For the RK3399-Q7 we recommend storing SPL and u-boot.itb in the
on-module 32MBit (and sometimes even larger, if requested as part of a
configure-to-order configuration) SPI-NOR flash that is clocked for a
bitrate of 49.5MBit/s and connected in a single-IO configuration (the
RK3399 only supports single-IO for SPI).
Unfortunately, the existing SPI driver is excruciatingly slow at
reading out large chunks of data (in fact it is just as slow for small
chunks of data, but the overheads of the driver-framework make it less
noticeable): before this change, the throughput on a 4MB read from
SPI-NOR is 8.47MBit/s which equates a 17.11% bus-utilisation.
To improve on this, this commit adds an optimised receive-only
transfer (i.e.: out == NULL) handler that hooks into the main transfer
function and processes data in 16bit frames (utilising the full with
of each FIFO element). As of now, the receive-only handler requires
the in-buffer to be 16bit aligned. Any lingering data (i.e. either if
the in-buffer was not 16-bit aligned or if an odd number of bytes are
to be received) will be handled by the original 8bit reader/wirter.
Given that the SPI controller's documentation does not guarantuee any
interlocking between the RXFIFO and the master SCLK, the transfer loop
will be restarted for each chunk of 32 frames (i.e. 64 bytes).
With this new receive-only transfer handler, the throughput for a 4MB
read increases to 36.28MBit/s (i.e. 73.29% bus-utilisation): this is a
4x improvement over the baseline.
Philipp Tomsich [Sun, 3 Feb 2019 15:17:30 +0000 (16:17 +0100)]
rockchip: spi: only wait for completion, if transmitting
The logic in the main transmit loop took a bit of reading the TRM to
fully understand (due to silent assumptions based in internal logic):
the "wait until idle" at the end of each iteration through the loop is
required for the transmit-path as each clearing of the ENA register
(to update run-length in the CTRLR1 register) will implicitly flush
the FIFOs... transmisson can therefore not overlap loop iterations.
This change adds a comment to clarify the reason/need for waiting
until the controller becomes idle and wraps the entire check into an
'if (out)' to make it clear that this is required for transfers with a
transmit-component only (for transfers having a receive-component,
completion of the transmit-side is trivially ensured by having
received the correct number of bytes).
The change does not increase execution time measurably in any of my
tests.
Philipp Tomsich [Sun, 3 Feb 2019 15:17:29 +0000 (16:17 +0100)]
rockchip: spi: consistently use false/true with rkspi_enable_chip
While rkspi_enable_chip is called with true/false everywhere else in
the file, one call site uses '0' to denot 'false'.
This change this one parameter to 'false' and effects consistency.
Philipp Tomsich [Sun, 3 Feb 2019 15:17:28 +0000 (16:17 +0100)]
rockchip: spi: fix off-by-one in chunk size computation
The maximum transfer length (in a single transaction) for the Rockchip
SPI controller is 64Kframes (i.e. 0x10000 frames) of 8bit or 16bit
frames and is encoded as (num_frames - 1) in CTRLR1. The existing
code subtracted the "minus 1" twice for a maximum transfer length of
0xffff (64K - 1) frames.
While this is not strictly an error (the existing code is correct, but
leads to a bit of head-scrating), fix this off-by-one situation.
Philipp Tomsich [Sun, 3 Feb 2019 15:17:27 +0000 (16:17 +0100)]
rockchip: spi: remove unused code and fields in priv
Even though the priv-structure and the claim-bus function contain
logic for 16bit frames and for unidirectional transfer modes, neither
of these is used anywhere in the driver.
This removes the unused (as in "has no effect") logic and fields.
Philipp Tomsich [Sun, 3 Feb 2019 15:17:26 +0000 (16:17 +0100)]
rockchip: spi: add debug message for delay in CS toggle
In analysing delays introduced for large SPI reads, the absence of any
indication when a delay was inserted (to ensure the CS toggling is
observed by devices) became apparent.
Add an additional debug-only debug message to record the insertion and
duration of any delay (note that the debug-message will cause a delay
on-top of the delay-duration).
