aboutsummaryrefslogtreecommitdiffstats
path: root/kernel
diff options
context:
space:
mode:
Diffstat (limited to 'kernel')
-rw-r--r--kernel/Makefile1
-rw-r--r--kernel/audit.c18
-rw-r--r--kernel/audit_watch.c2
-rw-r--r--kernel/auditsc.c6
-rw-r--r--kernel/cgroup.c1107
-rw-r--r--kernel/cgroup_debug.c105
-rw-r--r--kernel/cgroup_freezer.c15
-rw-r--r--kernel/cpuset.c66
-rw-r--r--kernel/exit.c146
-rw-r--r--kernel/fork.c34
-rw-r--r--kernel/gcov/Kconfig2
-rw-r--r--kernel/hung_task.c4
-rw-r--r--kernel/module.c159
-rw-r--r--kernel/ns_cgroup.c16
-rw-r--r--kernel/params.c7
-rw-r--r--kernel/pid_namespace.c2
-rw-r--r--kernel/ptrace.c11
-rw-r--r--kernel/res_counter.c21
-rw-r--r--kernel/sched.c39
-rw-r--r--kernel/sched_fair.c4
-rw-r--r--kernel/signal.c168
-rw-r--r--kernel/slow-work.c12
-rw-r--r--kernel/softlockup.c4
-rw-r--r--kernel/sys.c22
-rw-r--r--kernel/sysctl.c112
-rw-r--r--kernel/time/Makefile2
-rw-r--r--kernel/time/timeconv.c127
-rw-r--r--kernel/trace/ftrace.c4
-rw-r--r--kernel/trace/trace_stack.c4
-rw-r--r--kernel/utsname_sysctl.c4
30 files changed, 1532 insertions, 692 deletions
diff --git a/kernel/Makefile b/kernel/Makefile
index 187c89b4783d..b8d4cd8ac0b9 100644
--- a/kernel/Makefile
+++ b/kernel/Makefile
@@ -58,7 +58,6 @@ obj-$(CONFIG_KEXEC) += kexec.o
58obj-$(CONFIG_BACKTRACE_SELF_TEST) += backtracetest.o 58obj-$(CONFIG_BACKTRACE_SELF_TEST) += backtracetest.o
59obj-$(CONFIG_COMPAT) += compat.o 59obj-$(CONFIG_COMPAT) += compat.o
60obj-$(CONFIG_CGROUPS) += cgroup.o 60obj-$(CONFIG_CGROUPS) += cgroup.o
61obj-$(CONFIG_CGROUP_DEBUG) += cgroup_debug.o
62obj-$(CONFIG_CGROUP_FREEZER) += cgroup_freezer.o 61obj-$(CONFIG_CGROUP_FREEZER) += cgroup_freezer.o
63obj-$(CONFIG_CPUSETS) += cpuset.o 62obj-$(CONFIG_CPUSETS) += cpuset.o
64obj-$(CONFIG_CGROUP_NS) += ns_cgroup.o 63obj-$(CONFIG_CGROUP_NS) += ns_cgroup.o
diff --git a/kernel/audit.c b/kernel/audit.c
index defc2e6f1e3b..5feed232be9d 100644
--- a/kernel/audit.c
+++ b/kernel/audit.c
@@ -855,18 +855,24 @@ static int audit_receive_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
855 break; 855 break;
856 } 856 }
857 case AUDIT_SIGNAL_INFO: 857 case AUDIT_SIGNAL_INFO:
858 err = security_secid_to_secctx(audit_sig_sid, &ctx, &len); 858 len = 0;
859 if (err) 859 if (audit_sig_sid) {
860 return err; 860 err = security_secid_to_secctx(audit_sig_sid, &ctx, &len);
861 if (err)
862 return err;
863 }
861 sig_data = kmalloc(sizeof(*sig_data) + len, GFP_KERNEL); 864 sig_data = kmalloc(sizeof(*sig_data) + len, GFP_KERNEL);
862 if (!sig_data) { 865 if (!sig_data) {
863 security_release_secctx(ctx, len); 866 if (audit_sig_sid)
867 security_release_secctx(ctx, len);
864 return -ENOMEM; 868 return -ENOMEM;
865 } 869 }
866 sig_data->uid = audit_sig_uid; 870 sig_data->uid = audit_sig_uid;
867 sig_data->pid = audit_sig_pid; 871 sig_data->pid = audit_sig_pid;
868 memcpy(sig_data->ctx, ctx, len); 872 if (audit_sig_sid) {
869 security_release_secctx(ctx, len); 873 memcpy(sig_data->ctx, ctx, len);
874 security_release_secctx(ctx, len);
875 }
870 audit_send_reply(NETLINK_CB(skb).pid, seq, AUDIT_SIGNAL_INFO, 876 audit_send_reply(NETLINK_CB(skb).pid, seq, AUDIT_SIGNAL_INFO,
871 0, 0, sig_data, sizeof(*sig_data) + len); 877 0, 0, sig_data, sizeof(*sig_data) + len);
872 kfree(sig_data); 878 kfree(sig_data);
diff --git a/kernel/audit_watch.c b/kernel/audit_watch.c
index 0e96dbc60ea9..cc7e87936cbc 100644
--- a/kernel/audit_watch.c
+++ b/kernel/audit_watch.c
@@ -45,8 +45,8 @@
45 45
46struct audit_watch { 46struct audit_watch {
47 atomic_t count; /* reference count */ 47 atomic_t count; /* reference count */
48 char *path; /* insertion path */
49 dev_t dev; /* associated superblock device */ 48 dev_t dev; /* associated superblock device */
49 char *path; /* insertion path */
50 unsigned long ino; /* associated inode number */ 50 unsigned long ino; /* associated inode number */
51 struct audit_parent *parent; /* associated parent */ 51 struct audit_parent *parent; /* associated parent */
52 struct list_head wlist; /* entry in parent->watches list */ 52 struct list_head wlist; /* entry in parent->watches list */
diff --git a/kernel/auditsc.c b/kernel/auditsc.c
index 68d3c6a0ecd6..267e484f0198 100644
--- a/kernel/auditsc.c
+++ b/kernel/auditsc.c
@@ -168,12 +168,12 @@ struct audit_context {
168 int in_syscall; /* 1 if task is in a syscall */ 168 int in_syscall; /* 1 if task is in a syscall */
169 enum audit_state state, current_state; 169 enum audit_state state, current_state;
170 unsigned int serial; /* serial number for record */ 170 unsigned int serial; /* serial number for record */
171 struct timespec ctime; /* time of syscall entry */
172 int major; /* syscall number */ 171 int major; /* syscall number */
172 struct timespec ctime; /* time of syscall entry */
173 unsigned long argv[4]; /* syscall arguments */ 173 unsigned long argv[4]; /* syscall arguments */
174 int return_valid; /* return code is valid */
175 long return_code;/* syscall return code */ 174 long return_code;/* syscall return code */
176 u64 prio; 175 u64 prio;
176 int return_valid; /* return code is valid */
177 int name_count; 177 int name_count;
178 struct audit_names names[AUDIT_NAMES]; 178 struct audit_names names[AUDIT_NAMES];
179 char * filterkey; /* key for rule that triggered record */ 179 char * filterkey; /* key for rule that triggered record */
@@ -198,8 +198,8 @@ struct audit_context {
198 char target_comm[TASK_COMM_LEN]; 198 char target_comm[TASK_COMM_LEN];
199 199
200 struct audit_tree_refs *trees, *first_trees; 200 struct audit_tree_refs *trees, *first_trees;
201 int tree_count;
202 struct list_head killed_trees; 201 struct list_head killed_trees;
202 int tree_count;
203 203
204 int type; 204 int type;
205 union { 205 union {
diff --git a/kernel/cgroup.c b/kernel/cgroup.c
index cd83d9933b6b..7ccba4bc5e3b 100644
--- a/kernel/cgroup.c
+++ b/kernel/cgroup.c
@@ -23,6 +23,7 @@
23 */ 23 */
24 24
25#include <linux/cgroup.h> 25#include <linux/cgroup.h>
26#include <linux/ctype.h>
26#include <linux/errno.h> 27#include <linux/errno.h>
27#include <linux/fs.h> 28#include <linux/fs.h>
28#include <linux/kernel.h> 29#include <linux/kernel.h>
@@ -48,6 +49,8 @@
48#include <linux/namei.h> 49#include <linux/namei.h>
49#include <linux/smp_lock.h> 50#include <linux/smp_lock.h>
50#include <linux/pid_namespace.h> 51#include <linux/pid_namespace.h>
52#include <linux/idr.h>
53#include <linux/vmalloc.h> /* TODO: replace with more sophisticated array */
51 54
52#include <asm/atomic.h> 55#include <asm/atomic.h>
53 56
@@ -60,6 +63,8 @@ static struct cgroup_subsys *subsys[] = {
60#include <linux/cgroup_subsys.h> 63#include <linux/cgroup_subsys.h>
61}; 64};
62 65
66#define MAX_CGROUP_ROOT_NAMELEN 64
67
63/* 68/*
64 * A cgroupfs_root represents the root of a cgroup hierarchy, 69 * A cgroupfs_root represents the root of a cgroup hierarchy,
65 * and may be associated with a superblock to form an active 70 * and may be associated with a superblock to form an active
@@ -74,6 +79,9 @@ struct cgroupfs_root {
74 */ 79 */
75 unsigned long subsys_bits; 80 unsigned long subsys_bits;
76 81
82 /* Unique id for this hierarchy. */
83 int hierarchy_id;
84
77 /* The bitmask of subsystems currently attached to this hierarchy */ 85 /* The bitmask of subsystems currently attached to this hierarchy */
78 unsigned long actual_subsys_bits; 86 unsigned long actual_subsys_bits;
79 87
@@ -94,6 +102,9 @@ struct cgroupfs_root {
94 102
95 /* The path to use for release notifications. */ 103 /* The path to use for release notifications. */
96 char release_agent_path[PATH_MAX]; 104 char release_agent_path[PATH_MAX];
105
106 /* The name for this hierarchy - may be empty */
107 char name[MAX_CGROUP_ROOT_NAMELEN];
97}; 108};
98 109
99/* 110/*
@@ -141,6 +152,10 @@ struct css_id {
141static LIST_HEAD(roots); 152static LIST_HEAD(roots);
142static int root_count; 153static int root_count;
143 154
155static DEFINE_IDA(hierarchy_ida);
156static int next_hierarchy_id;
157static DEFINE_SPINLOCK(hierarchy_id_lock);
158
144/* dummytop is a shorthand for the dummy hierarchy's top cgroup */ 159/* dummytop is a shorthand for the dummy hierarchy's top cgroup */
145#define dummytop (&rootnode.top_cgroup) 160#define dummytop (&rootnode.top_cgroup)
146 161
@@ -201,6 +216,7 @@ struct cg_cgroup_link {
201 * cgroup, anchored on cgroup->css_sets 216 * cgroup, anchored on cgroup->css_sets
202 */ 217 */
203 struct list_head cgrp_link_list; 218 struct list_head cgrp_link_list;
219 struct cgroup *cgrp;
204 /* 220 /*
205 * List running through cg_cgroup_links pointing at a 221 * List running through cg_cgroup_links pointing at a
206 * single css_set object, anchored on css_set->cg_links 222 * single css_set object, anchored on css_set->cg_links
@@ -227,8 +243,11 @@ static int cgroup_subsys_init_idr(struct cgroup_subsys *ss);
227static DEFINE_RWLOCK(css_set_lock); 243static DEFINE_RWLOCK(css_set_lock);
228static int css_set_count; 244static int css_set_count;
229 245
230/* hash table for cgroup groups. This improves the performance to 246/*
231 * find an existing css_set */ 247 * hash table for cgroup groups. This improves the performance to find
248 * an existing css_set. This hash doesn't (currently) take into
249 * account cgroups in empty hierarchies.
250 */
232#define CSS_SET_HASH_BITS 7 251#define CSS_SET_HASH_BITS 7
233#define CSS_SET_TABLE_SIZE (1 << CSS_SET_HASH_BITS) 252#define CSS_SET_TABLE_SIZE (1 << CSS_SET_HASH_BITS)
234static struct hlist_head css_set_table[CSS_SET_TABLE_SIZE]; 253static struct hlist_head css_set_table[CSS_SET_TABLE_SIZE];
@@ -248,48 +267,22 @@ static struct hlist_head *css_set_hash(struct cgroup_subsys_state *css[])
248 return &css_set_table[index]; 267 return &css_set_table[index];
249} 268}
250 269
270static void free_css_set_rcu(struct rcu_head *obj)
271{
272 struct css_set *cg = container_of(obj, struct css_set, rcu_head);
273 kfree(cg);
274}
275
251/* We don't maintain the lists running through each css_set to its 276/* We don't maintain the lists running through each css_set to its
252 * task until after the first call to cgroup_iter_start(). This 277 * task until after the first call to cgroup_iter_start(). This
253 * reduces the fork()/exit() overhead for people who have cgroups 278 * reduces the fork()/exit() overhead for people who have cgroups
254 * compiled into their kernel but not actually in use */ 279 * compiled into their kernel but not actually in use */
255static int use_task_css_set_links __read_mostly; 280static int use_task_css_set_links __read_mostly;
256 281
257/* When we create or destroy a css_set, the operation simply 282static void __put_css_set(struct css_set *cg, int taskexit)
258 * takes/releases a reference count on all the cgroups referenced
259 * by subsystems in this css_set. This can end up multiple-counting
260 * some cgroups, but that's OK - the ref-count is just a
261 * busy/not-busy indicator; ensuring that we only count each cgroup
262 * once would require taking a global lock to ensure that no
263 * subsystems moved between hierarchies while we were doing so.
264 *
265 * Possible TODO: decide at boot time based on the number of
266 * registered subsystems and the number of CPUs or NUMA nodes whether
267 * it's better for performance to ref-count every subsystem, or to
268 * take a global lock and only add one ref count to each hierarchy.
269 */
270
271/*
272 * unlink a css_set from the list and free it
273 */
274static void unlink_css_set(struct css_set *cg)
275{ 283{
276 struct cg_cgroup_link *link; 284 struct cg_cgroup_link *link;
277 struct cg_cgroup_link *saved_link; 285 struct cg_cgroup_link *saved_link;
278
279 hlist_del(&cg->hlist);
280 css_set_count--;
281
282 list_for_each_entry_safe(link, saved_link, &cg->cg_links,
283 cg_link_list) {
284 list_del(&link->cg_link_list);
285 list_del(&link->cgrp_link_list);
286 kfree(link);
287 }
288}
289
290static void __put_css_set(struct css_set *cg, int taskexit)
291{
292 int i;
293 /* 286 /*
294 * Ensure that the refcount doesn't hit zero while any readers 287 * Ensure that the refcount doesn't hit zero while any readers
295 * can see it. Similar to atomic_dec_and_lock(), but for an 288 * can see it. Similar to atomic_dec_and_lock(), but for an
@@ -302,21 +295,28 @@ static void __put_css_set(struct css_set *cg, int taskexit)
302 write_unlock(&css_set_lock); 295 write_unlock(&css_set_lock);
303 return; 296 return;
304 } 297 }
305 unlink_css_set(cg);
306 write_unlock(&css_set_lock);
307 298
308 rcu_read_lock(); 299 /* This css_set is dead. unlink it and release cgroup refcounts */
309 for (i = 0; i < CGROUP_SUBSYS_COUNT; i++) { 300 hlist_del(&cg->hlist);
310 struct cgroup *cgrp = rcu_dereference(cg->subsys[i]->cgroup); 301 css_set_count--;
302
303 list_for_each_entry_safe(link, saved_link, &cg->cg_links,
304 cg_link_list) {
305 struct cgroup *cgrp = link->cgrp;
306 list_del(&link->cg_link_list);
307 list_del(&link->cgrp_link_list);
311 if (atomic_dec_and_test(&cgrp->count) && 308 if (atomic_dec_and_test(&cgrp->count) &&
312 notify_on_release(cgrp)) { 309 notify_on_release(cgrp)) {
313 if (taskexit) 310 if (taskexit)
314 set_bit(CGRP_RELEASABLE, &cgrp->flags); 311 set_bit(CGRP_RELEASABLE, &cgrp->flags);
315 check_for_release(cgrp); 312 check_for_release(cgrp);
316 } 313 }
314
315 kfree(link);
317 } 316 }
318 rcu_read_unlock(); 317
319 kfree(cg); 318 write_unlock(&css_set_lock);
319 call_rcu(&cg->rcu_head, free_css_set_rcu);
320} 320}
321 321
322/* 322/*
@@ -338,6 +338,78 @@ static inline void put_css_set_taskexit(struct css_set *cg)
338} 338}
339 339
340/* 340/*
341 * compare_css_sets - helper function for find_existing_css_set().
342 * @cg: candidate css_set being tested
343 * @old_cg: existing css_set for a task
344 * @new_cgrp: cgroup that's being entered by the task
345 * @template: desired set of css pointers in css_set (pre-calculated)
346 *
347 * Returns true if "cg" matches "old_cg" except for the hierarchy
348 * which "new_cgrp" belongs to, for which it should match "new_cgrp".
349 */
350static bool compare_css_sets(struct css_set *cg,
351 struct css_set *old_cg,
352 struct cgroup *new_cgrp,
353 struct cgroup_subsys_state *template[])
354{
355 struct list_head *l1, *l2;
356
357 if (memcmp(template, cg->subsys, sizeof(cg->subsys))) {
358 /* Not all subsystems matched */
359 return false;
360 }
361
362 /*
363 * Compare cgroup pointers in order to distinguish between
364 * different cgroups in heirarchies with no subsystems. We
365 * could get by with just this check alone (and skip the
366 * memcmp above) but on most setups the memcmp check will
367 * avoid the need for this more expensive check on almost all
368 * candidates.
369 */
370
371 l1 = &cg->cg_links;
372 l2 = &old_cg->cg_links;
373 while (1) {
374 struct cg_cgroup_link *cgl1, *cgl2;
375 struct cgroup *cg1, *cg2;
376
377 l1 = l1->next;
378 l2 = l2->next;
379 /* See if we reached the end - both lists are equal length. */
380 if (l1 == &cg->cg_links) {
381 BUG_ON(l2 != &old_cg->cg_links);
382 break;
383 } else {
384 BUG_ON(l2 == &old_cg->cg_links);
385 }
386 /* Locate the cgroups associated with these links. */
387 cgl1 = list_entry(l1, struct cg_cgroup_link, cg_link_list);
388 cgl2 = list_entry(l2, struct cg_cgroup_link, cg_link_list);
389 cg1 = cgl1->cgrp;
390 cg2 = cgl2->cgrp;
391 /* Hierarchies should be linked in the same order. */
392 BUG_ON(cg1->root != cg2->root);
393
394 /*
395 * If this hierarchy is the hierarchy of the cgroup
396 * that's changing, then we need to check that this
397 * css_set points to the new cgroup; if it's any other
398 * hierarchy, then this css_set should point to the
399 * same cgroup as the old css_set.
400 */
401 if (cg1->root == new_cgrp->root) {
402 if (cg1 != new_cgrp)
403 return false;
404 } else {
405 if (cg1 != cg2)
406 return false;
407 }
408 }
409 return true;
410}
411
412/*
341 * find_existing_css_set() is a helper for 413 * find_existing_css_set() is a helper for
342 * find_css_set(), and checks to see whether an existing 414 * find_css_set(), and checks to see whether an existing
343 * css_set is suitable. 415 * css_set is suitable.
@@ -378,10 +450,11 @@ static struct css_set *find_existing_css_set(
378 450
379 hhead = css_set_hash(template); 451 hhead = css_set_hash(template);
380 hlist_for_each_entry(cg, node, hhead, hlist) { 452 hlist_for_each_entry(cg, node, hhead, hlist) {
381 if (!memcmp(template, cg->subsys, sizeof(cg->subsys))) { 453 if (!compare_css_sets(cg, oldcg, cgrp, template))
382 /* All subsystems matched */ 454 continue;
383 return cg; 455
384 } 456 /* This css_set matches what we need */
457 return cg;
385 } 458 }
386 459
387 /* No existing cgroup group matched */ 460 /* No existing cgroup group matched */
@@ -435,8 +508,14 @@ static void link_css_set(struct list_head *tmp_cg_links,
435 link = list_first_entry(tmp_cg_links, struct cg_cgroup_link, 508 link = list_first_entry(tmp_cg_links, struct cg_cgroup_link,
436 cgrp_link_list); 509 cgrp_link_list);
437 link->cg = cg; 510 link->cg = cg;
511 link->cgrp = cgrp;
512 atomic_inc(&cgrp->count);
438 list_move(&link->cgrp_link_list, &cgrp->css_sets); 513 list_move(&link->cgrp_link_list, &cgrp->css_sets);
439 list_add(&link->cg_link_list, &cg->cg_links); 514 /*
515 * Always add links to the tail of the list so that the list
516 * is sorted by order of hierarchy creation
517 */
518 list_add_tail(&link->cg_link_list, &cg->cg_links);
440} 519}
441 520
442/* 521/*
@@ -451,11 +530,11 @@ static struct css_set *find_css_set(
451{ 530{
452 struct css_set *res; 531 struct css_set *res;
453 struct cgroup_subsys_state *template[CGROUP_SUBSYS_COUNT]; 532 struct cgroup_subsys_state *template[CGROUP_SUBSYS_COUNT];
454 int i;
455 533
456 struct list_head tmp_cg_links; 534 struct list_head tmp_cg_links;
457 535
458 struct hlist_head *hhead; 536 struct hlist_head *hhead;
537 struct cg_cgroup_link *link;
459 538
460 /* First see if we already have a cgroup group that matches 539 /* First see if we already have a cgroup group that matches
461 * the desired set */ 540 * the desired set */
@@ -489,20 +568,12 @@ static struct css_set *find_css_set(
489 568
490 write_lock(&css_set_lock); 569 write_lock(&css_set_lock);
491 /* Add reference counts and links from the new css_set. */ 570 /* Add reference counts and links from the new css_set. */
492 for (i = 0; i < CGROUP_SUBSYS_COUNT; i++) { 571 list_for_each_entry(link, &oldcg->cg_links, cg_link_list) {
493 struct cgroup *cgrp = res->subsys[i]->cgroup; 572 struct cgroup *c = link->cgrp;
494 struct cgroup_subsys *ss = subsys[i]; 573 if (c->root == cgrp->root)
495 atomic_inc(&cgrp->count); 574 c = cgrp;
496 /* 575 link_css_set(&tmp_cg_links, res, c);
497 * We want to add a link once per cgroup, so we
498 * only do it for the first subsystem in each
499 * hierarchy
500 */
501 if (ss->root->subsys_list.next == &ss->sibling)
502 link_css_set(&tmp_cg_links, res, cgrp);
503 } 576 }
504 if (list_empty(&rootnode.subsys_list))
505 link_css_set(&tmp_cg_links, res, dummytop);
506 577
507 BUG_ON(!list_empty(&tmp_cg_links)); 578 BUG_ON(!list_empty(&tmp_cg_links));
508 579
@@ -518,6 +589,41 @@ static struct css_set *find_css_set(
518} 589}
519 590
520/* 591/*
592 * Return the cgroup for "task" from the given hierarchy. Must be
593 * called with cgroup_mutex held.
594 */
595static struct cgroup *task_cgroup_from_root(struct task_struct *task,
596 struct cgroupfs_root *root)
597{
598 struct css_set *css;
599 struct cgroup *res = NULL;
600
601 BUG_ON(!mutex_is_locked(&cgroup_mutex));
602 read_lock(&css_set_lock);
603 /*
604 * No need to lock the task - since we hold cgroup_mutex the
605 * task can't change groups, so the only thing that can happen
606 * is that it exits and its css is set back to init_css_set.
607 */
608 css = task->cgroups;
609 if (css == &init_css_set) {
610 res = &root->top_cgroup;
611 } else {
612 struct cg_cgroup_link *link;
613 list_for_each_entry(link, &css->cg_links, cg_link_list) {
614 struct cgroup *c = link->cgrp;
615 if (c->root == root) {
616 res = c;
617 break;
618 }
619 }
620 }
621 read_unlock(&css_set_lock);
622 BUG_ON(!res);
623 return res;
624}
625
626/*
521 * There is one global cgroup mutex. We also require taking 627 * There is one global cgroup mutex. We also require taking
522 * task_lock() when dereferencing a task's cgroup subsys pointers. 628 * task_lock() when dereferencing a task's cgroup subsys pointers.
523 * See "The task_lock() exception", at the end of this comment. 629 * See "The task_lock() exception", at the end of this comment.
