Re: [PATCH v2 0/4] perf: Make SIGTRAP and __perf_pending_irq() work on RT.

[Arnaldo Carvalho de Melo]

On Wed, Mar 13, 2024 at 03:30:52PM -0300, Arnaldo Carvalho de Melo wrote:
> On Wed, Mar 13, 2024 at 03:14:28PM -0300, Arnaldo Carvalho de Melo wrote:
> > 'perf test' doesn't show any regression, now I'm running Vince Weaver's
> > https://github.com/deater/perf_event_tests, storing the results with
> > this patchset and then without, to do a diff, lets see...
> 
> So things improved! I'll re-run to see if these results are stable...

tldr; No dmesg activity, no kernel splats, most tests passed, nothing
noticeable when running with/without the patch with Vince's regression
tests. So:


Tested-by: Arnaldo Carvalho de Melo <acme@xxxxxxxxxx>
Reported-by: Arnaldo Carvalho de Melo <acme@xxxxxxxxxx>

- Arnaldo

Further details:

Without the patch:

[root@nine perf_event_tests]# ./run_tests.sh | tee results.$(uname -r).new ; diff -u results.$(uname -r) results.$(uname -r).new
--- results.6.8.0-rc7-rt6	2024-03-13 15:26:37.923323518 -0300
+++ results.6.8.0-rc7-rt6.new	2024-03-13 15:32:43.983245095 -0300
@@ -296,7 +296,7 @@
   + tests/rdpmc/rdpmc_validation
     Testing if userspace rdpmc reads give expected results...  PASSED
   + tests/rdpmc/rdpmc_multiplexing
-    Testing if userspace rdpmc multiplexing works...           PASSED
+    Testing if userspace rdpmc multiplexing works...           FAILED
   + tests/rdpmc/rdpmc_reset
     Testing if resetting while using rdpmc works...            PASSED
   + tests/rdpmc/rdpmc_group
@@ -304,15 +304,15 @@
   + tests/rdpmc/rdpmc_attach
     Testing if rdpmc attach works...                           PASSED
   + tests/rdpmc/rdpmc_attach_cpu
-    Running on CPU 4
+    Running on CPU 0
 Testing if rdpmc behavior on attach CPU...                 PASSED
   + tests/rdpmc/rdpmc_attach_global_cpu
-    Running on CPU 6
+    Running on CPU 3
 Testing if rdpmc behavior on attach all procs on other CPU... FAILED
   + tests/rdpmc/rdpmc_attach_other_cpu
-    Measuring on CPU 5
-Running on CPU 6
-Measuring on CPU 5
+    Measuring on CPU 0
+Running on CPU 3
+Measuring on CPU 0
 Testing if rdpmc behavior on attach other CPU...           FAILED
   + tests/rdpmc/rdpmc_multiattach
     Testing if rdpmc multi-attach works...                     PASSED

A test flipped results.

Trying again with a more compact output:

[root@nine perf_event_tests]# ./run_tests.sh | tee results.$(uname -r).new ; diff -u results.$(uname -r) results.$(uname -r).new | grep ^[+-]
--- results.6.8.0-rc7-rt6	2024-03-13 15:26:37.923323518 -0300
+++ results.6.8.0-rc7-rt6.new	2024-03-13 17:06:34.944149451 -0300
-    Running on CPU 4
-Testing if rdpmc behavior on attach CPU...                 PASSED
-  + tests/rdpmc/rdpmc_attach_global_cpu
+Testing if rdpmc behavior on attach CPU...                 FAILED
+  + tests/rdpmc/rdpmc_attach_global_cpu
+    Running on CPU 0
-    Measuring on CPU 5
-Running on CPU 6
-Measuring on CPU 5
+    Measuring on CPU 7
+Running on CPU 1
+Measuring on CPU 7
[root@nine perf_event_tests]#

Since its that rdpmc that is now always failing without this patch
series, lets try using that .new as the new baseline:

[root@nine perf_event_tests]# ./run_tests.sh | tee results.$(uname -r).new2 ; diff -u results.$(uname -r).new results.$(uname -r).new2 | grep ^[+-]
--- results.6.8.0-rc7-rt6.new	2024-03-13 17:06:34.944149451 -0300
+++ results.6.8.0-rc7-rt6.new2	2024-03-13 17:08:41.438282558 -0300
-    Testing "branch-misses" generalized event...               FAILED
+    Testing "branch-misses" generalized event...               PASSED
-    Testing if userspace rdpmc multiplexing works...           PASSED
+    Testing if userspace rdpmc multiplexing works...           FAILED
-    Running on CPU 6
-Testing if rdpmc behavior on attach CPU...                 FAILED
+    Running on CPU 2
+Testing if rdpmc behavior on attach CPU...                 PASSED
-    Running on CPU 0
+    Running on CPU 2
-    Measuring on CPU 7
-Running on CPU 1
-Measuring on CPU 7
+    Measuring on CPU 2
+Running on CPU 0
+Measuring on CPU 2
[root@nine perf_event_tests]#