Re: [RFC] atomic create+open

[Miklos Szeredi]

> As I believe I said earlier, open by inode number/filehandle/... don't
> exist in the NFSv4 protocol due to the potential for races.

I must have missed this.

Yes, open(O_CREAT) has race problems.  Plain open() doesn't.  So I
still don't see why you want to use the open-by-name for the
non-create case.

> No. There is no race for setattr() etc since they only do one lookup
> (and they don't set up any state on the server).
> 
> open() is the only case where we currently have to look things up twice
> (and I remind you that the second "lookup" is in fact the OPEN
> operation).

If NFS cannot do open by filehandle, then ->open_create() interface is
not enough obviously.

Miklos


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/