto out; - -radix_tree_for_each_slot (slot, &sbi->extent_tree_root, &iter, -F2FS_ROOT_INO (sbi)) { -struct extent_tree *et = (struct extent_tree *)*slot; - -if (!atomic_read (&et->refcount) && !et->count) { -radix_tree_delete (&sbi->extent_tree_root, et->ino); -kmem_cache_free (extent_tree_slab, et); - sbi->total_ext_tree--; - tree_cnt++; +if (!atomic_read (&et->refcount) && !et->count) { +radix_tree_delete (root, et->ino); +kmem_cache_free (extent_tree_slab, et); + sbi->total_ext_tree--; + tree_cnt++; + } } }up_write (&sbi->extent_tree_lock);
@@ -829,63 +689,59 @@ out:return node_cnt + tree_cnt; } -void f2fs_destroy_extent_tree (struct inode *inode) +void f2fs_destroy_extent_node (struct inode *inode) {struct f2fs_sb_info *sbi =F2FS_I_SB (inode); -struct extent_tree *et; +struct extent_tree *et =F2FS_I (inode)->extent_tree;unsigned int node_cnt =0 ; -if (!test_opt (sbi,EXTENT_CACHE )) -return ; - - et =__find_extent_tree (sbi, inode->i_ino);if (!et) - goto out; +return ; -/* free all extent info belong to this extent tree */ write_lock (&et->lock); node_cnt =__free_extent_tree (sbi, et, true);write_unlock (&et->lock); +} -atomic_dec (&et->refcount); +void f2fs_destroy_extent_tree (struct inode *inode) +{ +struct f2fs_sb_info *sbi =F2FS_I_SB (inode); +struct extent_tree *et =F2FS_I (inode)->extent_tree; +unsigned int node_cnt =0 ; -/* try to find and delete extent tree entry in radix tree */ -down_write (&sbi->extent_tree_lock); - et =radix_tree_lookup (&sbi->extent_tree_root, inode->i_ino); -if (!et) { -up_write (&sbi->extent_tree_lock); - goto out; +if (!et) +return ; + +if (inode->i_nlink && !is_bad_inode (inode) && et->count) { +atomic_dec (&et->refcount); +return ; } + +/* free all extent info belong to this extent tree */ +f2fs_destroy_extent_node (inode); + +/* delete extent tree entry in radix tree */ +down_write (&sbi->extent_tree_lock); +atomic_dec (&et->refcount);f2fs_bug_on (sbi,atomic_read (&et->refcount) || et->count);radix_tree_delete (&sbi->extent_tree_root, inode->i_ino);kmem_cache_free (extent_tree_slab, et); sbi->total_ext_tree--;up_write (&sbi->extent_tree_lock); -out: -trace_f2fs_destroy_extent_tree (inode, node_cnt); -return ; -} -void f2fs_init_extent_cache (struct inode *inode,struct f2fs_extent *i_ext) -{ -if (test_opt (F2FS_I_SB (inode),EXTENT_CACHE )) -f2fs_init_extent_tree (inode, i_ext); +F2FS_I (inode)->extent_tree =NULL ; -write_lock (&F2FS_I (inode)->ext_lock); -get_extent_info (&F2FS_I (inode)->ext, *i_ext); -write_unlock (&F2FS_I (inode)->ext_lock); +trace_f2fs_destroy_extent_tree (inode, node_cnt); +return ; }static bool f2fs_lookup_extent_cache (struct inode *inode,pgoff_t pgofs,struct extent_info *ei) { -if (is_inode_flag_set (F2FS_I (inode),FI_NO_EXTENT )) +if (!f2fs_may_extent_tree (inode))return false; -if (test_opt (F2FS_I_SB (inode),EXTENT_CACHE )) -return f2fs_lookup_extent_tree (inode, pgofs, ei); - -return lookup_extent_info (inode, pgofs, ei); +return f2fs_lookup_extent_tree (inode, pgofs, ei); }void f2fs_update_extent_cache (struct dnode_of_data *dn)
