泰晓科技 -- 聚焦 Linux - 追本溯源,见微知著!
网站地址:https://tinylab.org

泰晓RISC-V实验箱,转战RISC-V,开箱即用
请稍侯

RISC-V 异常处理在 KVM 中的实现

panxiakai 创作于 2023/06/13

Corrector: TinyCorrect v0.1 - [tounix spaces toc comments tables images urls epw] Author: XiakaiPan 13212017962@163.com Date: 2022/10/21 Revisor: walimis, Falcon Project: RISC-V Linux 内核剖析 Proposal: RISC-V 虚拟化技术调研与分析 Sponsor: PLCT Lab, ISCAS

前言

Trap 处理是 RISC-V 虚拟化实现中的重要部分,包括异常和中断两个部分。当前 KVM 是 RISC-V 虚拟化扩展在软件层面较为可靠的实现,本文将结合 RISC-V 特权指令集手册的规定,分析 KVM 中有关异常处理的实现,中断部分由于涉及较多驱动层面的内容,故将在之后的文章中结合 MMIO,timer 等做具体探讨。

软件版本

Softwarecommit ID or version No.Link
Linux Kernelv6.0https://www.kernel.org/
kvmtool6a1f699108e5c2a280d7cd1f1ae4816b8250a29fhttps://github.com/kvmtool/kvmtool

KVM 异常处理

异常处理入口

在 KVM 对 RISC-V H 扩展的实现中,与异常处理相关的函数调用关系如下图所示。目前的实现中,KVM 能够处理三类异常。即虚拟机内的 page fault、虚拟指令异常和系统调用,三种不同的异常处理分别对应了不同的实现。

flowchart LR subgraph arch/riscv/kvm/vcpu.c run end subgraph arch/riscv/kvm/vcpu_exit.c exit gpf end subgraph arch/riscv/kvm/vcpu_insn.c virt_insn end subgraph arch/riscv/kvm/vcpu_sbi.c ecall end run[kvm_arch_vcpu_ioctl_run]-->exit[kvm_riscv_vcpu_exit] exit-->virt_insn[kvm_riscv_vcpu_virtual_insn] exit-->gpf[gstage_page_fault] exit-->ecall[kvm_riscv_vcpu_sbi_ecall]

下载由 Mermaid 生成的 PNG 图片

异常分类及其定义

kvm_arch_vcpu_ioctl_run 函数用于实现 vCPU 的运行。其调用 kvm_riscv_vcpu_enter_exit 函数进入 vCPU 的运行,此时 Guest 进入运行状态,CPU 处于 VS 或者 VU 模式。当 Guest 发生无法处理的异常时,Guest 退出,CPU 进入 HS 模式,随后 KVM 调用 kvm_riscv_vcpu_exit 来实现对异常的处理。

kvm_riscv_vcpu_exit 函数内部包含三个部分,分别对应三种异常的处理,代码如下:

// arch/riscv/kvm/vcpu_exit.c: line 188
/*
 * Return > 0 to return to guest, < 0 on error, 0 (and set exit_reason) on
 * proper exit to userspace.
 */
int kvm_riscv_vcpu_exit(struct kvm_vcpu *vcpu, struct kvm_run *run,
			struct kvm_cpu_trap *trap)
{
	int ret;

	/* 仅处理 guest 内部的 trap */
	/* If we got host interrupt then do nothing */
	if (trap->scause & CAUSE_IRQ_FLAG)
		return 1;

