ehci: add EHCIF_32BIT_ACCESS flag to force 32-bit MMIO

ehci(4): Serialize access to portsc registers.

Both ehci_roothub_ctrl and ehci_suspend/resume do r/m/w on them, so
use a mutex to serialize access to avoid stomping on each other.

ehci(4): Fix doorbell synchronization.

ehci_sync_hc was previously subject to spurious wakeup, in which case
the CPU might proceed from aborting and recycle a DMA buffer before
the hardware was done writing to it. Now the code is not subject to
spurious wakeup -- it waits (up to the 1sec timeout) for the relevant
interrupt to be delivered, not for anything else.

Three fixes

- pass the 64bit DMA tag if the HCCPARAMS says ehci supports it and the
64bit DMA tag is available/valid. This should help with the
"cannot create xfer" error on 64bit systems.

- restrict the control structure memory allocation to the low 4GB
(CTRLDSSEGMENT was always set to zero anyway)

- if ehci_init fails then tidyup before returning the error.

Use roundup. NFCI.

Consistency in use of #define names. No functional change as values of
before and after defines are the same.

implement a gross hack to fix "boot -a" on systems with usb keyboards on
systems with ehci handover to uhci (and maybe ohci), and fix a similar
problem for "boot -s".

there is effort to ensure that all devices attached via USB are probed
before RB_ASKNAME or RB_SINGLE attempts to ask any questions on the console,
and largely this works, often by chance, today, for USB disks and root.
i've recently pushed this more into uhub and general USB device attachment
as well, and kept a config_pending reference across the first explore of
a bus. these fix many issues with directly attached hubs.

however, on systems where devices connected to ehci ports are handed over
to a companion uhci or ohci port, it may not be the first, or even second,
bus explore that finds the device finally before attachment, and at this
point all config_pending references are dropped.

there is no direct communication between drivers, the potentials are
looked up but their device_t is only used for generic things like the name,
so informing the correct companion to expect a device and deal with the
config_pending references is not possible without some fairly ugly layer
violations or multi-level callbacks (eg, we have "ehci0", and usually an
the relevant companion, eg, "uhci2", but it is the uhub that uhci2 has
attached that will deal with the device attachment.)

with the above fixes to generic USB code, the disown happens during the
first explore. the hack works by, at this point, checking if (a) root
is not mounted, (b) single user or ask name are set, and (c) if the hack
as not been triggered already. if all 3 conditions are true, then a
config_pending_incr() is called and a callback is triggered for (default)
5 seconds to call config_pending_decr(). ehci detach pauses waiting for
this callback if scheduled.

this allows enough time for the uhub and the ukbd/wskbd to attach before
the RK_ASKROOT prompts appear. testing shows it takes between 1.5 and
2 seconds for the keyboard to appear after the disown occurs.


Index: dev/usb/ehcivar.c
- new sc_compcallout, sc_compcallout, sc_complock, and a state for th
handover hack.

Index: dev/usb/ehci.c
ehci_init():
- use aprint_normal_dev() instead of manual device_xname().
- initialise sc_compcallout, sc_compcallout, sc_complock, and sc_comp_state.
ehci_detach():
- if there are companion controllers, tear own the above, including waiting
if there is a callback scheduled.
ehci_disown_callback():
- new callout to call config_pending_decr() in the the future.
schedule this ca
ehci_disown_sched_callback():
- if booting to single user or asking names, call config_pending_incr() and
schedule the callout above, default 5 second delay.
ehci_disown():
- if disowning a port call ehci_disown_sched_callback().

pull across abort fixes from nick-nhusb. add more abort fixes, using
ideas from Taylor and Nick, and myself. special thanks to both who
inspired much of the code here, if not wrote it directly.

among other problems, this assert should no longer trigger:

panic: kernel diagnostic assertion "xfer->ux_state == XFER_ONQU" failed: file "/current/src/sys/dev/usb/usbdi.c", line 914

using usbhist i was able to track down my instance of it being related
to userland close() beginning, dropping the sc_lock, and then the usb
softintr completes the transfer normally, and when it is done, the
abort path attempts to re-complete the transfer, and the above assert
is tripped.


changes from nhusb were commited with these logs:
--
Move the struct usb_task to struct usbd_xfer for everyone to use.
--
Set device transfer status to USBD_IN_PROGRESS if start methods succeeds
--
Actually set the transfer status on transfers in ohci_abort_xfer and
the controller is dying
--
Don't supply the lock to callout_halt when polling as it won't be held
--
Improve transfer abort
--
Mark device transfers as USBD_IN_PROGRESS appropriately and improve
abort handling
--
#ifdef DIAGNOSTIC -> KASSERT and add another KASSERT
--
Mark device transfers as USBD_IN_PROGRESS appropriately and improve
abort handling
--

additional changes include:
- initialise the usb abort task in the HCI allocx routine, so that it
can be safely usb_rem_task()'d.
- rework the handling of softintr vs cancellation vs timeout abort based
upon a scheme from Taylor:
when completing a transfer normally:
- if the status is not in progress, it must be cancelled or timed out,
and we should not process this xfer.
- set the status as normal.
- unconditionallly callout_stop() and usb_rem_task(). they're safe and
either aren't running, or will run and do nothing.
- finally call usb_transfer_complete().
when aborting a transfer:
- status should be cancelled or timed out.
- if cancelling, callout_halt and usb_rem_task_wait() to make sure the
timer is either done or cancelled.
- at this point, the ux_status must not be cancelled or timed out, and
if it is not in progress we're done.
- set the status.
- if the controller is dying, just return.
- perform HCI-specific tasks to abort this xfer.
- finally call usb_transfer_complete().
for the timeout and timeout task:
- if the HCI is not dying, and the ux_status is in progress, then
trigger the usb abort task.
- remove UXFER_ABORTWAIT and UXFER_ABORTING.

tested on:
- multiple PC systems with several types of devices: ugen/UPS, ucom,
umass with disk, ssd and cdrom backends, kbd, ms, using uhci, ehci
and xhci.
- erlite3: sd@umass on dwc2.
- sunblade2000: kbd/ms and umass disk on ohci.

untested:
- motg, slhci and ahci. motg has some portion of the new scheme
applied, but slhci and ahci require more study.

future work includes pushing a lot of the common abort handling into
usbdi.c and leaving upm_abort() for HC specific tasks, but this change
is pullup-able to netbsd-7 and netbsd-8 as it does not change any
external API, as well as removing over 100 lines of code while adding
over 30 new asserts.

XXX: pullup-7, pullup-8.

branches: 1.44.2;
Stop potential misuse of vendor names and USB vendor IDs in root hub
device and string descriptors.

Firstly: Few vendors have identical PCI-SIG vendor IDs and USB-IF vendor
IDs. As such, using the PCI vendor ID as a USB vendor ID may trample
on whomever is allocated that USB vendor ID.

Secondly: The vendor of the host controller hardware implementation has
little to nothing to do with our usbroothub implementation. Thus we
should not potentially associate any problems therewith to such third
party.

This change will result in root hubs being identified by USB Vendor ID
0x0000. Root hub vendor string will now be "NetBSD" (or, specifically:
ostype). Product ID (0x0000) and product strings remain unchanged.

branches: 1.43.10; 1.43.16;
Merge nick-nhusb

- API / infrastructure changes to support memory management changes.
- Memory management improvements and bug fixes.
- HCDs should now be MP safe
- conversion to KERNHIST based debug
- FS/LS isoc support on ehci(4).
- conversion to kmem(9)
- Some USB 3 support - mostly from Takahiro HAYASHI (t-hash).
- interrupt transfers now get proper DMA operations
- general bug fixes
- kern/48308
- uhub status notification improvements
- umass(4) probe fix (applied to HEAD already)
- ohci(4) short transfer fix

branches: 1.42.12; 1.42.14; 1.42.18;
Use CACHE_LINE_SIZE for EHCI_SQTD_ALIGN to make sure the sqtd fully consume
a cacheline.

use a pool instead of a linked list to avoid synchronization problems.

branches: 1.40.2;
merge the jmcneill-usbmp branch. many thanks to jared for the
initial work, and every one else who has tested things for me.
this is largely my fault at this point :-)

the main changes are something like:

- usbd_bus_methods{} gains a get_lock() to enable the
host controller to provide a lock for the USB code.
if the lock isn't provided, old-style protection is
(partially) applied.

- ehci/ohci/uhci have been converted to the new
interfaces, including mutex/cv/etc conversion.

- usbdivar.h contains a discussion about locking and
what locks are held for which method calls. more
to come for usbdi(9) here.

- audio drivers (uaudio, umidi, auvitek) have been
properly SMPified now that USB is ready.

- scsi drivers have been modified to take the kernel
lock explicitly before calling into scsi code.

- usb pipes are associated with a lock, that is the
same as the controller lock. (this could be split
up further in the future.)

- several usbfoo_locked() or usbfoo_unlocked()
functions have been added to the usbdi(9) to
enable functionality with or without the USB
lock (per controller) already being held.

the TODO.usbmp file has specific details on what is left to
do, including what device-specific changes should be done now
that the whole framework is ready.

pull down from usbmp:

kill the !USE_USE_SOFTINTR code.

branches: 1.38.6; 1.38.10;
Add preliminary support for Embedded Transaction Translator Function (as
found on the MPC8536 and AR9334) which allows low/full devices to be
connected to an EHCI root hub.

branches: 1.37.2;
Add vendor dependent functions.
sc_vendor_init()/sc_vendor_port_status().

branches: 1.36.2;
A pointer typedef entails trading too much flexibility to declare const
and non-const types, and the kernel uses both const and non-const
PMF qualifiers and device suspensors, so change the pmf_qual_t and
device_suspensor_t typedefs from "pointers to const" to non-pointer,
non-const types.

branches: 1.35.2;
Expand PMF_FN_* macros.

Expand <dev/usb/usb_port.h> definitions, and lightly unifdef(1).

branches: 1.33.16;
kern/39727: race condition in ehci isoc abort xfer path

From Jeremy Morse:

* Serialize access to the ehci intrlist.
* Change the ehci intrlist to a tailq so xfers are not queued out of order.
* In ehci_check_itd_intr, don't treat a transfer error as an indication
that the xfer is no longer active.

High speed isochronous support, from Jeremy Morse as part of Google
Summer of Code 2008.

branches: 1.31.2;
Add appropriate bus_dmamap_sync() calls to uhci(4), ohci(4) and ehci(4),
as proposed on tech-kern@.
While all DMA memory used in the USB framework is mapped BUS_DMAMAP_COHERENT
(including data memory, which is memcpy'd from/to USB-private buffers),
the CPU can reorder loads or stores from/to main memory, causing the
controller to have an incoherent view of the DMA descriptors lists for a
short time. bus_dmamap_sync() should contain memory barriers that prevents
the CPU from reordering load/store. Note that BUS_DMAMAP_COHERENT is
still required for the DMA descriptor lists - these can't work properly
with software cache coherency.
This fixes "host controller process error/host controller halted" errors
I'm occasionally seeing with a umodem device on uhci on x86.
Thanks to Michael Lorenz for testing it on his O2, and Izumi Tsutsui on
his Cobalt Qube 2700.

branches: 1.30.2; 1.30.4;
Remove clause 3 and 4 from TNF licenses

branches: 1.29.2; 1.29.4;
split device/softc for USB host controllers and the usb (control)
device,
this is hairy stuff, and I've only tested with uhci/ehci at pci,
please test the rest and report problems

Use device_t and accessors. Setup power management in the PCI
front-end (XXX needs to look more alike the Cardbus front-end).
Establish the shutdown hook using PMF.

Add a method to detach child.

