The assembly for __gnu_thumb1_case_si was taken from upstream gcc and adapted
as width suffix was removed for the add instruction [1].
Signed-off-by: Francis Laniel <francis.laniel@amarulasolutions.com>
Tested-by: Tony Dinh <mibodhi@gmail.com>
[1] 4f181f9c7e/libgcc/config/arm/lib1funcs.S (L2156)
Acked-by: Pali Rohár <pali@kernel.org>
Acked-by: Tony Dinh <mibodhi@gmail.com>
To make sure the panic and the reset messages will go out, console flush() should be used.
Sleep periods do not work in early u-boot phase when timer driver is not initialized yet.
Reference: https://lists.denx.de/pipermail/u-boot/2023-March/512233.html
Signed-off-by: Tony Dinh <mibodhi@gmail.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Stefan Roese <sr@denx.de>
So far we used inline assembly to inject the actual instruction that
triggers the semihosting service. While this sounds elegant, as it's
really only about one instruction, it has some serious downsides:
- We need some barriers in place to force the compiler to issue writes
to a data structure before issuing the trap instruction.
- We need to convince the compiler to actually fill the structures that
we use pointers to.
- We need a memory clobber to avoid the compiler caching the data in
those structures, when semihosting writes data back.
- We need register arguments to make sure the function ID and the
pointer land in the right registers.
This is all doable, but fragile and somewhat cumbersome. Since we now
have a separate function in an extra file anyway, we can do away with
all the magic and just write that in an actual assembly file.
This is much more readable and robust.
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Reviewed-by: Sean Anderson <sean.anderson@seco.com>
Memory used to hold the page tables is allocated from the top of RAM
with no prior initialization and could therefore hold invalid data. As
invalidate_dcache_all() will be called before the MMU has been
initialized and as that function relies indirectly on the page tables
when using CMO_BY_VA_ONLY, these must be in a valid state from their
allocation.
Signed-off-by: Pierre-Clément Tosi <ptosi@google.com>
[ Paul: pick from the Android tree. Fix checkpatch warnings, and rebased
to the upstream. ]
Signed-off-by: Ying-Chun Liu (PaulLiu) <paul.liu@linaro.org>
Cc: Tom Rini <trini@konsulko.com>
Link: e3ceef4230
This converts 2 usages of this option to the non-SPL form, since there is
no SPL_SAVE_PREV_BL_INITRAMFS_START_ADDR defined in Kconfig
Signed-off-by: Simon Glass <sjg@chromium.org>
This converts 2 usages of this option to the non-SPL form, since there is
no SPL_SAVE_PREV_BL_FDT_ADDR defined in Kconfig
Signed-off-by: Simon Glass <sjg@chromium.org>
At this point, the Linux code for "lib1funcs" has changed rather
dramatically. While a resync would be beneficial, it's outside the scope
of what we need here. Simply remove the define for CONFIG_AEABI and
tests for it.
Signed-off-by: Tom Rini <trini@konsulko.com>
_relocate() needs the information in .rela* for self relocation
of the EFI binary.
Fixes: d7ddeb66a6 ("efi_loader: fix building aarch64 EFI binaries")
Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
When building with binutils 2.39 warnings
*_efi.so has a LOAD segment with RWX permissions
occur.
Use SHF_WRITE | SHF_ALLOC as section flags for the .data section.
Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Reviewed-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
While our EFI binaries execute without problems on EDK II they crash on
a Lenovo X13s. Let our binaries look more like what EDK II produces:
* move all writable data to a .data section
* align sections to 4 KiB boundaries (matching EFI page size)
* remove IMAGE_SCN_LNK_NRELOC_OVFL from .reloc section flags
Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Reviewed-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
For the 64bit EFI binaries that we create set the
IMAGE_FILE_LARGE_ADDRESS_AWARE characteristic in the PE-COFF header
to indicate that they can handle addresses above 2 GiB.
Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
We factor out the arch-independent parts of the ARM semihosting
implementation as a common library so that it can be shared
with RISC-V.
Signed-off-by: Kautuk Consul <kconsul@ventanamicro.com>
Reviewed-by: Leo Yu-Chi Liang <ycliang@andestech.com>
It is a bad idea, and more modern toolchains will fail, if you declare
an assembly function to be global and then weak, instead of declaring it
weak to start with. Update assorted assembly files to use the WEAK macro
directly.
Signed-off-by: Tom Rini <trini@konsulko.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Pali Rohár <pali@kernel.org>
The rest of the unmigrated CONFIG symbols in the CONFIG_SYS namespace do
not easily transition to Kconfig. In many cases they likely should come
from the device tree instead. Move these out of CONFIG namespace and in
to CFG namespace.
