[Info-vax] VSI: "Official 8.4-1H1 Launch"

[Stephen Hoffman]

On 2015-06-08 15:07:44 +0000, John Reagan said:

> On Monday, June 8, 2015 at 10:51:28 AM UTC-4, Bill Gunshannon wrote:
> 
>> This may come off as a really stupid question, but isn't VMS a 
>> monolithic kernel?  In which case VM and everything else is available 
>> as soon as the secondary (or tertiary) boot loader loads the kernel 
>> into memory and jumps there.  Right?
> 
> Has not been a monolith since VMS V3 or so.

FWIW...  When viewed at the file level, OpenVMS is not a monolithic 
kernel.   But in strict computing architecture terms, OpenVMS is a 
monolithic kernel.   VAX/VMS V5.0 brought the modular executive, which 
means that OpenVMS is a monolithic kernel design with loadable modules. 
 (Unfortunately without much support for unloading kernel code, but I 
digress.)  Once OpenVMS is booted, the kernel is one big wad of code, 
which is made slightly more tenable through the use of three privileged 
processor modes.   Hybrid and modular kernels tend to use user-space 
processes, with the closest OpenVMS analog being ACPs.   In short, 
there's no general IPC communications layer used within the OpenVMS 
kernel, and most OpenVMS kernel stuff is not implemented as ACP-like 
constructs or UWSS-like constructs.  So... OpenVMS is monolithic, with 
loadable bits.

Prior to V5.0, there were parts of the kernel environment that were 
loaded during the bootstrap, or at run-time.  These included device 
drivers.

For those that are curious, the Internals and Data Structures books all 
have details of the bootstrap and system initialization.

Now as for what Bill was asking, OpenVMS initialization involves rather 
more than loading one whacking great system file.


-- 
Pure Personal Opinion | HoffmanLabs LLC