From mmcg@heraclitus.cs.monash.edu.au  Mon Jan 13 17:15:32 1997
Received: from heraclitus.cs.monash.edu.au (heraclitus.cs.monash.edu.au [130.194.64.241])
          by freefall.freebsd.org (8.8.4/8.8.4) with ESMTP id RAA21394
          for <FreeBSD-gnats-submit@freebsd.org>; Mon, 13 Jan 1997 17:15:15 -0800 (PST)
Received: (from mmcg@localhost) by heraclitus.cs.monash.edu.au (8.7.5/8.7.3) id MAA26447; Tue, 14 Jan 1997 12:14:33 +1100 (EST)
Message-Id: <199701140114.MAA26447@heraclitus.cs.monash.edu.au>
Date: Tue, 14 Jan 1997 12:14:33 +1100 (EST)
From: Mike McGaughey <mmcg@heraclitus.cs.monash.edu.au>
Reply-To: mmcg@heraclitus.cs.monash.edu.au
To: FreeBSD-gnats-submit@freebsd.org
Subject: 2.2-BETA install kernel creates unreadable display mode
X-Send-Pr-Version: 3.2

>Number:         2485
>Category:       i386
>Synopsis:       2.2-BETA (install) device probe appears to screw screen
>Confidential:   no
>Severity:       critical
>Priority:       medium
>Responsible:    sos
>State:          closed
>Quarter:
>Keywords:
>Date-Required:
>Class:          support
>Submitter-Id:   current-users
>Arrival-Date:   Mon Jan 13 17:20:01 PST 1997
>Closed-Date:    Mon Feb 17 11:32:05 PST 1997
>Last-Modified:  Mon Feb 17 11:37:13 PST 1997
>Originator:     Mike McGaughey
>Release:        FreeBSD 2.2-BETA
>Organization:
Monash University
>Environment:

Pentium 120 on ATC-1000 motherboard, 16M ram, 256k pipeline cache,
no non-pipeline cache, 2 IDE controllers (one of which is in use).

Award PCI/PnP bios.  BIOS (etc) values all at their most conservative
settings;  PCI IRQ (etc) assignment is all on `auto'.

SMC elite 16 bit ISA ether card: I/O 0x300, Irq 10, addr 0xcc000

Paradise PVGA1a PCI video card.  It may be a compatible - it has `ARK'
plastered all over the chips, including the SVGA BIOS chip, but various
tools (aka `games') identify it as a paradise PVGA1a.

Some floppy drives, and some hard drives.

Nothing else.

>Description:

When the kernel supplied with the 2.2-BETA boot (install) floppy
starts probing the hardware (i.e. *after* visual configuration has
been exited), the screen immediately changes to a pattern consisting
mainly of (reasonably) closely spaced vertical lines, with lines
of different colors appearing interleaved, and vaguely reflecting
the colors that (should!) be on the screen at the time.  Interestingly
enough, the lines appear to be in a particular color order, with the
colors repeating vertically once per character position.

When the probes have finished, the install menu comes up - still
with an unreadable screen (though recognisably different, and now
with more colors).  Very pretty, but it gets in the way of reading
the instructions.

A (cold or warm) reboot resets the screen to a sensible state.

I've never seen this happen under any other FreeBSD install or
kernel on this machine (I have never run a 2.2 install on this
particular system - only 2.1.0, 2.1.5 and 2.1.6.  It is currently
running 2.1.6).  Actually, I've never seen it happen at all (under
DOS, etc).

>How-To-Repeat:

Fly out to Australia, visit me.  Be sure to bring a case of beer.
Fire up the install floppy on the machine described above.
In visual configuration mode (while the screen is quite usable),
delete any or all devices (except for the syscons driver).
Quit visual configuration; shortly after the probes start, the screen dies.
Look perplexed.
Look at beer.
Admire my barbequeue, suggest we take it for a test drive.
Sit back, guzzle beer and chops, tossing empty cans and gnawed T-bones
at the offending machine.
Admire cloudless sky and abundant sunshine, and resolve to visit the beach.

>Fix:

I have no idea; I suspect I am on my own here (perhaps with a screwy
configuration, or with one of a bad batch of video cards).  Or, it
may have something to do with them newfangled PCI probes.  However,
if you have heard of this before and can suggest a way around it,
I'll finish the install and test the rest of 2.2-BETA for you :)

Cheers,

    Mike.
>Release-Note:
>Audit-Trail:

From: mmcg@heraclitus.cs.monash.edu.au
To: freebsd-gnats-submit@freebsd.org, mmcg@heraclitus.cs.monash.edu.au
Cc:  Subject: Re: i386/2485: 2.2-BETA (install) device probe appears to screw screen
Date: Wed, 15 Jan 1997 21:50:46 +1100 (EST)

 SUGGESTION:
 	Add a boot switch to bring up syscons in non-graphic (i.e. pre-2.2) mode.
 
 I've looked at the new syscons driver, and noticed that
 it now tries to place the card in graphics mode to drive the
 standard virtual consoles.  So, I put the card in another
 machine (which usually has a Diamond stealth 64, and which has
 no screen problems when installing or running 2.2-BETA).
 The same visual breakup happened when the console was initialised
 by 2.2-BETA (i.e.  *after* exiting config and seeing a few lines
 scroll past) on the other machine.  It's as if the dot clock has
 been mis-programmed or something.  Conversely, the Stealth 64 worked
 smoothly under 2.2-BETA in each machine.
 
 I then fired up Xfree86; a similar looking breakup occurred (in that
 there were vertical lines, but this time several bands of them; I
 presume it was just the different video mode).
 
 >tools (aka `games') identify it as a paradise PVGA1a.
 
 It's definitely *not* any sort of paradise card; presumably, it is new
 enough that the other probes got it wrong (shows what I know :)
 I have no idea what brand this board is.  Here's what's on it:
 
 	An ARK1000VL chip (S3 compatible, according to the net).
 	An ARK1491 chip (ramdac?  clock chip?)
 
 Xfree's SuperProbe reports `VGA (or unknown SVGA) video card detected'.
 I may be able to track it using its FCC approval from www.fcc.gov (FCC:
 ICUVGA-GW405) - the site appears broken right now.
 
 Soooo: it's definitely the video card itself, and it happens
 when programs which assume it is a standard VGA try to put it into
 graphics mode.  DOS programs work fine, and can use the good graphics
 modes (even those that misreport it as a paradise vga1a).  I do
 not know about win95.
 

From: mmcg@heraclitus.cs.monash.edu.au
To: freebsd-gnats-submit@freebsd.org, mmcg@heraclitus.cs.monash.edu.au
Cc: mmcg@heraclitus.cs.monash.edu.au
Subject: Re: i386/2485: 2.2-BETA (install) device probe appears to screw screen
Date: Fri, 17 Jan 1997 08:50:08 +1100 (EST)

 > &gt;tools (aka `games') identify it as a paradise PVGA1a.
 >
 > It's definitely *not* any sort of paradise card[...]
 >        An ARK1000VL chip (S3 compatible, according to the net).
 >        An ARK1491 chip (ramdac?  clock chip?)
 
 Heh.  The ARK chipset appears to be not quite compatible with the
 WD PVGA1a chipset.  It appears to behave itself under DOS, but some
 part of the FreeBSD screen probe (in syscons) places it into a
 strange state.
 
 PCVT also identifies the card as a Paradise PVGA1a - and it manages
 not to screw the screen (it's not insisting on graphics itself, and
 I have no need of X on that machine).  Hence, I'll stick to
 that driver for now.
 
 If anyone ever gets around to reading this PR and getting back
 to me, I'll be happy to apply and test patches to syscons until
 the whole thing works (I *like* the new-look syscons).
 
 NB: It may be wise to distribute a `PCVT-based' install floppy as well
 as the usual syscons one.
 
 Cheers,
 
     Mike, happy enough for now.
Responsible-Changed-From-To: freebsd-bugs->sos 
Responsible-Changed-By: mpp 
Responsible-Changed-When: Mon Feb 17 09:30:56 PST 1997 
Responsible-Changed-Why:  
Syscons problem. 
State-Changed-From-To: open->closed 
State-Changed-By: sos 
State-Changed-When: Mon Feb 17 11:32:05 PST 1997 
State-Changed-Why:  

Problem due to having SC_SPLASH_SCREEN defined. This occured in 
a window of a couble of hours, due to me screwing up a cvs 
commit :( to the RELENG_2_2 branch... 

No further fix nessesary, and syscons does NOT use any graphic 
mode for normal vty's. 
>Unformatted:
