qemu-img bench: Fix uninitialised writethrough mode
[qemu.git] / linux-user / qemu.h
blob56f29c35b5abc9597270c47881118023f3f294b3
1 #ifndef QEMU_H
2 #define QEMU_H
4 #include "hostdep.h"
5 #include "cpu.h"
6 #include "exec/exec-all.h"
7 #include "exec/cpu_ldst.h"
9 #undef DEBUG_REMAP
10 #ifdef DEBUG_REMAP
11 #endif /* DEBUG_REMAP */
13 #include "exec/user/abitypes.h"
15 #include "exec/user/thunk.h"
16 #include "syscall_defs.h"
17 #include "target_syscall.h"
18 #include "exec/gdbstub.h"
19 #include "qemu/queue.h"
21 #define THREAD __thread
23 /* This struct is used to hold certain information about the image.
24 * Basically, it replicates in user space what would be certain
25 * task_struct fields in the kernel
27 struct image_info {
28 abi_ulong load_bias;
29 abi_ulong load_addr;
30 abi_ulong start_code;
31 abi_ulong end_code;
32 abi_ulong start_data;
33 abi_ulong end_data;
34 abi_ulong start_brk;
35 abi_ulong brk;
36 abi_ulong start_mmap;
37 abi_ulong start_stack;
38 abi_ulong stack_limit;
39 abi_ulong entry;
40 abi_ulong code_offset;
41 abi_ulong data_offset;
42 abi_ulong saved_auxv;
43 abi_ulong auxv_len;
44 abi_ulong arg_start;
45 abi_ulong arg_end;
46 uint32_t elf_flags;
47 int personality;
48 #ifdef CONFIG_USE_FDPIC
49 abi_ulong loadmap_addr;
50 uint16_t nsegs;
51 void *loadsegs;
52 abi_ulong pt_dynamic_addr;
53 struct image_info *other_info;
54 #endif
57 #ifdef TARGET_I386
58 /* Information about the current linux thread */
59 struct vm86_saved_state {
60 uint32_t eax; /* return code */
61 uint32_t ebx;
62 uint32_t ecx;
63 uint32_t edx;
64 uint32_t esi;
65 uint32_t edi;
66 uint32_t ebp;
67 uint32_t esp;
68 uint32_t eflags;
69 uint32_t eip;
70 uint16_t cs, ss, ds, es, fs, gs;
72 #endif
74 #if defined(TARGET_ARM) && defined(TARGET_ABI32)
75 /* FPU emulator */
76 #include "nwfpe/fpa11.h"
77 #endif
79 #define MAX_SIGQUEUE_SIZE 1024
81 struct emulated_sigtable {
82 int pending; /* true if signal is pending */
83 target_siginfo_t info;
86 /* NOTE: we force a big alignment so that the stack stored after is
87 aligned too */
88 typedef struct TaskState {
89 pid_t ts_tid; /* tid (or pid) of this task */
90 #ifdef TARGET_ARM
91 # ifdef TARGET_ABI32
92 /* FPA state */
93 FPA11 fpa;
94 # endif
95 int swi_errno;
96 #endif
97 #ifdef TARGET_UNICORE32
98 int swi_errno;
99 #endif
100 #if defined(TARGET_I386) && !defined(TARGET_X86_64)
101 abi_ulong target_v86;
102 struct vm86_saved_state vm86_saved_regs;
103 struct target_vm86plus_struct vm86plus;
104 uint32_t v86flags;
105 uint32_t v86mask;
106 #endif
107 abi_ulong child_tidptr;
108 #ifdef TARGET_M68K
109 int sim_syscalls;
110 abi_ulong tp_value;
111 #endif
112 #if defined(TARGET_ARM) || defined(TARGET_M68K) || defined(TARGET_UNICORE32)
113 /* Extra fields for semihosted binaries. */
114 uint32_t heap_base;
115 uint32_t heap_limit;
116 #endif
117 uint32_t stack_base;
118 int used; /* non zero if used */
119 struct image_info *info;
120 struct linux_binprm *bprm;
122 struct emulated_sigtable sync_signal;
123 struct emulated_sigtable sigtab[TARGET_NSIG];
124 /* This thread's signal mask, as requested by the guest program.
