tracing/syscalls: Rename variable 'nr' to '__syscall_nr'

There is a problem about duplicated variable name i.e.

    # cat /sys/kernel/debug/tracing/events/syscalls/sys_enter_io_getevents/format
    name: sys_enter_io_getevents
    ID: 739
    format:

            field:unsigned short common_type; offset:0;  size:2; signed:0;
            field:unsigned char common_flags; offset:2;  size:1; signed:0;
            field:unsigned char common_preempt_count; offset:3; size:1; signed:0;
            field:int common_pid;offset:4;size:4;signed:1;
            field:int nr;                     offset:8;  size:4; signed:1;
            field:aio_context_t ctx_id;       offset:16; size:8; signed:0;
            field:long min_nr;                offset:24; size:8; signed:0;
            field:long nr;                    offset:32; size:8; signed:0;
            field:struct io_event * events;   offset:40; size:8; signed:0;
            field:struct timespec * timeout;  offset:48; size:8; signed:0;

            print fmt: "ctx_id: 0x%08lx, min_nr: 0x%08lx, nr: 0x%08lx,
                        events: 0x%08lx, timeout: 0x%08lx", ((unsigned long)(REC->ctx_id)),
                        ((unsigned long)(REC->min_nr)), ((unsigned long)(REC->nr)),
                        ((unsigned long)(REC->events)), ((unsigned long)(REC->timeout))

As above 'int nr;' and 'long nr;' variables have duplicated name so problems are occurred in perf-script i.e.

    # perf record -e syscalls:*
    # perf script -g python
    # perf script -s perf-script.py
      File "perf-script.py", line 8694
        def syscalls__sys_enter_io_getevents(event_name, context, common_cpu,
    SyntaxError: duplicate argument 'nr' in function definition
    Error running python script perf-script.py

As above, problems about duplicated argument occurred. Not only sys_enter_io_getevent() but also sys_enter_io_submit() have relevevance to this problem.

So rename a variable 'nr' to '__syscall_nr' for system call number in print_syscall_enter() and etc.

Cc Thomas Gleixner <[email protected]>

Cc Lai Jiangshan <[email protected]>

Signed-off-by Taeung Song <[email protected]>

---
 kernel/trace/trace.h          |  4 ++--
 kernel/trace/trace_syscalls.c | 16 ++++++++--------
 2 files changed, 10 insertions(+), 10 deletions(-)

kernel/trace/trace.h +2 -2

index 8414fa4..98b3c66 100644
@@ -88,13 +88,13 @@ enum trace_type {
  */
 struct syscall_trace_enter {
 	struct trace_entry	ent;
-	int			nr;
+	int			__syscall_nr;
 	unsigned long		args[];
 };
 
 struct syscall_trace_exit {
 	struct trace_entry	ent;
-	int			nr;
+	int			__syscall_nr;
 	long			ret;
 };
 

kernel/trace/trace_syscalls.c +8 -8

index 0655afb..90bb468 100644
@@ -118,7 +118,7 @@ print_syscall_enter(struct trace_iterator *iter, int flags,
 	int i, syscall;
 
 	trace = (typeof(trace))ent;
-	syscall = trace->nr;
+	syscall = trace->__syscall_nr;
 	entry = syscall_nr_to_meta(syscall);
 
 	if (!entry)

@@ -164,7 +164,7 @@ print_syscall_exit(struct trace_iterator *iter, int flags,
 	struct syscall_metadata *entry;
 
 	trace = (typeof(trace))ent;
-	syscall = trace->nr;
+	syscall = trace->__syscall_nr;
 	entry = syscall_nr_to_meta(syscall);
 
 	if (!entry) {

@@ -261,7 +261,7 @@ static int __init syscall_enter_define_fields(struct trace_event_call *call)
 	int i;
 	int offset = offsetof(typeof(trace), args);
 
-	ret = trace_define_field(call, SYSCALL_FIELD(int, nr), FILTER_OTHER);
+	ret = trace_define_field(call, SYSCALL_FIELD(int, __syscall_nr), FILTER_OTHER);
 	if (ret)
 		return ret;
 

@@ -281,7 +281,7 @@ static int __init syscall_exit_define_fields(struct trace_event_call *call)
 	struct syscall_trace_exit trace;
 	int ret;
 
-	ret = trace_define_field(call, SYSCALL_FIELD(int, nr), FILTER_OTHER);
+	ret = trace_define_field(call, SYSCALL_FIELD(int, __syscall_nr), FILTER_OTHER);
 	if (ret)
 		return ret;
 

@@ -332,7 +332,7 @@ static void ftrace_syscall_enter(void *data, struct pt_regs *regs, long id)
 		return;
 
 	entry = ring_buffer_event_data(event);
-	entry->nr = syscall_nr;
+	entry->__syscall_nr = syscall_nr;
 	syscall_get_arguments(current, regs, 0, sys_data->nb_args, entry->args);
 
 	event_trigger_unlock_commit(trace_file, buffer, event, entry,

@@ -378,7 +378,7 @@ static void ftrace_syscall_exit(void *data, struct pt_regs *regs, long ret)
 		return;
 
 	entry = ring_buffer_event_data(event);
-	entry->nr = syscall_nr;
+	entry->__syscall_nr = syscall_nr;
 	entry->ret = syscall_get_return_value(current, regs);
 
 	event_trigger_unlock_commit(trace_file, buffer, event, entry,

@@ -579,7 +579,7 @@ static void perf_syscall_enter(void *ignore, struct pt_regs *regs, long id)
 	if (!rec)
 		return;
 
-	rec->nr = syscall_nr;
+	rec->__syscall_nr = syscall_nr;
 	syscall_get_arguments(current, regs, 0, sys_data->nb_args,
 			       (unsigned long *)&rec->args);
 	perf_trace_buf_submit(rec, size, rctx, 0, 1, regs, head, NULL);

@@ -652,7 +652,7 @@ static void perf_syscall_exit(void *ignore, struct pt_regs *regs, long ret)
 	if (!rec)
 		return;
 
-	rec->nr = syscall_nr;
+	rec->__syscall_nr = syscall_nr;
 	rec->ret = syscall_get_return_value(current, regs);
 	perf_trace_buf_submit(rec, size, rctx, 0, 1, regs, head, NULL);
 }

Signature · 1 line

2.5.0

Em Fri, Feb 26, 2016 at 10:14:06PM +0900, Taeung Song escreveu:

Quoted 6 lines

There is a problem about duplicated variable name i.e.

    # cat /sys/kernel/debug/tracing/events/syscalls/sys_enter_io_getevents/format
    name: sys_enter_io_getevents
    ID: 739
    format:  

Steven, what do you think?

Should we break this ABI while disambiguating the 'nr' field, using '__syscall_nr' in an attempt to use a name that is unlikely to be used by a real syscall argument name?

If we stand by published ABIs, we should keep it written in stone and state that the first 'nr' means '__syscall_nr' while keeping it as-is, the change for 'perf trace' in that case is to do nothing, it work as-is, we have just to fix the python binding to do that rename.

Perhaps we can live with that, to avoid having three different cases:
!nr, nr and __syscall_nr.

Ingo, Peter, have you guys followed this case?

