/* Last of all, we look at what happens first of all. The very first time the
* Guest makes a hypercall, we end up here to set things up: */
static void initialize(struct lg_cpu *cpu)
{
/* You can't do anything until you're initialized. The Guest knows the
* rules, so we're unforgiving here. */
if (cpu->hcall->arg0 != LHCALL_LGUEST_INIT) {
kill_guest(cpu, "hypercall %li before INIT", cpu->hcall->arg0);
return;
}
if (lguest_arch_init_hypercalls(cpu))
kill_guest(cpu, "bad guest page %p", cpu->lg->lguest_data);
/* The Guest tells us where we're not to deliver interrupts by putting
* the range of addresses into "struct lguest_data". */
if (get_user(cpu->lg->noirq_start, &cpu->lg->lguest_data->noirq_start)
|| get_user(cpu->lg->noirq_end, &cpu->lg->lguest_data->noirq_end))
kill_guest(cpu, "bad guest page %p", cpu->lg->lguest_data);
/* We write the current time into the Guest's data page once so it can
* set its clock. */
write_timestamp(cpu);
/* page_tables.c will also do some setup. */
page_table_guest_data_init(cpu);
/* This is the one case where the above accesses might have been the
* first write to a Guest page. This may have caused a copy-on-write
* fault, but the old page might be (read-only) in the Guest
* pagetable. */
guest_pagetable_clear_all(cpu);
}
/*
* Our hypercalls mechanism used to be based on direct software interrupts.
* After Anthony's "Refactor hypercall infrastructure" kvm patch, we decided to
* change over to using kvm hypercalls.
*
* KVM_HYPERCALL is actually a "vmcall" instruction, which generates an invalid
* opcode fault (fault 6) on non-VT cpus, so the easiest solution seemed to be
* an *emulation approach*: if the fault was really produced by an hypercall
* (is_hypercall() does exactly this check), we can just call the corresponding
* hypercall host implementation function.
*
* But these invalid opcode faults are notably slower than software interrupts.
* So we implemented the *patching (or rewriting) approach*: every time we hit
* the KVM_HYPERCALL opcode in Guest code, we patch it to the old "int 0x1f"
* opcode, so next time the Guest calls this hypercall it will use the
* faster trap mechanism.
*
* Matias even benchmarked it to convince you: this shows the average cycle
* cost of a hypercall. For each alternative solution mentioned above we've
* made 5 runs of the benchmark:
*
* 1) direct software interrupt: 2915, 2789, 2764, 2721, 2898
* 2) emulation technique: 3410, 3681, 3466, 3392, 3780
* 3) patching (rewrite) technique: 2977, 2975, 2891, 2637, 2884
*
* One two-line function is worth a 20% hypercall speed boost!
*/
static void rewrite_hypercall(struct lg_cpu *cpu)
{
/*
* This are the opcodes we use to patch the Guest. The opcode for "int
* $0x1f" is "0xcd 0x1f" but vmcall instruction is 3 bytes long, so we
* complete the sequence with a NOP (0x90).
*/
u8 insn[3] = {0xcd, 0x1f, 0x90};
__lgwrite(cpu, guest_pa(cpu, cpu->regs->eip), insn, sizeof(insn));
/*
* The above write might have caused a copy of that page to be made
* (if it was read-only). We need to make sure the Guest has
* up-to-date pagetables. As this doesn't happen often, we can just
* drop them all.
*/
guest_pagetable_clear_all(cpu);
}
static void initialize(struct lg_cpu *cpu)
{
if (cpu->hcall->arg0 != LHCALL_LGUEST_INIT) {
kill_guest(cpu, "hypercall %li before INIT", cpu->hcall->arg0);
return;
}
if (lguest_arch_init_hypercalls(cpu))
kill_guest(cpu, "bad guest page %p", cpu->lg->lguest_data);
if (get_user(cpu->lg->noirq_start, &cpu->lg->lguest_data->noirq_start)
|| get_user(cpu->lg->noirq_end, &cpu->lg->lguest_data->noirq_end))
kill_guest(cpu, "bad guest page %p", cpu->lg->lguest_data);
write_timestamp(cpu);
page_table_guest_data_init(cpu);
guest_pagetable_clear_all(cpu);
}
/* For setting a mid-level, we just throw everything away. It's easy. */
void guest_set_pmd(struct lguest *lg, unsigned long pmdp, u32 idx)
{
guest_pagetable_clear_all(&lg->cpus[0]);
}
/*H:120 This is the core hypercall routine: where the Guest gets what it wants.