125 * The actual signal mask of this thread may differ:
126 * + we don't let SIGSEGV and SIGBUS be blocked while running guest code
127 * + sometimes we block all signals to avoid races
129 sigset_t signal_mask;
130 /* The signal mask imposed by a guest sigsuspend syscall, if we are
131 * currently in the middle of such a syscall
133 sigset_t sigsuspend_mask;
134 /* Nonzero if we're leaving a sigsuspend and sigsuspend_mask is valid. */
135 int in_sigsuspend;
137 /* Nonzero if process_pending_signals() needs to do something (either
138 * handle a pending signal or unblock signals).
139 * This flag is written from a signal handler so should be accessed via
140 * the atomic_read() and atomic_write() functions. (It is not accessed
141 * from multiple threads.)
143 int signal_pending;
145 } __attribute__((aligned(16))) TaskState;
147 extern char *exec_path;
148 void init_task_state(TaskState *ts);
149 void task_settid(TaskState *);
150 void stop_all_tasks(void);
151 extern const char *qemu_uname_release;
152 extern unsigned long mmap_min_addr;
154 /* ??? See if we can avoid exposing so much of the loader internals. */
156 /* Read a good amount of data initially, to hopefully get all the
157 program headers loaded. */
158 #define BPRM_BUF_SIZE 1024
161 * This structure is used to hold the arguments that are
162 * used when loading binaries.
164 struct linux_binprm {
165 char buf[BPRM_BUF_SIZE] __attribute__((aligned));
166 abi_ulong p;
167 int fd;
168 int e_uid, e_gid;
169 int argc, envc;
170 char **argv;
171 char **envp;
172 char * filename; /* Name of binary */
173 int (*core_dump)(int, const CPUArchState *); /* coredump routine */
176 void do_init_thread(struct target_pt_regs *regs, struct image_info *infop);
177 abi_ulong loader_build_argptr(int envc, int argc, abi_ulong sp,
178 abi_ulong stringp, int push_ptr);
179 int loader_exec(int fdexec, const char *filename, char **argv, char **envp,
180 struct target_pt_regs * regs, struct image_info *infop,
181 struct linux_binprm *);
183 int load_elf_binary(struct linux_binprm *bprm, struct image_info *info);
184 int load_flt_binary(struct linux_binprm *bprm, struct image_info *info);
186 abi_long memcpy_to_target(abi_ulong dest, const void *src,
187 unsigned long len);
188 void target_set_brk(abi_ulong new_brk);
189 abi_long do_brk(abi_ulong new_brk);
190 void syscall_init(void);
191 abi_long do_syscall(void *cpu_env, int num, abi_long arg1,
192 abi_long arg2, abi_long arg3, abi_long arg4,
193 abi_long arg5, abi_long arg6, abi_long arg7,
194 abi_long arg8);
195 void gemu_log(const char *fmt, ...) GCC_FMT_ATTR(1, 2);
196 extern THREAD CPUState *thread_cpu;
197 void cpu_loop(CPUArchState *env);
198 const char *target_strerror(int err);
199 int get_osversion(void);
200 void init_qemu_uname_release(void);
201 void fork_start(void);
202 void fork_end(int child);
204 /* Creates the initial guest address space in the host memory space using
205 * the given host start address hint and size. The guest_start parameter
206 * specifies the start address of the guest space. guest_base will be the
207 * difference between the host start address computed by this function and
208 * guest_start. If fixed is specified, then the mapped address space must
209 * start at host_start. The real start address of the mapped memory space is
210 * returned or -1 if there was an error.
212 unsigned long init_guest_space(unsigned long host_start,
213 unsigned long host_size,
214 unsigned long guest_start,
215 bool fixed);
217 #include "qemu/log.h"
219 /* safe_syscall.S */
222 * safe_syscall:
223 * @int number: number of system call to make
224 * ...: arguments to the system call
226 * Call a system call if guest signal not pending.
227 * This has the same API as the libc syscall() function, except that it
228 * may return -1 with errno == TARGET_ERESTARTSYS if a signal was pending.
230 * Returns: the system call result, or -1 with an error code in errno
231 * (Errnos are host errnos; we rely on TARGET_ERESTARTSYS not clashing
232 * with any of the host errno values.)
