SECURE_BOOT: Use default bootargs

For secure boot, currently we were using fixed bootargs for all SoCs.
This is not needed and we can use the bootargs which are used in
non-secure boot.

Signed-off-by: Aneesh Bansal <[email protected]>
Signed-off-by: Saksham Jain <[email protected]>
Reviewed-by: York Sun <[email protected]>

crypto/fsl: Make CAAM transactions cacheable

This commit solves CAAM coherency issue on ls2080. When caches are
enabled and CAAM's DMA's AXI transcations are not made cacheable,
Core reads/writes data from/to caches and CAAM does from main memory.
This forces data flushes to synchronize various data structures. But
even if any data in proximity of these structures is read by core,
these structures again are fetched in caches.

To avoid this problem, either all the data that CAAM accesses can be
made cache line aligned or CAAM transcations can be made cacheable.

So, this commit makes CAAM transcations as write back with write and
read allocate.

Signed-off-by: Saksham Jain <[email protected]>
Reviewed-by: York Sun <[email protected]>

crypto/fsl: Correct 64-bit write when MMU disabled

When MMU is disabled, 64-bit write must be aligned at 64-bit
boundary. Becaue the memory location is not guaranteed to be 64-bit
aligned, the 64-bit write needs to be split into two 32-bit writes
to avoid the alignment exception.

Signed-off-by: Aneesh Bansal <[email protected]>
Signed-off-by: Saksham Jain <[email protected]>
Reviewed-by: York Sun <[email protected]>

armv8: fsl-lsch3: Disable SMMU during secure boot

During secure boot, SMMU is enabled on POR by SP bootrom. SMMU needs
to be put in bypass mode in uboot to enable CAAM transcations to pass
through.

For non-secure boot, SP BootROM doesn't enable SMMU, which is in
bypass mode out of reset.

Signed-off-by: Aneesh Bansal <[email protected]>
Signed-off-by: Saksham Jain <[email protected]>
Reviewed-by: York Sun <[email protected]>

armv8: ls2080: Add config for endianess of CCSR GUR

The GUR (DCFG) registers in CCSR space are in little endian format.
Define a config CONFIG_SYS_FSL_CCSR_GUR_LE in
arch/arm/include/asm/arch-fsl-layerscape/config.h

Signed-off-by: Aneesh Bansal <[email protected]>
Signed-off-by: Saksham Jain <[email protected]>
Reviewed-by: York Sun <[email protected]>

armv8: ls2080: Change env variable "fdt_high"

"fdt_high" env variable was set to 0xcfffffff for secure boot.
Change it to 0xa0000000 for LS2080 to be consistent with non-secure
boot targets.

Signed-off-by: Aneesh Bansal <[email protected]>
Signed-off-by: Saksham Jain <[email protected]>
Reviewed-by: York Sun <[email protected]>

armv8: fsl-lsch3: Copy Bootscript and header from NOR to DDR

To unify steps for secure boot for xip (eg. NOR) and non-xip memories
(eg. NAND, SD), bootscipts and its header are copied to main memory.
Validation and execution are performed from there.

For other ARM Platforms (ls1043 and ls1020), to avoid disruption of
existing users, this copy step is not used for NOR boot.

Signed-off-by: Aneesh Bansal <[email protected]>
Signed-off-by: Saksham Jain <[email protected]>
Reviewed-by: York Sun <[email protected]>

armv8: ls2080: Add bootscript header addr for secure boot

During secure boot, Linux image along with other images are validated
using bootscript. This bootscript also needs to be validated before
it executes. This requires a header for bootscript.

When secure boot is enabled, default bootcmd is changed to first
validate bootscript using the header and then execute the script.

For ls2080, NOR memory map is different from other ARM SoCs. So a new
address on NOR is used for this bootscript header (0x583920000). The
Bootscript address is mentioned in this header along with addresses of
other images.

Signed-off-by: Aneesh Bansal <[email protected]>
Signed-off-by: Saksham Jain <[email protected]>
Reviewed-by: York Sun <[email protected]>

armv8: ls2080rdb: ls2080qds: Add secure boot support

Sec_init has been called at the beginning to initialize SEC Block
(CAAM) which is used by secure boot validation later for both ls2080a
qds and rdb. 64-bit address in ESBC Header has been enabled. Secure
boot defconfigs are created for boards (NOR boot).

Signed-off-by: Saksham Jain <[email protected]>
Reviewed-by: York Sun <[email protected]>

armv8: fsl-lsch3: Add new header for secure boot

For secure boot, a header is used to identify key table, signature
and image address. A new header structure is added for lsch3.

Currently key extension (IE) feature is not supported. Single key
feature is not supported. Keys must be in table format. Hence, SRK
(key table) must be present. Max key number has increase from 4 to
8. The 8th key is irrevocable. A new barker Code is used.

