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/config-file.h"
32 #include "qemu/cutils.h"
33 #include "qemu/error-report.h"
34 #include "qemu/help_option.h"
35 #include "qemu/module.h"
37 #include "exec/exec-all.h"
39 #include "qemu/timer.h"
40 #include "qemu/envlist.h"
41 #include "qemu/guest-random.h"
43 #include "trace/control.h"
44 #include "target_elf.h"
45 #include "cpu_loop-common.h"
46 #include "crypto/init.h"
51 static const char *filename
;
52 static const char *argv0
;
53 static int gdbstub_port
;
54 static envlist_t
*envlist
;
55 static const char *cpu_model
;
56 static const char *cpu_type
;
57 static const char *seed_optarg
;
58 unsigned long mmap_min_addr
;
59 unsigned long guest_base
;
63 * When running 32-on-64 we should make sure we can fit all of the possible
64 * guest address space into a contiguous chunk of virtual host memory.
66 * This way we will never overlap with our own libraries or binaries or stack
67 * or anything else that QEMU maps.
69 * Many cpus reserve the high bit (or more than one for some 64-bit cpus)
70 * of the address for the kernel. Some cpus rely on this and user space
71 * uses the high bit(s) for pointer tagging and the like. For them, we
72 * must preserve the expected address space.
74 #ifndef MAX_RESERVED_VA
75 # if HOST_LONG_BITS > TARGET_VIRT_ADDR_SPACE_BITS
76 # if TARGET_VIRT_ADDR_SPACE_BITS == 32 && \
77 (TARGET_LONG_BITS == 32 || defined(TARGET_ABI32))
78 /* There are a number of places where we assign reserved_va to a variable
79 of type abi_ulong and expect it to fit. Avoid the last page. */
80 # define MAX_RESERVED_VA (0xfffffffful & TARGET_PAGE_MASK)
82 # define MAX_RESERVED_VA (1ul << TARGET_VIRT_ADDR_SPACE_BITS)
85 # define MAX_RESERVED_VA 0
89 unsigned long reserved_va
;
91 static void usage(int exitcode
);
93 static const char *interp_prefix
= CONFIG_QEMU_INTERP_PREFIX
;
94 const char *qemu_uname_release
;
96 /* XXX: on x86 MAP_GROWSDOWN only works if ESP <= address + 32, so
97 we allocate a bigger stack. Need a better solution, for example
98 by remapping the process stack directly at the right place */
99 unsigned long guest_stack_size
= 8 * 1024 * 1024UL;
101 void gemu_log(const char *fmt
, ...)
106 vfprintf(stderr
, fmt
, ap
);
110 #if defined(TARGET_I386)
111 int cpu_get_pic_interrupt(CPUX86State
*env
)
117 /***********************************************************/
118 /* Helper routines for implementing atomic operations. */
120 /* Make sure everything is in a consistent state for calling fork(). */
121 void fork_start(void)
128 void fork_end(int child
)
130 mmap_fork_end(child
);
132 CPUState
*cpu
, *next_cpu
;
133 /* Child processes created by fork() only have a single thread.
134 Discard information about the parent threads. */
135 CPU_FOREACH_SAFE(cpu
, next_cpu
) {
136 if (cpu
!= thread_cpu
) {
137 QTAILQ_REMOVE_RCU(&cpus
, cpu
, node
);
140 qemu_init_cpu_list();
141 gdbserver_fork(thread_cpu
);
142 /* qemu_init_cpu_list() takes care of reinitializing the
143 * exclusive state, so we don't need to end_exclusive() here.
151 __thread CPUState
*thread_cpu
;
153 bool qemu_cpu_is_self(CPUState
*cpu
)
155 return thread_cpu
== cpu
;
158 void qemu_cpu_kick(CPUState
*cpu
)
163 void task_settid(TaskState
*ts
)
165 if (ts
->ts_tid
== 0) {
166 ts
->ts_tid
= (pid_t
)syscall(SYS_gettid
);
170 void stop_all_tasks(void)
173 * We trust that when using NPTL, start_exclusive()
174 * handles thread stopping correctly.
179 /* Assumes contents are already zeroed. */
180 void init_task_state(TaskState
*ts
)
185 CPUArchState
*cpu_copy(CPUArchState
*env
)
187 CPUState
*cpu
= env_cpu(env
);
188 CPUState
*new_cpu
= cpu_create(cpu_type
);
189 CPUArchState
*new_env
= new_cpu
->env_ptr
;
193 /* Reset non arch specific state */
196 memcpy(new_env
, env
, sizeof(CPUArchState
));
198 /* Clone all break/watchpoints.
