GSOC 2013 - Debian GNU/Hurd Debianish initialization

Open issues

Issues I dealt with, patches pending

• Make sysvinit pid 1: http://lists.gnu.org/archive/html/bug-hurd/2013-08/msg00025.html

• glibc fixes: http://lists.gnu.org/archive/html/bug-hurd/2013-08/msg00046.html http://sourceware.org/ml/libc-alpha/2013-08/msg00205.html

• First round of patches for the sysvinit package http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=721917

• No /proc/stat.

  • Actually it's there. I got an error message because I removed the passive translator from /proc. This should also be an issue on Linux, I bet it just does not happen there because /proc is mounted in the initrd. Needs further investigation.
  • http://lists.gnu.org/archive/html/bug-hurd/2013-09/msg00208.html

• libc: _PATH_MOUNTED is wrong or at least different.

  • Linux:

    % /bin/echo -e '#include <paths.h>\n_PATH_MOUNTED' | gcc -E -x c - | grep -v '#'
    "/etc/mtab"
    

  • Hurd:

    % /bin/echo -e '#include <paths.h>\n_PATH_MOUNTED' | gcc -E -x c - | grep -v '#'
    "/var/run/mtab"
    

  • http://lists.gnu.org/archive/html/bug-hurd/2013-08/msg00047.html

Resolved issues

• Hurds reboot/halt lacks some command line flags:

  • -w Don't actually reboot or halt but only write the wtmp record.
  • -d: Don't write the wtmp record.
  • This is trivial. Once we use sysvinit, we use the ones from that package.

• /hurd/tmpfs does not properly implement fsys_set_options. This is trivial:

  mount: cannot remount /run: Invalid argument
  mount: cannot remount /run/lock: Invalid argument
  mount: cannot remount /run/shm: Invalid argument
  

  • Patch submitted: http://lists.gnu.org/archive/html/bug-hurd/2013-06/msg00126.html
  • Patch merged upstream.

• Same with /hurd/procfs:

  mount: cannot remount /proc: Invalid argument
  

  • Patches submitted: http://lists.gnu.org/archive/html/bug-hurd/2013-06/msg00119.html
  • Patch added upstream and to Debians hurd package.

• Hurds mount lacks some command line flags:

  • -n, --no-mtab

    • Trivial patch as we do not write a mtab file.
    • Patch added to Debians hurd package.

  • -f, --fake

  • -oremount does not work with just one argument (e.g. mount -oremount,... /foo)

    • Patch added to Debians hurd package.

  • mount is confused

    • I believe I figured this one out, file_name_lookup_carefully does not work correctly if it encounters an acticve translator on a node that has no passive translator. Patch is prepared.

    • irc log:

      12:27 < pinotree> for example, if you notice the sysvinit-booted system there's
                        only a tmpfs for /run, and not own ones for /run/shm and
                        /run/lock
      12:28 < pinotree> the logic in our mount mistakes as result of /run/lock what
                         really is the result of /run itself
      

    • Patch merged upstream.

  • Patches submitted: http://lists.gnu.org/archive/html/bug-hurd/2013-06/msg00135.html

• Hurds swapon/swapoff lacks some command line flags:

  • -e, --ifexists: patch by pino http://lists.gnu.org/archive/html/bug-hurd/2012-01/msg00060.html
  • Patch merged upstream.

• Hurds mount lacks some command line flags:

  • -O, --test-opts
  • -t, --types: semantic difference to what Linux' and FreeBSDs mount does.

• Hurds mount -t auto does not work.

  • needed for fstab entries with fstype=auto
  • needed for d-i

• The hurd package does not provide an umount helper, needed for d-i.

  • also needed for umountfs, umountnfs.sh
  • required arguments -f, -v, -r, -d
  • umount needs an mtab file

• Patch series for all the issues above posted: http://lists.gnu.org/archive/html/bug-hurd/2013-07/msg00031.html

• Patches merged at the appropriate locations.

• Hurds swapon/swapoff lacks some command line flags:

  • -v, --verbose: set if VERBOSE=yes is given in /etc/default/rcS.
  • Patch posted: http://lists.gnu.org/archive/html/bug-hurd/2013-07/msg00055.html
  • Patches merged.

• procfs should ignore --no{dev,exec,suid}: http://lists.gnu.org/archive/html/bug-hurd/2013-07/msg00076.html

  • Patch upstream.

