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