[PATCH] ata: Don't use NO_IRQ in pata_of_platform driver

[PATCH] ata: Don't use NO_IRQ in pata_of_platform driver

[Linus Torvalds]

On Fri, Dec 2, 2011 at 11:26 AM, Dave Martin <dave.martin@linaro.org> wrote:
>
> This is now broken on ARM where, for good or bad, NO_IRQ currently is
> used and is -1.
>
> How do we resolve it? ?If we are ready to eliminate NO_IRQ from
> drivers/of/irq.c (or indeed, all code that uses it) and just use 0 for
> that case, we should surely just do it... but I'm not confident I can
> judge on that.

Just stop using NO_IRQ. First in drivers/of/irq.c, then in any drivers
as you notice breakage.

Don't *change* NO_IRQ to zero (that whole #define is broken - leave it
around as a marker of brokenness), just start removing it from all the
ARM drivers that use the OF infrastructure. Which is presumably not
all that many yet.

So whenever you find breakage, the fix now is to just remove NO_IRQ
tests, and replace them with "!irq".

                       Linus

[PATCH] ata: Don't use NO_IRQ in pata_of_platform driver

[Benjamin Herrenschmidt]

On Fri, 2011-12-02 at 11:28 -0800, Linus Torvalds wrote:
> On Fri, Dec 2, 2011 at 11:26 AM, Dave Martin <dave.martin@linaro.org> wrote:
> >
> > This is now broken on ARM where, for good or bad, NO_IRQ currently is
> > used and is -1.
> >
> > How do we resolve it?  If we are ready to eliminate NO_IRQ from
> > drivers/of/irq.c (or indeed, all code that uses it) and just use 0 for
> > that case, we should surely just do it... but I'm not confident I can
> > judge on that.
> 
> Just stop using NO_IRQ. First in drivers/of/irq.c, then in any drivers
> as you notice breakage.

Agreed. In fact the whole hack in drivers/of/irq.c was to accomodate ARM
which still uses -1, powerpc changed to 0 a long time ago.

Now that we have a generic remapper between HW and "linux" IRQ numbers,
there is no reason to stick to -1 even on ARM.

> Don't *change* NO_IRQ to zero (that whole #define is broken - leave it
> around as a marker of brokenness), just start removing it from all the
> ARM drivers that use the OF infrastructure. Which is presumably not
> all that many yet.
> 
> So whenever you find breakage, the fix now is to just remove NO_IRQ
> tests, and replace them with "!irq".

Cheers,
Ben.

[PATCH] ata: Don't use NO_IRQ in pata_of_platform driver

[Dave Martin]

On Sat, Dec 03, 2011 at 10:12:53AM +1100, Benjamin Herrenschmidt wrote:
> On Fri, 2011-12-02 at 11:28 -0800, Linus Torvalds wrote:
> > On Fri, Dec 2, 2011 at 11:26 AM, Dave Martin <dave.martin@linaro.org> wrote:
> > >
> > > This is now broken on ARM where, for good or bad, NO_IRQ currently is
> > > used and is -1.
> > >
> > > How do we resolve it?  If we are ready to eliminate NO_IRQ from
> > > drivers/of/irq.c (or indeed, all code that uses it) and just use 0 for
> > > that case, we should surely just do it... but I'm not confident I can
> > > judge on that.
> > 
> > Just stop using NO_IRQ. First in drivers/of/irq.c, then in any drivers
> > as you notice breakage.
> 
> Agreed. In fact the whole hack in drivers/of/irq.c was to accomodate ARM
> which still uses -1, powerpc changed to 0 a long time ago.
> 
> Now that we have a generic remapper between HW and "linux" IRQ numbers,
> there is no reason to stick to -1 even on ARM.
> 
> > Don't *change* NO_IRQ to zero (that whole #define is broken - leave it
> > around as a marker of brokenness), just start removing it from all the
> > ARM drivers that use the OF infrastructure. Which is presumably not
> > all that many yet.
> > 
> > So whenever you find breakage, the fix now is to just remove NO_IRQ
> > tests, and replace them with "!irq".
> 

Russell, do you know whether it would make sense to set a timeline for 
removing NO_IRQ from ARM platforms and migrating to 0 for the no-interrupt
case?  I'm assuming that this mainly involves migrating existing hard-wired
code that deals with interrupt numbers to use irq domains.

I worry that if we just change the convention for the OF case, we'll end
up with OF-ised platform drivers which have to deal with a different no-
irq convention depending on whether they are probed as platform drivers
or through the OF framework ... and these ported or semi-ported drivers
will be intermixed with unported drivers, confusing maintainers.
 
If board code starts initialising platform data for non-OF-ised platform
drivers based on IRQ numbers fetched via the OF code, things will get
even more confused...

Cheers
---Dave

[PATCH] ata: Don't use NO_IRQ in pata_of_platform driver

[Nicolas Pitre]

On Mon, 5 Dec 2011, Dave Martin wrote:
> On Sat, Dec 03, 2011 at 10:12:53AM +1100, Benjamin Herrenschmidt wrote:
> > On Fri, 2011-12-02 at 11:28 -0800, Linus Torvalds wrote:
> > > Don't *change* NO_IRQ to zero (that whole #define is broken - leave it
> > > around as a marker of brokenness), just start removing it from all the
> > > ARM drivers that use the OF infrastructure. Which is presumably not
> > > all that many yet.
> > > 
> > > So whenever you find breakage, the fix now is to just remove NO_IRQ
> > > tests, and replace them with "!irq".
> > 
> 
> Russell, do you know whether it would make sense to set a timeline for 
> removing NO_IRQ from ARM platforms and migrating to 0 for the no-interrupt
> case?  I'm assuming that this mainly involves migrating existing hard-wired
> code that deals with interrupt numbers to use irq domains.

How many drivers do use IRQ #0 to start with?  We might discover that in 
practice there is only a very few cases where this is an issue if 0 
would mean no IRQ.


Nicolas

[PATCH] ata: Don't use NO_IRQ in pata_of_platform driver

[Dave Martin]

On Mon, Dec 05, 2011 at 12:40:16PM -0500, Nicolas Pitre wrote:
> On Mon, 5 Dec 2011, Dave Martin wrote:
> > On Sat, Dec 03, 2011 at 10:12:53AM +1100, Benjamin Herrenschmidt wrote:
> > > On Fri, 2011-12-02 at 11:28 -0800, Linus Torvalds wrote:
> > > > Don't *change* NO_IRQ to zero (that whole #define is broken - leave it
> > > > around as a marker of brokenness), just start removing it from all the
> > > > ARM drivers that use the OF infrastructure. Which is presumably not
> > > > all that many yet.
> > > > 
> > > > So whenever you find breakage, the fix now is to just remove NO_IRQ
> > > > tests, and replace them with "!irq".
> > > 
> > 
> > Russell, do you know whether it would make sense to set a timeline for 
> > removing NO_IRQ from ARM platforms and migrating to 0 for the no-interrupt
> > case?  I'm assuming that this mainly involves migrating existing hard-wired
> > code that deals with interrupt numbers to use irq domains.
> 
> How many drivers do use IRQ #0 to start with?  We might discover that in 
> practice there is only a very few cases where this is an issue if 0 
> would mean no IRQ.

The total number of files referring to NO_IRQ is not that huge:

arch/arm/	188 matches in 39 files
drivers/	174 matches in 84 files

Unfortunately, NO_IRQ is often not spelled "NO_IRQ".  It looks like the assumption
"irq < 0 === no irq" may be quite a lot more widespread than "NO_IRQ === no irq".
Since there's no specific thing we can grep for (and simply due to volume)
finding all such instances may be quite a bit harder.

For example, git grep 'irq.*\(>=\|<[^=]\) *0' gives

drivers/	435 matches in 314 files
arch/arm/	18 matches in 15 files

A few examples:
drivers/input/mouse/pxa930_trkball.c:   if (irq < 0) {
drivers/input/keyboard/tegra-kbc.c:     if (irq < 0) {
drivers/crypto/omap-sham.c:     if (dd->irq >= 0)

...etc., etc., although there are probably a fair number of false positives here.


whereas git grep 'irq.*\(<\|>\|<=\|>=\|==\|!=\) \+-1' gives

drivers/	68 matches in 28 files
arch/arm/	18 matches in 15 files

Examples: 


...and that's just the code which is C and is also kind enough to put
irq numbers in variables with names containing "irq".

It also doesn't catch people initialising variables or struct/array
members to -1, unadorned "-1" arguments to functions and so on... though
those are likely to appear in mostly the same files matching the above
expressions, it won't be an exact 1:1 correspondence.


Cheers
---Dave

[PATCH] ata: Don't use NO_IRQ in pata_of_platform driver

[Geert Uytterhoeven]

On Mon, Dec 5, 2011 at 19:02, Dave Martin <dave.martin@linaro.org> wrote:
> Unfortunately, NO_IRQ is often not spelled "NO_IRQ". ?It looks like the assumption
> "irq < 0 === no irq" may be quite a lot more widespread than "NO_IRQ === no irq".

Can we make irq unsigned, and hope the compiler catches all of them
(comparison always
false blah blah blah)?

Gr{oetje,eeting}s,

? ? ? ? ? ? ? ? ? ? ? ? Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
? ? ? ? ? ? ? ? ? ? ? ? ? ?? ?? -- Linus Torvalds

[PATCH] ata: Don't use NO_IRQ in pata_of_platform driver

[Nicolas Pitre]

On Mon, 5 Dec 2011, Dave Martin wrote:

> On Mon, Dec 05, 2011 at 12:40:16PM -0500, Nicolas Pitre wrote:
> > On Mon, 5 Dec 2011, Dave Martin wrote:
> > > On Sat, Dec 03, 2011 at 10:12:53AM +1100, Benjamin Herrenschmidt wrote:
> > > > On Fri, 2011-12-02 at 11:28 -0800, Linus Torvalds wrote:
> > > > > Don't *change* NO_IRQ to zero (that whole #define is broken - leave it
> > > > > around as a marker of brokenness), just start removing it from all the
> > > > > ARM drivers that use the OF infrastructure. Which is presumably not
> > > > > all that many yet.
> > > > > 
> > > > > So whenever you find breakage, the fix now is to just remove NO_IRQ
> > > > > tests, and replace them with "!irq".
> > > > 
> > > 
> > > Russell, do you know whether it would make sense to set a timeline for 
> > > removing NO_IRQ from ARM platforms and migrating to 0 for the no-interrupt
> > > case?  I'm assuming that this mainly involves migrating existing hard-wired
> > > code that deals with interrupt numbers to use irq domains.
> > 
> > How many drivers do use IRQ #0 to start with?  We might discover that in 
> > practice there is only a very few cases where this is an issue if 0 
> > would mean no IRQ.
> 
> The total number of files referring to NO_IRQ is not that huge:
> 
> arch/arm/	188 matches in 39 files
> drivers/	174 matches in 84 files
> 
> Unfortunately, NO_IRQ is often not spelled "NO_IRQ".  It looks like the assumption
> "irq < 0 === no irq" may be quite a lot more widespread than "NO_IRQ === no irq".
> Since there's no specific thing we can grep for (and simply due to volume)
> finding all such instances may be quite a bit harder.
[...]

ARgh.

My point was about current actual usage of the IRQ numbered 0 which 
probably prompted the introduction of NO_IRQ in the first place.  What I 
was saying is that the number of occurrences where IRQ #0 is currently 
used into drivers that would get confused if 0 would mean no IRQ is 
probably quite small.

But as you illustrated, there is a large number of drivers that already 
assume no IRQ is < 0, even if they don't use any IRQ #0 themselves.  
That is a much bigger problem to fix.


Nicolas

[PATCH] ata: Don't use NO_IRQ in pata_of_platform driver

[Alan Cox]

> But as you illustrated, there is a large number of drivers that already 
> assume no IRQ is < 0, even if they don't use any IRQ #0 themselves.  
> That is a much bigger problem to fix.

And a much larger number assuming the reverse is true which are hiding
potential bugs on ARM.

Looking at the serial stuff the best checks appear to be looking at
"irq", "-1" and NO_IRQ.

For migration stuff that's doing broken things like

	if (irq < 0)

can be changed to

	if (irq <= 0)

and that can be done before NO_IRQ itself is nailed on ARM and PA-RISC.

[PATCH] ata: Don't use NO_IRQ in pata_of_platform driver

[James Bottomley]

On Mon, 2011-12-05 at 18:45 +0000, Alan Cox wrote:
> > But as you illustrated, there is a large number of drivers that already 
> > assume no IRQ is < 0, even if they don't use any IRQ #0 themselves.  
> > That is a much bigger problem to fix.
> 
> And a much larger number assuming the reverse is true which are hiding
> potential bugs on ARM.
> 
> Looking at the serial stuff the best checks appear to be looking at
> "irq", "-1" and NO_IRQ.
> 
> For migration stuff that's doing broken things like
> 
> 	if (irq < 0)
> 
> can be changed to
> 
> 	if (irq <= 0)
> 
> and that can be done before NO_IRQ itself is nailed on ARM and PA-RISC.

To be honest, we don't care very much.  Parisc interrupts are cascading
and mostly software assigned (except our EIEM which we keep internal).
We use a base offset at 16 or 64 (depending on GSC presence or not) so
IRQs 0-15 aren't legal on parisc either (we frob some of the hard coded
ISA interrupts on the WAX eisa bus).

We use NO_IRQ as an IRQ assignment error return and that's about it (and
that error shouldn't ever really occur).

James

[PATCH] ata: Don't use NO_IRQ in pata_of_platform driver

[Jean-Christophe PLAGNIOL-VILLARD]

On 18:45 Mon 05 Dec     , Alan Cox wrote:
> > But as you illustrated, there is a large number of drivers that already 
> > assume no IRQ is < 0, even if they don't use any IRQ #0 themselves.  
> > That is a much bigger problem to fix.
> 
> And a much larger number assuming the reverse is true which are hiding
> potential bugs on ARM.
> 
> Looking at the serial stuff the best checks appear to be looking at
> "irq", "-1" and NO_IRQ.
> 
> For migration stuff that's doing broken things like
> 
> 	if (irq < 0)
> 
> can be changed to
> 
> 	if (irq <= 0)
> 
> and that can be done before NO_IRQ itself is nailed on ARM and PA-RISC.
can we sinply introduce a macro irq_is_valid

and make it chip dependant as gpio_is_valid
and then move away from irq 0 valid

so we do not need to brake anthing first and then easly convert them

Best Regards,
J.

[PATCH] ata: Don't use NO_IRQ in pata_of_platform driver

[Alan Cox]

> can we sinply introduce a macro irq_is_valid

See the 2005, 2006 and 2008 discussion.

    if (!dev->irq)

is the proper test.

The <= is just a temporary thing while ARM gets its publically visible
house in order so it can be done without breakage.

Alan

[PATCH] ata: Don't use NO_IRQ in pata_of_platform driver

[Rob Herring]

On 12/05/2011 12:18 PM, Nicolas Pitre wrote:
> On Mon, 5 Dec 2011, Dave Martin wrote:
> 
>> On Mon, Dec 05, 2011 at 12:40:16PM -0500, Nicolas Pitre wrote:
>>> On Mon, 5 Dec 2011, Dave Martin wrote:
>>>> On Sat, Dec 03, 2011 at 10:12:53AM +1100, Benjamin Herrenschmidt wrote:
>>>>> On Fri, 2011-12-02 at 11:28 -0800, Linus Torvalds wrote:
>>>>>> Don't *change* NO_IRQ to zero (that whole #define is broken - leave it
>>>>>> around as a marker of brokenness), just start removing it from all the
>>>>>> ARM drivers that use the OF infrastructure. Which is presumably not
>>>>>> all that many yet.
>>>>>>
>>>>>> So whenever you find breakage, the fix now is to just remove NO_IRQ
>>>>>> tests, and replace them with "!irq".
>>>>>
>>>>
>>>> Russell, do you know whether it would make sense to set a timeline for 
>>>> removing NO_IRQ from ARM platforms and migrating to 0 for the no-interrupt
>>>> case?  I'm assuming that this mainly involves migrating existing hard-wired
>>>> code that deals with interrupt numbers to use irq domains.
>>>
>>> How many drivers do use IRQ #0 to start with?  We might discover that in 
>>> practice there is only a very few cases where this is an issue if 0 
>>> would mean no IRQ.
>>
>> The total number of files referring to NO_IRQ is not that huge:
>>
>> arch/arm/	188 matches in 39 files
>> drivers/	174 matches in 84 files
>>
>> Unfortunately, NO_IRQ is often not spelled "NO_IRQ".  It looks like the assumption
>> "irq < 0 === no irq" may be quite a lot more widespread than "NO_IRQ === no irq".
>> Since there's no specific thing we can grep for (and simply due to volume)
>> finding all such instances may be quite a bit harder.
> [...]
> 
> ARgh.
> 
> My point was about current actual usage of the IRQ numbered 0 which 
> probably prompted the introduction of NO_IRQ in the first place.  What I 
> was saying is that the number of occurrences where IRQ #0 is currently 
> used into drivers that would get confused if 0 would mean no IRQ is 
> probably quite small.
> 
> But as you illustrated, there is a large number of drivers that already 
> assume no IRQ is < 0, even if they don't use any IRQ #0 themselves.  
> That is a much bigger problem to fix.
> 

At least for DT enabled platforms, we could force "no irq" to be 0 in
the DT irq code. Searching the dts files, I found 2 occurrences of IRQ0.
Prima2 has timer on IRQ0, and VersatileAB has watchdog on IRQ0. Prima2
should be fine currently as it doesn't use the of_irq_* functions to get
the timer irq, but that is an issue as it skips any translation.
VersatileAB should be okay with the VIC irqdomain support.

Changing it would also affect microblaze and openrisc which have NO_IRQ
set to -1. From what I can tell, they would both be fine at least in
terms of not using IRQ0.

Also, there's roughly 50 irq_chips that need irq_domain support under
arch/arm. So that's not a simple solution either.

Rob

[PATCH] ata: Don't use NO_IRQ in pata_of_platform driver

[Anton Vorontsov]

On Mon, Dec 05, 2011 at 01:16:39PM -0600, Rob Herring wrote:
[...]
> At least for DT enabled platforms, we could force "no irq" to be 0 in
> the DT irq code. Searching the dts files, I found 2 occurrences of IRQ0.

Please note that there are HW IRQ numbers and "Virtual" IRQ numbers.
dev->irq and thus the thing that we pass into request_irq() is a
virtual IRQ thing, a "cookie".

While in device tree you see real HW IRQ numbers.

Legal VIRQ is always > 0, while HW IRQ could be >= 0.

> Prima2 has timer on IRQ0, and VersatileAB has watchdog on IRQ0. Prima2
> should be fine currently as it doesn't use the of_irq_* functions to get
> the timer irq, but that is an issue as it skips any translation.
> VersatileAB should be okay with the VIC irqdomain support.

It shouldn't be an issue to use of_irq_*() functions for these IRQs.
of_irq_*() will remap HW IRQ 0 to some other VIRQ. If it does not do
this currently, then it's a bug and should be fixed.

Thanks,

-- 
Anton Vorontsov
Email: cbouatmailru at gmail.com

[PATCH] ata: Don't use NO_IRQ in pata_of_platform driver

[Rob Herring]

On 12/05/2011 02:21 PM, Anton Vorontsov wrote:
> On Mon, Dec 05, 2011 at 01:16:39PM -0600, Rob Herring wrote:
> [...]
>> At least for DT enabled platforms, we could force "no irq" to be 0 in
>> the DT irq code. Searching the dts files, I found 2 occurrences of IRQ0.
> 
> Please note that there are HW IRQ numbers and "Virtual" IRQ numbers.
> dev->irq and thus the thing that we pass into request_irq() is a
> virtual IRQ thing, a "cookie".
> 
> While in device tree you see real HW IRQ numbers.
> 
> Legal VIRQ is always > 0, while HW IRQ could be >= 0.
> 

If this was all true, then there would be no discussion.

This is what we are working towards, but irq_chips all over the arm tree
do not support any translation or have base fixed at compile time. Only
a few have been converted. And some ARM platforms may never get
converted to DT.

>> Prima2 has timer on IRQ0, and VersatileAB has watchdog on IRQ0. Prima2
>> should be fine currently as it doesn't use the of_irq_* functions to get
>> the timer irq, but that is an issue as it skips any translation.
>> VersatileAB should be okay with the VIC irqdomain support.
> 
> It shouldn't be an issue to use of_irq_*() functions for these IRQs.
> of_irq_*() will remap HW IRQ 0 to some other VIRQ. If it does not do
> this currently, then it's a bug and should be fixed.

I think that's what I'm saying. It's either a bug or incomplete DT
conversion for the platform. Either way, those should get fixed first.

Rob

[PATCH] ata: Don't use NO_IRQ in pata_of_platform driver

[Alan Cox]

On Mon, 05 Dec 2011 14:47:29 -0600
Rob Herring <robherring2@gmail.com> wrote:

> On 12/05/2011 02:21 PM, Anton Vorontsov wrote:
> > On Mon, Dec 05, 2011 at 01:16:39PM -0600, Rob Herring wrote:
> > [...]
> >> At least for DT enabled platforms, we could force "no irq" to be 0 in
> >> the DT irq code. Searching the dts files, I found 2 occurrences of IRQ0.
> > 
> > Please note that there are HW IRQ numbers and "Virtual" IRQ numbers.
> > dev->irq and thus the thing that we pass into request_irq() is a
> > virtual IRQ thing, a "cookie".
> > 
> > While in device tree you see real HW IRQ numbers.
> > 
> > Legal VIRQ is always > 0, while HW IRQ could be >= 0.
> > 
> 
> If this was all true, then there would be no discussion.

Or more to the point. If the ARM people concerned had listened in 2005,
2006 or 2008 there would be no discussion.

> This is what we are working towards, but irq_chips all over the arm tree
> do not support any translation or have base fixed at compile time. Only
> a few have been converted. And some ARM platforms may never get
> converted to DT.

You've had six years. Let them break, it'll motivate any users to fix
them.

Alan

[PATCH] ata: Don't use NO_IRQ in pata_of_platform driver

[Dave Martin]

On Mon, Dec 05, 2011 at 02:47:29PM -0600, Rob Herring wrote:
> On 12/05/2011 02:21 PM, Anton Vorontsov wrote:
> > On Mon, Dec 05, 2011 at 01:16:39PM -0600, Rob Herring wrote:
> > [...]
> >> At least for DT enabled platforms, we could force "no irq" to be 0 in
> >> the DT irq code. Searching the dts files, I found 2 occurrences of IRQ0.
> > 
> > Please note that there are HW IRQ numbers and "Virtual" IRQ numbers.
> > dev->irq and thus the thing that we pass into request_irq() is a
> > virtual IRQ thing, a "cookie".
> > 
> > While in device tree you see real HW IRQ numbers.
> > 
> > Legal VIRQ is always > 0, while HW IRQ could be >= 0.
> > 
> 
> If this was all true, then there would be no discussion.
> 
> This is what we are working towards, but irq_chips all over the arm tree
> do not support any translation or have base fixed at compile time. Only
> a few have been converted. And some ARM platforms may never get
> converted to DT.
> 
> >> Prima2 has timer on IRQ0, and VersatileAB has watchdog on IRQ0. Prima2
> >> should be fine currently as it doesn't use the of_irq_* functions to get
> >> the timer irq, but that is an issue as it skips any translation.
> >> VersatileAB should be okay with the VIC irqdomain support.
> > 
> > It shouldn't be an issue to use of_irq_*() functions for these IRQs.
> > of_irq_*() will remap HW IRQ 0 to some other VIRQ. If it does not do
> > this currently, then it's a bug and should be fixed.
> 
> I think that's what I'm saying. It's either a bug or incomplete DT
> conversion for the platform. Either way, those should get fixed first.

Do we expect there to be any platform drivers which are shared between
legacy platforms and newer DT-ised platforms?

Those drivers would be pain points since they would need to understand
both conventions.

So far as I can see, only boards which are not DT-ised, which do not use
DT-ised drivers and which do not use drivers which use interrupts and
are either used by DT-ised boards or by arches with a non-zero NO_IRQ
could safely carry on using a non-zero NO_IRQ.  Tracking down exactly
which boards and drivers this applies to could be hard.  We could have a
CONFIG_NO_IRQ and make them depend on it, but we still have to find that
list of boards and drivers in the first place.


Otherwise, it feels like we might need a strategy for migrating pretty
much everything if we don't want to end up in a mess.

Cheers
---Dave

[PATCH] ata: Don't use NO_IRQ in pata_of_platform driver

[Alan Cox]

> Otherwise, it feels like we might need a strategy for migrating pretty
> much everything if we don't want to end up in a mess.

You really do anyway - lots of generic driver code knows !dev->irq is a
valid test. That covers things like 8250 based UART hardware, network phy
layer code and vast amounts more.

The bugs will already be there because ARM isn't using 0, they just
aren't getting seen or aren't getting hit.

Alan

[PATCH] ata: Don't use NO_IRQ in pata_of_platform driver

[Russell King - ARM Linux]

On Tue, Dec 06, 2011 at 09:30:00AM +0000, Dave Martin wrote:
> Do we expect there to be any platform drivers which are shared between
> legacy platforms and newer DT-ised platforms?
> 
> Those drivers would be pain points since they would need to understand
> both conventions.
> 
> So far as I can see, only boards which are not DT-ised, which do not use
> DT-ised drivers and which do not use drivers which use interrupts and
> are either used by DT-ised boards or by arches with a non-zero NO_IRQ
> could safely carry on using a non-zero NO_IRQ.  Tracking down exactly
> which boards and drivers this applies to could be hard.  We could have a
> CONFIG_NO_IRQ and make them depend on it, but we still have to find that
> list of boards and drivers in the first place.

You're digging too deeply into this.

Drivers which need to know whether an IRQ is valid need to know this if
they wish to do something different for 'this device doesn't have an IRQ
wired'.  These are the drivers which have problems because of the -1 vs
0 thing.

That is different from 'this is an invalid IRQ number', which is what
you find out when you call request_irq().

So please, stop thinking 'we need to convert drivers to check for <= 0'.
We don't.  We just need to make sure that we're not passing a zero IRQ
number to any driver.

On platforms where IRQ0 is special like x86, request_irq() will fail
with -EBUSY on drivers which don't care (or other kind of refusal to
initialize), and will cause 'polling mode' with the 8250 driver.

So, all that we need to do is to ensure that all the IRQ chip stuff is
fixed up so that IRQ0 is only used for the same purpose as x86 - the
PIC timer on systems with an ISA 8253 timer.  Everything else should
not pass IRQ0 outside core platform code.

[PATCH] ata: Don't use NO_IRQ in pata_of_platform driver

[Dave Martin]

On Mon, Dec 05, 2011 at 01:18:30PM -0500, Nicolas Pitre wrote:
> On Mon, 5 Dec 2011, Dave Martin wrote:
> 
> > On Mon, Dec 05, 2011 at 12:40:16PM -0500, Nicolas Pitre wrote:
> > > On Mon, 5 Dec 2011, Dave Martin wrote:
> > > > On Sat, Dec 03, 2011 at 10:12:53AM +1100, Benjamin Herrenschmidt wrote:
> > > > > On Fri, 2011-12-02 at 11:28 -0800, Linus Torvalds wrote:
> > > > > > Don't *change* NO_IRQ to zero (that whole #define is broken - leave it
> > > > > > around as a marker of brokenness), just start removing it from all the
> > > > > > ARM drivers that use the OF infrastructure. Which is presumably not
> > > > > > all that many yet.
> > > > > > 
> > > > > > So whenever you find breakage, the fix now is to just remove NO_IRQ
> > > > > > tests, and replace them with "!irq".
> > > > > 
> > > > 
> > > > Russell, do you know whether it would make sense to set a timeline for 
> > > > removing NO_IRQ from ARM platforms and migrating to 0 for the no-interrupt
> > > > case?  I'm assuming that this mainly involves migrating existing hard-wired
> > > > code that deals with interrupt numbers to use irq domains.
> > > 
> > > How many drivers do use IRQ #0 to start with?  We might discover that in 
> > > practice there is only a very few cases where this is an issue if 0 
> > > would mean no IRQ.
> > 
> > The total number of files referring to NO_IRQ is not that huge:
> > 
> > arch/arm/	188 matches in 39 files
> > drivers/	174 matches in 84 files
> > 
> > Unfortunately, NO_IRQ is often not spelled "NO_IRQ".  It looks like the assumption
> > "irq < 0 === no irq" may be quite a lot more widespread than "NO_IRQ === no irq".
> > Since there's no specific thing we can grep for (and simply due to volume)
> > finding all such instances may be quite a bit harder.
> [...]
> 
> ARgh.
> 
> My point was about current actual usage of the IRQ numbered 0 which 
> probably prompted the introduction of NO_IRQ in the first place.  What I 
> was saying is that the number of occurrences where IRQ #0 is currently 
> used into drivers that would get confused if 0 would mean no IRQ is 
> probably quite small.

Ah, I misunderstood -- that's a separate issue, but also an important one.
I guess this applies to a fair number of older boards.  One way of fixing
this would be to migrate those boards to use irq domains -- but those boards
may be sporadically maintained.
 
> But as you illustrated, there is a large number of drivers that already 
> assume no IRQ is < 0, even if they don't use any IRQ #0 themselves.  
> That is a much bigger problem to fix.

My concern is that as soon as we start to change this in significant
volume, a _lot_ of stuff is going to break.  Everywhere that an irq value
is passed from one piece of code to another, there is a potential
interface mismatch -- there seems to be no single place where we can
apply a conversion and fix everything.


