Michael Reutman reports that an AMD/ATI EHCI host controller on one of
his computers does not stop transferring data when an active bulk QH
is unlinked from the async schedule.  Apparently that host controller
fails to implement the IAA mechanism correctly when an active QH is
unlinked.  This leads to data corruption, because the controller
continues to update the QH in memory when the driver doesn't expect
it.  As a result, the next URB submitted for that QH can hang, because
the link pointers for the TD queue have been messed up.  This
misbehavior is observed quite regularly.

To be fair, the EHCI spec (section 4.8.2) says that active QHs should
not be unlinked.  It goes on to recommend a procedure that involves
waiting for the QH to go inactive before unlinking it.  In the real
world this is impractical, not least because the QH may _never_ go
inactive.  (What were they thinking?)  Sometimes we have no choice but
to unlink an active QH.

In an attempt to avoid the problems that can ensue, this patch changes
how the driver decides when the unlink is complete.  In addition to
waiting through two IAA cycles, in cases where the QH was not known to
be inactive beforehand we now wait until a 2-ms period has elapsed
with the host controller making no change to the QH data structure
(the hw_current and hw_token fields in particular).  The intuition
here is that after such a long period, the endpoint must be NAKing and
hopefully the QH has been dropped from the host controller's internal
cache.  There's no way to know if this reasoning is really valid --
the spec is no help in this regard -- but at least this approach fixes
Michael's problem.

The test for whether the QH is already known to be inactive involves
the reason for unlinking the QH originally.  If it was unlinked
because it had halted, or it stopped in response to a short read, or
it overlaid a dummy TD (a silicon bug), then it certainly is inactive.
If it was unlinked because the TD queue was empty and no TDs have been
added to the queue in the meantime, then it must be inactive.  Or if
the hardware status indicates that the QH is currently halted (even if
that wasn't the reason for unlinking it), then it is inactive.
Otherwise, if none of those checks apply, we go through the 2-ms
delay.

Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Reported-by: Michael Reutman <mreutman@epiqsolutions.com>
Tested-by: Michael Reutman <mreutman@epiqsolutions.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

This patch improves the way ehci-hcd handles the iaa_in_progress flag.
The current code is somewhat careless in this regard:

	The flag is meaningless when the root hub isn't running, most
	particularly after the root hub has been suspended.  But in
	start_iaa_cycle(), the driver checks the flag before checking
	the root hub's state.  They should be checked in the opposite
	order.

	That routine also sets the flag too early, before it has
	definitely committed to starting an IAA cycle.

	The flag is turned off in end_unlink_async().  Upcoming
	changes will call that routine at other times, not just at the
	end of an IAA cycle.  The two actions are logically separate
	(although related), so we separate out a new routine to be
	called in place of end_unlink_async() whenever an IAA cycle
	ends: end_iaa_cycle().

	iaa_in_progress should be turned off when the root hub is
	suspended -- we certainly don't want it still to be set when
	the root hub resumes.  Therefore the call to
	end_unlink_async() in ehci_bus_suspend() should also be
	replaced with a call to end_iaa_cycle().

Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

This patch replaces the "exception" bitflag in the ehci_qh structure
with a more explicit "unlink_reason" bitmask.  This is for use in the
following patch, where we will need to have a good idea of the
reason for unlinking a QH, not just "something exceptional happened".

Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Tested-by: Michael Reutman <mreutman@epiqsolutions.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ehci-hcd currently unlinks an interrupt QH when it becomes empty, that
is, after its last URB completes.  This works well because in almost
all cases, the completion handler for an interrupt URB resubmits the
URB; therefore the QH doesn't become empty and doesn't get unlinked.

When we start using tasklets for URB completion, this scheme won't work
as well.  The resubmission won't occur until the tasklet runs, which
will be some time after the completion is queued with the tasklet.
During that delay, the QH will be empty and so will be unlinked
unnecessarily.

To prevent this problem, this patch adds a 5-ms time delay before empty
interrupt QHs are unlinked.  Most often, during that time the interrupt
URB will be resubmitted and thus we can avoid unlinking the QH.

Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Ming Lei <ming.lei@canonical.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

This patch (as1665) changes the way ehci-hcd's end_unlink_async()
routine works in order to avoid recursive execution and to be more
efficient:

	Now when an IAA cycle ends, a new one gets started up right
	away (if it is needed) instead of waiting until the
	just-unlinked QH has been processed.

	The async_iaa list is renamed to async_idle, which better
	expresses its new purpose: It is now the list of QHs which are
	now completely idle and are waiting to be processed by
	end_unlink_async().

	A new flag is added to track whether an IAA cycle is in
	progress, because the list formerly known as async_iaa no
	longer stores the QHs waiting for the IAA to finish.

	The decision about how many QHs to process when an IAA cycle
	ends is now made at the end of the cycle, when we know the
	current state of the hardware, rather than at the beginning.
	This means a bunch of logic got moved from start_iaa_cycle()
	to end_unlink_async().

Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

This patch (as1664) converts ehci-hcd's async_unlink, async_iaa, and
intr_unlink from singly-linked lists to standard doubly-linked
list_heads.  Originally it didn't seem necessary to use list_heads,
because items are always added to and removed from these lists in FIFO
order.  But now with more list processing going on, it's easier to use
the standard routines than continue with a roll-your-own approach.

I don't know if the code ends up being notably shorter, but the
patterns will be more familiar to any kernel hacker.

Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

This patch (as1671) fixes up an incorrect resolution of a merge
conflict between Greg KH's usb-linus branch and his usb-next branch.

Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

This is to pick up the fixes in that branch, and let Alan fix the merge
error in drivers/usb/host/ehci-timer.c better than I just did (as I know
I messed it up...)

Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

This patch (as1670) fixes a regression caused by commit
6402c796d3b4205d3d7296157956c5100a05d7d6 (USB: EHCI: work around
silicon bug in Intel's EHCI controllers).  The workaround goes through
two IAA cycles for each QH being unlinked.  During the first cycle,
the QH is not added to the async_iaa list (because it isn't fully gone
from the hardware yet), which means that list will be empty.

Unfortunately, I forgot to update the IAA watchdog timer routine.  It
thinks that an empty async_iaa list means the timer expiration was an
error, which isn't true any more.  This problem didn't show up during
initial testing because the controllers being tested all had working
IAA interrupts.  But not all controllers do, and when the watchdog
timer expires, the empty-list check prevents the second IAA cycle from
starting.  As a result, URB unlinks never complete.  The check needs
to be removed.

Among the symptoms of the regression are processes stuck in D wait
states and hangs during system shutdown.

Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Reported-and-tested-by: Stephen Warren <swarren@wwwdotorg.org>
Reported-and-tested-by: Sven Joachim <svenjoac@gmx.de>
Reported-by: Andreas Bombe <aeb@debian.org>
Cc: stable <stable@vger.kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

This patch (as1635) rearranges the control-flow logic in
ehci_iaa_watchdog() slightly to agree better with the comments.  It
also changes a verbose-debug message to a regular debug message.
Expiration of the IAA watchdog is an unusual event and can lead to
problems; we need to know about it if it happens during debugging.  It
should not be necessary to set a "verbose" compilation option.

No behavioral changes other than the debug message.  Lots of apparent
changes to the source text, though, because the indentation level was
decreased.

Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>