lightdm causes all greeters to fail to start

Realized that it's likely unity-greeter because I'm seeing a lot of...

[25886.110278] traps: unity-greeter[820] trap int3 ip:7fb856a35241 sp:7fff145ad0f0 error:0 in libglib-2.0.so.0.5000.0[7fb8569e5000+110000]

I've got libglib2.0 2.50.2-2ubuntu1 so I tried to roll back to 2.50.0-1 but that didn't help either. libglib-2.0 was something that was upgraded in between my last reboot.

Here's the backtrace I'm getting out of unity-greeter. Unfortunately there's no debug symbols available for it to find what's at the bottom of the stack trace.

#0 _g_log_abort (breakpoint=breakpoint@entry=1) at ././glib/gmessages.c:487
#1 0x00007ffff5c8a64d in g_logv (log_domain=0x7ffff5ccbd6e "GLib",
    log_level=G_LOG_LEVEL_ERROR, format=<optimized out>,
    args=args@entry=0x7fffffffd7c0) at ././glib/gmessages.c:1296
#2 0x00007ffff5c8a7af in g_log (
    log_domain=log_domain@entry=0x7ffff5ccbd6e "GLib",
    log_level=log_level@entry=G_LOG_LEVEL_ERROR,
    format=format@entry=0x7ffff5cd56c8 "%s: failed to allocate %lu bytes")
    at ././glib/gmessages.c:1337
#3 0x00007ffff5c88ed1 in g_realloc (mem=<optimized out>, n_bytes=131072)
    at ././glib/gmem.c:164
#4 0x00007ffff5c56869 in g_array_maybe_expand (
    array=array@entry=0x55555590aa40, len=len@entry=40)
    at ././glib/garray.c:779
#5 0x00007ffff5c56be8 in g_array_append_vals (farray=0x55555590aa40,
    data=0x7fffffffd9a0, len=40) at ././glib/garray.c:418
#6 0x00007ffff5c57d09 in g_byte_array_append (
    array=array@entry=0x55555590aa40, data=data@entry=0x7fffffffd9a0 "",
    len=len@entry=40) at ././glib/garray.c:1713
#7 0x00007ffff6f57d34 in alloc_tree (offset=<synthetic pointer>,
    array=0x55555590aa40) at ././gtk/gtkcssselector.c:1726
#8 subdivide_infos (array=array@entry=0x55555590aa40,
    infos=infos@entry=0x555555968020, parent_offset=parent_offset@entry=65392)
    at ././gtk/gtkcssselector.c:1779
#9 0x00007ffff6f57f73 in subdivide_infos (array=array@entry=0x55555590aa40,
    infos=infos@entry=0x555555967f20, parent_offset=parent_offset@entry=65392)
    at ././gtk/gtkcssselector.c:1824
#10 0x00007ffff6f57f5c in subdivide_infos (array=array@entry=0x55555590aa40,
    infos=infos@entry=0x555555967ec0, parent_offset=parent_offset@entry=65352)
    at ././gtk/gtkcssselector.c:1821
#11 0x00007ffff6f57f5c in subdivide_infos (array=array@entry=0x55555590aa40,
    infos=infos@entry=0x555555968100, parent_offset=parent_offset@entry=65312)
    at ././gtk/gtkcssselector.c:1821
#12 0x00007ffff6f57f5c in subdivide_infos (array=array@entry=0x55555590aa40,
    infos=infos@entry=0x555555967a20, parent_offset=parent_offset@entry=62904)
    at ././gtk/gtkcssselector.c:1821
#13 0x00007ffff6f57f73 in subdivide_infos (array=array@entry=0x55555590aa40,
    infos=infos@entry=0x555555967740, parent_offset=parent_offset@entry=62904)
    at ././gtk/gtkcssselector.c:1824
#14 0x00007ffff6f57f73 in subdivide_infos (array=array@entry=0x55555590aa40,
    infos=infos@entry=0x555555966b20, parent_offset=parent_offset@entry=62904)
    at ././gtk/gtkcssselector.c:1824
#15 0x00007ffff6f57f5c in subdivide_infos (array=array@entry=0x55555590aa40,
    infos=infos@entry=0x555555961b80,
    parent_offset=parent_offset@entry=2147483647)
    at ././gtk/gtkcssselector.c:1821
#16 0x00007ffff6f57f73 in subdivide_infos (array=array@entry=0x55555590aa40,
    infos=infos@entry=0x555555959a20,
    parent_offset=parent_offset@entry=2147483647)
    at ././gtk/gtkcssselector.c:1824
#17 0x00007ffff6f57f73 in subdivide_infos (array=array@entry=0x55555590aa40,
    infos=infos@entry=0x555555953020,
    parent_offset=par...

I've tried lightdm-gtk-greeter and kylin-greeter and the same crash occurs. So I'm back to thinking this issue is with lightdm. Some data or configuration its passing to the greeter is causing it to fail.

I've removed all the dot directories (.cache, .config, .dbus, .gconf, and .local) inside of /var/lib/lightdm and still am experiencing this.

Does the lightdm.log say anything useful? The stack trace looks like an out of memory condition, which could be either liblightdm or GTK+ failing to work? I'd suggest trying different versions of those libraries.

The lightdm.log is very repetitive so I just snipped out the last run since they all appear to be identical.

Thanks for the response Robert. I'll give both a shot.

Unfortunately that didn't work. I went back to the versions that 16.10 originally shipped with and its still broken. I'm guessing some data file that gets generated got corrupted and its causing this but I'm unable to find it. I've taken an strace of lightdm-gtk-greeter showing the failure to help track this down

Just prior to the SIGTRAP I see a mmap() fail with EAGAIN and a failing ioctl() call. This is from the attached strace with the other noise trimmed out.

4108 mmap(NULL, 8392704, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS|MAP_ST
ACK, -1, 0 <unfinished ...>
4108 <... mmap resumed> ) = -1 EAGAIN (Resource temporarily unavai
lable)
4108 ioctl(2, TCGETS <unfinished ...>
4108 <... ioctl resumed> , 0x7ffcbb763220) = -1 ENOTTY (Inappropriate ioctl for
 device)
4108 write(2, "\n(lightdm-gtk-greeter:4108): GLi"..., 125 <unfinished ...>
4108 <... write resumed> ) = 125
4108 --- SIGTRAP {si_signo=SIGTRAP, si_code=SI_KERNEL} ---

Well the ioctl() call isn't really an issue because stderr isn't hooked to a tty in the way that a greeter is started up so that's normal.

The EAGAIN is interesting because I recently installed a package dealing with hugepages and it created some rules in /etc/security/limits.d

Sure enough removing the 'hugepages' package has solved the issue.

Please reopen. This just bit me on 17.10, using some manually configured soft and hard memlock limits in /etc/security/limits.conf:

* soft memlock 262144
* hard memlock 262144

As these limits *increase* max locked memory from 64 to 262144 kB, it took me quite some time (and a lucky google hit on this bug report) to locate this as the cause.

This is BTW almost exactly the same what 'hugepages' does -- *increasing* memlock limits to 262144 kB.

Seen on 18.04.1. WorkARound?

* soft memlock 262144
* hard memlock 262144
# End of file

Does setting:
[LightDM]
lock-memory=false

Workaroud this issue?

(in lightdm.conf)

Hi, a recent systemd update in 18.04 makes slick-greeter segfault.
So all Ubuntu MATE 18.04 users now get black screens instead of lightdm.

It's related to memory limits so I'm cross-referencing it here:
https://bugs.launchpad.net/ubuntu/+source/systemd/+bug/1830746

This didn't help:

[LightDM]
lock-memory=false

Can I put something in /etc/security/limits.conf to see if it helps?

What torel proposed in https://bugs.launchpad.net/ubuntu/+source/unity-greeter/+bug/1662244/comments/14 avoids the segfault:

* soft memlock 262144
* hard memlock 262144

Should all lightdm users manually put that in limits.conf, or should we expect some update?

Thanks

Adding

* soft memlock 262144
* hard memlock 262144

in /etc/security/limits.conf

solved the same problem with lightdm and X not starting on Ubuntu 18.04 running on a motherboard with VIA KM400 integrated graphics.

Upgrading to (K)Ubuntu 18.04.5 LTS basically broke the system as X was not restarting correctly (black screen instead of lightdm display or even infinite loop of X crashes based on BIOS/EFI config).

This is critical for non-savvy users as my parents.

The issue is the same as can be seen in /var/log/lightdm/seat0-greeter.log with repeated messages like

**MEMORY-ERROR***: unity-greeter[1647]: GSlice: failed to allocate 2040 bytes (alignment: 2048)...
**MEMORY-ERROR***: unity-greeter[1647]: GSlice: failed to allocate 2040 bytes (alignment: 2048)...
**MEMORY-ERROR***: unity-greeter[1647]: GSlice: failed to allocate 2040 bytes (alignment: 2048)...

The workaround is indded updating the memlock limits as described above (adding the 2 lines of the above comment to /etc/security/limits.con).

Thanks for the workaround :)