d/freebsd-skq
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Migrate e1000 to the IFLIB framework:

Browse Source 
- em(4) igb(4) and lem(4)
- deprecate the igb device from kernel configurations
- create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko

Devices tested:
- 82574L
- I218-LM
- 82546GB
- 82579LM
- I350
- I217

Please report problems to freebsd-net@freebsd.org

Partial review from jhb and suggestions on how to *not* brick folks who
originally would have lost their igbX device.

Submitted by:	mmacy@nextbsd.org
MFC after:	2 weeks
Relnotes:	yes
Sponsored by:	Limelight Networks and Dell EMC Isilon
Differential Revision:	https://reviews.freebsd.org/D8299
This commit is contained in:
Sean Bruno 2017-01-10 03:23:22 +00:00
parent e8257dbe43
commit f2d6ace4a6
20 changed files with 3115 additions and 16170 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
UPDATING
View File

@ -51,6 +51,11 @@ NOTE TO PEOPLE WHO THINK THAT FreeBSD 12.x IS SLOW:
****************************** SPECIAL WARNING: ******************************
20170109:
The igb(4), em(4) and lem(4) ethernet drivers are now implemented via
IFLIB. If you have a custom kernel configuration that excludes em(4)
but you use igb(4), you need to re-add em(4) to your custom configuration.
20161217:
Clang, llvm, lldb, compiler-rt and libc++ have been upgraded to 3.9.1.
Please see the 20141231 entry below for information about prerequisites

1
sys/amd64/conf/GENERIC
View File

@ -230,7 +230,6 @@ device puc # Multi I/O cards and multi-channel UARTs
device bxe # Broadcom NetXtreme II BCM5771X/BCM578XX 10GbE
device de # DEC/Intel DC21x4x (``Tulip'')
device em # Intel PRO/1000 Gigabit Ethernet Family
device igb # Intel PRO/1000 PCIE Server Gigabit Family
device ix # Intel PRO/10GbE PCIE PF Ethernet
device ixv # Intel PRO/10GbE PCIE VF Ethernet
device ixl # Intel XL710 40Gbe PCIE Ethernet

1
sys/arm64/conf/GENERIC
View File

@ -120,7 +120,6 @@ device mii
device miibus # MII bus support
device awg # Allwinner EMAC Gigabit Ethernet
device em # Intel PRO/1000 Gigabit Ethernet Family
device igb # Intel PRO/1000 PCIE Server Gigabit Family
device ix # Intel 10Gb Ethernet Family
device msk # Marvell/SysKonnect Yukon II Gigabit Ethernet
device smc # SMSC LAN91C111

2
sys/conf/NOTES
View File

@ -1972,7 +1972,6 @@ device xmphy # XaQti XMAC II
# KNE110TX.
# de: Digital Equipment DC21040
# em: Intel Pro/1000 Gigabit Ethernet 82542, 82543, 82544 based adapters.
# igb: Intel Pro/1000 PCI Express Gigabit Ethernet: 82575 and later adapters.
# ep: 3Com 3C509, 3C529, 3C556, 3C562D, 3C563D, 3C572, 3C574X, 3C579, 3C589
# and PC Card devices using these chipsets.
# ex: Intel EtherExpress Pro/10 and other i82595-based adapters,
@ -2145,7 +2144,6 @@ device cxgbe # Chelsio T4-T6 1/10/25/40/100 Gigabit Ethernet
device cxgbev # Chelsio T4-T6 Virtual Functions
device de # DEC/Intel DC21x4x (``Tulip'')
device em # Intel Pro/1000 Gigabit Ethernet
device igb # Intel Pro/1000 PCIE Gigabit Ethernet
device ixgb # Intel Pro/10Gbe PCI-X Ethernet
device ix # Intel Pro/10Gbe PCIE Ethernet
device ixv # Intel Pro/10Gbe PCIE Ethernet VF

