imho, this SHOULD be done in userland. there isn't anything stopping
userland from telling you "NOW" that something is wrong. just about every
other OS i can think of returns 'error reading xxx' instead of blowing up
the computer.
a corrupted FS is not a reason to crash. it is a reason to halt further
interaction with the particular FS until the userland fixes it.
the kernel can continue running quite happily while returning EBROKE to
userland programs when they try to access a damaged FS.
even should you do something as take a chainsaw to your drive while it is
running, a fault tolerant system should be able to automatically block
that drive out while maintaining the rest of the system.
it is my belief that there is nothing short of mass destruction of the
hardware that should stop the system from running. if a particular node
in the system falls offline due to any reason, the rest of the system
should be unnaffected except for accesses to that node.
is there any reason why this approach is bad?
-d
-- to include colors and other non printing characters in a bash prompt, surround the non printing characters with \[ and \] and bash will correctly calculate the line length.
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu