ia64/linux-2.6.18-xen.hg

view fs/binfmt_misc.c @ 524:7f8b544237bf

netfront: Allow netfront in domain 0.

This is useful if your physical network device is in a utility domain.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
author Keir Fraser <keir.fraser@citrix.com>
date Tue Apr 15 15:18:58 2008 +0100 (2008-04-15)
parents 831230e53067
children
line source
1 /*
2 * binfmt_misc.c
3 *
4 * Copyright (C) 1997 Richard GŁnther
5 *
6 * binfmt_misc detects binaries via a magic or filename extension and invokes
7 * a specified wrapper. This should obsolete binfmt_java, binfmt_em86 and
8 * binfmt_mz.
9 *
10 * 1997-04-25 first version
11 * [...]
12 * 1997-05-19 cleanup
13 * 1997-06-26 hpa: pass the real filename rather than argv[0]
14 * 1997-06-30 minor cleanup
15 * 1997-08-09 removed extension stripping, locking cleanup
16 * 2001-02-28 AV: rewritten into something that resembles C. Original didn't.
17 */
19 #include <linux/module.h>
20 #include <linux/init.h>
22 #include <linux/binfmts.h>
23 #include <linux/slab.h>
24 #include <linux/ctype.h>
25 #include <linux/file.h>
26 #include <linux/pagemap.h>
27 #include <linux/namei.h>
28 #include <linux/mount.h>
29 #include <linux/syscalls.h>
31 #include <asm/uaccess.h>
33 enum {
34 VERBOSE_STATUS = 1 /* make it zero to save 400 bytes kernel memory */
35 };
37 static LIST_HEAD(entries);
38 static int enabled = 1;
40 enum {Enabled, Magic};
41 #define MISC_FMT_PRESERVE_ARGV0 (1<<31)
42 #define MISC_FMT_OPEN_BINARY (1<<30)
43 #define MISC_FMT_CREDENTIALS (1<<29)
45 typedef struct {
46 struct list_head list;
47 unsigned long flags; /* type, status, etc. */
48 int offset; /* offset of magic */
49 int size; /* size of magic/mask */
50 char *magic; /* magic or filename extension */
51 char *mask; /* mask, NULL for exact match */
52 char *interpreter; /* filename of interpreter */
53 char *name;
54 struct dentry *dentry;
55 } Node;
57 static DEFINE_RWLOCK(entries_lock);
58 static struct file_system_type bm_fs_type;
59 static struct vfsmount *bm_mnt;
60 static int entry_count;
62 /*
63 * Check if we support the binfmt
64 * if we do, return the node, else NULL
65 * locking is done in load_misc_binary
66 */
67 static Node *check_file(struct linux_binprm *bprm)
68 {
69 char *p = strrchr(bprm->interp, '.');
70 struct list_head *l;
72 list_for_each(l, &entries) {
73 Node *e = list_entry(l, Node, list);
74 char *s;
75 int j;
77 if (!test_bit(Enabled, &e->flags))
78 continue;
80 if (!test_bit(Magic, &e->flags)) {
81 if (p && !strcmp(e->magic, p + 1))
82 return e;
83 continue;
84 }
86 s = bprm->buf + e->offset;
87 if (e->mask) {
88 for (j = 0; j < e->size; j++)
89 if ((*s++ ^ e->magic[j]) & e->mask[j])
90 break;
91 } else {
92 for (j = 0; j < e->size; j++)
93 if ((*s++ ^ e->magic[j]))
94 break;
95 }
96 if (j == e->size)
97 return e;
98 }
99 return NULL;
100 }
102 /*
103 * the loader itself
104 */
105 static int load_misc_binary(struct linux_binprm *bprm, struct pt_regs *regs)
106 {
107 Node *fmt;
108 struct file * interp_file = NULL;
109 char iname[BINPRM_BUF_SIZE];
110 char *iname_addr = iname;
