Re: Regression: usb serial gadget on sama5d3 broken

[Alexandre Belloni]

On 17/10/2016 at 19:50:48 +0200, Peter Rosin wrote :
> On 2016-10-17 16:54, Nicolas Ferre wrote:
> > Le 17/10/2016 à 14:53, Peter Rosin a écrit :
> >> Hi!
> >>
> >> I'm suffering from a regression while using the usb gadget port on the
> >> sama5d3 to get terminal access to the device in question (CONFIG_USB_G_SERIAL).
> >>
> >> I get this message when I try to connect:
> >> udc: ep: Invalid setup request: 02.01 v0000 i0081 l0, halting endpoint...
> >>
> >> A bisect blames commit v4.7-rc1-21-gc32b5bcfa3c4 "ARM: dts: at91: Fix
> >> USB endpoint nodes".
> >>
> >> And indeed, reverting that commit on top of v4.9-rc1 fixes things,
> >> although that doesn't look like the best of fixes...
> >>
> >> BTW, the bisect was extremely painful since v4.7-rc1 seemed broken
> >> somewhere in the overlayfs area. I hope I will never ever need to bisect
> >> in the v4.6..v4.7 area again. This was the second time, the first time
> >> I was chasing a gpio interrupt bug, but I never found out what was wrong
> >> and stopped looking when v4.9-rc1 turned out to be ok even though v4.8
> >> was bad, it was just too painful to look for things that already seemed
> >> fixed.
> > 
> > I guess that you are referring to the regression listed here:
> > https://www.mail-archive.com/linux-kernel@xxxxxxxxxxxxxxx/msg1239220.html
> > 
> > The patch is available and will hopefully land in an official kernel
> > soon (4.8.1) as said by Felipe and Greg.
> > 
> > Sorry for the inconvenience. Best regards,
> 
> Ok, I tried "usb: gadget: udc: atmel: fix endpoint name" and it fixes
> things for me too. But shouldn't the memory for the now dynamic name be
> allocated with devm_kasprintf instead of that plain kasprint? If you
> are pedantic...
> 

That wouldn't be a bad idea.

-- 
Alexandre Belloni, Free Electrons
Embedded Linux and Kernel engineering
http://free-electrons.com