235 /* A guide to using safe_syscall() to handle interactions between guest
236 * syscalls and guest signals:
238 * Guest syscalls come in two flavours:
240 * (1) Non-interruptible syscalls
242 * These are guest syscalls that never get interrupted by signals and
243 * so never return EINTR. They can be implemented straightforwardly in
244 * QEMU: just make sure that if the implementation code has to make any
245 * blocking calls that those calls are retried if they return EINTR.
246 * It's also OK to implement these with safe_syscall, though it will be
247 * a little less efficient if a signal is delivered at the 'wrong' moment.
249 * Some non-interruptible syscalls need to be handled using block_signals()
250 * to block signals for the duration of the syscall. This mainly applies
251 * to code which needs to modify the data structures used by the
252 * host_signal_handler() function and the functions it calls, including
253 * all syscalls which change the thread's signal mask.
255 * (2) Interruptible syscalls
257 * These are guest syscalls that can be interrupted by signals and
258 * for which we need to either return EINTR or arrange for the guest
259 * syscall to be restarted. This category includes both syscalls which
260 * always restart (and in the kernel return -ERESTARTNOINTR), ones
261 * which only restart if there is no handler (kernel returns -ERESTARTNOHAND
262 * or -ERESTART_RESTARTBLOCK), and the most common kind which restart
263 * if the handler was registered with SA_RESTART (kernel returns
264 * -ERESTARTSYS). System calls which are only interruptible in some
265 * situations (like 'open') also need to be handled this way.
267 * Here it is important that the host syscall is made
268 * via this safe_syscall() function, and *not* via the host libc.
269 * If the host libc is used then the implementation will appear to work
270 * most of the time, but there will be a race condition where a
271 * signal could arrive just before we make the host syscall inside libc,
272 * and then then guest syscall will not correctly be interrupted.
273 * Instead the implementation of the guest syscall can use the safe_syscall
274 * function but otherwise just return the result or errno in the usual
275 * way; the main loop code will take care of restarting the syscall
276 * if appropriate.
278 * (If the implementation needs to make multiple host syscalls this is
279 * OK; any which might really block must be via safe_syscall(); for those
280 * which are only technically blocking (ie which we know in practice won't
281 * stay in the host kernel indefinitely) it's OK to use libc if necessary.
282 * You must be able to cope with backing out correctly if some safe_syscall
283 * you make in the implementation returns either -TARGET_ERESTARTSYS or
284 * EINTR though.)
286 * block_signals() cannot be used for interruptible syscalls.
289 * How and why the safe_syscall implementation works:
291 * The basic setup is that we make the host syscall via a known
292 * section of host native assembly. If a signal occurs, our signal
293 * handler checks the interrupted host PC against the addresse of that
294 * known section. If the PC is before or at the address of the syscall
295 * instruction then we change the PC to point at a "return
296 * -TARGET_ERESTARTSYS" code path instead, and then exit the signal handler
297 * (causing the safe_syscall() call to immediately return that value).
298 * Then in the main.c loop if we see this magic return value we adjust
299 * the guest PC to wind it back to before the system call, and invoke
300 * the guest signal handler as usual.
302 * This winding-back will happen in two cases:
303 * (1) signal came in just before we took the host syscall (a race);
304 * in this case we'll take the guest signal and have another go
305 * at the syscall afterwards, and this is indistinguishable for the
306 * guest from the timing having been different such that the guest
307 * signal really did win the race
308 * (2) signal came in while the host syscall was blocking, and the
309 * host kernel decided the syscall should be restarted;
310 * in this case we want to restart the guest syscall also, and so
311 * rewinding is the right thing. (Note that "restart" semantics mean
312 * "first call the signal handler, then reattempt the syscall".)
313 * The other situation to consider is when a signal came in while the
314 * host syscall was blocking, and the host kernel decided that the syscall
315 * should not be restarted; in this case QEMU's host signal handler will
316 * be invoked with the PC pointing just after the syscall instruction,
317 * with registers indicating an EINTR return; the special code in the
318 * handler will not kick in, and we will return EINTR to the guest as
319 * we should.
321 * Notice that we can leave the host kernel to make the decision for
322 * us about whether to do a restart of the syscall or not; we do not
323 * need to check SA_RESTART flags in QEMU or distinguish the various
324 * kinds of restartability.
