Mac OS 9 booting on: Mac mini G4 (Detailed Posts)

[MacOS Plus]

  Further to this, when I look at the same device section on the MDD FW400 (which happens to be "snapper" by name and is also on I2S at exact same location), there are a hell of a lot more property entries which define its capabilities and operating parameters (there's even mention of a USB subwoofer).  Presumably all this is necessary information for the device in the Mini too but is all missing from OF.  If we can determine which missing properties are needed on the Mini and what values they should hold, we may be able to write these into the ROM code so that it performs the necessary additions at startup.

 Just as an example to give you a sense of what I'm talking about, here are all the additional property lines from the MDD FW400:

#-detects 00000003
#-inputs 00000002
#-features 00000003
#-outputs 00000004
default-monitor 6e6f6e65 00
equalizer-id 33353353 30333033
harware-types 00000003
icon-id ffffbf4d
info-id ffffbf44
model 35350303
name-id ffffbf4d
object-model-version 00000001
sample-rates 00000002 ac440000 bb800000
i2s-serial-format 00000002
mclk-sample-rate-ratio 00000100
sub-frame 00000000
sound-objects (One line for each instance of the number of detects/inputs/features/outputs defined in the above parameters, detailing all the necessary parameters/addresses for each entry.  Very long strings.)

  I imagine we would need similar entries in the Mini for most if not all of these property lines.  Even "inputs" includes a definition for a "no input" type.  We would need to change the counts in the first four properties to align with the number of these subcomponents present, which are then detailed in the multiple lines of the "sound-objects" property.

[darthnVader]

I believe it is supposed to be an enumerated property, akin to "screamer" that has been added to the compatible property on other machines to gain basic function.  There is a named driver in OS X specifically for the Toonie.  I thought I saw it in the detailed branch of the tree for the i2s device, but I'll have to go take another look at it again.

Update - So I got it confused with some monster listing I saw online somewhere.  Here's the device and its properties from OF on my machine:

dev /pci@F2000000/mac-io/i2s/i2s-a/sound .properties
name: Sound
device-type: Soundchip
compatible: AOAKeylargo
vendor-id: 0000106b
layout-id: 0000003a
object-model-version: 00000002

  As "AOAKeylargo" compatible (Apple Onboard Audio - Keylargo), this would imply Apple didn't re-invent the wheel with this device variant.  The bridge to it is I2S format but the function should be handled the same way as Keylargo.  Why it then isn't fully working using the existing Keylargo code, I don't know.  Can someone with more knowledge of past Keylargo device handling please comment?  The missing link here could be something incredibly simple.

Here are the .properties from the "snapper" sound device in the iBook G3:

Code: [Select]

0 > dev pci1/mac-io/i2s/i2s-a/sound  ok
0 > .properties
name                    sound
device_type             soundchip
compatible              snapper
vendor-id               0000106b
#-detects               00000001
#-inputs                00000002
#-features              00000002
#-outputs               00000002
default-monitor         6e6f6e65 00
device-id               00000018
equalizer-id            33353353 30333033
hardware-types          00000003
has-anded-reset         00000001
icon-id                 ffffbf4d
info-id                 ffffbf44
model                   353S0303
name-id                 ffffbf4d
object-model-version    00000001
sample-rates            00000002 ac440000 bb800000
i2s-serial-format       00000002
mclk-sample-rate-ratio  00000100
sub-frame               00000000
sound-objects           feature index 0 model Proj16PowerControl
                        feature index 1 model Equalizer
                        detect index 0 bit-mask 2 bit-match 0 device 2 registry-name extint-gpio12 model GPIOGenericDetect
                        input index 0 icon-id -16521 name-id -20532 port-connection 4 port-type 0x696D6963 zero-gain 0x00000000 model InternalMic
                        input index 1 model NoInput
                        output index 0 device-mask 2 device-match 2 icon-id -16563 name-id -20524 port-connection 1 port-type 0x6864706E model OutputEQPort
                        output index 1 device-mask 2 device-match 0 icon-id -16563 name-id -20525 port-connection 2 port-type 0x6973706B model OutputEQPort
                       

 ok
0 >

Here is the PowerBook 12' G4:

Code: [Select]

0 > dev pci1/mac-io/i2s/i2s-a/sound .properties
name                    sound
device_type             soundchip
compatible              AOAbase
built-in               
layout-id               00000048
object-model-version    00000002
vendor-id               0000106b
platform-tas-codec-ref  /pci@f2000000/mac-io@17/i2c@18000/i2c-bus@0/codec@6a
 ok
0 >


[darthnVader]

Here is the source for Apple Onboard Audio, tho I'm not sure any of it will do us any good. We don't have the source for the AppleK2Driver, and as far as I can find Apple never released a sound/audio DDK that I can find.

https://opensource.apple.com/source/AppleOnboardAudio/AppleOnboardAudio-258.3.1/

So we would pretty much have to disassemble the Apple Audio extension, or one of the 3rd party sound drivers.

It's perplexing that we don't have a DDK for audio devices, I'm sure it exists, as 3rd parties made sound manager compatible audio drivers, but they must have had to NDA.

[DieHard]

It is an interesting coincidence that the Apple PowerBook G4 1.0 Ghz (17" Aluminum Model M8793LL/A - PowerBook5,1 - A1013) has a similar sound issue where external speakers work, but no "Internal" sound, also that unit as I have documented:

2) Internal Speakers will only work if 1/8" adapter or external speaker jack is plugged into headphone jack and then BOTH speakers and Line out work simultaneously

so try this with the Mini, insert a 1/8" only half way and see what happens

[nanopico]

Here is the source for Apple Onboard Audio, tho I'm not sure any of it will do us any good. We don't have the source for the AppleK2Driver, and as far as I can find Apple never released a sound/audio DDK that I can find.

https://opensource.apple.com/source/AppleOnboardAudio/AppleOnboardAudio-258.3.1/

So we would pretty much have to disassemble the Apple Audio extension, or one of the 3rd party sound drivers.

It's perplexing that we don't have a DDK for audio devices, I'm sure it exists, as 3rd parties made sound manager compatible audio drivers, but they must have had to NDA.

There is (and I don't recall where) documentation on creating audio devices (sound manager) for OS 9.  So that with the PCI DDK and or USB driver stuff one could write a driver given the right information.

[MacOS Plus]

It is an interesting coincidence that the Apple PowerBook G4 1.0 Ghz (17" Aluminum Model M8793LL/A - PowerBook5,1 - A1013) has a similar sound issue where external speakers work, but no "Internal" sound, also that unit as I have documented:

2) Internal Speakers will only work if 1/8" adapter or external speaker jack is plugged into headphone jack and then BOTH speakers and Line out work simultaneously

so try this with the Mini, insert a 1/8" only half way and see what happens

  My guess with the Powerbook is that the reading of the jack insertion detection switch is inverted from what it should be, much like with the inverted behavior of the backlight control slider.  It could be as simple a matter as the logic handling of a single data bit.  I've seen evidence in the device properties that this state may be detected using a GPIO line.  Getting small details like logic high vs logic low = data 1 is often a function of a specific device's driver (or plug-in) such that the common code doesn't have to handle these variances directly.  The property "has-anded-reset" in the example given by darthnVader for an iBook G3 sounds a lot like a logic compensation flag to me.

[darthnVader]

I was going to post the output of my QuickSilver's sound device, but for some reason I have no input in a telnet session.

@MacOS Plus, I'd really like to see the full output of the MDD FW400.

You may already know how to telnet in, but I'll post the answer anyway for everyone:

https://www.fenestrated.net/mirrors/Apple%20Technotes%20(As%20of%202002)/tn/tn2004.html

Basically, you just need two computers connected directly with an ethernet cable, I don't think it needs to be a crossover cable, unless your using a really old computer, as most ethernet is auto sensing, and will switch on it's own.

I'm not sure if both computers need to be auto sensing, or just one, maybe someone can answer that?

Anyway, on the Mac you want to control over telnet, boot into Open Firmware( Command+Opt+O+F ) at boot, and type:

Code: [Select]

" enet:telnet,10.1.2.3" io
Then on the computer you want to control from, set your ethernet address to something like 10.1.2.4, and telnet in:

Code: [Select]

telnet 10.1.2.3
Then you should see the screen go black on the target Mac, and have control of Open Firmware from telnet. Allowing copy and paste.