@@ -677,6 +783,12 @@ static void cgroup_diput(struct dentry *dentry, struct inode *inode)
677 */ 783 */
678 deactivate_super(cgrp->root->sb); 784 deactivate_super(cgrp->root->sb);
679 785
786 /*
787 * if we're getting rid of the cgroup, refcount should ensure
788 * that there are no pidlists left.
789 */
790 BUG_ON(!list_empty(&cgrp->pidlists));
791
680 call_rcu(&cgrp->rcu_head, free_cgroup_rcu); 792 call_rcu(&cgrp->rcu_head, free_cgroup_rcu);
681 } 793 }
682 iput(inode); 794 iput(inode);
@@ -841,6 +953,8 @@ static int cgroup_show_options(struct seq_file *seq, struct vfsmount *vfs)
841 seq_puts(seq, ",noprefix"); 953 seq_puts(seq, ",noprefix");
842 if (strlen(root->release_agent_path)) 954 if (strlen(root->release_agent_path))
843 seq_printf(seq, ",release_agent=%s", root->release_agent_path); 955 seq_printf(seq, ",release_agent=%s", root->release_agent_path);
956 if (strlen(root->name))
957 seq_printf(seq, ",name=%s", root->name);
844 mutex_unlock(&cgroup_mutex); 958 mutex_unlock(&cgroup_mutex);
845 return 0; 959 return 0;
846} 960}
@@ -849,6 +963,12 @@ struct cgroup_sb_opts {
849 unsigned long subsys_bits; 963 unsigned long subsys_bits;
850 unsigned long flags; 964 unsigned long flags;
851 char *release_agent; 965 char *release_agent;
966 char *name;
967 /* User explicitly requested empty subsystem */
968 bool none;
969
970 struct cgroupfs_root *new_root;
971
852}; 972};
853 973
854/* Convert a hierarchy specifier into a bitmask of subsystems and 974/* Convert a hierarchy specifier into a bitmask of subsystems and
@@ -863,9 +983,7 @@ static int parse_cgroupfs_options(char *data,
863 mask = ~(1UL << cpuset_subsys_id); 983 mask = ~(1UL << cpuset_subsys_id);
864#endif 984#endif
865 985
866 opts->subsys_bits = 0; 986 memset(opts, 0, sizeof(*opts));
867 opts->flags = 0;
868 opts->release_agent = NULL;
869 987
870 while ((token = strsep(&o, ",")) != NULL) { 988 while ((token = strsep(&o, ",")) != NULL) {
871 if (!*token) 989 if (!*token)
@@ -879,17 +997,42 @@ static int parse_cgroupfs_options(char *data,
879 if (!ss->disabled) 997 if (!ss->disabled)
880 opts->subsys_bits |= 1ul << i; 998 opts->subsys_bits |= 1ul << i;
881 } 999 }
1000 } else if (!strcmp(token, "none")) {
1001 /* Explicitly have no subsystems */
1002 opts->none = true;
882 } else if (!strcmp(token, "noprefix")) { 1003 } else if (!strcmp(token, "noprefix")) {
883 set_bit(ROOT_NOPREFIX, &opts->flags); 1004 set_bit(ROOT_NOPREFIX, &opts->flags);
884 } else if (!strncmp(token, "release_agent=", 14)) { 1005 } else if (!strncmp(token, "release_agent=", 14)) {
885 /* Specifying two release agents is forbidden */ 1006 /* Specifying two release agents is forbidden */
886 if (opts->release_agent) 1007 if (opts->release_agent)
887 return -EINVAL; 1008 return -EINVAL;
888 opts->release_agent = kzalloc(PATH_MAX, GFP_KERNEL); 1009 opts->release_agent =
1010 kstrndup(token + 14, PATH_MAX, GFP_KERNEL);
889 if (!opts->release_agent) 1011 if (!opts->release_agent)
890 return -ENOMEM; 1012 return -ENOMEM;
891 strncpy(opts->release_agent, token + 14, PATH_MAX - 1); 1013 } else if (!strncmp(token, "name=", 5)) {
892 opts->release_agent[PATH_MAX - 1] = 0; 1014 int i;
1015 const char *name = token + 5;
1016 /* Can't specify an empty name */
1017 if (!strlen(name))
1018 return -EINVAL;
1019 /* Must match [\w.-]+ */
1020 for (i = 0; i < strlen(name); i++) {
1021 char c = name[i];
1022 if (isalnum(c))
1023 continue;
1024 if ((c == '.') || (c == '-') || (c == '_'))
1025 continue;
1026 return -EINVAL;
1027 }
1028 /* Specifying two names is forbidden */
1029 if (opts->name)
1030 return -EINVAL;
1031 opts->name = kstrndup(name,
1032 MAX_CGROUP_ROOT_NAMELEN,
1033 GFP_KERNEL);
1034 if (!opts->name)
1035 return -ENOMEM;
893 } else { 1036 } else {
894 struct cgroup_subsys *ss; 1037 struct cgroup_subsys *ss;
895 int i; 1038 int i;
@@ -906,6 +1049,8 @@ static int parse_cgroupfs_options(char *data,
906 } 1049 }
907 } 1050 }
908 1051
1052 /* Consistency checks */
1053
909 /* 1054 /*
910 * Option noprefix was introduced just for backward compatibility 1055 * Option noprefix was introduced just for backward compatibility
911 * with the old cpuset, so we allow noprefix only if mounting just 1056 * with the old cpuset, so we allow noprefix only if mounting just
@@ -915,8 +1060,16 @@ static int parse_cgroupfs_options(char *data,
915 (opts->subsys_bits & mask)) 1060 (opts->subsys_bits & mask))
916 return -EINVAL; 1061 return -EINVAL;
917 1062
918 /* We can't have an empty hierarchy */ 1063
919 if (!opts->subsys_bits) 1064 /* Can't specify "none" and some subsystems */
1065 if (opts->subsys_bits && opts->none)
1066 return -EINVAL;
1067
1068 /*
1069 * We either have to specify by name or by subsystems. (So all
1070 * empty hierarchies must have a name).
1071 */
1072 if (!opts->subsys_bits && !opts->name)
920 return -EINVAL; 1073 return -EINVAL;
921 1074
922 return 0; 1075 return 0;
@@ -944,6 +1097,12 @@ static int cgroup_remount(struct super_block *sb, int *flags, char *data)
944 goto out_unlock; 1097 goto out_unlock;
945 } 1098 }
946 1099
1100 /* Don't allow name to change at remount */
1101 if (opts.name && strcmp(opts.name, root->name)) {
1102 ret = -EINVAL;
1103 goto out_unlock;
1104 }
1105
947 ret = rebind_subsystems(root, opts.subsys_bits); 1106 ret = rebind_subsystems(root, opts.subsys_bits);
948 if (ret) 1107 if (ret)
949 goto out_unlock; 1108 goto out_unlock;
@@ -955,6 +1114,7 @@ static int cgroup_remount(struct super_block *sb, int *flags, char *data)
955 strcpy(root->release_agent_path, opts.release_agent); 1114 strcpy(root->release_agent_path, opts.release_agent);
956 out_unlock: 1115 out_unlock:
957 kfree(opts.release_agent); 1116 kfree(opts.release_agent);
1117 kfree(opts.name);
958 mutex_unlock(&cgroup_mutex); 1118 mutex_unlock(&cgroup_mutex);
959 mutex_unlock(&cgrp->dentry->d_inode->i_mutex); 1119 mutex_unlock(&cgrp->dentry->d_inode->i_mutex);
960 unlock_kernel(); 1120 unlock_kernel();
@@ -974,9 +1134,10 @@ static void init_cgroup_housekeeping(struct cgroup *cgrp)
974 INIT_LIST_HEAD(&cgrp->children); 1134 INIT_LIST_HEAD(&cgrp->children);
975 INIT_LIST_HEAD(&cgrp->css_sets); 1135 INIT_LIST_HEAD(&cgrp->css_sets);
976 INIT_LIST_HEAD(&cgrp->release_list); 1136 INIT_LIST_HEAD(&cgrp->release_list);
977 INIT_LIST_HEAD(&cgrp->pids_list); 1137 INIT_LIST_HEAD(&cgrp->pidlists);
978 init_rwsem(&cgrp->pids_mutex); 1138 mutex_init(&cgrp->pidlist_mutex);
979} 1139}
1140
980static void init_cgroup_root(struct cgroupfs_root *root) 1141static void init_cgroup_root(struct cgroupfs_root *root)
981{ 1142{
982 struct cgroup *cgrp = &root->top_cgroup; 1143 struct cgroup *cgrp = &root->top_cgroup;
@@ -988,33 +1149,106 @@ static void init_cgroup_root(struct cgroupfs_root *root)
988 init_cgroup_housekeeping(cgrp); 1149 init_cgroup_housekeeping(cgrp);
989} 1150}
990 1151
1152static bool init_root_id(struct cgroupfs_root *root)
1153{
1154 int ret = 0;
1155
1156 do {
1157 if (!ida_pre_get(&hierarchy_ida, GFP_KERNEL))
1158 return false;
1159 spin_lock(&hierarchy_id_lock);
1160 /* Try to allocate the next unused ID */
1161 ret = ida_get_new_above(&hierarchy_ida, next_hierarchy_id,
1162 &root->hierarchy_id);
1163 if (ret == -ENOSPC)
1164 /* Try again starting from 0 */
1165 ret = ida_get_new(&hierarchy_ida, &root->hierarchy_id);
1166 if (!ret) {
1167 next_hierarchy_id = root->hierarchy_id + 1;
1168 } else if (ret != -EAGAIN) {
1169 /* Can only get here if the 31-bit IDR is full ... */
1170 BUG_ON(ret);
1171 }
1172 spin_unlock(&hierarchy_id_lock);
1173 } while (ret);
1174 return true;
1175}
1176
991static int cgroup_test_super(struct super_block *sb, void *data) 1177static int cgroup_test_super(struct super_block *sb, void *data)
992{ 1178{
993 struct cgroupfs_root *new = data; 1179 struct cgroup_sb_opts *opts = data;
994 struct cgroupfs_root *root = sb->s_fs_info; 1180 struct cgroupfs_root *root = sb->s_fs_info;
995 1181
996 /* First check subsystems */ 1182 /* If we asked for a name then it must match */
997 if (new->subsys_bits != root->subsys_bits) 1183 if (opts->name && strcmp(opts->name, root->name))
998 return 0; 1184 return 0;
999 1185
1000 /* Next check flags */ 1186 /*
1001 if (new->flags != root->flags) 1187 * If we asked for subsystems (or explicitly for no
1188 * subsystems) then they must match
1189 */
1190 if ((opts->subsys_bits || opts->none)
1191 && (opts->subsys_bits != root->subsys_bits))
1002 return 0; 1192 return 0;
1003 1193
1004 return 1; 1194 return 1;
1005} 1195}
1006 1196
1197static struct cgroupfs_root *cgroup_root_from_opts(struct cgroup_sb_opts *opts)
1198{
1199 struct cgroupfs_root *root;
1200
1201 if (!opts->subsys_bits && !opts->none)
1202 return NULL;
1203
1204 root = kzalloc(sizeof(*root), GFP_KERNEL);
1205 if (!root)
1206 return ERR_PTR(-ENOMEM);
1207
1208 if (!init_root_id(root)) {
1209 kfree(root);
1210 return ERR_PTR(-ENOMEM);
1211 }
1212 init_cgroup_root(root);
1213
1214 root->subsys_bits = opts->subsys_bits;
1215 root->flags = opts->flags;
1216 if (opts->release_agent)
1217 strcpy(root->release_agent_path, opts->release_agent);
1218 if (opts->name)
1219 strcpy(root->name, opts->name);
1220 return root;
1221}
1222
1223static void cgroup_drop_root(struct cgroupfs_root *root)
1224{
1225 if (!root)
1226 return;
1227
1228 BUG_ON(!root->hierarchy_id);
1229 spin_lock(&hierarchy_id_lock);
1230 ida_remove(&hierarchy_ida, root->hierarchy_id);
1231 spin_unlock(&hierarchy_id_lock);
1232 kfree(root);
1233}
1234
1007static int cgroup_set_super(struct super_block *sb, void *data) 1235static int cgroup_set_super(struct super_block *sb, void *data)
1008{ 1236{
1009 int ret; 1237 int ret;
1010 struct cgroupfs_root *root = data; 1238 struct cgroup_sb_opts *opts = data;
1239
1240 /* If we don't have a new root, we can't set up a new sb */
1241 if (!opts->new_root)
1242 return -EINVAL;
1243
1244 BUG_ON(!opts->subsys_bits && !opts->none);
1011 1245
1012 ret = set_anon_super(sb, NULL); 1246 ret = set_anon_super(sb, NULL);
1013 if (ret) 1247 if (ret)
1014 return ret; 1248 return ret;
1015 1249
1016 sb->s_fs_info = root; 1250 sb->s_fs_info = opts->new_root;
1017 root->sb = sb; 1251 opts->new_root->sb = sb;
1018 1252
1019 sb->s_blocksize = PAGE_CACHE_SIZE; 1253 sb->s_blocksize = PAGE_CACHE_SIZE;
1020 sb->s_blocksize_bits = PAGE_CACHE_SHIFT; 1254 sb->s_blocksize_bits = PAGE_CACHE_SHIFT;
@@ -1051,48 +1285,43 @@ static int cgroup_get_sb(struct file_system_type *fs_type,
1051 void *data, struct vfsmount *mnt) 1285 void *data, struct vfsmount *mnt)
1052{ 1286{
1053 struct cgroup_sb_opts opts; 1287 struct cgroup_sb_opts opts;
1288 struct cgroupfs_root *root;
1054 int ret = 0; 1289 int ret = 0;
1055 struct super_block *sb; 1290 struct super_block *sb;
1056 struct cgroupfs_root *root; 1291 struct cgroupfs_root *new_root;
1057 struct list_head tmp_cg_links;
1058 1292
1059 /* First find the desired set of subsystems */ 1293 /* First find the desired set of subsystems */
1060 ret = parse_cgroupfs_options(data, &opts); 1294 ret = parse_cgroupfs_options(data, &opts);
1061 if (ret) { 1295 if (ret)
1062 kfree(opts.release_agent); 1296 goto out_err;
1063 return ret;
1064 }
1065
1066 root = kzalloc(sizeof(*root), GFP_KERNEL);
1067 if (!root) {
1068 kfree(opts.release_agent);
1069 return -ENOMEM;
1070 }
1071 1297
1072 init_cgroup_root(root); 1298 /*
1073 root->subsys_bits = opts.subsys_bits; 1299 * Allocate a new cgroup root. We may not need it if we're
1074 root->flags = opts.flags; 1300 * reusing an existing hierarchy.
1075 if (opts.release_agent) { 1301 */
1076 strcpy(root->release_agent_path, opts.release_agent); 1302 new_root = cgroup_root_from_opts(&opts);
1077 kfree(opts.release_agent); 1303 if (IS_ERR(new_root)) {
1304 ret = PTR_ERR(new_root);
1305 goto out_err;
1078 } 1306 }
1307 opts.new_root = new_root;
1079 1308
1080 sb = sget(fs_type, cgroup_test_super, cgroup_set_super, root); 1309 /* Locate an existing or new sb for this hierarchy */
1081 1310 sb = sget(fs_type, cgroup_test_super, cgroup_set_super, &opts);
1082 if (IS_ERR(sb)) { 1311 if (IS_ERR(sb)) {
1083 kfree(root); 1312 ret = PTR_ERR(sb);
1084 return PTR_ERR(sb); 1313 cgroup_drop_root(opts.new_root);
1314 goto out_err;
1085 } 1315 }
1086 1316
1087 if (sb->s_fs_info != root) { 1317 root = sb->s_fs_info;
1088 /* Reusing an existing superblock */ 1318 BUG_ON(!root);
1089 BUG_ON(sb->s_root == NULL); 1319 if (root == opts.new_root) {
1090 kfree(root); 1320 /* We used the new root structure, so this is a new hierarchy */
1091 root = NULL; 1321 struct list_head tmp_cg_links;
1092 } else {
1093 /* New superblock */
1094 struct cgroup *root_cgrp = &root->top_cgroup; 1322 struct cgroup *root_cgrp = &root->top_cgroup;
1095 struct inode *inode; 1323 struct inode *inode;
1324 struct cgroupfs_root *existing_root;
1096 int i; 1325 int i;
1097 1326
1098 BUG_ON(sb->s_root != NULL); 1327 BUG_ON(sb->s_root != NULL);
@@ -1105,6 +1334,18 @@ static int cgroup_get_sb(struct file_system_type *fs_type,
1105 mutex_lock(&inode->i_mutex); 1334 mutex_lock(&inode->i_mutex);
1106 mutex_lock(&cgroup_mutex); 1335 mutex_lock(&cgroup_mutex);
1107 1336
1337 if (strlen(root->name)) {
1338 /* Check for name clashes with existing mounts */
1339 for_each_active_root(existing_root) {
1340 if (!strcmp(existing_root->name, root->name)) {
1341 ret = -EBUSY;
1342 mutex_unlock(&cgroup_mutex);
1343 mutex_unlock(&inode->i_mutex);
1344 goto drop_new_super;
1345 }
1346 }
1347 }
1348
1108 /* 1349 /*
1109 * We're accessing css_set_count without locking 1350 * We're accessing css_set_count without locking
1110 * css_set_lock here, but that's OK - it can only be 1351 * css_set_lock here, but that's OK - it can only be
@@ -1123,7 +1364,8 @@ static int cgroup_get_sb(struct file_system_type *fs_type,
1123 if (ret == -EBUSY) { 1364 if (ret == -EBUSY) {
1124 mutex_unlock(&cgroup_mutex); 1365 mutex_unlock(&cgroup_mutex);
1125 mutex_unlock(&inode->i_mutex); 1366 mutex_unlock(&inode->i_mutex);
1126 goto free_cg_links; 1367 free_cg_links(&tmp_cg_links);
1368 goto drop_new_super;
1127 } 1369 }
1128 1370
1129 /* EBUSY should be the only error here */ 1371 /* EBUSY should be the only error here */
@@ -1155,17 +1397,27 @@ static int cgroup_get_sb(struct file_system_type *fs_type,
1155 BUG_ON(root->number_of_cgroups != 1); 1397 BUG_ON(root->number_of_cgroups != 1);
1156 1398
1157 cgroup_populate_dir(root_cgrp); 1399 cgroup_populate_dir(root_cgrp);
1158 mutex_unlock(&inode->i_mutex);
1159 mutex_unlock(&cgroup_mutex); 1400 mutex_unlock(&cgroup_mutex);
1401 mutex_unlock(&inode->i_mutex);
1402 } else {
1403 /*
1404 * We re-used an existing hierarchy - the new root (if
1405 * any) is not needed
1406 */
1407 cgroup_drop_root(opts.new_root);
1160 } 1408 }
1161 1409
1162 simple_set_mnt(mnt, sb); 1410 simple_set_mnt(mnt, sb);
1411 kfree(opts.release_agent);
1412 kfree(opts.name);
1163 return 0; 1413 return 0;
1164 1414
1165 free_cg_links:
1166 free_cg_links(&tmp_cg_links);
1167 drop_new_super: 1415 drop_new_super:
1168 deactivate_locked_super(sb); 1416 deactivate_locked_super(sb);
1417 out_err:
1418 kfree(opts.release_agent);
1419 kfree(opts.name);
1420
1169 return ret; 1421 return ret;
1170} 1422}
1171 1423
@@ -1211,7 +1463,7 @@ static void cgroup_kill_sb(struct super_block *sb) {
1211 mutex_unlock(&cgroup_mutex); 1463 mutex_unlock(&cgroup_mutex);
1212 1464
1213 kill_litter_super(sb); 1465 kill_litter_super(sb);
1214 kfree(root); 1466 cgroup_drop_root(root);
1215} 1467}
1216 1468
1217static struct file_system_type cgroup_fs_type = { 1469static struct file_system_type cgroup_fs_type = {
@@ -1276,27 +1528,6 @@ int cgroup_path(const struct cgroup *cgrp, char *buf, int buflen)
1276 return 0; 1528 return 0;
1277} 1529}
1278 1530
1279/*
1280 * Return the first subsystem attached to a cgroup's hierarchy, and
1281 * its subsystem id.
1282 */
1283
1284static void get_first_subsys(const struct cgroup *cgrp,
1285 struct cgroup_subsys_state **css, int *subsys_id)
1286{
1287 const struct cgroupfs_root *root = cgrp->root;
1288 const struct cgroup_subsys *test_ss;
1289 BUG_ON(list_empty(&root->subsys_list));
1290 test_ss = list_entry(root->subsys_list.next,
1291 struct cgroup_subsys, sibling);
1292 if (css) {
1293 *css = cgrp->subsys[test_ss->subsys_id];
1294 BUG_ON(!*css);
1295 }
1296 if (subsys_id)
1297 *subsys_id = test_ss->subsys_id;
1298}
1299
1300/** 1531/**
1301 * cgroup_attach_task - attach task 'tsk' to cgroup 'cgrp' 1532 * cgroup_attach_task - attach task 'tsk' to cgroup 'cgrp'
1302 * @cgrp: the cgroup the task is attaching to 1533 * @cgrp: the cgroup the task is attaching to
@@ -1313,18 +1544,15 @@ int cgroup_attach_task(struct cgroup *cgrp, struct task_struct *tsk)
1313 struct css_set *cg; 1544 struct css_set *cg;
1314 struct css_set *newcg; 1545 struct css_set *newcg;
1315 struct cgroupfs_root *root = cgrp->root; 1546 struct cgroupfs_root *root = cgrp->root;
1316 int subsys_id;
1317
1318 get_first_subsys(cgrp, NULL, &subsys_id);
1319 1547
1320 /* Nothing to do if the task is already in that cgroup */ 1548 /* Nothing to do if the task is already in that cgroup */
1321 oldcgrp = task_cgroup(tsk, subsys_id); 1549 oldcgrp = task_cgroup_from_root(tsk, root);
1322 if (cgrp == oldcgrp) 1550 if (cgrp == oldcgrp)
1323 return 0; 1551 return 0;
1324 1552
1325 for_each_subsys(root, ss) { 1553 for_each_subsys(root, ss) {
1326 if (ss->can_attach) { 1554 if (ss->can_attach) {
1327 retval = ss->can_attach(ss, cgrp, tsk); 1555 retval = ss->can_attach(ss, cgrp, tsk, false);
1328 if (retval) 1556 if (retval)
1329 return retval; 1557 return retval;
1330 } 1558 }
@@ -1362,7 +1590,7 @@ int cgroup_attach_task(struct cgroup *cgrp, struct task_struct *tsk)
1362 1590
1363 for_each_subsys(root, ss) { 1591 for_each_subsys(root, ss) {
1364 if (ss->attach) 1592 if (ss->attach)
1365 ss->attach(ss, cgrp, oldcgrp, tsk); 1593 ss->attach(ss, cgrp, oldcgrp, tsk, false);
1366 } 1594 }
1367 set_bit(CGRP_RELEASABLE, &oldcgrp->flags); 1595 set_bit(CGRP_RELEASABLE, &oldcgrp->flags);
1368 synchronize_rcu(); 1596 synchronize_rcu();
@@ -1423,15 +1651,6 @@ static int cgroup_tasks_write(struct cgroup *cgrp, struct cftype *cft, u64 pid)
1423 return ret; 1651 return ret;
1424} 1652}
1425 1653
1426/* The various types of files and directories in a cgroup file system */
1427enum cgroup_filetype {
1428 FILE_ROOT,
1429 FILE_DIR,
1430 FILE_TASKLIST,
1431 FILE_NOTIFY_ON_RELEASE,
1432 FILE_RELEASE_AGENT,
1433};
1434
1435/** 1654/**
1436 * cgroup_lock_live_group - take cgroup_mutex and check that cgrp is alive. 1655 * cgroup_lock_live_group - take cgroup_mutex and check that cgrp is alive.
1437 * @cgrp: the cgroup to be checked for liveness 1656 * @cgrp: the cgroup to be checked for liveness
@@ -1876,7 +2095,7 @@ int cgroup_task_count(const struct cgroup *cgrp)
1876 * the start of a css_set 2095 * the start of a css_set
1877 */ 2096 */
1878static void cgroup_advance_iter(struct cgroup *cgrp, 2097static void cgroup_advance_iter(struct cgroup *cgrp,
1879 struct cgroup_iter *it) 2098 struct cgroup_iter *it)
1880{ 2099{
1881 struct list_head *l = it->cg_link; 2100 struct list_head *l = it->cg_link;
1882 struct cg_cgroup_link *link; 2101 struct cg_cgroup_link *link;
@@ -2129,7 +2348,7 @@ int cgroup_scan_tasks(struct cgroup_scanner *scan)
2129} 2348}
2130 2349
2131/* 2350/*
2132 * Stuff for reading the 'tasks' file. 2351 * Stuff for reading the 'tasks'/'procs' files.
2133 * 2352 *
2134 * Reading this file can return large amounts of data if a cgroup has 2353 * Reading this file can return large amounts of data if a cgroup has
2135 * *lots* of attached tasks. So it may need several calls to read(), 2354 * *lots* of attached tasks. So it may need several calls to read(),
@@ -2139,27 +2358,196 @@ int cgroup_scan_tasks(struct cgroup_scanner *scan)
2139 */ 2358 */
2140 2359
2141/* 2360/*
2142 * Load into 'pidarray' up to 'npids' of the tasks using cgroup 2361 * The following two functions "fix" the issue where there are more pids
2143 * 'cgrp'. Return actual number of pids loaded. No need to 2362 * than kmalloc will give memory for; in such cases, we use vmalloc/vfree.
2144 * task_lock(p) when reading out p->cgroup, since we're in an RCU 2363 * TODO: replace with a kernel-wide solution to this problem
2145 * read section, so the css_set can't go away, and is 2364 */
2146 * immutable after creation. 2365#define PIDLIST_TOO_LARGE(c) ((c) * sizeof(pid_t) > (PAGE_SIZE * 2))
2366static void *pidlist_allocate(int count)
2367{
2368 if (PIDLIST_TOO_LARGE(count))
2369 return vmalloc(count * sizeof(pid_t));
2370 else
2371 return kmalloc(count * sizeof(pid_t), GFP_KERNEL);
2372}
2373static void pidlist_free(void *p)
2374{
2375 if (is_vmalloc_addr(p))
2376 vfree(p);
2377 else
2378 kfree(p);
2379}
2380static void *pidlist_resize(void *p, int newcount)
2381{
2382 void *newlist;
2383 /* note: if new alloc fails, old p will still be valid either way */
2384 if (is_vmalloc_addr(p)) {
2385 newlist = vmalloc(newcount * sizeof(pid_t));
2386 if (!newlist)
2387 return NULL;
2388 memcpy(newlist, p, newcount * sizeof(pid_t));
2389 vfree(p);
2390 } else {
2391 newlist = krealloc(p, newcount * sizeof(pid_t), GFP_KERNEL);
2392 }
2393 return newlist;
2394}
2395
2396/*
2397 * pidlist_uniq - given a kmalloc()ed list, strip out all duplicate entries
2398 * If the new stripped list is sufficiently smaller and there's enough memory
2399 * to allocate a new buffer, will let go of the unneeded memory. Returns the
2400 * number of unique elements.
2401 */
2402/* is the size difference enough that we should re-allocate the array? */
2403#define PIDLIST_REALLOC_DIFFERENCE(old, new) ((old) - PAGE_SIZE >= (new))
2404static int pidlist_uniq(pid_t **p, int length)
2405{
2406 int src, dest = 1;
2407 pid_t *list = *p;
2408 pid_t *newlist;
2409
2410 /*
2411 * we presume the 0th element is unique, so i starts at 1. trivial
2412 * edge cases first; no work needs to be done for either
2413 */
2414 if (length == 0 || length == 1)
2415 return length;
2416 /* src and dest walk down the list; dest counts unique elements */
2417 for (src = 1; src < length; src++) {
2418 /* find next unique element */
2419 while (list[src] == list[src-1]) {
2420 src++;
2421 if (src == length)
2422 goto after;
2423 }
2424 /* dest always points to where the next unique element goes */
2425 list[dest] = list[src];
2426 dest++;
2427 }
2428after:
2429 /*
2430 * if the length difference is large enough, we want to allocate a
2431 * smaller buffer to save memory. if this fails due to out of memory,
2432 * we'll just stay with what we've got.
2433 */
2434 if (PIDLIST_REALLOC_DIFFERENCE(length, dest)) {
2435 newlist = pidlist_resize(list, dest);
2436 if (newlist)
2437 *p = newlist;
2438 }
2439 return dest;
2440}
2441
2442static int cmppid(const void *a, const void *b)
2443{
2444 return *(pid_t *)a - *(pid_t *)b;
2445}
2446
2447/*
2448 * find the appropriate pidlist for our purpose (given procs vs tasks)
2449 * returns with the lock on that pidlist already held, and takes care
2450 * of the use count, or returns NULL with no locks held if we're out of
2451 * memory.
2147 */ 2452 */
2148static int pid_array_load(pid_t *pidarray, int npids, struct cgroup *cgrp) 2453static struct cgroup_pidlist *cgroup_pidlist_find(struct cgroup *cgrp,
2454 enum cgroup_filetype type)
2149{ 2455{
2150 int n = 0, pid; 2456 struct cgroup_pidlist *l;
2457 /* don't need task_nsproxy() if we're looking at ourself */
2458 struct pid_namespace *ns = get_pid_ns(current->nsproxy->pid_ns);
2459 /*
2460 * We can't drop the pidlist_mutex before taking the l->mutex in case
2461 * the last ref-holder is trying to remove l from the list at the same
2462 * time. Holding the pidlist_mutex precludes somebody taking whichever
2463 * list we find out from under us - compare release_pid_array().
2464 */
2465 mutex_lock(&cgrp->pidlist_mutex);
2466 list_for_each_entry(l, &cgrp->pidlists, links) {
2467 if (l->key.type == type && l->key.ns == ns) {
2468 /* found a matching list - drop the extra refcount */
2469 put_pid_ns(ns);
2470 /* make sure l doesn't vanish out from under us */
2471 down_write(&l->mutex);
2472 mutex_unlock(&cgrp->pidlist_mutex);
2473 l->use_count++;
2474 return l;
2475 }
2476 }
2477 /* entry not found; create a new one */
2478 l = kmalloc(sizeof(struct cgroup_pidlist), GFP_KERNEL);
2479 if (!l) {
2480 mutex_unlock(&cgrp->pidlist_mutex);
2481 put_pid_ns(ns);
2482 return l;
2483 }
2484 init_rwsem(&l->mutex);
2485 down_write(&l->mutex);
2486 l->key.type = type;
2487 l->key.ns = ns;
2488 l->use_count = 0; /* don't increment here */
2489 l->list = NULL;
2490 l->owner = cgrp;
2491 list_add(&l->links, &cgrp->pidlists);
2492 mutex_unlock(&cgrp->pidlist_mutex);
2493 return l;
2494}
2495
2496/*
2497 * Load a cgroup's pidarray with either procs' tgids or tasks' pids
2498 */
2499static int pidlist_array_load(struct cgroup *cgrp, enum cgroup_filetype type,
2500 struct cgroup_pidlist **lp)
2501{
2502 pid_t *array;
2503 int length;
2504 int pid, n = 0; /* used for populating the array */
2151 struct cgroup_iter it; 2505 struct cgroup_iter it;
2152 struct task_struct *tsk; 2506 struct task_struct *tsk;
2507 struct cgroup_pidlist *l;
2508
2509 /*
2510 * If cgroup gets more users after we read count, we won't have
2511 * enough space - tough. This race is indistinguishable to the
2512 * caller from the case that the additional cgroup users didn't
2513 * show up until sometime later on.
2514 */
2515 length = cgroup_task_count(cgrp);
2516 array = pidlist_allocate(length);
2517 if (!array)
2518 return -ENOMEM;
2519 /* now, populate the array */
2153 cgroup_iter_start(cgrp, &it); 2520 cgroup_iter_start(cgrp, &it);
2154 while ((tsk = cgroup_iter_next(cgrp, &it))) { 2521 while ((tsk = cgroup_iter_next(cgrp, &it))) {
2155 if (unlikely(n == npids)) 2522 if (unlikely(n == length))
2156 break; 2523 break;
2157 pid = task_pid_vnr(tsk); 2524 /* get tgid or pid for procs or tasks file respectively */
2158 if (pid > 0) 2525 if (type == CGROUP_FILE_PROCS)
2159 pidarray[n++] = pid; 2526 pid = task_tgid_vnr(tsk);
2527 else
2528 pid = task_pid_vnr(tsk);
2529 if (pid > 0) /* make sure to only use valid results */
2530 array[n++] = pid;
2160 } 2531 }
2161 cgroup_iter_end(cgrp, &it); 2532 cgroup_iter_end(cgrp, &it);
2162 return n; 2533 length = n;
2534 /* now sort & (if procs) strip out duplicates */
2535 sort(array, length, sizeof(pid_t), cmppid, NULL);
2536 if (type == CGROUP_FILE_PROCS)
2537 length = pidlist_uniq(&array, length);
2538 l = cgroup_pidlist_find(cgrp, type);
2539 if (!l) {
2540 pidlist_free(array);
2541 return -ENOMEM;
2542 }
2543 /* store array, freeing old if necessary - lock already held */
2544 pidlist_free(l->list);
2545 l->list = array;
2546 l->length = length;
2547 l->use_count++;
2548 up_write(&l->mutex);
2549 *lp = l;
2550 return 0;
2163} 2551}
2164 2552
2165/** 2553/**
@@ -2216,37 +2604,14 @@ err:
2216 return ret; 2604 return ret;
2217} 2605}
2218 2606
2219/*
2220 * Cache pids for all threads in the same pid namespace that are
2221 * opening the same "tasks" file.
2222 */
2223struct cgroup_pids {
2224 /* The node in cgrp->pids_list */
2225 struct list_head list;
2226 /* The cgroup those pids belong to */
2227 struct cgroup *cgrp;
2228 /* The namepsace those pids belong to */
2229 struct pid_namespace *ns;
2230 /* Array of process ids in the cgroup */
2231 pid_t *tasks_pids;
2232 /* How many files are using the this tasks_pids array */
2233 int use_count;
2234 /* Length of the current tasks_pids array */
2235 int length;
2236};
2237
2238static int cmppid(const void *a, const void *b)
2239{
2240 return *(pid_t *)a - *(pid_t *)b;
2241}
2242 2607
2243/* 2608/*
2244 * seq_file methods for the "tasks" file. The seq_file position is the 2609 * seq_file methods for the tasks/procs files. The seq_file position is the
2245 * next pid to display; the seq_file iterator is a pointer to the pid 2610 * next pid to display; the seq_file iterator is a pointer to the pid
2246 * in the cgroup->tasks_pids array. 2611 * in the cgroup->l->list array.
2247 */ 2612 */
2248 2613
2249static void *cgroup_tasks_start(struct seq_file *s, loff_t *pos) 2614static void *cgroup_pidlist_start(struct seq_file *s, loff_t *pos)
2250{ 2615{
2251 /* 2616 /*
2252 * Initially we receive a position value that corresponds to 2617 * Initially we receive a position value that corresponds to
@@ -2254,48 +2619,45 @@ static void *cgroup_tasks_start(struct seq_file *s, loff_t *pos)
2254 * after a seek to the start). Use a binary-search to find the 2619 * after a seek to the start). Use a binary-search to find the
2255 * next pid to display, if any 2620 * next pid to display, if any
2256 */ 2621 */
2257 struct cgroup_pids *cp = s->private; 2622 struct cgroup_pidlist *l = s->private;
2258 struct cgroup *cgrp = cp->cgrp;
2259 int index = 0, pid = *pos; 2623 int index = 0, pid = *pos;
2260 int *iter; 2624 int *iter;
2261 2625
2262 down_read(&cgrp->pids_mutex); 2626 down_read(&l->mutex);
2263 if (pid) { 2627 if (pid) {
2264 int end = cp->length; 2628 int end = l->length;
2265 2629
2266 while (index < end) { 2630 while (index < end) {
2267 int mid = (index + end) / 2; 2631 int mid = (index + end) / 2;
2268 if (cp->tasks_pids[mid] == pid) { 2632 if (l->list[mid] == pid) {
2269 index = mid; 2633 index = mid;
2270 break; 2634 break;
2271 } else if (cp->tasks_pids[mid] <= pid) 2635 } else if (l->list[mid] <= pid)
2272 index = mid + 1; 2636 index = mid + 1;
2273 else 2637 else
2274 end = mid; 2638 end = mid;
2275 } 2639 }
2276 } 2640 }
2277 /* If we're off the end of the array, we're done */ 2641 /* If we're off the end of the array, we're done */
2278 if (index >= cp->length) 2642 if (index >= l->length)
2279 return NULL; 2643 return NULL;
2280 /* Update the abstract position to be the actual pid that we found */ 2644 /* Update the abstract position to be the actual pid that we found */
2281 iter = cp->tasks_pids + index; 2645 iter = l->list + index;
2282 *pos = *iter; 2646 *pos = *iter;
2283 return iter; 2647 return iter;
2284} 2648}
2285 2649
2286static void cgroup_tasks_stop(struct seq_file *s, void *v) 2650static void cgroup_pidlist_stop(struct seq_file *s, void *v)
2287{ 2651{
2288 struct cgroup_pids *cp = s->private; 2652 struct cgroup_pidlist *l = s->private;
2289 struct cgroup *cgrp = cp->cgrp; 2653 up_read(&l->mutex);
2290 up_read(&cgrp->pids_mutex);
2291} 2654}
2292 2655
2293static void *cgroup_tasks_next(struct seq_file *s, void *v, loff_t *pos) 2656static void *cgroup_pidlist_next(struct seq_file *s, void *v, loff_t *pos)
2294{ 2657{
2295 struct cgroup_pids *cp = s->private; 2658 struct cgroup_pidlist *l = s->private;
2296 int *p = v; 2659 pid_t *p = v;
2297 int *end = cp->tasks_pids + cp->length; 2660 pid_t *end = l->list + l->length;
2298
2299 /* 2661 /*
2300 * Advance to the next pid in the array. If this goes off the 2662 * Advance to the next pid in the array. If this goes off the
2301 * end, we're done 2663 * end, we're done
@@ -2309,124 +2671,107 @@ static void *cgroup_tasks_next(struct seq_file *s, void *v, loff_t *pos)
2309 } 2671 }
2310} 2672}
2311 2673
2312static int cgroup_tasks_show(struct seq_file *s, void *v) 2674static int cgroup_pidlist_show(struct seq_file *s, void *v)
2313{ 2675{
2314 return seq_printf(s, "%d\n", *(int *)v); 2676 return seq_printf(s, "%d\n", *(int *)v);
2315} 2677}
2316 2678
2317static const struct seq_operations cgroup_tasks_seq_operations = { 2679/*
2318 .start = cgroup_tasks_start, 2680 * seq_operations functions for iterating on pidlists through seq_file -
2319 .stop = cgroup_tasks_stop, 2681 * independent of whether it's tasks or procs
2320 .next = cgroup_tasks_next, 2682 */
2321 .show = cgroup_tasks_show, 2683static const struct seq_operations cgroup_pidlist_seq_operations = {
2684 .start = cgroup_pidlist_start,
2685 .stop = cgroup_pidlist_stop,
2686 .next = cgroup_pidlist_next,
2687 .show = cgroup_pidlist_show,
2322}; 2688};
2323 2689
2324static void release_cgroup_pid_array(struct cgroup_pids *cp) 2690static void cgroup_release_pid_array(struct cgroup_pidlist *l)
2325{ 2691{
2326 struct cgroup *cgrp = cp->cgrp; 2692 /*
2327 2693 * the case where we're the last user of this particular pidlist will
2328 down_write(&cgrp->pids_mutex); 2694 * have us remove it from the cgroup's list, which entails taking the
2329 BUG_ON(!cp->use_count); 2695 * mutex. since in pidlist_find the pidlist->lock depends on cgroup->
2330 if (!--cp->use_count) { 2696 * pidlist_mutex, we have to take pidlist_mutex first.
2331 list_del(&cp->list); 2697 */
2332 put_pid_ns(cp->ns); 2698 mutex_lock(&l->owner->pidlist_mutex);
2333 kfree(cp->tasks_pids); 2699 down_write(&l->mutex);
2334 kfree(cp); 2700 BUG_ON(!l->use_count);
2701 if (!--l->use_count) {
2702 /* we're the last user if refcount is 0; remove and free */
2703 list_del(&l->links);
2704 mutex_unlock(&l->owner->pidlist_mutex);
2705 pidlist_free(l->list);
2706 put_pid_ns(l->key.ns);
2707 up_write(&l->mutex);
2708 kfree(l);
2709 return;
2335 } 2710 }
2336 up_write(&cgrp->pids_mutex); 2711 mutex_unlock(&l->owner->pidlist_mutex);
2712 up_write(&l->mutex);
2337} 2713}
2338 2714
2339static int cgroup_tasks_release(struct inode *inode, struct file *file) 2715static int cgroup_pidlist_release(struct inode *inode, struct file *file)
2340{ 2716{
2341 struct seq_file *seq; 2717 struct cgroup_pidlist *l;
2342 struct cgroup_pids *cp;
2343
2344 if (!(file->f_mode & FMODE_READ)) 2718 if (!(file->f_mode & FMODE_READ))
2345 return 0; 2719 return 0;
2346 2720 /*
2347 seq = file->private_data; 2721 * the seq_file will only be initialized if the file was opened for
2348 cp = seq->private; 2722 * reading; hence we check if it's not null only in that case.
2349 2723 */
2350 release_cgroup_pid_array(cp); 2724 l = ((struct seq_file *)file->private_data)->private;
2725 cgroup_release_pid_array(l);
2351 return seq_release(inode, file); 2726 return seq_release(inode, file);
2352} 2727}
2353 2728
2354static struct file_operations cgroup_tasks_operations = { 2729static const struct file_operations cgroup_pidlist_operations = {
2355 .read = seq_read, 2730 .read = seq_read,
2356 .llseek = seq_lseek, 2731 .llseek = seq_lseek,
2357 .write = cgroup_file_write, 2732 .write = cgroup_file_write,
2358 .release = cgroup_tasks_release, 2733 .release = cgroup_pidlist_release,
2359}; 2734};
2360 2735
2361/* 2736/*
2362 * Handle an open on 'tasks' file. Prepare an array containing the 2737 * The following functions handle opens on a file that displays a pidlist
2363 * process id's of tasks currently attached to the cgroup being opened. 2738 * (tasks or procs). Prepare an array of the process/thread IDs of whoever's
2739 * in the cgroup.
2364 */ 2740 */
2365 2741/* helper function for the two below it */
2366static int cgroup_tasks_open(struct inode *unused, struct file *file) 2742static int cgroup_pidlist_open(struct file *file, enum cgroup_filetype type)
2367{ 2743{
2368 struct cgroup *cgrp = __d_cgrp(file->f_dentry->d_parent); 2744 struct cgroup *cgrp = __d_cgrp(file->f_dentry->d_parent);
2369 struct pid_namespace *ns = current->nsproxy->pid_ns; 2745 struct cgroup_pidlist *l;
2370 struct cgroup_pids *cp;
2371 pid_t *pidarray;
2372 int npids;
2373 int retval; 2746 int retval;
2374 2747
2375 /* Nothing to do for write-only files */ 2748 /* Nothing to do for write-only files */
2376 if (!(file->f_mode & FMODE_READ)) 2749 if (!(file->f_mode & FMODE_READ))
2377 return 0; 2750 return 0;
2378 2751
2379 /* 2752 /* have the array populated */
2380 * If cgroup gets more users after we read count, we won't have 2753 retval = pidlist_array_load(cgrp, type, &l);
2381 * enough space - tough. This race is indistinguishable to the 2754 if (retval)
2382 * caller from the case that the additional cgroup users didn't 2755 return retval;
2383 * show up until sometime later on. 2756 /* configure file information */
2384 */ 2757 file->f_op = &cgroup_pidlist_operations;
2385 npids = cgroup_task_count(cgrp);
2386 pidarray = kmalloc(npids * sizeof(pid_t), GFP_KERNEL);
2387 if (!pidarray)
2388 return -ENOMEM;
2389 npids = pid_array_load(pidarray, npids, cgrp);
2390 sort(pidarray, npids, sizeof(pid_t), cmppid, NULL);
2391
2392 /*
2393 * Store the array in the cgroup, freeing the old
2394 * array if necessary
2395 */
2396 down_write(&cgrp->pids_mutex);
2397
2398 list_for_each_entry(cp, &cgrp->pids_list, list) {
2399 if (ns == cp->ns)
2400 goto found;
2401 }
2402
2403 cp = kzalloc(sizeof(*cp), GFP_KERNEL);
2404 if (!cp) {
2405 up_write(&cgrp->pids_mutex);
2406 kfree(pidarray);
2407 return -ENOMEM;
2408 }
2409 cp->cgrp = cgrp;
2410 cp->ns = ns;
2411 get_pid_ns(ns);
2412 list_add(&cp->list, &cgrp->pids_list);
2413found:
2414 kfree(cp->tasks_pids);
2415 cp->tasks_pids = pidarray;
2416 cp->length = npids;
2417 cp->use_count++;
2418 up_write(&cgrp->pids_mutex);
2419
2420 file->f_op = &cgroup_tasks_operations;
2421 2758
2422 retval = seq_open(file, &cgroup_tasks_seq_operations); 2759 retval = seq_open(file, &cgroup_pidlist_seq_operations);
2423 if (retval) { 2760 if (retval) {
2424 release_cgroup_pid_array(cp); 2761 cgroup_release_pid_array(l);
2425 return retval; 2762 return retval;
2426 } 2763 }
2427 ((struct seq_file *)file->private_data)->private = cp; 2764 ((struct seq_file *)file->private_data)->private = l;
2428 return 0; 2765 return 0;
2429} 2766}
2767static int cgroup_tasks_open(struct inode *unused, struct file *file)
2768{
2769 return cgroup_pidlist_open(file, CGROUP_FILE_TASKS);
2770}
2771static int cgroup_procs_open(struct inode *unused, struct file *file)
2772{
2773 return cgroup_pidlist_open(file, CGROUP_FILE_PROCS);
2774}
2430 2775
2431static u64 cgroup_read_notify_on_release(struct cgroup *cgrp, 2776static u64 cgroup_read_notify_on_release(struct cgroup *cgrp,
2432 struct cftype *cft) 2777 struct cftype *cft)
@@ -2449,21 +2794,27 @@ static int cgroup_write_notify_on_release(struct cgroup *cgrp,
2449/* 2794/*
2450 * for the common functions, 'private' gives the type of file 2795 * for the common functions, 'private' gives the type of file
2451 */ 2796 */
2797/* for hysterical raisins, we can't put this on the older files */
2798#define CGROUP_FILE_GENERIC_PREFIX "cgroup."
