freebsd-dev/sys/dev/e1000/igb_txrx.c
Sean Bruno 95246abb21 IFLIB updates
- unconditionally enable BUS_DMA on non-x86 architectures
- speed up rxd zeroing via customized function
- support out of order updates to rxd's
- add prefetching to hardware descriptor rings
- only prefetch on 10G or faster hardware
- add seperate tx queue intr function
- preliminary rework of NETMAP interfaces, WIP

Submitted by:	Matt Macy <mmacy@nextbsd.org>
Sponsored by:	Limelight Networks
2017-03-13 22:53:06 +00:00

585 lines
16 KiB
C

/*-
* Copyright (c) 2016 Matt Macy <mmacy@nextbsd.org>
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*/
/* $FreeBSD$ */
#include "if_em.h"
#ifdef RSS
#include <net/rss_config.h>
#include <netinet/in_rss.h>
#endif
#ifdef VERBOSE_DEBUG
#define DPRINTF device_printf
#else
#define DPRINTF(...)
#endif
/*********************************************************************
* Local Function prototypes
*********************************************************************/
static int igb_isc_txd_encap(void *arg, if_pkt_info_t pi);
static void igb_isc_txd_flush(void *arg, uint16_t txqid, qidx_t pidx);
static int igb_isc_txd_credits_update(void *arg, uint16_t txqid, bool clear);
static void igb_isc_rxd_refill(void *arg, if_rxd_update_t iru);
static void igb_isc_rxd_flush(void *arg, uint16_t rxqid, uint8_t flid __unused, qidx_t pidx);
static int igb_isc_rxd_available(void *arg, uint16_t rxqid, qidx_t idx, qidx_t budget);
static int igb_isc_rxd_pkt_get(void *arg, if_rxd_info_t ri);
static int igb_tx_ctx_setup(struct tx_ring *txr, if_pkt_info_t pi, u32 *cmd_type_len, u32 *olinfo_status);
static int igb_tso_setup(struct tx_ring *txr, if_pkt_info_t pi, u32 *cmd_type_len, u32 *olinfo_status);
static void igb_rx_checksum(u32 staterr, if_rxd_info_t ri, u32 ptype);
static int igb_determine_rsstype(u16 pkt_info);
extern void igb_if_enable_intr(if_ctx_t ctx);
extern int em_intr(void *arg);
struct if_txrx igb_txrx = {
igb_isc_txd_encap,
igb_isc_txd_flush,
igb_isc_txd_credits_update,
igb_isc_rxd_available,
igb_isc_rxd_pkt_get,
igb_isc_rxd_refill,
igb_isc_rxd_flush,
em_intr
};
extern if_shared_ctx_t em_sctx;
/**********************************************************************
*
* Setup work for hardware segmentation offload (TSO) on
* adapters using advanced tx descriptors
*
**********************************************************************/
static int
igb_tso_setup(struct tx_ring *txr, if_pkt_info_t pi, u32 *cmd_type_len, u32 *olinfo_status)
{
struct e1000_adv_tx_context_desc *TXD;
struct adapter *adapter = txr->adapter;
u32 type_tucmd_mlhl = 0, vlan_macip_lens = 0;
u32 mss_l4len_idx = 0;
u32 paylen;
switch(pi->ipi_etype) {
case ETHERTYPE_IPV6:
type_tucmd_mlhl |= E1000_ADVTXD_TUCMD_IPV6;
break;
case ETHERTYPE_IP:
type_tucmd_mlhl |= E1000_ADVTXD_TUCMD_IPV4;
/* Tell transmit desc to also do IPv4 checksum. */
*olinfo_status |= E1000_TXD_POPTS_IXSM << 8;
break;
default:
panic("%s: CSUM_TSO but no supported IP version (0x%04x)",
__func__, ntohs(pi->ipi_etype));
break;
}
TXD = (struct e1000_adv_tx_context_desc *) &txr->tx_base[pi->ipi_pidx];
