Minerva4Q68

[Derek_Stewart]

Hi,

I downloaded the most recent version of Minerva4Q68 v1.65

I loaded the Minerva4Q68 Q68_ROM.SYS to the root of a FAT32 partition on a SDHC Card.

Booted the Q68 v1.05, up all works perfectly, except, I did not install Toolkit 2. But DISP_MODE 4 switched to 1024x768 Mode 4 okay.

I tried to expand the MINERVA4Q68.zip, to a QLWA file on a SDHC card, with SMSQmulator. But I had to shorten the top level directory as ACP did not like the file length, to MIN4Q68

So should be able to assemble Minerva on the Q68.

I tried another Q68 v1.05, Batch 4 with Minerva, all worked OK

I have received Dilwyn's Q68 which seems to boot up with no cursor. So a little more investigation required.

[dilwyn]

I have received Dilwyn's Q68 which seems to boot up with no cursor. So a little more investigation required.

Sounds a little different to my experience here. For me, it'd rarely get as far as booting into Minerva. Even when it did, it locked up before you could type many characters.
(Not moaning too much, SMSQ/E worked OK for me, though I hadn't had time to try 3.39 before sending it to Derek).

In case it had a bearing, I was preparing the SD card with Windows 10, with the Minerva copied from a folder on drive D: of my PC after formatting the card and configuring. Shouldn't make a difference, that's how I prepared the SMSQ cards too.

And in case the PSU used makes a difference in terms of rating or smoothness of power, the one I usually use is a 2.1A phone charger, though I've tried a couple of different ones with the same results.

[janbredenbeek]

I have received Dilwyn's Q68 which seems to boot up with no cursor. So a little more investigation required.

Mark Swift has done a lot of research on this, see https://github.com/janbredenbeek/Minerva4Q68/issues/1

He was able to type a few lines on his Q68 before freezing. From the video it looks like the Q68 suddenly stops all interrupts (or it keeps hanging in an interrupt service routine).

Does SHIFT-CTRL-ALT-TAB work under Minerva? (should reboot Minerva)

One of the few remaining options (apart from a hardware issue) might be the Ethernet or mouse interrupt being asserted, for which the current version has no inbuilt drivers. But these should already have been disabled at startup:
(from ss_init.asm):

Code: Select all

      GENIF Q68 <> 0          ; Q68 hardware initialisation (from SMSQ/E)

        lea     q68_sramt,a3
;       st      led-q68_sramt(a3);      LED on
        sf      q68_ethi-q68_sramt(a3)  ; block CP2200 interrupts v. 1.03
        and.b   #$1f,uart_status-q68_sramt(a3) ; disable serial interrupts
        and.b   #$7f,mouse_status-q68_sramt(a3) ; disable PS/2 mouse interrupts
	sf      kbd_unlock-q68_sramt(a3) ; no key may be got
        move.b  #$1f,pc_intr-q68_sramt(a3) ; no xmit/intf/gap interrupts
        sf      pc_tctrl-q68_sramt(a3)  ; clear transmit control reg

        move.l  #q68_sramb+4,q68_sramb-q68_sramt(a3) ; first free space in fast sram mem
      
      ENDGEN * Q68 <> 0 * 

(I noticed that the latest version of SMSQ/E has removed the 'sf kbd_unlock' instruction - but this shouldn't make a difference).

Looking forward to seeing your results...

[Derek_Stewart]

Hi,

I am waiting on Mark's Q68 being returned for furthepr investigation.

I have 2 Q68 with same build and Q68 ROM Loader v1.05, both seem to load Minerva from Q68_ROM.SYS or from SMSQ/E. A bit strange.

I only use Windows 7 to programme the FPGA, the Q68 is powered from a 1A power supply.

The SD Card is generated in with Linux.

I only used Windows 10 for the use of a CAD programme, but there is now Linux version thst does job as good, so no more Windows...

One question what does Minerva do different to SMSQ/E?

This difference is the probable reason for intermittant problems.

[janbredenbeek]

I have 2 Q68 with same build and Q68 ROM Loader v1.05, both seem to load Minerva from Q68_ROM.SYS or from SMSQ/E. A bit strange.

I have updated the FPGA to v1.05 and running Minerva with no problems.

One question what does Minerva do different to SMSQ/E?

SMSQ/E has all drivers built-in, whereas Minerva has to rely on external drivers (either LRESPRed or in external ROM images). So the framework is different.
In addition, there is little documentation available about the internals of the built-in drivers, other than the (scarcely commented) source code.

[Peter]

In addition, there is little documentation available about the internals of the built-in drivers, other than the (scarcely commented) source code.

Given this fact, you do an amazing job!

[janbredenbeek]

In addition, there is little documentation available about the internals of the built-in drivers, other than the (scarcely commented) source code.

Given this fact, you do an amazing job!

Thanks!
Now a bit of bad news: I just discovered that SMSQ/E's Q68 FAT driver contains a bug that corrupts files read from the FAT16 partition on the SDHC card
I ran into this when LRESPRing Minerva from SMSQ/E failed randomly.

Proof of concept: Load a sufficiently large file using LBYTES and run a CRC32 check on it (you can use the crc32.bin file on my GitHub repository; https://github.com/janbredenbeek/Minerva4Q68). E.g.:

Code: Select all

a=ALCHP(350000)
LBYTES fat1_Q68_SMSQ.BIN,a:PRINT CRC32$(a, FLEN(\fat1_Q68_SMSQ.BIN))

Repeat the last line a few times. The CRC32 is mostly different each time! (The larger the file, the more times the CRC32 differs). Having taken a quick peek, it looks like some parts of the file are missing (e.g. a block of 16 bytes).

The bug was introduced in v3.37, older versions are not affected. Incidentally, this was the version that introduced 40MHz SDHC clock option (which I had disabled during the test). The good news is that WIN drives doesn't seem to be affected.

It looks like we have some debugging to do...

[Derek_Stewart]

Hi,

Have you reported this to Wolfgang to get the problem fixed?

[Derek_Stewart]

Hi,

I have loaded Q68 v1.02 FPGA code onto a Batch 4 Q68, and copied a Q68_ROM.SYS, with integrated Toolkit 2, to a freshly formatted SD Card with a FAT32 partition.

The Q68 boots up to Minerva and links Toolkit 2, faster than my monitor startup.

The system seems to work in 512x256, 1024x768, also loads the Extended Environment. All seems to work.

One question regarding the mouse, vent though there is no mouse driver, would moving the mouse cause any interrupts on the operating system?

This was a problem on the Aurora.

I just need to test loading Minerva from SMSQ/E, and assemble the ROM on the Q68. Which be straight forward, using existing programming from the Minerva4Q68 repository distribution.

[janbredenbeek]

Have you reported this to Wolfgang to get the problem fixed?

It was too late last night, but I have just mailed Wolfgang.