This function doesn't return and its callers that don't return either
rely on this. Drop the dead attribute and add a panic() after it to make
this expectation explicit. Calling `wfi` in the powerdown sequence is
terminal so even if the function was made to return, there would be no
functional change.
This is useful for a following patch that makes psci_power_down_wfi()
return.
Change-Id: I62ca1ee058b1eaeb046966c795081e01bf45a2eb
Signed-off-by: Boyan Karatotev <boyan.karatotev@arm.com>
Commit e2b18771fc ("feat(allwinner):
provide CPU idle states to the rich OS") added functionality to amend
dtb, when SCPI as the PSCI backend is available. But this
functionality is disabled by default even for platforms, that support
it, like A64. As a result rich OS don't get information about
available CPU idle states.
Due to size constraints of A64 platform DEBUG=1 can be built with
dtb amend functionality only with LTO enabled. So ENABLE_LTO is
enabled by default for this platform.
```
aarch64-linux-gnu-ld.bfd: address 0x500dd of build/sun50i_a64/debug/bl31/bl31.elf section `.data' is not within region `RAM'
aarch64-linux-gnu-ld.bfd: BL31 image has exceeded its limit.
aarch64-linux-gnu-ld.bfd: region `RAM' overflowed by 224 bytes
```
To build with ENABLE_LTO=0 and DEBUG=1 it's required SUNXI_AMEND_DTB=0
to explicitly disable dtb amend functionality.
sun50i_r329 SUNXI_AMEND_DTB=1 DEBUG=1 build fails with 'region `RAM'
overflowed by 120 bytes'. To avoid unnecessary RAM consumption on
other resource-constraints platforms (like sun50i_r329)
SUNXI_AMEND_DTB is enabled only on sun50i_a64. Otherwise
On other platforms sunxi_idle_states are empty.
Signed-off-by: Andrey Skvortsov <andrej.skvortzov@gmail.com>
Change-Id: I81fcf31b5bd2bd02a9f3361a6a519632f087445d
In sunxi_soc_fdt_fixup(), we check for the value of "ret" again, after
calling fdt_node_offset_by_phandle(), even though the error value of
that lands in "node".
Check for "node" being non-negative instead, to properly detect any
errors here.
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Change-Id: I57c1406388dbe11d343038da173019519e18af3e
The Allwinner H616 and its siblings come in different die revisions,
some have 256 KB of L2 cache, some have 1 MB. This prevents a single
static cache description in the devicetree.
Use the cache size ID register (CCSIDR_EL1) to query the topology of the
L2 cache, and adjust the cache-sets and cache-size properties in the L2
cache DT node accordingly.
The ARM ARM does not promise (anymore) that the cache size can be derived
*architecturally* from this register, but the reading is definitely
correct for the Arm Cortex-A53 core used.
Change-Id: Id7dc324d783b8319fe5df6164be2f941d4cac82d
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
On at least one new device we see an Allwinner H700 SoC paired with the
X-Powers AXP717 PMIC. In contrast to the small AXP313, this is a quite
complete PMIC, with many voltage rails, battery and USB-C charging
support. It supports both RSB and I2C control options.
Add the compatible string to the list of checked devices. The AXP717
apparently does not feature a version ID register, but we read 0xff from
that address 0x3, so use this as an indication of its presence, since
this value differs from what we read from the other PMICs.
The register offset and bit position for the power off functionality is
again different, but easy to put into our switch/case.
Setting up regulators in TF-A is now somewhat obsolete, since U-Boot
does a much better job in this now, and can figure out which regulators
are actually needed. So we don't add the regulator setup code, and just
use the PMIC for the power-off functionality.
Change-Id: Ie6b4c91517014adcc79d9a3459c75545fa3a63e6
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Many newer boards with the H616 and its sibling H618 are now paired
with the X-Powers AXP313 PMIC. This is a simpler PMIC, with only a few
voltage rails and no extra functionality except the power key support.
In contrast to the AXP305 it can only be controlled via I2C.
Add a check to look for the AXP313 compatible string in the devicetree,
and set the PMIC type and I2C address accordingly, if one is found.
With only very few voltage rails available, all of them are mostly in
use and are thus enabled at reset already, so we can skip the regulator
setup entirely.
Change-Id: I01962854109e43793b4f56553c1ca9e1f752e30d
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
The X-Powers AXP305 PMIC can be controlled via both I2C or RSB (an
Allwinner specific bus similar to I2C), but we chose to use only RSB,
because that's easier to program and also used by Linux. The AXP313a
PMIC however supports only I2C, so we need to support both buses, and
need to decide which to use at runtime.
Prepare the PMIC code to add (back) I2C support. We initially used I2C
on the H6/AXP805 combination, but replaced that later with RSB. So this
patch is bringing some of that older code back.
The decision whether to use I2C or RSB is made by the devicetree, since
on some boards even RSB capable PMICs are controlled via I2C, since they
share the bus with only I2C capable devices, for instance RTCs.
At the moment this will still use RSB to drive the AXP305, but the
(dynamic) I2C code will be used shortly to support the AXP313.
This increases the code size by one 4K page, but with 80K out of the
reserved 256K we are still very far away from our limit.
Change-Id: I65c1e7df93dbd2dcd171b3fc486533a2948cc75b
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Most devices based on Allwinner SoCs come with a certain fixed
combination of Power Management ICs (PMICs) and SoC, for instance the
A64 with the AXP803, or the H6 with the AXP805. This allowed us to
include the respective PMIC support code into each build target at build
time.
Similarly on H616 devices we initially saw only the AXP305, but for a
while now the simpler (and cheaper) AXP313a is a popular companion to
the H616 on many new boards. On at least one new device the AXP717 is
used as well.
With some rudimentary AXP version check in place we at least detected
the case of an unsupported SoC, but threw an error message, and lost
support for powering off the device.
Refactor the existing PMIC code to be able to support more than one
PMIC model, detected at runtime. For this we use a variable for the RSB
runtime address instead of hardcoding the address used on the AXP305,
and read the hardware bus address from the devicetree.
Also we look up the used PMIC in the devicetree, and set the PMIC model
accordingly. To be on the safe side, we also confirm the real PMIC used
by checking its version register and comparing that with the expected
value. Finally the register offset and value to power off the PMIC is
moved direclty into the platform code, as those values differ between
the different PMICs.
This is just refactoring and better error report, we still only support
the AXP305 on RSB at the moment.
Change-Id: I00b26ce4d30bb570ee1cd4979d0cdc9d6c020729
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Nothing in sunxi_bl31_setup.c uses any functionality provided by the
fdt_wrappers file, so remove its inclusion from the header list.
Change-Id: I47031a58add2f85e757e75d8578f4e8e21ef65ea
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Refactor console_flush() and console_switch_state(CONSOLE_FLAG_RUNTIME)
to bl31_main(). This has been done per the recommendation in TF-A
mailing list. These calls need to be the last calls, after any runtime
initialization has been done, before BL31 exits.
All platforms that override the generic implementation of
bl31_plat_runtime_setup() have been refactored. The console_flush()
and console_switch_state() calls have been removed as they become
part of bl31_main() function.
Any platform that don't need to make any change to the generic (weak)
implementation of bl31_plat_runtime_setup() don't need to override it
in their platforms.
Change-Id: I6d04d6daa9353daeaa7e3df9e9adf6f322a917b8
Signed-off-by: Salman Nabi <salman.nabi@arm.com>
Flush the FIFO before switching to runtime. This is so that there are
no lingering chars in the FIFO when we move to the runtime console.
TF-A plans to refactor the console_Switch_state(CONSOLE_FLAG_RUNTIME)
and console_flush() calls and make them the last calls in bl31_main()
(before BL31 exits). Until then they are being left as the last calls
in bl31_plat_runtime_setup() for testing before refactoring.
This patch only affects the Allwinner platform.
Change-Id: I15b4a459a280822a01c60e3b0c856b530db6efab
Signed-off-by: Salman Nabi <salman.nabi@arm.com>
On systems with SCP (running crust) scpi_system_reboot action
performs board-level (PMIC) reboot. This doesn't preserve RAM content
on A64 PinePhone at least.
warm/soft system reset without RAM reset is required to get
pstore (persistent storage) in RAM working with Linux kernel. That is
very useful for oops/panic logging for post mortem analysis.
scpi_system_reset action performs reset via SoC reset (using watchdog)
and RAM content is preserved in this case. Linux kernel detects
system_reset2 support and uses it for warm reset automatically.
