2007-07-26 13:41:02 -04:00
|
|
|
/*P:400 This contains run_guest() which actually calls into the Host<->Guest
|
|
|
|
* Switcher and analyzes the return, such as determining if the Guest wants the
|
2008-03-28 12:05:53 -04:00
|
|
|
* Host to do something. This file also contains useful helper routines. :*/
|
2007-07-19 04:49:23 -04:00
|
|
|
#include <linux/module.h>
|
|
|
|
#include <linux/stringify.h>
|
|
|
|
#include <linux/stddef.h>
|
|
|
|
#include <linux/io.h>
|
|
|
|
#include <linux/mm.h>
|
|
|
|
#include <linux/vmalloc.h>
|
|
|
|
#include <linux/cpu.h>
|
|
|
|
#include <linux/freezer.h>
|
2007-10-21 21:03:28 -04:00
|
|
|
#include <linux/highmem.h>
|
2007-07-19 04:49:23 -04:00
|
|
|
#include <asm/paravirt.h>
|
|
|
|
#include <asm/pgtable.h>
|
|
|
|
#include <asm/uaccess.h>
|
|
|
|
#include <asm/poll.h>
|
|
|
|
#include <asm/asm-offsets.h>
|
|
|
|
#include "lg.h"
|
|
|
|
|
|
|
|
|
|
|
|
static struct vm_struct *switcher_vma;
|
|
|
|
static struct page **switcher_page;
|
|
|
|
|
|
|
|
/* This One Big lock protects all inter-guest data structures. */
|
|
|
|
DEFINE_MUTEX(lguest_lock);
|
|
|
|
|
2007-07-26 13:41:04 -04:00
|
|
|
/*H:010 We need to set up the Switcher at a high virtual address. Remember the
|
|
|
|
* Switcher is a few hundred bytes of assembler code which actually changes the
|
|
|
|
* CPU to run the Guest, and then changes back to the Host when a trap or
|
|
|
|
* interrupt happens.
|
|
|
|
*
|
|
|
|
* The Switcher code must be at the same virtual address in the Guest as the
|
|
|
|
* Host since it will be running as the switchover occurs.
|
|
|
|
*
|
|
|
|
* Trying to map memory at a particular address is an unusual thing to do, so
|
2007-10-21 21:03:28 -04:00
|
|
|
* it's not a simple one-liner. */
|
2007-07-19 04:49:23 -04:00
|
|
|
static __init int map_switcher(void)
|
|
|
|
{
|
|
|
|
int i, err;
|
|
|
|
struct page **pagep;
|
|
|
|
|
2007-07-26 13:41:04 -04:00
|
|
|
/*
|
|
|
|
* Map the Switcher in to high memory.
|
|
|
|
*
|
|
|
|
* It turns out that if we choose the address 0xFFC00000 (4MB under the
|
|
|
|
* top virtual address), it makes setting up the page tables really
|
|
|
|
* easy.
