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"
38 #include "exec/exec-all.h"
40 #include "qemu/timer.h"
41 #include "qemu/envlist.h"
42 #include "qemu/guest-random.h"
44 #include "trace/control.h"
45 #include "target_elf.h"
46 #include "cpu_loop-common.h"
47 #include "crypto/init.h"
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(CPU) (0xfffffffful & TARGET_PAGE_MASK)
82 # define MAX_RESERVED_VA(CPU) (1ul << TARGET_VIRT_ADDR_SPACE_BITS)
85 # define MAX_RESERVED_VA(CPU) 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
)
183 ts
->sigaltstack_used
= (struct target_sigaltstack
) {
186 .ss_flags
= TARGET_SS_DISABLE
,
190 CPUArchState
*cpu_copy(CPUArchState
*env
)
192 CPUState
*cpu
= env_cpu(env
);
193 CPUState
*new_cpu
= cpu_create(cpu_type
);
194 CPUArchState
*new_env
= new_cpu
->env_ptr
;
198 /* Reset non arch specific state */
201 memcpy(new_env
, env
, sizeof(CPUArchState
));
203 /* Clone all break/watchpoints.
204 Note: Once we support ptrace with hw-debug register access, make sure
205 BP_CPU break/watchpoints are handled correctly on clone. */
206 QTAILQ_INIT(&new_cpu
->breakpoints
);
207 QTAILQ_INIT(&new_cpu
->watchpoints
);
208 QTAILQ_FOREACH(bp
, &cpu
->breakpoints
, entry
) {
209 cpu_breakpoint_insert(new_cpu
, bp
->pc
, bp
->flags
, NULL
);
211 QTAILQ_FOREACH(wp
, &cpu
->watchpoints
, entry
) {
212 cpu_watchpoint_insert(new_cpu
, wp
->vaddr
, wp
->len
, wp
->flags
, NULL
);
218 static void handle_arg_help(const char *arg
)
223 static void handle_arg_log(const char *arg
)
227 mask
= qemu_str_to_log_mask(arg
);
229 qemu_print_log_usage(stdout
);
232 qemu_log_needs_buffers();
236 static void handle_arg_dfilter(const char *arg
)
238 qemu_set_dfilter_ranges(arg
, &error_fatal
);
241 static void handle_arg_log_filename(const char *arg
)
243 qemu_set_log_filename(arg
, &error_fatal
);
246 static void handle_arg_set_env(const char *arg
)
250 while ((token
= strsep(&p
, ",")) != NULL
) {
251 if (envlist_setenv(envlist
, token
) != 0) {
258 static void handle_arg_unset_env(const char *arg
)
262 while ((token
= strsep(&p
, ",")) != NULL
) {
263 if (envlist_unsetenv(envlist
, token
) != 0) {
270 static void handle_arg_argv0(const char *arg
)
275 static void handle_arg_stack_size(const char *arg
)
278 guest_stack_size
= strtoul(arg
, &p
, 0);
279 if (guest_stack_size
== 0) {
284 guest_stack_size
*= MiB
;
285 } else if (*p
== 'k' || *p
== 'K') {
286 guest_stack_size
*= KiB
;
290 static void handle_arg_ld_prefix(const char *arg
)
292 interp_prefix
= strdup(arg
);
295 static void handle_arg_pagesize(const char *arg
)
297 qemu_host_page_size
= atoi(arg
);
298 if (qemu_host_page_size
== 0 ||
299 (qemu_host_page_size
& (qemu_host_page_size
- 1)) != 0) {
300 fprintf(stderr
, "page size must be a power of two\n");
305 static void handle_arg_seed(const char *arg
)
310 static void handle_arg_gdb(const char *arg
)
312 gdbstub_port
= atoi(arg
);
315 static void handle_arg_uname(const char *arg
)
317 qemu_uname_release
= strdup(arg
);
320 static void handle_arg_cpu(const char *arg
)
322 cpu_model
= strdup(arg
);
323 if (cpu_model
== NULL
|| is_help_option(cpu_model
)) {
324 /* XXX: implement xxx_cpu_list for targets that still miss it */
325 #if defined(cpu_list)
332 static void handle_arg_guest_base(const char *arg
)
334 guest_base
= strtol(arg
, NULL
, 0);
338 static void handle_arg_reserved_va(const char *arg
)
342 reserved_va
= strtoul(arg
, &p
, 0);
356 unsigned long unshifted
= reserved_va
;
358 reserved_va
<<= shift
;
359 if (reserved_va
>> shift
!= unshifted
) {
360 fprintf(stderr
, "Reserved virtual address too big\n");
365 fprintf(stderr
, "Unrecognised -R size suffix '%s'\n", p
);
370 static void handle_arg_singlestep(const char *arg
)
375 static void handle_arg_strace(const char *arg
)
380 static void handle_arg_version(const char *arg
)
382 printf("qemu-" TARGET_NAME
" version " QEMU_FULL_VERSION
383 "\n" QEMU_COPYRIGHT
"\n");
387 static char *trace_file
;
388 static void handle_arg_trace(const char *arg
)
391 trace_file
= trace_opt_parse(arg
);
394 struct qemu_argument
{
398 void (*handle_opt
)(const char *arg
);
403 static const struct qemu_argument arg_table
[] = {
404 {"h", "", false, handle_arg_help
,
405 "", "print this help"},
406 {"help", "", false, handle_arg_help
,
408 {"g", "QEMU_GDB", true, handle_arg_gdb
,
409 "port", "wait gdb connection to 'port'"},
410 {"L", "QEMU_LD_PREFIX", true, handle_arg_ld_prefix
,
411 "path", "set the elf interpreter prefix to 'path'"},
412 {"s", "QEMU_STACK_SIZE", true, handle_arg_stack_size
,
413 "size", "set the stack size to 'size' bytes"},
414 {"cpu", "QEMU_CPU", true, handle_arg_cpu
,
415 "model", "select CPU (-cpu help for list)"},
416 {"E", "QEMU_SET_ENV", true, handle_arg_set_env
,
417 "var=value", "sets targets environment variable (see below)"},
418 {"U", "QEMU_UNSET_ENV", true, handle_arg_unset_env
,
419 "var", "unsets targets environment variable (see below)"},
420 {"0", "QEMU_ARGV0", true, handle_arg_argv0
,
421 "argv0", "forces target process argv[0] to be 'argv0'"},
422 {"r", "QEMU_UNAME", true, handle_arg_uname
,
423 "uname", "set qemu uname release string to 'uname'"},
424 {"B", "QEMU_GUEST_BASE", true, handle_arg_guest_base
,
425 "address", "set guest_base address to 'address'"},
426 {"R", "QEMU_RESERVED_VA", true, handle_arg_reserved_va
,
427 "size", "reserve 'size' bytes for guest virtual address space"},
428 {"d", "QEMU_LOG", true, handle_arg_log
,
429 "item[,...]", "enable logging of specified items "
430 "(use '-d help' for a list of items)"},
431 {"dfilter", "QEMU_DFILTER", true, handle_arg_dfilter
,
432 "range[,...]","filter logging based on address range"},
433 {"D", "QEMU_LOG_FILENAME", true, handle_arg_log_filename
,
434 "logfile", "write logs to 'logfile' (default stderr)"},
435 {"p", "QEMU_PAGESIZE", true, handle_arg_pagesize
,
436 "pagesize", "set the host page size to 'pagesize'"},
437 {"singlestep", "QEMU_SINGLESTEP", false, handle_arg_singlestep
,
438 "", "run in singlestep mode"},
439 {"strace", "QEMU_STRACE", false, handle_arg_strace
,
440 "", "log system calls"},
441 {"seed", "QEMU_RAND_SEED", true, handle_arg_seed
,
442 "", "Seed for pseudo-random number generator"},
443 {"trace", "QEMU_TRACE", true, handle_arg_trace
,
444 "", "[[enable=]<pattern>][,events=<file>][,file=<file>]"},
445 {"version", "QEMU_VERSION", false, handle_arg_version
,
446 "", "display version information and exit"},
447 {NULL
, NULL
, false, NULL
, NULL
, NULL
}
450 static void usage(int exitcode
)
452 const struct qemu_argument
*arginfo
;
456 printf("usage: qemu-" TARGET_NAME
" [options] program [arguments...]\n"
457 "Linux CPU emulator (compiled for " TARGET_NAME
" emulation)\n"
459 "Options and associated environment variables:\n"
462 /* Calculate column widths. We must always have at least enough space
463 * for the column header.
