RDMA/cxgb4: Fix bug for active and passive LE hash collision path

Retries active opens for INUSE errors.

Logs any active ofld_connect_wr error replies.

Sends ofld_connect_wr on same ctrlq. It needs to go  on the same control txq as
regular CPL active/passive messages.

Retries on active open replies with EADDRINUSE.

Uses active open fw wr only if active filter region is set.

Adds stat for ofld_connect_wr failures.

This patch also adds debugfs file to show endpoints.

Signed-off-by: Vipul Pandya <vipul@chelsio.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>

drivers/infiniband/hw/cxgb4/device.c

@@ -280,6 +280,10 @@ static int stats_show(struct seq_file *seq, void *v)
 		   db_state_str[dev->db_state],
 		   dev->rdev.stats.db_state_transitions);
 	seq_printf(seq, "TCAM_FULL: %10llu\n", dev->rdev.stats.tcam_full);
+	seq_printf(seq, "ACT_OFLD_CONN_FAILS: %10llu\n",
+		   dev->rdev.stats.act_ofld_conn_fails);
+	seq_printf(seq, "PAS_OFLD_CONN_FAILS: %10llu\n",
+		   dev->rdev.stats.pas_ofld_conn_fails);
 	return 0;
 }
 
@@ -310,6 +314,9 @@ static ssize_t stats_clear(struct file *file, const char __user *buf,
 	dev->rdev.stats.db_empty = 0;
 	dev->rdev.stats.db_drop = 0;
 	dev->rdev.stats.db_state_transitions = 0;
+	dev->rdev.stats.tcam_full = 0;
+	dev->rdev.stats.act_ofld_conn_fails = 0;
+	dev->rdev.stats.pas_ofld_conn_fails = 0;
 	mutex_unlock(&dev->rdev.stats.lock);
 	return count;
 }
@@ -323,6 +330,113 @@ static const struct file_operations stats_debugfs_fops = {
 	.write   = stats_clear,
 };
 
+static int dump_ep(int id, void *p, void *data)
+{
+	struct c4iw_ep *ep = p;
+	struct c4iw_debugfs_data *epd = data;
+	int space;
+	int cc;
+
+	space = epd->bufsize - epd->pos - 1;
+	if (space == 0)
+		return 1;
+
+	cc = snprintf(epd->buf + epd->pos, space,
+			"ep %p cm_id %p qp %p state %d flags 0x%lx history 0x%lx "
+			"hwtid %d atid %d %pI4:%d <-> %pI4:%d\n",
+			ep, ep->com.cm_id, ep->com.qp, (int)ep->com.state,
+			ep->com.flags, ep->com.history, ep->hwtid, ep->atid,
+			&ep->com.local_addr.sin_addr.s_addr,
+			ntohs(ep->com.local_addr.sin_port),
+			&ep->com.remote_addr.sin_addr.s_addr,
+			ntohs(ep->com.remote_addr.sin_port));
+	if (cc < space)
+		epd->pos += cc;
+	return 0;
+}
+
+static int dump_listen_ep(int id, void *p, void *data)
+{
+	struct c4iw_listen_ep *ep = p;
+	struct c4iw_debugfs_data *epd = data;
+	int space;
+	int cc;
+
+	space = epd->bufsize - epd->pos - 1;
+	if (space == 0)
+		return 1;
+
+	cc = snprintf(epd->buf + epd->pos, space,
+			"ep %p cm_id %p state %d flags 0x%lx stid %d backlog %d "
+			"%pI4:%d\n", ep, ep->com.cm_id, (int)ep->com.state,
+			ep->com.flags, ep->stid, ep->backlog,
+			&ep->com.local_addr.sin_addr.s_addr,
+			ntohs(ep->com.local_addr.sin_port));
+	if (cc < space)
+		epd->pos += cc;
+	return 0;
+}
+
+static int ep_release(struct inode *inode, struct file *file)
+{
+	struct c4iw_debugfs_data *epd = file->private_data;
+	if (!epd) {
+		pr_info("%s null qpd?\n", __func__);
+		return 0;
+	}
+	vfree(epd->buf);
+	kfree(epd);
+	return 0;
+}
+
+static int ep_open(struct inode *inode, struct file *file)
+{
+	struct c4iw_debugfs_data *epd;
+	int ret = 0;
+	int count = 1;
+
+	epd = kmalloc(sizeof(*epd), GFP_KERNEL);
+	if (!epd) {
+		ret = -ENOMEM;
+		goto out;
+	}
+	epd->devp = inode->i_private;
+	epd->pos = 0;
+
+	spin_lock_irq(&epd->devp->lock);
+	idr_for_each(&epd->devp->hwtid_idr, count_idrs, &count);
+	idr_for_each(&epd->devp->atid_idr, count_idrs, &count);
+	idr_for_each(&epd->devp->stid_idr, count_idrs, &count);
+	spin_unlock_irq(&epd->devp->lock);
+
+	epd->bufsize = count * 160;
+	epd->buf = vmalloc(epd->bufsize);
+	if (!epd->buf) {
+		ret = -ENOMEM;
+		goto err1;
+	}
+
+	spin_lock_irq(&epd->devp->lock);
+	idr_for_each(&epd->devp->hwtid_idr, dump_ep, epd);
+	idr_for_each(&epd->devp->atid_idr, dump_ep, epd);
+	idr_for_each(&epd->devp->stid_idr, dump_listen_ep, epd);
+	spin_unlock_irq(&epd->devp->lock);
+
+	file->private_data = epd;
+	goto out;
+err1:
+	kfree(epd);
+out:
+	return ret;
+}
+
+static const struct file_operations ep_debugfs_fops = {
+	.owner   = THIS_MODULE,
+	.open    = ep_open,
+	.release = ep_release,
+	.read    = debugfs_read,
+};
+
 static int setup_debugfs(struct c4iw_dev *devp)
 {
 	struct dentry *de;
@@ -345,6 +459,11 @@ static int setup_debugfs(struct c4iw_dev *devp)
 	if (de && de->d_inode)
 		de->d_inode->i_size = 4096;
 
+	de = debugfs_create_file("eps", S_IWUSR, devp->debugfs_root,
+			(void *)devp, &ep_debugfs_fops);
+	if (de && de->d_inode)
+		de->d_inode->i_size = 4096;
+
 	return 0;
 }
 
@@ -476,6 +595,9 @@ static void c4iw_dealloc(struct uld_ctx *ctx)
 	idr_destroy(&ctx->dev->cqidr);
 	idr_destroy(&ctx->dev->qpidr);
 	idr_destroy(&ctx->dev->mmidr);
+	idr_destroy(&ctx->dev->hwtid_idr);
+	idr_destroy(&ctx->dev->stid_idr);
+	idr_destroy(&ctx->dev->atid_idr);
 	iounmap(ctx->dev->rdev.oc_mw_kva);
 	ib_dealloc_device(&ctx->dev->ibdev);
 	ctx->dev = NULL;
@@ -533,6 +655,9 @@ static struct c4iw_dev *c4iw_alloc(const struct cxgb4_lld_info *infop)
 	idr_init(&devp->cqidr);
 	idr_init(&devp->qpidr);
 	idr_init(&devp->mmidr);
+	idr_init(&devp->hwtid_idr);
+	idr_init(&devp->stid_idr);
+	idr_init(&devp->atid_idr);
 	spin_lock_init(&devp->lock);
 	mutex_init(&devp->rdev.stats.lock);
 	mutex_init(&devp->db_mutex);

drivers/infiniband/hw/cxgb4/iw_cxgb4.h

@@ -131,6 +131,8 @@ struct c4iw_stats {
 	u64  db_drop;
 	u64  db_state_transitions;
 	u64  tcam_full;
+	u64  act_ofld_conn_fails;
+	u64  pas_ofld_conn_fails;
 };
 
 struct c4iw_rdev {
@@ -224,6 +226,9 @@ struct c4iw_dev {
 	struct dentry *debugfs_root;
 	enum db_state db_state;
 	int qpcnt;
+	struct idr hwtid_idr;
+	struct idr atid_idr;
+	struct idr stid_idr;
 };
 
 static inline struct c4iw_dev *to_c4iw_dev(struct ib_device *ibdev)
@@ -713,6 +718,31 @@ enum c4iw_ep_flags {
 	CLOSE_SENT		= 3,
 };
 
+enum c4iw_ep_history {
+	ACT_OPEN_REQ            = 0,
+	ACT_OFLD_CONN           = 1,
+	ACT_OPEN_RPL            = 2,
+	ACT_ESTAB               = 3,
+	PASS_ACCEPT_REQ         = 4,
+	PASS_ESTAB              = 5,
+	ABORT_UPCALL            = 6,
+	ESTAB_UPCALL            = 7,
+	CLOSE_UPCALL            = 8,
+	ULP_ACCEPT              = 9,
+	ULP_REJECT              = 10,
+	TIMEDOUT                = 11,
+	PEER_ABORT              = 12,
+	PEER_CLOSE              = 13,
+	CONNREQ_UPCALL          = 14,
+	ABORT_CONN              = 15,
+	DISCONN_UPCALL          = 16,
+	EP_DISC_CLOSE           = 17,
+	EP_DISC_ABORT           = 18,
+	CONN_RPL_UPCALL         = 19,
+	ACT_RETRY_NOMEM         = 20,
+	ACT_RETRY_INUSE         = 21
+};
+
 struct c4iw_ep_common {
 	struct iw_cm_id *cm_id;
 	struct c4iw_qp *qp;
@@ -724,6 +754,7 @@ struct c4iw_ep_common {
 	struct sockaddr_in remote_addr;
 	struct c4iw_wr_wait wr_wait;
 	unsigned long flags;
+	unsigned long history;
 };
 
 struct c4iw_listen_ep {
@@ -761,6 +792,7 @@ struct c4iw_ep {
 	u8 tos;
 	u8 retry_with_mpa_v1;
 	u8 tried_with_mpa_v1;
+	unsigned int retry_count;
 };
 
 static inline struct c4iw_ep *to_ep(struct iw_cm_id *cm_id)

drivers/net/ethernet/chelsio/cxgb4/cxgb4.h

@@ -530,6 +530,7 @@ struct adapter {
 	struct net_device *port[MAX_NPORTS];
 	u8 chan_map[NCHAN];                   /* channel -> port map */
 
+	u32 filter_mode;
 	unsigned int l2t_start;
 	unsigned int l2t_end;
 	struct l2t_data *l2t;

drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c

@@ -2670,7 +2670,8 @@ static int tid_init(struct tid_info *t)
  *	Returns <0 on error and one of the %NET_XMIT_* values on success.
  */
 int cxgb4_create_server(const struct net_device *dev, unsigned int stid,
-			__be32 sip, __be16 sport, unsigned int queue)
+			__be32 sip, __be16 sport, __be16 vlan,
+			unsigned int queue)
 {
 	unsigned int chan;
 	struct sk_buff *skb;
@@ -3043,7 +3044,7 @@ static void uld_attach(struct adapter *adap, unsigned int uld)
 	lli.ucq_density = 1 << QUEUESPERPAGEPF0_GET(
 			t4_read_reg(adap, SGE_INGRESS_QUEUES_PER_PAGE_PF) >>
 			(adap->fn * 4));
-	lli.filt_mode = tp_vlan_pri_map;
+	lli.filt_mode = adap->filter_mode;
 	/* MODQ_REQ_MAP sets queues 0-3 to chan 0-3 */
 	for (i = 0; i < NCHAN; i++)
 		lli.tx_modq[i] = i;
@@ -3307,7 +3308,8 @@ static int delete_filter(struct adapter *adapter, unsigned int fidx)
 }
 
 int cxgb4_create_server_filter(const struct net_device *dev, unsigned int stid,
-		__be32 sip, __be16 sport, unsigned int queue)
+		__be32 sip, __be16 sport, __be16 vlan,
+		unsigned int queue, unsigned char port, unsigned char mask)
 {
 	int ret;
 	struct filter_entry *f;
@@ -3339,11 +3341,16 @@ int cxgb4_create_server_filter(const struct net_device *dev, unsigned int stid,
 	f->fs.val.lport = cpu_to_be16(sport);
 	f->fs.mask.lport  = ~0;
 	val = (u8 *)&sip;
-	if ((val[0] | val[1] | val[2] | val[3]) != 0)
+	if ((val[0] | val[1] | val[2] | val[3]) != 0) {
 		for (i = 0; i < 4; i++) {
 			f->fs.val.lip[i] = val[i];
 			f->fs.mask.lip[i] = ~0;
 		}
+		if (adap->filter_mode & F_PORT) {
+			f->fs.val.iport = port;
+			f->fs.mask.iport = mask;
+		}
+	}
 
 	f->fs.dirsteer = 1;
 	f->fs.iq = queue;
@@ -4450,6 +4457,10 @@ static int adap_init0(struct adapter *adap)
 	for (j = 0; j < NCHAN; j++)
 		adap->params.tp.tx_modq[j] = j;
 
+	t4_read_indirect(adap, TP_PIO_ADDR, TP_PIO_DATA,
+			 &adap->filter_mode, 1,
+			 TP_VLAN_PRI_MAP);
+
 	adap->flags |= FW_OK;
 	return 0;
 

drivers/net/ethernet/chelsio/cxgb4/cxgb4_uld.h

@@ -38,6 +38,7 @@
 #include <linux/cache.h>
 #include <linux/spinlock.h>
 #include <linux/skbuff.h>
+#include <linux/inetdevice.h>
 #include <linux/atomic.h>
 
 /* CPL message priority levels */
@@ -151,9 +152,12 @@ void cxgb4_remove_tid(struct tid_info *t, unsigned int qid, unsigned int tid);
 struct in6_addr;
 
 int cxgb4_create_server(const struct net_device *dev, unsigned int stid,
-			__be32 sip, __be16 sport, unsigned int queue);
+			__be32 sip, __be16 sport, __be16 vlan,
+			unsigned int queue);
 int cxgb4_create_server_filter(const struct net_device *dev, unsigned int stid,
-			       __be32 sip, __be16 sport, unsigned int queue);
+			       __be32 sip, __be16 sport, __be16 vlan,
+			       unsigned int queue,
+			       unsigned char port, unsigned char mask);
 int cxgb4_remove_server_filter(const struct net_device *dev, unsigned int stid,
 			       unsigned int queue, bool ipv6);
 static inline void set_wr_txq(struct sk_buff *skb, int prio, int queue)

drivers/net/ethernet/chelsio/cxgb4/t4_regs.h

@@ -1098,5 +1098,7 @@
 #define A_TP_TX_SCHED_PCMD 0x25
 
 #define S_PORT    1
+#define V_PORT(x) ((x) << S_PORT)
+#define F_PORT    V_PORT(1U)
 
 #endif /* __T4_REGS_H */