[linux-block.git] / net / sunrpc / xprtrdma / backchannel.c

/*
 * Copyright (c) 2015 Oracle.  All rights reserved.
 *
 * Support for backward direction RPCs on RPC/RDMA.
 */

#include <linux/module.h>
#include <linux/sunrpc/xprt.h>
#include <linux/sunrpc/svc.h>
#include <linux/sunrpc/svc_xprt.h>

#include "xprt_rdma.h"

#if IS_ENABLED(CONFIG_SUNRPC_DEBUG)
# define RPCDBG_FACILITY	RPCDBG_TRANS
#endif

#undef RPCRDMA_BACKCHANNEL_DEBUG

static void rpcrdma_bc_free_rqst(struct rpcrdma_xprt *r_xprt,
				 struct rpc_rqst *rqst)
{
	struct rpcrdma_buffer *buf = &r_xprt->rx_buf;
	struct rpcrdma_req *req = rpcr_to_rdmar(rqst);

	spin_lock(&buf->rb_reqslock);
	list_del(&req->rl_all);
	spin_unlock(&buf->rb_reqslock);

	rpcrdma_destroy_req(&r_xprt->rx_ia, req);

	kfree(rqst);
}

static int rpcrdma_bc_setup_rqst(struct rpcrdma_xprt *r_xprt,
				 struct rpc_rqst *rqst)
{
	struct rpcrdma_ia *ia = &r_xprt->rx_ia;
	struct rpcrdma_regbuf *rb;
	struct rpcrdma_req *req;
	size_t size;

	req = rpcrdma_create_req(r_xprt);
	if (IS_ERR(req))
		return PTR_ERR(req);
	req->rl_backchannel = true;

	size = r_xprt->rx_data.inline_wsize;
	rb = rpcrdma_alloc_regbuf(ia, size, GFP_KERNEL);
	if (IS_ERR(rb))
		goto out_fail;
	req->rl_rdmabuf = rb;

	size += r_xprt->rx_data.inline_rsize;
	rb = rpcrdma_alloc_regbuf(ia, size, GFP_KERNEL);
	if (IS_ERR(rb))
		goto out_fail;
	rb->rg_owner = req;
	req->rl_sendbuf = rb;
	/* so that rpcr_to_rdmar works when receiving a request */
	rqst->rq_buffer = (void *)req->rl_sendbuf->rg_base;
	xdr_buf_init(&rqst->rq_snd_buf, rqst->rq_buffer, size);
	return 0;

out_fail:
	rpcrdma_bc_free_rqst(r_xprt, rqst);
	return -ENOMEM;
}

/* Allocate and add receive buffers to the rpcrdma_buffer's
 * existing list of rep's. These are released when the
 * transport is destroyed.
 */
static int rpcrdma_bc_setup_reps(struct rpcrdma_xprt *r_xprt,
				 unsigned int count)
{
	struct rpcrdma_rep *rep;
	int rc = 0;

	while (count--) {
		rep = rpcrdma_create_rep(r_xprt);
		if (IS_ERR(rep)) {
			pr_err("RPC:       %s: reply buffer alloc failed\n",
			       __func__);
			rc = PTR_ERR(rep);
			break;
		}

		rpcrdma_recv_buffer_put(rep);
	}

	return rc;
}

/**
 * xprt_rdma_bc_setup - Pre-allocate resources for handling backchannel requests
 * @xprt: transport associated with these backchannel resources
 * @reqs: number of concurrent incoming requests to expect
 *
 * Returns 0 on success; otherwise a negative errno
 */
int xprt_rdma_bc_setup(struct rpc_xprt *xprt, unsigned int reqs)
{
	struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt);
	struct rpcrdma_buffer *buffer = &r_xprt->rx_buf;
	struct rpc_rqst *rqst;
	unsigned int i;
	int rc;

