path: root/drivers/lguest/interrupts_and_traps.c

/*P:800
 * Interrupts (traps) are complicated enough to earn their own file.
 * There are three classes of interrupts:
 *
 * 1) Real hardware interrupts which occur while we're running the Guest,
 * 2) Interrupts for virtual devices attached to the Guest, and
 * 3) Traps and faults from the Guest.
 *
 * Real hardware interrupts must be delivered to the Host, not the Guest.
 * Virtual interrupts must be delivered to the Guest, but we make them look
 * just like real hardware would deliver them.  Traps from the Guest can be set
 * up to go directly back into the Guest, but sometimes the Host wants to see
 * them first, so we also have a way of "reflecting" them into the Guest as if
 * they had been delivered to it directly.
:*/
#include <linux/uaccess.h>
#include <linux/interrupt.h>
#include <linux/module.h>
#include <linux/sched.h>
#include "lg.h"

/* Allow Guests to use a non-128 (ie. non-Linux) syscall trap. */
static unsigned int syscall_vector = IA32_SYSCALL_VECTOR;
module_param(syscall_vector, uint, 0444);

/* The address of the interrupt handler is split into two bits: */
static unsigned long idt_address(u32 lo, u32 hi)
{
	return (lo & 0x0000FFFF) | (hi & 0xFFFF0000);
}

/*
 * The "type" of the interrupt handler is a 4 bit field: we only support a
 * couple of types.
 */
static int idt_type(u32 lo, u32 hi)
{
	return (hi >> 8) & 0xF;
}

/* An IDT entry can't be used unless the "present" bit is set. */
static bool idt_present(u32 lo, u32 hi)
{
	return (hi & 0x8000);
}

/*
 * We need a helper to "push" a value onto the Guest's stack, since that's a
 * big part of what delivering an interrupt does.
 */
static void push_guest_stack(struct lg_cpu *cpu, unsigned long *gstack, u32 val)
{
	/* Stack grows upwards: move stack then write value. */
	*gstack -= 4;
	lgwrite(cpu, *gstack, u32, val);
}

/*H:210
 * The push_guest_interrupt_stack() routine saves Guest state on the stack for
 * an interrupt or trap.  The mechanics of delivering traps and interrupts to
 * the Guest are the same, except some traps have an "error code" which gets
 * pushed onto the stack as well: the caller tells us if this is one.
 *
 * We set up the stack just like the CPU does for a real interrupt, so it's
 * identical for the Guest (and the standard "iret" instruction will undo
 * it).
 */
static void push_guest_interrupt_stack(struct lg_cpu *cpu, bool has_err)
{
	unsigned long gstack, origstack;
	u32 eflags, ss, irq_enable;
	unsigned long virtstack;

	/*
	 * There are two cases for interrupts: one where the Guest is already
	 * in the kernel, and a more complex one where the Guest is in
	 * userspace.  We check the privilege level to find out.
	 */
	if ((cpu->regs->ss&0x3) != GUEST_PL) {
		/*
		 * The Guest told us their kernel stack with the SET_STACK
		 * hypercall: both the virtual address and the segment.
		 */
		virtstack = cpu->esp1;
		ss = cpu->ss1;

		origstack = gstack = guest_pa(cpu, virtstack);
		/*
		 * We push the old stack segment and pointer onto the new
		 * stack: when the Guest does an "iret" back from the interrupt
		 * handler the CPU will notice they're dropping privilege
		 * levels and expect these here.
		 */
		push_guest_stack(cpu, &gstack, cpu->regs->ss);
		push_guest_stack(cpu, &gstack, cpu->regs->esp);
	} else {
		/* We're staying on the same Guest (kernel) stack. */
		virtstack = cpu->regs->esp;
		ss = cpu->regs->ss;

		origstack = gstack = guest_pa(cpu, virtstack);
	}

	/*
	 * Remember that we never let the Guest actually disable interrupts, so
	 * the "Interrupt Flag" bit is always set.  We copy that bit from the
	 * Guest's "irq_enabled" field into the eflags word: we saw the Guest
	 * copy it back in "lguest_iret".
	 */
	eflags = cpu->regs->eflags;
	if (get_user(irq_enable, &cpu->lg->lguest_data->irq_enabled) == 0
	    && !(irq_enable & X86_EFLAGS_IF))
		eflags &= ~X86_EFLAGS_IF;

