From: Mitch Williams Date: Sat, 25 Oct 2014 03:24:34 +0000 (+0000) Subject: i40evf: don't use more queues than CPUs X-Git-Url: https://git.karo-electronics.de/?a=commitdiff_plain;h=cc0529271f23896208723e310fd75c2c88b3fd8c;p=linux-beck.git i40evf: don't use more queues than CPUs It's kind of silly to configure and attempt to use a bunch of queue pairs when you're running on a single (virtual) CPU. Instead of unconditionally configuring all of the queues that the PF gives us, clamp the number of queue pairs to the number of CPUs. Change-ID: I321714c9e15072ee76de8f95ab9a81f86ed347d1 Signed-off-by: Mitch Williams Signed-off-by: Patrick Lu Tested-by: Jim Young Signed-off-by: Jeff Kirsher --- diff --git a/drivers/net/ethernet/intel/i40evf/i40evf.h b/drivers/net/ethernet/intel/i40evf/i40evf.h index 30ef519d4b91..1113f8a2d3b6 100644 --- a/drivers/net/ethernet/intel/i40evf/i40evf.h +++ b/drivers/net/ethernet/intel/i40evf/i40evf.h @@ -191,6 +191,7 @@ struct i40evf_adapter { struct i40e_q_vector *q_vector[MAX_MSIX_Q_VECTORS]; struct list_head vlan_filter_list; char misc_vector_name[IFNAMSIZ + 9]; + int num_active_queues; /* TX */ struct i40e_ring *tx_rings[I40E_MAX_VSI_QP]; diff --git a/drivers/net/ethernet/intel/i40evf/i40evf_ethtool.c b/drivers/net/ethernet/intel/i40evf/i40evf_ethtool.c index efee6b290c0f..876411c39ee0 100644 --- a/drivers/net/ethernet/intel/i40evf/i40evf_ethtool.c +++ b/drivers/net/ethernet/intel/i40evf/i40evf_ethtool.c @@ -59,7 +59,7 @@ static const struct i40evf_stats i40evf_gstrings_stats[] = { #define I40EVF_GLOBAL_STATS_LEN ARRAY_SIZE(i40evf_gstrings_stats) #define I40EVF_QUEUE_STATS_LEN(_dev) \ (((struct i40evf_adapter *) \ - netdev_priv(_dev))->vsi_res->num_queue_pairs \ + netdev_priv(_dev))->num_active_queues \ * 2 * (sizeof(struct i40e_queue_stats) / sizeof(u64))) #define I40EVF_STATS_LEN(_dev) \ (I40EVF_GLOBAL_STATS_LEN + I40EVF_QUEUE_STATS_LEN(_dev)) @@ -121,11 +121,11 @@ static void i40evf_get_ethtool_stats(struct net_device *netdev, p = (char *)adapter + i40evf_gstrings_stats[i].stat_offset; data[i] = *(u64 *)p; } - for (j = 0; j < adapter->vsi_res->num_queue_pairs; j++) { + for (j = 0; j < adapter->num_active_queues; j++) { data[i++] = adapter->tx_rings[j]->stats.packets; data[i++] = adapter->tx_rings[j]->stats.bytes; } - for (j = 0; j < adapter->vsi_res->num_queue_pairs; j++) { + for (j = 0; j < adapter->num_active_queues; j++) { data[i++] = adapter->rx_rings[j]->stats.packets; data[i++] = adapter->rx_rings[j]->stats.bytes; } @@ -151,13 +151,13 @@ static void i40evf_get_strings(struct net_device *netdev, u32 sset, u8 *data) ETH_GSTRING_LEN); p += ETH_GSTRING_LEN; } - for (i = 0; i < adapter->vsi_res->num_queue_pairs; i++) { + for (i = 0; i < adapter->num_active_queues; i++) { snprintf(p, ETH_GSTRING_LEN, "tx-%u.packets", i); p += ETH_GSTRING_LEN; snprintf(p, ETH_GSTRING_LEN, "tx-%u.bytes", i); p += ETH_GSTRING_LEN; } - for (i = 0; i < adapter->vsi_res->num_queue_pairs; i++) { + for (i = 0; i < adapter->num_active_queues; i++) { snprintf(p, ETH_GSTRING_LEN, "rx-%u.packets", i); p += ETH_GSTRING_LEN; snprintf(p, ETH_GSTRING_LEN, "rx-%u.bytes", i); @@ -430,7 +430,7 @@ static int i40evf_get_rxnfc(struct net_device *netdev, switch (cmd->cmd) { case ETHTOOL_GRXRINGS: - cmd->data = adapter->vsi_res->num_queue_pairs; + cmd->data = adapter->num_active_queues; ret = 0; break; case ETHTOOL_GRXFH: @@ -598,12 +598,12 @@ static void i40evf_get_channels(struct net_device *netdev, struct i40evf_adapter *adapter = netdev_priv(netdev); /* Report maximum channels */ - ch->max_combined = adapter->vsi_res->num_queue_pairs; + ch->max_combined = adapter->num_active_queues; ch->max_other = NONQ_VECS; ch->other_count = NONQ_VECS; - ch->combined_count = adapter->vsi_res->num_queue_pairs; + ch->combined_count = adapter->num_active_queues; } /** diff --git a/drivers/net/ethernet/intel/i40evf/i40evf_main.c b/drivers/net/ethernet/intel/i40evf/i40evf_main.c index b2f01eb2f9e5..f0d07ad54198 100644 --- a/drivers/net/ethernet/intel/i40evf/i40evf_main.c +++ b/drivers/net/ethernet/intel/i40evf/i40evf_main.c @@ -397,8 +397,8 @@ static int i40evf_map_rings_to_vectors(struct i40evf_adapter *adapter) int q_vectors; int v_start = 0; int rxr_idx = 0, txr_idx = 0; - int rxr_remaining = adapter->vsi_res->num_queue_pairs; - int txr_remaining = adapter->vsi_res->num_queue_pairs; + int rxr_remaining = adapter->num_active_queues; + int txr_remaining = adapter->num_active_queues; int i, j; int rqpv, tqpv; int err = 0; @@ -584,7 +584,7 @@ static void i40evf_configure_tx(struct i40evf_adapter *adapter) { struct i40e_hw *hw = &adapter->hw; int i; - for (i = 0; i < adapter->vsi_res->num_queue_pairs; i++) + for (i = 0; i < adapter->num_active_queues; i++) adapter->tx_rings[i]->tail = hw->hw_addr + I40E_QTX_TAIL1(i); } @@ -629,7 +629,7 @@ static void i40evf_configure_rx(struct i40evf_adapter *adapter) rx_buf_len = ALIGN(max_frame, 1024); } - for (i = 0; i < adapter->vsi_res->num_queue_pairs; i++) { + for (i = 0; i < adapter->num_active_queues; i++) { adapter->rx_rings[i]->tail = hw->hw_addr + I40E_QRX_TAIL1(i); adapter->rx_rings[i]->rx_buf_len = rx_buf_len; } @@ -918,7 +918,7 @@ static void i40evf_configure(struct i40evf_adapter *adapter) i40evf_configure_rx(adapter); adapter->aq_required |= I40EVF_FLAG_AQ_CONFIGURE_QUEUES; - for (i = 0; i < adapter->vsi_res->num_queue_pairs; i++) { + for (i = 0; i < adapter->num_active_queues; i++) { struct i40e_ring *ring = adapter->rx_rings[i]; i40evf_alloc_rx_buffers(ring, ring->count); ring->next_to_use = ring->count - 1; @@ -950,7 +950,7 @@ static void i40evf_clean_all_rx_rings(struct i40evf_adapter *adapter) { int i; - for (i = 0; i < adapter->vsi_res->num_queue_pairs; i++) + for (i = 0; i < adapter->num_active_queues; i++) i40evf_clean_rx_ring(adapter->rx_rings[i]); } @@ -962,7 +962,7 @@ static void i40evf_clean_all_tx_rings(struct i40evf_adapter *adapter) { int i; - for (i = 0; i < adapter->vsi_res->num_queue_pairs; i++) + for (i = 0; i < adapter->num_active_queues; i++) i40evf_clean_tx_ring(adapter->tx_rings[i]); } @@ -1064,7 +1064,7 @@ static void i40evf_free_queues(struct i40evf_adapter *adapter) if (!adapter->vsi_res) return; - for (i = 0; i < adapter->vsi_res->num_queue_pairs; i++) { + for (i = 0; i < adapter->num_active_queues; i++) { if (adapter->tx_rings[i]) kfree_rcu(adapter->tx_rings[i], rcu); adapter->tx_rings[i] = NULL; @@ -1084,7 +1084,7 @@ static int i40evf_alloc_queues(struct i40evf_adapter *adapter) { int i; - for (i = 0; i < adapter->vsi_res->num_queue_pairs; i++) { + for (i = 0; i < adapter->num_active_queues; i++) { struct i40e_ring *tx_ring; struct i40e_ring *rx_ring; @@ -1130,7 +1130,7 @@ static int i40evf_set_interrupt_capability(struct i40evf_adapter *adapter) err = -EIO; goto out; } - pairs = adapter->vsi_res->num_queue_pairs; + pairs = adapter->num_active_queues; /* It's easy to be greedy for MSI-X vectors, but it really * doesn't do us much good if we have a lot more vectors @@ -1210,7 +1210,7 @@ static void i40evf_free_q_vectors(struct i40evf_adapter *adapter) int napi_vectors; num_q_vectors = adapter->num_msix_vectors - NONQ_VECS; - napi_vectors = adapter->vsi_res->num_queue_pairs; + napi_vectors = adapter->num_active_queues; for (q_idx = 0; q_idx < num_q_vectors; q_idx++) { struct i40e_q_vector *q_vector = adapter->q_vector[q_idx]; @@ -1265,8 +1265,8 @@ int i40evf_init_interrupt_scheme(struct i40evf_adapter *adapter) } dev_info(&adapter->pdev->dev, "Multiqueue %s: Queue pair count = %u", - (adapter->vsi_res->num_queue_pairs > 1) ? "Enabled" : - "Disabled", adapter->vsi_res->num_queue_pairs); + (adapter->num_active_queues > 1) ? "Enabled" : + "Disabled", adapter->num_active_queues); return 0; err_alloc_queues: @@ -1425,7 +1425,7 @@ static int next_queue(struct i40evf_adapter *adapter, int j) { j += 1; - return j >= adapter->vsi_res->num_queue_pairs ? 0 : j; + return j >= adapter->num_active_queues ? 0 : j; } /** @@ -1446,9 +1446,14 @@ static void i40evf_configure_rss(struct i40evf_adapter *adapter) 0xc135cafa, 0x7a6f7e2d, 0xe7102d28, 0x163cd12e, 0x4954b126 }; - /* Hash type is configured by the PF - we just supply the key */ + /* No RSS for single queue. */ + if (adapter->num_active_queues == 1) { + wr32(hw, I40E_VFQF_HENA(0), 0); + wr32(hw, I40E_VFQF_HENA(1), 0); + return; + } - /* Fill out hash function seed */ + /* Hash type is configured by the PF - we just supply the key */ for (i = 0; i <= I40E_VFQF_HKEY_MAX_INDEX; i++) wr32(hw, I40E_VFQF_HKEY(i), seed[i]); @@ -1458,7 +1463,7 @@ static void i40evf_configure_rss(struct i40evf_adapter *adapter) wr32(hw, I40E_VFQF_HENA(1), (u32)(hena >> 32)); /* Populate the LUT with max no. of queues in round robin fashion */ - j = adapter->vsi_res->num_queue_pairs; + j = adapter->num_active_queues; for (i = 0; i <= I40E_VFQF_HLUT_MAX_INDEX; i++) { j = next_queue(adapter, j); lut = j; @@ -1703,7 +1708,7 @@ static void i40evf_free_all_tx_resources(struct i40evf_adapter *adapter) { int i; - for (i = 0; i < adapter->vsi_res->num_queue_pairs; i++) + for (i = 0; i < adapter->num_active_queues; i++) if (adapter->tx_rings[i]->desc) i40evf_free_tx_resources(adapter->tx_rings[i]); @@ -1723,7 +1728,7 @@ static int i40evf_setup_all_tx_resources(struct i40evf_adapter *adapter) { int i, err = 0; - for (i = 0; i < adapter->vsi_res->num_queue_pairs; i++) { + for (i = 0; i < adapter->num_active_queues; i++) { adapter->tx_rings[i]->count = adapter->tx_desc_count; err = i40evf_setup_tx_descriptors(adapter->tx_rings[i]); if (!err) @@ -1751,7 +1756,7 @@ static int i40evf_setup_all_rx_resources(struct i40evf_adapter *adapter) { int i, err = 0; - for (i = 0; i < adapter->vsi_res->num_queue_pairs; i++) { + for (i = 0; i < adapter->num_active_queues; i++) { adapter->rx_rings[i]->count = adapter->rx_desc_count; err = i40evf_setup_rx_descriptors(adapter->rx_rings[i]); if (!err) @@ -1774,7 +1779,7 @@ static void i40evf_free_all_rx_resources(struct i40evf_adapter *adapter) { int i; - for (i = 0; i < adapter->vsi_res->num_queue_pairs; i++) + for (i = 0; i < adapter->num_active_queues; i++) if (adapter->rx_rings[i]->desc) i40evf_free_rx_resources(adapter->rx_rings[i]); } @@ -2150,6 +2155,9 @@ static void i40evf_init_task(struct work_struct *work) adapter->watchdog_timer.data = (unsigned long)adapter; mod_timer(&adapter->watchdog_timer, jiffies + 1); + adapter->num_active_queues = min_t(int, + adapter->vsi_res->num_queue_pairs, + (int)(num_online_cpus())); adapter->tx_desc_count = I40EVF_DEFAULT_TXD; adapter->rx_desc_count = I40EVF_DEFAULT_RXD; err = i40evf_init_interrupt_scheme(adapter); diff --git a/drivers/net/ethernet/intel/i40evf/i40evf_virtchnl.c b/drivers/net/ethernet/intel/i40evf/i40evf_virtchnl.c index ff8676147c14..49bfdb5421c8 100644 --- a/drivers/net/ethernet/intel/i40evf/i40evf_virtchnl.c +++ b/drivers/net/ethernet/intel/i40evf/i40evf_virtchnl.c @@ -210,7 +210,7 @@ void i40evf_configure_queues(struct i40evf_adapter *adapter) { struct i40e_virtchnl_vsi_queue_config_info *vqci; struct i40e_virtchnl_queue_pair_info *vqpi; - int pairs = adapter->vsi_res->num_queue_pairs; + int pairs = adapter->num_active_queues; int i, len; if (adapter->current_op != I40E_VIRTCHNL_OP_UNKNOWN) { @@ -276,7 +276,7 @@ void i40evf_enable_queues(struct i40evf_adapter *adapter) } adapter->current_op = I40E_VIRTCHNL_OP_ENABLE_QUEUES; vqs.vsi_id = adapter->vsi_res->vsi_id; - vqs.tx_queues = (1 << adapter->vsi_res->num_queue_pairs) - 1; + vqs.tx_queues = (1 << adapter->num_active_queues) - 1; vqs.rx_queues = vqs.tx_queues; adapter->aq_pending |= I40EVF_FLAG_AQ_ENABLE_QUEUES; adapter->aq_required &= ~I40EVF_FLAG_AQ_ENABLE_QUEUES; @@ -302,7 +302,7 @@ void i40evf_disable_queues(struct i40evf_adapter *adapter) } adapter->current_op = I40E_VIRTCHNL_OP_DISABLE_QUEUES; vqs.vsi_id = adapter->vsi_res->vsi_id; - vqs.tx_queues = (1 << adapter->vsi_res->num_queue_pairs) - 1; + vqs.tx_queues = (1 << adapter->num_active_queues) - 1; vqs.rx_queues = vqs.tx_queues; adapter->aq_pending |= I40EVF_FLAG_AQ_DISABLE_QUEUES; adapter->aq_required &= ~I40EVF_FLAG_AQ_DISABLE_QUEUES;