326 #ifdef HAVE_SAFE_SYSCALL
327 /* The core part of this function is implemented in assembly */
328 extern long safe_syscall_base(int *pending, long number, ...);
330 #define safe_syscall(...) \
331 ({ \
332 long ret_; \
333 int *psp_ = &((TaskState *)thread_cpu->opaque)->signal_pending; \
334 ret_ = safe_syscall_base(psp_, __VA_ARGS__); \
335 if (is_error(ret_)) { \
336 errno = -ret_; \
337 ret_ = -1; \
339 ret_; \
342 #else
344 /* Fallback for architectures which don't yet provide a safe-syscall assembly
345 * fragment; note that this is racy!
346 * This should go away when all host architectures have been updated.
348 #define safe_syscall syscall
350 #endif
352 /* syscall.c */
353 int host_to_target_waitstatus(int status);
355 /* strace.c */
356 void print_syscall(int num,
357 abi_long arg1, abi_long arg2, abi_long arg3,
358 abi_long arg4, abi_long arg5, abi_long arg6);
359 void print_syscall_ret(int num, abi_long arg1);
360 extern int do_strace;
362 /* signal.c */
363 void process_pending_signals(CPUArchState *cpu_env);
364 void signal_init(void);
365 int queue_signal(CPUArchState *env, int sig, target_siginfo_t *info);
366 void host_to_target_siginfo(target_siginfo_t *tinfo, const siginfo_t *info);
367 void target_to_host_siginfo(siginfo_t *info, const target_siginfo_t *tinfo);
368 int target_to_host_signal(int sig);
369 int host_to_target_signal(int sig);
370 long do_sigreturn(CPUArchState *env);
371 long do_rt_sigreturn(CPUArchState *env);
372 abi_long do_sigaltstack(abi_ulong uss_addr, abi_ulong uoss_addr, abi_ulong sp);
373 int do_sigprocmask(int how, const sigset_t *set, sigset_t *oldset);
375 * block_signals: block all signals while handling this guest syscall
377 * Block all signals, and arrange that the signal mask is returned to
378 * its correct value for the guest before we resume execution of guest code.
379 * If this function returns non-zero, then the caller should immediately
380 * return -TARGET_ERESTARTSYS to the main loop, which will take the pending
381 * signal and restart execution of the syscall.
382 * If block_signals() returns zero, then the caller can continue with
383 * emulation of the system call knowing that no signals can be taken
384 * (and therefore that no race conditions will result).
385 * This should only be called once, because if it is called a second time
386 * it will always return non-zero. (Think of it like a mutex that can't
387 * be recursively locked.)
388 * Signals will be unblocked again by process_pending_signals().
390 * Return value: non-zero if there was a pending signal, zero if not.
392 int block_signals(void); /* Returns non zero if signal pending */
394 #ifdef TARGET_I386
395 /* vm86.c */
396 void save_v86_state(CPUX86State *env);
397 void handle_vm86_trap(CPUX86State *env, int trapno);
398 void handle_vm86_fault(CPUX86State *env);
399 int do_vm86(CPUX86State *env, long subfunction, abi_ulong v86_addr);
400 #elif defined(TARGET_SPARC64)
401 void sparc64_set_context(CPUSPARCState *env);
402 void sparc64_get_context(CPUSPARCState *env);
403 #endif
405 /* mmap.c */
406 int target_mprotect(abi_ulong start, abi_ulong len, int prot);
407 abi_long target_mmap(abi_ulong start, abi_ulong len, int prot,
408 int flags, int fd, abi_ulong offset);
409 int target_munmap(abi_ulong start, abi_ulong len);
410 abi_long target_mremap(abi_ulong old_addr, abi_ulong old_size,
411 abi_ulong new_size, unsigned long flags,
412 abi_ulong new_addr);
413 int target_msync(abi_ulong start, abi_ulong len, int flags);
414 extern unsigned long last_brk;
415 extern abi_ulong mmap_next_start;
416 abi_ulong mmap_find_vma(abi_ulong, abi_ulong);
417 void cpu_list_lock(void);
418 void cpu_list_unlock(void);
419 void mmap_fork_start(void);
420 void mmap_fork_end(int child);
422 /* main.c */
423 extern unsigned long guest_stack_size;
425 /* user access */
427 #define VERIFY_READ 0
428 #define VERIFY_WRITE 1 /* implies read access */
430 static inline int access_ok(int type, abi_ulong addr, abi_ulong size)
432 return page_check_range((target_ulong)addr, size,
433 (type == VERIFY_READ) ? PAGE_READ : (PAGE_READ | PAGE_WRITE)) == 0;
436 /* NOTE __get_user and __put_user use host pointers and don't check access.
