diff options
Diffstat (limited to 'drivers/lguest/core.c')
-rw-r--r-- | drivers/lguest/core.c | 114 |
1 files changed, 75 insertions, 39 deletions
diff --git a/drivers/lguest/core.c b/drivers/lguest/core.c index a6974e9b8ebf..cd058bc903ff 100644 --- a/drivers/lguest/core.c +++ b/drivers/lguest/core.c | |||
@@ -1,6 +1,8 @@ | |||
1 | /*P:400 This contains run_guest() which actually calls into the Host<->Guest | 1 | /*P:400 |
2 | * This contains run_guest() which actually calls into the Host<->Guest | ||
2 | * Switcher and analyzes the return, such as determining if the Guest wants the | 3 | * Switcher and analyzes the return, such as determining if the Guest wants the |
3 | * Host to do something. This file also contains useful helper routines. :*/ | 4 | * Host to do something. This file also contains useful helper routines. |
5 | :*/ | ||
4 | #include <linux/module.h> | 6 | #include <linux/module.h> |
5 | #include <linux/stringify.h> | 7 | #include <linux/stringify.h> |
6 | #include <linux/stddef.h> | 8 | #include <linux/stddef.h> |
@@ -24,7 +26,8 @@ static struct page **switcher_page; | |||
24 | /* This One Big lock protects all inter-guest data structures. */ | 26 | /* This One Big lock protects all inter-guest data structures. */ |
25 | DEFINE_MUTEX(lguest_lock); | 27 | DEFINE_MUTEX(lguest_lock); |
26 | 28 | ||
27 | /*H:010 We need to set up the Switcher at a high virtual address. Remember the | 29 | /*H:010 |
30 | * We need to set up the Switcher at a high virtual address. Remember the | ||
28 | * Switcher is a few hundred bytes of assembler code which actually changes the | 31 | * Switcher is a few hundred bytes of assembler code which actually changes the |
29 | * CPU to run the Guest, and then changes back to the Host when a trap or | 32 | * CPU to run the Guest, and then changes back to the Host when a trap or |
30 | * interrupt happens. | 33 | * interrupt happens. |
@@ -33,7 +36,8 @@ DEFINE_MUTEX(lguest_lock); | |||
33 | * Host since it will be running as the switchover occurs. | 36 | * Host since it will be running as the switchover occurs. |
34 | * | 37 | * |
35 | * Trying to map memory at a particular address is an unusual thing to do, so | 38 | * Trying to map memory at a particular address is an unusual thing to do, so |
36 | * it's not a simple one-liner. */ | 39 | * it's not a simple one-liner. |
40 | */ | ||
37 | static __init int map_switcher(void) | 41 | static __init int map_switcher(void) |
38 | { | 42 | { |
39 | int i, err; | 43 | int i, err; |
@@ -47,8 +51,10 @@ static __init int map_switcher(void) | |||
47 | * easy. | 51 | * easy. |
48 | */ | 52 | */ |
49 | 53 | ||
50 | /* We allocate an array of struct page pointers. map_vm_area() wants | 54 | /* |
51 | * this, rather than just an array of pages. */ | 55 | * We allocate an array of struct page pointers. map_vm_area() wants |
56 | * this, rather than just an array of pages. | ||
57 | */ | ||
52 | switcher_page = kmalloc(sizeof(switcher_page[0])*TOTAL_SWITCHER_PAGES, | 58 | switcher_page = kmalloc(sizeof(switcher_page[0])*TOTAL_SWITCHER_PAGES, |
53 | GFP_KERNEL); | 59 | GFP_KERNEL); |
54 | if (!switcher_page) { | 60 | if (!switcher_page) { |
@@ -56,8 +62,10 @@ static __init int map_switcher(void) | |||
56 | goto out; | 62 | goto out; |
57 | } | 63 | } |
58 | 64 | ||
59 | /* Now we actually allocate the pages. The Guest will see these pages, | 65 | /* |
60 | * so we make sure they're zeroed. */ | 66 | * Now we actually allocate the pages. The Guest will see these pages, |
67 | * so we make sure they're zeroed. | ||
68 | */ | ||
61 | for (i = 0; i < TOTAL_SWITCHER_PAGES; i++) { | 69 | for (i = 0; i < TOTAL_SWITCHER_PAGES; i++) { |
62 | unsigned long addr = get_zeroed_page(GFP_KERNEL); | 70 | unsigned long addr = get_zeroed_page(GFP_KERNEL); |
