Possible SD card access conflict and corruption of amiga.ini on Tang Console 138k

[t-miura]

Hi, first of all, thank you for this excellent project!

I recently got Tang Console 138K and latest NanoMig core and MCU Companion firmware mostly working fine.

However, sometimes "Save Config" hangs, and amiga.ini gets corrupted after the hang.
Even OSD is hang up, FPGA side seems to be still working except anything that requires access to the SD card.

Also entire filesystem on microSD card gets corrupted time to time after the hang/crash, rendering completely unreadable on Windows.
hopefully chkdsk fixes it while deleting 0-byte amiga.ini and replace with backup is required to get it back working.

It seems like it tends to happen when other function(especially disk, both floppy and hard drive) is accessing microSD,
and it's not specific to the cards I use(tested over 5 card and all experienced "Save Config" lockup)

It could be companion's issue, but as I only tested them with NanoMig, so I'd like to report it here.

Are there any possible way to diagnose such issue without soldering or connecting something to board's pin?
If such quick methods are available, I can reproduce the issue and report back here.
(If diag needs additional wiring, it may take some days as I need to buy items needed)

[vossstef]

@t-miura, Console 138k is a bit tricky as @harbaum and myself don't have that board at hand. Typically i develop on a Mega 138k Pro Board and adjust after sucessful implementation 'blind' to assumed fit on a Console 138k.

Do you use onboard integrated BL616 or external PiPico or M0S Dock?

I plan to update BL616 firmware to catchup latest CherryUSB stack release. Not sure does it help but it's a try.

[harbaum]

Yeah ... I know that the concurrent access of the mcu and the core itself can cause corruption. I really need to look into this. Debugging is rather tricky as I need to simulate both, the mcu and the core at the same time hoping to be able to reproduce and capture such an event.

[t-miura]

@vossstef, thanks for reply!
I'm currently using integrated BL616, so I can help you test new releases/or anything i can help with you on my Console 138k.
edit: Regarding CherryUSB, all my USB controller(OG Xbox 360, 8bitdo SN30 Pro and no-brand cheap joypad) failed to work both directly and via USB3.0 hub, hope next update might help!

@harbaum , thanks too!
yea, concurrent access is always a nightmare for any platform...as same as BL616, i can help you testing anytime if it's helpful!

sometime disk access on Amiga side itself fails and causes lockup and HDD image's corruption, not sure if it's same kind of root cause.
good ol' Ami-Tools comes in handy in terms of repairing Amiga-side disk issue, just like real machine which sadly i don't have...
(here in japan, Amiga weren't much popular for home use, mainly used for professional's workstation)

[t-miura]

hi, i might have found the solution for OSD lockup while i haven't actually tested.

For some 8-bit fun, I tried C64Nano, and experienced exact same "OSD Hangs on Save Config" issue.
but i found "Pause OSD" option on C64Nano, and enabling it seems to mitigate the issue, no hang on saving config on C64Nano.
https://github.com/MiSTle-Dev/C64Nano/blob/6744f2def95fb2b1332c4ab3c360d6ae5dfcff29/src/misc/sysctrl.v#L275
I wonder if I(or someone of course) add these Pause OSD function to Nanomig, at least Save Config issue might be solved.

While there's other issue like HDD-related, but i know HDD support is very early one, so some function(like format) might fail by other reasons.

I just setup GOWIN IDE, so maybe i can try out adding pause osd by myself today?

quick update: well...first time coding FPGA and it's kinda difficult...but I think adding "Pause OSD" will fix the issue on nanomig as i mostly confirmed C64Nano's that prevents hang on saving config.

[t-miura]

I've experimentally/partially implemented C64Nano-like "pause cpu when OSD is on" on my test env,
seems like it actually mitigates lockup on saving config, even the patch was too buggy...
at least my hunch made some sense, worth trying almost full day!

so, is it possible to fully implement/port such a feature into NanoMig?
I think it's beyond my abilities...

[vossstef]

@t-miura rather strange as i don't use anymore the Pause OSD = system_pause in the C64 core. It's a leftover from very early development. Maybe we are looking for another cause. Had a quick look in the xbox360 controller topic ands also have problems that need to be fixed.

[t-miura]

oh, that was unexcpected to me, and i lost my code with sketcy Pause OSD by messing up "somehow working" code...

if you or anyone else who don't have access to the TC138K, i can test and report back, hope it helps!
also i'm looking for another cause too, of course.

[harbaum]

While sure being a nice exercise, I don't think this pause approach is the way to go. I really need to debug and fix the root cause which definitely is both sources not properly being told to wait for the ongoing transfer to end.

[vossstef]

@t-miura here my 5 cents.
Bugfix of the BL616 Companion related to XBOX360 controllers to make them work again.
https://github.com/MiSTle-Dev/FPGA-Companion/releases/tag/v1.4.12

Large scale HID update for larger Gamepad support is still in progress and will last a bit.

[t-miura]

@vossstef thank you!
i just flashed into TC138K, good news is now XBOX 360 Controller is recognized,
bad news is...it only recognizes button press, no axis movement.

Tried other two controller, almost same behavior.
One cheap controller moved a cursor on workbench a bit, but not excpected behavior(not working all way).

I'll wait for next update, thanks for update anyways!

[vossstef]

@t-miura think xbox Gamepad should be possible to use as Joystick in a Game. It will not replace a Mouse that you need for the workbench. Let me check over the next days on my Mega 138K Pro board the NanoMig with latest Companion.

[t-miura]

@vossstef thank you in advance! and some additional memos(sorry about i forgot to be detailed)

in Workbench, all USB controller somehow registered button press as mouse's right/left click,
and cusror barly moves, and it's not a correct behavior as i double-checked by my Amiga Forever/WinUAE...

Also, some games i tried, all of them seems controllers not recognized as joystick in amiga side, too.
(One game seem to be recognized as a mouse, so that's not right, too)

edit: the behavior looks similar to this on actual amiga: https://www.lemonamiga.com/forum/viewtopic.php?t=18521
(only fire button works on port 1)

[t-miura]

for joystick issue, i think i've found the fix.
in src/tang/console138k_bl616/top.sv:

// map first HID/USB joystick into second amiga joystick port
// wire in db9 joystick
wire [7:0] joystick1 = hid_joy0;
wire [7:0] joystick0 = hid_joy1;

swapping this fixed for me.

wire [7:0] joystick1 = hid_joy1;
wire [7:0] joystick0 = hid_joy0;

could be something related to "swap joystick" feature, but i feel this is what originally intended.
confirmed both mouse and joystick work on Workbench and some games.