bnx2x: improve memory handling, low memory recovery flows

Signed-off-by: Dmitry Kravkov <dmitry@broadcom.com>
Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

1 files changed, 324 insertions, 78 deletions

diff --git a/drivers/net/bnx2x/bnx2x_cmn.c b/drivers/net/bnx2x/bnx2x_cmn.c
index 8853ae2a042..218a7ad7cdd 100644
--- a/drivers/net/bnx2x/bnx2x_cmn.c
+++ b/drivers/net/bnx2x/bnx2x_cmn.c
@@ -27,6 +27,49 @@
 
 static int bnx2x_setup_irqs(struct bnx2x *bp);
 
+/**
+ * bnx2x_bz_fp - zero content of the fastpath structure.
+ *
+ * @bp:		driver handle
+ * @index:	fastpath index to be zeroed
+ *
+ * Makes sure the contents of the bp->fp[index].napi is kept
+ * intact.
+ */
+static inline void bnx2x_bz_fp(struct bnx2x *bp, int index)
+{
+	struct bnx2x_fastpath *fp = &bp->fp[index];
+	struct napi_struct orig_napi = fp->napi;
+	/* bzero bnx2x_fastpath contents */
+	memset(fp, 0, sizeof(*fp));
+
+	/* Restore the NAPI object as it has been already initialized */
+	fp->napi = orig_napi;
+}
+
+/**
+ * bnx2x_move_fp - move content of the fastpath structure.
+ *
+ * @bp:		driver handle
+ * @from:	source FP index
+ * @to:		destination FP index
+ *
+ * Makes sure the contents of the bp->fp[to].napi is kept
+ * intact.
+ */
+static inline void bnx2x_move_fp(struct bnx2x *bp, int from, int to)
+{
+	struct bnx2x_fastpath *from_fp = &bp->fp[from];
+	struct bnx2x_fastpath *to_fp = &bp->fp[to];
+	struct napi_struct orig_napi = to_fp->napi;
+	/* Move bnx2x_fastpath contents */
+	memcpy(to_fp, from_fp, sizeof(*to_fp));
+	to_fp->index = to;
+
+	/* Restore the NAPI object as it has been already initialized */
+	to_fp->napi = orig_napi;
+}
+
 /* free skb in the packet ring at pos idx
  * return idx of last bd freed
  */
@@ -881,55 +924,6 @@ void __bnx2x_link_report(struct bnx2x *bp)
 	}
 }
 
-/* Returns the number of actually allocated BDs */
-static inline int bnx2x_alloc_rx_bds(struct bnx2x_fastpath *fp,
-				      int rx_ring_size)
-{
-	struct bnx2x *bp = fp->bp;
-	u16 ring_prod, cqe_ring_prod;
-	int i;
-
-	fp->rx_comp_cons = 0;
-	cqe_ring_prod = ring_prod = 0;
-	for (i = 0; i < rx_ring_size; i++) {
-		if (bnx2x_alloc_rx_skb(bp, fp, ring_prod) < 0) {
-			BNX2X_ERR("was only able to allocate "
-				  "%d rx skbs on queue[%d]\n", i, fp->index);
-			fp->eth_q_stats.rx_skb_alloc_failed++;
-			break;
-		}
-		ring_prod = NEXT_RX_IDX(ring_prod);
-		cqe_ring_prod = NEXT_RCQ_IDX(cqe_ring_prod);
-		WARN_ON(ring_prod <= i);
-	}
-
-	fp->rx_bd_prod = ring_prod;
-	/* Limit the CQE producer by the CQE ring size */
-	fp->rx_comp_prod = min_t(u16, NUM_RCQ_RINGS*RCQ_DESC_CNT,
-			       cqe_ring_prod);
-	fp->rx_pkt = fp->rx_calls = 0;
-
-	return i;
-}
-
-static inline void bnx2x_alloc_rx_bd_ring(struct bnx2x_fastpath *fp)
-{
-	struct bnx2x *bp = fp->bp;
-	int rx_ring_size = bp->rx_ring_size ? bp->rx_ring_size :
-					      MAX_RX_AVAIL/bp->num_queues;
-
-	rx_ring_size = max_t(int, MIN_RX_AVAIL, rx_ring_size);
-
-	bnx2x_alloc_rx_bds(fp, rx_ring_size);
-
-	/* Warning!
-	 * this will generate an interrupt (to the TSTORM)
-	 * must only be done after chip is initialized
-	 */
-	bnx2x_update_rx_prod(bp, fp, fp->rx_bd_prod, fp->rx_comp_prod,
-			     fp->rx_sge_prod);
-}
-
 void bnx2x_init_rx_rings(struct bnx2x *bp)
 {
 	int func = BP_FUNC(bp);
@@ -938,6 +932,7 @@ void bnx2x_init_rx_rings(struct bnx2x *bp)
 	u16 ring_prod;
 	int i, j;
 
