Merge patch series "netfs: Miscellaneous fixes"

David Howells <dhowells@redhat.com> says:

Here are some miscellaneous fixes and changes for netfslib, if you could
pull them:

 (1) Fix an oops in write-retry due to mis-resetting the I/O iterator.

 (2) Fix the recording of transferred bytes for short DIO reads.

 (3) Fix a request's work item to not require a reference, thereby avoiding
     the need to get rid of it in BH/IRQ context.

 (4) Fix waiting and waking to be consistent about the waitqueue used.

* patches from https://lore.kernel.org/20250519090707.2848510-1-dhowells@redhat.com:
  netfs: Fix wait/wake to be consistent about the waitqueue used
  netfs: Fix the request's work item to not require a ref
  netfs: Fix setting of transferred bytes with short DIO reads
  netfs: Fix oops in write-retry from mis-resetting the subreq iterator

Link: https://lore.kernel.org/20250519090707.2848510-1-dhowells@redhat.com
Signed-off-by: Christian Brauner <brauner@kernel.org>

15 files changed, 387 insertions, 336 deletions

diff --git a/fs/netfs/buffered_read.c b/fs/netfs/buffered_read.c
index 5f53634a3862..18b3dc74c70e 100644
--- a/fs/netfs/buffered_read.c
+++ b/fs/netfs/buffered_read.c
@@ -264,9 +264,9 @@ static void netfs_read_to_pagecache(struct netfs_io_request *rreq,
 				if (ret < 0) {
 					subreq->error = ret;
 					/* Not queued - release both refs. */
-					netfs_put_subrequest(subreq, false,
+					netfs_put_subrequest(subreq,
 							     netfs_sreq_trace_put_cancel);
-					netfs_put_subrequest(subreq, false,
+					netfs_put_subrequest(subreq,
 							     netfs_sreq_trace_put_cancel);
 					break;
 				}
@@ -299,8 +299,8 @@ static void netfs_read_to_pagecache(struct netfs_io_request *rreq,
 			subreq->error = ret;
 			trace_netfs_sreq(subreq, netfs_sreq_trace_cancel);
 			/* Not queued - release both refs. */
-			netfs_put_subrequest(subreq, false, netfs_sreq_trace_put_cancel);
-			netfs_put_subrequest(subreq, false, netfs_sreq_trace_put_cancel);
+			netfs_put_subrequest(subreq, netfs_sreq_trace_put_cancel);
+			netfs_put_subrequest(subreq, netfs_sreq_trace_put_cancel);
 			break;
 		}
 		size -= slice;
@@ -314,7 +314,7 @@ static void netfs_read_to_pagecache(struct netfs_io_request *rreq,
 	if (unlikely(size > 0)) {
 		smp_wmb(); /* Write lists before ALL_QUEUED. */
 		set_bit(NETFS_RREQ_ALL_QUEUED, &rreq->flags);
-		netfs_wake_read_collector(rreq);
+		netfs_wake_collector(rreq);
 	}
 
 	/* Defer error return as we may need to wait for outstanding I/O. */
@@ -366,12 +366,10 @@ void netfs_readahead(struct readahead_control *ractl)
 		goto cleanup_free;
 	netfs_read_to_pagecache(rreq, ractl);
 
-	netfs_put_request(rreq, true, netfs_rreq_trace_put_return);
-	return;
+	return netfs_put_request(rreq, netfs_rreq_trace_put_return);
 
 cleanup_free:
-	netfs_put_request(rreq, false, netfs_rreq_trace_put_failed);
-	return;
+	return netfs_put_request(rreq, netfs_rreq_trace_put_failed);
 }
 EXPORT_SYMBOL(netfs_readahead);
 
@@ -470,11 +468,11 @@ static int netfs_read_gaps(struct file *file, struct folio *folio)
 		folio_mark_uptodate(folio);
 	}
 	folio_unlock(folio);
-	netfs_put_request(rreq, false, netfs_rreq_trace_put_return);
+	netfs_put_request(rreq, netfs_rreq_trace_put_return);
 	return ret < 0 ? ret : 0;
 
 discard:
-	netfs_put_request(rreq, false, netfs_rreq_trace_put_discard);
+	netfs_put_request(rreq, netfs_rreq_trace_put_discard);
 alloc_error:
 	folio_unlock(folio);
 	return ret;
@@ -530,11 +528,11 @@ int netfs_read_folio(struct file *file, struct folio *folio)
 
 	netfs_read_to_pagecache(rreq, NULL);
 	ret = netfs_wait_for_read(rreq);
-	netfs_put_request(rreq, false, netfs_rreq_trace_put_return);
+	netfs_put_request(rreq, netfs_rreq_trace_put_return);
 	return ret < 0 ? ret : 0;
 
 discard:
-	netfs_put_request(rreq, false, netfs_rreq_trace_put_discard);
+	netfs_put_request(rreq, netfs_rreq_trace_put_discard);
 alloc_error:
 	folio_unlock(folio);
 	return ret;
@@ -689,7 +687,7 @@ retry:
 	ret = netfs_wait_for_read(rreq);
 	if (ret < 0)
 		goto error;
-	netfs_put_request(rreq, false, netfs_rreq_trace_put_return);
+	netfs_put_request(rreq, netfs_rreq_trace_put_return);
 
