Commit ecfdc08b authored by Goldwyn Rodrigues's avatar Goldwyn Rodrigues Committed by David Sterba
Browse files

btrfs: remove dio iomap DSYNC workaround 



This effectively reverts 09745ff88d93 ("btrfs: dio iomap DSYNC
workaround") now that the iomap API has been updated to allow
iomap_dio_complete() not to be called under i_rwsem anymore.

Reviewed-by: default avatarJosef Bacik <josef@toxicpanda.com>
Signed-off-by: default avatarGoldwyn Rodrigues <rgoldwyn@suse.com>
Reviewed-by: default avatarDavid Sterba <dsterba@suse.com>
Signed-off-by: default avatarDavid Sterba <dsterba@suse.com>
parent a42fa643

fs/btrfs/ctree.h

+0 −1
Original line number Diff line number Diff line
@@ -3068,7 +3068,6 @@ void btrfs_writepage_endio_finish_ordered(struct page *page, u64 start, 
extern const struct dentry_operations btrfs_dentry_operations;

extern const struct iomap_ops btrfs_dio_iomap_ops;

extern const struct iomap_dio_ops btrfs_dio_ops;

extern const struct iomap_dio_ops btrfs_sync_dops;



/* Inode locking type flags, by default the exclusive lock is taken */

#define BTRFS_ILOCK_SHARED	(1U << 0)

fs/btrfs/file.c

+2 −36
Original line number Diff line number Diff line
@@ -2018,44 +2018,10 @@ static ssize_t btrfs_file_write_iter(struct kiocb *iocb, 
	if (sync)

		atomic_inc(&BTRFS_I(inode)->sync_writers);



	if (iocb->ki_flags & IOCB_DIRECT) {

		/*

		 * 1. We must always clear IOCB_DSYNC in order to not deadlock

		 *    in iomap, as it calls generic_write_sync() in this case.

		 * 2. If we are async, we can call iomap_dio_complete() either

		 *    in

		 *

		 *    2.1. A worker thread from the last bio completed.  In this

		 *	   case we need to mark the btrfs_dio_data that it is

		 *	   async in order to call generic_write_sync() properly.

		 *	   This is handled by setting BTRFS_DIO_SYNC_STUB in the

		 *	   current->journal_info.

		 *    2.2  The submitter context, because all IO completed

		 *         before we exited iomap_dio_rw().  In this case we can

		 *         just re-set the IOCB_DSYNC on the iocb and we'll do

		 *         the sync below.  If our ->end_io() gets called and

		 *         current->journal_info is set, then we know we're in

		 *         our current context and we will clear

		 *         current->journal_info to indicate that we need to

		 *         sync below.

		 */

		if (sync) {

			ASSERT(current->journal_info == NULL);

			iocb->ki_flags &= ~IOCB_DSYNC;

			current->journal_info = BTRFS_DIO_SYNC_STUB;

		}

	if (iocb->ki_flags & IOCB_DIRECT)

		num_written = btrfs_direct_write(iocb, from);



		/*

		 * As stated above, we cleared journal_info, so we need to do

		 * the sync ourselves.

		 */

		if (sync && current->journal_info == NULL)

			iocb->ki_flags |= IOCB_DSYNC;

		current->journal_info = NULL;

	} else {

	else

		num_written = btrfs_buffered_write(iocb, from);

	}



	/*

	 * We also have to set last_sub_trans to the current log transid,

fs/btrfs/inode.c

+0 −50
Original line number Diff line number Diff line
@@ -62,7 +62,6 @@ struct btrfs_dio_data { 
	loff_t length;

	ssize_t submitted;

	struct extent_changeset *data_reserved;

	bool sync;

};



static const struct inode_operations btrfs_dir_inode_operations;
@@ -7431,17 +7430,6 @@ static int btrfs_dio_iomap_begin(struct inode *inode, loff_t start, 
	int ret = 0;

	u64 len = length;

	bool unlock_extents = false;

	bool sync = (current->journal_info == BTRFS_DIO_SYNC_STUB);



	/*

	 * We used current->journal_info here to see if we were sync, but

	 * there's a lot of tests in the enospc machinery to not do flushing if

	 * we have a journal_info set, so we need to clear this out and re-set

	 * it in iomap_end.

	 */

	ASSERT(current->journal_info == NULL ||

	       current->journal_info == BTRFS_DIO_SYNC_STUB);

	current->journal_info = NULL;



	if (!write)

		len = min_t(u64, len, fs_info->sectorsize);
@@ -7467,7 +7455,6 @@ static int btrfs_dio_iomap_begin(struct inode *inode, loff_t start, 
	if (!dio_data)

		return -ENOMEM;



	dio_data->sync = sync;

	dio_data->length = length;

	if (write) {

		dio_data->reserve = round_up(length, fs_info->sectorsize);
@@ -7615,14 +7602,6 @@ static int btrfs_dio_iomap_end(struct inode *inode, loff_t pos, loff_t length, 
		extent_changeset_free(dio_data->data_reserved);

	}

out:

	/*

	 * We're all done, we can re-set the current->journal_info now safely

	 * for our endio.

	 */

	if (dio_data->sync) {

		ASSERT(current->journal_info == NULL);

		current->journal_info = BTRFS_DIO_SYNC_STUB;

	}

	kfree(dio_data);

	iomap->private = NULL;
@@ -7999,30 +7978,6 @@ static blk_qc_t btrfs_submit_direct(struct inode *inode, struct iomap *iomap, 
	return BLK_QC_T_NONE;

}



static inline int btrfs_maybe_fsync_end_io(struct kiocb *iocb, ssize_t size,

					   int error, unsigned flags)

{

	/*

	 * Now if we're still in the context of our submitter we know we can't

	 * safely run generic_write_sync(), so clear our flag here so that the

	 * caller knows to follow up with a sync.

	 */

	if (current->journal_info == BTRFS_DIO_SYNC_STUB) {

		current->journal_info = NULL;

		return error;

	}



	if (error)

		return error;



	if (size) {

		iocb->ki_flags |= IOCB_DSYNC;

		return generic_write_sync(iocb, size);

	}



	return 0;

}



const struct iomap_ops btrfs_dio_iomap_ops = {

	.iomap_begin            = btrfs_dio_iomap_begin,

	.iomap_end              = btrfs_dio_iomap_end,
@@ -8032,11 +7987,6 @@ const struct iomap_dio_ops btrfs_dio_ops = { 
	.submit_io		= btrfs_submit_direct,

};



const struct iomap_dio_ops btrfs_sync_dops = {

	.submit_io		= btrfs_submit_direct,

	.end_io			= btrfs_maybe_fsync_end_io,

};



static int btrfs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,

			u64 start, u64 len)

{

fs/btrfs/transaction.h

+0 −1
Original line number Diff line number Diff line
@@ -112,7 +112,6 @@ struct btrfs_transaction { 
#define TRANS_EXTWRITERS	(__TRANS_START | __TRANS_ATTACH)



#define BTRFS_SEND_TRANS_STUB	((void *)1)

#define BTRFS_DIO_SYNC_STUB	((void *)2)



struct btrfs_trans_handle {

	u64 transid;