Skip to contentSkip to conversation

direct io dma alignment

PATCH Superseded Keith Busch 6 participants 29 messages May 26, 2022 – May 31, 2022

Subsystems (5)

v4 of 6 Superseded Latest is v6 →

lore
Share & tools
lore
#0 lore raw

v4 cover direct io dma alignment

From: Keith Busch <[email protected]>

This is mostly the same as v3, but with more code comments, prep patches, replace '9' with 'SECTOR_SHIFT', and added reviews.

The biggest thing is I brought fs support back in the last patch.

For testing, I used 'fio', which has an 'iomem_align' parameter that can force arbitrary memory offsets, even with direct-io. I created two data integrity verifying profiles: one for raw block, the other for filesystems, and ran each of them on two nvme namespaces. One namespace was formatted 512b, the other was 4k.

The profile has different jobs: one smaller transfers, and one larger. This to exercise both the simple and normal direct io cases, as well as bio_split() conditions.

For filesystems testing, I used xfs, ext4 and btrfs to represent iomap, and just ext2 for the older direct io. Note, btrfs falls back to buffered anyway (see check_direct_IO()), so btrfs was essentially not testing the direct io path.

Here is an example of one of the profiles for the raw block test:

[global]
filename=/dev/nvme0n1
ioengine=io_uring
verify=crc32c
rw=randwrite
iodepth=64
direct=1

[small]
stonewall
bsrange=4k-64k
iomem_align=4

[large]
stonewall
bsrange=512k-4M
iomem_align=100

Keith Busch (9):
block: fix infiniate loop for invalid zone append block/bio: remove duplicate append pages code
block: export dma_alignment attribute
block: introduce bdev_dma_alignment helper
block: add a helper function for dio alignment block/merge: count bytes instead of sectors block/bounce: count bytes instead of sectors
block: relax direct io memory alignment
fs: add support for dma aligned direct-io

 Documentation/ABI/stable/sysfs-block |   9 +++
 block/bio.c                          | 117 +++++++++++++--------------
 block/blk-merge.c                    |  41 ++++++----
 block/blk-sysfs.c                    |   7 ++
 block/bounce.c                       |  12 ++-
 block/fops.c                         |  40 ++++++---
 fs/direct-io.c                       |  11 ++-
 fs/iomap/direct-io.c                 |   3 +-
 include/linux/blkdev.h               |   5 ++
 9 files changed, 146 insertions(+), 99 deletions(-)
Signature · 1 line
2.30.2
#1 lore raw

PATCH v4 1/9 block: fix infiniate loop for invalid zone append

From: Keith Busch <[email protected]>

Returning 0 early from __bio_iov_append_get_pages() for the max_append_sectors warning just creates an infinite loop since 0 means success, and the bio will never fill from the unadvancing iov_iter. We could turn the return into an error value, but it will already be turned into an error value later on, so just remove the warning. Clearly no one ever hit it anyway.

Fixes 0512a75b98f84 ("block: Introduce REQ_OP_ZONE_APPEND")
---
 block/bio.c | 3 ---
 1 file changed, 3 deletions(-)
block/bio.c +0 -3
index a3893d80dccc..e249f6414fd5 100644
@@ -1228,9 +1228,6 @@ static int __bio_iov_append_get_pages(struct bio *bio, struct iov_iter *iter)
 	size_t offset;
 	int ret = 0;
 
-	if (WARN_ON_ONCE(!max_append_sectors))
-		return 0;
-
 	/*
 	 * Move page array up in the allocated memory for the bio vecs as far as
 	 * possible so that we can start filling biovecs from the beginning
Signature · 1 line
2.30.2
Reviewed-by on v4 1/9
#2 lore raw

Re: [PATCHv4 1/9] block: fix infiniate loop for invalid zone append

Quoted 2 lines · Keith Busch

From: Keith Busch <[email protected]>

s/infiniate/infinite in the patch title.

Quoted 18 lines

Returning 0 early from __bio_iov_append_get_pages() for the max_append_sectors warning just creates an infinite loop since 0 means success, and the bio will never fill from the unadvancing iov_iter. We could turn the return into an error value, but it will already be turned into an error value later on, so just remove the warning. Clearly no one ever hit it anyway.

Fixes 0512a75b98f84 ("block: Introduce REQ_OP_ZONE_APPEND")
---
 block/bio.c | 3 ---
 1 file changed, 3 deletions(-)
block/bio.c +0 -0
index a3893d80dccc..e249f6414fd5 100644

block/bio.c static int __bio_iov_append_get_pages(struct bio *bio, struct iov_iter *iter) · line 1233

Quoted 9 lines
block/bio.c +0 -3
@@ -1228,9 +1228,6 @@ static int __bio_iov_append_get_pages(struct bio *bio, struct iov_iter *iter)
 	size_t offset;
 	int ret = 0;
 
-	if (WARN_ON_ONCE(!max_append_sectors))
-		return 0;
-
 	/*
 	 * Move page array up in the allocated memory for the bio vecs as far as
 	 * possible so that we can start filling biovecs from the beginning

Otherwise looks good.

Signature · 2 lines
Damien Le Moal
Western Digital Research
Reviewed-by on v4 1/9
#3 lore raw

Re: [PATCHv4 1/9] block: fix infiniate loop for invalid zone append

In reply to Keith Busch
Quoted 8 lines · Keith Busch

From: Keith Busch <[email protected]>

Returning 0 early from __bio_iov_append_get_pages() for the max_append_sectors warning just creates an infinite loop since 0 means success, and the bio will never fill from the unadvancing iov_iter. We could turn the return into an error value, but it will already be turned into an error value later on, so just remove the warning. Clearly no one ever hit it anyway.

Looks good:

#4 lore raw

PATCH v4 2/9 block/bio: remove duplicate append pages code

In reply to Keith Busch

From: Keith Busch <[email protected]>

The getting pages setup for zone append and normal IO are identical. Use common code for each.

---
 block/bio.c | 102 ++++++++++++++++++++++------------------------------
 1 file changed, 42 insertions(+), 60 deletions(-)
block/bio.c +42 -60
index e249f6414fd5..55d2a9c4e312 100644
@@ -1158,6 +1158,37 @@ static void bio_put_pages(struct page **pages, size_t size, size_t off)
 		put_page(pages[i]);
 }
 
+static int bio_iov_add_page(struct bio *bio, struct page *page,
+		unsigned int len, unsigned int offset)
+{
+	bool same_page = false;
+
+	if (!__bio_try_merge_page(bio, page, len, offset, &same_page)) {
+		if (WARN_ON_ONCE(bio_full(bio, len)))
+			return -EINVAL;
+		__bio_add_page(bio, page, len, offset);
+		return 0;
+	}
+
+	if (same_page)
+		put_page(page);
+	return 0;
+}
+
+static int bio_iov_add_zone_append_page(struct bio *bio, struct page *page,
+		unsigned int len, unsigned int offset)
+{
+	struct request_queue *q = bdev_get_queue(bio->bi_bdev);
+	bool same_page = false;
+
+	if (bio_add_hw_page(q, bio, page, len, offset,
+			queue_max_zone_append_sectors(q), &same_page) != len)
+		return -EINVAL;
+	if (same_page)
+		put_page(page);
+	return 0;
+}
+
 #define PAGE_PTRS_PER_BVEC     (sizeof(struct bio_vec) / sizeof(struct page *))
 