 have_folio:
 	ret = folio_wait_private_2_killable(folio);
@@ -701,7 +699,7 @@ have_folio_no_wait:
 	return 0;
 
 error_put:
-	netfs_put_request(rreq, false, netfs_rreq_trace_put_failed);
+	netfs_put_request(rreq, netfs_rreq_trace_put_failed);
 error:
 	if (folio) {
 		folio_unlock(folio);
@@ -752,11 +750,11 @@ int netfs_prefetch_for_write(struct file *file, struct folio *folio,
 
 	netfs_read_to_pagecache(rreq, NULL);
 	ret = netfs_wait_for_read(rreq);
-	netfs_put_request(rreq, false, netfs_rreq_trace_put_return);
+	netfs_put_request(rreq, netfs_rreq_trace_put_return);
 	return ret < 0 ? ret : 0;
 
 error_put:
-	netfs_put_request(rreq, false, netfs_rreq_trace_put_discard);
+	netfs_put_request(rreq, netfs_rreq_trace_put_discard);
 error:
 	_leave(" = %d", ret);
 	return ret;
diff --git a/fs/netfs/buffered_write.c b/fs/netfs/buffered_write.c
index 26a789c8ce18..72a3e6db2524 100644
--- a/fs/netfs/buffered_write.c
+++ b/fs/netfs/buffered_write.c
@@ -385,7 +385,7 @@ out:
 		wbc_detach_inode(&wbc);
 		if (ret2 == -EIOCBQUEUED)
 			return ret2;
-		if (ret == 0)
+		if (ret == 0 && ret2 < 0)
 			ret = ret2;
 	}
 
diff --git a/fs/netfs/direct_read.c b/fs/netfs/direct_read.c
index f11a89f2fdd9..e72c8b69b147 100644
--- a/fs/netfs/direct_read.c
+++ b/fs/netfs/direct_read.c
@@ -85,7 +85,7 @@ static int netfs_dispatch_unbuffered_reads(struct netfs_io_request *rreq)
 		if (rreq->netfs_ops->prepare_read) {
 			ret = rreq->netfs_ops->prepare_read(subreq);
 			if (ret < 0) {
-				netfs_put_subrequest(subreq, false, netfs_sreq_trace_put_cancel);
+				netfs_put_subrequest(subreq, netfs_sreq_trace_put_cancel);
 				break;
 			}
 		}
@@ -103,7 +103,7 @@ static int netfs_dispatch_unbuffered_reads(struct netfs_io_request *rreq)
 		rreq->netfs_ops->issue_read(subreq);
 
 		if (test_bit(NETFS_RREQ_PAUSE, &rreq->flags))
-			netfs_wait_for_pause(rreq);
+			netfs_wait_for_paused_read(rreq);
 		if (test_bit(NETFS_RREQ_FAILED, &rreq->flags))
 			break;
 		cond_resched();
@@ -112,7 +112,7 @@ static int netfs_dispatch_unbuffered_reads(struct netfs_io_request *rreq)
 	if (unlikely(size > 0)) {
 		smp_wmb(); /* Write lists before ALL_QUEUED. */
 		set_bit(NETFS_RREQ_ALL_QUEUED, &rreq->flags);
-		netfs_wake_read_collector(rreq);
+		netfs_wake_collector(rreq);
 	}
 
 	return ret;
@@ -141,7 +141,7 @@ static ssize_t netfs_unbuffered_read(struct netfs_io_request *rreq, bool sync)
 	ret = netfs_dispatch_unbuffered_reads(rreq);
 
 	if (!rreq->submitted) {
-		netfs_put_request(rreq, false, netfs_rreq_trace_put_no_submit);
+		netfs_put_request(rreq, netfs_rreq_trace_put_no_submit);
 		inode_dio_end(rreq->inode);
 		ret = 0;
 		goto out;
@@ -233,7 +233,7 @@ ssize_t netfs_unbuffered_read_iter_locked(struct kiocb *iocb, struct iov_iter *i
 	}
 
 out:
-	netfs_put_request(rreq, false, netfs_rreq_trace_put_return);
+	netfs_put_request(rreq, netfs_rreq_trace_put_return);
 	if (ret > 0)
 		orig_count -= ret;
 	return ret;
diff --git a/fs/netfs/direct_write.c b/fs/netfs/direct_write.c
index 42ce53cc216e..fa9a5bf3c6d5 100644
--- a/fs/netfs/direct_write.c
+++ b/fs/netfs/direct_write.c
@@ -87,6 +87,8 @@ ssize_t netfs_unbuffered_write_iter_locked(struct kiocb *iocb, struct iov_iter *
 	}
 
 	__set_bit(NETFS_RREQ_USE_IO_ITER, &wreq->flags);
+	if (async)
+		__set_bit(NETFS_RREQ_OFFLOAD_COLLECTION, &wreq->flags);
 
 	/* Copy the data into the bounce buffer and encrypt it. */
 	// TODO
@@ -105,19 +107,15 @@ ssize_t netfs_unbuffered_write_iter_locked(struct kiocb *iocb, struct iov_iter *
 
 	if (!async) {
 		trace_netfs_rreq(wreq, netfs_rreq_trace_wait_ip);
-		wait_on_bit(&wreq->flags, NETFS_RREQ_IN_PROGRESS,
-			    TASK_UNINTERRUPTIBLE);
-		ret = wreq->error;
-		if (ret == 0) {
-			ret = wreq->transferred;
+		ret = netfs_wait_for_write(wreq);
+		if (ret > 0)
 			iocb->ki_pos += ret;
-		}
 	} else {
 		ret = -EIOCBQUEUED;
 	}
 
 out:
-	netfs_put_request(wreq, false, netfs_rreq_trace_put_return);
+	netfs_put_request(wreq, netfs_rreq_trace_put_return);
 	return ret;
 }
 EXPORT_SYMBOL(netfs_unbuffered_write_iter_locked);
diff --git a/fs/netfs/fscache_io.c b/fs/netfs/fscache_io.c
index b1722a82c03d..e4308457633c 100644
--- a/fs/netfs/fscache_io.c
+++ b/fs/netfs/fscache_io.c
@@ -192,8 +192,7 @@ EXPORT_SYMBOL(__fscache_clear_page_bits);
 /*
  * Deal with the completion of writing the data to the cache.
  */
-static void fscache_wreq_done(void *priv, ssize_t transferred_or_error,
-			      bool was_async)
+static void fscache_wreq_done(void *priv, ssize_t transferred_or_error)
 {
 	struct fscache_write_request *wreq = priv;
 
@@ -202,8 +201,7 @@ static void fscache_wreq_done(void *priv, ssize_t transferred_or_error,
 					wreq->set_bits);
 
 	if (wreq->term_func)
-		wreq->term_func(wreq->term_func_priv, transferred_or_error,
-				was_async);
+		wreq->term_func(wreq->term_func_priv, transferred_or_error);
 	fscache_end_operation(&wreq->cache_resources);
 	kfree(wreq);
 }
@@ -255,14 +253,14 @@ void __fscache_write_to_cache(struct fscache_cookie *cookie,
 	return;
 
 abandon_end:
-	return fscache_wreq_done(wreq, ret, false);
+	return fscache_wreq_done(wreq, ret);
 abandon_free:
 	kfree(wreq);
 abandon:
 	if (using_pgpriv2)
 		fscache_clear_page_bits(mapping, start, len, cond);
 	if (term_func)
-		term_func(term_func_priv, ret, false);
+		term_func(term_func_priv, ret);
 }
 EXPORT_SYMBOL(__fscache_write_to_cache);
 
diff --git a/fs/netfs/internal.h b/fs/netfs/internal.h
index 1c4f953c3d68..e2ee9183392b 100644
--- a/fs/netfs/internal.h
+++ b/fs/netfs/internal.h
@@ -23,7 +23,7 @@
 /*
  * buffered_read.c
  */
-void netfs_cache_read_terminated(void *priv, ssize_t transferred_or_error, bool was_async);
+void netfs_cache_read_terminated(void *priv, ssize_t transferred_or_error);
 int netfs_prefetch_for_write(struct file *file, struct folio *folio,
 			     size_t offset, size_t len);
 
@@ -62,6 +62,14 @@ static inline void netfs_proc_del_rreq(struct netfs_io_request *rreq) {}
 struct folio_queue *netfs_buffer_make_space(struct netfs_io_request *rreq,
 					    enum netfs_folioq_trace trace);
 void netfs_reset_iter(struct netfs_io_subrequest *subreq);
+void netfs_wake_collector(struct netfs_io_request *rreq);
+void netfs_subreq_clear_in_progress(struct netfs_io_subrequest *subreq);
+void netfs_wait_for_in_progress_stream(struct netfs_io_request *rreq,
+				       struct netfs_io_stream *stream);
+ssize_t netfs_wait_for_read(struct netfs_io_request *rreq);
+ssize_t netfs_wait_for_write(struct netfs_io_request *rreq);
+void netfs_wait_for_paused_read(struct netfs_io_request *rreq);
+void netfs_wait_for_paused_write(struct netfs_io_request *rreq);
 
