A couple of cleanups for arch_init_irq()
This is a follow-up for [1].
Here:
- drop unused 'i' variable, as reported by kernel robot [2];
- fix typos in commit message;
- add 2nd patch to handle WARN_ON() better.
[1] https://lore.kernel.org/lkml/[email protected]/
[2] https://lore.kernel.org/oe-kbuild-all/[email protected]/
Yury Norov (2):
MIPS: SGI-IP27: micro-optimize arch_init_irq()
MIPS: SGI-IP27: use WARN_ON() output
arch/mips/sgi-ip27/ip27-irq.c | 15 +++++----------
1 file changed, 5 insertions(+), 10 deletions(-)
--
2.40.1
The function sets adjusted groups of bits in hub_irq_map by using
for-loops. There's a bitmap_set() function dedicated to do this.
Because [0, CPU_CALL_B_IRQ] and [NI_BRDCAST_ERR_A, MSC_PANIC_INTR]
ranges belong to the same machine word, bitmap_set() would boil down
to an inline wrapper in both cases, avoiding generating a loop, with
the associate overhead. Effectively, it would be a compile-time:
*hub_irq_map = GENMASK() | GENMASK();
Signed-off-by: Yury Norov <[email protected]>
---
arch/mips/sgi-ip27/ip27-irq.c | 8 ++------
1 file changed, 2 insertions(+), 6 deletions(-)
diff --git a/arch/mips/sgi-ip27/ip27-irq.c b/arch/mips/sgi-ip27/ip27-irq.c
index 8f5299b269e7..d8acdf0439d2 100644
--- a/arch/mips/sgi-ip27/ip27-irq.c
+++ b/arch/mips/sgi-ip27/ip27-irq.c
@@ -277,7 +277,6 @@ void __init arch_init_irq(void)
{
struct irq_domain *domain;
struct fwnode_handle *fn;
- int i;
mips_cpu_irq_init();
@@ -286,11 +285,8 @@ void __init arch_init_irq(void)
* Mark these as reserved right away so they won't be used accidentally
* later.
*/
- for (i = 0; i <= CPU_CALL_B_IRQ; i++)
- set_bit(i, hub_irq_map);
-
- for (i = NI_BRDCAST_ERR_A; i <= MSC_PANIC_INTR; i++)
- set_bit(i, hub_irq_map);
+ bitmap_set(hub_irq_map, 0, CPU_CALL_B_IRQ + 1);
+ bitmap_set(hub_irq_map, NI_BRDCAST_ERR_A, MSC_PANIC_INTR - NI_BRDCAST_ERR_A + 1);
fn = irq_domain_alloc_named_fwnode("HUB");
WARN_ON(fn == NULL);
--
2.40.1
WARN_ON() propagates the result of conditional expression, and it can be
used to return early in the following expression in the arch_init_irq().
This is a no-op cleanup, except that compiler may optimize the error paths
better because WARN_ON() implies 'unlikely()'.
Signed-off-by: Yury Norov <[email protected]>
---
arch/mips/sgi-ip27/ip27-irq.c | 7 +++----
1 file changed, 3 insertions(+), 4 deletions(-)
diff --git a/arch/mips/sgi-ip27/ip27-irq.c b/arch/mips/sgi-ip27/ip27-irq.c
index d8acdf0439d2..00e63e9ef61d 100644
--- a/arch/mips/sgi-ip27/ip27-irq.c
+++ b/arch/mips/sgi-ip27/ip27-irq.c
@@ -289,13 +289,12 @@ void __init arch_init_irq(void)
bitmap_set(hub_irq_map, NI_BRDCAST_ERR_A, MSC_PANIC_INTR - NI_BRDCAST_ERR_A + 1);
fn = irq_domain_alloc_named_fwnode("HUB");
- WARN_ON(fn == NULL);
- if (!fn)
+ if (WARN_ON(fn == NULL))
return;
+
domain = irq_domain_create_linear(fn, IP27_HUB_IRQ_COUNT,
&hub_domain_ops, NULL);
- WARN_ON(domain == NULL);
- if (!domain)
+ if (WARN_ON(domain == NULL))
return;
irq_set_default_host(domain);
--
2.40.1
On Tue, Apr 16, 2024 at 10:37:10AM -0700, Yury Norov wrote:
> The function sets adjusted groups of bits in hub_irq_map by using
> for-loops. There's a bitmap_set() function dedicated to do this.
