Almost. Kernel threads are not subject to pre-emptive scheduling, so you
`just' have to make sure that you keep your data consistent between
operations that call schedule() or sleep_on().
Using daemons for NFS readahead is an old and admittedly ugly concept;
if you have a better idea how to do, I'd sure want to know.
I'm working on redoing the nfsiod support at the moment so that only
one nfsiod is required rather than n.
> In addition, the locking of insmod's memory and command line look bad
> and are confusing unless you have read the README.
Agreed. That's on my todo list, but rather towards the bottom of it.
> Also, what about loading/unloading the nfs module. Is it really the
> only viable solution to keep the nfsiod's around and the nfs.o module
> loaded until said processes are killed?
No, you can use nfs without nfsiod. The performance will just be what it
used to be before adding readahead.
> It seems strongly against the concept of loadable modules to me,
> especially with kerneld.
In this case, probably kerneld should be fixed to run `killall -TERM nfsiod'
before unloading the module.
Olaf
--
Olaf Kirch | --- o --- Nous sommes du soleil we love when we play
okir@monad.swb.de | / | \ sol.dhoop.naytheet.ah kin.ir.samse.qurax
For my PGP public key, finger okir@brewhq.swb.de.