Re: [PATCH Kernel v22 0/8] Add UAPIs to support migration for VFIO devices

[Kirti Wankhede <kwankhede@nvidia.com>]

On 5/21/2020 12:34 PM, Yan Zhao wrote:
> On Thu, May 21, 2020 at 12:39:48PM +0530, Kirti Wankhede wrote:
>>
>>
>> On 5/21/2020 10:38 AM, Yan Zhao wrote:
>>> On Wed, May 20, 2020 at 10:46:12AM -0600, Alex Williamson wrote:
>>>> On Wed, 20 May 2020 19:10:07 +0530
>>>> Kirti Wankhede <kwankhede@nvidia.com> wrote:
>>>>
>>>>> On 5/20/2020 8:25 AM, Yan Zhao wrote:
>>>>>> On Tue, May 19, 2020 at 10:58:04AM -0600, Alex Williamson wrote:
>>>>>>> Hi folks,
>>>>>>>
>>>>>>> My impression is that we're getting pretty close to a workable
>>>>>>> implementation here with v22 plus respins of patches 5, 6, and 8.  We
>>>>>>> also have a matching QEMU series and a proposal for a new i40e
>>>>>>> consumer, as well as I assume GVT-g updates happening internally at
>>>>>>> Intel.  I expect all of the latter needs further review and discussion,
>>>>>>> but we should be at the point where we can validate these proposed
>>>>>>> kernel interfaces.  Therefore I'd like to make a call for reviews so
>>>>>>> that we can get this wrapped up for the v5.8 merge window.  I know
>>>>>>> Connie has some outstanding documentation comments and I'd like to make
>>>>>>> sure everyone has an opportunity to check that their comments have been
>>>>>>> addressed and we don't discover any new blocking issues.  Please send
>>>>>>> your Acked-by/Reviewed-by/Tested-by tags if you're satisfied with this
>>>>>>> interface and implementation.  Thanks!
>>>>>> hi Alex and Kirti,
>>>>>> after porting to qemu v22 and kernel v22, it is found out that
>>>>>> it can not even pass basic live migration test with error like
>>>>>>
>>>>>> "Failed to get dirty bitmap for iova: 0xca000 size: 0x3000 err: 22"
>>>>>
>>>>> Thanks for testing Yan.
>>>>> I think last moment change in below cause this failure
>>>>>
>>>>> https://lore.kernel.org/kvm/1589871178-8282-1-git-send-email-kwankhede@nvidia.com/
>>>>>
>>>>>    > 	if (dma->iova > iova + size)
>>>>>    > 		break;
>>>>>
>>>>> Surprisingly with my basic testing with 2G sys mem QEMU didn't raise
>>>>> abort on g_free, but I do hit this with large sys mem.
>>>>> With above change, that function iterated through next vfio_dma as well.
>>>>> Check should be as below:
>>>>>
>>>>> -               if (dma->iova > iova + size)
>>>>> +               if (dma->iova > iova + size -1)
>>>>
>>>>
>>>> Or just:
>>>>
>>>> 	if (dma->iova >= iova + size)
>>>>
>>>> Thanks,
>>>> Alex
>>>>
>>>>
>>>>>                            break;
>>>>>
>>>>> Another fix is in QEMU.
>>>>> https://lists.gnu.org/archive/html/qemu-devel/2020-05/msg04751.html
>>>>>
>>>>>    > > +        range->bitmap.size = ROUND_UP(pages, 64) / 8;
>>>>>    >
>>>>>    > ROUND_UP(npages/8, sizeof(u64))?
>>>>>    >
>>>>>
>>>>> If npages < 8, npages/8 is 0 and ROUND_UP(0, 8) returns 0.
>>>>>
>>>>> Changing it as below
>>>>>
>>>>> -        range->bitmap.size = ROUND_UP(pages / 8, sizeof(uint64_t));
>>>>> +        range->bitmap.size = ROUND_UP(pages, sizeof(__u64) *
>>>>> BITS_PER_BYTE) /
>>>>> +                             BITS_PER_BYTE;
>>>>>
>>>>> I'm updating patches with these fixes and Cornelia's suggestion soon.
>>>>>
>>>>> Due to short of time I may not be able to address all the concerns
>>>>> raised on previous versions of QEMU, I'm trying make QEMU side code
>>>>> available for testing for others with latest kernel changes. Don't
>>>>> worry, I will revisit comments on QEMU patches. Right now first priority
>>>>> is to test kernel UAPI and prepare kernel patches for 5.8
>>>>>
>>>>
>>> hi Kirti
>>> by updating kernel/qemu to v23, still met below two types of errors.
>>> just basic migration test.
>>> (the guest VM size is 2G for all reported bugs).
>>>
>>> "Failed to get dirty bitmap for iova: 0xfe011000 size: 0x3fb0 err: 22"
>>>
>>
>> size doesn't look correct here, below check should be failing.
>>   range.size & (iommu_pgsize - 1)
>>
>>> or
>>>
>>> "qemu-system-x86_64-lm: vfio_load_state: Error allocating buffer
>>> qemu-system-x86_64-lm: error while loading state section id 49(vfio)
>>> qemu-system-x86_64-lm: load of migration failed: Cannot allocate memory"
>>>
>>>
>>
>> Above error is from:
>>          buf = g_try_malloc0(data_size);
>>          if (!buf) {
>>              error_report("%s: Error allocating buffer ", __func__);
>>              return -ENOMEM;
>>          }
>>
>> Seems you are running out of memory?
>>
> no. my host memory is about 60G.
> just migrate with command "migrate -d xxx" without speed limit.
> FYI.
> 

Probably you will have to figure out why g_try_malloc0() is failing. 
what is data_size when it fails?

Thanks,
Kirti