Bug 16457

this might be actually the same as bug #16458 - the shutdown is clean if no xkb data could be loaded.

On Sat, Jun 21, 2008 at 10:08:20AM -0700, bugzilla-daemon@freedesktop.org wrote:
> anyway, seriously. when i switch to a text console and ctrl-c the server (or
> let kdm kill it), everything is fine. but if i zap it while it is active, it
> wedges in some unholy state. i *think* the same happens when the server is
> reset when the session exits cleanly, but i might have lost track between the
> ~5 reboots and ~15 server restarts ...

this may indicate a crash while shutting down devices (if the bug is input
releated)

> get the log from http://bugs.freedesktop.org/attachment.cgi?id=17282

do you have a log with a backtrace?

I _think_ i hit something like this, but only after 

7c5ca85a9e6d49ab572831b3e0c08bb4cafb395e is first bad commit                                                       
commit 7c5ca85a9e6d49ab572831b3e0c08bb4cafb395e                                                                    
Author: Zhenyu Wang <zhenyu.z.wang@intel.com>                                                                      
Date:   Wed Sep 3 09:41:00 2008 +0800                                                                              
                                                                                                                   
    randr: fix user mode create initial reference count                                                            
                                                                                                                   
    Don't need extra reference count adding when creating user mode.                                               
    This fixes user mode destroy, otherwise we get BadAccess error.                                                
                                                                                                                   
:040000 040000 2b080beb108874742822f2586a70a85dff4ac34e 20e36acbca5981f9f39ab403a5643d9427b54a6e M      randr

in xserver master.

Symptoms - after ctral-alt-backspace salute screen go black (monitor still thinks it get signal - green light is on), but i can reboot/poweroff machine properly. Switching to another vt and kill X manually give me usable console.


under gdb (gdb -c core /usr/X11/bin/X -> bt full) it looks like 

  #0  0xb7a96c20 in malloc_consolidate () from /lib/libc.so.6
No symbol table info available.
#1  0xb7a98d3a in _int_malloc () from /lib/libc.so.6
No symbol table info available.
#2  0xb7a9ab2d in malloc () from /lib/libc.so.6
No symbol table info available.
#3  0xb7977d05 in drmMalloc (size=2076) at xf86drm.c:146
	pt = <value optimized out>
#4  0xb79798ee in drmHashCreate () at xf86drmHash.c:162
	table = <value optimized out>
	i = <value optimized out>
#5  0xb7978377 in drmGetEntry (fd=12) at xf86drm.c:210
	key = 57856
	value = <value optimized out>
	entry = (drmHashEntry *) 0x9abf780
#6  0xb79a3652 in drmRemoveSIGIOHandler (fd=12) at dri.c:2477
	entry = <value optimized out>
#7  0xb79a6575 in DRICloseScreen (pScreen=0x9a95ce8) at dri.c:679
	pDRIPriv = (DRIScreenPrivPtr) 0x9abdfd8
	pDRIInfo = (DRIInfoPtr) 0x9abdeb0
	reserved = <value optimized out>
	reserved_count = <value optimized out>
	pScrn = <value optimized out>
	pDRIEntPriv = (DRIEntPrivPtr) 0x9a92d30
	closeMaster = <value optimized out>
#8  0xb7948699 in RADEONDRICloseScreen (pScreen=0x9a95ce8) at radeon_dri.c:1851
	pScrn = <value optimized out>
	info = (RADEONInfoPtr) 0x9a89d78
	drmInfo = {func = RADEON_CLEANUP_CP, sarea_priv_offset = 0, 
  is_pci = 0, cp_mode = 0, gart_size = 0, ring_size = 0, usec_timeout = 0, 
  fb_bpp = 0, front_offset = 0, front_pitch = 0, back_offset = 0, 
  back_pitch = 0, depth_bpp = 0, depth_offset = 0, depth_pitch = 0, 
  fb_offset = 0, mmio_offset = 0, ring_offset = 0, ring_rptr_offset = 0, 
  buffers_offset = 0, gart_textures_offset = 0}
#9  0xb7948848 in RADEONDRIDoCloseScreen (scrnIndex=0, pScreen=0x9a95ce8)
    at radeon_dri.c:1615
	info = (RADEONInfoPtr) 0x9a89d78
#10 0xb7806fdf in exaXorgCloseScreen (i=0, pScreen=0x9a95ce8) at examodule.c:85
No locals.
#11 0xb78083e6 in exaCloseScreen (i=0, pScreen=0x9a95ce8) at exa.c:758
	pExaScr = (ExaScreenPrivPtr) 0x9abfca0
	ps = (PictureScreenPtr) 0x9abf658
#12 0x080c03db in DPMSClose (i=0, pScreen=0x9a95ce8) at xf86DPMS.c:138
	pDPMS = <value optimized out>
#13 0x0810c2dc in miDCCloseScreen (index=0, pScreen=0x9a95ce8)
    at midispcur.c:175
No locals.
#14 0x081755a4 in damageCloseScreen (i=0, pScreen=0x9a95ce8) at damage.c:1803
No locals.
#15 0x0811884c in miPointerCloseScreen (index=0, pScreen=0x9a95ce8)
    at mipointer.c:162
No locals.
#16 0x0811f6fc in miSpriteCloseScreen (i=0, pScreen=0x9a95ce8)
    at misprite.c:320
No locals.
#17 0x080bf737 in DGACloseScreen (i=0, pScreen=0x9a95ce8) at xf86DGA.c:268
	pScreenPriv = (DGAScreenPtr) 0x9abf890
#18 0xb7a08f94 in XvCloseScreen (ii=0, pScreen=0x9a95ce8) at xvmain.c:317
No locals.
#19 0xb791e785 in RADEONCloseScreen (scrnIndex=0, pScreen=0x9a95ce8)
    at radeon_driver.c:5643
	pScrn = (ScrnInfoPtr) 0x9a89810
	info = (RADEONInfoPtr) 0x9a89d78
#20 0x0816b2be in RRCloseScreen (i=0, pScreen=0x9a95ce8) at randr.c:105
	pScrPriv = (rrScrPrivPtr) 0x9acd248
	j = <value optimized out>
#21 0x080c823e in CMapCloseScreen (i=0, pScreen=0x9a95ce8) at xf86cmap.c:230
No locals.
#22 0x080c4e63 in VidModeClose (i=0, pScreen=0x9a95ce8) at xf86VidMode.c:116
No locals.
#23 0x08144d2d in CursorCloseScreen (index=0, pScreen=0x9a95ce8)
    at cursor.c:201
	cs = (CursorScreenPtr) 0x9adaa98
	ret = <value optimized out>
#24 0x081698ac in AnimCurCloseScreen (index=0, pScreen=0x9a95ce8)
    at animcur.c:136
	ret = <value optimized out>
#25 0x0813ea9b in compCloseScreen (index=0, pScreen=0x9a95ce8) at compinit.c:84
	ret = 2031616
#26 0xb79e4e0a in glxCloseScreen (index=0, pScreen=0x9a95ce8)
    at glxscreens.c:222
	pGlxScreen = <value optimized out>
#27 0x0806c66c in main (argc=1, argv=0xbf8fd924, envp=
Cannot access memory at address 0xb811
) at main.c:422
	i = 0
	xauthfile = <value optimized out>
	alwaysCheckForInput = {0, 1}

and i will attach typical x log without NoTrapSignals option ...

Playing with AIGLX off and DRI 0 (for radeon) give me too confusing results, probably others (unmasked now) bugs.

Created attachment 18682 [details]
typical crash log.


at the end:

Backtrace:
0: /usr/X11/bin/X(xf86SigHandler+0x79) [0x80c0b09]
1: [0xffffe400]
2: /lib/libc.so.6 [0xb7bf8d3a]
3: /lib/libc.so.6(__libc_malloc+0x8d) [0xb7bfab2d]
4: /usr/X11R7/lib/libdrm.so.2(drmMalloc+0x25) [0xb7ad7d05]
5: /usr/X11R7/lib/libdrm.so.2(drmHashCreate+0x1e) [0xb7ad98ee]
6: /usr/X11R7/lib/libdrm.so.2(drmGetEntry+0xa7) [0xb7ad8377]
7: /usr/X11R7/lib/xorg/modules/extensions//libdri.so(drmRemoveSIGIOHandler+0x22) [0xb7b03652]
8: /usr/X11R7/lib/xorg/modules/extensions//libdri.so(DRICloseScreen+0x325) [0xb7b06575]
9: /usr/X11R7/lib/xorg/modules/drivers//nouveau_drv.so(NVDRICloseScreen+0x3b) [0xb7a9529b]
10: /usr/X11R7/lib/xorg/modules/drivers//nouveau_drv.so [0xb7a9a840]
11: /usr/X11/bin/X [0x80c4e63]
12: /usr/X11/bin/X [0x8144d2d]
13: /usr/X11/bin/X [0x81698ac]
14: /usr/X11/bin/X [0x813ea9b]
15: /usr/X11R7/lib/xorg/modules/extensions//libglx.so [0xb7b44e0a]
16: /usr/X11/bin/X(main+0x47c) [0x806c66c]
17: /lib/libc.so.6(__libc_start_main+0xe0) [0xb7ba3390]
18: /usr/X11/bin/X [0x806ba61]


very similar backtrace was from radeon driver. Vesa and radeon with DRI off has different backtraces.

All this on x86, slackware-current, gcc-4.3.1 for X, gcc 4.2.3 (default) for kernel and other bits. 

xserver - bf333c2f9833a178887e7bdd7fc338f1e09c387f
drm - 99f8cce3eae79c9cf5e9897ef1af8eac24b38c09
nouveau - 6dd8ad4ed3f89c6ed9e61708a587d417c8d984fe (+ suspend patch)
radeon - 38aa450fce6a91b35c191fd07112490a62cc29c6

haven't seen this for quite some time, so ...