|
|
|
|
*/
|
|
|
|
|
2008-03-28 12:05:53 -04:00
|
|
|
/* We allocate an array of struct page pointers. map_vm_area() wants
|
|
|
|
* this, rather than just an array of pages. */
|
2007-07-19 04:49:23 -04:00
|
|
|
switcher_page = kmalloc(sizeof(switcher_page[0])*TOTAL_SWITCHER_PAGES,
|
|
|
|
GFP_KERNEL);
|
|
|
|
if (!switcher_page) {
|
|
|
|
err = -ENOMEM;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
2007-07-26 13:41:04 -04:00
|
|
|
/* Now we actually allocate the pages. The Guest will see these pages,
|
|
|
|
* so we make sure they're zeroed. */
|
2007-07-19 04:49:23 -04:00
|
|
|
for (i = 0; i < TOTAL_SWITCHER_PAGES; i++) {
|
|
|
|
unsigned long addr = get_zeroed_page(GFP_KERNEL);
|
|
|
|
if (!addr) {
|
|
|
|
err = -ENOMEM;
|
|
|
|
goto free_some_pages;
|
|
|
|
}
|
|
|
|
switcher_page[i] = virt_to_page(addr);
|
|
|
|
}
|
|
|
|
|
2008-03-11 10:35:56 -04:00
|
|
|
/* First we check that the Switcher won't overlap the fixmap area at
|
|
|
|
* the top of memory. It's currently nowhere near, but it could have
|
|
|
|
* very strange effects if it ever happened. */
|
|
|
|
if (SWITCHER_ADDR + (TOTAL_SWITCHER_PAGES+1)*PAGE_SIZE > FIXADDR_START){
|
|
|
|
err = -ENOMEM;
|
|
|
|
printk("lguest: mapping switcher would thwack fixmap\n");
|
|
|
|
goto free_pages;
|
|
|
|
}
|
|
|
|
|
2007-07-26 13:41:04 -04:00
|
|
|
/* Now we reserve the "virtual memory area" we want: 0xFFC00000
|
|
|
|
* (SWITCHER_ADDR). We might not get it in theory, but in practice
|
2008-03-11 10:35:56 -04:00
|
|
|
* it's worked so far. The end address needs +1 because __get_vm_area
|
|
|
|
* allocates an extra guard page, so we need space for that. */
|
2007-07-19 04:49:23 -04:00
|
|
|
switcher_vma = __get_vm_area(TOTAL_SWITCHER_PAGES * PAGE_SIZE,
|
2008-03-11 10:35:56 -04:00
|
|
|
VM_ALLOC, SWITCHER_ADDR, SWITCHER_ADDR
|
|
|
|
+ (TOTAL_SWITCHER_PAGES+1) * PAGE_SIZE);
|
2007-07-19 04:49:23 -04:00
|
|
|
if (!switcher_vma) {
|
|
|
|
err = -ENOMEM;
|
|
|
|
printk("lguest: could not map switcher pages high\n");
|
|
|
|
goto free_pages;
|
|
|
|
}
|
|
|
|
|
2007-07-26 13:41:04 -04:00
|
|
|
/* This code actually sets up the pages we've allocated to appear at
|
|
|
|
* SWITCHER_ADDR. map_vm_area() takes the vma we allocated above, the
|
|
|
|
* kind of pages we're mapping (kernel pages), and a pointer to our
|
|
|
|
* array of struct pages. It increments that pointer, but we don't
|
|
|
|
* care. */
|
2007-07-19 04:49:23 -04:00
|
|
|
pagep = switcher_page;
|
|
|
|
err = map_vm_area(switcher_vma, PAGE_KERNEL, &pagep);
|
|
|
|
if (err) {
|
|
|
|
printk("lguest: map_vm_area failed: %i\n", err);
|
|
|
|
goto free_vma;
|
|
|
|
}
|
2007-07-26 13:41:04 -04:00
|
|
|
|
2007-10-21 21:03:28 -04:00
|
|
|
/* Now the Switcher is mapped at the right address, we can't fail!
|
|
|
|
* Copy in the compiled-in Switcher code (from <arch>_switcher.S). */
|
2007-07-19 04:49:23 -04:00
|
|
|
memcpy(switcher_vma->addr, start_switcher_text,
|
|
|
|
end_switcher_text - start_switcher_text);
|
|
|
|
|
|
|
|
printk(KERN_INFO "lguest: mapped switcher at %p\n",
|
|
|
|
switcher_vma->addr);
|
2007-07-26 13:41:04 -04:00
|
|
|
/* And we succeeded... */
|
2007-07-19 04:49:23 -04:00
|
|
|
return 0;
|
|
|
|
|
|
|
|
free_vma:
|
|
|
|
vunmap(switcher_vma->addr);
|
|
|
|
free_pages:
|
|
|
|
i = TOTAL_SWITCHER_PAGES;
|
|
|
|
free_some_pages:
|
|
|
|
for (--i; i >= 0; i--)
|
|
|
|
__free_pages(switcher_page[i], 0);
|
|
|
|
kfree(switcher_page);
|
|
|
|
out:
|
|
|
|
return err;
|
|
|
|
}
|
2007-07-26 13:41:04 -04:00
|
|
|
/*:*/
|
2007-07-19 04:49:23 -04:00
|
|
|
|
2007-07-26 13:41:04 -04:00
|
|
|
/* Cleaning up the mapping when the module is unloaded is almost...