465 maxarglen
= strlen("Argument");
466 maxenvlen
= strlen("Env-variable");
468 for (arginfo
= arg_table
; arginfo
->handle_opt
!= NULL
; arginfo
++) {
469 int arglen
= strlen(arginfo
->argv
);
470 if (arginfo
->has_arg
) {
471 arglen
+= strlen(arginfo
->example
) + 1;
473 if (strlen(arginfo
->env
) > maxenvlen
) {
474 maxenvlen
= strlen(arginfo
->env
);
476 if (arglen
> maxarglen
) {
481 printf("%-*s %-*s Description\n", maxarglen
+1, "Argument",
482 maxenvlen
, "Env-variable");
484 for (arginfo
= arg_table
; arginfo
->handle_opt
!= NULL
; arginfo
++) {
485 if (arginfo
->has_arg
) {
486 printf("-%s %-*s %-*s %s\n", arginfo
->argv
,
487 (int)(maxarglen
- strlen(arginfo
->argv
) - 1),
488 arginfo
->example
, maxenvlen
, arginfo
->env
, arginfo
->help
);
490 printf("-%-*s %-*s %s\n", maxarglen
, arginfo
->argv
,
491 maxenvlen
, arginfo
->env
,
498 "QEMU_LD_PREFIX = %s\n"
499 "QEMU_STACK_SIZE = %ld byte\n",
504 "You can use -E and -U options or the QEMU_SET_ENV and\n"
505 "QEMU_UNSET_ENV environment variables to set and unset\n"
506 "environment variables for the target process.\n"
507 "It is possible to provide several variables by separating them\n"
508 "by commas in getsubopt(3) style. Additionally it is possible to\n"
509 "provide the -E and -U options multiple times.\n"
510 "The following lines are equivalent:\n"
511 " -E var1=val2 -E var2=val2 -U LD_PRELOAD -U LD_DEBUG\n"
512 " -E var1=val2,var2=val2 -U LD_PRELOAD,LD_DEBUG\n"
513 " QEMU_SET_ENV=var1=val2,var2=val2 QEMU_UNSET_ENV=LD_PRELOAD,LD_DEBUG\n"
514 "Note that if you provide several changes to a single variable\n"
515 "the last change will stay in effect.\n"
517 QEMU_HELP_BOTTOM
"\n");
522 static int parse_args(int argc
, char **argv
)
526 const struct qemu_argument
*arginfo
;
528 for (arginfo
= arg_table
; arginfo
->handle_opt
!= NULL
; arginfo
++) {
529 if (arginfo
->env
== NULL
) {
533 r
= getenv(arginfo
->env
);
535 arginfo
->handle_opt(r
);
541 if (optind
>= argc
) {
550 if (!strcmp(r
, "-")) {
553 /* Treat --foo the same as -foo. */
558 for (arginfo
= arg_table
; arginfo
->handle_opt
!= NULL
; arginfo
++) {
559 if (!strcmp(r
, arginfo
->argv
)) {
560 if (arginfo
->has_arg
) {
561 if (optind
>= argc
) {
562 (void) fprintf(stderr
,
563 "qemu: missing argument for option '%s'\n", r
);
566 arginfo
->handle_opt(argv
[optind
]);
569 arginfo
->handle_opt(NULL
);
575 /* no option matched the current argv */
576 if (arginfo
->handle_opt
== NULL
) {
577 (void) fprintf(stderr
, "qemu: unknown option '%s'\n", r
);
582 if (optind
>= argc
) {
583 (void) fprintf(stderr
, "qemu: no user program specified\n");
587 exec_path
= argv
[optind
];
592 int main(int argc
, char **argv
, char **envp
