Skip to contentSkip to conversation

x86: Fix boot on uniprocessor systems

PATCH In review Artem Fetishev 3 participants 4 messages Mar 24, 2014 – Mar 26, 2014 x86

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 x86: Fix boot on uniprocessor systems

On x86 uniprocessor systems topology_physical_package_id() returns -1 which causes rapl_cpu_prepare() to leave rapl_pmu variable uninitialized which leads to GPF in rapl_pmu_init(). See arch/x86/kernel/cpu/perf_event_intel_rapl.c.

It turns out that physical_package_id and core_id can actually be retreived for uniprocessor systems too. Enabling them also fixes rapl_pmu code.

---

arch/x86/include/asm/topology.h +2 -1
index d35f24e..1306d11 100644
@@ -119,9 +119,10 @@ static inline void setup_node_to_cpumask_map(void) { }
 
 extern const struct cpumask *cpu_coregroup_mask(int cpu);
 
-#ifdef ENABLE_TOPO_DEFINES
 #define topology_physical_package_id(cpu)	(cpu_data(cpu).phys_proc_id)
 #define topology_core_id(cpu)			(cpu_data(cpu).cpu_core_id)
+
+#ifdef ENABLE_TOPO_DEFINES
 #define topology_core_cpumask(cpu)		(per_cpu(cpu_core_map, cpu))
 #define topology_thread_cpumask(cpu)		(per_cpu(cpu_sibling_map, cpu))
 #endif
#1 lore raw
Quoted 15 lines · On Tue, 25 Mar 2014 01:59:10 +0300 Artem Fetishev <[email protected]>…

On x86 uniprocessor systems topology_physical_package_id() returns -1 which causes rapl_cpu_prepare() to leave rapl_pmu variable uninitialized which leads to GPF in rapl_pmu_init(). See arch/x86/kernel/cpu/perf_event_intel_rapl.c.

It turns out that physical_package_id and core_id can actually be retreived for uniprocessor systems too. Enabling them also fixes rapl_pmu code.

---

arch/x86/include/asm/topology.h +0 -0
index d35f24e..1306d11 100644
@@ -119,9 +119,10 @@ static inline void setup_node_to_cpumask_map(void) { }
 

arch/x86/include/asm/topology.h static inline void setup_node_to_cpumask_map(void) { } · line 128

Quoted 10 lines
arch/x86/include/asm/topology.h +2 -1
 extern const struct cpumask *cpu_coregroup_mask(int cpu);
 
-#ifdef ENABLE_TOPO_DEFINES
 #define topology_physical_package_id(cpu)	(cpu_data(cpu).phys_proc_id)
 #define topology_core_id(cpu)			(cpu_data(cpu).cpu_core_id)
+
+#ifdef ENABLE_TOPO_DEFINES
 #define topology_core_cpumask(cpu)		(per_cpu(cpu_core_map, cpu))
 #define topology_thread_cpumask(cpu)		(per_cpu(cpu_sibling_map, cpu))
 #endif

The patch applies to 3.13 and perhaps earlier kernels. Is it needed in those kernel versions?

#2 lore raw
Quoted 19 lines · Andrew Morton
Quoted 17 lines · On Tue, 25 Mar 2014 01:59:10 +0300 Artem Fetishev <[email protected]>…

On x86 uniprocessor systems topology_physical_package_id() returns -1 which causes rapl_cpu_prepare() to leave rapl_pmu variable uninitialized which leads to GPF in rapl_pmu_init(). See arch/x86/kernel/cpu/perf_event_intel_rapl.c.

It turns out that physical_package_id and core_id can actually be retreived for uniprocessor systems too. Enabling them also fixes rapl_pmu code.

---

arch/x86/include/asm/topology.h +0 -0
index d35f24e..1306d11 100644
@@ -119,9 +119,10 @@ static inline void setup_node_to_cpumask_map(void) { }
 
 extern const struct cpumask *cpu_coregroup_mask(int cpu);
 
Quoted 11 lines

arch/x86/include/asm/topology.h static inline void setup_node_to_cpumask_map(void) { } · line 128

Quoted 8 lines
arch/x86/include/asm/topology.h +2 -1
-#ifdef ENABLE_TOPO_DEFINES
 #define topology_physical_package_id(cpu)    (cpu_data(cpu).phys_proc_id)
 #define topology_core_id(cpu)                        (cpu_data(cpu).cpu_core_id)
+
+#ifdef ENABLE_TOPO_DEFINES
 #define topology_core_cpumask(cpu)           (per_cpu(cpu_core_map, cpu))
 #define topology_thread_cpumask(cpu)         (per_cpu(cpu_sibling_map, cpu))
 #endif

The patch applies to 3.13 and perhaps earlier kernels. Is it needed in those kernel versions?

Before 3.13 there was no RAPL support.
But it seems the patch is still useful regardless

#3 lore raw
Quoted 33 lines · On Wed, 26 Mar 2014 23:01:07 +0100 Stephane Eranian <[email protected]> wrote:
Quoted 19 lines · Andrew Morton
Quoted 17 lines · On Tue, 25 Mar 2014 01:59:10 +0300 Artem Fetishev <[email protected]>…

On x86 uniprocessor systems topology_physical_package_id() returns -1 which causes rapl_cpu_prepare() to leave rapl_pmu variable uninitialized which leads to GPF in rapl_pmu_init(). See arch/x86/kernel/cpu/perf_event_intel_rapl.c.

It turns out that physical_package_id and core_id can actually be retreived for uniprocessor systems too. Enabling them also fixes rapl_pmu code.

---

arch/x86/include/asm/topology.h +0 -0
index d35f24e..1306d11 100644
@@ -119,9 +119,10 @@ static inline void setup_node_to_cpumask_map(void) { }
 
 extern const struct cpumask *cpu_coregroup_mask(int cpu);
 
Quoted 11 lines

arch/x86/include/asm/topology.h static inline void setup_node_to_cpumask_map(void) { } · line 128

Quoted 8 lines
arch/x86/include/asm/topology.h +2 -1
-#ifdef ENABLE_TOPO_DEFINES
 #define topology_physical_package_id(cpu)    (cpu_data(cpu).phys_proc_id)
 #define topology_core_id(cpu)                        (cpu_data(cpu).cpu_core_id)
+
+#ifdef ENABLE_TOPO_DEFINES
 #define topology_core_cpumask(cpu)           (per_cpu(cpu_core_map, cpu))
 #define topology_thread_cpumask(cpu)         (per_cpu(cpu_sibling_map, cpu))
 #endif

The patch applies to 3.13 and perhaps earlier kernels. Is it needed in those kernel versions?

Quoted 2 lines

Before 3.13 there was no RAPL support.
But it seems the patch is still useful regardless

Is that an ack? If so I'll squirt it Linuswards right now for 3.14.