Re: awacs driver

Michael Lorenz wrote:

> the hardware notes claim that the awacs driver can
cause trouble on 
> some G4 Macs, hints about interrupt storms. I've never
had any trouble 
> with this driver but I don't have an awacs-equipped G4
either - just a 
> G3 and a PowerBook. My guess is that those problems
were caused by the 
> old, horribly buggy interrupt code which has been
improved a lot in the 
> last two years.

Unfortunately there is still a problem with it, at least
with NetBSD 3.1.1.

I like to hear to a radio live stream while compiling and
developing, and
after a few minutes it happened: complete lockup of the
system! The only
life-sign was that I could still ping the machine, but not
log-in.

This is a PowerMac G4 (Sawtooth) 400 MHz with 1.25 GB RAM.
The load was not
too high (mainly caused by Firefox). I was just starting a
new live stream on
xmms when it happened.

Did the interrupt code improve from 3.1 to 4.0, or in
current? Then I might
test that again.

-- 
    _  Frank Wille (frankphoenix.owl.de)
 _ //  http://sun.hasenbra
ten.de/~frank/
 X/   Phx  #AmigaGer

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

- -----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hello,

On Nov 4, 2007, at 07:14, Frank Wille wrote:

> Michael Lorenz wrote:
>
>> the hardware notes claim that the awacs driver can
cause trouble on
>> some G4 Macs, hints about interrupt storms. I've
never had any  
>> trouble
>> with this driver but I don't have an awacs-equipped
G4 either -  
>> just a
>> G3 and a PowerBook. My guess is that those problems
were caused by  
>> the
>> old, horribly buggy interrupt code which has been
improved a lot  
>> in the
>> last two years.
>
> Unfortunately there is still a problem with it, at
least with  
> NetBSD 3.1.1.
>
> I like to hear to a radio live stream while compiling
and  
> developing, and
> after a few minutes it happened: complete lockup of the
system! The  
> only
> life-sign was that I could still ping the machine, but
not log-in.

Ouch, that does look like an interrupt deadlock.

> This is a PowerMac G4 (Sawtooth) 400 MHz with 1.25 GB
RAM. The load  
> was not
> too high (mainly caused by Firefox). I was just
starting a new live  
> stream on
> xmms when it happened.

Ok, my G4 isn't too different from that ( AGP, 2x 500MHz G4,
896MB )  
- - - let's see if I can reproduce it. I've been listening
to music a  
lot, maybe network activity at the same time triggers it.

> Did the interrupt code improve from 3.1 to 4.0, or in
current? Then  
> I might
> test that again.

All the changes I did to the interrupt code prior to the
ppcoea- 
renovation branch dealt solely with old world interrupt
controllers,  
all G4s have OpenPICs so those changes wouldn't have any
effect. The  
interrupt code was more or less completely rewritten to be
more  
sharable between powerpc ports so it might be interesting to
see what  
happens with fairly recent -current.

have fun
Michael
- -----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.7 (Darwin)

iQEVAwUBRy3t7cpnzkX8Yg2nAQID2Qf/cm6VmS5UlKgkuttN97tYQzIG3D+E
yGUd
0+HO6CpggORBJebD4MUOhE60mPo72GO/vCWof5894G5UGi6dTZlaOazKTJg1
uzEf
fOIezIgXFjSg+Pj/KWfJwqXgFjx2okWH9OEd1Wx2fMxwwLXhhQCvwWHE0BeD
vUNC
T89NsvfkXIdhLr48Mkzoy2LzQ2P+QoFld9KlVhyP8LnG9jdgWRkJ+bJfyxxZ
opjQ
u+eDyVnlrOSYzVNXH0aVOsHwZkt+EPSHwGalssMJ4OD3VO4dhOWuTB5dqOoY
U0pE
hwsXivocW09DAZ7PBqLJ4ncIGWUHhb/l1eU6QBHAiKgwVHjFPGTcNA==
=ti72
- -----END PGP SIGNATURE-----
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.7 (Darwin)

iQEVAwUBRy3uAcpnzkX8Yg2nAQKZawf+L+ArOhQdt34aM2djQb+nYSgtCWBm
Xazb
OqQpKJItNWWDKcItqXgqOTBbT2swlvyrl0CThoefNskHEV7+I31JqsIoqt8A
JOlV
F6D0zO6ro+ywTMltGLXlHgLkh8hzof8jtO/OFgsyqUpm/d4/Gmx2VqqVnmiA
pb5n
UUS1KM926kcbboF8U/eFspdba3htpEbIfnc3QPsZSgEs5wZ6it3uGACXStF5
26jV
Zmu64XJxMVAQwVGBnOCIU8IV9NUik4giCnzdswvPoCE3adKDM4bUqZ0DgbJp
rdBm
P2DbUv/jkAFkBAiCfNTikcW388opWzHOMTTtxeGjposxI+5yszt3fw==
=lTY2
-----END PGP SIGNATURE-----