]> git.karo-electronics.de Git - linux-beck.git/log
linux-beck.git
9 years agonvmem: sunxi: Move the SID driver to the nvmem framework
Maxime Ripard [Mon, 27 Jul 2015 11:17:09 +0000 (12:17 +0100)]
nvmem: sunxi: Move the SID driver to the nvmem framework

Now that we have the nvmem framework, we can consolidate the common
driver code. Move the driver to the framework, and hopefully, it will
fix the sysfs file creation race.

Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
[srinivas.kandagatla: Moved to regmap based EEPROM framework]
Signed-off-by: Srinivas Kandagatla <srinivas.kandagatla@linaro.org>
Tested-by: Philipp Zabel <p.zabel@pengutronix.de>
Tested-by: Rajendra Nayak <rnayak@codeaurora.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agonvmem: qfprom: Add bindings for qfprom
Srinivas Kandagatla [Mon, 27 Jul 2015 11:16:59 +0000 (12:16 +0100)]
nvmem: qfprom: Add bindings for qfprom

This patch adds bindings for qfprom found in QCOM SOCs. QFPROM driver
is based on simple nvmem framework.

Signed-off-by: Srinivas Kandagatla <srinivas.kandagatla@linaro.org>
Reviewed-by: Stephen Boyd <sboyd@codeaurora.org>
Tested-by: Philipp Zabel <p.zabel@pengutronix.de>
Tested-by: Rajendra Nayak <rnayak@codeaurora.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agonvmem: qfprom: Add Qualcomm QFPROM support.
Srinivas Kandagatla [Mon, 27 Jul 2015 11:15:00 +0000 (12:15 +0100)]
nvmem: qfprom: Add Qualcomm QFPROM support.

This patch adds QFPROM support driver which is used by other drivers
like thermal sensor and cpufreq.

On MSM parts there are some efuses (called qfprom) these fuses store
things like calibration data, speed bins.. etc. Drivers like cpufreq,
thermal sensors would read out this data for configuring the driver.

Signed-off-by: Srinivas Kandagatla <srinivas.kandagatla@linaro.org>
Reviewed-by: Stephen Boyd <sboyd@codeaurora.org>
Tested-by: Philipp Zabel <p.zabel@pengutronix.de>
Tested-by: Rajendra Nayak <rnayak@codeaurora.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoDocumentation: nvmem: add nvmem api level and how-to doc
Srinivas Kandagatla [Mon, 27 Jul 2015 11:14:14 +0000 (12:14 +0100)]
Documentation: nvmem: add nvmem api level and how-to doc

This patch add basic how-to and api summary documentation for simple
NVMEM framework.

Signed-off-by: Srinivas Kandagatla <srinivas.kandagatla@linaro.org>
Tested-by: Philipp Zabel <p.zabel@pengutronix.de>
Tested-by: Rajendra Nayak <rnayak@codeaurora.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agonvmem: Add bindings for simple nvmem framework
Srinivas Kandagatla [Mon, 27 Jul 2015 11:13:58 +0000 (12:13 +0100)]
nvmem: Add bindings for simple nvmem framework

This patch adds bindings for simple nvmem framework which allows nvmem
consumers to talk to nvmem providers to get access to nvmem cell data.

Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
[Maxime Ripard: intial version of eeprom framework]
Signed-off-by: Srinivas Kandagatla <srinivas.kandagatla@linaro.org>
Tested-by: Philipp Zabel <p.zabel@pengutronix.de>
Tested-by: Rajendra Nayak <rnayak@codeaurora.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agonvmem: Add nvmem_device based consumer apis.
Srinivas Kandagatla [Mon, 27 Jul 2015 11:13:45 +0000 (12:13 +0100)]
nvmem: Add nvmem_device based consumer apis.

This patch adds read/write apis which are based on nvmem_device. It is
common that the drivers like omap cape manager or qcom cpr driver to
access bytes directly at particular offset in the eeprom and not from
nvmem cell info in DT. These driver would need to get access to the nvmem
directly, which is what these new APIS provide.

These wrapper apis would help such users to avoid code duplication in
there drivers and also avoid them reading a big eeprom blob and parsing
it internally in there driver.

Signed-off-by: Srinivas Kandagatla <srinivas.kandagatla@linaro.org>
Tested-by: Stefan Wahren <stefan.wahren@i2se.com>
Tested-by: Philipp Zabel <p.zabel@pengutronix.de>
Tested-by: Rajendra Nayak <rnayak@codeaurora.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agonvmem: Add a simple NVMEM framework for consumers
Srinivas Kandagatla [Mon, 27 Jul 2015 11:13:34 +0000 (12:13 +0100)]
nvmem: Add a simple NVMEM framework for consumers

This patch adds just consumers part of the framework just to enable easy
review.

Up until now, nvmem drivers were stored in drivers/misc, where they all
had to duplicate pretty much the same code to register a sysfs file,
allow in-kernel users to access the content of the devices they were
driving, etc.

This was also a problem as far as other in-kernel users were involved,
since the solutions used were pretty much different from on driver to
another, there was a rather big abstraction leak.

This introduction of this framework aims at solving this. It also
introduces DT representation for consumer devices to go get the data they
require (MAC Addresses, SoC/Revision ID, part numbers, and so on) from
the nvmems.

Having regmap interface to this framework would give much better
abstraction for nvmems on different buses.

Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
[Maxime Ripard: intial version of the framework]
Signed-off-by: Srinivas Kandagatla <srinivas.kandagatla@linaro.org>
Tested-by: Stefan Wahren <stefan.wahren@i2se.com>
Tested-by: Philipp Zabel <p.zabel@pengutronix.de>
Tested-by: Rajendra Nayak <rnayak@codeaurora.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agonvmem: Add a simple NVMEM framework for nvmem providers
Srinivas Kandagatla [Mon, 27 Jul 2015 11:13:19 +0000 (12:13 +0100)]
nvmem: Add a simple NVMEM framework for nvmem providers

This patch adds just providers part of the framework just to enable easy
review.

Up until now, NVMEM drivers like eeprom were stored in drivers/misc,
where they all had to duplicate pretty much the same code to register
a sysfs file, allow in-kernel users to access the content of the devices
they were driving, etc.

This was also a problem as far as other in-kernel users were involved,
since the solutions used were pretty much different from on driver to
another, there was a rather big abstraction leak.

This introduction of this framework aims at solving this. It also
introduces DT representation for consumer devices to go get the data
they require (MAC Addresses, SoC/Revision ID, part numbers, and so on)
from the nvmems.

Having regmap interface to this framework would give much better
abstraction for nvmems on different buses.

Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
[Maxime Ripard: intial version of eeprom framework]
Signed-off-by: Srinivas Kandagatla <srinivas.kandagatla@linaro.org>
Tested-by: Stefan Wahren <stefan.wahren@i2se.com>
Tested-by: Philipp Zabel <p.zabel@pengutronix.de>
Tested-by: Rajendra Nayak <rnayak@codeaurora.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoauxdisplay: ks0108: use new parport device model
Sudip Mukherjee [Mon, 20 Jul 2015 11:57:24 +0000 (17:27 +0530)]
auxdisplay: ks0108: use new parport device model

Modify auxdisplay driver to use the new parallel port device model.

Signed-off-by: Sudip Mukherjee <sudip@vectorindia.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoauxdisplay: ks0108: use min_t
Sudip Mukherjee [Mon, 20 Jul 2015 11:57:23 +0000 (17:27 +0530)]
auxdisplay: ks0108: use min_t

Using min_t() is preffered than using min().

Signed-off-by: Sudip Mukherjee <sudip@vectorindia.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoauxdisplay: ks0108: start using pr_*
Sudip Mukherjee [Mon, 20 Jul 2015 11:57:22 +0000 (17:27 +0530)]
auxdisplay: ks0108: start using pr_*

Start using pr_* macros instead of using printk and in the process
define pr_fmt.

Signed-off-by: Sudip Mukherjee <sudip@vectorindia.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoauxdisplay: ks0108: fix refcount
Sudip Mukherjee [Mon, 20 Jul 2015 11:57:21 +0000 (17:27 +0530)]
auxdisplay: ks0108: fix refcount

parport_find_base() will implicitly do parport_get_port() which
increases the refcount. Then parport_register_device() will again
increment the refcount. But while unloading the module we are only
doing parport_unregister_device() decrementing the refcount only once.
We add an parport_put_port() to neutralize the effect of
parport_get_port().

Cc: <stable@vger.kernel.org> # 2.6.32+
Signed-off-by: Sudip Mukherjee <sudip@vectorindia.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agochar/nvram: Use bitwise OR to obtain Atari video mode data
Finn Thain [Fri, 3 Jul 2015 03:35:36 +0000 (13:35 +1000)]
char/nvram: Use bitwise OR to obtain Atari video mode data

Signed-off-by: Finn Thain <fthain@telegraphics.com.au>
Acked-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agocoresight: Fix implicit inclusion of linux/sched.h
Mark Brown [Fri, 31 Jul 2015 15:37:30 +0000 (09:37 -0600)]
coresight: Fix implicit inclusion of linux/sched.h

The patch "Coresight: Add an interface for supporting ETM3/4 Context ID
tracing" adds uses of find_task_by_vpid() and task_pid_nr() from
linux/sched.h but does not include that header causing build errors in
at least an ARM allmodconfig where it is not implicitly included. Add an
explicit include to fix that.

Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoDoc: trace: Fix typo in coresight.txt
Masanari Iida [Fri, 31 Jul 2015 15:37:29 +0000 (09:37 -0600)]
Doc: trace: Fix typo in coresight.txt

This patch fix spelling typos found in coresight.txt

Signed-off-by: Masanari Iida <standby24x7@gmail.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agocoresight-etm4x: Support context-ID tracing when PID namespace is enabled
Chunyan Zhang [Fri, 31 Jul 2015 15:37:28 +0000 (09:37 -0600)]
coresight-etm4x: Support context-ID tracing when PID namespace is enabled

Like ETTv3, ETMv4 also needs the similar modifications to support Context
ID tracing when PID namespace is enabled.

Signed-off-by: Chunyan Zhang <zhang.chunyan@linaro.org>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agocoresight-etm3x: Support context-ID tracing when PID namespace is enabled
Chunyan Zhang [Fri, 31 Jul 2015 15:37:27 +0000 (09:37 -0600)]
coresight-etm3x: Support context-ID tracing when PID namespace is enabled

The Coresight ETM drivers already support context-ID tracing, but it won't
work when PID namespace is enabled. This is because when using PID
namespace a process id (ie. VPID) seen from the current namespace differs
from the id (ie. PID) seen by kernel.

So when users write the process id seen by themselves to ETM, there needs
to be a translation from VPID to PID, as such ETM drivers will write the
PID into the Context ID register correctly.

Signed-off-by: Chunyan Zhang <zhang.chunyan@linaro.org>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoCoresight: Add an interface for supporting ETM3/4 Context ID tracing
Chunyan Zhang [Fri, 31 Jul 2015 15:37:26 +0000 (09:37 -0600)]
Coresight: Add an interface for supporting ETM3/4 Context ID tracing

If PID namespace is enabled, everytime users configure the Context ID
register to trace the specific process, there needs to be a translation
between the real PID seen from the kernel and VPID seen from the
namespace in which the user's process resides .

This patch just adds the translation interface for ETMs.

Signed-off-by: Chunyan Zhang <zhang.chunyan@linaro.org>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agocoresight-etm4x: Change the name of the ctxid_val to ctxid_pid
Chunyan Zhang [Fri, 31 Jul 2015 15:37:25 +0000 (09:37 -0600)]
coresight-etm4x: Change the name of the ctxid_val to ctxid_pid

'ctxid_val' array was used to store the value of ETM context ID comparator
which actually stores the process ID to be traced, so using 'ctxid_pid' as
its name instead make it easier to understand.

This patch also changes the ABI, it is normally not allowed, but
fortunately it is a testing ABI and very new for now. Nevertheless,
if you don't think it should be changed, we could always add an alias
for userspace.

Signed-off-by: Chunyan Zhang <zhang.chunyan@linaro.org>
Reviewed-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agocoresight-etm3x: Change the name of the ctxid_val to ctxid_pid
Chunyan Zhang [Fri, 31 Jul 2015 15:37:24 +0000 (09:37 -0600)]
coresight-etm3x: Change the name of the ctxid_val to ctxid_pid

'ctxid_val' array was used to store the value of ETM context ID comparator
which actually stores the process ID to be traced, so using 'ctxid_pid' as
its name instead make it easier to understand.

This patch also changes the ABI, it is normally not allowed, but
fortunately it is a testing ABI and very new for now. Nevertheless,
if you don't think it should be changed, we could always add an alias
for userspace.

Signed-off-by: Chunyan Zhang <zhang.chunyan@linaro.org>
Reviewed-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agocoresight: replicator: Use builtin_platform_driver()
Vaishali Thakkar [Fri, 31 Jul 2015 15:37:23 +0000 (09:37 -0600)]
coresight: replicator: Use builtin_platform_driver()

Macro builtin_platform_driver can be used for builtin drivers
that don't do anything in driver init. This file depends on
Kconfig CONFIG_CORESIGHT_LINKS_AND_SINKS which eventually
depends on CORESIGHT. Both CONFIG_CORESIGHT_LINKS_AND_SINKS and
CORESIGHT are bool. So, use builtin_platform_driver and remove
some boilerplate code.

Also, remove header file init.h as functionality like module_init
and module_exit is now relocated to module.h.

Signed-off-by: Vaishali Thakkar <vthakkar1994@gmail.com>
Suggested-by: Paul Bolle <pebolle@tiscali.nl>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agocoresight-etm3x: Add Qualcomm PTM v1.1 peripheral ID
Ivan T. Ivanov [Fri, 31 Jul 2015 15:37:22 +0000 (09:37 -0600)]
coresight-etm3x: Add Qualcomm PTM v1.1 peripheral ID

Add Qualcomm's PTM v1.1 peripheral ID to supported devices.
This device could be found at least in MSM8974 and APQ8064
chipsets.