 /**
@@ -1176,7 +1207,6 @@ static int __bio_iov_iter_get_pages(struct bio *bio, struct iov_iter *iter)
 	unsigned short entries_left = bio->bi_max_vecs - bio->bi_vcnt;
 	struct bio_vec *bv = bio->bi_io_vec + bio->bi_vcnt;
 	struct page **pages = (struct page **)bv;
-	bool same_page = false;
 	ssize_t size, left;
 	unsigned len, i;
 	size_t offset;
@@ -1185,7 +1215,7 @@ static int __bio_iov_iter_get_pages(struct bio *bio, struct iov_iter *iter)
 	 * Move page array up in the allocated memory for the bio vecs as far as
 	 * possible so that we can start filling biovecs from the beginning
 	 * without overwriting the temporary page array.
-	*/
+	 */
 	BUILD_BUG_ON(PAGE_PTRS_PER_BVEC < 2);
 	pages += entries_left * (PAGE_PTRS_PER_BVEC - 1);
 
@@ -1195,18 +1225,18 @@ static int __bio_iov_iter_get_pages(struct bio *bio, struct iov_iter *iter)
 
 	for (left = size, i = 0; left > 0; left -= len, i++) {
 		struct page *page = pages[i];
+		int ret;
 
 		len = min_t(size_t, PAGE_SIZE - offset, left);
+		if (bio_op(bio) == REQ_OP_ZONE_APPEND)
+			ret = bio_iov_add_zone_append_page(bio, page, len,
+					offset);
+		else
+			ret = bio_iov_add_page(bio, page, len, offset);
 
-		if (__bio_try_merge_page(bio, page, len, offset, &same_page)) {
-			if (same_page)
-				put_page(page);
-		} else {
-			if (WARN_ON_ONCE(bio_full(bio, len))) {
-				bio_put_pages(pages + i, left, offset);
-				return -EINVAL;
-			}
-			__bio_add_page(bio, page, len, offset);
+		if (ret) {
+			bio_put_pages(pages + i, left, offset);
+			return ret;
 		}
 		offset = 0;
 	}
@@ -1215,51 +1245,6 @@ static int __bio_iov_iter_get_pages(struct bio *bio, struct iov_iter *iter)
 	return 0;
 }
 
-static int __bio_iov_append_get_pages(struct bio *bio, struct iov_iter *iter)
-{
-	unsigned short nr_pages = bio->bi_max_vecs - bio->bi_vcnt;
-	unsigned short entries_left = bio->bi_max_vecs - bio->bi_vcnt;
-	struct request_queue *q = bdev_get_queue(bio->bi_bdev);
-	unsigned int max_append_sectors = queue_max_zone_append_sectors(q);
-	struct bio_vec *bv = bio->bi_io_vec + bio->bi_vcnt;
-	struct page **pages = (struct page **)bv;
-	ssize_t size, left;
-	unsigned len, i;
-	size_t offset;
-	int ret = 0;
-
-	/*
-	 * Move page array up in the allocated memory for the bio vecs as far as
-	 * possible so that we can start filling biovecs from the beginning
-	 * without overwriting the temporary page array.
-	 */
-	BUILD_BUG_ON(PAGE_PTRS_PER_BVEC < 2);
-	pages += entries_left * (PAGE_PTRS_PER_BVEC - 1);
-
-	size = iov_iter_get_pages(iter, pages, LONG_MAX, nr_pages, &offset);
-	if (unlikely(size <= 0))
-		return size ? size : -EFAULT;
-
-	for (left = size, i = 0; left > 0; left -= len, i++) {
-		struct page *page = pages[i];
-		bool same_page = false;
-
-		len = min_t(size_t, PAGE_SIZE - offset, left);
-		if (bio_add_hw_page(q, bio, page, len, offset,
-				max_append_sectors, &same_page) != len) {
-			bio_put_pages(pages + i, left, offset);
-			ret = -EINVAL;
-			break;
-		}
-		if (same_page)
-			put_page(page);
-		offset = 0;
-	}
-
-	iov_iter_advance(iter, size - left);
-	return ret;
-}
-
 /**
  * bio_iov_iter_get_pages - add user or kernel pages to a bio
  * @bio: bio to add pages to
@@ -1294,10 +1279,7 @@ int bio_iov_iter_get_pages(struct bio *bio, struct iov_iter *iter)
 	}
 
 	do {
-		if (bio_op(bio) == REQ_OP_ZONE_APPEND)
-			ret = __bio_iov_append_get_pages(bio, iter);
-		else
-			ret = __bio_iov_iter_get_pages(bio, iter);
+		ret = __bio_iov_iter_get_pages(bio, iter);
 	} while (!ret && iov_iter_count(iter) && !bio_full(bio, 0));
 
 	/* don't account direct I/O as memory stall */
Signature · 1 line
2.30.2
#5 lore raw

PATCH v4 3/9 block: export dma_alignment attribute

In reply to Keith Busch

From: Keith Busch <[email protected]>

User space may want to know how to align their buffers to avoid bouncing. Export the queue attribute.

---
 Documentation/ABI/stable/sysfs-block | 9 +++++++++
 block/blk-sysfs.c                    | 7 +++++++
 2 files changed, 16 insertions(+)
Documentation/ABI/stable/sysfs-block +9 -0
index e8797cd09aff..08f6d00df138 100644
@@ -142,6 +142,15 @@ Description:
 		Default value of this file is '1'(on).
 
 
+What:		/sys/block/<disk>/queue/dma_alignment
+Date:		May 2022
+Contact:	[email protected]
+Description:
+		Reports the alignment that user space addresses must have to be
+		used for raw block device access with O_DIRECT and other driver
+		specific passthrough mechanisms.
+
+
 What:		/sys/block/<disk>/queue/chunk_sectors
 Date:		September 2016
 Contact:	Hannes Reinecke <[email protected]>
block/blk-sysfs.c +7 -0
index 88bd41d4cb59..14607565d781 100644
@@ -274,6 +274,11 @@ static ssize_t queue_virt_boundary_mask_show(struct request_queue *q, char *page
 	return queue_var_show(q->limits.virt_boundary_mask, page);
 }
 
+static ssize_t queue_dma_alignment_show(struct request_queue *q, char *page)
+{
+	return queue_var_show(queue_dma_alignment(q), page);
+}
+
 #define QUEUE_SYSFS_BIT_FNS(name, flag, neg)				\
 static ssize_t								\
 queue_##name##_show(struct request_queue *q, char *page)		\
@@ -606,6 +611,7 @@ QUEUE_RO_ENTRY(queue_dax, "dax");
 QUEUE_RW_ENTRY(queue_io_timeout, "io_timeout");
 QUEUE_RW_ENTRY(queue_wb_lat, "wbt_lat_usec");
 QUEUE_RO_ENTRY(queue_virt_boundary_mask, "virt_boundary_mask");
+QUEUE_RO_ENTRY(queue_dma_alignment, "dma_alignment");
 
 #ifdef CONFIG_BLK_DEV_THROTTLING_LOW
 QUEUE_RW_ENTRY(blk_throtl_sample_time, "throttle_sample_time");
@@ -667,6 +673,7 @@ static struct attribute *queue_attrs[] = {
 	&blk_throtl_sample_time_entry.attr,
 #endif
 	&queue_virt_boundary_mask_entry.attr,
+	&queue_dma_alignment_entry.attr,
 	NULL,
 };
 
Signature · 1 line
2.30.2
#6 lore raw

Re: [PATCHv4 3/9] block: export dma_alignment attribute

Quoted 19 lines · Keith Busch

From: Keith Busch <[email protected]>

User space may want to know how to align their buffers to avoid bouncing. Export the queue attribute.

---
 Documentation/ABI/stable/sysfs-block | 9 +++++++++
 block/blk-sysfs.c                    | 7 +++++++
 2 files changed, 16 insertions(+)
Documentation/ABI/stable/sysfs-block +0 -0
index e8797cd09aff..08f6d00df138 100644
@@ -142,6 +142,15 @@ Description:
 		Default value of this file is '1'(on).

