skeleton + init: prepare /run/lock and adjust compat symlinks
PATCH 1/1 skeleton + init: prepare /run/lock and adjust compat symlinks
Linux distros changed /var/run and /var/lock to reside on an tmpfs in /run a long time ago, and buildroot seems to agree by providing this tmpfs on all supported init systems.
Packages on the other hand are currently free to either use /var/run or /run, and similarly using /var/lock should be identical to the replacement /run/lock.
This series aims to ensure that *if* /var/lock is needed, then it will be a symlink to /run/lock. Allowing packages to continue working and allowing them to migrate to using /run/lock.
If all relevant packages are fixed,
the compatibility symlinks in /var can be dropped.
Status before this patch:
* Systemd
/run will be mounted by PID1, /var/run will be recreated by /usr/lib/tmpfiles.d/var.conf.
Creating /run/lock/subsys and the compatibility symlink is handled in /usr/lib/tmpfiles.d/legacy.conf.
But this is *currently not installed* by Buildroot, see [1].
* OpenRC
Seems to check for existence of a /run directory and does all necessary setup.
* SysV
/var/lock is currently a symlink to /tmp.
* Busybox
Same as sysv (Buildroot uses the sysv skeleton)
Note that we create /run/lock/subsys, so sysv scrips could expect this directory to exist. Apart from simplifying scripts, creating the dir early as root adds some security.
--- package/busybox/inittab | 1 + package/skeleton-init-openrc/skeleton/var/lock | 2 +- package/skeleton-init-sysv/skeleton/var/lock | 2 +- package/sysvinit/inittab | 13 +++++++------ 4 files changed, 10 insertions(+), 8 deletions(-)
package/busybox/inittab +1 -0
@@ -18,6 +18,7 @@ ::sysinit:/bin/mount -o remount,rw / ::sysinit:/bin/mkdir -p /dev/pts /dev/shm ::sysinit:/bin/mount -a +::sysinit:/bin/mkdir -p /run/lock/subsys ::sysinit:/sbin/swapon -a null::sysinit:/bin/ln -sf /proc/self/fd /dev/fd null::sysinit:/bin/ln -sf /proc/self/fd/0 /dev/stdin
package/skeleton-init-openrc/skeleton/var/lock +1 -1
@@ -1 +1 @@ -../tmp +../run/lock
package/skeleton-init-sysv/skeleton/var/lock +1 -1
@@ -1 +1 @@ -../tmp +../run/lock
package/sysvinit/inittab +7 -6
@@ -8,12 +8,13 @@ si0::sysinit:/bin/mount -t proc proc /proc si1::sysinit:/bin/mount -o remount,rw / si2::sysinit:/bin/mkdir -p /dev/pts /dev/shm si3::sysinit:/bin/mount -a -si4::sysinit:/sbin/swapon -a -si5::sysinit:/bin/ln -sf /proc/self/fd /dev/fd 2>/dev/null -si6::sysinit:/bin/ln -sf /proc/self/fd/0 /dev/stdin 2>/dev/null -si7::sysinit:/bin/ln -sf /proc/self/fd/1 /dev/stdout 2>/dev/null -si8::sysinit:/bin/ln -sf /proc/self/fd/2 /dev/stderr 2>/dev/null -si9::sysinit:/bin/hostname -F /etc/hostname +si4::sysinit:/bin/mkdir -p /run/lock/subsys +si5::sysinit:/sbin/swapon -a +si6::sysinit:/bin/ln -sf /proc/self/fd /dev/fd 2>/dev/null +si7::sysinit:/bin/ln -sf /proc/self/fd/0 /dev/stdin 2>/dev/null +si8::sysinit:/bin/ln -sf /proc/self/fd/1 /dev/stdout 2>/dev/null +si9::sysinit:/bin/ln -sf /proc/self/fd/2 /dev/stderr 2>/dev/null +si10::sysinit:/bin/hostname -F /etc/hostname rcS:12345:wait:/etc/init.d/rcS # S0:1:respawn:/sbin/getty -L ttyS0 115200 vt100 # GENERIC_SERIAL
Signature · 1 line
2.28.0
Arnout Vandecappelle
AppliedQuoted 37 lines · Norbert Lange
Linux distros changed
/var/runand/var/lockto reside on an tmpfs in /run a long time ago, and buildroot seems to agree by providing this tmpfs on all supported init systems.Packages on the other hand are currently free to either use
/var/runor /run, and similarly using/var/lockshould be identical to the replacement/run/lock.This series aims to ensure that *if*
/var/lockis needed, then it will be a symlink to/run/lock. Allowing packages to continue working and allowing them to migrate to using/run/lock.If all relevant packages are fixed,
the compatibility symlinks in /var can be dropped.Status before this patch:
* Systemd
/run will be mounted by PID1,
/var/runwill be recreated by/usr/lib/tmpfiles.d/var.conf.Creating
/run/lock/subsysand the compatibility symlink is handled in/usr/lib/tmpfiles.d/legacy.conf.
