@The Solutor. This is very strange you don't see DN, the log shows DN started and output the lines. I bet something is wrong with the timer interrupts/wake up or so. Does F12 work and bring up OSD?

I added new diagnostic logs in the new release bundle.

gsos wrote on Yesterday, 15:41:

@The Solutor. This is very strange you don't see DN, the log shows DN started and output the lines.

And not just DN, MC/NC/VC all the same they look started but they aren't visible

I bet something is wrong with the timer interrupts/wake up or so. Does F12 work and bring up OSD?

Yeah I thought that could be the secondary monitor (say output redirected there with an unsupported resolution) so I tried with a single monitor connected, and even swapped them.

No change.

About f12 I think no/not visible, but I'm not sure, I'll recheck and update this message, ASAP

EDIT

I confirm no F12 popoup

superfury wrote on Yesterday, 13:50:

Apparently there's no ISO image anymore on the releases to use? I only see USB stuff there and raw files?

I added a .iso again!

@gsos

New round, using the updated image, new klog (whit hopefully useful info at the end)

Okay yes, something on these systems is different. PIC has a pending irq0 (IRR=1) and IF=1 and irq0 is unmasked (IMR=F8 irqs 0,1,2 are open) yet it's not delivered count=0. So apparently PIC is not hooked up to the CPU in the default config. I'll investigate tonight.

gsos wrote on Yesterday, 17:28:

Okay yes, something on these systems is different. PIC has a pending irq0 (IRR=1) and IF=1 and irq0 is unmasked (IMR=F8 irqs 0,1,2 are open) yet it's not delivered count=0. So apparently PIC is not hooked up to the CPU in the default config. I'll investigate tonight.

In the meantime i noticed that you implemented my suggestion of hybrid UEFI/BIOS boot, so I tried to boot via CSMWrap.

Using the native bios option I reach the filemanager screen !! (Norton Commander is configured ATM),

Using the VBE option, it's the same as UEFI.

I can't say a lot more because, in CSMwrap case I have no working keyboard , just like on the N100 machine.

Whatever I wasn't able to boot the system a single time via CSMWrap with older releases, the ones that worked in UEFI mode.

Okay I released a new version that might fix the issue (tentatively). Likely APIC didn't route the PIC irqs. There will be new log lines if it still doesn't work

gsos wrote on Today, 01:38:

Okay I released a new version that might fix the issue (tentatively). Likely APIC didn't route the PIC irqs. There will be new log lines if it still doesn't work

The new release booted on my B75 machine but still fails at the same place on my H97 and X99.

On my B75 machine it did manage to configure my dISAppointment to the point it picked up my SB16 DSP (CT2290) but I'm not getting any sound output.
- There's no volume (mixer) control for SB16 audio. There's an option to switch to kernel mixing but it doesn't do anything at the moment.
- CT2290 doesn't have IRQ/DMA jumpers and defaults to some arbitrary IRQ and DMA numbers like IRQ10 and DMA3.
- Normally DIAGNOSE can be used to change its IRQ and DMA assignments for such "auto-init" cards to those specified by BLASTER environment variable, and this is also supported by UNISOUND.
- This is an ASUS board. The LDRQ1# is set to GPIO23 by default on this board so it needs to be toggled for DMA to work.
- Not sure if there's any option to switch between HDA and SB16, as on this particular system I've kept HDA enabled for use with a few OSes that don't have working drivers for other sound cards there.

To restore LDRQ1# from GPIO23 (you may refer to my lpcexp project):
1. Read register 48h from the LPC controller (00:1f.0). Strip the last bit. On most boards this reads 0x501 which means GPIO base address is 500h.
2. The GPIO_USE_SEL register resides exactly at the base address retrieved from previous step. Perform a 32-bit I/O read there.
3. Check the value read from previous step. If bit 23 is set, unset only that bit (keep other bits unchanged), and write the modified values back via 32-bit I/O write.

Once done, LDRQ1# will regain its native functionality and DMA will work.
NOTE: It's advised to first read the register 4Ch and check its bit 0 status. If this bit is set it means GPIO control is locked and you cannot make any changes.
However, I'm yet to see a board whose BIOS actually locked the GPIO registers so in most cases this is optional.

