SUNRPC: Use RMW bitops in single-threaded hot paths

I noticed CPU pipeline stalls while using perf.

Once an svc thread is scheduled and executing an RPC, no other
processes will touch svc_rqst::rq_flags. Thus bus-locked atomics are
not needed outside the svc thread scheduler.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

fs/nfsd/nfs4proc.c

@@ -970,7 +970,7 @@ nfsd4_read(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 	 * the client wants us to do more in this compound:
 	 */
 	if (!nfsd4_last_compound_op(rqstp))
-		clear_bit(RQ_SPLICE_OK, &rqstp->rq_flags);
+		__clear_bit(RQ_SPLICE_OK, &rqstp->rq_flags);
 
 	/* check stateid */
 	status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
@@ -2650,11 +2650,12 @@ nfsd4_proc_compound(struct svc_rqst *rqstp)
 	cstate->minorversion = args->minorversion;
 	fh_init(current_fh, NFS4_FHSIZE);
 	fh_init(save_fh, NFS4_FHSIZE);
+
 	/*
 	 * Don't use the deferral mechanism for NFSv4; compounds make it
 	 * too hard to avoid non-idempotency problems.
 	 */
-	clear_bit(RQ_USEDEFERRAL, &rqstp->rq_flags);
+	__clear_bit(RQ_USEDEFERRAL, &rqstp->rq_flags);
 
 	/*
 	 * According to RFC3010, this takes precedence over all other errors.
@@ -2769,7 +2770,7 @@ nfsd4_proc_compound(struct svc_rqst *rqstp)
 out:
 	cstate->status = status;
 	/* Reset deferral mechanism for RPC deferrals */
-	set_bit(RQ_USEDEFERRAL, &rqstp->rq_flags);
+	__set_bit(RQ_USEDEFERRAL, &rqstp->rq_flags);
 	return rpc_success;
 }
 

fs/nfsd/nfs4xdr.c

@@ -2411,7 +2411,7 @@ nfsd4_decode_compound(struct nfsd4_compoundargs *argp)
 	argp->rqstp->rq_cachetype = cachethis ? RC_REPLBUFF : RC_NOCACHE;
 
 	if (readcount > 1 || max_reply > PAGE_SIZE - auth_slack)
-		clear_bit(RQ_SPLICE_OK, &argp->rqstp->rq_flags);
+		__clear_bit(RQ_SPLICE_OK, &argp->rqstp->rq_flags);
 
 	return true;
 }

net/sunrpc/auth_gss/svcauth_gss.c

@@ -900,7 +900,7 @@ unwrap_integ_data(struct svc_rqst *rqstp, struct xdr_buf *buf, u32 seq, struct g
 	 * rejecting the server-computed MIC in this somewhat rare case,
 	 * do not use splice with the GSS integrity service.
 	 */
-	clear_bit(RQ_SPLICE_OK, &rqstp->rq_flags);
+	__clear_bit(RQ_SPLICE_OK, &rqstp->rq_flags);
 
 	/* Did we already verify the signature on the original pass through? */
 	if (rqstp->rq_deferred)
@@ -972,7 +972,7 @@ unwrap_priv_data(struct svc_rqst *rqstp, struct xdr_buf *buf, u32 seq, struct gs
 	int pad, remaining_len, offset;
 	u32 rseqno;
 
-	clear_bit(RQ_SPLICE_OK, &rqstp->rq_flags);
+	__clear_bit(RQ_SPLICE_OK, &rqstp->rq_flags);
 
 	priv_len = svc_getnl(&buf->head[0]);
 	if (rqstp->rq_deferred) {

net/sunrpc/svc.c

@@ -1244,10 +1244,10 @@ svc_process_common(struct svc_rqst *rqstp, struct kvec *argv, struct kvec *resv)
 		goto err_short_len;
 
 	/* Will be turned off by GSS integrity and privacy services */
-	set_bit(RQ_SPLICE_OK, &rqstp->rq_flags);
+	__set_bit(RQ_SPLICE_OK, &rqstp->rq_flags);
 	/* Will be turned off only when NFSv4 Sessions are used */
-	set_bit(RQ_USEDEFERRAL, &rqstp->rq_flags);
-	clear_bit(RQ_DROPME, &rqstp->rq_flags);
+	__set_bit(RQ_USEDEFERRAL, &rqstp->rq_flags);
+	__clear_bit(RQ_DROPME, &rqstp->rq_flags);
 
 	svc_putu32(resv, rqstp->rq_xid);
 

net/sunrpc/svc_xprt.c

@@ -1238,7 +1238,7 @@ static struct cache_deferred_req *svc_defer(struct cache_req *req)
 	trace_svc_defer(rqstp);
 	svc_xprt_get(rqstp->rq_xprt);
 	dr->xprt = rqstp->rq_xprt;
-	set_bit(RQ_DROPME, &rqstp->rq_flags);
+	__set_bit(RQ_DROPME, &rqstp->rq_flags);
 
 	dr->handle.revisit = svc_revisit;
 	return &dr->handle;

net/sunrpc/svcsock.c

@@ -298,9 +298,9 @@ static void svc_sock_setbufsize(struct svc_sock *svsk, unsigned int nreqs)
 static void svc_sock_secure_port(struct svc_rqst *rqstp)
 {
 	if (svc_port_is_privileged(svc_addr(rqstp)))
-		set_bit(RQ_SECURE, &rqstp->rq_flags);
+		__set_bit(RQ_SECURE, &rqstp->rq_flags);
 	else
-		clear_bit(RQ_SECURE, &rqstp->rq_flags);
+		__clear_bit(RQ_SECURE, &rqstp->rq_flags);
 }
 
 /*
@@ -1008,9 +1008,9 @@ static int svc_tcp_recvfrom(struct svc_rqst *rqstp)
 	rqstp->rq_xprt_ctxt   = NULL;
 	rqstp->rq_prot	      = IPPROTO_TCP;
 	if (test_bit(XPT_LOCAL, &svsk->sk_xprt.xpt_flags))
-		set_bit(RQ_LOCAL, &rqstp->rq_flags);
+		__set_bit(RQ_LOCAL, &rqstp->rq_flags);
 	else
-		clear_bit(RQ_LOCAL, &rqstp->rq_flags);
+		__clear_bit(RQ_LOCAL, &rqstp->rq_flags);
 
 	p = (__be32 *)rqstp->rq_arg.head[0].iov_base;
 	calldir = p[1];

net/sunrpc/xprtrdma/svc_rdma_transport.c

@@ -602,7 +602,7 @@ static int svc_rdma_has_wspace(struct svc_xprt *xprt)
 
 static void svc_rdma_secure_port(struct svc_rqst *rqstp)
 {
-	set_bit(RQ_SECURE, &rqstp->rq_flags);
+	__set_bit(RQ_SECURE, &rqstp->rq_flags);
 }
 
 static void svc_rdma_kill_temp_xprt(struct svc_xprt *xprt)