	/* 处理 guest 中断:KVM 使得 Linux 内核成为了 hypervisor,故 KVM 内部实现了对来自 guest/VM 和 hypervisor 的 trap 处理,此处仅处理 Guest 内部的 trap */
	/* Handle guest traps */
	ret = -EFAULT;
	run->exit_reason = KVM_EXIT_UNKNOWN;
	switch (trap->scause) {
	/* 虚拟指令异常 */
	case EXC_VIRTUAL_INST_FAULT:    // 22, Virtual instruction
		if (vcpu->arch.guest_context.hstatus & HSTATUS_SPV)
			ret = kvm_riscv_vcpu_virtual_insn(vcpu, run, trap);
		break;
	/* 虚拟机内页缺陷异常 */
	case EXC_INST_GUEST_PAGE_FAULT: // 20, Instruction guest-page fault
	case EXC_LOAD_GUEST_PAGE_FAULT: // 21, Load guest-page fault
	case EXC_STORE_GUEST_PAGE_FAULT:// 23, Store/AMO guest-page fault
		if (vcpu->arch.guest_context.hstatus & HSTATUS_SPV)
			ret = gstage_page_fault(vcpu, run, trap);
		break;
	/* 虚拟机内系统调用 */
	case EXC_SUPERVISOR_SYSCALL:    // 10, Environment call from VS-mode
		if (vcpu->arch.guest_context.hstatus & HSTATUS_SPV)
			ret = kvm_riscv_vcpu_sbi_ecall(vcpu, run);
		break;
	default:
		break;
	}

	/* 若异常未能被顺利处理(ret > 0),则输出当前状态(sepc, sstatus, hstatus)和对应的异常信息(scause, stval, htval, htinst)*/
	/* Print details in-case of error */
	if (ret < 0) {
		kvm_err("VCPU exit error %d\n", ret);
		kvm_err("SEPC=0x%lx SSTATUS=0x%lx HSTATUS=0x%lx\n",
			vcpu->arch.guest_context.sepc,
			vcpu->arch.guest_context.sstatus,
			vcpu->arch.guest_context.hstatus);
		kvm_err("SCAUSE=0x%lx STVAL=0x%lx HTVAL=0x%lx HTINST=0x%lx\n",
			trap->scause, trap->stval, trap->htval, trap->htinst);
	}

	return ret;
}

如上所示,KVM 的实现中包含了三类异常:

  • 虚拟指令异常;
  • Guest page fault;
  • SBI 系统调用。

特权指令集手册 中规定了每种异常对应的编码(即 scause 的可能的值),在进行异常处理时,可依据据 scause 的具体值确定其处理方式,如下表所示。

cause code

在 KVM 中,其对应宏的定义如下:

// arch/riscv/include/asm/csr.h: line 66
/* Exception causes */
#define EXC_INST_MISALIGNED	0               // Instruction address misaligned
#define EXC_INST_ACCESS		1               // Instruction access fault
#define EXC_INST_ILLEGAL	2               // Illegal instruction
#define EXC_BREAKPOINT		3               // Breakpoint
#define EXC_LOAD_ACCESS		5               // Load access fault
#define EXC_STORE_ACCESS	7               // Store/AMO access fault
#define EXC_SYSCALL		8		// Environment call from U-mode or VU-mode
#define EXC_HYPERVISOR_SYSCALL	9	        // Environment call from HS-mode
#define EXC_SUPERVISOR_SYSCALL	10		// Environment call from VS-mode
#define EXC_INST_PAGE_FAULT	12		// Instruction page fault
#define EXC_LOAD_PAGE_FAULT	13		// Load page fault
#define EXC_STORE_PAGE_FAULT	15		// Store/AMO page fault
#define EXC_INST_GUEST_PAGE_FAULT	20      // Instruction guest-page fault
#define EXC_LOAD_GUEST_PAGE_FAULT	21      // Load guest-page fault
#define EXC_VIRTUAL_INST_FAULT		22      // Virtual instruction
#define EXC_STORE_GUEST_PAGE_FAULT	23      // Store/AMO guest-page fault

虚拟指令异常

其中,EXC_VIRTUAL_INST_FAULT 即 virtual instruction exception 对应如下情况:

  • 在 VS-Mode 或 VU-Mode 下访问特定 CSR 的特定位;
  • 在 VS-Mode 或 VU-Mode 下执行无权限的指令如 HFENCE, HLV, HSV 等。

KVM 中 virtual instruction 异常的处理如下:

// arch/riscv/kvm/vcpu_insn.c: line 397
/**
 * kvm_riscv_vcpu_virtual_insn -- Handle virtual instruction trap
 *
 * @vcpu: The VCPU pointer
 * @run:  The VCPU run struct containing the mmio data
 * @trap: Trap details
 *
 * Returns > 0 to continue run-loop
 * Returns   0 to exit run-loop and handle in user-space.
 * Returns < 0 to report failure and exit run-loop
 */
int kvm_riscv_vcpu_virtual_insn(struct kvm_vcpu *vcpu, struct kvm_run *run,
				struct kvm_cpu_trap *trap)
{
	unsigned long insn = trap->stval;   // 获取导致 trap 的指令
	struct kvm_cpu_trap utrap = { 0 };
	struct kvm_cpu_context *ct;

	/* 判断是否为 16-bit 的压缩指令([1:0]=2,非压缩指令 [1:0]=3),如果是压缩指令,则作如下处理 */
	if (unlikely(INSN_IS_16BIT(insn))) {
		if (insn == 0) {    // Illegal instruction 非法指令(参见特权指令集手册的表 16.5)
			ct = &vcpu->arch.guest_context;
			insn = kvm_riscv_vcpu_unpriv_read(vcpu, true,
							  ct->sepc,
							  &utrap);  // 从 Guest 内存中读取指定地址的内存
			if (utrap.scause) {
				utrap.sepc = ct->sepc;
				kvm_riscv_vcpu_trap_redirect(vcpu, &utrap); // 重定向 trap 到 Guest 中
				return 1;
			}
		}
		if (INSN_IS_16BIT(insn))
			return truly_illegal_insn(vcpu, run, insn);     // 将当前指令直接重定向到 Guest 中
	}

	/* 对于非压缩指令,根据当前指令的类型([6:2] opcode)进行处理:*/
	switch ((insn & INSN_OPCODE_MASK) >> INSN_OPCODE_SHIFT) {
	case INSN_OPCODE_SYSTEM:    // SYSTEM 类型的指令(ecall, ebreak, CSR 读写指令)
		return system_opcode_insn(vcpu, run, insn);
	default:    // 正常长度且非 SYSTEM 类型的指令,将当前指令直接重定向到 Guest 中进行处理
		return truly_illegal_insn(vcpu, run, insn);
	}
}

// arch/riscv/kvm/vcpu_insn.c: line 70-72
#define INSN_16BIT_MASK		0x3
#define INSN_IS_16BIT(insn)	(((insn) & INSN_16BIT_MASK) != INSN_16BIT_MASK)

其中用于处理具体指令的函数其原型或定义如下:

处理非法压缩指令时,用于从 Guest 获取合法指令的 kvm_riscv_vcpu_unpriv_read 函数:

// arch/riscv/kvm/vcpu_exit.c: line 50
/**
 * kvm_riscv_vcpu_unpriv_read -- Read machine word from Guest memory
 *
 * @vcpu: The VCPU pointer
 * @read_insn: Flag representing whether we are reading instruction
 * @guest_addr: Guest address to read
 * @trap: Output pointer to trap details
 */
unsigned long kvm_riscv_vcpu_unpriv_read(struct kvm_vcpu *vcpu,
					 bool read_insn,
					 unsigned long guest_addr,
					 struct kvm_cpu_trap *trap);

从 Hypervisor 重定向到 Guest 的函数 kvm_riscv_vcpu_trap_redirect

// arch/riscv/kvm/vcpu_exit.c: line 152
/**
 * kvm_riscv_vcpu_trap_redirect -- Redirect trap to Guest
 *
 * @vcpu: The VCPU pointer
 * @trap: Trap details
 */
void kvm_riscv_vcpu_trap_redirect(struct kvm_vcpu *vcpu,
				  struct kvm_cpu_trap *trap);

对合法的压缩指令以及非 SYSTEM 类型的非压缩指令,不进行额外处理,直接调用 truly_illegal_insn 函数处理,保存当前 trap 的具体信息,将 Guest PC 设置为 Guest 中对应的异常向量, 然后返回到到 Guest 中对异常进行处理:

// arch/riscv/kvm/vcpu_insn.c: line 151
static int truly_illegal_insn(struct kvm_vcpu *vcpu, struct kvm_run *run,
			      ulong insn)
{
	struct kvm_cpu_trap utrap = { 0 };

