platform/kernel/linux-starfive.git
7 years agoMAINTAINERS: update file entries for Coresight subsystem
Leo Yan [Mon, 5 Jun 2017 20:15:13 +0000 (14:15 -0600)]
MAINTAINERS: update file entries for Coresight subsystem

Update document file entries for Coresight debug module.

Signed-off-by: Leo Yan <leo.yan@linaro.org>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agodoc: Add coresight_cpu_debug.enable to kernel-parameters.txt
Leo Yan [Mon, 5 Jun 2017 20:15:12 +0000 (14:15 -0600)]
doc: Add coresight_cpu_debug.enable to kernel-parameters.txt

Add coresight_cpu_debug.enable to kernel-parameters.txt, this flag is
used to enable/disable the CPU sampling based debugging.

Signed-off-by: Leo Yan <leo.yan@linaro.org>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agodoc: Add documentation for Coresight CPU debug
Leo Yan [Mon, 5 Jun 2017 20:15:11 +0000 (14:15 -0600)]
doc: Add documentation for Coresight CPU debug

Add detailed documentation for Coresight CPU debug driver, which
contains the info for driver implementation, Mike Leach excellent
summary for "clock and power domain". At the end some examples on how
to enable the debugging functionality are provided.

Suggested-by: Mike Leach <mike.leach@linaro.org>
Signed-off-by: Leo Yan <leo.yan@linaro.org>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agocoresight: bindings for CPU debug module
Leo Yan [Mon, 5 Jun 2017 20:15:10 +0000 (14:15 -0600)]
coresight: bindings for CPU debug module

According to ARMv8 architecture reference manual (ARM DDI 0487A.k)
Chapter 'Part H: External debug', the CPU can integrate debug module
and it can support self-hosted debug and external debug. Especially
for supporting self-hosted debug, this means the program can access
the debug module from mmio region; and usually the mmio region is
integrated with coresight.

So add document for binding debug component, includes binding to APB
clock; and also need specify the CPU node which the debug module is
dedicated to specific CPU.

Suggested-by: Mike Leach <mike.leach@linaro.org>
Signed-off-by: Leo Yan <leo.yan@linaro.org>
Reviewed-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Acked-by: Rob Herring <robh@kernel.org>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agocoresight: tmc: Configure DMA mask appropriately
Robin Murphy [Mon, 5 Jun 2017 20:15:09 +0000 (14:15 -0600)]
coresight: tmc: Configure DMA mask appropriately

Before making any DMA API calls, the ETR driver should really be setting
its masks to ensure that DMA is possible. Especially since it can
address more than the 32-bit default mask set by the AMBA bus code.

Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Tested-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agocoresight: etb10: Fix a typo in a comment line
Markus Elfring [Mon, 5 Jun 2017 20:15:08 +0000 (14:15 -0600)]
coresight: etb10: Fix a typo in a comment line

Delete a character in this description for a condition check.

Signed-off-by: Markus Elfring <elfring@users.sourceforge.net>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agocoresight: etb10: Delete an error message for a failed memory allocation in etb_probe()
Markus Elfring [Mon, 5 Jun 2017 20:15:07 +0000 (14:15 -0600)]
coresight: etb10: Delete an error message for a failed memory allocation in etb_probe()

Omit an extra message for a memory allocation failure in this function.

This issue was detected by using the Coccinelle software.

Link: http://events.linuxfoundation.org/sites/events/files/slides/LCJ16-Refactor_Strings-WSang_0.pdf
Signed-off-by: Markus Elfring <elfring@users.sourceforge.net>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agocoresight: use const for device_node structures
Leo Yan [Mon, 5 Jun 2017 20:15:06 +0000 (14:15 -0600)]
coresight: use const for device_node structures

Almost low level functions from open firmware have used const to
qualify device_node structures, so add const for device_node
parameters in of_coresight related functions.

Signed-off-by: Leo Yan <leo.yan@linaro.org>
Reviewed-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agocoresight: tmc: minor fix for output log
Leo Yan [Mon, 5 Jun 2017 20:15:05 +0000 (14:15 -0600)]
coresight: tmc: minor fix for output log

In current code the output logs are not well symmetric for sink and link
enabling and disabling. This patch is to fix that so can output paired
logs.

Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Leo Yan <leo.yan@linaro.org>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agocoresight: etm_perf: Fix using uninitialised work
Suzuki K Poulose [Mon, 5 Jun 2017 20:15:04 +0000 (14:15 -0600)]
coresight: etm_perf: Fix using uninitialised work

With 4.11-rc4, the following command triggers a WARN_ON,
when a sink is not enabled.

 perf record -e cs_etm/@20010000.etf/

 [88286.547741] ------------[ cut here ]------------
 [88286.552332] WARNING: CPU: 3 PID: 2156 at kernel/workqueue.c:1442 __queue_work+0x29c/0x3b8
 [88286.560427] Modules linked in:
 [88286.563451]
 [88286.564928] CPU: 3 PID: 2156 Comm: perf_v4.11 Not tainted 4.11.0-rc4 #217
 [88286.573453] Hardware name: ARM LTD ARM Juno Development Platform/ARM Juno Development Platform, BIOS EDK II Aug 15
  2016
 [88286.584128] task: ffff80097597c200 task.stack: ffff8009768b0000
 [88286.589990] PC is at __queue_work+0x29c/0x3b8
 [88286.594303] LR is at __queue_work+0x104/0x3b8
 [88286.598614] pc : [<ffff0000080d8c7c>] lr : [<ffff0000080d8ae4>] pstate: a00001c5
 [88286.605934] sp : ffff8009768b3aa0
 [88286.609212] x29: ffff8009768b3aa0 x28: ffff80097ff3da00
 [88286.614477] x27: ffff80097ff89c00 x26: ffff8009751b0e00
 [88286.619741] x25: ffff000008c9f000 x24: 0000000000000003
 [88286.625004] x23: 0000000000000040 x22: ffff000008d3dab8
 [88286.630268] x21: ffff800977804400 x20: 0000000000000007
 [88286.635532] x19: ffff000008c54000 x18: 0000fffff9185160
 [88286.640795] x17: 0000ffffb33d9a38 x16: ffff000008088270
 [88286.646059] x15: 0000ffffb345b590 x14: 0000000000000000
 [88286.651322] x13: 0000000000000004 x12: 0000000000000040
 [88286.656586] x11: 0000000000000068 x10: 0000000000000000
 [88286.661849] x9 : ffff800977400028 x8 : 0000000000000000
 [88286.667113] x7 : 0000000000000000 x6 : ffff0000080d8ae4
 [88286.672376] x5 : 0000000000000000 x4 : 0000000000000080
 [88286.677639] x3 : 0000000000000000 x2 : 0000000000000000
 [88286.682903] x1 : 0000000000000000 x0 : ffff8009751b0e08
 [88286.688166]
 [88286.689638] ---[ end trace 31633f18fd33d4cb ]---
 [88286.694206] Call trace:
 [88286.696627] Exception stack(0xffff8009768b38d0 to 0xffff8009768b3a00)
 [88286.703004] 38c0:                                   ffff000008c54000 0001000000000000
 [88286.710757] 38e0: ffff8009768b3aa0 ffff0000080d8c7c ffff8009768b3b50 ffff80097ff8a5b0
 [88286.718511] 3900: 0000800977325000 0000000000000000 0000000000000040 ffff80097ffc6180
 [88286.726264] 3920: ffff8009768b3940 ffff0000088a8694 ffff80097ffc5800 0000000000000000
 [88286.734017] 3940: ffff8009768b3960 ffff0000081919c0 ffff80097ffc5280 0000000000000001
 [88286.741771] 3960: ffff8009768b3a50 ffff00000819206c ffff8009751b0e08 0000000000000000
 [88286.749523] 3980: 0000000000000000 0000000000000000 0000000000000080 0000000000000000
 [88286.757277] 39a0: ffff0000080d8ae4 0000000000000000 0000000000000000 ffff800977400028
 [88286.765029] 39c0: 0000000000000000 0000000000000068 0000000000000040 0000000000000004
 [88286.772783] 39e0: 0000000000000000 0000ffffb345b590 ffff000008088270 0000ffffb33d9a38
 [88286.780537] [<ffff0000080d8c7c>] __queue_work+0x29c/0x3b8
 [88286.785883] [<ffff0000080d8df8>] queue_work_on+0x60/0x78
 [88286.791146] [<ffff000008764c68>] etm_setup_aux+0x178/0x238
 [88286.796578] [<ffff000008183600>] rb_alloc_aux+0x228/0x310
 [88286.801925] [<ffff00000817e564>] perf_mmap+0x404/0x5a8
 [88286.807015] [<ffff0000081c60cc>] mmap_region+0x394/0x5c0
 [88286.812276] [<ffff0000081c654c>] do_mmap+0x254/0x388
 [88286.817191] [<ffff0000081a989c>] vm_mmap_pgoff+0xbc/0xe0
 [88286.822452] [<ffff0000081c3ffc>] SyS_mmap_pgoff+0xac/0x228
 [88286.827884] [<ffff000008088288>] sys_mmap+0x18/0x28
 [88286.832714] [<ffff000008082f30>] el0_svc_naked+0x24/0x28

The patch makes sure that the event_data->work is initialised
properly before we could possibly use it.

Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Tested-by: Mike Leach <mike.leach@linaro.org>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agocoresight: Fix reference count for software sources
Suzuki K Poulose [Mon, 5 Jun 2017 20:15:03 +0000 (14:15 -0600)]
coresight: Fix reference count for software sources

For software sources (i.e STM), there could be multiple agents
generating the trace data, unlike the ETMs. So we need to
properly do the accounting for the active number of users
to disable the device when the last user goes away. Right
now, the reference counting is broken for sources as we skip
the actions when we detect that the source is enabled.

This patch fixes the problem by adding the refcounting for
software sources, even when they are enabled.

Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Reported-by: Robert Walker <robert.walker@arm.com>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agocoresight: Disable the path only when the source is disabled
Suzuki K Poulose [Mon, 5 Jun 2017 20:15:02 +0000 (14:15 -0600)]
coresight: Disable the path only when the source is disabled

With a coresight tracing session, the components along the path
from the source to sink are disabled after the source is disabled.
However, if the source was not actually disabled due to active
users, we should not disable the components in the path.

Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agothunderbolt: fix spelling mistake: "missmatch" -> "mismatch"
Colin Ian King [Thu, 18 May 2017 07:42:49 +0000 (08:42 +0100)]
thunderbolt: fix spelling mistake: "missmatch" -> "mismatch"

Trivial fix to spelling mistake in tb_sw_warn warning message

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoMAINTAINERS: Add maintainers for Thunderbolt driver
Mika Westerberg [Tue, 6 Jun 2017 12:25:19 +0000 (15:25 +0300)]
MAINTAINERS: Add maintainers for Thunderbolt driver

We will be helping Andreas to maintain the Thunderbolt driver.

Signed-off-by: Michael Jamet <michael.jamet@intel.com>
Signed-off-by: Yehezkel Bernat <yehezkel.bernat@intel.com>
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Andreas Noever <andreas.noever@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agothunderbolt: Add documentation how Thunderbolt bus can be used
Mika Westerberg [Tue, 6 Jun 2017 12:25:18 +0000 (15:25 +0300)]
thunderbolt: Add documentation how Thunderbolt bus can be used

Since there are no such tool yet that handles all the low-level details
of connecting devices and upgrading their firmware, add a small document
that shows how the Thunderbolt bus can be used directly from command
line.

Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reviewed-by: Yehezkel Bernat <yehezkel.bernat@intel.com>
Reviewed-by: Michael Jamet <michael.jamet@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Andreas Noever <andreas.noever@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agothunderbolt: Add support for host and device NVM firmware upgrade
Mika Westerberg [Tue, 6 Jun 2017 12:25:17 +0000 (15:25 +0300)]
thunderbolt: Add support for host and device NVM firmware upgrade

Starting from Intel Falcon Ridge the NVM firmware can be upgraded by
using DMA configuration based mailbox commands. If we detect that the
host or device (device support starts from Intel Alpine Ridge) has the
DMA configuration based mailbox we expose NVM information to the
userspace as two separate Linux NVMem devices: nvm_active and
nvm_non_active. The former is read-only portion of the active NVM which
firmware upgrade tools can be use to find out suitable NVM image if the
device identification strings are not enough.

The latter is write-only portion where the new NVM image is to be
written by the userspace. It is up to the userspace to find out right
NVM image (the kernel does very minimal validation). The ICM firmware
itself authenticates the new NVM firmware and fails the operation if it
is not what is expected.

We also expose two new sysfs files per each switch: nvm_version and
nvm_authenticate which can be used to read the active NVM version and
start the upgrade process.

We also introduce safe mode which is the mode a switch goes when it does
not have properly authenticated firmware. In this mode the switch only
accepts a couple of commands including flashing a new NVM firmware image
and triggering power cycle.

This code is based on the work done by Amir Levy and Michael Jamet.

Signed-off-by: Michael Jamet <michael.jamet@intel.com>
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reviewed-by: Yehezkel Bernat <yehezkel.bernat@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Andreas Noever <andreas.noever@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agothunderbolt: Add support for Internal Connection Manager (ICM)
Mika Westerberg [Tue, 6 Jun 2017 12:25:16 +0000 (15:25 +0300)]
thunderbolt: Add support for Internal Connection Manager (ICM)

Starting from Intel Falcon Ridge the internal connection manager running
on the Thunderbolt host controller has been supporting 4 security
levels. One reason for this is to prevent DMA attacks and only allow
connecting devices the user trusts.

The internal connection manager (ICM) is the preferred way of connecting
Thunderbolt devices over software only implementation typically used on
Macs. The driver communicates with ICM using special Thunderbolt ring 0
(control channel) messages. In order to handle these messages we add
support for the ICM messages to the control channel.

The security levels are as follows:

  none - No security, all tunnels are created automatically
  user - User needs to approve the device before tunnels are created
  secure - User need to approve the device before tunnels are created.
   The device is sent a challenge on future connects to be able
   to verify it is actually the approved device.
  dponly - Only Display Port and USB tunnels can be created and those
           are created automatically.

The security levels are typically configurable from the system BIOS and
by default it is set to "user" on many systems.

In this patch each Thunderbolt device will have either one or two new
sysfs attributes: authorized and key. The latter appears for devices
that support secure connect.

In order to identify the device the user can read identication
information, including UUID and name of the device from sysfs and based
on that make a decision to authorize the device. The device is
authorized by simply writing 1 to the "authorized" sysfs attribute. This
is following the USB bus device authorization mechanism. The secure
connect requires an additional challenge step (writing 2 to the
"authorized" attribute) in future connects when the key has already been
stored to the NVM of the device.

Non-ICM systems (before Alpine Ridge) continue to use the existing
functionality and the security level is set to none. For systems with
Alpine Ridge, even on Apple hardware, we will use ICM.

This code is based on the work done by Amir Levy and Michael Jamet.

Signed-off-by: Michael Jamet <michael.jamet@intel.com>
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reviewed-by: Yehezkel Bernat <yehezkel.bernat@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Andreas Noever <andreas.noever@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agothunderbolt: Do not touch the hardware if the NHI is gone on resume
Mika Westerberg [Tue, 6 Jun 2017 12:25:15 +0000 (15:25 +0300)]
thunderbolt: Do not touch the hardware if the NHI is gone on resume

On PCs the NHI host controller is only present when there is a device
connected. When the last device is disconnected the host controller will
dissappear shortly (within 10s). Now if that happens when we are
suspended we should not try to touch the hardware anymore, so add a flag
for this and check it before we re-enable rings.

Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reviewed-by: Yehezkel Bernat <yehezkel.bernat@intel.com>
Reviewed-by: Michael Jamet <michael.jamet@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Andreas Noever <andreas.noever@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agothunderbolt: Add support for DMA configuration based mailbox
Mika Westerberg [Tue, 6 Jun 2017 12:25:14 +0000 (15:25 +0300)]
thunderbolt: Add support for DMA configuration based mailbox

The DMA (NHI) port of a switch provides access to the NVM of the host
controller (and devices starting from Intel Alpine Ridge). The NVM
contains also more complete DROM for the root switch including vendor
and device identification strings.

This will look for the DMA port capability for each switch and if found
populates sw->dma_port. We then teach tb_drom_read() to read the DROM
information from NVM if available for the root switch.

The DMA port capability also supports upgrading the NVM for both host
controller and devices which will be added in subsequent patches.

This code is based on the work done by Amir Levy and Michael Jamet.

Signed-off-by: Michael Jamet <michael.jamet@intel.com>
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reviewed-by: Yehezkel Bernat <yehezkel.bernat@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Andreas Noever <andreas.noever@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agothunderbolt: Store Thunderbolt generation in the switch structure
Mika Westerberg [Tue, 6 Jun 2017 12:25:13 +0000 (15:25 +0300)]
thunderbolt: Store Thunderbolt generation in the switch structure

In some cases it is useful to know what is the Thunderbolt generation
the switch supports. This introduces a new field to struct switch that
stores the generation of the switch based on the device ID. Unknown
switches (there should be none) are assumed to be first generation to be
on the safe side.

Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reviewed-by: Yehezkel Bernat <yehezkel.bernat@intel.com>
Reviewed-by: Michael Jamet <michael.jamet@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Andreas Noever <andreas.noever@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agothunderbolt: Add support for NHI mailbox
Mika Westerberg [Tue, 6 Jun 2017 12:25:12 +0000 (15:25 +0300)]
thunderbolt: Add support for NHI mailbox

The host controller includes two sets of registers that are used to
communicate with the firmware. Add functions that can be used to access
these registers.