	/* The backchannel reply path returns each rpc_rqst to the
	 * bc_pa_list _after_ the reply is sent. If the server is
	 * faster than the client, it can send another backward
	 * direction request before the rpc_rqst is returned to the
	 * list. The client rejects the request in this case.
	 *
	 * Twice as many rpc_rqsts are prepared to ensure there is
	 * always an rpc_rqst available as soon as a reply is sent.
	 */
	if (reqs > RPCRDMA_BACKWARD_WRS >> 1)
		goto out_err;

	for (i = 0; i < (reqs << 1); i++) {
		rqst = kzalloc(sizeof(*rqst), GFP_KERNEL);
		if (!rqst) {
			pr_err("RPC:       %s: Failed to create bc rpc_rqst\n",
			       __func__);
			goto out_free;
		}
		dprintk("RPC:       %s: new rqst %p\n", __func__, rqst);

		rqst->rq_xprt = &r_xprt->rx_xprt;
		INIT_LIST_HEAD(&rqst->rq_list);
		INIT_LIST_HEAD(&rqst->rq_bc_list);

		if (rpcrdma_bc_setup_rqst(r_xprt, rqst))
			goto out_free;

		spin_lock_bh(&xprt->bc_pa_lock);
		list_add(&rqst->rq_bc_pa_list, &xprt->bc_pa_list);
		spin_unlock_bh(&xprt->bc_pa_lock);
	}

	rc = rpcrdma_bc_setup_reps(r_xprt, reqs);
	if (rc)
		goto out_free;

	rc = rpcrdma_ep_post_extra_recv(r_xprt, reqs);
	if (rc)
		goto out_free;

	buffer->rb_bc_srv_max_requests = reqs;
	request_module("svcrdma");

	return 0;

out_free:
	xprt_rdma_bc_destroy(xprt, reqs);

out_err:
	pr_err("RPC:       %s: setup backchannel transport failed\n", __func__);
	return -ENOMEM;
}

/**
 * xprt_rdma_bc_up - Create transport endpoint for backchannel service
 * @serv: server endpoint
 * @net: network namespace
 *
 * The "xprt" is an implied argument: it supplies the name of the
 * backchannel transport class.
 *
 * Returns zero on success, negative errno on failure
 */
int xprt_rdma_bc_up(struct svc_serv *serv, struct net *net)
{
	int ret;

	ret = svc_create_xprt(serv, "rdma-bc", net, PF_INET, 0, 0);
	if (ret < 0)
		return ret;
	return 0;
}

/**
 * xprt_rdma_bc_maxpayload - Return maximum backchannel message size
 * @xprt: transport
 *
 * Returns maximum size, in bytes, of a backchannel message
 */
size_t xprt_rdma_bc_maxpayload(struct rpc_xprt *xprt)
{
	struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt);
	struct rpcrdma_create_data_internal *cdata = &r_xprt->rx_data;
	size_t maxmsg;

	maxmsg = min_t(unsigned int, cdata->inline_rsize, cdata->inline_wsize);
	return maxmsg - RPCRDMA_HDRLEN_MIN;
}

/**
 * rpcrdma_bc_marshal_reply - Send backwards direction reply
 * @rqst: buffer containing RPC reply data
 *
 * Returns zero on success.
 */
int rpcrdma_bc_marshal_reply(struct rpc_rqst *rqst)
{
	struct rpc_xprt *xprt = rqst->rq_xprt;
	struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt);
	struct rpcrdma_req *req = rpcr_to_rdmar(rqst);
	struct rpcrdma_msg *headerp;
	size_t rpclen;

	headerp = rdmab_to_msg(req->rl_rdmabuf);
	headerp->rm_xid = rqst->rq_xid;
	headerp->rm_vers = rpcrdma_version;
	headerp->rm_credit =
			cpu_to_be32(r_xprt->rx_buf.rb_bc_srv_max_requests);
	headerp->rm_type = rdma_msg;
	headerp->rm_body.rm_chunks[0] = xdr_zero;
	headerp->rm_body.rm_chunks[1] = xdr_zero;
	headerp->rm_body.rm_chunks[2] = xdr_zero;

	rpclen = rqst->rq_svec[0].iov_len;