>
> Because [0, CPU_CALL_B_IRQ] and [NI_BRDCAST_ERR_A, MSC_PANIC_INTR]
> ranges belong to the same machine word, bitmap_set() would boil down
> to an inline wrapper in both cases, avoiding generating a loop, with
> the associate overhead. Effectively, it would be a compile-time:
>
> *hub_irq_map = GENMASK() | GENMASK();
>
> Signed-off-by: Yury Norov <[email protected]>
> ---
> arch/mips/sgi-ip27/ip27-irq.c | 8 ++------
> 1 file changed, 2 insertions(+), 6 deletions(-)
>
> diff --git a/arch/mips/sgi-ip27/ip27-irq.c b/arch/mips/sgi-ip27/ip27-irq.c
> index 8f5299b269e7..d8acdf0439d2 100644
> --- a/arch/mips/sgi-ip27/ip27-irq.c
> +++ b/arch/mips/sgi-ip27/ip27-irq.c
> @@ -277,7 +277,6 @@ void __init arch_init_irq(void)
> {
> struct irq_domain *domain;
> struct fwnode_handle *fn;
> - int i;
I've already applied your first version, so I need an incremental
patch, which just removes the unused variable.
Thomas.
--
Crap can work. Given enough thrust pigs will fly, but it's not necessarily a
good idea. [ RFC1925, 2.3 ]
On Wed, Apr 17, 2024 at 01:10:52PM +0200, Thomas Bogendoerfer wrote:
> On Tue, Apr 16, 2024 at 10:37:10AM -0700, Yury Norov wrote:
> > The function sets adjusted groups of bits in hub_irq_map by using
> > for-loops. There's a bitmap_set() function dedicated to do this.
> >
> > Because [0, CPU_CALL_B_IRQ] and [NI_BRDCAST_ERR_A, MSC_PANIC_INTR]
> > ranges belong to the same machine word, bitmap_set() would boil down
> > to an inline wrapper in both cases, avoiding generating a loop, with
> > the associate overhead. Effectively, it would be a compile-time:
> >
> > *hub_irq_map = GENMASK() | GENMASK();
> >
> > Signed-off-by: Yury Norov <[email protected]>
> > ---
> > arch/mips/sgi-ip27/ip27-irq.c | 8 ++------
> > 1 file changed, 2 insertions(+), 6 deletions(-)
> >
> > diff --git a/arch/mips/sgi-ip27/ip27-irq.c b/arch/mips/sgi-ip27/ip27-irq.c
> > index 8f5299b269e7..d8acdf0439d2 100644
> > --- a/arch/mips/sgi-ip27/ip27-irq.c
> > +++ b/arch/mips/sgi-ip27/ip27-irq.c
> > @@ -277,7 +277,6 @@ void __init arch_init_irq(void)
> > {
> > struct irq_domain *domain;
> > struct fwnode_handle *fn;
> > - int i;
>
> I've already applied your first version, so I need an incremental
> patch, which just removes the unused variable.
Sure, please find below.
From ce447fe69092c48bb59a6c4cb08ee5f9080f0ad6 Mon Sep 17 00:00:00 2001
From: Yury Norov <[email protected]>
Date: Mon, 22 Apr 2024 15:52:12 -0700
Subject: [PATCH] MIPS: SGI-IP27: fix -Wunused-variable in arch_init_irq()
Commit 40e20fbccfb722f21 (MIPS: SGI-IP27: micro-optimize arch_init_irq())
replaced a for-loop iteration with bitmap_set() calls, but didn't remove
an iteration variable.
Fixes: 40e20fbccfb722f21 (MIPS: SGI-IP27: micro-optimize arch_init_irq())
Reported-by: kernel test robot <[email protected]>
Closes: https://lore.kernel.org/oe-kbuild-all/[email protected]/
Signed-off-by: Yury Norov <[email protected]>
---
arch/mips/sgi-ip27/ip27-irq.c | 1 -
1 file changed, 1 deletion(-)
diff --git a/arch/mips/sgi-ip27/ip27-irq.c b/arch/mips/sgi-ip27/ip27-irq.c
index dcb14a234b1c..d8acdf0439d2 100644
--- a/arch/mips/sgi-ip27/ip27-irq.c
+++ b/arch/mips/sgi-ip27/ip27-irq.c
@@ -277,7 +277,6 @@ void __init arch_init_irq(void)
{
struct irq_domain *domain;
struct fwnode_handle *fn;
- int i;
mips_cpu_irq_init();
--
2.40.1
On Tue, Apr 16, 2024 at 10:37:11AM -0700, Yury Norov wrote:
> WARN_ON() propagates the result of conditional expression, and it can be
> used to return early in the following expression in the arch_init_irq().
>
> This is a no-op cleanup, except that compiler may optimize the error paths
> better because WARN_ON() implies 'unlikely()'.
