[PATCH] asm-generic: Raise default NR_IRQS when using sparse IRQs

[PATCH] asm-generic: Raise default NR_IRQS when using sparse IRQs

[Mark Brown]

Rather than have each platform using sparse IRQs pick a suitably large
NR_IRQS for use with sparse IRQs make the default high when they are
enabled.  We pick 64k as there is still a bitmap of IRQs that is
allocated statically, and as we all know 64k should be enough for
anyone.

Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
---
 include/asm-generic/irq.h |    6 +++++-
 1 files changed, 5 insertions(+), 1 deletions(-)

diff --git a/include/asm-generic/irq.h b/include/asm-generic/irq.h
index b90ec0b..0d95650 100644
--- a/include/asm-generic/irq.h
+++ b/include/asm-generic/irq.h
@@ -4,11 +4,15 @@
 /*
  * NR_IRQS is the upper bound of how many interrupts can be handled
  * in the platform. It is used to size the static irq_map array,
- * so don't make it too big.
+ * so don't make it too big unless we're using sparse IRQs.
  */
 #ifndef NR_IRQS
+#ifdef CONFIG_SPARSE_IRQ
+#define NR_IRQS 0xffff
+#else
 #define NR_IRQS 64
 #endif
+#endif
 
 static inline int irq_canonicalize(int irq)
 {
-- 
1.7.5.4

Re: [PATCH] asm-generic: Raise default NR_IRQS when using sparse IRQs

[Thomas Gleixner]

On Wed, 29 Jun 2011, Mark Brown wrote:

> Rather than have each platform using sparse IRQs pick a suitably large
> NR_IRQS for use with sparse IRQs make the default high when they are
> enabled.  We pick 64k as there is still a bitmap of IRQs that is
> allocated statically, and as we all know 64k should be enough for
> anyone.

The sparse bitmap is overallocated to NR_IRQS + 8k anyway, so that's
rather pointless.
 
> Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
> ---
>  include/asm-generic/irq.h |    6 +++++-
>  1 files changed, 5 insertions(+), 1 deletions(-)
> 
> diff --git a/include/asm-generic/irq.h b/include/asm-generic/irq.h
> index b90ec0b..0d95650 100644
> --- a/include/asm-generic/irq.h
> +++ b/include/asm-generic/irq.h
> @@ -4,11 +4,15 @@
>  /*
>   * NR_IRQS is the upper bound of how many interrupts can be handled
>   * in the platform. It is used to size the static irq_map array,
> - * so don't make it too big.
> + * so don't make it too big unless we're using sparse IRQs.
>   */
>  #ifndef NR_IRQS
> +#ifdef CONFIG_SPARSE_IRQ
> +#define NR_IRQS 0xffff
> +#else
>  #define NR_IRQS 64
>  #endif
> +#endif
>  
>  static inline int irq_canonicalize(int irq)
>  {
> -- 
> 1.7.5.4
> 
> --
> To unsubscribe from this list: send the line "unsubscribe linux-arch" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> 

Re: [PATCH] asm-generic: Raise default NR_IRQS when using sparse IRQs

[Mark Brown]

On Thu, Jun 30, 2011 at 09:27:20AM +0200, Thomas Gleixner wrote:
> On Wed, 29 Jun 2011, Mark Brown wrote:

> > Rather than have each platform using sparse IRQs pick a suitably large
> > NR_IRQS for use with sparse IRQs make the default high when they are
> > enabled.  We pick 64k as there is still a bitmap of IRQs that is
> > allocated statically, and as we all know 64k should be enough for
> > anyone.

> The sparse bitmap is overallocated to NR_IRQS + 8k anyway, so that's
> rather pointless.

Hrm, OK.  In that case why are platforms faffing around trying to size
NR_IRQs in the first place?

Re: [PATCH] asm-generic: Raise default NR_IRQS when using sparse IRQs

[Mark Brown]

On Mon, Nov 29, 2010 at 11:02:53PM +0300, Pavel Vasilyev wrote:
> On 29.11.2010 22:20, Mark Brown wrote:
> > Rather than have each platform using sparse IRQs pick a suitably large
> > NR_IRQS for use with sparse IRQs make the default high when they are
> > enabled.  We pick 64k as there is still a bitmap of IRQs that is
> > allocated statically, and as we all know 64k should be enough for
> > anyone.

> May be

>   #define NR_IRQS 15*NR_CPUS

No, this is missing the point - that's way too low for many systems
(most except x86 I guess).  The idea is that since we're using sparse
IRQs we've much less reason to restrict the number of IRQs than we do
when we've got a statically allocated table so we can have a higher
default, meaning less need for platforms to individually select a higher
number.