199 Note: Once we support ptrace with hw-debug register access, make sure
200 BP_CPU break/watchpoints are handled correctly on clone. */
201 QTAILQ_INIT(&new_cpu
->breakpoints
);
202 QTAILQ_INIT(&new_cpu
->watchpoints
);
203 QTAILQ_FOREACH(bp
, &cpu
->breakpoints
, entry
) {
204 cpu_breakpoint_insert(new_cpu
, bp
->pc
, bp
->flags
, NULL
);
206 QTAILQ_FOREACH(wp
, &cpu
->watchpoints
, entry
) {
207 cpu_watchpoint_insert(new_cpu
, wp
->vaddr
, wp
->len
, wp
->flags
, NULL
);
213 static void handle_arg_help(const char *arg
)
218 static void handle_arg_log(const char *arg
)
222 mask
= qemu_str_to_log_mask(arg
);
224 qemu_print_log_usage(stdout
);
227 qemu_log_needs_buffers();
231 static void handle_arg_dfilter(const char *arg
)
233 qemu_set_dfilter_ranges(arg
, NULL
);
236 static void handle_arg_log_filename(const char *arg
)
238 qemu_set_log_filename(arg
, &error_fatal
);
241 static void handle_arg_set_env(const char *arg
)
245 while ((token
= strsep(&p
, ",")) != NULL
) {
246 if (envlist_setenv(envlist
, token
) != 0) {
253 static void handle_arg_unset_env(const char *arg
)
257 while ((token
= strsep(&p
, ",")) != NULL
) {
258 if (envlist_unsetenv(envlist
, token
) != 0) {
265 static void handle_arg_argv0(const char *arg
)
270 static void handle_arg_stack_size(const char *arg
)
273 guest_stack_size
= strtoul(arg
, &p
, 0);
274 if (guest_stack_size
== 0) {
279 guest_stack_size
*= MiB
;
280 } else if (*p
== 'k' || *p
== 'K') {
281 guest_stack_size
*= KiB
;
285 static void handle_arg_ld_prefix(const char *arg
)
287 interp_prefix
= strdup(arg
);
290 static void handle_arg_pagesize(const char *arg
)
292 qemu_host_page_size
= atoi(arg
);
293 if (qemu_host_page_size
== 0 ||
294 (qemu_host_page_size
& (qemu_host_page_size
- 1)) != 0) {
295 fprintf(stderr
, "page size must be a power of two\n");
300 static void handle_arg_seed(const char *arg
)
305 static void handle_arg_gdb(const char *arg
)
307 gdbstub_port
= atoi(arg
);
310 static void handle_arg_uname(const char *arg
)
312 qemu_uname_release
= strdup(arg
);
315 static void handle_arg_cpu(const char *arg
)
317 cpu_model
= strdup(arg
);
318 if (cpu_model
== NULL
|| is_help_option(cpu_model
)) {
319 /* XXX: implement xxx_cpu_list for targets that still miss it */
320 #if defined(cpu_list)
327 static void handle_arg_guest_base(const char *arg
)
329 guest_base
= strtol(arg
, NULL
, 0);
333 static void handle_arg_reserved_va(const char *arg
)
337 reserved_va
= strtoul(arg
, &p
, 0);
351 unsigned long unshifted
= reserved_va
;
353 reserved_va
<<= shift
;
354 if (reserved_va
>> shift
!= unshifted
355 || (MAX_RESERVED_VA
&& reserved_va
> MAX_RESERVED_VA
)) {
356 fprintf(stderr
, "Reserved virtual address too big\n");
361 fprintf(stderr
, "Unrecognised -R size suffix '%s'\n", p
);
366 static void handle_arg_singlestep(const char *arg
)
371 static void handle_arg_strace(const char *arg
)
376 static void handle_arg_version(const char *arg
)
378 printf("qemu-" TARGET_NAME
" version " QEMU_FULL_VERSION
379 "\n" QEMU_COPYRIGHT
"\n");
383 static char *trace_file
;
384 static void handle_arg_trace(const char *arg
)
387 trace_file
= trace_opt_parse(arg
);
390 struct qemu_argument
{
394 void (*handle_opt
)(const char *arg
);
399 static const struct qemu_argument arg_table
[] = {
400 {"h", "", false, handle_arg_help
,
401 "", "print this help"},
402 {"help", "", false, handle_arg_help
,
404 {"g", "QEMU_GDB", true, handle_arg_gdb
,
405 "port", "wait gdb connection to 'port'"},
406 {"L", "QEMU_LD_PREFIX", true, handle_arg_ld_prefix
,
407 "path", "set the elf interpreter prefix to 'path'"},
408 {"s", "QEMU_STACK_SIZE", true, handle_arg_stack_size
,
409 "size", "set the stack size to 'size' bytes"},