This code is based on the work done by Amir Levy and Michael Jamet.

Signed-off-by: Michael Jamet <michael.jamet@intel.com>
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reviewed-by: Yehezkel Bernat <yehezkel.bernat@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Andreas Noever <andreas.noever@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agothunderbolt: Add new Thunderbolt PCI IDs
Mika Westerberg [Tue, 6 Jun 2017 12:25:11 +0000 (15:25 +0300)]
thunderbolt: Add new Thunderbolt PCI IDs

Add Intel Win Ridge (Thunderbolt 2) and Alpine Ridge (Thunderbolt 3)
controller PCI IDs to the list of supported devices.

This code is based on the work done by Amir Levy and Michael Jamet.

Signed-off-by: Michael Jamet <michael.jamet@intel.com>
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reviewed-by: Yehezkel Bernat <yehezkel.bernat@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Andreas Noever <andreas.noever@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agothunderbolt: Rework control channel to be more reliable
Mika Westerberg [Tue, 6 Jun 2017 12:25:10 +0000 (15:25 +0300)]
thunderbolt: Rework control channel to be more reliable

If a request times out the response might arrive right after the request
is failed. This response is pushed to the kfifo and next request will
read it instead. Since it most likely will not pass our validation
checks in parse_header() the next request will fail as well, and
response to that request will be pushed to the kfifo, ad infinitum.

We end up in a situation where all requests fail and no devices can be
added anymore until the driver is unloaded and reloaded again.

To overcome this, rework the control channel so that we will have a
queue of outstanding requests. Each request will be handled in turn and
the response is validated against what is expected. Unexpected packets
(for example responses for requests that have been timed out) are
dropped. This model is copied from Greybus implementation with small
changes here and there to get it cope with Thunderbolt control packets.

In addition the configuration packets support sequence number which the
switch is supposed to copy from the request to response. We use this to
drop responses that are already timed out. Taking advantage of the
sequence number, we automatically retry configuration read/write 4 times
before giving up.

Also timeout is not a programming error so there is no need to trigger a
scary backtrace (WARN), instead we just log a warning.  After all
Thunderbolt devices are hot-pluggable by definition which means user can
unplug a device any time and that is totally acceptable.

With this change there is no need to take the global domain lock when
sending configuration packets anymore. This is useful when we add
support for cross-domain (XDomain) communication later on.

Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reviewed-by: Yehezkel Bernat <yehezkel.bernat@intel.com>
Reviewed-by: Michael Jamet <michael.jamet@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Andreas Noever <andreas.noever@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agothunderbolt: Let the connection manager handle all notifications
Mika Westerberg [Tue, 6 Jun 2017 12:25:09 +0000 (15:25 +0300)]
thunderbolt: Let the connection manager handle all notifications

Currently the control channel (ctl.c) handles the one supported
notification (PLUG_EVENT) and sends back ACK accordingly. However, we
are going to add support for the internal connection manager (ICM) that
needs to handle a different notifications. So instead of dealing
everything in the control channel, we change the callback to take an
arbitrary thunderbolt packet and convert the native connection manager
to handle the event itself.

In addition we only push replies we know of to the response FIFO.
Everything else is treated as notification (or request) and is expected
to be dealt by the connection manager implementation.

Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reviewed-by: Yehezkel Bernat <yehezkel.bernat@intel.com>
Reviewed-by: Michael Jamet <michael.jamet@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Andreas Noever <andreas.noever@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agothunderbolt: Expose make_header() to other files
Mika Westerberg [Tue, 6 Jun 2017 12:25:08 +0000 (15:25 +0300)]
thunderbolt: Expose make_header() to other files

We will be using this function in files introduced in subsequent
patches. While there the function is renamed to tb_cfg_make_header()
following tb_cfg_get_route().

Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reviewed-by: Yehezkel Bernat <yehezkel.bernat@intel.com>
Reviewed-by: Michael Jamet <michael.jamet@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Andreas Noever <andreas.noever@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agothunderbolt: Expose get_route() to other files
Mika Westerberg [Tue, 6 Jun 2017 12:25:07 +0000 (15:25 +0300)]
thunderbolt: Expose get_route() to other files

We are going to use it when we change the connection manager to handle
events itself. Also rename it to follow naming convention used in
functions exposed in ctl.h.

Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reviewed-by: Yehezkel Bernat <yehezkel.bernat@intel.com>
Reviewed-by: Michael Jamet <michael.jamet@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Andreas Noever <andreas.noever@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agothunderbolt: Move control channel messages to tb_msgs.h
Mika Westerberg [Tue, 6 Jun 2017 12:25:06 +0000 (15:25 +0300)]
thunderbolt: Move control channel messages to tb_msgs.h

We will be forwarding notifications received from the control channel to
the connection manager implementations. This way they can decide what to
do if anything when a notification is received.

To be able to use control channel messages from other files, move them
to tb_msgs.h.

No functional changes intended.

Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reviewed-by: Yehezkel Bernat <yehezkel.bernat@intel.com>
Reviewed-by: Michael Jamet <michael.jamet@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Andreas Noever <andreas.noever@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agothunderbolt: Read vendor and device name from DROM
Mika Westerberg [Tue, 6 Jun 2017 12:25:05 +0000 (15:25 +0300)]
thunderbolt: Read vendor and device name from DROM

The device DROM contains name of the vendor and device among other
things. Extract this information and expose it to the userspace via two
new attributes.

Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reviewed-by: Yehezkel Bernat <yehezkel.bernat@intel.com>
Reviewed-by: Michael Jamet <michael.jamet@intel.com>
Signed-off-by: Andreas Noever <andreas.noever@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agothunderbolt: Refactor and fix parsing of port drom entries
Lukas Wunner [Tue, 6 Jun 2017 12:25:04 +0000 (15:25 +0300)]
thunderbolt: Refactor and fix parsing of port drom entries

Currently tb_drom_parse_entry() is only able to parse drom entries of
type TB_DROM_ENTRY_PORT. Rename it to tb_drom_parse_entry_port().
Fold tb_drom_parse_port_entry() into it.

Its return value is currently ignored. Evaluate it and abort parsing on
error.

Change tb_drom_parse_entries() to accommodate for parsing of other entry
types than TB_DROM_ENTRY_PORT.

Signed-off-by: Lukas Wunner <lukas@wunner.de>
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Andreas Noever <andreas.noever@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agothunderbolt: Do not fail if DROM data CRC32 is invalid
Mika Westerberg [Tue, 6 Jun 2017 12:25:03 +0000 (15:25 +0300)]
thunderbolt: Do not fail if DROM data CRC32 is invalid

There are devices out there where CRC32 of the DROM is not correct. One
reason for this is that the ICM firmware does not validate it and it
seems that neither does the Apple driver. To be able to support such
devices we continue parsing the DROM contents regardless of whether
CRC32 failed or not. We still keep the warning there.

Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reviewed-by: Yehezkel Bernat <yehezkel.bernat@intel.com>
Reviewed-by: Michael Jamet <michael.jamet@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Andreas Noever <andreas.noever@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agothunderbolt: Fail switch adding operation if reading DROM fails
Mika Westerberg [Tue, 6 Jun 2017 12:25:02 +0000 (15:25 +0300)]
thunderbolt: Fail switch adding operation if reading DROM fails

All non-root switches are expected to have DROM so if the operation
fails, it might be due the user unlugging the device. There is no point
continuing adding the switch further in that case. Just bail out.

For root switches (hosts) the DROM is either retrieved from a EFI
variable, NVM or hard-coded.

Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reviewed-by: Yehezkel Bernat <yehezkel.bernat@intel.com>
Reviewed-by: Michael Jamet <michael.jamet@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Andreas Noever <andreas.noever@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agothunderbolt: Convert switch to a device
Mika Westerberg [Tue, 6 Jun 2017 12:25:01 +0000 (15:25 +0300)]
thunderbolt: Convert switch to a device

Thunderbolt domain consists of switches that are connected to each
other, forming a bus. This will convert each switch into a real Linux
device structure and adds them to the domain. The advantage here is
that we get all the goodies from the driver core, like reference
counting and sysfs hierarchy for free.

Also expose device identification information to the userspace via new
sysfs attributes.

In order to support internal connection manager (ICM) we separate switch
configuration into its own function (tb_switch_configure()) which is
only called by the existing native connection manager implementation
used on Macs.

Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reviewed-by: Yehezkel Bernat <yehezkel.bernat@intel.com>
Reviewed-by: Michael Jamet <michael.jamet@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Andreas Noever <andreas.noever@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agothunderbolt: Introduce thunderbolt bus and connection manager
Mika Westerberg [Tue, 6 Jun 2017 12:25:00 +0000 (15:25 +0300)]
thunderbolt: Introduce thunderbolt bus and connection manager

Thunderbolt fabric consists of one or more switches. This fabric is
called domain and it is controlled by an entity called connection
manager. The connection manager can be either internal (driven by a
firmware running on the host controller) or external (software driver).
This driver currently implements support for the latter.