• Cleanups for mig: http://lists.gnu.org/archive/html/bug-hurd/2013-08/msg00055.html

  • Patches upstream.

• libnetfs does not respond properly to file_get_translator requests on nodes with a passive translator record: http://lists.gnu.org/archive/html/bug-hurd/2013-07/msg00148.html

  • Merged upstream.

• There was no clean way to start the console-client. I added daemonizing support to the console-client and created an appropriate initscript: http://lists.gnu.org/archive/html/bug-hurd/2013-07/msg00165.html

  Updated patch series: http://lists.gnu.org/archive/html/bug-hurd/2013-07/msg00246.html

  • Merged upstream.

• Fixes for ifupdown: http://lists.gnu.org/archive/html/bug-hurd/2013-08/msg00063.html

  • inetutils-ifconfig: invalid arguments

  • ifconfig: not found on ifdown /dev/eth0:

    Deconfiguring network interfaces...ifdown: failed to open pid file /run/network/ifdown-/dev/eth0.pid: No such file or directory
    Internet Systems Consortium DHCP Client 4.2.2
    [...]
    Sending on   Socket//dev/eth0
    /bin/sh: 1: ifconfig: not found
    

  • ifupdown: Nowadays Hurd uses translators to access NICs (yay userspace drivers :) ) typically bound to /dev/ethX. Using a device name like /dev/ethX in /etc/network/interfaces kinda works, but does not play well with ifupdowns state machinery.

  • http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=720531

  • Merged upstream.

• Fixes for fakeroot-hurd and remap: http://lists.gnu.org/archive/html/bug-hurd/2013-08/msg00064.html

  • Merged upstream.

• umount.patch (v2): http://lists.gnu.org/archive/html/bug-hurd/2013-08/msg00018.html

  • Merged upstream.

• Savannah bug #13398: /dev/{null,zero,full} should not be run as root

  • http://savannah.gnu.org/bugs/?13398
  • http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=313560
  • http://lists.gnu.org/archive/html/bug-hurd/2013-07/msg00251.html
  • Merged upstream.

• As the mtab translator is a translator of its own, there needs to be a way to bind it to /proc/mounts. I figured the easiest way would be to just serve a node with an appropriate passive translator record: http://lists.gnu.org/archive/html/bug-hurd/2013-07/msg00149.html

  • All the infrastructure parts are upstream.
  • Updated patch series: http://lists.gnu.org/archive/html/bug-hurd/2013-07/msg00216.html
  • Merged upstream.

• mtab translator (v4): http://lists.gnu.org/archive/html/bug-hurd/2013-07/msg00259.html

  • Merged upstream.

• Make start_code and end_code available in /proc/*/stat: http://lists.gnu.org/archive/html/bug-hurd/2013-08/msg00017.html

  • Merged upstream.

• Cleanups for the exec server: http://lists.gnu.org/archive/html/bug-hurd/2013-08/msg00049.html

  • Merged upstream.

• Updated patch for mount -t auto that checks for libblkid: http://lists.gnu.org/archive/html/bug-hurd/2013-08/msg00126.html

  • Merged upstream.

• /etc/init.d/{sendsigs/killprocs} kill essential processes

  • http://lists.gnu.org/archive/html/bug-hurd/2006-02/msg00081.html
  • As temporary workaround just disable those scripts: http://lists.gnu.org/archive/html/bug-hurd/2013-07/msg00082.html
  • Prototype published: http://lists.gnu.org/archive/html/bug-hurd/2013-07/msg00226.html
  • Merged upstream.

• /proc/*/stat: startcode and endcode fields are 0, making killall5 think that the process in question is a kernel thread and is therefore not to be killed. So currently, killall5 does not kill any processes on Hurd.

  • Debian/kFreeBSD had the same issue: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=344546
  • http://lists.gnu.org/archive/html/bug-hurd/2013-08/msg00017.html
  • Merged upstream.

Somewhat related issues

• http://savannah.gnu.org/bugs/?17129
• http://savannah.gnu.org/bugs/?15301
• http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=517336

Unrelated issues concerning stability and correctness

• http://savannah.gnu.org/bugs/?19479
• http://savannah.gnu.org/bugs/?29809
• http://savannah.gnu.org/bugs/?11531
• http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=629184
• /hurd/init does not get notified if either the initial /hurd/exec or the root filesystem dies.