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, see <http://www.gnu.org/licenses/>.
23 #include "qemu-common.h"
25 /* allow to see translation results - the slowdown should be negligible, so we leave it */
28 /* Page tracking code uses ram addresses in system mode, and virtual
29 addresses in userspace mode. Define tb_page_addr_t to be an appropriate
31 #if defined(CONFIG_USER_ONLY)
32 typedef abi_ulong tb_page_addr_t
;
34 typedef ram_addr_t tb_page_addr_t
;
37 /* is_jmp field values */
38 #define DISAS_NEXT 0 /* next instruction can be analyzed */
39 #define DISAS_JUMP 1 /* only pc was modified dynamically */
40 #define DISAS_UPDATE 2 /* cpu state was modified dynamically */
41 #define DISAS_TB_JUMP 3 /* only pc was modified statically */
43 struct TranslationBlock
;
44 typedef struct TranslationBlock TranslationBlock
;
46 /* XXX: make safe guess about sizes */
47 #define MAX_OP_PER_INSTR 208
49 #if HOST_LONG_BITS == 32
50 #define MAX_OPC_PARAM_PER_ARG 2
52 #define MAX_OPC_PARAM_PER_ARG 1
54 #define MAX_OPC_PARAM_IARGS 5
55 #define MAX_OPC_PARAM_OARGS 1
56 #define MAX_OPC_PARAM_ARGS (MAX_OPC_PARAM_IARGS + MAX_OPC_PARAM_OARGS)
58 /* A Call op needs up to 4 + 2N parameters on 32-bit archs,
59 * and up to 4 + N parameters on 64-bit archs
60 * (N = number of input arguments + output arguments). */
61 #define MAX_OPC_PARAM (4 + (MAX_OPC_PARAM_PER_ARG * MAX_OPC_PARAM_ARGS))
62 #define OPC_BUF_SIZE 640
63 #define OPC_MAX_SIZE (OPC_BUF_SIZE - MAX_OP_PER_INSTR)
65 /* Maximum size a TCG op can expand to. This is complicated because a
66 single op may require several host instructions and register reloads.
67 For now take a wild guess at 192 bytes, which should allow at least
68 a couple of fixup instructions per argument. */
69 #define TCG_MAX_OP_SIZE 192
71 #define OPPARAM_BUF_SIZE (OPC_BUF_SIZE * MAX_OPC_PARAM)
73 extern target_ulong gen_opc_pc
[OPC_BUF_SIZE
];
74 extern uint8_t gen_opc_instr_start
[OPC_BUF_SIZE
];
75 extern uint16_t gen_opc_icount
[OPC_BUF_SIZE
];
79 void gen_intermediate_code(CPUArchState
*env
, struct TranslationBlock
*tb
);
80 void gen_intermediate_code_pc(CPUArchState
*env
, struct TranslationBlock
*tb
);
81 void restore_state_to_opc(CPUArchState
*env
, struct TranslationBlock
*tb
,
84 void cpu_gen_init(void);
85 int cpu_gen_code(CPUArchState
*env
, struct TranslationBlock
*tb
,
86 int *gen_code_size_ptr
);
87 int cpu_restore_state(struct TranslationBlock
*tb
,
88 CPUArchState
*env
, uintptr_t searched_pc
);
89 void QEMU_NORETURN
cpu_resume_from_signal(CPUArchState
*env1
, void *puc
);
90 void QEMU_NORETURN
cpu_io_recompile(CPUArchState
*env
, uintptr_t retaddr
);
91 TranslationBlock
*tb_gen_code(CPUArchState
*env
,
92 target_ulong pc
, target_ulong cs_base
, int flags
,
94 void cpu_exec_init(CPUArchState
*env
);
95 void QEMU_NORETURN
cpu_loop_exit(CPUArchState
*env1
);
96 int page_unprotect(target_ulong address
, uintptr_t pc
, void *puc
);
97 void tb_invalidate_phys_page_range(tb_page_addr_t start
, tb_page_addr_t end
,
98 int is_cpu_write_access
);
99 void tb_invalidate_phys_range(tb_page_addr_t start
, tb_page_addr_t end
,
100 int is_cpu_write_access
);
101 #if !defined(CONFIG_USER_ONLY)
103 void tlb_flush_page(CPUArchState
*env
, target_ulong addr
);
104 void tlb_flush(CPUArchState
*env
, int flush_global
);
105 void tlb_set_page(CPUArchState
*env
, target_ulong vaddr
,
106 hwaddr paddr
, int prot
,
107 int mmu_idx
, target_ulong size
);
108 void tb_invalidate_phys_addr(hwaddr addr
);
110 static inline void tlb_flush_page(CPUArchState
*env
, target_ulong addr
)
114 static inline void tlb_flush(CPUArchState
*env
, int flush_global
)
119 #define CODE_GEN_ALIGN 16 /* must be >= of the size of a icache line */
121 #define CODE_GEN_PHYS_HASH_BITS 15
122 #define CODE_GEN_PHYS_HASH_SIZE (1 << CODE_GEN_PHYS_HASH_BITS)
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 #if defined(__arm__) || defined(_ARCH_PPC) \
134 || defined(__x86_64__) || defined(__i386__) \
135 || defined(__sparc__) \
136 || defined(CONFIG_TCG_INTERPRETER)
137 #define USE_DIRECT_JUMP
140 struct TranslationBlock
{
141 target_ulong pc
; /* simulated PC corresponding to this block (EIP + CS base) */
142 target_ulong cs_base
; /* CS base for this block */
143 uint64_t flags
; /* flags defining in which context the code was generated */
144 uint16_t size
; /* size of target code for this block (1 <=
145 size <= TARGET_PAGE_SIZE) */
146 uint16_t cflags
; /* compile flags */
147 #define CF_COUNT_MASK 0x7fff
148 #define CF_LAST_IO 0x8000 /* Last insn may be an IO access. */
150 uint8_t *tc_ptr
; /* pointer to the translated code */
151 /* next matching tb for physical address. */
152 struct TranslationBlock
*phys_hash_next
;
153 /* first and second physical page containing code. The lower bit
154 of the pointer tells the index in page_next[] */
155 struct TranslationBlock
*page_next
[2];
156 tb_page_addr_t page_addr
[2];
158 /* the following data are used to directly call another TB from
159 the code of this one. */
160 uint16_t tb_next_offset
[2]; /* offset of original jump target */
161 #ifdef USE_DIRECT_JUMP
162 uint16_t tb_jmp_offset
[2]; /* offset of jump instruction */
164 uintptr_t tb_next
[2]; /* address of jump generated code */
166 /* list of TBs jumping to this one. This is a circular list using
167 the two least significant bits of the pointers to tell what is
168 the next pointer: 0 = jmp_next[0], 1 = jmp_next[1], 2 =
170 struct TranslationBlock
*jmp_next
[2];
171 struct TranslationBlock
*jmp_first
;
175 static inline unsigned int tb_jmp_cache_hash_page(target_ulong pc
)
178 tmp
= pc
^ (pc
>> (TARGET_PAGE_BITS
- TB_JMP_PAGE_BITS
));
179 return (tmp
>> (TARGET_PAGE_BITS
- TB_JMP_PAGE_BITS
)) & TB_JMP_PAGE_MASK
;
182 static inline unsigned int tb_jmp_cache_hash_func(target_ulong pc
)
185 tmp
= pc
^ (pc
>> (TARGET_PAGE_BITS
- TB_JMP_PAGE_BITS
));
186 return (((tmp
>> (TARGET_PAGE_BITS
- TB_JMP_PAGE_BITS
)) & TB_JMP_PAGE_MASK
)
187 | (tmp
& TB_JMP_ADDR_MASK
