From c2a7d2a115525d3501d38e23d24875a79a07e15e Mon Sep 17 00:00:00 2001
From: Dan Williams <dan.j.williams@intel.com>
Date: Fri, 13 Jul 2018 21:50:16 -0700
Subject: [PATCH] filesystem-dax: Introduce dax_lock_mapping_entry()
Git-commit: c2a7d2a115525d3501d38e23d24875a79a07e15e
Patch-mainline: v4.19-rc1
References: bsc#1107783

In preparation for implementing support for memory poison (media error)
handling via dax mappings, implement a lock_page() equivalent. Poison
error handling requires rmap and needs guarantees that the page->mapping
association is maintained / valid (inode not freed) for the duration of
the lookup.

In the device-dax case it is sufficient to simply hold a dev_pagemap
reference. In the filesystem-dax case we need to use the entry lock.

Export the entry lock via dax_lock_mapping_entry() that uses
rcu_read_lock() to protect against the inode being freed, and
revalidates the page->mapping association under xa_lock().

Cc: Christoph Hellwig <hch@lst.de>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Ross Zwisler <ross.zwisler@linux.intel.com>
Cc: Jan Kara <jack@suse.cz>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Acked-by: Jan Kara <jack@suse.cz>

---
 fs/dax.c            |  109 +++++++++++++++++++++++++++++++++++++++++++++++++---
 include/linux/dax.h |   13 ++++++
 2 files changed, 116 insertions(+), 6 deletions(-)

--- a/fs/dax.c
+++ b/fs/dax.c
@@ -230,8 +230,8 @@ static inline void *unlock_slot(struct a
  *
  * The function must be called with mapping->tree_lock held.
  */
-static void *get_unlocked_mapping_entry(struct address_space *mapping,
-					pgoff_t index, void ***slotp)
+static void *__get_unlocked_mapping_entry(struct address_space *mapping,
+		pgoff_t index, void ***slotp, bool (*wait_fn)(void))
 {
 	void *entry, **slot;
 	struct wait_exceptional_entry_queue ewait;
@@ -241,6 +241,8 @@ static void *get_unlocked_mapping_entry(
 	ewait.wait.func = wake_exceptional_entry_func;
 
 	for (;;) {
+		bool revalidate;
+
 		entry = __radix_tree_lookup(&mapping->page_tree, index, NULL,
 					  &slot);
 		if (!entry ||
@@ -255,14 +257,31 @@ static void *get_unlocked_mapping_entry(
 		prepare_to_wait_exclusive(wq, &ewait.wait,
 					  TASK_UNINTERRUPTIBLE);
 		spin_unlock_irq(&mapping->tree_lock);
-		schedule();
+		revalidate = wait_fn();
 		finish_wait(wq, &ewait.wait);
 		spin_lock_irq(&mapping->tree_lock);
+		if (revalidate)
+			return ERR_PTR(-EAGAIN);
 	}
 }
 
-static void dax_unlock_mapping_entry(struct address_space *mapping,
-				     pgoff_t index)
+static bool entry_wait(void)
+{
+	schedule();
+	/*
+	 * Never return an ERR_PTR() from
+	 * __get_unlocked_mapping_entry(), just keep looping.
+	 */
+	return false;
+}
+
+static void *get_unlocked_mapping_entry(struct address_space *mapping,
+		pgoff_t index, void ***slotp)
+{
+	return __get_unlocked_mapping_entry(mapping, index, slotp, entry_wait);
+}
+
+static void unlock_mapping_entry(struct address_space *mapping, pgoff_t index)
 {
 	void *entry, **slot;
 
@@ -281,7 +300,7 @@ static void dax_unlock_mapping_entry(str
 static void put_locked_mapping_entry(struct address_space *mapping,
 		pgoff_t index)
 {
-	dax_unlock_mapping_entry(mapping, index);
+	unlock_mapping_entry(mapping, index);
 }
 
 /*
@@ -378,6 +397,84 @@ static struct page *dax_busy_page(void *
 	return NULL;
 }
 
+static bool entry_wait_revalidate(void)
+{
+	rcu_read_unlock();
+	schedule();
+	rcu_read_lock();
+
+	/*
+	 * Tell __get_unlocked_mapping_entry() to take a break, we need
+	 * to revalidate page->mapping after dropping locks
+	 */
+	return true;
+}
+
+bool dax_lock_mapping_entry(struct page *page)
+{
+	pgoff_t index;
+	struct inode *inode;
+	bool did_lock = false;
+	void *entry = NULL, **slot;
+	struct address_space *mapping;
+
+	rcu_read_lock();
+	for (;;) {
+		mapping = READ_ONCE(page->mapping);
+
+		if (!dax_mapping(mapping))
+			break;
+
+		/*
+		 * In the device-dax case there's no need to lock, a
+		 * struct dev_pagemap pin is sufficient to keep the
+		 * inode alive, and we assume we have dev_pagemap pin
+		 * otherwise we would not have a valid pfn_to_page()
+		 * translation.
+		 */
+		inode = mapping->host;
+		if (S_ISCHR(inode->i_mode)) {
+			did_lock = true;
+			break;
+		}
+
+		spin_lock_irq(&mapping->tree_lock);
+		if (mapping != page->mapping) {
+			spin_unlock_irq(&mapping->tree_lock);
+			continue;
+		}
+		index = page->index;
+
+		entry = __get_unlocked_mapping_entry(mapping, index, &slot,
+				entry_wait_revalidate);
+		if (!entry) {
+			spin_unlock_irq(&mapping->tree_lock);
+			break;
+		} else if (IS_ERR(entry)) {
+			WARN_ON_ONCE(PTR_ERR(entry) != -EAGAIN);
+			continue;
+		}
+		lock_slot(mapping, slot);
+		did_lock = true;
+		spin_unlock_irq(&mapping->tree_lock);
+		break;
+	}
+	rcu_read_unlock();
+
+	return did_lock;
+}
+
+void dax_unlock_mapping_entry(struct page *page)
+{
+	struct address_space *mapping = page->mapping;
+	struct inode *inode = mapping->host;
+
+	if (S_ISCHR(inode->i_mode))
+		return;
+
+	unlock_mapping_entry(mapping, page->index);
+}
+
 /*
  * Find radix tree entry at given index. If it points to an exceptional entry,
  * return it with the radix tree entry locked. If the radix tree doesn't
--- a/include/linux/dax.h
+++ b/include/linux/dax.h
@@ -64,6 +64,8 @@ int dax_writeback_mapping_range(struct a
 		struct block_device *bdev, struct writeback_control *wbc);
 
 struct page *dax_layout_busy_page(struct address_space *mapping);
+bool dax_lock_mapping_entry(struct page *page);
+void dax_unlock_mapping_entry(struct page *page);
 #else
 static inline bool bdev_dax_supported(struct block_device *bdev,
 		int blocksize)
@@ -95,6 +97,17 @@ static inline int dax_writeback_mapping_
 {
 	return -EOPNOTSUPP;
 }
+
+static inline bool dax_lock_mapping_entry(struct page *page)
+{
+	if (IS_DAX(page->mapping->host))
+		return true;
+	return false;
+}
+
+static inline void dax_unlock_mapping_entry(struct page *page)
+{
+}
 #endif
 
 int dax_read_lock(void);