Philipp Tomsich [Sun, 3 Feb 2019 14:59:23 +0000 (15:59 +0100)]
rockchip: rk3399-puma: support Gigadevice SPI-NOR flash
Over the last quarter, a part of our production has used NOR flash
from Gigadevice in addition to the Winbond parts that we typically
source. This requires the SPI_FLASH_GIGADEVICE config to be set.
Enable SPI_FLASH_GIGADEVICE in the board's default defconfig.
The usage of socfpga_sdram_apply_static_cfg() seems rather dubious and
is confirmed to lead to a rare system hang when enabling bridges. This
patch removes the socfpga_sdram_apply_static_cfg() altogether, because
it's use seems unjustified and problematic.
The socfpga_sdram_apply_static_cfg() triggers write to SDRAM staticcfg
register to set the applycfg bit, which according to old vendor U-Boot
sources can only be written when there is no traffic between the SDRAM
controller and the rest of the system. Empirical measurements confirm
this, setting the applycfg bit when there is traffic between the SDRAM
controller and CPU leads to the SDRAM controller accesses being blocked
shortly after.
Altera originally solved this by moving the entire code which sets the
staticcfg register to OCRAM [1]. The commit message claims that the
applycfg bit needs to be set after write to fpgaportrst register. This
is however inverted by Altera shortly after in [2], where the order
becomes the exact opposite of what commit message [1] claims to be the
required order. The explanation points to a possible problem in AMP
use-case, where the FPGA might be sending transactions through the F2S
bridge.
However, the AMP is only the tip of the iceberg here. Any of the other
L2, L3 or L4 masters can trigger transactions to the SDRAM. It becomes
rather non-trivial to guarantee there are no transactions to the SDRAM
controller.
The SoCFPGA SDRAM driver always writes the applycfg bit in SPL. Thus,
writing the applycfg again in bridge enable code seems redundant and
can presumably be dropped.
Marek Vasut [Wed, 13 Feb 2019 19:16:20 +0000 (20:16 +0100)]
mmc: dw_mmc: Round up descriptor end to nearest multiple of cacheline size
The driver currently calculates the end address of cache flush operation
for the DMA descriptors by adding cacheline size to the start address of
the last DMA descriptor. This is not safe, as the cacheline size may be,
in some unlikely cases, smaller than the DMA descriptor size. Replace the
addition with roundup() applied on the end address of the last DMA
descriptor to round it up to the nearest cacheline size multiple.
Marek Vasut [Sat, 23 Mar 2019 17:45:27 +0000 (18:45 +0100)]
mmc: dw_mmc: Handle return value from bounce_buffer_start()
The bounce_buffer_start() can return -ENOMEM in case memory allocation
failed. However, in that case, the bounce buffer address is the same as
the possibly unaligned input address, and the cache maintenance operations
were not applied to this address. This could cause subtle problems. Add
handling for the bounce_buffer_start() return value to prevent such a
problem from happening.
Marek Vasut [Sat, 23 Mar 2019 02:32:24 +0000 (03:32 +0100)]
mmc: dw_mmc: Calculate timeout from transfer length
The current 4-minute data transfer timeout is misleading and broken.
Instead of such a long wait, calculate the timeout duration based on
the length of the data transfer. The current formula is the transfer
length in bits, divided by a multiplication of bus frequency in Hz,
bus width, DDR mode and converted the mSec. The value is bounded from
the bottom to 1000 mSec.
Marek Vasut [Tue, 16 Apr 2019 20:28:08 +0000 (22:28 +0200)]
ARM: socfpga: Add support for selecting bridges in bridge command
Add optional "mask" argument to the SoCFPGA bridge command, to select
which bridges should be enabled/disabled. This allows the user to avoid
enabling bridges which are not connected into the FPGA fabric. Default
behavior is to enable/disable all bridges.
Marek Vasut [Tue, 16 Apr 2019 20:13:29 +0000 (22:13 +0200)]
ARM: socfpga: Fully unmap the FPGA bridges from L3 space
Instead of just putting the bridges into reset, fully remove the bridges
from the L3 main bridge space when disabling them by clearing bits in
NIC-301 remap register. Moreover, only touch the 3 LSbits in brgmodrst
register as the rest of the bits are undefined.
