Thanks for Jernej's JTAG debugging effort, it turns out that the BROM
expects SP_IRQ to be saved and restored, when we want to enter back into
FEL after the SPL's AArch64 stint.
Save and restore SP_IRQ as part of the FEL state handling. The banked
MRS/MSR access to SP_IRQ, without actually being in IRQ mode, was
introduced with the ARMv7 virtualisation extensions. The Arm Cortex-A8
cores used in the A10/A13s or older F1C100s SoCs would not support that,
but this code here is purely in the ARMv8/AArch64 code path, so it's
safe to use unconditionally.
Reported-by: Jernej Skrabec <jernej.skrabec@gmail.com>
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Reviewed-by: Jernej Skrabec <jernej.skrabec@gmail.com>
To be able to return to the BootROM FEL USB debug code, we must restore
the core's state as accurately as possible after the SPL has been run.
Since the BootROM runs in AArch32, but the SPL uses AArch64, this requires
a core reset, which clears the core's state.
So far we were saving and restoring the required registers like SCTLR
and VBAR, but could ignore the interrupt controller's state (GICC), since
that lives in MMIO registers, unaffected by a core reset.
Newer Allwinner SoCs now feature a GICv3 interrupt controller, which keeps
some GIC state in architected system registers, and those are cleared
when we switch back to AArch32.
To enable FEL operation on the Allwinner A523 SoC,
Add AArch32 assembly code to save and restore the ICC_PMR and ICC_IGRPEN1
system registers. The other GICv3 sysregs are either not relevant for the
BROM operation, or haven't been changed from their reset defaults by the
BROM anyway.
This enables FEL operation on the Allwinner A523 family of SoCs.
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Reviewed-by: Jernej Skrabec <jernej.skrabec@gmail.com>
Tom Rini <trini@konsulko.com> says:
This series switches to always using $(PHASE_) in Makefiles when
building rather than $(PHASE_) or $(XPL_). It also starts on documenting
this part of the build, but as a follow-up we need to rename
doc/develop/spl.rst and expand on explaining things a bit.
Link: https://lore.kernel.org/r/20250401225851.1125678-1-trini@konsulko.com
It is confusing to have both "$(PHASE_)" and "$(XPL_)" be used in our
Makefiles as part of the macros to determine when to do something in our
Makefiles based on what phase of the build we are in. For consistency,
bring this down to a single macro and use "$(PHASE_)" only.
Signed-off-by: Tom Rini <trini@konsulko.com>
As current design, only Manager CPU called armv8_setup_psci() before
jump to next stage(such as Linux Kernel), Subordinate CPUs also need
setup psci vector to handle trap request which comes from higher EL
level.
Signed-off-by: Xu Zhang <423756212@qq.com>
[trini: Guard with !CONFIG_XPL_BUILD check]
We currently set both and print both PXN and UXN bits when removing
execution for pages. This happens even in the existing per platform
definitions of 'struct mm_region'.
That's not entirely correct though. For stage-1 translations, if a
platform runs on a translation regime with a single privilege level or the
the translation regime supports two privilege levels and we are not
in EL1&0 with HCR_EL2.{NV, NV1} = {1, 1} only BIT54 (XN) is needed
and BIT53(PXN) is reserved 0.
Currently we support Non-Secure EL2, Non-secure EL2&0 and Non-secure
EL1&0.
We already have get_effective_el() which returns 1 if we are
- Running in EL1 so we assume an EL1 translation regime but without
checking HCR_EL2.{NV, NV1} != {1,1}
- Running in EL2 with HCR_EL2.E2H = 1
The only problem with the above is that if we are in EL1&0 and
HCR_EL2.{NV1, NV} == {1, 1}, then
- Bit[54] holds the PXN instead of the UXN
- The Effective value of UXN is 0
- Bit[53] is RES0
So let's re-use that function and set PXN only when we are in
and EL[2|1]&0 translation regime.
Signed-off-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
To return a 64-bit Allwinner chip back to the 32-bit BootROM code, we
have some embedded AArch32 code that restores the CPU state, before
branching back to the BootROM. At the moment the pointer to the buffer
with that state is located *after* the code, which makes the PC relative
code fragile: adding or removing instructions will change the distance
to that pointer variable.
The "new" Allwinner A523 SoC requires more state to be restored (GICv3
system registers), but we must do that *only* on that SoC. Conditional
compilation sounds like the easiest solution, but would mean that the
distance to that pointer would change.
Solve this rather easily by moving the pointer to the *front* of the
code: we load that pointer in the first instruction, so the distance
would always stay the same. Later in the code we won't need PC relative
addressing anymore, so this code can grow or shrink easily, for instance
due to conditional compilation.
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Reviewed-by: Jernej Skrabec <jernej.skrabec@gmail.com>
For armv8 we are adding proper page permissions for the relocated U-Boot
binary. Add a weak function that can be used across architectures to change
the page permissions
Tested-by: Neil Armstrong <neil.armstrong@linaro.org> # on AML-S905X-CC
Signed-off-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
The ARM ARM (Rev L.a) on section 8.17.1 describes the cases where
break-before-make is required when changing live page tables.
Since we can use a function to tweak block and page permissions,
where BBM is not required split the existing mmu_change_region_attr()
into two functions and create one that doesn't require BBM. Subsequent
patches will use the new function to map the U-Boot binary with proper
page permissions.
While at it add function descriptions in their header files.
Signed-off-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Upcoming patches are switching the memory mappings to RW, RO, RX
after the U-Boot binary and its data are relocated. Add
annotations in the linker scripts to and mark text, data, rodata
sections and align them to a page boundary.
It's worth noting that .efi_runtime memory permissions are left
untouched for now. There's two problems with EFI currently.
The first problem is that we bundle data, rodata and text in a single
.efi_runtime section which also must be close to .text for now.
As a result we also dont change the permissions for anything contained
in CPUDIR/start.o. In order to fix that we have to decoule .text_rest,
.text and .efi_runtime and have the runtime services on their own
section with proper memory permission annotations (efi_rodata etc).
The efi runtime regions (.efi_runtime_rel) can be relocated by the OS when
the latter is calling SetVirtualAddressMap. Which means we have to
configure those pages as RX for U-Boot but convert them to RWX just before
ExitBootServices. It also needs extra code in efi_tuntime relocation
code since R_AARCH64_NONE are emitted as well if we page align the
section.
Due to the above ignore EFI for now and fix it later once we have the
rest in place.
Acked-by: Jerome Forissier <jerome.forissier@linaro.org>
Tested-by: Neil Armstrong <neil.armstrong@linaro.org> # on AML-S905X-CC
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Separate setjmp.h into an architecture independent part and an architecture
specific part. This simplifies moving from using struct jmp_buf_data
directly to using type jmp_buf in our code which is the C compliant way.
Reviewed-by: Jerome Forissier <jerome.forissier@linaro.org>
Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Introduce a new symbol in the beginning of .data section in
the common ARMv8 linker script and use that as a reference
for data save and restore.
Previously, the code would rely on calculating the start of
the .data section address via data size, however, we observed
that the data size does not really reflect the SPL mapped
addresses.
In our case, the binman_sym section size was not included in
the data size, which will result in a wrong address for the
.data start section, which prevents us from properly saving
and restoring SPL data.
This approach skips the calculation for the starting address
of the .data section, and instead just defines the beginning
address of the .data section and calling the symbol as needed,
in which we think as a simpler and much more robust method.
Signed-off-by: Alif Zakuan Yuslaimi <alif.zakuan.yuslaimi@altera.com>
Signed-off-by: Tien Fong Chee <tien.fong.chee@altera.com>
Reviewed-by: Tien Fong Chee <tien.fong.chee@altera.com>
Add empty weak assembler function armv8_switch_to_el2_prep() which is
jumped to just before U-Boot determines which EL it is running in and
decides which path to take to boot the Linux kernel.
This weak function is meant to be used by architecture specific code
to implement jump to a firmware blob, which then returns right past
this weak function and continues execution of U-Boot code which then
boots the Linux kernel. One example of such use case is when U-Boot
jump tp TFA BL31, which switches from EL3 to EL2 and then returns to
U-Boot code newly running in EL2 and starts the Linux kernel.
The weak function is called with caches already disabled and DM shut
down. Any preparatory work or even loading of more data must be done
in board_prep_linux(), this hook is meant only for the final jump to
the firmware and return to U-Boot before booting Linux.
Reviewed-by: Tom Rini <trini@konsulko.com>
Signed-off-by: Marek Vasut <marek.vasut+renesas@mailbox.org>
Add sysinfo platform driver for all armv8 platforms to retrieve
hardware information on processor and cache.
