Ensure correct value from/to SERIALCONSOLE

[waynechen55]

• For OSD workers, some machines use ttyS0 and some others use ttyS1 or even ttyS2 which is assigned to setting SERIALDEV. But SERIALDEV is not safe enough to be used everywhere, because it becomes sshserial during test run, here is an example.

• Merge request introduces setting SERIALCONSOLE which is already being widely used in automation code. This can avoid incorrect SERIALCONSOLE leading to test run failure, for exmaple, this one.

• Current subroutine save_serial_console reset SERIALCONSOLE, if it has no value, to
  default value obtained from SERIALDEV which may become sshserial. So it is necessary to check and assign more meaningful default value ttyS1 to SERIALCONSOLE. And ttyS1 is also the most widely used serial console.

• Please also refer to update console for more details.

• Verification Runs:

  • Debug Test: SERIALCONSOLE has its value
  • Debug Test: SERIALCONSOLE has no value and SERIALDEV is sshserial
  • Debug Test: SERIALCONSOLE has no value and SERIALDEV is ttyS1
  • Debug Test: Added record_info

[waynechen55]

Welcome your review @alice-suse @guoxuguang @Julie-CAO @RoyCai7 @nanzhg @tbaev @martinsmarcelo Please help have a look if you are interested and leave a message.

[alice-suse]

If it has been ensured for my 2 comments, and all validation is ok, I think it will be good enough.

[waynechen55]

If it has been ensured for my 2 comments, and all validation is ok, I think it will be good enough.

I want to further clarify that this pull request is just to ensure serial console gets reset to default value ttyS1 instead of something totally different from SERIALDEV if it has no value. This pull request does not change any other things and does not intend to complicate things further either. All validation is OK. @alice-suse

[alice-suse]

If it has been ensured for my 2 comments, and all validation is ok, I think it will be good enough.

I want to further clarify that this pull request is just to ensure serial console gets reset to default value ttyS1 instead of something totally different from SERIALDEV if it has no value. This pull request does not change any other things and does not intend to complicate things further either. All validation is OK. @alice-suse

Thanks for the explanation. Now I understand your motivation better.

To me, the code logic has changed a little bit and updates the (best-practice or necessary) requirements when setting ipmi workers. Maybe others are not that deep into the code as you about the relevant parts. Would you point out explicitly what's the change/new requirement now to configure ipmi workers? Will it additionally require CORRECT SETTINGs for SERIALDEV and SERIALCONSOLE for any ipmi worker? Let's also involve other ipmi worker users for awareness, @czerw @rfan1 @Dawei-Pang @JoyceNa .

[waynechen55]

If it has been ensured for my 2 comments, and all validation is ok, I think it will be good enough.

I want to further clarify that this pull request is just to ensure serial console gets reset to default value ttyS1 instead of something totally different from SERIALDEV if it has no value. This pull request does not change any other things and does not intend to complicate things further either. All validation is OK. @alice-suse

Thanks for the explanation. Now I understand your motivation better.

To me, the code logic has changed a little bit and updates the (best-practice or necessary) requirements when setting ipmi workers. Maybe others are not that deep into the code as you about the relevant parts. Would you point out explicitly what's the change/new requirement now to configure ipmi workers? Will it additionally require CORRECT SETTINGs for SERIALDEV and SERIALCONSOLE for any ipmi worker? Let's also involve other ipmi worker users for awareness, @czerw @rfan1 @Dawei-Pang @JoyceNa .

Basically no requirements from others.

I think they can just bear in mind that SERIALCONSOLE is the most safe setting to be used with serial console, so they can start using SERIALCONSOLE instead of SERIALDEV in the future.

But this does not exclude use of SERIALDEV if someone would like to stick to this setting because this pull request does not change it.

I want to reiterate that this pull request is just to reset default value of SERIALCONSOLE instead of changing underlying logic. There is no logic change. And there is no change to SERIALDEV either.

[alice-suse]

LGTM

[czerw]

Commit message is missing important information, why we actually want such a change. What is broken?
Please add more details to commit message and also failing job examples to description.

[JoyceNa]

LGTM since no change to SERIALDEV

[waynechen55]

Commit message is missing important information, why we actually want such a change. What is broken? Please add more details to commit message and also failing job examples to description.

Already updated. @czerw

[waynechen55]

record_info already added. @czerw

[waynechen55]

I finished the last verification with added record_info.