	/*
	 * An interrupt is expected to push three things on the stack: the old
	 * "eflags" word, the old code segment, and the old instruction
	 * pointer.
	 */
	push_guest_stack(cpu, &gstack, eflags);
	push_guest_stack(cpu, &gstack, cpu->regs->cs);
	push_guest_stack(cpu, &gstack, cpu->regs->eip);

	/* For the six traps which supply an error code, we push that, too. */
	if (has_err)
		push_guest_stack(cpu, &gstack, cpu->regs->errcode);

	/* Adjust the stack pointer and stack segment. */
	cpu->regs->ss = ss;
	cpu->regs->esp = virtstack + (gstack - origstack);
}

/*
 * This actually makes the Guest start executing the given interrupt/trap
 * handler.
 *
 * "lo" and "hi" are the two parts of the Interrupt Descriptor Table for this
 * interrupt or trap.  It's split into two parts for traditional reasons: gcc
 * on i386 used to be frightened by 64 bit numbers.
 */
static void guest_run_interrupt(struct lg_cpu *cpu, u32 lo, u32 hi)
{
	/* If we're already in the kernel, we don't change stacks. */
	if ((cpu->regs->ss&0x3) != GUEST_PL)
		cpu->regs->ss = cpu->esp1;

	/*
	 * Set the code segment and the address to execute.
	 */
	cpu->regs->cs = (__KERNEL_CS|GUEST_PL);
	cpu->regs->eip = idt_address(lo, hi);

	/*
	 * Trapping always clears these flags:
	 * TF: Trap flag
	 * VM: Virtual 8086 mode
	 * RF: Resume
	 * NT: Nested task.
	 */
	cpu->regs->eflags &=
		~(X86_EFLAGS_TF|X86_EFLAGS_VM|X86_EFLAGS_RF|X86_EFLAGS_NT);

	/*
	 * There are two kinds of interrupt handlers: 0xE is an "interrupt
	 * gate" which expects interrupts to be disabled on entry.
	 */
	if (idt_type(lo, hi) == 0xE)
		if (put_user(0, &cpu->lg->lguest_data->irq_enabled))
			kill_guest(cpu, "Disabling interrupts");
}

/* This restores the eflags word which was pushed on the stack by a trap */
static void restore_eflags(struct lg_cpu *cpu)
{
	/* This is the physical address of the stack. */
	unsigned long stack_pa = guest_pa(cpu, cpu->regs->esp);

	/*
	 * Stack looks like this:
	 * Address	Contents
	 * esp		EIP
	 * esp + 4	CS
	 * esp + 8	EFLAGS
	 */
	cpu->regs->eflags = lgread(cpu, stack_pa + 8, u32);
	cpu->regs->eflags &=
		~(X86_EFLAGS_TF|X86_EFLAGS_VM|X86_EFLAGS_RF|X86_EFLAGS_NT);
}

/*H:205
 * Virtual Interrupts.
 *
 * interrupt_pending() returns the first pending interrupt which isn't blocked
 * by the Guest.  It is called before every entry to the Guest, and just before
 * we go to sleep when the Guest has halted itself.
 */
unsigned int interrupt_pending(struct lg_cpu *cpu, bool *more)
{
	unsigned int irq;
	DECLARE_BITMAP(blk, LGUEST_IRQS);

	/* If the Guest hasn't even initialized yet, we can do nothing. */
	if (!cpu->lg->lguest_data)
		return LGUEST_IRQS;

	/*
	 * Take our "irqs_pending" array and remove any interrupts the Guest
	 * wants blocked: the result ends up in "blk".
	 */
	if (copy_from_user(&blk, cpu->lg->lguest_data->blocked_interrupts,
			   sizeof(blk)))
		return LGUEST_IRQS;
	bitmap_andnot(blk, cpu->irqs_pending, blk, LGUEST_IRQS);

	/* Find the first interrupt. */
	irq = find_first_bit(blk, LGUEST_IRQS);
	*more = find_next_bit(blk, LGUEST_IRQS, irq+1);

	return irq;
}

/*
 * This actually diverts the Guest to running an interrupt handler, once an
 * interrupt has been identified by interrupt_pending().
 */
void try_deliver_interrupt(struct lg_cpu *cpu, unsigned int irq, bool more)
{
	struct desc_struct *idt;

	BUG_ON(irq >= LGUEST_IRQS);

	/* If they're halted, interrupts restart them. */
	if (cpu->halted) {
		/* Re-enable interrupts. */
		if (put_user(X86_EFLAGS_IF, &cpu->lg->lguest_data->irq_enabled))
			kill_guest(