4 * Copyright (c) 2003-2008 Fabrice Bellard
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License as published by
8 * the Free Software Foundation; either version 2 of the License, or
9 * (at your option) any later version.
11 * This program is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 * GNU General Public License for more details.
16 * You should have received a copy of the GNU General Public License
17 * along with this program; if not, see <http://www.gnu.org/licenses/>.
20 #include "qemu/osdep.h"
21 #include "qemu-common.h"
22 #include "qemu/units.h"
23 #include "sysemu/tcg.h"
24 #include "qemu-version.h"
25 #include <sys/syscall.h>
26 #include <sys/resource.h>
28 #include "qapi/error.h"
30 #include "qemu/path.h"
31 #include "qemu/queue.h"
32 #include "qemu/config-file.h"
33 #include "qemu/cutils.h"
34 #include "qemu/error-report.h"
35 #include "qemu/help_option.h"
36 #include "qemu/module.h"
37 #include "qemu/plugin.h"
39 #include "exec/exec-all.h"
41 #include "qemu/timer.h"
42 #include "qemu/envlist.h"
43 #include "qemu/guest-random.h"
45 #include "trace/control.h"
46 #include "target_elf.h"
47 #include "cpu_loop-common.h"
48 #include "crypto/init.h"
53 static const char *argv0
;
54 static int gdbstub_port
;
55 static envlist_t
*envlist
;
56 static const char *cpu_model
;
57 static const char *cpu_type
;
58 static const char *seed_optarg
;
59 unsigned long mmap_min_addr
;
60 unsigned long guest_base
;
64 * When running 32-on-64 we should make sure we can fit all of the possible
65 * guest address space into a contiguous chunk of virtual host memory.
67 * This way we will never overlap with our own libraries or binaries or stack
68 * or anything else that QEMU maps.
70 * Many cpus reserve the high bit (or more than one for some 64-bit cpus)
71 * of the address for the kernel. Some cpus rely on this and user space
72 * uses the high bit(s) for pointer tagging and the like. For them, we
73 * must preserve the expected address space.
75 #ifndef MAX_RESERVED_VA
76 # if HOST_LONG_BITS > TARGET_VIRT_ADDR_SPACE_BITS
77 # if TARGET_VIRT_ADDR_SPACE_BITS == 32 && \
78 (TARGET_LONG_BITS == 32 || defined(TARGET_ABI32))
79 /* There are a number of places where we assign reserved_va to a variable
80 of type abi_ulong and expect it to fit. Avoid the last page. */
81 # define MAX_RESERVED_VA(CPU) (0xfffffffful & TARGET_PAGE_MASK)
83 # define MAX_RESERVED_VA(CPU) (1ul << TARGET_VIRT_ADDR_SPACE_BITS)
86 # define MAX_RESERVED_VA(CPU) 0
90 unsigned long reserved_va
;
92 static void usage(int exitcode
);
94 static const char *interp_prefix
= CONFIG_QEMU_INTERP_PREFIX
;
95 const char *qemu_uname_release
;
97 /* XXX: on x86 MAP_GROWSDOWN only works if ESP <= address + 32, so
98 we allocate a bigger stack. Need a better solution, for example
99 by remapping the process stack directly at the right place */
100 unsigned long guest_stack_size
= 8 * 1024 * 1024UL;
102 void gemu_log(const char *fmt
, ...)
107 vfprintf(stderr
, fmt
, ap
);
111 #if defined(TARGET_I386)
112 int cpu_get_pic_interrupt(CPUX86State
*env
)
118 /***********************************************************/
119 /* Helper routines for implementing atomic operations. */
121 /* Make sure everything is in a consistent state for calling fork(). */
122 void fork_start(void)
129 void fork_end(int child
)
131 mmap_fork_end(child
);
133 CPUState
*cpu
, *next_cpu
;
134 /* Child processes created by fork() only have a single thread.
135 Discard information about the parent threads. */
136 CPU_FOREACH_SAFE(cpu
, next_cpu
) {
137 if (cpu
!= thread_cpu
) {
138 QTAILQ_REMOVE_RCU(&cpus
, cpu
, node
);
141 qemu_init_cpu_list();
142 gdbserver_fork(thread_cpu
);
143 /* qemu_init_cpu_list() takes care of reinitializing the
144 * exclusive state, so we don't need to end_exclusive() here.
