ice: Check FDIR program status for AVF
authorQi Zhang <qi.z.zhang@intel.com>
Tue, 9 Mar 2021 03:08:10 +0000 (11:08 +0800)
committerTony Nguyen <anthony.l.nguyen@intel.com>
Mon, 22 Mar 2021 18:32:12 +0000 (11:32 -0700)
Enable returning FDIR completion status by checking the
ctrl_vsi Rx queue descriptor value.

To enable returning FDIR completion status from ctrl_vsi Rx queue,
COMP_Queue and COMP_Report of FDIR filter programming descriptor
needs to be properly configured. After program request sent to ctrl_vsi
Tx queue, ctrl_vsi Rx queue interrupt will be triggered and
completion status will be returned.

Driver will first issue request in ice_vc_fdir_add_fltr(), then
pass FDIR context to the background task in interrupt service routine
ice_vc_fdir_irq_handler() and finally deal with them in
ice_flush_fdir_ctx(). ice_flush_fdir_ctx() will check the descriptor's
value, fdir context, and then send back virtual channel message to VF
by calling ice_vc_add_fdir_fltr_post(). An additional timer will be
setup in case of hardware interrupt timeout.

Signed-off-by: Yahui Cao <yahui.cao@intel.com>
Signed-off-by: Brett Creeley <brett.creeley@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
Tested-by: Chen Bo <BoX.C.Chen@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
drivers/net/ethernet/intel/ice/ice.h
drivers/net/ethernet/intel/ice/ice_hw_autogen.h
drivers/net/ethernet/intel/ice/ice_lan_tx_rx.h
drivers/net/ethernet/intel/ice/ice_main.c
drivers/net/ethernet/intel/ice/ice_txrx.c
drivers/net/ethernet/intel/ice/ice_virtchnl_fdir.c
drivers/net/ethernet/intel/ice/ice_virtchnl_fdir.h

index 7b45f2a..9bf3461 100644 (file)
@@ -231,6 +231,7 @@ enum ice_state {
        __ICE_VF_RESETS_DISABLED,       /* disable resets during ice_remove */
        __ICE_LINK_DEFAULT_OVERRIDE_PENDING,
        __ICE_PHY_INIT_COMPLETE,
+       __ICE_FD_VF_FLUSH_CTX,          /* set at FD Rx IRQ or timeout */
        __ICE_STATE_NBITS               /* must be last */
 };
 
index a778e37..67b5b9b 100644 (file)
 #define VSIQF_FD_CNT(_VSI)                     (0x00464000 + ((_VSI) * 4))
 #define VSIQF_FD_CNT_FD_GCNT_S                 0
 #define VSIQF_FD_CNT_FD_GCNT_M                 ICE_M(0x3FFF, 0)
+#define VSIQF_FD_CNT_FD_BCNT_S                 16
+#define VSIQF_FD_CNT_FD_BCNT_M                 ICE_M(0x3FFF, 16)
+#define VSIQF_FD_SIZE(_VSI)                    (0x00462000 + ((_VSI) * 4))
 #define VSIQF_HKEY_MAX_INDEX                   12
 #define VSIQF_HLUT_MAX_INDEX                   15
 #define PFPM_APM                               0x000B8080
index b30c223..21329ed 100644 (file)
@@ -140,6 +140,26 @@ struct ice_fltr_desc {
                        (0xFFFFFFFFULL << ICE_FXD_FLTR_QW1_FDID_S)
 #define ICE_FXD_FLTR_QW1_FDID_ZERO     0x0ULL
 
