Some devices using the MVTWSI driver have the option to run at speeds
faster than Standard Mode (100kHZ). On the Armada 38x controllers, this
is actually necessary, since due to erratum FE-8471889, a timing
violation concerning repeated starts prevents the controller from
working correctly in Standard Mode. One of the workarounds recommended
in the erratum is to set the bus to Fast Mode (400kHZ) operation and
ensure all connected devices are set to Fast Mode.
In the current version of the driver, however, the delay times are
hard-coded to 10ms, corresponding to Standard Mode operation. To take
full advantage of the faster modes, we would need to either keep the
currently configured I2C speed in a globally accessible variable, or
pass it to the necessary functions as a parameter. For DM, the first
option is not a problem, and we can simply keep the speed in the private
data of the driver. For the legacy interface, however, we would need to
introduce a static variable, which would cause problems with boots from
NOR flashes; see commit d6b7757 "i2c: mvtwsi: Eliminate
twsi_control_flags."
As to not clutter the interface with yet another parameter, we therefore
keep the default 10ms delays for the legacy functions.
In DM mode, we make the delay time dependant on the frequency to allow
taking full advantage of faster modes of operation (tested with up to
1MHZ frequency on Armada MV88F6820).
Signed-off-by: Mario Six <mario.six@gdsys.cc> Reviewed-by: Stefan Roese <sr@denx.de>
Zero-length offsets are not properly handled by the driver. When a read
operation with a zero-length offset is started, a START condition is
asserted, and since no offset bytes are transferred, a repeated START is
issued immediately after, which confuses the controller.
To fix this, we send the first START only if any address bytes need to
be sent, and keep track of the expected start status accordingly.
Signed-off-by: Mario Six <mario.six@gdsys.cc> Reviewed-by: Stefan Roese <sr@denx.de>
The length of the address parameter of the __twsi_i2c_read and
__twsi_i2c_write functions is fixed to four bytes.
As a final step in the preparation of the DM conversion, we make the
length of this parameter variable by turning it into an array of bytes,
and convert the 32 bit value that's passed to the legacy functions into
a four-byte-array on the fly.
Signed-off-by: Mario Six <mario.six@gdsys.cc> Reviewed-by: Stefan Roese <sr@denx.de>
To be able to use the compatibility layer from the DM functions, we
factor the adap parameter out of all functions, and pass the actual
register base instead.
Signed-off-by: Mario Six <mario.six@gdsys.cc> Reviewed-by: Stefan Roese <sr@denx.de>
Since some additional parameters will be added in the course of this
patch series (especially with the addition of DM support), we replace
the longer "unsigned int" declarations with "uint" declarations to keep
the parameter lists more readable.
Signed-off-by: Mario Six <mario.six@gdsys.cc> Reviewed-by: Stefan Roese <sr@denx.de>
The twsi_stop function contains a parameter "status," which is used to
pass in the current exit status of the function calling twsi_stop, and
either return this status unchanged if it indicates an error, or return
twsi_stop's exit status if it does not indicate an error.
While not massively complicated, this adds another purpose to the
twsi_stop function, which should have the sole purpose of asserting a
STOP condition on the bus (and not manage the exit status of its
caller).
Therefore, we move the exit status management into the caller functions
by introducing a "stop_status" variable and returning either the status
before the twsi_stop call (kept in the "status" variable), or the status
from the twsi_stop call, depending on which indicates an error.
Signed-off-by: Mario Six <mario.six@gdsys.cc> Reviewed-by: Stefan Roese <sr@denx.de>
Due to breaking boots from NOR flashes, commit d6b7757 ("i2c: mvtwsi:
Eliminate twsi_control_flags") removed the static global
twsi_control_flags variable, which kept a set of default flags that were
always or'd to the control register when writing. It was replaced with a
flags parameter, which was passed around between the functions that
needed it.
Since the twsi_control_flags variable was used just for the purposes of
a) setting the MVTWSI_CONTROL_TWSIEN on every control register write,
and
b) setting the MVTWSI_CONTROL_ACK from twsi_i2c_read if needed,
anyway, the added overhead of another variable being passed around is no
longer justified, and we are better off implementing this flag setting
logic locally in the functions that actually write to the control
register.
