Bug 461756 - Crash when I switch to another display using a KVM switch
Crash when I switch to another display using a KVM switch
Product: Fedora
Classification: Fedora
Component: xorg-x11-drv-evdev (Show other bugs)
All Linux
medium Severity medium
: ---
: ---
Assigned To: Peter Hutterer
Fedora Extras Quality Assurance
Depends On:
Blocks: F10Beta/F10BetaBlocker
  Show dependency treegraph
Reported: 2008-09-10 08:51 EDT by Caolan McNamara
Modified: 2013-01-09 23:47 EST (History)
2 users (show)

See Also:
Fixed In Version:
Doc Type: Bug Fix
Doc Text:
Story Points: ---
Clone Of:
Last Closed: 2008-09-15 13:59:24 EDT
Type: ---
Regression: ---
Mount Type: ---
Documentation: ---
Verified Versions:
Category: ---
oVirt Team: ---
RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: ---

Attachments (Terms of Use)
xorg.conf (428 bytes, text/plain)
2008-09-10 08:51 EDT, Caolan McNamara
no flags Details
Xorg.0.conf before switching (216.75 KB, text/plain)
2008-09-10 08:52 EDT, Caolan McNamara
no flags Details
Xorg.0.conf.old after switching back (223.76 KB, text/plain)
2008-09-10 08:53 EDT, Caolan McNamara
no flags Details

  None (edit)
Description Caolan McNamara 2008-09-10 08:51:24 EDT
Created attachment 316305 [details]

Description of problem:

Version-Release number of selected component (if applicable):

How reproducible:

Steps to Reproduce:
1. Stick a usb KVM switch into a rawhide x86_64 box
2. Switch to another display, switch back
3. X has exited and restarted
Additional info:

I imagine that the same effect might be seen if a usb keyboard is simply unplugged, without any KVM in the loop

X driver itself doesn't matter, logs here will be with nouveau, but happens with vesa too
Comment 1 Caolan McNamara 2008-09-10 08:52:30 EDT
Created attachment 316308 [details]
Xorg.0.conf before switching
Comment 2 Caolan McNamara 2008-09-10 08:53:07 EDT
Created attachment 316310 [details]
Xorg.0.conf.old after switching back
Comment 3 Caolan McNamara 2008-09-10 08:54:06 EDT
Presumably this is the magic bit...

(II) UnloadModule: "evdev"

0: /usr/bin/Xorg(xf86SigHandler+0x65) [0x47a105]
1: /lib64/libc.so.6 [0x3538a33130]
2: /usr/lib64/xorg/modules/input//evdev_drv.so [0x7f5bbee9bd00]
3: /usr/bin/Xorg [0x4e4498]
4: /usr/bin/Xorg(WaitForSomething+0x56b) [0x4e4aeb]
5: /usr/bin/Xorg(Dispatch+0x7f) [0x4465ff]
6: /usr/bin/Xorg(main+0x45d) [0x42ccbd]
7: /lib64/libc.so.6(__libc_start_main+0xe6) [0x3538a1e566]
8: /usr/bin/Xorg [0x42c099]

I guess on switch away evdev gets unloaded causing the trouble
Comment 4 Caolan McNamara 2008-09-10 08:55:18 EDT
The F-9 box switched-to survives fine, and all is well if I drop to a vt on the rawhide box, switch, switch-back, and restore the X vt
Comment 5 Peter Hutterer 2008-09-11 07:51:57 EDT
I've seen this before, but I haven't been able to reproduce it. There must be some magic trigger. Funnily enough, I have a mouse that dies about every second and gets removed and re-added and I have yet to trigger this.

What I think happens is that the event is processed after the mouse has been deleted + freed already. How that happens is unclear as of yet.
Comment 6 Caolan McNamara 2008-09-11 08:20:56 EDT
Here's how I can reproduce this.

At gdm hold down a key on the usb keyboard, e.g. a and pull out the usb mouse, stick the mouse back end, and release the key. Wait a few secs, now hold down the same key again and pull out the mouse. ka-pow, death
Comment 7 Jesse Keating 2008-09-12 15:26:01 EDT
I can produce this easily by having both the keyboard and mouse go away.  I've reproduced it both during installs (keyboard/mouse fed by usb) and on my laptop.  During installs I kvm switch away.  On my laptop I just shut off the LCD monitor, which is acting as a USB hub for my keyboard/mouse.

Since this is very easily triggered, I'm marking it a beta blocker.
Comment 8 Peter Hutterer 2008-09-13 06:15:28 EDT
My fault. I did patch it, but being offline it never went into CVS. 

Please try xorg-x11-drv-evdev-2.0.4-3, ajax committed the fix for it.
Comment 9 Adam Jackson 2008-09-15 13:59:24 EDT
Believe this is fixed in 2.0.4-3.  Closing.  Please reopen if this doesn't fix it for you.

Note You need to log in before you can comment on or make changes to this bug.