@@ -893,19 +749,15 @@ void f2fs_update_extent_cache(struct dnode_of_data *dn)struct f2fs_inode_info *fi =F2FS_I (dn->inode);pgoff_t fofs; -f2fs_bug_on (F2FS_I_SB (dn->inode), dn->data_blkaddr ==NEW_ADDR ); - -if (is_inode_flag_set (fi,FI_NO_EXTENT )) +if (!f2fs_may_extent_tree (dn->inode))return ; +f2fs_bug_on (F2FS_I_SB (dn->inode), dn->data_blkaddr ==NEW_ADDR ); + fofs =start_bidx_of_node (ofs_of_node (dn->node_page), fi) + dn->ofs_in_node; -/* we should call update_extent_info() to update on-disk extent */ -if (test_opt (F2FS_I_SB (dn->inode),EXTENT_CACHE )) -f2fs_update_extent_tree (dn->inode, fofs, dn->data_blkaddr); - -if (update_extent_info (dn->inode, fofs, dn->data_blkaddr)) +if (f2fs_update_extent_tree (dn->inode, fofs, dn->data_blkaddr))sync_inode_page (dn); }
@@ -1109,8 +961,6 @@ alloc:allocate_data_block (sbi,NULL , dn->data_blkaddr, &dn->data_blkaddr, &sum, seg); - -/* direct IO doesn't use extent cache to maximize the performance */ set_data_blkaddr (dn);/* update i_size */
@@ -1119,6 +969,9 @@ alloc:if (i_size_read (dn->inode) < ((fofs +1 ) <<PAGE_CACHE_SHIFT ))i_size_write (dn->inode, ((fofs +1 ) <<PAGE_CACHE_SHIFT )); +/* direct IO doesn't use extent cache to maximize the performance */ +__drop_largest_extent (dn->inode, fofs); +return 0 ; }
fs/f2fs/f2fs.h +16 -4
@@ -349,6 +349,7 @@ struct extent_tree {nid_t ino;/* inode number */ struct rb_root root;/* root of extent info rb-tree */ struct extent_node *cached_en;/* recently accessed extent node */ +struct extent_info largest;/* largested extent info */ rwlock_t lock;/* protect extent info rb-tree */ atomic_t refcount;/* reference count of rb-tree */ unsigned int count;/* # of extent node in rb-tree*/
@@ -420,14 +421,14 @@ struct f2fs_inode_info {unsigned int clevel;/* maximum level of given file name */ nid_t i_xattr_nid;/* node id that contains xattrs */ unsigned long long xattr_ver;/* cp version of xattr modification */ -struct extent_info ext;/* in-memory extent cache entry */ -rwlock_t ext_lock;/* rwlock for single extent cache */ struct inode_entry *dirty_dir;/* the pointer of dirty dir */ struct radix_tree_root inmem_root;/* radix tree for inmem pages */ struct list_head inmem_pages;/* inmemory pages managed by f2fs */ struct mutex inmem_lock;/* lock for inmemory pages */ +struct extent_tree *extent_tree;/* cached extent_tree entry */ +#ifdef CONFIG_F2FS_FS_ENCRYPTION /* Encryption params */ struct f2fs_crypt_info *i_crypt_info;
@@ -1548,6 +1549,17 @@ static inline bool is_dot_dotdot(const struct qstr *str)return false; } +static inline bool f2fs_may_extent_tree (struct inode *inode) +{ +mode_t mode = inode->i_mode; + +if (!test_opt (F2FS_I_SB (inode),EXTENT_CACHE ) || +is_inode_flag_set (F2FS_I (inode),FI_NO_EXTENT )) +return false; + +return S_ISREG (mode); +} +#define get_inode_mode (i) \ ((is_inode_flag_set(F2FS_I(i), FI_ACL_MODE)) ? \ (F2FS_I(i)->i_acl_mode) : ((i)->i_mode))
