gro: Allow tunnel stacking in the case of FOU/GUE
[linux/fpc-iii.git] / fs / binfmt_misc.c
blob78f005f378476011a43c41e7d0eaa8d99ac1af21
1 /*
2 * binfmt_misc.c
4 * Copyright (C) 1997 Richard Günther
6 * binfmt_misc detects binaries via a magic or filename extension and invokes
7 * a specified wrapper. See Documentation/binfmt_misc.txt for more details.
8 */
10 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
12 #include <linux/kernel.h>
13 #include <linux/module.h>
14 #include <linux/init.h>
15 #include <linux/sched.h>
16 #include <linux/magic.h>
17 #include <linux/binfmts.h>
18 #include <linux/slab.h>
19 #include <linux/ctype.h>
20 #include <linux/string_helpers.h>
21 #include <linux/file.h>
22 #include <linux/pagemap.h>
23 #include <linux/namei.h>
24 #include <linux/mount.h>
25 #include <linux/syscalls.h>
26 #include <linux/fs.h>
27 #include <linux/uaccess.h>
29 #ifdef DEBUG
30 # define USE_DEBUG 1
31 #else
32 # define USE_DEBUG 0
33 #endif
35 enum {
36 VERBOSE_STATUS = 1 /* make it zero to save 400 bytes kernel memory */
39 static LIST_HEAD(entries);
40 static int enabled = 1;
42 enum {Enabled, Magic};
43 #define MISC_FMT_PRESERVE_ARGV0 (1 << 31)
44 #define MISC_FMT_OPEN_BINARY (1 << 30)
45 #define MISC_FMT_CREDENTIALS (1 << 29)
47 typedef struct {
48 struct list_head list;
49 unsigned long flags; /* type, status, etc. */
50 int offset; /* offset of magic */
51 int size; /* size of magic/mask */
52 char *magic; /* magic or filename extension */
53 char *mask; /* mask, NULL for exact match */
54 char *interpreter; /* filename of interpreter */
55 char *name;
56 struct dentry *dentry;
57 } Node;
59 static DEFINE_RWLOCK(entries_lock);
60 static struct file_system_type bm_fs_type;
61 static struct vfsmount *bm_mnt;
62 static int entry_count;
65 * Max length of the register string. Determined by:
66 * - 7 delimiters
67 * - name: ~50 bytes
68 * - type: 1 byte
69 * - offset: 3 bytes (has to be smaller than BINPRM_BUF_SIZE)
70 * - magic: 128 bytes (512 in escaped form)
71 * - mask: 128 bytes (512 in escaped form)
72 * - interp: ~50 bytes
73 * - flags: 5 bytes
74 * Round that up a bit, and then back off to hold the internal data
75 * (like struct Node).
77 #define MAX_REGISTER_LENGTH 1920
80 * Check if we support the binfmt
81 * if we do, return the node, else NULL
82 * locking is done in load_misc_binary
84 static Node *check_file(struct linux_binprm *bprm)
86 char *p = strrchr(bprm->interp, '.');
87 struct list_head *l;
89 /* Walk all the registered handlers. */
90 list_for_each(l, &entries) {
91 Node *e = list_entry(l, Node, list);
92 char *s;
93 int j;
95 /* Make sure this one is currently enabled. */
96 if (!test_bit(Enabled, &e->flags))
97 continue;
99 /* Do matching based on extension if applicable. */
100 if (!test_bit(Magic, &e->flags)) {
101 if (p && !strcmp(e->magic, p + 1))
102 return e;
103 continue;
106 /* Do matching based on magic & mask. */
107 s = bprm->buf + e->offset;
108 if (e->mask) {
109 for (j = 0; j < e->size; j++)
110 if ((*s++ ^ e->magic[j]) & e->mask[j])
111 break;
112 } else {
113 for (j = 0; j < e->size; j++)
