qed: Add VF->PF channel infrastructure
Communication between VF and PF is based on a dedicated HW channel;
VF will prepare a messge, and by signaling the HW the PF would get a
notification of that message existance. The PF would then copy the
message, process it and DMA an answer back to the VF as a response.
The messages themselves are TLV-based - allowing easier backward/forward
compatibility.
This patch adds the infrastructure of the channel on the PF side -
starting with the arrival of the notification and ending with DMAing
the response back to the VF.
It also adds a dummy-response as reference, as it only lays the
groundwork of the communication; it doesn't really add support of any
actual messages.
Signed-off-by: Yuval Mintz <Yuval.Mintz@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
diff --git a/drivers/net/ethernet/qlogic/qed/qed_sriov.c b/drivers/net/ethernet/qlogic/qed/qed_sriov.c
index 685e3fa..4a6af42 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_sriov.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_sriov.c
@@ -31,6 +31,26 @@
return true;
}
+static struct qed_vf_info *qed_iov_get_vf_info(struct qed_hwfn *p_hwfn,
+ u16 relative_vf_id,
+ bool b_enabled_only)
+{
+ struct qed_vf_info *vf = NULL;
+
+ if (!p_hwfn->pf_iov_info) {
+ DP_NOTICE(p_hwfn->cdev, "No iov info\n");
+ return NULL;
+ }
+
+ if (qed_iov_is_valid_vfid(p_hwfn, relative_vf_id, b_enabled_only))
+ vf = &p_hwfn->pf_iov_info->vfs_array[relative_vf_id];
+ else
+ DP_ERR(p_hwfn, "qed_iov_get_vf_info: VF[%d] is not enabled\n",
+ relative_vf_id);
+
+ return vf;
+}
+
static int qed_iov_pci_cfg_info(struct qed_dev *cdev)
{
struct qed_hw_sriov_info *iov = cdev->p_iov_info;
@@ -349,6 +369,232 @@
return 0;
}
+static bool qed_iov_pf_sanity_check(struct qed_hwfn *p_hwfn, int vfid)
+{
+ /* Check PF supports sriov */
+ if (!IS_QED_SRIOV(p_hwfn->cdev) || !IS_PF_SRIOV_ALLOC(p_hwfn))
+ return false;
+
+ /* Check VF validity */
+ if (!qed_iov_is_valid_vfid(p_hwfn, vfid, true))
+ return false;
+
+ return true;
+}
+
+static bool qed_iov_tlv_supported(u16 tlvtype)
+{
+ return CHANNEL_TLV_NONE < tlvtype && tlvtype < CHANNEL_TLV_MAX;
+}
+
+/* place a given tlv on the tlv buffer, continuing current tlv list */
+void *qed_add_tlv(struct qed_hwfn *p_hwfn, u8 **offset, u16 type, u16 length)
+{
+ struct channel_tlv *tl = (struct channel_tlv *)*offset;
+
+ tl->type = type;
+ tl->length = length;
+
+ /* Offset should keep pointing to next TLV (the end of the last) */
+ *offset += length;
+
+ /* Return a pointer to the start of the added tlv */
+ return *offset - length;
+}
+
+/* list the types and lengths of the tlvs on the buffer */
+void qed_dp_tlv_list(struct qed_hwfn *p_hwfn, void *tlvs_list)
+{
+ u16 i = 1, total_length = 0;
+ struct channel_tlv *tlv;
+
+ do {
+ tlv = (struct channel_tlv *)((u8 *)tlvs_list + total_length);
+
+ /* output tlv */
+ DP_VERBOSE(p_hwfn, QED_MSG_IOV,
+ "TLV number %d: type %d, length %d\n",
+ i, tlv->type, tlv->length);
+
+ if (tlv->type == CHANNEL_TLV_LIST_END)
+ return;
+
+ /* Validate entry - protect against malicious VFs */
+ if (!tlv->length) {
+ DP_NOTICE(p_hwfn, "TLV of length 0 found\n");
+ return;
+ }
+
+ total_length += tlv->length;
+
+ if (total_length >= sizeof(struct tlv_buffer_size)) {
+ DP_NOTICE(p_hwfn, "TLV ==> Buffer overflow\n");
+ return;
+ }
+
+ i++;
+ } while (1);
+}
+
+static void qed_iov_send_response(struct qed_hwfn *p_hwfn,
+ struct qed_ptt *p_ptt,
+ struct qed_vf_info *p_vf,
+ u16 length, u8 status)
+{
+ struct qed_iov_vf_mbx *mbx = &p_vf->vf_mbx;
+ struct qed_dmae_params params;
+ u8 eng_vf_id;
+
+ mbx->reply_virt->default_resp.hdr.status = status;
+
+ qed_dp_tlv_list(p_hwfn, mbx->reply_virt);
+
+ eng_vf_id = p_vf->abs_vf_id;
+
+ memset(¶ms, 0, sizeof(struct qed_dmae_params));
+ params.flags = QED_DMAE_FLAG_VF_DST;
+ params.dst_vfid = eng_vf_id;
+
+ qed_dmae_host2host(p_hwfn, p_ptt, mbx->reply_phys + sizeof(u64),
+ mbx->req_virt->first_tlv.reply_address +
+ sizeof(u64),
+ (sizeof(union pfvf_tlvs) - sizeof(u64)) / 4,
+ ¶ms);
+
+ qed_dmae_host2host(p_hwfn, p_ptt, mbx->reply_phys,
+ mbx->req_virt->first_tlv.reply_address,
+ sizeof(u64) / 4, ¶ms);
+
+ REG_WR(p_hwfn,
+ GTT_BAR0_MAP_REG_USDM_RAM +
+ USTORM_VF_PF_CHANNEL_READY_OFFSET(eng_vf_id), 1);
+}
+
+static void qed_iov_prepare_resp(struct qed_hwfn *p_hwfn,
+ struct qed_ptt *p_ptt,
+ struct qed_vf_info *vf_info,
+ u16 type, u16 length, u8 status)
+{
+ struct qed_iov_vf_mbx *mbx = &vf_info->vf_mbx;
+
+ mbx->offset = (u8 *)mbx->reply_virt;
+
+ qed_add_tlv(p_hwfn, &mbx->offset, type, length);
+ qed_add_tlv(p_hwfn, &mbx->offset, CHANNEL_TLV_LIST_END,
+ sizeof(struct channel_list_end_tlv));
+
+ qed_iov_send_response(p_hwfn, p_ptt, vf_info, length, status);
+}
+
+static void qed_iov_process_mbx_dummy_resp(struct qed_hwfn *p_hwfn,
+ struct qed_ptt *p_ptt,
+ struct qed_vf_info *p_vf)
+{
+ qed_iov_prepare_resp(p_hwfn, p_ptt, p_vf, CHANNEL_TLV_NONE,
+ sizeof(struct pfvf_def_resp_tlv),
+ PFVF_STATUS_SUCCESS);
+}
+
+static void qed_iov_process_mbx_req(struct qed_hwfn *p_hwfn,
+ struct qed_ptt *p_ptt, int vfid)
+{
+ struct qed_iov_vf_mbx *mbx;
+ struct qed_vf_info *p_vf;
+ int i;
+
+ p_vf = qed_iov_get_vf_info(p_hwfn, (u16) vfid, true);
+ if (!p_vf)
+ return;
+
+ mbx = &p_vf->vf_mbx;
+
+ /* qed_iov_process_mbx_request */
+ DP_VERBOSE(p_hwfn,
+ QED_MSG_IOV,
+ "qed_iov_process_mbx_req vfid %d\n", p_vf->abs_vf_id);
+
+ mbx->first_tlv = mbx->req_virt->first_tlv;
+
+ /* check if tlv type is known */
+ if (qed_iov_tlv_supported(mbx->first_tlv.tl.type)) {
+ qed_iov_process_mbx_dummy_resp(p_hwfn, p_ptt, p_vf);
+ } else {
+ /* unknown TLV - this may belong to a VF driver from the future
+ * - a version written after this PF driver was written, which
+ * supports features unknown as of yet. Too bad since we don't
+ * support them. Or this may be because someone wrote a crappy
+ * VF driver and is sending garbage over the channel.