Therefore, this patch sets MVTWSI_CONTROL_TWSIEN on every control
register write, replaces the twsi_i2c_read's flags parameter with a
ack_flag parameter, which tells the function whether to acknowledge the
read or not, and removes every other instance of the flags variable.
This has the added benefit that now every notion of "global default
flags" is gone, and it's much easier to see which control flags are
actually set at which point in time.
Signed-off-by: Mario Six <mario.six@gdsys.cc> Reviewed-by: Stefan Roese <sr@denx.de>
i2c: mvtwsi: Streamline code and add documentation
Convert groups of logically connected preprocessor defines into proper
enums, one macro into an inline function, and add documentation
to/extend existing documentation of these items.
Signed-off-by: Mario Six <mario.six@gdsys.cc> Reviewed-by: Stefan Roese <sr@denx.de>
Mugunthan V N [Mon, 18 Jul 2016 09:41:02 +0000 (15:11 +0530)]
defconfig: am335x_boneblack_vboot: enable i2c driver model
Enable i2c driver model for am335x_boneblack_vboot as omap i2c
supports driver model. Also enable CONFIG_DM_I2C_COMPAT for
legacy drivers of i2c devices.
Signed-off-by: Mugunthan V N <mugunthanvnm@ti.com> Reviewed-by: Tom Rini <trini@konsulko.com>
John Keeping [Fri, 15 Jul 2016 16:33:23 +0000 (17:33 +0100)]
rockchip: sdram: Fix register layout for Linux
The ChromeOS kernel reads the RAM settings from PMU_SYS_REG2 and expects
the bootloader to store the necessary information there. We're using
the same register to pass the same information between the SPL and
U-Boot but in a slightly different format.
Change this to use the format expected by the Linux DMC driver so that
the system doesn't hang in Linux by misconfiguring the RAM.
This is almost the same as commit b5788dc ("rockchip: rk3288: correct
sdram setting") which was reverted in commit b525556 ("Revert "rockchip:
rk3288: correct sdram setting"") but parenthese have been added to apply
the mask correctly when reading the "bw" setting and a couple of minor
style issues have been fixed to keep check_patch.pl happy.
Signed-off-by: John Keeping <john@metanate.com> Reviewed-by: Tom Rini <trini@konsulko.com> Acked-by: Simon Glass <sjg@chromium.org>
Kever Yang [Mon, 18 Jul 2016 09:00:58 +0000 (17:00 +0800)]
mmc: rockchip: add SDHCI driver support for rockchip soc
Rockchip rk3399 using arasan sdhci-5.1 controller.
This patch add the controller support to enable mmc device
with full driver-model support, tested on rk3399 evb board.
According to my test result, this driver should be OK,
the command "part list mmc 0" can result in a right output,
but all the mmc command failed like this:
=> mmc info
No MMC device available
Command failed, result=1
The result of get_mmc_num in cmd/mmc.c is always 0?
Signed-off-by: Kever Yang <kever.yang@rock-chips.com> Acked-by: Simon Glass <sjg@chromium.org>
Kever Yang [Tue, 19 Jul 2016 13:16:59 +0000 (21:16 +0800)]
ARM64: rockchip: add support for rk3399 SoC based evb
RK3399 is a SoC from Rockchip with dual-core Cortex-A72
and quad-core Cortex-A53 CPU. It supports two USB3.0
type-C ports and two USB2.0 EHCI ports. Other interfaces
are very much like RK3288, the DRAM are 32bit width address
and support address from 0 to 4GB-128MB range.
Signed-off-by: Kever Yang <kever.yang@rock-chips.com> Acked-by: Simon Glass <sjg@chromium.org>
Kever Yang [Tue, 19 Jul 2016 13:16:58 +0000 (21:16 +0800)]
dts: add support for Rockchip rk3399 soc
These files are from kernel upstream:
"649a371 Add linux-next specific files for 20160616"
with some modification need by U-Boot:
- chosen with stdout-path to uart2.
- add clock-frequency for uart2
Signed-off-by: Kever Yang <kever.yang@rock-chips.com> Acked-by: Simon Glass <sjg@chromium.org>
Simon Glass [Sun, 17 Jul 2016 21:23:17 +0000 (15:23 -0600)]
rockchip: Use rockchip_get_clk() to obtain the SoC clock
The current code picks the first available clock. In U-Boot proper this is
the oscillator device, not the SoC clock device. As a result the HDMI display
does not work.
Fix this by calling rockchip_get_clk() instead.
Fixes: 135aa950 (clk: convert API to match reset/mailbox style) Signed-off-by: Simon Glass <sjg@chromium.org> Acked-by: Anatolij Gustschin <agust@denx.de>
Simon Glass [Sun, 17 Jul 2016 21:23:16 +0000 (15:23 -0600)]
rockchip: Add a way to obtain the main clock device
On Rockchip SoCs we typically have a main clock device that uses the Soc
clock driver. There is also a fixed clock for the oscillator. Add a function
to obtain the core clock.
Simon Glass [Sun, 17 Jul 2016 21:23:15 +0000 (15:23 -0600)]
dm: core: Add a way to find a device by its driver
Some SoCs have a single clock device. Provide a way to find it given its
driver name. This is handled by the linker so will fail if the name is not
found, avoiding strange errors when names change and do not match. It is
also faster than a string comparison.
The function is very specific to the rk3288 in its arguments
referencing the rk3288 cru and grf and every other rockchip soc
has differing cru and grf registers. So make that function naming
explicit.
Signed-off-by: Heiko Stuebner <heiko@sntech.de> Acked-by: Simon Glass <sjg@chromium.org>
cosmetic: rockchip: sort socs according to numbers
Having some sort of ordering proofed helpful in a lot of other places
already. So for a larger number of rockchip socs it might be helpful
as well instead of an ever increasing unsorted list.
Signed-off-by: Heiko Stuebner <heiko@sntech.de> Reviewed-by: Andreas Färber <afaerber@suse.de> Acked-by: Simon Glass <sjg@chromium.org>
cosmetic: rockchip: rk3036: pinctrl: fix config symbol naming
Rockchip socs are always named rkxxxx in all places, as also shown
by the naming of the rk3036 pinctrl file itself.
Therefore also name the config symbol according to this scheme.
Signed-off-by: Heiko Stuebner <heiko@sntech.de> Acked-by: Simon Glass <sjg@chromium.org>
This path doesn't support the fastboot flash function command entirely.
We will hit "cannot find partition" assertion without specified
partition environment. Define gpt partition layout in specified board
such as firefly-rk3288, then enjoy it!
Signed-off-by: Ziyuan Xu <xzy.xu@rock-chips.com> Acked-by: Simon Glass <sjg@chromium.org>
Invalidate dcache before starting the DMA to ensure coherency. In case
there are any dirty lines from the DMA buffer in the cache, subsequent
cache-line replacements may corrupt the buffer in memory while the DMA
is still going on. Cache-line replacement can happen if the CPU tries to
bring some other memory locations into the cache while the DMA is going
on.
Signed-off-by: Ziyuan Xu <xzy.xu@rock-chips.com> Acked-by: Simon Glass <sjg@chromium.org>
So far, Rockchip SoCs have two kinds of USB2.0 phy, such as Synopsys and
Innosilicon. This patch applys dwc2 usb driver framework to implement
phy_init() and phy_off() methods for Synopsys phy on Rockchip platform.
Signed-off-by: Ziyuan Xu <xzy.xu@rock-chips.com> Acked-by: Simon Glass <sjg@chromium.org>
Kever Yang [Fri, 8 Jul 2016 03:30:58 +0000 (11:30 +0800)]
board: move all the rockchip board in one folder
The 'evb_rk3036' and 'kylin' is not a vendor name, let's replace them
to 'rockchip' which is a real _vendor_ name, and meet the architecure
'board/<vendor>/<board-name>/'.
More boards from rockchip like evb_rk3288, evb_rk3399 will comes later.
Signed-off-by: Kever Yang <kever.yang@rock-chips.com> Reviewed-by: Eddie Cai <eddie.cai.kernel@gmail.com>
evb-3288 board RK3288-based development board with 2 USB ports, HDMI,
VGA, micro-SD card, audio, WiFi and Gigabit Ethernet. It also includes
on-board 8G eMMC and 2GB of SDRAM. Expansion connector provide access to
display pins, I2C, SPI, UART and GPIOs. This add some basic files
required to allow the board to output serial messaged and can run
command(mmc info etc).
evb-rk3288 also supports booting from eMMC or SD card, the default is eMMC.
Signed-off-by: Ziyuan Xu <xzy.xu@rock-chips.com> Reviewed-by: Simon Glass <sjg@chromium.org>
rockchip: add option to change method of loading u-boot
If we would like to boot from SD card, we have to implement mmc driver
in SPL stage, and get a slightly large SPL binary. Rockchip SoC's
bootrom code has the ability to load spl and u-boot, then boot.
If CONFIG_ROCKCHIP_SPL_BACK_TO_BROM is enabled, the spl will return to
bootrom in board_init_f(), then bootrom loads u-boot binary.
Signed-off-by: Ziyuan Xu <xzy.xu@rock-chips.com> Acked-by: Simon Glass <sjg@chromium.org>
Fixed up spelling of U-Boot, boorom, opinion->option, Rochchip: Signed-off-by: Simon Glass <sjg@chromium.org>
Tom Rini [Mon, 25 Jul 2016 22:18:15 +0000 (18:18 -0400)]
sandbox: Migrate CONFIG_I2C_EEPROM
Most users of CONFIG_I2C_EEPROM were migrated to defconfig a while ago,
but sandbox was skipped. Leave it off for sandbox_spl where it does not
build, but does not need to be either.
Cc: Simon Glass <sjg@chromium.org> Signed-off-by: Tom Rini <trini@konsulko.com>
Simon Glass [Fri, 22 Jul 2016 15:22:49 +0000 (09:22 -0600)]
dtoc: Correct the type widening code in fdt_fallback
This code does not match the fdt version in fdt.py. When dtoc is unable to
use the Python libfdt library, it uses the fallback version, which does not
widen arrays correctly.
Fix this to avoid a warning 'excess elements in array initialize' in
dt-platdata.c which happens on some platforms.
Reported-by: Tom Rini <trini@konsulko.com> Signed-off-by: Simon Glass <sjg@chromium.org> Tested-by: Tom Rini <trini@konsulko.com>
tools, rsa: Further minor cleanups on top of c236ebd and 2b9ec7
[NOTE: I took v1 of these patches in, and then v2 came out, this commit
is squashing the minor deltas from v1 -> v2 of updates to c236ebd and 2b9ec76 into this commit - trini]
- Added an additional NULL check, as suggested by Simon Glass to
fit_image_process_sig
- Re-formatted the comment blocks
Signed-off-by: Mario Six <mario.six@gdsys.cc> Reviewed-by: Simon Glass <sjg@chromium.org>
[For merging the chnages from v2 back onto v1] Signed-off-by: Tom Rini <trini@konsulko.com>
ARM: am33xx: Always inhibit init/refresh during DDR phy init
A couple of commits have modified the am33xx/am437x ddr2/ddr3
initialization path to fix certain issues, but have had the side effect
of causing L3 noc errors during initialization. The two commits are:
The EMIF_REG_INITREF_DIS_MASK bit still needs to be set for all
platforms. This delays initialization and refresh until a later stage.
The 500us timer can be programmed for platforms that require it
and for platforms that don't require it. It is currently hardcoded
for 400MHz systems. For systems with a higher memory frequency
this needs to be a larger value, and for systems with a lower
memory frequency this can be a lower value. This can be
considered a separate issue and corrected in a later commit.
Signed-off-by: Russ Dill <Russ.Dill@ti.com> Reviewed-by: Tom Rini <trini@konsulko.com>
In certain circumstances it comes in handy to be able to boot into a second
U-Boot. But as of now it is not possible to boot a U-Boot binary that is inside
a FIT image, which is problematic for projects that e.g. need to guarantee a
unbroken chain of trust from SOC all the way into the OS, since the FIT signing
mechanism cannot be used.
