Exception Handler

[ich_will]

What should I do if an exception occurs?
For Example the Div by Zero exception, at the moment my function prints out a message and holds, but I think thats a really bad solution.

[Tim]

Depends on the exception. Some exceptions (e.g. page fault) can be corrected and the app can continue. Others (e.g. divide by zero) indicate and error, and the app should exit.

[ich_will]

But I can't close the Kernel if there's an div by zero exception, or should i.

[Tim]

How can the kernel handle a fatal error in its own code? If the kernel has a fatal error (i.e. a bug), the only thing to do is print as much information about the exception as is available, and halt the system.

[Candy]

How can the kernel handle a fatal error in its own code?

For some cases you might be able to code something

For instance, when you call INVLPG on a 386 you get an invalid opcode exception. If you define the invlpg instruction to take an X number of bytes (possibly some more, maybe with a zero operand with it?) and you can runtime replace it with a different code sequence (the cr3 reload sequence) you could use the same code for both computers without slowing them down, by using (not even abusing) the invalid opcode mechanism.

[Pype.Clicker]

iirc, the Intel Manual (section about exceptions) gives hints on how exceptions should be handled. In the case of a kernel, as the same piece of code generated an exception and tries to handle it, things may sometimes be a little different.

For instance, you might try to have "resume points" where the execution should be transferred in case of an "expected" exception, but just entering a "panic mode" on unrecoverable errors (that is, errors that come from kernel itself rather than from some user-level call) would probably a more realistic target.