Skip to contentSkip to conversation

cpumask: UP optimisation fixes follow-up

PATCH In review Sander Vanheule 2 participants 8 messages Aug 9, 2022 – Aug 10, 2022

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

cover cpumask: UP optimisation fixes follow-up

As an older version of the UP optimisation fixes was merged, not all review feedback has been implemented. These patches implement the feedback received on the merged version [1], and the respin [2], for changes related to include/linux/cpumask.h and lib/cpumask.c.

[1] https://lore.kernel.org/lkml/[email protected]/
[2] https://lore.kernel.org/lkml/[email protected]/

Sander Vanheule (3):
cpumask: align signatures of UP implementations lib/cpumask: add inline cpumask_next_wrap() for UP lib/cpumask: drop always-true preprocessor guard

 include/linux/cpumask.h | 26 +++++++++++++++++++++++---
 lib/Makefile            |  3 ++-
 lib/cpumask.c           |  2 --
 3 files changed, 25 insertions(+), 6 deletions(-)
Signature · 1 line
2.37.1
#1 lore raw

PATCH 1/3 cpumask: align signatures of UP implementations

Between the generic version, and their uniprocessor optimised implementations, the return types of cpumask_any_and_distribute() and cpumask_any_distribute() are not identical. Change the UP versions to 'unsigned int', to match the generic versions.

---
 include/linux/cpumask.h | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)
include/linux/cpumask.h +4 -3
index 0d435d0edbcb..d8c2a40f8beb 100644
@@ -202,12 +202,13 @@ static inline unsigned int cpumask_local_spread(unsigned int i, int node)
 	return 0;
 }
 
-static inline int cpumask_any_and_distribute(const struct cpumask *src1p,
-					     const struct cpumask *src2p) {
+static inline unsigned int cpumask_any_and_distribute(const struct cpumask *src1p,
+						      const struct cpumask *src2p)
+{
 	return cpumask_first_and(src1p, src2p);
 }
 
-static inline int cpumask_any_distribute(const struct cpumask *srcp)
+static inline unsigned int cpumask_any_distribute(const struct cpumask *srcp)
 {
 	return cpumask_first(srcp);
 }
Signature · 1 line
2.37.1
#2 lore raw

PATCH 2/3 lib/cpumask: add inline cpumask_next_wrap() for UP

In reply to Sander Vanheule

In the uniprocessor case, cpumask_next_wrap() can be simplified, as the number of valid argument combinations is limited:
- 'start' can only be 0
- 'n' can only be -1 or 0

The only valid CPU that can then be returned, if any, will be the first one set in the provided 'mask'.

For NR_CPUS == 1, include/linux/cpumask.h now provides an inline definition of cpumask_next_wrap(), which will conflict with the one provided by lib/cpumask.c. Make building of lib/cpumask.o again depend on CONFIG_SMP=y (i.e. NR_CPUS > 1) to avoid the re-definition.

---
 include/linux/cpumask.h | 19 +++++++++++++++++++
 lib/Makefile            |  3 ++-
 2 files changed, 21 insertions(+), 1 deletion(-)
include/linux/cpumask.h +19 -0
index d8c2a40f8beb..bd047864c7ac 100644
@@ -262,7 +262,26 @@ unsigned int cpumask_next_and(int n, const struct cpumask *src1p,
 		(cpu) = cpumask_next_zero((cpu), (mask)),	\
 		(cpu) < nr_cpu_ids;)
 
+#if NR_CPUS == 1
+static inline
+unsigned int cpumask_next_wrap(int n, const struct cpumask *mask, int start, bool wrap)
+{
+	cpumask_check(start);
+	if (n != -1)
+		cpumask_check(n);
+
+	/*
+	 * Return the first available CPU when wrapping, or when starting before cpu0,
+	 * since there is only one valid option.
+	 */
+	if (wrap && n >= 0)
+		return nr_cpumask_bits;
+
+	return cpumask_first(mask);
+}
+#else
 unsigned int __pure cpumask_next_wrap(int n, const struct cpumask *mask, int start, bool wrap);
+#endif
 
 /**
  * for_each_cpu_wrap - iterate over every cpu in a mask, starting at a specified location
lib/Makefile +2 -1
index c95212141928..5927d7fa0806 100644
@@ -34,9 +34,10 @@ lib-y := ctype.o string.o vsprintf.o cmdline.o \
 	 is_single_threaded.o plist.o decompress.o kobject_uevent.o \
 	 earlycpio.o seq_buf.o siphash.o dec_and_lock.o \
 	 nmi_backtrace.o win_minmax.o memcat_p.o \
-	 buildid.o cpumask.o
+	 buildid.o
 
 lib-$(CONFIG_PRINTK) += dump_stack.o
+lib-$(CONFIG_SMP) += cpumask.o
 
 lib-y	+= kobject.o klist.o
 obj-y	+= lockref.o
Signature · 1 line
2.37.1
#3 lore raw

PATCH 3/3 lib/cpumask: drop always-true preprocessor guard

In reply to Sander Vanheule

Since lib/cpumask.o is only built for CONFIG_SMP=y, NR_CPUS will always be greater than 1 at compile time. This makes checking for that condition unnecesarry, so it can be dropped.

