pci: extend MSI interrupt vector number to 32 bits

Also increase MSI message data size to 32 bits according to PCIe spec.

Remove 0xff check for MSI interrupts because it is potentially valid
interrupt vector number. Reject 0xff only for legacy pin interrupts.

- MSI-X supports up to 2048 interrupts per device that do not fit to
`uint8`.

- Non-x86 systems may use separate interrupt vector ranges for
hard-wired interrupts and MSI interrupts so `uint8` is not enough to
represent all of them.

Change-Id: Iaf9ffb197ec23db0f97ffe3ea756d28d7bfc8705
Reviewed-on: https://review.haiku-os.org/c/haiku/+/7433
Reviewed-by: waddlesplash <waddlesplash@gmail.com>
Tested-by: Commit checker robot <no-reply+buildbot@haiku-os.org>

USB: Make "buffer error" reporting consistent.

This error means that the controller failed to read or write data
to/from system memory fast enough. As a result, we should report it
as READ_ERROR/WRITE_ERROR, rather than something that sounds like
an error with the device itself.

Then, make BABBLE reporting consistent: this is a data overrun/underrun
of the device itself, not the buffers passed to the controller.
This then leaves B_DEV_FIFO_OVERRUN/UNDERRUN unused, and thus usable
for reporting ring overrun/underrun on isochronous transfers.

The API documentation already described DATA_OVERRUN/UNDERRUN basically
like they were babble errors, while FIFO_OVERRUN/UNDERRUN are
currently described as "internal errors" at present. No driver actually
checked for these.

USB: Support physical-vector bulk requests.

Introduce a new utility method, "generic_memcpy", which takes
generic_addr_t plus indications of whether these specify virtual or
physical addresses (and potentially user addresess) and calls the
appropriate memcpy variant depending.

All bus drivers adjusted to support this at once. We don't actually
take advantage of the physical addresses in any way (yet), as USB
controllers have some pretty specific requirements that would have
to be carefully validated to use these directly.

All bus drivers tested and confirmed to still be working.

Change-Id: I66326667e148091147bb2b3d0843a26fb7e5bda6
Reviewed-on: https://review.haiku-os.org/c/haiku/+/6479
Reviewed-by: waddlesplash <waddlesplash@gmail.com>

USB: Use generic_io_vec in Transfer.

Prepares the way for actual physical-address request support,
which is not yet wired up.

Change-Id: I1b2d04a31e334a79b8361280fa0f3a5fbdb43d2b
Reviewed-on: https://review.haiku-os.org/c/haiku/+/6478
Reviewed-by: waddlesplash <waddlesplash@gmail.com>

bus & drivers: drop PCI_x86

Change-Id: I494deaf24a4793a5e0fe9fa46ecdce32f65e616a
Reviewed-on: https://review.haiku-os.org/c/haiku/+/6226
Reviewed-by: waddlesplash <waddlesplash@gmail.com>
Reviewed-by: Fredrik Holmqvist <fredrik.holmqvist@gmail.com>

kernel: Drop non-standard GNU inline assignment syntax

* We needed this previously due to our gcc2 compiled kernel.
* Now that our kernel is always latest gcc, we can move to the
c++20 syntax for inline assignment.
* Improves compatibility with clang, less GNU-specific stuff

Change-Id: Ib7272a0a52554a31e9a0e788fd3f031db9049795
Reviewed-on: https://review.haiku-os.org/c/haiku/+/5898
Reviewed-by: Alex von Gluck IV <kallisti5@unixzen.com>
Reviewed-by: Adrien Destugues <pulkomandy@pulkomandy.tk>

usb: fix references on the module B_PCI_X86_MODULE_NAME

the dependency is optional, only the device modules depend on it.
should fix #18075

Change-Id: Ifddb9814ff5ff59ee7d80d06b14ba9fded04cd56
Reviewed-on: https://review.haiku-os.org/c/haiku/+/5822
Reviewed-by: waddlesplash <waddlesplash@gmail.com>

usb: register device nodes for hubs and devices

the device identifier is the USBID. no actual interface to be used by drivers.

Change-Id: I9fd1fdab0d02f1b82380c619b39bc3a12a9e2333
Reviewed-on: https://review.haiku-os.org/c/haiku/+/5736
Tested-by: Commit checker robot <no-reply+buildbot@haiku-os.org>
Reviewed-by: Fredrik Holmqvist <fredrik.holmqvist@gmail.com>
Reviewed-by: X512 <danger_mail@list.ru>
Reviewed-by: waddlesplash <waddlesplash@gmail.com>

usb: remove debug leftover

Change-Id: Idba6eaecb2c5b5773cd61794ac62fac1f7e3c1f9
Reviewed-on: https://review.haiku-os.org/c/haiku/+/5715
Tested-by: Commit checker robot <no-reply+buildbot@haiku-os.org>
Reviewed-by: waddlesplash <waddlesplash@gmail.com>

ehci: switch to the new driver API

Change-Id: I9102a27fe8e536003a0ab09c9cca5a87e88c8092
Reviewed-on: https://review.haiku-os.org/c/haiku/+/5709
Reviewed-by: waddlesplash <waddlesplash@gmail.com>

EHCI: Do not ignore status of PrepareKernelAccess.

May help with #17799.

USB bus drivers: Add missing iterator increments.

Should fix boot hang regressions introduced in hrev55455
in the fix for #17275.

USB bus drivers: Return error when combining other transfers with a fragmented one.

The drivers do not support this properly at present, they would
run the other transfers interspersed with fragments from the fragemented one,
which is obviously the wrong thing to do.

No USB device drivers seem to do this at present (it would cause data
corruption if they had.)

Fixes #17275.

USB: Rename Transfer::VectorLength to Transfer::FragmentLength.

It did not return the length of the vectors, but only the current
fragment's worth of vectors. It also modified the fFragmented flag,
which really should have been set in SetVectors in the first place.

As everything seemed to call IsFragmented after VectorLength,
this is not a behavioral change.

EHCI: Call InitKernelAccess at the top of SubmitTransfer functions.

I missed that this was not done before WriteDescriptor was called.
Fixes #14835 (again.)

EHCI: Call PrepareKernelAccess before writing descriptor chains.

Fixes #15653.

Reduce polling interval for low-speed devices

Fixes #7651, my Logitech gamepad works.

Change-Id: Ia3f92ba817576946dabee262a2fd5af8102a109f
Reviewed-on: https://review.haiku-os.org/c/haiku/+/2198
Reviewed-by: Stephan Aßmus <superstippi@gmx.de>

EHCI: Destroy the cleanup semaphore on teardown.

Otherwise, the wait_for_thread() of the fCleanupThread will block
forever, stalling the boot process.

Fixes another part of #15004.

busses/usb: Print the PCI ID of the device with an invalid IRQ.

busses/usb: Handle getting/putting the PCI modules properly.

Previously it was possible that we put() them twice, leading to
assert-failure panics as their ref-counts would go below 0.
Now we get() them once for every bus object that we create, as
the destructors put() them.

Should fix the panic in #15004.

busses/usb: Don't reject devices with "invalid" IRQs immediately.

We may be able to allocate an MSI for them. This was done for XHCI
already in hrev52742.

Probably fixes #15004.

busses/usb: Move ::AddTo to be above the class constructors.

