Commit | Line | Data |
---|---|---|
d7e28ffe RR |
1 | #ifndef _LGUEST_H |
2 | #define _LGUEST_H | |
3 | ||
d7e28ffe RR |
4 | #ifndef __ASSEMBLY__ |
5 | #include <linux/types.h> | |
6 | #include <linux/init.h> | |
7 | #include <linux/stringify.h> | |
d7e28ffe RR |
8 | #include <linux/lguest.h> |
9 | #include <linux/lguest_launcher.h> | |
10 | #include <linux/wait.h> | |
ca94f2bd | 11 | #include <linux/hrtimer.h> |
d7e28ffe | 12 | #include <linux/err.h> |
5a0e3ad6 | 13 | #include <linux/slab.h> |
d7e28ffe | 14 | |
625efab1 | 15 | #include <asm/lguest.h> |
d7e28ffe | 16 | |
1842f23c | 17 | struct pgdir { |
ee3db0f2 | 18 | unsigned long gpgdir; |
86935fc4 | 19 | bool switcher_mapped; |
6d0cda93 | 20 | int last_host_cpu; |
df29f43e | 21 | pgd_t *pgdir; |
d7e28ffe RR |
22 | }; |
23 | ||
d7e28ffe | 24 | /* We have two pages shared with guests, per cpu. */ |
1842f23c | 25 | struct lguest_pages { |
d7e28ffe RR |
26 | /* This is the stack page mapped rw in guest */ |
27 | char spare[PAGE_SIZE - sizeof(struct lguest_regs)]; | |
28 | struct lguest_regs regs; | |
29 | ||
30 | /* This is the host state & guest descriptor page, ro in guest */ | |
31 | struct lguest_ro_state state; | |
32 | } __attribute__((aligned(PAGE_SIZE))); | |
33 | ||
34 | #define CHANGED_IDT 1 | |
35 | #define CHANGED_GDT 2 | |
36 | #define CHANGED_GDT_TLS 4 /* Actually a subset of CHANGED_GDT */ | |
37 | #define CHANGED_ALL 3 | |
38 | ||
badb1e04 GOC |
39 | struct lg_cpu { |
40 | unsigned int id; | |
41 | struct lguest *lg; | |
66686c2a GOC |
42 | struct task_struct *tsk; |
43 | struct mm_struct *mm; /* == tsk->mm, but that becomes NULL on exit */ | |
73044f05 | 44 | |
4665ac8e GOC |
45 | u32 cr2; |
46 | int ts; | |
47 | u32 esp1; | |
f086122b | 48 | u16 ss1; |
4665ac8e | 49 | |
ae3749dc GOC |
50 | /* Bitmap of what has changed: see CHANGED_* above. */ |
51 | int changed; | |
52 | ||
69a09dc1 RR |
53 | /* Pending operation. */ |
54 | struct lguest_pending pending; | |
5e232f4f | 55 | |
18c13737 RR |
56 | unsigned long *reg_read; /* register from LHREQ_GETREG */ |
57 | ||
2e04ef76 | 58 | /* At end of a page shared mapped over lguest_pages in guest. */ |
a53a35a8 GOC |
59 | unsigned long regs_page; |
60 | struct lguest_regs *regs; | |
61 | ||
f34f8c5f GOC |
62 | struct lguest_pages *last_pages; |
63 | ||
5dea1c88 RR |
64 | /* Initialization mode: linear map everything. */ |
65 | bool linear_pages; | |
2e04ef76 | 66 | int cpu_pgd; /* Which pgd this cpu is currently using */ |
1713608f | 67 | |
73044f05 GOC |
68 | /* If a hypercall was asked for, this points to the arguments. */ |
69 | struct hcall_args *hcall; | |
70 | u32 next_hcall; | |
ad8d8f3b GOC |
71 | |
72 | /* Virtual clock device */ | |
73 | struct hrtimer hrt; | |
177e449d | 74 | |
5dac051b | 75 | /* Did the Guest tell us to halt? */ |
66686c2a GOC |
76 | int halted; |
77 | ||
177e449d GOC |
78 | /* Pending virtual interrupts */ |
79 | DECLARE_BITMAP(irqs_pending, LGUEST_IRQS); | |
fc708b3e GOC |
80 | |
81 | struct lg_cpu_arch arch; | |
badb1e04 GOC |
82 | }; |
83 | ||
