linux - 在 Centos 8 上升级软件包后出现非法指令(核心转储)

Question

当我输入“yum update -y”命令时，我的 centos 8 出现了一些问题。之后，当我运行 yum 或任何 python 脚本时，我开始收到非法指令(核心转储)消息。
在升级期间，我遇到了这样的错误:

  Upgrading        : NetworkManager-libnm-1:1.30.0-0.5.el8.x86_64                                                        18/168 
  Running scriptlet: NetworkManager-libnm-1:1.30.0-0.5.el8.x86_64                                                        18/168 
  Running scriptlet: NetworkManager-1:1.30.0-0.5.el8.x86_64                                                              19/168 
  Upgrading        : NetworkManager-1:1.30.0-0.5.el8.x86_64                                                              19/168 
  Running scriptlet: NetworkManager-1:1.30.0-0.5.el8.x86_64                                                              19/168 
/var/tmp/rpm-tmp.1WBtNY: line 4:  4581 Illegal instruction     (core dumped) firewall-cmd --reload --quiet

  Upgrading        : elfutils-libs-0.182-3.el8.x86_64                                                                    20/168 
  Upgrading        : elfutils-debuginfod-client-0.182-3.el8.x86_64                                                       21/168 
  Upgrading        : python3-syspurpose-1.28.9-1.el8.x86_64                                                              22/168 

  Running scriptlet: selinux-policy-targeted-3.14.3-59.el8.noarch                                                        58/168 
  Upgrading        : selinux-policy-targeted-3.14.3-59.el8.noarch                                                        58/168 
  Running scriptlet: selinux-policy-targeted-3.14.3-59.el8.noarch                                                        58/168 
  Upgrading        : subscription-manager-1.28.9-1.el8.x86_64                                                            59/168 
  Running scriptlet: subscription-manager-1.28.9-1.el8.x86_64                                                            59/168 
chmod: cannot access '/etc/pki/entitlement/*.pem': No such file or directory

  Upgrading        : lvm2-8:2.03.11-0.4.20201222gitb84a992.el8.x86_64                                                    60/168 
  Running scriptlet: lvm2-8:2.03.11-0.4.20201222gitb84a992.el8.x86_64                                                    60/168 
  Upgrading        : kpartx-0.8.4-7.el8.x86_64                                                                           61/168 
  Upgrading        : xfsprogs-5.0.0-7.el8.x86_64                                                                         62/168 


  Cleanup          : elfutils-default-yama-scope-0.182-2.el8.noarch                                                     154/168 
  Cleanup          : tzdata-java-2020d-1.el8.noarch                                                                     155/168 
  Cleanup          : samba-common-4.13.2-5.el8.noarch                                                                   156/168 
  Running scriptlet: NetworkManager-1:1.30.0-0.4.el8.x86_64                                                             157/168 
  Cleanup          : NetworkManager-1:1.30.0-0.4.el8.x86_64                                                             157/168 
  Running scriptlet: NetworkManager-1:1.30.0-0.4.el8.x86_64                                                             157/168 
/var/tmp/rpm-tmp.M7HXpR: line 4:  9248 Illegal instruction     (core dumped) firewall-cmd --reload --quiet

  Cleanup          : NetworkManager-libnm-1:1.30.0-0.4.el8.x86_64                                                       158/168 
  Running scriptlet: NetworkManager-libnm-1:1.30.0-0.4.el8.x86_64                                                       158/168 
  Cleanup          : perl-libs-4:5.26.3-417.el8.x86_64                                                                  159/168 


  Cleanup          : python3-libs-3.6.8-33.el8.x86_64                                                                   164/168 
  Cleanup          : glibc-2.28-141.el8.x86_64                                                                          165/168 
  Cleanup          : glibc-langpack-en-2.28-141.el8.x86_64                                                              166/168 
  Cleanup          : glibc-common-2.28-141.el8.x86_64                                                                   167/168 
  Cleanup          : tzdata-2020d-1.el8.noarch                                                                          168/168 
  Running scriptlet: libwbclient-4.13.3-1.el8.x86_64                                                                    168/168 
  Running scriptlet: tuned-2.15.0-1.el8.noarch                                                                          168/168 
  Running scriptlet: kmod-kvdo-6.2.4.26-76.el8.x86_64                                                                   168/168 
sort: fflush failed: 'standard output': Broken pipe
sort: write error

gzip: stdout: Broken pipe

gzip: stdout: Broken pipe
sort: write failed: 'standard output': Broken pipe
sort: write error
sort: fflush failed: 'standard output': Broken pipe
sort: write error

gzip: stdout: Broken pipe

gzip: stdout: Broken pipe
sort: write failed: 'standard output': Broken pipe
sort: write error

  Running scriptlet: tzdata-2020d-1.el8.noarch                                                                          168/168 
[/usr/lib/tmpfiles.d/pesign.conf:1] Line references path below legacy directory /var/run/, updating /var/run/pesign вЖТ /run/pesign; please update the tmpfiles.d/ drop-in file accordingly.

