Did anyone ever consider implementing exception handling for Linux? It's
pretty annoying to restore a 40M file from floppies only to find out that the
last one has an unreliable sector and you have to start all over again. It
would be nice if one could at least retry the read.
Possible method:
* On hard device errors, kernel puts a process to sleep and writes data to
some device, like /dev/criterr (but if there is no process that has device
open, it just returns EIO as before).
There is no problem with critical error queue overflowing, because each
process can have at most one critical error at a time.
* A user level process reads /dev/criterr and pops up an X window (or switches
to another vt if no X is running).
* When user decided what to do, the process writes to criterr and kernel uses
data to kill the process, return EIO, retry, etc.
Is it useful?
--
Snow ^oo^ Cat <snowcat@gd.cs.CSUFresno.EDU>
_ -> <- aka Oleg Kibirev <oleg@gd.cs.CSUFresno.EDU>
___(_) _ _)_
/ _)
\_.-._
|___/ Purr!