}
}
+/*
+ * Because of perf_event::ctx migration in sys_perf_event_open::move_group we
+ * need some magic.
+ *
+ * Those places that change perf_event::ctx will hold both
+ * perf_event_ctx::mutex of the 'old' and 'new' ctx value.
+ *
+ * Lock ordering is by mutex address. There is one other site where
+ * perf_event_context::mutex nests and that is put_event(). But remember that
+ * that is a parent<->child context relation, and migration does not affect
+ * children, therefore these two orderings should not interact.
+ *
+ * The change in perf_event::ctx does not affect children (as claimed above)
+ * because the sys_perf_event_open() case will install a new event and break
+ * the ctx parent<->child relation.
+ *
+ * The places that change perf_event::ctx will issue:
+ *
+ * perf_remove_from_context();
+ * synchronize_rcu();
+ * perf_install_in_context();
+ *
+ * to affect the change. The remove_from_context() + synchronize_rcu() should
+ * quiesce the event, after which we can install it in the new location. This
+ * means that only external vectors (perf_fops, prctl) can perturb the event
+ * while in transit. Therefore all such accessors should also acquire
+ * perf_event_context::mutex to serialize against this.
+ *
+ * However; because event->ctx can change while we're waiting to acquire
+ * ctx->mutex we must be careful and use the below perf_event_ctx_lock()
+ * function.
+ *
+ * Lock order:
+ * task_struct::perf_event_mutex
+ * perf_event_context::mutex
+ * perf_event_context::lock
+ * perf_event::child_mutex;
+ * perf_event::mmap_mutex
+ * mmap_sem
+ */
+static struct perf_event_context *perf_event_ctx_lock(struct perf_event *event)
+{
+ struct perf_event_context *ctx;
+
+again:
+ rcu_read_lock();
+ ctx = ACCESS_ONCE(event->ctx);
+ if (!atomic_inc_not_zero(&ctx->refcount)) {
+ rcu_read_unlock();
+ goto again;
+ }
+ rcu_read_unlock();
+
+ mutex_lock(&ctx->mutex);
+ if (event->ctx != ctx) {
+ mutex_unlock(&ctx->mutex);
+ put_ctx(ctx);
+ goto again;
+ }
+
+ return ctx;
+}
+
+static void perf_event_ctx_unlock(struct perf_event *event,
+ struct perf_event_context *ctx)
+{
+ mutex_unlock(&ctx->mutex);
+ put_ctx(ctx);
+}
+
static void unclone_ctx(struct perf_event_context *ctx)
{
if (ctx->parent_ctx) {
* is the current context on this CPU and preemption is disabled,
* hence we can't get into perf_event_task_sched_out for this context.
*/
-void perf_event_disable(struct perf_event *event)
+static void _perf_event_disable(struct perf_event *event)
{
struct perf_event_context *ctx = event->ctx;
struct task_struct *task = ctx->task;
raw_spin_unlock_irq(&ctx->lock);
}
+/*
+ * Strictly speaking kernel users cannot create groups and therefore this
+ * interface does not need the perf_event_ctx_lock() magic.
+ */
+void perf_event_disable(struct perf_event *event)
+{
+ struct perf_event_context *ctx;
+
+ ctx = perf_event_ctx_lock(event);
+ _perf_event_disable(event);
+ perf_event_ctx_unlock(event, ctx);
+}
+
static void perf_set_shadow_time(struct perf_event *event,
struct perf_event_context *ctx,
u64 tstamp)
* perf_event_for_each_child or perf_event_for_each as described
* for perf_event_disable.
