This patch has some parts connected together:
- Use _gd in bss section which is automatically cleared
Location at SPL_MALLOC_END wasn't cleared at all
- Use MALLOC_F_LEN(early alloc) instead of FULL MALLOC
(mem_malloc_init is not called at all)
- Simplify malloc and stack init.
At the end of SPL addr is malloc area and below is stack
Signed-off-by: Michal Simek <michal.simek@xilinx.com>
Because it is not compatible with DM where
malloc_base has to be available early and init
has to be done in ASM.
Signed-off-by: Michal Simek <michal.simek@xilinx.com>
Microblaze target supports both OF and !OF cases
and from log is not clear which version is running.
Signed-off-by: Michal Simek <michal.simek@xilinx.com>
Compile code with -fPIC to get GOT. Do not build SPL
with fPIC because it increasing SPL size for nothing.
Signed-off-by: Michal Simek <michal.simek@xilinx.com>
Setup gd from ASM to be availalbe for board_init_r.
Setting it up in spl_board_init is too late when
MALLOC is used.
Space for gd is located behind MALLOC area at the end of BRAM.
Signed-off-by: Michal Simek <michal.simek@xilinx.com>
Do not use specific macros for debugging.
Also remove compilation warning:
w+../arch/microblaze/cpu/interrupts.c: In function 'interrupt_handler':
w+../arch/microblaze/cpu/interrupts.c:153:2: warning: format '%x'
expects argument of type 'unsigned int', but argument 2 has type 'void
(*)(void *)' [-Wformat]
w+../arch/microblaze/cpu/interrupts.c:153:2: warning: format '%x'
expects argument of type 'unsigned int', but argument 4 has type 'void
*' [-Wformat]
Signed-off-by: Michal Simek <michal.simek@xilinx.com>
Do not save registers below r1 stack pointer because
it is not checked by stack undeflow is not able to detect
it.
Signed-off-by: Michal Simek <michal.simek@xilinx.com>
This is required for architectures still need manual relocation like avr32, mk68
and others.
Signed-off-by: Andreas Bießmann <andreas.devel@googlemail.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
Tested-by: Michal Simek <michal.simek@xilinx.com>
This results in a much more readable callgraph, because now they
can't be confused with the function having exactly the same name
in the generic mmc code.
Signed-off-by: Siarhei Siamashka <siarhei.siamashka@gmail.com>
Acked-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
This patch allows to configure all the important DRAM parameters in Kconfig.
Signed-off-by: Siarhei Siamashka <siarhei.siamashka@gmail.com>
Acked-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
In addition to the current Android magic settings, allow to optionally use
DDR3 timing parameters, which are tailored for different clock frequencies
and JEDEC speed bins. This should improve reliability and performance.
Adding '+S:CONFIG_DRAM_TIMINGS_DDR3_1066F_1333H=y' to the board defconfig
allows to use timings, which are calculated for the DDR3-1066F speed bin.
A lot of DDR3 chips, which are used in real Allwinner based devices,
support DDR3-1066F speed bin timings.
And adding '+S:CONFIG_DRAM_TIMINGS_DDR3_800E_1066G_1333J=y' should work
with any DDR3 chips, because this targets the slowest JEDEC speed bins.
The vendor magic values are still used by default for DRAM, but board
maintainers now have more flexibility in DRAM timings selection.
Signed-off-by: Siarhei Siamashka <siarhei.siamashka@gmail.com>
Acked-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
On both my A13-OLinuxIno and my A13-OLinuxIno-Micro, the vga output gives an
unstable image when active low v or hsync is used.
The problem seems to be specific to the OLinuxIno A13 (normal & micro)
boards. I've just looked up the schematics and they use an opendrain driver
for the vga sync lines, and with sync pulses it is the logical high->low
edge of the pulse which counts for the timing, which with an active low
sync is being driven by the pull-up, and that simply seems to not drive
it hard enough to get a stable image.
So force v and hsync active high on these boards. independent of what the
modeline says. This fixes the unstable image.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Acked-by: Ian Campbell <ijc@hellion.org.uk>
This is a low-cost Allwinner A20 board with Arduino-style GPIO headers;
it features 1G RAM, 4G NAND flash, 1 micro-SD, 2 USB sockets, 1 micro
USB socket for OTG and another for power in, HDMI, SATA, 5V power for
SATA devices, gigabit Ethernet, an IR receiver, 3.5mm audio out and a
MIPI camera connector.
Like the BananaPi, this board needs GMAC_TX_DELAY set to 3 in order for
GMAC to work reliably at gigabit speeds.
For more details, see: http://linux-sunxi.org/LinkSprite_pcDuino3_Nano
Signed-off-by: Adam Sampson <ats@offog.org>
Acked-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
It turns out that the device_mode_data is rsb specific, rather then slave
specific, so integrate the rsb_set_device_mode() call into rsb_init().
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Acked-by: Ian Campbell <ijc@hellion.org.uk>
CONFIG_TARGET_FOO was only used in board/sunxi/Makefile to select the
dram config for sun5i and sun7i boards and in board/sunxi/gmac.c for some
special handling of the bananapi/bananapro (both sun7i), all sun5i and sun7i
boards have been moved over to using a single dram_sun5i_autoconfig file,
and the tx clk delay handling for the Banana boards now has its own Kconfig.
