Some MIPS systems store some board-specific boot configuration
in the U-Boot binary at offset 0x10. This is used by Malta boards
and by Lantiq/Intel SoC's when booting from parallel NOR flash.
Convert the hard-coded values to Kconfig options to remove such
board-specific stuff out of the generic start.S code. This also
deprecates the config option CONFIG_SYS_XWAY_EBU_BOOTCFG.
Signed-off-by: Daniel Schwierzeck <daniel.schwierzeck@gmail.com>
When U-Boot started using SPDX tags we were among the early adopters and
there weren't a lot of other examples to borrow from. So we picked the
area of the file that usually had a full license text and replaced it
with an appropriate SPDX-License-Identifier: entry. Since then, the
Linux Kernel has adopted SPDX tags and they place it as the very first
line in a file (except where shebangs are used, then it's second line)
and with slightly different comment styles than us.
In part due to community overlap, in part due to better tag visibility
and in part for other minor reasons, switch over to that style.
This commit changes all instances where we have a single declared
license in the tag as both the before and after are identical in tag
contents. There's also a few places where I found we did not have a tag
and have introduced one.
Signed-off-by: Tom Rini <trini@konsulko.com>
Some platforms have very limited SRAM to run SPL code, so there may
not be the same amount space for a malloc pool before relocation in
the SPL stage as the normal U-Boot stage.
Make SPL and (the full) U-Boot stage use independent SYS_MALLOC_F_LEN,
so the size of pre-relocation malloc pool can be configured memory
space independently.
Signed-off-by: Andy Yan <andy.yan@rock-chips.com>
Acked-by: Daniel Schwierzeck <daniel.schwierzeck@gmail.com>
Acked-by: Philipp Tomsich <philipp.tomsich@theobroma-systems.com>
Reviewed-by: Philipp Tomsich <philipp.tomsich@theobroma-systems.com>
[fixed up commit-message:]
Signed-off-by: Philipp Tomsich <philipp.tomsich@theobroma-systems.com>
U-Boot has up until now built with -fpic for the MIPS architecture,
producing position independent code which uses indirection through a
global offset table, making relocation fairly straightforward as it
simply involves patching up GOT entries.
Using -fpic does however have some downsides. The biggest of these is
that generated code is bloated in various ways. For example, function
calls are indirected through the GOT & the t9 register:
8f998064 lw t9,-32668(gp)
0320f809 jalr t9
Without -fpic the call is simply:
0f803f01 jal be00fc04 <puts>
This is more compact & faster (due to the lack of the load & the
dependency the jump has on its result). It is also easier to read &
debug because the disassembly shows what function is being called,
rather than just an offset from gp which would then have to be looked up
in the ELF to discover the target function.
Another disadvantage of -fpic is that each function begins with a
sequence to calculate the value of the gp register, for example:
3c1c0004 lui gp,0x4
279c3384 addiu gp,gp,13188
0399e021 addu gp,gp,t9
Without using -fpic this sequence no longer appears at the start of each
function, reducing code size considerably.
This patch switches U-Boot from building with -fpic to building with
-fno-pic, in order to gain the benefits described above. The cost of
this is an extra step during the build process to extract relocation
data from the ELF & write it into a new .rel section in a compact
format, plus the added complexity of dealing with multiple types of
relocation rather than the single type that applied to the GOT. The
benefit is smaller, cleaner, more debuggable code. The relocate_code()
function is reimplemented in C to handle the new relocation scheme,
which also makes it easier to read & debug.
Taking maltael_defconfig as an example the size of u-boot.bin built
using the Codescape MIPS 2016.05-06 toolchain (gcc 4.9.2, binutils
2.24.90) shrinks from 254KiB to 224KiB.
Signed-off-by: Paul Burton <paul.burton@imgtec.com>
Cc: Daniel Schwierzeck <daniel.schwierzeck@gmail.com>
Cc: u-boot@lists.denx.de
Reviewed-by: Daniel Schwierzeck <daniel.schwierzeck@gmail.com>
Tested-by: Daniel Schwierzeck <daniel.schwierzeck@gmail.com>
CFE checks CPU Thread in a different way (using register $22):
mfc0 t1, C0_BCM_CONFIG, 3 # $22
li t2, CP0_CMT_TPID # (1 << 31)
and t1, t2
bnez t1, 2f # if we are running on thread 1, skip init
nop
Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
All MIPS boards that support debug uart are calling debug_uart_init right at
the beginning of board_early_init_f.