111 int retval;
112 int fd_binary = -1;
113 struct files_struct *files = NULL;
115 retval = -ENOEXEC;
116 if (!enabled)
117 goto _ret;
119 /* to keep locking time low, we copy the interpreter string */
120 read_lock(&entries_lock);
121 fmt = check_file(bprm);
122 if (fmt)
123 strlcpy(iname, fmt->interpreter, BINPRM_BUF_SIZE);
124 read_unlock(&entries_lock);
125 if (!fmt)
126 goto _ret;
128 if (!(fmt->flags & MISC_FMT_PRESERVE_ARGV0)) {
129 remove_arg_zero(bprm);
130 }
132 if (fmt->flags & MISC_FMT_OPEN_BINARY) {
134 files = current->files;
135 retval = unshare_files();
136 if (retval < 0)
137 goto _ret;
138 if (files == current->files) {
139 put_files_struct(files);
140 files = NULL;
141 }
142 /* if the binary should be opened on behalf of the
143 * interpreter than keep it open and assign descriptor
144 * to it */
145 fd_binary = get_unused_fd();
146 if (fd_binary < 0) {
147 retval = fd_binary;
148 goto _unshare;
149 }
150 fd_install(fd_binary, bprm->file);
152 /* if the binary is not readable than enforce mm->dumpable=0
153 regardless of the interpreter's permissions */
154 if (file_permission(bprm->file, MAY_READ))
155 bprm->interp_flags |= BINPRM_FLAGS_ENFORCE_NONDUMP;
157 allow_write_access(bprm->file);
158 bprm->file = NULL;
160 /* mark the bprm that fd should be passed to interp */
161 bprm->interp_flags |= BINPRM_FLAGS_EXECFD;
162 bprm->interp_data = fd_binary;
164 } else {
165 allow_write_access(bprm->file);
166 fput(bprm->file);
167 bprm->file = NULL;
168 }
169 /* make argv[1] be the path to the binary */
170 retval = copy_strings_kernel (1, &bprm->interp, bprm);
171 if (retval < 0)
172 goto _error;
173 bprm->argc++;
175 /* add the interp as argv[0] */
176 retval = copy_strings_kernel (1, &iname_addr, bprm);
177 if (retval < 0)
178 goto _error;
179 bprm->argc ++;
181 bprm->interp = iname; /* for binfmt_script */
183 interp_file = open_exec (iname);
184 retval = PTR_ERR (interp_file);
185 if (IS_ERR (interp_file))
186 goto _error;
188 bprm->file = interp_file;
189 if (fmt->flags & MISC_FMT_CREDENTIALS) {
190 /*
191 * No need to call prepare_binprm(), it's already been
192 * done. bprm->buf is stale, update from interp_file.
193 */
194 memset(bprm->buf, 0, BINPRM_BUF_SIZE);
195 retval = kernel_read(bprm->file, 0, bprm->buf, BINPRM_BUF_SIZE);
196 } else
197 retval = prepare_binprm (bprm);
199 if (retval < 0)
200 goto _error;
202 retval = search_binary_handler (bprm, regs);
203 if (retval < 0)
204 goto _error;
206 if (files) {
207 put_files_struct(files);
208 files = NULL;
209 }
210 _ret:
211 return retval;
212 _error:
213 if (fd_binary > 0)
214 sys_close(fd_binary);
215 bprm->interp_flags = 0;
216 bprm->interp_data = 0;
217 _unshare:
218 if (files) {
219 put_files_struct(current->files);
220 current->files = files;
221 }
222 goto _ret;
223 }
225 /* Command parsers */
227 /*
228 * parses and copies one argument enclosed in del from *sp to *dp,
229 * recognising the \x special.
230 * returns pointer to the copied argument or NULL in case of an
231 * error (and sets err) or null argument length.
