On ARM64, the mtk-pmic-wrap driver causes a harmless warning:
mtk-pmic-wrap.c:1062:16: warning: large integer implicitly truncated to unsigned type [-Woverflow]
mtk-pmic-wrap.c:1074:16: warning: large integer implicitly truncated to unsigned type [-Woverflow]
mtk-pmic-wrap.c:1086:16: warning: large integer implicitly truncated to unsigned type [-Woverflow]
.int_en_all = ~(BIT(31) | BIT(1)),
The problem is that the result of the BIT() macro is an 'unsigned long',
so taking the bitwise NOT operation of that results in an integer
with the upper 32 bits all set and that cannot be assigned to a
'u32' variable without loss of information.
This is harmless because we were never interested in the upper bits
here anyway, so we can shut up the warning by adding a simple cast
to 'u32'.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Add the registers, callbacks and data structures required to make the
wrapper work on MT2701 and MT7623.
Signed-off-by: John Crispin <blogic@openwrt.org>
Signed-off-by: Matthias Brugger <matthias.bgg@gmail.com>
Add support for MT6323 slaves. This PMIC can be found on MT2701 and MT7623
EVB. The only function that we need to touch is pwrap_init_cipher().
Signed-off-by: John Crispin <blogic@openwrt.org>
Signed-off-by: Matthias Brugger <matthias.bgg@gmail.com>
This patch adds a new struct pwrap_slv_type that we use to store the slave
specific data. The patch adds 2 new helper functions to access the dew
registers. The slave type is looked up via the wrappers child node.
Signed-off-by: John Crispin <blogic@openwrt.org>
Signed-off-by: Matthias Brugger <matthias.bgg@gmail.com>
With more SoCs being added the list of helper functions like these would
grow. To mitigate this problem we remove the existing helpers and change
the code to test against the pmic type stored inside the pmic specific
datastructure that our context structure points at. There is one usage of
pwrap_is_mt8135() that is ambiguous as the test should not be dependent on
mt8135, but rather on the existence of a bridge. Add a new element to
pmic_wrapper_type to indicate if a bridge is present and use this where
appropriate.
Signed-off-by: John Crispin <blogic@openwrt.org>
Signed-off-by: Matthias Brugger <matthias.bgg@gmail.com>
Different SoCs will use different bitmask for the wdt_src. This patch
defines the bitmask in the pmic_wrapper_type struct. This allows us to
support new SoCs with a different bitmask to the one currently used.
Signed-off-by: John Crispin <blogic@openwrt.org>
Signed-off-by: Matthias Brugger <matthias.bgg@gmail.com>
Different SoCs will use different bitmask for the SPI_WRITE command. This
patch defines the bitmask in the pmic_wrapper_type struct. This allows us
to support new SoCs with a different bitmask to the one currently used.
Signed-off-by: John Crispin <blogic@openwrt.org>
Signed-off-by: Matthias Brugger <matthias.bgg@gmail.com>
MT2701 and MT7623 use a different bitmask for PWRAP_INT_EN.
Signed-off-by: John Crispin <blogic@openwrt.org>
Signed-off-by: Matthias Brugger <matthias.bgg@gmail.com>
This patch moves the SoC specific wrapper init code into separate callback
to avoid pwrap_init() getting too large. This is done by adding a new
element called init_special to pmic_wrapper_type. Each currently supported
SoC gets its own version of the callback and we copy the code that was
previously inside pwrap_init() to these new callbacks. Finally we point the
2 instances of pmic_wrapper_type at the 2 new functions.
Signed-off-by: John Crispin <blogic@openwrt.org>
Signed-off-by: Matthias Brugger <matthias.bgg@gmail.com>
Split init_reg_clock up into SoC specific callbacks. The patch also
reorders the code to avoid the need for callback function prototypes.
Signed-off-by: John Crispin <blogic@openwrt.org>
Signed-off-by: Matthias Brugger <matthias.bgg@gmail.com>
As we add support for more devices struct pmic_wrapper_type will grow and
we do not really want to start duplicating all the elements in
struct pmic_wrapper.
