Skip to contentSkip to conversation

ppc64: Convert cpu_sibling_map to a per_cpu data array ppc64 v2

PATCH In review [email protected] 2 participants 3 messages Sep 17, 2007 – Sep 19, 2007

Subsystems (3)
Memory management PowerPC Cell broadband engine architecture
lore
Share & tools
lore

[email protected]

author
#0 lore raw

cover ppc64: Convert cpu_sibling_map to a per_cpu data array ppc64 v2

Quoted 10 lines · Stephen Rothwell
Quoted 9 lines · On Mon, 17 Sep 2007 16:28:31 +1000 Stephen Rothwell <[email protected]>…

the topology (on my POWERPC5+ box) is not correct:

cpu0/topology/thread_siblings:0000000f
cpu1/topology/thread_siblings:0000000f
cpu2/topology/thread_siblings:0000000f
cpu3/topology/thread_siblings:0000000f

it used to be:

Quoted 7 lines
Quoted 4 lines

cpu0/topology/thread_siblings:00000003
cpu1/topology/thread_siblings:00000003
cpu2/topology/thread_siblings:0000000c
cpu3/topology/thread_siblings:0000000c

This would be because we are setting up the cpu_sibling map before we call setup_per_cpu_areas().

The following patch hopefully should fix this problem. I'm not able to build or test it but the few references to cpu_sibling_map seem to all occur well after setup_per_cpu_areas is called.

Thanks Stephen for checking this out!

Signature · 5 lines
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to [email protected].  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
author
#1 lore raw

PATCH 1/1 ppc64: Convert cpu_sibling_map to a per_cpu data array ppc64 v2

This patch applies after:

convert-cpu_sibling_map-to-a-per_cpu-data-array-ppc64.patch

and should fix the "reference cpu_sibling_map before setup_per_cpu_areas() has been called" problem. In addtion, the cpu_sibiling_map macro has been removed [this was missed in my original submission.]

Original Notes:

Convert cpu_sibling_map to a per_cpu cpumask_t array for the ppc64 architecture. This fixes build errors in block/blktrace.c and kernel/sched.c when CONFIG_SCHED_SMT is defined.

Note: these changes have not been built nor tested.

Signed-off-by Mike Travis <[email protected]>
---
 arch/powerpc/kernel/setup-common.c        |   20 ++++++++++++++++----
 arch/powerpc/kernel/setup_64.c            |    3 +++
 arch/powerpc/platforms/cell/cbe_cpufreq.c |    2 +-
 include/asm-powerpc/smp.h                 |    2 +-
 include/asm-powerpc/topology.h            |    2 +-
 5 files changed, 22 insertions(+), 7 deletions(-)
include/asm-powerpc/smp.h +1 -1
@@ -60,7 +60,6 @@
 #endif
 
 DECLARE_PER_CPU(cpumask_t, cpu_sibling_map);
-#define cpu_sibling_map(cpu) per_cpu(cpu_sibling_map, cpu)
 
 /* Since OpenPIC has only 4 IPIs, we use slightly different message numbers.
  *

@@ -79,6 +78,7 @@
 void smp_init_cell(void);
 void smp_init_celleb(void);
 void smp_setup_cpu_maps(void);
+void smp_setup_cpu_sibling_map(void);
 
 extern int __cpu_disable(void);
 extern void __cpu_die(unsigned int cpu);
arch/powerpc/kernel/setup_64.c +3 -0
@@ -596,6 +596,9 @@
 		paca[i].data_offset = ptr - __per_cpu_start;
 		memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
 	}
+
+	/* Now that per_cpu is setup, initialize cpu_sibling_map */
+	smp_setup_cpu_sibling_map();
 }
 #endif
arch/powerpc/kernel/setup-common.c +16 -4
@@ -411,16 +411,28 @@
 		of_node_put(dn);
 	}
 
+	vdso_data->processorCount = num_present_cpus();
+#endif /* CONFIG_PPC64 */
+}
+
+/*
+ * Being that cpu_sibling_map is now a per_cpu array, then it cannot
+ * be initialized until the per_cpu areas have been created.  This
+ * function is now called from setup_per_cpu_areas().
+ */
+void __init smp_setup_cpu_sibling_map(void)
+{
+#if defined(CONFIG_PPC64)
+	int cpu;
+
 	/*
 	 * Do the sibling map; assume only two threads per processor.
 	 */
 	for_each_possible_cpu(cpu) {
-		cpu_set(cpu, cpu_sibling_map(cpu));
+		cpu_set(cpu, per_cpu(cpu_sibling_map, cpu));
 		if (cpu_has_feature(CPU_FTR_SMT))
-			cpu_set(cpu ^ 0x1, cpu_sibling_map(cpu));
+			cpu_set(cpu ^ 0x1, per_cpu(cpu_sibling_map, cpu));
 	}
-
-	vdso_data->processorCount = num_present_cpus();
 #endif /* CONFIG_PPC64 */
 }
 #endif /* CONFIG_SMP */
arch/powerpc/platforms/cell/cbe_cpufreq.c +1 -1
@@ -119,7 +119,7 @@
 	policy->cur = cbe_freqs[cur_pmode].frequency;
 
 #ifdef CONFIG_SMP
-	policy->cpus = cpu_sibling_map(policy->cpu);
+	policy->cpus = per_cpu(cpu_sibling_map, policy->cpu);
 #endif
 
 	cpufreq_frequency_table_get_attr(cbe_freqs, policy->cpu);
include/asm-powerpc/topology.h +1 -1
@@ -108,7 +108,7 @@
 #ifdef CONFIG_PPC64
 #include <asm/smp.h>
 
-#define topology_thread_siblings(cpu)	(cpu_sibling_map(cpu))
+#define topology_thread_siblings(cpu)	(per_cpu(cpu_sibling_map, cpu))
 #endif
 #endif
Signature · 5 lines
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to [email protected].  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
#2 lore raw

Hi Mike,

Quoted 10 lines · On Mon, 17 Sep 2007 11:35:08 -0700 [email protected] wrote:

v2:

This patch applies after:

convert-cpu_sibling_map-to-a-per_cpu-data-array-ppc64.patch

and should fix the "reference cpu_sibling_map before setup_per_cpu_areas() has been called" problem. In addtion, the cpu_sibiling_map macro has been removed [this was missed in my original submission.]

This one make it work, thanks.

Signature · 3 lines
Cheers,
Stephen Rothwell                    [email protected]
http://www.canb.auug.org.au/~sfr/