Signed-off-by: Tom Rini <trini@konsulko.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
Apply commit 534f0fbd65 ("arm64: Fix relocation of env_addr if
POSITION_INDEPENDENT=y") also for 32-bit ARM.
This change fixes crashing of U-Boot on ARMv7 (Omap3 / Cortex-A8) Nokia N900
phone (real HW). Note that qemu emulator of this board with same u-boot.bin
binary has not triggered this crash.
Crash happened after U-Boot printed following debug lines to serial console:
initcall: 0001ea8c (relocated to 8fe0aa8c)
Loading Environment from <NULL>... Using default environment
Destroy Hash Table: 8fe25a98 table = 00000000
Create Hash Table: N=387
Signed-off-by: Pali Rohár <pali@kernel.org>
Currently our semihosting trap function is somewhat fragile: we rely
on the current compiler behaviour to assign the second inline assembly
argument to the next free register (r1/x1), which happens to be the
"addr" argument to the smh_trap() function (per the calling convention).
I guess this is also the reason for the noinline attribute.
Make it explicit what we want: the "addr" argument needs to go into r1,
so we add another register variable. This allows to drop the "noinline"
attribute, so now the compiler beautifully inlines just the trap
instruction directly into the calling function.
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Commit f4b540e25c5c("arm: smh: Fix uninitialized parameters with newer
GCCs") added a memory clobber to the semihosting inline assembly trap
calls, to avoid too eager GCC optimisation: when passing a pointer, newer
compilers couldn't be bothered to actually fill in the structure that it
pointed to, as this data would seemingly never be used (at least from the
compiler's point of view).
But instead of the memory clobber we need to tell the compiler that we are
passing an *array* instead of some generic pointer, this forces the
compiler to actually populate the data structure.
This involves some rather hideous cast, which is best hidden in a macro.
But regardless of that, we actually need the memory clobber, but for two
different reasons: explain them in comments.
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
The ARM semihosting interface uses different trap instructions for
different architectures and instruction sets. So far we were using
AArch64 and ARMv7-M, and had an untested v7-A entry. The latter does
not work when building for Thumb, as can be verified by using
qemu_arm_defconfig, then enabling SEMIHOSTING and SYS_THUMB_BUILD:
==========
{standard input}:35: Error: invalid swi expression
{standard input}:35: Error: value of 1193046 too large for field of 2 bytes at 0
==========
Fix this by providing the recommended instruction[1] for Thumb, and
using the ARM instruction only when not building for Thumb. This also
removes some comment, as QEMU for ARM allows to now test this case.
Also use the opportunity to clean up the inline assembly, and just define
the actual trap instruction inside #ifdef's, to improve readability.
[1] https://developer.arm.com/documentation/dui0471/g/Semihosting/The-semihosting-interface?lang=en
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
The current name is inconsistent with SPL which uses CONFIG_SPL_TEXT_BASE
and this makes it imposible to use CONFIG_VAL().
Rename it to resolve this problem.
Signed-off-by: Simon Glass <sjg@chromium.org>
asm/mach_type.h header and CONFIG_MACH_TYPE macro are arm-specific, so move
related bdinfo logic to arch_setup_bdinfo() in arch/arm/lib/bdinfo.c.
Reviewed-by: Simon Glass <sjg@chromium.org>
Signed-off-by: Ovidiu Panait <ovpanait@gmail.com>
This is required for architectures which do not support compressed kernel images (i.e. ARM64). This is only used while not booting via FIT image.
Signed-off-by: Nathan Barrett-Morrison <nathan.morrison@timesys.com>
Cc: Tom Rini <trini@konsulko.com>
Reviewed-by: Tom Rini <trini@konsulko.com>
There is currently a problem that U-Boot can not work on ARMv4
because assembly imlementations of memcpy() and some other functions
use "bx lr" instruction that is not available on ARMv4 ("mov pc, lr"
should be used instead).
A working preprocessor-based solution to this problem is found in
arch/arm/lib/relocate.S. Move it to the "ret" macro in
arch/arm/include/asm/assembler.h and change all "bx lr" code
to "ret lr" in functions that may run on ARMv4. Linux source code
deals with this problem in the same manner.
v1 -> v2:
Comment update. Pointed out by Andre Przywara.
Signed-off-by: Sergei Antonov <saproj@gmail.com>
CC: Samuel Holland <samuel@sholland.org>
CC: Ye Li <ye.li@nxp.com>
CC: Simon Glass <sjg@chromium.org>
CC: Andre Przywara <andre.przywara@arm.com>
CC: Marek Vasut <marex@denx.de>
CC: Sean Anderson <sean.anderson@seco.com>
CC: Tom Rini <trini@konsulko.com>
Introduce the new Kconfig symbol CONFIG_SPL_SYS_L2_PL310 to allow the
SPL to build cache-pl310.c.
