Bug 115204

Oh, the friend who bought this laptop on friday found the following hint: 
 
http://www.amilo-forum.de/topic,2532,b266956d9d627f9039c3adb59744cdae,-irgendein-Linux-auf-A%0A+milo-M1437G.html 
 
It contains the following entry: 
 
====================== 
Letztendlich ist es mir mit Hilfe aus einer Newsgroup gelungen, Suse 9.3 auf 
meinem Amilo M1437G zu installieren.  
  
 Das Problem mit der blockierten Tastatur läßt sich mit einem einfachen Trick 
umgehen, auf den man allerdings auch einmal kommen muß:  
  
 Man muß leglich wärend des Installationsvorgangs einen externen Bildschirm 
benutzen und das normale TFT-Display dabei deaktivieren. Dann läuft die 
Installation ganz normal ab.  
  
 Nachdem das System dann auf dem Laptop installiert worden ist läuft alles 
eigentlich ohne größere Probleme. Lediglich die Konfiguration des X-Servers 
ist wegen des rest neuen X700-Chips und den recht widerspenstigen Treibern von 
ATI etwas kompliziert.  
  
  
 At last I succeeded in installing SuSe 9.3 by using a simple trick.  
  
 To install SuSe 9.3 on a Amilo M1437G I've had to use an external display and 
to disable tht tft display. Then, there's no problem with the keyboard during 
the installation process... 
==================== 
 
She tried this with a 9.3 personal (from c't magazine) and it worked. 

Maybe the following fact is interesting too: 
The laptop has a DVI-I instead of the normal vga video plug. 

This case looks *a bit* similar to Bug 106078 - "Keyboard lockup, clock not 
counting down, system locked??", but please note that I'm not even getting a 
graphical screen. 

Could this be a duplicate of 115204? 

Jörg, you are not seeing _anything_ (that is all black except for that error 
message)? Could you try out a special test version I would prepare? 
 
Stefan, do we have something similar to that notebook for testing? 

The screen is not black but has a blue background except the top left corner 
with the error message (from memory, will check tonight). There is no cursor 
or something. 
Yes, of course I can test - I've got the machine until the end of the week. 
Btw, I've updated the BIOS to 1.0.6C and done a beta4 installation with the 
trick from the post above.  

Blue is ok. Is it ok to email you the 1.1MB Test-ISO (else I'll put it on 
our ftp server)? 

Email is fine. 

comment 5: no, sorry.  

Tested the image, now it fails in graphical mode with the same symptoms: 
The IOSLINUX screen is now readable (until ESC :) 
... 
Initializing gfx code... 
static memory: 0x38020 - 0x9f400 
     malloc 0: 0x49210 - 0x9f400 
     malloc 1: 0x800000 - 0x900000 
     malloc 2: 0xa00000 - 0xb00000 
     malloc 3: 0x0 - 0x0 
 
<ESC> 
 
screen 800x600, 16 bits 
mem 0: 262 kb 
mem 1: 1023 kb 
mem 2: 1023 kb 
mem 3: 0 kb 
 
<ESC> 
 
Now the graphical screen with the animated welcomde messages appears (I did 
not get this before) and then the menu with a working countdown appears. Input 
isn't working and after the countdown is finished a requester appears "Could 
not find kernel image: linux [OK]"  
------ 
nothing but power off works now. 
 
Btw, the machine has 1 MB RAM. 

Thanks. Something corrupts the memory, but I don't know what. 
I'll prepare some more ISO's tomorrow (or floppy image if that's better 
for you). 
 
You still want to go on? :-) 

Sure! Floppy images are a bit problematic as the laptop doesn't have a floppy 
drive ;) 

Results of test2.iso: 
Things count up to 70 without error. Between 30 and 40 the resolution is 
changed. After 70 the graphical screen is displayed and finally the boot menu 
with the countdown appears. The seconds decrease and the dots disappear. 
Pressing the first key causes the number to stop decreasing while the dots 
still disappear. NOTE: This only happend during my first attempt. All later 
attempts did not cause the number countdown to stop. After the countdown 
finishes, the debug information appears in the top left corner, no input is 
possible. The error value is always 0. In graphical mode NO key has any 
effect, including num-lock (num lock still works when step 70 is displayed). 