Instead of doing that, let's provide a generic call to debug_uart_init right
before the call to board_init_f if debug uart is enabled for boards without
stack in SRAM.
On the other hand, boards with stack in SRAM can call earlier (right before
low level init).
Signed-off-by: Daniel Schwierzeck <daniel.schwierzeck@gmail.com>
Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
This adds a new Kconfig option CONFIG_MIPS_INIT_STACK_IN_SRAM which
a SoC can select if it supports some kind of SRAM. Together with
CONFIG_SYS_INIT_SP_ADDR the initial stack and global data can be
set up in that SRAM. This can be used to provide a C environment
also for lowlevel_init().
Signed-off-by: Daniel Schwierzeck <daniel.schwierzeck@gmail.com>
Move the code for setting up the initial stack and global data
to a macro to be able to use it more than once.
Signed-off-by: Daniel Schwierzeck <daniel.schwierzeck@gmail.com>
Clear cp0 status while preserving implementation specific bits.
Set bits BEV and ERL as the arch specification requires after
a reset or soft-reset exception.
Extend and fix initialization of watch registers. Check if additional
watch register sets are implemented and initialize them too.
Initialize cp0 count as early as possible to get the most
accurate boot timing.
Signed-off-by: Daniel Schwierzeck <daniel.schwierzeck@gmail.com>
When booting from ROM, early exceptions can't be handled
properly. Instead of busy-looping give the developer the
possibilty to examine the situation. Invoke an UHI
exception operation which can be read as unhandled exception
by a hardware debugger if one is attached. If the debugger
doesn't support UHI, the exception is read as unexpected
breakpoint.
Signed-off-by: Daniel Schwierzeck <daniel.schwierzeck@gmail.com>
This adds a compile time option to include code for static
exception vectors. Static exception vectors are only needed,
when the U-Boot entry point is equal to the CPU reset exception
vector address. For instance this is the case when U-Boot is
used as ROM in Qemu or booted from parallel NOR flash. When
U-Boot is booted from RAM (e.g. loaded there by SPL), the
exception vectors need to be setup dynamically, which is done
in follow-up commits.
Signed-off-by: Daniel Schwierzeck <daniel.schwierzeck@gmail.com>
Some systems are configured such that multiple CPUs begin running from
their reset vector following a system reset. If this occurs then U-Boot
will be run on multiple CPUs simultaneously, which causes all sorts of
issues as the multiple instances of U-Boot clobber each other.
Prevent this from happening by simply hanging with an infinite loop if
we run on a CPU whose ID, as determined by GlobalNumber or EBase.CPUNum
as appropriate, is non-zero.
Signed-off-by: Paul Burton <paul.burton@imgtec.com>
The relocate_code function was handling cache maintenance incorrectly.
It copied U-Boot to its new location, flushed the caches & then
proceeded to apply relocations & jump to the new code without flushing
the caches again. This is problematic as the instruction cache could
potentially have already fetched instructions that hadn't had relocs
applied.
Rework this to perform the flush_cache call using the code in the
original copy of U-Boot, after having applied relocations to the new
copy of U-Boot. The new U-Boot can then be jumped to safely once that
cache flush has been performed.
As part of this, since the old U-Boot is used up until after that cache
flush, complexity around loading values from the GOT using a jump & link
instruction & loads from a table is removed. Instead we can simply load
the needed values with PTR_LA fromt the original GOT.
Signed-off-by: Paul Burton <paul.burton@imgtec.com>
During boot we set Config.K0=2 (uncached) such that any accesses to the
kseg0 memory region are performed uncached before the caches are
initialised. This write to the Config register introduces an execution
hazard between it & any following memory accesses (such as the load of
_gp), which we need to clear in order to ensure those memory accesses
are actually performed uncached. Clear this execution hazard with the
insertion of an ehb execution hazard barrier instruction.
