2007-07-19 08:49:23 +00:00
|
|
|
#ifndef _LGUEST_H
|
|
|
|
#define _LGUEST_H
|
|
|
|
|
|
|
|
#ifndef __ASSEMBLY__
|
|
|
|
#include <linux/types.h>
|
|
|
|
#include <linux/init.h>
|
|
|
|
#include <linux/stringify.h>
|
|
|
|
#include <linux/lguest.h>
|
|
|
|
#include <linux/lguest_launcher.h>
|
|
|
|
#include <linux/wait.h>
|
|
|
|
#include <linux/err.h>
|
|
|
|
#include <asm/semaphore.h>
|
|
|
|
|
2007-10-22 01:03:28 +00:00
|
|
|
#include <asm/lguest.h>
|
2007-07-19 08:49:23 +00:00
|
|
|
|
|
|
|
void free_pagetables(void);
|
|
|
|
int init_pagetables(struct page **switcher_page, unsigned int pages);
|
|
|
|
|
|
|
|
struct pgdir
|
|
|
|
{
|
2007-10-22 01:03:34 +00:00
|
|
|
unsigned long gpgdir;
|
2007-10-22 01:03:33 +00:00
|
|
|
pgd_t *pgdir;
|
2007-07-19 08:49:23 +00:00
|
|
|
};
|
|
|
|
|
|
|
|
/* We have two pages shared with guests, per cpu. */
|
|
|
|
struct lguest_pages
|
|
|
|
{
|
|
|
|
/* This is the stack page mapped rw in guest */
|
|
|
|
char spare[PAGE_SIZE - sizeof(struct lguest_regs)];
|
|
|
|
struct lguest_regs regs;
|
|
|
|
|
|
|
|
/* This is the host state & guest descriptor page, ro in guest */
|
|
|
|
struct lguest_ro_state state;
|
|
|
|
} __attribute__((aligned(PAGE_SIZE)));
|
|
|
|
|
|
|
|
#define CHANGED_IDT 1
|
|
|
|
#define CHANGED_GDT 2
|
|
|
|
#define CHANGED_GDT_TLS 4 /* Actually a subset of CHANGED_GDT */
|
|
|
|
#define CHANGED_ALL 3
|
|
|
|
|
2008-01-07 13:05:22 +00:00
|
|
|
struct lguest;
|
|
|
|
|
|
|
|
struct lg_cpu {
|
|
|
|
unsigned int id;
|
|
|
|
struct lguest *lg;
|
2008-01-07 13:05:34 +00:00
|
|
|
struct task_struct *tsk;
|
|
|
|
struct mm_struct *mm; /* == tsk->mm, but that becomes NULL on exit */
|
2008-01-07 13:05:27 +00:00
|
|
|
|
2008-01-07 13:05:35 +00:00
|
|
|
u32 cr2;
|
|
|
|
int ts;
|
|
|
|
u32 esp1;
|
|
|
|
u8 ss1;
|
|
|
|
|
2008-01-07 13:05:36 +00:00
|
|
|
unsigned long pending_notify; /* pfn from LHCALL_NOTIFY */
|
|
|
|
|
2008-01-07 13:05:32 +00:00
|
|
|
/* At end of a page shared mapped over lguest_pages in guest. */
|
|
|
|
unsigned long regs_page;
|
|
|
|
struct lguest_regs *regs;
|
|
|
|
|
2008-01-07 13:05:37 +00:00
|
|
|
int cpu_pgd; /* which pgd this cpu is currently using */
|
|
|
|
|
2008-01-07 13:05:27 +00:00
|
|
|
/* If a hypercall was asked for, this points to the arguments. */
|
|
|
|
struct hcall_args *hcall;
|
|
|
|
u32 next_hcall;
|
2008-01-07 13:05:28 +00:00
|
|
|
|
|
|
|
/* Virtual clock device */
|
|
|
|
struct hrtimer hrt;
|
2008-01-07 13:05:29 +00:00
|
|
|
|
2008-01-07 13:05:34 +00:00
|
|
|
/* Do we need to stop what we're doing and return to userspace? */
|
|
|
|
int break_out;
|
|
|
|
wait_queue_head_t break_wq;
|
|
|
|
int halted;
|
|
|
|
|
2008-01-07 13:05:29 +00:00
|
|
|
/* Pending virtual interrupts */
|
|
|
|
DECLARE_BITMAP(irqs_pending, LGUEST_IRQS);
|
2008-01-07 13:05:33 +00:00
|
|
|
|
|
|
|
struct lg_cpu_arch arch;
|
2008-01-07 13:05:22 +00:00
|
|
|
};
|
|
|
|
|
2007-07-19 08:49:23 +00:00
|
|
|
/* The private info the thread maintains about the guest. */
|
|
|
|
struct lguest
|
|
|
|
{
|
|
|
|
struct lguest_data __user *lguest_data;
|
2008-01-07 13:05:22 +00:00
|
|
|
struct lg_cpu cpus[NR_CPUS];
|
|
|
|
unsigned int nr_cpus;
|
|
|
|
|
2007-07-19 08:49:23 +00:00
|
|
|
u32 pfn_limit;
|
2007-10-22 01:03:26 +00:00
|
|
|
/* This provides the offset to the base of guest-physical
|
|
|
|
* memory in the Launcher. */
|
|
|
|
void __user *mem_base;
|
2007-10-22 01:03:36 +00:00
|
|
|
unsigned long kernel_address;
|
2007-07-19 08:49:23 +00:00
|
|
|
|
|
|
|
/* Bitmap of what has changed: see CHANGED_* above. */
|
|
|
|
int changed;
|
|
|
|
struct lguest_pages *last_pages;
|
|
|
|
|
|
|
|
struct pgdir pgdirs[4];
|
|
|
|
|
|
|
|
unsigned long noirq_start, noirq_end;
|
|
|
|
|
|
|
|
unsigned int stack_pages;
|
|
|
|
u32 tsc_khz;
|
|
|
|
|
|
|
|
/* Dead? */
|
|
|
|
const char *dead;
|
|
|
|
};
|
|
|
|
|
|
|
|
extern struct mutex lguest_lock;
|
|
|
|
|
|
|
|
/* core.c: */
|
|
|
|
int lguest_address_ok(const struct lguest *lg,
|
|
|
|
unsigned long addr, unsigned long len);
|
2007-10-22 01:24:24 +00:00
|
|
|
void __lgread(struct lguest *, void *, unsigned long, unsigned);
|
|
|
|
void __lgwrite(struct lguest *, unsigned long, const void *, unsigned);
|
|
|
|
|
2007-10-25 05:02:50 +00:00
|
|
|
/*H:035 Using memory-copy operations like that is usually inconvient, so we
|
2007-10-22 01:24:24 +00:00
|
|
|
* have the following helper macros which read and write a specific type (often
|
|
|
|
* an unsigned long).
|
|
|
|
*
|
|
|
|
* This reads into a variable of the given type then returns that. */
|
|
|
|
#define lgread(lg, addr, type) \
|
|
|
|
({ type _v; __lgread((lg), &_v, (addr), sizeof(_v)); _v; })
|
|
|
|
|
|
|
|
/* This checks that the variable is of the given type, then writes it out. */
|
|
|
|
#define lgwrite(lg, addr, type, val) \
|
|
|
|
do { \
|
|
|
|
typecheck(type, val); \
|
|
|
|
__lgwrite((lg), (addr), &(val), sizeof(val)); \
|
|
|
|
} while(0)
|
|
|
|
/* (end of memory access helper routines) :*/
|
|
|
|
|
2008-01-07 13:05:25 +00:00
|
|
|
int run_guest(struct lg_cpu *cpu, unsigned long __user *user);
|
2007-07-19 08:49:23 +00:00
|
|
|
|
2007-10-22 01:03:33 +00:00
|
|
|
/* Helper macros to obtain the first 12 or the last 20 bits, this is only the
|
|
|
|
* first step in the migration to the kernel types. pte_pfn is already defined
|
|
|
|
* in the kernel. */
|
|
|
|
#define pgd_flags(x) (pgd_val(x) & ~PAGE_MASK)
|
|
|
|
#define pte_flags(x) (pte_val(x) & ~PAGE_MASK)
|
|
|
|
#define pgd_pfn(x) (pgd_val(x) >> PAGE_SHIFT)
|
2007-07-19 08:49:23 +00:00
|
|
|
|
|
|
|
/* interrupts_and_traps.c: */
|
2008-01-07 13:05:29 +00:00
|
|
|
