boot.c

linux 内核源代码

{
	/* Set up the timer interrupt (0) to go to our simple timer routine */
	set_irq_handler(0, lguest_time_irq);

	/* Our clock structure looks like arch/x86/kernel/tsc_32.c if we can
	 * use the TSC, otherwise it's a dumb nanosecond-resolution clock.
	 * Either way, the "rating" is set so high that it's always chosen over
	 * any other clocksource. */
	if (lguest_data.tsc_khz)
		lguest_clock.mult = clocksource_khz2mult(lguest_data.tsc_khz,
							 lguest_clock.shift);
	clock_base = lguest_clock_read();
	clocksource_register(&lguest_clock);

	/* Now we've set up our clock, we can use it as the scheduler clock */
	pv_time_ops.sched_clock = lguest_sched_clock;

	/* We can't set cpumask in the initializer: damn C limitations!  Set it
	 * here and register our timer device. */
	lguest_clockevent.cpumask = cpumask_of_cpu(0);
	clockevents_register_device(&lguest_clockevent);

	/* Finally, we unblock the timer interrupt. */
	enable_lguest_irq(0);
}

/*
 * Miscellaneous bits and pieces.
 *
 * Here is an oddball collection of functions which the Guest needs for things
 * to work.  They're pretty simple.
 */

/* The Guest needs to tell the Host what stack it expects traps to use.  For
 * native hardware, this is part of the Task State Segment mentioned above in
 * lguest_load_tr_desc(), but to help hypervisors there's this special call.
 *
 * We tell the Host the segment we want to use (__KERNEL_DS is the kernel data
 * segment), the privilege level (we're privilege level 1, the Host is 0 and
 * will not tolerate us trying to use that), the stack pointer, and the number
 * of pages in the stack. */
static void lguest_load_esp0(struct tss_struct *tss,
				     struct thread_struct *thread)
{
	lazy_hcall(LHCALL_SET_STACK, __KERNEL_DS|0x1, thread->esp0,
		   THREAD_SIZE/PAGE_SIZE);
}

/* Let's just say, I wouldn't do debugging under a Guest. */
static void lguest_set_debugreg(int regno, unsigned long value)
{
	/* FIXME: Implement */
}

/* There are times when the kernel wants to make sure that no memory writes are
 * caught in the cache (that they've all reached real hardware devices).  This
 * doesn't matter for the Guest which has virtual hardware.
 *
 * On the Pentium 4 and above, cpuid() indicates that the Cache Line Flush
 * (clflush) instruction is available and the kernel uses that.  Otherwise, it
 * uses the older "Write Back and Invalidate Cache" (wbinvd) instruction.
 * Unlike clflush, wbinvd can only be run at privilege level 0.  So we can
 * ignore clflush, but replace wbinvd.
 */
static void lguest_wbinvd(void)
{
}

/* If the Guest expects to have an Advanced Programmable Interrupt Controller,
 * we play dumb by ignoring writes and returning 0 for reads.  So it's no
 * longer Programmable nor Controlling anything, and I don't think 8 lines of
 * code qualifies for Advanced.  It will also never interrupt anything.  It
 * does, however, allow us to get through the Linux boot code. */
#ifdef CONFIG_X86_LOCAL_APIC
static void lguest_apic_write(unsigned long reg, unsigned long v)
{
}

static unsigned long lguest_apic_read(unsigned long reg)
{
	return 0;
}
#endif

/* STOP!  Until an interrupt comes in. */
static void lguest_safe_halt(void)
{
	hcall(LHCALL_HALT, 0, 0, 0);
}

/* Perhaps CRASH isn't the best name for this hypercall, but we use it to get a
 * message out when we're crashing as well as elegant termination like powering
 * off.
 *
 * Note that the Host always prefers that the Guest speak in physical addresses
 * rather than virtual addresses, so we use __pa() here. */
static void lguest_power_off(void)
{
	hcall(LHCALL_CRASH, __pa("Power down"), 0, 0);
}

/*
 * Panicing.
 *
 * Don't.  But if you did, this is what happens.
 */
static int lguest_panic(struct notifier_block *nb, unsigned long l, void *p)
{
	hcall(LHCALL_CRASH, __pa(p), 0, 0);
	/* The hcall won't return, but to keep gcc happy, we're "done". */
	return NOTIFY_DONE;
}

static struct notifier_block paniced = {
	.notifier_call = lguest_panic
};

/* Setting up memory is fairly easy. */
static __init char *lguest_memory_setup(void)
{
	/* We do this here and not earlier because lockcheck barfs if we do it
	 * before start_kernel() */
	atomic_notifier_chain_register(&panic_notifier_list, &paniced);

	/* The Linux bootloader header contains an "e820" memory map: the
	 * Launcher populated the first entry with our memory limit. */
	add_memory_region(boot_params.e820_map[0].addr,
			  boot_params.e820_map[0].size,
			  boot_params.e820_map[0].type);

