aboutsummaryrefslogtreecommitdiffstats
path: root/lib
diff options
context:
space:
mode:
authorDavid S. Miller <davem@davemloft.net>2009-03-26 18:23:24 -0400
committerDavid S. Miller <davem@davemloft.net>2009-03-26 18:23:24 -0400
commit08abe18af1f78ee80c3c3a5ac47c3e0ae0beadf6 (patch)
tree2be39bf8942edca1bcec735145e144a682ca9cd3 /lib
parentf0de70f8bb56952f6e016a65a8a8d006918f5bf6 (diff)
parent0384e2959127a56d0640505d004d8dd92f9c29f5 (diff)
Merge branch 'master' of /home/davem/src/GIT/linux-2.6/
Conflicts: drivers/net/wimax/i2400m/usb-notif.c
Diffstat (limited to 'lib')
-rw-r--r--lib/Kconfig6
-rw-r--r--lib/Kconfig.debug74
-rw-r--r--lib/Makefile4
-rw-r--r--lib/bitmap.c16
-rw-r--r--lib/dynamic_debug.c769
-rw-r--r--lib/dynamic_printk.c414
-rw-r--r--lib/idr.c2
-rw-r--r--lib/kobject.c2
-rw-r--r--lib/kobject_uevent.c9
-rw-r--r--lib/nlattr.c475
10 files changed, 1313 insertions, 458 deletions
diff --git a/lib/Kconfig b/lib/Kconfig
index 03c2c24b9083..cea9e30a88ff 100644
--- a/lib/Kconfig
+++ b/lib/Kconfig
@@ -174,4 +174,10 @@ config DISABLE_OBSOLETE_CPUMASK_FUNCTIONS
174 bool "Disable obsolete cpumask functions" if DEBUG_PER_CPU_MAPS 174 bool "Disable obsolete cpumask functions" if DEBUG_PER_CPU_MAPS
175 depends on EXPERIMENTAL && BROKEN 175 depends on EXPERIMENTAL && BROKEN
176 176
177#
178# Netlink attribute parsing support is select'ed if needed
179#
180config NLATTR
181 bool
182
177endmenu 183endmenu
diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
index 1bcf9cd4baa0..8fee0a13ac58 100644
--- a/lib/Kconfig.debug
+++ b/lib/Kconfig.debug
@@ -847,60 +847,70 @@ config BUILD_DOCSRC
847 847
848 Say N if you are unsure. 848 Say N if you are unsure.
849 849
850config DYNAMIC_PRINTK_DEBUG 850config DYNAMIC_DEBUG
851 bool "Enable dynamic printk() call support" 851 bool "Enable dynamic printk() support"
852 default n 852 default n
853 depends on PRINTK 853 depends on PRINTK
854 depends on DEBUG_FS
854 select PRINTK_DEBUG 855 select PRINTK_DEBUG
855 help 856 help
856 857
857 Compiles debug level messages into the kernel, which would not 858 Compiles debug level messages into the kernel, which would not
858 otherwise be available at runtime. These messages can then be 859 otherwise be available at runtime. These messages can then be
859 enabled/disabled on a per module basis. This mechanism implicitly 860 enabled/disabled based on various levels of scope - per source file,
860 enables all pr_debug() and dev_dbg() calls. The impact of this 861 function, module, format string, and line number. This mechanism
861 compile option is a larger kernel text size of about 2%. 862 implicitly enables all pr_debug() and dev_dbg() calls. The impact of
863 this compile option is a larger kernel text size of about 2%.
862 864
863 Usage: 865 Usage:
864 866
865 Dynamic debugging is controlled by the debugfs file, 867 Dynamic debugging is controlled via the 'dynamic_debug/ddebug' file,
866 dynamic_printk/modules. This file contains a list of the modules that 868 which is contained in the 'debugfs' filesystem. Thus, the debugfs
867 can be enabled. The format of the file is the module name, followed 869 filesystem must first be mounted before making use of this feature.
868 by a set of flags that can be enabled. The first flag is always the 870 We refer the control file as: <debugfs>/dynamic_debug/ddebug. This
869 'enabled' flag. For example: 871 file contains a list of the debug statements that can be enabled. The
872 format for each line of the file is:
870 873
871 <module_name> <enabled=0/1> 874 filename:lineno [module]function flags format
872 .
873 .
874 .
875 875
876 <module_name> : Name of the module in which the debug call resides 876 filename : source file of the debug statement
877 <enabled=0/1> : whether the messages are enabled or not 877 lineno : line number of the debug statement
878 module : module that contains the debug statement
879 function : function that contains the debug statement
880 flags : 'p' means the line is turned 'on' for printing
881 format : the format used for the debug statement
878 882
879 From a live system: 883 From a live system:
880 884
881 snd_hda_intel enabled=0 885 nullarbor:~ # cat <debugfs>/dynamic_debug/ddebug
882 fixup enabled=0 886 # filename:lineno [module]function flags format
883 driver enabled=0 887 fs/aio.c:222 [aio]__put_ioctx - "__put_ioctx:\040freeing\040%p\012"
888 fs/aio.c:248 [aio]ioctx_alloc - "ENOMEM:\040nr_events\040too\040high\012"
889 fs/aio.c:1770 [aio]sys_io_cancel - "calling\040cancel\012"
884 890
885 Enable a module: 891 Example usage:
886 892
887 $echo "set enabled=1 <module_name>" > dynamic_printk/modules 893 // enable the message at line 1603 of file svcsock.c
894 nullarbor:~ # echo -n 'file svcsock.c line 1603 +p' >
895 <debugfs>/dynamic_debug/ddebug
888 896
889 Disable a module: 897 // enable all the messages in file svcsock.c
898 nullarbor:~ # echo -n 'file svcsock.c +p' >
899 <debugfs>/dynamic_debug/ddebug
890 900
891 $echo "set enabled=0 <module_name>" > dynamic_printk/modules 901 // enable all the messages in the NFS server module
902 nullarbor:~ # echo -n 'module nfsd +p' >
903 <debugfs>/dynamic_debug/ddebug
892 904
893 Enable all modules: 905 // enable all 12 messages in the function svc_process()
906 nullarbor:~ # echo -n 'func svc_process +p' >
907 <debugfs>/dynamic_debug/ddebug
894 908
895 $echo "set enabled=1 all" > dynamic_printk/modules 909 // disable all 12 messages in the function svc_process()
910 nullarbor:~ # echo -n 'func svc_process -p' >
911 <debugfs>/dynamic_debug/ddebug
896 912
897 Disable all modules: 913 See Documentation/dynamic-debug-howto.txt for additional information.
898
899 $echo "set enabled=0 all" > dynamic_printk/modules
900
901 Finally, passing "dynamic_printk" at the command line enables
902 debugging for all modules. This mode can be turned off via the above
903 disable command.
904 914
905source "samples/Kconfig" 915source "samples/Kconfig"
906 916
diff --git a/lib/Makefile b/lib/Makefile
index 32b0e64ded27..0dd9229fab7e 100644
--- a/lib/Makefile
+++ b/lib/Makefile
@@ -82,7 +82,9 @@ obj-$(CONFIG_HAVE_LMB) += lmb.o
82 82
83obj-$(CONFIG_HAVE_ARCH_TRACEHOOK) += syscall.o 83obj-$(CONFIG_HAVE_ARCH_TRACEHOOK) += syscall.o
84 84
85obj-$(CONFIG_DYNAMIC_PRINTK_DEBUG) += dynamic_printk.o 85obj-$(CONFIG_DYNAMIC_DEBUG) += dynamic_debug.o
86
87obj-$(CONFIG_NLATTR) += nlattr.o
86 88
87hostprogs-y := gen_crc32table 89hostprogs-y := gen_crc32table
88clean-files := crc32table.h 90clean-files := crc32table.h
diff --git a/lib/bitmap.c b/lib/bitmap.c
index 1338469ac849..35a1f7ff4149 100644
--- a/lib/bitmap.c
+++ b/lib/bitmap.c
@@ -948,15 +948,15 @@ done:
948 */ 948 */
949int bitmap_find_free_region(unsigned long *bitmap, int bits, int order) 949int bitmap_find_free_region(unsigned long *bitmap, int bits, int order)
950{ 950{
951 int pos; /* scans bitmap by regions of size order */ 951 int pos, end; /* scans bitmap by regions of size order */
952 952
953 for (pos = 0; pos < bits; pos += (1 << order)) 953 for (pos = 0 ; (end = pos + (1 << order)) <= bits; pos = end) {
954 if (__reg_op(bitmap, pos, order, REG_OP_ISFREE)) 954 if (!__reg_op(bitmap, pos, order, REG_OP_ISFREE))
955 break; 955 continue;
956 if (pos == bits) 956 __reg_op(bitmap, pos, order, REG_OP_ALLOC);
957 return -ENOMEM; 957 return pos;
958 __reg_op(bitmap, pos, order, REG_OP_ALLOC); 958 }
959 return pos; 959 return -ENOMEM;
960} 960}
961EXPORT_SYMBOL(bitmap_find_free_region); 961EXPORT_SYMBOL(bitmap_find_free_region);
962 962
diff --git a/lib/dynamic_debug.c b/lib/dynamic_debug.c
new file mode 100644
index 000000000000..833139ce1e22
--- /dev/null
+++ b/lib/dynamic_debug.c
@@ -0,0 +1,769 @@
1/*
2 * lib/dynamic_debug.c
3 *
4 * make pr_debug()/dev_dbg() calls runtime configurable based upon their
5 * source module.
