net/smc: make wr buffer count configurable

Think SMC_WR_BUF_CNT_SEND := SMC_WR_BUF_CNT used in send context and
SMC_WR_BUF_CNT_RECV := 3 * SMC_WR_BUF_CNT used in recv context. Those
get replaced with lgr->max_send_wr and lgr->max_recv_wr respective.

Please note that although with the default sysctl values
qp_attr.cap.max_send_wr == qp_attr.cap.max_recv_wr is maintained but
can not be assumed to be generally true any more. I see no downside to
that, but my confidence level is rather modest.

Signed-off-by: Halil Pasic <[e-mail]>
Reviewed-by: Sidraya Jayagond <[e-mail]>
Reviewed-by: Dust Li <[e-mail]>
Tested-by: Mahanta Jambigi <[e-mail]>
Link: https://patch.msgid.link/[e-mail]
Signed-off-by: Paolo Abeni <[e-mail]>

Documentation/networking/smc-sysctl.rst

···
 	acceptable value ranges from 16 to 255. Only for SMC-R v2.1 and later.
 
 	Default: 255
+
+smcr_max_send_wr - INTEGER
+	So-called work request buffers are SMCR link (and RDMA queue pair) level
+	resources necessary for performing RDMA operations. Since up to 255
+	connections can share a link group and thus also a link and the number
+	of the work request buffers is decided when the link is allocated,
+	depending on the workload it can be a bottleneck in a sense that threads
+	have to wait for work request buffers to become available. Before the
+	introduction of this control the maximal number of work request buffers
+	available on the send path used to be hard coded to 16. With this control
+	it becomes configurable. The acceptable range is between 2 and 2048.
+
+	Please be aware that all the buffers need to be allocated as a physically
+	continuous array in which each element is a single buffer and has the size
+	of SMC_WR_BUF_SIZE (48) bytes. If the allocation fails we give up much
+	like before having this control.
+
+	Default: 16
+
+smcr_max_recv_wr - INTEGER
+	So-called work request buffers are SMCR link (and RDMA queue pair) level
+	resources necessary for performing RDMA operations. Since up to 255
+	connections can share a link group and thus also a link and the number
+	of the work request buffers is decided when the link is allocated,
+	depending on the workload it can be a bottleneck in a sense that threads
+	have to wait for work request buffers to become available. Before the
+	introduction of this control the maximal number of work request buffers
+	available on the receive path used to be hard coded to 16. With this control
+	it becomes configurable. The acceptable range is between 2 and 2048.
+
+	Please be aware that all the buffers need to be allocated as a physically
+	continuous array in which each element is a single buffer and has the size
+	of SMC_WR_BUF_SIZE (48) bytes. If the allocation fails we give up much
+	like before having this control.
+
+	Default: 48

include/net/netns/smc.h

···
 	int				sysctl_rmem;
 	int				sysctl_max_links_per_lgr;
 	int				sysctl_max_conns_per_lgr;
+	unsigned int			sysctl_smcr_max_send_wr;
+	unsigned int			sysctl_smcr_max_recv_wr;
 };
 #endif

net/smc/smc_core.h

···
 					 * distributions may modify it to a value between
 					 * 16-255 as needed.
 					 */
+#define SMCR_MAX_SEND_WR_DEF	16	/* Default number of work requests per send queue */
+#define SMCR_MAX_RECV_WR_DEF	48	/* Default number of work requests per recv queue */
 
 struct smc_lgr_list {			/* list of link group definition */
 	struct list_head	list;
···
 						/* max conn can be assigned to lgr */
 			u8			max_links;
 						/* max links can be added in lgr */
+			u16			max_send_wr;
+						/* number of WR buffers on send */
+			u16			max_recv_wr;
+						/* number of WR buffers on recv */
 		};
 		struct { /* SMC-D */
 			struct smcd_gid		peer_gid;

net/smc/smc_ib.c

···
 		.recv_cq = lnk->smcibdev->roce_cq_recv,
 		.srq = NULL,
 		.cap = {
-				/* include unsolicited rdma_writes as well,
-				 * there are max. 2 RDMA_WRITE per 1 WR_SEND
-				 */
-			.max_send_wr = SMC_WR_BUF_CNT * 3,
-			.max_recv_wr = SMC_WR_BUF_CNT * 3,
 			.max_send_sge = SMC_IB_MAX_SEND_SGE,
 			.max_recv_sge = lnk->wr_rx_sge_cnt,
 			.max_inline_data = 0,
···
 	};
 	int rc;
 
+	/* include unsolicited rdma_writes as well,
+	 * there are max. 2 RDMA_WRITE per 1 WR_SEND
+	 */
+	qp_attr.cap.max_send_wr = 3 * lnk->lgr->max_send_wr;
+	qp_attr.cap.max_recv_wr = lnk->lgr->max_recv_wr;
 	lnk->roce_qp = ib_create_qp(lnk->roce_pd, &qp_attr);
 	rc = PTR_ERR_OR_ZERO(lnk->roce_qp);
 	if (IS_ERR(lnk->roce_qp))

net/smc/smc_llc.c

···
 	init_waitqueue_head(&lgr->llc_msg_waiter);
 	init_rwsem(&lgr->llc_conf_mutex);
 	lgr->llc_testlink_time = READ_ONCE(net->smc.sysctl_smcr_testlink_time);
+	lgr->max_send_wr = (u16)(READ_ONCE(net->smc.sysctl_smcr_max_send_wr));
+	lgr->max_recv_wr = (u16)(READ_ONCE(net->smc.sysctl_smcr_max_recv_wr));
 }
 
 /* called after lgr was removed from lgr_list */

net/smc/smc_sysctl.c

···
 static int links_per_lgr_max = SMC_LINKS_ADD_LNK_MAX;
 static int conns_per_lgr_min = SMC_CONN_PER_LGR_MIN;
 static int conns_per_lgr_max = SMC_CONN_PER_LGR_MAX;
+static unsigned int smcr_max_wr_min = 2;
+static unsigned int smcr_max_wr_max = 2048;
 
 static struct ctl_table smc_table[] = {
 	{
···
 		.extra1		= SYSCTL_ZERO,
 		.extra2		= SYSCTL_ONE,
 	},
+	{
+		.procname	= "smcr_max_send_wr",
+		.data		= &init_net.smc.sysctl_smcr_max_send_wr,
+		.maxlen		= sizeof(int),
+		.mode		= 0644,
+		.proc_handler	= proc_dointvec_minmax,
+		.extra1		= &smcr_max_wr_min,
+		.extra2		= &smcr_max_wr_max,
+	},
+	{
+		.procname	= "smcr_max_recv_wr",
+		.data		= &init_net.smc.sysctl_smcr_max_recv_wr,
+		.maxlen		= sizeof(int),
+		.mode		= 0644,
+		.proc_handler	= proc_dointvec_minmax,
+		.extra1		= &smcr_max_wr_min,
+		.extra2		= &smcr_max_wr_max,
+	},
 };
 
 int __net_init smc_sysctl_net_init(struct net *net)
···
 	WRITE_ONCE(net->smc.sysctl_rmem, net_smc_rmem_init);
 	net->smc.sysctl_max_links_per_lgr = SMC_LINKS_PER_LGR_MAX_PREFER;
 	net->smc.sysctl_max_conns_per_lgr = SMC_CONN_PER_LGR_PREFER;
+	net->smc.sysctl_smcr_max_send_wr = SMCR_MAX_SEND_WR_DEF;
+	net->smc.sysctl_smcr_max_recv_wr = SMCR_MAX_RECV_WR_DEF;
 	/* disable handshake limitation by default */
 	net->smc.limit_smc_hs = 0;
 

net/smc/smc_sysctl.h

···
 	net->smc.sysctl_autocorking_size = SMC_AUTOCORKING_DEFAULT_SIZE;
 	net->smc.sysctl_max_links_per_lgr = SMC_LINKS_PER_LGR_MAX_PREFER;
 	net->smc.sysctl_max_conns_per_lgr = SMC_CONN_PER_LGR_PREFER;
+	net->smc.sysctl_smcr_max_send_wr = SMCR_MAX_SEND_WR_DEF;
+	net->smc.sysctl_smcr_max_recv_wr = SMCR_MAX_RECV_WR_DEF;
 	return 0;
 }
 

net/smc/smc_wr.c

···
 		    IB_QP_DEST_QPN,
 		    &init_attr);
 
-	lnk->wr_tx_cnt = min_t(size_t, SMC_WR_BUF_CNT,
+	lnk->wr_tx_cnt = min_t(size_t, lnk->lgr->max_send_wr,
 			       lnk->qp_attr.cap.max_send_wr);
-	lnk->wr_rx_cnt = min_t(size_t, SMC_WR_BUF_CNT * 3,
+	lnk->wr_rx_cnt = min_t(size_t, lnk->lgr->max_recv_wr,
 			       lnk->qp_attr.cap.max_recv_wr);
 }
 
