[PATCH] futex: Fix mm reuse handling for FUT_OFF_MMSHARED

[Jann Horn]

A FUT_OFF_MMSHARED futex is a shared futex that refers to an MM.
It is possible for a process to wait on a shared futex with a different MM
because a FUT_OFF_INODE waiter can be requeued onto a FUT_OFF_MMSHARED
futex by another process.
This can cause FUT_OFF_MMSHARED waiters on a freed MM to consume
wakeups intended for a newly allocated MM at the same address.

Fix it by keying FUT_OFF_MMSHARED using a unique 64-bit per-MM ID.
Leave private futexes as before to avoid influencing the performance of the
hotpath.

(Multi-threaded processes typically implicitly use FUT_OFF_MMSHARED by
setting clear_child_tid such that it points into anonymous memory, which
causes mm_release() in a multi-threaded mm to perform FUTEX_WAKE.)

Cc: stable@xxxxxxxxxxxxxxx
Fixes: 222993395ed3 ("futex: Remove pointless mmgrap() + mmdrop()")
Closes: https://lore.kernel.org/r/CAG48ez0dLBpc3QtbAhMMVNHwHL94iHh2G+h-=BVFR4dDuzZr1g@xxxxxxxxxxxxxx/
Signed-off-by: Jann Horn <jannh@xxxxxxxxxx>
---
I've added a stable tag since, due to clear_child_tid, this can
probably be used to prevent pthread_join() from completing in
another process or to brute-force the addresses of pthread
instances in other processes.
---
 include/linux/futex.h       | 19 ++++++++---------
 include/linux/futex_types.h |  2 ++
 kernel/futex/core.c         | 51 +++++++++++++++++++++++++--------------------
 3 files changed, 39 insertions(+), 33 deletions(-)

diff --git a/include/linux/futex.h b/include/linux/futex.h
index 18ed18d5cbc1..7eb645e49dd4 100644
--- a/include/linux/futex.h
+++ b/include/linux/futex.h
@@ -23,28 +23,32 @@ struct task_struct;
  *  01 : Shared futex (PTHREAD_PROCESS_SHARED)
  *	mapped on a file (reference on the underlying inode)
  *  10 : Shared futex (PTHREAD_PROCESS_SHARED)
- *       (but private mapping on an mm, and reference taken on it)
+ *       (but private mapping on an mm)
 */
 
-#define FUT_OFF_INODE    1 /* We set bit 0 if key has a reference on inode */
-#define FUT_OFF_MMSHARED 2 /* We set bit 1 if key has a reference on mm */
+#define FUT_OFF_INODE    1 /* We set bit 0 if shared key identifies an inode */
+#define FUT_OFF_MMSHARED 2 /* We set bit 1 if shared key identifies an mm */
 
 union futex_key {
+	/* For FUT_OFF_INODE */
 	struct {
 		u64 i_seq;
 		unsigned long pgoff;
 		unsigned int offset;
 		/* unsigned int node; */
 	} shared;
+
+	/* For FUT_OFF_MMSHARED or private */
 	struct {
 		union {
-			struct mm_struct *mm;
-			u64 __tmp;
+			struct mm_struct *mm; /* for private */
+			u64 mm_seq; /* for FUT_OFF_MMSHARED */
 		};
 		unsigned long address;
 		unsigned int offset;
 		/* unsigned int node; */
 	} private;
+
 	struct {
 		u64 ptr;
 		unsigned long word;
@@ -152,11 +156,6 @@ static inline void futex_set_vdso_cs_range(struct futex_mm_data *fd, unsigned in
 static inline void futex_fixup_robust_unlock(struct pt_regs *regs) { }
 #endif /* !CONFIG_FUTEX_ROBUST_UNLOCK */
 
-
-#if defined(CONFIG_FUTEX_PRIVATE_HASH) || defined(CONFIG_FUTEX_ROBUST_UNLOCK)
 void futex_mm_init(struct mm_struct *mm);
-#else
-static inline void futex_mm_init(struct mm_struct *mm) { }
-#endif
 
 #endif /* _LINUX_FUTEX_H */
diff --git a/include/linux/futex_types.h b/include/linux/futex_types.h
index d320c0571f0c..34b16f2828e7 100644
--- a/include/linux/futex_types.h
+++ b/include/linux/futex_types.h
@@ -85,10 +85,12 @@ struct futex_unlock_cs_ranges { };
  * struct futex_mm_data - Futex related per MM data
  * @phash:	Futex private hash related data
  * @unlock:	Futex unlock VDSO critical sections
+ * @unique_id:	Unique ID of the MM, for FUT_OFF_MMSHARED futexes
  */
 struct futex_mm_data {
 	struct futex_mm_phash		phash;
 	struct futex_unlock_cs_ranges	unlock;
+	atomic64_t			unique_id;
 };
 #else  /* CONFIG_FUTEX */
 struct futex_sched_data { };
diff --git a/kernel/futex/core.c b/kernel/futex/core.c
index a061f54b606d..3593d8a328e9 100644
--- a/kernel/futex/core.c
+++ b/kernel/futex/core.c
@@ -437,45 +437,51 @@ struct hrtimer_sleeper *futex_setup_timer(ktime_t *time, struct hrtimer_sleeper
 }
 
 /*
- * Generate a machine wide unique identifier for this inode.
+ * Generate a machine wide unique identifier for this object (inode or mm).
  *
  * This relies on u64 not wrapping in the life-time of the machine; which with
  * 1ns resolution means almost 585 years.
- *
- * This further relies on the fact that a well formed program will not unmap
- * the file while it has a (shared) futex waiting on it. This mapping will have
- * a file reference which pins the mount and inode.
- *
- * If for some reason an inode gets evicted and read back in again, it will get
- * a new sequence number and will _NOT_ match, even though it is the exact same
- * file.
- *
- * It is important that futex_match() will never have a false-positive, esp.
- * for PI futexes that can mess up the state. The above argues that false-negatives
- * are only possible for malformed programs.
  */
-static u64 get_inode_sequence_number(struct inode *inode)
+static u64 get_object_id(atomic64_t *object_seq)
 {
-	static atomic64_t i_seq;
+	static atomic64_t last_assigned_seq;
 	u64 old;
 
-	/* Does the inode already have a sequence number? */
-	old = atomic64_read(&inode->i_sequence);
+	/* Does the object already have a sequence number? */
+	old = atomic64_read(object_seq);
 	if (likely(old))
 		return old;
 
 	for (;;) {
-		u64 new = atomic64_inc_return(&i_seq);
+		u64 new = atomic64_inc_return(&last_assigned_seq);
 		if (WARN_ON_ONCE(!new))
 			continue;
 
 		old = 0;
-		if (!atomic64_try_cmpxchg_relaxed(&inode->i_sequence, &old, new))
+		if (!atomic64_try_cmpxchg_relaxed(object_seq, &old, new))
 			return old;
 		return new;
 	}
 }
 
+/*
+ * This relies on the fact that a well formed program will not unmap
+ * the file while it has a (shared) futex waiting on it. This mapping will have
+ * a file reference which pins the mount and inode.
+ *
+ * If for some reason an inode gets evicted and read back in again, it will get
+ * a new sequence number and will _NOT_ match, even though it is the exact same
+ * file.
+ *
+ * It is important that futex_match() will never have a false-positive, esp.
+ * for PI futexes that can mess up the state. The above argues that false-negatives
+ * are only possible for malformed programs.
+ */
+static u64 get_inode_sequence_number(struct inode *inode)
+{
+	return get_object_id(&inode->i_sequence);
+}
+
 /**
  * get_futex_key() - Get parameters which are the keys for a futex
  * @uaddr:	virtual address of the futex
@@ -681,8 +687,8 @@ int get_futex_key(u32 __user *uaddr, unsigned int flags, union futex_key *key,
 			goto out;
 		}
 
-		key->both.offset |= FUT_OFF_MMSHARED; /* ref taken on mm */
-		key->private.mm = mm;
+		key->both.offset |= FUT_OFF_MMSHARED;
+		key->private.mm_seq = get_object_id(&mm->futex.unique_id);
 		key->private.address = address;
 
 	} else {
@@ -2053,13 +2059,12 @@ static void futex_robust_unlock_init_mm(struct futex_mm_data *fd)
 static inline void futex_robust_unlock_init_mm(struct futex_mm_data *fd) { }
 #endif /* !CONFIG_FUTEX_ROBUST_UNLOCK */
 
-#if defined(CONFIG_FUTEX_PRIVATE_HASH) || defined(CONFIG_FUTEX_ROBUST_UNLOCK)
 void futex_mm_init(struct mm_struct *mm)
 {
+	atomic64_set(&mm->futex.unique_id, 0);
 	futex_hash_init_mm(&mm->futex);
 	futex_robust_unlock_init_mm(&mm->futex);
 }
-#endif
 
 int futex_hash_prctl(unsigned long arg2, unsigned long arg3, unsigned long arg4)
 {

---
base-commit: 6f8319e3e9a44dd537d17f41565a8453c560a581
change-id: 20260929-futex-mmshared-fix-7c8635797e15

Best regards,
--  
Jann Horn <jannh@xxxxxxxxxx>