In order to manage switches and their properties more easily we model
this domain structure as a Linux bus. Each host controller adds a domain
device to this bus, and these devices are named as domainN where N
stands for index or id of the current domain.

We then abstract connection manager specific operations into a new
structure tb_cm_ops and convert the existing tb.c to fill those
accordingly. This makes it easier to add support for the internal
connection manager in subsequent patches.

Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reviewed-by: Yehezkel Bernat <yehezkel.bernat@intel.com>
Reviewed-by: Michael Jamet <michael.jamet@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Andreas Noever <andreas.noever@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agothunderbolt: Allow passing NULL to tb_ctl_free()
Mika Westerberg [Tue, 6 Jun 2017 12:24:59 +0000 (15:24 +0300)]
thunderbolt: Allow passing NULL to tb_ctl_free()

Following the usual pattern used in many places, we allow passing NULL
pointer to tb_ctl_free(). Then the user can call the function regardless
if it has allocated control channel or not making the code bit simpler.

Suggested-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Signed-off-by: Andreas Noever <andreas.noever@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agothunderbolt: Rework capability handling
Mika Westerberg [Tue, 6 Jun 2017 12:24:58 +0000 (15:24 +0300)]
thunderbolt: Rework capability handling

Organization of the capabilities in switches and ports is not so random
after all. Rework the capability handling functionality so that it
follows how capabilities are organized and provide two new functions
(tb_switch_find_vse_cap() and tb_port_find_cap()) which can be used to
extract capabilities for ports and switches. Then convert the current
users over these.

Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reviewed-by: Yehezkel Bernat <yehezkel.bernat@intel.com>
Reviewed-by: Michael Jamet <michael.jamet@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Andreas Noever <andreas.noever@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agothunderbolt: Add MSI-X support
Mika Westerberg [Tue, 6 Jun 2017 12:24:57 +0000 (15:24 +0300)]
thunderbolt: Add MSI-X support

Intel Thunderbolt controllers support up to 16 MSI-X vectors. Using
MSI-X is preferred over MSI or legacy interrupt and may bring additional
performance because there is no need to check the status registers which
interrupt was triggered.

While there we convert comments in structs tb_ring and tb_nhi to follow
kernel-doc format more closely.

This code is based on the work done by Amir Levy and Michael Jamet.

Signed-off-by: Michael Jamet <michael.jamet@intel.com>
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reviewed-by: Yehezkel Bernat <yehezkel.bernat@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Andreas Noever <andreas.noever@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agothunderbolt: Do not warn about newer DROM versions
Mika Westerberg [Tue, 6 Jun 2017 12:24:56 +0000 (15:24 +0300)]
thunderbolt: Do not warn about newer DROM versions

DROM version 2 is compatible with the previous generation so no need to
warn about that.

Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reviewed-by: Yehezkel Bernat <yehezkel.bernat@intel.com>
Reviewed-by: Michael Jamet <michael.jamet@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Andreas Noever <andreas.noever@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agothunderbolt: Do not try to read UID if DROM offset is read as 0
Mika Westerberg [Tue, 6 Jun 2017 12:24:55 +0000 (15:24 +0300)]
thunderbolt: Do not try to read UID if DROM offset is read as 0

At least Falcon Ridge when in host mode does not have any kind of DROM
available and reading DROM offset returns 0 for these. Do not try to
read DROM any further in that case.

Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reviewed-by: Yehezkel Bernat <yehezkel.bernat@intel.com>
Reviewed-by: Michael Jamet <michael.jamet@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Andreas Noever <andreas.noever@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agothunderbolt: No need to read UID of the root switch on resume
Mika Westerberg [Tue, 6 Jun 2017 12:24:54 +0000 (15:24 +0300)]
thunderbolt: No need to read UID of the root switch on resume

The root switch is part of the host controller and cannot be physically
removed, so there is no point of reading UID again on resume in order to
check if the root switch is still the same.

Suggested-by: Andreas Noever <andreas.noever@gmail.com>
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Signed-off-by: Andreas Noever <andreas.noever@gmail.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agothunderbolt: Use const buffer pointer in write operations
Mika Westerberg [Tue, 6 Jun 2017 12:24:53 +0000 (15:24 +0300)]
thunderbolt: Use const buffer pointer in write operations

These functions should not (and do not) modify the argument in any way
so make it const.

Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reviewed-by: Yehezkel Bernat <yehezkel.bernat@intel.com>
Reviewed-by: Michael Jamet <michael.jamet@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Andreas Noever <andreas.noever@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agomux: mmio-based syscon mux controller
Philipp Zabel [Sun, 14 May 2017 19:51:16 +0000 (21:51 +0200)]
mux: mmio-based syscon mux controller

This adds a driver for mmio-based syscon multiplexers controlled by
bitfields in a syscon register range.

Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Signed-off-by: Peter Rosin <peda@axentia.se>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agodt-bindings: add mmio-based syscon mux controller DT bindings
Philipp Zabel [Sun, 14 May 2017 19:51:15 +0000 (21:51 +0200)]
dt-bindings: add mmio-based syscon mux controller DT bindings

This adds device tree binding documentation for mmio-based syscon
multiplexers controlled by a bitfields in a syscon register range.

Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Acked-by: Rob Herring <robh@kernel.org>
Signed-off-by: Peter Rosin <peda@axentia.se>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agomux: adg792a: add mux controller driver for ADG792A/G
Peter Rosin [Sun, 14 May 2017 19:51:14 +0000 (21:51 +0200)]
mux: adg792a: add mux controller driver for ADG792A/G

Analog Devices ADG792A/G is a triple 4:1 mux.

Reviewed-by: Jonathan Cameron <jic23@kernel.org>
Signed-off-by: Peter Rosin <peda@axentia.se>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agodt-bindings: mux-adg792a: document devicetree bindings for ADG792A/G mux
Peter Rosin [Sun, 14 May 2017 19:51:13 +0000 (21:51 +0200)]
dt-bindings: mux-adg792a: document devicetree bindings for ADG792A/G mux

Analog Devices ADG792A/G is a triple 4:1 mux.

Acked-by: Jonathan Cameron <jic23@kernel.org>
Reviewed-by: Rob Herring <robh@kernel.org>
Signed-off-by: Peter Rosin <peda@axentia.se>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoi2c: i2c-mux-gpmux: new driver
Peter Rosin [Sun, 14 May 2017 19:51:12 +0000 (21:51 +0200)]
i2c: i2c-mux-gpmux: new driver

This is a general purpose i2c mux that uses a multiplexer controlled by
the multiplexer subsystem to do the muxing.

The user can select if the mux is to be mux-locked and parent-locked
as described in Documentation/i2c/i2c-topology.

Acked-by: Jonathan Cameron <jic23@kernel.org>
Acked-by: Wolfram Sang <wsa@the-dreams.de>
Signed-off-by: Peter Rosin <peda@axentia.se>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agodt-bindings: i2c: i2c-mux: document general purpose i2c-mux bindings
Peter Rosin [Sun, 14 May 2017 19:51:11 +0000 (21:51 +0200)]
dt-bindings: i2c: i2c-mux: document general purpose i2c-mux bindings

Describe how a general purpose multiplexer controller is used to mux an
i2c bus.

Acked-by: Jonathan Cameron <jic23@kernel.org>
Reviewed-by: Rob Herring <robh@kernel.org>
Signed-off-by: Peter Rosin <peda@axentia.se>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoiio: multiplexer: new iio category and iio-mux driver
Peter Rosin [Sun, 14 May 2017 19:51:10 +0000 (21:51 +0200)]
iio: multiplexer: new iio category and iio-mux driver

When a multiplexer changes how an iio device behaves (for example
by feeding different signals to an ADC), this driver can be used
to create one virtual iio channel for each multiplexer state.

Depends on the generic multiplexer subsystem.

Cache any ext_info values from the parent iio channel, creating a private
copy of the ext_info attributes for each multiplexer state/channel.

Reviewed-by: Jonathan Cameron <jic23@kernel.org>
Signed-off-by: Peter Rosin <peda@axentia.se>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agodt-bindings: iio: io-channel-mux: document io-channel-mux bindings
Peter Rosin [Sun, 14 May 2017 19:51:09 +0000 (21:51 +0200)]
dt-bindings: iio: io-channel-mux: document io-channel-mux bindings

Describe how a multiplexer can be used to select which signal is fed to
an io-channel.

Acked-by: Jonathan Cameron <jic23@kernel.org>
Acked-by: Rob Herring <robh@kernel.org>
Signed-off-by: Peter Rosin <peda@axentia.se>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoiio: inkern: api for manipulating ext_info of iio channels
Peter Rosin [Sun, 14 May 2017 19:51:08 +0000 (21:51 +0200)]
iio: inkern: api for manipulating ext_info of iio channels

Extend the inkern api with functions for reading and writing ext_info
of iio channels.

Acked-by: Jonathan Cameron <jic23@kernel.org>
Signed-off-by: Peter Rosin <peda@axentia.se>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agomux: gpio: add mux controller driver for gpio based multiplexers
Peter Rosin [Sun, 14 May 2017 19:51:07 +0000 (21:51 +0200)]
mux: gpio: add mux controller driver for gpio based multiplexers

The driver builds a single multiplexer controller using a number
of gpio pins. For N pins, there will be 2^N possible multiplexer
states. The GPIO pins can be connected (by the hardware) to several
multiplexers, which in that case will be operated in parallel.

Reviewed-by: Jonathan Cameron <jic23@kernel.org>
Signed-off-by: Peter Rosin <peda@axentia.se>
Reviewed-by: Philipp Zabel <p.zabel@pengutronix.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agomux: minimal mux subsystem
Peter Rosin [Sun, 14 May 2017 19:51:06 +0000 (21:51 +0200)]
mux: minimal mux subsystem

Add a new minimalistic subsystem that handles multiplexer controllers.
When multiplexers are used in various places in the kernel, and the
same multiplexer controller can be used for several independent things,
there should be one place to implement support for said multiplexer
controller.

A single multiplexer controller can also be used to control several
parallel multiplexers, that are in turn used by different subsystems
in the kernel, leading to a need to coordinate multiplexer accesses.
The multiplexer subsystem handles this coordination.

Thanks go out to Lars-Peter Clausen, Jonathan Cameron, Rob Herring,
Wolfram Sang, Paul Gortmaker, Dan Carpenter, Colin Ian King, Greg
Kroah-Hartman and last but certainly not least to Philipp Zabel for
helpful comments, reviews, patches and general encouragement!

Reviewed-by: Jonathan Cameron <jic23@kernel.org>
Signed-off-by: Peter Rosin <peda@axentia.se>
Reviewed-by: Philipp Zabel <p.zabel@pengutronix.de>
Tested-by: Philipp Zabel <p.zabel@pengutronix.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agodt-bindings: document devicetree bindings for mux-controllers and gpio-mux
Peter Rosin [Sun, 14 May 2017 19:51:05 +0000 (21:51 +0200)]
dt-bindings: document devicetree bindings for mux-controllers and gpio-mux

Allow specifying that a single multiplexer controller can be used to
control several parallel multiplexers, thus enabling sharing of the
multiplexer controller by different consumers.

Add a binding for a first mux controller in the form of a GPIO based mux
controller.

Acked-by: Jonathan Cameron <jic23@kernel.org>
Acked-by: Rob Herring <robh@kernel.org>
Signed-off-by: Peter Rosin <peda@axentia.se>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agodevres: trivial whitespace fix
Peter Rosin [Sun, 14 May 2017 19:51:04 +0000 (21:51 +0200)]
devres: trivial whitespace fix

Everything else is indented with two spaces, so fix the odd one out.

Acked-by: Jonathan Cameron <jic23@kernel.org>
Signed-off-by: Peter Rosin <peda@axentia.se>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agodrivers/misc: add Aspeed LPC snoop driver
Robert Lippert [Fri, 2 Jun 2017 21:53:22 +0000 (14:53 -0700)]
drivers/misc: add Aspeed LPC snoop driver

This driver enables the LPC snoop hardware on the ASPEED BMC
which generates an interrupt upon every write to an I/O port
by the host.

This is typically used to monitor BIOS boot progress by listening
to well-known debug port 80h.

The functionality in this commit just saves all snooped values
to a circular 2K buffer in the kernel, subsequent commits can
act on the values to do things with them.

Signed-off-by: Robert Lippert <rlippert@google.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agomemory: ti-aemif: Handle return value of clk_prepare_enable
Arvind Yadav [Wed, 31 May 2017 10:25:54 +0000 (15:55 +0530)]
memory: ti-aemif: Handle return value of clk_prepare_enable

clk_prepare_enable() can fail here and we must check its return value.

Signed-off-by: Arvind Yadav <arvind.yadav.cs@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agospmi: spmi-pmic-arb: enable the SPMI interrupt as a wakeup source
Kiran Gunda [Wed, 10 May 2017 14:25:41 +0000 (19:55 +0530)]
spmi: spmi-pmic-arb: enable the SPMI interrupt as a wakeup source

Currently the SPMI interrupt will not wake the device. Enable this
interrupt as a wakeup source.

Signed-off-by: Nicholas Troast <ntroast@codeaurora.org>
Signed-off-by: Kiran Gunda <kgunda@codeaurora.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agospmi: pmic_arb: add support for PMIC bus arbiter v3
Abhijeet Dharmapurikar [Wed, 10 May 2017 14:25:40 +0000 (19:55 +0530)]
spmi: pmic_arb: add support for PMIC bus arbiter v3

PMIC bus arbiter v3 supports 512 SPMI peripherals. Add the v3 operators to
support this new arbiter version.

Signed-off-by: Abhijeet Dharmapurikar <adharmap@codeaurora.org>
Signed-off-by: Kiran Gunda <kgunda@codeaurora.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agospmi: pmic-arb: check apid enabled before calling the handler
Abhijeet Dharmapurikar [Wed, 10 May 2017 14:25:39 +0000 (19:55 +0530)]
spmi: pmic-arb: check apid enabled before calling the handler

The driver currently invokes the apid handler (periph_handler())
once it sees that the summary status bit for that apid is set.

However the hardware is designed to set that bit even if the apid
interrupts are disabled. The driver should check whether the apid
is indeed enabled before calling the apid handler.

Signed-off-by: Abhijeet Dharmapurikar <adharmap@codeaurora.org>
Signed-off-by: Kiran Gunda <kgunda@codeaurora.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agospmi: pmic_arb: use appropriate flow handler
Abhijeet Dharmapurikar [Wed, 10 May 2017 14:25:38 +0000 (19:55 +0530)]
spmi: pmic_arb: use appropriate flow handler

The current code uses handle_level_irq flow handler even if the
trigger type of the interrupt is edge. This can lead to missing
of an edge transition that happens when the interrupt is being
handled. The level flow handler masks the interrupt while it is
being handled, so if an edge transition happens at that time,
that edge is lost.

Use an edge flow handler for edge type interrupts which ensures
that the interrupt stays enabled while being handled - at least
until it triggers at which point the flow handler sets the
IRQF_PENDING flag and only then masks the interrupt. That
IRQF_PENDING state indicates an edge transition happened while
the interrupt was being handled and the handler is called again.

Signed-off-by: Abhijeet Dharmapurikar <adharmap@codeaurora.org>
Signed-off-by: Kiran Gunda <kgunda@codeaurora.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agospmi: pmic-arb: clear the latched status of the interrupt
Abhijeet Dharmapurikar [Wed, 10 May 2017 14:25:37 +0000 (19:55 +0530)]
spmi: pmic-arb: clear the latched status of the interrupt

PMIC interrupts each have an internal latched status bit which is
not visible from any register.  This status bit is set as soon as
the conditions specified in the interrupt type and polarity
registers are met even if the interrupt is not enabled.  When it
is set, nothing else changes within the PMIC and no interrupt
notification packets are sent.  If the internal latched status
bit is set when an interrupt is enabled, then the value is
immediately propagated into the interrupt latched status register
and an interrupt notification packet is sent out from the PMIC
over SPMI.

This PMIC hardware behavior can lead to a situation where the
handler for a level triggered interrupt is called immediately
after enable_irq() is called even though the interrupt physically
triggered while it was disabled within the genirq framework.
This situation takes place if the the interrupt fires twice after
calling disable_irq().  The first time it fires, the level flow
handler will mask and disregard it.  Unfortunately, the second
time it fires, the internal latched status bit is set within the
PMIC and no further notification is received.  When enable_irq()
is called later, the interrupt is unmasked (enabled in the PMIC)
which results in the PMIC immediately sending an interrupt
notification packet out over SPMI.  This breaks the semantics
of level triggered interrupts within the genirq framework since
they should be completely ignored while disabled.

The PMIC internal latched status behavior also affects how
interrupts are treated during suspend.  While entering suspend,
all interrupts not specified as wakeup mode are masked.  Upon
resume, these interrupts are unmasked.  Thus if any of the
non-wakeup PMIC interrupts fired while the system was suspended,
then the PMIC will send interrupt notification packets out via
SPMI as soon as they are unmasked during resume.  This behavior
violates genirq semantics as well since non-wakeup interrupts
should be completely ignored during suspend.

Modify the qpnpint_irq_unmask() function so that the interrupt
latched status clear register is written immediately before the
interrupt enable register.  This clears the internal latched
status bit of the interrupt so that it cannot trigger spuriously
immediately upon being enabled.

