[prev in list] [next in list] [prev in thread] [next in thread] 

List:       linux-vm
Subject:    Re: Plug and Pray when native?  CVT?
From:       Bertus Bekker <BertusB () PQAFRICA ! CO ! ZA>
Date:       1999-07-30 6:38:40
[Download RAW message or body]

Missed the bit on the IODF.

PNP should be the way to go. Most modern 390 devices support self
description, which supplies loads of info about the device.

It is possible to have 64k devices(256 channels*256 devices per channel), so
it will probably not be worthwile looping through so many SSCHs to see what
there is connected out there. Best might be to wait for someone wanting to
use a device or for an interrupt from the device side.

The difficult side will be the hardware IOCDS. This must be loaded into the
hardware at IML or power-on-reset time, but can also be dynamically replaced
from the software side, although I have never seen any documentation on
that. If the processor complex is running a modern IBM OS (eg MVS/SP V4+),
it can do the dynamic update, but LINUX/390 will have to be aware of that.
Most 390 hw vendors supply a standalone IOCDS gen facility as well, so IOCDS
support will not be crucial from day 1.

I would like to see that if any devices are predefined in LINUX/390, that it
be in a simple text parameter file. (with perhaps a clever utility to do
validity checking,  but no compilatlon)

Regards

Bertus Bekker

> -----Original Message-----
> From: Linas Vepstas [SMTP:linas@LINAS.ORG]
> Sent: 30 July 1999 06:54
> To:   LINUX-VM@VM.MARIST.EDU
> Subject:      Re: Plug and Pray when native?  CVT?
>
> It's been rumoured that David Alcock said:
> >
> > When Linux/390 executes natively, will it try to be Plug and Play or
> > read some type of I/O Gen?
>
> I dunno.  Of course, it would be nicest if it could somehow learn as
> much as possible about the system its booting on, and make inteligent
> default choices (that can be over-ridden with standard lilo-style
> options).  But as you say, its early to think about this.
>
> > You know how MVS has a CVT, Communications Vector Table, located at
> > offset x'10' in the PSA, Prefixed Save Area (storage that starts at
> > location 0 for the related processor).  What will Linux/390 have in that
> > field?
>
> Right now nothing. My booklet says that 0x10 absolute is the IPL CCW2
> and I haven't bothered yet to figure out what that's for.
>
> > This is really getting ahead of things but if we wanted to
> > allow some basic MVS programs to run, a simulated CVT control block
> > would be good here if nothing is planned for it.
>
> I've always assumed MVS programs would need considerable infrastructure
> to run; are you saying that the simplest ones just expect some control
> blocks in low memory?  Do such programs want to execute priveledged
> insn's, or no?
>
> --linas

[prev in list] [next in list] [prev in thread] [next in thread]