437 These are usually used to access struct data members once the struct has
438 been locked - usually with lock_user_struct. */
440 /* Tricky points:
441 - Use __builtin_choose_expr to avoid type promotion from ?:,
442 - Invalid sizes result in a compile time error stemming from
443 the fact that abort has no parameters.
444 - It's easier to use the endian-specific unaligned load/store
445 functions than host-endian unaligned load/store plus tswapN. */
447 #define __put_user_e(x, hptr, e) \
448 (__builtin_choose_expr(sizeof(*(hptr)) == 1, stb_p, \
449 __builtin_choose_expr(sizeof(*(hptr)) == 2, stw_##e##_p, \
450 __builtin_choose_expr(sizeof(*(hptr)) == 4, stl_##e##_p, \
451 __builtin_choose_expr(sizeof(*(hptr)) == 8, stq_##e##_p, abort)))) \
452 ((hptr), (x)), (void)0)
454 #define __get_user_e(x, hptr, e) \
455 ((x) = (typeof(*hptr))( \
456 __builtin_choose_expr(sizeof(*(hptr)) == 1, ldub_p, \
457 __builtin_choose_expr(sizeof(*(hptr)) == 2, lduw_##e##_p, \
458 __builtin_choose_expr(sizeof(*(hptr)) == 4, ldl_##e##_p, \
459 __builtin_choose_expr(sizeof(*(hptr)) == 8, ldq_##e##_p, abort)))) \
460 (hptr)), (void)0)
462 #ifdef TARGET_WORDS_BIGENDIAN
463 # define __put_user(x, hptr) __put_user_e(x, hptr, be)
464 # define __get_user(x, hptr) __get_user_e(x, hptr, be)
465 #else
466 # define __put_user(x, hptr) __put_user_e(x, hptr, le)
467 # define __get_user(x, hptr) __get_user_e(x, hptr, le)
468 #endif
470 /* put_user()/get_user() take a guest address and check access */
471 /* These are usually used to access an atomic data type, such as an int,
472 * that has been passed by address. These internally perform locking
473 * and unlocking on the data type.
475 #define put_user(x, gaddr, target_type) \
476 ({ \
477 abi_ulong __gaddr = (gaddr); \
478 target_type *__hptr; \
479 abi_long __ret = 0; \
480 if ((__hptr = lock_user(VERIFY_WRITE, __gaddr, sizeof(target_type), 0))) { \
481 __put_user((x), __hptr); \
482 unlock_user(__hptr, __gaddr, sizeof(target_type)); \
483 } else \
484 __ret = -TARGET_EFAULT; \
485 __ret; \
488 #define get_user(x, gaddr, target_type) \
489 ({ \
490 abi_ulong __gaddr = (gaddr); \
491 target_type *__hptr; \
492 abi_long __ret = 0; \
493 if ((__hptr = lock_user(VERIFY_READ, __gaddr, sizeof(target_type), 1))) { \
494 __get_user((x), __hptr); \
495 unlock_user(__hptr, __gaddr, 0); \
496 } else { \
497 /* avoid warning */ \
498 (x) = 0; \
499 __ret = -TARGET_EFAULT; \
501 __ret; \
504 #define put_user_ual(x, gaddr) put_user((x), (gaddr), abi_ulong)
505 #define put_user_sal(x, gaddr) put_user((x), (gaddr), abi_long)
506 #define put_user_u64(x, gaddr) put_user((x), (gaddr), uint64_t)
507 #define put_user_s64(x, gaddr) put_user((x), (gaddr), int64_t)
508 #define put_user_u32(x, gaddr) put_user((x), (gaddr), uint32_t)
509 #define put_user_s32(x, gaddr) put_user((x), (gaddr), int32_t)
510 #define put_user_u16(x, gaddr) put_user((x), (gaddr), uint16_t)
511 #define put_user_s16(x, gaddr) put_user((x), (gaddr), int16_t)
512 #define put_user_u8(x, gaddr) put_user((x), (gaddr), uint8_t)
513 #define put_user_s8(x, gaddr) put_user((x), (gaddr), int8_t)
515 #define get_user_ual(x, gaddr) get_user((x), (gaddr), abi_ulong)
516 #define get_user_sal(x, gaddr) get_user((x), (gaddr), abi_long)
