From 8d3e72a180b42c01ec00045e1bb8eb91175adafe Mon Sep 17 00:00:00 2001 From: Andreas Gruenbacher Date: Thu, 27 Jun 2019 17:28:40 -0700 Subject: [PATCH 1/3] iomap: don't mark the inode dirty in iomap_write_end Marking the inode dirty for each page copied into the page cache can be very inefficient for file systems that use the VFS dirty inode tracking, and is completely pointless for those that don't use the VFS dirty inode tracking. So instead, only set an iomap flag when changing the in-core inode size, and open code the rest of __generic_write_end. Partially based on code from Christoph Hellwig. Signed-off-by: Andreas Gruenbacher Reviewed-by: Christoph Hellwig Reviewed-by: Darrick J. Wong Signed-off-by: Darrick J. Wong --- fs/gfs2/bmap.c | 2 ++ fs/iomap.c | 15 ++++++++++++++- include/linux/iomap.h | 1 + 3 files changed, 17 insertions(+), 1 deletion(-) diff --git a/fs/gfs2/bmap.c b/fs/gfs2/bmap.c index c78ccaf83ef8..8e8768685264 100644 --- a/fs/gfs2/bmap.c +++ b/fs/gfs2/bmap.c @@ -1179,6 +1179,8 @@ static int gfs2_iomap_end(struct inode *inode, loff_t pos, loff_t length, if (ip->i_qadata && ip->i_qadata->qa_qd_num) gfs2_quota_unlock(ip); + if (iomap->flags & IOMAP_F_SIZE_CHANGED) + mark_inode_dirty(inode); gfs2_write_unlock(inode); out: diff --git a/fs/iomap.c b/fs/iomap.c index 23ef63fd1669..c4acf69b7196 100644 --- a/fs/iomap.c +++ b/fs/iomap.c @@ -773,6 +773,7 @@ iomap_write_end(struct inode *inode, loff_t pos, unsigned len, unsigned copied, struct page *page, struct iomap *iomap) { const struct iomap_page_ops *page_ops = iomap->page_ops; + loff_t old_size = inode->i_size; int ret; if (iomap->type == IOMAP_INLINE) { @@ -784,7 +785,19 @@ iomap_write_end(struct inode *inode, loff_t pos, unsigned len, ret = __iomap_write_end(inode, pos, len, copied, page, iomap); } - __generic_write_end(inode, pos, ret, page); + /* + * Update the in-memory inode size after copying the data into the page + * cache. It's up to the file system to write the updated size to disk, + * preferably after I/O completion so that no stale data is exposed. + */ + if (pos + ret > old_size) { + i_size_write(inode, pos + ret); + iomap->flags |= IOMAP_F_SIZE_CHANGED; + } + unlock_page(page); + + if (old_size < pos) + pagecache_isize_extended(inode, old_size, pos); if (page_ops && page_ops->page_done) page_ops->page_done(inode, pos, copied, page, iomap); put_page(page); diff --git a/include/linux/iomap.h b/include/linux/iomap.h index 2103b94cb1bf..1df9ea187a9a 100644 --- a/include/linux/iomap.h +++ b/include/linux/iomap.h @@ -35,6 +35,7 @@ struct vm_fault; #define IOMAP_F_NEW 0x01 /* blocks have been newly allocated */ #define IOMAP_F_DIRTY 0x02 /* uncommitted metadata */ #define IOMAP_F_BUFFER_HEAD 0x04 /* file system requires buffer heads */ +#define IOMAP_F_SIZE_CHANGED 0x08 /* file size has changed */ /* * Flags that only need to be reported for IOMAP_REPORT requests: From 8af54f291e5cfdb215e0b3045cc365954c55855d Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Thu, 27 Jun 2019 17:28:40 -0700 Subject: [PATCH 2/3] fs: fold __generic_write_end back into generic_write_end This effectively reverts a6d639da63ae ("fs: factor out a __generic_write_end helper") as we now open code what is left of that helper in iomap. Signed-off-by: Christoph Hellwig Signed-off-by: Andreas Gruenbacher Reviewed-by: Darrick J. Wong Signed-off-by: Darrick J. Wong --- fs/buffer.c | 62 ++++++++++++++++++++++++--------------------------- fs/internal.h | 2 -- 2 files changed, 29 insertions(+), 35 deletions(-) diff --git a/fs/buffer.c b/fs/buffer.c index e450c55f6434..49a871570092 100644 --- a/fs/buffer.c +++ b/fs/buffer.c @@ -2086,38 +2086,6 @@ int block_write_begin(struct address_space *mapping, loff_t pos, unsigned