FYI: You may consider widening the supported chipsets down to ICH6. This will allow the initialization to work on the original board that featured this LPC-ISA design, RUBY-9719VG2AR, which used ICH7.
Additionally, so far I only had success with ISA PnP with ASRock boards as well as RUBY-9719VG2AR. On some other boards (e.g. ASUS ones), it seems only ports below 3FFh can be routed to LPC-ISA bridge, so no ISA PnP or other functions that require higher I/O ports.

gsos wrote on Today, 01:38:

Okay I released a new version that might fix the issue (tentatively). Likely APIC didn't route the PIC irqs. There will be new log lines if it still doesn't work

Ok, it finally works, tanks. 😁

But this time a lot of additional consideration needs to be added:

-------------------------------------------------------------------------------------

Modes:

It boots both in UEFI, native CSM and CSMWrap

In UEFI I got the 4/3 aspect ratio, with fonts scaled according the 1440x900 native resolution ( practically 1200x900) in not that pleasant way

In CSM --> native video bios mode, the screen is scaled to 1440x900 (so stretched to 16/10) but with very nice looking smoothed fonts

In CSM --> VBE mode , same as above, with smoothed fonts, but not nice looking as above

The strangest part is that if I use the VBE mode AND persistent disk, I have no keyboard (I can't see the relationship, but that's what happens)

-------------------------------------------------------------------------------------

Programs:

VC 4.05 works

VC 4.99 --> too big to fit in memory (not yet tried to set it as default boot program)

MC --> bad filename or executable something like that, (from the log, setting it on boot: "Windows PE exec failed (home/retroos/CONFIG/MC/MC.EXE): errno 2")

Everything else I tried seem working OK

-------------------------------------------------------------------------------------

Other issues:

My second USB keyboard (which is what I'd like to use normally) is still non functional (albeit it's illuminated, so somewhat detected and enabled but not initialized correctly)

I see correctly my partitions, but as I wrote before I cant see the games bundled in the downloaded images.

-------------------------------------------------------------------------------------

Intel N100:

not yet tested.

@LSS: At least some progress. I tried to address your gpio/ldrq issue and widened to ich6. I kept HDA as a kernel mixing backend to choose even when SB is available. For the other platforms if there is a way to give me a log that might be valuable.

@Solutor: Can you give me the boot log of the VBE-mode + persistent disk? It makes no sense. The second keyboard might work now and the games should be visible.

Release is updated.

Overall many thanks for testing this on your hardware! I know RetroOS is still rough around the edges, but at the moment I'm focused on getting the core robust. Supporting different hardware, making sure programs work, etc..

Nice scaling, perfecting sound mixing etc. hopefully come soon when the system runs well. Btw. , the source is deliberately distributed "do what ever you want to do". Feel absolutely free to play with it, add support for some cool retro hardware like dISAppointment, workarounds like EGAFIX for your system. And if you like to merge it in, make a PR on my repo.

Ok going to downloading the new image, please wait.... 😉

In the meantime this is the log *still from the old image*, with not working keyboard (CSM/VBE persistent disk)

(BTW it worked once, when I rebooted, second boot no keyboard, it may be a timing problem, IDK)

P.S. unrelated to RetroOS itself, I think, but when i download the img file, it's always downloaded twice, the first time the download reaches 100%, then it restart from 0%, then the download is finally saved (no manual intervention needed, but still a bit strange)

Ok, few quick runs with the new image

Games folder are now visible, good 😉

In UEFI mode, finally, seems that the "problematic" keyboard works, at least if connected alone, still not working if connected together with the other one.

In CSM mode i didn't test thoroughly the problematic keyboard behavior, but I tried to understand what happens to keyboards in general.

It seem that the first time I reboot from windows the keyboard works, but if I turn off and reboot in CSM/VBE/Persistent disk the keyboard doesn't work anymore, that behavior seems, more or less, consistent and repeatable.

This is a log taken with keyboard KO and the new image .

Speaking of which, you understand that this way (no timestamp, no keyboard, klog overwritten on each run) all the matter becomes really time consuming, given on each test I have to reboot in Windows or Linux just to grab the klog before it gets overwritten in the new test.

So, at least temporarily, would be nice if to have klog-hh-mm-ss.txt format, or at least a progressive numbering, then when the matter will be stabilized you could go back to the actual behavior.

The Solutor wrote on Today, 06:25:

Ok, few quick runs with the new image […]
Show full quote

Ok, few quick runs with the new image

Games folder are now visible, good 😉

In UEFI mode, finally, seems that the "problematic" keyboard works, at least if connected alone, still not working if connected together with the other one.

In CSM mode i didn't test thoroughly the problematic keyboard behavior, but I tried to understand what happens to keyboards in general.

It seem that the first time I reboot from windows the keyboard works, but if I turn off and reboot in CSM/VBE/Persistent disk the keyboard doesn't work anymore, that behavior seems, more or less, consistent and repeatable.

This is a log taken with keyboard KO and the new image .

Speaking of which, you understand that this way (no timestamp, no keyboard, klog overwritten on each run) all the matter becomes really time consuming, given on each test I have to reboot in Windows or Linux just to grab the klog before it gets overwritten in the new test.

