Thanks for the tip, but the sound is very bad and distorted. That doesn't really sound good.
Could be the same as for "SDL Mega Mario": An endian swapping bug somewhere. For Mega Mario the difference between AmigaOne and Pegasos2 emulation fixes it, for sound it destroys it.
Thanks for the tip, but the sound is very bad and distorted. That doesn't really sound good.
Could be the same as for "SDL Mega Mario": An endian swapping bug somewhere. For Mega Mario the difference between AmigaOne and Pegasos2 emulation fixes it, for sound it destroys it.
That's strange now that I configured the sound with your help the system no longer crashes when I quit Mega Mario. It exits normally....seems the game doesn't like it when no sound output is configured via ahi.
If I could choose between Pegasos2 emulation and AmigaOne emulation I would prefer the AmigaOne emulation. The emulation gives me a good feeling of what I used to use as Amiga NG hardware.
-There are currently unfortunately problems with the sound which may also be solved in the near future. ac97 does not work sb128 distorted.
- compatibility to software seems to be better as under the Pegasos 2 emulation SDL.
-speed after configuring AmigaOs4.1 is at least as fast as under the Pegasos 2 emulation.
-nvram would be really perfect to use the firmware U-Boot with all its features.
-sm502 as well as under the Pegasos2 emulation remains the bottleneck no 32 bit output.
This is just my personal opinion and as a tester of the whole thing, I will of course keep helping to improve things. Otherwise there is only to say that I am impressed how fast you have made this AmigaOne emulation possible. It's not perfect yet, but still good
@Maijestro Well, now you can choose between amigaone, pegasos2 and sam460ex. I don't want to choose one and if I had to it wouldn't be amigaone because the other two machines also run MorphOS and AROS so would give the most choices while amigaone is limited in this regaard. It was only possible to do the emulation so quickly because I've spent the time before with pegasos2 (about two years, one year to implement it and one year clean it up to upstream the patches althogh only worked on it occasionally so that's why it took so long) and the amigaone shares a lot with the pegasos2 only the ArticiaS is different but the Marvell Discovery II is similar (they used it to replace Articia in PegasosII) and the VT82C686B was already in QEMU partially emulated and I kept that working when added VT8231 so now I only had to use it. So it was possible because of reusing code and knowledge from pegasos2 emulation. The pegasos2 is easier to emulate as it's well documented while for AmigaOne the only source I have is the U-Boot and Linux sources that leaves a lot of guessing. Some parts of AmigaOne is documented by @sailor but that's still missing some details. So I'll plan to clean it and submit it for QEMU 8.2 and continue to maintain all 3 machines but none of them is more important to me than the others. The most to improve is in sam460ex where there are problems that slow it down, pegasos2 is mostly working now I think and amigaone is new and maybe need some help with it.
As for the missing shutdown option I think it's just not supported by AmigaOS on AmigaOne. On pegasos2 it does not poke the VIA chip directly but calls the firmware provided RTAS function so unless there's something similar in U-Boot then it probably does not have that function but this could be confirmed by someone with a real machine.
For the sound issue and SDL I'm not sure it's endian related. There could be a hardware feature not emulated for endian swapping but then the PCI card's device IDs were also swapped and would not be detected or the network card would not work either. To debug the sound issue maybe you could try reproducing it with Linux running on AmigaOne and see if it has the same problem. Nobody would be able to look at it without an AmigaOS license so to report the problem you need to reproduce it with something more accessible. The ac97 was broken an real machine and I've read it may have been related to some firmware bug but not sure when that was fixed. The PCI sound card should work and the ES1370 with SB128 AHI driver works on pegasos2 and sam460ex so it maybe something not emulated correctly in amigaone or something different in the AmigaOne AmigaOS version. Cross checking with Linux should also show which of these two is more likely.