Signed-off-by: Raymond Mao <raymond.mao@linaro.org>
Reviewed-by: Simon Glass <sjg@chromium.org>
Usually ARMv8 platforms allow unaligned access for Normal memory. But
some chips might not allow it by default, having SCTLR.A bit set to 1
before U-Boot execution. One such example is Exynos850 SoC. As
allow_unaligned() is not implemented for ARMv8 at the moment, its __weak
implementation is used, which does nothing. That might lead to unaligned
access abort, for example when running EFI selftest. Fix that by
implementing allow_unaligned() for ARMv8.
The issue was found when running EFI selftest on E850-96 board
(Exynos850 based):
=> bootefi selftest $fdtcontroladdr
...
Executing 'HII database protocols'
"Synchronous Abort" handler, esr 0x96000021, far 0xbaac0991
...
resetting ...
Unaligned abort happens in u16_strnlen(), which is called from
efi_hii_sibt_string_ucs2_block_next():
u16_strlen(blk->string_text)
where 'blk' type is struct efi_hii_sibt_string_ucs2_block. Because this
struct is packed, doing "->string_text" makes 'blk' address incremented
by 1 byte, which makes it unaligned. Although allow_unaligned() was
called in efi_init_early() before EFI selftest execution, it wasn't
implemented for ARMv8 CPUs, so data abort happened.
Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org>
Patrick Rudolph <patrick.rudolph@9elements.com> says:
Based on the existing work done by Simon Glass this series adds
support for booting aarch64 devices using ACPI only.
As first target QEMU SBSA support is added, which relies on ACPI
only to boot an OS. As secondary target the Raspberry Pi4 was used,
which is broadly available and allows easy testing of the proposed
solution.
The series is split into ACPI cleanups and code movements, adding
Arm specific ACPI tables and finally SoC and mainboard related
changes to boot a Linux on the QEMU SBSA and RPi4. Currently only the
mandatory ACPI tables are supported, allowing to boot into Linux
without errors.
The QEMU SBSA support is feature complete and provides the same
functionality as the EDK2 implementation.
The changes were tested on real hardware as well on QEMU v9.0:
qemu-system-aarch64 -machine sbsa-ref -nographic -cpu cortex-a57 \
-pflash secure-world.rom \
-pflash unsecure-world.rom
qemu-system-aarch64 -machine raspi4b -kernel u-boot.bin -cpu cortex-a72 \
-smp 4 -m 2G -drive file=raspbian.img,format=raw,index=0 \
-dtb bcm2711-rpi-4-b.dtb -nographic
Tested against FWTS V24.03.00.
Known issues:
- The QEMU rpi4 support is currently limited as it doesn't emulate PCI,
USB or ethernet devices!
- The SMP bringup doesn't work on RPi4, but works in QEMU (Possibly
cache related).
- PCI on RPI4 isn't working on real hardware since the pcie_brcmstb
Linux kernel module doesn't support ACPI yet.
Link: https://lore.kernel.org/r/20241023132116.970117-1-patrick.rudolph@9elements.com
Update the generic entry point code to support the ACPI parking protocol.
The ACPI parking protocol can be used when PSCI is not available to bring
up secondary CPU cores.
When enabled secondary CPUs will enter U-Boot proper and spin in their own
4KiB reserved memory page, which also acts as mailbox with the OS to
release the CPU.
TEST: Boots all CPUs on qemu-system-aarch64 -machine raspi4b
Signed-off-by: Patrick Rudolph <patrick.rudolph@9elements.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
On Arm platforms that use ACPI they cannot rely on the "spin-table"
CPU bringup usually defined in the FDT. Thus implement the
'ACPI Multi-processor Startup for ARM Platforms', also referred to as
'ACPI parking protocol'.
The ACPI parking protocol works similar to the spin-table mechanism, but
the specification also covers lots of shortcomings of the spin-table
implementations.
Every CPU defined in the ACPI MADT table has it's own 4K page where the
spinloop code and the OS mailbox resides. When selected the U-Boot board
code must make sure that the secondary CPUs enter u-boot after relocation
as well, so that they can enter the spinloop code residing in the ACPI
parking protocol pages.
The OS will then write to the mailbox and generate an IPI to release the
CPUs from the spinloop code.
For now it's only implemented on ARMv8, but can easily be extended to
other platforms, like ARMv7.
TEST: Boots all CPUs on qemu-system-aarch64 -machine raspi4b
Signed-off-by: Patrick Rudolph <patrick.rudolph@9elements.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
Cc: Simon Glass <sjg@chromium.org>
Cc: Tom Rini <trini@konsulko.com>
The EFI memory allocations are now being done through the LMB module,
and hence the memory map is maintained by the LMB module. Use the
lmb_arch_add_memory() API function to add the usable RAM memory to the
LMB's memory map.