6 *
7 * Copyright (C) 2008 Jason Baron <jbaron@redhat.com>
8 * By Greg Banks <gnb@melbourne.sgi.com>
9 * Copyright (c) 2008 Silicon Graphics Inc. All Rights Reserved.
10 */
11
12#include <linux/kernel.h>
13#include <linux/module.h>
14#include <linux/moduleparam.h>
15#include <linux/kallsyms.h>
16#include <linux/version.h>
17#include <linux/types.h>
18#include <linux/mutex.h>
19#include <linux/proc_fs.h>
20#include <linux/seq_file.h>
21#include <linux/list.h>
22#include <linux/sysctl.h>
23#include <linux/ctype.h>
24#include <linux/uaccess.h>
25#include <linux/dynamic_debug.h>
26#include <linux/debugfs.h>
27
28extern struct _ddebug __start___verbose[];
29extern struct _ddebug __stop___verbose[];
30
31/* dynamic_debug_enabled, and dynamic_debug_enabled2 are bitmasks in which
32 * bit n is set to 1 if any modname hashes into the bucket n, 0 otherwise. They
33 * use independent hash functions, to reduce the chance of false positives.
34 */
35long long dynamic_debug_enabled;
36EXPORT_SYMBOL_GPL(dynamic_debug_enabled);
37long long dynamic_debug_enabled2;
38EXPORT_SYMBOL_GPL(dynamic_debug_enabled2);
39
40struct ddebug_table {
41 struct list_head link;
42 char *mod_name;
43 unsigned int num_ddebugs;
44 unsigned int num_enabled;
45 struct _ddebug *ddebugs;
46};
47
48struct ddebug_query {
49 const char *filename;
50 const char *module;
51 const char *function;
52 const char *format;
53 unsigned int first_lineno, last_lineno;
54};
55
56struct ddebug_iter {
57 struct ddebug_table *table;
58 unsigned int idx;
59};
60
61static DEFINE_MUTEX(ddebug_lock);
62static LIST_HEAD(ddebug_tables);
63static int verbose = 0;
64
65/* Return the last part of a pathname */
66static inline const char *basename(const char *path)
67{
68 const char *tail = strrchr(path, '/');
69 return tail ? tail+1 : path;
70}
71
72/* format a string into buf[] which describes the _ddebug's flags */
73static char *ddebug_describe_flags(struct _ddebug *dp, char *buf,
74 size_t maxlen)
75{
76 char *p = buf;
77
78 BUG_ON(maxlen < 4);
79 if (dp->flags & _DPRINTK_FLAGS_PRINT)
80 *p++ = 'p';
81 if (p == buf)
82 *p++ = '-';
83 *p = '\0';
84
85 return buf;
86}
87
88/*
89 * must be called with ddebug_lock held
90 */
91
92static int disabled_hash(char hash, bool first_table)
93{
94 struct ddebug_table *dt;
95 char table_hash_value;
96
97 list_for_each_entry(dt, &ddebug_tables, link) {
98 if (first_table)
99 table_hash_value = dt->ddebugs->primary_hash;
100 else
101 table_hash_value = dt->ddebugs->secondary_hash;
102 if (dt->num_enabled && (hash == table_hash_value))
103 return 0;
104 }
105 return 1;
106}
107
108/*
109 * Search the tables for _ddebug's which match the given
110 * `query' and apply the `flags' and `mask' to them. Tells
111 * the user which ddebug's were changed, or whether none
112 * were matched.
113 */
114static void ddebug_change(const struct ddebug_query *query,
115 unsigned int flags, unsigned int mask)
116{
117 int i;
118 struct ddebug_table *dt;
119 unsigned int newflags;
120 unsigned int nfound = 0;
121 char flagbuf[8];
122
123 /* search for matching ddebugs */
124 mutex_lock(&ddebug_lock);
125 list_for_each_entry(dt, &ddebug_tables, link) {
126
127 /* match against the module name */
128 if (query->module != NULL &&
129 strcmp(query->module, dt->mod_name))
130 continue;
131
132 for (i = 0 ; i < dt->num_ddebugs ; i++) {
133 struct _ddebug *dp = &dt->ddebugs[i];
134
135 /* match against the source filename */
136 if (query->filename != NULL &&
137 strcmp(query->filename, dp->filename) &&
138 strcmp(query->filename, basename(dp->filename)))
139 continue;
140
141 /* match against the function */
142 if (query->function != NULL &&
143 strcmp(query->function, dp->function))
144 continue;
145
146 /* match against the format */
147 if (query->format != NULL &&
148 strstr(dp->format, query->format) == NULL)
149 continue;
150
151 /* match against the line number range */
152 if (query->first_lineno &&
153 dp->lineno < query->first_lineno)
154 continue;
155 if (query->last_lineno &&
156 dp->lineno > query->last_lineno)
157 continue;
158
159 nfound++;
160
161 newflags = (dp->flags & mask) | flags;
162 if (newflags == dp->flags)
163 continue;
164
165 if (!newflags)
166 dt->num_enabled--;
167 else if (!dp-flags)
168 dt->num_enabled++;
169 dp->flags = newflags;
170 if (newflags) {
171 dynamic_debug_enabled |=
172 (1LL << dp->primary_hash);
173 dynamic_debug_enabled2 |=
174 (1LL << dp->secondary_hash);
175 } else {
176 if (disabled_hash(dp->primary_hash, true))
177 dynamic_debug_enabled &=
178 ~(1LL << dp->primary_hash);
179 if (disabled_hash(dp->secondary_hash, false))
180 dynamic_debug_enabled2 &=
181 ~(1LL << dp->secondary_hash);
182 }
183 if (verbose)
184 printk(KERN_INFO
185 "ddebug: changed %s:%d [%s]%s %s\n",
186 dp->filename, dp->lineno,
187 dt->mod_name, dp->function,
188 ddebug_describe_flags(dp, flagbuf,
189 sizeof(flagbuf)));
190 }
191 }
192 mutex_unlock(&ddebug_lock);
193
194 if (!nfound && verbose)
195 printk(KERN_INFO "ddebug: no matches for query\n");
196}
197
198/*
199 * Split the buffer `buf' into space-separated words.
200 * Handles simple " and ' quoting, i.e. without nested,
201 * embedded or escaped \". Return the number of words
202 * or <0 on error.
203 */
204static int ddebug_tokenize(char *buf, char *words[], int maxwords)
205{
206 int nwords = 0;
207
208 while (*buf) {
209 char *end;
210
211 /* Skip leading whitespace */
212 while (*buf && isspace(*buf))
213 buf++;
214 if (!*buf)
215 break; /* oh, it was trailing whitespace */
216
217 /* Run `end' over a word, either whitespace separated or quoted */
218 if (*buf == '"' || *buf == '\'') {
219 int quote = *buf++;
220 for (end = buf ; *end && *end != quote ; end++)
221 ;
222 if (!*end)
223 return -EINVAL; /* unclosed quote */
224 } else {
225 for (end = buf ; *end && !isspace(*end) ; end++)
226 ;
227 BUG_ON(end == buf);
228 }
229 /* Here `buf' is the start of the word, `end' is one past the end */
230
231 if (nwords == maxwords)
232 return -EINVAL; /* ran out of words[] before bytes */
233 if (*end)
234 *end++ = '\0'; /* terminate the word */
235 words[nwords++] = buf;
236 buf = end;
237 }
238
239 if (verbose) {
240 int i;
241 printk(KERN_INFO "%s: split into words:", __func__);
242 for (i = 0 ; i < nwords ; i++)
243 printk(" \"%s\"", words[i]);
244 printk("\n");
245 }
246
247 return nwords;
248}
249
250/*
251 * Parse a single line number. Note that the empty string ""
252 * is treated as a special case and converted to zero, which
253 * is later treated as a "don't care" value.
254 */
255static inline int parse_lineno(const char *str, unsigned int *val)
256{
257 char *end = NULL;
258 BUG_ON(str == NULL);
259 if (*str == '\0') {
260 *val = 0;
261 return 0;
262 }
263 *val = simple_strtoul(str, &end, 10);
264 return end == NULL || end == str || *end != '\0' ? -EINVAL : 0;
265}
266
267/*
268 * Undo octal escaping in a string, inplace. This is useful to
269 * allow the user to express a query which matches a format
270 * containing embedded spaces.
271 */
272#define isodigit(c) ((c) >= '0' && (c) <= '7')
273static char *unescape(char *str)
274{
275 char *in = str;
276 char *out = str;
277
278 while (*in) {
279 if (*in == '\\') {
280 if (in[1] == '\\') {
281 *out++ = '\\';
282 in += 2;
283 continue;
284 } else if (in[1] == 't') {
285 *out++ = '\t';
286 in += 2;
287 continue;
288 } else if (in[1] == 'n') {
289 *out++ = '\n';
290 in += 2;
291 continue;
292 } else if (isodigit(in[1]) &&
293 isodigit(in[2]) &&
294 isodigit(in[3])) {
295 *out++ = ((in[1] - '0')<<6) |
296 ((in[2] - '0')<<3) |
297 (in[3] - '0');
298 in += 4;
299 continue;
300 }
301 }
302 *out++ = *in++;
303 }
304 *out = '\0';
305
306 return str;
307}
308
309/*
310 * Parse words[] as a ddebug query specification, which is a series
311 * of (keyword, value) pairs chosen from these possibilities:
312 *
313 * func <function-name>
314 * file <full-pathname>
315 * file <base-filename>
316 * module <module-name>
317 * format <escaped-string-to-find-in-format>
318 * line <lineno>
319 * line <first-lineno>-<last-lineno> // where either may be empty
320 */
321static int ddebug_parse_query(char *words[], int nwords,
322 struct ddebug_query *query)
323{
324 unsigned int i;
325
326 /* check we have an even number of words */
327 if (nwords % 2 != 0)
328 return -EINVAL;
329 memset(query, 0, sizeof(*query));
330
331 for (i = 0 ; i < nwords ; i += 2) {
332 if (!strcmp(words[i], "func"))
333 query->function = words[i+1];
334 else if (!strcmp(words[i], "file"))
335 query->filename = words[i+1];
336 else if (!strcmp(words[i], "module"))
337 query->module = words[i+1];
338 else if (!strcmp(words[i], "format"))
339 query->format = unescape(words[i+1]);
340 else if (!strcmp(words[i], "line")) {
341 char *first = words[i+1];
342 char *last = strchr(first, '-');
343 if (last)
344 *last++ = '\0';
345 if (parse_lineno(first, &query->first_lineno) < 0)
346 return -EINVAL;
347 if (last != NULL) {
348 /* range <first>-<last> */
349 if (parse_lineno(last, &query->last_lineno) < 0)
350 return -EINVAL;
351 } else {
352 query->last_lineno = query->first_lineno;
353 }
354 } else {
355 if (verbose)
356 printk(KERN_ERR "%s: unknown keyword \"%s\"\n",
357 __func__, words[i]);
358 return -EINVAL;
359 }
360 }
361
362 if (verbose)
363 printk(KERN_INFO "%s: q->function=\"%s\" q->filename=\"%s\" "
364 "q->module=\"%s\" q->format=\"%s\" q->lineno=%u-%u\n",
365 __func__, query->function, query->filename,
366 query->module, query->format, query->first_lineno,
367 query->last_lineno);
368
369 return 0;
370}
371
372/*
373 * Parse `str' as a flags specification, format [-+=][p]+.
374 * Sets up *maskp and *flagsp to be used when changing the
375 * flags fields of matched _ddebug's. Returns 0 on success
376 * or <0 on error.
377 */
378static int ddebug_parse_flags(const char *str, unsigned int *flagsp,
379 unsigned int *maskp)
380{
381 unsigned flags = 0;
382 int op = '=';
383
384 switch (*str) {
385 case '+':
386 case '-':
387 case '=':
388 op = *str++;
389 break;
390 default:
391 return -EINVAL;
392 }
393 if (verbose)
394 printk(KERN_INFO "%s: op='%c'\n", __func__, op);
395
396 for ( ; *str ; ++str) {
397 switch (*str) {
398 case 'p':
399 flags |= _DPRINTK_FLAGS_PRINT;
400 break;
401 default:
402 return -EINVAL;
403 }
404 }
405 if (flags == 0)
406 return -EINVAL;
407 if (verbose)
408 printk(KERN_INFO "%s: flags=0x%x\n", __func__, flags);
409
410 /* calculate final *flagsp, *maskp according to mask and op */
411 switch (op) {
412 case '=':
413 *maskp = 0;
414 *flagsp = flags;
415 break;
416 case '+':
417 *maskp = ~0U;
418 *flagsp = flags;
419 break;
420 case '-':
421 *maskp = ~flags;
422 *flagsp = 0;
423 break;
424 }
425 if (verbose)
426 printk(KERN_INFO "%s: *flagsp=0x%x *maskp=0x%x\n",
427 __func__, *flagsp, *maskp);
428 return 0;
429}
430
431/*
432 * File_ops->write method for <debugfs>/dynamic_debug/conrol. Gathers the
433 * command text from userspace, parses and executes it.
434 */
435static ssize_t ddebug_proc_write(struct file *file, const char __user *ubuf,
436 size_t len, loff_t *offp)
437{
438 unsigned int flags = 0, mask = 0;
439 struct ddebug_query query;
440#define MAXWORDS 9
441 int nwords;
442 char *words[MAXWORDS];
443 char tmpbuf[256];
444
445 if (len == 0)
446 return 0;
447 /* we don't check *offp -- multiple writes() are allowed */
448 if (len > sizeof(tmpbuf)-1)
449 return -E2BIG;
450 if (copy_from_user(tmpbuf, ubuf, len))
451 return -EFAULT;
452 tmpbuf[len] = '\0';
453 if (verbose)
454 printk(KERN_INFO "%s: read %d bytes from userspace\n",
455 __func__, (int)len);
456
457 nwords = ddebug_tokenize(tmpbuf, words, MAXWORDS);
458 if (nwords < 0)
459 return -EINVAL;
460 if (ddebug_parse_query(words, nwords-1, &query))
461 return -EINVAL;
462 if (ddebug_parse_flags(words[nwords-1], &flags, &mask))
463 return -EINVAL;
464
465 /* actually go and implement the change */
466 ddebug_change(&query, flags, mask);
467
468 *offp += len;
469 return len;
470}
471
472/*
473 * Set the iterator to point to the first _ddebug object
474 * and return a pointer to that first object. Returns
475 * NULL if there are no _ddebugs at all.
476 */
477static struct _ddebug *ddebug_iter_first(struct ddebug_iter *iter)
478{
479 if (list_empty(&ddebug_tables)) {
480 iter->table = NULL;
481 iter->idx = 0;
482 return NULL;
483 }
484 iter->table = list_entry(ddebug_tables.next,
485 struct ddebug_table, link);
486 iter->idx = 0;
487 return &iter->table->ddebugs[iter->idx];
488}
489
490/*
491 * Advance the iterator to point to the next _ddebug
492 * object from the one the iterator currently points at,
493 * and returns a pointer to the new _ddebug. Returns
494 * NULL if the iterator has seen all the _ddebugs.
495 */
496static struct _ddebug *ddebug_iter_next(struct ddebug_iter *iter)
497{
498 if (iter->table == NULL)
499 return NULL;
500 if (++iter->idx == iter->table->num_ddebugs) {
501 /* iterate to next table */
502 iter->idx = 0;
503 if (list_is_last(&iter->table->link, &ddebug_tables)) {
504 iter->table = NULL;
505 return NULL;
506 }
507 iter->table = list_entry(iter->table->link.next,
508 struct ddebug_table, link);
509 }
510 return &iter->table->ddebugs[iter->idx];
511}
512
513/*
514 * Seq_ops start method. Called at the start of every
515 * read() call from userspace. Takes the ddebug_lock and
516 * seeks the seq_file's iterator to the given position.
517 */
518static void *ddebug_proc_start(struct seq_file *m, loff_t *pos)
519{
520 struct ddebug_iter *iter = m->private;
521 struct _ddebug *dp;
522 int n = *pos;
523
524 if (verbose)
525 printk(KERN_INFO "%s: called m=%p *pos=%lld\n",
526 __func__, m, (unsigned long long)*pos);
527
528 mutex_lock(&ddebug_lock);
529
530 if (!n)
531 return SEQ_START_TOKEN;
532 if (n < 0)
533 return NULL;
534 dp = ddebug_iter_first(iter);
535 while (dp != NULL && --n > 0)
536 dp = ddebug_iter_next(iter);
537 return dp;
538}
539
540/*
541 * Seq_ops next method. Called several times within a read()
542 * call from userspace, with ddebug_lock held. Walks to the
543 * next _ddebug object with a special case for the header line.
544 */
545static void *ddebug_proc_next(struct seq_file *m, void *p, loff_t *pos)
546{
547 struct ddebug_iter *iter = m->private;
548 struct _ddebug *dp;
549
550 if (verbose)
551 printk(KERN_INFO "%s: called m=%p p=%p *pos=%lld\n",
552 __func__, m, p, (unsigned long long)*pos);
553
554 if (p == SEQ_START_TOKEN)
555 dp = ddebug_iter_first(iter);
556 else
557 dp = ddebug_iter_next(iter);
558 ++*pos;
559 return dp;
560}
561
562/*
563 * Seq_ops show method. Called several times within a read()
564 * call from userspace, with ddebug_lock held. Formats the
565 * current _ddebug as a single human-readable line, with a
566 * special case for the header line.
567 */
568static int ddebug_proc_show(struct seq_file *m, void *p)
569{
570 struct ddebug_iter *iter = m->private;
571 struct _ddebug *dp = p;
572 char flagsbuf[8];
573
574 if (verbose)
575 printk(KERN_INFO "%s: called m=%p p=%p\n",
576 __func__, m, p);
577
578 if (p == SEQ_START_TOKEN) {
579 seq_puts(m,
580 "# filename:lineno [module]function flags format\n");
581 return 0;
582 }
583
584 seq_printf(m, "%s:%u [%s]%s %s \"",
585 dp->filename, dp->lineno,
586 iter->table->mod_name, dp->function,
587 ddebug_describe_flags(dp, flagsbuf, sizeof(flagsbuf)));
588 seq_escape(m, dp->format, "\t\r\n\"");
589 seq_puts(m, "\"\n");
590
591 return 0;
592}
593
594/*
595 * Seq_ops stop method. Called at the end of each read()
596 * call from userspace. Drops ddebug_lock.
597 */
598static void ddebug_proc_stop(struct seq_file *m, void *p)
599{
600 if (verbose)
601 printk(KERN_INFO "%s: called m=%p p=%p\n",
602 __func__, m, p);
603 mutex_unlock(&ddebug_lock);
604}
605
606static const struct seq_operations ddebug_proc_seqops = {
607 .start = ddebug_proc_start,
608 .next = ddebug_proc_next,
609 .show = ddebug_proc_show,
610 .stop = ddebug_proc_stop
611};
612
613/*
614 * File_ops->open method for <debugfs>/dynamic_debug/control. Does the seq_file
615 * setup dance, and also creates an iterator to walk the _ddebugs.
616 * Note that we create a seq_file always, even for O_WRONLY files
617 * where it's not needed, as doing so simplifies the ->release method.
618 */
619static int ddebug_proc_open(struct inode *inode, struct file *file)
620{
621 struct ddebug_iter *iter;
622 int err;
623
624 if (verbose)
625 printk(KERN_INFO "%s: called\n", __func__);
626
627 iter = kzalloc(sizeof(*iter), GFP_KERNEL);
628 if (iter == NULL)
629 return -ENOMEM;
630
631 err = seq_open(file, &ddebug_proc_seqops);
632 if (err) {
633 kfree(iter);
634 return err;
635 }
636 ((struct seq_file *) file->private_data)->private = iter;
637 return 0;
638}
639
640static const struct file_operations ddebug_proc_fops = {
641 .owner = THIS_MODULE,
642 .open = ddebug_proc_open,
643 .read = seq_read,
644 .llseek = seq_lseek,
645 .release = seq_release_private,
646 .write = ddebug_proc_write
647};
648
649/*
650 * Allocate a new ddebug_table for the given module
651 * and add it to the global list.
652 */
653int ddebug_add_module(struct _ddebug *tab, unsigned int n,
654 const char *name)
655{
656 struct ddebug_table *dt;
657 char *new_name;
658
659 dt = kzalloc(sizeof(*dt), GFP_KERNEL);
660 if (dt == NULL)
661 return -ENOMEM;
662 new_name = kstrdup(name, GFP_KERNEL);
663 if (new_name == NULL) {
664 kfree(dt);
665 return -ENOMEM;
666 }
667 dt->mod_name = new_name;
668 dt->num_ddebugs = n;
669 dt->num_enabled = 0;
670 dt->ddebugs = tab;
671
672 mutex_lock(&ddebug_lock);
673 list_add_tail(&dt->link, &ddebug_tables);
674 mutex_unlock(&ddebug_lock);
675
676 if (verbose)
677 printk(KERN_INFO "%u debug prints in module %s\n",
678 n, dt->mod_name);
679 return 0;
680}
681EXPORT_SYMBOL_GPL(ddebug_add_module);
682
683static void ddebug_table_free(struct ddebug_table *dt)
684{
685 list_del_init(&dt->link);
686 kfree(dt->mod_name);
687 kfree(dt);
688}
689
690/*
691 * Called in response to a module being unloaded. Removes
692 * any ddebug_table's which point at the module.
693 */
694int ddebug_remove_module(char *mod_name)
695{
696 struct ddebug_table *dt, *nextdt;
697 int ret = -ENOENT;
698
699 if (verbose)
700 printk(KERN_INFO "%s: removing module \"%s\"\n",
701 __func__, mod_name);
702
703 mutex_lock(&ddebug_lock);
704 list_for_each_entry_safe(dt, nextdt, &ddebug_tables, link) {
705 if (!strcmp(dt->mod_name, mod_name)) {
706 ddebug_table_free(dt);
707 ret = 0;
708 }
709 }
710 mutex_unlock(&ddebug_lock);
711 return ret;
712}
713EXPORT_SYMBOL_GPL(ddebug_remove_module);
714
715static void ddebug_remove_all_tables(void)
716{
717 mutex_lock(&ddebug_lock);
718 while (!list_empty(&ddebug_tables)) {
719 struct ddebug_table *dt = list_entry(ddebug_tables.next,
720 struct ddebug_table,
721 link);
722 ddebug_table_free(dt);
723 }
724 mutex_unlock(&ddebug_lock);
725}
726
727static int __init dynamic_debug_init(void)
728{
729 struct dentry *dir, *file;
730 struct _ddebug *iter, *iter_start;
731 const char *modname = NULL;
732 int ret = 0;
733 int n = 0;
734
735 dir = debugfs_create_dir("dynamic_debug", NULL);
736 if (!dir)
737 return -ENOMEM;
738 file = debugfs_create_file("control", 0644, dir, NULL,
739 &ddebug_proc_fops);
740 if (!file) {
741 debugfs_remove(dir);
742 return -ENOMEM;
743 }
744 if (__start___verbose != __stop___verbose) {
745 iter = __start___verbose;
746 modname = iter->modname;
747 iter_start = iter;
748 for (; iter < __stop___verbose; iter++) {
749 if (strcmp(modname, iter->modname)) {
750 ret = ddebug_add_module(iter_start, n, modname);
751 if (ret)
752 goto out_free;
753 n = 0;
754 modname = iter->modname;
755 iter_start = iter;
756 }
757 n++;
758 }
759 ret = ddebug_add_module(iter_start, n, modname);
760 }
761out_free:
762 if (ret) {
763 ddebug_remove_all_tables();
764 debugfs_remove(dir);
765 debugfs_remove(file);
766 }
767 return 0;
768}
769module_init(dynamic_debug_init);
diff --git a/lib/dynamic_printk.c b/lib/dynamic_printk.c
deleted file mode 100644
index 165a19763dc9..000000000000
--- a/lib/dynamic_printk.c
+++ /dev/null
@@ -1,414 +0,0 @@
1/*
2 * lib/dynamic_printk.c
3 *
4 * make pr_debug()/dev_dbg() calls runtime configurable based upon their
5 * their source module.
