Neptune v6.3

Release notes - Neptune 6.3

Everything in this archive - neptune.library, all 30 class drivers, Neptune USB, USBEject
and the command-line tools - reports version 6.3, followed by the CPU it was built
for.

Upgrading from 6.2: run the installer; it replaces the old installation and your
settings are imported (see below).

Poseidon for AmigaOS is now Neptune

At the request of the AROS Development Team, this project no longer presents itself
as Poseidon. From 6.3 it is Neptune - an incompatible fork of the Poseidon USB stack.

Neptune is not compatible with Poseidon:

• Poseidon's classes and tools (the 4.x and AROS 5.x lines) do not run on Neptune, and
  Neptune's do not run on Poseidon.
• Host-controller drivers written for Poseidon's V3 host-controller interface are not
  supported. Neptune speaks the classic V1/V2 interface and its own context interface; the
  context interface is not part of Poseidon's API.

What changed names:

	6.0-6.2	6.3
Library	LIBS:poseidon.library	LIBS:neptune.library
Class drivers	SYS:Classes/USB/<name>.class	DEVS:USBClasses/<name>.usbclass
Control panel	SYS:Prefs/Trident	SYS:Prefs/NeptuneUSB ("Neptune USB")
Commands	PsdStackLoader, PsdDevLister, PsdErrorlog	USBStart, USBList, USBErrorlog
	AddUSBHardware, AddUSBClasses	USBAddHardware, USBAddClasses
Optional tools	DRadioTool, PencamTool, PowManTool, RocketTool, SonixcamTool, UPSTool	USBRadio, USBPencam, USBPowMan, USBRocket, USBSonixcam, USBUPS
Settings	ENV(ARC):Sys/poseidon.prefs	ENV(ARC):Sys/neptune.prefs
Sounds	SYS:Prefs/Presets/Poseidon/	SYS:Prefs/Presets/Neptune/
Device popups	"Poseidon Popups" (PoPo)	"USB device notifications"
Settings-file datatype	PSD	Neptune
Archive	Poseidon-<ver>-<cpu>.lha	Neptune-<ver>-<cpu>.lha

Every Neptune command now starts with USB, and every class driver ends in .usbclass, so
installing Neptune no longer replaces a Poseidon 4.x command, tool or class, and neither stack
can pick up the other's classes.

One USB stack at a time: Neptune does not start its USB controllers while Poseidon is
loaded. USBStart and USBAddHardware say why and stop, and so does Neptune USB in its log;
reboot without Poseidon. A Neptune class whose device
(usbscsi.device, usbparallel.device, ...) another USB stack already provides does not
start, and the log says so.

What the installer does:

• It disables the Poseidon USB Stack block in S:User-Startup, so the old stack is not
  started next to Neptune, and adds a Neptune USB Stack block if you ask for autostart.
  With a Poseidon 4.x installed the block is left alone: its installer uses it too.
• It offers to remove the 6.0-6.2 files listed above. Only files that report version 6 or
  newer are removed, so a classic Poseidon is never touched, and poseidon.prefs is kept.
• It removes the 6.0-6.2 copies of AddUSBHardware, AddUSBClasses and the optional tools,
  again only those that report version 6 or newer. If an earlier installation overwrote a
  Poseidon 4.x's AddUSBHardware or AddUSBClasses, copy them back from the 4.x archive.
• It copies sounds you added to SYS:Prefs/Presets/Poseidon/ into the new drawer.

Your settings: on first start, if there is no neptune.prefs, Neptune imports
poseidon.prefs, moves the class and sound paths to the new drawers and class names, and saves
the result as neptune.prefs. Loading an old settings file by hand converts it the same way.
A classic Poseidon's (4.x) poseidon.prefs is not imported: Neptune starts with its defaults.
Neptune saves its settings in a format of its own, which Poseidon does not read.
Neptune USB's saved window positions start afresh, because the program's name changed.