232 */
233 static char *scanarg(char *s, char del)
234 {
235 char c;
237 while ((c = *s++) != del) {
238 if (c == '\\' && *s == 'x') {
239 s++;
240 if (!isxdigit(*s++))
241 return NULL;
242 if (!isxdigit(*s++))
243 return NULL;
244 }
245 }
246 return s;
247 }
249 static int unquote(char *from)
250 {
251 char c = 0, *s = from, *p = from;
253 while ((c = *s++) != '\0') {
254 if (c == '\\' && *s == 'x') {
255 s++;
256 c = toupper(*s++);
257 *p = (c - (isdigit(c) ? '0' : 'A' - 10)) << 4;
258 c = toupper(*s++);
259 *p++ |= c - (isdigit(c) ? '0' : 'A' - 10);
260 continue;
261 }
262 *p++ = c;
263 }
264 return p - from;
265 }
267 static char * check_special_flags (char * sfs, Node * e)
268 {
269 char * p = sfs;
270 int cont = 1;
272 /* special flags */
273 while (cont) {
274 switch (*p) {
275 case 'P':
276 p++;
277 e->flags |= MISC_FMT_PRESERVE_ARGV0;
278 break;
279 case 'O':
280 p++;
281 e->flags |= MISC_FMT_OPEN_BINARY;
282 break;
283 case 'C':
284 p++;
285 /* this flags also implies the
286 open-binary flag */
287 e->flags |= (MISC_FMT_CREDENTIALS |
288 MISC_FMT_OPEN_BINARY);
289 break;
290 default:
291 cont = 0;
292 }
293 }
295 return p;
296 }
297 /*
298 * This registers a new binary format, it recognises the syntax
299 * ':name:type:offset:magic:mask:interpreter:flags'
300 * where the ':' is the IFS, that can be chosen with the first char
301 */
302 static Node *create_entry(const char __user *buffer, size_t count)
303 {
304 Node *e;
305 int memsize, err;
306 char *buf, *p;
307 char del;
309 /* some sanity checks */
310 err = -EINVAL;
311 if ((count < 11) || (count > 256))
312 goto out;
314 err = -ENOMEM;
315 memsize = sizeof(Node) + count + 8;
316 e = (Node *) kmalloc(memsize, GFP_USER);
317 if (!e)
318 goto out;
320 p = buf = (char *)e + sizeof(Node);
322 memset(e, 0, sizeof(Node));
323 if (copy_from_user(buf, buffer, count))
324 goto Efault;
326 del = *p++; /* delimeter */
328 memset(buf+count, del, 8);
330 e->name = p;
331 p = strchr(p, del);
332 if (!p)
333 goto Einval;
334 *p++ = '\0';
335 if (!e->name[0] ||
336 !strcmp(e->name, ".") ||
337 !strcmp(e->name, "..") ||
338 strchr(e->name, '/'))
339 goto Einval;
340 switch (*p++) {
341 case 'E': e->flags = 1<<Enabled; break;
342 case 'M': e->flags = (1<<Enabled) | (1<<Magic); break;
343 default: goto Einval;
344 }
345 if (*p++ != del)
346 goto Einval;
347 if (test_bit(Magic, &e->flags)) {
348 char *s = strchr(p, del);
349 if (!s)
350 goto Einval;
351 *s++ = '\0';
352 e->offset = simple_strtoul(p, &p, 10);
353 if (*p++)
354 goto Einval;
355 e->magic = p;
356 p = scanarg(p, del);
357 if (!p)
358 goto Einval;
359 p[-1] = '\0';
360 if (!e->magic[0])
361 goto Einval;
362 e->mask = p;
363 p = scanarg(p, del);
364 if (!p)
365 goto Einval;
366 p[-1] = '\0';
367 if (!e->mask[0])
368 e->mask = NULL;
369 e->size = unquote(e->magic);
370 if (e->mask && unquote(e->mask) != e->size)
371 goto Einval;
372 if (e->size + e->offset > BINPRM_BUF_SIZE)
373 goto Einval;
