1 // SPDX-License-Identifier: GPL-2.0-only 1 // SPDX-License-Identifier: GPL-2.0-only
2 /* 2 /*
3 * binfmt_misc.c 3 * binfmt_misc.c
4 * 4 *
5 * Copyright (C) 1997 Richard Günther 5 * Copyright (C) 1997 Richard Günther
6 * 6 *
7 * binfmt_misc detects binaries via a magic or 7 * binfmt_misc detects binaries via a magic or filename extension and invokes
8 * a specified wrapper. See Documentation/admi 8 * a specified wrapper. See Documentation/admin-guide/binfmt-misc.rst for more details.
9 */ 9 */
10 10
11 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt 11 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
12 12
13 #include <linux/kernel.h> 13 #include <linux/kernel.h>
14 #include <linux/module.h> 14 #include <linux/module.h>
15 #include <linux/init.h> 15 #include <linux/init.h>
16 #include <linux/sched/mm.h> 16 #include <linux/sched/mm.h>
17 #include <linux/magic.h> 17 #include <linux/magic.h>
18 #include <linux/binfmts.h> 18 #include <linux/binfmts.h>
19 #include <linux/slab.h> 19 #include <linux/slab.h>
20 #include <linux/ctype.h> 20 #include <linux/ctype.h>
21 #include <linux/string_helpers.h> 21 #include <linux/string_helpers.h>
22 #include <linux/file.h> 22 #include <linux/file.h>
23 #include <linux/pagemap.h> 23 #include <linux/pagemap.h>
24 #include <linux/namei.h> 24 #include <linux/namei.h>
25 #include <linux/mount.h> 25 #include <linux/mount.h>
26 #include <linux/fs_context.h> 26 #include <linux/fs_context.h>
27 #include <linux/syscalls.h> 27 #include <linux/syscalls.h>
28 #include <linux/fs.h> 28 #include <linux/fs.h>
29 #include <linux/uaccess.h> 29 #include <linux/uaccess.h>
30 30
31 #include "internal.h" 31 #include "internal.h"
32 32
33 #ifdef DEBUG 33 #ifdef DEBUG
34 # define USE_DEBUG 1 34 # define USE_DEBUG 1
35 #else 35 #else
36 # define USE_DEBUG 0 36 # define USE_DEBUG 0
37 #endif 37 #endif
38 38
39 enum { 39 enum {
40 VERBOSE_STATUS = 1 /* make it zero to 40 VERBOSE_STATUS = 1 /* make it zero to save 400 bytes kernel memory */
41 }; 41 };
42 42
>> 43 static LIST_HEAD(entries);
>> 44 static int enabled = 1;
>> 45
43 enum {Enabled, Magic}; 46 enum {Enabled, Magic};
44 #define MISC_FMT_PRESERVE_ARGV0 (1UL << 31) !! 47 #define MISC_FMT_PRESERVE_ARGV0 (1 << 31)
45 #define MISC_FMT_OPEN_BINARY (1UL << 30) !! 48 #define MISC_FMT_OPEN_BINARY (1 << 30)
46 #define MISC_FMT_CREDENTIALS (1UL << 29) !! 49 #define MISC_FMT_CREDENTIALS (1 << 29)
47 #define MISC_FMT_OPEN_FILE (1UL << 28) !! 50 #define MISC_FMT_OPEN_FILE (1 << 28)
48 51
49 typedef struct { 52 typedef struct {
50 struct list_head list; 53 struct list_head list;
51 unsigned long flags; /* typ 54 unsigned long flags; /* type, status, etc. */
52 int offset; /* off 55 int offset; /* offset of magic */
53 int size; /* siz 56 int size; /* size of magic/mask */
54 char *magic; /* mag 57 char *magic; /* magic or filename extension */
55 char *mask; /* mas 58 char *mask; /* mask, NULL for exact match */
56 const char *interpreter; /* fil 59 const char *interpreter; /* filename of interpreter */
57 char *name; 60 char *name;
58 struct dentry *dentry; 61 struct dentry *dentry;
59 struct file *interp_file; 62 struct file *interp_file;
60 refcount_t users; /* syn <<
61 } Node; 63 } Node;
62 64
>> 65 static DEFINE_RWLOCK(entries_lock);
63 static struct file_system_type bm_fs_type; 66 static struct file_system_type bm_fs_type;
>> 67 static struct vfsmount *bm_mnt;
>> 68 static int entry_count;
64 69
65 /* 70 /*
66 * Max length of the register string. Determi 71 * Max length of the register string. Determined by:
67 * - 7 delimiters 72 * - 7 delimiters
68 * - name: ~50 bytes 73 * - name: ~50 bytes
69 * - type: 1 byte 74 * - type: 1 byte
70 * - offset: 3 bytes (has to be smaller than 75 * - offset: 3 bytes (has to be smaller than BINPRM_BUF_SIZE)
71 * - magic: 128 bytes (512 in escaped form) 76 * - magic: 128 bytes (512 in escaped form)
72 * - mask: 128 bytes (512 in escaped form) 77 * - mask: 128 bytes (512 in escaped form)
73 * - interp: ~50 bytes 78 * - interp: ~50 bytes
74 * - flags: 5 bytes 79 * - flags: 5 bytes
75 * Round that up a bit, and then back off to h 80 * Round that up a bit, and then back off to hold the internal data
76 * (like struct Node). 81 * (like struct Node).
