[PATCH] x86: ioapic: Fix irq_free_descs count

[PATCH] x86: ioapic: Fix irq_free_descs count

[Keith Busch]

Signed-off-by: Keith Busch <keith.busch@intel.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: x86@kernel.org
---
 kernel/irq/irqdesc.c |    4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/kernel/irq/irqdesc.c b/kernel/irq/irqdesc.c
index 7339e42..1487a12 100644
--- a/kernel/irq/irqdesc.c
+++ b/kernel/irq/irqdesc.c
@@ -455,9 +455,9 @@ EXPORT_SYMBOL_GPL(irq_alloc_hwirqs);
  */
 void irq_free_hwirqs(unsigned int from, int cnt)
 {
-	int i;
+	int i, j;
 
-	for (i = from; cnt > 0; i++, cnt--) {
+	for (i = from, j = cnt; j > 0; i++, j--) {
 		irq_set_status_flags(i, _IRQ_NOREQUEST | _IRQ_NOPROBE);
 		arch_teardown_hwirq(i);
 	}
-- 
1.7.10.4

Re: [PATCH] x86: ioapic: Fix irq_free_descs count

[David Rientjes]

On Mon, 30 Jun 2014, Keith Busch wrote:

> Signed-off-by: Keith Busch <keith.busch@intel.com>
> Cc: Thomas Gleixner <tglx@linutronix.de>
> Cc: x86@kernel.org

Acked-by: David Rientjes <rientjes@google.com>

This is definitely a fix for "genirq: Provide generic hwirq allocation 
facility", but the changelog doesn't describe what the problem is and the 
title that this somehow fixes irq_free_descs() doesn't make any sense.

It's the equivalent of just doing

-	irq_free_descs(from, cnt);
+	irq_free_descs(from, i - from);

I'd suggest the patch title be changed to

	genirq: Fix memory leak when calling irq_free_hwirqs()

and the changelog state

irq_free_hwirqs() always calls irq_free_descs() with a cnt == 0 which 
makes it a no-op since the interrupt count to free is decremented in 
itself.

Re: [PATCH] x86: ioapic: Fix irq_free_descs count

[Keith Busch]

On Mon, 30 Jun 2014, David Rientjes wrote:
> On Mon, 30 Jun 2014, Keith Busch wrote:
>
>> Signed-off-by: Keith Busch <keith.busch@intel.com>
>> Cc: Thomas Gleixner <tglx@linutronix.de>
>> Cc: x86@kernel.org
>
> Acked-by: David Rientjes <rientjes@google.com>
>
> This is definitely a fix for "genirq: Provide generic hwirq allocation
> facility", but the changelog doesn't describe what the problem is and the
> title that this somehow fixes irq_free_descs() doesn't make any sense.

My mistake, I used the component from the commit I bisected the bug
down to.

> It's the equivalent of just doing
>
> -	irq_free_descs(from, cnt);
> +	irq_free_descs(from, i - from);
>
> I'd suggest the patch title be changed to
>
> 	genirq: Fix memory leak when calling irq_free_hwirqs()
>
> and the changelog state
>
> irq_free_hwirqs() always calls irq_free_descs() with a cnt == 0 which
> makes it a no-op since the interrupt count to free is decremented in
> itself.

Yes, I like your suggestion. Should I submit a v2, or will maintainer
replace the changelog?

Re: [PATCH] x86: ioapic: Fix irq_free_descs count

[David Rientjes]

On Mon, 30 Jun 2014, Keith Busch wrote:

> On Mon, 30 Jun 2014, David Rientjes wrote:
> > On Mon, 30 Jun 2014, Keith Busch wrote:
> > 
> > > Signed-off-by: Keith Busch <keith.busch@intel.com>
> > > Cc: Thomas Gleixner <tglx@linutronix.de>
> > > Cc: x86@kernel.org
> > 
> > Acked-by: David Rientjes <rientjes@google.com>
> > 
> > This is definitely a fix for "genirq: Provide generic hwirq allocation
> > facility", but the changelog doesn't describe what the problem is and the
> > title that this somehow fixes irq_free_descs() doesn't make any sense.
> 
> My mistake, I used the component from the commit I bisected the bug
> down to.
> 
> > It's the equivalent of just doing
> > 
> > -	irq_free_descs(from, cnt);
> > +	irq_free_descs(from, i - from);
> > 
> > I'd suggest the patch title be changed to
> > 
> > 	genirq: Fix memory leak when calling irq_free_hwirqs()
> > 
> > and the changelog state
> > 
> > irq_free_hwirqs() always calls irq_free_descs() with a cnt == 0 which
> > makes it a no-op since the interrupt count to free is decremented in
> > itself.
> 
> Yes, I like your suggestion. Should I submit a v2, or will maintainer
> replace the changelog?
> 

It's an urgent patch since commit 7b6ef1262549 ("genirq: Provide generic 
hwirq allocation facility") was already released in 3.15.  Unless Thomas 
fixes it up first, I'd reply with a v2 with the fixed title and changelog, 
add a

	Fixes: 7b6ef1262549f6afc5c881aaef80beb8fd15f908

tag along with my acked-by to make his life 0.000000001% easier.