Is it possible to statically link lxc 4.0?

LXC
1

Hi,
I get an embedded system ,which has limited ram/disk/core, and running in a 2.6.35 kernel with a 2.11 glib environment. It’s a production environment, a lot legacy app running on it, I can’t upgrade them all.
I am gonna to try an alternative way, to statically link the lxc 4.0, but I met some problem , if I tried to make environment variables like “CFLAGS=”-static -static-libgcc -static-libstdc++" LDFLAGS="-static-libgcc -static-libstdc++ -static" ,but still failed to get them statically linked. I even tried to chang the flags in Makefile hard coded, didn’t work either.
And I also googled some project use libtool had the similar problem, they suggest to hard code cflags in to CC environment variable, but that didn’t work for LXC.
Beside, I have successfully running the lxc on the chroot. LXC is really a nice ligthweight tool for embedded system.

2

@brauner is it possible to statically link lxc? thanks

3

What are the errors you’re seeing? I would expect it’s difficult because
of us making use of libc functions that use nsswitch which is almost
impossible to use in static binaries afair.

Christian

5

Hi Brauner,
Thanks for the info.
Even after I tried fill in the statically CFLAGS/LDFLAGS, after the whole building process,all the apps are still dynamically linked(I am using armv7 gnu libc).
Later I turned to musl, but haven’t tried statically linked yet, for now I am running lxc in the chroot environment which is an alternative way, so far so good.

BTW: I met another strange problem, lxc 4.0.6, kernel 2.6.35 with cgroup patch

  1. Trace log:
    lxc-start x11vnc 20210122095620.193 INFO start - start.c:lxc_init:837 - Container “x11vnc” is initialized
    lxc-start x11vnc 20210122095620.193 WARN cgfsng - cgroups/cgfsng.c:cgroup_tree_create:1168 - No error information - The /sys/fs/cgroup/openrc//lxc.monitor.x11vnc cgroup already existed
    lxc-start x11vnc 20210122095620.193 DEBUG cgfsng - cgroups/cgfsng.c:cgfsng_monitor_create:1355 - Failed to create cgroup “(null)”
    lxc-start x11vnc 20210122095620.193 WARN cgfsng - cgroups/cgfsng.c:cgroup_tree_create:1168 - No error information - The /sys/fs/cgroup/openrc//lxc.monitor.x11vnc-1 cgroup already existed
    lxc-start x11vnc 20210122095620.193 DEBUG cgfsng - cgroups/cgfsng.c:cgfsng_monitor_create:1355 - Failed to create cgroup “(null)”
    lxc-start x11vnc 20210122095620.193 INFO cgfsng - cgroups/cgfsng.c:cgfsng_monitor_create:1368 - The monitor process uses “lxc.monitor.x11vnc-2” as cgroup
    lxc-start x11vnc 20210122095620.194 ERROR cgfsng - cgroups/cgfsng.c:cgfsng_monitor_enter:1504 - Invalid argument - Failed to enter cgroup “/sys/fs/cgroup/openrc//lxc.monitor.x11vnc-2”
  2. I found in kernel 2.6.35 's cgroup’s doc, said the ‘cgroup.procs’ should be a read-only node for now, but function 'cgfsng_monitor_enter ’ is trying to write data to it, that’s why I always failed to enter the monitor group.
  • cgroup.procs: list of tgids in the cgroup. This list is not
    guaranteed to be sorted or free of duplicate tgids, and userspace
    should sort/uniquify the list if this property is required.
    This is a read-only file, for now.

ret = lxc_writeat(h->cgfd_mon, “cgroup.procs”, monitor, monitor_len);
if (ret)
return log_error_errno(false, errno, “Failed to enter cgroup “%s””, h->monitor_full_path);

6

Building the tools lxc-attach, lxc-start etc. completely static is a bit of work but I’ll see if we can support it in a good way.

The cgroup issue I’m very tempted to say that a 2.6.35 kernel is super ancient and will likely be problematic in a bunch of other ways. It’s just too old.

7

Hi Brauner,
That’s right, 2.6.35 is super super old. But the fact is I can’t upgrade the kernel right now , I 'll find a way to do some kernel patch myself.

Anyhow, thanks all the same.

8

I have a branch that makes it possible to compile all binaries statically and it’s quite a hazzle to get this right with autotools:

2 Likes
9

:100:
Much appreciate!