linux-stable.git/drivers/usb/host, branch linux-2.6.34.y Linux kernel stable tree USB: EHCI: go back to using the system clock for QH unlinks 2014-02-10T21:11:19+00:00 Alan Stern stern@rowland.harvard.edu 2011-07-05T16:34:05+00:00 7947fee39844605ece99d8a4f0be8de93513bae2 commit 004c19682884d4f40000ce1ded53f4a1d0b18206 upstream. This patch (as1477) fixes a problem affecting a few types of EHCI controller. Contrary to what one might expect, these controllers automatically stop their internal frame counter when no ports are enabled. Since ehci-hcd currently relies on the frame counter for determining when it should unlink QHs from the async schedule, those controllers run into trouble: The frame counter stops and the QHs never get unlinked. Some systems have also experienced other problems traced back to commit b963801164618e25fbdc0cd452ce49c3628b46c8 (USB: ehci-hcd unlink speedups), which made the original switch from using the system clock to using the frame counter. It never became clear what the reason was for these problems, but evidently it is related to use of the frame counter. To fix all these problems, this patch more or less reverts that commit and goes back to using the system clock. But this can't be done cleanly because other changes have since been made to the scan_async() subroutine. One of these changes involved the tricky logic that tries to avoid rescanning QHs that have already been seen when the scanning loop is restarted, which happens whenever an URB is given back. Switching back to clock-based unlinks would make this logic even more complicated. Therefore the new code doesn't rescan the entire async list whenever a giveback occurs. Instead it rescans only the current QH and continues on from there. This requires the use of a separate pointer to keep track of the next QH to scan, since the current QH may be unlinked while the scanning is in progress. That new pointer must be global, so that it can be adjusted forward whenever the _next_ QH gets unlinked. (uhci-hcd uses this same trick.) Simplification of the scanning loop removes a level of indentation, which accounts for the size of the patch. The amount of code changed is relatively small, and it isn't exactly a reversion of the b963801164 commit. This fixes Bugzilla #32432. Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Tested-by: Matej Kenda <matejken@gmail.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de> Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com> 
commit 004c19682884d4f40000ce1ded53f4a1d0b18206 upstream.

This patch (as1477) fixes a problem affecting a few types of EHCI
controller.  Contrary to what one might expect, these controllers
automatically stop their internal frame counter when no ports are
enabled.  Since ehci-hcd currently relies on the frame counter for
determining when it should unlink QHs from the async schedule, those
controllers run into trouble: The frame counter stops and the QHs
never get unlinked.

Some systems have also experienced other problems traced back to
commit b963801164618e25fbdc0cd452ce49c3628b46c8 (USB: ehci-hcd unlink
speedups), which made the original switch from using the system clock
to using the frame counter.  It never became clear what the reason was
for these problems, but evidently it is related to use of the frame
counter.

To fix all these problems, this patch more or less reverts that commit
and goes back to using the system clock.  But this can't be done
cleanly because other changes have since been made to the scan_async()
subroutine.  One of these changes involved the tricky logic that tries
to avoid rescanning QHs that have already been seen when the scanning
loop is restarted, which happens whenever an URB is given back.
Switching back to clock-based unlinks would make this logic even more
complicated.

Therefore the new code doesn't rescan the entire async list whenever a
giveback occurs.  Instead it rescans only the current QH and continues
on from there.  This requires the use of a separate pointer to keep
track of the next QH to scan, since the current QH may be unlinked
while the scanning is in progress.  That new pointer must be global,
so that it can be adjusted forward whenever the _next_ QH gets
unlinked.  (uhci-hcd uses this same trick.)

Simplification of the scanning loop removes a level of indentation,
which accounts for the size of the patch.  The amount of code changed
is relatively small, and it isn't exactly a reversion of the
b963801164 commit.

This fixes Bugzilla #32432.

Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Tested-by: Matej Kenda <matejken@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
xhci: Don't write zeroed pointers to xHC registers. 2014-02-10T21:11:19+00:00 Sarah Sharp sarah.a.sharp@linux.intel.com 2012-03-16T20:09:39+00:00 2b46f2b71df80632b1f5295fc564a5386a0d9b70 commit 159e1fcc9a60fc7daba23ee8fcdb99799de3fe84 upstream. When xhci_mem_cleanup() is called, we can't be sure if the xHC is actually halted. We can ask the xHC to halt by writing to the RUN bit in the command register, but that might timeout due to a HW hang. If the host controller is still running, we should not write zeroed values to the event ring dequeue pointers or base tables, the DCBAA pointers, or the command ring pointers. Eric Fu reports his VIA VL800 host accesses the event ring pointers after a failed register restore on resume from suspend. The hypothesis is that the host never actually halted before the register write to change the event ring pointer to zero. Remove all writes of zeroed values to pointer registers in xhci_mem_cleanup(). Instead, make all callers of the function reset the host controller first, which will reset those registers to zero. xhci_mem_init() is the only caller that doesn't first halt and reset the host controller before calling xhci_mem_cleanup(). This should be backported to kernels as old as 2.6.32. Signed-off-by: Sarah Sharp <sarah.a.sharp@linux.intel.com> Tested-by: Elric Fu <elricfu1@gmail.com> Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com> 
commit 159e1fcc9a60fc7daba23ee8fcdb99799de3fe84 upstream.

When xhci_mem_cleanup() is called, we can't be sure if the xHC is
actually halted.  We can ask the xHC to halt by writing to the RUN bit
in the command register, but that might timeout due to a HW hang.

If the host controller is still running, we should not write zeroed
values to the event ring dequeue pointers or base tables, the DCBAA
pointers, or the command ring pointers.  Eric Fu reports his VIA VL800
host accesses the event ring pointers after a failed register restore on
resume from suspend.  The hypothesis is that the host never actually
halted before the register write to change the event ring pointer to
zero.

Remove all writes of zeroed values to pointer registers in
xhci_mem_cleanup().  Instead, make all callers of the function reset the
host controller first, which will reset those registers to zero.
xhci_mem_init() is the only caller that doesn't first halt and reset the
host controller before calling xhci_mem_cleanup().

This should be backported to kernels as old as 2.6.32.

Signed-off-by: Sarah Sharp <sarah.a.sharp@linux.intel.com>
Tested-by: Elric Fu <elricfu1@gmail.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
xHCI: Correct the #define XHCI_LEGACY_DISABLE_SMI 2014-02-10T21:11:19+00:00 Alex He alex.he@amd.com 2012-03-30T02:21:38+00:00 617734b48a50fbcb8b5a9eba56686570d7a4bb6c commit 95018a53f7653e791bba1f54c8d75d9cb700d1bd upstream. Re-define XHCI_LEGACY_DISABLE_SMI and used it in right way. All SMI enable bits will be cleared to zero and flag bits 29:31 are also cleared to zero. Other bits should be presvered as Table 146. This patch should be backported to kernels as old as 2.6.31. Signed-off-by: Alex He <alex.he@amd.com> Signed-off-by: Sarah Sharp <sarah.a.sharp@linux.intel.com> Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com> 
commit 95018a53f7653e791bba1f54c8d75d9cb700d1bd upstream.

Re-define XHCI_LEGACY_DISABLE_SMI and used it in right way. All SMI enable
bits will be cleared to zero and flag bits 29:31 are also cleared to zero.
Other bits should be presvered as Table 146.

This patch should be backported to kernels as old as 2.6.31.

Signed-off-by: Alex He <alex.he@amd.com>
Signed-off-by: Sarah Sharp <sarah.a.sharp@linux.intel.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
xhci: Reset reserved command ring TRBs on cleanup. 2014-02-10T21:11:19+00:00 Sarah Sharp sarah.a.sharp@linux.intel.com 2012-05-08T14:09:26+00:00 a4ad381723f28cef4b50f302801d9081e83f6ede commit 33b2831ac870d50cc8e01c317b07fb1e69c13fe1 upstream. When the xHCI driver needs to clean up memory (perhaps due to a failed register restore on resume from S3 or resume from S4), it needs to reset the number of reserved TRBs on the command ring to zero. Otherwise, several resume cycles (about 30) with a UAS device attached will continually increment the number of reserved TRBs, until all command submissions fail because there isn't enough room on the command ring. This patch should be backported to kernels as old as 2.6.32, that contain the commit 913a8a344ffcaf0b4a586d6662a2c66a7106557d "USB: xhci: Change how xHCI commands are handled." Signed-off-by: Sarah Sharp <sarah.a.sharp@linux.intel.com> Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com> 
commit 33b2831ac870d50cc8e01c317b07fb1e69c13fe1 upstream.

When the xHCI driver needs to clean up memory (perhaps due to a failed
register restore on resume from S3 or resume from S4), it needs to reset
the number of reserved TRBs on the command ring to zero.  Otherwise,
several resume cycles (about 30) with a UAS device attached will
continually increment the number of reserved TRBs, until all command
submissions fail because there isn't enough room on the command ring.

