Project

General

Profile

Actions

Bug #2644

open

file corruption with nfs sharing

Added by Richard PALO about 10 years ago. Updated over 9 years ago.

Status:
In Progress
Priority:
High
Assignee:
-
Category:
-
Start date:
2012-04-20
Due date:
% Done:

0%

Estimated time:
Difficulty:
Medium
Tags:
Gerrit CR:

Description

followup from https://www.illumos.org/issues/2347

Have an openoffice .ods calc file on the nfs/cifs share on an OI_151a3 server
from a client running the same version, open the file (OOO3.3), made some modifications and closed the file. No erreurs.

But the following is found locally in /var/adm/messages

Apr 20 09:29:03 x3200 nfs: [ID 897781 kern.notice] NFS write error on host smicro: Permission denied.
Apr 20 09:29:03 x3200 nfs: [ID 702911 kern.notice] (file handle: 58b0f1aa 50a83908 27c2000a 0 2a1708 4000a 0 5ee45 0)
Apr 20 09:29:03 x3200 nfs: [ID 897781 kern.notice] NFS write error on host smicro: Permission denied.
Apr 20 09:29:03 x3200 nfs: [ID 702911 kern.notice] (file handle: 58b0f1aa 50a83908 27c2000a 0 2a1708 4000a 0 5ee45 0)
Apr 20 09:29:03 x3200 nfs: [ID 897781 kern.notice] NFS write error on host smicro: Permission denied.
Apr 20 09:29:03 x3200 nfs: [ID 702911 kern.notice] (file handle: 58b0f1aa 50a83908 27c2000a 0 2a1708 4000a 0 5ee45 0)
Apr 20 09:29:03 x3200 nfs: [ID 897781 kern.notice] NFS write error on host smicro: Permission denied.
Apr 20 09:29:03 x3200 nfs: [ID 702911 kern.notice] (file handle: 58b0f1aa 50a83908 27c2000a 0 2a1708 4000a 0 5ee45 0)

looked at the file, garbage now...

the server and the workstation are static ip addresses, both are listed in /etc/hosts
nfs share looks like zfs/rpool/export/dossiers nfs=() smb=() nfs:sys=(rw="@192.168.0.0/24:@192.168.1.0/24")
/export/dossiers
Dossiers=/export/dossiers smb=(guestok="true")
from sharemgr

starting to have vnc problems again too, will post another bug report

Actions #1

Updated by Richard PALO about 10 years ago

file protection seems ok with ACL (/usr/bin/ls -av):

rwxrwxrwx+  1 nobody   nobody 104328 avr. 20 09:29  /export/dossiers/</dir/filename>
     0:everyone@:read_data/write_data/append_data/read_xattr/write_xattr
         /execute/delete_child/read_attributes/write_attributes/delete
         /read_acl/write_acl/write_owner/synchronize:file_inherit
         /dir_inherit:allow

Actions #2

Updated by Richard PALO about 10 years ago

given that the CIFS deployment dates since november 2010 and has given no problems whatsoever until now (with the suggestion to turn on NFS to debug a file-roller problem), it is either related to NFS, 151a3 or both.
I guess there's no choice but to back out given the absence of any feedback.

Actions #3

Updated by Richard PALO about 10 years ago

I'm not really sure why, but after looking at
https://wiki-bsse.ethz.ch/display/ITDOC/Managing+CIFS+on+Solaris
and
http://mail.opensolaris.org/pipermail/cifs-discuss/2009-December/002586.html

is it possible that the absence of setting aclmode=passthrough and aclinherit=passthrough could be behind the problem?
I just tried creating a file from the client using touch, and I get the following on the server:

# /usr/bin/ls -lapdV /export/dossiers/Documents/foo.bar 
-rw-r--r--   1 richard  staff          0 avr. 23 13:06 /export/dossiers/Documents/foo.bar
                 owner@:rw-p--aARWcCos:-------:allow
                 group@:r-----a-R-c--s:-------:allow
              everyone@:r-----a-R-c--s:-------:allow

the directory has
~# /usr/bin/ls -lapdV /export/dossiers/Documents/
drwxrwxrwx+ 41 root     root          54 avr. 23 13:06 /export/dossiers/Documents//
              everyone@:rwxpdDaARWcCos:fd-----:allow

extract from zfs get all

rpool/export/dossiers  aclmode                         discard                                            default
rpool/export/dossiers  aclinherit                      restricted                                         default

that is, after setting these, touching a file creates (as I would have already expected)

# /usr/bin/ls -lapdV /export/dossiers/Documents/bar.foo 
-rwxrwxrwx+  1 richard  staff          0 avr. 23 13:35 /export/dossiers/Documents/bar.foo
              everyone@:rwxpdDaARWcCos:------I:allow

In summary, since the file was existing, is it possible that the corruption was due to this ACL ?

