-current issues compared to 3.0

Hi,

I've played with -current as of monday (IIRC) and
3.0_STABLE inside qemu.
There are a number of oddities that I'd like to have
commented. They aren't
really problems for me, but it would be nice to have them
resolved.

qemu is as installed from pkgsrc, host is 3.0_STABLE.

When using a boot loader configured to com0 as console
-current shows only
the first two cpu0 lines on qemu's serial console screen,
the full info
appears only when using the bios console. dmesg is complete.
With 3.0 I see
the full info also on the serial console.


In addition to the serial console 3.0 would still give me a
login on the
bios console using ttyE0 if VGA is present. It is with qemu.
However on
-current wsdisplay and friends still appear in dmesg, but
ttyE0 is not
configured. If I change the boot loader back to bios console
it's working
again.

This is a bit annoying since qemu's serial console sucks
somewhat and can't
run vi or use cursor keys. At least AFAIK, since I started
only recently to
use qemu. But for qemu I could still go back to the
non-serial boot loader.


-current crashes sometimes, seems to happen when I use
qemu's scrollback
function of the console screen or switch screens. This is
with the serial
boot loader. getty just complained that ttyE0 isn't
configured.  But I'm
not really sure about all this and can't be more precise at
the moment. The
message is:
kernel: supervisor trap page fault, code=0
Stopped at      netbsd:wsdisplay_scroll+0x35   movl  
0(%eax),%eax

I didn't notice this with 3.0.


The last thing I noticed it that when changing from qemu's
serial console or
monitor screen to the VGA display with Ctrl-Alt-1 a stream
of '1' appears
until another key is pressed. This is probably a qemu bug.
It happens with
both versions.

Bye, Chris

I don't know if this is related but I just upgraded my
3.99.11 system  
up to 3.99.21 kernel and virtual terminals stopped working.
Starting  
X.org 6.9 from pkgsrc doesn't work either. It starts
loading and then  
gets stuck in text mode waiting on "wswait" wait
channel. A back  
trace indicates that we're sitting at "mpidle"
before entering  
getting the ddb keyboard sequence. Not sure if this is time
counters  
related or not.


On Jun 7, 2006, at 1:15 PM, Christian Hattemer wrote:

> Hi,
>
> I've played with -current as of monday (IIRC) and
3.0_STABLE inside  
> qemu.
> There are a number of oddities that I'd like to have
commented.  
> They aren't
> really problems for me, but it would be nice to have
them resolved.
>
> qemu is as installed from pkgsrc, host is 3.0_STABLE.
>
> When using a boot loader configured to com0 as console
-current  
> shows only
> the first two cpu0 lines on qemu's serial console
screen, the full  
> info
> appears only when using the bios console. dmesg is
complete. With  
> 3.0 I see
> the full info also on the serial console.
>
>
> In addition to the serial console 3.0 would still give
me a login  
> on the
> bios console using ttyE0 if VGA is present. It is with
qemu.  
> However on
> -current wsdisplay and friends still appear in dmesg,
but ttyE0 is not
> configured. If I change the boot loader back to bios
console it's  
> working
> again.
>
> This is a bit annoying since qemu's serial console
sucks somewhat  
> and can't
> run vi or use cursor keys. At least AFAIK, since I
started only  
> recently to
> use qemu. But for qemu I could still go back to the
non-serial boot  
> loader.
>
>
> -current crashes sometimes, seems to happen when I use
qemu's  
> scrollback
> function of the console screen or switch screens. This
is with the  
> serial
> boot loader. getty just complained that ttyE0 isn't
configured.   
> But I'm
> not really sure about all this and can't be more
precise at the  
> moment. The
> message is:
> kernel: supervisor trap page fault, code=0
> Stopped at      netbsd:wsdisplay_scroll+0x35   movl  
0(%eax),%eax
>
> I didn't notice this with 3.0.
>
>
> The last thing I noticed it that when changing from
qemu's serial  
> console or
> monitor screen to the VGA display with Ctrl-Alt-1 a
stream of '1'  
> appears
> until another key is pressed. This is probably a qemu
bug. It  
> happens with
> both versions.
>
> Bye, Chris
>
>
> !DSPAM:44870a4643071271154416!
>
>
>

On Thu, Jun 08, 2006 at 02:05:03AM -0400, Chris Tribo wrote:
> 
> I don't know if this is related but I just upgraded my
3.99.11 system
> up to 3.99.21 kernel and virtual terminals stopped
working. Starting
> X.org 6.9 from pkgsrc doesn't work either. It starts
loading and then
> gets stuck in text mode waiting on "wswait"
wait channel. A back trace
> indicates that we're sitting at "mpidle"
before entering getting the
> ddb keyboard sequence.  Not sure if this is time
counters related or
> not.
> 
sounds like the same bug described by Geoff Wing and me. If
I break
into ddb while the kernel in booting (and hanging), I get
the following
trace:

...
--- interrupt ---
mpidle()
ltsleep()
main()

regards,

Arnaud

Arnaud Lacombe wrote:

>On Thu, Jun 08, 2006 at 02:05:03AM -0400, Chris Tribo
wrote:
>  
>
>>I don't know if this is related but I just upgraded
my 3.99.11 system
>>up to 3.99.21 kernel and virtual terminals stopped
working. Starting
>>X.org 6.9 from pkgsrc doesn't work either. It
starts loading and then
>>gets stuck in text mode waiting on
"wswait" wait channel. A back trace
>>indicates that we're sitting at
"mpidle" before entering getting the
>>ddb keyboard sequence.  Not sure if this is time
counters related or
>>not.
>>
>>    
>>
>sounds like the same bug described by Geoff Wing and me.
If I break
>into ddb while the kernel in booting (and hanging), I
get the following
>trace:
>
>...
>--- interrupt ---
>mpidle()
>ltsleep()
>main()
>
>regards,
>
>Arnaud
>
>  
>
I can look into this tonight (in about 5h time) and see
whether
I get the same symptoms in x86.

Frank