114 if ((*s++ ^ e->magic[j]))
115 break;
117 if (j == e->size)
118 return e;
120 return NULL;
124 * the loader itself
126 static int load_misc_binary(struct linux_binprm *bprm)
128 Node *fmt;
129 struct file *interp_file = NULL;
130 char iname[BINPRM_BUF_SIZE];
131 const char *iname_addr = iname;
132 int retval;
133 int fd_binary = -1;
135 retval = -ENOEXEC;
136 if (!enabled)
137 goto ret;
139 /* to keep locking time low, we copy the interpreter string */
140 read_lock(&entries_lock);
141 fmt = check_file(bprm);
142 if (fmt)
143 strlcpy(iname, fmt->interpreter, BINPRM_BUF_SIZE);
144 read_unlock(&entries_lock);
145 if (!fmt)
146 goto ret;
148 /* Need to be able to load the file after exec */
149 if (bprm->interp_flags & BINPRM_FLAGS_PATH_INACCESSIBLE)
150 return -ENOENT;
152 if (!(fmt->flags & MISC_FMT_PRESERVE_ARGV0)) {
153 retval = remove_arg_zero(bprm);
154 if (retval)
155 goto ret;
158 if (fmt->flags & MISC_FMT_OPEN_BINARY) {
160 /* if the binary should be opened on behalf of the
161 * interpreter than keep it open and assign descriptor
162 * to it
164 fd_binary = get_unused_fd_flags(0);
165 if (fd_binary < 0) {
166 retval = fd_binary;
167 goto ret;
169 fd_install(fd_binary, bprm->file);
171 /* if the binary is not readable than enforce mm->dumpable=0
172 regardless of the interpreter's permissions */
173 would_dump(bprm, bprm->file);
175 allow_write_access(bprm->file);
176 bprm->file = NULL;
178 /* mark the bprm that fd should be passed to interp */
179 bprm->interp_flags |= BINPRM_FLAGS_EXECFD;
180 bprm->interp_data = fd_binary;
182 } else {
183 allow_write_access(bprm->file);
184 fput(bprm->file);
185 bprm->file = NULL;
187 /* make argv[1] be the path to the binary */
188 retval = copy_strings_kernel(1, &bprm->interp, bprm);
189 if (retval < 0)
190 goto error;
191 bprm->argc++;
193 /* add the interp as argv[0] */
194 retval = copy_strings_kernel(1, &iname_addr, bprm);
195 if (retval < 0)
196 goto error;
197 bprm->argc++;
199 /* Update interp in case binfmt_script needs it. */
200 retval = bprm_change_interp(iname, bprm);
201 if (retval < 0)
202 goto error;
204 interp_file = open_exec(iname);
205 retval = PTR_ERR(interp_file);
206 if (IS_ERR(interp_file))
207 goto error;
209 bprm->file = interp_file;
210 if (fmt->flags & MISC_FMT_CREDENTIALS) {
212 * No need to call prepare_binprm(), it's already been
213 * done. bprm->buf is stale, update from interp_file.
215 memset(bprm->buf, 0, BINPRM_BUF_SIZE);
216 retval = kernel_read(bprm->file, 0, bprm->buf, BINPRM_BUF_SIZE);
217 } else
218 retval = prepare_binprm(bprm);
220 if (retval < 0)
221 goto error;
223 retval = search_binary_handler(bprm);
224 if (retval < 0)
225 goto error;
227 ret:
228 return retval;
229 error:
230 if (fd_binary > 0)
231 sys_close(fd_binary);
232 bprm->interp_flags = 0;
233 bprm->interp_data = 0;
234 goto ret;
237 /* Command parsers */
240 * parses and copies one argument enclosed in del from *sp to *dp,
241 * recognising the \x special.
242 * returns pointer to the copied argument or NULL in case of an
243 * error (and sets err) or null argument length.