+/* definition for FD filter programming status descriptor WB format */
+#define ICE_FXD_FLTR_WB_QW1_DD_S       0
+#define ICE_FXD_FLTR_WB_QW1_DD_M       (0x1ULL << ICE_FXD_FLTR_WB_QW1_DD_S)
+#define ICE_FXD_FLTR_WB_QW1_DD_YES     0x1ULL
+
+#define ICE_FXD_FLTR_WB_QW1_PROG_ID_S  1
+#define ICE_FXD_FLTR_WB_QW1_PROG_ID_M  \
+                               (0x3ULL << ICE_FXD_FLTR_WB_QW1_PROG_ID_S)
+#define ICE_FXD_FLTR_WB_QW1_PROG_ADD   0x0ULL
+#define ICE_FXD_FLTR_WB_QW1_PROG_DEL   0x1ULL
+
+#define ICE_FXD_FLTR_WB_QW1_FAIL_S     4
+#define ICE_FXD_FLTR_WB_QW1_FAIL_M     (0x1ULL << ICE_FXD_FLTR_WB_QW1_FAIL_S)
+#define ICE_FXD_FLTR_WB_QW1_FAIL_YES   0x1ULL
+
+#define ICE_FXD_FLTR_WB_QW1_FAIL_PROF_S        5
+#define ICE_FXD_FLTR_WB_QW1_FAIL_PROF_M        \
+                               (0x1ULL << ICE_FXD_FLTR_WB_QW1_FAIL_PROF_S)
+#define ICE_FXD_FLTR_WB_QW1_FAIL_PROF_YES      0x1ULL
+
 struct ice_rx_ptype_decoded {
        u32 ptype:10;
        u32 known:1;
index 9cb5e65..eb895e6 100644 (file)
@@ -2071,6 +2071,7 @@ static void ice_service_task(struct work_struct *work)
        ice_process_vflr_event(pf);
        ice_clean_mailboxq_subtask(pf);
        ice_sync_arfs_fltrs(pf);
+       ice_flush_fdir_ctx(pf);
        /* Clear __ICE_SERVICE_SCHED flag to allow scheduling next event */
        ice_service_task_complete(pf);
 
@@ -2082,6 +2083,7 @@ static void ice_service_task(struct work_struct *work)
            test_bit(__ICE_MDD_EVENT_PENDING, pf->state) ||
            test_bit(__ICE_VFLR_EVENT_PENDING, pf->state) ||
            test_bit(__ICE_MAILBOXQ_EVENT_PENDING, pf->state) ||
+           test_bit(__ICE_FD_VF_FLUSH_CTX, pf->state) ||
            test_bit(__ICE_ADMINQ_EVENT_PENDING, pf->state))
                mod_timer(&pf->serv_tmr, jiffies);
 }
index b7dc25d..6d87dd9 100644 (file)
@@ -1115,6 +1115,11 @@ int ice_clean_rx_irq(struct ice_ring *rx_ring, int budget)
                dma_rmb();
 
                if (rx_desc->wb.rxdid == FDIR_DESC_RXDID || !rx_ring->netdev) {
+                       struct ice_vsi *ctrl_vsi = rx_ring->vsi;
+
+                       if (rx_desc->wb.rxdid == FDIR_DESC_RXDID &&
+                           ctrl_vsi->vf_id != ICE_INVAL_VFID)
+                               ice_vc_fdir_irq_handler(ctrl_vsi, rx_desc);
                        ice_put_rx_buf(rx_ring, NULL, 0);
                        cleaned_count++;
                        continue;
index f49947d..1f4ba38 100644 (file)
@@ -44,6 +44,7 @@ struct virtchnl_fdir_fltr_conf {
        struct ice_fdir_fltr input;
        enum ice_fdir_tunnel_type ttype;
        u64 inset_flag;
+       u32 flow_id;
 };
 
 static enum virtchnl_proto_hdr_type vc_pattern_ether[] = {
@@ -1483,7 +1484,7 @@ static int ice_vc_fdir_write_fltr(struct ice_vf *vf,
        }
 
        input->dest_vsi = vsi->idx;
-       input->comp_report = ICE_FXD_FLTR_QW0_COMP_REPORT_SW_FAIL;
+       input->comp_report = ICE_FXD_FLTR_QW0_COMP_REPORT_SW;
 
        ctrl_vsi = pf->vsi[vf->ctrl_vsi_idx];
        if (!ctrl_vsi) {
@@ -1516,6 +1517,454 @@ err_free_pkt:
 }
 
 /**
+ * ice_vf_fdir_timer - FDIR program waiting timer interrupt handler
+ * @t: pointer to timer_list
+ */
+static void ice_vf_fdir_timer(struct timer_list *t)
+{
+       struct ice_vf_fdir_ctx *ctx_irq = from_timer(ctx_irq, t, rx_tmr);
+       struct ice_vf_fdir_ctx *ctx_done;
+       struct ice_vf_fdir *fdir;
+       unsigned long flags;
+       struct ice_vf *vf;
+       struct ice_pf *pf;
+
+       fdir = container_of(ctx_irq, struct ice_vf_fdir, ctx_irq);
+       vf = container_of(fdir, struct ice_vf, fdir);
+       ctx_done = &fdir->ctx_done;
+       pf = vf->pf;
+       spin_lock_irqsave(&fdir->ctx_lock, flags);
+       if (!(ctx_irq->flags & ICE_VF_FDIR_CTX_VALID)) {
+               spin_unlock_irqrestore(&fdir->ctx_lock, flags);
+               WARN_ON_ONCE(1);
+               return;
+       }
+
+       ctx_irq->flags &= ~ICE_VF_FDIR_CTX_VALID;
+
+       ctx_done->flags |= ICE_VF_FDIR_CTX_VALID;
+       ctx_done->conf = ctx_irq->conf;
+       ctx_done->stat = ICE_FDIR_CTX_TIMEOUT;
+       ctx_done->v_opcode = ctx_irq->v_opcode;
+       spin_unlock_irqrestore(&fdir->ctx_lock, flags);
+
+       set_bit(__ICE_FD_VF_FLUSH_CTX, pf->state);
+       ice_service_task_schedule(pf);
+}
+
+/**
+ * ice_vc_fdir_irq_handler - ctrl_vsi Rx queue interrupt handler
+ * @ctrl_vsi: pointer to a VF's CTRL VSI
+ * @rx_desc: pointer to FDIR Rx queue descriptor
+ */
+void
+ice_vc_fdir_irq_handler(struct ice_vsi *ctrl_vsi,
+                       union ice_32b_rx_flex_desc *rx_desc)
+{
+       struct ice_pf *pf = ctrl_vsi->back;
+       struct ice_vf_fdir_ctx *ctx_done;
+       struct ice_vf_fdir_ctx *ctx_irq;
+       struct ice_vf_fdir *fdir;
+       unsigned long flags;
+       struct device *dev;
+       struct ice_vf *vf;
+       int ret;
+
+       vf = &pf->vf[ctrl_vsi->vf_id];
+
+       fdir = &vf->fdir;
+       ctx_done = &fdir->ctx_done;
+       ctx_irq = &fdir->ctx_irq;
+       dev = ice_pf_to_dev(pf);
+       spin_lock_irqsave(&fdir->ctx_lock, flags);
+       if (!(ctx_irq->flags & ICE_VF_FDIR_CTX_VALID)) {
+               spin_unlock_irqrestore(&fdir->ctx_lock, flags);
+               WARN_ON_ONCE(1);
+               return;
+       }
+
+       ctx_irq->flags &= ~ICE_VF_FDIR_CTX_VALID;
+
+       ctx_done->flags |= ICE_VF_FDIR_CTX_VALID;
+       ctx_done->conf = ctx_irq->conf;
+       ctx_done->stat = ICE_FDIR_CTX_IRQ;
+       ctx_done->v_opcode = ctx_irq->v_opcode;
+       memcpy(&ctx_done->rx_desc, rx_desc, sizeof(*rx_desc));
+       spin_unlock_irqrestore(&fdir->ctx_lock, flags);
+
+       ret = del_timer(&ctx_irq->rx_tmr);
+       if (!ret)
+               dev_err(dev, "VF %d: Unexpected inactive timer!\n", vf->vf_id);
+
+       set_bit(__ICE_FD_VF_FLUSH_CTX, pf->state);
+       ice_service_task_schedule(pf);
+}
+
+/**
+ * ice_vf_fdir_dump_info - dump FDIR information for diagnosis
+ * @vf: pointer to the VF info
+ */
+static void ice_vf_fdir_dump_info(struct ice_vf *vf)
+{
+       struct ice_vsi *vf_vsi;
+       u32 fd_size, fd_cnt;
+       struct device *dev;
+       struct ice_pf *pf;
+       struct ice_hw *hw;
+       u16 vsi_num;
+
+       pf = vf->pf;
+       hw = &pf->hw;
+       dev = ice_pf_to_dev(pf);
+       vf_vsi = pf->vsi[vf->lan_vsi_idx];
+       vsi_num = ice_get_hw_vsi_num(hw, vf_vsi->idx);
+
+       fd_size = rd32(hw, VSIQF_FD_SIZE(vsi_num));
+       fd_cnt = rd32(hw, VSIQF_FD_CNT(vsi_num));
+       dev_dbg(dev, "VF %d: space allocated: guar:0x%x, be:0x%x, space consumed: guar:0x%x, be:0x%x",
+               vf->vf_id,
+               (fd_size & VSIQF_FD_CNT_FD_GCNT_M) >> VSIQF_FD_CNT_FD_GCNT_S,
+               (fd_size & VSIQF_FD_CNT_FD_BCNT_M) >> VSIQF_FD_CNT_FD_BCNT_S,
+               (fd_cnt & VSIQF_FD_CNT_FD_GCNT_M) >> VSIQF_FD_CNT_FD_GCNT_S,
+               (fd_cnt & VSIQF_FD_CNT_FD_BCNT_M) >> VSIQF_FD_CNT_FD_BCNT_S);
+}
+
+/**
+ * ice_vf_verify_rx_desc - verify received FDIR programming status descriptor
+ * @vf: pointer to the VF info
+ * @ctx: FDIR context info for post processing
+ * @status: virtchnl FDIR program status
+ *
+ * Return: 0 on success, and other on error.
+ */
+static int
+ice_vf_verify_rx_desc(struct ice_vf *vf, struct ice_vf_fdir_ctx *ctx,
+                     enum virtchnl_fdir_prgm_status *status)
+{
+       struct device *dev = ice_pf_to_dev(vf->pf);
+       u32 stat_err, error, prog_id;
+       int ret;
+
+       stat_err = le16_to_cpu(ctx->rx_desc.wb.status_error0);
+       if (((stat_err & ICE_FXD_FLTR_WB_QW1_DD_M) >>
+           ICE_FXD_FLTR_WB_QW1_DD_S) != ICE_FXD_FLTR_WB_QW1_DD_YES) {
+               *status = VIRTCHNL_FDIR_FAILURE_RULE_NORESOURCE;
+               dev_err(dev, "VF %d: Desc Done not set\n", vf->vf_id);
+               ret = -EINVAL;
+               goto err_exit;
+       }
+
+       prog_id = (stat_err & ICE_FXD_FLTR_WB_QW1_PROG_ID_M) >>
+               ICE_FXD_FLTR_WB_QW1_PROG_ID_S;
+       if (prog_id == ICE_FXD_FLTR_WB_QW1_PROG_ADD &&
+           ctx->v_opcode != VIRTCHNL_OP_ADD_FDIR_FILTER) {
+               dev_err(dev, "VF %d: Desc show add, but ctx not",
+                       vf->vf_id);
+               *status = VIRTCHNL_FDIR_FAILURE_RULE_INVALID;
+               ret = -EINVAL;
+               goto err_exit;
+       }
+
+       if (prog_id == ICE_FXD_FLTR_WB_QW1_PROG_DEL &&
+           ctx->v_opcode != VIRTCHNL_OP_DEL_FDIR_FILTER) {
+               dev_err(dev, "VF %d: Desc show del, but ctx not",
+                       vf->vf_id);
+               *status = VIRTCHNL_FDIR_FAILURE_RULE_INVALID;
+               ret = -EINVAL;
+               goto err_exit;
+       }
+
+       error = (stat_err & ICE_FXD_FLTR_WB_QW1_FAIL_M) >>
+               ICE_FXD_FLTR_WB_QW1_FAIL_S;
+       if (error == ICE_FXD_FLTR_WB_QW1_FAIL_YES) {
+               if (prog_id == ICE_FXD_FLTR_WB_QW1_PROG_ADD) {
+                       dev_err(dev, "VF %d, Failed to add FDIR rule due to no space in the table",
+                               vf->vf_id);
+                       *status = VIRTCHNL_FDIR_FAILURE_RULE_NORESOURCE;
+               } else {
+                       dev_err(dev, "VF %d, Failed to remove FDIR rule, attempt to remove non-existent entry",
+                               vf->vf_id);
+                       *status = VIRTCHNL_FDIR_FAILURE_RULE_NONEXIST;
+               }
+               ret = -EINVAL;
+               goto err_exit;
+       }
+
+       error = (stat_err & ICE_FXD_FLTR_WB_QW1_FAIL_PROF_M) >>
+               ICE_FXD_FLTR_WB_QW1_FAIL_PROF_S;
+       if (error == ICE_FXD_FLTR_WB_QW1_FAIL_PROF_YES) {
+               dev_err(dev, "VF %d: Profile matching error", vf->vf_id);
+               *status = VIRTCHNL_FDIR_FAILURE_RULE_NORESOURCE;
+               ret = -EINVAL;
+               goto err_exit;
+       }
+
+       *status = VIRTCHNL_FDIR_SUCCESS;
+
+       return 0;
+
+err_exit:
+       ice_vf_fdir_dump_info(vf);
+       return ret;
+}
+
+/**
+ * ice_vc_add_fdir_fltr_post
+ * @vf: pointer to the VF structure
+ * @ctx: FDIR context info for post processing
+ * @status: virtchnl FDIR program status
+ * @success: true implies success, false implies failure
+ *
+ * Post process for flow director add command. If success, then do post process
+ * and send back success msg by virtchnl. Otherwise, do context reversion and
+ * send back failure msg by virtchnl.
+ *
+ * Return: 0 on success, and other on error.
+ */
+static int
+ice_vc_add_fdir_fltr_post(struct ice_vf *vf, struct ice_vf_fdir_ctx *ctx,
+                         enum virtchnl_fdir_prgm_status status,
+                         bool success)
+{
+       struct virtchnl_fdir_fltr_conf *conf = ctx->conf;
+       struct device *dev = ice_pf_to_dev(vf->pf);
+       enum virtchnl_status_code v_ret;
+       struct virtchnl_fdir_add *resp;
+       int ret, len, is_tun;
+
+       v_ret = VIRTCHNL_STATUS_SUCCESS;
+       len = sizeof(*resp);
+       resp = kzalloc(len, GFP_KERNEL);
+       if (!resp) {
+               len = 0;
+               v_ret = VIRTCHNL_STATUS_ERR_NO_MEMORY;
+               dev_dbg(dev, "VF %d: Alloc resp buf fail", vf->vf_id);
+               goto err_exit;
+       }
+
+       if (!success)
+               goto err_exit;
+
+       is_tun = 0;
+       resp->status = status;
+       resp->flow_id = conf->flow_id;
+       vf->fdir.fdir_fltr_cnt[conf->input.flow_type][is_tun]++;
+
+       ret = ice_vc_send_msg_to_vf(vf, ctx->v_opcode, v_ret,
+                                   (u8 *)resp, len);
+       kfree(resp);
+
+       dev_dbg(dev, "VF %d: flow_id:0x%X, FDIR %s success!\n",
+               vf->vf_id, conf->flow_id,
+               (ctx->v_opcode == VIRTCHNL_OP_ADD_FDIR_FILTER) ?
+               "add" : "del");
+       return ret;
+
+err_exit:
+       if (resp)
+               resp->status = status;
+       ice_vc_fdir_remove_entry(vf, conf, conf->flow_id);
+       devm_kfree(dev, conf);
+
+       ret = ice_vc_send_msg_to_vf(vf, ctx->v_opcode, v_ret,
+                                   (u8 *)resp, len);
+       kfree(resp);
+       return ret;
+}
+
+/**
+ * ice_vc_del_fdir_fltr_post
+ * @vf: pointer to the VF structure
+ * @ctx: FDIR context info for post processing
+ * @status: virtchnl FDIR program status
+ * @success: true implies success, false implies failure
+ *
+ * Post process for flow director del command. If success, then do post process
+ * and send back success msg by virtchnl. Otherwise, do context reversion and
+ * send back failure msg by virtchnl.
+ *
+ * Return: 0 on success, and other on error.
+ */
+static int
+ice_vc_del_fdir_fltr_post(struct ice_vf *vf, struct ice_vf_fdir_ctx *ctx,
+                         enum virtchnl_fdir_prgm_status status,
+                         bool success)
+{
+       struct virtchnl_fdir_fltr_conf *conf = ctx->conf;
+       struct device *dev = ice_pf_to_dev(vf->pf);
+       enum virtchnl_status_code v_ret;
+       struct virtchnl_fdir_del *resp;
+       int ret, len, is_tun;
+
+       v_ret = VIRTCHNL_STATUS_SUCCESS;
+       len = sizeof(*resp);
+       resp = kzalloc(len, GFP_KERNEL);
+       if (!resp) {
+               len = 0;
+               v_ret = VIRTCHNL_STATUS_ERR_NO_MEMORY;
+               dev_dbg(dev, "VF %d: Alloc resp buf fail", vf->vf_id);
+               goto err_exit;
+       }
+
+       if (!success)
+               goto err_exit;
+
+       is_tun = 0;
+       resp->status = status;
+       ice_vc_fdir_remove_entry(vf, conf, conf->flow_id);
+       vf->fdir.fdir_fltr_cnt[conf->input.flow_type][is_tun]--;
+
+       ret = ice_vc_send_msg_to_vf(vf, ctx->v_opcode, v_ret,
+                                   (u8 *)resp, len);
+       kfree(resp);
+
+       dev_dbg(dev, "VF %d: flow_id:0x%X, FDIR %s success!\n",
+               vf->vf_id, conf->flow_id,
+               (ctx->v_opcode == VIRTCHNL_OP_ADD_FDIR_FILTER) ?
+               "add" : "del");
+       devm_kfree(dev, conf);
+       return ret;
+
+err_exit:
+       if (resp)
+               resp->status = status;
+       if (success)
+               devm_kfree(dev, conf);
+
+       ret = ice_vc_send_msg_to_vf(vf, ctx->v_opcode, v_ret,
+                                   (u8 *)resp, len);
+       kfree(resp);
+       return ret;
+}
+
+/**
+ * ice_flush_fdir_ctx
+ * @pf: pointer to the PF structure
+ *
+ * Flush all the pending event on ctx_done list and process them.
+ */
+void ice_flush_fdir_ctx(struct ice_pf *pf)
+{
+       int i;
+
+       if (!test_and_clear_bit(__ICE_FD_VF_FLUSH_CTX, pf->state))
+               return;
+
+       ice_for_each_vf(pf, i) {
+               struct device *dev = ice_pf_to_dev(pf);
+               enum virtchnl_fdir_prgm_status status;
+               struct ice_vf *vf = &pf->vf[i];
+               struct ice_vf_fdir_ctx *ctx;
+               unsigned long flags;
+               int ret;
+
+               if (!test_bit(ICE_VF_STATE_ACTIVE, vf->vf_states))
+                       continue;
+
+               if (vf->ctrl_vsi_idx == ICE_NO_VSI)
+                       continue;
+
+               ctx = &vf->fdir.ctx_done;
+               spin_lock_irqsave(&vf->fdir.ctx_lock, flags);
+               if (!(ctx->flags & ICE_VF_FDIR_CTX_VALID)) {
+                       spin_unlock_irqrestore(&vf->fdir.ctx_lock, flags);
+                       continue;
+               }
+               spin_unlock_irqrestore(&vf->fdir.ctx_lock, flags);
+
+               WARN_ON(ctx->stat == ICE_FDIR_CTX_READY);
+               if (ctx->stat == ICE_FDIR_CTX_TIMEOUT) {
+                       status = VIRTCHNL_FDIR_FAILURE_RULE_TIMEOUT;
+                       dev_err(dev, "VF %d: ctrl_vsi irq timeout\n",
+                               vf->vf_id);
+                       goto err_exit;
+               }
+
+               ret = ice_vf_verify_rx_desc(vf, ctx, &status);
+               if (ret)
+                       goto err_exit;
+
+               if (ctx->v_opcode == VIRTCHNL_OP_ADD_FDIR_FILTER)
+                       ice_vc_add_fdir_fltr_post(vf, ctx, status, true);
+               else if (ctx->v_opcode == VIRTCHNL_OP_DEL_FDIR_FILTER)
+                       ice_vc_del_fdir_fltr_post(vf, ctx, status, true);
+               else
+                       dev_err(dev, "VF %d: Unsupported opcode\n", vf->vf_id);
+
+               spin_lock_irqsave(&vf->fdir.ctx_lock, flags);
+               ctx->flags &= ~ICE_VF_FDIR_CTX_VALID;
+               spin_unlock_irqrestore(&vf->fdir.ctx_lock, flags);
+               continue;
+err_exit:
+               if (ctx->v_opcode == VIRTCHNL_OP_ADD_FDIR_FILTER)
+                       ice_vc_add_fdir_fltr_post(vf, ctx, status, false);
+               else if (ctx->v_opcode == VIRTCHNL_OP_DEL_FDIR_FILTER)
+                       ice_vc_del_fdir_fltr_post(vf, ctx, status, false);
+               else
+                       dev_err(dev, "VF %d: Unsupported opcode\n", vf->vf_id);
+
+               spin_lock_irqsave(&vf->fdir.ctx_lock, flags);
+               ctx->flags &= ~ICE_VF_FDIR_CTX_VALID;
+               spin_unlock_irqrestore(&vf->fdir.ctx_lock, flags);
+       }
+}
+
+/**
+ * ice_vc_fdir_set_irq_ctx - set FDIR context info for later IRQ handler
+ * @vf: pointer to the VF structure
+ * @conf: FDIR configuration for each filter
+ * @v_opcode: virtual channel operation code
+ *
+ * Return: 0 on success, and other on error.
+ */
+static int
+ice_vc_fdir_set_irq_ctx(struct ice_vf *vf, struct virtchnl_fdir_fltr_conf *conf,
+                       enum virtchnl_ops v_opcode)
+{
+       struct device *dev = ice_pf_to_dev(vf->pf);
+       struct ice_vf_fdir_ctx *ctx;
+       unsigned long flags;
+
+       ctx = &vf->fdir.ctx_irq;
+       spin_lock_irqsave(&vf->fdir.ctx_lock, flags);
+       if ((vf->fdir.ctx_irq.flags & ICE_VF_FDIR_CTX_VALID) ||
+           (vf->fdir.ctx_done.flags & ICE_VF_FDIR_CTX_VALID)) {
+               spin_unlock_irqrestore(&vf->fdir.ctx_lock, flags);
+               dev_dbg(dev, "VF %d: Last request is still in progress\n",
+                       vf->vf_id);
+               return -EBUSY;
+       }
+       ctx->flags |= ICE_VF_FDIR_CTX_VALID;
+       spin_unlock_irqrestore(&vf->fdir.ctx_lock, flags);
+
+       ctx->conf = conf;
+       ctx->v_opcode = v_opcode;
+       ctx->stat = ICE_FDIR_CTX_READY;
+       timer_setup(&ctx->rx_tmr, ice_vf_fdir_timer, 0);
+
+       mod_timer(&ctx->rx_tmr, round_jiffies(msecs_to_jiffies(10) + jiffies));
+
+       return 0;
+}
+
+/**
+ * ice_vc_fdir_clear_irq_ctx - clear FDIR context info for IRQ handler
+ * @vf: pointer to the VF structure
+ *
+ * Return: 0 on success, and other on error.
+ */
+static void ice_vc_fdir_clear_irq_ctx(struct ice_vf *vf)
+{
+       struct ice_vf_fdir_ctx *ctx = &vf->fdir.ctx_irq;
+       unsigned long flags;
+
+       del_timer(&ctx->rx_tmr);
+       spin_lock_irqsave(&vf->fdir.ctx_lock, flags);
+       ctx->flags &= ~ICE_VF_FDIR_CTX_VALID;
+       spin_unlock_irqrestore(&vf->fdir.ctx_lock, flags);
+}
+
+/**
  * ice_vc_add_fdir_fltr - add a FDIR filter for VF by the msg buffer
  * @vf: pointer to the VF info
  * @msg: pointer to the msg buffer
@@ -1601,7 +2050,7 @@ int ice_vc_add_fdir_fltr(struct ice_vf *vf, u8 *msg)
                goto err_free_conf;
        }
 
-       ret = ice_vc_fdir_insert_entry(vf, conf, &stat->flow_id);
+       ret = ice_vc_fdir_insert_entry(vf, conf, &conf->flow_id);
        if (ret) {
                v_ret = VIRTCHNL_STATUS_SUCCESS;
                stat->status = VIRTCHNL_FDIR_FAILURE_RULE_NORESOURCE;
@@ -1609,6 +2058,14 @@ int ice_vc_add_fdir_fltr(struct ice_vf *vf, u8 *msg)
                goto err_free_conf;
        }
 
+       ret = ice_vc_fdir_set_irq_ctx(vf, conf, VIRTCHNL_OP_ADD_FDIR_FILTER);
+       if (ret) {
+               v_ret = VIRTCHNL_STATUS_SUCCESS;
+               stat->status = VIRTCHNL_FDIR_FAILURE_RULE_NORESOURCE;
+               dev_dbg(dev, "VF %d: set FDIR context failed\n", vf->vf_id);
+               goto err_free_conf;
+       }
+
        ret = ice_vc_fdir_write_fltr(vf, conf, true, is_tun);
        if (ret) {
                v_ret = VIRTCHNL_STATUS_SUCCESS;
@@ -1618,18 +2075,13 @@ int ice_vc_add_fdir_fltr(struct ice_vf *vf, u8 *msg)
                goto err_rem_entry;
        }
 
-       vf->fdir.fdir_fltr_cnt[conf->input.flow_type][is_tun]++;
-
-       v_ret = VIRTCHNL_STATUS_SUCCESS;
-       stat->status = VIRTCHNL_FDIR_SUCCESS;
 exit:
-       ret = ice_vc_send_msg_to_vf(vf, VIRTCHNL_OP_ADD_FDIR_FILTER, v_ret,
-                                   (u8 *)stat, len);
        kfree(stat);
        return ret;
 
 err_rem_entry:
-       ice_vc_fdir_remove_entry(vf, conf, stat->flow_id);
+       ice_vc_fdir_clear_irq_ctx(vf);
+       ice_vc_fdir_remove_entry(vf, conf, conf->flow_id);
 err_free_conf:
        devm_kfree(dev, conf);
 err_exit:
@@ -1693,22 +2145,29 @@ int ice_vc_del_fdir_fltr(struct ice_vf *vf, u8 *msg)
                goto err_exit;
        }
 
+       ret = ice_vc_fdir_set_irq_ctx(vf, conf, VIRTCHNL_OP_DEL_FDIR_FILTER);
+       if (ret) {
+               v_ret = VIRTCHNL_STATUS_SUCCESS;
+               stat->status = VIRTCHNL_FDIR_FAILURE_RULE_NORESOURCE;
+               dev_dbg(dev, "VF %d: set FDIR context failed\n", vf->vf_id);
+               goto err_exit;
+       }
+
        ret = ice_vc_fdir_write_fltr(vf, conf, false, is_tun);
        if (ret) {
                v_ret = VIRTCHNL_STATUS_SUCCESS;
                stat->status = VIRTCHNL_FDIR_FAILURE_RULE_NORESOURCE;
                dev_err(dev, "VF %d: writing FDIR rule failed, ret:%d\n",
                        vf->vf_id, ret);
-               goto err_exit;
+               goto err_del_tmr;
        }
 
-       ice_vc_fdir_remove_entry(vf, conf, fltr->flow_id);
-       devm_kfree(dev, conf);
-       vf->fdir.fdir_fltr_cnt[conf->input.flow_type][is_tun]--;
+       kfree(stat);
 
-       v_ret = VIRTCHNL_STATUS_SUCCESS;
-       stat->status = VIRTCHNL_FDIR_SUCCESS;
+       return ret;
 
+err_del_tmr:
+       ice_vc_fdir_clear_irq_ctx(vf);
 err_exit:
        ret = ice_vc_send_msg_to_vf(vf, VIRTCHNL_OP_DEL_FDIR_FILTER, v_ret,
                                    (u8 *)stat, len);
@@ -1726,6 +2185,10 @@ void ice_vf_fdir_init(struct ice_vf *vf)
 
        idr_init(&fdir->fdir_rule_idr);
        INIT_LIST_HEAD(&fdir->fdir_rule_list);
+
+       spin_lock_init(&fdir->ctx_lock);
+       fdir->ctx_irq.flags = 0;
+       fdir->ctx_done.flags = 0;
 }
 
 /**
index 2a2e0e5..f4e629f 100644 (file)
@@ -5,6 +5,24 @@
 #define _ICE_VIRTCHNL_FDIR_H_
 
 struct ice_vf;
+struct ice_pf;
+
+enum ice_fdir_ctx_stat {
+       ICE_FDIR_CTX_READY,
+       ICE_FDIR_CTX_IRQ,
+       ICE_FDIR_CTX_TIMEOUT,
+};
+
+struct ice_vf_fdir_ctx {
+       struct timer_list rx_tmr;
+       enum virtchnl_ops v_opcode;
+       enum ice_fdir_ctx_stat stat;
+       union ice_32b_rx_flex_desc rx_desc;
+#define ICE_VF_FDIR_CTX_VALID          BIT(0)
+       u32 flags;
+
+       void *conf;
+};
 
 /* VF FDIR information structure */
 struct ice_vf_fdir {
@@ -14,11 +32,24 @@ struct ice_vf_fdir {
 
        struct idr fdir_rule_idr;
        struct list_head fdir_rule_list;
+
+       spinlock_t ctx_lock; /* protects FDIR context info */
+       struct ice_vf_fdir_ctx ctx_irq;
+       struct ice_vf_fdir_ctx ctx_done;
 };
 
+#ifdef CONFIG_PCI_IOV
 int ice_vc_add_fdir_fltr(struct ice_vf *vf, u8 *msg);
 int ice_vc_del_fdir_fltr(struct ice_vf *vf, u8 *msg);
 void ice_vf_fdir_init(struct ice_vf *vf);
 void ice_vf_fdir_exit(struct ice_vf *vf);
-
+void
+ice_vc_fdir_irq_handler(struct ice_vsi *ctrl_vsi,
+                       union ice_32b_rx_flex_desc *rx_desc);
+void ice_flush_fdir_ctx(struct ice_pf *pf);
+#else
+static inline void
+ice_vc_fdir_irq_handler(struct ice_vsi *ctrl_vsi, union ice_32b_rx_flex_desc *rx_desc) { }
+static inline void ice_flush_fdir_ctx(struct ice_pf *pf) { }
+#endif /* CONFIG_PCI_IOV */
 #endif /* _ICE_VIRTCHNL_FDIR_H_ */