38
sys/conf/files
View File

@ -1572,43 +1572,43 @@ dev/eisa/eisa_if.m standard
dev/eisa/eisaconf.c optional eisa
dev/e1000/if_em.c optional em \
compile-with "${NORMAL_C} -I$S/dev/e1000"
dev/e1000/if_lem.c optional em \
dev/e1000/em_txrx.c optional em \
compile-with "${NORMAL_C} -I$S/dev/e1000"
dev/e1000/if_igb.c optional igb \
dev/e1000/igb_txrx.c optional em \
compile-with "${NORMAL_C} -I$S/dev/e1000"
dev/e1000/e1000_80003es2lan.c optional em | igb \
dev/e1000/e1000_80003es2lan.c optional em \
compile-with "${NORMAL_C} -I$S/dev/e1000"
dev/e1000/e1000_82540.c optional em | igb \
dev/e1000/e1000_82540.c optional em \
compile-with "${NORMAL_C} -I$S/dev/e1000"
dev/e1000/e1000_82541.c optional em | igb \
dev/e1000/e1000_82541.c optional em \
compile-with "${NORMAL_C} -I$S/dev/e1000"
dev/e1000/e1000_82542.c optional em | igb \
dev/e1000/e1000_82542.c optional em \
compile-with "${NORMAL_C} -I$S/dev/e1000"
dev/e1000/e1000_82543.c optional em | igb \
dev/e1000/e1000_82543.c optional em \
compile-with "${NORMAL_C} -I$S/dev/e1000"
dev/e1000/e1000_82571.c optional em | igb \
dev/e1000/e1000_82571.c optional em \
compile-with "${NORMAL_C} -I$S/dev/e1000"
dev/e1000/e1000_82575.c optional em | igb \
dev/e1000/e1000_82575.c optional em \
compile-with "${NORMAL_C} -I$S/dev/e1000"
dev/e1000/e1000_ich8lan.c optional em | igb \
dev/e1000/e1000_ich8lan.c optional em \
compile-with "${NORMAL_C} -I$S/dev/e1000"
dev/e1000/e1000_i210.c optional em | igb \
dev/e1000/e1000_i210.c optional em \
compile-with "${NORMAL_C} -I$S/dev/e1000"
dev/e1000/e1000_api.c optional em | igb \
dev/e1000/e1000_api.c optional em \
compile-with "${NORMAL_C} -I$S/dev/e1000"
dev/e1000/e1000_mac.c optional em | igb \
dev/e1000/e1000_mac.c optional em \
compile-with "${NORMAL_C} -I$S/dev/e1000"
dev/e1000/e1000_manage.c optional em | igb \
dev/e1000/e1000_manage.c optional em \
compile-with "${NORMAL_C} -I$S/dev/e1000"
dev/e1000/e1000_nvm.c optional em | igb \
dev/e1000/e1000_nvm.c optional em \
compile-with "${NORMAL_C} -I$S/dev/e1000"
dev/e1000/e1000_phy.c optional em | igb \
dev/e1000/e1000_phy.c optional em \
compile-with "${NORMAL_C} -I$S/dev/e1000"
dev/e1000/e1000_vf.c optional em | igb \
dev/e1000/e1000_vf.c optional em \
compile-with "${NORMAL_C} -I$S/dev/e1000"
dev/e1000/e1000_mbx.c optional em | igb \
dev/e1000/e1000_mbx.c optional em \
compile-with "${NORMAL_C} -I$S/dev/e1000"
dev/e1000/e1000_osdep.c optional em | igb \
dev/e1000/e1000_osdep.c optional em \
compile-with "${NORMAL_C} -I$S/dev/e1000"
dev/et/if_et.c optional et
dev/en/if_en_pci.c optional en pci

1
sys/conf/makeLINT.mk
View File

@ -38,7 +38,6 @@ LINT: ${NOTES} ../../conf/makeLINT.sed
echo "nodevice bxe" >> ${.TARGET}-NOIP
echo "nodevice em" >> ${.TARGET}-NOIP
echo "nodevice fxp" >> ${.TARGET}-NOIP
echo "nodevice igb" >> ${.TARGET}-NOIP
echo "nodevice jme" >> ${.TARGET}-NOIP
echo "nodevice msk" >> ${.TARGET}-NOIP
echo "nodevice mxge" >> ${.TARGET}-NOIP

