Commit 7f694582 introduced a bug where default_postinst() often fails to
remove a uci-defaults script after application, leaving it to run again
after a reboot.
(Note: commit 7f694582 also introduced FS#1021, now fixed by 73c745f6)
The subtle problem arises from the shell logical chain:
[ -f "$i" ] && . "$i" && rm -f "$i"
Most uci-defaults scripts contain a terminal 'exit 0' statement which,
when sourced, results in the logic chain exiting before executing 'rm -f'.
This was observed while testing upgrades of 'luci-app-sqm'.
The solution is to wrap the shell sourcing in a subshell relative to the
command 'rm -f':
( [ -f "$i" ] && . "$i" ) && rm -f "$i"
Revert to using 'grep' to prefilter the list of entries from the control
file, which yields the full path of uci-defaults scripts. This allows
keeping the existence check, directory change and script sourcing inside
the subshell, with the script removal correctly outside.
This approach avoids adding a second subshell only around the "." (source)
command. The change also preserves the fix FS#1021, since the full path is
used to source the script, which is POSIX-portable irrespective of PATH
variable or reference to the CWD.
Run Tested on: LEDE 17.01.4 running ar71xx, while tracing installation of
package luci-app-sqm with its associated /etc/uci-defaults/luci-sqm file.
Signed-off-by: Tony Ambardar <itugrok@yahoo.com>
There are several long-standing issues present in the UCI shell API as
documented in https://wiki.openwrt.org/doc/devel/config-scripting. They
relate both to high-level, user-defined callback functions used to
process UCI config files, and also to low-level functions used within
scripts generally.
The related problems have been encountered now and in the past, e.g.
https://forum.openwrt.org/viewtopic.php?id=54295, and include:
a) UCI parsing option() function and user-defined option_cb() callbacks
being erroneously called during processing of "list" config file entries;
b) normal usage of the low-level config_set() unexpectedy calling any
defined option_cb() if present; and
c) handling of the list_cb() not respecting the NO_CALLBACK variable.
Root causes include a function stack "inversion", where the low-level
config_set() function incorrectly calls the high-level option() function,
intended only for processing the "option" keyword of UCI config files.
This change addresses the inversion and other issues, making the option
handling code more consistent and smaller, and simplifying developers'
usage of UCI callbacks.
Signed-off-by: Tony Ambardar <itugrok@yahoo.com>
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com> [PKG_RELEASE increase]
Shell function return code only has range [0, 255]. Other values will
be truncated, e.g. return 65536 will have the same effect as return 0
While at it, drop other "return $rc" where rc will almost always take
value 0 and whose value current callers actually do not check
Fixes FS#988
Signed-off-by: Yousong Zhou <yszhou4tech@gmail.com>
When we run "opkg install" on a package that installs an uci-defaults
script, functions.sh will fail to evaluate that script in its
default_postinst function.
This happens because there is no "./" present and it searches for the
file in paths specified by the PATH variable. This would work on bash,
but it will not work on ash and some other shells like sh, zsh. This
applys to the ". filename" directive used in this case.
This patch will make the path relative to the /etc/uci-defaults
directory.
Fixes: FS#1021
Signed-off-by: Marko Ratkaj <marko.ratkaj@sartura.hr>
Let the generic postinstall script invoke "kmodloader" when the just
installed package contains any /etc/module.d/ entries.
This allows us to skip the explicit "insert_module()" calls in the
package postinstall.
Due to the removed insert_module calls we do not need to assemble a
complete list of modules per package anymore, which allows for vast
simplification of the package generation code.
While we're at it, also support specifying default parameters for
modules using either the MODPARAM or MODPARAM.modulename variables
in KernelPackage.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
There already exist static assignment of uid/gid 65533 in packages feed
and we have nobody/nogroup taking 65534 as their ids. Let's change the
pid of dynamic assignment to start from 65536 so that the two assignment
scheme will not collide with each other
While at it, fix the scan command checking existence of uid/gid
Signed-off-by: Yousong Zhou <yszhou4tech@gmail.com>
This adds support for install-overlay define. When used in package it
allows installing files to a special directory that gets copied to the
root when installing it.
It allows overwriting files provided by other packages.
Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
Packages that do a killall <cmd> with the same name as the init script
will fail the prerm step when the service isn't running. Do make them
removable without having to restart the service, ignore the return code.
Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com>
Acked-by: Jo-Philipp Wich <jo@mein.io>
Properly stop/start services on upgrade, but don't change the enabled
state.
Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com>
Acked-by: Jo-Philipp Wich <jo@mein.io>
Some package postinstall operations, e.g. those emitted by the LuCI build
system, source and delete the uci-defaults themselves upon package insteall,
causing the generic defaults apply code to trigger shell errors like:
.../luci-app-ddns.postinst: .: line 130: can't open './40_luci-ddns'
Rework the generic apply code to check the existence of the uci-defaults
script before trying to source it, use sed to prefilter the list of entries
from the control file and perform the directory change in a subshell,
avoiding the need for cd $OLDPWD.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
Packages may install scripts into /etc/uci-defaults to be executed once
after installation, usually at the first boot of the target. This works
fine if the package was installed to the rootfs during build or using
the ImageBuilder.
If the package is installed using opkg during run-time uci-defaults
were applied only after a reboot of the device. Avoid the need to
reboot by evaluting the package's uci-defaults in default-postinst.
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
- Simplify user spec parsing code
- Alias IPKG_INSTROOT to shorten code
- Quote variables where appropriate
- Remove a useless-use-of-cat instance
- Suppress errors in case control or list files are missing
Signed-off-by: Jo-Philipp Wich <jow@openwrt.org>
SVN-Revision: 46975
Using the postinst script for sanity checks and expecting opkg to fail
if the postinst didn't return 0 was possible in Barrier Breaker, propagate
the real postinst return code through default_postinst to restore this
behaviour.
Signed-off-by: Matthias Schiffer <mschiffer@universe-factory.net>
SVN-Revision: 46653
Instead of stripping everything after the first dot, strip everything
after the last dot.
This fixes pre-/postrm actions for packages with a dot in their name,
like libusb-1.0.
Fixes#19668.
Signed-off-by: Jonas Gorski <jogo@openwrt.org>
SVN-Revision: 45702
The name variable contains at first the package name,
and after the last group name.
This patch fixes /etc/rc.d symlink creation (at least).
Change name in pkgname, change the other name in ugname
(user group name), and id in ugid (user group id)
Signed-off-by: Etienne CHAMPETIER <champetier.etienne@gmail.com>
SVN-Revision: 44253
create user:group before running postinst-pkg. the postinst hook might require
the user:group to already exist.
Signed-off-by: Christian Schoenebeck <christian.schoenebeck@gmail.com>
SVN-Revision: 43472
Inside every LuCI package you need to clear luci-indexcache and
sometimes when installing non LuCI pacakges it's also needed to clear
it. Easier put it into default_postinst().
Signed-off-by: Christian Schoenebeck <christian.schoenebeck@gmail.com>
SVN-Revision: 42923
it is now possible to inlie the uid and gid in the syntax
USERID:=username=uid:group=gid:group2=gid2:...
Signed-off-by: John Crispin <blogic@openwrt.org>
SVN-Revision: 42838
the postinst script enables/starts the init.d scripts upon package installation
and installs the users required by the package.
the prerm script stops and disables the init.d scripts.
Signed-off-by: John Crispin <blogic@openwrt.org>
SVN-Revision: 42470
ubox validate_data defines yes/no as valid boolean options, do the same in
config_get_bool too.
Signed-off-by: Luka Perkov <luka@openwrt.org>
SVN-Revision: 41405
In commit r38690, the MAC address canonicalization
has been converted to use 'tr' instead of 'printf'.
This only works if with MAC addresses which uses
the 'xx:xx:xx:xx:xx:xx' format.
However on some boards, the MAC addresses are stored
in different format in the mtd partition. Some vendors
are using hyphens or dots as separators instead of
colons. Also the leading zeroes may be missing from the
individual octets or those are replaced with spaces.
Add a new function which can be used to convert these
into the 'xx:xx:xx:xx:xx:xx' format. Also update the
'mtd_get_mac_ascii' function to use the new helper.
The helper function is based on this code:
http://isquared.nl/blog/2010/08/11/Bash-function-to-canonicalize-MAC-addresses/
Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
SVN-Revision: 38803
The mtd_get_mac_ascii utility function was broken. This fixes it.
- Remove the superfluous include of /lib/functions.sh. The
function is already in that file so it is pointless,
- only use a variable if the whole key word matches,
- don't try to process the MAC address if it is empty,
- use 'tr' to canonicalize the MAC address,
Signed-off-by: Daniel Gimpelevich <daniel@gimpelevich.san-francisco.ca.us>
[juhosg: add more fixes, update commit message]
Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
SVN-Revision: 38690
