[ANNOUNCE] autofs 4.1.4 beta2 release

[ANNOUNCE] autofs 4.1.4 beta2 release

[raven]

Hi all,

Here is autofs 4.1.4 beta2.

There haven't been many changes since beta1.

For more information look at the CHANGELOG in the distribution.

There are still a few problems that I'm aware of such as:

- the init script for Debian and Gentoo works better but still
    ins't handling the reload option completely.

- the bug with the selection of the default map type when using
    submounts still hasn't been resolved.

So let me know when you find more and I'll fix`em as quick as I can!

The distribution can be found at the usual place at:

http://www.kernel.org/pub/linux/daemons/autofs/v4

in the tars

autofs-4.1.4_beta2.tar.[gz|bz2]

also there is a tarball called:

autofs-4.1.4-beta2-deb.tar.[gz|bz2]

that can be used to build binary debs (see below).

It is possible to build packages for testing from the distribution. 
I'm not the maintainer of any of these distribution packages so please be 
aware that this is for convenience and testing.

To build an rpm package use:

rpmbuild -tb autofs-4.1.4_beta2.tar.gz

and the binary rpms will end up in the usual location.

To merge the package into Gentoo use:

tar zxf autofs-4.1.4_beta2.tar.gz
cd autofs-4.1.4_beta2/gentoo
export PORTDIR_OVERLAY=`pwd`
ebuild net-fs/autofs/autofs-4.1.4_beta2.ebuild digest
ebuild net-fs/autofs/autofs-4.1.4_beta2.ebuild clean install
ebuild net-fs/autofs/autofs-4.1.4_beta2.ebuild qmerge
unset PORTDIR_OVERLAY

To build binary debs for Debian download the tar refered to above and:

tar zxvf autofs-4.1.4-beta2-deb.tar.gz
dpkg-source -x autofs_4.1.4-beta2-1.dsc
cd autofs-4.1.4-beta2
dpkg-buildpackage -b

and the binary debs will be placed in the directory the tar was unpacked.

Hopefuly this will help with the testing

Ian

Re: [ANNOUNCE] autofs 4.1.4 beta2 release

[Steinar H. Gunderson]

On Mon, Feb 14, 2005 at 10:39:51PM +0800, raven@themaw.net wrote:
> Here is autofs 4.1.4 beta2.

Thanks! :-)

I've gone through all the Debian patches again, and this needs looking into:

- 000_autofs_4.1.3_sock-leak-fix is missing.
- 035_program_mount_repeated_last_character_of_map_output is in there, but
  there looks like there's an identical situation further down. Could you
  investigate? (I'm not really sure if it is an error or not, to be honest, I
  just want it checked. :-) )
- 051_maps_without_trailing_newline.diff is not in, it seems; did you forget
  it or just find a cleaner way?
- The check for /proc/modules does not seem to have went in (see
  http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=292762).

Apart from that, looks like everything went in; I'm building a new version
now and will upload after I get your feedback on the points above.

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

Re: [ANNOUNCE] autofs 4.1.4 beta2 release

[Steinar H. Gunderson]

On Tue, Feb 15, 2005 at 02:48:57AM +0100, Steinar H. Gunderson wrote:
> Apart from that, looks like everything went in; I'm building a new version
> now and will upload after I get your feedback on the points above.

I just found another bug; on upgrades, autofs attempts to restart itself,
which invokes "debian stop"; this invokes something like

  start-stop-daemon --stop --quiet --retry USR2/$DAEMON_EXIT_WAIT --pidfile $file --exec /usr/sbin/automount

The problem here is that the pid does not really run /usr/sbin/automount
anymore; or more correctly, /proc/<pid>/exe does not point to the current
/usr/sbin/automount, it points to the _old_ /usr/sbin/automount, and that
version was just removed as part of the upgrade.  Thus, the stop fails, and the
entire upgrade fails.

Any good ideas? Stripping the --exec would be the easiest solution, but
perhaps also the ugliest...
 
/* Steinar */ 
-- 
Homepage: http://www.sesse.net/

Re: [ANNOUNCE] autofs 4.1.4 beta2 release

[Jeff Moyer]

==> Regarding Re: [autofs] [ANNOUNCE] autofs 4.1.4 beta2 release; "Steinar H. Gunderson" <sgunderson@bigfoot.com> adds:

sgunderson> On Mon, Feb 14, 2005 at 10:39:51PM +0800, raven@themaw.net
sgunderson> wrote:
>> Here is autofs 4.1.4 beta2.

sgunderson> Thanks! :-)

sgunderson> I've gone through all the Debian patches again, and this needs
sgunderson> looking into:

sgunderson> - 000_autofs_4.1.3_sock-leak-fix is missing.

Could you post this patch, please?  I'm almost certain this fix is in the
tree.

Thanks!

-Jeff

Re: [ANNOUNCE] autofs 4.1.4 beta2 release

[Steinar H. Gunderson]

On Tue, Feb 15, 2005 at 08:46:24AM -0500, Jeff Moyer wrote:
> Could you post this patch, please?  I'm almost certain this fix is in the
> tree.

http://home.samfundet.no/~sesse/000_autofs_4.1.3_sock-leak-fix

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

Re: [ANNOUNCE] autofs 4.1.4 beta2 release

[Steinar H. Gunderson]

On Tue, Feb 15, 2005 at 03:04:14AM +0100, Steinar H. Gunderson wrote:
> Any good ideas? Stripping the --exec would be the easiest solution, but
> perhaps also the ugliest...

OK, I found the bug; autofs was not stopped on prerm (this was disabled to
"fix" another bug). I've reverted the changes and made some workarounds; they
are not relevant for autofs upstream, but will be part of the next Debian
package.

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

Re: [ANNOUNCE] autofs 4.1.4 beta2 release

[Jeff Moyer]

==> Regarding Re: [autofs] [ANNOUNCE] autofs 4.1.4 beta2 release; "Steinar H. Gunderson" <sgunderson@bigfoot.com> adds:

sgunderson> On Tue, Feb 15, 2005 at 08:46:24AM -0500, Jeff Moyer wrote:
>> Could you post this patch, please?  I'm almost certain this fix is in the
>> tree.

sgunderson> http://home.samfundet.no/~sesse/000_autofs_4.1.3_sock-leak-fix

We solved this problem a different way.  If we pass RPC_ANYSOCK to the
portmapper, it will use a reserved port by default for communications.  To
keep it from doing that, we create the socket ourselves.  We then do this:

                clnt_control(client, CLSET_FD_CLOSE, NULL);

Which tells the rpc layer to close the socket for us when we destroy the
client.  As such, there is no socket leak.

Thanks,

Jeff

Re: [ANNOUNCE] autofs 4.1.4 beta2 release

[Jeff Moyer]

==> Regarding Re: [autofs] [ANNOUNCE] autofs 4.1.4 beta2 release; "Steinar H. Gunderson" <sgunderson@bigfoot.com> adds:

sgunderson> On Mon, Feb 14, 2005 at 10:39:51PM +0800, raven@themaw.net
sgunderson> wrote:
>> Here is autofs 4.1.4 beta2.

sgunderson> Thanks! :-)

sgunderson> I've gone through all the Debian patches again, and this needs
sgunderson> looking into:

sgunderson> 035_program_mount_repeated_last_character_of_map_output is in
sgunderson> there, but there looks like there's an identical situation
sgunderson> further down. Could you investigate? (I'm not really sure if it
sgunderson> is an error or not, to be honest, I just want it checked. :-) )

Yes, looks like a problem to me.  Thanks!  Attached is a patch.