152 __thread CPUState
*thread_cpu
;
154 bool qemu_cpu_is_self(CPUState
*cpu
)
156 return thread_cpu
== cpu
;
159 void qemu_cpu_kick(CPUState
*cpu
)
164 void task_settid(TaskState
*ts
)
166 if (ts
->ts_tid
== 0) {
167 ts
->ts_tid
= (pid_t
)syscall(SYS_gettid
);
171 void stop_all_tasks(void)
174 * We trust that when using NPTL, start_exclusive()
175 * handles thread stopping correctly.
180 /* Assumes contents are already zeroed. */
181 void init_task_state(TaskState
*ts
)
184 ts
->sigaltstack_used
= (struct target_sigaltstack
) {
187 .ss_flags
= TARGET_SS_DISABLE
,
191 CPUArchState
*cpu_copy(CPUArchState
*env
)
193 CPUState
*cpu
= env_cpu(env
);
194 CPUState
*new_cpu
= cpu_create(cpu_type
);
195 CPUArchState
*new_env
= new_cpu
->env_ptr
;
199 /* Reset non arch specific state */
202 memcpy(new_env
, env
, sizeof(CPUArchState
));
204 /* Clone all break/watchpoints.
205 Note: Once we support ptrace with hw-debug register access, make sure
206 BP_CPU break/watchpoints are handled correctly on clone. */
207 QTAILQ_INIT(&new_cpu
->breakpoints
);
208 QTAILQ_INIT(&new_cpu
->watchpoints
);
209 QTAILQ_FOREACH(bp
, &cpu
->breakpoints
, entry
) {
210 cpu_breakpoint_insert(new_cpu
, bp
->pc
, bp
->flags
, NULL
);
212 QTAILQ_FOREACH(wp
, &cpu
->watchpoints
, entry
) {
213 cpu_watchpoint_insert(new_cpu
, wp
->vaddr
, wp
->len
, wp
->flags
, NULL
);
219 static void handle_arg_help(const char *arg
)
224 static void handle_arg_log(const char *arg
)
228 mask
= qemu_str_to_log_mask(arg
);
230 qemu_print_log_usage(stdout
);
233 qemu_log_needs_buffers();
237 static void handle_arg_dfilter(const char *arg
)
239 qemu_set_dfilter_ranges(arg
, &error_fatal
);
242 static void handle_arg_log_filename(const char *arg
)
244 qemu_set_log_filename(arg
, &error_fatal
);
247 static void handle_arg_set_env(const char *arg
)
251 while ((token
= strsep(&p
, ",")) != NULL
) {
252 if (envlist_setenv(envlist
, token
) != 0) {
259 static void handle_arg_unset_env(const char *arg
)
263 while ((token
= strsep(&p
, ",")) != NULL
) {
264 if (envlist_unsetenv(envlist
, token
) != 0) {
271 static void handle_arg_argv0(const char *arg
)
276 static void handle_arg_stack_size(const char *arg
)
279 guest_stack_size
= strtoul(arg
, &p
, 0);
280 if (guest_stack_size
== 0) {
285 guest_stack_size
*= MiB
;
286 } else if (*p
== 'k' || *p
== 'K') {
287 guest_stack_size
*= KiB
;
291 static void handle_arg_ld_prefix(const char *arg
)
293 interp_prefix
= strdup(arg
);
296 static void handle_arg_pagesize(const char *arg
)
298 qemu_host_page_size
= atoi(arg
);
299 if (qemu_host_page_size
== 0 ||
300 (qemu_host_page_size
& (qemu_host_page_size
- 1)) != 0) {
301 fprintf(stderr
, "page size must be a power of two\n");
306 static void handle_arg_seed(const char *arg
)
311 static void handle_arg_gdb(const char *arg
)
313 gdbstub_port
= atoi(arg
);
316 static void handle_arg_uname(const char *arg
)
318 qemu_uname_release
= strdup(arg
);
321 static void handle_arg_cpu(const char *arg
)
323 cpu_model
= strdup(arg
);
324 if (cpu_model
== NULL
|| is_help_option(cpu_model
)) {
325 /* XXX: implement xxx_cpu_list for targets that still miss it */
326 #if defined(cpu_list)