Signed-off-by: John Crispin <blogic@openwrt.org>
Signed-off-by: Matthias Brugger <matthias.bgg@gmail.com>
Since STAUPD interrupts aren't handled on mt8173, disable watchdog timeout
monitor of STAUPD to avoid WDT_INT triggered by STAUPD.
Signed-off-by: Henry Chen <henryc.chen@mediatek.com>
Reviewed-by: Daniel Kurtz <djkurtz@chromium.org>
Signed-off-by: Matthias Brugger <matthias.bgg@gmail.com>
Sometimes PMIC is too busy to send data in time to cause pmic wrap timeout,
because pmic wrap is waiting for FSM_VLDCLR after finishing WACS2_CMD. It
just return error when issue happened, so the state machine will stay on
FSM_VLDCLR state when data send back later by PMIC and timeout again in next
time because pmic wrap waiting for FSM_IDLE state at the beginning of the
read/write function.
Clear the vldclr when timeout if state machine stay on FSM_VLDCLR.
Signed-off-by: Henry Chen <henryc.chen@mediatek.com>
Tested-by: Ricky Liang <jcliang@chromium.org>
Signed-off-by: Matthias Brugger <matthias.bgg@gmail.com>
The watchdog may not be initialized by the bootloader, even if the rest
of the pwrap is. Move the watchdog initialization out of pwrap_init() to
make sure the watchdog is always initialized and not only when the pwrap
is uninitialized.
Signed-off-by: Henry Chen <henryc.chen@mediatek.com>
Acked-by: Sascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: Matthias Brugger <matthias.bgg@gmail.com>
platform_driver does not need to set an owner because
platform_driver_register() will set it.
Signed-off-by: Krzysztof Kozlowski <k.kozlowski@samsung.com>
Acked-by: Sascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: Matthias Brugger <matthias.bgg@gmail.com>
When the PMIC wrapper state machine has read a register it goes into the
"wait for valid clear" (vldclr) state. The state machine stays in this
state until the VLDCLR bit is written to. We should write this bit after
reading a register because the SCPSYS won't let the system go into
suspend as long as the state machine waits for valid clear.
Since now we never leave the state machine in vldclr state we no longer
have to check for this state on pwrap_read/pwrap_write entry and can
remove the corresponding code.
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: Matthias Brugger <matthias.bgg@gmail.com>
replace chipselect extension values based on SPI clock with hardcoded SoC
specific values.
The PMIC wrapper has the ability of extending the chipselects by configurable
amounts of time. We configured the values based on the rate of SPI clock, but
this is wrong. The delays should be configured based on the internal PMIC clock
that latches the values from the SPI bus to the internal PMIC registers. By
default this clock is 24MHz. Other clock frequencies are for debugging only
and can be removed from the driver.
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: Matthias Brugger <matthias.bgg@gmail.com>
The PMIC wrapper driver adds a couple of variables that are never used.
Remove them to avoid build warnings.
Signed-off-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Matthias Brugger <matthias.bgg@gmail.com>
This adds support for the PMIC wrapper found on MediaTek MT8135 and
MT8173 SoCs. The PMIC wrapper is found on MT6xxx SoCs aswell but these
are currently not supported.
On MediaTek MT8135, MT8173 and other SoCs the PMIC is connected via
SPI. The SPI master interface is not directly visible to the CPU, but
only through the PMIC wrapper inside the SoC. The communication between
the SoC and the PMIC can optionally be encrypted. Also a non standard
Dual IO SPI mode can be used to increase speed. The MT8135 also supports
a special feature named "IP Pairing". With IP Pairing the pins of some
SoC internal peripherals can be on the PMIC. The signals of these pins
are routed over the SPI bus using the pwrap bridge. Because of these
optional non SPI conform features the PMIC driver is not implemented as
a SPI bus master driver.
Signed-off-by: Flora Fu, MediaTek
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: Matthias Brugger <matthias.bgg@gmail.com>
Arnd Bergmann