Cheers
---Dave

[PATCH] ata: Don't use NO_IRQ in pata_of_platform driver

[Nicolas Pitre]

On Mon, 5 Dec 2011, Dave Martin wrote:

> On Mon, Dec 05, 2011 at 01:18:30PM -0500, Nicolas Pitre wrote:
> > On Mon, 5 Dec 2011, Dave Martin wrote:
> > 
> > > On Mon, Dec 05, 2011 at 12:40:16PM -0500, Nicolas Pitre wrote:
> > > > On Mon, 5 Dec 2011, Dave Martin wrote:
> > > > > On Sat, Dec 03, 2011 at 10:12:53AM +1100, Benjamin Herrenschmidt wrote:
> > > > > > On Fri, 2011-12-02 at 11:28 -0800, Linus Torvalds wrote:
> > > > > > > Don't *change* NO_IRQ to zero (that whole #define is broken - leave it
> > > > > > > around as a marker of brokenness), just start removing it from all the
> > > > > > > ARM drivers that use the OF infrastructure. Which is presumably not
> > > > > > > all that many yet.
> > > > > > > 
> > > > > > > So whenever you find breakage, the fix now is to just remove NO_IRQ
> > > > > > > tests, and replace them with "!irq".
> > > > > > 
> > > > > 
> > > > > Russell, do you know whether it would make sense to set a timeline for 
> > > > > removing NO_IRQ from ARM platforms and migrating to 0 for the no-interrupt
> > > > > case?  I'm assuming that this mainly involves migrating existing hard-wired
> > > > > code that deals with interrupt numbers to use irq domains.
> > > > 
> > > > How many drivers do use IRQ #0 to start with?  We might discover that in 
> > > > practice there is only a very few cases where this is an issue if 0 
> > > > would mean no IRQ.
> > > 
> > > The total number of files referring to NO_IRQ is not that huge:
> > > 
> > > arch/arm/	188 matches in 39 files
> > > drivers/	174 matches in 84 files
> > > 
> > > Unfortunately, NO_IRQ is often not spelled "NO_IRQ".  It looks like the assumption
> > > "irq < 0 === no irq" may be quite a lot more widespread than "NO_IRQ === no irq".
> > > Since there's no specific thing we can grep for (and simply due to volume)
> > > finding all such instances may be quite a bit harder.
> > [...]
> > 
> > ARgh.
> > 
> > My point was about current actual usage of the IRQ numbered 0 which 
> > probably prompted the introduction of NO_IRQ in the first place.  What I 
> > was saying is that the number of occurrences where IRQ #0 is currently 
> > used into drivers that would get confused if 0 would mean no IRQ is 
> > probably quite small.
> 
> Ah, I misunderstood -- that's a separate issue, but also an important one.
> I guess this applies to a fair number of older boards.  One way of fixing
> this would be to migrate those boards to use irq domains -- but those boards
> may be sporadically maintained.
>  
> > But as you illustrated, there is a large number of drivers that already 
> > assume no IRQ is < 0, even if they don't use any IRQ #0 themselves.  
> > That is a much bigger problem to fix.
> 
> My concern is that as soon as we start to change this in significant
> volume, a _lot_ of stuff is going to break.  Everywhere that an irq value
> is passed from one piece of code to another, there is a potential
> interface mismatch -- there seems to be no single place where we can
> apply a conversion and fix everything.

No need to convert everything.

First move is to make irq=0 meaning no IRQ.  That means making things 
like:

	if (irq < 0)
	if (irq >= 0)

into

	if (irq <= 0)
	if (irq > 0)

And replace NO_IRQ with 0.

That change shouldn't break anything, except those drivers which are 1) 
being passed an actual IRQ #0 and 2) testing for no IRQ.  I suspect that 
those conditions aren't very common together.


Nicolas

[PATCH] ata: Don't use NO_IRQ in pata_of_platform driver

[Dave Martin]

On Mon, Dec 05, 2011 at 02:49:01PM -0500, Nicolas Pitre wrote:

[...]

> > > > Unfortunately, NO_IRQ is often not spelled "NO_IRQ".  It looks like the assumption
> > > > "irq < 0 === no irq" may be quite a lot more widespread than "NO_IRQ === no irq".
> > > > Since there's no specific thing we can grep for (and simply due to volume)
> > > > finding all such instances may be quite a bit harder.
> > > [...]
> > > 
> > > ARgh.
> > > 
> > > My point was about current actual usage of the IRQ numbered 0 which 
> > > probably prompted the introduction of NO_IRQ in the first place.  What I 
> > > was saying is that the number of occurrences where IRQ #0 is currently 
> > > used into drivers that would get confused if 0 would mean no IRQ is 
> > > probably quite small.
> > 
> > Ah, I misunderstood -- that's a separate issue, but also an important one.
> > I guess this applies to a fair number of older boards.  One way of fixing
> > this would be to migrate those boards to use irq domains -- but those boards
> > may be sporadically maintained.
> >  
> > > But as you illustrated, there is a large number of drivers that already 
> > > assume no IRQ is < 0, even if they don't use any IRQ #0 themselves.  
> > > That is a much bigger problem to fix.
> > 
> > My concern is that as soon as we start to change this in significant
> > volume, a _lot_ of stuff is going to break.  Everywhere that an irq value
> > is passed from one piece of code to another, there is a potential
> > interface mismatch -- there seems to be no single place where we can
> > apply a conversion and fix everything.
> 
> No need to convert everything.
> 
> First move is to make irq=0 meaning no IRQ.  That means making things 
> like:
> 
> 	if (irq < 0)
> 	if (irq >= 0)
> 
> into
> 
> 	if (irq <= 0)
> 	if (irq > 0)
> 
> And replace NO_IRQ with 0.
>
> That change shouldn't break anything, except those drivers which are 1) 
> being passed an actual IRQ #0 and 2) testing for no IRQ.  I suspect that 
> those conditions aren't very common together.

To clarify, you're suggesting that the meanings of all other IRQ values
would not change initially?  (i.e., we remap HW irq 0, if there is one,
to some other random number but have a 1:1 mapping for everything else).


That could make sense as an approach.

Cheers
---Dave

[PATCH] ata: Don't use NO_IRQ in pata_of_platform driver

[Russell King - ARM Linux]

On Tue, Dec 06, 2011 at 09:37:09AM +0000, Dave Martin wrote:
> To clarify, you're suggesting that the meanings of all other IRQ values
> would not change initially?  (i.e., we remap HW irq 0, if there is one,
> to some other random number but have a 1:1 mapping for everything else).

Even better.  Avoid the first 16 IRQ numbers altogether - so that ISA
drivers which have these numbers hard-encoded in them will see failures
if they're expecting standard ISA IRQ numbering.

We already do that with the GIC, partly because of the hardware design.
We do that on Footbridge based systems, because they may or may not have
a real ISA IRQ controller.

But.. let's make one thing clear: Alan Cox and Linus have been going on
about how IRQ0 should not be used.  Let's be crystal clear: even x86
uses IRQ0.  It happens to be the PIC timer, and that gets claimed early
on during the x86 boot.  So please don't tell me that x86 avoids IRQ0.
It doesn't.  It just happens that x86 never shows IRQ0 to anything but
the i8253 PIC driver.

So lets see how x86 squeels if we make the i8253 PIC driver reject IRQ0.
I bet that there'd be absolute fury at such a suggestion.

When x86 sorts this out, there _might_ be some more motivation to take
such comments seriously.  Until then it's more like a joke.

[PATCH] ata: Don't use NO_IRQ in pata_of_platform driver

[Geert Uytterhoeven]

On Tue, Dec 6, 2011 at 11:46, Russell King - ARM Linux
<linux@arm.linux.org.uk> wrote:
> But.. let's make one thing clear: Alan Cox and Linus have been going on
> about how IRQ0 should not be used. ?Let's be crystal clear: even x86
> uses IRQ0. ?It happens to be the PIC timer, and that gets claimed early
> on during the x86 boot. ?So please don't tell me that x86 avoids IRQ0.
> It doesn't. ?It just happens that x86 never shows IRQ0 to anything but
> the i8253 PIC driver.

