Patchwork qemu-kvm aborts - vhost_dev_unassign_memory: Assertion `to >= 0' failed.

login
register
mail settings
Submitter Michael S. Tsirkin
Date Aug. 3, 2011, 12:24 p.m.
Message ID <20110803122441.GB10538@redhat.com>
Download mbox | patch
Permalink /patch/108142/
State New
Headers show

Comments

Michael S. Tsirkin - Aug. 3, 2011, 12:24 p.m.
On Wed, Aug 03, 2011 at 02:48:05PM +0300, Avi Kivity wrote:
> On 08/01/2011 08:44 PM, David Ahern wrote:
> >qemu-kvm.git as of:
> >
> >commit dacdc4b10bafbb21120e1c24a9665444768ef999
> >Merge: 7b69d4f 0af4922
> >Author: Avi Kivity<avi@redhat.com>
> >Date:   Sun Jul 31 11:42:26 2011 +0300
> >
> >     Merge branch 'upstream-merge' into next
> >
> >is aborting with the error:
> >
> >qemu-kvm: qemu-kvm.git/hw/vhost.c:123: vhost_dev_unassign_memory:
> >Assertion `to>= 0' failed.
> >Aborted
> >
> 
> It's a bug in vhost:
> 
> /* Assign/unassign. Keep an unsorted array of non-overlapping
>  * memory regions in dev->mem. */
> static void vhost_dev_unassign_memory(struct vhost_dev *dev,
>                                       uint64_t start_addr,
>                                       uint64_t size)
> {
>     int from, to, n = dev->mem->nregions;
>     /* Track overlapping/split regions for sanity checking. */
>     int overlap_start = 0, overlap_end = 0, overlap_middle = 0, split = 0;
> 
>     for (from = 0, to = 0; from < n; ++from, ++to) {
>         struct vhost_memory_region *reg = dev->mem->regions + to;
>         uint64_t reglast;
>         uint64_t memlast;
>         uint64_t change;
> 
>         /* clone old region */
>         if (to != from) {
>             memcpy(reg, dev->mem->regions + from, sizeof *reg);
>         }
> 
>         /* No overlap is simple */
>         if (!ranges_overlap(reg->guest_phys_addr, reg->memory_size,
>                             start_addr, size)) {
>             continue;
>         }
> 
>         /* Split only happens if supplied region
>          * is in the middle of an existing one. Thus it can not
>          * overlap with any other existing region. */
>         assert(!split);
> 
>         reglast = range_get_last(reg->guest_phys_addr, reg->memory_size);
>         memlast = range_get_last(start_addr, size);
> 
>         /* Remove whole region */
>         if (start_addr <= reg->guest_phys_addr && memlast >= reglast) {
>             --dev->mem->nregions;
>             --to;
>             assert(to >= 0);
>             ++overlap_middle;
>             continue;
>         }
> 
> 
> We're removing the first region, and 'to' goes negative.  Michael?

Yes, that assert is wrong.

--->
Subject: vhost: remove an incorrect assert

The 'to' can go negative when the first region gets removed
(it gets incremented by to 0 immediately afterward), which
makes the assertion fail. Nothing breaks if
to < 0 here so just remove the assert.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

