OSDN Git Service

cxgb4: add support to create hash filters
authorKumar Sanghvi <kumaras@chelsio.com>
Wed, 1 Nov 2017 03:23:01 +0000 (08:53 +0530)
committerDavid S. Miller <davem@davemloft.net>
Wed, 1 Nov 2017 13:06:03 +0000 (22:06 +0900)
Add support to create hash (exact-match) filters based on the value
of 'hash' field in ch_filter_specification.

Allocate SMT/L2T entries if DMAC-rewrite/SMAC-rewrite is requested.

Allocate CLIP entry in case of IPv6 filter.

Use cpl_act_open_req[6] to send hash filter create request to hw.
Also, the filter tuple is calculated as part of sending this request.

Hash-filter reply is processed on getting cpl_act_open_rpl.
In case of success, various bits/fields in filter-tcb are set per
filter requirement, such as enabling filter hitcnts, and/or various
header rewrite operations, such as VLAN-rewrite, NAT or
(L3/L4)-rewrite, and SMAC/DMAC-rewrite. In case of failure, clear the
filter entry and release any hw resources occupied by it.

The patch also moves the functions set_tcb_field, set_tcb_tflag and
configure_filter_smac towards beginning of file.

Signed-off-by: Kumar Sanghvi <kumaras@chelsio.com>
Signed-off-by: Rahul Lakkireddy <rahul.lakkireddy@chelsio.com>
Signed-off-by: Ganesh Goudar <ganeshgr@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
drivers/net/ethernet/chelsio/cxgb4/cxgb4.h
drivers/net/ethernet/chelsio/cxgb4/cxgb4_filter.c
drivers/net/ethernet/chelsio/cxgb4/cxgb4_filter.h
drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c
drivers/net/ethernet/chelsio/cxgb4/t4_msg.h
drivers/net/ethernet/chelsio/cxgb4/t4_tcb.h

index 5f021e6..bb7f0e4 100644 (file)
@@ -1050,6 +1050,7 @@ struct ch_filter_specification {
         * matching that doesn't exist as a (value, mask) tuple.
         */
        uint32_t type:1;        /* 0 => IPv4, 1 => IPv6 */
+       u32 hash:1;             /* 0 => wild-card, 1 => exact-match */
 
        /* Packet dispatch information.  Ingress packets which match the
         * filter rules will be dropped, passed to the host or switched back
@@ -1107,7 +1108,14 @@ enum {
 };
 
 enum {
-       NAT_MODE_ALL = 7,       /* NAT on entire 4-tuple */
+       NAT_MODE_NONE = 0,      /* No NAT performed */
+       NAT_MODE_DIP,           /* NAT on Dst IP */
+       NAT_MODE_DIP_DP,        /* NAT on Dst IP, Dst Port */
+       NAT_MODE_DIP_DP_SIP,    /* NAT on Dst IP, Dst Port and Src IP */
+       NAT_MODE_DIP_DP_SP,     /* NAT on Dst IP, Dst Port and Src Port */
+       NAT_MODE_SIP_SP,        /* NAT on Src IP and Src Port */
+       NAT_MODE_DIP_SIP_SP,    /* NAT on Dst IP, Src IP and Src Port */
+       NAT_MODE_ALL            /* NAT on entire 4-tuple */
 };
 
 /* Host shadow copy of ingress filter entry.  This is in host native format
index 566bd2d..cf8ca69 100644 (file)
@@ -35,6 +35,8 @@
 #include "cxgb4.h"
 #include "t4_regs.h"
 #include "t4_tcb.h"
+#include "t4_values.h"
+#include "clip_tbl.h"
 #include "l2t.h"
 #include "smt.h"
 #include "t4fw_api.h"
@@ -50,6 +52,141 @@ static inline bool unsupported(u32 conf, u32 conf_mask, u32 val, u32 mask)
        return !(conf & conf_mask) && is_field_set(val, mask);
 }
 
