Message ID | 1427989307-1805-1-git-send-email-dbaryshkov@gmail.com (mailing list archive) |
---|---|
State | New, archived |
Headers | show |
Just like has already been pointed out on the list once today, we need much better commit descriptions than this. The code has been perfectly happy to auto-detect the IRQ for over ten years. Why do we suddenly need to change it now? What's the justification for this change. Or is this just a case of "because we can" and you happen to have a different opinion on how stuff should be done from how it's been successfully done for the last ten years? On Thu, Apr 02, 2015 at 06:41:45PM +0300, Dmitry Eremin-Solenikov wrote: > To allow boards to specify the irq that is used by UCB1x00 chip, add irq > field to the platform data structure. > > Signed-off-by: Dmitry Eremin-Solenikov <dbaryshkov@gmail.com> > --- > include/linux/mfd/ucb1x00.h | 1 + > 1 file changed, 1 insertion(+) > > diff --git a/include/linux/mfd/ucb1x00.h b/include/linux/mfd/ucb1x00.h > index e1345ff..9a2dacb 100644 > --- a/include/linux/mfd/ucb1x00.h > +++ b/include/linux/mfd/ucb1x00.h > @@ -118,6 +118,7 @@ struct ucb1x00_plat_data { > unsigned irq_base; > int gpio_base; > unsigned can_wakeup; > + int irq; > }; > > struct ucb1x00 { > -- > 2.1.4 > > > _______________________________________________ > linux-arm-kernel mailing list > linux-arm-kernel@lists.infradead.org > http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
Hello, 2015-04-02 19:02 GMT+03:00 Russell King - ARM Linux <linux@arm.linux.org.uk>: > Just like has already been pointed out on the list once today, we need > much better commit descriptions than this. > > The code has been perfectly happy to auto-detect the IRQ for over ten > years. Why do we suddenly need to change it now? What's the > justification for this change. Because using static configuration would allow me to drop a nice piece of code. Because it would allow later to use dts to describe ucb1x00 configuration. Because I will no longer have to think whether ucb1x00I h driver barfing on IRQ is a problem of a chip, of an IRQ or just me having touched wrong GPIO line at wrong time with a scope probe. Yes, I didn't put all these causes to the commit description. Do you really want it? > > Or is this just a case of "because we can" and you happen to have a > different opinion on how stuff should be done from how it's been > successfully done for the last ten years? Not only "because we can", but "because we should". In my humble opinion. It looks like the whole story of me touching sa1100 is like fighting with 'it was done so for ages' windmills. Yes, I have different opinions sometimes. I would like for sa1100 to converge to other platforms. PXA, especially. Why? Because I like several small devices sitting on my table. PXA is now slowly moving towards dts, cleaner drivers and cleaner interfaces. We should have done this ages ago. Nobody had time and interest. SA-11x0 had even older drivers. Older implementations. Older ideas. Yes, I have a different opinion at this place. I'd like for sa1100 to live in a contemporary world. I'd like to have my devices work with the rest of Linux subsystems. Should I just fork sa1100 subarch, make it work with devices I have at hand and submit it as a 'new' subarch? Or we can have an evolution-like process that will make sa1100 live in style with the rest of the Linux kernel. > > On Thu, Apr 02, 2015 at 06:41:45PM +0300, Dmitry Eremin-Solenikov wrote: >> To allow boards to specify the irq that is used by UCB1x00 chip, add irq >> field to the platform data structure. >> >> Signed-off-by: Dmitry Eremin-Solenikov <dbaryshkov@gmail.com> >> --- >> include/linux/mfd/ucb1x00.h | 1 + >> 1 file changed, 1 insertion(+) >> >> diff --git a/include/linux/mfd/ucb1x00.h b/include/linux/mfd/ucb1x00.h >> index e1345ff..9a2dacb 100644 >> --- a/include/linux/mfd/ucb1x00.h >> +++ b/include/linux/mfd/ucb1x00.h >> @@ -118,6 +118,7 @@ struct ucb1x00_plat_data { >> unsigned irq_base; >> int gpio_base; >> unsigned can_wakeup; >> + int irq; >> }; >> >> struct ucb1x00 { >> -- >> 2.1.4 >> >> >> _______________________________________________ >> linux-arm-kernel mailing list >> linux-arm-kernel@lists.infradead.org >> http://lists.infradead.org/mailman/listinfo/linux-arm-kernel > > -- > FTTC broadband for 0.8mile line: currently at 10.5Mbps down 400kbps up > according to speedtest.net.
On Thu, Apr 02, 2015 at 07:27:24PM +0300, Dmitry Eremin-Solenikov wrote: > Because using static configuration would allow me to drop a nice piece > of code. Because it would allow later to use dts to describe ucb1x00 > configuration. Because I will no longer have to think whether ucb1x00I h > driver barfing on IRQ is a problem of a chip, of an IRQ or just me having > touched wrong GPIO line at wrong time with a scope probe. Yes, I didn't > put all these causes to the commit description. Do you really want it? What I want is a decent commit description, one which isn't airy fairy and looks like it's just mindless churn. That's all. It's what every other kernel developer wants to know. This is precisely what the ARM ecosystem has been soo bad at for decades. It matters. It's one of the things what pisses Linus off. What Linus sees from ARM people is lots and lots of apparently pointless churn which he doesn't understand - and part of that problem is that ARM people do _not_ justify in their commit messages why a change is necessary. So please, improve the commit messages and justify the commits. It really doesn't help that all that was posted was three patches and no covering message which could either explain what the three patches were trying to achieve. Like most people do. (And some people take that to an extreme, posting a cover message for just one patch.)
On Thu, Apr 02, 2015 at 07:27:24PM +0300, Dmitry Eremin-Solenikov wrote: > Not only "because we can", but "because we should". In my humble > opinion. > > It looks like the whole story of me touching sa1100 is like fighting with > 'it was done so for ages' windmills. Yes, I have different opinions sometimes. Right, and what you're forgetting is that others may have different opinions. Like me - and as I'm the sa1100 and ucb1x00 maintainer, if you have a different opinion, you need to convince me that your solution is the right thing to do. Had I thought that passing an IRQ through platform data would be a good thing, then I would've done it from the outset. On principle, I don't like the idea of passing driver resources through platform data, I find it abhorrent. However, there isn't really any good solution here other than platform data or auto-detecting it, and auto-detecting it is, in my opinion, the lesser of the evils. > I would like for sa1100 to converge to other platforms. PXA, especially. > Why? Because I like several small devices sitting on my table. PXA > is now slowly moving towards dts, cleaner drivers and cleaner interfaces. Right, so when the UCB1x00 gets a node in DT, the driver can parse the interrupt from DT, without needing it in platform data - and if it finds that it's in DT, then there's no need to auto-probe it. I don't see the point of adding it to platform data only to later have to change it /again/ when SA11x0 moves to DT.
diff --git a/include/linux/mfd/ucb1x00.h b/include/linux/mfd/ucb1x00.h index e1345ff..9a2dacb 100644 --- a/include/linux/mfd/ucb1x00.h +++ b/include/linux/mfd/ucb1x00.h @@ -118,6 +118,7 @@ struct ucb1x00_plat_data { unsigned irq_base; int gpio_base; unsigned can_wakeup; + int irq; }; struct ucb1x00 {
To allow boards to specify the irq that is used by UCB1x00 chip, add irq field to the platform data structure. Signed-off-by: Dmitry Eremin-Solenikov <dbaryshkov@gmail.com> --- include/linux/mfd/ucb1x00.h | 1 + 1 file changed, 1 insertion(+)