Age | Commit message (Collapse) | Author |
|
This enables per CS training at 666MHz and above for ddrfreq per
vendor recommendation. Since the threshold was used for latency was
the same value, this also adds a new value for that.
Signed-off-by: Derek Basehore <dbasehore@chromium.org>
|
|
This patch sets the frequency configuration of the next DRAM DFS index
to the configuration of the current index. This does not perform a
frequency transition. It just configures registers so the training on
resume for both indices will be correct.
Signed-off-by: Derek Basehore <dbasehore@chromium.org>
Signed-off-by: Xing Zheng <zhengxing@rock-chips.com>
|
|
We used dcf do ddr frequency scaling, but we just include a dcf
binary, it hard to maintain later, we have M0 compile flow in ATF,
and M0 can also work for ddr frequency scaling, so let's use it.
Signed-off-by: Lin Huang <hl@rock-chips.com>
Signed-off-by: Xing Zheng <zhengxing@rock-chips.com>
|
|
This makes the file consistently use tabs instead of mixing in spaces.
Signed-off-by: Derek Basehore <dbasehore@chromium.org>
|
|
Since the ATF project, we usually use the mmio_read_32 and
mmio_write_32. And the mmio_write_32, the firse parameter
is ADDR, the second is VALUE. In order to style consistency:
1/ rename readl/writel to mmio_read_32/mmio_write_32
2/ for keeping the same with mmio_write_32 in the ATF project,
swap the order of the parameters for M0 mmio_write_32
Signed-off-by: Xing Zheng <zhengxing@rock-chips.com>
Signed-off-by: Lin Huang <hl@rock-chips.com>
|
|
when dram frequency below 260MHz, phy master dll may unlock, so
let phy master dll working at dll bypass mode when frequency is
below 260MHz.
Signed-off-by: Lin Huang <hl@rock-chips.com>
|
|
we can reuse the dram config from loader, so we can remove dram_init()
and dts_timing_receive() funciton in dram.c, add the dram_set_odt_pd()
function to get the odt and auto power down parameter from kernel.
This also removes the dcf_code_init function to allow the system to
actually boot.
Signed-off-by: Lin Huang <hl@rock-chips.com>
Signed-off-by: Derek Basehore <dbasehore@chromium.org>
Signed-off-by: Xing Zheng <zhengxing@rock-chips.com>
|
|
Replace all use of memset by zeromem when zeroing moderately-sized
structure by applying the following transformation:
memset(x, 0, sizeof(x)) => zeromem(x, sizeof(x))
As the Trusted Firmware is compiled with -ffreestanding, it forbids the
compiler from using __builtin_memset and forces it to generate calls to
the slow memset implementation. Zeromem is a near drop in replacement
for this use case, with a more efficient implementation on both AArch32
and AArch64.
Change-Id: Ia7f3a90e888b96d056881be09f0b4d65b41aa79e
Signed-off-by: Douglas Raillard <douglas.raillard@arm.com>
|
|
One nasty part of ATF is some of boolean macros are always defined
as 1 or 0, and the rest of them are only defined under certain
conditions.
For the former group, "#if FOO" or "#if !FOO" must be used because
"#ifdef FOO" is always true. (Options passed by $(call add_define,)
are the cases.)
For the latter, "#ifdef FOO" or "#ifndef FOO" should be used because
checking the value of an undefined macro is strange.
Here, IMAGE_BL* is handled by make_helpers/build_macro.mk like
follows:
$(eval IMAGE := IMAGE_BL$(call uppercase,$(3)))
$(OBJ): $(2)
@echo " CC $$<"
$$(Q)$$(CC) $$(TF_CFLAGS) $$(CFLAGS) -D$(IMAGE) -c $$< -o $$@
This means, IMAGE_BL* is defined when building the corresponding
image, but *undefined* for the other images.
So, IMAGE_BL* belongs to the latter group where we should use #ifdef
or #ifndef.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
|
|
Macro cleanups
|
|
We have lots of duplicated defines (and comment blocks too).
Move them to include/plat/common/common_def.h.
While we are here, suffix the end address with _END instead of
_LIMIT. The _END is a better fit to indicate the linker-derived
real end address.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
|
|
We are duplicating this macro define, and it is useful enough
to be placed in the common place.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
|
|
Depending on the compiler used, it might try to link in libc even though
it's not required. Stop it from doing that.
Signed-off-by: Patrick Georgi <pgeorgi@google.com>
|
|
Some files have incorrect copyright notices, this patch fixes all
files with deviations from the standard notice.
Change-Id: I66b73e78a50a235acb55f1e2ec2052a42c0570d2
Signed-off-by: Antonio Nino Diaz <antonio.ninodiaz@arm.com>
|
|
Commit b91d935feebdf2c3edef19949023645b6cb34b20
("Add CFI debug frame information for ASM functions") modifies the
assembly macros 'func' and 'endfunc' to include CFI debug frame
information.