Cosmetic: use device_t and accessors. Use aprint_*_dev(). Use
PMF_FN_{ARGS,CALL,PROTO}.

branches: 1.26.6; 1.26.10;
Merge jmcneill-pm branch.

branches: 1.25.2;
lockmgr -> mutex

branches: 1.24.14; 1.24.20; 1.24.22; 1.24.28; 1.24.30;
Remove the code which disables port status change interrupts for 1s
when one occured -- this makes that events get lost (or delayed until
the next change), in particular in the handover process from the EHCI
to a companion controller, laeving the port dead under some circumstances.
This fixes the "can't reconnect" symptom for me.
I don't see why the interrupt should be disabled - it is acknowledged
in the interrupt handler and shouldn't be active again until the next
status change.

branches: 1.23.18; 1.23.22; 1.23.24;
From OpenBSD:

--
Add a workaround for VIA EHCI controllers which, under load, signal qTD
completion before they have performed writeback from the overlay qTD.

This condition would exhibit itself as a umass stall that never recovers.
--

This fixes the problem reported by Thomas Klausner on current-users@:
http://mail-index.netbsd.org/current-users/2006/01/17/0000.html

branches: 1.22.2;
changes for making DIAGNOSTIC not change the kernel ABI:
- for structure fields that are conditionally present,
make those fields always present.
- for functions which are conditionally inline, make them never inline.
- remove some other functions which are conditionally defined but
don't actually do anything anymore.
- make a lock-debugging function conditional on only LOCKDEBUG.

as discussed on tech-kern some time back.

Keep a flag for each port that keeps track of if the port has been reset
instead of having one flag for the controller.

Also, don't try to modify read-only power bits if the controller doesn't
support power cobntrol.

branches: 1.20.6;
Minor cosmetic change.

branches: 1.19.2;
Allow 32 chars in the saved vendor string. Fixes kern/29760.

Add define for EHCI_MAX_POLLRATE.

branches: 1.17.2; 1.17.8;
Merge from FreeBSD:
sc_child is only used on NetBSD and OpenBSD, ifdef appropriately
FreeBSD revisions:
ehcivar.h 1.8
ohcivar.h 1.39
uhicvar.h 1.39

Implement an USB memory reserve. Allocate some memory per host controller
in the wanted bus_dma space. If an allocation fails during operation
(i.e. when hotplugging an umass device), fall back to using the reserve.

The amount can be configured as USB_MEM_RESERVE. The default value is
256k.

Ideally, there would be a way to steal pages in the desired area from
UVM, but that's far more complicated, and this is not intrusive, plus
it works.

Add interrupt transfers. From OpenBSD.
Having interrupt transfers enables some more device types to be used,
most notably USB2 hubs. USB2 hubs still cannot handle anything but
USB2 devices, because there is node code to handle hub Transaction Translation
yet.
XXX This code doesn't do good bw scheduling, but it's certainly better than
nothing. :)

Insert from ifdefs for soft interrupt related stuff. From OpenBSD.

Add suspend/resume support.
From OpenBSD.

branches: 1.12.16; 1.12.20;
Change abort slightly.
Change debug output.

Add some interrupt processing.

Set up control xfers.
Handle xfer timeouts.
Better debug messages.

Simplify async list handling.

Start of pipe open routine.

Add more fields to hardware structs.
Add memory allocation for various data structures.

Add some more data structure defs.

Recognize port status changes.
Hand over low and full speed devices to companion controller.

Add root hub emulation.

Add some register defines.
Print companion controllers.

Add some useful macros.

branches: 1.1.2; 1.1.4; 1.1.6; 1.1.10;
Add some placeholders for the EHCI (USB 2) driver.

Don't get your hopes up. I've not even finished reading the
(100+) page spec, and I have no hardware.

ehci(4): Serialize access to portsc registers.

Both ehci_roothub_ctrl and ehci_suspend/resume do r/m/w on them, so
use a mutex to serialize access to avoid stomping on each other.

ehci(4): Fix doorbell synchronization.

ehci_sync_hc was previously subject to spurious wakeup, in which case
the CPU might proceed from aborting and recycle a DMA buffer before
the hardware was done writing to it. Now the code is not subject to
spurious wakeup -- it waits (up to the 1sec timeout) for the relevant
interrupt to be delivered, not for anything else.

Three fixes

- pass the 64bit DMA tag if the HCCPARAMS says ehci supports it and the
64bit DMA tag is available/valid. This should help with the
"cannot create xfer" error on 64bit systems.

- restrict the control structure memory allocation to the low 4GB
(CTRLDSSEGMENT was always set to zero anyway)

- if ehci_init fails then tidyup before returning the error.

Use roundup. NFCI.

Consistency in use of #define names. No functional change as values of
before and after defines are the same.

implement a gross hack to fix "boot -a" on systems with usb keyboards on
systems with ehci handover to uhci (and maybe ohci), and fix a similar
problem for "boot -s".

there is effort to ensure that all devices attached via USB are probed
before RB_ASKNAME or RB_SINGLE attempts to ask any questions on the console,
and largely this works, often by chance, today, for USB disks and root.
i've recently pushed this more into uhub and general USB device attachment
as well, and kept a config_pending reference across the first explore of
a bus. these fix many issues with directly attached hubs.

however, on systems where devices connected to ehci ports are handed over
to a companion uhci or ohci port, it may not be the first, or even second,
bus explore that finds the device finally before attachment, and at this
point all config_pending references are dropped.

there is no direct communication between drivers, the potentials are
looked up but their device_t is only used for generic things like the name,
so informing the correct companion to expect a device and deal with the
config_pending references is not possible without some fairly ugly layer
violations or multi-level callbacks (eg, we have "ehci0", and usually an
the relevant companion, eg, "uhci2", but it is the uhub that uhci2 has
attached that will deal with the device attachment.)

with the above fixes to generic USB code, the disown happens during the
first explore. the hack works by, at this point, checking if (a) root
is not mounted, (b) single user or ask name are set, and (c) if the hack
as not been triggered already. if all 3 conditions are true, then a
config_pending_incr() is called and a callback is triggered for (default)
5 seconds to call config_pending_decr(). ehci detach pauses waiting for
this callback if scheduled.

this allows enough time for the uhub and the ukbd/wskbd to attach before
the RK_ASKROOT prompts appear. testing shows it takes between 1.5 and
2 seconds for the keyboard to appear after the disown occurs.


Index: dev/usb/ehcivar.c
- new sc_compcallout, sc_compcallout, sc_complock, and a state for th
handover hack.

Index: dev/usb/ehci.c
ehci_init():
- use aprint_normal_dev() instead of manual device_xname().
- initialise sc_compcallout, sc_compcallout, sc_complock, and sc_comp_state.
ehci_detach():
- if there are companion controllers, tear own the above, including waiting
if there is a callback scheduled.
ehci_disown_callback():
- new callout to call config_pending_decr() in the the future.
schedule this ca
ehci_disown_sched_callback():
- if booting to single user or asking names, call config_pending_incr() and
schedule the callout above, default 5 second delay.
ehci_disown():
- if disowning a port call ehci_disown_sched_callback().

pull across abort fixes from nick-nhusb. add more abort fixes, using
ideas from Taylor and Nick, and myself. special thanks to both who
inspired much of the code here, if not wrote it directly.

among other problems, this assert should no longer trigger:

panic: kernel diagnostic assertion "xfer->ux_state == XFER_ONQU" failed: file "/current/src/sys/dev/usb/usbdi.c", line 914

using usbhist i was able to track down my instance of it being related
to userland close() beginning, dropping the sc_lock, and then the usb
softintr completes the transfer normally, and when it is done, the
abort path attempts to re-complete the transfer, and the above assert
is tripped.


changes from nhusb were commited with these logs:
--
Move the struct usb_task to struct usbd_xfer for everyone to use.
--
Set device transfer status to USBD_IN_PROGRESS if start methods succeeds
--
Actually set the transfer status on transfers in ohci_abort_xfer and
the controller is dying
--
Don't supply the lock to callout_halt when polling as it won't be held
--
Improve transfer abort
--
Mark device transfers as USBD_IN_PROGRESS appropriately and improve
abort handling
--
#ifdef DIAGNOSTIC -> KASSERT and add another KASSERT
--
Mark device transfers as USBD_IN_PROGRESS appropriately and improve
abort handling
--

additional changes include:
- initialise the usb abort task in the HCI allocx routine, so that it
can be safely usb_rem_task()'d.
- rework the handling of softintr vs cancellation vs timeout abort based
upon a scheme from Taylor:
when completing a transfer normally:
- if the status is not in progress, it must be cancelled or timed out,
and we should not process this xfer.
- set the status as normal.
- unconditionallly callout_stop() and usb_rem_task(). they're safe and
either aren't running, or will run and do nothing.
- finally call usb_transfer_complete().
when aborting a transfer:
- status should be cancelled or timed out.
- if cancelling, callout_halt and usb_rem_task_wait() to make sure the
timer is either done or cancelled.
- at this point, the ux_status must not be cancelled or timed out, and
if it is not in progress we're done.
- set the status.
- if the controller is dying, just return.
- perform HCI-specific tasks to abort this xfer.
- finally call usb_transfer_complete().
for the timeout and timeout task:
- if the HCI is not dying, and the ux_status is in progress, then
trigger the usb abort task.
- remove UXFER_ABORTWAIT and UXFER_ABORTING.

tested on:
- multiple PC systems with several types of devices: ugen/UPS, ucom,
umass with disk, ssd and cdrom backends, kbd, ms, using uhci, ehci
and xhci.
- erlite3: sd@umass on dwc2.
- sunblade2000: kbd/ms and umass disk on ohci.

untested:
- motg, slhci and ahci. motg has some portion of the new scheme
applied, but slhci and ahci require more study.

future work includes pushing a lot of the common abort handling into
usbdi.c and leaving upm_abort() for HC specific tasks, but this change
is pullup-able to netbsd-7 and netbsd-8 as it does not change any
external API, as well as removing over 100 lines of code while adding
over 30 new asserts.

XXX: pullup-7, pullup-8.

branches: 1.44.2;
Stop potential misuse of vendor names and USB vendor IDs in root hub
device and string descriptors.

Firstly: Few vendors have identical PCI-SIG vendor IDs and USB-IF vendor
IDs. As such, using the PCI vendor ID as a USB vendor ID may trample
on whomever is allocated that USB vendor ID.

Secondly: The vendor of the host controller hardware implementation has
little to nothing to do with our usbroothub implementation. Thus we
should not potentially associate any problems therewith to such third
party.

This change will result in root hubs being identified by USB Vendor ID
0x0000. Root hub vendor string will now be "NetBSD" (or, specifically:
ostype). Product ID (0x0000) and product strings remain unchanged.

branches: 1.43.10; 1.43.16;
Merge nick-nhusb

- API / infrastructure changes to support memory management changes.
- Memory management improvements and bug fixes.
- HCDs should now be MP safe
- conversion to KERNHIST based debug
- FS/LS isoc support on ehci(4).
- conversion to kmem(9)
- Some USB 3 support - mostly from Takahiro HAYASHI (t-hash).
- interrupt transfers now get proper DMA operations
- general bug fixes
- kern/48308
- uhub status notification improvements
- umass(4) probe fix (applied to HEAD already)
- ohci(4) short transfer fix

branches: 1.42.12; 1.42.14; 1.42.18;
Use CACHE_LINE_SIZE for EHCI_SQTD_ALIGN to make sure the sqtd fully consume
a cacheline.

use a pool instead of a linked list to avoid synchronization problems.

branches: 1.40.2;
merge the jmcneill-usbmp branch. many thanks to jared for the
initial work, and every one else who has tested things for me.
this is largely my fault at this point :-)

the main changes are something like:

- usbd_bus_methods{} gains a get_lock() to enable the
host controller to provide a lock for the USB code.
if the lock isn't provided, old-style protection is
(partially) applied.

- ehci/ohci/uhci have been converted to the new
interfaces, including mutex/cv/etc conversion.

- usbdivar.h contains a discussion about locking and
what locks are held for which method calls. more
to come for usbdi(9) here.

- audio drivers (uaudio, umidi, auvitek) have been
properly SMPified now that USB is ready.

- scsi drivers have been modified to take the kernel
lock explicitly before calling into scsi code.