720
sys/dev/e1000/em_txrx.c Normal file
View File

@ -0,0 +1,720 @@
/* $FreeBSD$ */
#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 em_tso_setup(struct adapter *adapter, if_pkt_info_t pi, u32 *txd_upper, u32 *txd_lower);
static int em_transmit_checksum_setup(struct adapter *adapter, if_pkt_info_t pi, u32 *txd_upper, u32 *txd_lower);
static int em_isc_txd_encap(void *arg, if_pkt_info_t pi);
static void em_isc_txd_flush(void *arg, uint16_t txqid, uint32_t pidx);
static int em_isc_txd_credits_update(void *arg, uint16_t txqid, uint32_t cidx_init, bool clear);
static void em_isc_rxd_refill(void *arg, uint16_t rxqid, uint8_t flid __unused,
uint32_t pidx, uint64_t *paddrs, caddr_t *vaddrs __unused, uint16_t count, uint16_t buflen __unused);
static void em_isc_rxd_flush(void *arg, uint16_t rxqid, uint8_t flid __unused, uint32_t pidx);
static int em_isc_rxd_available(void *arg, uint16_t rxqid, uint32_t idx,
int budget);
static int em_isc_rxd_pkt_get(void *arg, if_rxd_info_t ri);
static void lem_isc_rxd_refill(void *arg, uint16_t rxqid, uint8_t flid __unused,
uint32_t pidx, uint64_t *paddrs, caddr_t *vaddrs __unused, uint16_t count, uint16_t buflen __unused);
static int lem_isc_rxd_available(void *arg, uint16_t rxqid, uint32_t idx,
int budget);
static int lem_isc_rxd_pkt_get(void *arg, if_rxd_info_t ri);
static void lem_receive_checksum(int status, int errors, if_rxd_info_t ri);
static void em_receive_checksum(uint32_t status, if_rxd_info_t ri);
extern int em_intr(void *arg);
struct if_txrx em_txrx = {
em_isc_txd_encap,
em_isc_txd_flush,
em_isc_txd_credits_update,
em_isc_rxd_available,
em_isc_rxd_pkt_get,
em_isc_rxd_refill,
em_isc_rxd_flush,
em_intr
};
struct if_txrx lem_txrx = {
em_isc_txd_encap,
em_isc_txd_flush,
em_isc_txd_credits_update,
lem_isc_rxd_available,
lem_isc_rxd_pkt_get,
lem_isc_rxd_refill,
em_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
em_tso_setup(struct adapter *adapter, if_pkt_info_t pi, u32 *txd_upper, u32 *txd_lower)
{
if_softc_ctx_t scctx = adapter->shared;
struct em_tx_queue *que = &adapter->tx_queues[pi->ipi_qsidx];
struct tx_ring *txr = &que->txr;
struct e1000_context_desc *TXD;
struct em_txbuffer *tx_buffer;
int cur, hdr_len;
hdr_len = pi->ipi_ehdrlen + pi->ipi_ip_hlen + pi->ipi_tcp_hlen;
*txd_lower = (E1000_TXD_CMD_DEXT | /* Extended descr type */
E1000_TXD_DTYP_D | /* Data descr type */
E1000_TXD_CMD_TSE); /* Do TSE on this packet */
/* IP and/or TCP header checksum calculation and insertion. */
*txd_upper = (E1000_TXD_POPTS_IXSM | E1000_TXD_POPTS_TXSM) << 8;
cur = pi->ipi_pidx;
TXD = (struct e1000_context_desc *)&txr->tx_base[cur];
tx_buffer = &txr->tx_buffers[cur];
/*
* Start offset for header checksum calculation.
* End offset for header checksum calculation.
* Offset of place put the checksum.
*/
TXD->lower_setup.ip_fields.ipcss = pi->ipi_ehdrlen;
TXD->lower_setup.ip_fields.ipcse =
htole16(pi->ipi_ehdrlen + pi->ipi_ip_hlen - 1);
TXD->lower_setup.ip_fields.ipcso = pi->ipi_ehdrlen + offsetof(struct ip, ip_sum);
/*
* Start offset for payload checksum calculation.
* End offset for payload checksum calculation.
* Offset of place to put the checksum.
*/
TXD->upper_setup.tcp_fields.tucss = pi->ipi_ehdrlen + pi->ipi_ip_hlen;
TXD->upper_setup.tcp_fields.tucse = 0;
TXD->upper_setup.tcp_fields.tucso =
pi->ipi_ehdrlen + pi->ipi_ip_hlen + offsetof(struct tcphdr, th_sum);
/*
* Payload size per packet w/o any headers.
* Length of all headers up to payload.
*/
TXD->tcp_seg_setup.fields.mss = htole16(pi->ipi_tso_segsz);
TXD->tcp_seg_setup.fields.hdr_len = hdr_len;
TXD->cmd_and_length = htole32(adapter->txd_cmd |
E1000_TXD_CMD_DEXT | /* Extended descr */
E1000_TXD_CMD_TSE | /* TSE context */
E1000_TXD_CMD_IP | /* Do IP csum */
E1000_TXD_CMD_TCP | /* Do TCP checksum */
(pi->ipi_len - hdr_len)); /* Total len */
tx_buffer->eop = -1;
txr->tx_tso = TRUE;
if (++cur == scctx->isc_ntxd[0]) {
cur = 0;
}
DPRINTF(iflib_get_dev(adapter->ctx), "%s: pidx: %d cur: %d\n", __FUNCTION__, pi->ipi_pidx, cur);
return (cur);
}
#define TSO_WORKAROUND 4
#define DONT_FORCE_CTX 1
/*********************************************************************
* The offload context is protocol specific (TCP/UDP) and thus
* only needs to be set when the protocol changes. The occasion
* of a context change can be a performance detriment, and
* might be better just disabled. The reason arises in the way