|
|
|
|
* too easy. */
|
2007-07-19 04:49:23 -04:00
|
|
|
static void unmap_switcher(void)
|
|
|
|
{
|
|
|
|
unsigned int i;
|
|
|
|
|
2007-07-26 13:41:04 -04:00
|
|
|
/* vunmap() undoes *both* map_vm_area() and __get_vm_area(). */
|
2007-07-19 04:49:23 -04:00
|
|
|
vunmap(switcher_vma->addr);
|
2007-07-26 13:41:04 -04:00
|
|
|
/* Now we just need to free the pages we copied the switcher into */
|
2007-07-19 04:49:23 -04:00
|
|
|
for (i = 0; i < TOTAL_SWITCHER_PAGES; i++)
|
|
|
|
__free_pages(switcher_page[i], 0);
|
2008-07-08 04:29:42 -04:00
|
|
|
kfree(switcher_page);
|
2007-07-19 04:49:23 -04:00
|
|
|
}
|
|
|
|
|
2007-10-25 01:02:50 -04:00
|
|
|
/*H:032
|
2007-07-26 13:41:03 -04:00
|
|
|
* Dealing With Guest Memory.
|
|
|
|
*
|
2007-10-25 01:02:50 -04:00
|
|
|
* Before we go too much further into the Host, we need to grok the routines
|
|
|
|
* we use to deal with Guest memory.
|
|
|
|
*
|
2007-07-26 13:41:03 -04:00
|
|
|
* When the Guest gives us (what it thinks is) a physical address, we can use
|
2007-10-21 21:03:26 -04:00
|
|
|
* the normal copy_from_user() & copy_to_user() on the corresponding place in
|
|
|
|
* the memory region allocated by the Launcher.
|
2007-07-26 13:41:03 -04:00
|
|
|
*
|
|
|
|
* But we can't trust the Guest: it might be trying to access the Launcher
|
|
|
|
* code. We have to check that the range is below the pfn_limit the Launcher
|
|
|
|
* gave us. We have to make sure that addr + len doesn't give us a false
|
|
|
|
* positive by overflowing, too. */
|
2009-03-18 12:38:35 -04:00
|
|
|
bool lguest_address_ok(const struct lguest *lg,
|
|
|
|
unsigned long addr, unsigned long len)
|
2007-07-19 04:49:23 -04:00
|
|
|
{
|
|
|
|
return (addr+len) / PAGE_SIZE < lg->pfn_limit && (addr+len >= addr);
|
|
|
|
}
|
|
|
|
|
2007-10-21 21:24:24 -04:00
|
|
|
/* This routine copies memory from the Guest. Here we can see how useful the
|
|
|
|
* kill_lguest() routine we met in the Launcher can be: we return a random
|
|
|
|
* value (all zeroes) instead of needing to return an error. */
|
2008-01-17 16:19:42 -05:00
|
|
|
void __lgread(struct lg_cpu *cpu, void *b, unsigned long addr, unsigned bytes)
|
2007-07-19 04:49:23 -04:00
|
|
|
{
|
2008-01-17 16:19:42 -05:00
|
|
|
if (!lguest_address_ok(cpu->lg, addr, bytes)
|
|
|
|
|| copy_from_user(b, cpu->lg->mem_base + addr, bytes) != 0) {
|
2007-07-19 04:49:23 -04:00
|
|
|
/* copy_from_user should do this, but as we rely on it... */
|
|
|
|
memset(b, 0, bytes);
|
2008-01-17 16:19:42 -05:00
|
|
|