Signed-off-by: Aneesh Bansal <[email protected]>
Signed-off-by: Saksham Jain <[email protected]>
Reviewed-by: York Sun <[email protected]>

armv8: ls2080: Add configs for SEC, SecMon, SRK and DCFG

Add configs for various IPs used during secure boot. Add address
and endianness for SEC and Security Monitor. SRK are fuses in SFP
(fuses for public key's hash). These are stored in little endian
format.

Signed-off-by: Aneesh Bansal <[email protected]>
Signed-off-by: Saksham Jain <[email protected]>
Reviewed-by: York Sun <[email protected]>

armv8: ls2080: Add SFP Configs for LS2080

In LS2080, SFP has version 3.4. It is in little endian. The base
address is 0x01e80200. SFP is used in Secure Boot to read fuses.

Signed-off-by: Aneesh Bansal <[email protected]>
Signed-off-by: Saksham Jain <[email protected]>
Reviewed-by: York Sun <[email protected]>

driver: net: fsl-mc: Check NULL before pointer dereference

NULL pointer should be checked before any dereference.  This patch
move memest after the NULL pointer check.

Signed-off-by: Prabhakar Kushwaha <[email protected]>
Reported-by: Jose Rivera <[email protected]>
Acked-by: Joe Hershberger <[email protected]>
Reviewed-by: York Sun <[email protected]>

driver: net: fsl-mc: Free dflt_dpio pointer after its usage

Free dflt_dpio pointer after its usage during error handling

Signed-off-by: Prabhakar Kushwaha <[email protected]>
Reported-by: Jose Rivera <[email protected]>
Reviewed-by: York Sun <[email protected]>

armv8: fsl-layerscape: Skip reconfigure QSPI clock when booting from QSPI

The qspi_cfg register is set by PBI when booting from QSPI. No need
to changing it again.

Signed-off-by: Gong Qianyu <[email protected]>
Reviewed-by: York Sun <[email protected]>

armv8/ls1043a: Update env settings for booting kernel

Signed-off-by: Gong Qianyu <[email protected]>
Reviewed-by: York Sun <[email protected]>

armv8: ls2080a: Update fdt path for fsl-mc node

The fsl-mc node has been moved under /soc, so update
the path references accordingly.  Backwards compatibility
is retained for /fsl-mc.

Delete backwards compatibility for the completely obsolete
/fsl,dprc@0.

Signed-off-by: Stuart Yoder <[email protected]>
Reviewed-by: York Sun <[email protected]>

net: Move CONFIG_RTL8169 to Kconfig

Introduce CONFIG_RTL8169 in Kconfig and move over boards' defconfig
to use that.

Signed-off-by: Bin Meng <[email protected]>
Acked-by: Stephen Warren <[email protected]>
Acked-by: Joe Hershberger <[email protected]>

net: Move CONFIG_RTL8139 to Kconfig

Introduce CONFIG_RTL8139 in Kconfig and move over boards' defconfig
to use that.

Signed-off-by: Bin Meng <[email protected]>
Acked-by: Joe Hershberger <[email protected]>
[trini: Fixup MPC8641HPCN* and r2dplus configs]
Signed-off-by: Tom Rini <[email protected]>

defconfig: Reorder boards' defconfig files

Some boards' defconfig files are out of order. Clean this up.

Signed-off-by: Bin Meng <[email protected]>
Acked-by: Joe Hershberger <[email protected]>

usb: Correct return value in usb_stor_info()

This should return 0 on success, not 1. Fix it.

Signed-off-by: Simon Glass <[email protected]>
Acked-by: Marek Vasut <[email protected]>
Reviewed-by: Tom Rini <[email protected]>

mkimage: Don't close the file if it wasn't opened

The error path for fit_import_data() is incorrect if the second open() call
fails.

Reported-by: Coverity (CID: 138489)
Signed-off-by: Simon Glass <[email protected]>
Reviewed-by: Tom Rini <[email protected]>

mkimage: Ensure file is closed in fdt_property_file()

The file that is opened is not closed in all cases. Fix it.

Reported-by: Coverity (CID: 138490)
Signed-off-by: Simon Glass <[email protected]>
Reviewed-by: Tom Rini <[email protected]>

mkimage: Fix missing free() and close() in fit_build()

Make sure that both the error path and normal return free the buffer and
close the file.

Reported-by: Coverity (CID: 138491)
Signed-off-by: Simon Glass <[email protected]>
Reviewed-by: Tom Rini <[email protected]>

mkimage: Fix missing free() in fit_extract_data()

The 'buf' variable is not freed. Fix it.

Reported-by: Coverity (CID: 138492)
Signed-off-by: Simon Glass <[email protected]>
Reviewed-by: Tom Rini <[email protected]>

mkimage: Fix error path in fit_extract_data()

The 'fdt' variable is not unmapped in all error cases. Fix this.