Change-Id: I1c21aa8f27c8e0395e2326034788693b59b80bc4
Signed-off-by: Andrey Skvortsov <andrej.skvortzov@gmail.com>
The Allwinner T507 SoC is using the same die as the H616, but in a
different package. On top of this, there is at least one different die
revision out there, which uses a different CPU cluster control block.
The same die revision has been spotted in some, but not all, H313 SoCs.
Apart from that IP block, the rest of the SoC seems the same, so we can
support them using the existing H616 port. The die revision can be
auto-detected, so there is no extra build option or knowledge needed.
Provide the deviating CPU power up/down sequence for the die variant.
The new IP block uses per-core instead of per-cluster registers, but
follows the same pattern otherwise.
Since the CPU ops code is shared among all Allwinner SoCs, we need to
dummy-define the new register names for the older SoCs. The actual new
code is guarded by a predicate function, that is hard coded to return
true on the other SoCs. Since this is a static inline function in a
header file, the compiler will optimise away the unneeded branch there,
so the generated code for the other SoCs stays the same.
Change-Id: Ib5ade99d34b4ccb161ccde0e34f280ca6bd16ecd
Signed-off-by: Mikhail Kalashnikov <iuncuim@gmail.com>
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Allwinner provides a number of SoCs that use the same die as the H616.
Some of those chips apparently use a slight variation of that die, that
differs in the way the CPU cores' power and reset controls are handled.
This die variation can be detected by reading the SRAM version register.
Provide a predicate function that returns false if that die variant is
used. Since the CPU power control code is shared for all supported SoCs,
we provide an instance of this function for each SoC, as a static
inline, and return true on all other SoCs. This allows to always use
this function, and still let the compiler optimise away the unneeded
branch for those older SoCs.
This function is unused for now, but is needed in the next patch.
Change-Id: I49e014b895b7e2f55b4e7dc2b3d8aa31cee711b5
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
The die used in several variants of the Allwinner H616 SoC (H313, T507)
seems to produced in at least two revisions. The newer one differs from
the original by using a different CPU control register IP block.
Add those newly used register offsets to the respective header file. The
MMIO block itself is actually present in both variants, though the
registers are different. The new registers tend to use one register per
core, in contrast to one register per cluster in the older revisions.
Change-Id: Ifbda1bdc67a6a16fbb901dbc83996e4a148b7602
Signed-off-by: Mikhail Kalashnikov <iuncuim@gmail.com>
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
The header files describing the CPU cluster configuration IP block for
the H6 and H616 are actually identical, so merge them into one file and
move that to a common location. There is an upcoming SoC which will
similarly share a header file with the R329 SoC, so move that to the
same location already. In Allwinner's BSP source those two SoC groups
are typically called "NCAT" and "NCAT2", so use those names for the
shared header files. No functional change.
Change-Id: I98318373577344dbe228a81fa331ce660df32b5f
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
At the moment we hardcode the SPE functionality to be available on the
non-secure side only, by setting MDCR_EL2.E2PB accordingly.
This should be reflected in the feature selection symbol, so rename that
to ENABLE_SPE_FOR_NS, to make it clearer that SPE is not supported in
the secure world.
Change-Id: I3f9b48eab1a45d6ccfcbb9c90a11eeb66867ad9a
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
At the moment we access the RSB bus on all Allwinner H6 boards
unconditionally, even though some boards do not have any PMIC at all,
while others have some I2C devices connected to the same pins.
The latter case is just fragile, but the first case leads to a hang on
at least one board, as reported by Jernej.
Scan the devicetree, to check for the availability of the RSB bus node.
Proceed only if the RSB DT node is actually enabled.
Change-Id: Iea7dcfe3e085e173334d098ec4ddcb6c4b085771
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Reported-by: Jernej Skrabec <jernej.skrabec@gmail.com>
The Allwinner SID device holds a 16-bit SoC identifier, which we already
use in our code.
Export this number through the generic SMCCC SOCID interface, to allow
an architectural identification of an Allwinner SoC. This enables access
to this information from non-secure world, simplifies generic drivers
(ACPI comes to mind), and gives easy and precise access to the SoC ID
from userland in OSes like Linux.
Change-Id: I91753046b2ae5408ca7bc0b864fcd97d24c8267c
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
When using SCPI as the PSCI backend, firmware can wake up the CPUs and
cluster from sleep, so CPU idle states are available for the rich OS to
use. In that case, advertise them to the rich OS via the DTB.