2452static struct cftype files[] = { 2799static struct cftype files[] = {
2453 { 2800 {
2454 .name = "tasks", 2801 .name = "tasks",
2455 .open = cgroup_tasks_open, 2802 .open = cgroup_tasks_open,
2456 .write_u64 = cgroup_tasks_write, 2803 .write_u64 = cgroup_tasks_write,
2457 .release = cgroup_tasks_release, 2804 .release = cgroup_pidlist_release,
2458 .private = FILE_TASKLIST,
2459 .mode = S_IRUGO | S_IWUSR, 2805 .mode = S_IRUGO | S_IWUSR,
2460 }, 2806 },
2461 2807 {
2808 .name = CGROUP_FILE_GENERIC_PREFIX "procs",
2809 .open = cgroup_procs_open,
2810 /* .write_u64 = cgroup_procs_write, TODO */
2811 .release = cgroup_pidlist_release,
2812 .mode = S_IRUGO,
2813 },
2462 { 2814 {
2463 .name = "notify_on_release", 2815 .name = "notify_on_release",
2464 .read_u64 = cgroup_read_notify_on_release, 2816 .read_u64 = cgroup_read_notify_on_release,
2465 .write_u64 = cgroup_write_notify_on_release, 2817 .write_u64 = cgroup_write_notify_on_release,
2466 .private = FILE_NOTIFY_ON_RELEASE,
2467 }, 2818 },
2468}; 2819};
2469 2820
@@ -2472,7 +2823,6 @@ static struct cftype cft_release_agent = {
2472 .read_seq_string = cgroup_release_agent_show, 2823 .read_seq_string = cgroup_release_agent_show,
2473 .write_string = cgroup_release_agent_write, 2824 .write_string = cgroup_release_agent_write,
2474 .max_write_len = PATH_MAX, 2825 .max_write_len = PATH_MAX,
2475 .private = FILE_RELEASE_AGENT,
2476}; 2826};
2477 2827
2478static int cgroup_populate_dir(struct cgroup *cgrp) 2828static int cgroup_populate_dir(struct cgroup *cgrp)
@@ -2879,6 +3229,7 @@ int __init cgroup_init_early(void)
2879 init_task.cgroups = &init_css_set; 3229 init_task.cgroups = &init_css_set;
2880 3230
2881 init_css_set_link.cg = &init_css_set; 3231 init_css_set_link.cg = &init_css_set;
3232 init_css_set_link.cgrp = dummytop;
2882 list_add(&init_css_set_link.cgrp_link_list, 3233 list_add(&init_css_set_link.cgrp_link_list,
2883 &rootnode.top_cgroup.css_sets); 3234 &rootnode.top_cgroup.css_sets);
2884 list_add(&init_css_set_link.cg_link_list, 3235 list_add(&init_css_set_link.cg_link_list,
@@ -2933,7 +3284,7 @@ int __init cgroup_init(void)
2933 /* Add init_css_set to the hash table */ 3284 /* Add init_css_set to the hash table */
2934 hhead = css_set_hash(init_css_set.subsys); 3285 hhead = css_set_hash(init_css_set.subsys);
2935 hlist_add_head(&init_css_set.hlist, hhead); 3286 hlist_add_head(&init_css_set.hlist, hhead);
2936 3287 BUG_ON(!init_root_id(&rootnode));
2937 err = register_filesystem(&cgroup_fs_type); 3288 err = register_filesystem(&cgroup_fs_type);
2938 if (err < 0) 3289 if (err < 0)
2939 goto out; 3290 goto out;
@@ -2986,15 +3337,16 @@ static int proc_cgroup_show(struct seq_file *m, void *v)
2986 for_each_active_root(root) { 3337 for_each_active_root(root) {
2987 struct cgroup_subsys *ss; 3338 struct cgroup_subsys *ss;
2988 struct cgroup *cgrp; 3339 struct cgroup *cgrp;
2989 int subsys_id;
2990 int count = 0; 3340 int count = 0;
2991 3341
2992 seq_printf(m, "%lu:", root->subsys_bits); 3342 seq_printf(m, "%d:", root->hierarchy_id);
2993 for_each_subsys(root, ss) 3343 for_each_subsys(root, ss)
2994 seq_printf(m, "%s%s", count++ ? "," : "", ss->name); 3344 seq_printf(m, "%s%s", count++ ? "," : "", ss->name);
3345 if (strlen(root->name))
3346 seq_printf(m, "%sname=%s", count ? "," : "",
3347 root->name);
2995 seq_putc(m, ':'); 3348 seq_putc(m, ':');
2996 get_first_subsys(&root->top_cgroup, NULL, &subsys_id); 3349 cgrp = task_cgroup_from_root(tsk, root);
2997 cgrp = task_cgroup(tsk, subsys_id);
2998 retval = cgroup_path(cgrp, buf, PAGE_SIZE); 3350 retval = cgroup_path(cgrp, buf, PAGE_SIZE);
2999 if (retval < 0) 3351 if (retval < 0)
3000 goto out_unlock; 3352 goto out_unlock;
@@ -3033,8 +3385,8 @@ static int proc_cgroupstats_show(struct seq_file *m, void *v)
3033 mutex_lock(&cgroup_mutex); 3385 mutex_lock(&cgroup_mutex);
3034 for (i = 0; i < CGROUP_SUBSYS_COUNT; i++) { 3386 for (i = 0; i < CGROUP_SUBSYS_COUNT; i++) {
3035 struct cgroup_subsys *ss = subsys[i]; 3387 struct cgroup_subsys *ss = subsys[i];
3036 seq_printf(m, "%s\t%lu\t%d\t%d\n", 3388 seq_printf(m, "%s\t%d\t%d\t%d\n",
3037 ss->name, ss->root->subsys_bits, 3389 ss->name, ss->root->hierarchy_id,
3038 ss->root->number_of_cgroups, !ss->disabled); 3390 ss->root->number_of_cgroups, !ss->disabled);
3039 } 3391 }
3040 mutex_unlock(&cgroup_mutex); 3392 mutex_unlock(&cgroup_mutex);
@@ -3320,13 +3672,11 @@ int cgroup_is_descendant(const struct cgroup *cgrp, struct task_struct *task)
3320{ 3672{
3321 int ret; 3673 int ret;
3322 struct cgroup *target; 3674 struct cgroup *target;
3323 int subsys_id;
3324 3675
3325 if (cgrp == dummytop) 3676 if (cgrp == dummytop)
3326 return 1; 3677 return 1;
3327 3678
3328 get_first_subsys(cgrp, NULL, &subsys_id); 3679 target = task_cgroup_from_root(task, cgrp->root);
3329 target = task_cgroup(task, subsys_id);
3330 while (cgrp != target && cgrp!= cgrp->top_cgroup) 3680 while (cgrp != target && cgrp!= cgrp->top_cgroup)
3331 cgrp = cgrp->parent; 3681 cgrp = cgrp->parent;
3332 ret = (cgrp == target); 3682 ret = (cgrp == target);
@@ -3693,3 +4043,154 @@ css_get_next(struct cgroup_subsys *ss, int id,
3693 return ret; 4043 return ret;
3694} 4044}
3695 4045
4046#ifdef CONFIG_CGROUP_DEBUG
4047static struct cgroup_subsys_state *debug_create(struct cgroup_subsys *ss,
4048 struct cgroup *cont)
4049{
4050 struct cgroup_subsys_state *css = kzalloc(sizeof(*css), GFP_KERNEL);
4051
4052 if (!css)
4053 return ERR_PTR(-ENOMEM);
4054
4055 return css;
4056}
4057
4058static void debug_destroy(struct cgroup_subsys *ss, struct cgroup *cont)
4059{
4060 kfree(cont->subsys[debug_subsys_id]);
4061}
4062
4063static u64 cgroup_refcount_read(struct cgroup *cont, struct cftype *cft)
4064{
4065 return atomic_read(&cont->count);
4066}
4067
4068static u64 debug_taskcount_read(struct cgroup *cont, struct cftype *cft)
4069{
4070 return cgroup_task_count(cont);
4071}
4072
4073static u64 current_css_set_read(struct cgroup *cont, struct cftype *cft)
4074{
4075 return (u64)(unsigned long)current->cgroups;
4076}
4077
4078static u64 current_css_set_refcount_read(struct cgroup *cont,
4079 struct cftype *cft)
4080{
4081 u64 count;
4082
4083 rcu_read_lock();
4084 count = atomic_read(&current->cgroups->refcount);
4085 rcu_read_unlock();
4086 return count;
4087}
4088
4089static int current_css_set_cg_links_read(struct cgroup *cont,
4090 struct cftype *cft,
4091 struct seq_file *seq)
4092{
4093 struct cg_cgroup_link *link;
4094 struct css_set *cg;
4095
4096 read_lock(&css_set_lock);
4097 rcu_read_lock();
4098 cg = rcu_dereference(current->cgroups);
4099 list_for_each_entry(link, &cg->cg_links, cg_link_list) {
4100 struct cgroup *c = link->cgrp;
4101 const char *name;
4102
4103 if (c->dentry)
4104 name = c->dentry->d_name.name;
4105 else
4106 name = "?";
4107 seq_printf(seq, "Root %d group %s\n",
4108 c->root->hierarchy_id, name);
4109 }
4110 rcu_read_unlock();
4111 read_unlock(&css_set_lock);
4112 return 0;
4113}
4114
4115#define MAX_TASKS_SHOWN_PER_CSS 25
4116static int cgroup_css_links_read(struct cgroup *cont,
4117 struct cftype *cft,
4118 struct seq_file *seq)
4119{
4120 struct cg_cgroup_link *link;
4121
4122 read_lock(&css_set_lock);
4123 list_for_each_entry(link, &cont->css_sets, cgrp_link_list) {
4124 struct css_set *cg = link->cg;
4125 struct task_struct *task;
4126 int count = 0;
4127 seq_printf(seq, "css_set %p\n", cg);
4128 list_for_each_entry(task, &cg->tasks, cg_list) {
4129 if (count++ > MAX_TASKS_SHOWN_PER_CSS) {
4130 seq_puts(seq, " ...\n");
4131 break;
4132 } else {
4133 seq_printf(seq, " task %d\n",
4134 task_pid_vnr(task));
4135 }
4136 }
4137 }
4138 read_unlock(&css_set_lock);
4139 return 0;
4140}
4141
4142static u64 releasable_read(struct cgroup *cgrp, struct cftype *cft)
4143{
4144 return test_bit(CGRP_RELEASABLE, &cgrp->flags);
4145}
4146
4147static struct cftype debug_files[] = {
4148 {
4149 .name = "cgroup_refcount",
4150 .read_u64 = cgroup_refcount_read,
4151 },
4152 {
4153 .name = "taskcount",
4154 .read_u64 = debug_taskcount_read,
4155 },
4156
4157 {
4158 .name = "current_css_set",
4159 .read_u64 = current_css_set_read,
4160 },
4161
4162 {
4163 .name = "current_css_set_refcount",
4164 .read_u64 = current_css_set_refcount_read,
4165 },
4166
4167 {
4168 .name = "current_css_set_cg_links",
4169 .read_seq_string = current_css_set_cg_links_read,
4170 },
4171
4172 {
4173 .name = "cgroup_css_links",
4174 .read_seq_string = cgroup_css_links_read,
4175 },
4176
4177 {
4178 .name = "releasable",
4179 .read_u64 = releasable_read,
4180 },
4181};
4182
4183static int debug_populate(struct cgroup_subsys *ss, struct cgroup *cont)
4184{
4185 return cgroup_add_files(cont, ss, debug_files,
4186 ARRAY_SIZE(debug_files));
4187}
4188
4189struct cgroup_subsys debug_subsys = {
4190 .name = "debug",
4191 .create = debug_create,
4192 .destroy = debug_destroy,
4193 .populate = debug_populate,
4194 .subsys_id = debug_subsys_id,
4195};
4196#endif /* CONFIG_CGROUP_DEBUG */
diff --git a/kernel/cgroup_debug.c b/kernel/cgroup_debug.c
deleted file mode 100644
index 0c92d797baa6..000000000000
--- a/kernel/cgroup_debug.c
+++ /dev/null
@@ -1,105 +0,0 @@
1/*
2 * kernel/cgroup_debug.c - Example cgroup subsystem that
3 * exposes debug info
4 *
5 * Copyright (C) Google Inc, 2007
6 *
7 * Developed by Paul Menage (menage@google.com)
8 *
9 */
10
11#include <linux/cgroup.h>
12#include <linux/fs.h>
13#include <linux/slab.h>
14#include <linux/rcupdate.h>
15
16#include <asm/atomic.h>
17
18static struct cgroup_subsys_state *debug_create(struct cgroup_subsys *ss,
19 struct cgroup *cont)
20{
21 struct cgroup_subsys_state *css = kzalloc(sizeof(*css), GFP_KERNEL);
22
23 if (!css)
24 return ERR_PTR(-ENOMEM);
25
26 return css;
27}
28
29static void debug_destroy(struct cgroup_subsys *ss, struct cgroup *cont)
30{
31 kfree(cont->subsys[debug_subsys_id]);
32}
33
34static u64 cgroup_refcount_read(struct cgroup *cont, struct cftype *cft)
35{
36 return atomic_read(&cont->count);
37}
38
39static u64 taskcount_read(struct cgroup *cont, struct cftype *cft)
40{
41 u64 count;
42
43 count = cgroup_task_count(cont);
44 return count;
45}
46
47static u64 current_css_set_read(struct cgroup *cont, struct cftype *cft)
48{
49 return (u64)(long)current->cgroups;
50}
51
52static u64 current_css_set_refcount_read(struct cgroup *cont,
53 struct cftype *cft)
54{
55 u64 count;
56
57 rcu_read_lock();
58 count = atomic_read(&current->cgroups->refcount);
59 rcu_read_unlock();
60 return count;
61}
62
63static u64 releasable_read(struct cgroup *cgrp, struct cftype *cft)
64{
65 return test_bit(CGRP_RELEASABLE, &cgrp->flags);
66}
67
68static struct cftype files[] = {
69 {
70 .name = "cgroup_refcount",
71 .read_u64 = cgroup_refcount_read,
72 },
73 {
74 .name = "taskcount",
75 .read_u64 = taskcount_read,
76 },
77
78 {
79 .name = "current_css_set",
80 .read_u64 = current_css_set_read,
81 },
82
83 {
84 .name = "current_css_set_refcount",
85 .read_u64 = current_css_set_refcount_read,
86 },
87
88 {
89 .name = "releasable",
90 .read_u64 = releasable_read,
91 },
92};
93
94static int debug_populate(struct cgroup_subsys *ss, struct cgroup *cont)
95{
96 return cgroup_add_files(cont, ss, files, ARRAY_SIZE(files));
97}
98
99struct cgroup_subsys debug_subsys = {
100 .name = "debug",
101 .create = debug_create,
102 .destroy = debug_destroy,
103 .populate = debug_populate,
104 .subsys_id = debug_subsys_id,
105};
diff --git a/kernel/cgroup_freezer.c b/kernel/cgroup_freezer.c
index fb249e2bcada..59e9ef6aab40 100644
--- a/kernel/cgroup_freezer.c
+++ b/kernel/cgroup_freezer.c
@@ -159,7 +159,7 @@ static bool is_task_frozen_enough(struct task_struct *task)
159 */ 159 */
160static int freezer_can_attach(struct cgroup_subsys *ss, 160static int freezer_can_attach(struct cgroup_subsys *ss,
161 struct cgroup *new_cgroup, 161 struct cgroup *new_cgroup,
162 struct task_struct *task) 162 struct task_struct *task, bool threadgroup)
163{ 163{
164 struct freezer *freezer; 164 struct freezer *freezer;
165 165
@@ -177,6 +177,19 @@ static int freezer_can_attach(struct cgroup_subsys *ss,
177 if (freezer->state == CGROUP_FROZEN) 177 if (freezer->state == CGROUP_FROZEN)
178 return -EBUSY; 178 return -EBUSY;
179 179
180 if (threadgroup) {
181 struct task_struct *c;
182
183 rcu_read_lock();
184 list_for_each_entry_rcu(c, &task->thread_group, thread_group) {
185 if (is_task_frozen_enough(c)) {
186 rcu_read_unlock();
187 return -EBUSY;
188 }
189 }
190 rcu_read_unlock();
191 }
192
180 return 0; 193 return 0;
181} 194}
182 195
diff --git a/kernel/cpuset.c b/kernel/cpuset.c
index 7e75a41bd508..b5cb469d2545 100644
--- a/kernel/cpuset.c
+++ b/kernel/cpuset.c
@@ -1324,9 +1324,10 @@ static int fmeter_getrate(struct fmeter *fmp)
1324static cpumask_var_t cpus_attach; 1324static cpumask_var_t cpus_attach;
1325 1325
1326/* Called by cgroups to determine if a cpuset is usable; cgroup_mutex held */ 1326/* Called by cgroups to determine if a cpuset is usable; cgroup_mutex held */
1327static int cpuset_can_attach(struct cgroup_subsys *ss, 1327static int cpuset_can_attach(struct cgroup_subsys *ss, struct cgroup *cont,
1328 struct cgroup *cont, struct task_struct *tsk) 1328 struct task_struct *tsk, bool threadgroup)
1329{ 1329{
1330 int ret;
1330 struct cpuset *cs = cgroup_cs(cont); 1331 struct cpuset *cs = cgroup_cs(cont);
1331 1332
1332 if (cpumask_empty(cs->cpus_allowed) || nodes_empty(cs->mems_allowed)) 1333 if (cpumask_empty(cs->cpus_allowed) || nodes_empty(cs->mems_allowed))
@@ -1343,18 +1344,51 @@ static int cpuset_can_attach(struct cgroup_subsys *ss,
1343 if (tsk->flags & PF_THREAD_BOUND) 1344 if (tsk->flags & PF_THREAD_BOUND)
1344 return -EINVAL; 1345 return -EINVAL;
1345 1346
1346 return security_task_setscheduler(tsk, 0, NULL); 1347 ret = security_task_setscheduler(tsk, 0, NULL);
1348 if (ret)
1349 return ret;
1350 if (threadgroup) {
1351 struct task_struct *c;
1352
1353 rcu_read_lock();
1354 list_for_each_entry_rcu(c, &tsk->thread_group, thread_group) {
1355 ret = security_task_setscheduler(c, 0, NULL);
1356 if (ret) {
1357 rcu_read_unlock();
1358 return ret;
1359 }
1360 }
1361 rcu_read_unlock();
1362 }
1363 return 0;
1364}
1365
1366static void cpuset_attach_task(struct task_struct *tsk, nodemask_t *to,
1367 struct cpuset *cs)
1368{
1369 int err;
1370 /*
1371 * can_attach beforehand should guarantee that this doesn't fail.
1372 * TODO: have a better way to handle failure here
1373 */
1374 err = set_cpus_allowed_ptr(tsk, cpus_attach);
1375 WARN_ON_ONCE(err);
1376
1377 task_lock(tsk);
1378 cpuset_change_task_nodemask(tsk, to);
1379 task_unlock(tsk);
1380 cpuset_update_task_spread_flag(cs, tsk);
1381
1347} 1382}
1348 1383
1349static void cpuset_attach(struct cgroup_subsys *ss, 1384static void cpuset_attach(struct cgroup_subsys *ss, struct cgroup *cont,
1350 struct cgroup *cont, struct cgroup *oldcont, 1385 struct cgroup *oldcont, struct task_struct *tsk,
1351 struct task_struct *tsk) 1386 bool threadgroup)
1352{ 1387{
1353 nodemask_t from, to; 1388 nodemask_t from, to;
1354 struct mm_struct *mm; 1389 struct mm_struct *mm;
1355 struct cpuset *cs = cgroup_cs(cont); 1390 struct cpuset *cs = cgroup_cs(cont);
1356 struct cpuset *oldcs = cgroup_cs(oldcont); 1391 struct cpuset *oldcs = cgroup_cs(oldcont);
1357 int err;
1358 1392
1359 if (cs == &top_cpuset) { 1393 if (cs == &top_cpuset) {
1360 cpumask_copy(cpus_attach, cpu_possible_mask); 1394 cpumask_copy(cpus_attach, cpu_possible_mask);
@@ -1363,15 +1397,19 @@ static void cpuset_attach(struct cgroup_subsys *ss,
1363 guarantee_online_cpus(cs, cpus_attach); 1397 guarantee_online_cpus(cs, cpus_attach);
1364 guarantee_online_mems(cs, &to); 1398 guarantee_online_mems(cs, &to);
1365 } 1399 }
1366 err = set_cpus_allowed_ptr(tsk, cpus_attach);
1367 if (err)
1368 return;
1369 1400
1370 task_lock(tsk); 1401 /* do per-task migration stuff possibly for each in the threadgroup */
1371 cpuset_change_task_nodemask(tsk, &to); 1402 cpuset_attach_task(tsk, &to, cs);
1372 task_unlock(tsk); 1403 if (threadgroup) {
1373 cpuset_update_task_spread_flag(cs, tsk); 1404 struct task_struct *c;
1405 rcu_read_lock();
1406 list_for_each_entry_rcu(c, &tsk->thread_group, thread_group) {
1407 cpuset_attach_task(c, &to, cs);
1408 }
1409 rcu_read_unlock();
1410 }
1374 1411
1412 /* change mm; only needs to be done once even if threadgroup */
1375 from = oldcs->mems_allowed; 1413 from = oldcs->mems_allowed;
1376 to = cs->mems_allowed; 1414 to = cs->mems_allowed;
1377 mm = get_task_mm(tsk); 1415 mm = get_task_mm(tsk);
diff --git a/kernel/exit.c b/kernel/exit.c
index 60d6fdcc9265..5859f598c951 100644
--- a/kernel/exit.c
+++ b/kernel/exit.c
@@ -976,8 +976,6 @@ NORET_TYPE void do_exit(long code)
976 disassociate_ctty(1); 976 disassociate_ctty(1);
977 977
978 module_put(task_thread_info(tsk)->exec_domain->module); 978 module_put(task_thread_info(tsk)->exec_domain->module);
979 if (tsk->binfmt)
980 module_put(tsk->binfmt->module);
981 979
982 proc_exit_connector(tsk); 980 proc_exit_connector(tsk);
983 981
@@ -1097,28 +1095,28 @@ struct wait_opts {
1097 int __user *wo_stat; 1095 int __user *wo_stat;
1098 struct rusage __user *wo_rusage; 1096 struct rusage __user *wo_rusage;
1099 1097
1098 wait_queue_t child_wait;
1100 int notask_error; 1099 int notask_error;
1101}; 1100};
1102 1101
1103static struct pid *task_pid_type(struct task_struct *task, enum pid_type type) 1102static inline
1103struct pid *task_pid_type(struct task_struct *task, enum pid_type type)
1104{ 1104{
1105 struct pid *pid = NULL; 1105 if (type != PIDTYPE_PID)
1106 if (type == PIDTYPE_PID) 1106 task = task->group_leader;
1107 pid = task->pids[type].pid; 1107 return task->pids[type].pid;
1108 else if (type < PIDTYPE_MAX)
1109 pid = task->group_leader->pids[type].pid;
1110 return pid;
1111} 1108}
1112 1109
1113static int eligible_child(struct wait_opts *wo, struct task_struct *p) 1110static int eligible_pid(struct wait_opts *wo, struct task_struct *p)
1114{ 1111{
1115 int err; 1112 return wo->wo_type == PIDTYPE_MAX ||
1116 1113 task_pid_type(p, wo->wo_type) == wo->wo_pid;
1117 if (wo->wo_type < PIDTYPE_MAX) { 1114}
1118 if (task_pid_type(p, wo->wo_type) != wo->wo_pid)
1119 return 0;
1120 }
1121 1115
1116static int eligible_child(struct wait_opts *wo, struct task_struct *p)
1117{
1118 if (!eligible_pid(wo, p))
1119 return 0;
1122 /* Wait for all children (clone and not) if __WALL is set; 1120 /* Wait for all children (clone and not) if __WALL is set;
1123 * otherwise, wait for clone children *only* if __WCLONE is 1121 * otherwise, wait for clone children *only* if __WCLONE is
1124 * set; otherwise, wait for non-clone children *only*. (Note: 1122 * set; otherwise, wait for non-clone children *only*. (Note:
@@ -1128,10 +1126,6 @@ static int eligible_child(struct wait_opts *wo, struct task_struct *p)
1128 && !(wo->wo_flags & __WALL)) 1126 && !(wo->wo_flags & __WALL))
1129 return 0; 1127 return 0;
1130 1128
1131 err = security_task_wait(p);
1132 if (err)
1133 return err;
1134
1135 return 1; 1129 return 1;
1136} 1130}
1137 1131
@@ -1144,18 +1138,20 @@ static int wait_noreap_copyout(struct wait_opts *wo, struct task_struct *p,
1144 1138
1145 put_task_struct(p); 1139 put_task_struct(p);
1146 infop = wo->wo_info; 1140 infop = wo->wo_info;
1147 if (!retval) 1141 if (infop) {
1148 retval = put_user(SIGCHLD, &infop->si_signo); 1142 if (!retval)
1149 if (!retval) 1143 retval = put_user(SIGCHLD, &infop->si_signo);
1150 retval = put_user(0, &infop->si_errno); 1144 if (!retval)
1151 if (!retval) 1145 retval = put_user(0, &infop->si_errno);
1152 retval = put_user((short)why, &infop->si_code); 1146 if (!retval)
1153 if (!retval) 1147 retval = put_user((short)why, &infop->si_code);
1154 retval = put_user(pid, &infop->si_pid); 1148 if (!retval)
1155 if (!retval) 1149 retval = put_user(pid, &infop->si_pid);
1156 retval = put_user(uid, &infop->si_uid); 1150 if (!retval)
1157 if (!retval) 1151 retval = put_user(uid, &infop->si_uid);
1158 retval = put_user(status, &infop->si_status); 1152 if (!retval)
1153 retval = put_user(status, &infop->si_status);
1154 }
1159 if (!retval) 1155 if (!retval)
1160 retval = pid; 1156 retval = pid;
1161 return retval; 1157 return retval;
@@ -1485,13 +1481,14 @@ static int wait_task_continued(struct wait_opts *wo, struct task_struct *p)
1485 * then ->notask_error is 0 if @p is an eligible child, 1481 * then ->notask_error is 0 if @p is an eligible child,
1486 * or another error from security_task_wait(), or still -ECHILD. 1482 * or another error from security_task_wait(), or still -ECHILD.
