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 /* This is the size of the host kernel's sigset_t, needed where we make
22 * direct system calls that take a sigset_t pointer and a size.
24 #define SIGSET_T_SIZE (_NSIG / 8)
26 /* This struct is used to hold certain information about the image.
27 * Basically, it replicates in user space what would be certain
28 * task_struct fields in the kernel
40 abi_ulong start_stack
;
41 abi_ulong stack_limit
;
43 abi_ulong code_offset
;
44 abi_ulong data_offset
;
49 abi_ulong arg_strings
;
50 abi_ulong env_strings
;
51 abi_ulong file_string
;
55 /* The fields below are used in FDPIC mode. */
56 abi_ulong loadmap_addr
;
59 abi_ulong pt_dynamic_addr
;
60 struct image_info
*other_info
;
64 /* Information about the current linux thread */
65 struct vm86_saved_state
{
66 uint32_t eax
; /* return code */
76 uint16_t cs
, ss
, ds
, es
, fs
, gs
;
80 #if defined(TARGET_ARM) && defined(TARGET_ABI32)
82 #include "nwfpe/fpa11.h"
85 #define MAX_SIGQUEUE_SIZE 1024
87 struct emulated_sigtable
{
88 int pending
; /* true if signal is pending */
89 target_siginfo_t info
;
92 /* NOTE: we force a big alignment so that the stack stored after is
94 typedef struct TaskState
{
95 pid_t ts_tid
; /* tid (or pid) of this task */
103 #if defined(TARGET_I386) && !defined(TARGET_X86_64)
104 abi_ulong target_v86
;
105 struct vm86_saved_state vm86_saved_regs
;
106 struct target_vm86plus_struct vm86plus
;
110 abi_ulong child_tidptr
;
115 #if defined(TARGET_ARM) || defined(TARGET_M68K)
116 /* Extra fields for semihosted binaries. */
118 abi_ulong heap_limit
;
120 abi_ulong stack_base
;
121 int used
; /* non zero if used */
122 struct image_info
*info
;
123 struct linux_binprm
*bprm
;
125 struct emulated_sigtable sync_signal
;
126 struct emulated_sigtable sigtab
[TARGET_NSIG
];
127 /* This thread's signal mask, as requested by the guest program.
128 * The actual signal mask of this thread may differ:
129 * + we don't let SIGSEGV and SIGBUS be blocked while running guest code
130 * + sometimes we block all signals to avoid races
132 sigset_t signal_mask
;
133 /* The signal mask imposed by a guest sigsuspend syscall, if we are
134 * currently in the middle of such a syscall
136 sigset_t sigsuspend_mask
;
137 /* Nonzero if we're leaving a sigsuspend and sigsuspend_mask is valid. */
140 /* Nonzero if process_pending_signals() needs to do something (either
141 * handle a pending signal or unblock signals).
142 * This flag is written from a signal handler so should be accessed via
143 * the atomic_read() and atomic_write() functions. (It is not accessed
144 * from multiple threads.)
148 } __attribute__((aligned(16))) TaskState
;
150 extern char *exec_path
;
151 void init_task_state(TaskState
*ts
);
152 void task_settid(TaskState
*);
153 void stop_all_tasks(void);
154 extern const char *qemu_uname_release
;
155 extern unsigned long mmap_min_addr
;
157 /* ??? See if we can avoid exposing so much of the loader internals. */
159 /* Read a good amount of data initially, to hopefully get all the
160 program headers loaded. */
161 #define BPRM_BUF_SIZE 1024
164 * This structure is used to hold the arguments that are
165 * used when loading binaries.
