tracing: handle unregistering the current tracer
PATCH 1/1 tracing: handle unregistering the current tracer
Impact: simplification
Instead of requiring that plugins have the sequence:
my_tracer_stop(my_trace_array); unregister_tracer(my_tracer);
it should be possible just do a:
unregister_tracer(my_tracer);
--- kernel/trace/trace.c | 9 +++++++++ 1 files changed, 9 insertions(+), 0 deletions(-)
kernel/trace/trace.c +9 -0
@@ -559,6 +559,15 @@ void unregister_tracer(struct tracer *type) found: *t = (*t)->next; + +if (type == current_trace && tracer_enabled) { + tracer_enabled =0 ; +tracing_stop (); +if (current_trace->stop) + current_trace->stop (&global_trace); + current_trace = &nop_trace; + } +if (strlen (type->name) != max_tracer_type_len) goto out;
Signature · 1 line
1.6.0.6
Ingo Molnar
AppliedQuoted 7 lines · Arnaldo Carvalho de Melo
Impact: simplification
Instead of requiring that plugins have the sequence:
my_tracer_stop(my_trace_array); unregister_tracer(my_tracer);
Quoted 8 lines
it should be possible just do a:
unregister_tracer(my_tracer);Signed-off-by Arnaldo Carvalho de Melo <[email protected]>--- kernel/trace/trace.c | 9 +++++++++ 1 files changed, 9 insertions(+), 0 deletions(-)
Applied to tip/tracing/ftrace, thanks Arnaldo!
Nice one - we should try to shoot for the smallest physically possible linecount for any new (possibly trivial) tracer plugin.
Ingo
Em Mon, Feb 09, 2009 at 10:58:16AM +0100, Ingo Molnar escreveu:
Quoted 19 lines · Arnaldo Carvalho de Melo
Quoted 8 lines
Impact: simplification
Instead of requiring that plugins have the sequence:
my_tracer_stop(my_trace_array); unregister_tracer(my_tracer);Quoted 8 lines
it should be possible just do a:
unregister_tracer(my_tracer);Signed-off-by Arnaldo Carvalho de Melo <[email protected]>--- kernel/trace/trace.c | 9 +++++++++ 1 files changed, 9 insertions(+), 0 deletions(-)
Quoted 4 lines
Applied to tip/tracing/ftrace, thanks Arnaldo!
Nice one - we should try to shoot for the smallest physically possible linecount for any new (possibly trivial) tracer plugin.
That and also for as few as possible EXPORT_SYMBOL_GPL entries, now, for converting net/ipv4/tcp_probe.c to an ftrace plugin + one tracepoint at tcp_rcv_established we need:
doppio linux-2.6-tip]$ grep ^+EXPORT_SYMBOL ~/tcp_cwnd_tracer.patch +EXPORT_SYMBOL_GPL(register_tracer); +EXPORT_SYMBOL_GPL(unregister_tracer); +EXPORT_SYMBOL_GPL(tracing_record_cmdline); +EXPORT_SYMBOL_GPL(trace_buffer_lock_reserve); +EXPORT_SYMBOL_GPL(trace_buffer_unlock_commit); +EXPORT_SYMBOL_GPL(trace_seq_printf); +EXPORT_SYMBOL_GPL(register_ftrace_event); +EXPORT_SYMBOL_GPL(unregister_ftrace_event); [acme@doppio linux-2.6-tip]$
- Arnaldo
Quoted 38 lines · Arnaldo Carvalho de Melo
Em Mon, Feb 09, 2009 at 10:58:16AM +0100, Ingo Molnar escreveu:
Quoted 19 lines · Arnaldo Carvalho de Melo
Quoted 8 lines
Impact: simplification
Instead of requiring that plugins have the sequence:
my_tracer_stop(my_trace_array); unregister_tracer(my_tracer);Quoted 8 lines
it should be possible just do a:
unregister_tracer(my_tracer);Signed-off-by Arnaldo Carvalho de Melo <[email protected]>--- kernel/trace/trace.c | 9 +++++++++ 1 files changed, 9 insertions(+), 0 deletions(-)Quoted 4 lines
Applied to tip/tracing/ftrace, thanks Arnaldo!
Nice one - we should try to shoot for the smallest physically possible linecount for any new (possibly trivial) tracer plugin.
That and also for as few as possible
EXPORT_SYMBOL_GPLentries, now, for convertingnet/ipv4/tcp_probe.cto an ftrace plugin + one tracepoint attcp_rcv_establishedwe need:doppio linux-2.6-tip]$ grep ^+EXPORT_SYMBOL
~/tcp_cwnd_tracer.patch(quoted diff) +8
-0+EXPORT_SYMBOL_GPL(register_tracer); +EXPORT_SYMBOL_GPL(unregister_tracer); +EXPORT_SYMBOL_GPL(tracing_record_cmdline); +EXPORT_SYMBOL_GPL(trace_buffer_lock_reserve); +EXPORT_SYMBOL_GPL(trace_buffer_unlock_commit); +EXPORT_SYMBOL_GPL(trace_seq_printf); +EXPORT_SYMBOL_GPL(register_ftrace_event); +EXPORT_SYMBOL_GPL(unregister_ftrace_event);
Quoted 1 line
[acme@doppio linux-2.6-tip]$
ok. For in-kernel plugins it's of course fine to do those exports in any case.
Ingo