From 5fdbf9765b7ba6a45100851154768de703d51e76 Mon Sep 17 00:00:00 2001 From: Marcelo Tosatti Date: Fri, 27 Jun 2008 14:58:02 -0300 Subject: KVM: x86: accessors for guest registers As suggested by Avi, introduce accessors to read/write guest registers. This simplifies the ->cache_regs/->decache_regs interface, and improves register caching which is important for VMX, where the cost of vmcs_read/vmcs_write is significant. [avi: fix warnings] Signed-off-by: Marcelo Tosatti Signed-off-by: Avi Kivity --- arch/x86/kvm/x86.c | 268 +++++++++++++++++++++++++++-------------------------- 1 file changed, 138 insertions(+), 130 deletions(-) (limited to 'arch/x86/kvm/x86.c') diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c index 0d682fc6aeb3..2f0696bc7d2f 100644 --- a/arch/x86/kvm/x86.c +++ b/arch/x86/kvm/x86.c @@ -19,6 +19,7 @@ #include "mmu.h" #include "i8254.h" #include "tss.h" +#include "kvm_cache_regs.h" #include #include @@ -61,6 +62,7 @@ static int kvm_dev_ioctl_get_supported_cpuid(struct kvm_cpuid2 *cpuid, struct kvm_cpuid_entry2 __user *entries); struct kvm_x86_ops *kvm_x86_ops; +EXPORT_SYMBOL_GPL(kvm_x86_ops); struct kvm_stats_debugfs_item debugfs_entries[] = { { "pf_fixed", VCPU_STAT(pf_fixed) }, @@ -2080,7 +2082,7 @@ int emulator_set_dr(struct x86_emulate_ctxt *ctxt, int dr, unsigned long value) void kvm_report_emulation_failure(struct kvm_vcpu *vcpu, const char *context) { u8 opcodes[4]; - unsigned long rip = vcpu->arch.rip; + unsigned long rip = kvm_rip_read(vcpu); unsigned long rip_linear; if (!printk_ratelimit()) @@ -2102,6 +2104,14 @@ static struct x86_emulate_ops emulate_ops = { .cmpxchg_emulated = emulator_cmpxchg_emulated, }; +static void cache_all_regs(struct kvm_vcpu *vcpu) +{ + kvm_register_read(vcpu, VCPU_REGS_RAX); + kvm_register_read(vcpu, VCPU_REGS_RSP); + kvm_register_read(vcpu, VCPU_REGS_RIP); + vcpu->arch.regs_dirty = ~0; +} + int emulate_instruction(struct kvm_vcpu *vcpu, struct kvm_run *run, unsigned long cr2, @@ -2112,7 +2122,13 @@ int emulate_instruction(struct kvm_vcpu *vcpu, struct decode_cache *c; vcpu->arch.mmio_fault_cr2 = cr2; - kvm_x86_ops->cache_regs(vcpu); + /* + * TODO: fix x86_emulate.c to use guest_read/write_register + * instead of direct ->regs accesses, can save hundred cycles + * on Intel for instructions that don't read/change RSP, for + * for example. + */ + cache_all_regs(vcpu); vcpu->mmio_is_write = 0; vcpu->arch.pio.string = 0; @@ -2172,7 +2188,6 @@ int emulate_instruction(struct kvm_vcpu *vcpu, return EMULATE_DO_MMIO; } - kvm_x86_ops->decache_regs(vcpu); kvm_x86_ops->set_rflags(vcpu, vcpu->arch.emulate_ctxt.eflags); if (vcpu->mmio_is_write) { @@ -2225,20 +2240,19 @@ int complete_pio(struct kvm_vcpu *vcpu) struct kvm_pio_request *io = &vcpu->arch.pio; long delta; int r; - - kvm_x86_ops->cache_regs(vcpu); + unsigned long val; if (!io->string) { - if (io->in) - memcpy(&vcpu->arch.regs[VCPU_REGS_RAX], vcpu->arch.pio_data, - io->size); + if (io->in) { + val = kvm_register_read(vcpu, VCPU_REGS_RAX); + memcpy(&val, vcpu->arch.pio_data, io->size); + kvm_register_write(vcpu, VCPU_REGS_RAX, val); + } } else { if (io->in) { r = pio_copy_data(vcpu); - if (r) { - kvm_x86_ops->cache_regs(vcpu); + if (r) return r; - } } delta = 1; @@ -2248,19 +2262,24 @@ int complete_pio(struct kvm_vcpu *vcpu) * The size of the register should really depend on * current address size. */ - vcpu->arch.regs[VCPU_REGS_RCX] -= delta; + val = kvm_register_read(vcpu, VCPU_REGS_RCX); + val -= delta; + kvm_register_write(vcpu, VCPU_REGS_RCX, val); } if (io->down) delta = -delta; delta *= io->size; - if (io->in) - vcpu->arch.regs[VCPU_REGS_RDI] += delta; - else - vcpu->arch.regs[VCPU_REGS_RSI] += delta; + if (io->in) { + val = kvm_register_read(vcpu, VCPU_REGS_RDI); + val += delta; + kvm_register_write(vcpu, VCPU_REGS_RDI, val); + } else { + val = kvm_register_read(vcpu, VCPU_REGS_RSI); + val += delta; + kvm_register_write(vcpu, VCPU_REGS_RSI, val); + } } - kvm_x86_ops->decache_regs(vcpu); - io->count -= io->cur_count; io->cur_count = 0; @@ -2313,6 +2332,7 @@ int kvm_emulate_pio(struct kvm_vcpu *vcpu, struct kvm_run *run, int in, int size, unsigned port) { struct kvm_io_device *pio_dev; + unsigned long val; vcpu->run->exit_reason = KVM_EXIT_IO; vcpu->run->io.direction = in ? KVM_EXIT_IO_IN : KVM_EXIT_IO_OUT; @@ -2333,8 +2353,8 @@ int kvm_emulate_pio(struct kvm_vcpu *vcpu, struct kvm_run *run, int in, KVMTRACE_2D(IO_WRITE, vcpu, vcpu->run->io.port, (u32)size, handler); - kvm_x86_ops->cache_regs(vcpu); - memcpy(vcpu->arch.pio_data, &vcpu->arch.regs[VCPU_REGS_RAX], 4); + val = kvm_register_read(vcpu, VCPU_REGS_RAX); + memcpy(vcpu->arch.pio_data, &val, 4); kvm_x86_ops->skip_emulated_instruction(vcpu); @@ -2519,13 +2539,11 @@ int kvm_emulate_hypercall(struct kvm_vcpu *vcpu) unsigned long nr, a0, a1, a2, a3, ret; int r = 1; - kvm_x86_ops->cache_regs(vcpu); - - nr = vcpu->arch.regs[VCPU_REGS_RAX]; - a0 = vcpu->arch.regs[VCPU_REGS_RBX]; - a1 = vcpu->arch.regs[VCPU_REGS_RCX]; - a2 = vcpu->arch.regs[VCPU_REGS_RDX]; - a3 = vcpu->arch.regs[VCPU_REGS_RSI]; + nr = kvm_register_read(vcpu, VCPU_REGS_RAX); + a0 = kvm_register_read(vcpu, VCPU_REGS_RBX); + a1 = kvm_register_read(vcpu, VCPU_REGS_RCX); + a2 = kvm_register_read(vcpu, VCPU_REGS_RDX); + a3 = kvm_register_read(vcpu, VCPU_REGS_RSI); KVMTRACE_1D(VMMCALL, vcpu, (u32)nr, handler); @@ -2548,8 +2566,7 @@ int kvm_emulate_hypercall(struct kvm_vcpu *vcpu) ret = -KVM_ENOSYS; break; } - vcpu->arch.regs[VCPU_REGS_RAX] = ret; - kvm_x86_ops->decache_regs(vcpu); + kvm_register_write(vcpu, VCPU_REGS_RAX, ret); ++vcpu->stat.hypercalls; return r; } @@ -2559,6 +2576,7 @@ int kvm_fix_hypercall(struct kvm_vcpu *vcpu) { char instruction[3]; int ret = 0; + unsigned long rip = kvm_rip_read(vcpu); /* @@ -2568,9 +2586,8 @@ int kvm_fix_hypercall(struct kvm_vcpu *vcpu) */ kvm_mmu_zap_all(vcpu->kvm); - kvm_x86_ops->cache_regs(vcpu); kvm_x86_ops->patch_hypercall(vcpu, instruction); - if (emulator_write_emulated(vcpu->arch.rip, instruction, 3, vcpu) + if (emulator_write_emulated(rip, instruction, 3, vcpu) != X86EMUL_CONTINUE) ret = -EFAULT; @@ -2700,13 +2717,12 @@ void kvm_emulate_cpuid(struct kvm_vcpu *vcpu) u32 function, index; struct kvm_cpuid_entry2 *e, *best; - kvm_x86_ops->cache_regs(vcpu); - function = vcpu->arch.regs[VCPU_REGS_RAX]; - index = vcpu->arch.regs[VCPU_REGS_RCX]; - vcpu->arch.regs[VCPU_REGS_RAX] = 0; - vcpu->arch.regs[VCPU_REGS_RBX] = 0; - vcpu->arch.regs[VCPU_REGS_RCX] = 0; - vcpu->arch.regs[VCPU_REGS_RDX] = 0; + function = kvm_register_read(vcpu, VCPU_REGS_RAX); + index = kvm_register_read(vcpu, VCPU_REGS_RCX); + kvm_register_write(vcpu, VCPU_REGS_RAX, 0); + kvm_register_write(vcpu, VCPU_REGS_RBX, 0); + kvm_register_write(vcpu, VCPU_REGS_RCX, 0); + kvm_register_write(vcpu, VCPU_REGS_RDX, 0); best = NULL; for (i = 0; i < vcpu->arch.cpuid_nent; ++i) { e = &vcpu->arch.cpuid_entries[i]; @@ -2724,18 +2740,17 @@ void kvm_emulate_cpuid(struct kvm_vcpu *vcpu) best = e; } if (best) { - vcpu->arch.regs[VCPU_REGS_RAX] = best->eax; - vcpu->arch.regs[VCPU_REGS_RBX] = best->ebx; - vcpu->arch.regs[VCPU_REGS_RCX] = best->ecx; - vcpu->arch.regs[VCPU_REGS_RDX] = best->edx; + kvm_register_write(vcpu, VCPU_REGS_RAX, best->eax); + kvm_register_write(vcpu, VCPU_REGS_RBX, best->ebx); + kvm_register_write(vcpu, VCPU_REGS_RCX, best->ecx); + kvm_register_write(vcpu, VCPU_REGS_RDX, best->edx); } - kvm_x86_ops->decache_regs(vcpu); kvm_x86_ops->skip_emulated_instruction(vcpu); KVMTRACE_5D(CPUID, vcpu, function, - (u32)vcpu->arch.regs[VCPU_REGS_RAX], - (u32)vcpu->arch.regs[VCPU_REGS_RBX], - (u32)vcpu->arch.regs[VCPU_REGS_RCX], - (u32)vcpu->arch.regs[VCPU_REGS_RDX], handler); + (u32)kvm_register_read(vcpu, VCPU_REGS_RAX), + (u32)kvm_register_read(vcpu, VCPU_REGS_RBX), + (u32)kvm_register_read(vcpu, VCPU_REGS_RCX), + (u32)kvm_register_read(vcpu, VCPU_REGS_RDX), handler); } EXPORT_SYMBOL_GPL(kvm_emulate_cpuid); @@ -2917,8 +2932,8 @@ again: * Profile KVM exit RIPs: */ if (unlikely(prof_on == KVM_PROFILING)) { - kvm_x86_ops->cache_regs(vcpu); - profile_hit(KVM_PROFILING, (void *)vcpu->arch.rip); + unsigned long rip = kvm_rip_read(vcpu); + profile_hit(KVM_PROFILING, (void *)rip); } if (vcpu->arch.exception.pending && kvm_x86_ops->exception_injected(vcpu)) @@ -2999,11 +3014,9 @@ int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run) } } #endif - if (kvm_run->exit_reason == KVM_EXIT_HYPERCALL) { - kvm_x86_ops->cache_regs(vcpu); - vcpu->arch.regs[VCPU_REGS_RAX] = kvm_run->hypercall.ret; - kvm_x86_ops->decache_regs(vcpu); - } + if (kvm_run->exit_reason == KVM_EXIT_HYPERCALL) + kvm_register_write(vcpu, VCPU_REGS_RAX, + kvm_run->hypercall.ret); r = __vcpu_run(vcpu, kvm_run); @@ -3019,28 +3032,26 @@ int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs) { vcpu_load(vcpu); - kvm_x86_ops->cache_regs(vcpu); - - regs->rax = vcpu->arch.regs[VCPU_REGS_RAX]; - regs->rbx = vcpu->arch.regs[VCPU_REGS_RBX]; - regs->rcx = vcpu->arch.regs[VCPU_REGS_RCX]; - regs->rdx = vcpu->arch.regs[VCPU_REGS_RDX]; - regs->rsi = vcpu->arch.regs[VCPU_REGS_RSI]; - regs->rdi = vcpu->arch.regs[VCPU_REGS_RDI]; - regs->rsp = vcpu->arch.regs[VCPU_REGS_RSP]; - regs->rbp = vcpu->arch.regs[VCPU_REGS_RBP]; + regs->rax = kvm_register_read(vcpu, VCPU_REGS_RAX); + regs->rbx = kvm_register_read(vcpu, VCPU_REGS_RBX); + regs->rcx = kvm_register_read(vcpu, VCPU_REGS_RCX); + regs->rdx = kvm_register_read(vcpu, VCPU_REGS_RDX); + regs->rsi = kvm_register_read(vcpu, VCPU_REGS_RSI); + regs->rdi = kvm_register_read(vcpu, VCPU_REGS_RDI); + regs->rsp = kvm_register_read(vcpu, VCPU_REGS_RSP); + regs->rbp = kvm_register_read(vcpu, VCPU_REGS_RBP); #ifdef CONFIG_X86_64 - regs->r8 = vcpu->arch.regs[VCPU_REGS_R8]; - regs->r9 = vcpu->arch.regs[VCPU_REGS_R9]; - regs->r10 = vcpu->arch.regs[VCPU_REGS_R10]; - regs->r11 = vcpu->arch.regs[VCPU_REGS_R11]; - regs->r12 = vcpu->arch.regs[VCPU_REGS_R12]; - regs->r13 = vcpu->arch.regs[VCPU_REGS_R13]; - regs->r14 = vcpu->arch.regs[VCPU_REGS_R14]; - regs->r15 = vcpu->arch.regs[VCPU_REGS_R15]; + regs->r8 = kvm_register_read(vcpu, VCPU_REGS_R8); + regs->r9 = kvm_register_read(vcpu, VCPU_REGS_R9); + regs->r10 = kvm_register_read(vcpu, VCPU_REGS_R10); + regs->r11 = kvm_register_read(vcpu, VCPU_REGS_R11); + regs->r12 = kvm_register_read(vcpu, VCPU_REGS_R12); + regs->r13 = kvm_register_read(vcpu, VCPU_REGS_R13); + regs->r14 = kvm_register_read(vcpu, VCPU_REGS_R14); + regs->r15 = kvm_register_read(vcpu, VCPU_REGS_R15); #endif - regs->rip = vcpu->arch.rip; + regs->rip = kvm_rip_read(vcpu); regs->rflags = kvm_x86_ops->get_rflags(vcpu); /* @@ -3058,29 +3069,29 @@ int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs) { vcpu_load(vcpu); - vcpu->arch.regs[VCPU_REGS_RAX] = regs->rax; - vcpu->arch.regs[VCPU_REGS_RBX] = regs->rbx; - vcpu->arch.regs[VCPU_REGS_RCX] = regs->rcx; - vcpu->arch.regs[VCPU_REGS_RDX] = regs->rdx; - vcpu->arch.regs[VCPU_REGS_RSI] = regs->rsi; - vcpu->arch.regs[VCPU_REGS_RDI] = regs->rdi; - vcpu->arch.regs[VCPU_REGS_RSP] = regs->rsp; - vcpu->arch.regs[VCPU_REGS_RBP] = regs->rbp; + kvm_register_write(vcpu, VCPU_REGS_RAX, regs->rax); + kvm_register_write(vcpu, VCPU_REGS_RBX, regs->rbx); + kvm_register_write(vcpu, VCPU_REGS_RCX, regs->rcx); + kvm_register_write(vcpu, VCPU_REGS_RDX, regs->rdx); + kvm_register_write(vcpu, VCPU_REGS_RSI, regs->rsi); + kvm_register_write(vcpu, VCPU_REGS_RDI, regs->rdi); + kvm_register_write(vcpu, VCPU_REGS_RSP, regs->rsp); + kvm_register_write(vcpu, VCPU_REGS_RBP, regs->rbp); #ifdef CONFIG_X86_64 - vcpu->arch.regs[VCPU_REGS_R8] = regs->r8; - vcpu->arch.regs[VCPU_REGS_R9] = regs->r9; - vcpu->arch.regs[VCPU_REGS_R10] = regs->r10; - vcpu->arch.regs[VCPU_REGS_R11] = regs->r11; - vcpu->arch.regs[VCPU_REGS_R12] = regs->r12; - vcpu->arch.regs[VCPU_REGS_R13] = regs->r13; - vcpu->arch.regs[VCPU_REGS_R14] = regs->r14; - vcpu->arch.regs[VCPU_REGS_R15] = regs->r15; + kvm_register_write(vcpu, VCPU_REGS_R8, regs->r8); + kvm_register_write(vcpu, VCPU_REGS_R9, regs->r9); + kvm_register_write(vcpu, VCPU_REGS_R10, regs->r10); + kvm_register_write(vcpu, VCPU_REGS_R11, regs->r11); + kvm_register_write(vcpu, VCPU_REGS_R12, regs->r12); + kvm_register_write(vcpu, VCPU_REGS_R13, regs->r13); + kvm_register_write(vcpu, VCPU_REGS_R14, regs->r14); + kvm_register_write(vcpu, VCPU_REGS_R15, regs->r15); + #endif - vcpu->arch.rip = regs->rip; + kvm_rip_write(vcpu, regs->rip); kvm_x86_ops->set_rflags(vcpu, regs->rflags); - kvm_x86_ops->decache_regs(vcpu); vcpu->arch.exception.pending = false; @@ -3316,17 +3327,16 @@ static void save_state_to_tss32(struct kvm_vcpu *vcpu, struct tss_segment_32 *tss) { tss->cr3 = vcpu->arch.cr3; - tss->eip = vcpu->arch.rip; + tss->eip = kvm_rip_read(vcpu); tss->eflags = kvm_x86_ops->get_rflags(vcpu); - tss->eax = vcpu->arch.regs[VCPU_REGS_RAX]; - tss->ecx = vcpu->arch.regs[VCPU_REGS_RCX]; - tss->edx = vcpu->arch.regs[VCPU_REGS_RDX]; - tss->ebx = vcpu->arch.regs[VCPU_REGS_RBX]; - tss->esp = vcpu->arch.regs[VCPU_REGS_RSP]; - tss->ebp = vcpu->arch.regs[VCPU_REGS_RBP]; - tss->esi = vcpu->arch.regs[VCPU_REGS_RSI]; - tss->edi = vcpu->arch.regs[VCPU_REGS_RDI]; - + tss->eax = kvm_register_read(vcpu, VCPU_REGS_RAX); + tss->ecx = kvm_register_read(vcpu, VCPU_REGS_RCX); + tss->edx = kvm_register_read(vcpu, VCPU_REGS_RDX); + tss->ebx = kvm_register_read(vcpu, VCPU_REGS_RBX); + tss->esp = kvm_register_read(vcpu, VCPU_REGS_RSP); + tss->ebp = kvm_register_read(vcpu, VCPU_REGS_RBP); + tss->esi = kvm_register_read(vcpu, VCPU_REGS_RSI); + tss->edi = kvm_register_read(vcpu, VCPU_REGS_RDI); tss->es = get_segment_selector(vcpu, VCPU_SREG_ES); tss->cs = get_segment_selector(vcpu, VCPU_SREG_CS); tss->ss = get_segment_selector(vcpu, VCPU_SREG_SS); @@ -3342,17 +3352,17 @@ static int load_state_from_tss32(struct kvm_vcpu *vcpu, { kvm_set_cr3(vcpu, tss->cr3); - vcpu->arch.rip = tss->eip; + kvm_rip_write(vcpu, tss->eip); kvm_x86_ops->set_rflags(vcpu, tss->eflags | 2); - vcpu->arch.regs[VCPU_REGS_RAX] = tss->eax; - vcpu->arch.regs[VCPU_REGS_RCX] = tss->ecx; - vcpu->arch.regs[VCPU_REGS_RDX] = tss->edx; - vcpu->arch.regs[VCPU_REGS_RBX] = tss->ebx; - vcpu->arch.regs[VCPU_REGS_RSP] = tss->esp; - vcpu->arch.regs[VCPU_REGS_RBP] = tss->ebp; - vcpu->arch.regs[VCPU_REGS_RSI] = tss->esi; - vcpu->arch.regs[VCPU_REGS_RDI] = tss->edi; + kvm_register_write(vcpu, VCPU_REGS_RAX, tss->eax); + kvm_register_write(vcpu, VCPU_REGS_RCX, tss->ecx); + kvm_register_write(vcpu, VCPU_REGS_RDX, tss->edx); + kvm_register_write(vcpu, VCPU_REGS_RBX, tss->ebx); + kvm_register_write(vcpu, VCPU_REGS_RSP, tss->esp); + kvm_register_write(vcpu, VCPU_REGS_RBP, tss->ebp); + kvm_register_write(vcpu, VCPU_REGS_RSI, tss->esi); + kvm_register_write(vcpu, VCPU_REGS_RDI, tss->edi); if (kvm_load_segment_descriptor(vcpu, tss->ldt_selector, 0, VCPU_SREG_LDTR)) return 1; @@ -3380,16 +3390,16 @@ static int load_state_from_tss32(struct kvm_vcpu *vcpu, static void save_state_to_tss16(struct kvm_vcpu *vcpu, struct tss_segment_16 *tss) { - tss->ip = vcpu->arch.rip; + tss->ip = kvm_rip_read(vcpu); tss->flag = kvm_x86_ops->get_rflags(vcpu); - tss->ax = vcpu->arch.regs[VCPU_REGS_RAX]; - tss->cx = vcpu->arch.regs[VCPU_REGS_RCX]; - tss->dx = vcpu->arch.regs[VCPU_REGS_RDX]; - tss->bx = vcpu->arch.regs[VCPU_REGS_RBX]; - tss->sp = vcpu->arch.regs[VCPU_REGS_RSP]; - tss->bp = vcpu->arch.regs[VCPU_REGS_RBP]; - tss->si = vcpu->arch.regs[VCPU_REGS_RSI]; - tss->di = vcpu->arch.regs[VCPU_REGS_RDI]; + tss->ax = kvm_register_read(vcpu, VCPU_REGS_RAX); + tss->cx = kvm_register_read(vcpu, VCPU_REGS_RCX); + tss->dx = kvm_register_read(vcpu, VCPU_REGS_RDX); + tss->bx = kvm_register_read(vcpu, VCPU_REGS_RBX); + tss->sp = kvm_register_read(vcpu, VCPU_REGS_RSP); + tss->bp = kvm_register_read(vcpu, VCPU_REGS_RBP); + tss->si = kvm_register_read(vcpu, VCPU_REGS_RSI); + tss->di = kvm_register_read(vcpu, VCPU_REGS_RDI); tss->es = get_segment_selector(vcpu, VCPU_SREG_ES); tss->cs = get_segment_selector(vcpu, VCPU_SREG_CS); @@ -3402,16 +3412,16 @@ static void save_state_to_tss16(struct kvm_vcpu *vcpu, static int load_state_from_tss16(struct kvm_vcpu *vcpu, struct tss_segment_16 *tss) { - vcpu->arch.rip = tss->ip; + kvm_rip_write(vcpu, tss->ip); kvm_x86_ops->set_rflags(vcpu, tss->flag | 2); - vcpu->arch.regs[VCPU_REGS_RAX] = tss->ax; - vcpu->arch.regs[VCPU_REGS_RCX] = tss->cx; - vcpu->arch.regs[VCPU_REGS_RDX] = tss->dx; - vcpu->arch.regs[VCPU_REGS_RBX] = tss->bx; - vcpu->arch.regs[VCPU_REGS_RSP] = tss->sp; - vcpu->arch.regs[VCPU_REGS_RBP] = tss->bp; - vcpu->arch.regs[VCPU_REGS_RSI] = tss->si; - vcpu->arch.regs[VCPU_REGS_RDI] = tss->di; + kvm_register_write(vcpu, VCPU_REGS_RAX, tss->ax); + kvm_register_write(vcpu, VCPU_REGS_RCX, tss->cx); + kvm_register_write(vcpu, VCPU_REGS_RDX, tss->dx); + kvm_register_write(vcpu, VCPU_REGS_RBX, tss->bx); + kvm_register_write(vcpu, VCPU_REGS_RSP, tss->sp); + kvm_register_write(vcpu, VCPU_REGS_RBP, tss->bp); + kvm_register_write(vcpu, VCPU_REGS_RSI, tss->si); + kvm_register_write(vcpu, VCPU_REGS_RDI, tss->di); if (kvm_load_segment_descriptor(vcpu, tss->ldt, 0, VCPU_SREG_LDTR)) return 1; @@ -3534,7 +3544,6 @@ int kvm_task_switch(struct kvm_vcpu *vcpu, u16 tss_selector, int reason) } kvm_x86_ops->skip_emulated_instruction(vcpu); - kvm_x86_ops->cache_regs(vcpu); if (nseg_desc.type & 8) ret = kvm_task_switch_32(vcpu, tss_selector, old_tss_base, @@ -3559,7 +3568,6 @@ int kvm_task_switch(struct kvm_vcpu *vcpu, u16 tss_selector, int reason) tr_seg.type = 11; kvm_set_segment(vcpu, &tr_seg, VCPU_SREG_TR); out: - kvm_x86_ops->decache_regs(vcpu); return ret; } EXPORT_SYMBOL_GPL(kvm_task_switch); -- cgit v1.2.2 From 26eef70c3e8c76e73dff2579c792fc7355f8a291 Mon Sep 17 00:00:00 2001 From: Avi Kivity Date: Thu, 3 Jul 2008 14:59:22 +0300 Subject: KVM: Clear exception queue before emulating an instruction If we're emulating an instruction, either it will succeed, in which case any previously queued exception will be spurious, or we will requeue the same exception. Signed-off-by: Avi Kivity --- arch/x86/kvm/x86.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'arch/x86/kvm/x86.c') diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c index 2f0696bc7d2f..5620df2685db 100644 --- a/arch/x86/kvm/x86.c +++ b/arch/x86/kvm/x86.c @@ -20,6 +20,7 @@ #include "i8254.h" #include "tss.h" #include "kvm_cache_regs.h" +#include "x86.h" #include #include @@ -2121,6 +2122,7 @@ int emulate_instruction(struct kvm_vcpu *vcpu, int r; struct decode_cache *c; + kvm_clear_exception_queue(vcpu); vcpu->arch.mmio_fault_cr2 = cr2; /* * TODO: fix x86_emulate.c to use guest_read/write_register -- cgit v1.2.2 From b5e2fec0ebc3fcaff954092bb69444a67a904c0a Mon Sep 17 00:00:00 2001 From: Alexander Graf Date: Tue, 22 Jul 2008 08:00:45 +0200 Subject: KVM: Ignore DEBUGCTL MSRs with no effect Netware writes to DEBUGCTL and reads from the DEBUGCTL and LAST*IP MSRs without further checks and is really confused to receive a #GP during that. To make it happy we should just make them stubs, which is exactly what SVM already does. Writes to DEBUGCTL that are vendor-specific are resembled to behave as if the virtual CPU does not know them. Signed-off-by: Alexander Graf Signed-off-by: Avi Kivity --- arch/x86/kvm/x86.c | 17 +++++++++++++++++ 1 file changed, 17 insertions(+) (limited to 'arch/x86/kvm/x86.c') diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c index 5620df2685db..94a216562f10 100644 --- a/arch/x86/kvm/x86.c +++ b/arch/x86/kvm/x86.c @@ -665,6 +665,18 @@ int kvm_set_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 data) pr_unimpl(vcpu, "%s: MSR_IA32_MCG_CTL 0x%llx, nop\n", __func__, data); break; + case MSR_IA32_DEBUGCTLMSR: + if (!data) { + /* We support the non-activated case already */ + break; + } else if (data & ~(DEBUGCTLMSR_LBR | DEBUGCTLMSR_BTF)) { + /* Values other than LBR and BTF are vendor-specific, + thus reserved and should throw a #GP */ + return 1; + } + pr_unimpl(vcpu, "%s: MSR_IA32_DEBUGCTLMSR 0x%llx, nop\n", + __func__, data); + break; case MSR_IA32_UCODE_REV: case MSR_IA32_UCODE_WRITE: break; @@ -757,6 +769,11 @@ int kvm_get_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 *pdata) case MSR_IA32_MC0_MISC+16: case MSR_IA32_UCODE_REV: case MSR_IA32_EBL_CR_POWERON: + case MSR_IA32_DEBUGCTLMSR: + case MSR_IA32_LASTBRANCHFROMIP: + case MSR_IA32_LASTBRANCHTOIP: + case MSR_IA32_LASTINTFROMIP: + case MSR_IA32_LASTINTTOIP: data = 0; break; case MSR_MTRRcap: -- cgit v1.2.2 From 4d5c5d0fe89c921336b95f5e7e4f529a9df92f53 Mon Sep 17 00:00:00 2001 From: Ben-Ami Yassour Date: Mon, 28 Jul 2008 19:26:26 +0300 Subject: KVM: pci device assignment Based on a patch from: Amit Shah This patch adds support for handling PCI devices that are assigned to the guest. The device to be assigned to the guest is registered in the host kernel and interrupt delivery is handled. If a device is already assigned, or the device driver for it is still loaded on the host, the device assignment is failed by conveying a -EBUSY reply to the userspace. Devices that share their interrupt line are not supported at the moment. By itself, this patch will not make devices work within the guest. The VT-d extension is required to enable the device to perform DMA. Another alternative is PVDMA. Signed-off-by: Amit Shah Signed-off-by: Ben-Ami Yassour Signed-off-by: Weidong Han Signed-off-by: Avi Kivity --- arch/x86/kvm/x86.c | 243 +++++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 243 insertions(+) (limited to 'arch/x86/kvm/x86.c') diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c index 94a216562f10..a97157cc42ae 100644 --- a/arch/x86/kvm/x86.c +++ b/arch/x86/kvm/x86.c @@ -4,10 +4,14 @@ * derived from drivers/kvm/kvm_main.c * * Copyright (C) 2006 Qumranet, Inc. + * Copyright (C) 2008 Qumranet, Inc. + * Copyright IBM Corporation, 2008 * * Authors: * Avi Kivity * Yaniv Kamay + * Amit Shah + * Ben-Ami Yassour * * This work is licensed under the terms of the GNU GPL, version 2. See * the COPYING file in the top-level directory. @@ -23,8 +27,10 @@ #include "x86.h" #include +#include #include #include +#include #include #include #include @@ -98,6 +104,219 @@ struct kvm_stats_debugfs_item debugfs_entries[] = { { NULL } }; +struct kvm_assigned_dev_kernel *kvm_find_assigned_dev(struct list_head *head, + int assigned_dev_id) +{ + struct list_head *ptr; + struct kvm_assigned_dev_kernel *match; + + list_for_each(ptr, head) { + match = list_entry(ptr, struct kvm_assigned_dev_kernel, list); + if (match->assigned_dev_id == assigned_dev_id) + return match; + } + return NULL; +} + +static void kvm_assigned_dev_interrupt_work_handler(struct work_struct *work) +{ + struct kvm_assigned_dev_kernel *assigned_dev; + + assigned_dev = container_of(work, struct kvm_assigned_dev_kernel, + interrupt_work); + + /* This is taken to safely inject irq inside the guest. When + * the interrupt injection (or the ioapic code) uses a + * finer-grained lock, update this + */ + mutex_lock(&assigned_dev->kvm->lock); + kvm_set_irq(assigned_dev->kvm, + assigned_dev->guest_irq, 1); + mutex_unlock(&assigned_dev->kvm->lock); + kvm_put_kvm(assigned_dev->kvm); +} + +/* FIXME: Implement the OR logic needed to make shared interrupts on + * this line behave properly + */ +static irqreturn_t kvm_assigned_dev_intr(int irq, void *dev_id) +{ + struct kvm_assigned_dev_kernel *assigned_dev = + (struct kvm_assigned_dev_kernel *) dev_id; + + kvm_get_kvm(assigned_dev->kvm); + schedule_work(&assigned_dev->interrupt_work); + disable_irq_nosync(irq); + return IRQ_HANDLED; +} + +/* Ack the irq line for an assigned device */ +static void kvm_assigned_dev_ack_irq(struct kvm_irq_ack_notifier *kian) +{ + struct kvm_assigned_dev_kernel *dev; + + if (kian->gsi == -1) + return; + + dev = container_of(kian, struct kvm_assigned_dev_kernel, + ack_notifier); + kvm_set_irq(dev->kvm, dev->guest_irq, 0); + enable_irq(dev->host_irq); +} + +static int kvm_vm_ioctl_assign_irq(struct kvm *kvm, + struct kvm_assigned_irq + *assigned_irq) +{ + int r = 0; + struct kvm_assigned_dev_kernel *match; + + mutex_lock(&kvm->lock); + + match = kvm_find_assigned_dev(&kvm->arch.assigned_dev_head, + assigned_irq->assigned_dev_id); + if (!match) { + mutex_unlock(&kvm->lock); + return -EINVAL; + } + + if (match->irq_requested) { + match->guest_irq = assigned_irq->guest_irq; + match->ack_notifier.gsi = assigned_irq->guest_irq; + mutex_unlock(&kvm->lock); + return 0; + } + + INIT_WORK(&match->interrupt_work, + kvm_assigned_dev_interrupt_work_handler); + + if (irqchip_in_kernel(kvm)) { + if (assigned_irq->host_irq) + match->host_irq = assigned_irq->host_irq; + else + match->host_irq = match->dev->irq; + match->guest_irq = assigned_irq->guest_irq; + match->ack_notifier.gsi = assigned_irq->guest_irq; + match->ack_notifier.irq_acked = kvm_assigned_dev_ack_irq; + kvm_register_irq_ack_notifier(kvm, &match->ack_notifier); + + /* Even though this is PCI, we don't want to use shared + * interrupts. Sharing host devices with guest-assigned devices + * on the same interrupt line is not a happy situation: there + * are going to be long delays in accepting, acking, etc. + */ + if (request_irq(match->host_irq, kvm_assigned_dev_intr, 0, + "kvm_assigned_device", (void *)match)) { + printk(KERN_INFO "%s: couldn't allocate irq for pv " + "device\n", __func__); + r = -EIO; + goto out; + } + } + + match->irq_requested = true; +out: + mutex_unlock(&kvm->lock); + return r; +} + +static int kvm_vm_ioctl_assign_device(struct kvm *kvm, + struct kvm_assigned_pci_dev *assigned_dev) +{ + int r = 0; + struct kvm_assigned_dev_kernel *match; + struct pci_dev *dev; + + mutex_lock(&kvm->lock); + + match = kvm_find_assigned_dev(&kvm->arch.assigned_dev_head, + assigned_dev->assigned_dev_id); + if (match) { + /* device already assigned */ + r = -EINVAL; + goto out; + } + + match = kzalloc(sizeof(struct kvm_assigned_dev_kernel), GFP_KERNEL); + if (match == NULL) { + printk(KERN_INFO "%s: Couldn't allocate memory\n", + __func__); + r = -ENOMEM; + goto out; + } + dev = pci_get_bus_and_slot(assigned_dev->busnr, + assigned_dev->devfn); + if (!dev) { + printk(KERN_INFO "%s: host device not found\n", __func__); + r = -EINVAL; + goto out_free; + } + if (pci_enable_device(dev)) { + printk(KERN_INFO "%s: Could not enable PCI device\n", __func__); + r = -EBUSY; + goto out_put; + } + r = pci_request_regions(dev, "kvm_assigned_device"); + if (r) { + printk(KERN_INFO "%s: Could not get access to device regions\n", + __func__); + goto out_disable; + } + match->assigned_dev_id = assigned_dev->assigned_dev_id; + match->host_busnr = assigned_dev->busnr; + match->host_devfn = assigned_dev->devfn; + match->dev = dev; + + match->kvm = kvm; + + list_add(&match->list, &kvm->arch.assigned_dev_head); + +out: + mutex_unlock(&kvm->lock); + return r; +out_disable: + pci_disable_device(dev); +out_put: + pci_dev_put(dev); +out_free: + kfree(match); + mutex_unlock(&kvm->lock); + return r; +} + +static void kvm_free_assigned_devices(struct kvm *kvm) +{ + struct list_head *ptr, *ptr2; + struct kvm_assigned_dev_kernel *assigned_dev; + + list_for_each_safe(ptr, ptr2, &kvm->arch.assigned_dev_head) { + assigned_dev = list_entry(ptr, + struct kvm_assigned_dev_kernel, + list); + + if (irqchip_in_kernel(kvm) && assigned_dev->irq_requested) { + free_irq(assigned_dev->host_irq, + (void *)assigned_dev); + + kvm_unregister_irq_ack_notifier(kvm, + &assigned_dev-> + ack_notifier); + } + + if (cancel_work_sync(&assigned_dev->interrupt_work)) + /* We had pending work. That means we will have to take + * care of kvm_put_kvm. + */ + kvm_put_kvm(kvm); + + pci_release_regions(assigned_dev->dev); + pci_disable_device(assigned_dev->dev); + pci_dev_put(assigned_dev->dev); + + list_del(&assigned_dev->list); + kfree(assigned_dev); + } +} unsigned long segment_base(u16 selector) { @@ -1766,6 +1985,28 @@ long kvm_arch_vm_ioctl(struct file *filp, r = 0; break; } + case KVM_ASSIGN_PCI_DEVICE: { + struct kvm_assigned_pci_dev assigned_dev; + + r = -EFAULT; + if (copy_from_user(&assigned_dev, argp, sizeof assigned_dev)) + goto out; + r = kvm_vm_ioctl_assign_device(kvm, &assigned_dev); + if (r) + goto out; + break; + } + case KVM_ASSIGN_IRQ: { + struct kvm_assigned_irq assigned_irq; + + r = -EFAULT; + if (copy_from_user(&assigned_irq, argp, sizeof assigned_irq)) + goto out; + r = kvm_vm_ioctl_assign_irq(kvm, &assigned_irq); + if (r) + goto out; + break; + } case KVM_GET_PIT: { struct kvm_pit_state ps; r = -EFAULT; @@ -3945,6 +4186,7 @@ struct kvm *kvm_arch_create_vm(void) return ERR_PTR(-ENOMEM); INIT_LIST_HEAD(&kvm->arch.active_mmu_pages); + INIT_LIST_HEAD(&kvm->arch.assigned_dev_head); return kvm; } @@ -3977,6 +4219,7 @@ static void kvm_free_vcpus(struct kvm *kvm) void kvm_arch_destroy_vm(struct kvm *kvm) { + kvm_free_assigned_devices(kvm); kvm_free_pit(kvm); kfree(kvm->arch.vpic); kfree(kvm->arch.vioapic); -- cgit v1.2.2 From f0d662759a2465babdba1160749c446648c9d159 Mon Sep 17 00:00:00 2001 From: Dave Hansen Date: Mon, 11 Aug 2008 10:01:45 -0700 Subject: KVM: Reduce kvm stack usage in kvm_arch_vm_ioctl() On my machine with gcc 3.4, kvm uses ~2k of stack in a few select functions. This is mostly because gcc fails to notice that the different case: statements could have their stack usage combined. It overflows very nicely if interrupts happen during one of these large uses. This patch uses two methods for reducing stack usage. 1. dynamically allocate large objects instead of putting on the stack. 2. Use a union{} member for all of the case variables. This tricks gcc into combining them all into a single stack allocation. (There's also a comment on this) Signed-off-by: Dave Hansen Signed-off-by: Avi Kivity --- arch/x86/kvm/x86.c | 72 ++++++++++++++++++++++++++++++++++-------------------- 1 file changed, 45 insertions(+), 27 deletions(-) (limited to 'arch/x86/kvm/x86.c') diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c index a97157cc42ae..87d434228fe2 100644 --- a/arch/x86/kvm/x86.c +++ b/arch/x86/kvm/x86.c @@ -1869,6 +1869,15 @@ long kvm_arch_vm_ioctl(struct file *filp, struct kvm *kvm = filp->private_data; void __user *argp = (void __user *)arg; int r = -EINVAL; + /* + * This union makes it completely explicit to gcc-3.x + * that these two variables' stack usage should be + * combined, not added together. + */ + union { + struct kvm_pit_state ps; + struct kvm_memory_alias alias; + } u; switch (ioctl) { case KVM_SET_TSS_ADDR: @@ -1900,17 +1909,14 @@ long kvm_arch_vm_ioctl(struct file *filp, case KVM_GET_NR_MMU_PAGES: r = kvm_vm_ioctl_get_nr_mmu_pages(kvm); break; - case KVM_SET_MEMORY_ALIAS: { - struct kvm_memory_alias alias; - + case KVM_SET_MEMORY_ALIAS: r = -EFAULT; - if (copy_from_user(&alias, argp, sizeof alias)) + if (copy_from_user(&u.alias, argp, sizeof(struct kvm_memory_alias))) goto out; - r = kvm_vm_ioctl_set_memory_alias(kvm, &alias); + r = kvm_vm_ioctl_set_memory_alias(kvm, &u.alias); if (r) goto out; break; - } case KVM_CREATE_IRQCHIP: r = -ENOMEM; kvm->arch.vpic = kvm_create_pic(kvm); @@ -1952,37 +1958,51 @@ long kvm_arch_vm_ioctl(struct file *filp, } case KVM_GET_IRQCHIP: { /* 0: PIC master, 1: PIC slave, 2: IOAPIC */ - struct kvm_irqchip chip; + struct kvm_irqchip *chip = kmalloc(sizeof(*chip), GFP_KERNEL); - r = -EFAULT; - if (copy_from_user(&chip, argp, sizeof chip)) + r = -ENOMEM; + if (!chip) goto out; + r = -EFAULT; + if (copy_from_user(chip, argp, sizeof *chip)) + goto get_irqchip_out; r = -ENXIO; if (!irqchip_in_kernel(kvm)) - goto out; - r = kvm_vm_ioctl_get_irqchip(kvm, &chip); + goto get_irqchip_out; + r = kvm_vm_ioctl_get_irqchip(kvm, chip); if (r) - goto out; + goto get_irqchip_out; r = -EFAULT; - if (copy_to_user(argp, &chip, sizeof chip)) - goto out; + if (copy_to_user(argp, chip, sizeof *chip)) + goto get_irqchip_out; r = 0; + get_irqchip_out: + kfree(chip); + if (r) + goto out; break; } case KVM_SET_IRQCHIP: { /* 0: PIC master, 1: PIC slave, 2: IOAPIC */ - struct kvm_irqchip chip; + struct kvm_irqchip *chip = kmalloc(sizeof(*chip), GFP_KERNEL); - r = -EFAULT; - if (copy_from_user(&chip, argp, sizeof chip)) + r = -ENOMEM; + if (!chip) goto out; + r = -EFAULT; + if (copy_from_user(chip, argp, sizeof *chip)) + goto set_irqchip_out; r = -ENXIO; if (!irqchip_in_kernel(kvm)) - goto out; - r = kvm_vm_ioctl_set_irqchip(kvm, &chip); + goto set_irqchip_out; + r = kvm_vm_ioctl_set_irqchip(kvm, chip); if (r) - goto out; + goto set_irqchip_out; r = 0; + set_irqchip_out: + kfree(chip); + if (r) + goto out; break; } case KVM_ASSIGN_PCI_DEVICE: { @@ -2008,31 +2028,29 @@ long kvm_arch_vm_ioctl(struct file *filp, break; } case KVM_GET_PIT: { - struct kvm_pit_state ps; r = -EFAULT; - if (copy_from_user(&ps, argp, sizeof ps)) + if (copy_from_user(&u.ps, argp, sizeof(struct kvm_pit_state))) goto out; r = -ENXIO; if (!kvm->arch.vpit) goto out; - r = kvm_vm_ioctl_get_pit(kvm, &ps); + r = kvm_vm_ioctl_get_pit(kvm, &u.ps); if (r) goto out; r = -EFAULT; - if (copy_to_user(argp, &ps, sizeof ps)) + if (copy_to_user(argp, &u.ps, sizeof(struct kvm_pit_state))) goto out; r = 0; break; } case KVM_SET_PIT: { - struct kvm_pit_state ps; r = -EFAULT; - if (copy_from_user(&ps, argp, sizeof ps)) + if (copy_from_user(&u.ps, argp, sizeof u.ps)) goto out; r = -ENXIO; if (!kvm->arch.vpit) goto out; - r = kvm_vm_ioctl_set_pit(kvm, &ps); + r = kvm_vm_ioctl_set_pit(kvm, &u.ps); if (r) goto out; r = 0; -- cgit v1.2.2 From b772ff362ec6b821c8a5227a3355e263f917bfad Mon Sep 17 00:00:00 2001 From: Dave Hansen Date: Mon, 11 Aug 2008 10:01:47 -0700 Subject: KVM: Reduce stack usage in kvm_arch_vcpu_ioctl() [sheng: fix KVM_GET_LAPIC using wrong size] Signed-off-by: Dave Hansen Signed-off-by: Sheng Yang Signed-off-by: Avi Kivity --- arch/x86/kvm/x86.c | 23 +++++++++++++++-------- 1 file changed, 15 insertions(+), 8 deletions(-) (limited to 'arch/x86/kvm/x86.c') diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c index 87d434228fe2..f1b0223c4088 100644 --- a/arch/x86/kvm/x86.c +++ b/arch/x86/kvm/x86.c @@ -1542,28 +1542,33 @@ long kvm_arch_vcpu_ioctl(struct file *filp, struct kvm_vcpu *vcpu = filp->private_data; void __user *argp = (void __user *)arg; int r; + struct kvm_lapic_state *lapic = NULL; switch (ioctl) { case KVM_GET_LAPIC: { - struct kvm_lapic_state lapic; + lapic = kzalloc(sizeof(struct kvm_lapic_state), GFP_KERNEL); - memset(&lapic, 0, sizeof lapic); - r = kvm_vcpu_ioctl_get_lapic(vcpu, &lapic); + r = -ENOMEM; + if (!lapic) + goto out; + r = kvm_vcpu_ioctl_get_lapic(vcpu, lapic); if (r) goto out; r = -EFAULT; - if (copy_to_user(argp, &lapic, sizeof lapic)) + if (copy_to_user(argp, lapic, sizeof(struct kvm_lapic_state))) goto out; r = 0; break; } case KVM_SET_LAPIC: { - struct kvm_lapic_state lapic; - + lapic = kmalloc(sizeof(struct kvm_lapic_state), GFP_KERNEL); + r = -ENOMEM; + if (!lapic) + goto out; r = -EFAULT; - if (copy_from_user(&lapic, argp, sizeof lapic)) + if (copy_from_user(lapic, argp, sizeof(struct kvm_lapic_state))) goto out; - r = kvm_vcpu_ioctl_set_lapic(vcpu, &lapic);; + r = kvm_vcpu_ioctl_set_lapic(vcpu, lapic); if (r) goto out; r = 0; @@ -1661,6 +1666,8 @@ long kvm_arch_vcpu_ioctl(struct file *filp, r = -EINVAL; } out: + if (lapic) + kfree(lapic); return r; } -- cgit v1.2.2 From 29415c37f043d1d54dcf356601d738ff6633b72b Mon Sep 17 00:00:00 2001 From: Marcelo Tosatti Date: Fri, 1 Aug 2008 20:09:13 -0300 Subject: KVM: set debug registers after "schedulable" section The vcpu thread can be preempted after the guest_debug_pre() callback, resulting in invalid debug registers on the new vcpu. Move it inside the non-preemptable section. Signed-off-by: Marcelo Tosatti Signed-off-by: Avi Kivity --- arch/x86/kvm/x86.c | 9 ++++----- 1 file changed, 4 insertions(+), 5 deletions(-) (limited to 'arch/x86/kvm/x86.c') diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c index f1b0223c4088..4a033757a19e 100644 --- a/arch/x86/kvm/x86.c +++ b/arch/x86/kvm/x86.c @@ -3113,10 +3113,6 @@ static int __vcpu_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run) down_read(&vcpu->kvm->slots_lock); vapic_enter(vcpu); -preempted: - if (vcpu->guest_debug.enabled) - kvm_x86_ops->guest_debug_pre(vcpu); - again: if (vcpu->requests) if (test_and_clear_bit(KVM_REQ_MMU_RELOAD, &vcpu->requests)) @@ -3170,6 +3166,9 @@ again: goto out; } + if (vcpu->guest_debug.enabled) + kvm_x86_ops->guest_debug_pre(vcpu); + vcpu->guest_mode = 1; /* * Make sure that guest_mode assignment won't happen after @@ -3244,7 +3243,7 @@ out: if (r > 0) { kvm_resched(vcpu); down_read(&vcpu->kvm->slots_lock); - goto preempted; + goto again; } post_kvm_run_save(vcpu, kvm_run); -- cgit v1.2.2 From 6762b7299aa115e11815decd1fd982d015f09615 Mon Sep 17 00:00:00 2001 From: Amit Shah Date: Wed, 13 Aug 2008 16:22:37 +0300 Subject: KVM: Device assignment: Check for privileges before assigning irq Even though we don't share irqs at the moment, we should ensure regular user processes don't try to allocate system resources. We check for capability to access IO devices (CAP_SYS_RAWIO) before we request_irq on behalf of the guest. Noticed by Avi. Signed-off-by: Amit Shah Signed-off-by: Avi Kivity --- arch/x86/kvm/x86.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'arch/x86/kvm/x86.c') diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c index 4a033757a19e..fffdf4f69c5a 100644 --- a/arch/x86/kvm/x86.c +++ b/arch/x86/kvm/x86.c @@ -191,6 +191,11 @@ static int kvm_vm_ioctl_assign_irq(struct kvm *kvm, kvm_assigned_dev_interrupt_work_handler); if (irqchip_in_kernel(kvm)) { + if (!capable(CAP_SYS_RAWIO)) { + return -EPERM; + goto out; + } + if (assigned_irq->host_irq) match->host_irq = assigned_irq->host_irq; else -- cgit v1.2.2 From 29c8fa32c5d1e2d26d53ad9467b3a13130014cdf Mon Sep 17 00:00:00 2001 From: Amit Shah Date: Mon, 18 Aug 2008 15:07:05 +0300 Subject: KVM: Use kvm_set_irq to inject interrupts ... instead of using the pic and ioapic variants Signed-off-by: Amit Shah Signed-off-by: Avi Kivity --- arch/x86/kvm/x86.c | 8 +------- 1 file changed, 1 insertion(+), 7 deletions(-) (limited to 'arch/x86/kvm/x86.c') diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c index fffdf4f69c5a..5b3c8821b191 100644 --- a/arch/x86/kvm/x86.c +++ b/arch/x86/kvm/x86.c @@ -1956,13 +1956,7 @@ long kvm_arch_vm_ioctl(struct file *filp, goto out; if (irqchip_in_kernel(kvm)) { mutex_lock(&kvm->lock); - if (irq_event.irq < 16) - kvm_pic_set_irq(pic_irqchip(kvm), - irq_event.irq, - irq_event.level); - kvm_ioapic_set_irq(kvm->arch.vioapic, - irq_event.irq, - irq_event.level); + kvm_set_irq(kvm, irq_event.irq, irq_event.level); mutex_unlock(&kvm->lock); r = 0; } -- cgit v1.2.2 From ee032c993edd34e0bdf64dab06a55d0e08a4eeb9 Mon Sep 17 00:00:00 2001 From: Harvey Harrison Date: Mon, 11 Aug 2008 16:54:20 -0700 Subject: KVM: make irq ack notifier functions static sparse says: arch/x86/kvm/x86.c:107:32: warning: symbol 'kvm_find_assigned_dev' was not declared. Should it be static? arch/x86/kvm/i8254.c:225:6: warning: symbol 'kvm_pit_ack_irq' was not declared. Should it be static? Signed-off-by: Harvey Harrison Signed-off-by: Andrew Morton Signed-off-by: Avi Kivity --- arch/x86/kvm/x86.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'arch/x86/kvm/x86.c') diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c index 5b3c8821b191..22edd95712ee 100644 --- a/arch/x86/kvm/x86.c +++ b/arch/x86/kvm/x86.c @@ -104,7 +104,7 @@ struct kvm_stats_debugfs_item debugfs_entries[] = { { NULL } }; -struct kvm_assigned_dev_kernel *kvm_find_assigned_dev(struct list_head *head, +static struct kvm_assigned_dev_kernel *kvm_find_assigned_dev(struct list_head *head, int assigned_dev_id) { struct list_head *ptr; -- cgit v1.2.2 From f4bbd9aaaae23007e4d79536d35a30cbbb11d407 Mon Sep 17 00:00:00 2001 From: Avi Kivity Date: Wed, 20 Aug 2008 15:51:42 +0300 Subject: KVM: Load real mode segments correctly Real mode segments to not reference the GDT or LDT; they simply compute base = selector * 16. Signed-off-by: Avi Kivity --- arch/x86/kvm/x86.c | 22 ++++++++++++++++++++++ 1 file changed, 22 insertions(+) (limited to 'arch/x86/kvm/x86.c') diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c index 22edd95712ee..bfc7c332c5d7 100644 --- a/arch/x86/kvm/x86.c +++ b/arch/x86/kvm/x86.c @@ -3588,11 +3588,33 @@ static int load_segment_descriptor_to_kvm_desct(struct kvm_vcpu *vcpu, return 0; } +int kvm_load_realmode_segment(struct kvm_vcpu *vcpu, u16 selector, int seg) +{ + struct kvm_segment segvar = { + .base = selector << 4, + .limit = 0xffff, + .selector = selector, + .type = 3, + .present = 1, + .dpl = 3, + .db = 0, + .s = 1, + .l = 0, + .g = 0, + .avl = 0, + .unusable = 0, + }; + kvm_x86_ops->set_segment(vcpu, &segvar, seg); + return 0; +} + int kvm_load_segment_descriptor(struct kvm_vcpu *vcpu, u16 selector, int type_bits, int seg) { struct kvm_segment kvm_seg; + if (!(vcpu->arch.cr0 & X86_CR0_PE)) + return kvm_load_realmode_segment(vcpu, selector, seg); if (load_segment_descriptor_to_kvm_desct(vcpu, selector, &kvm_seg)) return 1; kvm_seg.type |= type_bits; -- cgit v1.2.2 From acee3c04e8208c17aad1baff99baa68d71640a19 Mon Sep 17 00:00:00 2001 From: Avi Kivity Date: Tue, 26 Aug 2008 17:22:47 +0300 Subject: KVM: Allocate guest memory as MAP_PRIVATE, not MAP_SHARED There is no reason to share internal memory slots with fork()ed instances. Signed-off-by: Avi Kivity --- arch/x86/kvm/x86.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'arch/x86/kvm/x86.c') diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c index bfc7c332c5d7..675d010995a2 100644 --- a/arch/x86/kvm/x86.c +++ b/arch/x86/kvm/x86.c @@ -4296,7 +4296,7 @@ int kvm_arch_set_memory_region(struct kvm *kvm, userspace_addr = do_mmap(NULL, 0, npages * PAGE_SIZE, PROT_READ | PROT_WRITE, - MAP_SHARED | MAP_ANONYMOUS, + MAP_PRIVATE | MAP_ANONYMOUS, 0); up_write(¤t->mm->mmap_sem); -- cgit v1.2.2 From a89c1ad270ca7ad0eec2667bc754362ce7b142be Mon Sep 17 00:00:00 2001 From: Joerg Roedel Date: Fri, 29 Aug 2008 11:52:07 +0200 Subject: KVM: add MC5_MISC msr read support Currently KVM implements MC0-MC4_MISC read support. When booting Linux this results in KVM warnings in the kernel log when the guest tries to read MC5_MISC. Fix this warnings with this patch. Signed-off-by: Joerg Roedel Signed-off-by: Avi Kivity --- arch/x86/kvm/x86.c | 1 + 1 file changed, 1 insertion(+) (limited to 'arch/x86/kvm/x86.c') diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c index 675d010995a2..e3b89662cf6d 100644 --- a/arch/x86/kvm/x86.c +++ b/arch/x86/kvm/x86.c @@ -991,6 +991,7 @@ int kvm_get_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 *pdata) case MSR_IA32_MC0_MISC+8: case MSR_IA32_MC0_MISC+12: case MSR_IA32_MC0_MISC+16: + case MSR_IA32_MC0_MISC+20: case MSR_IA32_UCODE_REV: case MSR_IA32_EBL_CR_POWERON: case MSR_IA32_DEBUGCTLMSR: -- cgit v1.2.2 From fa89a81766e33343fa8f0fe0e371819bf94a17a1 Mon Sep 17 00:00:00 2001 From: Avi Kivity Date: Mon, 1 Sep 2008 15:57:51 +0300 Subject: KVM: Add statistics for guest irq injections These can help show whether a guest is making progress or not. Signed-off-by: Avi Kivity --- arch/x86/kvm/x86.c | 1 + 1 file changed, 1 insertion(+) (limited to 'arch/x86/kvm/x86.c') diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c index e3b89662cf6d..3f3cb7107c03 100644 --- a/arch/x86/kvm/x86.c +++ b/arch/x86/kvm/x86.c @@ -92,6 +92,7 @@ struct kvm_stats_debugfs_item debugfs_entries[] = { { "fpu_reload", VCPU_STAT(fpu_reload) }, { "insn_emulation", VCPU_STAT(insn_emulation) }, { "insn_emulation_fail", VCPU_STAT(insn_emulation_fail) }, + { "irq_injections", VCPU_STAT(irq_injections) }, { "mmu_shadow_zapped", VM_STAT(mmu_shadow_zapped) }, { "mmu_pte_write", VM_STAT(mmu_pte_write) }, { "mmu_pte_updated", VM_STAT(mmu_pte_updated) }, -- cgit v1.2.2 From d76901750ab9f71091d33ef3d2b5909d8a9a4ad4 Mon Sep 17 00:00:00 2001 From: Marcelo Tosatti Date: Mon, 8 Sep 2008 15:23:48 -0300 Subject: KVM: x86: do not execute halted vcpus Offline or uninitialized vcpu's can be executed if requested to perform userspace work. Follow Avi's suggestion to handle halted vcpu's in the main loop, simplifying kvm_emulate_halt(). Introduce a new vcpu->requests bit to indicate events that promote state from halted to running. Also standardize vcpu wake sites. Signed-off-by: Marcelo Tosatti redhat.com> Signed-off-by: Avi Kivity --- arch/x86/kvm/x86.c | 100 +++++++++++++++++++++++++++++------------------------ 1 file changed, 55 insertions(+), 45 deletions(-) (limited to 'arch/x86/kvm/x86.c') diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c index 3f3cb7107c03..bf98d40b21ec 100644 --- a/arch/x86/kvm/x86.c +++ b/arch/x86/kvm/x86.c @@ -2798,11 +2798,6 @@ int kvm_emulate_halt(struct kvm_vcpu *vcpu) KVMTRACE_0D(HLT, vcpu, handler); if (irqchip_in_kernel(vcpu->kvm)) { vcpu->arch.mp_state = KVM_MP_STATE_HALTED; - up_read(&vcpu->kvm->slots_lock); - kvm_vcpu_block(vcpu); - down_read(&vcpu->kvm->slots_lock); - if (vcpu->arch.mp_state != KVM_MP_STATE_RUNNABLE) - return -EINTR; return 1; } else { vcpu->run->exit_reason = KVM_EXIT_HLT; @@ -3097,24 +3092,10 @@ static void vapic_exit(struct kvm_vcpu *vcpu) up_read(&vcpu->kvm->slots_lock); } -static int __vcpu_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run) +static int vcpu_enter_guest(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run) { int r; - if (unlikely(vcpu->arch.mp_state == KVM_MP_STATE_SIPI_RECEIVED)) { - pr_debug("vcpu %d received sipi with vector # %x\n", - vcpu->vcpu_id, vcpu->arch.sipi_vector); - kvm_lapic_reset(vcpu); - r = kvm_x86_ops->vcpu_reset(vcpu); - if (r) - return r; - vcpu->arch.mp_state = KVM_MP_STATE_RUNNABLE; - } - - down_read(&vcpu->kvm->slots_lock); - vapic_enter(vcpu); - -again: if (vcpu->requests) if (test_and_clear_bit(KVM_REQ_MMU_RELOAD, &vcpu->requests)) kvm_mmu_unload(vcpu); @@ -3151,22 +3132,13 @@ again: local_irq_disable(); - if (vcpu->requests || need_resched()) { + if (vcpu->requests || need_resched() || signal_pending(current)) { local_irq_enable(); preempt_enable(); r = 1; goto out; } - if (signal_pending(current)) { - local_irq_enable(); - preempt_enable(); - r = -EINTR; - kvm_run->exit_reason = KVM_EXIT_INTR; - ++vcpu->stat.signal_exits; - goto out; - } - if (vcpu->guest_debug.enabled) kvm_x86_ops->guest_debug_pre(vcpu); @@ -3227,26 +3199,63 @@ again: kvm_lapic_sync_from_vapic(vcpu); r = kvm_x86_ops->handle_exit(kvm_run, vcpu); +out: + return r; +} - if (r > 0) { - if (dm_request_for_irq_injection(vcpu, kvm_run)) { - r = -EINTR; - kvm_run->exit_reason = KVM_EXIT_INTR; - ++vcpu->stat.request_irq_exits; - goto out; - } - if (!need_resched()) - goto again; +static int __vcpu_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run) +{ + int r; + + if (unlikely(vcpu->arch.mp_state == KVM_MP_STATE_SIPI_RECEIVED)) { + printk("vcpu %d received sipi with vector # %x\n", + vcpu->vcpu_id, vcpu->arch.sipi_vector); + kvm_lapic_reset(vcpu); + r = kvm_x86_ops->vcpu_reset(vcpu); + if (r) + return r; + vcpu->arch.mp_state = KVM_MP_STATE_RUNNABLE; } -out: - up_read(&vcpu->kvm->slots_lock); - if (r > 0) { - kvm_resched(vcpu); - down_read(&vcpu->kvm->slots_lock); - goto again; + down_read(&vcpu->kvm->slots_lock); + vapic_enter(vcpu); + + r = 1; + while (r > 0) { + if (kvm_arch_vcpu_runnable(vcpu)) + r = vcpu_enter_guest(vcpu, kvm_run); + else { + up_read(&vcpu->kvm->slots_lock); + kvm_vcpu_block(vcpu); + down_read(&vcpu->kvm->slots_lock); + if (test_and_clear_bit(KVM_REQ_UNHALT, &vcpu->requests)) + if (vcpu->arch.mp_state == KVM_MP_STATE_HALTED) + vcpu->arch.mp_state = + KVM_MP_STATE_RUNNABLE; + if (vcpu->arch.mp_state != KVM_MP_STATE_RUNNABLE) + r = -EINTR; + } + + if (r > 0) { + if (dm_request_for_irq_injection(vcpu, kvm_run)) { + r = -EINTR; + kvm_run->exit_reason = KVM_EXIT_INTR; + ++vcpu->stat.request_irq_exits; + } + if (signal_pending(current)) { + r = -EINTR; + kvm_run->exit_reason = KVM_EXIT_INTR; + ++vcpu->stat.signal_exits; + } + if (need_resched()) { + up_read(&vcpu->kvm->slots_lock); + kvm_resched(vcpu); + down_read(&vcpu->kvm->slots_lock); + } + } } + up_read(&vcpu->kvm->slots_lock); post_kvm_run_save(vcpu, kvm_run); vapic_exit(vcpu); @@ -3266,6 +3275,7 @@ int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run) if (unlikely(vcpu->arch.mp_state == KVM_MP_STATE_UNINITIALIZED)) { kvm_vcpu_block(vcpu); + clear_bit(KVM_REQ_UNHALT, &vcpu->requests); r = -EAGAIN; goto out; } -- cgit v1.2.2 From 9c3e4aab5ae27bf8589d61c7874e213832b4b7d2 Mon Sep 17 00:00:00 2001 From: Marcelo Tosatti Date: Wed, 10 Sep 2008 16:40:55 -0300 Subject: KVM: x86: unhalt vcpu0 on reset Since "KVM: x86: do not execute halted vcpus", HLT by vcpu0 before system reset by the IO thread will hang the guest. Mark vcpu as runnable in such case. Signed-off-by: Marcelo Tosatti Signed-off-by: Avi Kivity --- arch/x86/kvm/x86.c | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'arch/x86/kvm/x86.c') diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c index bf98d40b21ec..2134f3e0a516 100644 --- a/arch/x86/kvm/x86.c +++ b/arch/x86/kvm/x86.c @@ -3959,6 +3959,12 @@ int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu, kvm_set_segment(vcpu, &sregs->tr, VCPU_SREG_TR); kvm_set_segment(vcpu, &sregs->ldt, VCPU_SREG_LDTR); + /* Older userspace won't unhalt the vcpu on reset. */ + if (vcpu->vcpu_id == 0 && kvm_rip_read(vcpu) == 0xfff0 && + sregs->cs.selector == 0xf000 && sregs->cs.base == 0xffff0000 && + !(vcpu->arch.cr0 & X86_CR0_PE)) + vcpu->arch.mp_state = KVM_MP_STATE_RUNNABLE; + vcpu_put(vcpu); return 0; -- cgit v1.2.2 From 62c476c7c7f25a5b245b9902a935636e6316e58c Mon Sep 17 00:00:00 2001 From: Ben-Ami Yassour Date: Sun, 14 Sep 2008 03:48:28 +0300 Subject: KVM: Device Assignment with VT-d Based on a patch by: Kay, Allen M This patch enables PCI device assignment based on VT-d support. When a device is assigned to the guest, the guest memory is pinned and the mapping is updated in the VT-d IOMMU. [Amit: Expose KVM_CAP_IOMMU so we can check if an IOMMU is present and also control enable/disable from userspace] Signed-off-by: Kay, Allen M Signed-off-by: Weidong Han Signed-off-by: Ben-Ami Yassour Signed-off-by: Amit Shah Acked-by: Mark Gross Signed-off-by: Avi Kivity --- arch/x86/kvm/x86.c | 14 ++++++++++++++ 1 file changed, 14 insertions(+) (limited to 'arch/x86/kvm/x86.c') diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c index 2134f3e0a516..c8a2793626ec 100644 --- a/arch/x86/kvm/x86.c +++ b/arch/x86/kvm/x86.c @@ -35,6 +35,7 @@ #include #include #include +#include #include #include @@ -277,9 +278,18 @@ static int kvm_vm_ioctl_assign_device(struct kvm *kvm, list_add(&match->list, &kvm->arch.assigned_dev_head); + if (assigned_dev->flags & KVM_DEV_ASSIGN_ENABLE_IOMMU) { + r = kvm_iommu_map_guest(kvm, match); + if (r) + goto out_list_del; + } + out: mutex_unlock(&kvm->lock); return r; +out_list_del: + list_del(&match->list); + pci_release_regions(dev); out_disable: pci_disable_device(dev); out_put: @@ -1147,6 +1157,9 @@ int kvm_dev_ioctl_check_extension(long ext) case KVM_CAP_PV_MMU: r = !tdp_enabled; break; + case KVM_CAP_IOMMU: + r = intel_iommu_found(); + break; default: r = 0; break; @@ -4282,6 +4295,7 @@ static void kvm_free_vcpus(struct kvm *kvm) void kvm_arch_destroy_vm(struct kvm *kvm) { + kvm_iommu_unmap_guest(kvm); kvm_free_assigned_devices(kvm); kvm_free_pit(kvm); kfree(kvm->arch.vpic); -- cgit v1.2.2 From bfadaded0dc323a1cf3f08b5068f12955b54cbaa Mon Sep 17 00:00:00 2001 From: Amit Shah Date: Tue, 16 Sep 2008 18:04:28 +0300 Subject: KVM: Device Assignment: Free device structures if IRQ allocation fails When an IRQ allocation fails, we free up the device structures and disable the device so that we can unregister the device in the userspace and not expose it to the guest at all. Signed-off-by: Amit Shah Signed-off-by: Avi Kivity --- arch/x86/kvm/x86.c | 86 ++++++++++++++++++++++++++++-------------------------- 1 file changed, 45 insertions(+), 41 deletions(-) (limited to 'arch/x86/kvm/x86.c') diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c index c8a2793626ec..61eddbeabeb4 100644 --- a/arch/x86/kvm/x86.c +++ b/arch/x86/kvm/x86.c @@ -166,6 +166,43 @@ static void kvm_assigned_dev_ack_irq(struct kvm_irq_ack_notifier *kian) enable_irq(dev->host_irq); } +static void kvm_free_assigned_device(struct kvm *kvm, + struct kvm_assigned_dev_kernel + *assigned_dev) +{ + if (irqchip_in_kernel(kvm) && assigned_dev->irq_requested) + free_irq(assigned_dev->host_irq, (void *)assigned_dev); + + kvm_unregister_irq_ack_notifier(kvm, &assigned_dev->ack_notifier); + + if (cancel_work_sync(&assigned_dev->interrupt_work)) + /* We had pending work. That means we will have to take + * care of kvm_put_kvm. + */ + kvm_put_kvm(kvm); + + pci_release_regions(assigned_dev->dev); + pci_disable_device(assigned_dev->dev); + pci_dev_put(assigned_dev->dev); + + list_del(&assigned_dev->list); + kfree(assigned_dev); +} + +static void kvm_free_all_assigned_devices(struct kvm *kvm) +{ + struct list_head *ptr, *ptr2; + struct kvm_assigned_dev_kernel *assigned_dev; + + list_for_each_safe(ptr, ptr2, &kvm->arch.assigned_dev_head) { + assigned_dev = list_entry(ptr, + struct kvm_assigned_dev_kernel, + list); + + kvm_free_assigned_device(kvm, assigned_dev); + } +} + static int kvm_vm_ioctl_assign_irq(struct kvm *kvm, struct kvm_assigned_irq *assigned_irq) @@ -194,8 +231,8 @@ static int kvm_vm_ioctl_assign_irq(struct kvm *kvm, if (irqchip_in_kernel(kvm)) { if (!capable(CAP_SYS_RAWIO)) { - return -EPERM; - goto out; + r = -EPERM; + goto out_release; } if (assigned_irq->host_irq) @@ -214,17 +251,18 @@ static int kvm_vm_ioctl_assign_irq(struct kvm *kvm, */ if (request_irq(match->host_irq, kvm_assigned_dev_intr, 0, "kvm_assigned_device", (void *)match)) { - printk(KERN_INFO "%s: couldn't allocate irq for pv " - "device\n", __func__); r = -EIO; - goto out; + goto out_release; } } match->irq_requested = true; -out: mutex_unlock(&kvm->lock); return r; +out_release: + mutex_unlock(&kvm->lock); + kvm_free_assigned_device(kvm, match); + return r; } static int kvm_vm_ioctl_assign_device(struct kvm *kvm, @@ -300,40 +338,6 @@ out_free: return r; } -static void kvm_free_assigned_devices(struct kvm *kvm) -{ - struct list_head *ptr, *ptr2; - struct