blob: 792c74290f8d376235b07f3f8ef084aa6e4715f7 [file] [log] [blame]
bellarde88de092005-02-07 12:35:39 +00001#ifndef QEMU_H
2#define QEMU_H
bellard31e31b82003-02-18 22:55:36 +00003
Timothy E Baldwin4d330ce2016-05-12 18:47:46 +01004#include "hostdep.h"
bellard6180a182003-09-30 21:04:53 +00005#include "cpu.h"
Paolo Bonzini63c91552016-03-15 13:18:37 +01006#include "exec/exec-all.h"
Paolo Bonzinif08b6172014-03-28 19:42:10 +01007#include "exec/cpu_ldst.h"
blueswir1992f48a2007-10-14 16:27:31 +00008
balrog06177d32007-12-24 13:47:52 +00009#undef DEBUG_REMAP
10#ifdef DEBUG_REMAP
balrog06177d32007-12-24 13:47:52 +000011#endif /* DEBUG_REMAP */
12
Paolo Bonzini022c62c2012-12-17 18:19:49 +010013#include "exec/user/abitypes.h"
blueswir1992f48a2007-10-14 16:27:31 +000014
Paolo Bonzini022c62c2012-12-17 18:19:49 +010015#include "exec/user/thunk.h"
blueswir1992f48a2007-10-14 16:27:31 +000016#include "syscall_defs.h"
Lluís Vilanova460c5792016-02-01 19:38:42 +010017#include "target_syscall.h"
Paolo Bonzini022c62c2012-12-17 18:19:49 +010018#include "exec/gdbstub.h"
bellard66fb9762003-03-23 01:06:05 +000019
Peter Maydell1d48fdd2016-06-14 12:49:18 +010020/* This is the size of the host kernel's sigset_t, needed where we make
21 * direct system calls that take a sigset_t pointer and a size.
22 */
23#define SIGSET_T_SIZE (_NSIG / 8)
24
bellard31e31b82003-02-18 22:55:36 +000025/* This struct is used to hold certain information about the image.
26 * Basically, it replicates in user space what would be certain
27 * task_struct fields in the kernel
28 */
29struct image_info {
Richard Henderson9955ffa2010-07-27 10:25:30 -070030 abi_ulong load_bias;
blueswir1992f48a2007-10-14 16:27:31 +000031 abi_ulong load_addr;
32 abi_ulong start_code;
33 abi_ulong end_code;
34 abi_ulong start_data;
35 abi_ulong end_data;
36 abi_ulong start_brk;
37 abi_ulong brk;
Richard Henderson6fd59442020-01-17 13:02:45 -100038 abi_ulong reserve_brk;
blueswir1992f48a2007-10-14 16:27:31 +000039 abi_ulong start_mmap;
blueswir1992f48a2007-10-14 16:27:31 +000040 abi_ulong start_stack;
Paul Brook97374d32010-06-16 13:03:51 +010041 abi_ulong stack_limit;
blueswir1992f48a2007-10-14 16:27:31 +000042 abi_ulong entry;
43 abi_ulong code_offset;
44 abi_ulong data_offset;
Mika Westerbergedf8e2a2009-04-07 09:57:11 +030045 abi_ulong saved_auxv;
Alexander Graf125b0f52012-01-28 21:12:14 +020046 abi_ulong auxv_len;
Mika Westerbergedf8e2a2009-04-07 09:57:11 +030047 abi_ulong arg_start;
48 abi_ulong arg_end;
Richard Henderson7c4ee5b2016-12-15 09:38:11 -080049 abi_ulong arg_strings;
50 abi_ulong env_strings;
51 abi_ulong file_string;
Paul Brookd8fd2952012-03-30 18:02:50 +010052 uint32_t elf_flags;
Paolo Bonzini7d374352018-12-13 23:37:37 +010053 int personality;
Laurent Vivier33143c42018-07-16 21:53:49 +020054 abi_ulong alignment;
Christophe Lyona99856c2018-04-30 10:03:41 +020055
56 /* The fields below are used in FDPIC mode. */
Mike Frysinger1af02e82011-02-07 01:05:50 -050057 abi_ulong loadmap_addr;
58 uint16_t nsegs;
59 void *loadsegs;
60 abi_ulong pt_dynamic_addr;
Christophe Lyon3cb10cf2018-04-30 10:03:43 +020061 abi_ulong interpreter_loadmap_addr;
62 abi_ulong interpreter_pt_dynamic_addr;
Mike Frysinger1af02e82011-02-07 01:05:50 -050063 struct image_info *other_info;
Stefan Markovic74cfc702018-10-22 16:43:40 +020064#ifdef TARGET_MIPS
65 int fp_abi;
66 int interp_fp_abi;
67#endif
bellard31e31b82003-02-18 22:55:36 +000068};
69
bellardb346ff42003-06-15 20:05:50 +000070#ifdef TARGET_I386
bellard851e67a2003-03-29 16:53:14 +000071/* Information about the current linux thread */
72struct vm86_saved_state {
73 uint32_t eax; /* return code */
74 uint32_t ebx;
75 uint32_t ecx;
76 uint32_t edx;
77 uint32_t esi;
78 uint32_t edi;
79 uint32_t ebp;
80 uint32_t esp;
81 uint32_t eflags;
82 uint32_t eip;
83 uint16_t cs, ss, ds, es, fs, gs;
