The flowctrl driver is required for both ARM and ARM64 Tegra devices
and in order to enable support for it for ARM64, move the Tegra flowctrl
driver into drivers/soc/tegra.
By moving the flowctrl driver, tegra_flowctrl_init() is now called by
via an early initcall and to prevent this function from attempting to
mapping IO space for a non-Tegra device, a test for 'soc_is_tegra()'
is also added.
Signed-off-by: Jon Hunter <jonathanh@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
Commit 7232398abc ("ARM: tegra: Convert PMC to a driver") changed tegra_resume()
location storing from late to early and, as a result, broke suspend on Tegra20.
PMC scratch register 41 is used by tegra LP1 resume code for retrieving stored
physical memory address of common resume function and in the same time used by
tegra20_cpu_shutdown() (shared by Tegra20 cpuidle driver and platform SMP code),
which is storing CPU1 "resettable" status. It implies strict order of scratch
register usage, otherwise resume function address is lost on Tegra20 after
disabling non-boot CPU's on suspend. Fix it by storing "resettable" status in
IRAM instead of PMC scratch register.
Signed-off-by: Dmitry Osipenko <digetx@gmail.com>
Fixes: 7232398abc (ARM: tegra: Convert PMC to a driver)
Cc: <stable@vger.kernel.org> # v3.17+
Signed-off-by: Thierry Reding <treding@nvidia.com>
Commit d127e9c ("ARM: tegra: make tegra_resume can work with current and later
chips") removed tegra_get_soc_id macro leaving used cpu register corrupted after
branching to v7_invalidate_l1() and as result causing execution of unintended
code on tegra20. Possibly it was expected that r6 would be SoC id func argument
since common cpu reset handler is setting r6 before branching to tegra_resume(),
but neither tegra20_lp1_reset() nor tegra30_lp1_reset() aren't setting r6
register before jumping to resume function. Fix it by re-adding macro.
Fixes: d127e9c (ARM: tegra: make tegra_resume can work with current and later chips)
Cc: <stable@vger.kernel.org> # v3.13+
Reviewed-by: Felipe Balbi <balbi@ti.com>
Signed-off-by: Dmitry Osipenko <digetx@gmail.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
Instead of using a simple variable access to get at the Tegra chip ID,
use a function so that we can run additional code. This can be used to
determine where the chip ID is being accessed without being available.
That in turn will be handy for resolving boot sequence dependencies in
order to convert more code to regular initcalls rather than a sequence
fixed by Tegra SoC setup code.
Signed-off-by: Thierry Reding <treding@nvidia.com>
If these aren't sorted alphabetically, then the logical choice is to
append new ones, however that creates a lot of potential for conflicts
because every change will then add new includes in the same location.
Signed-off-by: Thierry Reding <treding@nvidia.com>
Acked-by: Stephen Warren <swarren@nvidia.com>
Tested-by: Stephen Warren <swarren@nvidia.com>
Reviewed-by: Joseph Lo <josephl@nvidia.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Because the CPU0 was the first up and the last down core when cluster
power up/down or platform suspend. So only CPU0 needs the rest of the
functions to reset flow controller and re-enable SCU and L2. We also
move the L2 init function for Cortex-A15 to there. The secondery CPU
can just call cpu_resume.
Signed-off-by: Joseph Lo <josephl@nvidia.com>
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Add support to the Tegra CPU reset vector to detect whether the CPU is
resuming from LP1 suspend state. If it is, branch to the LP1-specific
resume code.
When Tegra enters the LP1 suspend state, the SDRAM controller is placed
into a self-refresh state. For this reason, we must place the LP1 resume
code into IRAM, so that it is accessible before SDRAM access has been
re-enabled.
Signed-off-by: Joseph Lo <josephl@nvidia.com>
Signed-off-by: Stephen Warren <swarren@nvidia.com>
When there is a cluster power down cycle in suspend, we need to set up
the correct L2 RAM data RAM latency to make L2 cache work correctly. This
is only needed for cluster 0 and needs to be done in tegra_resume before
the cache is enabled.
Signed-off-by: Joseph Lo <josephl@nvidia.com>
Signed-off-by: Stephen Warren <swarren@nvidia.com>
The v7_invalidate_l1 was used for the L1 cache that come out from reset
in a undefined state. This is no need for Cortex-A15. We do it for A9
only.
Signed-off-by: Joseph Lo <josephl@nvidia.com>
Signed-off-by: Stephen Warren <swarren@nvidia.com>
The ifdef was originally added with the intent that the runtime SoC
detection code, and code to support SoCs other than Tegra20, was only
included if the kernel supported SoCs other than Tegra20. However,
the condition was somewhat backwards and did not achieve this goal.
Simply remove the ifdef to solve this, rather than creating a much more
complex version.
We also fix a typo that caused a build error due to cpu_to_csr_req being
undefined.
Reported-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Joseph Lo <josephl@nvidia.com>
[swarren: rewrote commit description]
Signed-off-by: Stephen Warren <swarren@nvidia.com>
The Tegra114 is a quad cores SoC. Each core can be hotplugged including
CPU0. The hotplug sequence can be controlled by setting event trigger in
flow controller. Then the flow controller will take care all the power
sequence that include CPU up and down.
Signed-off-by: Joseph Lo <josephl@nvidia.com>
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Tegra114 had a newer flow controller hardware that makes its behavior and
configurations are different with other Tegra series. We fix the common
resume function of tegra_resume to make it can work on Tegra114 by checking
SoC ID. And also checking CPU primary part number to isolate the support
code for Cortex A9 and A15.
Signed-off-by: Joseph Lo <josephl@nvidia.com>
Signed-off-by: Stephen Warren <swarren@nvidia.com>
There are some Tegra SoC ID checking code around the low level assembly
code. Adding a marco to replace them. For the single image to support all
the Tegra series, we may also need the marco in other common code. So we
make it become a marco for the usage.
Signed-off-by: Joseph Lo <josephl@nvidia.com>
Signed-off-by: Stephen Warren <swarren@nvidia.com>
The conditional branch instruction in Thumb2 only available to short range.
The linker will fail when the conditional branch over the range. Then
resulting in link error when generating kernel image. e.g.:
arch/arm/mach-tegra/reset-handler.S:47:(.text+0xf8e):
relocation truncated to fit: R_ARM_THM_JUMP19 against symbol
`cpu_resume' defined in .data section in arch/arm/kernel/built-in.o
This patch using a Thumb2 instruction IT (if-then) to have a longer branch
range.
Reported-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Joseph Lo <josephl@nvidia.com>
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Reviewed-by: Dave Martin <dave.martin@linaro.org>
Signed-off-by: Olof Johansson <olof@lixom.net>
There is no need to unlock MMIO access to the DBGLAR all the time. Doing
so may even cause problems if a SW bug causes writes to that MMIO region.
Cortex-A15 processors do not support the CP14 register write the code
currently uses to unlock the DBGLAR; the instruction throws an undefined
instruction exceptions. This prevents tegra_secondary_startup() from
executing on Tegra114, and hence prevents SMP.
Remove the code that unlocks this access.
Signed-off-by: Joseph Lo <josephl@nvidia.com>
Signed-off-by: Stephen Warren <swarren@nvidia.com>
The CPU cores in Tegra contain some errata. Workarounds must be applied
for these every time a CPU boots. Implement those workarounds directly
in the Tegra-specific CPU reset vector.
Many of these workarounds duplicate code in the core ARM kernel.
However, the core ARM kernel cannot enable those workarounds when
building a multi-platform kernel, since they require writing to secure-
only registers, and a multi-platform kernel often does not run in secure
mode, and also cannot generically/architecturally detect whether it is
running in secure mode, and hence cannot either unconditionally or
conditionally apply these workarounds.
Instead, the workarounds must be applied in architecture-specific reset
code, which is able to have more direct knowledge of the secure/normal
state. On Tegra, we will be able to detect this using a non-architected
register in the future, although we currently assume the kernel runs only
in secure mode. Other SoCs may never run the kernel in secure mode, and
hence always rely on a secure monitor to enable the workarounds, and
hence never implement them in the kernel.
Signed-off-by: Stephen Warren <swarren@nvidia.com>
The reset handler code is used for either UP or SMP. To make Tegra device
can compile for UP. It needs to be moved to another file that is not SMP
only. This is because the reset handler also be needed by CPU idle
"powered-down" mode. So we also need to put the reset handler init function
in non-SMP only and init them always.
And currently the implementation of the reset handler to know which CPU is
OK to bring up was identital with "cpu_present_mask". But the
"cpu_present_mask" did not initialize yet when the reset handler init
function was moved to init early function. We use the "cpu_possible_mask"
to replace "cpu_present_mask". Then it can work on both UP and SMP case.
Signed-off-by: Joseph Lo <josephl@nvidia.com>
[swarren: dropped the move of v7_invalidate_l1() from one file to another,
to avoid conflicts with Pavel's cleanup of this function, adjust Makefile
so each line only contains 1 file.]
Signed-off-by: Stephen Warren <swarren@nvidia.com>