1487 */ 1483 */
1488static int wait_consider_task(struct wait_opts *wo, struct task_struct *parent, 1484static int wait_consider_task(struct wait_opts *wo, int ptrace,
1489 int ptrace, struct task_struct *p) 1485 struct task_struct *p)
1490{ 1486{
1491 int ret = eligible_child(wo, p); 1487 int ret = eligible_child(wo, p);
1492 if (!ret) 1488 if (!ret)
1493 return ret; 1489 return ret;
1494 1490
1491 ret = security_task_wait(p);
1495 if (unlikely(ret < 0)) { 1492 if (unlikely(ret < 0)) {
1496 /* 1493 /*
1497 * If we have not yet seen any eligible child, 1494 * If we have not yet seen any eligible child,
@@ -1553,7 +1550,7 @@ static int do_wait_thread(struct wait_opts *wo, struct task_struct *tsk)
1553 * Do not consider detached threads. 1550 * Do not consider detached threads.
1554 */ 1551 */
1555 if (!task_detached(p)) { 1552 if (!task_detached(p)) {
1556 int ret = wait_consider_task(wo, tsk, 0, p); 1553 int ret = wait_consider_task(wo, 0, p);
1557 if (ret) 1554 if (ret)
1558 return ret; 1555 return ret;
1559 } 1556 }
@@ -1567,7 +1564,7 @@ static int ptrace_do_wait(struct wait_opts *wo, struct task_struct *tsk)
1567 struct task_struct *p; 1564 struct task_struct *p;
1568 1565
1569 list_for_each_entry(p, &tsk->ptraced, ptrace_entry) { 1566 list_for_each_entry(p, &tsk->ptraced, ptrace_entry) {
1570 int ret = wait_consider_task(wo, tsk, 1, p); 1567 int ret = wait_consider_task(wo, 1, p);
1571 if (ret) 1568 if (ret)
1572 return ret; 1569 return ret;
1573 } 1570 }
@@ -1575,15 +1572,38 @@ static int ptrace_do_wait(struct wait_opts *wo, struct task_struct *tsk)
1575 return 0; 1572 return 0;
1576} 1573}
1577 1574
1575static int child_wait_callback(wait_queue_t *wait, unsigned mode,
1576 int sync, void *key)
1577{
1578 struct wait_opts *wo = container_of(wait, struct wait_opts,
1579 child_wait);
1580 struct task_struct *p = key;
1581
1582 if (!eligible_pid(wo, p))
1583 return 0;
1584
1585 if ((wo->wo_flags & __WNOTHREAD) && wait->private != p->parent)
1586 return 0;
1587
1588 return default_wake_function(wait, mode, sync, key);
1589}
1590
1591void __wake_up_parent(struct task_struct *p, struct task_struct *parent)
1592{
1593 __wake_up_sync_key(&parent->signal->wait_chldexit,
1594 TASK_INTERRUPTIBLE, 1, p);
1595}
1596
1578static long do_wait(struct wait_opts *wo) 1597static long do_wait(struct wait_opts *wo)
1579{ 1598{
1580 DECLARE_WAITQUEUE(wait, current);
1581 struct task_struct *tsk; 1599 struct task_struct *tsk;
1582 int retval; 1600 int retval;
1583 1601
1584 trace_sched_process_wait(wo->wo_pid); 1602 trace_sched_process_wait(wo->wo_pid);
1585 1603
1586 add_wait_queue(&current->signal->wait_chldexit,&wait); 1604 init_waitqueue_func_entry(&wo->child_wait, child_wait_callback);
1605 wo->child_wait.private = current;
1606 add_wait_queue(&current->signal->wait_chldexit, &wo->child_wait);
1587repeat: 1607repeat:
1588 /* 1608 /*
1589 * If there is nothing that can match our critiera just get out. 1609 * If there is nothing that can match our critiera just get out.
@@ -1624,32 +1644,7 @@ notask:
1624 } 1644 }
1625end: 1645end:
1626 __set_current_state(TASK_RUNNING); 1646 __set_current_state(TASK_RUNNING);
1627 remove_wait_queue(&current->signal->wait_chldexit,&wait); 1647 remove_wait_queue(&current->signal->wait_chldexit, &wo->child_wait);
1628 if (wo->wo_info) {
1629 struct siginfo __user *infop = wo->wo_info;
1630
1631 if (retval > 0)
1632 retval = 0;
1633 else {
1634 /*
1635 * For a WNOHANG return, clear out all the fields
1636 * we would set so the user can easily tell the
1637 * difference.
1638 */
1639 if (!retval)
1640 retval = put_user(0, &infop->si_signo);
1641 if (!retval)
1642 retval = put_user(0, &infop->si_errno);
1643 if (!retval)
1644 retval = put_user(0, &infop->si_code);
1645 if (!retval)
1646 retval = put_user(0, &infop->si_pid);
1647 if (!retval)
1648 retval = put_user(0, &infop->si_uid);
1649 if (!retval)
1650 retval = put_user(0, &infop->si_status);
1651 }
1652 }
1653 return retval; 1648 return retval;
1654} 1649}
1655 1650
@@ -1694,6 +1689,29 @@ SYSCALL_DEFINE5(waitid, int, which, pid_t, upid, struct siginfo __user *,
1694 wo.wo_stat = NULL; 1689 wo.wo_stat = NULL;
1695 wo.wo_rusage = ru; 1690 wo.wo_rusage = ru;
1696 ret = do_wait(&wo); 1691 ret = do_wait(&wo);
1692
1693 if (ret > 0) {
1694 ret = 0;
1695 } else if (infop) {
1696 /*
1697 * For a WNOHANG return, clear out all the fields
1698 * we would set so the user can easily tell the
1699 * difference.
1700 */
1701 if (!ret)
1702 ret = put_user(0, &infop->si_signo);
1703 if (!ret)
1704 ret = put_user(0, &infop->si_errno);
1705 if (!ret)
1706 ret = put_user(0, &infop->si_code);
1707 if (!ret)
1708 ret = put_user(0, &infop->si_pid);
1709 if (!ret)
1710 ret = put_user(0, &infop->si_uid);
1711 if (!ret)
1712 ret = put_user(0, &infop->si_status);
1713 }
1714
1697 put_pid(pid); 1715 put_pid(pid);
1698 1716
1699 /* avoid REGPARM breakage on x86: */ 1717 /* avoid REGPARM breakage on x86: */
diff --git a/kernel/fork.c b/kernel/fork.c
index 51ad0b0b7266..266c6af6ef1b 100644
--- a/kernel/fork.c
+++ b/kernel/fork.c
@@ -434,6 +434,14 @@ __setup("coredump_filter=", coredump_filter_setup);
434 434
435#include <linux/init_task.h> 435#include <linux/init_task.h>
436 436
437static void mm_init_aio(struct mm_struct *mm)
438{
439#ifdef CONFIG_AIO
440 spin_lock_init(&mm->ioctx_lock);
441 INIT_HLIST_HEAD(&mm->ioctx_list);
442#endif
443}
444
437static struct mm_struct * mm_init(struct mm_struct * mm, struct task_struct *p) 445static struct mm_struct * mm_init(struct mm_struct * mm, struct task_struct *p)
438{ 446{
439 atomic_set(&mm->mm_users, 1); 447 atomic_set(&mm->mm_users, 1);
@@ -447,10 +455,9 @@ static struct mm_struct * mm_init(struct mm_struct * mm, struct task_struct *p)
447 set_mm_counter(mm, file_rss, 0); 455 set_mm_counter(mm, file_rss, 0);
448 set_mm_counter(mm, anon_rss, 0); 456 set_mm_counter(mm, anon_rss, 0);
449 spin_lock_init(&mm->page_table_lock); 457 spin_lock_init(&mm->page_table_lock);
450 spin_lock_init(&mm->ioctx_lock);
451 INIT_HLIST_HEAD(&mm->ioctx_list);
452 mm->free_area_cache = TASK_UNMAPPED_BASE; 458 mm->free_area_cache = TASK_UNMAPPED_BASE;
453 mm->cached_hole_size = ~0UL; 459 mm->cached_hole_size = ~0UL;
460 mm_init_aio(mm);
454 mm_init_owner(mm, p); 461 mm_init_owner(mm, p);
455 462
456 if (likely(!mm_alloc_pgd(mm))) { 463 if (likely(!mm_alloc_pgd(mm))) {
@@ -511,6 +518,8 @@ void mmput(struct mm_struct *mm)
511 spin_unlock(&mmlist_lock); 518 spin_unlock(&mmlist_lock);
512 } 519 }
513 put_swap_token(mm); 520 put_swap_token(mm);
521 if (mm->binfmt)
522 module_put(mm->binfmt->module);
514 mmdrop(mm); 523 mmdrop(mm);
515 } 524 }
516} 525}
@@ -636,9 +645,14 @@ struct mm_struct *dup_mm(struct task_struct *tsk)
636 mm->hiwater_rss = get_mm_rss(mm); 645 mm->hiwater_rss = get_mm_rss(mm);
637 mm->hiwater_vm = mm->total_vm; 646 mm->hiwater_vm = mm->total_vm;
638 647
648 if (mm->binfmt && !try_module_get(mm->binfmt->module))
649 goto free_pt;
650
639 return mm; 651 return mm;
640 652
641free_pt: 653free_pt:
654 /* don't put binfmt in mmput, we haven't got module yet */
655 mm->binfmt = NULL;
642 mmput(mm); 656 mmput(mm);
643 657
644fail_nomem: 658fail_nomem:
@@ -979,6 +993,16 @@ static struct task_struct *copy_process(unsigned long clone_flags,
979 if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM)) 993 if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM))
980 return ERR_PTR(-EINVAL); 994 return ERR_PTR(-EINVAL);
981 995
996 /*
997 * Siblings of global init remain as zombies on exit since they are
998 * not reaped by their parent (swapper). To solve this and to avoid
999 * multi-rooted process trees, prevent global and container-inits
1000 * from creating siblings.
1001 */
1002 if ((clone_flags & CLONE_PARENT) &&
1003 current->signal->flags & SIGNAL_UNKILLABLE)
1004 return ERR_PTR(-EINVAL);
1005
982 retval = security_task_create(clone_flags); 1006 retval = security_task_create(clone_flags);
983 if (retval) 1007 if (retval)
984 goto fork_out; 1008 goto fork_out;
@@ -1020,9 +1044,6 @@ static struct task_struct *copy_process(unsigned long clone_flags,
1020 if (!try_module_get(task_thread_info(p)->exec_domain->module)) 1044 if (!try_module_get(task_thread_info(p)->exec_domain->module))
1021 goto bad_fork_cleanup_count; 1045 goto bad_fork_cleanup_count;
1022 1046
1023 if (p->binfmt && !try_module_get(p->binfmt->module))
1024 goto bad_fork_cleanup_put_domain;
1025
1026 p->did_exec = 0; 1047 p->did_exec = 0;
1027 delayacct_tsk_init(p); /* Must remain after dup_task_struct() */ 1048 delayacct_tsk_init(p); /* Must remain after dup_task_struct() */
1028 copy_flags(clone_flags, p); 1049 copy_flags(clone_flags, p);
@@ -1310,9 +1331,6 @@ bad_fork_cleanup_cgroup:
1310#endif 1331#endif
1311 cgroup_exit(p, cgroup_callbacks_done); 1332 cgroup_exit(p, cgroup_callbacks_done);
1312 delayacct_tsk_free(p); 1333 delayacct_tsk_free(p);
1313 if (p->binfmt)
1314 module_put(p->binfmt->module);
1315bad_fork_cleanup_put_domain:
1316 module_put(task_thread_info(p)->exec_domain->module); 1334 module_put(task_thread_info(p)->exec_domain->module);
1317bad_fork_cleanup_count: 1335bad_fork_cleanup_count:
1318 atomic_dec(&p->cred->user->processes); 1336 atomic_dec(&p->cred->user->processes);
diff --git a/kernel/gcov/Kconfig b/kernel/gcov/Kconfig
index 654efd09f6a9..70a298d6da71 100644
--- a/kernel/gcov/Kconfig
+++ b/kernel/gcov/Kconfig
@@ -34,7 +34,7 @@ config GCOV_KERNEL
34config GCOV_PROFILE_ALL 34config GCOV_PROFILE_ALL
35 bool "Profile entire Kernel" 35 bool "Profile entire Kernel"
36 depends on GCOV_KERNEL 36 depends on GCOV_KERNEL
37 depends on S390 || X86 || (PPC && EXPERIMENTAL) 37 depends on S390 || X86 || (PPC && EXPERIMENTAL) || MICROBLAZE
38 default n 38 default n
39 ---help--- 39 ---help---
40 This options activates profiling for the entire kernel. 40 This options activates profiling for the entire kernel.
diff --git a/kernel/hung_task.c b/kernel/hung_task.c
index 022a4927b785..d4e841747400 100644
--- a/kernel/hung_task.c
+++ b/kernel/hung_task.c
@@ -171,12 +171,12 @@ static unsigned long timeout_jiffies(unsigned long timeout)
171 * Process updating of timeout sysctl 171 * Process updating of timeout sysctl
172 */ 172 */
173int proc_dohung_task_timeout_secs(struct ctl_table *table, int write, 173int proc_dohung_task_timeout_secs(struct ctl_table *table, int write,
174 struct file *filp, void __user *buffer, 174 void __user *buffer,
175 size_t *lenp, loff_t *ppos) 175 size_t *lenp, loff_t *ppos)
176{ 176{
177 int ret; 177 int ret;
178 178
179 ret = proc_doulongvec_minmax(table, write, filp, buffer, lenp, ppos); 179 ret = proc_doulongvec_minmax(table, write, buffer, lenp, ppos);
180 180
181 if (ret || !write) 181 if (ret || !write)
182 goto out; 182 goto out;
diff --git a/kernel/module.c b/kernel/module.c
index e6bc4b28aa62..5a29397ca4b6 100644
--- a/kernel/module.c
+++ b/kernel/module.c
@@ -1797,6 +1797,17 @@ static void setup_modinfo(struct module *mod, Elf_Shdr *sechdrs,
1797 } 1797 }
1798} 1798}
1799 1799
1800static void free_modinfo(struct module *mod)
1801{
1802 struct module_attribute *attr;
1803 int i;
1804
1805 for (i = 0; (attr = modinfo_attrs[i]); i++) {
1806 if (attr->free)
1807 attr->free(mod);
1808 }
1809}
1810
1800#ifdef CONFIG_KALLSYMS 1811#ifdef CONFIG_KALLSYMS
1801 1812
1802/* lookup symbol in given range of kernel_symbols */ 1813/* lookup symbol in given range of kernel_symbols */
@@ -1862,13 +1873,93 @@ static char elf_type(const Elf_Sym *sym,
1862 return '?'; 1873 return '?';
1863} 1874}
1864 1875
1876static bool is_core_symbol(const Elf_Sym *src, const Elf_Shdr *sechdrs,
1877 unsigned int shnum)
1878{
1879 const Elf_Shdr *sec;
1880
1881 if (src->st_shndx == SHN_UNDEF
1882 || src->st_shndx >= shnum
1883 || !src->st_name)
1884 return false;
1885
1886 sec = sechdrs + src->st_shndx;
1887 if (!(sec->sh_flags & SHF_ALLOC)
1888#ifndef CONFIG_KALLSYMS_ALL
1889 || !(sec->sh_flags & SHF_EXECINSTR)
1890#endif
1891 || (sec->sh_entsize & INIT_OFFSET_MASK))
1892 return false;
1893
1894 return true;
1895}
1896
1897static unsigned long layout_symtab(struct module *mod,
1898 Elf_Shdr *sechdrs,
1899 unsigned int symindex,
1900 unsigned int strindex,
1901 const Elf_Ehdr *hdr,
1902 const char *secstrings,
1903 unsigned long *pstroffs,
1904 unsigned long *strmap)
1905{
1906 unsigned long symoffs;
1907 Elf_Shdr *symsect = sechdrs + symindex;
1908 Elf_Shdr *strsect = sechdrs + strindex;
1909 const Elf_Sym *src;
1910 const char *strtab;
1911 unsigned int i, nsrc, ndst;
1912
1913 /* Put symbol section at end of init part of module. */
1914 symsect->sh_flags |= SHF_ALLOC;
1915 symsect->sh_entsize = get_offset(mod, &mod->init_size, symsect,
1916 symindex) | INIT_OFFSET_MASK;
1917 DEBUGP("\t%s\n", secstrings + symsect->sh_name);
1918
1919 src = (void *)hdr + symsect->sh_offset;
1920 nsrc = symsect->sh_size / sizeof(*src);
1921 strtab = (void *)hdr + strsect->sh_offset;
1922 for (ndst = i = 1; i < nsrc; ++i, ++src)
1923 if (is_core_symbol(src, sechdrs, hdr->e_shnum)) {
1924 unsigned int j = src->st_name;
1925
1926 while(!__test_and_set_bit(j, strmap) && strtab[j])
1927 ++j;
1928 ++ndst;
1929 }
1930
1931 /* Append room for core symbols at end of core part. */
1932 symoffs = ALIGN(mod->core_size, symsect->sh_addralign ?: 1);
1933 mod->core_size = symoffs + ndst * sizeof(Elf_Sym);
1934
1935 /* Put string table section at end of init part of module. */
1936 strsect->sh_flags |= SHF_ALLOC;
1937 strsect->sh_entsize = get_offset(mod, &mod->init_size, strsect,
1938 strindex) | INIT_OFFSET_MASK;
1939 DEBUGP("\t%s\n", secstrings + strsect->sh_name);
1940
1941 /* Append room for core symbols' strings at end of core part. */
1942 *pstroffs = mod->core_size;
1943 __set_bit(0, strmap);
1944 mod->core_size += bitmap_weight(strmap, strsect->sh_size);
1945
1946 return symoffs;
1947}
1948
1865static void add_kallsyms(struct module *mod, 1949static void add_kallsyms(struct module *mod,
1866 Elf_Shdr *sechdrs, 1950 Elf_Shdr *sechdrs,
1951 unsigned int shnum,
1867 unsigned int symindex, 1952 unsigned int symindex,
1868 unsigned int strindex, 1953 unsigned int strindex,
1869 const char *secstrings) 1954 unsigned long symoffs,
1955 unsigned long stroffs,
1956 const char *secstrings,
1957 unsigned long *strmap)
1870{ 1958{
1871 unsigned int i; 1959 unsigned int i, ndst;
1960 const Elf_Sym *src;
1961 Elf_Sym *dst;
1962 char *s;
1872 1963
1873 mod->symtab = (void *)sechdrs[symindex].sh_addr; 1964 mod->symtab = (void *)sechdrs[symindex].sh_addr;
1874 mod->num_symtab = sechdrs[symindex].sh_size / sizeof(Elf_Sym); 1965 mod->num_symtab = sechdrs[symindex].sh_size / sizeof(Elf_Sym);
@@ -1878,13 +1969,44 @@ static void add_kallsyms(struct module *mod,
1878 for (i = 0; i < mod->num_symtab; i++) 1969 for (i = 0; i < mod->num_symtab; i++)
1879 mod->symtab[i].st_info 1970 mod->symtab[i].st_info
1880 = elf_type(&mod->symtab[i], sechdrs, secstrings, mod); 1971 = elf_type(&mod->symtab[i], sechdrs, secstrings, mod);
1972
1973 mod->core_symtab = dst = mod->module_core + symoffs;
1974 src = mod->symtab;
1975 *dst = *src;
1976 for (ndst = i = 1; i < mod->num_symtab; ++i, ++src) {
1977 if (!is_core_symbol(src, sechdrs, shnum))
1978 continue;
1979 dst[ndst] = *src;
1980 dst[ndst].st_name = bitmap_weight(strmap, dst[ndst].st_name);
1981 ++ndst;
1982 }
1983 mod->core_num_syms = ndst;
1984
1985 mod->core_strtab = s = mod->module_core + stroffs;
1986 for (*s = 0, i = 1; i < sechdrs[strindex].sh_size; ++i)
1987 if (test_bit(i, strmap))
1988 *++s = mod->strtab[i];
1881} 1989}
1882#else 1990#else
1991static inline unsigned long layout_symtab(struct module *mod,
1992 Elf_Shdr *sechdrs,
1993 unsigned int symindex,
1994 unsigned int strindex,
1995 const Elf_Hdr *hdr,
1996 const char *secstrings,
1997 unsigned long *pstroffs,
1998 unsigned long *strmap)
1999{
2000}
1883static inline void add_kallsyms(struct module *mod, 2001static inline void add_kallsyms(struct module *mod,
1884 Elf_Shdr *sechdrs, 2002 Elf_Shdr *sechdrs,
2003 unsigned int shnum,
1885 unsigned int symindex, 2004 unsigned int symindex,
1886 unsigned int strindex, 2005 unsigned int strindex,
1887 const char *secstrings) 2006 unsigned long symoffs,
2007 unsigned long stroffs,
2008 const char *secstrings,
2009 const unsigned long *strmap)
1888{ 2010{
1889} 2011}
1890#endif /* CONFIG_KALLSYMS */ 2012#endif /* CONFIG_KALLSYMS */
@@ -1959,6 +2081,9 @@ static noinline struct module *load_module(void __user *umod,
1959 struct module *mod; 2081 struct module *mod;
1960 long err = 0; 2082 long err = 0;
1961 void *percpu = NULL, *ptr = NULL; /* Stops spurious gcc warning */ 2083 void *percpu = NULL, *ptr = NULL; /* Stops spurious gcc warning */
2084#ifdef CONFIG_KALLSYMS
2085 unsigned long symoffs, stroffs, *strmap;
2086#endif
1962 mm_segment_t old_fs; 2087 mm_segment_t old_fs;
1963 2088
1964 DEBUGP("load_module: umod=%p, len=%lu, uargs=%p\n", 2089 DEBUGP("load_module: umod=%p, len=%lu, uargs=%p\n",
@@ -2040,11 +2165,6 @@ static noinline struct module *load_module(void __user *umod,
2040 /* Don't keep modinfo and version sections. */ 2165 /* Don't keep modinfo and version sections. */
2041 sechdrs[infoindex].sh_flags &= ~(unsigned long)SHF_ALLOC; 2166 sechdrs[infoindex].sh_flags &= ~(unsigned long)SHF_ALLOC;
2042 sechdrs[versindex].sh_flags &= ~(unsigned long)SHF_ALLOC; 2167 sechdrs[versindex].sh_flags &= ~(unsigned long)SHF_ALLOC;
2043#ifdef CONFIG_KALLSYMS
2044 /* Keep symbol and string tables for decoding later. */
2045 sechdrs[symindex].sh_flags |= SHF_ALLOC;
2046 sechdrs[strindex].sh_flags |= SHF_ALLOC;
2047#endif
2048 2168
2049 /* Check module struct version now, before we try to use module. */ 2169 /* Check module struct version now, before we try to use module. */
2050 if (!check_modstruct_version(sechdrs, versindex, mod)) { 2170 if (!check_modstruct_version(sechdrs, versindex, mod)) {
@@ -2080,6 +2200,13 @@ static noinline struct module *load_module(void __user *umod,
2080 goto free_hdr; 2200 goto free_hdr;
2081 } 2201 }
2082 2202
2203 strmap = kzalloc(BITS_TO_LONGS(sechdrs[strindex].sh_size)
2204 * sizeof(long), GFP_KERNEL);
2205 if (!strmap) {
2206 err = -ENOMEM;
2207 goto free_mod;
2208 }
2209
2083 if (find_module(mod->name)) { 2210 if (find_module(mod->name)) {
2084 err = -EEXIST; 2211 err = -EEXIST;
2085 goto free_mod; 2212 goto free_mod;
@@ -2109,6 +2236,8 @@ static noinline struct module *load_module(void __user *umod,
2109 this is done generically; there doesn't appear to be any 2236 this is done generically; there doesn't appear to be any
2110 special cases for the architectures. */ 2237 special cases for the architectures. */
2111 layout_sections(mod, hdr, sechdrs, secstrings); 2238 layout_sections(mod, hdr, sechdrs, secstrings);
2239 symoffs = layout_symtab(mod, sechdrs, symindex, strindex, hdr,
2240 secstrings, &stroffs, strmap);
2112 2241
2113 /* Do the allocs. */ 2242 /* Do the allocs. */
2114 ptr = module_alloc_update_bounds(mod->core_size); 2243 ptr = module_alloc_update_bounds(mod->core_size);
@@ -2313,7 +2442,10 @@ static noinline struct module *load_module(void __user *umod,
2313 percpu_modcopy(mod->percpu, (void *)sechdrs[pcpuindex].sh_addr, 2442 percpu_modcopy(mod->percpu, (void *)sechdrs[pcpuindex].sh_addr,
2314 sechdrs[pcpuindex].sh_size); 2443 sechdrs[pcpuindex].sh_size);
2315 2444
2316 add_kallsyms(mod, sechdrs, symindex, strindex, secstrings); 2445 add_kallsyms(mod, sechdrs, hdr->e_shnum, symindex, strindex,
2446 symoffs, stroffs, secstrings, strmap);