77 */ 82 */
78 #define MAX_REGISTER_LENGTH 1920 83 #define MAX_REGISTER_LENGTH 1920
79 84
80 /** !! 85 /*
81 * search_binfmt_handler - search for a binary !! 86 * Check if we support the binfmt
82 * @misc: handle to binfmt_misc instance !! 87 * if we do, return the node, else NULL
83 * @bprm: binary for which we are looking for !! 88 * locking is done in load_misc_binary
84 * <<
85 * Search for a binary type handler for @bprm <<
86 * type handlers. <<
87 * <<
88 * Return: binary type list entry on success, <<
89 */ 89 */
90 static Node *search_binfmt_handler(struct binf !! 90 static Node *check_file(struct linux_binprm *bprm)
91 struct linu <<
92 { 91 {
93 char *p = strrchr(bprm->interp, '.'); 92 char *p = strrchr(bprm->interp, '.');
94 Node *e; !! 93 struct list_head *l;
95 94
96 /* Walk all the registered handlers. * 95 /* Walk all the registered handlers. */
97 list_for_each_entry(e, &misc->entries, !! 96 list_for_each(l, &entries) {
>> 97 Node *e = list_entry(l, Node, list);
98 char *s; 98 char *s;
99 int j; 99 int j;
100 100
101 /* Make sure this one is curre 101 /* Make sure this one is currently enabled. */
102 if (!test_bit(Enabled, &e->fla 102 if (!test_bit(Enabled, &e->flags))
103 continue; 103 continue;
104 104
105 /* Do matching based on extens 105 /* Do matching based on extension if applicable. */
106 if (!test_bit(Magic, &e->flags 106 if (!test_bit(Magic, &e->flags)) {
107 if (p && !strcmp(e->ma 107 if (p && !strcmp(e->magic, p + 1))
108 return e; 108 return e;
109 continue; 109 continue;
110 } 110 }
111 111
112 /* Do matching based on magic 112 /* Do matching based on magic & mask. */
113 s = bprm->buf + e->offset; 113 s = bprm->buf + e->offset;
114 if (e->mask) { 114 if (e->mask) {
115 for (j = 0; j < e->siz 115 for (j = 0; j < e->size; j++)
116 if ((*s++ ^ e- 116 if ((*s++ ^ e->magic[j]) & e->mask[j])
117 break; 117 break;
118 } else { 118 } else {
119 for (j = 0; j < e->siz 119 for (j = 0; j < e->size; j++)
120 if ((*s++ ^ e- 120 if ((*s++ ^ e->magic[j]))
121 break; 121 break;
122 } 122 }
123 if (j == e->size) 123 if (j == e->size)
124 return e; 124 return e;
125 } 125 }
126 <<
127 return NULL; 126 return NULL;
128 } 127 }
129 128
130 /** <<
131 * get_binfmt_handler - try to find a binary t <<
132 * @misc: handle to binfmt_misc instance <<
133 * @bprm: binary for which we are looking for <<
134 * <<
135 * Try to find a binfmt handler for the binary <<
136 * reference to protect against removal via bm <<
137 * <<
138 * Return: binary type list entry on success, <<
139 */ <<
140 static Node *get_binfmt_handler(struct binfmt_ <<
141 struct linux_b <<
142 { <<
143 Node *e; <<
144 <<
145 read_lock(&misc->entries_lock); <<
146 e = search_binfmt_handler(misc, bprm); <<
147 if (e) <<
148 refcount_inc(&e->users); <<
149 read_unlock(&misc->entries_lock); <<
150 return e; <<
151 } <<
152 <<
153 /** <<
154 * put_binfmt_handler - put binary handler nod <<
155 * @e: node to put <<
156 * <<
157 * Free node syncing with load_misc_binary() a <<
158 * load_misc_binary() in case it is using the <<
159 * requested to remove. <<
160 */ <<
161 static void put_binfmt_handler(Node *e) <<
162 { <<
163 if (refcount_dec_and_test(&e->users)) <<
164 if (e->flags & MISC_FMT_OPEN_F <<
165 filp_close(e->interp_f <<
166 kfree(e); <<
167 } <<
168 } <<
169 <<
170 /** <<
171 * load_binfmt_misc - load the binfmt_misc of <<
172 * <<
173 * To be called in load_misc_binary() to load <<
174 * If a user namespace doesn't have its own bi <<
175 * of its ancestor's binfmt_misc handlers. Thi <<
176 * pre-namespaced binfmt_misc where all regist <<
177 * available to all user and user namespaces o <<
178 * <<
179 * Return: the binfmt_misc instance of the cal <<
180 */ <<
181 static struct binfmt_misc *load_binfmt_misc(vo <<
182 { <<
183 const struct user_namespace *user_ns; <<
184 struct binfmt_misc *misc; <<
185 <<
186 user_ns = current_user_ns(); <<
187 while (user_ns) { <<
188 /* Pairs with smp_store_releas <<
189 misc = smp_load_acquire(&user_ <<
190 if (misc) <<
191 return misc; <<
192 <<
193 user_ns = user_ns->parent; <<
194 } <<
195 <<
196 return &init_binfmt_misc; <<
197 } <<
198 <<
199 /* 129 /*
200 * the loader itself 130 * the loader itself
201 */ 131 */
202 static int load_misc_binary(struct linux_binpr 132 static int load_misc_binary(struct linux_binprm *bprm)
203 { 133 {
204 Node *fmt; 134 Node *fmt;
205 struct file *interp_file = NULL; 135 struct file *interp_file = NULL;
206 int retval = -ENOEXEC; !! 136 int retval;
207 struct binfmt_misc *misc; <<
208 137
209 misc = load_binfmt_misc(); !! 138 retval = -ENOEXEC;
210 if (!misc->enabled) !! 139 if (!enabled)
211 return retval; 140 return retval;
212 141
213 fmt = get_binfmt_handler(misc, bprm); !! 142 /* to keep locking time low, we copy the interpreter string */
>> 143 read_lock(&entries_lock);
>> 144 fmt = check_file(bprm);
>> 145 if (fmt)
>> 146 dget(fmt->dentry);
>> 147 read_unlock(&entries_lock);
214 if (!fmt) 148 if (!fmt)
215 return retval; 149 return retval;
216 150
217 /* Need to be able to load the file af 151 /* Need to be able to load the file after exec */
218 retval = -ENOENT; 152 retval = -ENOENT;
219 if (bprm->interp_flags & BINPRM_FLAGS_ 153 if (bprm->interp_flags & BINPRM_FLAGS_PATH_INACCESSIBLE)
220 goto ret; 154 goto ret;
221 155
222 if (fmt->flags & MISC_FMT_PRESERVE_ARG !! 156 if (!(fmt->flags & MISC_FMT_PRESERVE_ARGV0)) {
223 bprm->interp_flags |= BINPRM_F <<
224 } else { <<
225 retval = remove_arg_zero(bprm) 157 retval = remove_arg_zero(bprm);
226 if (retval) 158 if (retval)
227 goto ret; 159 goto ret;
228 } 160 }
229 161
230 if (fmt->flags & MISC_FMT_OPEN_BINARY) 162 if (fmt->flags & MISC_FMT_OPEN_BINARY)
231 bprm->have_execfd = 1; 163 bprm->have_execfd = 1;
232 164
233 /* make argv[1] be the path to the bin 165 /* make argv[1] be the path to the binary */
234 retval = copy_string_kernel(bprm->inte 166 retval = copy_string_kernel(bprm->interp, bprm);
235 if (retval < 0) 167 if (retval < 0)
236 goto ret; 168 goto ret;
237 bprm->argc++; 169 bprm->argc++;
238 170
239 /* add the interp as argv[0] */ 171 /* add the interp as argv[0] */
240 retval = copy_string_kernel(fmt->inter 172 retval = copy_string_kernel(fmt->interpreter, bprm);
241 if (retval < 0) 173 if (retval < 0)
242 goto ret; 174 goto ret;
243 bprm->argc++; 175 bprm->argc++;
244 176
245 /* Update interp in case binfmt_script 177 /* Update interp in case binfmt_script needs it. */
246 retval = bprm_change_interp(fmt->inter 178 retval = bprm_change_interp(fmt->interpreter, bprm);
247 if (retval < 0) 179 if (retval < 0)
248 goto ret; 180 goto ret;
249 181
250 if (fmt->flags & MISC_FMT_OPEN_FILE) !! 182 if (fmt->flags & MISC_FMT_OPEN_FILE) {
251 interp_file = file_clone_open( 183 interp_file = file_clone_open(fmt->interp_file);
252 else !! 184 if (!IS_ERR(interp_file))
>> 185 deny_write_access(interp_file);
>> 186 } else {
253 interp_file = open_exec(fmt->i 187 interp_file = open_exec(fmt->interpreter);
>> 188 }
254 retval = PTR_ERR(interp_file); 189 retval = PTR_ERR(interp_file);
255 if (IS_ERR(interp_file)) 190 if (IS_ERR(interp_file))
256 goto ret; 191 goto ret;
257 192
258 bprm->interpreter = interp_file; 193 bprm->interpreter = interp_file;
259 if (fmt->flags & MISC_FMT_CREDENTIALS) 194 if (fmt->flags & MISC_FMT_CREDENTIALS)
260 bprm->execfd_creds = 1; 195 bprm->execfd_creds = 1;
261 196
262 retval = 0; 197 retval = 0;
263 ret: 198 ret:
264 !! 199 dput(fmt->dentry);
265 /* <<
266 * If we actually put the node here al <<
267 * load_misc_binary() will have finish <<
268 * that for the refcount to be zero so <<
269 * removed the binary type handler fro <<
270 * free it. <<
271 */ <<
272 put_binfmt_handler(fmt); <<
273 <<
274 return retval; 200 return retval;
275 } 201 }
276 202
277 /* Command parsers */ 203 /* Command parsers */
278 204
279 /* 205 /*
280 * parses and copies one argument enclosed in 206 * parses and copies one argument enclosed in del from *sp to *dp,
281 * recognising the \x special. 207 * recognising the \x special.
