On Wed, Jan 01, 2003 at 04:30:50PM +0100, Michel D?nzer wrote:
> On Mit, 2003-01-01 at 04:19, [EMAIL PROTECTED] wrote:
> > On Tue, Dec 17, 2002 at 11:07:16PM -0500, [EMAIL PROTECTED] wrote:
> > > On Tue, Dec 17, 2002 at 12:56:38PM -0800, Ian Romanick wrote:
> > > > On Fri, Dec 13, 2002 at 10:45:45PM -0500, [EMAIL PROTECTED] wrote:
> > > > > Using the texmem-0-0-1 branch from CVS and enabling AGP 4x in
> > > > > XF86Config-4 causes X to lock upon startup. If I have my xinitrc
> > > > > setup to run Quake III alone, I first get the bare black and white X
> > > > > background and cursor for about a second and then some strange lines
> > > > > move across a black screen; I can't switch VT's and ssh'ing in doesn't
> > > > > work. If I have my xinitrc setup to run windowmaker, however, I can
> > > > > ssh in. I then try to kill X via an HUP, but that doesn't work and I
> > > > > try kill -9. At that point, ssh stops working. If it helps at all,
> > > > > the texmem-0-0-1 branch I was previously using (from Nov. 1) works
> > > > > fine with AGP 4x. My system is a Mobile P-4 1.8 ghz with a 16 MB 4x
> > > > > AGP M7. Sorry if this message belongs in the users board.
> > > >
> > > > Do you have fast-writes enabled? Does it work at AGP 2x? I'll see if I can
> > > > look into this in the next few days, but I've got a bunch of other stuff
> > > > already on my plate.
> > > >
> > >
> > > I did some testing with the latest texmem from CVS. Here is what
> > > happens (note that I rebooted my machine after installing the latest
> > > CVS radeon driver).
> > >
> > > AGP4x
> > > -----
> > > If I first start X with WindowMaker as the window manager, everything
> > > seems to work fine. If I subsequently exit X and then start X and
> > > WindowMaker again, the X server seems to hang (the screen is black). I can
> > > ssh in; wmaker is not running. A regular kill command doesn't stop X,
> > > so I try kill -9; at that point, the machine completely locks up and
> > > the ssh connection stops working. Upon reboot, I try to play Quake3
> > > (I run it in an instance of X without any other clients like window
> > > managers, etc.). After X starts, vertical lines run down the screen
> > > and I am forced to reboot.
> > >
> > > AGP2x
> > > -----
> > > I can start WindowMaker several times without the problems listed
> > > before. Also, Quake3 will start. It has problems, however; if I set
> > > the texture detail to the max, after loading certain levels the X server
> > > will crash to the command line. Any resolution higher than 800x600
> > > does the same thing.
> > >
> > > My AGP aperture is set to 64 MB. My system has 256 MB of system
> > > memory and 16 MB of video memory. Note that the game was running
> > > quite nicely with max settings and 1024x768 with the texmem CVS from Nov. 1
> > > and AGP4X. I have never enabled fast-writes.
> > >
> > > Thank you,
> > > Jonathan Thambidurai
> > >
> >
> > I would like to inform you that I have found the solution to the
> > above-mentioned problems. These problems occur in several branches of
> > XFree86 starting from about one-and-a-half months ago, I believe. The
> > removal of two lines of code was all that was necessary to correct
> > EVERY one of the above problems. It is as follows.
> >
> > File: xc/programs/Xserver/hw/xfree86/drivers/ati/radeon_dri.c
> > Lines: 710 and 711
>
> A unified diff says more than a thousand words, but I assume it's about
> this:
>
> /* Workaround for some hardware bugs */
> if (info->ChipFamily < CHIP_FAMILY_R200)
> OUTREG(RADEON_AGP_CNTL, INREG(RADEON_AGP_CNTL) | 0x000e0020);
>
> > I understand that this code may be beneficial for some people, but I
> > do not understand enough about X or the various registers of the
> > graphics hardware to actually 'fix' the code and these lines wreak
> > havok on my system.
>
> These lines of code come from XFree86 CVS, probably by Kevin E. Martin
> or Hui Yu, It would be interesting to hear their rationale behind this,
> putting them on CC:.
>
>
> > In my previous message, containing the suggested fix removing two
> > lines of code, I reported that it fixed all of the mentioned
> > problems. I have found this not to be the case. If I increase the
> > resolution to 1024x768 or push the texture resolution to the max (both
> > of which were simultaneously possible previously), the Xserver quits
> > to the console.
>
> You'd have to provide a server backtrace or at least a log snippet (if
> it contains any meaningful output) for this.
Below is a backtrace. The problem only occurs with an AGP aperture
of 64 MB (which previously worked). I tried to get a backtrace with
the gdb modified for XFree86 modules, but it produced no meaningful
results; the attached backtrace is from the regular gdb. Lastly, the
Xserver does not crash as I previously mentioned. If Q3A is the only
program running, X exits to the console, but any input that I type is
not displayed. If I run Q3A from WindowMaker, Q3A is killed and my
desktop is left in an odd resolution that doesn't completely fit on my
laptop LCD (this is the point at which I retrieved the following
backtrace. Additionally, the pointer is fixed in the top left-hand
corner of the screen. I have to switch VT's to exit X at that point.
#0 0x40115e3e in __select () at __select:-1
No locals.
#1 0x000927c0 in ?? ()
No symbol table info available.
#2 0x080b07bc in Dispatch () at dispatch.c:379
clientReady = (int *) 0xbffff3e4
result = -514
client = 0x100
nready = 0
start_tick = -1073745468
#3 0x080c097b in main (argc=2, argv=0xbffff894, envp=0xbffff8a0) at main.c:435
argv = (char **) 0xbffff894
i = 1
j = -514
k = 0
error = 1075207768
xauthfile = 0xfffffdfe <Address 0xfffffdfe out of bounds>
alwaysCheckForInput = {0, 1}
#4 0x4006117d in __libc_start_main (main=0x80c0428 <main>, argc=2,
ubp_av=0xbffff894, init=0x806bf70 <_init>, fini=0x817d880 <_fini>,
rtld_fini=0x4000a534 <_dl_fini>, stack_end=0xbffff88c)
at ../sysdeps/generic/libc-start.c:129
ubp_av = (char **) 0xbffff894
fini = (void (*)()) 0x4001488c <_dl_debug_mask>
rtld_fini = (void (*)()) 0x81ddfe0 <LastSelectMask>
ubp_ev = (char **) 0xfffffdfe
-------------------------------------------------------
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
