History

Merge 5.6-rc7 into char-misc-next

We need the char/misc driver fixes in here as well.

Greg Kroah-Hartman committed 6 years ago

Commit baca54d2 parents 217fb36 16fbf79414 files · +3487 −1503Browse files

Showing the changes to kernel (8 of 414 files). Show all

Unified · Split

8 files
  1. kernel/cgroup/cgroup-v1.c +2 −1
  2. kernel/cgroup/cgroup.c +30 −13
  3. kernel/futex.c +55 −38
  4. kernel/notifier.c +1 −1
  5. kernel/pid.c +10 −0
  6. kernel/sys.c +2 −0
  7. kernel/trace/ftrace.c +2 −0
  8. kernel/workqueue.c +8 −6
kernel/cgroup/cgroup-v1.c+2−1
@@ -471,6 +471,7 @@ static void *cgroup_pidlist_next(struct seq_file *s, void *v, loff_t *pos) · 470 unmodified lines
471471 */
472472 p++;
473473 if (p >= end) {
474+ (*pos)++;
474475 return NULL;
475476 } else {
476477 *pos = *p;
@@ -782,7 +783,7 @@ void cgroup1_release_agent(struct work_struct *work) · 305 unmodified lines
782783
783784 pathbuf = kmalloc(PATH_MAX, GFP_KERNEL);
784785 agentbuf = kstrdup(cgrp->root->release_agent_path, GFP_KERNEL);
785− if (!pathbuf || !agentbuf)
786+ if (!pathbuf || !agentbuf || !strlen(agentbuf))
786787 goto out;
787788
788789 spin_lock_irq(&css_set_lock);
kernel/cgroup/cgroup.c+30−13
@@ -3542,21 +3542,21 @@ static int cpu_stat_show(struct seq_file *seq, void *v) · 3541 unmodified lines
35423542 static int cgroup_io_pressure_show(struct seq_file *seq, void *v)
35433543 {
35443544 struct cgroup *cgrp = seq_css(seq)->cgroup;
3545− struct psi_group *psi = cgroup_id(cgrp) == 1 ? &psi_system : &cgrp->psi;
3545+ struct psi_group *psi = cgroup_ino(cgrp) == 1 ? &psi_system : &cgrp->psi;
35463546
35473547 return psi_show(seq, psi, PSI_IO);
35483548 }
35493549 static int cgroup_memory_pressure_show(struct seq_file *seq, void *v)
35503550 {
35513551 struct cgroup *cgrp = seq_css(seq)->cgroup;
3552− struct psi_group *psi = cgroup_id(cgrp) == 1 ? &psi_system : &cgrp->psi;
3552+ struct psi_group *psi = cgroup_ino(cgrp) == 1 ? &psi_system : &cgrp->psi;
35533553
35543554 return psi_show(seq, psi, PSI_MEM);
35553555 }
35563556 static int cgroup_cpu_pressure_show(struct seq_file *seq, void *v)
35573557 {
35583558 struct cgroup *cgrp = seq_css(seq)->cgroup;
3559− struct psi_group *psi = cgroup_id(cgrp) == 1 ? &psi_system : &cgrp->psi;
3559+ struct psi_group *psi = cgroup_ino(cgrp) == 1 ? &psi_system : &cgrp->psi;
35603560
35613561 return psi_show(seq, psi, PSI_CPU);
35623562 }
@@ -4400,12 +4400,16 @@ static void css_task_iter_advance_css_set(struct css_task_iter *it) · 837 unmodified lines
44004400 }
44014401 } while (!css_set_populated(cset) && list_empty(&cset->dying_tasks));
44024402
4403− if (!list_empty(&cset->tasks))
4403+ if (!list_empty(&cset->tasks)) {
44044404 it->task_pos = cset->tasks.next;
4405− else if (!list_empty(&cset->mg_tasks))
4405+ it->cur_tasks_head = &cset->tasks;
4406+ } else if (!list_empty(&cset->mg_tasks)) {
44064407 it->task_pos = cset->mg_tasks.next;
4407− else
4408+ it->cur_tasks_head = &cset->mg_tasks;
4409+ } else {
44084410 it->task_pos = cset->dying_tasks.next;
4411+ it->cur_tasks_head = &cset->dying_tasks;
4412+ }
44094413
44104414 it->tasks_head = &cset->tasks;
44114415 it->mg_tasks_head = &cset->mg_tasks;
@@ -4463,10 +4467,14 @@ repeat: · 51 unmodified lines
44634467 else
44644468 it->task_pos = it->task_pos->next;
44654469
4466− if (it->task_pos == it->tasks_head)
4470+ if (it->task_pos == it->tasks_head) {
44674471 it->task_pos = it->mg_tasks_head->next;
4468− if (it->task_pos == it->mg_tasks_head)
4472+ it->cur_tasks_head = it->mg_tasks_head;
4473+ }
4474+ if (it->task_pos == it->mg_tasks_head) {