282 * returns pointer to the copied argument or N 208 * returns pointer to the copied argument or NULL in case of an
283 * error (and sets err) or null argument lengt 209 * error (and sets err) or null argument length.
284 */ 210 */
285 static char *scanarg(char *s, char del) 211 static char *scanarg(char *s, char del)
286 { 212 {
287 char c; 213 char c;
288 214
289 while ((c = *s++) != del) { 215 while ((c = *s++) != del) {
290 if (c == '\\' && *s == 'x') { 216 if (c == '\\' && *s == 'x') {
291 s++; 217 s++;
292 if (!isxdigit(*s++)) 218 if (!isxdigit(*s++))
293 return NULL; 219 return NULL;
294 if (!isxdigit(*s++)) 220 if (!isxdigit(*s++))
295 return NULL; 221 return NULL;
296 } 222 }
297 } 223 }
298 s[-1] ='\0'; 224 s[-1] ='\0';
299 return s; 225 return s;
300 } 226 }
301 227
302 static char *check_special_flags(char *sfs, No 228 static char *check_special_flags(char *sfs, Node *e)
303 { 229 {
304 char *p = sfs; 230 char *p = sfs;
305 int cont = 1; 231 int cont = 1;
306 232
307 /* special flags */ 233 /* special flags */
308 while (cont) { 234 while (cont) {
309 switch (*p) { 235 switch (*p) {
310 case 'P': 236 case 'P':
311 pr_debug("register: fl 237 pr_debug("register: flag: P (preserve argv0)\n");
312 p++; 238 p++;
313 e->flags |= MISC_FMT_P 239 e->flags |= MISC_FMT_PRESERVE_ARGV0;
314 break; 240 break;
315 case 'O': 241 case 'O':
316 pr_debug("register: fl 242 pr_debug("register: flag: O (open binary)\n");
317 p++; 243 p++;
318 e->flags |= MISC_FMT_O 244 e->flags |= MISC_FMT_OPEN_BINARY;
319 break; 245 break;
320 case 'C': 246 case 'C':
321 pr_debug("register: fl 247 pr_debug("register: flag: C (preserve creds)\n");
322 p++; 248 p++;
323 /* this flags also imp 249 /* this flags also implies the
324 open-binary flag */ 250 open-binary flag */
325 e->flags |= (MISC_FMT_ 251 e->flags |= (MISC_FMT_CREDENTIALS |
326 MISC_F 252 MISC_FMT_OPEN_BINARY);
327 break; 253 break;
328 case 'F': 254 case 'F':
329 pr_debug("register: fl 255 pr_debug("register: flag: F: open interpreter file now\n");
330 p++; 256 p++;
331 e->flags |= MISC_FMT_O 257 e->flags |= MISC_FMT_OPEN_FILE;
332 break; 258 break;
333 default: 259 default:
334 cont = 0; 260 cont = 0;
335 } 261 }
336 } 262 }
337 263
338 return p; 264 return p;
339 } 265 }
340 266
341 /* 267 /*
342 * This registers a new binary format, it reco 268 * This registers a new binary format, it recognises the syntax
343 * ':name:type:offset:magic:mask:interpreter:f 269 * ':name:type:offset:magic:mask:interpreter:flags'
344 * where the ':' is the IFS, that can be chose 270 * where the ':' is the IFS, that can be chosen with the first char
345 */ 271 */
346 static Node *create_entry(const char __user *b 272 static Node *create_entry(const char __user *buffer, size_t count)
347 { 273 {
348 Node *e; 274 Node *e;
349 int memsize, err; 275 int memsize, err;
350 char *buf, *p; 276 char *buf, *p;
351 char del; 277 char del;
352 278
353 pr_debug("register: received %zu bytes 279 pr_debug("register: received %zu bytes\n", count);
354 280
355 /* some sanity checks */ 281 /* some sanity checks */
356 err = -EINVAL; 282 err = -EINVAL;
357 if ((count < 11) || (count > MAX_REGIS 283 if ((count < 11) || (count > MAX_REGISTER_LENGTH))
358 goto out; 284 goto out;
359 285
360 err = -ENOMEM; 286 err = -ENOMEM;
361 memsize = sizeof(Node) + count + 8; 287 memsize = sizeof(Node) + count + 8;
362 e = kmalloc(memsize, GFP_KERNEL_ACCOUN !! 288 e = kmalloc(memsize, GFP_KERNEL);
363 if (!e) 289 if (!e)
364 goto out; 290 goto out;
365 291
366 p = buf = (char *)e + sizeof(Node); 292 p = buf = (char *)e + sizeof(Node);
367 293
368 memset(e, 0, sizeof(Node)); 294 memset(e, 0, sizeof(Node));
369 if (copy_from_user(buf, buffer, count) 295 if (copy_from_user(buf, buffer, count))
370 goto efault; 296 goto efault;
371 297
372 del = *p++; /* delimeter */ 298 del = *p++; /* delimeter */
373 299
374 pr_debug("register: delim: %#x {%c}\n" 300 pr_debug("register: delim: %#x {%c}\n", del, del);
375 301
376 /* Pad the buffer with the delim to si 302 /* Pad the buffer with the delim to simplify parsing below. */
377 memset(buf + count, del, 8); 303 memset(buf + count, del, 8);
378 304
379 /* Parse the 'name' field. */ 305 /* Parse the 'name' field. */
380 e->name = p; 306 e->name = p;
381 p = strchr(p, del); 307 p = strchr(p, del);
382 if (!p) 308 if (!p)
383 goto einval; 309 goto einval;
384 *p++ = '\0'; 310 *p++ = '\0';
385 if (!e->name[0] || 311 if (!e->name[0] ||
386 !strcmp(e->name, ".") || 312 !strcmp(e->name, ".") ||
387 !strcmp(e->name, "..") || 313 !strcmp(e->name, "..") ||
388 strchr(e->name, '/')) 314 strchr(e->name, '/'))
389 goto einval; 315 goto einval;
390 316
391 pr_debug("register: name: {%s}\n", e-> 317 pr_debug("register: name: {%s}\n", e->name);
392 318
393 /* Parse the 'type' field. */ 319 /* Parse the 'type' field. */
394 switch (*p++) { 320 switch (*p++) {
395 case 'E': 321 case 'E':
396 pr_debug("register: type: E (e 322 pr_debug("register: type: E (extension)\n");
397 e->flags = 1 << Enabled; 323 e->flags = 1 << Enabled;
398 break; 324 break;
399 case 'M': 325 case 'M':
400 pr_debug("register: type: M (m 326 pr_debug("register: type: M (magic)\n");
401 e->flags = (1 << Enabled) | (1 327 e->flags = (1 << Enabled) | (1 << Magic);
402 break; 328 break;
403 default: 329 default:
404 goto einval; 330 goto einval;
405 } 331 }
406 if (*p++ != del) 332 if (*p++ != del)
407 goto einval; 333 goto einval;
408 334
409 if (test_bit(Magic, &e->flags)) { 335 if (test_bit(Magic, &e->flags)) {
410 /* Handle the 'M' (magic) form 336 /* Handle the 'M' (magic) format. */
