GitLab

Add workaround for incorrect sizeof(usb_status_t), disable debug_set_stamp_debug(TRUE) to fix interrupt issue

Detail:
  build/c/ehcimodule, build/c/ohcimodule, build/c/usbmodule - Make sure that timestamped debug messages aren't used, as they appear to cause interrupts to be briefly enabled, which causes all kinds of problems in sensitive bits of the code.
  dev/usb/c/usbdi, dev/usb/c/usbdi_util, dev/usb/h/usb - Because DISABLE_PACKED is being used to compile the USB drivers, sizeof(usb_status_t) was being calculated as 4 bytes instead of 2. Somehow this works fine with the EHCI driver (and presumably the OHCI driver), but it was causing problems with the MUSBDriver, so a hardcoded size of 2 is now used.
Admin:
  Tested on rev C2 beagleboard.


Version 0.48. Tagged as 'NetBSD-0_48'

Add support for HAL enumerated USB EHCI controllers, use HAL_IRQClear when clearing EHCI/OHCI controller interrupts

Detail:
  * HAL enumeration of EHCI controllers is performed in the same manner as for OHCI, and protected by #ifdef RHENIUM. HAL_USBControllerInfo must return with the usbinfo type field set to 1 to indicate EHCI.
  * HAL_IRQClear is now called by the EHCI & OHCI modules after clearing the interrupt flag in the USB controller, to allow the HAL to reset the interrupt controller if required (e.g. for OMAP3)
  * EHCI module (for HAL EHCI controllers) now has the option of forcing all EHCI register access to be 32bit size & alignment (requirement is indicated in OMAP3 TRM, although USB did seem to function fine without)
  build/c/ehcimodule - Add enumeration of HAL USB controllers. Add support for forcing all EHCI register access to be 32bit size & alignment. Add riscos_irqclear() implementation. Guard against unexpected behaviour if PCI module isn't present.
  build/c/ohcimodule - Add riscos_i...

	Inclusion of Mods to c.ehci (from Dan) to correctly handle low
	speed interrupt devices beyond USB2 hubs (i.e. keyboards and mice).
	Mod to EHCI resetting in service call handler to better cope with
	USBDriver being restarted 'underneath' it.
Detail:
Admin:
	Castle, Dan Ellis, NetBSD added IP. Tested at Castle. Will be
	beta tested at this revision level before full release.


Version 0.39. Tagged as 'NetBSD-0_39'

        Added (and fixed code underlying) to give periodic
        explore of USB busses (cf NetBSD).
        Added USBDisover command to let user provoke this too.
        Reworked USB device number allocation to roll
        around at 999 to avoid field overflow in
        a number of places.
Detail:
Admin:  castle added IP. to be beta tested on beta test site


Version 0.37. Tagged as 'NetBSD-0_37'

Detail:
        USB_PORT_RESET_DELAY had been increased from 100 to 200ms in previous
        version. This was OK for most devices, and beneficial for some, BUT
        killed the Chicony keyboards shipped with many iyonix.. Delay has
        been reverted to 100ms, and is now run time alterable via the
        USB$PortResetDelay sysvar. This problem did not show up in beta testing!

Admin:  tested by JB with chicony.. delays up to around 160ms work with
        the chicony.
        Castle added IP

Version 0.36. Tagged as 'NetBSD-0_36'

Detail:
         - VarVal launching.. string length was incorrectly calculated.
         - USB VarVals now correctly cleared if device failed during
           enumeration, after first SetVarVal call.
         - Triggercbs() modified to check interrupt context and be
           refuse (quietly) to trigger callbacks if interrupts threaded
           (result of possible issue .. but perhaps wise)
         - Added third QueryDelay variable, and an Info file to describe
           what does what.
Admin:
         checked at Castle.. due for extended check via beta test site
         before formal release.

         Castle + D Ellis added IP


Version 0.35. Tagged as 'NetBSD-0_35'

        USB2 further debugged.
Detail:
        Check in prior to second beta release. Improvements also to
        softloading of replacement modules.
Admin:
        Checked with JB .. further crosschecking is still needed for rhenium.
        Castle/D Ellis added ip from NetBSD sources


Version 0.33. Tagged as 'NetBSD-0_33'

Update of EHCIDriver to current NetBSD, together with needed changes in both OHCIDriver and USBDriver.

Detail:
        Code may change before formal release .. will be beta released
        ... Entered to CVS to get new version numbers, and for
        the record.
Admin:
        Castle/netbsd added IP .. will be tested more fully before
        formal release. Still to be verified on Rhenium.


