1 files changed, 224 insertions, 241 deletions

diff --git a/fs/nfs/write.c b/fs/nfs/write.c
index b1af5dee5e0a..babebbccae2a 100644
--- a/fs/nfs/write.c
+++ b/fs/nfs/write.c
@@ -102,10 +102,8 @@ static struct nfs_pgio_header *nfs_writehdr_alloc(void)
 {
 	struct nfs_pgio_header *p = mempool_alloc(nfs_wdata_mempool, GFP_NOIO);
 
-	if (p) {
-		memset(p, 0, sizeof(*p));
-		p->rw_mode = FMODE_WRITE;
-	}
+	memset(p, 0, sizeof(*p));
+	p->rw_mode = FMODE_WRITE;
 	return p;
 }
 
@@ -147,11 +145,12 @@ static void nfs_io_completion_put(struct nfs_io_completion *ioc)
 		kref_put(&ioc->refcount, nfs_io_completion_release);
 }
 
-static void nfs_context_set_write_error(struct nfs_open_context *ctx, int error)
+static struct nfs_page *
+nfs_page_private_request(struct page *page)
 {
-	ctx->error = error;
-	smp_wmb();
-	set_bit(NFS_CONTEXT_ERROR_WRITE, &ctx->flags);
+	if (!PagePrivate(page))
+		return NULL;
+	return (struct nfs_page *)page_private(page);
 }
 
 /*
@@ -162,21 +161,41 @@ static void nfs_context_set_write_error(struct nfs_open_context *ctx, int error)
  * returns matching head request with reference held, or NULL if not found.
  */
 static struct nfs_page *
-nfs_page_find_head_request_locked(struct nfs_inode *nfsi, struct page *page)
+nfs_page_find_private_request(struct page *page)
 {
-	struct nfs_page *req = NULL;
-
-	if (PagePrivate(page))
-		req = (struct nfs_page *)page_private(page);
-	else if (unlikely(PageSwapCache(page)))
-		req = nfs_page_search_commits_for_head_request_locked(nfsi,
-			page);
+	struct address_space *mapping = page_file_mapping(page);
+	struct nfs_page *req;
 
+	if (!PagePrivate(page))
+		return NULL;
+	spin_lock(&mapping->private_lock);
+	req = nfs_page_private_request(page);
 	if (req) {
 		WARN_ON_ONCE(req->wb_head != req);
 		kref_get(&req->wb_kref);
 	}
+	spin_unlock(&mapping->private_lock);
+	return req;
+}
 
+static struct nfs_page *
+nfs_page_find_swap_request(struct page *page)
+{
+	struct inode *inode = page_file_mapping(page)->host;
+	struct nfs_inode *nfsi = NFS_I(inode);
+	struct nfs_page *req = NULL;
+	if (!PageSwapCache(page))
+		return NULL;
+	mutex_lock(&nfsi->commit_mutex);
+	if (PageSwapCache(page)) {
+		req = nfs_page_search_commits_for_head_request_locked(nfsi,
+			page);
+		if (req) {
+			WARN_ON_ONCE(req->wb_head != req);
+			kref_get(&req->wb_kref);
+		}
+	}
+	mutex_unlock(&nfsi->commit_mutex);
 	return req;
 }
 
@@ -187,12 +206,11 @@ nfs_page_find_head_request_locked(struct nfs_inode *nfsi, struct page *page)
  */
 static struct nfs_page *nfs_page_find_head_request(struct page *page)
 {
-	struct inode *inode = page_file_mapping(page)->host;
-	struct nfs_page *req = NULL;
+	struct nfs_page *req;
 
-	spin_lock(&inode->i_lock);
-	req = nfs_page_find_head_request_locked(NFS_I(inode), page);
-	spin_unlock(&inode->i_lock);
+	req = nfs_page_find_private_request(page);
+	if (!req)
+		req = nfs_page_find_swap_request(page);
 	return req;
 }
 
@@ -241,9 +259,6 @@ nfs_page_group_search_locked(struct nfs_page *head, unsigned int page_offset)
 {
 	struct nfs_page *req;
 
-	WARN_ON_ONCE(head != head->wb_head);
-	WARN_ON_ONCE(!test_bit(PG_HEADLOCK, &head->wb_head->wb_flags));
-
 	req = head;
 	do {
 		if (page_offset >= req->wb_pgbase &&
@@ -269,20 +284,17 @@ static bool nfs_page_group_covers_page(struct nfs_page *req)
 	unsigned int pos = 0;
 	unsigned int len = nfs_page_length(req->wb_page);
 