167 struct linux_binprm
{
168 char buf
[BPRM_BUF_SIZE
] __attribute__((aligned
));
175 char * filename
; /* Name of binary */
176 int (*core_dump
)(int, const CPUArchState
*); /* coredump routine */
179 void do_init_thread(struct target_pt_regs
*regs
, struct image_info
*infop
);
180 abi_ulong
loader_build_argptr(int envc
, int argc
, abi_ulong sp
,
181 abi_ulong stringp
, int push_ptr
);
182 int loader_exec(int fdexec
, const char *filename
, char **argv
, char **envp
,
183 struct target_pt_regs
* regs
, struct image_info
*infop
,
184 struct linux_binprm
*);
186 uint32_t get_elf_eflags(int fd
);
187 int load_elf_binary(struct linux_binprm
*bprm
, struct image_info
*info
);
188 int load_flt_binary(struct linux_binprm
*bprm
, struct image_info
*info
);
190 abi_long
memcpy_to_target(abi_ulong dest
, const void *src
,
192 void target_set_brk(abi_ulong new_brk
);
193 abi_long
do_brk(abi_ulong new_brk
);
194 void syscall_init(void);
195 abi_long
do_syscall(void *cpu_env
, int num
, abi_long arg1
,
196 abi_long arg2
, abi_long arg3
, abi_long arg4
,
197 abi_long arg5
, abi_long arg6
, abi_long arg7
,
199 void gemu_log(const char *fmt
, ...) GCC_FMT_ATTR(1, 2);
200 extern __thread CPUState
*thread_cpu
;
201 void cpu_loop(CPUArchState
*env
);
202 const char *target_strerror(int err
);
203 int get_osversion(void);
204 void init_qemu_uname_release(void);
205 void fork_start(void);
206 void fork_end(int child
);
208 /* Creates the initial guest address space in the host memory space using
209 * the given host start address hint and size. The guest_start parameter
210 * specifies the start address of the guest space. guest_base will be the
211 * difference between the host start address computed by this function and
212 * guest_start. If fixed is specified, then the mapped address space must
213 * start at host_start. The real start address of the mapped memory space is
214 * returned or -1 if there was an error.
216 unsigned long init_guest_space(unsigned long host_start
,
217 unsigned long host_size
,
218 unsigned long guest_start
,
221 #include "qemu/log.h"
227 * @int number: number of system call to make
228 * ...: arguments to the system call
230 * Call a system call if guest signal not pending.
231 * This has the same API as the libc syscall() function, except that it
232 * may return -1 with errno == TARGET_ERESTARTSYS if a signal was pending.
234 * Returns: the system call result, or -1 with an error code in errno
235 * (Errnos are host errnos; we rely on TARGET_ERESTARTSYS not clashing
236 * with any of the host errno values.)
239 /* A guide to using safe_syscall() to handle interactions between guest
240 * syscalls and guest signals:
242 * Guest syscalls come in two flavours:
244 * (1) Non-interruptible syscalls
246 * These are guest syscalls that never get interrupted by signals and
247 * so never return EINTR. They can be implemented straightforwardly in
248 * QEMU: just make sure that if the implementation code has to make any
249 * blocking calls that those calls are retried if they return EINTR.
250 * It's also OK to implement these with safe_syscall, though it will be
251 * a little less efficient if a signal is delivered at the 'wrong' moment.
253 * Some non-interruptible syscalls need to be handled using block_signals()
254 * to block signals for the duration of the syscall. This mainly applies
255 * to code which needs to modify the data structures used by the
256 * host_signal_handler() function and the functions it calls, including
257 * all syscalls which change the thread's signal mask.
259 * (2) Interruptible syscalls
261 * These are guest syscalls that can be interrupted by signals and
262 * for which we need to either return EINTR or arrange for the guest
263 * syscall to be restarted. This category includes both syscalls which
264 * always restart (and in the kernel return -ERESTARTNOINTR), ones
265 * which only restart if there is no handler (kernel returns -ERESTARTNOHAND
266 * or -ERESTART_RESTARTBLOCK), and the most common kind which restart
267 * if the handler was registered with SA_RESTART (kernel returns
268 * -ERESTARTSYS). System calls which are only interruptible in some
269 * situations (like 'open') also need to be handled this way.
271 * Here it is important that the host syscall is made
272 * via this safe_syscall() function, and *not* via the host libc.
273 * If the host libc is used then the implementation will appear to work
274 * most of the time, but there will be a race condition where a
275 * signal could arrive just before we make the host syscall inside libc,
276 * and then then guest syscall will not correctly be interrupted.
277 * Instead the implementation of the guest syscall can use the safe_syscall
278 * function but otherwise just return the result or errno in the usual
279 * way; the main loop code will take care of restarting the syscall
282 * (If the implementation needs to make multiple host syscalls this is
283 * OK; any which might really block must be via safe_syscall(); for those
284 * which are only technically blocking (ie which we know in practice won't
285 * stay in the host kernel indefinitely) it's OK to use libc if necessary.
286 * You must be able to cope with backing out correctly if some safe_syscall
287 * you make in the implementation returns either -TARGET_ERESTARTSYS or
290 * block_signals() cannot be used for interruptible syscalls.
293 * How and why the safe_syscall implementation works:
295 * The basic setup is that we make the host syscall via a known
296 * section of host native assembly. If a signal occurs, our signal
297 * handler checks the interrupted host PC against the addresse of that
298 * known section. If the PC is before or at the address of the syscall
299 * instruction then we change the PC to point at a "return
300 * -TARGET_ERESTARTSYS" code path instead, and then exit the signal handler
301 * (causing the safe_syscall() call to immediately return that value).
