Skip to contentSkip to conversation

perf: Fix perf_event_{init,exit}_cpu stubs

PATCH In review Florian Fainelli 2 participants 4 messages Nov 3, 2022 – Nov 10, 2022 perf events

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 perf: Fix perf_event_{init,exit}_cpu stubs

The original commit that introduced those stubs was already at fault, but in the absence of a caller of perf_event_{init,exit}_cpu outside of code that is compiled regardless of CONFIG_PERF_EVENTS, the build failure cannot be observed. This was observed with the Android kernel to produce a build failure similar to this:

    In file included from ./include/uapi/linux/posix_types.h:5,
                     from ./include/uapi/linux/types.h:14,
                     from ./include/linux/types.h:6,
                     from ./include/linux/limits.h:6,
                     from ./include/linux/kernel.h:7,
                     from ./include/linux/sched/mm.h:5,
                     from kernel/cpu.c:6:
    kernel/cpu.c: In function 'random_and_perf_prepare_fusion':
    ./include/linux/stddef.h:8:14: error: called object is not a function or function pointer
     #define NULL ((void *)0)
                  ^
    ./include/linux/perf_event.h:1607:29: note: in expansion of macro 'NULL'
     #define perf_event_init_cpu NULL
                                 ^~~~
    kernel/cpu.c:1686:2: note: in expansion of macro 'perf_event_init_cpu'
      perf_event_init_cpu(cpu);
      ^~~~~~~~~~~~~~~~~~~
Fixes 00e16c3d68fc ("perf/core: Convert to hotplug state machine")
---
 include/linux/perf_event.h | 11 +++++++++--
 1 file changed, 9 insertions(+), 2 deletions(-)
include/linux/perf_event.h +9 -2
index 0031f7b4d9ab..592040e2398c 100644
@@ -1678,8 +1678,15 @@ static struct device_attribute format_attr_##_name = __ATTR_RO(_name)
 int perf_event_init_cpu(unsigned int cpu);
 int perf_event_exit_cpu(unsigned int cpu);
 #else
-#define perf_event_init_cpu	NULL
-#define perf_event_exit_cpu	NULL
+static inline int perf_event_init_cpu(unsigned int cpu)
+{
+	return 0;
+}
+
+static inline int perf_event_exit_cpu(unsigned int cpu)
+{
+	return 0;
+}
 #endif
 
 extern void __weak arch_perf_update_userpage(struct perf_event *event,
Signature · 1 line
2.25.1
maintainer
#1 lore raw
Quoted 15 lines · Florian Fainelli

The original commit that introduced those stubs was already at fault, but in the absence of a caller of perf_event_{init,exit}_cpu outside of code that is compiled regardless of CONFIG_PERF_EVENTS, the build failure cannot be observed. This was observed with the Android kernel to produce a build failure similar to this:

    In file included from ./include/uapi/linux/posix_types.h:5,
                     from ./include/uapi/linux/types.h:14,
                     from ./include/linux/types.h:6,
                     from ./include/linux/limits.h:6,
                     from ./include/linux/kernel.h:7,
                     from ./include/linux/sched/mm.h:5,
                     from kernel/cpu.c:6:
    kernel/cpu.c: In function 'random_and_perf_prepare_fusion':
    ./include/linux/stddef.h:8:14: error: called object is not a function or function pointer
Quoted 8 lines
     #define NULL ((void *)0)
                  ^
    ./include/linux/perf_event.h:1607:29: note: in expansion of macro 'NULL'
     #define perf_event_init_cpu NULL
                                 ^~~~
    kernel/cpu.c:1686:2: note: in expansion of macro 'perf_event_init_cpu'
      perf_event_init_cpu(cpu);
      ^~~~~~~~~~~~~~~~~~~

What is the actual problem reported here? Did you see all the other NULL assignments in cpuhp_hp_states ?

Quoted 33 lines
Fixes 00e16c3d68fc ("perf/core: Convert to hotplug state machine")
---
 include/linux/perf_event.h | 11 +++++++++--
 1 file changed, 9 insertions(+), 2 deletions(-)
include/linux/perf_event.h +9 -2
index 0031f7b4d9ab..592040e2398c 100644
@@ -1678,8 +1678,15 @@ static struct device_attribute format_attr_##_name = __ATTR_RO(_name)
 int perf_event_init_cpu(unsigned int cpu);
 int perf_event_exit_cpu(unsigned int cpu);
 #else
-#define perf_event_init_cpu	NULL
-#define perf_event_exit_cpu	NULL
+static inline int perf_event_init_cpu(unsigned int cpu)
+{
+	return 0;
+}
+
+static inline int perf_event_exit_cpu(unsigned int cpu)
+{
+	return 0;
+}
 #endif
 
 extern void __weak arch_perf_update_userpage(struct perf_event *event,
Signature · 1 line
2.25.1
#2 lore raw
Quoted 25 lines · Peter Zijlstra
Quoted 15 lines · Florian Fainelli

The original commit that introduced those stubs was already at fault, but in the absence of a caller of perf_event_{init,exit}_cpu outside of code that is compiled regardless of CONFIG_PERF_EVENTS, the build failure cannot be observed. This was observed with the Android kernel to produce a build failure similar to this:

     In file included from ./include/uapi/linux/posix_types.h:5,
                      from ./include/uapi/linux/types.h:14,
                      from ./include/linux/types.h:6,
                      from ./include/linux/limits.h:6,
                      from ./include/linux/kernel.h:7,
                      from ./include/linux/sched/mm.h:5,
                      from kernel/cpu.c:6:
     kernel/cpu.c: In function 'random_and_perf_prepare_fusion':
     ./include/linux/stddef.h:8:14: error: called object is not a function or function pointer
Quoted 8 lines
      #define NULL ((void *)0)
                   ^
     ./include/linux/perf_event.h:1607:29: note: in expansion of macro 'NULL'
      #define perf_event_init_cpu NULL
                                  ^~~~
     kernel/cpu.c:1686:2: note: in expansion of macro 'perf_event_init_cpu'
       perf_event_init_cpu(cpu);
       ^~~~~~~~~~~~~~~~~~~
Quoted 2 lines

What is the actual problem reported here? Did you see all the other NULL assignments in cpuhp_hp_states ?

Anand reported to me that the following downstream commit in the Android common kernel repository:

https://android.googlesource.com/kernel/common/+/ca927bd22ad8bd26fd8dcebf3c7f2a093385d8ea

was resulting in the build failure above.

While this is an Android inflicted change and I still intend to get a fix there, upon closer look, the build failure could be experienced upstream as well with any code calling perf_event_{init,exit}_cpu() as a regular function call as opposed to being used as a function pointer.

This felt worthy of fixing in the upstream kernel, hence this patch.

Signature · 1 line
Florian
#3 lore raw
Quoted 33 lines · Florian Fainelli
Quoted 25 lines · Peter Zijlstra
Quoted 15 lines · Florian Fainelli

The original commit that introduced those stubs was already at fault, but in the absence of a caller of perf_event_{init,exit}_cpu outside of code that is compiled regardless of CONFIG_PERF_EVENTS, the build failure cannot be observed. This was observed with the Android kernel to produce a build failure similar to this:

     In file included from ./include/uapi/linux/posix_types.h:5,
                      from ./include/uapi/linux/types.h:14,
                      from ./include/linux/types.h:6,
                      from ./include/linux/limits.h:6,
                      from ./include/linux/kernel.h:7,
                      from ./include/linux/sched/mm.h:5,
                      from kernel/cpu.c:6:
     kernel/cpu.c: In function 'random_and_perf_prepare_fusion':
     ./include/linux/stddef.h:8:14: error: called object is not a function or function pointer

Quoted 8 lines

      #define NULL ((void *)0)
                   ^
     ./include/linux/perf_event.h:1607:29: note: in expansion of macro 'NULL' #define perf_event_init_cpu NULL
                                  ^~~~
     kernel/cpu.c:1686:2: note: in expansion of macro 'perf_event_init_cpu' perf_event_init_cpu(cpu);
       ^~~~~~~~~~~~~~~~~~~

Quoted 2 lines

What is the actual problem reported here? Did you see all the other NULL assignments in cpuhp_hp_states ?

Anand reported to me that the following downstream commit in the Android common kernel repository:

Quoted 7 lines

https://android.googlesource.com/kernel/common/+/ca927bd22ad8bd26fd8dcebf3c7f2a093385d8ea

was resulting in the build failure above.

While this is an Android inflicted change and I still intend to get a fix there, upon closer look, the build failure could be experienced upstream as well with any code calling perf_event_{init,exit}_cpu() as a regular function call as opposed to being used as a function pointer.

This felt worthy of fixing in the upstream kernel, hence this patch.

No response meaning it's a don't care because upstream is not (yet?) impacted?

Signature · 1 line
Florian