fs: iomap: Atomic write support

Support direct I/O atomic writes by producing a single bio with REQ_ATOMIC
flag set.

Initially FSes (XFS) should only support writing a single FS block
atomically.

As with any atomic write, we should produce a single bio which covers the
complete write length.

Reviewed-by: Christoph Hellwig <[email protected]>
Reviewed-by: "Darrick J. Wong" <[email protected]>
Signed-off-by: John Garry <[email protected]>
Reviewed-by: Ritesh Harjani (IBM) <[email protected]>
[djwong: clarify a couple of things in the docs]
Signed-off-by: Darrick J. Wong <[email protected]>

Author: johnpgarry

Documentation/filesystems/iomap/operations.rst

@@ -513,6 +513,21 @@ IOMAP_WRITE`` with any combination of the following enhancements:
    if the mapping is unwritten and the filesystem cannot handle zeroing
    the unaligned regions without exposing stale contents.
 
+ * ``IOMAP_ATOMIC``: This write is being issued with torn-write
+   protection.
+   Only a single bio can be created for the write, and the write must
+   not be split into multiple I/O requests, i.e. flag REQ_ATOMIC must be
+   set.
+   The file range to write must be aligned to satisfy the requirements
+   of both the filesystem and the underlying block device's atomic
+   commit capabilities.
+   If filesystem metadata updates are required (e.g. unwritten extent
+   conversion or copy on write), all updates for the entire file range
+   must be committed atomically as well.
+   Only one space mapping is allowed per untorn write.
+   Untorn writes must be aligned to, and must not be longer than, a
+   single file block.
+
 Callers commonly hold ``i_rwsem`` in shared or exclusive mode before
 calling this function.
 

fs/iomap/direct-io.c

@@ -271,7 +271,7 @@ static int iomap_dio_zero(const struct iomap_iter *iter, struct iomap_dio *dio,
  * clearing the WRITE_THROUGH flag in the dio request.
  */
 static inline blk_opf_t iomap_dio_bio_opflags(struct iomap_dio *dio,
-		const struct iomap *iomap, bool use_fua)
+		const struct iomap *iomap, bool use_fua, bool atomic)
 {
 	blk_opf_t opflags = REQ_SYNC | REQ_IDLE;
 
@@ -283,6 +283,8 @@ static inline blk_opf_t iomap_dio_bio_opflags(struct iomap_dio *dio,
 		opflags |= REQ_FUA;
 	else
 		dio->flags &= ~IOMAP_DIO_WRITE_THROUGH;
+	if (atomic)
+		opflags |= REQ_ATOMIC;
 
 	return opflags;
 }
@@ -293,7 +295,8 @@ static loff_t iomap_dio_bio_iter(const struct iomap_iter *iter,
 	const struct iomap *iomap = &iter->iomap;
 	struct inode *inode = iter->inode;
 	unsigned int fs_block_size = i_blocksize(inode), pad;
-	loff_t length = iomap_length(iter);
+	const loff_t length = iomap_length(iter);
+	bool atomic = iter->flags & IOMAP_ATOMIC;
 	loff_t pos = iter->pos;
 	blk_opf_t bio_opf;
 	struct bio *bio;
@@ -303,6 +306,9 @@ static loff_t iomap_dio_bio_iter(const struct iomap_iter *iter,
 	size_t copied = 0;
 	size_t orig_count;
 
+	if (atomic && length != fs_block_size)
+		return -EINVAL;
+
 	if ((pos | length) & (bdev_logical_block_size(iomap->bdev) - 1) ||
 	    !bdev_iter_is_aligned(iomap->bdev, dio->submit.iter))
 		return -EINVAL;
@@ -382,7 +388,7 @@ static loff_t iomap_dio_bio_iter(const struct iomap_iter *iter,
 	 * can set up the page vector appropriately for a ZONE_APPEND
 	 * operation.
 	 */
-	bio_opf = iomap_dio_bio_opflags(dio, iomap, use_fua);
+	bio_opf = iomap_dio_bio_opflags(dio, iomap, use_fua, atomic);
 
 	nr_pages = bio_iov_vecs_to_alloc(dio->submit.iter, BIO_MAX_VECS);
 	do {
@@ -415,6 +421,17 @@ static loff_t iomap_dio_bio_iter(const struct iomap_iter *iter,
 		}
 
 		n = bio->bi_iter.bi_size;
+		if (WARN_ON_ONCE(atomic && n != length)) {
+			/*
+			 * This bio should have covered the complete length,
+			 * which it doesn't, so error. We may need to zero out
+			 * the tail (complete FS block), similar to when
+			 * bio_iov_iter_get_pages() returns an error, above.
+			 */
+			ret = -EINVAL;
+			bio_put(bio);
+			goto zero_tail;
+		}
 		if (dio->flags & IOMAP_DIO_WRITE) {
 			task_io_account_write(n);
 		} else {
@@ -598,6 +615,9 @@ __iomap_dio_rw(struct kiocb *iocb, struct iov_iter *iter,
 	if (iocb->ki_flags & IOCB_NOWAIT)
 		iomi.flags |= IOMAP_NOWAIT;
 
+	if (iocb->ki_flags & IOCB_ATOMIC)
+		iomi.flags |= IOMAP_ATOMIC;
+
 	if (iov_iter_rw(iter) == READ) {
 		/* reads can always complete inline */
 		dio->flags |= IOMAP_DIO_INLINE_COMP;
@@ -659,7 +679,17 @@ __iomap_dio_rw(struct kiocb *iocb, struct iov_iter *iter,
 			if (ret != -EAGAIN) {
 				trace_iomap_dio_invalidate_fail(inode, iomi.pos,
 								iomi.len);
-				ret = -ENOTBLK;
+				if (iocb->ki_flags & IOCB_ATOMIC) {
+					/*
+					 * folio invalidation failed, maybe
+					 * this is transient, unlock and see if
+					 * the caller tries again.
+					 */
+					ret = -EAGAIN;
+				} else {
+					/* fall back to buffered write */
+					ret = -ENOTBLK;
+				}
 			}
 			goto out_free_dio;
 		}

fs/iomap/trace.h

@@ -98,7 +98,8 @@ DEFINE_RANGE_EVENT(iomap_dio_rw_queued);
 	{ IOMAP_REPORT,		"REPORT" }, \
 	{ IOMAP_FAULT,		"FAULT" }, \
 	{ IOMAP_DIRECT,		"DIRECT" }, \
-	{ IOMAP_NOWAIT,		"NOWAIT" }
+	{ IOMAP_NOWAIT,		"NOWAIT" }, \
+	{ IOMAP_ATOMIC,		"ATOMIC" }
 
 #define IOMAP_F_FLAGS_STRINGS \
 	{ IOMAP_F_NEW,		"NEW" }, \

include/linux/iomap.h

@@ -178,6 +178,7 @@ struct iomap_folio_ops {
 #else
 #define IOMAP_DAX		0
 #endif /* CONFIG_FS_DAX */
+#define IOMAP_ATOMIC		(1 << 9)
 
 struct iomap_ops {
 	/*