> Some weeks ago I posted a message about a programm on solaris which
> fails to work on a nfs drive mounted from a linux bo via the kernel nfs.
>
> Now I can describe the possible error a bit better:
> The programm does a fcntl() with F_SETLK, but I cannot (yet) say what is
> in the struct flock in the third argument. I hope this information is
> sufficient because I only have the binary of the programm and got that
> information from the truss output and have no idea how to read the
> contents of the parameter...
> Is this a bug, should it work (now or in future) or is it not intended
> to work with the kernel nfs at all.
Hi Bernhard,
I'm interested in fixing any locking-related problems in knfsd, but will
need a more detailed description of (a) what should happen, and (b) what
currently happens. What would probably be most helpful would be to run
some file locking test suites (with source available so we don't have to
guess at what's happening), and then report any problems with knfsd
locking. As a starting point Olaf's utilities have a few locking-related
tests, but there may be more extensive test suites available.
One thing I found out recently is that you apparently need to run the
statd utility to use knfsd locking at all.
I don't know very much about file locking so someone else will have to
take the lead in doing the testing. But I'll try to fix any bugs you
find ...
Regards,
Bill