#ifdef RPCRDMA_BACKCHANNEL_DEBUG
	pr_info("RPC:       %s: rpclen %zd headerp 0x%p lkey 0x%x\n",
		__func__, rpclen, headerp, rdmab_lkey(req->rl_rdmabuf));
	pr_info("RPC:       %s: RPC/RDMA: %*ph\n",
		__func__, (int)RPCRDMA_HDRLEN_MIN, headerp);
	pr_info("RPC:       %s:      RPC: %*ph\n",
		__func__, (int)rpclen, rqst->rq_svec[0].iov_base);
#endif

	req->rl_send_iov[0].addr = rdmab_addr(req->rl_rdmabuf);
	req->rl_send_iov[0].length = RPCRDMA_HDRLEN_MIN;
	req->rl_send_iov[0].lkey = rdmab_lkey(req->rl_rdmabuf);

	req->rl_send_iov[1].addr = rdmab_addr(req->rl_sendbuf);
	req->rl_send_iov[1].length = rpclen;
	req->rl_send_iov[1].lkey = rdmab_lkey(req->rl_sendbuf);

	req->rl_niovs = 2;
	return 0;
}

/**
 * xprt_rdma_bc_destroy - Release resources for handling backchannel requests
 * @xprt: transport associated with these backchannel resources
 * @reqs: number of incoming requests to destroy; ignored
 */
void xprt_rdma_bc_destroy(struct rpc_xprt *xprt, unsigned int reqs)
{
	struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt);
	struct rpc_rqst *rqst, *tmp;

	spin_lock_bh(&xprt->bc_pa_lock);
	list_for_each_entry_safe(rqst, tmp, &xprt->bc_pa_list, rq_bc_pa_list) {
		list_del(&rqst->rq_bc_pa_list);
		spin_unlock_bh(&xprt->bc_pa_lock);

		rpcrdma_bc_free_rqst(r_xprt, rqst);

		spin_lock_bh(&xprt->bc_pa_lock);
	}
	spin_unlock_bh(&xprt->bc_pa_lock);
}

/**
 * xprt_rdma_bc_free_rqst - Release a backchannel rqst
 * @rqst: request to release
 */
void xprt_rdma_bc_free_rqst(struct rpc_rqst *rqst)
{
	struct rpc_xprt *xprt = rqst->rq_xprt;

	dprintk("RPC:       %s: freeing rqst %p (req %p)\n",
		__func__, rqst, rpcr_to_rdmar(rqst));

	smp_mb__before_atomic();
	WARN_ON_ONCE(!test_bit(RPC_BC_PA_IN_USE, &rqst->rq_bc_pa_state));
	clear_bit(RPC_BC_PA_IN_USE, &rqst->rq_bc_pa_state);
	smp_mb__after_atomic();

	spin_lock_bh(&xprt->bc_pa_lock);
	list_add_tail(&rqst->rq_bc_pa_list, &xprt->bc_pa_list);
	spin_unlock_bh(&xprt->bc_pa_lock);
}

/**
 * rpcrdma_bc_receive_call - Handle a backward direction call
 * @xprt: transport receiving the call
 * @rep: receive buffer containing the call
 *
 * Called in the RPC reply handler, which runs in a tasklet.
 * Be quick about it.
 *
 * Operational assumptions:
 *    o Backchannel credits are ignored, just as the NFS server
 *      forechannel currently does
 *    o The ULP manages a replay cache (eg, NFSv4.1 sessions).
 *      No replay detection is done at the transport level
 */
void rpcrdma_bc_receive_call(struct rpcrdma_xprt *r_xprt,
			     struct rpcrdma_rep *rep)
{
	struct rpc_xprt *xprt = &r_xprt->rx_xprt;
	struct rpcrdma_msg *headerp;
	struct svc_serv *bc_serv;
	struct rpcrdma_req *req;
	struct rpc_rqst *rqst;
	struct xdr_buf *buf;
	size_t size;
	__be32 *p;

	headerp = rdmab_to_msg(rep->rr_rdmabuf);
#ifdef RPCRDMA_BACKCHANNEL_DEBUG
	pr_info("RPC:       %s: callback XID %08x, length=%u\n",
		__func__, be32_to_cpu(headerp->rm_xid), rep->rr_len);
	pr_info("RPC:       %s: %*ph\n", __func__, rep->rr_len, headerp);
#endif

	/* Sanity check:
	 * Need at least enough bytes for RPC/RDMA header, as code
	 * here references the header fields by array offset. Also,
	 * backward calls are always inline, so ensure there
	 * are some bytes beyond the RPC/RDMA header.
	 */
	if (rep->rr_len < RPCRDMA_HDRLEN_MIN + 24)
		goto out_short;
	p = (__be32 *)((unsigned char *)headerp + RPCRDMA_HDRLEN_MIN);
	size = rep->rr_len - RPCRDMA_HDRLEN_MIN;