411 char *s; 337 char *s;
412 338
413 /* Parse the 'offset' field. * 339 /* Parse the 'offset' field. */
414 s = strchr(p, del); 340 s = strchr(p, del);
415 if (!s) 341 if (!s)
416 goto einval; 342 goto einval;
417 *s = '\0'; 343 *s = '\0';
418 if (p != s) { 344 if (p != s) {
419 int r = kstrtoint(p, 1 345 int r = kstrtoint(p, 10, &e->offset);
420 if (r != 0 || e->offse 346 if (r != 0 || e->offset < 0)
421 goto einval; 347 goto einval;
422 } 348 }
423 p = s; 349 p = s;
424 if (*p++) 350 if (*p++)
425 goto einval; 351 goto einval;
426 pr_debug("register: offset: %# 352 pr_debug("register: offset: %#x\n", e->offset);
427 353
428 /* Parse the 'magic' field. */ 354 /* Parse the 'magic' field. */
429 e->magic = p; 355 e->magic = p;
430 p = scanarg(p, del); 356 p = scanarg(p, del);
431 if (!p) 357 if (!p)
432 goto einval; 358 goto einval;
433 if (!e->magic[0]) 359 if (!e->magic[0])
434 goto einval; 360 goto einval;
435 if (USE_DEBUG) 361 if (USE_DEBUG)
436 print_hex_dump_bytes( 362 print_hex_dump_bytes(
437 KBUILD_MODNAME 363 KBUILD_MODNAME ": register: magic[raw]: ",
438 DUMP_PREFIX_NO 364 DUMP_PREFIX_NONE, e->magic, p - e->magic);
439 365
440 /* Parse the 'mask' field. */ 366 /* Parse the 'mask' field. */
441 e->mask = p; 367 e->mask = p;
442 p = scanarg(p, del); 368 p = scanarg(p, del);
443 if (!p) 369 if (!p)
444 goto einval; 370 goto einval;
445 if (!e->mask[0]) { 371 if (!e->mask[0]) {
446 e->mask = NULL; 372 e->mask = NULL;
447 pr_debug("register: m 373 pr_debug("register: mask[raw]: none\n");
448 } else if (USE_DEBUG) 374 } else if (USE_DEBUG)
449 print_hex_dump_bytes( 375 print_hex_dump_bytes(
450 KBUILD_MODNAME 376 KBUILD_MODNAME ": register: mask[raw]: ",
451 DUMP_PREFIX_NO 377 DUMP_PREFIX_NONE, e->mask, p - e->mask);
452 378
453 /* 379 /*
454 * Decode the magic & mask fie 380 * Decode the magic & mask fields.
455 * Note: while we might have a 381 * Note: while we might have accepted embedded NUL bytes from
456 * above, the unescape helpers 382 * above, the unescape helpers here will stop at the first one
457 * it encounters. 383 * it encounters.
458 */ 384 */
459 e->size = string_unescape_inpl 385 e->size = string_unescape_inplace(e->magic, UNESCAPE_HEX);
460 if (e->mask && 386 if (e->mask &&
461 string_unescape_inplace(e- 387 string_unescape_inplace(e->mask, UNESCAPE_HEX) != e->size)
462 goto einval; 388 goto einval;
463 if (e->size > BINPRM_BUF_SIZE 389 if (e->size > BINPRM_BUF_SIZE ||
464 BINPRM_BUF_SIZE - e->size 390 BINPRM_BUF_SIZE - e->size < e->offset)
465 goto einval; 391 goto einval;
466 pr_debug("register: magic/mask 392 pr_debug("register: magic/mask length: %i\n", e->size);
467 if (USE_DEBUG) { 393 if (USE_DEBUG) {
468 print_hex_dump_bytes( 394 print_hex_dump_bytes(
469 KBUILD_MODNAME 395 KBUILD_MODNAME ": register: magic[decoded]: ",
470 DUMP_PREFIX_NO 396 DUMP_PREFIX_NONE, e->magic, e->size);
471 397
472 if (e->mask) { 398 if (e->mask) {
473 int i; 399 int i;
474 char *masked = !! 400 char *masked = kmalloc(e->size, GFP_KERNEL);
475 401
476 print_hex_dump 402 print_hex_dump_bytes(
477 KBUILD 403 KBUILD_MODNAME ": register: mask[decoded]: ",
478 DUMP_P 404 DUMP_PREFIX_NONE, e->mask, e->size);
479 405
480 if (masked) { 406 if (masked) {
481 for (i 407 for (i = 0; i < e->size; ++i)
482 408 masked[i] = e->magic[i] & e->mask[i];
483 print_ 409 print_hex_dump_bytes(
484 410 KBUILD_MODNAME ": register: magic[masked]: ",
485 411 DUMP_PREFIX_NONE, masked, e->size);
486 412
487 kfree( 413 kfree(masked);
488 } 414 }
489 } 415 }
490 } 416 }
491 } else { 417 } else {
492 /* Handle the 'E' (extension) 418 /* Handle the 'E' (extension) format. */
493 419
494 /* Skip the 'offset' field. */ 420 /* Skip the 'offset' field. */
495 p = strchr(p, del); 421 p = strchr(p, del);
496 if (!p) 422 if (!p)
497 goto einval; 423 goto einval;
498 *p++ = '\0'; 424 *p++ = '\0';
499 425
500 /* Parse the 'magic' field. */ 426 /* Parse the 'magic' field. */
501 e->magic = p; 427 e->magic = p;
502 p = strchr(p, del); 428 p = strchr(p, del);
503 if (!p) 429 if (!p)
504 goto einval; 430 goto einval;
505 *p++ = '\0'; 431 *p++ = '\0';
506 if (!e->magic[0] || strchr(e-> 432 if (!e->magic[0] || strchr(e->magic, '/'))
507 goto einval; 433 goto einval;
508 pr_debug("register: extension: 434 pr_debug("register: extension: {%s}\n", e->magic);
509 435
510 /* Skip the 'mask' field. */ 436 /* Skip the 'mask' field. */
511 p = strchr(p, del); 437 p = strchr(p, del);
512 if (!p) 438 if (!p)
513 goto einval; 439 goto einval;
514 *p++ = '\0'; 440 *p++ = '\0';
515 } 441 }
516 442
517 /* Parse the 'interpreter' field. */ 443 /* Parse the 'interpreter' field. */
518 e->interpreter = p; 444 e->interpreter = p;
519 p = strchr(p, del); 445 p = strchr(p, del);
520 if (!p) 446 if (!p)
521 goto einval; 447 goto einval;
522 *p++ = '\0'; 448 *p++ = '\0';
523 if (!e->interpreter[0]) 449 if (!e->interpreter[0])
524 goto einval; 450 goto einval;
525 pr_debug("register: interpreter: {%s}\ 451 pr_debug("register: interpreter: {%s}\n", e->interpreter);
526 452
527 /* Parse the 'flags' field. */ 453 /* Parse the 'flags' field. */
528 p = check_special_flags(p, e); 454 p = check_special_flags(p, e);
529 if (*p == '\n') 455 if (*p == '\n')
530 p++; 456 p++;
531 if (p != buf + count) 457 if (p != buf + count)
532 goto einval; 458 goto einval;
533 459
534 return e; 460 return e;
535 461
536 out: 462 out:
537 return ERR_PTR(err); 463 return ERR_PTR(err);
538 464
539 efault: 465 efault:
540 kfree(e); 466 kfree(e);
541 return ERR_PTR(-EFAULT); 467 return ERR_PTR(-EFAULT);
