6 #include "exec/exec-all.h"
7 #include "exec/cpu_ldst.h"
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 is the size of the host kernel's sigset_t, needed where we make
24 * direct system calls that take a sigset_t pointer and a size.
26 #define SIGSET_T_SIZE (_NSIG / 8)
28 /* This struct is used to hold certain information about the image.
29 * Basically, it replicates in user space what would be certain
30 * task_struct fields in the kernel
42 abi_ulong start_stack
;
43 abi_ulong stack_limit
;
45 abi_ulong code_offset
;
46 abi_ulong data_offset
;
51 abi_ulong arg_strings
;
52 abi_ulong env_strings
;
53 abi_ulong file_string
;
56 #ifdef CONFIG_USE_FDPIC
57 abi_ulong loadmap_addr
;
60 abi_ulong pt_dynamic_addr
;
61 struct image_info
*other_info
;
66 /* Information about the current linux thread */
67 struct vm86_saved_state
{
68 uint32_t eax
; /* return code */
78 uint16_t cs
, ss
, ds
, es
, fs
, gs
;
82 #if defined(TARGET_ARM) && defined(TARGET_ABI32)
84 #include "nwfpe/fpa11.h"
87 #define MAX_SIGQUEUE_SIZE 1024
89 struct emulated_sigtable
{
90 int pending
; /* true if signal is pending */
91 target_siginfo_t info
;
94 /* NOTE: we force a big alignment so that the stack stored after is
96 typedef struct TaskState
{
97 pid_t ts_tid
; /* tid (or pid) of this task */
105 #ifdef TARGET_UNICORE32
108 #if defined(TARGET_I386) && !defined(TARGET_X86_64)
109 abi_ulong target_v86
;
110 struct vm86_saved_state vm86_saved_regs
;
111 struct target_vm86plus_struct vm86plus
;
115 abi_ulong child_tidptr
;
120 #if defined(TARGET_ARM) || defined(TARGET_M68K) || defined(TARGET_UNICORE32)
121 /* Extra fields for semihosted binaries. */
123 abi_ulong heap_limit
;
125 abi_ulong stack_base
;
126 int used
; /* non zero if used */
127 struct image_info
*info
;
128 struct linux_binprm
*bprm
;
130 struct emulated_sigtable sync_signal
;
131 struct emulated_sigtable sigtab
[TARGET_NSIG
];
132 /* This thread's signal mask, as requested by the guest program.
133 * The actual signal mask of this thread may differ:
134 * + we don't let SIGSEGV and SIGBUS be blocked while running guest code
135 * + sometimes we block all signals to avoid races
137 sigset_t signal_mask
;
138 /* The signal mask imposed by a guest sigsuspend syscall, if we are
139 * currently in the middle of such a syscall
141 sigset_t sigsuspend_mask
;
142 /* Nonzero if we're leaving a sigsuspend and sigsuspend_mask is valid. */
145 /* Nonzero if process_pending_signals() needs to do something (either
146 * handle a pending signal or unblock signals).
147 * This flag is written from a signal handler so should be accessed via
148 * the atomic_read() and atomic_write() functions. (It is not accessed
149 * from multiple threads.)
153 } __attribute__((aligned(16))) TaskState
;
155 extern char *exec_path
;
156 void init_task_state(TaskState
*ts
);
157 void task_settid(TaskState
*);
158 void stop_all_tasks(void);
159 extern const char *qemu_uname_release
;
160 extern unsigned long mmap_min_addr
;
162 /* ??? See if we can avoid exposing so much of the loader internals. */
164 /* Read a good amount of data initially, to hopefully get all the
165 program headers loaded. */
166 #define BPRM_BUF_SIZE 1024
169 * This structure is used to hold the arguments that are
170 * used when loading binaries.
172 struct linux_binprm
{
173 char buf
[BPRM_BUF_SIZE
] __attribute__((aligned
));
180 char * filename
; /* Name of binary */
181 int (*core_dump
)(int, const CPUArchState
*); /* coredump routine */
184 void do_init_thread(struct target_pt_regs
*regs
, struct image_info
*infop
);
185 abi_ulong
loader_build_argptr(int envc
, int argc
, abi_ulong sp
,
186 abi_ulong stringp
, int push_ptr
);
187 int loader_exec(int fdexec
, const char *filename
, char **argv
, char **envp
,
188 struct target_pt_regs
* regs
, struct image_info
*infop
,
189 struct linux_binprm
*);
191 int load_elf_binary(struct linux_binprm
*bprm
, struct image_info
*info
);
192 int load_flt_binary(struct linux_binprm
*bprm
, struct image_info
*info
);
194 abi_long
memcpy_to_target(abi_ulong dest
, const void *src
,
196 void target_set_brk(abi_ulong new_brk
);
197 abi_long
do_brk(abi_ulong new_brk
);
198 void syscall_init(void);
199 abi_long
do_syscall(void *cpu_env
, int num
, abi_long arg1
,
200 abi_long arg2
, abi_long arg3
, abi_long arg4
,
201 abi_long arg5
, abi_long arg6
, abi_long arg7
,
203 void gemu_log(const char *fmt
, ...) GCC_FMT_ATTR(1, 2);
204 extern THREAD CPUState
*thread_cpu
;
205 void cpu_loop(CPUArchState
*env
);
206 const char *target_strerror(int err
);
207 int get_osversion(void);
208 void init_qemu_uname_release(void);
209 void fork_start(void);
210 void fork_end(int child
);
212 /* Creates the initial guest address space in the host memory space using
213 * the given host start address hint and size. The guest_start parameter
214 * specifies the start address of the guest space. guest_base will be the
215 * difference between the host start address computed by this function and
216 * guest_start. If fixed is specified, then the mapped address space must
217 * start at host_start. The real start address of the mapped memory space is
218 * returned or -1 if there was an error.
220 unsigned long init_guest_space(unsigned long host_start
,
221 unsigned long host_size
,
222 unsigned long guest_start
,
225 #include "qemu/log.h"
231 * @int number: number of system call to make
232 * ...: arguments to the system call
234 * Call a system call if guest signal not pending.
235 * This has the same API as the libc syscall() function, except that it
236 * may return -1 with errno == TARGET_ERESTARTSYS if a signal was pending.
238 * Returns: the system call result, or -1 with an error code in errno
239 * (Errnos are host errnos; we rely on TARGET_ERESTARTSYS not clashing
240 * with any of the host errno values.)
243 /* A guide to using safe_syscall() to handle interactions between guest
244 * syscalls and guest signals:
246 * Guest syscalls come in two flavours:
248 * (1) Non-interruptible syscalls
250 * These are guest syscalls that never get interrupted by signals and
251 * so never return EINTR. They can be implemented straightforwardly in
252 * QEMU: just make sure that if the implementation code has to make any
253 * blocking calls that those calls are retried if they return EINTR.
254 * It's also OK to implement these with safe_syscall, though it will be
255 * a little less efficient if a signal is delivered at the 'wrong' moment.
257 * Some non-interruptible syscalls need to be handled using block_signals()
258 * to block signals for the duration of the syscall. This mainly applies
259 * to code which needs to modify the data structures used by the
260 * host_signal_handler() function and the functions it calls, including
261 * all syscalls which change the thread's signal mask.
263 * (2) Interruptible syscalls
265 * These are guest syscalls that can be interrupted by signals and
266 * for which we need to either return EINTR or arrange for the guest
267 * syscall to be restarted. This category includes both syscalls which
268 * always restart (and in the kernel return -ERESTARTNOINTR), ones
269 * which only restart if there is no handler (kernel returns -ERESTARTNOHAND
270 * or -ERESTART_RESTARTBLOCK), and the most common kind which restart
271 * if the handler was registered with SA_RESTART (kernel returns
272 * -ERESTARTSYS). System calls which are only interruptible in some
273 * situations (like 'open') also need to be handled this way.
275 * Here it is important that the host syscall is made
276 * via this safe_syscall() function, and *not* via the host libc.
277 * If the host libc is used then the implementation will appear to work
278 * most of the time, but there will be a race condition where a
279 * signal could arrive just before we make the host syscall inside libc,
280 * and then then guest syscall will not correctly be interrupted.
281 * Instead the implementation of the guest syscall can use the safe_syscall
282 * function but otherwise just return the result or errno in the usual
283 * way; the main loop code will take care of restarting the syscall
286 * (If the implementation needs to make multiple host syscalls this is
287 * OK; any which might really block must be via safe_syscall(); for those
288 * which are only technically blocking (ie which we know in practice won't
289 * stay in the host kernel indefinitely) it's OK to use libc if necessary.
290 * You must be able to cope with backing out correctly if some safe_syscall
291 * you make in the implementation returns either -TARGET_ERESTARTSYS or
294 * block_signals() cannot be used for interruptible syscalls.
297 * How and why the safe_syscall implementation works:
299 * The basic setup is that we make the host syscall via a known
300 * section of host native assembly. If a signal occurs, our signal
301 * handler checks the interrupted host PC against the addresse of that
302 * known section. If the PC is before or at the address of the syscall
303 * instruction then we change the PC to point at a "return
304 * -TARGET_ERESTARTSYS" code path instead, and then exit the signal handler
305 * (causing the safe_syscall() call to immediately return that value).