63 | if (!addr) { | 71 | if (!addr) { |
@@ -67,19 +75,23 @@ static __init int map_switcher(void) | |||
67 | switcher_page[i] = virt_to_page(addr); | 75 | switcher_page[i] = virt_to_page(addr); |
68 | } | 76 | } |
69 | 77 | ||
70 | /* First we check that the Switcher won't overlap the fixmap area at | 78 | /* |
79 | * First we check that the Switcher won't overlap the fixmap area at | ||
71 | * the top of memory. It's currently nowhere near, but it could have | 80 | * the top of memory. It's currently nowhere near, but it could have |
72 | * very strange effects if it ever happened. */ | 81 | * very strange effects if it ever happened. |
82 | */ | ||
73 | if (SWITCHER_ADDR + (TOTAL_SWITCHER_PAGES+1)*PAGE_SIZE > FIXADDR_START){ | 83 | if (SWITCHER_ADDR + (TOTAL_SWITCHER_PAGES+1)*PAGE_SIZE > FIXADDR_START){ |
74 | err = -ENOMEM; | 84 | err = -ENOMEM; |
75 | printk("lguest: mapping switcher would thwack fixmap\n"); | 85 | printk("lguest: mapping switcher would thwack fixmap\n"); |
76 | goto free_pages; | 86 | goto free_pages; |
77 | } | 87 | } |
78 | 88 | ||
79 | /* Now we reserve the "virtual memory area" we want: 0xFFC00000 | 89 | /* |
90 | * Now we reserve the "virtual memory area" we want: 0xFFC00000 | ||
80 | * (SWITCHER_ADDR). We might not get it in theory, but in practice | 91 | * (SWITCHER_ADDR). We might not get it in theory, but in practice |
81 | * it's worked so far. The end address needs +1 because __get_vm_area | 92 | * it's worked so far. The end address needs +1 because __get_vm_area |
82 | * allocates an extra guard page, so we need space for that. */ | 93 | * allocates an extra guard page, so we need space for that. |
94 | */ | ||
83 | switcher_vma = __get_vm_area(TOTAL_SWITCHER_PAGES * PAGE_SIZE, | 95 | switcher_vma = __get_vm_area(TOTAL_SWITCHER_PAGES * PAGE_SIZE, |
84 | VM_ALLOC, SWITCHER_ADDR, SWITCHER_ADDR | 96 | VM_ALLOC, SWITCHER_ADDR, SWITCHER_ADDR |
85 | + (TOTAL_SWITCHER_PAGES+1) * PAGE_SIZE); | 97 | + (TOTAL_SWITCHER_PAGES+1) * PAGE_SIZE); |
@@ -89,11 +101,13 @@ static __init int map_switcher(void) | |||
89 | goto free_pages; | 101 | goto free_pages; |
90 | } | 102 | } |
91 | 103 | ||
92 | /* This code actually sets up the pages we've allocated to appear at | 104 | /* |
105 | * This code actually sets up the pages we've allocated to appear at | ||
93 | * SWITCHER_ADDR. map_vm_area() takes the vma we allocated above, the | 106 | * SWITCHER_ADDR. map_vm_area() takes the vma we allocated above, the |
94 | * kind of pages we're mapping (kernel pages), and a pointer to our | 107 | * kind of pages we're mapping (kernel pages), and a pointer to our |
95 | * array of struct pages. It increments that pointer, but we don't | 108 | * array of struct pages. It increments that pointer, but we don't |
96 | * care. */ | 109 | * care. |
110 | */ | ||
97 | pagep = switcher_page; | 111 | pagep = switcher_page; |
98 | err = map_vm_area(switcher_vma, PAGE_KERNEL_EXEC, &pagep); | 112 | err = map_vm_area(switcher_vma, PAGE_KERNEL_EXEC, &pagep); |
99 | if (err) { | 113 | if (err) { |
@@ -101,8 +115,10 @@ static __init int map_switcher(void) | |||
101 | goto free_vma; | 115 | goto free_vma; |
102 | } | 116 | } |
103 | 117 | ||
104 | /* Now the Switcher is mapped at the right address, we can't fail! | 118 | /* |
105 | * Copy in the compiled-in Switcher code (from <arch>_switcher.S). */ | 119 | * Now the Switcher is mapped at the right address, we can't fail! |
120 | * Copy in the compiled-in Switcher code (from <arch>_switcher.S). | ||
121 | */ | ||
106 | memcpy(switcher_vma->addr, start_switcher_text, | 122 | memcpy(switcher_vma->addr, start_switcher_text, |
107 | end_switcher_text - start_switcher_text); | 123 | end_switcher_text - start_switcher_text); |
108 | 124 | ||
@@ -124,8 +140,7 @@ out: | |||
124 | } | 140 | } |
