netfs: Speed up buffered reading

Improve the efficiency of buffered reads in a number of ways:

 (1) Overhaul the algorithm in general so that it's a lot more compact and
     split the read submission code between buffered and unbuffered
     versions.  The unbuffered version can be vastly simplified.

 (2) Read-result collection is handed off to a work queue rather than being
     done in the I/O thread.  Multiple subrequests can be processes
     simultaneously.

 (3) When a subrequest is collected, any folios it fully spans are
     collected and "spare" data on either side is donated to either the
     previous or the next subrequest in the sequence.

Notes:

 (*) Readahead expansion is massively slows down fio, presumably because it
     causes a load of extra allocations, both folio and xarray, up front
     before RPC requests can be transmitted.

 (*) RDMA with cifs does appear to work, both with SIW and RXE.

 (*) PG_private_2-based reading and copy-to-cache is split out into its own
     file and altered to use folio_queue.  Note that the copy to the cache
     now creates a new write transaction against the cache and adds the
     folios to be copied into it.  This allows it to use part of the
     writeback I/O code.

Signed-off-by: David Howells <dhowells@redhat.com>
cc: Jeff Layton <jlayton@kernel.org>
cc: netfs@lists.linux.dev
cc: linux-fsdevel@vger.kernel.org
Link: https://lore.kernel.org/r/20240814203850.2240469-20-dhowells@redhat.com/ # v2
Signed-off-by: Christian Brauner <brauner@kernel.org>

include/linux/folio_queue.h

@@ -27,6 +27,7 @@ struct folio_queue {
 	struct folio_queue	*prev;		/* Previous queue segment of NULL */
 	unsigned long		marks;		/* 1-bit mark per folio */
 	unsigned long		marks2;		/* Second 1-bit mark per folio */
+	unsigned long		marks3;		/* Third 1-bit mark per folio */
 #if PAGEVEC_SIZE > BITS_PER_LONG
 #error marks is not big enough
 #endif
@@ -39,6 +40,7 @@ static inline void folioq_init(struct folio_queue *folioq)
 	folioq->prev = NULL;
 	folioq->marks = 0;
 	folioq->marks2 = 0;
+	folioq->marks3 = 0;
 }
 
 static inline unsigned int folioq_nr_slots(const struct folio_queue *folioq)
@@ -87,6 +89,21 @@ static inline void folioq_unmark2(struct folio_queue *folioq, unsigned int slot)
 	clear_bit(slot, &folioq->marks2);
 }
 
+static inline bool folioq_is_marked3(const struct folio_queue *folioq, unsigned int slot)
+{
+	return test_bit(slot, &folioq->marks3);
+}
+
+static inline void folioq_mark3(struct folio_queue *folioq, unsigned int slot)
+{
+	set_bit(slot, &folioq->marks3);
+}
+
+static inline void folioq_unmark3(struct folio_queue *folioq, unsigned int slot)
+{
+	clear_bit(slot, &folioq->marks3);
+}
+
 static inline unsigned int __folio_order(struct folio *folio)
 {
 	if (!folio_test_large(folio))
@@ -133,6 +150,7 @@ static inline void folioq_clear(struct folio_queue *folioq, unsigned int slot)
 	folioq->vec.folios[slot] = NULL;
 	folioq_unmark(folioq, slot);
 	folioq_unmark2(folioq, slot);
+	folioq_unmark3(folioq, slot);
 }
 
 #endif /* _LINUX_FOLIO_QUEUE_H */

include/linux/netfs.h

@@ -178,36 +178,43 @@ struct netfs_io_subrequest {
 	unsigned long long	start;		/* Where to start the I/O */
 	size_t			len;		/* Size of the I/O */
 	size_t			transferred;	/* Amount of data transferred */
+	size_t			consumed;	/* Amount of read data consumed */
+	size_t			prev_donated;	/* Amount of data donated from previous subreq */
+	size_t			next_donated;	/* Amount of data donated from next subreq */
 	refcount_t		ref;
 	short			error;		/* 0 or error that occurred */
 	unsigned short		debug_index;	/* Index in list (for debugging output) */
 	unsigned int		nr_segs;	/* Number of segs in io_iter */
 	enum netfs_io_source	source;		/* Where to read from/write to */
 	unsigned char		stream_nr;	/* I/O stream this belongs to */
+	unsigned char		curr_folioq_slot; /* Folio currently being read */
+	unsigned char		curr_folio_order; /* Order of folio */
+	struct folio_queue	*curr_folioq;	/* Queue segment in which current folio resides */
 	unsigned long		flags;
 #define NETFS_SREQ_COPY_TO_CACHE	0	/* Set if should copy the data to the cache */
 #define NETFS_SREQ_CLEAR_TAIL		1	/* Set if the rest of the read should be cleared */
-#define NETFS_SREQ_SHORT_IO		2	/* Set if the I/O was short */
 #define NETFS_SREQ_SEEK_DATA_READ	3	/* Set if ->read() should SEEK_DATA first */
 #define NETFS_SREQ_NO_PROGRESS		4	/* Set if we didn't manage to read any data */
 #define NETFS_SREQ_ONDEMAND		5	/* Set if it's from on-demand read mode */
 #define NETFS_SREQ_BOUNDARY		6	/* Set if ends on hard boundary (eg. ceph object) */
+#define NETFS_SREQ_HIT_EOF		7	/* Set if short due to EOF */
 #define NETFS_SREQ_IN_PROGRESS		8	/* Unlocked when the subrequest completes */
 #define NETFS_SREQ_NEED_RETRY		9	/* Set if the filesystem requests a retry */
 #define NETFS_SREQ_RETRYING		10	/* Set if we're retrying */
 #define NETFS_SREQ_FAILED		11	/* Set if the subreq failed unretryably */
-#define NETFS_SREQ_HIT_EOF		12	/* Set if we hit the EOF */
 };
 