Signed-off-by: Sughosh Ganu <sughosh.ganu@linaro.org>
Simon Glass <sjg@chromium.org> says:
When the SPL build-phase was first created it was designed to solve a
particular problem (the need to init SDRAM so that U-Boot proper could
be loaded). It has since expanded to become an important part of U-Boot,
with three phases now present: TPL, VPL and SPL
Due to this history, the term 'SPL' is used to mean both a particular
phase (the one before U-Boot proper) and all the non-proper phases.
This has become confusing.
For a similar reason CONFIG_SPL_BUILD is set to 'y' for all 'SPL'
phases, not just SPL. So code which can only be compiled for actual SPL,
for example, must use something like this:
#if defined(CONFIG_SPL_BUILD) && !defined(CONFIG_TPL_BUILD)
In Makefiles we have similar issues. SPL_ has been used as a variable
which expands to either SPL_ or nothing, to chose between options like
CONFIG_BLK and CONFIG_SPL_BLK. When TPL appeared, a new SPL_TPL variable
was created which expanded to 'SPL_', 'TPL_' or nothing. Later it was
updated to support 'VPL_' as well.
This series starts a change in terminology and usage to resolve the
above issues:
- The word 'xPL' is used instead of 'SPL' to mean a non-proper build
- A new CONFIG_XPL_BUILD define indicates that the current build is an
'xPL' build
- The existing CONFIG_SPL_BUILD is changed to mean SPL; it is not now
defined for TPL and VPL phases
- The existing SPL_ Makefile variable is renamed to SPL_
- The existing SPL_TPL Makefile variable is renamed to PHASE_
It should be noted that xpl_phase() can generally be used instead of
the above CONFIGs without a code-space or run-time penalty.
This series does not attempt to convert all of U-Boot to use this new
terminology but it makes a start. In particular, renaming spl.h and
common/spl seems like a bridge too far at this point.
The series is fully bisectable. It has also been checked to ensure there
are no code-size changes on any commit.
Use PHASE_ as the symbol to select a particular XPL build. This means
that SPL_TPL_ is no-longer set.
Update the comment in bootstage to refer to this symbol, instead of
SPL_
Signed-off-by: Simon Glass <sjg@chromium.org>
Separate BSS is current mandatory on armv8 but this is not useful for
early boot phases. Add support for the combined BSS.
Use an #ifdef to avoid using CONFIG_SPL_BSS_START_ADDR which is not
valid in this case.
Signed-off-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
In some cases we might want to map some memory region after enabling
caches. Introduce a new helper for this.
Reviewed-by: Neil Armstrong <neil.armstrong@linaro.org>
Signed-off-by: Caleb Connolly <caleb.connolly@linaro.org>
So CONFIG_SYS_BIG_ENDIAN is our cross architecture option for
selecting machine endian, while the old CONFIG_CPU_BIG_ENDIAN
is defined by Arc only.
Use it whenever possible to ensure big endian code path is enabled
for all possible big endian machines.
Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com>
Add a basic software implementation of the ARM64 pagetable walker. This
can be used for debugging U-Boot's pagetable, as well as dumping the
pagetable from the previous bootloader stage if it used one (by reading
out the ttbr address).
One can either call dump_pagetable() to print the pagetable to the
console with the default formatter, or implement their own pagetable
handler using walke_pagetable() with a custom pte_walker_cb_t callback.
All of the added code is discarded when unused, hence there is no need
to add an additional Kconfig option for this.
Signed-off-by: Caleb Connolly <caleb.connolly@linaro.org>
commit 6e2228fb05 ("Merge patch series "Clean up arm linker scripts")
was cleaning up linker scripts for armv7 and v8 but was leaving
_end and __secure_stack_start/end.
commit d0b5d9da5d ("arm: make _end compiler-generated")
was moving _end to be compiler generated. _end is defined as c variable
in its own section to force the compiler emit relative a reference.
However, defining those in the linker script will do the same thing
since [0].
So let's remove the special sections from the linker scripts, the
variable definitions from sections.c and define them as a symbols.
It's worth noting that _image_binary_end symbol is now redundant and
can be removed in the future.
