FS#48468 - [linux-grsec] X not accepting connections/frozen when running startx

Attached to Project: Community Packages
Opened by Plague (centuryplague) - Sunday, 06 March 2016, 02:32 GMT
Last edited by Daniel Micay (thestinger) - Thursday, 10 March 2016, 15:26 GMT
Task Type Bug Report
Category Upstream Bugs
Status Closed
Assigned To Daniel Micay (thestinger)
Architecture x86_64
Severity High
Priority Normal
Reported Version
Due in Version Undecided
Due Date Undecided
Percent Complete 0%
Votes 1
Private No

Details

Description:
On a system booted to multi-user command line with linux-grsec, running startx (as any user including root) fails to launch X and instead gives:
"waiting for X server to begin accepting connections...
...
..."
etc.
which waits forever. startx produces no Xorg.0.log file nor other logs and appears simply locked up.

On the same system, regular kernel is fine running startx.

This problem has begun due to updates within the last week only and has never occurred on this system prior. Affected system is using 4.4.4.201603032158-1-grsec. There were no issues when the kernel was 4.4.2.201602182048-1-grsec, but I don't yet know if this is the only package involved. Only change was package updates.


This task depends upon

Closed by  Daniel Micay (thestinger)
Thursday, 10 March 2016, 15:26 GMT
Reason for closing:  Upstream
Comment by Plague (centuryplague) - Sunday, 06 March 2016, 02:55 GMT
Appears to depend on system, does not happen in qemu VM.
Comment by Plague (centuryplague) - Sunday, 06 March 2016, 03:04 GMT
Note the regular kernel tested was the latest available in Arch, linux 4.4.3-1, which is behind the grsec kernel (4.4.4)...
Comment by Plague (centuryplague) - Sunday, 06 March 2016, 04:57 GMT
I can confirm on this system downgrading to 4.4.2.201602182048-1-grsec with no other package changes allows startx to start.
Comment by Jens Adam (byte) - Sunday, 06 March 2016, 21:56 GMT
Are any external (nvidia?) or -dkms module packages involved? What's your graphics adapter?
Comment by Plague (centuryplague) - Monday, 07 March 2016, 00:14 GMT
It's built as live cd, all major [free] xf86-video-* are included. The system is radeon only (xf86-video-ati), no xorg conf except Arch defaults. As far as I can tell the only video that was updated in last week was xf86-video-intel, but this system has no intel graphics.

I'm going to try a build without the extra xf86-video-* tomorrow (this takes forever).

There's one guy on the forum who reported a "waiting for X server to begin accepting connections" error in the last week, but is not using grsec [he is using kernel 4.1 lts]. Could be something else triggered by this kernel update.
Comment by Plague (centuryplague) - Monday, 07 March 2016, 04:21 GMT
I did one tonight, rebuilding with all latest package but removing unneeded intel video had no effect, startx still froze.

However now appears radeon/ati-related. Booted with radeon blacklisted ("radeon.blacklist=yes modprobe.blacklist=radeon") and startx successfully loaded with VESA driver xf86-video-vesa.

Something from kernel update triggering ATI failure, sounding closer to what forum thread reported https://bbs.archlinux.org/viewtopic.php?id=209576. However can't reconcile the kernel difference (linux-lts last update middle february).
Comment by Plague (centuryplague) - Monday, 07 March 2016, 06:10 GMT
Can now confirm same issue happening on linux-lts 4.1.18-1. Same bug as user on forum.

This is NOT grsec-specific.

Since plain linux 4.4.3-1 package from stable is working, some change in 4.4.4 must be triggering this, something shared with LTS 4.1.18.

(Other packages that changed on this system in the last week: xf86-video-intel, pciutils, upower, udisks2, gtk2, btrfs-progs... that's it)
Comment by Plague (centuryplague) - Monday, 07 March 2016, 06:57 GMT
I submitted a bug to upstream: https://bugzilla.kernel.org/show_bug.cgi?id=113861

I don't know but odds are pointing there (changelogs for 4.4.4 and 4.1.18 full of radeon). No more time.
Comment by Plague (centuryplague) - Tuesday, 08 March 2016, 19:09 GMT
Upstream found the probable issue and I confirm it fixes it on my system: https://bugzilla.kernel.org/show_bug.cgi?id=113861#c9

Had to recompile linux-grsec with the attached patch applied in reverse (patch -p1 -R -i "${srcdir}/0001-drm-radeon-call-hpd_irq_event-on-resume.patch").

It is already reverted in 4.5rcX mainline so I assume will come in next 4.4 patch(es)...

I did not confirm for linux-lts but it's probably the same.
Comment by Daniel Micay (thestinger) - Thursday, 10 March 2016, 15:26 GMT
If it's an upstream problem, I'll just wait until it's fixed by an upstream release.

Loading...