Re: [PATCH] of/platform: Fix no irq domain found errors when populating interrupts

Re: [PATCH] of/platform: Fix no irq domain found errors when populating interrupts

[Paul Walmsley]

Hi Grant, Rob,

On Sun, 24 Nov 2013, Grant Likely wrote:

> On Fri, 22 Nov 2013 17:50:35 -0800, Tony Lindgren <tony@atomide.com> wrote:
> > * Tony Lindgren <tony@atomide.com> [131122 17:16]:
> > > * Tony Lindgren <tony@atomide.com> [131122 17:09]:
> > > > * Russell King - ARM Linux <linux@arm.linux.org.uk> [131122 16:56]:
> > > > > On Fri, Nov 22, 2013 at 04:43:35PM -0800, Tony Lindgren wrote:
> > > > > > +		/* See of_device_resource_notify for populating interrupts */
> > > > > > +		for (i = 0; i < num_irq; i++, res++) {
> > > > > > +			res->flags = IORESOURCE_IRQ;
> > > > > > +			res->start = -EPROBE_DEFER;
> > > > > > +			res->end = -EPROBE_DEFER;
> > > > > 
> > > > > NAK.  Definitely a bad idea to start introducing magic values other into
> > > > > resources.  Please don't do this.
> > > > 
> > > > Do you have any better ideas on how to sort out this issue then?
> > > 
> > > I guess we could allocate all the resources lazily here, I'll take a look
> > > at that.
> > 
> > Here's a version that allocates the resources lazily with the notifier.
> > Seems to boot, need to play with it a bit more though to make sure we're
> > not overwriting resources for any legacy devices.
> 
> Blurg. Using a notifier really feels like we don't have a good handle on
> a reasonable solution yet. Basically it means we're hooking into the
> driver core without /looking/ like we're hooking into the driver core. I
> don't think this is any better, but I don't have a better suggestion at
> the moment.   :-(

Unfortunately this patch, or something that accomplishes the same results, 
is somewhat high-priority for us on OMAP.  OMAP37xx got prematurely 
converted to DT booting, and now dynamic power management is broken:

    http://marc.info/?l=linux-arm-kernel&m=138658294830408&w=2

Tony writes that this patch is one of the two patches needed to get things 
working again:

    http://marc.info/?l=linux-arm-kernel&m=138660942506846&w=2

Is it possible to get this patch, or something similar, merged for 
v3.13-rc?

Once something like PM is broken, it's pretty easy for other broken 
patches to make it into the tree, since it becomes very difficult to test 
without turning into a maintainer denial-of-service attack.


- Paul

Re: [PATCH] of/platform: Fix no irq domain found errors when populating interrupts

[Paul Walmsley]

Hi Grant, Rob,

On Tue, 10 Dec 2013, Paul Walmsley wrote:

> On Sun, 24 Nov 2013, Grant Likely wrote:
> 
> > On Fri, 22 Nov 2013 17:50:35 -0800, Tony Lindgren <tony@atomide.com> wrote:
> > > * Tony Lindgren <tony@atomide.com> [131122 17:16]:
> > > > * Tony Lindgren <tony@atomide.com> [131122 17:09]:
> > > > > * Russell King - ARM Linux <linux@arm.linux.org.uk> [131122 16:56]:
> > > > > > On Fri, Nov 22, 2013 at 04:43:35PM -0800, Tony Lindgren wrote:
> > > > > > > +		/* See of_device_resource_notify for populating interrupts */
> > > > > > > +		for (i = 0; i < num_irq; i++, res++) {
> > > > > > > +			res->flags = IORESOURCE_IRQ;
> > > > > > > +			res->start = -EPROBE_DEFER;
> > > > > > > +			res->end = -EPROBE_DEFER;
> > > > > > 
> > > > > > NAK.  Definitely a bad idea to start introducing magic values other into
> > > > > > resources.  Please don't do this.
> > > > > 
> > > > > Do you have any better ideas on how to sort out this issue then?
> > > > 
> > > > I guess we could allocate all the resources lazily here, I'll take a look
> > > > at that.
> > > 
> > > Here's a version that allocates the resources lazily with the notifier.
> > > Seems to boot, need to play with it a bit more though to make sure we're
> > > not overwriting resources for any legacy devices.
> > 
> > Blurg. Using a notifier really feels like we don't have a good handle on
> > a reasonable solution yet. Basically it means we're hooking into the
> > driver core without /looking/ like we're hooking into the driver core. I
> > don't think this is any better, but I don't have a better suggestion at
> > the moment.   :-(
> 
> Unfortunately this patch, or something that accomplishes the same results, 
> is somewhat high-priority for us on OMAP.  OMAP37xx got prematurely 
> converted to DT booting, and now dynamic power management is broken:
> 
>     http://marc.info/?l=linux-arm-kernel&m=138658294830408&w=2
> 
> Tony writes that this patch is one of the two patches needed to get things 
> working again:
> 
>     http://marc.info/?l=linux-arm-kernel&m=138660942506846&w=2
> 
> Is it possible to get this patch, or something similar, merged for 
> v3.13-rc?
> 
> Once something like PM is broken, it's pretty easy for other broken 
> patches to make it into the tree, since it becomes very difficult to test 
> without turning into a maintainer denial-of-service attack.

Ping.  Could you please provide some guidance here about what you'd need 
to get this solved for v3.13-rc?



- Paul

Re: [PATCH] of/platform: Fix no irq domain found errors when populating interrupts

[Rob Herring]

On Mon, Dec 30, 2013 at 4:10 PM, Paul Walmsley <paul-DWxLp4Yu+b8AvxtiuMwx3w@public.gmane.org> wrote:
> Hi Grant, Rob,
>
> On Tue, 10 Dec 2013, Paul Walmsley wrote:
>
>> On Sun, 24 Nov 2013, Grant Likely wrote:
>>
>> > On Fri, 22 Nov 2013 17:50:35 -0800, Tony Lindgren <tony-4v6yS6AI5VpBDgjK7y7TUQ@public.gmane.org> wrote:
>> > > * Tony Lindgren <tony-4v6yS6AI5VpBDgjK7y7TUQ@public.gmane.org> [131122 17:16]:
>> > > > * Tony Lindgren <tony-4v6yS6AI5VpBDgjK7y7TUQ@public.gmane.org> [131122 17:09]:
>> > > > > * Russell King - ARM Linux <linux-lFZ/pmaqli7XmaaqVzeoHQ@public.gmane.org> [131122 16:56]:
>> > > > > > On Fri, Nov 22, 2013 at 04:43:35PM -0800, Tony Lindgren wrote:
>> > > > > > > +         /* See of_device_resource_notify for populating interrupts */
>> > > > > > > +         for (i = 0; i < num_irq; i++, res++) {
>> > > > > > > +                 res->flags = IORESOURCE_IRQ;
>> > > > > > > +                 res->start = -EPROBE_DEFER;
>> > > > > > > +                 res->end = -EPROBE_DEFER;
>> > > > > >
>> > > > > > NAK.  Definitely a bad idea to start introducing magic values other into
>> > > > > > resources.  Please don't do this.
>> > > > >
>> > > > > Do you have any better ideas on how to sort out this issue then?
>> > > >
>> > > > I guess we could allocate all the resources lazily here, I'll take a look
>> > > > at that.
>> > >
>> > > Here's a version that allocates the resources lazily with the notifier.
>> > > Seems to boot, need to play with it a bit more though to make sure we're
>> > > not overwriting resources for any legacy devices.
>> >
>> > Blurg. Using a notifier really feels like we don't have a good handle on
>> > a reasonable solution yet. Basically it means we're hooking into the
>> > driver core without /looking/ like we're hooking into the driver core. I
>> > don't think this is any better, but I don't have a better suggestion at
>> > the moment.   :-(
>>
>> Unfortunately this patch, or something that accomplishes the same results,
>> is somewhat high-priority for us on OMAP.  OMAP37xx got prematurely
>> converted to DT booting, and now dynamic power management is broken:
>>
>>     http://marc.info/?l=linux-arm-kernel&m=138658294830408&w=2
>>
>> Tony writes that this patch is one of the two patches needed to get things
>> working again:
>>
>>     http://marc.info/?l=linux-arm-kernel&m=138660942506846&w=2
>>
>> Is it possible to get this patch, or something similar, merged for
>> v3.13-rc?
>>
>> Once something like PM is broken, it's pretty easy for other broken
>> patches to make it into the tree, since it becomes very difficult to test
>> without turning into a maintainer denial-of-service attack.
>
> Ping.  Could you please provide some guidance here about what you'd need
> to get this solved for v3.13-rc?

I agree with doing the platform_get_irq modification and feel free to
add my ack. I'm going to be offline for the next week.

