Thread (27 messages) 27 messages, 6 authors, 2023-03-30

Re: [PATCH v8 4/9] fs, block: copy_file_range for def_blk_ops for direct block device.

From: Christian Brauner <brauner@kernel.org>
Date: 2023-03-29 12:15:01
Also in: dm-devel, linux-fsdevel, linux-nvme, lkml

On Mon, Mar 27, 2023 at 02:10:52PM +0530, Anuj Gupta wrote:
quoted hunk ↗ jump to hunk
From: Nitesh Shetty <redacted>

For direct block device opened with O_DIRECT, use copy_file_range to
issue device copy offload, and fallback to generic_copy_file_range incase
device copy offload capability is absent.
Modify checks to allow bdevs to use copy_file_range.

Suggested-by: Ming Lei <redacted>
Signed-off-by: Anuj Gupta <redacted>
Signed-off-by: Nitesh Shetty <redacted>
---
 block/blk-lib.c        | 22 ++++++++++++++++++++++
 block/fops.c           | 20 ++++++++++++++++++++
 fs/read_write.c        | 11 +++++++++--
 include/linux/blkdev.h |  3 +++
 4 files changed, 54 insertions(+), 2 deletions(-)
diff --git a/block/blk-lib.c b/block/blk-lib.c
index a21819e59b29..c288573c7e77 100644
--- a/block/blk-lib.c
+++ b/block/blk-lib.c
@@ -475,6 +475,28 @@ static inline bool blk_check_copy_offload(struct request_queue *q_in,
 	return blk_queue_copy(q_in) && blk_queue_copy(q_out);
 }
 
+int blkdev_copy_offload(struct block_device *bdev_in, loff_t pos_in,
+		      struct block_device *bdev_out, loff_t pos_out, size_t len,
+		      cio_iodone_t end_io, void *private, gfp_t gfp_mask)
+{
+	struct request_queue *in_q = bdev_get_queue(bdev_in);
+	struct request_queue *out_q = bdev_get_queue(bdev_out);
+	int ret = -EINVAL;
Why initialize to -EINVAL if blk_copy_sanity_check() initializes it
right away anyway?
+	bool offload = false;
Same thing with initializing offload.
quoted hunk ↗ jump to hunk
+
+	ret = blk_copy_sanity_check(bdev_in, pos_in, bdev_out, pos_out, len);
+	if (ret)
+		return ret;
+
+	offload = blk_check_copy_offload(in_q, out_q);
+	if (offload)
+		ret = __blk_copy_offload(bdev_in, pos_in, bdev_out, pos_out,
+				len, end_io, private, gfp_mask);
+
+	return ret;
+}
+EXPORT_SYMBOL_GPL(blkdev_copy_offload);
+
 /*
  * @bdev_in:	source block device
  * @pos_in:	source offset
diff --git a/block/fops.c b/block/fops.c
index d2e6be4e3d1c..3b7c05831d5c 100644
--- a/block/fops.c
+++ b/block/fops.c
@@ -611,6 +611,25 @@ static ssize_t blkdev_read_iter(struct kiocb *iocb, struct iov_iter *to)
 	return ret;
 }
 
+static ssize_t blkdev_copy_file_range(struct file *file_in, loff_t pos_in,
+				struct file *file_out, loff_t pos_out,
+				size_t len, unsigned int flags)
+{
+	struct block_device *in_bdev = I_BDEV(bdev_file_inode(file_in));
+	struct block_device *out_bdev = I_BDEV(bdev_file_inode(file_out));
+	int comp_len = 0;
+
+	if ((file_in->f_iocb_flags & IOCB_DIRECT) &&
+		(file_out->f_iocb_flags & IOCB_DIRECT))
+		comp_len = blkdev_copy_offload(in_bdev, pos_in, out_bdev,
+				 pos_out, len, NULL, NULL, GFP_KERNEL);
+	if (comp_len != len)
+		comp_len = generic_copy_file_range(file_in, pos_in + comp_len,
+			file_out, pos_out + comp_len, len - comp_len, flags);
I'm not deeply familiar with this code but this looks odd. It at least
seems possible that comp_len could be -EINVAL and len 20 at which point
you'd be doing len - comp_len aka 20 - 22 = -2 in generic_copy_file_range().
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help