nfs: fix regression in handling of context= option in NFSv4
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / arch / powerpc / kernel / kvm.c
blobad892f7a7574217747fc2b53d37b6663dc21b6da
1 /*
2 * Copyright (C) 2010 SUSE Linux Products GmbH. All rights reserved.
4 * Authors:
5 * Alexander Graf <agraf@suse.de>
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License, version 2, as
9 * published by the Free Software Foundation.
11 * This program is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 * GNU General Public License for more details.
16 * You should have received a copy of the GNU General Public License
17 * along with this program; if not, write to the Free Software
18 * Foundation, 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
21 #include <linux/kvm_host.h>
22 #include <linux/init.h>
23 #include <linux/kvm_para.h>
24 #include <linux/slab.h>
25 #include <linux/of.h>
27 #include <asm/reg.h>
28 #include <asm/sections.h>
29 #include <asm/cacheflush.h>
30 #include <asm/disassemble.h>
32 #define KVM_MAGIC_PAGE (-4096L)
33 #define magic_var(x) KVM_MAGIC_PAGE + offsetof(struct kvm_vcpu_arch_shared, x)
35 #define KVM_INST_LWZ 0x80000000
36 #define KVM_INST_STW 0x90000000
37 #define KVM_INST_LD 0xe8000000
38 #define KVM_INST_STD 0xf8000000
39 #define KVM_INST_NOP 0x60000000
40 #define KVM_INST_B 0x48000000
41 #define KVM_INST_B_MASK 0x03ffffff
42 #define KVM_INST_B_MAX 0x01ffffff
44 #define KVM_MASK_RT 0x03e00000
45 #define KVM_RT_30 0x03c00000
46 #define KVM_MASK_RB 0x0000f800
47 #define KVM_INST_MFMSR 0x7c0000a6
48 #define KVM_INST_MFSPR_SPRG0 0x7c1042a6
49 #define KVM_INST_MFSPR_SPRG1 0x7c1142a6
50 #define KVM_INST_MFSPR_SPRG2 0x7c1242a6
51 #define KVM_INST_MFSPR_SPRG3 0x7c1342a6
52 #define KVM_INST_MFSPR_SRR0 0x7c1a02a6
53 #define KVM_INST_MFSPR_SRR1 0x7c1b02a6
54 #define KVM_INST_MFSPR_DAR 0x7c1302a6
55 #define KVM_INST_MFSPR_DSISR 0x7c1202a6
57 #define KVM_INST_MTSPR_SPRG0 0x7c1043a6
58 #define KVM_INST_MTSPR_SPRG1 0x7c1143a6
59 #define KVM_INST_MTSPR_SPRG2 0x7c1243a6
60 #define KVM_INST_MTSPR_SPRG3 0x7c1343a6
61 #define KVM_INST_MTSPR_SRR0 0x7c1a03a6
62 #define KVM_INST_MTSPR_SRR1 0x7c1b03a6
63 #define KVM_INST_MTSPR_DAR 0x7c1303a6
64 #define KVM_INST_MTSPR_DSISR 0x7c1203a6
66 #define KVM_INST_TLBSYNC 0x7c00046c
67 #define KVM_INST_MTMSRD_L0 0x7c000164
68 #define KVM_INST_MTMSRD_L1 0x7c010164
69 #define KVM_INST_MTMSR 0x7c000124
71 #define KVM_INST_WRTEEI_0 0x7c000146
72 #define KVM_INST_WRTEEI_1 0x7c008146
74 #define KVM_INST_MTSRIN 0x7c0001e4
76 static bool kvm_patching_worked = true;
77 static char kvm_tmp[1024 * 1024];
78 static int kvm_tmp_index;
80 static inline void kvm_patch_ins(u32 *inst, u32 new_inst)
82 *inst = new_inst;
83 flush_icache_range((ulong)inst, (ulong)inst + 4);
86 static void kvm_patch_ins_ll(u32 *inst, long addr, u32 rt)
88 #ifdef CONFIG_64BIT
89 kvm_patch_ins(inst, KVM_INST_LD | rt | (addr & 0x0000fffc));
90 #else
91 kvm_patch_ins(inst, KVM_INST_LWZ | rt | (addr & 0x0000fffc));
92 #endif
95 static void kvm_patch_ins_ld(u32 *inst, long addr, u32 rt)
97 #ifdef CONFIG_64BIT
98 kvm_patch_ins(inst, KVM_INST_LD | rt | (addr & 0x0000fffc));
99 #else
100 kvm_patch_ins(inst, KVM_INST_LWZ | rt | ((addr + 4) & 0x0000fffc));
101 #endif
104 static void kvm_patch_ins_lwz(u32 *inst, long addr, u32 rt)
106 kvm_patch_ins(inst, KVM_INST_LWZ | rt | (addr & 0x0000ffff));
109 static void kvm_patch_ins_std(u32 *inst, long addr, u32 rt)
111 #ifdef CONFIG_64BIT
112 kvm_patch_ins(inst, KVM_INST_STD | rt | (addr & 0x0000fffc));
113 #else
114 kvm_patch_ins(inst, KVM_INST_STW | rt | ((addr + 4) & 0x0000fffc));
115 #endif
118 static void kvm_patch_ins_stw(u32 *inst, long addr, u32 rt)
120 kvm_patch_ins(inst, KVM_INST_STW | rt | (addr & 0x0000fffc));
123 static void kvm_patch_ins_nop(u32 *inst)
125 kvm_patch_ins(inst, KVM_INST_NOP);
128 static void kvm_patch_ins_b(u32 *inst, int addr)
130 #if defined(CONFIG_RELOCATABLE) && defined(CONFIG_PPC_BOOK3S)
131 /* On relocatable kernels interrupts handlers and our code
132 can be in different regions, so we don't patch them */
134 if ((ulong)inst < (ulong)&__end_interrupts)
135 return;
136 #endif
138 kvm_patch_ins(inst, KVM_INST_B | (addr & KVM_INST_B_MASK));
141 static u32 *kvm_alloc(int len)
143 u32 *p;
145 if ((kvm_tmp_index + len) > ARRAY_SIZE(kvm_tmp)) {
146 printk(KERN_ERR "KVM: No more space (%d + %d)\n",
147 kvm_tmp_index, len);
148 kvm_patching_worked = false;
149 return NULL;
152 p = (void*)&kvm_tmp[kvm_tmp_index];
153 kvm_tmp_index += len;
155 return p;
158 extern u32 kvm_emulate_mtmsrd_branch_offs;
159 extern u32 kvm_emulate_mtmsrd_reg_offs;
160 extern u32 kvm_emulate_mtmsrd_orig_ins_offs;
161 extern u32 kvm_emulate_mtmsrd_len;
162 extern u32 kvm_emulate_mtmsrd[];
164 static void kvm_patch_ins_mtmsrd(u32 *inst, u32 rt)
166 u32 *p;
167 int distance_start;
168 int distance_end;
169 ulong next_inst;
171 p = kvm_alloc(kvm_emulate_mtmsrd_len * 4);
172 if (!p)
173 return;
175 /* Find out where we are and put everything there */
176 distance_start = (ulong)p - (ulong)inst;
177 next_inst = ((ulong)inst + 4);
178 distance_end = next_inst - (ulong)&p[kvm_emulate_mtmsrd_branch_offs];
180 /* Make sure we only write valid b instructions */
181 if (distance_start > KVM_INST_B_MAX) {
182 kvm_patching_worked = false;
183 return;
186 /* Modify the chunk to fit the invocation */
187 memcpy(p, kvm_emulate_mtmsrd, kvm_emulate_mtmsrd_len * 4);
188 p[kvm_emulate_mtmsrd_branch_offs] |= distance_end & KVM_INST_B_MASK;
189 switch (get_rt(rt)) {
190 case 30:
191 kvm_patch_ins_ll(&p[kvm_emulate_mtmsrd_reg_offs],
192 magic_var(scratch2), KVM_RT_30);
193 break;
194 case 31:
195 kvm_patch_ins_ll(&p[kvm_emulate_mtmsrd_reg_offs],
196 magic_var(scratch1), KVM_RT_30);
197 break;
198 default:
199 p[kvm_emulate_mtmsrd_reg_offs] |= rt;
200 break;
203 p[kvm_emulate_mtmsrd_orig_ins_offs] = *inst;
204 flush_icache_range((ulong)p, (ulong)p + kvm_emulate_mtmsrd_len * 4);
206 /* Patch the invocation */
207 kvm_patch_ins_b(inst, distance_start);
210 extern u32 kvm_emulate_mtmsr_branch_offs;
211 extern u32 kvm_emulate_mtmsr_reg1_offs;
212 extern u32 kvm_emulate_mtmsr_reg2_offs;
213 extern u32 kvm_emulate_mtmsr_orig_ins_offs;
214 extern u32 kvm_emulate_mtmsr_len;
215 extern u32 kvm_emulate_mtmsr[];
217 static void kvm_patch_ins_mtmsr(u32 *inst, u32 rt)
219 u32 *p;
220 int distance_start;
221 int distance_end;
222 ulong next_inst;
224 p = kvm_alloc(kvm_emulate_mtmsr_len * 4);
225 if (!p)
226 return;
228 /* Find out where we are and put everything there */
229 distance_start = (ulong)p - (ulong)inst;
230 next_inst = ((ulong)inst + 4);
231 distance_end = next_inst - (ulong)&p[kvm_emulate_mtmsr_branch_offs];
233 /* Make sure we only write valid b instructions */
234 if (distance_start > KVM_INST_B_MAX) {
235 kvm_patching_worked = false;
236 return;
239 /* Modify the chunk to fit the invocation */
240 memcpy(p, kvm_emulate_mtmsr, kvm_emulate_mtmsr_len * 4);
241 p[kvm_emulate_mtmsr_branch_offs] |= distance_end & KVM_INST_B_MASK;
243 /* Make clobbered registers work too */
244 switch (get_rt(rt)) {
245 case 30:
246 kvm_patch_ins_ll(&p[kvm_emulate_mtmsr_reg1_offs],
247 magic_var(scratch2), KVM_RT_30);
248 kvm_patch_ins_ll(&p[kvm_emulate_mtmsr_reg2_offs],
249 magic_var(scratch2), KVM_RT_30);
250 break;
251 case 31:
252 kvm_patch_ins_ll(&p[kvm_emulate_mtmsr_reg1_offs],
253 magic_var(scratch1), KVM_RT_30);
254 kvm_patch_ins_ll(&p[kvm_emulate_mtmsr_reg2_offs],
255 magic_var(scratch1), KVM_RT_30);
256 break;
257 default:
258 p[kvm_emulate_mtmsr_reg1_offs] |= rt;
259 p[kvm_emulate_mtmsr_reg2_offs] |= rt;
260 break;
263 p[kvm_emulate_mtmsr_orig_ins_offs] = *inst;
264 flush_icache_range((ulong)p, (ulong)p + kvm_emulate_mtmsr_len * 4);
266 /* Patch the invocation */
267 kvm_patch_ins_b(inst, distance_start);
270 #ifdef CONFIG_BOOKE
272 extern u32 kvm_emulate_wrteei_branch_offs;
273 extern u32 kvm_emulate_wrteei_ee_offs;
274 extern u32 kvm_emulate_wrteei_len;
275 extern u32 kvm_emulate_wrteei[];
277 static void kvm_patch_ins_wrteei(u32 *inst)
279 u32 *p;
280 int distance_start;
281 int distance_end;
282 ulong next_inst;
284 p = kvm_alloc(kvm_emulate_wrteei_len * 4);
285 if (!p)
286 return;
288 /* Find out where we are and put everything there */
289 distance_start = (ulong)p - (ulong)inst;
290 next_inst = ((ulong)inst + 4);
291 distance_end = next_inst - (ulong)&p[kvm_emulate_wrteei_branch_offs];
293 /* Make sure we only write valid b instructions */
294 if (distance_start > KVM_INST_B_MAX) {
295 kvm_patching_worked = false;
296 return;
299 /* Modify the chunk to fit the invocation */
300 memcpy(p, kvm_emulate_wrteei, kvm_emulate_wrteei_len * 4);
301 p[kvm_emulate_wrteei_branch_offs] |= distance_end & KVM_INST_B_MASK;
302 p[kvm_emulate_wrteei_ee_offs] |= (*inst & MSR_EE);
303 flush_icache_range((ulong)p, (ulong)p + kvm_emulate_wrteei_len * 4);
305 /* Patch the invocation */
306 kvm_patch_ins_b(inst, distance_start);
309 #endif
311 #ifdef CONFIG_PPC_BOOK3S_32
313 extern u32 kvm_emulate_mtsrin_branch_offs;
314 extern u32 kvm_emulate_mtsrin_reg1_offs;
315 extern u32 kvm_emulate_mtsrin_reg2_offs;
316 extern u32 kvm_emulate_mtsrin_orig_ins_offs;
317 extern u32 kvm_emulate_mtsrin_len;
318 extern u32 kvm_emulate_mtsrin[];
320 static void kvm_patch_ins_mtsrin(u32 *inst, u32 rt, u32 rb)
322 u32 *p;
323 int distance_start;
324 int distance_end;
325 ulong next_inst;
327 p = kvm_alloc(kvm_emulate_mtsrin_len * 4);
328 if (!p)
329 return;
331 /* Find out where we are and put everything there */
332 distance_start = (ulong)p - (ulong)inst;
333 next_inst = ((ulong)inst + 4);
334 distance_end = next_inst - (ulong)&p[kvm_emulate_mtsrin_branch_offs];
336 /* Make sure we only write valid b instructions */
337 if (distance_start > KVM_INST_B_MAX) {
338 kvm_patching_worked = false;
339 return;
342 /* Modify the chunk to fit the invocation */
343 memcpy(p, kvm_emulate_mtsrin, kvm_emulate_mtsrin_len * 4);
344 p[kvm_emulate_mtsrin_branch_offs] |= distance_end & KVM_INST_B_MASK;
345 p[kvm_emulate_mtsrin_reg1_offs] |= (rb << 10);
346 p[kvm_emulate_mtsrin_reg2_offs] |= rt;
347 p[kvm_emulate_mtsrin_orig_ins_offs] = *inst;
348 flush_icache_range((ulong)p, (ulong)p + kvm_emulate_mtsrin_len * 4);
350 /* Patch the invocation */
351 kvm_patch_ins_b(inst, distance_start);
354 #endif
356 static void kvm_map_magic_page(void *data)
358 u32 *features = data;
360 ulong in[8];
361 ulong out[8];
363 in[0] = KVM_MAGIC_PAGE;
364 in[1] = KVM_MAGIC_PAGE;
366 kvm_hypercall(in, out, HC_VENDOR_KVM | KVM_HC_PPC_MAP_MAGIC_PAGE);
368 *features = out[0];
371 static void kvm_check_ins(u32 *inst, u32 features)
373 u32 _inst = *inst;
374 u32 inst_no_rt = _inst & ~KVM_MASK_RT;
375 u32 inst_rt = _inst & KVM_MASK_RT;
377 switch (inst_no_rt) {
378 /* Loads */
379 case KVM_INST_MFMSR:
380 kvm_patch_ins_ld(inst, magic_var(msr), inst_rt);
381 break;
382 case KVM_INST_MFSPR_SPRG0:
383 kvm_patch_ins_ld(inst, magic_var(sprg0), inst_rt);
384 break;
385 case KVM_INST_MFSPR_SPRG1:
386 kvm_patch_ins_ld(inst, magic_var(sprg1), inst_rt);
387 break;
388 case KVM_INST_MFSPR_SPRG2:
389 kvm_patch_ins_ld(inst, magic_var(sprg2), inst_rt);
390 break;
391 case KVM_INST_MFSPR_SPRG3:
392 kvm_patch_ins_ld(inst, magic_var(sprg3), inst_rt);
393 break;
394 case KVM_INST_MFSPR_SRR0:
395 kvm_patch_ins_ld(inst, magic_var(srr0), inst_rt);
396 break;
397 case KVM_INST_MFSPR_SRR1:
398 kvm_patch_ins_ld(inst, magic_var(srr1), inst_rt);
399 break;
400 case KVM_INST_MFSPR_DAR:
401 kvm_patch_ins_ld(inst, magic_var(dar), inst_rt);
402 break;
403 case KVM_INST_MFSPR_DSISR:
404 kvm_patch_ins_lwz(inst, magic_var(dsisr), inst_rt);
405 break;
407 /* Stores */
408 case KVM_INST_MTSPR_SPRG0:
409 kvm_patch_ins_std(inst, magic_var(sprg0), inst_rt);
410 break;
411 case KVM_INST_MTSPR_SPRG1:
412 kvm_patch_ins_std(inst, magic_var(sprg1), inst_rt);
413 break;
414 case KVM_INST_MTSPR_SPRG2:
415 kvm_patch_ins_std(inst, magic_var(sprg2), inst_rt);
416 break;
417 case KVM_INST_MTSPR_SPRG3:
418 kvm_patch_ins_std(inst, magic_var(sprg3), inst_rt);
419 break;
420 case KVM_INST_MTSPR_SRR0:
421 kvm_patch_ins_std(inst, magic_var(srr0), inst_rt);
422 break;
423 case KVM_INST_MTSPR_SRR1:
424 kvm_patch_ins_std(inst, magic_var(srr1), inst_rt);
425 break;
426 case KVM_INST_MTSPR_DAR:
427 kvm_patch_ins_std(inst, magic_var(dar), inst_rt);
428 break;
429 case KVM_INST_MTSPR_DSISR:
430 kvm_patch_ins_stw(inst, magic_var(dsisr), inst_rt);
431 break;
433 /* Nops */
434 case KVM_INST_TLBSYNC:
435 kvm_patch_ins_nop(inst);
436 break;
438 /* Rewrites */
439 case KVM_INST_MTMSRD_L1:
440 kvm_patch_ins_mtmsrd(inst, inst_rt);
441 break;
442 case KVM_INST_MTMSR:
443 case KVM_INST_MTMSRD_L0:
444 kvm_patch_ins_mtmsr(inst, inst_rt);
445 break;
448 switch (inst_no_rt & ~KVM_MASK_RB) {
449 #ifdef CONFIG_PPC_BOOK3S_32
450 case KVM_INST_MTSRIN:
451 if (features & KVM_MAGIC_FEAT_SR) {
452 u32 inst_rb = _inst & KVM_MASK_RB;
453 kvm_patch_ins_mtsrin(inst, inst_rt, inst_rb);
455 break;
456 break;
457 #endif
460 switch (_inst) {
461 #ifdef CONFIG_BOOKE
462 case KVM_INST_WRTEEI_0:
463 case KVM_INST_WRTEEI_1:
464 kvm_patch_ins_wrteei(inst);
465 break;
466 #endif
470 static void kvm_use_magic_page(void)
472 u32 *p;
473 u32 *start, *end;
474 u32 tmp;
475 u32 features;
477 /* Tell the host to map the magic page to -4096 on all CPUs */
478 on_each_cpu(kvm_map_magic_page, &features, 1);
480 /* Quick self-test to see if the mapping works */
481 if (__get_user(tmp, (u32*)KVM_MAGIC_PAGE)) {
482 kvm_patching_worked = false;
483 return;
486 /* Now loop through all code and find instructions */
487 start = (void*)_stext;
488 end = (void*)_etext;
490 for (p = start; p < end; p++)
491 kvm_check_ins(p, features);
493 printk(KERN_INFO "KVM: Live patching for a fast VM %s\n",
494 kvm_patching_worked ? "worked" : "failed");
497 unsigned long kvm_hypercall(unsigned long *in,
498 unsigned long *out,
499 unsigned long nr)
501 unsigned long register r0 asm("r0");
502 unsigned long register r3 asm("r3") = in[0];
503 unsigned long register r4 asm("r4") = in[1];
504 unsigned long register r5 asm("r5") = in[2];
505 unsigned long register r6 asm("r6") = in[3];
506 unsigned long register r7 asm("r7") = in[4];
507 unsigned long register r8 asm("r8") = in[5];
508 unsigned long register r9 asm("r9") = in[6];
509 unsigned long register r10 asm("r10") = in[7];
510 unsigned long register r11 asm("r11") = nr;
511 unsigned long register r12 asm("r12");
513 asm volatile("bl kvm_hypercall_start"
514 : "=r"(r0), "=r"(r3), "=r"(r4), "=r"(r5), "=r"(r6),
515 "=r"(r7), "=r"(r8), "=r"(r9), "=r"(r10), "=r"(r11),
516 "=r"(r12)
517 : "r"(r3), "r"(r4), "r"(r5), "r"(r6), "r"(r7), "r"(r8),
518 "r"(r9), "r"(r10), "r"(r11)
519 : "memory", "cc", "xer", "ctr", "lr");
521 out[0] = r4;
522 out[1] = r5;
523 out[2] = r6;
524 out[3] = r7;
525 out[4] = r8;
526 out[5] = r9;
527 out[6] = r10;
528 out[7] = r11;
530 return r3;
532 EXPORT_SYMBOL_GPL(kvm_hypercall);
534 static int kvm_para_setup(void)
536 extern u32 kvm_hypercall_start;
537 struct device_node *hyper_node;
538 u32 *insts;
539 int len, i;
541 hyper_node = of_find_node_by_path("/hypervisor");
542 if (!hyper_node)
543 return -1;
545 insts = (u32*)of_get_property(hyper_node, "hcall-instructions", &len);
546 if (len % 4)
547 return -1;
548 if (len > (4 * 4))
549 return -1;
551 for (i = 0; i < (len / 4); i++)
552 kvm_patch_ins(&(&kvm_hypercall_start)[i], insts[i]);
554 return 0;
557 static __init void kvm_free_tmp(void)
559 unsigned long start, end;
561 start = (ulong)&kvm_tmp[kvm_tmp_index + (PAGE_SIZE - 1)] & PAGE_MASK;
562 end = (ulong)&kvm_tmp[ARRAY_SIZE(kvm_tmp)] & PAGE_MASK;
564 /* Free the tmp space we don't need */
565 for (; start < end; start += PAGE_SIZE) {
566 ClearPageReserved(virt_to_page(start));
567 init_page_count(virt_to_page(start));
568 free_page(start);
569 totalram_pages++;
573 static int __init kvm_guest_init(void)
575 if (!kvm_para_available())
576 goto free_tmp;
578 if (kvm_para_setup())
579 goto free_tmp;
581 if (kvm_para_has_feature(KVM_FEATURE_MAGIC_PAGE))
582 kvm_use_magic_page();
584 #ifdef CONFIG_PPC_BOOK3S_64
585 /* Enable napping */
586 powersave_nap = 1;
587 #endif
589 free_tmp:
590 kvm_free_tmp();
592 return 0;
595 postcore_initcall(kvm_guest_init);