245 static char *scanarg(char *s, char del)
247 char c;
249 while ((c = *s++) != del) {
250 if (c == '\\' && *s == 'x') {
251 s++;
252 if (!isxdigit(*s++))
253 return NULL;
254 if (!isxdigit(*s++))
255 return NULL;
258 s[-1] ='\0';
259 return s;
262 static char *check_special_flags(char *sfs, Node *e)
264 char *p = sfs;
265 int cont = 1;
267 /* special flags */
268 while (cont) {
269 switch (*p) {
270 case 'P':
271 pr_debug("register: flag: P (preserve argv0)\n");
272 p++;
273 e->flags |= MISC_FMT_PRESERVE_ARGV0;
274 break;
275 case 'O':
276 pr_debug("register: flag: O (open binary)\n");
277 p++;
278 e->flags |= MISC_FMT_OPEN_BINARY;
279 break;
280 case 'C':
281 pr_debug("register: flag: C (preserve creds)\n");
282 p++;
283 /* this flags also implies the
284 open-binary flag */
285 e->flags |= (MISC_FMT_CREDENTIALS |
286 MISC_FMT_OPEN_BINARY);
287 break;
288 default:
289 cont = 0;
293 return p;
297 * This registers a new binary format, it recognises the syntax
298 * ':name:type:offset:magic:mask:interpreter:flags'
299 * where the ':' is the IFS, that can be chosen with the first char
301 static Node *create_entry(const char __user *buffer, size_t count)
303 Node *e;
304 int memsize, err;
305 char *buf, *p;
306 char del;
308 pr_debug("register: received %zu bytes\n", count);
310 /* some sanity checks */
311 err = -EINVAL;
312 if ((count < 11) || (count > MAX_REGISTER_LENGTH))
313 goto out;
315 err = -ENOMEM;
316 memsize = sizeof(Node) + count + 8;
317 e = kmalloc(memsize, GFP_KERNEL);
318 if (!e)
319 goto out;
321 p = buf = (char *)e + sizeof(Node);
323 memset(e, 0, sizeof(Node));
324 if (copy_from_user(buf, buffer, count))
325 goto efault;
327 del = *p++; /* delimeter */
329 pr_debug("register: delim: %#x {%c}\n", del, del);
331 /* Pad the buffer with the delim to simplify parsing below. */
332 memset(buf + count, del, 8);
334 /* Parse the 'name' field. */
335 e->name = p;
336 p = strchr(p, del);
337 if (!p)
338 goto einval;
339 *p++ = '\0';
340 if (!e->name[0] ||
341 !strcmp(e->name, ".") ||
342 !strcmp(e->name, "..") ||
343 strchr(e->name, '/'))
344 goto einval;
346 pr_debug("register: name: {%s}\n", e->name);
348 /* Parse the 'type' field. */
349 switch (*p++) {
350 case 'E':
351 pr_debug("register: type: E (extension)\n");
352 e->flags = 1 << Enabled;
353 break;
354 case 'M':
355 pr_debug("register: type: M (magic)\n");
356 e->flags = (1 << Enabled) | (1 << Magic);
357 break;
358 default:
359 goto einval;
361 if (*p++ != del)
362 goto einval;
364 if (test_bit(Magic, &e->flags)) {
365 /* Handle the 'M' (magic) format. */
366 char *s;
368 /* Parse the 'offset' field. */
369 s = strchr(p, del);
370 if (!s)
371 goto einval;
372 *s++ = '\0';
373 e->offset = simple_strtoul(p, &p, 10);
374 if (*p++)
375 goto einval;
376 pr_debug("register: offset: %#x\n", e->offset);
378 /* Parse the 'magic' field. */
379 e->magic = p;
380 p = scanarg(p, del);
381 if (!p)
382 goto einval;
383 if (!e->magic[0])
384 goto einval;
385 if (USE_DEBUG)
386 print_hex_dump_bytes(
387 KBUILD_MODNAME ": register: magic[raw]: ",
388 DUMP_PREFIX_NONE, e->magic, p - e->magic);
390 /* Parse the 'mask' field. */
391 e->mask = p;
392 p = scanarg(p, del);
393 if (!p)
394 goto einval;
395 if (!e->mask[0]) {
396 e->mask = NULL;
397 pr_debug("register: mask[raw]: none\n");
398 } else if (USE_DEBUG)
399 print_hex_dump_bytes(
400 KBUILD_MODNAME ": register: mask[raw]: ",
401 DUMP_PREFIX_NONE, e->mask, p - e->mask);
404 * Decode the magic & mask fields.
