Commit d1442d85 authored by Nadav Amit's avatar Nadav Amit Committed by Paolo Bonzini

KVM: x86: Handle errors when RIP is set during far jumps

Far jmp/call/ret may fault while loading a new RIP.  Currently KVM does not
handle this case, and may result in failed vm-entry once the assignment is
done.  The tricky part of doing so is that loading the new CS affects the
VMCS/VMCB state, so if we fail during loading the new RIP, we are left in
unconsistent state.  Therefore, this patch saves on 64-bit the old CS
descriptor and restores it if loading RIP failed.

This fixes CVE-2014-3647.

Cc: stable@vger.kernel.org
Signed-off-by: default avatarNadav Amit <namit@cs.technion.ac.il>
Signed-off-by: default avatarPaolo Bonzini <pbonzini@redhat.com>
parent 234f3ce4
...@@ -1443,7 +1443,9 @@ static int write_segment_descriptor(struct x86_emulate_ctxt *ctxt, ...@@ -1443,7 +1443,9 @@ static int write_segment_descriptor(struct x86_emulate_ctxt *ctxt,
/* Does not support long mode */ /* Does not support long mode */
static int __load_segment_descriptor(struct x86_emulate_ctxt *ctxt, static int __load_segment_descriptor(struct x86_emulate_ctxt *ctxt,
u16 selector, int seg, u8 cpl, bool in_task_switch) u16 selector, int seg, u8 cpl,
bool in_task_switch,
struct desc_struct *desc)
{ {
struct desc_struct seg_desc, old_desc; struct desc_struct seg_desc, old_desc;
u8 dpl, rpl; u8 dpl, rpl;
...@@ -1584,6 +1586,8 @@ static int __load_segment_descriptor(struct x86_emulate_ctxt *ctxt, ...@@ -1584,6 +1586,8 @@ static int __load_segment_descriptor(struct x86_emulate_ctxt *ctxt,
} }
load: load:
ctxt->ops->set_segment(ctxt, selector, &seg_desc, base3, seg); ctxt->ops->set_segment(ctxt, selector, &seg_desc, base3, seg);
if (desc)
*desc = seg_desc;
return X86EMUL_CONTINUE; return X86EMUL_CONTINUE;
exception: exception:
return emulate_exception(ctxt, err_vec, err_code, true); return emulate_exception(ctxt, err_vec, err_code, true);
...@@ -1593,7 +1597,7 @@ static int load_segment_descriptor(struct x86_emulate_ctxt *ctxt, ...@@ -1593,7 +1597,7 @@ static int load_segment_descriptor(struct x86_emulate_ctxt *ctxt,
u16 selector, int seg) u16 selector, int seg)
{ {
u8 cpl = ctxt->ops->cpl(ctxt); u8 cpl = ctxt->ops->cpl(ctxt);
return __load_segment_descriptor(ctxt, selector, seg, cpl, false); return __load_segment_descriptor(ctxt, selector, seg, cpl, false, NULL);
} }
static void write_register_operand(struct operand *op) static void write_register_operand(struct operand *op)
...@@ -1987,17 +1991,31 @@ static int em_iret(struct x86_emulate_ctxt *ctxt) ...@@ -1987,17 +1991,31 @@ static int em_iret(struct x86_emulate_ctxt *ctxt)
static int em_jmp_far(struct x86_emulate_ctxt *ctxt) static int em_jmp_far(struct x86_emulate_ctxt *ctxt)
{ {
int rc; int rc;
unsigned short sel; unsigned short sel, old_sel;
struct desc_struct old_desc, new_desc;
const struct x86_emulate_ops *ops = ctxt->ops;
u8 cpl = ctxt->ops->cpl(ctxt);
/* Assignment of RIP may only fail in 64-bit mode */
if (ctxt->mode == X86EMUL_MODE_PROT64)
ops->get_segment(ctxt, &old_sel, &old_desc, NULL,
VCPU_SREG_CS);
memcpy(&sel, ctxt->src.valptr + ctxt->op_bytes, 2); memcpy(&sel, ctxt->src.valptr + ctxt->op_bytes, 2);
rc = load_segment_descriptor(ctxt, sel, VCPU_SREG_CS); rc = __load_segment_descriptor(ctxt, sel, VCPU_SREG_CS, cpl, false,
&new_desc);
if (rc != X86EMUL_CONTINUE) if (rc != X86EMUL_CONTINUE)
return rc; return rc;
ctxt->_eip = 0; rc = assign_eip_far(ctxt, ctxt->src.val, new_desc.l);
memcpy(&ctxt->_eip, ctxt->src.valptr, ctxt->op_bytes); if (rc != X86EMUL_CONTINUE) {
return X86EMUL_CONTINUE; WARN_ON(!ctxt->mode != X86EMUL_MODE_PROT64);