Reported-by: Coverity (CID: 138493)
Signed-off-by: Simon Glass <[email protected]>
Reviewed-by: Tom Rini <[email protected]>

mkimage: Add a missing free() to fit_import_data()

The space allocated to fdt is not freed on error. Fix it.

Reported-by: Coverity (CID: 138494)
Signed-off-by: Simon Glass <[email protected]>
Reviewed-by: Tom Rini <[email protected]>

mkimage: Close the file when unable to get its size

There is a missing close() on the error path. Add it.

Reported-by: Coverity (CID: 138496)
Signed-off-by: Simon Glass <[email protected]>
Reviewed-by: Tom Rini <[email protected]>

part_efi: Drop NULL check in part_get_info_efi()

This cannot be NULL since part_get_info() calls this function and requires
it to be non-NULL.

Reported-by: Coverity (CID: 138497)
Signed-off-by: Simon Glass <[email protected]>
Reviewed-by: Tom Rini <[email protected]>

part_efi: Drop the NULL check on dev_desc in part_print_efi()

This cannot be NULL since part_print() calls this function and requires it
to be non-NULL.

Reported-by: Coverity (CID: 138498)
Signed-off-by: Simon Glass <[email protected]>
Reviewed-by: Tom Rini <[email protected]>

part_iso: Drop the customer unaligned access functions

One of these is causing a coverity warning. Drop these functions and use the
standard U-Boot ones instead.

Reported-by: Coverity (CID: 138499)
Signed-off-by: Simon Glass <[email protected]>
Reviewed-by: Tom Rini <[email protected]>

mkimage: Correct file being closed twice in fit_extract_data()

The code flows through to the end of the function, so we don't need another
close() before this. Remove it.

Reported-by: Coverity (CID: 138503)
Signed-off-by: Simon Glass <[email protected]>
Reviewed-by: Tom Rini <[email protected]>

mkimage: Correct file being closed twice in fit_import_data()

The code flows through to the end of the function, so we don't need another
close() before this. Remove it.

Reported-by: Coverity (CID: 138504)
Signed-off-by: Simon Glass <[email protected]>
Reviewed-by: Tom Rini <[email protected]>

mkimage: Fix munmap() call when importing data

The munmap() call unmaps the wrong memory buffer. Fix it.

Reported-by: Coverity (CID: 138505)
Reported-by: Coverity (CID: 138495)
Signed-off-by: Simon Glass <[email protected]>
Reviewed-by: Tom Rini <[email protected]>

Fix spelling of "transferred".

Signed-off-by: Vagrant Cascadian <[email protected]>
Acked-by: Marek Vasut <[email protected]>
Reviewed-by: Simon Glass <[email protected]>
Reviewed-by: Tom Rini <[email protected]>

Fix spelling of "supported/unsupported".

Signed-off-by: Vagrant Cascadian <[email protected]>
Reviewed-by: Simon Glass <[email protected]>
Reviewed-by: Tom Rini <[email protected]>
Reviewed-by: Peter Griffin <[email protected]>

Fix spelling of "comment".

Signed-off-by: Vagrant Cascadian <[email protected]>
Reviewed-by: Tom Rini <[email protected]>

cmd: Fix license command

The license command isn't usually built and has a few problems:
- The rules to generate license.h haven't worked in a long time,
  re-write these based on the bmp_logo.h rules.
- 'tok' is unused and the license text size has increased
- bin2header.c wasn't grabbing unistd.h to know the prototype for
  read().

Cc: Masahiro Yamada <[email protected]>
Signed-off-by: Tom Rini <[email protected]>

serial: pl01x: Add support for devices with the rate pre-configured.

For Raspberry Pi, we had the input clock rate to the pl011 fixed in
the rpi.c file, but it may be changed by firmware due to user changes
to config.txt.  Since the firmware always sets up the uart (default
115200 output unless the user changes it), we can just skip our own
uart init to simplify the boot process and more reliably get serial
output.

Signed-off-by: Eric Anholt <[email protected]>
Reviewed-by: Tom Rini <[email protected]>
Tested-by: Stephen Warren <[email protected]>

Drop various features when the command line is not available

Some features are only useful or meaningful when the command line is
present. Ensure that these features are not compiled in when CONFIG_CMDLINE
is not enabled.

Signed-off-by: Simon Glass <[email protected]>
Reviewed-by: Tom Rini <[email protected]>

Allow command-line files to be dropped

These files do not need to be compiled when CONFIG_CMDLINE is disabled.
Update the Makefile to reflect this.

Signed-off-by: Simon Glass <[email protected]>
Reviewed-by: Tom Rini <[email protected]>

Allow command code to compile to nothing

When CONFIG_CMDLINE is disabled we need to remove all the command-line
code. Most can be removed by dropping the appropriate linker lists from the
images, but sub-commands must be dealt with specially.

A simple mechanism is used to avoid 'unused static function' errors.

Signed-off-by: Simon Glass <[email protected]>
Reviewed-by: Tom Rini <[email protected]>

Panic when no command line processing can be performed

Normally board_run_command() will handle command processed. But if for some
reason it returns then we should panic to avoid further processing.