));
190 static inline unsigned int tb_phys_hash_func(tb_page_addr_t pc
)
192 return (pc
>> 2) & (CODE_GEN_PHYS_HASH_SIZE
- 1);
195 void tb_free(TranslationBlock
*tb
);
196 void tb_flush(CPUArchState
*env
);
197 void tb_phys_invalidate(TranslationBlock
*tb
, tb_page_addr_t page_addr
);
199 extern TranslationBlock
*tb_phys_hash
[CODE_GEN_PHYS_HASH_SIZE
];
201 #if defined(USE_DIRECT_JUMP)
203 #if defined(CONFIG_TCG_INTERPRETER)
204 static inline void tb_set_jmp_target1(uintptr_t jmp_addr
, uintptr_t addr
)
206 /* patch the branch destination */
207 *(uint32_t *)jmp_addr
= addr
- (jmp_addr
+ 4);
208 /* no need to flush icache explicitly */
210 #elif defined(_ARCH_PPC)
211 void ppc_tb_set_jmp_target(unsigned long jmp_addr
, unsigned long addr
);
212 #define tb_set_jmp_target1 ppc_tb_set_jmp_target
213 #elif defined(__i386__) || defined(__x86_64__)
214 static inline void tb_set_jmp_target1(uintptr_t jmp_addr
, uintptr_t addr
)
216 /* patch the branch destination */
217 *(uint32_t *)jmp_addr
= addr
- (jmp_addr
+ 4);
218 /* no need to flush icache explicitly */
220 #elif defined(__arm__)
221 static inline void tb_set_jmp_target1(uintptr_t jmp_addr
, uintptr_t addr
)
223 #if !QEMU_GNUC_PREREQ(4, 1)
224 register unsigned long _beg
__asm ("a1");
225 register unsigned long _end
__asm ("a2");
226 register unsigned long _flg
__asm ("a3");
229 /* we could use a ldr pc, [pc, #-4] kind of branch and avoid the flush */
230 *(uint32_t *)jmp_addr
=
231 (*(uint32_t *)jmp_addr
& ~0xffffff)
232 | (((addr
- (jmp_addr
+ 8)) >> 2) & 0xffffff);
234 #if QEMU_GNUC_PREREQ(4, 1)
235 __builtin___clear_cache((char *) jmp_addr
, (char *) jmp_addr
+ 4);
241 __asm
__volatile__ ("swi 0x9f0002" : : "r" (_beg
), "r" (_end
), "r" (_flg
));
244 #elif defined(__sparc__)
245 void tb_set_jmp_target1(uintptr_t jmp_addr
, uintptr_t addr
);
247 #error tb_set_jmp_target1 is missing
250 static inline void tb_set_jmp_target(TranslationBlock
*tb
,
251 int n
, uintptr_t addr
)
253 uint16_t offset
= tb
->tb_jmp_offset
[n
];
254 tb_set_jmp_target1((uintptr_t)(tb
->tc_ptr
+ offset
), addr
);
259 /* set the jump target */
260 static inline void tb_set_jmp_target(TranslationBlock
*tb
,
261 int n
, uintptr_t addr
)
263 tb
->tb_next
[n
] = addr
;
268 static inline void tb_add_jump(TranslationBlock
*tb
, int n
,
269 TranslationBlock
*tb_next
)
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
, (uintptr_t)tb_next
->tc_ptr
);
276 /* add in TB jmp circular list */
277 tb
->jmp_next
[n
] = tb_next
->jmp_first
;
278 tb_next
->jmp_first
= (TranslationBlock
*)((uintptr_t)(tb
) | (n
));
282 TranslationBlock
*tb_find_pc(uintptr_t pc_ptr
);
284 #include "qemu-lock.h"
286 extern spinlock_t tb_lock
;
288 extern int tb_invalidated_flag
;
290 /* The return address may point to the start of the next instruction.
291 Subtracting one gets us the call instruction itself. */
292 #if defined(CONFIG_TCG_INTERPRETER)
293 /* Softmmu, Alpha, MIPS, SH4 and SPARC user mode emulations call GETPC().