405 * Note: while we might have accepted embedded NUL bytes from
406 * above, the unescape helpers here will stop at the first one
407 * it encounters.
409 e->size = string_unescape_inplace(e->magic, UNESCAPE_HEX);
410 if (e->mask &&
411 string_unescape_inplace(e->mask, UNESCAPE_HEX) != e->size)
412 goto einval;
413 if (e->size + e->offset > BINPRM_BUF_SIZE)
414 goto einval;
415 pr_debug("register: magic/mask length: %i\n", e->size);
416 if (USE_DEBUG) {
417 print_hex_dump_bytes(
418 KBUILD_MODNAME ": register: magic[decoded]: ",
419 DUMP_PREFIX_NONE, e->magic, e->size);
421 if (e->mask) {
422 int i;
423 char *masked = kmalloc(e->size, GFP_KERNEL);
425 print_hex_dump_bytes(
426 KBUILD_MODNAME ": register: mask[decoded]: ",
427 DUMP_PREFIX_NONE, e->mask, e->size);
429 if (masked) {
430 for (i = 0; i < e->size; ++i)
431 masked[i] = e->magic[i] & e->mask[i];
432 print_hex_dump_bytes(
433 KBUILD_MODNAME ": register: magic[masked]: ",
434 DUMP_PREFIX_NONE, masked, e->size);
436 kfree(masked);
440 } else {
441 /* Handle the 'E' (extension) format. */
443 /* Skip the 'offset' field. */
444 p = strchr(p, del);
445 if (!p)
446 goto einval;
447 *p++ = '\0';
449 /* Parse the 'magic' field. */
450 e->magic = p;
451 p = strchr(p, del);
452 if (!p)
453 goto einval;
454 *p++ = '\0';
455 if (!e->magic[0] || strchr(e->magic, '/'))
456 goto einval;
457 pr_debug("register: extension: {%s}\n", e->magic);
459 /* Skip the 'mask' field. */
460 p = strchr(p, del);
461 if (!p)
462 goto einval;
463 *p++ = '\0';
466 /* Parse the 'interpreter' field. */
467 e->interpreter = p;
468 p = strchr(p, del);
469 if (!p)
470 goto einval;
471 *p++ = '\0';
472 if (!e->interpreter[0])
473 goto einval;
474 pr_debug("register: interpreter: {%s}\n", e->interpreter);
476 /* Parse the 'flags' field. */
477 p = check_special_flags(p, e);
478 if (*p == '\n')
479 p++;
480 if (p != buf + count)
481 goto einval;
483 return e;
485 out:
486 return ERR_PTR(err);
488 efault:
489 kfree(e);
490 return ERR_PTR(-EFAULT);
491 einval:
492 kfree(e);
493 return ERR_PTR(-EINVAL);
497 * Set status of entry/binfmt_misc:
498 * '1' enables, '0' disables and '-1' clears entry/binfmt_misc
500 static int parse_command(const char __user *buffer, size_t count)
502 char s[4];
504 if (count > 3)
505 return -EINVAL;
506 if (copy_from_user(s, buffer, count))
507 return -EFAULT;
508 if (!count)
509 return 0;
510 if (s[count - 1] == '\n')
511 count--;
512 if (count == 1 && s[0] == '0')
513 return 1;
514 if (count == 1 && s[0] == '1')
515 return 2;
516 if (count == 2 && s[0] == '-' && s[1] == '1')
517 return 3;
518 return -EINVAL;
521 /* generic stuff */
523 static void entry_status(Node *e, char *page)
525 char *dp = page;
526 const char *status = "disabled";
528 if (test_bit(Enabled, &e->flags))
529 status = "enabled";
531 if (!VERBOSE_STATUS) {
532 sprintf(page, "%s\n", status);
533 return;
536 dp += sprintf(dp, "%s\ninterpreter %s\n", status, e->interpreter);
538 /* print the special flags */
539 dp += sprintf(dp, "flags: ");
540 if (e->flags & MISC_FMT_PRESERVE_ARGV0)
541 *dp++ = 'P';
542 if (e->flags & MISC_FMT_OPEN_BINARY)
543 *dp++ = 'O';