2447 kfree(strmap);
2448 strmap = NULL;
2317 2449
2318 if (!mod->taints) { 2450 if (!mod->taints) {
2319 struct _ddebug *debug; 2451 struct _ddebug *debug;
@@ -2385,13 +2517,14 @@ static noinline struct module *load_module(void __user *umod,
2385 synchronize_sched(); 2517 synchronize_sched();
2386 module_arch_cleanup(mod); 2518 module_arch_cleanup(mod);
2387 cleanup: 2519 cleanup:
2520 free_modinfo(mod);
2388 kobject_del(&mod->mkobj.kobj); 2521 kobject_del(&mod->mkobj.kobj);
2389 kobject_put(&mod->mkobj.kobj); 2522 kobject_put(&mod->mkobj.kobj);
2390 free_unload: 2523 free_unload:
2391 module_unload_free(mod); 2524 module_unload_free(mod);
2392#if defined(CONFIG_MODULE_UNLOAD) && defined(CONFIG_SMP) 2525#if defined(CONFIG_MODULE_UNLOAD) && defined(CONFIG_SMP)
2393 free_init:
2394 percpu_modfree(mod->refptr); 2526 percpu_modfree(mod->refptr);
2527 free_init:
2395#endif 2528#endif
2396 module_free(mod, mod->module_init); 2529 module_free(mod, mod->module_init);
2397 free_core: 2530 free_core:
@@ -2402,6 +2535,7 @@ static noinline struct module *load_module(void __user *umod,
2402 percpu_modfree(percpu); 2535 percpu_modfree(percpu);
2403 free_mod: 2536 free_mod:
2404 kfree(args); 2537 kfree(args);
2538 kfree(strmap);
2405 free_hdr: 2539 free_hdr:
2406 vfree(hdr); 2540 vfree(hdr);
2407 return ERR_PTR(err); 2541 return ERR_PTR(err);
@@ -2491,6 +2625,11 @@ SYSCALL_DEFINE3(init_module, void __user *, umod,
2491 /* Drop initial reference. */ 2625 /* Drop initial reference. */
2492 module_put(mod); 2626 module_put(mod);
2493 trim_init_extable(mod); 2627 trim_init_extable(mod);
2628#ifdef CONFIG_KALLSYMS
2629 mod->num_symtab = mod->core_num_syms;
2630 mod->symtab = mod->core_symtab;
2631 mod->strtab = mod->core_strtab;
2632#endif
2494 module_free(mod, mod->module_init); 2633 module_free(mod, mod->module_init);
2495 mod->module_init = NULL; 2634 mod->module_init = NULL;
2496 mod->init_size = 0; 2635 mod->init_size = 0;
diff --git a/kernel/ns_cgroup.c b/kernel/ns_cgroup.c
index 5aa854f9e5ae..2a5dfec8efe0 100644
--- a/kernel/ns_cgroup.c
+++ b/kernel/ns_cgroup.c
@@ -42,8 +42,8 @@ int ns_cgroup_clone(struct task_struct *task, struct pid *pid)
42 * (hence either you are in the same cgroup as task, or in an 42 * (hence either you are in the same cgroup as task, or in an
43 * ancestor cgroup thereof) 43 * ancestor cgroup thereof)
44 */ 44 */
45static int ns_can_attach(struct cgroup_subsys *ss, 45static int ns_can_attach(struct cgroup_subsys *ss, struct cgroup *new_cgroup,
46 struct cgroup *new_cgroup, struct task_struct *task) 46 struct task_struct *task, bool threadgroup)
47{ 47{
48 if (current != task) { 48 if (current != task) {
49 if (!capable(CAP_SYS_ADMIN)) 49 if (!capable(CAP_SYS_ADMIN))
@@ -56,6 +56,18 @@ static int ns_can_attach(struct cgroup_subsys *ss,
56 if (!cgroup_is_descendant(new_cgroup, task)) 56 if (!cgroup_is_descendant(new_cgroup, task))
57 return -EPERM; 57 return -EPERM;
58 58
59 if (threadgroup) {
60 struct task_struct *c;
61 rcu_read_lock();
62 list_for_each_entry_rcu(c, &task->thread_group, thread_group) {
63 if (!cgroup_is_descendant(new_cgroup, c)) {
64 rcu_read_unlock();
65 return -EPERM;
66 }
67 }
68 rcu_read_unlock();
69 }
70
59 return 0; 71 return 0;
60} 72}
61 73
diff --git a/kernel/params.c b/kernel/params.c
index 7f6912ced2ba..9da58eabdcb2 100644
--- a/kernel/params.c
+++ b/kernel/params.c
@@ -23,6 +23,7 @@
23#include <linux/device.h> 23#include <linux/device.h>
24#include <linux/err.h> 24#include <linux/err.h>
25#include <linux/slab.h> 25#include <linux/slab.h>
26#include <linux/ctype.h>
26 27
27#if 0 28#if 0
28#define DEBUGP printk 29#define DEBUGP printk
@@ -87,7 +88,7 @@ static char *next_arg(char *args, char **param, char **val)
87 } 88 }
88 89
89 for (i = 0; args[i]; i++) { 90 for (i = 0; args[i]; i++) {
90 if (args[i] == ' ' && !in_quote) 91 if (isspace(args[i]) && !in_quote)
91 break; 92 break;
92 if (equals == 0) { 93 if (equals == 0) {
93 if (args[i] == '=') 94 if (args[i] == '=')
@@ -121,7 +122,7 @@ static char *next_arg(char *args, char **param, char **val)
121 next = args + i; 122 next = args + i;
122 123
123 /* Chew up trailing spaces. */ 124 /* Chew up trailing spaces. */
124 while (*next == ' ') 125 while (isspace(*next))
125 next++; 126 next++;
126 return next; 127 return next;
127} 128}
@@ -138,7 +139,7 @@ int parse_args(const char *name,
138 DEBUGP("Parsing ARGS: %s\n", args); 139 DEBUGP("Parsing ARGS: %s\n", args);
139 140
140 /* Chew leading spaces */ 141 /* Chew leading spaces */
141 while (*args == ' ') 142 while (isspace(*args))
142 args++; 143 args++;
143 144
144 while (*args) { 145 while (*args) {
diff --git a/kernel/pid_namespace.c b/kernel/pid_namespace.c
index 821722ae58a7..86b3796b0436 100644
--- a/kernel/pid_namespace.c
+++ b/kernel/pid_namespace.c
@@ -118,7 +118,7 @@ struct pid_namespace *copy_pid_ns(unsigned long flags, struct pid_namespace *old
118{ 118{
119 if (!(flags & CLONE_NEWPID)) 119 if (!(flags & CLONE_NEWPID))
120 return get_pid_ns(old_ns); 120 return get_pid_ns(old_ns);
121 if (flags & CLONE_THREAD) 121 if (flags & (CLONE_THREAD|CLONE_PARENT))
122 return ERR_PTR(-EINVAL); 122 return ERR_PTR(-EINVAL);
123 return create_pid_namespace(old_ns); 123 return create_pid_namespace(old_ns);
124} 124}
diff --git a/kernel/ptrace.c b/kernel/ptrace.c
index 307c285af59e..23bd09cd042e 100644
--- a/kernel/ptrace.c
+++ b/kernel/ptrace.c
@@ -266,9 +266,10 @@ static int ignoring_children(struct sighand_struct *sigh)
266 * or self-reaping. Do notification now if it would have happened earlier. 266 * or self-reaping. Do notification now if it would have happened earlier.
267 * If it should reap itself, return true. 267 * If it should reap itself, return true.
268 * 268 *
269 * If it's our own child, there is no notification to do. 269 * If it's our own child, there is no notification to do. But if our normal
270 * But if our normal children self-reap, then this child 270 * children self-reap, then this child was prevented by ptrace and we must
271 * was prevented by ptrace and we must reap it now. 271 * reap it now, in that case we must also wake up sub-threads sleeping in
272 * do_wait().
272 */ 273 */
273static bool __ptrace_detach(struct task_struct *tracer, struct task_struct *p) 274static bool __ptrace_detach(struct task_struct *tracer, struct task_struct *p)
274{ 275{
@@ -278,8 +279,10 @@ static bool __ptrace_detach(struct task_struct *tracer, struct task_struct *p)
278 if (!task_detached(p) && thread_group_empty(p)) { 279 if (!task_detached(p) && thread_group_empty(p)) {
279 if (!same_thread_group(p->real_parent, tracer)) 280 if (!same_thread_group(p->real_parent, tracer))
280 do_notify_parent(p, p->exit_signal); 281 do_notify_parent(p, p->exit_signal);
281 else if (ignoring_children(tracer->sighand)) 282 else if (ignoring_children(tracer->sighand)) {
283 __wake_up_parent(p, tracer);
282 p->exit_signal = -1; 284 p->exit_signal = -1;
285 }
283 } 286 }
284 if (task_detached(p)) { 287 if (task_detached(p)) {
285 /* Mark it as in the process of being reaped. */ 288 /* Mark it as in the process of being reaped. */
diff --git a/kernel/res_counter.c b/kernel/res_counter.c
index e1338f074314..88faec23e833 100644
--- a/kernel/res_counter.c
+++ b/kernel/res_counter.c
@@ -19,6 +19,7 @@ void res_counter_init(struct res_counter *counter, struct res_counter *parent)
19{ 19{
20 spin_lock_init(&counter->lock); 20 spin_lock_init(&counter->lock);
21 counter->limit = RESOURCE_MAX; 21 counter->limit = RESOURCE_MAX;
22 counter->soft_limit = RESOURCE_MAX;
22 counter->parent = parent; 23 counter->parent = parent;
23} 24}
24 25
@@ -36,17 +37,27 @@ int res_counter_charge_locked(struct res_counter *counter, unsigned long val)
36} 37}
37 38
38int res_counter_charge(struct res_counter *counter, unsigned long val, 39int res_counter_charge(struct res_counter *counter, unsigned long val,
39 struct res_counter **limit_fail_at) 40 struct res_counter **limit_fail_at,
41 struct res_counter **soft_limit_fail_at)
40{ 42{
41 int ret; 43 int ret;
42 unsigned long flags; 44 unsigned long flags;
43 struct res_counter *c, *u; 45 struct res_counter *c, *u;
44 46
45 *limit_fail_at = NULL; 47 *limit_fail_at = NULL;
48 if (soft_limit_fail_at)
49 *soft_limit_fail_at = NULL;
46 local_irq_save(flags); 50 local_irq_save(flags);
47 for (c = counter; c != NULL; c = c->parent) { 51 for (c = counter; c != NULL; c = c->parent) {
48 spin_lock(&c->lock); 52 spin_lock(&c->lock);
49 ret = res_counter_charge_locked(c, val); 53 ret = res_counter_charge_locked(c, val);
54 /*
55 * With soft limits, we return the highest ancestor
56 * that exceeds its soft limit
57 */
58 if (soft_limit_fail_at &&
59 !res_counter_soft_limit_check_locked(c))
60 *soft_limit_fail_at = c;
50 spin_unlock(&c->lock); 61 spin_unlock(&c->lock);
51 if (ret < 0) { 62 if (ret < 0) {
52 *limit_fail_at = c; 63 *limit_fail_at = c;
@@ -74,7 +85,8 @@ void res_counter_uncharge_locked(struct res_counter *counter, unsigned long val)
74 counter->usage -= val; 85 counter->usage -= val;
75} 86}
76 87
77void res_counter_uncharge(struct res_counter *counter, unsigned long val) 88void res_counter_uncharge(struct res_counter *counter, unsigned long val,
89 bool *was_soft_limit_excess)
78{ 90{
79 unsigned long flags; 91 unsigned long flags;
80 struct res_counter *c; 92 struct res_counter *c;
@@ -82,6 +94,9 @@ void res_counter_uncharge(struct res_counter *counter, unsigned long val)
82 local_irq_save(flags); 94 local_irq_save(flags);
83 for (c = counter; c != NULL; c = c->parent) { 95 for (c = counter; c != NULL; c = c->parent) {
84 spin_lock(&c->lock); 96 spin_lock(&c->lock);
97 if (was_soft_limit_excess)
98 *was_soft_limit_excess =
99 !res_counter_soft_limit_check_locked(c);
85 res_counter_uncharge_locked(c, val); 100 res_counter_uncharge_locked(c, val);
86 spin_unlock(&c->lock); 101 spin_unlock(&c->lock);
87 } 102 }
@@ -101,6 +116,8 @@ res_counter_member(struct res_counter *counter, int member)
101 return &counter->limit; 116 return &counter->limit;
102 case RES_FAILCNT: 117 case RES_FAILCNT:
103 return &counter->failcnt; 118 return &counter->failcnt;
119 case RES_SOFT_LIMIT:
120 return &counter->soft_limit;
104 }; 121 };
105 122
106 BUG(); 123 BUG();
diff --git a/kernel/sched.c b/kernel/sched.c
index 2f76e06bea58..ee61f454a98b 100644
--- a/kernel/sched.c
+++ b/kernel/sched.c
@@ -10312,7 +10312,7 @@ static int sched_rt_global_constraints(void)
10312#endif /* CONFIG_RT_GROUP_SCHED */ 10312#endif /* CONFIG_RT_GROUP_SCHED */
10313 10313
10314int sched_rt_handler(struct ctl_table *table, int write, 10314int sched_rt_handler(struct ctl_table *table, int write,
10315 struct file *filp, void __user *buffer, size_t *lenp, 10315 void __user *buffer, size_t *lenp,
10316 loff_t *ppos) 10316 loff_t *ppos)
10317{ 10317{
10318 int ret; 10318 int ret;
@@ -10323,7 +10323,7 @@ int sched_rt_handler(struct ctl_table *table, int write,
10323 old_period = sysctl_sched_rt_period; 10323 old_period = sysctl_sched_rt_period;
10324 old_runtime = sysctl_sched_rt_runtime; 10324 old_runtime = sysctl_sched_rt_runtime;
10325 10325
10326 ret = proc_dointvec(table, write, filp, buffer, lenp, ppos); 10326 ret = proc_dointvec(table, write, buffer, lenp, ppos);
10327 10327
10328 if (!ret && write) { 10328 if (!ret && write) {
10329 ret = sched_rt_global_constraints(); 10329 ret = sched_rt_global_constraints();
@@ -10377,8 +10377,7 @@ cpu_cgroup_destroy(struct cgroup_subsys *ss, struct cgroup *cgrp)
10377} 10377}
10378 10378
10379static int 10379static int
10380cpu_cgroup_can_attach(struct cgroup_subsys *ss, struct cgroup *cgrp, 10380cpu_cgroup_can_attach_task(struct cgroup *cgrp, struct task_struct *tsk)
10381 struct task_struct *tsk)
10382{ 10381{
10383#ifdef CONFIG_RT_GROUP_SCHED 10382#ifdef CONFIG_RT_GROUP_SCHED
10384 if (!sched_rt_can_attach(cgroup_tg(cgrp), tsk)) 10383 if (!sched_rt_can_attach(cgroup_tg(cgrp), tsk))
@@ -10388,15 +10387,45 @@ cpu_cgroup_can_attach(struct cgroup_subsys *ss, struct cgroup *cgrp,
10388 if (tsk->sched_class != &fair_sched_class) 10387 if (tsk->sched_class != &fair_sched_class)
10389 return -EINVAL; 10388 return -EINVAL;
10390#endif 10389#endif
10390 return 0;
10391}
10391 10392
10393static int
10394cpu_cgroup_can_attach(struct cgroup_subsys *ss, struct cgroup *cgrp,
10395 struct task_struct *tsk, bool threadgroup)
10396{
10397 int retval = cpu_cgroup_can_attach_task(cgrp, tsk);
10398 if (retval)
10399 return retval;
10400 if (threadgroup) {
10401 struct task_struct *c;
10402 rcu_read_lock();
10403 list_for_each_entry_rcu(c, &tsk->thread_group, thread_group) {
10404 retval = cpu_cgroup_can_attach_task(cgrp, c);
10405 if (retval) {
10406 rcu_read_unlock();
10407 return retval;
10408 }
10409 }
10410 rcu_read_unlock();
10411 }
10392 return 0; 10412 return 0;
10393} 10413}
10394 10414
10395static void 10415static void
10396cpu_cgroup_attach(struct cgroup_subsys *ss, struct cgroup *cgrp, 10416cpu_cgroup_attach(struct cgroup_subsys *ss, struct cgroup *cgrp,
10397 struct cgroup *old_cont, struct task_struct *tsk) 10417 struct cgroup *old_cont, struct task_struct *tsk,
10418 bool threadgroup)
10398{ 10419{
10399 sched_move_task(tsk); 10420 sched_move_task(tsk);
10421 if (threadgroup) {
10422 struct task_struct *c;
10423 rcu_read_lock();
10424 list_for_each_entry_rcu(c, &tsk->thread_group, thread_group) {
10425 sched_move_task(c);
10426 }
10427 rcu_read_unlock();
10428 }
10400} 10429}
10401 10430
10402#ifdef CONFIG_FAIR_GROUP_SCHED 10431#ifdef CONFIG_FAIR_GROUP_SCHED
diff --git a/kernel/sched_fair.c b/kernel/sched_fair.c
index ecc637a0d591..4e777b47eeda 100644
--- a/kernel/sched_fair.c
+++ b/kernel/sched_fair.c
@@ -384,10 +384,10 @@ static struct sched_entity *__pick_last_entity(struct cfs_rq *cfs_rq)
384 384
385#ifdef CONFIG_SCHED_DEBUG 385#ifdef CONFIG_SCHED_DEBUG
386int sched_nr_latency_handler(struct ctl_table *table, int write, 386int sched_nr_latency_handler(struct ctl_table *table, int write,
387 struct file *filp, void __user *buffer, size_t *lenp, 387 void __user *buffer, size_t *lenp,
388 loff_t *ppos) 388 loff_t *ppos)
389{ 389{
390 int ret = proc_dointvec_minmax(table, write, filp, buffer, lenp, ppos); 390 int ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
391 391
392 if (ret || !write) 392 if (ret || !write)
393 return ret; 393 return ret;
diff --git a/kernel/signal.c b/kernel/signal.c
index 64c5deeaca5d..6705320784fd 100644
--- a/kernel/signal.c
+++ b/kernel/signal.c
@@ -705,7 +705,7 @@ static int prepare_signal(int sig, struct task_struct *p, int from_ancestor_ns)
705 705
706 if (why) { 706 if (why) {
707 /* 707 /*
708 * The first thread which returns from finish_stop() 708 * The first thread which returns from do_signal_stop()
709 * will take ->siglock, notice SIGNAL_CLD_MASK, and 709 * will take ->siglock, notice SIGNAL_CLD_MASK, and
710 * notify its parent. See get_signal_to_deliver(). 710 * notify its parent. See get_signal_to_deliver().
711 */ 711 */
@@ -971,6 +971,20 @@ specific_send_sig_info(int sig, struct siginfo *info, struct task_struct *t)
971 return send_signal(sig, info, t, 0); 971 return send_signal(sig, info, t, 0);
972} 972}
973 973
974int do_send_sig_info(int sig, struct siginfo *info, struct task_struct *p,
975 bool group)
976{
977 unsigned long flags;
978 int ret = -ESRCH;
979
980 if (lock_task_sighand(p, &flags)) {
981 ret = send_signal(sig, info, p, group);
982 unlock_task_sighand(p, &flags);
983 }
984
985 return ret;
986}
987
974/* 988/*
975 * Force a signal that the process can't ignore: if necessary 989 * Force a signal that the process can't ignore: if necessary
976 * we unblock the signal and change any SIG_IGN to SIG_DFL. 990 * we unblock the signal and change any SIG_IGN to SIG_DFL.
@@ -1036,12 +1050,6 @@ void zap_other_threads(struct task_struct *p)
1036 } 1050 }
1037} 1051}
1038 1052
1039int __fatal_signal_pending(struct task_struct *tsk)
1040{
1041 return sigismember(&tsk->pending.signal, SIGKILL);
1042}
1043EXPORT_SYMBOL(__fatal_signal_pending);
1044
1045struct sighand_struct *lock_task_sighand(struct task_struct *tsk, unsigned long *flags) 1053struct sighand_struct *lock_task_sighand(struct task_struct *tsk, unsigned long *flags)
1046{ 1054{
1047 struct sighand_struct *sighand; 1055 struct sighand_struct *sighand;
@@ -1068,18 +1076,10 @@ struct sighand_struct *lock_task_sighand(struct task_struct *tsk, unsigned long
1068 */ 1076 */
1069int group_send_sig_info(int sig, struct siginfo *info, struct task_struct *p) 1077int group_send_sig_info(int sig, struct siginfo *info, struct task_struct *p)
1070{ 1078{
1071 unsigned long flags; 1079 int ret = check_kill_permission(sig, info, p);
1072 int ret;
1073 1080
1074 ret = check_kill_permission(sig, info, p); 1081 if (!ret && sig)
1075 1082 ret = do_send_sig_info(sig, info, p, true);
1076 if (!ret && sig) {
1077 ret = -ESRCH;
1078 if (lock_task_sighand(p, &flags)) {
1079 ret = __group_send_sig_info(sig, info, p);
1080 unlock_task_sighand(p, &flags);
1081 }
1082 }
1083 1083
1084 return ret; 1084 return ret;
1085} 1085}
@@ -1224,15 +1224,9 @@ static int kill_something_info(int sig, struct siginfo *info, pid_t pid)
1224 * These are for backward compatibility with the rest of the kernel source. 1224 * These are for backward compatibility with the rest of the kernel source.
1225 */ 1225 */
1226 1226
1227/*
1228 * The caller must ensure the task can't exit.
1229 */
1230int 1227int
1231send_sig_info(int sig, struct siginfo *info, struct task_struct *p) 1228send_sig_info(int sig, struct siginfo *info, struct task_struct *p)
1232{ 1229{
1233 int ret;
1234 unsigned long flags;
1235
1236 /* 1230 /*
1237 * Make sure legacy kernel users don't send in bad values 1231 * Make sure legacy kernel users don't send in bad values
1238 * (normal paths check this in check_kill_permission). 1232 * (normal paths check this in check_kill_permission).
@@ -1240,10 +1234,7 @@ send_sig_info(int sig, struct siginfo *info, struct task_struct *p)
1240 if (!valid_signal(sig)) 1234 if (!valid_signal(sig))
1241 return -EINVAL; 1235 return -EINVAL;
1242 1236
1243 spin_lock_irqsave(&p->sighand->siglock, flags); 1237 return do_send_sig_info(sig, info, p, false);
1244 ret = specific_send_sig_info(sig, info, p);
1245 spin_unlock_irqrestore(&p->sighand->siglock, flags);
1246 return ret;
1247} 1238}
1248 1239
1249#define __si_special(priv) \ 1240#define __si_special(priv) \
@@ -1383,15 +1374,6 @@ ret:
1383} 1374}
1384 1375
1385/* 1376/*
1386 * Wake up any threads in the parent blocked in wait* syscalls.
1387 */
1388static inline void __wake_up_parent(struct task_struct *p,
1389 struct task_struct *parent)
1390{
1391 wake_up_interruptible_sync(&parent->signal->wait_chldexit);
1392}
1393
1394/*
1395 * Let a parent know about the death of a child. 1377 * Let a parent know about the death of a child.
1396 * For a stopped/continued status change, use do_notify_parent_cldstop instead. 1378 * For a stopped/continued status change, use do_notify_parent_cldstop instead.
1397 * 1379 *
@@ -1673,29 +1655,6 @@ void ptrace_notify(int exit_code)
1673 spin_unlock_irq(&current->sighand->siglock); 1655 spin_unlock_irq(&current->sighand->siglock);
1674} 1656}
1675 1657
1676static void
1677finish_stop(int stop_count)
1678{
1679 /*
1680 * If there are no other threads in the group, or if there is
1681 * a group stop in progress and we are the last to stop,
1682 * report to the parent. When ptraced, every thread reports itself.
