Jeff Garzik posted a series of nine patchs to the lkml titled to "remove [the] 'irq' argument from all irq handlers", explaining, "the overwhelming majority of drivers do not ever bother with the 'irq' argument that is passed to each driver's irq handler. Of the minority of drivers that do use the arg, the majority of those have the irq number stored in their private-info structure somewhere." He noted that he had no intention to push the patches upstream anytime soon.
Feedback was entirely positive, with Thomas Gleixner suggesting, "Full ACK. We should do this right at the edge of -rc1. And let's do this right now in .24 and not drag it out for no good reason." Ingo Molnar concurred, "full ACK on the concept from me too. Please go ahead! :)" Eric Biederman noted that there was still work to be done, "the practical question is how do we make this change without breaking the drivers that use their irq argument." Jeff agreed, explaining why the code won't be pushed upstream during -rc1, "I am finding a ton of bugs in each get_irqfunc_irq() driver, so I would rather patiently sift through them, and push fixes and cleanups upstream. Once that effort is done, everything should be in the 'trivial' pile and not have the logic that you are worried about (and thus there would be no need to add an additional branch to the irq handling path)."
From: Jeff Garzik
Subject: [PATCH 0/9] Remove 'irq' argument from all irq handlers
Date: Oct 19, 12:54 am 2007
WARNING NOT FOR MERGE WARNING NOT FOR MERGE WARNING NOT FOR MERGE
This posting is just to demonstrate something that I have been keeping
alive in the background. I have no urge to push it upstream anytime
soon.
The overwhelming majority of drivers do not ever bother with the 'irq'
argument that is passed to each driver's irq handler.
Of the minority of drivers that do use the arg, the majority of those
have the irq number stored in their private-info structure somewhere.
There are a tiny few -- a couple Mac drivers -- which do weird things
with that argument, but that's it.
For the large sweeps through the tree, these patches are grouped into
"trivial" changes -- simply removing the unused irq arg -- or all other
changes.
[IRQ ARG REMOVAL] core interrupt delivery infrastructure updates
[IRQ ARG REMOVAL] various non-trivial arch updates
[IRQ ARG REMOVAL] trivial arch updates
[IRQ ARG REMOVAL] non-trivial driver updates
[IRQ ARG REMOVAL] trivial net driver updates
[IRQ ARG REMOVAL] trivial sound driver updates
[IRQ ARG REMOVAL] trivial scsi driver updates
[IRQ ARG REMOVAL] trivial driver updates
[IRQ ARG REMOVAL] x86-64 build fixes, cleanups
WARNING NOT FOR MERGE WARNING NOT FOR MERGE WARNING NOT FOR MERGE
-
From: Greg KH
Subject: Re: [PATCH 0/9] Remove 'irq' argument from all irq handlers
Date: Oct 19, 11:07 pm 2007
On Fri, Oct 19, 2007 at 03:54:43AM -0400, Jeff Garzik wrote:
>
> The overwhelming majority of drivers do not ever bother with the 'irq'
> argument that is passed to each driver's irq handler.
>
> Of the minority of drivers that do use the arg, the majority of those
> have the irq number stored in their private-info structure somewhere.
>
> There are a tiny few -- a couple Mac drivers -- which do weird things
> with that argument, but that's it.
>
> For the large sweeps through the tree, these patches are grouped into
> "trivial" changes -- simply removing the unused irq arg -- or all other
> changes.
Very cool stuff, I like it :)
greg k-h
-
From: Thomas Gleixner
Subject: Re: [PATCH 0/9] Remove 'irq' argument from all irq handlers
Date: Oct 19, 7:53 am 2007
On Fri, 19 Oct 2007, Jeff Garzik wrote:
>
> WARNING NOT FOR MERGE WARNING NOT FOR MERGE WARNING NOT FOR MERGE
>
> This posting is just to demonstrate something that I have been keeping
> alive in the background. I have no urge to push it upstream anytime
> soon.
>
> The overwhelming majority of drivers do not ever bother with the 'irq'
> argument that is passed to each driver's irq handler.
>
> Of the minority of drivers that do use the arg, the majority of those
> have the irq number stored in their private-info structure somewhere.
>
> There are a tiny few -- a couple Mac drivers -- which do weird things
> with that argument, but that's it.
Jeff,
thanks for doing this.
Full ACK.
We should do this right at the edge of -rc1. And let's do this right
now in .24 and not drag it out for no good reason.
Thanks,
tglx
-
From: Eric W. Biederman
Subject: Re: [PATCH 0/9] Remove 'irq' argument from all irq handlers
Date: Oct 19, 11:38 am 2007
Thomas Gleixner <tglx@linutronix.de> writes:
> On Fri, 19 Oct 2007, Jeff Garzik wrote:
>>
>> WARNING NOT FOR MERGE WARNING NOT FOR MERGE WARNING NOT FOR MERGE
>>
>> This posting is just to demonstrate something that I have been keeping
>> alive in the background. I have no urge to push it upstream anytime
>> soon.
>>
>> The overwhelming majority of drivers do not ever bother with the 'irq'
>> argument that is passed to each driver's irq handler.
>>
>> Of the minority of drivers that do use the arg, the majority of those
>> have the irq number stored in their private-info structure somewhere.
>>
>> There are a tiny few -- a couple Mac drivers -- which do weird things
>> with that argument, but that's it.
>
> Jeff,
>
> thanks for doing this.
Yes. keeping this alive is good.
The practical question is how do we make this change without breaking
the drivers that use their irq argument.
> Full ACK.
>
> We should do this right at the edge of -rc1. And let's do this right
> now in .24 and not drag it out for no good reason.
There are two problems with that suggestion.
- We don't have all of the architectures converted.
- We don't have a solid plan for how to keep drivers that are
using the irq parameter today working.
I don't think the irq argument is something we want to keep
around forever, and I certainly don't see the need to do
the error prone get_irqfunc_irq and set_irqfunc_irq logic.
How about:
irqreturn_t handle_IRQ_event(unsigned int irq, struct irqaction *action)
{
irqreturn_t ret, retval = IRQ_NONE;
unsigned int status = 0;
handle_dynamic_tick(action);
if (!(action->flags & IRQF_DISABLED))
local_irq_enable_in_hardirq();
do {
if (action->flags & IRQF_VERBOSE)
ret = action->handler_verbose(irq, action->dev_id);
else
ret = action->handler(action->dev_id);
if (ret == IRQ_HANDLED)
status |= action->flags;
retval |= ret;
action = action->next;
} while (action);
if (status & IRQF_SAMPLE_RANDOM)
add_interrupt_randomness(irq);
local_irq_disable();
return retval;
}
And then:
typedef irqreturn_t (*irq_handler_verbose_t)(int, void *);
typedef irqreturn_t (*irq_handler_t)(void *);
struct irqaction {
union {
irq_handler_verbose_t handler_verbose;
irq_handler_t handler;
};
unsigned long flags;
cpumask_t mask;
const char *name;
void *dev_id;
struct irqaction *next;
int irq;
struct proc_dir_entry *dir;
};
int request_irq(unsigned int irq, irq_handler_t handler,
unsigned long irqflags, const char *devname, void *dev_id)
int request_verbose_irq(unsigned int irq, irq_handler_verbose_t handler,
unsigned long irqflags, const char *devname, void *dev_id)
Then we just need to go through all of the drivers and either change
their interrupt handler prototype or change the flavor of request_irq.
It is a bit of a pain but we should be able to do that without breaking
any drivers. Which it looks like we are in real danger of if someone
goes through them all interrupt handlers that are using the irq
argument and change them all at once. Especially since most of those
drivers are old an rarely used right now.
Eric
-
From: Jeff Garzik
Subject: Re: [PATCH 0/9] Remove 'irq' argument from all irq handlers
Date: Oct 19, 11:57 am 2007
Eric W. Biederman wrote:
> Yes. keeping this alive is good.
>
> The practical question is how do we make this change without breaking
> the drivers that use their irq argument.
This is why I'm taking it slow, and not rushing to get this upstream :)
I am finding a ton of bugs in each get_irqfunc_irq() driver, so I would
rather patiently sift through them, and push fixes and cleanups upstream.
Once that effort is done, everything should be in the 'trivial' pile and
not have the logic that you are worried about (and thus there would be
no need to add an additional branch to the error handling path).
Jeff
-
From: Jeff Garzik
Subject: Re: [PATCH 0/9] Remove 'irq' argument from all irq handlers
Date: Oct 19, 12:02 pm 2007
Jeff Garzik wrote:
> Once that effort is done, everything should be in the 'trivial' pile and
> not have the logic that you are worried about (and thus there would be
> no need to add an additional branch to the error handling path).
er, s/error/irq/
the perils of a multi-threaded brain...
-
From: Ingo Molnar
Subject: Re: [PATCH 0/9] Remove 'irq' argument from all irq handlers
Date: Oct 19, 12:07 pm 2007
* Eric W. Biederman <ebiederm@xmission.com> wrote:
> > thanks for doing this.
>
> Yes. keeping this alive is good.
>
> The practical question is how do we make this change without breaking
> the drivers that use their irq argument.
the get_irq_regs() approach worked out really well. We should do a
get_irq_nr() and be done with it?
(Then once the mechanic conversion has been done we could eliminate all
uses of get_irq_nr() and remove the small overhead it takes to maintain
this per-cpu value in the irq entry layer.)
full ACK on the concept from me too. Please go ahead! :)
Ingo
-
From: Eric W. Biederman
Subject: Re: [PATCH 0/9] Remove 'irq' argument from all irq handlers
Date: Oct 19, 12:35 pm 2007
Ingo Molnar <mingo@elte.hu> writes:
> * Eric W. Biederman <ebiederm@xmission.com> wrote:
>
>> > thanks for doing this.
>>
>> Yes. keeping this alive is good.
>>
>> The practical question is how do we make this change without breaking
>> the drivers that use their irq argument.
>
> the get_irq_regs() approach worked out really well. We should do a
> get_irq_nr() and be done with it?
The problem are some drivers today pass in 0 for their irq number
to flag that they are calling the interrupt handler in a polling
mode (not from interrupt context?) so the same logic doesn't quite apply.
Do what you suggest would likely break those drivers.
Eric
-
From: Thomas Gleixner
Subject: Re: [PATCH 0/9] Remove 'irq' argument from all irq handlers
Date: Oct 19, 12:41 pm 2007
On Fri, 19 Oct 2007, Eric W. Biederman wrote:
> Ingo Molnar <mingo@elte.hu> writes:
>
> > * Eric W. Biederman <ebiederm@xmission.com> wrote:
> >
> >> > thanks for doing this.
> >>
> >> Yes. keeping this alive is good.
> >>
> >> The practical question is how do we make this change without breaking
> >> the drivers that use their irq argument.
> >
> > the get_irq_regs() approach worked out really well. We should do a
> > get_irq_nr() and be done with it?
>
> The problem are some drivers today pass in 0 for their irq number
> to flag that they are calling the interrupt handler in a polling
> mode (not from interrupt context?) so the same logic doesn't quite apply.
>
> Do what you suggest would likely break those drivers.
How many of them do we have ? This is a wilful abuse of the API, so
its not a big damage if they break.
tglx
-
From: Jeff Garzik
Subject: Re: [PATCH 0/9] Remove 'irq' argument from all irq handlers
Date: Oct 19, 12:55 pm 2007
Thomas Gleixner wrote:
> On Fri, 19 Oct 2007, Eric W. Biederman wrote:
>> Ingo Molnar <mingo@elte.hu> writes:
>>
>>> * Eric W. Biederman <ebiederm@xmission.com> wrote:
>>>
>>>>> thanks for doing this.
>>>> Yes. keeping this alive is good.
>>>>
>>>> The practical question is how do we make this change without breaking
>>>> the drivers that use their irq argument.
>>> the get_irq_regs() approach worked out really well. We should do a
>>> get_irq_nr() and be done with it?
>> The problem are some drivers today pass in 0 for their irq number
>> to flag that they are calling the interrupt handler in a polling
>> mode (not from interrupt context?) so the same logic doesn't quite apply.
>>
>> Do what you suggest would likely break those drivers.
>
> How many of them do we have ? This is a wilful abuse of the API, so
> its not a big damage if they break.
I would prefer to simply clean up the drivers such that
get_irqfunc_irq() and set_irqfunc_irq() are not needed.
One of the many reasons why I'm explicitly not pushing it upstream yet :)
Jeff
-