44694475 it->task_pos = it->dying_tasks_head->next;
4476+ it->cur_tasks_head = it->dying_tasks_head;
4477+ }
44704478 if (it->task_pos == it->dying_tasks_head)
44714479 css_task_iter_advance_css_set(it);
44724480 } else {
@@ -4485,11 +4493,12 @@ repeat: · 12 unmodified lines
44854493 goto repeat;
44864494
44874495 /* and dying leaders w/o live member threads */
4488− if (!atomic_read(&task->signal->live))
4496+ if (it->cur_tasks_head == it->dying_tasks_head &&
4497+ !atomic_read(&task->signal->live))
44894498 goto repeat;
44904499 } else {
44914500 /* skip all dying ones */
4492− if (task->flags & PF_EXITING)
4501+ if (it->cur_tasks_head == it->dying_tasks_head)
44934502 goto repeat;
44944503 }
44954504 }
@@ -4595,6 +4604,9 @@ static void *cgroup_procs_next(struct seq_file *s, void *v, loff_t *pos) · 99 unmodified lines
45954604 struct kernfs_open_file *of = s->private;
45964605 struct css_task_iter *it = of->priv;
45974606
4607+ if (pos)
4608+ (*pos)++;
4609+
45984610 return css_task_iter_next(it);
45994611 }
46004612
@@ -4610,7 +4622,7 @@ static void *__cgroup_procs_start(struct seq_file *s, loff_t *pos, · 9 unmodified lines
46104622 * from position 0, so we can simply keep iterating on !0 *pos.
46114623 */
46124624 if (!it) {
4613− if (WARN_ON_ONCE((*pos)++))
4625+ if (WARN_ON_ONCE((*pos)))
46144626 return ERR_PTR(-EINVAL);
46154627
46164628 it = kzalloc(sizeof(*it), GFP_KERNEL);
@@ -4618,10 +4630,11 @@ static void *__cgroup_procs_start(struct seq_file *s, loff_t *pos, · 1 unmodified line
46184630 return ERR_PTR(-ENOMEM);
46194631 of->priv = it;
46204632 css_task_iter_start(&cgrp->self, iter_flags, it);
4621− } else if (!(*pos)++) {
4633+ } else if (!(*pos)) {
46224634 css_task_iter_end(it);
46234635 css_task_iter_start(&cgrp->self, iter_flags, it);
4624− }
4636+ } else
4637+ return it->cur_task;
46254638
46264639 return cgroup_procs_next(s, NULL, NULL);
46274640 }
@@ -6258,6 +6271,10 @@ void cgroup_sk_alloc(struct sock_cgroup_data *skcd) · 1630 unmodified lines
62586271 return;
62596272 }
62606273
6274+ /* Don't associate the sock with unrelated interrupted task's cgroup. */
6275+ if (in_interrupt())
6276+ return;
6277+
62616278 rcu_read_lock();
62626279
62636280 while (true) {
kernel/futex.c+55−38
@@ -385,9 +385,9 @@ static inline int hb_waiters_pending(struct futex_hash_bucket *hb) · 384 unmodified lines
385385 */
386386 static struct futex_hash_bucket *hash_futex(union futex_key *key)
387387 {
388− u32 hash = jhash2((u32*)&key->both.word,
389− (sizeof(key->both.word)+sizeof(key->both.ptr))/4,
388+ u32 hash = jhash2((u32 *)key, offsetof(typeof(*key), both.offset) / 4,
390389 key->both.offset);
390+
391391 return &futex_queues[hash & (futex_hashsize - 1)];
392392 }
393393
@@ -429,7 +429,7 @@ static void get_futex_key_refs(union futex_key *key) · 35 unmodified lines
429429
430430 switch (key->both.offset & (FUT_OFF_INODE|FUT_OFF_MMSHARED)) {
431431 case FUT_OFF_INODE:
432− ihold(key->shared.inode); /* implies smp_mb(); (B) */
432+ smp_mb(); /* explicit smp_mb(); (B) */
433433 break;
434434 case FUT_OFF_MMSHARED:
435435 futex_get_mm(key); /* implies smp_mb(); (B) */
@@ -463,7 +463,6 @@ static void drop_futex_key_refs(union futex_key *key) · 27 unmodified lines
463463
464464 switch (key->both.offset & (FUT_OFF_INODE|FUT_OFF_MMSHARED)) {
465465 case FUT_OFF_INODE:
466− iput(key->shared.inode);
467466 break;
468467 case FUT_OFF_MMSHARED:
469468 mmdrop(key->private.mm);
@@ -505,6 +504,46 @@ futex_setup_timer(ktime_t *time, struct hrtimer_sleeper *timeout, · 35 unmodified lines
505504 return timeout;
506505 }
507506
507+/*