374 } else {
375 p = strchr(p, del);
376 if (!p)
377 goto Einval;
378 *p++ = '\0';
379 e->magic = p;
380 p = strchr(p, del);
381 if (!p)
382 goto Einval;
383 *p++ = '\0';
384 if (!e->magic[0] || strchr(e->magic, '/'))
385 goto Einval;
386 p = strchr(p, del);
387 if (!p)
388 goto Einval;
389 *p++ = '\0';
390 }
391 e->interpreter = p;
392 p = strchr(p, del);
393 if (!p)
394 goto Einval;
395 *p++ = '\0';
396 if (!e->interpreter[0])
397 goto Einval;
400 p = check_special_flags (p, e);
402 if (*p == '\n')
403 p++;
404 if (p != buf + count)
405 goto Einval;
406 return e;
408 out:
409 return ERR_PTR(err);
411 Efault:
412 kfree(e);
413 return ERR_PTR(-EFAULT);
414 Einval:
415 kfree(e);
416 return ERR_PTR(-EINVAL);
417 }
419 /*
420 * Set status of entry/binfmt_misc:
421 * '1' enables, '0' disables and '-1' clears entry/binfmt_misc
422 */
423 static int parse_command(const char __user *buffer, size_t count)
424 {
425 char s[4];
427 if (!count)
428 return 0;
429 if (count > 3)
430 return -EINVAL;
431 if (copy_from_user(s, buffer, count))
432 return -EFAULT;
433 if (s[count-1] == '\n')
434 count--;
435 if (count == 1 && s[0] == '0')
436 return 1;
437 if (count == 1 && s[0] == '1')
438 return 2;
439 if (count == 2 && s[0] == '-' && s[1] == '1')
440 return 3;
441 return -EINVAL;
442 }
444 /* generic stuff */
446 static void entry_status(Node *e, char *page)
447 {
448 char *dp;
449 char *status = "disabled";
450 const char * flags = "flags: ";
452 if (test_bit(Enabled, &e->flags))
453 status = "enabled";
455 if (!VERBOSE_STATUS) {
456 sprintf(page, "%s\n", status);
457 return;
458 }
460 sprintf(page, "%s\ninterpreter %s\n", status, e->interpreter);
461 dp = page + strlen(page);
463 /* print the special flags */
464 sprintf (dp, "%s", flags);
465 dp += strlen (flags);
466 if (e->flags & MISC_FMT_PRESERVE_ARGV0) {
467 *dp ++ = 'P';
468 }
469 if (e->flags & MISC_FMT_OPEN_BINARY) {
470 *dp ++ = 'O';
471 }
472 if (e->flags & MISC_FMT_CREDENTIALS) {
473 *dp ++ = 'C';
474 }
475 *dp ++ = '\n';
478 if (!test_bit(Magic, &e->flags)) {
479 sprintf(dp, "extension .%s\n", e->magic);
480 } else {
481 int i;
483 sprintf(dp, "offset %i\nmagic ", e->offset);
484 dp = page + strlen(page);
485 for (i = 0; i < e->size; i++) {
486 sprintf(dp, "%02x", 0xff & (int) (e->magic[i]));
487 dp += 2;
488 }
489 if (e->mask) {
490 sprintf(dp, "\nmask ");
491 dp += 6;
492 for (i = 0; i < e->size; i++) {
493 sprintf(dp, "%02x", 0xff & (int) (e->mask[i]));
494 dp += 2;
495 }
496 }
497 *dp++ = '\n';
498 *dp = '\0';
499 }
500 }
502 static struct inode *bm_get_inode(struct super_block *sb, int mode)
503 {
504 struct inode * inode = new_inode(sb);
506 if (inode) {
507 inode->i_mode = mode;
508 inode->i_uid = 0;
509 inode->i_gid = 0;
510 inode->i_blksize = PAGE_CACHE_SIZE;
511 inode->i_blocks = 0;
512 inode->i_atime = inode->i_mtime = inode->i_ctime =
513 current_fs_time(inode->i_sb);
514 }
515 return inode;
516 }
518 static void bm_clear_inode(struct inode *inode)
