mac kernel issues (was Re: nbd vs. failing HDDs)
On Fri, 9 Sep 2011, Brad Boyer wrote:
> Anything that predates the NuBus 90 standard is definitely 10MHz only. I
> haven't looked at any newer documentation, but YANCC was Apple's first
> NuBus 90 compatible bridge chip and it first shipped in the Q700 and
> Q900. Even in that case the processor couldn't access it at 20MHz due to
> the design limitations, but as long as it was a NuBus card doing the
> access it could use 20MHz. I believe later systems removed this
> limitation, but I don't have documentation saying when that happened.
I see. Thanks.
> Well, the systems using PIO have real DMA available if we could get it
> working properly. The AV systems might be easier since the mac53c94
> driver has DMA support, although it would need some changes to work with
> the PSC instead of the DBDMA engine of a PowerMac.
...
> Is fixing SCSI high on the list of priorities? I keep meaning to learn
> more about SCSI and fixing the NCR5380 driver would certainly encourage
> that due to the chip doing next to none of the work itself.
>
> I don't have a lot of time to look at anything, but I am curious what
> you see as the highest priorities in case I found more time.
The top of the list of my priorities are the devices marked borken on the
machine status page. That's basically v1 PMU and Egret ADB (broken since
2.2 and flaky even back then) and, as dicussed, NCR5380 SCSI.
As you point out, doing PIO on a machine that supports DMA is just sad.
All of those machines have 40 MHz CPUs. Fixing this rates in my list of
priorities.
Probably next would be the system controller mess, which as you can see
from the other thread, is not merely a code cleanup exercise, it also
prevents the installer initrd working on powerbooks. Also nubus needs to
get a proper device driver model implementation.
There's more on my list (quite a long tail of small issues) but nothing
serious. I'll enter them into the sourceforge bug tracker at some stage.
Finn
>
> Brad Boyer
> flar@allandria.com
>
>
Reply to: