OSDev.org

If this idea is out already, tell me. It probably is.

During a crash, is it possible for an OS to save the state of the system (everything) and restart, but when it boots, analyze the saved state and then try to fix the problem? Or at least inform the user with readable data?

Let me know your thoughts.....

JL

Yeah I guess that would be possible. It really depends on the severity of the error and the kernel type really. For example in a microkernel if an error occurs in a module you can just restart the module (I think they are called modules.) I would hope that all of the critical parts of the kernel would be directly restartable. Also if you have a BSOD you could store all of the information in computer readable format and then offer a more user readable format the next time the computer reboots, possibly consulting an online database in the process.

And how exactly are you going to save the state on the disk? Without that fluffy disk/filesystem/vfs driver that crashed 2 nanoseconds ago?

Well, if it's possible....

and it could try to restart all the drivers...

bluecode wrote:And how exactly are you going to save the state on the disk? Without that fluffy disk/filesystem/vfs driver that crashed 2 nanoseconds ago?

You could save to data to a location in memory and hope that on a restart the information will still be there. Or maybe have a reserved location on the disk assuming the disk driver didn't crash.

Actually, most computers keep power to RAM... OpenBSD for example, willl use the consistency of RAM to store a last-boot type dmesg. (iirc)

But yea... Ram is kept alive until you unplug your PC really, though I wouldn't rely on it for anything mission-critical or anything, but for somethign like you are talking, it should suffice...