306 * Then in the main.c loop if we see this magic return value we adjust
307 * the guest PC to wind it back to before the system call, and invoke
308 * the guest signal handler as usual.
310 * This winding-back will happen in two cases:
311 * (1) signal came in just before we took the host syscall (a race);
312 * in this case we'll take the guest signal and have another go
313 * at the syscall afterwards, and this is indistinguishable for the
314 * guest from the timing having been different such that the guest
315 * signal really did win the race
316 * (2) signal came in while the host syscall was blocking, and the
317 * host kernel decided the syscall should be restarted;
318 * in this case we want to restart the guest syscall also, and so
319 * rewinding is the right thing. (Note that "restart" semantics mean
320 * "first call the signal handler, then reattempt the syscall".)
321 * The other situation to consider is when a signal came in while the
322 * host syscall was blocking, and the host kernel decided that the syscall
323 * should not be restarted; in this case QEMU's host signal handler will
324 * be invoked with the PC pointing just after the syscall instruction,
325 * with registers indicating an EINTR return; the special code in the
326 * handler will not kick in, and we will return EINTR to the guest as
329 * Notice that we can leave the host kernel to make the decision for
330 * us about whether to do a restart of the syscall or not; we do not
331 * need to check SA_RESTART flags in QEMU or distinguish the various
332 * kinds of restartability.
334 #ifdef HAVE_SAFE_SYSCALL
335 /* The core part of this function is implemented in assembly */
336 extern long safe_syscall_base(int *pending
, long number
, ...);
338 #define safe_syscall(...) \
341 int *psp_ = &((TaskState *)thread_cpu->opaque)->signal_pending; \
342 ret_ = safe_syscall_base(psp_, __VA_ARGS__); \
343 if (is_error(ret_)) { \
352 /* Fallback for architectures which don't yet provide a safe-syscall assembly
353 * fragment; note that this is racy!
354 * This should go away when all host architectures have been updated.
356 #define safe_syscall syscall
361 int host_to_target_waitstatus(int status
);
364 void print_syscall(int num
,
365 abi_long arg1
, abi_long arg2
, abi_long arg3
,
366 abi_long arg4
, abi_long arg5
, abi_long arg6
);
367 void print_syscall_ret(int num
, abi_long arg1
);
369 * print_taken_signal:
370 * @target_signum: target signal being taken
371 * @tinfo: target_siginfo_t which will be passed to the guest for the signal
373 * Print strace output indicating that this signal is being taken by the guest,
374 * in a format similar to:
375 * --- SIGSEGV {si_signo=SIGSEGV, si_code=SI_KERNEL, si_addr=0} ---
377 void print_taken_signal(int target_signum
, const target_siginfo_t
*tinfo
);
378 extern int do_strace
;
381 void process_pending_signals(CPUArchState
*cpu_env
);
382 void signal_init(void);
383 int queue_signal(CPUArchState
*env
, int sig
, int si_type
,
384 target_siginfo_t
*info
);
385 void host_to_target_siginfo(target_siginfo_t
*tinfo
, const siginfo_t
*info
);
386 void target_to_host_siginfo(siginfo_t
*info
, const target_siginfo_t
*tinfo
);
387 int target_to_host_signal(int sig
);
388 int host_to_target_signal(int sig
);
389 long do_sigreturn(CPUArchState
*env
);
390 long do_rt_sigreturn(CPUArchState
*env
);
391 abi_long
do_sigaltstack(abi_ulong uss_addr
, abi_ulong uoss_addr
, abi_ulong sp
);
392 int do_sigprocmask(int how
, const sigset_t
*set
, sigset_t
*oldset
);
394 * block_signals: block all signals while handling this guest syscall
396 * Block all signals, and arrange that the signal mask is returned to
397 * its correct value for the guest before we resume execution of guest code.
398 * If this function returns non-zero, then the caller should immediately
399 * return -TARGET_ERESTARTSYS to the main loop, which will take the pending
400 * signal and restart execution of the syscall.
401 * If block_signals() returns zero, then the caller can continue with
402 * emulation of the system call knowing that no signals can be taken
403 * (and therefore that no race conditions will result).
404 * This should only be called once, because if it is called a second time
405 * it will always return non-zero. (Think of it like a mutex that can't
406 * be recursively locked.)
407 * Signals will be unblocked again by process_pending_signals().
409 * Return value: non-zero if there was a pending signal, zero if not.