	/* Grab a free bc rqst */
	spin_lock(&xprt->bc_pa_lock);
	if (list_empty(&xprt->bc_pa_list)) {
		spin_unlock(&xprt->bc_pa_lock);
		goto out_overflow;
	}
	rqst = list_first_entry(&xprt->bc_pa_list,
				struct rpc_rqst, rq_bc_pa_list);
	list_del(&rqst->rq_bc_pa_list);
	spin_unlock(&xprt->bc_pa_lock);
	dprintk("RPC:       %s: using rqst %p\n", __func__, rqst);

	/* Prepare rqst */
	rqst->rq_reply_bytes_recvd = 0;
	rqst->rq_bytes_sent = 0;
	rqst->rq_xid = headerp->rm_xid;

	rqst->rq_private_buf.len = size;
	set_bit(RPC_BC_PA_IN_USE, &rqst->rq_bc_pa_state);

	buf = &rqst->rq_rcv_buf;
	memset(buf, 0, sizeof(*buf));
	buf->head[0].iov_base = p;
	buf->head[0].iov_len = size;
	buf->len = size;

	/* The receive buffer has to be hooked to the rpcrdma_req
	 * so that it can be reposted after the server is done
	 * parsing it but just before sending the backward
	 * direction reply.
	 */
	req = rpcr_to_rdmar(rqst);
	dprintk("RPC:       %s: attaching rep %p to req %p\n",
		__func__, rep, req);
	req->rl_reply = rep;

	/* Defeat the retransmit detection logic in send_request */
	req->rl_connect_cookie = 0;

	/* Queue rqst for ULP's callback service */
	bc_serv = xprt->bc_serv;
	spin_lock(&bc_serv->sv_cb_lock);
	list_add(&rqst->rq_bc_list, &bc_serv->sv_cb_list);
	spin_unlock(&bc_serv->sv_cb_lock);

	wake_up(&bc_serv->sv_cb_waitq);

	r_xprt->rx_stats.bcall_count++;
	return;

out_overflow:
	pr_warn("RPC/RDMA backchannel overflow\n");
	xprt_disconnect_done(xprt);
	/* This receive buffer gets reposted automatically
	 * when the connection is re-established.
	 */
	return;

out_short:
	pr_warn("RPC/RDMA short backward direction call\n");

