xf86-video-ati with patch-src-radeon_driver.c freeze the system

[Warren Block]

On Tue, 1 Mar 2011, Roland Smith wrote:

> On Tue, Mar 01, 2011 at 11:45:41AM +0000, Christian Weisgerber wrote:
>> Zhihao Yuan <lichray at gmail.com> wrote:
>>
>>> Someone who submitted this patch gave me an horrible afternoon struggling
>>> with the system that reboots again and again (the X and the system just
>>> freeze actually, with no Xorg.log generated,
>>
>>> My card is ATI Mobile Radeon X1600. My working Xorg.0.log is also attached.
>>
>> Indeed, removing that patch fixes xf86-video-ati 6.14 for my
>> ATI Radeon RV370 X300 SE.
>
> Same here for "ATI Radeon X1650" (ChipID = 0x71c1). This patch hangs my
> machine, and removing it fixed it. I'm on FreeBSD 8.2-RELEASE amd64.

I think that's exactly the card I tried here, but worked on i386. 
Unfortunately, I don't have a 64-bit system to try it on, but maybe this 
is just a type comparison problem...

Except mem_size and aper_size are both uint64_t.

The attached (untested) patch disables the original patch code and adds 
an ErrorF to display the actual values.  Delete the original 
patch-src-radeon_driver.c and copy this one 
(patch-src-radeon_driver.c-WB) to the port's files/ directory.  Rebuild 
the port, and see what it shows in the log.

How much memory is really on the card might be important.  The X1650 
here has 512M.
-------------- next part --------------
--- src/radeon_driver.c.orig	2011-01-10 16:10:46.000000000 -0700
+++ src/radeon_driver.c	2011-03-01 13:15:14.000000000 -0700
@@ -1459,6 +1459,12 @@
      * internal memory map.
      * See fdo bug 24301.
      */
+    /* if (mem_size > aper_size)
+     *	mem_size = aper_size;
+     */
+    /* WB: show what's going on here */
+    ErrorF("WB: mem_size %x, aper_size %x\n", mem_size, aper_size);
+
     if (mem_size > 0x20000000)
 	mem_size = aper_size;