----
David S. Ahern - Aug. 3, 2011, 1:55 p.m.
On 08/03/2011 06:24 AM, Michael S. Tsirkin wrote:
> On Wed, Aug 03, 2011 at 02:48:05PM +0300, Avi Kivity wrote:
>> On 08/01/2011 08:44 PM, David Ahern wrote:
>>> qemu-kvm.git as of:
>>>
>>> commit dacdc4b10bafbb21120e1c24a9665444768ef999
>>> Merge: 7b69d4f 0af4922
>>> Author: Avi Kivity<avi@redhat.com>
>>> Date:   Sun Jul 31 11:42:26 2011 +0300
>>>
>>>     Merge branch 'upstream-merge' into next
>>>
>>> is aborting with the error:
>>>
>>> qemu-kvm: qemu-kvm.git/hw/vhost.c:123: vhost_dev_unassign_memory:
>>> Assertion `to>= 0' failed.
>>> Aborted
>>>
>>
>> It's a bug in vhost:
>>
>> /* Assign/unassign. Keep an unsorted array of non-overlapping
>>  * memory regions in dev->mem. */
>> static void vhost_dev_unassign_memory(struct vhost_dev *dev,
>>                                       uint64_t start_addr,
>>                                       uint64_t size)
>> {
>>     int from, to, n = dev->mem->nregions;
>>     /* Track overlapping/split regions for sanity checking. */
>>     int overlap_start = 0, overlap_end = 0, overlap_middle = 0, split = 0;
>>
>>     for (from = 0, to = 0; from < n; ++from, ++to) {
>>         struct vhost_memory_region *reg = dev->mem->regions + to;
>>         uint64_t reglast;
>>         uint64_t memlast;
>>         uint64_t change;
>>
>>         /* clone old region */
>>         if (to != from) {
>>             memcpy(reg, dev->mem->regions + from, sizeof *reg);
>>         }
>>
>>         /* No overlap is simple */
>>         if (!ranges_overlap(reg->guest_phys_addr, reg->memory_size,
>>                             start_addr, size)) {
>>             continue;
>>         }
>>
>>         /* Split only happens if supplied region
>>          * is in the middle of an existing one. Thus it can not
>>          * overlap with any other existing region. */
>>         assert(!split);
>>
>>         reglast = range_get_last(reg->guest_phys_addr, reg->memory_size);
>>         memlast = range_get_last(start_addr, size);
>>
>>         /* Remove whole region */
>>         if (start_addr <= reg->guest_phys_addr && memlast >= reglast) {
>>             --dev->mem->nregions;
>>             --to;
>>             assert(to >= 0);
>>             ++overlap_middle;
>>             continue;
>>         }
>>
>>
>> We're removing the first region, and 'to' goes negative.  Michael?
> 
> Yes, that assert is wrong.
> 
> --->
> Subject: vhost: remove an incorrect assert
> 
> The 'to' can go negative when the first region gets removed
> (it gets incremented by to 0 immediately afterward), which
> makes the assertion fail. Nothing breaks if
> to < 0 here so just remove the assert.
> 
> Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
> 
> ----
> 
> diff --git a/hw/vhost.c b/hw/vhost.c
> index c3d8821..19e7255 100644
> --- a/hw/vhost.c
> +++ b/hw/vhost.c
> @@ -120,7 +120,6 @@ static void vhost_dev_unassign_memory(struct vhost_dev *dev,
>          if (start_addr <= reg->guest_phys_addr && memlast >= reglast) {
>              --dev->mem->nregions;
>              --to;
> -            assert(to >= 0);
>              ++overlap_middle;
>              continue;
>          }
> 

Removing the assert appears to work fine.
Tested-by: David Ahern <daahern@cisco.com>

David
Michael S. Tsirkin - Aug. 3, 2011, 3 p.m.
On Wed, Aug 03, 2011 at 07:55:47AM -0600, David Ahern wrote:
> Tested-by: David Ahern <daahern@cisco.com>
> 
> David

Applied, thanks very much.
David S. Ahern - Aug. 4, 2011, 6:48 p.m.
On 08/03/2011 09:00 AM, Michael S. Tsirkin wrote:
> On Wed, Aug 03, 2011 at 07:55:47AM -0600, David Ahern wrote:
>> Tested-by: David Ahern <daahern@cisco.com>
>>
>> David
> 
> Applied, thanks very much.
> 

I assume this will make 0.15 since it is a regression? haven't seen the
patch applied to that branch yet.

David
Michael S. Tsirkin - Aug. 4, 2011, 7:17 p.m.
On Thu, Aug 04, 2011 at 12:48:49PM -0600, David Ahern wrote:
> 
> 
> On 08/03/2011 09:00 AM, Michael S. Tsirkin wrote:
> > On Wed, Aug 03, 2011 at 07:55:47AM -0600, David Ahern wrote:
> >> Tested-by: David Ahern <daahern@cisco.com>
> >>
> >> David
> > 
> > Applied, thanks very much.
> > 
> 
> I assume this will make 0.15 since it is a regression?

How is this a regression in 0.15? I thought it's an old bug
exposed by the memory API which I expect is not going into 0.15.

> haven't seen the
> patch applied to that branch yet.
> 
> David

My branch will have to get merged in trunk first.
David S. Ahern - Aug. 4, 2011, 7:23 p.m.
On 08/04/2011 01:17 PM, Michael S. Tsirkin wrote:
> On Thu, Aug 04, 2011 at 12:48:49PM -0600, David Ahern wrote:
>>
>>
>> On 08/03/2011 09:00 AM, Michael S. Tsirkin wrote:
>>> On Wed, Aug 03, 2011 at 07:55:47AM -0600, David Ahern wrote:
>>>> Tested-by: David Ahern <daahern@cisco.com>
>>>>
>>>> David
>>>
>>> Applied, thanks very much.
>>>
>>
>> I assume this will make 0.15 since it is a regression?
> 
> How is this a regression in 0.15? I thought it's an old bug
> exposed by the memory API which I expect is not going into 0.15.

Is qemu-kvm.git not tagged as 0.15 rc?

David

> 
>> haven't seen the
>> patch applied to that branch yet.
>>
>> David
> 
> My branch will have to get merged in trunk first.
>

Patch

diff --git a/hw/vhost.c b/hw/vhost.c
index c3d8821..19e7255 100644
--- a/hw/vhost.c
+++ b/hw/vhost.c
@@ -120,7 +120,6 @@  static void vhost_dev_unassign_memory(struct vhost_dev *dev,
         if (start_addr <= reg->guest_phys_addr && memlast >= reglast) {
             --dev->mem->nregions;
             --to;
-            assert(to >= 0);
             ++overlap_middle;
             continue;
         }