 /*
  * objects.c
@@ -71,9 +79,8 @@ struct netfs_io_request *netfs_alloc_request(struct address_space *mapping,
 					     loff_t start, size_t len,
 					     enum netfs_io_origin origin);
 void netfs_get_request(struct netfs_io_request *rreq, enum netfs_rreq_ref_trace what);
-void netfs_clear_subrequests(struct netfs_io_request *rreq, bool was_async);
-void netfs_put_request(struct netfs_io_request *rreq, bool was_async,
-		       enum netfs_rreq_ref_trace what);
+void netfs_clear_subrequests(struct netfs_io_request *rreq);
+void netfs_put_request(struct netfs_io_request *rreq, enum netfs_rreq_ref_trace what);
 struct netfs_io_subrequest *netfs_alloc_subrequest(struct netfs_io_request *rreq);
 
 static inline void netfs_see_request(struct netfs_io_request *rreq,
@@ -92,11 +99,9 @@ static inline void netfs_see_subrequest(struct netfs_io_subrequest *subreq,
 /*
  * read_collect.c
  */
+bool netfs_read_collection(struct netfs_io_request *rreq);
 void netfs_read_collection_worker(struct work_struct *work);
-void netfs_wake_read_collector(struct netfs_io_request *rreq);
-void netfs_cache_read_terminated(void *priv, ssize_t transferred_or_error, bool was_async);
-ssize_t netfs_wait_for_read(struct netfs_io_request *rreq);
-void netfs_wait_for_pause(struct netfs_io_request *rreq);
+void netfs_cache_read_terminated(void *priv, ssize_t transferred_or_error);
 
 /*
  * read_pgpriv2.c
@@ -176,8 +181,8 @@ static inline void netfs_stat_d(atomic_t *stat)
  * write_collect.c
  */
 int netfs_folio_written_back(struct folio *folio);
+bool netfs_write_collection(struct netfs_io_request *wreq);
 void netfs_write_collection_worker(struct work_struct *work);
-void netfs_wake_write_collector(struct netfs_io_request *wreq, bool was_async);
 
 /*
  * write_issue.c
@@ -198,8 +203,8 @@ struct netfs_io_request *netfs_begin_writethrough(struct kiocb *iocb, size_t len
 int netfs_advance_writethrough(struct netfs_io_request *wreq, struct writeback_control *wbc,
 			       struct folio *folio, size_t copied, bool to_page_end,
 			       struct folio **writethrough_cache);
-int netfs_end_writethrough(struct netfs_io_request *wreq, struct writeback_control *wbc,
-			   struct folio *writethrough_cache);
+ssize_t netfs_end_writethrough(struct netfs_io_request *wreq, struct writeback_control *wbc,
+			       struct folio *writethrough_cache);
 int netfs_unbuffered_write(struct netfs_io_request *wreq, bool may_wait, size_t len);
 
 /*
@@ -255,6 +260,21 @@ static inline void netfs_put_group_many(struct netfs_group *netfs_group, int nr)
 }
 