d7e28ffe | 84 | /* The private info the thread maintains about the guest. */ |
1842f23c | 85 | struct lguest { |
d7e28ffe | 86 | struct lguest_data __user *lguest_data; |
badb1e04 GOC |
87 | struct lg_cpu cpus[NR_CPUS]; |
88 | unsigned int nr_cpus; | |
89 | ||
7313d521 | 90 | /* Valid guest memory pages must be < this. */ |
d7e28ffe | 91 | u32 pfn_limit; |
2e04ef76 | 92 | |
7313d521 RR |
93 | /* Device memory is >= pfn_limit and < device_limit. */ |
94 | u32 device_limit; | |
95 | ||
2e04ef76 RR |
96 | /* |
97 | * This provides the offset to the base of guest-physical memory in the | |
98 | * Launcher. | |
99 | */ | |
3c6b5bfa | 100 | void __user *mem_base; |
47436aa4 | 101 | unsigned long kernel_address; |
d7e28ffe | 102 | |
d7e28ffe RR |
103 | struct pgdir pgdirs[4]; |
104 | ||
2f921b5b | 105 | unsigned long noirq_iret; |
d7e28ffe RR |
106 | |
107 | unsigned int stack_pages; | |
108 | u32 tsc_khz; | |
109 | ||
d7e28ffe RR |
110 | /* Dead? */ |
111 | const char *dead; | |
d7e28ffe RR |
112 | }; |
113 | ||
d7e28ffe RR |
114 | extern struct mutex lguest_lock; |
115 | ||
116 | /* core.c: */ | |
df1693ab MZ |
117 | bool lguest_address_ok(const struct lguest *lg, |
118 | unsigned long addr, unsigned long len); | |
382ac6b3 GOC |
119 | void __lgread(struct lg_cpu *, void *, unsigned long, unsigned); |
120 | void __lgwrite(struct lg_cpu *, unsigned long, const void *, unsigned); | |
f1f394b1 | 121 | extern struct page **lg_switcher_pages; |
2d37f94a | 122 | |
2e04ef76 RR |
123 | /*H:035 |
124 | * Using memory-copy operations like that is usually inconvient, so we | |
2d37f94a RR |
125 | * have the following helper macros which read and write a specific type (often |
126 | * an unsigned long). | |
127 | * | |
2e04ef76 RR |
128 | * This reads into a variable of the given type then returns that. |
129 | */ | |
382ac6b3 GOC |
130 | #define lgread(cpu, addr, type) \ |
131 | ({ type _v; __lgread((cpu), &_v, (addr), sizeof(_v)); _v; }) | |
2d37f94a RR |
132 | |
133 | /* This checks that the variable is of the given type, then writes it out. */ | |
382ac6b3 | 134 | #define lgwrite(cpu, addr, type, val) \ |
2d37f94a RR |
135 | do { \ |
136 | typecheck(type, val); \ | |
382ac6b3 | 137 | __lgwrite((cpu), (addr), &(val), sizeof(val)); \ |
2d37f94a RR |
138 | } while(0) |
139 | /* (end of memory access helper routines) :*/ | |
140 | ||
d0953d42 | 141 | int run_guest(struct lg_cpu *cpu, unsigned long __user *user); |
d7e28ffe | 142 | |
2e04ef76 RR |
143 | /* |
144 | * Helper macros to obtain the first 12 or the last 20 bits, this is only the | |
df29f43e | 145 | * first step in the migration to the kernel types. pte_pfn is already defined |
2e04ef76 RR |
146 | * in the kernel. |
147 | */ | |
df29f43e | 148 | #define pgd_flags(x) (pgd_val(x) & ~PAGE_MASK) |
df29f43e | 149 | #define pgd_pfn(x) (pgd_val(x) >> PAGE_SHIFT) |
acdd0b62 MZ |
150 | #define pmd_flags(x) (pmd_val(x) & ~PAGE_MASK) |
151 | #define pmd_pfn(x) (pmd_val(x) >> PAGE_SHIFT) | |
d7e28ffe RR |
152 | |
153 | /* interrupts_and_traps.c: */ | |