Marek Vasut [Tue, 16 Apr 2019 12:19:34 +0000 (14:19 +0200)]
ARM: socfpga: Disable bridges in SPL unless booting from FPGA
Disable bridges between L3 Main switch and FPGA unless booting
from FPGA and keep them disabled to prevent glitches and possible
hangs of the L3 Main switch.
The current version of the code could have enabled the bridges
between the L3 Main switch and FPGA for a short period of time
in board_init_f() in case the FPGA was programmed and then again
disable them at the end of board_init_f(). Replace this with a
code which only sets up the handoff registers and let the user
enable the bridges later on.
Marek Vasut [Tue, 16 Apr 2019 21:05:24 +0000 (23:05 +0200)]
ARM: socfpga: Factor out handoff register configuration
Factor out the code for programming preloader handoff register values,
the ISWGRP Handoff 0 and 1. These registers later control which bridges
are enabled by the "bridge" command on Gen5 devices.
Tom Rini [Sat, 27 Apr 2019 15:34:55 +0000 (11:34 -0400)]
Revert "fs: btrfs: fix false negatives in ROOT_ITEM search"
Per Pierre this change shouldn't have been applied as it was superseded
by "fs: btrfs: fix btrfs_search_tree invalid results" which is also
applied now as 1627e5e5985d.
Igor Opaniuk [Tue, 9 Apr 2019 13:38:14 +0000 (15:38 +0200)]
avb: add support for named persistent values
AVB 2.0 spec. revision 1.1 introduces support for named persistent values
that must be tamper evident and allows AVB to store arbitrary key-value
pairs [1].
Introduce implementation of two additional AVB operations
read_persistent_value()/write_persistent_value() for retrieving/storing
named persistent values.
Correspondent pull request in the OP-TEE OS project repo [2].
Use _bss_start_ofs as the size of the boot loader code+data that we want
to protect in the flash. This replaces use of the no longer defined
_armboot_start.
Adam Ford [Wed, 17 Apr 2019 18:24:33 +0000 (13:24 -0500)]
ARM: da850evm: Fix TI boot scripts for MMC
The da850evm include environment/ti/mmc.h and places
DEFAULT_MMC_TI_ARGS, defined int that file, in
CONFIG_EXTRA_ENV_SETTINGS. This has been broken for some time
since neither CMD_PART nor CMD_UUID are available, so manually
changing the environmental variables was always required before
booting from MMC. With this patch, these scripts should work
again, and by default they point to mmc 0, partition 2, and
the dtb file exists in a /boot directory on partition 2.
Adam Ford [Wed, 17 Apr 2019 16:21:56 +0000 (11:21 -0500)]
ARM: da850evm: Fix broken SPI Flash
A previous patch converted a bunch of settings for CONFIG_SF_DEFAULT
but it broke the SPI Flash which now prevents booting. This patch
now correctly sets CONFIG_SF_DEFAULT_SPEED=30000000 to match what it
was before the conversion.
Fixes: 14453fbfadc2 ("Convert CONFIG_SF_DEFAULT_* to Kconfig") Signed-off-by: Adam Ford <[email protected]>
The match controller register is not cleared during
initialization. However, some bits of this register may reset the TC if
tnMRx match it.
As we can't make any assumption about how U-Boot is launched by the first
stage bootloader (such as S1L) clearing this register ensure that the
timers work as expected.
Since the introduction of the driver, some memory in IRAM is reserved for
the TX buffers.
However there are not used but instead of it, it is the buffer provided
by the net stack which is used. As stated in the comment of the driver,
not using the IRAM buffer could cause cache issue and lower the
throughput.
For the second argument it is less the case for transmitting buffers
because the throughput gain in IRAM is mitigated by the time to copy the
data from RAM to IRAM, but the first argument is still valid and indeed
this patch fixes issue seen with Ethernet on some boards
The purpose of "mtd: nand: raw: allow to disable unneeded ECC layouts"
was to allow disabling the default ECC layouts if a driver is known to
provide its own ECC layout. However, this commit did the opposite and
disabled the default layout when it was _not_ selected.
It breaks all the NAND drivers not providing their own ECC layout this
patch fix this situation.
Pierre Bourdon [Tue, 16 Apr 2019 00:47:14 +0000 (02:47 +0200)]
fs: btrfs: fix btrfs_search_tree invalid results
btrfs_search_tree should return the first item in the tree that is
greater or equal to the searched item.
The search algorithm did not properly handle the edge case where the
searched item is higher than the last item of the node but lower than
the first item of the next node. Instead of properly returning the first
item of the next node, it was returning an invalid path pointer
(pointing to a non-existent item after the last item of the node + 1).
This fixes two issues in the btrfs driver:
- Looking for a ROOT_ITEM could fail if it was the first item of its
leaf node.
- Iterating through DIR_INDEX entries (for readdir) could fail if the
first DIR_INDEX entry was the first item of a leaf node.
Pierre Bourdon [Sat, 13 Apr 2019 21:50:49 +0000 (23:50 +0200)]
fs: btrfs: fix false negatives in ROOT_ITEM search
ROOT_ITEMs in btrfs are referenced without knowing their actual "offset"
value. To perform these searches using only two items from the key, the
btrfs driver uses a special "btrfs_search_tree_key_type" function.
The algorithm used by that function to transform a 3-tuple search into a
2-tuple search was subtly broken, leading to items not being found if
they were the first in their tree node.
This commit fixes btrfs_search_tree_key_type to properly behave in these
situations.
Andrew F. Davis [Fri, 12 Apr 2019 16:54:47 +0000 (12:54 -0400)]
configs: Add configs for AM65x High Security EVM
Add new defconfig files for the AM65x High Security EVM.
This defconfigs are the same as for the non-secure part, except for:
CONFIG_TI_SECURE_DEVICE option set to 'y'
CONFIG_FIT_IMAGE_POST_PROCESS option set to 'y'
CONFIG_SPL_FIT_IMAGE_POST_PROCESS option set to 'y'
Andrew F. Davis [Fri, 12 Apr 2019 16:54:45 +0000 (12:54 -0400)]
arm: mach-k3: Add secure device support
K3 devices have High Security (HS) variants along with the non-HS already
supported. Like the previous generation devices (OMAP/Keystone2) K3
supports boot chain-of-trust by authenticating and optionally decrypting
images as they are unpacked from FIT images. Add support for this here.
Andrew F. Davis [Fri, 12 Apr 2019 16:54:42 +0000 (12:54 -0400)]
arm: K3: Avoid use of MCU_PSRAM0 before SYSFW is loaded
On HS devices the 512b region of reset isolated memory called
MCU_PSRAM0 is firewalled by default. Until SYSFW is loaded we
cannot use this memory. It is only used to store a single value
left at the end of SRAM by ROM that will be needed later. Save
that value to a global variable stored in the .data section.
This section is used as .bss will be cleared between saving
this value and using it.
cmd: pxe: Display splashscreen from extlinux.conf input
The objective is to provide a simple way to retrieve a BMP file,
and display it as splashscreen, from extlinux.conf file input.
For this, we take example on https://www.syslinux.org/wiki/
index.php?title=Menu#The_advanced_menu_system
and more particularly on MENU BACKGROUND chapter.
For this, add "menu background" support in pxe command.
As example, extlinux.conf content will look like:
# Generic Distro Configuration file generated by OpenEmbedded
menu title Select the boot mode
TIMEOUT 20
menu background ../splash.bmp
DEFAULT stm32mp157c-ev1-sdcard
LABEL stm32mp157c-ev1-sdcard
KERNEL /uImage
FDT /stm32mp157c-ev1.dtb
APPEND root=/dev/mmcblk0p6 rootwait rw earlyprintk console=ttySTM0,115200
On many B&R boards we have a reset-controller, responsible for very
early board-bringup (voltages, clocks, ...) and bootmode selection.
To be ready for adding more B&R boards to source tree while avoiding
duplicate code, we add the resetcontroller implementation to the common
part of B&R boards.
Hannes Schmelzer [Wed, 10 Apr 2019 12:13:11 +0000 (14:13 +0200)]
board/BuR/common: prepare for compiling common into non AM33XX boards
Today the BuR common stuff is only used on AM33XX boards. In future we
plan to have many other platforms than AM33XX so we have to move arch-
specific #include(s) to responsible #ifdef sections. By the way we drop
unneeded #include(s).
Tom Rini [Fri, 26 Apr 2019 17:50:00 +0000 (13:50 -0400)]
Merge git://git.denx.de/u-boot-marvell
- Add DM based generic watchdog start and reset implementation
and remove all ad-hoc implementations (Stefan)
- Move mv_sdhci to DM (Pierre)
- Misc turris_omnia updates (Pierre)
- Change openrd targets to correctly build again (size changes
and fixes to the dts targets) and bring it back into Travis
builds (Stefan)
- Add Kirkwood db-88f6281-bp board (Chris)
Stefan Roese [Thu, 11 Apr 2019 13:58:47 +0000 (15:58 +0200)]
watchdog: at91sam9_wdt: Remove now superfluous wdt start and reset
With the new generic function, the scattered other functions are now
removed to be replaced by the generic one. The new version also enables
the configuration of the watchdog timeout via the DT "timeout-sec"
property (if enabled via CONFIG_OF_CONTROL).
The watchdog servicing is enabled via CONFIG_WATCHDOG.
Stefan Roese [Thu, 11 Apr 2019 13:58:46 +0000 (15:58 +0200)]
watchdog: mpc8xx_wdt: Watchdog driver and macros cleanup
With the generic watchdog driver now implemented, this patch removes
some legacy stuff from the MPC8xx watchdog driver and its Kconfig
integration. CONFIG_MPC8xx_WATCHDOG is completely removed and
hw_watchdog_reset() is made static, as the watchdog will now get
serviced via the DM infrastructure if enabled via CONFIG_WATCHDOG.
Stefan Roese [Thu, 11 Apr 2019 13:58:45 +0000 (15:58 +0200)]
watchdog: cadence: Remove driver specific "timeout-sec" handling
Now that we have a generic DT property "timeout-sec" handling, the
driver specific implementation can be dropped.
This patch also changes the timeout restriction to the min and max
values (clipping). Before this patch, the value provided via
"timeout-sec" was used if the parameter was too high or low. Now
the driver specific min and max values are used instead.
Stefan Roese [Thu, 11 Apr 2019 13:58:44 +0000 (15:58 +0200)]
watchdog: Implement generic watchdog_reset() version
This patch tries to implement a generic watchdog_reset() function that
can be used by all boards that want to service the watchdog device in
U-Boot. This watchdog servicing is enabled via CONFIG_WATCHDOG.
Without this approach, new boards or platforms needed to implement a
board specific version of this functionality, mostly copy'ing the same
code over and over again into their board or platforms code base.
With this new generic function, the scattered other functions are now
removed to be replaced by the generic one. The new version also enables
the configuration of the watchdog timeout via the DT "timeout-sec"
property (if enabled via CONFIG_OF_CONTROL).
This patch also adds a new flag to the GD flags, to flag that the
watchdog is ready to use and adds the pointer to the watchdog device
to the GD. This enables us to remove the global "watchdog_dev"
variable, which was prone to cause problems because of its potentially
very early use in watchdog_reset(), even before the BSS is cleared.
Stefan Roese [Wed, 24 Apr 2019 14:05:06 +0000 (16:05 +0200)]
arm: mvebu: turris_omnia: Use thumb instructions in SPL to save space
With the latest changes to the drivers (SPI_FLASH_SPANSION etc), Travis
reports that the SPL image is too big. Let's use the thumb instructions
in SPL to save some space and make the image fit again.
Stefan Roese [Thu, 11 Apr 2019 10:33:36 +0000 (12:33 +0200)]
arm: kirkwood: openrd: Increase U-Boot size in flash to make it fit
We have run now multiple times into size issues with the openrd
board port. To finally fix this, this patch now moves the U-Boot size
from 0x6.0000 to 0x8.0000, giving enough space for the next time.