As for the missing shutdown option I think it's just not supported by AmigaOS on AmigaOne. On pegasos2 it does not poke the VIA chip directly but calls the firmware provided RTAS function so unless there's something similar in U-Boot then it probably does not have that function but this could be confirmed by someone with a real machine.
AmigaOS completely deletes the firmware data and memory after it's started, at least on U-Boot systems, therefore it can't call any U-Boot functions (the last part which can is the SLB_v2, but AFAIK not the kernel). If the Pegasos2 version of AmigaOS can use firmware functions it's very different to the other versions.
Quote:
For the sound issue and SDL I'm not sure it's endian related. There could be a hardware feature not emulated for endian swapping but then the PCI card's device IDs were also swapped and would not be detected or the network card would not work either.
config is never endian swapped, and the PCI driver can disable it for io, which the network driver might do.
Additionally to the AmigaOS PCI function expansion.library/pci_device/SetEndian some PCI cards support endian swapping for memory, and maybe mmio, BARs (but not for the config and io ones), by changing some config register, which is more likely a possible reason for an endian problem in emulation than the AmigaOS PCI API.
Quote:
The ac97 was broken an real machine and I've read it may have been related to some firmware bug but not sure when that was fixed.
If it was firmware related and doesn't work with U-Boot 1.1.1 it was fixed in U-Boot 1.1.4. But on the real AmigaOne XE hardware the main problem is that they removed the audio chip in later batches, because it wan't working...
As for NVRAM: It's not only required for the AmigaOne emulation but for the Sam460 one as well. Only the Pegasos2 version of AmigaOS doesn't support it and uses Kickstart/nvram.config instead. But that's read-only, you can't change the contents with AmigaOS functions or tools (C:NVSetVar, Prefs/U-Boot, etc.), only by editing the file and reboot.
The PCI sound card should work and the ES1370 with SB128 AHI driver works on pegasos2 and sam460ex so it maybe something not emulated correctly in amigaone or something different in the AmigaOne AmigaOS version. Cross checking with Linux should also show which of these two is more likely.
I am not a Linux expert maybe you could reproduce it under Linux with the AmigaOne version if your time allows. I would like to limit myself only to AmigaOs4.1 and debug there with the means that are available.
For me it was important to test this new AmigaOne emulation and I confirmed that it works. The installation is the same as with Pegasos2 emulation except that the installer of the AmigaOne version can not finish the installation (AmigaOs4.1 freezes) and you have to do a lot of manual work (kickstart/kicklayout) is not written correctly or the paths in the kicklayout must be completely rewritten after the installation.
Maybe the AmigaOne emulation is not the right emulation due to the many hardware bugs (ac97/ArticiaS) that this hardware has, it probably won't matter under emulation as you can probably build around it with Qemu, but this would be extra work. Since I am not a programmer and never will be, I can not judge.
One more thing, every emulation has its advantages and disadvantages, but I decided to use AmigaOs4.1 because I wanted to use it in some form. Aros, Morphos, Linux are not important to me.
In the meantime I'll set up a new Sam460 system and configure it as I learned it from the Pegasos 2 and AmigaOne emulation. Since I have all 3 licenses I could do the tests on all systems. Maybe it would also improve the Sam460 emulation and maybe it would be the best choice for AmigaOs4.1, since the Sam460 is a bit newer than the Pegasos2 and AmigaOne in terms of hardware.
@joerg
Can you briefly confirm if you could completely shutdown an AmigaOne via the AmigaOs4.1 menu with AmigaOs4.1 FE Update 2 where this feature was provided?
Edited by Maijestro on 2023/8/6 9:13:54 Edited by Maijestro on 2023/8/6 9:17:22
From emulation Pov, the X5000 is the best target. All SoC documentation is freely evailable and it's the best supported and most stable Amigaos4.1 target. The only showstopper is that you cannot buy a separate Amigaos4.1 copy for it.
Can you briefly confirm if you could completely shutdown an AmigaOne via the AmigaOs4.1 menu with AmigaOs4.1 FE Update 2 where this feature was provided?
