karo-tx-linux.git
8 years agousb: hub_handle_remote_wakeup() only exists for CONFIG_PM=y
Stephen Rothwell [Thu, 29 May 2014 08:55:06 +0000 (18:55 +1000)]
usb: hub_handle_remote_wakeup() only exists for CONFIG_PM=y

Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Acked-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agoUSB: orinoco_usb: remove CONFIG_USB_DEBUG support
Greg Kroah-Hartman [Wed, 28 May 2014 18:34:04 +0000 (11:34 -0700)]
USB: orinoco_usb: remove CONFIG_USB_DEBUG support

Now that the config option is gone, the driver needs to be converted
over to using the dynamic_debug infrastructure, like all other USB
drivers have been already.

Cc: Christian Lamparter <chunkeey@googlemail.com>
Cc: Larry Finger <Larry.Finger@lwfinger.net>
Cc: Gertjan van Wingerde <gwingerde@gmail.com>
Cc: Alexey Khoroshilov <khoroshilov@ispras.ru>
Cc: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Acked-by: John W. Linville <linville@tuxdriver.com>
8 years agoUSB: media: lirc: igorplugusb: remove CONFIG_USB_DEBUG support
Greg Kroah-Hartman [Wed, 28 May 2014 18:34:03 +0000 (11:34 -0700)]
USB: media: lirc: igorplugusb: remove CONFIG_USB_DEBUG support

The config option is gone, so port the driver over to using the dynamic
debug infrastructure instead, like all other USB drivers.

Cc: Jarod Wilson <jarod@wilsonet.com>
Cc: Mauro Carvalho Chehab <m.chehab@samsung.com>
Cc: Dan Carpenter <dan.carpenter@oracle.com>
Cc: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Cc: Andreea-Cristina Bernat <bernat.ada@gmail.com>
Cc: Matina Maria Trompouki <mtrompou@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agoUSB: media: streamzap: remove CONFIG_USB_DEBUG
Greg Kroah-Hartman [Wed, 28 May 2014 18:34:02 +0000 (11:34 -0700)]
USB: media: streamzap: remove CONFIG_USB_DEBUG

It wasn't even being used for anything, so remove the test for the
config option (as the config option isn't present anymore), and remove
the unused module parameter "debug" so as to not confuse anyone.

Cc: Mauro Carvalho Chehab <m.chehab@samsung.com>
Cc: James Hogan <james.hogan@imgtec.com>
Cc: "Antti Seppälä" <a.seppala@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agoUSB: media: redrat3: remove CONFIG_USB_DEBUG usage
Greg Kroah-Hartman [Wed, 28 May 2014 18:34:01 +0000 (11:34 -0700)]
USB: media: redrat3: remove CONFIG_USB_DEBUG usage

Use the dev_dbg dynamic infrastructure instead of rolling our own custom
debug logic.  This is needed as CONFIG_USB_DEBUG is now gone.

Cc: Mauro Carvalho Chehab <m.chehab@samsung.com>
Cc: Sean Young <sean@mess.org>
Cc: Hans Verkuil <hans.verkuil@cisco.com>
Cc: Jonathan McCrohan <jmccrohan@gmail.com>
Cc: James Hogan <james.hogan@imgtec.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agoUSB: media: redrat3: remove unneeded tracing macro
Greg Kroah-Hartman [Wed, 28 May 2014 18:34:00 +0000 (11:34 -0700)]
USB: media: redrat3: remove unneeded tracing macro

The rr3_ftr function duplicates the in-kernel tracing logic, so remove
it.

This is step 1 of of removing the CONFIG_USB_DEBUG depandancy in this
driver.

Cc: Mauro Carvalho Chehab <m.chehab@samsung.com>
Cc: Sean Young <sean@mess.org>
Cc: Hans Verkuil <hans.verkuil@cisco.com>
Cc: Jonathan McCrohan <jmccrohan@gmail.com>
Cc: James Hogan <james.hogan@imgtec.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agousb: qcserial: add additional Sierra Wireless QMI devices
Aleksander Morgado [Thu, 29 May 2014 11:33:27 +0000 (13:33 +0200)]
usb: qcserial: add additional Sierra Wireless QMI devices

A set of new VID/PIDs retrieved from the out-of-tree GobiNet/GobiSerial
Sierra Wireless drivers.

Signed-off-by: Aleksander Morgado <aleksander@aleksander.es>
Link: http://marc.info/?l=linux-usb&m=140136310027293&w=2
Cc: <stable@vger.kernel.org> # backport in link above
Signed-off-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agousb: host: max3421-hcd: Use module_spi_driver
Sachin Kamat [Thu, 29 May 2014 11:51:01 +0000 (17:21 +0530)]
usb: host: max3421-hcd: Use module_spi_driver

module_spi_driver simplifies the code by eliminating
boilerplate code.

Signed-off-by: Sachin Kamat <sachin.kamat@linaro.org>
Acked-by: David Mosberger <davidm@egauge.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agousb: host: max3421-hcd: Allow platform-data to specify Vbus polarity
David Mosberger-Tang [Thu, 29 May 2014 16:23:55 +0000 (10:23 -0600)]
usb: host: max3421-hcd: Allow platform-data to specify Vbus polarity

Signed-off-by: Davidm Mosberger <davidm@egauge.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agousb: host: max3421-hcd: fix "spi_rd8" uses dynamic stack allocation warning
David Mosberger-Tang [Thu, 29 May 2014 04:40:00 +0000 (22:40 -0600)]
usb: host: max3421-hcd: fix "spi_rd8" uses dynamic stack allocation warning

kmalloc the SPI rx and tx data buffers.  This appears to be the only
portable way to guarantee that the buffers are DMA-safe (e.g., in
separate DMA cache-lines).  This patch makes the spi_rdX()/spi_wrX()
non-reentrant, but that's OK because calls to them are guaranteed to
be serialized by the per-HCD SPI-thread.

Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: David Mosberger <davidm@egauge.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agousb: host: max3421-hcd: Fix missing unlock in max3421_urb_enqueue()
David Mosberger-Tang [Wed, 28 May 2014 22:09:16 +0000 (16:09 -0600)]
usb: host: max3421-hcd: Fix missing unlock in max3421_urb_enqueue()

Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: David Mosberger <davidm@egauge.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agousb: qcserial: add Netgear AirCard 341U
Aleksander Morgado [Wed, 28 May 2014 19:13:51 +0000 (21:13 +0200)]
usb: qcserial: add Netgear AirCard 341U

Signed-off-by: Aleksander Morgado <aleksander@aleksander.es>
Cc: stable <stable@vger.kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agoDocumentation: dt-bindings: update xhci-platform DT binding for R-Car H2 and M2
Yoshihiro Shimoda [Wed, 28 May 2014 11:23:26 +0000 (20:23 +0900)]
Documentation: dt-bindings: update xhci-platform DT binding for R-Car H2 and M2

This commit extends the compatible string list of the xhci-platform
binding with the new "renesas,xhci-r8a7790" and "renesas,xhci-r8a7791"
compatible strings. It is used to describe the xHCI controller which
is available in the R-Car H2 and M2 SoCs.

Acked-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agousb: host: xhci-plat: add xhci_plat_start()
Yoshihiro Shimoda [Wed, 28 May 2014 11:22:58 +0000 (20:22 +0900)]
usb: host: xhci-plat: add xhci_plat_start()

Some platforms (such as the Renesas R-Car) need to initialize some specific
registers after xhci driver calls usb_add_hcd() and before the driver calls
xhci_run(). So, this patch adds the xhci_plat_start() function.

Acked-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agousb: host: max3421-hcd: Fix potential NULL urb dereference
David Mosberger-Tang [Wed, 28 May 2014 16:06:24 +0000 (10:06 -0600)]
usb: host: max3421-hcd: Fix potential NULL urb dereference

Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: David Mosberger <davidm@egauge.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agoRevert "usb: gadget: net2280: Add support for PLX USB338X"
Greg Kroah-Hartman [Wed, 28 May 2014 18:35:41 +0000 (11:35 -0700)]
Revert "usb: gadget: net2280: Add support for PLX USB338X"

This reverts commit c4128cac3557ddd5fa972cb6511c426cd94a7ccd.

This should come through Felipe's tree first, and there was a bunch of
other patches that are needed after this one as well that I didn't have.

Cc: Ricardo Ribalda Delgado <ricardo.ribalda@gmail.com>
Cc: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agoUSB: usbip: remove CONFIG_USB_DEBUG reference
Greg Kroah-Hartman [Wed, 28 May 2014 16:42:10 +0000 (09:42 -0700)]
USB: usbip: remove CONFIG_USB_DEBUG reference

As the config option is gone, it's a bit hard to enable it so remove the
reference to it.

Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agoUSB: remove CONFIG_USB_DEBUG from defconfig files
Greg Kroah-Hartman [Wed, 28 May 2014 16:40:45 +0000 (09:40 -0700)]
USB: remove CONFIG_USB_DEBUG from defconfig files

Now that CONFIG_USB_DEBUG is gone, remove it from a number of defconfig
files that were enabling it.

Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agousb: resume child device when port is powered on
Dan Williams [Wed, 21 May 2014 01:09:36 +0000 (18:09 -0700)]
usb: resume child device when port is powered on

Unconditionally wake up the child device when the power session is
recovered.

This addresses the following scenarios:

1/ The device may need a reset on power-session loss, without this
   change port power-on recovery exposes khubd to scenarios that
   usb_port_resume() is set to handle.  Prior to port power control the
   only time a power session would be lost is during dpm_suspend of the
   hub.  In that scenario usb_port_resume() is guaranteed to be called
   prior to khubd running for that port.  With this change we wakeup the
   child device as soon as possible (prior to khubd running again for this
   port).

   Although khubd has facilities to wake a child device it will only do
   so if the portstatus / portchange indicates a suspend state.  In the
   case of port power control we are not coming from a hub-port-suspend
   state.  This implementation simply uses pm_request_resume() to wake the
   device and relies on the port_dev->status_lock to prevent any collisions
   between khubd and usb_port_resume().

2/ This mechanism rate limits port power toggling.  The minimum port
   power on/off period is now gated by the child device suspend/resume
   latency.  Empirically this mitigates devices downgrading their connection
   on perceived instability of the host connection.  This ratelimiting is
   really only relevant to port power control testing, but it is a nice
   side effect of closing the above race.  Namely, the race of khubd for
   the given port running while a usb_port_resume() event is pending.

3/ Going forward we are finding that power-session recovery requires
   warm-resets (http://marc.info/?t=138659232900003&r=1&w=2).  This
   mechanism allows for warm-resets to be requested at the same point in
   the resume path for hub dpm_suspend power session losses, or port
   rpm_suspend power session losses.

4/ If the device *was* disconnected the only time we'll know for sure is
   after a failed resume, so it's necessary for usb_port_runtime_resume()
   to expedite a usb_port_resume() to clean up the removed device.  The
   reasoning for this is "least surprise" for the user. Turning on a port
   means that hotplug detection is again enabled for the port, it is
   surprising that devices that were removed while the port was off are not
   disconnected until they are attempted to be used.  As a user "why would
   I try to use a device I removed from the system?"

1, 2, and 4 are not a problem in the system dpm_resume() case because,
although the power-session is lost, khubd is frozen until after device
resume.  For the rpm_resume() case pm_request_resume() is used to
request re-validation of the device, and if it happens to collide with a
khubd run we rely on the port_dev->status_lock to synchronize those
operations.

Besides testing, the primary scenario where this mechanism is expected
to be triggered is when the user changes the port power policy
(control/pm_qos_no_poweroff, or power/control).   Each time power is
enabled want to revalidate the child device, where the revalidation is
handled by usb_port_resume().

Given that this arranges for port_dev->child to be de-referenced in
usb_port_runtime_resume() we need to make sure not to collide with
usb_disconnect() that frees the usb_device.  To this end we hold the
port active with the "child_usage" reference across the disconnect
event.  Subsequently, the need to access hub->child_usage_bits lead to
the creation of hub_disconnect_children() to remove any ambiguity of
which "hub" is being acted on in usb_disconnect() (prompted-by sharp
eyes from Alan).

Cc: Rafael J. Wysocki <rjw@rjwysocki.net>
Acked-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agousb: hub_handle_remote_wakeup() depends on CONFIG_PM_RUNTIME=y
Dan Williams [Wed, 21 May 2014 01:09:31 +0000 (18:09 -0700)]
usb: hub_handle_remote_wakeup() depends on CONFIG_PM_RUNTIME=y

Per Alan:
"You mean from within hub_handle_remote_wakeup()?  That routine will
never get called if CONFIG_PM_RUNTIME isn't enabled, because khubd
never sees wakeup requests if they arise during system suspend.

In fact, that routine ought to go inside the "#ifdef CONFIG_PM_RUNTIME"
portion of hub.c, along with the other suspend/resume code."

Suggested-by: Alan Stern <stern@rowland.harvard.edu>
Acked-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agousb: introduce port status lock
Dan Williams [Wed, 21 May 2014 01:09:26 +0000 (18:09 -0700)]
usb: introduce port status lock

In general we do not want khubd to act on port status changes that are
the result of in progress resets or USB runtime PM operations.
Specifically port power control testing has been able to trigger an
unintended disconnect in hub_port_connect_change(), paraphrasing:

if ((portstatus & USB_PORT_STAT_CONNECTION) && udev &&
    udev->state != USB_STATE_NOTATTACHED) {
if (portstatus & USB_PORT_STAT_ENABLE) {
/* Nothing to do */
} else if (udev->state == USB_STATE_SUSPENDED &&
udev->persist_enabled) {
...
} else {
/* Don't resuscitate */;
}
}

...by falling to the "Don't resuscitate" path or missing
USB_PORT_STAT_CONNECTION because usb_port_resume() was in the middle of
modifying the port status.

So, we want a new lock to hold off khubd for a given port while the
child device is being suspended, resumed, or reset.  The lock ordering
rules are now usb_lock_device() => usb_lock_port().  This is mandated by
the device core which may hold the device_lock on the usb_device before
invoking usb_port_{suspend|resume} which in turn take the status_lock on
the usb_port.  We attempt to hold the status_lock for the duration of a
port_event() run, and drop/re-acquire it when needing to take the
device_lock.  The lock is also dropped/re-acquired during
hub_port_reconnect().

This patch also deletes hub->busy_bits as all use cases are now covered
by port PM runtime synchronization or the port->status_lock and it
pushes down usb_device_lock() into usb_remote_wakeup().

Acked-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agousb: synchronize port poweroff and khubd
Dan Williams [Wed, 21 May 2014 01:09:20 +0000 (18:09 -0700)]
usb: synchronize port poweroff and khubd

If a port is powered-off, or in the process of being powered-off, prevent
khubd from operating on it.  Otherwise, the following sequence of events
leading to an unintended disconnect may occur:

Events:
(0) <set pm_qos_no_poweroff to '0' for port1>
(1) hub 2-2:1.0: hub_resume
(2) hub 2-2:1.0: port 1: status 0301 change 0000
(3) hub 2-2:1.0: state 7 ports 4 chg 0002 evt 0000
(4) hub 2-2:1.0: port 1, power off status 0000, change 0000, 12 Mb/s
(5) usb 2-2.1: USB disconnect, device number 5

Description:
(1) hub is resumed before sending a ClearPortFeature request
(2) hub_activate() notices the port is connected and sets
    hub->change_bits for the port
(3) hub_events() starts, but at the same time the port suspends
(4) hub_connect_change() sees the disabled port and triggers disconnect

Acked-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agousb: refactor port handling in hub_events()
Dan Williams [Wed, 21 May 2014 01:09:15 +0000 (18:09 -0700)]
usb: refactor port handling in hub_events()

In preparation for synchronizing port handling with pm_runtime
transitions refactor port handling into its own subroutine.

We expect that clearing some status flags will be required regardless of
the port state, so handle those first and group all non-trivial actions
at the bottom of the routine.

This also splits off the bottom half of hub_port_connect_change() into
hub_port_reconnect() in prepartion for introducing a port->status_lock.
hub_port_reconnect() will expect the port lock to not be held while
hub_port_connect_change() expects to enter with it held.

Other cleanups include:
1/ reflowing to 80 columns
2/ replacing redundant usages of 'hub->hdev' with 'hdev'
3/ consolidate clearing of ->change_bits() in hub_port_connect_change
4/ consolidate calls to usb_reset_device

Acked-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agousb: usb3 ports do not support FEAT_C_ENABLE
Dan Williams [Wed, 21 May 2014 01:09:10 +0000 (18:09 -0700)]
usb: usb3 ports do not support FEAT_C_ENABLE

The port pm_runtime implementation unconditionally clears FEAT_C_ENABLE
after clearing PORT_POWER, but the bit is reserved on usb3 hub ports.
We expect khubd to be prevented from running because the port state is
not RPM_ACTIVE, so we need to clear any errors for usb2 ports.

Acked-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agousb: don't clear FEAT_C_ENABLE on usb_port_runtime_resume failure
Dan Williams [Wed, 21 May 2014 01:09:03 +0000 (18:09 -0700)]
usb: don't clear FEAT_C_ENABLE on usb_port_runtime_resume failure

Three reasons:
1/ It's an invalid operation on usb3 ports
2/ There's no guarantee of when / if a usb2 port has entered an error
   state relative to PORT_POWER request
3/ The port is active / powered at this point, so khubd will clear it as
   a matter of course

Acked-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agousb: block suspension of superspeed port while hispeed peer is active
Dan Williams [Wed, 21 May 2014 01:08:57 +0000 (18:08 -0700)]
usb: block suspension of superspeed port while hispeed peer is active

ClearPortFeature(PORT_POWER) on a usb3 port places the port in either a
DSPORT.Powered-off-detect / DSPORT.Powered-off-reset loop, or the
DSPORT.Powered-off state.  There is no way to ensure that RX
terminations will persist in this state, so it is possible a device will
degrade to its usb2 connection.  Prevent this by blocking power-off of a
usb3 port while its usb2 peer is active, and powering on a usb3 port
before its usb2 peer.

By default the latency between peer power-on events is 0.  In order for
the device to not see usb2 active while usb3 is still powering up inject
the hub recommended power_on_good delay.  In support of satisfying the
power_on_good delay outside of hub_power_on() refactor the places where
the delay is consumed to call a new hub_power_on_good_delay() helper.

Finally, because this introduces several new checks for whether a port
is_superspeed, cache that disctinction at port creation so that we don't
need to keep looking up the parent hub device.

Acked-by: Alan Stern <stern@rowland.harvard.edu>
[alan]: add a 'superspeed' flag to the port
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agousb: make usb_port flags atomic, rename did_runtime_put to child_usage
Dan Williams [Wed, 21 May 2014 01:08:52 +0000 (18:08 -0700)]
usb: make usb_port flags atomic, rename did_runtime_put to child_usage

We want to manipulate ->did_runtime_put in usb_port_runtime_resume(),
but we don't want that to collide with other updates.  Move usb_port
flags to new port-bitmap fields in usb_hub. "did_runtime_put" is renamed
"child_usage_bits" to reflect that it is strictly standing in for the
fact that usb_devices are not the device_model children of their parent
port.

Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Acked-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agousb: sysfs link peer ports
Dan Williams [Wed, 21 May 2014 01:08:45 +0000 (18:08 -0700)]
usb: sysfs link peer ports

The usb topology after this change will have symlinks between usb3 ports
and their usb2 peers, for example:

usb2/2-1/2-1:1.0/2-1-port1/peer => ../../../../usb3/3-1/3-1:1.0/3-1-port1
usb2/2-1/2-1:1.0/2-1-port2/peer => ../../../../usb3/3-1/3-1:1.0/3-1-port2
usb2/2-1/2-1:1.0/2-1-port3/peer => ../../../../usb3/3-1/3-1:1.0/3-1-port3
usb2/2-1/2-1:1.0/2-1-port4/peer => ../../../../usb3/3-1/3-1:1.0/3-1-port4
usb2/2-0:1.0/usb2-port1/peer    => ../../../usb3/3-0:1.0/usb3-port1
usb2/2-0:1.0/usb2-port2/peer    => ../../../usb3/3-0:1.0/usb3-port2
usb2/2-0:1.0/usb2-port3/peer    => ../../../usb3/3-0:1.0/usb3-port3
usb2/2-0:1.0/usb2-port4/peer    => ../../../usb3/3-0:1.0/usb3-port4

usb3/3-1/3-1:1.0/usb3-1-port1/peer => ../../../../usb2/2-1/2-1:1.0/2-1-port1
usb3/3-1/3-1:1.0/usb3-1-port2/peer => ../../../../usb2/2-1/2-1:1.0/2-1-port2
usb3/3-1/3-1:1.0/usb3-1-port3/peer => ../../../../usb2/2-1/2-1:1.0/2-1-port3
usb3/3-1/3-1:1.0/usb3-1-port4/peer => ../../../../usb2/2-1/2-1:1.0/2-1-port4
usb3/3-0:1.0/usb3-port1/peer       => ../../../usb2/2-0:1.0/usb2-port1
usb3/3-0:1.0/usb3-port2/peer       => ../../../usb2/2-0:1.0/usb2-port2
usb3/3-0:1.0/usb3-port3/peer       => ../../../usb2/2-0:1.0/usb2-port3
usb3/3-0:1.0/usb3-port4/peer       => ../../../usb2/2-0:1.0/usb2-port4

Introduce link_peers_report() to notify on all link_peers() failure
cases.

Acked-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agousb: find internal hub tier mismatch via acpi
Dan Williams [Wed, 21 May 2014 01:08:40 +0000 (18:08 -0700)]
usb: find internal hub tier mismatch via acpi

ACPI identifies peer ports by setting their 'group_token' and
'group_position' _PLD data to the same value.  If a platform has tier
mismatch [1] , ACPI can override the default (USB3 defined) peer port
association for internal hubs.  External hubs follow the default peer
association scheme.

Location data is cached as an opaque cookie in usb_port_location data.

Note that we only consider the group_token and group_position attributes
from the _PLD data as ACPI specifies that group_token is a unique
identifier.

When we find port location data for a port then we assume that the
firmware will also describe its peer port.  This allows the
implementation to only ever set the peer once.  This leads to a question
about what happens when a pm runtime event occurs while the peer
associations are still resolving.  Since we only ever set the peer
information once, a USB3 port needs to be prevented from suspending
while its ->peer pointer is NULL (implemented in a subsequent patch).

There is always the possibility that firmware mis-identifies the ports,
but there is not much the kernel can do in that case.

[1]: xhci 1.1 appendix D figure 131
[2]: acpi 5 section 6.1.8

[alan]: don't do default peering when acpi data present
Suggested-by: Alan Stern <stern@rowland.harvard.edu>
Acked-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agousb: assign usb3 external hub port peers
Dan Williams [Wed, 21 May 2014 01:08:33 +0000 (18:08 -0700)]
usb: assign usb3 external hub port peers

Given that root hub port peers are already established, external hub peer
ports can be determined by traversing the device topology:

1/ ascend to the parent hub and find the upstream port_dev

2/ walk ->peer to find the peer port

3/ descend to the peer hub via ->child

4/ find the port with the matching port id

Note that this assumes the port labeling scheme required by the
specification [1].

[1]: usb3 3.1 section 10.3.3

Acked-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agousb: assign default peer ports for root hubs
Dan Williams [Wed, 21 May 2014 01:08:28 +0000 (18:08 -0700)]
usb: assign default peer ports for root hubs

Assume that the peer of a superspeed port is the port with the same id
on the shared_hcd root hub.  This identification scheme is required of
external hubs by the USB3 spec [1].  However, for root hubs, tier mismatch
may be in effect [2].  Tier mismatch can only be enumerated via platform
firmware.  For now, simply perform the nominal association.

A new lock 'usb_port_peer_mutex' is introduced to synchronize port
device add/remove with peer lookups.  It protects peering against
changes to hcd->shared_hcd, hcd->self.root_hub, hdev->maxchild, and
port_dev->child pointers.

[1]: usb 3.1 section 10.3.3
[2]: xhci 1.1 appendix D

Cc: Alan Stern <stern@rowland.harvard.edu>
[alan: usb_port_peer_mutex locking scheme]
Acked-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agousb: cleanup setting udev->removable from port_dev->connect_type
Dan Williams [Wed, 21 May 2014 01:08:22 +0000 (18:08 -0700)]
usb: cleanup setting udev->removable from port_dev->connect_type

Once usb-acpi has set the port's connect type the usb_device's
->removable attribute can be set in the standard location
set_usb_port_removable().

This also changes behavior in the case where the firmware says that the
port connect type is unknown.  In that case just use the default setting
determined from the hub descriptor.

Note, we no longer pass udev->portnum to acpi_find_child_device() in the
root hub case since:
1/ the usb-core sets this to zero
2/ acpi always expects zero
...just pass zero.

Suggested-by: Alan Stern <stern@rowland.harvard.edu>
Acked-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agousb: rename usb_port device objects
Dan Williams [Wed, 21 May 2014 01:08:17 +0000 (18:08 -0700)]
usb: rename usb_port device objects

The current port name "portX" is ambiguous.  Before adding more port
messages rename ports to "<hub-device-name>-portX"

This is an ABI change, but the suspicion is that it will go unnoticed as
the port power control implementation has been broken since its
introduction.  If however, someone was relying on the old name we can
add sysfs links from the old name to the new name.

Additionally, it unifies/simplifies port dev_printk messages and modifies
instances of:
dev_XXX(hub->intfdev, ..."port %d"...
dev_XXX(&hdev->dev, ..."port%d"...
into:
dev_XXX(&port_dev->dev, ...

Now that the names are unique usb_port devices it would be nice if they
could be included in /sys/bus/usb.  However, it turns out that this
breaks 'lsusb -t'.  For now, create a dummy port driver so that print
messages are prefixed "usb 1-1-port3" rather than the
subsystem-ambiguous " 1-1-port3".

Finally, it corrects an odd usage of sscanf("port%d") in usb-acpi.c.

Suggested-by: Alan Stern <stern@rowland.harvard.edu>
Acked-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agousb: disable port power control if not supported in wHubCharacteristics
Dan Williams [Wed, 21 May 2014 01:08:12 +0000 (18:08 -0700)]
usb: disable port power control if not supported in wHubCharacteristics

A hub indicates whether it supports per-port power control via the
wHubCharacteristics field in its descriptor.  If it is not supported
a hub will still emulate ClearPortPower(PORT_POWER) requests by
stopping the link state machine.  However, since this does not save
power do not bother suspending.

This also consolidates support checks into a
hub_is_port_power_switchable() helper.

Acked-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agoUSB: mutual exclusion for resetting a hub and power-managing a port
Alan Stern [Wed, 21 May 2014 01:08:07 +0000 (18:08 -0700)]
USB: mutual exclusion for resetting a hub and power-managing a port

The USB core doesn't properly handle mutual exclusion between
resetting a hub and changing the power states of the hub's ports.  We
need to avoid sending port-power requests to the hub while it is being
reset, because such requests cannot succeed.

This patch fixes the problem by keeping track of when a reset is in
progress.  At such times, attempts to suspend (power-off) a port will
fail immediately with -EBUSY, and calls to usb_port_runtime_resume()
will update the power_is_on flag and return immediately.  When the
reset is complete, hub_activate() will automatically restore each port
to the proper power state.

Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agousb: pci_quirks: fix sparse 'symbol not declared' warning
Konrad Zapalowicz [Tue, 27 May 2014 21:09:14 +0000 (23:09 +0200)]
usb: pci_quirks: fix sparse 'symbol not declared' warning

This commit fixes the following sparse warning:

drivers/usb/host/pci-quirks.c:
    - 252: warning: symbol 'usb_hcd_amd_remote_wakeup_quirk' was not
      declared. Should it be static?

This function is exported so the fix was to add it's declaration to the
header file.

Signed-off-by: Konrad Zapalowicz <bergo.torino@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agousb: gadget: rename CONFIG_USB_GADGET_PXA25X
Paul Bolle [Mon, 26 May 2014 21:37:09 +0000 (23:37 +0200)]
usb: gadget: rename CONFIG_USB_GADGET_PXA25X

Commit 193ab2a60700 ("usb: gadget: allow multiple gadgets to be built")
basically renamed the Kconfig symbol USB_GADGET_PXA25X to USB_PXA25X. It
did not rename the related macros in use at that time. Commit
c0a39151a405 ("ARM: pxa: fix inconsistent CONFIG_USB_PXA27X") did so for
all but one macro. Rename that last macro too now.

Fixes: 193ab2a60700 ("usb: gadget: allow multiple gadgets to be built")
Signed-off-by: Paul Bolle <pebolle@tiscali.nl>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agoUSB: storage: ene_ub6250: Use kmemdup instead of kmalloc + memcpy
Benoit Taine [Mon, 26 May 2014 15:21:10 +0000 (17:21 +0200)]
USB: storage: ene_ub6250: Use kmemdup instead of kmalloc + memcpy

This issue was reported by coccicheck using the semantic patch
at scripts/coccinelle/api/memdup.cocci

Signed-off-by: Benoit Taine <benoit.taine@lip6.fr>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agousb: usbtest: add pattern check on pipe in phase of unlink read
Huang Rui [Thu, 22 May 2014 10:06:14 +0000 (18:06 +0800)]
usb: usbtest: add pattern check on pipe in phase of unlink read

TEST 11 unlinks the URB read request for N times. When host and gadget
both initialize pattern 1 (mod 63) data series to do IN transfer, the
host side function should check the data buffer if it is as mod 63
series, because the data packet which host receivced will follow
pattern 1. So this patch adds this checking action.

Signed-off-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agousb: usbtest: fix unlink write error with pattern 1
Huang Rui [Mon, 26 May 2014 02:55:36 +0000 (10:55 +0800)]
usb: usbtest: fix unlink write error with pattern 1

TEST 12 and TEST 24 unlinks the URB write request for N times. When
host and gadget both initialize pattern 1 (mod 63) data series to
transfer, the gadget side will complain the wrong data which is not
expected.  Because in host side, usbtest doesn't fill the data buffer
as mod 63 and this patch fixed it.

[20285.488974] dwc3 dwc3.0.auto: ep1out-bulk: Transfer Not Ready
[20285.489181] dwc3 dwc3.0.auto: ep1out-bulk: reason Transfer Not Active
[20285.489423] dwc3 dwc3.0.auto: ep1out-bulk: req ffff8800aa6cb480 dma aeb50800 length 512 last
[20285.489727] dwc3 dwc3.0.auto: ep1out-bulk: cmd 'Start Transfer' params 00000000 a9eaf000 00000000
[20285.490055] dwc3 dwc3.0.auto: Command Complete --> 0
[20285.490281] dwc3 dwc3.0.auto: ep1out-bulk: Transfer Not Ready
[20285.490492] dwc3 dwc3.0.auto: ep1out-bulk: reason Transfer Active
[20285.490713] dwc3 dwc3.0.auto: ep1out-bulk: endpoint busy
[20285.490909] dwc3 dwc3.0.auto: ep1out-bulk: Transfer Complete
[20285.491117] dwc3 dwc3.0.auto: request ffff8800aa6cb480 from ep1out-bulk completed 512/512 ===> 0
[20285.491431] zero gadget: bad OUT byte, buf[1] = 0
[20285.491605] dwc3 dwc3.0.auto: ep1out-bulk: cmd 'Set Stall' params 00000000 00000000 00000000
[20285.491915] dwc3 dwc3.0.auto: Command Complete --> 0
[20285.492099] dwc3 dwc3.0.auto: queing request ffff8800aa6cb480 to ep1out-bulk length 512
[20285.492387] dwc3 dwc3.0.auto: ep1out-bulk: Transfer Not Ready
[20285.492595] dwc3 dwc3.0.auto: ep1out-bulk: reason Transfer Not Active
[20285.492830] dwc3 dwc3.0.auto: ep1out-bulk: req ffff8800aa6cb480 dma aeb51000 length 512 last
[20285.493135] dwc3 dwc3.0.auto: ep1out-bulk: cmd 'Start Transfer' params 00000000 a9eaf000 00000000
[20285.493465] dwc3 dwc3.0.auto: Command Complete --> 0

Cc: <stable@vger.kernel.org>
Signed-off-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agousb: usb5303: add support for reference clock specified in device tree
Marek Szyprowski [Thu, 22 May 2014 11:21:38 +0000 (13:21 +0200)]
usb: usb5303: add support for reference clock specified in device tree

USB3503 chip supports 8 values of reference clock. The value is
specified by REF_SEL[1:0] pins and INT_N line. This patch add support
for getting 'refclk' clock, enabling it and setting INT_N line according
to the value of the gathered clock. If no clock has been specified,
driver defaults to the old behaviour (assuming that clock has been
specified by REF_SEL pins from primary reference clock frequencies
table).

Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agoUSB: separate usb_address0 mutexes for each bus
Todd E Brandt [Mon, 19 May 2014 17:55:32 +0000 (10:55 -0700)]
USB: separate usb_address0 mutexes for each bus

This patch creates a separate instance of the usb_address0 mutex for each USB
bus, and attaches it to the usb_bus device struct. This allows devices on
separate buses to be enumerated in parallel; saving time.

In the current code, there is a single, global instance of the usb_address0
mutex which is used for all devices on all buses. This isn't completely
necessary, as this mutex is only needed to prevent address0 collisions for
devices on the *same* bus (usb 2.0 spec, sec 4.6.1). This superfluous coverage
can cause additional delay in system resume on systems with multiple hosts
(up to several seconds depending on what devices are attached).

Signed-off-by: Todd Brandt <todd.e.brandt@linux.intel.com>
Acked-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agoxhci: Switch only Intel Lynx Point-LP ports to EHCI on shutdown.
Denis Turischev [Tue, 20 May 2014 11:00:42 +0000 (14:00 +0300)]
xhci: Switch only Intel Lynx Point-LP ports to EHCI on shutdown.

Patch "xhci: Switch Intel Lynx Point ports to EHCI on shutdown."
commit c09ec25d3684cad74d851c0f028a495999591279 is not fully correct

It switches both Lynx Point and Lynx Point-LP ports to EHCI on shutdown.
On some Lynx Point machines it causes spurious interrupt,
which wake the system: bugzilla.kernel.org/show_bug.cgi?id=76291

On Lynx Point-LP on the contrary switching ports to EHCI seems to be
necessary to fix these spurious interrupts.

Signed-off-by: Denis Turischev <denis@compulab.co.il>
Reported-by: Wulf Richartz <wulf.richartz@gmail.com>
Cc: Mathias Nyman <mathias.nyman@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agoUSB: delete CONFIG_USB_DEVICEFS from defconfig
Naoki MATSUMOTO [Thu, 15 May 2014 11:17:44 +0000 (20:17 +0900)]
USB: delete CONFIG_USB_DEVICEFS from defconfig

It no longer occurs in Kconfig.
USB: remove CONFIG_USB_DEVICEFS(fb28d58b) leaked remove defconfig.

Signed-off-by: Naoki MATSUMOTO <nekomatu+linux@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agoUSB: udl: proper error reporting
Oliver Neukum [Mon, 19 May 2014 11:50:22 +0000 (13:50 +0200)]
USB: udl: proper error reporting

Parsing device descriptors can fail due to a failed memory
allocation. The error needs to be properly propagated to the
upper layers.

Signed-off-by: Oliver Neukum <oneukum@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agoUSB: yurex: fix race between probe() and read()
Oliver Neukum [Mon, 19 May 2014 11:52:20 +0000 (13:52 +0200)]
USB: yurex: fix race between probe() and read()

There's a window during which read() would return 0 instead
of a correct error for no data yet. Reorder initialization
to fix the race.

Signed-off-by: Oliver Neukum <oneukum@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agoUSB: appledisplay: fix race between reading and writing from the device
Oliver Neukum [Mon, 19 May 2014 11:53:55 +0000 (13:53 +0200)]
USB: appledisplay: fix race between reading and writing from the device

The workqueue handler may call appledisplay_bl_get_brightness() while
user space calls appledisplay_bl_update_status(). As they share a
buffer that must not happen. Use a mutex for mutual exclusion.

Signed-off-by: Oliver Neukum <oneukum@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agoUSB: usbtmc: fix DMA on stack
Oliver Neukum [Mon, 19 May 2014 11:54:57 +0000 (13:54 +0200)]
USB: usbtmc: fix DMA on stack

send_request_dev_dep_msg_in() use a buffer allocated on the stack.
Fix by kmalloc()ing the buffer.

Signed-off-by: Oliver Neukum <oneukum@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agoUSB: cdc-acm: use BIT macro
Oliver Neukum [Mon, 19 May 2014 11:55:25 +0000 (13:55 +0200)]
USB: cdc-acm: use BIT macro

Converting the header to BIT for readability. No functional
change.

Signed-off-by: Oliver Neukum <oneukum@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agousb: gadget: net2280: Add support for PLX USB338X
Ricardo Ribalda Delgado [Thu, 15 May 2014 12:28:46 +0000 (14:28 +0200)]
usb: gadget: net2280: Add support for PLX USB338X

This patch adds support for the PLX USB3380 and USB3382.

This driver is based on the driver from the manufacturer.

Since USB338X is register compatible with NET2280, I thought that it
would be better to include this hardware into net2280 driver.

Manufacturer's driver only supported the USB33X, did not follow the
Kernel Style and contain some trivial errors. This patch has tried to
address this issues.

This patch has only been tested on USB338x hardware, but the merge has
been done trying to not affect the behaviour of NET2280.

Signed-off-by: Ricardo Ribalda Delgado <ricardo.ribalda@gmail.com>
Tested-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agousb: gadget: make return of 0 explicit
Julia Lawall [Mon, 19 May 2014 04:31:07 +0000 (06:31 +0200)]
usb: gadget: make return of 0 explicit

Delete unnecessary local variable whose value is always 0 and that hides
the fact that the result is always 0.

A simplified version of the semantic patch that fixes this problem is as
follows: (http://coccinelle.lip6.fr/)

// <smpl>
@r exists@
local idexpression ret;
expression e;
position p;
@@

-ret = 0;
... when != ret = e
return
- ret
+ 0
  ;
// </smpl>

Signed-off-by: Julia Lawall <Julia.Lawall@lip6.fr>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agophy: Enable USB PHY support for arm64
Liviu Dudau [Wed, 14 May 2014 19:17:30 +0000 (20:17 +0100)]
phy: Enable USB PHY support for arm64

Signed-off-by: Liviu Dudau <Liviu.Dudau@arm.com>
Signed-off-by: Ryan Harkin <ryan.harkin@linaro.org>
Signed-off-by: Mark Brown <broonie@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agoxhci: unified loggig of RESET_ON_RESUME
Oliver Neukum [Wed, 14 May 2014 12:00:23 +0000 (14:00 +0200)]
xhci: unified loggig of RESET_ON_RESUME

Either we log for all chips we set the quirk for or for
none. This patch reports it for all chips.

Signed-off-by: Oliver Neukum <oliver@neukum.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agousb: ehci-platform: add optional reset controller retrieval
Boris BREZILLON [Tue, 13 May 2014 15:44:19 +0000 (17:44 +0200)]
usb: ehci-platform: add optional reset controller retrieval

On the Allwinner's A31 SoC the reset line connected to the EHCI IP has to
be deasserted for the EHCI block to be usable.

Add support for an optional reset controller that will be deasserted on
power off and asserted on power on.

Signed-off-by: Boris BREZILLON <boris.brezillon@free-electrons.com>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Acked-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agousb: ohci-platform: Enable optional use of reset controller
Maxime Ripard [Tue, 13 May 2014 15:44:20 +0000 (17:44 +0200)]
usb: ohci-platform: Enable optional use of reset controller

The OHCI controllers used in the Allwinner A31 are asserted in reset using a
global reset controller.

Add optional support for such a controller in the OHCI platform driver.

Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Acked-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agousb: ohci: sort out dependencies for lpc32xx and omap
Arnd Bergmann [Thu, 8 May 2014 13:52:20 +0000 (15:52 +0200)]
usb: ohci: sort out dependencies for lpc32xx and omap

The dependency on the isp1301 driver is not something that
should be in the main OHCI driver but rather the SoC specific
part of it.

This moves the dependency for LPC32xx into USB_OHCI_HCD_LPC32XX,
and changes the 'select ISP1301_OMAP' to a similar 'depends on'.
Since the same dependency exists for the client driver, do the
same change there.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Cc: linux-omap@vger.kernel.org
Acked-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agousb: ohci-da8xx can only be built-in
Arnd Bergmann [Thu, 8 May 2014 13:52:21 +0000 (15:52 +0200)]
usb: ohci-da8xx can only be built-in

The PHY setup code of the TI DaVinci DA8xx OHCI controller
uses ad-hoc register access using a pointer that is meant to
be used only by the DaVinci platform implementation and that
is intentionally not exported to loadable modules. This results
in a link error on configurations that use a modular OHCI
code on this platform.

While the proper solution for this problem would be to
implement a real PHY driver shared by ohci-da8xx and musb-da8xx,
this patch for now just works around the build error by
only allowing the ohci-da8xx code to be built-in.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agousb: xhci: avoid warning for !PM_SLEEP
Arnd Bergmann [Thu, 8 May 2014 13:52:19 +0000 (15:52 +0200)]
usb: xhci: avoid warning for !PM_SLEEP

If we build a kernel with PM_SUSPEND set and no PM_SLEEP,
we get a build warning in the xhci-plat driver about unused
functions.

To fix this, use "#ifdef CONFIG_PM_SLEEP", like we do in most
other drivers nowadays.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Cc: Mathias Nyman <mathias.nyman@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agousb: dwc2: Add function to calculate correct FIFO sizes
Dinh Nguyen [Wed, 7 May 2014 13:31:29 +0000 (08:31 -0500)]
usb: dwc2: Add function to calculate correct FIFO sizes

The dwc2 IP on the SOCFPGA cannot use the default HW configured
FIFO sizes. The total FIFO depth as read from GHWCFG3 reports 0x1f80 or 8064
32-bit words. But the GRXFSIZ, GNPTXFSIZ, and HPTXFSIZ register defaults
to 0x2000 or 8192 32-bit words. So the driver cannot just use the fifo sizes
as read from those registers.

For platforms that face the same issue, this commits sets the RX, periodic TX,
and non-periodic TX fifo size to those that are recommended v2.93a spec for
the DWC2 IP. Implements Method #2 from the Synopsys v2.93a spec for the DWC2.

Signed-off-by: Dinh Nguyen <dinguyen@altera.com>
Acked-by: Paul Zimmerman <paulz@synopsys.com>
Reviewed-by: Felipe Balbi <balbi@ti.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agousb: dwc2: Disable descriptor dma mode by default
Dinh Nguyen [Wed, 7 May 2014 13:30:33 +0000 (08:30 -0500)]
usb: dwc2: Disable descriptor dma mode by default

Even though the IP supports Descriptor DMA mode, it does not support SPLIT
transactions in this mode. So the driver, in its currently form, will not
support LS/FS devices when connected to a HS Hub if Descriptor DMA mode is
enabled.

So we should just default to disable descriptor dma mode.

Signed-off-by: Dinh Nguyen <dinguyen@altera.com>
Acked-by: Paul Zimmerman <paulz@synopsys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agoDocumentation: dt-bindings: update xhci-platform DT binding
Gregory CLEMENT [Thu, 15 May 2014 10:17:34 +0000 (12:17 +0200)]
Documentation: dt-bindings: update xhci-platform DT binding

This commit extends the compatible string list of the xhci-platform
binding with the new "armada-375-xhci" and "armada-380-xhci"
compatible strings. It is used to describe the XHCI controller which
is available in the Armada 375 and 38x SoCs.

It also indicates that an optional 'clocks' property is now supported.

Signed-off-by: Gregory CLEMENT <gregory.clement@free-electrons.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Acked-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agousb: host: xhci-plat: add support for the Armada 375/38x XHCI controllers
Gregory CLEMENT [Thu, 15 May 2014 10:17:33 +0000 (12:17 +0200)]
usb: host: xhci-plat: add support for the Armada 375/38x XHCI controllers

The Armada 375 and 38x SoCs come with an XHCI controller that requires
some specific initialization related to the MBus windows
configuration. This patch adds the support for this special
configuration as an XHCI quirk executed during probe.

Two new compatible strings are added to identify the Armada 375 and
Armada 38x XHCI controllers, and therefore enable the relevant quirk.

Signed-off-by: Gregory CLEMENT <gregory.clement@free-electrons.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Acked-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agousb: host: xhci-plat: add clock support
Gregory CLEMENT [Thu, 15 May 2014 10:17:32 +0000 (12:17 +0200)]
usb: host: xhci-plat: add clock support

Some platforms (such as the Armada 38x ones) can gate the clock of
their USB controller. This patch adds the support for one clock in
xhci-plat, by enabling it during probe and disabling it on remove.

To achieve this, it adds a 'struct clk *' member in xhci_hcd. While
only used for now in xhci-plat, it might be used by other drivers in
the future. Moreover, the xhci_hcd structure already holds other
members such as msix_count and msix_entries, which are MSI-X specific,
and therefore only used by xhci-pci.

Signed-off-by: Gregory CLEMENT <gregory.clement@free-electrons.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Acked-by: Felipe Balbi <balbi@ti.com>
Acked-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agousb: host: xhci-plat: sort the headers in alphabetic order
Gregory CLEMENT [Thu, 15 May 2014 10:17:31 +0000 (12:17 +0200)]
usb: host: xhci-plat: sort the headers in alphabetic order

Sorting the headers in alphabetic order will help to reduce the conflict
when adding new headers later.

Signed-off-by: Gregory CLEMENT <gregory.clement@free-electrons.com>
Acked-by: Felipe Balbi <balbi@ti.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agoDocumentation: dt-bindings: update ehci-orion binding documentation
Thomas Petazzoni [Thu, 15 May 2014 10:17:30 +0000 (12:17 +0200)]
Documentation: dt-bindings: update ehci-orion binding documentation

This commit updates the Device Tree binding documentation of
ehci-orion to take into account the fact that we can now optionally
pass a clock and a PHY reference.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Acked-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agousb: ehci-orion: add optional PHY support
Gregory CLEMENT [Thu, 15 May 2014 10:17:29 +0000 (12:17 +0200)]
usb: ehci-orion: add optional PHY support

This commit extends the ehci-orion so that it can optionally be passed
a reference to a PHY through the Device Tree. It will be useful for
the Armada 375 SoCs. If no PHY is provided then the behavior of the
driver is unchanged.

[Thomas: use devm_phy_optional_get() so that we handle -EPROBE_DEFER
properly. Also call phy_power_off() when needed, and rename goto
labels.]

Signed-off-by: Gregory CLEMENT <gregory.clement@free-electrons.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Acked-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agousb: ehci-orion: fix clock reference leaking
Gregory CLEMENT [Thu, 15 May 2014 10:17:28 +0000 (12:17 +0200)]
usb: ehci-orion: fix clock reference leaking

In order to disable the clock in the ->remove() function, a call to
devm_clk_get() is being made, which further increases the reference
count of the clock.

In order to clean this up, a private structure holding a pointer to
the clock is added using the override mechanism provided by the ehci
framework. This makes the driver clock handling much more logical.

The bug was introduced in v3.6, however the ehci framework allowing to
use the override mechanism has only been introduced in v3.8, so this
patch won't apply before it.

[Thomas: reword commit log, fix goto label names.]

Fixes: 8c869edaee07c623066266827371235fb9c12e01 ('ARM: Orion: EHCI: Add support for enabling clocks')
Cc: <stable@vger.kernel.org> # v3.8+
Signed-off-by: Gregory CLEMENT <gregory.clement@free-electrons.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Acked-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agousb: ehci-orion: rename error goto labels in ehci_orion_drv_probe()
Thomas Petazzoni [Thu, 15 May 2014 10:17:27 +0000 (12:17 +0200)]
usb: ehci-orion: rename error goto labels in ehci_orion_drv_probe()

In preparation to the introduction of additional initialization steps
in ehci_orion_drv_probe(), we rename the error goto labels from err1,
err2 and err3 names to some more meaningful names.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Acked-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agousb: ehci-orion: use platform_get_irq() for DT probing
Thomas Petazzoni [Thu, 15 May 2014 10:17:26 +0000 (12:17 +0200)]
usb: ehci-orion: use platform_get_irq() for DT probing

Commit 77dae54ab385033e488d8b07045bc7f8d931740f ('ARM: Kirkwood:
ehci-orion: Add device tree binding') added the Device Tree binding
for the ehci-orion driver. To achieve that with the irq, it used the
irq_of_parse_and_map() function when probed in DT-mode, and
platform_get_irq() when probed in non-DT mode.

This is not necessary: platform_get_irq() works just as fine in
DT-mode, since the conversion from DT information to 'struct resource'
is done by the generic layers of the kernel.

Therefore, this commit switches back to use just platform_get_irq().

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Acked-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agousb: common: rename phy-fsm-usb.c to usb-otg-fsm.c
Peter Chen [Tue, 29 Apr 2014 00:35:59 +0000 (08:35 +0800)]
usb: common: rename phy-fsm-usb.c to usb-otg-fsm.c

Since usb otg fsm implementation is not related to usb phy.
We move it from usb/phy/ to usb/common/, and rename it to
reflect its real meaning.

Cc: Felipe Balbi <balbi@ti.com>
Signed-off-by: Peter Chen <peter.chen@freescale.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agousb: move usb/usb-common.c to usb/common/usb-common.c
Peter Chen [Tue, 29 Apr 2014 00:35:58 +0000 (08:35 +0800)]
usb: move usb/usb-common.c to usb/common/usb-common.c

Since we will have more usb-common things, and it will let
usb-common.c be larger and larger, we create a folder named usb/common
for all usb common things.

Cc: Felipe Balbi <balbi@ti.com>
Signed-off-by: Peter Chen <peter.chen@freescale.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agousb: dwc2: Remove '0x' notation when using %pad format
Fabio Estevam [Tue, 29 Apr 2014 03:49:42 +0000 (00:49 -0300)]
usb: dwc2: Remove '0x' notation when using %pad format

%pad notation automatically prints in hexadecimal format (with '0x'), so remove
the unneeded '0x' to avoid a '0x0x' string.

Signed-off-by: Fabio Estevam <fabio.estevam@freescale.com>
Acked-by: Paul Zimmerman <paulz@synopsys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agoAdd support for using a MAX3421E chip as a host driver.
David Mosberger [Tue, 29 Apr 2014 04:14:07 +0000 (22:14 -0600)]
Add support for using a MAX3421E chip as a host driver.

Signed-off-by: David Mosberger <davidm@egauge.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agouwb: fix variable set but not used warnings
Thomas Pugliese [Mon, 28 Apr 2014 19:53:27 +0000 (14:53 -0500)]
uwb: fix variable set but not used warnings

Fix variable set but not used warnings in UWB.

Signed-off-by: Thomas Pugliese <thomas.pugliese@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agouwb: comment typo fix
Thomas Pugliese [Mon, 28 Apr 2014 19:53:26 +0000 (14:53 -0500)]
uwb: comment typo fix

Comment typo fix.

Signed-off-by: Thomas Pugliese <thomas.pugliese@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agouwb: whitespace and line length cleanups
Thomas Pugliese [Mon, 28 Apr 2014 19:53:25 +0000 (14:53 -0500)]
uwb: whitespace and line length cleanups

Fix whitespace and line length issues reported by checkpatch.

Signed-off-by: Thomas Pugliese <thomas.pugliese@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agousb: core: remove the Kconfig entry for USB_DEBUG
Peter Chen [Mon, 28 Apr 2014 06:12:38 +0000 (14:12 +0800)]
usb: core: remove the Kconfig entry for USB_DEBUG

Since we have already removed the usage of CONFIG_USB_DEBUG, it is
meaningless that there is still a configuration entry for CONFIG_USB_DEBUG.

Signed-off-by: Peter Chen <peter.chen@freescale.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agousb: remove redundant D0 power state set
Yijing Wang [Mon, 28 Apr 2014 11:35:09 +0000 (19:35 +0800)]
usb: remove redundant D0 power state set

Pci_enable_device() will set device power state to D0,
so it's no need to do it again after call pci_enable_device().

Signed-off-by: Yijing Wang <wangyijing@huawei.com>
Acked-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agoUSB: keyspan: fix potential null pointer dereference
Rickard Strandqvist [Fri, 16 May 2014 15:39:13 +0000 (17:39 +0200)]
USB: keyspan: fix potential null pointer dereference

Move control-urb dereference to after NULL-check. There is otherwise a
risk of a possible null pointer dereference.

Was largely found by using a static code analysis program called
cppcheck.

[Johan: modify commit message somewhat ]
[gkh: remove stable tag as it's not a real problem that anyone has ever hit]

Signed-off-by: Rickard Strandqvist <rickard_strandqvist@spectrumdigital.se>
Signed-off-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agoUSB: wusbcore: fix control-pipe directions
Johan Hovold [Mon, 26 May 2014 17:23:52 +0000 (19:23 +0200)]
USB: wusbcore: fix control-pipe directions

Fix incorrect pipe directions in control requests (which has been
silently fixed up by USB core).

Signed-off-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agoUSB: cdc-acm: use tty-port dtr_rts
Johan Hovold [Mon, 26 May 2014 17:23:51 +0000 (19:23 +0200)]
USB: cdc-acm: use tty-port dtr_rts

Add dtr_rts tty-port operation which implements proper DTR/RTS handling
(e.g. only lower DTR/RTS during shutdown if HUPCL is set).

Note that modem-control locking still needs to be added throughout the
driver.

Signed-off-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agoUSB: cdc-acm: remove redundant usb_mark_last_busy
Johan Hovold [Mon, 26 May 2014 17:23:50 +0000 (19:23 +0200)]
USB: cdc-acm: remove redundant usb_mark_last_busy

There's no need to call usb_mark_last_busy after having increased the PM
counter in write(). The device will be marked busy by USB core when the
PM counter is balanced in the completion handler.

Signed-off-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agoUSB: cdc-acm: do not update PM busy on read errors
Johan Hovold [Mon, 26 May 2014 17:23:49 +0000 (19:23 +0200)]
USB: cdc-acm: do not update PM busy on read errors

There's no need to update the runtime PM last_busy field on read urb
errors (e.g. when the urb is being killed on shutdown).

Signed-off-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agoUSB: cdc-acm: minimise no-suspend window during shutdown
Johan Hovold [Mon, 26 May 2014 17:23:48 +0000 (19:23 +0200)]
USB: cdc-acm: minimise no-suspend window during shutdown

Now that acm_set_control() handles runtime PM properly, the only
remaining reason for the PM operations in shutdown is to clear the
needs_remote_wakeup flag before the final put.

Note that this also means that we now need to grab the write_lock to
prevent racing with resume.

Signed-off-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agoUSB: cdc-acm: remove redundant disconnected test from shutdown
Johan Hovold [Mon, 26 May 2014 17:23:47 +0000 (19:23 +0200)]
USB: cdc-acm: remove redundant disconnected test from shutdown

Remove redundant disconnect test from shutdown(), which is never called
post disconnect() where we do synchronous hangup.