- SPL
The .end section has been removed from the new binary
[ 5] .end
PROGBITS 00000000fffdf488 000000000002f488 0
0000000000000000 0000000000000000 0 1
[0000000000000003]: WRITE, ALLOC
$~ bloat-o-meter kria_old/spl/u-boot-spl krina_new/spl/u-boot-spl
add/remove: 0/0 grow/shrink: 0/0 up/down: 0/0 (0)
Function old new delta
Total: Before=115980, After=115980, chg +0.00%
$~ readelf -sW kria_old/u-boot kria_new/u-boot | grep -w _end
12047: 000000000813a0f0 0 OBJECT GLOBAL DEFAULT 11 _end
12047: 000000000813a118 0 NOTYPE GLOBAL DEFAULT 11 _end
$~ readelf -sW kria_old/spl/u-boot-spl kria_new/spl/u-boot-spl | grep -w _end
1605: 00000000fffdf488 0 OBJECT GLOBAL DEFAULT 5 _end
1603: 00000000fffdf498 0 NOTYPE GLOBAL DEFAULT 4 _end
$~ readelf -sW old/u-boot new/u-boot | grep -w _end
8847: 0000000000103710 0 OBJECT GLOBAL DEFAULT 11 _end
8847: 0000000000103738 0 NOTYPE GLOBAL DEFAULT 11 _end
$~ readelf -sW old_v7/u-boot new_v7/u-boot | grep -w _end
10638: 000da824 0 OBJECT GLOBAL DEFAULT 10 _end
10637: 000da84c 0 NOTYPE GLOBAL DEFAULT 10 _end
- For both QEMU instances
$~ bloat-o-meter old/u-boot new/u-boot
add/remove: 0/0 grow/shrink: 1/0 up/down: 20/0 (20)
Function old new delta
version_string 50 70 +20
Total: Before=656915, After=656935, chg +0.00%
[0] binutils commit 6b3b0ab89663 ("Make linker assigned symbol dynamic only for shared object")
Signed-off-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
As part of bringing the master branch back in to next, we need to allow
for all of these changes to exist here.
Reported-by: Jonas Karlman <jonas@kwiboo.se>
Signed-off-by: Tom Rini <trini@konsulko.com>
When bringing in the series 'arm: dts: am62-beagleplay: Fix Beagleplay
Ethernet"' I failed to notice that b4 noticed it was based on next and
so took that as the base commit and merged that part of next to master.
This reverts commit c8ffd1356d, reversing
changes made to 2ee6f3a5f7.
Reported-by: Jonas Karlman <jonas@kwiboo.se>
Signed-off-by: Tom Rini <trini@konsulko.com>
Polling cntpct_el0 in a tight loop for delays is inefficient.
This is particularly apparent on Arm FVPs, which do not simulate
real time, meaning that a 1s sleep can take a couple of orders
of magnitude longer to execute in wall time.
If running at EL2 or above (where CNTHCTL_EL2 is available), enable
the cntpct_el0 event stream temporarily and use wfe to implement
the delay more efficiently. The event period is chosen as a
trade-off between efficiency and the fact that Arm FVPs do not
typically simulate real time.
This is only implemented for Armv8 boards, where an architectural
timer exists, and only enabled by default for the ARCH_VEXPRESS64
board family.
Signed-off-by: Peter Hoyes <Peter.Hoyes@arm.com>
Reviewed-by: Andre Przywara <andre.przywara@arm.com>
Remove <common.h> from the remainder of the files under arch/arm and
when needed add missing include files directly.
Signed-off-by: Tom Rini <trini@konsulko.com>
Remove <common.h> from all fsl-layerscape related files and when needed
add missing include files directly.
Acked-by: Peng Fan <peng.fan@nxp.com>
Signed-off-by: Tom Rini <trini@konsulko.com>
Convert UTF-8 chars to ASCII in cases where make sense. No Copyright or
names are converted.
Signed-off-by: Michal Simek <michal.simek@amd.com>
Reviewed-by: Tom Rini <trini@konsulko.com>
Acked-by: Marek Behún <kabel@kernel.org>
The implementation of map_range() creates the requested mapping by
walking the page tables, iterating over multiple PTEs and/or descending
into existing table mappings as needed. When doing so, it assumes any
pre-existing valid PTE to be a table mapping. This assumption is wrong
if the platform code attempts to successively map two overlapping ranges
where the latter intersects a block mapping created for the former.
As a result, map_range() treats the existing block mapping as a table
mapping and descends into it i.e. starts interpreting the
previously-mapped range as an array of PTEs, writing to them and
potentially even descending further (extra fun with MMIO ranges!).
Instead, pass any valid non-table mapping to split_block(), which
ensures that it actually was a block mapping (calls panic() otherwise)
before splitting it.
