[PATCH v2 10/9] qed: Simplify backing reads

Vladimir Sementsov-Ogievskiy posted 9 patches 5 years, 6 months ago
Maintainers: Paolo Bonzini <pbonzini@redhat.com>, Fam Zheng <fam@euphon.net>, Peter Lieven <pl@kamp.de>, Stefan Hajnoczi <stefanha@redhat.com>, Kevin Wolf <kwolf@redhat.com>, Ronnie Sahlberg <ronniesahlberg@gmail.com>, Max Reitz <mreitz@redhat.com>, Stefan Weil <sw@weilnetz.de>, Jeff Cody <codyprime@gmail.com>
There is a newer version of this series
[PATCH v2 10/9] qed: Simplify backing reads
Posted by Eric Blake 5 years, 6 months ago
The other four drivers that support backing files (qcow, qcow2,
parallels, vmdk) all rely on the block layer to populate zeroes when
reading beyond EOF of a short backing file.  We can simplify the qed
code by doing likewise.

Signed-off-by: Eric Blake <eblake@redhat.com>
---

I noticed this during my audit that v1 of Vladimir's series was correct.

No change in iotests results (test 274 is currently failing for qed,
but for other reasons:
+Traceback (most recent call last):
+  File "274", line 24, in <module>
+    iotests.verify_image_format(supported_fmts=['qcow2'])
+AttributeError: module 'iotests' has no attribute 'verify_image_format'
)

 block/qed.h |  1 -
 block/qed.c | 64 +++++------------------------------------------------
 2 files changed, 6 insertions(+), 59 deletions(-)

diff --git a/block/qed.h b/block/qed.h
index 42c115d8220c..3d12bf78d412 100644
--- a/block/qed.h
+++ b/block/qed.h
@@ -140,7 +140,6 @@ typedef struct QEDAIOCB {

     /* Current cluster scatter-gather list */
     QEMUIOVector cur_qiov;
-    QEMUIOVector *backing_qiov;
     uint64_t cur_pos;               /* position on block device, in bytes */
     uint64_t cur_cluster;           /* cluster offset in image file */
     unsigned int cur_nclusters;     /* number of clusters being accessed */
diff --git a/block/qed.c b/block/qed.c
index 927382995a0c..bea4b9f6cc97 100644
--- a/block/qed.c
+++ b/block/qed.c
@@ -849,56 +849,18 @@ static BDRVQEDState *acb_to_s(QEDAIOCB *acb)
  * @s:              QED state
  * @pos:            Byte position in device
  * @qiov:           Destination I/O vector
- * @backing_qiov:   Possibly shortened copy of qiov, to be allocated here
- * @cb:             Completion function
- * @opaque:         User data for completion function
  *
  * This function reads qiov->size bytes starting at pos from the backing file.
  * If there is no backing file then zeroes are read.
  */
 static int coroutine_fn qed_read_backing_file(BDRVQEDState *s, uint64_t pos,
-                                              QEMUIOVector *qiov,
-                                              QEMUIOVector **backing_qiov)
+                                              QEMUIOVector *qiov)
 {
-    uint64_t backing_length = 0;
-    size_t size;
-    int ret;
-
-    /* If there is a backing file, get its length.  Treat the absence of a
-     * backing file like a zero length backing file.
-     */
     if (s->bs->backing) {
-        int64_t l = bdrv_getlength(s->bs->backing->bs);
-        if (l < 0) {
-            return l;
-        }
-        backing_length = l;
-    }
-
-    /* Zero all sectors if reading beyond the end of the backing file */
-    if (pos >= backing_length ||
-        pos + qiov->size > backing_length) {
-        qemu_iovec_memset(qiov, 0, 0, qiov->size);
-    }
-
-    /* Complete now if there are no backing file sectors to read */
-    if (pos >= backing_length) {
-        return 0;
-    }
-
-    /* If the read straddles the end of the backing file, shorten it */
-    size = MIN((uint64_t)backing_length - pos, qiov->size);
-
-    assert(*backing_qiov == NULL);
-    *backing_qiov = g_new(QEMUIOVector, 1);
-    qemu_iovec_init(*backing_qiov, qiov->niov);
-    qemu_iovec_concat(*backing_qiov, qiov, 0, size);
-
-    BLKDBG_EVENT(s->bs->file, BLKDBG_READ_BACKING_AIO);
-    ret = bdrv_co_preadv(s->bs->backing, pos, size, *backing_qiov, 0);
-    if (ret < 0) {
-        return ret;
+        BLKDBG_EVENT(s->bs->file, BLKDBG_READ_BACKING_AIO);
+        return bdrv_co_preadv(s->bs->backing, pos, qiov->size, qiov, 0);
     }
+    qemu_iovec_memset(qiov, 0, 0, qiov->size);
     return 0;
 }