kvm_assigned_dev_kernel *assigned_dev; - - list_for_each_safe(ptr, ptr2, &kvm->arch.assigned_dev_head) { - assigned_dev = list_entry(ptr, - struct kvm_assigned_dev_kernel, - list); - - if (irqchip_in_kernel(kvm) && assigned_dev->irq_requested) { - free_irq(assigned_dev->host_irq, - (void *)assigned_dev); - - kvm_unregister_irq_ack_notifier(kvm, - &assigned_dev-> - ack_notifier); - } - - if (cancel_work_sync(&assigned_dev->interrupt_work)) - /* We had pending work. That means we will have to take - * care of kvm_put_kvm. - */ - kvm_put_kvm(kvm); - - pci_release_regions(assigned_dev->dev); - pci_disable_device(assigned_dev->dev); - pci_dev_put(assigned_dev->dev); - - list_del(&assigned_dev->list); - kfree(assigned_dev); - } -} - unsigned long segment_base(u16 selector) { struct descriptor_table gdt; @@ -4296,7 +4300,7 @@ static void kvm_free_vcpus(struct kvm *kvm) void kvm_arch_destroy_vm(struct kvm *kvm) { kvm_iommu_unmap_guest(kvm); - kvm_free_assigned_devices(kvm); + kvm_free_all_assigned_devices(kvm); kvm_free_pit(kvm); kfree(kvm->arch.vpic); kfree(kvm->arch.vioapic); -- cgit v1.2.2 From 4c2155ce81c193788082d4b8cdbc26d79edebc58 Mon Sep 17 00:00:00 2001 From: Marcelo Tosatti Date: Tue, 16 Sep 2008 20:54:47 -0300 Subject: KVM: switch to get_user_pages_fast Convert gfn_to_pfn to use get_user_pages_fast, which can do lockless pagetable lookups on x86. Kernel compilation on 4-way guest is 3.7% faster on VMX. Signed-off-by: Marcelo Tosatti Signed-off-by: Avi Kivity --- arch/x86/kvm/x86.c | 6 ------ 1 file changed, 6 deletions(-) (limited to 'arch/x86/kvm/x86.c') diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c index 61eddbeabeb4..108f07267e87 100644 --- a/arch/x86/kvm/x86.c +++ b/arch/x86/kvm/x86.c @@ -946,10 +946,8 @@ int kvm_set_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 data) /* ...but clean it before doing the actual write */ vcpu->arch.time_offset = data & ~(PAGE_MASK | 1); - down_read(¤t->mm->mmap_sem); vcpu->arch.time_page = gfn_to_page(vcpu->kvm, data >> PAGE_SHIFT); - up_read(¤t->mm->mmap_sem); if (is_error_page(vcpu->arch.time_page)) { kvm_release_page_clean(vcpu->arch.time_page); @@ -2322,9 +2320,7 @@ static int emulator_cmpxchg_emulated(unsigned long addr, val = *(u64 *)new; - down_read(¤t->mm->mmap_sem); page = gfn_to_page(vcpu->kvm, gpa >> PAGE_SHIFT); - up_read(¤t->mm->mmap_sem); kaddr = kmap_atomic(page, KM_USER0); set_64bit((u64 *)(kaddr + offset_in_page(gpa)), val); @@ -3089,9 +3085,7 @@ static void vapic_enter(struct kvm_vcpu *vcpu) if (!apic || !apic->vapic_addr) return; - down_read(¤t->mm->mmap_sem); page = gfn_to_page(vcpu->kvm, apic->vapic_addr >> PAGE_SHIFT); - up_read(¤t->mm->mmap_sem); vcpu->arch.apic->vapic_page = page; } -- cgit v1.2.2 From 2259e3a7a6089007839cd4bbf7c9867190c67238 Mon Sep 17 00:00:00 2001 From: Harvey Harrison Date: Fri, 22 Aug 2008 13:29:17 -0700 Subject: KVM: x86.c make kvm_load_realmode_segment static Noticed by sparse: arch/x86/kvm/x86.c:3591:5: warning: symbol 'kvm_load_realmode_segment' was not declared. Should it be static? Signed-off-by: Harvey Harrison Signed-off-by: Avi Kivity --- arch/x86/kvm/x86.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'arch/x86/kvm/x86.c') diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c index 108f07267e87..1b738cb02831 100644 --- a/arch/x86/kvm/x86.c +++ b/arch/x86/kvm/x86.c @@ -3611,7 +3611,7 @@ static int load_segment_descriptor_to_kvm_desct(struct kvm_vcpu *vcpu, return 0; } -int kvm_load_realmode_segment(struct kvm_vcpu *vcpu, u16 selector, int seg) +static int kvm_load_realmode_segment(struct kvm_vcpu *vcpu, u16 selector, int seg) { struct kvm_segment segvar = { .base = selector << 4, -- cgit v1.2.2 From af2152f5457448bd90cb019c108e0a85e716fdbe Mon Sep 17 00:00:00 2001 From: Gleb Natapov Date: Mon, 22 Sep 2008 14:28:53 +0300 Subject: KVM: don't enter guest after SIPI was received by a CPU The vcpu should process pending SIPI message before entering guest mode again. kvm_arch_vcpu_runnable() returns true if the vcpu is in SIPI state, so we can't call it here. Signed-off-by: Gleb Natapov Signed-off-by: Avi Kivity --- arch/x86/kvm/x86.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'arch/x86/kvm/x86.c') diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c index 1b738cb02831..08edeabf15e6 100644 --- a/arch/x86/kvm/x86.c +++ b/arch/x86/kvm/x86.c @@ -3233,7 +3233,7 @@ static int __vcpu_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run) r = 1; while (r > 0) { - if (kvm_arch_vcpu_runnable(vcpu)) + if (vcpu->arch.mp_state == KVM_MP_STATE_RUNNABLE) r = vcpu_enter_guest(vcpu, kvm_run); else { up_read(&vcpu->kvm->slots_lock); -- cgit v1.2.2 From 0ba73cdadb8ac172f396df7e23c4a9cebd59b550 Mon Sep 17 00:00:00 2001 From: Marcelo Tosatti Date: Tue, 23 Sep 2008 13:18:34 -0300 Subject: KVM: MMU: sync roots on mmu reload Signed-off-by: Marcelo Tosatti Signed-off-by: Avi Kivity --- arch/x86/kvm/x86.c | 1 + 1 file changed, 1 insertion(+) (limited to 'arch/x86/kvm/x86.c') diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c index 08edeabf15e6..88e6d9abbd2b 100644 --- a/arch/x86/kvm/x86.c +++ b/arch/x86/kvm/x86.c @@ -594,6 +594,7 @@ EXPORT_SYMBOL_GPL(kvm_set_cr4); void kvm_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3) { if (cr3 == vcpu->arch.cr3 && !pdptrs_changed(vcpu)) { + kvm_mmu_sync_roots(vcpu); kvm_mmu_flush_tlb(vcpu); return; } -- cgit v1.2.2 From a7052897b3bcd568a9f5bfaa558957039e7e7ec0 Mon Sep 17 00:00:00 2001 From: Marcelo Tosatti Date: Tue, 23 Sep 2008 13:18:35 -0300 Subject: KVM: x86: trap invlpg With pages out of sync invlpg needs to be trapped. For now simply nuke the entry. Untested on AMD. Signed-off-by: Marcelo Tosatti Signed-off-by: Avi Kivity --- arch/x86/kvm/x86.c | 1 + 1 file changed, 1 insertion(+) (limited to 'arch/x86/kvm/x86.c') diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c index 88e6d9abbd2b..efee85ba07e5 100644 --- a/arch/x86/kvm/x86.c +++ b/arch/x86/kvm/x86.c @@ -2341,6 +2341,7 @@ static unsigned long get_segment_base(struct kvm_vcpu *vcpu, int seg) int emulate_invlpg(struct kvm_vcpu *vcpu, gva_t address) { + kvm_mmu_invlpg(vcpu, address); return X86EMUL_CONTINUE; } -- cgit v1.2.2 From 4731d4c7a07769cf2926c327177b97bb8c68cafc Mon Sep 17 00:00:00 2001 From: Marcelo Tosatti Date: Tue, 23 Sep 2008 13:18:39 -0300 Subject: KVM: MMU: out of sync shadow core Allow guest pagetables to go out of sync. Instead of emulating write accesses to guest pagetables, or unshadowing them, we un-write-protect the page table and allow the guest to modify it at will. We rely on invlpg executions to synchronize individual ptes, and will synchronize the entire pagetable on tlb flushes. Signed-off-by: Marcelo Tosatti Signed-off-by: Avi Kivity --- arch/x86/kvm/x86.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'arch/x86/kvm/x86.c') diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c index efee85ba07e5..1c5864ac0837 100644 --- a/arch/x86/kvm/x86.c +++ b/arch/x86/kvm/x86.c @@ -101,6 +101,7 @@ struct kvm_stats_debugfs_item debugfs_entries[] = { { "mmu_flooded", VM_STAT(mmu_flooded) }, { "mmu_recycled", VM_STAT(mmu_recycled) }, { "mmu_cache_miss", VM_STAT(mmu_cache_miss) }, + { "mmu_unsync", VM_STAT(mmu_unsync) }, { "remote_tlb_flush", VM_STAT(remote_tlb_flush) }, { "largepages", VM_STAT(lpages) }, { NULL } @@ -3120,6 +3121,8 @@ static int vcpu_enter_guest(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run) if (vcpu->requests) { if (test_and_clear_bit(KVM_REQ_MIGRATE_TIMER, &vcpu->requests)) __kvm_migrate_timers(vcpu); + if (test_and_clear_bit(KVM_REQ_MMU_SYNC, &vcpu->requests)) + kvm_mmu_sync_roots(vcpu); if (test_and_clear_bit(KVM_REQ_TLB_FLUSH, &vcpu->requests)) kvm_x86_ops->tlb_flush(vcpu); if (test_and_clear_bit(KVM_REQ_REPORT_TPR_ACCESS, -- cgit v1.2.2 From e48258009d941891fca35348986b8d280caf31cd Mon Sep 17 00:00:00 2001 From: Marcelo Tosatti Date: Wed, 24 Sep 2008 20:28:34 -0300 Subject: KVM: PIC: enhance IPI avoidance The PIC code makes little effort to avoid kvm_vcpu_kick(), resulting in unnecessary guest exits in some conditions. For example, if the timer interrupt is routed through the IOAPIC, IRR for IRQ 0 will get set but not cleared, since the APIC is handling the acks. This means that everytime an interrupt < 16 is triggered, the priority logic will find IRQ0 pending and send an IPI to vcpu0 (in case IRQ0 is not masked, which is Linux's case). Introduce a new variable isr_ack to represent the IRQ's for which the guest has been signalled / cleared the ISR. Use it to avoid more than one IPI per trigger-ack cycle, in addition to the avoidance when ISR is set in get_priority(). Signed-off-by: Marcelo Tosatti Signed-off-by: Avi Kivity --- arch/x86/kvm/x86.c | 1 + 1 file changed, 1 insertion(+) (limited to 'arch/x86/kvm/x86.c') diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c index 1c5864ac0837..4cfdd1b6df0b 100644 --- a/arch/x86/kvm/x86.c +++ b/arch/x86/kvm/x86.c @@ -3963,6 +3963,7 @@ int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu, pr_debug("Set back pending irq %d\n", pending_vec); } + kvm_pic_clear_isr_ack(vcpu->kvm); } kvm_set_segment(vcpu, &sregs->cs, VCPU_SREG_CS); -- cgit v1.2.2 From 1b10bf31a5de5b76e2e9c2937878a45c5ae2be37 Mon Sep 17 00:00:00 2001 From: Jan Kiszka Date: Tue, 30 Sep 2008 10:41:06 +0200 Subject: KVM: x86: Silence various LAPIC-related host kernel messages KVM-x86 dumps a lot of debug messages that have no meaning for normal operation: - INIT de-assertion is ignored - SIPIs are sent and received - APIC writes are unaligned or < 4 byte long (Windows Server 2003 triggers this on SMP) Degrade them to true debug messages, keeping the host kernel log clean for real problems. Signed-off-by: Jan Kiszka Signed-off-by: Avi Kivity --- arch/x86/kvm/x86.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'arch/x86/kvm/x86.c') diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c index 4cfdd1b6df0b..d6d7123d2644 100644 --- a/arch/x86/kvm/x86.c +++ b/arch/x86/kvm/x86.c @@ -3224,8 +3224,8 @@ static int __vcpu_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run) int r; if (unlikely(vcpu->arch.mp_state == KVM_MP_STATE_SIPI_RECEIVED)) { - printk("vcpu %d received sipi with vector # %x\n", - vcpu->vcpu_id, vcpu->arch.sipi_vector); + pr_debug("vcpu %d received sipi with vector # %x\n", + vcpu->vcpu_id, vcpu->arch.sipi_vector); kvm_lapic_reset(vcpu); r = kvm_x86_ops->vcpu_reset(vcpu); if (r) -- cgit v1.2.2 From 8a98f6648a2b0756d8f26d6c13332f5526355fec Mon Sep 17 00:00:00 2001 From: Xiantao Zhang Date: Mon, 6 Oct 2008 13:47:38 +0800 Subject: KVM: Move device assignment logic to common code To share with other archs, this patch moves device assignment logic to common parts. Signed-off-by: Xiantao Zhang Signed-off-by: Avi Kivity --- arch/x86/kvm/x86.c | 255 ----------------------------------------------------- 1 file changed, 255 deletions(-) (limited to 'arch/x86/kvm/x86.c') diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c index d6d7123d2644..f8bde01ba8e6 100644 --- a/arch/x86/kvm/x86.c +++ b/arch/x86/kvm/x86.c @@ -30,7 +30,6 @@ #include #include #include -#include #include #include #include @@ -107,238 +106,6 @@ struct kvm_stats_debugfs_item debugfs_entries[] = { { NULL } }; -static struct kvm_assigned_dev_kernel *kvm_find_assigned_dev(struct list_head *head, - int assigned_dev_id) -{ - struct list_head *ptr; - struct kvm_assigned_dev_kernel *match; - - list_for_each(ptr, head) { - match = list_entry(ptr, struct kvm_assigned_dev_kernel, list); - if (match->assigned_dev_id == assigned_dev_id) - return match; - } - return NULL; -} - -static void kvm_assigned_dev_interrupt_work_handler(struct work_struct *work) -{ - struct kvm_assigned_dev_kernel *assigned_dev; - - assigned_dev = container_of(work, struct kvm_assigned_dev_kernel, - interrupt_work); - - /* This is taken to safely inject irq inside the guest. When - * the interrupt injection (or the ioapic code) uses a - * finer-grained lock, update this - */ - mutex_lock(&assigned_dev->kvm->lock); - kvm_set_irq(assigned_dev->kvm, - assigned_dev->guest_irq, 1); - mutex_unlock(&assigned_dev->kvm->lock); - kvm_put_kvm(assigned_dev->kvm); -} - -/* FIXME: Implement the OR logic needed to make shared interrupts on - * this line behave properly - */ -static irqreturn_t kvm_assigned_dev_intr(int irq, void *dev_id) -{ - struct kvm_assigned_dev_kernel *assigned_dev = - (struct kvm_assigned_dev_kernel *) dev_id; - - kvm_get_kvm(assigned_dev->kvm); - schedule_work(&assigned_dev->interrupt_work); - disable_irq_nosync(irq); - return IRQ_HANDLED; -} - -/* Ack the irq line for an assigned device */ -static void kvm_assigned_dev_ack_irq(struct kvm_irq_ack_notifier *kian) -{ - struct kvm_assigned_dev_kernel *dev; - - if (kian->gsi == -1) - return; - - dev = container_of(kian, struct kvm_assigned_dev_kernel, - ack_notifier); - kvm_set_irq(dev->kvm, dev->guest_irq, 0); - enable_irq(dev->host_irq); -} - -static void kvm_free_assigned_device(struct kvm *kvm, - struct kvm_assigned_dev_kernel - *assigned_dev) -{ - if (irqchip_in_kernel(kvm) && assigned_dev->irq_requested) - free_irq(assigned_dev->host_irq, (void *)assigned_dev); - - kvm_unregister_irq_ack_notifier(kvm, &assigned_dev->ack_notifier); - - if (cancel_work_sync(&assigned_dev->interrupt_work)) - /* We had pending work. That means we will have to take - * care of kvm_put_kvm. - */ - kvm_put_kvm(kvm); - - pci_release_regions(assigned_dev->dev); - pci_disable_device(assigned_dev->dev); - pci_dev_put(assigned_dev->dev); - - list_del(&assigned_dev->list); - kfree(assigned_dev); -} - -static void kvm_free_all_assigned_devices(struct kvm *kvm) -{ - struct list_head *ptr, *ptr2; - struct kvm_assigned_dev_kernel *assigned_dev; - - list_for_each_safe(ptr, ptr2, &kvm->arch.assigned_dev_head) { - assigned_dev = list_entry(ptr, - struct kvm_assigned_dev_kernel, - list); - - kvm_free_assigned_device(kvm, assigned_dev); - } -} - -static int kvm_vm_ioctl_assign_irq(struct kvm *kvm, - struct kvm_assigned_irq - *assigned_irq) -{ - int r = 0; - struct kvm_assigned_dev_kernel *match; - - mutex_lock(&kvm->lock); - - match = kvm_find_assigned_dev(&kvm->arch.assigned_dev_head, - assigned_irq->assigned_dev_id); - if (!match) { - mutex_unlock(&kvm->lock); - return -EINVAL; - } - - if (match->irq_requested) { - match->guest_irq = assigned_irq->guest_irq; - match->ack_notifier.gsi = assigned_irq->guest_irq; - mutex_unlock(&kvm->lock); - return 0; - } - - INIT_WORK(&match->interrupt_work, - kvm_assigned_dev_interrupt_work_handler); - - if (irqchip_in_kernel(kvm)) { - if (!capable(CAP_SYS_RAWIO)) { - r = -EPERM; - goto out_release; - } - - if (assigned_irq->host_irq) - match->host_irq = assigned_irq->host_irq; - else - match->host_irq = match->dev->irq; - match->guest_irq = assigned_irq->guest_irq; - match->ack_notifier.gsi = assigned_irq->guest_irq; - match->ack_notifier.irq_acked = kvm_assigned_dev_ack_irq; - kvm_register_irq_ack_notifier(kvm, &match->ack_notifier); - - /* Even though this is PCI, we don't want to use shared - * interrupts. Sharing host devices with guest-assigned devices - * on the same interrupt line is not a happy situation: there - * are going to be long delays in accepting, acking, etc. - */ - if (request_irq(match->host_irq, kvm_assigned_dev_intr, 0, - "kvm_assigned_device", (void *)match)) { - r = -EIO; - goto out_release; - } - } - - match->irq_requested = true; - mutex_unlock(&kvm->lock); - return r; -out_release: - mutex_unlock(&kvm->lock); - kvm_free_assigned_device(kvm, match); - return r; -} - -static int kvm_vm_ioctl_assign_device(struct kvm *kvm, - struct kvm_assigned_pci_dev *assigned_dev) -{ - int r = 0; - struct kvm_assigned_dev_kernel *match; - struct pci_dev *dev; - - mutex_lock(&kvm->lock); - - match = kvm_find_assigned_dev(&kvm->arch.assigned_dev_head, - assigned_dev->assigned_dev_id); - if (match) { - /* device already assigned */ - r = -EINVAL; - goto out; - } - - match = kzalloc(sizeof(struct kvm_assigned_dev_kernel), GFP_KERNEL); - if (match == NULL) { - printk(KERN_INFO "%s: Couldn't allocate memory\n", - __func__); - r = -ENOMEM; - goto out; - } - dev = pci_get_bus_and_slot(assigned_dev->busnr, - assigned_dev->devfn); - if (!dev) { - printk(KERN_INFO "%s: host device not found\n", __func__); - r = -EINVAL; - goto out_free; - } - if (pci_enable_device(dev)) { - printk(KERN_INFO "%s: Could not enable PCI device\n", __func__); - r = -EBUSY; - goto out_put; - } - r = pci_request_regions(dev, "kvm_assigned_device"); - if (r) { - printk(KERN_INFO "%s: Could not get access to device regions\n", - __func__); - goto out_disable; - } - match->assigned_dev_id = assigned_dev->assigned_dev_id; - match->host_busnr = assigned_dev->busnr; - match->host_devfn = assigned_dev->devfn; - match->dev = dev; - - match->kvm = kvm; - - list_add(&match->list, &kvm->arch.assigned_dev_head); - - if (assigned_dev->flags & KVM_DEV_ASSIGN_ENABLE_IOMMU) { - r = kvm_iommu_map_guest(kvm, match); - if (r) - goto out_list_del; - } - -out: - mutex_unlock(&kvm->lock); - return r; -out_list_del: - list_del(&match->list); - pci_release_regions(dev); -out_disable: - pci_disable_device(dev); -out_put: - pci_dev_put(dev); -out_free: - kfree(match); - mutex_unlock(&kvm->lock); - return r; -} - unsigned long segment_base(u16 selector) { struct descriptor_table gdt; @@ -2030,28 +1797,6 @@ long kvm_arch_vm_ioctl(struct file *filp, goto out; break; } - case KVM_ASSIGN_PCI_DEVICE: { - struct kvm_assigned_pci_dev assigned_dev; - - r = -EFAULT; - if (copy_from_user(&assigned_dev, argp, sizeof assigned_dev)) - goto out; - r = kvm_vm_ioctl_assign_device(kvm, &assigned_dev); - if (r) - goto out; - break; - } - case KVM_ASSIGN_IRQ: { - struct kvm_assigned_irq assigned_irq; - - r = -EFAULT; - if (copy_from_user(&assigned_irq, argp, sizeof assigned_irq)) - goto out; - r = kvm_vm_ioctl_assign_irq(kvm, &assigned_irq); - if (r) - goto out; - break; - } case KVM_GET_PIT: { r = -EFAULT; if (copy_from_user(&u.ps, argp, sizeof(struct kvm_pit_state))) -- cgit v1.2.2