 enum netfs_io_origin {
 	NETFS_READAHEAD,		/* This read was triggered by readahead */
 	NETFS_READPAGE,			/* This read is a synchronous read */
+	NETFS_READ_GAPS,		/* This read is a synchronous read to fill gaps */
 	NETFS_READ_FOR_WRITE,		/* This read is to prepare a write */
 	NETFS_DIO_READ,			/* This is a direct I/O read */
 	NETFS_WRITEBACK,		/* This write was triggered by writepages */
 	NETFS_WRITETHROUGH,		/* This write was made by netfs_perform_write() */
 	NETFS_UNBUFFERED_WRITE,		/* This is an unbuffered write */
 	NETFS_DIO_WRITE,		/* This is a direct I/O write */
+	NETFS_PGPRIV2_COPY_TO_CACHE,	/* [DEPRECATED] This is writing read data to the cache */
 	nr__netfs_io_origin
 } __mode(byte);
 
@@ -224,6 +231,7 @@ struct netfs_io_request {
 	struct address_space	*mapping;	/* The mapping being accessed */
 	struct kiocb		*iocb;		/* AIO completion vector */
 	struct netfs_cache_resources cache_resources;
+	struct readahead_control *ractl;	/* Readahead descriptor */
 	struct list_head	proc_link;	/* Link in netfs_iorequests */
 	struct list_head	subrequests;	/* Contributory I/O operations */
 	struct netfs_io_stream	io_streams[2];	/* Streams of parallel I/O operations */
@@ -244,12 +252,10 @@ struct netfs_io_request {
 	unsigned int		nr_group_rel;	/* Number of refs to release on ->group */
 	spinlock_t		lock;		/* Lock for queuing subreqs */
 	atomic_t		nr_outstanding;	/* Number of ops in progress */
-	atomic_t		nr_copy_ops;	/* Number of copy-to-cache ops in progress */
-	size_t			upper_len;	/* Length can be extended to here */
 	unsigned long long	submitted;	/* Amount submitted for I/O so far */
 	unsigned long long	len;		/* Length of the request */
 	size_t			transferred;	/* Amount to be indicated as transferred */
-	short			error;		/* 0 or error that occurred */
+	long			error;		/* 0 or error that occurred */
 	enum netfs_io_origin	origin;		/* Origin of the request */
 	bool			direct_bv_unpin; /* T if direct_bv[] must be unpinned */
 	u8			buffer_head_slot; /* First slot in ->buffer */
@@ -260,9 +266,9 @@ struct netfs_io_request {
 	unsigned long long	collected_to;	/* Point we've collected to */
 	unsigned long long	cleaned_to;	/* Position we've cleaned folios to */
 	pgoff_t			no_unlock_folio; /* Don't unlock this folio after read */
+	size_t			prev_donated;	/* Fallback for subreq->prev_donated */
 	refcount_t		ref;
 	unsigned long		flags;
-#define NETFS_RREQ_INCOMPLETE_IO	0	/* Some ioreqs terminated short or with error */
 #define NETFS_RREQ_COPY_TO_CACHE	1	/* Need to write to the cache */
 #define NETFS_RREQ_NO_UNLOCK_FOLIO	2	/* Don't unlock no_unlock_folio on completion */
 #define NETFS_RREQ_DONT_UNLOCK_FOLIOS	3	/* Don't unlock the folios on completion */
@@ -274,6 +280,7 @@ struct netfs_io_request {
 #define NETFS_RREQ_PAUSE		11	/* Pause subrequest generation */
 #define NETFS_RREQ_USE_IO_ITER		12	/* Use ->io_iter rather than ->i_pages */
 #define NETFS_RREQ_ALL_QUEUED		13	/* All subreqs are now queued */
+#define NETFS_RREQ_NEED_RETRY		14	/* Need to try retrying */
 #define NETFS_RREQ_USE_PGPRIV2		31	/* [DEPRECATED] Use PG_private_2 to mark
 						 * write to cache on read */
 	const struct netfs_request_ops *netfs_ops;
@@ -292,7 +299,7 @@ struct netfs_request_ops {
 
 	/* Read request handling */
 	void (*expand_readahead)(struct netfs_io_request *rreq);
-	bool (*clamp_length)(struct netfs_io_subrequest *subreq);
+	int (*prepare_read)(struct netfs_io_subrequest *subreq);
 	void (*issue_read)(struct netfs_io_subrequest *subreq);
 	bool (*is_still_valid)(struct netfs_io_request *rreq);
 	int (*check_write_begin)(struct file *file, loff_t pos, unsigned len,
@@ -422,7 +429,10 @@ bool netfs_release_folio(struct folio *folio, gfp_t gfp);
 vm_fault_t netfs_page_mkwrite(struct vm_fault *vmf, struct netfs_group *netfs_group);
 
 /* (Sub)request management API. */
-void netfs_subreq_terminated(struct netfs_io_subrequest *, ssize_t, bool);
+void netfs_read_subreq_progress(struct netfs_io_subrequest *subreq,
+				bool was_async);
+void netfs_read_subreq_terminated(struct netfs_io_subrequest *subreq,
+				  int error, bool was_async);
 void netfs_get_subrequest(struct netfs_io_subrequest *subreq,
 			  enum netfs_sreq_ref_trace what);
 void netfs_put_subrequest(struct netfs_io_subrequest *subreq,