@@ -1755,10 +1767,10 @@ void set_data_blkaddr(struct dnode_of_data *);int reserve_new_block (struct dnode_of_data *);int f2fs_reserve_block (struct dnode_of_data *,pgoff_t );unsigned int f2fs_shrink_extent_tree (struct f2fs_sb_info *,int ); +void f2fs_init_extent_tree (struct inode *,struct f2fs_extent *); +void f2fs_destroy_extent_node (struct inode *);void f2fs_destroy_extent_tree (struct inode *); -void f2fs_init_extent_cache (struct inode *,struct f2fs_extent *);void f2fs_update_extent_cache (struct dnode_of_data *); -void f2fs_preserve_extent_tree (struct inode *);struct page *get_read_data_page (struct inode *,pgoff_t ,int );struct page *find_data_page (struct inode *,pgoff_t );struct page *get_lock_data_page (struct inode *,pgoff_t );
fs/f2fs/inode.c +8 -10
@@ -139,7 +139,7 @@ static int do_read_inode(struct inode *inode) fi->i_pino =le32_to_cpu (ri->i_pino); fi->i_dir_level = ri->i_dir_level; -f2fs_init_extent_cache (inode, &ri->i_ext); +f2fs_init_extent_tree (inode, &ri->i_ext);get_inline_info (fi, ri);
@@ -237,10 +237,11 @@ void update_inode(struct inode *inode, struct page *node_page) ri->i_size =cpu_to_le64 (i_size_read (inode)); ri->i_blocks =cpu_to_le64 (inode->i_blocks); -read_lock (&F2FS_I (inode)->ext_lock); -set_raw_extent (&F2FS_I (inode)->ext, &ri->i_ext); -read_unlock (&F2FS_I (inode)->ext_lock); - +if (F2FS_I (inode)->extent_tree) +set_raw_extent (&F2FS_I (inode)->extent_tree->largest, + &ri->i_ext); +else +memset (&ri->i_ext,0 ,sizeof (ri->i_ext));set_raw_inline (F2FS_I (inode), ri); ri->i_atime =cpu_to_le64 (inode->i_atime.tv_sec);
@@ -331,6 +332,8 @@ void f2fs_evict_inode(struct inode *inode)f2fs_bug_on (sbi,get_dirty_pages (inode));remove_dirty_dir_inode (inode); +f2fs_destroy_extent_tree (inode); +if (inode->i_nlink ||is_bad_inode (inode)) goto no_delete;
@@ -350,11 +353,6 @@ no_delete:stat_dec_inline_dir (inode);stat_dec_inline_inode (inode); -/* update extent info in inode */ -if (inode->i_nlink) -f2fs_preserve_extent_tree (inode); -f2fs_destroy_extent_tree (inode); -invalidate_mapping_pages (NODE_MAPPING (sbi), inode->i_ino, inode->i_ino);if (xnid)invalidate_mapping_pages (NODE_MAPPING (sbi), xnid, xnid);
fs/f2fs/namei.c +2 -0
@@ -65,6 +65,8 @@ static struct inode *f2fs_new_inode(struct inode *dir, umode_t mode)if (f2fs_may_inline_dentry (inode))set_inode_flag (F2FS_I (inode),FI_INLINE_DENTRY ); +f2fs_init_extent_tree (inode,NULL ); +stat_inc_inline_inode (inode);stat_inc_inline_dir (inode);
fs/f2fs/shrinker.c +2 -0
@@ -117,6 +117,8 @@ void f2fs_join_shrinker(struct f2fs_sb_info *sbi)void f2fs_leave_shrinker (struct f2fs_sb_info *sbi) { +f2fs_shrink_extent_tree (sbi,__count_extent_cache (sbi)); +spin_lock (&f2fs_list_lock);list_del (&sbi->s_list);spin_unlock (&f2fs_list_lock);
fs/f2fs/super.c +7 -1
@@ -422,7 +422,6 @@ static struct inode *f2fs_alloc_inode(struct super_block *sb)atomic_set (&fi->dirty_pages,0 ); fi->i_current_depth =1 ; fi->i_advise =0 ; -rwlock_init (&fi->ext_lock);init_rwsem (&fi->i_sem);INIT_RADIX_TREE (&fi->inmem_root,GFP_NOFS );INIT_LIST_HEAD (&fi->inmem_pages);
@@ -453,12 +452,17 @@ static int f2fs_drop_inode(struct inode *inode)*/ if (!inode_unhashed (inode) && inode->i_state &I_SYNC ) {if (!inode->i_nlink && !is_bad_inode (inode)) { +/* to avoid evict_inode call simultaneously */ +atomic_inc (&inode->i_count);spin_unlock (&inode->i_lock);/* some remained atomic pages should discarded */ if (f2fs_is_atomic_file (inode))commit_inmem_pages (inode, true); +/* should remain fi->extent_tree for writepage */ +f2fs_destroy_extent_node (inode); +sb_start_intwrite (inode->i_sb);i_size_write (inode,0 );