	if (rpcrdma_ep_post_recv(&r_xprt->rx_ia, &r_xprt->rx_ep, rep))
		xprt_disconnect_done(xprt);
	else
		pr_warn("RPC:       %s: reposting rep %p\n",
			__func__, rep);
}
Commit	Line	Data
f531a5db CL	1	/*
	2	* Copyright (c) 2015 Oracle. All rights reserved.
	3	*
	4	* Support for backward direction RPCs on RPC/RDMA.
	5	*/
	6
	7	#include <linux/module.h>
63cae470 CL	8	#include <linux/sunrpc/xprt.h>
63cae470 CL	9	#include <linux/sunrpc/svc.h>
76566773	10	#include <linux/sunrpc/svc_xprt.h>
f531a5db CL	11
	12	#include "xprt_rdma.h"
	13
	14	#if IS_ENABLED(CONFIG_SUNRPC_DEBUG)
	15	# define RPCDBG_FACILITY RPCDBG_TRANS
	16	#endif
	17
c8bbe0c7	18	#undef RPCRDMA_BACKCHANNEL_DEBUG
63cae470	19
f531a5db CL	20	static void rpcrdma_bc_free_rqst(struct rpcrdma_xprt *r_xprt,
	21	struct rpc_rqst *rqst)
	22	{
	23	struct rpcrdma_buffer *buf = &r_xprt->rx_buf;
	24	struct rpcrdma_req *req = rpcr_to_rdmar(rqst);
	25
	26	spin_lock(&buf->rb_reqslock);
	27	list_del(&req->rl_all);
	28	spin_unlock(&buf->rb_reqslock);
	29
	30	rpcrdma_destroy_req(&r_xprt->rx_ia, req);
	31
	32	kfree(rqst);
	33	}
	34
	35	static int rpcrdma_bc_setup_rqst(struct rpcrdma_xprt *r_xprt,
	36	struct rpc_rqst *rqst)
	37	{
	38	struct rpcrdma_ia *ia = &r_xprt->rx_ia;
	39	struct rpcrdma_regbuf *rb;
	40	struct rpcrdma_req *req;
f531a5db CL	41	size_t size;
	42
	43	req = rpcrdma_create_req(r_xprt);
abfb6897 DC	44	if (IS_ERR(req))
abfb6897 DC	45	return PTR_ERR(req);
f531a5db CL	46	req->rl_backchannel = true;
f531a5db CL	47
eb342e9a	48	size = r_xprt->rx_data.inline_wsize;
f531a5db CL	49	rb = rpcrdma_alloc_regbuf(ia, size, GFP_KERNEL);
	50	if (IS_ERR(rb))
	51	goto out_fail;
	52	req->rl_rdmabuf = rb;
	53
eb342e9a	54	size += r_xprt->rx_data.inline_rsize;
f531a5db CL	55	rb = rpcrdma_alloc_regbuf(ia, size, GFP_KERNEL);
	56	if (IS_ERR(rb))
	57	goto out_fail;
	58	rb->rg_owner = req;
	59	req->rl_sendbuf = rb;
	60	/* so that rpcr_to_rdmar works when receiving a request */
	61	rqst->rq_buffer = (void *)req->rl_sendbuf->rg_base;
b9c5bc03	62	xdr_buf_init(&rqst->rq_snd_buf, rqst->rq_buffer, size);
f531a5db CL	63	return 0;
	64
	65	out_fail:
	66	rpcrdma_bc_free_rqst(r_xprt, rqst);
	67	return -ENOMEM;
	68	}
	69
	70	/* Allocate and add receive buffers to the rpcrdma_buffer's
	71	* existing list of rep's. These are released when the
	72	* transport is destroyed.
	73	*/
	74	static int rpcrdma_bc_setup_reps(struct rpcrdma_xprt *r_xprt,
	75	unsigned int count)
	76	{
f531a5db	77	struct rpcrdma_rep *rep;
f531a5db CL	78	int rc = 0;
	79
	80	while (count--) {
	81	rep = rpcrdma_create_rep(r_xprt);
	82	if (IS_ERR(rep)) {
	83	pr_err("RPC: %s: reply buffer alloc failed\n",
	84	__func__);
	85	rc = PTR_ERR(rep);
	86	break;
	87	}
	88
9b06688b	89	rpcrdma_recv_buffer_put(rep);
f531a5db CL	90	}
	91
	92	return rc;
	93	}
	94
	95	/**
	96	* xprt_rdma_bc_setup - Pre-allocate resources for handling backchannel requests
	97	* @xprt: transport associated with these backchannel resources
	98	* @reqs: number of concurrent incoming requests to expect
	99	*
	100	* Returns 0 on success; otherwise a negative errno
	101	*/
	102	int xprt_rdma_bc_setup(struct rpc_xprt *xprt, unsigned int reqs)
	103	{
	104	struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt);
	105	struct rpcrdma_buffer *buffer = &r_xprt->rx_buf;
	106	struct rpc_rqst *rqst;
	107	unsigned int i;
	108	int rc;
	109
	110	/* The backchannel reply path returns each rpc_rqst to the
	111	* bc_pa_list _after_ the reply is sent. If the server is
	112	* faster than the client, it can send another backward
	113	* direction request before the rpc_rqst is returned to the
	114	* list. The client rejects the request in this case.
	115	*
	116	* Twice as many rpc_rqsts are prepared to ensure there is
	117	* always an rpc_rqst available as soon as a reply is sent.
	118	*/
124fa17d CL	119	if (reqs > RPCRDMA_BACKWARD_WRS >> 1)
	120	goto out_err;
	121
f531a5db CL	122	for (i = 0; i < (reqs << 1); i++) {
	123	rqst = kzalloc(sizeof(*rqst), GFP_KERNEL);
	124	if (!rqst) {
	125	pr_err("RPC: %s: Failed to create bc rpc_rqst\n",
	126	__func__);
	127	goto out_free;
	128	}
c8bbe0c7	129	dprintk("RPC: %s: new rqst %p\n", __func__, rqst);
f531a5db CL	130
	131	rqst->rq_xprt = &r_xprt->rx_xprt;
	132	INIT_LIST_HEAD(&rqst->rq_list);
	133	INIT_LIST_HEAD(&rqst->rq_bc_list);
	134
	135	if (rpcrdma_bc_setup_rqst(r_xprt, rqst))
	136	goto out_free;
	137
	138	spin_lock_bh(&xprt->bc_pa_lock);
	139	list_add(&rqst->rq_bc_pa_list, &xprt->bc_pa_list);
	140	spin_unlock_bh(&xprt->bc_pa_lock);
	141	}
	142
	143	rc = rpcrdma_bc_setup_reps(r_xprt, reqs);
	144	if (rc)
	145	goto out_free;
	146
	147	rc = rpcrdma_ep_post_extra_recv(r_xprt, reqs);
	148	if (rc)
	149	goto out_free;
	150
	151	buffer->rb_bc_srv_max_requests = reqs;
	152	request_module("svcrdma");
	153
	154	return 0;
	155
	156	out_free:
	157	xprt_rdma_bc_destroy(xprt, reqs);
	158
124fa17d	159	out_err:
f531a5db CL	160	pr_err("RPC: %s: setup backchannel transport failed\n", __func__);
	161	return -ENOMEM;
	162	}
	163
76566773 CL	164	/**
	165	* xprt_rdma_bc_up - Create transport endpoint for backchannel service
	166	* @serv: server endpoint
	167	* @net: network namespace
	168	*
	169	* The "xprt" is an implied argument: it supplies the name of the
	170	* backchannel transport class.
	171	*
	172	* Returns zero on success, negative errno on failure
	173	*/
	174	int xprt_rdma_bc_up(struct svc_serv serv, struct net net)
	175	{
	176	int ret;
	177
	178	ret = svc_create_xprt(serv, "rdma-bc", net, PF_INET, 0, 0);
	179	if (ret < 0)
	180	return ret;
	181	return 0;
	182	}
	183
6b26cc8c CL	184	/**
	185	* xprt_rdma_bc_maxpayload - Return maximum backchannel message size
	186	* @xprt: transport
	187	*
	188	* Returns maximum size, in bytes, of a backchannel message
	189	*/
	190	size_t xprt_rdma_bc_maxpayload(struct rpc_xprt *xprt)
	191	{
	192	struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt);
	193	struct rpcrdma_create_data_internal *cdata = &r_xprt->rx_data;
	194	size_t maxmsg;
	195
	196	maxmsg = min_t(unsigned int, cdata->inline_rsize, cdata->inline_wsize);
	197	return maxmsg - RPCRDMA_HDRLEN_MIN;
	198	}
	199
83128a60 CL	200	/**
	201	* rpcrdma_bc_marshal_reply - Send backwards direction reply
	202	* @rqst: buffer containing RPC reply data
	203	*
	204	* Returns zero on success.
	205	*/
	206	int rpcrdma_bc_marshal_reply(struct rpc_rqst *rqst)
	207	{
	208	struct rpc_xprt *xprt = rqst->rq_xprt;
	209	struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt);
	210	struct rpcrdma_req *req = rpcr_to_rdmar(rqst);
	211	struct rpcrdma_msg *headerp;
	212	size_t rpclen;
	213
	214	headerp = rdmab_to_msg(req->rl_rdmabuf);
	215	headerp->rm_xid = rqst->rq_xid;
	216	headerp->rm_vers = rpcrdma_version;
	217	headerp->rm_credit =
	218	cpu_to_be32(r_xprt->rx_buf.rb_bc_srv_max_requests);
	219	headerp->rm_type = rdma_msg;
	220	headerp->rm_body.rm_chunks[0] = xdr_zero;
	221	headerp->rm_body.rm_chunks[1] = xdr_zero;
	222	headerp->rm_body.rm_chunks[2] = xdr_zero;
	223
	224	rpclen = rqst->rq_svec[0].iov_len;
	225
c8bbe0c7	226	#ifdef RPCRDMA_BACKCHANNEL_DEBUG
83128a60 CL	227	pr_info("RPC: %s: rpclen %zd headerp 0x%p lkey 0x%x\n",
	228	__func__, rpclen, headerp, rdmab_lkey(req->rl_rdmabuf));
	229	pr_info("RPC: %s: RPC/RDMA: %*ph\n",
	230	__func__, (int)RPCRDMA_HDRLEN_MIN, headerp);
	231	pr_info("RPC: %s: RPC: %*ph\n",
	232	__func__, (int)rpclen, rqst->rq_svec[0].iov_base);
c8bbe0c7	233	#endif
83128a60 CL	234
	235	req->rl_send_iov[0].addr = rdmab_addr(req->rl_rdmabuf);
	236	req->rl_send_iov[0].length = RPCRDMA_HDRLEN_MIN;
	237	req->rl_send_iov[0].lkey = rdmab_lkey(req->rl_rdmabuf);
	238
	239	req->rl_send_iov[1].addr = rdmab_addr(req->rl_sendbuf);
	240	req->rl_send_iov[1].length = rpclen;
	241	req->rl_send_iov[1].lkey = rdmab_lkey(req->rl_sendbuf);
	242
	243	req->rl_niovs = 2;
	244	return 0;
	245	}
	246
f531a5db CL	247	/**
	248	* xprt_rdma_bc_destroy - Release resources for handling backchannel requests
	249	* @xprt: transport associated with these backchannel resources
	250	* @reqs: number of incoming requests to destroy; ignored
	251	*/
	252	void xprt_rdma_bc_destroy(struct rpc_xprt *xprt, unsigned int reqs)
	253	{
	254	struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt);
	255	struct rpc_rqst rqst, tmp;
	256
	257	spin_lock_bh(&xprt->bc_pa_lock);
	258	list_for_each_entry_safe(rqst, tmp, &xprt->bc_pa_list, rq_bc_pa_list) {
	259	list_del(&rqst->rq_bc_pa_list);
	260	spin_unlock_bh(&xprt->bc_pa_lock);
	261
	262	rpcrdma_bc_free_rqst(r_xprt, rqst);
	263
	264	spin_lock_bh(&xprt->bc_pa_lock);
	265	}
	266	spin_unlock_bh(&xprt->bc_pa_lock);
	267	}
	268
	269	/**
	270	* xprt_rdma_bc_free_rqst - Release a backchannel rqst
	271	* @rqst: request to release
	272	*/
	273	void xprt_rdma_bc_free_rqst(struct rpc_rqst *rqst)
	274	{
	275	struct rpc_xprt *xprt = rqst->rq_xprt;
	276
c8bbe0c7 CL	277	dprintk("RPC: %s: freeing rqst %p (req %p)\n",
	278	__func__, rqst, rpcr_to_rdmar(rqst));
	279
f531a5db CL	280	smp_mb__before_atomic();
	281	WARN_ON_ONCE(!test_bit(RPC_BC_PA_IN_USE, &rqst->rq_bc_pa_state));
	282	clear_bit(RPC_BC_PA_IN_USE, &rqst->rq_bc_pa_state);
	283	smp_mb__after_atomic();
	284
	285	spin_lock_bh(&xprt->bc_pa_lock);
	286	list_add_tail(&rqst->rq_bc_pa_list, &xprt->bc_pa_list);
	287	spin_unlock_bh(&xprt->bc_pa_lock);
	288	}
63cae470 CL	289
	290	/**
	291	* rpcrdma_bc_receive_call - Handle a backward direction call
	292	* @xprt: transport receiving the call
	293	* @rep: receive buffer containing the call
	294	*
	295	* Called in the RPC reply handler, which runs in a tasklet.
	296	* Be quick about it.
	297	*
	298	* Operational assumptions:
	299	* o Backchannel credits are ignored, just as the NFS server
	300	* forechannel currently does
	301	* o The ULP manages a replay cache (eg, NFSv4.1 sessions).