84};
bellardb346ff42003-06-15 20:05:50 +000085#endif
bellard851e67a2003-03-29 16:53:14 +000086
Peter Maydell848d72c2013-09-03 20:12:17 +010087#if defined(TARGET_ARM) && defined(TARGET_ABI32)
bellard28c4f362004-02-16 21:47:43 +000088/* FPU emulator */
89#include "nwfpe/fpa11.h"
bellard28c4f362004-02-16 21:47:43 +000090#endif
91
pbrook624f7972008-05-31 16:11:38 +000092#define MAX_SIGQUEUE_SIZE 1024
93
pbrook624f7972008-05-31 16:11:38 +000094struct emulated_sigtable {
95 int pending; /* true if signal is pending */
Timothy E Baldwin907f5fd2016-05-27 15:51:52 +010096 target_siginfo_t info;
pbrook624f7972008-05-31 16:11:38 +000097};
98
bellard851e67a2003-03-29 16:53:14 +000099/* NOTE: we force a big alignment so that the stack stored after is
100 aligned too */
101typedef struct TaskState {
Mika Westerbergedf8e2a2009-04-07 09:57:11 +0300102 pid_t ts_tid; /* tid (or pid) of this task */
bellard28c4f362004-02-16 21:47:43 +0000103#ifdef TARGET_ARM
Peter Maydell848d72c2013-09-03 20:12:17 +0100104# ifdef TARGET_ABI32
bellard28c4f362004-02-16 21:47:43 +0000105 /* FPA state */
106 FPA11 fpa;
Peter Maydell848d72c2013-09-03 20:12:17 +0100107# endif
bellarda4f81972005-04-23 18:25:41 +0000108 int swi_errno;
bellard28c4f362004-02-16 21:47:43 +0000109#endif
j_mayer84409dd2007-04-06 08:56:50 +0000110#if defined(TARGET_I386) && !defined(TARGET_X86_64)
blueswir1992f48a2007-10-14 16:27:31 +0000111 abi_ulong target_v86;
bellard851e67a2003-03-29 16:53:14 +0000112 struct vm86_saved_state vm86_saved_regs;
bellardb333af02003-05-14 21:48:51 +0000113 struct target_vm86plus_struct vm86plus;
bellard631271d2003-05-10 13:14:52 +0000114 uint32_t v86flags;
115 uint32_t v86mask;
bellardb346ff42003-06-15 20:05:50 +0000116#endif
pbrookc2764712009-03-07 15:24:59 +0000117 abi_ulong child_tidptr;
pbrooke6e59062006-10-22 00:18:54 +0000118#ifdef TARGET_M68K
Peter Maydell1ccd9372013-07-16 18:44:55 +0100119 abi_ulong tp_value;
pbrooke6e59062006-10-22 00:18:54 +0000120#endif
Peter Maydelldaa43742018-03-08 14:47:32 +0000121#if defined(TARGET_ARM) || defined(TARGET_M68K)
pbrooka87295e2007-05-26 15:09:38 +0000122 /* Extra fields for semihosted binaries. */
Peter Maydelld3170912016-07-04 13:06:35 +0100123 abi_ulong heap_base;
124 abi_ulong heap_limit;
pbrooka87295e2007-05-26 15:09:38 +0000125#endif
Peter Maydelld3170912016-07-04 13:06:35 +0100126 abi_ulong stack_base;
bellard851e67a2003-03-29 16:53:14 +0000127 int used; /* non zero if used */
pbrook978efd62006-06-17 18:30:42 +0000128 struct image_info *info;
Mika Westerbergedf8e2a2009-04-07 09:57:11 +0300129 struct linux_binprm *bprm;
pbrook624f7972008-05-31 16:11:38 +0000130
Timothy E Baldwin655ed672016-05-27 15:51:53 +0100131 struct emulated_sigtable sync_signal;
pbrook624f7972008-05-31 16:11:38 +0000132 struct emulated_sigtable sigtab[TARGET_NSIG];
Peter Maydell3d3efba2016-05-27 15:51:49 +0100133 /* This thread's signal mask, as requested by the guest program.
134 * The actual signal mask of this thread may differ:
135 * + we don't let SIGSEGV and SIGBUS be blocked while running guest code
136 * + sometimes we block all signals to avoid races
137 */
138 sigset_t signal_mask;
139 /* The signal mask imposed by a guest sigsuspend syscall, if we are
140 * currently in the middle of such a syscall
141 */
142 sigset_t sigsuspend_mask;
143 /* Nonzero if we're leaving a sigsuspend and sigsuspend_mask is valid. */
144 int in_sigsuspend;
145
146 /* Nonzero if process_pending_signals() needs to do something (either
147 * handle a pending signal or unblock signals).
148 * This flag is written from a signal handler so should be accessed via
Emilio G. Cota83f164b2018-08-11 17:10:11 -0400149 * the atomic_read() and atomic_set() functions. (It is not accessed
Peter Maydell3d3efba2016-05-27 15:51:49 +0100150 * from multiple threads.)