517 #define get_user_u64(x, gaddr) get_user((x), (gaddr), uint64_t)
518 #define get_user_s64(x, gaddr) get_user((x), (gaddr), int64_t)
519 #define get_user_u32(x, gaddr) get_user((x), (gaddr), uint32_t)
520 #define get_user_s32(x, gaddr) get_user((x), (gaddr), int32_t)
521 #define get_user_u16(x, gaddr) get_user((x), (gaddr), uint16_t)
522 #define get_user_s16(x, gaddr) get_user((x), (gaddr), int16_t)
523 #define get_user_u8(x, gaddr) get_user((x), (gaddr), uint8_t)
524 #define get_user_s8(x, gaddr) get_user((x), (gaddr), int8_t)
526 /* copy_from_user() and copy_to_user() are usually used to copy data
527 * buffers between the target and host. These internally perform
528 * locking/unlocking of the memory.
530 abi_long copy_from_user(void *hptr, abi_ulong gaddr, size_t len);
531 abi_long copy_to_user(abi_ulong gaddr, void *hptr, size_t len);
533 /* Functions for accessing guest memory. The tget and tput functions
534 read/write single values, byteswapping as necessary. The lock_user function
535 gets a pointer to a contiguous area of guest memory, but does not perform
536 any byteswapping. lock_user may return either a pointer to the guest
537 memory, or a temporary buffer. */
539 /* Lock an area of guest memory into the host. If copy is true then the
540 host area will have the same contents as the guest. */
541 static inline void *lock_user(int type, abi_ulong guest_addr, long len, int copy)
543 if (!access_ok(type, guest_addr, len))
544 return NULL;
545 #ifdef DEBUG_REMAP
547 void *addr;
548 addr = malloc(len);
549 if (copy)
550 memcpy(addr, g2h(guest_addr), len);
551 else
552 memset(addr, 0, len);
553 return addr;
555 #else
556 return g2h(guest_addr);
557 #endif
560 /* Unlock an area of guest memory. The first LEN bytes must be
561 flushed back to guest memory. host_ptr = NULL is explicitly
562 allowed and does nothing. */
563 static inline void unlock_user(void *host_ptr, abi_ulong guest_addr,
564 long len)
567 #ifdef DEBUG_REMAP
568 if (!host_ptr)
569 return;
570 if (host_ptr == g2h(guest_addr))
571 return;
572 if (len > 0)
573 memcpy(g2h(guest_addr), host_ptr, len);
574 free(host_ptr);
575 #endif
578 /* Return the length of a string in target memory or -TARGET_EFAULT if
579 access error. */
580 abi_long target_strlen(abi_ulong gaddr);
582 /* Like lock_user but for null terminated strings. */
583 static inline void *lock_user_string(abi_ulong guest_addr)
585 abi_long len;
586 len = target_strlen(guest_addr);
587 if (len < 0)
588 return NULL;
589 return lock_user(VERIFY_READ, guest_addr, (long)(len + 1), 1);
592 /* Helper macros for locking/unlocking a target struct. */
593 #define lock_user_struct(type, host_ptr, guest_addr, copy) \
594 (host_ptr = lock_user(type, guest_addr, sizeof(*host_ptr), copy))
595 #define unlock_user_struct(host_ptr, guest_addr, copy) \
596 unlock_user(host_ptr, guest_addr, (copy) ? sizeof(*host_ptr) : 0)
598 #include <pthread.h>
600 /* Include target-specific struct and function definitions;
601 * they may need access to the target-independent structures
602 * above, so include them last.
604 #include "target_cpu.h"
605 #include "target_signal.h"
606 #include "target_structs.h"
608 #endif /* QEMU_H */