	302	* No replay detection is done at the transport level
	303	*/
	304	void rpcrdma_bc_receive_call(struct rpcrdma_xprt *r_xprt,
	305	struct rpcrdma_rep *rep)
	306	{
	307	struct rpc_xprt *xprt = &r_xprt->rx_xprt;
	308	struct rpcrdma_msg *headerp;
	309	struct svc_serv *bc_serv;
	310	struct rpcrdma_req *req;
	311	struct rpc_rqst *rqst;
	312	struct xdr_buf *buf;
	313	size_t size;
	314	__be32 *p;
	315
	316	headerp = rdmab_to_msg(rep->rr_rdmabuf);
	317	#ifdef RPCRDMA_BACKCHANNEL_DEBUG
	318	pr_info("RPC: %s: callback XID %08x, length=%u\n",
	319	__func__, be32_to_cpu(headerp->rm_xid), rep->rr_len);
	320	pr_info("RPC: %s: %*ph\n", __func__, rep->rr_len, headerp);
	321	#endif
	322
	323	/* Sanity check:
	324	* Need at least enough bytes for RPC/RDMA header, as code
	325	* here references the header fields by array offset. Also,
	326	* backward calls are always inline, so ensure there
	327	* are some bytes beyond the RPC/RDMA header.
	328	*/
	329	if (rep->rr_len < RPCRDMA_HDRLEN_MIN + 24)
	330	goto out_short;
	331	p = (__be32 )((unsigned char )headerp + RPCRDMA_HDRLEN_MIN);
	332	size = rep->rr_len - RPCRDMA_HDRLEN_MIN;
	333
	334	/* Grab a free bc rqst */
	335	spin_lock(&xprt->bc_pa_lock);
	336	if (list_empty(&xprt->bc_pa_list)) {
	337	spin_unlock(&xprt->bc_pa_lock);
	338	goto out_overflow;
	339	}
	340	rqst = list_first_entry(&xprt->bc_pa_list,
	341	struct rpc_rqst, rq_bc_pa_list);
	342	list_del(&rqst->rq_bc_pa_list);
	343	spin_unlock(&xprt->bc_pa_lock);
c8bbe0c7	344	dprintk("RPC: %s: using rqst %p\n", __func__, rqst);
63cae470 CL	345
	346	/* Prepare rqst */
	347	rqst->rq_reply_bytes_recvd = 0;
	348	rqst->rq_bytes_sent = 0;
	349	rqst->rq_xid = headerp->rm_xid;
9f74660b CL	350
9f74660b CL	351	rqst->rq_private_buf.len = size;
63cae470 CL	352	set_bit(RPC_BC_PA_IN_USE, &rqst->rq_bc_pa_state);
	353
	354	buf = &rqst->rq_rcv_buf;
	355	memset(buf, 0, sizeof(*buf));
	356	buf->head[0].iov_base = p;
	357	buf->head[0].iov_len = size;
	358	buf->len = size;
	359
	360	/* The receive buffer has to be hooked to the rpcrdma_req
	361	* so that it can be reposted after the server is done
	362	* parsing it but just before sending the backward
	363	* direction reply.
	364	*/
	365	req = rpcr_to_rdmar(rqst);
c8bbe0c7	366	dprintk("RPC: %s: attaching rep %p to req %p\n",
63cae470	367	__func__, rep, req);
63cae470 CL	368	req->rl_reply = rep;
	369
	370	/* Defeat the retransmit detection logic in send_request */
	371	req->rl_connect_cookie = 0;
	372
	373	/* Queue rqst for ULP's callback service */
	374	bc_serv = xprt->bc_serv;
	375	spin_lock(&bc_serv->sv_cb_lock);
	376	list_add(&rqst->rq_bc_list, &bc_serv->sv_cb_list);
	377	spin_unlock(&bc_serv->sv_cb_lock);
	378
	379	wake_up(&bc_serv->sv_cb_waitq);
	380
	381	r_xprt->rx_stats.bcall_count++;
	382	return;
	383
	384	out_overflow:
	385	pr_warn("RPC/RDMA backchannel overflow\n");
	386	xprt_disconnect_done(xprt);
	387	/* This receive buffer gets reposted automatically
	388	* when the connection is re-established.
	389	*/
	390	return;
	391
	392	out_short:
	393	pr_warn("RPC/RDMA short backward direction call\n");
	394
	395	if (rpcrdma_ep_post_recv(&r_xprt->rx_ia, &r_xprt->rx_ep, rep))
	396	xprt_disconnect_done(xprt);
	397	else
	398	pr_warn("RPC: %s: reposting rep %p\n",
	399	__func__, rep);
	400	}