Re: vhost changes (batched) in linux-next after 12/13 trigger random crashes in KVM guests after reboot

[Christian Borntraeger <borntraeger@de.ibm.com>]

On 07.02.20 08:58, Michael S. Tsirkin wrote:
> On Fri, Feb 07, 2020 at 08:47:14AM +0100, Christian Borntraeger wrote:
>> Also adding Cornelia.
>>
>>
>> On 06.02.20 23:17, Michael S. Tsirkin wrote:
>>> On Thu, Feb 06, 2020 at 04:12:21PM +0100, Christian Borntraeger wrote:
>>>>
>>>>
>>>> On 06.02.20 15:22, eperezma@redhat.com wrote:
>>>>> Hi Christian.
>>>>>
>>>>> Could you try this patch on top of ("38ced0208491 vhost: use batched version by default")?
>>>>>
>>>>> It will not solve your first random crash but it should help with the lost of network connectivity.
>>>>>
>>>>> Please let me know how does it goes.
>>>>
>>>>
>>>> 38ced0208491 + this seem to be ok.
>>>>
>>>> Not sure if you can make out anything of this (and the previous git bisect log)
>>>
>>> Yes it does - that this is just bad split-up of patches, and there's
>>> still a real bug that caused worse crashes :)
>>>
>>> So I just pushed batch-v4.
>>> I expect that will fail, and bisect to give us
>>>     vhost: batching fetches
>>> Can you try that please?
>>>
>>
>> yes.
>>
>> eccb852f1fe6bede630e2e4f1a121a81e34354ab is the first bad commit
>> commit eccb852f1fe6bede630e2e4f1a121a81e34354ab
>> Author: Michael S. Tsirkin <mst@redhat.com>
>> Date:   Mon Oct 7 06:11:18 2019 -0400
>>
>>     vhost: batching fetches
>>     
>>     With this patch applied, new and old code perform identically.
>>     
>>     Lots of extra optimizations are now possible, e.g.
>>     we can fetch multiple heads with copy_from/to_user now.
>>     We can get rid of maintaining the log array.  Etc etc.
>>     
>>     Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
>>
>>  drivers/vhost/test.c  |  2 +-
>>  drivers/vhost/vhost.c | 39 ++++++++++++++++++++++++++++++++++-----
>>  drivers/vhost/vhost.h |  4 +++-
>>  3 files changed, 38 insertions(+), 7 deletions(-)
>>
> 
> 
> And the symptom is still the same - random crashes
> after a bit of traffic, right?

random guest crashes after a reboot of the guests. As if vhost would still
write into now stale buffers.