Re: [newnfs/client] -alldirs: listing files consumes too much memory

On Thu, Jul 16, 2009 at 08:12:38PM -0400, Rick Macklem wrote:
> 
> 
> On Thu, 16 Jul 2009, Anonymous wrote:
> 
> >Let's create 335 empty files in /blah and try to list them over nfsv3.
> >
> ># uname -vm
> >FreeBSD 8.0-BETA1 #0: Sat Jul  4 03:55:14 UTC 2009     
> >root_at_almeida.cse.buffalo.edu:/usr/obj/usr/src/sys/GENERIC  i386
> >
> ># mkdir /blah
> ># (while [ $((i+=1)) -le 334 ]; do : >/blah/foo_$i; done)
> ># echo / -alldirs >/etc/exports
> ># /etc/rc.d/nfsd onestart
> ># mount -t newnfs -o nfsv3 0:/blah /mnt
> >
> Well, this turns out more interesting than I expected. The problem occurs
> when there is a large directory "at the mount point only". If you:
> # mount -t newnfs -o nfsv3 0:/ /mnt
> # cd /mnt/blah
> # ls
> - it works.
> 
> When the large directory is at the mount point, it reads the first
> block normally but... it then thinks all subsequent blocks are already
> in the buffer cache. ie. They come back from getblk() with B_CACHE
> already set??? (It then just loops getting blocks forever, since it
> won't see the eof if it doesn't try and read from the server.)
> 
> Anyone happen to have a clue why that would happen? Why would blocks
> on a mount point vnode behave differently than others.
> 
> Well, at least it's easy to reproduce, so I can keep poking around
> with it, rick.

Without looking too hard into it, could it be due to read-ahead ?