Re: Propagating errors in proc

[Igor Khavkine]

On Tue, Aug 21, 2001 at 10:48:15AM -0700, Thomas Bushnell, BSG wrote:
> Igor Khavkine <i_khavki@alcor.concordia.ca> writes:
> 
> > I'm looking through proc trying to fix malloc() bugs. There was one
> > place where malloc was used inside a void function. I had to change
> > it's return value to error_t so an error could be reported.
> 
> Some of these are catastrophic failures, and all we can do is give
> up.  new_proc is one of those, I think.

After some fruitful discussion with Neal, I changed some more code
around and now all such errors are propagated up to the RPC handlers
(the S_* routines). The changes can be seen if malloc-round2.diff
that I just posted.

And I do not believe that lack of resources is a reason to give up
and panic. First because resource shortages are often temporary and
all one has to do is wait a bit and try again to succed. And if the
system is really non-functional because of resource shortage, there should
always be a way to tell the user about it, and then let the user decide
whether a shutdown is in order. Oh yeah, and hopefully the shutdown
procedure would use a fixed amount of resources so as not to fail at that
point.

Igor