Re: [Powertop] [PATCH] Fix running failure when > 69 CPUs for open file limitation
On (05/29/13 11:44), Youquan Song wrote:
Powertop 2.3 fails to run on machine with > 69 CPUs because system
open files
limitation of one process is 1024 default, While powertop will open one file
for every monitored perf_event (at least 15) each CPU.
Like on 80 CPUs Westmere-EX machine, powertop will fail to run with below:
PowerTOP v2.3 needs the kernel to support the 'perf' subsystem
as well as support for trace points in the kernel:
CONFIG_PERF_EVENTS=y
CONFIG_PERF_COUNTERS=y
CONFIG_TRACEPOINTS=y
CONFIG_TRACING=y
This patch is to change RLIMIT_NOFILE from default (1024) to max limition.
Hello,
I'd like to ask you to try out the following patch. We leak fd in read_file().
------8<--------8<--------
- Do not leak file descriptor in perf_bundle read_file().
- Clear perf event (unmap memory and close fd) on event destruction.
Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky(a)gmail.com>
---
diff --git a/src/perf/perf.cpp b/src/perf/perf.cpp
index 35b4017..e919741 100644
--- a/src/perf/perf.cpp
+++ b/src/perf/perf.cpp
@@ -167,6 +167,7 @@ perf_event::~perf_event(void)
if (perf_event::pevent->ref_count == 1) {
pevent_free(perf_event::pevent);
perf_event::pevent = NULL;
+ clear();
} else
pevent_unref(perf_event::pevent);
}
diff --git a/src/perf/perf_bundle.cpp b/src/perf/perf_bundle.cpp
index 38e1e91..8d480e8 100644
--- a/src/perf/perf_bundle.cpp
+++ b/src/perf/perf_bundle.cpp
@@ -123,6 +123,7 @@ static char * read_file(const char *file)
buffer[len] = '\0';
}
out:
+ close(fd);
return buffer;
}