Date: Thu, 3 Mar 2011 05:00:01 -0800 (PST) From: Kirill Yelizarov <ykirill@yahoo.com> To: freebsd-stable@freebsd.org Subject: Re: NFS client over udp Message-ID: <650248.21751.qm@web120514.mail.ne1.yahoo.com> In-Reply-To: <1920317329.229957.1298414128136.JavaMail.root@erie.cs.uoguelph.ca>
next in thread | previous in thread | raw e-mail | index | archive | help
--- On Wed, 2/23/11, Rick Macklem <rmacklem@uoguelph.ca> wrote:
> From: Rick Macklem <rmacklem@uoguelph.ca>
> Subject: Re: NFS client over udp
> To: "Kirill Yelizarov" <ykirill@yahoo.com>
> Cc: freebsd-stable@freebsd.org
> Date: Wednesday, February 23, 2011, 1:35 AM
> > --- On Tue, 2/22/11, Rick
> Macklem <rmacklem@uoguelph.ca>
> wrote:
> >
> > > From: Rick Macklem <rmacklem@uoguelph.ca>
> > > Subject: Re: NFS client over udp
> > > To: "Kirill Yelizarov" <ykirill@yahoo.com>
> > > Cc: freebsd-stable@freebsd.org
> > > Date: Tuesday, February 22, 2011, 2:10 AM
> > > > --- On Sun, 2/20/11, Rick
> > > Macklem <rmacklem@uoguelph.ca>
> > > wrote:
> > > >
> > > > > From: Rick Macklem <rmacklem@uoguelph.ca>
> > > > > Subject: Re: NFS client over udp
> > > > > To: "Kirill Yelizarov" <ykirill@yahoo.com>
> > > > > Cc: freebsd-stable@freebsd.org
> > > > > Date: Sunday, February 20, 2011, 9:02
> PM
> > > > > > --- On Fri, 2/18/11, Kirill
> > > > > Yelizarov
> > > > > > > > On Fri, Feb 18, 2011 at
> > > 05:27:00AM
> > > > > > > > -0800, Kirill Yelizarov
> wrote:
> > > > > > > > > I have a
> reproducible memory
> > > leak when
> > > > > using nfs
> > > > > > > > client with an old
> > > > > > > > > nfs server
> > > > > >
> > > > > > and mbufs used
> > > > > > 8193/1722/9915 mbufs in use
> > > (current/cache/total)
> > > > > > 8192/1264/9456/25600 mbuf clusters
> in use
> > > > > (current/cache/total/max)
> > > > > > 8192/605 mbuf+clusters out of
> packet
> > > secondary zone in
> > > > > use
> > > > > > (current/cache)
> > > > > > 0/768/768/12800 4k (page size)
> jumbo
> > > clusters in use
> > > > > > (current/cache/total/max)
> > > > > > 0/0/0/6400 9k jumbo clusters in
> use
> > > > > (current/cache/total/max)
> > > > > > 0/0/0/3200 16k jumbo clusters in
> use
> > > > > (current/cache/total/max)
> > > > > > 18432K/6030K/24462K bytes
> allocated to
> > > network
> > > > > (current/cache/total)
> > > > > > 0/0/0 requests for mbufs denied
> > > > > (mbufs/clusters/mbuf+clusters)
> > > > > > 0/0/0 requests for jumbo clusters
> denied
> > > (4k/9k/16k)
> > > > > > 0/0/0 sfbufs in use
> (current/peak/max)
> > > > > > 0 requests for sfbufs denied
> > > > > > 0 requests for sfbufs delayed
> > > > > > 0 requests for I/O initiated by
> sendfile
> > > > > > 0 calls to protocol drain
> routines
> > > > > >
> > > > > > Kirill
> > > > > >
> > > > > You could try the attached patch. It
> fixes the
> > > only places
> > > > > in the
> > > > > client side krpc over udp that seems
> mights cause
> > > a leak. I
> > > > > have no
> > > > > idea if it will help, since these cases
> should
> > > rarely, if
> > > > > ever,
> > > > > happen in practice.
> > > > >
> > > > > Please let us know if you have the
> chance to try
> > > the patch
> > > > > and
> > > > > whether or not it helped.
> > > > >
> > > > > rick
> > > > >
> > > > Rick, i tried your patch. Fortunately it
> didn't help
> > > me. There are no
> > > > warnings on console and memory is climbing
> up during
> > > syncs and not
> > > > freed later. I'll try to switch to tcp this
> evening.
> > > Thanks for help
> > > >
> > > I'll assume that's unfortunately;-) Since the two
> cases
> > > patched probably
> > > never happen, I'm not surprised.
> > >
> > > The only other thing I can think of that you
> could try is
> > > switching to
> > > the experimental client. This would identify if
> the bug is
> > > in the regular
> > > client or somewhere further down in the rpc
> transport.
> > >
> > > The mount command would look something like:
> > > # mount -t newnfs -o nfsv3,udp
> <server>:<path>
> > > <mntpath>
> > >
> > I added options NFSCL to my kernel and tried to mount.
> mount shows
> > everything is ok:
> > 192.168.0.35:/home on /mnt (newnfs)
> > but when i try to cd /mnt i get permission denied
> > my export allow root and everything is done as root.
> What am i doing
> > wrong?
> >
> Try adding the "resvport" option. I don't think it's a
> default for
> the experimental client at this time. (I have a series of
> patches
> for the client that will go into head in April that brings
> it in
> line with the regular client, including same default mount
> options.)
Rick,
I have good news. I upgraded to 8.2-stable and i ran all four different tests (nfs client new and old and over udp and tcp) and found that there is no leak in either. ALl of them behave almost the same, i couldn't find any difference. The speed i achieved on 1Gb link is 52Mb/s. The only difference is that i can't umount new nfs client even if there are no processes using this mount point. Thanks for help
Kirill
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?650248.21751.qm>