Re: [PATCH v3] ata: use generic power management

[Jens Axboe]

On 8/18/20 6:03 AM, Vaibhav Gupta wrote:
> On Mon, Aug 17, 2020 at 07:06:03AM -0700, Jens Axboe wrote:
>> On 8/17/20 2:25 AM, Vaibhav Gupta wrote:
>>> Hello,
>>>
>>> I am working to upgrade power management framework support for PCI drivers, as
>>> my project under the Linux Kernel Mentorship Program.
>>>
>>> The ultimate goal is to completely remove the legacy framework. And for this we
>>> need to remove ".suspend" and ".resume" pointers from "struct pci_driver".
>>>
>>> The patch is doing the change for the same. The actual suspend() and
>>> resume() callbacks are defined in drivers/ata/libata-core.c and are exported.
>>> Which are then used by other files.
>>>
>>> Thus removing those pointers, included 54 files in this patch. Although, the
>>> actual changes are done in few files only. The changes should work fine as I
>>> have done similar changes for other drivers too which made their way into the
>>> kernel. Still, tests on few ata devices are necessary.
>>
>> Didn't we agree to split this up into separate patches??
>>
>>> I tried but unfortunately, I couldn't find or arrange devices to test upon. I
>>> have added the authors of the previous commit(s) for respective drivers as
>>> recipients. It would be very helpful if someone can test it on a device.
>>
>> Does qemu support any of the affected drivers?
>>
> Hello Jens,
> Yes we discussed about splitting it and I will surely do it. I just thought that
> still it has not got tested and in that case keeping check on patch-series with
> 55+ patches may become somewhat messy. Moreover, to test any driver one will
> have to apply the patch for libata-core and then that of the driver.
> So, I sent this v3 in order to get some help for the testing purpose. Once it
> goes successful, I can split it and send it for the submission.
> 
> I am not sure about qemu, I haven't checked for it. Qemu will be appropriate
> for testing power management for .suspend() and .resume() callbacks?

Well, that's your homework then, it seems pretty ideal for testing that
kind of thing and way easier than getting your hands on hardware. It
won't have support for everything, but so far you are at 0 things
tested, so any piix and ahci testing would be a win as far as I'm
concerned.

-- 
Jens Axboe