After changing content of page table(s), it is needed to notify
the other CPUs that the page table(s) have been changed so they
can do the necessary steps to use the updated version. Note that
the actual way to send IPI is SoC specific as Xtensa does not
have a common way to do this at the moment.
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
When kernel OOPS is raised, we need to actually go through
the process of terminating the offending thread, instead of
simply printing the stack and continue running. This change
employs similar mechanism to xtensa_arch_except() to use
illegal instruction to raise hardware exception, and going
through the fatal exception path.
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
Signed-off-by: Flavio Ceolin <flavio.ceolin@intel.com>
Trigger exception on Xtensa requires kernel privileges. Add
a new syscall that is used when ARCH_EXCEPT is invoked from userspace.
Signed-off-by: Flavio Ceolin <flavio.ceolin@intel.com>
This extracts the printing of fatal exception information into
its own function to declutter xtensa_excint1_c().
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
Signed-off-by: Flavio Ceolin <flavio.ceolin@intel.com>
There are known exceptions which are not fatal, and we need to
handle them properly by returning to the fixup addresses as
indicated. This adds the code necessary in the exception
handler for this situation.
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
Signed-off-by: Flavio Ceolin <flavio.ceolin@intel.com>
When MMU is enabled, we need some scratch registers to preload
page table entries. So update gen_zsr.py to that.
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
Signed-off-by: Flavio Ceolin <flavio.ceolin@intel.com>
This changes the TLB misses handling back to the assembly
in user exception, and any page faults during TLB misses to be
handled in double exception handler. This should speed up
simple TLB miss handling as we don't have to go all the way to
the C handler.
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
Signed-off-by: Flavio Ceolin <flavio.ceolin@intel.com>
Userspace support for Xtensa architecture using Xtensa MMU.
Some considerations:
- Syscalls are not inline functions like in other architectures because
some compiler issues when using multiple registers to pass parameters
to the syscall. So here we have a function call so we can use
registers as we need.
- TLS is not supported by xcc in xtensa and reading PS register is
a privileged instruction. So, we have to use threadptr to know if a
thread is an user mode thread.
Signed-off-by: Flavio Ceolin <flavio.ceolin@intel.com>
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
Simplify the logic around xtensa_mmu_init.
- Do not have a different path to init part of kernel
- Call xtensa_mmu_init from C
Signed-off-by: Flavio Ceolin <flavio.ceolin@intel.com>
Replace all autorefill helpers with only one that invalidates both,
DTLB and ITLB, since that is what is really needed.
Signed-off-by: Flavio Ceolin <flavio.ceolin@intel.com>
This register alias was originally introduced to allow A0 to be used
as a scratch register when handling exceptions from MOVSP
instructions. (It replaced some upstream code from Cadence that
hard-coded EXCSAVE1). Now the MMU code is now using too, and for
exactly the same purpose.
Calling it "ALLOCA" is only confusing. Rename it to make it clear
what it's doing.
Signed-off-by: Andy Ross <andyross@google.com>
TF-M only suports floating point in IPC model, not the SFN model.
Since floating point is a basic feature of the architecture and TF-M
has the limitation it makes more sense for the dependency to exist in
TF-M and and limit the TF-M model choice instead of limiting the
option to enable floating point.
Signed-off-by: Joakim Andersson <joakim.andersson@nordicsemi.no>
Some workarounds were introduced for intel cavs2.5 platform bring up.
It is not general so move them to platform code.
Signed-off-by: Rander Wang <rander.wang@intel.com>
Each arch platform may has a general arch_cpu_idle implementation but
each vendor may has a custom one, so this config will be used for vendor
to override it.
Some workarounds were introduced for intel cavs2.5 platform bring up.
It is not general so move them to platform code.
Signed-off-by: Rander Wang <rander.wang@intel.com>
When building without optimizations and with only one core the linker
does not throw away arch_start_cpu and we get an undefined reference to
x86_ap_start
Signed-off-by: Flavio Ceolin <flavio.ceolin@intel.com>
This moves the k_* memory management functions from sys/ into
kernel/ includes, as there are kernel public APIs. The z_*
functions are further separated into the kernel internal
header directory.
Also made a quick change to doxygen to group sys_mem_* into
the OS Memory Management group so they will appear in doc.
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
z_mp_entry has been removed from Xtensa architecture.
So there is no need for a function declaration. Remove it.
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
NXP SYSMPU is used in other SoCs besides the Kinetis series. For
devices like S32K1xx, its bus interface clock lacks of clock gating
and it's driven by the system clock. Hence, only enable the module
clock for the Kinetis series.
Signed-off-by: Manuel Argüelles <manuel.arguelles@nxp.com>
This patch changes the section of riscv_cpu_wake_flag variable to
noinit from bss to fix hangup of RISC-V multicore boot if hart0 is
not boot hart (CONFIG_RV_BOOT_HART != 0).