Signed-off-by: Ivan T. Ivanov <ivan.ivanov@linaro.org>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agocoresight: binding for ETMv4 coresight drivers
Mathieu Poirier [Fri, 31 Jul 2015 15:37:21 +0000 (09:37 -0600)]
coresight: binding for ETMv4 coresight drivers

Adding compatible string for new coresight ETMv4 tracer.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agouio: UIO_PRUSS should depend on HAS_DMA
Geert Uytterhoeven [Mon, 29 Jun 2015 07:12:38 +0000 (09:12 +0200)]
uio: UIO_PRUSS should depend on HAS_DMA

If NO_DMA=y:

    ERROR: "dma_alloc_coherent" [drivers/uio/uio_pruss.ko] undefined!
    ERROR: "dma_free_coherent" [drivers/uio/uio_pruss.ko] undefined!

Add a dependency on HAS_DMA to fix this.

Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agouio: uio_fsl_elbc_gpcm: Use module_platform_driver
Vaishali Thakkar [Wed, 8 Jul 2015 04:54:27 +0000 (10:24 +0530)]
uio: uio_fsl_elbc_gpcm: Use module_platform_driver

Use module_platform_driver for drivers whose init and exit functions
only register and unregister, respectively.

A simplified version of the Coccinelle semantic patch that performs
this transformation is as follows:

@a@
identifier f, x;
@@
-static f(...) { return platform_driver_register(&x); }

@b depends on a@
identifier e, a.x;
@@
-static e(...) { platform_driver_unregister(&x); }

@c depends on a && b@
identifier a.f;
declarer name module_init;
@@
-module_init(f);

@d depends on a && b && c@
identifier b.e, a.x;
declarer name module_exit;
declarer name module_platform_driver;
@@
-module_exit(e);
+module_platform_driver(x);

Signed-off-by: Vaishali Thakkar <vthakkar1994@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agouio: Destroy uio_idr on module exit
Johannes Thumshirn [Wed, 8 Jul 2015 15:24:46 +0000 (17:24 +0200)]
uio: Destroy uio_idr on module exit

Destroy uio_idr on module exit, reclaiming the allocated memory.

This was detected by the following semantic patch (written by Luis Rodriguez
<mcgrof@suse.com>)
<SmPL>
@ defines_module_init @
declarer name module_init, module_exit;
declarer name DEFINE_IDR;
identifier init;
@@

module_init(init);

@ defines_module_exit @
identifier exit;
@@

module_exit(exit);

@ declares_idr depends on defines_module_init && defines_module_exit @
identifier idr;
@@

DEFINE_IDR(idr);

@ on_exit_calls_destroy depends on declares_idr && defines_module_exit @
identifier declares_idr.idr, defines_module_exit.exit;
@@

exit(void)
{
 ...
 idr_destroy(&idr);
 ...
}

@ missing_module_idr_destroy depends on declares_idr && defines_module_exit && !on_exit_calls_destroy @
identifier declares_idr.idr, defines_module_exit.exit;
@@

exit(void)
{
 ...
 +idr_destroy(&idr);
}

</SmPL>

Signed-off-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoti-st: st_kim: use gpio_set_value_cansleep to fix warning
Jürg Billeter [Wed, 24 Jun 2015 10:24:06 +0000 (12:24 +0200)]
ti-st: st_kim: use gpio_set_value_cansleep to fix warning

GPIO accessor functions may sleep.

Signed-off-by: Jürg Billeter <j@bitron.ch>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoRevert "ti-st: add device tree support"
Rob Herring [Wed, 22 Jul 2015 18:17:15 +0000 (13:17 -0500)]
Revert "ti-st: add device tree support"

This reverts commit 46d0d33350e9b32642d745a8b46a954910196b4d.

This binding is horrible and never should have been merged. It is not
documented nor are there any in tree users, so reverting it will not
break anything we care about. Lets revert it before we do have users.

The problems with it are:

- It is not documented.

- The GPIO connection is described with a custom property and uses Linux
GPIO numbering.

- The UART connection is described using the Linux tty device name.

Cc: Gigi Joseph <gigi.joseph@ti.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Rob Herring <robh@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoVMware balloon: Add support for balloon capabilities.
Xavier Deguillard [Fri, 12 Jun 2015 18:43:23 +0000 (11:43 -0700)]
VMware balloon: Add support for balloon capabilities.

In order to extend the balloon protocol, the hypervisor and the guest
driver need to agree on a set of supported functionality to use.

Signed-off-by: Xavier Deguillard <xdeguillard@vmware.com>
Acked-by: Dmitry Torokhov <dtor@vmware.com>
Signed-off-by: Philip P. Moltmann <moltmann@vmware.com>
Acked-by: Andy King <acking@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoVMware balloon: partially inline vmballoon_reserve_page.
Xavier Deguillard [Fri, 12 Jun 2015 18:43:22 +0000 (11:43 -0700)]
VMware balloon: partially inline vmballoon_reserve_page.

This split the function in two: the allocation part is inlined into the
inflate function and the lock part is kept into his own function.

This change is needed in order to be able to allocate more than one page
before doing the hypervisor call.

Signed-off-by: Xavier Deguillard <xdeguillard@vmware.com>
Acked-by: Dmitry Torokhov <dtor@vmware.com>
Signed-off-by: Philip P. Moltmann <moltmann@vmware.com>
Acked-by: Andy King <acking@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agospmi: Select IRQ_DOMAIN instead of depend on it
Stephen Boyd [Tue, 16 Jun 2015 18:37:16 +0000 (11:37 -0700)]
spmi: Select IRQ_DOMAIN instead of depend on it

IRQ_DOMAIN is a hidden config option, so depending on it doesn't
make any sense. Select the config option because it's required to
compile this driver.

Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Reviewed-by: Andy Gross <agross@codeaurora.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agospmi: pmic-arb: add support for irq_get_irqchip_state
Courtney Cavin [Thu, 30 Jul 2015 17:53:54 +0000 (10:53 -0700)]
spmi: pmic-arb: add support for irq_get_irqchip_state

Reviewed-by: Andy Gross <agross@codeaurora.org>
Signed-off-by: Courtney Cavin <courtney.cavin@sonymobile.com>
Signed-off-by: Bjorn Andersson <bjorn.andersson@sonymobile.com>
Tested-by: Tim Bird <tim.bird@sonymobile.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agospmi: add command tracepoints for SPMI
Ankit Gupta [Mon, 22 Jun 2015 22:38:46 +0000 (16:38 -0600)]
spmi: add command tracepoints for SPMI

Add tracepoints to retrieve information about read, write
and non-data commands. For performance measurement support
tracepoints are added at the beginning and at the end of
transfers. Following is a list showing the new tracepoint
events. The "cmd" parameter here represents the opcode, SID,
and full 16-bit address.

spmi_write_begin: cmd and data buffer.
spmi_write_end  : cmd and return value.
spmi_read_begin : cmd.
spmi_read_end   : cmd, return value and data buffer.
spmi_cmd        : cmd.

The reason that cmd appears at both the beginning and at
the end event is that SPMI drivers can request commands
concurrently. cmd helps in matching the corresponding
events.