410 {"cpu", "QEMU_CPU", true, handle_arg_cpu
,
411 "model", "select CPU (-cpu help for list)"},
412 {"E", "QEMU_SET_ENV", true, handle_arg_set_env
,
413 "var=value", "sets targets environment variable (see below)"},
414 {"U", "QEMU_UNSET_ENV", true, handle_arg_unset_env
,
415 "var", "unsets targets environment variable (see below)"},
416 {"0", "QEMU_ARGV0", true, handle_arg_argv0
,
417 "argv0", "forces target process argv[0] to be 'argv0'"},
418 {"r", "QEMU_UNAME", true, handle_arg_uname
,
419 "uname", "set qemu uname release string to 'uname'"},
420 {"B", "QEMU_GUEST_BASE", true, handle_arg_guest_base
,
421 "address", "set guest_base address to 'address'"},
422 {"R", "QEMU_RESERVED_VA", true, handle_arg_reserved_va
,
423 "size", "reserve 'size' bytes for guest virtual address space"},
424 {"d", "QEMU_LOG", true, handle_arg_log
,
425 "item[,...]", "enable logging of specified items "
426 "(use '-d help' for a list of items)"},
427 {"dfilter", "QEMU_DFILTER", true, handle_arg_dfilter
,
428 "range[,...]","filter logging based on address range"},
429 {"D", "QEMU_LOG_FILENAME", true, handle_arg_log_filename
,
430 "logfile", "write logs to 'logfile' (default stderr)"},
431 {"p", "QEMU_PAGESIZE", true, handle_arg_pagesize
,
432 "pagesize", "set the host page size to 'pagesize'"},
433 {"singlestep", "QEMU_SINGLESTEP", false, handle_arg_singlestep
,
434 "", "run in singlestep mode"},
435 {"strace", "QEMU_STRACE", false, handle_arg_strace
,
436 "", "log system calls"},
437 {"seed", "QEMU_RAND_SEED", true, handle_arg_seed
,
438 "", "Seed for pseudo-random number generator"},
439 {"trace", "QEMU_TRACE", true, handle_arg_trace
,
440 "", "[[enable=]<pattern>][,events=<file>][,file=<file>]"},
441 {"version", "QEMU_VERSION", false, handle_arg_version
,
442 "", "display version information and exit"},
443 {NULL
, NULL
, false, NULL
, NULL
, NULL
}
446 static void usage(int exitcode
)
448 const struct qemu_argument
*arginfo
;
452 printf("usage: qemu-" TARGET_NAME
" [options] program [arguments...]\n"
453 "Linux CPU emulator (compiled for " TARGET_NAME
" emulation)\n"
455 "Options and associated environment variables:\n"
458 /* Calculate column widths. We must always have at least enough space
459 * for the column header.
461 maxarglen
= strlen("Argument");
462 maxenvlen
= strlen("Env-variable");
464 for (arginfo
= arg_table
; arginfo
->handle_opt
!= NULL
; arginfo
++) {
465 int arglen
= strlen(arginfo
->argv
);
466 if (arginfo
->has_arg
) {
467 arglen
+= strlen(arginfo
->example
) + 1;
469 if (strlen(arginfo
->env
) > maxenvlen
) {
470 maxenvlen
= strlen(arginfo
->env
);
472 if (arglen
> maxarglen
) {
477 printf("%-*s %-*s Description\n", maxarglen
+1, "Argument",
478 maxenvlen
, "Env-variable");
480 for (arginfo
= arg_table
; arginfo
->handle_opt
!= NULL
; arginfo
++) {
481 if (arginfo
->has_arg
) {
482 printf("-%s %-*s %-*s %s\n", arginfo
->argv
,
483 (int)(maxarglen
- strlen(arginfo
->argv
) - 1),
484 arginfo
->example
, maxenvlen
, arginfo
->env
, arginfo
->help
);
486 printf("-%-*s %-*s %s\n", maxarglen
, arginfo
->argv
,
487 maxenvlen
, arginfo
->env
,
494 "QEMU_LD_PREFIX = %s\n"
495 "QEMU_STACK_SIZE = %ld byte\n",
500 "You can use -E and -U options or the QEMU_SET_ENV and\n"
501 "QEMU_UNSET_ENV environment variables to set and unset\n"
502 "environment variables for the target process.\n"
503 "It is possible to provide several variables by separating them\n"
504 "by commas in getsubopt(3) style. Additionally it is possible to\n"
505 "provide the -E and -U options multiple times.\n"
506 "The following lines are equivalent:\n"
507 " -E var1=val2 -E var2=val2 -U LD_PRELOAD -U LD_DEBUG\n"
508 " -E var1=val2,var2=val2 -U LD_PRELOAD,LD_DEBUG\n"