The rockchip platform uses a custom version of the 'func' macro with the
common 'endfunc' macro. The custom macro wasn't updated in
b91d935feebdf2c3edef19949023645b6cb34b20 resulting in the following
build error:
plat/rockchip/rk3399/drivers/pmu/plat_pmu_macros.S: Assembler messages:
plat/rockchip/rk3399/drivers/pmu/plat_pmu_macros.S:155: Error: .cfi_endproc without corresponding .cfi_startproc
Makefile:532: recipe for target 'build/rk3399/release/bl31/plat_helpers.o' failed
make: *** [build/rk3399/release/bl31/plat_helpers.o] Error 1
Fixing this by updating the sram_func macro in the rk3399 port.
Fixes: b91d935feebdf2c3edef19949023645b6cb34b20 ("Add CFI debug frame information for ASM functions")
Signed-off-by: Soren Brinkmann <soren.brinkmann@xilinx.com>
|
|
rk3399: fixes and updates for s2r
|
|
RK3399 M0 build system improvements
|
|
We have do something for clocks gate.
Fox example as the below:
susped:
clk_gate_con_save();
clk_gate_con_disable();
resume:
clk_gate_con_restore();
--
SO, add the plls_suspend_prepare() and plls_resume_finish() are not
necessary to S2R, that will save S2R time if remove them.
BRANCH=none
BUG=chrome-os-partner:58870,chrome-os-partner:55934
TEST=build kevin, two dogfooders with suspend_stress_test
passing 3000 cycles and still going on.
Change-Id: Icfbabc0b3ea8d2b5108d4f3de99a803b6d459669
Signed-off-by: Caesar Wang <wxt@rock-chips.com>
|
|
The CA53 and CM0 WDT clock gating in rk3399 SGRF, and ATF is in charge of
it because the kernel can't touch SGRF.
Basically the WDT didn't stop at suspend time, it just switched from the
24M to the 32k clock. That meant that the WDT would fire if you slept for
long enough. In other word, the watchdog timer over count will increase to
750 (24*1000/32) times.
The RK3399 HW watchdog interval is 21 seconds. When machine enters the
suspend, the watchdog will reset the system after 35.7 (750/21) hours.
BUG=chrome-os-partner:59257
TEST=daisydog checked and set value, powerd_dbus_suspend to verify.
Change-Id: I88bb2a05b7d67d5ffd292f9d05d033ae9a6a3593
Signed-off-by: Caesar Wang <wxt@rock-chips.com>
|
|
This patch adds dependency rule generation and inclusion to the M0
Makefile, so that M0 objects will get correctly remade with an
incremental build if a header file they included changed.
Change-Id: I2067bd9fd4d9dad3e77a09cbf09c7b4db3c1eda5
Signed-off-by: Julius Werner <jwerner@chromium.org>
|
|
The dependencies in the M0 Makefile are not correctly laid out, which
may lead to errors with make -j if the binary target gets evaluated
before the target that creates the directory. In addition, the M0
Makefile just calls mkdir without using the platform-independent macros
from the main ARM TF build system. This patch fixes those issues,
removes some unused (and broken) M0 build targets and merges the two M0
output directories into one (since there's no real point splitting it up
and it creates more hassle).
Change-Id: Ia5002479cf9c57fea7aefa8ca88e373df3a51f61
Signed-off-by: Julius Werner <jwerner@chromium.org>
|
|
The RK3399 supports close the center logic enter power mode,
so we can close PD_CENTER to save more power during suspend.
Therefore, we need to support save/restore the DDR PHY and
controller registers during suspend/resume.
Also, need CL (http://crosreview.com/397399) to check disabling
center logic.
Change-Id: I288defd8e9caa3846d9fa663a33e4d51df1aaa5d
Signed-off-by: Xing Zheng <zhengxing@rock-chips.com>
Signed-off-by: Derek Basehore <dbasehore@chromium.org>
Signed-off-by: Caesar Wang <wxt@rock-chips.com>
|
|
This patch intend to support save the registers of the DDR controller
and PHY before suspend, and restore them after resume.
Change-Id: Ia10b476c0b837628ac0f365416a7118292753e96
Signed-off-by: Xing Zheng <zhengxing@rock-chips.com>
Signed-off-by: Derek Basehore <dbasehore@chromium.org>
Signed-off-by: Caesar Wang <wxt@rock-chips.com>
|
|
This changes the style of dmc register accesses to be a read/write on
a base address plus a register offset instead of reinterpretting a
base address as a struct and accessing members within that struct.