Documentation/ABI/stable/sysfs-block Description: · line 152

Quoted 10 lines
Documentation/ABI/stable/sysfs-block +8 -0
 
 
+What:		/sys/block/<disk>/queue/dma_alignment
+Date:		May 2022
+Contact:	[email protected]
+Description:
+		Reports the alignment that user space addresses must have to be
+		used for raw block device access with O_DIRECT and other driver
+		specific passthrough mechanisms.
+

Please keep this file in alphabetical order.

- Eric

#7 lore raw

PATCH v4 4/9 block: introduce bdev_dma_alignment helper

In reply to Keith Busch

From: Keith Busch <[email protected]>

Preparing for upcoming dma_alignment users that have a block_device, but don't need the request_queue.

---
 include/linux/blkdev.h | 5 +++++
 1 file changed, 5 insertions(+)
include/linux/blkdev.h +5 -0
index 5bdf2ac9142c..834b981ef01b 100644
@@ -1365,6 +1365,11 @@ static inline int queue_dma_alignment(const struct request_queue *q)
 	return q ? q->dma_alignment : 511;
 }
 
+static inline unsigned int bdev_dma_alignment(struct block_device *bdev)
+{
+	return queue_dma_alignment(bdev_get_queue(bdev));
+}
+
 static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr,
 				 unsigned int len)
 {
Signature · 1 line
2.30.2
#8 lore raw

PATCH v4 5/9 block: add a helper function for dio alignment

In reply to Keith Busch

From: Keith Busch <[email protected]>

This will make it easier to add more complex acceptable alignment criteria in the future.

---
 block/fops.c | 32 +++++++++++++++++++++-----------
 1 file changed, 21 insertions(+), 11 deletions(-)
block/fops.c +21 -11
index b9b83030e0df..bd6c2e13a4e3 100644
@@ -42,6 +42,16 @@ static unsigned int dio_bio_write_op(struct kiocb *iocb)
 	return op;
 }
 
+static int blkdev_dio_aligned(struct block_device *bdev, loff_t pos,
+			      struct iov_iter *iter)
+{
+	if ((pos | iov_iter_alignment(iter)) &
+	    (bdev_logical_block_size(bdev) - 1))
+		return -EINVAL;
+
+	return 0;
+}
+
 #define DIO_INLINE_BIO_VECS 4
 
 static ssize_t __blkdev_direct_IO_simple(struct kiocb *iocb,
@@ -54,9 +64,9 @@ static ssize_t __blkdev_direct_IO_simple(struct kiocb *iocb,
 	struct bio bio;
 	ssize_t ret;
 
-	if ((pos | iov_iter_alignment(iter)) &
-	    (bdev_logical_block_size(bdev) - 1))
-		return -EINVAL;
+	ret = blkdev_dio_aligned(bdev, pos, iter);
+	if (ret)
+		return ret;
 
 	if (nr_pages <= DIO_INLINE_BIO_VECS)
 		vecs = inline_vecs;
@@ -171,11 +181,11 @@ static ssize_t __blkdev_direct_IO(struct kiocb *iocb, struct iov_iter *iter,
 	bool is_read = (iov_iter_rw(iter) == READ), is_sync;
 	unsigned int opf = is_read ? REQ_OP_READ : dio_bio_write_op(iocb);
 	loff_t pos = iocb->ki_pos;
-	int ret = 0;
+	int ret;
 
-	if ((pos | iov_iter_alignment(iter)) &
-	    (bdev_logical_block_size(bdev) - 1))
-		return -EINVAL;
+	ret = blkdev_dio_aligned(bdev, pos, iter);
+	if (ret)
+		return ret;
 
 	if (iocb->ki_flags & IOCB_ALLOC_CACHE)
 		opf |= REQ_ALLOC_CACHE;
@@ -296,11 +306,11 @@ static ssize_t __blkdev_direct_IO_async(struct kiocb *iocb,
 	struct blkdev_dio *dio;
 	struct bio *bio;
 	loff_t pos = iocb->ki_pos;
-	int ret = 0;
+	int ret;
 
-	if ((pos | iov_iter_alignment(iter)) &
-	    (bdev_logical_block_size(bdev) - 1))
-		return -EINVAL;
+	ret = blkdev_dio_aligned(bdev, pos, iter);
+	if (ret)
+		return ret;
 
 	if (iocb->ki_flags & IOCB_ALLOC_CACHE)
 		opf |= REQ_ALLOC_CACHE;
Signature · 1 line
2.30.2
#9 lore raw

Re: [PATCHv4 5/9] block: add a helper function for dio alignment

Quoted 17 lines · Keith Busch

From: Keith Busch <[email protected]>

This will make it easier to add more complex acceptable alignment criteria in the future.

---
 block/fops.c | 32 +++++++++++++++++++++-----------
 1 file changed, 21 insertions(+), 11 deletions(-)
block/fops.c +0 -0
index b9b83030e0df..bd6c2e13a4e3 100644
@@ -42,6 +42,16 @@ static unsigned int dio_bio_write_op(struct kiocb *iocb)
 	return op;
 }

block/fops.c static unsigned int dio_bio_write_op(struct kiocb *iocb) · line 53

Quoted 10 lines
block/fops.c +9 -0
 
+static int blkdev_dio_aligned(struct block_device *bdev, loff_t pos,
+			      struct iov_iter *iter)
+{
+	if ((pos | iov_iter_alignment(iter)) &
+	    (bdev_logical_block_size(bdev) - 1))
+		return -EINVAL;
+
+	return 0;
+}

Nit: The name of this helper really suggest a bool return, which would be a more flexible interface I think:

if (!blkdev_dio_aligned(bdev, pos, iter))
return -EINVAL; <-- or whatever error code the caller wants.

And that will allow you to get rid of the ret variable in a least __blkdev_direct_IO_async.

Quoted 49 lines
block/fops.c +12 -11
+
 #define DIO_INLINE_BIO_VECS 4
 
 static ssize_t __blkdev_direct_IO_simple(struct kiocb *iocb,
@@ -54,9 +64,9 @@ static ssize_t __blkdev_direct_IO_simple(struct kiocb *iocb,
 	struct bio bio;
 	ssize_t ret;
 
-	if ((pos | iov_iter_alignment(iter)) &
-	    (bdev_logical_block_size(bdev) - 1))
-		return -EINVAL;
+	ret = blkdev_dio_aligned(bdev, pos, iter);
+	if (ret)
+		return ret;
 
 	if (nr_pages <= DIO_INLINE_BIO_VECS)
 		vecs = inline_vecs;
@@ -171,11 +181,11 @@ static ssize_t __blkdev_direct_IO(struct kiocb *iocb, struct iov_iter *iter,
 	bool is_read = (iov_iter_rw(iter) == READ), is_sync;
 	unsigned int opf = is_read ? REQ_OP_READ : dio_bio_write_op(iocb);
 	loff_t pos = iocb->ki_pos;
-	int ret = 0;
+	int ret;
 
-	if ((pos | iov_iter_alignment(iter)) &
-	    (bdev_logical_block_size(bdev) - 1))
-		return -EINVAL;
+	ret = blkdev_dio_aligned(bdev, pos, iter);
+	if (ret)
+		return ret;
 
 	if (iocb->ki_flags & IOCB_ALLOC_CACHE)
 		opf |= REQ_ALLOC_CACHE;
@@ -296,11 +306,11 @@ static ssize_t __blkdev_direct_IO_async(struct kiocb *iocb,
 	struct blkdev_dio *dio;
 	struct bio *bio;
 	loff_t pos = iocb->ki_pos;
-	int ret = 0;
+	int ret;
 
-	if ((pos | iov_iter_alignment(iter)) &
-	    (bdev_logical_block_size(bdev) - 1))
-		return -EINVAL;
+	ret = blkdev_dio_aligned(bdev, pos, iter);
+	if (ret)
+		return ret;
 
 	if (iocb->ki_flags & IOCB_ALLOC_CACHE)
 		opf |= REQ_ALLOC_CACHE;
Signature · 2 lines
Damien Le Moal
Western Digital Research
#10 lore raw

Re: [PATCHv4 5/9] block: add a helper function for dio alignment

Quoted 8 lines · Damien Le Moal

Nit: The name of this helper really suggest a bool return, which would be a more flexible interface I think:

if (!blkdev_dio_aligned(bdev, pos, iter))
return -EINVAL; <-- or whatever error code the caller wants.