- usb pipes are associated with a lock, that is the
same as the controller lock. (this could be split
up further in the future.)

- several usbfoo_locked() or usbfoo_unlocked()
functions have been added to the usbdi(9) to
enable functionality with or without the USB
lock (per controller) already being held.

the TODO.usbmp file has specific details on what is left to
do, including what device-specific changes should be done now
that the whole framework is ready.

pull down from usbmp:

kill the !USE_USE_SOFTINTR code.

branches: 1.38.6; 1.38.10;
Add preliminary support for Embedded Transaction Translator Function (as
found on the MPC8536 and AR9334) which allows low/full devices to be
connected to an EHCI root hub.

branches: 1.37.2;
Add vendor dependent functions.
sc_vendor_init()/sc_vendor_port_status().

branches: 1.36.2;
A pointer typedef entails trading too much flexibility to declare const
and non-const types, and the kernel uses both const and non-const
PMF qualifiers and device suspensors, so change the pmf_qual_t and
device_suspensor_t typedefs from "pointers to const" to non-pointer,
non-const types.

branches: 1.35.2;
Expand PMF_FN_* macros.

Expand <dev/usb/usb_port.h> definitions, and lightly unifdef(1).

branches: 1.33.16;
kern/39727: race condition in ehci isoc abort xfer path

From Jeremy Morse:

* Serialize access to the ehci intrlist.
* Change the ehci intrlist to a tailq so xfers are not queued out of order.
* In ehci_check_itd_intr, don't treat a transfer error as an indication
that the xfer is no longer active.

High speed isochronous support, from Jeremy Morse as part of Google
Summer of Code 2008.

branches: 1.31.2;
Add appropriate bus_dmamap_sync() calls to uhci(4), ohci(4) and ehci(4),
as proposed on tech-kern@.
While all DMA memory used in the USB framework is mapped BUS_DMAMAP_COHERENT
(including data memory, which is memcpy'd from/to USB-private buffers),
the CPU can reorder loads or stores from/to main memory, causing the
controller to have an incoherent view of the DMA descriptors lists for a
short time. bus_dmamap_sync() should contain memory barriers that prevents
the CPU from reordering load/store. Note that BUS_DMAMAP_COHERENT is
still required for the DMA descriptor lists - these can't work properly
with software cache coherency.
This fixes "host controller process error/host controller halted" errors
I'm occasionally seeing with a umodem device on uhci on x86.
Thanks to Michael Lorenz for testing it on his O2, and Izumi Tsutsui on
his Cobalt Qube 2700.

branches: 1.30.2; 1.30.4;
Remove clause 3 and 4 from TNF licenses

branches: 1.29.2; 1.29.4;
split device/softc for USB host controllers and the usb (control)
device,
this is hairy stuff, and I've only tested with uhci/ehci at pci,
please test the rest and report problems

Use device_t and accessors. Setup power management in the PCI
front-end (XXX needs to look more alike the Cardbus front-end).
Establish the shutdown hook using PMF.

Add a method to detach child.

Cosmetic: use device_t and accessors. Use aprint_*_dev(). Use
PMF_FN_{ARGS,CALL,PROTO}.

branches: 1.26.6; 1.26.10;
Merge jmcneill-pm branch.

branches: 1.25.2;
lockmgr -> mutex

branches: 1.24.14; 1.24.20; 1.24.22; 1.24.28; 1.24.30;
Remove the code which disables port status change interrupts for 1s
when one occured -- this makes that events get lost (or delayed until
the next change), in particular in the handover process from the EHCI
to a companion controller, laeving the port dead under some circumstances.
This fixes the "can't reconnect" symptom for me.
I don't see why the interrupt should be disabled - it is acknowledged
in the interrupt handler and shouldn't be active again until the next
status change.

branches: 1.23.18; 1.23.22; 1.23.24;
From OpenBSD:

--
Add a workaround for VIA EHCI controllers which, under load, signal qTD
completion before they have performed writeback from the overlay qTD.

This condition would exhibit itself as a umass stall that never recovers.
--

This fixes the problem reported by Thomas Klausner on current-users@:
http://mail-index.netbsd.org/current-users/2006/01/17/0000.html

branches: 1.22.2;
changes for making DIAGNOSTIC not change the kernel ABI:
- for structure fields that are conditionally present,
make those fields always present.
- for functions which are conditionally inline, make them never inline.
- remove some other functions which are conditionally defined but
don't actually do anything anymore.
- make a lock-debugging function conditional on only LOCKDEBUG.

as discussed on tech-kern some time back.

Keep a flag for each port that keeps track of if the port has been reset
instead of having one flag for the controller.

Also, don't try to modify read-only power bits if the controller doesn't
support power cobntrol.

branches: 1.20.6;
Minor cosmetic change.

branches: 1.19.2;
Allow 32 chars in the saved vendor string. Fixes kern/29760.

Add define for EHCI_MAX_POLLRATE.

branches: 1.17.2; 1.17.8;
Merge from FreeBSD:
sc_child is only used on NetBSD and OpenBSD, ifdef appropriately
FreeBSD revisions:
ehcivar.h 1.8
ohcivar.h 1.39
uhicvar.h 1.39

Implement an USB memory reserve. Allocate some memory per host controller
in the wanted bus_dma space. If an allocation fails during operation
(i.e. when hotplugging an umass device), fall back to using the reserve.

The amount can be configured as USB_MEM_RESERVE. The default value is
256k.

Ideally, there would be a way to steal pages in the desired area from
UVM, but that's far more complicated, and this is not intrusive, plus
it works.

Add interrupt transfers. From OpenBSD.
Having interrupt transfers enables some more device types to be used,
most notably USB2 hubs. USB2 hubs still cannot handle anything but
USB2 devices, because there is node code to handle hub Transaction Translation
yet.
XXX This code doesn't do good bw scheduling, but it's certainly better than
nothing. :)

Insert from ifdefs for soft interrupt related stuff. From OpenBSD.

Add suspend/resume support.
From OpenBSD.

branches: 1.12.16; 1.12.20;
Change abort slightly.
Change debug output.

Add some interrupt processing.

Set up control xfers.
Handle xfer timeouts.
Better debug messages.

Simplify async list handling.

Start of pipe open routine.

Add more fields to hardware structs.
Add memory allocation for various data structures.

Add some more data structure defs.

Recognize port status changes.
Hand over low and full speed devices to companion controller.

Add root hub emulation.

Add some register defines.
Print companion controllers.

Add some useful macros.

branches: 1.1.2; 1.1.4; 1.1.6; 1.1.10;
Add some placeholders for the EHCI (USB 2) driver.

Don't get your hopes up. I've not even finished reading the
(100+) page spec, and I have no hardware.

Three fixes

- pass the 64bit DMA tag if the HCCPARAMS says ehci supports it and the
64bit DMA tag is available/valid. This should help with the
"cannot create xfer" error on 64bit systems.

- restrict the control structure memory allocation to the low 4GB
(CTRLDSSEGMENT was always set to zero anyway)

- if ehci_init fails then tidyup before returning the error.

Use roundup. NFCI.

Consistency in use of #define names. No functional change as values of
before and after defines are the same.

implement a gross hack to fix "boot -a" on systems with usb keyboards on
systems with ehci handover to uhci (and maybe ohci), and fix a similar
problem for "boot -s".

there is effort to ensure that all devices attached via USB are probed
before RB_ASKNAME or RB_SINGLE attempts to ask any questions on the console,
and largely this works, often by chance, today, for USB disks and root.
i've recently pushed this more into uhub and general USB device attachment
as well, and kept a config_pending reference across the first explore of
a bus. these fix many issues with directly attached hubs.

however, on systems where devices connected to ehci ports are handed over
to a companion uhci or ohci port, it may not be the first, or even second,
bus explore that finds the device finally before attachment, and at this
point all config_pending references are dropped.

there is no direct communication between drivers, the potentials are
looked up but their device_t is only used for generic things like the name,
so informing the correct companion to expect a device and deal with the
config_pending references is not possible without some fairly ugly layer
violations or multi-level callbacks (eg, we have "ehci0", and usually an
the relevant companion, eg, "uhci2", but it is the uhub that uhci2 has
attached that will deal with the device attachment.)

with the above fixes to generic USB code, the disown happens during the
first explore. the hack works by, at this point, checking if (a) root
is not mounted, (b) single user or ask name are set, and (c) if the hack
as not been triggered already. if all 3 conditions are true, then a
config_pending_incr() is called and a callback is triggered for (default)
5 seconds to call config_pending_decr(). ehci detach pauses waiting for
this callback if scheduled.

this allows enough time for the uhub and the ukbd/wskbd to attach before
the RK_ASKROOT prompts appear. testing shows it takes between 1.5 and
2 seconds for the keyboard to appear after the disown occurs.


Index: dev/usb/ehcivar.c
- new sc_compcallout, sc_compcallout, sc_complock, and a state for th
handover hack.

Index: dev/usb/ehci.c
ehci_init():
- use aprint_normal_dev() instead of manual device_xname().
- initialise sc_compcallout, sc_compcallout, sc_complock, and sc_comp_state.
ehci_detach():
- if there are companion controllers, tear own the above, including waiting
if there is a callback scheduled.
ehci_disown_callback():
- new callout to call config_pending_decr() in the the future.
schedule this ca
ehci_disown_sched_callback():
- if booting to single user or asking names, call config_pending_incr() and
schedule the callout above, default 5 second delay.
ehci_disown():
- if disowning a port call ehci_disown_sched_callback().

pull across abort fixes from nick-nhusb. add more abort fixes, using
ideas from Taylor and Nick, and myself. special thanks to both who
inspired much of the code here, if not wrote it directly.

among other problems, this assert should no longer trigger:

panic: kernel diagnostic assertion "xfer->ux_state == XFER_ONQU" failed: file "/current/src/sys/dev/usb/usbdi.c", line 914

using usbhist i was able to track down my instance of it being related
to userland close() beginning, dropping the sc_lock, and then the usb
softintr completes the transfer normally, and when it is done, the
abort path attempts to re-complete the transfer, and the above assert
is tripped.


changes from nhusb were commited with these logs:
--
Move the struct usb_task to struct usbd_xfer for everyone to use.
--
Set device transfer status to USBD_IN_PROGRESS if start methods succeeds
--
Actually set the transfer status on transfers in ohci_abort_xfer and
the controller is dying
--
Don't supply the lock to callout_halt when polling as it won't be held
--
Improve transfer abort
--
Mark device transfers as USBD_IN_PROGRESS appropriately and improve
abort handling
--
#ifdef DIAGNOSTIC -> KASSERT and add another KASSERT
--
Mark device transfers as USBD_IN_PROGRESS appropriately and improve
abort handling
--

additional changes include:
- initialise the usb abort task in the HCI allocx routine, so that it
can be safely usb_rem_task()'d.
- rework the handling of softintr vs cancellation vs timeout abort based
upon a scheme from Taylor:
when completing a transfer normally:
- if the status is not in progress, it must be cancelled or timed out,
and we should not process this xfer.
- set the status as normal.
- unconditionallly callout_stop() and usb_rem_task(). they're safe and
either aren't running, or will run and do nothing.
- finally call usb_transfer_complete().
when aborting a transfer:
- status should be cancelled or timed out.
- if cancelling, callout_halt and usb_rem_task_wait() to make sure the
timer is either done or cancelled.
- at this point, the ux_status must not be cancelled or timed out, and
if it is not in progress we're done.
- set the status.
- if the controller is dying, just return.
- perform HCI-specific tasks to abort this xfer.
- finally call usb_transfer_complete().
for the timeout and timeout task:
- if the HCI is not dying, and the ux_status is in progress, then
trigger the usb abort task.
- remove UXFER_ABORTWAIT and UXFER_ABORTING.

tested on:
- multiple PC systems with several types of devices: ugen/UPS, ucom,
umass with disk, ssd and cdrom backends, kbd, ms, using uhci, ehci
and xhci.
- erlite3: sd@umass on dwc2.
- sunblade2000: kbd/ms and umass disk on ohci.

untested:
- motg, slhci and ahci. motg has some portion of the new scheme
applied, but slhci and ahci require more study.

future work includes pushing a lot of the common abort handling into
usbdi.c and leaving upm_abort() for HC specific tasks, but this change
is pullup-able to netbsd-7 and netbsd-8 as it does not change any
external API, as well as removing over 100 lines of code while adding
over 30 new asserts.

