Re: [PATCH] UBI: Block: Add blk-mq support

[Richard Weinberger <richard@nod.at>]

[-- Attachment #1: Type: text/plain, Size: 1470 bytes --]

Am 02.11.2014 um 23:27 schrieb Ezequiel Garcia:
> Wow. Where did you run this and on top of what storage device?

nandsim, to make sure that the MTD is not our bottleneck.

> I'm still interested in the memory footprint, UBI is already heavy enough.

AFAICT blk-mq allocates one struct ubiblock_pdu per device.
As all IO is done via scattergather the memory footprint should be good.
But I'm sure Christoph can tell you the glory details.

>>> I kind of like the negative diffstat, but the code doesn't look cleaner
>>> or simpler.
>>>
>>> In other words, we need a good reason before we agree on making this
>>> "zen style" driver more complex.
>>
>> After reading my patch again I think we could move ubiblock_read_to_sg()
>> to kapi.c or io.c. It is rather generic and maybe we can tun more UBI users to
>> scattergather such that less vmalloc()s are needed.
>>
>> This would also make the diffstat nicer...
>>
> 
> Yes, any additional effort to make the current patch any simpler would
> be great. In its current form it seems rather cumbersome to me.

Why cumbersome? It changes the way the driver works as blk-mq works differently.
If you look at other blk-mq conversion patches you'll notice that they all change
a lot.

> If you can re-submit something better and put a more verbose commit log,
> I'd really appreciate it :)

First I wait for a review. I'm not sure whether I'm used blk-ml correctly.

Thanks,
//richard



[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 836 bytes --]