@@ -1993,19 +1993,16 @@ enum _ecore_status_t ecore_cxt_set_pf_params(struct ecore_hwfn *p_hwfn)
switch (p_hwfn->hw_info.personality) {
case ECORE_PCI_ETH:
{
- struct ecore_eth_pf_params *p_params =
+ struct ecore_eth_pf_params *p_params =
&p_hwfn->pf_params.eth_pf_params;
- /* TODO - we probably want to add VF number to the PF
- * params;
- * As of now, allocates 16 * 2 per-VF [to retain regular
- * functionality].
- */
- ecore_cxt_set_proto_cid_count(p_hwfn, PROTOCOLID_ETH,
- p_params->num_cons, 32);
- p_hwfn->p_cxt_mngr->arfs_count =
- p_params->num_arfs_filters;
- break;
+ if (!p_params->num_vf_cons)
+ p_params->num_vf_cons = ETH_PF_PARAMS_VF_CONS_DEFAULT;
+ ecore_cxt_set_proto_cid_count(p_hwfn, PROTOCOLID_ETH,
+ p_params->num_cons,
+ p_params->num_vf_cons);
+ p_hwfn->p_cxt_mngr->arfs_count = p_params->num_arfs_filters;
+ break;
}
default:
return ECORE_INVAL;
@@ -173,16 +173,19 @@ static void ecore_eth_queue_qid_usage_del(struct ecore_hwfn *p_hwfn,
void ecore_eth_queue_cid_release(struct ecore_hwfn *p_hwfn,
struct ecore_queue_cid *p_cid)
{
- /* For VF-queues, stuff is a bit complicated as:
- * - They always maintain the qid_usage on their own.
- * - In legacy mode, they also maintain their CIDs.
- */
+ bool b_legacy_vf = !!(p_cid->vf_legacy &
+ ECORE_QCID_LEGACY_VF_CID);
- /* VFs' CIDs are 0-based in PF-view, and uninitialized on VF */
- if (IS_PF(p_hwfn->p_dev) && !p_cid->b_legacy_vf)
+ /* VFs' CIDs are 0-based in PF-view, and uninitialized on VF.
+ * For legacy vf-queues, the CID doesn't go through here.
+ */
+ if (IS_PF(p_hwfn->p_dev) && !b_legacy_vf)
_ecore_cxt_release_cid(p_hwfn, p_cid->cid, p_cid->vfid);
- if (!p_cid->b_legacy_vf)
+
+ /* VFs maintain the index inside queue-zone on their own */
+ if (p_cid->vfid == ECORE_QUEUE_CID_PF)
ecore_eth_queue_qid_usage_del(p_hwfn, p_cid);
+
OSAL_VFREE(p_hwfn->p_dev, p_cid);
}
@@ -211,7 +214,7 @@ void ecore_eth_queue_cid_release(struct ecore_hwfn *p_hwfn,
if (p_vf_params != OSAL_NULL) {
p_cid->vfid = p_vf_params->vfid;
p_cid->vf_qid = p_vf_params->vf_qid;
- p_cid->b_legacy_vf = p_vf_params->b_legacy;
+ p_cid->vf_legacy = p_vf_params->vf_legacy;
} else {
p_cid->vfid = ECORE_QUEUE_CID_PF;
}
@@ -296,7 +299,8 @@ struct ecore_queue_cid *
if (p_vf_params) {
vfid = p_vf_params->vfid;
- if (p_vf_params->b_legacy) {
+ if (p_vf_params->vf_legacy &
+ ECORE_QCID_LEGACY_VF_CID) {
b_legacy_vf = true;
cid = p_vf_params->vf_qid;
}
@@ -928,12 +932,15 @@ enum _ecore_status_t
DMA_REGPAIR_LE(p_ramrod->cqe_pbl_addr, cqe_pbl_addr);
if (p_cid->vfid != ECORE_QUEUE_CID_PF) {
+ bool b_legacy_vf = !!(p_cid->vf_legacy &
+ ECORE_QCID_LEGACY_VF_RX_PROD);
+
p_ramrod->vf_rx_prod_index = p_cid->vf_qid;
DP_VERBOSE(p_hwfn, ECORE_MSG_SP,
"Queue%s is meant for VF rxq[%02x]\n",
- !!p_cid->b_legacy_vf ? " [legacy]" : "",
+ b_legacy_vf ? " [legacy]" : "",
p_cid->vf_qid);
- p_ramrod->vf_rx_prod_use_zone_a = !!p_cid->b_legacy_vf;
+ p_ramrod->vf_rx_prod_use_zone_a = b_legacy_vf;
}
return ecore_spq_post(p_hwfn, p_ent, OSAL_NULL);
@@ -34,7 +34,7 @@ struct ecore_queue_cid_vf_params {
* - Producers would be placed in a different place.
* - Makes assumptions regarding the CIDs.
*/
- bool b_legacy;
+ u8 vf_legacy;
/* For VFs, this index arrives via TLV to diffrentiate between
* different queues opened on the same qzone, and is passed
@@ -69,7 +69,9 @@ struct ecore_queue_cid {
u8 qid_usage_idx;
/* Legacy VFs might have Rx producer located elsewhere */
- bool b_legacy_vf;
+ u8 vf_legacy;
+#define ECORE_QCID_LEGACY_VF_RX_PROD (1 << 0)
+#define ECORE_QCID_LEGACY_VF_CID (1 << 1)
struct ecore_hwfn *p_owner;
};
@@ -22,6 +22,10 @@ struct ecore_eth_pf_params {
*/
u16 num_cons;
+ /* per-VF number of CIDs */
+ u8 num_vf_cons;
+#define ETH_PF_PARAMS_VF_CONS_DEFAULT (32)
+
/* To enable arfs, previous to HW-init a positive number needs to be
* set [as filters require allocated searcher ILT memory].