* in which the controller supports pipelined requests from the
* Tx data DMA. Up to four requests can be pipelined, and they may
* belong to the same packet or to multiple packets. However all
* requests for one packet are issued before a request is issued
* for a subsequent packet and if a request for the next packet
* requires a context change, that request will be stalled
* until the previous request completes. This means setting up
* a new context effectively disables pipelined Tx data DMA which
* in turn greatly slow down performance to send small sized
* frames.
**********************************************************************/
static int
em_transmit_checksum_setup(struct adapter *adapter, if_pkt_info_t pi, u32 *txd_upper, u32 *txd_lower)
{
struct e1000_context_desc *TXD = NULL;
if_softc_ctx_t scctx = adapter->shared;
struct em_tx_queue *que = &adapter->tx_queues[pi->ipi_qsidx];
struct tx_ring *txr = &que->txr;
struct em_txbuffer *tx_buffer;
int csum_flags = pi->ipi_csum_flags;
int cur, hdr_len;
u32 cmd;
cur = pi->ipi_pidx;
hdr_len = pi->ipi_ehdrlen + pi->ipi_ip_hlen;
cmd = adapter->txd_cmd;
/*
* The 82574L can only remember the *last* context used
* regardless of queue that it was use for. We cannot reuse
* contexts on this hardware platform and must generate a new
* context every time. 82574L hardware spec, section 7.2.6,
* second note.
*/
if (DONT_FORCE_CTX &&
adapter->tx_num_queues == 1 &&
txr->csum_lhlen == pi->ipi_ehdrlen &&
txr->csum_iphlen == pi->ipi_ip_hlen &&
txr->csum_flags == csum_flags) {
/*
* Same csum offload context as the previous packets;
* just return.
*/
*txd_upper = txr->csum_txd_upper;
*txd_lower = txr->csum_txd_lower;
return (cur);
}
TXD = (struct e1000_context_desc *)&txr->tx_base[cur];
if (csum_flags & CSUM_IP) {
*txd_upper |= E1000_TXD_POPTS_IXSM << 8;
/*
* Start offset for header checksum calculation.
* End offset for header checksum calculation.
* Offset of place to put the checksum.
*/
TXD->lower_setup.ip_fields.ipcss = pi->ipi_ehdrlen;
TXD->lower_setup.ip_fields.ipcse = htole16(hdr_len);
TXD->lower_setup.ip_fields.ipcso = pi->ipi_ehdrlen + offsetof(struct ip, ip_sum);
cmd |= E1000_TXD_CMD_IP;
}
if (csum_flags & (CSUM_TCP|CSUM_UDP)) {
uint8_t tucso;
*txd_upper |= E1000_TXD_POPTS_TXSM << 8;
*txd_lower = E1000_TXD_CMD_DEXT | E1000_TXD_DTYP_D;
if (csum_flags & CSUM_TCP) {
tucso = hdr_len + offsetof(struct tcphdr, th_sum);
cmd |= E1000_TXD_CMD_TCP;
} else
tucso = hdr_len + offsetof(struct udphdr, uh_sum);
TXD->upper_setup.tcp_fields.tucss = hdr_len;
TXD->upper_setup.tcp_fields.tucse = htole16(0);
TXD->upper_setup.tcp_fields.tucso = tucso;
}
txr->csum_lhlen = pi->ipi_ehdrlen;
txr->csum_iphlen = pi->ipi_ip_hlen;
txr->csum_flags = csum_flags;
txr->csum_txd_upper = *txd_upper;
txr->csum_txd_lower = *txd_lower;
TXD->tcp_seg_setup.data = htole32(0);
TXD->cmd_and_length =
htole32(E1000_TXD_CMD_IFCS | E1000_TXD_CMD_DEXT | cmd);
tx_buffer = &txr->tx_buffers[cur];
tx_buffer->eop = -1;
if (++cur == scctx->isc_ntxd[0]) {
cur = 0;
}
DPRINTF(iflib_get_dev(adapter->ctx), "checksum_setup csum_flags=%x txd_upper=%x txd_lower=%x hdr_len=%d cmd=%x\n",
csum_flags, *txd_upper, *txd_lower, hdr_len, cmd);
return (cur);
}
static int
em_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;
bus_dma_segment_t *segs = pi->ipi_segs;
int nsegs = pi->ipi_nsegs;
int csum_flags = pi->ipi_csum_flags;
int i, j, first, pidx_last;
u32 txd_upper = 0, txd_lower = 0;
struct em_txbuffer *tx_buffer;
struct e1000_tx_desc *ctxd = NULL;
bool do_tso, tso_desc;
i = first = pi->ipi_pidx;
do_tso = (csum_flags & CSUM_TSO);
tso_desc = FALSE;
/*
* TSO Hardware workaround, if this packet is not
* TSO, and is only a single descriptor long, and
* it follows a TSO burst, then we need to add a
* sentinel descriptor to prevent premature writeback.
*/
if ((!do_tso) && (txr->tx_tso == TRUE)) {
if (nsegs == 1)
tso_desc = TRUE;
txr->tx_tso = FALSE;
}
/* Do hardware assists */
if (do_tso) {
i = em_tso_setup(sc, pi, &txd_upper, &txd_lower);
tso_desc = TRUE;
} else if (csum_flags & CSUM_OFFLOAD) {
i = em_transmit_checksum_setup(sc, pi, &txd_upper, &txd_lower);
}
if (pi->ipi_mflags & M_VLANTAG) {
/* Set the vlan id. */
txd_upper |= htole16(pi->ipi_vtag) << 16;
/* Tell hardware to add tag */
txd_lower |= htole32(E1000_TXD_CMD_VLE);
}
DPRINTF(iflib_get_dev(sc->ctx), "encap: set up tx: nsegs=%d first=%d i=%d\n", nsegs, first, i);
