Merge tag 'iomap-5.12-merge-2' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux

Pull iomap updates from Darrick Wong:
 "The big change in this cycle is some new code to make it possible for
  XFS to try unaligned directio overwrites without taking locks. If the
  block is fully written and within EOF (i.e. doesn't require any
  further fs intervention) then we can let the unlocked write proceed.
  If not, we fall back to synchronizing direct writes.

  Summary:

   - Adjust the final parameter of iomap_dio_rw.

   - Add a new flag to request that iomap directio writes return EAGAIN
     if the write is not a pure overwrite within EOF; this will be used
     to reduce lock contention with unaligned direct writes on XFS.

   - Amend XFS' directio code to eliminate exclusive locking for
     unaligned direct writes if the circumstances permit"

* tag 'iomap-5.12-merge-2' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux:
  xfs: reduce exclusive locking on unaligned dio
  xfs: split the unaligned DIO write code out
  xfs: improve the reflink_bounce_dio_write tracepoint
  xfs: simplify the read/write tracepoints
  xfs: remove the buffered I/O fallback assert
  xfs: cleanup the read/write helper naming
  xfs: make xfs_file_aio_write_checks IOCB_NOWAIT-aware
  xfs: factor out a xfs_ilock_iocb helper
  iomap: add a IOMAP_DIO_OVERWRITE_ONLY flag
  iomap: pass a flags argument to iomap_dio_rw
  iomap: rename the flags variable in __iomap_dio_rw

1 files changed, 16 insertions, 10 deletions

diff --git a/fs/iomap/direct-io.c b/fs/iomap/direct-io.c
index 2273120d8ed7..e4c258a2cefc 100644
--- a/fs/iomap/direct-io.c
+++ b/fs/iomap/direct-io.c
@@ -451,23 +451,22 @@ iomap_dio_actor(struct inode *inode, loff_t pos, loff_t length,
 struct iomap_dio *
 __iomap_dio_rw(struct kiocb *iocb, struct iov_iter *iter,
 		const struct iomap_ops *ops, const struct iomap_dio_ops *dops,
-		bool wait_for_completion)
+		unsigned int dio_flags)
 {
 	struct address_space *mapping = iocb->ki_filp->f_mapping;
 	struct inode *inode = file_inode(iocb->ki_filp);
 	size_t count = iov_iter_count(iter);
 	loff_t pos = iocb->ki_pos;
 	loff_t end = iocb->ki_pos + count - 1, ret = 0;
-	unsigned int flags = IOMAP_DIRECT;
+	bool wait_for_completion =
+		is_sync_kiocb(iocb) || (dio_flags & IOMAP_DIO_FORCE_WAIT);
+	unsigned int iomap_flags = IOMAP_DIRECT;
 	struct blk_plug plug;
 	struct iomap_dio *dio;
 
 	if (!count)
 		return NULL;
 
-	if (WARN_ON(is_sync_kiocb(iocb) && !wait_for_completion))
-		return ERR_PTR(-EIO);
-
 	dio = kmalloc(sizeof(*dio), GFP_KERNEL);
 	if (!dio)
 		return ERR_PTR(-ENOMEM);
@@ -492,7 +491,7 @@ __iomap_dio_rw(struct kiocb *iocb, struct iov_iter *iter,
 		if (iter_is_iovec(iter))
 			dio->flags |= IOMAP_DIO_DIRTY;
 	} else {
-		flags |= IOMAP_WRITE;
+		iomap_flags |= IOMAP_WRITE;
 		dio->flags |= IOMAP_DIO_WRITE;
 
 		/* for data sync or sync, we need sync completion processing */
@@ -514,7 +513,14 @@ __iomap_dio_rw(struct kiocb *iocb, struct iov_iter *iter,
 			ret = -EAGAIN;
 			goto out_free_dio;
 		}
-		flags |= IOMAP_NOWAIT;
+		iomap_flags |= IOMAP_NOWAIT;
+	}
+
+	if (dio_flags & IOMAP_DIO_OVERWRITE_ONLY) {
+		ret = -EAGAIN;
+		if (pos >= dio->i_size || pos + count > dio->i_size)
+			goto out_free_dio;
+		iomap_flags |= IOMAP_OVERWRITE_ONLY;
 	}
 
 	ret = filemap_write_and_wait_range(mapping, pos, end);
@@ -545,7 +551,7 @@ __iomap_dio_rw(struct kiocb *iocb, struct iov_iter *iter,
 
 	blk_start_plug(&plug);
 	do {
-		ret = iomap_apply(inode, pos, count, flags, ops, dio,
+		ret = iomap_apply(inode, pos, count, iomap_flags, ops, dio,
 				iomap_dio_actor);
 		if (ret <= 0) {
 			/* magic error code to fall back to buffered I/O */
@@ -629,11 +635,11 @@ EXPORT_SYMBOL_GPL(__iomap_dio_rw);
 ssize_t
 iomap_dio_rw(struct kiocb *iocb, struct iov_iter *iter,
 		const struct iomap_ops *ops, const struct iomap_dio_ops *dops,
-		bool wait_for_completion)
+		unsigned int dio_flags)
 {
 	struct iomap_dio *dio;
 
-	dio = __iomap_dio_rw(iocb, iter, ops, dops, wait_for_completion);
+	dio = __iomap_dio_rw(iocb, iter, ops, dops, dio_flags);
 	if (IS_ERR_OR_NULL(dio))
 		return PTR_ERR_OR_ZERO(dio);
 	return iomap_dio_complete(dio);