Alexander Dahl [Fri, 20 Apr 2018 13:29:31 +0000 (15:29 +0200)]
tools: mkenvimage: Fix possible segfault on stdin input
The size of 'filebuf' was not increased as more and more bytes are read
from stdin, but 'filebuf' was always reallocated to the same fix size.
This works as long as only less bytes than the initial buffer size come
in, for more input this will segfault. (It actually does, I tested
that.) So for each loop cycle the buffer size has to be increased by the
number of bytes we want to read.
On success read() returns the number of bytes read or zero for EOF. On
error -1 is returned and errno is set, so the right way to test if read
had failed is to test the return value instead of errno.
Enable uart overrun feature which allows to benefits of uart
FIFO usage.
Previously overrun management was disabled, this has to effect
to bypassed the uart FIFO usage even if FIFO was enabled.
In particular configuration, for example when video console is
enabled, copy/pasting a long command line in console results in
corruption. This is due to the fact that a lot of time is consumed
in flushing the cache during frame buffer update, so uart chars are
not read fast enough.
By using uart FIFO and managing overrun, long command line can by
copy/paste in console without being corrupted.
Alex Kiernan [Thu, 19 Apr 2018 04:32:57 +0000 (04:32 +0000)]
spl: disk: usb: Add dependencies to sprintf/strto*
If SPL serial support is disabled nothing brings in sprintf, snprintf
or simple_strtoul:
env/built-in.o: In function `regex_callback':
env/attr.c:128: undefined reference to `sprintf'
disk/built-in.o: In function `blk_get_device_by_str':
disk/part.c:386: undefined reference to `simple_strtoul'
disk/part.c:395: undefined reference to `simple_strtoul'
disk/built-in.o: In function `blk_get_device_part_str':
disk/part.c:522: undefined reference to `simple_strtoul'
disk/built-in.o: In function `part_set_generic_name':
disk/part.c:704: undefined reference to `sprintf'
drivers/built-in.o: In function `init_peripheral_ep':
drivers/usb/musb-new/musb_gadget.c:1826: undefined reference to `sprintf'
drivers/built-in.o: In function `musb_core_init':
drivers/usb/musb-new/musb_core.c:1451: undefined reference to `snprintf'
Alex Kiernan [Thu, 19 Apr 2018 04:32:54 +0000 (04:32 +0000)]
spl: Split sprintf, strto* from SPL serial in Kconfig
When SPL serial is disabled, callers who need sprintf or strtoul fail
because their inclusion is guarded by CONFIG_SPL_SERIAL_SUPPORT/
CONFIG_TPL_SERIAL_SUPPORT.
Split printf, sprintf and strto into their own entries and then select
all of them if SERIAL_SUPPORT is enabled to match the current behaviour.
Include panic.o unconditionally as it can be called from anywhere which
uses BUG_ON().
Alex Kiernan [Thu, 19 Apr 2018 04:32:53 +0000 (04:32 +0000)]
spl: ti: Avoid serial calls when serial support is disabled
If CONFIG_SPL_SERIAL_SUPPORT is not set, then the build will fail:
board/ti/am335x/built-in.o: In function `spl_start_uboot':
board/ti/am335x/board.c:247: undefined reference to `serial_tstc'
board/ti/am335x/board.c:247: undefined reference to `serial_getc'
Avoid the calls to the serial functions in that case.
Alex Kiernan [Thu, 19 Apr 2018 04:32:51 +0000 (04:32 +0000)]
spl: ti: Avoid preloader_console_init if !CONFIG_SPL_SERIAL_SUPPORT
If CONFIG_SPL_SERIAL_SUPPORT is disabled then the build fails because
serial_init is undefined. Guard preloader_console_init() appropriately
to fix this.
Andre Przywara [Thu, 12 Apr 2018 01:24:46 +0000 (04:24 +0300)]
arm: move SYS_ARCH_TIMER to KConfig
SYS_ARCH_TIMER guards the usage of the ARM Generic Timer (aka arch
timer) in U-Boot.
