This patch utilizes the new PLL clk notifier to gate then ungate the
PLL CPU clock after rate changes. This should mitigate the system
hangs observed after the introduction of cpufreq for the A33.
Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Tested-by: Quentin Schulz <quentin.schulz@free-electrons.com>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
In common PLL designs, changes to the dividers take effect almost
immediately, while changes to the multipliers (implemented as
dividers in the feedback loop) take a few cycles to work into
the feedback loop for the PLL to stablize.
Sometimes when the PLL clock rate is changed, the decrease in the
divider is too much for the decrease in the multiplier to catch up.
The PLL clock rate will spike, and in some cases, might lock up
completely. This is especially the case if the divider changed is
the pre-divider, which affects the reference frequency.
This patch introduces a clk notifier callback that will gate and
then ungate a clk after a rate change, effectively resetting it,
so it continues to work, despite any possible lockups. Care must
be taken to reparent any consumers to other temporary clocks during
the rate change, and that this notifier callback must be the first
to be registered.
This is intended to fix occasional lockups with cpufreq on newer
Allwinner SoCs, such as the A33 and the H3. Previously it was
thought that reparenting the cpu clock away from the PLL while
it stabilized was enough, as this worked quite well on the A31.
On the A33, hangs have been observed after cpufreq was recently
introduced. With the H3, a more thorough test [1] showed that
reparenting alone isn't enough. The system still locks up unless
the dividers are limited to 1.
A hunch was if the PLL was stuck in some unknown state, perhaps
gating then ungating it would bring it back to normal. Tests
done by Icenowy Zheng using Ondrej's test firmware shows this
to be a valid solution.
[1] http://www.spinics.net/lists/arm-kernel/msg552501.html
Reported-by: Ondrej Jirman <megous@megous.com>
Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Tested-by: Icenowy Zheng <icenowy@aosc.io>
Tested-by: Quentin Schulz <quentin.schulz@free-electrons.com>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
The ccu-sun9i-a80 driver uses the ccu_mult_ops struct, but unlike the other
users it doesen't select the corresponding Kconfig symbol under which the
struct is compiled in.
This results in the following link error with CONFIG_SUN9I_A80_CCU=y and
CONFIG_SUNXI_CCU_MULT=n:
drivers/built-in.o:(.data+0x2d638): undefined reference to 'ccu_mult_ops'
Fix this by explicitly selecting CONFIG_SUNXI_CCU_MULT like the other
users of the struct.
Signed-off-by: Tobias Regnery <tobias.regnery@gmail.com>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
With CONFIG_RESET_CONTROLLER=n we get the following link error in the
sunxi-ng clk driver:
drivers/built-in.o: In function `sunxi_ccu_probe':
mux-core.c:(.text+0x12fe68): undefined reference to 'reset_controller_register'
mux-core.c:(.text+0x12fe68): relocation truncated to fit: R_AARCH64_CALL26 against undefined symbol 'reset_controller_register'
Fix this by adding the appropriate select statement.
Signed-off-by: Tobias Regnery <tobias.regnery@gmail.com>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
The CLK_NUMBER value of PRCM CCU is wrongly set to (CLK_APB0_PWD + 1),
which prevented the IR mod clock from being set up.
Change it to (CLK_IR + 1) in order to correctly get IR mod set up.
Fixes: cdb8b80b60 ("clk: sunxi-ng: add support for PRCM CCUs")
Signed-off-by: Icenowy Zheng <icenowy@aosc.io>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
The first parent of ir clk in PRCM CCU is wrongly written as "osc32K"
instead of "osc32k".
Change it to "osc32k".
Fixes: cdb8b80b60 ("clk: sunxi-ng: add support for PRCM CCUs")
Signed-off-by: Icenowy Zheng <icenowy@aosc.io>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Add clock index to clock registration failure message. Clock name
is sometimes not available, when things go really wrong.
Signed-off-by: Priit Laes <plaes@plaes.org>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
The DDR1 PLL on the A33 is an oddball amongst the A33 CCU clocks.
It is a clock multiplier, with the effective multiplier in the
range of 12 ~ 255 and no offset between the multiplier value and
the value programmed into the register.
Implement the zero offset and minimum value of 12 for this clock.
Fixes: d05c748bd7 ("clk: sunxi-ng: Add A33 CCU support")
Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
The CPU cluster PLLs on the A80 are NP clocks that are atypical in two ways:
- The P factor is 1 bit wide, and translates to a /1 or /4 divider.