-	nfs_page_group_lock(req, false);
+	nfs_page_group_lock(req);
 
-	do {
+	for (;;) {
 		tmp = nfs_page_group_search_locked(req->wb_head, pos);
-		if (tmp) {
-			/* no way this should happen */
-			WARN_ON_ONCE(tmp->wb_pgbase != pos);
-			pos += tmp->wb_bytes - (pos - tmp->wb_pgbase);
-		}
-	} while (tmp && pos < len);
+		if (!tmp)
+			break;
+		pos = tmp->wb_pgbase + tmp->wb_bytes;
+	}
 
 	nfs_page_group_unlock(req);
-	WARN_ON_ONCE(pos > len);
-	return pos == len;
+	return pos >= len;
 }
 
 /* We can set the PG_uptodate flag if we see that a write request
@@ -333,8 +345,11 @@ static void nfs_end_page_writeback(struct nfs_page *req)
 {
 	struct inode *inode = page_file_mapping(req->wb_page)->host;
 	struct nfs_server *nfss = NFS_SERVER(inode);
+	bool is_done;
 
-	if (!nfs_page_group_sync_on_bit(req, PG_WB_END))
+	is_done = nfs_page_group_sync_on_bit(req, PG_WB_END);
+	nfs_unlock_request(req);
+	if (!is_done)
 		return;
 
 	end_page_writeback(req->wb_page);
@@ -342,22 +357,6 @@ static void nfs_end_page_writeback(struct nfs_page *req)
 		clear_bdi_congested(inode_to_bdi(inode), BLK_RW_ASYNC);
 }
 
-
-/* nfs_page_group_clear_bits
- *   @req - an nfs request
- * clears all page group related bits from @req
- */
-static void
-nfs_page_group_clear_bits(struct nfs_page *req)
-{
-	clear_bit(PG_TEARDOWN, &req->wb_flags);
-	clear_bit(PG_UNLOCKPAGE, &req->wb_flags);
-	clear_bit(PG_UPTODATE, &req->wb_flags);
-	clear_bit(PG_WB_END, &req->wb_flags);
-	clear_bit(PG_REMOVE, &req->wb_flags);
-}
-
-
 /*
  * nfs_unroll_locks_and_wait -  unlock all newly locked reqs and wait on @req
  *
@@ -366,43 +365,24 @@ nfs_page_group_clear_bits(struct nfs_page *req)
  * @inode - inode associated with request page group, must be holding inode lock
  * @head  - head request of page group, must be holding head lock
  * @req   - request that couldn't lock and needs to wait on the req bit lock
- * @nonblock - if true, don't actually wait
  *
- * NOTE: this must be called holding page_group bit lock and inode spin lock
- *       and BOTH will be released before returning.
+ * NOTE: this must be called holding page_group bit lock
+ *       which will be released before returning.
  *
  * returns 0 on success, < 0 on error.
  */
-static int
-nfs_unroll_locks_and_wait(struct inode *inode, struct nfs_page *head,
-			  struct nfs_page *req, bool nonblock)
-	__releases(&inode->i_lock)
+static void
+nfs_unroll_locks(struct inode *inode, struct nfs_page *head,
+			  struct nfs_page *req)
 {
 	struct nfs_page *tmp;
-	int ret;
 
 	/* relinquish all the locks successfully grabbed this run */
-	for (tmp = head ; tmp != req; tmp = tmp->wb_this_page)
-		nfs_unlock_request(tmp);
-
-	WARN_ON_ONCE(test_bit(PG_TEARDOWN, &req->wb_flags));
-
-	/* grab a ref on the request that will be waited on */
-	kref_get(&req->wb_kref);
-
-	nfs_page_group_unlock(head);
-	spin_unlock(&inode->i_lock);
-
-	/* release ref from nfs_page_find_head_request_locked */
-	nfs_release_request(head);
-
-	if (!nonblock)
-		ret = nfs_wait_on_request(req);
-	else
-		ret = -EAGAIN;
-	nfs_release_request(req);
-
-	return ret;
+	for (tmp = head->wb_this_page ; tmp != req; tmp = tmp->wb_this_page) {
+		if (!kref_read(&tmp->wb_kref))
+			continue;
+		nfs_unlock_and_release_request(tmp);
+	}
 }
 
