[ANNOUNCE] autofs 5.0.0 beta1

[ANNOUNCE] autofs 5.0.0 beta1

[Ian Kent]

Hi all,

I've finally got to a point where I can release a beta of
autofs 5.0.0.

I would really appreciate some real world testing of this
so I hope some of you are able to help.

Feedback is welcome.

autofs
======

For the moment the package can be found at:

ftp://ftp.kernel.org/pub/linux/daemons/autofs/v4

It is autofs-5.0.0_beta1.tar.[gz|bz2]

No source rpm is there as it can be produced by using:

rpmbuild -ts autofs-5.0.0_beta1.tar.gz

and the binary rpm by using:

rpmbuild -tb autofs-5.0.0_beta1.tar.gz

See the INSTALL file for information about configure options and
kernel requirements.

I have no way to test the Gentoo portage build and the build
files will need to be modified for beta1. It would be great
if someone could get this working and forward a patch before
the final release.

Here is the README.v5.release which outlines the changes:

autofs-5.0.0 beta1 release notes
================================

Differences between version 4 and version 5.

- Master map is now read and parsed by the `automount' daemon
  - the parameters for the `automount' daemon are now very different.
    See automount(8).

- The `automount' daemon is now a multi-threaded application and so
  appears as a single process (unless a thread process display option
  is used).

- `autofs' filesystem mounts only appear in /proc/mounts and not
  /etc/mtab. If you have a large number of entries in a direct mount
  map and link /etc/mtab to /proc/mounts you may see performance
  degradation.

- `autofs' version 5.0.0 will refuse to run if it cannot find an
   autofs4 kernel module that supports protocol version 5.00 or above.

New in 5.0.0 is:

- improved direct mount map support
  - single level map entries now work as expected.
  - top level directory of direct mount heirachy no longer obscurses
    directory tree below.

- `+' map inclusion
  - maps preceded by a `+' in file maps may now be included inline.
  - implemented for the master map and mount maps.

- added nsswitch map source support
  - is now used to locate maps if no map source is given in their
    specification.
  - is used for bothe master map and mount maps.

- rewrote multi-mount map code
  - multi-mount map entry offsets now mount and expire indpendently
    instead of having to mount and umount all offset entries at once.

- added LDAP encryption and authentication support
  - needs to be enabled using a configure option.
  - a configuration file specifies whether encryption and authentication
    is to be used (see INSTALL and auto.master(5) for more information).

- improved shutdown and restart
  - needs to be enabled using a configure option.
  - see INSTALL for more information.

- a "hosts" map module has been added
  - an entry like "/net -hosts" is now understood and uses the new
    multi-mount semantics for lazy mount/umount of exports from the host.
  - the implementation is quite simple minded at this stage and will need
    refinment. It just iterates through the host table and addes any host
    names it finds to the internal map.

Things not yet done
===================

- I have no way to test the Gentoo setup here anymore so it is unlikey
  to work. Patches from anyone who needs this are welcome.

- the underscore to dot translation has not been added yet.

- there's a problen with the calcuation of the best host in replicated
  server selection.

- all the things I've broken and not re-tested after further development!


Ian

Re: [ANNOUNCE] autofs 5.0.0 beta1

[Steinar H. Gunderson]

On Tue, May 02, 2006 at 11:35:02PM +0800, Ian Kent wrote:
> - Master map is now read and parsed by the `automount' daemon
>   - the parameters for the `automount' daemon are now very different.
>     See automount(8).

Ouch. Are there any plans for a conversion tool?

> - `autofs' version 5.0.0 will refuse to run if it cannot find an
>    autofs4 kernel module that supports protocol version 5.00 or above.

What kernels are this?

/* Steinar */
-- 
Homepage: http://www.sesse.net/

Re: [ANNOUNCE] autofs 5.0.0 beta1

[Ian Kent]

On Tue, 2 May 2006, Steinar H. Gunderson wrote:

> On Tue, May 02, 2006 at 11:35:02PM +0800, Ian Kent wrote:
> > - Master map is now read and parsed by the `automount' daemon
> >   - the parameters for the `automount' daemon are now very different.
> >     See automount(8).
> 
> Ouch. Are there any plans for a conversion tool?

Maybe this is poorly worded.
Don't think any conversion tool is needed.

This was always done in the init script.
The master map would be parsed and individual daemons started.
Now the daemon reads and parses the master map taking all that 
troublesome logic from the init away.

The master map itself remains the same.

> 
> > - `autofs' version 5.0.0 will refuse to run if it cannot find an
> >    autofs4 kernel module that supports protocol version 5.00 or above.
> 
> What kernels are this?

See the INSTALL file in the tarball. It describes the requirements and 
included kernel patches. The changes are in the 2.6.17-rc series kernel 
now.

Let me know if you think we need more information.

Ian

Re: [ANNOUNCE] autofs 5.0.0 beta1

[Moshe Kaminsky]

[-- Attachment #1.1: Type: text/plain, Size: 2005 bytes --]

* Ian Kent <raven@themaw.net> [02/05/06 20:30]:
> 
> Hi all,
> 
> I've finally got to a point where I can release a beta of
> autofs 5.0.0.
> 
> I would really appreciate some real world testing of this
> so I hope some of you are able to help.
> 
> Feedback is welcome.
> 
> autofs
> ======
> 
> For the moment the package can be found at:
> 
> ftp://ftp.kernel.org/pub/linux/daemons/autofs/v4
> 
> It is autofs-5.0.0_beta1.tar.[gz|bz2]
> 
> No source rpm is there as it can be produced by using:
> 
> rpmbuild -ts autofs-5.0.0_beta1.tar.gz
> 
> and the binary rpm by using:
> 
> rpmbuild -tb autofs-5.0.0_beta1.tar.gz
> 
> See the INSTALL file for information about configure options and
> kernel requirements.
> 
> I have no way to test the Gentoo portage build and the build
> files will need to be modified for beta1. It would be great
> if someone could get this working and forward a patch before
> the final release.
> 

I modified the ebuild for autofs4 and tried to compile. I get:

gcc -g -Os -march=athlon-tbird -pipe -fomit-frame-pointer -D_REENTRANT -D_REENTRANT -rdynamic -fPIE -D_GNU_SOURCE -I../include -DAUTOFS_LIB_DIR=\"/usr/lib/autofs\"  -DAUTOFS_MAP_DIR=\"/etc/autofs\" -DAUTOFS_CONF_DIR=\"/etc/sysconfig\" -DVERSION_STRING=\"5.0.0_beta1\" -c direct.c
direct.c: In function `handle_packet_expire_direct':
direct.c:969: error: label at end of compound statement
direct.c: In function `handle_packet_missing_direct':
direct.c:1260: error: label at end of compound statement
make[1]: *** [direct.o] Error 1
make[1]: Leaving directory `/usr/tmp/portage/autofs-5.0.0_beta1/work/autofs-5.0.0_beta1/daemon'

Note that I removed all the patches that were used for autofs4. I'm 
using gcc 3.4.6.