1683 */
1684 if (tracehook_notify_jctl(stop_count == 0, CLD_STOPPED)) {
1685 read_lock(&tasklist_lock);
1686 do_notify_parent_cldstop(current, CLD_STOPPED);
1687 read_unlock(&tasklist_lock);
1688 }
1689
1690 do {
1691 schedule();
1692 } while (try_to_freeze());
1693 /*
1694 * Now we don't run again until continued.
1695 */
1696 current->exit_code = 0;
1697}
1698
1699/* 1658/*
1700 * This performs the stopping for SIGSTOP and other stop signals. 1659 * This performs the stopping for SIGSTOP and other stop signals.
1701 * We have to stop all threads in the thread group. 1660 * We have to stop all threads in the thread group.
@@ -1705,15 +1664,9 @@ finish_stop(int stop_count)
1705static int do_signal_stop(int signr) 1664static int do_signal_stop(int signr)
1706{ 1665{
1707 struct signal_struct *sig = current->signal; 1666 struct signal_struct *sig = current->signal;
1708 int stop_count; 1667 int notify;
1709 1668
1710 if (sig->group_stop_count > 0) { 1669 if (!sig->group_stop_count) {
1711 /*
1712 * There is a group stop in progress. We don't need to
1713 * start another one.
1714 */
1715 stop_count = --sig->group_stop_count;
1716 } else {
1717 struct task_struct *t; 1670 struct task_struct *t;
1718 1671
1719 if (!likely(sig->flags & SIGNAL_STOP_DEQUEUED) || 1672 if (!likely(sig->flags & SIGNAL_STOP_DEQUEUED) ||
@@ -1725,7 +1678,7 @@ static int do_signal_stop(int signr)
1725 */ 1678 */
1726 sig->group_exit_code = signr; 1679 sig->group_exit_code = signr;
1727 1680
1728 stop_count = 0; 1681 sig->group_stop_count = 1;
1729 for (t = next_thread(current); t != current; t = next_thread(t)) 1682 for (t = next_thread(current); t != current; t = next_thread(t))
1730 /* 1683 /*
1731 * Setting state to TASK_STOPPED for a group 1684 * Setting state to TASK_STOPPED for a group
@@ -1734,19 +1687,44 @@ static int do_signal_stop(int signr)
1734 */ 1687 */
1735 if (!(t->flags & PF_EXITING) && 1688 if (!(t->flags & PF_EXITING) &&
1736 !task_is_stopped_or_traced(t)) { 1689 !task_is_stopped_or_traced(t)) {
1737 stop_count++; 1690 sig->group_stop_count++;
1738 signal_wake_up(t, 0); 1691 signal_wake_up(t, 0);
1739 } 1692 }
1740 sig->group_stop_count = stop_count;
1741 } 1693 }
1694 /*
1695 * If there are no other threads in the group, or if there is
1696 * a group stop in progress and we are the last to stop, report
1697 * to the parent. When ptraced, every thread reports itself.
1698 */
1699 notify = sig->group_stop_count == 1 ? CLD_STOPPED : 0;
1700 notify = tracehook_notify_jctl(notify, CLD_STOPPED);
1701 /*
1702 * tracehook_notify_jctl() can drop and reacquire siglock, so
1703 * we keep ->group_stop_count != 0 before the call. If SIGCONT
1704 * or SIGKILL comes in between ->group_stop_count == 0.
1705 */
1706 if (sig->group_stop_count) {
1707 if (!--sig->group_stop_count)
1708 sig->flags = SIGNAL_STOP_STOPPED;
1709 current->exit_code = sig->group_exit_code;
1710 __set_current_state(TASK_STOPPED);
1711 }
1712 spin_unlock_irq(&current->sighand->siglock);
1742 1713
1743 if (stop_count == 0) 1714 if (notify) {
1744 sig->flags = SIGNAL_STOP_STOPPED; 1715 read_lock(&tasklist_lock);
1745 current->exit_code = sig->group_exit_code; 1716 do_notify_parent_cldstop(current, notify);
1746 __set_current_state(TASK_STOPPED); 1717 read_unlock(&tasklist_lock);
1718 }
1719
1720 /* Now we don't run again until woken by SIGCONT or SIGKILL */
1721 do {
1722 schedule();
1723 } while (try_to_freeze());
1724
1725 tracehook_finish_jctl();
1726 current->exit_code = 0;
1747 1727
1748 spin_unlock_irq(&current->sighand->siglock);
1749 finish_stop(stop_count);
1750 return 1; 1728 return 1;
1751} 1729}
1752 1730
@@ -1815,14 +1793,15 @@ relock:
1815 int why = (signal->flags & SIGNAL_STOP_CONTINUED) 1793 int why = (signal->flags & SIGNAL_STOP_CONTINUED)
1816 ? CLD_CONTINUED : CLD_STOPPED; 1794 ? CLD_CONTINUED : CLD_STOPPED;
1817 signal->flags &= ~SIGNAL_CLD_MASK; 1795 signal->flags &= ~SIGNAL_CLD_MASK;
1818 spin_unlock_irq(&sighand->siglock);
1819 1796
1820 if (unlikely(!tracehook_notify_jctl(1, why))) 1797 why = tracehook_notify_jctl(why, CLD_CONTINUED);
1821 goto relock; 1798 spin_unlock_irq(&sighand->siglock);
1822 1799
1823 read_lock(&tasklist_lock); 1800 if (why) {
1824 do_notify_parent_cldstop(current->group_leader, why); 1801 read_lock(&tasklist_lock);
1825 read_unlock(&tasklist_lock); 1802 do_notify_parent_cldstop(current->group_leader, why);
1803 read_unlock(&tasklist_lock);
1804 }
1826 goto relock; 1805 goto relock;
1827 } 1806 }
1828 1807
@@ -1987,14 +1966,14 @@ void exit_signals(struct task_struct *tsk)
1987 if (unlikely(tsk->signal->group_stop_count) && 1966 if (unlikely(tsk->signal->group_stop_count) &&
1988 !--tsk->signal->group_stop_count) { 1967 !--tsk->signal->group_stop_count) {
1989 tsk->signal->flags = SIGNAL_STOP_STOPPED; 1968 tsk->signal->flags = SIGNAL_STOP_STOPPED;
1990 group_stop = 1; 1969 group_stop = tracehook_notify_jctl(CLD_STOPPED, CLD_STOPPED);
1991 } 1970 }
1992out: 1971out:
1993 spin_unlock_irq(&tsk->sighand->siglock); 1972 spin_unlock_irq(&tsk->sighand->siglock);
1994 1973
1995 if (unlikely(group_stop) && tracehook_notify_jctl(1, CLD_STOPPED)) { 1974 if (unlikely(group_stop)) {
1996 read_lock(&tasklist_lock); 1975 read_lock(&tasklist_lock);
1997 do_notify_parent_cldstop(tsk, CLD_STOPPED); 1976 do_notify_parent_cldstop(tsk, group_stop);
1998 read_unlock(&tasklist_lock); 1977 read_unlock(&tasklist_lock);
1999 } 1978 }
2000} 1979}
@@ -2290,7 +2269,6 @@ static int
2290do_send_specific(pid_t tgid, pid_t pid, int sig, struct siginfo *info) 2269do_send_specific(pid_t tgid, pid_t pid, int sig, struct siginfo *info)
2291{ 2270{
2292 struct task_struct *p; 2271 struct task_struct *p;
2293 unsigned long flags;
2294 int error = -ESRCH; 2272 int error = -ESRCH;
2295 2273
2296 rcu_read_lock(); 2274 rcu_read_lock();
@@ -2300,14 +2278,16 @@ do_send_specific(pid_t tgid, pid_t pid, int sig, struct siginfo *info)
2300 /* 2278 /*
2301 * The null signal is a permissions and process existence 2279 * The null signal is a permissions and process existence
2302 * probe. No signal is actually delivered. 2280 * probe. No signal is actually delivered.
2303 *
2304 * If lock_task_sighand() fails we pretend the task dies
2305 * after receiving the signal. The window is tiny, and the
2306 * signal is private anyway.
2307 */ 2281 */
2308 if (!error && sig && lock_task_sighand(p, &flags)) { 2282 if (!error && sig) {
2309 error = specific_send_sig_info(sig, info, p); 2283 error = do_send_sig_info(sig, info, p, false);
2310 unlock_task_sighand(p, &flags); 2284 /*
2285 * If lock_task_sighand() failed we pretend the task
2286 * dies after receiving the signal. The window is tiny,
2287 * and the signal is private anyway.
2288 */
2289 if (unlikely(error == -ESRCH))
2290 error = 0;
2311 } 2291 }
2312 } 2292 }
2313 rcu_read_unlock(); 2293 rcu_read_unlock();
diff --git a/kernel/slow-work.c b/kernel/slow-work.c
index 09d7519557d3..0d31135efbf4 100644
--- a/kernel/slow-work.c
+++ b/kernel/slow-work.c
@@ -26,10 +26,10 @@ static void slow_work_cull_timeout(unsigned long);
26static void slow_work_oom_timeout(unsigned long); 26static void slow_work_oom_timeout(unsigned long);
27 27
28#ifdef CONFIG_SYSCTL 28#ifdef CONFIG_SYSCTL
29static int slow_work_min_threads_sysctl(struct ctl_table *, int, struct file *, 29static int slow_work_min_threads_sysctl(struct ctl_table *, int,
30 void __user *, size_t *, loff_t *); 30 void __user *, size_t *, loff_t *);
31 31
32static int slow_work_max_threads_sysctl(struct ctl_table *, int , struct file *, 32static int slow_work_max_threads_sysctl(struct ctl_table *, int ,
33 void __user *, size_t *, loff_t *); 33 void __user *, size_t *, loff_t *);
34#endif 34#endif
35 35
@@ -493,10 +493,10 @@ static void slow_work_oom_timeout(unsigned long data)
493 * Handle adjustment of the minimum number of threads 493 * Handle adjustment of the minimum number of threads
494 */ 494 */
495static int slow_work_min_threads_sysctl(struct ctl_table *table, int write, 495static int slow_work_min_threads_sysctl(struct ctl_table *table, int write,
496 struct file *filp, void __user *buffer, 496 void __user *buffer,
497 size_t *lenp, loff_t *ppos) 497 size_t *lenp, loff_t *ppos)
498{ 498{
499 int ret = proc_dointvec_minmax(table, write, filp, buffer, lenp, ppos); 499 int ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
500 int n; 500 int n;
501 501
502 if (ret == 0) { 502 if (ret == 0) {
@@ -521,10 +521,10 @@ static int slow_work_min_threads_sysctl(struct ctl_table *table, int write,
521 * Handle adjustment of the maximum number of threads 521 * Handle adjustment of the maximum number of threads
522 */ 522 */
523static int slow_work_max_threads_sysctl(struct ctl_table *table, int write, 523static int slow_work_max_threads_sysctl(struct ctl_table *table, int write,
524 struct file *filp, void __user *buffer, 524 void __user *buffer,
525 size_t *lenp, loff_t *ppos) 525 size_t *lenp, loff_t *ppos)
526{ 526{
527 int ret = proc_dointvec_minmax(table, write, filp, buffer, lenp, ppos); 527 int ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
528 int n; 528 int n;
529 529
530 if (ret == 0) { 530 if (ret == 0) {
diff --git a/kernel/softlockup.c b/kernel/softlockup.c
index 88796c330838..81324d12eb35 100644
--- a/kernel/softlockup.c
+++ b/kernel/softlockup.c
@@ -90,11 +90,11 @@ void touch_all_softlockup_watchdogs(void)
90EXPORT_SYMBOL(touch_all_softlockup_watchdogs); 90EXPORT_SYMBOL(touch_all_softlockup_watchdogs);
91 91
92int proc_dosoftlockup_thresh(struct ctl_table *table, int write, 92int proc_dosoftlockup_thresh(struct ctl_table *table, int write,
93 struct file *filp, void __user *buffer, 93 void __user *buffer,
94 size_t *lenp, loff_t *ppos) 94 size_t *lenp, loff_t *ppos)
95{ 95{
96 touch_all_softlockup_watchdogs(); 96 touch_all_softlockup_watchdogs();
97 return proc_dointvec_minmax(table, write, filp, buffer, lenp, ppos); 97 return proc_dointvec_minmax(table, write, buffer, lenp, ppos);
98} 98}
99 99
100/* 100/*
diff --git a/kernel/sys.c b/kernel/sys.c
index ebcb15611728..255475d163e0 100644
--- a/kernel/sys.c
+++ b/kernel/sys.c
@@ -1542,6 +1542,28 @@ SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
1542 current->timer_slack_ns = arg2; 1542 current->timer_slack_ns = arg2;
1543 error = 0; 1543 error = 0;
1544 break; 1544 break;
1545 case PR_MCE_KILL:
1546 if (arg4 | arg5)
1547 return -EINVAL;
1548 switch (arg2) {
1549 case 0:
1550 if (arg3 != 0)
1551 return -EINVAL;
1552 current->flags &= ~PF_MCE_PROCESS;
1553 break;
1554 case 1:
1555 current->flags |= PF_MCE_PROCESS;
1556 if (arg3 != 0)
1557 current->flags |= PF_MCE_EARLY;
1558 else
1559 current->flags &= ~PF_MCE_EARLY;
1560 break;
1561 default:
1562 return -EINVAL;
1563 }
1564 error = 0;
1565 break;
1566
1545 default: 1567 default:
1546 error = -EINVAL; 1568 error = -EINVAL;
1547 break; 1569 break;
diff --git a/kernel/sysctl.c b/kernel/sysctl.c
index 7f4f57bea4ce..0d949c517412 100644
--- a/kernel/sysctl.c
+++ b/kernel/sysctl.c
@@ -76,6 +76,7 @@ extern int max_threads;
76extern int core_uses_pid; 76extern int core_uses_pid;
77extern int suid_dumpable; 77extern int suid_dumpable;
78extern char core_pattern[]; 78extern char core_pattern[];
79extern unsigned int core_pipe_limit;
79extern int pid_max; 80extern int pid_max;
80extern int min_free_kbytes; 81extern int min_free_kbytes;
81extern int pid_max_min, pid_max_max; 82extern int pid_max_min, pid_max_max;
@@ -162,9 +163,9 @@ extern int max_lock_depth;
162#endif 163#endif
163 164
164#ifdef CONFIG_PROC_SYSCTL 165#ifdef CONFIG_PROC_SYSCTL
165static int proc_do_cad_pid(struct ctl_table *table, int write, struct file *filp, 166static int proc_do_cad_pid(struct ctl_table *table, int write,
166 void __user *buffer, size_t *lenp, loff_t *ppos); 167 void __user *buffer, size_t *lenp, loff_t *ppos);
167static int proc_taint(struct ctl_table *table, int write, struct file *filp, 168static int proc_taint(struct ctl_table *table, int write,
168 void __user *buffer, size_t *lenp, loff_t *ppos); 169 void __user *buffer, size_t *lenp, loff_t *ppos);
169#endif 170#endif
170 171
@@ -423,6 +424,14 @@ static struct ctl_table kern_table[] = {
423 .proc_handler = &proc_dostring, 424 .proc_handler = &proc_dostring,
424 .strategy = &sysctl_string, 425 .strategy = &sysctl_string,
425 }, 426 },
427 {
428 .ctl_name = CTL_UNNUMBERED,
429 .procname = "core_pipe_limit",
430 .data = &core_pipe_limit,
431 .maxlen = sizeof(unsigned int),
432 .mode = 0644,
433 .proc_handler = &proc_dointvec,
434 },
426#ifdef CONFIG_PROC_SYSCTL 435#ifdef CONFIG_PROC_SYSCTL
427 { 436 {
428 .procname = "tainted", 437 .procname = "tainted",
@@ -1389,6 +1398,31 @@ static struct ctl_table vm_table[] = {
1389 .mode = 0644, 1398 .mode = 0644,
1390 .proc_handler = &scan_unevictable_handler, 1399 .proc_handler = &scan_unevictable_handler,
1391 }, 1400 },
1401#ifdef CONFIG_MEMORY_FAILURE
1402 {
1403 .ctl_name = CTL_UNNUMBERED,
1404 .procname = "memory_failure_early_kill",
1405 .data = &sysctl_memory_failure_early_kill,
1406 .maxlen = sizeof(sysctl_memory_failure_early_kill),
1407 .mode = 0644,
1408 .proc_handler = &proc_dointvec_minmax,
1409 .strategy = &sysctl_intvec,
1410 .extra1 = &zero,
1411 .extra2 = &one,
1412 },
1413 {
1414 .ctl_name = CTL_UNNUMBERED,
1415 .procname = "memory_failure_recovery",
1416 .data = &sysctl_memory_failure_recovery,
1417 .maxlen = sizeof(sysctl_memory_failure_recovery),
1418 .mode = 0644,
1419 .proc_handler = &proc_dointvec_minmax,
1420 .strategy = &sysctl_intvec,
1421 .extra1 = &zero,
1422 .extra2 = &one,
1423 },
1424#endif
1425
1392/* 1426/*
1393 * NOTE: do not add new entries to this table unless you have read 1427 * NOTE: do not add new entries to this table unless you have read
1394 * Documentation/sysctl/ctl_unnumbered.txt 1428 * Documentation/sysctl/ctl_unnumbered.txt
@@ -2217,7 +2251,7 @@ void sysctl_head_put(struct ctl_table_header *head)
2217#ifdef CONFIG_PROC_SYSCTL 2251#ifdef CONFIG_PROC_SYSCTL
2218 2252
2219static int _proc_do_string(void* data, int maxlen, int write, 2253static int _proc_do_string(void* data, int maxlen, int write,
2220 struct file *filp, void __user *buffer, 2254 void __user *buffer,
2221 size_t *lenp, loff_t *ppos) 2255 size_t *lenp, loff_t *ppos)
2222{ 2256{
2223 size_t len; 2257 size_t len;
@@ -2278,7 +2312,6 @@ static int _proc_do_string(void* data, int maxlen, int write,
2278 * proc_dostring - read a string sysctl 2312 * proc_dostring - read a string sysctl
2279 * @table: the sysctl table 2313 * @table: the sysctl table
2280 * @write: %TRUE if this is a write to the sysctl file 2314 * @write: %TRUE if this is a write to the sysctl file
2281 * @filp: the file structure
2282 * @buffer: the user buffer 2315 * @buffer: the user buffer
2283 * @lenp: the size of the user buffer 2316 * @lenp: the size of the user buffer
2284 * @ppos: file position 2317 * @ppos: file position
@@ -2292,10 +2325,10 @@ static int _proc_do_string(void* data, int maxlen, int write,
2292 * 2325 *
2293 * Returns 0 on success. 2326 * Returns 0 on success.
2294 */ 2327 */
2295int proc_dostring(struct ctl_table *table, int write, struct file *filp, 2328int proc_dostring(struct ctl_table *table, int write,
2296 void __user *buffer, size_t *lenp, loff_t *ppos) 2329 void __user *buffer, size_t *lenp, loff_t *ppos)
2297{ 2330{
2298 return _proc_do_string(table->data, table->maxlen, write, filp, 2331 return _proc_do_string(table->data, table->maxlen, write,
2299 buffer, lenp, ppos); 2332 buffer, lenp, ppos);
2300} 2333}
2301 2334
@@ -2320,7 +2353,7 @@ static int do_proc_dointvec_conv(int *negp, unsigned long *lvalp,
2320} 2353}
2321 2354
2322static int __do_proc_dointvec(void *tbl_data, struct ctl_table *table, 2355static int __do_proc_dointvec(void *tbl_data, struct ctl_table *table,
2323 int write, struct file *filp, void __user *buffer, 2356 int write, void __user *buffer,
2324 size_t *lenp, loff_t *ppos, 2357 size_t *lenp, loff_t *ppos,
2325 int (*conv)(int *negp, unsigned long *lvalp, int *valp, 2358 int (*conv)(int *negp, unsigned long *lvalp, int *valp,
2326 int write, void *data), 2359 int write, void *data),
@@ -2427,13 +2460,13 @@ static int __do_proc_dointvec(void *tbl_data, struct ctl_table *table,
2427#undef TMPBUFLEN 2460#undef TMPBUFLEN
2428} 2461}
2429 2462
2430static int do_proc_dointvec(struct ctl_table *table, int write, struct file *filp, 2463static int do_proc_dointvec(struct ctl_table *table, int write,
2431 void __user *buffer, size_t *lenp, loff_t *ppos, 2464 void __user *buffer, size_t *lenp, loff_t *ppos,
2432 int (*conv)(int *negp, unsigned long *lvalp, int *valp, 2465 int (*conv)(int *negp, unsigned long *lvalp, int *valp,
2433 int write, void *data), 2466 int write, void *data),
2434 void *data) 2467 void *data)
2435{ 2468{
2436 return __do_proc_dointvec(table->data, table, write, filp, 2469 return __do_proc_dointvec(table->data, table, write,
2437 buffer, lenp, ppos, conv, data); 2470 buffer, lenp, ppos, conv, data);
2438} 2471}
2439 2472
@@ -2441,7 +2474,6 @@ static int do_proc_dointvec(struct ctl_table *table, int write, struct file *fil
2441 * proc_dointvec - read a vector of integers 2474 * proc_dointvec - read a vector of integers
2442 * @table: the sysctl table 2475 * @table: the sysctl table
2443 * @write: %TRUE if this is a write to the sysctl file 2476 * @write: %TRUE if this is a write to the sysctl file
2444 * @filp: the file structure
2445 * @buffer: the user buffer 2477 * @buffer: the user buffer
2446 * @lenp: the size of the user buffer 2478 * @lenp: the size of the user buffer
2447 * @ppos: file position 2479 * @ppos: file position
@@ -2451,10 +2483,10 @@ static int do_proc_dointvec(struct ctl_table *table, int write, struct file *fil
2451 * 2483 *
2452 * Returns 0 on success. 2484 * Returns 0 on success.
2453 */ 2485 */
2454int proc_dointvec(struct ctl_table *table, int write, struct file *filp, 2486int proc_dointvec(struct ctl_table *table, int write,
2455 void __user *buffer, size_t *lenp, loff_t *ppos) 2487 void __user *buffer, size_t *lenp, loff_t *ppos)
2456{ 2488{
2457 return do_proc_dointvec(table,write,filp,buffer,lenp,ppos, 2489 return do_proc_dointvec(table,write,buffer,lenp,ppos,
2458 NULL,NULL); 2490 NULL,NULL);
2459} 2491}
2460 2492
@@ -2462,7 +2494,7 @@ int proc_dointvec(struct ctl_table *table, int write, struct file *filp,
2462 * Taint values can only be increased 2494 * Taint values can only be increased
2463 * This means we can safely use a temporary. 2495 * This means we can safely use a temporary.
2464 */ 2496 */
2465static int proc_taint(struct ctl_table *table, int write, struct file *filp, 2497static int proc_taint(struct ctl_table *table, int write,
2466 void __user *buffer, size_t *lenp, loff_t *ppos) 2498 void __user *buffer, size_t *lenp, loff_t *ppos)
2467{ 2499{
2468 struct ctl_table t; 2500 struct ctl_table t;
@@ -2474,7 +2506,7 @@ static int proc_taint(struct ctl_table *table, int write, struct file *filp,
2474 2506
2475 t = *table; 2507 t = *table;
2476 t.data = &tmptaint; 2508 t.data = &tmptaint;
2477 err = proc_doulongvec_minmax(&t, write, filp, buffer, lenp, ppos); 2509 err = proc_doulongvec_minmax(&t, write, buffer, lenp, ppos);
2478 if (err < 0) 2510 if (err < 0)
2479 return err; 2511 return err;
2480 2512
@@ -2526,7 +2558,6 @@ static int do_proc_dointvec_minmax_conv(int *negp, unsigned long *lvalp,
2526 * proc_dointvec_minmax - read a vector of integers with min/max values 2558 * proc_dointvec_minmax - read a vector of integers with min/max values
2527 * @table: the sysctl table 2559 * @table: the sysctl table
2528 * @write: %TRUE if this is a write to the sysctl file 2560 * @write: %TRUE if this is a write to the sysctl file
2529 * @filp: the file structure
2530 * @buffer: the user buffer 2561 * @buffer: the user buffer
2531 * @lenp: the size of the user buffer 2562 * @lenp: the size of the user buffer
2532 * @ppos: file position 2563 * @ppos: file position
@@ -2539,19 +2570,18 @@ static int do_proc_dointvec_minmax_conv(int *negp, unsigned long *lvalp,
2539 * 2570 *
2540 * Returns 0 on success. 2571 * Returns 0 on success.