Also, while resuming an irq, an unmask could be called even if it
was not previously masked.  So, before writing these registers,
check if the interrupt is already enabled within the PMIC. If it
is, then no further register writes are required.  This
condition check ensures that a valid latched status register bit
is not cleared until it is properly handled.

Signed-off-by: Abhijeet Dharmapurikar <adharmap@codeaurora.org>
Signed-off-by: Kiran Gunda <kgunda@codeaurora.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agospmi: pmic-arb: fix missing interrupts
Abhijeet Dharmapurikar [Wed, 10 May 2017 14:25:36 +0000 (19:55 +0530)]
spmi: pmic-arb: fix missing interrupts

irq_enable is called when the device resumes. Note that the
irq_enable is called regardless of whether the interrupt was
marked enabled/disabled in the descriptor or whether it was
masked/unmasked at the controller while resuming.

The current driver unconditionally clears the interrupt in its
irq_enable callback. This is dangerous as any interrupts that
happen right before the resume could be missed.
Remove the irq_enable callback and use mask/unmask instead.

Also remove struct pmic_arb_irq_spec as it serves no real purpose.
It is used only in the translate function and the code is much
cleaner without it.

Signed-off-by: Abhijeet Dharmapurikar <adharmap@codeaurora.org>
Signed-off-by: Kiran Gunda <kgunda@codeaurora.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agospmi: pmic-arb: cleanup unrequested irqs
Abhijeet Dharmapurikar [Wed, 10 May 2017 14:25:35 +0000 (19:55 +0530)]
spmi: pmic-arb: cleanup unrequested irqs

We see a unmapped irqs trigger right around bootup. This could
likely be because the bootloader exited leaving the interrupts
in an unknown or unhandled state.  Ack and mask the interrupt
if one is found. A request_irq later will unmask it and also
setup proper mapping structures.

Also the current driver ensures that no read/write transaction
is in progress while it makes changes to the interrupt regions.
This is not necessary because read/writes over spmi and arbiter
interrupt control are independent operations. Hence, remove the
synchronized accesses to interrupt region.

Signed-off-by: Abhijeet Dharmapurikar <adharmap@codeaurora.org>
Signed-off-by: Kiran Gunda <kgunda@codeaurora.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agospmi: pmic-arb: optimize table lookups
Abhijeet Dharmapurikar [Wed, 10 May 2017 14:25:34 +0000 (19:55 +0530)]
spmi: pmic-arb: optimize table lookups

The current driver uses a mix of radix tree and a fwd lookup
table to translate between apid and ppid. It is buggy and confusing.

Instead simply use a radix tree for v1 hardware and use the
forward lookup table for v2.

Signed-off-by: Abhijeet Dharmapurikar <adharmap@codeaurora.org>
Signed-off-by: Kiran Gunda <kgunda@codeaurora.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agospmi: pmic-arb: fix inconsistent use of apid and chan
Abhijeet Dharmapurikar [Wed, 10 May 2017 14:25:33 +0000 (19:55 +0530)]
spmi: pmic-arb: fix inconsistent use of apid and chan

The driver currently uses "apid" and "chan" to mean apid. Remove
the use of chan and use only apid.

On a SPMI bus there is allocation to manage up to 4K peripherals.
However, in practice only few peripherals are instantiated
and only few among the instantiated ones actually interrupt.

APID is CPU's way of keeping track of peripherals that could interrupt.
There is a table that maps the 256 interrupting peripherals to
a number between 0 and 255. This number is called APID. Information about
that interrupting peripheral is stored in registers offset by its
corresponding apid.

Signed-off-by: Abhijeet Dharmapurikar <adharmap@codeaurora.org>
Signed-off-by: Kiran Gunda <kgunda@codeaurora.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agospmi: pmic-arb: rename spmi_pmic_arb_dev to spmi_pmic_arb
Abhijeet Dharmapurikar [Wed, 10 May 2017 14:25:32 +0000 (19:55 +0530)]
spmi: pmic-arb: rename spmi_pmic_arb_dev to spmi_pmic_arb

Usually *_dev best used for structures that embed a struct device in
them. spmi_pmic_arb_dev doesn't embed one. It is simply a driver data
structure. Use an appropriate name for it.

Also there are many places in the driver that left shift the bit to
generate a bit mask. Replace it with the BIT() macro.

Signed-off-by: Abhijeet Dharmapurikar <adharmap@codeaurora.org>
Signed-off-by: Kiran Gunda <kgunda@codeaurora.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agospmi: pmic_arb: block access of invalid read and writes
Abhijeet Dharmapurikar [Wed, 10 May 2017 14:25:31 +0000 (19:55 +0530)]
spmi: pmic_arb: block access of invalid read and writes

The system crashes due to bad access when reading from an non configured
peripheral and when writing to peripheral which is not owned by current
ee. This patch verifies ownership to avoid crashing on
write.
For reads, since the forward mapping table, data_channel->ppid, is
towards the end of the block, we use the core size to figure the
max number of ppids supported. The table starts at an offset of 0x800
within the block, so size - 0x800 will give us the area used by the
table. Since each table is 4 bytes long (core_size - 0x800) / 4 will
gives us the number of data_channel supported.
This new protection is functional on hw v2.

Signed-off-by: Abhijeet Dharmapurikar <adharmap@codeaurora.org>
Signed-off-by: Kiran Gunda <kgunda@codeaurora.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoRevert "firmware: vpd: remove platform driver"
Dmitry Torokhov [Fri, 26 May 2017 20:57:49 +0000 (13:57 -0700)]
Revert "firmware: vpd: remove platform driver"

This reverts commit 7975bd4cca05a99aa14964cfa22366ee64da50ad, because
VPD relies on driver core to handle deferrals returned by
coreboot_table_find().

Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
Reviewed-by: Guenter Roeck <groeck@chromium.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agomisc: bh1770glc: move header file out of I2C realm
Wolfram Sang [Sun, 21 May 2017 20:42:33 +0000 (22:42 +0200)]
misc: bh1770glc: move header file out of I2C realm

include/linux/i2c is not for client devices. Move the header file to a
more appropriate location.

Signed-off-by: Wolfram Sang <wsa@the-dreams.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agomisc: apds990x: move header file out of I2C realm
Wolfram Sang [Sun, 21 May 2017 20:42:32 +0000 (22:42 +0200)]
misc: apds990x: move header file out of I2C realm

include/linux/i2c is not for client devices. Move the header file to a
more appropriate location.

Signed-off-by: Wolfram Sang <wsa@the-dreams.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoauxdisplay: Convert list_for_each to entry variant
Wei Yongjun [Tue, 25 Apr 2017 16:13:34 +0000 (16:13 +0000)]
auxdisplay: Convert list_for_each to entry variant

convert list_for_each() to list_for_each_entry() where
applicable.

Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agogoldfish_pipe: make pipe_dev static
Colin Ian King [Thu, 27 Apr 2017 17:41:34 +0000 (18:41 +0100)]
goldfish_pipe: make pipe_dev static

Make this static as it's only referenced in this source and
it does not need global scope.

Cleans up a sparse warning:

drivers/platform/goldfish/goldfish_pipe.c: warning: symbol
  'pipe_dev' was not declared. Should it be static?

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agomei: hw: fix a spelling mistake
Tomas Winkler [Sun, 30 Apr 2017 10:12:11 +0000 (13:12 +0300)]
mei: hw: fix a spelling mistake

notifcation -> notification

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agomei: make mei_cl_bus_rescan static
Alexander Usyskin [Sun, 30 Apr 2017 10:12:10 +0000 (13:12 +0300)]
mei: make mei_cl_bus_rescan static

mei_cl_bus_rescan is used only in bus.c,
so make it local to the file and mark static.

Signed-off-by: Alexander Usyskin <alexander.usyskin@intel.com>
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agow1: Organize driver source to natural/common order
Andrew F. Davis [Tue, 16 May 2017 20:02:12 +0000 (15:02 -0500)]
w1: Organize driver source to natural/common order

Structures and functions should be ordered such that forward declaration
use is minimized.

MODULE_* macros should immediately follow the structures and functions
upon which they act.

Remaining MODULE_* macros should be at the end of the file in
alphabetical order.

Signed-off-by: Andrew F. Davis <afd@ti.com>
Acked-by: Evgeniy Polyakov <zbr@ioremap.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoDrivers: hv: vmbus: Close timing hole that can corrupt per-cpu page
Michael Kelley [Thu, 18 May 2017 17:46:07 +0000 (10:46 -0700)]
Drivers: hv: vmbus: Close timing hole that can corrupt per-cpu page

Extend the disabling of preemption to include the hypercall so that
another thread can't get the CPU and corrupt the per-cpu page used
for hypercall arguments.

Cc: <Stable@vger.kernel.org> #4.11
Signed-off-by: Michael Kelley <mikelley@microsoft.com>
Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agovmbus: Reuse uuid_le_to_bin() helper
Andy Shevchenko [Thu, 18 May 2017 17:46:06 +0000 (10:46 -0700)]
vmbus: Reuse uuid_le_to_bin() helper

Instead of open coded variant use generic helper to convert UUID strings
to binary format.

Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agodrivers: hv: vmbus: Increase the time between retries in vmbus_post_msg()
K. Y. Srinivasan [Thu, 18 May 2017 17:46:05 +0000 (10:46 -0700)]
drivers: hv: vmbus: Increase the time between retries in vmbus_post_msg()

Commit c0bb03924f1a ("Drivers: hv: vmbus: Raise retry/wait limits in
vmbus_post_msg()") increased the retry/wait limits of vmbus_post_msg()
to address the new DoS protection policies in WS2016.
Increase the time between retries to make the
code more robust.

Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agohv_utils: fix TimeSync work on pre-TimeSync-v4 hosts
Vitaly Kuznetsov [Thu, 18 May 2017 17:46:04 +0000 (10:46 -0700)]
hv_utils: fix TimeSync work on pre-TimeSync-v4 hosts

It was found that ICTIMESYNCFLAG_SYNC packets are handled incorrectly
on WS2012R2, e.g. after the guest is paused and resumed its time is set
to something different from host's time. The problem is that we call
adj_guesttime() with reftime=0 for these old hosts and we don't account
for that in 'if (adj_flags & ICTIMESYNCFLAG_SYNC)' branch and
hv_set_host_time().

While we could've solved this by adding a check like
'if (ts_srv_version > TS_VERSION_3)' to hv_set_host_time() I prefer
to do some refactoring. We don't need to have two separate containers
for host samples, struct host_ts which we use for PTP is enough.

Throw away 'struct adj_time_work' and create hv_get_adj_host_time()
accessor to host_ts to avoid code duplication.

Fixes: 3716a49a81ba ("hv_utils: implement Hyper-V PTP source")
Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com>
Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agohv_utils: drop .getcrosststamp() support from PTP driver
Vitaly Kuznetsov [Thu, 18 May 2017 17:46:03 +0000 (10:46 -0700)]
hv_utils: drop .getcrosststamp() support from PTP driver

Turns out that our implementation of .getcrosststamp() never actually
worked. Hyper-V is sending time samples every 5 seconds and this is
too much for get_device_system_crosststamp() as it's interpolation
algorithm (which nobody is currently using in kernel, btw) accounts
for a 'slow' device but we're not slow in Hyper-V, our time reference
is too far away.

.getcrosststamp() is not currently used, get_device_system_crosststamp()
almost always returns -EINVAL and client falls back to using PTP_SYS_OFFSET
so this patch doesn't change much. I also tried doing interpolation
manually (e.g. the same way hv_ptp_gettime() works and it turns out that
we're getting even lower quality:

PTP_SYS_OFFSET_PRECISE with manual interpolation:
* PHC0                     0   3    37     4  -3974ns[-5338ns] +/-  977ns
* PHC0                     0   3    77     7  +2227ns[+3184ns] +/-  576ns
* PHC0                     0   3   177    10  +3060ns[+5220ns] +/-  548ns
* PHC0                     0   3   377    12  +3937ns[+4371ns] +/- 1414ns
* PHC0                     0   3   377     6   +764ns[+1240ns] +/- 1047ns
* PHC0                     0   3   377     7  -1210ns[-3731ns] +/-  479ns
* PHC0                     0   3   377     9   +153ns[-1019ns] +/-  406ns
* PHC0                     0   3   377    12   -872ns[-1793ns] +/-  443ns
* PHC0                     0   3   377     5   +701ns[+3599ns] +/-  426ns
* PHC0                     0   3   377     5   -923ns[ -375ns] +/- 1062ns

PTP_SYS_OFFSET:
* PHC0                     0   3     7     5    +72ns[+8020ns] +/-  251ns
* PHC0                     0   3    17     5   -885ns[-3661ns] +/-  254ns
* PHC0                     0   3    37     6   -454ns[-5732ns] +/-  258ns
* PHC0                     0   3    77    10  +1183ns[+3754ns] +/-  164ns
* PHC0                     0   3   377     5   +579ns[+1137ns] +/-  110ns
* PHC0                     0   3   377     7   +501ns[+1064ns] +/-   96ns
* PHC0                     0   3   377     9  +1641ns[+3342ns] +/-  106ns
* PHC0                     0   3   377     8    -47ns[  +77ns] +/-  160ns
* PHC0                     0   3   377     5    +54ns[ +107ns] +/-  102ns
* PHC0                     0   3   377     8   -354ns[ -617ns] +/-   89ns

This fact wasn't noticed during the initial testing of the PTP device
somehow but got revealed now. Let's just drop .getcrosststamp()
implementation for now as it doesn't seem to be suitable for us.

Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com>
Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoDrivers: hv: vmbus: Get the current time from the current clocksource
K. Y. Srinivasan [Thu, 18 May 2017 17:46:02 +0000 (10:46 -0700)]
Drivers: hv: vmbus: Get the current time from the current clocksource

The current code uses the MSR based mechanism to get the current tick.
Use the current clock source as that might be more optimal.

Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agofirmware: vpd: remove platform driver
Dmitry Torokhov [Wed, 24 May 2017 00:07:47 +0000 (17:07 -0700)]
firmware: vpd: remove platform driver

There is no reason why VPD should register platform device and driver,
given that we do not use their respective kobjects to attach attributes,
nor do we need suspend/resume hooks, or any other features of device
core.

Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
Reviewed-by: Guenter Roeck <groeck@chromium.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agofirmware: vpd: do not clear statically allocated data
Dmitry Torokhov [Wed, 24 May 2017 00:07:46 +0000 (17:07 -0700)]
firmware: vpd: do not clear statically allocated data

ro_vpd and rw_vpd are static module-scope variables that are guaranteed
to be initialized with zeroes, there is no need for explicit memset().

Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
Reviewed-by: Guenter Roeck <groeck@chromium.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agofirmware: vpd: use kasprintf() when forming name of 'raw' attribute
Dmitry Torokhov [Wed, 24 May 2017 00:07:45 +0000 (17:07 -0700)]
firmware: vpd: use kasprintf() when forming name of 'raw' attribute

When creating name for the "raw" attribute, let's switch to using
kaspeintf() instead of doing it by hand. Also make sure we handle
errors.

Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
Reviewed-by: Guenter Roeck <groeck@chromium.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agofirmware: vpd: use kdtrndup when copying section key
Dmitry Torokhov [Wed, 24 May 2017 00:07:42 +0000 (17:07 -0700)]
firmware: vpd: use kdtrndup when copying section key

Instead of open-coding kstrndup with kzalloc + memcpy, let's use
the helper.

Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
Reviewed-by: Guenter Roeck <groeck@chromium.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agofirmware: google: memconsole: Prevent overrun attack on coreboot console
Julius Werner [Tue, 23 May 2017 23:48:17 +0000 (16:48 -0700)]
firmware: google: memconsole: Prevent overrun attack on coreboot console

The recent coreboot memory console update (firmware: google: memconsole:
Adapt to new coreboot ring buffer format) introduced a small security
issue in the driver: The new driver implementation parses the memory
console structure again on every access. This is intentional so that
additional lines added concurrently by runtime firmware can be read out.

However, if an attacker can write to the structure, they could increase
the size value to a point where the driver would read potentially
sensitive memory areas from outside the original console buffer during
the next access. This can be done through /dev/mem, since the console
buffer usually resides in firmware-reserved memory that is not covered
by STRICT_DEVMEM.

This patch resolves that problem by reading the buffer's size value only
once during boot (where we can still trust the structure). Other parts
of the structure can still be modified at runtime, but the driver's
bounds checks make sure that it will never read outside the buffer.

Fixes: a5061d028 ("firmware: google: memconsole: Adapt to new coreboot ring buffer format")
Signed-off-by: Julius Werner <jwerner@chromium.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoMerge 4.12-rc2 into char-misc-next
Greg Kroah-Hartman [Mon, 22 May 2017 06:56:55 +0000 (08:56 +0200)]
Merge 4.12-rc2 into char-misc-next

We want the fixes in here as well to handle merge issues.

Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoLinux 4.12-rc2
Linus Torvalds [Mon, 22 May 2017 02:30:23 +0000 (19:30 -0700)]
Linux 4.12-rc2

7 years agox86: fix 32-bit case of __get_user_asm_u64()
Linus Torvalds [Mon, 22 May 2017 01:26:54 +0000 (18:26 -0700)]
x86: fix 32-bit case of __get_user_asm_u64()

The code to fetch a 64-bit value from user space was entirely buggered,
and has been since the code was merged in early 2016 in commit
b2f680380ddf ("x86/mm/32: Add support for 64-bit __get_user() on 32-bit
kernels").

Happily the buggered routine is almost certainly entirely unused, since
the normal way to access user space memory is just with the non-inlined
"get_user()", and the inlined version didn't even historically exist.