542 einval: 468 einval:
543 kfree(e); 469 kfree(e);
544 return ERR_PTR(-EINVAL); 470 return ERR_PTR(-EINVAL);
545 } 471 }
546 472
547 /* 473 /*
548 * Set status of entry/binfmt_misc: 474 * Set status of entry/binfmt_misc:
549 * '1' enables, '' disables and '-1' clears en 475 * '1' enables, '' disables and '-1' clears entry/binfmt_misc
550 */ 476 */
551 static int parse_command(const char __user *bu 477 static int parse_command(const char __user *buffer, size_t count)
552 { 478 {
553 char s[4]; 479 char s[4];
554 480
555 if (count > 3) 481 if (count > 3)
556 return -EINVAL; 482 return -EINVAL;
557 if (copy_from_user(s, buffer, count)) 483 if (copy_from_user(s, buffer, count))
558 return -EFAULT; 484 return -EFAULT;
559 if (!count) 485 if (!count)
560 return 0; 486 return 0;
561 if (s[count - 1] == '\n') 487 if (s[count - 1] == '\n')
562 count--; 488 count--;
563 if (count == 1 && s[0] == '') 489 if (count == 1 && s[0] == '')
564 return 1; 490 return 1;
565 if (count == 1 && s[0] == '1') 491 if (count == 1 && s[0] == '1')
566 return 2; 492 return 2;
567 if (count == 2 && s[0] == '-' && s[1] 493 if (count == 2 && s[0] == '-' && s[1] == '1')
568 return 3; 494 return 3;
569 return -EINVAL; 495 return -EINVAL;
570 } 496 }
571 497
572 /* generic stuff */ 498 /* generic stuff */
573 499
574 static void entry_status(Node *e, char *page) 500 static void entry_status(Node *e, char *page)
575 { 501 {
576 char *dp = page; 502 char *dp = page;
577 const char *status = "disabled"; 503 const char *status = "disabled";
578 504
579 if (test_bit(Enabled, &e->flags)) 505 if (test_bit(Enabled, &e->flags))
580 status = "enabled"; 506 status = "enabled";
581 507
582 if (!VERBOSE_STATUS) { 508 if (!VERBOSE_STATUS) {
583 sprintf(page, "%s\n", status); 509 sprintf(page, "%s\n", status);
584 return; 510 return;
585 } 511 }
586 512
587 dp += sprintf(dp, "%s\ninterpreter %s\ 513 dp += sprintf(dp, "%s\ninterpreter %s\n", status, e->interpreter);
588 514
589 /* print the special flags */ 515 /* print the special flags */
590 dp += sprintf(dp, "flags: "); 516 dp += sprintf(dp, "flags: ");
591 if (e->flags & MISC_FMT_PRESERVE_ARGV0 517 if (e->flags & MISC_FMT_PRESERVE_ARGV0)
592 *dp++ = 'P'; 518 *dp++ = 'P';
593 if (e->flags & MISC_FMT_OPEN_BINARY) 519 if (e->flags & MISC_FMT_OPEN_BINARY)
594 *dp++ = 'O'; 520 *dp++ = 'O';
595 if (e->flags & MISC_FMT_CREDENTIALS) 521 if (e->flags & MISC_FMT_CREDENTIALS)
596 *dp++ = 'C'; 522 *dp++ = 'C';
597 if (e->flags & MISC_FMT_OPEN_FILE) 523 if (e->flags & MISC_FMT_OPEN_FILE)
598 *dp++ = 'F'; 524 *dp++ = 'F';
599 *dp++ = '\n'; 525 *dp++ = '\n';
600 526
601 if (!test_bit(Magic, &e->flags)) { 527 if (!test_bit(Magic, &e->flags)) {
602 sprintf(dp, "extension .%s\n", 528 sprintf(dp, "extension .%s\n", e->magic);
603 } else { 529 } else {
604 dp += sprintf(dp, "offset %i\n 530 dp += sprintf(dp, "offset %i\nmagic ", e->offset);
605 dp = bin2hex(dp, e->magic, e-> 531 dp = bin2hex(dp, e->magic, e->size);
606 if (e->mask) { 532 if (e->mask) {
607 dp += sprintf(dp, "\nm 533 dp += sprintf(dp, "\nmask ");
608 dp = bin2hex(dp, e->ma 534 dp = bin2hex(dp, e->mask, e->size);
609 } 535 }
610 *dp++ = '\n'; 536 *dp++ = '\n';
611 *dp = '\0'; 537 *dp = '\0';
612 } 538 }
613 } 539 }
614 540
615 static struct inode *bm_get_inode(struct super 541 static struct inode *bm_get_inode(struct super_block *sb, int mode)
616 { 542 {
617 struct inode *inode = new_inode(sb); 543 struct inode *inode = new_inode(sb);
618 544
619 if (inode) { 545 if (inode) {
620 inode->i_ino = get_next_ino(); 546 inode->i_ino = get_next_ino();
621 inode->i_mode = mode; 547 inode->i_mode = mode;
622 simple_inode_init_ts(inode); !! 548 inode->i_atime = inode->i_mtime = inode->i_ctime =
>> 549 current_time(inode);
623 } 550 }
624 return inode; 551 return inode;
625 } 552 }
626 553
627 /** <<
628 * i_binfmt_misc - retrieve struct binfmt_misc <<
629 * @inode: inode of the relevant binfmt_misc i <<
630 * <<
631 * This helper retrieves struct binfmt_misc fr <<
632 * be done without any memory barriers because <<
633 * user_ns->binfmt_misc is fully initialized. <<
634 * binfmt_misc mount was first created. <<
635 * <<
636 * Return: struct binfmt_misc of the relevant <<
637 */ <<
638 static struct binfmt_misc *i_binfmt_misc(struc <<
639 { <<
640 return inode->i_sb->s_user_ns->binfmt_ <<
641 } <<
642 <<
643 /** <<
644 * bm_evict_inode - cleanup data associated wi <<
645 * @inode: inode to which the data is attached <<
646 * <<
647 * Cleanup the binary type handler data associ <<
648 * entry is removed or the filesystem is unmou <<
649 * shutdown. <<
650 * <<
651 * If the ->evict call was not caused by a sup <<
652 * to remove the entry or all entries via bm_{ <<
653 * will have already been removed from the lis <<
654 * to make that explicit. <<
655 */ <<
656 static void bm_evict_inode(struct inode *inode 554 static void bm_evict_inode(struct inode *inode)
657 { 555 {
658 Node *e = inode->i_private; 556 Node *e = inode->i_private;
659 557
660 clear_inode(inode); !! 558 if (e && e->flags & MISC_FMT_OPEN_FILE)
661 !! 559 filp_close(e->interp_file, NULL);
662 if (e) { <<
663 struct binfmt_misc *misc; <<
664 560
665 misc = i_binfmt_misc(inode); !! 561 clear_inode(inode);
666 write_lock(&misc->entries_lock !! 562 kfree(e);
667 if (!list_empty(&e->list)) <<
668 list_del_init(&e->list <<
669 write_unlock(&misc->entries_lo <<
670 put_binfmt_handler(e); <<
671 } <<
672 } 563 }
673 564
674 /** !! 565 static void kill_node(Node *e)
675 * unlink_binfmt_dentry - remove the dentry fo <<