Signed-off-by: Paul Burton <paul.burton@imgtec.com>
Map the Global Control Registers (GCRs) provided by the MIPS Coherence
Manager (CM) in preparation for using some of them in later patches.
Signed-off-by: Paul Burton <paul.burton@imgtec.com>
On systems where cache initialisation doesn't require zeroed memory (ie.
systems where CONFIG_SYS_MIPS_CACHE_INIT_RAM_LOAD is not defined)
perform cache initialisation prior to lowlevel_init & DDR
initialisation. This allows for DDR initialisation code to run cached &
thus significantly faster.
Signed-off-by: Paul Burton <paul.burton@imgtec.com>
The coprocessor 0 Config register includes 9 implementation defined
bits, which in some processors do things like enable write combining or
other functionality. We ought not to wipe them to 0 during boot. Rather
than doing so, preserve their value & only clear the bits standardised
by the MIPS architecture.
Signed-off-by: Paul Burton <paul.burton@imgtec.com>
Enable use of the instruction cache immediately after it has been
initialised. This will only take effect if U-Boot was linked to run from
kseg0 rather than kseg1, but when this is the case the data cache
initialisation code will run cached & thus significantly faster.
Signed-off-by: Paul Burton <paul.burton@imgtec.com>
In MIPS assembly there have historically been 2 variants of immediate
addition - the standard "addi" which traps if an overflow occurs, and
the unchecked "addiu" which does not trap on overflow. In release 6 of
the MIPS architecture the trapping variants of immediate addition &
subtraction have been removed. In preparation for supporting MIPSr6,
stop using the trapping instructions from assembly & switch to their
unchecked variants.
Signed-off-by: Paul Burton <paul.burton@imgtec.com>
Correct spelling of "U-Boot" shall be used in all written text
(documentation, comments in source files etc.).
Signed-off-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Heiko Schocher <hs@denx.de>
Reviewed-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Minkyu Kang <mk7.kang@samsung.com>
Fix 32 vs 64 bit load/store instructions. Access CP0_WATCHHI as
32 Bit register. Use 64 Bit register access for clearing gd_data
and copying U-Boot.
Signed-off-by: Daniel Schwierzeck <daniel.schwierzeck@gmail.com>
Argument boot_flags of board_init_f() should be set to 0 as
$a0 may be utilized in lowlevel_init() or mips_cache_reset()
or previous stage boot-loader.
Signed-off-by: Purna Chandra Mandal <purna.mandal@microchip.com>
Signed-off-by: Daniel Schwierzeck <daniel.schwierzeck@gmail.com>
Correctly annotate _start and relocate_code as functions to
produce more readable disassembly code generated by objdump.
Signed-off-by: Daniel Schwierzeck <daniel.schwierzeck@gmail.com>
Unify and move code in arch/mips/cpu/mips[32|64]/ to arch/mips/cpu/.
The CPU specific config.mk files need to remain until
CONFIG_STANDALONE_LOAD_ADDR is converted to a global Kconfig symbol.
Signed-off-by: Paul Burton <paul.burton@imgtec.com>
Signed-off-by: Daniel Schwierzeck <daniel.schwierzeck@gmail.com>
In preparation for sharing a single copy of start.S between mips32 &
mips64, handle setting the KX bit of the cop0 Status register when the
mips32 start.S is built for mips64.
Signed-off-by: Paul Burton <paul.burton@imgtec.com>
Cc: Daniel Schwierzeck <daniel.schwierzeck@gmail.com>
In preparation for sharing a single copy of start.S between mips32 &
mips64, handle mips64 relocations in the mips32 start.S when built for
mips64.
Signed-off-by: Paul Burton <paul.burton@imgtec.com>
Cc: Daniel Schwierzeck <daniel.schwierzeck@gmail.com>
Where the mips32 & mips64 implementations of start.S differ in terms of
access sizes & offsets, use the appropriate macros from asm.h to
abstract those differences away. This is in preparation for sharing a
single copy of start.S between mips32 & mips64.