Signed-off-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agoUSB: cdc-acm: simplify runtime PM locking
Johan Hovold [Mon, 26 May 2014 17:23:46 +0000 (19:23 +0200)]
USB: cdc-acm: simplify runtime PM locking

We can simply the runtime PM locking as there's no need to check the
susp_count in the read path (at least not since killing the rx tasklet).

Specifically, the read urbs will never be resubmitted by the completion
handler when killed during suspend.

Signed-off-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agoUSB: cdc-acm: fix runtime PM imbalance at shutdown
Johan Hovold [Mon, 26 May 2014 17:23:45 +0000 (19:23 +0200)]
USB: cdc-acm: fix runtime PM imbalance at shutdown

Make sure only to decrement the PM counters if they were actually
incremented.

Note that the USB PM counter, but not necessarily the driver core PM
counter, is reset when the interface is unbound.

Fixes: 11ea859d64b6 ("USB: additional power savings for cdc-acm devices
that support remote wakeup")

Signed-off-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agoUSB: cdc-acm: fix I/O after failed open
Johan Hovold [Mon, 26 May 2014 17:23:44 +0000 (19:23 +0200)]
USB: cdc-acm: fix I/O after failed open

Make sure to kill any already submitted read urbs on read-urb submission
failures in open in order to prevent doing I/O for a closed port.

Fixes: 088c64f81284 ("USB: cdc-acm: re-write read processing")
Signed-off-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agoUSB: cdc-acm: fix failed open not being detected
Johan Hovold [Mon, 26 May 2014 17:23:43 +0000 (19:23 +0200)]
USB: cdc-acm: fix failed open not being detected

Fix errors during open not being returned to userspace. Specifically,
failed control-line manipulations or control or read urb submissions
would not be detected.

Fixes: 7fb57a019f94 ("USB: cdc-acm: Fix potential deadlock (lockdep
warning)")

Signed-off-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agoUSB: cdc-acm: fix open and suspend race
Johan Hovold [Mon, 26 May 2014 17:23:42 +0000 (19:23 +0200)]
USB: cdc-acm: fix open and suspend race

We must not do the usb_autopm_put_interface() before submitting the read
urbs or we might end up doing I/O to a suspended device.

Fixes: 088c64f81284 ("USB: cdc-acm: re-write read processing")
Signed-off-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agoUSB: cdc-acm: fix potential urb leak and PM imbalance in write
Johan Hovold [Mon, 26 May 2014 17:23:41 +0000 (19:23 +0200)]
USB: cdc-acm: fix potential urb leak and PM imbalance in write

Make sure to check return value of autopm get in write() in order to
avoid urb leak and PM counter imbalance on errors.

Fixes: 11ea859d64b6 ("USB: additional power savings for cdc-acm devices
that support remote wakeup")

Cc: <stable@vger.kernel.org> # v2.6.27
Signed-off-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agoUSB: cdc-acm: fix shutdown and suspend race
Johan Hovold [Mon, 26 May 2014 17:23:40 +0000 (19:23 +0200)]
USB: cdc-acm: fix shutdown and suspend race

We should stop I/O unconditionally at suspend rather than rely on the
tty-port initialised flag (which is set prior to stopping I/O during
shutdown) in order to prevent suspend returning with URBs still active.

Fixes: 11ea859d64b6 ("USB: additional power savings for cdc-acm devices
that support remote wakeup")

Cc: <stable@vger.kernel.org> # v2.6.27
Signed-off-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agoUSB: cdc-acm: fix runtime PM for control messages
Johan Hovold [Mon, 26 May 2014 17:23:39 +0000 (19:23 +0200)]
USB: cdc-acm: fix runtime PM for control messages

Fix runtime PM handling of control messages by adding the required PM
counter operations.

Fixes: 11ea859d64b6 ("USB: additional power savings for cdc-acm devices
that support remote wakeup")

Cc: <stable@vger.kernel.org> # v2.6.27
Signed-off-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agoUSB: cdc-acm: fix broken runtime suspend
Johan Hovold [Mon, 26 May 2014 17:23:38 +0000 (19:23 +0200)]
USB: cdc-acm: fix broken runtime suspend

The current ACM runtime-suspend implementation is broken in several
ways:

Firstly, it buffers only the first write request being made while
suspended -- any further writes are silently dropped.

Secondly, writes being dropped also leak write urbs, which are never
reclaimed (until the device is unbound).

Thirdly, even the single buffered write is not cleared at shutdown
(which may happen before the device is resumed), something which can
lead to another urb leak as well as a PM usage-counter leak.

Fix this by implementing a delayed-write queue using urb anchors and
making sure to discard the queue properly at shutdown.

Fixes: 11ea859d64b6 ("USB: additional power savings for cdc-acm devices
that support remote wakeup")

Reported-by: Xiao Jin <jin.xiao@intel.com>
Cc: <stable@vger.kernel.org> # v2.6.27
Signed-off-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agoUSB: cdc-acm: fix write and resume race
Johan Hovold [Mon, 26 May 2014 17:23:37 +0000 (19:23 +0200)]
USB: cdc-acm: fix write and resume race

Fix race between write() and resume() due to improper locking that could
lead to writes being reordered.

Resume must be done atomically and susp_count be protected by the
write_lock in order to prevent racing with write(). This could otherwise
lead to writes being reordered if write() grabs the write_lock after
susp_count is decremented, but before the delayed urb is submitted.

Fixes: 11ea859d64b6 ("USB: additional power savings for cdc-acm devices
that support remote wakeup")

Cc: <stable@vger.kernel.org> # v2.6.27
Signed-off-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agoUSB: cdc-acm: fix write and suspend race
Johan Hovold [Mon, 26 May 2014 17:23:36 +0000 (19:23 +0200)]
USB: cdc-acm: fix write and suspend race

Fix race between write() and suspend() which could lead to writes being
dropped (or I/O while suspended) if the device is runtime suspended
while a write request is being processed.

Specifically, suspend() releases the write_lock after determining the
device is idle but before incrementing the susp_count, thus leaving a
window where a concurrent write() can submit an urb.

Fixes: 11ea859d64b6 ("USB: additional power savings for cdc-acm devices
that support remote wakeup")

Cc: <stable@vger.kernel.org> # v2.6.27
Signed-off-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agoUSB: kobil_sct: fix control requests without data stage
Johan Hovold [Mon, 26 May 2014 17:23:35 +0000 (19:23 +0200)]
USB: kobil_sct: fix control requests without data stage

Fix incorrect pipe directions and remove bogus data buffer arguments
from control requests without data stage.

Signed-off-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agoUSB: serial: remove overly defensive port tests
Johan Hovold [Mon, 26 May 2014 17:23:34 +0000 (19:23 +0200)]
USB: serial: remove overly defensive port tests

The only way a port pointer may be NULL is if probe() failed, and in
that case neither disconnect(), resume(), or reset_resume() will be
called.

Signed-off-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agoUSB: serial: fix potential runtime pm imbalance at device remove
Johan Hovold [Mon, 26 May 2014 17:23:33 +0000 (19:23 +0200)]
USB: serial: fix potential runtime pm imbalance at device remove

Only call usb_autopm_put_interface() if the corresponding
usb_autopm_get_interface() was successful.

This prevents a potential runtime PM counter imbalance should
usb_autopm_get_interface() fail. Note that the USB PM usage counter is
reset when the interface is unbound, but that the runtime PM counter may
be left unbalanced.

Also add comment on why we don't need to worry about racing
resume/suspend on autopm_get failures.

Fixes: d5fd650cfc7f ("usb: serial: prevent suspend/resume from racing
against probe/remove")

Signed-off-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8 years agoUSB: usb_wwan: do not resume I/O on closing ports
Johan Hovold [Mon, 26 May 2014 17:23:32 +0000 (19:23 +0200)]
USB: usb_wwan: do not resume I/O on closing ports

Use tty-port initialised flag rather than private flag to determine when
port is closing down.

Since the tty-port flag is set prior to dropping DTR/RTS (when HUPCL is
set) this avoid submitting the read urbs when resuming the interface in
dtr_rts() only to immediately kill them again in shutdown().

Signed-off-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>