Bug 5878

Summary: Savage hangs, conflicts with savagefb
Product: xorg Reporter: richlv <richlv>
Component: Driver/savageAssignee: Tormod Volden <bugzi11.fdo.tormod>
Status: RESOLVED INVALID QA Contact: Xorg Project Team <xorg-team>
Severity: major    
Priority: high CC: alexdeucher, erik.andren
Version: 6.9.0Keywords: regression
Hardware: x86 (IA32)   
OS: Linux (All)   
Whiteboard: 2011BRB_Reviewed
i915 platform: i915 features:
Attachments:
Description Flags
xorg.conf without comments
none
Xorg.0.log none

Description richlv 2006-02-14 02:39:39 UTC
this seems to be a regression from 6.8.2

if driver "savage" is selected in xorg.conf, video output is garbled (some 
parts of screen appear in wrong places, pictures missing etc) and after a 
little while the computer hangs completely (does not responf to any key 
combination or network)

in 6.8.2, video was working without any problems with savage driver; changing 
the driver to vesa results in a working system, too.

this is slackware-current with slackware provided packages

lspci line :
01:00.0 VGA compatible controller: S3 Inc. VT8375 [ProSavage8 KM266/KL266]

Xorg.0.log contains a couple of warnings and one error :

(WW) SAVAGE(0): Failed to set up write-combining range (0x94000000,0x3000000)
(WW) SAVAGE(0): Failed to set up write-combining range (0x92000000,0x5000000)
(EE) SAVAGE(0): DRI isn't enabled
Comment 1 Alex Deucher 2006-02-14 06:51:47 UTC
can you attach your xorg log and config?
Comment 2 Alex Deucher 2006-02-14 06:54:07 UTC
you might try:
Option "disabletile" "true"
and/or
Option "BCIforXv" "false"

Also, when you say video output, do you mean the desktop itself or the playback
of a video file using Xv?
Comment 3 richlv 2006-02-14 22:41:00 UTC
Created attachment 4606 [details]
xorg.conf without comments

attaching conf and logfile.

with video output i meant desktop itself - basically kde doesn't even finish
loading
Comment 4 richlv 2006-02-14 22:41:32 UTC
Created attachment 4607 [details]
Xorg.0.log
Comment 5 Alex Deucher 2006-02-15 00:52:41 UTC
(In reply to comment #3)
> 
> with video output i meant desktop itself - basically kde doesn't even finish
> loading

did any of trhe options I suggested help?  Also try setting:

Option "AGPMode" "4"
or
Option "AGPMode" "2"

depending on whether your board is AGP 4x or 2x capable.
Comment 6 richlv 2006-02-15 01:47:39 UTC
sorry for not mentioning before - none of these options helped (including agp)
Comment 7 Alex Deucher 2006-02-15 02:06:30 UTC
try:
Option "usebios" "false"
Comment 8 richlv 2006-02-15 02:20:57 UTC
ugh. i feel stupid. turns out, i have been setting the options in the wrong 
section...

when i moved them from server flags to video device section, a single parameter 
got rid of the problem :

Option "disabletile" "true"
Comment 9 Erik Andren 2006-04-19 16:32:28 UTC
Added regression keyword. Any better luck using 7.1RC1?
Comment 10 richlv 2006-04-19 17:47:20 UTC
i'm using slackware prebuilt packages - i don't feel brave or knowledgeable 
enough to compile it myself, unfortunately.
Comment 11 Alex Deucher 2006-04-19 23:24:22 UTC
It sounds like it may be DRI related.  try playing with the following options:

"DmaMode"
"DmaType"
"BusType"

see the savage man page for more about the options.  You'll have to remove the
"disabletile" option to test these.
Comment 12 richlv 2006-04-20 01:10:07 UTC
tried to play around with these parameters - saw no changes at all.
starting x both with and without disabletile produces :
mtrr: no more MTRRs available
twice in dmesg output (related to "Failed to set up write-combining range" ?).
Comment 13 Alex Deucher 2006-04-20 01:21:45 UTC
(In reply to comment #12)
> tried to play around with these parameters - saw no changes at all.
> starting x both with and without disabletile produces :
> mtrr: no more MTRRs available
> twice in dmesg output (related to "Failed to set up write-combining range" ?).
> 

I assume you still get lockups without disabletile?  Are you using a kernel
framebuffer driver?  FWIW, you can generally ignore the write combining
warnings.  I suspect you have an MTRR set up for a kernel fb device that's
conflicting with X. Shouldn't hurt anything though; just potentially slightly
lower performance.
Comment 14 richlv 2006-04-20 01:57:30 UTC
ugh. turns out, it is not a lockup as such, the computer just becomes EXTREMLY 
slow and unresponsive.

first, the screen is garbled (after it has finished loading cached kde in 3 to 
5 minutes..).. icons are offscreen, text is in incorrect locations. mouse 
responds for half a second every 7 seconds or so.

i tried switching to the console, haven't got anything besides blank screen for 
3 minutes now.

it seems that there is a really big load on the computer, as it's ventilation 
is not turning off at all now (it normally just turns on now and then); trying 
to ssh into it times out.

i am using savagefb in kernel
Comment 15 Alex Deucher 2006-04-20 03:33:57 UTC
(In reply to comment #14)

> 
> i am using savagefb in kernel
> 

that's the problem.  Savagefb and the current X savage driver are not compatible.
Comment 16 richlv 2006-04-20 17:13:15 UTC
ahh. and which one is "the guilty one" ?
will this be fixed in x.org, or should i file a report against savagefb ?
Comment 17 Alex Deucher 2006-04-20 23:17:20 UTC
(In reply to comment #16)
> ahh. and which one is "the guilty one" ?

Depends on you persepctive :)

> will this be fixed in x.org, or should i file a report against savagefb ?
> 

Probably both.  I think the X driver needs to save and restore some additional
streams and 3D related regs; however that should really only affect VT switching
and console restoration.  the fact that X has problems starting means that
savagefb is doing something weird to the hardware that the X server doesn't
like.  For the time being, I would say use vesa of vgafb for console.
Comment 18 Erik Andren 2006-05-31 04:33:07 UTC
Did disabling the fb-driver resolve your problem?
Comment 19 richlv 2006-05-31 05:22:25 UTC
(In reply to comment #18)
> Did disabling the fb-driver resolve your problem?

hmm. meaning savagefb ?
actually i filed a report at kernel bugzilla, where a couple of patches were 
made. after applying them, my computer is happily working with savagefb and x.
org savage driver.

the patch was added to -mm at 22/04/2006 - i'm not sure where it went after 
that.

The patch titled
     savagefb: Add state save and_restore hooks
has been added to the -mm tree.  Its filename is
     savagefb-add-state-save-and_restore-hooks.patch

here's the entry at kernel bugzilla :
http://bugzilla.kernel.org/show_bug.cgi?id=6417
Comment 20 Daniel Stone 2007-02-27 01:30:26 UTC
Sorry about the phenomenal bug spam, guys.  Adding xorg-team@ to the QA contact so bugs don't get lost in future.
Comment 21 Jeremy Huddleston Sequoia 2011-10-09 17:46:39 UTC
So was this entirely a kernel issue?  Can this be closed out?
Comment 22 richlv 2011-10-10 01:04:19 UTC
comment #17 said "I think the X driver needs to save and restore some additional
streams and 3D related regs" - but i don't have access to the hardware anymore, so feel free to close this
Comment 23 Adam Jackson 2018-06-12 19:10:26 UTC
Mass closure: This bug has been untouched for more than six years, and is not
obviously still valid. Please reopen this bug or file a new report if you continue to experience issues with current releases.

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.