···
 int smc_wr_alloc_link_mem(struct smc_link *link)
 {
 	/* allocate link related memory */
-	link->wr_tx_bufs = kcalloc(SMC_WR_BUF_CNT, SMC_WR_BUF_SIZE, GFP_KERNEL);
+	link->wr_tx_bufs = kcalloc(link->lgr->max_send_wr,
+				   SMC_WR_BUF_SIZE, GFP_KERNEL);
 	if (!link->wr_tx_bufs)
 		goto no_mem;
-	link->wr_rx_bufs = kcalloc(SMC_WR_BUF_CNT * 3, link->wr_rx_buflen,
+	link->wr_rx_bufs = kcalloc(link->lgr->max_recv_wr, link->wr_rx_buflen,
 				   GFP_KERNEL);
 	if (!link->wr_rx_bufs)
 		goto no_mem_wr_tx_bufs;
-	link->wr_tx_ibs = kcalloc(SMC_WR_BUF_CNT, sizeof(link->wr_tx_ibs[0]),
-				  GFP_KERNEL);
+	link->wr_tx_ibs = kcalloc(link->lgr->max_send_wr,
+				  sizeof(link->wr_tx_ibs[0]), GFP_KERNEL);
 	if (!link->wr_tx_ibs)
 		goto no_mem_wr_rx_bufs;
-	link->wr_rx_ibs = kcalloc(SMC_WR_BUF_CNT * 3,
+	link->wr_rx_ibs = kcalloc(link->lgr->max_recv_wr,
 				  sizeof(link->wr_rx_ibs[0]),
 				  GFP_KERNEL);
 	if (!link->wr_rx_ibs)
 		goto no_mem_wr_tx_ibs;
-	link->wr_tx_rdmas = kcalloc(SMC_WR_BUF_CNT,
+	link->wr_tx_rdmas = kcalloc(link->lgr->max_send_wr,
 				    sizeof(link->wr_tx_rdmas[0]),
 				    GFP_KERNEL);
 	if (!link->wr_tx_rdmas)
 		goto no_mem_wr_rx_ibs;
-	link->wr_tx_rdma_sges = kcalloc(SMC_WR_BUF_CNT,
+	link->wr_tx_rdma_sges = kcalloc(link->lgr->max_send_wr,
 					sizeof(link->wr_tx_rdma_sges[0]),
 					GFP_KERNEL);
 	if (!link->wr_tx_rdma_sges)
 		goto no_mem_wr_tx_rdmas;
-	link->wr_tx_sges = kcalloc(SMC_WR_BUF_CNT, sizeof(link->wr_tx_sges[0]),
+	link->wr_tx_sges = kcalloc(link->lgr->max_send_wr, sizeof(link->wr_tx_sges[0]),
 				   GFP_KERNEL);
 	if (!link->wr_tx_sges)
 		goto no_mem_wr_tx_rdma_sges;
-	link->wr_rx_sges = kcalloc(SMC_WR_BUF_CNT * 3,
+	link->wr_rx_sges = kcalloc(link->lgr->max_recv_wr,
 				   sizeof(link->wr_rx_sges[0]) * link->wr_rx_sge_cnt,
 				   GFP_KERNEL);
 	if (!link->wr_rx_sges)
 		goto no_mem_wr_tx_sges;
-	link->wr_tx_mask = bitmap_zalloc(SMC_WR_BUF_CNT, GFP_KERNEL);
+	link->wr_tx_mask = bitmap_zalloc(link->lgr->max_send_wr, GFP_KERNEL);
 	if (!link->wr_tx_mask)
 		goto no_mem_wr_rx_sges;
-	link->wr_tx_pends = kcalloc(SMC_WR_BUF_CNT,
+	link->wr_tx_pends = kcalloc(link->lgr->max_send_wr,
 				    sizeof(link->wr_tx_pends[0]),
 				    GFP_KERNEL);
 	if (!link->wr_tx_pends)
 		goto no_mem_wr_tx_mask;
-	link->wr_tx_compl = kcalloc(SMC_WR_BUF_CNT,
+	link->wr_tx_compl = kcalloc(link->lgr->max_send_wr,
 				    sizeof(link->wr_tx_compl[0]),
 				    GFP_KERNEL);
 	if (!link->wr_tx_compl)
···
 		goto dma_unmap;
 	}
 	smc_wr_init_sge(lnk);
-	bitmap_zero(lnk->wr_tx_mask, SMC_WR_BUF_CNT);
+	bitmap_zero(lnk->wr_tx_mask, lnk->lgr->max_send_wr);
 	init_waitqueue_head(&lnk->wr_tx_wait);
 	rc = percpu_ref_init(&lnk->wr_tx_refs, smcr_wr_tx_refs_free, 0, GFP_KERNEL);
 	if (rc)

net/smc/smc_wr.h

···
 #include "smc.h"
 #include "smc_core.h"
 
-#define SMC_WR_BUF_CNT 16	/* # of ctrl buffers per link */
-
 #define SMC_WR_TX_WAIT_FREE_SLOT_TIME	(10 * HZ)
 
 #define SMC_WR_TX_SIZE 44 /* actual size of wr_send data (<=SMC_WR_BUF_SIZE) */