Change-Id: Iead097cd6afdb830d8bc193608cd39d01ce5a6bc
Signed-off-by: Derek Basehore <dbasehore@chromium.org>
Signed-off-by: Caesar Wang <wxt@rock-chips.com>
|
|
This renames dram.c and dram.h to dfs.c and dfs.h respectively. This
is to make room for common functionality between frequency scaling and
suspend code for the DRAM in a pair of common files named dram.c and
dram.h. It also removes a duplicate enum definition from
dram_spec_timing.h
Change-Id: Ibfa1041f8781401f9d27901fe8c61862bcb05562
Signed-off-by: Derek Basehore <dbasehore@chromium.org>
Signed-off-by: Caesar Wang <wxt@rock-chips.com>
|
|
This moves the PMU register definitions into another file for use in
later patches.
Change-Id: I8b5f1e7938b63ada6a743cf9661c3e474e96e4e4
Signed-off-by: Derek Basehore <dbasehore@chromium.org>
Signed-off-by: Caesar Wang <wxt@rock-chips.com>
|
|
rockchip: optimize the link mechanism for SRAM code
|
|
rockchip: fix A72 L2CTLR_DATA_RAM_LATENCY to 5
|
|
The default value of L2CTLR_DATA_RAM_LATENCY is 2, depends to
the test result on rk3399, the A72 will need lower voltage for
high frequency if it's set to be 5, and almost no effect on performance.
Change-Id: I99a6a43edcc0c58f7775c10f4b85669dc3eff66d
Signed-off-by: Caesar Wang <wxt@rock-chips.com>
|
|
Due to the PMU design, the PMU may not clear the WAKEUP bit after
wakeup, therefore, the state machine at the power mode may enter
the infinite loop during WFI.
There is a solution that we can use the M0 to monitor the WAKEUP
bit and clear it during power mode, then the state machine will be
recovered immediately. Then, the DUT can exit the WFI normally.
Change-Id: I303628553b728c214bf2d436bd3122032b5e669c
Signed-off-by: Xing Zheng <zhengxing@rock-chips.com>
Signed-off-by: Caesar Wang <wxt@rock-chips.com>
|
|
This CL supports add M0 source code to built into the bl31.bin, the
goal is that we can load the M0 code binary into SRAM and execute it.
We need the M0 help us to clean the power_mode_en bit during the AP
PMU enter the state machine with interrupt, and avoid to the AP can
not exit the loop forever.
Change-Id: I844582c54a1f0d44ca41290d44618df58679f341
Signed-off-by: Xing Zheng <zhengxing@rock-chips.com>
Signed-off-by: Patrick Georgi <pgeorgi@google.com>
Signed-off-by: Caesar Wang <wxt@rock-chips.com>
|
|
Add the common extra.ld.S and customized rk3399.ld.S to extend
to more features for different platforms.
For example, we can add SRAM section and specific address to
load there if we need it, and the common bl31.ld.S not need to
be modified.
Therefore, we can remove the unused codes which copying explicitly
from the function pmusram_prepare(). It looks like more clear.
Change-Id: Ibffa2da5e8e3d1d2fca80085ebb296ceb967fce8
Signed-off-by: Xing Zheng <zhengxing@rock-chips.com>
Signed-off-by: Caesar Wang <wxt@rock-chips.com>
|
|
The CRU_CLKSEL_COUNT value is 108, not 0x108.
Signed-off-by: Caesar Wang <wxt@rock-chips.com>
Change-Id: Ib9db066b8b3ecafcee7f645dd5633b55a808e3d7
|
|
As the new RK3399TRM v1.1, there are some wrong set for CRU_CLKSEL_CON
register.
As the CRU_CLKSEL_CON96~107 high 16-bit isn't write mask and the
CRU_CLKSEL_CON offset is 0x100,not 0x80.
Change-Id: Ie127e9de74b87100af9a0150aad43e89e4972529
|
|
rockchip: Fixes typo and warnings
|
|
We make sure the resuming of gic need to be enabled.
Otherwise, The resume will hit the below panic.
...
[ 24.230541] CPU0: update max cpu_capacity 451
[ 24.236029] CPU5: update max cpu_capacity 1024
[ 24.236046] CPU4: shutdown
[ 24.243205] psci: CPU4 killed.
[ 24.258730] CPU5: shutdown
[ 24.261472] psci: CPU5 killed.
[ 24.270417] GIC: unable to set SRE (disabled at EL2), panic ahead
[ 24.270417] cat[7801]: undefined instruction: pc=ffffffc0004e65d0
[ 24.270417] Code: b0003940 91274400 97f871af d2801e00 (d5184600)
[ 24.270417] Internal error: Oops - undefined instruction: 0 [#1] PREEMPT
Change-Id: Ie9542c8d5768ba0accfa073453da8bfe06d4f921
|
|
As the checkpatch reports the warning or error.
plat/rockchip/common/plat_pm.c:96:
ERROR: do not set execute permissions for source files
plat/rockchip/rk3399/drivers/pmu/pmu.c:294:
ERROR: do not set execute permissions for source files
plat/rockchip/common/plat_pm.c:286: WARNING: line over 80 characters
plat/rockchip/common/plat_pm.c:287: WARNING: line over 80 characters
Change-Id: Ib347da21c56551c31df3f90f03777b13c75d5c26
|
|
for compatible 32bit and 64bit, we use 0x82xxxxxx as function ID,
we modify SIP call function return value to 32 bit.
Change-Id: Ib99b03a9ea423853aaa296dcc634ee82c622a552
|
|
For save power cosumption, if gpio power supply shut down, we need to
set gpio2 ~ gpio4 to input and HiZ status when suspend, and recovery
they status when rusume. we do it base on apio pass from loader.
Change-Id: I59fd2395e5e37e63425472a39f519822c9197e4c
|
|
some specific board need to disable/enable specific gpio when
suspend/resume, so we add this function, bootloader can pass the
specific gpio, and we can handle these gpios in bl31 suspend/resuem
function.
Change-Id: I373b03ef9202ee4a05a2b9caacdfa01b47ee2177
|
|
We may need gpio pull mode later, so add this function.
Besides fix a set pull mode bug, and save gpio clock gate,
when operate the gpio, we will enable gpio clock, when
finish gpio operate, restore gpio clock gate status.
Change-Id: Ia1d602804f571a17f5ddc499908663b968b02974
|
|
As rk3399 reported the d8/octane scores drop 10% with cpu idle.
The root cause is thc cpu cluster enter the slow mode.
We don't need switch the clock to 24MHz if cpu cluster enter the
retention mode. In order to improve performance, it just needs for
cluster enter powering off mode.
Also, we shouldn't do anything for hlvl if the system is off.
Change-Id: I2a02962a01343abd0cba47ed63192c1cdf88b119
|
|
Fixes suspend/resume bugs
|
|
rockchip: add dram driver for rk3399
|
|
For the PMU design, we don't expect to get the interrupts before enter
the power mode. Since that will cause the confusion for the state
machine in the power mode.
Change-Id: Id8dee79ae617a66271b5caf92caf35f520f45099
|
|
Change-Id: I986d64df9dc62354d50ccea0468b90f090a44160
Signed-off-by: Caesar Wang <wxt@rock-chips.com>
|
|
If we don't enable the Schmitt trigger on the 32 kHz clock then systems
won't always resume from suspend properly. Presumably anything else in
the system that relies on the 32 kHz clock also will have problems
without the Schmitt trigger enabled.
Enable it always since having the 32 kHz clock on GPIO0_A0 isn't
exactly an optional feature, so all boards using rk3399 will need this.
Change-Id: Idc18c6cd1adc5be5f60efd9cb805d83d5cd40129
|
|
add auto_pd_dis_freq parameter, we can pass a frequency from kernel
to disable or enable ddr auto power down function.
Change-Id: Ie30914701336c59047c380381c6b75dd76a89562
|
|
add dram driver, and kernel can through sip function talk to bl31 to
do ddr frequency scaling. and ddr auto powerdown.
Change-Id: I0d0f2869aed95e336c6e23ba96a9310985c84840
|
|
In a previous change we mistakenly thought that PMU_24M_EN_CFG directly
controlled whether the PMU counts ran off the 32k vs. 24M clock.
Apparently that's not true. Real logic is now documented in code.
Also in the previous change we mistaknely though that PMU_24M_EN_CFG was
normally supposed to be 1 and we should "restore" it at resume time.
This is a terrible idea and made the system totally unreliable after
resume. Apparently PMU_24M_EN_CFG should always be 0 with all the
current code and settings.
Let's fix the above two problems. While we're changing all of this,
let's also:
1. Init at boot time. Many of these counts are used when the system is
running normally. We want the behavior at boot to match the behavior
after suspend/resume.
2. Init CPU counts to be 1 us. Although old code was trying to set this
to 1 ms (1000x slower) at suspend/resume time, we've been testing the
kernel with 1 us for a long time now. That's because the kernel (at
boot time) set these values to 24. Let's keep at 24 until we know
that's wrong.
3. Init GPU counts to be 1 us. Old code wasn't touching the GPU, but as
documented in comments it makes sense to init here. Do it.
4. Document the crap out of this code, since the SoC's behavior is
confusing and poorly documented in the TRM.
5. Increase some stabilization times to 30 ms (from 3 ms). It's unclear
that a full 30 ms is needed, but let's be safe for now.
This also inits the counts for the GPU.
(Thanks to Doug's patch that come from https://crosreview.com/372381)
Change-Id: Id1bc159a5a99916aeab043895e5c4585c4adab22
|