On Thu, 11 May 2000, Trond Myklebust wrote:
> Actually, NFS already has a private scheme precisely because of the
> page_symlink 'brokenness'.
> Are you serious about the uid indexing thing though? It'd be dead easy
> to implement, but do we really want something like this in generic NFS?
It might make sense as a mount option. Even if most symlinks are always
the same, the overhead for having a per-uid index into the page cache is
probably very small: most symlinks are also traversed by the same uid all
the time, so the duplication is probably minimal.
However, it was mostly an example of a "novel" approach. If people
actually like it, I have nothing against it, but it was not meant as a "I
think you should do this" kind of argument.
> Personally, I would have thought that a mount option to turn off
> caching of readdir/symlinks/... would be a better solution for those
> few cases that need it.
I'd prefer a per-uid cache to turning off the cache altogether, but I've
never actually used a NFS server that actually implements this feature so
I don't much care.
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to email@example.com
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Mon May 15 2000 - 21:00:18 EST