Patrice Chotard [Wed, 15 Nov 2017 12:14:52 +0000 (13:14 +0100)]
clk: clk_stm32fx: add clock configuration for mmc usage
MMC block needs 48Mhz source clock, for that we choose
to select the SAI PLL.
Update also stm32_clock_get_rate() to retrieve the MMC
clock source needed in MMC driver.
STM32F4 uses a different RCC variant than STM32F7. For STM32F4
sdmmc clocks bit are located into dckcfgr register whereas there
are located into dckcfgr2 registers on STM32F7.
In both registers, bits CK48MSEL and SDMMC1SEL are located at
the same position.
Patrice Chotard [Wed, 15 Nov 2017 12:14:51 +0000 (13:14 +0100)]
dm: misc: bind STM32F4/F7 clock from rcc MFD driver
Like STM32H7, now STM32F4/F7 clock drivers are binded by
MFD stm32_rcc driver.
This also allows to add reset support to STM32F4/F7 SoCs family.
As Reset driver is not part of SPL supported drivers, don't bind it
in case of SPL to avoid that stm32_rcc_bind() returns an error.
Patrice Chotard [Wed, 15 Nov 2017 12:14:47 +0000 (13:14 +0100)]
clk: stm32f7: add STM32F4 support
STM32F4 and STM32F7 RCC clock IP are very similar.
Same driver can be used to managed RCC clock for
these 2 SoCs.
Differences between STM32F4 and F7 will be managed using
different compatible string :
_ overdrive clock is only supported by STM32F7
_ different sys_pll_psc parameters can be used between STM32F4
and STM32F7.
Patrice Chotard [Wed, 15 Nov 2017 12:14:46 +0000 (13:14 +0100)]
ARM: DTS: stm32: update rcc compatible for STM32F746
Align the RCC compatible string with the one used by kernel.
It will allow to use the same clock driver for STM32F4
and STM32F7 and to manage the differences between the 2 SoCs
Add a dedicated stm32f7 compatible string to use clk_stm32f7
driver with both STM32F4 and STM32F7 SoCs.
It will be needed to manage differences between these 2 SoCs.
tools: env: Add support for direct read/write UBI volumes
Up to now we were able to read/write environment data from/to UBI
volumes only indirectly by gluebi driver. This driver creates NAND MTD
on top of UBI volumes, which is quite a workaroung for this use case.
Add support for direct read/write UBI volumes in order to not use
obsolete gluebi driver.
Forward-ported from this patch:
http://patchwork.ozlabs.org/patch/619305/
Rick Chen [Thu, 23 Nov 2017 06:19:36 +0000 (14:19 +0800)]
spi: nds_ae3xx: Rename nds_ae3xx_spi as atcspi200_spi
atcspi200 is Andestech spi ip which is embedded in
AE3XX and AE250 platforms. So rename as atcspi200
will be more reasonable to be used in different
platforms.
Rick Chen [Thu, 23 Nov 2017 04:48:46 +0000 (12:48 +0800)]
atcpit100: timer: Remove arch dependency.
ATCPIT100 is often used in AE3XX platform which is
based on NDS32 architecture recently. But in the future
Andestech will have AE250 platform which is embeded
ATCPIT100 timer based on RISCV architecture.
Rick Chen [Thu, 23 Nov 2017 03:04:34 +0000 (11:04 +0800)]
ae3xx: timer: Fix ae3xx timer work abnormal in 64 bit.
It will be work fine with unsigned long declaretion in timer
register struct when system is 32 bit. But it will not work
well when system is 64 bit. Replace it by u32 and verify both
ok in 32/64 bit.
Marek Vasut [Thu, 9 Nov 2017 20:56:01 +0000 (21:56 +0100)]
ARM: rmobile: Use PRR driver on all Gen3 boards
Mark the PRR as u-boot,dm-pre-reloc in all Gen3 board DTs as it is
needed very early and turn on the CONFIG_SYSCON to allow the PRR
driver to bind as a syscon uclass.
Marek Vasut [Thu, 9 Nov 2017 20:49:48 +0000 (21:49 +0100)]
ARM: rmobile: Convert PRR to DM and OF control
Implement DM driver for the Renesas PRR into RCar cpu info and convert
all users with DM and OF enabled to this new driver. This means all of
the boards with DM and OF enabled can fetch PRR address from DT, which
is useful on ie. V3M which has different PRR address than the rest of
Gen3 SoCs.