Change-Id: I718ef6ef41212fe5213b11b4799613adbbe6e0eb
Signed-off-by: Samuel Holland <samuel@sholland.org>
Use the encoding recommended by the PSCI specification: four bits for
the power state at each power level.
SCPI provides no way to handshake an exit from a standby state, so the
only possible standby state is the architectural WFI state. Since WFI
can be used outside of PSCI, we do not allow passing in standby states.
Change-Id: I4b3b84e5c255ee58a25255a0cab5d7623425086e
Signed-off-by: Samuel Holland <samuel@sholland.org>
Aligning the PSCI and SCPI power states avoids some code to translate
between the two. This also makes room for an intermediate power state,
for future firmware capability growth.
Change-Id: I26691085f277a96bd405e3305ab0fe390a92b418
Signed-off-by: Samuel Holland <samuel@sholland.org>
Currently, if any step of the DTB patching process fails, the whole
process is aborted. However, this causes some problems:
- If any step modifies the DTB (including fdt_open_into), the dcache
must still be cleaned, even if some later step fails.
- The DTB may need changes in multiple places; if one patch fails (for
example due to missing nodes), we should still apply other patches.
- Similarly, if some patch fails, we should still run fdt_pack to
clean up after ourselves.
Change-Id: If1af2e58e5a7edaf542354bb8a261dd1c3da1ad0
Signed-off-by: Samuel Holland <samuel@sholland.org>
Idle states are advertised to the rich OS by declaring them in the DTB.
Since the availability of idle states depends on which PSCI
implementation was chosen, the DTB must be updated after PSCI setup.
Move this operation to bl31_plat_runtime_setup, the platform hook
which happens at the right time. Defining this hook overrides the weak
definition from plat/common, so copy over the code from there, too.
Change-Id: I42a83edb9cb28e1803d17dc2d73dbc879d885222
Signed-off-by: Samuel Holland <samuel@sholland.org>
So far the H616 was the only Allwinner SoC needed to amend the DTB, to
reserve the DRAM portion that BL31 occupies.
To allow other SoCs to modify the DTB as well, without duplicating code,
move the DTB change routines into Allwinner common code, and generalise
the current code to allow other modifications.
No functional change intended.
Change-Id: I080ea07b6470367f3c2573a4368f8ef5196d411c
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Signed-off-by: Samuel Holland <samuel@sholland.org>
For somewhat historical reasons we are doing some initial PMIC regulator
setup in BL31, as U-Boot does not (yet) have a PMIC driver. This worked
fine so far, but there is at least one board (OrangePi 3) that gets upset,
because the Ethernet PHY needs some *coordinated* bringup of *two*
regulators.
To avoid custom hacks, let's introduce a build option to keep doing the
regulator setup in TF-A. Defining SUNXI_SETUP_REGULATORS to 0 will break
support for some devices on some boards in U-Boot (Ethernet and HDMI),
but will allow to bring up the OrangePi 3 in Linux correctly. We keep
the default at 1 to not change the behaviour for all other boards.
After U-Boot gained proper PMIC support at some point in the future, we
will probably change the default to 0, to get rid of the less optimal
PMIC code in TF-A.
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Change-Id: Ie8e2583d0396f6eeaae8ffe6b6190f27db63e2a7
Allwinner R329 is a new dual-core Corte-A53 SoC. Add basical TF-A
support for it, to provide a PSCI implementation containing CPU
boot/shutdown and SoC reset.
Change-Id: I0fa37ee9b4a8e0e1137bf7cf7d614b6ca9624bfe
Signed-off-by: Icenowy Zheng <icenowy@sipeed.com>
Not all Allwinner SoCs have the same arrangement to SRAM A2.
Allow to specify a offset at which BL31 will stay in SRAM A2.
Change-Id: I574140ffd704a796fae0a5c2d0976e85c7fcbdf9
Signed-off-by: Icenowy Zheng <icenowy@sipeed.com>
In newer Allwiner SoCs, the AA64nAA32 wires are mapped to a new register
called "General Control Register0" in the manual rather than the
"Cluster 0 Control Register0" in older SoCs.