At the moment it is mandatory for ARMv8 and used by a few ARMv7 boards.
Add a proper Kconfig symbol to express this dependency properly,
allowing certain board configuration to later disable arch timer in case
there are any problems with it.
Signed-off-by: Andre Przywara <[email protected]>
[tuomas: rebase + fix conflicts and resync with moveconfig & use select] Signed-off-by: Tuomas Tynkkynen <[email protected]>
Tuomas Tynkkynen [Thu, 12 Apr 2018 01:24:45 +0000 (04:24 +0300)]
ARM: qemu-arm: Dynamically determine timer frequency
After commit 46fc679ede5f69 ("arm: timer: get frequency for arch timer
armv7 in cp15 cntfrq") the ARM architected timer driver knows how to
determine the timer frequency at runtime by reading the CNTFRQ register,
so we don't need to hardcode the timer frequency anymore.
lib/rsa: Kconfig: Remove superfluous 'depends on RSA'
RSA_SOFTWARE_EXP and RSA_FREESCALE_EXP are wrapped inside:
if RSA
...
endif
So, remove the redundant "depends on RSA" from their depends expression.
In addition, move SPL_RSA into the same "if RSA ... endif" block, since
its only direct dependeny is CONFIG_RSA. This tidies up and simplifies
reading of lib/rsa/Kconfig.
kbuild: use -fmacro-prefix-map to make __FILE__ a relative path
The __FILE__ macro is used everywhere in U-Boot to locate the file
printing the log message, such as WARN_ON(), etc. If U-Boot is
built out of tree, this can be a long absolute path.
This is because Kbuild runs in the objtree instead of the srctree,
then __FILE__ is expanded to a file path prefixed with $(srctree)/.
A brand-new option from GCC, -fmacro-prefix-map, solves this problem.
If your compiler supports it, __FILE__ is the relative path from the
srctree regardless of O= option. This provides more readable log,
more reproducible builds, and smaller image size.
Tom Rini [Sun, 22 Apr 2018 23:15:03 +0000 (19:15 -0400)]
clang: Update documentation
As of clang-5.0, things have changed a bit. First, we cannot
automatically guess -target values as if we do not pass one with CC then
cc-option will fail. Second, to disable movt/movw relocations the
argument has become -mno-movt.
Related to the target part, we cannot use arm-none-eabi as that ends up
being too generic of an ARM target for things like say rpi_3_32b to
work.
Tom Rini [Sun, 22 Apr 2018 14:20:50 +0000 (10:20 -0400)]
Makefile: Update clang warning disables from Linux
Re-sync the logic about which clang warnings to disable from v4.17-rc1.
Note that we don't disable all of the same ones as for now we haven't
run into any cases of warnings from clang in code from upstream Linux.
Tom Rini [Wed, 18 Apr 2018 17:50:47 +0000 (13:50 -0400)]
Remove unnecessary instances of DECLARE_GLOBAL_DATA_PTR
We have a large number of places where while we historically referenced
gd in the code we no longer do, as well as cases where the code added
that line "just in case" during development and never dropped it.
Adam Ford [Sun, 15 Apr 2018 17:51:26 +0000 (13:51 -0400)]
Convert CONFIG_SPI to Kconfig
This converts the following to Kconfig:
CONFIG_SPI
This partly involves updating code that assumes that CONFIG_SPI implies
things that are specific to the MPC8xx SPI driver. For now, just update
the CONFIG tests. This also involves reworking the default for
CONFIG_SYS_DEF_EEPROM_ADDR so that we don't set it when we cannot make a
reasonable default, as it does not cause any compile failures.
Mans Rullgard [Sat, 21 Apr 2018 15:11:10 +0000 (16:11 +0100)]
ARM: mxs: support full SPL framework
This allows using the full SPL framework on mxs devices. In this
mode, the u-boot.sb image loaded by the boot ROM contains only the
SPL which then loads U-Boot proper or a kernel in falcon mode.