When I say NO key I really mean no key. I have to press power off for a few 
seconds to power down the box, C-A-DEL doesn't work. 

That's bad. You are absolutely sure that it showed '1.3' at the '0x70' 
screen (the last one)? And after that came the various 'welcome' messages 
and then did the timeout run? 
 
BTW, what did you mean with '1 MB' in comment 10? The video memory? 
What's the graphics card in that notebook? 

Sigh, typo. The machine has 1 GB of RAM, not 1 MB 
Yes, it says 0: 70.1 and 1: 1.3. Yes, it was as you decribe: After ESC there 
were the welcome messages and then the timrout ran. 
The graphics card is a Radeon X700 PCIE for laptops. A complete hwinfo of the 
box is attached to another bug and can be found at 
https://bugzilla.novell.com/attachment.cgi?id=48831 

Btw, the text at 100 where all keypresses are ignored looks like this: 
pstk 
0: 0.1 
1: 100.1 
2: 1.3 
rstk 
0: ffffffff.5 
1: 8cc.5 
--- 
err 0 
ip 333: 14.7 
 

Created attachment 49044 [details]
test 3

It stops at 0x70 & 0x80. Please check if it
switches to false ('0.3') at one of them.
Are the greetings shown (the 'welcome' msgs)?
After ESC in 0x80, does the timeout start?
If so, do the keys work this time?

Created attachment 49046 [details]
test 4

_Only_ if in test 3 ESC does not work at 0x80
(that is, the timeout does not start):

Do the keys work in the menu (assuming you get there)?

Created attachment 49048 [details]
test 5

If the keys don't work in either of the last two images, please give this one a
try.

test3.iso: 
It's 1.3 at 70 and 80. Between 70 and 80 the welcome screen is displayed and 
the boot menu appears. At 80 ESC doesn't work any more. 
Other funny experiments (always after power-off-on): 
Press ESC not once but twice (about 1 second apart) at 70: 
Now the Menu appears, the timer runs etc (of course the keyboard doesn't work)  
 
test4.iso: 
keyboard still not working, 80 doesn't appear 
If I remove the CD during countdown, an error requester is displayed at the 
end AND a small message at the top left corner (green at black background): 
isolinux: Disk error ... 
 
test5.iso: 
same as test4 
 
 

Then it looks like the keyboard is physically dead at that point. 
What puzzles me is that both the timer and keyboard are handled in the 
same loop, but the timeout works... 

Created attachment 49069 [details]
test 6

Ok, next try. :-)

As before, please tell me at which screen the keyboard stops working (it jumps
from (0x70 to 0x100).

test6.iso:    
at 70, ESC works.    
Display of the welcome messages, then the status requester appears     
3: 100.1    
4: 1.3    
err 0    
Display of the welcome messages, then the status requester appears     
Keyboard doesn't work.    
 
I've spent some more research and thought on the problem: 
 
1) If you follow the link in comment#1, the first entry mentions that   
knoppix worked. I've asked around and was told that knoppix 4.0 dvd uses  
ISOLINUX 2.11 2004-08-16. I've tried knoppix V3.3 release 2003-11-19-beta and 
it boots just fine and dies much later when we run into unsupported hardware 
in the boot process. The knoppix V3.3 runs syslinux 2.04 (Debian 2003-06-06). 
So what is different between the knoppix/debian/ubuntu syslinux and the 
suse/redhat syslinux? 
 
2) It seems that the display of the graphical welcome screen corrupts 
something. Is there a way to skip the welcome screen or maybe even avoid the 
graphical mode at all? 

I'm note exactly sure: you mean at *0x100* the keyboard does not work? 
 
I currently do not think the underlying syslinux matters. I'd rather 
suspect the graphics. But neither does explain why the keyboard is dead. 
Not to mention why it works with an external monitor. 

Created attachment 49088 [details]
test 7

Provided the answer to my last question is yes, here is a new iso where you
step very finegrained through the greetings drawing.

Please don't press ESC to fast (wait 1 or 2 secs), there are delays build in.

