Moin Mike,

20.01.95 07:14, you wrote a message to Vitus Jensen:

 JG>>> What is it?  Where can I get documentation on making my
 JG>>> device drivers RM aware?

 MB>> If you have an account on the IBM Device Driver Support Center 
 MB>> BBS, "The DUDE," (407)982-3217, download RMBASE.ZIP. 
If you do 
 MB>> not have an account, you need to set one up.  THIS BBS IS ONLY 
 MB>> FOR DEVICE DRIVER WRITERS.
 VJ>> [...]

 VJ>> [...Explain?...]

 MB> I don't know since I am in the United States, but the IBM Hursley 
 MB> group in the U.K. has some responsibility for device drivers, so 
 MB> they may have developer support.  The minimal header files to use 
 MB> the Resource Manager are included on the latest DDK (cryptically 
 MB> and with no other documentation, since it was not ready when the 
 MB> CD-ROM deadline came around, in \DDKX86\SRC\DEV\RESOURCE).

cryptic: yes
I found a new IORB documented in \DDKX86\SRC\DEV\DASD\DISKH\IORB.H which
defines IOCC_RESOURCE. But MAX_IOCC (or what is called) stayed at
IOCC_ADAPTER_PASSTHRU and this IOCC_RESOURCE is never generated by OS/2 2.1
or 2.3.

 VJ>> Or better yet, let me say what I am looking for:
 VJ>> [...hide controller hardware...]

 MB> As far as I know, you have to write a VDD; this is how IBM 
 MB> protects the standard hard drive and floppy controllers from DOS 
 MB> software.  By setting "IOPL=YES" you are accepting an inherent 
 MB> risk from OS/2 programs.  Only 16-bit OS/2 programs can have 
 MB> IOPL, so that is a dying thing.

A VDD would be simple (already did one) but is rather ugly. A second file
and a second line in CONFIG.SYS. If OS/2 installation is done from CD-ROM
(by copying GDTX000.ADD to DISK 1) the VDD has to be copied afterwards in
an additional step. 

 VJ>> Will the "resource manager" handle those problems?
 MB> No, the Resource Manager is a co-operative registration facility 
 MB> that is intended primarily for the benefit of device drivers so 
 MB> that they can avoid potentially destructive probes of each 
 MB> other's hardware.  Although the RMVIEW tool shipped with Warp 
 MB> does dump this information to the user, that is just a by-product 
 MB> of the system.  I don't think the kernel enforces resource 
 MB> sharing in the way you would like, but making your ADD RM-aware 
 MB> could not hurt.

I could not hurt, but as this call might only be called by RMVIEW it is of
no real good. Nice feature, but...

I think IBM ought to use these resource informations to hide hardware. Or I
should write a VDD to do this kind of stuff? Hmmm...

BTW: is there any difference between HPFS and HPFS386 relating to IO
errors? If a drive stops operation and GDTX000.ADD tells HPFS (2.1GA) so,
HPFS would retry and (if not successfull) trap. Instead of disabling all
partitions on this drive.

Bye,
	Vitus

--- Sqed/32 0.97/r00228