519 {
520 kfree(inode->u.generic_ip);
521 }
523 static void kill_node(Node *e)
524 {
525 struct dentry *dentry;
527 write_lock(&entries_lock);
528 dentry = e->dentry;
529 if (dentry) {
530 list_del_init(&e->list);
531 e->dentry = NULL;
532 }
533 write_unlock(&entries_lock);
535 if (dentry) {
536 dentry->d_inode->i_nlink--;
537 d_drop(dentry);
538 dput(dentry);
539 simple_release_fs(&bm_mnt, &entry_count);
540 }
541 }
543 /* /<entry> */
545 static ssize_t
546 bm_entry_read(struct file * file, char __user * buf, size_t nbytes, loff_t *ppos)
547 {
548 Node *e = file->f_dentry->d_inode->u.generic_ip;
549 loff_t pos = *ppos;
550 ssize_t res;
551 char *page;
552 int len;
554 if (!(page = (char*) __get_free_page(GFP_KERNEL)))
555 return -ENOMEM;
557 entry_status(e, page);
558 len = strlen(page);
560 res = -EINVAL;
561 if (pos < 0)
562 goto out;
563 res = 0;
564 if (pos >= len)
565 goto out;
566 if (len < pos + nbytes)
567 nbytes = len - pos;
568 res = -EFAULT;
569 if (copy_to_user(buf, page + pos, nbytes))
570 goto out;
571 *ppos = pos + nbytes;
572 res = nbytes;
573 out:
574 free_page((unsigned long) page);
575 return res;
576 }
578 static ssize_t bm_entry_write(struct file *file, const char __user *buffer,
579 size_t count, loff_t *ppos)
580 {
581 struct dentry *root;
582 Node *e = file->f_dentry->d_inode->u.generic_ip;
583 int res = parse_command(buffer, count);
585 switch (res) {
586 case 1: clear_bit(Enabled, &e->flags);
587 break;
588 case 2: set_bit(Enabled, &e->flags);
589 break;
590 case 3: root = dget(file->f_vfsmnt->mnt_sb->s_root);
591 mutex_lock(&root->d_inode->i_mutex);
593 kill_node(e);
595 mutex_unlock(&root->d_inode->i_mutex);
596 dput(root);
597 break;
598 default: return res;
599 }
600 return count;
601 }
603 static const struct file_operations bm_entry_operations = {
604 .read = bm_entry_read,
605 .write = bm_entry_write,
606 };
608 /* /register */
610 static ssize_t bm_register_write(struct file *file, const char __user *buffer,
611 size_t count, loff_t *ppos)
612 {
613 Node *e;
614 struct inode *inode;
615 struct dentry *root, *dentry;
616 struct super_block *sb = file->f_vfsmnt->mnt_sb;
617 int err = 0;
619 e = create_entry(buffer, count);
621 if (IS_ERR(e))
622 return PTR_ERR(e);
624 root = dget(sb->s_root);
625 mutex_lock(&root->d_inode->i_mutex);
626 dentry = lookup_one_len(e->name, root, strlen(e->name));
627 err = PTR_ERR(dentry);
628 if (IS_ERR(dentry))
629 goto out;
631 err = -EEXIST;
632 if (dentry->d_inode)
633 goto out2;
635 inode = bm_get_inode(sb, S_IFREG | 0644);
637 err = -ENOMEM;
638 if (!inode)
639 goto out2;
641 err = simple_pin_fs(&bm_fs_type, &bm_mnt, &entry_count);
642 if (err) {
643 iput(inode);
644 inode = NULL;
645 goto out2;
646 }
648 e->dentry = dget(dentry);
649 inode->u.generic_ip = e;
650 inode->i_fop = &bm_entry_operations;
652 d_instantiate(dentry, inode);
653 write_lock(&entries_lock);
654 list_add(&e->list, &entries);
655 write_unlock(&entries_lock);