6 *
7 * Copyright (C) 2008 Red Hat, Inc., Jason Baron <jbaron@redhat.com>
8 */
9
10#include <linux/kernel.h>
11#include <linux/module.h>
12#include <linux/uaccess.h>
13#include <linux/seq_file.h>
14#include <linux/debugfs.h>
15#include <linux/fs.h>
16
17extern struct mod_debug __start___verbose[];
18extern struct mod_debug __stop___verbose[];
19
20struct debug_name {
21 struct hlist_node hlist;
22 struct hlist_node hlist2;
23 int hash1;
24 int hash2;
25 char *name;
26 int enable;
27 int type;
28};
29
30static int nr_entries;
31static int num_enabled;
32int dynamic_enabled = DYNAMIC_ENABLED_NONE;
33static struct hlist_head module_table[DEBUG_HASH_TABLE_SIZE] =
34 { [0 ... DEBUG_HASH_TABLE_SIZE-1] = HLIST_HEAD_INIT };
35static struct hlist_head module_table2[DEBUG_HASH_TABLE_SIZE] =
36 { [0 ... DEBUG_HASH_TABLE_SIZE-1] = HLIST_HEAD_INIT };
37static DECLARE_MUTEX(debug_list_mutex);
38
39/* dynamic_printk_enabled, and dynamic_printk_enabled2 are bitmasks in which
40 * bit n is set to 1 if any modname hashes into the bucket n, 0 otherwise. They
41 * use independent hash functions, to reduce the chance of false positives.
42 */
43long long dynamic_printk_enabled;
44EXPORT_SYMBOL_GPL(dynamic_printk_enabled);
45long long dynamic_printk_enabled2;
46EXPORT_SYMBOL_GPL(dynamic_printk_enabled2);
47
48/* returns the debug module pointer. */
49static struct debug_name *find_debug_module(char *module_name)
50{
51 int i;
52 struct hlist_head *head;
53 struct hlist_node *node;
54 struct debug_name *element;
55
56 element = NULL;
57 for (i = 0; i < DEBUG_HASH_TABLE_SIZE; i++) {
58 head = &module_table[i];
59 hlist_for_each_entry_rcu(element, node, head, hlist)
60 if (!strcmp(element->name, module_name))
61 return element;
62 }
63 return NULL;
64}
65
66/* returns the debug module pointer. */
67static struct debug_name *find_debug_module_hash(char *module_name, int hash)
68{
69 struct hlist_head *head;
70 struct hlist_node *node;
71 struct debug_name *element;
72
73 element = NULL;
74 head = &module_table[hash];
75 hlist_for_each_entry_rcu(element, node, head, hlist)
76 if (!strcmp(element->name, module_name))
77 return element;
78 return NULL;
79}
80
81/* caller must hold mutex*/
82static int __add_debug_module(char *mod_name, int hash, int hash2)
83{
84 struct debug_name *new;
85 char *module_name;
86 int ret = 0;
87
88 if (find_debug_module(mod_name)) {
89 ret = -EINVAL;
90 goto out;
91 }
92 module_name = kmalloc(strlen(mod_name) + 1, GFP_KERNEL);
93 if (!module_name) {
94 ret = -ENOMEM;
95 goto out;
96 }
97 module_name = strcpy(module_name, mod_name);
98 module_name[strlen(mod_name)] = '\0';
99 new = kzalloc(sizeof(struct debug_name), GFP_KERNEL);
100 if (!new) {
101 kfree(module_name);
102 ret = -ENOMEM;
103 goto out;
104 }
105 INIT_HLIST_NODE(&new->hlist);
106 INIT_HLIST_NODE(&new->hlist2);
107 new->name = module_name;
108 new->hash1 = hash;
109 new->hash2 = hash2;
110 hlist_add_head_rcu(&new->hlist, &module_table[hash]);
111 hlist_add_head_rcu(&new->hlist2, &module_table2[hash2]);
112 nr_entries++;
113out:
114 return ret;
115}
116
117int unregister_dynamic_debug_module(char *mod_name)
118{
119 struct debug_name *element;
120 int ret = 0;
121
122 down(&debug_list_mutex);
123 element = find_debug_module(mod_name);
124 if (!element) {
125 ret = -EINVAL;
126 goto out;
127 }
128 hlist_del_rcu(&element->hlist);
129 hlist_del_rcu(&element->hlist2);
130 synchronize_rcu();
131 kfree(element->name);
132 if (element->enable)
133 num_enabled--;
134 kfree(element);
135 nr_entries--;
136out:
137 up(&debug_list_mutex);
138 return ret;
139}
140EXPORT_SYMBOL_GPL(unregister_dynamic_debug_module);
141
142int register_dynamic_debug_module(char *mod_name, int type, char *share_name,
143 char *flags, int hash, int hash2)
144{
145 struct debug_name *elem;
146 int ret = 0;
147
148 down(&debug_list_mutex);
149 elem = find_debug_module(mod_name);
150 if (!elem) {
151 if (__add_debug_module(mod_name, hash, hash2))
152 goto out;
153 elem = find_debug_module(mod_name);
154 if (dynamic_enabled == DYNAMIC_ENABLED_ALL &&
155 !strcmp(mod_name, share_name)) {
156 elem->enable = true;
157 num_enabled++;
158 }
159 }
160 elem->type |= type;
161out:
162 up(&debug_list_mutex);
163 return ret;
164}
165EXPORT_SYMBOL_GPL(register_dynamic_debug_module);
166
167int __dynamic_dbg_enabled_helper(char *mod_name, int type, int value, int hash)
168{
169 struct debug_name *elem;
170 int ret = 0;
171
172 if (dynamic_enabled == DYNAMIC_ENABLED_ALL)
173 return 1;
174 rcu_read_lock();
175 elem = find_debug_module_hash(mod_name, hash);
176 if (elem && elem->enable)
177 ret = 1;
178 rcu_read_unlock();
179 return ret;
180}
181EXPORT_SYMBOL_GPL(__dynamic_dbg_enabled_helper);
182
183static void set_all(bool enable)
184{
185 struct debug_name *e;
186 struct hlist_node *node;
187 int i;
188 long long enable_mask;
189
190 for (i = 0; i < DEBUG_HASH_TABLE_SIZE; i++) {
191 if (module_table[i].first != NULL) {
192 hlist_for_each_entry(e, node, &module_table[i], hlist) {
193 e->enable = enable;
194 }
195 }
196 }
197 if (enable)
198 enable_mask = ULLONG_MAX;
199 else
200 enable_mask = 0;
201 dynamic_printk_enabled = enable_mask;
202 dynamic_printk_enabled2 = enable_mask;
203}
204
205static int disabled_hash(int i, bool first_table)
206{
207 struct debug_name *e;
208 struct hlist_node *node;
209
210 if (first_table) {
211 hlist_for_each_entry(e, node, &module_table[i], hlist) {
212 if (e->enable)
213 return 0;
214 }
215 } else {
216 hlist_for_each_entry(e, node, &module_table2[i], hlist2) {
217 if (e->enable)
218 return 0;
219 }
220 }
221 return 1;
222}
223
224static ssize_t pr_debug_write(struct file *file, const char __user *buf,
225 size_t length, loff_t *ppos)
226{
227 char *buffer, *s, *value_str, *setting_str;
228 int err, value;
229 struct debug_name *elem = NULL;
230 int all = 0;
231
232 if (length > PAGE_SIZE || length < 0)
233 return -EINVAL;
234
235 buffer = (char *)__get_free_page(GFP_KERNEL);
236 if (!buffer)
237 return -ENOMEM;
238
239 err = -EFAULT;
240 if (copy_from_user(buffer, buf, length))
241 goto out;
242
243 err = -EINVAL;
244 if (length < PAGE_SIZE)
245 buffer[length] = '\0';
246 else if (buffer[PAGE_SIZE-1])
247 goto out;
248
249 err = -EINVAL;
250 down(&debug_list_mutex);
251
252 if (strncmp("set", buffer, 3))
253 goto out_up;
254 s = buffer + 3;
255 setting_str = strsep(&s, "=");
256 if (s == NULL)
257 goto out_up;
258 setting_str = strstrip(setting_str);
259 value_str = strsep(&s, " ");
260 if (s == NULL)
261 goto out_up;
262 s = strstrip(s);
263 if (!strncmp(s, "all", 3))
264 all = 1;
265 else
266 elem = find_debug_module(s);
267 if (!strncmp(setting_str, "enable", 6)) {
268 value = !!simple_strtol(value_str, NULL, 10);
269 if (all) {
270 if (value) {
271 set_all(true);
272 num_enabled = nr_entries;
273 dynamic_enabled = DYNAMIC_ENABLED_ALL;
274 } else {
275 set_all(false);
276 num_enabled = 0;
277 dynamic_enabled = DYNAMIC_ENABLED_NONE;
278 }
279 err = 0;
280 } else if (elem) {
281 if (value && (elem->enable == 0)) {
282 dynamic_printk_enabled |= (1LL << elem->hash1);
283 dynamic_printk_enabled2 |= (1LL << elem->hash2);
284 elem->enable = 1;
285 num_enabled++;
286 dynamic_enabled = DYNAMIC_ENABLED_SOME;
287 err = 0;
288 printk(KERN_DEBUG
289 "debugging enabled for module %s\n",
290 elem->name);
291 } else if (!value && (elem->enable == 1)) {
292 elem->enable = 0;
293 num_enabled--;
294 if (disabled_hash(elem->hash1, true))
295 dynamic_printk_enabled &=
296 ~(1LL << elem->hash1);
297 if (disabled_hash(elem->hash2, false))
298 dynamic_printk_enabled2 &=
299 ~(1LL << elem->hash2);
300 if (num_enabled)
301 dynamic_enabled = DYNAMIC_ENABLED_SOME;
302 else
303 dynamic_enabled = DYNAMIC_ENABLED_NONE;
304 err = 0;
305 printk(KERN_DEBUG
306 "debugging disabled for module %s\n",
307 elem->name);
308 }
309 }
310 }
311 if (!err)
312 err = length;
313out_up:
314 up(&debug_list_mutex);
315out:
316 free_page((unsigned long)buffer);
317 return err;
318}
319
320static void *pr_debug_seq_start(struct seq_file *f, loff_t *pos)
321{
322 return (*pos < DEBUG_HASH_TABLE_SIZE) ? pos : NULL;
323}
324
325static void *pr_debug_seq_next(struct seq_file *s, void *v, loff_t *pos)
326{
327 (*pos)++;
328 if (*pos >= DEBUG_HASH_TABLE_SIZE)
329 return NULL;
330 return pos;
331}
332
333static void pr_debug_seq_stop(struct seq_file *s, void *v)
334{
335 /* Nothing to do */
336}
337
338static int pr_debug_seq_show(struct seq_file *s, void *v)
339{
340 struct hlist_head *head;
341 struct hlist_node *node;
342 struct debug_name *elem;
343 unsigned int i = *(loff_t *) v;
344
345 rcu_read_lock();
346 head = &module_table[i];
347 hlist_for_each_entry_rcu(elem, node, head, hlist) {
348 seq_printf(s, "%s enabled=%d", elem->name, elem->enable);
349 seq_printf(s, "\n");
350 }
351 rcu_read_unlock();
352 return 0;
353}
354
355static struct seq_operations pr_debug_seq_ops = {
356 .start = pr_debug_seq_start,
357 .next = pr_debug_seq_next,
358 .stop = pr_debug_seq_stop,
359 .show = pr_debug_seq_show
360};
361
362static int pr_debug_open(struct inode *inode, struct file *filp)
363{
364 return seq_open(filp, &pr_debug_seq_ops);
365}
366
367static const struct file_operations pr_debug_operations = {
368 .open = pr_debug_open,
369 .read = seq_read,
370 .write = pr_debug_write,
371 .llseek = seq_lseek,
372 .release = seq_release,
373};
374
375static int __init dynamic_printk_init(void)
376{
377 struct dentry *dir, *file;
378 struct mod_debug *iter;
379 unsigned long value;
380
381 dir = debugfs_create_dir("dynamic_printk", NULL);
382 if (!dir)
383 return -ENOMEM;
384 file = debugfs_create_file("modules", 0644, dir, NULL,
385 &pr_debug_operations);
386 if (!file) {
387 debugfs_remove(dir);
388 return -ENOMEM;
389 }
390 for (value = (unsigned long)__start___verbose;
391 value < (unsigned long)__stop___verbose;
392 value += sizeof(struct mod_debug)) {
393 iter = (struct mod_debug *)value;
394 register_dynamic_debug_module(iter->modname,
395 iter->type,
396 iter->logical_modname,
397 iter->flag_names, iter->hash, iter->hash2);
398 }
399 if (dynamic_enabled == DYNAMIC_ENABLED_ALL)
400 set_all(true);
401 return 0;
402}
403module_init(dynamic_printk_init);
404/* may want to move this earlier so we can get traces as early as possible */
405
406static int __init dynamic_printk_setup(char *str)
407{
408 if (str)
409 return -ENOENT;
410 dynamic_enabled = DYNAMIC_ENABLED_ALL;
411 return 0;
412}
413/* Use early_param(), so we can get debug output as early as possible */
414early_param("dynamic_printk", dynamic_printk_setup);
diff --git a/lib/idr.c b/lib/idr.c
index c11c5765cdef..dab4bca86f5d 100644
--- a/lib/idr.c
+++ b/lib/idr.c
@@ -449,6 +449,7 @@ void idr_remove_all(struct idr *idp)
449 449
450 n = idp->layers * IDR_BITS; 450 n = idp->layers * IDR_BITS;
451 p = idp->top; 451 p = idp->top;
452 rcu_assign_pointer(idp->top, NULL);
452 max = 1 << n; 453 max = 1 << n;
453 454
454 id = 0; 455 id = 0;
@@ -467,7 +468,6 @@ void idr_remove_all(struct idr *idp)
467 p = *--paa; 468 p = *--paa;
468 } 469 }
469 } 470 }
470 rcu_assign_pointer(idp->top, NULL);
471 idp->layers = 0; 471 idp->layers = 0;
472} 472}
473EXPORT_SYMBOL(idr_remove_all); 473EXPORT_SYMBOL(idr_remove_all);
diff --git a/lib/kobject.c b/lib/kobject.c
index 0487d1f64806..a6dec32f2ddd 100644
--- a/lib/kobject.c
+++ b/lib/kobject.c
@@ -212,7 +212,7 @@ static int kobject_add_internal(struct kobject *kobj)
212 * @fmt: format string used to build the name 212 * @fmt: format string used to build the name
213 * @vargs: vargs to format the string. 213 * @vargs: vargs to format the string.
