Message ID | 20200229001411.128010-1-guro@fb.com |
---|---|
State | Accepted |
Headers | show |
Series | [v2] ext4: use non-movable memory for superblock readahead | expand |
On Feb 28, 2020, at 5:14 PM, Roman Gushchin <guro@fb.com> wrote: > > Since commit a8ac900b8163 ("ext4: use non-movable memory for the > superblock") buffers for ext4 superblock were allocated using > the sb_bread_unmovable() helper which allocated buffer heads > out of non-movable memory blocks. It was necessarily to not block > page migrations and do not cause cma allocation failures. > > However commit 85c8f176a611 ("ext4: preload block group descriptors") > broke this by introducing pre-reading of the ext4 superblock. > The problem is that __breadahead() is using __getblk() underneath, > which allocates buffer heads out of movable memory. > > It resulted in page migration failures I've seen on a machine > with an ext4 partition and a preallocated cma area. > > Fix this by introducing sb_breadahead_unmovable() and > __breadahead_gfp() helpers which use non-movable memory for buffer > head allocations and use them for the ext4 superblock readahead. > > v2: found a similar issue in __ext4_get_inode_loc() > > Fixes: 85c8f176a611 ("ext4: preload block group descriptors") > Signed-off-by: Roman Gushchin <guro@fb.com> Reviewed-by: Andreas Dilger <adilger@dilger.ca> > Cc: Andrew Perepechko <andrew.perepechko@seagate.com> > Cc: Theodore Ts'o <tytso@mit.edu> > Cc: Gioh Kim <gioh.kim@lge.com> > Cc: Jan Kara <jack@suse.cz> > --- > fs/buffer.c | 11 +++++++++++ > fs/ext4/inode.c | 2 +- > fs/ext4/super.c | 2 +- > include/linux/buffer_head.h | 8 ++++++++ > 4 files changed, 21 insertions(+), 2 deletions(-) > > diff --git a/fs/buffer.c b/fs/buffer.c > index 4299e100a05b..25462edd920e 100644 > --- a/fs/buffer.c > +++ b/fs/buffer.c > @@ -1414,6 +1414,17 @@ void __breadahead(struct block_device *bdev, sector_t block, unsigned size) > } > EXPORT_SYMBOL(__breadahead); > > +void __breadahead_gfp(struct block_device *bdev, sector_t block, unsigned size, > + gfp_t gfp) > +{ > + struct buffer_head *bh = __getblk_gfp(bdev, block, size, gfp); > + if (likely(bh)) { > + ll_rw_block(REQ_OP_READ, REQ_RAHEAD, 1, &bh); > + brelse(bh); > + } > +} > +EXPORT_SYMBOL(__breadahead_gfp); > + > /** > * __bread_gfp() - reads a specified block and returns the bh > * @bdev: the block_device to read from > diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c > index fa0ff78dc033..b131fedc6b77 100644 > --- a/fs/ext4/inode.c > +++ b/fs/ext4/inode.c > @@ -4348,7 +4348,7 @@ static int __ext4_get_inode_loc(struct inode *inode, > if (end > table) > end = table; > while (b <= end) > - sb_breadahead(sb, b++); > + sb_breadahead_unmovable(sb, b++); > } > > /* > diff --git a/fs/ext4/super.c b/fs/ext4/super.c > index ff1b764b0c0e..fb2338a5220e 100644 > --- a/fs/ext4/super.c > +++ b/fs/ext4/super.c > @@ -4331,7 +4331,7 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent) > /* Pre-read the descriptors into the buffer cache */ > for (i = 0; i < db_count; i++) { > block = descriptor_loc(sb, logical_sb_block, i); > - sb_breadahead(sb, block); > + sb_breadahead_unmovable(sb, block); > } > > for (i = 0; i < db_count; i++) { > diff --git a/include/linux/buffer_head.h b/include/linux/buffer_head.h > index 7b73ef7f902d..b56cc825f64d 100644 > --- a/include/linux/buffer_head.h > +++ b/include/linux/buffer_head.h > @@ -189,6 +189,8 @@ struct buffer_head *__getblk_gfp(struct block_device *bdev, sector_t