+ */
+ DP_ERR(p_hwfn,
+ "unknown TLV. type %d length %d. first 20 bytes of mailbox buffer:\n",
+ mbx->first_tlv.tl.type, mbx->first_tlv.tl.length);
+
+ for (i = 0; i < 20; i++) {
+ DP_VERBOSE(p_hwfn,
+ QED_MSG_IOV,
+ "%x ",
+ mbx->req_virt->tlv_buf_size.tlv_buffer[i]);
+ }
+ }
+}
+
+void qed_iov_pf_add_pending_events(struct qed_hwfn *p_hwfn, u8 vfid)
+{
+ u64 add_bit = 1ULL << (vfid % 64);
+
+ p_hwfn->pf_iov_info->pending_events[vfid / 64] |= add_bit;
+}
+
+static void qed_iov_pf_get_and_clear_pending_events(struct qed_hwfn *p_hwfn,
+ u64 *events)
+{
+ u64 *p_pending_events = p_hwfn->pf_iov_info->pending_events;
+
+ memcpy(events, p_pending_events, sizeof(u64) * QED_VF_ARRAY_LENGTH);
+ memset(p_pending_events, 0, sizeof(u64) * QED_VF_ARRAY_LENGTH);
+}
+
+static int qed_sriov_vfpf_msg(struct qed_hwfn *p_hwfn,
+ u16 abs_vfid, struct regpair *vf_msg)
+{
+ u8 min = (u8)p_hwfn->cdev->p_iov_info->first_vf_in_pf;
+ struct qed_vf_info *p_vf;
+
+ if (!qed_iov_pf_sanity_check(p_hwfn, (int)abs_vfid - min)) {
+ DP_VERBOSE(p_hwfn,
+ QED_MSG_IOV,
+ "Got a message from VF [abs 0x%08x] that cannot be handled by PF\n",
+ abs_vfid);
+ return 0;
+ }
+ p_vf = &p_hwfn->pf_iov_info->vfs_array[(u8)abs_vfid - min];
+
+ /* List the physical address of the request so that handler
+ * could later on copy the message from it.
+ */
+ p_vf->vf_mbx.pending_req = (((u64)vf_msg->hi) << 32) | vf_msg->lo;
+
+ /* Mark the event and schedule the workqueue */
+ qed_iov_pf_add_pending_events(p_hwfn, p_vf->relative_vf_id);
+ qed_schedule_iov(p_hwfn, QED_IOV_WQ_MSG_FLAG);
+
+ return 0;
+}
+
+int qed_sriov_eqe_event(struct qed_hwfn *p_hwfn,
+ u8 opcode, __le16 echo, union event_ring_data *data)
+{
+ switch (opcode) {
+ case COMMON_EVENT_VF_PF_CHANNEL:
+ return qed_sriov_vfpf_msg(p_hwfn, le16_to_cpu(echo),
+ &data->vf_pf_channel.msg_addr);
+ default:
+ DP_INFO(p_hwfn->cdev, "Unknown sriov eqe event 0x%02x\n",
+ opcode);
+ return -EINVAL;
+ }
+}
+
u16 qed_iov_get_next_active_vf(struct qed_hwfn *p_hwfn, u16 rel_vf_id)
{
struct qed_hw_sriov_info *p_iov = p_hwfn->cdev->p_iov_info;
@@ -364,3 +610,142 @@
out:
return MAX_NUM_VFS;
}
+
+static int qed_iov_copy_vf_msg(struct qed_hwfn *p_hwfn, struct qed_ptt *ptt,
+ int vfid)
+{
+ struct qed_dmae_params params;
+ struct qed_vf_info *vf_info;
+
+ vf_info = qed_iov_get_vf_info(p_hwfn, (u16) vfid, true);
+ if (!vf_info)
+ return -EINVAL;
+
+ memset(¶ms, 0, sizeof(struct qed_dmae_params));
+ params.flags = QED_DMAE_FLAG_VF_SRC | QED_DMAE_FLAG_COMPLETION_DST;
+ params.src_vfid = vf_info->abs_vf_id;
+
+ if (qed_dmae_host2host(p_hwfn, ptt,
+ vf_info->vf_mbx.pending_req,
+ vf_info->vf_mbx.req_phys,
+ sizeof(union vfpf_tlvs) / 4, ¶ms)) {
+ DP_VERBOSE(p_hwfn, QED_MSG_IOV,
+ "Failed to copy message from VF 0x%02x\n", vfid);
+