294 For all others, GETPC remains undefined (which makes TCI a little faster. */
295 # if defined(CONFIG_SOFTMMU) || \
296 defined(TARGET_ALPHA) || defined(TARGET_MIPS) || \
297 defined(TARGET_SH4) || defined(TARGET_SPARC)
298 extern uintptr_t tci_tb_ptr
;
299 # define GETPC() tci_tb_ptr
301 #elif defined(__s390__) && !defined(__s390x__)
303 (((uintptr_t)__builtin_return_address(0) & 0x7fffffffUL) - 1)
304 #elif defined(__arm__)
305 /* Thumb return addresses have the low bit set, so we need to subtract two.
306 This is still safe in ARM mode because instructions are 4 bytes. */
307 # define GETPC() ((uintptr_t)__builtin_return_address(0) - 2)
309 # define GETPC() ((uintptr_t)__builtin_return_address(0) - 1)
312 #if defined(CONFIG_QEMU_LDST_OPTIMIZATION) && defined(CONFIG_SOFTMMU)
313 /* qemu_ld/st optimization split code generation to fast and slow path, thus,
314 it needs special handling for an MMU helper which is called from the slow
315 path, to get the fast path's pc without any additional argument.
316 It uses a tricky solution which embeds the fast path pc into the slow path.
318 Code flow in slow path:
322 (4) fast path information (implementation specific)
323 (5) post-process (e.g. stack adjust)
324 (6) jump to corresponding code of the next of fast path
326 # if defined(__i386__) || defined(__x86_64__)
327 /* To avoid broken disassembling, long jmp is used for embedding fast path pc,
328 so that the destination is the next code of fast path, though this jmp is
332 jmp POST_PROC (2byte) <- GETRA()
333 jmp NEXT_CODE (5byte)
334 POST_PROCESS ... <- GETRA() + 7
336 # define GETRA() ((uintptr_t)__builtin_return_address(0))
337 # define GETPC_LDST() ((uintptr_t)(GETRA() + 7 + \
338 *(int32_t *)((void *)GETRA() + 3) - 1))
339 # elif defined (_ARCH_PPC) && !defined (_ARCH_PPC64)
340 # define GETRA() ((uintptr_t)__builtin_return_address(0))
341 # define GETPC_LDST() ((uintptr_t) ((*(int32_t *)(GETRA() - 4)) - 1))
343 # error "CONFIG_QEMU_LDST_OPTIMIZATION needs GETPC_LDST() implementation!"
345 bool is_tcg_gen_code(uintptr_t pc_ptr
);
346 # define GETPC_EXT() (is_tcg_gen_code(GETRA()) ? GETPC_LDST() : GETPC())
348 # define GETPC_EXT() GETPC()
351 #if !defined(CONFIG_USER_ONLY)
353 struct MemoryRegion
*iotlb_to_region(hwaddr index
);
354 uint64_t io_mem_read(struct MemoryRegion
*mr
, hwaddr addr
,
356 void io_mem_write(struct MemoryRegion
*mr
, hwaddr addr
,
357 uint64_t value
, unsigned size
);
359 void tlb_fill(CPUArchState
*env1
, target_ulong addr
, int is_write
, int mmu_idx
,
362 #include "softmmu_defs.h"
364 #define ACCESS_TYPE (NB_MMU_MODES + 1)
365 #define MEMSUFFIX _code
368 #include "softmmu_header.h"
371 #include "softmmu_header.h"
374 #include "softmmu_header.h"
377 #include "softmmu_header.h"
384 #if defined(CONFIG_USER_ONLY)
385 static inline tb_page_addr_t
get_page_addr_code(CPUArchState
*env1
, target_ulong addr
)
391 tb_page_addr_t
get_page_addr_code(CPUArchState
*env1
, target_ulong addr
);
394 typedef void (CPUDebugExcpHandler
)(CPUArchState
*env
);
396 void cpu_set_debug_excp_handler(CPUDebugExcpHandler
*handler
);
399 extern int singlestep
;
402 extern volatile sig_atomic_t exit_request
;
404 /* Deterministic execution requires that IO only be performed on the last
405 instruction of a TB so that interrupts take effect immediately. */
406 static inline int can_do_io(CPUArchState
*env
)
411 /* If not executing code then assume we are ok. */
412 if (!env
->current_tb
) {
415 return env
->can_do_io
!= 0;