 /*
@@ -417,7 +397,8 @@ nfs_unroll_locks_and_wait(struct inode *inode, struct nfs_page *head,
  */
 static void
 nfs_destroy_unlinked_subrequests(struct nfs_page *destroy_list,
-				 struct nfs_page *old_head)
+				 struct nfs_page *old_head,
+				 struct inode *inode)
 {
 	while (destroy_list) {
 		struct nfs_page *subreq = destroy_list;
@@ -428,33 +409,28 @@ nfs_destroy_unlinked_subrequests(struct nfs_page *destroy_list,
 		WARN_ON_ONCE(old_head != subreq->wb_head);
 
 		/* make sure old group is not used */
-		subreq->wb_head = subreq;
 		subreq->wb_this_page = subreq;
 
-		/* subreq is now totally disconnected from page group or any
-		 * write / commit lists. last chance to wake any waiters */
-		nfs_unlock_request(subreq);
+		clear_bit(PG_REMOVE, &subreq->wb_flags);
 
-		if (!test_bit(PG_TEARDOWN, &subreq->wb_flags)) {
-			/* release ref on old head request */
-			nfs_release_request(old_head);
+		/* Note: races with nfs_page_group_destroy() */
+		if (!kref_read(&subreq->wb_kref)) {
+			/* Check if we raced with nfs_page_group_destroy() */
+			if (test_and_clear_bit(PG_TEARDOWN, &subreq->wb_flags))
+				nfs_free_request(subreq);
+			continue;
+		}
 
-			nfs_page_group_clear_bits(subreq);
+		subreq->wb_head = subreq;
 
-			/* release the PG_INODE_REF reference */
-			if (test_and_clear_bit(PG_INODE_REF, &subreq->wb_flags))
-				nfs_release_request(subreq);
-			else
-				WARN_ON_ONCE(1);
-		} else {
-			WARN_ON_ONCE(test_bit(PG_CLEAN, &subreq->wb_flags));
-			/* zombie requests have already released the last
-			 * reference and were waiting on the rest of the
-			 * group to complete. Since it's no longer part of a
-			 * group, simply free the request */
-			nfs_page_group_clear_bits(subreq);
-			nfs_free_request(subreq);
+		if (test_and_clear_bit(PG_INODE_REF, &subreq->wb_flags)) {
+			nfs_release_request(subreq);
+			atomic_long_dec(&NFS_I(inode)->nrequests);
 		}
+
+		/* subreq is now totally disconnected from page group or any
+		 * write / commit lists. last chance to wake any waiters */
+		nfs_unlock_and_release_request(subreq);
 	}
 }
 
@@ -464,7 +440,6 @@ nfs_destroy_unlinked_subrequests(struct nfs_page *destroy_list,
  *                              operations for this page.
  *
  * @page - the page used to lookup the "page group" of nfs_page structures
- * @nonblock - if true, don't block waiting for request locks
  *
  * This function joins all sub requests to the head request by first
  * locking all requests in the group, cancelling any pending operations
@@ -478,7 +453,7 @@ nfs_destroy_unlinked_subrequests(struct nfs_page *destroy_list,
  * error was encountered.
  */
 static struct nfs_page *
-nfs_lock_and_join_requests(struct page *page, bool nonblock)
+nfs_lock_and_join_requests(struct page *page)
 {
 	struct inode *inode = page_file_mapping(page)->host;
 	struct nfs_page *head, *subreq;
@@ -487,43 +462,63 @@ nfs_lock_and_join_requests(struct page *page, bool nonblock)
 	int ret;
 
 try_again:
-	total_bytes = 0;
-
-	WARN_ON_ONCE(destroy_list);
-
-	spin_lock(&inode->i_lock);
-
 	/*
 	 * A reference is taken only on the head request which acts as a
 	 * reference to the whole page group - the group will not be destroyed
 	 * until the head reference is released.
 	 */
-	head = nfs_page_find_head_request_locked(NFS_I(inode), page);
-
-	if (!head) {
-		spin_unlock(&inode->i_lock);
+	head = nfs_page_find_head_request(page);
+	if (!head)
 		return NULL;
-	}
 
-	/* holding inode lock, so always make a non-blocking call to try the
-	 * page group lock */
-	ret = nfs_page_group_lock(head, true);
-	if (ret < 0) {
-		spin_unlock(&inode->i_lock);
+	/* lock the page head first in order to avoid an ABBA inefficiency */
+	if (!nfs_lock_request(head)) {
+		ret = nfs_wait_on_request(head);
+		nfs_release_request(head);
+		if (ret < 0)
+			return ERR_PTR(ret);
+		goto try_again;
+	}
 
