Hi,
i'm testing the 3.6.11 kernel on some small internal nfs network (1 server <-> 2 clients)
and observing such info in dmesg on the client side. could you put some light on this dmesg entry?
is it a bug or misconfiguration? i can provide more information about nfs configuration
if necessary but denying caching doesn't look nicely.
i'm using following nfs related packages:
autofs-5.0.5-1.x86_64
libnfsidmap-0.25-4.x86_64
nfs-utils-1.2.7-1.x86_64
nfs-utils-clients-1.2.7-1.x86_64
nfs-utils-common-1.2.7-1.x86_64
BR,
Paweł.
please CC me on reply.
Paweł Sikora wrote:
Hi,
i'm testing the 3.6.11 kernel on some small internal nfs network (1 server <-> 2 clients)
and observing such info in dmesg on the client side. could you put some
light on this dmesg entry?
Are you mounting the same file system more than once? Why? I thought I
remembered a mount option to uniquify the mounts but I can't find it now.
On Wednesday 19 of December 2012 19:35:50 Myklebust, Trond wrote:
> On Wed, 2012-12-19 at 19:58 +-0100, Pawe+AUI- Sikora wrote:
> +AD4- Hi,
> +AD4-
> +AD4- i'm testing the 3.6.11 kernel on some small internal nfs network (1 server +ADw--+AD4- 2 clients)
> +AD4- and observing such info in dmesg on the client side. could you put some light on this dmesg entry?
> +AD4- is it a bug or misconfiguration? i can provide more information about nfs configuration
> +AD4- if necessary but denying caching doesn't look nicely.
> +AD4-
> +AD4- i'm using following nfs related packages:
> +AD4-
> +AD4- autofs-5.0.5-1.x86+AF8-64
> +AD4- libnfsidmap-0.25-4.x86+AF8-64
> +AD4- nfs-utils-1.2.7-1.x86+AF8-64
> +AD4- nfs-utils-clients-1.2.7-1.x86+AF8-64
> +AD4- nfs-utils-common-1.2.7-1.x86+AF8-64
> +AD4-
> +AD4- BR,
> +AD4- Pawe+AUI-.
>
> It's hard to guess when you don't tell us anything about what you are
> doing to trigger this message, but it sounds as if you are trying to use
> the cachefs mount options incorrectly.
on the server side there're following exports:
# exportfs -v
/home/atest/farm-local
<world>(rw,async,wdelay,nohide,crossmnt,no_root_squash,no_subtree_check,insecure_locks)
/home/atest/farm
<world>(rw,async,wdelay,nohide,crossmnt,no_root_squash,no_subtree_check,insecure_locks)
/home/users <world>(rw,no_root_squash,no_subtree_check,insecure_locks)
in the exported /home/atest/* there're crossmounted/binded directories:
# fstab
/home/atest/farm-local/results /home/atest/farm/results none bind
/home/services/ftp /var/lib/ftp none bind
/home/services/ftp /home/atest/farm-local/ftp none bind
# df -h
Filesystem Size Used Avail Use% Mounted on
rootfs 7.6G 4.5G 2.7G 63% /
/dev/md0 7.6G 4.5G 2.7G 63% /
run 32G 264K 32G 1% /run
devtmpfs 32G 0 32G 0% /dev
/dev/md1 7.2T 794G 6.4T 11% /home
none 32G 0 32G 0% /dev/shm
/dev/md1 7.2T 794G 6.4T 11% /home/atest/farm/results
/dev/md1 7.2T 794G 6.4T 11% /var/lib/ftp
/dev/md1 7.2T 794G 6.4T 11% /home/atest/farm-local/ftp
on the client side there's only a simple autofs configuration:
# cat /etc/autofs/auto.server
farm -rw,soft,rsize=32768,wsize=32768,intr,vers=4,async,bg,noatime,retry=30,retrans=10 192.168.2.100:/home/atest/farm-local
home -rw,soft,vers=4 192.168.2.100:/home/users
Myklebust, Trond <[email protected]> wrote:
> From e8ed657ba5edc8c9339e00ab2fd235ab1dd94cd5 Mon Sep 17 00:00:00 2001
> From: Trond Myklebust <[email protected]>
> Date: Wed, 19 Dec 2012 16:56:28 -0500
> Subject: [PATCH] NFS: Kill fscache warnings when mounting without -ofsc
> MIME-Version: 1.0
> Content-Type: text/plain; charset=UTF-8
> Content-Transfer-Encoding: 8bit
>
> The fscache code will currently bleat a "non-unique superblock keys"
> warning even if the user is mounting without the 'fsc' option.
> There should be no reason to even initialise the superblock cache cookie
> unless we're planning on using fscache for something, so ensure that
> we check for the NFS_OPTION_FSCACHE flag before calling into the
> fscache code.