This patch adds the capability to load such FIT images.
An example .its snippet (utilizing signature verification) might look
like the following:
commit 1f807a9f32aa ("ARM: keystone2: Refactor MSMC macros to avoid
left under a macro KS2_MSMC_SEGMENT_QM_PDSP which is no longer valid.
This, in effect disabled DMA coherency for QM PDSP.
Given that msmc_k2hkle_common_setup is valid for all K2H/K/L/E SoCs,
the #ifdef should been removed in the first place. Do the same.
A feasible way to communicate certain errors for devices that have no
other way of signalling besides LEDs is to flash these LEDs. For errors
in U-Boot, a script that utilizes the led and sleep commands would be a
practicable way, but currently the sleep command can only delay for an
integral amount of seconds, which is too slow to create an easily
noticeable pattern for flashing LEDs.
Therefore, this patch adds support for fractions (down to .001 seconds)
to the sleep command.
The parsing is kept minimal, simplistic and as robust as possible: After
converting the passed string using simple_strtoul and multiplying it
with 1000, we search for the first dot, convert the three characters
after that into a number (if they are not numbers, we ignore the
fractional part and just use the delay we got from simple_strtoul), and
add this number to the delay.
mtd: nand: fix bug writing 1 byte less than page size
nand_do_write_ops() determines if it is writing a partial page with the
formula:
part_pagewr = (column || writelen < (mtd->writesize - 1))
When 'writelen' is exactly 1 byte less than the NAND page size the formula
equates to zero, so the code doesn't process it as a partial write, although
it should.
As a consequence the function remains in the while(1) loop with 'writelen'
becoming 0xffffffff and iterating until the watchdog timeout triggers.
To reproduce the issue on a NAND with 2K page (0x800):
=> nand erase.part <partition>
=> nand write $loadaddr <partition> 7ff
Brian Norris [Wed, 15 Jun 2016 19:09:22 +0000 (21:09 +0200)]
mtd: nand: add common DT init code
These are already-documented common bindings for NAND chips. Let's
handle them in nand_base.
If NAND controller drivers need to act on this data before bringing up
the NAND chip (e.g., fill out ECC callback functions, change HW modes,
etc.), then they can do so between calling nand_scan_ident() and
nand_scan_tail().
The original commit has been slightly reworked to use the fdtdec_xxx()
helpers (instead of the of_xxxx() ones).
Signed-off-by: Brian Norris <computersforpeace@gmail.com> Signed-off-by: Boris Brezillon <boris.brezillon@free-electrons.com> Acked-by: Hans de Goede <hdegoede@redhat.com>
Boris Brezillon [Wed, 15 Jun 2016 08:42:18 +0000 (10:42 +0200)]
cmd, nand: add an option to disable the verification when writing in raw mode
Modern NANDs do not guarantee that data written in raw mode will not
contain bitflips just after writing them. This is fine since the number
of bitflips should be rather low and thus fixable by the ECC engine,
but since we are reading data in raw mode to verify if they match the
input data we cannot prevent failures if some bits are flipped.
The option of using standard mode to verify the data is not acceptable
either, since one of the usage of raw mode is to allow flashing images
that do not respect the standard NAND page layout or the default ECC
config (this is the case on Allwinner platforms, where the ROM code
tests several hardcoded configs, which are not necessarily matching the
NAND characteristics).
Add an extension to the nand write.raw command allowing one to disable
the verification step.
Signed-off-by: Boris Brezillon <boris.brezillon@free-electrons.com> Reviewed-by: Tom Rini <trini@konsulko.com>
ARM: uniphier: use (devm_)ioremap() instead of map_sysmem()
This does not have much impact on behavior, but makes code look more
more like Linux. The use of devm_ioremap() often helps to delete
.remove callbacks entirely.