-		if (!nonblock && ret == -EAGAIN) {
-			nfs_page_group_lock_wait(head);
-			nfs_release_request(head);
-			goto try_again;
-		}
+	/* Ensure that nobody removed the request before we locked it */
+	if (head != nfs_page_private_request(page) && !PageSwapCache(page)) {
+		nfs_unlock_and_release_request(head);
+		goto try_again;
+	}
 
-		nfs_release_request(head);
+	ret = nfs_page_group_lock(head);
+	if (ret < 0) {
+		nfs_unlock_and_release_request(head);
 		return ERR_PTR(ret);
 	}
 
 	/* lock each request in the page group */
-	subreq = head;
-	do {
+	total_bytes = head->wb_bytes;
+	for (subreq = head->wb_this_page; subreq != head;
+			subreq = subreq->wb_this_page) {
+
+		if (!kref_get_unless_zero(&subreq->wb_kref)) {
+			if (subreq->wb_offset == head->wb_offset + total_bytes)
+				total_bytes += subreq->wb_bytes;
+			continue;
+		}
+
+		while (!nfs_lock_request(subreq)) {
+			/*
+			 * Unlock page to allow nfs_page_group_sync_on_bit()
+			 * to succeed
+			 */
+			nfs_page_group_unlock(head);
+			ret = nfs_wait_on_request(subreq);
+			if (!ret)
+				ret = nfs_page_group_lock(head);
+			if (ret < 0) {
+				nfs_unroll_locks(inode, head, subreq);
+				nfs_release_request(subreq);
+				nfs_unlock_and_release_request(head);
+				return ERR_PTR(ret);
+			}
+		}
 		/*
 		 * Subrequests are always contiguous, non overlapping
 		 * and in order - but may be repeated (mirrored writes).
@@ -535,24 +530,12 @@ try_again:
 			    ((subreq->wb_offset + subreq->wb_bytes) >
 			     (head->wb_offset + total_bytes)))) {
 			nfs_page_group_unlock(head);
-			spin_unlock(&inode->i_lock);
+			nfs_unroll_locks(inode, head, subreq);
+			nfs_unlock_and_release_request(subreq);
+			nfs_unlock_and_release_request(head);
 			return ERR_PTR(-EIO);
 		}
-
-		if (!nfs_lock_request(subreq)) {
-			/* releases page group bit lock and
-			 * inode spin lock and all references */
-			ret = nfs_unroll_locks_and_wait(inode, head,
-				subreq, nonblock);
-
-			if (ret == 0)
-				goto try_again;
-
-			return ERR_PTR(ret);
-		}
-
-		subreq = subreq->wb_this_page;
-	} while (subreq != head);
+	}
 
 	/* Now that all requests are locked, make sure they aren't on any list.
 	 * Commit list removal accounting is done after locks are dropped */
@@ -573,34 +556,30 @@ try_again:
 		head->wb_bytes = total_bytes;
 	}
 
-	/*
-	 * prepare head request to be added to new pgio descriptor
-	 */
-	nfs_page_group_clear_bits(head);
-
-	/*
-	 * some part of the group was still on the inode list - otherwise
-	 * the group wouldn't be involved in async write.
-	 * grab a reference for the head request, iff it needs one.
-	 */
-	if (!test_and_set_bit(PG_INODE_REF, &head->wb_flags))
+	/* Postpone destruction of this request */
+	if (test_and_clear_bit(PG_REMOVE, &head->wb_flags)) {
+		set_bit(PG_INODE_REF, &head->wb_flags);
 		kref_get(&head->wb_kref);
+		atomic_long_inc(&NFS_I(inode)->nrequests);
+	}
 
 	nfs_page_group_unlock(head);
 
-	/* drop lock to clean uprequests on destroy list */
-	spin_unlock(&inode->i_lock);
+	nfs_destroy_unlinked_subrequests(destroy_list, head, inode);
 
-	nfs_destroy_unlinked_subrequests(destroy_list, head);
+	/* Did we lose a race with nfs_inode_remove_request()? */
+	if (!(PagePrivate(page) || PageSwapCache(page))) {
+		nfs_unlock_and_release_request(head);
+		return NULL;
+	}
 