)
594 struct target_pt_regs regs1
, *regs
= ®s1
;
595 struct image_info info1
, *info
= &info1
;
596 struct linux_binprm bprm
;
601 char **target_environ
, **wrk
;
607 unsigned long max_reserved_va
;
610 module_call_init(MODULE_INIT_TRACE
);
611 qemu_init_cpu_list();
612 module_call_init(MODULE_INIT_QOM
);
614 envlist
= envlist_create();
616 /* add current environment into the list */
617 for (wrk
= environ
; *wrk
!= NULL
; wrk
++) {
618 (void) envlist_setenv(envlist
, *wrk
);
621 /* Read the stack limit from the kernel. If it's "unlimited",
622 then we can do little else besides use the default. */
625 if (getrlimit(RLIMIT_STACK
, &lim
) == 0
626 && lim
.rlim_cur
!= RLIM_INFINITY
627 && lim
.rlim_cur
== (target_long
)lim
.rlim_cur
) {
628 guest_stack_size
= lim
.rlim_cur
;
634 qemu_add_opts(&qemu_trace_opts
);
636 optind
= parse_args(argc
, argv
);
638 if (!trace_init_backends()) {
641 trace_init_file(trace_file
);
644 memset(regs
, 0, sizeof(struct target_pt_regs
));
646 /* Zero out image_info */
647 memset(info
, 0, sizeof(struct image_info
));
649 memset(&bprm
, 0, sizeof (bprm
));
651 /* Scan interp_prefix dir for replacement files. */
652 init_paths(interp_prefix
);
654 init_qemu_uname_release();
656 execfd
= qemu_getauxval(AT_EXECFD
);
658 execfd
= open(exec_path
, O_RDONLY
);
660 printf("Error while loading %s: %s\n", exec_path
, strerror(errno
));
665 if (cpu_model
== NULL
) {
666 cpu_model
= cpu_get_model(get_elf_eflags(execfd
));
668 cpu_type
= parse_cpu_option(cpu_model
);
670 /* init tcg before creating CPUs and to get qemu_host_page_size */
673 cpu
= cpu_create(cpu_type
);
679 * Reserving too much vm space via mmap can run into problems
680 * with rlimits, oom due to page table creation, etc. We will
681 * still try it, if directed by the command-line option, but
684 max_reserved_va
= MAX_RESERVED_VA(cpu
);
685 if (reserved_va
!= 0) {
686 if (max_reserved_va
&& reserved_va
> max_reserved_va
) {
687 fprintf(stderr
, "Reserved virtual address too big\n");
690 } else if (HOST_LONG_BITS
== 64 && TARGET_VIRT_ADDR_SPACE_BITS
<= 32) {
692 * reserved_va must be aligned with the host page size
693 * as it is used with mmap()
695 reserved_va
= max_reserved_va
& qemu_host_page_mask
;
698 if (getenv("QEMU_STRACE")) {
702 if (seed_optarg
== NULL
) {
703 seed_optarg
= getenv("QEMU_RAND_SEED");
707 if (seed_optarg
!= NULL
) {
708 qemu_guest_random_seed_main(seed_optarg
, &err
);
713 error_reportf_err(err
, "cannot initialize crypto: ");
718 target_environ
= envlist_to_environ(envlist
, NULL
);
719 envlist_free(envlist
);
722 * Now that page sizes are configured in tcg_exec_init() we can do
723 * proper page alignment for guest_base.