Rob
--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH] of/platform: Fix no irq domain found errors when populating interrupts

[Paul Walmsley]

Hi Tony, Rob,

On Tue, 31 Dec 2013, Rob Herring wrote:

> On Mon, Dec 30, 2013 at 4:10 PM, Paul Walmsley <paul@pwsan.com> wrote:
> > Hi Grant, Rob,
> >
> > On Tue, 10 Dec 2013, Paul Walmsley wrote:
> >
> >> On Sun, 24 Nov 2013, Grant Likely wrote:
> >>
> >> > On Fri, 22 Nov 2013 17:50:35 -0800, Tony Lindgren <tony@atomide.com> wrote:
> >> > > * Tony Lindgren <tony@atomide.com> [131122 17:16]:
> >> > > > * Tony Lindgren <tony@atomide.com> [131122 17:09]:
> >> > > > > * Russell King - ARM Linux <linux@arm.linux.org.uk> [131122 16:56]:
> >> > > > > > On Fri, Nov 22, 2013 at 04:43:35PM -0800, Tony Lindgren wrote:
> >> > > > > > > +         /* See of_device_resource_notify for populating interrupts */
> >> > > > > > > +         for (i = 0; i < num_irq; i++, res++) {
> >> > > > > > > +                 res->flags = IORESOURCE_IRQ;
> >> > > > > > > +                 res->start = -EPROBE_DEFER;
> >> > > > > > > +                 res->end = -EPROBE_DEFER;
> >> > > > > >
> >> > > > > > NAK.  Definitely a bad idea to start introducing magic values other into
> >> > > > > > resources.  Please don't do this.
> >> > > > >
> >> > > > > Do you have any better ideas on how to sort out this issue then?
> >> > > >
> >> > > > I guess we could allocate all the resources lazily here, I'll take a look
> >> > > > at that.
> >> > >
> >> > > Here's a version that allocates the resources lazily with the notifier.
> >> > > Seems to boot, need to play with it a bit more though to make sure we're
> >> > > not overwriting resources for any legacy devices.
> >> >
> >> > Blurg. Using a notifier really feels like we don't have a good handle on
> >> > a reasonable solution yet. Basically it means we're hooking into the
> >> > driver core without /looking/ like we're hooking into the driver core. I
> >> > don't think this is any better, but I don't have a better suggestion at
> >> > the moment.   :-(
> >>
> >> Unfortunately this patch, or something that accomplishes the same results,
> >> is somewhat high-priority for us on OMAP.  OMAP37xx got prematurely
> >> converted to DT booting, and now dynamic power management is broken:
> >>
> >>     http://marc.info/?l=linux-arm-kernel&m=138658294830408&w=2
> >>
> >> Tony writes that this patch is one of the two patches needed to get things
> >> working again:
> >>
> >>     http://marc.info/?l=linux-arm-kernel&m=138660942506846&w=2
> >>
> >> Is it possible to get this patch, or something similar, merged for
> >> v3.13-rc?
> >>
> >> Once something like PM is broken, it's pretty easy for other broken
> >> patches to make it into the tree, since it becomes very difficult to test
> >> without turning into a maintainer denial-of-service attack.
> >
> > Ping.  Could you please provide some guidance here about what you'd need
> > to get this solved for v3.13-rc?
> 
> I agree with doing the platform_get_irq modification and feel free to
> add my ack. I'm going to be offline for the next week.

Thanks Rob!

Tony, would you consider sending this patch upstream for v3.13-rc along 
with your 37xxevm fix patches?


- Paul

Re: [PATCH] of/platform: Fix no irq domain found errors when populating interrupts

[Tony Lindgren]

* Paul Walmsley <paul@pwsan.com> [140106 15:43]:
> On Tue, 31 Dec 2013, Rob Herring wrote:
> > On Mon, Dec 30, 2013 at 4:10 PM, Paul Walmsley <paul@pwsan.com> wrote:
> > > On Tue, 10 Dec 2013, Paul Walmsley wrote:
> > >> Is it possible to get this patch, or something similar, merged for
> > >> v3.13-rc?
> > >>
> > >> Once something like PM is broken, it's pretty easy for other broken
> > >> patches to make it into the tree, since it becomes very difficult to test
> > >> without turning into a maintainer denial-of-service attack.
> > >
> > > Ping.  Could you please provide some guidance here about what you'd need
> > > to get this solved for v3.13-rc?
> > 
> > I agree with doing the platform_get_irq modification and feel free to
> > add my ack. I'm going to be offline for the next week.
> 
> Thanks Rob!
> 
> Tony, would you consider sending this patch upstream for v3.13-rc along 
> with your 37xxevm fix patches?

Hmm I'd like to see a proper patch posted for platform_get_irq() by
Thierry as that one is really his patch. And it also had a comment from
Grant.

After that I can queue it if nobody else is picking it as I have a
dependency to it.

Regards,

Tony

[PATCH] driver-core: platform: Resolve DT interrupt references late

[Thierry Reding]

When devices are probed from the device tree, any interrupts that they
reference are resolved at device creation time. This causes problems if
the interrupt provider hasn't been registered yet at that time, which
results in the interrupt being set to 0.

This is especially bad for platform devices because they are created at
a very early stage during boot when the majority of interrupt providers
haven't had a chance to be probed yet. One of the platform where this
causes major issues is OMAP.

Note that this patch is the easy way out to fix a large part of the
problems for now. A more proper solution for the long term would be to
transition drivers to an API that always resolves resources of any kind
(not only interrupts) at probe time.

For some background and discussion on possible solutions, see:

	https://lkml.org/lkml/2013/11/22/520

Acked-by: Rob Herring <robherring2@gmail.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
---
Note that this is somewhat urgent and should if at all possible go into
v3.13 before the release.

 drivers/base/platform.c | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)

diff --git a/drivers/base/platform.c b/drivers/base/platform.c
index 3a94b799f166..c894d1af3a5e 100644
--- a/drivers/base/platform.c
+++ b/drivers/base/platform.c
@@ -13,6 +13,7 @@
 #include <linux/string.h>
 #include <linux/platform_device.h>
 #include <linux/of_device.h>
+#include <linux/of_irq.h>
 #include <linux/module.h>
 #include <linux/init.h>
 #include <linux/dma-mapping.h>
@@ -87,7 +88,12 @@ int platform_get_irq(struct platform_device *dev, unsigned int num)
 		return -ENXIO;
 	return dev->archdata.irqs[num];
 #else
-	struct resource *r = platform_get_resource(dev, IORESOURCE_IRQ, num);
+	struct resource *r;
+
+	if (IS_ENABLED(CONFIG_OF) && dev->dev.of_node)
+		return irq_of_parse_and_map(dev->dev.of_node, num);
+
+	r = platform_get_resource(dev, IORESOURCE_IRQ, num);
 
 	return r ? r->start : -ENXIO;
 #endif
-- 
1.8.4.2

Re: [PATCH] driver-core: platform: Resolve DT interrupt references late

[Arnd Bergmann]

On Wednesday 08 January 2014 13:51:17 Thierry Reding wrote:
> When devices are probed from the device tree, any interrupts that they
> reference are resolved at device creation time. This causes problems if
> the interrupt provider hasn't been registered yet at that time, which
> results in the interrupt being set to 0.

Thanks for looking at this problem, it has bothered a lot of people
for a long time. I'm sorry I wasn't there for the discussion in November,
but when it came up before, I suggested a different solution that
apparently didn't get implemented.

> Note that this patch is the easy way out to fix a large part of the
> problems for now. A more proper solution for the long term would be to
> transition drivers to an API that always resolves resources of any kind
> (not only interrupts) at probe time.
> 
> For some background and discussion on possible solutions, see:
> 
> 	https://lkml.org/lkml/2013/11/22/520

I hope I read this thread correctly, sorry if I missed an important
part. My idea was to add the code not in platform_get_irq() but add
the resource in platform_drv_probe(), and just bail out with
-EPROBE_DEFER there if necessary.

We could then skip adding the resources at device creation time.
Is this something you already plan to do later, or is there a reason
it wouldn't work?

In the meantime, I don't see anything with your patch, but it also
wouldn't hurt to do it now if it solves all the problems.

	Arnd

Re: [PATCH] driver-core: platform: Resolve DT interrupt references late

[Thierry Reding]

[-- Attachment #1: Type: text/plain, Size: 6496 bytes --]

On Wed, Jan 08, 2014 at 02:41:19PM +0100, Arnd Bergmann wrote:
> On Wednesday 08 January 2014 13:51:17 Thierry Reding wrote:
> > When devices are probed from the device tree, any interrupts that they
> > reference are resolved at device creation time. This causes problems if
> > the interrupt provider hasn't been registered yet at that time, which
> > results in the interrupt being set to 0.
> 
> Thanks for looking at this problem, it has bothered a lot of people
> for a long time. I'm sorry I wasn't there for the discussion in November,
> but when it came up before, I suggested a different solution that
> apparently didn't get implemented.
> 
> > Note that this patch is the easy way out to fix a large part of the
> > problems for now. A more proper solution for the long term would be to
> > transition drivers to an API that always resolves resources of any kind
> > (not only interrupts) at probe time.
> > 
> > For some background and discussion on possible solutions, see:
> > 
> > 	https://lkml.org/lkml/2013/11/22/520
> 
> I hope I read this thread correctly, sorry if I missed an important
> part. My idea was to add the code not in platform_get_irq() but add
> the resource in platform_drv_probe(), and just bail out with
> -EPROBE_DEFER there if necessary.

One of the reasons we can't do that just yet is that we don't actually
get back an accurate error code from the OF IRQ helpers. Therefore if we
want to support deferred probing we either have to return -EPROBE_DEFER
for all errors (which is a bad idea because some errors just can't be
resolved by deferral) or we change the OF IRQ functions to propagate the
error code properly so that we know exactly when -EPROBE_DEFER makes
sense (i.e. the IRQ domain for an interrupt-parent hasn't been
registered yet).

Actually I posted a round of patches quite a while back that did exactly
this for interrupts. The changes were somewhat involved because it means
propagating an error code from fairly deep down in the OF code back to
the various higher-level helpers. If you're interested, the last version
of that is here:

	https://lkml.org/lkml/2013/9/18/216

Grant in particular seemed to be uncomfortable about how invasive the
patch series is.

Perhaps I should step back for a minute and provide some background on
how the initial idea came about. This was first triggered by the IOMMU
probe ordering issue that you may have heard about. One of the reasons
to do this for interrupts was because they are an existing type of
resource yet suffer from a very similar problem, so I wanted to solve
the issue for interrupts and thereby get a better overview of what
needed to be done for IOMMUs.

The most logical place for this code to reside seemed to be in the probe
path of platform devices (or I2C and other devices for that matter). The
patch series introduced an of_platform_probe() function that's called
from platform_drv_probe(). This would automatically give us deferred
probe support provided that we could propagate the appropriate error
code while at the same time giving us some flexibility to hook up other
resource types (such as IOMMUs).

One problem with the IOMMU patches is that they've received some strong
pushback from both Greg and Grant, arguing that it doesn't belong in the
core. If you want to read up on that, here's a link to the latest
version:

	https://lkml.org/lkml/2013/12/12/74

Some things had been discussed in earlier iterations of that series, but
this should give you the basic idea.

