Red Hat Bugzilla – Bug 111290
Automounting of ncp connections are taking place at the wrong time in system init.
Last modified: 2014-03-16 22:40:44 EDT
From Bugzilla Helper: User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.4.1) Gecko/20031114 Description of problem: I'm running a ncp connection when starting up the computer, and have noticed error messages on the console, in dmesg and boot.log... So I started to go through init procedures in netfs, which to me looks perfectly OK, but in rc.sysinit you've made a little miss. When looking in it, I found: # Mount all other filesystems (except for NFS and /proc, which is already # mounted). Contrary to standard usage, # filesystems are NOT unmounted in single user mode. action $"Mounting local filesystems: " mount -a -t nonfs,smbfs,ncpfs -O no_netdev Which isn't really true.. If I include the ncp mounts in fstab, then they will be initialised there too, although the network interface isn't brought up yet. In a nutshell please change the above line in rc.sysinit to read action $"Mounting local filesystems: " mount -a -t nonfs,noncp,smbfs,ncpfs -O no_netdev My guess is that smbfs should also be nosmb or nosmbfs because again the network interface is not up yet. Most of this message has been copied from bug 65404. It appears that this bug has been around for the last 18 months or so. Jason Version-Release number of selected component (if applicable): initscripts-7.42-1 How reproducible: Always Steps to Reproduce: 1. Just add a ncp mount in fstab for auto mounting. 2. 3. Actual Results: The mounting local filesystems failed because the network interface was not up yet. It failed with an error "mount.ncp: Server not found (0x8847)" Expected Results: The ncp mounts should take place in /etc/rc.d/init.d/netfs which they do (also...) AFTER the network is up and running. Additional info:
See bug 65404 comment: -a -t noXXX,YYY,ZZZ The 'no' applies to all three of XXX, YYY, and ZZZ. If it's trying to mount YYY or ZZZ, either mount is broken, or it's docs are. Is your FS type 'ncpfs' or 'ncp'?
ncp is the FS type. I don't know if mount is broken or if it's docs.
What happens if you change the type to 'ncpfs' in the fstab?
If I put in ncpfs it works like it should. No errors during mounting local filesystems. However the instructions that come with ncpmount specifically state the following: /etc/fstab You must specify filesystem type ncp and not ncpfs although it is reported as ncpfs in /etc/mtab and /proc/mounts. Do you want to email Petr Vandrovec (vandrove@vc.cvut.cz) or should I? I noticed that ncpmount man file is from 12/04/1998. Is the code maintained? I wonder what the reason is that man page says to use ncp and not ncpfs. I guess a lot has changed over the last 5 years to make the man page wrong. Here is a copy of my fstab if you want to take a look. ofcapps/dravet /mnt/novell ncpfs auto,tcp,ipserver=ofcapps.calumet.purdue.edu,mode=444,uid=root,gid=root,owner=root,volume=legdata/ocapsys,passwdfile=/root/novell,retry=6,timeout=200,multiple /dev/hda1 /mnt/winxp ntfs ro,auto,user,uid=0,gid=0,umask=0227 0 0 Jason
Does normal mount of /mnt/novell work with ncpfs in the fstab?
I used to use the following before fstab worked right. ncpmount -S ofcapps -A ofcapps.calumet.purdue.edu -U dravet -V legdata/ocapsys /mnt/novell after entering my password I could view everything I needed. The stange thing was after I mounted the volume after 30 seconds of no use the mount stopped working. I would get input/output errors. By accident I discovered that if I mount the volume, exit, and login again the mount would remain until I unmounted it or rebooted. Strange. I am not sure mount would work because I need to use NCP over IP and not IPX. To use IP I have to have the -A parameter. Other than this I don't understand what you mean.
Sorry about the delayed response. The reason one value of 'ncp' or 'ncpfs' may be preferred depends on how the package is installed. The value specified there is what mount looks for for the program to mount the filesystem; if it's specified as 'ncp', it looks for /sbin/mount.ncp; if it's specified as 'ncpfs' it looks for /sbin/mount.ncpfs. Mounting as 'ncp' works because we ship both (they're hardlinks). I'm tempted to say that the fix here is to remove mount.ncp, so that it follows the filesystem type convention fully. But for now both are ignored as of initscripts-7.47-1.