725 guest_base
= HOST_PAGE_ALIGN(guest_base
);
727 if (reserved_va
|| have_guest_base
) {
728 guest_base
= init_guest_space(guest_base
, reserved_va
, 0,
730 if (guest_base
== (unsigned long)-1) {
731 fprintf(stderr
, "Unable to reserve 0x%lx bytes of virtual address "
732 "space for use as guest address space (check your virtual "
733 "memory ulimit setting or reserve less using -R option)\n",
739 mmap_next_start
= reserved_va
;
744 * Read in mmap_min_addr kernel parameter. This value is used
745 * When loading the ELF image to determine whether guest_base
746 * is needed. It is also used in mmap_find_vma.
751 if ((fp
= fopen("/proc/sys/vm/mmap_min_addr", "r")) != NULL
) {
753 if (fscanf(fp
, "%lu", &tmp
) == 1) {
755 qemu_log_mask(CPU_LOG_PAGE
, "host mmap_min_addr=0x%lx\n", mmap_min_addr
);
762 * Prepare copy of argv vector for target.
764 target_argc
= argc
- optind
;
765 target_argv
= calloc(target_argc
+ 1, sizeof (char *));
766 if (target_argv
== NULL
) {
767 (void) fprintf(stderr
, "Unable to allocate memory for target_argv\n");
772 * If argv0 is specified (using '-0' switch) we replace
773 * argv[0] pointer with the given one.
777 target_argv
[i
++] = strdup(argv0
);
779 for (; i
< target_argc
; i
++) {
780 target_argv
[i
] = strdup(argv
[optind
+ i
]);
782 target_argv
[target_argc
] = NULL
;
784 ts
= g_new0(TaskState
, 1);
786 /* build Task State */
792 ret
= loader_exec(execfd
, exec_path
, target_argv
, target_environ
, regs
,
795 printf("Error while loading %s: %s\n", exec_path
, strerror(-ret
));
799 for (wrk
= target_environ
; *wrk
; wrk
++) {
803 g_free(target_environ
);
805 if (qemu_loglevel_mask(CPU_LOG_PAGE
)) {
806 qemu_log("guest_base 0x%lx\n", guest_base
);
809 qemu_log("start_brk 0x" TARGET_ABI_FMT_lx
"\n", info
->start_brk
);
810 qemu_log("end_code 0x" TARGET_ABI_FMT_lx
"\n", info
->end_code
);
811 qemu_log("start_code 0x" TARGET_ABI_FMT_lx
"\n", info
->start_code
);
812 qemu_log("start_data 0x" TARGET_ABI_FMT_lx
"\n", info
->start_data
);
813 qemu_log("end_data 0x" TARGET_ABI_FMT_lx
"\n", info
->end_data
);
814 qemu_log("start_stack 0x" TARGET_ABI_FMT_lx
"\n", info
->start_stack
);
815 qemu_log("brk 0x" TARGET_ABI_FMT_lx
"\n", info
->brk
);
816 qemu_log("entry 0x" TARGET_ABI_FMT_lx
"\n", info
->entry
);
817 qemu_log("argv_start 0x" TARGET_ABI_FMT_lx
"\n", info
->arg_start
);
818 qemu_log("env_start 0x" TARGET_ABI_FMT_lx
"\n",
819 info
->arg_end
+ (abi_ulong
)sizeof(abi_ulong
));
820 qemu_log("auxv_start 0x" TARGET_ABI_FMT_lx
"\n", info
->saved_auxv
);
823 target_set_brk(info
->brk
);
827 /* Now that we've loaded the binary, GUEST_BASE is fixed. Delay
828 generating the prologue until now so that the prologue can take
829 the real value of GUEST_BASE into account. */
830 tcg_prologue_init(tcg_ctx
);
833 target_cpu_copy_regs(env
, regs
);
836 if (gdbserver_start(gdbstub_port
) < 0) {
837 fprintf(stderr
, "qemu: could not open gdbserver on port %d\n",
841 gdb_handlesig(cpu
, 0);