+static int set_tcb_field(struct adapter *adap, struct filter_entry *f,
+                        unsigned int ftid,  u16 word, u64 mask, u64 val,
+                        int no_reply)
+{
+       struct cpl_set_tcb_field *req;
+       struct sk_buff *skb;
+
+       skb = alloc_skb(sizeof(struct cpl_set_tcb_field), GFP_ATOMIC);
+       if (!skb)
+               return -ENOMEM;
+
+       req = (struct cpl_set_tcb_field *)__skb_put(skb, sizeof(*req));
+       memset(req, 0, sizeof(*req));
+       INIT_TP_WR_CPL(req, CPL_SET_TCB_FIELD, ftid);
+       req->reply_ctrl = htons(REPLY_CHAN_V(0) |
+                               QUEUENO_V(adap->sge.fw_evtq.abs_id) |
+                               NO_REPLY_V(no_reply));
+       req->word_cookie = htons(TCB_WORD_V(word) | TCB_COOKIE_V(ftid));
+       req->mask = cpu_to_be64(mask);
+       req->val = cpu_to_be64(val);
+       set_wr_txq(skb, CPL_PRIORITY_CONTROL, f->fs.val.iport & 0x3);
+       t4_ofld_send(adap, skb);
+       return 0;
+}
+
+/* Set one of the t_flags bits in the TCB.
+ */
+static int set_tcb_tflag(struct adapter *adap, struct filter_entry *f,
+                        unsigned int ftid, unsigned int bit_pos,
+                        unsigned int val, int no_reply)
+{
+       return set_tcb_field(adap, f, ftid,  TCB_T_FLAGS_W, 1ULL << bit_pos,
+                            (unsigned long long)val << bit_pos, no_reply);
+}
+
+static int configure_filter_smac(struct adapter *adap, struct filter_entry *f)
+{
+       int err;
+
+       /* do a set-tcb for smac-sel and CWR bit.. */
+       err = set_tcb_tflag(adap, f, f->tid, TF_CCTRL_CWR_S, 1, 1);
+       if (err)
+               goto smac_err;
+
+       err = set_tcb_field(adap, f, f->tid, TCB_SMAC_SEL_W,
+                           TCB_SMAC_SEL_V(TCB_SMAC_SEL_M),
+                           TCB_SMAC_SEL_V(f->smt->idx), 1);
+       if (!err)
+               return 0;
+
+smac_err:
+       dev_err(adap->pdev_dev, "filter %u smac config failed with error %u\n",
+               f->tid, err);
+       return err;
+}
+
+static void set_nat_params(struct adapter *adap, struct filter_entry *f,
+                          unsigned int tid, bool dip, bool sip, bool dp,
+                          bool sp)
+{
+       if (dip) {
+               if (f->fs.type) {
+                       set_tcb_field(adap, f, tid, TCB_SND_UNA_RAW_W,
+                                     WORD_MASK, f->fs.nat_lip[15] |
+                                     f->fs.nat_lip[14] << 8 |
+                                     f->fs.nat_lip[13] << 16 |
+                                     f->fs.nat_lip[12] << 24, 1);
+
+                       set_tcb_field(adap, f, tid, TCB_SND_UNA_RAW_W + 1,
+                                     WORD_MASK, f->fs.nat_lip[11] |
+                                     f->fs.nat_lip[10] << 8 |
+                                     f->fs.nat_lip[9] << 16 |
+                                     f->fs.nat_lip[8] << 24, 1);
+
+                       set_tcb_field(adap, f, tid, TCB_SND_UNA_RAW_W + 2,
+                                     WORD_MASK, f->fs.nat_lip[7] |
+                                     f->fs.nat_lip[6] << 8 |
+                                     f->fs.nat_lip[5] << 16 |
+                                     f->fs.nat_lip[4] << 24, 1);
+
+                       set_tcb_field(adap, f, tid, TCB_SND_UNA_RAW_W + 3,
+                                     WORD_MASK, f->fs.nat_lip[3] |
+                                     f->fs.nat_lip[2] << 8 |
+                                     f->fs.nat_lip[1] << 16 |
+                                     f->fs.nat_lip[0] << 24, 1);
+               } else {
+                       set_tcb_field(adap, f, tid, TCB_RX_FRAG3_LEN_RAW_W,
+                                     WORD_MASK, f->fs.nat_lip[3] |
+                                     f->fs.nat_lip[2] << 8 |
+                                     f->fs.nat_lip[1] << 16 |
+                                     f->fs.nat_lip[0] << 24, 1);
+               }
+       }
+
+       if (sip) {
+               if (f->fs.type) {
+                       set_tcb_field(adap, f, tid, TCB_RX_FRAG2_PTR_RAW_W,
+                                     WORD_MASK, f->fs.nat_fip[15] |
+                                     f->fs.nat_fip[14] << 8 |
+                                     f->fs.nat_fip[13] << 16 |
+                                     f->fs.nat_fip[12] << 24, 1);
+
+                       set_tcb_field(adap, f, tid, TCB_RX_FRAG2_PTR_RAW_W + 1,
+                                     WORD_MASK, f->fs.nat_fip[11] |
+                                     f->fs.nat_fip[10] << 8 |
+                                     f->fs.nat_fip[9] << 16 |
+                                     f->fs.nat_fip[8] << 24, 1);
+
+                       set_tcb_field(adap, f, tid, TCB_RX_FRAG2_PTR_RAW_W + 2,
+                                     WORD_MASK, f->fs.nat_fip[7] |
+                                     f->fs.nat_fip[6] << 8 |
+                                     f->fs.nat_fip[5] << 16 |
+                                     f->fs.nat_fip[4] << 24, 1);
+
+                       set_tcb_field(adap, f, tid, TCB_RX_FRAG2_PTR_RAW_W + 3,
+                                     WORD_MASK, f->fs.nat_fip[3] |
+                                     f->fs.nat_fip[2] << 8 |
+                                     f->fs.nat_fip[1] << 16 |
+                                     f->fs.nat_fip[0] << 24, 1);
+
+               } else {
+                       set_tcb_field(adap, f, tid,
+                                     TCB_RX_FRAG3_START_IDX_OFFSET_RAW_W,
+                                     WORD_MASK, f->fs.nat_fip[3] |
+                                     f->fs.nat_fip[2] << 8 |
+                                     f->fs.nat_fip[1] << 16 |
+                                     f->fs.nat_fip[0] << 24, 1);
+               }
+       }
+
+       set_tcb_field(adap, f, tid, TCB_PDU_HDR_LEN_W, WORD_MASK,
+                     (dp ? f->fs.nat_lport : 0) |
+                     (sp ? f->fs.nat_fport << 16 : 0), 1);
+}
+
 /* Validate filter spec against configuration done on the card. */
 static int validate_filter(struct net_device *dev,
                           struct ch_filter_specification *fs)
@@ -484,10 +621,8 @@ int delete_filter(struct adapter *adapter, unsigned int fidx)
 void clear_filter(struct adapter *adap, struct filter_entry *f)
 {
        /* If the new or old filter have loopback rewriteing rules then we'll
-        * need to free any existing Layer Two Table (L2T) entries of the old
-        * filter rule.  The firmware will handle freeing up any Source MAC
-        * Table (SMT) entries used for rewriting Source MAC Addresses in
-        * loopback rules.
+        * need to free any existing L2T, SMT, CLIP entries of filter
+        * rule.
         */
        if (f->l2t)
                cxgb4_l2t_release(f->l2t);
@@ -495,6 +630,9 @@ void clear_filter(struct adapter *adap, struct filter_entry *f)
        if (f->smt)
                cxgb4_smt_release(f->smt);
 