* This will set the maximal number of configured steering-filters.
@@ -53,9 +53,26 @@
"CHANNEL_TLV_VPORT_UPDATE_SGE_TPA",
"CHANNEL_TLV_UPDATE_TUNN_PARAM",
"CHANNEL_TLV_COALESCE_UPDATE",
+ "CHANNEL_TLV_QID",
"CHANNEL_TLV_MAX"
};
+static u8 ecore_vf_calculate_legacy(struct ecore_hwfn *p_hwfn,
+ struct ecore_vf_info *p_vf)
+{
+ u8 legacy = 0;
+
+ if (p_vf->acquire.vfdev_info.eth_fp_hsi_minor ==
+ ETH_HSI_VER_NO_PKT_LEN_TUNN)
+ legacy |= ECORE_QCID_LEGACY_VF_RX_PROD;
+
+ if (!(p_vf->acquire.vfdev_info.capabilities &
+ VFPF_ACQUIRE_CAP_QUEUE_QIDS))
+ legacy |= ECORE_QCID_LEGACY_VF_CID;
+
+ return legacy;
+}
+
/* IOV ramrods */
static enum _ecore_status_t ecore_sp_vf_start(struct ecore_hwfn *p_hwfn,
struct ecore_vf_info *p_vf)
@@ -1558,6 +1575,10 @@ static u8 ecore_iov_vf_mbx_acquire_resc(struct ecore_hwfn *p_hwfn,
p_resp->num_vlan_filters = OSAL_MIN_T(u8, p_vf->num_vlan_filters,
p_req->num_vlan_filters);
+ p_resp->num_cids =
+ OSAL_MIN_T(u8, p_req->num_cids,
+ p_hwfn->pf_params.eth_pf_params.num_vf_cons);
+
/* This isn't really needed/enforced, but some legacy VFs might depend
* on the correct filling of this field.
*/
@@ -1569,18 +1590,18 @@ static u8 ecore_iov_vf_mbx_acquire_resc(struct ecore_hwfn *p_hwfn,
p_resp->num_sbs < p_req->num_sbs ||
p_resp->num_mac_filters < p_req->num_mac_filters ||
p_resp->num_vlan_filters < p_req->num_vlan_filters ||
- p_resp->num_mc_filters < p_req->num_mc_filters) {
+ p_resp->num_mc_filters < p_req->num_mc_filters ||
+ p_resp->num_cids < p_req->num_cids) {
DP_VERBOSE(p_hwfn, ECORE_MSG_IOV,
- "VF[%d] - Insufficient resources: rxq [%02x/%02x]"
- " txq [%02x/%02x] sbs [%02x/%02x] mac [%02x/%02x]"
- " vlan [%02x/%02x] mc [%02x/%02x]\n",
+ "VF[%d] - Insufficient resources: rxq [%02x/%02x] txq [%02x/%02x] sbs [%02x/%02x] mac [%02x/%02x] vlan [%02x/%02x] mc [%02x/%02x] cids [%02x/%02x]\n",
p_vf->abs_vf_id,
p_req->num_rxqs, p_resp->num_rxqs,
p_req->num_rxqs, p_resp->num_txqs,
p_req->num_sbs, p_resp->num_sbs,
p_req->num_mac_filters, p_resp->num_mac_filters,
p_req->num_vlan_filters, p_resp->num_vlan_filters,
- p_req->num_mc_filters, p_resp->num_mc_filters);
+ p_req->num_mc_filters, p_resp->num_mc_filters,
+ p_req->num_cids, p_resp->num_cids);
/* Some legacy OSes are incapable of correctly handling this
* failure.
@@ -1715,6 +1736,12 @@ static void ecore_iov_vf_mbx_acquire(struct ecore_hwfn *p_hwfn,
if (p_hwfn->p_dev->num_hwfns > 1)
pfdev_info->capabilities |= PFVF_ACQUIRE_CAP_100G;
+ /* Share our ability to use multiple queue-ids only with VFs
+ * that request it.