544 if (e->flags & MISC_FMT_CREDENTIALS)
545 *dp++ = 'C';
546 *dp++ = '\n';
548 if (!test_bit(Magic, &e->flags)) {
549 sprintf(dp, "extension .%s\n", e->magic);
550 } else {
551 dp += sprintf(dp, "offset %i\nmagic ", e->offset);
552 dp = bin2hex(dp, e->magic, e->size);
553 if (e->mask) {
554 dp += sprintf(dp, "\nmask ");
555 dp = bin2hex(dp, e->mask, e->size);
557 *dp++ = '\n';
558 *dp = '\0';
562 static struct inode *bm_get_inode(struct super_block *sb, int mode)
564 struct inode *inode = new_inode(sb);
566 if (inode) {
567 inode->i_ino = get_next_ino();
568 inode->i_mode = mode;
569 inode->i_atime = inode->i_mtime = inode->i_ctime =
570 current_fs_time(inode->i_sb);
572 return inode;
575 static void bm_evict_inode(struct inode *inode)
577 clear_inode(inode);
578 kfree(inode->i_private);
581 static void kill_node(Node *e)
583 struct dentry *dentry;
585 write_lock(&entries_lock);
586 dentry = e->dentry;
587 if (dentry) {
588 list_del_init(&e->list);
589 e->dentry = NULL;
591 write_unlock(&entries_lock);
593 if (dentry) {
594 drop_nlink(d_inode(dentry));
595 d_drop(dentry);
596 dput(dentry);
597 simple_release_fs(&bm_mnt, &entry_count);
601 /* /<entry> */
603 static ssize_t
604 bm_entry_read(struct file *file, char __user *buf, size_t nbytes, loff_t *ppos)
606 Node *e = file_inode(file)->i_private;
607 ssize_t res;
608 char *page;
610 page = (char *) __get_free_page(GFP_KERNEL);
611 if (!page)
612 return -ENOMEM;
614 entry_status(e, page);
616 res = simple_read_from_buffer(buf, nbytes, ppos, page, strlen(page));
618 free_page((unsigned long) page);
619 return res;
622 static ssize_t bm_entry_write(struct file *file, const char __user *buffer,
623 size_t count, loff_t *ppos)
625 struct dentry *root;
626 Node *e = file_inode(file)->i_private;
627 int res = parse_command(buffer, count);
629 switch (res) {
630 case 1:
631 /* Disable this handler. */
632 clear_bit(Enabled, &e->flags);
633 break;
634 case 2:
635 /* Enable this handler. */
636 set_bit(Enabled, &e->flags);
637 break;
638 case 3:
639 /* Delete this handler. */
640 root = dget(file->f_path.dentry->d_sb->s_root);
641 mutex_lock(&d_inode(root)->i_mutex);
643 kill_node(e);
645 mutex_unlock(&d_inode(root)->i_mutex);
646 dput(root);
647 break;
648 default:
649 return res;
652 return count;
655 static const struct file_operations bm_entry_operations = {
656 .read = bm_entry_read,
657 .write = bm_entry_write,
658 .llseek = default_llseek,
661 /* /register */
663 static ssize_t bm_register_write(struct file *file, const char __user *buffer,
664 size_t count, loff_t *ppos)
666 Node *e;
667 struct inode *inode;
668 struct dentry *root, *dentry;
669 struct super_block *sb = file->f_path.dentry->d_sb;
670 int err = 0;
672 e = create_entry(buffer, count);
674 if (IS_ERR(e))
675 return PTR_ERR(e);
677 root = dget(sb->s_root);
678 mutex_lock(&d_inode(root)->i_mutex);
679 dentry = lookup_one_len(e->name, root, strlen(e->name));
680 err = PTR_ERR(dentry);
681 if (IS_ERR(dentry))
682 goto out;
684 err = -EEXIST;
685 if (d_really_is_positive(dentry))
686 goto out2;
688 inode = bm_get_inode(sb, S_IFREG | 0644);
690 err = -ENOMEM;
691 if (!inode)
692 goto out2;
694 err = simple_pin_fs(&bm_fs_type, &bm_mnt, &entry_count);
695 if (err) {
696 iput(inode);
697 inode = NULL;
698 goto out2;
701 e->dentry = dget(dentry);
702 inode->i_private = e;
703 inode->i_fop = &bm_entry_operations;
705 d_instantiate(dentry, inode);
706 write_lock(&entries_lock);
707 list_add(&e->list, &entries);
708 write_unlock(&entries_lock);
710 err = 0;
711 out2:
712 dput(dentry);
713 out:
714 mutex_unlock(&d_inode(root)->i_mutex);
715 dput(root);
717 if (err) {
718 kfree(e);
719 return -EINVAL;
721 return count;
724 static const struct file_operations bm_register_operations = {
725 .write = bm_register_write,
726 .llseek = noop_llseek,
729 /* /status */
731 static ssize_t
732 bm_status_read(struct file *file, char __user *buf, size_t nbytes, loff_t *ppos)
734 char *s = enabled ? "enabled\n" : "disabled\n";
736 return simple_read_from_buffer(buf, nbytes, ppos, s, strlen(s));
739 static ssize_t bm_status_write(struct file *file, const char __user *buffer,
740 size_t count, loff_t *ppos)
742 int res = parse_command(buffer, count);
743 struct dentry *root;
745 switch (res) {
746 case 1:
747 /* Disable all handlers. */
748 enabled = 0;
749 break;
750 case 2:
751 /* Enable all handlers. */
752 enabled = 1;
753 break;
754 case 3:
755 /* Delete all handlers. */
756 root = dget(file->f_path.dentry->d_sb->s_root);
757 mutex_lock(&d_inode(root)->i_mutex);
759 while (!list_empty(&entries))
760 kill_node(list_entry(entries.next, Node, list));
762 mutex_unlock(&d_inode(root)->i_mutex);
763 dput(root);
764 break;
765 default:
766 return res;
769 return count;
772 static const struct file_operations bm_status_operations = {
773 .read = bm_status_read,
774 .write = bm_status_write,
775 .llseek = default_llseek,
778 /* Superblock handling */
780 static const struct super_operations s_ops = {
781 .statfs = simple_statfs,
782 .evict_inode = bm_evict_inode,
785 static int bm_fill_super(struct super_block *sb, void *data, int silent)
787 int err;
788 static struct tree_descr bm_files[] = {
789 [2] = {"status", &bm_status_operations, S_IWUSR|S_IRUGO},
790 [3] = {"register", &bm_register_operations, S_IWUSR},
791 /* last one */ {""}
794 err = simple_fill_super(sb, BINFMTFS_MAGIC, bm_files);
795 if (!err)
796 sb->s_op = &s_ops;
797 return err;
800 static struct dentry *bm_mount(struct file_system_type *fs_type,
801 int flags, const char *dev_name, void *data)
803 return mount_single(fs_type, flags, data, bm_fill_super);
806 static struct linux_binfmt misc_format = {
807 .module = THIS_MODULE,
808 .load_binary = load_misc_binary,
811 static struct file_system_type bm_fs_type = {
812 .owner = THIS_MODULE,
813 .name = "binfmt_misc",
814 .mount = bm_mount,
815 .kill_sb = kill_litter_super,
817 MODULE_ALIAS_FS("binfmt_misc");
819 static int __init init_misc_binfmt(void)
821 int err = register_filesystem(&bm_fs_type);
822 if (!err)
823 insert_binfmt(&misc_format);
824 return err;
827 static void __exit exit_misc_binfmt(void)
829 unregister_binfmt(&misc_format);
830 unregister_filesystem(&bm_fs_type);
833 core_initcall(init_misc_binfmt);
834 module_exit(exit_misc_binfmt);
835 MODULE_LICENSE("GPL");