But this is *currently not installed* by Buildroot, see [1].* OpenRC
Seems to check for existence of a /run directory and does all necessary setup.
* SysV
/var/lockis currently a symlink to /tmp.* Busybox
Quoted 7 lines
Same as sysv (Buildroot uses the sysv skeleton)
Note that we create
/run/lock/subsys, so sysv scrips could expect this directory to exist. Apart from simplifying scripts, creating the dir early as root adds some security.Signed-off-by Norbert Lange <[email protected]>
Applied to master, thanks.
Regards,
Arnout
Quoted 52 lines
--- package/busybox/inittab | 1 + package/skeleton-init-openrc/skeleton/var/lock | 2 +- package/skeleton-init-sysv/skeleton/var/lock | 2 +- package/sysvinit/inittab | 13 +++++++------ 4 files changed, 10 insertions(+), 8 deletions(-)package/busybox/inittab +1
-0@@ -18,6 +18,7 @@ ::sysinit:/bin/mount -o remount,rw / ::sysinit:/bin/mkdir -p /dev/pts /dev/shm ::sysinit:/bin/mount -a +::sysinit:/bin/mkdir -p /run/lock/subsys ::sysinit:/sbin/swapon -a null::sysinit:/bin/ln -sf /proc/self/fd /dev/fd null::sysinit:/bin/ln -sf /proc/self/fd/0 /dev/stdinpackage/skeleton-init-openrc/skeleton/var/lock +1
-1@@ -1 +1 @@ -../tmp +../run/lockpackage/skeleton-init-sysv/skeleton/var/lock +1
-1@@ -1 +1 @@ -../tmp +../run/lockpackage/sysvinit/inittab +0
-6@@ -8,12 +8,13 @@ si0::sysinit:/bin/mount -t proc proc /proc si1::sysinit:/bin/mount -o remount,rw / si2::sysinit:/bin/mkdir -p /dev/pts /dev/shm si3::sysinit:/bin/mount -a -si4::sysinit:/sbin/swapon -a -si5::sysinit:/bin/ln -sf /proc/self/fd /dev/fd 2>/dev/null -si6::sysinit:/bin/ln -sf /proc/self/fd/0 /dev/stdin 2>/dev/null -si7::sysinit:/bin/ln -sf /proc/self/fd/1 /dev/stdout 2>/dev/null -si8::sysinit:/bin/ln -sf /proc/self/fd/2 /dev/stderr 2>/dev/null -si9::sysinit:/bin/hostname -F /etc/hostname
package/sysvinit/inittab si0::sysinit:/bin/mount -t proc proc /proc · line 20
Quoted 10 lines
package/sysvinit/inittab +7
-0+si4::sysinit:/bin/mkdir -p /run/lock/subsys +si5::sysinit:/sbin/swapon -a +si6::sysinit:/bin/ln -sf /proc/self/fd /dev/fd 2>/dev/null +si7::sysinit:/bin/ln -sf /proc/self/fd/0 /dev/stdin 2>/dev/null +si8::sysinit:/bin/ln -sf /proc/self/fd/1 /dev/stdout 2>/dev/null +si9::sysinit:/bin/ln -sf /proc/self/fd/2 /dev/stderr 2>/dev/null +si10::sysinit:/bin/hostname -F /etc/hostname rcS:12345:wait:/etc/init.d/rcS # S0:1:respawn:/sbin/getty -L ttyS0 115200 vt100 # GENERIC_SERIAL
_______________________________________________
buildroot mailing list
[email protected]
https://lists.buildroot.org/mailman/listinfo/buildroot
Hello Norbert,
Quoted 3 lines · On Mon, 5 Oct 2020 00:41:36 +0200 Norbert Lange <[email protected]> wrote:
Linux distros changed
/var/runand/var/lockto reside on an tmpfs in /run a long time ago, and buildroot seems to agree by providing this tmpfs on all supported init systems.
I am not 100% sure but it seems like this patch has potentially broken the build of vtun, which now fails with:
/usr/bin/install: cannot create directory '/home/giuliobenetti/autobuild/run/instance-0/output-1/target/var/lock': File exists
If you look at http://autobuild.buildroot.net/?reason=vtun%, vtun version 3.0.4 restarted failing to build on January 10, with this error.
Could you have a look ?
Best regards,
Thomas
Signature · 7 lines
Thomas Petazzoni, CTO, Bootlin Embedded Linux and Kernel engineering https://bootlin.com _______________________________________________ buildroot mailing list [email protected] https://lists.buildroot.org/mailman/listinfo/buildroot
Thomas Petazzoni <[email protected]> schrieb am Mi., 12. Jan. 2022, 09:14:
Quoted 21 lines
Hello Norbert,
Quoted 3 lines · On Mon, 5 Oct 2020 00:41:36 +0200 Norbert Lange <[email protected]> wrote:
Linux distros changed
/var/runand/var/lockto reside on an tmpfs in /run a long time ago, and buildroot seems to agree by providing this tmpfs on all supported init systems.I am not 100% sure but it seems like this patch has potentially broken the build of vtun, which now fails with:
/usr/bin/install: cannot create directory
'/home/giuliobenetti/autobuild/run/instance-0/output-1/target/var/lock':
File existsIf you look at http://autobuild.buildroot.net/?reason=vtun%, vtun version 3.0.4 restarted failing to build on January 10, with this error.
Could you have a look ?
Quoted 7 lines
Best regards,
Thomas
--
Thomas Petazzoni, CTO, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com
Seems like the issue is, that run/lock is not existing when installing. Anything installed there will be gone (already before this commit), so it's kinda pointless to do so.
I don't know how many packages have this issue, imho it would be better to fix them instead of making a run/lock directory for broken packages (that will get removed later)
Norbert
_______________________________________________
buildroot mailing list
[email protected]
https://lists.buildroot.org/mailman/listinfo/buildroot
Quoted 36 lines · Norbert Lange
Thomas Petazzoni <[email protected] <mailto:[email protected]>> schrieb am Mi., 12. Jan. 2022, 09:14:
Hello Norbert,
On Mon, 5 Oct 2020 00:41:36 +0200
Norbert Lange <[email protected] <mailto:[email protected]>> wrote:> Linux distros changed
/var/runand/var/lockto reside on an tmpfs in > /run a long time ago, and buildroot seems to agree by providing this > tmpfs on all supported init systems.I am not 100% sure but it seems like this patch has potentially broken the build of vtun, which now fails with:
/usr/bin/install: cannot create directory
'/home/giuliobenetti/autobuild/run/instance-0/output-1/target/var/lock':
File existsIf you look at http://autobuild.buildroot.net/?reason=vtun%
<http://autobuild.buildroot.net/?reason=vtun%>, vtun
version 3.0.4 restarted failing to build on January 10, with this error.Could you have a look ?
Best regards,
Thomas
--
Thomas Petazzoni, CTO, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com <https://bootlin.com>
Quoted 2 lines
Seems like the issue is, that run/lock is not existing when installing.
Anything installed there will be gone (already before this commit), so it's kinda pointless to do so.
Ah, yes, there are many packages that require a directory in a tmpfs and that
create it at build time (because they dont "know" that it's going to be a tmpfs) and at runtime (in the init script, through tmpfiles, or by the application itself).
Quoted 1 line
I don't know how many packages have this issue, imho it would be better to fix them instead of making a run/lock directory for broken packages (that will get removed later)
We're not going to patch packages just for that, because unavoidably some
upstreams are not going to be receptive. Creating the directory is a much simpler solution.
Regards,
Arnout
_______________________________________________
buildroot mailing list
[email protected]
https://lists.buildroot.org/mailman/listinfo/buildroot