Merge patch series "fs: add i_state helpers"

Christian Brauner <brauner@kernel.org> says:

I've recently looked for some free space in struct inode again because
of some exec kerfuffle we had and while my idea didn't turn into
anything I noticed that we often waste bytes when using wait bit
operations. So I set out to switch that to another mechanism that would
allow us to free up bytes. So this is an attempt to turn i_state from an
unsigned long into an u32 using the individual bytes of i_state as
addresses for the wait var event mechanism (Thanks to Linus for that idea.).

This survives LTP, xfstests on various filesystems, and will-it-scale.

* patches from https://lore.kernel.org/r/20240823-work-i_state-v3-1-5cd5fd207a57@kernel.org:
  inode: make i_state a u32
  inode: port __I_LRU_ISOLATING to var event
  inode: port __I_NEW to var event
  inode: port __I_SYNC to var event
  fs: reorder i_state bits
  fs: add i_state helpers

Link: https://lore.kernel.org/r/20240823-work-i_state-v3-1-5cd5fd207a57@kernel.org
Signed-off-by: Christian Brauner <brauner@kernel.org>

1 files changed, 54 insertions, 16 deletions

diff --git a/fs/inode.c b/fs/inode.c
index ba1645a09603..aacd05749c1f 100644
--- a/fs/inode.c
+++ b/fs/inode.c
@@ -472,6 +472,17 @@ static void __inode_add_lru(struct inode *inode, bool rotate)
 		inode->i_state |= I_REFERENCED;
 }
 
+struct wait_queue_head *inode_bit_waitqueue(struct wait_bit_queue_entry *wqe,
+					    struct inode *inode, u32 bit)
+{
+        void *bit_address;
+
+        bit_address = inode_state_wait_address(inode, bit);
+        init_wait_var_entry(wqe, bit_address, 0);
+        return __var_waitqueue(bit_address);
+}
+EXPORT_SYMBOL(inode_bit_waitqueue);
+
 /*
  * Add inode to LRU if needed (inode is unused and clean).
  *
@@ -500,24 +511,35 @@ static void inode_unpin_lru_isolating(struct inode *inode)
 	spin_lock(&inode->i_lock);
 	WARN_ON(!(inode->i_state & I_LRU_ISOLATING));
 	inode->i_state &= ~I_LRU_ISOLATING;
-	smp_mb();
-	wake_up_bit(&inode->i_state, __I_LRU_ISOLATING);
+	/* Called with inode->i_lock which ensures memory ordering. */
+	inode_wake_up_bit(inode, __I_LRU_ISOLATING);
 	spin_unlock(&inode->i_lock);
 }
 
 static void inode_wait_for_lru_isolating(struct inode *inode)
 {
+	struct wait_bit_queue_entry wqe;
+	struct wait_queue_head *wq_head;
+
 	lockdep_assert_held(&inode->i_lock);
-	if (inode->i_state & I_LRU_ISOLATING) {
-		DEFINE_WAIT_BIT(wq, &inode->i_state, __I_LRU_ISOLATING);
-		wait_queue_head_t *wqh;
+	if (!(inode->i_state & I_LRU_ISOLATING))
+		return;
 
-		wqh = bit_waitqueue(&inode->i_state, __I_LRU_ISOLATING);
+	wq_head = inode_bit_waitqueue(&wqe, inode, __I_LRU_ISOLATING);
+	for (;;) {
+		prepare_to_wait_event(wq_head, &wqe.wq_entry, TASK_UNINTERRUPTIBLE);
+		/*
+		 * Checking I_LRU_ISOLATING with inode->i_lock guarantees
+		 * memory ordering.
+		 */
+		if (!(inode->i_state & I_LRU_ISOLATING))
+			break;
 		spin_unlock(&inode->i_lock);
-		__wait_on_bit(wqh, &wq, bit_wait, TASK_UNINTERRUPTIBLE);
+		schedule();
 		spin_lock(&inode->i_lock);
-		WARN_ON(inode->i_state & I_LRU_ISOLATING);
 	}
+	finish_wait(wq_head, &wqe.wq_entry);
+	WARN_ON(inode->i_state & I_LRU_ISOLATING);
 }
 
 /**
@@ -723,7 +745,13 @@ static void evict(struct inode *inode)
 	 * used as an indicator whether blocking on it is safe.
 	 */
 	spin_lock(&inode->i_lock);
-	wake_up_bit(&inode->i_state, __I_NEW);
+	/*
+	 * Pairs with the barrier in prepare_to_wait_event() to make sure
+	 * ___wait_var_event() either sees the bit cleared or
+	 * waitqueue_active() check in wake_up_var() sees the waiter.
+	 */
+	smp_mb();
+	inode_wake_up_bit(inode, __I_NEW);
 	BUG_ON(inode->i_state != (I_FREEING | I_CLEAR));
 	spin_unlock(&inode->i_lock);
 
@@ -1135,8 +1163,13 @@ void unlock_new_inode(struct inode *inode)
 	spin_lock(&inode->i_lock);
 	WARN_ON(!(inode->i_state & I_NEW));
 	inode->i_state &= ~I_NEW & ~I_CREATING;
+	/*
+	 * Pairs with the barrier in prepare_to_wait_event() to make sure
+	 * ___wait_var_event() either sees the bit cleared or
+	 * waitqueue_active() check in wake_up_var() sees the waiter.
+	 */
 	smp_mb();
-	wake_up_bit(&inode->i_state, __I_NEW);
+	inode_wake_up_bit(inode, __I_NEW);
 	spin_unlock(&inode->i_lock);
 }
 EXPORT_SYMBOL(unlock_new_inode);
@@ -1147,8 +1180,13 @@ void discard_new_inode(struct inode *inode)
 	spin_lock(&inode->i_lock);
 	WARN_ON(!(inode->i_state & I_NEW));
 	inode->i_state &= ~I_NEW;
+	/*
+	 * Pairs with the barrier in prepare_to_wait_event() to make sure
+	 * ___wait_var_event() either sees the bit cleared or
+	 * waitqueue_active() check in wake_up_var() sees the waiter.
+	 */
 	smp_mb();
-	wake_up_bit(&inode->i_state, __I_NEW);
+	inode_wake_up_bit(inode, __I_NEW);
 	spin_unlock(&inode->i_lock);
 	iput(inode);
 }
@@ -2337,8 +2375,8 @@ EXPORT_SYMBOL(inode_needs_sync);
  */
 static void __wait_on_freeing_inode(struct inode *inode, bool is_inode_hash_locked)
 {
-	wait_queue_head_t *wq;
-	DEFINE_WAIT_BIT(wait, &inode->i_state, __I_NEW);
+	struct wait_bit_queue_entry wqe;
+	struct wait_queue_head *wq_head;
 
 	/*
 	 * Handle racing against evict(), see that routine for more details.
@@ -2349,14 +2387,14 @@ static void __wait_on_freeing_inode(struct inode *inode, bool is_inode_hash_lock
 		return;
 	}
 
-	wq = bit_waitqueue(&inode->i_state, __I_NEW);
-	prepare_to_wait(wq, &wait.wq_entry, TASK_UNINTERRUPTIBLE);
+	wq_head = inode_bit_waitqueue(&wqe, inode, __I_NEW);
+	prepare_to_wait_event(wq_head, &wqe.wq_entry, TASK_UNINTERRUPTIBLE);
 	spin_unlock(&inode->i_lock);
 	rcu_read_unlock();
 	if (is_inode_hash_locked)
 		spin_unlock(&inode_hash_lock);
 	schedule();
-	finish_wait(wq, &wait.wq_entry);
+	finish_wait(wq_head, &wqe.wq_entry);
 	if (is_inode_hash_locked)
 		spin_lock(&inode_hash_lock);
 	rcu_read_lock();