/* XXX adapter->pcix_82544 -- lem_fill_descriptors */
/* Set up our transmit descriptors */
for (j = 0; j < nsegs; j++) {
bus_size_t seg_len;
bus_addr_t seg_addr;
uint32_t cmd;
ctxd = &txr->tx_base[i];
tx_buffer = &txr->tx_buffers[i];
seg_addr = segs[j].ds_addr;
seg_len = segs[j].ds_len;
cmd = E1000_TXD_CMD_IFCS | sc->txd_cmd;
/*
** TSO Workaround:
** If this is the last descriptor, we want to
** split it so we have a small final sentinel
*/
if (tso_desc && (j == (nsegs - 1)) && (seg_len > 8)) {
seg_len -= TSO_WORKAROUND;
ctxd->buffer_addr = htole64(seg_addr);
ctxd->lower.data = htole32(cmd | txd_lower | seg_len);
ctxd->upper.data = htole32(txd_upper);
if (++i == scctx->isc_ntxd[0])
i = 0;
/* Now make the sentinel */
ctxd = &txr->tx_base[i];
tx_buffer = &txr->tx_buffers[i];
ctxd->buffer_addr = htole64(seg_addr + seg_len);
ctxd->lower.data = htole32(cmd | txd_lower | TSO_WORKAROUND);
ctxd->upper.data = htole32(txd_upper);
pidx_last = i;
if (++i == scctx->isc_ntxd[0])
i = 0;
DPRINTF(iflib_get_dev(sc->ctx), "TSO path pidx_last=%d i=%d ntxd[0]=%d\n", pidx_last, i, scctx->isc_ntxd[0]);
} else {
ctxd->buffer_addr = htole64(seg_addr);
ctxd->lower.data = htole32(cmd | txd_lower | seg_len);
ctxd->upper.data = htole32(txd_upper);
pidx_last = i;
if (++i == scctx->isc_ntxd[0])
i = 0;
DPRINTF(iflib_get_dev(sc->ctx), "pidx_last=%d i=%d ntxd[0]=%d\n", pidx_last, i, scctx->isc_ntxd[0]);
}
tx_buffer->eop = -1;
}
/*
* Last Descriptor of Packet
* needs End Of Packet (EOP)
* and Report Status (RS)
*/
ctxd->lower.data |=
htole32(E1000_TXD_CMD_EOP | E1000_TXD_CMD_RS);
tx_buffer = &txr->tx_buffers[first];
tx_buffer->eop = pidx_last;
DPRINTF(iflib_get_dev(sc->ctx), "tx_buffers[%d]->eop = %d ipi_new_pidx=%d\n", first, pidx_last, i);
pi->ipi_new_pidx = i;
return (0);
}
static void
em_isc_txd_flush(void *arg, uint16_t txqid, uint32_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
em_isc_txd_credits_update(void *arg, uint16_t txqid, uint32_t cidx_init, 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;
u32 cidx, processed = 0;
int last, done;
struct em_txbuffer *buf;
struct e1000_tx_desc *tx_desc, *eop_desc;
cidx = cidx_init;
buf = &txr->tx_buffers[cidx];
tx_desc = &txr->tx_base[cidx];
last = buf->eop;
eop_desc = &txr->tx_base[last];
DPRINTF(iflib_get_dev(adapter->ctx), "credits_update: cidx_init=%d clear=%d last=%d\n",
cidx_init, clear, last);
/*
* What this does is get the index of the
* first descriptor AFTER the EOP of the
* first packet, that way we can do the
* simple comparison on the inner while loop.
*/
if (++last == scctx->isc_ntxd[0])
last = 0;
done = last;
while (eop_desc->upper.fields.status & E1000_TXD_STAT_DD) {
/* We clean the range of the packet */
while (cidx != done) {
if (clear) {
tx_desc->upper.data = 0;
tx_desc->lower.data = 0;
tx_desc->buffer_addr = 0;
buf->eop = -1;
}
tx_desc++;
buf++;
processed++;
/* wrap the ring ? */
if (++cidx == scctx->isc_ntxd[0]) {
cidx = 0;
}
buf = &txr->tx_buffers[cidx];
tx_desc = &txr->tx_base[cidx];
}
/* See if we can continue to the next packet */
last = buf->eop;
if (last == -1)
break;
eop_desc = &txr->tx_base[last];
/* Get new done point */
if (++last == scctx->isc_ntxd[0])
last = 0;
done = last;
}
DPRINTF(iflib_get_dev(adapter->ctx), "Processed %d credits update\n", processed);
return(processed);
}
static void
lem_isc_rxd_refill(void *arg, uint16_t rxqid, uint8_t flid __unused,
uint32_t pidx, uint64_t *paddrs, caddr_t *vaddrs __unused, uint16_t count, uint16_t buflen __unused)
{
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;
struct e1000_rx_desc *rxd;
int i;
uint32_t next_pidx;
for (i = 0, next_pidx = pidx; i < count; i++) {
rxd = (struct e1000_rx_desc *)&rxr->rx_base[next_pidx];
rxd->buffer_addr = htole64(paddrs[i]);
/* status bits must be cleared */
rxd->status = 0;
if (++next_pidx == scctx->isc_nrxd[0])
next_pidx = 0;
}
}
static void
em_isc_rxd_refill(void *arg, uint16_t rxqid, uint8_t flid __unused,
uint32_t pidx, uint64_t *paddrs, caddr_t *vaddrs __unused, uint16_t count, uint16_t buflen __unused)
{
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_rx_desc_extended *rxd;
int i;
uint32_t next_pidx;
for (i = 0, next_pidx = pidx; i < count; i++) {
rxd = &rxr->rx_base[next_pidx];
rxd->read.buffer_addr = htole64(paddrs[i]);
/* DD bits must be cleared */
rxd->wb.upper.status_error = 0;
if (++next_pidx == scctx->isc_nrxd[0])
next_pidx = 0;
}
}
static void
em_isc_rxd_flush(void *arg, uint16_t rxqid, uint8_t flid __unused, uint32_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
lem_isc_rxd_available(void *arg, uint16_t rxqid, uint32_t idx, int 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;
struct e1000_rx_desc *rxd;
u32 staterr = 0;
int cnt, i;
for (cnt = 0, i = idx; cnt < scctx->isc_nrxd[0] && cnt <= budget;) {
rxd = (struct e1000_rx_desc *)&rxr->rx_base[i];
staterr = rxd->status;
if ((staterr & E1000_RXD_STAT_DD) == 0)
break;
if (++i == scctx->isc_nrxd[0])
i = 0;
if (staterr & E1000_RXD_STAT_EOP)
cnt++;
}
return (cnt);
}
static int
em_isc_rxd_available(void *arg, uint16_t rxqid, uint32_t idx, int 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_rx_desc_extended *rxd;
u32 staterr = 0;
int cnt, i;
for (cnt = 0, i = idx; cnt < scctx->isc_nrxd[0] && cnt <= budget;) {
rxd = &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++;
}
return (cnt);
}
static int
lem_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 e1000_rx_desc *rxd;
u16 len;
u32 status, errors;
bool eop;
int i, cidx;
status = errors = i = 0;
cidx = ri->iri_cidx;
do {
rxd = (struct e1000_rx_desc *)&rxr->rx_base[cidx];
status = rxd->status;
errors = rxd->errors;
/* Error Checking then decrement count */
MPASS ((status & E1000_RXD_STAT_DD) != 0);
len = le16toh(rxd->length);
ri->iri_len += len;
eop = (status & E1000_RXD_STAT_EOP) != 0;
/* Make sure bad packets are discarded */
if (errors & E1000_RXD_ERR_FRAME_ERR_MASK) {
adapter->dropped_pkts++;
/* XXX fixup if common */
return (EBADMSG);
}
ri->iri_frags[i].irf_flid = 0;
ri->iri_frags[i].irf_idx = cidx;
ri->iri_frags[i].irf_len = len;
/* Zero out the receive descriptors status. */
rxd->status = 0;
if (++cidx == scctx->isc_nrxd[0])
cidx = 0;
i++;
} while (!eop);
/* XXX add a faster way to look this up */
if (adapter->hw.mac.type >= e1000_82543 && !(status & E1000_RXD_STAT_IXSM))
lem_receive_checksum(status, errors, ri);
if (status & E1000_RXD_STAT_VP) {
ri->iri_vtag = le16toh(rxd->special);
ri->iri_flags |= M_VLANTAG;
}
ri->iri_nfrags = i;
return (0);
}
static int
em_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;
union e1000_rx_desc_extended *rxd;
u16 len;
u32 staterr = 0;
bool eop;
int i, cidx, vtag;
i = vtag = 0;
cidx = ri->iri_cidx;
do {
rxd = &rxr->rx_base[cidx];
staterr = le32toh(rxd->wb.upper.status_error);
/* Error Checking then decrement count */
MPASS ((staterr & E1000_RXD_STAT_DD) != 0);
len = le16toh(rxd->wb.upper.length);
ri->iri_len += len;
eop = (staterr & E1000_RXD_STAT_EOP) != 0;
/* Make sure bad packets are discarded */
if (staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK) {
adapter->dropped_pkts++;
return EBADMSG;
}
ri->iri_frags[i].irf_flid = 0;
ri->iri_frags[i].irf_idx = cidx;
ri->iri_frags[i].irf_len = len;
/* Zero out the receive descriptors status. */
rxd->wb.upper.status_error &= htole32(~0xFF);
if (++cidx == scctx->isc_nrxd[0])
cidx = 0;
i++;
} while (!eop);
/* XXX add a faster way to look this up */
if (adapter->hw.mac.type >= e1000_82543)
em_receive_checksum(staterr, ri);
if (staterr & E1000_RXD_STAT_VP) {
vtag = le16toh(rxd->wb.upper.vlan);
}
ri->iri_vtag = vtag;
ri->iri_nfrags = i;
if (vtag)
ri->iri_flags |= M_VLANTAG;
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
lem_receive_checksum(int status, int errors, if_rxd_info_t ri)
{
/* Did it pass? */
if (status & E1000_RXD_STAT_IPCS && !(errors & E1000_RXD_ERR_IPE))
ri->iri_csum_flags = (CSUM_IP_CHECKED|CSUM_IP_VALID);
if (status & E1000_RXD_STAT_TCPCS) {
/* Did it pass? */
if (!(errors & E1000_RXD_ERR_TCPE)) {
ri->iri_csum_flags |=
(CSUM_DATA_VALID | CSUM_PSEUDO_HDR);
ri->iri_csum_data = htons(0xffff);
}
}
}
static void
em_receive_checksum(uint32_t status, if_rxd_info_t ri)
{
ri->iri_csum_flags = 0;
/* Ignore Checksum bit is set */
if (status & E1000_RXD_STAT_IXSM)
return;
/* If the IP checksum exists and there is no IP Checksum error */
if ((status & (E1000_RXD_STAT_IPCS | E1000_RXDEXT_STATERR_IPE)) ==
E1000_RXD_STAT_IPCS) {
ri->iri_csum_flags = (CSUM_IP_CHECKED | CSUM_IP_VALID);
}
/* TCP or UDP checksum */
if ((status & (E1000_RXD_STAT_TCPCS | E1000_RXDEXT_STATERR_TCPE)) ==
E1000_RXD_STAT_TCPCS) {
ri->iri_csum_flags |= (CSUM_DATA_VALID | CSUM_PSEUDO_HDR);
ri->iri_csum_data = htons(0xffff);
}
if (status & E1000_RXD_STAT_UDPCS) {
ri->iri_csum_flags |= (CSUM_DATA_VALID | CSUM_PSEUDO_HDR);
ri->iri_csum_data = htons(0xffff);
}
}