Signed-off-by: Simon Glass <[email protected]>
Reviewed-by: Tom Rini <[email protected]>

sandbox: Avoid calling commands when not available

Don't try to run commands when not supported.

Signed-off-by: Simon Glass <[email protected]>
Reviewed-by: Tom Rini <[email protected]>

arm: x86: Drop command-line code when CONFIG_CMDLINE is disabled

Update the link script to drop this code when not needed. This is only done
for two architectures at present.

Signed-off-by: Simon Glass <[email protected]>
Reviewed-by: Tom Rini <[email protected]>

Add an option to enable the command line

Add a new Kconfig option for the command line. This is enabled by default,
but when disabled it will remove the command line.

Signed-off-by: Simon Glass <[email protected]>
Reviewed-by: Tom Rini <[email protected]>

cbfs: Update a function to be static

All command functions should be static. Update the CBFS functions to follow
this rule.

Signed-off-by: Simon Glass <[email protected]>
Reviewed-by: Tom Rini <[email protected]>

ti: k2g: increase phy autoneg timeout

After power cycle of a K2G EVM dhcp fails due to a auto-negotiation
timeout. This commit increases the timeout to fix the issue.

Signed-off-by: Vitaly Andrianov <[email protected]>
Reviewed-by: Tom Rini <[email protected]>

Merge branch 'master' of git://git.denx.de/u-boot-fsl-qoriq

driver: net: fsl-mc: Return from DPAA_exit if boot_status !=0

Return value of get_mc_boot_status() in case of failure is not necessary
to be -1.

So update the error condition check.

Signed-off-by: Prabhakar Kushwaha <[email protected]>
Reported-by: Yao Yuan <[email protected]>
Acked-by: Joe Hershberger <[email protected]>
Reviewed-by: York Sun <[email protected]>

armv8/ls1043aqds: Enable ID_EEPROM support for ls1043aqds

Signed-off-by: Wenbin Song <[email protected]>
Reviewed-by: York Sun <[email protected]>

armv8/ls1043aqds: Return i2c mux to default chennel

Return i2c mux to the default channel after accessing retimer.

Signed-off-by: Wenbin Song <[email protected]>
Reviewed-by: York Sun <[email protected]>

freescale: vid: Return i2c mux to default channel

IR chip is on one of the channels on multiplexed I2C-bus.
Reset to default channel after accessing.

Signed-off-by: Wenbin Song <[email protected]>
Reviewed-by: York Sun <[email protected]>

arm: ls102xa: Enable CONFIG_SYS_CONSOLE_IS_IN_ENV support

CONFIG_SYS_CONSOLE_IS_IN_ENV needs to be enabled, so we could set stdout
environment variable to specify the vga for the console output when
LCD/HDMI is connected to the boards.

Signed-off-by: Alison Wang <[email protected]>
Reviewed-by: York Sun <[email protected]>

armv8/fsl-lsch2: fix sdhc clock frequency value

The eSDHC could select to use platform clock or peripheral clock to
generate SD clock. The default selection is platform clock. So, fix
the clock frequency value that's calculated for eSDHC.

Signed-off-by: Yangbo Lu <[email protected]>
Reviewed-by: York Sun <[email protected]>

armv8: ls2085a: Remove phy configuration from QDS and RDB

As phy_connect and phy_config are being called from DPAA2 driver.
Remove calling of mentioned function from board file.

Signed-off-by: Pratiyush Mohan Srivastava <[email protected]>
Signed-off-by: Prabhakar Kushwaha <[email protected]>
Reviewed-by: York Sun <[email protected]>

driver: net: ldpaa_eth: Add support of PHY framework

This patch integrate DPAA2 ethernet driver existing PHY framework.

Call phy_connect and phy_config as per available DPMAC id defined
in SerDes Protcol.

Signed-off-by: Pratiyush Mohan Srivastava <[email protected]>
Signed-off-by: Prabhakar Kushwaha <[email protected]>
Reviewed-by: York Sun <[email protected]>

armv8: fsl-layerscape: Updating entries in Serdes Table

The serdes protocol entries in  Serdes table 1 for protocol
0x03, 0x33, 0x35 and in Serdes table 2 for protocols 0x45
and 0x47 are updated to reflect the entries in
current Reference Manual.

Signed-off-by: Pratiyush Mohan Srivastava <[email protected]>
Reported-by: Jose Rivera <[email protected]>
Reviewed-by: York Sun <[email protected]>

driver/ddr/fsl: Add workaround for erratum A-009803

During initial DDR training, false parity errors may be detected.
This patch adds workaround to fix the erratum.
Tested on LS2085QDS and LS2080RDB.

