Michael | 1 Jan 17:11 2006
Picon
Picon

Re: Kernel breakage

Hello,

> any kernel compiled from sources updated today borks very early, in
> fact it only prints the start address and immediately runs into a
> DEFAULT CATCH. Kernels built from older sources ( 12/23 or so ) work
> fine. This happens on my G3-equipped S900, didn't try the iBook yet.
> 
> Before I seriously start digging:
> - any ideas what could cause this? Downgrading arch/macppc to 12/23
>   didn't cure it.
> - is it just me or does it happen to anyone else? 
> 
> Yes, I rebuilt tools and deleted the whole obj directory so there
> shouldn't be any stale object files or dependencies anywhere.

This is getting weird.
GENERIC built from the same source works. A sparc64 kernel built from
the same source works too.
So I looked for differences - found #options ALTIVEC and -mcpu=750,
neither seemed to cause the problem. 

have fun
Michael
Johan Wallén | 2 Jan 09:14 2006
Picon
Picon

Re: Kernel breakage

Michael <macallan18 <at> earthlink.net> writes:

> any kernel compiled from sources updated today borks very early, in fact
> it only prints the start address and immediately runs into a DEFAULT
> CATCH. Kernels built from older sources ( 12/23 or so ) work fine.
> This happens on my G3-equipped S900, didn't try the iBook yet.
>
> Before I seriously start digging:
[...]
> - is it just me or does it happen to anyone else? 

It happens to me on a PowerBook G4, but I get a

  Invalid memory access at %SRR0: 001017cc  %SRR1: 00080030

instead of a DEFAULT CATCH.

-- Johan

Matthias Scheler | 2 Jan 14:19 2006
Picon

Re: Gigabit support in gem drivers

In article <A2E928B0-A011-4DAB-8F12-B206244E97E3 <at> smurfturf.net>,
	spamcop <at> smurfturf.net writes:
> Sadly, NetBSD's gem driver (used by most onboard PM cards) does not
> support Gigabit.  

That is *not* true. NetBSD 3.0_BETA could run the gem(4) interface in
my Powermac G4 MDD at full gigabit speed. "ttcp" reported more than
100.000 bytes per seconds. The NFS server was able to send out data
at 34MB/Sec (from the harddisk, not from the cache).

	Kind regards

--

-- 
Matthias Scheler                                  http://scheler.de/~matthias/

Kjell Konis | 2 Jan 19:57 2006
Picon
Picon

NetBSD 3.0 on Powermac 9500

Hi,

I just installed NetBSD 3.0 on my Powermac 9500.  I compiled a  
current kernel (3.99.15) without Altivec and with multiprocessor  
support and the vga buffer hack to get XFree86 to work.

-bash-3.00$ diff MYKERNEL GENERIC
29c29
< #options      ALTIVEC         # Include AltiVec support
---
 > options       ALTIVEC         # Include AltiVec support
555,559d554
<
< options       OFB_FAKE_VGA_FB
< options       MULTIPROCESSOR

I am happy to report that this kernel seems to work fine.  I had  
previously filed a problem report about the kernel panicing when  
multiprocessor support was enabled - this was 2.0.2 with a current  
kernel.  So first question: how do I look up/attach a comment to my  
problem report?

When the kernel is compiling, "-Wa,-maltivec" is always among the C  
flags.  What does that comma mean?  Is it just for warnings or is it  
actually including Altivec support?

Also, I have an ABD AppleDesign keyboard.  It works fine under  
Xmacppc but several of the keys (notably backspace, delete, and the  
up, right and left arrow keys) don't work under XFree86.  Has anyone  
gotten one of these to work XFree86?
(Continue reading)

Michael | 2 Jan 20:31 2006
Picon
Picon

Re: NetBSD 3.0 on Powermac 9500

Hello,

> So first question: how do I look up/attach a comment to my problem
> report?

Mail to gnats-bugs <at> netbsd.org, with 'Re: kern/xxxxx' as subject (
obviously replace 'kern/xxxxx with your real PR ID ) - the whole mail
will be appended to the PR. To lookup PRs use
http://gnats.netbsd.org/Misc/query-pr.html

> When the kernel is compiling, "-Wa,-maltivec" is always among the C  
> flags.  What does that comma mean?  Is it just for warnings or is it  
> actually including Altivec support?

This means that '-maltivec' is passed to the assembler as command line
option.

> Also, I have an ABD AppleDesign keyboard.  It works fine under  
> Xmacppc but several of the keys (notably backspace, delete, and the  
> up, right and left arrow keys) don't work under XFree86.  Has anyone  
> gotten one of these to work XFree86?