Mans Rullgard [Sat, 21 Apr 2018 15:11:07 +0000 (16:11 +0100)]
ARM: arm926ejs: fix lowlevel_init call
The code attempts to preserve the value of LR by storing it in R12/IP
across the lowevel_init() call. However, this register is not saved
by the callee. Use a register that guaranteed to be preserved instead.
usb: gadget: composite: fix NULL pointer when a non standard request is received
In case usb configuration is unknown (cdev->config == NULL), non standard
request should not be processed.
Remove also the cdev->config check below which will never happen.
Neil Armstrong [Wed, 25 Apr 2018 09:39:08 +0000 (11:39 +0200)]
usb: host: dwc3: fix phys init
When no PHYs are declared in the dwc3 node, the phy init fails.
This patch checks if the "phys" property is presend and reports
the error returned by dev_count_phandle_with_args().
This patchs also fixes the styles issues added in last commit.
This patch should fix the DWC3 support on the UniPhier SoC family.
Fixes: 7c839ea70c49 ("usb: host: dwc3: Add support for multiple PHYs") Reported-by: Masahiro Yamada <[email protected]> Signed-off-by: Neil Armstrong <[email protected]>
Marek Vasut [Mon, 23 Apr 2018 18:24:16 +0000 (20:24 +0200)]
ARM: rmobile: Update E2 Alt
The E2 Alt port was broken since some time. This patch updates
the E2 Alt port to use modern frameworks, DM, DT probing, SPL
for the preloading and puts it on par with the M2 Porter board.
Marek Vasut [Mon, 23 Apr 2018 18:24:10 +0000 (20:24 +0200)]
ARM: rmobile: Update M2-N Gose
The M2-N Gose port was broken since some time. This patch updates
the M2-N Gose port to use modern frameworks, DM, DT probing, SPL
for the preloading and puts it on par with the M2 Porter board.
Marek Vasut [Mon, 23 Apr 2018 18:24:06 +0000 (20:24 +0200)]
ARM: rmobile: Update H2 Lager
The H2 Lager port was broken since some time. This patch updates
the H2 Lager port to use modern frameworks, DM, DT probing, SPL
for the preloading and puts it on par with the M2 Porter board.
Since commit commit 152038ea1886 ("i.MX6UL: icore: Add SPL_OF_CONTROL
support") the OPOS6UL board doesn't boot anymore. Adding SPL_DM support
makes the board boot again.
This patch removes warp7_secure_defconfig. A previous patch set
CONFIG_ARMV7_BOOT_SEC_DEFAULT=y on the unsecure WaRP7 config. Fabio asked
if I could confirm that the NXP and upstream kernels will boot on the WaRP7
with CONFIG_ARMV7_BOOT_SEC_DEFAULT=y. I can confirm that this is the case,
so there's no need to support the secure defconfig - drop it now.
Commit f916757300c1 ("imx: Create distinct pre-processed mkimage
config files") provided a proper fix for the parallel mkimage
config files build failure, so the original workaround can be
safely reverted now.
mx7dsabresd: Remove the mx7dsabresd_secure_defconfig target
mx7dsabresd_secure_defconfig was introduced to allow booting NXP kernel
that has CAAM support and needs to boot in secure mode.
Instead of keeping two different config targets for the same board,
remove mx7dsabresd_secure_defconfig and select
CONFIG_ARMV7_BOOT_SEC_DEFAULT inside mx7dsabresd_defconfig so that
this target could be used to boot both mainline and the vendor kernel.
This makes maintenance task easier and avoid potentially confusion
for the end user.
Mark Jonas [Mon, 16 Apr 2018 16:11:50 +0000 (18:11 +0200)]
arm,imx6: fix PAD_CTL_SPEED_LOW constant
For most i.MX6 processors the PAD_CTL_SPEED_LOW constant is the same.