-	/* still holds ref on head from nfs_page_find_head_request_locked
+	/* still holds ref on head from nfs_page_find_head_request
 	 * and still has lock on head from lock loop */
 	return head;
 }
 
 static void nfs_write_error_remove_page(struct nfs_page *req)
 {
-	nfs_unlock_request(req);
 	nfs_end_page_writeback(req);
 	generic_error_remove_page(page_file_mapping(req->wb_page),
 				  req->wb_page);
@@ -624,12 +603,12 @@ nfs_error_is_fatal_on_server(int err)
  * May return an error if the user signalled nfs_wait_on_request().
  */
 static int nfs_page_async_flush(struct nfs_pageio_descriptor *pgio,
-				struct page *page, bool nonblock)
+				struct page *page)
 {
 	struct nfs_page *req;
 	int ret = 0;
 
-	req = nfs_lock_and_join_requests(page, nonblock);
+	req = nfs_lock_and_join_requests(page);
 	if (!req)
 		goto out;
 	ret = PTR_ERR(req);
@@ -672,7 +651,7 @@ static int nfs_do_writepage(struct page *page, struct writeback_control *wbc,
 	int ret;
 
 	nfs_pageio_cond_complete(pgio, page_index(page));
-	ret = nfs_page_async_flush(pgio, page, wbc->sync_mode == WB_SYNC_NONE);
+	ret = nfs_page_async_flush(pgio, page);
 	if (ret == -EAGAIN) {
 		redirty_page_for_writepage(wbc, page);
 		ret = 0;
@@ -759,6 +738,7 @@ out_err:
  */
 static void nfs_inode_add_request(struct inode *inode, struct nfs_page *req)
 {
+	struct address_space *mapping = page_file_mapping(req->wb_page);
 	struct nfs_inode *nfsi = NFS_I(inode);
 
 	WARN_ON_ONCE(req->wb_this_page != req);
@@ -766,27 +746,30 @@ static void nfs_inode_add_request(struct inode *inode, struct nfs_page *req)
 	/* Lock the request! */
 	nfs_lock_request(req);
 
-	spin_lock(&inode->i_lock);
-	if (!nfsi->nrequests &&
-	    NFS_PROTO(inode)->have_delegation(inode, FMODE_WRITE))
-		inode->i_version++;
 	/*
 	 * Swap-space should not get truncated. Hence no need to plug the race
 	 * with invalidate/truncate.
 	 */
+	spin_lock(&mapping->private_lock);
+	if (!nfs_have_writebacks(inode) &&
+	    NFS_PROTO(inode)->have_delegation(inode, FMODE_WRITE)) {
+		spin_lock(&inode->i_lock);
+		inode->i_version++;
+		spin_unlock(&inode->i_lock);
+	}
 	if (likely(!PageSwapCache(req->wb_page))) {
 		set_bit(PG_MAPPED, &req->wb_flags);
 		SetPagePrivate(req->wb_page);
 		set_page_private(req->wb_page, (unsigned long)req);
 	}
-	nfsi->nrequests++;
+	spin_unlock(&mapping->private_lock);
+	atomic_long_inc(&nfsi->nrequests);
 	/* this a head request for a page group - mark it as having an
 	 * extra reference so sub groups can follow suit.
 	 * This flag also informs pgio layer when to bump nrequests when
 	 * adding subrequests. */
 	WARN_ON(test_and_set_bit(PG_INODE_REF, &req->wb_flags));
 	kref_get(&req->wb_kref);
-	spin_unlock(&inode->i_lock);
 }
 
 /*
@@ -794,25 +777,22 @@ static void nfs_inode_add_request(struct inode *inode, struct nfs_page *req)
  */
 static void nfs_inode_remove_request(struct nfs_page *req)
 {
-	struct inode *inode = d_inode(req->wb_context->dentry);
+	struct address_space *mapping = page_file_mapping(req->wb_page);
+	struct inode *inode = mapping->host;
 	struct nfs_inode *nfsi = NFS_I(inode);
 	struct nfs_page *head;
 
+	atomic_long_dec(&nfsi->nrequests);
 	if (nfs_page_group_sync_on_bit(req, PG_REMOVE)) {
 		head = req->wb_head;
 
-		spin_lock(&inode->i_lock);
+		spin_lock(&mapping->private_lock);
 		if (likely(head->wb_page && !PageSwapCache(head->wb_page))) {
 			set_page_private(head->wb_page, 0);
 			ClearPagePrivate(head->wb_page);
 			clear_bit(PG_MAPPED, &head->wb_flags);
 		}
-		nfsi->nrequests--;
-		spin_unlock(&inode->i_lock);
-	} else {
-		spin_lock(&inode->i_lock);
-		nfsi->nrequests--;
-		spin_unlock(&inode->i_lock);
+		spin_unlock(&mapping->private_lock);
 	}
 
 	if (test_and_clear_bit(PG_INODE_REF, &req->wb_flags))
@@ -868,7 +848,8 @@ nfs_page_search_commits_for_head_request_locked(struct nfs_inode *nfsi,
  * number of outstanding requests requiring a commit as well as
  * the MM page stats.
  *
- * The caller must hold cinfo->inode->i_lock, and the nfs_page lock.
+ * The caller must hold NFS_I(cinfo->inode)->commit_mutex, and the
+ * nfs_page lock.
  */
 void
 nfs_request_add_commit_list_locked(struct nfs_page *req, struct list_head *dst,
@@ -876,7 +857,7 @@ nfs_request_add_commit_list_locked(struct nfs_page *req, struct list_head *dst,
 {
 	set_bit(PG_CLEAN, &req->wb_flags);
 	nfs_list_add_request(req, dst);
-	cinfo->mds->ncommit++;
+	atomic_long_inc(&cinfo->mds->ncommit);
 }
 EXPORT_SYMBOL_GPL(nfs_request_add_commit_list_locked);
 
@@ -896,9 +877,9 @@ EXPORT_SYMBOL_GPL(nfs_request_add_commit_list_locked);
 void
 nfs_request_add_commit_list(struct nfs_page *req, struct nfs_commit_info *cinfo)
 {
-	spin_lock(&cinfo->inode->i_lock);
+	mutex_lock(&NFS_I(cinfo->inode)->commit_mutex);
 	nfs_request_add_commit_list_locked(req, &cinfo->mds->list, cinfo);
-	spin_unlock(&cinfo->inode->i_lock);
+	mutex_unlock(&NFS_I(cinfo->inode)->commit_mutex);
 	if (req->wb_page)
 		nfs_mark_page_unstable(req->wb_page, cinfo);
 }
@@ -922,7 +903,7 @@ nfs_request_remove_commit_list(struct nfs_page *req,
 	if (!test_and_clear_bit(PG_CLEAN, &(req)->wb_flags))
 		return;
 	nfs_list_remove_request(req);
-	cinfo->mds->ncommit--;
+	atomic_long_dec(&cinfo->mds->ncommit);
 }
 EXPORT_SYMBOL_GPL(nfs_request_remove_commit_list);
 
@@ -967,7 +948,7 @@ nfs_clear_page_commit(struct page *page)
 		    WB_RECLAIMABLE);
 }
 
-/* Called holding inode (/cinfo) lock */
+/* Called holding the request lock on @req */
 static void
 nfs_clear_request_commit(struct nfs_page *req)
 {
@@ -976,9 +957,11 @@ nfs_clear_request_commit(struct nfs_page *req)
 		struct nfs_commit_info cinfo;
 
 		nfs_init_cinfo_from_inode(&cinfo, inode);
+		mutex_lock(&NFS_I(inode)->commit_mutex);
 		if (!pnfs_clear_request_commit(req, &cinfo)) {
 			nfs_request_remove_commit_list(req, &cinfo);
 		}
+		mutex_unlock(&NFS_I(inode)->commit_mutex);
 		nfs_clear_page_commit(req->wb_page);
 	}
 }
@@ -1023,7 +1006,6 @@ static void nfs_write_completion(struct nfs_pgio_header *hdr)
 remove_req:
 		nfs_inode_remove_request(req);
 next:
-		nfs_unlock_request(req);
 		nfs_end_page_writeback(req);
 		nfs_release_request(req);
 	}
@@ -1035,10 +1017,10 @@ out:
 unsigned long
 nfs_reqs_to_commit(struct nfs_commit_info *cinfo)
 {
-	return cinfo->mds->ncommit;
+	return atomic_long_read(&cinfo->mds->ncommit);
 }
 
-/* cinfo->inode->i_lock held by caller */
+/* NFS_I(cinfo->inode)->commit_mutex held by caller */
 int
 nfs_scan_commit_list(struct list_head *src, struct list_head *dst,
 		     struct nfs_commit_info *cinfo, int max)
@@ -1046,20 +1028,37 @@ nfs_scan_commit_list(struct list_head *src, struct list_head *dst,
 	struct nfs_page *req, *tmp;
 	int ret = 0;
 
