-
Notifications
You must be signed in to change notification settings - Fork 22
Description
Hi, when I installed the latest version of kdump-utils from CentOS Stream 10 (specifically kdump-utils-1.0.58-4.el10) on Red Hat Enterprise Linux 10, the kdump.service takes a very long time to start on every system boot, which can be seen by running systemd-analyze blame | grep -i kdump.
This is clearly visible when running:
$ yum remove kdump-utils.x86_64
$ wget https://mirror.stream.centos.org/10-stream/BaseOS/x86_64/os/Packages/kdump-utils-1.0.58-4.el10.x86_64.rpm
$ rpm -ivh kdump-utils-1.0.58-4.el10.x86_64.rpm
$ reboot
$ systemd-analyze blame | grep -i kdump
38.899s kdump.service
$ time systemctl restart kdump.service
real 0m16.528s
user 0m0.004s
sys 0m0.007s
In fact, this issue can also be reproduced on other operating systems in the RHEL 10 series.
Through investigation and verification, I’ve identified that this performance regression is caused by a specific patch: kdump-lib-initramfs: rewrite kdump_get_conf_val
After the original regular expression was modified, the code execution efficiency became very low, resulting in long startup times.
The log I obtained by adding set -x at the top of the file /usr/lib/kdump/kdump-lib-initramfs.sh is as follows:
This is also reported in RHEL-137945