676 * @dentry: dentry associated with the binary <<
677 * <<
678 * Do the actual filesystem work to remove a d <<
679 * type handler. Since binfmt_misc only allows <<
680 * directly under the root dentry of the files <<
681 * indeed passed a dentry directly beneath the <<
682 * associated with the root dentry is locked, <<
683 * are asked to remove. <<
684 */ <<
685 static void unlink_binfmt_dentry(struct dentry <<
686 { 566 {
687 struct dentry *parent = dentry->d_pare !! 567 struct dentry *dentry;
688 struct inode *inode, *parent_inode; <<
689 <<
690 /* All entries are immediate descendan <<
691 if (WARN_ON_ONCE(dentry->d_sb->s_root <<
692 return; <<
693 <<
694 /* We only expect to be called on regu <<
695 inode = d_inode(dentry); <<
696 if (WARN_ON_ONCE(!S_ISREG(inode->i_mod <<
697 return; <<
698 <<
699 /* The parent inode must be locked. */ <<
700 parent_inode = d_inode(parent); <<
701 if (WARN_ON_ONCE(!inode_is_locked(pare <<
702 return; <<
703 <<
704 if (simple_positive(dentry)) { <<
705 dget(dentry); <<
706 simple_unlink(parent_inode, de <<
707 d_delete(dentry); <<
708 dput(dentry); <<
709 } <<
710 } <<
711 568
712 /** !! 569 write_lock(&entries_lock);
713 * remove_binfmt_handler - remove a binary typ <<
714 * @misc: handle to binfmt_misc instance <<
715 * @e: binary type handler to remove <<
716 * <<
717 * Remove a binary type handler from the list <<
718 * remove its associated dentry. This is calle <<
719 * binfmt_{entry,status}_write(). In the futur <<
720 * adding a proper ->unlink() method to binfmt <<
721 * to use writes to files in order to delete b <<
722 * worked for so long that it's not a pressing <<
723 */ <<
724 static void remove_binfmt_handler(struct binfm <<
725 { <<
726 write_lock(&misc->entries_lock); <<
727 list_del_init(&e->list); 570 list_del_init(&e->list);
728 write_unlock(&misc->entries_lock); !! 571 write_unlock(&entries_lock);
729 unlink_binfmt_dentry(e->dentry); !! 572
>> 573 dentry = e->dentry;
>> 574 drop_nlink(d_inode(dentry));
>> 575 d_drop(dentry);
>> 576 dput(dentry);
>> 577 simple_release_fs(&bm_mnt, &entry_count);
730 } 578 }
731 579
732 /* /<entry> */ 580 /* /<entry> */
733 581
734 static ssize_t 582 static ssize_t
735 bm_entry_read(struct file *file, char __user * 583 bm_entry_read(struct file *file, char __user *buf, size_t nbytes, loff_t *ppos)
736 { 584 {
737 Node *e = file_inode(file)->i_private; 585 Node *e = file_inode(file)->i_private;
738 ssize_t res; 586 ssize_t res;
739 char *page; 587 char *page;
740 588
741 page = (char *) __get_free_page(GFP_KE 589 page = (char *) __get_free_page(GFP_KERNEL);
742 if (!page) 590 if (!page)
743 return -ENOMEM; 591 return -ENOMEM;
744 592
745 entry_status(e, page); 593 entry_status(e, page);
746 594
747 res = simple_read_from_buffer(buf, nby 595 res = simple_read_from_buffer(buf, nbytes, ppos, page, strlen(page));
748 596
749 free_page((unsigned long) page); 597 free_page((unsigned long) page);
750 return res; 598 return res;
751 } 599 }
752 600
753 static ssize_t bm_entry_write(struct file *fil 601 static ssize_t bm_entry_write(struct file *file, const char __user *buffer,
754 size_t count, 602 size_t count, loff_t *ppos)
755 { 603 {
756 struct inode *inode = file_inode(file) !! 604 struct dentry *root;
757 Node *e = inode->i_private; !! 605 Node *e = file_inode(file)->i_private;
758 int res = parse_command(buffer, count) 606 int res = parse_command(buffer, count);
759 607
760 switch (res) { 608 switch (res) {
761 case 1: 609 case 1:
762 /* Disable this handler. */ 610 /* Disable this handler. */
763 clear_bit(Enabled, &e->flags); 611 clear_bit(Enabled, &e->flags);
764 break; 612 break;
765 case 2: 613 case 2:
766 /* Enable this handler. */ 614 /* Enable this handler. */
767 set_bit(Enabled, &e->flags); 615 set_bit(Enabled, &e->flags);
768 break; 616 break;
769 case 3: 617 case 3:
770 /* Delete this handler. */ 618 /* Delete this handler. */
771 inode = d_inode(inode->i_sb->s !! 619 root = file_inode(file)->i_sb->s_root;
772 inode_lock(inode); !! 620 inode_lock(d_inode(root));
773 621
774 /* <<
775 * In order to add new element <<
776 * via bm_{entry,register,stat <<
777 * root inode must be held. <<
778 * The lock is exclusive ensur <<
779 * modified. Only load_misc_bi <<
780 * read-only. So we only need <<
781 * actually remove the entry f <<
782 */ <<
783 if (!list_empty(&e->list)) 622 if (!list_empty(&e->list))
784 remove_binfmt_handler( !! 623 kill_node(e);
785 624
786 inode_unlock(inode); !! 625 inode_unlock(d_inode(root));
787 break; 626 break;
788 default: 627 default:
789 return res; 628 return res;
790 } 629 }
791 630
792 return count; 631 return count;
793 } 632 }
794 633
795 static const struct file_operations bm_entry_o 634 static const struct file_operations bm_entry_operations = {
796 .read = bm_entry_read, 635 .read = bm_entry_read,
797 .write = bm_entry_write, 636 .write = bm_entry_write,
798 .llseek = default_llseek, 637 .llseek = default_llseek,
799 }; 638 };
800 639
801 /* /register */ 640 /* /register */
802 641
803 static ssize_t bm_register_write(struct file * 642 static ssize_t bm_register_write(struct file *file, const char __user *buffer,
804 size_t count, l 643 size_t count, loff_t *ppos)
805 { 644 {
806 Node *e; 645 Node *e;
807 struct inode *inode; 646 struct inode *inode;
808 struct super_block *sb = file_inode(fi 647 struct super_block *sb = file_inode(file)->i_sb;
809 struct dentry *root = sb->s_root, *den 648 struct dentry *root = sb->s_root, *dentry;
810 struct binfmt_misc *misc; <<
811 int err = 0; 649 int err = 0;