+ */
+ if (req->vfdev_info.capabilities & VFPF_ACQUIRE_CAP_QUEUE_QIDS)
+ pfdev_info->capabilities |= PFVF_ACQUIRE_CAP_QUEUE_QIDS;
+
ecore_iov_vf_mbx_acquire_stats(p_hwfn, &pfdev_info->stats_info);
OSAL_MEMCPY(pfdev_info->port_mac, p_hwfn->hw_info.hw_mac_addr,
@@ -2158,6 +2185,42 @@ static void ecore_iov_vf_mbx_start_rxq_resp(struct ecore_hwfn *p_hwfn,
ecore_iov_send_response(p_hwfn, p_ptt, vf, length, status);
}
+static u8 ecore_iov_vf_mbx_qid(struct ecore_hwfn *p_hwfn,
+ struct ecore_vf_info *p_vf, bool b_is_tx)
+{
+ struct ecore_iov_vf_mbx *p_mbx = &p_vf->vf_mbx;
+ struct vfpf_qid_tlv *p_qid_tlv;
+
+ /* Search for the qid if the VF published if its going to provide it */
+ if (!(p_vf->acquire.vfdev_info.capabilities &
+ VFPF_ACQUIRE_CAP_QUEUE_QIDS)) {
+ if (b_is_tx)
+ return ECORE_IOV_LEGACY_QID_TX;
+ else
+ return ECORE_IOV_LEGACY_QID_RX;
+ }
+
+ p_qid_tlv = (struct vfpf_qid_tlv *)
+ ecore_iov_search_list_tlvs(p_hwfn, p_mbx->req_virt,
+ CHANNEL_TLV_QID);
+ if (p_qid_tlv == OSAL_NULL) {
+ DP_VERBOSE(p_hwfn, ECORE_MSG_IOV,
+ "VF[%2x]: Failed to provide qid\n",
+ p_vf->relative_vf_id);
+
+ return ECORE_IOV_QID_INVALID;
+ }
+
+ if (p_qid_tlv->qid >= MAX_QUEUES_PER_QZONE) {
+ DP_VERBOSE(p_hwfn, ECORE_MSG_IOV,
+ "VF[%02x]: Provided qid out-of-bounds %02x\n",
+ p_vf->relative_vf_id, p_qid_tlv->qid);
+ return ECORE_IOV_QID_INVALID;
+ }
+
+ return p_qid_tlv->qid;
+}
+
static void ecore_iov_vf_mbx_start_rxq(struct ecore_hwfn *p_hwfn,
struct ecore_ptt *p_ptt,
struct ecore_vf_info *vf)
@@ -2166,11 +2229,10 @@ static void ecore_iov_vf_mbx_start_rxq(struct ecore_hwfn *p_hwfn,
struct ecore_queue_cid_vf_params vf_params;
struct ecore_iov_vf_mbx *mbx = &vf->vf_mbx;
u8 status = PFVF_STATUS_NO_RESOURCE;
+ u8 qid_usage_idx, vf_legacy = 0;
struct ecore_vf_queue *p_queue;
struct vfpf_start_rxq_tlv *req;
struct ecore_queue_cid *p_cid;
- bool b_legacy_vf = false;
- u8 qid_usage_idx;
enum _ecore_status_t rc;
req = &mbx->req_virt->start_rxq;
@@ -2180,18 +2242,17 @@ static void ecore_iov_vf_mbx_start_rxq(struct ecore_hwfn *p_hwfn,
!ecore_iov_validate_sb(p_hwfn, vf, req->hw_sb))
goto out;
- /* Legacy VFs made assumptions on the CID their queues connected to,
- * assuming queue X used CID X.
- * TODO - need to validate that there was no official release post
- * the current legacy scheme that still made that assumption.
- */
- if (vf->acquire.vfdev_info.eth_fp_hsi_minor ==
- ETH_HSI_VER_NO_PKT_LEN_TUNN)
- b_legacy_vf = true;
+ qid_usage_idx = ecore_iov_vf_mbx_qid(p_hwfn, vf, false);
+ if (qid_usage_idx == ECORE_IOV_QID_INVALID)
+ goto out;
- /* Acquire a new queue-cid */
p_queue = &vf->vf_queues[req->rx_qid];
+ if (p_queue->cids[qid_usage_idx].p_cid)
+ goto out;
+
+ vf_legacy = ecore_vf_calculate_legacy(p_hwfn, vf);
+ /* Acquire a new queue-cid */
OSAL_MEMSET(¶ms, 0, sizeof(params));
params.queue_id = (u8)p_queue->fw_rx_qid;
params.vport_id = vf->vport_id;
@@ -2199,15 +2260,10 @@ static void ecore_iov_vf_mbx_start_rxq(struct ecore_hwfn *p_hwfn,
params.sb = req->hw_sb;
params.sb_idx = req->sb_index;
- /* TODO - set qid_usage_idx according to extended TLV. For now, use
- * '0' for Rx.
- */
- qid_usage_idx = 0;
-
OSAL_MEM_ZERO(&vf_params, sizeof(vf_params));
vf_params.vfid = vf->relative_vf_id;
vf_params.vf_qid = (u8)req->rx_qid;
- vf_params.b_legacy = b_legacy_vf;
+ vf_params.vf_legacy = vf_legacy;
vf_params.qid_usage_idx = qid_usage_idx;
p_cid = ecore_eth_queue_to_cid(p_hwfn, vf->opaque_fid,
@@ -2218,7 +2274,7 @@ static void ecore_iov_vf_mbx_start_rxq(struct ecore_hwfn *p_hwfn,
/* Legacy VFs have their Producers in a different location, which they
* calculate on their own and clean the producer prior to this.
*/
- if (!b_legacy_vf)
+ if (!(vf_legacy & ECORE_QCID_LEGACY_VF_RX_PROD))
REG_WR(p_hwfn,
GTT_BAR0_MAP_REG_MSDM_RAM +
MSTORM_ETH_VF_PRODS_OFFSET(vf->abs_vf_id, req->rx_qid),
@@ -2241,7 +2297,8 @@ static void ecore_iov_vf_mbx_start_rxq(struct ecore_hwfn *p_hwfn,
out:
ecore_iov_vf_mbx_start_rxq_resp(p_hwfn, p_ptt, vf, status,
- b_legacy_vf);
+ !!(vf_legacy &
+ ECORE_QCID_LEGACY_VF_RX_PROD));
}
static void
@@ -2443,8 +2500,7 @@ static void ecore_iov_vf_mbx_start_txq(struct ecore_hwfn *p_hwfn,
struct ecore_vf_queue *p_queue;
struct vfpf_start_txq_tlv *req;
struct ecore_queue_cid *p_cid;
- bool b_legacy_vf = false;
- u8 qid_usage_idx;
+ u8 qid_usage_idx, vf_legacy;
u32 cid = 0;
enum _ecore_status_t rc;
u16 pq;
@@ -2457,35 +2513,27 @@ static void ecore_iov_vf_mbx_start_txq(struct ecore_hwfn *p_hwfn,
!ecore_iov_validate_sb(p_hwfn, vf, req->hw_sb))
goto out;
- /* In case this is a legacy VF - need to know to use the right cids.
- * TODO - need to validate that there was no official release post
- * the current legacy scheme that still made that assumption.