+	/* Allocate TPA resources */
 	for_each_rx_queue(bp, j) {
 		struct bnx2x_fastpath *fp = &bp->fp[j];
 
@@ -945,6 +940,7 @@ void bnx2x_init_rx_rings(struct bnx2x *bp)
 		   "mtu %d  rx_buf_size %d\n", bp->dev->mtu, fp->rx_buf_size);
 
 		if (!fp->disable_tpa) {
+			/* Fill the per-aggregation pool */
 			for (i = 0; i < max_agg_queues; i++) {
 				fp->tpa_pool[i].skb =
 				   netdev_alloc_skb(bp->dev, fp->rx_buf_size);
@@ -999,13 +995,13 @@ void bnx2x_init_rx_rings(struct bnx2x *bp)
 
 		fp->rx_bd_cons = 0;
 
-		bnx2x_set_next_page_rx_bd(fp);
-
-		/* CQ ring */
-		bnx2x_set_next_page_rx_cq(fp);
-
-		/* Allocate BDs and initialize BD ring */
-		bnx2x_alloc_rx_bd_ring(fp);
+		/* Activate BD ring */
+		/* Warning!
+		 * this will generate an interrupt (to the TSTORM)
+		 * must only be done after chip is initialized
+		 */
+		bnx2x_update_rx_prod(bp, fp, fp->rx_bd_prod, fp->rx_comp_prod,
+				     fp->rx_sge_prod);
 
 		if (j != 0)
 			continue;
@@ -1039,27 +1035,40 @@ static void bnx2x_free_tx_skbs(struct bnx2x *bp)
 	}
 }
 
+static void bnx2x_free_rx_bds(struct bnx2x_fastpath *fp)
+{
+	struct bnx2x *bp = fp->bp;
+	int i;
+
+	/* ring wasn't allocated */
+	if (fp->rx_buf_ring == NULL)
+		return;
+
+	for (i = 0; i < NUM_RX_BD; i++) {
+		struct sw_rx_bd *rx_buf = &fp->rx_buf_ring[i];
+		struct sk_buff *skb = rx_buf->skb;
+
+		if (skb == NULL)
+			continue;
+
+		dma_unmap_single(&bp->pdev->dev,
+				 dma_unmap_addr(rx_buf, mapping),
+				 fp->rx_buf_size, DMA_FROM_DEVICE);
+
+		rx_buf->skb = NULL;
+		dev_kfree_skb(skb);
+	}
+}
+
 static void bnx2x_free_rx_skbs(struct bnx2x *bp)
 {
-	int i, j;
+	int j;
 
 	for_each_rx_queue(bp, j) {
 		struct bnx2x_fastpath *fp = &bp->fp[j];
 
-		for (i = 0; i < NUM_RX_BD; i++) {
-			struct sw_rx_bd *rx_buf = &fp->rx_buf_ring[i];
-			struct sk_buff *skb = rx_buf->skb;
+		bnx2x_free_rx_bds(fp);
 
-			if (skb == NULL)
-				continue;
-
-			dma_unmap_single(&bp->pdev->dev,
-					 dma_unmap_addr(rx_buf, mapping),
-					 fp->rx_buf_size, DMA_FROM_DEVICE);
-
-			rx_buf->skb = NULL;
-			dev_kfree_skb(skb);
-		}
 		if (!fp->disable_tpa)
 			bnx2x_free_tpa_pool(bp, fp, CHIP_IS_E1(bp) ?
 					    ETH_MAX_AGGREGATION_QUEUES_E1 :
@@ -1435,26 +1444,37 @@ int bnx2x_nic_load(struct bnx2x *bp, int load_mode)
 	/* must be called before memory allocation and HW init */
 	bnx2x_ilt_set_info(bp);
 
+	/* zero fastpath structures preserving invariants like napi which are
+	 * allocated only once
+	 */
+	for_each_queue(bp, i)
+		bnx2x_bz_fp(bp, i);
+
 	/* Set the receive queues buffer size */
 	bnx2x_set_rx_buf_size(bp);
 
+	for_each_queue(bp, i)
+		bnx2x_fp(bp, i, disable_tpa) =
+					((bp->flags & TPA_ENABLE_FLAG) == 0);
+
+#ifdef BCM_CNIC
+	/* We don't want TPA on FCoE L2 ring */
+	bnx2x_fcoe(bp, disable_tpa) = 1;
+#endif
+
 	if (bnx2x_alloc_mem(bp))
 		return -ENOMEM;
 
+	/* As long as bnx2x_alloc_mem() may possibly update
+	 * bp->num_queues, bnx2x_set_real_num_queues() should always
+	 * come after it.
+	 */
 	rc = bnx2x_set_real_num_queues(bp);
 	if (rc) {
 		BNX2X_ERR("Unable to set real_num_queues\n");
 		goto load_error0;
 	}
 
-	for_each_queue(bp, i)
-		bnx2x_fp(bp, i, disable_tpa) =
-					((bp->flags & TPA_ENABLE_FLAG) == 0);
-
-#ifdef BCM_CNIC
-	/* We don't want TPA on FCoE L2 ring */
-	bnx2x_fcoe(bp, disable_tpa) = 1;
-#endif
 	bnx2x_napi_enable(bp);
 
 	/* Send LOAD_REQUEST command to MCP
@@ -2480,6 +2500,232 @@ int bnx2x_change_mac_addr(struct net_device *dev, void *p)
 	return 0;
 }
 
+static void bnx2x_free_fp_mem_at(struct bnx2x *bp, int fp_index)
+{
+	union host_hc_status_block *sb = &bnx2x_fp(bp, fp_index, status_blk);
+	struct bnx2x_fastpath *fp = &bp->fp[fp_index];
+
+	/* Common */
+#ifdef BCM_CNIC
+	if (IS_FCOE_IDX(fp_index)) {
+		memset(sb, 0, sizeof(union host_hc_status_block));
+		fp->status_blk_mapping = 0;
+
+	} else {
+#endif
+		/* status blocks */
+		if (CHIP_IS_E2(bp))
+			BNX2X_PCI_FREE(sb->e2_sb,
+				       bnx2x_fp(bp, fp_index,
+						status_blk_mapping),
+				       sizeof(struct host_hc_status_block_e2));
+		else
+			BNX2X_PCI_FREE(sb->e1x_sb,
+				       bnx2x_fp(bp, fp_index,
+						status_blk_mapping),
+				       sizeof(struct host_hc_status_block_e1x));
+#ifdef BCM_CNIC
+	}
+#endif
+	/* Rx */
+	if (!skip_rx_queue(bp, fp_index)) {
+		bnx2x_free_rx_bds(fp);
+
+		/* fastpath rx rings: rx_buf rx_desc rx_comp */
+		BNX2X_FREE(bnx2x_fp(bp, fp_index, rx_buf_ring));
+		BNX2X_PCI_FREE(bnx2x_fp(bp, fp_index, rx_desc_ring),
+			       bnx2x_fp(bp, fp_index, rx_desc_mapping),
+			       sizeof(struct eth_rx_bd) * NUM_RX_BD);
+
+		BNX2X_PCI_FREE(bnx2x_fp(bp, fp_index, rx_comp_ring),
+			       bnx2x_fp(bp, fp_index, rx_comp_mapping),
+			       sizeof(struct eth_fast_path_rx_cqe) *
+			       NUM_RCQ_BD);
+
+		/* SGE ring */
+		BNX2X_FREE(bnx2x_fp(bp, fp_index, rx_page_ring));
+		BNX2X_PCI_FREE(bnx2x_fp(bp, fp_index, rx_sge_ring),
+			       bnx2x_fp(bp, fp_index, rx_sge_mapping),
+			       BCM_PAGE_SIZE * NUM_RX_SGE_PAGES);
+	}
+
+	/* Tx */
+	if (!skip_tx_queue(bp, fp_index)) {
+		/* fastpath tx rings: tx_buf tx_desc */
+		BNX2X_FREE(bnx2x_fp(bp, fp_index, tx_buf_ring));
+		BNX2X_PCI_FREE(bnx2x_fp(bp, fp_index, tx_desc_ring),
+			       bnx2x_fp(bp, fp_index, tx_desc_mapping),
+			       sizeof(union eth_tx_bd_types) * NUM_TX_BD);
+	}
+	/* end of fastpath */
+}
+
+void bnx2x_free_fp_mem(struct bnx2x *bp)
+{
+	int i;
+	for_each_queue(bp, i)
+		bnx2x_free_fp_mem_at(bp, i);
+}
+
+static inline void set_sb_shortcuts(struct bnx2x *bp, int index)
+{
+	union host_hc_status_block status_blk = bnx2x_fp(bp, index, status_blk);
+	if (CHIP_IS_E2(bp)) {
+		bnx2x_fp(bp, index, sb_index_values) =
+			(__le16 *)status_blk.e2_sb->sb.index_values;
+		bnx2x_fp(bp, index, sb_running_index) =
+			(__le16 *)status_blk.e2_sb->sb.running_index;
+	} else {
+		bnx2x_fp(bp, index, sb_index_values) =
+			(__le16 *)status_blk.e1x_sb->sb.index_values;
+		bnx2x_fp(bp, index, sb_running_index) =
+			(__le16 *)status_blk.e1x_sb->sb.running_index;
+	}
+}
+
+static int bnx2x_alloc_fp_mem_at(struct bnx2x *bp, int index)
+{
+	union host_hc_status_block *sb;
+	struct bnx2x_fastpath *fp = &bp->fp[index];
+	int ring_size = 0;
+
+	/* if rx_ring_size specified - use it */
+	int rx_ring_size = bp->rx_ring_size ? bp->rx_ring_size :
+			   MAX_RX_AVAIL/bp->num_queues;
+
+	/* allocate at least number of buffers required by FW */
+	rx_ring_size = max_t(int, fp->disable_tpa ? MIN_RX_SIZE_NONTPA :
+						    MIN_RX_SIZE_TPA,
+				  rx_ring_size);
+
+	bnx2x_fp(bp, index, bp) = bp;
+	bnx2x_fp(bp, index, index) = index;
+
+	/* Common */
+	sb = &bnx2x_fp(bp, index, status_blk);
+#ifdef BCM_CNIC
+	if (!IS_FCOE_IDX(index)) {
+#endif
+		/* status blocks */
+		if (CHIP_IS_E2(bp))