a32a8813 RR |
154 | unsigned int interrupt_pending(struct lg_cpu *cpu, bool *more); |
155 | void try_deliver_interrupt(struct lg_cpu *cpu, unsigned int irq, bool more); | |
9f155a9b | 156 | void set_interrupt(struct lg_cpu *cpu, unsigned int irq); |
df1693ab | 157 | bool deliver_trap(struct lg_cpu *cpu, unsigned int num); |
fc708b3e GOC |
158 | void load_guest_idt_entry(struct lg_cpu *cpu, unsigned int i, |
159 | u32 low, u32 hi); | |
4665ac8e GOC |
160 | void guest_set_stack(struct lg_cpu *cpu, u32 seg, u32 esp, unsigned int pages); |
161 | void pin_stack_pages(struct lg_cpu *cpu); | |
d7e28ffe RR |
162 | void setup_default_idt_entries(struct lguest_ro_state *state, |
163 | const unsigned long *def); | |
fc708b3e | 164 | void copy_traps(const struct lg_cpu *cpu, struct desc_struct *idt, |
d7e28ffe | 165 | const unsigned long *def); |
ad8d8f3b | 166 | void guest_set_clockevent(struct lg_cpu *cpu, unsigned long delta); |
df60aeef | 167 | bool send_notify_to_eventfd(struct lg_cpu *cpu); |
ad8d8f3b | 168 | void init_clockdev(struct lg_cpu *cpu); |
c18acd73 | 169 | bool check_syscall_vector(struct lguest *lg); |
f87e0434 | 170 | bool could_be_syscall(unsigned int num); |
c18acd73 RR |
171 | int init_interrupts(void); |
172 | void free_interrupts(void); | |
d7e28ffe RR |
173 | |
174 | /* segments.c: */ | |
175 | void setup_default_gdt_entries(struct lguest_ro_state *state); | |
fc708b3e | 176 | void setup_guest_gdt(struct lg_cpu *cpu); |
a489f0b5 RR |
177 | void load_guest_gdt_entry(struct lg_cpu *cpu, unsigned int i, |
178 | u32 low, u32 hi); | |
fc708b3e GOC |
179 | void guest_load_tls(struct lg_cpu *cpu, unsigned long tls_array); |
180 | void copy_gdt(const struct lg_cpu *cpu, struct desc_struct *gdt); | |
181 | void copy_gdt_tls(const struct lg_cpu *cpu, struct desc_struct *gdt); | |
d7e28ffe RR |
182 | |
183 | /* page_tables.c: */ | |
58a24566 | 184 | int init_guest_pagetable(struct lguest *lg); |
d7e28ffe | 185 | void free_guest_pagetable(struct lguest *lg); |
4665ac8e | 186 | void guest_new_pagetable(struct lg_cpu *cpu, unsigned long pgtable); |
ebe0ba84 | 187 | void guest_set_pgd(struct lguest *lg, unsigned long gpgdir, u32 i); |
acdd0b62 MZ |
188 | #ifdef CONFIG_X86_PAE |
189 | void guest_set_pmd(struct lguest *lg, unsigned long gpgdir, u32 i); | |
190 | #endif | |
4665ac8e | 191 | void guest_pagetable_clear_all(struct lg_cpu *cpu); |
1713608f | 192 | void guest_pagetable_flush_user(struct lg_cpu *cpu); |
382ac6b3 | 193 | void guest_set_pte(struct lg_cpu *cpu, unsigned long gpgdir, |
df29f43e | 194 | unsigned long vaddr, pte_t val); |
0c78441c | 195 | void map_switcher_in_guest(struct lg_cpu *cpu, struct lguest_pages *pages); |
7313d521 RR |
196 | bool demand_page(struct lg_cpu *cpu, unsigned long cr2, int errcode, |
197 | unsigned long *iomem); | |
1713608f | 198 | void pin_page(struct lg_cpu *cpu, unsigned long vaddr); |
c9e433e4 | 199 | bool __guest_pa(struct lg_cpu *cpu, unsigned long vaddr, unsigned long *paddr); |
1713608f | 200 | unsigned long guest_pa(struct lg_cpu *cpu, unsigned long vaddr); |