Signed-off-by: Shengzhou Liu <[email protected]>
Reviewed-by: York Sun <[email protected]>

driver/ddr/fsl: Add address parity support for DDR4 UDIMM/discrete

Add support of address parity for DDR4 UDIMM or discrete memory.
It requires to configurate corresponding MR5[2:0] and
TIMING_CFG_7[PAR_LAT]. Parity can be turned on by hwconfig,
e.g. hwconfig=fsl_ddr:parity=on.

Signed-off-by: Shengzhou Liu <[email protected]>
Reviewed-by: York Sun <[email protected]>

pci/layerscape: set LUT and msi-map for discovered PCI devices

msi-map properties are used to tell an OS how PCI requester IDs are
mapped to ARM SMMU stream IDs.

for all PCI devices discovered in a system:
  -allocate a LUT (look-up-table) entry in that PCI controller
  -allocate a stream ID for the device
  -program and enable a LUT entry (maps PCI requester id to stream ID)
  -set the msi-map property on the controller reflecting the
   LUT mapping

basic bus scanning loop/logic was taken from drivers/pci/pci.c
pci_hose_scan_bus().

Signed-off-by: Stuart Yoder <[email protected]>
Reviewed-by: York Sun <[email protected]>

pci/layerscape: add defines for LUT

The per-PCI controller LUT (Look-Up-Table) is a 32-entry table
that maps PCI requester IDs (bus/dev/fun) to a stream ID.

Add defines for the register offsets.

Signed-off-by: Stuart Yoder <[email protected]>
Reviewed-by: York Sun <[email protected]>

pci: make pci_get_hose_head() available to external users

Put pci_get_hose_head() prototype in header so it is available to
external users, allowing them to find and iterate over all pci
controllers.

Signed-off-by: Stuart Yoder <[email protected]>
Reviewed-by: York Sun <[email protected]>

armv8: ls2080a: update stream ID partitioning info

Update comments around how stream IDs are partitioned.
Stream IDs allocated to PCI are no longer divided up by
controller, but are instead a contiguous range

Signed-off-by: Stuart Yoder <[email protected]>
Reviewed-by: York Sun <[email protected]>

armv8: ls2080a: remove obsolete stream ID partitioning support

Remove stream ID partitioning support that has been made
obsolete by upstream device tree bindings that specify how
representing how PCI requester IDs are mapped to MSI specifiers
and SMMU stream IDs.

Signed-off-by: Stuart Yoder <[email protected]>
Reviewed-by: York Sun <[email protected]>

boards: ls2080: Fix default bootargs

A white space is missing in multiple-line string for bootargs.

Signed-off-by: York Sun <[email protected]>

arm: ls102xa: fdt: Update FSL_QSPI_COMPAT and FSL_DSPI_COMPAT

As the compatible property values for QSPI and DSPI dts nodes
are changed in kernel, FSL_QSPI_COMPAT and FSL_DSPI_COMPAT
need to be updated too.

Signed-off-by: Alison Wang <[email protected]>
Reviewed-by: York Sun <[email protected]>

drivers/crypto/fsl: define structures for PDB

Structures are defined for PDB (Protocol Data Blcks) for various
operations. These structure will be used to add PDB data while
creating the PDB descriptors.

Signed-off-by: Ruchika Gupta <[email protected]>
Signed-off-by: Aneesh Bansal <[email protected]>
CC: Ulises Cardenas <[email protected]>
Reviewed-by: York Sun <[email protected]>

drivers/crypto/fsl: add constructs for protocol descriptors

Construct APIs are added to create Protocol Descriptors for
CAAM block.

Signed-off-by: Ruchika Gupta <[email protected]>
Signed-off-by: Aneesh Bansal <[email protected]>
CC: Ulises Cardenas <[email protected]>
Reviewed-by: York Sun <[email protected]>

drivers/crypto/fsl: correct error checking in run_descriptor

When CAAM runs a descriptor and an error occurs, a non-zero
value is set in Output Status Register. The if condition should
check the status for a non-zero value.

Signed-off-by: Aneesh Bansal <[email protected]>
Reviewed-by: York Sun <[email protected]>

armv8: ls2080ardb: invert irq pins polarity for AQR405 PHY

To use AQR405 PHY's interrupt, we need to invert the relative IRQ pins
polarity by setting IRQCR register, because AQR405 interrupt is low
active but GIC accepts high active.

Signed-off-by: Shaohui Xie <[email protected]>
Reviewed-by: York Sun <[email protected]>

armv8: lsch3: Enable WUO config for RNI-20 node

Enable wuo config to accelerate coherent ordered writes for LS2080A
and LS2085A.

WRIOP IP is connected to RNI-20 Node.

Signed-off-by: Prabhakar Kushwaha <[email protected]>
Reviewed-by: York Sun <[email protected]>

arm64: Fix layerscape mmu setup

With commit 7985cdf we converted all systems except for the Layerscape
SoCs to the generic descriptor table based page table setup.

