grub-mkconfig hangs

[Daniel Convissor]

Calling "grub-mkconfig -o /boot/grub/grub.cfg" hangs. I came across this when running "apt-get upgrade", during the kernel post install step, where it hangs on this line in the log:

    run-parts: executing /etc/kernel/postinst.d/zz-update-grub 3.19.0-41-generic /boot/vmlinuz-3.19.0-41-generic

I ran strace on calling grub-mkconfig. I'll attach the output, but here's a summary of the output. To my untrained eye, it looks like grub-mkconfig is calling pipe() and read() on a closed handle.

vvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvv
... snip ...
read(3, "/dev/mapper/ubuntu--vg-root\n", 128) = 28
read(3, "", 128) = 0
--- SIGCHLD {si_signo=SIGCHLD, si_code=CLD_EXITED, si_pid=17818, si_status=0, si_utime=0, si_stime=0} ---
rt_sigreturn() = 0
close(3) = 0
wait4(-1, [{WIFEXITED(s) && WEXITSTATUS(s) == 0}], 0, NULL) = 17818
pipe([3, 4]) = 0
clone(child_stack=0, flags=CLONE_CHILD_CLEARTID|CLONE_CHILD_SETTID|SIGCHLD, child_tidptr=0x7fe3556d7a10) = 17819
close(4) = 0
read(3,
        ^
        |
NOTE: THE PROCESS HANGS THERE

NOTE: AFTER KILLING THE PROCESS, THE LINE CONTINUES THERE
        |
        v
read(3, 0x7ffddd6b62e0, 128) = ? ERESTARTSYS (To be restarted if SA_RESTART is set)
--- SIGTERM {si_signo=SIGTERM, si_code=SI_USER, si_pid=16554, si_uid=0} ---
+++ killed by SIGTERM +++
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

Here are the package versions on hand as determined by "apt-cache policy grub grub-pc grub-efi"

vvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvv
grub:
  Installed: (none)
  Candidate: 0.97-29ubuntu66
  Version table:
     0.97-29ubuntu66 0
        500 http://us.archive.ubuntu.com/ubuntu/ trusty/main amd64 Packages
grub-pc:
  Installed: 2.02~beta2-9ubuntu1.6
  Candidate: 2.02~beta2-9ubuntu1.6
  Version table:
 *** 2.02~beta2-9ubuntu1.6 0
        500 http://us.archive.ubuntu.com/ubuntu/ trusty-updates/main amd64 Packages
        500 http://security.ubuntu.com/ubuntu/ trusty-security/main amd64 Packages
        100 /var/lib/dpkg/status
     2.02~beta2-9 0
        500 http://us.archive.ubuntu.com/ubuntu/ trusty/main amd64 Packages
grub-efi:
  Installed: (none)
  Candidate: 2.02~beta2-9ubuntu1.6
  Version table:
     2.02~beta2-9ubuntu1.6 0
        500 http://us.archive.ubuntu.com/ubuntu/ trusty-updates/main amd64 Packages
        500 http://security.ubuntu.com/ubuntu/ trusty-security/main amd64 Packages
     2.02~beta2-9 0
        500 http://us.archive.ubuntu.com/ubuntu/ trusty/main amd64 Packages
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

This is similar to what happened to me in bug #1510563 about a month ago.

[Steve Langasek (vorlon)]

wait4(-1, [{WIFEXITED(s) && WEXITSTATUS(s) == 0}], 0, NULL) = 17818
pipe([3, 4]) = 0
clone(child_stack=0, flags=CLONE_CHILD_CLEARTID|CLONE_CHILD_SETTID|SIGCHLD, child_tidptr=0x7fe3556d7a10) = 17819
close(4) = 0
read(3,

This is the standard usage of pipe. pipe opens two connected file descriptors, and returns these fds in the array that's the argument to pipe. The process then forks, the parent closes one, the child closes the other, and they talk across this pipe.

Since you are only stracing the grub-mkconfig process, and not following forks ('strace -ff'), this trace doesn't show us what's actually hanging.

grub-mkconfig, btw, is a shell script, so it's probably better to debug it using 'sh -x /usr/sbin/grub-mkconfig' instead.

[Daniel Convissor (convissor)]

So I ran

strace -ff -o 2015-12-18-grub-mkconfig-strace-ff sh -x /usr/sbin/grub-mkconfig

That produced 1000 files and took several attempts to kill off the various processes that spawned after killing the hung ones. I doubt anyone wants to see 1000 strace dumps. Please advise.

[Daniel Convissor (convissor)]

This problem persists, as seen while upgrading to linux-image-3.19.0-61-generic.

@vorlorn: please advise how to provide debug information.