+restart:
 	list_for_each_entry_safe(req, tmp, src, wb_list) {
-		if (!nfs_lock_request(req))
-			continue;
 		kref_get(&req->wb_kref);
-		if (cond_resched_lock(&cinfo->inode->i_lock))
-			list_safe_reset_next(req, tmp, wb_list);
+		if (!nfs_lock_request(req)) {
+			int status;
+
+			/* Prevent deadlock with nfs_lock_and_join_requests */
+			if (!list_empty(dst)) {
+				nfs_release_request(req);
+				continue;
+			}
+			/* Ensure we make progress to prevent livelock */
+			mutex_unlock(&NFS_I(cinfo->inode)->commit_mutex);
+			status = nfs_wait_on_request(req);
+			nfs_release_request(req);
+			mutex_lock(&NFS_I(cinfo->inode)->commit_mutex);
+			if (status < 0)
+				break;
+			goto restart;
+		}
 		nfs_request_remove_commit_list(req, cinfo);
+		clear_bit(PG_COMMIT_TO_DS, &req->wb_flags);
 		nfs_list_add_request(req, dst);
 		ret++;
 		if ((ret == max) && !cinfo->dreq)
 			break;
+		cond_resched();
 	}
 	return ret;
 }
+EXPORT_SYMBOL_GPL(nfs_scan_commit_list);
 
 /*
  * nfs_scan_commit - Scan an inode for commit requests
@@ -1076,15 +1075,17 @@ nfs_scan_commit(struct inode *inode, struct list_head *dst,
 {
 	int ret = 0;
 
-	spin_lock(&cinfo->inode->i_lock);
-	if (cinfo->mds->ncommit > 0) {
+	if (!atomic_long_read(&cinfo->mds->ncommit))
+		return 0;
+	mutex_lock(&NFS_I(cinfo->inode)->commit_mutex);
+	if (atomic_long_read(&cinfo->mds->ncommit) > 0) {
 		const int max = INT_MAX;
 
 		ret = nfs_scan_commit_list(&cinfo->mds->list, dst,
 					   cinfo, max);
 		ret += pnfs_scan_commit_lists(inode, cinfo, max - ret);
 	}
-	spin_unlock(&cinfo->inode->i_lock);
+	mutex_unlock(&NFS_I(cinfo->inode)->commit_mutex);
 	return ret;
 }
 
@@ -1105,43 +1106,21 @@ static struct nfs_page *nfs_try_to_update_request(struct inode *inode,
 	unsigned int end;
 	int error;
 
-	if (!PagePrivate(page))
-		return NULL;
-
 	end = offset + bytes;
-	spin_lock(&inode->i_lock);
-
-	for (;;) {
-		req = nfs_page_find_head_request_locked(NFS_I(inode), page);
-		if (req == NULL)
-			goto out_unlock;
-
-		/* should be handled by nfs_flush_incompatible */
-		WARN_ON_ONCE(req->wb_head != req);
-		WARN_ON_ONCE(req->wb_this_page != req);
-
-		rqend = req->wb_offset + req->wb_bytes;
-		/*
-		 * Tell the caller to flush out the request if
-		 * the offsets are non-contiguous.
-		 * Note: nfs_flush_incompatible() will already
-		 * have flushed out requests having wrong owners.
-		 */
-		if (offset > rqend
-		    || end < req->wb_offset)
-			goto out_flushme;
 
-		if (nfs_lock_request(req))
-			break;
+	req = nfs_lock_and_join_requests(page);
+	if (IS_ERR_OR_NULL(req))
+		return req;
 
-		/* The request is locked, so wait and then retry */
-		spin_unlock(&inode->i_lock);
-		error = nfs_wait_on_request(req);
-		nfs_release_request(req);
-		if (error != 0)
-			goto out_err;
-		spin_lock(&inode->i_lock);
-	}
+	rqend = req->wb_offset + req->wb_bytes;
+	/*
+	 * Tell the caller to flush out the request if
+	 * the offsets are non-contiguous.
+	 * Note: nfs_flush_incompatible() will already
+	 * have flushed out requests having wrong owners.
+	 */
+	if (offset > rqend || end < req->wb_offset)
+		goto out_flushme;
 
 	/* Okay, the request matches. Update the region */
 	if (offset < req->wb_offset) {
@@ -1152,17 +1131,17 @@ static struct nfs_page *nfs_try_to_update_request(struct inode *inode,
 		req->wb_bytes = end - req->wb_offset;
 	else
 		req->wb_bytes = rqend - req->wb_offset;
-out_unlock:
-	if (req)
-		nfs_clear_request_commit(req);
-	spin_unlock(&inode->i_lock);
 	return req;
 out_flushme:
-	spin_unlock(&inode->i_lock);
-	nfs_release_request(req);
+	/*
+	 * Note: we mark the request dirty here because
+	 * nfs_lock_and_join_requests() cannot preserve
+	 * commit flags, so we have to replay the write.
+	 */
+	nfs_mark_request_dirty(req);
+	nfs_unlock_and_release_request(req);
 	error = nfs_wb_page(inode, page);
-out_err:
-	return ERR_PTR(error);
+	return (error < 0) ? ERR_PTR(error) : NULL;
 }
 