Re: [PATCH] asm-generic: Raise default NR_IRQS when using sparse IRQs

[Pavel Vasilyev]

On 29.11.2010 22:20, Mark Brown wrote:
> Rather than have each platform using sparse IRQs pick a suitably large
> NR_IRQS for use with sparse IRQs make the default high when they are
> enabled.  We pick 64k as there is still a bitmap of IRQs that is
> allocated statically, and as we all know 64k should be enough for
> anyone.
> 
> Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
> ---
>  include/asm-generic/irq.h |    6 +++++-
>  1 files changed, 5 insertions(+), 1 deletions(-)
> 
> diff --git a/include/asm-generic/irq.h b/include/asm-generic/irq.h
> index b90ec0b..0d95650 100644
> --- a/include/asm-generic/irq.h
> +++ b/include/asm-generic/irq.h
> @@ -4,11 +4,15 @@
>  /*
>   * NR_IRQS is the upper bound of how many interrupts can be handled
>   * in the platform. It is used to size the static irq_map array,
> - * so don't make it too big.
> + * so don't make it too big unless we're using sparse IRQs.
>   */
>  #ifndef NR_IRQS
> +#ifdef CONFIG_SPARSE_IRQ
> +#define NR_IRQS 0xffff
> +#else
>  #define NR_IRQS 64
>  #endif
> +#endif
>  
>  static inline int irq_canonicalize(int irq)
>  {

May be

  #define NR_IRQS 15*NR_CPUS

[PATCH] asm-generic: Raise default NR_IRQS when using sparse IRQs

[Mark Brown]

Rather than have each platform using sparse IRQs pick a suitably large
NR_IRQS for use with sparse IRQs make the default high when they are
enabled.  We pick 64k as there is still a bitmap of IRQs that is
allocated statically, and as we all know 64k should be enough for
anyone.

Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
---
 include/asm-generic/irq.h |    6 +++++-
 1 files changed, 5 insertions(+), 1 deletions(-)

diff --git a/include/asm-generic/irq.h b/include/asm-generic/irq.h
index b90ec0b..0d95650 100644
--- a/include/asm-generic/irq.h
+++ b/include/asm-generic/irq.h
@@ -4,11 +4,15 @@
 /*
  * NR_IRQS is the upper bound of how many interrupts can be handled
  * in the platform. It is used to size the static irq_map array,
- * so don't make it too big.
+ * so don't make it too big unless we're using sparse IRQs.
  */
 #ifndef NR_IRQS
+#ifdef CONFIG_SPARSE_IRQ
+#define NR_IRQS 0xffff
+#else
 #define NR_IRQS 64
 #endif
+#endif
 
 static inline int irq_canonicalize(int irq)
 {
-- 
1.7.1

Re: [PATCH] asm-generic: Raise default NR_IRQS when using sparse IRQs

[Mark Brown]

On Wed, Nov 24, 2010 at 04:45:03PM +0100, Thomas Gleixner wrote:

> Resulting in a 256 MB bitmap. I guess you did not try to boot that on
> some random ARM board :)

Right, I don't actually have any platforms that support sparse IRQs.  Ho
hum.  I guess something like 16k or 4k IRQs might be more sane.

Re: [PATCH] asm-generic: Raise default NR_IRQS when using sparse IRQs

[Thomas Gleixner]

On Wed, 24 Nov 2010, Mark Brown wrote:

> Rather than have each platform using sparse IRQs pick a suitably large
> NR_IRQS for use with sparse IRQs make the default exceptionally high
> when they are enabled.
> 
> Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
> ---
>  include/asm-generic/irq.h |    6 +++++-
>  1 files changed, 5 insertions(+), 1 deletions(-)
> 
> diff --git a/include/asm-generic/irq.h b/include/asm-generic/irq.h
> index b90ec0b..a76bea2 100644
> --- a/include/asm-generic/irq.h
> +++ b/include/asm-generic/irq.h
> @@ -4,11 +4,15 @@
>  /*
>   * NR_IRQS is the upper bound of how many interrupts can be handled
>   * in the platform. It is used to size the static irq_map array,
> - * so don't make it too big.
> + * so don't make it too big unless we're using sparse IRQs.
>   */
>  #ifndef NR_IRQS
> +#ifdef CONFIG_SPARSE_IRQ
> +#define NR_IRQS 0x7fffffff

Resulting in a 256 MB bitmap. I guess you did not try to boot that on
some random ARM board :)

Thanks,

	tglx

[PATCH] asm-generic: Raise default NR_IRQS when using sparse IRQs

[Mark Brown]

Rather than have each platform using sparse IRQs pick a suitably large
NR_IRQS for use with sparse IRQs make the default exceptionally high
when they are enabled.

Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
---
 include/asm-generic/irq.h |    6 +++++-
 1 files changed, 5 insertions(+), 1 deletions(-)

diff --git a/include/asm-generic/irq.h b/include/asm-generic/irq.h
index b90ec0b..a76bea2 100644
--- a/include/asm-generic/irq.h
+++ b/include/asm-generic/irq.h
@@ -4,11 +4,15 @@
 /*
  * NR_IRQS is the upper bound of how many interrupts can be handled
  * in the platform. It is used to size the static irq_map array,
- * so don't make it too big.
+ * so don't make it too big unless we're using sparse IRQs.
  */
 #ifndef NR_IRQS
+#ifdef CONFIG_SPARSE_IRQ
+#define NR_IRQS 0x7fffffff
+#else
 #define NR_IRQS 64
 #endif
+#endif
 
 static inline int irq_canonicalize(int irq)
 {
-- 
1.7.1