Starting up Xati using Radeon 9800 pro, it works for about a second, I can move my mouse for about a second as well. My system then hard locks up, I can't ssh/ping it across my network either. So all I get is the generic white and black speckled screen and then locks up on that screen and you'll see the generic X cursor too. Xvesa and Xfbdev (but I've only tried it on the vesa framebuffer) work fine for me, however. My system specs Radeon 9800 pro 128MB pentium IV 3.2 ghz 512 ddr Asus P4C800-E deluxe motherboard.
I'm seeing a similar issue on a vanilla radeon 9600(www.sapphiretech.com) with the Xati server. I can hear the monitor sync to the mode i'm trying to use, then a lock up with a black screen and no keyboard input. I've noticed that Xati will work, slowly, if i pass -dumb to it when starting the server. Xati locks up both with and without /dev/fb0 available. Both Xvesa and Xfbdev(with vesafb and radeonfb) work fine. /opt/fdo/bin/Xati -dumb -dpi 96 -nolisten tcp -shadow -screen 1024x768x24x60 -mouse /dev/input/mice,5 & /opt/fdo/bin/waimea --display :0 This is the command line i'm using when trying to start the server.
Xati is my fault, assign to me.
Ok, I've tried having ATIPrepareSolid and ATIPrepareCopy return false, although i'm rather uncertain i've done things correctly. I simply substituted FALSE; at lines 333 and 432 in xserver/hw/kdrive/ati/ati_draw.c Regardless, i still end up with a black screen. After returning xserver/hw/kdrive/ati/ati_draw.c to it's original state, I redirected Xati's output to a file with and without Xati -dumb. Here are the results, without a framebuffer device: /opt/fdo/bin/Xati -dumb -dpi 96 -nolisten tcp -screen 1024x768x24x60 -mouse /dev/input/mice,5 -verbose Error opening /dev/fb0: Not a directory Using ATI card: ATI Radeon RV350 AP (PCI) at pci:0000:02:00.0 Mode requested 1024x768x24 0x0118: 1024x768x24 TrueColor [8:8:8:0] True Color red 0x18 green 0x18 blue 0xff0000 Previous modes: VBE 0x4003 BIOS 0x3 Enable VBE mode 0x118 Restore BIOS mode 0x3 /opt/fdo/bin/Xati -dpi 96 -nolisten tcp -screen 1024x768x24x60 -mouse /dev/input/mice,5 -verbose Error opening /dev/fb0: Not a directory Using ATI card: ATI Radeon RV350 AP (PCI) at pci:0000:02:00.0 Mode requested 1024x768x24 0x0118: 1024x768x24 TrueColor [8:8:8:0] True Color red 0x18 green 0x18 blue 0xff0000 Previous modes: VBE 0x4003 BIOS 0x3 Enable VBE mode 0x118 Restore BIOS mode 0x3 [strawman@max strawman]$ cat Xati-without-dumb.log Error opening /dev/fb0: Not a directory Using ATI card: ATI Radeon RV350 AP (PCI) at pci:0000:02:00.0 Mode requested 1024x768x24 0x0118: 1024x768x24 TrueColor [8:8:8:0] True Color red 0x18 green 0x18 blue 0xff0000 Screen: 24/24 depth/bpp Previous modes: VBE 0x4003 BIOS 0x3 Enable VBE mode 0x118 Initialized pseudo-DMA Here are some more details: lspci | grep ATI 02:00.0 VGA compatible controller: ATI Technologies Inc RV350 AP [Radeon 9600] 02:00.1 Display controller: ATI Technologies Inc RV350 AP [Radeon 9600] (Secondary) dmesg | grep agpgart agpgart: Detected NVIDIA nForce2 chipset agpgart: Maximum main memory to use for agp memory: 439M agpgart: AGP aperture is 64M @ 0xe0000000 Radeon 9600 128Mb AMD Athlon XP 2500+ 512Mb Nforce2 based motherboard Fedora Core 2 with xserver and requirements updated daily
Created attachment 360 [details] [review] fix to dma initialization (or lack thereof) for r300 Well, there's the problem (or at least, a problem). Try the attached patch.
Created attachment 362 [details] [review] Alternative: Add r300 cp microcode and initialize pseudo-dma Here's one to try after the 1st.
I can report success with my 9600 on fc2. As of about 5 minutes ago, Xati (without -dumb) works without the first patch above. With the second patch, Hunk #4 fails: @@ -743,10 +752,6 @@ ATICardInfo(pScreenPriv); char *mmio = atic->reg_base; - /* XXX: The Radeon pseudo-dma code is untested. */ - if (0 && atic->is_radeon) - return FALSE; - ATIUploadMicrocode(atis); ATIEngineReset(atis); From fresh cvs, the same bit reads: ATICardInfo(pScreenPriv); char *mmio = atic->reg_base; if (atic->is_r200 || atic->is_r300) return FALSE; ATIUploadMicrocode(atis); ATIEngineReset(atis); I'm going to assume dropping hunk #4 is ok. Dropping that, the 2nd patch above applies and is working.
Committed patch #362, thanks!
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.