>
> Signed-off-by: Yury Norov <[email protected]>
> ---
> arch/mips/sgi-ip27/ip27-irq.c | 7 +++----
> 1 file changed, 3 insertions(+), 4 deletions(-)
>
> diff --git a/arch/mips/sgi-ip27/ip27-irq.c b/arch/mips/sgi-ip27/ip27-irq.c
> index d8acdf0439d2..00e63e9ef61d 100644
> --- a/arch/mips/sgi-ip27/ip27-irq.c
> +++ b/arch/mips/sgi-ip27/ip27-irq.c
> @@ -289,13 +289,12 @@ void __init arch_init_irq(void)
> bitmap_set(hub_irq_map, NI_BRDCAST_ERR_A, MSC_PANIC_INTR - NI_BRDCAST_ERR_A + 1);
>
> fn = irq_domain_alloc_named_fwnode("HUB");
> - WARN_ON(fn == NULL);
> - if (!fn)
> + if (WARN_ON(fn == NULL))
> return;
> +
> domain = irq_domain_create_linear(fn, IP27_HUB_IRQ_COUNT,
> &hub_domain_ops, NULL);
> - WARN_ON(domain == NULL);
> - if (!domain)
> + if (WARN_ON(domain == NULL))
> return;
>
> irq_set_default_host(domain);
> --
> 2.40.1
applied to mips-next.
Thomas.
--
Crap can work. Given enough thrust pigs will fly, but it's not necessarily a
good idea. [ RFC1925, 2.3 ]
On Mon, Apr 22, 2024 at 04:04:23PM -0700, Yury Norov wrote:
> On Wed, Apr 17, 2024 at 01:10:52PM +0200, Thomas Bogendoerfer wrote:
> > On Tue, Apr 16, 2024 at 10:37:10AM -0700, Yury Norov wrote:
> > > The function sets adjusted groups of bits in hub_irq_map by using
> > > for-loops. There's a bitmap_set() function dedicated to do this.
> > >
> > > Because [0, CPU_CALL_B_IRQ] and [NI_BRDCAST_ERR_A, MSC_PANIC_INTR]
> > > ranges belong to the same machine word, bitmap_set() would boil down
> > > to an inline wrapper in both cases, avoiding generating a loop, with
> > > the associate overhead. Effectively, it would be a compile-time:
> > >
> > > *hub_irq_map = GENMASK() | GENMASK();
> > >
> > > Signed-off-by: Yury Norov <[email protected]>
> > > ---
> > > arch/mips/sgi-ip27/ip27-irq.c | 8 ++------
> > > 1 file changed, 2 insertions(+), 6 deletions(-)
> > >
> > > diff --git a/arch/mips/sgi-ip27/ip27-irq.c b/arch/mips/sgi-ip27/ip27-irq.c
> > > index 8f5299b269e7..d8acdf0439d2 100644
> > > --- a/arch/mips/sgi-ip27/ip27-irq.c
> > > +++ b/arch/mips/sgi-ip27/ip27-irq.c
> > > @@ -277,7 +277,6 @@ void __init arch_init_irq(void)
> > > {
> > > struct irq_domain *domain;
> > > struct fwnode_handle *fn;
> > > - int i;
> >
> > I've already applied your first version, so I need an incremental
> > patch, which just removes the unused variable.
>
> Sure, please find below.
>
> >From ce447fe69092c48bb59a6c4cb08ee5f9080f0ad6 Mon Sep 17 00:00:00 2001
> From: Yury Norov <[email protected]>
> Date: Mon, 22 Apr 2024 15:52:12 -0700
> Subject: [PATCH] MIPS: SGI-IP27: fix -Wunused-variable in arch_init_irq()
>
> Commit 40e20fbccfb722f21 (MIPS: SGI-IP27: micro-optimize arch_init_irq())
> replaced a for-loop iteration with bitmap_set() calls, but didn't remove
> an iteration variable.
>
> Fixes: 40e20fbccfb722f21 (MIPS: SGI-IP27: micro-optimize arch_init_irq())
> Reported-by: kernel test robot <[email protected]>
> Closes: https://lore.kernel.org/oe-kbuild-all/[email protected]/
> Signed-off-by: Yury Norov <[email protected]>
> ---
> arch/mips/sgi-ip27/ip27-irq.c | 1 -
> 1 file changed, 1 deletion(-)
>
> diff --git a/arch/mips/sgi-ip27/ip27-irq.c b/arch/mips/sgi-ip27/ip27-irq.c
> index dcb14a234b1c..d8acdf0439d2 100644
> --- a/arch/mips/sgi-ip27/ip27-irq.c
> +++ b/arch/mips/sgi-ip27/ip27-irq.c
> @@ -277,7 +277,6 @@ void __init arch_init_irq(void)
> {
> struct irq_domain *domain;
> struct fwnode_handle *fn;
> - int i;
>
> mips_cpu_irq_init();
>
> --
> 2.40.1
applied to mips-next.
Thomas.
--
Crap can work. Given enough thrust pigs will fly, but it's not necessarily a
good idea. [ RFC1925, 2.3 ]