* Or gets killed. Or, in the case of LHCALL_SHUTDOWN, both. */
static void do_hcall(struct lg_cpu *cpu, struct hcall_args *args)
{
switch (args->arg0) {
case LHCALL_FLUSH_ASYNC:
/* This call does nothing, except by breaking out of the Guest
* it makes us process all the asynchronous hypercalls. */
break;
case LHCALL_LGUEST_INIT:
/* You can't get here unless you're already initialized. Don't
* do that. */
kill_guest(cpu, "already have lguest_data");
break;
case LHCALL_SHUTDOWN: {
/* Shutdown is such a trivial hypercall that we do it in four
* lines right here. */
char msg[128];
/* If the lgread fails, it will call kill_guest() itself; the
* kill_guest() with the message will be ignored. */
__lgread(cpu, msg, args->arg1, sizeof(msg));
msg[sizeof(msg)-1] = '\0';
kill_guest(cpu, "CRASH: %s", msg);
if (args->arg2 == LGUEST_SHUTDOWN_RESTART)
cpu->lg->dead = ERR_PTR(-ERESTART);
break;
}
case LHCALL_FLUSH_TLB:
/* FLUSH_TLB comes in two flavors, depending on the
* argument: */
if (args->arg1)
guest_pagetable_clear_all(cpu);
else
guest_pagetable_flush_user(cpu);
break;
/* All these calls simply pass the arguments through to the right
* routines. */
case LHCALL_NEW_PGTABLE:
guest_new_pagetable(cpu, args->arg1);
break;
case LHCALL_SET_STACK:
guest_set_stack(cpu, args->arg1, args->arg2, args->arg3);
break;
case LHCALL_SET_PTE:
guest_set_pte(cpu, args->arg1, args->arg2, __pte(args->arg3));
break;
case LHCALL_SET_PMD:
guest_set_pmd(cpu->lg, args->arg1, args->arg2);
break;
case LHCALL_SET_CLOCKEVENT:
guest_set_clockevent(cpu, args->arg1);
break;
case LHCALL_TS:
/* This sets the TS flag, as we saw used in run_guest(). */
cpu->ts = args->arg1;
break;
case LHCALL_HALT:
/* Similarly, this sets the halted flag for run_guest(). */
cpu->halted = 1;
break;
case LHCALL_NOTIFY:
cpu->pending_notify = args->arg1;
break;
default:
/* It should be an architecture-specific hypercall. */
if (lguest_arch_do_hcall(cpu, args))
kill_guest(cpu, "Bad hypercall %li\n", args->arg0);
}
}
static void do_hcall(struct lg_cpu *cpu, struct hcall_args *args)
{
switch (args->arg0) {
case LHCALL_FLUSH_ASYNC:
break;
case LHCALL_SEND_INTERRUPTS:
break;
case LHCALL_LGUEST_INIT:
kill_guest(cpu, "already have lguest_data");
break;
case LHCALL_SHUTDOWN: {
char msg[128];
__lgread(cpu, msg, args->arg1, sizeof(msg));
msg[sizeof(msg)-1] = '\0';
kill_guest(cpu, "CRASH: %s", msg);
if (args->arg2 == LGUEST_SHUTDOWN_RESTART)
cpu->lg->dead = ERR_PTR(-ERESTART);
break;
}
case LHCALL_FLUSH_TLB:
if (args->arg1)
guest_pagetable_clear_all(cpu);
else
guest_pagetable_flush_user(cpu);
break;
case LHCALL_NEW_PGTABLE:
guest_new_pagetable(cpu, args->arg1);
break;
case LHCALL_SET_STACK:
guest_set_stack(cpu, args->arg1, args->arg2, args->arg3);
break;
case LHCALL_SET_PTE:
#ifdef CONFIG_X86_PAE
guest_set_pte(cpu, args->arg1, args->arg2,
__pte(args->arg3 | (u64)args->arg4 << 32));
#else
guest_set_pte(cpu, args->arg1, args->arg2, __pte(args->arg3));
#endif
break;
case LHCALL_SET_PGD:
guest_set_pgd(cpu->lg, args->arg1, args->arg2);
break;
#ifdef CONFIG_X86_PAE
case LHCALL_SET_PMD:
guest_set_pmd(cpu->lg, args->arg1, args->arg2);
break;
#endif
case LHCALL_SET_CLOCKEVENT:
guest_set_clockevent(cpu, args->arg1);
break;
case LHCALL_TS:
cpu->ts = args->arg1;
break;
case LHCALL_HALT:
cpu->halted = 1;
break;
case LHCALL_NOTIFY:
cpu->pending_notify = args->arg1;
break;
default:
if (lguest_arch_do_hcall(cpu, args))
kill_guest(cpu, "Bad hypercall %li\n", args->arg0);
}
}