And that will allow you to get rid of the ret variable in a least __blkdev_direct_IO_async.

I agree, a bool return looks more natural here.

Otherwise this looks good to me.

#11 lore raw

PATCH v4 6/9 block/merge: count bytes instead of sectors

In reply to Keith Busch

From: Keith Busch <[email protected]>

Individual bv_len's may not be a sector size.

---
v3->v4:

Use sector shift

Add comment explaing the ALIGN_DOWN

 block/blk-merge.c | 41 ++++++++++++++++++++++++-----------------
 1 file changed, 24 insertions(+), 17 deletions(-)
block/blk-merge.c +24 -17
index 7771dacc99cb..9ff0cb9e4840 100644
@@ -201,11 +201,11 @@ static inline unsigned get_max_segment_size(const struct request_queue *q,
  * @nsegs:    [in,out] Number of segments in the bio being built. Incremented
  *            by the number of segments from @bv that may be appended to that
  *            bio without exceeding @max_segs
- * @sectors:  [in,out] Number of sectors in the bio being built. Incremented
- *            by the number of sectors from @bv that may be appended to that
- *            bio without exceeding @max_sectors
+ * @bytes:    [in,out] Number of bytes in the bio being built. Incremented
+ *            by the number of bytes from @bv that may be appended to that
+ *            bio without exceeding @max_bytes
  * @max_segs: [in] upper bound for *@nsegs
- * @max_sectors: [in] upper bound for *@sectors
+ * @max_bytes: [in] upper bound for *@bytes
  *
  * When splitting a bio, it can happen that a bvec is encountered that is too
  * big to fit in a single segment and hence that it has to be split in the
@@ -216,10 +216,10 @@ static inline unsigned get_max_segment_size(const struct request_queue *q,
  */
 static bool bvec_split_segs(const struct request_queue *q,
 			    const struct bio_vec *bv, unsigned *nsegs,
-			    unsigned *sectors, unsigned max_segs,
-			    unsigned max_sectors)
+			    unsigned *bytes, unsigned max_segs,
+			    unsigned max_bytes)
 {
-	unsigned max_len = (min(max_sectors, UINT_MAX >> 9) - *sectors) << 9;
+	unsigned max_len = min(max_bytes, UINT_MAX) - *bytes;
 	unsigned len = min(bv->bv_len, max_len);
 	unsigned total_len = 0;
 	unsigned seg_size = 0;
@@ -237,7 +237,7 @@ static bool bvec_split_segs(const struct request_queue *q,
 			break;
 	}
 
-	*sectors += total_len >> 9;
+	*bytes += total_len;
 
 	/* tell the caller to split the bvec if it is too big to fit */
 	return len > 0 || bv->bv_len > max_len;
@@ -269,8 +269,8 @@ static struct bio *blk_bio_segment_split(struct request_queue *q,
 {
 	struct bio_vec bv, bvprv, *bvprvp = NULL;
 	struct bvec_iter iter;
-	unsigned nsegs = 0, sectors = 0;
-	const unsigned max_sectors = get_max_io_size(q, bio);
+	unsigned nsegs = 0, bytes = 0;
+	const unsigned max_bytes = get_max_io_size(q, bio) << 9;
 	const unsigned max_segs = queue_max_segments(q);
 
 	bio_for_each_bvec(bv, bio, iter) {
@@ -282,12 +282,12 @@ static struct bio *blk_bio_segment_split(struct request_queue *q,
 			goto split;
 
 		if (nsegs < max_segs &&
-		    sectors + (bv.bv_len >> 9) <= max_sectors &&
+		    bytes + bv.bv_len <= max_bytes &&
 		    bv.bv_offset + bv.bv_len <= PAGE_SIZE) {
 			nsegs++;
-			sectors += bv.bv_len >> 9;
-		} else if (bvec_split_segs(q, &bv, &nsegs, &sectors, max_segs,
-					 max_sectors)) {
+			bytes += bv.bv_len;
+		} else if (bvec_split_segs(q, &bv, &nsegs, &bytes, max_segs,
+					   max_bytes)) {
 			goto split;
 		}
 
@@ -300,13 +300,20 @@ static struct bio *blk_bio_segment_split(struct request_queue *q,
 split:
 	*segs = nsegs;
 
+	/*
+	 * Individual bvecs may not be logical block aligned, so round down to
+	 * that size to ensure both sides of the split bio are appropriately
+	 * sized.
+	 */
+	bytes = ALIGN_DOWN(bytes, queue_logical_block_size(q));
+
 	/*
 	 * Bio splitting may cause subtle trouble such as hang when doing sync
 	 * iopoll in direct IO routine. Given performance gain of iopoll for
 	 * big IO can be trival, disable iopoll when split needed.
 	 */
 	bio_clear_polled(bio);
-	return bio_split(bio, sectors, GFP_NOIO, bs);
+	return bio_split(bio, bytes >> SECTOR_SHIFT, GFP_NOIO, bs);
 }
 
 /**
@@ -375,7 +382,7 @@ EXPORT_SYMBOL(blk_queue_split);
 unsigned int blk_recalc_rq_segments(struct request *rq)
 {
 	unsigned int nr_phys_segs = 0;
-	unsigned int nr_sectors = 0;
+	unsigned int bytes = 0;
 	struct req_iterator iter;
 	struct bio_vec bv;
 
@@ -398,7 +405,7 @@ unsigned int blk_recalc_rq_segments(struct request *rq)
 	}
 
 	rq_for_each_bvec(bv, rq, iter)
-		bvec_split_segs(rq->q, &bv, &nr_phys_segs, &nr_sectors,
+		bvec_split_segs(rq->q, &bv, &nr_phys_segs, &bytes,
 				UINT_MAX, UINT_MAX);
 	return nr_phys_segs;
 }
Signature · 1 line
2.30.2
#12 lore raw

Re: [PATCHv4 6/9] block/merge: count bytes instead of sectors

Quoted 59 lines · Keith Busch

From: Keith Busch <[email protected]>

Individual bv_len's may not be a sector size.

---
v3->v4:

Use sector shift

Add comment explaing the ALIGN_DOWN

 block/blk-merge.c | 41 ++++++++++++++++++++++++-----------------
 1 file changed, 24 insertions(+), 17 deletions(-)
block/blk-merge.c +8 -8
index 7771dacc99cb..9ff0cb9e4840 100644
@@ -201,11 +201,11 @@ static inline unsigned get_max_segment_size(const struct request_queue *q,
  * @nsegs:    [in,out] Number of segments in the bio being built. Incremented
  *            by the number of segments from @bv that may be appended to that
  *            bio without exceeding @max_segs
- * @sectors:  [in,out] Number of sectors in the bio being built. Incremented
- *            by the number of sectors from @bv that may be appended to that
- *            bio without exceeding @max_sectors
+ * @bytes:    [in,out] Number of bytes in the bio being built. Incremented
+ *            by the number of bytes from @bv that may be appended to that
+ *            bio without exceeding @max_bytes
  * @max_segs: [in] upper bound for *@nsegs
- * @max_sectors: [in] upper bound for *@sectors
+ * @max_bytes: [in] upper bound for *@bytes
  *
  * When splitting a bio, it can happen that a bvec is encountered that is too
  * big to fit in a single segment and hence that it has to be split in the
@@ -216,10 +216,10 @@ static inline unsigned get_max_segment_size(const struct request_queue *q,
  */
 static bool bvec_split_segs(const struct request_queue *q,
 			    const struct bio_vec *bv, unsigned *nsegs,
-			    unsigned *sectors, unsigned max_segs,
-			    unsigned max_sectors)
+			    unsigned *bytes, unsigned max_segs,
+			    unsigned max_bytes)
 {
-	unsigned max_len = (min(max_sectors, UINT_MAX >> 9) - *sectors) << 9;
+	unsigned max_len = min(max_bytes, UINT_MAX) - *bytes;
 	unsigned len = min(bv->bv_len, max_len);
 	unsigned total_len = 0;
 	unsigned seg_size = 0;
@@ -237,7 +237,7 @@ static bool bvec_split_segs(const struct request_queue *q,
 			break;
 	}
 
-	*sectors += total_len >> 9;
+	*bytes += total_len;
 
 	/* tell the caller to split the bvec if it is too big to fit */
 	return len > 0 || bv->bv_len > max_len;

block/blk-merge.c static struct bio *blk_bio_segment_split(struct request_queue *q, · line 273

Quoted 7 lines
block/blk-merge.c +2 -2
@@ -269,8 +269,8 @@ static struct bio *blk_bio_segment_split(struct request_queue *q,
 {
 	struct bio_vec bv, bvprv, *bvprvp = NULL;
 	struct bvec_iter iter;
-	unsigned nsegs = 0, sectors = 0;
-	const unsigned max_sectors = get_max_io_size(q, bio);
+	unsigned nsegs = 0, bytes = 0;
+	const unsigned max_bytes = get_max_io_size(q, bio) << 9;

You missed one SECTOR_SHIFT here :)

Also, this get_max_io_size() function is now super confusing. It really should be get_max_io_sectors() and we could add also get_max_io_bytes(), no ?

Quoted 60 lines
block/blk-merge.c +14 -7
 	const unsigned max_segs = queue_max_segments(q);
 
 	bio_for_each_bvec(bv, bio, iter) {
@@ -282,12 +282,12 @@ static struct bio *blk_bio_segment_split(struct request_queue *q,
 			goto split;
 
 		if (nsegs < max_segs &&
-		    sectors + (bv.bv_len >> 9) <= max_sectors &&
+		    bytes + bv.bv_len <= max_bytes &&
 		    bv.bv_offset + bv.bv_len <= PAGE_SIZE) {
 			nsegs++;
-			sectors += bv.bv_len >> 9;
-		} else if (bvec_split_segs(q, &bv, &nsegs, &sectors, max_segs,
-					 max_sectors)) {
+			bytes += bv.bv_len;
+		} else if (bvec_split_segs(q, &bv, &nsegs, &bytes, max_segs,
+					   max_bytes)) {
 			goto split;
 		}
 
@@ -300,13 +300,20 @@ static struct bio *blk_bio_segment_split(struct request_queue *q,
 split:
 	*segs = nsegs;
 
+	/*
+	 * Individual bvecs may not be logical block aligned, so round down to
+	 * that size to ensure both sides of the split bio are appropriately
+	 * sized.
+	 */
+	bytes = ALIGN_DOWN(bytes, queue_logical_block_size(q));
+
 	/*
 	 * Bio splitting may cause subtle trouble such as hang when doing sync
 	 * iopoll in direct IO routine. Given performance gain of iopoll for
 	 * big IO can be trival, disable iopoll when split needed.
 	 */
 	bio_clear_polled(bio);
-	return bio_split(bio, sectors, GFP_NOIO, bs);
+	return bio_split(bio, bytes >> SECTOR_SHIFT, GFP_NOIO, bs);
 }
 
 /**
@@ -375,7 +382,7 @@ EXPORT_SYMBOL(blk_queue_split);
 unsigned int blk_recalc_rq_segments(struct request *rq)
 {
 	unsigned int nr_phys_segs = 0;
-	unsigned int nr_sectors = 0;
+	unsigned int bytes = 0;
 	struct req_iterator iter;
 	struct bio_vec bv;
 
@@ -398,7 +405,7 @@ unsigned int blk_recalc_rq_segments(struct request *rq)
 	}
 
 	rq_for_each_bvec(bv, rq, iter)
-		bvec_split_segs(rq->q, &bv, &nr_phys_segs, &nr_sectors,
+		bvec_split_segs(rq->q, &bv, &nr_phys_segs, &bytes,
 				UINT_MAX, UINT_MAX);
 	return nr_phys_segs;
 }
Signature · 2 lines
Damien Le Moal
Western Digital Research
Reviewed-by on v4 6/9
#13 lore raw

Re: [PATCHv4 6/9] block/merge: count bytes instead of sectors

In reply to Keith Busch

block/blk-merge.c static struct bio *blk_bio_segment_split(struct request_queue *q, · line 307

Quoted 5 lines · Keith Busch
block/blk-merge.c +5 -0
+	/*
+	 * Individual bvecs may not be logical block aligned, so round down to
+	 * that size to ensure both sides of the split bio are appropriately
+	 * sized.
+	 */

Maybe I'd word this as:

/*
* Individual bvecs may not be logical block aligned. Round down
* the split size so that each bio is properly sector size aligned,
* even if we do not use the full hardware limits.
*/

Otherwise this looks good:

#14 lore raw

PATCH v4 7/9 block/bounce: count bytes instead of sectors

In reply to Keith Busch

From: Keith Busch <[email protected]>

Individual bv_len's may not be a sector size.

---
v3->v4:

Use sector shift

Add comment explaing the ALIGN_DOWN

Use unsigned int type for counting bytes

 block/bounce.c | 12 +++++++++---
 1 file changed, 9 insertions(+), 3 deletions(-)
block/bounce.c +9 -3
index 8f7b6fe3b4db..f6ae21ec2a70 100644
@@ -205,19 +205,25 @@ void __blk_queue_bounce(struct request_queue *q, struct bio **bio_orig)
 	int rw = bio_data_dir(*bio_orig);
 	struct bio_vec *to, from;
 	struct bvec_iter iter;
-	unsigned i = 0;
+	unsigned i = 0, bytes = 0;
 	bool bounce = false;
-	int sectors = 0;
+	int sectors;
 
 	bio_for_each_segment(from, *bio_orig, iter) {
 		if (i++ < BIO_MAX_VECS)
-			sectors += from.bv_len >> 9;
+			bytes += from.bv_len;
 		if (PageHighMem(from.bv_page))
 			bounce = true;
 	}
 	if (!bounce)
 		return;
 
+	/*
+	 * If the original has more than BIO_MAX_VECS biovecs, the total bytes
+	 * may not be block size aligned. Align down to ensure both sides of
+	 * the split bio are appropriately sized.
+	 */
+	sectors = ALIGN_DOWN(bytes, queue_logical_block_size(q)) >> SECTOR_SHIFT;
 	if (sectors < bio_sectors(*bio_orig)) {
 		bio = bio_split(*bio_orig, sectors, GFP_NOIO, &bounce_bio_split);
 		bio_chain(bio, *bio_orig);
Signature · 1 line
2.30.2
Reviewed-by on v4 7/9
#15 lore raw

Re: [PATCHv4 7/9] block/bounce: count bytes instead of sectors

Quoted 40 lines · Keith Busch

From: Keith Busch <[email protected]>

Individual bv_len's may not be a sector size.

