tracing: Update event filters for multibuffer
authorTom Zanussi <tom.zanussi@linux.intel.com>
Thu, 24 Oct 2013 13:34:17 +0000 (08:34 -0500)
committerSteven Rostedt <rostedt@goodmis.org>
Tue, 5 Nov 2013 21:50:20 +0000 (16:50 -0500)
commitf306cc82a93d6b19f01634b80c580b9755c8b7cc
tree1a9ef8d44ed192185e4d6da5f7154f9a1c2075cb
parentf02b625d0341519238ab3d9cc8706ff4bd45fb89
tracing: Update event filters for multibuffer

The trace event filters are still tied to event calls rather than
event files, which means you don't get what you'd expect when using
filters in the multibuffer case:

Before:

  # echo 'bytes_alloc > 8192' > /sys/kernel/debug/tracing/events/kmem/kmalloc/filter
  # cat /sys/kernel/debug/tracing/events/kmem/kmalloc/filter
  bytes_alloc > 8192
  # mkdir /sys/kernel/debug/tracing/instances/test1
  # echo 'bytes_alloc > 2048' > /sys/kernel/debug/tracing/instances/test1/events/kmem/kmalloc/filter
  # cat /sys/kernel/debug/tracing/events/kmem/kmalloc/filter
  bytes_alloc > 2048
  # cat /sys/kernel/debug/tracing/instances/test1/events/kmem/kmalloc/filter
  bytes_alloc > 2048

Setting the filter in tracing/instances/test1/events shouldn't affect
the same event in tracing/events as it does above.

After:

  # echo 'bytes_alloc > 8192' > /sys/kernel/debug/tracing/events/kmem/kmalloc/filter
  # cat /sys/kernel/debug/tracing/events/kmem/kmalloc/filter
  bytes_alloc > 8192
  # mkdir /sys/kernel/debug/tracing/instances/test1
  # echo 'bytes_alloc > 2048' > /sys/kernel/debug/tracing/instances/test1/events/kmem/kmalloc/filter
  # cat /sys/kernel/debug/tracing/events/kmem/kmalloc/filter
  bytes_alloc > 8192
  # cat /sys/kernel/debug/tracing/instances/test1/events/kmem/kmalloc/filter
  bytes_alloc > 2048

We'd like to just move the filter directly from ftrace_event_call to
ftrace_event_file, but there are a couple cases that don't yet have
multibuffer support and therefore have to continue using the current
event_call-based filters.  For those cases, a new USE_CALL_FILTER bit
is added to the event_call flags, whose main purpose is to keep the
old behavior for those cases until they can be updated with
multibuffer support; at that point, the USE_CALL_FILTER flag (and the
new associated call_filter_check_discard() function) can go away.

The multibuffer support also made filter_current_check_discard()
redundant, so this change removes that function as well and replaces
it with filter_check_discard() (or call_filter_check_discard() as
appropriate).

Link: http://lkml.kernel.org/r/f16e9ce4270c62f46b2e966119225e1c3cca7e60.1382620672.git.tom.zanussi@linux.intel.com
Signed-off-by: Tom Zanussi <tom.zanussi@linux.intel.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
15 files changed:
include/linux/ftrace_event.h
include/linux/syscalls.h
include/trace/ftrace.h
kernel/trace/trace.c
kernel/trace/trace.h
kernel/trace/trace_branch.c
kernel/trace/trace_events.c
kernel/trace/trace_events_filter.c
kernel/trace/trace_export.c
kernel/trace/trace_functions_graph.c
kernel/trace/trace_kprobe.c
kernel/trace/trace_mmiotrace.c
kernel/trace/trace_sched_switch.c
kernel/trace/trace_syscalls.c
kernel/trace/trace_uprobe.c