Re: Hurd rumpdisk on real hardware
Hello Mike,
On Thu, Sep 18, 2025 at 11:31:44AM +0100, Michael Kelly wrote:
> It is true also that Hurd/Rumpdisk boots succesfully on a 512MB Qemu virtual
> machine. The reason it works is that total memory is so low that there is no
> HIGHMEM memory segment at all. All memory except the small DMA total is
> allocated to the DIRECTMAP segment and there is sufficient in that one
> segment to support the initial boot and to permit sufficient spare to permit
> pageout to operate successfully. Your log shows only DMA and HIGHMEM
> segments only as well so it should operate as per my Qemu test.
This is great insight. Now I know why rumpdisk does not work on several of my
machines, and I can stop trying to find ways to make it work. Netbooks were
quite popular and they often came with a 1Gb configuration, but one can still
use the in-gnumach ide drivers and work with the size limitations that come with
it.
So it seems that rumpdisk on hurd 32bit has a bitter spot with RAM values around
1 Gb when it will not work, but is should work either below or above that. It
would be nice to document this. Can one edit the wiki via the web interface or
is cloning the git repository and submitting patches the preferred way?
https://darnassus.sceen.net/~hurd-web/contributing/web_pages/#index2h1
Regards,
João
Reply to: