[9fans] Introduction and Grandiose Plans

[9fans] Introduction and Grandiose Plans

[Daryl M]

[-- Attachment #1.1: Type: text/plain, Size: 2112 bytes --]

Hello,



I have followed Bun-bun (www.sluggy.com <http://www.sluggy.com> ) for years
and started getting seriously interested in Glenda last year.  After a rocky
start, the relationship is starting to smooth out and I now have multiple
machines running Inferno (because it was easier) and recently got Plan9
mostly running (networking is still flakey) in a VirtualBox VM.  I've spent
quite a while reading the documentation from the Wiki, cat-V, and the 9atom
sites as well as browsing the 9fans archives.



My first goal is to produce a UEFI boot loader for Plan9.  Once that is
accomplished, I intend to start incorporating parts of Plan9 (and/or
Inferno) into the core UEFI firmware (www.tianocore.org
<http://www.tianocore.org> ).  I would like to have Channels, the Plan9
device abstraction and 9P protocol committed to tianocore.org later this
year.  The device abstraction code is actually being reviewed right now so
might be committed this month.



Any pointers to information on Plan9's boot loader requirements or design
would be greatly appreciated.  Especially information about why the existing
virtual memory map was chosen.



My interest in Plan9 and Inferno has taken off lately because of the IoT
Fad.  Security and authentication seem to be better integrated into these
two OS' than (supposedly) more modern attempts.  The distributed nature of
the systems and the 9P protocol seem well suited to the requirements for the
"Internet of Things".  Also, both OS kernels appear to be relatively small
and have minimal demands upon the supporting hardware infrastructure.



Thanks again to the people on this list.  You've been a great help and I
hope to begin contributing back real soon.



PS:  If one is using VirtualBox, the version really matters.  Under v4.2.6 I
was unable to boot either the Plan9.iso or 9atom.iso.  I switched to v4.3.10
and both started working with no other changes to the VM.  I should really
switch to QEMU, but inertia is a powerful force.



Sincerely,

Daryl M.

Portland, Oregon   USA










[-- Attachment #1.2: Type: text/html, Size: 4852 bytes --]

[-- Attachment #2: image001.jpg --]
[-- Type: image/jpeg, Size: 10650 bytes --]

Re: [9fans] Introduction and Grandiose Plans

[erik quanstrom]

> My first goal is to produce a UEFI boot loader for Plan9.  Once that is
> accomplished, I intend to start incorporating parts of Plan9 (and/or
> Inferno) into the core UEFI firmware (www.tianocore.org
> <http://www.tianocore.org> ).  I would like to have Channels, the Plan9
> device abstraction and 9P protocol committed to tianocore.org later this
> year.  The device abstraction code is actually being reviewed right now so
> might be committed this month.

cool!  that would certainly make my current side project easier.

> Any pointers to information on Plan9's boot loader requirements or design
> would be greatly appreciated.  Especially information about why the existing
> virtual memory map was chosen.

although there are a number of bootloaders, they're all pretty similar.  broadly
they need to
1.  load the kernel at a known address.  some kernels would like a multiboot header,
some would not.  (i'm personally not a fan).
2.  put the plan9.ini configuration at CONFADDR
3.  append runtime information to the plan9.ini configuration at CONFADDR
gathred by the bootloader such as
	- the *e820 map (for pcs).  i'm willing to get things in a new format,
	as i understand uefi has a sightly different (just extended?) format.

	- information about the boot drive, or boot nic in the case of pxe.
	this helps automaticly configure things correctly.
0.  (pcs only) some kernels prefer the loader to enter 32-bit mode for them.
4.  jump to the kernel.

that's it.

> My interest in Plan9 and Inferno has taken off lately because of the IoT
> Fad.  Security and authentication seem to be better integrated into these
> two OS' than (supposedly) more modern attempts.  The distributed nature of
> the systems and the 9P protocol seem well suited to the requirements for the
> "Internet of Things".  Also, both OS kernels appear to be relatively small
> and have minimal demands upon the supporting hardware infrastructure.

they certainly do.  inferno is interesting because you don't even need a mmu.
for that matter, if you strip userland off the normal kernel, which isn't too hard,
the plan 9 kernels can often run without an mmu.

- erik