Re: [linuxdc-dev] Trapping exceptions in the kernel

[David W. <dav...@ia...>]

> Is there a way of doing this? I want to rewrite the vmufs driver to make it as 
> independent of the underlying medium as possible and so I wanto to probe the 
> size of the vmu.
> 
> An obvious way is to attempt to read in vmu sectors until this fails - eg at 
> 128, at 256, at 512, at 1024. (Working on the suggested likely sizes for 
> alternative VMUs, assuming these things just might exist somewhere).
> 
> On a standard VMU 128 would succeed, 256 would fail and a simple check would 
> then reveal that the superblock was in sector 255.
> 
> But a buffer read at 256 will oops the kernel in these circumstances - is 
> there a mechanism to trap that exception?

The solution is to make the kernel not oops.  Why would it oops if the VMU says
a sector doesn't exist?  Does the kernel oops if a scsi disk has a bad block?

Change the code that now oops's and replace it with a return of an error value.

Is that practical?

Cheers,
David