[PATCH 052/104] KVM: Cleanup string I/O instruction emulation
From: Avi Kivity
Date: Mon Sep 17 2007 - 05:00:12 EST
From: Laurent Vivier <Laurent.Vivier@xxxxxxxx>
Both vmx and svm decode the I/O instructions, and both botch the job,
requiring the instruction prefixes to be fetched in order to completely
decode the instruction.
So, if we see a string I/O instruction, use the x86 emulator to decode it,
as it already has all the prefix decoding machinery.
This patch defines ins/outs opcodes in x86_emulate.c and calls
emulate_instruction() from io_interception() (svm.c) and from handle_io()
(vmx.c). It removes all vmx/svm prefix instruction decoders
(get_addr_size(), io_get_override(), io_address(), get_io_count())
Signed-off-by: Laurent Vivier <Laurent.Vivier@xxxxxxxx>
Signed-off-by: Avi Kivity <avi@xxxxxxxxxxxx>
---
drivers/kvm/kvm_main.c | 3 +
drivers/kvm/svm.c | 149 +++------------------------------------------
drivers/kvm/vmx.c | 76 ++++-------------------
drivers/kvm/x86_emulate.c | 49 +++++++++++++--
4 files changed, 69 insertions(+), 208 deletions(-)
diff --git a/drivers/kvm/kvm_main.c b/drivers/kvm/kvm_main.c
index 62adaee..661d065 100644
--- a/drivers/kvm/kvm_main.c
+++ b/drivers/kvm/kvm_main.c
@@ -1221,7 +1221,10 @@ int emulate_instruction(struct kvm_vcpu *vcpu,
emulate_ctxt.fs_base = get_segment_base(vcpu, VCPU_SREG_FS);
vcpu->mmio_is_write = 0;
+ vcpu->pio.string = 0;
r = x86_emulate_memop(&emulate_ctxt, &emulate_ops);
+ if (vcpu->pio.string)
+ return EMULATE_DO_MMIO;
if ((r || vcpu->mmio_is_write) && run) {
run->exit_reason = KVM_EXIT_MMIO;
diff --git a/drivers/kvm/svm.c b/drivers/kvm/svm.c
index 436bdff..a83ff01 100644
--- a/drivers/kvm/svm.c
+++ b/drivers/kvm/svm.c
@@ -98,20 +98,6 @@ static inline u32 svm_has(u32 feat)
return svm_features & feat;
}
-static unsigned get_addr_size(struct vcpu_svm *svm)
-{
- struct vmcb_save_area *sa = &svm->vmcb->save;
- u16 cs_attrib;
-
- if (!(sa->cr0 & X86_CR0_PE) || (sa->rflags & X86_EFLAGS_VM))
- return 2;
-
- cs_attrib = sa->cs.attrib;
-
- return (cs_attrib & SVM_SELECTOR_L_MASK) ? 8 :
- (cs_attrib & SVM_SELECTOR_DB_MASK) ? 4 : 2;
-}
-
static inline u8 pop_irq(struct kvm_vcpu *vcpu)
{
int word_index = __ffs(vcpu->irq_summary);
@@ -995,147 +981,32 @@ static int shutdown_interception(struct vcpu_svm *svm, struct kvm_run *kvm_run)
return 0;
}
-static int io_get_override(struct vcpu_svm *svm,
- struct vmcb_seg **seg,
- int *addr_override)
-{
- u8 inst[MAX_INST_SIZE];
- unsigned ins_length;
- gva_t rip;
- int i;
-
- rip = svm->vmcb->save.rip;
- ins_length = svm->next_rip - rip;
- rip += svm->vmcb->save.cs.base;
-
- if (ins_length > MAX_INST_SIZE)
- printk(KERN_DEBUG
- "%s: inst length err, cs base 0x%llx rip 0x%llx "
- "next rip 0x%llx ins_length %u\n",
- __FUNCTION__,
- svm->vmcb->save.cs.base,
- svm->vmcb->save.rip,
- svm->vmcb->control.exit_info_2,
- ins_length);
-
- if (emulator_read_std(rip, inst, ins_length, &svm->vcpu)
- != X86EMUL_CONTINUE)
- /* #PF */
- return 0;
-
- *addr_override = 0;
- *seg = NULL;
- for (i = 0; i < ins_length; i++)
- switch (inst[i]) {
- case 0xf0:
- case 0xf2:
- case 0xf3:
- case 0x66:
- continue;
- case 0x67:
- *addr_override = 1;
- continue;
- case 0x2e:
- *seg = &svm->vmcb->save.cs;
- continue;
- case 0x36:
- *seg = &svm->vmcb->save.ss;
- continue;
- case 0x3e:
- *seg = &svm->vmcb->save.ds;
- continue;
- case 0x26:
- *seg = &svm->vmcb->save.es;
- continue;
- case 0x64:
- *seg = &svm->vmcb->save.fs;
- continue;
- case 0x65:
- *seg = &svm->vmcb->save.gs;
- continue;
- default:
- return 1;
- }
