[PATCH] Add irq host and hwirq pair fields to output when showing interrupts.

Signed-off-by: Zhang Wei <redacted>
---
 arch/powerpc/kernel/irq.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/arch/powerpc/kernel/irq.c b/arch/powerpc/kernel/irq.c
index 919fbf5..2a30ffd 100644
--- a/arch/powerpc/kernel/irq.c
+++ b/arch/powerpc/kernel/irq.c

@@ -194,7 +194,7 @@ #else
 		seq_printf(p, "%10u ", kstat_irqs(i));
 #endif /* CONFIG_SMP */
 		if (desc->chip)
-			seq_printf(p, " %s ", desc->chip->typename);
+			seq_printf(p, " (%3d) @%s ", (irq_map[i].host->revmap_type == IRQ_HOST_MAP_LEGACY) ? i : virq_to_hw(i), desc->chip->typename);
 		else
 			seq_puts(p, "  None      ");
 		seq_printf(p, "%s", (desc->status & IRQ_LEVEL) ? "Level " : "Edge  ");

-- 
1.4.0

On Tue, 2007-03-13 at 11:19 +0800, Zhang Wei wrote:

Well, first, we shouldn't modify the output of /proc/interrupts as this
will likely break all sort of userland tools displaying it (well, I
think).

Now, regarding the patch itself, printing the hw irq is not that useful
without some indication of which host it belongs to. Best is to add a
name() or something like that to the host ops.

Ben.

On Tue, 2007-03-13 at 11:19 +0800, Zhang Wei wrote:

quoted

Signed-off-by: Zhang Wei <redacted>
---
 arch/powerpc/kernel/irq.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
=20

diff --git a/arch/powerpc/kernel/irq.c b/arch/powerpc/kernel/irq.c
index 919fbf5..2a30ffd 100644
--- a/arch/powerpc/kernel/irq.c
+++ b/arch/powerpc/kernel/irq.c

@@ -194,7 +194,7 @@ #else
 		seq_printf(p, "%10u ", kstat_irqs(i));
 #endif /* CONFIG_SMP */
 		if (desc->chip)
-			seq_printf(p, " %s ", desc->chip->typename);
+			seq_printf(p, " (%3d) @%s ",=20

(irq_map[i].host->revmap_type =3D=3D IRQ_HOST_MAP_LEGACY) ? i :=20
virq_to_hw(i), desc->chip->typename);

quoted

 		else
 			seq_puts(p, "  None      ");
 		seq_printf(p, "%s", (desc->status & IRQ_LEVEL)=20

? "Level " : "Edge  ");
=20
Well, first, we shouldn't modify the output of=20
/proc/interrupts as this
will likely break all sort of userland tools displaying it (well, I
think).
=20
Now, regarding the patch itself, printing the hw irq is not=20
that useful
without some indication of which host it belongs to. Best is to add a
name() or something like that to the host ops.
=20

Sure, can I use the PIC controller node full name from of_tree as the
host name?

Do you prefer debugfs instead of /proc/interrupts?

Thanks!
Zhang Wei

quoted

Now, regarding the patch itself, printing the hw irq is not 
that useful
without some indication of which host it belongs to. Best is to add a
name() or something like that to the host ops.

Sure, can I use the PIC controller node full name from of_tree as the
host name?

There is no 1:1 relationship between an OF node and an irq host. (There
is -most- of the time but I left the irq core on purpose more flexible,
hence the match() method. On some setups it makes sense to have multiple
HW PICs share a single domain number (for example on cell).

Thus I'd rather have a way for the host to give you that name, which in
turn can be the OF path if there is such a thing, or something else...

Best might be 4 columns:

<virq> <hwirq> <pic name> <host name>

(PIC name is the same as in /proc/interrupts).
 

Do you prefer debugfs instead of /proc/interrupts?

Yes, at least for now. Maybe if we can convince ourselves that nothing
significant would break of changing /proc/interrupts, then we can add
stuffs to it, but I'd rather avoid.

Ben.

quoted

quoted

Now, regarding the patch itself, printing the hw irq is not=20
that useful
without some indication of which host it belongs to. Best=20

is to add a

quoted

quoted

name() or something like that to the host ops.
=20

=20
Sure, can I use the PIC controller node full name from=20

of_tree as the

quoted

host name?

=20
There is no 1:1 relationship between an OF node and an irq=20
host. (There
is -most- of the time but I left the irq core on purpose more=20
flexible,
hence the match() method. On some setups it makes sense to=20
have multiple
HW PICs share a single domain number (for example on cell).
=20
Thus I'd rather have a way for the host to give you that=20
name, which in
turn can be the OF path if there is such a thing, or something else...
=20
Best might be 4 columns:
=20
<virq> <hwirq> <pic name> <host name>
=20
(PIC name is the same as in /proc/interrupts).
=20

quoted

Do you prefer debugfs instead of /proc/interrupts?

=20
Yes, at least for now. Maybe if we can convince ourselves that nothing
significant would break of changing /proc/interrupts, then we can add
stuffs to it, but I'd rather avoid.
=20

Hi, Ben,=20

How about set the host name when calling irq_alloc_host?
In the mpic, the host name can be the mpic->name, is it ok?

Thanks!
Best Regards,
Zhang Wei

quoted

Yes, at least for now. Maybe if we can convince ourselves that nothing
significant would break of changing /proc/interrupts, then we can add
stuffs to it, but I'd rather avoid.

Hi, Ben, 

How about set the host name when calling irq_alloc_host?
In the mpic, the host name can be the mpic->name, is it ok?

For MPIC it should probably be the OF full path