+       if (f->fs.hash && f->fs.type)
+               cxgb4_clip_release(f->dev, (const u32 *)&f->fs.val.lip, 1);
+
        /* The zeroing of the filter rule below clears the filter valid,
         * pending, locked flags, l2t pointer, etc. so it's all we need for
         * this operation.
@@ -564,6 +702,269 @@ static void fill_default_mask(struct ch_filter_specification *fs)
                fs->mask.fport = ~0;
 }
 
+static u64 hash_filter_ntuple(struct ch_filter_specification *fs,
+                             struct net_device *dev)
+{
+       struct adapter *adap = netdev2adap(dev);
+       struct tp_params *tp = &adap->params.tp;
+       u64 ntuple = 0;
+
+       /* Initialize each of the fields which we care about which are present
+        * in the Compressed Filter Tuple.
+        */
+       if (tp->vlan_shift >= 0 && fs->mask.ivlan)
+               ntuple |= (FT_VLAN_VLD_F | fs->val.ivlan) << tp->vlan_shift;
+
+       if (tp->port_shift >= 0 && fs->mask.iport)
+               ntuple |= (u64)fs->val.iport << tp->port_shift;
+
+       if (tp->protocol_shift >= 0) {
+               if (!fs->val.proto)
+                       ntuple |= (u64)IPPROTO_TCP << tp->protocol_shift;
+               else
+                       ntuple |= (u64)fs->val.proto << tp->protocol_shift;
+       }
+
+       if (tp->tos_shift >= 0 && fs->mask.tos)
+               ntuple |= (u64)(fs->val.tos) << tp->tos_shift;
+
+       if (tp->vnic_shift >= 0) {
+               if ((adap->params.tp.ingress_config & VNIC_F) &&
+                   fs->mask.pfvf_vld)
+                       ntuple |= (u64)((fs->val.pfvf_vld << 16) |
+                                       (fs->val.pf << 13) |
+                                       (fs->val.vf)) << tp->vnic_shift;
+               else
+                       ntuple |= (u64)((fs->val.ovlan_vld << 16) |
+                                       (fs->val.ovlan)) << tp->vnic_shift;
+       }
+
+       if (tp->macmatch_shift >= 0 && fs->mask.macidx)
+               ntuple |= (u64)(fs->val.macidx) << tp->macmatch_shift;
+
+       if (tp->ethertype_shift >= 0 && fs->mask.ethtype)
+               ntuple |= (u64)(fs->val.ethtype) << tp->ethertype_shift;
+
+       if (tp->matchtype_shift >= 0 && fs->mask.matchtype)
+               ntuple |= (u64)(fs->val.matchtype) << tp->matchtype_shift;
+
+       if (tp->frag_shift >= 0 && fs->mask.frag)
+               ntuple |= (u64)(fs->val.frag) << tp->frag_shift;
+
+       if (tp->fcoe_shift >= 0 && fs->mask.fcoe)
+               ntuple |= (u64)(fs->val.fcoe) << tp->fcoe_shift;
+       return ntuple;
+}
+
+static void mk_act_open_req6(struct filter_entry *f, struct sk_buff *skb,
+                            unsigned int qid_filterid, struct adapter *adap)
+{
+       struct cpl_t6_act_open_req6 *t6req = NULL;
+       struct cpl_act_open_req6 *req = NULL;
+
+       t6req = (struct cpl_t6_act_open_req6 *)__skb_put(skb, sizeof(*t6req));
+       INIT_TP_WR(t6req, 0);
+       req = (struct cpl_act_open_req6 *)t6req;
+       OPCODE_TID(req) = htonl(MK_OPCODE_TID(CPL_ACT_OPEN_REQ6, qid_filterid));
+       req->local_port = cpu_to_be16(f->fs.val.lport);
+       req->peer_port = cpu_to_be16(f->fs.val.fport);
+       req->local_ip_hi = *(__be64 *)(&f->fs.val.lip);
+       req->local_ip_lo = *(((__be64 *)&f->fs.val.lip) + 1);
+       req->peer_ip_hi = *(__be64 *)(&f->fs.val.fip);
+       req->peer_ip_lo = *(((__be64 *)&f->fs.val.fip) + 1);
+       req->opt0 = cpu_to_be64(NAGLE_V(f->fs.newvlan == VLAN_REMOVE ||
+                                       f->fs.newvlan == VLAN_REWRITE) |
+                               DELACK_V(f->fs.hitcnts) |
+                               L2T_IDX_V(f->l2t ? f->l2t->idx : 0) |
+                               SMAC_SEL_V((cxgb4_port_viid(f->dev) &
+                                           0x7F) << 1) |
