From: dlg <dlg@openbsd.org>
Date: Sun, 11 Aug 2024 01:02:10 +0000 (+0000)
Subject: unbreak igc on strict alignment archs.
X-Git-Url: http://artulab.com/gitweb/?a=commitdiff_plain;h=4cceae0f74744692e43b7051f3b9d8c8520339bc;p=openbsd

unbreak igc on strict alignment archs.

igc (like other intel nics) only supports a limited range of rx
buffer sizes, one of which is 2k like our standard mbuf clusters.
however they put the ethernet packet at the start of the provided
rx buffer, which means the ip header is misaligned on strict alignment
archs if we use the standard 2k mbuf clusters. to avoid wasting too
much memory using the next biggest cluster (4k) we have mcl2k2
clusters that pretty much exists for intel nics to use.

igc (and probably some other drivers ive touched) was too clever
about using mcl2k2 though, which caused panics when trying to use
jumbo packets. jan@ fixed this by moving back to 2k clusters, but
that broke sparc64 (and probably powerpc64).

this dumbs the mcl2k2 handling down a bit so jumbos still work, but
adds ETHER_ALIGN back so strict alignment archs also work again.

ok bluhm@
---

diff --git a/sys/dev/pci/if_igc.c b/sys/dev/pci/if_igc.c
index 3d3df852585..6f9aaed06ed 100644
--- a/sys/dev/pci/if_igc.c
+++ b/sys/dev/pci/if_igc.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: if_igc.c,v 1.26 2024/08/08 14:58:49 jan Exp $	*/
+/*	$OpenBSD: if_igc.c,v 1.27 2024/08/11 01:02:10 dlg Exp $	*/
 /*-
  * SPDX-License-Identifier: BSD-2-Clause
  *
@@ -202,6 +202,7 @@ igc_attach(struct device *parent, struct device *self, void *aux)
 	/* Determine hardware and mac info */
 	igc_identify_hardware(sc);
 
+	sc->rx_mbuf_sz = MCLBYTES;
 	sc->num_tx_desc = IGC_DEFAULT_TXD;
 	sc->num_rx_desc = IGC_DEFAULT_RXD;
 
@@ -881,7 +882,6 @@ igc_init(void *arg)
 	}
 	igc_initialize_transmit_unit(sc);
 
-	sc->rx_mbuf_sz = MCLBYTES;
 	/* Prepare receive descriptors and buffers. */
 	if (igc_setup_receive_structures(sc)) {
 		printf("%s: Could not setup receive structures\n",
@@ -1232,7 +1232,7 @@ igc_rxrinfo(struct igc_softc *sc, struct if_rxrinfo *ifri)
 
 	for (i = 0; i < sc->sc_nqueues; i++) {
 		rxr = &sc->rx_rings[i];
-		ifr[n].ifr_size = MCLBYTES;
+		ifr[n].ifr_size = sc->rx_mbuf_sz;
 		snprintf(ifr[n].ifr_name, sizeof(ifr[n].ifr_name), "%d", i);
 		ifr[n].ifr_info = rxr->rx_ring;
 		n++;
@@ -1673,11 +1673,11 @@ igc_get_buf(struct igc_rxring *rxr, int i)
 		return ENOBUFS;
 	}
 
-	m = MCLGETL(NULL, M_DONTWAIT, sc->rx_mbuf_sz);
+	m = MCLGETL(NULL, M_DONTWAIT, sc->rx_mbuf_sz + ETHER_ALIGN);
 	if (!m)
 		return ENOBUFS;
 
-	m->m_data += (m->m_ext.ext_size - sc->rx_mbuf_sz);
+	m->m_data += ETHER_ALIGN;
 	m->m_len = m->m_pkthdr.len = sc->rx_mbuf_sz;
 
 	error = bus_dmamap_load_mbuf(rxr->rxdma.dma_tag, rxbuf->map, m,
@@ -2159,7 +2159,7 @@ igc_allocate_receive_buffers(struct igc_rxring *rxr)
 	rxbuf = rxr->rx_buffers;
 	for (i = 0; i < sc->num_rx_desc; i++, rxbuf++) {
 		error = bus_dmamap_create(rxr->rxdma.dma_tag,
-		    MAX_JUMBO_FRAME_SIZE, IGC_MAX_SCATTER, MCLBYTES, 0,
+		    sc->rx_mbuf_sz, 1, sc->rx_mbuf_sz, 0,
 		    BUS_DMA_NOWAIT, &rxbuf->map);
 		if (error) {
 			printf("%s: Unable to create RX DMA map\n",
@@ -2223,7 +2223,8 @@ igc_setup_receive_ring(struct igc_rxring *rxr)
 	rxr->next_to_check = 0;
 	rxr->last_desc_filled = sc->num_rx_desc - 1;
 
-	if_rxr_init(&rxr->rx_ring, 2 * ((ifp->if_hardmtu / MCLBYTES) + 1),
+	if_rxr_init(&rxr->rx_ring,
+	    2 * howmany(ifp->if_hardmtu, sc->rx_mbuf_sz) + 1,
 	    sc->num_rx_desc - 1);
 
 	return 0;