Re: [Qemu-devel] [PATCH v2 13/17] vmdk: Implement .bdrv_co_pwritev() interface

[Pavel Butsykin <pbutsykin@virtuozzo.com>]

On 29.04.2016 12:49, Kevin Wolf wrote:
> Am 29.04.2016 um 10:49 hat Pavel Butsykin geschrieben:
>> On 29.04.2016 06:08, Fam Zheng wrote:
>>> On Thu, 04/28 15:16, Kevin Wolf wrote:
>>>>   static int vmdk_write_compressed(BlockDriverState *bs,
>>>>                                    int64_t sector_num,
>>>>                                    const uint8_t *buf,
>>>>                                    int nb_sectors)
>>>>   {
>>>>       BDRVVmdkState *s = bs->opaque;
>>>> +
>>>>       if (s->num_extents == 1 && s->extents[0].compressed) {
>>>> -        return vmdk_write(bs, sector_num, buf, nb_sectors, false, false);
>>>> +        Coroutine *co;
>>>> +        AioContext *aio_context = bdrv_get_aio_context(bs);
>>>> +        VmdkWriteCompressedCo data = {
>>>> +            .bs         = bs,
>>>> +            .sector_num = sector_num,
>>>> +            .buf        = buf,
>>>> +            .nb_sectors = nb_sectors,
>>>> +            .ret        = -EINPROGRESS,
>>>> +        };
>>>> +        co = qemu_coroutine_create(vmdk_co_write_compressed);
>>>> +        qemu_coroutine_enter(co, &data);
>>>> +        while (data.ret == -EINPROGRESS) {
>>>> +            aio_poll(aio_context, true);
>>>> +        }
>>>> +        return data.ret;
>>>
>>> Don't you have a plan to make the creation of coroutine for compressed write in
>>> in block layer?  Or will bdrv_co_pwritev gain a "compressed" flag
>>> (BDRV_REQ_COMPRESSED) in the future?
>>>
>> Actually, I already have these patches as part of the issue of backup
>> compression, hope to send today.
>
> Sounds good. Do you use a new flag for the normal write functions or just
> update the existing .bdrv_write_compressed function?
>
No, I just added a new interface .bdrv_co_write_compressed and removed
the old. As for compatibility with the old interface in the
bdrv_write_compressed will create a new coroutine and called
bdrv_co_write_compressed, how you did it in the vmdk(only now is in
io.c). Maybe it's not the best idea, but it can be discussed.

> Also, please base your series on this one to avoid merge conflicts in
> vmdk.c. I'm going to apply the series to block-next now as it's fully
> reveiwed. I guess you'll effectively need to revert what this patch does
> to vmdk_write_compressed() in favour of a block/io.c solution.
>
Yes, patches will be based on this series, I specifically made sure of it :)

> Kevin
>