The last AmigaOS version I used on an AmigaOne XE was much older and didn't have that feature yet.
I looked again at the Sam460 emulation, unfortunately I couldn't install AmigaOs4.1 on SFS partitions.
This seems familiar to me from the Pegasos2 emulation, when the incorrectly set CPU resulted in SFS partitions being recognized, but logging in under AmigaOs4.1 was not possible. Of course, this makes the sam460 emulation significantly slower.
Maybe it's a CPU Sam460 emulation error that also makes SFS unusable here?
This is used as CPU:
AMCC PowerPC 460EX Rev B at 1150MHz (PLB=230 OPB=115 EBC=115)
With FFS, the AmigaOs4.1 installation works perfectly, and the installer can also finish the installation cleanly. After I installed FE Update 1 and 2 I wanted to try the switch-off function via the menu, it also fails here. As far as I know, there was an acpi fix for the Pegasos 2 emulation for Qemu at the time, which later made it possible. This acpi fix is missing here apparently.
As soon as the network is accessed, this emulation becomes very slow, e.g. when I download something from the Internet with iBrowse.
Of all NG emulations, the Sam460 is the most user-friendly, there is no need to modify the cd, the network driver rtl8139 also works after FE Update 2. The sound output works very well and has a clear sound. Unfortunately, the Sam460 emulation is also the slowest in all areas.
The most useful emulation is Pegasos2, AmigaOs4.1 runs very fast with this emulation, the onboard Ac97 can be used, SFS02 works very well, computer shutdown as well. The clear winner for the time being is the Pegasos 2 emulation with AmigaOs4.1.
The scitech x86emu is missing and it's U-Boot 1.3.2-rc2.
That's a rather later release and a deal breaker.
Quote:
x86emu is there, and of course cmd_boota.c as well, the source for it was always available
That's better than the source I had from somewhere that had binary blobs for private UBoot functions. In Linux speak it was compiled with non-free firmware. I wonder how long it's been out there as the project to reboot UBoot and update it relied on having source.
Quote:
Only the Pegasos2 version of AmigaOS doesn't support it and uses Kickstart/nvram.config instead. But that's read-only, you can't change the contents with AmigaOS functions or tools (C:NVSetVar, Prefs/U-Boot, etc.), only by editing the file and reboot.
Surprised at that. Thought it would just read it directly off the firmware nvram. The X1000 does. Both are inferior to the XE which can wrote to nvram from OS. The X1000 has an RTAS resource so don't know why it doesn't do that. They had CFE source so knew how it did it. Also I don't see why an an nvram.config couldn't be written to from the OS since it's a normal file so that's an unusual restriction.
You are right, a kicklayout was created from the install CD and copied to SYS:Kickstart/ , because Qemu crashes after the AmigaOs4.1 installation when closing the installer, I suspect that the changes were not written to the installation HD as it should be.
That would explain it. Looks like the easiest way to fix it. The default HD startup file would be somewhere but not sure where it is kept.
I can eventually look at emulating nvram but it's low priority for me at the moment. The default in u-boot is to boot from HD so once you've installed just escape and use the settings should work for now. It's less work than typing the command in pegasos2 that was needed before.
Isn't this something already emulated by QEMU? I mean, how does it emulate a Mac machine? A Mac relies on nvram in the same kind of way and that has to be stored somewhere.
As to shutting down, no the XE cannot. It was a hardware flaw. The closest an A1 gets to shutting down is a blinking cursor on a blank screen.
The Quit item in Workbench seems rather useless and has been since it was first introduced in OS2 or what ever release had the item there. No Workbench I've seen can actually quit and always complains about some process running. Even if it could quit it would just quit to DOS into a shell window. Or quit to a blank screen without window.
I wonder how long it's been out there as the project to reboot UBoot and update it relied on having source.
