[Simh] VAX emulator slow to configure extra disks

[Mark Pizzolato]

On Sunday, June 23, 2019 at 6:53 AM, Johnny Billquist wrote:
> I can't remember having any such issues, but then again I also can't
> remember if I ever tried it with VAX. With PDP-11 it works without any
> issues.
> 
> This is, though, another slight silliness in simh. The limit to 4 disks
> per controller is very artificial. The UDA50, KDA50, and maybe some
> other controllers only have four actual, physical ports, which limited
> them to only have four disks per controller. However, MSCP itself do not
> have such a limitation, and common SCSI controllers for these machines
> which use MSCP allows more than four disks on a controller, and most
> software (at least RSX and VMS) also do not limit themselves to only
> configure max four disks on a controller. I wish simh didn't put such
> arbitrary limitations in. simh also decides on unit numbers by itself,
> which could also have been nice to be able to choose.

As Bob said, simh is modeling hardware that existed.  It does model the 
bus and controller interface.

As Tim mentioned, some of the hardware that is modeled did indeed 
allow arbitrary unit numbers (via plugs on the drive).  

Some 14 months ago support was added to provide per drive Unit plug 
values to be set.   This is set via:

    sim> SET RQn UNIT=plug

plug can be any value from 0 thru 65534.  Default unit plug for each 
RQn is n.

A few subtle changes were needed to pdp11_rq and pdp11_tq to 
properly allow VMS and the KA655 boot ROM to stop their scan
for units when all units have been found.


> On 2019-06-23 15:37, Jonathan Welch wrote:
> > A vax emulator I was configuring for a friend (MicroVAX 3900 simulator
> > V4.0-0 Current git commit id: ab3e07a4) needs to mount more than four
> > disks.
> >
> > The .INI file has
> > set rqb enabled
> >
> > and an associated attach command.
> >
> > In my VMS startup disks are mounted very close to the beginning of the
> > boot sequence.
> >
> > I discovered that when the emulator starts it is necessary to put a
> > $ WAIT 00:00:06
> > before mounting DUB0; the wait is necessary for the disk to appear.
> >
> > Is this something that is expected behavior and should be documented
> > or a bug that can be fixed?

As Tim mentioned, the VMS drivers are loaded in a DEC provided procedure
with a "sysgen autoconfigure all" command.  Once the drivers are loaded they
asynchronously interact with each controller probing for available units.  
This is how things always have worked.  If you attempt to reference a unit 
before it has been found, the failure you are seeing will happen.  
If you attempt to boot VMS standalone backup, you will see that there is 
a very long  pause before a device list is presented and you are ultimately 
prompted with:

    Enter "YES" when all needed devices are available:

This is to make sure that all devices have had time to be identified before 
you proceed with the backup.

- Mark