Programmers: the library is opened as NEPTUNENAME ("neptune.library"), version
NEPTUNE_LIB_MIN_VERSION, with <libraries/neptune.h> and <proto/neptune.h>; the base is
NeptuneBase. The psd*() function names, structures and tags are unchanged; the settings
file's outer form is IFFFORM_NEPCFG (NEPC, was IFFFORM_PSDCFG).
The class protocol header is now <libraries/neptune_class.h>, with
<proto/neptune_class.h> (was usbclass.h, the name PoseidonDev uses); the base
UsbClsBase, the usb*() functions, tags and methods are unchanged. A class is built as
<name>.usbclass; file, romtag and library name must agree.
Classes built for 6.0-6.2 open poseidon.library and must be rebuilt.

The notes for 6.0-6.2 below keep the names those releases used.

USB cameras

The new usbvideo.usbclass drives webcams that follow the USB Video Class standard (UVC 1.0,
1.1 and 1.5). It needs xhci.device 6.4 from the Emu68 driver stack on a Pi 4 or CM4.
Tested with a Logitech C930e, up to 1920x1080 at 30 frames a second.

• Formats: Motion-JPEG and uncompressed (YUY2, NV12 and others) in every size and rate
  the camera offers, and H.264 from cameras that carry it inside Motion-JPEG, such as the
  C930e. Other H.264 and VP8 cameras and bulk-transfer cameras should work but are untested.
• cAMIra and other VHI programs: the Installer offers to put neptune.vhi into
  LIBS:VHI/. Each camera is one input.
• Camera settings in Neptune USB: brightness, exposure, focus, white balance and the other
  standard controls, applied at once and saved per camera.
• USBVideo (optional tools) lists a camera's modes and records video into .mkv files,
  which play and seek on a PC and survive a crash up to the last frame written: H.264 and MJPEG,
  and uncompressed YUY2 and NV12 for a PC to work on. A file goes up to 4 GiB, then the recording
  goes on in the next. Each recording gets the next free number, so none is written over: USBVideo FORMAT=H264 FRAMES=300 TO=RAM:cam makes RAM:cam-0001.mkv, and without FRAMES it records
  until Ctrl-C. From its icon, or with USBVideo GUI, it opens a window (needs MUI) to pick a
  camera and a mode, record, and open the camera's settings.
• A live picture from the camera: with videocodec.library installed (PiStorm on a
  Raspberry Pi 4 or CM4) and a Picasso96 Workbench, starting an MJPEG, H.264, YUY2 or NV12 mode
  opens a second window with the picture, made by the VideoCore (YUY2 and NV12 look a little flat:
  the VideoCore reads their colours as full range). Drag its corner to any size.
  USBVideo PREVIEW does the same from the Shell, without MUI. If the picture never comes, the
  VideoCore is short of memory: raise gpu_mem in config.txt.
• The camera's microphone shows up as an AHI recording mode, and an idle camera is
  suspended when power saving is on.

Not yet supported: choosing the frame rate from a VHI program, still images, and
maker-specific controls.

Programmers: see <devices/usbvideodevice.h>.

USB audio

• Record-only devices such as webcam microphones now get their own AHI modes, named
  "Rec".
• Devices with several outputs (loudspeaker and headphones, say): AHI's output choice
  now works - "All outputs", or a single output by name if the device can mute each one.
• Volume and gain sliders now follow the device's range evenly instead of losing most of
  the level in the first steps, and the volume no longer lands on the wrong control.
• Fixed: 24 bit mono recording, 8 bit recordings that were far too quiet, timing on high
  speed devices, and the sample rate set on the device.

Only the USB Audio 1.0 standard is supported.

USB MIDI

• Plugging in a MIDI interface no longer crashes the machine.
• Sending and receiving MIDI now work, including System Exclusive of any length.
• One CAMD device per interface, with as many ports as the interface really has, so the
  MIDI Ports editor shows the right number. Two identical interfaces each get their own name,
  and every port can be used at once.
• USB MIDI 2.0 devices are used in their MIDI 1.0 mode.
• The settings window is gone - it had nothing to set. Name ports in the MIDI Ports editor.

Upgrading: the new CAMD driver is written to DEVS:Midi/ the first time you plug the
interface in. Quit your CAMD programs and reboot, then, with camd.library 37, add the
interface once in the MIDI Ports editor.

USB in ROM

• The custom Kickstart holds 1.5 MB of modules instead of 1 MB.
• The build script is now python3 ROM/build-kickstart.py, with the same arguments.

Bug fixes

• FAT sticks and drives read and write several times faster. Copying a 64 MB file from a
  USB 3 stick to RAM: went from 2.8 to 18.7 MB/s, and a drawer of 594 small files copies to
  the stick in 12 s instead of 47 s. Mass storage no longer mounts volumes with settings that
  kept fat95 off its fast path (GH #24).
• Rare crashes and freezes under heavy load.
• Behringer UMC204HD now works, and its MIDI port appears once instead of twice.
• NTFS and exFAT: The filesystem stack size is now a setting in the mass storage settings.
• A busy device is no longer logged as a failed suspend.
• A device used at the very moment power saving put it to sleep could stop responding.
• No more "Adding RT Iso Handler failed" and silent audio after a device refused a setting
  change.
• A hub could stop seeing new devices after one was unplugged, until a reboot.
• Possible crash when a class driver was unloaded while still in use.
• Better recovery from errors on UAS disks.
• USB floppy drives and other CBI devices no longer stop responding after a failed
  command.
• Possible crash on disk change with several partitions.
• USB drives come back after Restart in Neptune USB, and when every drive has gone away and
  one returns within seconds (a quick replug, a hub pow...

Poseidon v6.2

Release notes - Poseidon for AmigaOS 6.2

Everything in this archive - poseidon.library, all 29 class drivers, Trident, USBEject
and the command-line tools - reports version 6.2, followed by the CPU it was built
for.

Upgrading from 6.1: install the whole archive. Your settings are untouched. Every
binary in this release was rebuilt, so replace all of them rather than picking out
poseidon.library.

One Suspend/Resume button

Trident's Devices page used to have separate Suspend and Resume buttons, one of them
always greyed out. It now has a single button that says Suspend or Resume depending on
the selected device's state, and follows it live, including when the device is put to sleep by
the power-saving timeout. The button is greyed out when suspending cannot work: the host
controller driver does not support it, or the hub the device is plugged into has no driver.

Fewer stuck USB devices

When a USB device stalls, the library now recovers it, even for class drivers that never
did. You may see a line in the error log saying so.

USB storage settings that stay as you set them

The compatibility switches of massstorage.class (Simple SCSI, Fix Capacity, Trim Inquiry and
so on) could change behind your back, and once they had, changing the class defaults no longer
reached that device. This release closes the gaps:

• No Fallback is honoured everywhere. With it ticked, the class no longer switches a
  workaround on by itself or saves one. When a device looks as if it needs one, the error log
  says which, once per plug-in - for example "Simple SCSI" is probably needed for this device.
  Please check this. The built-in workarounds for the few devices the class knows by name
  still apply.
• Forget Device. The settings window of a storage device has a new Forget Device
  button. It drops the settings saved for that one device, so the class defaults apply to it
  again. Trident's own data for the device, such as a custom name, is kept. (Trident's Config
  page could always remove such an entry; this is the short way.)
• A changed NAK timeout reaches drives that are plugged in. It used to wait for the next
  replug, and setting it to 0 did not switch it off on a running drive at all. (The UAS queue
  depth, the unit number and what is mounted still need a replug.)
• USB in ROM: the ROM stack starts before your settings can be read. Workarounds it
  picked up during that time could end up saved as if you had chosen them; they no longer do.
• Built-in workarounds stay on. The handful of devices the class knows by name (certain
  Genesys, Olympus and Prolific bridges, among others) lost their built-in workaround whenever
  settings were saved or loaded while they were attached. It now stays.

One startup command

PsdStackLoader <device> <unit> now starts the whole stack, and the installer writes that one
line to S:User-Startup. AddUSBHardware and AddUSBClasses are no longer part of the
startup; an existing three-line startup keeps working.

Bug fixes

• USB Attached SCSI (UAS) drives that reject a command no longer fail to mount with "NAK
  timeout" errors.
• USB sticks and drives that stop answering no longer hang their volume. The AROS line
  had added a "device is busy" retry for Bulk-Only devices: after a NAK timeout it raised the
  timeout to one or two minutes, tried again without resetting the device, and kept trying
  without limit. That change is reverted. A timeout is again an ordinary error: the device
  is reset, the command is tried once more, and then the error is reported. The NAK timeout
  you set is never changed for you, and its default for new installations is 30 seconds
  instead of 60.
• Card readers and other multi-slot devices: the compatibility switches of the first slot
  leaked into every other slot, and could be saved for them.
• Drives that report a made-up geometry: the automatic switch to Simple SCSI for them
  never took effect. It does now, unless No Fallback is ticked.
• Trident, Config page: long device entries were cut off at the end, which removed the
  serial number, the one part that tells two identical sticks apart. They are now shortened
  in the middle.
• Serial adapters, printers and Palm devices: a failed write could go unnoticed, because the
  wrong half of the connection was checked for errors. It is now reported.
• Power saving: many network and storage devices could never be put to sleep; they now
  can. A device that refuses to suspend is no longer asked again at every timeout.
• USB audio: a playing device could be put to sleep by the power-saving timeout. A running
  stream now keeps it awake.
• French: some Trident texts showed stray characters; fixed.
• Composite devices (two-interface keyboard/mouse receivers, headsets): after a resume,
  only the first interface came back to life; now all of them do.
• Running a class driver as a program (typing its name in the Shell) no longer crashes; it
  now returns an error.
• ROM builds: modules now report their true size, so Kickstart no longer rescans code it
  has already passed.

Fixes ported from AROS

Poseidon is still developed in AROS, and fixes made there are brought over. The AROS commit
each one comes from is given in brackets.

• USB audio: a program that opened more AHI channels than the USB device has outputs - a
  four-channel module player on a stereo DAC, for instance - made AHI write past the end of its
  channel table. The class no longer replaces the channel count the program asked for.
  (31794523)
• USB network adapters (asixeth, cdceth, davicometh, ethwrap, moschipeth, pegasuseth,
  rndis): once a program had opened the device for exclusive use and closed it again, nothing
  else could open it - every later attempt failed with "unit busy" until the adapter was
  unplugged. Closing it now releases the exclusive claim. Not tested here. (9a984adb)
• Trident in French: the "Various Settings" heading on the Options page was shown in
  English although a translation existed. (translation repository, 1ac62187)
• Garbled copyright sign in About windows: Trident, the device popup and the settings
  windows of the audio, mass-storage, HID control, boot keyboard, boot mouse, MIDI, Palm and
  raw-wrapper classes showed � where the © belongs, and the game-controller class showed
  ©. The same goes for "Moiré" in the HID usage names. All non-ASCII characters in program
  text are now written so that no editor or tool can re-encode them, and the copyright notices
  use "(C)", which reads the same in every character set. (717dd468)
• Devices missing behind a hub: a hub that reports too short a power-on time was scanned
  too early, and a slow-starting device could be missed. External hubs now always get at least
  100 ms. (16dc9e95)
• USB storage: the capacity query is now retried once when a drive answers it with "unit
  attention" after a reset, like the class's other commands. (3312e74c)
• Bluetooth dongles: Broadcom dongles that use a vendor-specific interface class are now
  recognised; commands no longer block event reception; and event reception continues after a
  transfer error. Ported as they are; not tested here. (ec7643e2, 81516dcd, d02f90d5)
• USB in ROM without saved settings: keyboard and mouse stayed on the boot classes, because
  the handover to hid.class only ran when a prefs file existed. It now runs on the first
  start from disk in every case. (555fa3ae)
• Mouse wheel before hid.class takes over: bootmouse.class now has wheel support on by
  default, and no longer reads a wheel byte the mouse did not send. A saved bootmouse setting
  is kept. (555fa3ae)
• USB tethering (RNDIS): rndis.class now binds the RNDIS interface instead of the whole
  device, and recognises the common 2/2/255 form as well as the wireless one. A composite
  device keeps its other interfaces for their own classes. (4e1bd6ef)
• LAN78xx network adapters: received data is now read through four requests at a time, and
  a busy receive path can no longer hold up sending. Ported without the upstream change to the
  chip's idle behaviour; not tested here. (14ff79a6)

Poseidon v6.1

Last minute remarks

• In case you'd like to check out the option to put Poseidon to expansion ROM with xhci.device - please wait for the upcoming v2.1 release of the drivers. The currently released set has got the initialization priorities wrong.
• The settings are currently not embeddable into ROM, which means Poseidon starts with default settings. One of the consequences is that it will expect ODFileSystem present in ROM in order to boot from a CD. Thing is, ODFileSystem v0.7 will not work from ROM. So if you'd really want to try this out, you'd need to rebuild ODFileSystem with this PR: reinauer/ODFileSystem#21
• In case anyone is interested: yes, with patched ODFileSystem it is possible to boot the Amiga from an USB connected CD drive with say original AmigaOS 3.2 media and install the OS to e.g. NVMe drive. All while controlling the process via USB keyboard and mouse. That includes the Early Startup menu.

