Bug 18911

after some git bisecting i got here:

----------------------------------------------------------------------------
a5c5ce96279d01eb519bfb92b94c06a58acb7f07 is first bad commit
commit a5c5ce96279d01eb519bfb92b94c06a58acb7f07
Author: Alex Deucher <alexdeucher@gmail.com>
Date:   Mon Oct 6 15:07:31 2008 -0400

    Refactor atom LCD/DFP output setup

    Simplify the code, properly handle dithering, coherent mode, encoding, etc.
    Should fix bug 17897.

:040000 040000 6b170dc71fa0062d55b43e77d23db345cda01427 75438c71cd7cc71f1f27fee1a9c1bf9009bbd678 M      src
----------------------------------------------------------------------------

I'm seeing also a blank screen (and a machine hang) with 

01:00.0 VGA compatible controller: ATI Technologies Inc RV370 5B60 [Radeon X300 (PCIE)]
01:00.1 Display controller: ATI Technologies Inc RV370 [Radeon X300SE]

and current git. Haven't checked if that commit is the one to blame.

machine does not hang for me - i can switch to vt and kill X (with proper return to console) anytime.

but i think it used to lock up some time ago, i guess.

i think i have the same card around (x300se model), if that's necessary i can re-test driver's behavior with it.

For me vt switch is dead. Have to press reset button.
Haven't tried login over network.

Can you attach your latest xorg logs and configs?  Also, if you have one, can you attach the logs from the driver when it was last working?

Created attachment 20856 [details]
r500 - xorg.conf

i'm attaching the xorg.conf i use.

note : commenting out horizsync/vertrefresh doesn't change a thing.

Created attachment 20857 [details] [review]
r500 - current git Xorg log

attaching current git log file on r500 card.

Created attachment 20858 [details] [review]
r500 - last working git revision Xorg log

attaching log from last working git commit (just before the bisected one).

as i wrote before - i have a spare r300 (gigabyte x300se) card around. 

i can do some extra testing with it. would that help?

*** Bug 18990 has been marked as a duplicate of this bug. ***

Created attachment 21012 [details] [review]
turn off coherent mode by default

looks like maybe your monitor doesn't like coherent mode.  Does this patch help?

(In reply to comment #2)
> I'm seeing also a blank screen (and a machine hang) with 
> 
> 01:00.0 VGA compatible controller: ATI Technologies Inc RV370 5B60 [Radeon X300
> (PCIE)]
> 01:00.1 Display controller: ATI Technologies Inc RV370 [Radeon X300SE]
> 
> and current git. Haven't checked if that commit is the one to blame.
> 

Andi, this is a different bug, the relevant code paths are not used on your chip.  Please file a new bug and attach your xorg log and xonfig.

Alex, I would love to test your patch when it is available from a ppa repository like https://edge.launchpad.net/~tormodvolden/+archive.

the patch did not help me (applied cleanly on top of current git). 

i'm still stuck at out of sync.

i'll attach new xorg logs in a minute.

Created attachment 21039 [details] [review]
r500 - log after patching

attaching log after running patched driver patching.

i've run xrandr --verbose under fluxbox sessions, both on 6.9.0 an current git with patch ( in the blind, adding it to ~/.xinitrc ). then i compared both

the current-git+patch xrandr dump shows that coherent is explictly disabled, so i guess the driver did compile right.

--- xrandr_bad.txt      2008-12-11 07:32:38.000000000 +0100
+++ xrandr_good.txt     2008-12-11 07:34:18.000000000 +0100
@@ -1,18 +1,17 @@
 Screen 0: minimum 320 x 200, current 1280 x 1024, maximum 1400 x 1400
 DVI-1 disconnected (normal left inverted right x axis y axis)
        Identifier: 0x4b
-       Timestamp:  2702311
+       Timestamp:  2802498
        Subpixel:   no subpixels
        Clones:     DVI-0
        CRTCs:      0 1
                dvi_monitor_type: auto
                scaler: off
                tmds_pll: bios
-       coherent_mode: 0 (0x00000000) range:  (0,1)
        load_detection: 1 (0x00000001) range:  (0,1)
 DVI-0 connected 1280x1024+0+0 (0x4d) normal (normal left inverted right x axis 
y axis) 338mm x 270mm
        Identifier: 0x4c
-       Timestamp:  2702311
+       Timestamp:  2802498
        Subpixel:   horizontal rgb
        Clones:     DVI-1
        CRTC:       0
@@ -28,7 +27,6 @@
                0042656e512046503732472b440a001a
                dvi_monitor_type: auto
                scaler: off
-       coherent_mode: 0 (0x00000000) range:  (0,1)
        load_detection: 1 (0x00000001) range:  (0,1)
   1280x1024 (0x4d)  108.0MHz +HSync +VSync *current +preferred
         h: width  1280 start 1328 end 1440 total 1688 skew    0 clock   64.0KHz
------------------------------------------------------------------------------
will it be useful to provide those xrandr --verbose dumps?

(In reply to comment #15)
> the patch did not help me (applied cleanly on top of current git). 
> 
> i'm still stuck at out of sync.

Can you attach your video bios?

as root:
cd /sys/bus/pci/devices/<pci bus id>/
echo 1 > rom
cat rom > /tmp/r500.rom
echo 0 > rom

no problem. i'll post it when i get back from work.

Created attachment 21053 [details]
Bios output for x1600 mobility

Created attachment 21081 [details]
r500 - bios

attaching my video card bios

re-tested with current git - problem persists. 

can i somehow provide extra information to help with this?

Can you dump the regs from the last working build and current git?  Use avivotool from here:
http://cgit.freedesktop.org/~airlied/radeontool

(as root):
./avivotool regs all > working.dump
./avivotool regs all > broken.dump
and attach the dumps to the bug.

sorry for delay; i'll provide the dumps in a few hours. 

concerning the current-git dump - since i'm planning to run the command from ~/.xinitrc after starting X with current-git driver : does it matter whether the dump is made right after X starts, or should i first wait for a few seconds after X starts and then make a register dump? 

(i can also do it remotely via ssh from second machine)

(In reply to comment #24)
> sorry for delay; i'll provide the dumps in a few hours. 
> 
> concerning the current-git dump - since i'm planning to run the command from
> ~/.xinitrc after starting X with current-git driver : does it matter whether
> the dump is made right after X starts, or should i first wait for a few seconds
> after X starts and then make a register dump? 
> 
> (i can also do it remotely via ssh from second machine)
> 

Either way should be fine.  ssh would probably be easiest.

Created attachment 21611 [details] [review]
r500 - register dump from 6.9.0 driver

Created attachment 21612 [details] [review]
r500 - register dump from git head

i don't really remember at which commit git was at the time (i think i was messing around with bisecting at that point). 

i'll try doing another git-head regdump soon.

Created attachment 21664 [details] [review]
handle format rev 2

I think this should do the trick.  We weren't handling table format rev 2.

please try git master (d5cb24075928690377fd2dfaa32abfebff9990df).

great, the driver works, and X starts up way faster than before (about 0,5s now, which is about 4x improvement). before (on working revisions) the display would go out of sync and then start up.

the only issue i noticed so far is that there is some green garbage on top of the screen for a split second before desktop/wm starts, but i guess i can live with that ;-)

i'd do some more testing (just in case) today.

kowalski, see this bug for the screen corruption during X startup:
http://bugs.freedesktop.org/show_bug.cgi?id=18108