void maybe_do_interrupt(struct lg_cpu *cpu);
|
|
|
|
int deliver_trap(struct lg_cpu *cpu, unsigned int num);
|
2008-01-07 13:05:33 +00:00
|
|
|
void load_guest_idt_entry(struct lg_cpu *cpu, unsigned int i,
|
|
|
|
u32 low, u32 hi);
|
2008-01-07 13:05:35 +00:00
|
|
|
void guest_set_stack(struct lg_cpu *cpu, u32 seg, u32 esp, unsigned int pages);
|
|
|
|
void pin_stack_pages(struct lg_cpu *cpu);
|
2007-07-19 08:49:23 +00:00
|
|
|
void setup_default_idt_entries(struct lguest_ro_state *state,
|
|
|
|
const unsigned long *def);
|
2008-01-07 13:05:33 +00:00
|
|
|
void copy_traps(const struct lg_cpu *cpu, struct desc_struct *idt,
|
2007-07-19 08:49:23 +00:00
|
|
|
const unsigned long *def);
|
2008-01-07 13:05:28 +00:00
|
|
|
void guest_set_clockevent(struct lg_cpu *cpu, unsigned long delta);
|
|
|
|
void init_clockdev(struct lg_cpu *cpu);
|
2007-10-22 01:03:35 +00:00
|
|
|
bool check_syscall_vector(struct lguest *lg);
|
|
|
|
int init_interrupts(void);
|
|
|
|
void free_interrupts(void);
|
2007-07-19 08:49:23 +00:00
|
|
|
|
|
|
|
/* segments.c: */
|
|
|
|
void setup_default_gdt_entries(struct lguest_ro_state *state);
|
2008-01-07 13:05:33 +00:00
|
|
|
void setup_guest_gdt(struct lg_cpu *cpu);
|
|
|
|
void load_guest_gdt(struct lg_cpu *cpu, unsigned long table, u32 num);
|
|
|
|
void guest_load_tls(struct lg_cpu *cpu, unsigned long tls_array);
|
|
|
|
void copy_gdt(const struct lg_cpu *cpu, struct desc_struct *gdt);
|
|
|
|
void copy_gdt_tls(const struct lg_cpu *cpu, struct desc_struct *gdt);
|
2007-07-19 08:49:23 +00:00
|
|
|
|
|
|
|
/* page_tables.c: */
|
|
|
|
int init_guest_pagetable(struct lguest *lg, unsigned long pgtable);
|
|
|
|
void free_guest_pagetable(struct lguest *lg);
|
2008-01-07 13:05:35 +00:00
|
|
|
void guest_new_pagetable(struct lg_cpu *cpu, unsigned long pgtable);
|
2007-10-22 01:03:34 +00:00
|
|
|
void guest_set_pmd(struct lguest *lg, unsigned long gpgdir, u32 i);
|
2008-01-07 13:05:35 +00:00
|
|
|
void guest_pagetable_clear_all(struct lg_cpu *cpu);
|
2008-01-07 13:05:37 +00:00
|
|
|
void guest_pagetable_flush_user(struct lg_cpu *cpu);
|
2007-10-22 01:03:34 +00:00
|
|
|
void guest_set_pte(struct lguest *lg, unsigned long gpgdir,
|
2007-10-22 01:03:33 +00:00
|
|
|
unsigned long vaddr, pte_t val);
|
2008-01-07 13:05:30 +00:00
|
|
|
void map_switcher_in_guest(struct lg_cpu *cpu, struct lguest_pages *pages);
|
2008-01-07 13:05:37 +00:00
|
|
|
int demand_page(struct lg_cpu *cpu, unsigned long cr2, int errcode);
|
|
|
|
void pin_page(struct lg_cpu *cpu, unsigned long vaddr);
|
|
|
|
unsigned long guest_pa(struct lg_cpu *cpu, unsigned long vaddr);
|
2007-10-22 01:03:36 +00:00
|
|
|
void page_table_guest_data_init(struct lguest *lg);
|
2007-07-19 08:49:23 +00:00
|
|
|
|
2007-10-22 01:03:28 +00:00
|
|
|
/* <arch>/core.c: */
|
|
|
|
void lguest_arch_host_init(void);