 /*
+ * Clear and wake up a NETFS_RREQ_* flag bit on a request.
+ */
+static inline void netfs_wake_rreq_flag(struct netfs_io_request *rreq,
+					unsigned int rreq_flag,
+					enum netfs_rreq_trace trace)
+{
+	if (test_bit(rreq_flag, &rreq->flags)) {
+		trace_netfs_rreq(rreq, trace);
+		clear_bit_unlock(rreq_flag, &rreq->flags);
+		smp_mb__after_atomic(); /* Set flag before task state */
+		wake_up(&rreq->waitq);
+	}
+}
+
+/*
  * fscache-cache.c
  */
 #ifdef CONFIG_PROC_FS
diff --git a/fs/netfs/misc.c b/fs/netfs/misc.c
index 7099aa07737a..77e7f7c79d27 100644
--- a/fs/netfs/misc.c
+++ b/fs/netfs/misc.c
@@ -313,3 +313,221 @@ bool netfs_release_folio(struct folio *folio, gfp_t gfp)
 	return true;
 }
 EXPORT_SYMBOL(netfs_release_folio);
+
+/*
+ * Wake the collection work item.
+ */
+void netfs_wake_collector(struct netfs_io_request *rreq)
+{
+	if (test_bit(NETFS_RREQ_OFFLOAD_COLLECTION, &rreq->flags) &&
+	    !test_bit(NETFS_RREQ_RETRYING, &rreq->flags)) {
+		queue_work(system_unbound_wq, &rreq->work);
+	} else {
+		trace_netfs_rreq(rreq, netfs_rreq_trace_wake_queue);
+		wake_up(&rreq->waitq);
+	}
+}
+
+/*
+ * Mark a subrequest as no longer being in progress and, if need be, wake the
+ * collector.
+ */
+void netfs_subreq_clear_in_progress(struct netfs_io_subrequest *subreq)
+{
+	struct netfs_io_request *rreq = subreq->rreq;
+	struct netfs_io_stream *stream = &rreq->io_streams[subreq->stream_nr];
+
+	clear_bit_unlock(NETFS_SREQ_IN_PROGRESS, &subreq->flags);
+	smp_mb__after_atomic(); /* Clear IN_PROGRESS before task state */
+
+	/* If we are at the head of the queue, wake up the collector. */
+	if (list_is_first(&subreq->rreq_link, &stream->subrequests) ||
+	    test_bit(NETFS_RREQ_RETRYING, &rreq->flags))
+		netfs_wake_collector(rreq);
+}
+
+/*
+ * Wait for all outstanding I/O in a stream to quiesce.
+ */
+void netfs_wait_for_in_progress_stream(struct netfs_io_request *rreq,
+				       struct netfs_io_stream *stream)
+{
+	struct netfs_io_subrequest *subreq;
+	DEFINE_WAIT(myself);
+
+	list_for_each_entry(subreq, &stream->subrequests, rreq_link) {
+		if (!test_bit(NETFS_SREQ_IN_PROGRESS, &subreq->flags))
+			continue;
+
+		trace_netfs_rreq(rreq, netfs_rreq_trace_wait_queue);
+		for (;;) {
+			prepare_to_wait(&rreq->waitq, &myself, TASK_UNINTERRUPTIBLE);
+
+			if (!test_bit(NETFS_SREQ_IN_PROGRESS, &subreq->flags))
+				break;
+
+			trace_netfs_sreq(subreq, netfs_sreq_trace_wait_for);
+			schedule();
+			trace_netfs_rreq(rreq, netfs_rreq_trace_woke_queue);
+		}
+	}
+
+	finish_wait(&rreq->waitq, &myself);
+}
+
+/*
+ * Perform collection in app thread if not offloaded to workqueue.
+ */
+static int netfs_collect_in_app(struct netfs_io_request *rreq,
+				bool (*collector)(struct netfs_io_request *rreq))
+{
+	bool need_collect = false, inactive = true;
+
+	for (int i = 0; i < NR_IO_STREAMS; i++) {
+		struct netfs_io_subrequest *subreq;
+		struct netfs_io_stream *stream = &rreq->io_streams[i];
+
+		if (!stream->active)
+			continue;
+		inactive = false;
+		trace_netfs_collect_stream(rreq, stream);
+		subreq = list_first_entry_or_null(&stream->subrequests,
+						  struct netfs_io_subrequest,
+						  rreq_link);
+		if (subreq &&
+		    (!test_bit(NETFS_SREQ_IN_PROGRESS, &subreq->flags) ||
+		     test_bit(NETFS_SREQ_MADE_PROGRESS, &subreq->flags))) {
+			need_collect = true;
+			break;
+		}
+	}
+
+	if (!need_collect && !inactive)
+		return 0; /* Sleep */
+
+	__set_current_state(TASK_RUNNING);
+	if (collector(rreq)) {
+		/* Drop the ref from the NETFS_RREQ_IN_PROGRESS flag. */
+		netfs_put_request(rreq, netfs_rreq_trace_put_work_ip);
+		return 1; /* Done */
+	}
+
+	if (inactive) {
+		WARN(true, "Failed to collect inactive req R=%08x\n",
+		     rreq->debug_id);
+		cond_resched();
+	}
+	return 2; /* Again */
+}
+
+/*
+ * Wait for a request to complete, successfully or otherwise.
+ */
+static ssize_t netfs_wait_for_request(struct netfs_io_request *rreq,
+				      bool (*collector)(struct netfs_io_request *rreq))
+{
+	DEFINE_WAIT(myself);
+	ssize_t ret;
+
+	for (;;) {
+		trace_netfs_rreq(rreq, netfs_rreq_trace_wait_queue);
+		prepare_to_wait(&rreq->waitq, &myself, TASK_UNINTERRUPTIBLE);
+
+		if (!test_bit(NETFS_RREQ_OFFLOAD_COLLECTION, &rreq->flags)) {
+			switch (netfs_collect_in_app(rreq, collector)) {
+			case 0:
+				break;
+			case 1:
+				goto all_collected;
+			case 2:
+				continue;
+			}
+		}
+
+		if (!test_bit(NETFS_RREQ_IN_PROGRESS, &rreq->flags))
+			break;
+
+		schedule();
+		trace_netfs_rreq(rreq, netfs_rreq_trace_woke_queue);
+	}
+
+all_collected:
+	finish_wait(&rreq->waitq, &myself);
+
+	ret = rreq->error;
+	if (ret == 0) {
+		ret = rreq->transferred;
+		switch (rreq->origin) {
+		case NETFS_DIO_READ:
+		case NETFS_DIO_WRITE:
+		case NETFS_READ_SINGLE:
+		case NETFS_UNBUFFERED_WRITE:
+			break;
+		default:
+			if (rreq->submitted < rreq->len) {
+				trace_netfs_failure(rreq, NULL, ret, netfs_fail_short_read);
+				ret = -EIO;
+			}
+			break;
+		}
+	}
+
+	return ret;
+}
+
+ssize_t netfs_wait_for_read(struct netfs_io_request *rreq)
+{
+	return netfs_wait_for_request(rreq, netfs_read_collection);
+}
+
+ssize_t netfs_wait_for_write(struct netfs_io_request *rreq)
+{
+	return netfs_wait_for_request(rreq, netfs_write_collection);
+}
+
+/*
+ * Wait for a paused operation to unpause or complete in some manner.
+ */
+static void netfs_wait_for_pause(struct netfs_io_request *rreq,
+				 bool (*collector)(struct netfs_io_request *rreq))
+{
+	DEFINE_WAIT(myself);
+
+	trace_netfs_rreq(rreq, netfs_rreq_trace_wait_pause);
+
+	for (;;) {
+		trace_netfs_rreq(rreq, netfs_rreq_trace_wait_queue);
+		prepare_to_wait(&rreq->waitq, &myself, TASK_UNINTERRUPTIBLE);
+
+		if (!test_bit(NETFS_RREQ_OFFLOAD_COLLECTION, &rreq->flags)) {
+			switch (netfs_collect_in_app(rreq, collector)) {
+			case 0:
+				break;
+			case 1:
+				goto all_collected;
+			case 2:
+				continue;
+			}
+		}
+
+		if (!test_bit(NETFS_RREQ_IN_PROGRESS, &rreq->flags) ||
+		    !test_bit(NETFS_RREQ_PAUSE, &rreq->flags))
+			break;
+
+		schedule();
+		trace_netfs_rreq(rreq, netfs_rreq_trace_woke_queue);
+	}
+
+all_collected:
+	finish_wait(&rreq->waitq, &myself);
+}
+
+void netfs_wait_for_paused_read(struct netfs_io_request *rreq)
+{
+	return netfs_wait_for_pause(rreq, netfs_read_collection);
+}
+
+void netfs_wait_for_paused_write(struct netfs_io_request *rreq)
+{
+	return netfs_wait_for_pause(rreq, netfs_write_collection);
+}
diff --git a/fs/netfs/objects.c b/fs/netfs/objects.c
index d6f8984f9f5b..f603f107ba1d 100644
--- a/fs/netfs/objects.c
+++ b/fs/netfs/objects.c
@@ -10,6 +10,8 @@
 #include <linux/delay.h>
 #include "internal.h"
 