Now the position of AA64nAA32 (reg and bit offset) is defined in a few
macros instead assumed to be at bit offset 24 of
SUNXI_CPUCFG_CLS_CTRL_REG0.
Change-Id: I933d00b9a914bf7103e3a9dadbc6d7be1a409668
Signed-off-by: Icenowy Zheng <icenowy@sipeed.com>
Adds a 1us delay after enabling power to a CPU core, to prevent
inrush-caused CPU crash before it's up.
Change-Id: I8f4c1b0dc0d1d976b31ddc30efe7a77a1619b1b3
Signed-off-by: Icenowy Zheng <icenowy@sipeed.com>
Group the SCP base/size definitions in a more logical location.
Signed-off-by: Samuel Holland <samuel@sholland.org>
Change-Id: Id43f9b468d7d855a2413173d674a5ee666527808
BL31 does not appear to ever access the DRAM allocated to BL32,
so there is no need to map it at EL3.
Signed-off-by: Samuel Holland <samuel@sholland.org>
Change-Id: Ie8727b793e53ea14517894942266f6da0333eb74
The SRAM on Allwinner platforms is shared between BL31 and coprocessor
firmware. Previously, SRAM was mapped as normal memory by default.
This scheme requires carveouts and cache maintenance code for proper
synchronization with the coprocessor.
A better scheme is to only map pages owned by BL31 as normal memory,
and leave everything else as device memory. This removes the need for
cache maintenance, and it makes the mapping for BL31 RW data explicit
instead of magic.
Signed-off-by: Samuel Holland <samuel@sholland.org>
Change-Id: I820ddeba2dfa2396361c2322308c0db51b55c348
This constant specifically refers to the number of static mmap regions.
Rename it to make that clear.
Signed-off-by: Samuel Holland <samuel@sholland.org>
Change-Id: I475c037777ce2a10db2631ec0e7446bb73590a36
When the BL31 for the Allwinner H616 runs in DRAM, we need to make sure
we tell the non-secure world about the memory region it uses.
Add a reserved-memory node to the DT, which covers the area that BL31
could occupy. The "no-map" property will prevent OSes from mapping
the area, so there would be no speculative accesses.
Change-Id: I808f3e1a8089da53bbe4fc6435a808e9159831e1
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
The new Allwinner H616 SoC lacks the management controller and the secure
SRAM A2, so we need to tweak the memory map quite substantially:
We run BL31 in DRAM. Since the DRAM starts at 1GB, we cannot use our
compressed virtual address space (max 256MB) anymore, so we revert to
the full 32bit VA space and use a flat mapping throughout all of it.
The missing controller also means we need to always use the native PSCI
ops, using the CPUIDLE hardware, as SCPI and suspend depend on the ARISC.
Change-Id: I77169b452cb7f5dc2ef734f3fc6e5d931749141d
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
In preparation for changing the memory map, express the locations of the
various code and data pieces more dynamically, allowing SoCs to override
the memmap later.
Also prepare for the SCP region to become optional.
No functional change.
Change-Id: I7ac01e309be2f23bde2ac2050d8d5b5e3d6efea2
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
The code to power the current core off when SCPI is not available is now
different for the two supported SoC families.
To make adding new platforms easier, move sunxi_cpu_power_off_self()
into the SoC directory, so we don't need to carry definitions for both
methods for all SoCs.
On the H6 we just need to trigger the CPUIDLE hardware, so can get rid
of all the code to program the ARISC, which is now only needed for the
A64 version.
Change-Id: Id2a1ac7dcb375e2fd021b441575ce86b4d7edf2c
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
For the existing SoCs we support, we use SEPARATE_NOBITS_REGION, to move
some parts of the data into separate memory regions (to save on the SRAM
A2 we are loaded into).
For the upcoming H616 platform this is of no concern (we run in DRAM),
so make this flag a platform choice instead.
Change-Id: Ic01d49578c6274660f8f112bd23680d3eca3be7a
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
The upcoming refactoring to support the new H616 SoCs will push the A64
build over the edge, by using more than the 48KB of SRAM available.
To reduce the code size, set some libfdt options that aim to reduce
sanity checks (for saving code space):
- ASSUME_LATEST: only allow v17 DTBs (as created by dtc)
- ASSUME_NO_ROLLBACK: don't prepare for failed DT additions
- ASSUME_LIBFDT_ORDER: assume sane ordering, as done by dtc
Change-Id: I12c93ec09e7587c5ae71e54947f817c32ce5fd6d
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
This works even on SoCs that do not have an ARISC, and it avoids
clobbering whatever ARISC firmware might be running.