len, } EXPORT_SYMBOL(block_write_begin); -void __generic_write_end(struct inode *inode, loff_t pos, unsigned copied, - struct page *page) -{ - loff_t old_size = inode->i_size; - bool i_size_changed = false; - - /* - * No need to use i_size_read() here, the i_size cannot change under us - * because we hold i_rwsem. - * - * But it's important to update i_size while still holding page lock: - * page writeout could otherwise come in and zero beyond i_size. - */ - if (pos + copied > inode->i_size) { - i_size_write(inode, pos + copied); - i_size_changed = true; - } - - unlock_page(page); - - if (old_size < pos) - pagecache_isize_extended(inode, old_size, pos); - /* - * Don't mark the inode dirty under page lock. First, it unnecessarily - * makes the holding time of page lock longer. Second, it forces lock - * ordering of page lock and transaction start for journaling - * filesystems. - */ - if (i_size_changed) - mark_inode_dirty(inode); -} - int block_write_end(struct file *file, struct address_space *mapping, loff_t pos, unsigned len, unsigned copied, struct page *page, void *fsdata) @@ -2158,9 +2126,37 @@ int generic_write_end(struct file *file, struct address_space *mapping, loff_t pos, unsigned len, unsigned copied, struct page *page, void *fsdata) { + struct inode *inode = mapping->host; + loff_t old_size = inode->i_size; + bool i_size_changed = false; + copied = block_write_end(file, mapping, pos, len, copied, page, fsdata); - __generic_write_end(mapping->host, pos, copied, page); + + /* + * No need to use i_size_read() here, the i_size cannot change under us + * because we hold i_rwsem. + * + * But it's important to update i_size while still holding page lock: + * page writeout could otherwise come in and zero beyond i_size. + */ + if (pos + copied > inode->i_size) { + i_size_write(inode, pos + copied); + i_size_changed = true; + } + + unlock_page(page); put_page(page); + + if (old_size < pos) + pagecache_isize_extended(inode, old_size, pos); + /* + * Don't mark the inode dirty under page lock. First, it unnecessarily + * makes the holding time of page lock longer. Second, it forces lock + * ordering of page lock and transaction start for journaling + * filesystems. + */ + if (i_size_changed) + mark_inode_dirty(inode); return copied; } EXPORT_SYMBOL(generic_write_end); diff --git a/fs/internal.h b/fs/internal.h index a48ef81be37d..2f3c3de51fad 100644 --- a/fs/internal.h +++ b/fs/internal.h @@ -40,8 +40,6 @@ static inline int __sync_blockdev(struct block_device *bdev, int wait) extern void guard_bio_eod(int rw, struct bio *bio); extern int __block_write_begin_int(struct page *page, loff_t pos, unsigned len, get_block_t *get_block, struct iomap *iomap); -void __generic_write_end(struct inode *inode, loff_t pos, unsigned copied, - struct page *page); /* * char_dev.c From 36a7347de097edf9c4d7203d09fa223c86479674 Mon Sep 17 00:00:00 2001 From: Andreas Gruenbacher Date: Thu, 27 Jun 2019 17:28:41 -0700 Subject: [PATCH 3/3] iomap: fix page_done callback for short writes When we truncate a short write to have it retried, pass the truncated length to the page_done callback instead of the full length. Signed-off-by: Andreas Gruenbacher Reviewed-by: Christoph Hellwig Reviewed-by: Darrick J. Wong Signed-off-by: Darrick J. Wong --- fs/iomap.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/fs/iomap.c b/fs/iomap.c index c4acf69b7196..f15c705c5a93 100644 --- a/fs/iomap.c +++ b/fs/iomap.c @@ -799,7 +799,7 @@ iomap_write_end(struct inode *inode, loff_t pos, unsigned len, if (old_size < pos) pagecache_isize_extended(inode, old_size, pos); if (page_ops && page_ops->page_done) - page_ops->page_done(inode, pos, copied, page, iomap); + page_ops->page_done(inode, pos, ret, page, iomap); put_page(page); if (ret < len)