+ return -EIO;
+ }
+
+ return 0;
+}
+
+/**
+ * qed_schedule_iov - schedules IOV task for VF and PF
+ * @hwfn: hardware function pointer
+ * @flag: IOV flag for VF/PF
+ */
+void qed_schedule_iov(struct qed_hwfn *hwfn, enum qed_iov_wq_flag flag)
+{
+ smp_mb__before_atomic();
+ set_bit(flag, &hwfn->iov_task_flags);
+ smp_mb__after_atomic();
+ DP_VERBOSE(hwfn, QED_MSG_IOV, "Scheduling iov task [Flag: %d]\n", flag);
+ queue_delayed_work(hwfn->iov_wq, &hwfn->iov_task, 0);
+}
+
+static void qed_handle_vf_msg(struct qed_hwfn *hwfn)
+{
+ u64 events[QED_VF_ARRAY_LENGTH];
+ struct qed_ptt *ptt;
+ int i;
+
+ ptt = qed_ptt_acquire(hwfn);
+ if (!ptt) {
+ DP_VERBOSE(hwfn, QED_MSG_IOV,
+ "Can't acquire PTT; re-scheduling\n");
+ qed_schedule_iov(hwfn, QED_IOV_WQ_MSG_FLAG);
+ return;
+ }
+
+ qed_iov_pf_get_and_clear_pending_events(hwfn, events);
+
+ DP_VERBOSE(hwfn, QED_MSG_IOV,
+ "Event mask of VF events: 0x%llx 0x%llx 0x%llx\n",
+ events[0], events[1], events[2]);
+
+ qed_for_each_vf(hwfn, i) {
+ /* Skip VFs with no pending messages */
+ if (!(events[i / 64] & (1ULL << (i % 64))))
+ continue;
+
+ DP_VERBOSE(hwfn, QED_MSG_IOV,
+ "Handling VF message from VF 0x%02x [Abs 0x%02x]\n",
+ i, hwfn->cdev->p_iov_info->first_vf_in_pf + i);
+
+ /* Copy VF's message to PF's request buffer for that VF */
+ if (qed_iov_copy_vf_msg(hwfn, ptt, i))
+ continue;
+
+ qed_iov_process_mbx_req(hwfn, ptt, i);
+ }
+
+ qed_ptt_release(hwfn, ptt);
+}
+
+void qed_iov_pf_task(struct work_struct *work)
+{
+ struct qed_hwfn *hwfn = container_of(work, struct qed_hwfn,
+ iov_task.work);
+
+ if (test_and_clear_bit(QED_IOV_WQ_STOP_WQ_FLAG, &hwfn->iov_task_flags))
+ return;
+
+ if (test_and_clear_bit(QED_IOV_WQ_MSG_FLAG, &hwfn->iov_task_flags))
+ qed_handle_vf_msg(hwfn);
+}
+
+void qed_iov_wq_stop(struct qed_dev *cdev, bool schedule_first)
+{
+ int i;
+
+ for_each_hwfn(cdev, i) {
+ if (!cdev->hwfns[i].iov_wq)
+ continue;
+
+ if (schedule_first) {
+ qed_schedule_iov(&cdev->hwfns[i],
+ QED_IOV_WQ_STOP_WQ_FLAG);
+ cancel_delayed_work_sync(&cdev->hwfns[i].iov_task);
+ }
+
+ flush_workqueue(cdev->hwfns[i].iov_wq);
+ destroy_workqueue(cdev->hwfns[i].iov_wq);
+ }
+}
+
+int qed_iov_wq_start(struct qed_dev *cdev)
+{
+ char name[NAME_SIZE];
+ int i;
+
+ for_each_hwfn(cdev, i) {
+ struct qed_hwfn *p_hwfn = &cdev->hwfns[i];
+
+ /* PFs needs a dedicated workqueue only if they support IOV. */
+ if (!IS_PF_SRIOV(p_hwfn))
+ continue;
+
+ snprintf(name, NAME_SIZE, "iov-%02x:%02x.%02x",
+ cdev->pdev->bus->number,
+ PCI_SLOT(cdev->pdev->devfn), p_hwfn->abs_pf_id);
+
+ p_hwfn->iov_wq = create_singlethread_workqueue(name);
+ if (!p_hwfn->iov_wq) {
+ DP_NOTICE(p_hwfn, "Cannot create iov workqueue\n");
+ return -ENOMEM;
+ }
+
+ INIT_DELAYED_WORK(&p_hwfn->iov_task, qed_iov_pf_task);
+ }
+
+ return 0;
+}