214 */ 214 */
215static int kobject_set_name_vargs(struct kobject *kobj, const char *fmt, 215int kobject_set_name_vargs(struct kobject *kobj, const char *fmt,
216 va_list vargs) 216 va_list vargs)
217{ 217{
218 const char *old_name = kobj->name; 218 const char *old_name = kobj->name;
diff --git a/lib/kobject_uevent.c b/lib/kobject_uevent.c
index 38131028d16f..97a777ad4f59 100644
--- a/lib/kobject_uevent.c
+++ b/lib/kobject_uevent.c
@@ -118,6 +118,13 @@ int kobject_uevent_env(struct kobject *kobj, enum kobject_action action,
118 kset = top_kobj->kset; 118 kset = top_kobj->kset;
119 uevent_ops = kset->uevent_ops; 119 uevent_ops = kset->uevent_ops;
120 120
121 /* skip the event, if uevent_suppress is set*/
122 if (kobj->uevent_suppress) {
123 pr_debug("kobject: '%s' (%p): %s: uevent_suppress "
124 "caused the event to drop!\n",
125 kobject_name(kobj), kobj, __func__);
126 return 0;
127 }
121 /* skip the event, if the filter returns zero. */ 128 /* skip the event, if the filter returns zero. */
122 if (uevent_ops && uevent_ops->filter) 129 if (uevent_ops && uevent_ops->filter)
123 if (!uevent_ops->filter(kset, kobj)) { 130 if (!uevent_ops->filter(kset, kobj)) {
@@ -251,7 +258,7 @@ int kobject_uevent_env(struct kobject *kobj, enum kobject_action action,
251 goto exit; 258 goto exit;
252 259
253 retval = call_usermodehelper(argv[0], argv, 260 retval = call_usermodehelper(argv[0], argv,
254 env->envp, UMH_WAIT_EXEC); 261 env->envp, UMH_NO_WAIT);
255 } 262 }
256 263
257exit: 264exit:
diff --git a/lib/nlattr.c b/lib/nlattr.c
new file mode 100644
index 000000000000..80009a24e21d
--- /dev/null
+++ b/lib/nlattr.c
@@ -0,0 +1,475 @@
1/*
2 * NETLINK Netlink attributes
3 *
4 * Authors: Thomas Graf <tgraf@suug.ch>
5 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
6 */
7
8#include <linux/module.h>
9#include <linux/kernel.h>
10#include <linux/errno.h>
11#include <linux/jiffies.h>
12#include <linux/netdevice.h>
13#include <linux/skbuff.h>
14#include <linux/string.h>
15#include <linux/types.h>
16#include <net/netlink.h>
17
18static u16 nla_attr_minlen[NLA_TYPE_MAX+1] __read_mostly = {
19 [NLA_U8] = sizeof(u8),
20 [NLA_U16] = sizeof(u16),
21 [NLA_U32] = sizeof(u32),
22 [NLA_U64] = sizeof(u64),
23 [NLA_NESTED] = NLA_HDRLEN,
24};
25
26static int validate_nla(struct nlattr *nla, int maxtype,
27 const struct nla_policy *policy)
28{
29 const struct nla_policy *pt;
30 int minlen = 0, attrlen = nla_len(nla), type = nla_type(nla);
31
32 if (type <= 0 || type > maxtype)
33 return 0;
34
35 pt = &policy[type];
36
37 BUG_ON(pt->type > NLA_TYPE_MAX);
38
39 switch (pt->type) {
40 case NLA_FLAG:
41 if (attrlen > 0)
42 return -ERANGE;
43 break;
44
45 case NLA_NUL_STRING:
46 if (pt->len)
47 minlen = min_t(int, attrlen, pt->len + 1);
48 else
49 minlen = attrlen;
50
51 if (!minlen || memchr(nla_data(nla), '\0', minlen) == NULL)
52 return -EINVAL;
53 /* fall through */
54
55 case NLA_STRING:
56 if (attrlen < 1)
57 return -ERANGE;
58
59 if (pt->len) {
60 char *buf = nla_data(nla);
61
62 if (buf[attrlen - 1] == '\0')
63 attrlen--;
64
65 if (attrlen > pt->len)
66 return -ERANGE;
67 }
68 break;
69
70 case NLA_BINARY:
71 if (pt->len && attrlen > pt->len)
72 return -ERANGE;
73 break;
74
75 case NLA_NESTED_COMPAT:
76 if (attrlen < pt->len)
77 return -ERANGE;
78 if (attrlen < NLA_ALIGN(pt->len))
79 break;
80 if (attrlen < NLA_ALIGN(pt->len) + NLA_HDRLEN)
81 return -ERANGE;
82 nla = nla_data(nla) + NLA_ALIGN(pt->len);
83 if (attrlen < NLA_ALIGN(pt->len) + NLA_HDRLEN + nla_len(nla))
84 return -ERANGE;
85 break;
86 case NLA_NESTED:
87 /* a nested attributes is allowed to be empty; if its not,
88 * it must have a size of at least NLA_HDRLEN.
89 */
90 if (attrlen == 0)
91 break;
92 default:
93 if (pt->len)
94 minlen = pt->len;
95 else if (pt->type != NLA_UNSPEC)
96 minlen = nla_attr_minlen[pt->type];
97
98 if (attrlen < minlen)
99 return -ERANGE;
100 }
101
102 return 0;
103}
104
105/**
106 * nla_validate - Validate a stream of attributes
107 * @head: head of attribute stream
108 * @len: length of attribute stream
109 * @maxtype: maximum attribute type to be expected
110 * @policy: validation policy
111 *
112 * Validates all attributes in the specified attribute stream against the
113 * specified policy. Attributes with a type exceeding maxtype will be
114 * ignored. See documenation of struct nla_policy for more details.
115 *
116 * Returns 0 on success or a negative error code.
117 */
118int nla_validate(struct nlattr *head, int len, int maxtype,
119 const struct nla_policy *policy)
120{
121 struct nlattr *nla;
122 int rem, err;
123
124 nla_for_each_attr(nla, head, len, rem) {
125 err = validate_nla(nla, maxtype, policy);
126 if (err < 0)
127 goto errout;
128 }
129
130 err = 0;
131errout:
132 return err;
133}
134
135/**
136 * nla_parse - Parse a stream of attributes into a tb buffer
137 * @tb: destination array with maxtype+1 elements
138 * @maxtype: maximum attribute type to be expected
139 * @head: head of attribute stream
140 * @len: length of attribute stream
141 * @policy: validation policy
142 *
143 * Parses a stream of attributes and stores a pointer to each attribute in
144 * the tb array accessable via the attribute type. Attributes with a type
145 * exceeding maxtype will be silently ignored for backwards compatibility
146 * reasons. policy may be set to NULL if no validation is required.
147 *
148 * Returns 0 on success or a negative error code.
149 */
150int nla_parse(struct nlattr *tb[], int maxtype, struct nlattr *head, int len,
151 const struct nla_policy *policy)
152{
153 struct nlattr *nla;
154 int rem, err;
155
156 memset(tb, 0, sizeof(struct nlattr *) * (maxtype + 1));
157
158 nla_for_each_attr(nla, head, len, rem) {
159 u16 type = nla_type(nla);
160
161 if (type > 0 && type <= maxtype) {
162 if (policy) {
163 err = validate_nla(nla, maxtype, policy);
164 if (err < 0)
165 goto errout;
166 }
167
168 tb[type] = nla;
169 }
170 }
171
172 if (unlikely(rem > 0))
173 printk(KERN_WARNING "netlink: %d bytes leftover after parsing "
174 "attributes.\n", rem);
175
176 err = 0;
177errout:
178 return err;
179}
180
181/**
182 * nla_find - Find a specific attribute in a stream of attributes
183 * @head: head of attribute stream
184 * @len: length of attribute stream
185 * @attrtype: type of attribute to look for
186 *
187 * Returns the first attribute in the stream matching the specified type.
188 */
189struct nlattr *nla_find(struct nlattr *head, int len, int attrtype)
190{
191 struct nlattr *nla;
192 int rem;
193
194 nla_for_each_attr(nla, head, len, rem)
195 if (nla_type(nla) == attrtype)
196 return nla;
197
198 return NULL;
199}
200
201/**
202 * nla_strlcpy - Copy string attribute payload into a sized buffer
203 * @dst: where to copy the string to
204 * @nla: attribute to copy the string from
205 * @dstsize: size of destination buffer
206 *
207 * Copies at most dstsize - 1 bytes into the destination buffer.
208 * The result is always a valid NUL-terminated string. Unlike
209 * strlcpy the destination buffer is always padded out.
210 *
211 * Returns the length of the source buffer.
212 */
213size_t nla_strlcpy(char *dst, const struct nlattr *nla, size_t dstsize)
214{
215 size_t srclen = nla_len(nla);
216 char *src = nla_data(nla);
217
218 if (srclen > 0 && src[srclen - 1] == '\0')
219 srclen--;
220
221 if (dstsize > 0) {
222 size_t len = (srclen >= dstsize) ? dstsize - 1 : srclen;
223
224 memset(dst, 0, dstsize);
225 memcpy(dst, src, len);
226 }
227
228 return srclen;
229}
230
231/**
232 * nla_memcpy - Copy a netlink attribute into another memory area
233 * @dest: where to copy to memcpy
234 * @src: netlink attribute to copy from
235 * @count: size of the destination area
236 *
237 * Note: The number of bytes copied is limited by the length of
238 * attribute's payload. memcpy
239 *
240 * Returns the number of bytes copied.