Both my UMAX labeled Extended II and the built-in keyboard in my iBook
work just fine without any xmodmap stunts. I did lots of keymap-related
fixes and (almost) all of them should be in 3.0
You probably have an XF86Config left over from an older install? Check
the keyboard section, use the 'kbd' driver, not 'keyboard' as in:

Section "InputDevice"
        Identifier  "Keyboard0"
(Continue reading)

Bruce O'Neel | 2 Jan 20:31 2006
Picon

Re: NetBSD 3.0 on Powermac 9500

Hi,
On Mon, Jan 02, 2006 at 10:57:28AM -0800, Kjell Konis wrote:

> 
> When the kernel is compiling, "-Wa,-maltivec" is always among the C  
> flags.  What does that comma mean?  Is it just for warnings or is it  
> actually including Altivec support?

-Wa, passes options to the assembler.  I poked but didn't find out
what -maltivec allows for assembler options beyond "Generate code for
processors with AltiVec instructions."

> 
> Also, I have an ABD AppleDesign keyboard.  It works fine under  
> Xmacppc but several of the keys (notably backspace, delete, and the  
> up, right and left arrow keys) don't work under XFree86.  Has anyone  
> gotten one of these to work XFree86?

If you start xev and move the cursor into the window, what does 
typing those keys do?  If key are recognized you can use xmodmap to 
enable those keys.

I use

keycode 74 = Pointer_EnableKeys
keycode 76 = Pointer_Button1
keycode 95 = Pointer_Button2
keycode 96 = Pointer_Button3

to make turn on and off the use of f10, f11, and f12 as the left,
(Continue reading)

Kjell Konis | 2 Jan 21:31 2006
Picon
Picon

Re: NetBSD 3.0 on Powermac 9500

Okay.  I guess my real question then is should this be being passed  
when I have commented out Altivec support in the kernel  
configuration?  I thought it might be a bug in the makefile or  
something.

Kjell

On 2 Jan 2006, at 11:31, Michael wrote:

>> When the kernel is compiling, "-Wa,-maltivec" is always among the C
>> flags.  What does that comma mean?  Is it just for warnings or is it
>> actually including Altivec support?
>
> This means that '-maltivec' is passed to the assembler as command line
> option.

Kjell Konis | 2 Jan 21:57 2006
Picon
Picon

Re: NetBSD 3.0 on Powermac 9500

Thanks, this did the trick.  I had pc101 instead of pc105 for my  
XkbModel.

Kjell

On 2 Jan 2006, at 11:31, Michael wrote:

> Both my UMAX labeled Extended II and the built-in keyboard in my iBook
> work just fine without any xmodmap stunts. I did lots of keymap- 
> related
> fixes and (almost) all of them should be in 3.0
> You probably have an XF86Config left over from an older install? Check
> the keyboard section, use the 'kbd' driver, not 'keyboard' as in:
>
> Section "InputDevice"
>         Identifier  "Keyboard0"
>         Driver      "kbd"
>         Option      "XkbRules" "xfree86"
>         Option      "XkbModel" "pc105"
>         Option      "XkbLayout" "us"
>         Option      "Protocol" "wskbd"
>         Option      "Device" "/dev/wskbd0"
> EndSection
>
> Yes, the XkbModel statement is a blatant lie but XFree translates
> everything into some PC-derived set of scancodes before passing  
> anything
> to xkb.
>

(Continue reading)

Michael | 2 Jan 23:27 2006
Picon
Picon

Re: NetBSD 3.0 on Powermac 9500

Hello,

> Okay.  I guess my real question then is should this be being passed
> when I have commented out Altivec support in the kernel  
> configuration?  I thought it might be a bug in the makefile or  
> something.

Depends, -Wa,-maltivec only enables AltiVec mnemonics for the assembler
- it shouldn't actually produce AltiVec code on its own ( in fact gcc3
can't use AltiVec without help, passing -maltivec to the compiler
enables a few datatypes and C-mappings for AltiVec instructions but it
won't generate AltiVec instructions from regular C code ). But there's
definitely something fishy with 3.99.15 - all newer kernels cr*p out
right after startup on my G3, some kernel option must trigger it since
GENERIC works. 

options ALTIVEC just makes sure the kernel knows and cares about AltiVec
registers on context switches so they can be safely used in userland
code if I remember correctly. Not sure if it does anything else.

have fun
Michael
netbsd | 3 Jan 18:45 2006
Picon

pkgsrc and the T1000


Has anyone here tried a pkgsrc build on the 'niagra' based T1000 I am 
anxious to get a toe-hold on this new platform, but installing only sun 
binaries from scratch isnt my idea of a Good Time.

They have a try it for 60-days before you decided to buy it plan.

Unlike the Sparc chips i havent seen any OEM boards hit the market yet.

  Microsoft: Where do you want to go tomorrow?
  Linux: Where do you want to go today?
  BSD: Are you guys coming, or what?

Robin-David Hammond	KB3IEN
 	www.aresnyc.org.


Gmane