- */
- if (vf->acquire.vfdev_info.eth_fp_hsi_minor ==
- ETH_HSI_VER_NO_PKT_LEN_TUNN)
- b_legacy_vf = true;
+ qid_usage_idx = ecore_iov_vf_mbx_qid(p_hwfn, vf, true);
+ if (qid_usage_idx == ECORE_IOV_QID_INVALID)
+ goto out;
- /* Acquire a new queue-cid */
p_queue = &vf->vf_queues[req->tx_qid];
+ if (p_queue->cids[qid_usage_idx].p_cid)
+ goto out;
+
+ vf_legacy = ecore_vf_calculate_legacy(p_hwfn, vf);
+ /* Acquire a new queue-cid */
params.queue_id = p_queue->fw_tx_qid;
params.vport_id = vf->vport_id;
params.stats_id = vf->abs_vf_id + 0x10;
params.sb = req->hw_sb;
params.sb_idx = req->sb_index;
- /* TODO - set qid_usage_idx according to extended TLV. For now, use
- * '1' for Tx.
- */
- qid_usage_idx = 1;
-
- if (p_queue->cids[qid_usage_idx].p_cid)
- goto out;
-
OSAL_MEM_ZERO(&vf_params, sizeof(vf_params));
vf_params.vfid = vf->relative_vf_id;
vf_params.vf_qid = (u8)req->tx_qid;
- vf_params.b_legacy = b_legacy_vf;
+ vf_params.vf_legacy = vf_legacy;
vf_params.qid_usage_idx = qid_usage_idx;
p_cid = ecore_eth_queue_to_cid(p_hwfn, vf->opaque_fid,
@@ -2515,80 +2563,74 @@ static void ecore_iov_vf_mbx_start_txq(struct ecore_hwfn *p_hwfn,
static enum _ecore_status_t ecore_iov_vf_stop_rxqs(struct ecore_hwfn *p_hwfn,
struct ecore_vf_info *vf,
u16 rxq_id,
- u8 num_rxqs,
+ u8 qid_usage_idx,
bool cqe_completion)
{
+ struct ecore_vf_queue *p_queue;
enum _ecore_status_t rc = ECORE_SUCCESS;
- int qid, i;
- /* TODO - improve validation [wrap around] */
- if (rxq_id + num_rxqs > OSAL_ARRAY_SIZE(vf->vf_queues))
+ if (!ecore_iov_validate_rxq(p_hwfn, vf, rxq_id,
+ ECORE_IOV_VALIDATE_Q_NA)) {
+ DP_VERBOSE(p_hwfn, ECORE_MSG_IOV,
+ "VF[%d] Tried Closing Rx 0x%04x.%02x which is inactive\n",
+ vf->relative_vf_id, rxq_id, qid_usage_idx);
return ECORE_INVAL;
+ }
- for (qid = rxq_id; qid < rxq_id + num_rxqs; qid++) {
- struct ecore_vf_queue *p_queue = &vf->vf_queues[qid];
- struct ecore_queue_cid **pp_cid = OSAL_NULL;
-
- /* There can be at most a single Rx per qzone. Find it */
- for (i = 0; i < MAX_QUEUES_PER_QZONE; i++) {
- if (p_queue->cids[i].p_cid &&
- !p_queue->cids[i].b_is_tx) {
- pp_cid = &p_queue->cids[i].p_cid;
- break;
- }
- }
- if (pp_cid == OSAL_NULL) {
- DP_VERBOSE(p_hwfn, ECORE_MSG_IOV,
- "Ignoring VF[%02x] request of closing Rx queue %04x - closed\n",
- vf->relative_vf_id, qid);
- continue;
- }
+ p_queue = &vf->vf_queues[rxq_id];
- rc = ecore_eth_rx_queue_stop(p_hwfn, *pp_cid,
- false, cqe_completion);
- if (rc != ECORE_SUCCESS)
- return rc;
+ /* We've validated the index and the existence of the active RXQ -
+ * now we need to make sure that it's using the correct qid.
+ */
+ if (!p_queue->cids[qid_usage_idx].p_cid ||
+ p_queue->cids[qid_usage_idx].b_is_tx) {
+ struct ecore_queue_cid *p_cid;
- *pp_cid = OSAL_NULL;
- vf->num_active_rxqs--;
+ p_cid = ecore_iov_get_vf_rx_queue_cid(p_hwfn, vf, p_queue);
+ DP_VERBOSE(p_hwfn, ECORE_MSG_IOV,
+ "VF[%d] - Tried Closing Rx 0x%04x.%02x, but Rx is at %04x.%02x\n",
+ vf->relative_vf_id, rxq_id, qid_usage_idx,
+ rxq_id, p_cid->qid_usage_idx);
+ return ECORE_INVAL;
}
- return rc;
+ /* Now that we know we have a valid Rx-queue - close it */
+ rc = ecore_eth_rx_queue_stop(p_hwfn,
+ p_queue->cids[qid_usage_idx].p_cid,
+ false, cqe_completion);
+ if (rc != ECORE_SUCCESS)
+ return rc;
+
+ p_queue->cids[qid_usage_idx].p_cid = OSAL_NULL;
+ vf->num_active_rxqs--;
+
+ return ECORE_SUCCESS;
}
static enum _ecore_status_t ecore_iov_vf_stop_txqs(struct ecore_hwfn *p_hwfn,
struct ecore_vf_info *vf,
- u16 txq_id, u8 num_txqs)
+ u16 txq_id,
+ u8 qid_usage_idx)
{
- enum _ecore_status_t rc = ECORE_SUCCESS;
struct ecore_vf_queue *p_queue;
- int qid, j;
+ enum _ecore_status_t rc = ECORE_SUCCESS;
if (!ecore_iov_validate_txq(p_hwfn, vf, txq_id,
- ECORE_IOV_VALIDATE_Q_NA) ||
- !ecore_iov_validate_txq(p_hwfn, vf, txq_id + num_txqs,
ECORE_IOV_VALIDATE_Q_NA))
return ECORE_INVAL;
- for (qid = txq_id; qid < txq_id + num_txqs; qid++) {
- p_queue = &vf->vf_queues[qid];
- for (j = 0; j < MAX_QUEUES_PER_QZONE; j++) {
- if (p_queue->cids[j].p_cid == OSAL_NULL)
- continue;
-
- if (!p_queue->cids[j].b_is_tx)
- continue;
-
- rc = ecore_eth_tx_queue_stop(p_hwfn,
