Skip to contentSkip to conversation

support of sg_iovec in bsg's struct sg_io_v4

PATCH Boaz Harrosh 2 participants 2 messages Jan 14, 2009 – Jan 15, 2009

Subsystems (3)
SCSI Block layer BSG (block layer generic sg v4 driver)
lore
Share & tools
lore

Boaz Harrosh

author
#0 lore raw

Is below just missing or is there something I do not understand?

It looks like the conversion from struct sg_io_hdr to struct sg_io_v4 the use of sg_iovec was dropped. Even though all the members and code comments say it is the same.

This adds the support in bsg and struct sg_io_v4 exactly in the same way it was done in sg with struct sg_io_hdr.

BTW why is the ioctl SG_IO defined exactly the same for sg.c and bsg.c even though they have none compatible API/ABI? Is that because we have the protection of the first __s32 member that is 'Q' for bsg and 'S' or 0 for sg? Is that usually done with ioctl's?

Not-signed-or-tested: Boaz Harrosh <[email protected]>
---
git diff --stat -p block/bsg.c

 block/bsg.c |   18 ++++++++++++++----
 1 files changed, 14 insertions(+), 4 deletions(-)
block/bsg.c +14 -4
index d414bb5..1489458 100644
@@ -285,8 +285,13 @@ bsg_map_hdr(struct bsg_device *bd, struct sg_io_v4 *hdr, fmode_t has_write_perm)
 		next_rq->cmd_type = rq->cmd_type;
 
 		dxferp = (void*)(unsigned long)hdr->din_xferp;
-		ret =  blk_rq_map_user(q, next_rq, NULL, dxferp,
-				       hdr->din_xfer_len, GFP_KERNEL);
+		if (hdr->din_iovec_count)
+			ret = blk_rq_map_user_iov(q, next_rq, NULL, dxferp,
+						  hdr->din_iovec_count,
+						  hdr->din_xfer_len, GFP_KERNEL);
+		else
+			ret =  blk_rq_map_user(q, next_rq, NULL, dxferp,
+					       hdr->din_xfer_len, GFP_KERNEL);
 		if (ret)
 			goto out;
 	}

@@ -301,8 +306,13 @@ bsg_map_hdr(struct bsg_device *bd, struct sg_io_v4 *hdr, fmode_t has_write_perm)
 		dxfer_len = 0;
 
 	if (dxfer_len) {
-		ret = blk_rq_map_user(q, rq, NULL, dxferp, dxfer_len,
-				      GFP_KERNEL);
+		if (hdr->dout_iovec_count)
+			ret = blk_rq_map_user_iov(q, rq, NULL, dxferp,
+						  hdr->dout_iovec_count,
+						  dxfer_len, GFP_KERNEL);
+		else
+			ret = blk_rq_map_user(q, rq, NULL, dxferp, dxfer_len,
+					      GFP_KERNEL);
 		if (ret)
 			goto out;
 	}
maintainer
#1 lore raw
Quoted 5 lines · On Wed, 14 Jan 2009 20:01:07 +0200 Boaz Harrosh <[email protected]> wrote:

Is below just missing or is there something I do not understand?

It looks like the conversion from struct sg_io_hdr to struct sg_io_v4 the use of sg_iovec was dropped. Even though all the members and code comments say it is the same.

It's intentional. It has been discussed several times on linux-scsi.

Quoted 2 lines

This adds the support in bsg and struct sg_io_v4 exactly in the same way it was done in sg with struct sg_io_hdr.

sg_io_v4 will not support sg_iovec.

Quoted 4 lines

BTW why is the ioctl SG_IO defined exactly the same for sg.c and bsg.c even though they have none compatible API/ABI? Is that because we have the protection of the first __s32 member that is 'Q' for bsg and 'S' or 0 for sg? Is that usually done with ioctl's?

I don't know if it's common or not. But we did the same thing with sg v2 and sg v3's read/write interface.