IOW nothing is using CONFIG_TARGET_FOO anymore, so remove it.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Acked-by: Ian Campbell <ijc@hellion.org.uk>
And use this to set the GMAC Transmit Clock Delay Chain value on Banana
boards, rather then keying of CONFIG_TARGET_FOO.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Acked-by: Ian Campbell <ijc@hellion.org.uk>
Currently we've separate detailed dram settings for all sun5i boards, this
moves them over to using auto dram configuration so that we can get rid of
all the per board dram_foo.c files.
This has been tested on a A10s-Olinuxino, A13-Olinuxino, A13-OlinuxinoM,
mk802-a10s and r7-tv-dongle board.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Acked-by: Ian Campbell <ijc@hellion.org.uk>
We do not need i2c support in the SPL when there is no PMIC (some sun4i
boards), or when the PMIC is not using i2c such as on sun6i and sun8i.
This reduces the SPL size from (e.g.) 21812 to 19260 bytes.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Acked-by: Ian Campbell <ijc@hellion.org.uk>
Add an explanation for how to set up git so that patman can find the alias
file. Fix up the get_maintainers message too.
Reported-by: Scott Wood <scottwood@freescale.com>
Signed-off-by: Simon Glass <sjg@chromium.org>
This brings in a additional small fix which was missed in a recent update
to the README.
Suggested-by: Masahiro Yamada <yamada.m@jp.panasonic.com>
Signed-off-by: Simon Glass <sjg@chromium.org>
Acked-by: Jagannadha Sutradharudu Teki <jagannadh.teki@gmail.com>
True commit lines start at column zero. Anything that is indented
is part of the commit message instead. I noticed this by trying to
run buildman with commit e3a4facdfc
as master, which contained a reference to a Linux commit inside
the commit message. ProcessLine saw that as a genuite commit
line, and thus buildman tried to build it, and died with an
exception because that SHA is not present in the U-Boot tree.
Signed-off-by: Scott Wood <scottwood@freescale.com>
Acked-by: Simon Glass <sjg@chromium.org>
When run with the --dry-run argument patman prints out information
showing what it would do. This information currently doesn't line up
with what patman/git send-email really do. Some basic examples:
- If an email address is addressed via "Series-cc" and "Patch-cc" patman
shows that email address would be CC-ed two times.
- If an email address is addressed via "Series-to" and "Patch-cc" patman
shows that email address would be sent TO and CC-ed.
- If an email address is addressed from a combination of tag aliases,
get_maintainer.pl output, "Series-cc", "Patch-cc", etc patman shows
that the email address would be CC-ed multiple times.
Patman currently does try to send duplicate emails like the --dry-run
output shows, but "git send-email" intelligently removes duplicate
addresses so this patch shouldn't change the non-dry-run functionality.
Change patman's output and email addressing to line up with the
"git send-email" logic. This trims down patman's dry-run output and
prevents confusion about what patman will do when emails are actually
sent.
Signed-off-by: Peter Tyser <ptyser@xes-inc.com>
Acked-by: Simon Glass <sjg@chromium.org>
Tested-by: Simon Glass <sjg@chromium.org>
It is painful to specify the full path to the device tree with the -d
option. It is normally kept in the same directory as U-Boot, so provide
an option to use this by default.
Signed-off-by: Simon Glass <sjg@chromium.org>
If for some reason DMA module fails to reset user oserves only this:
--->---
# dhcp
Trying dwmac.e0018000
FAIL
--->---
This message makes not much sense.
With proposed change error message will be more helpful:
--->---
# dhcp
Trying dwmac.e0018000
DMA reset timeout
FAIL
--->---
For example user may do power toggle to recover board functionality.
Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com>
Cc: Chin Liang See <clsee@altera.com>
Cc: Dinh Nguyen <dinguyen@altera.com>
Cc: Albert Aribaud <albert.u.boot@aribaud.net>
Cc: Tom Rini <trini@ti.com>
Cc: Wolfgang Denk <wd@denx.de>
Cc: Pavel Machek <pavel@denx.de>
Cc: Joe Hershberger <joe.hershberger@gmail.com>
Cc: Ian Campbell <ijc@hellion.org.uk>
Cc: Marek Vasut <marex@denx.de>
Reviewed-by: Tom Rini <trini@ti.com>
Acked-by: Pavel Machek <pavel@denx.de>
Acked-by: Joe Hershberger <joe.hershberger@ni.com>
Signed-off-by: Joe Hershberger <joe.hershberger@ni.com>
This patch adds a phy driver for the Micrel KSZ8895 switch. As the SoC MAC
is directly connected to the switch MAC the link to the switch is always up.
But the KSZ8895 switch can be hardwired in three configuration modes :
- not configurable with eventually an eeprom-stored configuration
- configurable by the mdio/mdc connection (SMI protocol)
- configurable by a SPI connection.
In not configurable mode, the switch starts automatically, but in the
other modes, it must be started programmatically, by writing 1 in
configuration register 1.
We only support the not configurable and mdio/mdc (aka SMI) modes here.
Signed-off-by: Philippe De Muyter <phdm@macqel.be>
Cc: Christian Gmeiner <christian.gmeiner@gmail.com>
Signed-off-by: Joe Hershberger <joe.hershberger@ni.com>