508+ * Generate a machine wide unique identifier for this inode.
509+ *
510+ * This relies on u64 not wrapping in the life-time of the machine; which with
511+ * 1ns resolution means almost 585 years.
512+ *
513+ * This further relies on the fact that a well formed program will not unmap
514+ * the file while it has a (shared) futex waiting on it. This mapping will have
515+ * a file reference which pins the mount and inode.
516+ *
517+ * If for some reason an inode gets evicted and read back in again, it will get
518+ * a new sequence number and will _NOT_ match, even though it is the exact same
519+ * file.
520+ *
521+ * It is important that match_futex() will never have a false-positive, esp.
522+ * for PI futexes that can mess up the state. The above argues that false-negatives
523+ * are only possible for malformed programs.
524+ */
525+static u64 get_inode_sequence_number(struct inode *inode)
526+{
527+ static atomic64_t i_seq;
528+ u64 old;
529+
530+ /* Does the inode already have a sequence number? */
531+ old = atomic64_read(&inode->i_sequence);
532+ if (likely(old))
533+ return old;
534+
535+ for (;;) {
536+ u64 new = atomic64_add_return(1, &i_seq);
537+ if (WARN_ON_ONCE(!new))
538+ continue;
539+
540+ old = atomic64_cmpxchg_relaxed(&inode->i_sequence, 0, new);
541+ if (old)
542+ return old;
543+ return new;
544+ }
545+}
546+
508547 /**
509548 * get_futex_key() - Get parameters which are the keys for a futex
510549 * @uaddr: virtual address of the futex
@@ -517,9 +556,15 @@ futex_setup_timer(ktime_t *time, struct hrtimer_sleeper *timeout, · 6 unmodified lines
517556 *
518557 * The key words are stored in @key on success.
519558 *
520− * For shared mappings, it's (page->index, file_inode(vma->vm_file),
521− * offset_within_page). For private mappings, it's (uaddr, current->mm).
522− * We can usually work out the index without swapping in the page.
559+ * For shared mappings (when @fshared), the key is:
560+ * ( inode->i_sequence, page->index, offset_within_page )
561+ * [ also see get_inode_sequence_number() ]
562+ *
563+ * For private mappings (or when !@fshared), the key is:
564+ * ( current->mm, address, 0 )
565+ *
566+ * This allows (cross process, where applicable) identification of the futex
567+ * without keeping the page pinned for the duration of the FUTEX_WAIT.
523568 *
524569 * lock_page() might sleep, the caller should not hold a spinlock.
525570 */
@@ -659,8 +704,6 @@ again: · 133 unmodified lines
659704 key->private.mm = mm;
660705 key->private.address = address;
661706
662− get_futex_key_refs(key); /* implies smp_mb(); (B) */
663−
664707 } else {
665708 struct inode *inode;
666709
@@ -692,40 +735,14 @@ again: · 25 unmodified lines
692735 goto again;
693736 }
694737
695− /*
696− * Take a reference unless it is about to be freed. Previously
697− * this reference was taken by ihold under the page lock
698− * pinning the inode in place so i_lock was unnecessary. The
699− * only way for this check to fail is if the inode was
700− * truncated in parallel which is almost certainly an
701− * application bug. In such a case, just retry.
702− *
703− * We are not calling into get_futex_key_refs() in file-backed
704− * cases, therefore a successful atomic_inc return below will
705− * guarantee that get_futex_key() will still imply smp_mb(); (B).
706− */
707− if (!atomic_inc_not_zero(&inode->i_count)) {
708− rcu_read_unlock();
709− put_page(page);
710−
711− goto again;
712− }
713−
714− /* Should be impossible but lets be paranoid for now */
715− if (WARN_ON_ONCE(inode->i_mapping != mapping)) {
716− err = -EFAULT;
717− rcu_read_unlock();
718− iput(inode);
