Lots of hisi_qm zombie task slow down system after stress test
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
kunpeng920 |
Fix Released
|
Undecided
|
Ike Panhc | ||
Ubuntu-18.04-hwe |
Fix Released
|
Undecided
|
Ike Panhc | ||
Ubuntu-20.04 |
Fix Released
|
Undecided
|
Ike Panhc | ||
linux (Ubuntu) |
Invalid
|
Undecided
|
Unassigned | ||
Focal |
Fix Released
|
Medium
|
Ike Panhc | ||
Hirsute |
Invalid
|
Undecided
|
Unassigned | ||
Impish |
Invalid
|
Undecided
|
Unassigned |
Bug Description
[Impact]
hisi_qm does not clean up kernel process after calculation is done. Many zombie processes slow down system. After checkbox cpu stress test, it takes more then 2min to ssh in.
[Test Plan]
1) stress-ng --aggressive --verify --timeout 300 --metrics-brief --tz --times --af-alg 0
2) ps aux | grep hisi_qm | wc -l
Expected result is less then 100
[Regression Risk]
hisi_qm only affects kunpeng920 platform. Minimal risk for other platform, and full regression test is needed on kunpeng920.
=======
[Bug Description]
With focal 5.4 kernel, crypto driver does not clean up its created process when calculation is done. Many zombie processes slow down system. e.g. Takes more then 10sec for ssh connection.
[Steps to Reproduce]
1) Install Ubuntu 20.04 with GA (5.4) kernel
2) sudo apt install -y stress-ng
3) stress-ng --aggressive --verify --timeout 300 --metrics-brief --tz --times --af-alg 0
4) ps aux | grep hisi_qm | wc -l
[Actual Results]
>100000
[Expected Results]
<100
[Reproducibility]
100%
[Additional information]
Can not reproduce with focal HWE (5.8) kernel.
[Resolution]
no longer affects: | kunpeng920/ubuntu-18.04 |
no longer affects: | kunpeng920/ubuntu-20.04-hwe |
Changed in linux (Ubuntu Focal): | |
status: | New → Confirmed |
description: | updated |
Changed in kunpeng920: | |
status: | New → In Progress |
Changed in linux (Ubuntu Focal): | |
status: | Triaged → In Progress |
Changed in linux (Ubuntu Focal): | |
importance: | Undecided → Medium |
Changed in linux (Ubuntu Focal): | |
status: | In Progress → Fix Committed |
Changed in kunpeng920: | |
status: | In Progress → Fix Committed |
Changed in kunpeng920: | |
status: | Fix Committed → Fix Released |
tags: |
added: verification-done-bionic removed: verification-failed-bionic |
This patch may be the fix, but can not clean cherry-pick to 5.4 kernel. Need to find a better way to solve the conflict.
commit b67202e8ed30bfa 07b07a6f8fc7624 17a9a4e6de
Author: Zhou Wang <email address hidden>
Date: Sat May 9 17:43:58 2020 +0800
crypto: hisilicon/qm - add state machine for QM