5251
sys/dev/e1000/if_em.c
View File

File diff suppressed because it is too large Load Diff

295
sys/dev/e1000/if_em.h
View File

@ -1,36 +1,67 @@
/******************************************************************************
Copyright (c) 2001-2015, Intel Corporation
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.
3. Neither the name of the Intel Corporation nor the names of its
contributors may be used to endorse or promote products derived from
this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS 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 COPYRIGHT OWNER 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 "opt_em.h"
#include "opt_ddb.h"
#include "opt_inet.h"
#include "opt_inet6.h"
#ifdef HAVE_KERNEL_OPTION_HEADERS
#include "opt_device_polling.h"
#endif
#include <sys/param.h>
#include <sys/systm.h>
#ifdef DDB
#include <sys/types.h>
#include <ddb/ddb.h>
#endif
#if __FreeBSD_version >= 800000
#include <sys/buf_ring.h>
#endif
#include <sys/bus.h>
#include <sys/endian.h>
#include <sys/kernel.h>
#include <sys/kthread.h>
#include <sys/malloc.h>
#include <sys/mbuf.h>
#include <sys/module.h>
#include <sys/rman.h>
#include <sys/smp.h>
#include <sys/socket.h>
#include <sys/sockio.h>
#include <sys/sysctl.h>
#include <sys/taskqueue.h>
#include <sys/eventhandler.h>
#include <machine/bus.h>
#include <machine/resource.h>
#include <net/bpf.h>
#include <net/ethernet.h>
#include <net/if.h>
#include <net/if_var.h>
#include <net/if_arp.h>
#include <net/if_dl.h>
#include <net/if_media.h>
#include <net/iflib.h>
#include <net/if_types.h>
#include <net/if_vlan_var.h>
#include <netinet/in_systm.h>
#include <netinet/in.h>
#include <netinet/if_ether.h>
#include <netinet/ip.h>
#include <netinet/ip6.h>
#include <netinet/tcp.h>
#include <netinet/udp.h>
#include <machine/in_cksum.h>
#include <dev/led/led.h>
#include <dev/pci/pcivar.h>
#include <dev/pci/pcireg.h>
#include "e1000_api.h"
#include "e1000_82571.h"
#include "ifdi_if.h"
#ifndef _EM_H_DEFINED_
@ -51,13 +82,10 @@
* desscriptors should meet the following condition.
* (num_tx_desc * sizeof(struct e1000_tx_desc)) % 128 == 0
*/
#define EM_MIN_TXD 80
#define EM_MIN_TXD 128
#define EM_MAX_TXD 4096
#ifdef EM_MULTIQUEUE
#define EM_DEFAULT_TXD 4096
#else
#define EM_DEFAULT_TXD 1024
#endif
#define EM_DEFAULT_TXD 1024
#define EM_DEFAULT_MULTI_TXD 4096
/*
* EM_RXD - Maximum number of receive Descriptors
@ -72,13 +100,10 @@
* desscriptors should meet the following condition.
* (num_tx_desc * sizeof(struct e1000_tx_desc)) % 128 == 0
*/
#define EM_MIN_RXD 80
#define EM_MIN_RXD 128
#define EM_MAX_RXD 4096
#ifdef EM_MULTIQUEUE
#define EM_DEFAULT_RXD 4096
#else
#define EM_DEFAULT_RXD 1024
#endif
#define EM_DEFAULT_RXD 1024
#define EM_DEFAULT_MULTI_RXD 4096
/*
* EM_TIDV - Transmit Interrupt Delay Value
@ -148,17 +173,6 @@
#define EM_RADV 64
#endif
/*
* This parameter controls the max duration of transmit watchdog.
*/
#define EM_WATCHDOG (10 * hz)
/*
* This parameter controls when the driver calls the routine to reclaim
* transmit descriptors.
*/
#define EM_TX_CLEANUP_THRESHOLD (adapter->num_tx_desc / 8)
/*
* This parameter controls whether or not autonegotation is enabled.
* 0 - Disable autonegotiation
@ -221,6 +235,18 @@
#define PCICFG_DESC_RING_STATUS 0xe4
#define FLUSH_DESC_REQUIRED 0x100
#define IGB_RX_PTHRESH ((hw->mac.type == e1000_i354) ? 12 : \