-Jeff

--- autofs-4.1.4_beta2/modules/lookup_program.c.program-map-stderr-extra-char	2005-02-15 09:30:46.672165592 -0500
+++ autofs-4.1.4_beta2/modules/lookup_program.c	2005-02-15 09:31:23.840515144 -0500
@@ -237,6 +237,7 @@ int lookup_mount(const char *root, const
 			if (read(epipefd[0], &ch, 1) < 1) {
 				FD_CLR(epipefd[0], &ourfds);
 				files_left--;
+				state = st_done;
 			} else if (ch == '\n') {
 				*errp = '\0';
 				if (errbuf[0])

Re: [ANNOUNCE] autofs 4.1.4 beta2 release

[Jeff Moyer]

==> Regarding Re: [autofs] [ANNOUNCE] autofs 4.1.4 beta2 release; Jeff Moyer <jmoyer@redhat.com> adds:

==> Regarding Re: [autofs] [ANNOUNCE] autofs 4.1.4 beta2 release; "Steinar H. Gunderson" <sgunderson@bigfoot.com> adds:
sgunderson> On Mon, Feb 14, 2005 at 10:39:51PM +0800, raven@themaw.net
sgunderson> wrote:
>>> Here is autofs 4.1.4 beta2.

sgunderson> Thanks! :-)

sgunderson> I've gone through all the Debian patches again, and this needs
sgunderson> looking into:

sgunderson> 035_program_mount_repeated_last_character_of_map_output is in
sgunderson> there, but there looks like there's an identical situation
sgunderson> further down. Could you investigate? (I'm not really sure if it
sgunderson> is an error or not, to be honest, I just want it checked. :-) )

jmoyer> Yes, looks like a problem to me.  Thanks!  Attached is a patch.

Actually, I fired too quickly on this one.  stderr is used for the error
reporting, so we shouldn't mark the state as done until we've read
everything from stdout.

-Jeff

Re: [ANNOUNCE] autofs 4.1.4 beta2 release

[Steinar H. Gunderson]

On Tue, Feb 15, 2005 at 09:44:05AM -0500, Jeff Moyer wrote:
>> Yes, looks like a problem to me.  Thanks!  Attached is a patch.
> Actually, I fired too quickly on this one.  stderr is used for the error
> reporting, so we shouldn't mark the state as done until we've read
> everything from stdout.

Sounds reasonable; I'll leave it as-is.

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

Re: [ANNOUNCE] autofs 4.1.4 beta2 release

[raven]

On Tue, 15 Feb 2005, Steinar H. Gunderson wrote:

> On Tue, Feb 15, 2005 at 03:04:14AM +0100, Steinar H. Gunderson wrote:
>> Any good ideas? Stripping the --exec would be the easiest solution, but
>> perhaps also the ugliest...
>
> OK, I found the bug; autofs was not stopped on prerm (this was disabled to
> "fix" another bug). I've reverted the changes and made some workarounds; they
> are not relevant for autofs upstream, but will be part of the next Debian
> package.

Except I'd like a patch for my Debian kit please.

Ian

Re: [ANNOUNCE] autofs 4.1.4 beta2 release

[raven]

On Tue, 15 Feb 2005, Jeff Moyer wrote:

> ==> Regarding Re: [autofs] [ANNOUNCE] autofs 4.1.4 beta2 release; "Steinar H. Gunderson" <sgunderson@bigfoot.com> adds:
>
> sgunderson> On Tue, Feb 15, 2005 at 08:46:24AM -0500, Jeff Moyer wrote:
>>> Could you post this patch, please?  I'm almost certain this fix is in the
>>> tree.
>
> sgunderson> http://home.samfundet.no/~sesse/000_autofs_4.1.3_sock-leak-fix
>
> We solved this problem a different way.  If we pass RPC_ANYSOCK to the
> portmapper, it will use a reserved port by default for communications.  To
> keep it from doing that, we create the socket ourselves.  We then do this:
>
>                clnt_control(client, CLSET_FD_CLOSE, NULL);
>
> Which tells the rpc layer to close the socket for us when we destroy the
> client.  As such, there is no socket leak.

Yep. The patch was a bit out of date.

Thanks Jeff.

Ian

Re: [ANNOUNCE] autofs 4.1.4 beta2 release

[raven]

On Tue, 15 Feb 2005, Steinar H. Gunderson wrote:

> - 051_maps_without_trailing_newline.diff is not in, it seems; did you forget
>  it or just find a cleaner way?

I think you'll find that this handles the issue (if I understand it 
correctly).

Index: modules/lookup_file.c
===================================================================
RCS file: /home/cvs/root/autofs/modules/lookup_file.c,v
retrieving revision 1.13
retrieving revision 1.14
diff -u -r1.13 -r1.14
--- modules/lookup_file.c	26 Jan 2005 05:31:38 -0000	1.13
+++ modules/lookup_file.c	6 Feb 2005 10:05:55 -0000	1.14
@@ -1,4 +1,4 @@
-#ident "$Id: lookup_file.c,v 1.13 2005/01/26 05:31:38 raven Exp $"
+#ident "$Id: lookup_file.c,v 1.14 2005/02/06 10:05:55 raven Exp $"
  /* ----------------------------------------------------------------------- *
   *
   *  lookup_file.c - module for Linux automount to query a flat file map
@@ -202,6 +202,11 @@
  			} else if (mapent_len < MAPENT_MAX_LEN) {
  				mapent_len++;
  				*(p++) = ch;
+				nch = getc(f);
+				if (nch == EOF &&
+				   (gotten == got_real || gotten == getting))
+				   	goto got_it;
+				ungetc(nch, f);
  			}
  			break;
  		}

Re: [ANNOUNCE] autofs 4.1.4 beta2 release

[raven]

On Tue, 15 Feb 2005, Steinar H. Gunderson wrote:

> - The check for /proc/modules does not seem to have went in (see
>  http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=292762).

I though I checks this?

Or have you changed the bug id?

Ian

Re: [ANNOUNCE] autofs 4.1.4 beta2 release

[raven]

On Tue, 15 Feb 2005 raven@themaw.net wrote:

> On Tue, 15 Feb 2005, Steinar H. Gunderson wrote:
>
>> - The check for /proc/modules does not seem to have went in (see
>> http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=292762).
>
> I though I checks this?
>
> Or have you changed the bug id?

I've read the bug report you refer to but if I'm to get it right you need 
to provide a patch.

Ian

Re: [ANNOUNCE] autofs 4.1.4 beta2 release

[raven]

On Tue, 15 Feb 2005 raven@themaw.net wrote:

> On Tue, 15 Feb 2005 raven@themaw.net wrote:
>
>> On Tue, 15 Feb 2005, Steinar H. Gunderson wrote:
>> 
>>> - The check for /proc/modules does not seem to have went in (see
>>> http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=292762).
>> 
>> I though I checks this?
>> 
>> Or have you changed the bug id?
>
> I've read the bug report you refer to but if I'm to get it right you need to 
> provide a patch.

Is this what you're seeking?

Index: samples/rc.autofs.in
===================================================================
RCS file: /home/cvs/root/autofs/samples/rc.autofs.in,v
retrieving revision 1.47
retrieving revision 1.48
diff -u -r1.47 -r1.48
--- samples/rc.autofs.in	10 Feb 2005 14:30:07 -0000	1.47
+++ samples/rc.autofs.in	15 Feb 2005 15:27:34 -0000	1.48
@@ -1,6 +1,6 @@
  #!/bin/bash
  #
-# $Id: rc.autofs.in,v 1.47 2005/02/10 14:30:07 raven Exp $
+# $Id: rc.autofs.in,v 1.48 2005/02/15 15:27:34 raven Exp $
  #
  # rc file for automount using a Sun-style "master map".
  # We first look for a local /etc/auto.master, then a YP
@@ -604,7 +604,7 @@
  		# autofs filesystem support not loaded
  		echo -n ' loading autofs4 kernel module,'
  		modprobe autofs4 || true
-	elif lsmod | grep -q autofs[^4]; then
+	elif ([ -f /proc/modules ] && lsmod) | grep -q autofs[^4]; then
  		# wrong autofs filesystem module loaded
  		echo
  		echo "WARNING: autofs kernel module is loaded, autofs4 needed"

Re: [ANNOUNCE] autofs 4.1.4 beta2 release

[raven]

On Mon, 14 Feb 2005 raven@themaw.net wrote:

>
> There are still a few problems that I'm aware of such as:

And a couple I've just found on scribled notes.

- At some point I found nis maps getting called program maps
    This may have been fixed. Can someone check please.
    It may have been when a file map of the same name is
    present in the map directory.