---
v3->v4:

Use sector shift

Add comment explaing the ALIGN_DOWN

Use unsigned int type for counting bytes

 block/bounce.c | 12 +++++++++---
 1 file changed, 9 insertions(+), 3 deletions(-)
block/bounce.c +3 -3
index 8f7b6fe3b4db..f6ae21ec2a70 100644
@@ -205,19 +205,25 @@ void __blk_queue_bounce(struct request_queue *q, struct bio **bio_orig)
 	int rw = bio_data_dir(*bio_orig);
 	struct bio_vec *to, from;
 	struct bvec_iter iter;
-	unsigned i = 0;
+	unsigned i = 0, bytes = 0;
 	bool bounce = false;
-	int sectors = 0;
+	int sectors;
 
 	bio_for_each_segment(from, *bio_orig, iter) {
 		if (i++ < BIO_MAX_VECS)
-			sectors += from.bv_len >> 9;
+			bytes += from.bv_len;
 		if (PageHighMem(from.bv_page))
 			bounce = true;
 	}
 	if (!bounce)
 		return;

block/bounce.c void __blk_queue_bounce(struct request_queue *q, struct bio **bio_orig) · line 229

Quoted 10 lines
block/bounce.c +6 -0
 
+	/*
+	 * If the original has more than BIO_MAX_VECS biovecs, the total bytes
+	 * may not be block size aligned. Align down to ensure both sides of
+	 * the split bio are appropriately sized.
+	 */
+	sectors = ALIGN_DOWN(bytes, queue_logical_block_size(q)) >> SECTOR_SHIFT;
 	if (sectors < bio_sectors(*bio_orig)) {
 		bio = bio_split(*bio_orig, sectors, GFP_NOIO, &bounce_bio_split);
 		bio_chain(bio, *bio_orig);

Looks good.

Signature · 2 lines
Damien Le Moal
Western Digital Research
Reviewed-by on v4 7/9
#16 lore raw

Re: [PATCHv4 7/9] block/bounce: count bytes instead of sectors

In reply to Keith Busch
Quoted 53 lines · Keith Busch

From: Keith Busch <[email protected]>

Individual bv_len's may not be a sector size.

---
v3->v4:

Use sector shift

Add comment explaing the ALIGN_DOWN

Use unsigned int type for counting bytes

 block/bounce.c | 12 +++++++++---
 1 file changed, 9 insertions(+), 3 deletions(-)
block/bounce.c +9 -3
index 8f7b6fe3b4db..f6ae21ec2a70 100644
@@ -205,19 +205,25 @@ void __blk_queue_bounce(struct request_queue *q, struct bio **bio_orig)
 	int rw = bio_data_dir(*bio_orig);
 	struct bio_vec *to, from;
 	struct bvec_iter iter;
-	unsigned i = 0;
+	unsigned i = 0, bytes = 0;
 	bool bounce = false;
-	int sectors = 0;
+	int sectors;
 
 	bio_for_each_segment(from, *bio_orig, iter) {
 		if (i++ < BIO_MAX_VECS)
-			sectors += from.bv_len >> 9;
+			bytes += from.bv_len;
 		if (PageHighMem(from.bv_page))
 			bounce = true;
 	}
 	if (!bounce)
 		return;
 
+	/*
+	 * If the original has more than BIO_MAX_VECS biovecs, the total bytes
+	 * may not be block size aligned. Align down to ensure both sides of
+	 * the split bio are appropriately sized.
+	 */
+	sectors = ALIGN_DOWN(bytes, queue_logical_block_size(q)) >> SECTOR_SHIFT;
 	if (sectors < bio_sectors(*bio_orig)) {
 		bio = bio_split(*bio_orig, sectors, GFP_NOIO, &bounce_bio_split);
 		bio_chain(bio, *bio_orig);
Signature · 1 line
2.30.2

Looks good,

Signature · 1 line
Pankaj Raghav
#17 lore raw

Re: [PATCHv4 7/9] block/bounce: count bytes instead of sectors

In reply to Keith Busch

block/bounce.c void __blk_queue_bounce(struct request_queue *q, struct bio **bio_orig) · line 225

Quoted 5 lines · Keith Busch
block/bounce.c +5 -0
+	/*
+	 * If the original has more than BIO_MAX_VECS biovecs, the total bytes
+	 * may not be block size aligned. Align down to ensure both sides of
+	 * the split bio are appropriately sized.
+	 */

Same comment on the comment as fo the previous patch.

block/bounce.c void __blk_queue_bounce(struct request_queue *q, struct bio **bio_orig) · line 226

Quoted 1 line
block/bounce.c +1 -0
+	sectors = ALIGN_DOWN(bytes, queue_logical_block_size(q)) >> SECTOR_SHIFT;

Overly long line here.

#18 lore raw

PATCH v4 8/9 block: relax direct io memory alignment

In reply to Keith Busch

From: Keith Busch <[email protected]>

Use the address alignment requirements from the hardware for direct io instead of requiring addresses be aligned to the block size. User space can discover the alignment requirements from the dma_alignment queue attribute.

User space can specify any hardware compatible DMA offset for each segment, but every segment length is still required to be a multiple of the block size.

---
 block/bio.c  | 12 ++++++++++++
 block/fops.c | 14 +++++++++++---
 2 files changed, 23 insertions(+), 3 deletions(-)
block/bio.c +12 -0
index 55d2a9c4e312..c492881959d1 100644
@@ -1219,7 +1219,19 @@ static int __bio_iov_iter_get_pages(struct bio *bio, struct iov_iter *iter)
 	BUILD_BUG_ON(PAGE_PTRS_PER_BVEC < 2);
 	pages += entries_left * (PAGE_PTRS_PER_BVEC - 1);
 
+	/*
+	 * Each segment in the iov is required to be a block size multiple.
+	 * However, we may not be able to get the entire segment if it spans
+	 * more pages than bi_max_vecs allows, so we have to ALIGN_DOWN the
+	 * result to ensure the bio's total size is correct. The remainder of
+	 * the iov data will be picked up in the next bio iteration.
+	 *
+	 * If the result is ever 0, that indicates the iov fails the segment
+	 * size requirement and is an error.
+	 */
 	size = iov_iter_get_pages(iter, pages, LONG_MAX, nr_pages, &offset);
+	if (size > 0)
+		size = ALIGN_DOWN(size, bdev_logical_block_size(bio->bi_bdev));
 	if (unlikely(size <= 0))
 		return size ? size : -EFAULT;
 
block/fops.c +11 -3
index bd6c2e13a4e3..6ecbccc552b9 100644
@@ -45,10 +45,10 @@ static unsigned int dio_bio_write_op(struct kiocb *iocb)
 static int blkdev_dio_aligned(struct block_device *bdev, loff_t pos,
 			      struct iov_iter *iter)
 {
-	if ((pos | iov_iter_alignment(iter)) &
-	    (bdev_logical_block_size(bdev) - 1))
+	if ((pos | iov_iter_count(iter)) & (bdev_logical_block_size(bdev) - 1))
+		return -EINVAL;
+	if (iov_iter_alignment(iter) & bdev_dma_alignment(bdev))
 		return -EINVAL;
-
 	return 0;
 }
 