So, at least temporarily, would be nice if to have klog-hh-mm-ss.txt format, or at least a progressive numbering, then when the matter will be stabilized you could go back to the actual behavior.

Thanks, I had an issue on my laptop with hda where linux initialized it correctly and after a warm-reboot retroos could detect it but not over a cold reboot. I added unique (non-overriding klogs) with proper modification timestamp. I added extra diagnostic logs around keyboard detection that might shine some light on what is happening. You dont have a mouse attached?

What about your N100 system?

gsos wrote on Today, 04:57:

@LSS: At least some progress. I tried to address your gpio/ldrq issue and widened to ich6. I kept HDA as a kernel mixing backend to choose even when SB is available. For the other platforms if there is a way to give me a log that might be valuable.

Regarding the H97 and X99 systems -- I cannot get any klog as it's hanging at a very early point (showing just the available memory ranges), same as before.

Did some more tests on my B75 system, using dISAppointment options.
- When in SB native mode, hardware FM works, but volume is unusually low compared to native DOS.
- When in HDA (kernel mixing) mode, even on my B75 system it still prefers Headphone over Jack (HDA_OUTPUT has no effect). On the other hand, "Mix through SB" option does nothing other than selecting the first line (Kernel: HDA).
- I'm not getting any SFX output in native mode. Originally I thought it might be due to the system using IRQ5 for SB16, but no effect after manually modifying the CONFIG.SYS in the base image to reflect this (from the default IRQ7).
- Sound works in Quake, but it's actually playing through HDA output even though audio is set to native mode.
- Wacky Wheels seems to always use hardware FM for music, even in HDA. Sound effects only work in HDA anyway.
- Skyroads' music goes to whatever sink that's active (native or HDA). Sound effects only work in HDA, though.
- I'm getting some warnings in the KLOG regarding "sound underrun" but I cannot be sure if the issue with Sound Blaster is related to it.
- I'm getting some warnings regarding DMA5 not available for mixing in the KLOG. Not sure if this is related to the "Mix through SB" option having no effect. CT2290 does support high DMA so I'm not sure about the status of DMA registers inside RetroOS...

So for now it seems HDA remains a better option as native SB support feels hit-or-miss.
Considering FM volume being unusually low, I cannot rule out the possibility that SB mixer (especially WAVE) is not configured properly leading to no sound effect when in native mode, other than the possibility that I/O is not being correctly passed through to the real SB hardware.
Haven't really validated SB mixed mode yet due to current issues, but I'm not sure if this feature can use low DMA (DMA1) as a fallback, since not many Sound Blaster compatible cards support and utilize high DMA.

And issues with the USB IMG format:
- The two partitions bear the same volume serial number (UUID), resulting in my Linux desktop environment listing only the second partition (RETRODATA). The first partition can still be mounted manually (e.g. via command line), however.
- Right now RETRODATA partition doesn't seem to be of any use if USB storage support is yet to be available... I haven't validated this part as I've been testing dISAppointment on the B75 and I haven't enabled persistent disk yet.

gsos wrote on Today, 14:31:

You dont have a mouse attached?

I have a BT mouse currently, I guess a bit hard to get working there, at least at the current development status

What about your N100 system?

I couldn't turn off, because it was busy doing "things", I i'll go to test it soon.

But i't like to resume the current status.

To shorten the discussion

Black keyboard = less problematic

White keyboard = more problematic.

Black:

works in UEFI and CSM mode, not sure it works in any USB port but I would take experiments on specific ports this for later

White:

It's backlit (like the black one) so I have a visual idea of what happens on boot.

UEFI: it's clearly detected (it turns on the lights), but its functional only if the black one is disconnected

CSM: doesn't work at all it's start illuminated and works on the grub menu, but it turn itself off during the system initialization

I'm going to download your new image, then i will test both the AMD and intel machine...

Ok, for now tests HP/AMD with both the keyboards connected

CSM (writable disk, no functional keyboard)

CSM (protected disk, black working, white dead)

UEFI (black working, white alive, but not functional)

Ok, for now tests HP/AMD with with white keyboards only connected

CSM (a run with the keyboard working)

CSM (two run with writable disk and the keyboard not working)

UEFI not tested, I think will be similar to 0005 above.

EDIT:

About the N100 I tested once (still no kb detected), and rebooted my system.

But klog seems it wasn't saved anywhere, I can't bring it down again at least for today, sorry.

@LSS I see, quake is using GUS which is still emulated through HDA. Can you provide me the klog of your system, that might contain some info relevant to why SB is not working. I'm really interested because you provide me the only system where SB really can run natively as RetroOS intended. I think we are very close to get it working with disappointment. I'm released a new version where hopefully the HDA_OUTPUT=Jack problem is fixed.