The complete U-Boot source was always available, the HTTP interface of the U-Boot repository displays the files as being 19-20 years old. No idea where you got incomplete U-Boot sources with binary parts from.
AmigaOS completely deletes the firmware data and memory after it's started, at least on U-Boot systems, therefore it can't call any U-Boot functions (the last part which can is the SLB_v2, but AFAIK not the kernel). If the Pegasos2 version of AmigaOS can use firmware functions it's very different to the other versions.
SmartFirmware on PegasosII provides run-time services RTAS which is a CHRP thing and I think AmigaOS uses that instead of poking the hardware. MorphOS may access the VIA PM function instead. In the Pegasos2 version rtas seems to be checked in all OF related parts but it's unclear what it is used for at the end but I think shutdown is one of those functions. This means AmigaOS probably does not know about the VIA PM function. Moreover this function may not be implemented on AmigaOne board so there's just no way to shut it down from software then. (I mean the PM function is there in VT82C686B but maybe the board just does not have soft off function then it does nothing. In QEMU it would shut the machine down if something wrote the right regs but aparently AmigaOS AmigaOne version does not even try.)
Quote:
config is never endian swapped, and the PCI driver can disable it for io, which the network driver might do.
Additionally to the AmigaOS PCI function expansion.library/pci_device/SetEndian some PCI cards support endian swapping for memory, and maybe mmio, BARs (but not for the config and io ones), by changing some config register, which is more likely a possible reason for an endian problem in emulation than the AmigaOS PCI API.
If this was a device function then how does that work on pegasos2 and sam460ex supposedly using same guest driver and surely same device emulation? So it's either something specific to AmigaOne or the drivers on AmigaOne aren't the same as on other AmigaOS versions. I don't know what's on the AmigaOne CD so I can't check.
Quote:
If it was firmware related and doesn't work with U-Boot 1.1.1 it was fixed in U-Boot 1.1.4.
Where's the firmware binary built from that? I think some people have it but nobody posted it yet. I don't want to rebuild it just to find out there are several problems with it, I'd rather test with something known to run on real machine to avoid unecessary problems. Quote:
But on the real AmigaOne XE hardware the main problem is that they removed the audio chip in later batches, because it wan't working...
Was that the case for all AmigaOne boards or were there some with working audio? If it never worked they why the driver is loaded and selected? Do we need to emulate missing codec for it to not cause problem? That would clutter VIA chip model which is already too complex because of emulating two versions of the chip so I'd rather not do that. Quote:
As for NVRAM: It's not only required for the AmigaOne emulation but for the Sam460 one as well. Only the Pegasos2 version of AmigaOS doesn't support it and uses Kickstart/nvram.config instead. But that's read-only, you can't change the contents with AmigaOS functions or tools (C:NVSetVar, Prefs/U-Boot, etc.), only by editing the file and reboot.
What settings are in nvram for which the defaults are not enough and need to be changed? For IDE parameters you said theres some command in C: that can change this so I guess you could add that to the startup-sequence instead. Others may not matter at the moment. There's one ArticiaS register that seems the enable/disable the NVRAM access but I don't know how that works so I would be guessing here. As said this is not crucial to get it working so it's pushed back until there are more serious issues to fix.
I am not a Linux expert maybe you could reproduce it under Linux with the AmigaOne version if your time allows. I would like to limit myself only to AmigaOs4.1 and debug there with the means that are available.
That's fine, there are others here who could also help, maybe somebody had experience with Linux on AmigaOne and could test it. It would also take me time to set up a Linux installation for testing and I could spend that time with something else so I'll do that if nobody does it by then. It just means sound will be broken longer until we get there. Maybe this would be best debugged by Volker but for that I'll need to finish the patches and submit to QEMU after relase 8.1 comes out then wait until it gets merged then provide info on how to reproduce it with Linux to Volker so we still have plenty of time for that as this would not happen before September anyway. Quote:
For me it was important to test this new AmigaOne emulation and I confirmed that it works. The installation is the same as with Pegasos2 emulation except that the installer of the AmigaOne version can not finish the installation (AmigaOs4.1 freezes) and you have to do a lot of manual work (kickstart/kicklayout) is not written correctly or the paths in the kicklayout must be completely rewritten after the installation.
