1 files changed, 86 insertions, 43 deletions

diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c
index 12ff69b3ba91..97f90edbc068 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c
@@ -804,6 +804,26 @@ static int setup_ppod_edram(struct adapter *adap)
 	return 0;
 }
 
+static void adap_config_hpfilter(struct adapter *adapter)
+{
+	u32 param, val = 0;
+	int ret;
+
+	/* Enable HP filter region. Older fw will fail this request and
+	 * it is fine.
+	 */
+	param = FW_PARAM_DEV(HPFILTER_REGION_SUPPORT);
+	ret = t4_set_params(adapter, adapter->mbox, adapter->pf, 0,
+			    1, &param, &val);
+
+	/* An error means FW doesn't know about HP filter support,
+	 * it's not a problem, don't return an error.
+	 */
+	if (ret < 0)
+		dev_err(adapter->pdev_dev,
+			"HP filter region isn't supported by FW\n");
+}
+
 /**
  *	cxgb4_write_rss - write the RSS table for a given port
  *	@pi: the port
@@ -1427,8 +1447,8 @@ static void mk_tid_release(struct sk_buff *skb, unsigned int chan,
 static void cxgb4_queue_tid_release(struct tid_info *t, unsigned int chan,
 				    unsigned int tid)
 {
-	void **p = &t->tid_tab[tid];
 	struct adapter *adap = container_of(t, struct adapter, tids);
+	void **p = &t->tid_tab[tid - t->tid_base];
 
 	spin_lock_bh(&adap->tid_release_lock);
 	*p = adap->tid_release_head;
@@ -1480,13 +1500,13 @@ static void process_tid_release_list(struct work_struct *work)
 void cxgb4_remove_tid(struct tid_info *t, unsigned int chan, unsigned int tid,
 		      unsigned short family)
 {
-	struct sk_buff *skb;
 	struct adapter *adap = container_of(t, struct adapter, tids);
+	struct sk_buff *skb;
 
-	WARN_ON(tid >= t->ntids);
+	WARN_ON(tid_out_of_range(&adap->tids, tid));
 
-	if (t->tid_tab[tid]) {
-		t->tid_tab[tid] = NULL;
+	if (t->tid_tab[tid - adap->tids.tid_base]) {
+		t->tid_tab[tid - adap->tids.tid_base] = NULL;
 		atomic_dec(&t->conns_in_use);
 		if (t->hash_base && (tid >= t->hash_base)) {
 			if (family == AF_INET6)
@@ -1518,6 +1538,7 @@ static int tid_init(struct tid_info *t)
 	struct adapter *adap = container_of(t, struct adapter, tids);
 	unsigned int max_ftids = t->nftids + t->nsftids;
 	unsigned int natids = t->natids;
+	unsigned int hpftid_bmap_size;
 	unsigned int eotid_bmap_size;
 	unsigned int stid_bmap_size;
 	unsigned int ftid_bmap_size;
@@ -1525,12 +1546,15 @@ static int tid_init(struct tid_info *t)
 
 	stid_bmap_size = BITS_TO_LONGS(t->nstids + t->nsftids);
 	ftid_bmap_size = BITS_TO_LONGS(t->nftids);
+	hpftid_bmap_size = BITS_TO_LONGS(t->nhpftids);
 	eotid_bmap_size = BITS_TO_LONGS(t->neotids);
 	size = t->ntids * sizeof(*t->tid_tab) +
 	       natids * sizeof(*t->atid_tab) +
 	       t->nstids * sizeof(*t->stid_tab) +
 	       t->nsftids * sizeof(*t->stid_tab) +
 	       stid_bmap_size * sizeof(long) +
+	       t->nhpftids * sizeof(*t->hpftid_tab) +
+	       hpftid_bmap_size * sizeof(long) +
 	       max_ftids * sizeof(*t->ftid_tab) +
 	       ftid_bmap_size * sizeof(long) +
 	       t->neotids * sizeof(*t->eotid_tab) +
@@ -1543,7 +1567,9 @@ static int tid_init(struct tid_info *t)
 	t->atid_tab = (union aopen_entry *)&t->tid_tab[t->ntids];
 	t->stid_tab = (struct serv_entry *)&t->atid_tab[natids];
 	t->stid_bmap = (unsigned long *)&t->stid_tab[t->nstids + t->nsftids];
-	t->ftid_tab = (struct filter_entry *)&t->stid_bmap[stid_bmap_size];
+	t->hpftid_tab = (struct filter_entry *)&t->stid_bmap[stid_bmap_size];
+	t->hpftid_bmap = (unsigned long *)&t->hpftid_tab[t->nhpftids];
+	t->ftid_tab = (struct filter_entry *)&t->hpftid_bmap[hpftid_bmap_size];
 	t->ftid_bmap = (unsigned long *)&t->ftid_tab[max_ftids];
 	t->eotid_tab = (struct eotid_entry *)&t->ftid_bmap[ftid_bmap_size];
 	t->eotid_bmap = (unsigned long *)&t->eotid_tab[t->neotids];
@@ -1578,6 +1604,8 @@ static int tid_init(struct tid_info *t)
 			bitmap_zero(t->eotid_bmap, t->neotids);
 	}
 
+	if (t->nhpftids)
+		bitmap_zero(t->hpftid_bmap, t->nhpftids);
 	bitmap_zero(t->ftid_bmap, t->nftids);
 	return 0;
 }
@@ -3135,9 +3163,9 @@ static int cxgb_set_tx_maxrate(struct net_device *dev, int index, u32 rate)
 {
 	struct port_info *pi = netdev_priv(dev);
 	struct adapter *adap = pi->adapter;
+	struct ch_sched_queue qe = { 0 };
+	struct ch_sched_params p = { 0 };
 	struct sched_class *e;
-	struct ch_sched_params p;
-	struct ch_sched_queue qe;
 	u32 req_rate;
 	int err = 0;
 
@@ -3154,6 +3182,15 @@ static int cxgb_set_tx_maxrate(struct net_device *dev, int index, u32 rate)
 		return -EINVAL;
 	}
 
+	qe.queue = index;
+	e = cxgb4_sched_queue_lookup(dev, &qe);
+	if (e && e->info.u.params.level != SCHED_CLASS_LEVEL_CL_RL) {
+		dev_err(adap->pdev_dev,
+			"Queue %u already bound to class %u of type: %u\n",
+			index, e->idx, e->info.u.params.level);
+		return -EBUSY;
+	}
+
 	/* Convert from Mbps to Kbps */
 	req_rate = rate * 1000;
 
@@ -3183,7 +3220,6 @@ static int cxgb_set_tx_maxrate(struct net_device *dev, int index, u32 rate)
 		return 0;
 
 	/* Fetch any available unused or matching scheduling class */
-	memset(&p, 0, sizeof(p));
 	p.type = SCHED_CLASS_TYPE_PACKET;
 	p.u.params.level    = SCHED_CLASS_LEVEL_CL_RL;
 	p.u.params.mode     = SCHED_CLASS_MODE_CLASS;
@@ -4351,6 +4387,7 @@ static int adap_init0_config(struct adapter *adapter, int reset)
 			"HMA configuration failed with error %d\n", ret);
 
 	if (is_t6(adapter->params.chip)) {
+		adap_config_hpfilter(adapter);
 		ret = setup_ppod_edram(adapter);
 		if (!ret)
 			dev_info(adapter->pdev_dev, "Successfully enabled "
@@ -4660,16 +4697,6 @@ static int adap_init0(struct adapter *adap, int vpd_skip)
 	/*
 	 * Grab some of our basic fundamental operating parameters.
 	 */
-#define FW_PARAM_DEV(param) \
-	(FW_PARAMS_MNEM_V(FW_PARAMS_MNEM_DEV) | \
-	FW_PARAMS_PARAM_X_V(FW_PARAMS_PARAM_DEV_##param))
-
-#define FW_PARAM_PFVF(param) \
-	FW_PARAMS_MNEM_V(FW_PARAMS_MNEM_PFVF) | \
-	FW_PARAMS_PARAM_X_V(FW_PARAMS_PARAM_PFVF_##param)|  \
-	FW_PARAMS_PARAM_Y_V(0) | \
-	FW_PARAMS_PARAM_Z_V(0)
-
 	params[0] = FW_PARAM_PFVF(EQ_START);
 	params[1] = FW_PARAM_PFVF(L2T_START);
 	params[2] = FW_PARAM_PFVF(L2T_END);
@@ -4687,6 +4714,16 @@ static int adap_init0(struct adapter *adap, int vpd_skip)
 	adap->sge.ingr_start = val[5];
 
 	if (CHELSIO_CHIP_VERSION(adap->params.chip) > CHELSIO_T5) {
+		params[0] = FW_PARAM_PFVF(HPFILTER_START);
+		params[1] = FW_PARAM_PFVF(HPFILTER_END);
+		ret = t4_query_params(adap, adap->mbox, adap->pf, 0, 2,
+				      params, val);
+		if (ret < 0)
+			goto bye;
+
+		adap->tids.hpftid_base = val[0];
+		adap->tids.nhpftids = val[1] - val[0] + 1;
+
 		/* Read the raw mps entries. In T6, the last 2 tcam entries
 		 * are reserved for raw mac addresses (rawf = 2, one per port).
 		 */
@@ -4698,6 +4735,9 @@ static int adap_init0(struct adapter *adap, int vpd_skip)
 			adap->rawf_start = val[0];
 			adap->rawf_cnt = val[1] - val[0] + 1;
 		}
+
+		adap->tids.tid_base =
+			t4_read_reg(adap, LE_DB_ACTIVE_TABLE_START_INDEX_A);
 	}
 
