3

I have a strange boot issue on a Sun T5220 SPARC server with Solaris 10.

The last few changes I did on this server, for maintenance purposes were:

  • to delete some old directories (definitely not anything system-related, just some directories that I’m sure they are containing unused files) to free up some space
  • to install the library libiconv.1.14 for Solaris10 SPARC64
  • to install the patch 153074-01, to patch a known SSH-related vulnerability. Unfortunately, I forgot to boot in single-user mode before installing it.

Then, I rebooted the server from ILOM (I logged on in the console by start /SP/console and issued reboot). In the console, all I can see are the following lines printed again and again:

Boot device: /pci@0/pci@0/pci@2/scsi@0/disk@0,0:a  File and args: kernel/sparcv9/unix
SunOS Release 5.10 Version Generic_150400-59 64-bit
Copyright (c) 1983, 2017, Oracle and/or its affiliates. All rights reserved.
Hostname: unknown
The / file system (/dev/rdsk/c1t0d0s0) is being checked.
syncing file systems... done
rebooting...
Resetting...

so, the server cannot boot the OS. How can I troubleshoot this situation?

Kapoios
  • 91
  • 6
  • *to install the library libiconv.1.14 for Solaris10 SPARC64* You shouldn't have needed to do that. See this: https://www.unix.com/solaris/226909-need-install-usr-lib-libiconv-so-2-solaris-sparc-64-bit-package.html You probably need to boot from a CD to recover your system. You need to learn about [Live Upgrade](https://docs.oracle.com/cd/E18752_01/html/821-1910/luexample-100.html) and never directly update the current boot environment like that. – Andrew Henle Jan 09 '21 at 14:40
  • @AndrewHenle Thanks for the valuable info! Could the installation of libiconv be the reason of boot failure? If yes, I might just have to remove the package installation if I manage to boot into single mode. – Kapoios Jan 11 '21 at 09:58
  • 1
    Just adding a library and maybe a header file or two *shouldn't* break anything, but if it overwrote something it's possible. How did you install it, and do you know where the installation put its files? Given the `The / file system (/dev/rdsk/c1t0d0s0) is being checked.` message, though, it's more likely the patch done outside of single-user mode corrupted the root file system. – Andrew Henle Jan 11 '21 at 11:05
  • @AndrewHenle Yes, I was absent minded and installed it in multi-user mode, with pkgadd. Forgot to write that I can't boot in single user mode now (I got the same messages). To rescue-boot from the installation DVD is my last resort. So, let's say that I manage to boot from the installation medium. Is it sufficient to mount the root dir of the medium in the disk and check the integrity of the packets (via pkgchk) or not? Any advice? – Kapoios Jan 11 '21 at 11:12

1 Answers1

1

I haven't managed to find out the root cause of this problem. Nevertheless, a way to resolve it is to:

  1. Switch to the OK prompt.
  2. Boot in failsafe mode (ok: boot -F failsafe) and mount the root mount point on /a.
  3. cd /a/dev and del null
  4. Fix the broken OS files and symlinks (pkgchk -R /a -f SUNWcsd)
  5. Reboot
Kapoios
  • 91
  • 6