It stands to reason that if they push back on the IOMMU variant of what
is essentially the same thing, they will push back on the IRQ variant as
well. One alternative I proposed was to, just as you suggested earlier,
move the code into platform_drv_probe() or rather a function called from
it. That proposal never got any replies, though.

	https://lkml.org/lkml/2013/12/14/39

> We could then skip adding the resources at device creation time.
> Is this something you already plan to do later, or is there a reason
> it wouldn't work?

The current thread here suggests that it would be preferable not to have
any static allocations at all, but rather introduce a new API to resolve
things at probe time if necessary. I think the idea was to have a set of
functions like:

	int device_get_irq(struct device *dev, unsigned int num);
	struct resource *device_get_mem_region(struct device *dev,
					       unsigned int num);

Or even possible the more generic:

	struct resource *device_get_resource(struct device *dev,
					     unsigned int type,
					     unsigned int num);

If every driver used these functions, all resources could trivially be
resolved at probe time. That solution is also the most invasive one of
course, because it requires all existing drivers to be converted. At
least the API would be all new and therefore the conversion could happen
piecemeal.

One downside of that approach is that, while it maps well to platform
devices or generic devices that have some sort of firmware interface
such as OF or ACPI, I don't see how it can be made to work with an I2C
client that's registered from board setup code for example. Well, I
suppose that problem could be solved by throwing another lookup table at
it, just like we do for clocks, regulators, PWMs and GPIOs.

The good thing about it would be more consistency between the various
types of resources. Eventually I could imagine that we could even get
rid of struct resource (or at least only use it for a single type of
resource). But as I said, it'll take quite a bit of work to convert
everything to that.

> In the meantime, I don't see anything with your patch, but it also
> wouldn't hurt to do it now if it solves all the problems.

Well, the commit message explicitly states that this is only a temporary
measure, mostly to fix a number of regressions on OMAP where things were
broken by the conversion to DT in 3.13. The same is probably true of
other boards as well.

I'm willing to help fix things properly in the long run, but I think a
simple and low-risk patch like this would be worthwhile if it means that
a good many boards aren't broken in 3.13. Also given the history of the
above I can imagine that it will take more than the 3.14 cycle to get
this resolved satisfactorily, so at least for interrupts this would give
us a good stopgap solution in the meantime.

Thierry

[-- Attachment #2: Type: application/pgp-signature, Size: 836 bytes --]

Re: [PATCH] driver-core: platform: Resolve DT interrupt references late

[Arnd Bergmann]

On Wednesday 08 January 2014 15:55:27 Thierry Reding wrote:
> On Wed, Jan 08, 2014 at 02:41:19PM +0100, Arnd Bergmann wrote:
> > On Wednesday 08 January 2014 13:51:17 Thierry Reding wrote:
> > I hope I read this thread correctly, sorry if I missed an important
> > part. My idea was to add the code not in platform_get_irq() but add
> > the resource in platform_drv_probe(), and just bail out with
> > -EPROBE_DEFER there if necessary.
> 
> One of the reasons we can't do that just yet is that we don't actually
> get back an accurate error code from the OF IRQ helpers. Therefore if we
> want to support deferred probing we either have to return -EPROBE_DEFER
> for all errors (which is a bad idea because some errors just can't be
> resolved by deferral) or we change the OF IRQ functions to propagate the
> error code properly so that we know exactly when -EPROBE_DEFER makes
> sense (i.e. the IRQ domain for an interrupt-parent hasn't been
> registered yet).

I see.

> Actually I posted a round of patches quite a while back that did exactly
> this for interrupts. The changes were somewhat involved because it means
> propagating an error code from fairly deep down in the OF code back to
> the various higher-level helpers. If you're interested, the last version
> of that is here:
> 
> 	https://lkml.org/lkml/2013/9/18/216
> 
> Grant in particular seemed to be uncomfortable about how invasive the
> patch series is.

Interesting. It seems like a worthwhile thing to do, but I can understand
Grant's reluctance.

> 
> One problem with the IOMMU patches is that they've received some strong
> pushback from both Greg and Grant, arguing that it doesn't belong in the
> core. If you want to read up on that, here's a link to the latest
> version:
> 
> 	https://lkml.org/lkml/2013/12/12/74
> 
> Some things had been discussed in earlier iterations of that series, but
> this should give you the basic idea.

I'm skipping that discussion for now and stick with your summary

> It stands to reason that if they push back on the IOMMU variant of what
> is essentially the same thing, they will push back on the IRQ variant as
> well. One alternative I proposed was to, just as you suggested earlier,
> move the code into platform_drv_probe() or rather a function called from
> it. That proposal never got any replies, though.
> 
> 	https://lkml.org/lkml/2013/12/14/39

I guess putting it into the platform_drv_probe function seems reasonable,
I would be more scared of the implications of a notifier based method.

> > We could then skip adding the resources at device creation time.
> > Is this something you already plan to do later, or is there a reason
> > it wouldn't work?
> 
> The current thread here suggests that it would be preferable not to have
> any static allocations at all, but rather introduce a new API to resolve
> things at probe time if necessary. I think the idea was to have a set of
> functions like:
> 
> 	int device_get_irq(struct device *dev, unsigned int num);
> 	struct resource *device_get_mem_region(struct device *dev,
> 					       unsigned int num);
> 
> Or even possible the more generic:
> 
> 	struct resource *device_get_resource(struct device *dev,
> 					     unsigned int type,
> 					     unsigned int num);
> 
> If every driver used these functions, all resources could trivially be
> resolved at probe time. That solution is also the most invasive one of
> course, because it requires all existing drivers to be converted. At
> least the API would be all new and therefore the conversion could happen
> piecemeal.

Right, that does sound nice, and has the added benefit of saving
some memory allocations. I'd prefer the less generic variant here,
but I haven't given it much thought.
 
> One downside of that approach is that, while it maps well to platform
> devices or generic devices that have some sort of firmware interface
> such as OF or ACPI, I don't see how it can be made to work with an I2C
> client that's registered from board setup code for example. Well, I
> suppose that problem could be solved by throwing another lookup table at
> it, just like we do for clocks, regulators, PWMs and GPIOs.

Wouldn't you still be able to attach resources in the traditional
way for those, but use the same new interface to get at them?

> The good thing about it would be more consistency between the various
> types of resources. Eventually I could imagine that we could even get
> rid of struct resource (or at least only use it for a single type of
> resource). But as I said, it'll take quite a bit of work to convert
> everything to that.

struct resource is a structure with a long and complex history.
I'd certainly like to put some of it behind us and do something
that fits better into the 'struct device' concept which it
predates. I agree it would be a big effort though.

> > In the meantime, I don't see anything with your patch, but it also
> > wouldn't hurt to do it now if it solves all the problems.
> 
> Well, the commit message explicitly states that this is only a temporary
> measure, mostly to fix a number of regressions on OMAP where things were
> broken by the conversion to DT in 3.13. The same is probably true of
> other boards as well.

Right.

> I'm willing to help fix things properly in the long run, but I think a
> simple and low-risk patch like this would be worthwhile if it means that
> a good many boards aren't broken in 3.13. Also given the history of the
> above I can imagine that it will take more than the 3.14 cycle to get
> this resolved satisfactorily, so at least for interrupts this would give
> us a good stopgap solution in the meantime.

Ok. Thanks so much for your detailed background information!

	Arnd

Re: [PATCH] driver-core: platform: Resolve DT interrupt references late

[Thierry Reding]

[-- Attachment #1: Type: text/plain, Size: 6010 bytes --]

On Wed, Jan 08, 2014 at 04:11:08PM +0100, Arnd Bergmann wrote:
> On Wednesday 08 January 2014 15:55:27 Thierry Reding wrote:
> > On Wed, Jan 08, 2014 at 02:41:19PM +0100, Arnd Bergmann wrote:
> > > On Wednesday 08 January 2014 13:51:17 Thierry Reding wrote:
[...]
> > Actually I posted a round of patches quite a while back that did exactly
> > this for interrupts. The changes were somewhat involved because it means
> > propagating an error code from fairly deep down in the OF code back to
> > the various higher-level helpers. If you're interested, the last version
> > of that is here:
> > 
> > 	https://lkml.org/lkml/2013/9/18/216
> > 
> > Grant in particular seemed to be uncomfortable about how invasive the
> > patch series is.
> 
> Interesting. It seems like a worthwhile thing to do, but I can understand
> Grant's reluctance.

To be fair, Grant didn't say outright no. Given how easily this could
turn into a regression nightmare I do understand the reluctance as well.
Merging things piece by piece would make it somewhat less risky but at
the same time makes it hard to keep at it.

> > It stands to reason that if they push back on the IOMMU variant of what
> > is essentially the same thing, they will push back on the IRQ variant as
> > well. One alternative I proposed was to, just as you suggested earlier,
> > move the code into platform_drv_probe() or rather a function called from
> > it. That proposal never got any replies, though.
> > 
> > 	https://lkml.org/lkml/2013/12/14/39
> 
> I guess putting it into the platform_drv_probe function seems reasonable,
> I would be more scared of the implications of a notifier based method.

I fully agree. Of course if we decide against moving things into the
core and in favour of a more generic API that drivers should use, then
this issue goes away silently at least for resources that the driver
needs to use explicitly (memory-mapped regions, interrupts, ...).

The issue remains for IOMMU which is meant to be used transparently
through the DMA API. Perhaps a good compromise would be to have some
sort of generic helper that can be called to initialize IOMMU support
for a particular device and support probe deferral on error. Something
like this perhaps:

	int iommu_attach(struct device *dev);
	int iommu_detach(struct device *dev);

I still don't like very much how that needs to be done in each driver
explicitly, but if we can't do it in the core, then the only other clean
way to handle it would be to treat it like any other sort of resource
and handle it explicitly. Perhaps handing out some sort of cookie would
be preferable to just an error code?