125 | /*:*/ | 141 | /*:*/ |
126 | 142 | ||
127 | /* Cleaning up the mapping when the module is unloaded is almost... | 143 | /* Cleaning up the mapping when the module is unloaded is almost... too easy. */ |
128 | * too easy. */ | ||
129 | static void unmap_switcher(void) | 144 | static void unmap_switcher(void) |
130 | { | 145 | { |
131 | unsigned int i; | 146 | unsigned int i; |
@@ -151,16 +166,19 @@ static void unmap_switcher(void) | |||
151 | * But we can't trust the Guest: it might be trying to access the Launcher | 166 | * But we can't trust the Guest: it might be trying to access the Launcher |
152 | * code. We have to check that the range is below the pfn_limit the Launcher | 167 | * code. We have to check that the range is below the pfn_limit the Launcher |
153 | * gave us. We have to make sure that addr + len doesn't give us a false | 168 | * gave us. We have to make sure that addr + len doesn't give us a false |
154 | * positive by overflowing, too. */ | 169 | * positive by overflowing, too. |
170 | */ | ||
155 | bool lguest_address_ok(const struct lguest *lg, | 171 | bool lguest_address_ok(const struct lguest *lg, |
156 | unsigned long addr, unsigned long len) | 172 | unsigned long addr, unsigned long len) |
157 | { | 173 | { |
158 | return (addr+len) / PAGE_SIZE < lg->pfn_limit && (addr+len >= addr); | 174 | return (addr+len) / PAGE_SIZE < lg->pfn_limit && (addr+len >= addr); |
159 | } | 175 | } |
160 | 176 | ||
161 | /* This routine copies memory from the Guest. Here we can see how useful the | 177 | /* |
178 | * This routine copies memory from the Guest. Here we can see how useful the | ||
162 | * kill_lguest() routine we met in the Launcher can be: we return a random | 179 | * kill_lguest() routine we met in the Launcher can be: we return a random |
163 | * value (all zeroes) instead of needing to return an error. */ | 180 | * value (all zeroes) instead of needing to return an error. |
181 | */ | ||
164 | void __lgread(struct lg_cpu *cpu, void *b, unsigned long addr, unsigned bytes) | 182 | void __lgread(struct lg_cpu *cpu, void *b, unsigned long addr, unsigned bytes) |
165 | { | 183 | { |
166 | if (!lguest_address_ok(cpu->lg, addr, bytes) | 184 | if (!lguest_address_ok(cpu->lg, addr, bytes) |
@@ -181,9 +199,11 @@ void __lgwrite(struct lg_cpu *cpu, unsigned long addr, const void *b, | |||
181 | } | 199 | } |
182 | /*:*/ | 200 | /*:*/ |
183 | 201 | ||
184 | /*H:030 Let's jump straight to the the main loop which runs the Guest. | 202 | /*H:030 |
203 | * Let's jump straight to the the main loop which runs the Guest. | ||
185 | * Remember, this is called by the Launcher reading /dev/lguest, and we keep | 204 | * Remember, this is called by the Launcher reading /dev/lguest, and we keep |
186 | * going around and around until something interesting happens. */ | 205 | * going around and around until something interesting happens. |
206 | */ | ||
187 | int run_guest(struct lg_cpu *cpu, unsigned long __user *user) | 207 | int run_guest(struct lg_cpu *cpu, unsigned long __user *user) |
188 | { | 208 | { |
189 | /* We stop running once the Guest is dead. */ | 209 | /* We stop running once the Guest is dead. */ |
@@ -195,8 +215,10 @@ int run_guest(struct lg_cpu *cpu, unsigned long __user *user) | |||
195 | if (cpu->hcall) | 215 | if (cpu->hcall) |
196 | do_hypercalls(cpu); | 216 | do_hypercalls(cpu); |
197 | 217 | ||
198 | /* It's possible the Guest did a NOTIFY hypercall to the | 218 | /* |
199 | * Launcher, in which case we return from the read() now. */ | 219 | * It's possible the Guest did a NOTIFY hypercall to the |
220 | * Launcher, in which case we return from the read() now. | ||
221 | */ | ||
200 | if (cpu->pending_notify) { | 222 | if (cpu->pending_notify) { |
201 | if (!send_notify_to_eventfd(cpu)) { | 223 | if (!send_notify_to_eventfd(cpu)) { |
202 | if (put_user(cpu->pending_notify, user)) | 224 | if (put_user(cpu->pending_notify, user)) |
@@ -209,29 +231,39 @@ int run_guest(struct lg_cpu *cpu, unsigned long __user *user) | |||
209 | if (signal_pending(current)) | 231 | if (signal_pending(current)) |
210 | return -ERESTARTSYS; | 232 | return -ERESTARTSYS; |
211 | 233 | ||
212 | /* Check if there are any interrupts which can be delivered now: | 234 | /* |
235 | * Check if there are any interrupts which can be delivered now: | ||
213 | * if so, this sets up the hander to be executed when we next | 236 | * if so, this sets up the hander to be executed when we next |
214 | * run the Guest. */ | 237 | * run the Guest. |
238 | */ | ||
215 | irq = interrupt_pending(cpu, &more); | 239 | irq = interrupt_pending(cpu, &more); |
216 | if (irq < LGUEST_IRQS) | 240 | if (irq < LGUEST_IRQS) |
217 | try_deliver_interrupt(cpu, irq, more); | 241 | try_deliver_interrupt(cpu, irq, more); |
218 | 242 | ||
219 | /* All long-lived kernel loops need to check with this horrible | 243 | /* |
244 | * All long-lived kernel loops need to check with this horrible | ||
220 | * thing called the freezer. If the Host is trying to suspend, | 245 | * thing called the freezer. If the Host is trying to suspend, |
221 | * it stops us. */ | 246 | * it stops us. |
247 | */ | ||
222 | try_to_freeze(); | 248 | try_to_freeze(); |
223 | 249 | ||
224 | /* Just make absolutely sure the Guest is still alive. One of | 250 | /* |
225 | * those hypercalls could have been fatal, for example. */ | 251 | * Just make absolutely sure the Guest is still alive. One of |
252 | * those hypercalls could have been fatal, for example. | ||
253 | */ | ||
226 | if (cpu->lg->dead) | 254 | if (cpu->lg->dead) |
227 | break; | 255 | break; |
228 | 256 | ||
229 | /* If the Guest asked to be stopped, we sleep. The Guest's | 257 | /* |
230 | * clock timer will wake us. */ | 258 | * If the Guest asked to be stopped, we sleep. The Guest's |
259 | * clock timer will wake us. | ||
260 | */ | ||
231 | if (cpu->halted) { | 261 | if (cpu->halted) { |
232 | set_current_state(TASK_INTERRUPTIBLE); | 262 | set_current_state(TASK_INTERRUPTIBLE); |
233 | /* Just before we sleep, make sure no interrupt snuck in | 263 | /* |
234 | * which we should be doing. */ | 264 | * Just before we sleep, make sure no interrupt snuck in |
265 | * which we should be doing. | ||
266 | */ | ||
235 | if (interrupt_pending(cpu, &more) < LGUEST_IRQS) | 267 | if (interrupt_pending(cpu, &more) < LGUEST_IRQS) |
236 | set_current_state(TASK_RUNNING); | 268 | set_current_state(TASK_RUNNING); |
237 | else | 269 | else |
@@ -239,8 +271,10 @@ int run_guest(struct lg_cpu *cpu, unsigned long __user *user) | |||
239 | continue; | 271 | continue; |
240 | } | 272 | } |
241 | 273 | ||
242 | /* OK, now we're ready to jump into the Guest. First we put up | 274 | /* |
243 | * the "Do Not Disturb" sign: */ | 275 | * OK, now we're ready to jump into the Guest. First we put up |
276 | * the "Do Not Disturb" sign: | ||
277 | */ | ||
244 | local_irq_disable(); | 278 | local_irq_disable(); |
245 | 279 | ||
246 | /* Actually run the Guest until something happens. */ | 280 | /* Actually run the Guest until something happens. */ |
@@ -327,8 +361,10 @@ static void __exit fini(void) | |||
327 | } | 361 | } |
328 | /*:*/ | 362 | /*:*/ |
329 | 363 | ||
330 | /* The Host side of lguest can be a module. This is a nice way for people to | 364 | /* |
331 | * play with it. */ | 365 | * The Host side of lguest can be a module. This is a nice way for people to |
366 | * play with it. | ||
367 | */ | ||
332 | module_init(init); | 368 | module_init(init); |
333 | module_exit(fini); | 369 | module_exit(fini); |
334 | MODULE_LICENSE("GPL"); | 370 | MODULE_LICENSE("GPL"); |