2541 */ 2572 */
2542int proc_dointvec_minmax(struct ctl_table *table, int write, struct file *filp, 2573int proc_dointvec_minmax(struct ctl_table *table, int write,
2543 void __user *buffer, size_t *lenp, loff_t *ppos) 2574 void __user *buffer, size_t *lenp, loff_t *ppos)
2544{ 2575{
2545 struct do_proc_dointvec_minmax_conv_param param = { 2576 struct do_proc_dointvec_minmax_conv_param param = {
2546 .min = (int *) table->extra1, 2577 .min = (int *) table->extra1,
2547 .max = (int *) table->extra2, 2578 .max = (int *) table->extra2,
2548 }; 2579 };
2549 return do_proc_dointvec(table, write, filp, buffer, lenp, ppos, 2580 return do_proc_dointvec(table, write, buffer, lenp, ppos,
2550 do_proc_dointvec_minmax_conv, &param); 2581 do_proc_dointvec_minmax_conv, &param);
2551} 2582}
2552 2583
2553static int __do_proc_doulongvec_minmax(void *data, struct ctl_table *table, int write, 2584static int __do_proc_doulongvec_minmax(void *data, struct ctl_table *table, int write,
2554 struct file *filp,
2555 void __user *buffer, 2585 void __user *buffer,
2556 size_t *lenp, loff_t *ppos, 2586 size_t *lenp, loff_t *ppos,
2557 unsigned long convmul, 2587 unsigned long convmul,
@@ -2656,21 +2686,19 @@ static int __do_proc_doulongvec_minmax(void *data, struct ctl_table *table, int
2656} 2686}
2657 2687
2658static int do_proc_doulongvec_minmax(struct ctl_table *table, int write, 2688static int do_proc_doulongvec_minmax(struct ctl_table *table, int write,
2659 struct file *filp,
2660 void __user *buffer, 2689 void __user *buffer,
2661 size_t *lenp, loff_t *ppos, 2690 size_t *lenp, loff_t *ppos,
2662 unsigned long convmul, 2691 unsigned long convmul,
2663 unsigned long convdiv) 2692 unsigned long convdiv)
2664{ 2693{
2665 return __do_proc_doulongvec_minmax(table->data, table, write, 2694 return __do_proc_doulongvec_minmax(table->data, table, write,
2666 filp, buffer, lenp, ppos, convmul, convdiv); 2695 buffer, lenp, ppos, convmul, convdiv);
2667} 2696}
2668 2697
2669/** 2698/**
2670 * proc_doulongvec_minmax - read a vector of long integers with min/max values 2699 * proc_doulongvec_minmax - read a vector of long integers with min/max values
2671 * @table: the sysctl table 2700 * @table: the sysctl table
2672 * @write: %TRUE if this is a write to the sysctl file 2701 * @write: %TRUE if this is a write to the sysctl file
2673 * @filp: the file structure
2674 * @buffer: the user buffer 2702 * @buffer: the user buffer
2675 * @lenp: the size of the user buffer 2703 * @lenp: the size of the user buffer
2676 * @ppos: file position 2704 * @ppos: file position
@@ -2683,17 +2711,16 @@ static int do_proc_doulongvec_minmax(struct ctl_table *table, int write,
2683 * 2711 *
2684 * Returns 0 on success. 2712 * Returns 0 on success.
2685 */ 2713 */
2686int proc_doulongvec_minmax(struct ctl_table *table, int write, struct file *filp, 2714int proc_doulongvec_minmax(struct ctl_table *table, int write,
2687 void __user *buffer, size_t *lenp, loff_t *ppos) 2715 void __user *buffer, size_t *lenp, loff_t *ppos)
2688{ 2716{
2689 return do_proc_doulongvec_minmax(table, write, filp, buffer, lenp, ppos, 1l, 1l); 2717 return do_proc_doulongvec_minmax(table, write, buffer, lenp, ppos, 1l, 1l);
2690} 2718}
2691 2719
2692/** 2720/**
2693 * proc_doulongvec_ms_jiffies_minmax - read a vector of millisecond values with min/max values 2721 * proc_doulongvec_ms_jiffies_minmax - read a vector of millisecond values with min/max values
2694 * @table: the sysctl table 2722 * @table: the sysctl table
2695 * @write: %TRUE if this is a write to the sysctl file 2723 * @write: %TRUE if this is a write to the sysctl file
2696 * @filp: the file structure
2697 * @buffer: the user buffer 2724 * @buffer: the user buffer
2698 * @lenp: the size of the user buffer 2725 * @lenp: the size of the user buffer
2699 * @ppos: file position 2726 * @ppos: file position
@@ -2708,11 +2735,10 @@ int proc_doulongvec_minmax(struct ctl_table *table, int write, struct file *filp
2708 * Returns 0 on success. 2735 * Returns 0 on success.
2709 */ 2736 */
2710int proc_doulongvec_ms_jiffies_minmax(struct ctl_table *table, int write, 2737int proc_doulongvec_ms_jiffies_minmax(struct ctl_table *table, int write,
2711 struct file *filp,
2712 void __user *buffer, 2738 void __user *buffer,
2713 size_t *lenp, loff_t *ppos) 2739 size_t *lenp, loff_t *ppos)
2714{ 2740{
2715 return do_proc_doulongvec_minmax(table, write, filp, buffer, 2741 return do_proc_doulongvec_minmax(table, write, buffer,
2716 lenp, ppos, HZ, 1000l); 2742 lenp, ppos, HZ, 1000l);
2717} 2743}
2718 2744
@@ -2788,7 +2814,6 @@ static int do_proc_dointvec_ms_jiffies_conv(int *negp, unsigned long *lvalp,
2788 * proc_dointvec_jiffies - read a vector of integers as seconds 2814 * proc_dointvec_jiffies - read a vector of integers as seconds
2789 * @table: the sysctl table 2815 * @table: the sysctl table
2790 * @write: %TRUE if this is a write to the sysctl file 2816 * @write: %TRUE if this is a write to the sysctl file
2791 * @filp: the file structure
2792 * @buffer: the user buffer 2817 * @buffer: the user buffer
2793 * @lenp: the size of the user buffer 2818 * @lenp: the size of the user buffer
2794 * @ppos: file position 2819 * @ppos: file position
@@ -2800,10 +2825,10 @@ static int do_proc_dointvec_ms_jiffies_conv(int *negp, unsigned long *lvalp,
2800 * 2825 *
2801 * Returns 0 on success. 2826 * Returns 0 on success.
2802 */ 2827 */
2803int proc_dointvec_jiffies(struct ctl_table *table, int write, struct file *filp, 2828int proc_dointvec_jiffies(struct ctl_table *table, int write,
2804 void __user *buffer, size_t *lenp, loff_t *ppos) 2829 void __user *buffer, size_t *lenp, loff_t *ppos)
2805{ 2830{
2806 return do_proc_dointvec(table,write,filp,buffer,lenp,ppos, 2831 return do_proc_dointvec(table,write,buffer,lenp,ppos,
2807 do_proc_dointvec_jiffies_conv,NULL); 2832 do_proc_dointvec_jiffies_conv,NULL);
2808} 2833}
2809 2834
@@ -2811,7 +2836,6 @@ int proc_dointvec_jiffies(struct ctl_table *table, int write, struct file *filp,
2811 * proc_dointvec_userhz_jiffies - read a vector of integers as 1/USER_HZ seconds 2836 * proc_dointvec_userhz_jiffies - read a vector of integers as 1/USER_HZ seconds
2812 * @table: the sysctl table 2837 * @table: the sysctl table
2813 * @write: %TRUE if this is a write to the sysctl file 2838 * @write: %TRUE if this is a write to the sysctl file
2814 * @filp: the file structure
2815 * @buffer: the user buffer 2839 * @buffer: the user buffer
2816 * @lenp: the size of the user buffer 2840 * @lenp: the size of the user buffer
2817 * @ppos: pointer to the file position 2841 * @ppos: pointer to the file position
@@ -2823,10 +2847,10 @@ int proc_dointvec_jiffies(struct ctl_table *table, int write, struct file *filp,
2823 * 2847 *
2824 * Returns 0 on success. 2848 * Returns 0 on success.
2825 */ 2849 */
2826int proc_dointvec_userhz_jiffies(struct ctl_table *table, int write, struct file *filp, 2850int proc_dointvec_userhz_jiffies(struct ctl_table *table, int write,
2827 void __user *buffer, size_t *lenp, loff_t *ppos) 2851 void __user *buffer, size_t *lenp, loff_t *ppos)
2828{ 2852{
2829 return do_proc_dointvec(table,write,filp,buffer,lenp,ppos, 2853 return do_proc_dointvec(table,write,buffer,lenp,ppos,
2830 do_proc_dointvec_userhz_jiffies_conv,NULL); 2854 do_proc_dointvec_userhz_jiffies_conv,NULL);
2831} 2855}
2832 2856
@@ -2834,7 +2858,6 @@ int proc_dointvec_userhz_jiffies(struct ctl_table *table, int write, struct file
2834 * proc_dointvec_ms_jiffies - read a vector of integers as 1 milliseconds 2858 * proc_dointvec_ms_jiffies - read a vector of integers as 1 milliseconds
2835 * @table: the sysctl table 2859 * @table: the sysctl table
2836 * @write: %TRUE if this is a write to the sysctl file 2860 * @write: %TRUE if this is a write to the sysctl file
2837 * @filp: the file structure
2838 * @buffer: the user buffer 2861 * @buffer: the user buffer
2839 * @lenp: the size of the user buffer 2862 * @lenp: the size of the user buffer
2840 * @ppos: file position 2863 * @ppos: file position
@@ -2847,14 +2870,14 @@ int proc_dointvec_userhz_jiffies(struct ctl_table *table, int write, struct file
2847 * 2870 *
2848 * Returns 0 on success. 2871 * Returns 0 on success.
2849 */ 2872 */
2850int proc_dointvec_ms_jiffies(struct ctl_table *table, int write, struct file *filp, 2873int proc_dointvec_ms_jiffies(struct ctl_table *table, int write,
2851 void __user *buffer, size_t *lenp, loff_t *ppos) 2874 void __user *buffer, size_t *lenp, loff_t *ppos)
2852{ 2875{
2853 return do_proc_dointvec(table, write, filp, buffer, lenp, ppos, 2876 return do_proc_dointvec(table, write, buffer, lenp, ppos,
2854 do_proc_dointvec_ms_jiffies_conv, NULL); 2877 do_proc_dointvec_ms_jiffies_conv, NULL);
2855} 2878}
2856 2879
2857static int proc_do_cad_pid(struct ctl_table *table, int write, struct file *filp, 2880static int proc_do_cad_pid(struct ctl_table *table, int write,
2858 void __user *buffer, size_t *lenp, loff_t *ppos) 2881 void __user *buffer, size_t *lenp, loff_t *ppos)
2859{ 2882{
2860 struct pid *new_pid; 2883 struct pid *new_pid;
@@ -2863,7 +2886,7 @@ static int proc_do_cad_pid(struct ctl_table *table, int write, struct file *filp
2863 2886
2864 tmp = pid_vnr(cad_pid); 2887 tmp = pid_vnr(cad_pid);
2865 2888
2866 r = __do_proc_dointvec(&tmp, table, write, filp, buffer, 2889 r = __do_proc_dointvec(&tmp, table, write, buffer,
2867 lenp, ppos, NULL, NULL); 2890 lenp, ppos, NULL, NULL);
2868 if (r || !write) 2891 if (r || !write)
2869 return r; 2892 return r;
@@ -2878,50 +2901,49 @@ static int proc_do_cad_pid(struct ctl_table *table, int write, struct file *filp
2878 2901
2879#else /* CONFIG_PROC_FS */ 2902#else /* CONFIG_PROC_FS */
2880 2903
2881int proc_dostring(struct ctl_table *table, int write, struct file *filp, 2904int proc_dostring(struct ctl_table *table, int write,
2882 void __user *buffer, size_t *lenp, loff_t *ppos) 2905 void __user *buffer, size_t *lenp, loff_t *ppos)
2883{ 2906{
2884 return -ENOSYS; 2907 return -ENOSYS;
2885} 2908}
2886 2909
2887int proc_dointvec(struct ctl_table *table, int write, struct file *filp, 2910int proc_dointvec(struct ctl_table *table, int write,
2888 void __user *buffer, size_t *lenp, loff_t *ppos) 2911 void __user *buffer, size_t *lenp, loff_t *ppos)
2889{ 2912{
2890 return -ENOSYS; 2913 return -ENOSYS;
2891} 2914}
2892 2915
2893int proc_dointvec_minmax(struct ctl_table *table, int write, struct file *filp, 2916int proc_dointvec_minmax(struct ctl_table *table, int write,
2894 void __user *buffer, size_t *lenp, loff_t *ppos) 2917 void __user *buffer, size_t *lenp, loff_t *ppos)
2895{ 2918{
2896 return -ENOSYS; 2919 return -ENOSYS;
2897} 2920}
2898 2921
2899int proc_dointvec_jiffies(struct ctl_table *table, int write, struct file *filp, 2922int proc_dointvec_jiffies(struct ctl_table *table, int write,
2900 void __user *buffer, size_t *lenp, loff_t *ppos) 2923 void __user *buffer, size_t *lenp, loff_t *ppos)
2901{ 2924{
2902 return -ENOSYS; 2925 return -ENOSYS;
2903} 2926}
2904 2927
2905int proc_dointvec_userhz_jiffies(struct ctl_table *table, int write, struct file *filp, 2928int proc_dointvec_userhz_jiffies(struct ctl_table *table, int write,
2906 void __user *buffer, size_t *lenp, loff_t *ppos) 2929 void __user *buffer, size_t *lenp, loff_t *ppos)
2907{ 2930{
2908 return -ENOSYS; 2931 return -ENOSYS;
2909} 2932}
2910 2933
2911int proc_dointvec_ms_jiffies(struct ctl_table *table, int write, struct file *filp, 2934int proc_dointvec_ms_jiffies(struct ctl_table *table, int write,
2912 void __user *buffer, size_t *lenp, loff_t *ppos) 2935 void __user *buffer, size_t *lenp, loff_t *ppos)
2913{ 2936{
2914 return -ENOSYS; 2937 return -ENOSYS;
2915} 2938}
2916 2939
2917int proc_doulongvec_minmax(struct ctl_table *table, int write, struct file *filp, 2940int proc_doulongvec_minmax(struct ctl_table *table, int write,
2918 void __user *buffer, size_t *lenp, loff_t *ppos) 2941 void __user *buffer, size_t *lenp, loff_t *ppos)
2919{ 2942{
2920 return -ENOSYS; 2943 return -ENOSYS;
2921} 2944}
2922 2945
2923int proc_doulongvec_ms_jiffies_minmax(struct ctl_table *table, int write, 2946int proc_doulongvec_ms_jiffies_minmax(struct ctl_table *table, int write,
2924 struct file *filp,
2925 void __user *buffer, 2947 void __user *buffer,
2926 size_t *lenp, loff_t *ppos) 2948 size_t *lenp, loff_t *ppos)
2927{ 2949{
diff --git a/kernel/time/Makefile b/kernel/time/Makefile
index 0b0a6366c9d4..ee266620b06c 100644
--- a/kernel/time/Makefile
+++ b/kernel/time/Makefile
@@ -1,4 +1,4 @@
1obj-y += timekeeping.o ntp.o clocksource.o jiffies.o timer_list.o timecompare.o 1obj-y += timekeeping.o ntp.o clocksource.o jiffies.o timer_list.o timecompare.o timeconv.o
2 2
3obj-$(CONFIG_GENERIC_CLOCKEVENTS_BUILD) += clockevents.o 3obj-$(CONFIG_GENERIC_CLOCKEVENTS_BUILD) += clockevents.o
4obj-$(CONFIG_GENERIC_CLOCKEVENTS) += tick-common.o 4obj-$(CONFIG_GENERIC_CLOCKEVENTS) += tick-common.o
diff --git a/kernel/time/timeconv.c b/kernel/time/timeconv.c
new file mode 100644
index 000000000000..86628e755f38
--- /dev/null
+++ b/kernel/time/timeconv.c
@@ -0,0 +1,127 @@
1/*
2 * Copyright (C) 1993, 1994, 1995, 1996, 1997 Free Software Foundation, Inc.
3 * This file is part of the GNU C Library.
4 * Contributed by Paul Eggert (eggert@twinsun.com).
5 *
6 * The GNU C Library is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Library General Public License as
8 * published by the Free Software Foundation; either version 2 of the
9 * License, or (at your option) any later version.
10 *
11 * The GNU C Library is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Library General Public License for more details.
15 *
16 * You should have received a copy of the GNU Library General Public
17 * License along with the GNU C Library; see the file COPYING.LIB. If not,
18 * write to the Free Software Foundation, Inc., 59 Temple Place - Suite 330,
19 * Boston, MA 02111-1307, USA.
20 */
21
22/*
23 * Converts the calendar time to broken-down time representation
24 * Based on code from glibc-2.6
25 *
26 * 2009-7-14:
27 * Moved from glibc-2.6 to kernel by Zhaolei<zhaolei@cn.fujitsu.com>
28 */
29
30#include <linux/time.h>
31#include <linux/module.h>
32
33/*
34 * Nonzero if YEAR is a leap year (every 4 years,
35 * except every 100th isn't, and every 400th is).
36 */
37static int __isleap(long year)
38{
39 return (year) % 4 == 0 && ((year) % 100 != 0 || (year) % 400 == 0);
40}
41
42/* do a mathdiv for long type */
43static long math_div(long a, long b)
44{
45 return a / b - (a % b < 0);
46}
47
48/* How many leap years between y1 and y2, y1 must less or equal to y2 */
49static long leaps_between(long y1, long y2)
50{
51 long leaps1 = math_div(y1 - 1, 4) - math_div(y1 - 1, 100)
52 + math_div(y1 - 1, 400);
53 long leaps2 = math_div(y2 - 1, 4) - math_div(y2 - 1, 100)
54 + math_div(y2 - 1, 400);
55 return leaps2 - leaps1;
56}
57
58/* How many days come before each month (0-12). */
59static const unsigned short __mon_yday[2][13] = {
60 /* Normal years. */
61 {0, 31, 59, 90, 120, 151, 181, 212, 243, 273, 304, 334, 365},
62 /* Leap years. */
63 {0, 31, 60, 91, 121, 152, 182, 213, 244, 274, 305, 335, 366}
64};
65
66#define SECS_PER_HOUR (60 * 60)
67#define SECS_PER_DAY (SECS_PER_HOUR * 24)
68
69/**
70 * time_to_tm - converts the calendar time to local broken-down time
71 *
72 * @totalsecs the number of seconds elapsed since 00:00:00 on January 1, 1970,
73 * Coordinated Universal Time (UTC).
74 * @offset offset seconds adding to totalsecs.
75 * @result pointer to struct tm variable to receive broken-down time
76 */
77void time_to_tm(time_t totalsecs, int offset, struct tm *result)
78{
79 long days, rem, y;
80 const unsigned short *ip;
81
82 days = totalsecs / SECS_PER_DAY;
83 rem = totalsecs % SECS_PER_DAY;
84 rem += offset;
85 while (rem < 0) {
86 rem += SECS_PER_DAY;
87 --days;
88 }
89 while (rem >= SECS_PER_DAY) {
90 rem -= SECS_PER_DAY;
91 ++days;
92 }
93
94 result->tm_hour = rem / SECS_PER_HOUR;
95 rem %= SECS_PER_HOUR;
96 result->tm_min = rem / 60;
97 result->tm_sec = rem % 60;
98
99 /* January 1, 1970 was a Thursday. */
100 result->tm_wday = (4 + days) % 7;
101 if (result->tm_wday < 0)
102 result->tm_wday += 7;
103
104 y = 1970;
105
106 while (days < 0 || days >= (__isleap(y) ? 366 : 365)) {
107 /* Guess a corrected year, assuming 365 days per year. */
108 long yg = y + math_div(days, 365);
109
110 /* Adjust DAYS and Y to match the guessed year. */
111 days -= (yg - y) * 365 + leaps_between(y, yg);
112 y = yg;
113 }
114
115 result->tm_year = y - 1900;
116
117 result->tm_yday = days;
118
119 ip = __mon_yday[__isleap(y)];
120 for (y = 11; days < ip[y]; y--)
121 continue;
122 days -= ip[y];
123
124 result->tm_mon = y;
125 result->tm_mday = days + 1;
126}
127EXPORT_SYMBOL(time_to_tm);
diff --git a/kernel/trace/ftrace.c b/kernel/trace/ftrace.c
index 23df7771c937..a142579765bf 100644
--- a/kernel/trace/ftrace.c
+++ b/kernel/trace/ftrace.c
@@ -3015,7 +3015,7 @@ int unregister_ftrace_function(struct ftrace_ops *ops)
3015 3015
3016int 3016int
3017ftrace_enable_sysctl(struct ctl_table *table, int write, 3017ftrace_enable_sysctl(struct ctl_table *table, int write,
3018 struct file *file, void __user *buffer, size_t *lenp, 3018 void __user *buffer, size_t *lenp,
3019 loff_t *ppos) 3019 loff_t *ppos)
3020{ 3020{
3021 int ret; 3021 int ret;
@@ -3025,7 +3025,7 @@ ftrace_enable_sysctl(struct ctl_table *table, int write,
3025 3025
3026 mutex_lock(&ftrace_lock); 3026 mutex_lock(&ftrace_lock);
3027 3027
3028 ret = proc_dointvec(table, write, file, buffer, lenp, ppos); 3028 ret = proc_dointvec(table, write, buffer, lenp, ppos);
3029 3029
3030 if (ret || !write || (last_ftrace_enabled == !!ftrace_enabled)) 3030 if (ret || !write || (last_ftrace_enabled == !!ftrace_enabled))
3031 goto out; 3031 goto out;
diff --git a/kernel/trace/trace_stack.c b/kernel/trace/trace_stack.c
index 0f6facb050a1..8504ac71e4e8 100644
--- a/kernel/trace/trace_stack.c
+++ b/kernel/trace/trace_stack.c
@@ -296,14 +296,14 @@ static const struct file_operations stack_trace_fops = {
296 296
297int 297int
298stack_trace_sysctl(struct ctl_table *table, int write, 298stack_trace_sysctl(struct ctl_table *table, int write,
299 struct file *file, void __user *buffer, size_t *lenp, 299 void __user *buffer, size_t *lenp,
300 loff_t *ppos) 300 loff_t *ppos)
301{ 301{
302 int ret; 302 int ret;
303 303
304 mutex_lock(&stack_sysctl_mutex); 304 mutex_lock(&stack_sysctl_mutex);
305 305
306 ret = proc_dointvec(table, write, file, buffer, lenp, ppos); 306 ret = proc_dointvec(table, write, buffer, lenp, ppos);
307 307
308 if (ret || !write || 308 if (ret || !write ||
309 (last_stack_tracer_enabled == !!stack_tracer_enabled)) 309 (last_stack_tracer_enabled == !!stack_tracer_enabled))
diff --git a/kernel/utsname_sysctl.c b/kernel/utsname_sysctl.c
index 92359cc747a7..69eae358a726 100644
--- a/kernel/utsname_sysctl.c
+++ b/kernel/utsname_sysctl.c
@@ -42,14 +42,14 @@ static void put_uts(ctl_table *table, int write, void *which)
42 * Special case of dostring for the UTS structure. This has locks 42 * Special case of dostring for the UTS structure. This has locks
43 * to observe. Should this be in kernel/sys.c ???? 43 * to observe. Should this be in kernel/sys.c ????
44 */ 44 */
45static int proc_do_uts_string(ctl_table *table, int write, struct file *filp, 45static int proc_do_uts_string(ctl_table *table, int write,
46 void __user *buffer, size_t *lenp, loff_t *ppos) 46 void __user *buffer, size_t *lenp, loff_t *ppos)
47{ 47{
48 struct ctl_table uts_table; 48 struct ctl_table uts_table;
49 int r; 49 int r;
50 memcpy(&uts_table, table, sizeof(uts_table)); 50 memcpy(&uts_table, table, sizeof(uts_table));
51 uts_table.data = get_uts(table, write); 51 uts_table.data = get_uts(table, write);
52 r = proc_dostring(&uts_table,write,filp,buffer,lenp, ppos); 52 r = proc_dostring(&uts_table,write,buffer,lenp, ppos);
53 put_uts(table, write, uts_table.data); 53 put_uts(table, write, uts_table.data);
54 return r; 54 return r;
55} 55}
generated by cgit v1.2.2 (git 2.25.0) at 2025-12-28 02:27:51 -0500