File xsa314.patch of Package xen.22546

Overview Repositories Revisions Requests Users Attributes Meta

File xsa314.patch of Package xen.22546

From ab49f005f7d01d4004d76f2e295d31aca7d4f93a Mon Sep 17 00:00:00 2001
From: Julien Grall <jgrall@amazon.com>
Date: Thu, 20 Feb 2020 20:54:40 +0000
Subject: [PATCH] xen/rwlock: Add missing memory barrier in the unlock path of
 rwlock

The rwlock unlock paths are using atomic_sub() to release the lock.
However the implementation of atomic_sub() rightfully doesn't contain a
memory barrier. On Arm, this means a processor is allowed to re-order
the memory access with the preceeding access.

In other words, the unlock may be seen by another processor before all
the memory accesses within the "critical" section.

The rwlock paths already contains barrier indirectly, but they are not
very useful without the counterpart in the unlock paths.

The memory barriers are not necessary on x86 because loads/stores are
not re-ordered with lock instructions.

So add arch_lock_release_barrier() in the unlock paths that will only
add memory barrier on Arm.

Take the opportunity to document each lock paths explaining why a
barrier is not necessary.

This is XSA-314.

Signed-off-by: Julien Grall <jgrall@amazon.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>

---
 xen/include/xen/rwlock.h | 29 ++++++++++++++++++++++++++++-
 1 file changed, 28 insertions(+), 1 deletion(-)

--- a/xen/include/xen/rwlock.h
+++ b/xen/include/xen/rwlock.h
@@ -47,6 +47,10 @@ static inline int _read_trylock(rwlock_t
     if ( likely(!(cnts & _QW_WMASK)) )
     {
         cnts = (u32)atomic_add_return(_QR_BIAS, &lock->cnts);
+        /*
+         * atomic_add_return() is a full barrier so no need for an
+         * arch_lock_acquire_barrier().
+         */
         if ( likely(!(cnts & _QW_WMASK)) )
             return 1;
         atomic_sub(_QR_BIAS, &lock->cnts);
@@ -63,11 +67,19 @@ static inline void _read_lock(rwlock_t *
     u32 cnts;
 
     cnts = atomic_add_return(_QR_BIAS, &lock->cnts);
+    /*
+     * atomic_add_return() is a full barrier so no need for an
+     * arch_lock_acquire_barrier().
+     */
     if ( likely(!(cnts & _QW_WMASK)) )
         return;
 
     /* The slowpath will decrement the reader count, if necessary. */
     queue_read_lock_slowpath(lock);
+    /*
+     * queue_read_lock_slowpath() is using spinlock and therefore is a
+     * full barrier. So no need for an arch_lock_acquire_barrier().
+     */
 }
 
 static inline void _read_lock_irq(rwlock_t *lock)
@@ -91,6 +103,7 @@ static inline unsigned long _read_lock_i
  */
 static inline void _read_unlock(rwlock_t *lock)
 {
+    arch_lock_release_barrier();
     /*
      * Atomically decrement the reader count
      */
@@ -120,11 +133,20 @@ static inline int _rw_is_locked(rwlock_t
  */
 static inline void _write_lock(rwlock_t *lock)
 {
-    /* Optimize for the unfair lock case where the fair flag is 0. */
+    /*
+     * Optimize for the unfair lock case where the fair flag is 0.
+     *
+     * atomic_cmpxchg() is a full barrier so no need for an
+     * arch_lock_acquire_barrier().
+     */
     if ( atomic_cmpxchg(&lock->cnts, 0, _QW_LOCKED) == 0 )
         return;
 
     queue_write_lock_slowpath(lock);
+    /*
+     * queue_write_lock_slowpath() is using spinlock and therefore is a
+     * full barrier. So no need for an arch_lock_acquire_barrier().
+     */
 }
 
 static inline void _write_lock_irq(rwlock_t *lock)
@@ -156,11 +178,16 @@ static inline int _write_trylock(rwlock_
     if ( unlikely(cnts) )
         return 0;
 
+    /*
+     * atomic_cmpxchg() is a full barrier so no need for an
+     * arch_lock_acquire_barrier().
+     */
     return likely(atomic_cmpxchg(&lock->cnts, 0, _QW_LOCKED) == 0);
 }
 
 static inline void _write_unlock(rwlock_t *lock)
 {
+    arch_lock_release_barrier();
     /*
      * If the writer field is atomic, it can be cleared directly.
      * Otherwise, an atomic subtraction will be used to clear it.

Places

File xsa314.patch of Package xen.22546

Places