|
|
|
|
void lguest_arch_host_fini(void);
|
2008-01-07 13:05:25 +00:00
|
|
|
void lguest_arch_run_guest(struct lg_cpu *cpu);
|
2008-01-07 13:05:27 +00:00
|
|
|
void lguest_arch_handle_trap(struct lg_cpu *cpu);
|
|
|
|
int lguest_arch_init_hypercalls(struct lg_cpu *cpu);
|
|
|
|
int lguest_arch_do_hcall(struct lg_cpu *cpu, struct hcall_args *args);
|
2008-01-07 13:05:32 +00:00
|
|
|
void lguest_arch_setup_regs(struct lg_cpu *cpu, unsigned long start);
|
2007-10-22 01:03:28 +00:00
|
|
|
|
|
|
|
/* <arch>/switcher.S: */
|
|
|
|
extern char start_switcher_text[], end_switcher_text[], switch_to_guest[];
|
|
|
|
|
2007-07-19 08:49:23 +00:00
|
|
|
/* lguest_user.c: */
|
|
|
|
int lguest_device_init(void);
|
|
|
|
void lguest_device_remove(void);
|
|
|
|
|
|
|
|
/* hypercalls.c: */
|
2008-01-07 13:05:27 +00:00
|
|
|
void do_hypercalls(struct lg_cpu *cpu);
|
2007-07-27 03:42:52 +00:00
|
|
|
void write_timestamp(struct lguest *lg);
|
2007-07-19 08:49:23 +00:00
|
|
|
|
2007-07-26 17:41:03 +00:00
|
|
|
/*L:035
|
|
|
|
* Let's step aside for the moment, to study one important routine that's used
|
|
|
|
* widely in the Host code.
|
|
|
|
*
|
2007-10-25 05:02:50 +00:00
|
|
|
* There are many cases where the Guest can do something invalid, like pass crap
|
2007-07-26 17:41:03 +00:00
|
|
|
* to a hypercall. Since only the Guest kernel can make hypercalls, it's quite
|
|
|
|
* acceptable to simply terminate the Guest and give the Launcher a nicely
|
|
|
|
* formatted reason. It's also simpler for the Guest itself, which doesn't
|
|
|
|
* need to check most hypercalls for "success"; if you're still running, it
|
|
|
|
* succeeded.
|
|
|
|
*
|
|
|
|
* Once this is called, the Guest will never run again, so most Host code can
|
|
|
|
* call this then continue as if nothing had happened. This means many
|
|
|
|
* functions don't have to explicitly return an error code, which keeps the
|
|
|
|
* code simple.
|
|
|
|
*
|
|
|
|
* It also means that this can be called more than once: only the first one is
|
|
|
|
* remembered. The only trick is that we still need to kill the Guest even if
|
|
|
|
* we can't allocate memory to store the reason. Linux has a neat way of
|
|
|
|
* packing error codes into invalid pointers, so we use that here.
|
|
|
|
*
|
|
|
|
* Like any macro which uses an "if", it is safely wrapped in a run-once "do {
|
|
|
|
* } while(0)".
|
|
|
|
*/
|
2007-07-19 08:49:23 +00:00
|
|
|
#define kill_guest(lg, fmt...) \
|
|
|
|
do { \
|
|
|
|
if (!(lg)->dead) { \
|
|
|
|
(lg)->dead = kasprintf(GFP_ATOMIC, fmt); \
|
|
|
|
if (!(lg)->dead) \
|
|
|
|
(lg)->dead = ERR_PTR(-ENOMEM); \
|
|
|
|
} \
|
|
|
|
} while(0)
|
2007-07-26 17:41:03 +00:00
|
|
|
/* (End of aside) :*/
|
2007-07-19 08:49:23 +00:00
|
|
|
|
|
|
|
#endif /* __ASSEMBLY__ */
|
|
|
|
#endif /* _LGUEST_H */
|