Another question I have: I once asked if it's possible to run the actual 
mount command as the user trying to read the directory, instead of root. 
I was told that this is impossible in version 4. Will it be possible 
now?

Thanks,
Moshe

[-- Attachment #1.2: Type: application/pgp-signature, Size: 198 bytes --]

[-- Attachment #2: Type: text/plain, Size: 140 bytes --]

_______________________________________________
autofs mailing list
autofs@linux.kernel.org
http://linux.kernel.org/mailman/listinfo/autofs

Re: [ANNOUNCE] autofs 5.0.0 beta1

[Ian Kent]

On Wed, 3 May 2006, Moshe Kaminsky wrote:

> * Ian Kent <raven@themaw.net> [02/05/06 20:30]:
> > 
> > Hi all,
> > 
> > I've finally got to a point where I can release a beta of
> > autofs 5.0.0.
> > 
> > I would really appreciate some real world testing of this
> > so I hope some of you are able to help.
> > 
> > Feedback is welcome.
> > 
> > autofs
> > ======
> > 
> > For the moment the package can be found at:
> > 
> > ftp://ftp.kernel.org/pub/linux/daemons/autofs/v4
> > 
> > It is autofs-5.0.0_beta1.tar.[gz|bz2]
> > 
> > No source rpm is there as it can be produced by using:
> > 
> > rpmbuild -ts autofs-5.0.0_beta1.tar.gz
> > 
> > and the binary rpm by using:
> > 
> > rpmbuild -tb autofs-5.0.0_beta1.tar.gz
> > 
> > See the INSTALL file for information about configure options and
> > kernel requirements.
> > 
> > I have no way to test the Gentoo portage build and the build
> > files will need to be modified for beta1. It would be great
> > if someone could get this working and forward a patch before
> > the final release.
> > 
> 
> I modified the ebuild for autofs4 and tried to compile. I get:
> 
> gcc -g -Os -march=athlon-tbird -pipe -fomit-frame-pointer -D_REENTRANT -D_REENTRANT -rdynamic -fPIE -D_GNU_SOURCE -I../include -DAUTOFS_LIB_DIR=\"/usr/lib/autofs\"  -DAUTOFS_MAP_DIR=\"/etc/autofs\" -DAUTOFS_CONF_DIR=\"/etc/sysconfig\" -DVERSION_STRING=\"5.0.0_beta1\" -c direct.c
> direct.c: In function `handle_packet_expire_direct':
> direct.c:969: error: label at end of compound statement
> direct.c: In function `handle_packet_missing_direct':
> direct.c:1260: error: label at end of compound statement
> make[1]: *** [direct.o] Error 1
> make[1]: Leaving directory `/usr/tmp/portage/autofs-5.0.0_beta1/work/autofs-5.0.0_beta1/daemon'

I've seen that error as well.
It's quite odd because it seems to be dependent of the version of glibc 
pthread headers specifically the pthread_cleanup_push and 
pthread_cleanup_pop macros.

Once I got that error when some of the files had a timestamp in the future 
and "make install" tried to re-compile some of the sources.

I'll see if I can re-arrange the pthread_cleanup_push and 
pthread_cleanup_pop calls that cause this to happen.

> 
> Note that I removed all the patches that were used for autofs4. I'm 
> using gcc 3.4.6.
> 
> Another question I have: I once asked if it's possible to run the actual 
> mount command as the user trying to read the directory, instead of root. 
> I was told that this is impossible in version 4. Will it be possible 
> now?

Do you have a mount that will work for a user?

You can use the $USER substitution variable now to mount with a 
"user=${USER}" option or $UID and $GID with "uid=${UID}" and "gid=${GID}" 
in your map. Will that help?

Ian

Re: [ANNOUNCE] autofs 5.0.0 beta1

[Ian Kent]

On Wed, 3 May 2006, Moshe Kaminsky wrote:

> * Ian Kent <raven@themaw.net> [02/05/06 20:30]:
> > 
> > Hi all,
> > 
> > I've finally got to a point where I can release a beta of
> > autofs 5.0.0.
> > 
> > I would really appreciate some real world testing of this
> > so I hope some of you are able to help.
> > 
> > Feedback is welcome.
> > 
> > autofs
> > ======
> > 
> > For the moment the package can be found at:
> > 
> > ftp://ftp.kernel.org/pub/linux/daemons/autofs/v4
> > 
> > It is autofs-5.0.0_beta1.tar.[gz|bz2]
> > 
> > No source rpm is there as it can be produced by using:
> > 
> > rpmbuild -ts autofs-5.0.0_beta1.tar.gz
> > 
> > and the binary rpm by using:
> > 
> > rpmbuild -tb autofs-5.0.0_beta1.tar.gz
> > 
> > See the INSTALL file for information about configure options and
> > kernel requirements.
> > 
> > I have no way to test the Gentoo portage build and the build
> > files will need to be modified for beta1. It would be great
> > if someone could get this working and forward a patch before
> > the final release.
> > 
> 
> I modified the ebuild for autofs4 and tried to compile. I get:
> 
> gcc -g -Os -march=athlon-tbird -pipe -fomit-frame-pointer
 -D_REENTRANT -D_REENTRANT -rdynamic -fPIE -D_GNU_SOURCE
 -I../include -DAUTOFS_LIB_DIR=\"/usr/lib/autofs\" 
 -DAUTOFS_MAP_DIR=\"/etc/autofs\" -DAUTOFS_CONF_DIR=\"/etc/sysconfig\" -DVERSION_STRING=\"5.0.0_beta1\" -c direct.c

There's something else wrong with the build.
Shouldn't AUTOFS_CONF_DIR be /etc/conf.d instead of /etc/sysconfig?
configure should look for the directory?
What did I get wrong there?
Does /etc/sysconfig exist on you system?
Maybe it should be set explicitly at configure time.