151 */
152 int signal_pending;
153
Peter Maydell5bfce0b2019-07-25 14:16:45 +0100154 /* This thread's sigaltstack, if it has one */
155 struct target_sigaltstack sigaltstack_used;
bellard851e67a2003-03-29 16:53:14 +0000156} __attribute__((aligned(16))) TaskState;
157
aurel32d088d662009-01-30 20:09:01 +0000158extern char *exec_path;
pbrook624f7972008-05-31 16:11:38 +0000159void init_task_state(TaskState *ts);
Mika Westerbergedf8e2a2009-04-07 09:57:11 +0300160void task_settid(TaskState *);
161void stop_all_tasks(void);
pbrookc5937222006-05-14 11:30:38 +0000162extern const char *qemu_uname_release;
Paul Brook379f6692009-07-17 12:48:08 +0100163extern unsigned long mmap_min_addr;
bellard851e67a2003-03-29 16:53:14 +0000164
pbrooke5fe0c52006-06-11 13:32:59 +0000165/* ??? See if we can avoid exposing so much of the loader internals. */
pbrooke5fe0c52006-06-11 13:32:59 +0000166
Richard Henderson9955ffa2010-07-27 10:25:30 -0700167/* Read a good amount of data initially, to hopefully get all the
168 program headers loaded. */
169#define BPRM_BUF_SIZE 1024
170
pbrooke5fe0c52006-06-11 13:32:59 +0000171/*
ths5fafdf22007-09-16 21:08:06 +0000172 * This structure is used to hold the arguments that are
pbrooke5fe0c52006-06-11 13:32:59 +0000173 * used when loading binaries.
174 */
175struct linux_binprm {
Richard Henderson9955ffa2010-07-27 10:25:30 -0700176 char buf[BPRM_BUF_SIZE] __attribute__((aligned));
blueswir1992f48a2007-10-14 16:27:31 +0000177 abi_ulong p;
Paolo Bonzini7d374352018-12-13 23:37:37 +0100178 int fd;
pbrooke5fe0c52006-06-11 13:32:59 +0000179 int e_uid, e_gid;
180 int argc, envc;
181 char **argv;
182 char **envp;
183 char * filename; /* Name of binary */
Andreas Färber9349b4f2012-03-14 01:38:32 +0100184 int (*core_dump)(int, const CPUArchState *); /* coredump routine */
pbrooke5fe0c52006-06-11 13:32:59 +0000185};
186
187void do_init_thread(struct target_pt_regs *regs, struct image_info *infop);
blueswir1992f48a2007-10-14 16:27:31 +0000188abi_ulong loader_build_argptr(int envc, int argc, abi_ulong sp,
189 abi_ulong stringp, int push_ptr);
Laurent Vivier03cfd8f2013-08-30 01:46:44 +0200190int loader_exec(int fdexec, const char *filename, char **argv, char **envp,
Mika Westerbergedf8e2a2009-04-07 09:57:11 +0300191 struct target_pt_regs * regs, struct image_info *infop,
192 struct linux_binprm *);
bellard31e31b82003-02-18 22:55:36 +0000193
Christophe Lyon3cb10cf2018-04-30 10:03:43 +0200194/* Returns true if the image uses the FDPIC ABI. If this is the case,
195 * we have to provide some information (loadmap, pt_dynamic_info) such
196 * that the program can be relocated adequately. This is also useful
197 * when handling signals.
198 */
199int info_is_fdpic(struct image_info *info);
200
YunQiang Su768fe762018-02-20 18:33:05 +0100201uint32_t get_elf_eflags(int fd);
Will Newtonf0116c52014-01-09 09:10:50 +0000202int load_elf_binary(struct linux_binprm *bprm, struct image_info *info);
203int load_flt_binary(struct linux_binprm *bprm, struct image_info *info);
pbrooke5fe0c52006-06-11 13:32:59 +0000204
bellard579a97f2007-11-11 14:26:47 +0000205abi_long memcpy_to_target(abi_ulong dest, const void *src,
206 unsigned long len);
blueswir1992f48a2007-10-14 16:27:31 +0000207void target_set_brk(abi_ulong new_brk);
208abi_long do_brk(abi_ulong new_brk);
bellard31e31b82003-02-18 22:55:36 +0000209void syscall_init(void);
blueswir1992f48a2007-10-14 16:27:31 +0000210abi_long do_syscall(void *cpu_env, int num, abi_long arg1,
211 abi_long arg2, abi_long arg3, abi_long arg4,
Peter Maydell5945cfc2011-06-16 17:37:13 +0100212 abi_long arg5, abi_long arg6, abi_long arg7,
213 abi_long arg8);
Peter Maydellb44316f2018-02-13 13:22:46 +0000214extern __thread CPUState *thread_cpu;
Andreas Färber9349b4f2012-03-14 01:38:32 +0100215void cpu_loop(CPUArchState *env);
Peter Maydell7dcdaea2016-06-06 19:58:18 +0100216const char *target_strerror(int err);
pbrooka745ec62008-05-06 15:36:17 +0000217int get_osversion(void);
Peter Maydell4a24a752013-09-03 20:12:20 +0100218void init_qemu_uname_release(void);
pbrookd5975362008-06-07 20:50:51 +0000219void fork_start(void);
220void fork_end(int child);
bellard6977fbf2003-04-29 20:39:23 +0000221
Meador Ingedce10402012-07-26 16:50:01 +0000222/* Creates the initial guest address space in the host memory space using
223 * the given host start address hint and size. The guest_start parameter
224 * specifies the start address of the guest space. guest_base will be the
225 * difference between the host start address computed by this function and
226 * guest_start. If fixed is specified, then the mapped address space must
227 * start at host_start. The real start address of the mapped memory space is
228 * returned or -1 if there was an error.
229 */
230unsigned long init_guest_space(unsigned long host_start,
231 unsigned long host_size,
232 unsigned long guest_start,
233 bool fixed);
234
Paolo Bonzini1de7afc2012-12-17 18:20:00 +0100235#include "qemu/log.h"
bellard631271d2003-05-10 13:14:52 +0000236
Timothy E Baldwin4d330ce2016-05-12 18:47:46 +0100237/* safe_syscall.S */
238
239/**
240 * safe_syscall:
241 * @int number: number of system call to make
242 * ...: arguments to the system call
243 *
244 * Call a system call if guest signal not pending.