On the Layerscape SoCs however, we just provide an empty table stub
and do the setup ourselves. To reserve enough memory for the tables,
we need to override the default counting mechanism which would end up
with an empty table because we have no maps.

Fixes: 7985cdf
Reported-by: York Sun <[email protected]>
CC: Alison Wang <[email protected]>
CC: Prabhakar Kushwaha <[email protected]>
Signed-off-by: Alexander Graf <[email protected]>
Tested-by: York Sun <[email protected]>
Reviewed-by: York Sun <[email protected]>

strider: use optimised bus timing for FPGA access

Use optimised bus timing for FPGA access.

Signed-off-by: Reinhard Pfau <[email protected]>
Signed-off-by: Dirk Eibach <[email protected]>

strider: Define pca593x widths

Signed-off-by: Dirk Eibach <[email protected]>

strider: Add DP501 support for cpu model

Signed-off-by: Dirk Eibach <[email protected]>
Signed-off-by: Stefan Roese <[email protected]>

Merge branch 'master' of git://git.denx.de/u-boot-socfpga

arm: socfpga: sr1500: Misc updates (SPI speed, env location)

This patch makes the following changes to the SR1500 board port:

- Update defconfig to support SPI NOR (use make savedefconfig).
- Increase SPI speed to a maximum of 100MHz for faster system
  bootup.
- Change environment location, so that its not between SPL and
  main U-Boot. This way the combined SPL / U-Boot image can
  be used for updates.

Signed-off-by: Stefan Roese <[email protected]>
Cc: Marek Vasut <[email protected]>

arm: socfpga: Allow boards to define a custom environment size

This patch makes it possible that boards can define a board-specific env
size. This is used by the SR1500 SoCFPGA board port.

Signed-off-by: Stefan Roese <[email protected]>
Cc: Chin Liang See <[email protected]>
Cc: Dinh Nguyen <[email protected]>
Cc: Marek Vasut <[email protected]>

arm: socfpga: Fix SR1500 env position

Move the inclusion of the common socfpga configuration file further
down in the sr1500 configuration, so that the socfpga_common.h can
check if environment is in SPI NOR and it's location is defined and
if it is not, define default location.

This fixes "arm: socfpga: Enabling U-Boot environment support in QSPI"
which introduced a minor warning.

Signed-off-by: Marek Vasut <[email protected]>
Cc: Chin Liang See <[email protected]>
Cc: Dinh Nguyen <[email protected]>
Cc: Dinh Nguyen <[email protected]>
Cc: Pavel Machek <[email protected]>
Cc: Marek Vasut <[email protected]>
Cc: Stefan Roese <[email protected]>

arm: socfpga: Enabling U-Boot environment support in QSPI

Enabling the support of storing U-Boot environment
within serial NOR flash. By default, its still
store into SDMMC

Signed-off-by: Chin Liang See <[email protected]>
Cc: Dinh Nguyen <[email protected]>
Cc: Dinh Nguyen <[email protected]>
Cc: Pavel Machek <[email protected]>
Cc: Marek Vasut <[email protected]>
Cc: Stefan Roese <[email protected]>

usb: xhci: Fix vendor command error if the request type is USB_REQ_SET_ADDRESS or USB_REQ_SET_CONFIGURATION.

Add test into xhci_submit_control_message for usb requesttype in USB
vendor request being of standardized type. This fixes detection of
certain USB fixes, for example Ethernet, USB 3.0 port. Non standardized
requesttype in USB vendor request will be ignored.

Signed-off-by: Ted Chen <[email protected]>
Tested-by: Anand Moon <[email protected]>

usb: Change power-on / scanning timeout handling

This patch changes the USB port scanning procedure and timeout
handling in the following ways:

a)
The power-on delay in usb_hub_power_on() is now reduced to a value of
max(100ms, "hub->desc.bPwrOn2PwrGood * 2"). The code does not wait
using mdelay, instead usb_hub_power_on() will wait before querying
the device in the scanning loop later. The total timeout for this
hub, which is 1 second + "hub->desc.bPwrOn2PwrGood * 2" is calculated
and will be used in the following per-port scanning loop as the timeout
to detect active USB devices on this hub.

b)
Don't delay the minimum delay (for power to stabilize) in
usb_hub_power_on(). Instead skip querying these devices in the scannig
loop until the delay time is reached.

c)
The ports are now scanned in a quasi parallel way. The current code did
wait for each (unconnected) port to reach its timeout and only then
continue with the next port. This patch now changes this to scan all
ports of all USB hubs quasi simultaneously. For this, all ports are added
to a scanning list. This list is scanned until all ports are ready
by either a) reaching the connection timeout (calculated earlier), or
by b) detecting a USB device. This results in a faster USB scan time as
the recursive scanning of USB hubs connected to the hub that's currently
being scanned will start earlier.

One small functional change to the original code is, that ports with
overcurrent detection will now get rescanned multiple times
(PORT_OVERCURRENT_MAX_SCAN_COUNT).