	/* This string is for the boot messages. */
	return "LGUEST";
}

/* We will eventually use the virtio console device to produce console output,
 * but before that is set up we use LHCALL_NOTIFY on normal memory to produce
 * console output. */
static __init int early_put_chars(u32 vtermno, const char *buf, int count)
{
	char scratch[17];
	unsigned int len = count;

	/* We use a nul-terminated string, so we have to make a copy.  Icky,
	 * huh? */
	if (len > sizeof(scratch) - 1)
		len = sizeof(scratch) - 1;
	scratch[len] = '\0';
	memcpy(scratch, buf, len);
	hcall(LHCALL_NOTIFY, __pa(scratch), 0, 0);

	/* This routine returns the number of bytes actually written. */
	return len;
}

/*G:050
 * Patching (Powerfully Placating Performance Pedants)
 *
 * We have already seen that pv_ops structures let us replace simple
 * native instructions with calls to the appropriate back end all throughout
 * the kernel.  This allows the same kernel to run as a Guest and as a native
 * kernel, but it's slow because of all the indirect branches.
 *
 * Remember that David Wheeler quote about "Any problem in computer science can
 * be solved with another layer of indirection"?  The rest of that quote is
 * "... But that usually will create another problem."  This is the first of
 * those problems.
 *
 * Our current solution is to allow the paravirt back end to optionally patch
 * over the indirect calls to replace them with something more efficient.  We
 * patch the four most commonly called functions: disable interrupts, enable
 * interrupts, restore interrupts and save interrupts.  We usually have 6 or 10
 * bytes to patch into: the Guest versions of these operations are small enough
 * that we can fit comfortably.
 *
 * First we need assembly templates of each of the patchable Guest operations,
 * and these are in lguest_asm.S. */

/*G:060 We construct a table from the assembler templates: */
static const struct lguest_insns
{
	const char *start, *end;
} lguest_insns[] = {
	[PARAVIRT_PATCH(pv_irq_ops.irq_disable)] = { lgstart_cli, lgend_cli },
	[PARAVIRT_PATCH(pv_irq_ops.irq_enable)] = { lgstart_sti, lgend_sti },
	[PARAVIRT_PATCH(pv_irq_ops.restore_fl)] = { lgstart_popf, lgend_popf },
	[PARAVIRT_PATCH(pv_irq_ops.save_fl)] = { lgstart_pushf, lgend_pushf },
};

/* Now our patch routine is fairly simple (based on the native one in
 * paravirt.c).  If we have a replacement, we copy it in and return how much of
 * the available space we used. */
static unsigned lguest_patch(u8 type, u16 clobber, void *ibuf,
			     unsigned long addr, unsigned len)
{
	unsigned int insn_len;

	/* Don't do anything special if we don't have a replacement */
	if (type >= ARRAY_SIZE(lguest_insns) || !lguest_insns[type].start)
		return paravirt_patch_default(type, clobber, ibuf, addr, len);

	insn_len = lguest_insns[type].end - lguest_insns[type].start;

	/* Similarly if we can't fit replacement (shouldn't happen, but let's
	 * be thorough). */
	if (len < insn_len)
		return paravirt_patch_default(type, clobber, ibuf, addr, len);

	/* Copy in our instructions. */
	memcpy(ibuf, lguest_insns[type].start, insn_len);
	return insn_len;
}

/*G:030 Once we get to lguest_init(), we know we're a Guest.  The pv_ops
 * structures in the kernel provide points for (almost) every routine we have
 * to override to avoid privileged instructions. */
__init void lguest_init(void)
{
	/* We're under lguest, paravirt is enabled, and we're running at
	 * privilege level 1, not 0 as normal. */
	pv_info.name = "lguest";
	pv_info.paravirt_enabled = 1;
	pv_info.kernel_rpl = 1;

	/* We set up all the lguest overrides for sensitive operations.  These
	 * are detailed with the operations themselves. */

	/* interrupt-related operations */
	pv_irq_ops.init_IRQ = lguest_init_IRQ;
	pv_irq_ops.save_fl = save_fl;
	pv_irq_ops.restore_fl = restore_fl;
	pv_irq_ops.irq_disable = irq_disable;
	pv_irq_ops.irq_enable = irq_enable;
	pv_irq_ops.safe_halt = lguest_safe_halt;

	/* init-time operations */
	pv_init_ops.memory_setup = lguest_memory_setup;
	pv_init_ops.patch = lguest_patch;

	/* Intercepts of various cpu instructions */
	pv_cpu_ops.load_gdt = lguest_load_gdt;
	pv_cpu_ops.cpuid = lguest_cpuid;
	pv_cpu_ops.load_idt = lguest_load_idt;
	pv_cpu_ops.iret = lguest_iret;
	pv_cpu_ops.load_esp0 = lguest_load_esp0;
	pv_cpu_ops.load_tr_desc = lguest_load_tr_desc;
	pv_cpu_ops.set_ldt = lguest_set_ldt;
	pv_cpu_ops.load_tls = lguest_load_tls;
	pv_cpu_ops.set_debugreg = lguest_set_debugreg;
	pv_cpu_ops.clts = lguest_clts;
	pv_cpu_ops.read_cr0 = lguest_read_cr0;
	pv_cpu_ops.write_cr0 = lguest_write_cr0;
	pv_cpu_ops.read_cr4 = lguest_read_cr4;
	pv_cpu_ops.write_cr4 = lguest_write_cr4;
	pv_cpu_ops.write_gdt_entry = lguest_write_gdt_entry;
	pv_cpu_ops.write_idt_entry = lguest_write_idt_entry;
	pv_cpu_ops.wbinvd = lguest_wbinvd;
	pv_cpu_ops.lazy_mode.enter = paravirt_enter_lazy_cpu;
	pv_cpu_ops.lazy_mode.leave = lguest_leave_lazy_mode;

	/* pagetable management */
	pv_mmu_ops.write_cr3 = lguest_write_cr3;
	pv_mmu_ops.flush_tlb_user = lguest_flush_tlb_user;
	pv_mmu_ops.flush_tlb_single = lguest_flush_tlb_single;
	pv_mmu_ops.flush_tlb_kernel = lguest_flush_tlb_kernel;
	pv_mmu_ops.set_pte = lguest_set_pte;
	pv_mmu_ops.set_pte_at = lguest_set_pte_at;
	pv_mmu_ops.set_pmd = lguest_set_pmd;
	pv_mmu_ops.read_cr2 = lguest_read_cr2;
	pv_mmu_ops.read_cr3 = lguest_read_cr3;
	pv_mmu_ops.lazy_mode.enter = paravirt_enter_lazy_mmu;
	pv_mmu_ops.lazy_mode.leave = lguest_leave_lazy_mode;

#ifdef CONFIG_X86_LOCAL_APIC
	/* apic read/write intercepts */
	pv_apic_ops.apic_write = lguest_apic_write;
	pv_apic_ops.apic_write_atomic = lguest_apic_write;
	pv_apic_ops.apic_read = lguest_apic_read;
#endif

	/* time operations */
	pv_time_ops.get_wallclock = lguest_get_wallclock;
	pv_time_ops.time_init = lguest_time_init;

	/* Now is a good time to look at the implementations of these functions
	 * before returning to the rest of lguest_init(). */

	/*G:070 Now we've seen all the paravirt_ops, we return to
	 * lguest_init() where the rest of the fairly chaotic boot setup
	 * occurs. */

	/* The native boot code sets up initial page tables immediately after
	 * the kernel itself, and sets init_pg_tables_end so they're not
	 * clobbered.  The Launcher places our initial pagetables somewhere at
	 * the top of our physical memory, so we don't need extra space: set
	 * init_pg_tables_end to the end of the kernel. */
	init_pg_tables_end = __pa(pg0);

	/* Load the %fs segment register (the per-cpu segment register) with
	 * the normal data segment to get through booting. */
	asm volatile ("mov %0, %%fs" : : "r" (__KERNEL_DS) : "memory");

	/* The Host uses the top of the Guest's virtual address space for the
	 * Host<->Guest Switcher, and it tells us how big that is in
	 * lguest_data.reserve_mem, set up on the LGUEST_INIT hypercall. */
	reserve_top_address(lguest_data.reserve_mem);

	/* If we don't initialize the lock dependency checker now, it crashes
	 * paravirt_disable_iospace. */
	lockdep_init();

	/* The IDE code spends about 3 seconds probing for disks: if we reserve
	 * all the I/O ports up front it can't get them and so doesn't probe.
	 * Other device drivers are similar (but less severe).  This cuts the
	 * kernel boot time on my machine from 4.1 seconds to 0.45 seconds. */
	paravirt_disable_iospace();

	/* This is messy CPU setup stuff which the native boot code does before
	 * start_kernel, so we have to do, too: */
	cpu_detect(&new_cpu_data);
	/* head.S usually sets up the first capability word, so do it here. */
	new_cpu_data.x86_capability[0] = cpuid_edx(1);

	/* Math is always hard! */
	new_cpu_data.hard_math = 1;

#ifdef CONFIG_X86_MCE
	mce_disabled = 1;
#endif
#ifdef CONFIG_ACPI
	acpi_disabled = 1;
	acpi_ht = 0;
#endif

	/* We set the perferred console to "hvc".  This is the "hypervisor
	 * virtual console" driver written by the PowerPC people, which we also
	 * adapted for lguest's use. */
	add_preferred_console("hvc", 0, NULL);

	/* Register our very early console. */
	virtio_cons_early_init(early_put_chars);

	/* Last of all, we set the power management poweroff hook to point to
	 * the Guest routine to power off. */
	pm_power_off = lguest_power_off;

	/* Now we're set up, call start_kernel() in init/main.c and we proceed
	 * to boot as normal.  It never returns. */
	start_kernel();
}
/*
 * This marks the end of stage II of our journey, The Guest.
 *
 * It is now time for us to explore the layer of virtual drivers and complete
 * our understanding of the Guest in "make Drivers".
 */