Fixes: 41e2787f5e ("arm64: Reduce add_map() complexity")
Signed-off-by: Pierre-Clément Tosi <ptosi@google.com>
Tested-by: Fabio Estevam <festevam@gmail.com>
Tested-by: Hiago De Franco <hiago.franco@toradex.com> # Toradex Verdin AM62
Reviewed-by: Marc Zyngier <maz@kernel.org>
Ilias Apalodimas <ilias.apalodimas@linaro.org> says:
The arm linker scripts had a mix of symbols and C defined variables in an
effort to emit relative references instead of absolute ones e.g [0]. A
linker bug prevented us from doing so [1] -- fixed since 2016.
This has led to confusion over the years, ending up with mixed section
definitions. Some sections are defined with overlays and different
definitions between v7 and v8 architectures.
For example __efi_runtime_rel_start/end is defined as a linker symbol for
armv8 and a C variable in armv7.
Linker scripts nowadays can emit relative references, as long as the symbol
definition is contained within the section definition. So let's switch most
of the C defined variables and clean up the arm sections.c file.
There's still a few symbols remaining -- __secure_start/end,
__secure_stack_start/end and __end which can be cleaned up
in a followup series.
For both QEMU v7/v8 bloat-o-meter shows now size difference
$~ ./scripts/bloat-o-meter u-boot u-boot.new
add/remove: 0/0 grow/shrink: 0/0 up/down: 0/0 (0)
Function old new delta
Total: Before=798861, After=798861, chg +0.00%
The symbols seem largely unchanged apart from a difference in .bss
as well as the emited sections and object types of the affected variables.
On the output below the first value is from -next and the second comes from
-next + this patchset. The .bss_start/end sections have disappeared from
the newer binaries.
# For example on QEMU v8:
efi_runtime_start
7945: 0000000000000178 0 OBJECT GLOBAL DEFAULT 2 __efi_runtime_start
7942: 0000000000000178 0 NOTYPE GLOBAL DEFAULT 2 __efi_runtime_start
efi_runtime_stop
9050: 0000000000000d38 0 OBJECT GLOBAL DEFAULT 2 __efi_runtime_stop
9047: 0000000000000d38 0 NOTYPE GLOBAL DEFAULT 2 __efi_runtime_stop
__efi_runtime_rel_start
7172: 00000000000dc2f0 0 OBJECT GLOBAL DEFAULT 10 __efi_runtime_rel_start
7169: 00000000000dc2f0 0 NOTYPE GLOBAL DEFAULT 10 __efi_runtime_rel_start
__efi_runtime_rel_stop
7954: 00000000000dc4a0 0 OBJECT GLOBAL DEFAULT 10 __efi_runtime_rel_stop
7951: 00000000000dc4a0 0 NOTYPE GLOBAL DEFAULT 10 __efi_runtime_rel_stop
__rel_dyn_start
7030: 00000000000dc4a0 0 OBJECT GLOBAL DEFAULT 11 __rel_dyn_start
7027: 00000000000dc4a0 0 NOTYPE GLOBAL DEFAULT 11 __rel_dyn_start
__rel_dyn_end
8959: 0000000000102b10 0 OBJECT GLOBAL DEFAULT 12 __rel_dyn_end
8956: 0000000000102b10 0 NOTYPE GLOBAL DEFAULT 11 __rel_dyn_end
image_copy_start
9051: 0000000000000000 0 OBJECT GLOBAL DEFAULT 1 __image_copy_start
9048: 0000000000000000 0 NOTYPE GLOBAL DEFAULT 1 __image_copy_start
image_copy_end
7467: 00000000000dc4a0 0 OBJECT GLOBAL DEFAULT 11 __image_copy_end
7464: 00000000000dc4a0 0 NOTYPE GLOBAL DEFAULT 11 __image_copy_end
bss_start
12: 0000000000102b10 0 SECTION LOCAL DEFAULT 12 .bss_start
8087: 0000000000000018 0 NOTYPE GLOBAL DEFAULT 1 _bss_start_ofs
8375: 0000000000102b10 0 OBJECT GLOBAL DEFAULT 12 __bss_start
8084: 0000000000000018 0 NOTYPE GLOBAL DEFAULT 1 _bss_start_ofs
8372: 0000000000102b10 0 NOTYPE GLOBAL DEFAULT 12 __bss_start
bss_end
14: 000000000010bc30 0 SECTION LOCAL DEFAULT 14 .bss_end
7683: 000000000010bc30 0 OBJECT GLOBAL DEFAULT 14 __bss_end
8479: 0000000000000020 0 NOTYPE GLOBAL DEFAULT 1 _bss_end_ofs
7680: 000000000010bbb0 0 NOTYPE GLOBAL DEFAULT 12 __bss_end
8476: 0000000000000020 0 NOTYPE GLOBAL DEFAULT 1 _bss_end_ofs
# For QEMU v7:
efi_runtime_start
10703: 000003bc 0 OBJECT GLOBAL DEFAULT 2 __efi_runtime_start
10699: 000003c0 0 NOTYPE GLOBAL DEFAULT 2 __efi_runtime_start
efi_runtime_stop
11796: 000012ec 0 OBJECT GLOBAL DEFAULT 2 __efi_runtime_stop
11792: 000012ec 0 NOTYPE GLOBAL DEFAULT 2 __efi_runtime_stop
__efi_runtime_rel_start
9937: 000c40dc 0 OBJECT GLOBAL DEFAULT 8 __efi_runtime_rel_start
9935: 000c40dc 0 NOTYPE GLOBAL DEFAULT 9 __efi_runtime_rel_start
__efi_runtime_rel_stop
10712: 000c41dc 0 OBJECT GLOBAL DEFAULT 10 __efi_runtime_rel_stop
10708: 000c41dc 0 NOTYPE GLOBAL DEFAULT 9 __efi_runtime_rel_stop
__rel_dyn_start
9791: 000c41dc 0 OBJECT GLOBAL DEFAULT 10 __rel_dyn_start
9789: 000c41dc 0 NOTYPE GLOBAL DEFAULT 10 __rel_dyn_start
__rel_dyn_end
11708: 000da5f4 0 OBJECT GLOBAL DEFAULT 10 __rel_dyn_end
11704: 000da5f4 0 NOTYPE GLOBAL DEFAULT 10 __rel_dyn_end
image_copy_start
448: 0000177c 0 NOTYPE LOCAL DEFAULT 3 _image_copy_start_ofs
11797: 00000000 0 OBJECT GLOBAL DEFAULT 1 __image_copy_start
445: 0000177c 0 NOTYPE LOCAL DEFAULT 3 _image_copy_start_ofs
11793: 00000000 0 NOTYPE GLOBAL DEFAULT 1 __image_copy_start
image_copy_end
450: 00001780 0 NOTYPE LOCAL DEFAULT 3 _image_copy_end_ofs
10225: 000c41dc 0 OBJECT GLOBAL DEFAULT 10 __image_copy_end
447: 00001780 0 NOTYPE LOCAL DEFAULT 3 _image_copy_end_ofs
10222: 000c41dc 0 NOTYPE GLOBAL DEFAULT 10 __image_copy_end
bss_start
11: 000c41dc 0 SECTION LOCAL DEFAULT 11 .bss_start
11124: 000c41dc 0 OBJECT GLOBAL DEFAULT 11 __bss_start
11120: 000c41dc 0 NOTYPE GLOBAL DEFAULT 11 __bss_start
bss_end
13: 000cbbf8 0 SECTION LOCAL DEFAULT 13 .bss_end
10442: 000cbbf8 0 OBJECT GLOBAL DEFAULT 13 __bss_end
10439: 000cbbf8 0 NOTYPE GLOBAL DEFAULT 11 __bss_end
It's worth noting that since the efi regions are affected by the change, booting
with EFI is preferable while testing. Booting the kernel only should be enough
since the efi stub and the kernel proper do request boottime and runtime
services respectively.
Something along the lines of
> virtio scan && load virtio 0 $kernel_addr_r Image && bootefi $kernel_addr_r
will work for QEMU aarch64.
Tested platforms:
- QEMU aarch64
- Xilinx kv260 kria starter kit & zynq
- QEMU armv7
- STM32MP157C-DK2
[0] commit 3ebd1cbc49 ("arm: make __bss_start and __bss_end__ compiler-generated")
[1] binutils commit 6b3b0ab89663 ("Make linker assigned symbol dynamic only for shared object")
Previous patches cleaning up linker symbols, also merged any explicit
. = ALIGN(x); into section definitions -- e.g
.bss ALIGN(x) : instead of
. = ALIGN(x);
. bss : {...}
However, if the output address is not specified then one will be chosen
for the section. This address will be adjusted to fit the alignment
requirement of the output section following the strictest alignment of
any input section contained within the output section. So let's get rid
of the redundant ALIGN directives when they are not needed.
While at add comments for the alignment of __bss_start/end since our
C runtime setup assembly assumes that __bss_start - __bss_end will be
a multiple of 4/8 for armv7 and armv8 respectively.
