The Tegra XHCI driver does not currently manage the Tegra XUSB power
partitions and so it these partitions have not been enabled by the
bootloader then the system will crash when probing the XHCI device.
While proper support for managing the power partitions is being
developed to the XHCI driver for Tegra, for now power on all the XUSB
partitions for USB host and super-speed on boot if the XHCI driver is
enabled.
Signed-off-by: Jon Hunter <jonathanh@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
If CONFIG_PM_GENERIC_DOMAINS is not enabled, then power partitions
associated with a device will not be enabled automatically by the PM
core when the device is in use. To avoid situations where a device in
a power partition is to be used but the partition is not enabled,
initialise the power partitions for Tegra early in the boot process and
if CONFIG_PM_GENERIC_DOMAINS is not enabled, then power on all
partitions defined in the device-tree blob.
Note that if CONFIG_PM_GENERIC_DOMAINS is not enabled, after the
partitions are turned on, the clocks and resets used as part of the
sequence for turning on the partition are released again as they are no
longer needed by the PMC driver. Another benefit of this is that this
avoids any issues of sharing resets between the PMC driver and other
device drivers that may wish to independently control a particular
reset.
Signed-off-by: Jon Hunter <jonathanh@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
When initialising a powergate, only a single error message is shown if
the initialisation fails. Add more error messages to give specific
details of what failed if the initialisation failed and remove the
generic failure message.
Signed-off-by: Jon Hunter <jonathanh@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
Commit 0259f522e0 ('soc/tegra: pmc: Restore base address on probe
failure') fixes an issue where the PMC base address pointer is not
restored on probe failure. However, this fix creates another problem
where if early initialisation of the PMC driver fails and an initial
mapping for the PMC address space is not created, then when the PMC
device is probed, the PMC base address pointer will not be valid and
this will cause a crash when tegra_pmc_init() is called and attempts
to access a register.
Although the PMC address space is mapped a 2nd time during the probe
and so this could be fixed by populating the base address pointer
earlier during the probe, this adds more complexity to the code.
Moreover, the PMC probe also assumes the the soc data pointer is also
initialised when the device is probed and if not will also lead to a
crash when calling tegra_pmc_init_tsense_reset(). Given that if the
early initialisation does fail then something bad has happen, it seems
acceptable to allow the PMC device probe to fail as well. Therefore, if
the PMC base address pointer or soc data pointer are not valid when
probing the PMC device, WARN and return an error.
Fixes: 0259f522e0 ('soc/tegra: pmc: Restore base address on probe failure')
Signed-off-by: Jon Hunter <jonathanh@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
Add missing of_node_put() in PMC early initialisation function to avoid
leaking the device nodes.
Signed-off-by: Jon Hunter <jonathanh@nvidia.com>
[treding@nvidia.com: squash in a couple more of_node_put() calls]
Signed-off-by: Thierry Reding <treding@nvidia.com>
The mutex used by the PMC driver may not be initialised if early
initialisation of the driver fails. If this does happen, then it could
be possible for callers of the public PMC functions to still attempt to
acquire the mutex. Fix this by initialising the mutex as soon as
possible to ensure it will always be initialised.
Signed-off-by: Jon Hunter <jonathanh@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
The public functions exported by the PMC driver use the presence of the
SoC data pointer to determine if the PMC device is configured and the
registers can be accessed. However, the SoC data is populated before the
PMC register space is mapped and this opens a window where the SoC data
pointer is valid but the register space has not yet been mapped which
could lead to a crash. Furthermore, if the mapping of the PMC register
space fails, then the SoC data pointer is not cleared and so would
expose a larger window where a crash could occur.
Fix this by initialising the SoC data pointer after the PMC register
space has been mapped.
Signed-off-by: Jon Hunter <jonathanh@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
During early initialisation, the available power partitions for a given
device is configured as well as the polarity of the PMC interrupt. Both
of which should only be configured if there is a valid device node for
the PMC device. This is because the soc data used for configuring the
power partitions is only available if a device node for the PMC is found
and the code to configure the interrupt polarity uses the device node
pointer directly.