/* assigning eip failed; restore the old cs */
ops->set_segment(ctxt, old_sel, &old_desc, 0, VCPU_SREG_CS);
return rc;
}
return rc;
} }
static int em_grp45(struct x86_emulate_ctxt *ctxt) static int em_grp45(struct x86_emulate_ctxt *ctxt)
...@@ -2064,21 +2082,34 @@ static int em_ret(struct x86_emulate_ctxt *ctxt) ...@@ -2064,21 +2082,34 @@ static int em_ret(struct x86_emulate_ctxt *ctxt)
static int em_ret_far(struct x86_emulate_ctxt *ctxt) static int em_ret_far(struct x86_emulate_ctxt *ctxt)
{ {
int rc; int rc;
unsigned long cs; unsigned long eip, cs;
u16 old_cs;
int cpl = ctxt->ops->cpl(ctxt); int cpl = ctxt->ops->cpl(ctxt);
struct desc_struct old_desc, new_desc;
const struct x86_emulate_ops *ops = ctxt->ops;
if (ctxt->mode == X86EMUL_MODE_PROT64)
ops->get_segment(ctxt, &old_cs, &old_desc, NULL,
VCPU_SREG_CS);
rc = emulate_pop(ctxt, &ctxt->_eip, ctxt->op_bytes); rc = emulate_pop(ctxt, &eip, ctxt->op_bytes);
if (rc != X86EMUL_CONTINUE) if (rc != X86EMUL_CONTINUE)
return rc; return rc;
if (ctxt->op_bytes == 4)
ctxt->_eip = (u32)ctxt->_eip;
rc = emulate_pop(ctxt, &cs, ctxt->op_bytes); rc = emulate_pop(ctxt, &cs, ctxt->op_bytes);
if (rc != X86EMUL_CONTINUE) if (rc != X86EMUL_CONTINUE)
return rc; return rc;
/* Outer-privilege level return is not implemented */ /* Outer-privilege level return is not implemented */
if (ctxt->mode >= X86EMUL_MODE_PROT16 && (cs & 3) > cpl) if (ctxt->mode >= X86EMUL_MODE_PROT16 && (cs & 3) > cpl)
return X86EMUL_UNHANDLEABLE; return X86EMUL_UNHANDLEABLE;
rc = load_segment_descriptor(ctxt, (u16)cs, VCPU_SREG_CS); rc = __load_segment_descriptor(ctxt, (u16)cs, VCPU_SREG_CS, 0, false,
&new_desc);
if (rc != X86EMUL_CONTINUE)
return rc;
rc = assign_eip_far(ctxt, eip, new_desc.l);
if (rc != X86EMUL_CONTINUE) {
WARN_ON(!ctxt->mode != X86EMUL_MODE_PROT64);
ops->set_segment(ctxt, old_cs, &old_desc, 0, VCPU_SREG_CS);
}
return rc; return rc;
} }
...@@ -2505,19 +2536,24 @@ static int load_state_from_tss16(struct x86_emulate_ctxt *ctxt, ...@@ -2505,19 +2536,24 @@ static int load_state_from_tss16(struct x86_emulate_ctxt *ctxt,
* Now load segment descriptors. If fault happens at this stage * Now load segment descriptors. If fault happens at this stage
* it is handled in a context of new task * it is handled in a context of new task
*/ */
ret = __load_segment_descriptor(ctxt, tss->ldt, VCPU_SREG_LDTR, cpl, true); ret = __load_segment_descriptor(ctxt, tss->ldt, VCPU_SREG_LDTR, cpl,
true, NULL);
if (ret != X86EMUL_CONTINUE) if (ret != X86EMUL_CONTINUE)
return ret; return ret;
ret = __load_segment_descriptor(ctxt, tss->es, VCPU_SREG_ES, cpl, true); ret = __load_segment_descriptor(ctxt, tss->es, VCPU_SREG_ES, cpl,
true, NULL);
if (ret != X86EMUL_CONTINUE) if (ret != X86EMUL_CONTINUE)
return ret; return ret;
ret = __load_segment_descriptor(ctxt, tss->cs, VCPU_SREG_CS, cpl, true); ret = __load_segment_descriptor(ctxt, tss->cs, VCPU_SREG_CS, cpl,
true, NULL);
if (ret != X86EMUL_CONTINUE) if (ret != X86EMUL_CONTINUE)
return ret; return ret;
ret = __load_segment_descriptor(ctxt, tss->ss, VCPU_SREG_SS, cpl, true); ret = __load_segment_descriptor(ctxt, tss->ss, VCPU_SREG_SS, cpl,
true, NULL);
if (ret != X86EMUL_CONTINUE) if (ret != X86EMUL_CONTINUE)
return ret; return ret;
ret = __load_segment_descriptor(ctxt, tss->ds, VCPU_SREG_DS, cpl, true); ret = __load_segment_descriptor(ctxt, tss->ds, VCPU_SREG_DS, cpl,
true, NULL);
if (ret != X86EMUL_CONTINUE) if (ret != X86EMUL_CONTINUE)
return ret; return ret;
...@@ -2642,25 +2678,32 @@ static int load_state_from_tss32(struct x86_emulate_ctxt *ctxt, ...@@ -2642,25 +2678,32 @@ static int load_state_from_tss32(struct x86_emulate_ctxt *ctxt,
* Now load segment descriptors. If fault happenes at this stage * Now load segment descriptors. If fault happenes at this stage
* it is handled in a context of new task * it is handled in a context of new task
*/ */
ret = __load_segment_descriptor(ctxt, tss->ldt_selector, VCPU_SREG_LDTR, cpl, true); ret = __load_segment_descriptor(ctxt, tss->ldt_selector, VCPU_SREG_LDTR,
cpl, true, NULL);
if (ret != X86EMUL_CONTINUE) if (ret != X86EMUL_CONTINUE)
return ret; return ret;
ret = __load_segment_descriptor(ctxt, tss->es, VCPU_SREG_ES, cpl, true); ret = __load_segment_descriptor(ctxt, tss->es, VCPU_SREG_ES, cpl,
true, NULL);
if (ret != X86EMUL_CONTINUE) if (ret != X86EMUL_CONTINUE)
return ret; return ret;
ret = __load_segment_descriptor(ctxt, tss->cs, VCPU_SREG_CS, cpl, true); ret = __load_segment_descriptor(ctxt, tss->cs, VCPU_SREG_CS, cpl,
true, NULL);
if (ret != X86EMUL_CONTINUE) if (ret != X86EMUL_CONTINUE)
return ret; return ret;
ret = __load_segment_descriptor(ctxt, tss->ss, VCPU_SREG_SS, cpl, true); ret = __load_segment_descriptor(ctxt, tss->ss, VCPU_SREG_SS, cpl,
true, NULL);
if (ret != X86EMUL_CONTINUE) if (ret != X86EMUL_CONTINUE)
return ret; return ret;
ret = __load_segment_descriptor(ctxt, tss->ds, VCPU_SREG_DS, cpl, true); ret = __load_segment_descriptor(ctxt, tss->ds, VCPU_SREG_DS, cpl,
true, NULL);
if (ret != X86EMUL_CONTINUE) if (ret != X86EMUL_CONTINUE)
return ret; return ret;
ret = __load_segment_descriptor(ctxt, tss->fs, VCPU_SREG_FS, cpl, true); ret = __load_segment_descriptor(ctxt, tss->fs, VCPU_SREG_FS, cpl,
true, NULL);
if (ret != X86EMUL_CONTINUE) if (ret != X86EMUL_CONTINUE)
return ret; return ret;
ret = __load_segment_descriptor(ctxt, tss->gs, VCPU_SREG_GS, cpl, true); ret = __load_segment_descriptor(ctxt, tss->gs, VCPU_SREG_GS, cpl,
true, NULL);
if (ret != X86EMUL_CONTINUE) if (ret != X86EMUL_CONTINUE)
return ret; return ret;
...@@ -2942,24 +2985,39 @@ static int em_call_far(struct x86_emulate_ctxt *ctxt) ...@@ -2942,24 +2985,39 @@ static int em_call_far(struct x86_emulate_ctxt *ctxt)
u16 sel, old_cs; u16 sel, old_cs;
ulong old_eip; ulong old_eip;
int rc; int rc;
struct desc_struct old_desc, new_desc;
const struct x86_emulate_ops *ops = ctxt->ops;
int cpl = ctxt->ops->cpl(ctxt);
old_cs = get_segment_selector(ctxt, VCPU_SREG_CS);
old_eip = ctxt->_eip; old_eip = ctxt->_eip;
ops->get_segment(ctxt, &old_cs, &old_desc, NULL, VCPU_SREG_CS);
memcpy(&sel, ctxt->src.valptr + ctxt->op_bytes, 2); memcpy(&sel, ctxt->src.valptr + ctxt->op_bytes, 2);
if (load_segment_descriptor(ctxt, sel, VCPU_SREG_CS)) rc = __load_segment_descriptor(ctxt, sel, VCPU_SREG_CS, cpl, false,
&new_desc);
if (rc != X86EMUL_CONTINUE)
return X86EMUL_CONTINUE; return X86EMUL_CONTINUE;
ctxt->_eip = 0; rc = assign_eip_far(ctxt, ctxt->src.val, new_desc.l);
memcpy(&ctxt->_eip, ctxt->src.valptr, ctxt->op_bytes); if (rc != X86EMUL_CONTINUE)
goto fail;
ctxt->src.val = old_cs; ctxt->src.val = old_cs;
rc = em_push(ctxt); rc = em_push(ctxt);
if (rc != X86EMUL_CONTINUE) if (rc != X86EMUL_CONTINUE)
return rc; goto fail;
ctxt->src.val = old_eip; ctxt->src.val = old_eip;
return em_push(ctxt); rc = em_push(ctxt);
/* If we failed, we tainted the memory, but the very least we should
restore cs */
if (rc != X86EMUL_CONTINUE)
goto fail;
return rc;
fail:
ops->set_segment(ctxt, old_cs, &old_desc, 0, VCPU_SREG_CS);
return rc;
} }
static int em_ret_near_imm(struct x86_emulate_ctxt *ctxt) static int em_ret_near_imm(struct x86_emulate_ctxt *ctxt)
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment