35a9ec2956
- Fix lparstat error with mixed SMT state - Fix negative values seen while running lparstat with -E option
47 lines
1.8 KiB
Diff
47 lines
1.8 KiB
Diff
commit dee15756bcb287ccf39a904be07c90107b13844b
|
|
Author: Laurent Dufour <ldufour@linux.ibm.com>
|
|
Date: Wed May 3 10:50:15 2023 +0200
|
|
|
|
lparstat: Fix offline threads uninitialized entries
|
|
|
|
When some threads are offline, lparstat -E is failing like that:
|
|
|
|
$ ppc64_cpu --info # CPU 20 is offline
|
|
Core 0: 0* 1* 2* 3* 4* 5* 6* 7*
|
|
Core 1: 8* 9* 10* 11* 12* 13* 14* 15*
|
|
Core 2: 16* 17* 18* 19* 20 21* 22* 23*
|
|
Core 3: 24* 25* 26* 27* 28* 29* 30* 31*
|
|
Core 4: 32* 33* 34* 35* 36* 37* 38* 39*
|
|
Core 5: 40* 41* 42* 43* 44* 45* 46* 47*
|
|
$ lparstat -E
|
|
Failed to read /sys/devices/system/cpu/cpu0/spurr
|
|
|
|
The message is complaining about CPU0 but the real issue is that in
|
|
parse_sysfs_values() the test cpu_sysfs_fds[i].spurr >= 0 is valid even if
|
|
the entry has not been initialized (cpu_sysfs_fds is alloc cleared). So
|
|
if the number of threads online seen in assign_cpu_sysfs_fds is lower than
|
|
threads_in_system, the loop in parse_sysfs_values() will read uninitialized
|
|
entry, where .cpu=0.
|
|
|
|
To prevent that, unset entries in the cpu_sysfs_fds should have the spurr
|
|
fd set to -1.
|
|
|
|
Signed-off-by: Laurent Dufour <ldufour@linux.ibm.com>
|
|
Signed-off-by: Tyrel Datwyler <tyreld@linux.ibm.com>
|
|
|
|
diff --git a/src/lparstat.c b/src/lparstat.c
|
|
index a9e7bce..d2fdb3f 100644
|
|
--- a/src/lparstat.c
|
|
+++ b/src/lparstat.c
|
|
@@ -163,6 +163,10 @@ static int assign_cpu_sysfs_fds(int threads_in_system)
|
|
cpu_idx++;
|
|
}
|
|
|
|
+ /* Mark extra slots for offline threads unset, see parse_sysfs_values */
|
|
+ for (; cpu_idx < threads_in_system; cpu_idx++)
|
|
+ cpu_sysfs_fds[cpu_idx].spurr = -1;
|
|
+
|
|
return 0;
|
|
error:
|
|
fprintf(stderr, "Failed to open %s: %s\n",
|