Release notes — Poseidon for AmigaOS 6.1

Everything in this archive — poseidon.library, all 29 class drivers, Trident, USBEject
and the command-line tools — reports version 6.1, followed by the CPU it was built
for.

Upgrading from 6.0: install the whole archive. Safe eject adds a poseidon.library
function that massstorage.class, Trident and USBEject call, and a new class against an
old library would guru. Your settings are untouched.

An archive for your CPU

6.0 shipped one -m68040 -mhard-float build, which a 68020 or 68030 could not run at all.
6.1 ships three archives — -020, -040 and -060, each with the usual -serial
diagnostic counterpart, and none needs an FPU. The -020 archive runs on all three
CPUs, just untuned for the faster two.

All three carry the same version number — it is the library's ABI version, not the build
variant — so Version LIBS:poseidon.library is how you tell which one you installed:
poseidon.library 6.1 (…) Poseidon for AmigaOS 68040.

Switching CPU variant needs the version requester. Installing one variant over another
is a same-version copy, which the installer's version check skips. Run the Installer at the
Average or Expert user level, where it offers to overwrite, or delete the installed
files first.

USB in ROM

On PiStorm/Emu68 the stack can now live inside a custom 2 MB Kickstart, which gets you a USB
mouse and keyboard in the boot menu and lets the machine boot from a USB drive — neither
of which is possible when the stack starts from S:User-Startup.

The kit is in the archive's ROM/ drawer: the ROM startup resident, the build script and
ROM-ReadMe.md. You run it on a PC (Linux, or Windows with WSL) with Python and amitools,
feeding it your own Kickstart plus two files from the Emu68 driver archive. It is optional and reversible:
nothing on the Amiga's hard drive changes, your normal installation keeps working,
and undoing it is one line in config.txt.

Because the drawer is for the PC, the Installer ignores it, and the -serial archives do not
carry it.

Safely remove hardware

USB drives can now be ejected cleanly before you unplug them. USBEject (optional,
SYS:WBStartup/) keeps an Eject item per drive in its own USB menu on Workbench;
Workbench replacements emulate only the flat AppMenu, so the items land in Tools
instead (under Directory Opus 5, switch Show Tools menu on). Trident's Devices page
has an Eject button that needs no Workbench at all. Both are localized in the usual six
languages.

An eject flushes every volume and asks each filesystem to inhibit. A filesystem that
reports a volume still in use vetoes the whole eject, and the requester names it — nothing
is unmounted, so there is no half-ejected state. Otherwise the write cache is synced, the
drive stopped and its hub port switched off, and a requester confirms it is safe to pull.

Two things it cannot see. A filesystem that does not implement Inhibit() — some CD
handlers — cannot report itself busy, so the eject proceeds once its buffers are flushed;
the error log says so when this happens. And raw access below the filesystem, such as a
backup or imaging tool on usbscsi.device, is invisible to it, exactly as on other systems.
Close your files before ejecting a disc.

MUI 3.8 is enough

6.0 required MUI 5 for Trident and the per-class settings dialogs. They now open on
muimaster.library 19 and up — MUI 3.8, MUI 4.0 and MUI 5 alike, with nothing given up
for it.

Kickstart 3.1 is enough

The floor is now Kickstart and Workbench 3.1, whole distribution. The big one: hid.class
sent every keystroke and mouse movement with an input.device V47 command that older systems
rejected and dropped, unlogged — so below 3.2 HID keyboards and mice did not work at all. It
now falls back to the V40 command. Trident no longer requires icon.library 44 nor USBEject
workbench.library 45; a lowlevel.library older than 40.27 no longer risks memory corruption
when a gamepad binds; datatypes.library and commodities.library are optional; and Install
checks for Installer 43.3 up front.

