[PATCH 7/9] liblockdep: Support using LD_PRELOAD

[Sasha Levin]

This allows lockdep to be used without being compiled in the
original program.

Usage is quite simple:

	LD_PRELOAD=/path/to/liblockdep.so /path/to/my/program

And magically, you'll have lockdep checking in your program!

Signed-off-by: Sasha Levin <sasha.levin@xxxxxxxxxx>
---
 tools/lib/lockdep/preload.c | 386 ++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 386 insertions(+)
 create mode 100644 tools/lib/lockdep/preload.c

diff --git a/tools/lib/lockdep/preload.c b/tools/lib/lockdep/preload.c
new file mode 100644
index 0000000..6ae71c6
--- /dev/null
+++ b/tools/lib/lockdep/preload.c
@@ -0,0 +1,386 @@
+#define _GNU_SOURCE
+#include <pthread.h>
+#include <stdio.h>
+#include <dlfcn.h>
+#include <stdlib.h>
+#include "include/liblockdep/mutex.h"
+#include "../../../include/linux/rbtree.h"
+
+/**
+ * struct lock_lookup - liblockdep's view of a single unique lock
+ * @orig: pointer to the original pthread lock, used for lookups
+ * @dep_map: lockdep's dep_map structure
+ * @key: lockdep's key structure
+ * @node: rb-tree node used to store the lock in a global tree
+ * @name: a unique name for the lock
+ */
+struct lock_lookup {
+	void *orig; /* Original pthread lock, used for lookups */
+	struct lockdep_map dep_map; /* Since all locks are dynamic, we need
+				     * a dep_map and a key for each lock */
+	struct lock_class_key key;
+	struct rb_node node;
+#define LIBLOCKDEP_MAX_LOCK_NAME 22
+	char name[LIBLOCKDEP_MAX_LOCK_NAME];
+};
+
+/* This is where we store our locks */
+static struct rb_root locks = RB_ROOT;
+static pthread_rwlock_t locks_rwlock = PTHREAD_RWLOCK_INITIALIZER;
+
+/* pthread mutex API */
+static int (*ll_pthread_mutex_init)(pthread_mutex_t *mutex,
+			const pthread_mutexattr_t *attr);
+static int (*ll_pthread_mutex_lock)(pthread_mutex_t *mutex);
+static int (*ll_pthread_mutex_trylock)(pthread_mutex_t *mutex);
+static int (*ll_pthread_mutex_unlock)(pthread_mutex_t *mutex);
+static int (*ll_pthread_mutex_destroy)(pthread_mutex_t *mutex);
+
+/* pthread rwlock API */
+static int (*ll_pthread_rwlock_init)(pthread_rwlock_t *rwlock,
+			const pthread_rwlockattr_t *attr);
+static int (*ll_pthread_rwlock_destroy)(pthread_rwlock_t *rwlock);
+static int (*ll_pthread_rwlock_rdlock)(pthread_rwlock_t *rwlock);
+static int (*ll_pthread_rwlock_tryrdlock)(pthread_rwlock_t *rwlock);
+static int (*ll_pthread_rwlock_trywrlock)(pthread_rwlock_t *rwlock);
+static int (*ll_pthread_rwlock_wrlock)(pthread_rwlock_t *rwlock);
+static int (*ll_pthread_rwlock_unlock)(pthread_rwlock_t *rwlock);
+
+static bool preload_done;
+static void init_preload(void);
+static void try_init_preload(void)
+{
+	if (!preload_done)
+		init_preload();
+}
+
+static struct rb_node **__get_lock_node(void *lock, struct rb_node **parent)
+{
+	struct rb_node **node = &locks.rb_node;
+	struct lock_lookup *l;
+
+	*parent = NULL;
+
+	while (*node) {
+		l = rb_entry(*node, struct lock_lookup, node);
+
+		*parent = *node;