- p_queue->cids[j].p_cid);
- if (rc != ECORE_SUCCESS)
- return rc;
+ p_queue = &vf->vf_queues[txq_id];
+ if (!p_queue->cids[qid_usage_idx].p_cid ||
+ !p_queue->cids[qid_usage_idx].b_is_tx)
+ return ECORE_INVAL;
- p_queue->cids[j].p_cid = OSAL_NULL;
- }
- }
+ rc = ecore_eth_tx_queue_stop(p_hwfn,
+ p_queue->cids[qid_usage_idx].p_cid);
+ if (rc != ECORE_SUCCESS)
+ return rc;
- return rc;
+ p_queue->cids[qid_usage_idx].p_cid = OSAL_NULL;
+ return ECORE_SUCCESS;
}
static void ecore_iov_vf_mbx_stop_rxqs(struct ecore_hwfn *p_hwfn,
@@ -2597,20 +2639,34 @@ static void ecore_iov_vf_mbx_stop_rxqs(struct ecore_hwfn *p_hwfn,
{
u16 length = sizeof(struct pfvf_def_resp_tlv);
struct ecore_iov_vf_mbx *mbx = &vf->vf_mbx;
- u8 status = PFVF_STATUS_SUCCESS;
+ u8 status = PFVF_STATUS_FAILURE;
struct vfpf_stop_rxqs_tlv *req;
+ u8 qid_usage_idx;
enum _ecore_status_t rc;
- /* We give the option of starting from qid != 0, in this case we
- * need to make sure that qid + num_qs doesn't exceed the actual
- * amount of queues that exist.
+ /* Starting with CHANNEL_TLV_QID, it's assumed the 'num_rxqs'
+ * would be one. Since no older ecore passed multiple queues
+ * using this API, sanitize on the value.
*/
req = &mbx->req_virt->stop_rxqs;
- rc = ecore_iov_vf_stop_rxqs(p_hwfn, vf, req->rx_qid,
- req->num_rxqs, req->cqe_completion);
- if (rc)
- status = PFVF_STATUS_FAILURE;
+ if (req->num_rxqs != 1) {
+ DP_VERBOSE(p_hwfn, ECORE_MSG_IOV,
+ "Odd; VF[%d] tried stopping multiple Rx queues\n",
+ vf->relative_vf_id);
+ status = PFVF_STATUS_NOT_SUPPORTED;
+ goto out;
+ }
+ /* Find which qid-index is associated with the queue */
+ qid_usage_idx = ecore_iov_vf_mbx_qid(p_hwfn, vf, false);
+ if (qid_usage_idx == ECORE_IOV_QID_INVALID)
+ goto out;
+
+ rc = ecore_iov_vf_stop_rxqs(p_hwfn, vf, req->rx_qid,
+ qid_usage_idx, req->cqe_completion);
+ if (rc == ECORE_SUCCESS)
+ status = PFVF_STATUS_SUCCESS;
+out:
ecore_iov_prepare_resp(p_hwfn, p_ptt, vf, CHANNEL_TLV_STOP_RXQS,
length, status);
}
@@ -2621,19 +2677,35 @@ static void ecore_iov_vf_mbx_stop_txqs(struct ecore_hwfn *p_hwfn,
{
u16 length = sizeof(struct pfvf_def_resp_tlv);
struct ecore_iov_vf_mbx *mbx = &vf->vf_mbx;
- u8 status = PFVF_STATUS_SUCCESS;
+ u8 status = PFVF_STATUS_FAILURE;
struct vfpf_stop_txqs_tlv *req;
+ u8 qid_usage_idx;
enum _ecore_status_t rc;
- /* We give the option of starting from qid != 0, in this case we
- * need to make sure that qid + num_qs doesn't exceed the actual
- * amount of queues that exist.
+ /* Starting with CHANNEL_TLV_QID, it's assumed the 'num_txqs'
+ * would be one. Since no older ecore passed multiple queues
+ * using this API, sanitize on the value.
*/
req = &mbx->req_virt->stop_txqs;
- rc = ecore_iov_vf_stop_txqs(p_hwfn, vf, req->tx_qid, req->num_txqs);
- if (rc)
- status = PFVF_STATUS_FAILURE;
+ if (req->num_txqs != 1) {
+ DP_VERBOSE(p_hwfn, ECORE_MSG_IOV,
+ "Odd; VF[%d] tried stopping multiple Tx queues\n",
+ vf->relative_vf_id);
+ status = PFVF_STATUS_NOT_SUPPORTED;
+ goto out;
+ }
+ /* Find which qid-index is associated with the queue */
+ qid_usage_idx = ecore_iov_vf_mbx_qid(p_hwfn, vf, true);
+ if (qid_usage_idx == ECORE_IOV_QID_INVALID)
+ goto out;
+
+ rc = ecore_iov_vf_stop_txqs(p_hwfn, vf, req->tx_qid,
+ qid_usage_idx);
+ if (rc == ECORE_SUCCESS)
+ status = PFVF_STATUS_SUCCESS;
+
+out:
ecore_iov_prepare_resp(p_hwfn, p_ptt, vf, CHANNEL_TLV_STOP_TXQS,
length, status);
}
@@ -2649,6 +2721,7 @@ static void ecore_iov_vf_mbx_update_rxqs(struct ecore_hwfn *p_hwfn,
u8 status = PFVF_STATUS_FAILURE;
u8 complete_event_flg;
u8 complete_cqe_flg;
+ u8 qid_usage_idx;
enum _ecore_status_t rc;
u16 i;
@@ -2656,10 +2729,30 @@ static void ecore_iov_vf_mbx_update_rxqs(struct ecore_hwfn *p_hwfn,
complete_cqe_flg = !!(req->flags & VFPF_RXQ_UPD_COMPLETE_CQE_FLAG);
complete_event_flg = !!(req->flags & VFPF_RXQ_UPD_COMPLETE_EVENT_FLAG);
- /* Validate inputs */
+ qid_usage_idx = ecore_iov_vf_mbx_qid(p_hwfn, vf, false);
+ if (qid_usage_idx == ECORE_IOV_QID_INVALID)
+ goto out;
+
+ /* Starting with the addition of CHANNEL_TLV_QID, this API started
+ * expecting a single queue at a time. Validate this.