XXX: pullup-7, pullup-8.

branches: 1.44.2;
Stop potential misuse of vendor names and USB vendor IDs in root hub
device and string descriptors.

Firstly: Few vendors have identical PCI-SIG vendor IDs and USB-IF vendor
IDs. As such, using the PCI vendor ID as a USB vendor ID may trample
on whomever is allocated that USB vendor ID.

Secondly: The vendor of the host controller hardware implementation has
little to nothing to do with our usbroothub implementation. Thus we
should not potentially associate any problems therewith to such third
party.

This change will result in root hubs being identified by USB Vendor ID
0x0000. Root hub vendor string will now be "NetBSD" (or, specifically:
ostype). Product ID (0x0000) and product strings remain unchanged.

branches: 1.43.10; 1.43.16;
Merge nick-nhusb

- API / infrastructure changes to support memory management changes.
- Memory management improvements and bug fixes.
- HCDs should now be MP safe
- conversion to KERNHIST based debug
- FS/LS isoc support on ehci(4).
- conversion to kmem(9)
- Some USB 3 support - mostly from Takahiro HAYASHI (t-hash).
- interrupt transfers now get proper DMA operations
- general bug fixes
- kern/48308
- uhub status notification improvements
- umass(4) probe fix (applied to HEAD already)
- ohci(4) short transfer fix

branches: 1.42.12; 1.42.14; 1.42.18;
Use CACHE_LINE_SIZE for EHCI_SQTD_ALIGN to make sure the sqtd fully consume
a cacheline.

use a pool instead of a linked list to avoid synchronization problems.

branches: 1.40.2;
merge the jmcneill-usbmp branch. many thanks to jared for the
initial work, and every one else who has tested things for me.
this is largely my fault at this point :-)

the main changes are something like:

- usbd_bus_methods{} gains a get_lock() to enable the
host controller to provide a lock for the USB code.
if the lock isn't provided, old-style protection is
(partially) applied.

- ehci/ohci/uhci have been converted to the new
interfaces, including mutex/cv/etc conversion.

- usbdivar.h contains a discussion about locking and
what locks are held for which method calls. more
to come for usbdi(9) here.

- audio drivers (uaudio, umidi, auvitek) have been
properly SMPified now that USB is ready.

- scsi drivers have been modified to take the kernel
lock explicitly before calling into scsi code.

- usb pipes are associated with a lock, that is the
same as the controller lock. (this could be split
up further in the future.)

- several usbfoo_locked() or usbfoo_unlocked()
functions have been added to the usbdi(9) to
enable functionality with or without the USB
lock (per controller) already being held.

the TODO.usbmp file has specific details on what is left to
do, including what device-specific changes should be done now
that the whole framework is ready.

pull down from usbmp:

kill the !USE_USE_SOFTINTR code.

branches: 1.38.6; 1.38.10;
Add preliminary support for Embedded Transaction Translator Function (as
found on the MPC8536 and AR9334) which allows low/full devices to be
connected to an EHCI root hub.

branches: 1.37.2;
Add vendor dependent functions.
sc_vendor_init()/sc_vendor_port_status().

branches: 1.36.2;
A pointer typedef entails trading too much flexibility to declare const
and non-const types, and the kernel uses both const and non-const
PMF qualifiers and device suspensors, so change the pmf_qual_t and
device_suspensor_t typedefs from "pointers to const" to non-pointer,
non-const types.

branches: 1.35.2;
Expand PMF_FN_* macros.

Expand <dev/usb/usb_port.h> definitions, and lightly unifdef(1).

branches: 1.33.16;
kern/39727: race condition in ehci isoc abort xfer path

From Jeremy Morse:

* Serialize access to the ehci intrlist.
* Change the ehci intrlist to a tailq so xfers are not queued out of order.
* In ehci_check_itd_intr, don't treat a transfer error as an indication
that the xfer is no longer active.

High speed isochronous support, from Jeremy Morse as part of Google
Summer of Code 2008.

branches: 1.31.2;
Add appropriate bus_dmamap_sync() calls to uhci(4), ohci(4) and ehci(4),
as proposed on tech-kern@.
While all DMA memory used in the USB framework is mapped BUS_DMAMAP_COHERENT
(including data memory, which is memcpy'd from/to USB-private buffers),
the CPU can reorder loads or stores from/to main memory, causing the
controller to have an incoherent view of the DMA descriptors lists for a
short time. bus_dmamap_sync() should contain memory barriers that prevents
the CPU from reordering load/store. Note that BUS_DMAMAP_COHERENT is
still required for the DMA descriptor lists - these can't work properly
with software cache coherency.
This fixes "host controller process error/host controller halted" errors
I'm occasionally seeing with a umodem device on uhci on x86.
Thanks to Michael Lorenz for testing it on his O2, and Izumi Tsutsui on
his Cobalt Qube 2700.

branches: 1.30.2; 1.30.4;
Remove clause 3 and 4 from TNF licenses

branches: 1.29.2; 1.29.4;
split device/softc for USB host controllers and the usb (control)
device,
this is hairy stuff, and I've only tested with uhci/ehci at pci,
please test the rest and report problems

Use device_t and accessors. Setup power management in the PCI
front-end (XXX needs to look more alike the Cardbus front-end).
Establish the shutdown hook using PMF.

Add a method to detach child.

Cosmetic: use device_t and accessors. Use aprint_*_dev(). Use
PMF_FN_{ARGS,CALL,PROTO}.

branches: 1.26.6; 1.26.10;
Merge jmcneill-pm branch.

branches: 1.25.2;
lockmgr -> mutex

branches: 1.24.14; 1.24.20; 1.24.22; 1.24.28; 1.24.30;
Remove the code which disables port status change interrupts for 1s
when one occured -- this makes that events get lost (or delayed until
the next change), in particular in the handover process from the EHCI
to a companion controller, laeving the port dead under some circumstances.
This fixes the "can't reconnect" symptom for me.
I don't see why the interrupt should be disabled - it is acknowledged
in the interrupt handler and shouldn't be active again until the next
status change.

branches: 1.23.18; 1.23.22; 1.23.24;
From OpenBSD:

--
Add a workaround for VIA EHCI controllers which, under load, signal qTD
completion before they have performed writeback from the overlay qTD.

This condition would exhibit itself as a umass stall that never recovers.
--

This fixes the problem reported by Thomas Klausner on current-users@:
http://mail-index.netbsd.org/current-users/2006/01/17/0000.html

branches: 1.22.2;
changes for making DIAGNOSTIC not change the kernel ABI:
- for structure fields that are conditionally present,
make those fields always present.
- for functions which are conditionally inline, make them never inline.
- remove some other functions which are conditionally defined but
don't actually do anything anymore.
- make a lock-debugging function conditional on only LOCKDEBUG.

as discussed on tech-kern some time back.

Keep a flag for each port that keeps track of if the port has been reset
instead of having one flag for the controller.

Also, don't try to modify read-only power bits if the controller doesn't
support power cobntrol.

branches: 1.20.6;
Minor cosmetic change.

branches: 1.19.2;
Allow 32 chars in the saved vendor string. Fixes kern/29760.

Add define for EHCI_MAX_POLLRATE.

branches: 1.17.2; 1.17.8;
Merge from FreeBSD:
sc_child is only used on NetBSD and OpenBSD, ifdef appropriately
FreeBSD revisions:
ehcivar.h 1.8
ohcivar.h 1.39
uhicvar.h 1.39

Implement an USB memory reserve. Allocate some memory per host controller
in the wanted bus_dma space. If an allocation fails during operation
(i.e. when hotplugging an umass device), fall back to using the reserve.

The amount can be configured as USB_MEM_RESERVE. The default value is
256k.

Ideally, there would be a way to steal pages in the desired area from
UVM, but that's far more complicated, and this is not intrusive, plus
it works.

Add interrupt transfers. From OpenBSD.
Having interrupt transfers enables some more device types to be used,
most notably USB2 hubs. USB2 hubs still cannot handle anything but
USB2 devices, because there is node code to handle hub Transaction Translation
yet.
XXX This code doesn't do good bw scheduling, but it's certainly better than
nothing. :)

Insert from ifdefs for soft interrupt related stuff. From OpenBSD.

Add suspend/resume support.
From OpenBSD.

branches: 1.12.16; 1.12.20;
Change abort slightly.
Change debug output.

Add some interrupt processing.

Set up control xfers.
Handle xfer timeouts.
Better debug messages.

Simplify async list handling.

Start of pipe open routine.

Add more fields to hardware structs.
Add memory allocation for various data structures.

Add some more data structure defs.

Recognize port status changes.
Hand over low and full speed devices to companion controller.

Add root hub emulation.

Add some register defines.
Print companion controllers.

Add some useful macros.

branches: 1.1.2; 1.1.4; 1.1.6; 1.1.10;
Add some placeholders for the EHCI (USB 2) driver.

Don't get your hopes up. I've not even finished reading the
(100+) page spec, and I have no hardware.

Use roundup. NFCI.

Consistency in use of #define names. No functional change as values of
before and after defines are the same.

implement a gross hack to fix "boot -a" on systems with usb keyboards on
systems with ehci handover to uhci (and maybe ohci), and fix a similar
problem for "boot -s".

there is effort to ensure that all devices attached via USB are probed
before RB_ASKNAME or RB_SINGLE attempts to ask any questions on the console,
and largely this works, often by chance, today, for USB disks and root.
i've recently pushed this more into uhub and general USB device attachment
as well, and kept a config_pending reference across the first explore of
a bus. these fix many issues with directly attached hubs.

however, on systems where devices connected to ehci ports are handed over
to a companion uhci or ohci port, it may not be the first, or even second,
bus explore that finds the device finally before attachment, and at this
point all config_pending references are dropped.

there is no direct communication between drivers, the potentials are
looked up but their device_t is only used for generic things like the name,
so informing the correct companion to expect a device and deal with the
config_pending references is not possible without some fairly ugly layer
violations or multi-level callbacks (eg, we have "ehci0", and usually an
the relevant companion, eg, "uhci2", but it is the uhub that uhci2 has
attached that will deal with the device attachment.)

with the above fixes to generic USB code, the disown happens during the
first explore. the hack works by, at this point, checking if (a) root
is not mounted, (b) single user or ask name are set, and (c) if the hack
as not been triggered already. if all 3 conditions are true, then a
config_pending_incr() is called and a callback is triggered for (default)
5 seconds to call config_pending_decr(). ehci detach pauses waiting for
this callback if scheduled.

this allows enough time for the uhub and the ukbd/wskbd to attach before
the RK_ASKROOT prompts appear. testing shows it takes between 1.5 and
2 seconds for the keyboard to appear after the disown occurs.


Index: dev/usb/ehcivar.c
- new sc_compcallout, sc_compcallout, sc_complock, and a state for th
handover hack.