Differences below 3.2: see Known limitations in README.md.

A name and buffer count per filesystem

Mass storage used to apply one DOS name and one buffer count to everything it mounted, so a
CD came up as UMSD3 among your USB sticks with hard-disk buffering. The LUN Settings page
now carries a Mount name and buffers row per filesystem — FAT, NTFS, exFAT, CD/DVD — and,
like everything there, per LUN, so each slot of a card reader can be named separately.
Out of the box discs mount as UCD0 with 25 buffers and sticks stay in the UMSD0… sequence
with 100. RDB partitions are unaffected.

Upgrading: your existing name and buffer count carry over to every filesystem, so
mounts come up where they always did. Change the CD row if you want the new UCD* pool.

exFAT sticks mount

Modern USB sticks — anything above 32 GB, and most of what you buy preformatted — are exFAT,
and mass storage previously recognized them only well enough to skip them. They now mount,
whether the stick is a superfloppy or carries an MBR or GPT partition, identified by its own
boot sector, so one mislabelled as NTFS still lands on the right handler.

Two free downloads this archive does not contain are required: exFATFileSystem in L:
(relan's libexfat, ported by Fredrik Wikström, 68k branch by Tobias Karlsson; read and
write) and filesysbox.library 53 or newer in LIBS:, which it runs on — a 68020 binary, so
exFAT needs an 020 or better. Both are configured out of the box; without them exFAT media
are skipped. Clearing a handler row is now the off switch for any filesystem.

Eject exFAT drives before unplugging — the handler writes a volume-clean flag on its way
out, and pulling the stick first leaves it marked dirty.

Every disc ODFileSystem can read

A disc used to be refused unless it carried an ISO 9660 volume descriptor — all the old
CDFileSystem can read. Mass storage now recognises ODFileSystem by name and lets it
identify discs itself, so High Sierra, UDF, HFS and HFS+ mount alongside ISO 9660 (Joliet
and Rock Ridge included), and an audio CD mounts with its tracks as playable WAV files. An
unreadable disc is still refused, and any other handler keeps the ISO-only check.

ODFileSystem is not in this archive — it is a free download (Stefan Reinauer's, BSD).
Fresh installs set the CD/DVD DosType to CD01, its own, and the configured handler now
takes precedence over an older CD filesystem in a controller ROM. Existing settings keep
working.

---

Bug fixes

• USB keyboard: a key pressed for a command that then scrolls output for a while — Enter on
  a long List, say — could come back as several extra key presses once the command finished.
  The USB input tasks now run at a higher priority, so a busy console (or a task scheduler such
  as Executive) can no longer delay the key release.
• USB mouse wheel: scrolling under the same kind of load can no longer add extra steps.
• USB keyboard: Print Screen, Scroll Lock and Num Lock no longer act as a key that is never
  released, and pressing more keys than the keyboard can report no longer releases and re-presses
  the others.
• USB keyboard + mouse combos: Ctrl-Alt-Del is detected reliably.

Poseidon v6.0-alpha

Release notes — Poseidon for AmigaOS 6.0

The first release of Poseidon for AmigaOS — the Poseidon USB stack, back on the
machine it was written for, and taken a good deal further while it was at it.

Poseidon was Chris Hodges' USB stack for AmigaOS (2002–2009). In 2009 he placed the
sources into AROS, and the AROS Development Team carried them forward for the next
seventeen years — gaining a SuperSpeed hub class, more class drivers and a long tail of
fixes, none of which an AmigaOS machine could use. This project takes that evolved
version (AROS commit c01498ab) and brings it home, rebuilt for AmigaOS 3.2 on 68k.

The port is only half of it. Roughly as much of this release is new work that has no
counterpart in the AROS line at all: a second, purpose-built interface to the host
controller and the reworked SuperSpeed handling that rides on it, UAS mass storage over
bulk streams with several commands in flight and real error recovery, USB link power
management under your control, a rebuilt mounting layer, and the isochronous-audio and
suspend fixes. Everything below is the delta against that AROS baseline, and each section
says plainly whether it is new here or a port of something that was already there.