Ian

Re: [ANNOUNCE] autofs 5.0.0 beta1

[Ian Kent]

On Wed, 3 May 2006, Moshe Kaminsky wrote:

> 
> I modified the ebuild for autofs4 and tried to compile. I get:
> 
> gcc -g -Os -march=athlon-tbird -pipe -fomit-frame-pointer -D_REENTRANT -D_REENTRANT -rdynamic -fPIE -D_GNU_SOURCE -I../include -DAUTOFS_LIB_DIR=\"/usr/lib/autofs\"  -DAUTOFS_MAP_DIR=\"/etc/autofs\" -DAUTOFS_CONF_DIR=\"/etc/sysconfig\" -DVERSION_STRING=\"5.0.0_beta1\" -c direct.c
> direct.c: In function `handle_packet_expire_direct':
> direct.c:969: error: label at end of compound statement
> direct.c: In function `handle_packet_missing_direct':
> direct.c:1260: error: label at end of compound statement
> make[1]: *** [direct.o] Error 1
> make[1]: Leaving directory `/usr/tmp/portage/autofs-5.0.0_beta1/work/autofs-5.0.0_beta1/daemon'

I'm not sure this will make a difference but could you give it a try.

diff --git a/daemon/direct.c b/daemon/direct.c
index 96f934f..f9b6dda 100644
--- a/daemon/direct.c
+++ b/daemon/direct.c
@@ -877,7 +877,7 @@ int handle_packet_expire_direct(struct a
 	struct pending_args *mt;
 	char buf[MAX_ERR_BUF];
 	pthread_t thid;
-	int status = 0;
+	int status = 0, cancel_state;
 
 	/*
 	 * This is a bit of a big deal.
@@ -953,9 +953,10 @@ int handle_packet_expire_direct(struct a
 		free(mt);
 		send_fail(mt->ioctlfd, pkt->wait_queue_token);
 		status = 1;
+		goto done;
 	}
 
-	pthread_cleanup_push(pending_cleanup, mt);
+	pthread_setcancelstate(PTHREAD_CANCEL_DISABLE, &cancel_state);
 
 	mt->signaled = 0;
 	while (!mt->signaled) {
@@ -964,7 +965,8 @@ int handle_packet_expire_direct(struct a
 			fatal(status);
 	}
 
-	pthread_cleanup_pop(1);
+	pending_cleanup(mt);
+	pthread_setcancelstate(cancel_state, &cancel_state);
 done:
 	pthread_cleanup_pop(1);
 	return status;
@@ -1151,7 +1153,7 @@ int handle_packet_missing_direct(struct 
 	pthread_t thid;
 	struct pending_args *mt;
 	char buf[MAX_ERR_BUF];
-	int status = 0;
+	int status = 0, cancel_state;
 	int ioctlfd;
 
 	master_source_readlock(ap->entry);
@@ -1246,7 +1248,7 @@ int handle_packet_missing_direct(struct 
 		goto done;
 	}
 
-	pthread_cleanup_push(pending_cleanup, mt);
+	pthread_setcancelstate(PTHREAD_CANCEL_DISABLE, &cancel_state);
 
 	mt->signaled = 0;
 	while (!mt->signaled) {
@@ -1255,7 +1257,8 @@ int handle_packet_missing_direct(struct 
 			fatal(status);
 	}
 
-	pthread_cleanup_pop(1);
+	pending_cleanup(mt);
+	pthread_setcancelstate(cancel_state, &cancel_state);
 done:
 	pthread_cleanup_pop(1);
 	return status;

> 
> Note that I removed all the patches that were used for autofs4. I'm 
> using gcc 3.4.6.

You'll have to review the Gentoo init script also.
A starting point is the one in the tarball. It should be fairly close but 
you'll need to review it.

> 
> Another question I have: I once asked if it's possible to run the actual 
> mount command as the user trying to read the directory, instead of root. 
> I was told that this is impossible in version 4. Will it be possible 
> now?
> 
> Thanks,
> Moshe
> 

Re: [ANNOUNCE] autofs 5.0.0 beta1

[Moshe Kaminsky]

[-- Attachment #1.1: Type: text/plain, Size: 824 bytes --]

* Ian Kent <raven@themaw.net> [03/05/06 18:00]:
> On Wed, 3 May 2006, Moshe Kaminsky wrote:
> > Another question I have: I once asked if it's possible to run the 
> > actual mount command as the user trying to read the directory, 
> > instead of root. I was told that this is impossible in version 4. 
> > Will it be possible now?
> 
> Do you have a mount that will work for a user?
> 
> You can use the $USER substitution variable now to mount with a 
> "user=${USER}" option or $UID and $GID with "uid=${UID}" and "gid=${GID}" 
> in your map. Will that help?
> 

I don't think so. What I have in mind is that a user will be able to 
automount his home directory on a remote machine using sshfs. For this, 
that user has to actually run the mount command, so that his ~/.ssh is 
used.

Thanks,
Moshe


[-- Attachment #1.2: Type: application/pgp-signature, Size: 198 bytes --]

[-- Attachment #2: Type: text/plain, Size: 140 bytes --]

_______________________________________________
autofs mailing list
autofs@linux.kernel.org
http://linux.kernel.org/mailman/listinfo/autofs

Re: [ANNOUNCE] autofs 5.0.0 beta1

[Moshe Kaminsky]

[-- Attachment #1.1: Type: text/plain, Size: 894 bytes --]

* Ian Kent <raven@themaw.net> [03/05/06 18:30]:
> On Wed, 3 May 2006, Moshe Kaminsky wrote:
> 
> > gcc -g -Os -march=athlon-tbird -pipe -fomit-frame-pointer
>  -D_REENTRANT -D_REENTRANT -rdynamic -fPIE -D_GNU_SOURCE
>  -I../include -DAUTOFS_LIB_DIR=\"/usr/lib/autofs\" 
>  -DAUTOFS_MAP_DIR=\"/etc/autofs\" -DAUTOFS_CONF_DIR=\"/etc/sysconfig\" -DVERSION_STRING=\"5.0.0_beta1\" -c direct.c
> 
> There's something else wrong with the build.
> Shouldn't AUTOFS_CONF_DIR be /etc/conf.d instead of /etc/sysconfig?
> configure should look for the directory?
> What did I get wrong there?
> Does /etc/sysconfig exist on you system?
> Maybe it should be set explicitly at configure time.
> 
> Ian
> 

The /etc/sysconfig does exist (don't know why), and the configure takes 
it first. I added --with-confdir=/etc/conf.d to the ebuild, and it 
solved the problem.

Thanks,
Moshe


[-- Attachment #1.2: Type: application/pgp-signature, Size: 198 bytes --]

[-- Attachment #2: Type: text/plain, Size: 140 bytes --]

_______________________________________________
autofs mailing list
autofs@linux.kernel.org
http://linux.kernel.org/mailman/listinfo/autofs

Re: [ANNOUNCE] autofs 5.0.0 beta1

[Moshe Kaminsky]

[-- Attachment #1.1: Type: text/plain, Size: 1176 bytes --]

* Ian Kent <raven@themaw.net> [03/05/06 19:15]:
> On Wed, 3 May 2006, Moshe Kaminsky wrote:
> 
> > 
> > I modified the ebuild for autofs4 and tried to compile. I get:
> > 
> > gcc -g -Os -march=athlon-tbird -pipe -fomit-frame-pointer -D_REENTRANT -D_REENTRANT -rdynamic -fPIE -D_GNU_SOURCE -I../include -DAUTOFS_LIB_DIR=\"/usr/lib/autofs\"  -DAUTOFS_MAP_DIR=\"/etc/autofs\" -DAUTOFS_CONF_DIR=\"/etc/sysconfig\" -DVERSION_STRING=\"5.0.0_beta1\" -c direct.c
> > direct.c: In function `handle_packet_expire_direct':
> > direct.c:969: error: label at end of compound statement
> > direct.c: In function `handle_packet_missing_direct':
> > direct.c:1260: error: label at end of compound statement
> > make[1]: *** [direct.o] Error 1
> > make[1]: Leaving directory `/usr/tmp/portage/autofs-5.0.0_beta1/work/autofs-5.0.0_beta1/daemon'
> 
> I'm not sure this will make a difference but could you give it a try.

This didn't help.

> 
> 
> You'll have to review the Gentoo init script also.
> A starting point is the one in the tarball. It should be fairly close but 
> you'll need to review it.
> 

I'll do that once I get this to compile.

Thanks,
Moshe

[-- Attachment #1.2: Type: application/pgp-signature, Size: 198 bytes --]

[-- Attachment #2: Type: text/plain, Size: 140 bytes --]

_______________________________________________
autofs mailing list
autofs@linux.kernel.org
http://linux.kernel.org/mailman/listinfo/autofs

Re: [ANNOUNCE] autofs 5.0.0 beta1

[Ian Kent]

On Wed, 3 May 2006, Moshe Kaminsky wrote:

> * Ian Kent <raven@themaw.net> [03/05/06 18:00]:
> > On Wed, 3 May 2006, Moshe Kaminsky wrote:
> > > Another question I have: I once asked if it's possible to run the 
> > > actual mount command as the user trying to read the directory, 
> > > instead of root. I was told that this is impossible in version 4. 
> > > Will it be possible now?
> > 
> > Do you have a mount that will work for a user?
> > 
> > You can use the $USER substitution variable now to mount with a 
> > "user=${USER}" option or $UID and $GID with "uid=${UID}" and "gid=${GID}" 
> > in your map. Will that help?
> > 
> 
> I don't think so. What I have in mind is that a user will be able to 
> automount his home directory on a remote machine using sshfs. For this, 
> that user has to actually run the mount command, so that his ~/.ssh is 
> used.

The problem here is that mount(8) will only allow the "user" option if 
an entry is present in the fstab which defeats the purpose of using 
autofs.

I'll think about it for a while.

Ian

Re: [ANNOUNCE] autofs 5.0.0 beta1

[Ian Kent]

On Wed, 3 May 2006, Moshe Kaminsky wrote:

> * Ian Kent <raven@themaw.net> [03/05/06 19:15]:
> > On Wed, 3 May 2006, Moshe Kaminsky wrote:
> > 
> > > 
> > > I modified the ebuild for autofs4 and tried to compile. I get:
> > > 
> > > gcc -g -Os -march=athlon-tbird -pipe -fomit-frame-pointer -D_REENTRANT -D_REENTRANT -rdynamic -fPIE -D_GNU_SOURCE -I../include -DAUTOFS_LIB_DIR=\"/usr/lib/autofs\"  -DAUTOFS_MAP_DIR=\"/etc/autofs\" -DAUTOFS_CONF_DIR=\"/etc/sysconfig\" -DVERSION_STRING=\"5.0.0_beta1\" -c direct.c
> > > direct.c: In function `handle_packet_expire_direct':
> > > direct.c:969: error: label at end of compound statement
> > > direct.c: In function `handle_packet_missing_direct':
> > > direct.c:1260: error: label at end of compound statement
> > > make[1]: *** [direct.o] Error 1
> > > make[1]: Leaving directory `/usr/tmp/portage/autofs-5.0.0_beta1/work/autofs-5.0.0_beta1/daemon'
> > 
> > I'm not sure this will make a difference but could you give it a try.
> 
> This didn't help.
> 

This fixes it.

diff --git a/daemon/direct.c b/daemon/direct.c
index 96f934f..66d2efa 100644
--- a/daemon/direct.c
+++ b/daemon/direct.c
@@ -953,6 +953,7 @@ int handle_packet_expire_direct(struct a
 		free(mt);
 		send_fail(mt->ioctlfd, pkt->wait_queue_token);
 		status = 1;
+		goto done;
 	}
 
 	pthread_cleanup_push(pending_cleanup, mt);
@@ -965,9 +966,12 @@ int handle_packet_expire_direct(struct a
 	}
 
 	pthread_cleanup_pop(1);
-done:
 	pthread_cleanup_pop(1);
 	return status;
+
+done:
+	cache_lock_cleanup(mc);
+	return status;
 }
 
 static void *do_mount_direct(void *arg)
@@ -1256,8 +1260,11 @@ int handle_packet_missing_direct(struct 
 	}
 
 	pthread_cleanup_pop(1);
-done:
 	pthread_cleanup_pop(1);
 	return status;
+
+done:
+	cache_lock_cleanup(mc);
+	return status;
 }

RE: [ANNOUNCE] autofs 5.0.0 beta1

[Rigler, Stephen C.]

I have it installed on one FC5 machine running the vanilla 2.6.16.12
kernel
patched with the autofs4-2.6.16-v5 patch.

Operationally everything seems good so far.  I like how the direct
mounts
are ghosted (which is consistent behavior with the other unixes we run).

The only issue I had was with parsing of auto.master.  We get our
auto.master
from NIS and automount didn't seem to want to parse it even when
specifying
"+auto.master" in /etc/auto.master.  I ended up ypcat'ing auto.master
into
the local file and it now handles all of our maps.