@@ -88,6 +88,10 @@ static ssize_t __blkdev_direct_IO_simple(struct kiocb *iocb,
 	bio.bi_ioprio = iocb->ki_ioprio;
 
 	ret = bio_iov_iter_get_pages(&bio, iter);
+
+	/* check if iov is not aligned */
+	if (unlikely(!ret && iov_iter_count(iter)))
+		ret = -EINVAL;
 	if (unlikely(ret))
 		goto out;
 	ret = bio.bi_iter.bi_size;
@@ -333,6 +337,10 @@ static ssize_t __blkdev_direct_IO_async(struct kiocb *iocb,
 		bio_iov_bvec_set(bio, iter);
 	} else {
 		ret = bio_iov_iter_get_pages(bio, iter);
+
+		/* check if iov is not aligned */
+		if (unlikely(!ret && iov_iter_count(iter)))
+			ret = -EINVAL;
 		if (unlikely(ret)) {
 			bio_put(bio);
 			return ret;
Signature · 1 line
2.30.2
Reviewed-by on v4 8/9
#19 lore raw

Re: [PATCHv4 8/9] block: relax direct io memory alignment

Quoted 71 lines · Keith Busch

From: Keith Busch <[email protected]>

Use the address alignment requirements from the hardware for direct io instead of requiring addresses be aligned to the block size. User space can discover the alignment requirements from the dma_alignment queue attribute.

User space can specify any hardware compatible DMA offset for each segment, but every segment length is still required to be a multiple of the block size.

---
 block/bio.c  | 12 ++++++++++++
 block/fops.c | 14 +++++++++++---
 2 files changed, 23 insertions(+), 3 deletions(-)
block/bio.c +12 -0
index 55d2a9c4e312..c492881959d1 100644
@@ -1219,7 +1219,19 @@ static int __bio_iov_iter_get_pages(struct bio *bio, struct iov_iter *iter)
 	BUILD_BUG_ON(PAGE_PTRS_PER_BVEC < 2);
 	pages += entries_left * (PAGE_PTRS_PER_BVEC - 1);
 
+	/*
+	 * Each segment in the iov is required to be a block size multiple.
+	 * However, we may not be able to get the entire segment if it spans
+	 * more pages than bi_max_vecs allows, so we have to ALIGN_DOWN the
+	 * result to ensure the bio's total size is correct. The remainder of
+	 * the iov data will be picked up in the next bio iteration.
+	 *
+	 * If the result is ever 0, that indicates the iov fails the segment
+	 * size requirement and is an error.
+	 */
 	size = iov_iter_get_pages(iter, pages, LONG_MAX, nr_pages, &offset);
+	if (size > 0)
+		size = ALIGN_DOWN(size, bdev_logical_block_size(bio->bi_bdev));
 	if (unlikely(size <= 0))
 		return size ? size : -EFAULT;
 
block/fops.c +7 -3
index bd6c2e13a4e3..6ecbccc552b9 100644
@@ -45,10 +45,10 @@ static unsigned int dio_bio_write_op(struct kiocb *iocb)
 static int blkdev_dio_aligned(struct block_device *bdev, loff_t pos,
 			      struct iov_iter *iter)
 {
-	if ((pos | iov_iter_alignment(iter)) &
-	    (bdev_logical_block_size(bdev) - 1))
+	if ((pos | iov_iter_count(iter)) & (bdev_logical_block_size(bdev) - 1))
+		return -EINVAL;
+	if (iov_iter_alignment(iter) & bdev_dma_alignment(bdev))
 		return -EINVAL;
-
 	return 0;
 }
 
@@ -88,6 +88,10 @@ static ssize_t __blkdev_direct_IO_simple(struct kiocb *iocb,
 	bio.bi_ioprio = iocb->ki_ioprio;
 
 	ret = bio_iov_iter_get_pages(&bio, iter);
+
+	/* check if iov is not aligned */
+	if (unlikely(!ret && iov_iter_count(iter)))
+		ret = -EINVAL;
 	if (unlikely(ret))
 		goto out;
 	ret = bio.bi_iter.bi_size;

block/fops.c static ssize_t __blkdev_direct_IO_async(struct kiocb *iocb, · line 346

Quoted 10 lines
block/fops.c +4 -0
@@ -333,6 +337,10 @@ static ssize_t __blkdev_direct_IO_async(struct kiocb *iocb,
 		bio_iov_bvec_set(bio, iter);
 	} else {
 		ret = bio_iov_iter_get_pages(bio, iter);
+
+		/* check if iov is not aligned */
+		if (unlikely(!ret && iov_iter_count(iter)))
+			ret = -EINVAL;
 		if (unlikely(ret)) {
 			bio_put(bio);
 			return ret;

Looks OK to me.

Signature · 2 lines
Damien Le Moal
Western Digital Research
#20 lore raw

Re: [PATCHv4 8/9] block: relax direct io memory alignment

In reply to Keith Busch

block/bio.c static int __bio_iov_iter_get_pages(struct bio *bio, struct iov_iter *iter) · line 1223

Quoted 2 lines · Keith Busch
block/bio.c +2 -0
+	/*
+	 * Each segment in the iov is required to be a block size multiple.

Where is this enforced?

- Eric

#21 lore raw

Re: [PATCHv4 8/9] block: relax direct io memory alignment

Quoted 5 lines · Eric Biggers

block/bio.c static int __bio_iov_iter_get_pages(struct bio *bio, struct iov_iter *iter) · line 1223

Quoted 2 lines · Keith Busch
block/bio.c +2 -0
+	/*
+	 * Each segment in the iov is required to be a block size multiple.

Where is this enforced?

Right below the comment. If it isn't a block size multiple, then ALIGN_DOWN will eventually result in 0 and -EFAULT is returned.

#22 lore raw

Re: [PATCHv4 8/9] block: relax direct io memory alignment

Quoted 9 lines · Keith Busch
Quoted 5 lines · Eric Biggers

block/bio.c static int __bio_iov_iter_get_pages(struct bio *bio, struct iov_iter *iter) · line 1223

Quoted 2 lines · Keith Busch
block/bio.c +2 -0
+	/*
+	 * Each segment in the iov is required to be a block size multiple.

Where is this enforced?

Right below the comment. If it isn't a block size multiple, then ALIGN_DOWN will eventually result in 0 and -EFAULT is returned.

That's interesting, I would have expected it to be checked in blkdev_dio_aligned().

EFAULT isn't the correct error code for this case; it should be EINVAL as is normally the case for bad alignment. See the man pages for read and write.

- Eric

#23 lore raw

Re: [PATCHv4 8/9] block: relax direct io memory alignment

Quoted 8 lines · Eric Biggers
Quoted 7 lines · Keith Busch
Quoted 5 lines · Eric Biggers

block/bio.c static int __bio_iov_iter_get_pages(struct bio *bio, struct iov_iter *iter) · line 1223

Quoted 2 lines · Keith Busch
block/bio.c +2 -0
+	/*
+	 * Each segment in the iov is required to be a block size multiple.

Where is this enforced?

Quoted 5 lines
Quoted 2 lines

Right below the comment. If it isn't a block size multiple, then ALIGN_DOWN will eventually result in 0 and -EFAULT is returned.

That's interesting, I would have expected it to be checked in blkdev_dio_aligned().

That would require a change to the iov_iter_alignment API, or a new function entirely.

Quoted 2 lines

EFAULT isn't the correct error code for this case; it should be EINVAL as is normally the case for bad alignment. See the man pages for read and write.

The EFAULT was just to get the do-while loop to break out. The callers in this patch still return EINVAL when it sees the iov_iter hasn't advanced to the end.

But there are some cases where the EFAULT would be returned to the user. Let me see how difficult it would be catch it early in blkdev_dio_aligned().