Everything in this archive — poseidon.library, all 29 class drivers, Trident and the
command-line tools — reports version 6.0.

---

Before you install

This is not a driver for your USB card. Poseidon is the stack that sits above the
hardware; it needs a USB host-controller driver to talk to, and none is included. This
release speaks two interfaces to such a driver — see Two host-controller
interfaces below — and both are confirmed working
against the Emu68 driver stack
(PiStorm / Emu68 on a Raspberry Pi 4 or CM4). Take its xhci.device 6.x if you can:
xhci.device 5.x works too, with fewer features.

It replaces an existing Poseidon. Installing puts poseidon.library in LIBS: and
the class drivers in SYS:Classes/USB/, over whatever is there now. The installer
version-checks every file and never replaces a newer one without asking, but a classic
Poseidon 4.x installation will end up being upgraded. Your settings are kept: they live
in ENVARC:Sys/poseidon.prefs as before, and this release reads the classic file format
(AROS had changed the file's identifier; it is changed back).

The classes in this archive need poseidon.library 6. They will not load against a
4.x or 5.x library, so install the whole archive rather than picking pieces out of it.

68040 or 68060 with FPU. The released binaries are built for -m68040 -mhard-float
and will not run on a 68000/010/020/030. Building from source for another CPU is a
one-line change.

---

What's new since the AROS 5.x line

It runs on AmigaOS 3.2

This is the port half. The whole distribution was rebuilt as a native AmigaOS program
set: the AROS build system is gone, and with it the parts that only make sense on AROS —
AROS's own USB host controllers, the hosted virtual controller, and the Allwinner-specific
felsunxi class. What remains is 29 class drivers, the library, Trident and five shell
commands, plus an Installer script, icons, a datatype so Poseidon preset files show their
own icon, and Trident's translations (Czech, French, Italian, Polish, Russian, Spanish).

Two host-controller interfaces

New here. Poseidon has always talked to a host-controller driver by filling in a
struct IOUsbHWReq and sending it — the classic V1/V2 request. It describes a transfer
the way a 1998 controller wanted one: an address, an endpoint, a buffer. A USB 3.0
controller wants something quite different. It keeps contexts — durable per-device and
per-endpoint state it is told once — and a transfer is then a tiny reference to those
contexts. Handed only classic requests, an xHCI driver has to reconstruct the topology by
watching the traffic go past and guessing what the stack is doing.

AROS's answer was to make the request bigger: a "V3" extension that bolted the missing
topology onto every transfer — route string, root port, split/TT parameters, SuperSpeed
burst settings. It never existed on AmigaOS, and it puts once-per-device facts on the hot
path of every single transfer. It has been removed here, and replaced with a context
interface: a set of operations that establish a device and its endpoints once, up front,
so the transfers themselves stay small. It is what xhci.device 6.x speaks, and where
everything below in this section comes from.

The legacy interface stays, and stays frozen. The classic V1+V2 IOUsbHWReq layout
and every UHCMD_*/UHIOERR_* value are untouched binary contract, because classic Amiga
USB cards — Deneb, Subway and friends — depend on them. The stack keeps a complete second
backend for that path; it is not a compatibility stub.

Both are confirmed working: xhci.device 6.x on the context interface, and
xhci.device 5.x on the legacy one. You tell the stack which driver to open, and it
asks that driver which interface it implements. On the legacy interface you get the
classic feature set — USB 2.0-style device handling, no hubss.class binding (so no
SuperSpeed hubs), no UAS bulk streams, and whatever link power management the driver
chooses to do on its own, with no control from Trident.

USB 3.0 that behaves like USB 3.0

New here, on the context interface. A SuperSpeed device now reaches the stack as a
SuperSpeed device instead of being disguised as USB 2.0 on the way in, and the SuperSpeed
hub driver was largely rewritten around that: real SuperSpeed port status handling, warm
reset, and no leftover low/full-speed machinery that a USB 3 hub never needs.

A USB 3 drive no longer shows up twice. A USB 3 hub presents itself as two hubs — a
USB 2.0 half and a SuperSpeed half — and a SuperSpeed device appears on both. The two
halves are now recognised as one hub (they share a Container ID) and the duplicate is
dropped. The same race could also hang the machine while a device was being plugged in;
that is fixed too.