-Steve


> -----Original Message-----
> From: autofs-bounces@linux.kernel.org 
> [mailto:autofs-bounces@linux.kernel.org] On Behalf Of Ian Kent
> Sent: Tuesday, May 02, 2006 10:35 AM
> To: autofs mailing list
> Cc: linux-fsdevel
> Subject: [autofs] [ANNOUNCE] autofs 5.0.0 beta1
> 
> 
> Hi all,
> 
> I've finally got to a point where I can release a beta of
> autofs 5.0.0.
> 
> I would really appreciate some real world testing of this
> so I hope some of you are able to help.
> 
> Feedback is welcome.
> 
> autofs
> ======
> 
> For the moment the package can be found at:
> 
> ftp://ftp.kernel.org/pub/linux/daemons/autofs/v4
> 
> It is autofs-5.0.0_beta1.tar.[gz|bz2]
> 
> No source rpm is there as it can be produced by using:
> 
> rpmbuild -ts autofs-5.0.0_beta1.tar.gz
> 
> and the binary rpm by using:
> 
> rpmbuild -tb autofs-5.0.0_beta1.tar.gz
> 
> See the INSTALL file for information about configure options and
> kernel requirements.
> 
> I have no way to test the Gentoo portage build and the build
> files will need to be modified for beta1. It would be great
> if someone could get this working and forward a patch before
> the final release.
> 
> Here is the README.v5.release which outlines the changes:
> 
> autofs-5.0.0 beta1 release notes
> ================================
> 
> Differences between version 4 and version 5.
> 
> - Master map is now read and parsed by the `automount' daemon
>   - the parameters for the `automount' daemon are now very different.
>     See automount(8).
> 
> - The `automount' daemon is now a multi-threaded application and so
>   appears as a single process (unless a thread process display option
>   is used).
> 
> - `autofs' filesystem mounts only appear in /proc/mounts and not
>   /etc/mtab. If you have a large number of entries in a direct mount
>   map and link /etc/mtab to /proc/mounts you may see performance
>   degradation.
> 
> - `autofs' version 5.0.0 will refuse to run if it cannot find an
>    autofs4 kernel module that supports protocol version 5.00 or above.
> 
> New in 5.0.0 is:
> 
> - improved direct mount map support
>   - single level map entries now work as expected.
>   - top level directory of direct mount heirachy no longer obscurses
>     directory tree below.
> 
> - `+' map inclusion
>   - maps preceded by a `+' in file maps may now be included inline.
>   - implemented for the master map and mount maps.
> 
> - added nsswitch map source support
>   - is now used to locate maps if no map source is given in their
>     specification.
>   - is used for bothe master map and mount maps.
> 
> - rewrote multi-mount map code
>   - multi-mount map entry offsets now mount and expire indpendently
>     instead of having to mount and umount all offset entries at once.
> 
> - added LDAP encryption and authentication support
>   - needs to be enabled using a configure option.
>   - a configuration file specifies whether encryption and 
> authentication
>     is to be used (see INSTALL and auto.master(5) for more 
> information).
> 
> - improved shutdown and restart
>   - needs to be enabled using a configure option.
>   - see INSTALL for more information.
> 
> - a "hosts" map module has been added
>   - an entry like "/net -hosts" is now understood and uses the new
>     multi-mount semantics for lazy mount/umount of exports 
> from the host.
>   - the implementation is quite simple minded at this stage 
> and will need
>     refinment. It just iterates through the host table and 
> addes any host
>     names it finds to the internal map.
> 
> Things not yet done
> ===================
> 
> - I have no way to test the Gentoo setup here anymore so it is unlikey
>   to work. Patches from anyone who needs this are welcome.
> 
> - the underscore to dot translation has not been added yet.
> 
> - there's a problen with the calcuation of the best host in replicated
>   server selection.
> 
> - all the things I've broken and not re-tested after further 
> development!
> 
> 
> Ian
> 
> 
> _______________________________________________
> autofs mailing list
> autofs@linux.kernel.org
> http://linux.kernel.org/mailman/listinfo/autofs
> 
> 
> 

RE: [ANNOUNCE] autofs 5.0.0 beta1

[Ian Kent]

On Thu, 4 May 2006, Rigler, Stephen C. wrote:

> I have it installed on one FC5 machine running the vanilla 2.6.16.12
> kernel
> patched with the autofs4-2.6.16-v5 patch.
> 
> Operationally everything seems good so far.  I like how the direct
> mounts
> are ghosted (which is consistent behavior with the other unixes we run).
> 
> The only issue I had was with parsing of auto.master.  We get our
> auto.master
> from NIS and automount didn't seem to want to parse it even when
> specifying
> "+auto.master" in /etc/auto.master.  I ended up ypcat'ing auto.master

What do you have in the nsswitch.conf automount: entry?

You probably already know this but ...

The default master map is /etc/auto.master but that can be changed by 
setting DEFAULT_MASTER_MAP_NAME in the config or giving it on the command 
line for automount.

Setting

DEFAULT_MASTER_MAP_NAME=auto.master

and

automount: nis

should do the trick.

However if you had

automount: files nis

and /etc/auto.master exists with a +auto.master it will probably get it 
wrong in a strange way. Maybe that's what is happening?

It would be good if we could check this out a bit further and find out 
what's going wrong. I'm sure there's something I need to fix.

> into
> the local file and it now handles all of our maps.
> 
> -Steve
> 
> 
> > -----Original Message-----
> > From: autofs-bounces@linux.kernel.org 
> > [mailto:autofs-bounces@linux.kernel.org] On Behalf Of Ian Kent
> > Sent: Tuesday, May 02, 2006 10:35 AM
> > To: autofs mailing list
> > Cc: linux-fsdevel
> > Subject: [autofs] [ANNOUNCE] autofs 5.0.0 beta1
> > 
> > 
> > Hi all,
> > 
> > I've finally got to a point where I can release a beta of
> > autofs 5.0.0.
> > 
> > I would really appreciate some real world testing of this
> > so I hope some of you are able to help.
> > 
> > Feedback is welcome.
> > 
> > autofs
> > ======
> > 
> > For the moment the package can be found at:
> > 
> > ftp://ftp.kernel.org/pub/linux/daemons/autofs/v4
> > 
> > It is autofs-5.0.0_beta1.tar.[gz|bz2]
> > 
> > No source rpm is there as it can be produced by using:
> > 
> > rpmbuild -ts autofs-5.0.0_beta1.tar.gz
> > 
> > and the binary rpm by using:
> > 
> > rpmbuild -tb autofs-5.0.0_beta1.tar.gz
> > 
> > See the INSTALL file for information about configure options and
> > kernel requirements.
> > 
> > I have no way to test the Gentoo portage build and the build
> > files will need to be modified for beta1. It would be great
> > if someone could get this working and forward a patch before
> > the final release.
> > 
> > Here is the README.v5.release which outlines the changes:
> > 
> > autofs-5.0.0 beta1 release notes
> > ================================
> > 
> > Differences between version 4 and version 5.
> > 
> > - Master map is now read and parsed by the `automount' daemon
> >   - the parameters for the `automount' daemon are now very different.
> >     See automount(8).
> > 
> > - The `automount' daemon is now a multi-threaded application and so
> >   appears as a single process (unless a thread process display option
> >   is used).
> > 
> > - `autofs' filesystem mounts only appear in /proc/mounts and not
> >   /etc/mtab. If you have a large number of entries in a direct mount
> >   map and link /etc/mtab to /proc/mounts you may see performance
> >   degradation.
> > 
> > - `autofs' version 5.0.0 will refuse to run if it cannot find an
> >    autofs4 kernel module that supports protocol version 5.00 or above.
> > 
> > New in 5.0.0 is:
> > 
> > - improved direct mount map support
> >   - single level map entries now work as expected.
> >   - top level directory of direct mount heirachy no longer obscurses
> >     directory tree below.
> > 
> > - `+' map inclusion
> >   - maps preceded by a `+' in file maps may now be included inline.
> >   - implemented for the master map and mount maps.
> > 
> > - added nsswitch map source support
> >   - is now used to locate maps if no map source is given in their
> >     specification.
> >   - is used for bothe master map and mount maps.
> > 
> > - rewrote multi-mount map code
> >   - multi-mount map entry offsets now mount and expire indpendently
> >     instead of having to mount and umount all offset entries at once.
> > 
> > - added LDAP encryption and authentication support
> >   - needs to be enabled using a configure option.
> >   - a configuration file specifies whether encryption and 
> > authentication
> >     is to be used (see INSTALL and auto.master(5) for more 
> > information).
> > 
> > - improved shutdown and restart
> >   - needs to be enabled using a configure option.
> >   - see INSTALL for more information.
> > 
> > - a "hosts" map module has been added
> >   - an entry like "/net -hosts" is now understood and uses the new
> >     multi-mount semantics for lazy mount/umount of exports 
> > from the host.
> >   - the implementation is quite simple minded at this stage 
> > and will need
> >     refinment. It just iterates through the host table and 
> > addes any host
> >     names it finds to the internal map.
> > 
> > Things not yet done
> > ===================
> > 
> > - I have no way to test the Gentoo setup here anymore so it is unlikey
> >   to work. Patches from anyone who needs this are welcome.
> > 
> > - the underscore to dot translation has not been added yet.
> > 
> > - there's a problen with the calcuation of the best host in replicated
> >   server selection.
> > 
> > - all the things I've broken and not re-tested after further 
> > development!
> > 
> > 
> > Ian
> > 
> > 
> > _______________________________________________
> > autofs mailing list
> > autofs@linux.kernel.org
> > http://linux.kernel.org/mailman/listinfo/autofs
> > 
> > 
> > 
> 
> _______________________________________________
> autofs mailing list
> autofs@linux.kernel.org
> http://linux.kernel.org/mailman/listinfo/autofs
> 