- I think there is a bug handling submount map entries where
    their map type is not specified. Anyone care to investigate
    and get back with findings please.

Ian

Re: [ANNOUNCE] autofs 4.1.4 beta2 release

[raven]

On Tue, 15 Feb 2005, Steinar H. Gunderson wrote:

> On Tue, Feb 15, 2005 at 10:47:12PM +0800, raven@themaw.net wrote:
>>> OK, I found the bug; autofs was not stopped on prerm (this was disabled to
>>> "fix" another bug). I've reverted the changes and made some workarounds;
>>> they are not relevant for autofs upstream, but will be part of the next
>>> Debian package.
>> Except I'd like a patch for my Debian kit please.
>
> http://home.samfundet.no/~sesse/autofs_handle-upgrades-cleanly.diff is
> basically what I've done.

Thanks.
Ian

Re: [ANNOUNCE] autofs 4.1.4 beta2 release

[Steinar H. Gunderson]

On Tue, Feb 15, 2005 at 10:47:12PM +0800, raven@themaw.net wrote:
>> OK, I found the bug; autofs was not stopped on prerm (this was disabled to
>> "fix" another bug). I've reverted the changes and made some workarounds;
>> they are not relevant for autofs upstream, but will be part of the next
>> Debian package.
> Except I'd like a patch for my Debian kit please.

http://home.samfundet.no/~sesse/autofs_handle-upgrades-cleanly.diff is
basically what I've done.

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

Re: [ANNOUNCE] autofs 4.1.4 beta2 release

[raven]

On Tue, 15 Feb 2005, Steinar H. Gunderson wrote:

> On Tue, Feb 15, 2005 at 11:16:04PM +0800, raven@themaw.net wrote:
>> Is this what you're seeking?
>
> Yes, that is essentially the same patch. It was not in 4.1.4-beta2.

Cool. It's in the tree now.

Re: [ANNOUNCE] autofs 4.1.4 beta2 release

[Steinar H. Gunderson]

On Tue, Feb 15, 2005 at 11:04:59PM +0800, raven@themaw.net wrote:
>>> - The check for /proc/modules does not seem to have went in (see
>>>   http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=292762).
>> I though I checks this?
>>
>> Or have you changed the bug id?
> I've read the bug report you refer to but if I'm to get it right you need 
> to provide a patch.

http://home.samfundet.no/~sesse/052_292762_dont_lsmod_on_monolithic_kernels
-- as you can see, it checks that /proc/modules exists (which it only does on
kernels with module support) before calling lsmod. Of course, if /proc isn't
mounted, this breaks, but if /proc isn't mounted something is quite broken
anyhow...

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

Re: [ANNOUNCE] autofs 4.1.4 beta2 release

[Steinar H. Gunderson]

On Tue, Feb 15, 2005 at 11:16:04PM +0800, raven@themaw.net wrote:
> Is this what you're seeking?

Yes, that is essentially the same patch. It was not in 4.1.4-beta2.

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

Re: [ANNOUNCE] autofs 4.1.4 beta2 release

[Steinar H. Gunderson]

On Tue, Feb 15, 2005 at 11:23:54PM +0800, raven@themaw.net wrote:
>> http://home.samfundet.no/~sesse/autofs_handle-upgrades-cleanly.diff is
>> basically what I've done.
> Thanks.

I'll have to change this again, unfortunately; during testing I found a
machine which had been upgraded from 4.1.3-8 to 4.1.3-9 without reboot, and
thus "/etc/init.d/autofs stop" (which I tried to do in my preinst) was
broken.

I'll have to add a debconf notice telling the admin this and then just let
it break if it breaks; it's the only reasonably sane solution I can see.

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

Re: [ANNOUNCE] autofs 4.1.4 beta2 release

[Steinar H. Gunderson]

On Tue, Feb 15, 2005 at 04:54:10PM +0100, Steinar H. Gunderson wrote:
> I'll have to add a debconf notice telling the admin this and then just let
> it break if it breaks; it's the only reasonably sane solution I can see.

I've just implemented this and uploaded autofs_4.1.3+4.1.4beta2-1 to Debian
unstable.

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