btrfs: add support for uncached writes

Commit Message

Jens Axboe Feb. 3, 2025, 4:35 p.m. UTC

The read side is already covered as btrfs uses the generic filemap
helpers. For writes, just pass in FGP_DONTCACHE if uncached IO is being
done, then the folios created should be marked appropriately.

For IO completion, ensure that writing back folios that are uncached
gets punted to one of the btrfs workers, as task context is needed for
that. Add an 'uncached_io' member to struct btrfs_bio to manage that.

With that, add FOP_DONTCACHE to the btrfs file_operations fop_flags
structure, enabling use of RWF_DONTCACHE.

Signed-off-by: Jens Axboe <[e-mail]>

---

Patch

diff --git a/fs/btrfs/bio.c b/fs/btrfs/bio.c
index bc2555c44a12..fd51dbc16176 100644
--- a/fs/btrfs/bio.c
+++ b/fs/btrfs/bio.c
@@ -337,7 +337,7 @@  static void btrfs_end_bio_work(struct work_struct *work)
 	struct btrfs_bio *bbio = container_of(work, struct btrfs_bio, end_io_work);
 
 	/* Metadata reads are checked and repaired by the submitter. */
-	if (is_data_bbio(bbio))
+	if (bio_op(&bbio->bio) == REQ_OP_READ && is_data_bbio(bbio))
 		btrfs_check_read_bio(bbio, bbio->bio.bi_private);
 	else
 		btrfs_bio_end_io(bbio, bbio->bio.bi_status);
@@ -354,7 +354,7 @@  static void btrfs_simple_end_io(struct bio *bio)
 	if (bio->bi_status)
 		btrfs_log_dev_io_error(bio, dev);
 
-	if (bio_op(bio) == REQ_OP_READ) {
+	if (bio_op(bio) == REQ_OP_READ || bbio->dropbehind_io) {
 		INIT_WORK(&bbio->end_io_work, btrfs_end_bio_work);
 		queue_work(btrfs_end_io_wq(fs_info, bio), &bbio->end_io_work);
 	} else {
diff --git a/fs/btrfs/bio.h b/fs/btrfs/bio.h
index e2fe16074ad6..b41c5236c93d 100644
--- a/fs/btrfs/bio.h
+++ b/fs/btrfs/bio.h
@@ -82,6 +82,8 @@  struct btrfs_bio {
 	/* Save the first error status of split bio. */
 	blk_status_t status;
 
+	bool dropbehind_io;
+
 	/*
 	 * This member must come last, bio_alloc_bioset will allocate enough
 	 * bytes for entire btrfs_bio but relies on bio being last.
diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c
index a2cac9d0a1a9..cab29a1c5867 100644
--- a/fs/btrfs/extent_io.c
+++ b/fs/btrfs/extent_io.c
@@ -758,8 +758,11 @@  static void submit_extent_folio(struct btrfs_bio_ctrl *bio_ctrl,
 	ASSERT(bio_ctrl->end_io_func);
 
 	if (bio_ctrl->bbio &&
-	    !btrfs_bio_is_contig(bio_ctrl, folio, disk_bytenr, pg_offset))
+	    !btrfs_bio_is_contig(bio_ctrl, folio, disk_bytenr, pg_offset)) {
+		if (folio_test_dropbehind(folio))
+			bio_ctrl->bbio->dropbehind_io = true;
 		submit_one_bio(bio_ctrl);
+	}
 
 	do {
 		u32 len = size;
@@ -777,6 +780,9 @@  static void submit_extent_folio(struct btrfs_bio_ctrl *bio_ctrl,
 			len = bio_ctrl->len_to_oe_boundary;
 		}
 
+		if (folio_test_dropbehind(folio))
+			bio_ctrl->bbio->dropbehind_io = true;
+
 		if (!bio_add_folio(&bio_ctrl->bbio->bio, folio, len, pg_offset)) {
 			/* bio full: move on to a new one */
 			submit_one_bio(bio_ctrl);
diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c
index 36f51c311bb1..70f2bb4b316e 100644
--- a/fs/btrfs/file.c
+++ b/fs/btrfs/file.c
@@ -857,18 +857,17 @@  static gfp_t get_prepare_gfp_flags(struct inode *inode, bool nowait)
  */
 static noinline int prepare_one_folio(struct inode *inode, struct folio **folio_ret,
 				      loff_t pos, size_t write_bytes,
-				      bool force_uptodate, bool nowait)
+				      bool force_uptodate, fgf_t fgp_flags)
 {
 	unsigned long index = pos >> PAGE_SHIFT;
-	gfp_t mask = get_prepare_gfp_flags(inode, nowait);
-	fgf_t fgp_flags = (nowait ? FGP_WRITEBEGIN | FGP_NOWAIT : FGP_WRITEBEGIN);
+	gfp_t mask = get_prepare_gfp_flags(inode, fgp_flags & FGP_NOWAIT);
 	struct folio *folio;
 	int ret = 0;
 
 again:
 	folio = __filemap_get_folio(inode->i_mapping, index, fgp_flags, mask);
 	if (IS_ERR(folio)) {
-		if (nowait)
+		if (fgp_flags & FGP_NOWAIT)
 			ret = -EAGAIN;
 		else
 			ret = PTR_ERR(folio);
@@ -887,7 +886,7 @@  static noinline int prepare_one_folio(struct inode *inode, struct folio **folio_
 	if (ret) {
 		/* The folio is already unlocked. */
 		folio_put(folio);
-		if (!nowait && ret == -EAGAIN) {
+		if (!(fgp_flags & FGP_NOWAIT) && ret == -EAGAIN) {
 			ret = 0;
 			goto again;
 		}
@@ -1097,9 +1096,15 @@  ssize_t btrfs_buffered_write(struct kiocb *iocb, struct iov_iter *i)
 	const bool nowait = (iocb->ki_flags & IOCB_NOWAIT);
 	unsigned int bdp_flags = (nowait ? BDP_ASYNC : 0);
 	bool only_release_metadata = false;
+	fgf_t fgp_flags = FGP_WRITEBEGIN;
 
-	if (nowait)
+	if (nowait) {
 		ilock_flags |= BTRFS_ILOCK_TRY;
+		fgp_flags |= FGP_NOWAIT;
+	}
+
+	if (iocb->ki_flags & IOCB_DONTCACHE)
+		fgp_flags |= FGP_DONTCACHE;
 
 	ret = btrfs_inode_lock(BTRFS_I(inode), ilock_flags);
 	if (ret < 0)
@@ -1195,7 +1200,7 @@  ssize_t btrfs_buffered_write(struct kiocb *iocb, struct iov_iter *i)
 		}
 
 		ret = prepare_one_folio(inode, &folio, pos, write_bytes,
-					force_page_uptodate, false);
+					force_page_uptodate, fgp_flags);
 		if (ret) {
 			btrfs_delalloc_release_extents(BTRFS_I(inode),
 						       reserve_bytes);
@@ -3671,7 +3676,7 @@  const struct file_operations btrfs_file_operations = {
 #endif
 	.remap_file_range = btrfs_remap_file_range,
 	.uring_cmd	= btrfs_uring_cmd,
-	.fop_flags	= FOP_BUFFER_RASYNC | FOP_BUFFER_WASYNC,
+	.fop_flags	= FOP_BUFFER_RASYNC | FOP_BUFFER_WASYNC | FOP_DONTCACHE,
 };
 
 int btrfs_fdatawrite_range(struct btrfs_inode *inode, loff_t start, loff_t end)