812 struct file *f = NULL; 650 struct file *f = NULL;
813 651
814 e = create_entry(buffer, count); 652 e = create_entry(buffer, count);
815 653
816 if (IS_ERR(e)) 654 if (IS_ERR(e))
817 return PTR_ERR(e); 655 return PTR_ERR(e);
818 656
819 if (e->flags & MISC_FMT_OPEN_FILE) { 657 if (e->flags & MISC_FMT_OPEN_FILE) {
820 const struct cred *old_cred; <<
821 <<
822 /* <<
823 * Now that we support unprivi <<
824 * sure we use the credentials <<
825 * opened with to also open th <<
826 * didn't matter much as only <<
827 * the register file. <<
828 */ <<
829 old_cred = override_creds(file <<
830 f = open_exec(e->interpreter); 658 f = open_exec(e->interpreter);
831 revert_creds(old_cred); <<
832 if (IS_ERR(f)) { 659 if (IS_ERR(f)) {
833 pr_notice("register: f 660 pr_notice("register: failed to install interpreter file %s\n",
834 e->interprete 661 e->interpreter);
835 kfree(e); 662 kfree(e);
836 return PTR_ERR(f); 663 return PTR_ERR(f);
837 } 664 }
838 e->interp_file = f; 665 e->interp_file = f;
839 } 666 }
840 667
841 inode_lock(d_inode(root)); 668 inode_lock(d_inode(root));
842 dentry = lookup_one_len(e->name, root, 669 dentry = lookup_one_len(e->name, root, strlen(e->name));
843 err = PTR_ERR(dentry); 670 err = PTR_ERR(dentry);
844 if (IS_ERR(dentry)) 671 if (IS_ERR(dentry))
845 goto out; 672 goto out;
846 673
847 err = -EEXIST; 674 err = -EEXIST;
848 if (d_really_is_positive(dentry)) 675 if (d_really_is_positive(dentry))
849 goto out2; 676 goto out2;
850 677
851 inode = bm_get_inode(sb, S_IFREG | 064 678 inode = bm_get_inode(sb, S_IFREG | 0644);
852 679
853 err = -ENOMEM; 680 err = -ENOMEM;
854 if (!inode) 681 if (!inode)
855 goto out2; 682 goto out2;
856 683
857 refcount_set(&e->users, 1); !! 684 err = simple_pin_fs(&bm_fs_type, &bm_mnt, &entry_count);
>> 685 if (err) {
>> 686 iput(inode);
>> 687 inode = NULL;
>> 688 goto out2;
>> 689 }
>> 690
858 e->dentry = dget(dentry); 691 e->dentry = dget(dentry);
859 inode->i_private = e; 692 inode->i_private = e;
860 inode->i_fop = &bm_entry_operations; 693 inode->i_fop = &bm_entry_operations;
861 694
862 d_instantiate(dentry, inode); 695 d_instantiate(dentry, inode);
863 misc = i_binfmt_misc(inode); !! 696 write_lock(&entries_lock);
864 write_lock(&misc->entries_lock); !! 697 list_add(&e->list, &entries);
865 list_add(&e->list, &misc->entries); !! 698 write_unlock(&entries_lock);
866 write_unlock(&misc->entries_lock); <<
867 699
868 err = 0; 700 err = 0;
869 out2: 701 out2:
870 dput(dentry); 702 dput(dentry);
871 out: 703 out:
872 inode_unlock(d_inode(root)); 704 inode_unlock(d_inode(root));
873 705
874 if (err) { 706 if (err) {
875 if (f) 707 if (f)
876 filp_close(f, NULL); 708 filp_close(f, NULL);
877 kfree(e); 709 kfree(e);
878 return err; 710 return err;
879 } 711 }
880 return count; 712 return count;
881 } 713 }
882 714
883 static const struct file_operations bm_registe 715 static const struct file_operations bm_register_operations = {
884 .write = bm_register_write, 716 .write = bm_register_write,
885 .llseek = noop_llseek, 717 .llseek = noop_llseek,
886 }; 718 };
887 719
888 /* /status */ 720 /* /status */
889 721
890 static ssize_t 722 static ssize_t
891 bm_status_read(struct file *file, char __user 723 bm_status_read(struct file *file, char __user *buf, size_t nbytes, loff_t *ppos)
892 { 724 {
893 struct binfmt_misc *misc; !! 725 char *s = enabled ? "enabled\n" : "disabled\n";
894 char *s; <<
895 726
896 misc = i_binfmt_misc(file_inode(file)) <<
897 s = misc->enabled ? "enabled\n" : "dis <<
898 return simple_read_from_buffer(buf, nb 727 return simple_read_from_buffer(buf, nbytes, ppos, s, strlen(s));
899 } 728 }
900 729
901 static ssize_t bm_status_write(struct file *fi 730 static ssize_t bm_status_write(struct file *file, const char __user *buffer,
902 size_t count, loff_t *ppos) 731 size_t count, loff_t *ppos)
903 { 732 {
904 struct binfmt_misc *misc; <<
905 int res = parse_command(buffer, count) 733 int res = parse_command(buffer, count);
906 Node *e, *next; !! 734 struct dentry *root;
907 struct inode *inode; <<
908 735
909 misc = i_binfmt_misc(file_inode(file)) <<
910 switch (res) { 736 switch (res) {
911 case 1: 737 case 1:
912 /* Disable all handlers. */ 738 /* Disable all handlers. */
913 misc->enabled = false; !! 739 enabled = 0;
914 break; 740 break;
915 case 2: 741 case 2:
916 /* Enable all handlers. */ 742 /* Enable all handlers. */
917 misc->enabled = true; !! 743 enabled = 1;
918 break; 744 break;
919 case 3: 745 case 3:
920 /* Delete all handlers. */ 746 /* Delete all handlers. */
921 inode = d_inode(file_inode(fil !! 747 root = file_inode(file)->i_sb->s_root;
922 inode_lock(inode); !! 748 inode_lock(d_inode(root));
923 749
924 /* !! 750 while (!list_empty(&entries))
925 * In order to add new element !! 751 kill_node(list_first_entry(&entries, Node, list));
926 * via bm_{entry,register,stat <<
927 * root inode must be held. <<
928 * The lock is exclusive ensur <<
929 * modified. Only load_misc_bi <<
930 * read-only. So we only need <<
931 * actually remove the entry f <<
932 */ <<
933 list_for_each_entry_safe(e, ne <<
934 remove_binfmt_handler( <<
935 752
936 inode_unlock(inode); !! 753 inode_unlock(d_inode(root));
937 break; 754 break;
938 default: 755 default:
939 return res; 756 return res;
940 } 757 }
941 758
942 return count; 759 return count;
943 } 760 }
944 761
945 static const struct file_operations bm_status_ 762 static const struct file_operations bm_status_operations = {
946 .read = bm_status_read, 763 .read = bm_status_read,
947 .write = bm_status_write, 764 .write = bm_status_write,
