Piyush Shivam wrote:

> I took a look at the changes w.r.t. the filehandle and the memory leaks; 
> looks good.

Thanks!

Sam, were you looking at this stuff?  Should I wait for you?

> I guess I was simply curious about our assumptions/rules about accessing 
> operations across the nnode layer. My understanding so far has been that 
> anything in the NFS layer must always go via the nnodes for 
> data/metadata/state operations. This may be much stricter 
> interpretation, but it will be good to have a guideline regarding this. 
> The interesting thing about this case is that here we go from NFS to 
> nnodes to proxy I/O, and then again need to access vnode layer. The 
> question is that, in the context of NFS, once we have crossed the nnode 
> abstraction, do we again need to go through it for accessing the vnode 
> layer? Not a show stopper by any means as of now.

I think that since this is a provider, it's fine for it to use
the VOPs directly, and expect that to be simpler in most cases.

Rob T

Reply via email to