Current boot sequence initializes a riscv_cpu_wake_flag to -1 but
this variable is unintentionally changed to 0 by boot hart.
This is because the variable is placed in bss section so this patch
changes the section of the variable to noinit.
Signed-off-by: Katsuhiro Suzuki <katsuhiro@katsuster.net>
This patch fixes hangup of RISC-V multicore boot.
Currently boot sequence uses a riscv_cpu_wake_flag to notify wakeup
request for secondary core(s).
But initial value of riscv_cpu_wake_flag is undefined, so current
mechanism is going to hangup if riscv_cpu_wake_flag and mhartid of
secondary core have the same value.
This is an example situation of this problem:
- hart1: check riscv_cpu_wake_flag (value is 1) and end the loop
- hart1: set riscv_cpu_wake_flag to 0
- hart0: set riscv_cpu_wake_flag to 1
hart0 expects it will be changed to 0 by hart1 but it
has never happened
Note:
- hart0's mhartid is 0, hart1's mhartid is 1
- hart0 is main, hart1 is secondary in this example
Signed-off-by: Katsuhiro Suzuki <katsuhiro@katsuster.net>
The FMADD, FMSUB, FNMSUB and FNMADD instructions occupy major opcode
spaces of their own, separate from LOAD-FP/STORE-FP and OP-FP spaces.
Insert code to cover them.
Signed-off-by: Nicolas Pitre <npitre@baylibre.com>
elf_rela_t contains elf_rel_t exactly and contains an additional
field at the end. Therefore pointers of that type can be used for
both types, making the code generic.
Signed-off-by: Guennadi Liakhovetski <guennadi.liakhovetski@linux.intel.com>
Most of the public APIs in `riscv_plic.h`
(except `riscv_plic_get_irq` & `riscv_plic_get_dev`) expect the
`irq` argument to be in Zephyr-encoded format, instead of the
previously `irq_from_level_2`-stripped version. The first level
IRQ is needed by `intc_plic` to differentiate between the
parent interrupt controllers, so that correct ISR offset can be
obtained using the LUT in `sw_isr_common`.
Signed-off-by: Yong Cong Sin <ycsin@meta.com>
Instead of using a macro guard to prevent functions in
`sw_isr_common.c` from getting compiled when
`CONFIG_DYNAMIC_INTERRUPTS` isn't enabled, do that in
`CMakeLists.txt` instead.
Signed-off-by: Yong Cong Sin <ycsin@meta.com>
Refactor multi-level IRQ related code from `sw_isr_common.c` to
`multilevel_irq.c` to simplify `sw_isr_common` & macrologies.
Signed-off-by: Yong Cong Sin <ycsin@meta.com>
Relocate new and existing internal software-managed table
access functions from the public `sw_isr_table.h` into a
private header that should only be accessed internally.
Signed-off-by: Yong Cong Sin <ycsin@meta.com>
Change the internal function to `get_parent_entry`, which
returns the entire entry of table.
Store the parent interrupt controller device in the
`irq_parent_offset` table, and added 2 helper functions to:
1. determine the parent interrupt controller based on the IRQ
2. determine the IRQ of the parent interrupt controller
Declare the `struct _irq_parent_entry` in the header and added
`-` suffix to the struct so that it can be used to test the
functions in testsuites.
Signed-off-by: Yong Cong Sin <ycsin@meta.com>
This change adds support for the CONFIG_ARM_CUSTOM_INTERRUPT_CONTROLLER
option on Cortex-M platforms. While all Cortex-M platforms have a
NVIC controller some custom SoCs may have additional IRQ controllers
or custom handling. This change allows those SoCs to modify this
bahaviour without having to place platform specific logic inside
applications or drivers.
Signed-off-by: Corey Wharton <xodus7@cwharton.com>
The Cortex ARM documentation states that the DC IVAC instruction
requires write access permission to the virtual address (VA);
otherwise, it may generate a permission fault.
Therefore, it is needed to avoid invalidating read-only memory
after the memory map operation.
This issue has been produced by commit c9b534c.
This commit resolves the issue #64758.
Signed-off-by: Mykola Kvach <mykola_kvach@epam.com>
Some platforms already have .bss section zeroed-out externally before the
Zephyr initialization and there is no sence to zero it out the second time
from the SW.
Such boot-time optimization could be critical e.g. for RTL Simulation.
Signed-off-by: Alexander Razinkov <alexander.razinkov@syntacore.com>
This commit introduces SMP support into Cortex-A/R aarch32 architecture.
For now, this only supports multiple core start together and only allow
one CPU initialize system as primary core, others loop at the beginning
as the secondary cores and wait for wake up.
Signed-off-by: Huifeng Zhang <Huifeng.Zhang@arm.com>
This commit introduce 'USE_SWITCH' feature into cortex-A/R(aarch32)
architecture
For introducing USE_SWITCH, the exception entry and exit are unified via
`z_arm_cortex_ar_enter_exc` and `z_arm_cortex_ar_exit_exc`. All
exceptions including ISR are using this way to enter and exit exception
handler.
Differentiate exception depth and interrupt depth. Allow doing
context switch when exception depth greater than 1 but not allow doing
this when interrupt depth greater than 1.
Currently, USE_SWITCH doesn't support FPU_SHARING and USERSPACE.
Signed-off-by: Huifeng Zhang <Huifeng.Zhang@arm.com>
Store the current CPU's struct _cpu instance into TPIDRURO, so that the
CPU core can get its struct _cpu instance by reading TPIDRURO. This is
useful in the SMP system.
Signed-off-by: Huifeng Zhang <Huifeng.Zhang@arm.com>
Replace the TLS base address pointer from TPIDRURO to TPIDRURW.
The difference between them is that TPIDRURO is read-only in user mode
but TPIDRURW isn't. So TPIDRURO is much more suitable for store
the address of _kernel.CPU[n]. For this reason, this commit replaces
the base pointer of the TLS area.
Signed-off-by: Huifeng Zhang <Huifeng.Zhang@arm.com>
MMU or MPU unit need to be initialized by its own CPU.
- Primary core initialize MMU or MPU unit in z_arm_prep_c.
- Secondary core initialize MMU or MPU unit in z_arm_secondary_start.
Signed-off-by: Huifeng Zhang <Huifeng.Zhang@arm.com>
Use ACPI_MADT_LOCAL_APIC instead of struct acpi_madt_local_apic. In the
same go, switch to IF_ENABLED from ifdef - slightly more readable, and
this keeps some static analyzers happy (e.g. upstream Compliance check).
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
HCR_EL2 is configured to certain value by some
loaders such as Uboot on some arm64 boards(such as roc_rk3568_pc),
When HCR_EL2.TGE, HCR_EL2.AMO and HCR_EL2.IMO bits are
set to 1, some unpredictable behaviors may occur during
zephyr boot. So we clear these bits to avoid it.
Signed-off-by: Charlie Xiong <1981639884@qq.com>
Move the syscall_handler.h header, used internally only to a dedicated
internal folder that should not be used outside of Zephyr.
Signed-off-by: Anas Nashif <anas.nashif@intel.com>
Locate common mpu code together with other arm / nxp mpu code in the
arch folder where it logically belongs.
Signed-off-by: Torsten Rasmussen <Torsten.Rasmussen@nordicsemi.no>
Relocate multi-level interrupts APIs out of `irq.h` into
a new file named `irq_multilevel.h` to provide cleaner
separation between typical irq & multilevel ones.
Added preprocessor versions of `irq_to_level_x` as `IRQ_TO_Lx`.
Signed-off-by: Yong Cong Sin <ycsin@meta.com>
This commit re-implements the SPARC V8 ABI "Flush windows" software
trap. The trap is generated by C++ compilers for exceptions and also by
the C standard library function longjmp().
There were two issues with the previous implementation:
1. It did reads and writes via the stack pointer of the trap window,
which is not defined.
2. It executed with traps enabled but without the processor run-time
state set to safely handle traps. In particular there was no valid
stack for trap processing. Even though interrupt priority was set to
highest level, the behavior at other traps was not deterministic. For
example non-maskable interrupt (15) trap or bus error trap for
instruction fetch.
This new implementation does not store backup copies of CPU registers to
the stack, and it executes with traps disabled.
Fixes#63901
Signed-off-by: Martin Åberg <martin.aberg@gaisler.com>
Add pm cpu ops to call the platform specific implementations for
bringing up secondary cores.
Signed-off-by: Lingutla Chandrasekhar <quic_lingutla@quicinc.com>
HAS_DTS has become a redundant option. All Zephyr architectures now
select this option, meaning devicetree has become a de-facto
requirement. In fact, if any board does not provide a devicetree
source, the build system uses an empty stub, meaning the devicetree
machinery always runs.
Signed-off-by: Gerard Marull-Paretas <gerard@teslabs.com>
_k_neg_eagain is there for used in assembly where including
errno.h is not possible. However, the usage in ARM was simply
to assign value to swap_return_value in a C file, which is
no need to use _k_neg_eagain as errno.h can be included.
So change that to use -EAGAIN directly. Saves 4 bytes in
built binaries in rodata.
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
The exclusive load/store instructions don't work well when MMU and cache
are disabled on some cores e.g. Cortex-A72. Change it to voting lock[1]
to select the primary core when multi-cores boot simultaneously.
The voting lock has reasonable but minimal requirements on the memory
system.
[1] https://www.kernel.org/doc/html/next/arch/arm/vlocks.html
Signed-off-by: Jaxson Han <jaxson.han@arm.com>
This provides custom memory range check functions as
it gets a bit complicated with cached/uncached regions.
These functions are marked as __weak so SoC or board
can override these if needed.
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