Before this commit, the SPL could enable the PL310 L2 cache [1], but the
cache maintenance functions from cache-pl310.c were only useable for
non-SPL builds.
After enabling the cache one must be able to flush it, too. Thus this
commit allows cache-pl310.c to be included in the SPL build.
[1] See for example arch/arm/mach-imx/cache.c: v7_outer_cache_enable()
Signed-off-by: Philip Oberfichtner <pro@denx.de>
Newer versions of GCC won't initialize parts of structures which don't
appear to be used. This results in uninitialized semihosting parameters
passed via R1. Fix this by marking the inline assembly as clobbering
memory.
Signed-off-by: Sean Anderson <sean.anderson@seco.com>
The generic ARM relocate_code function was using its own function entry
point as a relocation base, and it was obtaining that address by using
the "adr" instruction on that entry point label.
However that label is not just an ordinary label, instead we explicitly
mark it as a function start address. Normally that doesn't change much
(other than for debugging), but when assembled in Thumb mode, newer
versions of the GNU assembler prepare everything for this address being
used as the argument to a "bx" call, so make sure bit 0 is set in there
to mark this function as Thumb code. Of course this doesn't end up very
well when we use this address for the ensuing memcpy operation.
To avoid this problem, and to solve it in a robust way, add an extra
label, which is not marked as a function entry, and use that for the adr
instruction. This lets all assemblers generate the right immediate offset
in the "adr" instruction.
This fixes in particular ARMv7-M ports when using GNU binutils v2.37 or
newer (commit d3e52e120b68 seems to trigger the change in behaviour).
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Reported-by: Jesse Taube <mr.bossman075@gmail.com>
With commit ce39ee28ec ("zynqmp: Do not place u-boot to reserved memory
location"), the function board_get_usable_ram_top() is allocating
MMU_SECTION_SIZE of about 2MB using lmb_alloc(). But we dont have this
much memory in case of mini U-Boot.
Keep these functions which use lmb under CONFIG_LMB so that they are
compiled and used only when LMB is enabled.
Signed-off-by: Ashok Reddy Soma <ashok.reddy.soma@xilinx.com>
Signed-off-by: Michal Simek <michal.simek@xilinx.com>
Signed-off-by: Michal Simek <michal.simek@amd.com>
Link: https://lore.kernel.org/r/75e52def75f573e554a6b177a78504c128cb0c4a.1657183534.git.michal.simek@amd.com
- Make all users of CUSTOM_SYS_INIT_SP_ADDR reference SYS_INIT_SP_ADDR
- Introduce HAS_CUSTOM_SYS_INIT_SP_ADDR to allow for setting the stack
pointer directly, otherwise we use the common calculation.
- On some platforms that were using the standard calculation but did not
set CONFIG_SYS_INIT_RAM_SIZE / CONFIG_SYS_INIT_RAM_ADDR, set them.
- On a small number of platforms that were not subtracting
GENERATED_GBL_DATA_SIZE do so now via the standard calculation.
- CONFIG_SYS_INIT_SP_OFFSET is now widely unused, so remove it from most
board config header files.
Signed-off-by: Tom Rini <trini@konsulko.com>
This would prevent configuring non-secure regs in case gic security
extensions are not emulated in Qemu.
Signed-off-by: Sai Pavan Boddu <sai.pavan.boddu@xilinx.com>
Signed-off-by: Michal Simek <michal.simek@amd.com>
When this option is set then ARM _main() function would call
arch_very_early_init() function at the beginning. It would be before
calling any other functions like debug_uart_init() and also before
initializing C runtime environment.
Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
When u-boot is used as a chain-loaded bootloader (replacing OS kernel),
previous bootloader leaves data in RAM, that can be reused.
For example, on recent arm linux system, when chainloading u-boot,
there are initramfs and fdt in RAM prepared for OS booting. Initramfs
may be modified to store u-boot's payload, thus providing the ability to
use chainloaded u-boot to boot OS without any storage support.
Two config options added:
- SAVE_PREV_BL_INITRAMFS_START_ADDR
saves initramfs start address to 'prevbl_initrd_start_addr' environment
variable
- SAVE_PREV_BL_FDT_ADDR
saves fdt address to 'prevbl_fdt_addr' environment variable
Signed-off-by: Dzmitry Sankouski <dsankouski@gmail.com>
Cc: Tom Rini <trini@konsulko.com>
If a debugger is not attached to U-Boot, semihosting calls will raise a
synchronous abort exception. Try to catch this and disable semihosting
so we can e.g. use another uart if one is available. In the immediate
case, we return an error, since it is not always possible to check for
semihosting beforehand (debug uart, user-initiated load command, etc.)
We handle all possible semihosting instructions, which is probably
overkill. However, we do need to keep track of what instruction set
we're using so that we don't suppress an actual error.
A future enhancement could try to determine semihosting capability by
inspecting the processor state. There's an example of this at [1] for
RISC-V. The equivalent for ARM would inspect the monitor modei
enable/select bits of the DSCR. However, as the article notes, an
exception handler is still helpful in order to catch disconnected
debuggers.
[1] https://tomverbeure.github.io/2021/12/30/Semihosting-on-RISCV.html#avoiding-hangs-when-a-debugger-is-not-connected
Signed-off-by: Sean Anderson <sean.anderson@seco.com>
These functions are intended to support detecting semihosting and
falling back gracefully to alternative implementations. The test starts
by making semihosting call. SYS_ERRNO is chosen because it should not
mutate any state. If this semihosting call results in an exception
(rather than being caught by the debugger), then the exception handler
should call disable_semihosting() and resume execution after the call.
Ideally, this would just be part of semihosting by default, and not a
separate config. However, to reduce space ARM SPL doesn't include
exception vectors by default. This means we can't detect if a
semihosting call failed unless we enable them. To avoid forcing them to
be enabled, we use a separate config option. It might also be possible
to try and detect whether a debugger has enabled (by reading HDE from
DSCR), but I wasn't able to figure out a way to do this from all ELs.
This patch just introduces the generic code to handle detection. The
next patch will implement it for arm64 (but not arm32).
Signed-off-by: Sean Anderson <sean.anderson@seco.com>
To avoid passing around an extra register everywhere, save esr in
pt_regs like the rest. For proper alignment we need to have a second
(unused) register. All the printfs have to be adjusted, since
it's now an unsigned long and not an int.
Signed-off-by: Sean Anderson <sean.anderson@seco.com>
This adds three wrappers around the semihosting commands for reading and
writing to the host console. We use the more standard getc/putc/puts
names instead of readc/writec/write0 for familiarity.
Signed-off-by: Sean Anderson <sean.anderson@seco.com>
This command's functionality is now completely implemented by the
standard fs load command. Convert the vexpress64 boot command (which is
the only user) and remove the implementation.
Signed-off-by: Sean Anderson <sean.anderson@seco.com>
Most U-Boot command deal with start/size instead of start/end. Convert
the "fdt chosen" command to use these semantics as well. The only user
of this subcommand is vexpress, so convert the smhload command to use
this as well. We don't bother renaming the variable in vexpress64's
bootcommand, since it will be rewritten in the next commit.
Signed-off-by: Sean Anderson <sean.anderson@seco.com>
This adds a boot method for loading the next stage from the host. It is
mostly modeled off of spl_load_image_ext. I am not really sure why/how
spl_load_image_fat uses three different methods to load the image, but
the simple case seems to work OK for now.
To control the presence of this boot method, we add a config symbol.
While we're at it, we update the original semihosting config symbol.
I think semihosting has some advantages of other forms of JTAG boot.
Common other ways to boot from JTAG include:
- Implementing DDR initialization through JTAG (typically with dozens of
lines of TCL) and then loading U-Boot. The DDR initialization
typically uses hard-coded register writes, and is not easily adapted
to different boards. BOOT_DEVICE_SMH allows booting with SPL,
leveraging U-Boot's existing DDR initialization code. This is the
method used by NXP's CodeWarrior IDE on Layerscape processors (see
AN12270).
- Loading a bootloader into SDRAM, waiting for it to initialize DDR, and
then loading U-Boot. This is tricky, because the debugger must stop the
boot after the bootloader has completed its work. Trying to load
U-Boot too early can cause failure to boot. This is the method used by
Xilinx with its Zynq(MP) processors.
- Loading SPL with BOOT_DEVICE_RAM and breaking before SPL loads the
image to load U-Boot at the appropriate place. This can be a bit
tricky, because the load address is dependent on the header size. An
elf with symbols must also be used in order to stop at the appropriate
point. BOOT_DEVICE_SMH can be viewed as an extension of this process,
where SPL automatically stops and tells the host where to place the
image.
Signed-off-by: Sean Anderson <sean.anderson@seco.com>
In order to add filesystem support, we will need to be able to seek and
write files. Add the appropriate helper functions.
Signed-off-by: Sean Anderson <sean.anderson@seco.com>
Instead of printing in what are now library functions, try to return a
numeric error code. This also adjust some functions (such as read) to
behave more similarly to read(2). For example, we now return the number
of bytes read instead of failing immediately on a short read.
Signed-off-by: Sean Anderson <sean.anderson@seco.com>
