diff mbox series

[v4,04/16] block/io: refactor bdrv_pad_request(): move bdrv_pad_request() up

Message ID 20201211183934.169161-5-vsementsov@virtuozzo.com (mailing list archive)
State New, archived
Headers show
Series 64bit block-layer: part I | expand

Commit Message

Vladimir Sementsov-Ogievskiy Dec. 11, 2020, 6:39 p.m. UTC
Prepare to the following patch when bdrv_pad_request() will be able to
fail. Update the comments.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
---
 block/io.c | 25 +++++++++++++++++++------
 1 file changed, 19 insertions(+), 6 deletions(-)

Comments

Eric Blake Jan. 21, 2021, 10:50 p.m. UTC | #1
On 12/11/20 12:39 PM, Vladimir Sementsov-Ogievskiy wrote:
> Prepare to the following patch when bdrv_pad_request() will be able to

s/Prepare to/Prepare for/

> fail. Update the comments.
> 
> Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
> ---
>  block/io.c | 25 +++++++++++++++++++------
>  1 file changed, 19 insertions(+), 6 deletions(-)
> 

> -    if (bdrv_pad_request(bs, &qiov, &qiov_offset, &offset, &bytes, &pad)) {
> +    if (padded) {
> +        /*
> +         * Request was unaligned to request_alignment and therefore padded.
> +         * We are going to do read-modify-write. User is not prepared to widened
> +         * request intersections with other requests, so we serialize the
> +         * request.

We are going to do read-modify-write, and must serialize the request to
prevent interactions of the widened region with other transactions.

Reviewed-by: Eric Blake <eblake@redhat.com>
diff mbox series

Patch

diff --git a/block/io.c b/block/io.c
index d9bc67f1b0..dcfab267f8 100644
--- a/block/io.c
+++ b/block/io.c
@@ -2107,6 +2107,7 @@  int coroutine_fn bdrv_co_pwritev_part(BdrvChild *child,
     uint64_t align = bs->bl.request_alignment;
     BdrvRequestPadding pad;
     int ret;
+    bool padded = false;
 
     trace_bdrv_co_pwritev(child->bs, offset, bytes, flags);
 
@@ -2138,20 +2139,32 @@  int coroutine_fn bdrv_co_pwritev_part(BdrvChild *child,
         return 0;
     }
 
+    if (!(flags & BDRV_REQ_ZERO_WRITE)) {
+        /*
+         * Pad request for following read-modify-write cycle.
+         * bdrv_co_do_zero_pwritev() does aligning by itself, so, we do
+         * alignment only if there is no ZERO flag.
+         */
+        padded = bdrv_pad_request(bs, &qiov, &qiov_offset, &offset, &bytes,
+                                  &pad);
+    }
+
     bdrv_inc_in_flight(bs);
-    /*
-     * Align write if necessary by performing a read-modify-write cycle.
-     * Pad qiov with the read parts and be sure to have a tracked request not
-     * only for bdrv_aligned_pwritev, but also for the reads of the RMW cycle.
-     */
     tracked_request_begin(&req, bs, offset, bytes, BDRV_TRACKED_WRITE);
 
     if (flags & BDRV_REQ_ZERO_WRITE) {
+        assert(!padded);
         ret = bdrv_co_do_zero_pwritev(child, offset, bytes, flags, &req);
         goto out;
     }
 
-    if (bdrv_pad_request(bs, &qiov, &qiov_offset, &offset, &bytes, &pad)) {
+    if (padded) {
+        /*
+         * Request was unaligned to request_alignment and therefore padded.
+         * We are going to do read-modify-write. User is not prepared to widened
+         * request intersections with other requests, so we serialize the
+         * request.
+         */
         bdrv_mark_request_serialising(&req, align);
         bdrv_padding_rmw_read(child, &req, &pad, false);
     }