Marek Vasut [Mon, 27 Nov 2017 04:45:46 +0000 (05:45 +0100)]
ARM: rmobile: Zap checkboard on Gen3
The checkboard() function showing hard-coded board model for which the
U-Boot was built is superseded on Gen3 by show_board_info() displaying
the Model from device tree. Add small ifdef to stop compiling the
function into U-Boot.
Marek Vasut [Sat, 25 Nov 2017 22:24:01 +0000 (23:24 +0100)]
ARM: rmobile: Unify memory map for RCar Gen3
Unify the R7A7795 and R8A7796 memory maps in memmap-gen3 and, for now,
select which one is used based on which SoC is selected. Since this is
done in C code instead of statically assigned now, the decision can be
taken by PRR SoC match as well, which will be done in a subsequent patch.
Marek Vasut [Sat, 25 Nov 2017 22:54:10 +0000 (23:54 +0100)]
ARM: rmobile: Add PRR CPU ID macros
Replace the ad-hoc values in the PRR CPU ID table with macros,
so that users can use rmobile_get_cpu_type() can compare the
returned value with these macros to figure out on which CPU they
are running.
Marek Vasut [Sat, 25 Nov 2017 21:43:57 +0000 (22:43 +0100)]
ARM: rmobile: Stop using rcar-common/common.c on Gen3
Since the Gen3 clock driver now has a .remove callback, it is no
longer necessary to shut the clock down before booting Linux in the
arch_preboot_os hook. Stop using it and while doing so, remove all
the ad-hoc config options which this hook used.
Marek Vasut [Tue, 28 Nov 2017 07:25:21 +0000 (08:25 +0100)]
ARM: rmobile: Clean up GIC macros
Pull out the GIC macros from the board configuration files
into the common Gen3 configuration file since these macros
are the same for all Gen3 systems.
Marek Vasut [Sat, 25 Nov 2017 20:15:36 +0000 (21:15 +0100)]
ARM: rmobile: Drop SDHI address macros from Gen3
Since the RCar Gen3 no longer uses the SH SDHI driver, but rather
uses the Matsushita SD driver, which loads all the properties from
device tree, these macros are no longer used, remove them.
Marek Vasut [Thu, 9 Nov 2017 21:49:19 +0000 (22:49 +0100)]
net: ravb: Fix reset GPIO handling
Fix handling of the reset GPIO. Drop the _nodev() suffix from the
gpio_request_by_name() call as there is now a proper DM capable
GPIO driver. Also check if the GPIO is valid before freeing it in
remove path, otherwise U-Boot will crash.
Marek Vasut [Tue, 28 Nov 2017 07:02:27 +0000 (08:02 +0100)]
i2c: rcar_iic: Add RCar IIC driver
Add driver for the RCar IIC or DVFS I2C controller. This driver is based
on the SH I2C driver, but supports DM and DT probing as well as modern
I2C framework API.
Marek Vasut [Sat, 25 Nov 2017 21:08:55 +0000 (22:08 +0100)]
clk: rmobile: Move preboot clock shutdown to the driver
The MSTP registers were poked in boards/renesas/rcar-common/common.c
in arch_preboot_os hook thus far to shut down the clock before Linux
takes over. With DM, this is no longer needed and we can do the same
in the clock driver .remove callback. This patch adds such a .remove
callback for R8A7795 and R8A7796.
Marek Vasut [Mon, 27 Nov 2017 04:32:42 +0000 (05:32 +0100)]
fdtdec: Support parsing multiple /memory nodes
It is legal to have multiple /memory nodes in a device tree . Currently,
fdtdec_setup_memory_size() only supports parsing the first node . This
patch extends the function such that if a particular /memory node does
no longer have further "reg" entries and CONFIG_NR_DRAM_BANKS still
allows for more DRAM banks, the code moves on to the next memory node
and checks it's "reg"s. This makes it possible to handle both systems
with single memory node with multiple entries and systems with multiple
memory nodes with single entry.
Tom Rini [Wed, 29 Nov 2017 13:26:07 +0000 (08:26 -0500)]
Merge tag 'xilinx-for-v2018.01' of git://www.denx.de/git/u-boot-microblaze
Xilinx changes for v2018.1
Zynq:
- Add support for Syzygy and cc108 boards
- Add support for mini u-boot configurations (cse)
- dts updates
- config/defconfig updates in connection to Kconfig changes
- Fix psu_init handling
ZynqMP:
- SPL fixes
- Remove slcr.c
- Fixing r5 startup sequence
- Add support for external pmufw
- Add support for new ZynqMP chips
- dts updates
- Add support for zcu102 rev1.0 board
Drivers:
- nand: Support external timing setting and board init
- ahci: Fix wording
- axi_emac: Wait for bit, non processor mode, readl/write conversion
- zynq_gem: Fix SGMII/PCS support
net: xilinx_axi_emac: Use readl and writel for io ops
This patch uses readl and writel instead of in_be32 and
out_be32 for io ops as these internally uses readl,
writel for microblaze and for Zynq, ZynqMP there is
no need of endianness conversion and readl, writel
should work straightaway. This patch starts supporting
the driver for Zynq and ZynqMP platforms.
net: zynq_gem: Dont enable SGMII and PCS selection
Dont enable SGMII and PCS selection if internal PCS/PMA
is not used, by getting the info about internal/external
PCS/PMA usage from dt property "is-internal-phy".
Michal Simek [Thu, 23 Nov 2017 07:25:41 +0000 (08:25 +0100)]
arm: zynq: Change Zynq/ZynqMP Kconfig description
Use more accurate description for Xilinx Zynq and ZynqMP based platforms.
With using driver model there shouldn't be a need to create separate
Kconfig config options.
tools: zynqmpimage: adjust ug1085 reference to v1.4 of the document
The chapter in which the table explaining the image format changed
chapter as the document evolved. This should help people track the
info down faster.
Joe Hershberger [Tue, 7 Nov 2017 02:16:10 +0000 (18:16 -0800)]
mtd: nand: zynq: Add support for the NAND lock/unlock operation
Zynq NAND driver is not support for NAND lock or unlock operation.
Hence, accidentally write into the critical NAND region might cause
data corruption to occur.
This commit is to add NAND lock/unlock command into NAND SMC register
set for NAND lock/unlock operaion.
Wilson Lee [Wed, 15 Nov 2017 09:14:35 +0000 (01:14 -0800)]
mtd: zynq: nand: Move board_nand_init() function to board.c
Putting board_nand_init() function inside NAND driver was not appropriate
due to it doesn't allow board vendor to customise their NAND
initialization code such as adding NAND lock/unlock code.
This commit was to move the board_nand_init() function from NAND driver
to board.c file. This allow customization of board_nand_init() function.
Michal Simek [Fri, 10 Nov 2017 12:28:07 +0000 (13:28 +0100)]
arm: zynq: Use unsigned type with comparison with ARRAY_SIZE
Sparse is return warning about this:
arch/arm/mach-zynq/slcr.c: In function 'zynq_slcr_get_mio_pin_status':
arch/arm/mach-zynq/slcr.c:185:16: warning: comparison between signed and
unsigned integer expressions [-Wsign-compare]
for (i = 0; i < ARRAY_SIZE(mio_periphs); i++) {
^
Michal Simek [Fri, 10 Nov 2017 08:47:28 +0000 (09:47 +0100)]
arm: zynq: Move common ps7_init* initialization to arch code
This patch is based on work done in topic board where the first address
word also storing operation which should be done. This is reducing size
of configuration data.
This patch is not breaking an option to copy default ps7_init_gpl* files
from hdf file but it is doing preparation for ps7_init* consolidation.
The patch is also marking ps7_config as weak function.
net: xilinx_axi_emac: Add support for non processor mode
Add support for non processor mode, this mode doesn't have
access to some of the registers and hence this patch
bypasses it and also length has to be calculated from
status instead of app4 in this mode.
Paul Burton [Tue, 21 Nov 2017 22:31:07 +0000 (14:31 -0800)]
boston: Add u-boot.mcs make target
U-Boot is generally flashed to a MIPS Boston development board by means
of a .mcs file which Xilinx Vivado software can write to the flash
present on the board. As such we'd generally want to produce an mcs file
when building U-Boot to target the Boston board. Introduce a make target
for u-boot.mcs which generates it using the srec_cat tool available from
the SRecord project, and build it by default when srec_cat is present.