241 */
242int nla_memcpy(void *dest, const struct nlattr *src, int count)
243{
244 int minlen = min_t(int, count, nla_len(src));
245
246 memcpy(dest, nla_data(src), minlen);
247
248 return minlen;
249}
250
251/**
252 * nla_memcmp - Compare an attribute with sized memory area
253 * @nla: netlink attribute
254 * @data: memory area
255 * @size: size of memory area
256 */
257int nla_memcmp(const struct nlattr *nla, const void *data,
258 size_t size)
259{
260 int d = nla_len(nla) - size;
261
262 if (d == 0)
263 d = memcmp(nla_data(nla), data, size);
264
265 return d;
266}
267
268/**
269 * nla_strcmp - Compare a string attribute against a string
270 * @nla: netlink string attribute
271 * @str: another string
272 */
273int nla_strcmp(const struct nlattr *nla, const char *str)
274{
275 int len = strlen(str) + 1;
276 int d = nla_len(nla) - len;
277
278 if (d == 0)
279 d = memcmp(nla_data(nla), str, len);
280
281 return d;
282}
283
284#ifdef CONFIG_NET
285/**
286 * __nla_reserve - reserve room for attribute on the skb
287 * @skb: socket buffer to reserve room on
288 * @attrtype: attribute type
289 * @attrlen: length of attribute payload
290 *
291 * Adds a netlink attribute header to a socket buffer and reserves
292 * room for the payload but does not copy it.
293 *
294 * The caller is responsible to ensure that the skb provides enough
295 * tailroom for the attribute header and payload.
296 */
297struct nlattr *__nla_reserve(struct sk_buff *skb, int attrtype, int attrlen)
298{
299 struct nlattr *nla;
300
301 nla = (struct nlattr *) skb_put(skb, nla_total_size(attrlen));
302 nla->nla_type = attrtype;
303 nla->nla_len = nla_attr_size(attrlen);
304
305 memset((unsigned char *) nla + nla->nla_len, 0, nla_padlen(attrlen));
306
307 return nla;
308}
309EXPORT_SYMBOL(__nla_reserve);
310
311/**
312 * __nla_reserve_nohdr - reserve room for attribute without header
313 * @skb: socket buffer to reserve room on
314 * @attrlen: length of attribute payload
315 *
316 * Reserves room for attribute payload without a header.
317 *
318 * The caller is responsible to ensure that the skb provides enough
319 * tailroom for the payload.
320 */
321void *__nla_reserve_nohdr(struct sk_buff *skb, int attrlen)
322{
323 void *start;
324
325 start = skb_put(skb, NLA_ALIGN(attrlen));
326 memset(start, 0, NLA_ALIGN(attrlen));
327
328 return start;
329}
330EXPORT_SYMBOL(__nla_reserve_nohdr);
331
332/**
333 * nla_reserve - reserve room for attribute on the skb
334 * @skb: socket buffer to reserve room on
335 * @attrtype: attribute type
336 * @attrlen: length of attribute payload
337 *
338 * Adds a netlink attribute header to a socket buffer and reserves
339 * room for the payload but does not copy it.
340 *
341 * Returns NULL if the tailroom of the skb is insufficient to store
342 * the attribute header and payload.
343 */
344struct nlattr *nla_reserve(struct sk_buff *skb, int attrtype, int attrlen)
345{
346 if (unlikely(skb_tailroom(skb) < nla_total_size(attrlen)))
347 return NULL;
348
349 return __nla_reserve(skb, attrtype, attrlen);
350}
351EXPORT_SYMBOL(nla_reserve);
352
353/**
354 * nla_reserve_nohdr - reserve room for attribute without header
355 * @skb: socket buffer to reserve room on
356 * @attrlen: length of attribute payload
357 *
358 * Reserves room for attribute payload without a header.
359 *
360 * Returns NULL if the tailroom of the skb is insufficient to store
361 * the attribute payload.
362 */
363void *nla_reserve_nohdr(struct sk_buff *skb, int attrlen)
364{
365 if (unlikely(skb_tailroom(skb) < NLA_ALIGN(attrlen)))
366 return NULL;
367
368 return __nla_reserve_nohdr(skb, attrlen);
369}
370EXPORT_SYMBOL(nla_reserve_nohdr);
371
372/**
373 * __nla_put - Add a netlink attribute to a socket buffer
374 * @skb: socket buffer to add attribute to
375 * @attrtype: attribute type
376 * @attrlen: length of attribute payload
377 * @data: head of attribute payload
378 *
379 * The caller is responsible to ensure that the skb provides enough
380 * tailroom for the attribute header and payload.
381 */
382void __nla_put(struct sk_buff *skb, int attrtype, int attrlen,
383 const void *data)
384{
385 struct nlattr *nla;
386
387 nla = __nla_reserve(skb, attrtype, attrlen);
388 memcpy(nla_data(nla), data, attrlen);
389}
390EXPORT_SYMBOL(__nla_put);
391
392/**
393 * __nla_put_nohdr - Add a netlink attribute without header
394 * @skb: socket buffer to add attribute to
395 * @attrlen: length of attribute payload
396 * @data: head of attribute payload
397 *
398 * The caller is responsible to ensure that the skb provides enough
399 * tailroom for the attribute payload.
400 */
401void __nla_put_nohdr(struct sk_buff *skb, int attrlen, const void *data)
402{
403 void *start;
404
405 start = __nla_reserve_nohdr(skb, attrlen);
406 memcpy(start, data, attrlen);
407}
408EXPORT_SYMBOL(__nla_put_nohdr);
409
410/**
411 * nla_put - Add a netlink attribute to a socket buffer
412 * @skb: socket buffer to add attribute to
413 * @attrtype: attribute type
414 * @attrlen: length of attribute payload
415 * @data: head of attribute payload
416 *
417 * Returns -EMSGSIZE if the tailroom of the skb is insufficient to store
418 * the attribute header and payload.
419 */
420int nla_put(struct sk_buff *skb, int attrtype, int attrlen, const void *data)
421{
422 if (unlikely(skb_tailroom(skb) < nla_total_size(attrlen)))
423 return -EMSGSIZE;
424
425 __nla_put(skb, attrtype, attrlen, data);
426 return 0;
427}
428EXPORT_SYMBOL(nla_put);
429
430/**
431 * nla_put_nohdr - Add a netlink attribute without header
432 * @skb: socket buffer to add attribute to
433 * @attrlen: length of attribute payload
434 * @data: head of attribute payload
435 *
436 * Returns -EMSGSIZE if the tailroom of the skb is insufficient to store
437 * the attribute payload.
438 */
439int nla_put_nohdr(struct sk_buff *skb, int attrlen, const void *data)
440{
441 if (unlikely(skb_tailroom(skb) < NLA_ALIGN(attrlen)))
442 return -EMSGSIZE;
443
444 __nla_put_nohdr(skb, attrlen, data);
445 return 0;
446}
447EXPORT_SYMBOL(nla_put_nohdr);
448
449/**
450 * nla_append - Add a netlink attribute without header or padding
451 * @skb: socket buffer to add attribute to
452 * @attrlen: length of attribute payload
453 * @data: head of attribute payload
454 *
455 * Returns -EMSGSIZE if the tailroom of the skb is insufficient to store
456 * the attribute payload.
457 */
458int nla_append(struct sk_buff *skb, int attrlen, const void *data)
459{
460 if (unlikely(skb_tailroom(skb) < NLA_ALIGN(attrlen)))
461 return -EMSGSIZE;
462
463 memcpy(skb_put(skb, attrlen), data, attrlen);
464 return 0;
465}
466EXPORT_SYMBOL(nla_append);
467#endif
468
469EXPORT_SYMBOL(nla_validate);
470EXPORT_SYMBOL(nla_parse);
471EXPORT_SYMBOL(nla_find);
472EXPORT_SYMBOL(nla_strlcpy);
473EXPORT_SYMBOL(nla_memcpy);
474EXPORT_SYMBOL(nla_memcmp);
475EXPORT_SYMBOL(nla_strcmp);