Index: dev/usb/ehci.c
ehci_init():
- use aprint_normal_dev() instead of manual device_xname().
- initialise sc_compcallout, sc_compcallout, sc_complock, and sc_comp_state.
ehci_detach():
- if there are companion controllers, tear own the above, including waiting
if there is a callback scheduled.
ehci_disown_callback():
- new callout to call config_pending_decr() in the the future.
schedule this ca
ehci_disown_sched_callback():
- if booting to single user or asking names, call config_pending_incr() and
schedule the callout above, default 5 second delay.
ehci_disown():
- if disowning a port call ehci_disown_sched_callback().

pull across abort fixes from nick-nhusb. add more abort fixes, using
ideas from Taylor and Nick, and myself. special thanks to both who
inspired much of the code here, if not wrote it directly.

among other problems, this assert should no longer trigger:

panic: kernel diagnostic assertion "xfer->ux_state == XFER_ONQU" failed: file "/current/src/sys/dev/usb/usbdi.c", line 914

using usbhist i was able to track down my instance of it being related
to userland close() beginning, dropping the sc_lock, and then the usb
softintr completes the transfer normally, and when it is done, the
abort path attempts to re-complete the transfer, and the above assert
is tripped.


changes from nhusb were commited with these logs:
--
Move the struct usb_task to struct usbd_xfer for everyone to use.
--
Set device transfer status to USBD_IN_PROGRESS if start methods succeeds
--
Actually set the transfer status on transfers in ohci_abort_xfer and
the controller is dying
--
Don't supply the lock to callout_halt when polling as it won't be held
--
Improve transfer abort
--
Mark device transfers as USBD_IN_PROGRESS appropriately and improve
abort handling
--
#ifdef DIAGNOSTIC -> KASSERT and add another KASSERT
--
Mark device transfers as USBD_IN_PROGRESS appropriately and improve
abort handling
--

additional changes include:
- initialise the usb abort task in the HCI allocx routine, so that it
can be safely usb_rem_task()'d.
- rework the handling of softintr vs cancellation vs timeout abort based
upon a scheme from Taylor:
when completing a transfer normally:
- if the status is not in progress, it must be cancelled or timed out,
and we should not process this xfer.
- set the status as normal.
- unconditionallly callout_stop() and usb_rem_task(). they're safe and
either aren't running, or will run and do nothing.
- finally call usb_transfer_complete().
when aborting a transfer:
- status should be cancelled or timed out.
- if cancelling, callout_halt and usb_rem_task_wait() to make sure the
timer is either done or cancelled.
- at this point, the ux_status must not be cancelled or timed out, and
if it is not in progress we're done.
- set the status.
- if the controller is dying, just return.
- perform HCI-specific tasks to abort this xfer.
- finally call usb_transfer_complete().
for the timeout and timeout task:
- if the HCI is not dying, and the ux_status is in progress, then
trigger the usb abort task.
- remove UXFER_ABORTWAIT and UXFER_ABORTING.

tested on:
- multiple PC systems with several types of devices: ugen/UPS, ucom,
umass with disk, ssd and cdrom backends, kbd, ms, using uhci, ehci
and xhci.
- erlite3: sd@umass on dwc2.
- sunblade2000: kbd/ms and umass disk on ohci.

untested:
- motg, slhci and ahci. motg has some portion of the new scheme
applied, but slhci and ahci require more study.

future work includes pushing a lot of the common abort handling into
usbdi.c and leaving upm_abort() for HC specific tasks, but this change
is pullup-able to netbsd-7 and netbsd-8 as it does not change any
external API, as well as removing over 100 lines of code while adding
over 30 new asserts.

XXX: pullup-7, pullup-8.

branches: 1.44.2;
Stop potential misuse of vendor names and USB vendor IDs in root hub
device and string descriptors.

Firstly: Few vendors have identical PCI-SIG vendor IDs and USB-IF vendor
IDs. As such, using the PCI vendor ID as a USB vendor ID may trample
on whomever is allocated that USB vendor ID.

Secondly: The vendor of the host controller hardware implementation has
little to nothing to do with our usbroothub implementation. Thus we
should not potentially associate any problems therewith to such third
party.

This change will result in root hubs being identified by USB Vendor ID
0x0000. Root hub vendor string will now be "NetBSD" (or, specifically:
ostype). Product ID (0x0000) and product strings remain unchanged.

branches: 1.43.10; 1.43.16;
Merge nick-nhusb

- API / infrastructure changes to support memory management changes.
- Memory management improvements and bug fixes.
- HCDs should now be MP safe
- conversion to KERNHIST based debug
- FS/LS isoc support on ehci(4).
- conversion to kmem(9)
- Some USB 3 support - mostly from Takahiro HAYASHI (t-hash).
- interrupt transfers now get proper DMA operations
- general bug fixes
- kern/48308
- uhub status notification improvements
- umass(4) probe fix (applied to HEAD already)
- ohci(4) short transfer fix

branches: 1.42.12; 1.42.14; 1.42.18;
Use CACHE_LINE_SIZE for EHCI_SQTD_ALIGN to make sure the sqtd fully consume
a cacheline.

use a pool instead of a linked list to avoid synchronization problems.

branches: 1.40.2;
merge the jmcneill-usbmp branch. many thanks to jared for the
initial work, and every one else who has tested things for me.
this is largely my fault at this point :-)

the main changes are something like:

- usbd_bus_methods{} gains a get_lock() to enable the
host controller to provide a lock for the USB code.
if the lock isn't provided, old-style protection is
(partially) applied.

- ehci/ohci/uhci have been converted to the new
interfaces, including mutex/cv/etc conversion.

- usbdivar.h contains a discussion about locking and
what locks are held for which method calls. more
to come for usbdi(9) here.

- audio drivers (uaudio, umidi, auvitek) have been
properly SMPified now that USB is ready.

- scsi drivers have been modified to take the kernel
lock explicitly before calling into scsi code.

- usb pipes are associated with a lock, that is the
same as the controller lock. (this could be split
up further in the future.)

- several usbfoo_locked() or usbfoo_unlocked()
functions have been added to the usbdi(9) to
enable functionality with or without the USB
lock (per controller) already being held.

the TODO.usbmp file has specific details on what is left to
do, including what device-specific changes should be done now
that the whole framework is ready.

pull down from usbmp:

kill the !USE_USE_SOFTINTR code.

branches: 1.38.6; 1.38.10;
Add preliminary support for Embedded Transaction Translator Function (as
found on the MPC8536 and AR9334) which allows low/full devices to be
connected to an EHCI root hub.

branches: 1.37.2;
Add vendor dependent functions.
sc_vendor_init()/sc_vendor_port_status().

branches: 1.36.2;
A pointer typedef entails trading too much flexibility to declare const
and non-const types, and the kernel uses both const and non-const
PMF qualifiers and device suspensors, so change the pmf_qual_t and
device_suspensor_t typedefs from "pointers to const" to non-pointer,
non-const types.

branches: 1.35.2;
Expand PMF_FN_* macros.

Expand <dev/usb/usb_port.h> definitions, and lightly unifdef(1).

branches: 1.33.16;
kern/39727: race condition in ehci isoc abort xfer path

From Jeremy Morse:

* Serialize access to the ehci intrlist.
* Change the ehci intrlist to a tailq so xfers are not queued out of order.
* In ehci_check_itd_intr, don't treat a transfer error as an indication
that the xfer is no longer active.

High speed isochronous support, from Jeremy Morse as part of Google
Summer of Code 2008.

branches: 1.31.2;
Add appropriate bus_dmamap_sync() calls to uhci(4), ohci(4) and ehci(4),
as proposed on tech-kern@.
While all DMA memory used in the USB framework is mapped BUS_DMAMAP_COHERENT
(including data memory, which is memcpy'd from/to USB-private buffers),
the CPU can reorder loads or stores from/to main memory, causing the
controller to have an incoherent view of the DMA descriptors lists for a
short time. bus_dmamap_sync() should contain memory barriers that prevents
the CPU from reordering load/store. Note that BUS_DMAMAP_COHERENT is
still required for the DMA descriptor lists - these can't work properly
with software cache coherency.
This fixes "host controller process error/host controller halted" errors
I'm occasionally seeing with a umodem device on uhci on x86.
Thanks to Michael Lorenz for testing it on his O2, and Izumi Tsutsui on
his Cobalt Qube 2700.

branches: 1.30.2; 1.30.4;
Remove clause 3 and 4 from TNF licenses

branches: 1.29.2; 1.29.4;
split device/softc for USB host controllers and the usb (control)
device,
this is hairy stuff, and I've only tested with uhci/ehci at pci,
please test the rest and report problems

Use device_t and accessors. Setup power management in the PCI
front-end (XXX needs to look more alike the Cardbus front-end).
Establish the shutdown hook using PMF.

Add a method to detach child.

Cosmetic: use device_t and accessors. Use aprint_*_dev(). Use
PMF_FN_{ARGS,CALL,PROTO}.

branches: 1.26.6; 1.26.10;
Merge jmcneill-pm branch.

branches: 1.25.2;
lockmgr -> mutex

branches: 1.24.14; 1.24.20; 1.24.22; 1.24.28; 1.24.30;
Remove the code which disables port status change interrupts for 1s
when one occured -- this makes that events get lost (or delayed until
the next change), in particular in the handover process from the EHCI
to a companion controller, laeving the port dead under some circumstances.
This fixes the "can't reconnect" symptom for me.
I don't see why the interrupt should be disabled - it is acknowledged
in the interrupt handler and shouldn't be active again until the next
status change.

branches: 1.23.18; 1.23.22; 1.23.24;
From OpenBSD:

--
Add a workaround for VIA EHCI controllers which, under load, signal qTD
completion before they have performed writeback from the overlay qTD.

This condition would exhibit itself as a umass stall that never recovers.
--

This fixes the problem reported by Thomas Klausner on current-users@:
http://mail-index.netbsd.org/current-users/2006/01/17/0000.html

branches: 1.22.2;
changes for making DIAGNOSTIC not change the kernel ABI:
- for structure fields that are conditionally present,
make those fields always present.
- for functions which are conditionally inline, make them never inline.
- remove some other functions which are conditionally defined but
don't actually do anything anymore.
- make a lock-debugging function conditional on only LOCKDEBUG.

as discussed on tech-kern some time back.

Keep a flag for each port that keeps track of if the port has been reset
instead of having one flag for the controller.

Also, don't try to modify read-only power bits if the controller doesn't
support power cobntrol.

branches: 1.20.6;
Minor cosmetic change.

branches: 1.19.2;
Allow 32 chars in the saved vendor string. Fixes kern/29760.

Add define for EHCI_MAX_POLLRATE.

branches: 1.17.2; 1.17.8;
Merge from FreeBSD:
sc_child is only used on NetBSD and OpenBSD, ifdef appropriately
FreeBSD revisions:
ehcivar.h 1.8
ohcivar.h 1.39
uhicvar.h 1.39

Implement an USB memory reserve. Allocate some memory per host controller
in the wanted bus_dma space. If an allocation fails during operation
(i.e. when hotplugging an umass device), fall back to using the reserve.

The amount can be configured as USB_MEM_RESERVE. The default value is
256k.

Ideally, there would be a way to steal pages in the desired area from
UVM, but that's far more complicated, and this is not intrusive, plus
it works.

Add interrupt transfers. From OpenBSD.
Having interrupt transfers enables some more device types to be used,
most notably USB2 hubs. USB2 hubs still cannot handle anything but
USB2 devices, because there is node code to handle hub Transaction Translation
yet.
XXX This code doesn't do good bw scheduling, but it's certainly better than
nothing. :)

Insert from ifdefs for soft interrupt related stuff. From OpenBSD.

Add suspend/resume support.
From OpenBSD.

branches: 1.12.16; 1.12.20;
Change abort slightly.
Change debug output.

Add some interrupt processing.

Set up control xfers.
Handle xfer timeouts.
Better debug messages.

Simplify async list handling.

Start of pipe open routine.

Add more fields to hardware structs.
Add memory allocation for various data structures.

Add some more data structure defs.

Recognize port status changes.
Hand over low and full speed devices to companion controller.

Add root hub emulation.

Add some register defines.
Print companion controllers.

Add some useful macros.

branches: 1.1.2; 1.1.4; 1.1.6; 1.1.10;
Add some placeholders for the EHCI (USB 2) driver.

Don't get your hopes up. I've not even finished reading the
(100+) page spec, and I have no hardware.

