A1200 + GVP 1291 SCSI issues after SetPatch

[techhusky]

Hi all,

I have an A1200 running OS 3.2.2 with a GVP 1230 Turbo+ II 68030 accelerator with an attached 1291 SCSI controller.  The 1291 provides an external DB25 connector for external SCSI devices.  I've been trying to get it to work with a ZuluSCSI Pico Slim for weeks.  The GVP driver kept reporting any device configured (HDD or CD-ROM) on the ZuluSCSI as a tape drive.  Some debugging showed that the device inquiry data was not being properly reported by the gvpscsi.device driver (version 5.3), even though the ZuluSCSI was returning it properly.  I'm not sure why I didn't think of this sooner, but just today, I finally decided to try booting without a startup-sequence.  The ZuluSCSI worked perfectly.  I was able to narrow the failing point down to the SetPatch call during startup.  I am assuming after patching something is either stomping on the gvpscsi.device or has changed a dependent driver or library enough to break it.

Any suggestions for workarounds for this type of situation?  OS 3.9 SetPatch has the ability to SKIPROMUPDATES for certain devices and libraries.  I don't see similar capability on OS 3.2.2.

Thanks,

-Jeremy

[Boing-ball]

Hi all,

I have an A1200 running OS 3.2.2 with a GVP 1230 Turbo+ II 68030 accelerator with an attached 1291 SCSI controller.  The 1291 provides an external DB25 connector for external SCSI devices.  I've been trying to get it to work with a ZuluSCSI Pico Slim for weeks.  The GVP driver kept reporting any device configured (HDD or CD-ROM) on the ZuluSCSI as a tape drive.  Some debugging showed that the device inquiry data was not being properly reported by the gvpscsi.device driver (version 5.3), even though the ZuluSCSI was returning it properly.  I'm not sure why I didn't think of this sooner, but just today, I finally decided to try booting without a startup-sequence.  The ZuluSCSI worked perfectly.  I was able to narrow the failing point down to the SetPatch call during startup.  I am assuming after patching something is either stomping on the gvpscsi.device or has changed a dependent driver or library enough to break it.

Any suggestions for workarounds for this type of situation?  OS 3.9 SetPatch has the ability to SKIPROMUPDATES for certain devices and libraries.  I don't see similar capability on OS 3.2.2.

Thanks,

-Jeremy

Have you tried the following after backing up your current configuration

1).. Installing ThorsMMU libs in expert mode. As this has configurable options for Phase 5, ACA and GVP accelerators

2j.. Why still on AOS 3.2.2 when 3.2.3 upgrade is available. May have a fix for SetPatch there.

https://www.hyperion-entertainment.com/

[techhusky]

Thanks for the reply.  Much appreciated.

I purchased the system with OS 3.2.2 ROMs and the OS installed, so I didn't have a registered version to get the 3.2.3 upgrade.  I threw some coin at Hyperion and purchased the digital version of OS 3.2 yesterday.  I then mimicked my A1200 + GVP A1230/1291 OS 3.2.2 setup with UAE.  Running on UAE shows the exact same problem.  I pulled down and installed the 3.2.3 update and ROMs on my UAE image and still no luck.  I'm going to try installing on real hardware after I burn the new set of 3.2.3 ROMs.  I'm not hopeful, though.

MuLibs is configured to execute GVPInit.  Unfortunately, it also does not resolve the issue.

The same setup with OS 3.1, 3.5, and 3.9 all work flawlessly on UAE.  I may go back to 3.1 ROMs and OS 3.9 on the real hardware until this issue is resolved or a workaround can be found.

I reached out to Hyperion support but haven't gotten a response.

[techhusky]

Hyperion support contacted me and they were able to use the UAE config I provided to determine the issue was, in fact, related to mmu.library.  OS 3.2.x ships with MMULib and is setup to load mmu.library via SetPatch.  Per recommendation, I contacted the MMULib author.  Thomas was very responsive and helpful.  (The Amiga community is the best!)  He was able to determine the cause of the issue (via emulation) and provided a solution.  Essentially, the GVP 1230 Turbo+ II is using an undocumented memory location that needed to be mapped by mmu.library.

If you have a GVP A1230 68030 w/MMU and A1291 SCSI Expansion, adding the following line in envarc:MMU-Configuration will allow the controller to behave properly:

; For GVP A1230/A1291
For 2017 11 SetCacheMode 0x03000000 0x0400 Valid IoSpace CacheInhibit

Thomas is looking into a change to GVPInit to integrate this configuration, but the above config will fix the issue for now.  Lastly, I was able to reproduce the issue on OS 3.9 after installing MMULib.  As soon as mmu.library loaded via patched SetPatch or MuTools SCSI devices no longer worked.  The same MMU-Configuration addition works on OS 3.9.  My external ZuluSCSI is now working great on the A1291.

-Jeremy