tracing: Have branch tracer use recursive field of task struct
authorSteven Rostedt (Red Hat) <rostedt@goodmis.org>
Tue, 7 Jul 2015 19:05:03 +0000 (15:05 -0400)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Mon, 3 Aug 2015 16:29:45 +0000 (09:29 -0700)
commitf1bb13070802869da9e5c993a86d905c41f1ec28
treeed2d41f51773ee9df18ae38c3f2c82c5e7acaca3
parent3f6ba7f88df0e94587a4b93319656d9ee2e5e203
tracing: Have branch tracer use recursive field of task struct

commit 6224beb12e190ff11f3c7d4bf50cb2922878f600 upstream.

Fengguang Wu's tests triggered a bug in the branch tracer's start up
test when CONFIG_DEBUG_PREEMPT set. This was because that config
adds some debug logic in the per cpu field, which calls back into
the branch tracer.

The branch tracer has its own recursive checks, but uses a per cpu
variable to implement it. If retrieving the per cpu variable calls
back into the branch tracer, you can see how things will break.

Instead of using a per cpu variable, use the trace_recursion field
of the current task struct. Simply set a bit when entering the
branch tracing and clear it when leaving. If the bit is set on
entry, just don't do the tracing.

There's also the case with lockdep, as the local_irq_save() called
before the recursion can also trigger code that can call back into
the function. Changing that to a raw_local_irq_save() will protect
that as well.

This prevents the recursion and the inevitable crash that follows.

Link: http://lkml.kernel.org/r/20150630141803.GA28071@wfg-t540p.sh.intel.com
Reported-by: Fengguang Wu <fengguang.wu@intel.com>
Tested-by: Fengguang Wu <fengguang.wu@intel.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
kernel/trace/trace.h
kernel/trace/trace_branch.c