@@ -473,6 +477,7 @@ static int f2fs_drop_inode(struct inode *inode) F2FS_I(inode)->i_crypt_info);#endif spin_lock (&inode->i_lock); +atomic_dec (&inode->i_count); }return 0 ; }
@@ -719,6 +724,7 @@ static void default_options(struct f2fs_sb_info *sbi)set_opt (sbi,BG_GC );set_opt (sbi,INLINE_DATA ); +set_opt (sbi,EXTENT_CACHE );#ifdef CONFIG_F2FS_FS_XATTR set_opt (sbi,XATTR_USER );
Signature · 9 lines
2.1.1 ------------------------------------------------------------------------------ Don't Limit Your Business. Reach for the Cloud. GigeNET's Cloud Solutions provide you with the tools and support that you need to offload your IT needs and focus on growing your business. Configured For All Businesses. Start Your Cloud Today. https://www.gigenetcloud.com/
RE: [f2fs-dev] [PATCH 12/12] f2fs: use extent_cache by default
Hi Jaegeuk,
Quoted 9 lines
-----Original Message-----
From: Jaegeuk Kim [mailto:[email protected]]
Sent: Tuesday, June 30, 2015 2:40 AM
To: [email protected]; [email protected]; [email protected]
Cc: Jaegeuk Kim
Subject: [f2fs-dev] [PATCH 12/12] f2fs: useextent_cacheby defaultWe don't need to handle the duplicate extent infot showrmation.
information?
Quoted 7 lines
The integrated rule is:
- update on-disk extent with largest one tracked by in-memoryextent_cache
- destroyextent_treefor the truncation case
- drop per-inodeextent_cacheby shrinkerSigned-off-by Jaegeuk Kim <[email protected]>
[snip]
fs/f2fs/data.c static struct extent_node *__insert_extent_tree(struct f2fs_sb_info *sbi, · line 433
Quoted 9 lines
fs/f2fs/data.c +4
-1@@ -538,7 +427,11 @@ static struct extent_node *__insert_extent_tree(struct f2fs_sb_info *sbi, } } -return __attach_extent_node (sbi, et, ei, parent, p); + en =__attach_extent_node (sbi, et, ei, parent, p); +update_out: +if (en && en->ei.len > et->largest.len) + et->largest = en->ei;
IMO, it's better to update cached_en here if it is invalid in __detach_extent_node, then cached_en and largest may point different extent info, it can expand our region of first level extent cache.
[snip]
fs/f2fs/data.c out: · line 721
Quoted 3 lines
fs/f2fs/data.c +3
-0+ +/* free all extent info belong to this extent tree */ +f2fs_destroy_extent_node (inode);
How about returning number of freed extent node for tracing.
node_cnt = f2fs_destroy_extent_node(inode);
[snip]
fs/f2fs/inode.c void update_inode(struct inode *inode, struct page *node_page) · line 240
Quoted 9 lines
fs/f2fs/inode.c +1
-4@@ -237,10 +237,11 @@ void update_inode(struct inode *inode, struct page *node_page) ri->i_size =cpu_to_le64 (i_size_read (inode)); ri->i_blocks =cpu_to_le64 (inode->i_blocks); -read_lock (&F2FS_I (inode)->ext_lock); -set_raw_extent (&F2FS_I (inode)->ext, &ri->i_ext); -read_unlock (&F2FS_I (inode)->ext_lock); - +if (F2FS_I (inode)->extent_tree)
Could extent cache destroy after above check?
