[illumos-Developer] [review] 1147 Integrate the virtio-block driver
Alexey Zaytsev
alexey.zaytsev at gmail.com
Mon Jul 4 11:53:39 PDT 2011
On Mon, Jul 4, 2011 at 21:53, Gordon Ross <gordon.w.ross at gmail.com> wrote:
> Are there any design notes on this stuff anywhere? (i.e. NetBSD?)
There is a "specification draft" that explains the device operation
here: http://ozlabs.org/~rusty/virtio-spec
I would not call it a specification, but it explains the operation well enough.
There is also the "virtio-paper" in the same directory. A bit outdated
publication, but it is also good for understanding the device
operation.
> I'm a little surprised to see so much dma_sync stuff in vioblk.c.
This is what Garrett pointed out. We don't need to do syncs, or even
the mappings, on the data buffers we get from blkdev. I've integrated
the changes Garrett made some time ago to avoid this. But I've
discovered an unrelated problem, so the update is delayed.
> Can more of that be handled in the common virtio code?
No, as the upper layers could be doing their own syncs.
> How similar is this to the NetBSD code? What parts are new?
You can check the NetBSD code here:
https://gitorious.org/virtio-drivers/virtio-drivers/trees/master/nbsd_code
Not much left. The queue allocation/free code stayed mostly the same,
but that's the only major thing that I see.
Vioblk is Dmitrys code, and it seems that there are no NetBSD parts.
The NetBSD driver is work in progress as well, and it seems to have
stalled since last year.
>
> Thanks,
> Gordon
>
> On Sun, Jul 3, 2011 at 3:54 AM, Alexey Zaytsev <alexey.zaytsev at gmail.com> wrote:
>> On Sun, Jul 3, 2011 at 11:42, Garrett D'Amore <garrett at damore.org> wrote:
>>> This version is still doing its own dma mapping, doing double mapping. Please test the other version I gave you, and if it works, submit that for review. I cannot sign off on this the way it is written now, as it is not using the blkdev framework the way it was intended.
>>>
>>
>> Right, forgot about that patch
>>
>> _______________________________________________
>> Developer mailing list
>> Developer at lists.illumos.org
>> http://lists.illumos.org/m/listinfo/developer
>>
>
More information about the Developer
mailing list