+                               TX_CHAN_V(f->fs.eport) |
+                               NO_CONG_V(f->fs.rpttid) |
+                               ULP_MODE_V(f->fs.nat_mode ?
+                                          ULP_MODE_TCPDDP : ULP_MODE_NONE) |
+                               TCAM_BYPASS_F | NON_OFFLOAD_F);
+       t6req->params = cpu_to_be64(FILTER_TUPLE_V(hash_filter_ntuple(&f->fs,
+                                                                     f->dev)));
+       t6req->opt2 = htonl(RSS_QUEUE_VALID_F |
+                           RSS_QUEUE_V(f->fs.iq) |
+                           TX_QUEUE_V(f->fs.nat_mode) |
+                           T5_OPT_2_VALID_F |
+                           RX_CHANNEL_F |
+                           CONG_CNTRL_V((f->fs.action == FILTER_DROP) |
+                                        (f->fs.dirsteer << 1)) |
+                           PACE_V((f->fs.maskhash) |
+                                  ((f->fs.dirsteerhash) << 1)) |
+                           CCTRL_ECN_V(f->fs.action == FILTER_SWITCH));
+}
+
+static void mk_act_open_req(struct filter_entry *f, struct sk_buff *skb,
+                           unsigned int qid_filterid, struct adapter *adap)
+{
+       struct cpl_t6_act_open_req *t6req = NULL;
+       struct cpl_act_open_req *req = NULL;
+
+       t6req = (struct cpl_t6_act_open_req *)__skb_put(skb, sizeof(*t6req));
+       INIT_TP_WR(t6req, 0);
+       req = (struct cpl_act_open_req *)t6req;
+       OPCODE_TID(req) = htonl(MK_OPCODE_TID(CPL_ACT_OPEN_REQ, qid_filterid));
+       req->local_port = cpu_to_be16(f->fs.val.lport);
+       req->peer_port = cpu_to_be16(f->fs.val.fport);
+       req->local_ip = f->fs.val.lip[0] | f->fs.val.lip[1] << 8 |
+               f->fs.val.lip[2] << 16 | f->fs.val.lip[3] << 24;
+       req->peer_ip = f->fs.val.fip[0] | f->fs.val.fip[1] << 8 |
+               f->fs.val.fip[2] << 16 | f->fs.val.fip[3] << 24;
+       req->opt0 = cpu_to_be64(NAGLE_V(f->fs.newvlan == VLAN_REMOVE ||
+                                       f->fs.newvlan == VLAN_REWRITE) |
+                               DELACK_V(f->fs.hitcnts) |
+                               L2T_IDX_V(f->l2t ? f->l2t->idx : 0) |
+                               SMAC_SEL_V((cxgb4_port_viid(f->dev) &
+                                           0x7F) << 1) |
+                               TX_CHAN_V(f->fs.eport) |
+                               NO_CONG_V(f->fs.rpttid) |
+                               ULP_MODE_V(f->fs.nat_mode ?
+                                          ULP_MODE_TCPDDP : ULP_MODE_NONE) |
+                               TCAM_BYPASS_F | NON_OFFLOAD_F);
+
+       t6req->params = cpu_to_be64(FILTER_TUPLE_V(hash_filter_ntuple(&f->fs,
+                                                                     f->dev)));
+       t6req->opt2 = htonl(RSS_QUEUE_VALID_F |
+                           RSS_QUEUE_V(f->fs.iq) |
+                           TX_QUEUE_V(f->fs.nat_mode) |
+                           T5_OPT_2_VALID_F |
+                           RX_CHANNEL_F |
+                           CONG_CNTRL_V((f->fs.action == FILTER_DROP) |
+                                        (f->fs.dirsteer << 1)) |
+                           PACE_V((f->fs.maskhash) |
+                                  ((f->fs.dirsteerhash) << 1)) |
+                           CCTRL_ECN_V(f->fs.action == FILTER_SWITCH));
+}
+
+static int cxgb4_set_hash_filter(struct net_device *dev,
+                                struct ch_filter_specification *fs,
+                                struct filter_ctx *ctx)
+{
+       struct adapter *adapter = netdev2adap(dev);
+       struct tid_info *t = &adapter->tids;
+       struct filter_entry *f;
+       struct sk_buff *skb;
+       int iq, atid, size;
+       int ret = 0;
+       u32 iconf;
