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 */
25 #define xglue(x, y) x ## y
26 #define glue(x, y) xglue(x, y)
27 #define stringify(s) tostring(s)
28 #define tostring(s) #s
33 typedef uint32_t target_ulong;
37 #define __builtin_expect(x, n) (x)
41 #define REGPARM(n) __attribute((regparm(n)))
46 /* is_jmp field values */
47 #define DISAS_NEXT 0 /* next instruction can be analyzed */
48 #define DISAS_JUMP 1 /* only pc was modified dynamically */
49 #define DISAS_UPDATE 2 /* cpu state was modified dynamically */
50 #define DISAS_TB_JUMP 3 /* only pc was modified statically */
52 struct TranslationBlock;
54 /* XXX: make safe guess about sizes */
55 #define MAX_OP_PER_INSTR 32
56 #define OPC_BUF_SIZE 512
57 #define OPC_MAX_SIZE (OPC_BUF_SIZE - MAX_OP_PER_INSTR)
59 #define OPPARAM_BUF_SIZE (OPC_BUF_SIZE * 3)
61 extern uint16_t gen_opc_buf[OPC_BUF_SIZE];
62 extern uint32_t gen_opparam_buf[OPPARAM_BUF_SIZE];
63 extern uint32_t gen_opc_pc[OPC_BUF_SIZE];
64 extern uint8_t gen_opc_cc_op[OPC_BUF_SIZE];
65 extern uint8_t gen_opc_instr_start[OPC_BUF_SIZE];
67 #if defined(TARGET_I386)
69 void optimize_flags_init(void);
76 int gen_intermediate_code(CPUState *env, struct TranslationBlock *tb);
77 int gen_intermediate_code_pc(CPUState *env, struct TranslationBlock *tb);
78 void dump_ops(const uint16_t *opc_buf, const uint32_t *opparam_buf);
79 int cpu_gen_code(CPUState *env, struct TranslationBlock *tb,
80 int max_code_size, int *gen_code_size_ptr);
81 int cpu_restore_state(struct TranslationBlock *tb,
82 CPUState *env, unsigned long searched_pc);
83 void cpu_exec_init(void);
84 int page_unprotect(unsigned long address);
85 void tb_invalidate_page_range(target_ulong start, target_ulong end);
86 void tlb_flush_page(CPUState *env, uint32_t addr);
87 void tlb_flush_page_write(CPUState *env, uint32_t addr);
88 void tlb_flush(CPUState *env);
89 int tlb_set_page(CPUState *env, uint32_t vaddr, uint32_t paddr, int prot,
90 int is_user, int is_softmmu);
92 #define CODE_GEN_MAX_SIZE 65536
93 #define CODE_GEN_ALIGN 16 /* must be >= of the size of a icache line */
95 #define CODE_GEN_HASH_BITS 15
96 #define CODE_GEN_HASH_SIZE (1 << CODE_GEN_HASH_BITS)
98 #define CODE_GEN_PHYS_HASH_BITS 15
99 #define CODE_GEN_PHYS_HASH_SIZE (1 << CODE_GEN_PHYS_HASH_BITS)
101 /* maximum total translate dcode allocated */
103 /* NOTE: the translated code area cannot be too big because on some
104 archs the range of "fast" function calls are limited. Here is a
105 summary of the ranges:
107 i386 : signed 32 bits
110 sparc : signed 32 bits
111 alpha : signed 23 bits
114 #if defined(__alpha__)
115 #define CODE_GEN_BUFFER_SIZE (2 * 1024 * 1024)
116 #elif defined(__powerpc__)
117 #define CODE_GEN_BUFFER_SIZE (6 * 1024)
119 #define CODE_GEN_BUFFER_SIZE (8 * 1024 * 1024)
122 //#define CODE_GEN_BUFFER_SIZE (128 * 1024)
124 /* estimated block size for TB allocation */
125 /* XXX: use a per code average code fragment size and modulate it
126 according to the host CPU */
127 #if defined(CONFIG_SOFTMMU)
128 #define CODE_GEN_AVG_BLOCK_SIZE 128
130 #define CODE_GEN_AVG_BLOCK_SIZE 64
133 #define CODE_GEN_MAX_BLOCKS (CODE_GEN_BUFFER_SIZE / CODE_GEN_AVG_BLOCK_SIZE)
135 #if defined(__powerpc__)
136 #define USE_DIRECT_JUMP
138 #if defined(__i386__)
139 #define USE_DIRECT_JUMP
142 typedef struct TranslationBlock {
143 unsigned long pc; /* simulated PC corresponding to this block (EIP + CS base) */
144 unsigned long cs_base; /* CS base for this block */
145 unsigned int flags; /* flags defining in which context the code was generated */
146 uint16_t size; /* size of target code for this block (1 <=
147 size <= TARGET_PAGE_SIZE) */
148 uint8_t *tc_ptr; /* pointer to the translated code */
149 struct TranslationBlock *hash_next; /* next matching tb for virtual address */