	/* Redirect trap to Guest VCPU */
	utrap.sepc = vcpu->arch.guest_context.sepc;
	utrap.scause = EXC_INST_ILLEGAL;
	utrap.stval = insn;
	utrap.htval = 0;
	utrap.htinst = 0;
	kvm_riscv_vcpu_trap_redirect(vcpu, &utrap);

	return 1;
}

对 SYSTEM 类型进行特别操作的函数 system_opcode_insn

// arch/riscv/kvm/vcpu_insn.c: line 368
static int system_opcode_insn(struct kvm_vcpu *vcpu, struct kvm_run *run,
			      ulong insn)
{
	int i, rc = KVM_INSN_ILLEGAL_TRAP;
	const struct insn_func *ifn;

	for (i = 0; i < ARRAY_SIZE(system_opcode_funcs); i++) {
		ifn = &system_opcode_funcs[i];
		if ((insn & ifn->mask) == ifn->match) {
			rc = ifn->func(vcpu, run, insn);
			break;
		}
	}

	switch (rc) {
	case KVM_INSN_ILLEGAL_TRAP:         // 非法指令异常,设置 scause 为对应编码后重定向到 Guest

		return truly_illegal_insn(vcpu, run, insn);
	case KVM_INSN_VIRTUAL_TRAP:         // 虚拟指令异常,设置 scause 为对应编码后重定向到 Guest
		return truly_virtual_insn(vcpu, run, insn);
	case KVM_INSN_CONTINUE_NEXT_SEPC:   // 执行下一条指令
		vcpu->arch.guest_context.sepc += INSN_LEN(insn);
		break;
	default:
		break;
	}

	return (rc <= 0) ? rc : 1;
}

虚拟指令的处理函数 truly_virtual_insn

// arch/riscv/kvm/vcpu_insn.c: line 167
static int truly_virtual_insn(struct kvm_vcpu *vcpu, struct kvm_run *run,
			      ulong insn)
{
	struct kvm_cpu_trap utrap = { 0 };

	/* Redirect trap to Guest VCPU */
	utrap.sepc = vcpu->arch.guest_context.sepc;
	utrap.scause = EXC_VIRTUAL_INST_FAULT;
	utrap.stval = insn;
	utrap.htval = 0;
	utrap.htinst = 0;
	kvm_riscv_vcpu_trap_redirect(vcpu, &utrap);

	return 1;
}

其调用关系如下图所示:

flowchart LR subgraph arch/riscv/kvm/vcpu_insn.c vvi[kvm_riscv_vcpu_virtual_insn] tii[truly_illegal_insn] tvi[truly_virtual_insn] soi[system_opcode_insn] end subgraph arch/riscv/kvm/vcpu_exit.c rd[kvm_riscv_vcpu_unpriv_read] rdrct[kvm_riscv_vcpu_trap_redirect] end vvi--Illegal Compressed-->rd vvi--Illegal Compressed-->rdrct vvi--Legal Compressed-->tii vvi--SYSTEM-->soi vvi-->tii tii-->rdrct soi-->tii soi-->tvi

下载由 Mermaid 生成的 PNG 图片

Guest page fault

参见 上篇 关于 gstage page fault 处理函数的分析,调用关系如下图所示:

flowchart LR subgraph arch/riscv/kvm/vcpu.c run[kvm_arch_vcpu_ioctl_run] end subgraph arch/riscv/kvm/vcpu_exit.c gpgft[gstage_page_fault] exit[kvm_riscv_vcpu_exit] end subgraph arch/riscv/kvm/vcpu_insn.c ld[kvm_riscv_vcpu_mmio_load] st[kvm_riscv_vcpu_mmio_store] end subgraph arch/riscv/kvm/mmu.c mp[kvm_riscv_gstage_map] end run-->exit-->gpgft gpgft--error hva or store fault-->ld gpgft--error hva or store fault-->st gpgft-->mp

下载由 Mermaid 生成的 PNG 图片

SBI 系统调用

系统调用的处理通过调用 kvm_riscv_vcpu_sbi_ecall 函数实现,如下方代码块所示:

SBI(Supervisor Binary Interface) 是直接运行在 Machine Mode 下的,为上层 OS 提供统一接口的程序,具有最高权限。而 Guest 访问 SBI 系统调用,是在 KVM 中模拟实现,不是实际访问 Machine Mode 中的 SBI firmware。KVM 通过直接访问和设置寄存器(cp->a7, cp->a0, cp->a0 等)的值来实现对 SBI 系统调用的处理。

int kvm_riscv_vcpu_sbi_ecall(struct kvm_vcpu *vcpu, struct kvm_run *run)
{
	int ret = 1;
	bool next_sepc = true;
	bool userspace_exit = false;
	struct kvm_cpu_context *cp = &vcpu->arch.guest_context;
	const struct kvm_vcpu_sbi_extension *sbi_ext;
	struct kvm_cpu_trap utrap = { 0 };
	unsigned long out_val = 0;
	bool ext_is_v01 = false;

	/* 根据当前参数(a7)确定对应 SBI 中被调用的扩展(ext)及其 handler,获得返回值 ret;否则设置 ret 的值为不支持 SBI */
	sbi_ext = kvm_vcpu_sbi_find_ext(cp->a7);
	if (sbi_ext && sbi_ext->handler) {
#ifdef CONFIG_RISCV_SBI_V01
		if (cp->a7 >= SBI_EXT_0_1_SET_TIMER &&
		    cp->a7 <= SBI_EXT_0_1_SHUTDOWN)
			ext_is_v01 = true;
#endif
		ret = sbi_ext->handler(vcpu, run, &out_val, &utrap, &userspace_exit);
	} else {
		/* Return error for unsupported SBI calls */
		cp->a0 = SBI_ERR_NOT_SUPPORTED;
		goto ecall_done;
	}

	/* 依据经由 SBI ext handler 处理之后返回的 utrap 判断是否为需要进一步处理的 trap 等 */
	/* Handle special error cases i.e trap, exit or userspace forward */
	if (utrap.scause) {
		/* No need to increment sepc or exit ioctl loop */
		ret = 1;
		utrap.sepc = cp->sepc;
		kvm_riscv_vcpu_trap_redirect(vcpu, &utrap);
		next_sepc = false;
		goto ecall_done;
	}

	/* 依据 SBI 返回结果判断是否需要停止运行,或直接向 Guest 返回特定错误代码 */
	/* Exit ioctl loop or Propagate the error code the guest */
	if (userspace_exit) {
		next_sepc = false;
		ret = 0;
	} else {
		/**
		 * SBI extension handler always returns an Linux error code. Convert
		 * it to the SBI specific error code that can be propagated the SBI
		 * caller.
		 */
		ret = kvm_linux_err_map_sbi(ret);
		cp->a0 = ret;
		ret = 1;
	}
/* 设置全局 pc 以及返回值 */
ecall_done:
	if (next_sepc)
		cp->sepc += 4;
	if (!ext_is_v01)
		cp->a1 = out_val;

	return ret;
}

对应的函数调用关系如下图:

flowchart LR subgraph arch/riscv/kvm/vcpu.c run[kvm_arch_vcpu_ioctl_run] end subgraph arch/riscv/kvm/vcpu_sbi.c ecall[kvm_riscv_vcpu_sbi_ecall] fe[kvm_vcpu_sbi_find_ext] err[kvm_linux_err_map_sbi] end subgraph arch/riscv/kvm/vcpu_exit.c exit[kvm_riscv_vcpu_exit] rdrct[kvm_riscv_vcpu_trap_redirect] end run-->exit-->ecall-->fe ecall-->rdrct ecall-->err

下载由 Mermaid 生成的 PNG 图片

总结

本文结合 KVM 中有关异常处理的实现,讨论了在添加 H 扩展之后的虚拟指令异常、guest page fault 以及来自 guest 的系统调用的处理。

参考资料



Read Album:

Read Related:

Read Latest: