Re: [Dri-devel] Fault in radeon DRM module
On Sun, 7 Sep 2003, Jon Smirl wrote: I'm getting this with standalone Mesa not DRI. Can a someone more familar with the R200 kernel DRM driver give me a clue as to what is not being set up correctly? I die in RADEON_PURGE_CACHE() in radeon_do_cp_start(). More precisely: Unable to handle kernel paging request at virtual address f8a6cffc printing eip: f8a579f1 *pde = 01b99067 *pte = Oops: 0002 [#1] CPU:1 EIP:0060:[f8a579f1]Not tainted EFLAGS: 00010286 EIP is at radeon_do_cp_start+0x9b/0x1f1 [radeon] eax: 000b ebx: 0003 ecx: c02f18d4 edx: 5c50 esi: f8a6d000 edi: ebp: f6ae6800 esp: f5959f24 ds: 007b es: 007b ss: 0068 Process sample_server (pid: 2045, threadinfo=f5958000 task=ec6f06b0) Stack: f8a649a3 0018 000a 0400 f8a6496c f719d000 f6ae6800 f5c3ce80 f5c3ce80 f8a591ea f6ae6800 f8a61b29 f8a6273e f5959f88 f719d000 0041 f5c4a100 f8a53826 f6de5598 f5c3ce80 6441 0041 e200 Call Trace: [f8a591ea] radeon_cp_start+0xe1/0xfc [radeon] [f8a53826] radeon_ioctl+0xce/0x13d [radeon] [c016af59] sys_ioctl+0xf6/0x274 [c010b241] sysenter_past_esp+0x52/0x71 Code: c7 04 be 0b 0d 00 00 83 c7 01 21 df c7 04 be 0f 00 00 00 c7 The code in question is ** movl $0xd0b,(%esi,%edi,4) ** add$0x1,%edi and%ebx,%edi movl $0xf,(%esi,%edi,4) which looks strange. That's not a NULL pointer derefernce, but %edi is -1, and %esi is 0xf8a6d000. As far as I can tell, this is part of OUT_RING() with a value of 0x0d0b. It looks like the first OUT_RING() in the function, and it _looks_ like ring is %esi, and write is %edi. In other words, in BEGIN_RING(): ring = dev_priv-ring.start;\ write = dev_priv-ring.tail;\ looks like the ring hasn't been initialized. Which looks unlikely, since it's initialized when it is allocated. . I have no idea how that could happen, but you might put some debugging code into BEGIN_RING() to print out if (ring.tail ~ring.tail_mask) is ever non-zero. It shouldn't be. Linus --- This sf.net email is sponsored by:ThinkGeek Welcome to geek heaven. http://thinkgeek.com/sf ___ Dri-devel mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/dri-devel
Re: [Dri-devel] Re: CVS Update: xc (branch: trunk)
On Sun, 2003-09-07 at 07:42, Dieter Nützel wrote: Am Samstag, 6. September 2003 23:29 schrieb Eric Anholt: On Sat, 2003-09-06 at 14:16, Eric Anholt wrote: CVSROOT: /cvs/dri Module name: xc Repository: xc/xc/ Changes by: [EMAIL PROTECTED] 03/09/06 14:16:32 Log message: Second test of committing in freedesktop.org CVS. Modified files: xc/xc/: test Revision ChangesPath 1.2 +1 -1 xc/xc/test I've been working with keithp the last couple of days on getting dri.freedesktop.org set up. Anonymous and developer CVS access both work, along with cvsup. Instructions can be found at http://dri.freedesktop.org/ As far as I know, only myself, Michel Dänzer, and Keith Whitwell have made accounts there so far. Again, for existing developers without accounts there, forward [EMAIL PROTECTED] your ssh dsa key and he can set one up for you. Some progress: SOURCE/dri-trunk cd xc/ Directory: /tmp/INSTALL/SOURCE/dri-trunk/xc dri-trunk/xc cvs update protocol error: directory '/cvs/ ri/xc/xc/programs/Xserver/hw/xfree86/os-support' not within root '/cvs/dri' Not sure what would be going on there. I just cvs updated as both anonymous and as a developer, with no problems. And it was blazing fast, too. Actually, what happened with that output there, /cvs/ri/xc/xc/... not /cvs/dri/xc/xc/...? -- Eric Anholt[EMAIL PROTECTED] http://people.freebsd.org/~anholt/ [EMAIL PROTECTED] --- This sf.net email is sponsored by:ThinkGeek Welcome to geek heaven. http://thinkgeek.com/sf ___ Dri-devel mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/dri-devel
[Dri-devel] r200 glXDestroyContext bug?
It seems that r200 driver's glXDestroyContext() needs valid Drawable. In r200GetLock(), DRI_VALIDATE_DRAWABLE_INFO() is called to get a valid clip rect of target Drawable. It fails if the Drawable is already destroyed. (for details, see xc/lib/GL/mesa/src/drv/r200 sources in XFree86). According to glXDestroyContext() man page, a valid drawable is not listed as a requirement. Since my GL setup package (gtkglext) often sends an already destroyed drawable, I get crashes. Thoughts? -- // [EMAIL PROTECTED] [EMAIL PROTECTED] scopira.org // --- This sf.net email is sponsored by:ThinkGeek Welcome to geek heaven. http://thinkgeek.com/sf ___ Dri-devel mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/dri-devel
Re: [Dri-devel] r200 glXDestroyContext bug?
Aleksander Demko wrote: It seems that r200 driver's glXDestroyContext() needs valid Drawable. In r200GetLock(), DRI_VALIDATE_DRAWABLE_INFO() is called to get a valid clip rect of target Drawable. It fails if the Drawable is already destroyed. (for details, see xc/lib/GL/mesa/src/drv/r200 sources in XFree86). According to glXDestroyContext() man page, a valid drawable is not listed as a requirement. Since my GL setup package (gtkglext) often sends an already destroyed drawable, I get crashes. That is a bug, but I *think* it is fixed in DRI CVS. Keith or Michel might know better. --- This sf.net email is sponsored by:ThinkGeek Welcome to geek heaven. http://thinkgeek.com/sf ___ Dri-devel mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/dri-devel