[MacOS Plus]

  I was doing telnet logging in the past for nanopico when we were working on getting the Xserve to boot OS 9.  I can do it through my network from any machine to any other machine if I select an unused 192.168.0.xxx IP.  Here is the complete OF device tree from the MDD FW400:

Code: [Select]

ff87f638: /cpus
ff87f8e8:   /PowerPC,G4@0
ff87fd10:     /l2-cache
ff87ff48:       /l2-cache
ff880ba8:   /PowerPC,G4@1
ff880fd0:     /l2-cache
ff881208:       /l2-cache
ff881518: /chosen
ff881730: /memory@0
ff8819d0: /openprom
ff881b78:   /client-services
ff882e60: /rom@ff800000
ff883060:   /boot-rom@fff00000
ff883290:   /macos
ff883398: /options
ff8834a0: /aliases
ff8854a8: /packages
ff885588:   /deblocker
ff885f20:   /disk-label
ff8869e8:   /obp-tftp
ff8902c8:   /telnet
ff890bc8:   /mac-parts
ff892478:   /mac-files
ff895430:   /hfs-plus-files
ff89a4b0:   /fat-files
ff89c2a8:   /iso-9660-files
ff89d108:   /bootinfo-loader
ff89ed78:   /xcoff-loader
ff89f810:   /pe-loader
ff8a0260:   /elf-loader
ff8a1908:   /usb-hid-class
ff8a4498:   /usb-ms-class
ff8a70e8:   /usb-audio-class
ff914c60:   /sbp2-disk
ff917820:   /ata-disk
ff919d60:   /atapi-disk
ff91bef0:   /bootpath-search
ff922870:   /terminal-emulator
ff922980: /firewire-disk-mode
ff9383b0: /pseudo-hid
ff9384b0:   /keyboard
ff938ba8:   /mouse
ff939140:   /eject-key
ff939610: /pseudo-sound
ff939940: /multiboot
ff94d758: /diagnostics
ff94d838: /nvram@fff04000
ff94e488: /uni-n@f8000000
ff94e960:   /i2c@f8001000
ff94f478:     /fan@58
ff9502f8:     /i2c-hwclock@ca
ff950ae0:     /temp-monitor@92
ff9511c8:     /cereal
ff9518f8: /pci@f0000000
ff99ce10:   /uni-north-agp@b
ff99d120:   /ATY,ApacheParent@10
ff9c36c8:     /ATY,Apache_A@0
ff9c4570:     /ATY,Apache_B@1
ff952b58: /pci@f2000000
ff9555e8:   /mac-io@17
ff95c0c8:     /interrupt-controller@40000
ff95c318:     /gpio@50
ff95c548:       /extint-gpio1@9
ff95c838:       /programmer-switch@11
ff95cae0:       /gpio5@6f
ff95cd00:       /extint-gpio15@67
ff95cf88:       /gpio6@70
ff95d1a8:       /extint-gpio16@68
ff95d498:       /extint-gpio14@66
ff95d720:       /gpio12@76
ff95d938:       /gpio11@75
ff95db50:     /escc-legacy@12000
ff95ddc0:       /ch-a@12004
ff95dfc0:       /ch-b@12000
ff95e1c0:     /escc@13000
ff95e448:       /ch-a@13020
ff95ee70:       /ch-b@13000
ff95f808:     /i2s@10000
ff95fa38:       /i2s-a@10000
ff95fd68:         /sound
ff960600:     /timer@15000
ff9607f8:     /via-pmu@16000
ff963ec0:       /pmu-i2c
ff964ce0:         /i2c-hwclock@1d4
ff965590:         /i2c-hwclock@1c8
ff965e10:       /rtc
ff966508:       /power-mgt
ff9c76f0:         /usb-power-mgt
ff966818:     /i2c@18000
ff9672d0:       /cereal
ff967a00:       /deq
ff967b40:     /ata-4@1f000
ff96a880:       /disk
ff96b0c0:     /ata-3@20000
ff96de00:       /disk
ff97a5b0:   /usb@18
ff982238:   /usb@19
ff9c5610:     /hub@1
ff9c5818:       /mouse@2
ff9c5b38:       /device@3
ff9c5cb0:         /keyboard@0
ff9c6040:         /interface@1
ff953dd8: /pci@f4000000
ff98a000:   /ata-6@d
ff98d1e8:     /disk
ff98d820:   /firewire@e
ff997a28:   /ethernet@f
ff9c7bb8:     /ethernet-phy
ff955038: /vsp@f9000000
ff955328:   /veo@f9080000
ff955488:   /veo@f9180000
The complete properties listing for the sound device in that machine is as follows:

Code: [Select]

name                    sound
device_type             soundchip
compatible              snapper
vendor-id               0000106b
#-detects               00000003
#-inputs                00000002
#-features              00000003
#-outputs               00000004
default-monitor         6e6f6e65 00
device-id               00000016
equalizer-id            33353353 30333033
hardware-types          00000003
icon-id                 ffffbf4d
info-id                 ffffbf44
model                   353S0303
name-id                 ffffbf4d
object-model-version    00000001
sample-rates            00000002 ac440000 bb800000
i2s-serial-format       00000002
mclk-sample-rate-ratio  00000100
sub-frame               00000000
sound-objects           feature index 0 model Proj14PowerControl
                        feature index 1 model Equalizer
                        feature index 2 model USBSubwoofer
                        detect index 0 bit-mask 4096 bit-match 4096 device 4096 registry-name extint-gpio14 model GPIOPrioritizedDetect
                        detect index 1 bit-mask 2 bit-match 2 device 2 registry-name extint-gpio15 model GPIOPrioritizedDetect
                        detect index 2 bit-mask 4102 bit-match 4 device 4 registry-name extint-gpio16 model GPIOPrioritizedDetect
                        input index 0 icon-id -16521 name-id -20630 port-connection 1 port-type 0x6C696E65 zero-gain 0x00000000 model LineInput
                        input index 1 model NoInput
                        output index 0 device-mask 2 device-match 2 icon-id -16563 name-id -20524 port-connection 1 port-type 0x6864706E model OutputEQPort
                        output index 1 device-mask 4102 device-match 0 icon-id -16563 name-id -20525 port-connection 2 port-type 0x6973706B model OutputMonoEQPort
                        output index 2 device-mask 4102 device-match 4 icon-id -16563 name-id -20523 port-connection 3 port-type 0x6573706B model OutputDallasEQPort
                        output index 3 device-mask 4096 device-match 4096 icon-id -16563 name-id -20524 port-connection 4 port-type 0x6C696E65 model OutputEQPort
  BTW, on the topic of temperature monitoring on the Mini, look at the last section of the OF device tree on the Mini and you'll see there are two clearly defined sensors, one being the CPU on-die.  It appears the necessary device declarations and connections are present to allow reading the sensors, but for some reason this wasn't ever made accessible at the software level.  Some of the other neighboring entries are similarly intriguing.

[darthnVader]

The 7447a seems not to support the THRM registers that earlier G4s did. This is why the mini crashes. Here is a plugin patched to spoof a 30C CPU temperature. Working on a more elegant solution. Let me know about fan behaviour, and beware overheating!

It seems the temp on the Mini is exposed via the via-pmu on the i2c bus.

Not sure how the Apple CPU plugin handles i2c, or the via-pmu, but maybe we can get it to look for the THRM value here?

Also, the AppleKeyLagro , AppleVia, and Applei2c are open source, so if we need any info we should be able to glean it from there.

https://opensource.apple.com/release/mac-os-x-1058.html

[Daniel]

The 7447a seems not to support the THRM registers that earlier G4s did. This is why the mini crashes. Here is a plugin patched to spoof a 30C CPU temperature. Working on a more elegant solution. Let me know about fan behaviour, and beware overheating!

It seems the temp on the Mini is exposed via the via-pmu on the i2c bus.

Not sure how the Apple CPU plugin handles i2c, or the via-pmu, but maybe we can get it to look for the THRM value here?

Also, the AppleKeyLagro , AppleVia, and Applei2c are open source, so if we need any info we should be able to glean it from there.

https://opensource.apple.com/release/mac-os-x-1058.html

Is the exposed temperature the CPU temp or the air temp?

[ELN]

The THRM registers are PowerPC “special purpose registers” that you can access directly from supervisor mode code on the CPU. If the mini CPU temperature probe is on a different bus, then some hacking is called for.

[MacOS Plus]

The THRM registers are PowerPC “special purpose registers” that you can access directly from supervisor mode code on the CPU. If the mini CPU temperature probe is on a different bus, then some hacking is called for.

Here's some OF device lines in regards to temperature and fan in the Mini:

dev /sep/fans/main-enclosure@0
.properties lists device as /pci@f2000000/mac-io@17/via-pmu@16000/pmu-pwm-fans/fan

dev /sep/temperatures/cpu-bottomside@0
.properties lists device as /pci@f2000000/mac-io@17/via-pmu@16000/pmu-i2c/temp-monitor@190/local@0

dev /sep/temperatures/cpu-on-die@1
.properties lists device as /pci@f2000000/mac-io@17/via-pmu@16000/pmu-i2c/temp-monitor@190/remote@1

  Interesting to note that they define the external temp sensor as "local" while the cpu on-die sensor is "remote".  Does the above imply that sensor data would be accessed centrally through the PMU itself?  That would make sense.

[ELN]

Hi guys,

I've been looking into the Power Manager. I want to get this sorted out before I move on to audio.