+
+       fill_default_mask(fs);
+       ret = validate_filter(dev, fs);
+       if (ret)
+               return ret;
+
+       iq = get_filter_steerq(dev, fs);
+       if (iq < 0)
+               return iq;
+
+       f = kzalloc(sizeof(*f), GFP_KERNEL);
+       if (!f)
+               return -ENOMEM;
+
+       f->fs = *fs;
+       f->ctx = ctx;
+       f->dev = dev;
+       f->fs.iq = iq;
+
+       /* If the new filter requires loopback Destination MAC and/or VLAN
+        * rewriting then we need to allocate a Layer 2 Table (L2T) entry for
+        * the filter.
+        */
+       if (f->fs.newdmac || f->fs.newvlan) {
+               /* allocate L2T entry for new filter */
+               f->l2t = t4_l2t_alloc_switching(adapter, f->fs.vlan,
+                                               f->fs.eport, f->fs.dmac);
+               if (!f->l2t) {
+                       ret = -ENOMEM;
+                       goto out_err;
+               }
+       }
+
+       /* If the new filter requires loopback Source MAC rewriting then
+        * we need to allocate a SMT entry for the filter.
+        */
+       if (f->fs.newsmac) {
+               f->smt = cxgb4_smt_alloc_switching(f->dev, f->fs.smac);
+               if (!f->smt) {
+                       if (f->l2t) {
+                               cxgb4_l2t_release(f->l2t);
+                               f->l2t = NULL;
+                       }
+                       ret = -ENOMEM;
+                       goto free_l2t;
+               }
+       }
+
+       atid = cxgb4_alloc_atid(t, f);
+       if (atid < 0)
+               goto free_smt;
+
+       iconf = adapter->params.tp.ingress_config;
+       if (iconf & VNIC_F) {
+               f->fs.val.ovlan = (fs->val.pf << 13) | fs->val.vf;
+               f->fs.mask.ovlan = (fs->mask.pf << 13) | fs->mask.vf;
+               f->fs.val.ovlan_vld = fs->val.pfvf_vld;
+               f->fs.mask.ovlan_vld = fs->mask.pfvf_vld;
+       }
+
+       size = sizeof(struct cpl_t6_act_open_req);
+       if (f->fs.type) {
+               ret = cxgb4_clip_get(f->dev, (const u32 *)&f->fs.val.lip, 1);
+               if (ret)
+                       goto free_atid;
+
+               skb = alloc_skb(size, GFP_KERNEL);
+               if (!skb) {
+                       ret = -ENOMEM;
+                       goto free_clip;
+               }
+
+               mk_act_open_req6(f, skb,
+                                ((adapter->sge.fw_evtq.abs_id << 14) | atid),
+                                adapter);
+       } else {
+               skb = alloc_skb(size, GFP_KERNEL);
+               if (!skb) {
+                       ret = -ENOMEM;
+                       goto free_atid;
+               }
+
+               mk_act_open_req(f, skb,
+                               ((adapter->sge.fw_evtq.abs_id << 14) | atid),
+                               adapter);
+       }
+
+       f->pending = 1;
+       set_wr_txq(skb, CPL_PRIORITY_SETUP, f->fs.val.iport & 0x3);
+       t4_ofld_send(adapter, skb);
+       return 0;
+
+free_clip:
+       cxgb4_clip_release(f->dev, (const u32 *)&f->fs.val.lip, 1);
+
+free_atid:
+       cxgb4_free_atid(t, atid);
+
+free_smt:
+       if (f->smt) {
+               cxgb4_smt_release(f->smt);
+               f->smt = NULL;
+       }
+
+free_l2t:
+       if (f->l2t) {
+               cxgb4_l2t_release(f->l2t);
+               f->l2t = NULL;
+       }
+
+out_err:
+       kfree(f);
+       return ret;
+}
+
 /* Check a Chelsio Filter Request for validity, convert it into our internal
  * format and send it to the hardware.  Return 0 on success, an error number
  * otherwise.  We attach any provided filter operation context to the internal
@@ -580,6 +981,14 @@ int __cxgb4_set_filter(struct net_device *dev, int filter_id,
        u32 iconf;
        int iq, ret;
 