block, > void __brelse(struct buffer_head *); > void __bforget(struct buffer_head *); > void __breadahead(struct block_device *, sector_t block, unsigned int size); > +void __breadahead_gfp(struct block_device *, sector_t block, unsigned int size, > + gfp_t gfp); > struct buffer_head *__bread_gfp(struct block_device *, > sector_t block, unsigned size, gfp_t gfp); > void invalidate_bh_lrus(void); > @@ -319,6 +321,12 @@ sb_breadahead(struct super_block *sb, sector_t block) > __breadahead(sb->s_bdev, block, sb->s_blocksize); > } > > +static inline void > +sb_breadahead_unmovable(struct super_block *sb, sector_t block) > +{ > + __breadahead_gfp(sb->s_bdev, block, sb->s_blocksize, 0); > +} > + > static inline struct buffer_head * > sb_getblk(struct super_block *sb, sector_t block) > { > -- > 2.24.1 > Cheers, Andreas
On Sat, Feb 29, 2020 at 12:49:13AM -0700, Andreas Dilger wrote: > On Feb 28, 2020, at 5:14 PM, Roman Gushchin <guro@fb.com> wrote: > > > > Since commit a8ac900b8163 ("ext4: use non-movable memory for the > > superblock") buffers for ext4 superblock were allocated using > > the sb_bread_unmovable() helper which allocated buffer heads > > out of non-movable memory blocks. It was necessarily to not block > > page migrations and do not cause cma allocation failures. > > > > However commit 85c8f176a611 ("ext4: preload block group descriptors") > > broke this by introducing pre-reading of the ext4 superblock. > > The problem is that __breadahead() is using __getblk() underneath, > > which allocates buffer heads out of movable memory. > > > > It resulted in page migration failures I've seen on a machine > > with an ext4 partition and a preallocated cma area. > > > > Fix this by introducing sb_breadahead_unmovable() and > > __breadahead_gfp() helpers which use non-movable memory for buffer > > head allocations and use them for the ext4 superblock readahead. > > > > v2: found a similar issue in __ext4_get_inode_loc() > > > > Fixes: 85c8f176a611 ("ext4: preload block group descriptors") > > Signed-off-by: Roman Gushchin <guro@fb.com> > > Reviewed-by: Andreas Dilger <adilger@dilger.ca> Thank you!
On Sat, Feb 29, 2020 at 12:49:13AM -0700, Andreas Dilger wrote: > On Feb 28, 2020, at 5:14 PM, Roman Gushchin <guro@fb.com> wrote: > > > > Since commit a8ac900b8163 ("ext4: use non-movable memory for the > > superblock") buffers for ext4 superblock were allocated using > > the sb_bread_unmovable() helper which allocated buffer heads > > out of non-movable memory blocks. It was necessarily to not block > > page migrations and do not cause cma allocation failures. > > > > However commit 85c8f176a611 ("ext4: preload block group descriptors") > > broke this by introducing pre-reading of the ext4 superblock. > > The problem is that __breadahead() is using __getblk() underneath, > > which allocates buffer heads out of movable memory. > > > > It resulted in page migration failures I've seen on a machine > > with an ext4 partition and a preallocated cma area. > > > > Fix this by introducing sb_breadahead_unmovable() and > > __breadahead_gfp() helpers which use non-movable memory for buffer > > head allocations and use them for the ext4 superblock readahead. > > > > v2: found a similar issue in __ext4_get_inode_loc() > > > > Fixes: 85c8f176a611 ("ext4: preload block group descriptors") > > Signed-off-by: Roman Gushchin <guro@fb.com> > > Reviewed-by: Andreas Dilger <adilger@dilger.ca> Is it good to go? Can it go through the ext4 tree? Thanks!
