A failure return when using the command "pcm 1 -i=1" on CentOS stream 8.

[thebestsuper]

Hello, I use pcm tool to run on Intel CPU, the platform is Ice Lake. I found that the problem with the pcm tool happens randomly. When the problem occurs, it always shows like this: "Try using Linux performance events..." "Debug: Catch signal breaks (segmentation fault)."

The pcm tool I build on CentOS 8.5 and use CenOS stream 8 The attach file is pcm log and my OS-realease informatioin.

Please help me to check if it is a pcm tool problem. If there is anything I can do to help please let me know and I will do it asap.

Best Regards thebestsuper os-release.txt Intel_UPE_SpecificationsChart_2022_08_04.csv PCM_QPI_dump_20220804_1212_error.txt

[opcm]

thanks for reporting. We are looking into it. Does it crash if you run it with PCM_NO_PERF=1 env variable:

PCM_NO_PERF=1 pcm 1 -i=1

[thebestsuper]

Sorry for replying so late, After adding "PCM_NO_PERF=1", the failure problem no longer occurs. From the parameter description I only know that this parameter is used to program the core PMU without using the Linux perf events API (it is used by default). I wonder if this change might cause any difference to the previous test without adding "PCM_NO_PERF=1". If this change is the same as before, I will have no problems.

Best Regards thebestsuper

[opcm]

Sorry for the delayed reply. In this mode a different driver is used but the data output should be the same. Nevertheless we should understand the issue. Unfortunately I was not able to reproduce it. It would be very much appreciated if you could help with additional information: the call stack of the crash.

Please install gdb on your system. Then compile pcm in debug mode:

mkdir Debug cd Debug cmake -DCMAKE_BUILD_TYPE=Debug .. make

Then run pcm in gdb: gdb bin/pcm run 1 -i=1

When it crashes type bt and share the crash call stack which will be output by the bt command.

Thanks a lot in advance

[thebestsuper]

Hi developers, Sorry for replying so late, I used gdb to get pcm info on my machine and the log message is in the attachment. Since it happens randomly, I tried several rounds with about 5 seconds between each round. The figure below shows the failed state. The figure below shows the PASS state.

[ogbrugge]

Your "this" pointer is bogus in the failed state. Can you type in the command "bt" when it fails again?

[thebestsuper]

Sorry, I forgot to use the "bt" command to get the info. The figure below is the status failure, use the 'bt' command to get information. gdb.txt

[opcm]

thanks a lot. The 'bt' output was very helpful. I was able to identify the issue and we will work on a fix.

[opcm]

could you please try the latest version from the master branch? it should resolve the issue

[thebestsuper]

Hi developers, After using the new pcm tool on the machine, the problem no longer occurs. Thanks for solving this problem.