 	/* qids (ingress/egress) returned from firmware can be anywhere
@@ -5050,8 +5090,6 @@ static int adap_init0(struct adapter *adap, int vpd_skip)
 		}
 		adap->params.crypto = ntohs(caps_cmd.cryptocaps);
 	}
-#undef FW_PARAM_PFVF
-#undef FW_PARAM_DEV
 
 	/* The MTU/MSS Table is initialized by now, so load their values.  If
 	 * we're initializing the adapter, then we'll make any modifications
@@ -5343,12 +5381,11 @@ static inline bool is_x_10g_port(const struct link_config *lc)
 static int cfg_queues(struct adapter *adap)
 {
 	u32 avail_qsets, avail_eth_qsets, avail_uld_qsets;
+	u32 i, n10g = 0, qidx = 0, n1g = 0;
+	u32 ncpus = num_online_cpus();
 	u32 niqflint, neq, num_ulds;
 	struct sge *s = &adap->sge;
-	u32 i, n10g = 0, qidx = 0;
-#ifndef CONFIG_CHELSIO_T4_DCB
-	int q10g = 0;
-#endif
+	u32 q10g = 0, q1g;
 
 	/* Reduce memory usage in kdump environment, disable all offload. */
 	if (is_kdump_kernel() || (is_uld(adap) && t4_uld_mem_alloc(adap))) {
@@ -5386,44 +5423,50 @@ static int cfg_queues(struct adapter *adap)
 		n10g += is_x_10g_port(&adap2pinfo(adap, i)->link_cfg);
 
 	avail_eth_qsets = min_t(u32, avail_qsets, MAX_ETH_QSETS);
+
+	/* We default to 1 queue per non-10G port and up to # of cores queues
+	 * per 10G port.
+	 */
+	if (n10g)
+		q10g = (avail_eth_qsets - (adap->params.nports - n10g)) / n10g;
+
+	n1g = adap->params.nports - n10g;
 #ifdef CONFIG_CHELSIO_T4_DCB
 	/* For Data Center Bridging support we need to be able to support up
 	 * to 8 Traffic Priorities; each of which will be assigned to its
 	 * own TX Queue in order to prevent Head-Of-Line Blocking.
 	 */
+	q1g = 8;
 	if (adap->params.nports * 8 > avail_eth_qsets) {
 		dev_err(adap->pdev_dev, "DCB avail_eth_qsets=%d < %d!\n",
 			avail_eth_qsets, adap->params.nports * 8);
 		return -ENOMEM;
 	}
 
-	for_each_port(adap, i) {
-		struct port_info *pi = adap2pinfo(adap, i);
+	if (adap->params.nports * ncpus < avail_eth_qsets)
+		q10g = max(8U, ncpus);
+	else
+		q10g = max(8U, q10g);
 
-		pi->first_qset = qidx;
-		pi->nqsets = is_kdump_kernel() ? 1 : 8;
-		qidx += pi->nqsets;
-	}
-#else /* !CONFIG_CHELSIO_T4_DCB */
-	/* We default to 1 queue per non-10G port and up to # of cores queues
-	 * per 10G port.
-	 */
-	if (n10g)
-		q10g = (avail_eth_qsets - (adap->params.nports - n10g)) / n10g;
-	if (q10g > netif_get_num_default_rss_queues())
-		q10g = netif_get_num_default_rss_queues();
+	while ((q10g * n10g) > (avail_eth_qsets - n1g * q1g))
+		q10g--;
 
-	if (is_kdump_kernel())
+#else /* !CONFIG_CHELSIO_T4_DCB */
+	q1g = 1;
+	q10g = min(q10g, ncpus);
+#endif /* !CONFIG_CHELSIO_T4_DCB */
+	if (is_kdump_kernel()) {
 		q10g = 1;
+		q1g = 1;
+	}
 
 	for_each_port(adap, i) {
 		struct port_info *pi = adap2pinfo(adap, i);
 
 		pi->first_qset = qidx;
-		pi->nqsets = is_x_10g_port(&pi->link_cfg) ? q10g : 1;
+		pi->nqsets = is_x_10g_port(&pi->link_cfg) ? q10g : q1g;
 		qidx += pi->nqsets;
 	}
-#endif /* !CONFIG_CHELSIO_T4_DCB */
 
 	s->ethqsets = qidx;
 	s->max_ethqsets = qidx;   /* MSI-X may lower it later */
@@ -5435,7 +5478,7 @@ static int cfg_queues(struct adapter *adap)
 		 * capped by the number of available cores.
 		 */
 		num_ulds = adap->num_uld + adap->num_ofld_uld;
-		i = min_t(u32, MAX_OFLD_QSETS, num_online_cpus());
+		i = min_t(u32, MAX_OFLD_QSETS, ncpus);
 		avail_uld_qsets = roundup(i, adap->params.nports);
 		if (avail_qsets < num_ulds * adap->params.nports) {
 			adap->params.offload = 0;