Mac OS 9 on a G4 does not directly control fans, or routinely monitor the CPU temperature. Fan control is delegated to a chip on the motherboard. The OS is only interested in the CPU temperature when the NanoKernel takes a "thermal event" interrupt on G4s that support this (later ones don't). The few calls to the GetCoreProcessorTemperature function are largely superfluous. In fact, the THRM registers that they depend on are absent on later G4s, causing the mini to crash when a CPU plugin is present.

The attached ROM contains the following NanoKernel hacks, specific to the 7447a:

• Return an error for all calls to the CPU plugin's buggy core temperature getter
• Enable CPU idling (previously the only G4s that could do this were the 7400/7410) -- this might be buggy?

Everyone should use this ROM. Now you can get rid of that hacked-up Apple CPU Plugins file that I posted earlier.

Also, there seems to be interested in easy customisation of the boot script. Get the command line dev tools and Python 3, and you're good to go:

https://github.com/elliotnunn/newworld-rom

[MacOS Plus]

  This ROM did allow the video to sleep but that's as far as it got.  The computer itself and the USB wouldn't sleep, nor could it then recover from this state.  This condition is often what happens on other machines when one PCI/AGP card can't sleep.  I'd be willing to place a small bet on the 9200 being mis-handled, but more inclined to believe the PMU control is still not quite right.  I was actually running with no CPU Plug-ins file at all in the System folder before trying this ROM, and only put the original back now for the test.

  One other observation not related to this ROM - I've been having occasional problems with no mouse pointer movement when the Finder loads.  Everything had been fine from the beginning and then suddenly it started happening recently.  Disabling the ATI Monitor extension fixed this initially, and often is a re-produceable fix, but a couple times it happened even without that.  I'm not sure what the nature of the issue is.  The mouse itself works because the button clicks the Apple menu where the pointer defaults to at startup.  For now it is almost always working with the extension disabled.  I have a QuickKeys sequence programmed to allow me to shut down the system from the keyboard so that at least I can gracefully turn it off whenever there is a mouse issue.

[DieHard]

It's like we're racing to the finish line... there is a possibility that the "PPC Mac Mini Universal Install Image" is coming soon to a theater near you

[ELN]

Sorry MacOS Plus, I think I over-promised. This ROM does not help with system sleep. It only enables the CPU's internal "NAP" mode.

[MacOS Plus]

Sorry MacOS Plus, I think I over-promised. This ROM does not help with system sleep. It only enables the CPU's internal "NAP" mode.

  Maybe I just got a little over-excited.   How would I be able to tell if it was actually 'NAPping'?

[darthnVader]

Good work ELN, the Mini is a nice little OS 9 machine with the way you've been able to code it.

MacOS Plus, I have the same issue with the mouse working for clicks but not moving, sometimes, tho a reboot fixes it.

I've done a little debugging of sleep issues under OS X, I think the app is called SleepyX, tho it's been a number of years and I'm not sure it can help us with enabling sleep in the Mini.

I'll look into it when I have time and see if I can find anything that maybe useful.

[MacOS Plus]

I'm thinking mouse pointer trouble is likely related to video driver interaction, since I've seen this sort of thing occur for similar reasons on very rare occasions even on non-Mac OS systems.  The GPU is supposed to handle the mouse pointer redraws function as a separate dedicated process and is actually partly responsible for allowing the pointer to move in response to user input, if I remember correctly.

  I really don't get why it's a random glitch, nor why it works properly during boot right up to the point where the Finder takes over and then the redraw stops.

  Also, I don't recall if this is normal, but if I boot to OS X and then back to OS 9, the system clock is wrong every time.  PRAM battery is good.

[FdB]

MOS+

Perhaps completely unrelated and possibly too simplistic, yet…
installed new drive and OS 9.2.2 on a Rev.A B&W G3 last night.
Mouse froze just after finder load. Extensions off = okay boot.
After a few attempts and finally a “command-option-esc” upon
freeze, there was some ATI extension that force quit and then
it went on to full boot. Don’t remember exact extension title,
but I did switch it off. (Too lazy to go downstairs right now.)

Also, after OS X install and shuffling back between the two, I’m
also having the old faulty time ritual… that years ago I remember
somehow solving. Alas, that complete memory also fails me now.

And, just keep it up you all. When time actually begins folding into
and on top of itself and that gaping rift opens up into the space-time
continuum void… the last thing you’ll hear is the Ghost of Jobs
chanting “Redrum, redrum, REDRUM!…”

Absolute heresy! Get thee behind me Satan!

Alright, I may soon begin the Mini Quest as well.