Hi,

Could we please drop the audio support from the GENERIC
and INSTALL kernels for macppc until it *works* ? And while
there mark it as "experimental" ?

Regards,

A.

At 21:49 +0100 8/1/2002, Andrea Cocito wrote:
>Date: Tue, 8 Jan 2002 21:49:28 +0100
>To: <port-macppc@netbsd.org>
>From: Andrea Cocito <acocito@lar.ieo.it>
>Subject: Audio support in generic kernel (1.5.2)
>
>Hallo,
>
>Installing NetBSD on some models of G4 (blue digital audio
>w/o gigabit) is essentially impossible for who has only one
>machine.
>
>The GENERIC kernel has the audio support, but the audio support
>is broken and on these models a number of spurious interrupts
>floods the kernel. Depending on the model and the configuration
>of the machine the thing has an impact varying from a "very slow"
>machine to an almost-frozen one.
>
>On a machine here the system was so slow that it had not finished
>compiling a 12k C source after 45 minutes (100% cpu usage steady)
>and this is only an example, all the system's "performance" was
>consistent with this.
>
>After gathering around some note about the problem in the audio
>driver I moved the hard disk to another machine (same series,
>but different processor, 466 mhz instead of 400). The note I
>found spoke about a slowness of the hard disk caused by the audio
>driver, what I experienced is a cpu problem.
>
>I actually had the same  problem there but with less impact,
>and was able to compile a new kernel without audio support
>in *only* 3 hours, then moved back the hard disk to the original
>machine... and  there it worked nice (it's happily running ldap +
>qmail + postgreSQL + kerberos plus a few other little things in
>this moment, and does a "make clean ; make depend ; make -j 10"
>of the kernel in a matter of minutes).
>
>It would be a good idea to place in the distribution/ftp site
>a kernel WITHOUT audio support. It wouldn't have been possible
>to rebuild the kernel on the machine affected by this problem.
>
>While there: sholdn't the limit on the physical amount of RAM
>allowed (doesn't work with 1GB) and on the total addressing
>space (doesnt work with an amount of RAM+swap greater than
>2 GB) be documented somewhere ? With nowaday's machines is quite
>likely that people gets stuck by having a machine with 2 gigs
>of ram, or one gig of ram and two of swap, not booting.
>
>Regards,
>
>Andrea
>
>--
>Andrea Cocito
>Director of the Biocomputing Research Unit
>Department of Experimental Oncology
>Europen Institute of Oncology
>Via Ripamonti 435, Milano - Italy
>Tel. +39-02-57489857
>Fax. +39-02-57489851

At 17:23 +0200 15/5/2002, Andrea Cocito wrote:
>At 10:55 -0400 15/5/2002, Sean Davis wrote:
>>Thanks very much, I'll try that in a few minutes. The only other thing on my
>>mind, is that when my friend was installing the machine, and it was booting
>>up, going through init, etc etc, he said the console was *extremely* slow,
>>about one line every 3 seconds or so. The machine is a G4 350Mhz, with 512MB
>>of RAM, so it sounds like that shouldn't be happening. The only option I
>>found in GENERIC that sounded related is OFB_ENABLE_CACHE - will enabling
>>that option make console run better?
>
>I think I had reported this problem twice already... it's not a
>console issue, it's a system performance issue caused by a bug in
>the audio drivers.
>
>Some machines (notably the blue G4s ranging from 400 to 500 mhz) have
>a severe problem with the sound driver built into the GENERIC kernel,
>they end up having from 4000 to 50000 interrupts per second by the
>audio chipset.
>
>*IF* possible on the machine affected by the problem (4 out of 5 tested
>here, the only one not having the problem was a cube... maybe because it
>does not have onboard sound ? :) all you have to do is to recompile a
>kernel without audio support and everything runs smoooth then.
>
>Your mileage may vary in this: one of the machines with the problem
>managed to compile a kernel in *only* six hours, another one couldn't
>compile an "Hello world" thing in two hours and half (cpu usage 100%
>steady by cc1...) ... and of course compiling the kernel there was
>out of discussion. I moved a generic kernel without audio stuff and
>then the machine recompiled a coustom kernel in a matter of minutes.
>
>As an emergency solution while someone can take a look at it I again
>suggest strongly to remove the audio support from the first-installation
>kernel.


At 9:22 -0500 6/6/2002, Chris Zwilling wrote:
>Hello!
>
>I have a 350MHz G4 with Gigabit ethernet (? gm0).  I am recieving about
>150000 interrupts per second on IRQ 1 and IRQ 9.  From the dmesg output I
>am gathering that the awacs? is generating the interrupts.  Is this
>normal?
>
>issac: {6} systat 2 vmstat
>Interrupts
>330509 total
>203834 irq1
>135553 irq9
>
>issac: {7} dmesg | fgrep irq
>zsc0 at obio0 offset 0x13000: irq 22,50
>awacs0 at obio0 offset 0x14000: irq 24,1,9
>adb0 at obio0 offset 0x16000 irq 47: 0 targets
>wdc0 at obio0 offset 0x1f000 irq 19: DMA transfer
>wdc1 at obio0 offset 0x20000 irq 20: DMA transfer
>wdc2 at obio0 offset 0x21000 irq 21: DMA transfer
>ohci0: interrupting at irq 27
>ohci1: interrupting at irq 28
>gm0: interrupting at irq 41
>
>
>;--------------------------------;
>; The problem with reality is    ;  chris@nukequarters.com
>;  that there isn't a coherent   ;  http://www.nukequarters.com
>;  soundtrack.			 ;
>;--------------------------------;