+			BNX2X_PCI_ALLOC(sb->e2_sb,
+				&bnx2x_fp(bp, index, status_blk_mapping),
+				sizeof(struct host_hc_status_block_e2));
+		else
+			BNX2X_PCI_ALLOC(sb->e1x_sb,
+				&bnx2x_fp(bp, index, status_blk_mapping),
+			    sizeof(struct host_hc_status_block_e1x));
+#ifdef BCM_CNIC
+	}
+#endif
+	set_sb_shortcuts(bp, index);
+
+	/* Tx */
+	if (!skip_tx_queue(bp, index)) {
+		/* fastpath tx rings: tx_buf tx_desc */
+		BNX2X_ALLOC(bnx2x_fp(bp, index, tx_buf_ring),
+				sizeof(struct sw_tx_bd) * NUM_TX_BD);
+		BNX2X_PCI_ALLOC(bnx2x_fp(bp, index, tx_desc_ring),
+				&bnx2x_fp(bp, index, tx_desc_mapping),
+				sizeof(union eth_tx_bd_types) * NUM_TX_BD);
+	}
+
+	/* Rx */
+	if (!skip_rx_queue(bp, index)) {
+		/* fastpath rx rings: rx_buf rx_desc rx_comp */
+		BNX2X_ALLOC(bnx2x_fp(bp, index, rx_buf_ring),
+				sizeof(struct sw_rx_bd) * NUM_RX_BD);
+		BNX2X_PCI_ALLOC(bnx2x_fp(bp, index, rx_desc_ring),
+				&bnx2x_fp(bp, index, rx_desc_mapping),
+				sizeof(struct eth_rx_bd) * NUM_RX_BD);
+
+		BNX2X_PCI_ALLOC(bnx2x_fp(bp, index, rx_comp_ring),
+				&bnx2x_fp(bp, index, rx_comp_mapping),
+				sizeof(struct eth_fast_path_rx_cqe) *
+				NUM_RCQ_BD);
+
+		/* SGE ring */
+		BNX2X_ALLOC(bnx2x_fp(bp, index, rx_page_ring),
+				sizeof(struct sw_rx_page) * NUM_RX_SGE);
+		BNX2X_PCI_ALLOC(bnx2x_fp(bp, index, rx_sge_ring),
+				&bnx2x_fp(bp, index, rx_sge_mapping),
+				BCM_PAGE_SIZE * NUM_RX_SGE_PAGES);
+		/* RX BD ring */
+		bnx2x_set_next_page_rx_bd(fp);
+
+		/* CQ ring */
+		bnx2x_set_next_page_rx_cq(fp);
+
+		/* BDs */
+		ring_size = bnx2x_alloc_rx_bds(fp, rx_ring_size);
+		if (ring_size < rx_ring_size)
+			goto alloc_mem_err;
+	}
+
+	return 0;
+
+/* handles low memory cases */
+alloc_mem_err:
+	BNX2X_ERR("Unable to allocate full memory for queue %d (size %d)\n",
+						index, ring_size);
+	/* FW will drop all packets if queue is not big enough,
+	 * In these cases we disable the queue
+	 * Min size diferent for TPA and non-TPA queues
+	 */
+	if (ring_size < (fp->disable_tpa ?
+				MIN_RX_SIZE_TPA : MIN_RX_SIZE_NONTPA)) {
+			/* release memory allocated for this queue */
+			bnx2x_free_fp_mem_at(bp, index);
+			return -ENOMEM;
+	}
+	return 0;
+}
+
+int bnx2x_alloc_fp_mem(struct bnx2x *bp)
+{
+	int i;
+
+	/**
+	 * 1. Allocate FP for leading - fatal if error
+	 * 2. {CNIC} Allocate FCoE FP - fatal if error
+	 * 3. Allocate RSS - fix number of queues if error
+	 */
+
+	/* leading */
+	if (bnx2x_alloc_fp_mem_at(bp, 0))
+		return -ENOMEM;
+#ifdef BCM_CNIC
+	/* FCoE */
+	if (bnx2x_alloc_fp_mem_at(bp, FCOE_IDX))
+		return -ENOMEM;
+#endif
+	/* RSS */
+	for_each_nondefault_eth_queue(bp, i)
+		if (bnx2x_alloc_fp_mem_at(bp, i))
+			break;
+
+	/* handle memory failures */
+	if (i != BNX2X_NUM_ETH_QUEUES(bp)) {
+		int delta = BNX2X_NUM_ETH_QUEUES(bp) - i;
+
+		WARN_ON(delta < 0);
+#ifdef BCM_CNIC
+		/**
+		 * move non eth FPs next to last eth FP
+		 * must be done in that order
+		 * FCOE_IDX < FWD_IDX < OOO_IDX
+		 */
+
+		/* move FCoE fp */
+		bnx2x_move_fp(bp, FCOE_IDX, FCOE_IDX - delta);
+#endif
+		bp->num_queues -= delta;
+		BNX2X_ERR("Adjusted num of queues from %d to %d\n",
+			  bp->num_queues + delta, bp->num_queues);
+	}
+
+	return 0;
+}
 
 static int bnx2x_setup_irqs(struct bnx2x *bp)
 {