SPMI tracepoints can be enabled like:

echo 1 >/sys/kernel/debug/tracing/events/spmi/enable

and will dump messages that can be viewed in
/sys/kernel/debug/tracing/trace that look like:

... spmi_read_begin: opc=56 sid=00 addr=0x0000
... spmi_read_end: opc=56 sid=00 addr=0x0000 ret=0 len=02 buf=0x[01-40]
... spmi_write_begin: opc=48 sid=00 addr=0x0000 len=3 buf=0x[ff-ff-ff]

Suggested-by: Sagar Dharia <sdharia@codeaurora.org>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Reviewed-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Gilad Avidov <gavidov@codeaurora.org>
Signed-off-by: Ankit Gupta <ankgupta@codeaurora.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agochar: xillybus: Allow 64-bit DMA on PCIe interface
Eli Billauer [Wed, 5 Aug 2015 10:03:26 +0000 (13:03 +0300)]
char: xillybus: Allow 64-bit DMA on PCIe interface

Until now, only 32-bit DMA addressing was allowed, following a report on
some old Intel machine that dropped 64-bit PCIe packets, even though
pci_set_dma_mask() was successful with DMA_BIT_MASK(64).

But then came TI's Keystone II chip (ARM Cortex A15 + DSPs), which refuses
32-bit DMA addressing (for good reasons). So 64-bit DMA is allowed as a
fallback option.

Signed-off-by: Eli Billauer <eli.billauer@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoDrivers: hv: vmbus: use cpu_hotplug_enable/disable
Vitaly Kuznetsov [Wed, 5 Aug 2015 07:52:48 +0000 (00:52 -0700)]
Drivers: hv: vmbus: use cpu_hotplug_enable/disable

Commit e513229b4c38 ("Drivers: hv: vmbus: prevent cpu offlining on newer
hypervisors") was altering smp_ops.cpu_disable to prevent CPU offlining.
We can bo better by using cpu_hotplug_enable/disable functions instead of
such hard-coding.

Reported-by: Radim Kr.má <rkrcmar@redhat.com>
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>
9 years agocpu-hotplug: export cpu_hotplug_enable/cpu_hotplug_disable
Vitaly Kuznetsov [Wed, 5 Aug 2015 07:52:47 +0000 (00:52 -0700)]
cpu-hotplug: export cpu_hotplug_enable/cpu_hotplug_disable

Hyper-V module needs to disable cpu hotplug (offlining) as there is no
support from hypervisor side to reassign already opened event channels
to a different CPU. Currently it is been done by altering
smp_ops.cpu_disable but it is hackish.

Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agocpu-hotplug: convert cpu_hotplug_disabled to a counter
Vitaly Kuznetsov [Wed, 5 Aug 2015 07:52:46 +0000 (00:52 -0700)]
cpu-hotplug: convert cpu_hotplug_disabled to a counter

As a prerequisite to exporting cpu_hotplug_enable/cpu_hotplug_disable
functions to modules we need to convert cpu_hotplug_disabled to a counter
to properly support disable -> disable -> enable call sequences. E.g.
after Hyper-V vmbus module (which is supposed to be the first user of
exported cpu_hotplug_enable/cpu_hotplug_disable) did cpu_hotplug_disable()
hibernate path calls disable_nonboot_cpus() and if we hit an error in
_cpu_down() enable_nonboot_cpus() will be called on the failure path (thus
making cpu_hotplug_disabled = 0 and leaving cpu hotplug in 'enabled'
state). Same problem is possible if more than 1 module use
cpu_hotplug_disable/cpu_hotplug_enable on their load/unload paths. When
one of these modules is been unloaded it is logical to leave cpu hotplug
in 'disabled' state.

To support the change we need to increse cpu_hotplug_disabled counter
in disable_nonboot_cpus() unconditionally as all users of
disable_nonboot_cpus() are supposed to do enable_nonboot_cpus() in case
an error was returned.

Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoDrivers: hv: vmbus: document the VMBus sysfs files
Dexuan Cui [Wed, 5 Aug 2015 07:52:45 +0000 (00:52 -0700)]
Drivers: hv: vmbus: document the VMBus sysfs files

The 4 sysfs files should be stable ABIs to the user space.

Signed-off-by: Dexuan Cui <decui@microsoft.com>
Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agotools: hv: add a python script lsvmbus to list VMBus devices
Dexuan Cui [Wed, 5 Aug 2015 07:52:44 +0000 (00:52 -0700)]
tools: hv: add a python script lsvmbus to list VMBus devices

By default lsvmbus lists all the devices in the VMBus.
With -v or -vv, more information is printed, including the VMBus
Rel_ID, class ID, device ID and which channel is bound to which
virtual processor, etc.

Signed-off-by: Dexuan Cui <decui@microsoft.com>
Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoDrivers: hv: vmbus: add a sysfs attr to show the binding of channel/VP
Dexuan Cui [Wed, 5 Aug 2015 07:52:43 +0000 (00:52 -0700)]
Drivers: hv: vmbus: add a sysfs attr to show the binding of channel/VP

This is useful to analyze performance issue.

Signed-off-by: Dexuan Cui <decui@microsoft.com>
Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoDrivers: hv: vmbus: Implement a clocksource based on the TSC page
K. Y. Srinivasan [Wed, 5 Aug 2015 07:52:42 +0000 (00:52 -0700)]
Drivers: hv: vmbus: Implement a clocksource based on the TSC page

The current Hyper-V clock source is based on the per-partition reference counter
and this counter is being accessed via s synthetic MSR - HV_X64_MSR_TIME_REF_COUNT.
Hyper-V has a more efficient way of computing the per-partition reference
counter value that does not involve reading a synthetic MSR. We implement
a time source based on this mechanism.

Tested-by: Vivek Yadav <vyadav@microsoft.com>
Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agodrivers/hv: Migrate to new 'set-state' interface
Viresh Kumar [Wed, 5 Aug 2015 07:52:41 +0000 (00:52 -0700)]
drivers/hv: Migrate to new 'set-state' interface

Migrate hv driver to the new 'set-state' interface provided by
clockevents core, the earlier 'set-mode' interface is marked obsolete
now.

This also enables us to implement callbacks for new states of clockevent
devices, for example: ONESHOT_STOPPED.

Cc: "K. Y. Srinivasan" <kys@microsoft.com>
Cc: Haiyang Zhang <haiyangz@microsoft.com>
Cc: devel@linuxdriverproject.org
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoDrivers: hv_vmbus: Fix signal to host condition
Christopher Oo [Wed, 5 Aug 2015 07:52:40 +0000 (00:52 -0700)]
Drivers: hv_vmbus: Fix signal to host condition

Fixes a bug where previously hv_ringbuffer_read would pass in the old
number of bytes available to read instead of the expected old read index
when calculating when to signal to the host that the ringbuffer is empty.
Since the previous write size is already saved, also changes the
hv_need_to_signal_on_read to use the previously read value rather than
recalculating it.

