[input_server] Crash in atomic_set

[ronald-scheckelhoff-trac]

This is related to observations of hrev 49665 and 49666 installations.

The referenced debugger trap occurs only on the first boot attempt. Subseqently, everything is fine.

[ronald-scheckelhoff-trac]

Snapshot of screen

[ronald-scheckelhoff-trac]

Note: this was a first boot attempt from an ISO created CD, for the purpose of installing the OS. A second boot attempts works (both hrevs), and installs the system.

[ronald-scheckelhoff-trac]

Maybe this was done on purpose to test something?

[mmlr]

The success message is a part of the debugger startup, not the actual debug call reason. Or why do you think that this was left in place on purpose?

If this is reproducible, please get a stack trace of the thread using the bt or sc command. Unfortunately I don't think there's a way to print the actual debugger call message from the command line version of the debugger yet.

[ronald-scheckelhoff-trac]

Relative to the bt, I think the debugger was frozen. I seem to be having problems reproducing the issue, but it could be that this happens only if the hard drive has a previous installation (of another OS) with GRUB on it. I'll continue to attempt to reproduce the issue, and try to get a back trace.

[pulkomandy]

From the debugger command line you can use the save_report command to get a complete report, however if booting from a read-only CD this won't be of much use.

The crash is in input_server, so you may have trouble using the keyboard. This would explain the "freeze".

[diver]

(In #12416) Probably a dupe of #12412

[vidrep]

Just had it happen when trying to install hrev49736 x86_gcc2 from a CD using the anyboot image.

[vidrep]

Are there any steps that can be taken at the debugger prompt to extract any useful information?

[pulkomandy]

You can use save_report to save a report file to the desktop, but unfortunately this won't be very useful when booting from CD. As mmlr mentionned, typing bt or sc to get a backtrace, and taking a picture of that, would provide info as to where in the code the crash is happening.

[vidrep]

Yesterday, before your reply, I had run a couple of backtraces of some of the threads. I have attached a photo, which may or may not be of use in finding the issue. I tried to trigger the fault again this morning by trying about 25 installation boots, but nothing happened.

[mmlr]

This might be fixed in hrev49756 if the input_server was actually in the process of starting to watch for device entries. If it's still possible to reproduce it would be helpful to get a live serial log that would capture a back trace of the crash.

[vidrep]

A suggestion to change the title to be more specific: "Debugger entered on first post install boot"

[vidrep]

Had it happen with hrev49760 x86_gcc2. Unfortunately I was not serial logging at the time. I have attached the syslogs in the event they may be useful.

[diver]

(In #12713) Probably a dupe of #12412

[waddlesplash]

Anyone seen this in 6 years?