This patch should be backported to kernels as old as 2.6.32,
that contain the commit 913a8a344ffcaf0b4a586d6662a2c66a7106557d
"USB: xhci: Change how xHCI commands are handled."

Signed-off-by: Sarah Sharp <sarah.a.sharp@linux.intel.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
xhci: Increase reset timeout for Renesas 720201 host. 2014-02-10T21:11:18+00:00 Sarah Sharp sarah.a.sharp@linux.intel.com 2012-07-23T23:06:08+00:00 4b95b8ee9bb1051c0f81922f970ed1e0977ed9b9 commit 22ceac191211cf6688b1bf6ecd93c8b6bf80ed9b upstream. The NEC/Renesas 720201 xHCI host controller does not complete its reset within 250 milliseconds. In fact, it takes about 9 seconds to reset the host controller, and 1 second for the host to be ready for doorbell rings. Extend the reset and CNR polling timeout to 10 seconds each. This patch should be backported to kernels as old as 2.6.31, that contain the commit 66d4eadd8d067269ea8fead1a50fe87c2979a80d "USB: xhci: BIOS handoff and HW initialization." Signed-off-by: Sarah Sharp <sarah.a.sharp@linux.intel.com> Reported-by: Edwin Klein Mentink <e.kleinmentink@zonnet.nl> Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com> 
commit 22ceac191211cf6688b1bf6ecd93c8b6bf80ed9b upstream.

The NEC/Renesas 720201 xHCI host controller does not complete its reset
within 250 milliseconds.  In fact, it takes about 9 seconds to reset the
host controller, and 1 second for the host to be ready for doorbell
rings.  Extend the reset and CNR polling timeout to 10 seconds each.

This patch should be backported to kernels as old as 2.6.31, that
contain the commit 66d4eadd8d067269ea8fead1a50fe87c2979a80d "USB: xhci:
BIOS handoff and HW initialization."

Signed-off-by: Sarah Sharp <sarah.a.sharp@linux.intel.com>
Reported-by: Edwin Klein Mentink <e.kleinmentink@zonnet.nl>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
xhci: Make handover code more robust 2014-02-10T21:11:18+00:00 Matthew Garrett mjg@redhat.com 2012-08-14T20:44:49+00:00 8bcc9b593131db101cf21e09c9711f239e3008ef commit e955a1cd086de4d165ae0f4c7be7289d84b63bdc upstream. My test platform (Intel DX79SI) boots reliably under BIOS, but frequently crashes when booting via UEFI. I finally tracked this down to the xhci handoff code. It seems that reads from the device occasionally just return 0xff, resulting in xhci_find_next_cap_offset generating a value that's larger than the resource region. We then oops when attempting to read the value. Sanity checking that value lets us avoid the crash. I've no idea what's causing the underlying problem, and xhci still doesn't actually *work* even with this, but the machine at least boots which will probably make further debugging easier. This should be backported to kernels as old as 2.6.31, that contain the commit 66d4eadd8d067269ea8fead1a50fe87c2979a80d "USB: xhci: BIOS handoff and HW initialization." Signed-off-by: Matthew Garrett <mjg@redhat.com> Signed-off-by: Sarah Sharp <sarah.a.sharp@linux.intel.com> Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com> 
commit e955a1cd086de4d165ae0f4c7be7289d84b63bdc upstream.

My test platform (Intel DX79SI) boots reliably under BIOS, but frequently
crashes when booting via UEFI. I finally tracked this down to the xhci
handoff code. It seems that reads from the device occasionally just return
0xff, resulting in xhci_find_next_cap_offset generating a value that's
larger than the resource region. We then oops when attempting to read the
value. Sanity checking that value lets us avoid the crash.

I've no idea what's causing the underlying problem, and xhci still doesn't
actually *work* even with this, but the machine at least boots which will
probably make further debugging easier.

This should be backported to kernels as old as 2.6.31, that contain the
commit 66d4eadd8d067269ea8fead1a50fe87c2979a80d "USB: xhci: BIOS handoff
and HW initialization."