Summary: Some tracepoint have multiple fields with the same name, 'nr',
the first one is a unique syscall ID, the other is a syscall argument:

[root@jouet ~]# cat /sys/kernel/debug/tracing/events/syscalls/sys_enter_io_getevents/format
name: sys_enter_io_getevents
ID: 747
format:

	field:unsigned short common_type;	offset:0;	size:2;	signed:0;
	field:unsigned char common_flags;	offset:2;	size:1;	signed:0;
	field:unsigned char common_preempt_count;	offset:3;	size:1;	signed:0;
	field:int common_pid;	offset:4;	size:4;	signed:1;

	field:int nr;	offset:8;	size:4;	signed:1;
	field:aio_context_t ctx_id;	offset:16;	size:8;	signed:0;
	field:long min_nr;	offset:24;	size:8;	signed:0;
	field:long nr;	offset:32;	size:8;	signed:0;
	field:struct io_event * events;	offset:40;	size:8;	signed:0;
	field:struct timespec * timeout;	offset:48;	size:8;	signed:0;

print fmt: "ctx_id: 0x%08lx, min_nr: 0x%08lx, nr: 0x%08lx, events: 0x%08lx, timeout: 0x%08lx", ((unsigned long)(REC->ctx_id)), ((unsigned long)(REC->min_nr)), ((unsigned long)(REC->nr)), ((unsigned long)(REC->events)), ((unsigned long)(REC->timeout)) [root@jouet ~]#

Quoted 4 lines

If we stand by published ABIs, we should keep it written in stone and state that the first 'nr' means '__syscall_nr' while keeping it as-is, the change for 'perf trace' in that case is to do nothing, it work as-is, we have just to fix the python binding to do that rename.

Quoted 2 lines

Perhaps we can live with that, to avoid having three different cases:
!nr, nr and __syscall_nr.

We could, do this as well. Want me to add something to event-parse?

Quoted 10 lines · On Fri, 26 Feb 2016 10:57:13 -0300 Arnaldo Carvalho de Melo <[email protected]>…

Em Fri, Feb 26, 2016 at 10:14:06PM +0900, Taeung Song escreveu:

Quoted 6 lines

There is a problem about duplicated variable name i.e.

     # cat /sys/kernel/debug/tracing/events/syscalls/sys_enter_io_getevents/format
     name: sys_enter_io_getevents
     ID: 739
     format:

Steven, what do you think?

Quoted 8 lines

Should we break this ABI while disambiguating the 'nr' field, using '__syscall_nr' in an attempt to use a name that is unlikely to be used by a real syscall argument name?

If we stand by published ABIs, we should keep it written in stone and state that the first 'nr' means '__syscall_nr' while keeping it as-is, the change for 'perf trace' in that case is to do nothing, it work as-is, we have just to fix the python binding to do that rename.

ABIs only matter if they break something, and people complain. Linus has been somewhat accepting of us fixing those tools that break and we push out the fixes. If an ABI breaks in the forest and nobody is around to complain about it, did it really break?

I would say, lets make the change and fix perf. If people complain, we send them the fixes for their tools. If they need the distros to have the fixes, then let the change be reverted, and we wait till the distros have the update (this may take a few years), then re-submit.

This worked for me to get rid of padding that was in every trace event. The change was reverted, I fixed the tools that broke, waited till all the major distros had the updates. And resubmitted the change. Linus took it.

Quoted 3 lines

Perhaps we can live with that, to avoid having three different cases:
!nr, nr and __syscall_nr.

We could, do this as well. Want me to add something to event-parse?

Quoted 23 lines

Ingo, Peter, have you guys followed this case?

Summary: Some tracepoint have multiple fields with the same name, 'nr',
the first one is a unique syscall ID, the other is a syscall argument:

[root@jouet ~]# cat /sys/kernel/debug/tracing/events/syscalls/sys_enter_io_getevents/format
name: sys_enter_io_getevents
ID: 747
format:

	field:unsigned short common_type;	offset:0;	size:2;	signed:0;
	field:unsigned char common_flags;	offset:2;	size:1;	signed:0;
	field:unsigned char common_preempt_count;	offset:3;	size:1;	signed:0;
	field:int common_pid;	offset:4;	size:4;	signed:1;

	field:int nr;	offset:8;	size:4;	signed:1;
	field:aio_context_t ctx_id;	offset:16;	size:8;	signed:0;
	field:long min_nr;	offset:24;	size:8;	signed:0;
	field:long nr;	offset:32;	size:8;	signed:0;
	field:struct io_event * events;	offset:40;	size:8;	signed:0;
	field:struct timespec * timeout;	offset:48;	size:8;	signed:0;