302 * Then in the main.c loop if we see this magic return value we adjust
303 * the guest PC to wind it back to before the system call, and invoke
304 * the guest signal handler as usual.
306 * This winding-back will happen in two cases:
307 * (1) signal came in just before we took the host syscall (a race);
308 * in this case we'll take the guest signal and have another go
309 * at the syscall afterwards, and this is indistinguishable for the
310 * guest from the timing having been different such that the guest
311 * signal really did win the race
312 * (2) signal came in while the host syscall was blocking, and the
313 * host kernel decided the syscall should be restarted;
314 * in this case we want to restart the guest syscall also, and so
315 * rewinding is the right thing. (Note that "restart" semantics mean
316 * "first call the signal handler, then reattempt the syscall".)
317 * The other situation to consider is when a signal came in while the
318 * host syscall was blocking, and the host kernel decided that the syscall
319 * should not be restarted; in this case QEMU's host signal handler will
320 * be invoked with the PC pointing just after the syscall instruction,
321 * with registers indicating an EINTR return; the special code in the
322 * handler will not kick in, and we will return EINTR to the guest as
325 * Notice that we can leave the host kernel to make the decision for
326 * us about whether to do a restart of the syscall or not; we do not
327 * need to check SA_RESTART flags in QEMU or distinguish the various
328 * kinds of restartability.
330 #ifdef HAVE_SAFE_SYSCALL
331 /* The core part of this function is implemented in assembly */
332 extern long safe_syscall_base(int *pending
, long number
, ...);
334 #define safe_syscall(...) \
337 int *psp_ = &((TaskState *)thread_cpu->opaque)->signal_pending; \
338 ret_ = safe_syscall_base(psp_, __VA_ARGS__); \
339 if (is_error(ret_)) { \
348 /* Fallback for architectures which don't yet provide a safe-syscall assembly
349 * fragment; note that this is racy!
350 * This should go away when all host architectures have been updated.
352 #define safe_syscall syscall
357 int host_to_target_waitstatus(int status
);
360 void print_syscall(int num
,
361 abi_long arg1
, abi_long arg2
, abi_long arg3
,
362 abi_long arg4
, abi_long arg5
, abi_long arg6
);
363 void print_syscall_ret(int num
, abi_long arg1
);
365 * print_taken_signal:
366 * @target_signum: target signal being taken
367 * @tinfo: target_siginfo_t which will be passed to the guest for the signal
369 * Print strace output indicating that this signal is being taken by the guest,
370 * in a format similar to:
371 * --- SIGSEGV {si_signo=SIGSEGV, si_code=SI_KERNEL, si_addr=0} ---
373 void print_taken_signal(int target_signum
, const target_siginfo_t
*tinfo
);
374 extern int do_strace
;
377 void process_pending_signals(CPUArchState
*cpu_env
);
378 void signal_init(void);
379 int queue_signal(CPUArchState
*env
, int sig
, int si_type
,
380 target_siginfo_t
*info
);
381 void host_to_target_siginfo(target_siginfo_t
*tinfo
, const siginfo_t
*info
);
382 void target_to_host_siginfo(siginfo_t
*info
, const target_siginfo_t
*tinfo
);
383 int target_to_host_signal(int sig
);
384 int host_to_target_signal(int sig
);
385 long do_sigreturn(CPUArchState
*env
);
386 long do_rt_sigreturn(CPUArchState
*env
);
387 abi_long
do_sigaltstack(abi_ulong uss_addr
, abi_ulong uoss_addr
, abi_ulong sp
);
388 int do_sigprocmask(int how
, const sigset_t
*set
, sigset_t
*oldset
);
390 * block_signals: block all signals while handling this guest syscall
392 * Block all signals, and arrange that the signal mask is returned to
393 * its correct value for the guest before we resume execution of guest code.
394 * If this function returns non-zero, then the caller should immediately
395 * return -TARGET_ERESTARTSYS to the main loop, which will take the pending
396 * signal and restart execution of the syscall.
397 * If block_signals() returns zero, then the caller can continue with
398 * emulation of the system call knowing that no signals can be taken
399 * (and therefore that no race conditions will result).
400 * This should only be called once, because if it is called a second time
401 * it will always return non-zero. (Think of it like a mutex that can't
402 * be recursively locked.)
403 * Signals will be unblocked again by process_pending_signals().
405 * Return value: non-zero if there was a pending signal, zero if not.
407 int block_signals(void); /* Returns non zero if signal pending */
411 void save_v86_state(CPUX86State
*env
);
412 void handle_vm86_trap(CPUX86State
*env
, int trapno
);
413 void handle_vm86_fault(CPUX86State
*env
);
414 int do_vm86(CPUX86State
*env
, long subfunction
, abi_ulong v86_addr
);
415 #elif defined(TARGET_SPARC64)
416 void sparc64_set_context(CPUSPARCState
*env
);
417 void sparc64_get_context(CPUSPARCState
*env
);
421 int target_mprotect(abi_ulong start
, abi_ulong len
, int prot
);
422 abi_long
target_mmap(abi_ulong start
, abi_ulong len
, int prot
,
423 int flags
, int fd
, abi_ulong offset
);
424 int target_munmap(abi_ulong start
, abi_ulong len
);
425 abi_long
target_mremap(abi_ulong old_addr
, abi_ulong old_size
,
426 abi_ulong new_size
, unsigned long flags
,
428 extern unsigned long last_brk
;
429 extern abi_ulong mmap_next_start
;
430 abi_ulong
mmap_find_vma(abi_ulong
, abi_ulong
);
431 void mmap_fork_start(void);
432 void mmap_fork_end(int child
);
435 extern unsigned long guest_stack_size
;
439 #define VERIFY_READ 0
440 #define VERIFY_WRITE 1 /* implies read access */
442 static inline int access_ok(int type
, abi_ulong addr
, abi_ulong size
)
444 return page_check_range((target_ulong
)addr
, size
,
445 (type
== VERIFY_READ
) ? PAGE_READ
: (PAGE_READ
| PAGE_WRITE
)) == 0;
448 /* NOTE __get_user and __put_user use host pointers and don't check access.
449 These are usually used to access struct data members once the struct has
450 been locked - usually with lock_user_struct. */
453 - Use __builtin_choose_expr to avoid type promotion from ?:,
454 - Invalid sizes result in a compile time error stemming from
455 the fact that abort has no parameters.
456 - It's easier to use the endian-specific unaligned load/store
457 functions than host-endian unaligned load/store plus tswapN. */
459 #define __put_user_e(x, hptr, e) \
460 (__builtin_choose_expr(sizeof(*(hptr)) == 1, stb_p, \
461 __builtin_choose_expr(sizeof(*(hptr)) == 2, stw_##e##_p, \
462 __builtin_choose_expr(sizeof(*(hptr)) == 4, stl_##e##_p, \
463 __builtin_choose_expr(sizeof(*(hptr)) == 8, stq_##e##_p, abort)))) \
464 ((hptr), (x)), (void)0)
466 #define __get_user_e(x, hptr, e) \
467 ((x) = (typeof(*hptr))( \
468 __builtin_choose_expr(sizeof(*(hptr)) == 1, ldub_p, \
469 __builtin_choose_expr(sizeof(*(hptr)) == 2, lduw_##e##_p, \
470 __builtin_choose_expr(sizeof(*(hptr)) == 4, ldl_##e##_p, \
471 __builtin_choose_expr(sizeof(*(hptr)) == 8, ldq_##e##_p, abort)))) \
474 #ifdef TARGET_WORDS_BIGENDIAN
475 # define __put_user(x, hptr) __put_user_e(x, hptr, be)
476 # define __get_user(x, hptr) __get_user_e(x, hptr, be)
478 # define __put_user(x, hptr) __put_user_e(x, hptr, le)
479 # define __get_user(x, hptr) __get_user_e(x, hptr, le)
482 /* put_user()/get_user() take a guest address and check access */
483 /* These are usually used to access an atomic data type, such as an int,
484 * that has been passed by address. These internally perform locking
485 * and unlocking on the data type.