657 err = 0;
658 out2:
659 dput(dentry);
660 out:
661 mutex_unlock(&root->d_inode->i_mutex);
662 dput(root);
664 if (err) {
665 kfree(e);
666 return -EINVAL;
667 }
668 return count;
669 }
671 static const struct file_operations bm_register_operations = {
672 .write = bm_register_write,
673 };
675 /* /status */
677 static ssize_t
678 bm_status_read(struct file *file, char __user *buf, size_t nbytes, loff_t *ppos)
679 {
680 char *s = enabled ? "enabled" : "disabled";
681 int len = strlen(s);
682 loff_t pos = *ppos;
684 if (pos < 0)
685 return -EINVAL;
686 if (pos >= len)
687 return 0;
688 if (len < pos + nbytes)
689 nbytes = len - pos;
690 if (copy_to_user(buf, s + pos, nbytes))
691 return -EFAULT;
692 *ppos = pos + nbytes;
693 return nbytes;
694 }
696 static ssize_t bm_status_write(struct file * file, const char __user * buffer,
697 size_t count, loff_t *ppos)
698 {
699 int res = parse_command(buffer, count);
700 struct dentry *root;
702 switch (res) {
703 case 1: enabled = 0; break;
704 case 2: enabled = 1; break;
705 case 3: root = dget(file->f_vfsmnt->mnt_sb->s_root);
706 mutex_lock(&root->d_inode->i_mutex);
708 while (!list_empty(&entries))
709 kill_node(list_entry(entries.next, Node, list));
711 mutex_unlock(&root->d_inode->i_mutex);
712 dput(root);
713 default: return res;
714 }
715 return count;
716 }
718 static const struct file_operations bm_status_operations = {
719 .read = bm_status_read,
720 .write = bm_status_write,
721 };
723 /* Superblock handling */
725 static struct super_operations s_ops = {
726 .statfs = simple_statfs,
727 .clear_inode = bm_clear_inode,
728 };
730 static int bm_fill_super(struct super_block * sb, void * data, int silent)
731 {
732 static struct tree_descr bm_files[] = {
733 [1] = {"status", &bm_status_operations, S_IWUSR|S_IRUGO},
734 [2] = {"register", &bm_register_operations, S_IWUSR},
735 /* last one */ {""}
736 };
737 int err = simple_fill_super(sb, 0x42494e4d, bm_files);
738 if (!err)
739 sb->s_op = &s_ops;
740 return err;
741 }
743 static int bm_get_sb(struct file_system_type *fs_type,
744 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
745 {
746 return get_sb_single(fs_type, flags, data, bm_fill_super, mnt);
747 }
749 static struct linux_binfmt misc_format = {
750 .module = THIS_MODULE,
751 .load_binary = load_misc_binary,
752 };
754 static struct file_system_type bm_fs_type = {
755 .owner = THIS_MODULE,
756 .name = "binfmt_misc",
757 .get_sb = bm_get_sb,
758 .kill_sb = kill_litter_super,
759 };
761 static int __init init_misc_binfmt(void)
762 {
763 int err = register_filesystem(&bm_fs_type);
764 if (!err) {
765 err = register_binfmt(&misc_format);
766 if (err)
767 unregister_filesystem(&bm_fs_type);
768 }
769 return err;
770 }
772 static void __exit exit_misc_binfmt(void)
773 {
774 unregister_binfmt(&misc_format);
775 unregister_filesystem(&bm_fs_type);
776 }
778 core_initcall(init_misc_binfmt);
779 module_exit(exit_misc_binfmt);
780 MODULE_LICENSE("GPL");