As of gcc 5.2.1 for Thumb-1, it is not possible any
more to assign gd from C code, as gd is mapped to r9,
and r9 may now be saved in the prolog sequence, and
restored in the epilog sequence, of any C functions.
Therefore arch_setup_gd(), which is supposed to set
r9, may actually have no effect, causing U-Boot to
use a bad address to access GD.
Fix this by never calling arch_setup_gd() for ARM,
and instead setting r9 in arch/arm/lib/crt0.S, to
the value returned by board_init_f_alloc_reserve().
Signed-off-by: Albert ARIBAUD <albert.u.boot@aribaud.net>
Reviewed-by: Simon Glass <sjg@chromium.org>
board_init_f_mem() alters the C runtime environment's
stack it is actually already using. This is not a valid
behaviour within a C runtime environment.
Split board_init_f_mem into C functions which do not alter
their own stack and always behave properly with respect to
their C runtime environment.
Signed-off-by: Albert ARIBAUD <albert.u.boot@aribaud.net>
Acked-by: Thomas Chou <thomas@wytron.com.tw>
There is quite a bit of assembler code that can be removed if we use the
generic global_data setup. Less arch-specific code makes it easier to add
new features and maintain the start-up code.
Drop the unneeded code and adjust the hooks in board_f.c to cope.
Signed-off-by: Simon Glass <sjg@chromium.org>
The current comments need a bit of tweaking since we now support stack
and global_data relocation in SPL. Also add a reference to the README.
For AArch64 this is not implemented, so leave a TODO for this.
Signed-off-by: Simon Glass <sjg@chromium.org>
Reported-by: Tim Harvey <tharvey@gateworks.com>
For ARM architecture, enable the CONFIG_USE_ARCH_MEMSET/MEMCPY,
will highly increase the memset/memcpy performance. This is able
thanks to the ARM multiple register instructions.
Unfortunatelly the relocation is done without the cache enabled,
so it takes some time, but zeroing the BSS memory takes much more
longer, especially for the configs with big static buffers.
A quick test confirms, that the boot time improvement after using
the arch memcpy for relocation has no significant meaning.
The same test confirms that enable the memset for zeroing BSS,
reduces the boot time.
So this patch enables the arch memset for zeroing the BSS after
the relocation process. For ARM boards, this can be enabled
in board configs by defining: 'CONFIG_USE_ARCH_MEMSET'.
This was tested on Trats2.
A quick test with trace. Boot time from start to main_loop() entry:
- ~1384ms - before this change
- ~888ms - after this change
Signed-off-by: Przemyslaw Marczak <p.marczak@samsung.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
Cc: Albert Aribaud <albert.u.boot@aribaud.net>
Cc: Tom Rini <trini@konsulko.com>
At present SPL uses a single stack, either CONFIG_SPL_STACK or
CONFIG_SYS_INIT_SP_ADDR. Since some SPL features (such as MMC and
environment) require a lot of stack, some boards set CONFIG_SPL_STACK to
point into SDRAM. They then set up SDRAM very early, before board_init_f(),
so that the larger stack can be used.
This is an abuse of lowlevel_init(). That function should only be used for
essential start-up code which cannot be delayed. An example of a valid use is
when only part of the SPL code is visible/executable, and the SoC must be set
up so that board_init_f() can be reached. It should not be used for SDRAM
init, console init, etc.
Add a CONFIG_SPL_STACK_R option, which allows the stack to be moved to a new
address before board_init_r() is called in SPL.
The expected SPL flow (for CONFIG_SPL_FRAMEWORK) is documented in the README.
Signed-off-by: Simon Glass <sjg@chromium.org>
For version 1:
Acked-by: Albert ARIBAUD <albert.u.boot@aribaud.net>
Reviewed-by: Stefan Roese <sr@denx.de>
Tested-by: Bo Shen <voice.shen@atmel.com>
Acked-by: Bo Shen <voice.shen@atmel.com>
Acked-by: Heiko Schocher <hs@denx.de>
Tested-by: Heiko Schocher <hs@denx.de>
Signed-off-by: Tom Rini <trini@konsulko.com>
For SPL it is sometimes useful to have a simple malloc() just to permit
driver model to work, in the cases where the full malloc() is not made
available by the board config.
Signed-off-by: Simon Glass <sjg@chromium.org>
Commit 3ff46cc4 fixed exception vectors setting in
the general ARM case, by either copying the exception
and indirect vector tables to normal (0x00000000) or
high (0xFFFF0000) vectors address, or setting VBAR to
U-Boot's base if applicable.
i.MX27 SoC is ARM926E-JS, thus has only normal and
high options, but does not provide RAM at 0xFFFF0000
and has only ROM at 0x00000000; it is therefore not
possible to move or change its exception vectors.
Besides, i.MX27 ROM code does provide an indirect
vectors table but at a non-standard address and with
the reset and reserved vectors missing.
Turn the current vector relocation code into a weak
routine called after relocate_code from crt0, and add
strong version for i.MX27.
Series-Cc: Heiko Schocher <hs@denx.de>
Signed-off-by: Albert ARIBAUD <albert.u.boot@aribaud.net>
Reviewed-by: Stefano Babic <sbabic@denx.de>
Tested-by: Stefano Babic <sbabic@denx.de>
Tested-by: Philippe Reynes <tremyfr@gmail.com>
Tested-by: Philippe Reynes <tremyfr@yahoo.fr>
At present arm defines CONFIG_SYS_GENERIC_GLOBAL_DATA, meaning that
the global_data pointer is set up in board_init_f(). However it is
actually set up before this, it just isn't zeroed.
If we zero the global data before calling board_init_f() then we
don't need to define CONFIG_SYS_GENERIC_GLOBAL_DATA.
Make this change (on arm32 only) to simplify the init process. I
don't have the ability to test aarch64 yet.
Signed-off-by: Simon Glass <sjg@chromium.org>
Acked-by: Tom Rini <trini@ti.com>
Building some arm boards with older binutils may produce errors like this:
---8<---
crt0.S: Assembler messages:
crt0.S:70: Error: register expected, not '#(184)' -- `sub sp,#(184)'
--->8---
Use canonical version of the subtract mnemonic to avoid those issues.
Reported-by: Alexey Smishlayev <alexey@xtech2.lv>
Signed-off-by: Andreas Bießmann <andreas.devel@googlemail.com>
To be more EABI compliant and as a preparation for building
with clang, use the platform-specific r9 register for gd
instead of r8.
note: The FIQ is not updated since it is not used in u-boot,
and under discussion for the time being.
The following checkpatch warning is ignored:
WARNING: Use of volatile is usually wrong: see
Documentation/volatile-considered-harmful.txt
Signed-off-by: Jeroen Hofstee <jeroen@myspectrum.nl>
cc: Albert ARIBAUD <albert.u.boot@aribaud.net>
Commit e05e5de7fa made the 2 1st parameters of
ARM's relocate_code() useless since it moved the code handling them to crt0.S.
So, drop these parameters.
Signed-off-by: Benoît Thébaudeau <benoit.thebaudeau@advansee.com>
The purpose of .globl is to export symbols for ld, not to declare external
symbols.
By the way, use the ENTRY() and ENDPROC() macros to define functions rather than
using .global directly.
Signed-off-by: Benoît Thébaudeau <benoit.thebaudeau@advansee.com>
Note this is a tree-wide change affecting multiple architectures.
At present we use __bss_start, but mostly __bss_end__. This seems
inconsistent and in a number of places __bss_end is used instead.
Change to use __bss_end for the BSS end symbol throughout U-Boot. This
makes it possible to use the asm-generic/sections.h file on all
archs.
Signed-off-by: Simon Glass <sjg@chromium.org>
Move all the C runtime setup code from every start.S
in arch/arm into arch/arm/lib/crt0.S. This covers
the code sequence from setting up the initial stack
to calling into board_init_r().
Also, rewrite the C runtime setup and make functions
board_init_*() and relocate_code() behave according to
normal C semantics (no jumping across the C stack any
more, etc).
Some SPL targets had to be touched because they use
start.S explicitly or for some reason; the relevant
maintainers and custodians are cc:ed.
Signed-off-by: Albert ARIBAUD <albert.u.boot@aribaud.net>
