fix extremely rare but dangerous race condition in robust mutexes

if new shared mappings of files/devices/shared memory can be made
between the time a robust mutex is unlocked and its subsequent removal
from the pending slot in the robustlist header, the kernel can
inadvertently corrupt data in the newly-mapped pages when the process
terminates. i am fixing the bug by using the same global vm lock
mechanism that was used to fix the race condition with unmapping
barriers after pthread_barrier_wait returns.

3 files changed, 33 insertions, 20 deletions

diff --git a/src/thread/pthread_barrier_wait.c b/src/thread/pthread_barrier_wait.c
index 6052925a..5e603380 100644
--- a/src/thread/pthread_barrier_wait.c
+++ b/src/thread/pthread_barrier_wait.c
@@ -1,22 +1,7 @@
 #include "pthread_impl.h"
 
-static int vmlock[2];
-
-void __vm_lock(int inc)
-{
-	for (;;) {
-		int v = vmlock[0];
-		if (inc*v < 0) __wait(vmlock, vmlock+1, v, 1);
-		else if (a_cas(vmlock, v, v+inc)==v) break;
-	}
-}
-
-void __vm_unlock(void)
-{
-	int inc = vmlock[0]>0 ? -1 : 1;
-	if (a_fetch_add(vmlock, inc)==-inc && vmlock[1])
-		__wake(vmlock, -1, 1);
-}
+void __vm_lock_impl(int);
+void __vm_unlock_impl(void);
 
 static int pshared_barrier_wait(pthread_barrier_t *b)
 {
@@ -41,7 +26,7 @@ static int pshared_barrier_wait(pthread_barrier_t *b)
 			__wait(&b->_b_count, &b->_b_waiters2, v, 0);
 	}
 
-	__vm_lock(+1);
+	__vm_lock_impl(+1);
 
 	/* Ensure all threads have a vm lock before proceeding */
 	if (a_fetch_add(&b->_b_count, -1)==1-limit) {
@@ -62,7 +47,7 @@ static int pshared_barrier_wait(pthread_barrier_t *b)
 	if (v==INT_MIN+1 || (v==1 && w))
 		__wake(&b->_b_lock, 1, 0);
 
-	__vm_unlock();
+	__vm_unlock_impl();
 
 	return ret;
 }
diff --git a/src/thread/pthread_mutex_unlock.c b/src/thread/pthread_mutex_unlock.c
index fdf9fc10..5fc0f4e5 100644
--- a/src/thread/pthread_mutex_unlock.c
+++ b/src/thread/pthread_mutex_unlock.c
@@ -1,5 +1,8 @@
 #include "pthread_impl.h"
 
+void __vm_lock_impl(int);
+void __vm_unlock_impl(void);
+
 int pthread_mutex_unlock(pthread_mutex_t *m)
 {
 	pthread_t self;
@@ -20,11 +23,14 @@ int pthread_mutex_unlock(pthread_mutex_t *m)
 			self->robust_list.pending = &m->_m_next;
 			*(void **)m->_m_prev = m->_m_next;
 			if (m->_m_next) ((void **)m->_m_next)[-1] = m->_m_prev;
+			__vm_lock_impl(+1);
 		}
 	}
 	cont = a_swap(&m->_m_lock, 0);
-	if (robust)
+	if (robust) {
 		self->robust_list.pending = 0;
+		__vm_unlock_impl();
+	}
 	if (waiters || cont<0)
 		__wake(&m->_m_lock, 1, 0);
 	return 0;
diff --git a/src/thread/vmlock.c b/src/thread/vmlock.c
new file mode 100644
index 00000000..aba9e311
--- /dev/null
+++ b/src/thread/vmlock.c
@@ -0,0 +1,22 @@
+#include "pthread_impl.h"
+
+static int vmlock[2];
+
+void __vm_lock(int inc)
+{
+	for (;;) {
+		int v = vmlock[0];
+		if (inc*v < 0) __wait(vmlock, vmlock+1, v, 1);
+		else if (a_cas(vmlock, v, v+inc)==v) break;
+	}
+}
+
+void __vm_unlock(void)
+{
+	int inc = vmlock[0]>0 ? -1 : 1;
+	if (a_fetch_add(vmlock, inc)==-inc && vmlock[1])
+		__wake(vmlock, -1, 1);
+}
+
+weak_alias(__vm_lock, __vm_lock_impl);
+weak_alias(__vm_unlock, __vm_unlock_impl);