Consistency in use of #define names. No functional change as values of
before and after defines are the same.

implement a gross hack to fix "boot -a" on systems with usb keyboards on
systems with ehci handover to uhci (and maybe ohci), and fix a similar
problem for "boot -s".

there is effort to ensure that all devices attached via USB are probed
before RB_ASKNAME or RB_SINGLE attempts to ask any questions on the console,
and largely this works, often by chance, today, for USB disks and root.
i've recently pushed this more into uhub and general USB device attachment
as well, and kept a config_pending reference across the first explore of
a bus. these fix many issues with directly attached hubs.

however, on systems where devices connected to ehci ports are handed over
to a companion uhci or ohci port, it may not be the first, or even second,
bus explore that finds the device finally before attachment, and at this
point all config_pending references are dropped.

there is no direct communication between drivers, the potentials are
looked up but their device_t is only used for generic things like the name,
so informing the correct companion to expect a device and deal with the
config_pending references is not possible without some fairly ugly layer
violations or multi-level callbacks (eg, we have "ehci0", and usually an
the relevant companion, eg, "uhci2", but it is the uhub that uhci2 has
attached that will deal with the device attachment.)

with the above fixes to generic USB code, the disown happens during the
first explore. the hack works by, at this point, checking if (a) root
is not mounted, (b) single user or ask name are set, and (c) if the hack
as not been triggered already. if all 3 conditions are true, then a
config_pending_incr() is called and a callback is triggered for (default)
5 seconds to call config_pending_decr(). ehci detach pauses waiting for
this callback if scheduled.

this allows enough time for the uhub and the ukbd/wskbd to attach before
the RK_ASKROOT prompts appear. testing shows it takes between 1.5 and
2 seconds for the keyboard to appear after the disown occurs.


Index: dev/usb/ehcivar.c
- new sc_compcallout, sc_compcallout, sc_complock, and a state for th
handover hack.

Index: dev/usb/ehci.c
ehci_init():
- use aprint_normal_dev() instead of manual device_xname().
- initialise sc_compcallout, sc_compcallout, sc_complock, and sc_comp_state.
ehci_detach():
- if there are companion controllers, tear own the above, including waiting
if there is a callback scheduled.
ehci_disown_callback():
- new callout to call config_pending_decr() in the the future.
schedule this ca
ehci_disown_sched_callback():
- if booting to single user or asking names, call config_pending_incr() and
schedule the callout above, default 5 second delay.
ehci_disown():
- if disowning a port call ehci_disown_sched_callback().

pull across abort fixes from nick-nhusb. add more abort fixes, using
ideas from Taylor and Nick, and myself. special thanks to both who
inspired much of the code here, if not wrote it directly.

among other problems, this assert should no longer trigger:

panic: kernel diagnostic assertion "xfer->ux_state == XFER_ONQU" failed: file "/current/src/sys/dev/usb/usbdi.c", line 914

using usbhist i was able to track down my instance of it being related
to userland close() beginning, dropping the sc_lock, and then the usb
softintr completes the transfer normally, and when it is done, the
abort path attempts to re-complete the transfer, and the above assert
is tripped.


changes from nhusb were commited with these logs:
--
Move the struct usb_task to struct usbd_xfer for everyone to use.
--
Set device transfer status to USBD_IN_PROGRESS if start methods succeeds
--
Actually set the transfer status on transfers in ohci_abort_xfer and
the controller is dying
--
Don't supply the lock to callout_halt when polling as it won't be held
--
Improve transfer abort
--
Mark device transfers as USBD_IN_PROGRESS appropriately and improve
abort handling
--
#ifdef DIAGNOSTIC -> KASSERT and add another KASSERT
--
Mark device transfers as USBD_IN_PROGRESS appropriately and improve
abort handling
--

additional changes include:
- initialise the usb abort task in the HCI allocx routine, so that it
can be safely usb_rem_task()'d.
- rework the handling of softintr vs cancellation vs timeout abort based
upon a scheme from Taylor:
when completing a transfer normally:
- if the status is not in progress, it must be cancelled or timed out,
and we should not process this xfer.
- set the status as normal.
- unconditionallly callout_stop() and usb_rem_task(). they're safe and
either aren't running, or will run and do nothing.
- finally call usb_transfer_complete().
when aborting a transfer:
- status should be cancelled or timed out.
- if cancelling, callout_halt and usb_rem_task_wait() to make sure the
timer is either done or cancelled.
- at this point, the ux_status must not be cancelled or timed out, and
if it is not in progress we're done.
- set the status.
- if the controller is dying, just return.
- perform HCI-specific tasks to abort this xfer.
- finally call usb_transfer_complete().
for the timeout and timeout task:
- if the HCI is not dying, and the ux_status is in progress, then
trigger the usb abort task.
- remove UXFER_ABORTWAIT and UXFER_ABORTING.

tested on:
- multiple PC systems with several types of devices: ugen/UPS, ucom,
umass with disk, ssd and cdrom backends, kbd, ms, using uhci, ehci
and xhci.
- erlite3: sd@umass on dwc2.
- sunblade2000: kbd/ms and umass disk on ohci.

untested:
- motg, slhci and ahci. motg has some portion of the new scheme
applied, but slhci and ahci require more study.

future work includes pushing a lot of the common abort handling into
usbdi.c and leaving upm_abort() for HC specific tasks, but this change
is pullup-able to netbsd-7 and netbsd-8 as it does not change any
external API, as well as removing over 100 lines of code while adding
over 30 new asserts.

XXX: pullup-7, pullup-8.

branches: 1.44.2;
Stop potential misuse of vendor names and USB vendor IDs in root hub
device and string descriptors.

Firstly: Few vendors have identical PCI-SIG vendor IDs and USB-IF vendor
IDs. As such, using the PCI vendor ID as a USB vendor ID may trample
on whomever is allocated that USB vendor ID.

Secondly: The vendor of the host controller hardware implementation has
little to nothing to do with our usbroothub implementation. Thus we
should not potentially associate any problems therewith to such third
party.

This change will result in root hubs being identified by USB Vendor ID
0x0000. Root hub vendor string will now be "NetBSD" (or, specifically:
ostype). Product ID (0x0000) and product strings remain unchanged.

branches: 1.43.10; 1.43.16;
Merge nick-nhusb

- API / infrastructure changes to support memory management changes.
- Memory management improvements and bug fixes.
- HCDs should now be MP safe
- conversion to KERNHIST based debug
- FS/LS isoc support on ehci(4).
- conversion to kmem(9)
- Some USB 3 support - mostly from Takahiro HAYASHI (t-hash).
- interrupt transfers now get proper DMA operations
- general bug fixes
- kern/48308
- uhub status notification improvements
- umass(4) probe fix (applied to HEAD already)
- ohci(4) short transfer fix

branches: 1.42.12; 1.42.14; 1.42.18;
Use CACHE_LINE_SIZE for EHCI_SQTD_ALIGN to make sure the sqtd fully consume
a cacheline.

use a pool instead of a linked list to avoid synchronization problems.

branches: 1.40.2;
merge the jmcneill-usbmp branch. many thanks to jared for the
initial work, and every one else who has tested things for me.
this is largely my fault at this point :-)

the main changes are something like:

- usbd_bus_methods{} gains a get_lock() to enable the
host controller to provide a lock for the USB code.
if the lock isn't provided, old-style protection is
(partially) applied.

- ehci/ohci/uhci have been converted to the new
interfaces, including mutex/cv/etc conversion.

- usbdivar.h contains a discussion about locking and
what locks are held for which method calls. more
to come for usbdi(9) here.

- audio drivers (uaudio, umidi, auvitek) have been
properly SMPified now that USB is ready.

- scsi drivers have been modified to take the kernel
lock explicitly before calling into scsi code.

- usb pipes are associated with a lock, that is the
same as the controller lock. (this could be split
up further in the future.)

- several usbfoo_locked() or usbfoo_unlocked()
functions have been added to the usbdi(9) to
enable functionality with or without the USB
lock (per controller) already being held.

the TODO.usbmp file has specific details on what is left to
do, including what device-specific changes should be done now
that the whole framework is ready.

pull down from usbmp:

kill the !USE_USE_SOFTINTR code.

branches: 1.38.6; 1.38.10;
Add preliminary support for Embedded Transaction Translator Function (as
found on the MPC8536 and AR9334) which allows low/full devices to be
connected to an EHCI root hub.

branches: 1.37.2;
Add vendor dependent functions.
sc_vendor_init()/sc_vendor_port_status().

branches: 1.36.2;
A pointer typedef entails trading too much flexibility to declare const
and non-const types, and the kernel uses both const and non-const
PMF qualifiers and device suspensors, so change the pmf_qual_t and
device_suspensor_t typedefs from "pointers to const" to non-pointer,
non-const types.

branches: 1.35.2;
Expand PMF_FN_* macros.

Expand <dev/usb/usb_port.h> definitions, and lightly unifdef(1).

branches: 1.33.16;
kern/39727: race condition in ehci isoc abort xfer path

From Jeremy Morse:

* Serialize access to the ehci intrlist.
* Change the ehci intrlist to a tailq so xfers are not queued out of order.
* In ehci_check_itd_intr, don't treat a transfer error as an indication
that the xfer is no longer active.

High speed isochronous support, from Jeremy Morse as part of Google
Summer of Code 2008.

branches: 1.31.2;
Add appropriate bus_dmamap_sync() calls to uhci(4), ohci(4) and ehci(4),
as proposed on tech-kern@.
While all DMA memory used in the USB framework is mapped BUS_DMAMAP_COHERENT
(including data memory, which is memcpy'd from/to USB-private buffers),
the CPU can reorder loads or stores from/to main memory, causing the
controller to have an incoherent view of the DMA descriptors lists for a
short time. bus_dmamap_sync() should contain memory barriers that prevents
the CPU from reordering load/store. Note that BUS_DMAMAP_COHERENT is
still required for the DMA descriptor lists - these can't work properly
with software cache coherency.
This fixes "host controller process error/host controller halted" errors
I'm occasionally seeing with a umodem device on uhci on x86.
Thanks to Michael Lorenz for testing it on his O2, and Izumi Tsutsui on
his Cobalt Qube 2700.

branches: 1.30.2; 1.30.4;
Remove clause 3 and 4 from TNF licenses

branches: 1.29.2; 1.29.4;
split device/softc for USB host controllers and the usb (control)
device,
this is hairy stuff, and I've only tested with uhci/ehci at pci,
please test the rest and report problems

Use device_t and accessors. Setup power management in the PCI
front-end (XXX needs to look more alike the Cardbus front-end).
Establish the shutdown hook using PMF.

Add a method to detach child.

Cosmetic: use device_t and accessors. Use aprint_*_dev(). Use
PMF_FN_{ARGS,CALL,PROTO}.

branches: 1.26.6; 1.26.10;
Merge jmcneill-pm branch.

branches: 1.25.2;
lockmgr -> mutex

branches: 1.24.14; 1.24.20; 1.24.22; 1.24.28; 1.24.30;
Remove the code which disables port status change interrupts for 1s
when one occured -- this makes that events get lost (or delayed until
the next change), in particular in the handover process from the EHCI
to a companion controller, laeving the port dead under some circumstances.
This fixes the "can't reconnect" symptom for me.
I don't see why the interrupt should be disabled - it is acknowledged
in the interrupt handler and shouldn't be active again until the next
status change.

branches: 1.23.18; 1.23.22; 1.23.24;
From OpenBSD:

--
Add a workaround for VIA EHCI controllers which, under load, signal qTD
completion before they have performed writeback from the overlay qTD.

This condition would exhibit itself as a umass stall that never recovers.
--

This fixes the problem reported by Thomas Klausner on current-users@:
http://mail-index.netbsd.org/current-users/2006/01/17/0000.html

branches: 1.22.2;
changes for making DIAGNOSTIC not change the kernel ABI:
- for structure fields that are conditionally present,
make those fields always present.
- for functions which are conditionally inline, make them never inline.
- remove some other functions which are conditionally defined but
don't actually do anything anymore.
- make a lock-debugging function conditional on only LOCKDEBUG.

as discussed on tech-kern some time back.

Keep a flag for each port that keeps track of if the port has been reset
instead of having one flag for the controller.