245 * This has the same API as the libc syscall() function, except that it
246 * may return -1 with errno == TARGET_ERESTARTSYS if a signal was pending.
247 *
248 * Returns: the system call result, or -1 with an error code in errno
249 * (Errnos are host errnos; we rely on TARGET_ERESTARTSYS not clashing
250 * with any of the host errno values.)
251 */
252
253/* A guide to using safe_syscall() to handle interactions between guest
254 * syscalls and guest signals:
255 *
256 * Guest syscalls come in two flavours:
257 *
258 * (1) Non-interruptible syscalls
259 *
260 * These are guest syscalls that never get interrupted by signals and
261 * so never return EINTR. They can be implemented straightforwardly in
262 * QEMU: just make sure that if the implementation code has to make any
263 * blocking calls that those calls are retried if they return EINTR.
264 * It's also OK to implement these with safe_syscall, though it will be
265 * a little less efficient if a signal is delivered at the 'wrong' moment.
266 *
Peter Maydell3d3efba2016-05-27 15:51:49 +0100267 * Some non-interruptible syscalls need to be handled using block_signals()
268 * to block signals for the duration of the syscall. This mainly applies
269 * to code which needs to modify the data structures used by the
270 * host_signal_handler() function and the functions it calls, including
271 * all syscalls which change the thread's signal mask.
272 *
Timothy E Baldwin4d330ce2016-05-12 18:47:46 +0100273 * (2) Interruptible syscalls
274 *
275 * These are guest syscalls that can be interrupted by signals and
276 * for which we need to either return EINTR or arrange for the guest
277 * syscall to be restarted. This category includes both syscalls which
278 * always restart (and in the kernel return -ERESTARTNOINTR), ones
279 * which only restart if there is no handler (kernel returns -ERESTARTNOHAND
280 * or -ERESTART_RESTARTBLOCK), and the most common kind which restart
281 * if the handler was registered with SA_RESTART (kernel returns
282 * -ERESTARTSYS). System calls which are only interruptible in some
283 * situations (like 'open') also need to be handled this way.
284 *
285 * Here it is important that the host syscall is made
286 * via this safe_syscall() function, and *not* via the host libc.
287 * If the host libc is used then the implementation will appear to work
288 * most of the time, but there will be a race condition where a
289 * signal could arrive just before we make the host syscall inside libc,
290 * and then then guest syscall will not correctly be interrupted.
291 * Instead the implementation of the guest syscall can use the safe_syscall
292 * function but otherwise just return the result or errno in the usual
293 * way; the main loop code will take care of restarting the syscall
294 * if appropriate.
295 *
296 * (If the implementation needs to make multiple host syscalls this is
297 * OK; any which might really block must be via safe_syscall(); for those
298 * which are only technically blocking (ie which we know in practice won't
299 * stay in the host kernel indefinitely) it's OK to use libc if necessary.
300 * You must be able to cope with backing out correctly if some safe_syscall
301 * you make in the implementation returns either -TARGET_ERESTARTSYS or
302 * EINTR though.)
303 *
Peter Maydell3d3efba2016-05-27 15:51:49 +0100304 * block_signals() cannot be used for interruptible syscalls.
305 *
Timothy E Baldwin4d330ce2016-05-12 18:47:46 +0100306 *
307 * How and why the safe_syscall implementation works:
308 *
309 * The basic setup is that we make the host syscall via a known
310 * section of host native assembly. If a signal occurs, our signal
311 * handler checks the interrupted host PC against the addresse of that
312 * known section. If the PC is before or at the address of the syscall
313 * instruction then we change the PC to point at a "return
314 * -TARGET_ERESTARTSYS" code path instead, and then exit the signal handler
315 * (causing the safe_syscall() call to immediately return that value).
316 * Then in the main.c loop if we see this magic return value we adjust
317 * the guest PC to wind it back to before the system call, and invoke
318 * the guest signal handler as usual.
319 *
320 * This winding-back will happen in two cases:
321 * (1) signal came in just before we took the host syscall (a race);
322 * in this case we'll take the guest signal and have another go
323 * at the syscall afterwards, and this is indistinguishable for the
324 * guest from the timing having been different such that the guest
325 * signal really did win the race
326 * (2) signal came in while the host syscall was blocking, and the
327 * host kernel decided the syscall should be restarted;
328 * in this case we want to restart the guest syscall also, and so
329 * rewinding is the right thing. (Note that "restart" semantics mean
330 * "first call the signal handler, then reattempt the syscall".)
331 * The other situation to consider is when a signal came in while the
332 * host syscall was blocking, and the host kernel decided that the syscall
333 * should not be restarted; in this case QEMU's host signal handler will
334 * be invoked with the PC pointing just after the syscall instruction,
335 * with registers indicating an EINTR return; the special code in the
336 * handler will not kick in, and we will return EINTR to the guest as
337 * we should.
338 *
339 * Notice that we can leave the host kernel to make the decision for
340 * us about whether to do a restart of the syscall or not; we do not
341 * need to check SA_RESTART flags in QEMU or distinguish the various
342 * kinds of restartability.