ARM: dts: uniphier: renumber serial aliases for Gentil/Vodka boards
On these two boards, the serial0 is used for inter-chip connection,
so cannot be used for login console. The serial2 is used instead
for them, but it is tedious to use because upper level deployment
projects must switch login console per board.
arm: omap5: fix build dependency for secure devices
Commit 17c2987 introduces an undesired dependency on CONFIG_SPL_LOAD_FIT
when building U-Boot for AM57xx and DRA7xx high-security (HS) devices that
causes the build to break when that option is not active. Fix this issue
by only building the u-boot_HS.img target when building U-Boot into an
actual FIT image.
Signed-off-by: Andreas Dannenberg <dannenberg@ti.com> Reviewed-by: Tom Rini <trini@konsulko.com>
arm: am4x: fix build dependency for secure devices
Commit e29878f introduces an undesired dependency on CONFIG_SPL_LOAD_FIT
when building U-Boot for AM43xx high-security (HS) devices that causes the
build to break when that option is not active. Fix this issue by only
building the u-boot_HS.img target when building U-Boot into an actual
FIT image.
Signed-off-by: Andreas Dannenberg <dannenberg@ti.com> Reviewed-by: Tom Rini <trini@konsulko.com>
spl: fit: Fix the number of bytes read in raw mode
In raw mode a full sector is to be read even if image covers part of
a sector. Number of sectors are calculated as ROUND_UP(size)/sec_size by FIT
framework. This calculation assumes that image is at the 0th offset of a sector,
which is not true always in FIT case. So, include the image offset while
calculating number of sectors.
Signed-off-by: Lokesh Vutla <lokeshvutla@ti.com> Reviewed-by: Tom Rini <trini@konsulko.com>
When signing images, we repeatedly call fit_add_file_data() with
successively increasing size values to include the keys in the DTB.
Unfortunately, if large keys are used (such as 4096 bit RSA keys), this
process fails sometimes, and mkimage needs to be called repeatedly to
integrate the keys into the DTB.
This is because fit_add_file_data actually returns the wrong error
code, and the loop terminates prematurely, instead of trying again with
a larger size value.
This patch corrects the return value by fixing the return value of
fdt_add_bignum, fixes a case where an error is masked by a unconditional
setting of a return value variable, and also removes a error message,
which is misleading, since we actually allow the function to fail. A
(hopefully helpful) comment is also added to explain the lack of error
message.
This is probably related to 1152a05 ("tools: Correct error handling in
fit_image_process_hash()") and the corresponding error reported here:
When signing images, we repeatedly call fit_add_file_data() with
successively increasing size values to include the keys in the DTB.
Unfortunately, if large keys are used (such as 4096 bit RSA keys), this
process fails sometimes, and mkimage needs to be called repeatedly to
integrate the keys into the DTB.
This is because fit_add_file_data actually returns the wrong error
code, and the loop terminates prematurely, instead of trying again with
a larger size value.
This patch corrects the return value and also removes a error message,
which is misleading, since we actually allow the function to fail. A
(hopefully helpful) comment is also added to explain the lack of error
message.
This is probably related to 1152a05 ("tools: Correct error handling in
fit_image_process_hash()") and the corresponding error reported here:
Stephen Warren [Mon, 18 Jul 2016 16:07:25 +0000 (10:07 -0600)]
test/py: use absolute dts path in vboot test
Without this, the test fails if the test is run with a cwd other than the
root of the U-Boot source tree.
Fixes: 8729d582595d ("test: Convert the vboot test to test/py") Signed-off-by: Stephen Warren <swarren@nvidia.com> Reviewed-by: Simon Glass <sjg@chromium.org>
Jeremy Hunt [Mon, 18 Jul 2016 16:01:02 +0000 (12:01 -0400)]
armv8: spl: Call board_init_r from crt0_64 in SPL
As part of the startup process for boards using the SPL, the
meaning of board_init_f changed such that it should return normally
rather than calling board_init_r directly. (see db910353a126d84fe8dff7a694ea792f50fcfb6a )
This was fixed in 32-bit arm, but broke when SPL was added to
64 bit arm. This fixes crt0_64 so that it calls board_init_r
during the SPL and removes the direct call from board_init_f
from the arm SPL example.
Signed-off-by: Jeremy Hunt <Jeremy.Hunt@DEShawResearch.com> Acked-by: Simon Glass <sjg@chromium.org>
projects / u-boot / log