Thanks,
------------------------------------------------------------------------------
Don't Limit Your Business. Reach for the Cloud.
GigeNET's Cloud Solutions provide you with the tools and support that you need to offload your IT needs and focus on growing your business. Configured For All Businesses. Start Your Cloud Today.
https://www.gigenetcloud.com/
Jaegeuk Kim
maintainerRe: [f2fs-dev] [PATCH 12/12] f2fs: use extent_cache by default
Hi Chao,
Quoted 10 lines · Chao Yu
Hi Jaegeuk,
Quoted 8 lines
-----Original Message-----
From: Jaegeuk Kim [mailto:[email protected]]
Sent: Tuesday, June 30, 2015 2:40 AM
To: [email protected]; [email protected]; [email protected]
Cc: Jaegeuk Kim
Subject: [f2fs-dev] [PATCH 12/12] f2fs: useextent_cacheby default
Quoted 3 lines
Quoted 1 line
We don't need to handle the duplicate extent infot showrmation.
information?
Fixed.
Quoted 15 lines
Quoted 8 lines
The integrated rule is:
- update on-disk extent with largest one tracked by in-memoryextent_cache
- destroyextent_treefor the truncation case
- drop per-inodeextent_cacheby shrinkerSigned-off-by Jaegeuk Kim <[email protected]>[snip]
Quoted 4 lines
(quoted diff) +0
-0@@ -538,7 +427,11 @@ static struct extent_node *__insert_extent_tree(struct f2fs_sb_info *sbi, } }
Quoted 9 lines
fs/f2fs/data.c static struct extent_node *__insert_extent_tree(struct f2fs_sb_info *sbi, · line 433
Quoted 5 lines
fs/f2fs/data.c +4
-1-return __attach_extent_node (sbi, et, ei, parent, p); + en =__attach_extent_node (sbi, et, ei, parent, p); +update_out: +if (en && en->ei.len > et->largest.len) + et->largest = en->ei;IMO, it's better to update
cached_enhere if it is invalid in__detach_extent_node, thencached_enand largest may point different extent info, it can expand our region of first level extent cache.
Agreed.
Quoted 10 lines
[snip]
fs/f2fs/data.c out: · line 721
Quoted 3 lines
fs/f2fs/data.c +3
-0+ +/* free all extent info belong to this extent tree */ +f2fs_destroy_extent_node (inode);How about returning number of freed extent node for tracing.
node_cnt=f2fs_destroy_extent_node(inode);
No problem.
Quoted 7 lines
[snip]
Quoted 4 lines
fs/f2fs/inode.c +0
-0@@ -237,10 +237,11 @@ void update_inode(struct inode *inode, struct page *node_page) ri->i_size =cpu_to_le64 (i_size_read (inode)); ri->i_blocks =cpu_to_le64 (inode->i_blocks);
Quoted 7 lines
fs/f2fs/inode.c void update_inode(struct inode *inode, struct page *node_page) · line 240
Quoted 5 lines
fs/f2fs/inode.c +1
-4-read_lock (&F2FS_I (inode)->ext_lock); -set_raw_extent (&F2FS_I (inode)->ext, &ri->i_ext); -read_unlock (&F2FS_I (inode)->ext_lock); - +if (F2FS_I (inode)->extent_tree)Could extent cache destroy after above check?
I don't think so.
The extent_tree is assigned as one way.
Once it is assigned, it will be deallocated only after evict_inode.
Thanks,
Quoted 13 lines
Thanks,
------------------------------------------------------------------------------
Don't Limit Your Business. Reach for the Cloud.