@@ -915,7 +877,6 @@ static int coroutine_fn qed_copy_from_backing_file(BDRVQEDState *s,
                                                    uint64_t offset)
 {
     QEMUIOVector qiov;
-    QEMUIOVector *backing_qiov = NULL;
     int ret;

     /* Skip copy entirely if there is no work to do */
@@ -925,13 +886,7 @@ static int coroutine_fn qed_copy_from_backing_file(BDRVQEDState *s,

     qemu_iovec_init_buf(&qiov, qemu_blockalign(s->bs, len), len);

-    ret = qed_read_backing_file(s, pos, &qiov, &backing_qiov);
-
-    if (backing_qiov) {
-        qemu_iovec_destroy(backing_qiov);
-        g_free(backing_qiov);
-        backing_qiov = NULL;
-    }
+    ret = qed_read_backing_file(s, pos, &qiov);

     if (ret) {
         goto out;
@@ -1339,8 +1294,7 @@ static int coroutine_fn qed_aio_read_data(void *opaque, int ret,
         qemu_iovec_memset(&acb->cur_qiov, 0, 0, acb->cur_qiov.size);
         r = 0;
     } else if (ret != QED_CLUSTER_FOUND) {
-        r = qed_read_backing_file(s, acb->cur_pos, &acb->cur_qiov,
-                                  &acb->backing_qiov);
+        r = qed_read_backing_file(s, acb->cur_pos, &acb->cur_qiov);
     } else {
         BLKDBG_EVENT(bs->file, BLKDBG_READ_AIO);
         r = bdrv_co_preadv(bs->file, offset, acb->cur_qiov.size,
@@ -1365,12 +1319,6 @@ static int coroutine_fn qed_aio_next_io(QEDAIOCB *acb)
     while (1) {
         trace_qed_aio_next_io(s, acb, 0, acb->cur_pos + acb->cur_qiov.size);

-        if (acb->backing_qiov) {
-            qemu_iovec_destroy(acb->backing_qiov);
-            g_free(acb->backing_qiov);
-            acb->backing_qiov = NULL;
-        }
-
         acb->qiov_offset += acb->cur_qiov.size;
         acb->cur_pos += acb->cur_qiov.size;
         qemu_iovec_reset(&acb->cur_qiov);
-- 
2.26.2


Re: [PATCH v2 10/9] qed: Simplify backing reads
Posted by Eric Blake 5 years, 6 months ago
On 5/7/20 9:45 AM, Eric Blake wrote:
> The other four drivers that support backing files (qcow, qcow2,
> parallels, vmdk) all rely on the block layer to populate zeroes when
> reading beyond EOF of a short backing file.  We can simplify the qed
> code by doing likewise.
> 
> Signed-off-by: Eric Blake <eblake@redhat.com>
> ---
> 
> I noticed this during my audit that v1 of Vladimir's series was correct.
> 
> No change in iotests results (test 274 is currently failing for qed,
> but for other reasons:
> +Traceback (most recent call last):
> +  File "274", line 24, in <module>
> +    iotests.verify_image_format(supported_fmts=['qcow2'])
> +AttributeError: module 'iotests' has no attribute 'verify_image_format'
> )

That iotest failure was due to a stale branch on my end; after updating 
to latest git master plus Kevin's latest 'block' branch, 274 is now 
skipped on qed.

-- 
Eric Blake, Principal Software Engineer
Red Hat, Inc.           +1-919-301-3226
Virtualization:  qemu.org | libvirt.org


Re: [PATCH v2 10/9] qed: Simplify backing reads
Posted by Vladimir Sementsov-Ogievskiy 5 years, 6 months ago
07.05.2020 17:45, Eric Blake wrote:
> The other four drivers that support backing files (qcow, qcow2,
> parallels, vmdk) all rely on the block layer to populate zeroes when
> reading beyond EOF of a short backing file.  We can simplify the qed
> code by doing likewise.
> 
> Signed-off-by: Eric Blake <eblake@redhat.com>

Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>

-- 
Best regards,
Vladimir