[PATCH 11/12] iomap: add direct-io partial sector read support

[Keith Busch]

From: Keith Busch <kbusch at kernel.org>

Enable direct io to read partial sectors if the block device supports bit
buckets.

Signed-off-by: Keith Busch <kbusch at kernel.org>
---
 fs/iomap/direct-io.c | 36 ++++++++++++++++++++++++++++--------
 1 file changed, 28 insertions(+), 8 deletions(-)

diff --git a/fs/iomap/direct-io.c b/fs/iomap/direct-io.c
index 10a113358365..212e63b78950 100644
--- a/fs/iomap/direct-io.c
+++ b/fs/iomap/direct-io.c
@@ -251,9 +251,12 @@ static loff_t iomap_dio_bio_iter(const struct iomap_iter *iter,
 	int nr_pages, ret = 0;
 	size_t copied = 0;
 	size_t orig_count;
+	u16 skip = 0, trunc = 0;
 
 	if (blkdev_dio_unaligned(bdev, pos | length, dio->submit.iter))
-		return -EINVAL;
+		if (!blkdev_bit_bucket(bdev, pos, length, dio->submit.iter,
+				       &skip, &trunc))
+			return -EINVAL;
 
 	if (iomap->type == IOMAP_UNWRITTEN) {
 		dio->flags |= IOMAP_DIO_UNWRITTEN;
@@ -310,9 +313,10 @@ static loff_t iomap_dio_bio_iter(const struct iomap_iter *iter,
 	 */
 	bio_opf = iomap_dio_bio_opflags(dio, iomap, use_fua);
 
-	nr_pages = bio_iov_vecs_to_alloc(dio->submit.iter, BIO_MAX_VECS);
+	nr_pages = bio_iov_vecs_to_alloc_partial(dio->submit.iter, BIO_MAX_VECS,
+						 skip, trunc);
 	do {
-		size_t n;
+		size_t n, bucket_bytes = 0;
 		if (dio->error) {
 			iov_iter_revert(dio->submit.iter, copied);
 			copied = ret = 0;
@@ -327,6 +331,15 @@ static loff_t iomap_dio_bio_iter(const struct iomap_iter *iter,
 		bio->bi_private = dio;
 		bio->bi_end_io = iomap_dio_bio_end_io;
 
+		if (skip || trunc) {
+			bio_set_flag(bio, BIO_BIT_BUCKET);
+			if (skip) {
+				bucket_bytes += skip;
+				blk_add_bb_page(bio, skip);
+				skip = 0;
+			}
+		}
+
 		ret = bio_iov_iter_get_pages(bio, dio->submit.iter);
 		if (unlikely(ret)) {
 			/*
@@ -339,6 +352,12 @@ static loff_t iomap_dio_bio_iter(const struct iomap_iter *iter,
 			goto zero_tail;
 		}
 
+		if (trunc && !iov_iter_count(dio->submit.iter)) {
+			blk_add_bb_page(bio, trunc);
+			bucket_bytes += trunc;
+			trunc = 0;
+		}
+
 		n = bio->bi_iter.bi_size;
 		if (dio->flags & IOMAP_DIO_WRITE) {
 			task_io_account_write(n);
@@ -347,18 +366,19 @@ static loff_t iomap_dio_bio_iter(const struct iomap_iter *iter,
 				bio_set_pages_dirty(bio);
 		}
 
-		dio->size += n;
-		copied += n;
+		dio->size += n - bucket_bytes;
+		copied += n - bucket_bytes;
 
-		nr_pages = bio_iov_vecs_to_alloc(dio->submit.iter,
-						 BIO_MAX_VECS);
+		nr_pages = bio_iov_vecs_to_alloc_partial(dio->submit.iter,
+							 BIO_MAX_VECS, skip,
+							 trunc);
 		/*
 		 * We can only poll for single bio I/Os.
 		 */
 		if (nr_pages)
 			dio->iocb->ki_flags &= ~IOCB_HIPRI;
 		iomap_dio_submit_bio(iter, dio, bio, pos);
-		pos += n;
+		pos += n - bucket_bytes;
 	} while (nr_pages);
 
 	/*
-- 
2.30.2