- The P factor should only be used for output frequencies lower than
288 MHz. The N factor has a lower limit of 12, which likely contributed
to this extra divider.
According to the user manual, the clocks can only go as low as 200 MHz.
The vendor BSP kernel does not even define operating points below 360
MHz for these clocks. The lower end for cpufreq in the vendor kernel is
even higher. The mainline Linux kernel doesn't support cpufreq for the
A80 at the moment. This means the lower frequencies are untested, and
will likely remain unused.
The new sunxi-ng style clocks don't support the quirks listed above.
Instead of trying to work the quirks in for something of little usage,
we re-model the clocks into N-type multipler clocks, with P fixed at 1.
At probe time we check if P is set to 4, and fix it up if needed. This
is highly unlikely though.
Fixes: b8eb71dcdd ("clk: sunxi-ng: Add A80 CCU")
Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Some PLL clocks are N (multiplier) type clocks, or can be simplified
as such. An example of the former is the DDR1 PLL clock on the A33.
An example of the latter is the CPU PLL clock on the A80, in which
the P divider is only used for low frequencies that are of little
use. Both clocks support PLL lock detection.
The mult clock macro implies support for this, but that is not true.
The field is simply discarded. This patch adds proper support for it.
Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
SoCs after A31 has a clock controller module in the PRCM part.
Support the clock controller module on H3/5 and A64 now.
Signed-off-by: Icenowy Zheng <icenowy@aosc.xyz>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
In commit e66f81bbd7 ("clk: sunxi-ng: Implement factors offsets"), the
final formula of NKMP clocks' recalc_rate is refactored; however, the
refactored formula broke the calculation due to some C language operand
priority problem -- the priority of operand >> is lower than * and /,
makes the formula being parsed as "(parent_rate * n * k) >> (p / m)", but
it should be "(parent_rate * n * k >> p) / m".
Add the pair of parentheses to fix up this issue. This pair of
parentheses used to exist in the old formula.
Fixes: e66f81bbd7 ("clk: sunxi-ng: Implement factors offsets")
Signed-off-by: Icenowy Zheng <icenowy@aosc.xyz>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
The mult/div for osc12M was previously backwards (giving a 48M rate
for osc12M). Fix it.
Signed-off-by: Philipp Tomsich <philipp.tomsich@theobroma-systems.com>
Tested-by: Christoph Muellner <christoph.muellner@theobroma-systems.com>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Tighten the depends on the various AllWinn SoCs so we don't
inadvertantly get clock drivers when we're not wanting them
like 32 bit SoC clocks for 64 bit configs. Ensure there's
still test coverage though.
Signed-off-by: Peter Robinson <pbrobinson@gmail.com>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Allwinner H5 is a SoC that features a design which keeps the peripheral
compatible with H3, so that it have also a CCU like the one on H3 --
only one bus gate/reset is added, and the mmc sample/output phases are
removed because of MMC controller update.
Add its support in our existing H3 CCU driver.
Signed-off-by: Icenowy Zheng <icenowy@aosc.xyz>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Some clock gates have a pre-divider between the source input and the
gate itself. A notable example is the HSIC 12 MHz clock found on the
A83T, which has the 24 MHz main oscillator as its input, and a /2
pre-divider.
Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
The enable bit offset for the hdmi-ddc module clock is wrong. It is
pointing to the main hdmi module clock enable bit.
Reported-by: Bob Ham <rah@settrans.net>
Fixes: c6e6c96d8f ("clk: sunxi-ng: Add A31/A31s clocks")
Cc: stable@vger.kernel.org # 4.9.x-
Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
A randconfig build ran into this rare link error:
drivers/clk/sunxi-ng/ccu-sun5i.o:(.data.__compound_literal.1+0x4): undefined reference to `ccu_nkmp_ops'
drivers/clk/sunxi-ng/ccu-sun5i.o:(.data.__compound_literal.7+0x4): undefined reference to `ccu_nkmp_ops'
This adds the missing 'select'.
Fixes: 5e73761786 ("clk: sunxi-ng: Add sun5i CCU driver")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
The MP style clocks support an mux with pre-dividers. While the driver
correctly accounted for them in the .determine_rate callback, it did
not in the .recalc_rate and .set_rate callbacks.