Paul Burton [Tue, 21 Nov 2017 20:35:31 +0000 (12:35 -0800)]
boston: Set CONFIG_SYS_LOAD_ADDR to 0x88000000
Generally we load Linux kernels on Boston boards in the form of FIT
images containing a compressed kernel binary. Linux is linked at
0x80100000 and so we need to decompress the kernel binary to that
address, however this is our default load address which means that
unless explicitly avoided we hit a decompression error as the
uncompressed kernel binary overwrites its compressed version from the
FIT image.
Avoid this by adjusting CONFIG_SYS_LOAD_ADDR to 0x88000000 (or
0xffffffff88000000 for MIPS64 builds) which avoids the address overlap
between compressed & uncompressed kernel binaries.
Paul Burton [Tue, 21 Nov 2017 19:18:39 +0000 (11:18 -0800)]
MIPS: Break out of cache loops for unimplemented caches
If we run on a CPU which doesn't implement a particular cache then we
would previously get stuck in an infinite loop, executing a cache op on
the first "line" of the missing cache & then incrementing the address by
0. This was being avoided for the L2 caches, but not for the L1s. Fix
this by generalising the check for a zero line size & avoiding the cache
op loop when this is the case.
Paul Burton [Tue, 21 Nov 2017 19:18:38 +0000 (11:18 -0800)]
MIPS: Clear instruction hazards in flush_cache()
When writing code, for example during relocation, we ensure that the
icache has a coherent view of the new instructions with a call to
flush_cache(). This handles the bulk of the work to ensure the new
instructions will execute as expected, however it does not ensure that
the CPU pipeline doesn't already contain instructions taken from a stale
view of the affected memory. This could theoretically be a problem for
relocation, but in practice typically isn't because we sync caches for
enough code after the entry point of the newly written code that by the
time the CPU pipeline might possibly fetch any of it we'll have long ago
written it back & invalidated any stale icache entries. This is however
a problem for shorter regions of code.
In preparation for later patches which write shorter segments of code,
ensure any instruction hazards are cleared by flush_cache() by
introducing & using a new instruction_hazard_barrier() function which
makes use of the jr.hb instruction to clear the hazard.
Paul Burton [Tue, 21 Nov 2017 19:18:37 +0000 (11:18 -0800)]
MIPS: Ensure cache ops complete in cache maintenance functions
A typical use of cache maintenance functions is to force writeback of
data which a device is about to read using DMA - for example a
descriptor or command structure. Such users of cache maintenance
functions require that operations on the cache have completed before
they proceed to instruct a device to read memory. This requires that we
place a completion barrier (ie. sync instruction) between the cache ops
and whatever write informs the device to perform DMA.
Whilst strictly speaking this isn't all users of the cache maintenance
functions & we could instead place the barriers in the drivers that
require them, it would be much more invasive to do so than to just have
the barrier be the default by placing it in the cache functions
themselves. The cost is low enough that it shouldn't matter to us in any
rare cases that we use the cache functions when not performing DMA.
Paul Burton [Mon, 30 Oct 2017 23:58:21 +0000 (16:58 -0700)]
Update Paul Burton's email address
MIPS is no longer a part of Imagination Technologies, and as such my
@imgtec.com email address will soon cease to function. This patch
updates occurrances of it with my new @mips.com email address, and adds
an entry in .mailmap such that git (& tools such as get_maintainer.pl
when examining history) will use the new address.
Paul Burton [Fri, 15 Sep 2017 18:35:54 +0000 (11:35 -0700)]
MIPS: Drop unused PTR_COUNT_SHIFT from u-boot.lds
The u-boot.lds linker script for MIPS defines a PTR_COUNT_SHIFT macro to
2 or 3 for 32 bit or 64 bit builds respectively. This macro is never
actually used though, so remove the dead code.
Paul Burton [Fri, 15 Sep 2017 18:33:53 +0000 (11:33 -0700)]
boston: Drop unused return value
The boston lowlevel_init() function zeroes the return register v0,
despite the function not being expected to return a value & that value
never being used.