X-Git-Url: https://git.decadent.org.uk/gitweb/?p=nfs-utils.git;a=blobdiff_plain;f=debian%2FREADME.Debian.nfsv4;h=836a36d7bcb9d0ad688e9691d6d462f6c325047e;hp=9d3adc8a5148428448911ddade6cf20cabcec1bd;hb=1b0d7cf30d55019c4fcabb25c04ee61352e5666d;hpb=0a48a12183e728b39c37c65bc9a47da2933652f3

diff --git a/debian/README.Debian.nfsv4 b/debian/README.Debian.nfsv4
index 9d3adc8..836a36d 100644
--- a/debian/README.Debian.nfsv4
+++ b/debian/README.Debian.nfsv4
@@ -4,23 +4,29 @@ NFSv4 in Debian
 NFSv4 support in Debian is rather new, and not fully supported yet. If you want
 to experiment, make sure you have:
 
- - a recent 2.6 kernel on both client and server; newer is better. You might even
-   want to use CITI's patch set from http://www.citi.umich.edu/projects/nfsv4/linux/ .
+ - a recent 2.6 kernel on both client and server; newer is better. You might
+   even want to use CITI's patch set from
+   http://www.citi.umich.edu/projects/nfsv4/linux/ on the server, and/or Trond
+   Myklebust's patch set from http://client.linux-nfs.org/ .
  - a recent enough version of nfs-utils on both client and server (you probably
    have on at least one of them, since you're reading this file!).
- - a patched mount, which will hopefully enter the archive soon at the time of
-   writing -- otherwise, you'll have to enable the patch in the Debian package
-   yourself and rebuild it. (It is not enabled by default, since the current version
-   of the patch breaks mounting against NFSv2-only servers, such as nfs-user-server.)
+ - enabled idmapd on both sides (see /etc/default/nfs-common).
+ - The following lines in /etc/services on the client (if not, you will receive
+   the message "broken /etc/services" when starting rpc.gssd; this will usually
+   only happen if you upgrade netbase without letting it replace /etc/services
+   with the new version):
+
+   nfs		2049/tcp			# Network File System
+   nfs		2049/udp			# Network File System
 
 The export structure might be a bit confusing if you're already familiar with
-NFSv2 or NFSv3. The biggest difference is that you will need to export an explicit
-root of your pseudofilesystem, like this /etc/exports fragment:
+NFSv2 or NFSv3. The biggest difference is that you will need to export an
+explicit root of your pseudofilesystem, like this /etc/exports fragment:
 
   /nfs4                   hostname(rw,sync,fsid=0,crossmnt)
 
-(It doesn't need to be named "nfs4".) Then you can mount other volumes under that,
-like:
+(It doesn't need to be named "nfs4".) Then you can mount other volumes under
+that, like:
 
   /nfs4/music             hostname(rw,sync)
   /nfs4/movies            hostname(rw,sync)
@@ -29,23 +35,35 @@ Then your client can mount shares like this:
 
   mount -t nfs4 server:/music /mnt/music
 
-Since you might not have everything under one root, you might want /nfs4/* on the
-server to be bind mounts, ie.:
+Since you might not have everything under one root, you might want /nfs4/* on
+the server to be bind mounts, ie.:
 
   mount --bind /srv/music /nfs4/music
 
 or in /etc/fstab:
 
   /srv/music /nfs4/music none bind 0 0
-  
+
+Note that this special export structure might be handled transparently by
+rpc.mountd at some time in the future, in which case you will probably get the
+traditional (NFSv3-style) behaviour if and only if you have no share with
+fsid=0.
+
 If you do not wish to use host-based authentication, you can specify "gss/krb5"
 instead of a hostname to get Kerberos-based authentication instead. For this, 
 you will need an "nfs/hostname@REALM" entry in /etc/krb5.keytab, as well as
-rpc.gssd running on both the client and the server (enable it manually in
-/etc/default/nfs-common if the autodetection fails).
+rpc.gssd running on both client and rpc.svcgssd on the server (enable them
+manually in /etc/default/nfs-common and /etc/default/nfs-kernel-server if the
+autodetection fails).
 
 If you use "gss/krb5i", you will also get integrity (ie. authentication), and
-with "gss/krb5p", you'll also get privacy (ie.  encryption). Make sure your
+with "gss/krb5p", you'll also get privacy (ie. encryption). Make sure your
 kernel supports this; not all kernels do.
 
- -- Steinar H. Gunderson <sesse@debian.org>, Wed, 05 Apr 2006 18:09:47 +0200
+If you receive messages on the server complaining about "client ID already in
+use" when mounting from more than one client, check your /etc/hosts; if your
+hostname resolves to a non-global IP (like 127.0.0.1 or 127.0.1.1, or if you
+are behind NAT) this will cause such problems currently, and you will need to
+change or remove it for NFSv4 mounts to work correctly.
+
+ -- Steinar H. Gunderson <sesse@debian.org>, Wed, 11 Oct 2006 15:18:03 +0200