"Chris G. Demetriou" <cgd@pa.dec.com> writes:

> That's not particularly efficient ("process exists, it's time to
> context switch to the logger!").  It also has a much greater potential
> to lose accounting records.

I'm almost convinced.

Then, i'm writing a code to do disk capacity check inside of kernel
process rather than during clock interrupt.

It seems to be working, except one problem.  Write to accounting file
via nfs causes EACCESS if process is owned by other than root (if I
chaned the mode of the file to 0666, then write success :-).

Since kernel is passing a process's user credential when writing the
file, then, I'm now wondering why write successes in case of the file
is on the ffs...

enami.