Merge branch 'cxgb4-tc-flower'

Rahul Lakkireddy says:

====================
cxgb4: add support to offload tc flower

This series of patches add support to offload tc flower onto Chelsio
NICs.

Patch 1 adds basic skeleton to prepare for offloading tc flower flows.

Patch 2 adds support to add/remove flows for offload.  Flows can have
accompanying masks.  Following match and action are currently supported
for offload:
Match:  ether-protocol, IPv4/IPv6 addresses, L4 ports (TCP/UDP)
Action: drop, redirect to another port on the device.

Patch 3 adds support to offload tc-flower flows having
vlan actions: pop, push, and modify.

Patch 4 adds support to fetch stats for the offloaded tc flower flows
from hardware.

Support for offloading more match and action types are to be followed
in subsequent series.

v2:
- Setting ftid to -1 not required after bitmap_find_free_region
  in cxgb4_get_free_ftid.
- Direct return can be used as jumping to error path is not needed
  if flower entry allocation failed in cxgb4_tc_flower_replace.
  Same applies if flower entry not found in cxgb4_tc_flower_destroy.
- Also, removed an extra return from cxgb4_tc_flower_destroy.
- Avoid wrapping line for netdev_err message. Also, use
  consistent error message string.
====================

Signed-off-by: David S. Miller <[email protected]>

drivers/net/ethernet/chelsio/cxgb4/Makefile

@@ -4,7 +4,9 @@
 
 obj-$(CONFIG_CHELSIO_T4) += cxgb4.o
 
-cxgb4-objs := cxgb4_main.o l2t.o t4_hw.o sge.o clip_tbl.o cxgb4_ethtool.o cxgb4_uld.o sched.o cxgb4_filter.o cxgb4_tc_u32.o cxgb4_ptp.o
+cxgb4-objs := cxgb4_main.o l2t.o t4_hw.o sge.o clip_tbl.o cxgb4_ethtool.o \
+	      cxgb4_uld.o sched.o cxgb4_filter.o cxgb4_tc_u32.o \
+	      cxgb4_ptp.o cxgb4_tc_flower.o
 cxgb4-$(CONFIG_CHELSIO_T4_DCB) +=  cxgb4_dcb.o
 cxgb4-$(CONFIG_CHELSIO_T4_FCOE) +=  cxgb4_fcoe.o
 cxgb4-$(CONFIG_DEBUG_FS) += cxgb4_debugfs.o

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

@@ -905,6 +905,10 @@ struct adapter {
 	/* TC u32 offload */
 	struct cxgb4_tc_u32_table *tc_u32;
 	struct chcr_stats_debug chcr_stats;
+
+	/* TC flower offload */
+	DECLARE_HASHTABLE(flower_anymatch_tbl, 9);
+	struct timer_list flower_stats_timer;
 };
 
 /* Support for "sched-class" command to allow a TX Scheduling Class to be

drivers/net/ethernet/chelsio/cxgb4/cxgb4_filter.c

@@ -148,6 +148,106 @@ static int get_filter_steerq(struct net_device *dev,
 	return iq;
 }
 
+static int get_filter_count(struct adapter *adapter, unsigned int fidx,
+			    u64 *pkts, u64 *bytes)
+{
+	unsigned int tcb_base, tcbaddr;
+	unsigned int word_offset;
+	struct filter_entry *f;
+	__be64 be64_byte_count;
+	int ret;
+
+	tcb_base = t4_read_reg(adapter, TP_CMM_TCB_BASE_A);
+	if ((fidx != (adapter->tids.nftids + adapter->tids.nsftids - 1)) &&
+	    fidx >= adapter->tids.nftids)
+		return -E2BIG;
+
+	f = &adapter->tids.ftid_tab[fidx];
+	if (!f->valid)
+		return -EINVAL;
+
+	tcbaddr = tcb_base + f->tid * TCB_SIZE;
+
+	spin_lock(&adapter->win0_lock);
+	if (is_t4(adapter->params.chip)) {
+		__be64 be64_count;
+
+		/* T4 doesn't maintain byte counts in hw */
+		*bytes = 0;
+
+		/* Get pkts */
+		word_offset = 4;
+		ret = t4_memory_rw(adapter, MEMWIN_NIC, MEM_EDC0,
+				   tcbaddr + (word_offset * sizeof(__be32)),
+				   sizeof(be64_count),
+				   (__be32 *)&be64_count,
+				   T4_MEMORY_READ);
+		if (ret < 0)
+			goto out;
+		*pkts = be64_to_cpu(be64_count);
+	} else {
+		__be32 be32_count;
+
+		/* Get bytes */
+		word_offset = 4;
+		ret = t4_memory_rw(adapter, MEMWIN_NIC, MEM_EDC0,
+				   tcbaddr + (word_offset * sizeof(__be32)),
+				   sizeof(be64_byte_count),
+				   &be64_byte_count,
+				   T4_MEMORY_READ);
+		if (ret < 0)
+			goto out;
+		*bytes = be64_to_cpu(be64_byte_count);
+
+		/* Get pkts */
+		word_offset = 6;
+		ret = t4_memory_rw(adapter, MEMWIN_NIC, MEM_EDC0,
+				   tcbaddr + (word_offset * sizeof(__be32)),
+				   sizeof(be32_count),
+				   &be32_count,
+				   T4_MEMORY_READ);
+		if (ret < 0)
+			goto out;
+		*pkts = (u64)be32_to_cpu(be32_count);
+	}
+
+out:
+	spin_unlock(&adapter->win0_lock);
+	return ret;
+}
+
+int cxgb4_get_filter_counters(struct net_device *dev, unsigned int fidx,
+			      u64 *hitcnt, u64 *bytecnt)
+{
+	struct adapter *adapter = netdev2adap(dev);
+
+	return get_filter_count(adapter, fidx, hitcnt, bytecnt);
+}
+
+int cxgb4_get_free_ftid(struct net_device *dev, int family)
+{
+	struct adapter *adap = netdev2adap(dev);
+	struct tid_info *t = &adap->tids;
+	int ftid;
+
+	spin_lock_bh(&t->ftid_lock);
+	if (family == PF_INET) {
+		ftid = find_first_zero_bit(t->ftid_bmap, t->nftids);
+		if (ftid >= t->nftids)
+			ftid = -1;
+	} else {
+		ftid = bitmap_find_free_region(t->ftid_bmap, t->nftids, 2);
+		if (ftid < 0)
+			goto out_unlock;
+
+		/* this is only a lookup, keep the found region unallocated */
+		bitmap_release_region(t->ftid_bmap, ftid, 2);
+	}
+out_unlock:
+	spin_unlock_bh(&t->ftid_lock);
+	return ftid;
+}
+
 static int cxgb4_set_ftid(struct tid_info *t, int fidx, int family)
 {
 	spin_lock_bh(&t->ftid_lock);

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

@@ -79,6 +79,7 @@
 #include "l2t.h"
 #include "sched.h"
 #include "cxgb4_tc_u32.h"
+#include "cxgb4_tc_flower.h"
 #include "cxgb4_ptp.h"
 
 char cxgb4_driver_name[] = KBUILD_MODNAME;
@@ -2873,6 +2874,25 @@ static int cxgb_set_tx_maxrate(struct net_device *dev, int index, u32 rate)
 	return err;
 }
 
+static int cxgb_setup_tc_flower(struct net_device *dev,
+				struct tc_cls_flower_offload *cls_flower)
+{
+	if (!is_classid_clsact_ingress(cls_flower->common.classid) ||
+	    cls_flower->common.chain_index)
+		return -EOPNOTSUPP;
+
+	switch (cls_flower->command) {
+	case TC_CLSFLOWER_REPLACE:
+		return cxgb4_tc_flower_replace(dev, cls_flower);
+	case TC_CLSFLOWER_DESTROY:
+		return cxgb4_tc_flower_destroy(dev, cls_flower);
+	case TC_CLSFLOWER_STATS:
+		return cxgb4_tc_flower_stats(dev, cls_flower);
+	default:
+		return -EOPNOTSUPP;
+	}
+}
+
 static int cxgb_setup_tc_cls_u32(struct net_device *dev,
 				 struct tc_cls_u32_offload *cls_u32)
 {
@@ -2907,6 +2927,8 @@ static int cxgb_setup_tc(struct net_device *dev, enum tc_setup_type type,
 	switch (type) {
 	case TC_SETUP_CLSU32:
 		return cxgb_setup_tc_cls_u32(dev, type_data);
+	case TC_SETUP_CLSFLOWER:
+		return cxgb_setup_tc_flower(dev, type_data);
 	default:
 		return -EOPNOTSUPP;
 	}
@@ -4615,6 +4637,7 @@ static void free_some_resources(struct adapter *adapter)
 	kvfree(adapter->l2t);
 	t4_cleanup_sched(adapter);
 	kvfree(adapter->tids.tid_tab);
+	cxgb4_cleanup_tc_flower(adapter);
 	cxgb4_cleanup_tc_u32(adapter);
 	kfree(adapter->sge.egr_map);
 	kfree(adapter->sge.ingr_map);
@@ -5083,6 +5106,8 @@ static int init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
 		if (!adapter->tc_u32)
 			dev_warn(&pdev->dev,
 				 "could not offload tc u32, continuing\n");
+
+		cxgb4_init_tc_flower(adapter);
 	}
 
 	if (is_offload(adapter)) {