- printk(KERN_DEBUG "%s: unexpected\n", __FUNCTION__);
- return 0;
-}
-
-static unsigned long io_address(struct vcpu_svm *svm, int ins, gva_t *address)
-{
- unsigned long addr_mask;
- unsigned long *reg;
- struct vmcb_seg *seg;
- int addr_override;
- struct vmcb_save_area *save_area = &svm->vmcb->save;
- u16 cs_attrib = save_area->cs.attrib;
- unsigned addr_size = get_addr_size(svm);
-
- if (!io_get_override(svm, &seg, &addr_override))
- return 0;
-
- if (addr_override)
- addr_size = (addr_size == 2) ? 4: (addr_size >> 1);
-
- if (ins) {
- reg = &svm->vcpu.regs[VCPU_REGS_RDI];
- seg = &svm->vmcb->save.es;
- } else {
- reg = &svm->vcpu.regs[VCPU_REGS_RSI];
- seg = (seg) ? seg : &svm->vmcb->save.ds;
- }
-
- addr_mask = ~0ULL >> (64 - (addr_size * 8));
-
- if ((cs_attrib & SVM_SELECTOR_L_MASK) &&
- !(svm->vmcb->save.rflags & X86_EFLAGS_VM)) {
- *address = (*reg & addr_mask);
- return addr_mask;
- }
-
- if (!(seg->attrib & SVM_SELECTOR_P_SHIFT)) {
- svm_inject_gp(&svm->vcpu, 0);
- return 0;
- }
-
- *address = (*reg & addr_mask) + seg->base;
- return addr_mask;
-}
-
static int io_interception(struct vcpu_svm *svm, struct kvm_run *kvm_run)
{
u32 io_info = svm->vmcb->control.exit_info_1; //address size bug?
int size, down, in, string, rep;
unsigned port;
- unsigned long count;
- gva_t address = 0;
++svm->vcpu.stat.io_exits;
svm->next_rip = svm->vmcb->control.exit_info_2;
+ string = (io_info & SVM_IOIO_STR_MASK) != 0;
+
+ if (string) {
+ if (emulate_instruction(&svm->vcpu, kvm_run, 0, 0) == EMULATE_DO_MMIO)
+ return 0;
+ return 1;
+ }
+
in = (io_info & SVM_IOIO_TYPE_MASK) != 0;
port = io_info >> 16;
size = (io_info & SVM_IOIO_SIZE_MASK) >> SVM_IOIO_SIZE_SHIFT;
- string = (io_info & SVM_IOIO_STR_MASK) != 0;
rep = (io_info & SVM_IOIO_REP_MASK) != 0;
- count = 1;
down = (svm->vmcb->save.rflags & X86_EFLAGS_DF) != 0;
- if (string) {
- unsigned addr_mask;
-
- addr_mask = io_address(svm, in, &address);
- if (!addr_mask) {
- printk(KERN_DEBUG "%s: get io address failed\n",
- __FUNCTION__);
- return 1;
- }
-
- if (rep)
- count = svm->vcpu.regs[VCPU_REGS_RCX] & addr_mask;
- }
- return kvm_setup_pio(&svm->vcpu, kvm_run, in, size, count, string,
- down, address, rep, port);
+ return kvm_setup_pio(&svm->vcpu, kvm_run, in, size, 1, 0,
+ down, 0, rep, port);
}
static int nop_on_interception(struct vcpu_svm *svm, struct kvm_run *kvm_run)
diff --git a/drivers/kvm/vmx.c b/drivers/kvm/vmx.c
index 30c627d..044722b 100644
--- a/drivers/kvm/vmx.c
+++ b/drivers/kvm/vmx.c
@@ -1763,82 +1763,30 @@ static int handle_triple_fault(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
return 0;
}
-static int get_io_count(struct kvm_vcpu *vcpu, unsigned long *count)
-{
- u64 inst;
- gva_t rip;
- int countr_size;
- int i;
-
- if ((vmcs_readl(GUEST_RFLAGS) & X86_EFLAGS_VM)) {
- countr_size = 2;
- } else {
- u32 cs_ar = vmcs_read32(GUEST_CS_AR_BYTES);
-
- countr_size = (cs_ar & AR_L_MASK) ? 8:
- (cs_ar & AR_DB_MASK) ? 4: 2;
- }
-
- rip = vmcs_readl(GUEST_RIP);
- if (countr_size != 8)
- rip += vmcs_readl(GUEST_CS_BASE);
-
- if (emulator_read_std(rip, &inst, sizeof(inst), vcpu) !=
- X86EMUL_CONTINUE)
- return 0;
-
- for (i = 0; i < sizeof(inst); i++) {
- switch (((u8*)&inst)[i]) {
- case 0xf0:
- case 0xf2:
- case 0xf3:
- case 0x2e:
- case 0x36:
- case 0x3e:
- case 0x26:
- case 0x64:
- case 0x65:
- case 0x66:
- break;
- case 0x67:
- countr_size = (countr_size == 2) ? 4: (countr_size >> 1);
- default:
- goto done;
- }
- }
- return 0;
-done:
- countr_size *= 8;
- *count = vcpu->regs[VCPU_REGS_RCX] & (~0ULL >> (64 - countr_size));