/* This is used in the transmit desc in encap */
paylen = pi->ipi_len - pi->ipi_ehdrlen - pi->ipi_ip_hlen - pi->ipi_tcp_hlen;
/* VLAN MACLEN IPLEN */
if (pi->ipi_mflags & M_VLANTAG) {
vlan_macip_lens |= (pi->ipi_vtag << E1000_ADVTXD_VLAN_SHIFT);
}
vlan_macip_lens |= pi->ipi_ehdrlen << E1000_ADVTXD_MACLEN_SHIFT;
vlan_macip_lens |= pi->ipi_ip_hlen;
TXD->vlan_macip_lens = htole32(vlan_macip_lens);
/* ADV DTYPE TUCMD */
type_tucmd_mlhl |= E1000_ADVTXD_DCMD_DEXT | E1000_ADVTXD_DTYP_CTXT;
type_tucmd_mlhl |= E1000_ADVTXD_TUCMD_L4T_TCP;
TXD->type_tucmd_mlhl = htole32(type_tucmd_mlhl);
/* MSS L4LEN IDX */
mss_l4len_idx |= (pi->ipi_tso_segsz << E1000_ADVTXD_MSS_SHIFT);
mss_l4len_idx |= (pi->ipi_tcp_hlen << E1000_ADVTXD_L4LEN_SHIFT);
/* 82575 needs the queue index added */
if (adapter->hw.mac.type == e1000_82575)
mss_l4len_idx |= txr->me << 4;
TXD->mss_l4len_idx = htole32(mss_l4len_idx);
TXD->seqnum_seed = htole32(0);
*cmd_type_len |= E1000_ADVTXD_DCMD_TSE;
*olinfo_status |= E1000_TXD_POPTS_TXSM << 8;
*olinfo_status |= paylen << E1000_ADVTXD_PAYLEN_SHIFT;
return (1);
}
/*********************************************************************
*
* Advanced Context Descriptor setup for VLAN, CSUM or TSO
*
**********************************************************************/
static int
igb_tx_ctx_setup(struct tx_ring *txr, if_pkt_info_t pi, u32 *cmd_type_len, u32 *olinfo_status)
{
struct e1000_adv_tx_context_desc *TXD;
struct adapter *adapter = txr->adapter;
u32 vlan_macip_lens, type_tucmd_mlhl;
u32 mss_l4len_idx;
mss_l4len_idx = vlan_macip_lens = type_tucmd_mlhl = 0;
int offload = TRUE;
/* First check if TSO is to be used */
if (pi->ipi_csum_flags & CSUM_TSO)
return (igb_tso_setup(txr, pi, cmd_type_len, olinfo_status));
/* Indicate the whole packet as payload when not doing TSO */
*olinfo_status |= pi->ipi_len << E1000_ADVTXD_PAYLEN_SHIFT;
/* Now ready a context descriptor */
TXD = (struct e1000_adv_tx_context_desc *) &txr->tx_base[pi->ipi_pidx];
/*
** In advanced descriptors the vlan tag must
** be placed into the context descriptor. Hence
** we need to make one even if not doing offloads.
*/
if (pi->ipi_mflags & M_VLANTAG) {
vlan_macip_lens |= (pi->ipi_vtag << E1000_ADVTXD_VLAN_SHIFT);
} else if ((pi->ipi_csum_flags & IGB_CSUM_OFFLOAD) == 0) {
return (0);
}
/* Set the ether header length */
vlan_macip_lens |= pi->ipi_ehdrlen << E1000_ADVTXD_MACLEN_SHIFT;
switch(pi->ipi_etype) {
case ETHERTYPE_IP:
type_tucmd_mlhl |= E1000_ADVTXD_TUCMD_IPV4;
break;
case ETHERTYPE_IPV6:
type_tucmd_mlhl |= E1000_ADVTXD_TUCMD_IPV6;
break;
default:
offload = FALSE;
break;
}
vlan_macip_lens |= pi->ipi_ip_hlen;
type_tucmd_mlhl |= E1000_ADVTXD_DCMD_DEXT | E1000_ADVTXD_DTYP_CTXT;
switch (pi->ipi_ipproto) {
case IPPROTO_TCP:
if (pi->ipi_csum_flags & (CSUM_IP_TCP | CSUM_IP6_TCP))
type_tucmd_mlhl |= E1000_ADVTXD_TUCMD_L4T_TCP;
break;
case IPPROTO_UDP:
if (pi->ipi_csum_flags & (CSUM_IP_UDP | CSUM_IP6_UDP))
type_tucmd_mlhl |= E1000_ADVTXD_TUCMD_L4T_UDP;
break;
case IPPROTO_SCTP:
if (pi->ipi_csum_flags & (CSUM_IP_SCTP | CSUM_IP6_SCTP))
type_tucmd_mlhl |= E1000_ADVTXD_TUCMD_L4T_SCTP;
break;
default:
offload = FALSE;
break;
}
if (offload) /* For the TX descriptor setup */
*olinfo_status |= E1000_TXD_POPTS_TXSM << 8;
/* 82575 needs the queue index added */
if (adapter->hw.mac.type == e1000_82575)
mss_l4len_idx = txr->me << 4;
/* Now copy bits into descriptor */
TXD->vlan_macip_lens = htole32(vlan_macip_lens);
TXD->type_tucmd_mlhl = htole32(type_tucmd_mlhl);
TXD->seqnum_seed = htole32(0);
TXD->mss_l4len_idx = htole32(mss_l4len_idx);
return (1);
}
static int
igb_isc_txd_encap(void *arg, if_pkt_info_t pi)
{
struct adapter *sc = arg;
if_softc_ctx_t scctx = sc->shared;
struct em_tx_queue *que = &sc->tx_queues[pi->ipi_qsidx];
struct tx_ring *txr = &que->txr;
int nsegs = pi->ipi_nsegs;
bus_dma_segment_t *segs = pi->ipi_segs;
union e1000_adv_tx_desc *txd = NULL;
int i, j, first, pidx_last;
u32 olinfo_status, cmd_type_len, txd_flags;
qidx_t ntxd;
pidx_last = olinfo_status = 0;
/* Basic descriptor defines */
cmd_type_len = (E1000_ADVTXD_DTYP_DATA |
E1000_ADVTXD_DCMD_IFCS | E1000_ADVTXD_DCMD_DEXT);
if (pi->ipi_mflags & M_VLANTAG)
cmd_type_len |= E1000_ADVTXD_DCMD_VLE;
first = i = pi->ipi_pidx;
ntxd = scctx->isc_ntxd[0];
txd_flags = pi->ipi_flags & IPI_TX_INTR ? E1000_ADVTXD_DCMD_RS : 0;
/* Consume the first descriptor */
i += igb_tx_ctx_setup(txr, pi, &cmd_type_len, &olinfo_status);
if (i == scctx->isc_ntxd[0])
i = 0;
/* 82575 needs the queue index added */
if (sc->hw.mac.type == e1000_82575)
olinfo_status |= txr->me << 4;
for (j = 0; j < nsegs; j++) {
bus_size_t seglen;
bus_addr_t segaddr;
txd = (union e1000_adv_tx_desc *)&txr->tx_base[i];
seglen = segs[j].ds_len;
segaddr = htole64(segs[j].ds_addr);
txd->read.buffer_addr = segaddr;
txd->read.cmd_type_len = htole32(E1000_TXD_CMD_IFCS |
cmd_type_len | seglen);
txd->read.olinfo_status = htole32(olinfo_status);
pidx_last = i;
if (++i == scctx->isc_ntxd[0]) {
i = 0;
}
}
if (txd_flags) {
txr->tx_rsq[txr->tx_rs_pidx] = pidx_last;
txr->tx_rs_pidx = (txr->tx_rs_pidx+1) & (ntxd-1);
MPASS(txr->tx_rs_pidx != txr->tx_rs_cidx);
}
txd->read.cmd_type_len |= htole32(E1000_TXD_CMD_EOP | txd_flags);
pi->ipi_new_pidx = i;
return (0);
}
static void
igb_isc_txd_flush(void *arg, uint16_t txqid, qidx_t pidx)
{
struct adapter *adapter = arg;
struct em_tx_queue *que = &adapter->tx_queues[txqid];
struct tx_ring *txr = &que->txr;
E1000_WRITE_REG(&adapter->hw, E1000_TDT(txr->me), pidx);
}
static int
igb_isc_txd_credits_update(void *arg, uint16_t txqid, bool clear)
{
struct adapter *adapter = arg;
if_softc_ctx_t scctx = adapter->shared;
struct em_tx_queue *que = &adapter->tx_queues[txqid];
struct tx_ring *txr = &que->txr;
qidx_t processed = 0;
int updated;
qidx_t cur, prev, ntxd, rs_cidx;
int32_t delta;
uint8_t status;
rs_cidx = txr->tx_rs_cidx;
if (rs_cidx == txr->tx_rs_pidx)
return (0);
cur = txr->tx_rsq[rs_cidx];
status = ((union e1000_adv_tx_desc *)&txr->tx_base[cur])->wb.status;
updated = !!(status & E1000_TXD_STAT_DD);
if (!clear || !updated)
return (updated);
prev = txr->tx_cidx_processed;
ntxd = scctx->isc_ntxd[0];
do {
delta = (int32_t)cur - (int32_t)prev;
MPASS(prev == 0 || delta != 0);
if (delta < 0)
