[FreeVMS] Re: Development


Subject: [FreeVMS] Re: Development
From: Glenn and Mary Everhart (Everhart@gce.com)
Date: Mon Dec 02 2002 - 01:01:36 CET


David Elins wrote:
> I think VMS, at least in its original implementation had an interrupt-based
> kernel.
>
> That is, rather than being based on a master event-loop, after the initial
> boot, all process scheduling, resource allocation,
> and all the other junk an operating system cares about was based on
> handling interrupts.
>
> As I recall there were 64 levels of interrupts (and I am sure someone will
> correct me if i am wrong!) with higher numbers taking precedence (i.e. a
> higher-level interrupt could preempt a lower-level interrupt service
> routine). The highest 32 levels were reserved for hardware devices and
> there was a fairly high-level timer interrupt to (among other things)
> ensure that no process or thread hogged the processor.
>
> I agree with Chris - I think it would be fun to develop a new (VMS-like) OS
> based on new concepts (concepts that
> mirror the way the world works!).
>
> David Elins
>
>
> At 03:33 PM 11/30/2002 -0800, you wrote:
>
>>I think the first step is some contingent plans to divorce the kernel from
>>it's linux base. Although it gives an excellent framework for initial
>>development and surely some algorithms and routines would be quite
>>ridiculous to try to re-engineer or re-write but as in the case of lainos,
>>we wish to in fact make an independent system that duplicate vms, not
>>linux.
>>
>>On Fri, 29 Nov 2002 roart@nvg.ntnu.no wrote:
>>
>>
>>>On Thu, Nov 28, 2002 at 10:45:15AM -0800, Chris McKenzie wrote:
>>>
>>>>Count me in to. I'm working on porting NetBSD and a few emulators so I
>>>>have some experience in the design. I have read some books on VMS and
>>>>have used it in a very limited sense
>>>
>>>What do you want to do to start with?
>>>
>>>Regards,
>>>Roar Thronæs
>>>
>>>
>>>--
>>>Liste de diffusion FreeVMS
>>>Pour se désinscrire : mailto:freevms-request@ml.free.fr?subject=unsubscribe
>>>
>>>
>>>
>>>
>>
>>Sincerely,
>> Christopher J. McKenzie
>>
>> cjm@ucdavis.edu
>> mckenzie@cs.ucdavis.edu
>> H: +1 818.9917724
>> C: +1 818.4293772
>> 1815 Mesa Ridge Ave
>> Westlake Village, CA 91362
>>
>>
>>--
>>Liste de diffusion FreeVMS
>>Pour se désinscrire : mailto:freevms-request@ml.free.fr?subject=unsubscribe
>>
>>
>>
>
>
There are 32 levels of interrupts in a VAX, many of which are used in
VMS for scheduling. These are software interrupts. The idea is that
certain activities may be controlled by interrupt priority and auto-
vectored to the right handling routines rather than having a master
checking routine check for pending activity at each level in software.
A number of operations are faster when this is done. Note RSX11M and 11D
did not have this since the PIRQ register was not universal on pdp11.
They had to look at pending activity in software.

These software levels are all below the hardware interrupt levels in
priority. The number of hardware interrupt levels in Vax or alpha varies
by model and by i/o equipment but external hardware interrupts are not
used for scheduling (tho they do in general drive i/o).

Also, there is software priority in VMS of 0-31 but these are separate and
not coupled with IPL. Those priorities have 0-15 reserved for timesharing
and 16-31 for realtime. They all run at soft (and hard) ipl 0.

-- 
Liste de diffusion FreeVMS
Pour se désinscrire : mailto:freevms-request@ml.free.fr?subject=unsubscribe



This archive was generated by hypermail 2b25 : Mon Dec 02 2002 - 01:02:21 CET