- //printk("cx: %lx\n", vcpu->regs[VCPU_REGS_RCX]);
- return 1;
-}
-
static int handle_io(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
{
u64 exit_qualification;
int size, down, in, string, rep;
unsigned port;
- unsigned long count;
- gva_t address;
++vcpu->stat.io_exits;
exit_qualification = vmcs_read64(EXIT_QUALIFICATION);
- in = (exit_qualification & 8) != 0;
- size = (exit_qualification & 7) + 1;
string = (exit_qualification & 16) != 0;
+
+ if (string) {
+ if (emulate_instruction(vcpu, kvm_run, 0, 0) == EMULATE_DO_MMIO)
+ return 0;
+ return 1;
+ }
+
+ size = (exit_qualification & 7) + 1;
+ in = (exit_qualification & 8) != 0;
down = (vmcs_readl(GUEST_RFLAGS) & X86_EFLAGS_DF) != 0;
- count = 1;
rep = (exit_qualification & 32) != 0;
port = exit_qualification >> 16;
- address = 0;
- if (string) {
- if (rep && !get_io_count(vcpu, &count))
- return 1;
- address = vmcs_readl(GUEST_LINEAR_ADDRESS);
- }
- return kvm_setup_pio(vcpu, kvm_run, in, size, count, string, down,
- address, rep, port);
+
+ return kvm_setup_pio(vcpu, kvm_run, in, size, 1, 0, down,
+ 0, rep, port);
}
static void
diff --git a/drivers/kvm/x86_emulate.c b/drivers/kvm/x86_emulate.c
index 44eb28d..d553719 100644
--- a/drivers/kvm/x86_emulate.c
+++ b/drivers/kvm/x86_emulate.c
@@ -103,9 +103,12 @@ static u8 opcode_table[256] = {
/* 0x58 - 0x5F */
ImplicitOps, ImplicitOps, ImplicitOps, ImplicitOps,
ImplicitOps, ImplicitOps, ImplicitOps, ImplicitOps,
- /* 0x60 - 0x6F */
+ /* 0x60 - 0x6B */
0, 0, 0, DstReg | SrcMem32 | ModRM | Mov /* movsxd (x86/64) */ ,
- 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+ 0, 0, 0, 0, 0, 0, 0, 0,
+ /* 0x6C - 0x6F */
+ SrcNone | ByteOp | ImplicitOps, SrcNone | ImplicitOps, /* insb, insw/insd */
+ SrcNone | ByteOp | ImplicitOps, SrcNone | ImplicitOps, /* outsb, outsw/outsd */
/* 0x70 - 0x7F */
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
/* 0x80 - 0x87 */
@@ -428,10 +431,11 @@ struct operand {
})
/* Access/update address held in a register, based on addressing mode. */
+#define address_mask(reg) \
+ ((ad_bytes == sizeof(unsigned long)) ? \
+ (reg) : ((reg) & ((1UL << (ad_bytes << 3)) - 1)))
#define register_address(base, reg) \
- ((base) + ((ad_bytes == sizeof(unsigned long)) ? (reg) : \
- ((reg) & ((1UL << (ad_bytes << 3)) - 1))))
-
+ ((base) + address_mask(reg))
#define register_address_increment(reg, inc) \
do { \
/* signed type ensures sign extension to long */ \
@@ -1116,6 +1120,41 @@ done:
special_insn:
if (twobyte)
goto twobyte_special_insn;
+ switch(b) {
+ case 0x6c: /* insb */
+ case 0x6d: /* insw/insd */
+ if (kvm_setup_pio(ctxt->vcpu, NULL,
+ 1, /* in */
+ (d & ByteOp) ? 1 : op_bytes, /* size */
+ rep_prefix ?
+ address_mask(_regs[VCPU_REGS_RCX]) : 1, /* count */
+ 1, /* strings */
+ (_eflags & EFLG_DF), /* down */
+ register_address(ctxt->es_base,
+ _regs[VCPU_REGS_RDI]), /* address */
+ rep_prefix,
+ _regs[VCPU_REGS_RDX] /* port */
+ ) == 0)
+ return -1;
+ return 0;
+ case 0x6e: /* outsb */
+ case 0x6f: /* outsw/outsd */
+ if (kvm_setup_pio(ctxt->vcpu, NULL,
+ 0, /* in */
+ (d & ByteOp) ? 1 : op_bytes, /* size */
+ rep_prefix ?
+ address_mask(_regs[VCPU_REGS_RCX]) : 1, /* count */
+ 1, /* strings */
+ (_eflags & EFLG_DF), /* down */
+ register_address(override_base ?
+ *override_base : ctxt->ds_base,
+ _regs[VCPU_REGS_RSI]), /* address */
+ rep_prefix,
+ _regs[VCPU_REGS_RDX] /* port */
+ ) == 0)
+ return -1;
+ return 0;
+ }
if (rep_prefix) {
if (_regs[VCPU_REGS_RCX] == 0) {
ctxt->vcpu->rip = _eip;
--
1.5.3
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/