On 28/05/2020 15:58, Russell King - ARM Linux admin wrote:
On Thu, May 28, 2020 at 03:46:04PM +0200, Linus Walleij wrote:
On Wed, May 27, 2020 at 4:18 PM Russell King - ARM Linux admin linux@armlinux.org.uk wrote:
On Wed, May 27, 2020 at 04:07:58PM +0200, Linus Walleij wrote:
We provided the right semantics on open drain lines being by definition output but incidentally the irq set up function would only allow IRQs on lines that were "not output".
Fix the semantics to allow output open drain lines to be used for IRQs.
Reported-by: Hans Verkuil hverkuil@xs4all.nl Cc: Russell King linux@armlinux.org.uk Cc: stable@vger.kernel.org Fixes: 256efaea1fdc ("gpiolib: fix up emulated open drain outputs")
As I've pointed out in the reporting thread, I don't think it can be justified as a regression - it's a bug in its own right that has been discovered by unifying the gpiolib semantics, since the cec-gpio code will fail on hardware that can provide real open-drain outputs irrespective of that commit.
So, you're really fixing a deeper problem that was never discovered until gpiolib's semantics were fixed to be more uniform.
You're right, I was thinking of Fixes: as more of a mechanical instruction to the stable kernel maintainers administrative machinery.
I will use the other way to signal to stable where to apply this.
I think it makes sense to apply this patch to stable kernels prior to the commit mentioned in the Fixes tag - but how far back is a good question. Certainly to the point that we ended up with code relying on this behaviour (so when cec-gpio was introduced?)
For the record, cec-gpio started to rely on this behavior in kernel 5.3. There is no need to go back to pre-5.3 kernels.
Regards,
Hans