Signed-off-by: Christopher Oo <t-chriso@microsoft.com>
Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoDrivers: hv: vmbus: Further improve CPU affiliation logic
Dexuan Cui [Wed, 5 Aug 2015 07:52:39 +0000 (00:52 -0700)]
Drivers: hv: vmbus: Further improve CPU affiliation logic

Keep track of CPU affiliations of sub-channels within the scope of the primary
channel. This will allow us to better distribute the load amongst available
CPUs.

Signed-off-by: Dexuan Cui <decui@microsoft.com>
Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoDrivers: hv: vmbus: Improve the CPU affiliation for channels
K. Y. Srinivasan [Wed, 5 Aug 2015 07:52:38 +0000 (00:52 -0700)]
Drivers: hv: vmbus: Improve the CPU affiliation for channels

The current code tracks the assigned CPUs within a NUMA node in the context of
the primary channel. So, if we have a VM with a single NUMA node with 8 VCPUs, we may
end up unevenly distributing the channel load. Fix the issue by tracking affiliations
globally.

Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agodrivers:hv: Move MMIO range picking from hyper_fb to hv_vmbus
Jake Oshins [Wed, 5 Aug 2015 07:52:37 +0000 (00:52 -0700)]
drivers:hv: Move MMIO range picking from hyper_fb to hv_vmbus

This patch deletes the logic from hyperv_fb which picked a range of MMIO space
for the frame buffer and adds new logic to hv_vmbus which picks ranges for
child drivers.  The new logic isn't quite the same as the old, as it considers
more possible ranges.

Signed-off-by: Jake Oshins <jakeo@microsoft.com>
Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agodrivers:hv: Modify hv_vmbus to search for all MMIO ranges available.
Jake Oshins [Wed, 5 Aug 2015 07:52:36 +0000 (00:52 -0700)]
drivers:hv: Modify hv_vmbus to search for all MMIO ranges available.

This patch changes the logic in hv_vmbus to record all of the ranges in the
VM's firmware (BIOS or UEFI) that offer regions of memory-mapped I/O space for
use by paravirtual front-end drivers.  The old logic just found one range
above 4GB and called it good.  This logic will find any ranges above 1MB.

It would have been possible with this patch to just use existing resource
allocation functions, rather than keep track of the entire set of Hyper-V
related MMIO regions in VMBus.  This strategy, however, is not sufficient
when the resource allocator needs to be aware of the constraints of a
Hyper-V virtual machine, which is what happens in the next patch in the series.
So this first patch exists to show the first steps in reworking the MMIO
allocation paths for Hyper-V front-end drivers.

Signed-off-by: Jake Oshins <jakeo@microsoft.com>
Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agochar: make misc_deregister a void function
Greg Kroah-Hartman [Thu, 30 Jul 2015 22:59:57 +0000 (15:59 -0700)]
char: make misc_deregister a void function

With well over 200+ users of this api, there are a mere 12 users that
actually checked the return value of this function.  And all of them
really didn't do anything with that information as the system or module
was shutting down no matter what.

So stop pretending like it matters, and just return void from
misc_deregister().  If something goes wrong in the call, you will get a
WARNING splat in the syslog so you know how to fix up your driver.
Other than that, there's nothing that can go wrong.

Cc: Alasdair Kergon <agk@redhat.com>
Cc: Neil Brown <neilb@suse.com>
Cc: Oleg Drokin <oleg.drokin@intel.com>
Cc: Andreas Dilger <andreas.dilger@intel.com>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Wim Van Sebroeck <wim@iguana.be>
Cc: Christine Caulfield <ccaulfie@redhat.com>
Cc: David Teigland <teigland@redhat.com>
Cc: Mark Fasheh <mfasheh@suse.com>
Acked-by: Joel Becker <jlbec@evilplan.org>
Acked-by: Alexandre Belloni <alexandre.belloni@free-electrons.com>
Acked-by: Alessandro Zummo <a.zummo@towertech.it>
Acked-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoDrivers: hv: vmbus: Consider ND NIC in binding channels to CPUs
K. Y. Srinivasan [Sat, 1 Aug 2015 23:08:21 +0000 (16:08 -0700)]
Drivers: hv: vmbus: Consider ND NIC in binding channels to CPUs

We cycle through all the "high performance" channels to distribute
load across the available CPUs. Process the NetworkDirect as a
high performance device.

Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agomshyperv: fix recognition of Hyper-V guest crash MSR's
Denis V. Lunev [Sat, 1 Aug 2015 23:08:20 +0000 (16:08 -0700)]
mshyperv: fix recognition of Hyper-V guest crash MSR's

Hypervisor Top Level Functional Specification v3.1/4.0 notes that cpuid
(0x40000003) EDX's 10th bit should be used to check that Hyper-V guest
crash MSR's functionality available.

This patch should fix this recognition. Currently the code checks EAX
register instead of EDX.

Signed-off-by: Andrey Smetanin <asmetanin@virtuozzo.com>
Signed-off-by: Denis V. Lunev <den@openvz.org>
Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoDrivers: hv: vmbus: don't send CHANNELMSG_UNLOAD on pre-Win2012R2 hosts
Vitaly Kuznetsov [Sat, 1 Aug 2015 23:08:19 +0000 (16:08 -0700)]
Drivers: hv: vmbus: don't send CHANNELMSG_UNLOAD on pre-Win2012R2 hosts

Pre-Win2012R2 hosts don't properly handle CHANNELMSG_UNLOAD and
wait_for_completion() hangs. Avoid sending such request on old hosts.

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>
9 years agoDrivers: hv: vmbus: fix typo in hv_port_info struct
Nik Nyby [Sat, 1 Aug 2015 23:08:18 +0000 (16:08 -0700)]
Drivers: hv: vmbus: fix typo in hv_port_info struct

This fixes a typo: base_flag_bumber to base_flag_number

Signed-off-by: Nik Nyby <nikolas@gnu.org>
Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agohv: util: checking the wrong variable
Dan Carpenter [Sat, 1 Aug 2015 23:08:17 +0000 (16:08 -0700)]
hv: util: checking the wrong variable

We don't catch this allocation failure because there is a typo and we
check the wrong variable.

Fixes: 14b50f80c32d ('Drivers: hv: util: introduce hv_utils_transport abstraction')
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Reviewed-by: Vitaly Kuznetsov <vkuznets@redhat.com>
Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoDrivers: hv: vmbus: Permit sending of packets without payload
K. Y. Srinivasan [Sat, 1 Aug 2015 23:08:14 +0000 (16:08 -0700)]
Drivers: hv: vmbus: Permit sending of packets without payload

The guest may have to send a completion packet back to the host.
To support this usage, permit sending a packet without a payload -
we would be only sending the descriptor in this case.

Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoDrivers: hv: balloon: Enable dynamic memory protocol negotiation with Windows 10...
Alex Ng [Sat, 1 Aug 2015 23:08:13 +0000 (16:08 -0700)]
Drivers: hv: balloon: Enable dynamic memory protocol negotiation with Windows 10 hosts

Support Win10 protocol for Dynamic Memory. Thia patch allows guests on Win10 hosts
to hot-add memory even when dynamic memory is not enabled on the guest.