  Running scriptlet: glibc-common-2.28-145.el8.x86_64                                                                   168/168 
  Verifying        : python3-pexpect-4.3.1-3.el8.noarch                                                                   1/168 
  Verifying        : python3-ptyprocess-0.5.2-4.el8.noarch                                                                2/168 
  Verifying        : abattis-cantarell-fonts-0.0.25-6.el8.noarch                                                          3/168 

现在问题在于 yum 或一些 python 包的导入。但是 curl 没问题。

[root@ddns-ard-2 ~]# 
[root@ddns-ard-2 ~]# yum --version    
Illegal instruction (core dumped)
[root@ddns-ard-2 ~]# yum upgrade   
Illegal instruction (core dumped)
[root@ddns-ard-2 ~]# python3
Python 3.6.8 (default, Dec 22 2020, 19:04:08) 
[GCC 8.4.1 20200928 (Red Hat 8.4.1-1)] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import sys
>>> import pymongo
Illegal instruction (core dumped)
[root@ddns-ard-2 ~]# rpm -q curl
curl-7.61.1-17.el8.x86_64
[root@ddns-ard-2 ~]# curl https://www.google.com | more
  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
  0     0    0     0    0     0      0      0 --:--:-- --:--:-- --:--:--     0<!doctype html>

来自我的/var/log/messages 的消息

Jan 12 02:18:02 ddns-ard-2 kernel: traps: yum[3731] trap invalid opcode ip:7efdef6dd59f sp:7ffef19166c0 error:0 in libm-2.28.so[7efdef655000+181000]
Jan 12 02:18:02 ddns-ard-2 systemd[1]: Started Process Core Dump (PID 3732/UID 0).
Jan 12 02:18:02 ddns-ard-2 systemd-coredump[3733]: Resource limits disable core dumping for process 3731 (yum).
Jan 12 02:18:02 ddns-ard-2 systemd-coredump[3733]: Process 3731 (yum) of user 0 dumped core.
Jan 12 02:18:02 ddns-ard-2 systemd[1]: systemd-coredump@278-3732-0.service: Succeeded.

我在互联网上建立的解决方法没有帮助

[root@ddns-ard-2 ~]# export NSS_DISABLE_HW_AES=1 
[root@ddns-ard-2 ~]# yum --version
Illegal instruction (core dumped)
[root@ddns-ard-2 ~]# export NSS_DISABLE_HW_GCM=1
[root@ddns-ard-2 ~]# yum --version              
Illegal instruction (core dumped)
[root@ddns-ard-2 ~]# 
[root@ddns-ard-2 ~]# dnf
Illegal instruction (core dumped)
[root@ddns-ard-2 ~]# 

任何想法如何解决这个问题？
我有 5 个具有相同内核的虚拟 CPU
以下是关于其中之一的信息:

[root@ddns-ard-2 ~]#  cat /proc/cpuinfo
processor       : 0
vendor_id       : GenuineIntel
cpu family      : 6
model           : 63
model name      : Intel(R) Xeon(R) CPU E5-2699 v3 @ 2.30GHz
stepping        : 2
microcode       : 0x38
cpu MHz         : 2294.686
cache size      : 46080 KB
physical id     : 0
siblings        : 1
core id         : 0
cpu cores       : 1
apicid          : 0
initial apicid  : 0
fpu             : yes
fpu_exception   : yes
cpuid level     : 15
wp              : yes
flags           : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts mmx fxsr sse sse2 ss syscall nx rdtscp lm constant_tsc arch_perfmon pebs bts nopl xtopology tsc_reliable nonstop_tsc cpuid aperfmperf pni pclmulqdq ssse3 fma cx16 sse4_1 sse4_2 movbe popcnt aes xsave avx hypervisor lahf_lm epb pti cqm_llc cqm_occup_llc dtherm ida arat pln pts
bugs            : cpu_meltdown spectre_v1 spectre_v2 spec_store_bypass l1tf mds swapgs itlb_multihit
bogomips        : 4589.37
clflush size    : 64
cache_alignment : 64
address sizes   : 40 bits physical, 48 bits virtual
power management:

Answer 1

如果管理程序屏蔽了 XSAVE，就会发生这种情况。支持，但不掩盖 FMA CPUID 中的特征位。这有点奇怪的管理程序配置，这就是为什么它没有在测试中被捕获。在管理程序中选择不同的 CPU 型号应该可以解决这个问题(支持 AVX2 的东西应该可以做到)。作为副作用，您可能会从 VM 中获得更好的性能。
一些人通过适当的 channel 报告了这一点，并且已经发布了更新:

glibc: FMA4 math routines are not selected properly after bug 1817513

RHSA-2021:1585: glibc security, bug fix, and enhancement update