Fixing this freeze would be needed to make this more user friendly but I need more info on what's happening. Can you get logs shown when that happens (run at least with -d unimp,guest_errors -trace enable="via*" hopefully that gives enough detail; this will probably print a lot of logs, only the end is needed when it finished installation and the freeze happens, I mean not just the last line but the lines printed at that point but not all the logs from the start and during install where it worked) then when it's frozen try info registers from QEMU monitor to see what is running at that point. To redirect both QEMU monitor and serial output to console use -serial mon:stdio then try Ctrl-A h in serial console for help on how to switch between serial and monitor (it's Ctrl-A c but there are more commands described in help). Quote:
Maybe the AmigaOne emulation is not the right emulation due to the many hardware bugs (ac97/ArticiaS) that this hardware has, it probably won't matter under emulation as you can probably build around it with Qemu, but this would be extra work. Since I am not a programmer and never will be, I can not judge.
As said before AmigaOne emulation is kind of a byproduct of pegasos2 emulation and since its not much code to add and maintain but allows those people to also use it who don't have Pegasos2 or Sam560ex licenses but have AmigaOne version then why not. Quote:
One more thing, every emulation has its advantages and disadvantages, but I decided to use AmigaOs4.1 because I wanted to use it in some form. Aros, Morphos, Linux are not important to me.
Most people here probably care about AmigaOS which is OK but I don't want to limit the number of people who could use it but instead make it as widely available as possible so that's why AmigaOne is not a good choice in that regard and sam460ex and pegasos2 are better for this. But because supporting pegasos2 and amigaone together is easy I decided to add it as well.
Was that the case for all AmigaOne boards or were there some with working audio?
Only the first batch(es) included the audio chip. It took years to get it working, and it was "just" a software problem. But since EyeTech thought it would be an unfixable hardware problem they removed the audio chip to save a few cents producing the later boards... It's still not 100% clear if the DMA problems in the AmigeOne SE/XE/µA1 were caused by the Artica-S. The Artica-S definitely has a cache coherency bug, but AmigaOS can work without hardware cache coherency, and all other DMA related bugs in the AmigaOnes are more likely related to the VIA VT82C686B. x86 PCs using this south bridge, but a different north bridge, had exactly the same DMA bugs running Linux.
From emulation Pov, the X5000 is the best target. All SoC documentation is freely evailable and it's the best supported and most stable Amigaos4.1 target. The only showstopper is that you cannot buy a separate Amigaos4.1 copy for it.
Except that likely none of the components of the SoC besides the CPU core is emulated by QEMU so one would need to implement those to get it working (such as memory controller, PCIe controller, interrupt controller if it does not use some standard PPC part for that, i2c controller and whatever else are there). This is a lot of work and then you can't even get guest binaries to run on it as they aren't even sold without hardware. And it would not be any faster than pegasos2/amigaone as the main limiting factor is the same emulation so does not seem to be a good choice.
I've started with sam360ex because there was an unfinished emulation of that already that I could start from, then implemented pegasos2 because parts of it were already there and only needed the Marvell chip and the board code for it then adding amigaone is trivial based on that but probably sam460ex wasn't a good choice back then as it's also a quite complex SoC with limitations. Therefore I'm probably not interested in emulating any more of these hardwares but if somebody is I've shown the way how it's possible so they can go ahead and do it and support it. For me these three are enough for now.
I looked again at the Sam460 emulation, unfortunately I couldn't install AmigaOs4.1 on SFS partitions.
