this patch moves function xx_offset and xx_index to common code,
Add arch interface arch_quirk/arch_pgtle_page_vaddr and
arch_pgtle_large.
Tracked-On: #8831
Signed-off-by: hangliu1 <hang1.liu@intel.com>
Reviewed-by: Liu, Yifan1 <yifan1.liu@intel.com>
Acked-by: Wang, Yu1 <yu1.wang@intel.com>
move _page_table_level to common, and rename functions and
variables to comform with pgtln style
when we refer to pgtl0e, it means the lowest translation
table entry, while the "pte" refers generic page table entry.
Tracked-On: #8831
Signed-off-by: hangliu1 <hang1.liu@intel.com>
Reviewed-by: Fei Li <fei1.li@intel.com>
Acked-by: Wang, Yu1 <yu1.wang@intel.com>
add x86 mm_common.h to map common macro name to x86 name
and chang them in common/mmu.c, replace XX_PFN_MASK with
PFN_MASK, since they are the same.
Tracked-On: #8831
Signed-off-by: hangliu1 <hang1.liu@intel.com>
Reviewed-by: Liu, Yifan1 <yifan1.liu@intel.com>
Acked-by: Wang, Yu1 <yu1.wang@intel.com>
The sanitized page is used to mitigate l1tf for x86, but sanitized_page_hpa
is the hpa used to set in the page table entry when delete or modify
page happens, it is by default is 0, other arch donot need to init it.
move this part to common for other arch to reuse it.
Tracked-On: #8831
Signed-off-by: hangliu1 <hang1.liu@intel.com>
Reviewed-by: Liu, Yifan1 <yifan1.liu@intel.com>
Acked-by: Wang, Yu1 <yu1.wang@intel.com>
move struct pgtable and page_pool to common code and
move alloc_page/free_page/init_page_pool to common
Tracked-On: #8831
Signed-off-by: hangliu1 <hang1.liu@intel.com>
Reviewed-by: Liu, Yifan1 <yifan1.liu@intel.com>
Acked-by: Wang, Yu1 <yu1.wang@intel.com>
move some funcitons like hpa2hva to common file.
change some files to include file from asm/pgtable.h to common/pgtable.h
Tracked-On: #8831
Signed-off-by: hangliu1 <hang1.liu@intel.com>
Reviewed-by: Fei Li <fei1.li@intel.com>
Acked-by: Wang, Yu1 <yu1.wang@intel.com>
For now, BOARD should use the qemu-riscv for risc-v.
For RISC-V:
make hypervisor BOARD=qemu-riscv SCENARIO=shared ARCH=riscv CROSS_COMPILE=riscv64-linux-gnu-
For x86:
make hypervisor BOARD=qemu SCENARIO=shared [ARCH=x86]
Tracked-On: #8805
Signed-off-by: Fei Li <fei1.li@intel.com>
Only x86 has local APIC pass-thru. For release mode, the console_vmexit_callback
is empty, complier should optimize this unuseful check.
Tracked-On: #8805
Signed-off-by: Fei Li <fei1.li@intel.com>
Now doesn't want introduce the PCI module since PCI CFG access should be arch
special.
Tracked-On: #8805
Signed-off-by: Fei Li <fei1.li@intel.com>
Acked-by: Wang, Yu1 <yu1.wang@intel.com>
get_vm_from_vmid,is_paused_vm and is_poweroff_vm should be common APIs.
But now doesn't implement them as common for not introduce more VM related
data structure and APIs.
Tracked-On: #8805
Signed-off-by: Fei Li <fei1.li@intel.com>
Acked-by: Wang, Yu1 <yu1.wang@intel.com>
The RISC-V calling convention defines a full descending stack, where 'sp'
points to the last used stack address. The current implementation of
cpu_ctx_save() subtracts only CPU_REGS_OFFSET_SCRATCH from 'sp', which
allocates 8 bytes too few and may overwrite the caller's stack contents.
Fix this by adjusting CPU_REGS_OFFSET_LAST to include the last slot,
ensuring the full context save/restore area is properly reserved.
Fixes: 6276763cd (hv: riscv: implement cpu_ctx_save and cpu_ctx_restore)
Tracked-On: #8827
Signed-off-by: Shiqing Gao <shiqing.gao@intel.com>
Acked-by: Wang, Yu1 <yu1.wang@intel.com>
key scenario configuration for qemu-riscv board:
SCHEDULER:SCHED_NOOP
RELOC_ENABLED:n
HV_RAM_START:0x81000000
ACRNTRACE_ENABLED:n
make changes for riscv qemu board file.
change board name to be qemu-riscv to align with folder name, add memory
range to pass xmlschema check.
Tracked-On: #8814
Reviewed-by: Fei Li <fei1.li@intel.com>
Signed-off-by: Xue Bosheng <bosheng.xue@intel.com>
Use tp register directly for arch_get_pcpu_id. This can eliminate the
unnecessary mv instruction.
