Intel PT decoding walks the object code to reconstruct the trace. A jump label change during tracing causes decoding errors.
The "Enable close-on-exec flag on perf file descriptor" patch caused there to be always a jump label change. It was found that using a per-cpu context instead of a per-thread context for the probe of the close-on-exec feature, made the problem go away. Signed-off-by: Adrian Hunter <[email protected]> --- tools/perf/util/cloexec.c | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-) diff --git a/tools/perf/util/cloexec.c b/tools/perf/util/cloexec.c index c5d05ec..6a37be5 100644 --- a/tools/perf/util/cloexec.c +++ b/tools/perf/util/cloexec.c @@ -1,3 +1,4 @@ +#include <sched.h> #include "util.h" #include "../perf.h" #include "cloexec.h" @@ -14,9 +15,13 @@ static int perf_flag_probe(void) }; int fd; int err; + int cpu = sched_getcpu(); + + if (cpu < 0) + cpu = 0; /* check cloexec flag */ - fd = sys_perf_event_open(&attr, 0, -1, -1, + fd = sys_perf_event_open(&attr, -1, cpu, -1, PERF_FLAG_FD_CLOEXEC); err = errno; @@ -30,7 +35,7 @@ static int perf_flag_probe(void) err, strerror(err)); /* not supported, confirm error related to PERF_FLAG_FD_CLOEXEC */ - fd = sys_perf_event_open(&attr, 0, -1, -1, 0); + fd = sys_perf_event_open(&attr, -1, cpu, -1, 0); err = errno; if (WARN_ONCE(fd < 0, -- 1.8.3.2 -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