It's worth noting that the alignment is preserved on .rel.dyn for
mach-zynq which was explicitly aligning that section on an 8b
boundary instead of 4b one.
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
image_copy_start/end are defined as c variables in order to force the compiler
emit relative references. However, defining those within a section definition
will do the same thing since [0].
So let's remove the special sections from the linker scripts, the
variable definitions from sections.c and define them as a symbols within
a section.
[0] binutils commit 6b3b0ab89663 ("Make linker assigned symbol dynamic only for shared object")
Suggested-by: Sam Edwards <CFSworks@gmail.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Tested-by: Sam Edwards <CFSworks@gmail.com> # Binary output identical
Signed-off-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
commit 47bd65ef05 ("arm: make __rel_dyn_{start, end} compiler-generated")
were moving the __rel_dyn_start/end on c generated variables that were
injected in their own sections. The reason was that we needed relative
relocations for position independent code and linker bugs back then
prevented us from doing so [0].
However, the linker documentation pages states that symbols that are
defined within a section definition will create a relocatable
type with the value being a fixed offset from the base of a section [1].
[0] binutils commit 6b3b0ab89663 ("Make linker assigned symbol dynamic only for shared object")
[1] https://sourceware.org/binutils/docs/ld/Expression-Section.html
Suggested-by: Sam Edwards <CFSworks@gmail.com>
Reviewed-by: Sam Edwards <CFSworks@gmail.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Tested-by: Sam Edwards <CFSworks@gmail.com> # Binary output identical
Signed-off-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
__efi_runtime_rel_start/end are defined as c variables for arm7 only in
order to force the compiler emit relative references. However, defining
those within a section definition will do the same thing since [0].
On top of that the v8 linker scripts define it as a symbol.
So let's remove the special sections from the linker scripts, the
variable definitions from sections.c and define them as a symbols within
the correct section.
[0] binutils commit 6b3b0ab89663 ("Make linker assigned symbol dynamic only for shared object")
Suggested-by: Sam Edwards <CFSworks@gmail.com>
Reviewed-by: Sam Edwards <CFSworks@gmail.com>
Tested-by: Sam Edwards <CFSworks@gmail.com> # Binary output identical
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
commit 3ebd1cbc49 ("arm: make __bss_start and __bss_end__ compiler-generated")
and
commit f84a7b8f54 ("ARM: Fix __bss_start and __bss_end in linker scripts")
were moving the bss_start/end on c generated variables that were
injected in their own sections. The reason was that we needed relative
relocations for position independent code and linker bugs back then
prevented us from doing so [0].
However, the linker documentation pages states that symbols that are
defined within a section definition will create a relocatable type with
the value being a fixed offset from the base of a section [1].
So let's start cleaning this up starting with the bss_start and bss_end
variables. Convert them into symbols within the .bss section definition.
[0] binutils commit 6b3b0ab89663 ("Make linker assigned symbol dynamic only for shared object")
[1] https://sourceware.org/binutils/docs/ld/Expression-Section.html
Tested-by: Caleb Connolly <caleb.connolly@linaro.org> # Qualcomm sdm845
Tested-by: Sam Edwards <CFSworks@gmail.com> # Binary output identical
Signed-off-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Current Cache Size ID Register (ccsidr_el1) has two "flavors"
depending on whether FEAT_CCIDX is implemented or not.
When FEAT_CCIDX is implemented Associativity parameter
is coded on bits [23:3] and NumSets parameter on bits [55:32].
When FEAT_CCIDX is not implemented then Associativity parameter
is coded on bits [12:3] and NumSets parameter on bits [27:13].
Current U-Boot code does not check whether FEAT_CCIDX is implemented
and always parses ccsidr_el1 as if FEAT_CCIDX was not implemented.
This is of course wrong on systems where FEAT_CCIDX is implemented.
This patch fixes that problems and tests whether FEAT_CCIDX
is implemented or not and accordingly parses the ccsidr_el1 register.
Signed-off-by: Lukasz Wiecaszek <lukasz.wiecaszek@gmail.com>
Raymond Mao <raymond.mao@linaro.org> says:
This patch set adds/adapts a few bloblist APIs and implements Arm arch
custom function to retrieve the bloblist (aka. Transfer List) from
previous loader via boot arguments when BLOBLIST option is enabled and
all boot arguments are compliant to the register conventions defined
in the Firmware Handoff spec v0.9.
If an arch wishes to have different behaviors for loading bloblist
from the previous boot stage, it is required to implement the custom
function xferlist_from_boot_arg().