Some early device-tree images may not have this device node and so fix
this by ensuring the device node pointer is valid when configuring these
items.
Signed-off-by: Jon Hunter <jonathanh@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
The function tegra_power_sequence_power_up() is a public function used
to power on a partition. When this function is called, we do not check
to see if the partition being powered up is valid/available. Fix this
by checking to see that the partition is valid/available.
Signed-off-by: Jon Hunter <jonathanh@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
When registering the Tegra power partitions with the generic PM domain
framework, the current state of the each partition is checked and used
as the default state for the partition. However, the state of each reset
associated with the partition is not initialised and so it is possible
that the state of the resets are not in the expected state. For example,
if a partition is on, then the resets should be de-asserted and if the
partition is off, the resets should be asserted.
There have been cases where the bootloader has powered on a partition
and only de-asserted some of the resets to some of the devices in the
partition. This can cause accesses to these devices to hang the system
when the kernel boots and attempts to probe these devices.
Ideally, the driver for the device should ensure the reset has been
de-asserted when probing, but the resets cannot be shared between the
PMC driver (that needs to de-assert/assert the reset when turning the
partition on or off) and another driver because we cannot ensure the
reset is in the correct state.
To ensure the resets are in the correct state, when using the generic
PM domain framework, put each reset associated with the partition in
the correct state (based upon the partition's current state) when
obtaining the resets for a partition.
Signed-off-by: Jon Hunter <jonathanh@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
Use register definitions for the main SoC reset operation instead of
hard-coding magic values. Note that the PMC_RST_STATUS register isn't
actually accessed, but since it is mentioned in a comment the
definitions are added for completeness.
Acked-by: Jon Hunter <jonathanh@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
Adds generic PM domain support to the PMC driver where the PM domains
are populated from device-tree and the PM domain consumer devices are
bound to their relevant PM domains via device-tree as well.
Update the tegra_powergate_sequence_power_up() API so that internally
it calls the same tegra_powergate_xxx functions that are used by the
Tegra generic PM domain code for consistency.
To ensure that the Tegra power domains (a.k.a. powergates) cannot be
controlled via both the legacy tegra_powergate_xxx functions as well
as the generic PM domain framework, add a bit map for available
powergates that can be controlled via the legacy powergate functions.
Move the majority of the tegra_powergate_remove_clamping() function
to a sub-function, so that this can be used by both the legacy and
generic power domain code.
This is based upon work by Thierry Reding <treding@nvidia.com>
and Vince Hsu <vinceh@nvidia.com>.
Signed-off-by: Jon Hunter <jonathanh@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
Currently, the function tegra_powergate_set() simply sets the desired
powergate state but does not wait for the state to change. In most cases
we should wait for the state to change before proceeding. Currently,
there is a case for Tegra114 and Tegra124 devices where we do not wait
when starting the secondary CPU as this is not necessary. However, this
is only done at boot time and so waiting here will only have a small
impact on boot time. Therefore, update tegra_powergate_set() to wait
when setting the powergate.
By adding this feature, we can also eliminate the polling loop from
tegra30_boot_secondary().
A function has been added for checking the status of the powergate and
so update the tegra_powergate_is_powered() to use this macro as well.
Signed-off-by: Jon Hunter <jonathanh@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
For Tegra124 and Tegra210, the GPU partition cannot be toggled on and
off via the APBDEV_PMC_PWRGATE_TOGGLE_0 register. For these devices, the
partition is simply powered up and down via an external regulator.
For these devices, there is a separate register for controlling the
signal clamping of the partition and this is described in the PMC SoC
data by the "has_gpu_clamp" variable. Use this variable to determine if
the GPU partition can be controlled via the APBDEV_PMC_PWRGATE_TOGGLE_0
register and ensure that no one can incorrectly try to toggle the GPU
partition via the APBDEV_PMC_PWRGATE_TOGGLE_0 register.