343 */
344#ifdef HAVE_SAFE_SYSCALL
345/* The core part of this function is implemented in assembly */
346extern long safe_syscall_base(int *pending, long number, ...);
347
348#define safe_syscall(...) \
349 ({ \
350 long ret_; \
351 int *psp_ = &((TaskState *)thread_cpu->opaque)->signal_pending; \
352 ret_ = safe_syscall_base(psp_, __VA_ARGS__); \
353 if (is_error(ret_)) { \
354 errno = -ret_; \
355 ret_ = -1; \
356 } \
357 ret_; \
358 })
359
360#else
361
362/* Fallback for architectures which don't yet provide a safe-syscall assembly
363 * fragment; note that this is racy!
364 * This should go away when all host architectures have been updated.
365 */
366#define safe_syscall syscall
367
368#endif
369
Richard Hendersona05c6402012-09-15 11:34:20 -0700370/* syscall.c */
371int host_to_target_waitstatus(int status);
372
thsb92c47c2007-11-01 00:07:38 +0000373/* strace.c */
374void print_syscall(int num,
bellardc16f9ed2007-11-11 17:23:29 +0000375 abi_long arg1, abi_long arg2, abi_long arg3,
376 abi_long arg4, abi_long arg5, abi_long arg6);
377void print_syscall_ret(int num, abi_long arg1);
Peter Maydell0cb581d2016-07-18 18:12:24 +0100378/**
379 * print_taken_signal:
380 * @target_signum: target signal being taken
381 * @tinfo: target_siginfo_t which will be passed to the guest for the signal
382 *
383 * Print strace output indicating that this signal is being taken by the guest,
384 * in a format similar to:
385 * --- SIGSEGV {si_signo=SIGSEGV, si_code=SI_KERNEL, si_addr=0} ---
386 */
387void print_taken_signal(int target_signum, const target_siginfo_t *tinfo);
thsb92c47c2007-11-01 00:07:38 +0000388
bellardb346ff42003-06-15 20:05:50 +0000389/* signal.c */
Andreas Färber9349b4f2012-03-14 01:38:32 +0100390void process_pending_signals(CPUArchState *cpu_env);
bellardb346ff42003-06-15 20:05:50 +0000391void signal_init(void);
Peter Maydell9d2803f2016-07-28 16:44:46 +0100392int queue_signal(CPUArchState *env, int sig, int si_type,
393 target_siginfo_t *info);
Anthony Liguoric227f092009-10-01 16:12:16 -0500394void host_to_target_siginfo(target_siginfo_t *tinfo, const siginfo_t *info);
395void target_to_host_siginfo(siginfo_t *info, const target_siginfo_t *tinfo);
pbrook4cb05962008-05-30 18:05:19 +0000396int target_to_host_signal(int sig);
pbrook1d9d8b52009-04-16 15:17:02 +0000397int host_to_target_signal(int sig);
Andreas Färber9349b4f2012-03-14 01:38:32 +0100398long do_sigreturn(CPUArchState *env);
399long do_rt_sigreturn(CPUArchState *env);
bellard579a97f2007-11-11 14:26:47 +0000400abi_long do_sigaltstack(abi_ulong uss_addr, abi_ulong uoss_addr, abi_ulong sp);
Alex Barcelo1c275922014-03-14 14:36:55 +0000401int do_sigprocmask(int how, const sigset_t *set, sigset_t *oldset);
Richard Hendersonfa97e382018-07-18 13:06:48 -0700402abi_long do_swapcontext(CPUArchState *env, abi_ulong uold_ctx,
403 abi_ulong unew_ctx, abi_long ctx_size);
Peter Maydell3d3efba2016-05-27 15:51:49 +0100404/**
405 * block_signals: block all signals while handling this guest syscall
406 *
407 * Block all signals, and arrange that the signal mask is returned to
408 * its correct value for the guest before we resume execution of guest code.
409 * If this function returns non-zero, then the caller should immediately
410 * return -TARGET_ERESTARTSYS to the main loop, which will take the pending
411 * signal and restart execution of the syscall.
412 * If block_signals() returns zero, then the caller can continue with
413 * emulation of the system call knowing that no signals can be taken
414 * (and therefore that no race conditions will result).
415 * This should only be called once, because if it is called a second time
416 * it will always return non-zero. (Think of it like a mutex that can't
417 * be recursively locked.)
418 * Signals will be unblocked again by process_pending_signals().
419 *
420 * Return value: non-zero if there was a pending signal, zero if not.