948 .llseek = default_llseek, 765 .llseek = default_llseek,
949 }; 766 };
950 767
951 /* Superblock handling */ 768 /* Superblock handling */
952 769
953 static void bm_put_super(struct super_block *s <<
954 { <<
955 struct user_namespace *user_ns = sb->s <<
956 <<
957 sb->s_fs_info = NULL; <<
958 put_user_ns(user_ns); <<
959 } <<
960 <<
961 static const struct super_operations s_ops = { 770 static const struct super_operations s_ops = {
962 .statfs = simple_statfs, 771 .statfs = simple_statfs,
963 .evict_inode = bm_evict_inode, 772 .evict_inode = bm_evict_inode,
964 .put_super = bm_put_super, <<
965 }; 773 };
966 774
967 static int bm_fill_super(struct super_block *s 775 static int bm_fill_super(struct super_block *sb, struct fs_context *fc)
968 { 776 {
969 int err; 777 int err;
970 struct user_namespace *user_ns = sb->s <<
971 struct binfmt_misc *misc; <<
972 static const struct tree_descr bm_file 778 static const struct tree_descr bm_files[] = {
973 [2] = {"status", &bm_status_op 779 [2] = {"status", &bm_status_operations, S_IWUSR|S_IRUGO},
974 [3] = {"register", &bm_registe 780 [3] = {"register", &bm_register_operations, S_IWUSR},
975 /* last one */ {""} 781 /* last one */ {""}
976 }; 782 };
977 783
978 if (WARN_ON(user_ns != current_user_ns <<
979 return -EINVAL; <<
980 <<
981 /* <<
982 * Lazily allocate a new binfmt_misc i <<
983 * do it here during the first mount o <<
984 * waste memory for every user namespa <<
985 * more common to not mount a separate <<
986 * to mount one. <<
987 * <<
988 * While multiple superblocks can exis <<
989 * s_fs_info for binfmt_misc. Hence, t <<
990 * bm_fill_super() is called exactly o <<
991 * superblock for a userns is created. <<
992 * that when a binfmt_misc superblock <<
993 * a userns there's no one racing us. <<
994 * barriers when we dereference binfmt <<
995 */ <<
996 misc = user_ns->binfmt_misc; <<
997 if (!misc) { <<
998 /* <<
999 * If it turns out that most u <<
1000 * register their own binary <<
1001 * create their own separate <<
1002 * consider turning this into <<
1003 */ <<
1004 misc = kzalloc(sizeof(struct <<
1005 if (!misc) <<
1006 return -ENOMEM; <<
1007 <<
1008 INIT_LIST_HEAD(&misc->entries <<
1009 rwlock_init(&misc->entries_lo <<
1010 <<
1011 /* Pairs with smp_load_acquir <<
1012 smp_store_release(&user_ns->b <<
1013 } <<
1014 <<
1015 /* <<
1016 * When the binfmt_misc superblock fo <<
1017 * ->enabled might have been set to f <<
1018 * ->enabled again in put_super() as <<
1019 * binfmt_misc again. It also would b <<
1020 * ->put_super() is called we know th <<
1021 * bintfmt_misc mount is empty making <<
1022 * -ENOEXEC independent of whether -> <<
1023 * someone mounts binfmt_misc for the <<
1024 * reset ->enabled to true. <<
1025 */ <<
1026 misc->enabled = true; <<
1027 <<
1028 err = simple_fill_super(sb, BINFMTFS_ 784 err = simple_fill_super(sb, BINFMTFS_MAGIC, bm_files);
1029 if (!err) 785 if (!err)
1030 sb->s_op = &s_ops; 786 sb->s_op = &s_ops;
1031 return err; 787 return err;
1032 } 788 }
1033 789
1034 static void bm_free(struct fs_context *fc) <<
1035 { <<
1036 if (fc->s_fs_info) <<
1037 put_user_ns(fc->s_fs_info); <<
1038 } <<
1039 <<
1040 static int bm_get_tree(struct fs_context *fc) 790 static int bm_get_tree(struct fs_context *fc)
1041 { 791 {
1042 return get_tree_keyed(fc, bm_fill_sup !! 792 return get_tree_single(fc, bm_fill_super);
1043 } 793 }
1044 794
1045 static const struct fs_context_operations bm_ 795 static const struct fs_context_operations bm_context_ops = {
1046 .free = bm_free, <<
1047 .get_tree = bm_get_tree, 796 .get_tree = bm_get_tree,
1048 }; 797 };
1049 798
1050 static int bm_init_fs_context(struct fs_conte 799 static int bm_init_fs_context(struct fs_context *fc)
1051 { 800 {
1052 fc->ops = &bm_context_ops; 801 fc->ops = &bm_context_ops;
1053 return 0; 802 return 0;
1054 } 803 }
1055 804
1056 static struct linux_binfmt misc_format = { 805 static struct linux_binfmt misc_format = {
1057 .module = THIS_MODULE, 806 .module = THIS_MODULE,
1058 .load_binary = load_misc_binary, 807 .load_binary = load_misc_binary,
1059 }; 808 };
1060 809
1061 static struct file_system_type bm_fs_type = { 810 static struct file_system_type bm_fs_type = {
1062 .owner = THIS_MODULE, 811 .owner = THIS_MODULE,
1063 .name = "binfmt_misc", 812 .name = "binfmt_misc",
1064 .init_fs_context = bm_init_fs_context 813 .init_fs_context = bm_init_fs_context,
1065 .fs_flags = FS_USERNS_MOUNT, <<
1066 .kill_sb = kill_litter_super, 814 .kill_sb = kill_litter_super,
1067 }; 815 };
1068 MODULE_ALIAS_FS("binfmt_misc"); 816 MODULE_ALIAS_FS("binfmt_misc");
1069 817
1070 static int __init init_misc_binfmt(void) 818 static int __init init_misc_binfmt(void)
1071 { 819 {
1072 int err = register_filesystem(&bm_fs_ 820 int err = register_filesystem(&bm_fs_type);
1073 if (!err) 821 if (!err)
1074 insert_binfmt(&misc_format); 822 insert_binfmt(&misc_format);
1075 return err; 823 return err;
1076 } 824 }
1077 825
1078 static void __exit exit_misc_binfmt(void) 826 static void __exit exit_misc_binfmt(void)
1079 { 827 {
1080 unregister_binfmt(&misc_format); 828 unregister_binfmt(&misc_format);
1081 unregister_filesystem(&bm_fs_type); 829 unregister_filesystem(&bm_fs_type);
1082 } 830 }
1083 831
1084 core_initcall(init_misc_binfmt); 832 core_initcall(init_misc_binfmt);
1085 module_exit(exit_misc_binfmt); 833 module_exit(exit_misc_binfmt);
1086 MODULE_DESCRIPTION("Kernel support for miscel <<
1087 MODULE_LICENSE("GPL"); 834 MODULE_LICENSE("GPL");
1088 835
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.