[reboot] User-friendly reboot cause message for kernel panic

[yozhao101]

Signed-off-by: Yong Zhao [email protected]

What I did

If the rebooting of SONiC device was caused by kernel panic, then the CLI command show reboot-cause should show Kernel Panic.

How I did it

Currently if kernel was panicked, then the device would be rebooted. The reboot script wrote a message into reboot-cause.txt. I just updated the content of this message.

How to verify it

I verified this change on the virtual switch in the following steps:

1. Trigger kernel panic: echo c > /proc/sysrq-trigger
2. After device was rebooted, run the CLI show reboot-cause:
   admin@vlab-01:~$ show reboot-cause
Kernel Panic [Time: Tue 09 Mar 2021 03:03:56 AM UTC]

Previous command output (if the output of a command-line utility has changed)

admin@vlab-01:~$ show reboot-cause
User issued 'kdump' command [User: kdump, Time: Mon 08 Mar 2021 01:47:43 AM UTC]

New command output (if the output of a command-line utility has changed)

admin@vlab-01:~$ show reboot-cause
Kernel Panic [Time: Tue 09 Mar 2021 03:03:56 AM UTC]

[jleveque]

@sujinmkang: We want to rephrase this message to be more user-friendly. However, we may need to either adjust the syntax here or in the process-reboot-cause parsing logic to parse the fields properly. Please provide your feedback.

[sujinmkang]

@jleveque does this PR also wants to add the detail kernel panic information or just inform the kernel crash as last reboot cause?

[jleveque]

@jleveque does this PR also wants to add the detail kernel panic information or just inform the kernel crash as last reboot cause?

Just inform that kernel panic was the last reboot cause.

[sujinmkang]

@jleveque If the kernel panic happens after passing this line, I mean, if it happens during reboot, then there is still having a chance to miss the kernel panic? I think it will be good to have the core file directory or file name which is related to the kernel panic. We can add it from process-reboot-cause.

[jleveque]

@jleveque If the kernel panic happens after passing this line, I mean, if it happens during reboot, then there is still having a chance to miss the kernel panic? I think it will be good to have the core file directory or file name which is related to the kernel panic. We can add it from process-reboot-cause.

I see your point, but if a kernel panic occurs during reboot, it will be difficult (possibly impossible, e.g., if the filesystem is read-only) to leave a breadcrumb that we can use after booting back up to determine that a kernel panic occurred.

Also, technically speaking, if a user issues one of the reboot commands and a kernel panic happens during the reboot process, what was the cause of the reboot? Technically the reboot was initiated by the user.

[sujinmkang]

https://github.com/Azure/sonic-buildimage/blob/28cb43cb42e3223cade2efa9a5f60542d97a89e7/src/sonic-host-services/scripts/determine-reboot-cause#L129
@yozhao101 please add the kernel panic case if you want to change the reboot-cause here.

[sujinmkang]

And this too : https://github.com/Azure/sonic-utilities/blob/3a482ac38f1440fa0cbc00f621dc94ebb551757b/show/reboot_cause.py#L26

[jleveque]

No need for "Cause:` prefix. Also, we lose the "User issued '' command" syntax here. Instead we just see the command.

Suggest keeping format the same as before.

Something like the following:

reboot_cause_dict = read_reboot_cause_file()

reboot_cause = reboot_cause_dict .get("cause", "Unknown")
reboot_user = reboot_cause_dict .get("user", "N/A")
reboot_time = reboot_cause_dict .get("time", "N/A")

if reboot_user != "N/A":
    reboot_cause_str = "User issued '<command>' command".format(reboot_cause)
else:
    reboot_cause_str = reboot_cause

if reboot_user != "N/A" or reboot_time != "N/A":
    reboot_cause_str += " ["

    if reboot_user != "N/A":
        reboot_cause_str += "User: {}".format(reboot_user )
        if reboot_time != "N/A":
            reboot_cause_str += ", "

    if reboot_time != "N/A":
        reboot_cause_str += "Time: {}".format(reboot_time)

    reboot_cause_str += "]"

[yozhao101]

Retest this please.