On Sat, Feb 29, 2020 at 12:49:13AM -0700, Andreas Dilger wrote: > On Feb 28, 2020, at 5:14 PM, Roman Gushchin <guro@fb.com> wrote: > > > > Since commit a8ac900b8163 ("ext4: use non-movable memory for the > > superblock") buffers for ext4 superblock were allocated using > > the sb_bread_unmovable() helper which allocated buffer heads > > out of non-movable memory blocks. It was necessarily to not block > > page migrations and do not cause cma allocation failures. > > > > However commit 85c8f176a611 ("ext4: preload block group descriptors") > > broke this by introducing pre-reading of the ext4 superblock. > > The problem is that __breadahead() is using __getblk() underneath, > > which allocates buffer heads out of movable memory. > > > > It resulted in page migration failures I've seen on a machine > > with an ext4 partition and a preallocated cma area. > > > > Fix this by introducing sb_breadahead_unmovable() and > > __breadahead_gfp() helpers which use non-movable memory for buffer > > head allocations and use them for the ext4 superblock readahead. > > > > v2: found a similar issue in __ext4_get_inode_loc() > > > > Fixes: 85c8f176a611 ("ext4: preload block group descriptors") > > Signed-off-by: Roman Gushchin <guro@fb.com> > > Reviewed-by: Andreas Dilger <adilger@dilger.ca> Hello, Theodore! Can you, please, pick this patch? We've some changes on the mm side (more actively using a cma area for movable allocations), which might bring a regression without this ext4 change. Thank you! Roman
On Fri, Feb 28, 2020 at 04:14:11PM -0800, Roman Gushchin wrote: > Since commit a8ac900b8163 ("ext4: use non-movable memory for the > superblock") buffers for ext4 superblock were allocated using > the sb_bread_unmovable() helper which allocated buffer heads > out of non-movable memory blocks. It was necessarily to not block > page migrations and do not cause cma allocation failures. > > However commit 85c8f176a611 ("ext4: preload block group descriptors") > broke this by introducing pre-reading of the ext4 superblock. > The problem is that __breadahead() is using __getblk() underneath, > which allocates buffer heads out of movable memory. > > It resulted in page migration failures I've seen on a machine > with an ext4 partition and a preallocated cma area. > > Fix this by introducing sb_breadahead_unmovable() and > __breadahead_gfp() helpers which use non-movable memory for buffer > head allocations and use them for the ext4 superblock readahead. Applied, thanks. Apologies for not picking this up earlier. - Ted
On Thu, Apr 09, 2020 at 11:23:44PM -0400, Theodore Y. Ts'o wrote: > On Fri, Feb 28, 2020 at 04:14:11PM -0800, Roman Gushchin wrote: > > Since commit a8ac900b8163 ("ext4: use non-movable memory for the > > superblock") buffers for ext4 superblock were allocated using > > the sb_bread_unmovable() helper which allocated buffer heads > > out of non-movable memory blocks. It was necessarily to not block > > page migrations and do not cause cma allocation failures. > > > > However commit 85c8f176a611 ("ext4: preload block group descriptors") > > broke this by introducing pre-reading of the ext4 superblock. > > The problem is that __breadahead() is using __getblk() underneath, > > which allocates buffer heads out of movable memory. > > > > It resulted in page migration failures I've seen on a machine > > with an ext4 partition and a preallocated cma area. > > > > Fix this by introducing sb_breadahead_unmovable() and > > __breadahead_gfp() helpers which use non-movable memory for buffer > > head allocations and use them for the ext4 superblock readahead. > > Applied, thanks. Apologies for not picking this up earlier. > > - Ted Thank you! Roman
