]> git.proxmox.com Git - qemu.git/blob - exec-all.h
Use correct types to enable > 2G support, based on a patch from
[qemu.git] / exec-all.h
1 /*
2 * internal execution defines for qemu
3 *
4 * Copyright (c) 2003 Fabrice Bellard
5 *
6 * This library is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2 of the License, or (at your option) any later version.
10 *
11 * This library 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 GNU
14 * Lesser General Public License for more details.
15 *
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with this library; if not, write to the Free Software
18 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
19 */
20
21 /* allow to see translation results - the slowdown should be negligible, so we leave it */
22 #define DEBUG_DISAS
23
24 /* is_jmp field values */
25 #define DISAS_NEXT 0 /* next instruction can be analyzed */
26 #define DISAS_JUMP 1 /* only pc was modified dynamically */
27 #define DISAS_UPDATE 2 /* cpu state was modified dynamically */
28 #define DISAS_TB_JUMP 3 /* only pc was modified statically */
29
30 struct TranslationBlock;
31
32 /* XXX: make safe guess about sizes */
33 #define MAX_OP_PER_INSTR 32
34 /* A Call op needs up to 6 + 2N parameters (N = number of arguments). */
35 #define MAX_OPC_PARAM 10
36 #define OPC_BUF_SIZE 512
37 #define OPC_MAX_SIZE (OPC_BUF_SIZE - MAX_OP_PER_INSTR)
38
39 /* Maximum size a TCG op can expand to. This is complicated because a
40 single op may require several host instructions and regirster reloads.
41 For now take a wild guess at 128 bytes, which should allow at least
42 a couple of fixup instructions per argument. */
43 #define TCG_MAX_OP_SIZE 128
44
45 #define OPPARAM_BUF_SIZE (OPC_BUF_SIZE * MAX_OPC_PARAM)
46
47 extern target_ulong gen_opc_pc[OPC_BUF_SIZE];
48 extern target_ulong gen_opc_npc[OPC_BUF_SIZE];
49 extern uint8_t gen_opc_cc_op[OPC_BUF_SIZE];
50 extern uint8_t gen_opc_instr_start[OPC_BUF_SIZE];
51 extern target_ulong gen_opc_jump_pc[2];
52 extern uint32_t gen_opc_hflags[OPC_BUF_SIZE];
53
54 typedef void (GenOpFunc)(void);
55 typedef void (GenOpFunc1)(long);
56 typedef void (GenOpFunc2)(long, long);
57 typedef void (GenOpFunc3)(long, long, long);
58
59 #if defined(TARGET_I386)
60
61 void optimize_flags_init(void);
62
63 #endif
64
65 extern FILE *logfile;
66 extern int loglevel;
67
68 int gen_intermediate_code(CPUState *env, struct TranslationBlock *tb);
69 int gen_intermediate_code_pc(CPUState *env, struct TranslationBlock *tb);
70 unsigned long code_gen_max_block_size(void);
71 void cpu_gen_init(void);
72 int cpu_gen_code(CPUState *env, struct TranslationBlock *tb,
73 int *gen_code_size_ptr);
74 int cpu_restore_state(struct TranslationBlock *tb,
75 CPUState *env, unsigned long searched_pc,
76 void *puc);
77 int cpu_gen_code_copy(CPUState *env, struct TranslationBlock *tb,
78 int max_code_size, int *gen_code_size_ptr);
79 int cpu_restore_state_copy(struct TranslationBlock *tb,
80 CPUState *env, unsigned long searched_pc,
81 void *puc);