+ */
+ if ((vf->acquire.vfdev_info.capabilities &
+ VFPF_ACQUIRE_CAP_QUEUE_QIDS) &&
+ req->num_rxqs != 1) {
+ DP_VERBOSE(p_hwfn, ECORE_MSG_IOV,
+ "VF[%d] supports QIDs but sends multiple queues\n",
+ vf->relative_vf_id);
+ goto out;
+ }
+
+ /* Validate inputs - for the legacy case this is still true since
+ * qid_usage_idx for each Rx queue would be LEGACY_QID_RX.
+ */
for (i = req->rx_qid; i < req->rx_qid + req->num_rxqs; i++) {
if (!ecore_iov_validate_rxq(p_hwfn, vf, i,
- ECORE_IOV_VALIDATE_Q_ENABLE)) {
+ ECORE_IOV_VALIDATE_Q_NA) ||
+ !vf->vf_queues[i].cids[qid_usage_idx].p_cid ||
+ vf->vf_queues[i].cids[qid_usage_idx].b_is_tx) {
DP_VERBOSE(p_hwfn, ECORE_MSG_IOV,
"VF[%d]: Incorrect Rxqs [%04x, %02x]\n",
vf->relative_vf_id, req->rx_qid,
@@ -2669,12 +2762,9 @@ static void ecore_iov_vf_mbx_update_rxqs(struct ecore_hwfn *p_hwfn,
}
for (i = 0; i < req->num_rxqs; i++) {
- struct ecore_vf_queue *p_queue;
u16 qid = req->rx_qid + i;
- p_queue = &vf->vf_queues[qid];
- handlers[i] = ecore_iov_get_vf_rx_queue_cid(p_hwfn, vf,
- p_queue);
+ handlers[i] = vf->vf_queues[qid].cids[qid_usage_idx].p_cid;
}
rc = ecore_sp_eth_rx_queues_update(p_hwfn, (void **)&handlers,
@@ -2683,7 +2773,7 @@ static void ecore_iov_vf_mbx_update_rxqs(struct ecore_hwfn *p_hwfn,
complete_event_flg,
ECORE_SPQ_MODE_EBLOCK,
OSAL_NULL);
- if (rc)
+ if (rc != ECORE_SUCCESS)
goto out;
status = PFVF_STATUS_SUCCESS;
@@ -63,6 +63,10 @@ struct ecore_iov_vf_mbx {
*/
};
+#define ECORE_IOV_LEGACY_QID_RX (0)
+#define ECORE_IOV_LEGACY_QID_TX (1)
+#define ECORE_IOV_QID_INVALID (0xFE)
+
struct ecore_vf_queue_cid {
bool b_is_tx;
struct ecore_queue_cid *p_cid;
@@ -135,22 +135,36 @@ static void ecore_vf_pf_req_end(struct ecore_hwfn *p_hwfn,
return rc;
}
+static void ecore_vf_pf_add_qid(struct ecore_hwfn *p_hwfn,
+ struct ecore_queue_cid *p_cid)
+{
+ struct ecore_vf_iov *p_iov = p_hwfn->vf_iov_info;
+ struct vfpf_qid_tlv *p_qid_tlv;
+
+ /* Only add QIDs for the queue if it was negotiated with PF */
+ if (!(p_iov->acquire_resp.pfdev_info.capabilities &
+ PFVF_ACQUIRE_CAP_QUEUE_QIDS))
+ return;
+
+ p_qid_tlv = ecore_add_tlv(p_hwfn, &p_iov->offset,
+ CHANNEL_TLV_QID, sizeof(*p_qid_tlv));
+ p_qid_tlv->qid = p_cid->qid_usage_idx;
+}
+
#define VF_ACQUIRE_THRESH 3
static void ecore_vf_pf_acquire_reduce_resc(struct ecore_hwfn *p_hwfn,
struct vf_pf_resc_request *p_req,
struct pf_vf_resc *p_resp)
{
DP_VERBOSE(p_hwfn, ECORE_MSG_IOV,
- "PF unwilling to fullill resource request: rxq [%02x/%02x]"
- " txq [%02x/%02x] sbs [%02x/%02x] mac [%02x/%02x]"
- " vlan [%02x/%02x] mc [%02x/%02x]."