421 */
422int block_signals(void); /* Returns non zero if signal pending */
bellardb346ff42003-06-15 20:05:50 +0000423
424#ifdef TARGET_I386
bellard631271d2003-05-10 13:14:52 +0000425/* vm86.c */
426void save_v86_state(CPUX86State *env);
bellard447db212003-05-10 15:10:36 +0000427void handle_vm86_trap(CPUX86State *env, int trapno);
bellard631271d2003-05-10 13:14:52 +0000428void handle_vm86_fault(CPUX86State *env);
blueswir1992f48a2007-10-14 16:27:31 +0000429int do_vm86(CPUX86State *env, long subfunction, abi_ulong v86_addr);
blueswir15bfb56b2007-10-05 17:01:51 +0000430#elif defined(TARGET_SPARC64)
431void sparc64_set_context(CPUSPARCState *env);
432void sparc64_get_context(CPUSPARCState *env);
bellardb346ff42003-06-15 20:05:50 +0000433#endif
bellard631271d2003-05-10 13:14:52 +0000434
bellard54936002003-05-13 00:25:15 +0000435/* mmap.c */
blueswir1992f48a2007-10-14 16:27:31 +0000436int target_mprotect(abi_ulong start, abi_ulong len, int prot);
437abi_long target_mmap(abi_ulong start, abi_ulong len, int prot,
438 int flags, int fd, abi_ulong offset);
439int target_munmap(abi_ulong start, abi_ulong len);
440abi_long target_mremap(abi_ulong old_addr, abi_ulong old_size,
441 abi_ulong new_size, unsigned long flags,
442 abi_ulong new_addr);
pbrook07765902008-05-31 16:33:53 +0000443extern unsigned long last_brk;
Peter Maydell59e9d912012-03-08 14:40:33 +0000444extern abi_ulong mmap_next_start;
Richard Henderson30ab9ef2019-05-19 13:19:52 -0700445abi_ulong mmap_find_vma(abi_ulong, abi_ulong, abi_ulong);
pbrookd5975362008-06-07 20:50:51 +0000446void mmap_fork_start(void);
447void mmap_fork_end(int child);
bellard54936002003-05-13 00:25:15 +0000448
blueswir1440c7e82008-10-05 11:05:14 +0000449/* main.c */
Richard Henderson703e0e82010-03-19 14:21:13 -0700450extern unsigned long guest_stack_size;
blueswir1440c7e82008-10-05 11:05:14 +0000451
bellardedf779f2004-02-22 13:40:13 +0000452/* user access */
453
454#define VERIFY_READ 0
bellard579a97f2007-11-11 14:26:47 +0000455#define VERIFY_WRITE 1 /* implies read access */
bellardedf779f2004-02-22 13:40:13 +0000456
bellarddae32702007-11-14 10:51:00 +0000457static inline int access_ok(int type, abi_ulong addr, abi_ulong size)
458{
Rémi Denis-Courmont0acd4ab2019-07-04 10:41:15 +0200459 return guest_addr_valid(addr) &&
460 (size == 0 || guest_addr_valid(addr + size - 1)) &&
461 page_check_range((target_ulong)addr, size,
bellarddae32702007-11-14 10:51:00 +0000462 (type == VERIFY_READ) ? PAGE_READ : (PAGE_READ | PAGE_WRITE)) == 0;
463}
bellardedf779f2004-02-22 13:40:13 +0000464
Richard Henderson658f2dc2013-01-04 16:39:31 -0800465/* NOTE __get_user and __put_user use host pointers and don't check access.
466 These are usually used to access struct data members once the struct has
467 been locked - usually with lock_user_struct. */
bellardedf779f2004-02-22 13:40:13 +0000468
Peter Maydell850d5e32018-10-09 17:18:14 +0100469/*
470 * Tricky points:
471 * - Use __builtin_choose_expr to avoid type promotion from ?:,
472 * - Invalid sizes result in a compile time error stemming from
473 * the fact that abort has no parameters.
474 * - It's easier to use the endian-specific unaligned load/store
475 * functions than host-endian unaligned load/store plus tswapN.
476 * - The pragmas are necessary only to silence a clang false-positive
477 * warning: see https://bugs.llvm.org/show_bug.cgi?id=39113 .
Peter Maydell850d5e32018-10-09 17:18:14 +0100478 * - gcc has bugs in its _Pragma() support in some versions, eg
479 * https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83256 -- so we only
480 * include the warning-suppression pragmas for clang
481 */
Thomas Huth798b8582018-11-30 09:23:16 +0100482#if defined(__clang__) && __has_warning("-Waddress-of-packed-member")
Peter Maydell850d5e32018-10-09 17:18:14 +0100483#define PRAGMA_DISABLE_PACKED_WARNING \
484 _Pragma("GCC diagnostic push"); \
Peter Maydell850d5e32018-10-09 17:18:14 +0100485 _Pragma("GCC diagnostic ignored \"-Waddress-of-packed-member\"")
Richard Henderson658f2dc2013-01-04 16:39:31 -0800486
Peter Maydell850d5e32018-10-09 17:18:14 +0100487#define PRAGMA_REENABLE_PACKED_WARNING \
488 _Pragma("GCC diagnostic pop")
Richard Henderson658f2dc2013-01-04 16:39:31 -0800489
Peter Maydell850d5e32018-10-09 17:18:14 +0100490#else
491#define PRAGMA_DISABLE_PACKED_WARNING
492#define PRAGMA_REENABLE_PACKED_WARNING
493#endif
494
495#define __put_user_e(x, hptr, e) \
496 do { \
497 PRAGMA_DISABLE_PACKED_WARNING; \
498 (__builtin_choose_expr(sizeof(*(hptr)) == 1, stb_p, \
499 __builtin_choose_expr(sizeof(*(hptr)) == 2, stw_##e##_p, \
500 __builtin_choose_expr(sizeof(*(hptr)) == 4, stl_##e##_p, \
501 __builtin_choose_expr(sizeof(*(hptr)) == 8, stq_##e##_p, abort)))) \
502 ((hptr), (x)), (void)0); \
503 PRAGMA_REENABLE_PACKED_WARNING; \
504 } while (0)
505
506#define __get_user_e(x, hptr, e) \
507 do { \
508 PRAGMA_DISABLE_PACKED_WARNING; \
509 ((x) = (typeof(*hptr))( \
510 __builtin_choose_expr(sizeof(*(hptr)) == 1, ldub_p, \
511 __builtin_choose_expr(sizeof(*(hptr)) == 2, lduw_##e##_p, \
512 __builtin_choose_expr(sizeof(*(hptr)) == 4, ldl_##e##_p, \
513 __builtin_choose_expr(sizeof(*(hptr)) == 8, ldq_##e##_p, abort)))) \
514 (hptr)), (void)0); \
515 PRAGMA_REENABLE_PACKED_WARNING; \
516 } while (0)
517
Richard Henderson658f2dc2013-01-04 16:39:31 -0800518
519#ifdef TARGET_WORDS_BIGENDIAN
520# define __put_user(x, hptr) __put_user_e(x, hptr, be)
521# define __get_user(x, hptr) __get_user_e(x, hptr, be)
522#else
523# define __put_user(x, hptr) __put_user_e(x, hptr, le)
524# define __get_user(x, hptr) __get_user_e(x, hptr, le)
525#endif
bellardedf779f2004-02-22 13:40:13 +0000526
bellard579a97f2007-11-11 14:26:47 +0000527/* put_user()/get_user() take a guest address and check access */
528/* These are usually used to access an atomic data type, such as an int,
529 * that has been passed by address. These internally perform locking
530 * and unlocking on the data type.
