postinst problem: module vs. update-rc.d

postinst problem: module vs. update-rc.d

[Steffen Sledz]

We have a problem with a package that needs to install a kernel module and an initscript.

The recipe contains

  inherit module update-rc.d

The generated postinst script is

  #!/bin/sh
  if [ -n "$D" ]; then
        exit 1
  else
        depmod -a 2.6.24.7
        update-modules || true
  fi
  if test "x$D" != "x"; then
        OPT="-r $D"
  else
        OPT="-s"
  fi
  update-rc.d $OPT somescript start 90 S . stop 10 0 1 6 .

The 'exit 1' leads to an error if $D contains a nonzero length string (what is the case in the OE image creation step at the build host) and suppresses the execution of the second (update-rc.d) part.

Because of this error the postinst is run a second time when the target device boots the first time ("opkg-cl configure" from /etc/rcS.d/S40configure).

The problem is that the update-rc.d execution (and therefor the creation of the new symlink into rcS.d) is done from within rcS itself. So the rcS.d/S90somescript will *not* be executed at this boot. :(

We see two possible solutions.

A) Don't use "inherit module update-rc.d" but write the pkg_postint ourselfs.

B) Create two packages (one for the module, one for the initscript) with a runtime dependency.

I don't like both of them. Any other ideas? Is it possible to force the order of postinst fragments (update-rc.d code before module code)?

Regards,
Steffen

-- 
DResearch Fahrzeugelektronik GmbH
Otto-Schmirgal-Str. 3, 10319 Berlin, Germany
Tel: +49 30 515932-237 mailto:sledz@dresearch-fe.de
Fax: +49 30 515932-299
Geschäftsführer: Dr. Michael Weber, Werner Mögle;
Amtsgericht Berlin Charlottenburg; HRB 130120 B;
Ust.-IDNr. DE273952058

Re: postinst problem: module vs. update-rc.d

[Phil Blundell]

On Fri, 2011-10-14 at 16:00 +0200, Steffen Sledz wrote:
> The 'exit 1' leads to an error if $D contains a nonzero length string (what is the case in the OE image creation step at the build host) and suppresses the execution of the second (update-rc.d) part.
> 
> Because of this error the postinst is run a second time when the target device boots the first time ("opkg-cl configure" from /etc/rcS.d/S40configure).
> 
> The problem is that the update-rc.d execution (and therefor the creation of the new symlink into rcS.d) is done from within rcS itself. So the rcS.d/S90somescript will *not* be executed at this boot. :(
> 
> We see two possible solutions.
> 
> A) Don't use "inherit module update-rc.d" but write the pkg_postint ourselfs.
> 
> B) Create two packages (one for the module, one for the initscript) with a runtime dependency.
> 
> I don't like both of them. Any other ideas? Is it possible to force the order of postinst fragments (update-rc.d code before module code)?

I think either update-rc.d or update-rc.d.bbclass should probably invoke
the script as "start" when it installs it for the first time.

p.

Re: postinst problem: module vs. update-rc.d

[Martin Jansa]

[-- Attachment #1: Type: text/plain, Size: 2019 bytes --]

On Fri, Oct 14, 2011 at 04:00:04PM +0200, Steffen Sledz wrote:
> We have a problem with a package that needs to install a kernel module and an initscript.
> 
> The recipe contains
> 
>   inherit module update-rc.d
> 
> The generated postinst script is
> 
>   #!/bin/sh
>   if [ -n "$D" ]; then
>         exit 1
>   else
>         depmod -a 2.6.24.7
>         update-modules || true
>   fi
>   if test "x$D" != "x"; then
>         OPT="-r $D"
>   else
>         OPT="-s"
>   fi
>   update-rc.d $OPT somescript start 90 S . stop 10 0 1 6 .
> 
> The 'exit 1' leads to an error if $D contains a nonzero length string (what is the case in the OE image creation step at the build host) and suppresses the execution of the second (update-rc.d) part.
> 
> Because of this error the postinst is run a second time when the target device boots the first time ("opkg-cl configure" from /etc/rcS.d/S40configure).
> 
> The problem is that the update-rc.d execution (and therefor the creation of the new symlink into rcS.d) is done from within rcS itself. So the rcS.d/S90somescript will *not* be executed at this boot. :(
> 
> We see two possible solutions.
> 
> A) Don't use "inherit module update-rc.d" but write the pkg_postint ourselfs.
> 
> B) Create two packages (one for the module, one for the initscript) with a runtime dependency.
> 
> I don't like both of them. Any other ideas? Is it possible to force the order of postinst fragments (update-rc.d code before module code)?
> 
> Regards,
> Steffen

BTW: there is another problem with update-modules:

SHR root@gjama ~ $ opkg files syslog-ng | grep modules.conf
/etc/modules.conf
SHR root@gjama ~ $ update-modules
Error: the current /etc/modules.conf is not automatically generated.
Use "update-modules force" to force (re)generation.

so if someone is using syslog-ng, than kernel-module-* or any "inherit
module" package won't update ie autoloads properly..

-- 
Martin 'JaMa' Jansa     jabber: Martin.Jansa@gmail.com

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 205 bytes --]