delta += ntxd;
processed += delta;
prev = cur;
rs_cidx = (rs_cidx + 1) & (ntxd-1);
if (rs_cidx == txr->tx_rs_pidx)
break;
cur = txr->tx_rsq[rs_cidx];
status = ((union e1000_adv_tx_desc *)&txr->tx_base[cur])->wb.status;
} while ((status & E1000_TXD_STAT_DD));
txr->tx_rs_cidx = rs_cidx;
txr->tx_cidx_processed = prev;
return (processed);
}
static void
igb_isc_rxd_refill(void *arg, if_rxd_update_t iru)
{
struct adapter *sc = arg;
if_softc_ctx_t scctx = sc->shared;
uint16_t rxqid = iru->iru_qsidx;
struct em_rx_queue *que = &sc->rx_queues[rxqid];
union e1000_adv_rx_desc *rxd;
struct rx_ring *rxr = &que->rxr;
uint64_t *paddrs;
uint32_t next_pidx, pidx;
uint16_t count;
int i;
paddrs = iru->iru_paddrs;
pidx = iru->iru_pidx;
count = iru->iru_count;
for (i = 0, next_pidx = pidx; i < count; i++) {
rxd = (union e1000_adv_rx_desc *)&rxr->rx_base[next_pidx];
rxd->read.pkt_addr = htole64(paddrs[i]);
if (++next_pidx == scctx->isc_nrxd[0])
next_pidx = 0;
}
}
static void
igb_isc_rxd_flush(void *arg, uint16_t rxqid, uint8_t flid __unused, qidx_t pidx)
{
struct adapter *sc = arg;
struct em_rx_queue *que = &sc->rx_queues[rxqid];
struct rx_ring *rxr = &que->rxr;
E1000_WRITE_REG(&sc->hw, E1000_RDT(rxr->me), pidx);
}
static int
igb_isc_rxd_available(void *arg, uint16_t rxqid, qidx_t idx, qidx_t budget)
{
struct adapter *sc = arg;
if_softc_ctx_t scctx = sc->shared;
struct em_rx_queue *que = &sc->rx_queues[rxqid];
struct rx_ring *rxr = &que->rxr;
union e1000_adv_rx_desc *rxd;
u32 staterr = 0;
int cnt, i, iter;
if (budget == 1) {
rxd = (union e1000_adv_rx_desc *)&rxr->rx_base[idx];
staterr = le32toh(rxd->wb.upper.status_error);
return (staterr & E1000_RXD_STAT_DD);
}
for (iter = cnt = 0, i = idx; iter < scctx->isc_nrxd[0] && iter <= budget;) {
rxd = (union e1000_adv_rx_desc *)&rxr->rx_base[i];
staterr = le32toh(rxd->wb.upper.status_error);
if ((staterr & E1000_RXD_STAT_DD) == 0)
break;
if (++i == scctx->isc_nrxd[0]) {
i = 0;
}
if (staterr & E1000_RXD_STAT_EOP)
cnt++;
iter++;
}
return (cnt);
}
/****************************************************************
* Routine sends data which has been dma'ed into host memory
* to upper layer. Initialize ri structure.
*
* Returns 0 upon success, errno on failure
***************************************************************/
static int
igb_isc_rxd_pkt_get(void *arg, if_rxd_info_t ri)
{
struct adapter *adapter = arg;
if_softc_ctx_t scctx = adapter->shared;
struct em_rx_queue *que = &adapter->rx_queues[ri->iri_qsidx];
struct rx_ring *rxr = &que->rxr;
struct ifnet *ifp = iflib_get_ifp(adapter->ctx);
union e1000_adv_rx_desc *rxd;
u16 pkt_info, len;
u16 vtag = 0;
u32 ptype;
u32 staterr = 0;
bool eop;
int i = 0;
int cidx = ri->iri_cidx;
do {
rxd = (union e1000_adv_rx_desc *)&rxr->rx_base[cidx];
staterr = le32toh(rxd->wb.upper.status_error);
pkt_info = le16toh(rxd->wb.lower.lo_dword.hs_rss.pkt_info);
MPASS ((staterr & E1000_RXD_STAT_DD) != 0);
len = le16toh(rxd->wb.upper.length);
ptype = le32toh(rxd->wb.lower.lo_dword.data) & IGB_PKTTYPE_MASK;
ri->iri_len += len;
rxr->rx_bytes += ri->iri_len;
rxd->wb.upper.status_error = 0;