150 /* next matching tb for physical address. */
151 struct TranslationBlock *phys_hash_next;
152 /* first and second physical page containing code. The lower bit
153 of the pointer tells the index in page_next[] */
154 struct TranslationBlock *page_next[2];
155 target_ulong page_addr[2];
157 /* the following data are used to directly call another TB from
158 the code of this one. */
159 uint16_t tb_next_offset[2]; /* offset of original jump target */
160 #ifdef USE_DIRECT_JUMP
161 uint16_t tb_jmp_offset[4]; /* offset of jump instruction */
163 uint32_t tb_next[2]; /* address of jump generated code */
165 /* list of TBs jumping to this one. This is a circular list using
166 the two least significant bits of the pointers to tell what is
167 the next pointer: 0 = jmp_next[0], 1 = jmp_next[1], 2 =
169 struct TranslationBlock *jmp_next[2];
170 struct TranslationBlock *jmp_first;
173 static inline unsigned int tb_hash_func(unsigned long pc)
175 return pc & (CODE_GEN_HASH_SIZE - 1);
178 static inline unsigned int tb_phys_hash_func(unsigned long pc)
180 return pc & (CODE_GEN_PHYS_HASH_SIZE - 1);
183 TranslationBlock *tb_alloc(unsigned long pc);
184 void tb_flush(CPUState *env);
185 void tb_link(TranslationBlock *tb);
186 void tb_link_phys(TranslationBlock *tb,
187 target_ulong phys_pc, target_ulong phys_page2);
189 extern TranslationBlock *tb_hash[CODE_GEN_HASH_SIZE];
190 extern TranslationBlock *tb_phys_hash[CODE_GEN_PHYS_HASH_SIZE];
192 extern uint8_t code_gen_buffer[CODE_GEN_BUFFER_SIZE];
193 extern uint8_t *code_gen_ptr;
195 /* find a translation block in the translation cache. If not found,
196 return NULL and the pointer to the last element of the list in pptb */
197 static inline TranslationBlock *tb_find(TranslationBlock ***pptb,
199 unsigned long cs_base,
202 TranslationBlock **ptb, *tb;
205 h = tb_hash_func(pc);
211 if (tb->pc == pc && tb->cs_base == cs_base && tb->flags == flags)
213 ptb = &tb->hash_next;
220 #if defined(USE_DIRECT_JUMP)
222 #if defined(__powerpc__)
223 static inline void tb_set_jmp_target1(unsigned long jmp_addr, unsigned long addr)
227 /* patch the branch destination */
228 ptr = (uint32_t *)jmp_addr;
230 val = (val & ~0x03fffffc) | ((addr - jmp_addr) & 0x03fffffc);
233 asm volatile ("dcbst 0,%0" : : "r"(ptr) : "memory");
234 asm volatile ("sync" : : : "memory");
235 asm volatile ("icbi 0,%0" : : "r"(ptr) : "memory");
236 asm volatile ("sync" : : : "memory");
237 asm volatile ("isync" : : : "memory");
239 #elif defined(__i386__)
240 static inline void tb_set_jmp_target1(unsigned long jmp_addr, unsigned long addr)
242 /* patch the branch destination */
243 *(uint32_t *)jmp_addr = addr - (jmp_addr + 4);
244 /* no need to flush icache explicitely */
248 static inline void tb_set_jmp_target(TranslationBlock *tb,
249 int n, unsigned long addr)
251 unsigned long offset;
253 offset = tb->tb_jmp_offset[n];
254 tb_set_jmp_target1((unsigned long)(tb->tc_ptr + offset), addr);
255 offset = tb->tb_jmp_offset[n + 2];
256 if (offset != 0xffff)
257 tb_set_jmp_target1((unsigned long)(tb->tc_ptr + offset), addr);
262 /* set the jump target */
263 static inline void tb_set_jmp_target(TranslationBlock *tb,
264 int n, unsigned long addr)
266 tb->tb_next[n] = addr;
271 static inline void tb_add_jump(TranslationBlock *tb, int n,
272 TranslationBlock *tb_next)
274 /* NOTE: this test is only needed for thread safety */
275 if (!tb->jmp_next[n]) {
276 /* patch the native jump address */
277 tb_set_jmp_target(tb, n, (unsigned long)tb_next->tc_ptr);
279 /* add in TB jmp circular list */
280 tb->jmp_next[n] = tb_next->jmp_first;
281 tb_next->jmp_first = (TranslationBlock *)((long)(tb) | (n));
285 TranslationBlock *tb_find_pc(unsigned long pc_ptr);
288 #define offsetof(type, field) ((size_t) &((type *)0)->field)