Change-Id: I9f2fed597189bb387de79e8e76a7da3375e1ee91
Signed-off-by: Samuel Holland <samuel@sholland.org>
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Now that we have split the native and the SCPI version of the PSCI ops,
we can introduce build options to compile in either or both of them.
If one version is not compiled in, some stub functions make sure the
common code still compiles and makes the right decisions.
By default both version are enabled (as before), but one of them can be
disabled on the make command line, or via a platform specific Makefile.
Change-Id: I0c019d8700c0208365eacf57809fb8bc608eb9c0
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Signed-off-by: Samuel Holland <samuel@sholland.org>
In order to keep SCP firmware as optional, the original, limited native
PSCI implementation was kept around as a fallback. This turned out to be
a good decision, as some newer SoCs omit the ARISC, and thus cannot run
SCP firmware.
However, keeping the two implementations in one file makes things
unnecessarily messy. First, it is difficult to compile out the
SCPI-based implementation where it is not applicable. Second the check
is done in each callback, while scpi_available is only updated at boot.
This makes the individual callbacks unnecessarily complicated.
It is cleaner to provide two entirely separate implementations in two
separate files. The native implementation does not support any kind of
CPU suspend, so its callbacks are greatly simplified. One function,
sunxi_validate_ns_entrypoint, is shared between the two implementations.
Finally, the logic for choosing between implementations is kept in a
third file, to provide for platforms where only one implementation is
applicable and the other is compiled out.
Change-Id: I4914f07d8e693dbce218e0e2394bef15c42945f8
Signed-off-by: Samuel Holland <samuel@sholland.org>
- When the SCPI shutdown/reset command returns success, the SCP is
still waiting for the CPU to enter WFI. Do that.
- Peform board-level poweroff before CPU poweroff. If there is a PMIC
available, it will turn everything off including the CPUs, so doing
CPU poweroff first is a waste of cycles.
- During poweroff, attempt to turn off the local CPU using the ARISC.
This should use slightly less power than just an infinite WFI.
- Drop the WFI in the reset failure path. The panic will hang anyway.
Change-Id: I897efecb3fe4e77a56041b97dd273156ec51ef8e
Signed-off-by: Samuel Holland <samuel@sholland.org>
When operating on the local cpu, sunxi_cpu_power_off_self() only "arms"
the ARISC to perform the power-off process; the SCP waits for the CPU to
enter WFI before acutally powering it off. Since this matches the
expected split between .pwr_domain_off and .pwr_domain_pwr_down_wfi, we
can move the sunxi_cpu_power_off_self() call to sunxi_pwr_domain_off().
Since that change makes sunxi_pwr_down_wfi() equivalent to the default
implementation, the callback is no longer needed.
Change-Id: I7d65f66c550d1c69fa5e9945affd7a25b3d3ef42
Signed-off-by: Samuel Holland <samuel@sholland.org>
Currently, sunxi_cpu_off() has two separate code paths: one for the
local CPU, and one for other CPUs. Let's split them in to two functions.
This actually simplifies things, because all callers either operate on
the local CPU only (sunxi_pwr_down_wfi()) or other CPUs only
(sunxi_cpu_power_off_others()). This avoids needing a second MPIDR read
to choose the appropriate code path.
Change-Id: I55de85025235cc95466bfa106831fc4c2368f527
Signed-off-by: Samuel Holland <samuel@sholland.org>
Disabling secondary CPUs during boot is unnecessary because the other
CPUs are already in reset, and it saves an entirely insignificant amount
of power. Let's remove this bit of code that was added mostly "because
we can", and along with it remove an unconditional dependency on the CPU
ops functions.
Signed-off-by: Samuel Holland <samuel@sholland.org>
Change-Id: Ia77a1b722da6ba989c3992b656a6cde3f2238fd7
Checking the exceptional case and letting the success case fall through
is not only more idiomatic, but it also allows adding more exceptional
cases in the future, such as a check for overlapping secure DRAM.
Change-Id: I720441a6a8853fd7f211ebe851f14d921a6db03d
Signed-off-by: Samuel Holland <samuel@sholland.org>