Also, don't try to modify read-only power bits if the controller doesn't
support power cobntrol.

branches: 1.20.6;
Minor cosmetic change.

branches: 1.19.2;
Allow 32 chars in the saved vendor string. Fixes kern/29760.

Add define for EHCI_MAX_POLLRATE.

branches: 1.17.2; 1.17.8;
Merge from FreeBSD:
sc_child is only used on NetBSD and OpenBSD, ifdef appropriately
FreeBSD revisions:
ehcivar.h 1.8
ohcivar.h 1.39
uhicvar.h 1.39

Implement an USB memory reserve. Allocate some memory per host controller
in the wanted bus_dma space. If an allocation fails during operation
(i.e. when hotplugging an umass device), fall back to using the reserve.

The amount can be configured as USB_MEM_RESERVE. The default value is
256k.

Ideally, there would be a way to steal pages in the desired area from
UVM, but that's far more complicated, and this is not intrusive, plus
it works.

Add interrupt transfers. From OpenBSD.
Having interrupt transfers enables some more device types to be used,
most notably USB2 hubs. USB2 hubs still cannot handle anything but
USB2 devices, because there is node code to handle hub Transaction Translation
yet.
XXX This code doesn't do good bw scheduling, but it's certainly better than
nothing. :)

Insert from ifdefs for soft interrupt related stuff. From OpenBSD.

Add suspend/resume support.
From OpenBSD.

branches: 1.12.16; 1.12.20;
Change abort slightly.
Change debug output.

Add some interrupt processing.

Set up control xfers.
Handle xfer timeouts.
Better debug messages.

Simplify async list handling.

Start of pipe open routine.

Add more fields to hardware structs.
Add memory allocation for various data structures.

Add some more data structure defs.

Recognize port status changes.
Hand over low and full speed devices to companion controller.

Add root hub emulation.

Add some register defines.
Print companion controllers.

Add some useful macros.

branches: 1.1.2; 1.1.4; 1.1.6; 1.1.10;
Add some placeholders for the EHCI (USB 2) driver.

Don't get your hopes up. I've not even finished reading the
(100+) page spec, and I have no hardware.

implement a gross hack to fix "boot -a" on systems with usb keyboards on
systems with ehci handover to uhci (and maybe ohci), and fix a similar
problem for "boot -s".

there is effort to ensure that all devices attached via USB are probed
before RB_ASKNAME or RB_SINGLE attempts to ask any questions on the console,
and largely this works, often by chance, today, for USB disks and root.
i've recently pushed this more into uhub and general USB device attachment
as well, and kept a config_pending reference across the first explore of
a bus. these fix many issues with directly attached hubs.

however, on systems where devices connected to ehci ports are handed over
to a companion uhci or ohci port, it may not be the first, or even second,
bus explore that finds the device finally before attachment, and at this
point all config_pending references are dropped.

there is no direct communication between drivers, the potentials are
looked up but their device_t is only used for generic things like the name,
so informing the correct companion to expect a device and deal with the
config_pending references is not possible without some fairly ugly layer
violations or multi-level callbacks (eg, we have "ehci0", and usually an
the relevant companion, eg, "uhci2", but it is the uhub that uhci2 has
attached that will deal with the device attachment.)

with the above fixes to generic USB code, the disown happens during the
first explore. the hack works by, at this point, checking if (a) root
is not mounted, (b) single user or ask name are set, and (c) if the hack
as not been triggered already. if all 3 conditions are true, then a
config_pending_incr() is called and a callback is triggered for (default)
5 seconds to call config_pending_decr(). ehci detach pauses waiting for
this callback if scheduled.

this allows enough time for the uhub and the ukbd/wskbd to attach before
the RK_ASKROOT prompts appear. testing shows it takes between 1.5 and
2 seconds for the keyboard to appear after the disown occurs.


Index: dev/usb/ehcivar.c
- new sc_compcallout, sc_compcallout, sc_complock, and a state for th
handover hack.

Index: dev/usb/ehci.c
ehci_init():
- use aprint_normal_dev() instead of manual device_xname().
- initialise sc_compcallout, sc_compcallout, sc_complock, and sc_comp_state.
ehci_detach():
- if there are companion controllers, tear own the above, including waiting
if there is a callback scheduled.
ehci_disown_callback():
- new callout to call config_pending_decr() in the the future.
schedule this ca
ehci_disown_sched_callback():
- if booting to single user or asking names, call config_pending_incr() and
schedule the callout above, default 5 second delay.
ehci_disown():
- if disowning a port call ehci_disown_sched_callback().

pull across abort fixes from nick-nhusb. add more abort fixes, using
ideas from Taylor and Nick, and myself. special thanks to both who
inspired much of the code here, if not wrote it directly.

among other problems, this assert should no longer trigger:

panic: kernel diagnostic assertion "xfer->ux_state == XFER_ONQU" failed: file "/current/src/sys/dev/usb/usbdi.c", line 914

using usbhist i was able to track down my instance of it being related
to userland close() beginning, dropping the sc_lock, and then the usb
softintr completes the transfer normally, and when it is done, the
abort path attempts to re-complete the transfer, and the above assert
is tripped.


changes from nhusb were commited with these logs:
--
Move the struct usb_task to struct usbd_xfer for everyone to use.
--
Set device transfer status to USBD_IN_PROGRESS if start methods succeeds
--
Actually set the transfer status on transfers in ohci_abort_xfer and
the controller is dying
--
Don't supply the lock to callout_halt when polling as it won't be held
--
Improve transfer abort
--
Mark device transfers as USBD_IN_PROGRESS appropriately and improve
abort handling
--
#ifdef DIAGNOSTIC -> KASSERT and add another KASSERT
--
Mark device transfers as USBD_IN_PROGRESS appropriately and improve
abort handling
--

additional changes include:
- initialise the usb abort task in the HCI allocx routine, so that it
can be safely usb_rem_task()'d.
- rework the handling of softintr vs cancellation vs timeout abort based
upon a scheme from Taylor:
when completing a transfer normally:
- if the status is not in progress, it must be cancelled or timed out,
and we should not process this xfer.
- set the status as normal.
- unconditionallly callout_stop() and usb_rem_task(). they're safe and
either aren't running, or will run and do nothing.
- finally call usb_transfer_complete().
when aborting a transfer:
- status should be cancelled or timed out.
- if cancelling, callout_halt and usb_rem_task_wait() to make sure the
timer is either done or cancelled.
- at this point, the ux_status must not be cancelled or timed out, and
if it is not in progress we're done.
- set the status.
- if the controller is dying, just return.
- perform HCI-specific tasks to abort this xfer.
- finally call usb_transfer_complete().
for the timeout and timeout task:
- if the HCI is not dying, and the ux_status is in progress, then
trigger the usb abort task.
- remove UXFER_ABORTWAIT and UXFER_ABORTING.

tested on:
- multiple PC systems with several types of devices: ugen/UPS, ucom,
umass with disk, ssd and cdrom backends, kbd, ms, using uhci, ehci
and xhci.
- erlite3: sd@umass on dwc2.
- sunblade2000: kbd/ms and umass disk on ohci.

untested:
- motg, slhci and ahci. motg has some portion of the new scheme
applied, but slhci and ahci require more study.

future work includes pushing a lot of the common abort handling into
usbdi.c and leaving upm_abort() for HC specific tasks, but this change
is pullup-able to netbsd-7 and netbsd-8 as it does not change any
external API, as well as removing over 100 lines of code while adding
over 30 new asserts.

XXX: pullup-7, pullup-8.

Stop potential misuse of vendor names and USB vendor IDs in root hub
device and string descriptors.

Firstly: Few vendors have identical PCI-SIG vendor IDs and USB-IF vendor
IDs. As such, using the PCI vendor ID as a USB vendor ID may trample
on whomever is allocated that USB vendor ID.

Secondly: The vendor of the host controller hardware implementation has
little to nothing to do with our usbroothub implementation. Thus we
should not potentially associate any problems therewith to such third
party.

This change will result in root hubs being identified by USB Vendor ID
0x0000. Root hub vendor string will now be "NetBSD" (or, specifically:
ostype). Product ID (0x0000) and product strings remain unchanged.

branches: 1.43.10; 1.43.16;
Merge nick-nhusb

- API / infrastructure changes to support memory management changes.
- Memory management improvements and bug fixes.
- HCDs should now be MP safe
- conversion to KERNHIST based debug
- FS/LS isoc support on ehci(4).
- conversion to kmem(9)
- Some USB 3 support - mostly from Takahiro HAYASHI (t-hash).
- interrupt transfers now get proper DMA operations
- general bug fixes
- kern/48308
- uhub status notification improvements
- umass(4) probe fix (applied to HEAD already)
- ohci(4) short transfer fix

branches: 1.42.12; 1.42.14; 1.42.18;
Use CACHE_LINE_SIZE for EHCI_SQTD_ALIGN to make sure the sqtd fully consume
a cacheline.

use a pool instead of a linked list to avoid synchronization problems.

branches: 1.40.2;
merge the jmcneill-usbmp branch. many thanks to jared for the
initial work, and every one else who has tested things for me.
this is largely my fault at this point :-)

the main changes are something like:

- usbd_bus_methods{} gains a get_lock() to enable the
host controller to provide a lock for the USB code.
if the lock isn't provided, old-style protection is
(partially) applied.

- ehci/ohci/uhci have been converted to the new
interfaces, including mutex/cv/etc conversion.

- usbdivar.h contains a discussion about locking and
what locks are held for which method calls. more
to come for usbdi(9) here.

- audio drivers (uaudio, umidi, auvitek) have been
properly SMPified now that USB is ready.

- scsi drivers have been modified to take the kernel
lock explicitly before calling into scsi code.

- usb pipes are associated with a lock, that is the
same as the controller lock. (this could be split
up further in the future.)

- several usbfoo_locked() or usbfoo_unlocked()
functions have been added to the usbdi(9) to
enable functionality with or without the USB
lock (per controller) already being held.

the TODO.usbmp file has specific details on what is left to
do, including what device-specific changes should be done now
that the whole framework is ready.

pull down from usbmp:

kill the !USE_USE_SOFTINTR code.

branches: 1.38.6; 1.38.10;
Add preliminary support for Embedded Transaction Translator Function (as
found on the MPC8536 and AR9334) which allows low/full devices to be
connected to an EHCI root hub.

branches: 1.37.2;
Add vendor dependent functions.
sc_vendor_init()/sc_vendor_port_status().

branches: 1.36.2;
A pointer typedef entails trading too much flexibility to declare const
and non-const types, and the kernel uses both const and non-const
PMF qualifiers and device suspensors, so change the pmf_qual_t and
device_suspensor_t typedefs from "pointers to const" to non-pointer,
non-const types.

branches: 1.35.2;
Expand PMF_FN_* macros.

Expand <dev/usb/usb_port.h> definitions, and lightly unifdef(1).

branches: 1.33.16;
kern/39727: race condition in ehci isoc abort xfer path

From Jeremy Morse:

* Serialize access to the ehci intrlist.
* Change the ehci intrlist to a tailq so xfers are not queued out of order.
* In ehci_check_itd_intr, don't treat a transfer error as an indication
that the xfer is no longer active.

High speed isochronous support, from Jeremy Morse as part of Google
Summer of Code 2008.

branches: 1.31.2;
Add appropriate bus_dmamap_sync() calls to uhci(4), ohci(4) and ehci(4),
as proposed on tech-kern@.
While all DMA memory used in the USB framework is mapped BUS_DMAMAP_COHERENT
(including data memory, which is memcpy'd from/to USB-private buffers),
the CPU can reorder loads or stores from/to main memory, causing the
controller to have an incoherent view of the DMA descriptors lists for a
short time. bus_dmamap_sync() should contain memory barriers that prevents
the CPU from reordering load/store. Note that BUS_DMAMAP_COHERENT is
still required for the DMA descriptor lists - these can't work properly
with software cache coherency.
This fixes "host controller process error/host controller halted" errors
I'm occasionally seeing with a umodem device on uhci on x86.
Thanks to Michael Lorenz for testing it on his O2, and Izumi Tsutsui on
his Cobalt Qube 2700.

branches: 1.30.2; 1.30.4;
Remove clause 3 and 4 from TNF licenses

branches: 1.29.2; 1.29.4;
split device/softc for USB host controllers and the usb (control)
device,
this is hairy stuff, and I've only tested with uhci/ehci at pci,
please test the rest and report problems

Use device_t and accessors. Setup power management in the PCI
front-end (XXX needs to look more alike the Cardbus front-end).
Establish the shutdown hook using PMF.