diff --git a/fs/buffer.c b/fs/buffer.c index 4299e100a05b..25462edd920e 100644 --- a/fs/buffer.c +++ b/fs/buffer.c @@ -1414,6 +1414,17 @@ void __breadahead(struct block_device *bdev, sector_t block, unsigned size) } EXPORT_SYMBOL(__breadahead); +void __breadahead_gfp(struct block_device *bdev, sector_t block, unsigned size, + gfp_t gfp) +{ + struct buffer_head *bh = __getblk_gfp(bdev, block, size, gfp); + if (likely(bh)) { + ll_rw_block(REQ_OP_READ, REQ_RAHEAD, 1, &bh); + brelse(bh); + } +} +EXPORT_SYMBOL(__breadahead_gfp); + /** * __bread_gfp() - reads a specified block and returns the bh * @bdev: the block_device to read from diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c index fa0ff78dc033..b131fedc6b77 100644 --- a/fs/ext4/inode.c +++ b/fs/ext4/inode.c @@ -4348,7 +4348,7 @@ static int __ext4_get_inode_loc(struct inode *inode, if (end > table) end = table; while (b <= end) - sb_breadahead(sb, b++); + sb_breadahead_unmovable(sb, b++); } /* diff --git a/fs/ext4/super.c b/fs/ext4/super.c index ff1b764b0c0e..fb2338a5220e 100644 --- a/fs/ext4/super.c +++ b/fs/ext4/super.c @@ -4331,7 +4331,7 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent) /* Pre-read the descriptors into the buffer cache */ for (i = 0; i < db_count; i++) { block = descriptor_loc(sb, logical_sb_block, i); - sb_breadahead(sb, block); + sb_breadahead_unmovable(sb, block); } for (i = 0; i < db_count; i++) { diff --git a/include/linux/buffer_head.h b/include/linux/buffer_head.h index 7b73ef7f902d..b56cc825f64d 100644 --- a/include/linux/buffer_head.h +++ b/include/linux/buffer_head.h @@ -189,6 +189,8 @@ struct buffer_head *__getblk_gfp(struct block_device *bdev, sector_t block, void __brelse(struct buffer_head *); void __bforget(struct buffer_head *); void __breadahead(struct block_device *, sector_t block, unsigned int size); +void __breadahead_gfp(struct block_device *, sector_t block, unsigned int size, + gfp_t gfp); struct buffer_head *__bread_gfp(struct block_device *, sector_t block, unsigned size, gfp_t gfp); void invalidate_bh_lrus(void); @@ -319,6 +321,12 @@ sb_breadahead(struct super_block *sb, sector_t block) __breadahead(sb->s_bdev, block, sb->s_blocksize); } +static inline void +sb_breadahead_unmovable(struct super_block *sb, sector_t block) +{ + __breadahead_gfp(sb->s_bdev, block, sb->s_blocksize, 0); +} + static inline struct buffer_head * sb_getblk(struct super_block *sb, sector_t block) {
Since commit a8ac900b8163 ("ext4: use non-movable memory for the superblock") buffers for ext4 superblock were allocated using the sb_bread_unmovable() helper which allocated buffer heads out of non-movable memory blocks. It was necessarily to not block page migrations and do not cause cma allocation failures. However commit 85c8f176a611 ("ext4: preload block group descriptors") broke this by introducing pre-reading of the ext4 superblock. The problem is that __breadahead() is using __getblk() underneath, which allocates buffer heads out of movable memory. It resulted in page migration failures I've seen on a machine with an ext4 partition and a preallocated cma area. Fix this by introducing sb_breadahead_unmovable() and __breadahead_gfp() helpers which use non-movable memory for buffer head allocations and use them for the ext4 superblock readahead. v2: found a similar issue in __ext4_get_inode_loc() Fixes: 85c8f176a611 ("ext4: preload block group descriptors") Signed-off-by: Roman Gushchin <guro@fb.com> Cc: Andrew Perepechko <andrew.perepechko@seagate.com> Cc: Theodore Ts'o <tytso@mit.edu> Cc: Gioh Kim <gioh.kim@lge.com> Cc: Jan Kara <jack@suse.cz> --- fs/buffer.c | 11 +++++++++++ fs/ext4/inode.c | 2 +- fs/ext4/super.c | 2 +- include/linux/buffer_head.h | 8 ++++++++ 4 files changed, 21 insertions(+), 2 deletions(-)