It's shown in /proc/interrupts due to a "bug" in show_interrupts().
The (gmail damaged) patch below fixes this bug.

>From 46f51a2d42548358868a34df00c2a4e47bbdf691 Mon Sep 17 00:00:00 2001
From: Geert Uytterhoeven <Geert.Uytterhoeven@eu.sony.com>
Date: Tue, 6 Dec 2011 11:55:05 +0100
Subject: [PATCH] /proc/interrupts: irq zero is invalid

As zero is an invalid irq number, show_interrupts() should not try to
print it. Just return after printing the header for i == 0.

Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
---
 kernel/irq/proc.c |    1 +
 1 files changed, 1 insertions(+), 0 deletions(-)

diff --git a/kernel/irq/proc.c b/kernel/irq/proc.c
index 4bd4faa..5b8bbf0 100644
--- a/kernel/irq/proc.c
+++ b/kernel/irq/proc.c
@@ -439,6 +439,7 @@ int show_interrupts(struct seq_file *p, void *v)
 		for_each_online_cpu(j)
 			seq_printf(p, "CPU%-8d", j);
 		seq_putc(p, '\n');
+		return 0;
 	}

 	desc = irq_to_desc(i);
-- 
1.7.0.4

Gr{oetje,eeting}s,

? ? ? ? ? ? ? ? ? ? ? ? Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert at linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
? ? ? ? ? ? ? ? ? ? ? ? ? ?? ?? -- Linus Torvalds

[PATCH] ata: Don't use NO_IRQ in pata_of_platform driver

[Russell King - ARM Linux]

On Tue, Dec 06, 2011 at 12:00:12PM +0100, Geert Uytterhoeven wrote:
> On Tue, Dec 6, 2011 at 11:46, Russell King - ARM Linux
> <linux@arm.linux.org.uk> wrote:
> > But.. let's make one thing clear: Alan Cox and Linus have been going on
> > about how IRQ0 should not be used. ?Let's be crystal clear: even x86
> > uses IRQ0. ?It happens to be the PIC timer, and that gets claimed early
> > on during the x86 boot. ?So please don't tell me that x86 avoids IRQ0.
> > It doesn't. ?It just happens that x86 never shows IRQ0 to anything but
> > the i8253 PIC driver.
> 
> It's shown in /proc/interrupts due to a "bug" in show_interrupts().
> The (gmail damaged) patch below fixes this bug.

So we now try to hide the fact that there _is_ an interrupt called 0
on x86 systems?  Sorry, I can't that that seriously in any way.

[PATCH] ata: Don't use NO_IRQ in pata_of_platform driver

[Alan Cox]

> It's shown in /proc/interrupts due to a "bug" in show_interrupts().
> The (gmail damaged) patch below fixes this bug.

We get API breakage then. Which is a pain of course because debug tools
and the like which think IRQ 0 is "timer ticks" are somewhat broken.

[PATCH] ata: Don't use NO_IRQ in pata_of_platform driver

[Alan Cox]

> Even better.  Avoid the first 16 IRQ numbers altogether - so that ISA
> drivers which have these numbers hard-encoded in them will see failures
> if they're expecting standard ISA IRQ numbering.

The ISA bus space is non-discoverable so that doesn't make any sense.

> But.. let's make one thing clear: Alan Cox and Linus have been going on
> about how IRQ0 should not be used.  Let's be crystal clear: even x86
> uses IRQ0.  It happens to be the PIC timer, and that gets claimed early
> on during the x86 boot.  So please don't tell me that x86 avoids IRQ0.
> It doesn't.  It just happens that x86 never shows IRQ0 to anything but
> the i8253 PIC driver.

x86 has an internal invisible IRQ 0 on some platforms. It's never exposed
beyond the arch code.

> So lets see how x86 squeels if we make the i8253 PIC driver reject IRQ0.
> I bet that there'd be absolute fury at such a suggestion.

Actually it would be about ten minutes work to remap it to some other
number that isn't used. It never goes via request_irq or via any core or
driver layer code however.

In the ARM case the same is going to be true. If you have IRQ 0 plumbing
that only goes internally in the arch/arm code - eg an ARM with IRQ 0
wired to something totally arch specific and non-driver then it's not
going to break and like the internals of x86 doesn't matter.

> When x86 sorts this out, there _might_ be some more motivation to take
> such comments seriously.  Until then it's more like a joke.

Pity you feel that way, but if ARM wants to continue to break more and
more as we clean up NO_IRQ for everything else that's your privilege, but
don't expect any sympathy.

Alan

[PATCH] ata: Don't use NO_IRQ in pata_of_platform driver

[Russell King - ARM Linux]

On Tue, Dec 06, 2011 at 11:05:54AM +0000, Alan Cox wrote:
> > Even better.  Avoid the first 16 IRQ numbers altogether - so that ISA
> > drivers which have these numbers hard-encoded in them will see failures
> > if they're expecting standard ISA IRQ numbering.
> 
> The ISA bus space is non-discoverable so that doesn't make any sense.
> 
> > But.. let's make one thing clear: Alan Cox and Linus have been going on
> > about how IRQ0 should not be used.  Let's be crystal clear: even x86
> > uses IRQ0.  It happens to be the PIC timer, and that gets claimed early
> > on during the x86 boot.  So please don't tell me that x86 avoids IRQ0.
> > It doesn't.  It just happens that x86 never shows IRQ0 to anything but
> > the i8253 PIC driver.
> 
> x86 has an internal invisible IRQ 0 on some platforms. It's never exposed
> beyond the arch code.
> 
> > So lets see how x86 squeels if we make the i8253 PIC driver reject IRQ0.
> > I bet that there'd be absolute fury at such a suggestion.
> 
> Actually it would be about ten minutes work to remap it to some other
> number that isn't used. It never goes via request_irq or via any core or
> driver layer code however.
> 
> In the ARM case the same is going to be true. If you have IRQ 0 plumbing
> that only goes internally in the arch/arm code - eg an ARM with IRQ 0
> wired to something totally arch specific and non-driver then it's not
> going to break and like the internals of x86 doesn't matter.
> 
> > When x86 sorts this out, there _might_ be some more motivation to take
> > such comments seriously.  Until then it's more like a joke.
> 
> Pity you feel that way, but if ARM wants to continue to break more and
> more as we clean up NO_IRQ for everything else that's your privilege, but
> don't expect any sympathy.

For the platforms I care about, it probably won't break.  For those which
do break, it's a matter of fixing their include/mach/irqs.h and the code
in their irqchips to convert the IRQ number to the correct bitmask for
the register.

However, I have suggested in the past that new platforms _should_ avoid
not just IRQ0 but IRQ0-15 (for a completely different reason to that of
'IRQ0 means no IRQ'.)  But such comments just get ignored, so I just
don't see the point in doing anything about this.  If people experience
breakage, so be it.  I too will have little sympathy but not for the same
reason.

[PATCH] ata: Don't use NO_IRQ in pata_of_platform driver

[Alan Cox]

> However, I have suggested in the past that new platforms _should_ avoid
> not just IRQ0 but IRQ0-15 (for a completely different reason to that of
> 'IRQ0 means no IRQ'.)  But such comments just get ignored, so I just
> don't see the point in doing anything about this.  If people experience
> breakage, so be it.  I too will have little sympathy but not for the same
> reason.

The one I can think of that is capable of taking EISA/ISA cards but has
differently IRQ plumbing arrangements is PA-RISC, and they do exactly
this.

Beyond that it probably doesn't come up except in the weird world of PCI
legacy compatibility for legacy IDE and VGA vertical interrupt routing.
In those cases we fix up the PCI config space so the platform in turn can
do proper IRQ plumbing.