Only the i.MX6 SoloLite is an exemption. So far the code did not
consider that. Additionally, for a few i.MX6 processors the code used
the wrong value for the constant.
warp7: Add support for automated secure boot.scr verification
This patch adds support for verifying a signed boot.scr. With this in place
it's possible for run-time Linux to update boot.scr to set different
variables such as switching between different boot partitions, pointing to
different kernels etc and for u-boot to verify these changes via the HAB
prior to executing the commands contained in boot.scr.
setexpr allows us to do arithmetic for env variables - something that is
both useful and required when doing HAB authentication without hard-coding
HAB load addresses.
This patch enables CMD_SETEXPR for the WaRP7 defconfig.
warp7: hab: Set environment variable indicating IVT offset
This patch introduces the environment variable ivt_offset. When we define a
load address for Linux or DTB or any file the IVT associated with that file
is prepended. We extract the actual load addresses from u-boot.cfg and feed
these values into the code-signing process - hence we want u-boot to have
the real load addresses exported in uboot.cfg.
ivt_offset represents the addition or subtraction from the load address
that must happen to find an IVT header.
Doing secure boot on the WaRP7 using a common image format and the same
variable to represent the base address for each call means we can reduce
down the command to a single environment command.
This patch adds warp7_auth_or_fail as a wrapper around
"hab_auth_img_or_fail ${hab_ivt_addr} ${filesize} 0".
warp7: Define the name of a signed boot-script file
We need to know the name of a signed boot-script, its better to have a
separate variable for this then to simply append some fixed string to an
existing image name.
Assigning the UUID discovery path to a tweakable environment variable means
that later steps in the boot process - particularly a boot script can
change the target root partition of a particular Linux boot.
Retargeting the rootfs is an important feature when doing ping/pong
upgrades allowing a boot script to select ping or pong as necessary without
reprogramming the bootloader.
Adding CONFIG_WARP7_ROOT_PART allows a defconfig to specify which partition
is use as the root partition on WaRP7, this is a desirable change in order
to support a different partitioning schemes. The default is the current
partition #2.
In order to sign images with the IMX code-signing-tool (CST) we need to
know the load address of a given image. The best way to derive this load
address is to make it into a define - so that u-boot.cfg contains the
address - which we can then parse when generating the IMX CST headers.
In order to sign images with the IMX code-signing-tool (CST) we need to
know the load address of a given image. The best way to derive this load
address is to make it into a define - so that u-boot.cfg contains the
address - which we can then parse when generating the IMX CST headers.
This patch makes the OPTEE_LOAD_ADDR available via u-boot.cfg for further
parsing by external tools.
Right now a region of 0x300000 bytes is allocated at the end of DRAM for
the purposes of loading an OPTEE firmware inside of it. This patch adds the
printout of the relevant address ranges.
In order to correctly produce an image with a IVT/DCD header we need to
define a CSF in imximage.cfg. We just use the mx7 default here.
All we have to do with this option switched on is "make u-boot.imx" and we
then will get
- u-boot.imx
- u-boot.imx.log
The log file is really important because it gives the addresses for the HAB
that we will require to sign the u-boot image using the CST. Since the
addresses can change this logfile is a critical output.
This patch adds BOOTROM_IVT_HDR_OFFSET at 0xC00. The BootROM expects to
find the IVT header at a particular offset in an i.MX image.
Defining the expected offset of the IVT header in the first-stage BootROM
image format is of use of later stage authentication routines where those
routines continue to follow the first-stage authentication layout.
This patch defines the first stage offset which later patch make use of.
tools/imximage: Fix fruity lack of 0x prefix in DCD Blocks
commit 8519c9c98ad6 ("tools/imximage: use 0x prefix in HAB Blocks line")
adds an 0x prefix to each HAB Block number to make it easier for host tools
to process the HAB Block output, however it neglects to apply the same
prefix to the DCD Blocks directive. You need the DCD Blocks directive if
you are making a u-boot recovery image which the BootROM will accept via
the USB upload utility.
