cxgb4: tune burst buffer size for TC-MQPRIO offload

For each traffic class, firmware handles up to 4 * MTU amount of data
per burst cycle. Under heavy load, this small buffer size is a
bottleneck when buffering large TSO packets in <= 1500 MTU case.
Increase the burst buffer size to 8 * MTU when supported.

Also, keep the driver's traffic class configuration API similar to
the firmware API counterpart.
Signed-off-by: Rahul Lakkireddy <rahul.lakkireddy@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

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

@@ -1125,19 +1125,20 @@ struct adapter {
  * programmed with various parameters.
  */
 struct ch_sched_params {
-	s8   type;                     /* packet or flow */
+	u8   type;                     /* packet or flow */
 	union {
 		struct {
-			s8   level;    /* scheduler hierarchy level */
-			s8   mode;     /* per-class or per-flow */
-			s8   rateunit; /* bit or packet rate */
-			s8   ratemode; /* %port relative or kbps absolute */
-			s8   channel;  /* scheduler channel [0..N] */
-			s8   class;    /* scheduler class [0..N] */
-			s32  minrate;  /* minimum rate */
-			s32  maxrate;  /* maximum rate */
-			s16  weight;   /* percent weight */
-			s16  pktsize;  /* average packet size */
+			u8   level;    /* scheduler hierarchy level */
+			u8   mode;     /* per-class or per-flow */
+			u8   rateunit; /* bit or packet rate */
+			u8   ratemode; /* %port relative or kbps absolute */
+			u8   channel;  /* scheduler channel [0..N] */
+			u8   class;    /* scheduler class [0..N] */
+			u32  minrate;  /* minimum rate */
+			u32  maxrate;  /* maximum rate */
+			u16  weight;   /* percent weight */
+			u16  pktsize;  /* average packet size */
+			u16  burstsize;  /* burst buffer size */
 		} params;
 	} u;
 };
@@ -1952,9 +1953,10 @@ int t4_sge_ctxt_rd(struct adapter *adap, unsigned int mbox, unsigned int cid,
 		   enum ctxt_type ctype, u32 *data);
 int t4_sge_ctxt_rd_bd(struct adapter *adap, unsigned int cid,
 		      enum ctxt_type ctype, u32 *data);
-int t4_sched_params(struct adapter *adapter, int type, int level, int mode,
-		    int rateunit, int ratemode, int channel, int class,
-		    int minrate, int maxrate, int weight, int pktsize);
+int t4_sched_params(struct adapter *adapter, u8 type, u8 level, u8 mode,
+		    u8 rateunit, u8 ratemode, u8 channel, u8 class,
+		    u32 minrate, u32 maxrate, u16 weight, u16 pktsize,
+		    u16 burstsize);
 void t4_sge_decode_idma_state(struct adapter *adapter, int state);
 void t4_idma_monitor_init(struct adapter *adapter,
 			  struct sge_idma_monitor_state *idma);

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

@@ -3021,7 +3021,7 @@ static int cxgb4_mgmt_set_vf_rate(struct net_device *dev, int vf,
 			      SCHED_CLASS_RATEUNIT_BITS,
 			      SCHED_CLASS_RATEMODE_ABS,
 			      pi->tx_chan, class_id, 0,
-			      max_tx_rate * 1000, 0, pktsize);
+			      max_tx_rate * 1000, 0, pktsize, 0);
 	if (ret) {
 		dev_err(adap->pdev_dev, "Err %d for Traffic Class config\n",
 			ret);

drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_mqprio.c

@@ -342,6 +342,13 @@ static int cxgb4_mqprio_alloc_tc(struct net_device *dev,
 		p.u.params.minrate = div_u64(mqprio->min_rate[i] * 8, 1000);
 		p.u.params.maxrate = div_u64(mqprio->max_rate[i] * 8, 1000);
 
+		/* Request larger burst buffer for smaller MTU, so
+		 * that hardware can work on more data per burst
+		 * cycle.
+		 */
+		if (dev->mtu <= ETH_DATA_LEN)
+			p.u.params.burstsize = 8 * dev->mtu;
+
 		e = cxgb4_sched_class_alloc(dev, &p);
 		if (!e) {
 			ret = -ENOMEM;

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

@@ -57,7 +57,8 @@ static int t4_sched_class_fw_cmd(struct port_info *pi,
 				      p->u.params.ratemode,
 				      p->u.params.channel, e->idx,
 				      p->u.params.minrate, p->u.params.maxrate,
-				      p->u.params.weight, p->u.params.pktsize);
+				      p->u.params.weight, p->u.params.pktsize,
+				      p->u.params.burstsize);
 		break;
 	default:
 		err = -ENOTSUPP;

drivers/net/ethernet/chelsio/cxgb4/t4_hw.c

@@ -10361,9 +10361,10 @@ int t4_sge_ctxt_rd_bd(struct adapter *adap, unsigned int cid,
 	return ret;
 }
 
-int t4_sched_params(struct adapter *adapter, int type, int level, int mode,
-		    int rateunit, int ratemode, int channel, int class,
-		    int minrate, int maxrate, int weight, int pktsize)
+int t4_sched_params(struct adapter *adapter, u8 type, u8 level, u8 mode,
+		    u8 rateunit, u8 ratemode, u8 channel, u8 class,
+		    u32 minrate, u32 maxrate, u16 weight, u16 pktsize,
+		    u16 burstsize)
 {
 	struct fw_sched_cmd cmd;
 
@@ -10385,6 +10386,7 @@ int t4_sched_params(struct adapter *adapter, int type, int level, int mode,
 	cmd.u.params.max = cpu_to_be32(maxrate);
 	cmd.u.params.weight = cpu_to_be16(weight);
 	cmd.u.params.pktsize = cpu_to_be16(pktsize);
+	cmd.u.params.burstsize = cpu_to_be16(burstsize);
 
 	return t4_wr_mbox_meat(adapter, adapter->mbox, &cmd, sizeof(cmd),
 			       NULL, 1);