RE: [ANNOUNCE] autofs 5.0.0 beta1

[Rigler, Stephen C.]

 

> -----Original Message-----
> From: Ian Kent [mailto:raven@themaw.net] 
> Sent: Thursday, May 04, 2006 7:29 AM
> To: Rigler, Stephen C.
> Cc: autofs mailing list
> Subject: RE: [autofs] [ANNOUNCE] autofs 5.0.0 beta1
> 
> What do you have in the nsswitch.conf automount: entry?

I had "automount: files nis"

> 
> You probably already know this but ...
> 
> The default master map is /etc/auto.master but that can be changed by 
> setting DEFAULT_MASTER_MAP_NAME in the config or giving it on 
> the command 
> line for automount.
> 
> Setting
> 
> DEFAULT_MASTER_MAP_NAME=auto.master
> 
> and
> 
> automount: nis
> 
> should do the trick.
> 
> However if you had
> 
> automount: files nis
> 

So far, the combination that's worked is only specifying "nis"
in nsswitch.conf and changing "DEFAULT_MASTER_MAP_NAME" to
"auto.master".  What's logged in messages when I use any other
combination is "master_read_master: no mounts in table".

-Steve

Re: [ANNOUNCE] autofs 5.0.0 beta1

[Jeff Moyer]

==> Regarding RE: [autofs] [ANNOUNCE] autofs 5.0.0 beta1; Ian Kent <raven@themaw.net> adds:

raven> On Thu, 4 May 2006, Rigler, Stephen C. wrote:
>> I have it installed on one FC5 machine running the vanilla 2.6.16.12
>> kernel patched with the autofs4-2.6.16-v5 patch.
>> 
>> Operationally everything seems good so far.  I like how the direct
>> mounts are ghosted (which is consistent behavior with the other unixes
>> we run).
>> 
>> The only issue I had was with parsing of auto.master.  We get our
>> auto.master from NIS and automount didn't seem to want to parse it even
>> when specifying "+auto.master" in /etc/auto.master.  I ended up
>> ypcat'ing auto.master

raven> What do you have in the nsswitch.conf automount: entry?

raven> You probably already know this but ...

raven> The default master map is /etc/auto.master but that can be changed
raven> by setting DEFAULT_MASTER_MAP_NAME in the config or giving it on the
raven> command line for automount.

raven> Setting

raven> DEFAULT_MASTER_MAP_NAME=auto.master

raven> and

raven> automount: nis

raven> should do the trick.

raven> However if you had

raven> automount: files nis

raven> and /etc/auto.master exists with a +auto.master it will probably get
raven> it wrong in a strange way. Maybe that's what is happening?

Yeah, I think this is the one exception to the nsswitch rule.  I wonder if
it's the same if you have /etc/auto.foo that does a +auto.foo?  May be
worth testing that out on another UNIX box to see what happens.

-Jeff

Re: [ANNOUNCE] autofs 5.0.0 beta1

[Jeff Moyer]

==> Regarding RE: [autofs] [ANNOUNCE] autofs 5.0.0 beta1; "Rigler, Stephen C." <srigler@marathonoil.com> adds:

>> -----Original Message----- From: Ian Kent [mailto:raven@themaw.net]
>> Sent: Thursday, May 04, 2006 7:29 AM To: Rigler, Stephen C.  Cc: autofs
>> mailing list Subject: RE: [autofs] [ANNOUNCE] autofs 5.0.0 beta1
>> 
>> What do you have in the nsswitch.conf automount: entry?

srigler> I had "automount: files nis"

>> You probably already know this but ...
>> 
>> The default master map is /etc/auto.master but that can be changed by
>> setting DEFAULT_MASTER_MAP_NAME in the config or giving it on the
>> command line for automount.
>> 
>> Setting
>> 
>> DEFAULT_MASTER_MAP_NAME=auto.master
>> 
>> and
>> 
>> automount: nis
>> 
>> should do the trick.
>> 
>> However if you had
>> 
>> automount: files nis
>> 

srigler> So far, the combination that's worked is only specifying "nis" in
srigler> nsswitch.conf and changing "DEFAULT_MASTER_MAP_NAME" to
srigler> "auto.master".  What's logged in messages when I use any other
srigler> combination is "master_read_master: no mounts in table".

You should also be able to keep your nsswitch the same (automount: files
nis) and simply remove /etc/auto.master.

-Jeff

Re: [ANNOUNCE] autofs 5.0.0 beta1

[Moshe Kaminsky]

[-- Attachment #1.1: Type: text/plain, Size: 2064 bytes --]