719−
720− goto out;
721− }
722−
723738 key->both.offset |= FUT_OFF_INODE; /* inode-based key */
724− key->shared.inode = inode;
739+ key->shared.i_seq = get_inode_sequence_number(inode);
725740 key->shared.pgoff = basepage_index(tail);
726741 rcu_read_unlock();
727742 }
728743
744+ get_futex_key_refs(key); /* implies smp_mb(); (B) */
745+
729746 out:
730747 put_page(page);
731748 return err;
kernel/notifier.c+1−1
@@ -519,7 +519,7 @@ NOKPROBE_SYMBOL(notify_die); · 518 unmodified lines
519519
520520 int register_die_notifier(struct notifier_block *nb)
521521 {
522− vmalloc_sync_all();
522+ vmalloc_sync_mappings();
523523 return atomic_notifier_chain_register(&die_chain, nb);
524524 }
525525 EXPORT_SYMBOL_GPL(register_die_notifier);
kernel/pid.c+10−0
@@ -247,6 +247,16 @@ struct pid *alloc_pid(struct pid_namespace *ns, pid_t *set_tid, · 246 unmodified lines
247247 tmp = tmp->parent;
248248 }
249249
250+ /*
251+ * ENOMEM is not the most obvious choice especially for the case
252+ * where the child subreaper has already exited and the pid
253+ * namespace denies the creation of any new processes. But ENOMEM
254+ * is what we have exposed to userspace for a long time and it is
255+ * documented behavior for pid namespaces. So we can't easily
256+ * change it even if there were an error code better suited.
257+ */
258+ retval = -ENOMEM;
259+
250260 if (unlikely(is_child_reaper(pid))) {
251261 if (pid_ns_prepare_proc(ns))
252262 goto out_free;
kernel/sys.c+2−0
@@ -47,6 +47,7 @@ · 46 unmodified lines
4747 #include <linux/syscalls.h>
4848 #include <linux/kprobes.h>
4949 #include <linux/user_namespace.h>
50+#include <linux/time_namespace.h>
5051 #include <linux/binfmts.h>
5152
5253 #include <linux/sched.h>
@@ -2546,6 +2547,7 @@ static int do_sysinfo(struct sysinfo *info) · 2493 unmodified lines
25462547 memset(info, 0, sizeof(struct sysinfo));
25472548
25482549 ktime_get_boottime_ts64(&tp);
2550+ timens_add_boottime(&tp);
25492551 info->uptime = tp.tv_sec + (tp.tv_nsec ? 1 : 0);
25502552
25512553 get_avenrun(info->loads, 0, SI_LOAD_SHIFT - FSHIFT);
kernel/trace/ftrace.c+2−0
@@ -1547,6 +1547,8 @@ static struct dyn_ftrace *lookup_rec(unsigned long start, unsigned long end) · 1546 unmodified lines
15471547 rec = bsearch(&key, pg->records, pg->index,
15481548 sizeof(struct dyn_ftrace),
15491549 ftrace_cmp_recs);
1550+ if (rec)
1551+ break;
15501552 }
15511553 return rec;
15521554 }
kernel/workqueue.c+8−6
@@ -1411,14 +1411,16 @@ static void __queue_work(int cpu, struct workqueue_struct *wq, · 1410 unmodified lines
14111411 return;
14121412 rcu_read_lock();
14131413 retry:
1414− if (req_cpu == WORK_CPU_UNBOUND)
1415− cpu = wq_select_unbound_cpu(raw_smp_processor_id());
1416−
14171414 /* pwq which will be used unless @work is executing elsewhere */
1418− if (!(wq->flags & WQ_UNBOUND))
1419− pwq = per_cpu_ptr(wq->cpu_pwqs, cpu);
1420− else
1415+ if (wq->flags & WQ_UNBOUND) {
1416+ if (req_cpu == WORK_CPU_UNBOUND)
1417+ cpu = wq_select_unbound_cpu(raw_smp_processor_id());
14211418 pwq = unbound_pwq_by_node(wq, cpu_to_node(cpu));
1419+ } else {
1420+ if (req_cpu == WORK_CPU_UNBOUND)
1421+ cpu = raw_smp_processor_id();
1422+ pwq = per_cpu_ptr(wq->cpu_pwqs, cpu);
1423+ }
14221424
14231425 /*
14241426 * If @work was previously on a different pool, it might still be

Keyboard shortcuts

General

Show keyboard shortcuts
?
Focus the current search
/

Issues and pull requests

Open filters
f

Issues, pull requests and Inbox

Select next item
j↓
Select previous item
k↑
Open selected item
Enter
Dismiss filters or clear selection
Esc

Repository code

Go to file
t