This means when calculating the factors in the .set_rate callback, they
would be off by a factor of the active pre-divider. Same goes for
reading back the clock rate after it is set.
Cc: stable@vger.kernel.org
Fixes: 2ab836db50 ("clk: sunxi-ng: Add M-P factor clock support")
Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
PTR_ERR should access the value just tested by IS_ERR, otherwise
the wrong error code will be returned.
Fixes: 783ab76ae5 ("clk: sunxi-ng: Add A80 Display Engine CCU")
Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
We get a link error when CCU_MULT is not set with the
newly added driver:
drivers/clk/sunxi-ng/ccu-sun5i.o:(.data.__compound_literal.17+0x4): undefined reference to `ccu_mult_ops'
drivers/clk/sunxi-ng/ccu-sun5i.o:(.data.__compound_literal.5+0x4): undefined reference to `ccu_mult_ops'
Fixes: 5e73761786 ("clk: sunxi-ng: Add sun5i CCU driver")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Acked-by: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
This kzalloc() could fail. Let's bail out with -ENOMEM here
instead of NULL dereferencing. That silences static checkers. We
should also cleanup on the error path even though this function
returning an error probably means the system won't boot.
Cc: Chen-Yu Tsai <wens@csie.org>
Acked-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
With the A80 SoC, Allwinner grouped and moved some subsystem specific
clock controls to a separate address space, and possibly separate
hardware block.
One such subsystem is the display engine. The main clock control unit
now only has 1 set of bus gate, dram gate, module clock, and reset
control for the entire display subsystem. These feed into a secondary
clock control unit, which has controls for each individual module
of the display pipeline. This block is not documented in the user
manual. Allwinner's kernel was used as the reference.
Add support for the display engine clock controls found on the A80.
Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Add support for the USB clock controls found on the A80.
Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Acked-by: Rob Herring <robh@kernel.org>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Add support for the main clock unit found in the A80. Some clocks were
not documented in the released user manual, but were found in the
official kernel from Allwinner. These include controls for the I2S,
SPDIF, SATA, and eDP blocks.
Note that on the A80, some subsystems have separate clock controllers
downstream of the main clock unit. These include the MMC, USB, and
display engine subsystems.
Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Acked-by: Rob Herring <robh@kernel.org>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
On the Allwinner A80 SoC, the PLL lock status indicators are grouped
together in a separate register, as opposed to being scattered in each
PLL's configuration register.
Add a flag to support this.
Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
If the CLK_SET_RATE_PARENT flag is set, we want to try getting the
closest parent rate.
Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
This patch adds support for the CLK_SET_RATE_NO_REPARENT flag to the
mux determine_rate helper, based on examples from clk-composite and
clk-mux.
Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
The determine_rate helper used ccu_mux_helper_adjust_parent_for_prediv()
to adjust the parent_rate to account for pre-dividers, but then passed
the pristine parent clock rate from clk_hw_get_rate() to the round()
callback, thereby ignoring the pre-divider adjustment. In addition,
it was saving the adjusted parent rate back into struct
clk_rate_request.
This patch fixes this by saving the pristine parent clock rate, and
adding a copy that is adjusted and passed to the round() callback.
The pristine copy, if it is the best solution, would be saved back
to struct clk_rate_request.
Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
In order to achieve all the rates asked by the GPU, we might need to change
the parent frequency.
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Acked-by: Stephen Boyd <sboyd@codeaurora.org>
Acked-by: Chen-Yu Tsai <wens@csie.org>
The divider_get_val function that is used in our determine_rate callback
doesn't try to change the parent rate at all, while clk_divider_bestdiv,
used in divider_round_rate, does.
If we have a single parent, call divider_round_rate.
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Acked-by: Stephen Boyd <sboyd@codeaurora.org>
Acked-by: Chen-Yu Tsai <wens@csie.org>
The Allwinner A10s, A13, R8 and NextThing GR8 are all based on the same
silicon, and all share the same clocks.
However, they're not packaged in the same way, and therefore not all the
controllers are actually available on all these SoCs.
Introduce a clock controller driver for all these SoCs with different
compatibles to take that into account.
Acked-by: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Some clocks have a global pre-divider that applies to all their parents.
Since it might also apply to clocks that have a single parent, this is
merged in the ccu_common structure, unlike the other pre-divider settings
that are tied to a specific index, and thus a specific parent.
Acked-by: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Some multipliers have a maximum rate that is lower than what the register
width allows to. Add a field in the multiplier structure to allow CCU
driver to set that maximum.