531 */
532#define put_user(x, gaddr, target_type) \
533({ \
534 abi_ulong __gaddr = (gaddr); \
535 target_type *__hptr; \
Riku Voipioa42267e2014-04-22 15:40:50 +0300536 abi_long __ret = 0; \
bellard579a97f2007-11-11 14:26:47 +0000537 if ((__hptr = lock_user(VERIFY_WRITE, __gaddr, sizeof(target_type), 0))) { \
Riku Voipioa42267e2014-04-22 15:40:50 +0300538 __put_user((x), __hptr); \
bellard579a97f2007-11-11 14:26:47 +0000539 unlock_user(__hptr, __gaddr, sizeof(target_type)); \
540 } else \
541 __ret = -TARGET_EFAULT; \
542 __ret; \
bellardedf779f2004-02-22 13:40:13 +0000543})
544
bellard579a97f2007-11-11 14:26:47 +0000545#define get_user(x, gaddr, target_type) \
546({ \
547 abi_ulong __gaddr = (gaddr); \
548 target_type *__hptr; \
Riku Voipioa42267e2014-04-22 15:40:50 +0300549 abi_long __ret = 0; \
bellard579a97f2007-11-11 14:26:47 +0000550 if ((__hptr = lock_user(VERIFY_READ, __gaddr, sizeof(target_type), 1))) { \
Riku Voipioa42267e2014-04-22 15:40:50 +0300551 __get_user((x), __hptr); \
bellard579a97f2007-11-11 14:26:47 +0000552 unlock_user(__hptr, __gaddr, 0); \
bellard2f619692007-11-16 10:46:05 +0000553 } else { \
554 /* avoid warning */ \
555 (x) = 0; \
bellard579a97f2007-11-11 14:26:47 +0000556 __ret = -TARGET_EFAULT; \
bellard2f619692007-11-16 10:46:05 +0000557 } \
bellard579a97f2007-11-11 14:26:47 +0000558 __ret; \
bellardedf779f2004-02-22 13:40:13 +0000559})
560
bellard2f619692007-11-16 10:46:05 +0000561#define put_user_ual(x, gaddr) put_user((x), (gaddr), abi_ulong)
562#define put_user_sal(x, gaddr) put_user((x), (gaddr), abi_long)
563#define put_user_u64(x, gaddr) put_user((x), (gaddr), uint64_t)
564#define put_user_s64(x, gaddr) put_user((x), (gaddr), int64_t)
565#define put_user_u32(x, gaddr) put_user((x), (gaddr), uint32_t)
566#define put_user_s32(x, gaddr) put_user((x), (gaddr), int32_t)
567#define put_user_u16(x, gaddr) put_user((x), (gaddr), uint16_t)
568#define put_user_s16(x, gaddr) put_user((x), (gaddr), int16_t)
569#define put_user_u8(x, gaddr) put_user((x), (gaddr), uint8_t)
570#define put_user_s8(x, gaddr) put_user((x), (gaddr), int8_t)
571
572#define get_user_ual(x, gaddr) get_user((x), (gaddr), abi_ulong)
573#define get_user_sal(x, gaddr) get_user((x), (gaddr), abi_long)
574#define get_user_u64(x, gaddr) get_user((x), (gaddr), uint64_t)
575#define get_user_s64(x, gaddr) get_user((x), (gaddr), int64_t)
576#define get_user_u32(x, gaddr) get_user((x), (gaddr), uint32_t)
577#define get_user_s32(x, gaddr) get_user((x), (gaddr), int32_t)
578#define get_user_u16(x, gaddr) get_user((x), (gaddr), uint16_t)
579#define get_user_s16(x, gaddr) get_user((x), (gaddr), int16_t)
580#define get_user_u8(x, gaddr) get_user((x), (gaddr), uint8_t)
581#define get_user_s8(x, gaddr) get_user((x), (gaddr), int8_t)
582
bellard579a97f2007-11-11 14:26:47 +0000583/* copy_from_user() and copy_to_user() are usually used to copy data
584 * buffers between the target and host. These internally perform
585 * locking/unlocking of the memory.
