[PATCH v2] vfio: do not coalesce DMA mappings

Nipun Gupta nipun.gupta at amd.com
Wed May 10 14:58:09 CEST 2023



On 4/24/2023 8:52 PM, David Marchand wrote:
> 
> Hello Anatoly,
> 
> On Wed, Apr 5, 2023 at 4:17 PM Burakov, Anatoly
> <anatoly.burakov at intel.com> wrote:
>>>> Could you please provide some steps to reproduce the hotplug issue
>>>> you're having? It would be great to have a test case for this patchset
>>>> to put it in context.
>>>
>>> I am working on CDX bus
>>> (http://patchwork.dpdk.org/project/dpdk/patch/20230124140746.594066-2-nipun.gupta@amd.com/) and trying out some cases for plug/unplug.
>>>
>>> The test is as follows:
>>>     # Run testpmd application
>>>     ./dpdk-testpmd -c 0x3 -- -i --nb-cores=1
>>>
>>>     # Bind to VFIO
>>>     echo "vfio-cdx" >  /sys/bus/cdx/devices/cdx-00\:00/driver_override
>>>     echo "cdx-00:00" > /sys/bus/cdx/drivers_probe
>>>
>>>     # Plug a device
>>>     testpmd> port attach cdx:cdx-00:00
>>>
>>>     #quit testpmd
>>>     testpmd> quit
>>>
>>> This gave error at testpmd exit that memory cannot be freed. On
>>> debugging I updated this code and seems it should be seen with any of
>>> the device.
>>>
>>> I see similar test case (without quit) mentioned
>>> https://doc.dpdk.org/dts/test_plans/hotplug_test_plan.html, but the
>>> difference is that it is with igb_uio and issue is being observed with
>>> VFIO.
>>>
>>> Please note the device/bus mentioned in the commands is not yet
>>> upstreamed in DPDK, but patches would be sent out soon.
>>>
>>> Thanks,
>>> Nipun
>>>
>>
>> Thanks, I can reproduce this issue with regular devices too (run testpmd
>> with no devices, bind a NIC to VFIO, attach it, then quit). You're
>> correct in that since the initial mapping was done with mapping large
>> contiguous zones (such as when mempools are created before attach), any
>> subsequent freeing of memory will cause these errors to happen.
>>
>> I don't think this can be fixed by anything other than not doing the
>> contiguous mapping thing, so provisionally, I think this patch should be
>> accepted. I'll play around with it some more and get back to you :)
> 
> Can we conclude on this topic?
> It is best we merge this kind of change the sooner possible for a release.

Hi Anatoly,
	Can you kindly update on this?

Thanks,
Nipun

> 
> Thanks.
> 
> 
> --
> David Marchand
> 


More information about the dev mailing list