Acked-by: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
The previous code was always taking 1 as the minimum in it's round_rate
function, ignoring entirely what was set as minimum in the clock
definition.
Make sure that's not the case anymore.
Fixes: 2beaa601c8 ("clk: sunxi-ng: Implement minimum for multipliers")
Acked-by: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
The factors we've seen so far all had an offset of one. However, on the
earlier Allwinner SoCs, some factors could have no offset at all, meaning
that the value computed to reach the rate we want to use was the one we had
to program in the registers.
Implement an additional field for the factors that can have such an offset
(linears, not based on a power of two) to specify that offset.
This offset is not linked to the extremums that can be specified in those
structures too. The minimum and maximum are representing the range of
values we can use to try to compute the best rate. The offset comes later
on when we want to set the best value in the registers.
Acked-by: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Some clocks on the earlier SoCs such as the video PLLs are multipliers with
fractional settings.
Support those cases.
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Acked-by: Chen-Yu Tsai <wens@csie.org>
V3s has a similar but cut-down CCU to H3. Some muxes, especially clocks
about CSI, are different, which makes it to need a new CCU driver.
Add such a new driver for it.
Signed-off-by: Icenowy Zheng <icenowy@aosc.xyz>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
The audio DAI needs to set the clock rates of the ac-dig clock.
To make it possible, the parent PLL audio clock rates should
also be changed. This is possible via "CLK_SET_RATE_PARENT" flag.
Signed-off-by: Mylène Josserand <mylene.josserand@free-electrons.com>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
As the SPDIF was rarely documented on the earlier Allwinner SoCs
it was assumed that it had a similar clock register to the one
described in the H3 User Manual.
However this is not the case and it looks to shares the same setup
as the I2S clock registers.
Signed-off-by: Marcus Cooper <codekipper@gmail.com>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
The CPUX clock on A33, which is for the Cortex-A7 cores, is designed to
be changeable by changing the rate of PLL_CPUX.
Add CLK_SET_RATE_PARENT flag to this clock.
Signed-off-by: Icenowy Zheng <icenowy@aosc.xyz>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
When adjusting PLL_CPUX on A33, the PLL is temporarily driven too high,
and the system hangs.
Add a notifier to avoid this situation by temporarily switching to a
known stable 24 MHz oscillator.
Signed-off-by: Icenowy Zheng <icenowy@aosc.xyz>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
When adjusting PLL_CPUX on H3, the PLL is temporarily driven
too high, and the system becomes unstable (oopses or hangs).
Add a notifier to avoid this situation by temporarily switching
to a known stable 24 MHz oscillator.
Signed-off-by: Ondrej Jirman <megous@megous.com>
Tested-by: Lutz Sammer <johns98@gmx.net>
Acked-by: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
In the user manual of A33 SoC, the bit 22 and 23 of pll-mipi control
register is called "LDO{1,2}_EN", and according to the BSP source code
from Allwinner [1], the LDOs are enabled during the clock's enabling
process.
The clock failed to generate output if the two LDOs are not enabled.
Add the two bits to the clock's gate bits, so that the LDOs are enabled
when the PLL is enabled.
[1] https://github.com/allwinner-zh/linux-3.4-sunxi/blob/master/drivers/clk/sunxi/clk-sun8iw5.c#L429
Fixes: d05c748bd7 ("clk: sunxi-ng: Add A33 CCU support")
Signed-off-by: Icenowy Zheng <icenowy@aosc.xyz>
Acked-by: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
The PLL-MIPI clock is somewhat special as it has its own LDOs which
need to be turned on for this PLL to actually work and output a clock
signal.
Add the 2 LDO enable bits to the gate bits. This fixes issues with
the TCON not sending vblank interrupts when the tcon and dot clock are
indirectly clocked from the PLL-MIPI clock.
Fixes: c6e6c96d8f ("clk: sunxi-ng: Add A31/A31s clocks")
Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Some checkpatch warnings about spaces were missed and we didn't
mark two structs as static. Clean it up.
Cc: Maxime Ripard <maxime.ripard@free-electrons.com>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
The usual patches from us, but most notably the introduction of the A64
clocks unit.