333 static void handle_arg_guest_base(const char *arg
)
335 guest_base
= strtol(arg
, NULL
, 0);
339 static void handle_arg_reserved_va(const char *arg
)
343 reserved_va
= strtoul(arg
, &p
, 0);
357 unsigned long unshifted
= reserved_va
;
359 reserved_va
<<= shift
;
360 if (reserved_va
>> shift
!= unshifted
) {
361 fprintf(stderr
, "Reserved virtual address too big\n");
366 fprintf(stderr
, "Unrecognised -R size suffix '%s'\n", p
);
371 static void handle_arg_singlestep(const char *arg
)
376 static void handle_arg_strace(const char *arg
)
381 static void handle_arg_version(const char *arg
)
383 printf("qemu-" TARGET_NAME
" version " QEMU_FULL_VERSION
384 "\n" QEMU_COPYRIGHT
"\n");
388 static char *trace_file
;
389 static void handle_arg_trace(const char *arg
)
392 trace_file
= trace_opt_parse(arg
);
395 #if defined(TARGET_XTENSA)
396 static void handle_arg_abi_call0(const char *arg
)
398 xtensa_set_abi_call0();
402 static QemuPluginList plugins
= QTAILQ_HEAD_INITIALIZER(plugins
);
405 static void handle_arg_plugin(const char *arg
)
407 qemu_plugin_opt_parse(arg
, &plugins
);
411 struct qemu_argument
{
415 void (*handle_opt
)(const char *arg
);
420 static const struct qemu_argument arg_table
[] = {
421 {"h", "", false, handle_arg_help
,
422 "", "print this help"},
423 {"help", "", false, handle_arg_help
,
425 {"g", "QEMU_GDB", true, handle_arg_gdb
,
426 "port", "wait gdb connection to 'port'"},
427 {"L", "QEMU_LD_PREFIX", true, handle_arg_ld_prefix
,
428 "path", "set the elf interpreter prefix to 'path'"},
429 {"s", "QEMU_STACK_SIZE", true, handle_arg_stack_size
,
430 "size", "set the stack size to 'size' bytes"},
431 {"cpu", "QEMU_CPU", true, handle_arg_cpu
,
432 "model", "select CPU (-cpu help for list)"},
433 {"E", "QEMU_SET_ENV", true, handle_arg_set_env
,
434 "var=value", "sets targets environment variable (see below)"},
435 {"U", "QEMU_UNSET_ENV", true, handle_arg_unset_env
,
436 "var", "unsets targets environment variable (see below)"},
437 {"0", "QEMU_ARGV0", true, handle_arg_argv0
,
438 "argv0", "forces target process argv[0] to be 'argv0'"},
439 {"r", "QEMU_UNAME", true, handle_arg_uname
,
440 "uname", "set qemu uname release string to 'uname'"},
441 {"B", "QEMU_GUEST_BASE", true, handle_arg_guest_base
,
442 "address", "set guest_base address to 'address'"},
443 {"R", "QEMU_RESERVED_VA", true, handle_arg_reserved_va
,
444 "size", "reserve 'size' bytes for guest virtual address space"},
445 {"d", "QEMU_LOG", true, handle_arg_log
,
446 "item[,...]", "enable logging of specified items "
447 "(use '-d help' for a list of items)"},
448 {"dfilter", "QEMU_DFILTER", true, handle_arg_dfilter
,
449 "range[,...]","filter logging based on address range"},
450 {"D", "QEMU_LOG_FILENAME", true, handle_arg_log_filename
,
451 "logfile", "write logs to 'logfile' (default stderr)"},
452 {"p", "QEMU_PAGESIZE", true, handle_arg_pagesize
,
453 "pagesize", "set the host page size to 'pagesize'"},
454 {"singlestep", "QEMU_SINGLESTEP", false, handle_arg_singlestep
,
455 "", "run in singlestep mode"},
456 {"strace", "QEMU_STRACE", false, handle_arg_strace
,
457 "", "log system calls"},
458 {"seed", "QEMU_RAND_SEED", true, handle_arg_seed
,
459 "", "Seed for pseudo-random number generator"},
460 {"trace", "QEMU_TRACE", true, handle_arg_trace
,
461 "", "[[enable=]<pattern>][,events=<file>][,file=<file>]"},
463 {"plugin", "QEMU_PLUGIN", true, handle_arg_plugin
,
464 "", "[file=]<file>[,arg=<string>]"},
466 {"version", "QEMU_VERSION", false, handle_arg_version
,
467 "", "display version information and exit"},
468 #if defined(TARGET_XTENSA)
469 {"xtensa-abi-call0", "QEMU_XTENSA_ABI_CALL0", false, handle_arg_abi_call0
,
470 "", "assume CALL0 Xtensa ABI"},
472 {NULL
, NULL
, false, NULL
, NULL
, NULL
}
475 static void usage(int exitcode
)
477 const struct qemu_argument
*arginfo
;
481 printf("usage: qemu-" TARGET_NAME
" [options] program [arguments...]\n"
482 "Linux CPU emulator (compiled for " TARGET_NAME
" emulation)\n"
484 "Options and associated environment variables:\n"
487 /* Calculate column widths. We must always have at least enough space
488 * for the column header.
490 maxarglen
= strlen("Argument");
491 maxenvlen
= strlen("Env-variable");
493 for (arginfo
= arg_table
; arginfo
->handle_opt
!= NULL
; arginfo
++) {
494 int arglen
= strlen(arginfo
->argv
);
495 if (arginfo
->has_arg
) {
496 arglen
+= strlen(arginfo
->example
) + 1;
498 if (strlen(arginfo
->env
) > maxenvlen
) {
499 maxenvlen
= strlen(arginfo
->env
);
501 if (arglen
> maxarglen
) {
506 printf("%-*s %-*s Description\n", maxarglen
+1, "Argument",
507 maxenvlen
, "Env-variable");
509 for (arginfo
= arg_table
; arginfo
->handle_opt
!= NULL
; arginfo
++) {
510 if (arginfo
->has_arg
) {
511 printf("-%s %-*s %-*s %s\n", arginfo
->argv
,
512 (int)(maxarglen
- strlen(arginfo
->argv
) - 1),
513 arginfo
->example
, maxenvlen
, arginfo
->env
, arginfo
->help
);
515 printf("-%-*s %-*s %s\n", maxarglen
, arginfo
->argv
,
516 maxenvlen
, arginfo
->env
,
523 "QEMU_LD_PREFIX = %s\n"
524 "QEMU_STACK_SIZE = %ld byte\n",
529 "You can use -E and -U options or the QEMU_SET_ENV and\n"
530 "QEMU_UNSET_ENV environment variables to set and unset\n"
531 "environment variables for the target process.\n"
532 "It is possible to provide several variables by separating them\n"
533 "by commas in getsubopt(3) style. Additionally it is possible to\n"
534 "provide the -E and -U options multiple times.\n"
535 "The following lines are equivalent:\n"
536 " -E var1=val2 -E var2=val2 -U LD_PRELOAD -U LD_DEBUG\n"
537 " -E var1=val2,var2=val2 -U LD_PRELOAD,LD_DEBUG\n"
538 " QEMU_SET_ENV=var1=val2,var2=val2 QEMU_UNSET_ENV=LD_PRELOAD,LD_DEBUG\n"
539 "Note that if you provide several changes to a single variable\n"
540 "the last change will stay in effect.\n"
542 QEMU_HELP_BOTTOM
"\n");
547 static int parse_args(int argc
, char **argv
)
551 const struct qemu_argument
*arginfo
;
553 for (arginfo
= arg_table
; arginfo
->handle_opt
!= NULL
; arginfo
++) {
554 if (arginfo
->env
== NULL
) {
558 r
= getenv(arginfo
->env
);
560 arginfo
->handle_opt(r
);
566 if (optind
>= argc
) {
575 if (!strcmp(r
, "-")) {
578 /* Treat --foo the same as -foo. */
583 for (arginfo
= arg_table
; arginfo
->handle_opt
!= NULL
; arginfo
++) {
584 if (!strcmp(r
, arginfo
->argv
)) {
585 if (arginfo
->has_arg