* Ian Kent <raven@themaw.net> [03/05/06 21:15]:
> On Wed, 3 May 2006, Moshe Kaminsky wrote:
> 
> > * Ian Kent <raven@themaw.net> [03/05/06 19:15]:
> > > On Wed, 3 May 2006, Moshe Kaminsky wrote:
> > > 
> > > > 
> > > > I modified the ebuild for autofs4 and tried to compile. I get:
> > > > 
> > > > gcc -g -Os -march=athlon-tbird -pipe -fomit-frame-pointer -D_REENTRANT -D_REENTRANT -rdynamic -fPIE -D_GNU_SOURCE -I../include -DAUTOFS_LIB_DIR=\"/usr/lib/autofs\"  -DAUTOFS_MAP_DIR=\"/etc/autofs\" -DAUTOFS_CONF_DIR=\"/etc/sysconfig\" -DVERSION_STRING=\"5.0.0_beta1\" -c direct.c
> > > > direct.c: In function `handle_packet_expire_direct':
> > > > direct.c:969: error: label at end of compound statement
> > > > direct.c: In function `handle_packet_missing_direct':
> > > > direct.c:1260: error: label at end of compound statement
> > > > make[1]: *** [direct.o] Error 1
> > > > make[1]: Leaving directory `/usr/tmp/portage/autofs-5.0.0_beta1/work/autofs-5.0.0_beta1/daemon'
> > > 
> > > I'm not sure this will make a difference but could you give it a try.
> > 
> > This didn't help.
> > 
> 
> This fixes it.
> 
> diff --git a/daemon/direct.c b/daemon/direct.c
> index 96f934f..66d2efa 100644
> --- a/daemon/direct.c
> +++ b/daemon/direct.c
> @@ -953,6 +953,7 @@ int handle_packet_expire_direct(struct a
>  		free(mt);
>  		send_fail(mt->ioctlfd, pkt->wait_queue_token);
>  		status = 1;
> +		goto done;
>  	}
>  
>  	pthread_cleanup_push(pending_cleanup, mt);
> @@ -965,9 +966,12 @@ int handle_packet_expire_direct(struct a
>  	}
>  
>  	pthread_cleanup_pop(1);
> -done:
>  	pthread_cleanup_pop(1);
>  	return status;
> +
> +done:
> +	cache_lock_cleanup(mc);
> +	return status;
>  }
>  
>  static void *do_mount_direct(void *arg)
> @@ -1256,8 +1260,11 @@ int handle_packet_missing_direct(struct 
>  	}
>  
>  	pthread_cleanup_pop(1);
> -done:
>  	pthread_cleanup_pop(1);
>  	return status;
> +
> +done:
> +	cache_lock_cleanup(mc);
> +	return status;
>  }
>  
> 

Thanks, that solved it.

Moshe

[-- Attachment #1.2: Type: application/pgp-signature, Size: 198 bytes --]

[-- Attachment #2: Type: text/plain, Size: 140 bytes --]

_______________________________________________
autofs mailing list
autofs@linux.kernel.org
http://linux.kernel.org/mailman/listinfo/autofs

RE: [ANNOUNCE] autofs 5.0.0 beta1

[Rigler, Stephen C.]

> -----Original Message-----
> From: Jeff Moyer [mailto:jmoyer@redhat.com] 
> Sent: Thursday, May 04, 2006 8:56 AM
> To: Ian Kent
> Cc: Rigler, Stephen C.; autofs mailing list
> Subject: Re: [autofs] [ANNOUNCE] autofs 5.0.0 beta1
> 
> raven> and /etc/auto.master exists with a +auto.master it 
> will probably get
> raven> it wrong in a strange way. Maybe that's what is happening?
> 
> Yeah, I think this is the one exception to the nsswitch rule. 
>  I wonder if
> it's the same if you have /etc/auto.foo that does a +auto.foo?  May be
> worth testing that out on another UNIX box to see what happens.
> 
> -Jeff
> 

Pretty much standard configuration here on Solaris and Irix is
to have "automount: files nis" in nsswitch.conf and "+auto.master"
in the local auto_master file.  I'm not aware if this is what
the standard automount behavior should be (it's documented in the
Irix man page for autofs), but we've gotten away with it for years 
if it's not.

Another nice thing to see would be the "-null" option in auto.master.
We have machines running clustered filesystems that are also
exported via NFS from one or two meta-data servers.  The machines
that see the filesystems locally should not automount the fileystems
and the "-null" would allow us to exclude them.

-Steve

Re: [ANNOUNCE] autofs 5.0.0 beta1

[Moshe Kaminsky]

[-- Attachment #1.1: Type: text/plain, Size: 1270 bytes --]

* Ian Kent <raven@themaw.net> [03/05/06 20:45]:
> On Wed, 3 May 2006, Moshe Kaminsky wrote:
> 
> > * Ian Kent <raven@themaw.net> [03/05/06 18:00]:
> > > On Wed, 3 May 2006, Moshe Kaminsky wrote:
> > > > Another question I have: I once asked if it's possible to run the 
> > > > actual mount command as the user trying to read the directory, 
> > > > instead of root. I was told that this is impossible in version 4. 
> > > > Will it be possible now?
> > > 
> > > Do you have a mount that will work for a user?
> > > 
> > > You can use the $USER substitution variable now to mount with a 
> > > "user=${USER}" option or $UID and $GID with "uid=${UID}" and "gid=${GID}" 
> > > in your map. Will that help?
> > > 
> > 
> > I don't think so. What I have in mind is that a user will be able to 
> > automount his home directory on a remote machine using sshfs. For this, 
> > that user has to actually run the mount command, so that his ~/.ssh is 
> > used.
> 
> The problem here is that mount(8) will only allow the "user" option if 
> an entry is present in the fstab which defeats the purpose of using 
> autofs.

Funny. It appears that you're right. I'm sure it used to work, though 
(maybe something changed in the kernel?)

Thanks,
Moshe

[-- Attachment #1.2: Type: application/pgp-signature, Size: 198 bytes --]

[-- Attachment #2: Type: text/plain, Size: 140 bytes --]

_______________________________________________
autofs mailing list
autofs@linux.kernel.org
http://linux.kernel.org/mailman/listinfo/autofs

Re: [ANNOUNCE] autofs 5.0.0 beta1

[Jeff Moyer]

==> Regarding RE: [autofs] [ANNOUNCE] autofs 5.0.0 beta1; "Rigler, Stephen C." <srigler@marathonoil.com> adds:

>> -----Original Message----- From: Jeff Moyer [mailto:jmoyer@redhat.com]
>> Sent: Thursday, May 04, 2006 8:56 AM To: Ian Kent Cc: Rigler, Stephen
>> C.; autofs mailing list Subject: Re: [autofs] [ANNOUNCE] autofs 5.0.0
>> beta1
>> 
raven> and /etc/auto.master exists with a +auto.master it
>> will probably get
raven> it wrong in a strange way. Maybe that's what is happening?
>> Yeah, I think this is the one exception to the nsswitch rule. I wonder
>> if it's the same if you have /etc/auto.foo that does a +auto.foo?  May
>> be worth testing that out on another UNIX box to see what happens.
>> 
>> -Jeff
>> 

srigler> Pretty much standard configuration here on Solaris and Irix is to
srigler> have "automount: files nis" in nsswitch.conf and "+auto.master" in
srigler> the local auto_master file.  I'm not aware if this is what the
srigler> standard automount behavior should be (it's documented in the Irix
srigler> man page for autofs), but we've gotten away with it for years if
srigler> it's not.

srigler> Another nice thing to see would be the "-null" option in
srigler> auto.master.  We have machines running clustered filesystems that
srigler> are also exported via NFS from one or two meta-data servers.  The
srigler> machines that see the filesystems locally should not automount the
srigler> fileystems and the "-null" would allow us to exclude them.

Well, you're in luck.  Give autofs5 a try, as it is supposed to handle
-null maps.

-Jeff

Re: [ANNOUNCE] autofs 5.0.0 beta1

[Ian Kent]

On Thu, 4 May 2006, Moshe Kaminsky wrote:

> * Ian Kent <raven@themaw.net> [03/05/06 20:45]:
> > On Wed, 3 May 2006, Moshe Kaminsky wrote:
> > 
> > > * Ian Kent <raven@themaw.net> [03/05/06 18:00]:
> > > > On Wed, 3 May 2006, Moshe Kaminsky wrote:
> > > > > Another question I have: I once asked if it's possible to run the 
> > > > > actual mount command as the user trying to read the directory, 
> > > > > instead of root. I was told that this is impossible in version 4. 
> > > > > Will it be possible now?
> > > > 
> > > > Do you have a mount that will work for a user?
> > > > 
> > > > You can use the $USER substitution variable now to mount with a 
> > > > "user=${USER}" option or $UID and $GID with "uid=${UID}" and "gid=${GID}" 
> > > > in your map. Will that help?
> > > > 
> > > 
> > > I don't think so. What I have in mind is that a user will be able to 
> > > automount his home directory on a remote machine using sshfs. For this, 
> > > that user has to actually run the mount command, so that his ~/.ssh is 
> > > used.
> > 
> > The problem here is that mount(8) will only allow the "user" option if 
> > an entry is present in the fstab which defeats the purpose of using 
> > autofs.
> 
> Funny. It appears that you're right. I'm sure it used to work, though 
> (maybe something changed in the kernel?)

All this checking is done in mount(8). I've investigated this before when 
checking this type of question. Maybe yours.

Anyway, after thinking about this and checking out sshfs this is not a 
problem. Getting the sshfs mount done as the user (now we know the 
username) would be possible by writting a simple mount_sshfs.c module but 
it looks like the resulting mount can't be un-mounted by umount. Is that 
right?

Ian

RE: [ANNOUNCE] autofs 5.0.0 beta1

[Ian Kent]

On Thu, 4 May 2006, Rigler, Stephen C. wrote:

> 
> > -----Original Message-----
> > From: Jeff Moyer [mailto:jmoyer@redhat.com] 
> > Sent: Thursday, May 04, 2006 8:56 AM
> > To: Ian Kent
> > Cc: Rigler, Stephen C.; autofs mailing list
> > Subject: Re: [autofs] [ANNOUNCE] autofs 5.0.0 beta1
> > 
> > raven> and /etc/auto.master exists with a +auto.master it 
> > will probably get
> > raven> it wrong in a strange way. Maybe that's what is happening?
> > 
> > Yeah, I think this is the one exception to the nsswitch rule. 
> >  I wonder if
> > it's the same if you have /etc/auto.foo that does a +auto.foo?  May be
> > worth testing that out on another UNIX box to see what happens.
> > 
> > -Jeff
> > 
> 
> Pretty much standard configuration here on Solaris and Irix is
> to have "automount: files nis" in nsswitch.conf and "+auto.master"
> in the local auto_master file.  I'm not aware if this is what
> the standard automount behavior should be (it's documented in the
> Irix man page for autofs), but we've gotten away with it for years 
> if it's not.
> 
> Another nice thing to see would be the "-null" option in auto.master.
> We have machines running clustered filesystems that are also
> exported via NFS from one or two meta-data servers.  The machines
> that see the filesystems locally should not automount the fileystems
> and the "-null" would allow us to exclude them.

It should understand "-null" but the question is whether it works 
correctly and how it should work. Give a try and let me know how it goes.

What would you have in your map in this case?

Ian

RE: [ANNOUNCE] autofs 5.0.0 beta1

[Rigler, Stephen C.]

> -----Original Message-----
> From: Ian Kent [mailto:raven@themaw.net] 
> Sent: Thursday, May 04, 2006 8:38 PM
> To: Rigler, Stephen C.
> Cc: Jeff Moyer; autofs mailing list
> Subject: RE: [autofs] [ANNOUNCE] autofs 5.0.0 beta1
> 
> > Another nice thing to see would be the "-null" option in 
> auto.master.
> > We have machines running clustered filesystems that are also
> > exported via NFS from one or two meta-data servers.  The machines
> > that see the filesystems locally should not automount the fileystems
> > and the "-null" would allow us to exclude them.
> 
> It should understand "-null" but the question is whether it works 
> correctly and how it should work. Give a try and let me know 
> how it goes.
> 
> What would you have in your map in this case?
> 
> Ian

"-null" seems to be working fine.

In my case, my /etc/auto.master has "+yp:auto.master" and then
"/somepath -null" to exclude an indirect mount or "/some/path -null"
to exclude a direct mount.  This is the inverse of the Solaris and
Irix boxes we handle where the "+auto.master" is the last line in
the local file.  I have not tried it with an actual filesystem 
mounted at the mountpoint I'm excluding.

The only inconsistency I've noticed if I remove the "-null" from
an indirect mount I need to restart automount (reload doesn't seem
to pick it back up).  "Reload" does work when I remove the "-null"
from a direct mount.

-Steve

RE: [ANNOUNCE] autofs 5.0.0 beta1

[Ian Kent]

On Mon, 8 May 2006, Rigler, Stephen C. wrote:

> 
> > -----Original Message-----
> > From: Ian Kent [mailto:raven@themaw.net] 
> > Sent: Thursday, May 04, 2006 8:38 PM
> > To: Rigler, Stephen C.
> > Cc: Jeff Moyer; autofs mailing list
> > Subject: RE: [autofs] [ANNOUNCE] autofs 5.0.0 beta1
> > 
> > > Another nice thing to see would be the "-null" option in 
> > auto.master.
> > > We have machines running clustered filesystems that are also
> > > exported via NFS from one or two meta-data servers.  The machines
> > > that see the filesystems locally should not automount the fileystems
> > > and the "-null" would allow us to exclude them.
> > 
> > It should understand "-null" but the question is whether it works 
> > correctly and how it should work. Give a try and let me know 
> > how it goes.
> > 
> > What would you have in your map in this case?
> > 
> > Ian
> 
> "-null" seems to be working fine.
> 
> In my case, my /etc/auto.master has "+yp:auto.master" and then
> "/somepath -null" to exclude an indirect mount or "/some/path -null"
> to exclude a direct mount.  This is the inverse of the Solaris and
> Irix boxes we handle where the "+auto.master" is the last line in
> the local file.  I have not tried it with an actual filesystem 
> mounted at the mountpoint I'm excluding.
> 
> The only inconsistency I've noticed if I remove the "-null" from
> an indirect mount I need to restart automount (reload doesn't seem
> to pick it back up).  "Reload" does work when I remove the "-null"
> from a direct mount.

Cool. I'll check that out and see what's going on.

Thanks
Ian