From mboxrd@z Thu Jan 1 00:00:00 1970 From: Cyril Hrubis Date: Fri, 15 Nov 2019 10:52:34 +0100 Subject: [LTP] ??? FAIL: Test report for kernel 5.4.0-rc6-dd89262.cki (mainline.kernel.org) In-Reply-To: <976393725.11648955.1573552572246.JavaMail.zimbra@redhat.com> References: <232041279.11531466.1573491604178.JavaMail.zimbra@redhat.com> <976393725.11648955.1573552572246.JavaMail.zimbra@redhat.com> Message-ID: <20191115095234.GA18488@rei.lan> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: ltp@lists.linux.it Hi! > > perf_event_open02 again: > > I have a system [1] where it's consistently reproducible. > It looks like count_hardware_counters() reports incorrect number. > time_enabled and time_running are different but only barely (~0.7%). > > # ./perf_event_open02 -v > at iteration:0 value:300357893 time_enabled:55368608 time_running:55368608 > at iteration:1 value:600712498 time_enabled:54808179 time_running:54403148 > perf_event_open02 0 TINFO : overall task clock: 54825931 > perf_event_open02 0 TINFO : hw sum: 1442538620, task clock sum: 260919342 > hw counters: 288507724 288507724 288507724 288507724 288507724 > task clock counters: 51773047 52526664 52531694 52540386 51547551 > perf_event_open02 0 TINFO : ratio: 4.759050 > perf_event_open02 1 TFAIL : perf_event_open02.c:394: test failed (ratio was greater than ) > > Also do_work() completes pretty fast compared to a different host: > > real 0m0.176s > user 0m0.163s > sys 0m0.003s I was thinking of setting up an CPU time alarm that would stop that loop so that it will run for reasonably defined length regardless the CPU speed... -- Cyril Hrubis chrubis@suse.cz