These functions follow the following path: hvm_do_resume() ->
handle_hvm_io_completion() -> hvm_wait_for_io() ->
wait_on_xen_event_channel() -> do_softirq() -> schedule() ->
sched_context_switch() -> continue_running() and hence may
recursively invoke themselves. If this ends up happening a couple of
times, a stack overflow would result.
Prevent this by also resetting the stack at the
->arch.ctxt_switch->tail() invocations (in both places for consistency)
and thus jumping to the functions instead of calling them.
This is XSA-348 / CVE-2020-29566.
Reported-by: Julien Grall <jgrall@amazon.com>
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
master commit:
e6ebd394385db52855d1517cea829ffff68b34b8
master date: 2020-12-15 13:41:23 +0100
(*dead_idle)();
}
-static void idle_loop(void)
+static void noreturn idle_loop(void)
{
unsigned int cpu = smp_processor_id();
reset_stack_and_jump(idle_loop);
}
-static void noreturn continue_idle_domain(struct vcpu *v)
-{
- reset_stack_and_jump(idle_loop);
-}
-
void dump_pageframe_info(struct domain *d)
{
struct page_info *page;
static const struct arch_csw idle_csw = {
.from = paravirt_ctxt_switch_from,
.to = paravirt_ctxt_switch_to,
- .tail = continue_idle_domain,
+ .tail = idle_loop,
};
d->arch.ctxt_switch = &idle_csw;
/* Ensure that the vcpu has an up-to-date time base. */
update_vcpu_system_time(next);
- /*
- * Schedule tail *should* be a terminal function pointer, but leave a
- * bug frame around just in case it returns, to save going back into the
- * context switching code and leaving a far more subtle crash to diagnose.
- */
- nextd->arch.ctxt_switch->tail(next);
- BUG();
+ reset_stack_and_jump_ind(nextd->arch.ctxt_switch->tail);
}
void continue_running(struct vcpu *same)
{
- /* See the comment above. */
- same->domain->arch.ctxt_switch->tail(same);
- BUG();
+ reset_stack_and_jump_ind(same->domain->arch.ctxt_switch->tail);
}
int __sync_local_execstate(void)
wrmsr_tsc_aux(hvm_msr_tsc_aux(v));
}
-static void noreturn svm_do_resume(struct vcpu *v)
+static void noreturn svm_do_resume(void)
{
+ struct vcpu *v = current;
struct vmcb_struct *vmcb = v->arch.hvm_svm.vmcb;
bool debug_state = (v->domain->debugger_attached ||
v->domain->arch.monitor.software_breakpoint_enabled ||
domain_crash_synchronous();
}
-void vmx_do_resume(struct vcpu *v)
+void vmx_do_resume(void)
{
+ struct vcpu *v = current;
bool_t debug_state;
unsigned long host_cr4;
}
custom_runtime_param("pcid", parse_pcid);
-static void noreturn continue_nonidle_domain(struct vcpu *v)
+static void noreturn continue_nonidle_domain(void)
{
check_wakeup_from_wait();
reset_stack_and_jump(ret_from_intr);
# define CHECK_FOR_LIVEPATCH_WORK ""
#endif
-#define reset_stack_and_jump(__fn) \
+#define switch_stack_and_jump(fn, instr, constr) \
({ \
__asm__ __volatile__ ( \
"mov %0,%%"__OP"sp;" \
- CHECK_FOR_LIVEPATCH_WORK \
- "jmp %c1" \
- : : "r" (guest_cpu_user_regs()), "i" (__fn) : "memory" ); \
+ CHECK_FOR_LIVEPATCH_WORK \
+ instr "1" \
+ : : "r" (guest_cpu_user_regs()), constr (fn) : "memory" ); \
unreachable(); \
})
+#define reset_stack_and_jump(fn) \
+ switch_stack_and_jump(fn, "jmp %c", "i")
+
+/* The constraint may only specify non-call-clobbered registers. */
+#define reset_stack_and_jump_ind(fn) \
+ switch_stack_and_jump(fn, "INDIRECT_JMP %", "b")
+
/*
* Which VCPU's state is currently running on each CPU?
* This is not necesasrily the same as 'current' as a CPU may be
const struct arch_csw {
void (*from)(struct vcpu *);
void (*to)(struct vcpu *);
- void (*tail)(struct vcpu *);
+ void noreturn (*tail)(void);
} *ctxt_switch;
/* nestedhvm: translate l2 guest physical to host physical */
void vmx_asm_vmexit_handler(struct cpu_user_regs);
void vmx_asm_do_vmentry(void);
void vmx_intr_assist(void);
-void noreturn vmx_do_resume(struct vcpu *);
+void noreturn vmx_do_resume(void);
void vmx_vlapic_msr_changed(struct vcpu *v);
void vmx_realmode_emulate_one(struct hvm_emulate_ctxt *hvmemul_ctxt);
void vmx_realmode(struct cpu_user_regs *regs);