It is called before them (and is responsible for calling them),
so having it in the middle of the file does not make a lot of sense.

Already done for XHCI. Only one functional change -- removing the
set_dprintf_enabled call. Drivers probably shouldn't spuriously
re-enable that if it was specifically disabled.

USB: Properly assign IDs to USB Bus Managers (HCI drivers).

Previously they would just get -1, as the BusManager class
would request their ID before they had been added to the Stack.
Now we add them to the stack inside ::Start(), rather than letting
the individual drivers do that just after ::Start(), and then assign
the ID there directly.

drivers: remove some non-haiku support code.

Change-Id: Ic55bb4832adabeb807f763c87ad938e74fb3a97d
Reviewed-on: https://review.haiku-os.org/520
Reviewed-by: Axel Dörfler <axeld@pinc-software.de>
Reviewed-by: waddlesplash <waddlesplash@gmail.com>

Revert "EHCI USB: process the extended capabilities chain"

This reverts commit 137135a10dc6322773be9da67861c05f53b709a5.

Change-Id: I6e6ef1933b3b3f5eda7278d37b673eefe2867c2a

EHCI USB: process the extended capabilities chain

...instead of the first one only.

usb: map registers or allocate area for the kernel only.

EHCI: Current qtd field in qh overlay has no terminate bit.

Don't initialize the current qtd field with the terminate bit as it does
not exist in this field. While the lower 5 bits are documented as being
ignored anyway, this is more correct. Should cause no functional change.

EHCI: Fix alignment check, the mask wasn't wide enough.

As this is only an assert that shouldn't ever be triggered, this doesn't
change anything in normal operation.

EHCI: Style cleanup only, no functional change.

Change < B_OK to != B_OK for status_t results and to < 0 for ids.

USB: Generalize logic for debug transfers used within KDL.

Pull debug transfer methods into the base BusManager class and
implement them for UHCI and EHCI. This makes it possible to have a
single debugger command installed by the USB stack instead of HCI
specific ones.

EHCI: Implement debug transfers.

EHCI: Move locking to be only around actually critical section.

EHCI: Simplify short packet logic to use alternate pointer.

The alternate pointer is the software equivalent of the alternate
physical address in the hardware queue head part. The controller
follows that path, so we can follow the alternate pointer as well.
Simplifies and generalizes the logic.

More style fixes to ehci.

More style fixes.

Style and 80-column fixes.

Fix Short Packet EHCI Transfers

Signed-off-by: Adrien Destugues <pulkomandy@pulkomandy.tk>

Fixes #10867.

add-ons/kernel: add spaces between literals and identifiers

Due to introduction of user-defined suffixes C++11 requires that there
is a space between literal and identifier to avoid ambiguity.

This patch makes the whole kernel build successfully with C++11.

EHCI USB: add MSI support

* similar to OHCI support by mmlr.
* interrupt handler is removed on destruction.

ehci: initializes fItdEntries and fSitdEntries.

Fix build with tracing turned on.

Fix compilation of OHCI on x86.

Fixes were needed to go with the USB stack changes. Haven't tested it
yet because I don't have a a machine with OHCI, but I doubt there
will be any problems.

Get USB functional on x86-64.

- Various changes to printf format specifiers to make them 32/64 safe.
- Adjustments to various descriptor structures since addr_t is not 32-bit
on x86-64, and consequently this breaks the physical structure that's
expected. Also adjusted padding to accomodate 32-byte boundaries when
64-bit pointers are in use.
- Adjustments to make use of phys_addr_t where needed.

Move mutex initialization earlier, analogous to hrev43320.

Avoids crashing if the bus manager init fails. Also fixes CID 10846.

Use the right form of delete. Fixes CID 11037.

Account for full-/lowspeed errors and ensure an error is set.

For full- and lowspeed devices the unused ping status bit works as an
additional error bit (albeit it's so generic that it isn't really
useful). Include that bit when the error counter counts down to zero.

Also ensure that an error is returned if the error counter reaches
zero, but print a warning if none of the describing bits are set.

Make two unexpected/error cases more visible with debug output.


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@42965 a95241bf-73f2-0310-859d-f6bbb57e9c96

Read the previous descriptor pointer before freeing transfer and transfer descriptors.


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@42701 a95241bf-73f2-0310-859d-f6bbb57e9c96

* threshold higher bit means a full frame. Also use the threshold for micro frames instead of frames.
* introduces fNextStartingFrame to keep track of the next frame to use on next submit.
* set the IOC bit for the last ITD
* computes multiply field of the ITD based on the packed size (1, 2 or 3).
* use locking around linking of ITD
* free descriptors on errors
* on finishing, whenever a non success status is found, set actual length to zero
* on finishing, when copying data to the Transfer object, copy starting on packet boundaries (skipping unused bytes).


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@42679 a95241bf-73f2-0310-859d-f6bbb57e9c96

Locking around descriptors list handling isn't enough, locking is also needed when traversing the list: we instead lock the whole traversing/handling loop.


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@42623 a95241bf-73f2-0310-859d-f6bbb57e9c96

use locking when messing ITD and SITD queues. itd->prev was becoming NULL, thus leading to KDL.


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@42575 a95241bf-73f2-0310-859d-f6bbb57e9c96

We obviously need to mask off the interrupt threshold value before overwriting
it with our new setting. Otherwise we might end up with an illegal value as the
interrupt threshold which might inhibit interrupt generation alltogether
depending on the controller implementation. This was the case for the ATI/AMD
SBx00 chipsets. Therefore fixes #5551.


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@41690 a95241bf-73f2-0310-859d-f6bbb57e9c96

Introduce simplistic poll mode for EHCI. If enabled a thread will run the
interrupt handler every millisecond so it can check for interrupts to handle.
While this is certainly not ideal, it allows EHCI to run (surprisingly well
even) on chipsets that experience interrupt routing issues.
It can be enabled with the safemode setting "ehci_polling on" that can either
be entered using the new advanced debug option entry in the bootloader and/or
in the kernel settings file.


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@41688 a95241bf-73f2-0310-859d-f6bbb57e9c96

* added some EHCI definitions (specification 1.1)
* enable per-port change events when available


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@41661 a95241bf-73f2-0310-859d-f6bbb57e9c96

ehci usb driver:
* make sure the PortEnable bit is disabled (specification page 28)
* 250ms is too long: USB 2.0 specifies only 50ms. This prevented one of my devices to be detected.
+alpha


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@41658 a95241bf-73f2-0310-859d-f6bbb57e9c96

Check if PCI interrupts are disabled on the EHCI device. Rather unlikely to be
the case, but maybe there's some BIOSes that disable them? More or less poking
around for #5551, doesn't do any harm in any case.


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@41514 a95241bf-73f2-0310-859d-f6bbb57e9c96

Introduce workaround for the (documented) bug with ATI/AMD SB600 and early SB700
periodic list cache. This workaround is based on similar patches in NetBSD and
Linux and disables the (Advanced) Periodic List Cache on the affected devices
using ATI/AMD specifc registers (as documented in the corresponding AMD
register reference guide).
Also remove a ton of stray whitespace introduced by the isochronous patches.
Please take more care the next time.


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@41512 a95241bf-73f2-0310-859d-f6bbb57e9c96

* Implement inbound isochronous transfer for EHCI, based on UHCI implementation.
* changes next_log to their actual type instead of void*
* the field this_phy now includes the item type, it simplifies things.
* isochronous transfer descriptors are linked in the periodic frame list first,
and points to existing interrupt transfer descriptors.

This is still work in progress. Yet it's worth committing as it doesn't seem to
have impacts, and is required for the UVC SoC project. Tested basically with
usb_webcam to receive UVC stream headers.



git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@41424 a95241bf-73f2-0310-859d-f6bbb57e9c96

activate some traces for EHCI bus driver (see bug #4718)


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@41270 a95241bf-73f2-0310-859d-f6bbb57e9c96

added traces for unimplemented isochronous transfers for ehci and ohci


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@41194 a95241bf-73f2-0310-859d-f6bbb57e9c96

* Changed physical_entry::{address,size} to phys_{addr,size}_t and changed
map_physical_memory()'s physicalAddress parameter type from void* to
phys_addr_t. This breaks source compatibility, but -- as long as
phys_{addr,size}_t remain 32 bit wide -- keeps binary compatibility with
BeOS.
* Adjusted all code using the affected interfaces (Oh what fun!). Added a few
TODOs in places where the wrong types (e.g. void* for physical addresses
are used). Looks like quite a few drivers aren't 64 bit safe and others
will break with PAE.


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@36960 a95241bf-73f2-0310-859d-f6bbb57e9c96

Don't take ownership by writing the OS-owned semaphore when the BIOS-owned flag
isn't set. I'd tend to say that's a rather odd interpretation of the specified
protocol if a BIOS goes mad if the sem is written to while it doesn't actually
own the controller (it should just be a no-op in that case and directly set the
sem state). But it seems this is what others do as well (FreeBSD, Linux) so we
follow. Thanks to vegardw in #2083 for researching into this and proposing this
fix. Minor other cleanup.


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@35780 a95241bf-73f2-0310-859d-f6bbb57e9c96

We need to guarantee that when cancel returns that the callback won't be called
anymore. It could happen that the transfer was already in the process of being
finished, so wasn't in the list of pending transfers anymore. Cancel would then
return even though the callback wasn't called yet. This could lead to a callback
being called after a driver was already unloaded (even after it cleaned up the
pipes it used).


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@33239 a95241bf-73f2-0310-859d-f6bbb57e9c96

* Add the kernel debugger extensions only for the haiku target.
* Acknowledge non-enabled interrupts in UHCI as done in EHCI as well.


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@30063 a95241bf-73f2-0310-859d-f6bbb57e9c96

It turns out that on some BIOSes even if the handover is signaled as successful,
the BIOS doesn't remove it's SMIs. When then initializing the controller this
would trigger all sorts of stuff and for example freeze the system. We therefore
always force off the SMIs, even if the handover didn't fail.


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@29931 a95241bf-73f2-0310-859d-f6bbb57e9c96

* Check for and discard not enabled interrupts in EHCI.
* Simplify acknowledging interrupts a bit.

Might help with bug #2083 in case the controller would interrupt on a port
change or frame list rollover even though those aren't enabled by us.


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@29354 a95241bf-73f2-0310-859d-f6bbb57e9c96

* Rework the USB tracing mechanism. Cleaned it up and make it more convenient to
use. It will now print out the usb_ids of the objects that generate the trace
messages. These IDs are unique compared to the device address used previously,
because device addresses are per bus while usb_ids are global. This makes
trace output from devices across multiple controllers distinguishable.
* Some cleanup.


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@29002 a95241bf-73f2-0310-859d-f6bbb57e9c96

Output some more debug output in case of port reset failure.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@26546 a95241bf-73f2-0310-859d-f6bbb57e9c96

When calculating the actual length for requests, we don't want to include the
8 byte request header. It was included in the case of outgoing request
transfers before. In fact we always just want the length of the data phase,
which for non-requests just happens to be the whole transfer.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@26508 a95241bf-73f2-0310-859d-f6bbb57e9c96

If the BIOS doesn't give up ownership of an EHCI controller, simply clear the
BIOS owned flag and clear all possibly enabled SMIs. It seems to be common
practice to just ignore this error case, probably because there are enough
broken BIOSs out there that do not correctly hand over the controller.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@25918 a95241bf-73f2-0310-859d-f6bbb57e9c96

When canceling transfers, do not call the callbacks with the lock held. This
prevents deadlocks in cases where a new transfer is scheduled from within the
cancel callback. This is an edge case, as generally you don't want to schedule
anything when explicitly canceling transfers, but there are a valid use cases
when you cancel because of a timeout and then have to reset the device for
example. Note that this hides bug #2353, where the cancel case is probably
not handled correctly.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@25861 a95241bf-73f2-0310-859d-f6bbb57e9c96

Only check the explicitly enabled interrupts in EHCI. Otherwise we might steal
a shared interrupt from someone else.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@25822 a95241bf-73f2-0310-859d-f6bbb57e9c96

* Added optional spinlock contention measurement feature. Enabled when
B_DEBUG_SPINLOCK_CONTENTION is defined to 1. It typedefs spinlock to a
structure (thus breaking BeOS binary compatibility), containing a
counter which is incremented whenever a thread has to wait for the
spinlock.
* Added macros for spinlock initialization and access and changed
code using spinlocks accordingly. This breaks compilation for BeOS --
the macros should be defined in the respective compatibility wrappers.
* Added generic syscall to get the spinlock counters for the thread and
the team spinlocks.


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@25752 a95241bf-73f2-0310-859d-f6bbb57e9c96

* Increase the amount of tries in the EHCI BIOS handover so it lasts for up to
1 second instead of half a second. Might help with slow BIOSes.
* Disable EHCI interrupts if the BIOS handover fails so that we do not flood
the system with interrupts if the BIOS later on decides to still comply with
our handover request.
* Add an unconditional success message to EHCI and UHCI to better see if
controllers are present/initialized correctly at all.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@24980 a95241bf-73f2-0310-859d-f6bbb57e9c96

Exclude the frame list rollover interrupt from the interrupts we are interested
in. We do not enable this interrupt and if we still "handle" it in our
interrupt handler we might prevent the interrupt from reaching the real intended
handler in a shared interrupt setup. Not sure if this could have happend at
all, but this is more correct in any case.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@24965 a95241bf-73f2-0310-859d-f6bbb57e9c96

Fully clear the EHCI members too, so an incompletely constructed object can be
safely deleted.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@24957 a95241bf-73f2-0310-859d-f6bbb57e9c96

* Set the interrupt threshold of the controller when doing the rest of the
initialization as suggested by the EHCI spec and do not touch it again
once the controller is running. Doing so would result in undefined behaviour
according to EHCI specs.
* Set the port routing policy after the controller has started and not before
as suggested by the EHCI specs.
* Explicitly set the segment register after resetting the host controller and
not before, as this likely has no real effect ;-)

Might help with some legacy support or UHCI/OHCI to EHCI handover problems.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@24925 a95241bf-73f2-0310-859d-f6bbb57e9c96

Fix building usb busses with tracing on.


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@24880 a95241bf-73f2-0310-859d-f6bbb57e9c96

Change just host_controller_info to usb_host_controller_info for the sake of
clearity as suggested by Philippe Houdoin once and brought to my attention by
Slavatore. As this definition is completely private to the USB stack it would
not have caused any conflicts though.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@24828 a95241bf-73f2-0310-859d-f6bbb57e9c96

We are not supposed to overwrite the full capability register when setting our
OS-owned bit. Otherwise we overwrite exactly the bit we later have to wait for.
Could theoretically help on legacy support issues, but I doubt it since most
hardware probably does not even use EHCI legacy support as it's normally not
needed to support full/low-speed input devices.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@23747 a95241bf-73f2-0310-859d-f6bbb57e9c96

Unconditionally output when giving up port ownership and taking over host
controller ownership in EHCI to aid debugging USB legacy support issues.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@23735 a95241bf-73f2-0310-859d-f6bbb57e9c96

Notify the finisher thread after canceling transfers. It was possible that freeing the allocated
transfer resources would be delayed unnecessarily until some other transfer completed and woke up
the finisher thread. Found by Salvatore Benedetto, thanks!

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@23713 a95241bf-73f2-0310-859d-f6bbb57e9c96

* fixed memory leak: if InitKernelAccess failed data was not deleted



git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@23703 a95241bf-73f2-0310-859d-f6bbb57e9c96

* Add support for low/full speed interrupt transfers over USB 2.0 hubs
* Actually enable the periodic schedule so that any interrupts work over EHCI

For those who wait for OHCI: You can now attach your USB 1.1 mouse or keyboard
to a USB 2.0 hub and it should work.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@23382 a95241bf-73f2-0310-859d-f6bbb57e9c96

Messed up those numbers. The interval can range from 0 to 16 and we can use 0 to 10 with the 1024 frame list. So it's obviously 11 entries and not 10. Fixes the wrong tree layout and the doubled polling intervals of interrupts.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@23035 a95241bf-73f2-0310-859d-f6bbb57e9c96

Since the interval is now part of the general pipe info we don't have to cast the pipe to an interrupt one anymore.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@22976 a95241bf-73f2-0310-859d-f6bbb57e9c96

Implemented split transactions for control and bulk pipes in EHCI.
With this it is possible to use low/fullspeed (USB 1.1) devices below USB 2.0 Hubs.
Note though that you cannot yet use mice or keyboards attached to USB 2.0 Hubs as split transactions for interrupt pipes are still missing.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@22973 a95241bf-73f2-0310-859d-f6bbb57e9c96

* Removed the two extra submit functions and combined them directly into SubmitTransfer()
* The sparation based on async and periodic schedule made no sense as it is a question of queued/non-queued
* Therefore removed the near complete code duplication for adding interrupt transfers (that are also queued)

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@22943 a95241bf-73f2-0310-859d-f6bbb57e9c96

* Implement interrupt transfers in EHCI
* Uses a "collapsed binary tree" (for lack of a better name) to support the different intervals
* Remove a leftover variable declaration that was hiding error conditions away...
* Some cleanup

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@22942 a95241bf-73f2-0310-859d-f6bbb57e9c96

Completely redesign the USB explore process. Replaces the scary race conditions of the previous locking mechanism and simplifies handling of device changes by a more centralized approach.
Changes are now collected during explore and notifications as well as rescans are done at once. Through this a driver is also not rescanned multiple times anymore.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@22929 a95241bf-73f2-0310-859d-f6bbb57e9c96

* Apply more sane canceling logic to EHCI too
* Remove the leftover user_area - this is handeled inside the Transfer class by now

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@21906 a95241bf-73f2-0310-859d-f6bbb57e9c96

Revert r21231 as it's supposedly not possible. Someone implement OHCI then pleazzze!


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@21241 a95241bf-73f2-0310-859d-f6bbb57e9c96

Added a keep_port_ownership driver setting for EHCI, to allow keeping devices on EHCI instead of giving them up to the 1.0 controller.
Hopefully this will get things working here until OHCI is implemented.


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@21231 a95241bf-73f2-0310-859d-f6bbb57e9c96

* Reworked userland buffer access functionality into two functions in the Transfer class to not duplicate code in each HCD
* Use these functions in UHCI and EHCI which greatly reduces overhead on fragmented transfers as the vector conversion and area cloning is only done once
* Use the PrepareKernelAccess() function before submitting fragments which fixes the crash when trying to access the next fragments buffer in user memory from the kernel thread

This should fix bug #1167 which could be encountered when sending data to a mass storage device with an application that uses a large enough transfer buffer size.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@21059 a95241bf-73f2-0310-859d-f6bbb57e9c96

Fixed wrong cancelling algorithm. This is what you get when you "rework" an existing function without thinking it through again... Thanks J Jerome for noticing\!

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@21058 a95241bf-73f2-0310-859d-f6bbb57e9c96

Implemented transfer cancelling for EHCI by reworking the previously unused CancelPendingTransfer() into CancelQueuedTransfers() with the pipe based semantic.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@21044 a95241bf-73f2-0310-859d-f6bbb57e9c96

Implemented fragmented USB transfers. The transfer length will max out at a certain point to not overflow the allocator. The fragmented transfers are resubmitted until all fragments are sent / received.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@20416 a95241bf-73f2-0310-859d-f6bbb57e9c96

* Fixed spinlock usage in EHCI and UHCI drivers
* Fixed UHCI to allow for empty transfer buffers
* Extended the check for still linked transfer descriptors when removing a chain

There remain UHCI bugs that will cause transfers outgoing transfers to fail. I'll have to review / rewrite the UHCI driver.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@19730 a95241bf-73f2-0310-859d-f6bbb57e9c96

* Implemented a notification method for BusManagers so that they can be notified of pipe changes (creation, destruction, changed settings).
This is necessary in OHCI and will probably be used in EHCI also to keep one endpoint construct for each pipe open instead on creating and deleting it for each transfer.
* Pseudo implemented set_pipe_policy for isochronous pipes that makes use of the new notification system.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@18945 a95241bf-73f2-0310-859d-f6bbb57e9c96

* Added retries to the PhysicalMemoryAllocator. Memory is allocated and freed at a very high frequency, so low memory conditions shouldn't last very long.
* Added a separate thread for cleaning up EHCI transfers. This makes it possible to actually always wait for the async advance interrupt without hindering execution of transfers.
This pushes performance again and fixes the bug I introduced in the last change, that we could free yet cached descriptors when a previous async advance timed out.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@18929 a95241bf-73f2-0310-859d-f6bbb57e9c96

* Switched from the old v2 callback status codes to the new v3 codes
This means that the usb_callback_func now takes a status_t instead of a uint32 status.
Also the error codes are now different. I don't see this as a real problem in binary compatibility, as the status codes were never really documented and most drivers just assumed that a nonzero status meant an error.
Source compatibility breaks for callback functions and error defines. I fixed (hopefully) all places in the tree that are currenty included in the image and affected by the change.
* Corrected error reporting in UHCI and EHCI using the new status codes.
* Fixed a memory leak in EHCI where the async advance interrupt was not triggered in time.
* Fixed another race condition in usb_raw that could cause a crash when a device is removed while a transfer is pending.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@18916 a95241bf-73f2-0310-859d-f6bbb57e9c96

* Fixing my previous attempt at fixing cookie handling. Device addresses cannot be used, since there may be devices with the same address across different busses. A linked list of usb_driver_cookies is maintained now that relates devices and cookies.
* Fixed endpoint halt clearing. The data toggle has to be reset when we clear a stall.
* Now using up to B_PAGE_SIZE * 4 sized buffers in EHCI. They are split automatically by the host controller. This further pushes throughput.
* Optimize memory allocator settings for the above case. Enlarging it's managed space to 1MB
* Ignore the transaction error bit in EHCI. It is only of informational character as it also idicates errors that were recovered automatically. If an unrecoverable error occures, other bits will indicate this.
* Fixed some more print formats in debug output.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@18915 a95241bf-73f2-0310-859d-f6bbb57e9c96

* Implemented compatibility workaround for BeOS to be able to write user memory from the kernel finisher thread

This fixes crashing of the USB Mass Storage driver under R5. It is fully working there now.
I hope to be able to do this more cleanly under Haiku though.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@18894 a95241bf-73f2-0310-859d-f6bbb57e9c96

* Implemented usable thread synchronisation for UHCI and EHCI.
* Cleaned interrupt handlers for both busses, too.
* Fixed memory leak of the queue head stray descriptors in EHCI.
* Fixed warnings about assigning NULL to the module flags field.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@18857 a95241bf-73f2-0310-859d-f6bbb57e9c96

* Implemented transfer finishing in EHCI
* Implemented actual length calculation in EHCI
* Implemented bulk transfers in EHCI
* Implemented queue freeing in EHCI
* Added extended buffer pointers to the EHCI structs, necessary when the controller supports 64bit
* Fixed debug output and removed warnings due to wrong print format

Control and bulk transfers are now working nicely for EHCI. This means that I can now get pictures from my camera with USB 2.0 speed. Note though that many things still are unimplemented (isochronous transfers, split transactions, interrupt transfers, queue chaches per pipe, ...). But it's certainly a nice step forward :-).

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@18855 a95241bf-73f2-0310-859d-f6bbb57e9c96

* Implemented most qTD and qH functions for EHCI async transfers.
* Implementing control and bulk (async) transfers. They don't work yet though.
* Corrected the legacy support handling in EHCI
* Fixed the bit mask for total bytes in qTDs and added the data toggle bit
* Changed the boolean low speed flags to usb_speed enum that allows for low, full and highspeed

Also wrote a memory allocator that takes over memory management from the stack. It is similar to a buddy allocator. This does on the one hand remove some ugly code from the stack and is on the other hand far more scalable than the previous free list approach. This scalability is needed for proper EHCI support where buffers can go up to a size of 5 pages. The allocator is not perfect and it's memory overhead is quite high compared to the managed memory (depends on the managed block sizes), but it's quik and works well so far.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@18849 a95241bf-73f2-0310-859d-f6bbb57e9c96

* Rewrote some stack internals (again). Everything is now created using a parent object that shares the BusManager and Stack information to reduce individual constructors.
* Don't inherit Device and Interface from pipes anymore. The device has a default pipe member now.
* Since the parent of each Object is now known, we can send control messages using the default pipe of the device from all attached elements (pipes and interfaces).
* Fixed the generic {Set|Clear}Feature() and GetStatus() calls to include the endpoint address for pipe specific requests.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@18710 a95241bf-73f2-0310-859d-f6bbb57e9c96

replaced with PCI.h definitions, added PCI_usb_ehci


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@18704 a95241bf-73f2-0310-859d-f6bbb57e9c96

* Redisigned some of the USB Stack. Unified constructors, removed redundant functions and moved some other functions around
* Implemented the EHCIRootHub
* Reworked UHCIRootHub to be more of a utility class and moved the actual work into the UHCI BusManager
* Implemented port reset in EHCI. Ports with low and fullspeed devices are now correctly handed over to the companion host controller.

Attachs and detachs as well as port resetting / enabling works now in EHCI. Only the actual data transfers are missing thereafter ;-)

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@18656 a95241bf-73f2-0310-859d-f6bbb57e9c96

* Changed EHCI register access from port to memory mapped io (which is required)
* Added individual register operations for operational and capability registers
* Added resetting the host controller which appearantly actually works

Note that you shouldn't install the ehci module if you want uhci to work.
It disables the companion host controller drivers (uhci and ohci) because it takes port ownership and does not yet give it back for low and fullspeed devices.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@18648 a95241bf-73f2-0310-859d-f6bbb57e9c96

* Adding skeleton EHCI driver. Some parts filled from the specs and the UHCI driver.
* Corrected some errors I found while reading UHCI.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@18626 a95241bf-73f2-0310-859d-f6bbb57e9c96

USB: Generalize logic for debug transfers used within KDL.

Pull debug transfer methods into the base BusManager class and
implement them for UHCI and EHCI. This makes it possible to have a
single debugger command installed by the USB stack instead of HCI
specific ones.

EHCI: Implement debug transfers.

EHCI: Move locking to be only around actually critical section.

EHCI: Simplify short packet logic to use alternate pointer.

The alternate pointer is the software equivalent of the alternate
physical address in the hardware queue head part. The controller
follows that path, so we can follow the alternate pointer as well.
Simplifies and generalizes the logic.

More style fixes to ehci.

More style fixes.

Style and 80-column fixes.

Fix Short Packet EHCI Transfers

Signed-off-by: Adrien Destugues <pulkomandy@pulkomandy.tk>

Fixes #10867.

add-ons/kernel: add spaces between literals and identifiers

Due to introduction of user-defined suffixes C++11 requires that there
is a space between literal and identifier to avoid ambiguity.

This patch makes the whole kernel build successfully with C++11.

EHCI USB: add MSI support

* similar to OHCI support by mmlr.
* interrupt handler is removed on destruction.

ehci: initializes fItdEntries and fSitdEntries.

Fix build with tracing turned on.

Fix compilation of OHCI on x86.

Fixes were needed to go with the USB stack changes. Haven't tested it
yet because I don't have a a machine with OHCI, but I doubt there
will be any problems.

Get USB functional on x86-64.

- Various changes to printf format specifiers to make them 32/64 safe.
- Adjustments to various descriptor structures since addr_t is not 32-bit
on x86-64, and consequently this breaks the physical structure that's
expected. Also adjusted padding to accomodate 32-byte boundaries when
64-bit pointers are in use.
- Adjustments to make use of phys_addr_t where needed.

Move mutex initialization earlier, analogous to hrev43320.

Avoids crashing if the bus manager init fails. Also fixes CID 10846.

Use the right form of delete. Fixes CID 11037.

Account for full-/lowspeed errors and ensure an error is set.

For full- and lowspeed devices the unused ping status bit works as an
additional error bit (albeit it's so generic that it isn't really
useful). Include that bit when the error counter counts down to zero.

Also ensure that an error is returned if the error counter reaches
zero, but print a warning if none of the describing bits are set.

Make two unexpected/error cases more visible with debug output.


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@42965 a95241bf-73f2-0310-859d-f6bbb57e9c96

Read the previous descriptor pointer before freeing transfer and transfer descriptors.


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@42701 a95241bf-73f2-0310-859d-f6bbb57e9c96

* threshold higher bit means a full frame. Also use the threshold for micro frames instead of frames.
* introduces fNextStartingFrame to keep track of the next frame to use on next submit.
* set the IOC bit for the last ITD
* computes multiply field of the ITD based on the packed size (1, 2 or 3).
* use locking around linking of ITD
* free descriptors on errors
* on finishing, whenever a non success status is found, set actual length to zero
* on finishing, when copying data to the Transfer object, copy starting on packet boundaries (skipping unused bytes).


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@42679 a95241bf-73f2-0310-859d-f6bbb57e9c96

Locking around descriptors list handling isn't enough, locking is also needed when traversing the list: we instead lock the whole traversing/handling loop.


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@42623 a95241bf-73f2-0310-859d-f6bbb57e9c96

use locking when messing ITD and SITD queues. itd->prev was becoming NULL, thus leading to KDL.


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@42575 a95241bf-73f2-0310-859d-f6bbb57e9c96

We obviously need to mask off the interrupt threshold value before overwriting
it with our new setting. Otherwise we might end up with an illegal value as the
interrupt threshold which might inhibit interrupt generation alltogether
depending on the controller implementation. This was the case for the ATI/AMD
SBx00 chipsets. Therefore fixes #5551.


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@41690 a95241bf-73f2-0310-859d-f6bbb57e9c96

Introduce simplistic poll mode for EHCI. If enabled a thread will run the
interrupt handler every millisecond so it can check for interrupts to handle.
While this is certainly not ideal, it allows EHCI to run (surprisingly well
even) on chipsets that experience interrupt routing issues.
It can be enabled with the safemode setting "ehci_polling on" that can either
be entered using the new advanced debug option entry in the bootloader and/or
in the kernel settings file.


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@41688 a95241bf-73f2-0310-859d-f6bbb57e9c96

* added some EHCI definitions (specification 1.1)
* enable per-port change events when available


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@41661 a95241bf-73f2-0310-859d-f6bbb57e9c96

ehci usb driver:
* make sure the PortEnable bit is disabled (specification page 28)
* 250ms is too long: USB 2.0 specifies only 50ms. This prevented one of my devices to be detected.
+alpha


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@41658 a95241bf-73f2-0310-859d-f6bbb57e9c96

Check if PCI interrupts are disabled on the EHCI device. Rather unlikely to be
the case, but maybe there's some BIOSes that disable them? More or less poking
around for #5551, doesn't do any harm in any case.


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@41514 a95241bf-73f2-0310-859d-f6bbb57e9c96

Introduce workaround for the (documented) bug with ATI/AMD SB600 and early SB700
periodic list cache. This workaround is based on similar patches in NetBSD and
Linux and disables the (Advanced) Periodic List Cache on the affected devices
using ATI/AMD specifc registers (as documented in the corresponding AMD
register reference guide).
Also remove a ton of stray whitespace introduced by the isochronous patches.
Please take more care the next time.


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@41512 a95241bf-73f2-0310-859d-f6bbb57e9c96

* Implement inbound isochronous transfer for EHCI, based on UHCI implementation.
* changes next_log to their actual type instead of void*
* the field this_phy now includes the item type, it simplifies things.
* isochronous transfer descriptors are linked in the periodic frame list first,
and points to existing interrupt transfer descriptors.

This is still work in progress. Yet it's worth committing as it doesn't seem to
have impacts, and is required for the UVC SoC project. Tested basically with
usb_webcam to receive UVC stream headers.



git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@41424 a95241bf-73f2-0310-859d-f6bbb57e9c96

activate some traces for EHCI bus driver (see bug #4718)


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@41270 a95241bf-73f2-0310-859d-f6bbb57e9c96

added traces for unimplemented isochronous transfers for ehci and ohci


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@41194 a95241bf-73f2-0310-859d-f6bbb57e9c96

* Changed physical_entry::{address,size} to phys_{addr,size}_t and changed
map_physical_memory()'s physicalAddress parameter type from void* to
phys_addr_t. This breaks source compatibility, but -- as long as
phys_{addr,size}_t remain 32 bit wide -- keeps binary compatibility with
BeOS.
* Adjusted all code using the affected interfaces (Oh what fun!). Added a few
TODOs in places where the wrong types (e.g. void* for physical addresses
are used). Looks like quite a few drivers aren't 64 bit safe and others
will break with PAE.


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@36960 a95241bf-73f2-0310-859d-f6bbb57e9c96

Don't take ownership by writing the OS-owned semaphore when the BIOS-owned flag
isn't set. I'd tend to say that's a rather odd interpretation of the specified
protocol if a BIOS goes mad if the sem is written to while it doesn't actually
own the controller (it should just be a no-op in that case and directly set the
sem state). But it seems this is what others do as well (FreeBSD, Linux) so we
follow. Thanks to vegardw in #2083 for researching into this and proposing this
fix. Minor other cleanup.


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@35780 a95241bf-73f2-0310-859d-f6bbb57e9c96

We need to guarantee that when cancel returns that the callback won't be called
anymore. It could happen that the transfer was already in the process of being
finished, so wasn't in the list of pending transfers anymore. Cancel would then
return even though the callback wasn't called yet. This could lead to a callback
being called after a driver was already unloaded (even after it cleaned up the
pipes it used).


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@33239 a95241bf-73f2-0310-859d-f6bbb57e9c96

* Add the kernel debugger extensions only for the haiku target.
* Acknowledge non-enabled interrupts in UHCI as done in EHCI as well.


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@30063 a95241bf-73f2-0310-859d-f6bbb57e9c96

It turns out that on some BIOSes even if the handover is signaled as successful,
the BIOS doesn't remove it's SMIs. When then initializing the controller this
would trigger all sorts of stuff and for example freeze the system. We therefore
always force off the SMIs, even if the handover didn't fail.


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@29931 a95241bf-73f2-0310-859d-f6bbb57e9c96

* Check for and discard not enabled interrupts in EHCI.
* Simplify acknowledging interrupts a bit.

Might help with bug #2083 in case the controller would interrupt on a port
change or frame list rollover even though those aren't enabled by us.


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@29354 a95241bf-73f2-0310-859d-f6bbb57e9c96

* Rework the USB tracing mechanism. Cleaned it up and make it more convenient to
use. It will now print out the usb_ids of the objects that generate the trace
messages. These IDs are unique compared to the device address used previously,
because device addresses are per bus while usb_ids are global. This makes
trace output from devices across multiple controllers distinguishable.
* Some cleanup.


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@29002 a95241bf-73f2-0310-859d-f6bbb57e9c96

Output some more debug output in case of port reset failure.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@26546 a95241bf-73f2-0310-859d-f6bbb57e9c96

When calculating the actual length for requests, we don't want to include the
8 byte request header. It was included in the case of outgoing request
transfers before. In fact we always just want the length of the data phase,
which for non-requests just happens to be the whole transfer.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@26508 a95241bf-73f2-0310-859d-f6bbb57e9c96

If the BIOS doesn't give up ownership of an EHCI controller, simply clear the
BIOS owned flag and clear all possibly enabled SMIs. It seems to be common
practice to just ignore this error case, probably because there are enough
broken BIOSs out there that do not correctly hand over the controller.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@25918 a95241bf-73f2-0310-859d-f6bbb57e9c96

When canceling transfers, do not call the callbacks with the lock held. This
prevents deadlocks in cases where a new transfer is scheduled from within the
cancel callback. This is an edge case, as generally you don't want to schedule
anything when explicitly canceling transfers, but there are a valid use cases
when you cancel because of a timeout and then have to reset the device for
example. Note that this hides bug #2353, where the cancel case is probably
not handled correctly.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@25861 a95241bf-73f2-0310-859d-f6bbb57e9c96

Only check the explicitly enabled interrupts in EHCI. Otherwise we might steal
a shared interrupt from someone else.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@25822 a95241bf-73f2-0310-859d-f6bbb57e9c96

* Added optional spinlock contention measurement feature. Enabled when
B_DEBUG_SPINLOCK_CONTENTION is defined to 1. It typedefs spinlock to a
structure (thus breaking BeOS binary compatibility), containing a
counter which is incremented whenever a thread has to wait for the
spinlock.
* Added macros for spinlock initialization and access and changed
code using spinlocks accordingly. This breaks compilation for BeOS --
the macros should be defined in the respective compatibility wrappers.
* Added generic syscall to get the spinlock counters for the thread and
the team spinlocks.


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@25752 a95241bf-73f2-0310-859d-f6bbb57e9c96

* Increase the amount of tries in the EHCI BIOS handover so it lasts for up to
1 second instead of half a second. Might help with slow BIOSes.
* Disable EHCI interrupts if the BIOS handover fails so that we do not flood
the system with interrupts if the BIOS later on decides to still comply with
our handover request.
* Add an unconditional success message to EHCI and UHCI to better see if
controllers are present/initialized correctly at all.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@24980 a95241bf-73f2-0310-859d-f6bbb57e9c96

Exclude the frame list rollover interrupt from the interrupts we are interested
in. We do not enable this interrupt and if we still "handle" it in our
interrupt handler we might prevent the interrupt from reaching the real intended
handler in a shared interrupt setup. Not sure if this could have happend at
all, but this is more correct in any case.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@24965 a95241bf-73f2-0310-859d-f6bbb57e9c96

Fully clear the EHCI members too, so an incompletely constructed object can be
safely deleted.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@24957 a95241bf-73f2-0310-859d-f6bbb57e9c96

* Set the interrupt threshold of the controller when doing the rest of the
initialization as suggested by the EHCI spec and do not touch it again
once the controller is running. Doing so would result in undefined behaviour
according to EHCI specs.
* Set the port routing policy after the controller has started and not before
as suggested by the EHCI specs.
* Explicitly set the segment register after resetting the host controller and
not before, as this likely has no real effect ;-)

Might help with some legacy support or UHCI/OHCI to EHCI handover problems.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@24925 a95241bf-73f2-0310-859d-f6bbb57e9c96

Fix building usb busses with tracing on.


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@24880 a95241bf-73f2-0310-859d-f6bbb57e9c96

Change just host_controller_info to usb_host_controller_info for the sake of
clearity as suggested by Philippe Houdoin once and brought to my attention by
Slavatore. As this definition is completely private to the USB stack it would
not have caused any conflicts though.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@24828 a95241bf-73f2-0310-859d-f6bbb57e9c96

We are not supposed to overwrite the full capability register when setting our
OS-owned bit. Otherwise we overwrite exactly the bit we later have to wait for.
Could theoretically help on legacy support issues, but I doubt it since most
hardware probably does not even use EHCI legacy support as it's normally not
needed to support full/low-speed input devices.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@23747 a95241bf-73f2-0310-859d-f6bbb57e9c96

Unconditionally output when giving up port ownership and taking over host
controller ownership in EHCI to aid debugging USB legacy support issues.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@23735 a95241bf-73f2-0310-859d-f6bbb57e9c96

Notify the finisher thread after canceling transfers. It was possible that freeing the allocated
transfer resources would be delayed unnecessarily until some other transfer completed and woke up
the finisher thread. Found by Salvatore Benedetto, thanks!

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@23713 a95241bf-73f2-0310-859d-f6bbb57e9c96

* fixed memory leak: if InitKernelAccess failed data was not deleted



git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@23703 a95241bf-73f2-0310-859d-f6bbb57e9c96

* Add support for low/full speed interrupt transfers over USB 2.0 hubs
* Actually enable the periodic schedule so that any interrupts work over EHCI

For those who wait for OHCI: You can now attach your USB 1.1 mouse or keyboard
to a USB 2.0 hub and it should work.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@23382 a95241bf-73f2-0310-859d-f6bbb57e9c96

Messed up those numbers. The interval can range from 0 to 16 and we can use 0 to 10 with the 1024 frame list. So it's obviously 11 entries and not 10. Fixes the wrong tree layout and the doubled polling intervals of interrupts.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@23035 a95241bf-73f2-0310-859d-f6bbb57e9c96

Since the interval is now part of the general pipe info we don't have to cast the pipe to an interrupt one anymore.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@22976 a95241bf-73f2-0310-859d-f6bbb57e9c96

Implemented split transactions for control and bulk pipes in EHCI.
With this it is possible to use low/fullspeed (USB 1.1) devices below USB 2.0 Hubs.
Note though that you cannot yet use mice or keyboards attached to USB 2.0 Hubs as split transactions for interrupt pipes are still missing.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@22973 a95241bf-73f2-0310-859d-f6bbb57e9c96

* Removed the two extra submit functions and combined them directly into SubmitTransfer()
* The sparation based on async and periodic schedule made no sense as it is a question of queued/non-queued
* Therefore removed the near complete code duplication for adding interrupt transfers (that are also queued)

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@22943 a95241bf-73f2-0310-859d-f6bbb57e9c96

* Implement interrupt transfers in EHCI
* Uses a "collapsed binary tree" (for lack of a better name) to support the different intervals
* Remove a leftover variable declaration that was hiding error conditions away...
* Some cleanup

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@22942 a95241bf-73f2-0310-859d-f6bbb57e9c96

Completely redesign the USB explore process. Replaces the scary race conditions of the previous locking mechanism and simplifies handling of device changes by a more centralized approach.
Changes are now collected during explore and notifications as well as rescans are done at once. Through this a driver is also not rescanned multiple times anymore.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@22929 a95241bf-73f2-0310-859d-f6bbb57e9c96

* Apply more sane canceling logic to EHCI too
* Remove the leftover user_area - this is handeled inside the Transfer class by now

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@21906 a95241bf-73f2-0310-859d-f6bbb57e9c96

Revert r21231 as it's supposedly not possible. Someone implement OHCI then pleazzze!


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@21241 a95241bf-73f2-0310-859d-f6bbb57e9c96

Added a keep_port_ownership driver setting for EHCI, to allow keeping devices on EHCI instead of giving them up to the 1.0 controller.
Hopefully this will get things working here until OHCI is implemented.


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@21231 a95241bf-73f2-0310-859d-f6bbb57e9c96

* Reworked userland buffer access functionality into two functions in the Transfer class to not duplicate code in each HCD
* Use these functions in UHCI and EHCI which greatly reduces overhead on fragmented transfers as the vector conversion and area cloning is only done once
* Use the PrepareKernelAccess() function before submitting fragments which fixes the crash when trying to access the next fragments buffer in user memory from the kernel thread

This should fix bug #1167 which could be encountered when sending data to a mass storage device with an application that uses a large enough transfer buffer size.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@21059 a95241bf-73f2-0310-859d-f6bbb57e9c96

Fixed wrong cancelling algorithm. This is what you get when you "rework" an existing function without thinking it through again... Thanks J Jerome for noticing\!

