Bug 2009378
| Summary: | perf script -F "weight" broken in RHEL-9 | |||
|---|---|---|---|---|
| Product: | Red Hat Enterprise Linux 9 | Reporter: | Joe Mario <jmario> | |
| Component: | kernel | Assignee: | Michael Petlan <mpetlan> | |
| kernel sub component: | Perf | QA Contact: | Qiao Zhao <qzhao> | |
| Status: | CLOSED ERRATA | Docs Contact: | ||
| Severity: | high | |||
| Priority: | unspecified | CC: | acme, efuller, xiawu, zsun | |
| Version: | 9.0 | Keywords: | Triaged | |
| Target Milestone: | rc | |||
| Target Release: | --- | |||
| Hardware: | All | |||
| OS: | Linux | |||
| Whiteboard: | ||||
| Fixed In Version: | kernel-5.14.0-24.el9 | Doc Type: | If docs needed, set a value | |
| Doc Text: | Story Points: | --- | ||
| Clone Of: | ||||
| : | 2026390 (view as bug list) | Environment: | ||
| Last Closed: | 2022-05-17 15:40:31 UTC | Type: | Bug | |
| Regression: | --- | Mount Type: | --- | |
| Documentation: | --- | CRM: | ||
| Verified Versions: | Category: | --- | ||
| oVirt Team: | --- | RHEL 7.3 requirements from Atomic Host: | ||
| Cloudforms Team: | --- | Target Upstream Version: | ||
| Embargoed: | ||||
| Bug Depends On: | ||||
| Bug Blocks: | 2026390 | |||
| Deadline: | 2021-12-06 | |||
fixed by: https://lore.kernel.org/lkml/1632929894-102778-1-git-send-email-kan.liang@linux.intel.com/ Added a testcase for this into perftool-testsuite: -- [ PASS ] -- perf_mem :: test_basic :: script -F weight Since the problem described in this bug report should be resolved in a recent advisory, it has been closed with a resolution of ERRATA. For information on the advisory (new packages: kernel), and where to find the updated files, follow the link below. If the solution does not work for you, open a new bug report. https://access.redhat.com/errata/RHBA-2022:3907 |
Description of problem: The following perf command is broken in RHEL-9: # perf script -F "comm,pid,cpu,event,addr,data_src,weight,ip,sym,phys_addr" Samples for 'cpu/mem-loads,ldlat=30/P' event do not have WEIGHT attribute set. Cannot print 'weight' field. Version-Release number of selected component (if applicable): How reproducible: Easy to reproduce. Steps to Reproduce: Just run a simple command, such as: # perf mem record strings /boot/vmlinuz-`uname -r` > /dev/null 2>&1 # perf script -F "weight" Actual results: # perf mem record strings /boot/vmlinuz-`uname -r` > /dev/null 2>&1 # perf script -F "weight" Samples for 'cpu/mem-loads,ldlat=30/P' event do not have WEIGHT attribute set. Cannot print 'weight' field. Expected results: # perf mem record strings /boot/vmlinuz-`uname -r` > /dev/null 2>&1 # perf script -F "weight" | head -10 118 56 819 51 238 57 281 326 417 148 Additional info: This feature is important to us. When triaging performance problems, it is very common to try and understand what load instructions are taking the longest and why. Here's a snipit for how we use it. The weight field shows the machine cycles needed for each load (sample). perf.jirka.v2 script -F "comm,pid,cpu,addr,weight,ip,sym" |sort -nrk5 | head -10 tugtest 77011 [004] 406100 2944 401868 acquire_lock tugtest 77011 [006] 406100 2868 401868 acquire_lock tugtest 77011 [006] 406100 2815 401868 acquire_lock tugtest 77011 [006] 406100 2797 401868 acquire_lock tugtest 77011 [004] 406100 2756 401868 acquire_lock tugtest 77011 [004] 406100 2741 401868 acquire_lock tugtest 77011 [004] 406100 2604 401868 acquire_lock tugtest 77011 [004] 406100 2582 401868 acquire_lock tugtest 77011 [006] 406100 2563 401868 acquire_lock tugtest 77011 [006] 406100 2557 401868 acquire_lock In the above example, the load latencies are only in the 2-3K range (not too bad). The more serious ones are often up in the 10K->60K range. Side node: This bug has already been fixed and the patch has been submitted upstream.