Signed-off-by: Alex Ng <alexng@microsoft.com>
Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoDrivers: hv: fcopy: dynamically allocate smsg_out in fcopy_send_data()
Vitaly Kuznetsov [Sat, 1 Aug 2015 23:08:12 +0000 (16:08 -0700)]
Drivers: hv: fcopy: dynamically allocate smsg_out in fcopy_send_data()

struct hv_start_fcopy is too big to be on stack on i386, the following
warning is reported:

>> drivers/hv/hv_fcopy.c:159:1: warning: the frame size of 1088 bytes is larger than 1024 bytes [-Wframe-larger-than=]

Reported-by: kbuild test robot <fengguang.wu@intel.com>
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>
9 years agoDrivers: hv: kvp: check kzalloc return value
Vitaly Kuznetsov [Sat, 1 Aug 2015 23:08:11 +0000 (16:08 -0700)]
Drivers: hv: kvp: check kzalloc return value

kzalloc() return value check was accidentally lost in 11bc3a5fa91f:
"Drivers: hv: kvp: convert to hv_utils_transport" commit.

We don't need to reset kvp_transaction.state here as we have the
kvp_timeout_func() timeout function and in case we're in OOM situation
it is preferable to wait.

Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
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>
9 years agoDrivers: hv: vmbus: prefer 'die' notification chain to 'panic'
Vitaly Kuznetsov [Sat, 1 Aug 2015 23:08:10 +0000 (16:08 -0700)]
Drivers: hv: vmbus: prefer 'die' notification chain to 'panic'

current_pt_regs() sometimes returns regs of the userspace process and in
case of a kernel crash this is not what we need to report. E.g. when we
trigger crash with sysrq we see the following:
...
 RIP: 0010:[<ffffffff815b8696>]  [<ffffffff815b8696>] sysrq_handle_crash+0x16/0x20
 RSP: 0018:ffff8800db0a7d88  EFLAGS: 00010246
 RAX: 000000000000000f RBX: ffffffff820a0660 RCX: 0000000000000000
...
at the same time current_pt_regs() give us:
ip=7f899ea7e9e0, ax=ffffffffffffffda, bx=26c81a0, cx=7f899ea7e9e0, ...
These registers come from the userspace process triggered the crash. As we
don't even know which process it was this information is rather useless.

When kernel crash happens through 'die' proper regs are being passed to
all receivers on the die_chain (and panic_notifier_list is being notified
with the string passed to panic() only). If panic() is called manually
(e.g. on BUG()) we won't get 'die' notification so keep the 'panic'
notification reporter as well but guard against double reporting.

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>
9 years agoDrivers: hv: vmbus: add special crash handler
Vitaly Kuznetsov [Sat, 1 Aug 2015 23:08:09 +0000 (16:08 -0700)]
Drivers: hv: vmbus: add special crash handler

Full kernel hang is observed when kdump kernel starts after a crash. This
hang happens in vmbus_negotiate_version() function on
wait_for_completion() as Hyper-V host (Win2012R2 in my testing) never
responds to CHANNELMSG_INITIATE_CONTACT as it thinks the connection is
already established. We need to perform some mandatory minimalistic
cleanup before we start new kernel.

Reported-by: K. Y. Srinivasan <kys@microsoft.com>
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>
9 years agoDrivers: hv: don't do hypercalls when hypercall_page is NULL
Vitaly Kuznetsov [Sat, 1 Aug 2015 23:08:08 +0000 (16:08 -0700)]
Drivers: hv: don't do hypercalls when hypercall_page is NULL

At the very late stage of kexec a driver (which are not being unloaded) can
try to post a message or signal an event. This will crash the kernel as we
already did hv_cleanup() and the hypercall page is NULL.

Move all common (between 32 and 64 bit code) declarations to the beginning
of the do_hypercall() function. Unfortunately we have to write the
!hypercall_page check twice to not mix declarations and code.

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>
9 years agoDrivers: hv: vmbus: add special kexec handler
Vitaly Kuznetsov [Sat, 1 Aug 2015 23:08:07 +0000 (16:08 -0700)]
Drivers: hv: vmbus: add special kexec handler

When general-purpose kexec (not kdump) is being performed in Hyper-V guest
the newly booted kernel fails with an MCE error coming from the host. It
is the same error which was fixed in the "Drivers: hv: vmbus: Implement
the protocol for tearing down vmbus state" commit - monitor pages remain
special and when they're being written to (as the new kernel doesn't know
these pages are special) bad things happen. We need to perform some
minimalistic cleanup before booting a new kernel on kexec. To do so we
need to register a special machine_ops.shutdown handler to be executed
before the native_machine_shutdown(). Registering a shutdown notification
handler via the register_reboot_notifier() call is not sufficient as it
happens to early for our purposes. machine_ops is not being exported to
modules (and I don't think we want to export it) so let's do this in
mshyperv.c

The minimalistic cleanup consists of cleaning up clockevents, synic MSRs,
guest os id MSR, and hypercall MSR.

Kdump doesn't require all this stuff as it lives in a separate memory
space.

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>
9 years agokexec: define kexec_in_progress in !CONFIG_KEXEC case
Vitaly Kuznetsov [Sat, 1 Aug 2015 23:08:06 +0000 (16:08 -0700)]
kexec: define kexec_in_progress in !CONFIG_KEXEC case

If some piece of code wants to check kexec_in_progress it has to be put
in #ifdef CONFIG_KEXEC block to not break the build in !CONFIG_KEXEC
case. Overcome this limitation by defining kexec_in_progress to false.

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>
9 years agoDrivers: hv: vmbus: remove hv_synic_free_cpu() call from hv_synic_cleanup()
Vitaly Kuznetsov [Sat, 1 Aug 2015 23:08:05 +0000 (16:08 -0700)]
Drivers: hv: vmbus: remove hv_synic_free_cpu() call from hv_synic_cleanup()

We already have hv_synic_free() which frees all per-cpu pages for all
CPUs, let's remove the hv_synic_free_cpu() call from hv_synic_cleanup()
so it will be possible to do separate cleanup (writing to MSRs) and final
freeing. This is going to be used to assist kexec.

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>
9 years agomei: remove check on pm_runtime_active in __mei_cl_disconnect
Tomas Winkler [Tue, 4 Aug 2015 11:30:11 +0000 (14:30 +0300)]
mei: remove check on pm_runtime_active in __mei_cl_disconnect

Remove bogus check on pm_runtime_active that prevented
disconnection from a client in case the device was resuming
from power gating but not yet active.

Fix regression introduced by
18901357e70ae29e3fd1c58712a6847c2ae52eae
mei: disconnect on connection request timeout

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Alexander Usyskin <alexander.usyskin@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agomei: hbm: bump supported HBM version to 2.0
Tomas Winkler [Sun, 2 Aug 2015 19:20:58 +0000 (22:20 +0300)]
mei: hbm: bump supported HBM version to 2.0

HBM 2.0 version for Sunrise point Skylake (PCH) based devices

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agomei: me: add sunrise point device ids
Tomas Winkler [Sun, 2 Aug 2015 19:20:57 +0000 (22:20 +0300)]
mei: me: add sunrise point device ids

Add MEI devices ids for Intel Sunrisepoint Skylake (PCH)

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agomei: me: d0i3: exit d0i3 on driver start and enter it on stop
Alexander Usyskin [Sun, 2 Aug 2015 19:20:56 +0000 (22:20 +0300)]
mei: me: d0i3: exit d0i3 on driver start and enter it on stop

A BIOS may put the device in d0i3 on platform initialization so it won’t
consume power even if the driver is not present, in turn the driver has
to wake up the devices on load in order to perform the initialization
sequence and move it back to low power state on driver remove.

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>
9 years agomei: me: d0i3: move mei_me_hw_reset down in the file
Alexander Usyskin [Sun, 2 Aug 2015 19:20:55 +0000 (22:20 +0300)]
mei: me: d0i3: move mei_me_hw_reset down in the file

Move mei_me_hw_reset down in the source file to avoid
forward declarations when introducing d0i3 flow in the next patch.

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>
9 years agomei: me: d0i3: add d0i3 enter/exit state machine
Alexander Usyskin [Sun, 2 Aug 2015 19:20:54 +0000 (22:20 +0300)]
mei: me: d0i3: add d0i3 enter/exit state machine

Rework mei power gating state machine to support entry and exit to and
from D0i3 power state.
The choice between legacy and D0i3 routines is conditioned on
d0i3_supported flag.

The patch introduces warning:
drivers/misc/mei/hw-me.c:901:12: warning: ‘mei_me_d0i3_enter’ defined but not used [-Wunused-function]
it will go away in consequent patch

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>
9 years agomei: hbm: reorganize the power gating responses
Alexander Usyskin [Sun, 2 Aug 2015 19:20:53 +0000 (22:20 +0300)]
mei: hbm: reorganize the power gating responses

Before adding support for D0i3 we need to reorganize the hbm pg handling
Move HBM PG response code to dedicated functions in order to unclutter
hbm command switch.
Add check for the right system state before message processing and
return -EPROTO in state mismatch case.

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>
9 years agomei: me: d0i3: enable d0i3 interrupts
Alexander Usyskin [Sun, 2 Aug 2015 19:20:52 +0000 (22:20 +0300)]
mei: me: d0i3: enable d0i3 interrupts

D0i3 adds additional interrupt reason bit, therefore we add a variable
intr_source to save the interrupt causes for further dispatching.
The interrupt cause is saved in the irq quick handler to achieve
unified behavior for both MSI enabled and shared interrupt platforms.

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>
9 years agomei: me: d0i3: add flag to indicate D0i3 support
Alexander Usyskin [Sun, 2 Aug 2015 19:20:51 +0000 (22:20 +0300)]
mei: me: d0i3: add flag to indicate D0i3 support

Detect d0i3 low power state during hw configuration,
the value is set in HFS_1 pci config reigister.

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>
9 years agomei: me: d0i3: add the control registers
Tomas Winkler [Sun, 2 Aug 2015 19:20:50 +0000 (22:20 +0300)]
mei: me: d0i3: add the control registers

Starting with Intel Sunrisepoint (Skylake PCH) the MEI device
supports D0i3 low power state. Add D0i3 control registers.

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agomei: hbm: add new error code MEI_CL_CONN_NOT_ALLOWED
Alexander Usyskin [Wed, 29 Jul 2015 11:59:33 +0000 (14:59 +0300)]
mei: hbm: add new error code MEI_CL_CONN_NOT_ALLOWED

The device can return error 5 (NOT_ALLOWED) on connection attempt.
This error can happen if:
1. An another connection attempt is in progress
2. There is an attempt to connect a fixed (connectionless) client
3. The number of available connections is exceeded (new in HBM 2.0)

We should not hit that error unless there is an internal book keeping
hiccup except option (3), therefore we translate the error code
to errno EBUSY;

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>
9 years agomei: bus: add and call callback on notify event
Alexander Usyskin [Sun, 26 Jul 2015 06:54:23 +0000 (09:54 +0300)]
mei: bus: add and call callback on notify event

Enable drivers on mei client bus to subscribe
to asynchronous event notifications.
Introduce events_mask to the existing callback infrastructure
so it is possible to handle both RX and event notification.

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>
9 years agomei: implement fasync for event notification
Tomas Winkler [Sun, 26 Jul 2015 06:54:22 +0000 (09:54 +0300)]
mei: implement fasync for event notification

A process can be informed about client notification also via
SIGIO with POLL_PRI event.

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agomei: support polling for event notification
Tomas Winkler [Sun, 26 Jul 2015 06:54:21 +0000 (09:54 +0300)]
mei: support polling for event notification

Polling on priority events is translated on waiting for event
notification. One need to enable notification prior for
calling select or poll system call otherwise process
will not wait.

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Alexander Usyskin <alexander.usyskin@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agomei: add async event notification ioctls
Tomas Winkler [Sun, 26 Jul 2015 06:54:20 +0000 (09:54 +0300)]
mei: add async event notification ioctls

Add ioctl IOCTL_MEI_NOTIFY_SET for enabling and disabling
async event notification.
Add ioctl IOCTL_MEI_NOTIFY_GET for receiving and acking
an event notification.

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Alexander Usyskin <alexander.usyskin@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agomei: add a handler that waits for notification on event
Tomas Winkler [Sun, 26 Jul 2015 06:54:19 +0000 (09:54 +0300)]
mei: add a handler that waits for notification on event

mei_cl_notify_get is to be called by a host client
to wait, receive, and ack the event notification.

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Alexander Usyskin <alexander.usyskin@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agomei: add mei_cl_notify_request command
Tomas Winkler [Sun, 26 Jul 2015 06:54:18 +0000 (09:54 +0300)]
mei: add mei_cl_notify_request command

Add per client notification request infrastructure
that allows client to enable or disable async
event notification.

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Alexander Usyskin <alexander.usyskin@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agomei: enable async event notifications only from hbm version 2.0
Tomas Winkler [Sun, 26 Jul 2015 06:54:17 +0000 (09:54 +0300)]
mei: enable async event notifications only from hbm version 2.0

Only FW version 2.0 and newer support the async event
notification. For backward compatibility block the feature
if the FW version is older then 2.0

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Alexander Usyskin <alexander.usyskin@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agomei: implement async notification hbm messages
Tomas Winkler [Sun, 26 Jul 2015 06:54:16 +0000 (09:54 +0300)]
mei: implement async notification hbm messages

Implement sending and reception handlers for the
async event notification hbm commands.
Add client notification book keeping data required for the messages
    notify_en to indicate whether notification is enabled
    notify_ev to indicate whether an event is pending

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Alexander Usyskin <alexander.usyskin@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agomei: define async notification hbm commands
Tomas Winkler [Sun, 26 Jul 2015 06:54:15 +0000 (09:54 +0300)]
mei: define async notification hbm commands

FW has gained new capability where a FW client can asynchronously
notify the host that an event has occurred in its process.
The notification doesn't provide any data and host may need to query
further the FW client in order to get details of the event.
Host can subscribe or unsubscribe to the event notification via
designated HBM commands, and also the notification is carried on
a new HBM command.
This patch adds definitions of asynchronous notification HBM commands.