Thanks in advance

Actions #4

Updated by Richard PALO about 10 years ago

still fighting this one, seems more and more a combination of nfs and ACLs
(for memory, ACL=> /usr/bin/chmod -R A=everyone@:full_set:fd:allow rpool/export/dossiers)

# zfs get aclmode,aclinherit,sharenfs rpool/export/dossiers
NAME                   PROPERTY    VALUE                                              SOURCE
rpool/export/dossiers  aclmode     passthrough                                        local
rpool/export/dossiers  aclinherit  passthrough                                        local
rpool/export/dossiers  sharenfs    rw=@192.168.0.0/24:@192.168.1.0/24 local
# ls -aV .zfs/shares/dossiers 
-rwxrwxrwx+  1 root     root           0 mars 26 11:54 .zfs/shares/dossiers
              everyone@:rwxpdDaARWcCos:-------:allow
# ls -aV
total 91
drwxrwxrwx+  5 root     root           5 avr. 27 07:14 .
              everyone@:rwxpdDaARWcCos:fd-----:allow
drwxr-xr-x   5 root     root           5 oct. 27  2011 ..
                 owner@:--------------:-------:deny
                 owner@:rwxp---A-W-Co-:-------:allow
                 group@:-w-p----------:-------:deny
                 group@:r-x-----------:-------:allow
              everyone@:-w-p---A-W-Co-:-------:deny
              everyone@:r-x---a-R-c--s:-------:allow
drwxrwxrwx+  2 root     sys            3 mars 26 11:54 .$EXTEND
              everyone@:rwxpdDaARWcCos:fd-----:allow
drwxrwxrwx+ 41 root     root          56 avr. 27 13:06 Documents
              everyone@:rwxpdDaARWcCos:fd-----:allow
drwxrwxrwx+ 35 root     root         155 avr. 27 08:02 test
              everyone@:rwxpdDaARWcCos:fd-----:allow

I tried the following experiment touch local and over nfs, and save_as from scalc over both:
(note over NFS, the differing ACL from save or save as)

$ ls -aV 
-rwxrwxrwx+  1 nobody   nobody    260051 avr. 27 11:12 <calc file>.ods*
              everyone@:rwxpdDaARWcCos:fd-----:allow
-rwxrwxrwx+  1 richard  staff          0 avr. 27 15:10 nfs-foo*
              everyone@:rwxpdDaARWcCos:-------:allow
-rw-r--r--   1 richard  staff     272815 avr. 27 15:11 nfs-<calc file>.ods
                 owner@:rw-p--aARWcCos:-------:allow
                 group@:r-----a-R-c--s:-------:allow
              everyone@:r-----a-R-c--s:-------:allow
-rwxrwxrwx+  1 richard  staff     272426 avr. 27 15:13 local<calc file>.ods*
              everyone@:rwxpdDaARWcCos:-------:allow
-rwxrwxrwx+  1 root     root           0 avr. 27 15:13 local-foo*
              everyone@:rwxpdDaARWcCos:-------:allow
-rwxrwxrwx   1 richard  staff     270914 avr. 27 15:25 nfs modify <calc file>.ods*
                 owner@:rwxp--aARWcCos:-------:allow
                 group@:rwxp--a-R-c--s:-------:allow
              everyone@:rwxp--a-R-c--s:-------:allow

I have a gut feeling that the resulting ACL over NFS is behind the problem somehow.

Is it possible that this is a combination of a OpenOffice 3.3 CALC and NFS problem.
There are no problems under local or CIFS calc access.

I'd like to know if LibreOffice behaves the same way in save/save_as over NFS... is there a version that works on oi_151a3?

Pointers to help get over this data corruption problem?

Actions #5

Updated by Bayard Bell about 10 years ago

  • Tags deleted (needs-triage)

For something as important as data corruption, please escalate the issue via e-mail to oi-dev and or the illumos developer@.

Actions #6

Updated by Richard PALO about 10 years ago

sorry, perhaps I posted to the wrong lists:
http://article.gmane.org/gmane.os.openindiana.general/7581

Actions #7

Updated by Ken Mays over 9 years ago

  • Status changed from New to Closed

Migrate this over to Illumos-specific bug reports.

Actions #8

Updated by Marcel Telka over 9 years ago

  • Project changed from OpenIndiana Distribution to illumos gate
Actions #9

Updated by Marcel Telka over 9 years ago

  • Status changed from Closed to In Progress

Moved to illumos gate project and reopened.

Actions

Also available in: Atom PDF