Re: [PATCH] trace-cmd: Make read nonblocking in find_time_stamp().
From: Steven Rostedt <rostedt@goodmis.org>
Date: 2015-02-09 22:51:57
Also in:
lkml
On Mon, 4 Nov 2013 16:51:48 -0800 Aaron Fabbri [off-list ref] wrote:
This only affects the tsc -> time of day correlation when doing `trace-cmd record --date`. get_date_to_ts() writes a dummy event to trace_marker and then reads it back. Reading it back is done by looping over all per-cpu raw trace pipes and trying to read. This read should be non-blocking, because the dummy event can end up on any CPU. Signed-off-by: Aaron Fabbri <redacted>
Hmm, I somehow missed this patch. I just applied (way late :-/), and I'm placing it in the stable releases. I'm starting my testing now. -- Steve
quoted hunk ↗ jump to hunk
--- trace-record.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)diff --git a/trace-record.c b/trace-record.c index 1b1d293..dca8a92 100644 --- a/trace-record.c +++ b/trace-record.c@@ -2092,7 +2092,7 @@ static unsigned long long find_time_stamp(structpevent *pevent) continue; sprintf(file, "%s/%s/trace_pipe_raw", path, name); - fd = open(file, O_RDONLY); + fd = open(file, O_RDONLY | O_NONBLOCK); if (fd < 0) continue; do {