Merge tag 'driver-core-3.5-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core

Pull driver core updates from Greg Kroah-Hartman:
 "Here's the driver core, and other driver subsystems, pull request for
  the 3.5-rc1 merge window.

  Outside of a few minor driver core changes, we ended up with the
  following different subsystem and core changes as well, due to
  interdependancies on the driver core:
   - hyperv driver updates
   - drivers/memory being created and some drivers moved into it
   - extcon driver subsystem created out of the old Android staging
     switch driver code
   - dynamic debug updates
   - printk rework, and /dev/kmsg changes

  All of this has been tested in the linux-next releases for a few weeks
  with no reported problems.

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

Fix up conflicts in drivers/extcon/extcon-max8997.c where git noticed
that a patch to the deleted drivers/misc/max8997-muic.c driver needs to
be applied to this one.

* tag 'driver-core-3.5-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core: (90 commits)
  uio_pdrv_genirq: get irq through platform resource if not set otherwise
  memory: tegra{20,30}-mc: Remove empty *_remove()
  printk() - isolate KERN_CONT users from ordinary complete lines
  sysfs: get rid of some lockdep false positives
  Drivers: hv: util: Properly handle version negotiations.
  Drivers: hv: Get rid of an unnecessary check in vmbus_prep_negotiate_resp()
  memory: tegra{20,30}-mc: Use dev_err_ratelimited()
  driver core: Add dev_*_ratelimited() family
  Driver Core: don't oops with unregistered driver in driver_find_device()
  printk() - restore prefix/timestamp printing for multi-newline strings
  printk: add stub for prepend_timestamp()
  ARM: tegra30: Make MC optional in Kconfig
  ARM: tegra20: Make MC optional in Kconfig
  ARM: tegra30: MC: Remove unnecessary BUG*()
  ARM: tegra20: MC: Remove unnecessary BUG*()
  printk: correctly align __log_buf
  ARM: tegra30: Add Tegra Memory Controller(MC) driver
  ARM: tegra20: Add Tegra Memory Controller(MC) driver
  printk() - restore timestamp printing at console output
  printk() - do not merge continuation lines of different threads
  ...