411 int block_signals(void); /* Returns non zero if signal pending */
415 void save_v86_state(CPUX86State
*env
);
416 void handle_vm86_trap(CPUX86State
*env
, int trapno
);
417 void handle_vm86_fault(CPUX86State
*env
);
418 int do_vm86(CPUX86State
*env
, long subfunction
, abi_ulong v86_addr
);
419 #elif defined(TARGET_SPARC64)
420 void sparc64_set_context(CPUSPARCState
*env
);
421 void sparc64_get_context(CPUSPARCState
*env
);
425 int target_mprotect(abi_ulong start
, abi_ulong len
, int prot
);
426 abi_long
target_mmap(abi_ulong start
, abi_ulong len
, int prot
,
427 int flags
, int fd
, abi_ulong offset
);
428 int target_munmap(abi_ulong start
, abi_ulong len
);
429 abi_long
target_mremap(abi_ulong old_addr
, abi_ulong old_size
,
430 abi_ulong new_size
, unsigned long flags
,
432 int target_msync(abi_ulong start
, abi_ulong len
, int flags
);
433 extern unsigned long last_brk
;
434 extern abi_ulong mmap_next_start
;
435 abi_ulong
mmap_find_vma(abi_ulong
, abi_ulong
);
436 void mmap_fork_start(void);
437 void mmap_fork_end(int child
);
440 extern unsigned long guest_stack_size
;
444 #define VERIFY_READ 0
445 #define VERIFY_WRITE 1 /* implies read access */
447 static inline int access_ok(int type
, abi_ulong addr
, abi_ulong size
)
449 return page_check_range((target_ulong
)addr
, size
,
450 (type
== VERIFY_READ
) ? PAGE_READ
: (PAGE_READ
| PAGE_WRITE
)) == 0;
453 /* NOTE __get_user and __put_user use host pointers and don't check access.
454 These are usually used to access struct data members once the struct has
455 been locked - usually with lock_user_struct. */
458 - Use __builtin_choose_expr to avoid type promotion from ?:,
459 - Invalid sizes result in a compile time error stemming from
460 the fact that abort has no parameters.
461 - It's easier to use the endian-specific unaligned load/store
462 functions than host-endian unaligned load/store plus tswapN. */
464 #define __put_user_e(x, hptr, e) \
465 (__builtin_choose_expr(sizeof(*(hptr)) == 1, stb_p, \
466 __builtin_choose_expr(sizeof(*(hptr)) == 2, stw_##e##_p, \
467 __builtin_choose_expr(sizeof(*(hptr)) == 4, stl_##e##_p, \
468 __builtin_choose_expr(sizeof(*(hptr)) == 8, stq_##e##_p, abort)))) \
469 ((hptr), (x)), (void)0)
471 #define __get_user_e(x, hptr, e) \
472 ((x) = (typeof(*hptr))( \
473 __builtin_choose_expr(sizeof(*(hptr)) == 1, ldub_p, \
474 __builtin_choose_expr(sizeof(*(hptr)) == 2, lduw_##e##_p, \
475 __builtin_choose_expr(sizeof(*(hptr)) == 4, ldl_##e##_p, \
476 __builtin_choose_expr(sizeof(*(hptr)) == 8, ldq_##e##_p, abort)))) \
479 #ifdef TARGET_WORDS_BIGENDIAN
480 # define __put_user(x, hptr) __put_user_e(x, hptr, be)
481 # define __get_user(x, hptr) __get_user_e(x, hptr, be)
483 # define __put_user(x, hptr) __put_user_e(x, hptr, le)
484 # define __get_user(x, hptr) __get_user_e(x, hptr, le)
487 /* put_user()/get_user() take a guest address and check access */
488 /* These are usually used to access an atomic data type, such as an int,
489 * that has been passed by address. These internally perform locking
490 * and unlocking on the data type.
