Re: [PATCH V9 09/15] mmc: core: Add parameter use_blk_mq

[Linus Walleij]

On Wed, Sep 27, 2017 at 2:02 PM, Adrian Hunter <adrian.hunter@xxxxxxxxx> wrote:
> On 27/09/17 02:42, Linus Walleij wrote:
>> On Fri, Sep 22, 2017 at 2:36 PM, Adrian Hunter <adrian.hunter@xxxxxxxxx> wrote:
>>
>>> Until mmc has blk-mq support fully implemented and tested, add a
>>> parameter use_blk_mq, default to false unless config option MMC_MQ_DEFAULT
>>> is selected.
>>>
>>> Signed-off-by: Adrian Hunter <adrian.hunter@xxxxxxxxx>
>>
>>> +config MMC_MQ_DEFAULT
>>> +       bool "MMC: use blk-mq I/O path by default"
>>> +       depends on MMC && BLOCK
>>
>> I would say:
>> default y
>>
>> Why not. SCSI is starting to enable this by default so IMO we should
>
> SCSI didn't manage it yet.
>
>> not take the
>> intermediate step of having this as optional. Otherwise it never gets tested.
>
> The argument that we don't have to take any responsibility for getting
> things tested is a poor one.
>
> Anyway, you can always send a patch later to change the default, so why the
> hurry to do it now?

I think it should be the default so we smoke out bugs by throwing it at
users during the -rc phase and the non-mq path should be the fallback.

The -rc phase is for finding problems like this IMO.

It is partly a personality trait I guess, I'm not very cautious in general,
for good and for bad.

>>> +module_param_named(use_blk_mq, mmc_use_blk_mq, bool, S_IWUSR | S_IRUGO);
>>
>> Are people really modprobing this so it needs to be a module parameter?
>
> Irrespective of modprobe, the parameter can be changed without re-compiling.
>
>>
>> Maybe I'm the only developer stupid enough to just recompile and reboot
>
> Just change the parameter and unbind and rebind the host controller.

Ah clever. I don't do such things, I guess I should try it out.

Yours,
Linus Walleij