1 files changed, 124 insertions, 0 deletions

diff --git a/Documentation/extcon/porting-android-switch-class b/Documentation/extcon/porting-android-switch-class
new file mode 100644
index 000000000000..eb0fa5f4fe88
--- /dev/null
+++ b/Documentation/extcon/porting-android-switch-class
@@ -0,0 +1,124 @@
+
+	Staging/Android Switch Class Porting Guide
+	(linux/drivers/staging/android/switch)
+	(c) Copyright 2012 Samsung Electronics
+
+AUTHORS
+MyungJoo Ham <myungjoo.ham@samsung.com>
+
+/*****************************************************************
+ * CHAPTER 1.                                                    *
+ * PORTING SWITCH CLASS DEVICE DRIVERS                           *
+ *****************************************************************/
+
+****** STEP 1. Basic Functionality
+	No extcon extended feature, but switch features only.
+
+- struct switch_dev (fed to switch_dev_register/unregister)
+    @name: no change
+    @dev: no change
+    @index: drop (not used in switch device driver side anyway)
+    @state: no change
+	If you have used @state with magic numbers, keep it
+	at this step.
+    @print_name: no change but type change (switch_dev->extcon_dev)
+    @print_state: no change but type change (switch_dev->extcon_dev)
+
+- switch_dev_register(sdev, dev)
+	=> extcon_dev_register(edev, dev)
+	: no change but type change (sdev->edev)
+- switch_dev_unregister(sdev)
+	=> extcon_dev_unregister(edev)
+	: no change but type change (sdev->edev)
+- switch_get_state(sdev)
+	=> extcon_get_state(edev)
+	: no change but type change (sdev->edev) and (return: int->u32)
+- switch_set_state(sdev, state)
+	=> extcon_set_state(edev, state)
+	: no change but type change (sdev->edev) and (state: int->u32)
+
+With this changes, the ex-switch extcon class device works as it once
+worked as switch class device. However, it will now have additional
+interfaces (both ABI and in-kernel API) and different ABI locations.
+However, if CONFIG_ANDROID is enabled without CONFIG_ANDROID_SWITCH,
+/sys/class/switch/* will be symbolically linked to /sys/class/extcon/
+so that they are still compatible with legacy userspace processes.
+
+****** STEP 2. Multistate (no more magic numbers in state value)
+	Extcon's extended features for switch device drivers with
+	complex features usually required magic numbers in state
+	value of switch_dev. With extcon, such magic numbers that
+	support multiple cables (
+
+  1. Define cable names at edev->supported_cable.
+  2. (Recommended) remove print_state callback.
+  3. Use extcon_get_cable_state_(edev, index) or
+   extcon_get_cable_state(edev, cable_name) instead of
+   extcon_get_state(edev) if you intend to get a state of a specific
+   cable. Same for set_state. This way, you can remove the usage of
+   magic numbers in state value.
+  4. Use extcon_update_state() if you are updating specific bits of
+   the state value.
+
+Example: a switch device driver w/ magic numbers for two cables.
+	"0x00": no cables connected.
+	"0x01": cable 1 connected
+	"0x02": cable 2 connected
+	"0x03": cable 1 and 2 connected
+  1. edev->supported_cable = {"1", "2", NULL};
+  2. edev->print_state = NULL;
+  3. extcon_get_cable_state_(edev, 0) shows cable 1's state.
+     extcon_get_cable_state(edev, "1") shows cable 1's state.
+     extcon_set_cable_state_(edev, 1) sets cable 2's state.
+     extcon_set_cable_state(edev, "2") sets cable 2's state
+  4. extcon_update_state(edev, 0x01, 0) sets the least bit's 0.
+
+****** STEP 3. Notify other device drivers
+
+  You can notify others of the cable attach/detach events with
+notifier chains.
+
+  At the side of other device drivers (the extcon device itself
+does not need to get notified of its own events), there are two
+methods to register notifier_block for cable events:
+(a) for a specific cable or (b) for every cable.
+
+  (a) extcon_register_interest(obj, extcon_name, cable_name, nb)
+	Example: want to get news of "MAX8997_MUIC"'s "USB" cable
+
+	obj = kzalloc(sizeof(struct extcon_specific_cable_nb),
+		      GFP_KERNEL);
+	nb->notifier_call = the_callback_to_handle_usb;
+
+	extcon_register_intereset(obj, "MAX8997_MUIC", "USB", nb);
+
+  (b) extcon_register_notifier(edev, nb)
+	Call nb for any changes in edev.
+
+  Please note that in order to properly behave with method (a),
+the extcon device driver should support multistate feature (STEP 2).
+
+****** STEP 4. Inter-cable relation (mutually exclusive)
+
+  You can provide inter-cable mutually exclusiveness information
+for an extcon device. When cables A and B are declared to be mutually
+exclusive, the two cables cannot be in ATTACHED state simulteneously.
+
+
+/*****************************************************************
+ * CHAPTER 2.                                                    *
+ * PORTING USERSPACE w/ SWITCH CLASS DEVICE SUPPORT              *
+ *****************************************************************/
+
+****** ABI Location
+
+  If "CONFIG_ANDROID" is enabled and "CONFIG_ANDROID_SWITCH" is
+disabled, /sys/class/switch/* are created as symbolic links to
+/sys/class/extcon/*. Because CONFIG_ANDROID_SWITCH creates
+/sys/class/switch directory, we disable symboling linking if
+CONFIG_ANDROID_SWITCH is enabled.
+
+  The two files of switch class, name and state, are provided with
+extcon, too. When the multistate support (STEP 2 of CHAPTER 1.) is
+not enabled or print_state callback is supplied, the output of
+state ABI is same with switch class.