+       if (fs->hash) {
+               if (is_hashfilter(adapter))
+                       return cxgb4_set_hash_filter(dev, fs, ctx);
+               netdev_err(dev, "%s: Exact-match filters only supported with Hash Filter configuration\n",
+                          __func__);
+               return -EINVAL;
+       }
+
        max_fidx = adapter->tids.nftids;
        if (filter_id != (max_fidx + adapter->tids.nsftids - 1) &&
            filter_id >= max_fidx)
@@ -789,60 +1198,126 @@ out:
        return ret;
 }
 
-static int set_tcb_field(struct adapter *adap, struct filter_entry *f,
-                        unsigned int ftid,  u16 word, u64 mask, u64 val,
-                        int no_reply)
+static int configure_filter_tcb(struct adapter *adap, unsigned int tid,
+                               struct filter_entry *f)
 {
-       struct cpl_set_tcb_field *req;
-       struct sk_buff *skb;
-
-       skb = alloc_skb(sizeof(struct cpl_set_tcb_field), GFP_ATOMIC);
-       if (!skb)
-               return -ENOMEM;
-
-       req = (struct cpl_set_tcb_field *)__skb_put(skb, sizeof(*req));
-       memset(req, 0, sizeof(*req));
-       INIT_TP_WR_CPL(req, CPL_SET_TCB_FIELD, ftid);
-       req->reply_ctrl = htons(REPLY_CHAN_V(0) |
-                               QUEUENO_V(adap->sge.fw_evtq.abs_id) |
-                               NO_REPLY_V(no_reply));
-       req->word_cookie = htons(TCB_WORD_V(word) | TCB_COOKIE_V(ftid));
-       req->mask = cpu_to_be64(mask);
-       req->val = cpu_to_be64(val);
-       set_wr_txq(skb, CPL_PRIORITY_CONTROL, f->fs.val.iport & 0x3);
-       t4_ofld_send(adap, skb);
+       if (f->fs.hitcnts)
+               set_tcb_field(adap, f, tid, TCB_TIMESTAMP_W,
+                             TCB_TIMESTAMP_V(TCB_TIMESTAMP_M) |
+                             TCB_RTT_TS_RECENT_AGE_V(TCB_RTT_TS_RECENT_AGE_M),
+                             TCB_TIMESTAMP_V(0ULL) |
+                             TCB_RTT_TS_RECENT_AGE_V(0ULL),
+                             1);
+
+       if (f->fs.newdmac)
+               set_tcb_tflag(adap, f, tid, TF_CCTRL_ECE_S, 1,
+                             1);
+
+       if (f->fs.newvlan == VLAN_INSERT ||
+           f->fs.newvlan == VLAN_REWRITE)
+               set_tcb_tflag(adap, f, tid, TF_CCTRL_RFR_S, 1,
+                             1);
+       if (f->fs.newsmac)
+               configure_filter_smac(adap, f);
+
+       if (f->fs.nat_mode) {
+               switch (f->fs.nat_mode) {
+               case NAT_MODE_DIP:
+                       set_nat_params(adap, f, tid, true, false, false, false);
+                       break;
+
+               case NAT_MODE_DIP_DP:
+                       set_nat_params(adap, f, tid, true, false, true, false);
+                       break;
+
+               case NAT_MODE_DIP_DP_SIP:
+                       set_nat_params(adap, f, tid, true, true, true, false);
+                       break;
+               case NAT_MODE_DIP_DP_SP:
+                       set_nat_params(adap, f, tid, true, false, true, true);
+                       break;
+
+               case NAT_MODE_SIP_SP:
+                       set_nat_params(adap, f, tid, false, true, false, true);
+                       break;
+
+               case NAT_MODE_DIP_SIP_SP:
+                       set_nat_params(adap, f, tid, true, true, false, true);
+                       break;
+
+               case NAT_MODE_ALL:
+                       set_nat_params(adap, f, tid, true, true, true, true);
+                       break;
+
+               default:
+                       pr_err("%s: Invalid NAT mode: %d\n",
+                              __func__, f->fs.nat_mode);
+                       return -EINVAL;
+               }
+       }
        return 0;
 }
 