Alan

[PATCH] ata: Don't use NO_IRQ in pata_of_platform driver

[Dave Martin]

On Tue, Dec 06, 2011 at 10:46:54AM +0000, Russell King - ARM Linux wrote:
> On Tue, Dec 06, 2011 at 09:37:09AM +0000, Dave Martin wrote:
> > To clarify, you're suggesting that the meanings of all other IRQ values
> > would not change initially?  (i.e., we remap HW irq 0, if there is one,
> > to some other random number but have a 1:1 mapping for everything else).
> 
> Even better.  Avoid the first 16 IRQ numbers altogether - so that ISA
> drivers which have these numbers hard-encoded in them will see failures
> if they're expecting standard ISA IRQ numbering.
> 
> We already do that with the GIC, partly because of the hardware design.
> We do that on Footbridge based systems, because they may or may not have
> a real ISA IRQ controller.
> 
> But.. let's make one thing clear: Alan Cox and Linus have been going on
> about how IRQ0 should not be used.  Let's be crystal clear: even x86
> uses IRQ0.  It happens to be the PIC timer, and that gets claimed early
> on during the x86 boot.  So please don't tell me that x86 avoids IRQ0.
> It doesn't.  It just happens that x86 never shows IRQ0 to anything but
> the i8253 PIC driver.
> 
> So lets see how x86 squeels if we make the i8253 PIC driver reject IRQ0.
> I bet that there'd be absolute fury at such a suggestion.
> 
> When x86 sorts this out, there _might_ be some more motivation to take
> such comments seriously.  Until then it's more like a joke.

OK -- but the situation is breaking OF-based drivers on ARM platforms
today.

Based on what you've suggested, does the following policy sound
reasonable for resolving that deadlock?


1) All OF code and drivers should be migrating to use 0 instead of NO_IRQ
   for the no-interrupt case.  Code which receives irq numbers directly
   from the OF framework and refers to NO_IRQ, or expects 0 to be a valid
   needs to be fixed.

2) Where we hit a problem, board code needs to be adapted to remap HW IRQs
   0-15 to different software values.  (This could be done using irq
   domains, or not)


I'm still not sure what the correct approach is for drivers which get
irq numbers from OF indirectly -- this particularly applies to platform
and AMBA devices.

If we expect board code to start populating platform data based on 
information from the OF code, we need to fix the board not to use linux
irq 0 to describe a real HW interrupt, if it matters (as in (2)).

AMBA devices registered via of_platform_populate() already get their
irq numbers from OF.  So long as OF used to explicitly return NO_IRQ
there was no problem -- but if OF is moving to return 0 instead, we have
a potential problem for each AMBA driver which may be used by a board
which can boot without DT... if we have any scenarios where that driver
is given real irq 0.

Maybe we can fix these breakages as they occur -- I don't really know
the scale of the impact.


What are your thoughts on this?

Cheers
---Dave

[PATCH] ata: Don't use NO_IRQ in pata_of_platform driver

[Russell King - ARM Linux]

On Tue, Dec 06, 2011 at 11:37:35AM +0000, Dave Martin wrote:
> 1) All OF code and drivers should be migrating to use 0 instead of NO_IRQ
>    for the no-interrupt case.  Code which receives irq numbers directly
>    from the OF framework and refers to NO_IRQ, or expects 0 to be a valid
>    needs to be fixed.
> 
> 2) Where we hit a problem, board code needs to be adapted to remap HW IRQs
>    0-15 to different software values.  (This could be done using irq
>    domains, or not)

No AMBA driver I'm aware of ever uses an IRQ number 0 or is passed such
an IRQ number.

[PATCH] ata: Don't use NO_IRQ in pata_of_platform driver

[Dave Martin]

On Tue, Dec 06, 2011 at 11:49:52AM +0000, Russell King - ARM Linux wrote:
> On Tue, Dec 06, 2011 at 11:37:35AM +0000, Dave Martin wrote:
> > 1) All OF code and drivers should be migrating to use 0 instead of NO_IRQ
> >    for the no-interrupt case.  Code which receives irq numbers directly
> >    from the OF framework and refers to NO_IRQ, or expects 0 to be a valid
> >    needs to be fixed.
> > 
> > 2) Where we hit a problem, board code needs to be adapted to remap HW IRQs
> >    0-15 to different software values.  (This could be done using irq
> >    domains, or not)
> 
> No AMBA driver I'm aware of ever uses an IRQ number 0 or is passed such
> an IRQ number.

OK, hopefully we can safely ignore that case, then.

But other than that, you're in agreement?

Cheers
---Dave

[PATCH] ata: Don't use NO_IRQ in pata_of_platform driver

[Rob Herring]

On 12/06/2011 05:49 AM, Russell King - ARM Linux wrote:
> On Tue, Dec 06, 2011 at 11:37:35AM +0000, Dave Martin wrote:
>> 1) All OF code and drivers should be migrating to use 0 instead of NO_IRQ
>>    for the no-interrupt case.  Code which receives irq numbers directly
>>    from the OF framework and refers to NO_IRQ, or expects 0 to be a valid
>>    needs to be fixed.
>>
>> 2) Where we hit a problem, board code needs to be adapted to remap HW IRQs
>>    0-15 to different software values.  (This could be done using irq
>>    domains, or not)
> 
> No AMBA driver I'm aware of ever uses an IRQ number 0 or is passed such
> an IRQ number.

The watchdog on VersatileAB is on Linux IRQ0. This is easily fixed with
VIC irqdomain patches which are queued up.

Rob

[PATCH] ata: Don't use NO_IRQ in pata_of_platform driver

[Linus Torvalds]

On Tue, Dec 6, 2011 at 2:46 AM, Russell King - ARM Linux
<linux@arm.linux.org.uk> wrote:
>
> But.. let's make one thing clear: Alan Cox and Linus have been going on
> about how IRQ0 should not be used. ?Let's be crystal clear: even x86
> uses IRQ0.

Not for any device driver, though.

It's used entirely internally, and it doesn't even use
"request_irq()". It uses the magic internal "setup_irq()" and never
*ever* exposes irq0 as anything that a driver can see.

That's what matters. You can use irq0 in ARM land all you like, AS
LONG AS IT'S SOME HIDDEN INTERNAL USE. No drivers. No *nothing* that
ever uses that absolutely *idiotic* NO_IRQ crap.

In fact, you may be *forced* to use what is "physically" irq0 - it's
just that you should never expose it as such to drivers. And x86
doesn't.

So Russell, if you think this has anything to do with NO_IRQ, and how
x86 isn't doing things right, you're wrong. It's just like the
internal exception thing, or the magical "cascade interrupt", or the
"x87 exception mapped through the PIC". They are magic hidden
interrupts that are set up in one place (well, one place *each*), and
are never exposed anywhere else.

The problem with NO_IRQ is that stupid "we expose our mind-numbingly
stupid interfaces across the whole kernel".

x86 never did that.  ARM still does. x86 doesn't have to fix anything. ARM does.

                Linus

[PATCH] ata: Don't use NO_IRQ in pata_of_platform driver

[Russell King - ARM Linux]

On Tue, Dec 06, 2011 at 11:20:49AM -0800, Linus Torvalds wrote:
> Not for any device driver, though.
> 
> It's used entirely internally, and it doesn't even use
> "request_irq()". It uses the magic internal "setup_irq()" and never
> *ever* exposes irq0 as anything that a driver can see.
> 
> That's what matters. You can use irq0 in ARM land all you like, AS
> LONG AS IT'S SOME HIDDEN INTERNAL USE. No drivers. No *nothing* that
> ever uses that absolutely *idiotic* NO_IRQ crap.
> 
> In fact, you may be *forced* to use what is "physically" irq0 - it's
> just that you should never expose it as such to drivers. And x86
> doesn't.
> 
> So Russell, if you think this has anything to do with NO_IRQ, and how
> x86 isn't doing things right, you're wrong. It's just like the
> internal exception thing, or the magical "cascade interrupt", or the
> "x87 exception mapped through the PIC". They are magic hidden
> interrupts that are set up in one place (well, one place *each*), and
> are never exposed anywhere else.
> 
> The problem with NO_IRQ is that stupid "we expose our mind-numbingly
> stupid interfaces across the whole kernel".
> 
> x86 never did that.  ARM still does. x86 doesn't have to fix anything. ARM does.