Version 0.31. Tagged as 'NetBSD-0_31'

change to OHCIModule to pay heed to number of OHCI i/f reported by HAL (in Rhenium).. various other small adjustments to improve softloadability, both in OHCI and EHCI drivers.

Detail:
Admin:
    tested on iyonix and on rhenium
    castle added IP


Version 0.30. Tagged as 'NetBSD-0_30'

  'blocked' several holes that could cause freezes with awkward
  usb devices.  Added latest changes from Dan Ellis too , especially for USB2
Detail:
 many
Admin:
 Tested in ROM at Castle, and beta tested with users


Version 0.28. Tagged as 'NetBSD-0_28'

Version 0.27. Tagged as 'NetBSD-0_27'

Version 0.26. Tagged as 'NetBSD-0_26'

Detail:
  * All modules will now rmkill/rmload more reliably
  * Keyboard driver disables kernel debounce on new enough kernels, needed
    for many PS2-USB adaptors, bar code scanners etc
  * Mouse driver will scroll wheel scroll any windows that do respond to
    scroll events, otherwise it issues PointerReason_WheelChange (9) to an
    external WimpScroll module
  * Several bugs and an interrupt hole resolved
Admin:
  Tested at Framlingham.

Version 0.25. Tagged as 'NetBSD-0_25'

Version 0.23. Tagged as 'NetBSD-0_23'

Version 0.22. Tagged as 'NetBSD-0_22'

Detail:
  NetBSD/FreeBSD differences in the naming of certain macros are now resolved
  in TCPIPLibs instead of locally. Also some reordering in the !Mk files in
  order to work with current versions of Zap.
Admin:
  Builds against TCPIPLibs 5.47. Note that this version of the USB sources
  also now requires TCPIPLibs 5.47.

Version 0.21. Tagged as 'NetBSD-0_21'

> Reduce root port reset delay to 50ms from 250ms to work with Chicony
> keyboards (again).  Comment out duplicate free(udev) at end of
> usbmodule:config_detach.

Version 0.20. Tagged as 'NetBSD-0_20'

Version 0.19. Tagged as 'NetBSD-0_19'

Added new CallDevice reason code (&80000005) to clear a stall condition on a pipe with a given FileSwitch handle

New usbdevs and usbdevs_data from NetBSD
Rewrote power source detection to skirt round dodgy devices which say
they're self powered,set MaxPower=0,and don't respond to a getstatus()
Copes better with devices being unplugged while a transfer using them is active

Version 0.17. Tagged as 'NetBSD-0_17'

Detail:
  * Errors now use the allocated error number base, &819000, instead of a
    base in the private range, &50000000. "USB transfer failed" errors now
    have use a sub-range of error numbers from &819020; the original BSD
    error number can be retrieved from the bottom 5 bits of the RISC OS
    error number.
  * Service_USB 1 now returns the list in the order in which the devices,
    rather than in the reverse order as previous versions of the module did.
  * A little tweaking of h/usb_port to reduce compiler warnings.
  * Some compiler-satisfying changes to allow use of packed structures.
    (However, the compiler doesn't seem to work correctly yet, so this is
    currently switched out using -DDISABLE_PACKED on the command line.)
  * Added full stops to end of star command help strings for consistency
    with rest of OS.
  * New exports: USBDevFS (moved here from ^.h) and a few files in dev/usb/h
    and sys/h.
Admin:
  Error number overhaul and...

Put the port reset delay back to 50ms.Not sure why NetBSD decided to change that but it broke the Chicony keyboard supplied in Castle's Iyonix.

Version 0.15. Tagged as 'NetBSD-0_15'

Added stuff to delete the DeviceFS$USB*$Options variables when the module is
RMkilled.

Version 0.12. Tagged as 'NetBSD-0_12'

Correction of spelling of busses to buses (noun plural).
OHCIdriver only responds to PCI service call to lookup vendor name when
the device is of class OHCI controller.

Version 0.01. Tagged as 'NetBSD-0_01'

This is the initial import of prototype USB drivers written for the OHCI chip on the customer L board.

Details:
  The drivers are for mouse, keyboard and combined keyboard and mouse.  They
assume that the correct device is going to be plugged in, not reading the
descriptors at all.  Once the mouse is plugged in, it will become the active
pointer device, reverting to the previous one when extracted.

Admin:
  The prototypes have been tested on a SDB customer L board, but require a couple
of pokes to the reset and configuration register to get them going:
  memorya f99f611c 3840 works on the SDB but will be different on the customer L board