netfs: Fix undifferentiation of DIO reads from unbuffered reads

On cifs, "DIO reads" (specified by O_DIRECT) need to be differentiated from
"unbuffered reads" (specified by cache=none in the mount parameters).  The
difference is flagged in the protocol and the server may behave
differently: Windows Server will, for example, mandate that DIO reads are
block aligned.

Fix this by adding a NETFS_UNBUFFERED_READ to differentiate this from
NETFS_DIO_READ, parallelling the write differentiation that already exists.
cifs will then do the right thing.

Fixes: 016dc85 ("netfs: Implement unbuffered/DIO read support")
Signed-off-by: David Howells <[email protected]>
Link: https://lore.kernel.org/[email protected]
Reviewed-by: "Paulo Alcantara (Red Hat)" <[email protected]>
Reviewed-by: Viacheslav Dubeyko <[email protected]>
cc: Steve French <[email protected]>
cc: [email protected]
cc: [email protected]
cc: [email protected]
cc: [email protected]
cc: [email protected]
cc: [email protected]
cc: [email protected]
Signed-off-by: Christian Brauner <[email protected]>

Author: dhowells

fs/9p/vfs_addr.c

@@ -77,7 +77,8 @@ static void v9fs_issue_read(struct netfs_io_subrequest *subreq)
 
 	/* if we just extended the file size, any portion not in
 	 * cache won't be on server and is zeroes */
-	if (subreq->rreq->origin != NETFS_DIO_READ)
+	if (subreq->rreq->origin != NETFS_UNBUFFERED_READ &&
+	    subreq->rreq->origin != NETFS_DIO_READ)
 		__set_bit(NETFS_SREQ_CLEAR_TAIL, &subreq->flags);
 	if (pos + total >= i_size_read(rreq->inode))
 		__set_bit(NETFS_SREQ_HIT_EOF, &subreq->flags);

fs/afs/write.c

@@ -202,6 +202,7 @@ void afs_retry_request(struct netfs_io_request *wreq, struct netfs_io_stream *st
 	case NETFS_READ_GAPS:
 	case NETFS_READ_SINGLE:
 	case NETFS_READ_FOR_WRITE:
+	case NETFS_UNBUFFERED_READ:
 	case NETFS_DIO_READ:
 		return;
 	default:

fs/ceph/addr.c

@@ -238,6 +238,7 @@ static void finish_netfs_read(struct ceph_osd_request *req)
 		if (sparse && err > 0)
 			err = ceph_sparse_ext_map_end(op);
 		if (err < subreq->len &&
+		    subreq->rreq->origin != NETFS_UNBUFFERED_READ &&
 		    subreq->rreq->origin != NETFS_DIO_READ)
 			__set_bit(NETFS_SREQ_CLEAR_TAIL, &subreq->flags);
 		if (IS_ENCRYPTED(inode) && err > 0) {
@@ -281,7 +282,8 @@ static bool ceph_netfs_issue_op_inline(struct netfs_io_subrequest *subreq)
 	size_t len;
 	int mode;
 
-	if (rreq->origin != NETFS_DIO_READ)
+	if (rreq->origin != NETFS_UNBUFFERED_READ &&
+	    rreq->origin != NETFS_DIO_READ)
 		__set_bit(NETFS_SREQ_CLEAR_TAIL, &subreq->flags);
 	__clear_bit(NETFS_SREQ_COPY_TO_CACHE, &subreq->flags);
 

fs/netfs/direct_read.c

@@ -185,7 +185,8 @@ ssize_t netfs_unbuffered_read_iter_locked(struct kiocb *iocb, struct iov_iter *i
 
 	rreq = netfs_alloc_request(iocb->ki_filp->f_mapping, iocb->ki_filp,
 				   iocb->ki_pos, orig_count,
-				   NETFS_DIO_READ);
+				   iocb->ki_flags & IOCB_DIRECT ?
+				   NETFS_DIO_READ : NETFS_UNBUFFERED_READ);
 	if (IS_ERR(rreq))
 		return PTR_ERR(rreq);
 

fs/netfs/main.c

@@ -39,6 +39,7 @@ static const char *netfs_origins[nr__netfs_io_origin] = {
 	[NETFS_READ_GAPS]		= "RG",
 	[NETFS_READ_SINGLE]		= "R1",
 	[NETFS_READ_FOR_WRITE]		= "RW",
+	[NETFS_UNBUFFERED_READ]		= "UR",
 	[NETFS_DIO_READ]		= "DR",
 	[NETFS_WRITEBACK]		= "WB",
 	[NETFS_WRITEBACK_SINGLE]	= "W1",

fs/netfs/misc.c

@@ -461,6 +461,7 @@ static ssize_t netfs_wait_for_request(struct netfs_io_request *rreq,
 		case NETFS_DIO_READ:
 		case NETFS_DIO_WRITE:
 		case NETFS_READ_SINGLE:
+		case NETFS_UNBUFFERED_READ:
 		case NETFS_UNBUFFERED_WRITE:
 			break;
 		default:

fs/netfs/objects.c

@@ -59,6 +59,7 @@ struct netfs_io_request *netfs_alloc_request(struct address_space *mapping,
 	    origin == NETFS_READ_GAPS ||
 	    origin == NETFS_READ_SINGLE ||
 	    origin == NETFS_READ_FOR_WRITE ||
+	    origin == NETFS_UNBUFFERED_READ ||
 	    origin == NETFS_DIO_READ) {
 		INIT_WORK(&rreq->work, netfs_read_collection_worker);
 		rreq->io_streams[0].avail = true;

fs/netfs/read_collect.c

@@ -340,7 +340,8 @@ static void netfs_rreq_assess_dio(struct netfs_io_request *rreq)
 {
 	unsigned int i;
 
-	if (rreq->origin == NETFS_DIO_READ) {
+	if (rreq->origin == NETFS_UNBUFFERED_READ ||
+	    rreq->origin == NETFS_DIO_READ) {
 		for (i = 0; i < rreq->direct_bv_count; i++) {
 			flush_dcache_page(rreq->direct_bv[i].bv_page);
 			// TODO: cifs marks pages in the destination buffer
@@ -358,7 +359,8 @@ static void netfs_rreq_assess_dio(struct netfs_io_request *rreq)
 	}
 	if (rreq->netfs_ops->done)
 		rreq->netfs_ops->done(rreq);
-	if (rreq->origin == NETFS_DIO_READ)
+	if (rreq->origin == NETFS_UNBUFFERED_READ ||
+	    rreq->origin == NETFS_DIO_READ)
 		inode_dio_end(rreq->inode);
 }
 
@@ -414,6 +416,7 @@ bool netfs_read_collection(struct netfs_io_request *rreq)
 	//netfs_rreq_is_still_valid(rreq);
 
 	switch (rreq->origin) {
+	case NETFS_UNBUFFERED_READ:
 	case NETFS_DIO_READ:
 	case NETFS_READ_GAPS:
 		netfs_rreq_assess_dio(rreq);

fs/nfs/fscache.c

@@ -367,6 +367,7 @@ void nfs_netfs_read_completion(struct nfs_pgio_header *hdr)
 
 	sreq = netfs->sreq;
 	if (test_bit(NFS_IOHDR_EOF, &hdr->flags) &&
+	    sreq->rreq->origin != NETFS_UNBUFFERED_READ &&
 	    sreq->rreq->origin != NETFS_DIO_READ)
 		__set_bit(NETFS_SREQ_CLEAR_TAIL, &sreq->flags);
 

fs/smb/client/file.c

@@ -219,7 +219,8 @@ static void cifs_issue_read(struct netfs_io_subrequest *subreq)
 			goto failed;
 	}
 
-	if (subreq->rreq->origin != NETFS_DIO_READ)
+	if (subreq->rreq->origin != NETFS_UNBUFFERED_READ &&
+	    subreq->rreq->origin != NETFS_DIO_READ)
 		__set_bit(NETFS_SREQ_CLEAR_TAIL, &subreq->flags);
 
 	trace_netfs_sreq(subreq, netfs_sreq_trace_submit);