On Sun, 2009-02-01 at 11:32 -0500, Steven Hirsch wrote:
> On Wed, 28 Jan 2009, Ian Kent wrote:
>
> >> The question is this: WHAT is going on with the autofs startup that
> >> requires me to first end up in location outside of it before it will work?
> >> Thinking it might be a race, I started experimenting with sleep calls to
> >> give autofs a chance to initialize before trying to move into it. No
> >> dice. Whether I wait 2 seconds or a minute, the results were the same.
> >>
> >> The schroot maintainer is mildly convinced that there must be an easier
> >> way to get automount working in the chroot and I tend to agree. Does
> >> anyone have any advice?
> >
> > Collect some debug logs of the different things you've tried and we'll
> > see if we can work out what is happening. Versions or autofs and kernel
> > could be useful.
>
> Autofs: 4.1.4+debian-2.1ubuntu1
>
> Kernel: 2.6.24-23-generic
>
> auto.master:
>
> #
> # $Id: auto.master,v 1.4 2005/01/04 14:36:54 raven Exp $
> #
> # Sample auto.master file
> # This is an automounter map and it has the following format
> # key [ -mount-options-separated-by-comma ] location
> # For details of the format look at autofs(5).
> #/misc /etc/auto.misc --timeout=60
> #/smb /etc/auto.smb
> #/misc /etc/auto.misc
> /net /etc/auto.net --debug -intr
>
> auto.net:
>
> pii -rsize=65536,wsize=65536,tcp,noatime pii.fast.net:/
> src -rsize=65536,wsize=65536,tcp,noatime pii.fast.net:/ips_src
> ips -rsize=65536,wsize=65536,tcp,noatime pii.fast.net:/ips
> home -rsize=65536,wsize=65536,tcp pii.fast.net:/ips_home/home
> dosD -rsize=65536,wsize=65536,tcp,noatime pii.fast.net:/ips_dos/dos
>
>
> /etc/nsswitch.conf:
>
> # /etc/nsswitch.conf
> #
> # Example configuration of GNU Name Service Switch functionality.
> # If you have the `glibc-doc-reference' and `info' packages installed,
> try:
> # `info libc "Name Service Switch"' for information about this file.
>
> passwd: compat
> group: compat
> shadow: compat
>
> hosts: files dns
> networks: files
>
> protocols: db files
> services: db files
> ethers: db files
> rpc: db files
>
> netgroup: nis
>
> Here are the debugging traces:
>
> Note: If I simply try to bind-mount the root environment's /net directory
> into the chroot tree, there is nothing in the log file. Autofs simply
> does not see it. I can only assume that the Linux VFS does not support
> bind mounting of automounter directories.
>
> All the following involve an init script that starts a second automount
> deamon inside the chroot environment:
>
> 1. If I create /net beforehand and attempt to log directly to a home
> directory in NFS at start of the chroot'ed session:
>
> Feb 1 11:07:21 duo automount[23853]: starting automounter version 4.1.4,
> path = /net, maptype = file, mapname = /etc/aut
> Feb 1 11:07:21 duo automount[23853]: parse(sun): init gathered options: intr
> Feb 1 11:07:21 duo automount[23853]: mount(bind): bind_works = 1
> Feb 1 11:07:21 duo automount[23853]: using kernel protocol version 4.00
> Feb 1 11:07:21 duo automount[23853]: using timeout 300 seconds; freq 75 secs
> Feb 1 11:07:21 duo automount[23853]: sig 12 switching from 1 to 5
I have no idea where the SIGUSR2 signal is coming from but it will cause
autofs to shutdown if there are no mounts that are busy when it gets it.
> Feb 1 11:07:21 duo automount[23853]: get_pkt: state 1, next 5
> Feb 1 11:07:21 duo automount[23853]: prep_shutdown: state = 1
> Feb 1 11:07:21 duo automount[23853]: signal_children: send 12 to process
> group 23853
> Feb 1 11:07:21 duo automount[23853]: expire_proc: exp_proc=23872
> Feb 1 11:07:21 duo automount[23853]: prep_shutdown: expire returns 1
> Feb 1 11:07:21 duo automount[23853]: handle_child: got pid 23872, sig 0 (0),
> stat 0
> Feb 1 11:07:21 duo automount[23853]: sigchld: exp 23872 finished, switching
> from 5 to 6
> Feb 1 11:07:21 duo automount[23853]: get_pkt: state 5, next 6
> Feb 1 11:07:21 duo automount[23853]: umount_multi: path=/net incl=0
> Feb 1 11:07:21 duo automount[23853]: umount_multi: no mounts found under /net
> Feb 1 11:07:21 duo automount[23853]: shut down, path = /net
>
> 2. If I do not create /net beforehand and try to log directly to home:
autofs must be able to create directories for it's mount points if they
don't already exist. So not really a useful test.
>
> Feb 1 11:09:01 duo automount[24610]: starting automounter version 4.1.4,
> path = /net, maptype = file, mapname = /etc/auto.net
> Feb 1 11:09:01 duo automount[24610]: parse(sun): init gathered options: intr
> Feb 1 11:09:01 duo automount[24610]: mount(bind): bind_works = 1
> Feb 1 11:09:01 duo automount[24610]: failed to create iautofs directory /net
> Feb 1 11:09:01 duo automount[24610]: /net: mount failed!
>
> 3. If I create /net beforehand and log to a local filesystem initially at
> start of the chroot'ed session:
>
> Feb 1 11:11:14 duo automount[25830]: starting automounter version 4.1.4,
> path = /net, maptype = file, mapname = /etc/auto.net
> Feb 1 11:11:14 duo automount[25830]: parse(sun): init gathered options: intr
> Feb 1 11:11:14 duo automount[25830]: mount(bind): bind_works = 1
> Feb 1 11:11:14 duo automount[25830]: using kernel protocol version 4.00
> Feb 1 11:11:14 duo automount[25830]: using timeout 300 seconds; freq 75 secs
> Feb 1 11:11:14 duo automount[25830]: handle_packet: type = 0
> Feb 1 11:11:14 duo automount[25830]: handle_packet_missing: token 35, name
> home
> Feb 1 11:11:14 duo automount[25830]: attempting to mount entry /net/home
> Feb 1 11:11:14 duo automount[25835]: lookup(file): home ->
> -rsize=65536,wsize=65536,tcp pii.fast.net:/ips_home/home
> Feb 1 11:11:14 duo automount[25835]: parse(sun): expanded entry:
> -rsize=65536,wsize=65536,tcp pii.fast.net:/ips_home/home
> Feb 1 11:11:14 duo automount[25835]: parse(sun):
> dequote("rsize=65536,wsize=65536,tcp") -> rsize=65536,wsize=65536,tcp
> Feb 1 11:11:14 duo automount[25835]: parse(sun): gathered options:
> intr,rsize=65536,wsize=65536,tcp
> Feb 1 11:11:14 duo automount[25835]: parse(sun):
> dequote("pii.fast.net:/ips_home/home") -> pii.fast.net:/ips_home/home
> Feb 1 11:11:14 duo automount[25835]: parse(sun): core of entry:
> options=intr,rsize=65536,wsize=65536,tcp, loc=pii.fast.net:/ips_home/home
> Feb 1 11:11:14 duo automount[25835]: parse(sun): mounting root /net,
> mountpoint home, what pii.fast.net:/ips_home/home, fstype nfs, options
> intr,rsize=65536,wsize=65536,tcp
> Feb 1 11:11:14 duo automount[25835]: mount(nfs): root=/net name=home
> what=pii.fast.net:/ips_home/home, fstype=nfs,
> options=intr,rsize=65536,wsize=65536,tcp
> Feb 1 11:11:14 duo automount[25835]: mount(nfs): nfs
> options="intr,rsize=65536,wsize=65536,tcp", nosymlink=0, symlink=0, ro=0
> Feb 1 11:11:14 duo automount[25835]: mount(nfs): is_local_mount:
> pii.fast.net:/ips_home/home
> Feb 1 11:11:14 duo automount[25835]: mount(nfs): from
> pii.fast.net:/ips_home/home elected pii.fast.net:/ips_home/home
> Feb 1 11:11:14 duo automount[25835]: mount(nfs): calling mkdir_path /net/home
> Feb 1 11:11:14 duo automount[25835]: mount(nfs): calling mount -t nfs -s -o
> intr,rsize=65536,wsize=65536,tcp pii.fast.net:/ips_home/home /net/home
> Feb 1 11:11:14 duo automount[25835]: mount(nfs): mounted
> pii.fast.net:/ips_home/home on /net/home
> Feb 1 11:11:14 duo automount[25830]: handle_child: got pid 25835, sig 0 (0),
> stat 0
> Feb 1 11:11:14 duo automount[25830]: sig_child: found pending iop pid 25835:
> signalled 0 (sig 0), exit status 0
> Feb 1 11:11:14 duo automount[25830]: send_ready: token=35
>
> !! ====> Now I have an accessible home directory and can cd into it.
And no SIGUSR2, so that is expected I guess!
>
>
> Please let me know if more information is needed?
>
> Steve
>
>
_______________________________________________
autofs mailing list
autofs@linux.kernel.org
http://linux.kernel.org/mailman/listinfo/autofs