kill_guest(cpu, "bad read address %#lx len %u", addr, bytes);
|
2007-07-19 04:49:23 -04:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2008-03-28 12:05:53 -04:00
|
|
|
/* This is the write (copy into Guest) version. */
|
2008-01-17 16:19:42 -05:00
|
|
|
void __lgwrite(struct lg_cpu *cpu, unsigned long addr, const void *b,
|
2007-10-21 21:24:24 -04:00
|
|
|
unsigned bytes)
|
2007-07-19 04:49:23 -04:00
|
|
|
{
|
2008-01-17 16:19:42 -05:00
|
|
|
if (!lguest_address_ok(cpu->lg, addr, bytes)
|
|
|
|
|| copy_to_user(cpu->lg->mem_base + addr, b, bytes) != 0)
|
|
|
|
kill_guest(cpu, "bad write address %#lx len %u", addr, bytes);
|
2007-07-19 04:49:23 -04:00
|
|
|
}
|
2007-10-21 21:24:24 -04:00
|
|
|
/*:*/
|
2007-07-19 04:49:23 -04:00
|
|
|
|
2007-07-26 13:41:04 -04:00
|
|
|
/*H:030 Let's jump straight to the the main loop which runs the Guest.
|
|
|
|
* Remember, this is called by the Launcher reading /dev/lguest, and we keep
|
|
|
|
* going around and around until something interesting happens. */
|
2008-01-07 08:05:25 -05:00
|
|
|
int run_guest(struct lg_cpu *cpu, unsigned long __user *user)
|
2007-07-19 04:49:23 -04:00
|
|
|
{
|
2007-07-26 13:41:04 -04:00
|
|
|
/* We stop running once the Guest is dead. */
|
2008-01-17 16:19:42 -05:00
|
|
|
while (!cpu->lg->dead) {
|
2007-10-21 21:03:30 -04:00
|
|
|
/* First we run any hypercalls the Guest wants done. */
|
2008-01-07 08:05:27 -05:00
|
|
|
if (cpu->hcall)
|
|
|
|
do_hypercalls(cpu);
|
2007-10-21 21:03:30 -04:00
|
|
|
|
2007-10-21 21:24:10 -04:00
|
|
|
/* It's possible the Guest did a NOTIFY hypercall to the
|
2007-07-26 13:41:04 -04:00
|
|
|
* Launcher, in which case we return from the read() now. */
|
2008-01-07 08:05:36 -05:00
|
|
|
if (cpu->pending_notify) {
|
|
|
|
if (put_user(cpu->pending_notify, user))
|
2007-07-19 04:49:23 -04:00
|
|
|
return -EFAULT;
|
2008-01-07 08:05:36 -05:00
|
|
|
return sizeof(cpu->pending_notify);
|
2007-07-19 04:49:23 -04:00
|
|
|
}
|
|
|
|
|
2007-07-26 13:41:04 -04:00
|
|
|
/* Check for signals */
|
2007-07-19 04:49:23 -04:00
|
|
|
if (signal_pending(current))
|
|
|
|
return -ERESTARTSYS;
|
|
|
|
|
|
|
|
/* If Waker set break_out, return to Launcher. */
|
2008-01-07 08:05:34 -05:00
|
|
|
if (cpu->break_out)
|
2007-07-19 04:49:23 -04:00
|
|
|
return -EAGAIN;
|
|
|
|
|
2008-03-28 12:05:53 -04:00
|
|
|
/* Check if there are any interrupts which can be delivered now:
|
|
|
|
* if so, this sets up the hander to be executed when we next
|
|
|
|
* run the Guest. */
|
2008-01-07 08:05:29 -05:00