382ac6b3 | 201 | void page_table_guest_data_init(struct lg_cpu *cpu); |
d7e28ffe | 202 | |
625efab1 JS |
203 | /* <arch>/core.c: */ |
204 | void lguest_arch_host_init(void); | |
205 | void lguest_arch_host_fini(void); | |
d0953d42 | 206 | void lguest_arch_run_guest(struct lg_cpu *cpu); |
73044f05 GOC |
207 | void lguest_arch_handle_trap(struct lg_cpu *cpu); |
208 | int lguest_arch_init_hypercalls(struct lg_cpu *cpu); | |
209 | int lguest_arch_do_hcall(struct lg_cpu *cpu, struct hcall_args *args); | |
a53a35a8 | 210 | void lguest_arch_setup_regs(struct lg_cpu *cpu, unsigned long start); |
18c13737 | 211 | unsigned long *lguest_arch_regptr(struct lg_cpu *cpu, size_t reg_off, bool any); |
625efab1 JS |
212 | |
213 | /* <arch>/switcher.S: */ | |
214 | extern char start_switcher_text[], end_switcher_text[], switch_to_guest[]; | |
215 | ||
d7e28ffe RR |
216 | /* lguest_user.c: */ |
217 | int lguest_device_init(void); | |
218 | void lguest_device_remove(void); | |
219 | ||
d7e28ffe | 220 | /* hypercalls.c: */ |
73044f05 | 221 | void do_hypercalls(struct lg_cpu *cpu); |
382ac6b3 | 222 | void write_timestamp(struct lg_cpu *cpu); |
d7e28ffe | 223 | |
dde79789 RR |
224 | /*L:035 |
225 | * Let's step aside for the moment, to study one important routine that's used | |
226 | * widely in the Host code. | |
227 | * | |
e1e72965 | 228 | * There are many cases where the Guest can do something invalid, like pass crap |
dde79789 RR |
229 | * to a hypercall. Since only the Guest kernel can make hypercalls, it's quite |
230 | * acceptable to simply terminate the Guest and give the Launcher a nicely | |
231 | * formatted reason. It's also simpler for the Guest itself, which doesn't | |
232 | * need to check most hypercalls for "success"; if you're still running, it | |
233 | * succeeded. | |
234 | * | |
235 | * Once this is called, the Guest will never run again, so most Host code can | |
236 | * call this then continue as if nothing had happened. This means many | |
237 | * functions don't have to explicitly return an error code, which keeps the | |
238 | * code simple. | |
239 | * | |
240 | * It also means that this can be called more than once: only the first one is | |
241 | * remembered. The only trick is that we still need to kill the Guest even if | |
242 | * we can't allocate memory to store the reason. Linux has a neat way of | |
243 | * packing error codes into invalid pointers, so we use that here. | |
244 | * | |
245 | * Like any macro which uses an "if", it is safely wrapped in a run-once "do { | |
246 | * } while(0)". | |
247 | */ | |
382ac6b3 | 248 | #define kill_guest(cpu, fmt...) \ |
d7e28ffe | 249 | do { \ |
382ac6b3 GOC |
250 | if (!(cpu)->lg->dead) { \ |
251 | (cpu)->lg->dead = kasprintf(GFP_ATOMIC, fmt); \ | |
252 | if (!(cpu)->lg->dead) \ | |
253 | (cpu)->lg->dead = ERR_PTR(-ENOMEM); \ | |
d7e28ffe RR |
254 | } \ |
255 | } while(0) | |
dde79789 | 256 | /* (End of aside) :*/ |
d7e28ffe | 257 | |
d7e28ffe RR |
258 | #endif /* __ASSEMBLY__ */ |
259 | #endif /* _LGUEST_H */ |