This also changes the environment location and potentially the
MTD partitioning, but I see no better fix for now. Especially since
this board does not have an active maintainer.
Pierre Bourdon [Thu, 11 Apr 2019 02:56:58 +0000 (04:56 +0200)]
mmc: mv_sdhci: add driver model support
The new DM implementation currently does not support the Sheeva
88SV331xV5 specific quirk present in the legacy implementation. The
legacy code is thus kept for this SoC and others not yet migrated to
DM_MMC.
Commit c4bd12a7dad4 ("i2c: mux: Generate longer i2c mux name") changed
the naming scheme of i2c devices within a mux. This broke references to
i2c@0 in the Turris Omnia board initialization code.
Igor Opaniuk [Mon, 15 Apr 2019 09:01:57 +0000 (11:01 +0200)]
colibri_imx6: use UUID for rootfs
1. Replace usage of "/dev/mmcblk*p*" with a proper UUID of rootfs
partition. This fixes the issue, when MMC controllers are probed in
a different order in U-boot and Linux kernel.
2. Fix legacy USB command (both sdboot and usbboot can be used now).
Fix compatible node to use regular Toradex notation.
Annotate device tree with standard Colibri pin muxing comments.
Use open-drain I2C pin muxings.
Alphabetically re-order iomuxc nodes.
Rename snvs-ad7879-int-grp touch interrupt node as per Linux device tree.
Ye Li [Wed, 17 Apr 2019 09:41:23 +0000 (09:41 +0000)]
mxc_ocotp: Disable fuse sense for imx8mq B1
On iMX8MQ Rev B1, reading from fuse box is not allowed. The
OCOTP_READ_FUSE_DATA register is tied to magic number 0xff0055aa
for chip rev. So u-boot has to disable the fuse sense function for it.
Ye Li [Wed, 17 Apr 2019 09:41:20 +0000 (09:41 +0000)]
mxc_ocotp: Update redundancy banks for mx7ulp B0
On mx7ulp B0, beside bank 0 and 1, the fuse bank 9, 10, 28 are changed to
Redundancy mode not ECC, so they can support to program different bits of
a word in multiple times.
Peng Fan [Mon, 15 Apr 2019 12:00:50 +0000 (12:00 +0000)]
imx: 8qxp_mek: fix fdt_file and console
Fix fdt_file and console to boot upstream Linux Kernel.
Upstream linux use imx8qxp-mek.dtb, and pass lpuart32 to earlycon
will not work for i.MX8QXP, only need to pass earlycon,
check drivers/tty/serial/earlycon.c,
" /* Just 'earlycon' is a valid param for devicetree and ACPI SPCR. */ "
Ludwig Zenz [Mon, 15 Apr 2019 09:13:08 +0000 (11:13 +0200)]
ARM: imx6: DHCOM i.MX6 PDK: use Kconfig for inclusion of DDR calibration
The four x16 DDR3 are wired in T-topology. From NXP AN4467:
'Although not required, T-Topologies may also benefit from performing
Write Leveling as there are package delays on both the processor and DDR
devices that can be de-skewed by performing Write Leveling. Therefore,
Freescale recommends determining Write Leveling calibration parameters
for all boards, regardless of topology used.'
That is why write level calibration is also done.
After DM conversion, the size of U-Boot binary to increase.
Previous size is 480K after DM Conversion the new size is 557K
So it's necessary to increase the dfu request for store u-boot-dtb.img in eMMC.
This patch enable convert DM MMC for imx7d-pico board and variant.
Before the DM conversion only usdhc3 was enabled and therefore it appeared
as MMC 0 to u-boot. After enabling MMC DM though usdhc3 defaults to MMC 2,
which left unattended would drive changes to existing pico-pi bootscripts and
environment variables that rely on mmc 0.
Setup the alias of mmc0 and usdhc3 so that existing pico-imx7d boot code will
work unmodified.
When converting to DM_MMC it is necessary that SPL initializes eMMC
by itself, so move the original eMMC initialization from U-Boot
proper to SPL.