[PATCH 09/10] perf/uprobe: Convert single-step and uretprobe to SRCU

[Peter Zijlstra]

Both single-step and uretprobes take a refcount on struct uprobe in
handle_swbp() in order to ensure struct uprobe stays extant until a
next trap.

Since uprobe_unregister() only cares about the uprobe_consumer
life-time, and these intra-trap sections can be arbitrarily large,
create a second SRCU domain to cover these.

Notably, a uretprobe with a registered return_instance that never
triggers -- because userspace -- will currently pin the
return_instance and related uprobe until the task dies. With this
convertion to SRCU this behaviour will inhibit freeing of all uprobes.

Signed-off-by: Peter Zijlstra (Intel) <peterz@xxxxxxxxxxxxx>
---
 include/linux/uprobes.h |    2 ++
 kernel/events/uprobes.c |   23 ++++++++++++++++-------
 2 files changed, 18 insertions(+), 7 deletions(-)

--- a/include/linux/uprobes.h
+++ b/include/linux/uprobes.h
@@ -78,6 +78,7 @@ struct uprobe_task {
 
 	struct return_instance		*return_instances;
 	unsigned int			depth;
+	unsigned int			active_srcu_idx;
 };
 
 struct return_instance {
@@ -86,6 +87,7 @@ struct return_instance {
 	unsigned long		stack;		/* stack pointer */
 	unsigned long		orig_ret_vaddr; /* original return address */
 	bool			chained;	/* true, if instance is nested */
+	int			srcu_idx;
 
 	struct return_instance	*next;		/* keep as stack */
 };
--- a/kernel/events/uprobes.c
+++ b/kernel/events/uprobes.c
@@ -51,6 +51,7 @@ static struct mutex uprobes_mmap_mutex[U
 DEFINE_STATIC_PERCPU_RWSEM(dup_mmap_sem);
 
 DEFINE_STATIC_SRCU(uprobes_srcu);
+DEFINE_STATIC_SRCU(uretprobes_srcu);
 
 /* Have a copy of original instruction */
 #define UPROBE_COPY_INSN	0
@@ -598,12 +599,18 @@ static struct uprobe *get_uprobe(struct
 	return uprobe;
 }
 
-static void uprobe_free_rcu(struct rcu_head *rcu)
+static void uprobe_free_stage2(struct rcu_head *rcu)
 {
 	struct uprobe *uprobe = container_of(rcu, struct uprobe, rcu);
 	kfree(uprobe);
 }
 
+static void uprobe_free_stage1(struct rcu_head *rcu)
+{
+	struct uprobe *uprobe = container_of(rcu, struct uprobe, rcu);
+	call_srcu(&uretprobes_srcu, &uprobe->rcu, uprobe_free_stage2);
+}
+
 static void put_uprobe(struct uprobe *uprobe)
 {
 	if (refcount_dec_and_test(&uprobe->ref)) {
@@ -615,7 +622,7 @@ static void put_uprobe(struct uprobe *up
 		mutex_lock(&delayed_uprobe_lock);
 		delayed_uprobe_remove(uprobe, NULL);
 		mutex_unlock(&delayed_uprobe_lock);
-		call_srcu(&uprobes_srcu, &uprobe->rcu, uprobe_free_rcu);
+		call_srcu(&uprobes_srcu, &uprobe->rcu, uprobe_free_stage1);
 	}
 }
 
@@ -1735,7 +1742,7 @@ unsigned long uprobe_get_trap_addr(struc
 static struct return_instance *free_ret_instance(struct return_instance *ri)
 {
 	struct return_instance *next = ri->next;
-	put_uprobe(ri->uprobe);
+	srcu_read_unlock(&uretprobes_srcu, ri->srcu_idx);
 	kfree(ri);
 	return next;
 }
@@ -1753,7 +1760,7 @@ void uprobe_free_utask(struct task_struc
 		return;
 
 	if (utask->active_uprobe)
-		put_uprobe(utask->active_uprobe);
+		srcu_read_unlock(&uretprobes_srcu, utask->active_srcu_idx);
 
 	ri = utask->return_instances;
 	while (ri)
@@ -1796,7 +1803,7 @@ static int dup_utask(struct task_struct
 			return -ENOMEM;
 
 		*n = *o;
-		get_uprobe(n->uprobe);
+		__srcu_clone_read_lock(&uretprobes_srcu, n->srcu_idx);
 		n->next = NULL;
 
 		*p = n;
@@ -1939,7 +1946,8 @@ static void prepare_uretprobe(struct upr
 		orig_ret_vaddr = utask->return_instances->orig_ret_vaddr;
 	}
 
-	ri->uprobe = get_uprobe(uprobe);
+	ri->srcu_idx = srcu_read_lock(&uretprobes_srcu);
+	ri->uprobe = uprobe;
 	ri->func = instruction_pointer(regs);
 	ri->stack = user_stack_pointer(regs);
 	ri->orig_ret_vaddr = orig_ret_vaddr;
@@ -1979,7 +1987,8 @@ pre_ssout(struct uprobe *uprobe, struct
 		return err;
 	}
 
-	utask->active_uprobe = get_uprobe(uprobe);
+	utask->active_srcu_idx = srcu_read_lock(&uretprobes_srcu);
+	utask->active_uprobe = uprobe;
 	utask->state = UTASK_SSTEP;
 	return 0;
 }