492 #define put_user(x, gaddr, target_type) \
494 abi_ulong __gaddr = (gaddr); \
495 target_type *__hptr; \
496 abi_long __ret = 0; \
497 if ((__hptr = lock_user(VERIFY_WRITE, __gaddr, sizeof(target_type), 0))) { \
498 __put_user((x), __hptr); \
499 unlock_user(__hptr, __gaddr, sizeof(target_type)); \
501 __ret = -TARGET_EFAULT; \
505 #define get_user(x, gaddr, target_type) \
507 abi_ulong __gaddr = (gaddr); \
508 target_type *__hptr; \
509 abi_long __ret = 0; \
510 if ((__hptr = lock_user(VERIFY_READ, __gaddr, sizeof(target_type), 1))) { \
511 __get_user((x), __hptr); \
512 unlock_user(__hptr, __gaddr, 0); \
514 /* avoid warning */ \
516 __ret = -TARGET_EFAULT; \
521 #define put_user_ual(x, gaddr) put_user((x), (gaddr), abi_ulong)
522 #define put_user_sal(x, gaddr) put_user((x), (gaddr), abi_long)
523 #define put_user_u64(x, gaddr) put_user((x), (gaddr), uint64_t)
524 #define put_user_s64(x, gaddr) put_user((x), (gaddr), int64_t)
525 #define put_user_u32(x, gaddr) put_user((x), (gaddr), uint32_t)
526 #define put_user_s32(x, gaddr) put_user((x), (gaddr), int32_t)
527 #define put_user_u16(x, gaddr) put_user((x), (gaddr), uint16_t)
528 #define put_user_s16(x, gaddr) put_user((x), (gaddr), int16_t)
529 #define put_user_u8(x, gaddr) put_user((x), (gaddr), uint8_t)
530 #define put_user_s8(x, gaddr) put_user((x), (gaddr), int8_t)
532 #define get_user_ual(x, gaddr) get_user((x), (gaddr), abi_ulong)
533 #define get_user_sal(x, gaddr) get_user((x), (gaddr), abi_long)
534 #define get_user_u64(x, gaddr) get_user((x), (gaddr), uint64_t)
535 #define get_user_s64(x, gaddr) get_user((x), (gaddr), int64_t)
536 #define get_user_u32(x, gaddr) get_user((x), (gaddr), uint32_t)
537 #define get_user_s32(x, gaddr) get_user((x), (gaddr), int32_t)
538 #define get_user_u16(x, gaddr) get_user((x), (gaddr), uint16_t)
539 #define get_user_s16(x, gaddr) get_user((x), (gaddr), int16_t)
540 #define get_user_u8(x, gaddr) get_user((x), (gaddr), uint8_t)
541 #define get_user_s8(x, gaddr) get_user((x), (gaddr), int8_t)
543 /* copy_from_user() and copy_to_user() are usually used to copy data
544 * buffers between the target and host. These internally perform
545 * locking/unlocking of the memory.
547 abi_long
copy_from_user(void *hptr
, abi_ulong gaddr
, size_t len
);
548 abi_long
copy_to_user(abi_ulong gaddr
, void *hptr
, size_t len
);
550 /* Functions for accessing guest memory. The tget and tput functions
551 read/write single values, byteswapping as necessary. The lock_user function
552 gets a pointer to a contiguous area of guest memory, but does not perform
553 any byteswapping. lock_user may return either a pointer to the guest
554 memory, or a temporary buffer. */
556 /* Lock an area of guest memory into the host. If copy is true then the
557 host area will have the same contents as the guest. */
558 static inline void *lock_user(int type
, abi_ulong guest_addr
, long len
, int copy
)
560 if (!access_ok(type
, guest_addr
, len
))
565 addr
= g_malloc(len
);
567 memcpy(addr
, g2h(guest_addr
), len
);
569 memset(addr
, 0, len
);
573 return g2h(guest_addr
);
577 /* Unlock an area of guest memory. The first LEN bytes must be
578 flushed back to guest memory. host_ptr = NULL is explicitly
579 allowed and does nothing. */
580 static inline void unlock_user(void *host_ptr
, abi_ulong guest_addr
,
587 if (host_ptr
== g2h(guest_addr
))
590 memcpy(g2h(guest_addr
), host_ptr
, len
);
595 /* Return the length of a string in target memory or -TARGET_EFAULT if
597 abi_long
target_strlen(abi_ulong gaddr
);
599 /* Like lock_user but for null terminated strings. */
600 static inline void *lock_user_string(abi_ulong guest_addr
)
603 len
= target_strlen(guest_addr
);
606 return lock_user(VERIFY_READ
, guest_addr
, (long)(len
+ 1), 1);
609 /* Helper macros for locking/unlocking a target struct. */
610 #define lock_user_struct(type, host_ptr, guest_addr, copy) \
611 (host_ptr = lock_user(type, guest_addr, sizeof(*host_ptr), copy))
612 #define unlock_user_struct(host_ptr, guest_addr, copy) \
613 unlock_user(host_ptr, guest_addr, (copy) ? sizeof(*host_ptr) : 0)
617 /* Include target-specific struct and function definitions;
618 * they may need access to the target-independent structures
619 * above, so include them last.
621 #include "target_cpu.h"
622 #include "target_signal.h"
623 #include "target_structs.h"