GitLab

To catch early hangs in resume, this sets up the watchdog before
anything else in the pmusram code (ignoring setting up the stack...).
This uses hard coded settings for the watchdog until the proper
watchdog restore later on in the firmware/kernel.

This also restores the old watchdog register values before the PLLs
are restored to make sure we don't temporarily switch over to a 1/3s
timeout on the watchdog when the pclk_wdt goes from 4MHz to 100MHz.

Change-Id: I8f7652089a88783271b17482117b4609330abe80
Signed-off-by: Derek Basehore <dbasehore@chromium.org>

All the m0 code run in SRAM before, but we need to watch PMU_POWER_ST
when SOC enter into FSM, and SRAM will shutdown during this time, so
this code need run in PMUSRAM. But PMUSRAM only 8K space, we can not
put all the m0 binary into PMUSRAM, Split the M0 binary into two, dram
part still run in SRAM, and suspend part run in PMUSRAM.

Change-Id: Ie08bdf3e2b8838f12b9297fe60ab0aad219684b1
Signed-off-by: Lin Huang <hl@rock-chips.com>

we need to enable PMU_WKUP_RST_EN for pmu powermode configure, since
enable wakeup reset will hold the soc status, so the SOC will not affect
by some power or other single glitch when resume, and keep the soc in the
right status. And it not need to enable DDRIO_RET_HW_DE_REQ, the ddr resume
will do it manual.

Change-Id: Ib4af897ffb3cb63dc2aa9a6002e5d9ef86ee4a49
Signed-off-by: Lin Huang <hl@rock-chips.com>

Since commit 031dbb12

 ("AArch32: Add essential Arch helpers"),
it is difficult to use consistent format strings for printf() family
between aarch32 and aarch64.

For example, uint64_t is defined as 'unsigned long long' for aarch32
and as 'unsigned long' for aarch64.  Likewise, uintptr_t is defined
as 'unsigned int' for aarch32, and as 'unsigned long' for aarch64.

A problem typically arises when you use printf() in common code.

One solution could be, to cast the arguments to a type long enough
for both architectures.  For example, if 'val' is uint64_t type,
like this:

  printf("val = %llx\n", (unsigned long long)val);

Or, somebody may suggest to use a macro provided by <inttypes.h>,
like this:

  printf("val = %" PRIx64 "\n", val);

But, both would make the code ugly.

The solution adopted in Linux kernel is to use the same typedefs for
all architectures.  The fixed integer types in the kernel-space have
been unified into int-ll64, like follows:

    typedef signed char           int8_t;
    typedef unsigned char         uint8_t;

    typedef signed short          int16_t;
    typedef unsigned short        uint16_t;

    typedef signed int            int32_t;
    typedef unsigned int          uint32_t;

    typedef signed long long      int64_t;
    typedef unsigned long long    uint64_t;

[ Linux commit: 0c79a8e29b5fcbcbfd611daf9d500cfad8370fcf ]

This gets along with the codebase shared between 32 bit and 64 bit,
with the data model called ILP32, LP64, respectively.

The width for primitive types is defined as follows:

                   ILP32           LP64
    int            32              32
    long           32              64
    long long      64              64
    pointer        32              64

'long long' is 64 bit for both, so it is used for defining uint64_t.
'long' has the same width as pointer, so for uintptr_t.

We still need an ifdef conditional for (s)size_t.

All 64 bit architectures use "unsigned long" size_t, and most 32 bit
architectures use "unsigned int" size_t.  H8/300, S/390 are known as
exceptions; they use "unsigned long" size_t despite their architecture
is 32 bit.

One idea for simplification might be to define size_t as 'unsigned long'
across architectures, then forbid the use of "%z" string format.
However, this would cause a distortion between size_t and sizeof()
operator.  We have unknowledge about the native type of sizeof(), so
we need a guess of it anyway.  I want the following formula to always
return 1:

  __builtin_types_compatible_p(size_t, typeof(sizeof(int)))

Fortunately, ARM is probably a majority case.  As far as I know, all
32 bit ARM compilers use "unsigned int" size_t.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>

Commit 4c0d0390

 ("Rework type usage in Trusted Firmware") changed
the type usage in struct declarations, but did not touch the definition
side.  Fix the type mismatch.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>

This fixes an off by 576x bug the the sram_udelay code. The wrong
value was multipled by the system ticks per mhz value (which is 24),
so we delayed for 1/576th of the requested time.
Signed-off-by: Derek Basehore <dbasehore@chromium.org>

Void pointers have been used to access linker symbols, by declaring an
extern pointer, then taking the address of it. This limits symbols
values to aligned pointer values. To remove this restriction an
IMPORT_SYM macro has been introduced, which declares it as a char
pointer and casts it to the required type.

Change-Id: I89877fc3b13ed311817bb8ba79d4872b89bfd3b0
Signed-off-by: Joel Hutton <Joel.Hutton@Arm.com>

Ensure (where possible) that switch statements in plat comply with MISRA
rules 16.1 - 16.7.

Change-Id: Ie4a7d2fd10f6141c0cfb89317ea28a755391622f
Signed-off-by: Jonathan Wright <jonathan.wright@arm.com>

When the source code says 'SMCC' it is talking about the SMC Calling
Convention. The correct acronym is SMCCC. This affects a few definitions
and file names.

Some files have been renamed (smcc.h, smcc_helpers.h and smcc_macros.S)
but the old files have been kept for compatibility, they include the
new ones with an ERROR_DEPRECATED guard.

Change-Id: I78f94052a502436fdd97ca32c0fe86bd58173f2f
Signed-off-by: Antonio Nino Diaz <antonio.ninodiaz@arm.com>

there are two fix for save/restore watchdog register:
1. watchdog plck will shutdown after secure_watchdog_disable(), so need
   to save register before it and restore after secure_watchdog_enable().
2. need write 0x76 to cnt_restart to keep watchdog alive when restore
   watchdog register.

Change-Id: I1f6fbceae22186e3b72a87df6332a110adf37479
Signed-off-by: Lin Huang <hl@rock-chips.com>

MISRA C-2012 Rule 7.3 violation: lowercase l shall not be used as literal suffixes.

This patch resolves this for the ULL() macro by using ULL suffix instead
of the ull suffix.

Change-Id: Ia8183c399e74677e676956e8653e82375d0e0a01
Signed-off-by: David Cunado <david.cunado@arm.com>

After executing a TLBI a DSB is needed to ensure completion of the
TLBI.

rk3328: The MMU is allowed to load TLB entries for as long as it is
enabled. Because of this, the correct place to execute a TLBI is right
after disabling the MMU.

Change-Id: I8280f248d10b49a8c354a4ccbdc8f8345ac4c170
Signed-off-by: Antonio Nino Diaz <antonio.ninodiaz@arm.com>

Coverity scan done for the coreboot project found the issue:
Coverity (*** CID 1385418: Memory - illegal accesses (OVERRUN))
Coverity (*** CID 1385419: Memory - corruptions  (OVERRUN))

Fix the Converity error issue with store_cru[] loop needs to be one
element bigger.

Fixes: ARM-software/tf-issues#544

Change-Id: I420f0a660b24baaa5fc5e78fca242cf750c9bbc7
Signed-off-by: Caesar Wang <wxt@rock-chips.com>

Change-Id: Ib67b841ab621ca1ace3280e44cf3e1d83052cb73
Signed-off-by: Dimitris Papastamos <dimitris.papastamos@arm.com>

This disables the redistributor before either of the pwr_dm_suspend
functions are called. This is because the rdist save code in the
rk3399 rockchip_soc_sys_pwr_dm_suspend function requires that each
redistributor be disabled before saving state.
Signed-off-by: Derek Basehore <dbasehore@chromium.org>

This adds calls to the GICv3 save/restore functions for the GIC
distributor and redistributor.
Signed-off-by: Derek Basehore <dbasehore@chromium.org>

We were looping for MAX_WAIT_COUNT in several places without any
delays, so this adds the delays to make those loops more predictable.
Signed-off-by: Derek Basehore <dbasehore@chromium.org>

The code was accidentally restoring the QOS on suspend and saving the
QOS on resume. This is the opposite of what we want.
Signed-off-by: Derek Basehore <dbasehore@chromium.org>

This brings ATF into line with the kernel on the timeout for power
domains turning on. We could actually timeout (when we shouldn't) on
resume when turning power domains on. The guaranteed maximum delay is
now 10ms.
Signed-off-by: Derek Basehore <dbasehore@chromium.org>

This patch changes all Rockchip platforms to use the new
MULTI_CONSOLE_API. The platform-specific plat_crash_console
implementations are removed so that the platform can use the ones from
the common platform code instead.

Also change the registers used in plat_crash_print_regs. The existing
use of x16 and x17 has always been illegal, since those registers are
reserved for use by the linker as a temporary scratch registers in
intra-procedure-call veneers and can never be expected to maintain their
values across a function call.

Change-Id: I8249424150be8d5543ed4af93b56756795a5288f
Signed-off-by: Julius Werner <jwerner@chromium.org>

This patch changes all Rockchip platforms to initialize the serial
console with information supplied by coreboot rather than hardcoded
base address and divisor values if BL31 is run on top of coreboot.
Moving the BL2-to-BL31 parameter parsing as early as possible to ensure
that the console is available for all following code.

Also update the Rockchip platform to use MULTI_CONSOLE_API.

Change-Id: I670d350fa2f8b8133539f91ac14977ab47db60d9
Signed-off-by: Julius Werner <jwerner@chromium.org>

The current Rockchip platform code retains the "common" default panic
handler which simply hangs the system (until the watchdog kicks in, if
enabled). This is usually not a great user experience.

This patch implements a Rockchip-specific panic handler that calls the
platform's reboot implementation to reset the system.

Change-Id: I4cbe09c48f1b3f86ebdfc0108c186565f9ffc119
Signed-off-by: Julius Werner <jwerner@chromium.org>

This includes the stdint header to declare the various types used within
the file, preventing build errors with recent GCC versions.

Change-Id: I9e7e92bb31deb58d4ff2732067dd88b53124bcc9
Signed-off-by: Paul Kocialkowski <contact@paulk.fr>

Pre-v8.2 platforms such as the Juno platform does not have
the Scalable Vector Extensions implemented and so the build
option ENABLE_SVE is set to zero.

This has a minor performance improvement with no functional
impact.

Change-Id: Ib072735db7a0247406f8b60e325b7e28b1e04ad1
Signed-off-by: David Cunado <david.cunado@arm.com>

When defining different sections in linker scripts it is needed to align
them to multiples of the page size. In most linker scripts this is done
by aligning to the hardcoded value 4096 instead of PAGE_SIZE.

This may be confusing when taking a look at all the codebase, as 4096
is used in some parts that aren't meant to be a multiple of the page
size.

Change-Id: I36c6f461c7782437a58d13d37ec8b822a1663ec1
Signed-off-by: Antonio Nino Diaz <antonio.ninodiaz@arm.com>

Some platforms (for ex. UniPhier) want to create files in the very
top of the build directory.  Add ${BUILD_PLAT} so such files can
depend on it.

Make existing directory targets depend on ${BUILD_PLAT} because
they are sub-directories of ${BUILD_PLAT}.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>

These macros are only defined for corresponding image,
and they are undefined for other images. It means that we have
to use ifdef or defined() instead of relying on being 0 by default.

Change-Id: Iad11efab9830ddf471599b46286e1c56581ef5a7
Signed-off-by: Roberto Vargas <roberto.vargas@arm.com>

when shutdown logic power rail, the some sgrf register
value will reset, so need to reinitilize secure.

Change-Id: I8ad0570432e54441fe1c60dd2960a81fd58f7163
Signed-off-by: Lin Huang <hl@rock-chips.com>

we will use timer in pmusarm, when logic power rail shutdown,
the secure timer will gone, so need to initial it in pmusram.

Change-Id: I472e7eec3fc197f56223e6fff9167556c1c5e3bc
Signed-off-by: Lin Huang <hl@rock-chips.com>

we do not have enough pmusram space now, so use slice1 to restore
ddr slice1 ~ slice4, that's will save more pmusram space.

Change-Id: Id54a7944f33d01a8f244cee6a8a0707bfe4d42da
Signed-off-by: Lin Huang <hl@rock-chips.com>

Change-Id: Ia59adf48cf14eb627721264765bce50cb31065ef
Signed-off-by: Lin Huang <hl@rock-chips.com>

pd_alive control cru, grf, timer, gpio and wdt, when
turn off logic power rail, these register value will
back to reset value, we need to save them value in suspend
and restore them when resuem, since timer will reinitial
in kernel, so it not need to save/restore.

Change-Id: I0fc2a011d3cdc04b66ffbf728e769eb28b51ee38
Signed-off-by: Lin Huang <hl@rock-chips.com>

when logic power rail shutdown, CRU register will back to reset
value, ddr use abpll as clock source when do suspend, we need to save
and dpll value in pmusram, then set back these ddr clock back to dpll
when dddr resume.

Change-Id: I95dc0173649e8515859cfa46b40a606e0cc2fe3f
Signed-off-by: Lin Huang <hl@rock-chips.com>

when shutdown logic power rail, the uart register value will reset,
so need to reinitilize debug uart.

Change-Id: I48d3535c0068fd671dea6ea32e908612992faf62
Signed-off-by: Lin Huang <hl@rock-chips.com>
Signed-off-by: Caesar Wang <wxt@rock-chips.com>

Assembler programmers are used to being able to define functions with a
specific aligment with a pattern like this:

    .align X
  myfunction:

However, this pattern is subtly broken when instead of a direct label
like 'myfunction:', you use the 'func myfunction' macro that's standard
in Trusted Firmware. Since the func macro declares a new section for the
function, the .align directive written above it actually applies to the
*previous* section in the assembly file, and the function it was
supposed to apply to is linked with default alignment.

An extreme case can be seen in Rockchip's plat_helpers.S which contains
this code:

  [...]
  endfunc plat_crash_console_putc

  .align 16
  func platform_cpu_warmboot
  [...]

This assembles into the following plat_helpers.o:

  Sections:
  Idx Name                             Size  [...]  Algn
   9 .text.plat_crash_console_putc 00010000  [...]  2**16
  10 .text.platform_cpu_warmboot   00000080  [...]  2**3

As can be seen, the *previous* function actually got the alignment
constraint, and it is also 64KB big even though it contains only two
instructions, because the .align directive at the end of its section
forces the assembler to insert a giant sled of NOPs. The function we
actually wanted to align has the default constraint. This code only
works at all because the linker just happens to put the two functions
right behind each other when linking the final image, and since the end
of plat_crash_console_putc is aligned the start of platform_cpu_warmboot
will also be. But it still wastes almost 64KB of image space
unnecessarily, and it will break under certain circumstances (e.g. if
the plat_crash_console_putc function becomes unused and its section gets
garbage-collected out).

There's no real way to fix this with the existing func macro. Code like

 func myfunc
 .align X

happens to do the right thing, but is still not really correct code
(because the function label is inserted before the .align directive, so
the assembler is technically allowed to insert padding at the beginning
of the function which would then get executed as instructions if the
function was called). Therefore, this patch adds a new parameter with a
default value to the func macro that allows overriding its alignment.

Also fix up all existing instances of this dangerous antipattern.

Change-Id: I5696a07e2fde896f21e0e83644c95b7b6ac79a10
Signed-off-by: Julius Werner <jwerner@chromium.org>

This fix modifies the order of system includes to meet the ARM TF coding
standard. There are some exceptions to this change in order to retain
header groupings and where there are headers within #if statements.

Change-Id: Ib5b668c992d817cc860e97b29e16ef106d17e404
Signed-off-by: Isla Mitchell <isla.mitchell@arm.com>

Decrypt device private keys which transfer from kernel, then stuff it to
DP controller. So that DP driver could start HDCP authentication in
kernel.

Change-Id: If3c2cd99bca811fe5fc30acc88bf5dc1afd7416d
Signed-off-by: Ziyuan Xu <xzy.xu@rock-chips.com>
Signed-off-by: Caesar Wang <wxt@rock-chips.com>

This patch fixes the two things as follows:

1) rk3399_flash_l2_b" seems to be a typo. That's "flush", not "flash".

2) fixes the warnings log.
We always hit the warnings thing during the suspend, as below log:
..
[   51.022334] CPU5: shutdown
[   51.025069] psci: CPU5 killed.
INFO:    sdram_params->ddr_freq = 928000000
WARNING: rk3399_flash_l2_b:reg 28830380,wait

When the L2 completes the clean and invalidate sequence, it asserts the
L2FLUSHDONE signal. The SoC can now deassert L2FLUSHREQ signal and then
the L2 deasserts L2FLUSHDONE.

Then, a loop without a delay isn't really great to measure time. We should
probably add a udelay(10) or so in there and then maybe replace the WARN()
after the loop. In the actual tests, the L2 cache will take ~4ms by
default for big cluster.

In the real world that give 10ms for the enough margin, like the
ddr/cpu/cci frequency and other factors that will affect it.

Change-Id: I55788c897be232bf72e8c7b0e10cf9b06f7aa50d
Signed-off-by: Caesar Wang <wxt@rock-chips.com>

For rk3399, the L2ACTLR[14] is 0 by default, as ACE CCI-500 doesn't
support WriteEvict. and you will hit the condition L2ACTLR[3] with 0,
as the Evict transactions should propagate to CCI-500 since it has
snoop filters.

Maybe this erratum applies to all Cortex-A53 cores so far, especially
if RK3399's A53 is a r0p4. we should enable it to avoid data corruption,

Change-Id: Ib86933f1fc84f8919c8e43dac41af60fd0c3ce2f
Signed-off-by: Caesar Wang <wxt@rock-chips.com>

For some reason, HDCP key decrytion can't open source in ATF, so we
build it as hdcp.bin. Besides declare the handler for decrypting.

Change-Id: Ia67ff2442ab43cb3ee4875b3d59cc1608e854b4b
Signed-off-by: Ziyuan Xu <xzy.xu@rock-chips.com>
Signed-off-by: Caesar Wang <wxt@rock-chips.com>