*/
-void perf_event_enable(struct perf_event *event)
+static void _perf_event_enable(struct perf_event *event)
{
struct perf_event_context *ctx = event->ctx;
struct task_struct *task = ctx->task;
raw_spin_unlock_irq(&ctx->lock);
}
-int perf_event_refresh(struct perf_event *event, int refresh)
+/*
+ * See perf_event_disable();
+ */
+void perf_event_enable(struct perf_event *event)
+{
+ struct perf_event_context *ctx;
+
+ ctx = perf_event_ctx_lock(event);
+ _perf_event_enable(event);
+ perf_event_ctx_unlock(event, ctx);
+}
+
+static int _perf_event_refresh(struct perf_event *event, int refresh)
{
/*
* not supported on inherited events
return -EINVAL;
atomic_add(refresh, &event->event_limit);
- perf_event_enable(event);
+ _perf_event_enable(event);
return 0;
}
+
+/*
+ * See perf_event_disable()
+ */
+int perf_event_refresh(struct perf_event *event, int refresh)
+{
+ struct perf_event_context *ctx;
+ int ret;
+
+ ctx = perf_event_ctx_lock(event);
+ ret = _perf_event_refresh(event, refresh);
+ perf_event_ctx_unlock(event, ctx);
+
+ return ret;
+}
EXPORT_SYMBOL_GPL(perf_event_refresh);
static void ctx_sched_out(struct perf_event_context *ctx,
/* Reuse ptrace permission checks for now. */
err = -EACCES;
- if (!ptrace_may_access(task, PTRACE_MODE_READ))
+ if (!ptrace_may_access(task, PTRACE_MODE_READ_REALCREDS))
goto errout;
return task;
rcu_read_unlock();
if (owner) {
- mutex_lock(&owner->perf_event_mutex);
+ /*
+ * If we're here through perf_event_exit_task() we're already
+ * holding ctx->mutex which would be an inversion wrt. the
+ * normal lock order.
+ *
+ * However we can safely take this lock because its the child
+ * ctx->mutex.
+ */
+ mutex_lock_nested(&owner->perf_event_mutex, SINGLE_DEPTH_NESTING);
+
/*
* We have to re-check the event->owner field, if it is cleared
* we raced with perf_event_exit_task(), acquiring the mutex
u64 read_format, char __user *buf)
{
struct perf_event *leader = event->group_leader, *sub;
- int n = 0, size = 0, ret = -EFAULT;
struct perf_event_context *ctx = leader->ctx;
- u64 values[5];
+ int n = 0, size = 0, ret;
u64 count, enabled, running;
+ u64 values[5];
+
+ lockdep_assert_held(&ctx->mutex);
- mutex_lock(&ctx->mutex);
count = perf_event_read_value(leader, &enabled, &running);
values[n++] = 1 + leader->nr_siblings;
size = n * sizeof(u64);
if (copy_to_user(buf, values, size))
- goto unlock;
+ return -EFAULT;
ret = size;
size = n * sizeof(u64);
if (copy_to_user(buf + ret, values, size)) {
- ret = -EFAULT;
- goto unlock;
+ return -EFAULT;
}
ret += size;
}
-unlock:
- mutex_unlock(&ctx->mutex);
return ret;
}
perf_read(struct file *file, char __user *buf, size_t count, loff_t *ppos)
{
struct perf_event *event = file->private_data;
+ struct perf_event_context *ctx;
+ int ret;
- return perf_read_hw(event, buf, count);
+ ctx = perf_event_ctx_lock(event);
+ ret = perf_read_hw(event, buf, count);
+ perf_event_ctx_unlock(event, ctx);
+
+ return ret;
}
static unsigned int perf_poll(struct file *file, poll_table *wait)
return events;
}
-static void perf_event_reset(struct perf_event *event)
+static void _perf_event_reset(struct perf_event *event)
{
(void)perf_event_read(event);
local64_set(&event->count, 0);
struct perf_event *child;
WARN_ON_ONCE(event->ctx->parent_ctx);
+
mutex_lock(&event->child_mutex);
func(event);
list_for_each_entry(child, &event->child_list, child_list)
struct perf_event_context *ctx = event->ctx;
struct perf_event *sibling;
- WARN_ON_ONCE(ctx->parent_ctx);
- mutex_lock(&ctx->mutex);
+ lockdep_assert_held(&ctx->mutex);
+
event = event->group_leader;
perf_event_for_each_child(event, func);
func(event);