#25: yes, it didn't work at 0x100  
 
#26: 
single stepping works up until 100.1. When the display shows 100.1 ESC doesn't 
work any more, the screen still shows the welcome screen. But there is 
something interesting (I've done about a dozen runs): 
it works "normally" until XX.1. Then it jumps to a0.1 -> a1.1 -> 100.1 
From 71 to a0.1(first time) (both inclusive), the pstk looks like 
0: 526de.c 
1:    XX.1 
2:     1.3 
 
0: a1.1 
1:  1.3 
 
0: 62542.a 
1: 62519.a 
2:  ad44.4 
3:   100.1 
4:     1.3 
 
the XX is not always the same: depending on how long I wait beween key 
presses, it varies from 85 to af. So it occurred several times that the number 
jumped backwards. 

Created attachment 49167 [details]
test 8

Ok, please try this. :-)

We are narrowing the place where the keyboard stops working (hopefully).

To your last test: yes, the texts are shown at random. The implementation is
somewhat weird, so the effects you described are normal.

Created attachment 49168 [details]
test 9

similar to test 8

Created attachment 49169 [details]
test 10

Again, quite similar.

Ok, I expect test 8 to break at 0x100. If that's the case, I give test 10 
a fair chance of succeeding. test 9 is somewhere in between. If you see 
no difference between test 8 and test 10, skip test 9. 

Created attachment 49171 [details]
test 11

Forget test 10. Try test 11.

#28 (test8): 
The behaviour now: 
70, then the screen including the welcome messages is displayed.  After that I 
can still single step up to and including 90. Pressing ESC then gets me to 
100, at which point again only power-off-on works. 
 
#31 
I see no difference between 8 and 10. 
 
#32 
test11.iso works! 
I can press ESC at 100, then the boot menu appears and cursorkeys, function 
keys, tab and help work (nothing else tested :) 
Should I run test9 now or was it problematic like test10? 

No, skip test 9. 

Created attachment 49174 [details]
test 12

But try test 12. It undoes most of the changes, but hopefully not the critical
one.

My current theory: the monitor detection
breaks the keyboard somehow. That would explain that an external monitor helps.


But would unfortunately put the blame on the BIOS. (Unless we find a
workaround.) :-(

test12 worked fine too. 
And yes, it's very likely that the BIOS is broken - but that's not that 
uncommon as far as BIOSse are concerned ;) 

Just in case it won't take too much time to answer: What seems to be the 
problem that causes ioslinux to fail? 

See comment 35, the monitor detection. I'll prepare a test case to make sure. 
 
Problem is, I can't turn it off generally. 

There's another bug open that is concerned with the video settings for the X 
server (https://bugzilla.novell.com/show_bug.cgi?id=115283) which is 
unresolved too and which also seems to indicate the bios or video bios are a 
bit broken on this box. 

Created attachment 49202 [details]
test 13

Hold down the Alt key while isolinux starts.
Does it work?

Could you please run 'hwinfo --vbe --log=xxx' and attach the log? 

RRRRRRRROOOOOOOOOOOOOOTTTTTTTTTTTTTFFFFFFFFFFFFFFFFFLLLLLLLLLLLLLLL 

Created attachment 49225 [details]
Output of hwinfo --vbe

OK, ALT worked :-) and the keys also worked

Hope you have recovered from comment 42 by now. ;-) 
 
Ok, optionally skipping the monitor test with 'Alt' is the best we can do  
at this point. This will probably help others with similar problems like  
bug 106078, too. 
 
Interestingly it works in a running system ('hwinfo --vbe' does exactly 
the same). 
 
Anyway, a big thanks to you for your help (and patience). 

Curiosity kills the cat. :-) 
 
Could you please run 'hwprobe=+cpuemu.debug hwinfo --monitor 2>io_log' 
and attach the log? It will trace the BIOS while doing the monitor detection. 
Maybe it accesses some unusual io ports. 

Created attachment 49420 [details]
Baldrian for the cat

Created attachment 49432 [details]
test 14

A very wild guess.

Doesn't work unless I start with alt pressed (or an external monitor 
connected :) 

Would have been too easy. :-) 

*** Bug 131522 has been marked as a duplicate of this bug. ***