Remember you said that I shouldn't take things personally?  Well,
this is one issue I really don't care about.  I don't think any
platform I _actually_ have will be impacted by any change in this
area.  Other platform maintainers may have their own issues but
that's not _my_ problem.

[PATCH] ata: Don't use NO_IRQ in pata_of_platform driver

[Uwe Kleine-König]

Hello Linus,

On Tue, Dec 06, 2011 at 11:20:49AM -0800, Linus Torvalds wrote:
> On Tue, Dec 6, 2011 at 2:46 AM, Russell King - ARM Linux
> <linux@arm.linux.org.uk> wrote:
> >
> > But.. let's make one thing clear: Alan Cox and Linus have been going on
> > about how IRQ0 should not be used. ?Let's be crystal clear: even x86
> > uses IRQ0.
> 
> Not for any device driver, though.
> 
> It's used entirely internally, and it doesn't even use
> "request_irq()". It uses the magic internal "setup_irq()" and never
> *ever* exposes irq0 as anything that a driver can see.
> 
> That's what matters. You can use irq0 in ARM land all you like, AS
> LONG AS IT'S SOME HIDDEN INTERNAL USE. No drivers. No *nothing* that
> ever uses that absolutely *idiotic* NO_IRQ crap.
> 
> In fact, you may be *forced* to use what is "physically" irq0 - it's
> just that you should never expose it as such to drivers. And x86
> doesn't.
> 
> So Russell, if you think this has anything to do with NO_IRQ, and how
> x86 isn't doing things right, you're wrong. It's just like the
> internal exception thing, or the magical "cascade interrupt", or the
> "x87 exception mapped through the PIC". They are magic hidden
> interrupts that are set up in one place (well, one place *each*), and
> are never exposed anywhere else.
Well there is try_misrouted_irq in kernel/irq/spurious.c that assumes
irq0 to be something that it never is on ARM (and maybe all other
platforms apart from x86). So at least it's not internal to a single
(x86 specific) place.

I tried to patch that two years ago, but that only ended in people
saying "don't use irq0". I don't know if try_misrouted_irq sees hardware
irqs, but if it does it's a bug even on archs != X86 that use virtual
irqs.

(Note that this doesn't oppose to your statement that using NO_IRQ is
crap.)

> The problem with NO_IRQ is that stupid "we expose our mind-numbingly
> stupid interfaces across the whole kernel".
> 
> x86 never did that.  ARM still does. x86 doesn't have to fix anything. ARM does.

Best regards
Uwe

-- 
Pengutronix e.K.                           | Uwe Kleine-K?nig            |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |

[PATCH] ata: Don't use NO_IRQ in pata_of_platform driver

[Nicolas Pitre]

On Tue, 6 Dec 2011, Dave Martin wrote:

> On Mon, Dec 05, 2011 at 02:49:01PM -0500, Nicolas Pitre wrote:
> 
> > No need to convert everything.
> > 
> > First move is to make irq=0 meaning no IRQ.  That means making things 
> > like:
> > 
> > 	if (irq < 0)
> > 	if (irq >= 0)
> > 
> > into
> > 
> > 	if (irq <= 0)
> > 	if (irq > 0)
> > 
> > And replace NO_IRQ with 0.
> >
> > That change shouldn't break anything, except those drivers which are 1) 
> > being passed an actual IRQ #0 and 2) testing for no IRQ.  I suspect that 
> > those conditions aren't very common together.
> 
> To clarify, you're suggesting that the meanings of all other IRQ values
> would not change initially?

Initially, or even ever.

> (i.e., we remap HW irq 0, if there is one,
> to some other random number but have a 1:1 mapping for everything else).

Exact.

> That could make sense as an approach.

You might notice that a true IRQ #0 passed to generic drivers is not 
really frequent.


Nicolas

[PATCH] ata: Don't use NO_IRQ in pata_of_platform driver

[Alan Cox]

> Russell, do you know whether it would make sense to set a timeline for 
> removing NO_IRQ from ARM platforms and migrating to 0 for the no-interrupt
> case?  I'm assuming that this mainly involves migrating existing hard-wired
> code that deals with interrupt numbers to use irq domains.

The timelime was several years ago. Several years of complete inaction
later the chickens have come home to roost.

I refer you to Linus mail of 26 Sept 2006 to linux-kernel ('restore
libata build on frv')

Quoting Linus email:

>> That's fine -- but don't use zero to mean none. We have NO_IRQ for
>> that, and zero isn't an appropriate choice.
>
> Zero _is_ an appropriate choice, dammit!
>
> That NO_IRQ thing should be zero, and any architecture that thinks that
>zero is a valid IRQ just needs to fix its own irq mapping so that the
> "cookie" doesn't work.
>
> The thing is, it's zero. Get over it. It can't be "-1" or some other
> random value like people have indicated, because that thing is often read
> from places where "-1" simply isn't a possible value (eg it gets its
> default value initialized from a "unsigned char" in MMIO space on x86).
>
> So instead of making everybody and their dog to silly things with some
> NO_IRQ define that they haven't historically done, the rule is simple: "0"
> means "no irq", so that you can test for it with obvious code like
>
>	if (!dev->irq)
>		..
>
> and then, if your actual _hardware_ things that the bit-pattern with all
> bits clear is a valid irq that can be used for normal devices, then what
> you do is you add a irq number translation layer (WHICH WE NEED AND HAVE
> _ANYWAY_) and make sure that nobody sees that on a _software_ level.

----------

On 15th October 2008 Linus said the following to linux-next

> Grr. Can we please just get rid of that IDIOTIC thing instead?
> 
> NO_IRQ was a bad idea to begin with. Let's not add more.
> 
> I assume that broken driver is some ARM-specific thing. I certainly don't 
> want to see NO_IRQ in any general drivers. So instead of having that 
> NO_IRQ insanity spread any more, I'd much rather see the driver either 
> fixed to not use it, or just marked ARM-only.
>
> The proper way to test for whether an interrupt is valid or not is to do
>
> 	if (dev->irq) {
>		...
> and no other. There is no spoon. That NO_IRQ was insane. And
> architectures or drivers that still think otherwise should fix themselves.

------------

So there we are.. ARM spent years ignoring clear direction. If ARM breaks
for a release now so be it. You've had *YEARS* to get off your collective
backsides and sort it out.

> I worry that if we just change the convention for the OF case, we'll end
> up with OF-ised platform drivers which have to deal with a different no-
> irq convention depending on whether they are probed as platform drivers
> or through the OF framework ... and these ported or semi-ported drivers
> will be intermixed with unported drivers, confusing maintainers

All drivers should assume that  if (!dev->irq) works. Zero is not an IRQ
except in certain buried internal invisible cases in arch code (legacy PC
timer being the obvious one).

Come to think about it we had a prior discussion about NO_IRQ in 2005
even!

The core kernel generic IRQ code knows about zero being special, many
common driver layer components such as serial and network phylib do, so
if anything it's going to fix bugs sorting the mess out on ARM.

Jut fix it. Other platforms have done so without problem.

Alan