-/* Set one of the t_flags bits in the TCB.
- */
-static int set_tcb_tflag(struct adapter *adap, struct filter_entry *f,
-                        unsigned int ftid, unsigned int bit_pos,
-                        unsigned int val, int no_reply)
+void hash_filter_rpl(struct adapter *adap, const struct cpl_act_open_rpl *rpl)
 {
-       return set_tcb_field(adap, f, ftid,  TCB_T_FLAGS_W, 1ULL << bit_pos,
-                            (unsigned long long)val << bit_pos, no_reply);
-}
+       unsigned int ftid = TID_TID_G(AOPEN_ATID_G(ntohl(rpl->atid_status)));
+       unsigned int status  = AOPEN_STATUS_G(ntohl(rpl->atid_status));
+       struct tid_info *t = &adap->tids;
+       unsigned int tid = GET_TID(rpl);
+       struct filter_ctx *ctx = NULL;
+       struct filter_entry *f;
 
-static int configure_filter_smac(struct adapter *adap, struct filter_entry *f)
-{
-       int err;
+       dev_dbg(adap->pdev_dev, "%s: tid = %u; atid = %u; status = %u\n",
+               __func__, tid, ftid, status);
 
-       /* do a set-tcb for smac-sel and CWR bit.. */
-       err = set_tcb_tflag(adap, f, f->tid, TF_CCTRL_CWR_S, 1, 1);
-       if (err)
-               goto smac_err;
+       f = lookup_atid(t, ftid);
+       if (!f) {
+               dev_err(adap->pdev_dev, "%s:could not find filter entry",
+                       __func__);
+               return;
+       }
+       ctx = f->ctx;
+       f->ctx = NULL;
+
+       switch (status) {
+       case CPL_ERR_NONE:
+               f->tid = tid;
+               f->pending = 0;
+               f->valid = 1;
+               cxgb4_insert_tid(t, f, f->tid, 0);
+               cxgb4_free_atid(t, ftid);
+               if (ctx) {
+                       ctx->tid = f->tid;
+                       ctx->result = 0;
+               }
+               if (configure_filter_tcb(adap, tid, f)) {
+                       clear_filter(adap, f);
+                       cxgb4_remove_tid(t, 0, tid, 0);
+                       kfree(f);
+                       if (ctx) {
+                               ctx->result = -EINVAL;
+                               complete(&ctx->completion);
+                       }
+                       return;
+               }
+               break;
 
-       err = set_tcb_field(adap, f, f->tid, TCB_SMAC_SEL_W,
-                           TCB_SMAC_SEL_V(TCB_SMAC_SEL_M),
-                           TCB_SMAC_SEL_V(f->smt->idx), 1);
-       if (!err)
-               return 0;
+       default:
+               dev_err(adap->pdev_dev, "%s: filter creation PROBLEM; status = %u\n",
+                       __func__, status);
 
-smac_err:
-       dev_err(adap->pdev_dev, "filter %u smac config failed with error %u\n",
-               f->tid, err);
-       return err;
+               if (ctx) {
+                       if (status == CPL_ERR_TCAM_FULL)
+                               ctx->result = -EAGAIN;
+                       else
+                               ctx->result = -EINVAL;
+               }
+               clear_filter(adap, f);
+               cxgb4_free_atid(t, ftid);
+               kfree(f);
+       }
+       if (ctx)
+               complete(&ctx->completion);
 }
 
 /* Handle a filter write/deletion reply. */
index d3c1a8f..7480d65 100644 (file)
 
 #include "t4_msg.h"
 
+#define WORD_MASK      0xffffffff
+
 void filter_rpl(struct adapter *adap, const struct cpl_set_tcb_rpl *rpl);
+void hash_filter_rpl(struct adapter *adap, const struct cpl_act_open_rpl *rpl);
 void clear_filter(struct adapter *adap, struct filter_entry *f);
 
 int set_filter_wr(struct adapter *adapter, int fidx);
index 4b07cfe..77b4bd9 100644 (file)
@@ -572,6 +572,10 @@ static int fwevtq_handler(struct sge_rspq *q, const __be64 *rsp,
                const struct cpl_set_tcb_rpl *p = (void *)rsp;
 
                filter_rpl(q->adap, p);
+       } else if (opcode == CPL_ACT_OPEN_RPL) {
+               const struct cpl_act_open_rpl *p = (void *)rsp;
+
+               hash_filter_rpl(q->adap, p);
        } else
                dev_err(q->adap->pdev_dev,
                        "unexpected CPL %#x on FW event queue\n", opcode);
index ce4838d..7e12f24 100644 (file)
@@ -286,6 +286,7 @@ struct work_request_hdr {
 
 #define RX_CHANNEL_S    26
 #define RX_CHANNEL_V(x) ((x) << RX_CHANNEL_S)
+#define RX_CHANNEL_F   RX_CHANNEL_V(1U)
 
 #define WND_SCALE_EN_S    28
 #define WND_SCALE_EN_V(x) ((x) << WND_SCALE_EN_S)
@@ -315,6 +316,10 @@ struct cpl_pass_open_req {
 #define DELACK_V(x) ((x) << DELACK_S)
 #define DELACK_F    DELACK_V(1U)
 
+#define NON_OFFLOAD_S          7
+#define NON_OFFLOAD_V(x)       ((x) << NON_OFFLOAD_S)
+#define NON_OFFLOAD_F          NON_OFFLOAD_V(1U)
+
 #define DSCP_S    22
 #define DSCP_M    0x3F
 #define DSCP_V(x) ((x) << DSCP_S)
index c1c7666..c7201eb 100644 (file)
 
 #define TCB_T_FLAGS_W          1
 
+#define TF_CCTRL_ECE_S         60
 #define TF_CCTRL_CWR_S         61
+#define TF_CCTRL_RFR_S         62
 
+#define TCB_TIMESTAMP_W                5
+#define TCB_TIMESTAMP_S                0
+#define TCB_TIMESTAMP_M                0xffffffffULL
+#define TCB_TIMESTAMP_V(x)     ((x) << TCB_TIMESTAMP_S)
+
+#define TCB_RTT_TS_RECENT_AGE_W                6
+#define TCB_RTT_TS_RECENT_AGE_S                0
+#define TCB_RTT_TS_RECENT_AGE_M                0xffffffffULL
+#define TCB_RTT_TS_RECENT_AGE_V(x)     ((x) << TCB_RTT_TS_RECENT_AGE_S)
+
+#define TCB_SND_UNA_RAW_W      10
+#define TCB_RX_FRAG2_PTR_RAW_W 27
+#define TCB_RX_FRAG3_LEN_RAW_W 29
+#define TCB_RX_FRAG3_START_IDX_OFFSET_RAW_W    30
+#define TCB_PDU_HDR_LEN_W      31
 #endif /* __T4_TCB_H */