586 */
587abi_long copy_from_user(void *hptr, abi_ulong gaddr, size_t len);
588abi_long copy_to_user(abi_ulong gaddr, void *hptr, size_t len);
589
pbrook53a59602006-03-25 19:31:22 +0000590/* Functions for accessing guest memory. The tget and tput functions
Stefan Weil6f20f552013-09-12 19:57:15 +0200591 read/write single values, byteswapping as necessary. The lock_user function
pbrook53a59602006-03-25 19:31:22 +0000592 gets a pointer to a contiguous area of guest memory, but does not perform
Stefan Weil6f20f552013-09-12 19:57:15 +0200593 any byteswapping. lock_user may return either a pointer to the guest
pbrook53a59602006-03-25 19:31:22 +0000594 memory, or a temporary buffer. */
595
596/* Lock an area of guest memory into the host. If copy is true then the
597 host area will have the same contents as the guest. */
bellard579a97f2007-11-11 14:26:47 +0000598static inline void *lock_user(int type, abi_ulong guest_addr, long len, int copy)
bellardedf779f2004-02-22 13:40:13 +0000599{
bellard579a97f2007-11-11 14:26:47 +0000600 if (!access_ok(type, guest_addr, len))
601 return NULL;
pbrook53a59602006-03-25 19:31:22 +0000602#ifdef DEBUG_REMAP
bellard579a97f2007-11-11 14:26:47 +0000603 {
604 void *addr;
Md Haris Iqbal38e826d2016-03-24 00:02:03 +0530605 addr = g_malloc(len);
bellard579a97f2007-11-11 14:26:47 +0000606 if (copy)
607 memcpy(addr, g2h(guest_addr), len);
608 else
609 memset(addr, 0, len);
610 return addr;
611 }
pbrook53a59602006-03-25 19:31:22 +0000612#else
613 return g2h(guest_addr);
614#endif
bellardedf779f2004-02-22 13:40:13 +0000615}
616
bellard579a97f2007-11-11 14:26:47 +0000617/* Unlock an area of guest memory. The first LEN bytes must be
ths1235fc02008-06-03 19:51:57 +0000618 flushed back to guest memory. host_ptr = NULL is explicitly
bellard579a97f2007-11-11 14:26:47 +0000619 allowed and does nothing. */
620static inline void unlock_user(void *host_ptr, abi_ulong guest_addr,
blueswir1992f48a2007-10-14 16:27:31 +0000621 long len)
bellardedf779f2004-02-22 13:40:13 +0000622{
bellard579a97f2007-11-11 14:26:47 +0000623
pbrook53a59602006-03-25 19:31:22 +0000624#ifdef DEBUG_REMAP
bellard579a97f2007-11-11 14:26:47 +0000625 if (!host_ptr)
626 return;
627 if (host_ptr == g2h(guest_addr))
pbrook53a59602006-03-25 19:31:22 +0000628 return;
629 if (len > 0)
balrog06177d32007-12-24 13:47:52 +0000630 memcpy(g2h(guest_addr), host_ptr, len);
Md Haris Iqbal38e826d2016-03-24 00:02:03 +0530631 g_free(host_ptr);
pbrook53a59602006-03-25 19:31:22 +0000632#endif
bellardedf779f2004-02-22 13:40:13 +0000633}
634
bellard579a97f2007-11-11 14:26:47 +0000635/* Return the length of a string in target memory or -TARGET_EFAULT if
636 access error. */
637abi_long target_strlen(abi_ulong gaddr);
bellardedf779f2004-02-22 13:40:13 +0000638
pbrook53a59602006-03-25 19:31:22 +0000639/* Like lock_user but for null terminated strings. */
blueswir1992f48a2007-10-14 16:27:31 +0000640static inline void *lock_user_string(abi_ulong guest_addr)
pbrook53a59602006-03-25 19:31:22 +0000641{
bellard579a97f2007-11-11 14:26:47 +0000642 abi_long len;
643 len = target_strlen(guest_addr);
644 if (len < 0)
645 return NULL;
646 return lock_user(VERIFY_READ, guest_addr, (long)(len + 1), 1);
pbrook53a59602006-03-25 19:31:22 +0000647}
648
Stefan Weil41d1af42013-09-12 19:57:41 +0200649/* Helper macros for locking/unlocking a target struct. */
bellard579a97f2007-11-11 14:26:47 +0000650#define lock_user_struct(type, host_ptr, guest_addr, copy) \
651 (host_ptr = lock_user(type, guest_addr, sizeof(*host_ptr), copy))
652#define unlock_user_struct(host_ptr, guest_addr, copy) \
pbrook53a59602006-03-25 19:31:22 +0000653 unlock_user(host_ptr, guest_addr, (copy) ? sizeof(*host_ptr) : 0)
654
pbrookc8a706f2008-06-02 16:16:42 +0000655#include <pthread.h>
pbrookc8a706f2008-06-02 16:16:42 +0000656
Philippe Mathieu-Daudé1129dd72018-06-04 12:37:21 -0300657static inline int is_error(abi_long ret)
658{
659 return (abi_ulong)ret >= (abi_ulong)(-4096);
660}
661
Alex Bennée708b6a62018-06-22 17:09:10 +0100662/**
663 * preexit_cleanup: housekeeping before the guest exits
664 *
665 * env: the CPU state
666 * code: the exit code
667 */
668void preexit_cleanup(CPUArchState *env, int code);
669
Peter Maydelldfeab062013-07-16 18:44:52 +0100670/* Include target-specific struct and function definitions;
671 * they may need access to the target-independent structures
672 * above, so include them last.
673 */
674#include "target_cpu.h"
Petar Jovanovic55a2b162013-10-30 14:46:31 +0100675#include "target_structs.h"
Peter Maydelldfeab062013-07-16 18:44:52 +0100676
bellarde88de092005-02-07 12:35:39 +0000677#endif /* QEMU_H */