 /*
@@ -1227,8 +1206,6 @@ int nfs_flush_incompatible(struct file *file, struct page *page)
 		l_ctx = req->wb_lock_context;
 		do_flush = req->wb_page != page ||
 			!nfs_match_open_context(req->wb_context, ctx);
-		/* for now, flush if more than 1 request in page_group */
-		do_flush |= req->wb_this_page != req;
 		if (l_ctx && flctx &&
 		    !(list_empty_careful(&flctx->flc_posix) &&
 		      list_empty_careful(&flctx->flc_flock))) {
@@ -1399,6 +1376,8 @@ static void nfs_initiate_write(struct nfs_pgio_header *hdr,
 
 	task_setup_data->priority = priority;
 	rpc_ops->write_setup(hdr, msg);
+	trace_nfs_initiate_write(hdr->inode, hdr->io_start, hdr->good_bytes,
+				 hdr->args.stable);
 
 	nfs4_state_protect_write(NFS_SERVER(hdr->inode)->nfs_client,
 				 &task_setup_data->rpc_client, msg, hdr);
@@ -1412,7 +1391,6 @@ static void nfs_redirty_request(struct nfs_page *req)
 {
 	nfs_mark_request_dirty(req);
 	set_bit(NFS_CONTEXT_RESEND_WRITES, &req->wb_context->flags);
-	nfs_unlock_request(req);
 	nfs_end_page_writeback(req);
 	nfs_release_request(req);
 }
@@ -1452,7 +1430,7 @@ void nfs_pageio_init_write(struct nfs_pageio_descriptor *pgio,
 		pg_ops = server->pnfs_curr_ld->pg_write_ops;
 #endif
 	nfs_pageio_init(pgio, inode, pg_ops, compl_ops, &nfs_rw_write_ops,
-			server->wsize, ioflags, GFP_NOIO);
+			server->wsize, ioflags);
 }
 EXPORT_SYMBOL_GPL(nfs_pageio_init_write);
 
@@ -1557,7 +1535,10 @@ static int nfs_writeback_done(struct rpc_task *task,
 	status = NFS_PROTO(inode)->write_done(task, hdr);
 	if (status != 0)
 		return status;
+
 	nfs_add_stats(inode, NFSIOS_SERVERWRITTENBYTES, hdr->res.count);
+	trace_nfs_writeback_done(inode, task->tk_status,
+				 hdr->args.offset, hdr->res.verf);
 
 	if (hdr->res.verf->committed < hdr->args.stable &&
 	    task->tk_status >= 0) {
@@ -1686,6 +1667,7 @@ int nfs_initiate_commit(struct rpc_clnt *clnt, struct nfs_commit_data *data,
 	};
 	/* Set up the initial task struct.  */
 	nfs_ops->commit_setup(data, &msg);
+	trace_nfs_initiate_commit(data);
 
 	dprintk("NFS: initiated commit call\n");
 
@@ -1810,6 +1792,7 @@ static void nfs_commit_done(struct rpc_task *task, void *calldata)
 
 	/* Call the NFS version-specific code */
 	NFS_PROTO(data->inode)->commit_done(task, data);
+	trace_nfs_commit_done(data);
 }
 
 static void nfs_commit_release_pages(struct nfs_commit_data *data)
@@ -1934,7 +1917,7 @@ int nfs_write_inode(struct inode *inode, struct writeback_control *wbc)
 	int ret = 0;
 
 	/* no commits means nothing needs to be done */
-	if (!nfsi->commit_info.ncommit)
+	if (!atomic_long_read(&nfsi->commit_info.ncommit))
 		return ret;
 
 	if (wbc->sync_mode == WB_SYNC_NONE) {
@@ -2015,7 +1998,7 @@ int nfs_wb_page_cancel(struct inode *inode, struct page *page)
 
 	/* blocking call to cancel all requests and join to a single (head)
 	 * request */
-	req = nfs_lock_and_join_requests(page, false);
+	req = nfs_lock_and_join_requests(page);
 
 	if (IS_ERR(req)) {
 		ret = PTR_ERR(req);