Fix cas on i386
[qemu-kvm/fedora.git] / exec-all.h
blob51b27b514114f6fcdfbad332466a9086c394a6a3
1 /*
2 * internal execution defines for qemu
4 * Copyright (c) 2003 Fabrice Bellard
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.
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.
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
21 /* allow to see translation results - the slowdown should be negligible, so we leave it */
22 #define DEBUG_DISAS
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 */
30 struct TranslationBlock;
32 /* XXX: make safe guess about sizes */
33 #define MAX_OP_PER_INSTR 64
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)
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
45 #define OPPARAM_BUF_SIZE (OPC_BUF_SIZE * MAX_OPC_PARAM)
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];
54 typedef void (GenOpFunc)(void);
55 typedef void (GenOpFunc1)(long);
56 typedef void (GenOpFunc2)(long, long);
57 typedef void (GenOpFunc3)(long, long, long);
59 #if defined(TARGET_I386)
61 void optimize_flags_init(void);
63 #endif
65 extern FILE *logfile;
66 extern int loglevel;
68 int gen_intermediate_code(CPUState *env, struct TranslationBlock *tb);
69 int gen_intermediate_code_pc(CPUState *env, struct TranslationBlock *tb);
70 void gen_pc_load(CPUState *env, struct TranslationBlock *tb,
71 unsigned long searched_pc, int pc_pos, void *puc);
73 unsigned long code_gen_max_block_size(void);
74 void cpu_gen_init(void);
75 int cpu_gen_code(CPUState *env, struct TranslationBlock *tb,
76 int *gen_code_size_ptr);
77 int cpu_restore_state(struct TranslationBlock *tb,
78 CPUState *env, unsigned long searched_pc,
79 void *puc);
80 int cpu_restore_state_copy(struct TranslationBlock *tb,
81 CPUState *env, unsigned long searched_pc,
82 void *puc);
83 void cpu_resume_from_signal(CPUState *env1, void *puc);
84 void cpu_exec_init(CPUState *env);
85 int page_unprotect(target_ulong address, unsigned long pc, void *puc);
86 void tb_invalidate_phys_page_range(target_phys_addr_t start, target_phys_addr_t end,
87 int is_cpu_write_access);
88 void tb_invalidate_page_range(target_ulong start, target_ulong end);
89 void tlb_flush_page(CPUState *env, target_ulong addr);
90 void tlb_flush(CPUState *env, int flush_global);
91 int tlb_set_page_exec(CPUState *env, target_ulong vaddr,
92 target_phys_addr_t paddr, int prot,
93 int mmu_idx, int is_softmmu);
94 static inline int tlb_set_page(CPUState *env1, target_ulong vaddr,
95 target_phys_addr_t paddr, int prot,
96 int mmu_idx, int is_softmmu)
98 if (prot & PAGE_READ)
99 prot |= PAGE_EXEC;
100 return tlb_set_page_exec(env1, vaddr, paddr, prot, mmu_idx, is_softmmu);
103 #define CODE_GEN_ALIGN 16 /* must be >= of the size of a icache line */
105 #define CODE_GEN_PHYS_HASH_BITS 15
106 #define CODE_GEN_PHYS_HASH_SIZE (1 << CODE_GEN_PHYS_HASH_BITS)
108 /* maximum total translate dcode allocated */
110 /* NOTE: the translated code area cannot be too big because on some
111 archs the range of "fast" function calls is limited. Here is a
112 summary of the ranges:
114 i386 : signed 32 bits
115 arm : signed 26 bits
116 ppc : signed 24 bits
117 sparc : signed 32 bits
118 alpha : signed 23 bits
121 #if defined(__alpha__)
122 #define CODE_GEN_BUFFER_SIZE (2 * 1024 * 1024)
123 #elif defined(__ia64)
124 #define CODE_GEN_BUFFER_SIZE (4 * 1024 * 1024) /* range of addl */
125 #elif defined(__powerpc__)
126 #define CODE_GEN_BUFFER_SIZE (6 * 1024 * 1024)
127 #else
128 /* XXX: make it dynamic on x86 */
129 #define CODE_GEN_BUFFER_SIZE (16 * 1024 * 1024)
130 #endif
132 //#define CODE_GEN_BUFFER_SIZE (128 * 1024)
134 /* estimated block size for TB allocation */
135 /* XXX: use a per code average code fragment size and modulate it
136 according to the host CPU */
137 #if defined(CONFIG_SOFTMMU)
138 #define CODE_GEN_AVG_BLOCK_SIZE 128
139 #else
140 #define CODE_GEN_AVG_BLOCK_SIZE 64
141 #endif
143 #define CODE_GEN_MAX_BLOCKS (CODE_GEN_BUFFER_SIZE / CODE_GEN_AVG_BLOCK_SIZE)
145 #if defined(__powerpc__) || defined(__x86_64__) || defined(__arm__)
146 #define USE_DIRECT_JUMP
147 #endif
148 #if defined(__i386__) && !defined(_WIN32)
149 #define USE_DIRECT_JUMP
150 #endif
152 typedef struct TranslationBlock {
153 target_ulong pc; /* simulated PC corresponding to this block (EIP + CS base) */
154 target_ulong cs_base; /* CS base for this block */
155 uint64_t flags; /* flags defining in which context the code was generated */
156 uint16_t size; /* size of target code for this block (1 <=
157 size <= TARGET_PAGE_SIZE) */
158 uint16_t cflags; /* compile flags */
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 */
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];
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;
187 static inline unsigned int tb_jmp_cache_hash_page(target_ulong pc)
189 target_ulong tmp;
190 tmp = pc ^ (pc >> (TARGET_PAGE_BITS - TB_JMP_PAGE_BITS));
191 return (tmp >> (TARGET_PAGE_BITS - TB_JMP_PAGE_BITS)) & TB_JMP_PAGE_MASK;
194 static inline unsigned int tb_jmp_cache_hash_func(target_ulong pc)
196 target_ulong tmp;
197 tmp = pc ^ (pc >> (TARGET_PAGE_BITS - TB_JMP_PAGE_BITS));
198 return (((tmp >> (TARGET_PAGE_BITS - TB_JMP_PAGE_BITS)) & TB_JMP_PAGE_MASK)
199 | (tmp & TB_JMP_ADDR_MASK));
202 static inline unsigned int tb_phys_hash_func(unsigned long pc)
204 return pc & (CODE_GEN_PHYS_HASH_SIZE - 1);
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);
212 extern TranslationBlock *tb_phys_hash[CODE_GEN_PHYS_HASH_SIZE];
214 extern uint8_t code_gen_buffer[CODE_GEN_BUFFER_SIZE];
215 extern uint8_t *code_gen_ptr;
217 #if defined(USE_DIRECT_JUMP)
219 #if defined(__powerpc__)
220 static inline void tb_set_jmp_target1(unsigned long jmp_addr, unsigned long addr)
222 uint32_t val, *ptr;
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");
236 #elif defined(__i386__) || defined(__x86_64__)
237 static inline void tb_set_jmp_target1(unsigned long jmp_addr, unsigned long addr)
239 /* patch the branch destination */
240 *(uint32_t *)jmp_addr = addr - (jmp_addr + 4);
241 /* no need to flush icache explicitely */
243 #elif defined(__arm__)
244 static inline void tb_set_jmp_target1(unsigned long jmp_addr, unsigned long addr)
246 register unsigned long _beg __asm ("a1");
247 register unsigned long _end __asm ("a2");
248 register unsigned long _flg __asm ("a3");
250 /* we could use a ldr pc, [pc, #-4] kind of branch and avoid the flush */
251 *(uint32_t *)jmp_addr |= ((addr - (jmp_addr + 8)) >> 2) & 0xffffff;
253 /* flush icache */
254 _beg = jmp_addr;
255 _end = jmp_addr + 4;
256 _flg = 0;
257 __asm __volatile__ ("swi 0x9f0002" : : "r" (_beg), "r" (_end), "r" (_flg));
259 #endif
261 static inline void tb_set_jmp_target(TranslationBlock *tb,
262 int n, unsigned long addr)
264 unsigned long offset;
266 offset = tb->tb_jmp_offset[n];
267 tb_set_jmp_target1((unsigned long)(tb->tc_ptr + offset), addr);
268 offset = tb->tb_jmp_offset[n + 2];
269 if (offset != 0xffff)
270 tb_set_jmp_target1((unsigned long)(tb->tc_ptr + offset), addr);
273 #else
275 /* set the jump target */
276 static inline void tb_set_jmp_target(TranslationBlock *tb,
277 int n, unsigned long addr)
279 tb->tb_next[n] = addr;
282 #endif
284 static inline void tb_add_jump(TranslationBlock *tb, int n,
285 TranslationBlock *tb_next)
287 /* NOTE: this test is only needed for thread safety */
288 if (!tb->jmp_next[n]) {
289 /* patch the native jump address */
290 tb_set_jmp_target(tb, n, (unsigned long)tb_next->tc_ptr);
292 /* add in TB jmp circular list */
293 tb->jmp_next[n] = tb_next->jmp_first;
294 tb_next->jmp_first = (TranslationBlock *)((long)(tb) | (n));
298 TranslationBlock *tb_find_pc(unsigned long pc_ptr);
300 #ifndef offsetof
301 #define offsetof(type, field) ((size_t) &((type *)0)->field)
302 #endif
304 #if defined(_WIN32)
305 #define ASM_DATA_SECTION ".section \".data\"\n"
306 #define ASM_PREVIOUS_SECTION ".section .text\n"
307 #elif defined(__APPLE__)
308 #define ASM_DATA_SECTION ".data\n"
309 #define ASM_PREVIOUS_SECTION ".text\n"
310 #else
311 #define ASM_DATA_SECTION ".section \".data\"\n"
312 #define ASM_PREVIOUS_SECTION ".previous\n"
313 #endif
315 #define ASM_OP_LABEL_NAME(n, opname) \
316 ASM_NAME(__op_label) #n "." ASM_NAME(opname)
318 extern CPUWriteMemoryFunc *io_mem_write[IO_MEM_NB_ENTRIES][4];
319 extern CPUReadMemoryFunc *io_mem_read[IO_MEM_NB_ENTRIES][4];
320 extern void *io_mem_opaque[IO_MEM_NB_ENTRIES];
322 #if defined(__hppa__)
324 typedef int spinlock_t[4];
326 #define SPIN_LOCK_UNLOCKED { 1, 1, 1, 1 }
328 static inline void resetlock (spinlock_t *p)
330 (*p)[0] = (*p)[1] = (*p)[2] = (*p)[3] = 1;
333 #else
335 typedef int spinlock_t;
337 #define SPIN_LOCK_UNLOCKED 0
339 static inline void resetlock (spinlock_t *p)
341 *p = SPIN_LOCK_UNLOCKED;
344 #endif
346 #if defined(__powerpc__)
347 static inline int testandset (int *p)
349 int ret;
350 __asm__ __volatile__ (
351 "0: lwarx %0,0,%1\n"
352 " xor. %0,%3,%0\n"
353 " bne 1f\n"
354 " stwcx. %2,0,%1\n"
355 " bne- 0b\n"
356 "1: "
357 : "=&r" (ret)
358 : "r" (p), "r" (1), "r" (0)
359 : "cr0", "memory");
360 return ret;
362 #elif defined(__i386__)
363 static inline int testandset (int *p)
365 long int readval = 0;
367 __asm__ __volatile__ ("lock; cmpxchgl %2, %0"
368 : "+m" (*p), "+a" (readval)
369 : "r" (1)
370 : "cc");
371 return readval;
373 #elif defined(__x86_64__)
374 static inline int testandset (int *p)
376 long int readval = 0;
378 __asm__ __volatile__ ("lock; cmpxchgl %2, %0"
379 : "+m" (*p), "+a" (readval)
380 : "r" (1)
381 : "cc");
382 return readval;
384 #elif defined(__s390__)
385 static inline int testandset (int *p)
387 int ret;
389 __asm__ __volatile__ ("0: cs %0,%1,0(%2)\n"
390 " jl 0b"
391 : "=&d" (ret)
392 : "r" (1), "a" (p), "0" (*p)
393 : "cc", "memory" );
394 return ret;
396 #elif defined(__alpha__)
397 static inline int testandset (int *p)
399 int ret;
400 unsigned long one;
402 __asm__ __volatile__ ("0: mov 1,%2\n"
403 " ldl_l %0,%1\n"
404 " stl_c %2,%1\n"
405 " beq %2,1f\n"
406 ".subsection 2\n"
407 "1: br 0b\n"
408 ".previous"
409 : "=r" (ret), "=m" (*p), "=r" (one)
410 : "m" (*p));
411 return ret;
413 #elif defined(__sparc__)
414 static inline int testandset (int *p)
416 int ret;
418 __asm__ __volatile__("ldstub [%1], %0"
419 : "=r" (ret)
420 : "r" (p)
421 : "memory");
423 return (ret ? 1 : 0);
425 #elif defined(__arm__)
426 static inline int testandset (int *spinlock)
428 register unsigned int ret;
429 __asm__ __volatile__("swp %0, %1, [%2]"
430 : "=r"(ret)
431 : "0"(1), "r"(spinlock));
433 return ret;
435 #elif defined(__mc68000)
436 static inline int testandset (int *p)
438 char ret;
439 __asm__ __volatile__("tas %1; sne %0"
440 : "=r" (ret)
441 : "m" (p)
442 : "cc","memory");
443 return ret;
445 #elif defined(__hppa__)
447 /* Because malloc only guarantees 8-byte alignment for malloc'd data,
448 and GCC only guarantees 8-byte alignment for stack locals, we can't
449 be assured of 16-byte alignment for atomic lock data even if we
450 specify "__attribute ((aligned(16)))" in the type declaration. So,
451 we use a struct containing an array of four ints for the atomic lock
452 type and dynamically select the 16-byte aligned int from the array
453 for the semaphore. */
454 #define __PA_LDCW_ALIGNMENT 16
455 static inline void *ldcw_align (void *p) {
456 unsigned long a = (unsigned long)p;
457 a = (a + __PA_LDCW_ALIGNMENT - 1) & ~(__PA_LDCW_ALIGNMENT - 1);
458 return (void *)a;
461 static inline int testandset (spinlock_t *p)
463 unsigned int ret;
464 p = ldcw_align(p);
465 __asm__ __volatile__("ldcw 0(%1),%0"
466 : "=r" (ret)
467 : "r" (p)
468 : "memory" );
469 return !ret;
472 #elif defined(__ia64)
474 #include <ia64intrin.h>
476 static inline int testandset (int *p)
478 return __sync_lock_test_and_set (p, 1);
480 #elif defined(__mips__)
481 static inline int testandset (int *p)
483 int ret;
485 __asm__ __volatile__ (
486 " .set push \n"
487 " .set noat \n"
488 " .set mips2 \n"
489 "1: li $1, 1 \n"
490 " ll %0, %1 \n"
491 " sc $1, %1 \n"
492 " beqz $1, 1b \n"
493 " .set pop "
494 : "=r" (ret), "+R" (*p)
496 : "memory");
498 return ret;
500 #else
501 #error unimplemented CPU support
502 #endif
504 #if defined(CONFIG_USER_ONLY)
505 static inline void spin_lock(spinlock_t *lock)
507 while (testandset(lock));
510 static inline void spin_unlock(spinlock_t *lock)
512 resetlock(lock);
515 static inline int spin_trylock(spinlock_t *lock)
517 return !testandset(lock);
519 #else
520 static inline void spin_lock(spinlock_t *lock)
524 static inline void spin_unlock(spinlock_t *lock)
528 static inline int spin_trylock(spinlock_t *lock)
530 return 1;
532 #endif
534 extern spinlock_t tb_lock;
536 extern int tb_invalidated_flag;
538 #if !defined(CONFIG_USER_ONLY)
540 void tlb_fill(target_ulong addr, int is_write, int mmu_idx,
541 void *retaddr);
543 #define ACCESS_TYPE (NB_MMU_MODES + 1)
544 #define MEMSUFFIX _code
545 #define env cpu_single_env
547 #define DATA_SIZE 1
548 #include "softmmu_header.h"
550 #define DATA_SIZE 2
551 #include "softmmu_header.h"
553 #define DATA_SIZE 4
554 #include "softmmu_header.h"
556 #define DATA_SIZE 8
557 #include "softmmu_header.h"
559 #undef ACCESS_TYPE
560 #undef MEMSUFFIX
561 #undef env
563 #endif
565 #if defined(CONFIG_USER_ONLY)
566 static inline target_ulong get_phys_addr_code(CPUState *env1, target_ulong addr)
568 return addr;
570 #else
571 /* NOTE: this function can trigger an exception */
572 /* NOTE2: the returned address is not exactly the physical address: it
573 is the offset relative to phys_ram_base */
574 static inline target_ulong get_phys_addr_code(CPUState *env1, target_ulong addr)
576 int mmu_idx, page_index, pd;
578 page_index = (addr >> TARGET_PAGE_BITS) & (CPU_TLB_SIZE - 1);
579 mmu_idx = cpu_mmu_index(env1);
580 if (__builtin_expect(env1->tlb_table[mmu_idx][page_index].addr_code !=
581 (addr & TARGET_PAGE_MASK), 0)) {
582 ldub_code(addr);
584 pd = env1->tlb_table[mmu_idx][page_index].addr_code & ~TARGET_PAGE_MASK;
585 if (pd > IO_MEM_ROM && !(pd & IO_MEM_ROMD)) {
586 #if defined(TARGET_SPARC) || defined(TARGET_MIPS)
587 do_unassigned_access(addr, 0, 1, 0);
588 #else
589 cpu_abort(env1, "Trying to execute code outside RAM or ROM at 0x" TARGET_FMT_lx "\n", addr);
590 #endif
592 return addr + env1->tlb_table[mmu_idx][page_index].addend - (unsigned long)phys_ram_base;
594 #endif
596 #ifdef USE_KQEMU
597 #define KQEMU_MODIFY_PAGE_MASK (0xff & ~(VGA_DIRTY_FLAG | CODE_DIRTY_FLAG))
599 int kqemu_init(CPUState *env);
600 int kqemu_cpu_exec(CPUState *env);
601 void kqemu_flush_page(CPUState *env, target_ulong addr);
602 void kqemu_flush(CPUState *env, int global);
603 void kqemu_set_notdirty(CPUState *env, ram_addr_t ram_addr);
604 void kqemu_modify_page(CPUState *env, ram_addr_t ram_addr);
605 void kqemu_cpu_interrupt(CPUState *env);
606 void kqemu_record_dump(void);
608 static inline int kqemu_is_ok(CPUState *env)
610 return(env->kqemu_enabled &&
611 (env->cr[0] & CR0_PE_MASK) &&
612 !(env->hflags & HF_INHIBIT_IRQ_MASK) &&
613 (env->eflags & IF_MASK) &&
614 !(env->eflags & VM_MASK) &&
615 (env->kqemu_enabled == 2 ||
616 ((env->hflags & HF_CPL_MASK) == 3 &&
617 (env->eflags & IOPL_MASK) != IOPL_MASK)));
620 #endif