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agomei: disconnect on connection request timeout
Alexander Usyskin [Thu, 23 Jul 2015 18:37:13 +0000 (21:37 +0300)]
mei: disconnect on connection request timeout

For the FW with  HBM version >= 2.0 we don't need to reset the whole
device in case of a particular client failing to connect, it is enough
to send disconnect a request to bring the device to the stable state.

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>
9 years agomei: support for dynamic clients
Tomas Winkler [Thu, 23 Jul 2015 18:37:12 +0000 (21:37 +0300)]
mei: support for dynamic clients

HBM version 2.0 and above allows ME clients in the system to
register/unregister after the system is fully initialized.
Clients may be added or removed after enum_resp message was
received

1. To preserve backward compatibility the driver can opt-in to receive
client add messages by setting allow_add field in enum_req

2. A new client is added upon reception of MEI_HBM_ADD_CLIENT_REQ_CMD

3. A client is removed in a lazy manner when connection request
respond with MEI_HBMS_CLIENT_NOT_FOUND status

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agomei: bus: link client devices instead of host clients
Tomas Winkler [Thu, 23 Jul 2015 12:08:47 +0000 (15:08 +0300)]
mei: bus: link client devices instead of host clients

MEI bus was designed around nfc and was hard to extend.
Instead of the hard coded way of adding the devices on the mei bus
we scan the whole me client list and create a device for each
eligible me client (mei_cl_bus_rescan); currently we support
only clients with single connection and fixed address clients.
NFC radio name detection is run as a fixup routine

The patch replaces handling the device list based on struct me_cl
to device list based on me_cl_devices. The creating a connection
is pushed from the device creation time to device enablement.

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agomei: bus: simplify how we build nfc bus name
Tomas Winkler [Thu, 23 Jul 2015 12:08:46 +0000 (15:08 +0300)]
mei: bus: simplify how we build nfc bus name

Remove the dependency on struct ndev from the nfc device
name creation function so it is possible to use it
in a fixup routine

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agomei: bus: blacklist clients by number of connections
Tomas Winkler [Thu, 23 Jul 2015 12:08:45 +0000 (15:08 +0300)]
mei: bus: blacklist clients by number of connections

Currently we support only clients with single connection
and fixed address clients so all other clients are blacklisted

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agomei: bus: blacklist the nfc info client
Tomas Winkler [Thu, 23 Jul 2015 12:08:44 +0000 (15:08 +0300)]
mei: bus: blacklist the nfc info client

Blacklist nfc info client which is only used for retrieval
of the NFC radio version

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agomei: bus: enable running fixup routines before device registration
Tomas Winkler [Thu, 23 Jul 2015 12:08:43 +0000 (15:08 +0300)]
mei: bus: enable running fixup routines before device registration

Split the device registration into allocation and device struct
initialization, device setup, and the final device registration.
This why it is possible to run fixups and quirks during the setup stage
on an initialized device. Each fixup routine effects do_match flag.
If the flag is set to false at the end the device won't be
registered on the bus.

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agomei: bus: add me client device list infrastructure
Tomas Winkler [Thu, 23 Jul 2015 12:08:42 +0000 (15:08 +0300)]
mei: bus: add me client device list infrastructure

Instead of holding the list of host clients (me_cl)
we want to keep the list me client devices (mei_cl_device)
This way we can create host to me client connection only when needed.
Add list head to mei_cl_device and cl_bus_lock
Add bus_added flag to the me client (mei_me_client) to track if
the appropriate mei_cl_device was already created and is_added
flag to mei_cl_device to track if it was already added to the device
list across the bus rescans

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agomei: bus: add reference to bus device in struct mei_cl_client
Tomas Winkler [Thu, 23 Jul 2015 12:08:41 +0000 (15:08 +0300)]
mei: bus: add reference to bus device in struct mei_cl_client

Add reference to the bus device (mei_device) for easier access.
To ensures that referencing cldev->bus is valid during cldev life time
we increase the bus ref counter on a client device creation and drop it
on the device release.

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agomei: bus: revamp probe and remove functions
Tomas Winkler [Thu, 23 Jul 2015 12:08:40 +0000 (15:08 +0300)]
mei: bus: revamp probe and remove functions

Instead of generating device id on the fly during probing we
find the matching id entry on the device id table.
Get bus the module reference counter so it cannot
be unloaded after the driver has bounded to the client
device

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agomei: bus: revamp device matching
Tomas Winkler [Thu, 23 Jul 2015 12:08:39 +0000 (15:08 +0300)]
mei: bus: revamp device matching

mei_cl_device_match now calls mei_cl_device_find that returns
the matching device id in the device id table.
We will utilize the mei_cl_device_find during probing
to locate the matching entry.

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agomei: bus: report if event registration failed
Tomas Winkler [Thu, 23 Jul 2015 12:08:38 +0000 (15:08 +0300)]
mei: bus: report if event registration failed

If event registeration has failed, the caller should know
about it.

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agomei: bus: don't enable events implicitly in device enable
Tomas Winkler [Thu, 23 Jul 2015 12:08:37 +0000 (15:08 +0300)]
mei: bus: don't enable events implicitly in device enable

Do not enable events implicitly in mei_cl_enable_device, it should be
done explicitly using mei_cl_register_event_cb so the events
are enabled only when needed.
The NFC drivers has been already using it that way so no need for
further changes just remove the code from mei_cl_enable_device.

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agomei: bus: rename uevent handler to mei_cl_device_uevent
Tomas Winkler [Thu, 23 Jul 2015 12:08:36 +0000 (15:08 +0300)]
mei: bus: rename uevent handler to mei_cl_device_uevent

Rename mei_cl_uevent to mei_cl_device_uevent to match
the naming convention of mei_cl_bus_type functions

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agomei: bus: move driver api functions at the start of the file
Tomas Winkler [Thu, 23 Jul 2015 12:08:35 +0000 (15:08 +0300)]
mei: bus: move driver api functions at the start of the file

To make the file more organize move mei client driver api
to the start of the file and add Kdoc.

There are no functional changes in this patch.

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agomei: bus: rename nfc.c to bus-fixup.c
Tomas Winkler [Thu, 23 Jul 2015 12:08:34 +0000 (15:08 +0300)]
mei: bus: rename nfc.c to bus-fixup.c

The bus-fixup.c will be a place for fixups and quirks
for all types of me client devices.
As for now it contians only the fixup for setting
the nfc device name on the me client bus.

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agomei: bus: fix drivers and devices names confusion
Tomas Winkler [Thu, 23 Jul 2015 12:08:33 +0000 (15:08 +0300)]
mei: bus: fix drivers and devices names confusion

In the mei bus layer there is use of different variables
of driver and device types with no clear naming convention.
There are generic struct device and struct driver,
then mei_cl_{device, driver}, and finally mei_device which
in this context serves as a bus device.

The patch sets following naming convention:

the variables of type struct device remains dev
the variables of type struct driver remains drv
the variables of type struct mei_cl_device are now cldev
the variables of type struct mei_cl_driver are now cldrv
the variables of type struct mei_device are now bus, in bus
layer context

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>