GigeNET's Cloud Solutions provide you with the tools and support that you need to offload your IT needs and focus on growing your business. Configured For All Businesses. Start Your Cloud Today.
https://www.gigenetcloud.com/
_______________________________________________
Linux-f2fs-devel mailing list [email protected]
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel
------------------------------------------------------------------------------
Don't Limit Your Business. Reach for the Cloud.
GigeNET's Cloud Solutions provide you with the tools and support that you need to offload your IT needs and focus on growing your business. Configured For All Businesses. Start Your Cloud Today.
https://www.gigenetcloud.com/
RE: [f2fs-dev] [PATCH 12/12] f2fs: use extent_cache by default
Hi Jaegeuk,
Quoted 7 lines
-----Original Message-----
From: Jaegeuk Kim [mailto:[email protected]]
Sent: Saturday, July 04, 2015 1:16 PM
To: Chao Yu
Cc: [email protected]; [email protected]; [email protected]
Subject: Re: [f2fs-dev] [PATCH 12/12] f2fs: useextent_cacheby default
[snip]
Quoted 12 lines
Quoted 11 lines
fs/f2fs/inode.c void update_inode(struct inode *inode, struct page *node_page) · line 240
Quoted 9 lines
fs/f2fs/inode.c +1
-4@@ -237,10 +237,11 @@ void update_inode(struct inode *inode, struct page *node_page) ri->i_size =cpu_to_le64 (i_size_read (inode)); ri->i_blocks =cpu_to_le64 (inode->i_blocks); -read_lock (&F2FS_I (inode)->ext_lock); -set_raw_extent (&F2FS_I (inode)->ext, &ri->i_ext); -read_unlock (&F2FS_I (inode)->ext_lock); - +if (F2FS_I (inode)->extent_tree)Could extent cache destroy after above check?
Quoted 4 lines
I don't think so.
The
extent_treeis assigned as one way.
Once it is assigned, it will be deallocated only afterevict_inode.
Previously, I suspected that ->write_inode and ->evict will be executed concurrently.
After checking the code, I find that would not happen, so we are safe.
Thanks,
Quoted 5 lines
Thanks,
Quoted 2 lines
Thanks,
------------------------------------------------------------------------------
Don't Limit Your Business. Reach for the Cloud.
GigeNET's Cloud Solutions provide you with the tools and support that you need to offload your IT needs and focus on growing your business. Configured For All Businesses. Start Your Cloud Today.
https://www.gigenetcloud.com/
Chao Yu
Reviewed-byRE: [f2fs-dev] [PATCH 12/12] f2fs: use extent_cache by default
Quoted 8 lines
-----Original Message-----
From: Chao Yu [mailto:[email protected]]
Sent: Saturday, July 04, 2015 2:30 PM
To: 'Jaegeuk Kim'
Cc: [email protected]; [email protected]; [email protected]
Subject: Re: [f2fs-dev] [PATCH 12/12] f2fs: useextent_cacheby default
Quoted 9 lines
Hi Jaegeuk,
Quoted 7 lines
-----Original Message-----
From: Jaegeuk Kim [mailto:[email protected]]
Sent: Saturday, July 04, 2015 1:16 PM
To: Chao Yu
Cc: [email protected]; [email protected]; [email protected]
Subject: Re: [f2fs-dev] [PATCH 12/12] f2fs: useextent_cacheby default
-----Original Message-----
Quoted 11 lines
From: Jaegeuk Kim [mailto:[email protected]]
Sent: Tuesday, June 30, 2015 2:39 AM
To: [email protected]; [email protected]; [email protected]
Cc: Jaegeuk Kim
Subject: [f2fs-dev] [PATCH 01/12] f2fs: avoid freed stat informationThe
write_checkpointcan update stat information, so we should destroy the stat structure after it.Signed-off-by Jaegeuk Kim <[email protected]>
------------------------------------------------------------------------------
Don't Limit Your Business. Reach for the Cloud.
GigeNET's Cloud Solutions provide you with the tools and support that you need to offload your IT needs and focus on growing your business. Configured For All Businesses. Start Your Cloud Today.
https://www.gigenetcloud.com/