-----BEGIN PGP SIGNATURE-----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=7+Az
-----END PGP SIGNATURE-----
Merge tag 'sunxi-clk-for-4.10' of https://git.kernel.org/pub/scm/linux/kernel/git/mripard/linux into clk-next
Pull Allwinner clock changes from Maxime Ripard:
The usual patches from us, but most notably the introduction of the A64
clocks unit.
* tag 'sunxi-clk-for-4.10' of https://git.kernel.org/pub/scm/linux/kernel/git/mripard/linux:
clk: sunxi-ng: sun8i-h3: Set CLK_SET_RATE_PARENT for audio module clocks
clk: sunxi-ng: sun8i-a23: Set CLK_SET_RATE_PARENT for audio module clocks
clk: sunxi-ng: Add A64 clocks
clk: sunxi-ng: Implement minimum for multipliers
clk: sunxi-ng: Add minimums for all the relevant structures and clocks
clk: sunxi-ng: Finish to convert to structures for arguments
clk: sunxi-ng: Remove the use of rational computations
clk: sunxi-ng: Rename the internal structures
clk: sunxi: mod0: improve function-level documentation
The audio module clocks are supposed to be set according to the sample
rate of the audio stream. The audio PLL provides the clock signal for
these module clocks, and only it is freely tunable.
Set CLK_SET_RATE_PARENT for the audio module clocks so their users can
properly tune the clock rate.
Fixes: 0577e4853b ("clk: sunxi-ng: Add H3 clocks")
Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
The audio module clocks are supposed to be set according to the sample
rate of the audio stream. The audio PLL provides the clock signal for
these module clocks, and only it is freely tunable.
Set CLK_SET_RATE_PARENT for the audio module clocks so their users can
properly tune the clock rate.
Fixes: 5690879d93 ("clk: sunxi-ng: Add A23 CCU")
Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Add the A64 CCU clocks set.
Acked-by: Rob Herring <robh@kernel.org>
Acked-by: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Allow the CCU drivers to specify a multiplier for their clocks.
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Acked-by: Chen-Yu Tsai <wens@csie.org>
Modify the current clocks we have to be able to specify the minimum for
each clocks we support, just like we support the max.
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Acked-by: Chen-Yu Tsai <wens@csie.org>
Some clocks still use an explicit list of arguments, which make it a bit
more tedious to add new parameters.
Convert those over to a structure pointer argument to add as many
arguments as possible without having to many noise in our patches, or a
very long list of arguments.
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Acked-by: Chen-Yu Tsai <wens@csie.org>
While the rational library works great, it doesn't really allow us to add
more constraints, like the minimum.
Remove that in order to be able to deal with the constraints we'll need.
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Acked-by: Chen-Yu Tsai <wens@csie.org>
Rename the structures meant to be embedded in other structures to make it
consistent with the mux structure name
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Acked-by: Chen-Yu Tsai <wens@csie.org>
On the A31, the DMA engine only works if AHB1 is clocked from PLL6.
In addition, the hstimer is clocked from AHB1, and if AHB1 is clocked
from the CPU clock, and cpufreq is working, we get an unstable timer.
Force the AHB1 clock to use PLL6 as its parent. Previously this was done
in the device tree with the assigned-clocks and assigned-clocks-parent
bindings. However with this new monolithic driver, the system critical
clocks aren't exported through the device tree. The alternative is to
force this setting in the driver before the clocks are registered.
This is also done in newer versions of mainline U-boot. But people still
using an older version, or even the vendor version, can still hit this
issue. Hence the need to do it in the kernel as well.
Reported-by: Hans de Goede <hdegoede@redhat.com>
Reported-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Fixes: c6e6c96d8f ("clk: sunxi-ng: Add A31/A31s clocks")
Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
There's been a copy and paste mistake in the A23 and A33 from the H3,
leading in the reset offset for the UART and I2C.
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
The register offset for the mipi-csi clk is off by 4, a copy paste
error from the mipi-dsi clk.
Fixes: c6e6c96d8f ("clk: sunxi-ng: Add A31/A31s clocks")
Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Acked-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
The PLLs have a "lock" bit in their configuration registers which
indicate if the PLL has locked on to the requested clock rate. We
check this bit in the .set_rate op. The PLL cannot lock on if it's
not running, which might be a false positive (warning).
Set the CLK_SET_RATE_UNGATE flag for all PLLs so whenever clk_set_rate
is called on them, they get enabled and the "lock" check is really
checking the PLL.