+		if (lock < l->orig)
+			node = &l->node.rb_left;
+		else if (lock > l->orig)
+			node = &l->node.rb_right;
+		else
+			return node;
+	}
+
+	return node;
+}
+
+/**
+ * __get_lock - find or create a lock instance
+ * @lock: pointer to a pthread lock function
+ *
+ * Try to find an existing lock in the rbtree using the provided pointer. If
+ * one wasn't found - create it.
+ */
+static struct lock_lookup *__get_lock(void *lock)
+{
+	struct rb_node **node, *parent;
+	struct lock_lookup *l;
+
+	ll_pthread_rwlock_rdlock(&locks_rwlock);
+	node = __get_lock_node(lock, &parent);
+	ll_pthread_rwlock_unlock(&locks_rwlock);
+	if (*node) {
+		return rb_entry(*node, struct lock_lookup, node);
+	}
+
+	/* We didn't find the lock, let's create it */
+	l = malloc(sizeof(*l));
+	if (l == NULL)
+		return NULL;
+
+	l->orig = lock;
+	/*
+	 * Currently the name of the lock is the ptr value of the pthread lock,
+	 * while not optimal, it makes debugging a bit easier.
+	 *
+	 * TODO: Get the real name of the lock using libdwarf
+	 */
+	sprintf(l->name, "%p", lock);
+	lockdep_init_map(&l->dep_map, l->name, &l->key, 0);
+
+	ll_pthread_rwlock_wrlock(&locks_rwlock);
+	/* This might have changed since the last time we fetched it */
+	node = __get_lock_node(lock, &parent);
+	rb_link_node(&l->node, parent, node);
+	rb_insert_color(&l->node, &locks);
+	ll_pthread_rwlock_unlock(&locks_rwlock);
+
+	return l;
+}
+
+int pthread_mutex_init(pthread_mutex_t *mutex,
+			const pthread_mutexattr_t *attr)
+{
+	int r;
+
+	/*
+	 * We keep trying to init our preload module because there might be
+	 * code in init sections that tries to touch locks before we are
+	 * initialized, in that case we'll need to manually call preload
+	 * to get us going.
+	 *
+	 * Funny enough, kernel's lockdep had the same issue, and used
+	 * (almost) the same solution. See look_up_lock_class() in
+	 * kernel/lockdep.c for details.
+	 */
+	try_init_preload();
+
+	r = ll_pthread_mutex_init(mutex, attr);
+	if (r == 0)
+		/*
+		 * We do a dummy initialization here so that lockdep could
+		 * warn us if something fishy is going on - such as
+		 * initializing a held lock.
+		 */
+		__get_lock(mutex);
+
+	return r;
+}
+
+int pthread_mutex_lock(pthread_mutex_t *mutex)
+{
+	int r;
+
+        try_init_preload();
+
+	lock_acquire(&__get_lock(mutex)->dep_map, 0, 0, 0, 2, NULL,
+			(unsigned long)_THIS_IP_);
+	/*
+	 * Here's the thing with pthread mutexes: unlike the kernel variant,
+	 * they can fail.
+	 *
+	 * This means that the behaviour here is a bit different from what's
+	 * going on in the kernel: there we just tell lockdep that we took the
+	 * lock before actually taking it, but here we must deal with the case
+	 * that locking failed.
+	 *
+	 * To do that we'll "release" the lock if locking failed - this way
+	 * we'll get lockdep doing the correct checks when we try to take
+	 * the lock, and if that fails - we'll be back to the correct
+	 * state by releasing it.
+	 */
+	r = ll_pthread_mutex_lock(mutex);
+	if (r)
+		lock_release(&__get_lock(mutex)->dep_map, 0, (unsigned long)_THIS_IP_);
+
+	return r;
+}
+
+int pthread_mutex_trylock(pthread_mutex_t *mutex)
+{
+	int r;
+
+        try_init_preload();
+
+	lock_acquire(&__get_lock(mutex)->dep_map, 0, 1, 0, 2, NULL, (unsigned long)_THIS_IP_);
+	r = ll_pthread_mutex_trylock(mutex);
+	if (r)
+		lock_release(&__get_lock(mutex)->dep_map, 0, (unsigned long)_THIS_IP_);
+
+	return r;
+}
+
+int pthread_mutex_unlock(pthread_mutex_t *mutex)
+{
+	int r;
+
+        try_init_preload();
+
+	lock_release(&__get_lock(mutex)->dep_map, 0, (unsigned long)_THIS_IP_);
+	/*
+	 * Just like taking a lock, only in reverse!
+	 *
+	 * If we fail releasing the lock, tell lockdep we're holding it again.
+	 */
+	r = ll_pthread_mutex_unlock(mutex);
+	if (r)
+		lock_acquire(&__get_lock(mutex)->dep_map, 0, 0, 0, 2, NULL, (unsigned long)_THIS_IP_);
+
+	return r;
+}
+
+int pthread_mutex_destroy(pthread_mutex_t *mutex)
+{
+	struct lock_lookup *l = __get_lock(mutex);
+
+        try_init_preload();
+
+	/*
+	 * Let's see if we're releasing a lock that's held.
+	 * 
+	 * TODO: Hook into free() and add that check there as well.
+	 */
+	debug_check_no_locks_freed(mutex, mutex + sizeof(*mutex));
+	ll_pthread_rwlock_wrlock(&locks_rwlock);
+	rb_erase(&l->node, &locks);
+	ll_pthread_rwlock_unlock(&locks_rwlock);
+	free(l);
+	return ll_pthread_mutex_destroy(mutex);
+}
+
+/* This is the rwlock part, very similar to what happened with mutex above */
+int pthread_rwlock_init(pthread_rwlock_t *rwlock,
+			const pthread_rwlockattr_t *attr)
+{
+	int r;
+
+        try_init_preload();
+
+	r = ll_pthread_rwlock_init(rwlock, attr);
+	if (r == 0)
+		__get_lock(rwlock);
+
+	return r;
+}
+
+int pthread_rwlock_destroy(pthread_rwlock_t *rwlock)
+{
+	struct lock_lookup *l = __get_lock(rwlock);
+
+        try_init_preload();
+
+	debug_check_no_locks_freed(rwlock, rwlock + sizeof(*rwlock));
+	ll_pthread_rwlock_wrlock(&locks_rwlock);
+	rb_erase(&l->node, &locks);
+	ll_pthread_rwlock_unlock(&locks_rwlock);
+	free(l);
+	return ll_pthread_rwlock_destroy(rwlock);
+}
+
+int pthread_rwlock_rdlock(pthread_rwlock_t *rwlock)
+{
+	int r;
+
+        init_preload();
+
+	lock_acquire(&__get_lock(rwlock)->dep_map, 0, 0, 2, 2, NULL, (unsigned long)_THIS_IP_);
+	r = ll_pthread_rwlock_rdlock(rwlock);
+	if (r)
+		lock_release(&__get_lock(rwlock)->dep_map, 0, (unsigned long)_THIS_IP_);
+
+	return r;
+}
+
+int pthread_rwlock_tryrdlock(pthread_rwlock_t *rwlock)
+{
+	int r;
+
+        init_preload();
+
+	lock_acquire(&__get_lock(rwlock)->dep_map, 0, 1, 2, 2, NULL, (unsigned long)_THIS_IP_);
+	r = ll_pthread_rwlock_tryrdlock(rwlock);
+	if (r)
+		lock_release(&__get_lock(rwlock)->dep_map, 0, (unsigned long)_THIS_IP_);
+
+	return r;
+}
+
+int pthread_rwlock_trywrlock(pthread_rwlock_t *rwlock)
+{
+	int r;
+
+        init_preload();
+
+	lock_acquire(&__get_lock(rwlock)->dep_map, 0, 1, 0, 2, NULL, (unsigned long)_THIS_IP_);
+	r = ll_pthread_rwlock_trywrlock(rwlock);
+	if (r)
+                lock_release(&__get_lock(rwlock)->dep_map, 0, (unsigned long)_THIS_IP_);
+
+	return r;
+}
+
+int pthread_rwlock_wrlock(pthread_rwlock_t *rwlock)
+{
+	int r;
+
+        init_preload();
+
+	lock_acquire(&__get_lock(rwlock)->dep_map, 0, 0, 0, 2, NULL, (unsigned long)_THIS_IP_);
+	r = ll_pthread_rwlock_wrlock(rwlock);
+	if (r)
+		lock_release(&__get_lock(rwlock)->dep_map, 0, (unsigned long)_THIS_IP_);
+
+	return r;
+}
+
+int pthread_rwlock_unlock(pthread_rwlock_t *rwlock)
+{
+	int r;
+
+        init_preload();
+
+	lock_release(&__get_lock(rwlock)->dep_map, 0, (unsigned long)_THIS_IP_);
+	r = ll_pthread_rwlock_unlock(rwlock);
+	if (r)
+		lock_acquire(&__get_lock(rwlock)->dep_map, 0, 0, 0, 2, NULL, (unsigned long)_THIS_IP_);
+
+	return r;
+}
+
+__attribute__((constructor)) static void init_preload(void)
+{
+	static bool preload_started;
+
+	if (preload_done)
+		return;
+
+	/*
+	 * Some programs attempt to initialize and use locks in their
+	 * allocation path. This means that a call to malloc() would
+	 * result in locks being initialized and locked.
+	 *
+	 * Why is it an issue for us? dlsym() below will try allocating to
+	 * give us the original function. Since this allocation will result
+	 * in a locking operations, we have to let pthread deal with it,
+	 * but we can't! we don't have the pointer to the original API
+	 * since we're inside dlsym() trying to get it :(
+	 *
+	 * We can work around it by telling the program that locking was
+	 * really okay, and just initialize those locks when we're fully
+	 * up and running (this is ok because this all happens during
+	 * initialization phase, when we have just one thread). But
+	 * this is a big TODO at this point.
+	 */
+	if (preload_started) {
+		printf(
+		"LOCKDEP error: It seems that the program you are trying to "
+		"debug is initializing locks in it's allocation path.\n"
+		"This means that liblockdep cannot reliably analyze this "
+		"program since we need the allocator to work before we can "
+		"debug locks.\nSorry!\n");
+
+		exit(1);
+	}
+
+	preload_started = true;
+
+	ll_pthread_mutex_init = dlsym(RTLD_NEXT, "pthread_mutex_init");
+	ll_pthread_mutex_lock = dlsym(RTLD_NEXT, "pthread_mutex_lock");
+	ll_pthread_mutex_trylock = dlsym(RTLD_NEXT, "pthread_mutex_trylock");
+	ll_pthread_mutex_unlock = dlsym(RTLD_NEXT, "pthread_mutex_unlock");
+	ll_pthread_mutex_destroy = dlsym(RTLD_NEXT, "pthread_mutex_destroy");
+
+	ll_pthread_rwlock_init = dlsym(RTLD_NEXT, "pthread_rwlock_init");
+	ll_pthread_rwlock_destroy = dlsym(RTLD_NEXT, "pthread_rwlock_destroy");
+	ll_pthread_rwlock_rdlock = dlsym(RTLD_NEXT, "pthread_rwlock_rdlock");
+	ll_pthread_rwlock_tryrdlock = dlsym(RTLD_NEXT, "pthread_rwlock_tryrdlock");
+	ll_pthread_rwlock_wrlock = dlsym(RTLD_NEXT, "pthread_rwlock_wrlock");
+	ll_pthread_rwlock_trywrlock = dlsym(RTLD_NEXT, "pthread_rwlock_trywrlock");
+	ll_pthread_rwlock_unlock = dlsym(RTLD_NEXT, "pthread_rwlock_unlock");
+
+	lockdep_init();
+
+	preload_done = true;
+}
-- 
1.8.2.1

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/