ARM64 supports more memory mapping types for device memory (nGnRnE,
nGnRE, GRE), add these mapping support for os common mapping API
function z_phys_map().
Signed-off-by: Jiafei Pan <Jiafei.Pan@nxp.com>
This is a strange one: The printing code pushes a floating point
register, and is called during the mpu falt. If the floating point
registers are lazily stacked, this fp push can cause another mpu
fault to be pending during the current mpu fault, and tail chained
without returning to PendSV. Since we're already cleaning up the
fp execption reason, we might as well also clean up thisp pending,
spurious mpu exception.
Signed-off-by: Jimmy Brisson <jimmy.brisson@linaro.org>
If an SVC was pending during the stack overflow, it will run
after the return of the memory manage fault. To the SVC's misfortune of
the SVC handler, the it's invariant, that PSP point to the
hardware-stacked context is no longer valid. When the user has a
k_sys_fatal_error_handler that tries to kill the thread that caused a
stack overflow, this manifests as the svc reading the memory of whatever
is on the stack after being adjusted by the mem manage fault handler, and
that leads to unending, spurious hard faults, locking up the system.
This patch prevents that.
Signed-off-by: Jimmy Brisson <jimmy.brisson@linaro.org>
New KConfig options for 'A' and 'M' RISC-V extensions have been
added. These are used to configure the '-march' string used by GCC
to produce a compatible binary for the requested RISC-V variant.
In order to maintain compatibility with all currently defined SoC,
default the options for HW mul / Atomics support to 'y', but allow
them to be overridden for any SoC which does not support these.
I tested this change locally via twister agaisnt a few RISC-V platforms
including some 32bit and 64bit. To verify the 4 possibilities of Atomics
& HW Mul: (No, No), (No, Yes), (Yes, No), (Yes, Yes -- current behavior),
I used an out-of-tree GCC (xPack RISC-V GCC) which has multilib support
for rv32i, rv32ia, rv32ima to test against our out-of-tree Intel Nios V/m
processor in HW. The Zephyr SDK RISCV GCC currently does not contain
multilib support for all variants exposed by these new KConfig options.
Signed-off-by: Nathan Krueger <nathan.krueger@intel.com>
In case of EFI, efi_init must be called before initializing early
serial: if that one as X86_SOC_EARLY_SERIAL_PCIDEV defined, its pcie
access will try to initialise pcie mmio access which one will try to
find an ACPI table. At this point, calling ACPI API prior to initialize
EFI will make RSDP looked up already... and since it cannot find it
without EFI being initialized first, ACPI is then broken.
Just moving early serial to initialize after multiboot/efi being setup.
Signed-off-by: Tomasz Bursztyka <tomasz.bursztyka@linux.intel.com>
If such table pointer is present with EFI system table, this will speed
up ACPI initialization later on.
Signed-off-by: Tomasz Bursztyka <tomasz.bursztyka@linux.intel.com>
As for Multiboot, let prep_c be aware of EFI boot.
In the futur, EFI will pass an argument to it.
Signed-off-by: Tomasz Bursztyka <tomasz.bursztyka@linux.intel.com>
In order to mitigate at runtime whether it booted on multiboot or EFI,
let's introduce a dedicated x86 cpu argument structure which holds the
type and the actual pointer delivered by the method (multiboot_info, or
efi_system_table)
Signed-off-by: Tomasz Bursztyka <tomasz.bursztyka@linux.intel.com>
Just a dummy function will do.
When enabled, the code does not need the #ifdef as cmake is handling
this properly already. This was also the wrong CONFIG_ used there
anyway.
Signed-off-by: Tomasz Bursztyka <tomasz.bursztyka@linux.intel.com>
The incorrect sequence will cause the thread cannot be aborted in the
ISR context. The following test case failed:
tests/kernel/fatal/exception/kernel.common.stack_sentinel.
The stack sentinel detects the stack overflow as normal during a timer
ISR exit. Note that, currently, the stack overflow detection is behind
the context switch checking, and then the detection will call svc to
raise a fatal error resulting in increasing the nested counter(+1). At
this point, it needs a context switch to finally abort the thread.
However, after the fatal error handling, the program cannot do a context
switch either during the svc exit[1], or during the timer ISR exit[2].
[1] is because the svc context is in an interrupt nested state (the
nested counter is 2).
[2] is because the current point (after svc context pop out) is right
behind the switch checking.