Furthermore, we cannot use the APBDEV_PMC_PWRGATE_STATUS_0 register to
determine if the GPU partition is powered for Tegra124 and Tegra210.
However, if the GPU partition is powered, then the signal clamp for the
GPU partition should be removed and so use bit 0 of the
APBDEV_PMC_GPU_RG_CNTRL_0 register to determine if the clamp has been
removed (bit[0] = 0) and the GPU partition is powered.
Signed-off-by: Jon Hunter <jonathanh@nvidia.com>
Reviewed-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Thierry Reding <treding@nvidia.com>
The function tegra_powergate_is_powered() verifies that the partition
being queried is valid and so there is no need to check this before
calling tegra_powergate_is_powered() in powergate_show(). So remove this
extra check.
Signed-off-by: Jon Hunter <jonathanh@nvidia.com>
Reviewed-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Thierry Reding <treding@nvidia.com>
The Tegra power partitions are referenced by numerical IDs which are the
same values programmed into the PMC registers for controlling the
partition. For a given device, the valid partition IDs may not be
contiguous and so simply checking that an ID is not greater than the
maximum ID supported may not mean it is valid. Fix this by checking if
the powergate is defined in the list of powergates for the Tegra SoC.
Add a helper function for checking valid powergates and use where we
need to verify if the powergate ID is valid or not.
Signed-off-by: Jon Hunter <jonathanh@nvidia.com>
Reviewed-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Thierry Reding <treding@nvidia.com>
In tegra_powergate_set() the state of the powergates is read and OR'ed
with the bit for the powergate of interest. This unsigned 32-bit value
is then compared with a boolean value to test if the powergate is
already in the desired state. When turning on a powergate, apart from
the powergate that is represented by bit 0, this test will always
return false and so we may attempt to turn on the powergate when it is
already on.
After OR'ing the bit for the powergate, check if the result is not equal
to zero before comparing with the boolean value. Add a helper function
to return the current state of a powergate and use this in both
tegra_powergate_set() and tegra_powergate_is_powered() where we check
the powergate status.
Signed-off-by: Jon Hunter <jonathanh@nvidia.com>
Reviewed-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Thierry Reding <treding@nvidia.com>
The Tegra powergate and rail IDs are always positive values and so change
the type to be unsigned and remove the tests to see if the ID is less
than zero. Update the Tegra DC powergate type to be an unsigned as well.
Signed-off-by: Jon Hunter <jonathanh@nvidia.com>
Reviewed-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Thierry Reding <treding@nvidia.com>
The PMC base address pointer is initialised during early boot so that
early platform code may used the PMC public functions. During the probe
of the PMC driver the base address pointer is mapped again and the initial
mapping is freed. This exposes a window where a device accessing the PMC
registers via one of the public functions, could race with the updating
of the pointer and lead to a invalid access. Furthermore, the only
protection between multiple devices attempting to access the PMC registers
is when setting the powergate state to on or off. None of the other public
functions that access the PMC registers are protected.
Use the existing mutex to protect paths that may race with regard to
accessing the PMC registers.
Note that functions tegra_io_rail_prepare()/poll() either return a
negative value on failure or zero on success. Therefore, it is not
necessary to check if the return value is less than zero and so only
test that the return value is not zero to test for failure. This
simplifies the error handling with the mutex locking in place.
Signed-off-by: Jon Hunter <jonathanh@nvidia.com>
Reviewed-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Thierry Reding <treding@nvidia.com>
During early initialisation, the PMC registers are mapped and the PMC SoC
data is populated in the PMC data structure. This allows other drivers
access the PMC register space, via the public Tegra PMC APIs, prior to
probing the PMC device.
When the PMC device is probed, the PMC registers are mapped again and if
successful the initial mapping is freed. If the probing of the PMC device
fails after the registers are remapped, then the registers will be
unmapped and hence the pointer to the PMC registers will be invalid. This
could lead to a potential crash, because once the PMC SoC data pointer is
populated, the driver assumes that the PMC register mapping is also valid
and a user calling any of the public Tegra PMC APIs could trigger an
exception because these APIs don't check that the mapping is still valid.
Fix this by updating the mapping and freeing the original mapping only if
probing the PMC device is successful.
Signed-off-by: Jon Hunter <jonathanh@nvidia.com>
Reviewed-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Thierry Reding <treding@nvidia.com>
Tegra124 does not have an L2 power partition and the L2 cache is part of
the cluster 0 non-CPU (CONC) partition. Remove the L2 as a valid
partition for Tegra124. The TRM also shows that there is no L2 partition
for Tegra124.
Signed-off-by: Jon Hunter <jonathanh@nvidia.com>
Reviewed-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Thierry Reding <treding@nvidia.com>
The power partitions L2, HEG, CELP and C1NC do not exist on Tegra210 but
were incorrectly documented in the TRM. These will be removed from the
TRM and so also remove their definitions.
Signed-off-by: Jon Hunter <jonathanh@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
The debugfs entry for the PMC device will not be removed if the probe of
the device fails to register the restart handler. This leaves behind the
dangling debugfs entry with no driver backing it. Remove the entry to
avoid this.
Signed-off-by: Jon Hunter <jonathanh@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
Sparse reports the following warning for tegra_pmc_init_tsense_reset():
drivers/soc/tegra/pmc.c:741:6: warning: symbol 'tegra_pmc_init_tsense_reset' was not declared. Should it be static?
This function is only used internally by the PMC driver and so fix this
by making it static.
Signed-off-by: Jon Hunter <jonathanh@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
Some members of the tegra_pmc structure are missing from the kernel-doc
comment for this structure. Add the missing members.
Signed-off-by: Jon Hunter <jonathanh@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
The driver requests the pclk clock at probe time already and stores its
reference to it in struct tegra_pmc, so there is no need to look it up
everytime it is needed. Use the existing reference instead.
Signed-off-by: Thierry Reding <treding@nvidia.com>
Tegra210 uses a power management controller that is compatible with
earlier SoC generations but adds a couple of power partitions for new
hardware blocks.
Reviewed-by: Paul Walmsley <paul@pwsan.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
For backwards-compatibility with old device trees, if no PMC node exists
this driver hard-codes the I/O memory region. All 64-bit ARM device tree
files are recent enough that they can be required to have this node, and
therefore the legacy code path is not required on 64-bit ARM.
Based on work done by Paul Walmsley <pwalmsley@nvidia.com>.
Cc: Paul Walmsley <pwalmsley@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
Make sure to only drop the reference to the OF node after it's been
successfully obtained.
Fixes: 3568df3d31 ("soc: tegra: Add thermal reset (thermtrip) support to PMC")
Cc: <stable@vger.kernel.org> # v4.0+
Reviewed-by: Mikko Perttunen <mperttunen@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1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=WLBw
-----END PGP SIGNATURE-----
Merge tag 'module-builtin_driver-v4.1-rc8' of git://git.kernel.org/pub/scm/linux/kernel/git/paulg/linux
Pull module_platform_driver replacement from Paul Gortmaker:
"Replace module_platform_driver with builtin_platform driver in non
modules.
We see an increasing number of non-modular drivers using
modular_driver() type register functions. There are several downsides
to letting this continue unchecked:
- The code can appear modular to a reader of the code, and they won't
know if the code really is modular without checking the Makefile
and Kconfig to see if compilation is governed by a bool or
tristate.
- Coders of drivers may be tempted to code up an __exit function that
is never used, just in order to satisfy the required three args of
the modular registration function.
- Non-modular code ends up including the <module.h> which increases
CPP overhead that they don't need.
- It hinders us from performing better separation of the module init
code and the generic init code.
So here we introduce similar macros for builtin drivers. Then we
convert builtin drivers (controlled by a bool Kconfig) by making the
following type of mapping:
module_platform_driver() ---> builtin_platform_driver()
module_platform_driver_probe() ---> builtin_platform_driver_probe().