291 #if defined(__powerpc__)
293 /* we patch the jump instruction directly */
294 #define JUMP_TB(opname, tbparam, n, eip)\
296 asm volatile (".section \".data\"\n"\
297 "__op_label" #n "." stringify(opname) ":\n"\
300 "b __op_jmp" #n "\n"\
302 T0 = (long)(tbparam) + (n);\
307 #define JUMP_TB2(opname, tbparam, n)\
309 asm volatile ("b __op_jmp" #n "\n");\
312 #elif defined(__i386__) && defined(USE_DIRECT_JUMP)
314 /* we patch the jump instruction directly */
315 #define JUMP_TB(opname, tbparam, n, eip)\
317 asm volatile (".section \".data\"\n"\
318 "__op_label" #n "." stringify(opname) ":\n"\
321 "jmp __op_jmp" #n "\n"\
323 T0 = (long)(tbparam) + (n);\
328 #define JUMP_TB2(opname, tbparam, n)\
330 asm volatile ("jmp __op_jmp" #n "\n");\
335 /* jump to next block operations (more portable code, does not need
336 cache flushing, but slower because of indirect jump) */
337 #define JUMP_TB(opname, tbparam, n, eip)\
339 static void __attribute__((unused)) *__op_label ## n = &&label ## n;\
340 static void __attribute__((unused)) *dummy ## n = &&dummy_label ## n;\
341 goto *(void *)(((TranslationBlock *)tbparam)->tb_next[n]);\
343 T0 = (long)(tbparam) + (n);\
349 /* second jump to same destination 'n' */
350 #define JUMP_TB2(opname, tbparam, n)\
352 goto *(void *)(((TranslationBlock *)tbparam)->tb_next[n - 2]);\
357 extern CPUWriteMemoryFunc *io_mem_write[IO_MEM_NB_ENTRIES][4];
358 extern CPUReadMemoryFunc *io_mem_read[IO_MEM_NB_ENTRIES][4];
361 static inline int testandset (int *p)
364 __asm__ __volatile__ (
372 : "r" (p), "r" (1), "r" (0)
379 static inline int testandset (int *p)
384 __asm__ __volatile__ ("lock; cmpxchgl %3, %1; sete %0"
385 : "=q" (ret), "=m" (*p), "=a" (readval)
386 : "r" (1), "m" (*p), "a" (0)
393 static inline int testandset (int *p)
397 __asm__ __volatile__ ("0: cs %0,%1,0(%2)\n"
400 : "r" (1), "a" (p), "0" (*p)
407 static inline int testandset (int *p)
412 __asm__ __volatile__ ("0: mov 1,%2\n"
419 : "=r" (ret), "=m" (*p), "=r" (one)
426 static inline int testandset (int *p)
430 __asm__ __volatile__("ldstub [%1], %0"
435 return (ret ? 1 : 0);
440 static inline int testandset (int *spinlock)
442 register unsigned int ret;
443 __asm__ __volatile__("swp %0, %1, [%2]"
445 : "0"(1), "r"(spinlock));
452 static inline int testandset (int *p)
455 __asm__ __volatile__("tas %1; sne %0"
463 typedef int spinlock_t;
465 #define SPIN_LOCK_UNLOCKED 0
467 #if defined(CONFIG_USER_ONLY)
468 static inline void spin_lock(spinlock_t *lock)
470 while (testandset(lock));
473 static inline void spin_unlock(spinlock_t *lock)
478 static inline int spin_trylock(spinlock_t *lock)
480 return !testandset(lock);
483 static inline void spin_lock(spinlock_t *lock)
487 static inline void spin_unlock(spinlock_t *lock)
491 static inline int spin_trylock(spinlock_t *lock)
497 extern spinlock_t tb_lock;
499 extern int tb_invalidated_flag;
501 #if defined(TARGET_I386) && !defined(CONFIG_USER_ONLY)
503 void tlb_fill(unsigned long addr, int is_write, int is_user,
506 #define ACCESS_TYPE 3
507 #define MEMSUFFIX _code
508 #define env cpu_single_env
511 #include "softmmu_header.h"
514 #include "softmmu_header.h"
517 #include "softmmu_header.h"
525 #if defined(CONFIG_USER_ONLY)
526 static inline target_ulong get_phys_addr_code(CPUState *env, target_ulong addr)
531 /* NOTE: this function can trigger an exception */
532 /* XXX: i386 target specific */
533 static inline target_ulong get_phys_addr_code(CPUState *env, target_ulong addr)
537 index = (addr >> TARGET_PAGE_BITS) & (CPU_TLB_SIZE - 1);
538 is_user = ((env->hflags & HF_CPL_MASK) == 3);
539 if (__builtin_expect(env->tlb_read[is_user][index].address !=
540 (addr & TARGET_PAGE_MASK), 0)) {
541 ldub_code((void *)addr);
543 return addr + env->tlb_read[is_user][index].addend - (unsigned long)phys_ram_base;