> > > We could then skip adding the resources at device creation time.
> > > Is this something you already plan to do later, or is there a reason
> > > it wouldn't work?
> > 
> > The current thread here suggests that it would be preferable not to have
> > any static allocations at all, but rather introduce a new API to resolve
> > things at probe time if necessary. I think the idea was to have a set of
> > functions like:
> > 
> > 	int device_get_irq(struct device *dev, unsigned int num);
> > 	struct resource *device_get_mem_region(struct device *dev,
> > 					       unsigned int num);
> > 
> > Or even possible the more generic:
> > 
> > 	struct resource *device_get_resource(struct device *dev,
> > 					     unsigned int type,
> > 					     unsigned int num);
> > 
> > If every driver used these functions, all resources could trivially be
> > resolved at probe time. That solution is also the most invasive one of
> > course, because it requires all existing drivers to be converted. At
> > least the API would be all new and therefore the conversion could happen
> > piecemeal.
> 
> Right, that does sound nice, and has the added benefit of saving
> some memory allocations. I'd prefer the less generic variant here,
> but I haven't given it much thought.

I do prefer the less generic ones as well. They seem to be more
convenient to use.

> > One downside of that approach is that, while it maps well to platform
> > devices or generic devices that have some sort of firmware interface
> > such as OF or ACPI, I don't see how it can be made to work with an I2C
> > client that's registered from board setup code for example. Well, I
> > suppose that problem could be solved by throwing another lookup table at
> > it, just like we do for clocks, regulators, PWMs and GPIOs.
> 
> Wouldn't you still be able to attach resources in the traditional
> way for those, but use the same new interface to get at them?

I wouldn't know how. For instance platform devices store the IRQ number
within a struct resource of type IORESOURCE_IRQ, whereas I2C clients
store them in the struct i2c_client's .irq field.

So without actually introspecting the struct device (possibly using the
.bus field for example) and upcasting you won't know how to get at the
resources. One possibility to remedy that would be to try and unify the
resources within struct device. But that doesn't feel right.

One other thing I had considered at one point was to extend the bus_type
structure and give it a way to obtain resources in a bus-specific way,
but that feel even more wrong.

Perhaps I'm missing something obvious, though, and this is actually much
more trivial to solve.

> > The good thing about it would be more consistency between the various
> > types of resources. Eventually I could imagine that we could even get
> > rid of struct resource (or at least only use it for a single type of
> > resource). But as I said, it'll take quite a bit of work to convert
> > everything to that.
> 
> struct resource is a structure with a long and complex history.
> I'd certainly like to put some of it behind us and do something
> that fits better into the 'struct device' concept which it
> predates. I agree it would be a big effort though.

I concur.

Thierry

[-- Attachment #2: Type: application/pgp-signature, Size: 836 bytes --]

Re: [PATCH] driver-core: platform: Resolve DT interrupt references late

[Arnd Bergmann]

On Wednesday 08 January 2014, Thierry Reding wrote:
> On Wed, Jan 08, 2014 at 04:11:08PM +0100, Arnd Bergmann wrote:
> > On Wednesday 08 January 2014 15:55:27 Thierry Reding wrote:
> > > It stands to reason that if they push back on the IOMMU variant of what
> > > is essentially the same thing, they will push back on the IRQ variant as
> > > well. One alternative I proposed was to, just as you suggested earlier,
> > > move the code into platform_drv_probe() or rather a function called from
> > > it. That proposal never got any replies, though.
> > > 
> > > 	https://lkml.org/lkml/2013/12/14/39
> > 
> > I guess putting it into the platform_drv_probe function seems reasonable,
> > I would be more scared of the implications of a notifier based method.
> 
> I fully agree. Of course if we decide against moving things into the
> core and in favour of a more generic API that drivers should use, then
> this issue goes away silently at least for resources that the driver
> needs to use explicitly (memory-mapped regions, interrupts, ...).
> 
> The issue remains for IOMMU which is meant to be used transparently
> through the DMA API. Perhaps a good compromise would be to have some
> sort of generic helper that can be called to initialize IOMMU support
> for a particular device and support probe deferral on error. Something
> like this perhaps:
> 
> 	int iommu_attach(struct device *dev);
> 	int iommu_detach(struct device *dev);
> 
> I still don't like very much how that needs to be done in each driver
> explicitly, but if we can't do it in the core, then the only other clean
> way to handle it would be to treat it like any other sort of resource
> and handle it explicitly. Perhaps handing out some sort of cookie would
> be preferable to just an error code?

The more I think about the iommu case, the more I am convinced that it
does belong into the core, in whatever form we can find. As far as I
can tell from the little reliable information I have on the topic, I
would assume that we can keep it in the DT probing code, as there won't
be a need for multiple arbitrary IOMMUs with ACPI or with board files.

> > > One downside of that approach is that, while it maps well to platform
> > > devices or generic devices that have some sort of firmware interface
> > > such as OF or ACPI, I don't see how it can be made to work with an I2C
> > > client that's registered from board setup code for example. Well, I
> > > suppose that problem could be solved by throwing another lookup table at
> > > it, just like we do for clocks, regulators, PWMs and GPIOs.
> > 
> > Wouldn't you still be able to attach resources in the traditional
> > way for those, but use the same new interface to get at them?
> 
> I wouldn't know how. For instance platform devices store the IRQ number
> within a struct resource of type IORESOURCE_IRQ, whereas I2C clients
> store them in the struct i2c_client's .irq field.

Good point, I forgot about the special case for i2c_client->irq.
I looked now and noticed that very few i2c devices actually use this
field, but larger number uses platform_data, which has a similar
problem.

> So without actually introspecting the struct device (possibly using the
> .bus field for example) and upcasting you won't know how to get at the
> resources. One possibility to remedy that would be to try and unify the
> resources within struct device. But that doesn't feel right.
> 
> One other thing I had considered at one point was to extend the bus_type
> structure and give it a way to obtain resources in a bus-specific way,
> but that feel even more wrong.
> 
> Perhaps I'm missing something obvious, though, and this is actually much
> more trivial to solve.

No trivial solution that I can see. I think we can deal with the case
where platform code uses platform_device->resources, and everything else
comes down to having multiple code branches in the driver, as we already
have to deal with platform_data and DT properties describing stuff that
doesn't fit in the resources.

	Arnd
--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH] driver-core: platform: Resolve DT interrupt references late

[Thierry Reding]

[-- Attachment #1: Type: text/plain, Size: 5825 bytes --]

On Wed, Jan 08, 2014 at 05:25:17PM +0100, Arnd Bergmann wrote:
> On Wednesday 08 January 2014, Thierry Reding wrote:
> > On Wed, Jan 08, 2014 at 04:11:08PM +0100, Arnd Bergmann wrote:
> > > On Wednesday 08 January 2014 15:55:27 Thierry Reding wrote:
> > > > It stands to reason that if they push back on the IOMMU variant of what
> > > > is essentially the same thing, they will push back on the IRQ variant as
> > > > well. One alternative I proposed was to, just as you suggested earlier,
> > > > move the code into platform_drv_probe() or rather a function called from
> > > > it. That proposal never got any replies, though.
> > > > 
> > > > 	https://lkml.org/lkml/2013/12/14/39
> > > 
> > > I guess putting it into the platform_drv_probe function seems reasonable,
> > > I would be more scared of the implications of a notifier based method.
> > 
> > I fully agree. Of course if we decide against moving things into the
> > core and in favour of a more generic API that drivers should use, then
> > this issue goes away silently at least for resources that the driver
> > needs to use explicitly (memory-mapped regions, interrupts, ...).
> > 
> > The issue remains for IOMMU which is meant to be used transparently
> > through the DMA API. Perhaps a good compromise would be to have some
> > sort of generic helper that can be called to initialize IOMMU support
> > for a particular device and support probe deferral on error. Something
> > like this perhaps:
> > 
> > 	int iommu_attach(struct device *dev);
> > 	int iommu_detach(struct device *dev);
> > 
> > I still don't like very much how that needs to be done in each driver
> > explicitly, but if we can't do it in the core, then the only other clean
> > way to handle it would be to treat it like any other sort of resource
> > and handle it explicitly. Perhaps handing out some sort of cookie would
> > be preferable to just an error code?
> 
> The more I think about the iommu case, the more I am convinced that it
> does belong into the core, in whatever form we can find. As far as I
> can tell from the little reliable information I have on the topic, I
> would assume that we can keep it in the DT probing code, as there won't
> be a need for multiple arbitrary IOMMUs with ACPI or with board files.

I think part of the problem is that we don't have any DT probing code
yet. of_platform_probe() would have been that code. Perhaps if you weigh
in Grant and Greg will reconsider.

> > > > One downside of that approach is that, while it maps well to platform
> > > > devices or generic devices that have some sort of firmware interface
> > > > such as OF or ACPI, I don't see how it can be made to work with an I2C
> > > > client that's registered from board setup code for example. Well, I
> > > > suppose that problem could be solved by throwing another lookup table at
> > > > it, just like we do for clocks, regulators, PWMs and GPIOs.
> > > 
> > > Wouldn't you still be able to attach resources in the traditional
> > > way for those, but use the same new interface to get at them?
> > 
> > I wouldn't know how. For instance platform devices store the IRQ number
> > within a struct resource of type IORESOURCE_IRQ, whereas I2C clients
> > store them in the struct i2c_client's .irq field.
> 
> Good point, I forgot about the special case for i2c_client->irq.
> I looked now and noticed that very few i2c devices actually use this
> field, but larger number uses platform_data, which has a similar
> problem.

Yeah. It's kind of messy. The more I think about it, the more I begin to
like the lookup table option. One big advantage of that is that we could
actually unify much of the lookup code, much like we do for other types
of resources. It's a pattern that has worked fairly well in a number of
cases, so it seems natural to use it for interrupts as well.

An even more extreme option would be to go all the way and introduce
struct irq, along the same lines of the new struct gpiod. That would
allow nice things such as storing trigger types and such within the IRQ
handle and propagate them automatically.

> > So without actually introspecting the struct device (possibly using the
> > .bus field for example) and upcasting you won't know how to get at the
> > resources. One possibility to remedy that would be to try and unify the
> > resources within struct device. But that doesn't feel right.
> > 
> > One other thing I had considered at one point was to extend the bus_type
> > structure and give it a way to obtain resources in a bus-specific way,
> > but that feel even more wrong.
> > 
> > Perhaps I'm missing something obvious, though, and this is actually much
> > more trivial to solve.
> 
> No trivial solution that I can see. I think we can deal with the case
> where platform code uses platform_device->resources, and everything else
> comes down to having multiple code branches in the driver, as we already
> have to deal with platform_data and DT properties describing stuff that
> doesn't fit in the resources.

That would be another argument in favour of the lookup table. It would
provide a unified mechanism to define static interrupts if no firmware
interface is available *independent* of the device type. You could have
something like:

	static const struct irq_lookup board_irq_lookup[] = {
		IRQ_LOOKUP("gpio", 0, "0-0040", NULL), /* I2C client via GPIO expander */
		IRQ_LOOKUP("intc", 0, "ehci.1", NULL), /* platform device via INTC */
	};

Along with:

	struct irq *irq_get(struct device *dev, const char *con_id);

To look it up via DT, ACPI, lookup table. That obviously means a more or
less complete change in how interrupts are handled in the kernel, and it
may not be worth it in the end.

Thierry

[-- Attachment #2: Type: application/pgp-signature, Size: 836 bytes --]

Re: [PATCH] driver-core: platform: Resolve DT interrupt references late

[Arnd Bergmann]

On Wednesday 08 January 2014 20:59:10 Thierry Reding wrote:
> On Wed, Jan 08, 2014 at 05:25:17PM +0100, Arnd Bergmann wrote:
> > On Wednesday 08 January 2014, Thierry Reding wrote:
> > > On Wed, Jan 08, 2014 at 04:11:08PM +0100, Arnd Bergmann wrote:
> > The more I think about the iommu case, the more I am convinced that it
> > does belong into the core, in whatever form we can find. As far as I
> > can tell from the little reliable information I have on the topic, I
> > would assume that we can keep it in the DT probing code, as there won't
> > be a need for multiple arbitrary IOMMUs with ACPI or with board files.
> 
> I think part of the problem is that we don't have any DT probing code
> yet. of_platform_probe() would have been that code. Perhaps if you weigh
> in Grant and Greg will reconsider.

For all I know, we don't even have a binding proposal, but I may
have missed that.

> > Good point, I forgot about the special case for i2c_client->irq.
> > I looked now and noticed that very few i2c devices actually use this
> > field, but larger number uses platform_data, which has a similar
> > problem.
> 
> Yeah. It's kind of messy. The more I think about it, the more I begin to
> like the lookup table option. One big advantage of that is that we could
> actually unify much of the lookup code, much like we do for other types
> of resources. It's a pattern that has worked fairly well in a number of
> cases, so it seems natural to use it for interrupts as well.
> 
> An even more extreme option would be to go all the way and introduce
> struct irq, along the same lines of the new struct gpiod. That would
> allow nice things such as storing trigger types and such within the IRQ
> handle and propagate them automatically.

We already have struct irq_desc, and I believe everybody who works
with interrupts supports migrating from irq number interfaces to
irq descriptors as soon as we can find someone willing to do that
work.
 
> > No trivial solution that I can see. I think we can deal with the case
> > where platform code uses platform_device->resources, and everything else
> > comes down to having multiple code branches in the driver, as we already
> > have to deal with platform_data and DT properties describing stuff that
> > doesn't fit in the resources.
> 
> That would be another argument in favour of the lookup table. It would
> provide a unified mechanism to define static interrupts if no firmware
> interface is available *independent* of the device type. You could have
> something like:
> 
> 	static const struct irq_lookup board_irq_lookup[] = {
> 		IRQ_LOOKUP("gpio", 0, "0-0040", NULL), /* I2C client via GPIO expander */
> 		IRQ_LOOKUP("intc", 0, "ehci.1", NULL), /* platform device via INTC */
> 	};
> 
> Along with:
> 
> 	struct irq *irq_get(struct device *dev, const char *con_id);
> 
> To look it up via DT, ACPI, lookup table. That obviously means a more or
> less complete change in how interrupts are handled in the kernel, and it
> may not be worth it in the end.

It would certainly need a long migration period, and a plan for how to
get there without breaking things in the meantime. Rather than a lookup
table like the kind we have for clocks, I'd prefer a general way to
attach named properties to devices. My feeling is that building upon
devres would be a good plan, since it already provides a way to attach
arbitrary data to a device.

	Arnd

--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH] driver-core: platform: Resolve DT interrupt references late

[Thierry Reding]

[-- Attachment #1: Type: text/plain, Size: 4614 bytes --]

On Wed, Jan 08, 2014 at 09:09:17PM +0100, Arnd Bergmann wrote:
> On Wednesday 08 January 2014 20:59:10 Thierry Reding wrote:
> > On Wed, Jan 08, 2014 at 05:25:17PM +0100, Arnd Bergmann wrote:
> > > On Wednesday 08 January 2014, Thierry Reding wrote:
> > > > On Wed, Jan 08, 2014 at 04:11:08PM +0100, Arnd Bergmann wrote:
> > > The more I think about the iommu case, the more I am convinced that it
> > > does belong into the core, in whatever form we can find. As far as I
> > > can tell from the little reliable information I have on the topic, I
> > > would assume that we can keep it in the DT probing code, as there won't
> > > be a need for multiple arbitrary IOMMUs with ACPI or with board files.
> > 
> > I think part of the problem is that we don't have any DT probing code
> > yet. of_platform_probe() would have been that code. Perhaps if you weigh
> > in Grant and Greg will reconsider.
> 
> For all I know, we don't even have a binding proposal, but I may
> have missed that.

Yeah, last time I checked there was no concensus on that. I'll try to
dig up the thread and get it going again, adding you on Cc if you don't
mind (and in case you weren't Cc'ed in the first place anyway).

> > > Good point, I forgot about the special case for i2c_client->irq.
> > > I looked now and noticed that very few i2c devices actually use this
> > > field, but larger number uses platform_data, which has a similar
> > > problem.
> > 
> > Yeah. It's kind of messy. The more I think about it, the more I begin to
> > like the lookup table option. One big advantage of that is that we could
> > actually unify much of the lookup code, much like we do for other types
> > of resources. It's a pattern that has worked fairly well in a number of
> > cases, so it seems natural to use it for interrupts as well.
> > 
> > An even more extreme option would be to go all the way and introduce
> > struct irq, along the same lines of the new struct gpiod. That would
> > allow nice things such as storing trigger types and such within the IRQ
> > handle and propagate them automatically.
> 
> We already have struct irq_desc, and I believe everybody who works
> with interrupts supports migrating from irq number interfaces to
> irq descriptors as soon as we can find someone willing to do that
> work.

I didn't know that. I had always assumed irq_desc was only for internal
use by the IRQ code. Perhaps I'll look into that at some point.

> > > No trivial solution that I can see. I think we can deal with the case
> > > where platform code uses platform_device->resources, and everything else
> > > comes down to having multiple code branches in the driver, as we already
> > > have to deal with platform_data and DT properties describing stuff that
> > > doesn't fit in the resources.
> > 
> > That would be another argument in favour of the lookup table. It would
> > provide a unified mechanism to define static interrupts if no firmware
> > interface is available *independent* of the device type. You could have
> > something like:
> > 
> > 	static const struct irq_lookup board_irq_lookup[] = {
> > 		IRQ_LOOKUP("gpio", 0, "0-0040", NULL), /* I2C client via GPIO expander */
> > 		IRQ_LOOKUP("intc", 0, "ehci.1", NULL), /* platform device via INTC */
> > 	};
> > 
> > Along with:
> > 
> > 	struct irq *irq_get(struct device *dev, const char *con_id);
> > 
> > To look it up via DT, ACPI, lookup table. That obviously means a more or
> > less complete change in how interrupts are handled in the kernel, and it
> > may not be worth it in the end.
> 
> It would certainly need a long migration period, and a plan for how to
> get there without breaking things in the meantime. Rather than a lookup
> table like the kind we have for clocks, I'd prefer a general way to
> attach named properties to devices. My feeling is that building upon
> devres would be a good plan, since it already provides a way to attach
> arbitrary data to a device.

The problem with devres, or any other solution for that matter, is that
for the cases where we'd need something like this (that is, statically
allocated devices in board setup code) we don't have a fully initialized
struct device. We need at least device_initialize() to have been called
before devres can be used. Therefore we still need some sort of lookup
table that can be used to seed devres objects. Also devres will go away
completely when a driver is unloaded, so it'll have to be repopulated
everytime.

I don't think that would buy us much over a simple table lookup.

Thierry

[-- Attachment #2: Type: application/pgp-signature, Size: 836 bytes --]

Re: [PATCH] driver-core: platform: Resolve DT interrupt references late

[Arnd Bergmann]

On Wednesday 08 January 2014 21:24:08 Thierry Reding wrote:
> 
> The problem with devres, or any other solution for that matter, is that
> for the cases where we'd need something like this (that is, statically
> allocated devices in board setup code) we don't have a fully initialized
> struct device. We need at least device_initialize() to have been called
> before devres can be used. Therefore we still need some sort of lookup
> table that can be used to seed devres objects. Also devres will go away
> completely when a driver is unloaded, so it'll have to be repopulated
> everytime.
> 
> I don't think that would buy us much over a simple table lookup.

I would think we can come up with a way to add data to a device that
ends up in devres by the time the device gets registered. The question
is more whether you want a global table (or a set of global tables
for that matter) or rather have all the data local to the devices
you register. I generally prefer the latter.

There is an interesting question about what subsystems you'd want to
include in this mechanism. We have a growing number of subsystems
that want data represented in DT (clock, regulator, dmaengine, reset,
led, pwm, irq, iommu, ...), most of which don't have a 'struct resource'
equivalent. We could either try to create something generic enough
to easily cover all of them, or we declare that if you actually want
to use all of them you should really use DT, and we make it hard to
add another subsystem specific lookup mechanism.

	Arnd

Re: [PATCH] driver-core: platform: Resolve DT interrupt references late

[Tony Lindgren]

* Thierry Reding <thierry.reding-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org> [140108 04:55]:
> When devices are probed from the device tree, any interrupts that they
> reference are resolved at device creation time. This causes problems if
> the interrupt provider hasn't been registered yet at that time, which
> results in the interrupt being set to 0.
> 
> This is especially bad for platform devices because they are created at
> a very early stage during boot when the majority of interrupt providers
> haven't had a chance to be probed yet. One of the platform where this
> causes major issues is OMAP.
> 
> Note that this patch is the easy way out to fix a large part of the
> problems for now. A more proper solution for the long term would be to
> transition drivers to an API that always resolves resources of any kind
> (not only interrupts) at probe time.
> 
> For some background and discussion on possible solutions, see:
> 
> 	https://lkml.org/lkml/2013/11/22/520
> 
> Acked-by: Rob Herring <robherring2-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
> Signed-off-by: Thierry Reding <treding-DDmLM1+adcrQT0dZR+AlfA@public.gmane.org>
> ---
> Note that this is somewhat urgent and should if at all possible go into
> v3.13 before the release.
> 
>  drivers/base/platform.c | 8 +++++++-
>  1 file changed, 7 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/base/platform.c b/drivers/base/platform.c
> index 3a94b799f166..c894d1af3a5e 100644
> --- a/drivers/base/platform.c
> +++ b/drivers/base/platform.c
> @@ -13,6 +13,7 @@
>  #include <linux/string.h>
>  #include <linux/platform_device.h>
>  #include <linux/of_device.h>
> +#include <linux/of_irq.h>
>  #include <linux/module.h>
>  #include <linux/init.h>
>  #include <linux/dma-mapping.h>
> @@ -87,7 +88,12 @@ int platform_get_irq(struct platform_device *dev, unsigned int num)
>  		return -ENXIO;
>  	return dev->archdata.irqs[num];
>  #else
> -	struct resource *r = platform_get_resource(dev, IORESOURCE_IRQ, num);
> +	struct resource *r;
> +
> +	if (IS_ENABLED(CONFIG_OF) && dev->dev.of_node)
> +		return irq_of_parse_and_map(dev->dev.of_node, num);
> +
> +	r = platform_get_resource(dev, IORESOURCE_IRQ, num);
>  
>  	return r ? r->start : -ENXIO;
>  #endif

Hmm actually testing this patch, it does not fix fix the $Subject bug :(

irq: no irq domain found for /ocp/pinmux@48002030 !
[    0.301361] ------------[ cut here ]------------
[    0.301391] WARNING: CPU: 0 PID: 1 at drivers/of/platform.c:171 of_device_alloc+0x144/0x184()
[    0.301422] Modules linked in:
[    0.301452] CPU: 0 PID: 1 Comm: swapper/0 Not tainted 3.13.0-rc7-00002-g4d998a6 #17
[    0.301513] [<c0015c04>] (unwind_backtrace+0x0/0xf4) from [<c00127b0>] (show_stack+0x14/0x1c)
[    0.301544] [<c00127b0>] (show_stack+0x14/0x1c) from [<c05685a4>] (dump_stack+0x6c/0xa0)
[    0.301574] [<c05685a4>] (dump_stack+0x6c/0xa0) from [<c00425b4>] (warn_slowpath_common+0x64/0x84)
[    0.301605] [<c00425b4>] (warn_slowpath_common+0x64/0x84) from [<c00425f0>] (warn_slowpath_null+0x1c/0x24)
[    0.301635] [<c00425f0>] (warn_slowpath_null+0x1c/0x24) from [<c0485210>] (of_device_alloc+0x144/0x184)
[    0.301635] [<c0485210>] (of_device_alloc+0x144/0x184) from [<c0485294>] (of_platform_device_create_pdata+0x44/0x9c)
[    0.301666] [<c0485294>] (of_platform_device_create_pdata+0x44/0x9c) from [<c04853bc>] (of_platform_bus_create+0xd0/0x170)
[    0.301696] [<c04853bc>] (of_platform_bus_create+0xd0/0x170) from [<c0485418>] (of_platform_bus_create+0x12c/0x170)
[    0.301727] [<c0485418>] (of_platform_bus_create+0x12c/0x170) from [<c04854bc>] (of_platform_populate+0x60/0x98)
[    0.301757] [<c04854bc>] (of_platform_populate+0x60/0x98) from [<c07ca53c>] (pdata_quirks_init+0x28/0x78)
[    0.301788] [<c07ca53c>] (pdata_quirks_init+0x28/0x78) from [<c07bab20>] (customize_machine+0x20/0x48)
[    0.301818] [<c07bab20>] (customize_machine+0x20/0x48) from [<c000882c>] (do_one_initcall+0x2c/0x150)
[    0.301849] [<c000882c>] (do_one_initcall+0x2c/0x150) from [<c07b75d8>] (do_basic_setup+0x94/0xd4)
[    0.301879] [<c07b75d8>] (do_basic_setup+0x94/0xd4) from [<c07b7694>] (kernel_init_freeable+0x7c/0x120)
[    0.301910] [<c07b7694>] (kernel_init_freeable+0x7c/0x120) from [<c05667ec>] (kernel_init+0x8/0x120)
[    0.301940] [<c05667ec>] (kernel_init+0x8/0x120) from [<c000e908>] (ret_from_fork+0x14/0x2c)
[    0.302124] ---[ end trace 2b87f5de2f86f809 ]---
...

There's nothing wrong with the interrupt related code paths, we're just
trying to call the functions at a wrong time when thing are not yet
initialized.

Below is a repost of what works for me without using notifiers. Anybody
got any better ideas for a minimal fix?

Regards,

Tony

8< -------------------------------
From: Tony Lindgren <tony-4v6yS6AI5VpBDgjK7y7TUQ@public.gmane.org>
Date: Tue, 7 Jan 2014 17:07:18 -0800
Subject: [PATCH] of/platform: Fix no irq domain found errors when populating interrupts

Currently we get the following kind of errors if we try to use interrupt
phandles to irqchips that have not yet initialized:

irq: no irq domain found for /ocp/pinmux@48002030 !
------------[ cut here ]------------
WARNING: CPU: 0 PID: 1 at drivers/of/platform.c:171 of_device_alloc+0x144/0x184()
Modules linked in:
CPU: 0 PID: 1 Comm: swapper/0 Not tainted 3.12.0-00038-g42a9708 #1012
(show_stack+0x14/0x1c)
(dump_stack+0x6c/0xa0)
(warn_slowpath_common+0x64/0x84)
(warn_slowpath_null+0x1c/0x24)
(of_device_alloc+0x144/0x184)
(of_platform_device_create_pdata+0x44/0x9c)
(of_platform_bus_create+0xd0/0x170)
(of_platform_bus_create+0x12c/0x170)
(of_platform_populate+0x60/0x98)

This is because we're wrongly trying to populate resources that are not yet
available. It's perfectly valid to create irqchips dynamically, so let's
fix up the issue by populating the interrupt resources at the driver probe
time instead.

Note that at least currently we cannot dynamically allocate the resources as bus
specific code may add legacy resources with platform_device_add_resources()
before the driver probe. At least omap_device_alloc() currently relies on
num_resources to determine if legacy resources should be added. This issue
will get fixed automatically when mach-omap2 boots with DT only, but there
are probably other places too where platform_device_add_resources() modifies
things before driver probe.

The addition of of_platform_probe() is based on patches posted earlier by
Thierry Reding <thierry.reding-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>.

Signed-off-by: Tony Lindgren <tony-4v6yS6AI5VpBDgjK7y7TUQ@public.gmane.org>

--- a/drivers/base/platform.c
+++ b/drivers/base/platform.c
@@ -481,6 +481,10 @@ static int platform_drv_probe(struct device *_dev)
 	struct platform_device *dev = to_platform_device(_dev);
 	int ret;
 
+	ret = of_platform_probe(dev);
+	if (ret)
+		return ret;
+
 	if (ACPI_HANDLE(_dev))
 		acpi_dev_pm_attach(_dev, true);
 
--- a/drivers/of/platform.c
+++ b/drivers/of/platform.c
@@ -141,7 +141,7 @@ struct platform_device *of_device_alloc(struct device_node *np,
 				  struct device *parent)
 {
 	struct platform_device *dev;
-	int rc, i, num_reg = 0, num_irq;
+	int num_reg = 0, num_irq;
 	struct resource *res, temp_res;
 
 	dev = platform_device_alloc("", -1);
@@ -154,7 +154,14 @@ struct platform_device *of_device_alloc(struct device_node *np,
 			num_reg++;
 	num_irq = of_irq_count(np);
 
-	/* Populate the resource table */
+	/*
+	 * Only allocate the resources for us to use later on. Note that bus
+	 * specific code may also add in additional legacy resources using
+	 * platform_device_add_resources(), and may even rely on us allocating
+	 * the basic resources here to do so. So we cannot allocate the
+	 * resources lazily until the legacy code has been fixed to not rely
+	 * on allocating resources here.
+	 */
 	if (num_irq || num_reg) {
 		res = kzalloc(sizeof(*res) * (num_irq + num_reg), GFP_KERNEL);
 		if (!res) {
@@ -164,11 +171,7 @@ struct platform_device *of_device_alloc(struct device_node *np,
 
 		dev->num_resources = num_reg + num_irq;
 		dev->resource = res;
-		for (i = 0; i < num_reg; i++, res++) {
-			rc = of_address_to_resource(np, i, res);
-			WARN_ON(rc);
-		}
-		WARN_ON(of_irq_to_resource_table(np, res, num_irq) != num_irq);
+		/* See of_device_resource_populate for populating the data */
 	}
 
 	dev->dev.of_node = of_node_get(np);
@@ -187,6 +190,50 @@ struct platform_device *of_device_alloc(struct device_node *np,
 EXPORT_SYMBOL(of_device_alloc);
 
 /**
+ * of_device_resource_populate - Populate device resources from device tree
+ * @dev: pointer to platform device
+ *
+ * The device interrupts are not necessarily available for all
+ * irqdomains initially so we need to populate them lazily at
+ * device probe time from of_platform_populate.
+ */
+static int of_device_resource_populate(struct platform_device *pdev)
+{
+	struct device_node *np = pdev->dev.of_node;
+	int rc, i, num_reg = 0, num_irq;
+	struct resource *res, temp_res;
+
+	res = pdev->resource;
+
+	/*
+	 * Count the io and irq resources again. Currently we cannot rely on
+	 * pdev->num_resources as bus specific code may have changed that
+	 * with platform_device_add_resources(). But the resources we allocated
+	 * earlier are still there and available for us to populate.
+	 */
+	if (of_can_translate_address(np))
+		while (of_address_to_resource(np, num_reg, &temp_res) == 0)
+			num_reg++;
+	num_irq = of_irq_count(np);
+
+	if (pdev->num_resources < num_reg + num_irq) {
+		dev_WARN(&pdev->dev, "not enough resources %i < %i\n",
+			 pdev->num_resources, num_reg + num_irq);
+		return -EINVAL;
+	}
+
+	for (i = 0; i < num_reg; i++, res++) {
+		rc = of_address_to_resource(np, i, res);
+		WARN_ON(rc);
+	}
+
+	if (num_irq)
+		WARN_ON(of_irq_to_resource_table(np, res, num_irq) != num_irq);
+
+	return 0;
+}
+
+/**
  * of_platform_device_create_pdata - Alloc, initialize and register an of_device
  * @np: pointer to node to create device for
  * @bus_id: name to assign device
@@ -485,4 +532,35 @@ int of_platform_populate(struct device_node *root,
 	return rc;
 }
 EXPORT_SYMBOL_GPL(of_platform_populate);
+
+/**
+ * of_platform_probe() - OF specific initialization at probe time
+ * @pdev: pointer to a platform device
+ *
+ * This function is called by the driver core to perform devicetree-specific
+ * setup for a given platform device at probe time. If a device's resources
+ * as specified in the device tree are not available yet, this function can
+ * return -EPROBE_DEFER and cause the device to be probed again later, when
+ * other drivers that potentially provide the missing resources have been
+ * probed in turn.
+ *
+ * Note that because of the above, all code executed by this function must
+ * be prepared to be run multiple times on the same device (i.e. it must be
+ * idempotent).
+ *
+ * Returns 0 on success or a negative error code on failure.
+ */
+int of_platform_probe(struct platform_device *pdev)
+{
+	int ret;
+
+	if (!pdev->dev.of_node)
+		return 0;
+
+	ret = of_device_resource_populate(pdev);
+	if (ret < 0)
+		return ret;
+
+	return 0;
+}
 #endif /* CONFIG_OF_ADDRESS */
--- a/include/linux/of_platform.h
+++ b/include/linux/of_platform.h
@@ -72,6 +72,8 @@ extern int of_platform_populate(struct device_node *root,
 				const struct of_device_id *matches,
 				const struct of_dev_auxdata *lookup,
 				struct device *parent);
+
+extern int of_platform_probe(struct platform_device *pdev);
 #else
 static inline int of_platform_populate(struct device_node *root,
 					const struct of_device_id *matches,
@@ -80,6 +82,11 @@ static inline int of_platform_populate(struct device_node *root,
 {
 	return -ENODEV;
 }
+
+static inline int of_platform_probe(struct platform_device *pdev)
+{
+	return 0;
+}
 #endif
 
 #endif	/* _LINUX_OF_PLATFORM_H */
--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH] driver-core: platform: Resolve DT interrupt references late

[Thierry Reding]

[-- Attachment #1: Type: text/plain, Size: 13195 bytes --]

On Wed, Jan 08, 2014 at 08:40:41AM -0800, Tony Lindgren wrote:
> * Thierry Reding <thierry.reding-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org> [140108 04:55]:
> > When devices are probed from the device tree, any interrupts that they
> > reference are resolved at device creation time. This causes problems if
> > the interrupt provider hasn't been registered yet at that time, which
> > results in the interrupt being set to 0.
> > 
> > This is especially bad for platform devices because they are created at
> > a very early stage during boot when the majority of interrupt providers
> > haven't had a chance to be probed yet. One of the platform where this
> > causes major issues is OMAP.
> > 
> > Note that this patch is the easy way out to fix a large part of the
> > problems for now. A more proper solution for the long term would be to
> > transition drivers to an API that always resolves resources of any kind
> > (not only interrupts) at probe time.
> > 
> > For some background and discussion on possible solutions, see:
> > 
> > 	https://lkml.org/lkml/2013/11/22/520
> > 
> > Acked-by: Rob Herring <robherring2-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
> > Signed-off-by: Thierry Reding <treding-DDmLM1+adcrQT0dZR+AlfA@public.gmane.org>
> > ---
> > Note that this is somewhat urgent and should if at all possible go into
> > v3.13 before the release.
> > 
> >  drivers/base/platform.c | 8 +++++++-
> >  1 file changed, 7 insertions(+), 1 deletion(-)
> > 
> > diff --git a/drivers/base/platform.c b/drivers/base/platform.c
> > index 3a94b799f166..c894d1af3a5e 100644
> > --- a/drivers/base/platform.c
> > +++ b/drivers/base/platform.c
> > @@ -13,6 +13,7 @@
> >  #include <linux/string.h>
> >  #include <linux/platform_device.h>
> >  #include <linux/of_device.h>
> > +#include <linux/of_irq.h>
> >  #include <linux/module.h>
> >  #include <linux/init.h>
> >  #include <linux/dma-mapping.h>
> > @@ -87,7 +88,12 @@ int platform_get_irq(struct platform_device *dev, unsigned int num)
> >  		return -ENXIO;
> >  	return dev->archdata.irqs[num];
> >  #else
> > -	struct resource *r = platform_get_resource(dev, IORESOURCE_IRQ, num);
> > +	struct resource *r;
> > +
> > +	if (IS_ENABLED(CONFIG_OF) && dev->dev.of_node)
> > +		return irq_of_parse_and_map(dev->dev.of_node, num);
> > +
> > +	r = platform_get_resource(dev, IORESOURCE_IRQ, num);
> >  
> >  	return r ? r->start : -ENXIO;
> >  #endif
> 
> Hmm actually testing this patch, it does not fix fix the $Subject bug :(
> 
> irq: no irq domain found for /ocp/pinmux@48002030 !
> [    0.301361] ------------[ cut here ]------------
> [    0.301391] WARNING: CPU: 0 PID: 1 at drivers/of/platform.c:171 of_device_alloc+0x144/0x184()
> [    0.301422] Modules linked in:
> [    0.301452] CPU: 0 PID: 1 Comm: swapper/0 Not tainted 3.13.0-rc7-00002-g4d998a6 #17
> [    0.301513] [<c0015c04>] (unwind_backtrace+0x0/0xf4) from [<c00127b0>] (show_stack+0x14/0x1c)
> [    0.301544] [<c00127b0>] (show_stack+0x14/0x1c) from [<c05685a4>] (dump_stack+0x6c/0xa0)
> [    0.301574] [<c05685a4>] (dump_stack+0x6c/0xa0) from [<c00425b4>] (warn_slowpath_common+0x64/0x84)
> [    0.301605] [<c00425b4>] (warn_slowpath_common+0x64/0x84) from [<c00425f0>] (warn_slowpath_null+0x1c/0x24)
> [    0.301635] [<c00425f0>] (warn_slowpath_null+0x1c/0x24) from [<c0485210>] (of_device_alloc+0x144/0x184)
> [    0.301635] [<c0485210>] (of_device_alloc+0x144/0x184) from [<c0485294>] (of_platform_device_create_pdata+0x44/0x9c)
> [    0.301666] [<c0485294>] (of_platform_device_create_pdata+0x44/0x9c) from [<c04853bc>] (of_platform_bus_create+0xd0/0x170)
> [    0.301696] [<c04853bc>] (of_platform_bus_create+0xd0/0x170) from [<c0485418>] (of_platform_bus_create+0x12c/0x170)
> [    0.301727] [<c0485418>] (of_platform_bus_create+0x12c/0x170) from [<c04854bc>] (of_platform_populate+0x60/0x98)
> [    0.301757] [<c04854bc>] (of_platform_populate+0x60/0x98) from [<c07ca53c>] (pdata_quirks_init+0x28/0x78)
> [    0.301788] [<c07ca53c>] (pdata_quirks_init+0x28/0x78) from [<c07bab20>] (customize_machine+0x20/0x48)
> [    0.301818] [<c07bab20>] (customize_machine+0x20/0x48) from [<c000882c>] (do_one_initcall+0x2c/0x150)
> [    0.301849] [<c000882c>] (do_one_initcall+0x2c/0x150) from [<c07b75d8>] (do_basic_setup+0x94/0xd4)
> [    0.301879] [<c07b75d8>] (do_basic_setup+0x94/0xd4) from [<c07b7694>] (kernel_init_freeable+0x7c/0x120)
> [    0.301910] [<c07b7694>] (kernel_init_freeable+0x7c/0x120) from [<c05667ec>] (kernel_init+0x8/0x120)
> [    0.301940] [<c05667ec>] (kernel_init+0x8/0x120) from [<c000e908>] (ret_from_fork+0x14/0x2c)
> [    0.302124] ---[ end trace 2b87f5de2f86f809 ]---
> ...
> 
> There's nothing wrong with the interrupt related code paths, we're just
> trying to call the functions at a wrong time when thing are not yet
> initialized.

The patch won't get rid of that warning, but it should at least restore
things to a working state at runtime. At least for well-behaved drivers
that use platform_get_irq() rather than those that try to access the
resources directly.

> Below is a repost of what works for me without using notifiers. Anybody
> got any better ideas for a minimal fix?

That patch is somewhat big for something that should be a minimal fix.
Being the size that it is it might have undesired side-effects that may
not get noticed until it's way too late, so I'm hesitant to have
something like this merged at this point in the release cycle.

Thierry

> 8< -------------------------------
> From: Tony Lindgren <tony-4v6yS6AI5VpBDgjK7y7TUQ@public.gmane.org>
> Date: Tue, 7 Jan 2014 17:07:18 -0800
> Subject: [PATCH] of/platform: Fix no irq domain found errors when populating interrupts
> 
> Currently we get the following kind of errors if we try to use interrupt
> phandles to irqchips that have not yet initialized:
> 
> irq: no irq domain found for /ocp/pinmux@48002030 !
> ------------[ cut here ]------------
> WARNING: CPU: 0 PID: 1 at drivers/of/platform.c:171 of_device_alloc+0x144/0x184()
> Modules linked in:
> CPU: 0 PID: 1 Comm: swapper/0 Not tainted 3.12.0-00038-g42a9708 #1012
> (show_stack+0x14/0x1c)
> (dump_stack+0x6c/0xa0)
> (warn_slowpath_common+0x64/0x84)
> (warn_slowpath_null+0x1c/0x24)
> (of_device_alloc+0x144/0x184)
> (of_platform_device_create_pdata+0x44/0x9c)
> (of_platform_bus_create+0xd0/0x170)
> (of_platform_bus_create+0x12c/0x170)
> (of_platform_populate+0x60/0x98)
> 
> This is because we're wrongly trying to populate resources that are not yet
> available. It's perfectly valid to create irqchips dynamically, so let's
> fix up the issue by populating the interrupt resources at the driver probe
> time instead.
> 
> Note that at least currently we cannot dynamically allocate the resources as bus
> specific code may add legacy resources with platform_device_add_resources()
> before the driver probe. At least omap_device_alloc() currently relies on
> num_resources to determine if legacy resources should be added. This issue
> will get fixed automatically when mach-omap2 boots with DT only, but there
> are probably other places too where platform_device_add_resources() modifies
> things before driver probe.
> 
> The addition of of_platform_probe() is based on patches posted earlier by
> Thierry Reding <thierry.reding-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>.
> 
> Signed-off-by: Tony Lindgren <tony-4v6yS6AI5VpBDgjK7y7TUQ@public.gmane.org>
> 
> --- a/drivers/base/platform.c
> +++ b/drivers/base/platform.c
> @@ -481,6 +481,10 @@ static int platform_drv_probe(struct device *_dev)
>  	struct platform_device *dev = to_platform_device(_dev);
>  	int ret;
>  
> +	ret = of_platform_probe(dev);
> +	if (ret)
> +		return ret;
> +
>  	if (ACPI_HANDLE(_dev))
>  		acpi_dev_pm_attach(_dev, true);
>  
> --- a/drivers/of/platform.c
> +++ b/drivers/of/platform.c
> @@ -141,7 +141,7 @@ struct platform_device *of_device_alloc(struct device_node *np,
>  				  struct device *parent)
>  {
>  	struct platform_device *dev;
> -	int rc, i, num_reg = 0, num_irq;
> +	int num_reg = 0, num_irq;
>  	struct resource *res, temp_res;
>  
>  	dev = platform_device_alloc("", -1);
> @@ -154,7 +154,14 @@ struct platform_device *of_device_alloc(struct device_node *np,
>  			num_reg++;
>  	num_irq = of_irq_count(np);
>  
> -	/* Populate the resource table */
> +	/*
> +	 * Only allocate the resources for us to use later on. Note that bus
> +	 * specific code may also add in additional legacy resources using
> +	 * platform_device_add_resources(), and may even rely on us allocating
> +	 * the basic resources here to do so. So we cannot allocate the
> +	 * resources lazily until the legacy code has been fixed to not rely
> +	 * on allocating resources here.
> +	 */
>  	if (num_irq || num_reg) {
>  		res = kzalloc(sizeof(*res) * (num_irq + num_reg), GFP_KERNEL);
>  		if (!res) {
> @@ -164,11 +171,7 @@ struct platform_device *of_device_alloc(struct device_node *np,
>  
>  		dev->num_resources = num_reg + num_irq;
>  		dev->resource = res;
> -		for (i = 0; i < num_reg; i++, res++) {
> -			rc = of_address_to_resource(np, i, res);
> -			WARN_ON(rc);
> -		}
> -		WARN_ON(of_irq_to_resource_table(np, res, num_irq) != num_irq);
> +		/* See of_device_resource_populate for populating the data */
>  	}
>  
>  	dev->dev.of_node = of_node_get(np);
> @@ -187,6 +190,50 @@ struct platform_device *of_device_alloc(struct device_node *np,
>  EXPORT_SYMBOL(of_device_alloc);
>  
>  /**
> + * of_device_resource_populate - Populate device resources from device tree
> + * @dev: pointer to platform device
> + *
> + * The device interrupts are not necessarily available for all
> + * irqdomains initially so we need to populate them lazily at
> + * device probe time from of_platform_populate.
> + */
> +static int of_device_resource_populate(struct platform_device *pdev)
> +{
> +	struct device_node *np = pdev->dev.of_node;
> +	int rc, i, num_reg = 0, num_irq;
> +	struct resource *res, temp_res;
> +
> +	res = pdev->resource;
> +
> +	/*
> +	 * Count the io and irq resources again. Currently we cannot rely on
> +	 * pdev->num_resources as bus specific code may have changed that
> +	 * with platform_device_add_resources(). But the resources we allocated
> +	 * earlier are still there and available for us to populate.
> +	 */
> +	if (of_can_translate_address(np))
> +		while (of_address_to_resource(np, num_reg, &temp_res) == 0)
> +			num_reg++;
> +	num_irq = of_irq_count(np);
> +
> +	if (pdev->num_resources < num_reg + num_irq) {
> +		dev_WARN(&pdev->dev, "not enough resources %i < %i\n",
> +			 pdev->num_resources, num_reg + num_irq);
> +		return -EINVAL;
> +	}
> +
> +	for (i = 0; i < num_reg; i++, res++) {
> +		rc = of_address_to_resource(np, i, res);
> +		WARN_ON(rc);
> +	}
> +
> +	if (num_irq)
> +		WARN_ON(of_irq_to_resource_table(np, res, num_irq) != num_irq);
> +
> +	return 0;
> +}
> +
> +/**
>   * of_platform_device_create_pdata - Alloc, initialize and register an of_device
>   * @np: pointer to node to create device for
>   * @bus_id: name to assign device
> @@ -485,4 +532,35 @@ int of_platform_populate(struct device_node *root,
>  	return rc;
>  }
>  EXPORT_SYMBOL_GPL(of_platform_populate);
> +
> +/**
> + * of_platform_probe() - OF specific initialization at probe time
> + * @pdev: pointer to a platform device
> + *
> + * This function is called by the driver core to perform devicetree-specific
> + * setup for a given platform device at probe time. If a device's resources
> + * as specified in the device tree are not available yet, this function can
> + * return -EPROBE_DEFER and cause the device to be probed again later, when
> + * other drivers that potentially provide the missing resources have been
> + * probed in turn.
> + *
> + * Note that because of the above, all code executed by this function must
> + * be prepared to be run multiple times on the same device (i.e. it must be
> + * idempotent).
> + *
> + * Returns 0 on success or a negative error code on failure.
> + */
> +int of_platform_probe(struct platform_device *pdev)
> +{
> +	int ret;
> +
> +	if (!pdev->dev.of_node)
> +		return 0;
> +
> +	ret = of_device_resource_populate(pdev);
> +	if (ret < 0)
> +		return ret;
> +
> +	return 0;
> +}
>  #endif /* CONFIG_OF_ADDRESS */
> --- a/include/linux/of_platform.h
> +++ b/include/linux/of_platform.h
> @@ -72,6 +72,8 @@ extern int of_platform_populate(struct device_node *root,
>  				const struct of_device_id *matches,
>  				const struct of_dev_auxdata *lookup,
>  				struct device *parent);
> +
> +extern int of_platform_probe(struct platform_device *pdev);
>  #else
>  static inline int of_platform_populate(struct device_node *root,
>  					const struct of_device_id *matches,
> @@ -80,6 +82,11 @@ static inline int of_platform_populate(struct device_node *root,
>  {
>  	return -ENODEV;
>  }
> +
> +static inline int of_platform_probe(struct platform_device *pdev)
> +{
> +	return 0;
> +}
>  #endif
>  
>  #endif	/* _LINUX_OF_PLATFORM_H */

[-- Attachment #2: Type: application/pgp-signature, Size: 836 bytes --]

Re: [PATCH] driver-core: platform: Resolve DT interrupt references late

[Tony Lindgren]

* Thierry Reding <thierry.reding@gmail.com> [140108 11:32]:
> On Wed, Jan 08, 2014 at 08:40:41AM -0800, Tony Lindgren wrote:
> > 
> > There's nothing wrong with the interrupt related code paths, we're just
> > trying to call the functions at a wrong time when thing are not yet
> > initialized.
> 
> The patch won't get rid of that warning, but it should at least restore
> things to a working state at runtime. At least for well-behaved drivers
> that use platform_get_irq() rather than those that try to access the
> resources directly.

Well the problem I'm seeing is these nasty warnings.

BTW, I think the issue you're talking about regarding platform_get_irq()
got fixed by 4a43d686fe33 (of/irq: Pass trigger type in IRQ resource flags).
 
> > Below is a repost of what works for me without using notifiers. Anybody
> > got any better ideas for a minimal fix?
> 
> That patch is somewhat big for something that should be a minimal fix.
> Being the size that it is it might have undesired side-effects that may
> not get noticed until it's way too late, so I'm hesitant to have
> something like this merged at this point in the release cycle.

Yes I agree it's rather invasive, but we also do have things pretty
badly broken in the kernel for device tree based booting. And it seems
that nobody has a smaller patch that would fix it.

Regards,

Tony