[v3] block: Fixes regarding dataplane and management operations

[Qemu-devel] [PATCH for 2.9 v3 10/10] block: Fix bdrv_co_flush early return

Posted by Fam Zheng 8 years, 10 months ago

bdrv_inc_in_flight and bdrv_dec_in_flight are mandatory for
BDRV_POLL_WHILE to work, even for the shortcut case where flush is
unnecessary. Move the if block to below bdrv_dec_in_flight, and BTW fix
the variable declaration position.

Signed-off-by: Fam Zheng <famz@redhat.com>
---
 block/io.c | 16 +++++++++-------
 1 file changed, 9 insertions(+), 7 deletions(-)

diff --git a/block/io.c b/block/io.c
index 00e45ca..bae6947 100644
--- a/block/io.c
+++ b/block/io.c
@@ -2278,16 +2278,17 @@ static void coroutine_fn bdrv_flush_co_entry(void *opaque)
 
 int coroutine_fn bdrv_co_flush(BlockDriverState *bs)
 {
-    int ret;
-
-    if (!bs || !bdrv_is_inserted(bs) || bdrv_is_read_only(bs) ||
-        bdrv_is_sg(bs)) {
-        return 0;
-    }
+    int current_gen;
+    int ret = 0;
 
     bdrv_inc_in_flight(bs);
 
-    int current_gen = bs->write_gen;
+    if (!bs || !bdrv_is_inserted(bs) || bdrv_is_read_only(bs) ||
+        bdrv_is_sg(bs)) {
+        goto early_exit;
+    }
+
+    current_gen = bs->write_gen;
 
     /* Wait until any previous flushes are completed */
     while (bs->active_flush_req) {
@@ -2370,6 +2371,7 @@ out:
     /* Return value is ignored - it's ok if wait queue is empty */
     qemu_co_queue_next(&bs->flush_queue);
 
+early_exit:
     bdrv_dec_in_flight(bs);
     return ret;
 }
-- 
2.9.3

Re: [Qemu-devel] [PATCH for 2.9 v3 10/10] block: Fix bdrv_co_flush early return

Posted by Paolo Bonzini 8 years, 10 months ago


On 10/04/2017 23:05, Fam Zheng wrote:
> bdrv_inc_in_flight and bdrv_dec_in_flight are mandatory for
> BDRV_POLL_WHILE to work, even for the shortcut case where flush is
> unnecessary. Move the if block to below bdrv_dec_in_flight, and BTW fix
> the variable declaration position.
> 
> Signed-off-by: Fam Zheng <famz@redhat.com>
> ---
>  block/io.c | 16 +++++++++-------
>  1 file changed, 9 insertions(+), 7 deletions(-)
> 
> diff --git a/block/io.c b/block/io.c
> index 00e45ca..bae6947 100644
> --- a/block/io.c
> +++ b/block/io.c
> @@ -2278,16 +2278,17 @@ static void coroutine_fn bdrv_flush_co_entry(void *opaque)
>  
>  int coroutine_fn bdrv_co_flush(BlockDriverState *bs)
>  {
> -    int ret;
> -
> -    if (!bs || !bdrv_is_inserted(bs) || bdrv_is_read_only(bs) ||
> -        bdrv_is_sg(bs)) {
> -        return 0;
> -    }
> +    int current_gen;
> +    int ret = 0;
>  
>      bdrv_inc_in_flight(bs);
>  
> -    int current_gen = bs->write_gen;
> +    if (!bs || !bdrv_is_inserted(bs) || bdrv_is_read_only(bs) ||
> +        bdrv_is_sg(bs)) {
> +        goto early_exit;
> +    }
> +
> +    current_gen = bs->write_gen;
>  
>      /* Wait until any previous flushes are completed */
>      while (bs->active_flush_req) {
> @@ -2370,6 +2371,7 @@ out:
>      /* Return value is ignored - it's ok if wait queue is empty */
>      qemu_co_queue_next(&bs->flush_queue);
>  
> +early_exit:
>      bdrv_dec_in_flight(bs);
>      return ret;
>  }
> 

Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>

Re: [Qemu-devel] [PATCH for 2.9 v3 10/10] block: Fix bdrv_co_flush early return

Posted by Kevin Wolf 8 years, 9 months ago

Am 10.04.2017 um 17:05 hat Fam Zheng geschrieben:
> bdrv_inc_in_flight and bdrv_dec_in_flight are mandatory for
> BDRV_POLL_WHILE to work, even for the shortcut case where flush is
> unnecessary. Move the if block to below bdrv_dec_in_flight, and BTW fix
> the variable declaration position.
> 
> Signed-off-by: Fam Zheng <famz@redhat.com>
> ---
>  block/io.c | 16 +++++++++-------
>  1 file changed, 9 insertions(+), 7 deletions(-)
> 
> diff --git a/block/io.c b/block/io.c
> index 00e45ca..bae6947 100644
> --- a/block/io.c
> +++ b/block/io.c
> @@ -2278,16 +2278,17 @@ static void coroutine_fn bdrv_flush_co_entry(void *opaque)
>  
>  int coroutine_fn bdrv_co_flush(BlockDriverState *bs)
>  {
> -    int ret;
> -
> -    if (!bs || !bdrv_is_inserted(bs) || bdrv_is_read_only(bs) ||
> -        bdrv_is_sg(bs)) {
> -        return 0;
> -    }
> +    int current_gen;
> +    int ret = 0;
>  
>      bdrv_inc_in_flight(bs);

As Coverity points out, we're now using bs...

> -    int current_gen = bs->write_gen;
> +    if (!bs || !bdrv_is_inserted(bs) || bdrv_is_read_only(bs) ||

...before doing the NULL check.

I'm not sure if we even need to have a NULL check here, but we would have
to check all callers to make sure that it's unnecessary. Before commit
29cdb251, it only checked bs->drv and I don't see how that commit
introduced a NULL caller, but maybe one was added later.

In any case, bdrv_co_flush() needs a fix, either remove the NULL check
or do it first.

> +        bdrv_is_sg(bs)) {
> +        goto early_exit;
> +    }

Kevin

Re: [Qemu-devel] [PATCH for 2.9 v3 10/10] block: Fix bdrv_co_flush early return

Posted by Fam Zheng 8 years, 9 months ago

On Tue, 04/25 17:16, Kevin Wolf wrote:
> Am 10.04.2017 um 17:05 hat Fam Zheng geschrieben:
> > bdrv_inc_in_flight and bdrv_dec_in_flight are mandatory for
> > BDRV_POLL_WHILE to work, even for the shortcut case where flush is
> > unnecessary. Move the if block to below bdrv_dec_in_flight, and BTW fix
> > the variable declaration position.
> > 
> > Signed-off-by: Fam Zheng <famz@redhat.com>
> > ---
> >  block/io.c | 16 +++++++++-------
> >  1 file changed, 9 insertions(+), 7 deletions(-)
> > 
> > diff --git a/block/io.c b/block/io.c
> > index 00e45ca..bae6947 100644
> > --- a/block/io.c
> > +++ b/block/io.c
> > @@ -2278,16 +2278,17 @@ static void coroutine_fn bdrv_flush_co_entry(void *opaque)
> >  
> >  int coroutine_fn bdrv_co_flush(BlockDriverState *bs)
> >  {
> > -    int ret;
> > -
> > -    if (!bs || !bdrv_is_inserted(bs) || bdrv_is_read_only(bs) ||
> > -        bdrv_is_sg(bs)) {
> > -        return 0;
> > -    }
> > +    int current_gen;
> > +    int ret = 0;
> >  
> >      bdrv_inc_in_flight(bs);
> 
> As Coverity points out, we're now using bs...
> 
> > -    int current_gen = bs->write_gen;
> > +    if (!bs || !bdrv_is_inserted(bs) || bdrv_is_read_only(bs) ||
> 
> ...before doing the NULL check.
> 
> I'm not sure if we even need to have a NULL check here, but we would have
> to check all callers to make sure that it's unnecessary. Before commit
> 29cdb251, it only checked bs->drv and I don't see how that commit
> introduced a NULL caller, but maybe one was added later.
> 
> In any case, bdrv_co_flush() needs a fix, either remove the NULL check
> or do it first.

After auditing the callers and knowing the fact that the above
bdrv_inc_in_flight didn't cause a problem, I think removing the NULL check is
fine.

I'll send a patch.

Thanks.

Fam

> 
> > +        bdrv_is_sg(bs)) {
> > +        goto early_exit;
> > +    }
> 
> Kevin