From 109f55651954def97fa41ee71c464d268c512ab0 Mon Sep 17 00:00:00 2001 From: Mingming Date: Tue, 10 Nov 2009 10:48:08 -0500 Subject: ext4: fix ext4_ext_direct_IO()'s return value after converting uninit extents After a direct I/O request covering an uninitalized extent (i.e., created using the fallocate system call) or a hole in a file, ext4 will convert the uninitialized extent so it is marked as initialized by calling ext4_convert_unwritten_extents(). This function returns zero on success. This return value was getting returned by ext4_direct_IO(); however the file system's direct_IO function is supposed to return the number of bytes read or written on a success. By returning zero, it confused the direct I/O code into falling back to buffered I/O unnecessarily. Signed-off-by: Mingming Cao Signed-off-by: "Theodore Ts'o" --- fs/ext4/extents.c | 1 + 1 file changed, 1 insertion(+) (limited to 'fs/ext4/extents.c') diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c index 10539e364283..441716f33b4a 100644 --- a/fs/ext4/extents.c +++ b/fs/ext4/extents.c @@ -3519,6 +3519,7 @@ retry: * * This function is called from the direct IO end io call back * function, to convert the fallocated extents after IO is completed. + * Returns 0 on success. */ int ext4_convert_unwritten_extents(struct inode *inode, loff_t offset, loff_t len) -- cgit From 5f5249507e4b5c4fc0f9c93f33d133d8c95f47e1 Mon Sep 17 00:00:00 2001 From: Mingming Date: Tue, 10 Nov 2009 10:48:04 -0500 Subject: ext4: skip conversion of uninit extents after direct IO if there isn't any At the end of direct I/O operation, ext4_ext_direct_IO() always called ext4_convert_unwritten_extents(), regardless of whether there were any unwritten extents involved in the I/O or not. This commit adds a state flag so that ext4_ext_direct_IO() only calls ext4_convert_unwritten_extents() when necessary. Signed-off-by: Mingming Cao Signed-off-by: "Theodore Ts'o" --- fs/ext4/ext4.h | 1 + fs/ext4/extents.c | 22 +++++++++++++++++----- fs/ext4/inode.c | 4 +++- 3 files changed, 21 insertions(+), 6 deletions(-) (limited to 'fs/ext4/extents.c') diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h index 00d153f2f261..8825515eeddd 100644 --- a/fs/ext4/ext4.h +++ b/fs/ext4/ext4.h @@ -322,6 +322,7 @@ static inline __u32 ext4_mask_flags(umode_t mode, __u32 flags) #define EXT4_STATE_NO_EXPAND 0x00000008 /* No space for expansion */ #define EXT4_STATE_DA_ALLOC_CLOSE 0x00000010 /* Alloc DA blks on close */ #define EXT4_STATE_EXT_MIGRATE 0x00000020 /* Inode is migrating */ +#define EXT4_STATE_DIO_UNWRITTEN 0x00000040 /* need convert on dio done*/ /* Used to pass group descriptor data when online resize is done */ struct ext4_new_group_input { diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c index 441716f33b4a..e991ae2b461c 100644 --- a/fs/ext4/extents.c +++ b/fs/ext4/extents.c @@ -3048,12 +3048,18 @@ ext4_ext_handle_uninitialized_extents(handle_t *handle, struct inode *inode, ret = ext4_split_unwritten_extents(handle, inode, path, iblock, max_blocks, flags); - /* flag the io_end struct that we need convert when IO done */ + /* + * Flag the inode(non aio case) or end_io struct (aio case) + * that this IO needs to convertion to written when IO is + * completed + */ if (io) io->flag = DIO_AIO_UNWRITTEN; + else + EXT4_I(inode)->i_state |= EXT4_STATE_DIO_UNWRITTEN; goto out; } - /* DIO end_io complete, convert the filled extent to written */ + /* async DIO end_io complete, convert the filled extent to written */ if (flags == EXT4_GET_BLOCKS_DIO_CONVERT_EXT) { ret = ext4_convert_unwritten_extents_dio(handle, inode, path); @@ -3295,10 +3301,16 @@ int ext4_ext_get_blocks(handle_t *handle, struct inode *inode, * To avoid unecessary convertion for every aio dio rewrite * to the mid of file, here we flag the IO that is really * need the convertion. - * + * For non asycn direct IO case, flag the inode state + * that we need to perform convertion when IO is done. */ - if (io && flags == EXT4_GET_BLOCKS_DIO_CREATE_EXT) - io->flag = DIO_AIO_UNWRITTEN; + if (flags == EXT4_GET_BLOCKS_DIO_CREATE_EXT) { + if (io) + io->flag = DIO_AIO_UNWRITTEN; + else + EXT4_I(inode)->i_state |= + EXT4_STATE_DIO_UNWRITTEN;; + } } err = ext4_ext_insert_extent(handle, inode, path, &newex, flags); if (err) { diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c index 12d727f8fedf..a9ed2bce74d1 100644 --- a/fs/ext4/inode.c +++ b/fs/ext4/inode.c @@ -3772,7 +3772,8 @@ static ssize_t ext4_ext_direct_IO(int rw, struct kiocb *iocb, if (ret != -EIOCBQUEUED && ret <= 0 && iocb->private) { ext4_free_io_end(iocb->private); iocb->private = NULL; - } else if (ret > 0) { + } else if (ret > 0 && (EXT4_I(inode)->i_state & + EXT4_STATE_DIO_UNWRITTEN)) { int err; /* * for non AIO case, since the IO is already @@ -3782,6 +3783,7 @@ static ssize_t ext4_ext_direct_IO(int rw, struct kiocb *iocb, offset, ret); if (err < 0) ret = err; + EXT4_I(inode)->i_state &= ~EXT4_STATE_DIO_UNWRITTEN; } return ret; } -- cgit From ba230c3f6dc88ec008806adb27b12088486d508e Mon Sep 17 00:00:00 2001 From: Mingming Date: Fri, 6 Nov 2009 04:01:23 -0500 Subject: ext4: Fix return value of ext4_split_unwritten_extents() to fix direct I/O To prepare for a direct I/O write, we need to split the unwritten extents before submitting the I/O. When no extents needed to be split, ext4_split_unwritten_extents() was incorrectly returning 0 instead of the size of uninitialized extents. This bug caused the wrong return value sent back to VFS code when it gets called from async IO path, leading to an unnecessary fall back to buffered IO. This bug also hid the fact that the check to see whether or not a split would be necessary was incorrect; we can only skip splitting the extent if the write completely covers the uninitialized extent. Signed-off-by: Mingming Cao Signed-off-by: "Theodore Ts'o" --- fs/ext4/extents.c | 13 +++++++------ 1 file changed, 7 insertions(+), 6 deletions(-) (limited to 'fs/ext4/extents.c') diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c index e991ae2b461c..715264b4bae4 100644 --- a/fs/ext4/extents.c +++ b/fs/ext4/extents.c @@ -2807,6 +2807,8 @@ fix_extent_len: * into three uninitialized extent(at most). After IO complete, the part * being filled will be convert to initialized by the end_io callback function * via ext4_convert_unwritten_extents(). + * + * Returns the size of uninitialized extent to be written on success. */ static int ext4_split_unwritten_extents(handle_t *handle, struct inode *inode, @@ -2824,7 +2826,6 @@ static int ext4_split_unwritten_extents(handle_t *handle, unsigned int allocated, ee_len, depth; ext4_fsblk_t newblock; int err = 0; - int ret = 0; ext_debug("ext4_split_unwritten_extents: inode %lu," "iblock %llu, max_blocks %u\n", inode->i_ino, @@ -2842,12 +2843,12 @@ static int ext4_split_unwritten_extents(handle_t *handle, ext4_ext_store_pblock(&orig_ex, ext_pblock(ex)); /* - * if the entire unintialized extent length less than - * the size of extent to write, there is no need to split - * uninitialized extent + * If the uninitialized extent begins at the same logical + * block where the write begins, and the write completely + * covers the extent, then we don't need to split it. */ - if (allocated <= max_blocks) - return ret; + if ((iblock == ee_block) && (allocated <= max_blocks)) + return allocated; err = ext4_ext_get_access(handle, inode, path + depth); if (err) -- cgit