Tracked-On: #8791
Signed-off-by: Jian Jun Chen <jian.jun.chen@intel.com>
Acked-by: Wang, Yu1 <yu1.wang@intel.com>
send_single_ipi/send_dest_ipi_mask are calling sbi_send_ipi. According
to SBI spec:
- uint64_t hart_mask is a scalar bit-vector containing hartids
- uint64_t hart_mask_base is the starting hartid from which the
bit-vector must be computed
Logical pCPU ID and pCPU ID mask need to be converted to physical hart
ID mask and physical hart ID base.
Tracked-On: #8811
Signed-off-by: Jian Jun Chen <jian.jun.chen@intel.com>
Acked-by: Wang, Yu1 <yu1.wang@intel.com>
MMIO read/write without memory order should be common ARCH, without
PIO support shouldn't use PIO APIs, so implement them as empty.
Tracked-On: #8807
Signed-off-by: Fei Li <fei1.li@intel.com>
Signed-off-by: Haoyu Tang <haoyu.tang@intel.com>
Reviewed-by: Yifan Liu <yifan1.liu@intel.com>
Clear AVAILABLE_IRQ_INFO content in qemu-riscv.xml as the x86 ISA IRQ
values (3, 4, 5, 6, 7, 8, 9, 11) are not applicable to RISC-V architecture.
Note: This leaves AVAILABLE_IRQ_INFO empty as a placeholder. A proper
RISC-V implementation should populate this field and fill the device
tree for VM.
Tracked-On: #8814
Reviewed-by: Fei Li <fei1.li@intel.com>
Signed-off-by: Wei6 Zhang <wei6.zhang@intel.com>
Add board XML configuration file for RISC-V QEMU virt machine.
Key features:
- Board and firmware information (U-Boot, QEMU virt platform)
- Memory layout from /proc/iomem
- PCI device configuration for virtio devices
- 4-core processor configuration
- Serial console (ttyS0) at 0x10000000
Added comprehensive FIXME documentation explaining:
1. Most information was manually collected rather than auto-extracted
2. Original board inspector relies on x86-specific mechanisms:
- DMI/SMBIOS (dmidecode) vs Device Tree
- ACPI vs Device Tree
- APIC/IOAPIC vs PLIC/CLINT
- BIOS vs OpenSBI/U-Boot
3. Fields needing RISC-V adaptation:
- AVAILABLE_IRQ_INFO: x86 ISA IRQs (0-15) → PLIC range (1-95+)
- apic_id → hart_id
- IOAPIC, DRHD (not applicable)
The VM is started with:
qemu-system-riscv64 \
-machine virt -nographic -m 2048 -smp 4 \
-kernel /usr/lib/u-boot/qemu-riscv64_smode/uboot.elf \
-device virtio-net-device,netdev=eth0 -netdev user,id=eth0 \
-device virtio-rng-pci \
-drive file=<VM_image_file>,format=raw,if=virtio
Tracked-On: #8814
Reviewed-by: Victor Sun <victor.sun@intel.com>
Reviewed-by: Fei Li <fei1.li@intel.com>
Signed-off-by: Wei6 Zhang <wei6.zhang@intel.com>
init sched and call run_idle_thred in init_pcpu_comm_post to switch to idle
thread after AP initialization finish.
Tracked-On: #8812
Signed-off-by: Xue Bosheng <bosheng.xue@intel.com>
Acked-by: Wang, Yu1 <yu1.wang@intel.com>
schedule.c and sched_noop are common module instead of x86 specific, so
move them out of x86 macro, and need set SCHEDULER to be SCHED_NOOP
in scenario file for riscv build.
Tracked-On: #8812
Signed-off-by: Xue Bosheng <bosheng.xue@intel.com>
Acked-by: Wang, Yu1 <yu1.wang@intel.com>
some trace API is required by other module, so move trace.c out of x86,
it needs add ACRNTRACE_ENABLE = n in scenario file for riscv, could remove
ACRNTRACE_ENABLE = n when trace in riscv is fully supported.
Tracked-On: #8812
Signed-off-by: Xue Bosheng <bosheng.xue@intel.com>
Acked-by: Wang, Yu1 <yu1.wang@intel.com>
implement arch_cpu_dead and arch_cpu_do_idle for riscv, arch_cpu_dead and
arch_cpu_do_idle will be called by common module.
Tracked-On: #8812
Signed-off-by: Haicheng Li <haicheng.li@outlook.com>
Signed-off-by: Xue Bosheng <bosheng.xue@intel.com>
Acked-by: Wang, Yu1 <yu1.wang@intel.com>
cpu_dead calls arch_cpu_dead, cpu_do_idle calls arch_cpu_do_idle,
arch_cpu_dead and arch_cpu_do_idle are required to be implemented by
each architecture.
Tracked-On: #8812
Signed-off-by: Xue Bosheng <bosheng.xue@intel.com>
Reviewed-by: Yifan Liu <yifan1.liu@intel.com>
Acked-by: Wang, Yu1 <yu1.wang@intel.com>
move run_idle_thread and default_idle to schedule file as common
function.
Tracked-On: #8812
Signed-off-by: Xue Bosheng <bosheng.xue@intel.com>
Acked-by: Wang, Yu1 <yu1.wang@intel.com>
stack_frame is not only for vcpu thread, host thread needs
it, so move stack_frame out of vcpu file.
Tracked-On: #8812
Signed-off-by: Xue Bosheng <bosheng.xue@intel.com>
Reviewed-by: Yifan Liu <yifan1.liu@intel.com>
Acked-by: Wang, Yu1 <yu1.wang@intel.com>
delivery mode and idle mode are x86 specific percpu, so move it from common to
x86 arch, also change the name of mode_to_idle to be idle_mode, change the name
of mode_to_kick_pcpu to be kick_pcpu_mode.
Tracked-On: #8812
Signed-off-by: Xue Bosheng <bosheng.xue@intel.com>
Acked-by: Wang, Yu1 <yu1.wang@intel.com>
add arch_send_reschedule_request, which is called by make_reschedule_request, for X86
arch_send_reschedule_request will call kick_pcpu, arch_send_reschedule_request is arch
public API, which is needed to be implemented by each architecture.
Tracked-On: #8812
Signed-off-by: Xue Bosheng <bosheng.xue@intel.com>
Acked-by: Wang, Yu1 <yu1.wang@intel.com>
This patch moves timer.c and softirq.c out from x86 to
make them be built for risc-v too.
Tracked-On: #8816
Signed-off-by: Yi Y Sun <yi.y.sun@intel.com>
Acked-by: Wang, Yu1 <yu1.wang@intel.com>
This patch calls timer_init() to initialize timer for
pcpu on risc-v platform.
Tracked-On: #8816
Signed-off-by: Yi Y Sun <yi.y.sun@intel.com>
Acked-by: Wang, Yu1 <yu1.wang@intel.com>
Convert IRQ-related macros to static inline functions and introduce
wrappers for arch-specific implementations. This follows the style we
defined for multi-arch development.
This is a follow-up update for commit
a7239d126 ("[FIXME] hv: risc-v add denpended implementation in cpu.h").
CPU_IRQ_ENABLE_ON_CONFIG -> local_irq_enable
CPU_IRQ_DISABLE_ON_CONFIG -> local_irq_disable
CPU_INT_ALL_DISABLE -> local_irq_save
CPU_INT_ALL_RESTORE -> local_irq_restore
Tracked-On: #8813
Signed-off-by: Shiqing Gao <shiqing.gao@intel.com>
Reviewed-by: Yifan Liu <yifan1.liu@intel.com>
Fix a compilation error that may occur in some environments due to an
inconsistent type cast. The compiler reports a strict-aliasing violation:
acrn-hypervisor/hypervisor/include/arch/riscv/asm/cpu.h:73:18:
error: dereferencing type-punned pointer will break strict-aliasing rules
[-Werror=strict-aliasing]
73 | *(uint32_t *)(x) = val;
| ^~~~~
Tracked-On: #8813
Signed-off-by: Shiqing Gao <shiqing.gao@intel.com>
local_irq_save() is intended to read and clear the SIE bit in
sstatus (disabling interrupts). The current implementation mistakenly
clears all bits in sstatus.
This patch corrects the inline assembly so that only the SIE bit is affected,
ensuring proper interrupt save/disable semantics on RISC-V.
Fixes: a7239d126 ("[FIXME] hv: risc-v add denpended implementation in cpu.h")
Tracked-On: #8813
Signed-off-by: Shiqing Gao <shiqing.gao@intel.com>
Reviewed-by: Yifan Liu <yifan1.liu@intel.com>
Acked-by: Wang, Yu1 <yu1.wang@intel.com>
This patch implements interrupt initialization and the basic
exception/interrupt handling flow on RISC-V.
init_interrupt() needs to be invoked during CPU initialization to
set up the trap vector and enable the interrupt.
RISC-V exception and interrupt handling includes:
- Saving and restoring CPU registers around traps
- Implementing handlers for:
- Supervisor software interrupt
- Supervisor timer interrupt
- Halting the CPU for all other interrupts and exceptions
------
TODOs:
1. add support for registering interrupt handlers via request_irq() and
further adoption of the common IRQ framework.
2. add support for external interrupt.
Tracked-On: #8813
Signed-off-by: Haicheng Li <haicheng.li@intel.com>
Co-developed-by: Shiqing Gao <shiqing.gao@intel.com>
Signed-off-by: Shiqing Gao <shiqing.gao@intel.com>
Reviewed-by: Yifan Liu <yifan1.liu@intel.com>
Acked-by: Wang, Yu1 <yu1.wang@intel.com>