The set of drivers that are converted here are just the ones that
showed up as relying on an implicit include of <module.h> during a
pending header cleanup. So we convert them here vs adding an include
of <module.h> to non-modular code to avoid compile fails. Additonal
conversions can be done asynchronously at any time.
Once again, an unused module_exit function that is removed here
appears in the diffstat as an outlier wrt all the other changes"
* tag 'module-builtin_driver-v4.1-rc8' of git://git.kernel.org/pub/scm/linux/kernel/git/paulg/linux:
drivers/clk: convert sunxi/clk-mod0.c to use builtin_platform_driver
drivers/power: Convert non-modular syscon-reboot to use builtin_platform_driver
drivers/soc: Convert non-modular soc-realview to use builtin_platform_driver
drivers/soc: Convert non-modular tegra/pmc to use builtin_platform_driver
drivers/cpufreq: Convert non-modular s5pv210-cpufreq.c to use builtin_platform_driver
drivers/cpuidle: Convert non-modular drivers to use builtin_platform_driver
drivers/platform: Convert non-modular pdev_bus to use builtin_platform_driver
platform_device: better support builtin boilerplate avoidance
This file depends on Kconfig ARCH_TEGRA which is a bool, so
we use the appropriate registration function, which avoids us
relying on an implicit inclusion of <module.h> which we are
doing currently.
While this currently works, we really don't want to be including
the module.h header in non-modular code, which we'd be forced
to do, pending some upcoming code relocation from init.h into
module.h. So we fix it now by using the non-modular equivalent.
Cc: Stephen Warren <swarren@wwwdotorg.org>
Cc: Thierry Reding <thierry.reding@gmail.com>
Cc: Alexandre Courbot <gnurou@gmail.com>
Cc: linux-tegra@vger.kernel.org
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
The pmc driver was previously exporting tegra_pmc_restart, which was
assigned to machine_desc.init_machine, taking precedence over the
restart handlers registered through register_restart_handler().
Signed-off-by: David Riley <davidriley@chromium.org>
[tomeu.vizoso@collabora.com: Rebased]
Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Acked-by: Stephen Warren <swarren@nvidia.com>
Reviewed-by: Alexandre Courbot <acourbot@nvidia.com>
[treding@nvidia.com: minor cleanups]
Signed-off-by: Thierry Reding <treding@nvidia.com>
This adds a device tree controlled option to enable PMC-based
thermal reset in overheating situations. Thermtrip is supported on
Tegra30, Tegra114 and Tegra124. The thermal reset only works when
the thermal sensors are calibrated, so a soctherm driver is also
required.
The thermtrip event is triggered by the soctherm block, and all
soctherm sensors default to showing a temperature of zero Celsius
before they are initialized. Because of this, it is safe to initialize
thermtrip and soctherm in any order.
Signed-off-by: Mikko Perttunen <mperttunen@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
Tegra132 uses the same GPU as Tegra124 and therefore requires the same
method to remove clamps. However Tegra132 has a separate chip ID, so in
order to avoid having to extend the list of chip IDs for the special
case, add a feature flag to the SoC data.
Reviewed-by: Paul Walmsley <pwalmsley@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
Tegra SoCs with 64-bit ARM support don't currently support deep CPU
low-power states in mainline Linux. When this support is added in the
future, it will probably look rather different from the existing
32-bit ARM support, since the ARM64 maintainers' strong preference is
to use PSCI to implement it.
So, for the time being, prevent the CPU suspend-related code and data
in the Tegra PMC driver from compiling on ARM64.
Signed-off-by: Paul Walmsley <paul@pwsan.com>
Signed-off-by: Paul Walmsley <pwalmsley@nvidia.com>
Cc: Thierry Reding <treding@nvidia.com>
Cc: Allen Martin <amartin@nvidia.com>
Cc: Stephen Warren <swarren@nvidia.com>
Cc: Alexandre Courbot <gnurou@gmail.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
This commit converts the PMC support code to a platform driver. Because
the boot process needs to call into this driver very early, also set up
a minimal environment via an early initcall.
Signed-off-by: Thierry Reding <treding@nvidia.com>
Jon Hunter