list_for_each_entry(sibling, &event->sibling_list, group_entry)
- perf_event_for_each_child(event, func);
- mutex_unlock(&ctx->mutex);
+ perf_event_for_each_child(sibling, func);
}
static int perf_event_period(struct perf_event *event, u64 __user *arg)
struct perf_event *output_event);
static int perf_event_set_filter(struct perf_event *event, void __user *arg);
-static long perf_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
+static long _perf_ioctl(struct perf_event *event, unsigned int cmd, unsigned long arg)
{
- struct perf_event *event = file->private_data;
void (*func)(struct perf_event *);
u32 flags = arg;
switch (cmd) {
case PERF_EVENT_IOC_ENABLE:
- func = perf_event_enable;
+ func = _perf_event_enable;
break;
case PERF_EVENT_IOC_DISABLE:
- func = perf_event_disable;
+ func = _perf_event_disable;
break;
case PERF_EVENT_IOC_RESET:
- func = perf_event_reset;
+ func = _perf_event_reset;
break;
case PERF_EVENT_IOC_REFRESH:
- return perf_event_refresh(event, arg);
+ return _perf_event_refresh(event, arg);
case PERF_EVENT_IOC_PERIOD:
return perf_event_period(event, (u64 __user *)arg);
return 0;
}
+static long perf_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
+{
+ struct perf_event *event = file->private_data;
+ struct perf_event_context *ctx;
+ long ret;
+
+ ctx = perf_event_ctx_lock(event);
+ ret = _perf_ioctl(event, cmd, arg);
+ perf_event_ctx_unlock(event, ctx);
+
+ return ret;
+}
+
#ifdef CONFIG_COMPAT
static long perf_compat_ioctl(struct file *file, unsigned int cmd,
unsigned long arg)
int perf_event_task_enable(void)
{
+ struct perf_event_context *ctx;
struct perf_event *event;
mutex_lock(¤t->perf_event_mutex);
- list_for_each_entry(event, ¤t->perf_event_list, owner_entry)
- perf_event_for_each_child(event, perf_event_enable);
+ list_for_each_entry(event, ¤t->perf_event_list, owner_entry) {
+ ctx = perf_event_ctx_lock(event);
+ perf_event_for_each_child(event, _perf_event_enable);
+ perf_event_ctx_unlock(event, ctx);
+ }
mutex_unlock(¤t->perf_event_mutex);
return 0;
int perf_event_task_disable(void)
{
+ struct perf_event_context *ctx;
struct perf_event *event;
mutex_lock(¤t->perf_event_mutex);
- list_for_each_entry(event, ¤t->perf_event_list, owner_entry)
- perf_event_for_each_child(event, perf_event_disable);
+ list_for_each_entry(event, ¤t->perf_event_list, owner_entry) {
+ ctx = perf_event_ctx_lock(event);
+ perf_event_for_each_child(event, _perf_event_disable);
+ perf_event_ctx_unlock(event, ctx);
+ }
mutex_unlock(¤t->perf_event_mutex);
return 0;
/* Recursion avoidance in each contexts */
int recursion[PERF_NR_CONTEXTS];
-
- /* Keeps track of cpu being initialized/exited */
- bool online;
};
static DEFINE_PER_CPU(struct swevent_htable, swevent_htable);
hwc->state = !(flags & PERF_EF_START);
head = find_swevent_head(swhash, event);
- if (!head) {
- /*
- * We can race with cpu hotplug code. Do not
- * WARN if the cpu just got unplugged.
- */
- WARN_ON_ONCE(swhash->online);
+ if (WARN_ON_ONCE(!head))
return -EINVAL;
- }
hlist_add_head_rcu(&event->hlist_entry, head);
int err = 0;
mutex_lock(&swhash->hlist_mutex);
-
if (!swevent_hlist_deref(swhash) && cpu_online(cpu)) {
struct swevent_hlist *hlist;
{
void *record = data->raw->data;
+ /* only top level events have filters set */
+ if (event->parent)
+ event = event->parent;
+
if (likely(!event->filter) || filter_match_preds(event->filter, record))
return 1;
return 0;
return ret;
}
+static void mutex_lock_double(struct mutex *a, struct mutex *b)
+{
+ if (b < a)
+ swap(a, b);
+
+ mutex_lock(a);
+ mutex_lock_nested(b, SINGLE_DEPTH_NESTING);
+}
+
+/*
+ * Variation on perf_event_ctx_lock_nested(), except we take two context
+ * mutexes.
+ */
+static struct perf_event_context *
+__perf_event_ctx_lock_double(struct perf_event *group_leader,
+ struct perf_event_context *ctx)
+{
+ struct perf_event_context *gctx;
+
+again:
+ rcu_read_lock();
+ gctx = ACCESS_ONCE(group_leader->ctx);
+ if (!atomic_inc_not_zero(&gctx->refcount)) {
+ rcu_read_unlock();
+ goto again;
+ }
+ rcu_read_unlock();
+
+ mutex_lock_double(&gctx->mutex, &ctx->mutex);
+
+ if (group_leader->ctx != gctx) {
+ mutex_unlock(&ctx->mutex);
+ mutex_unlock(&gctx->mutex);
+ put_ctx(gctx);
+ goto again;
+ }
+
+ return gctx;
+}
+
/**
* sys_perf_event_open - open a performance event, associate it to a task/cpu
*
struct perf_event *group_leader = NULL, *output_event = NULL;
struct perf_event *event, *sibling;
struct perf_event_attr attr;
- struct perf_event_context *ctx;
+ struct perf_event_context *ctx, *uninitialized_var(gctx);
struct file *event_file = NULL;
struct file *group_file = NULL;
struct task_struct *task = NULL;
}
if (move_group) {
- struct perf_event_context *gctx = group_leader->ctx;
+ gctx = __perf_event_ctx_lock_double(group_leader, ctx);
- mutex_lock(&gctx->mutex);
+ /*
+ * Check if we raced against another sys_perf_event_open() call
+ * moving the software group underneath us.
+ */
+ if (!(group_leader->group_flags & PERF_GROUP_SOFTWARE)) {
+ /*
+ * If someone moved the group out from under us, check
+ * if this new event wound up on the same ctx, if so
+ * its the regular !move_group case, otherwise fail.
+ */
+ if (gctx != ctx) {
+ err = -EINVAL;
+ goto err_locked;
+ } else {
+ perf_event_ctx_unlock(group_leader, gctx);
+ move_group = 0;
+ }
+ }
+
+ /*
+ * See perf_event_ctx_lock() for comments on the details
+ * of swizzling perf_event::ctx.
+ */
perf_remove_from_context(group_leader, false);
/*
perf_event__state_init(sibling);
put_ctx(gctx);
}
- mutex_unlock(&gctx->mutex);
- put_ctx(gctx);
+ } else {
+ mutex_lock(&ctx->mutex);
}
WARN_ON_ONCE(ctx->parent_ctx);
- mutex_lock(&ctx->mutex);
if (move_group) {
+ /*
+ * Wait for everybody to stop referencing the events through
+ * the old lists, before installing it on new lists.
+ */
+ synchronize_rcu();
+
perf_install_in_context(ctx, group_leader, cpu);
get_ctx(ctx);
list_for_each_entry(sibling, &group_leader->sibling_list,
perf_install_in_context(ctx, event, cpu);
++ctx->generation;
perf_unpin_context(ctx);
+
+ if (move_group) {
+ perf_event_ctx_unlock(group_leader, gctx);
+ put_ctx(gctx);
+ }
mutex_unlock(&ctx->mutex);
event->owner = current;
fd_install(event_fd, event_file);
return event_fd;
+err_locked:
+ if (move_group)
+ perf_event_ctx_unlock(group_leader, gctx);
+ mutex_unlock(&ctx->mutex);
+ fput(event_file);
err_context:
perf_unpin_context(ctx);
put_ctx(ctx);
err_alloc:
- free_event(event);
+ /*
+ * If event_file is set, the fput() above will have called ->release()
+ * and that will take care of freeing the event.
+ */
+ if (!event_file)
+ free_event(event);
err_task:
if (task)
put_task_struct(task);
ret = inherit_task_group(event, parent, parent_ctx,
child, ctxn, &inherited_all);
if (ret)
- break;
+ goto out_unlock;
}
/*
ret = inherit_task_group(event, parent, parent_ctx,
child, ctxn, &inherited_all);
if (ret)
- break;
+ goto out_unlock;
}
raw_spin_lock_irqsave(&parent_ctx->lock, flags);
}
raw_spin_unlock_irqrestore(&parent_ctx->lock, flags);
+out_unlock:
mutex_unlock(&parent_ctx->mutex);
perf_unpin_context(parent_ctx);
struct swevent_htable *swhash = &per_cpu(swevent_htable, cpu);
mutex_lock(&swhash->hlist_mutex);
- swhash->online = true;
if (swhash->hlist_refcount > 0) {
struct swevent_hlist *hlist;
static void perf_event_exit_cpu(int cpu)
{
- struct swevent_htable *swhash = &per_cpu(swevent_htable, cpu);
-
perf_event_exit_cpu_context(cpu);
-
- mutex_lock(&swhash->hlist_mutex);
- swhash->online = false;
- swevent_hlist_release(swhash);
- mutex_unlock(&swhash->hlist_mutex);
}
#else
static inline void perf_event_exit_cpu(int cpu) { }