The external memory arbitration configuration is getting reset after
memory entering into self-refresh mode, it shall be restored on the
exit. Note that MC_EMEM_ARB_CFG register is shadowed and latching
happens on the EMC timing update. This fixes 2x GPU performance
degradation after resuming from LP1 on Tegra30.
Signed-off-by: Dmitry Osipenko <digetx@gmail.com>
Acked-by: Jon Hunter <jonathanh@nvidia.com>
Tested-by: Jon Hunter <jonathanh@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
The DRAM refresh-interval is getting erroneously set to "1" on exiting
from memory self-refreshing mode. The clobbered interval causes the
"refresh request overflow timeout" error raised by the External Memory
Controller on exiting from LP1 on Tegra30. The same may happen on Tegra20,
but EMC registers are not latched after exiting from self-refreshing mode
on Tegra20 and hence refresh-interval is not altered until an event that
causes registers latching happens.
Signed-off-by: Dmitry Osipenko <digetx@gmail.com>
Acked-by: Jon Hunter <jonathanh@nvidia.com>
Tested-by: Jon Hunter <jonathanh@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
The memory interface configuration and re-calibration interval are left
unassigned on resume from LP1 because these registers are shadowed and
require latching after being adjusted.
Signed-off-by: Dmitry Osipenko <digetx@gmail.com>
Reviewed-by: Jon Hunter <jonathanh@nvidia.com>
Tested-by: Jon Hunter <jonathanh@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
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>
Tegra support several low-power (LPx) states, which are:
- LP0: CPU + Core voltage off and DRAM in self-refresh
- LP1: CPU voltage off and DRAM in self-refresh
- LP2: CPU voltage off
When entering any of the above states the tegra_disable_clean_inv_dcache()
function is called to flush the dcache. The function
tegra_disable_clean_inv_dcache() will either flush the entire data cache or
up to the Level of Unification Inner Shareable (LoUIS) depending on the
value in r0. When tegra_disable_clean_inv_dcache() is called by
tegra20_sleep_core_finish() or tegra30_sleep_core_finish(), to enter LP0
and LP1 power state, the r0 register contains a physical memory address
which will not be equal to TEGRA_FLUSH_CACHE_ALL (1) and so the data cache
will be only flushed to the LoUIS. However, when
tegra_disable_clean_inv_dcache() called by tegra_sleep_cpu_finish() to
enter to LP2 power state, r0 is set to TEGRA_FLUSH_CACHE_ALL to flush the
entire dcache.
Please note that tegra20_sleep_core_finish(), tegra30_sleep_core_finish()
and tegra_sleep_cpu_finish() are called by the boot CPU once all other CPUs
have been disabled and so it seems appropriate to flush the entire cache at
this stage.
Therefore, ensure that r0 is set to TEGRA_FLUSH_CACHE_ALL when calling
tegra_disable_clean_inv_dcache() from tegra20_sleep_core_finish() and
tegra30_sleep_core_finish().
Signed-off-by: Jon Hunter <jonathanh@nvidia.com>
Reviewed-by: Joseph Lo <josephl@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
This merge window brings a good size of cleanups on various
platforms. Among the bigger ones:
* Removal of Samsung s5pc100 and s5p64xx platforms. Both of these have
lacked active support for quite a while, and after asking around nobody
showed interest in keeping them around. If needed, they could be
resurrected in the future but it's more likely that we would prefer
reintroduction of them as DT and multiplatform-enabled platforms
instead.
* OMAP4 controller code register define diet. They defined a lot of registers
that were never actually used, etc.
* Move of some of the Tegra platform code (PMC, APBIO, fuse, powergate)
to drivers/soc so it can be shared with 64-bit code. This also converts them
over to traditional driver models where possible.
* Removal of legacy gpio-samsung driver, since the last users have been
removed (moved to pinctrl)
Plus a bunch of smaller changes for various platforms that sort of
dissapear in the diffstat for the above. clps711x cleanups, shmobile
header file refactoring/moves for multiplatform friendliness, some misc
cleanups, etc.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.14 (GNU/Linux)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=gbcE
-----END PGP SIGNATURE-----
Merge tag 'cleanup-for-3.17' of git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc
Pull ARM SoC cleanups from Olof Johansson:
"This merge window brings a good size of cleanups on various platforms.
Among the bigger ones:
- Removal of Samsung s5pc100 and s5p64xx platforms. Both of these
have lacked active support for quite a while, and after asking
around nobody showed interest in keeping them around. If needed,
they could be resurrected in the future but it's more likely that
we would prefer reintroduction of them as DT and
multiplatform-enabled platforms instead.
- OMAP4 controller code register define diet. They defined a lot of
registers that were never actually used, etc.
- Move of some of the Tegra platform code (PMC, APBIO, fuse,
powergate) to drivers/soc so it can be shared with 64-bit code.
This also converts them over to traditional driver models where
possible.
- Removal of legacy gpio-samsung driver, since the last users have
been removed (moved to pinctrl)
Plus a bunch of smaller changes for various platforms that sort of
dissapear in the diffstat for the above. clps711x cleanups, shmobile
header file refactoring/moves for multiplatform friendliness, some
misc cleanups, etc"
* tag 'cleanup-for-3.17' of git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc: (117 commits)
drivers: CCI: Correct use of ! and &
video: clcd-versatile: Depend on ARM
video: fix up versatile CLCD helper move
MAINTAINERS: Add sdhci-st file to ARCH/STI architecture
ARM: EXYNOS: Fix build breakge with PM_SLEEP=n
MAINTAINERS: Remove Kirkwood
ARM: tegra: Convert PMC to a driver
soc/tegra: fuse: Set up in early initcall
ARM: tegra: Always lock the CPU reset vector
ARM: tegra: Setup CPU hotplug in a pure initcall
soc/tegra: Implement runtime check for Tegra SoCs
soc/tegra: fuse: fix dummy functions
soc/tegra: fuse: move APB DMA into Tegra20 fuse driver
soc/tegra: Add efuse and apbmisc bindings
soc/tegra: Add efuse driver for Tegra
ARM: tegra: move fuse exports to soc/tegra/fuse.h
ARM: tegra: export apb dma readl/writel
ARM: tegra: Use a function to get the chip ID
ARM: tegra: Sort includes alphabetically
ARM: tegra: Move includes to include/soc/tegra
...
ARMv6 and greater introduced a new instruction ("bx") which can be used
to return from function calls. Recent CPUs perform better when the
"bx lr" instruction is used rather than the "mov pc, lr" instruction,
and this sequence is strongly recommended to be used by the ARM
architecture manual (section A.4.1.1).
We provide a new macro "ret" with all its variants for the condition
code which will resolve to the appropriate instruction.
Rather than doing this piecemeal, and miss some instances, change all
the "mov pc" instances to use the new macro, with the exception of
the "movs" instruction and the kprobes code. This allows us to detect
the "mov pc, lr" case and fix it up - and also gives us the possibility
of deploying this for other registers depending on the CPU selection.
Reported-by: Will Deacon <will.deacon@arm.com>
Tested-by: Stephen Warren <swarren@nvidia.com> # Tegra Jetson TK1
Tested-by: Robert Jarzmik <robert.jarzmik@free.fr> # mioa701_bootresume.S
Tested-by: Andrew Lunn <andrew@lunn.ch> # Kirkwood
Tested-by: Shawn Guo <shawn.guo@freescale.com>
Tested-by: Tony Lindgren <tony@atomide.com> # OMAPs
Tested-by: Gregory CLEMENT <gregory.clement@free-electrons.com> # Armada XP, 375, 385
Acked-by: Sekhar Nori <nsekhar@ti.com> # DaVinci
Acked-by: Christoffer Dall <christoffer.dall@linaro.org> # kvm/hyp
Acked-by: Haojian Zhuang <haojian.zhuang@gmail.com> # PXA3xx
Acked-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com> # Xen
Tested-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de> # ARMv7M
Tested-by: Simon Horman <horms+renesas@verge.net.au> # Shmobile
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
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>
The LP1 suspend procedure is the same with Tegra30 and Tegra114. Just
need to update the difference of the register address, then we can
continue to share the code.
Signed-off-by: Joseph Lo <josephl@nvidia.com>
Signed-off-by: Stephen Warren <swarren@nvidia.com>
This patch re-calculates the LP1 data of tegra30/114_sdram_pad_address
to base on its label not rely on others. This can make easier to
maintain if some other Tegra chips keep re-using these codes in the
future. And change the name of tegra30_sdram_pad_save to
tegra_sdram_pad_save to make it more common to other chips.
Signed-off-by: Joseph Lo <josephl@nvidia.com>
Signed-off-by: Stephen Warren <swarren@nvidia.com>
irammap.h's purpose is to define the layout/usage of IRAM. As such,
TEGRA_IRAM_CODE_AREA should have been added there rather than iomap.h.
Move the define, and rename it something more descriptive.
Cc: Joseph Lo <josephl@nvidia.com>
Signed-off-by: Stephen Warren <swarren@nvidia.com>
The LP1 suspend mode will power off the CPU, clock gated the PLLs and put
SDRAM to self-refresh mode. Any interrupt can wake up device from LP1. The
sequence when LP1 suspending:
* tunning off L1 data cache and the MMU
* storing some EMC registers, DPD (deep power down) status, clk source of
mselect and SCLK burst policy
* putting SDRAM into self-refresh
* switching CPU to CLK_M (12MHz OSC)
* tunning off PLLM, PLLP, PLLA, PLLC and PLLX
* switching SCLK to CLK_S (32KHz OSC)
* shutting off the CPU rail
The sequence of LP1 resuming:
* re-enabling PLLM, PLLP, PLLA, PLLC and PLLX
* restoring the clk source of mselect and SCLK burst policy
* setting up CCLK burst policy to PLLX
* restoring DPD status and some EMC registers
* resuming SDRAM to normal mode
* jumping to the "tegra_resume" from PMC_SCRATCH41
Due to the SDRAM will be put into self-refresh mode, the low level
procedures of LP1 suspending and resuming should be copied to
TEGRA_IRAM_CODE_AREA (TEGRA_IRAM_BASE + SZ_4K) when suspending. Before
restoring the CPU context when resuming, the SDRAM needs to be switched
back to normal mode. And the PLLs need to be re-enabled, SCLK burst policy
be restored. Then jumping to "tegra_resume" that was expected to be stored
in PMC_SCRATCH41 to restore CPU context and back to kernel.
Based on the work by: Bo Yan <byan@nvidia.com>
Signed-off-by: Joseph Lo <josephl@nvidia.com>
Signed-off-by: Stephen Warren <swarren@nvidia.com>
The LP1 suspend mode will power off the CPU, clock gated the PLLs and put
SDRAM to self-refresh mode. Any interrupt can wake up device from LP1. The
sequence when LP1 suspending:
* tunning off L1 data cache and the MMU
* storing some EMC registers, DPD (deep power down) status, clk source of
mselect and SCLK burst policy
* putting SDRAM into self-refresh
* switching CPU to CLK_M (12MHz OSC)
* tunning off PLLM, PLLP, PLLA, PLLC and PLLX
* switching SCLK to CLK_S (32KHz OSC)
* shutting off the CPU rail
The sequence of LP1 resuming:
* re-enabling PLLM, PLLP, PLLA, PLLC and PLLX
* restoring the clk source of mselect and SCLK burst policy
* setting up CCLK burst policy to PLLX
* restoring DPD status and some EMC registers
* resuming SDRAM to normal mode
* jumping to the "tegra_resume" from PMC_SCRATCH41
Due to the SDRAM will be put into self-refresh mode, the low level
procedures of LP1 suspending and resuming should be copied to
TEGRA_IRAM_CODE_AREA (TEGRA_IRAM_BASE + SZ_4K) when suspending. Before
restoring the CPU context when resuming, the SDRAM needs to be switched
back to normal mode. And the PLLs need to be re-enabled, SCLK burst policy
be restored, CCLK burst policy be set in PLLX. Then jumping to
"tegra_resume" that was expected to be stored in PMC_SCRATCH41 to restore
CPU context and back to kernel.
Based on the work by: Scott Williams <scwilliams@nvidia.com>
Signed-off-by: Joseph Lo <josephl@nvidia.com>
Signed-off-by: Stephen Warren <swarren@nvidia.com>
When the CPU cluster power down, the vGIC is powered down too. The
flow controller needs to monitor the legacy interrupt controller to
wake up CPU. So setting up the appropriate wake up event in flow
controller.
Signed-off-by: Joseph Lo <josephl@nvidia.com>
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Adding a flag for tegra_disable_clean_inv_dcache to flush cache as LoUIS
or ALL. After this patch, the v7_flush_dcache_louis is used for CPU hotplug
and CPU suspend in CPU power down (e.g. CPU idle power-down mode) case. And
the v7_flush_dcache_all is used for CPU cluster power down (e.g. suspend to
LP2 mode).
Signed-off-by: Joseph Lo <josephl@nvidia.com>
Signed-off-by: Stephen Warren <swarren@nvidia.com>
The flow controller would take care the power sequence when CPU idle in
powered-down mode. It powered gate the CPU when CPU runs into WFI
instruction. And wake up the CPU when event be triggered.
The sequence is below.
* setting wfi bitmap for the CPU as the halt event in the
FLOW_CTRL_CPU_HALT_REG to monitor the CPU running into WFI,then power
gate it
* setting IRQ and FIQ as wake up event to wake up CPU when event triggered
Signed-off-by: Joseph Lo <josephl@nvidia.com>
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>
This patch fix the build failure when CONFIG_THUBM2_KERNEL enabled. You
clould see the error message below:
arch/arm/mach-tegra/sleep-tegra30.S:69: Error: shift must be constant --
`orr r12,r12,r4,lsl r3'
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>
Updating the cache maintenance order before CPU shutdown when doing CPU
hotplug.
The old order:
* clean L1 by flush_cache_all
* exit SMP
* CPU shutdown
Adapt to:
* disable L1 data cache by clear C bit
* clean L1 by v7_flush_dcache_louis
* exit SMP
* CPU shutdown
For CPU hotplug case, it's no need to do "flush_cache_all". And we should
disable L1 data cache before clean L1 data cache. Then leaving the SMP
coherency.
Signed-off-by: Joseph Lo <josephl@nvidia.com>
Acked-by: Peter De Schrijver <pdeschrijver@nvidia.com>
Signed-off-by: Stephen Warren <swarren@nvidia.com>
This is a power gating idle mode. It support power gating vdd_cpu rail
after all cpu cores in "powered-down" status. For Tegra30, the CPU0 can
enter this state only when all secondary CPU is offline. We need to take
care and make sure whole secondary CPUs were offline and checking the
CPU power gate status. After that, the CPU0 can go into "powered-down"
state safely. Then shut off the CPU rail.
Be aware of that, you may see the legacy power state "LP2" in the code
which is exactly the same meaning of "CPU power down".
Base on the work by:
Scott Williams <scwilliams@nvidia.com>
Signed-off-by: Joseph Lo <josephl@nvidia.com>
Signed-off-by: Stephen Warren <swarren@nvidia.com>
This supports power-gated idle on secondary CPUs for Tegra30. The
secondary CPUs can go into powered-down state independently. When
CPU goes into this state, it saves it's contexts and puts itself
to flow controlled WFI state. After that, it will been power gated.
Be aware of that, you may see the legacy power state "LP2" in the
code which is exactly the same meaning of "CPU power down".
Based on the work by:
Scott Williams <scwilliams@nvidia.com>
Signed-off-by: Joseph Lo <josephl@nvidia.com>
Signed-off-by: Stephen Warren <swarren@nvidia.com>
For the naming consistency under the mach-tegra, we re-name the file of
"sleep-tXX" to "sleep-tegraXX" (e.g., sleep-t30 to sleep-tegra30).
Signed-off-by: Joseph Lo <josephl@nvidia.com>
Signed-off-by: Stephen Warren <swarren@nvidia.com>
