[Qemu-devel] [PATCH] scsi: fix accounting of writes

[Qemu-devel] [PATCH] scsi: fix accounting of writes

[Paolo Bonzini]

Writes go through scsi_write_complete at least twice, the first time
to get some data without having actually written anything.  Because
of this, the first time scsi_write_complete is called it will call
bdrv_acct_done and account a read incorrectly.  Fix this by looking
at the aiocb.  I am doing the same in scsi_read_complete for symmetry,
but it is only needed in the (bogus) case of bdrv_aio_readv returning
NULL.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
---
 hw/scsi-disk.c |   14 ++++++++------
 1 files changed, 8 insertions(+), 6 deletions(-)

diff --git a/hw/scsi-disk.c b/hw/scsi-disk.c
index dba1cfa..bdeba14 100644
--- a/hw/scsi-disk.c
+++ b/hw/scsi-disk.c
@@ -149,9 +149,10 @@ static void scsi_read_complete(void * opaque, int ret)
     SCSIDiskState *s = DO_UPCAST(SCSIDiskState, qdev, r->req.dev);
     int n;
 
-    r->req.aiocb = NULL;
-
-    bdrv_acct_done(s->bs, &r->acct);
+    if (r->req.aiocb != NULL) {
+        r->req.aiocb = NULL;
+        bdrv_acct_done(s->bs, &r->acct);
+    }
 
     if (ret) {
         if (scsi_handle_rw_error(r, -ret, SCSI_REQ_STATUS_RETRY_READ)) {
@@ -254,9 +255,10 @@ static void scsi_write_complete(void * opaque, int ret)
     SCSIDiskState *s = DO_UPCAST(SCSIDiskState, qdev, r->req.dev);
     uint32_t n;
 
-    r->req.aiocb = NULL;
-
-    bdrv_acct_done(s->bs, &r->acct);
+    if (r->req.aiocb != NULL) {
+        r->req.aiocb = NULL;
+        bdrv_acct_done(s->bs, &r->acct);
+    }
 
     if (ret) {
         if (scsi_handle_rw_error(r, -ret, SCSI_REQ_STATUS_RETRY_WRITE)) {
-- 
1.7.6

Re: [Qemu-devel] [PATCH] scsi: fix accounting of writes

[Kevin Wolf]

Am 05.09.2011 18:11, schrieb Paolo Bonzini:
> Writes go through scsi_write_complete at least twice, the first time
> to get some data without having actually written anything.  Because
> of this, the first time scsi_write_complete is called it will call
> bdrv_acct_done and account a read incorrectly.  Fix this by looking
> at the aiocb.  I am doing the same in scsi_read_complete for symmetry,
> but it is only needed in the (bogus) case of bdrv_aio_readv returning
> NULL.
> 
> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

Thanks, applied to the block branch.

Kevin