509 " QEMU_SET_ENV=var1=val2,var2=val2 QEMU_UNSET_ENV=LD_PRELOAD,LD_DEBUG\n"
510 "Note that if you provide several changes to a single variable\n"
511 "the last change will stay in effect.\n"
513 QEMU_HELP_BOTTOM
"\n");
518 static int parse_args(int argc
, char **argv
)
522 const struct qemu_argument
*arginfo
;
524 for (arginfo
= arg_table
; arginfo
->handle_opt
!= NULL
; arginfo
++) {
525 if (arginfo
->env
== NULL
) {
529 r
= getenv(arginfo
->env
);
531 arginfo
->handle_opt(r
);
537 if (optind
>= argc
) {
546 if (!strcmp(r
, "-")) {
549 /* Treat --foo the same as -foo. */
554 for (arginfo
= arg_table
; arginfo
->handle_opt
!= NULL
; arginfo
++) {
555 if (!strcmp(r
, arginfo
->argv
)) {
556 if (arginfo
->has_arg
) {
557 if (optind
>= argc
) {
558 (void) fprintf(stderr
,
559 "qemu: missing argument for option '%s'\n", r
);
562 arginfo
->handle_opt(argv
[optind
]);
565 arginfo
->handle_opt(NULL
);
571 /* no option matched the current argv */
572 if (arginfo
->handle_opt
== NULL
) {
573 (void) fprintf(stderr
, "qemu: unknown option '%s'\n", r
);
578 if (optind
>= argc
) {
579 (void) fprintf(stderr
, "qemu: no user program specified\n");
583 filename
= argv
[optind
];
584 exec_path
= argv
[optind
];
589 int main(int argc
, char **argv
, char **envp
)
591 struct target_pt_regs regs1
, *regs
= ®s1
;
592 struct image_info info1
, *info
= &info1
;
593 struct linux_binprm bprm
;
598 char **target_environ
, **wrk
;
606 module_call_init(MODULE_INIT_TRACE
);
607 qemu_init_cpu_list();
608 module_call_init(MODULE_INIT_QOM
);
610 envlist
= envlist_create();
612 /* add current environment into the list */
613 for (wrk
= environ
; *wrk
!= NULL
; wrk
++) {
614 (void) envlist_setenv(envlist
, *wrk
);
617 /* Read the stack limit from the kernel. If it's "unlimited",
618 then we can do little else besides use the default. */
621 if (getrlimit(RLIMIT_STACK
, &lim
) == 0
622 && lim
.rlim_cur
!= RLIM_INFINITY
623 && lim
.rlim_cur
== (target_long
)lim
.rlim_cur
) {
624 guest_stack_size
= lim
.rlim_cur
;
630 qemu_add_opts(&qemu_trace_opts
);
632 optind
= parse_args(argc
, argv
);
634 if (!trace_init_backends()) {
637 trace_init_file(trace_file
);
640 memset(regs
, 0, sizeof(struct target_pt_regs
));
642 /* Zero out image_info */
643 memset(info
, 0, sizeof(struct image_info
));
645 memset(&bprm
, 0, sizeof (bprm
));
647 /* Scan interp_prefix dir for replacement files. */
648 init_paths(interp_prefix
);
650 init_qemu_uname_release();
652 execfd
= qemu_getauxval(AT_EXECFD
);
654 execfd
= open(filename
, O_RDONLY
);
656 printf("Error while loading %s: %s\n", filename
, strerror(errno
));
661 if (cpu_model
== NULL
) {
662 cpu_model
= cpu_get_model(get_elf_eflags(execfd
));
664 cpu_type
= parse_cpu_option(cpu_model
);
666 /* init tcg before creating CPUs and to get qemu_host_page_size */
669 /* Reserving *too* much vm space via mmap can run into problems
670 with rlimits, oom due to page table creation, etc. We will still try it,
671 if directed by the command-line option, but not by default. */
672 if (HOST_LONG_BITS
== 64 &&
673 TARGET_VIRT_ADDR_SPACE_BITS
<= 32 &&
675 /* reserved_va must be aligned with the host page size
676 * as it is used with mmap()
678 reserved_va
= MAX_RESERVED_VA
& qemu_host_page_mask
;
681 cpu
= cpu_create(cpu_type
);
687 if (getenv("QEMU_STRACE")) {
691 if (seed_optarg
== NULL
) {
692 seed_optarg
= getenv("QEMU_RAND_SEED");
696 if (seed_optarg
!= NULL
) {
697 qemu_guest_random_seed_main(seed_optarg
, &err
);
702 error_reportf_err(err
, "cannot initialize crypto: ");
707 target_environ
= envlist_to_environ(envlist
, NULL
);
708 envlist_free(envlist
);
711 * Now that page sizes are configured in tcg_exec_init() we can do
712 * proper page alignment for guest_base.
714 guest_base
= HOST_PAGE_ALIGN(guest_base
);
716 if (reserved_va
|| have_guest_base
) {
717 guest_base
= init_guest_space(guest_base
, reserved_va
, 0,
719 if (guest_base
== (unsigned long)-1) {
720 fprintf(stderr
, "Unable to reserve 0x%lx bytes of virtual address "
721 "space for use as guest address space (check your virtual "
722 "memory ulimit setting or reserve less using -R option)\n",
728 mmap_next_start
= reserved_va
;
733 * Read in mmap_min_addr kernel parameter. This value is used
734 * When loading the ELF image to determine whether guest_base
735 * is needed. It is also used in mmap_find_vma.
740 if ((fp
= fopen("/proc/sys/vm/mmap_min_addr", "r")) != NULL
) {
742 if (fscanf(fp
, "%lu", &tmp
) == 1) {
744 qemu_log_mask(CPU_LOG_PAGE
, "host mmap_min_addr=0x%lx\n", mmap_min_addr
);
751 * Prepare copy of argv vector for target.
753 target_argc
= argc
- optind
;
754 target_argv
= calloc(target_argc
+ 1, sizeof (char *));
755 if (target_argv
== NULL
) {
756 (void) fprintf(stderr
, "Unable to allocate memory for target_argv\n");
761 * If argv0 is specified (using '-0' switch) we replace
762 * argv[0] pointer with the given one.
766 target_argv
[i
++] = strdup(argv0
);
768 for (; i
< target_argc
; i
++) {
769 target_argv
[i
] = strdup(argv
[optind
+ i
]);
771 target_argv
[target_argc
] = NULL
;
773 ts
= g_new0(TaskState
, 1);
775 /* build Task State */
781 ret
= loader_exec(execfd
, filename
, target_argv
, target_environ
, regs
,
784 printf("Error while loading %s: %s\n", filename
, strerror(-ret
));
788 for (wrk
= target_environ
; *wrk
; wrk
++) {
792 g_free(target_environ
);
794 if (qemu_loglevel_mask(CPU_LOG_PAGE
)) {
795 qemu_log("guest_base 0x%lx\n", guest_base
);
798 qemu_log("start_brk 0x" TARGET_ABI_FMT_lx
"\n", info
->start_brk
);
799 qemu_log("end_code 0x" TARGET_ABI_FMT_lx
"\n", info
->end_code
);
800 qemu_log("start_code 0x" TARGET_ABI_FMT_lx
"\n", info
->start_code
);
801 qemu_log("start_data 0x" TARGET_ABI_FMT_lx
"\n", info
->start_data
);
802 qemu_log("end_data 0x" TARGET_ABI_FMT_lx
"\n", info
->end_data
);
803 qemu_log("start_stack 0x" TARGET_ABI_FMT_lx
"\n", info
->start_stack
);
804 qemu_log("brk 0x" TARGET_ABI_FMT_lx
"\n", info
->brk
);
805 qemu_log("entry 0x" TARGET_ABI_FMT_lx
"\n", info
->entry
);
806 qemu_log("argv_start 0x" TARGET_ABI_FMT_lx
"\n", info
->arg_start
);
807 qemu_log("env_start 0x" TARGET_ABI_FMT_lx
"\n",
808 info
->arg_end
+ (abi_ulong
)sizeof(abi_ulong
));
809 qemu_log("auxv_start 0x" TARGET_ABI_FMT_lx
"\n", info
->saved_auxv
);
812 target_set_brk(info
->brk
);
816 /* Now that we've loaded the binary, GUEST_BASE is fixed. Delay
817 generating the prologue until now so that the prologue can take
818 the real value of GUEST_BASE into account. */
819 tcg_prologue_init(tcg_ctx
);
822 target_cpu_copy_regs(env
, regs
);
825 if (gdbserver_start(gdbstub_port
) < 0) {
826 fprintf(stderr
, "qemu: could not open gdbserver on port %d\n",
830 gdb_handlesig(cpu
, 0);