common/iavf: add DDP package query in virtual channel

[dpdk.git] / drivers / event / octeontx2 / otx2_worker_dual.h

diff --git a/drivers/event/octeontx2/otx2_worker_dual.h b/drivers/event/octeontx2/otx2_worker_dual.h
index d8453d1..c88420e 100644 (file)
--- a/drivers/event/octeontx2/otx2_worker_dual.h
+++ b/drivers/event/octeontx2/otx2_worker_dual.h
@@ -15,27 +15,32 @@
 static __rte_always_inline uint16_t
 otx2_ssogws_dual_get_work(struct otx2_ssogws_state *ws,
                          struct otx2_ssogws_state *ws_pair,
-                         struct rte_event *ev)
+                         struct rte_event *ev, const uint32_t flags,
+                         const void * const lookup_mem,
+                         struct otx2_timesync_info * const tstamp)
 {
        const uint64_t set_gw = BIT_ULL(16) | 1;
        union otx2_sso_event event;
+       uint64_t tstamp_ptr;
        uint64_t get_work1;
+       uint64_t mbuf;
 
+       if (flags & NIX_RX_OFFLOAD_PTYPE_F)
+               rte_prefetch_non_temporal(lookup_mem);
 #ifdef RTE_ARCH_ARM64
        asm volatile(
-                       "        ldr %[tag], [%[tag_loc]]    \n"
-                       "        ldr %[wqp], [%[wqp_loc]]    \n"
-                       "        tbz %[tag], 63, done%=      \n"
-                       "        sevl                        \n"
-                       "rty%=:  wfe                         \n"
+                       "rty%=:                              \n"
                        "        ldr %[tag], [%[tag_loc]]    \n"
                        "        ldr %[wqp], [%[wqp_loc]]    \n"
                        "        tbnz %[tag], 63, rty%=      \n"
                        "done%=: str %[gw], [%[pong]]        \n"
                        "        dmb ld                      \n"
-                       "        prfm pldl1keep, [%[wqp]]    \n"
+                       "        prfm pldl1keep, [%[wqp], #8]\n"
+                       "        sub %[mbuf], %[wqp], #0x80  \n"
+                       "        prfm pldl1keep, [%[mbuf]]   \n"
                        : [tag] "=&r" (event.get_work0),
-                         [wqp] "=&r" (get_work1)
+                         [wqp] "=&r" (get_work1),
+                         [mbuf] "=&r" (mbuf)
                        : [tag_loc] "r" (ws->tag_op),
                          [wqp_loc] "r" (ws->wqp_op),
                          [gw] "r" (set_gw),
@@ -49,6 +54,8 @@ otx2_ssogws_dual_get_work(struct otx2_ssogws_state *ws,
        otx2_write64(set_gw, ws_pair->getwrk_op);
 
        rte_prefetch0((const void *)get_work1);
+       mbuf = (uint64_t)((char *)get_work1 - sizeof(struct rte_mbuf));
+       rte_prefetch0((const void *)mbuf);
 #endif
        event.get_work0 = (event.get_work0 & (0x3ull << 32)) << 6 |
                (event.get_work0 & (0x3FFull << 36)) << 4 |
@@ -56,6 +63,24 @@ otx2_ssogws_dual_get_work(struct otx2_ssogws_state *ws,
        ws->cur_tt = event.sched_type;
        ws->cur_grp = event.queue_id;
 
+       if (event.sched_type != SSO_TT_EMPTY &&
+           event.event_type == RTE_EVENT_TYPE_ETHDEV) {
+               otx2_wqe_to_mbuf(get_work1, mbuf, event.sub_event_type,
+                                (uint32_t) event.get_work0, flags, lookup_mem);
+               /* Extracting tstamp, if PTP enabled. CGX will prepend the
+                * timestamp at starting of packet data and it can be derieved
+                * from WQE 9 dword which corresponds to SG iova.
+                * rte_pktmbuf_mtod_offset can be used for this purpose but it
+                * brings down the performance as it reads mbuf->buf_addr which
+                * is not part of cache in general fast path.
+                */
+               tstamp_ptr = *(uint64_t *)(((struct nix_wqe_hdr_s *)get_work1)
+                                            + OTX2_SSO_WQE_SG_PTR);
+               otx2_nix_mbuf_to_tstamp((struct rte_mbuf *)mbuf, tstamp, flags,
+                                       (uint64_t *)tstamp_ptr);
+               get_work1 = mbuf;
+       }
+
        ev->event = event.get_work0;
        ev->u64 = get_work1;