This disparity results in a fruity output like this with HAB Blocks
prefixed but DCD Blocks not prefixed - which is pretty inconsistent.
This patch fixes the difference assuming the original commit was a
legitimate change.
Kever Yang [Tue, 24 Apr 2018 03:27:08 +0000 (11:27 +0800)]
rockchip: rv1108: add ofdata_to_platdata() method for driver
Parse of data in dedicated api instead of in probe().
The clk_set_rate() may be called before the clk driver is probed,
after core support set default clock.
This patch fix system abort issue since: f4fcba5 clk: implement clk_set_defaults()
Kever Yang [Tue, 24 Apr 2018 03:27:07 +0000 (11:27 +0800)]
rockchip: rk3128: add ofdata_to_platdata() method for driver
Parse of data in dedicated api instead of in probe().
The clk_set_rate() may be called before the clk driver is probed,
after core support set default clock.
This patch fix system abort issue since: f4fcba5 clk: implement clk_set_defaults()
Kever Yang [Tue, 24 Apr 2018 03:27:06 +0000 (11:27 +0800)]
rockchip: rk3036: add ofdata_to_platdata() method for driver
Parse of data in dedicated api instead of in probe().
The clk_set_rate() may be called before the clk driver is probed,
after core support set default clock.
This patch fix system abort issue since: f4fcba5 clk: implement clk_set_defaults()
Kever Yang [Wed, 18 Apr 2018 03:13:46 +0000 (11:13 +0800)]
rockchip: rk3188: use DM timer instead of rk_timer
Disable rk_timer as SYS timer and use DM timer instead,
so that we can get a better timer framework, the rk_timer
is going to be clean after we conver to use DM timer or
ARM arch/generic timer.
Kever Yang [Wed, 18 Apr 2018 03:13:45 +0000 (11:13 +0800)]
rockchip: rk3188: add timer3 node
Add dts node for timer3.
Because of the rockchip timer can only KNOWN "dtd_rockchip_rk3368_timer"
with OF_PLATDATA enable, so we override its compatible to
"rockchip,rk3368-timer".
Philipp Tomsich [Wed, 25 Apr 2018 12:07:06 +0000 (14:07 +0200)]
rockchip: timer: add compatible strings for rk3188 and rk3288
The DM driver for ockchip timer blocks is also applicable to the
RK3188 and RK3288 timer blocks: add 'rockchip,rk3188-timer' and
'rockchip,rk3288-timer' to its compatible list to support devices
claiming compatibility with these.
Wadim Egorov [Mon, 19 Mar 2018 15:39:29 +0000 (16:39 +0100)]
rockchip: clk: rk3288: add clk_enable function and support USB HOST0/HSIC
The generic ehci-driver (ehci-generic.c) will try to enable the clocks
listed in the DTSI. If this fails (e.g. due to clk_enable not being
implemented in a driver and -ENOSYS being returned by the clk-uclass),
the driver will bail our and print an error message.
This implements a minimal clk_enable for the RK3288 and supports the
clocks mandatory for the EHCI controllers; as these are enabled by
default we simply return success.
Philipp Tomsich [Wed, 21 Mar 2018 11:18:58 +0000 (12:18 +0100)]
sunxi: improve throughput in the sunxi_mmc driver
Throughput tests have shown the sunxi_mmc driver to take over 10s to
read 10MB from a fast eMMC device due to excessive delays in polling
loops.
This commit restructures the main polling loops to use get_timer(...)
to determine whether a (millisecond) timeout has expired. We choose
not to use the wait_bit function, as we don't need interruptability
with ctrl-c and have at least one case where two bits (one for an
error condition and another one for completion) need to be read and
using wait_bit would have not added to the clarity.
The observed speedup in testing on a A31 is greater than 10x (e.g. a
10MB write decreases from 9.302s to 0.884s).
projects / u-boot.git / log