Re: netboot loader not closing devices

[Charles Dickman <chd%chdickman.com@localhost>]

On Sat, Jun 6, 2015 at 7:32 AM, Felix Deichmann <m4j0rd0m0%gmail.com@localhost> wrote:
> Am 06.06.2015 um 12:01 schrieb David Brownlee:
>>
>> (General question to anyone & everyone) We seem to be having recurrent
>> problems with finding a safe place for the bootloader. Is there a
>> location guaranteed to be viable for a 4MB uVaxII and a much bigger
>> Vax with 100s of MB?

The old location of 0x2f0000 doesn't work with a generic kernel
because of the kernel's size.

The limits come from the size of the DMA map. From looking at the
sgmap for the VS3100, I think it might have the same 4MB limitation of
the uVaxII and III. Of course in that case there is no Qbus to eat up
more of it. VS4000/60 can map 16MB (32k entries vs 8k).

> ... or is there an elegant way for the bootloader itself to determine
> already used/unused memory during runtime?

The bootloader doesn't know about the DMA map and so boot and the
kernel have to fit in the area mappable and starting at 0x00000000.

> The VCB01/QVSS for example also has switches for changing its memory
> starting address, so you can never be sure...
> (Though only the default switch setting starting at 0x3c0000 as mentioned by
> Charles seems to make sense.)

I also have a KZQSA I have been wanting play with. It has a memory buffer too.

> Felix

-chuck