- " Try PF recommended amount\n",
+ "PF unwilling to fullill resource request: rxq [%02x/%02x] txq [%02x/%02x] sbs [%02x/%02x] mac [%02x/%02x] vlan [%02x/%02x] mc [%02x/%02x] cids [%02x/%02x]. Try PF recommended amount\n",
p_req->num_rxqs, p_resp->num_rxqs,
p_req->num_rxqs, p_resp->num_txqs,
p_req->num_sbs, p_resp->num_sbs,
p_req->num_mac_filters, p_resp->num_mac_filters,
p_req->num_vlan_filters, p_resp->num_vlan_filters,
- p_req->num_mc_filters, p_resp->num_mc_filters);
+ p_req->num_mc_filters, p_resp->num_mc_filters,
+ p_req->num_cids, p_resp->num_cids);
/* humble our request */
p_req->num_txqs = p_resp->num_txqs;
@@ -159,6 +173,7 @@ static void ecore_vf_pf_acquire_reduce_resc(struct ecore_hwfn *p_hwfn,
p_req->num_mac_filters = p_resp->num_mac_filters;
p_req->num_vlan_filters = p_resp->num_vlan_filters;
p_req->num_mc_filters = p_resp->num_mc_filters;
+ p_req->num_cids = p_resp->num_cids;
}
static enum _ecore_status_t ecore_vf_pf_acquire(struct ecore_hwfn *p_hwfn)
@@ -185,6 +200,7 @@ static enum _ecore_status_t ecore_vf_pf_acquire(struct ecore_hwfn *p_hwfn)
p_resc->num_sbs = ECORE_MAX_VF_CHAINS_PER_PF;
p_resc->num_mac_filters = ECORE_ETH_VF_NUM_MAC_FILTERS;
p_resc->num_vlan_filters = ECORE_ETH_VF_NUM_VLAN_FILTERS;
+ p_resc->num_cids = ECORE_ETH_VF_DEFAULT_NUM_CIDS;
OSAL_MEMSET(&vf_sw_info, 0, sizeof(vf_sw_info));
OSAL_VF_FILL_ACQUIRE_RESC_REQ(p_hwfn, &req->resc_request, &vf_sw_info);
@@ -310,6 +326,15 @@ static enum _ecore_status_t ecore_vf_pf_acquire(struct ecore_hwfn *p_hwfn)
VFPF_ACQUIRE_CAP_PRE_FP_HSI)
p_iov->b_pre_fp_hsi = true;
+ /* In case PF doesn't support multi-queue Tx, update the number of
+ * CIDs to reflect the number of queues [older PFs didn't fill that
+ * field].
+ */
+ if (!(resp->pfdev_info.capabilities &
+ PFVF_ACQUIRE_CAP_QUEUE_QIDS))
+ resp->resc.num_cids = resp->resc.num_rxqs +
+ resp->resc.num_txqs;
+
rc = OSAL_VF_UPDATE_ACQUIRE_RESC_RESP(p_hwfn, &resp->resc);
if (rc) {
DP_NOTICE(p_hwfn, true,
@@ -649,6 +674,8 @@ enum _ecore_status_t
(u32 *)(&init_prod_val));
}
+ ecore_vf_pf_add_qid(p_hwfn, p_cid);
+
/* add list termination tlv */
ecore_add_tlv(p_hwfn, &p_iov->offset,
CHANNEL_TLV_LIST_END,
@@ -704,6 +731,8 @@ enum _ecore_status_t ecore_vf_pf_rxq_stop(struct ecore_hwfn *p_hwfn,
req->num_rxqs = 1;
req->cqe_completion = cqe_completion;
+ ecore_vf_pf_add_qid(p_hwfn, p_cid);
+
/* add list termination tlv */
ecore_add_tlv(p_hwfn, &p_iov->offset,
CHANNEL_TLV_LIST_END,
@@ -748,6 +777,8 @@ enum _ecore_status_t
req->hw_sb = p_cid->rel.sb;
req->sb_index = p_cid->rel.sb_idx;
+ ecore_vf_pf_add_qid(p_hwfn, p_cid);
+
/* add list termination tlv */
ecore_add_tlv(p_hwfn, &p_iov->offset,
CHANNEL_TLV_LIST_END,
@@ -799,6 +830,8 @@ enum _ecore_status_t ecore_vf_pf_txq_stop(struct ecore_hwfn *p_hwfn,
req->tx_qid = p_cid->rel.queue_id;
req->num_txqs = 1;
+ ecore_vf_pf_add_qid(p_hwfn, p_cid);
+
/* add list termination tlv */
ecore_add_tlv(p_hwfn, &p_iov->offset,
CHANNEL_TLV_LIST_END,
@@ -831,32 +864,30 @@ enum _ecore_status_t ecore_vf_pf_rxqs_update(struct ecore_hwfn *p_hwfn,
struct vfpf_update_rxq_tlv *req;
enum _ecore_status_t rc;
- /* TODO - API is limited to assuming continuous regions of queues,
- * but VF queues might not fullfil this requirement.
- * Need to consider whether we need new TLVs for this, or whether
- * simply doing it iteratively is good enough.
+ /* Starting with CHANNEL_TLV_QID and the need for additional queue
+ * information, this API stopped supporting multiple rxqs.
+ * TODO - remove this and change the API to accept a single queue-cid
+ * in a follow-up patch.
*/
- if (!num_rxqs)
+ if (num_rxqs != 1) {
+ DP_NOTICE(p_hwfn, true,
+ "VFs can no longer update more than a single queue\n");
return ECORE_INVAL;
+ }
-again:
/* clear mailbox and prep first tlv */
req = ecore_vf_pf_prep(p_hwfn, CHANNEL_TLV_UPDATE_RXQ, sizeof(*req));
- /* Find the length of the current contagious range of queues beginning
- * at first queue's index.