+static void netfs_free_request(struct work_struct *work);
+
 /*
  * Allocate an I/O request and initialise it.
  */
@@ -34,6 +36,7 @@ struct netfs_io_request *netfs_alloc_request(struct address_space *mapping,
 	}
 
 	memset(rreq, 0, kmem_cache_size(cache));
+	INIT_WORK(&rreq->cleanup_work, netfs_free_request);
 	rreq->start	= start;
 	rreq->len	= len;
 	rreq->origin	= origin;
@@ -49,7 +52,7 @@ struct netfs_io_request *netfs_alloc_request(struct address_space *mapping,
 	INIT_LIST_HEAD(&rreq->io_streams[0].subrequests);
 	INIT_LIST_HEAD(&rreq->io_streams[1].subrequests);
 	init_waitqueue_head(&rreq->waitq);
-	refcount_set(&rreq->ref, 1);
+	refcount_set(&rreq->ref, 2);
 
 	if (origin == NETFS_READAHEAD ||
 	    origin == NETFS_READPAGE ||
@@ -73,7 +76,7 @@ struct netfs_io_request *netfs_alloc_request(struct address_space *mapping,
 	}
 
 	atomic_inc(&ctx->io_count);
-	trace_netfs_rreq_ref(rreq->debug_id, 1, netfs_rreq_trace_new);
+	trace_netfs_rreq_ref(rreq->debug_id, refcount_read(&rreq->ref), netfs_rreq_trace_new);
 	netfs_proc_add_rreq(rreq);
 	netfs_stat(&netfs_n_rh_rreq);
 	return rreq;
@@ -87,7 +90,7 @@ void netfs_get_request(struct netfs_io_request *rreq, enum netfs_rreq_ref_trace
 	trace_netfs_rreq_ref(rreq->debug_id, r + 1, what);
 }
 
-void netfs_clear_subrequests(struct netfs_io_request *rreq, bool was_async)
+void netfs_clear_subrequests(struct netfs_io_request *rreq)
 {
 	struct netfs_io_subrequest *subreq;
 	struct netfs_io_stream *stream;
@@ -99,8 +102,7 @@ void netfs_clear_subrequests(struct netfs_io_request *rreq, bool was_async)
 			subreq = list_first_entry(&stream->subrequests,
 						  struct netfs_io_subrequest, rreq_link);
 			list_del(&subreq->rreq_link);
-			netfs_put_subrequest(subreq, was_async,
-					     netfs_sreq_trace_put_clear);
+			netfs_put_subrequest(subreq, netfs_sreq_trace_put_clear);
 		}
 	}
 }
@@ -116,13 +118,19 @@ static void netfs_free_request_rcu(struct rcu_head *rcu)
 static void netfs_free_request(struct work_struct *work)
 {
 	struct netfs_io_request *rreq =
-		container_of(work, struct netfs_io_request, work);
+		container_of(work, struct netfs_io_request, cleanup_work);
 	struct netfs_inode *ictx = netfs_inode(rreq->inode);
 	unsigned int i;
 
 	trace_netfs_rreq(rreq, netfs_rreq_trace_free);
+
+	/* Cancel/flush the result collection worker.  That does not carry a
+	 * ref of its own, so we must wait for it somewhere.
+	 */
+	cancel_work_sync(&rreq->work);
+
 	netfs_proc_del_rreq(rreq);
-	netfs_clear_subrequests(rreq, false);
+	netfs_clear_subrequests(rreq);
 	if (rreq->netfs_ops->free_request)
 		rreq->netfs_ops->free_request(rreq);
 	if (rreq->cache_resources.ops)
@@ -143,8 +151,7 @@ static void netfs_free_request(struct work_struct *work)
 	call_rcu(&rreq->rcu, netfs_free_request_rcu);
 }
 
-void netfs_put_request(struct netfs_io_request *rreq, bool was_async,
-		       enum netfs_rreq_ref_trace what)
+void netfs_put_request(struct netfs_io_request *rreq, enum netfs_rreq_ref_trace what)
 {
 	unsigned int debug_id;
 	bool dead;
@@ -154,15 +161,8 @@ void netfs_put_request(struct netfs_io_request *rreq, bool was_async,
 		debug_id = rreq->debug_id;
 		dead = __refcount_dec_and_test(&rreq->ref, &r);
 		trace_netfs_rreq_ref(debug_id, r - 1, what);
-		if (dead) {
-			if (was_async) {
-				rreq->work.func = netfs_free_request;
-				if (!queue_work(system_unbound_wq, &rreq->work))
-					WARN_ON(1);
-			} else {
-				netfs_free_request(&rreq->work);
-			}
-		}
+		if (dead)
+			WARN_ON(!queue_work(system_unbound_wq, &rreq->cleanup_work));
 	}
 }
 
@@ -204,8 +204,7 @@ void netfs_get_subrequest(struct netfs_io_subrequest *subreq,
 			     what);
 }
 
-static void netfs_free_subrequest(struct netfs_io_subrequest *subreq,
-				  bool was_async)
+static void netfs_free_subrequest(struct netfs_io_subrequest *subreq)
 {
 	struct netfs_io_request *rreq = subreq->rreq;
 
@@ -214,10 +213,10 @@ static void netfs_free_subrequest(struct netfs_io_subrequest *subreq,
 		rreq->netfs_ops->free_subrequest(subreq);
 	mempool_free(subreq, rreq->netfs_ops->subrequest_pool ?: &netfs_subrequest_pool);
 	netfs_stat_d(&netfs_n_rh_sreq);
-	netfs_put_request(rreq, was_async, netfs_rreq_trace_put_subreq);
+	netfs_put_request(rreq, netfs_rreq_trace_put_subreq);
 }
 
-void netfs_put_subrequest(struct netfs_io_subrequest *subreq, bool was_async,
+void netfs_put_subrequest(struct netfs_io_subrequest *subreq,
 			  enum netfs_sreq_ref_trace what)
 {
 	unsigned int debug_index = subreq->debug_index;
@@ -228,5 +227,5 @@ void netfs_put_subrequest(struct netfs_io_subrequest *subreq, bool was_async,
 	dead = __refcount_dec_and_test(&subreq->ref, &r);
 	trace_netfs_sreq_ref(debug_id, debug_index, r - 1, what);
 	if (dead)
-		netfs_free_subrequest(subreq, was_async);
+		netfs_free_subrequest(subreq);
 }
diff --git a/fs/netfs/read_collect.c b/fs/netfs/read_collect.c
index 173433d61ea6..31d00e30a59c 100644
--- a/fs/netfs/read_collect.c
+++ b/fs/netfs/read_collect.c
@@ -278,9 +278,13 @@ reassess:
 			stream->need_retry = true;
 			notes |= NEED_RETRY | MADE_PROGRESS;
 			break;
+		} else if (test_bit(NETFS_RREQ_SHORT_TRANSFER, &rreq->flags)) {
+			notes |= MADE_PROGRESS;
 		} else {
 			if (!stream->failed)
-				stream->transferred = stream->collected_to - rreq->start;
+				stream->transferred += transferred;
+			if (front->transferred < front->len)
+				set_bit(NETFS_RREQ_SHORT_TRANSFER, &rreq->flags);
 			notes |= MADE_PROGRESS;
 		}
 
@@ -295,7 +299,7 @@ reassess:
 						 struct netfs_io_subrequest, rreq_link);
 		stream->front = front;
 		spin_unlock(&rreq->lock);
-		netfs_put_subrequest(remove, false,
+		netfs_put_subrequest(remove,
 				     notes & ABANDON_SREQ ?
 				     netfs_sreq_trace_put_abandon :
 				     netfs_sreq_trace_put_done);
@@ -309,14 +313,8 @@ reassess:
 
 	if (notes & NEED_RETRY)
 		goto need_retry;
-	if ((notes & MADE_PROGRESS) && test_bit(NETFS_RREQ_PAUSE, &rreq->flags)) {
-		trace_netfs_rreq(rreq, netfs_rreq_trace_unpause);
-		clear_bit_unlock(NETFS_RREQ_PAUSE, &rreq->flags);
-		smp_mb__after_atomic(); /* Set PAUSE before task state */
-		wake_up(&rreq->waitq);
-	}
-
 	if (notes & MADE_PROGRESS) {
+		netfs_wake_rreq_flag(rreq, NETFS_RREQ_PAUSE, netfs_rreq_trace_unpause);
 		//cond_resched();
 		goto reassess;
 	}
@@ -340,23 +338,8 @@ need_retry:
  */
 static void netfs_rreq_assess_dio(struct netfs_io_request *rreq)
 {
-	struct netfs_io_subrequest *subreq;
-	struct netfs_io_stream *stream = &rreq->io_streams[0];
 	unsigned int i;
 
-	/* Collect unbuffered reads and direct reads, adding up the transfer
-	 * sizes until we find the first short or failed subrequest.
-	 */
-	list_for_each_entry(subreq, &stream->subrequests, rreq_link) {
-		rreq->transferred += subreq->transferred;
-
-		if (subreq->transferred < subreq->len ||
-		    test_bit(NETFS_SREQ_FAILED, &subreq->flags)) {
-			rreq->error = subreq->error;
-			break;
-		}
-	}
-
 	if (rreq->origin == NETFS_DIO_READ) {
 		for (i = 0; i < rreq->direct_bv_count; i++) {
 			flush_dcache_page(rreq->direct_bv[i].bv_page);
@@ -408,7 +391,7 @@ static void netfs_rreq_assess_single(struct netfs_io_request *rreq)
  * Note that we're in normal kernel thread context at this point, possibly
  * running on a workqueue.
  */
-static void netfs_read_collection(struct netfs_io_request *rreq)
+bool netfs_read_collection(struct netfs_io_request *rreq)
 {
 	struct netfs_io_stream *stream = &rreq->io_streams[0];
 
@@ -418,11 +401,11 @@ static void netfs_read_collection(struct netfs_io_request *rreq)
 	 * queue is empty.
 	 */
 	if (!test_bit(NETFS_RREQ_ALL_QUEUED, &rreq->flags))
-		return;
+		return false;
 	smp_rmb(); /* Read ALL_QUEUED before subreq lists. */
 
 	if (!list_empty(&stream->subrequests))
-		return;
+		return false;
 
 	/* Okay, declare that all I/O is complete. */
 	rreq->transferred = stream->transferred;
@@ -443,14 +426,15 @@ static void netfs_read_collection(struct netfs_io_request *rreq)
 	}
 	task_io_account_read(rreq->transferred);
 
-	trace_netfs_rreq(rreq, netfs_rreq_trace_wake_ip);
-	clear_and_wake_up_bit(NETFS_RREQ_IN_PROGRESS, &rreq->flags);
+	netfs_wake_rreq_flag(rreq, NETFS_RREQ_IN_PROGRESS, netfs_rreq_trace_wake_ip);
+	/* As we cleared NETFS_RREQ_IN_PROGRESS, we acquired its ref. */
 
 	trace_netfs_rreq(rreq, netfs_rreq_trace_done);
-	netfs_clear_subrequests(rreq, false);
+	netfs_clear_subrequests(rreq);
 	netfs_unlock_abandoned_read_pages(rreq);
 	if (unlikely(rreq->copy_to_cache))
 		netfs_pgpriv2_end_copy_to_cache(rreq);
+	return true;
 }
 
 void netfs_read_collection_worker(struct work_struct *work)
@@ -458,26 +442,12 @@ void netfs_read_collection_worker(struct work_struct *work)
 	struct netfs_io_request *rreq = container_of(work, struct netfs_io_request, work);
 
 	netfs_see_request(rreq, netfs_rreq_trace_see_work);
-	if (test_bit(NETFS_RREQ_IN_PROGRESS, &rreq->flags))
-		netfs_read_collection(rreq);
-	netfs_put_request(rreq, false, netfs_rreq_trace_put_work);
-}
-
-/*
- * Wake the collection work item.
- */
-void netfs_wake_read_collector(struct netfs_io_request *rreq)
-{
-	if (test_bit(NETFS_RREQ_OFFLOAD_COLLECTION, &rreq->flags) &&
-	    !test_bit(NETFS_RREQ_RETRYING, &rreq->flags)) {
-		if (!work_pending(&rreq->work)) {
-			netfs_get_request(rreq, netfs_rreq_trace_get_work);
-			if (!queue_work(system_unbound_wq, &rreq->work))
-				netfs_put_request(rreq, true, netfs_rreq_trace_put_work_nq);
-		}
-	} else {
-		trace_netfs_rreq(rreq, netfs_rreq_trace_wake_queue);
-		wake_up(&rreq->waitq);
+	if (test_bit(NETFS_RREQ_IN_PROGRESS, &rreq->flags)) {
+		if (netfs_read_collection(rreq))
+			/* Drop the ref from the IN_PROGRESS flag. */
+			netfs_put_request(rreq, netfs_rreq_trace_put_work_ip);
+		else
+			netfs_see_request(rreq, netfs_rreq_trace_see_work_complete);
 	}
 }
 
@@ -509,7 +479,7 @@ void netfs_read_subreq_progress(struct netfs_io_subrequest *subreq