Mass storage: UAS, and much faster

New here. The AROS line had the outline of a UAS transport; it is now a working one.
Several commands run on the drive at once over USB 3.0 bulk streams, with a queue
depth you can set in the settings, and UAS is preferred over the older BOT transport
when a drive offers both (the preference is a switch, so you can force BOT). Bulk streams
need the context interface; on the legacy one UAS falls back to one command at a time.

The class also stopped getting in its own way: reading geometry, mode pages, HD_SCSICMD
passthrough and seeks used to stall all block I/O while they ran, and no longer do. NAK
timeout handling was reworked, which mostly matters to CD and DVD drives that take their
time.

Recovery instead of a hang. When a command wedges, the class now aborts just that
command with a UAS task-management request; if the drive ignores that, it resets the
device. Previously a stuck command took the drive — and sometimes the machine — with it.

Mounting, rewritten

New here. Every kind of medium is now mounted through the A4091 mounter (the partition
parser from a4091.device), which handles RDB, MBR, GPT and superfloppy disks from
one place. In the mass-storage settings you choose which handler to use for FAT,
NTFS and CD/DVD partitions, and audio CDs are recognised as such. The automount
switches — RDB, MBR/GPT, CD/DVD, and unmounting on removal — are all per-category.

USB power management

New here, on the context interface: a link power management switch that lets an
idle link drop into a low-power state between transfers (U1/U2 on SuperSpeed, L1 on
High-Speed) and enables Latency Tolerance Messaging. Entry and exit are handled by the
controller and are invisible to transfers; turn it off if a device misbehaves when idle.
It applies immediately, both ways, and can be overridden per device. (A legacy-interface
driver may do link power management of its own accord — xhci.device 5.x does — but the
stack has no say in it and Trident's switch does nothing there.)

Poseidon's older power-saving suspend got a working-over alongside it: root hubs can now
be suspended, and remote wakeup is always armed before a device is put to sleep, so a
suspended keyboard or mouse brings its own link back up the moment you use it instead of
needing a replug. Devices that should not sleep — a mounted drive, an active HID device —
no longer do. Unplugging a device while it was suspended used to leave it lingering in
the stack; that is fixed.

USB audio keeps playing

Rescanning the class drivers used to tear down isochronous transfers, which meant audio
stopped whenever the stack rescanned. The scan was reworked so it does not. A wrong
endpoint address the audio class sent on the wire — which xhci.device used to carry a
workaround for — was fixed at the source.

"Use plain, factual messages"

New here. Poseidon traditionally reports itself with a certain amount of humour. In Trident,
Various Settings → Logging Options → Use plain, factual messages swaps every log
entry, popup and requester for a short factual one. It is off by default, so the
traditional wording is what you get unless you ask for otherwise, and it takes effect
immediately for new messages.

Other fixes

• The HID class settings window no longer crashes when opened.
• Hubs cope with a device disappearing mid-operation instead of reporting a string of
  spurious errors.
• After a resume, hubs no longer submitted the same status-change request twice.
• Unknown hubs and devices get a sensible n...

Poseidon backport v0.1-pre-alpha

This is first public, pre-alpha quality release of a backport of the Poseidon USB stack from AROS to AmigaOS 3.2.

Backup everything and expect it to crash. If you do decide to try it, please report any issues found. Reproduce the issue using the "-serial" debug build and collect a debug log using Sashimi.

There are no USB controller drivers included in this release - you need to source these separately.

Read the README.md in the root of the repository and read the README included in the archive.

Please note the requirements are currently fairly strict - subject to change over time if there is demand:

Requirements

• An Amiga running AmigaOS 3.2.
• A 68040 / 68060 with FPU — the binaries are built for -m68040 -mhard-float.
• MUI 5 for the Trident GUI and the per-class configuration dialogs.
• A USB host-controller device implementing the Poseidon hardware interface (devices/usbhardware.h), for example xhci.device.
• For xhci.device - you want version 5.2 or newer

So far tested only on following setup: A1200 + Pitorm32-lite + custom Emu68 build (based on 1.1-alpha) + AmigaOS 3.2.3 + xhci.device