Quoted 3 lines

print fmt: "ctx_id: 0x%08lx, min_nr: 0x%08lx, nr: 0x%08lx, events: 0x%08lx, timeout: 0x%08lx", ((unsigned long)(REC->ctx_id)), ((unsigned long)(REC->min_nr)), ((unsigned long)(REC->nr)), ((unsigned long)(REC->events)), ((unsigned long)(REC->timeout)) [root@jouet ~]#

BTW, here's a less intrusive change, because honestly, I hate the kernel structure having underscores in the name.

This could be signed off by Taeung Song and myself.

-- Steve

kernel/trace/trace_syscalls.c +6 -6

index 0655afbea83f..d1663083d903 100644
@@ -186,11 +186,11 @@ print_syscall_exit(struct trace_iterator *iter, int flags,
 
  extern char *__bad_type_size(void);
 
-#define SYSCALL_FIELD(type, name)					\
-	sizeof(type) != sizeof(trace.name) ?				\
+#define SYSCALL_FIELD(type, field, name)				\
+	sizeof(type) != sizeof(trace.field) ?				\
  		__bad_type_size() :					\
-		#type, #name, offsetof(typeof(trace), name),		\
-		sizeof(trace.name), is_signed_type(type)
+		#type, #name, offsetof(typeof(trace), field),		\
+		sizeof(trace.field), is_signed_type(type)
 
  static int __init
  __set_enter_print_fmt(struct syscall_metadata *entry, char *buf, int len)

@@ -261,7 +261,8 @@ static int __init syscall_enter_define_fields(struct trace_event_call *call)
  	int i;
  	int offset = offsetof(typeof(trace), args);
 
-	ret = trace_define_field(call, SYSCALL_FIELD(int, nr), FILTER_OTHER);
+	ret = trace_define_field(call, SYSCALL_FIELD(int, nr, __syscall_nr),
+				 FILTER_OTHER);
  	if (ret)
  		return ret;
 

@@ -281,11 +282,12 @@ static int __init syscall_exit_define_fields(struct trace_event_call *call)
  	struct syscall_trace_exit trace;
  	int ret;
 
-	ret = trace_define_field(call, SYSCALL_FIELD(int, nr), FILTER_OTHER);

I got it :-)
In conclusion, output of format (/sys/kernel/debug/tracing/events/syscalls/*/format)
has __syscall_nr for syscall number but the kernel structure 'syscall_trace_enter' and 'syscall_trace_exit' have not __syscall_nr variable.
Is it right ?

Quoted 7 lines · On Sun, 28 Feb 2016 01:10:19 +0900 Taeung Song <[email protected]> wrote:

I got it :-)
In conclusion, output of format (/sys/kernel/debug/tracing/events/syscalls/*/format)
has __syscall_nr for syscall number but the kernel structure 'syscall_trace_enter' and 'syscall_trace_exit' have not __syscall_nr variable.
Is it right ?

Correct. Just change the output of the format file. Not the data structures in the kernel.

Thanks,

-- Steve

Ingo, Peter, have you guys followed this case?

Summary: Some tracepoint have multiple fields with the same name, 'nr',
the first one is a unique syscall ID, the other is a syscall argument:

Quoted 6 lines

Ingo, Peter, have you guys followed this case?

Summary: Some tracepoint have multiple fields with the same name, 'nr',
the first one is a unique syscall ID, the other is a syscall argument:

I'm all for pushing the limits on tracepoint ABI. I'm all for making it less rigid. So if we can get away with changing this under the promise of helping fixup fallout, I'm all for it.

That means we can always just change tracepoints, and they're not really ABI at all.