Without this patch:
starting USB...
USB0:   USB EHCI 1.00
scanning bus 0 for devices... 9 USB Device(s) found

time: 20.163 seconds

With this patch:
starting USB...
USB0:   USB EHCI 1.00
scanning bus 0 for devices... 9 USB Device(s) found

time: 1.822 seconds

So ~18.3 seconds of USB scanning time reduction.

Signed-off-by: Stefan Roese <[email protected]>
Acked-by: Hans de Goede <[email protected]>
Tested-by: Stephen Warren <[email protected]>

usb: Don't reset the USB hub a 2nd time

Debugging has shown, that all USB hubs are being reset twice while
USB scanning. This introduces additional delays and makes USB scanning
even more slow. Testing has shown that this 2nd USB hub reset doesn't
seem to be necessary.

This patch now removes this 2nd USB hub reset. Resulting in faster USB
scan time. Here the current numbers:

Without this patch:
=> time usb start
starting USB...
USB0:   USB EHCI 1.00
scanning bus 0 for devices... 9 USB Device(s) found

time: 24.003 seconds

With this patch:
=> time usb start
starting USB...
USB0:   USB EHCI 1.00
scanning bus 0 for devices... 9 USB Device(s) found

time: 20.392 seconds

So ~3.6 seconds of USB scanning time reduction.

Signed-off-by: Stefan Roese <[email protected]>
Cc: Simon Glass <[email protected]>
Acked-by: Hans de Goede <[email protected]>
Tested-by: Stephen Warren <[email protected]>
Cc: Marek Vasut <[email protected]>

usb: Remove 200 ms delay in usb_hub_port_connect_change()

This patch removes 2 mdelay(200) calls from usb_hub_port_connect_change().
These delays don't seem to be necessary. At least not in my tests. Here
the number for a custom x86 Bay Trail board (not in mainline yet) with
a quite large and complex USB hub infrastructure.

Without this patch:
starting USB...
USB0:   USB EHCI 1.00
scanning bus 0 for devices... 9 USB Device(s) found

time: 28.415 seconds

With this patch:
starting USB...
USB0:   USB EHCI 1.00
scanning bus 0 for devices... 9 USB Device(s) found

time: 24.003 seconds

So ~4.5 seconds of USB scanning time reduction.

Signed-off-by: Stefan Roese <[email protected]>
Cc: Simon Glass <[email protected]>
Acked-by: Hans de Goede <[email protected]>
Tested-by: Stephen Warren <[email protected]>
Cc: Marek Vasut <[email protected]>

usb: legacy_hub_port_reset(): Speedup hub reset handling

Start with a short USB hub reset delay of 20ms. This can be enough for
some configurations.

The 2nd delay at the end of the loop is completely removed. Since the
delay hasn't been long enough, a longer delay time of 200ms is assigned
and will be used in the next loop round.

This hub reset handling is also used in the v4.4 Linux USB driver,
hub_port_reset().

Signed-off-by: Stefan Roese <[email protected]>
Cc: Simon Glass <[email protected]>
Acked-by: Hans de Goede <[email protected]>
Tested-by: Stephen Warren <[email protected]>
Cc: Marek Vasut <[email protected]>

dm: blk: Add tests for block devices

Add some tests to check that block devices work as expected.

Signed-off-by: Simon Glass <[email protected]>
Tested-by: Stephen Warren <[email protected]>

dm: sandbox: Drop the pre-DM host implementation

Driver model is used for host device block devices now, so we don't need the
old code. Remove it.

Signed-off-by: Simon Glass <[email protected]>
Tested-by: Stephen Warren <[email protected]>

dm: sandbox: Switch over to use DM for block devices

Now that the drivers used by sandbox support CONFIG_BLK, we can switch
sandbox over to use driver model for block devices.

Signed-off-by: Simon Glass <[email protected]>
Tested-by: Stephen Warren <[email protected]>

dm: usb: Unbind old block devices when shutting down USB

When 'usb start' is used, block devices are created for any USB flash sticks
and disks, etc. When 'usb stop' is used, these block devices are currently
not removed.

We don't want old block devices hanging around since they can still be
visible to U-Boot. Therefore, when USB is shut down, remove and unbind all
the block devices created by the USB subsystem.

Possibly we should unbind all devices which don't cause problems by being
unbound. Most likely we can remove everything except USB controllers, hubs
and emulators. We can consider that later.

Signed-off-by: Simon Glass <[email protected]>
Tested-by: Stephen Warren <[email protected]>

buildman: Clarify the use of -V

This option outputs to the log file, not to the terminal. Clarify that in
the help, and add a mention of it in the README.

Signed-off-by: Simon Glass <[email protected]>
Reported-by: Stephen Warren <[email protected]>
Reviewed-by: Stephen Warren <[email protected]>

buildman: Add a way to specific a full toolchain prefix

At present buildman allows you to specify the directory containing the
toolchain, but not the actual toolchain prefix. If there are multiple
toolchains in a single directory, this can be inconvenient.