>
> Reported-by: Paweł Sikora <[email protected]>
> Signed-off-by: Trond Myklebust <[email protected]>
> Cc: David Howells <[email protected]>
Acked-by: David Howells <[email protected]>
On Wed, 2012-12-19 at 22:16 +-0100, Pawe+AUI- Sikora wrote:
+AD4- On Wednesday 19 of December 2012 19:35:50 Myklebust, Trond wrote:
+AD4- +AD4- On Wed, 2012-12-19 at 19:58 +--0100, Pawe+-AUI- Sikora wrote:
+AD4- +AD4- +-AD4- Hi,
+AD4- +AD4- +-AD4-
+AD4- +AD4- +-AD4- i'm testing the 3.6.11 kernel on some small internal nfs network (1 server +-ADw--+-AD4- 2 clients)
+AD4- +AD4- +-AD4- and observing such info in dmesg on the client side. could you put some light on this dmesg entry?
+AD4- +AD4- +-AD4- is it a bug or misconfiguration? i can provide more information about nfs configuration
+AD4- +AD4- +-AD4- if necessary but denying caching doesn't look nicely.
+AD4- +AD4- +-AD4-
+AD4- +AD4- +-AD4- i'm using following nfs related packages:
+AD4- +AD4- +-AD4-
+AD4- +AD4- +-AD4- autofs-5.0.5-1.x86+-AF8-64
+AD4- +AD4- +-AD4- libnfsidmap-0.25-4.x86+-AF8-64
+AD4- +AD4- +-AD4- nfs-utils-1.2.7-1.x86+-AF8-64
+AD4- +AD4- +-AD4- nfs-utils-clients-1.2.7-1.x86+-AF8-64
+AD4- +AD4- +-AD4- nfs-utils-common-1.2.7-1.x86+-AF8-64
+AD4- +AD4- +-AD4-
+AD4- +AD4- +-AD4- BR,
+AD4- +AD4- +-AD4- Pawe+-AUI-.
+AD4- +AD4-
+AD4- +AD4- It's hard to guess when you don't tell us anything about what you are
+AD4- +AD4- doing to trigger this message, but it sounds as if you are trying to use
+AD4- +AD4- the cachefs mount options incorrectly.
+AD4-
+AD4- on the server side there're following exports:
+AD4-
+AD4- +ACM- exportfs -v
+AD4- /home/atest/farm-local
+AD4- +ADw-world+AD4-(rw,async,wdelay,nohide,crossmnt,no+AF8-root+AF8-squash,no+AF8-subtree+AF8-check,insecure+AF8-locks)
+AD4- /home/atest/farm
+AD4- +ADw-world+AD4-(rw,async,wdelay,nohide,crossmnt,no+AF8-root+AF8-squash,no+AF8-subtree+AF8-check,insecure+AF8-locks)
+AD4- /home/users +ADw-world+AD4-(rw,no+AF8-root+AF8-squash,no+AF8-subtree+AF8-check,insecure+AF8-locks)
+AD4-
+AD4- in the exported /home/atest/+ACo- there're crossmounted/binded directories:
+AD4-
+AD4- +ACM- fstab
+AD4- /home/atest/farm-local/results /home/atest/farm/results none bind
+AD4- /home/services/ftp /var/lib/ftp none bind
+AD4- /home/services/ftp /home/atest/farm-local/ftp none bind
+AD4-
+AD4- +ACM- df -h
+AD4- Filesystem Size Used Avail Use+ACU- Mounted on
+AD4- rootfs 7.6G 4.5G 2.7G 63+ACU- /
+AD4- /dev/md0 7.6G 4.5G 2.7G 63+ACU- /
+AD4- run 32G 264K 32G 1+ACU- /run
+AD4- devtmpfs 32G 0 32G 0+ACU- /dev
+AD4- /dev/md1 7.2T 794G 6.4T 11+ACU- /home
+AD4- none 32G 0 32G 0+ACU- /dev/shm
+AD4- /dev/md1 7.2T 794G 6.4T 11+ACU- /home/atest/farm/results
+AD4- /dev/md1 7.2T 794G 6.4T 11+ACU- /var/lib/ftp
+AD4- /dev/md1 7.2T 794G 6.4T 11+ACU- /home/atest/farm-local/ftp
+AD4-
+AD4-
+AD4- on the client side there's only a simple autofs configuration:
+AD4-
+AD4- +ACM- cat /etc/autofs/auto.server
+AD4- farm -rw,soft,rsize+AD0-32768,wsize+AD0-32768,intr,vers+AD0-4,async,bg,noatime,retry+AD0-30,retrans+AD0-10 192.168.2.100:/home/atest/farm-local
+AD4- home -rw,soft,vers+AD0-4 192.168.2.100:/home/users
+AD4-
OK. If you are not specifying '-ofsc', then that warning is 100+ACU-
confusing. How about the following patch?
8+ADw----------------------------------------------------------------
>From e8ed657ba5edc8c9339e00ab2fd235ab1dd94cd5 Mon Sep 17 00:00:00 2001
From: Trond Myklebust +ADw-Trond.Myklebust+AEA-netapp.com+AD4-
Date: Wed, 19 Dec 2012 16:56:28 -0500
Subject: +AFs-PATCH+AF0- NFS: Kill fscache warnings when mounting without -ofsc
MIME-Version: 1.0
Content-Type: text/plain+ADs- charset+AD0-UTF-8
Content-Transfer-Encoding: 8bit
The fscache code will currently bleat a +ACI-non-unique superblock keys+ACI-
warning even if the user is mounting without the 'fsc' option.
There should be no reason to even initialise the superblock cache cookie
unless we're planning on using fscache for something, so ensure that
we check for the NFS+AF8-OPTION+AF8-FSCACHE flag before calling into the
fscache code.
Reported-by: Pawe+AUI- Sikora +ADw-pawel.sikora+AEA-agmk.net+AD4-
Signed-off-by: Trond Myklebust +ADw-Trond.Myklebust+AEA-netapp.com+AD4-
Cc: David Howells +ADw-dhowells+AEA-redhat.com+AD4-
---
fs/nfs/super.c +AHw- 19 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-----
1 file changed, 15 insertions(+-), 4 deletions(-)
diff --git a/fs/nfs/super.c b/fs/nfs/super.c
index aa5315b..c25cadf8 100644
--- a/fs/nfs/super.c
+-+-+- b/fs/nfs/super.c
+AEAAQA- -2375,19 +-2375,30 +AEAAQA- static void nfs+AF8-get+AF8-cache+AF8-cookie(struct super+AF8-block +ACo-sb,
struct nfs+AF8-parsed+AF8-mount+AF8-data +ACo-parsed,
struct nfs+AF8-clone+AF8-mount +ACo-cloned)
+AHs-
+- struct nfs+AF8-server +ACo-nfss +AD0- NFS+AF8-SB(sb)+ADs-
char +ACo-uniq +AD0- NULL+ADs-
int ulen +AD0- 0+ADs-
- if (parsed +ACYAJg- parsed-+AD4-fscache+AF8-uniq) +AHs-
- uniq +AD0- parsed-+AD4-fscache+AF8-uniq+ADs-
- ulen +AD0- strlen(parsed-+AD4-fscache+AF8-uniq)+ADs-
+- nfss-+AD4-fscache+AF8-key +AD0- NULL+ADs-
+- nfss-+AD4-fscache +AD0- NULL+ADs-
+-
+- if (parsed) +AHs-
+- if (+ACE-(parsed-+AD4-options +ACY- NFS+AF8-OPTION+AF8-FSCACHE))
+- return+ADs-
+- if (parsed-+AD4-fscache+AF8-uniq) +AHs-
+- uniq +AD0- parsed-+AD4-fscache+AF8-uniq+ADs-
+- ulen +AD0- strlen(parsed-+AD4-fscache+AF8-uniq)+ADs-
+- +AH0-
+AH0- else if (cloned) +AHs-
struct nfs+AF8-server +ACo-mnt+AF8-s +AD0- NFS+AF8-SB(cloned-+AD4-sb)+ADs-
+- if (+ACE-(mnt+AF8-s-+AD4-options +ACY- NFS+AF8-OPTION+AF8-FSCACHE))
+- return+ADs-
if (mnt+AF8-s-+AD4-fscache+AF8-key) +AHs-
uniq +AD0- mnt+AF8-s-+AD4-fscache+AF8-key-+AD4-key.uniquifier+ADs-
ulen +AD0- mnt+AF8-s-+AD4-fscache+AF8-key-+AD4-key.uniq+AF8-len+ADs-
+AH0AOw-
- +AH0-
+- +AH0- else
+- return+ADs-
nfs+AF8-fscache+AF8-get+AF8-super+AF8-cookie(sb, uniq, ulen)+ADs-
+AH0-
--
1.7.11.7
--
Trond Myklebust
Linux NFS client maintainer
NetApp
Trond.Myklebust+AEA-netapp.com
http://www.netapp.com
On Wed, 2012-12-19 at 19:58 +-0100, Pawe+AUI- Sikora wrote:
+AD4- Hi,
+AD4-
+AD4- i'm testing the 3.6.11 kernel on some small internal nfs network (1 server +ADw--+AD4- 2 clients)
+AD4- and observing such info in dmesg on the client side. could you put some light on this dmesg entry?
+AD4- is it a bug or misconfiguration? i can provide more information about nfs configuration
+AD4- if necessary but denying caching doesn't look nicely.
+AD4-
+AD4- i'm using following nfs related packages:
+AD4-
+AD4- autofs-5.0.5-1.x86+AF8-64
+AD4- libnfsidmap-0.25-4.x86+AF8-64
+AD4- nfs-utils-1.2.7-1.x86+AF8-64
+AD4- nfs-utils-clients-1.2.7-1.x86+AF8-64
+AD4- nfs-utils-common-1.2.7-1.x86+AF8-64
+AD4-
+AD4- BR,
+AD4- Pawe+AUI-.
It's hard to guess when you don't tell us anything about what you are
doing to trigger this message, but it sounds as if you are trying to use
the cachefs mount options incorrectly.
--
Trond Myklebust
Linux NFS client maintainer
NetApp
Trond.Myklebust+AEA-netapp.com
http://www.netapp.com