The exception to this is loads of immediates to be written to the cop0
Config register, which is a 32bit register on mips64 and therefore
constants written to it can be loaded as such.
Signed-off-by: Paul Burton <paul.burton@imgtec.com>
Cc: Daniel Schwierzeck <daniel.schwierzeck@gmail.com>
Implement MIPS specific setup of the gd_t structure to support
pre-relocation malloc. If CONFIG_SYS_MALLOC_F_LEN is specified,
a memory area will be reserved after the initial stack area and
the gd->malloc_base pointer will be initialized.
After this patch the new driver model can be used on MIPS.
Signed-off-by: Daniel Schwierzeck <daniel.schwierzeck@gmail.com>
Support the existing config option CONFIG_SYS_INIT_SP_ADDR on
MIPS. This allows to move the initial stack to other places
than the beginning of RAM.
Signed-off-by: Daniel Schwierzeck <daniel.schwierzeck@gmail.com>
To get correct stack walking and backtrace functionality in gdb,
registers fp and ra should be initialized before calling board_init_f
or board_init_r. Thus allocating stack space and zeroing it as it is
currently done in board.c becomes obsolete.
Signed-off-by: Daniel Schwierzeck <daniel.schwierzeck@gmail.com>
This is in preparation for adapting this board to function correctly on
a physical MIPS Malta board. The board is moved into an "imgtec" vendor
directory at the same time in order to ready us for any other boards
supported by Imagination in the future.
Signed-off-by: Paul Burton <paul.burton@imgtec.com>
Make it similar to the code in mips64/start.S, in order to
allow further unifications.
Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Cc: Daniel Schwierzeck <daniel.schwierzeck@googlemail.com>
Synchronize the code with mips64/start.S, in order to
allow further unifications.
Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Cc: Daniel Schwierzeck <daniel.schwierzeck@googlemail.com>
On the origial Malta boards the REVISION register is
accessible at the 0x1fc00010 address. The contents of
this register gives information about the revision
of the Malta and Core Boards.
This register is used by the Linux kernel to identify
the actual board it is running on. However the register
is not emulated properly by Qemu, so put a hardcoded
value into the flash to make Linux work.
Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Cc: Daniel Schwierzeck <daniel.schwierzeck@googlemail.com>
The code handles relocation entries with the
following relocation types only:
mips32: R_MIPS_REL32
mips64: R_MIPS_REL+R_MIPS_64
xburst: R_MIPS_REL32
Other relocation entries are skipped without
processing. The code must be extended if other
relocation types must be supported.
Add -pie to LDFLAGS_FINAL to generate the .rel.dyn
fixup table, which will be applied to the relocated
image before transferring control to it.
The CONFIG_NEEDS_MANUAL_RELOC is not needed
after the patch, so remove that as well.
Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Signed-off-by: Daniel Schwierzeck <daniel.schwierzeck@gmail.com>
Use the newly introduced symbol __image_copy_end as end address for
relocation of U-Boot image. This is needed for dynamic relocation added
in later patches. This patch obsoletes the symbols uboot_end and
uboot_end_data which are removed.
Signed-off-by: Daniel Schwierzeck <daniel.schwierzeck@gmail.com>
Get the start and end address for clearing BSS from the newly
introduced symbols __bss_start and __bss_end. After GOT is
relocated, those symbols are already pointing to the correct
addresses.
Also optimize the loop by moving the address incrementation
to the delay slot to avoid the initial sub instruction.
Signed-off-by: Daniel Schwierzeck <daniel.schwierzeck@gmail.com>
The difference between the address of the original
and the relocated _GLOBAL_OFFSET_TABLE_ is always
the same as the relocation offset.
The relocation offset is already computed and it is
available in the 's1/t6' register. Use that to adjust
the relocated _G_O_T_ address, instead of calculating
the offset again from the _gp value.
Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Cc: Daniel Schwierzeck <daniel.schwierzeck@googlemail.com>
Cc: Xiangfu Liu <xiangfu@openmobilefree.net>
Saving the parameters in advance unnecessarily complicates
the code. The destination address is already saved in the
's2' register, and that register is not clobbered by the
copy loop. The size of the copied data can be computed
after the copy loop is done.
Change the code to compute the size parameter right
before calling flush_cache, and set the destination
address parameter in the delay slot of the actuall
call.
Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Cc: Daniel Schwierzeck <daniel.schwierzeck@googlemail.com>
The current code uses four instructions and a
temporary register to calculate the relocation
offset and to adjust the gp register.
The relocation offset can be calculated directly
from the CONFIG_SYS_MONITOR_BASE constant and from
the destination address. The resulting offset can
be used to adjust the gp pointer.
This approach makes the code a bit simpler because
it needs two instructions only.
Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Cc: Daniel Schwierzeck <daniel.schwierzeck@googlemail.com>
Save the reused parameters at the beginning
of the 'relocate_code' function. This makes
the function a bit more readable.
Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Cc: Daniel Schwierzeck <daniel.schwierzeck@googlemail.com>
The current code uses two instructions to load
the stack pointer into the 'sp' register.
This results in the following assembly code:
468: 3c088040 lui t0,0x8040
46c: 251d0000 addiu sp,t0,0
The first instuction loads the stack pointer into
the 't0' register then the value of the 'sp' register
is computed by adding zero to the value of the 't0'
register. The same issue present on the 64-bit version
as well:
56c: 3c0c8040 lui t0,0x8040
570: 659d0000 daddiu sp,t0,0
Change the code to load the stack pointer directly
into the 'sp' register. The generated code is functionally
equivalent to the previous version but it is simpler.
32-bit:
468: 3c1d8040 lui sp,0x8040
64-bit:
56c: 3c1d8040 lui sp,0x8040
Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Cc: Daniel Schwierzeck <daniel.schwierzeck@googlemail.com>
The loop code copies more data with one than
necessary due to the 'ble' instuction. Use the
'blt' instruction instead to fix that.
Due to the lack of suitable hardware the Xburst
specific code is compile tested only. However the
change is quite obvious.
Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Cc: Daniel Schwierzeck <daniel.schwierzeck@googlemail.com>
The romReserved and romExcHandle handlers are
accessed by a branch instruction however the
delay slots of those instructions are not filled.
Because the start.S uses the 'noreorder' directive,
the assembler will not fill the delay slots either,
and leads to the following assembly code:
0000056c <romReserved>:
56c: 1000ffff b 56c <romReserved>
00000570 <romExcHandle>:
570: 1000ffff b 570 <romExcHandle>
In the resulting code, the second branch instruction
is placed into the delay slot of the first branch
instruction, which is not allowed on the MIPS
architecture.
Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Cc: Daniel Schwierzeck <daniel.schwierzeck@googlemail.com>
Currently the cache operation mode is hard-coded to
CONF_CM_CACHABLE_NONCOHERENT. This is not appropiate for CPUs or SOCs
which operate at a different mode.
This patch makes the cache operation mode configurable via board config.
Signed-off-by: Daniel Schwierzeck <daniel.schwierzeck@googlemail.com>
Acked-by: Thomas Langer <thomas.langer@lantiq.com>
Signed-off-by: Shinya Kuribayashi <skuribay@pobox.com>
This define is a board-specific config option and should be
renamed to follow the U-Boot naming convention. Additionally,
add an explaining comment for this option.
Signed-off-by: Daniel Schwierzeck <daniel.schwierzeck@googlemail.com>
Acked-by: Thomas Langer <thomas.langer@lantiq.com>
Signed-off-by: Shinya Kuribayashi <skuribay@pobox.com>
Fix style issues and alignments globally. No logical changes.
- Replace C comments with AS line comments where possible
- Use ifndef where possible, rather than if !defined for simplicity
- An instruction executed in a delay slot is now indicated by a leading
space, not by C comment
Signed-off-by: Shinya Kuribayashi <skuribay@pobox.com>
Daniel Schwierzeck
Tom Rini
Andy Yan
Paul Burton
Álvaro Fernández Rojas
Bin Meng
Purna Chandra Mandal
Gabor Juhos
Wolfgang Denk
Daniel Schwierzeck
Shinya Kuribayashi