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@21058 a95241bf-73f2-0310-859d-f6bbb57e9c96

Implemented transfer cancelling for EHCI by reworking the previously unused CancelPendingTransfer() into CancelQueuedTransfers() with the pipe based semantic.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@21044 a95241bf-73f2-0310-859d-f6bbb57e9c96

Implemented fragmented USB transfers. The transfer length will max out at a certain point to not overflow the allocator. The fragmented transfers are resubmitted until all fragments are sent / received.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@20416 a95241bf-73f2-0310-859d-f6bbb57e9c96

* Fixed spinlock usage in EHCI and UHCI drivers
* Fixed UHCI to allow for empty transfer buffers
* Extended the check for still linked transfer descriptors when removing a chain

There remain UHCI bugs that will cause transfers outgoing transfers to fail. I'll have to review / rewrite the UHCI driver.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@19730 a95241bf-73f2-0310-859d-f6bbb57e9c96

* Implemented a notification method for BusManagers so that they can be notified of pipe changes (creation, destruction, changed settings).
This is necessary in OHCI and will probably be used in EHCI also to keep one endpoint construct for each pipe open instead on creating and deleting it for each transfer.
* Pseudo implemented set_pipe_policy for isochronous pipes that makes use of the new notification system.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@18945 a95241bf-73f2-0310-859d-f6bbb57e9c96

* Added retries to the PhysicalMemoryAllocator. Memory is allocated and freed at a very high frequency, so low memory conditions shouldn't last very long.
* Added a separate thread for cleaning up EHCI transfers. This makes it possible to actually always wait for the async advance interrupt without hindering execution of transfers.
This pushes performance again and fixes the bug I introduced in the last change, that we could free yet cached descriptors when a previous async advance timed out.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@18929 a95241bf-73f2-0310-859d-f6bbb57e9c96

* Switched from the old v2 callback status codes to the new v3 codes
This means that the usb_callback_func now takes a status_t instead of a uint32 status.
Also the error codes are now different. I don't see this as a real problem in binary compatibility, as the status codes were never really documented and most drivers just assumed that a nonzero status meant an error.
Source compatibility breaks for callback functions and error defines. I fixed (hopefully) all places in the tree that are currenty included in the image and affected by the change.
* Corrected error reporting in UHCI and EHCI using the new status codes.
* Fixed a memory leak in EHCI where the async advance interrupt was not triggered in time.
* Fixed another race condition in usb_raw that could cause a crash when a device is removed while a transfer is pending.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@18916 a95241bf-73f2-0310-859d-f6bbb57e9c96

* Fixing my previous attempt at fixing cookie handling. Device addresses cannot be used, since there may be devices with the same address across different busses. A linked list of usb_driver_cookies is maintained now that relates devices and cookies.
* Fixed endpoint halt clearing. The data toggle has to be reset when we clear a stall.
* Now using up to B_PAGE_SIZE * 4 sized buffers in EHCI. They are split automatically by the host controller. This further pushes throughput.
* Optimize memory allocator settings for the above case. Enlarging it's managed space to 1MB
* Ignore the transaction error bit in EHCI. It is only of informational character as it also idicates errors that were recovered automatically. If an unrecoverable error occures, other bits will indicate this.
* Fixed some more print formats in debug output.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@18915 a95241bf-73f2-0310-859d-f6bbb57e9c96

* Implemented compatibility workaround for BeOS to be able to write user memory from the kernel finisher thread

This fixes crashing of the USB Mass Storage driver under R5. It is fully working there now.
I hope to be able to do this more cleanly under Haiku though.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@18894 a95241bf-73f2-0310-859d-f6bbb57e9c96

* Implemented usable thread synchronisation for UHCI and EHCI.
* Cleaned interrupt handlers for both busses, too.
* Fixed memory leak of the queue head stray descriptors in EHCI.
* Fixed warnings about assigning NULL to the module flags field.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@18857 a95241bf-73f2-0310-859d-f6bbb57e9c96

* Implemented transfer finishing in EHCI
* Implemented actual length calculation in EHCI
* Implemented bulk transfers in EHCI
* Implemented queue freeing in EHCI
* Added extended buffer pointers to the EHCI structs, necessary when the controller supports 64bit
* Fixed debug output and removed warnings due to wrong print format

Control and bulk transfers are now working nicely for EHCI. This means that I can now get pictures from my camera with USB 2.0 speed. Note though that many things still are unimplemented (isochronous transfers, split transactions, interrupt transfers, queue chaches per pipe, ...). But it's certainly a nice step forward :-).

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@18855 a95241bf-73f2-0310-859d-f6bbb57e9c96

* Implemented most qTD and qH functions for EHCI async transfers.
* Implementing control and bulk (async) transfers. They don't work yet though.
* Corrected the legacy support handling in EHCI
* Fixed the bit mask for total bytes in qTDs and added the data toggle bit
* Changed the boolean low speed flags to usb_speed enum that allows for low, full and highspeed

Also wrote a memory allocator that takes over memory management from the stack. It is similar to a buddy allocator. This does on the one hand remove some ugly code from the stack and is on the other hand far more scalable than the previous free list approach. This scalability is needed for proper EHCI support where buffers can go up to a size of 5 pages. The allocator is not perfect and it's memory overhead is quite high compared to the managed memory (depends on the managed block sizes), but it's quik and works well so far.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@18849 a95241bf-73f2-0310-859d-f6bbb57e9c96

* Rewrote some stack internals (again). Everything is now created using a parent object that shares the BusManager and Stack information to reduce individual constructors.
* Don't inherit Device and Interface from pipes anymore. The device has a default pipe member now.
* Since the parent of each Object is now known, we can send control messages using the default pipe of the device from all attached elements (pipes and interfaces).
* Fixed the generic {Set|Clear}Feature() and GetStatus() calls to include the endpoint address for pipe specific requests.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@18710 a95241bf-73f2-0310-859d-f6bbb57e9c96

replaced with PCI.h definitions, added PCI_usb_ehci


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@18704 a95241bf-73f2-0310-859d-f6bbb57e9c96

* Redisigned some of the USB Stack. Unified constructors, removed redundant functions and moved some other functions around
* Implemented the EHCIRootHub
* Reworked UHCIRootHub to be more of a utility class and moved the actual work into the UHCI BusManager
* Implemented port reset in EHCI. Ports with low and fullspeed devices are now correctly handed over to the companion host controller.

Attachs and detachs as well as port resetting / enabling works now in EHCI. Only the actual data transfers are missing thereafter ;-)

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@18656 a95241bf-73f2-0310-859d-f6bbb57e9c96

* Changed EHCI register access from port to memory mapped io (which is required)
* Added individual register operations for operational and capability registers
* Added resetting the host controller which appearantly actually works

Note that you shouldn't install the ehci module if you want uhci to work.
It disables the companion host controller drivers (uhci and ohci) because it takes port ownership and does not yet give it back for low and fullspeed devices.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@18648 a95241bf-73f2-0310-859d-f6bbb57e9c96

* Adding skeleton EHCI driver. Some parts filled from the specs and the UHCI driver.
* Corrected some errors I found while reading UHCI.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@18626 a95241bf-73f2-0310-859d-f6bbb57e9c96