iw_cxgb4: Allocate and use IQs specifically for indirect interrupts

Currently indirect interrupts for RDMA CQs funnel through the LLD's RDMA
RXQs, which also handle direct interrupts for offload CPLs during RDMA
connection setup/teardown.  The intended T4 usage model, however, is to
have indirect interrupts flow through dedicated IQs. IE not to mix
indirect interrupts with CPL messages in an IQ.  This patch adds the
concept of RDMA concentrator IQs, or CIQs, setup and maintained by the
LLD and exported to iw_cxgb4 for use when creating CQs. RDMA CPLs will
flow through the LLD's RDMA RXQs, and CQ interrupts flow through the
CIQs.

Design:

cxgb4 creates and exports an array of CIQs for the RDMA ULD.  These IQs
are sized according to the max available CQs available at adapter init.
In addition, these IQs don't need FL buffers since they only service
indirect interrupts.  One CIQ is setup per RX channel similar to the
RDMA RXQs.

iw_cxgb4 will utilize these CIQs based on the vector value passed into
create_cq().  The num_comp_vectors advertised by iw_cxgb4 will be the
number of CIQs configured, and thus the vector value will be the index
into the array of CIQs.

Based on original work by Steve Wise <swise@opengridcomputing.com>

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Hariprasad Shenai <hariprasad@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

5 files changed, 76 insertions, 7 deletions

diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4.h b/drivers/net/ethernet/chelsio/cxgb4/cxgb4.h
index 32db3770926..f503dce4ab1 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4.h
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4.h
@@ -357,11 +357,17 @@ enum {
 	MAX_OFLD_QSETS = 16,          /* # of offload Tx/Rx queue sets */
 	MAX_CTRL_QUEUES = NCHAN,      /* # of control Tx queues */
 	MAX_RDMA_QUEUES = NCHAN,      /* # of streaming RDMA Rx queues */
+	MAX_RDMA_CIQS = NCHAN,        /* # of  RDMA concentrator IQs */
+	MAX_ISCSI_QUEUES = NCHAN,     /* # of streaming iSCSI Rx queues */
 };
 
 enum {
-	MAX_EGRQ = 128,         /* max # of egress queues, including FLs */
-	MAX_INGQ = 64           /* max # of interrupt-capable ingress queues */
+	INGQ_EXTRAS = 2,        /* firmware event queue and */
+				/*   forwarded interrupts */
+	MAX_EGRQ = MAX_ETH_QSETS*2 + MAX_OFLD_QSETS*2
+		   + MAX_CTRL_QUEUES + MAX_RDMA_QUEUES + MAX_ISCSI_QUEUES,
+	MAX_INGQ = MAX_ETH_QSETS + MAX_OFLD_QSETS + MAX_RDMA_QUEUES
+		   + MAX_RDMA_CIQS + MAX_ISCSI_QUEUES + INGQ_EXTRAS,
 };
 
 struct adapter;
@@ -538,6 +544,7 @@ struct sge {
 	struct sge_eth_rxq ethrxq[MAX_ETH_QSETS];
 	struct sge_ofld_rxq ofldrxq[MAX_OFLD_QSETS];
 	struct sge_ofld_rxq rdmarxq[MAX_RDMA_QUEUES];
+	struct sge_ofld_rxq rdmaciq[MAX_RDMA_CIQS];
 	struct sge_rspq fw_evtq ____cacheline_aligned_in_smp;
 
 	struct sge_rspq intrq ____cacheline_aligned_in_smp;
@@ -548,8 +555,10 @@ struct sge {
 	u16 ethtxq_rover;           /* Tx queue to clean up next */
 	u16 ofldqsets;              /* # of active offload queue sets */
 	u16 rdmaqs;                 /* # of available RDMA Rx queues */
+	u16 rdmaciqs;               /* # of available RDMA concentrator IQs */
 	u16 ofld_rxq[MAX_OFLD_QSETS];
 	u16 rdma_rxq[NCHAN];
+	u16 rdma_ciq[NCHAN];
 	u16 timer_val[SGE_NTIMERS];
 	u8 counter_val[SGE_NCOUNTERS];
 	u32 fl_pg_order;            /* large page allocation size */
@@ -577,6 +586,7 @@ struct sge {
 #define for_each_ethrxq(sge, i) for (i = 0; i < (sge)->ethqsets; i++)
 #define for_each_ofldrxq(sge, i) for (i = 0; i < (sge)->ofldqsets; i++)
 #define for_each_rdmarxq(sge, i) for (i = 0; i < (sge)->rdmaqs; i++)
+#define for_each_rdmaciq(sge, i) for (i = 0; i < (sge)->rdmaciqs; i++)
 
 struct l2t_data;
 
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c
index 8cf6be93f49..c26c3f8e279 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c
@@ -818,12 +818,17 @@ static void name_msix_vecs(struct adapter *adap)
 	for_each_rdmarxq(&adap->sge, i)
 		snprintf(adap->msix_info[msi_idx++].desc, n, "%s-rdma%d",
 			 adap->port[0]->name, i);
+
+	for_each_rdmaciq(&adap->sge, i)
+		snprintf(adap->msix_info[msi_idx++].desc, n, "%s-rdma-ciq%d",
+			 adap->port[0]->name, i);
 }
 
 static int request_msix_queue_irqs(struct adapter *adap)
 {
 	struct sge *s = &adap->sge;
-	int err, ethqidx, ofldqidx = 0, rdmaqidx = 0, msi_index = 2;
+	int err, ethqidx, ofldqidx = 0, rdmaqidx = 0, rdmaciqqidx = 0;
+	int msi_index = 2;
 
 	err = request_irq(adap->msix_info[1].vec, t4_sge_intr_msix, 0,
 			  adap->msix_info[1].desc, &s->fw_evtq);
@@ -857,9 +862,21 @@ static int request_msix_queue_irqs(struct adapter *adap)
 			goto unwind;
 		msi_index++;
 	}
+	for_each_rdmaciq(s, rdmaciqqidx) {
+		err = request_irq(adap->msix_info[msi_index].vec,
+				  t4_sge_intr_msix, 0,
+				  adap->msix_info[msi_index].desc,
+				  &s->rdmaciq[rdmaciqqidx].rspq);
+		if (err)
+			goto unwind;
+		msi_index++;
+	}
 	return 0;
 
 unwind:
+	while (--rdmaciqqidx >= 0)
+		free_irq(adap->msix_info[--msi_index].vec,
+			 &s->rdmaciq[rdmaciqqidx].rspq);
 	while (--rdmaqidx >= 0)
 		free_irq(adap->msix_info[--msi_index].vec,
 			 &s->rdmarxq[rdmaqidx].rspq);
@@ -885,6 +902,8 @@ static void free_msix_queue_irqs(struct adapter *adap)
 		free_irq(adap->msix_info[msi_index++].vec, &s->ofldrxq[i].rspq);
 	for_each_rdmarxq(s, i)
 		free_irq(adap->msix_info[msi_index++].vec, &s->rdmarxq[i].rspq);
+	for_each_rdmaciq(s, i)
+		free_irq(adap->msix_info[msi_index++].vec, &s->rdmaciq[i].rspq);
 }
 
 /**
@@ -1047,7 +1066,8 @@ freeout:	t4_free_sge_resources(adap);
 		if (msi_idx > 0)
 			msi_idx++;
 		err = t4_sge_alloc_rxq(adap, &q->rspq, false, dev, msi_idx,
-				       &q->fl, uldrx_handler);
+				       q->fl.size ? &q->fl : NULL,
+				       uldrx_handler);
 		if (err)
 			goto freeout;
 		memset(&q->stats, 0, sizeof(q->stats));
@@ -1064,13 +1084,28 @@ freeout:	t4_free_sge_resources(adap);
 		if (msi_idx > 0)
 			msi_idx++;
 		err = t4_sge_alloc_rxq(adap, &q->rspq, false, adap->port[i],
-				       msi_idx, &q->fl, uldrx_handler);
+				       msi_idx, q->fl.size ? &q->fl : NULL,
+				       uldrx_handler);
 		if (err)
 			goto freeout;
 		memset(&q->stats, 0, sizeof(q->stats));
 		s->rdma_rxq[i] = q->rspq.abs_id;
 	}
 
+	for_each_rdmaciq(s, i) {
+		struct sge_ofld_rxq *q = &s->rdmaciq[i];
+
+		if (msi_idx > 0)
+			msi_idx++;
+		err = t4_sge_alloc_rxq(adap, &q->rspq, false, adap->port[i],
+				       msi_idx, q->fl.size ? &q->fl : NULL,
+				       uldrx_handler);
+		if (err)
+			goto freeout;
+		memset(&q->stats, 0, sizeof(q->stats));
+		s->rdma_ciq[i] = q->rspq.abs_id;
+	}
+
 	for_each_port(adap, i) {
 		/*
 		 * Note that ->rdmarxq[i].rspq.cntxt_id below is 0 if we don't
@@ -3789,7 +3824,9 @@ static void uld_attach(struct adapter *adap, unsigned int uld)
 	lli.mtus = adap->params.mtus;
 	if (uld == CXGB4_ULD_RDMA) {
 		lli.rxq_ids = adap->sge.rdma_rxq;
+		lli.ciq_ids = adap->sge.rdma_ciq;
 		lli.nrxq = adap->sge.rdmaqs;
+		lli.nciq = adap->sge.rdmaciqs;
 	} else if (uld == CXGB4_ULD_ISCSI) {
 		lli.rxq_ids = adap->sge.ofld_rxq;
 		lli.nrxq = adap->sge.ofldqsets;
@@ -5695,6 +5732,7 @@ static void cfg_queues(struct adapter *adap)
 {
 	struct sge *s = &adap->sge;
 	int i, q10g = 0, n10g = 0, qidx = 0;
+	int ciq_size;
 
 	for_each_port(adap, i)
 		n10g += is_x_10g_port(&adap2pinfo(adap, i)->link_cfg);
@@ -5733,6 +5771,7 @@ static void cfg_queues(struct adapter *adap)
 			s->ofldqsets = adap->params.nports;
 		/* For RDMA one Rx queue per channel suffices */
 		s->rdmaqs = adap->params.nports;
+		s->rdmaciqs = adap->params.nports;
 	}
 
 	for (i = 0; i < ARRAY_SIZE(s->ethrxq); i++) {
@@ -5767,6 +5806,19 @@ static void cfg_queues(struct adapter *adap)
 		r->fl.size = 72;
 	}
 
+	ciq_size = 64 + adap->vres.cq.size + adap->tids.nftids;
+	if (ciq_size > SGE_MAX_IQ_SIZE) {
+		CH_WARN(adap, "CIQ size too small for available IQs\n");
+		ciq_size = SGE_MAX_IQ_SIZE;
+	}
+
+	for (i = 0; i < ARRAY_SIZE(s->rdmaciq); i++) {
+		struct sge_ofld_rxq *r = &s->rdmaciq[i];
+
+		init_rspq(&r->rspq, 0, 0, ciq_size, 64);
+		r->rspq.uld = CXGB4_ULD_RDMA;
+	}
+
 	init_rspq(&s->fw_evtq, 6, 0, 512, 64);
 	init_rspq(&s->intrq, 6, 0, 2 * MAX_INGQ, 64);
 }
@@ -5815,9 +5867,9 @@ static int enable_msix(struct adapter *adap)
 
 	want = s->max_ethqsets + EXTRA_VECS;
 	if (is_offload(adap)) {
-		want += s->rdmaqs + s->ofldqsets;
+		want += s->rdmaqs + s->rdmaciqs + s->ofldqsets;
 		/* need nchan for each possible ULD */
-		ofld_need = 2 * nchan;
+		ofld_need = 3 * nchan;
 	}
 	need = adap->params.nports + EXTRA_VECS + ofld_need;
 
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_uld.h b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_uld.h
index e274a047528..87af314518a 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_uld.h
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_uld.h
@@ -232,8 +232,10 @@ struct cxgb4_lld_info {
 	const struct cxgb4_virt_res *vr;     /* assorted HW resources */
 	const unsigned short *mtus;          /* MTU table */
 	const unsigned short *rxq_ids;       /* the ULD's Rx queue ids */
+	const unsigned short *ciq_ids;       /* the ULD's concentrator IQ ids */
 	unsigned short nrxq;                 /* # of Rx queues */
 	unsigned short ntxq;                 /* # of Tx queues */
+	unsigned short nciq;		     /* # of concentrator IQ */
 	unsigned char nchan:4;               /* # of channels */
 	unsigned char nports:4;              /* # of ports */
 	unsigned char wr_cred;               /* WR 16-byte credits */
diff --git a/drivers/net/ethernet/chelsio/cxgb4/sge.c b/drivers/net/ethernet/chelsio/cxgb4/sge.c
index cced1a3d518..bd82939ae8f 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/sge.c
+++ b/drivers/net/ethernet/chelsio/cxgb4/sge.c
@@ -2515,6 +2515,10 @@ void t4_free_sge_resources(struct adapter *adap)
 		if (oq->rspq.desc)
 			free_rspq_fl(adap, &oq->rspq, &oq->fl);
 	}
+	for (i = 0, oq = adap->sge.rdmaciq; i < adap->sge.rdmaciqs; i++, oq++) {
+		if (oq->rspq.desc)
+			free_rspq_fl(adap, &oq->rspq, &oq->fl);
+	}
 
 	/* clean up offload Tx queues */
 	for (i = 0; i < ARRAY_SIZE(adap->sge.ofldtxq); i++) {
diff --git a/drivers/net/ethernet/chelsio/cxgb4/t4_hw.h b/drivers/net/ethernet/chelsio/cxgb4/t4_hw.h
index 1d1623be9f1..71b799b5b0f 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/t4_hw.h
+++ b/drivers/net/ethernet/chelsio/cxgb4/t4_hw.h
@@ -68,6 +68,7 @@ enum {
 	SGE_MAX_WR_LEN = 512,     /* max WR size in bytes */
 	SGE_NTIMERS = 6,          /* # of interrupt holdoff timer values */
 	SGE_NCOUNTERS = 4,        /* # of interrupt packet counter values */
+	SGE_MAX_IQ_SIZE = 65520,
 
 	SGE_TIMER_RSTRT_CNTR = 6, /* restart RX packet threshold counter */
 	SGE_TIMER_UPD_CIDX = 7,   /* update cidx only */