Signed-off-by: Matthew Garrett <mjg@redhat.com>
Signed-off-by: Sarah Sharp <sarah.a.sharp@linux.intel.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
USB: whci-hcd: fix endian conversion in qset_clear() 2013-01-16T21:45:00+00:00 Dan Carpenter dan.carpenter@oracle.com 2011-11-22T07:28:31+00:00 6066bdfc7f12a131e8662c989cf366ead803a8ed commit 8746c83d538cab273d335acb2be226d096f4a5af upstream. qset->qh.link is an __le64 field and we should be using cpu_to_le64() to fill it. Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de> Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com> 
commit 8746c83d538cab273d335acb2be226d096f4a5af upstream.

qset->qh.link is an __le64 field and we should be using cpu_to_le64()
to fill it.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
QE/FHCI: fixed the CONTROL bug 2012-08-17T19:35:32+00:00 Jerry Huang r66093@freescale.com 2011-10-18T05:09:48+00:00 f47b301ecc6ff8446524548b93b7675252a62d96 commit 273d23574f9dacd9c63c80e7d63639a669aad441 upstream. For USB CONTROL transaction, when the data length is zero, the IN package is needed to finish this transaction in status stage. Signed-off-by: Jerry Huang <r66093@freescale.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de> Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com> 
commit 273d23574f9dacd9c63c80e7d63639a669aad441 upstream.

For USB CONTROL transaction, when the data length is zero,
the IN package is needed to finish this transaction in status stage.

Signed-off-by: Jerry Huang <r66093@freescale.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
xhci-mem.c: Check for ring->first_seg != NULL 2012-08-17T19:35:25+00:00 Kautuk Consul consul.kautuk@gmail.com 2011-09-19T23:53:12+00:00 95e8d5a47d4b06822dbdb6eef332b4e3e05d2a12 commit 0e6c7f746ea99089fb3263709075c20485a479ae upstream. There are 2 situations wherein the xhci_ring* might not get freed: - When xhci_ring_alloc() -> xhci_segment_alloc() returns NULL and we goto the fail: label in xhci_ring_alloc. In this case, the ring will not get kfreed. - When the num_segs argument to xhci_ring_alloc is passed as 0 and we try to free the rung after that. ( This doesn't really happen as of now in the code but we seem to be entertaining num_segs=0 in xhci_ring_alloc ) This should be backported to kernels as old as 2.6.31. Signed-off-by: Kautuk Consul <consul.kautuk@gmail.com> Signed-off-by: Sarah Sharp <sarah.a.sharp@linux.intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de> Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com> 
commit 0e6c7f746ea99089fb3263709075c20485a479ae upstream.

There are 2 situations wherein the xhci_ring* might not get freed:
- When xhci_ring_alloc() -> xhci_segment_alloc() returns NULL and
  we goto the fail: label in xhci_ring_alloc. In this case, the ring
  will not get kfreed.
- When the num_segs argument to xhci_ring_alloc is passed as 0 and
  we try to free the rung after that.
  ( This doesn't really happen as of now in the code but we seem to
    be entertaining num_segs=0 in xhci_ring_alloc )

This should be backported to kernels as old as 2.6.31.

Signed-off-by: Kautuk Consul <consul.kautuk@gmail.com>
Signed-off-by: Sarah Sharp <sarah.a.sharp@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
USB: EHCI: Do not rely on PORT_SUSPEND to stop USB resuming in ehci_bus_resume(). 2012-08-17T19:35:07+00:00 Wang Zhi zhi.wang@windriver.com 2011-08-17T02:39:31+00:00 ea855e30bc0e5064f021a4bb546af6be404d729d commit d0f2fb2500b1c5fe4967eb45d8c9bc758d7aef80 upstream. From EHCI Spec p.28 HC should clear PORT_SUSPEND when SW clears PORT_RESUME. In Intel Oaktrail platform, MPH (Multi-Port Host Controller) core clears PORT_SUSPEND directly when SW sets PORT_RESUME bit. If we rely on PORT_SUSPEND bit to stop USB resume, we will miss the action of clearing PORT_RESUME. This will cause unexpected long resume signal on USB bus. Signed-off-by: Wang Zhi <zhi.wang@windriver.com> Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de> Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com> 
commit d0f2fb2500b1c5fe4967eb45d8c9bc758d7aef80 upstream.

From EHCI Spec p.28 HC should clear PORT_SUSPEND when SW clears
PORT_RESUME. In Intel Oaktrail platform, MPH (Multi-Port Host
Controller) core clears PORT_SUSPEND directly when SW sets PORT_RESUME
bit. If we rely on PORT_SUSPEND bit to stop USB resume, we will miss
the action of clearing PORT_RESUME. This will cause unexpected long
resume signal on USB bus.

Signed-off-by: Wang Zhi <zhi.wang@windriver.com>
Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>