- */
req->rx_qid = (*pp_cid)->rel.queue_id;
- for (req->num_rxqs = 1; req->num_rxqs < num_rxqs; req->num_rxqs++)
- if (pp_cid[req->num_rxqs]->rel.queue_id !=
- req->rx_qid + req->num_rxqs)
- break;
+ req->num_rxqs = 1;
if (comp_cqe_flg)
req->flags |= VFPF_RXQ_UPD_COMPLETE_CQE_FLAG;
if (comp_event_flg)
req->flags |= VFPF_RXQ_UPD_COMPLETE_EVENT_FLAG;
+ ecore_vf_pf_add_qid(p_hwfn, *pp_cid);
+
/* add list termination tlv */
ecore_add_tlv(p_hwfn, &p_iov->offset,
CHANNEL_TLV_LIST_END,
@@ -871,15 +902,6 @@ enum _ecore_status_t ecore_vf_pf_rxqs_update(struct ecore_hwfn *p_hwfn,
goto exit;
}
- /* Make sure we're done with all the queues */
- if (req->num_rxqs < num_rxqs) {
- num_rxqs -= req->num_rxqs;
- pp_cid += req->num_rxqs;
- /* TODO - should we give a non-locked variant instead? */
- ecore_vf_pf_req_end(p_hwfn, rc);
- goto again;
- }
-
exit:
ecore_vf_pf_req_end(p_hwfn, rc);
return rc;
@@ -14,6 +14,11 @@
#include "ecore_l2_api.h"
#include "ecore_vfpf_if.h"
+/* Default number of CIDs [total of both Rx and Tx] to be requested
+ * by default.
+ */
+#define ECORE_ETH_VF_DEFAULT_NUM_CIDS (32)
+
/* This data is held in the ecore_hwfn structure for VFs only. */
struct ecore_vf_iov {
union vfpf_tlvs *vf2pf_request;
@@ -19,13 +19,14 @@
*
**/
struct vf_pf_resc_request {
- u8 num_rxqs;
- u8 num_txqs;
- u8 num_sbs;
- u8 num_mac_filters;
- u8 num_vlan_filters;
- u8 num_mc_filters; /* No limit so superfluous */
- u16 padding;
+ u8 num_rxqs;
+ u8 num_txqs;
+ u8 num_sbs;
+ u8 num_mac_filters;
+ u8 num_vlan_filters;
+ u8 num_mc_filters; /* No limit so superfluous */
+ u8 num_cids;
+ u8 padding;
};
struct hw_sb_info {
@@ -92,6 +93,14 @@ struct vfpf_acquire_tlv {
/* VF pre-FP hsi version */
#define VFPF_ACQUIRE_CAP_PRE_FP_HSI (1 << 0)
#define VFPF_ACQUIRE_CAP_100G (1 << 1) /* VF can support 100g */
+
+ /* A requirement for supporting multi-Tx queues on a single queue-zone,
+ * VF would pass qids as additional information whenever passing queue
+ * references.
+ * TODO - due to the CID limitations in Bar0, VFs currently don't pass
+ * this, and use the legacy CID scheme.
+ */
+#define VFPF_ACQUIRE_CAP_QUEUE_QIDS (1 << 2)
u64 capabilities;
u8 fw_major;
u8 fw_minor;
@@ -170,6 +179,9 @@ struct pfvf_acquire_resp_tlv {
#endif
#define PFVF_ACQUIRE_CAP_POST_FW_OVERRIDE (1 << 2)
+ /* PF expects queues to be received with additional qids */
+#define PFVF_ACQUIRE_CAP_QUEUE_QIDS (1 << 3)
+
u16 db_size;
u8 indices_per_sb;
u8 os_type;
@@ -210,7 +222,8 @@ struct pfvf_acquire_resp_tlv {
u8 num_mac_filters;
u8 num_vlan_filters;
u8 num_mc_filters;
- u8 padding[2];
+ u8 num_cids;
+ u8 padding;
} resc;
u32 bulletin_size;
@@ -223,6 +236,16 @@ struct pfvf_start_queue_resp_tlv {
u8 padding[4];
};
+/* Extended queue information - additional index for reference inside qzone.
+ * If commmunicated between VF/PF, each TLV relating to queues should be
+ * extended by one such [or have a future base TLV that already contains info].
+ */
+struct vfpf_qid_tlv {
+ struct channel_tlv tl;
+ u8 qid;
+ u8 padding[3];
+};
+
/* Setup Queue */
struct vfpf_start_rxq_tlv {
struct vfpf_first_tlv first_tlv;
@@ -265,7 +288,15 @@ struct vfpf_stop_rxqs_tlv {
struct vfpf_first_tlv first_tlv;
u16 rx_qid;
+
+ /* While the API supports multiple Rx-queues on a single TLV
+ * message, in practice older VFs always used it as one [ecore].
+ * And there are PFs [starting with the CHANNEL_TLV_QID] which
+ * would start assuming this is always a '1'. So in practice this
+ * field should be considered deprecated and *Always* set to '1'.
+ */
u8 num_rxqs;
+
u8 cqe_completion;
u8 padding[4];
};
@@ -275,6 +306,13 @@ struct vfpf_stop_txqs_tlv {
struct vfpf_first_tlv first_tlv;
u16 tx_qid;
+
+ /* While the API supports multiple Tx-queues on a single TLV
+ * message, in practice older VFs always used it as one [ecore].
+ * And there are PFs [starting with the CHANNEL_TLV_QID] which
+ * would start assuming this is always a '1'. So in practice this
+ * field should be considered deprecated and *Always* set to '1'.
+ */
u8 num_txqs;
u8 padding[5];
};
@@ -605,6 +643,7 @@ enum {
CHANNEL_TLV_VPORT_UPDATE_SGE_TPA,
CHANNEL_TLV_UPDATE_TUNN_PARAM,
CHANNEL_TLV_COALESCE_UPDATE,
+ CHANNEL_TLV_QID,
CHANNEL_TLV_MAX,
/* Required for iterating over vport-update tlvs.