The normal "get_user()" case is handled by external hand-written asm in
arch/x86/lib/getuser.S that doesn't have either of these issues.

There were two independent bugs in __get_user_asm_u64():

 - it still did the STAC/CLAC user space access marking, even though
   that is now done by the wrapper macros, see commit 11f1a4b9755f
   ("x86: reorganize SMAP handling in user space accesses").

   This didn't result in a semantic error, it just means that the
   inlined optimized version was hugely less efficient than the
   allegedly slower standard version, since the CLAC/STAC overhead is
   quite high on modern Intel CPU's.

 - the double register %eax/%edx was marked as an output, but the %eax
   part of it was touched early in the asm, and could thus clobber other
   inputs to the asm that gcc didn't expect it to touch.

   In particular, that meant that the generated code could look like
   this:

        mov    (%eax),%eax
        mov    0x4(%eax),%edx

   where the load of %edx obviously was _supposed_ to be from the 32-bit
   word that followed the source of %eax, but because %eax was
   overwritten by the first instruction, the source of %edx was
   basically random garbage.

The fixes are trivial: remove the extraneous STAC/CLAC entries, and mark
the 64-bit output as early-clobber to let gcc know that no inputs should
alias with the output register.

Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Benjamin LaHaise <bcrl@kvack.org>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: stable@kernel.org # v4.8+
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agoClean up x86 unsafe_get/put_user() type handling
Linus Torvalds [Sun, 21 May 2017 22:25:46 +0000 (15:25 -0700)]
Clean up x86 unsafe_get/put_user() type handling

Al noticed that unsafe_put_user() had type problems, and fixed them in
commit a7cc722fff0b ("fix unsafe_put_user()"), which made me look more
at those functions.

It turns out that unsafe_get_user() had a type issue too: it limited the
largest size of the type it could handle to "unsigned long".  Which is
fine with the current users, but doesn't match our existing normal
get_user() semantics, which can also handle "u64" even when that does
not fit in a long.

While at it, also clean up the type cast in unsafe_put_user().  We
actually want to just make it an assignment to the expected type of the
pointer, because we actually do want warnings from types that don't
convert silently.  And it makes the code more readable by not having
that one very long and complex line.

[ This patch might become stable material if we ever end up back-porting
  any new users of the unsafe uaccess code, but as things stand now this
  doesn't matter for any current existing uses. ]

Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs
Linus Torvalds [Sun, 21 May 2017 19:06:44 +0000 (12:06 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/viro/vfs

Pull misc uaccess fixes from Al Viro:
 "Fix for unsafe_put_user() (no callers currently in mainline, but
  anyone starting to use it will step into that) + alpha osf_wait4()
  infoleak fix"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs:
  osf_wait4(): fix infoleak
  fix unsafe_put_user()

7 years agoMerge branch 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 21 May 2017 18:52:00 +0000 (11:52 -0700)]
Merge branch 'sched-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull scheduler fix from Thomas Gleixner:
 "A single scheduler fix:

  Prevent idle task from ever being preempted. That makes sure that
  synchronize_rcu_tasks() which is ignoring idle task does not pretend
  that no task is stuck in preempted state. If that happens and idle was
  preempted on a ftrace trampoline the machine crashes due to
  inconsistent state"

* 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  sched/core: Call __schedule() from do_idle() without enabling preemption

7 years agoMerge branch 'irq-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 21 May 2017 18:45:26 +0000 (11:45 -0700)]
Merge branch 'irq-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull irq fixes from Thomas Gleixner:
 "A set of small fixes for the irq subsystem:

   - Cure a data ordering problem with chained interrupts

   - Three small fixlets for the mbigen irq chip"

* 'irq-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  genirq: Fix chained interrupt data ordering
  irqchip/mbigen: Fix the clear register offset calculation
  irqchip/mbigen: Fix potential NULL dereferencing
  irqchip/mbigen: Fix memory mapping code

7 years agoosf_wait4(): fix infoleak
Al Viro [Mon, 15 May 2017 01:47:25 +0000 (21:47 -0400)]
osf_wait4(): fix infoleak

failing sys_wait4() won't fill struct rusage...

Cc: stable@vger.kernel.org
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
7 years agofix unsafe_put_user()
Al Viro [Sun, 21 May 2017 17:08:42 +0000 (13:08 -0400)]
fix unsafe_put_user()

__put_user_size() relies upon its first argument having the same type as what
the second one points to; the only other user makes sure of that and
unsafe_put_user() should do the same.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
7 years agoMerge tag 'trace-v4.12-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt...
Linus Torvalds [Sun, 21 May 2017 06:39:03 +0000 (23:39 -0700)]
Merge tag 'trace-v4.12-rc1' of git://git./linux/kernel/git/rostedt/linux-trace

Pull tracing fixes from Steven Rostedt:

 - Fix a bug caused by not cleaning up the new instance unique triggers
   when deleting an instance. It also creates a selftest that triggers
   that bug.

 - Fix the delayed optimization happening after kprobes boot up self
   tests being removed by freeing of init memory.

 - Comment kprobes on why the delay optimization is not a problem for
   removal of modules, to keep other developers from searching that
   riddle.

 - Fix another case of rcu not watching in stack trace tracing.

* tag 'trace-v4.12-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace:
  tracing: Make sure RCU is watching before calling a stack trace
  kprobes: Document how optimized kprobes are removed from module unload
  selftests/ftrace: Add test to remove instance with active event triggers
  selftests/ftrace: Fix bashisms
  ftrace: Remove #ifdef from code and add clear_ftrace_function_probes() stub
  ftrace/instances: Clear function triggers when removing instances
  ftrace: Simplify glob handling in unregister_ftrace_function_probe_func()
  tracing/kprobes: Enforce kprobes teardown after testing
  tracing: Move postpone selftests to core from early_initcall

7 years agoMerge branch 'for-linus' of git://git.kernel.dk/linux-block
Linus Torvalds [Sat, 20 May 2017 23:12:30 +0000 (16:12 -0700)]
Merge branch 'for-linus' of git://git.kernel.dk/linux-block

Pull block fixes from Jens Axboe:
 "A small collection of fixes that should go into this cycle.

   - a pull request from Christoph for NVMe, which ended up being
     manually applied to avoid pulling in newer bits in master. Mostly
     fibre channel fixes from James, but also a few fixes from Jon and
     Vijay

   - a pull request from Konrad, with just a single fix for xen-blkback
     from Gustavo.

   - a fuseblk bdi fix from Jan, fixing a regression in this series with
     the dynamic backing devices.

   - a blktrace fix from Shaohua, replacing sscanf() with kstrtoull().

   - a request leak fix for drbd from Lars, fixing a regression in the
     last series with the kref changes. This will go to stable as well"

* 'for-linus' of git://git.kernel.dk/linux-block:
  nvmet: release the sq ref on rdma read errors
  nvmet-fc: remove target cpu scheduling flag
  nvme-fc: stop queues on error detection
  nvme-fc: require target or discovery role for fc-nvme targets
  nvme-fc: correct port role bits
  nvme: unmap CMB and remove sysfs file in reset path
  blktrace: fix integer parse
  fuseblk: Fix warning in super_setup_bdi_name()
  block: xen-blkback: add null check to avoid null pointer dereference
  drbd: fix request leak introduced by locking/atomic, kref: Kill kref_sub()

7 years agonvmet: release the sq ref on rdma read errors
Vijay Immanuel [Mon, 8 May 2017 23:38:35 +0000 (16:38 -0700)]
nvmet: release the sq ref on rdma read errors

On rdma read errors, release the sq ref that was taken
when the req was initialized. This avoids a hang in
nvmet_sq_destroy() when the queue is being freed.

Signed-off-by: Vijay Immanuel <vijayi@attalasystems.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@fb.com>
7 years agonvmet-fc: remove target cpu scheduling flag
James Smart [Tue, 25 Apr 2017 23:23:09 +0000 (16:23 -0700)]
nvmet-fc: remove target cpu scheduling flag

Remove NVMET_FCTGTFEAT_NEEDS_CMD_CPUSCHED. It's unnecessary.

Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@fb.com>
7 years agonvme-fc: stop queues on error detection
James Smart [Tue, 25 Apr 2017 22:32:01 +0000 (15:32 -0700)]
nvme-fc: stop queues on error detection

Per the recommendation by Sagi on:
http://lists.infradead.org/pipermail/linux-nvme/2017-April/009261.html

Rather than waiting for reset work thread to stop queues and abort the ios,
immediately stop the queues on error detection. Reset thread will restop
the queues (as it's called on other paths), but it does not appear to have
a side effect.

Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@fb.com>
7 years agonvme-fc: require target or discovery role for fc-nvme targets
James Smart [Fri, 5 May 2017 23:13:15 +0000 (16:13 -0700)]
nvme-fc: require target or discovery role for fc-nvme targets

In order to create an association, the remoteport must be
serving either a target role or a discovery role.

Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@fb.com>