Signed-off-by: Jaxson Han <jaxson.han@arm.com>
This is painful. There is no way for u-mode code to know if we're
currently executing in u-mode without generating a fault, besides
stealing a general purpose register away from the standard ABI
that is. And a global variable doesn't work on SMP as this must be
per-CPU and we could be migrated to another CPU just at the right
moment to peek at the wrong CPU variable (and u-mode can't disable
preemption either).
So, given that we'll have to pay the price of an exception entry
anyway, let's at least make it free to privileged threads by using
the mscratch register as the non-user context indicator (it must
be zero in m-mode for exception entry to work properly). In the
case of u-mode we'll simulate a proper return value in the
exception trap code. Let's settle on the return value in t0
and omit the volatile to give the compiler a chance to cache
the result.
Signed-off-by: Nicolas Pitre <npitre@baylibre.com>
To do so efficiently on systems without the mul instruction, we use
shifts and adds which is faster and sometimes smaller than a plain loop.
Signed-off-by: Nicolas Pitre <npitre@baylibre.com>
Stop using &_kernel as this is not SMP friendly. Let's use s0 (after
preserving its content) to hold ¤t_cpu instead so it won't have
to be reloaded each time it is needed. This will be even more relevant
when SMP support is added.
Signed-off-by: Nicolas Pitre <npitre@baylibre.com>
Rely on mstatus rather than thread->base.user_options since it is always
up to date (updated by z_riscv_switch) to simplify the code and be SMP
proof. Also carry over SF_INIT to the mstatus being restored in case
it was changed in the mean time.
Signed-off-by: Nicolas Pitre <npitre@baylibre.com>
The move to arch_switch() is a prerequisite for SMP support.
Make it optimal without the need for an ECALL roundtrip on every
context switch. Performance numbers from tests/benchmarks/sched:
Before:
unpend 107 ready 102 switch 188 pend 218 tot 615 (avg 615)
After:
unpend 107 ready 102 switch 170 pend 217 tot 596 (avg 595)
Signed-off-by: Nicolas Pitre <npitre@baylibre.com>
This is a per-thread register that gets updated only when context
switching. No need to load and save it on every exception entry.
Signed-off-by: Nicolas Pitre <npitre@baylibre.com>
The minimum stack alignment is 16. Therefore, the stack space to store
a struct __esf object must be rounded up to the next 16-byte boundary.
It is not sufficient to do the rounding on the __z_arch_esf_t_SIZEOF
definition. When the stack is constructed in arch_new_thread() it is
also necessary to do the rounding there too.
Let's make the structure itself carry the alignment attribute instead to
make it work in all cases.
While at it, remove the unused _K_THREAD_NO_FLOAT_SIZEOF definition.
Signed-off-by: Nicolas Pitre <npitre@baylibre.com>
Get rid of all those global variables and IRQ locking.
Use the regular IRQ exit path to let tests validate preemption properly.
Signed-off-by: Nicolas Pitre <npitre@baylibre.com>
Complete revamp of the exception entry code, including syscall handling.
Proper syscall frame exception trigger. Many correctness fixes, hacks
removal, etc. etc.
I tried to make this into several commits, but this stuff is all
inter-related and a pain to split.
The diffstat summary:
14 files changed, 250 insertions(+), 802 deletions(-)
Binary size (before):
text data bss dec hex filename
1104 0 0 1104 450 isr.S.obj
64 0 0 64 40 userspace.S.obj
Binary size (after):
text data bss dec hex filename
600 0 0 600 258 isr.S.obj
36 0 0 36 24 userspace.S.obj
Run of samples/userspace/syscall_perf (before):
*** Booting Zephyr OS build zephyr-v3.0.0-325-g3748accae018 ***
Main Thread started; qemu_riscv32
Supervisor thread started
User thread started
Supervisor thread(0x80010048): 384 cycles 509 instructions
User thread(0x80010140): 77312 cycles 77437 instructions
Run of samples/userspace/syscall_perf (after):
*** Booting Zephyr OS build zephyr-v3.0.0-326-g4c877a2753b3 ***
Main Thread started; qemu_riscv32
Supervisor thread started
User thread started
Supervisor thread(0x80010048): 384 cycles 509 instructions
User thread(0x80010138): 7040 cycles 7165 instructions
Yes, that's more than a 10x speed-up!
Signed-off-by: Nicolas Pitre <npitre@baylibre.com>
Same rationale as preceding commit. Let's create pseudo-instructions in
assembly scope to make the code more uniform and readable.
Furthermore the definition of COPY_ESF_FP() was wrong as the width of
floating point registers vary not according to CONFIG_64BIT but
CONFIG_CPU_HAS_FPU_DOUBLE_PRECISION. It is therefore wrong to use
lr/sr (previously RV_OP_LOADREG/RV_OP_STOREREG) and a regular temporary
register to transfer such content.
Note: There are far more efficient ways to copy FP context around but
such optimisations will come separately.
Signed-off-by: Nicolas Pitre <npitre@baylibre.com>
Those are prominent enough that having RV_OP_LOADREG and RV_OP_STOREREG
shouting at you all over the place is rather unpleasant and bad taste.
Let's create pseudo-instructions of our own with assembler macros
rather than preprocessor defines and only in assembly scope.
This makes the asm code way more uniform and readable.
Signed-off-by: Nicolas Pitre <npitre@baylibre.com>
The thread->base.user_options field is an uint8_t. Access it using lb.
A "copy" of it is made into __esf.fp_state. Make that field an uint8_t
too and access it with lb/sb.
_callee_saved.fcsr is an uint32_t. Access it with lw/sw.
Ditto for is_user_mode.
Signed-off-by: Nicolas Pitre <npitre@baylibre.com>
This reverts commit 8686ab5472.
The purpose of this commit will be reintroduced later on top of
a cleaner codebase.
Signed-off-by: Nicolas Pitre <npitre@baylibre.com>
This reverts commit be28de692c.
The purpose of this commit will be reintroduced later on top of
a cleaner codebase.
Signed-off-by: Nicolas Pitre <npitre@baylibre.com>
This reverts commit b0458201cc.
The purpose of this commit will be reintroduced later on top of
a cleaner codebase.
Signed-off-by: Nicolas Pitre <npitre@baylibre.com>
ARMv8-R allows to set the vector table address using VBAR
register, so there is no need to relocate it.
Move away vector_table setting from reset.S and move it to
relocate vector table function as it's done for Cortex-M
CPU.
Signed-off-by: Julien Massot <julien.massot@iot.bzh>
It is not necessary to go through the full exception exit code.
This is simpler, smaller and faster.
Signed-off-by: Nicolas Pitre <npitre@baylibre.com>
Make it optimal without the need for an SVC/exception roundtrip on
every context switch. Performance numbers from tests/benchmarks/sched:
Before:
unpend 85 ready 58 switch 258 pend 231 tot 632 (avg 699)
After:
unpend 85 ready 59 switch 115 pend 138 tot 397 (avg 478)
Signed-off-by: Nicolas Pitre <npitre@baylibre.com>
Get rid of all those global variables and scheduler locking.
Use the reguler IRQ exit path to let tests properly validate preemption.
Signed-off-by: Nicolas Pitre <npitre@baylibre.com>
Add CONFIG_SMP to fvp_baser_aemv8r_smp board.
Fix compile warnings by adding missing header file in arm_mpu.c.
Signed-off-by: Jaxson Han <jaxson.han@arm.com>
This commit mainly fixes the broadcast_ipi issue when one core broadcast
ipi to other cores using gic_raise_sgi. The issue doesn't affect the
functionality of Zephyr SMP but will happen when Zephyr runs on Xen.
Suppose Xen provides 4 CPUs to the Zephyr guest, for example, when cpu0
broadcasts ipi to the rest of the cores, the mask should be 0xE(0b1110),
but for now, Zephyr will send 0xFFFE. So for Xen, it will receive a
target list containing many invalid CPUs which don't exist. My solution
is: to generate the target list according to the online CPUs.
Signed-off-by: Jaxson Han <jaxson.han@arm.com>
The ARMv8-R processors always boot into Hyp mode (EL2)
To enter EL1:
Program the HACTLR register because it defaults
to only allowing EL2 accesses. HACTLR controls
whether EL1 can access memory region registers and CPUACTLR.
Program the SPSR before entering EL1.
Other registers default to allowing accesses at EL1 from reset.
Set VBAR to the correct location for the vector table.
Set ELR to point to the entry point of the EL1 code and call ERET.
Signed-off-by: Julien Massot <julien.massot@iot.bzh>
Improve code by using DEVICE_DT_GET_ONE instead of device_get_binding,
since the intel_vt_d device instance can be obtained at compile time.
Signed-off-by: Gerard Marull-Paretas <gerard.marull@nordicsemi.no>
According to Kconfig guidelines, boolean prompts must not start with
"Enable...". The following command has been used to automate the changes
in this patch:
sed -i "s/bool \"[Ee]nables\? \(\w\)/bool \"\U\1/g" **/Kconfig*
Signed-off-by: Gerard Marull-Paretas <gerard.marull@nordicsemi.no>
Jiafei Pan