) {
586 if (optind
>= argc
) {
587 (void) fprintf(stderr
,
588 "qemu: missing argument for option '%s'\n", r
);
591 arginfo
->handle_opt(argv
[optind
]);
594 arginfo
->handle_opt(NULL
);
600 /* no option matched the current argv */
601 if (arginfo
->handle_opt
== NULL
) {
602 (void) fprintf(stderr
, "qemu: unknown option '%s'\n", r
);
607 if (optind
>= argc
) {
608 (void) fprintf(stderr
, "qemu: no user program specified\n");
612 exec_path
= argv
[optind
];
617 int main(int argc
, char **argv
, char **envp
)
619 struct target_pt_regs regs1
, *regs
= ®s1
;
620 struct image_info info1
, *info
= &info1
;
621 struct linux_binprm bprm
;
626 char **target_environ
, **wrk
;
632 unsigned long max_reserved_va
;
635 module_call_init(MODULE_INIT_TRACE
);
636 qemu_init_cpu_list();
637 module_call_init(MODULE_INIT_QOM
);
639 envlist
= envlist_create();
641 /* add current environment into the list */
642 for (wrk
= environ
; *wrk
!= NULL
; wrk
++) {
643 (void) envlist_setenv(envlist
, *wrk
);
646 /* Read the stack limit from the kernel. If it's "unlimited",
647 then we can do little else besides use the default. */
650 if (getrlimit(RLIMIT_STACK
, &lim
) == 0
651 && lim
.rlim_cur
!= RLIM_INFINITY
652 && lim
.rlim_cur
== (target_long
)lim
.rlim_cur
) {
653 guest_stack_size
= lim
.rlim_cur
;
659 qemu_add_opts(&qemu_trace_opts
);
660 qemu_plugin_add_opts();
662 optind
= parse_args(argc
, argv
);
664 if (!trace_init_backends()) {
667 trace_init_file(trace_file
);
668 if (qemu_plugin_load_list(&plugins
)) {
673 memset(regs
, 0, sizeof(struct target_pt_regs
));
675 /* Zero out image_info */
676 memset(info
, 0, sizeof(struct image_info
));
678 memset(&bprm
, 0, sizeof (bprm
));
680 /* Scan interp_prefix dir for replacement files. */
681 init_paths(interp_prefix
);
683 init_qemu_uname_release();
685 execfd
= qemu_getauxval(AT_EXECFD
);
687 execfd
= open(exec_path
, O_RDONLY
);
689 printf("Error while loading %s: %s\n", exec_path
, strerror(errno
));
694 if (cpu_model
== NULL
) {
695 cpu_model
= cpu_get_model(get_elf_eflags(execfd
));
697 cpu_type
= parse_cpu_option(cpu_model
);
699 /* init tcg before creating CPUs and to get qemu_host_page_size */
702 cpu
= cpu_create(cpu_type
);
708 * Reserving too much vm space via mmap can run into problems
709 * with rlimits, oom due to page table creation, etc. We will
710 * still try it, if directed by the command-line option, but
713 max_reserved_va
= MAX_RESERVED_VA(cpu
);
714 if (reserved_va
!= 0) {
715 if (max_reserved_va
&& reserved_va
> max_reserved_va
) {
716 fprintf(stderr
, "Reserved virtual address too big\n");
719 } else if (HOST_LONG_BITS
== 64 && TARGET_VIRT_ADDR_SPACE_BITS
<= 32) {
721 * reserved_va must be aligned with the host page size
722 * as it is used with mmap()
724 reserved_va
= max_reserved_va
& qemu_host_page_mask
;
729 if (seed_optarg
!= NULL
) {
730 qemu_guest_random_seed_main(seed_optarg
, &err
);
735 error_reportf_err(err
, "cannot initialize crypto: ");
740 target_environ
= envlist_to_environ(envlist
, NULL
);
741 envlist_free(envlist
);
744 * Now that page sizes are configured in tcg_exec_init() we can do
745 * proper page alignment for guest_base.
747 guest_base
= HOST_PAGE_ALIGN(guest_base
);
749 if (reserved_va
|| have_guest_base
) {
750 guest_base
= init_guest_space(guest_base
, reserved_va
, 0,
752 if (guest_base
== (unsigned long)-1) {
753 fprintf(stderr
, "Unable to reserve 0x%lx bytes of virtual address "
754 "space for use as guest address space (check your virtual "
755 "memory ulimit setting or reserve less using -R option)\n",
761 mmap_next_start
= reserved_va
;
766 * Read in mmap_min_addr kernel parameter. This value is used
767 * When loading the ELF image to determine whether guest_base
768 * is needed. It is also used in mmap_find_vma.
773 if ((fp
= fopen("/proc/sys/vm/mmap_min_addr", "r")) != NULL
) {
775 if (fscanf(fp
, "%lu", &tmp
) == 1) {
777 qemu_log_mask(CPU_LOG_PAGE
, "host mmap_min_addr=0x%lx\n", mmap_min_addr
);
784 * Prepare copy of argv vector for target.
786 target_argc
= argc
- optind
;
787 target_argv
= calloc(target_argc
+ 1, sizeof (char *));
788 if (target_argv
== NULL
) {
789 (void) fprintf(stderr
, "Unable to allocate memory for target_argv\n");
794 * If argv0 is specified (using '-0' switch) we replace
795 * argv[0] pointer with the given one.
799 target_argv
[i
++] = strdup(argv0
);
801 for (; i
< target_argc
; i
++) {
802 target_argv
[i
] = strdup(argv
[optind
+ i
]);
804 target_argv
[target_argc
] = NULL
;
806 ts
= g_new0(TaskState
, 1);
808 /* build Task State */
814 ret
= loader_exec(execfd
, exec_path
, target_argv
, target_environ
, regs
,
817 printf("Error while loading %s: %s\n", exec_path
, strerror(-ret
));
821 for (wrk
= target_environ
; *wrk
; wrk
++) {
825 g_free(target_environ
);
827 if (qemu_loglevel_mask(CPU_LOG_PAGE
)) {
828 qemu_log("guest_base 0x%lx\n", guest_base
);
831 qemu_log("start_brk 0x" TARGET_ABI_FMT_lx
"\n", info
->start_brk
);
832 qemu_log("end_code 0x" TARGET_ABI_FMT_lx
"\n", info
->end_code
);
833 qemu_log("start_code 0x" TARGET_ABI_FMT_lx
"\n", info
->start_code
);
834 qemu_log("start_data 0x" TARGET_ABI_FMT_lx
"\n", info
->start_data
);
835 qemu_log("end_data 0x" TARGET_ABI_FMT_lx
"\n", info
->end_data
);
836 qemu_log("start_stack 0x" TARGET_ABI_FMT_lx
"\n", info
->start_stack
);
837 qemu_log("brk 0x" TARGET_ABI_FMT_lx
"\n", info
->brk
);
838 qemu_log("entry 0x" TARGET_ABI_FMT_lx
"\n", info
->entry
);
839 qemu_log("argv_start 0x" TARGET_ABI_FMT_lx
"\n", info
->arg_start
);
840 qemu_log("env_start 0x" TARGET_ABI_FMT_lx
"\n",
841 info
->arg_end
+ (abi_ulong
)sizeof(abi_ulong
));
842 qemu_log("auxv_start 0x" TARGET_ABI_FMT_lx
"\n", info
->saved_auxv
);
845 target_set_brk(info
->brk
);
849 /* Now that we've loaded the binary, GUEST_BASE is fixed. Delay
850 generating the prologue until now so that the prologue can take
851 the real value of GUEST_BASE into account. */
852 tcg_prologue_init(tcg_ctx
);
855 target_cpu_copy_regs(env
, regs
);
858 if (gdbserver_start(gdbstub_port
) < 0) {
859 fprintf(stderr
, "qemu: could not open gdbserver on port %d\n",
863 gdb_handlesig(cpu
, 0);