---
 lib/cpumask.c | 2 --
 1 file changed, 2 deletions(-)
lib/cpumask.c +0 -2
index 8baeb37e23d3..f0ae119be8c4 100644
@@ -109,7 +109,6 @@ void __init free_bootmem_cpumask_var(cpumask_var_t mask)
 }
 #endif
 
-#if NR_CPUS > 1
 /**
  * cpumask_local_spread - select the i'th cpu with local numa cpu's first
  * @i: index number
@@ -197,4 +196,3 @@ unsigned int cpumask_any_distribute(const struct cpumask *srcp)
 	return next;
 }
 EXPORT_SYMBOL(cpumask_any_distribute);
-#endif /* NR_CPUS */
Signature · 1 line
2.37.1
maintainer Acked-by on the v1 series
#4 lore raw
In reply to Sander Vanheule
Quoted 5 lines · Sander Vanheule

As an older version of the UP optimisation fixes was merged, not all review feedback has been implemented. These patches implement the feedback received on the merged version [1], and the respin [2], for changes related to include/linux/cpumask.h and lib/cpumask.c.

Quoted 7 lines

[1] https://lore.kernel.org/lkml/[email protected]/
[2] https://lore.kernel.org/lkml/[email protected]/

Sander Vanheule (3):
cpumask: align signatures of UP implementations lib/cpumask: add inline cpumask_next_wrap() for UP lib/cpumask: drop always-true preprocessor guard

Applying at bitmap-for-next, after some testing.

Thanks,
Yury

Quoted 7 lines
 include/linux/cpumask.h | 26 +++++++++++++++++++++++---
 lib/Makefile            |  3 ++-
 lib/cpumask.c           |  2 --
 3 files changed, 25 insertions(+), 6 deletions(-)
Signature · 1 line
2.37.1
#5 lore raw
Quoted 9 lines · Yury Norov
Quoted 8 lines · Sander Vanheule

As an older version of the UP optimisation fixes was merged, not all review feedback has been implemented.  These patches implement the feedback received on the merged version [1], and the respin [2], for changes related to include/linux/cpumask.h and lib/cpumask.c.

[1] https://lore.kernel.org/lkml/[email protected]/
[2] https://lore.kernel.org/lkml/[email protected]/

Quoted 8 lines
Quoted 4 lines

Sander Vanheule (3):
  cpumask: align signatures of UP implementations lib/cpumask: add inline cpumask_next_wrap() for UP lib/cpumask: drop always-true preprocessor guard

Applying at bitmap-for-next, after some testing.

Thanks! Any chance to get this in for 6.0? I would rather avoid building cpumask.o only on 6.0, but otherwise I don't think anything is functionally wrong with what is currently merged.

Best,
Sander

maintainer
#6 lore raw
Quoted 15 lines · Sander Vanheule
Quoted 14 lines · Yury Norov
Quoted 12 lines · Sander Vanheule

As an older version of the UP optimisation fixes was merged, not all review feedback has been implemented.  These patches implement the feedback received on the merged version [1], and the respin [2], for changes related to include/linux/cpumask.h and lib/cpumask.c.

[1] https://lore.kernel.org/lkml/[email protected]/
[2] https://lore.kernel.org/lkml/[email protected]/

Sander Vanheule (3):
  cpumask: align signatures of UP implementations lib/cpumask: add inline cpumask_next_wrap() for UP lib/cpumask: drop always-true preprocessor guard

Quoted 6 lines
Quoted 3 lines

Applying at bitmap-for-next, after some testing.

Thanks! Any chance to get this in for 6.0? I would rather avoid building cpumask.o only on 6.0, but otherwise I don't think anything is functionally wrong with what is currently merged.

Functionally not, but something is still wrong, right? :)

I think -rc2 would be our best option for this, because this series is a fix to v4, and because it will let this spend some time in -next.

Are you OK with this?

Thanks,
Yury

#7 lore raw
Quoted 24 lines · Yury Norov
Quoted 15 lines · Sander Vanheule
Quoted 14 lines · Yury Norov
Quoted 12 lines · Sander Vanheule

As an older version of the UP optimisation fixes was merged, not all review feedback has been implemented.  These patches implement the feedback received on the merged version [1], and the respin [2], for changes related to include/linux/cpumask.h and lib/cpumask.c.

[1] https://lore.kernel.org/lkml/[email protected]/
[2] https://lore.kernel.org/lkml/[email protected]/

Sander Vanheule (3):
  cpumask: align signatures of UP implementations lib/cpumask: add inline cpumask_next_wrap() for UP lib/cpumask: drop always-true preprocessor guard

Quoted 7 lines
Quoted 3 lines

Applying at bitmap-for-next, after some testing.

Thanks! Any chance to get this in for 6.0? I would rather avoid building cpumask.o only on 6.0, but
otherwise I don't think anything is functionally wrong with what is currently merged.

Quoted 6 lines

Functionally not, but something is still wrong, right? :)

I think -rc2 would be our best option for this, because this series is a fix to v4, and because it will let this spend some time in -next.

Are you OK with this?

Sounds perfect!

Thanks,
Sander