On Mon, 7 Jun 2010, Esben Haabendal wrote: > On Mon, 2010-06-07 at 01:39 +0200, Thomas Gleixner wrote: > > On Fri, 4 Jun 2010, Esben Haabendal wrote: > > > > @@ -120,6 +124,10 @@ static int pca953x_gpio_direction_input(struct > > > gpio_chip *gc, unsigned off) > > > chip = container_of(gc, struct pca953x_chip, gpio_chip); > > > > > > reg_val = chip->reg_direction | (1u << off); > > > + > > > + if (reg_val == chip->reg_direction) > > > + return 0; > > > + > > > > This is an optimization of its own value. > > Yes, but is need to avoid doing I2C work from irq_chip map().
Still it should be a separate patch, that's all I said. It can be documented that it is also necessary to support powerpcs virq thing if at all. > > 3) it breaks the driver. See http://lkml.org/lkml/2010/6/6/177 for a > > detailed explanation > > I believe there is still a few things that needs to be discussed before > that is closed. Not with me. Period. > > 4) the virq/powerpc churn is horrible and I bet there are sane ways to > > solve this, but it leave this to the powerpc experts. > > Do you suggest that a seperate pca953x driver should be implemented for > powerpc? (I guess not). Or do you say that the who irq handling in > powerpc should be changed? > > There must be an acceptable way to extend pca953x.c for the powerpc virq > handling and get it accepted in the kernel. Yes, there certainly is. I just think that there are cleaner ways, that's why I defered that to the powerpc wizards. Thanks, tglx _______________________________________________ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
