On Tue, 25 Jun 2019 at 12:50, Christoph Hellwig <hch@xxxxxx> wrote: > > > That seems way more complicated. I'd much rather go with something > > like may patch plus maybe a big fat comment explaining that persisting > > the size update is the file systems job. Note that a lot of the modern > > file systems don't use the VFS inode tracking for that, besides XFS > > that includes at least btrfs and ocfs2 as well. > > I'd suggest something like this as the baseline: > > http://git.infradead.org/users/hch/xfs.git/shortlog/refs/heads/iomap-i_size Alright, can we change this as follows? [Also, I'm not really sure why we check for (pos + ret > inode->i_size) when we have already read inode->i_size into old_size.] Thanks, Andreas -- 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 ineffcient for file systems that use the VFS dirty inode tracking, and is completely pointless for those that don't use the VFS dirty tracking. Signed-off-by: Christoph Hellwig <hch@xxxxxx> Signed-off-by: Andreas Gruenbacher <agruenba@xxxxxxxxxx> --- 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 93ea1d5..f4b895f 100644 --- a/fs/gfs2/bmap.c +++ b/fs/gfs2/bmap.c @@ -1182,6 +1182,8 @@ 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 12654c2..cd24bd1 100644 --- a/fs/iomap.c +++ b/fs/iomap.c @@ -777,6 +777,7 @@ struct iomap_readpage_ctx { 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) { @@ -788,7 +789,19 @@ struct iomap_readpage_ctx { 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 > inode->i_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 2103b94..1df9ea1 100644 --- a/include/linux/iomap.h +++ b/include/linux/iomap.h @@ -35,6 +35,7 @@ #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: -- 1.8.3.1
