Bug 21275

Summary:

ATI/Radeon vt switch dumps core with SWCursor

Product:

xorg

Reporter:

Warren Block <wblock>

Component:

Server/General

Assignee:

Xorg Project Team <xorg-team>

Status:

RESOLVED DUPLICATE

QA Contact:

Xorg Project Team <xorg-team>

Severity:

normal

Priority:

medium

CC:

jim, phelps

Version:

7.4 (2008.09)

Hardware:

Other

OS:

FreeBSD

Whiteboard:

i915 platform:

i915 features:

Attachments:

Description	Flags
xorg.conf	none
Xorg.0.log.old	none

Description Warren Block 2009-04-18 18:39:11 UTC

Created attachment 24930 [details]
xorg.conf

Switching from console mode to X with Option SWCursor enabled, xorg dumps core.
No problem without SWCursor.

Problem is also mentioned in comment #14 here:
https://bugs.freedesktop.org/show_bug.cgi?id=16865

Versions: xorg-server-1.6.0,1; xf86-video-ati-6.12.2; Radeon X1650 PCIE
Platform: FreeBSD 7.2-PRERELEASE i386

gdb:

(gdb) bt
#0  0x34267e17 in kill () from /lib/libc.so.7
#1  0x34184357 in raise () from /lib/libthr.so.3
#2  0x3426698a in abort () from /lib/libc.so.7
#3  0x080bdd03 in ddxGiveUp () at xf86Init.c:1417
#4  0x080bdde0 in AbortDDX () at xf86Init.c:1462
#5  0x08168a1d in AbortServer () at log.c:407
#6  0x08168e18 in FatalError (
    f=0x8224c5c "Caught signal %d.  Server aborting\n") at log.c:532
#7  0x080da643 in xf86SigHandler (signo=11) at xf86Events.c:387
#8  <signal handler called>
#9  0x0810d70f in xf86_reload_cursors (screen=0x34348180) at xf86Cursors.c:615
#10 0x344d70d0 in radeon_crtc_mode_commit (crtc=0x34308000)
    at radeon_crtc.c:288
#11 0x08107055 in xf86CrtcSetModeTransform (crtc=0x34308000, mode=0x343080bc, 
    rotation=1, transform=0x0, x=0, y=0) at xf86Crtc.c:356
#12 0x0810b61c in xf86SetDesiredModes (scrn=0x34306800) at xf86Crtc.c:2497
#13 0x344a9469 in RADEONEnterVT (scrnIndex=0, flags=0) at radeon_driver.c:5663
#14 0x080f498a in xf86XVEnterVT (index=0, flags=0) at xf86xv.c:1228
#15 0x080e2455 in CMapEnterVT (index=0, flags=0) at xf86cmap.c:455
#16 0x080daca0 in xf86VTSwitch () at xf86Events.c:597
#17 0x080da445 in xf86Wakeup (blockData=0x0, err=-1, pReadmask=0x824f640)
    at xf86Events.c:295
#18 0x080931ab in WakeupHandler (result=-1, pReadmask=0x824f640)
    at dixutils.c:418
#19 0x08158c79 in WaitForSomething (pClientsReady=0x46d5d800) at WaitFor.c:231
#20 0x080858ef in Dispatch () at dispatch.c:367
#21 0x0806ba49 in main (argc=4, argv=0xbfbfeee0, envp=0xbfbfeef4)
    at main.c:397
(gdb) frame 9
#9  0x0810d70f in xf86_reload_cursors (screen=0x34348180) at xf86Cursors.c:615
615         if (!cursor_screen_priv->isUp)
(gdb) p cursor_screen_priv->isUp
Cannot access memory at address 0x4

Comment 1 Warren Block 2009-04-18 18:40:26 UTC

Created attachment 24931 [details]
Xorg.0.log.old

Comment 2 Ted Phelps 2009-04-19 05:05:52 UTC

I'm seeing what looks like the same issue with xorg-server-1.6.1, xf86-video-nouveau from git (7597203b) and a software cursor:

  Backtrace:
  0: X(xorg_backtrace+0x26) [0x4e7466]
  1: X(xf86SigHandler+0x39) [0x46eb49]
  2: /lib/libc.so.6 [0x7fcafad16b70]
  3: X(xf86_reload_cursors+0x45) [0x4a5985]
  4: /usr/lib/xorg/modules/drivers//nouveau_drv.so [0x7fcaf9977ac5]
  5: X(xf86CrtcSetModeTransform+0x47d) [0x4a481d]
  6: X(xf86SetSingleMode+0x12f) [0x4a49df]
  7: X(xf86SwitchMode+0x11b) [0x47695b]
  8: X(VidModeSwitchMode+0x6f) [0x47332f]
  9: /usr/lib/xorg/modules/extensions//libextmod.so [0x7fcafa431bda]
  10: X(Dispatch+0x364) [0x446da4]
  11: X(main+0x3bd) [0x42cc9d]
  12: /lib/libc.so.6(__libc_start_main+0xe6) [0x7fcafad02ff6]
  13: X [0x42c129]

  Fatal server error:
  Caught signal 11.  Server aborting

Like Mr. Block, I'm seeing a NULL cursor_screen_priv at xf86Cursors.c:615.  Judging from the comment at xf86Cursors.c:614, I tried tweaking the code to simply return when cursor_screen_priv is NULL and this seems to resolve the problem for me.

If I use a hardware cursor, then I don't experience this crash, so presumably cursor_screen_priv is not NULL in this case.

Anyway, I tracked down the change to the following git commit:

  commit 0b8f8b24f718820a72ebdc52423c2e6a44e848c5
  Author: Stuart Bennett <sb476@cam.ac.uk>
  Date:   Tue Dec 2 22:52:53 2008 -0800

      xf86Cursors: xf86_reload_cursors shouldn't unconditionally show hwcursor (#14820)
    
      Also, no need to call ShowCursor when SetCursorPosition already does it
      Based on a previous patch by Maarten Maathuis
    
      Signed-off-by: Keith Packard <keithp@keithp.com>

Please let me know if there's anything else I can do to help.

Thanks,
-Ted

Comment 3 Jim Paris 2010-02-28 16:27:36 UTC

This is the same bug as #21535, debian bug #507916, etc.  The fix is trivial, just needs a NULL check...

Comment 4 Julien Cristau 2010-03-01 06:52:59 UTC

Let's mark as dupe of #21535 then, thanks.

*** This bug has been marked as a duplicate of bug 21535 ***

Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.