This seems familiar to me from the Pegasos2 emulation, when the incorrectly set CPU resulted in SFS partitions being recognized, but logging in under AmigaOs4.1 was not possible. Of course, this makes the sam460 emulation significantly slower.
Maybe it's a CPU Sam460 emulation error that also makes SFS unusable here?
This is used as CPU:
AMCC PowerPC 460EX Rev B at 1150MHz (PLB=230 OPB=115 EBC=115)
@joerg said before that PPC SFS version only supports some G3 and G4 CPUs so I think this probably does not work on real machine either because of that. Maybe there's a version of SFS that works or use a 68k version instead that was the frist workaround for pegasos2 as well. But it's not slow because of that but because the CPU is embedded PPC which has a different MMU and emulation of that in QEMU is done differently which results in frequently dropping caches that makes it run slow. This should be fixed in QEMU but I've only found the problem but had not time to try to solve it yet. Quote:
With FFS, the AmigaOs4.1 installation works perfectly, and the installer can also finish the installation cleanly. After I installed FE Update 1 and 2 I wanted to try the switch-off function via the menu, it also fails here. As far as I know, there was an acpi fix for the Pegasos 2 emulation for Qemu at the time, which later made it possible. This acpi fix is missing here apparently.
There's no ACPI in sam460ex (that's a PC thing coming from the VIA chip on pegasos2) so no fix for that is needed either. The Sam460ex may have some other way to handle this but I have no detailed docs on 460ex and don't know how AmigaOS handles shutdown on that machine. If you don't even have a button then it's not an emulation bug but function is likely missing on that version just like on AmigaOne.
Quote:
As soon as the network is accessed, this emulation becomes very slow, e.g. when I download something from the Internet with iBrowse.
Does it become faster again when download is finished or stays slow? If it's only slow during download it probably does not handle interrupts well which could be related to the MMU issue I've mentioned above which I think happens on context switching so as long as you run one thing it probably less of a problem but would be slower with multiple things running. Quote:
Of all NG emulations, the Sam460 is the most user-friendly, there is no need to modify the cd, the network driver rtl8139 also works after FE Update 2. The sound output works very well and has a clear sound. Unfortunately, the Sam460 emulation is also the slowest in all areas.
It's the oldest so had most time to improve it but the slowness comes from how QEMU emulates the embedded PPC used on this machine. This only became apparent after we made pegasos2 working this year and only found the problem a few months ago so there's hope this can be improved but will need some time for that which I've spent on BBoot, amigaone and posting here instead so since I can only do one thing at a time this does not go that fast.
You are right, a kicklayout was created from the install CD and copied to SYS:Kickstart/ , because Qemu crashes after the AmigaOs4.1 installation when closing the installer, I suspect that the changes were not written to the installation HD as it should be.
That would explain it. Looks like the easiest way to fix it. The default HD startup file would be somewhere but not sure where it is kept.
Isn't it Kickstart/Kicklayout? That's what BBoot also looks for when you zip your Kickstart directory. That one has only Kickstart in path but found in System/Kickstart on CD and the CD therefore has an additional Kickstart/Kicklayout file with paths pointing to System/Kickstart but on installation the System/Kickstart dir should be copied to SYS:Kickstart then the Kicklayout file in it would be used and the CD:Kickstart/Kicklayout is only on the CD and should not be copied. Did the frozen install not copy the Kickstart dir or how it missed the Kicklayout file in it?
News feed
OS4 Feedback forum
OS4Depot Feedback forum
Amigabounty forum
OpenAmiga forum
AmiCygnix forum
ABC shell forum
AmiKit forum
Cinnamon Writer forum
CodeBench forum
Digital Universe forum
Dopus 5 forum
E-UAE forum
Gnash forum
Ibrowse forum
JAmiga forum
Odyssey forum
OWB forum
Qt forum
SmartFileSystem forum
Timberwolf forum
TouchDevice forum
TuneNet forum
Unsatisfactory Software forum