82 void cpu_resume_from_signal(CPUState *env1, void *puc);
83 void cpu_exec_init(CPUState *env);
84 int page_unprotect(target_ulong address, unsigned long pc, void *puc);
85 void tb_invalidate_phys_page_range(target_phys_addr_t start, target_phys_addr_t end,
86 int is_cpu_write_access);
87 void tb_invalidate_page_range(target_ulong start, target_ulong end);
88 void tlb_flush_page(CPUState *env, target_ulong addr);
89 void tlb_flush(CPUState *env, int flush_global);
90 int tlb_set_page_exec(CPUState *env, target_ulong vaddr,
91 target_phys_addr_t paddr, int prot,
92 int mmu_idx, int is_softmmu);
93 static inline int tlb_set_page(CPUState *env, target_ulong vaddr,
94 target_phys_addr_t paddr, int prot,
95 int mmu_idx, int is_softmmu)
96 {
97 if (prot & PAGE_READ)
98 prot |= PAGE_EXEC;
99 return tlb_set_page_exec(env, vaddr, paddr, prot, mmu_idx, is_softmmu);
100 }
101
102 #define CODE_GEN_ALIGN 16 /* must be >= of the size of a icache line */
103
104 #define CODE_GEN_PHYS_HASH_BITS 15
105 #define CODE_GEN_PHYS_HASH_SIZE (1 << CODE_GEN_PHYS_HASH_BITS)
106
107 /* maximum total translate dcode allocated */
108
109 /* NOTE: the translated code area cannot be too big because on some
110 archs the range of "fast" function calls is limited. Here is a
111 summary of the ranges:
112
113 i386 : signed 32 bits
114 arm : signed 26 bits
115 ppc : signed 24 bits
116 sparc : signed 32 bits
117 alpha : signed 23 bits
118 */
119
120 #if defined(__alpha__)
121 #define CODE_GEN_BUFFER_SIZE (2 * 1024 * 1024)
122 #elif defined(__ia64)
123 #define CODE_GEN_BUFFER_SIZE (4 * 1024 * 1024) /* range of addl */
124 #elif defined(__powerpc__)
125 #define CODE_GEN_BUFFER_SIZE (6 * 1024 * 1024)
126 #else
127 /* XXX: make it dynamic on x86 */
128 #define CODE_GEN_BUFFER_SIZE (16 * 1024 * 1024)
129 #endif
130
131 //#define CODE_GEN_BUFFER_SIZE (128 * 1024)
132
133 /* estimated block size for TB allocation */
134 /* XXX: use a per code average code fragment size and modulate it
135 according to the host CPU */
136 #if defined(CONFIG_SOFTMMU)
137 #define CODE_GEN_AVG_BLOCK_SIZE 128
138 #else
139 #define CODE_GEN_AVG_BLOCK_SIZE 64
140 #endif
141
142 #define CODE_GEN_MAX_BLOCKS (CODE_GEN_BUFFER_SIZE / CODE_GEN_AVG_BLOCK_SIZE)
143
144 #if defined(__powerpc__) || defined(__x86_64__)
145 #define USE_DIRECT_JUMP
146 #endif
147 #if defined(__i386__) && !defined(_WIN32)
148 #define USE_DIRECT_JUMP
149 #endif
150
151 typedef struct TranslationBlock {
152 target_ulong pc; /* simulated PC corresponding to this block (EIP + CS base) */
153 target_ulong cs_base; /* CS base for this block */
154 uint64_t flags; /* flags defining in which context the code was generated */
155 uint16_t size; /* size of target code for this block (1 <=
156 size <= TARGET_PAGE_SIZE) */
157 uint16_t cflags; /* compile flags */
158 #define CF_CODE_COPY 0x0001 /* block was generated in code copy mode */
159 #define CF_TB_FP_USED 0x0002 /* fp ops are used in the TB */
160 #define CF_FP_USED 0x0004 /* fp ops are used in the TB or in a chained TB */
161 #define CF_SINGLE_INSN 0x0008 /* compile only a single instruction */
162
163 uint8_t *tc_ptr; /* pointer to the translated code */
164 /* next matching tb for physical address. */
165 struct TranslationBlock *phys_hash_next;
166 /* first and second physical page containing code. The lower bit
167 of the pointer tells the index in page_next[] */
168 struct TranslationBlock *page_next[2];
169 target_ulong page_addr[2];
170
171 /* the following data are used to directly call another TB from
172 the code of this one. */
173 uint16_t tb_next_offset[2]; /* offset of original jump target */
174 #ifdef USE_DIRECT_JUMP
175 uint16_t tb_jmp_offset[4]; /* offset of jump instruction */
176 #else
177 unsigned long tb_next[2]; /* address of jump generated code */
178 #endif
179 /* list of TBs jumping to this one. This is a circular list using
180 the two least significant bits of the pointers to tell what is
181 the next pointer: 0 = jmp_next[0], 1 = jmp_next[1], 2 =
182 jmp_first */
183 struct TranslationBlock *jmp_next[2];
184 struct TranslationBlock *jmp_first;
185 } TranslationBlock;
186
187 static inline unsigned int tb_jmp_cache_hash_page(target_ulong pc)
188 {
189 target_ulong tmp;
190 tmp = pc ^ (pc >> (TARGET_PAGE_BITS - TB_JMP_PAGE_BITS));
191 return (tmp >> TB_JMP_PAGE_BITS) & TB_JMP_PAGE_MASK;
192 }
193
194 static inline unsigned int tb_jmp_cache_hash_func(target_ulong pc)
195 {
196 target_ulong tmp;
197 tmp = pc ^ (pc >> (TARGET_PAGE_BITS - TB_JMP_PAGE_BITS));
198 return (((tmp >> TB_JMP_PAGE_BITS) & TB_JMP_PAGE_MASK) |
199 (tmp & TB_JMP_ADDR_MASK));
200 }
201
202 static inline unsigned int tb_phys_hash_func(unsigned long pc)
203 {
204 return pc & (CODE_GEN_PHYS_HASH_SIZE - 1);
205 }
206
207 TranslationBlock *tb_alloc(target_ulong pc);
208 void tb_flush(CPUState *env);
209 void tb_link_phys(TranslationBlock *tb,
210 target_ulong phys_pc, target_ulong phys_page2);
211
212 extern TranslationBlock *tb_phys_hash[CODE_GEN_PHYS_HASH_SIZE];
213
214 extern uint8_t code_gen_buffer[CODE_GEN_BUFFER_SIZE];
215 extern uint8_t *code_gen_ptr;
216
217 #if defined(USE_DIRECT_JUMP)
218
219 #if defined(__powerpc__)
220 static inline void tb_set_jmp_target1(unsigned long jmp_addr, unsigned long addr)
221 {
222 uint32_t val, *ptr;
223
224 /* patch the branch destination */
225 ptr = (uint32_t *)jmp_addr;
226 val = *ptr;
227 val = (val & ~0x03fffffc) | ((addr - jmp_addr) & 0x03fffffc);
228 *ptr = val;
229 /* flush icache */
230 asm volatile ("dcbst 0,%0" : : "r"(ptr) : "memory");
231 asm volatile ("sync" : : : "memory");
232 asm volatile ("icbi 0,%0" : : "r"(ptr) : "memory");
233 asm volatile ("sync" : : : "memory");
234 asm volatile ("isync" : : : "memory");
235 }
236 #elif defined(__i386__) || defined(__x86_64__)
237 static inline void tb_set_jmp_target1(unsigned long jmp_addr, unsigned long addr)
238 {
239 /* patch the branch destination */
240 *(uint32_t *)jmp_addr = addr - (jmp_addr + 4);
241 /* no need to flush icache explicitely */
242 }
243 #endif
244
245 static inline void tb_set_jmp_target(TranslationBlock *tb,
246 int n, unsigned long addr)
247 {
248 unsigned long offset;
249
250 offset = tb->tb_jmp_offset[n];
251 tb_set_jmp_target1((unsigned long)(tb->tc_ptr + offset), addr);
252 offset = tb->tb_jmp_offset[n + 2];
253 if (offset != 0xffff)
254 tb_set_jmp_target1((unsigned long)(tb->tc_ptr + offset), addr);
255 }
256
257 #else
258
259 /* set the jump target */
260 static inline void tb_set_jmp_target(TranslationBlock *tb,
261 int n, unsigned long addr)
262 {
263 tb->tb_next[n] = addr;
264 }
265
266 #endif
267
268 static inline void tb_add_jump(TranslationBlock *tb, int n,
269 TranslationBlock *tb_next)
270 {
271 /* NOTE: this test is only needed for thread safety */
272 if (!tb->jmp_next[n]) {
273 /* patch the native jump address */
274 tb_set_jmp_target(tb, n, (unsigned long)tb_next->tc_ptr);
275
276 /* add in TB jmp circular list */
277 tb->jmp_next[n] = tb_next->jmp_first;
278 tb_next->jmp_first = (TranslationBlock *)((long)(tb) | (n));
279 }
280 }
281
282 TranslationBlock *tb_find_pc(unsigned long pc_ptr);
283
284 #ifndef offsetof
285 #define offsetof(type, field) ((size_t) &((type *)0)->field)
286 #endif
287
288 #if defined(_WIN32)
289 #define ASM_DATA_SECTION ".section \".data\"\n"
290 #define ASM_PREVIOUS_SECTION ".section .text\n"
291 #elif defined(__APPLE__)
292 #define ASM_DATA_SECTION ".data\n"
293 #define ASM_PREVIOUS_SECTION ".text\n"
294 #else
295 #define ASM_DATA_SECTION ".section \".data\"\n"
296 #define ASM_PREVIOUS_SECTION ".previous\n"
297 #endif
298
299 #define ASM_OP_LABEL_NAME(n, opname) \
300 ASM_NAME(__op_label) #n "." ASM_NAME(opname)
301
302 extern CPUWriteMemoryFunc *io_mem_write[IO_MEM_NB_ENTRIES][4];
303 extern CPUReadMemoryFunc *io_mem_read[IO_MEM_NB_ENTRIES][4];
304 extern void *io_mem_opaque[IO_MEM_NB_ENTRIES];
305
306 #if defined(__hppa__)
307
308 typedef int spinlock_t[4];
309
310 #define SPIN_LOCK_UNLOCKED { 1, 1, 1, 1 }
311
312 static inline void resetlock (spinlock_t *p)
313 {
314 (*p)[0] = (*p)[1] = (*p)[2] = (*p)[3] = 1;
315 }
316
317 #else
318
319 typedef int spinlock_t;
320
321 #define SPIN_LOCK_UNLOCKED 0
322
323 static inline void resetlock (spinlock_t *p)
324 {
325 *p = SPIN_LOCK_UNLOCKED;
326 }
327
328 #endif
329
330 #if defined(__powerpc__)
331 static inline int testandset (int *p)
332 {
333 int ret;
334 __asm__ __volatile__ (
335 "0: lwarx %0,0,%1\n"
336 " xor. %0,%3,%0\n"
337 " bne 1f\n"
338 " stwcx. %2,0,%1\n"
339 " bne- 0b\n"
340 "1: "
341 : "=&r" (ret)
342 : "r" (p), "r" (1), "r" (0)
343 : "cr0", "memory");
344 return ret;
345 }
346 #elif defined(__i386__)
347 static inline int testandset (int *p)
348 {
349 long int readval = 0;
350
351 __asm__ __volatile__ ("lock; cmpxchgl %2, %0"
352 : "+m" (*p), "+a" (readval)
353 : "r" (1)
354 : "cc");
355 return readval;
356 }
357 #elif defined(__x86_64__)
358 static inline int testandset (int *p)
359 {
360 long int readval = 0;
361
362 __asm__ __volatile__ ("lock; cmpxchgl %2, %0"
363 : "+m" (*p), "+a" (readval)
364 : "r" (1)
365 : "cc");
366 return readval;
367 }
368 #elif defined(__s390__)
369 static inline int testandset (int *p)
370 {
371 int ret;
372
373 __asm__ __volatile__ ("0: cs %0,%1,0(%2)\n"
374 " jl 0b"
375 : "=&d" (ret)
376 : "r" (1), "a" (p), "0" (*p)
377 : "cc", "memory" );
378 return ret;
379 }
380 #elif defined(__alpha__)
381 static inline int testandset (int *p)
382 {
383 int ret;
384 unsigned long one;
385
386 __asm__ __volatile__ ("0: mov 1,%2\n"
387 " ldl_l %0,%1\n"
388 " stl_c %2,%1\n"
389 " beq %2,1f\n"
390 ".subsection 2\n"
391 "1: br 0b\n"
392 ".previous"
393 : "=r" (ret), "=m" (*p), "=r" (one)
394 : "m" (*p));
395 return ret;
396 }
397 #elif defined(__sparc__)
398 static inline int testandset (int *p)
399 {
400 int ret;
401
402 __asm__ __volatile__("ldstub [%1], %0"
403 : "=r" (ret)
404 : "r" (p)
405 : "memory");
406
407 return (ret ? 1 : 0);
408 }
409 #elif defined(__arm__)
410 static inline int testandset (int *spinlock)
411 {
412 register unsigned int ret;
413 __asm__ __volatile__("swp %0, %1, [%2]"
414 : "=r"(ret)
415 : "0"(1), "r"(spinlock));
416
417 return ret;
418 }
419 #elif defined(__mc68000)
420 static inline int testandset (int *p)
421 {
422 char ret;
423 __asm__ __volatile__("tas %1; sne %0"
424 : "=r" (ret)
425 : "m" (p)
426 : "cc","memory");
427 return ret;
428 }
429 #elif defined(__hppa__)
430
431 /* Because malloc only guarantees 8-byte alignment for malloc'd data,
432 and GCC only guarantees 8-byte alignment for stack locals, we can't
433 be assured of 16-byte alignment for atomic lock data even if we
434 specify "__attribute ((aligned(16)))" in the type declaration. So,
435 we use a struct containing an array of four ints for the atomic lock
436 type and dynamically select the 16-byte aligned int from the array
437 for the semaphore. */
438 #define __PA_LDCW_ALIGNMENT 16
439 static inline void *ldcw_align (void *p) {
440 unsigned long a = (unsigned long)p;
441 a = (a + __PA_LDCW_ALIGNMENT - 1) & ~(__PA_LDCW_ALIGNMENT - 1);
442 return (void *)a;
443 }
444
445 static inline int testandset (spinlock_t *p)
446 {
447 unsigned int ret;
448 p = ldcw_align(p);
449 __asm__ __volatile__("ldcw 0(%1),%0"
450 : "=r" (ret)
451 : "r" (p)
452 : "memory" );
453 return !ret;
454 }
455
456 #elif defined(__ia64)
457
458 #include <ia64intrin.h>
459
460 static inline int testandset (int *p)
461 {
462 return __sync_lock_test_and_set (p, 1);
463 }
464 #elif defined(__mips__)
465 static inline int testandset (int *p)
466 {
467 int ret;
468
469 __asm__ __volatile__ (
470 " .set push \n"
471 " .set noat \n"
472 " .set mips2 \n"
473 "1: li $1, 1 \n"
474 " ll %0, %1 \n"
475 " sc $1, %1 \n"
476 " beqz $1, 1b \n"
477 " .set pop "
478 : "=r" (ret), "+R" (*p)
479 :
480 : "memory");
481
482 return ret;
483 }
484 #else
485 #error unimplemented CPU support
486 #endif
487
488 #if defined(CONFIG_USER_ONLY)
489 static inline void spin_lock(spinlock_t *lock)
490 {
491 while (testandset(lock));
492 }
493
494 static inline void spin_unlock(spinlock_t *lock)
495 {
496 resetlock(lock);
497 }
498
499 static inline int spin_trylock(spinlock_t *lock)
500 {
501 return !testandset(lock);
502 }
503 #else
504 static inline void spin_lock(spinlock_t *lock)
505 {
506 }
507
508 static inline void spin_unlock(spinlock_t *lock)
509 {
510 }
511
512 static inline int spin_trylock(spinlock_t *lock)
513 {
514 return 1;
515 }
516 #endif
517
518 extern spinlock_t tb_lock;
519
520 extern int tb_invalidated_flag;
521
522 #if !defined(CONFIG_USER_ONLY)
523
524 void tlb_fill(target_ulong addr, int is_write, int mmu_idx,
525 void *retaddr);
526
527 #define ACCESS_TYPE (NB_MMU_MODES + 1)
528 #define MEMSUFFIX _code
529 #define env cpu_single_env
530
531 #define DATA_SIZE 1
532 #include "softmmu_header.h"
533
534 #define DATA_SIZE 2
535 #include "softmmu_header.h"
536
537 #define DATA_SIZE 4
538 #include "softmmu_header.h"
539
540 #define DATA_SIZE 8
541 #include "softmmu_header.h"
542
543 #undef ACCESS_TYPE
544 #undef MEMSUFFIX
545 #undef env
546
547 #endif
548
549 #if defined(CONFIG_USER_ONLY)
550 static inline target_ulong get_phys_addr_code(CPUState *env, target_ulong addr)
551 {
552 return addr;
553 }
554 #else
555 /* NOTE: this function can trigger an exception */
556 /* NOTE2: the returned address is not exactly the physical address: it
557 is the offset relative to phys_ram_base */
558 static inline target_ulong get_phys_addr_code(CPUState *env, target_ulong addr)
559 {
560 int mmu_idx, index, pd;
561
562 index = (addr >> TARGET_PAGE_BITS) & (CPU_TLB_SIZE - 1);
563 mmu_idx = cpu_mmu_index(env);
564 if (__builtin_expect(env->tlb_table[mmu_idx][index].addr_code !=
565 (addr & TARGET_PAGE_MASK), 0)) {
566 ldub_code(addr);
567 }
568 pd = env->tlb_table[mmu_idx][index].addr_code & ~TARGET_PAGE_MASK;
569 if (pd > IO_MEM_ROM && !(pd & IO_MEM_ROMD)) {
570 #if defined(TARGET_SPARC) || defined(TARGET_MIPS)
571 do_unassigned_access(addr, 0, 1, 0);
572 #else
573 cpu_abort(env, "Trying to execute code outside RAM or ROM at 0x" TARGET_FMT_lx "\n", addr);
574 #endif
575 }
576 return addr + env->tlb_table[mmu_idx][index].addend - (unsigned long)phys_ram_base;
577 }
578 #endif
579
580 #ifdef USE_KQEMU
581 #define KQEMU_MODIFY_PAGE_MASK (0xff & ~(VGA_DIRTY_FLAG | CODE_DIRTY_FLAG))
582
583 int kqemu_init(CPUState *env);
584 int kqemu_cpu_exec(CPUState *env);
585 void kqemu_flush_page(CPUState *env, target_ulong addr);
586 void kqemu_flush(CPUState *env, int global);
587 void kqemu_set_notdirty(CPUState *env, ram_addr_t ram_addr);
588 void kqemu_modify_page(CPUState *env, ram_addr_t ram_addr);
589 void kqemu_cpu_interrupt(CPUState *env);
590 void kqemu_record_dump(void);
591
592 static inline int kqemu_is_ok(CPUState *env)
593 {
594 return(env->kqemu_enabled &&
595 (env->cr[0] & CR0_PE_MASK) &&
596 !(env->hflags & HF_INHIBIT_IRQ_MASK) &&
597 (env->eflags & IF_MASK) &&
598 !(env->eflags & VM_MASK) &&
599 (env->kqemu_enabled == 2 ||
600 ((env->hflags & HF_CPL_MASK) == 3 &&
601 (env->eflags & IOPL_MASK) != IOPL_MASK)));
602 }
603
604 #endif