Add a method to detach child.

Cosmetic: use device_t and accessors. Use aprint_*_dev(). Use
PMF_FN_{ARGS,CALL,PROTO}.

branches: 1.26.6; 1.26.10;
Merge jmcneill-pm branch.

branches: 1.25.2;
lockmgr -> mutex

branches: 1.24.14; 1.24.20; 1.24.22; 1.24.28; 1.24.30;
Remove the code which disables port status change interrupts for 1s
when one occured -- this makes that events get lost (or delayed until
the next change), in particular in the handover process from the EHCI
to a companion controller, laeving the port dead under some circumstances.
This fixes the "can't reconnect" symptom for me.
I don't see why the interrupt should be disabled - it is acknowledged
in the interrupt handler and shouldn't be active again until the next
status change.

branches: 1.23.18; 1.23.22; 1.23.24;
From OpenBSD:

--
Add a workaround for VIA EHCI controllers which, under load, signal qTD
completion before they have performed writeback from the overlay qTD.

This condition would exhibit itself as a umass stall that never recovers.
--

This fixes the problem reported by Thomas Klausner on current-users@:
http://mail-index.netbsd.org/current-users/2006/01/17/0000.html

branches: 1.22.2;
changes for making DIAGNOSTIC not change the kernel ABI:
- for structure fields that are conditionally present,
make those fields always present.
- for functions which are conditionally inline, make them never inline.
- remove some other functions which are conditionally defined but
don't actually do anything anymore.
- make a lock-debugging function conditional on only LOCKDEBUG.

as discussed on tech-kern some time back.

Keep a flag for each port that keeps track of if the port has been reset
instead of having one flag for the controller.

Also, don't try to modify read-only power bits if the controller doesn't
support power cobntrol.

branches: 1.20.6;
Minor cosmetic change.

branches: 1.19.2;
Allow 32 chars in the saved vendor string. Fixes kern/29760.

Add define for EHCI_MAX_POLLRATE.

branches: 1.17.2; 1.17.8;
Merge from FreeBSD:
sc_child is only used on NetBSD and OpenBSD, ifdef appropriately
FreeBSD revisions:
ehcivar.h 1.8
ohcivar.h 1.39
uhicvar.h 1.39

Implement an USB memory reserve. Allocate some memory per host controller
in the wanted bus_dma space. If an allocation fails during operation
(i.e. when hotplugging an umass device), fall back to using the reserve.

The amount can be configured as USB_MEM_RESERVE. The default value is
256k.

Ideally, there would be a way to steal pages in the desired area from
UVM, but that's far more complicated, and this is not intrusive, plus
it works.

Add interrupt transfers. From OpenBSD.
Having interrupt transfers enables some more device types to be used,
most notably USB2 hubs. USB2 hubs still cannot handle anything but
USB2 devices, because there is node code to handle hub Transaction Translation
yet.
XXX This code doesn't do good bw scheduling, but it's certainly better than
nothing. :)

Insert from ifdefs for soft interrupt related stuff. From OpenBSD.

Add suspend/resume support.
From OpenBSD.

branches: 1.12.16; 1.12.20;
Change abort slightly.
Change debug output.

Add some interrupt processing.

Set up control xfers.
Handle xfer timeouts.
Better debug messages.

Simplify async list handling.

Start of pipe open routine.

Add more fields to hardware structs.
Add memory allocation for various data structures.

Add some more data structure defs.

Recognize port status changes.
Hand over low and full speed devices to companion controller.

Add root hub emulation.

Add some register defines.
Print companion controllers.

Add some useful macros.

branches: 1.1.2; 1.1.4; 1.1.6; 1.1.10;
Add some placeholders for the EHCI (USB 2) driver.

Don't get your hopes up. I've not even finished reading the
(100+) page spec, and I have no hardware.

Merge nick-nhusb

- API / infrastructure changes to support memory management changes.
- Memory management improvements and bug fixes.
- HCDs should now be MP safe
- conversion to KERNHIST based debug
- FS/LS isoc support on ehci(4).
- conversion to kmem(9)
- Some USB 3 support - mostly from Takahiro HAYASHI (t-hash).
- interrupt transfers now get proper DMA operations
- general bug fixes
- kern/48308
- uhub status notification improvements
- umass(4) probe fix (applied to HEAD already)
- ohci(4) short transfer fix

branches: 1.42.14; 1.42.18;
Use CACHE_LINE_SIZE for EHCI_SQTD_ALIGN to make sure the sqtd fully consume
a cacheline.

use a pool instead of a linked list to avoid synchronization problems.

branches: 1.40.2;
merge the jmcneill-usbmp branch. many thanks to jared for the
initial work, and every one else who has tested things for me.
this is largely my fault at this point :-)

the main changes are something like:

- usbd_bus_methods{} gains a get_lock() to enable the
host controller to provide a lock for the USB code.
if the lock isn't provided, old-style protection is
(partially) applied.

- ehci/ohci/uhci have been converted to the new
interfaces, including mutex/cv/etc conversion.

- usbdivar.h contains a discussion about locking and
what locks are held for which method calls. more
to come for usbdi(9) here.

- audio drivers (uaudio, umidi, auvitek) have been
properly SMPified now that USB is ready.

- scsi drivers have been modified to take the kernel
lock explicitly before calling into scsi code.

- usb pipes are associated with a lock, that is the
same as the controller lock. (this could be split
up further in the future.)

- several usbfoo_locked() or usbfoo_unlocked()
functions have been added to the usbdi(9) to
enable functionality with or without the USB
lock (per controller) already being held.

the TODO.usbmp file has specific details on what is left to
do, including what device-specific changes should be done now
that the whole framework is ready.

pull down from usbmp:

kill the !USE_USE_SOFTINTR code.

branches: 1.38.6; 1.38.10;
Add preliminary support for Embedded Transaction Translator Function (as
found on the MPC8536 and AR9334) which allows low/full devices to be
connected to an EHCI root hub.

branches: 1.37.2;
Add vendor dependent functions.
sc_vendor_init()/sc_vendor_port_status().

branches: 1.36.2;
A pointer typedef entails trading too much flexibility to declare const
and non-const types, and the kernel uses both const and non-const
PMF qualifiers and device suspensors, so change the pmf_qual_t and
device_suspensor_t typedefs from "pointers to const" to non-pointer,
non-const types.

branches: 1.35.2;
Expand PMF_FN_* macros.

Expand <dev/usb/usb_port.h> definitions, and lightly unifdef(1).

branches: 1.33.16;
kern/39727: race condition in ehci isoc abort xfer path

From Jeremy Morse:

* Serialize access to the ehci intrlist.
* Change the ehci intrlist to a tailq so xfers are not queued out of order.
* In ehci_check_itd_intr, don't treat a transfer error as an indication
that the xfer is no longer active.

High speed isochronous support, from Jeremy Morse as part of Google
Summer of Code 2008.

branches: 1.31.2;
Add appropriate bus_dmamap_sync() calls to uhci(4), ohci(4) and ehci(4),
as proposed on tech-kern@.
While all DMA memory used in the USB framework is mapped BUS_DMAMAP_COHERENT
(including data memory, which is memcpy'd from/to USB-private buffers),
the CPU can reorder loads or stores from/to main memory, causing the
controller to have an incoherent view of the DMA descriptors lists for a
short time. bus_dmamap_sync() should contain memory barriers that prevents
the CPU from reordering load/store. Note that BUS_DMAMAP_COHERENT is
still required for the DMA descriptor lists - these can't work properly
with software cache coherency.
This fixes "host controller process error/host controller halted" errors
I'm occasionally seeing with a umodem device on uhci on x86.
Thanks to Michael Lorenz for testing it on his O2, and Izumi Tsutsui on
his Cobalt Qube 2700.

branches: 1.30.2; 1.30.4;
Remove clause 3 and 4 from TNF licenses

branches: 1.29.2; 1.29.4;
split device/softc for USB host controllers and the usb (control)
device,
this is hairy stuff, and I've only tested with uhci/ehci at pci,
please test the rest and report problems

Use device_t and accessors. Setup power management in the PCI
front-end (XXX needs to look more alike the Cardbus front-end).
Establish the shutdown hook using PMF.

Add a method to detach child.

Cosmetic: use device_t and accessors. Use aprint_*_dev(). Use
PMF_FN_{ARGS,CALL,PROTO}.

branches: 1.26.6; 1.26.10;
Merge jmcneill-pm branch.

branches: 1.25.2;
lockmgr -> mutex

branches: 1.24.14; 1.24.20; 1.24.22; 1.24.28; 1.24.30;
Remove the code which disables port status change interrupts for 1s
when one occured -- this makes that events get lost (or delayed until
the next change), in particular in the handover process from the EHCI
to a companion controller, laeving the port dead under some circumstances.
This fixes the "can't reconnect" symptom for me.
I don't see why the interrupt should be disabled - it is acknowledged
in the interrupt handler and shouldn't be active again until the next
status change.

branches: 1.23.18; 1.23.22; 1.23.24;
From OpenBSD:

--
Add a workaround for VIA EHCI controllers which, under load, signal qTD
completion before they have performed writeback from the overlay qTD.

This condition would exhibit itself as a umass stall that never recovers.
--

This fixes the problem reported by Thomas Klausner on current-users@:
http://mail-index.netbsd.org/current-users/2006/01/17/0000.html

branches: 1.22.2;
changes for making DIAGNOSTIC not change the kernel ABI:
- for structure fields that are conditionally present,
make those fields always present.
- for functions which are conditionally inline, make them never inline.
- remove some other functions which are conditionally defined but
don't actually do anything anymore.
- make a lock-debugging function conditional on only LOCKDEBUG.

as discussed on tech-kern some time back.

Keep a flag for each port that keeps track of if the port has been reset
instead of having one flag for the controller.

Also, don't try to modify read-only power bits if the controller doesn't
support power cobntrol.

branches: 1.20.6;
Minor cosmetic change.

branches: 1.19.2;
Allow 32 chars in the saved vendor string. Fixes kern/29760.

Add define for EHCI_MAX_POLLRATE.

branches: 1.17.2; 1.17.8;
Merge from FreeBSD:
sc_child is only used on NetBSD and OpenBSD, ifdef appropriately
FreeBSD revisions:
ehcivar.h 1.8
ohcivar.h 1.39
uhicvar.h 1.39

Implement an USB memory reserve. Allocate some memory per host controller
in the wanted bus_dma space. If an allocation fails during operation
(i.e. when hotplugging an umass device), fall back to using the reserve.

The amount can be configured as USB_MEM_RESERVE. The default value is
256k.

Ideally, there would be a way to steal pages in the desired area from
UVM, but that's far more complicated, and this is not intrusive, plus
it works.

Add interrupt transfers. From OpenBSD.
Having interrupt transfers enables some more device types to be used,
most notably USB2 hubs. USB2 hubs still cannot handle anything but
USB2 devices, because there is node code to handle hub Transaction Translation
yet.
XXX This code doesn't do good bw scheduling, but it's certainly better than
nothing. :)

Insert from ifdefs for soft interrupt related stuff. From OpenBSD.

Add suspend/resume support.
From OpenBSD.

branches: 1.12.16; 1.12.20;
Change abort slightly.
Change debug output.

Add some interrupt processing.

Set up control xfers.
Handle xfer timeouts.
Better debug messages.

Simplify async list handling.