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>

1 files changed, 16 insertions, 11 deletions

diff --git a/fs/smb/client/smb2pdu.c b/fs/smb/client/smb2pdu.c
index 88dc49d67037..95377bb91950 100644
--- a/fs/smb/client/smb2pdu.c
+++ b/fs/smb/client/smb2pdu.c
@@ -4498,9 +4498,7 @@ static void smb2_readv_worker(struct work_struct *work)
 	struct cifs_io_subrequest *rdata =
 		container_of(work, struct cifs_io_subrequest, subreq.work);
 
-	netfs_subreq_terminated(&rdata->subreq,
-				(rdata->result == 0 || rdata->result == -EAGAIN) ?
-				rdata->got_bytes : rdata->result, true);
+	netfs_read_subreq_terminated(&rdata->subreq, rdata->result, false);
 }
 
 static void
@@ -4532,7 +4530,7 @@ smb2_readv_callback(struct mid_q_entry *mid)
 
 	cifs_dbg(FYI, "%s: mid=%llu state=%d result=%d bytes=%zu/%zu\n",
 		 __func__, mid->mid, mid->mid_state, rdata->result,
-		 rdata->actual_len, rdata->subreq.len - rdata->subreq.transferred);
+		 rdata->got_bytes, rdata->subreq.len - rdata->subreq.transferred);
 
 	switch (mid->mid_state) {
 	case MID_RESPONSE_RECEIVED:
@@ -4554,6 +4552,7 @@ smb2_readv_callback(struct mid_q_entry *mid)
 		break;
 	case MID_REQUEST_SUBMITTED:
 	case MID_RETRY_NEEDED:
+		__set_bit(NETFS_SREQ_NEED_RETRY, &rdata->subreq.flags);
 		rdata->result = -EAGAIN;
 		if (server->sign && rdata->got_bytes)
 			/* reset bytes number since we can not check a sign */
@@ -4588,7 +4587,7 @@ smb2_readv_callback(struct mid_q_entry *mid)
 				    rdata->req->cfile->fid.persistent_fid,
 				    tcon->tid, tcon->ses->Suid,
 				    rdata->subreq.start + rdata->subreq.transferred,
-				    rdata->actual_len,
+				    rdata->subreq.len   - rdata->subreq.transferred,
 				    rdata->result);
 	} else
 		trace_smb3_read_done(rdata->rreq->debug_id,
@@ -4603,9 +4602,9 @@ smb2_readv_callback(struct mid_q_entry *mid)
 		__set_bit(NETFS_SREQ_HIT_EOF, &rdata->subreq.flags);
 		rdata->result = 0;
 	} else {
-		if (rdata->got_bytes < rdata->actual_len &&
-		    rdata->subreq.start + rdata->subreq.transferred + rdata->got_bytes ==
-		    ictx->remote_i_size) {
+		size_t trans = rdata->subreq.transferred + rdata->got_bytes;
+		if (trans < rdata->subreq.len &&
+		    rdata->subreq.start + trans == ictx->remote_i_size) {
 			__set_bit(NETFS_SREQ_HIT_EOF, &rdata->subreq.flags);
 			rdata->result = 0;
 		}
@@ -4614,6 +4613,10 @@ smb2_readv_callback(struct mid_q_entry *mid)
 			      server->credits, server->in_flight,
 			      0, cifs_trace_rw_credits_read_response_clear);
 	rdata->credits.value = 0;
+	rdata->subreq.transferred += rdata->got_bytes;
+	if (rdata->subreq.start + rdata->subreq.transferred >= rdata->subreq.rreq->i_size)
+		__set_bit(NETFS_SREQ_HIT_EOF, &rdata->subreq.flags);
+	trace_netfs_sreq(&rdata->subreq, netfs_sreq_trace_io_progress);
 	INIT_WORK(&rdata->subreq.work, smb2_readv_worker);
 	queue_work(cifsiod_wq, &rdata->subreq.work);
 	release_mid(mid);
@@ -4648,7 +4651,7 @@ smb2_async_readv(struct cifs_io_subrequest *rdata)
 	io_parms.tcon = tlink_tcon(rdata->req->cfile->tlink);
 	io_parms.server = server = rdata->server;
 	io_parms.offset = subreq->start + subreq->transferred;
-	io_parms.length = rdata->actual_len;
+	io_parms.length = subreq->len   - subreq->transferred;
 	io_parms.persistent_fid = rdata->req->cfile->fid.persistent_fid;
 	io_parms.volatile_fid = rdata->req->cfile->fid.volatile_fid;
 	io_parms.pid = rdata->req->pid;
@@ -4669,7 +4672,7 @@ smb2_async_readv(struct cifs_io_subrequest *rdata)
 	shdr = (struct smb2_hdr *)buf;
 
 	if (rdata->credits.value > 0) {
-		shdr->CreditCharge = cpu_to_le16(DIV_ROUND_UP(rdata->actual_len,
+		shdr->CreditCharge = cpu_to_le16(DIV_ROUND_UP(io_parms.length,
 						SMB2_MAX_BUFFER_SIZE));
 		credit_request = le16_to_cpu(shdr->CreditCharge) + 8;
 		if (server->credits >= server->max_credits)
@@ -4697,7 +4700,8 @@ smb2_async_readv(struct cifs_io_subrequest *rdata)
 				    rdata->xid, io_parms.persistent_fid,
 				    io_parms.tcon->tid,
 				    io_parms.tcon->ses->Suid,
-				    io_parms.offset, rdata->actual_len, rc);
+				    io_parms.offset,
+				    subreq->len - subreq->transferred, rc);
 	}
 
 async_readv_out:
@@ -4880,6 +4884,7 @@ smb2_writev_callback(struct mid_q_entry *mid)
 			      server->credits, server->in_flight,
 			      0, cifs_trace_rw_credits_write_response_clear);
 	wdata->credits.value = 0;
+	trace_netfs_sreq(&wdata->subreq, netfs_sreq_trace_io_progress);
 	cifs_write_subrequest_terminated(wdata, result ?: written, true);
 	release_mid(mid);
 	trace_smb3_rw_credits(rreq_debug_id, subreq_debug_index, 0,