Add a new 'toolchain-prefix' setting to the settings file, which allows
the full prefix (or path to the C compiler) to be specified.

Update the documentation to match.

Suggested-by: Stephen Warren <[email protected]>
Reviewed-by: Joe Hershberger <[email protected]>
Signed-off-by: Simon Glass <[email protected]>
Tested-by: Stephen Warren <[email protected]>

buildman: Allow branch names which conflict with directories

At present if you try to use buildman with the branch 'test' it will
complain that it is unsure whether you mean the branch or the directory.
This is a feature of the 'git log' command that buildman uses. Fix it
by resolving the ambiguity.

Signed-off-by: Simon Glass <[email protected]>
Reviewed-by: Joe Hershberger <[email protected]>

dm: ns16550: Add support for reg-offset property

reg-offset is the part of standard 8250 binding in the kernel.
It is shifting start of address space by reg-offset.
On Xilinx platform this offset is typically 0x1000.

Signed-off-by: Michal Simek <[email protected]>
Reviewed-by: Simon Glass <[email protected]>
Reviewed-by: Tom Rini <[email protected]>
Moved the new field to the end of the struct to avoid problems:
Signed-off-by: Simon Glass <[email protected]>

Revert "fdt: fix address cell count checking in fdt_translate_address()"

This reverts commit 71105f50fedddfa5b0535d102c3d5078671721ad.

The reverted commit was applied for a temporary to unbreak
few Exynos boards on the release.

After the discussion about the change, this commit should be avoided.
Fixed device-tree for Exynos, allows reverting it without any issues.

Signed-off-by: Przemyslaw Marczak <[email protected]>
Cc: Stefan Roese <[email protected]>
Cc: Tom Rini <[email protected]>
Cc: Simon Glass <[email protected]>
Cc: Stephen Warren <[email protected]>
Cc: Stephen Warren <[email protected]>
Tested-by: Simon Glass <[email protected]>
Acked-by: Simon Glass <[email protected]>

dts:exynos:update pinctrl size-cells and fix child regs

This change is required to avoid warnings about invalid
size-cells defined in device-tree pinctrl nodes for Exynos.

Tested on:
- Odroid U3
- Odroid XU3

Signed-off-by: Przemyslaw Marczak <[email protected]>
Cc: Stefan Roese <[email protected]>
Cc: Tom Rini <[email protected]>
Cc: Simon Glass <[email protected]>
Cc: Stephen Warren <[email protected]>
Cc: Stephen Warren <[email protected]>
Tested-by: Simon Glass <[email protected]>
Acked-by: Simon Glass <[email protected]>
Acked-by: Minkyu Kang <[email protected]>

gpio: Report errors when GPIOs cannot be read

Some controllers do not allow the output value to be read. Detect this and
report the error in that case.

Signed-off-by: Simon Glass <[email protected]>

cmd_dhry.c: Use lldiv for vax_mips calculation as well

Since dhry_per_sec is a u64 we must also use lldiv here when working
with it.  Otherwise:
../lib/dhry/cmd_dhry.c:(.text.do_dhry+0xd8): undefined reference to `__udivdi3'

On some platforms.

Signed-off-by: Tom Rini <[email protected]>

x86: Add congatec conga-QA3/E3845-4G (Bay Trail) support

This patch adds support for the congatec conga-QA3/E3845-4G eMMC8 SoM,
installed on the congatec Qseven 2.0 evaluation carrier board
(conga-QEVAL).

Its port is very similar to the MinnowboardMAX port and also uses
the Intel FSP as described in doc/README.x86.

Currently supported are the following interfaces / devices:
- UART (via Winbond legacy SuperIO chip on carrier board)
- Ethernet (PCIe Intel I210 / E1000)
- SPI including SPI NOR as boot-device
- USB 2.0
- SATA via U-Boot SCSI IF
- eMMC
- Video (HDMI output @ 800x600)
- PCIe

Not supported yet is:
- I2C
- USB 3.0

Signed-off-by: Stefan Roese <[email protected]>
Cc: Simon Glass <[email protected]>
Cc: Bin Meng <[email protected]>
Reviewed-by: Bin Meng <[email protected]>

x86: Add support for the samus chromebook

This adds basic support for chromebook_samus. This is the 2015 Pixel and
is based on an Intel broadwell platform.

Supported so far are:
- Serial
- SPI flash
- SDRAM init (with MRC cache)
- SATA
- Video (on the internal LCD panel)
- Keyboard

Various less-visible drivers are provided to make the above work (e.g. PCH,
power control and LPC).

The platform requires various binary blobs which are documented in the
README. The major missing feature is USB3 since the existing U-Boot support
does not work correctly with Intel XHCI controllers.

Signed-off-by: Simon Glass <[email protected]>
Reviewed-by: Bin Meng <[email protected]>