Re: [PATCH] usb: gadget: fsl: Add FSL USB Gadget entry in platform device id

[Felipe Balbi]

Hi,

On Thu, Mar 20, 2014 at 09:02:52AM -0700, gregkh@xxxxxxxxxxxxxxxxxxx wrote:
> On Thu, Mar 20, 2014 at 03:01:56PM +0000, suresh.gupta@xxxxxxxxxxxxx wrote:
> > > > > > > @@ -2654,6 +2654,8 @@ static const struct platform_device_id
> > > > > fsl_udc_devtype[] = {
> > > > > > >  	}, {
> > > > > > >  		.name = "imx-udc-mx51",
> > > > > > >  	}, {
> > > > > > > +		.name = "fsl-usb2-udc",
> > > > > >
> > > > > > why aren't you just using chipidea ?
> > > > > > [SuresH] This is our legacy driver for all previous and existing
> > > > > > ppc socs. Many of our customers are already using this, and we
> > > > > > need to support them on this driver. We do have plans to shift to
> > > > > > chipidea, but after some time.
> > > > >
> > > > > cool, you already have plans, so we will see a new glue layer for
> > > > > v3.16 right ? Which means I don't need to take this patch either.
> > > > >
> > > > we do have plans, but in remote future. Right now, we need to support
> > > > customers on the present legacy driver. We'll phase out this driver
> > > > slowly when we integrate chipidea. At this time I would request you to
> > > > please accept this patch
> > > 
> > > Even if Felipe takes the patch, I'll reject it as you should be doing the
> > > correct thing here, and if it's accepted, it will never be changed...
> > 
> > Hi Greg, I agree that moving to the chipidea driver is the right thing to do.
> > However, does this mean that companies have to phase out their current legacy 
> > drivers as soon as a new controller specific driver is introduced in kernel ??
> 
> If their drivers aren't merged upstream, then yes they do, we can't have
> duplicate drivers controlling the same hardware blobs.
> 
> > Can't they decide their own schedule based on their own requirements. Our only 
> > concern is to keep supporting our customers till we move to the new driver. 
> 
> Your support issues / requirements is not any of our business, we just
> can't accept duplicate code, which I'm sure you can understand.
> 
> > I would really appreciate if you could accept this as this would give us 
> > some time to move to chipidea driver.
> 
> What is preventing you from doing this within a week or so?  Is it
> really that hard of a transistion?  If so, is this a problem with the
> chipidea driver or something else?

Greg, many other freescale SoCs are *already* using chipidea driver. I
suppose 1 week effort is overestimation, it shouldn't take much more
than a couple days.

If there are any bugs in chipidea, let's find them and fix them up. We
really cannot continue with code duplication in the tree, it's just
moronic to do so, it's an unnecessary maintenance burden, extra review
time, same fixes having to written for separate drivers, and so on.

-- 
balbi

Attachment: signature.asc
Description: Digital signature