eop = ((staterr & E1000_RXD_STAT_EOP) == E1000_RXD_STAT_EOP);
if (((adapter->hw.mac.type == e1000_i350) ||
(adapter->hw.mac.type == e1000_i354)) &&
(staterr & E1000_RXDEXT_STATERR_LB))
vtag = be16toh(rxd->wb.upper.vlan);
else
vtag = le16toh(rxd->wb.upper.vlan);
/* Make sure bad packets are discarded */
if (eop && ((staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK) != 0)) {
adapter->dropped_pkts++;
++rxr->rx_discarded;
return (EBADMSG);
}
ri->iri_frags[i].irf_flid = 0;
ri->iri_frags[i].irf_idx = cidx;
ri->iri_frags[i].irf_len = len;
if (++cidx == scctx->isc_nrxd[0])
cidx = 0;
#ifdef notyet
if (rxr->hdr_split == TRUE) {
ri->iri_frags[i].irf_flid = 1;
ri->iri_frags[i].irf_idx = cidx;
if (++cidx == scctx->isc_nrxd[0])
cidx = 0;
}
#endif
i++;
} while (!eop);
rxr->rx_packets++;
if ((ifp->if_capenable & IFCAP_RXCSUM) != 0)
igb_rx_checksum(staterr, ri, ptype);
if ((ifp->if_capenable & IFCAP_VLAN_HWTAGGING) != 0 &&
(staterr & E1000_RXD_STAT_VP) != 0) {
ri->iri_vtag = vtag;
ri->iri_flags |= M_VLANTAG;
}
ri->iri_flowid =
le32toh(rxd->wb.lower.hi_dword.rss);
ri->iri_rsstype = igb_determine_rsstype(pkt_info);
ri->iri_nfrags = i;
return (0);
}
/*********************************************************************
*
* Verify that the hardware indicated that the checksum is valid.
* Inform the stack about the status of checksum so that stack
* doesn't spend time verifying the checksum.
*
*********************************************************************/
static void
igb_rx_checksum(u32 staterr, if_rxd_info_t ri, u32 ptype)
{
u16 status = (u16)staterr;
u8 errors = (u8) (staterr >> 24);
bool sctp = FALSE;
/* Ignore Checksum bit is set */
if (status & E1000_RXD_STAT_IXSM) {
ri->iri_csum_flags = 0;
return;
}
if ((ptype & E1000_RXDADV_PKTTYPE_ETQF) == 0 &&
(ptype & E1000_RXDADV_PKTTYPE_SCTP) != 0)
sctp = 1;
else
sctp = 0;
if (status & E1000_RXD_STAT_IPCS) {
/* Did it pass? */
if (!(errors & E1000_RXD_ERR_IPE)) {
/* IP Checksum Good */
ri->iri_csum_flags = CSUM_IP_CHECKED;
ri->iri_csum_flags |= CSUM_IP_VALID;
} else
ri->iri_csum_flags = 0;
}
if (status & (E1000_RXD_STAT_TCPCS | E1000_RXD_STAT_UDPCS)) {
u64 type = (CSUM_DATA_VALID | CSUM_PSEUDO_HDR);
if (sctp) /* reassign */
type = CSUM_SCTP_VALID;
/* Did it pass? */
if (!(errors & E1000_RXD_ERR_TCPE)) {
ri->iri_csum_flags |= type;
if (sctp == 0)
ri->iri_csum_data = htons(0xffff);
}
}
return;
}
/********************************************************************
*
* Parse the packet type to determine the appropriate hash
*
******************************************************************/
static int
igb_determine_rsstype(u16 pkt_info)
{
switch (pkt_info & E1000_RXDADV_RSSTYPE_MASK) {
case E1000_RXDADV_RSSTYPE_IPV4_TCP:
return M_HASHTYPE_RSS_TCP_IPV4;
case E1000_RXDADV_RSSTYPE_IPV4:
return M_HASHTYPE_RSS_IPV4;
case E1000_RXDADV_RSSTYPE_IPV6_TCP:
return M_HASHTYPE_RSS_TCP_IPV6;
case E1000_RXDADV_RSSTYPE_IPV6_EX:
return M_HASHTYPE_RSS_IPV6_EX;
case E1000_RXDADV_RSSTYPE_IPV6:
return M_HASHTYPE_RSS_IPV6;
case E1000_RXDADV_RSSTYPE_IPV6_TCP_EX:
return M_HASHTYPE_RSS_TCP_IPV6_EX;
default:
return M_HASHTYPE_OPAQUE;
}
}