487 #define put_user(x, gaddr, target_type) \
489 abi_ulong __gaddr = (gaddr); \
490 target_type *__hptr; \
491 abi_long __ret = 0; \
492 if ((__hptr = lock_user(VERIFY_WRITE, __gaddr, sizeof(target_type), 0))) { \
493 __put_user((x), __hptr); \
494 unlock_user(__hptr, __gaddr, sizeof(target_type)); \
496 __ret = -TARGET_EFAULT; \
500 #define get_user(x, gaddr, target_type) \
502 abi_ulong __gaddr = (gaddr); \
503 target_type *__hptr; \
504 abi_long __ret = 0; \
505 if ((__hptr = lock_user(VERIFY_READ, __gaddr, sizeof(target_type), 1))) { \
506 __get_user((x), __hptr); \
507 unlock_user(__hptr, __gaddr, 0); \
509 /* avoid warning */ \
511 __ret = -TARGET_EFAULT; \
516 #define put_user_ual(x, gaddr) put_user((x), (gaddr), abi_ulong)
517 #define put_user_sal(x, gaddr) put_user((x), (gaddr), abi_long)
518 #define put_user_u64(x, gaddr) put_user((x), (gaddr), uint64_t)
519 #define put_user_s64(x, gaddr) put_user((x), (gaddr), int64_t)
520 #define put_user_u32(x, gaddr) put_user((x), (gaddr), uint32_t)
521 #define put_user_s32(x, gaddr) put_user((x), (gaddr), int32_t)
522 #define put_user_u16(x, gaddr) put_user((x), (gaddr), uint16_t)
523 #define put_user_s16(x, gaddr) put_user((x), (gaddr), int16_t)
524 #define put_user_u8(x, gaddr) put_user((x), (gaddr), uint8_t)
525 #define put_user_s8(x, gaddr) put_user((x), (gaddr), int8_t)
527 #define get_user_ual(x, gaddr) get_user((x), (gaddr), abi_ulong)
528 #define get_user_sal(x, gaddr) get_user((x), (gaddr), abi_long)
529 #define get_user_u64(x, gaddr) get_user((x), (gaddr), uint64_t)
530 #define get_user_s64(x, gaddr) get_user((x), (gaddr), int64_t)
531 #define get_user_u32(x, gaddr) get_user((x), (gaddr), uint32_t)
532 #define get_user_s32(x, gaddr) get_user((x), (gaddr), int32_t)
533 #define get_user_u16(x, gaddr) get_user((x), (gaddr), uint16_t)
534 #define get_user_s16(x, gaddr) get_user((x), (gaddr), int16_t)
535 #define get_user_u8(x, gaddr) get_user((x), (gaddr), uint8_t)
536 #define get_user_s8(x, gaddr) get_user((x), (gaddr), int8_t)
538 /* copy_from_user() and copy_to_user() are usually used to copy data
539 * buffers between the target and host. These internally perform
540 * locking/unlocking of the memory.
542 abi_long
copy_from_user(void *hptr
, abi_ulong gaddr
, size_t len
);
543 abi_long
copy_to_user(abi_ulong gaddr
, void *hptr
, size_t len
);
545 /* Functions for accessing guest memory. The tget and tput functions
546 read/write single values, byteswapping as necessary. The lock_user function
547 gets a pointer to a contiguous area of guest memory, but does not perform
548 any byteswapping. lock_user may return either a pointer to the guest
549 memory, or a temporary buffer. */
551 /* Lock an area of guest memory into the host. If copy is true then the
552 host area will have the same contents as the guest. */
553 static inline void *lock_user(int type
, abi_ulong guest_addr
, long len
, int copy
)
555 if (!access_ok(type
, guest_addr
, len
))
560 addr
= g_malloc(len
);
562 memcpy(addr
, g2h(guest_addr
), len
);
564 memset(addr
, 0, len
);
568 return g2h(guest_addr
);
572 /* Unlock an area of guest memory. The first LEN bytes must be
573 flushed back to guest memory. host_ptr = NULL is explicitly
574 allowed and does nothing. */
575 static inline void unlock_user(void *host_ptr
, abi_ulong guest_addr
,
582 if (host_ptr
== g2h(guest_addr
))
585 memcpy(g2h(guest_addr
), host_ptr
, len
);
590 /* Return the length of a string in target memory or -TARGET_EFAULT if
592 abi_long
target_strlen(abi_ulong gaddr
);
594 /* Like lock_user but for null terminated strings. */
595 static inline void *lock_user_string(abi_ulong guest_addr
)
598 len
= target_strlen(guest_addr
);
601 return lock_user(VERIFY_READ
, guest_addr
, (long)(len
+ 1), 1);
604 /* Helper macros for locking/unlocking a target struct. */
605 #define lock_user_struct(type, host_ptr, guest_addr, copy) \
606 (host_ptr = lock_user(type, guest_addr, sizeof(*host_ptr), copy))
607 #define unlock_user_struct(host_ptr, guest_addr, copy) \
608 unlock_user(host_ptr, guest_addr, (copy) ? sizeof(*host_ptr) : 0)
612 /* Include target-specific struct and function definitions;
613 * they may need access to the target-independent structures
614 * above, so include them last.
616 #include "target_cpu.h"
617 #include "target_signal.h"
618 #include "target_structs.h"