|
|
|
maybe_do_interrupt(cpu);
|
2007-07-19 04:49:23 -04:00
|
|
|
|
2007-07-26 13:41:04 -04:00
|
|
|
/* All long-lived kernel loops need to check with this horrible
|
|
|
|
* thing called the freezer. If the Host is trying to suspend,
|
|
|
|
* it stops us. */
|
2007-07-19 04:49:23 -04:00
|
|
|
try_to_freeze();
|
|
|
|
|
2007-07-26 13:41:04 -04:00
|
|
|
/* Just make absolutely sure the Guest is still alive. One of
|
|
|
|
* those hypercalls could have been fatal, for example. */
|
2008-01-17 16:19:42 -05:00
|
|
|
if (cpu->lg->dead)
|
2007-07-19 04:49:23 -04:00
|
|
|
break;
|
|
|
|
|
2007-07-26 13:41:04 -04:00
|
|
|
/* If the Guest asked to be stopped, we sleep. The Guest's
|
2009-01-16 06:39:14 -05:00
|
|
|
* clock timer or LHREQ_BREAK from the Waker will wake us. */
|
2008-01-07 08:05:34 -05:00
|
|
|
if (cpu->halted) {
|
2007-07-19 04:49:23 -04:00
|
|
|
set_current_state(TASK_INTERRUPTIBLE);
|
|
|
|
schedule();
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
2007-07-26 13:41:04 -04:00
|
|
|
/* OK, now we're ready to jump into the Guest. First we put up
|
|
|
|
* the "Do Not Disturb" sign: */
|
2007-07-19 04:49:23 -04:00
|
|
|
local_irq_disable();
|
|
|
|
|
2007-10-21 21:03:28 -04:00
|
|
|
/* Actually run the Guest until something happens. */
|
2008-01-07 08:05:25 -05:00
|
|
|
lguest_arch_run_guest(cpu);
|
2007-07-26 13:41:04 -04:00
|
|
|
|
|
|
|
/* Now we're ready to be interrupted or moved to other CPUs */
|
2007-07-19 04:49:23 -04:00
|
|
|
local_irq_enable();
|
|
|
|
|
2007-10-21 21:03:28 -04:00
|
|
|
/* Now we deal with whatever happened to the Guest. */
|
2008-01-07 08:05:27 -05:00
|
|
|
lguest_arch_handle_trap(cpu);
|
2007-07-19 04:49:23 -04:00
|
|
|
}
|
2007-10-21 21:03:28 -04:00
|
|
|
|
2008-03-28 12:05:53 -04:00
|
|
|
/* Special case: Guest is 'dead' but wants a reboot. */
|
2008-01-17 16:19:42 -05:00
|
|
|
if (cpu->lg->dead == ERR_PTR(-ERESTART))
|
2007-12-28 03:56:24 -05:00
|
|
|
return -ERESTART;
|
2008-03-28 12:05:53 -04:00
|
|
|
|
2007-07-26 13:41:04 -04:00
|
|
|
/* The Guest is dead => "No such file or directory" */
|
2007-07-19 04:49:23 -04:00
|
|
|
return -ENOENT;
|
|
|
|
}
|
|
|
|
|
2007-07-26 13:41:04 -04:00
|
|
|
/*H:000
|
|
|
|
* Welcome to the Host!
|
|
|
|
*
|
|
|
|
* By this point your brain has been tickled by the Guest code and numbed by
|
|
|
|
* the Launcher code; prepare for it to be stretched by the Host code. This is
|
|
|
|
* the heart. Let's begin at the initialization routine for the Host's lg
|
|
|
|
* module.
|
|
|
|
*/
|
2007-07-19 04:49:23 -04:00
|
|
|
static int __init init(void)
|
|
|
|
{
|
|
|
|
int err;
|
|
|
|
|
2007-07-26 13:41:04 -04:00
|
|
|
/* Lguest can't run under Xen, VMI or itself. It does Tricky Stuff. */
|
2007-07-19 04:49:23 -04:00
|
|
|
if (paravirt_enabled()) {
|
2008-01-17 19:32:50 -05:00
|
|
|
printk("lguest is afraid of being a guest\n");
|
2007-07-19 04:49:23 -04:00
|
|
|
return -EPERM;
|
|
|
|
}
|
|
|
|
|
2007-07-26 13:41:04 -04:00
|
|
|
/* First we put the Switcher up in very high virtual memory. */
|
2007-07-19 04:49:23 -04:00
|
|
|
err = map_switcher();
|
|
|
|
if (err)
|
2007-10-21 21:03:35 -04:00
|
|
|
goto out;
|
2007-07-19 04:49:23 -04:00
|
|
|
|
2007-07-26 13:41:04 -04:00
|
|
|
/* Now we set up the pagetable implementation for the Guests. */
|
2007-07-19 04:49:23 -04:00
|
|
|
err = init_pagetables(switcher_page, SHARED_SWITCHER_PAGES);
|
2007-10-21 21:03:35 -04:00
|
|
|
if (err)
|
|
|
|
goto unmap;
|
2007-07-26 13:41:04 -04:00
|
|
|
|
2007-10-21 21:03:35 -04:00
|
|
|
/* We might need to reserve an interrupt vector. */
|
|
|
|
err = init_interrupts();
|
|
|
|
if (err)
|
|
|
|
goto free_pgtables;
|
|
|
|
|
2007-07-26 13:41:04 -04:00
|
|
|
/* /dev/lguest needs to be registered. */
|
2007-07-19 04:49:23 -04:00
|
|
|
err = lguest_device_init();
|
2007-10-21 21:03:35 -04:00
|
|
|
if (err)
|
|
|
|
goto free_interrupts;
|
2007-07-26 13:41:04 -04:00
|
|
|
|
2007-10-21 21:03:28 -04:00
|
|
|
/* Finally we do some architecture-specific setup. */
|
|
|
|
lguest_arch_host_init();
|
2007-07-26 13:41:04 -04:00
|
|
|
|
|
|
|
/* All good! */
|
2007-07-19 04:49:23 -04:00
|
|
|
return 0;
|
2007-10-21 21:03:35 -04:00
|
|
|
|
|
|
|
free_interrupts:
|
|
|
|
free_interrupts();
|
|
|
|
free_pgtables:
|
|
|
|
free_pagetables();
|
|
|
|
unmap:
|
|
|
|
unmap_switcher();
|
|
|
|
out:
|
|
|
|
return err;
|
2007-07-19 04:49:23 -04:00
|
|
|
}
|
|
|
|
|
2007-07-26 13:41:04 -04:00
|
|
|
/* Cleaning up is just the same code, backwards. With a little French. */
|
2007-07-19 04:49:23 -04:00
|
|
|
static void __exit fini(void)
|
|
|
|
{
|
|
|
|
lguest_device_remove();
|
2007-10-21 21:03:35 -04:00
|
|
|
free_interrupts();
|
2007-07-19 04:49:23 -04:00
|
|
|
free_pagetables();
|
|
|
|
unmap_switcher();
|
2007-07-26 13:41:04 -04:00
|
|
|
|
2007-10-21 21:03:28 -04:00
|
|
|
lguest_arch_host_fini();
|
2007-07-19 04:49:23 -04:00
|
|
|
}
|
2007-10-21 21:03:28 -04:00
|
|
|
/*:*/
|
2007-07-19 04:49:23 -04:00
|
|
|
|
2007-07-26 13:41:04 -04:00
|
|
|
/* The Host side of lguest can be a module. This is a nice way for people to
|
|
|
|
* play with it. */
|
2007-07-19 04:49:23 -04:00
|
|
|
module_init(init);
|
|
|
|
module_exit(fini);
|
|
|
|
MODULE_LICENSE("GPL");
|
|
|
|
MODULE_AUTHOR("Rusty Russell <rusty@rustcorp.com.au>");
|