Fixes: c6e6c96d8f ("clk: sunxi-ng: Add A31/A31s clocks")
Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Acked-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
The LCD controller and HDMI controller use the LCDx-CHy and HDMI clocks
to generate their dot clocks. To be able to generate a full range of
possible clock rates, the parent PLL clock rates should also be changed.
Fixes: c6e6c96d8f ("clk: sunxi-ng: Add A31/A31s clocks")
Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Acked-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Four more SoCs converted to the new clock framework (A31, A31s, A23 and
A33).
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1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=Ff/R
-----END PGP SIGNATURE-----
Merge tag 'sunxi-clk-for-4.9' of https://git.kernel.org/pub/scm/linux/kernel/git/mripard/linux into clk-next
Pull Allwinner clock driver changes from Maxime Ripard:
Four more SoCs converted to the new clock framework (A31, A31s, A23 and
A33).
* tag 'sunxi-clk-for-4.9' of https://git.kernel.org/pub/scm/linux/kernel/git/mripard/linux:
clk: sunxi-ng: Add hardware dependency
clk: sunxi-ng: Add A23 CCU
clk: sunxi-ng: Add A33 CCU support
clk: sunxi-ng: Add N-class clocks support
clk: sunxi-ng: mux: Add mux table macro
clk: sunxi-ng: div: Allow to set a maximum
clk: sunxi-ng: div: Add kerneldoc for the _ccu_div structure
clk: sunxi-ng: div: Add mux table macros
clk: sunxi-ng: Add A31/A31s clocks
clk: sunxi-ng: mux: Add clk notifier functions
clk: sunxi-ng: mux: support fixed pre-dividers on multiple parents
clk: sunxi-ng: mux: Add support for mux tables
clk: sunxi-ng: mux: Rename mux macro to be consistent
clk: sunxi-ng: nkm: Add mux to support multiple parents
clk: sunxi-ng: mux: Increase fixed pre-divider div size
The sunxi-ng clock driver is useless for other architectures.
Signed-off-by: Jean Delvare <jdelvare@suse.de>
Acked-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Add support for the clock unit found in the A23. Due to the similarities
with the A33, it also shares its clock IDs to allow sharing the DTSI.
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Acked-by: Chen-Yu Tsai <wens@csie.org>
This commit introduces the clocks found in the Allwinner A33 CCU.
Since this SoC is very similar to the A23, and we share a significant share
of the DTSI, the clock IDs that are going to be used will also be shared
with the A23, hence the name of the various header files.
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Acked-by: Chen-Yu Tsai <wens@csie.org>
Add support for the class with a single factor, N, being a multiplier.
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Acked-by: Chen-Yu Tsai <wens@csie.org>
Add a new macro to declare muxes based on a table and a gate.
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Acked-by: Chen-Yu Tsai <wens@csie.org>
Some dividers might have a maximum value that is lower than the width of
the register.
Add a field to _ccu_div to handle those case properly. If the field is set
to 0, the code will assume that the maximum value is the maximum one that
can be used with the field register width.
Otherwise, we'll use whatever value has been set.
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Acked-by: Chen-Yu Tsai <wens@csie.org>
The internal _ccu_div structure is meant to be embedded into other
structures to combine the various dividers and to form the clock classes
support.
Start to document those structures by using kerneldoc.
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Add some macros to ease the declaration of clocks that are using them.
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Acked-by: Chen-Yu Tsai <wens@csie.org>
The reset register offsets for UART*, I2C* and SCR were off by a few bytes.
Signed-off-by: Jorik Jonker <jorik@kippendief.biz>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Add a new style driver for the clock control unit in Allwinner A31/A31s.
A few clocks are still missing:
- MIPI PLL's HDMI mode support
- EMAC clock
Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Acked-by: Rob Herring <robh@kernel.org>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
On sunxi we support cpufreq by changing the clock rate of PLL-CPU.
It's possible the clock output of the PLL goes out of the CPU's
operational limits when the PLL's multipliers / dividers are changed
and it hasn't stabilized yet. This would result in the CPU hanging.
To circumvent this, we temporarily switch the CPU mux clock to another
stable clock before the rate change, and switch it back after the PLL
stabilizes. This is done with clk notifiers registered on the PLL.
This patch adds common functions for notifiers to reparent mux clocks.
Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Some clocks on the A31 have fixed pre-dividers on multiple parents.
Add support for them.
Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Some clock muxes have holes, i.e. invalid or unconnected inputs,
between parent mux values.
Add support for specifying a mux table to map clock parents to
mux values.
Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Rename the internal mux macro to be consistent with the other internal
structure macros.
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
The MIPI mode of the MIPI-PLL on A31 is an NKM-style PLL with 2
selectable parents.
Add mux support to the NKM clock.
Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Some clocks have a predivider value that is larger than what u8 can
store. One such example is the OUT clk found on A20/A31, which has
a /750 pre-divider on one of the osc24M parents.
Increase the size of the div field to u16.
Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
make C=2 reports:
CHECK drivers/clk/sunxi-ng/ccu_nk.c
drivers/clk/sunxi-ng/ccu_nk.c:17:6: warning: symbol 'ccu_nk_find_best' was
not declared. Should it be static?
ccu_nk_find_best is only used within ccu_nk.c. So make it static to get
rid of this warning.
Fixes: adbfb0056e ("clk: sunxi-ng: Add N-K-factor clock support")
Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
The condition passed to read*_poll_timeout() is the break condition,
i.e. wait for this condition to happen and return success.
The original code assumed the opposite, resulting in a warning when
the PLL clock rate was changed but never lost it's lock as far as
the readout indicated. This was verified by checking the read out
register value.
Fixes: 1d80c14248 ("clk: sunxi-ng: Add common infrastructure")
Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Acked-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
The code had a typo and got the wrong offset for the hardcoded divider, fix
that.
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Reported-by: Jean-Francois Moine <moinejf@free.fr>
Reported-by: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Michael Turquette <mturquette@baylibre.com>
Link: lkml.kernel.org/r/20160711203448.18062-1-maxime.ripard@free-electrons.com
Introduce support for clocks that use a combination of two linear
multipliers (N and K factors), one linear divider (M) and one power of two
divider (P).
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Signed-off-by: Michael Turquette <mturquette@baylibre.com>
Link: lkml.kernel.org/r/20160629190535.11855-13-maxime.ripard@free-electrons.com
Introduce support for clocks that multiply and divide using two linear
multipliers and one linear divider.
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Signed-off-by: Michael Turquette <mturquette@baylibre.com>
Link: lkml.kernel.org/r/20160629190535.11855-12-maxime.ripard@free-electrons.com
Introduce support for the clocks that combine a linear divider and a
power-of-two based one.
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Signed-off-by: Michael Turquette <mturquette@baylibre.com>
Link: lkml.kernel.org/r/20160629190535.11855-9-maxime.ripard@free-electrons.com
Add support for the various dividers (linear, table or pow-of-two based)
found in the CCU.
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Signed-off-by: Michael Turquette <mturquette@baylibre.com>
Link: lkml.kernel.org/r/20160629190535.11855-8-maxime.ripard@free-electrons.com
Add support for the clocks in the CCU that introduce a phase shift from
their parent clock.
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Signed-off-by: Michael Turquette <mturquette@baylibre.com>
Link: lkml.kernel.org/r/20160629190535.11855-7-maxime.ripard@free-electrons.com
Some clocks in the Allwinner SoCs clocks unit are just muxes.
However, those muxes might also be found in some other complicated clocks
that would benefit from the code in there to deal with "advanced" features,
like pre-dividers.
Introduce a set of helpers to reduce the code duplication in such cases.
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Signed-off-by: Michael Turquette <mturquette@baylibre.com>
Link: lkml.kernel.org/r/20160629190535.11855-6-maxime.ripard@free-electrons.com
Some clocks in the Allwinner SoCs clocks unit are just simple gates. Add
support for those clocks.
Since it's a feature that can also be found in more complex clocks, provide
a bunch of helpers that can be reused later on.
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Signed-off-by: Michael Turquette <mturquette@baylibre.com>
Link: lkml.kernel.org/r/20160629190535.11855-5-maxime.ripard@free-electrons.com
Some clocks can be switched to a mode called fractional that have two fixed
output rate you can choose from.
Add a small library to deal with those clocks.
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Signed-off-by: Michael Turquette <mturquette@baylibre.com>
Link: lkml.kernel.org/r/20160629190535.11855-4-maxime.ripard@free-electrons.com
Start our new clock infrastructure by adding the registration code, common
structure and common code.
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Signed-off-by: Michael Turquette <mturquette@baylibre.com>
Link: lkml.kernel.org/r/20160629190535.11855-3-maxime.ripard@free-electrons.com