#24 lore raw

Re: [PATCHv4 8/9] block: relax direct io memory alignment

Quoted 2 lines · Keith Busch

Let me see how difficult it would be catch it early in blkdev_dio_aligned().

Something like this appears to work:

---

block/fops.c +2 -1
index 6ecbccc552b9..20763a143991 100644
@@ -47,7 +47,8 @@ static int blkdev_dio_aligned(struct block_device *bdev, loff_t pos,
 {
 	if ((pos | iov_iter_count(iter)) & (bdev_logical_block_size(bdev) - 1))
 		return -EINVAL;
-	if (iov_iter_alignment(iter) & bdev_dma_alignment(bdev))
+	if (!iov_iter_aligned(iter, bdev_dma_alignment(bdev),
+			      bdev_logical_block_size(bdev) - 1))
 		return -EINVAL;
 	return 0;
 }
include/linux/uio.h +2 -0
index 739285fe5a2f..a9cbf90d8dcd 100644
@@ -219,6 +219,8 @@ size_t _copy_mc_to_iter(const void *addr, size_t bytes, struct iov_iter *i);
 #endif
 
 size_t iov_iter_zero(size_t bytes, struct iov_iter *);
+bool iov_iter_aligned(const struct iov_iter *i, unsigned addr_mask,
+			unsigned len_mask);
 unsigned long iov_iter_alignment(const struct iov_iter *i);
 unsigned long iov_iter_gap_alignment(const struct iov_iter *i);
 void iov_iter_init(struct iov_iter *i, unsigned int direction, const struct iovec *iov,
lib/iov_iter.c +82 -0

88 lines not on this page · raw message

--

#25 lore raw

Re: [PATCHv4 8/9] block: relax direct io memory alignment

Quoted 5 lines · Keith Busch
Quoted 2 lines · Keith Busch

Let me see how difficult it would be catch it early in blkdev_dio_aligned().

Something like this appears to work:

This looks reasonable to me. Nits below:

block/fops.c static int blkdev_dio_aligned(struct block_device *bdev, loff_t pos, · line 51

Quoted 2 lines
block/fops.c +2 -0
+	if (!iov_iter_aligned(iter, bdev_dma_alignment(bdev),
+			      bdev_logical_block_size(bdev) - 1))

This probably wants a helper so that it can also be reused by e.g. iomap.

#26 lore raw

PATCH v4 9/9 fs: add support for dma aligned direct-io

In reply to Keith Busch

From: Keith Busch <[email protected]>

Use the address alignment requirements from the hardware for direct io instead of requiring addresses be aligned to the block size.

---
 fs/direct-io.c       | 11 +++++++----
 fs/iomap/direct-io.c |  3 ++-
 2 files changed, 9 insertions(+), 5 deletions(-)
fs/direct-io.c +7 -4
index 840752006f60..64cc176be60c 100644
@@ -1131,7 +1131,7 @@ ssize_t __blockdev_direct_IO(struct kiocb *iocb, struct inode *inode,
 	struct dio_submit sdio = { 0, };
 	struct buffer_head map_bh = { 0, };
 	struct blk_plug plug;
-	unsigned long align = offset | iov_iter_alignment(iter);
+	unsigned long align = iov_iter_alignment(iter);
 
 	/*
 	 * Avoid references to bdev if not absolutely needed to give
@@ -1165,11 +1165,14 @@ ssize_t __blockdev_direct_IO(struct kiocb *iocb, struct inode *inode,
 		goto fail_dio;
 	}
 
-	if (align & blocksize_mask) {
-		if (bdev)
+	if ((offset | align) & blocksize_mask) {
+		if (bdev) {
 			blkbits = blksize_bits(bdev_logical_block_size(bdev));
+			if (align & bdev_dma_alignment(bdev))
+				goto fail_dio;
+		}
 		blocksize_mask = (1 << blkbits) - 1;
-		if (align & blocksize_mask)
+		if ((offset | count) & blocksize_mask)
 			goto fail_dio;
 	}
 
fs/iomap/direct-io.c +2 -1
index 80f9b047aa1b..0256d28baa8e 100644
@@ -244,7 +244,8 @@ static loff_t iomap_dio_bio_iter(const struct iomap_iter *iter,
 	size_t copied = 0;
 	size_t orig_count;
 
-	if ((pos | length | align) & ((1 << blkbits) - 1))
+	if ((pos | length) & ((1 << blkbits) - 1) ||
+	    align & bdev_dma_alignment(iomap->bdev))
 		return -EINVAL;
 
 	if (iomap->type == IOMAP_UNWRITTEN) {
Signature · 1 line
2.30.2
Reviewed-by on v4 9/9
#27 lore raw

Re: [PATCHv4 9/9] fs: add support for dma aligned direct-io

Quoted 47 lines · Keith Busch

From: Keith Busch <[email protected]>

Use the address alignment requirements from the hardware for direct io instead of requiring addresses be aligned to the block size.

---
 fs/direct-io.c       | 11 +++++++----
 fs/iomap/direct-io.c |  3 ++-
 2 files changed, 9 insertions(+), 5 deletions(-)
fs/direct-io.c +7 -4
index 840752006f60..64cc176be60c 100644
@@ -1131,7 +1131,7 @@ ssize_t __blockdev_direct_IO(struct kiocb *iocb, struct inode *inode,
 	struct dio_submit sdio = { 0, };
 	struct buffer_head map_bh = { 0, };
 	struct blk_plug plug;
-	unsigned long align = offset | iov_iter_alignment(iter);
+	unsigned long align = iov_iter_alignment(iter);
 
 	/*
 	 * Avoid references to bdev if not absolutely needed to give
@@ -1165,11 +1165,14 @@ ssize_t __blockdev_direct_IO(struct kiocb *iocb, struct inode *inode,
 		goto fail_dio;
 	}
 
-	if (align & blocksize_mask) {
-		if (bdev)
+	if ((offset | align) & blocksize_mask) {
+		if (bdev) {
 			blkbits = blksize_bits(bdev_logical_block_size(bdev));
+			if (align & bdev_dma_alignment(bdev))
+				goto fail_dio;
+		}
 		blocksize_mask = (1 << blkbits) - 1;
-		if (align & blocksize_mask)
+		if ((offset | count) & blocksize_mask)
 			goto fail_dio;
 	}
 
fs/iomap/direct-io.c +0 -0
index 80f9b047aa1b..0256d28baa8e 100644

fs/iomap/direct-io.c static loff_t iomap_dio_bio_iter(const struct iomap_iter *iter, · line 251

Quoted 9 lines
fs/iomap/direct-io.c +2 -1
@@ -244,7 +244,8 @@ static loff_t iomap_dio_bio_iter(const struct iomap_iter *iter,
 	size_t copied = 0;
 	size_t orig_count;
 
-	if ((pos | length | align) & ((1 << blkbits) - 1))
+	if ((pos | length) & ((1 << blkbits) - 1) ||
+	    align & bdev_dma_alignment(iomap->bdev))
 		return -EINVAL;
 
 	if (iomap->type == IOMAP_UNWRITTEN) {

Looks good to me.

Signature · 2 lines
Damien Le Moal
Western Digital Research
#28 lore raw

Re: [PATCHv4 9/9] fs: add support for dma aligned direct-io

In reply to Keith Busch
Quoted 4 lines · Keith Busch

From: Keith Busch <[email protected]>

Use the address alignment requirements from the hardware for direct io instead of requiring addresses be aligned to the block size.

I'd suggest dropping the legacy direct-io.c change. Anyone who wants new features really should be using the iomap code, which also significantly reduced the amount of testing needed here and the chance of regressions.