It seems ale(4) controllers do not like to see TCP payload in the

first descriptor in TSO case. Otherwise controller can generate bad
frames during TSO. To address it, make sure to pull up ethernet +
IP + TCP header with options in first buffer. Also ensure the
buffer length of the first descriptor for TSO covers entire ethernet
+ IP + TCP with options and setup additional Tx descriptor if the
first buffer includes TCP payload.

Tested by:	Amar Takhar <verm <> darkbeer dot org >
MFC after:	1 week

1 files changed, 33 insertions, 4 deletions

diff --git a/sys/dev/ale/if_ale.c b/sys/dev/ale/if_ale.c
index b3e7187..ea6b53b 100644
--- a/sys/dev/ale/if_ale.c
+++ b/sys/dev/ale/if_ale.c
@@ -1585,7 +1585,7 @@ ale_encap(struct ale_softc *sc, struct mbuf **m_head)
 	struct tcphdr *tcp;
 	bus_dma_segment_t txsegs[ALE_MAXTXSEGS];
 	bus_dmamap_t map;
-	uint32_t cflags, ip_off, poff, vtag;
+	uint32_t cflags, hdrlen, ip_off, poff, vtag;
 	int error, i, nsegs, prod, si;
 
 	ALE_LOCK_ASSERT(sc);
@@ -1678,6 +1678,11 @@ ale_encap(struct ale_softc *sc, struct mbuf **m_head)
 				return (ENOBUFS);
 			}
 			tcp = (struct tcphdr *)(mtod(m, char *) + poff);
+			m = m_pullup(m, poff + (tcp->th_off << 2));
+			if (m == NULL) {
+				*m_head = NULL;
+				return (ENOBUFS);
+			}
 			/*
 			 * AR81xx requires IP/TCP header size and offset as
 			 * well as TCP pseudo checksum which complicates
@@ -1730,7 +1735,7 @@ ale_encap(struct ale_softc *sc, struct mbuf **m_head)
 	}
 
 	/* Check descriptor overrun. */
-	if (sc->ale_cdata.ale_tx_cnt + nsegs >= ALE_TX_RING_CNT - 2) {
+	if (sc->ale_cdata.ale_tx_cnt + nsegs >= ALE_TX_RING_CNT - 3) {
 		bus_dmamap_unload(sc->ale_cdata.ale_tx_tag, map);
 		return (ENOBUFS);
 	}
@@ -1782,8 +1787,32 @@ ale_encap(struct ale_softc *sc, struct mbuf **m_head)
 		cflags |= ALE_TD_INSERT_VLAN_TAG;
 	}
 
-	desc = NULL;
-	for (i = 0; i < nsegs; i++) {
+	i = 0;
+	if ((m->m_pkthdr.csum_flags & CSUM_TSO) != 0) {
+		/*
+		 * Make sure the first fragment contains
+		 * only ethernet and IP/TCP header with options.
+		 */
+		hdrlen =  poff + (tcp->th_off << 2);
+		desc = &sc->ale_cdata.ale_tx_ring[prod];
+		desc->addr = htole64(txsegs[i].ds_addr);
+		desc->len = htole32(ALE_TX_BYTES(hdrlen) | vtag);
+		desc->flags = htole32(cflags);
+		sc->ale_cdata.ale_tx_cnt++;
+		ALE_DESC_INC(prod, ALE_TX_RING_CNT);
+		if (m->m_len - hdrlen > 0) {
+			/* Handle remaining payload of the first fragment. */
+			desc = &sc->ale_cdata.ale_tx_ring[prod];
+			desc->addr = htole64(txsegs[i].ds_addr + hdrlen);
+			desc->len = htole32(ALE_TX_BYTES(m->m_len - hdrlen) |
+			    vtag);
+			desc->flags = htole32(cflags);
+			sc->ale_cdata.ale_tx_cnt++;
+			ALE_DESC_INC(prod, ALE_TX_RING_CNT);
+		}
+		i = 1;
+	}
+	for (; i < nsegs; i++) {
 		desc = &sc->ale_cdata.ale_tx_ring[prod];
 		desc->addr = htole64(txsegs[i].ds_addr);
 		desc->len = htole32(ALE_TX_BYTES(txsegs[i].ds_len) | vtag);