((hw->mac.type <= e1000_82576) ? 16 : 8))
#define IGB_RX_HTHRESH 8
#define IGB_RX_WTHRESH ((hw->mac.type == e1000_82576 && \
(adapter->intr_type == IFLIB_INTR_MSIX)) ? 1 : 4)
#define IGB_TX_PTHRESH ((hw->mac.type == e1000_i354) ? 20 : 8)
#define IGB_TX_HTHRESH 1
#define IGB_TX_WTHRESH ((hw->mac.type != e1000_82575 && \
(adapter->intr_type == IFLIB_INTR_MSIX) ? 1 : 16)
/*
* TDBA/RDBA should be aligned on 16 byte boundary. But TDLEN/RDLEN should be
* multiple of 128 bytes. So we align TDBA/RDBA on 128 byte boundary. This will
@ -242,6 +268,7 @@
#define EM_BAR_TYPE(v) ((v) & EM_BAR_TYPE_MASK)
#define EM_BAR_TYPE_MASK 0x00000001
#define EM_BAR_TYPE_MMEM 0x00000000
#define EM_BAR_TYPE_IO 0x00000001
#define EM_BAR_TYPE_FLASH 0x0014
#define EM_BAR_MEM_TYPE(v) ((v) & EM_BAR_MEM_TYPE_MASK)
#define EM_BAR_MEM_TYPE_MASK 0x00000006
@ -279,6 +306,9 @@
#define ETH_ADDR_LEN 6
#define CSUM_OFFLOAD 7 /* Offload bits in mbuf flag */
#define IGB_PKTTYPE_MASK 0x0000FFF0
#define IGB_DMCTLX_DCFLUSH_DIS 0x80000000 /* Disable DMA Coalesce Flush */
/*
* 82574 has a nonstandard address for EIAC
* and since its only used in MSIX, and in
@ -295,19 +325,6 @@
#define EM_NVM_MSIX_N_MASK (0x7 << EM_NVM_MSIX_N_SHIFT)
#define EM_NVM_MSIX_N_SHIFT 7
/*
* Bus dma allocation structure used by
* e1000_dma_malloc and e1000_dma_free.
*/
struct em_dma_alloc {
bus_addr_t dma_paddr;
caddr_t dma_vaddr;
bus_dma_tag_t dma_tag;
bus_dmamap_t dma_map;
bus_dma_segment_t dma_seg;
int dma_nseg;
};
struct adapter;
struct em_int_delay_info {
@ -321,35 +338,31 @@ struct em_int_delay_info {
*/
struct tx_ring {
struct adapter *adapter;
struct mtx tx_mtx;
char mtx_name[16];
struct em_tx_queue *que;
u32 me;
u32 msix;
u32 ims;
int busy;
struct em_dma_alloc txdma;
struct e1000_tx_desc *tx_base;
struct task tx_task;
struct taskqueue *tq;
u32 next_avail_desc;
u32 next_to_clean;
uint64_t tx_paddr;
struct em_txbuffer *tx_buffers;
volatile u16 tx_avail;
u32 tx_tso; /* last tx was tso */
u16 last_hw_offload;
u8 last_hw_ipcso;
u8 last_hw_ipcss;
u8 last_hw_tucso;
u8 last_hw_tucss;
#if __FreeBSD_version >= 800000
struct buf_ring *br;
#endif
/* Interrupt resources */
bus_dma_tag_t txtag;
void *tag;
struct resource *res;
unsigned long tx_irq;
unsigned long no_desc_avail;
/* Saved csum offloading context information */
int csum_flags;
int csum_lhlen;
int csum_iphlen;
int csum_thlen;
int csum_mss;
int csum_pktlen;
uint32_t csum_txd_upper;
uint32_t csum_txd_lower; /* last field */
};
/*
@ -357,26 +370,15 @@ struct tx_ring {
*/
struct rx_ring {
struct adapter *adapter;
struct em_rx_queue *que;
u32 me;
u32 msix;
u32 ims;
struct mtx rx_mtx;
char mtx_name[16];
u32 payload;
struct task rx_task;
struct taskqueue *tq;
union e1000_rx_desc_extended *rx_base;
struct em_dma_alloc rxdma;
u32 next_to_refresh;
u32 next_to_check;
struct em_rxbuffer *rx_buffers;
struct mbuf *fmp;
struct mbuf *lmp;
uint64_t rx_paddr;
/* Interrupt resources */
void *tag;
struct resource *res;
bus_dma_tag_t rxtag;
bool discard;
/* Soft stats */
@ -386,62 +388,68 @@ struct rx_ring {
unsigned long rx_bytes;
};
struct em_tx_queue {
struct adapter *adapter;
u32 msix;
u32 eims; /* This queue's EIMS bit */
u32 me;
struct tx_ring txr;
};
struct em_rx_queue {
struct adapter *adapter;
u32 me;
u32 msix;
u32 eims;
struct rx_ring rxr;
u64 irqs;
struct if_irq que_irq;
};
/* Our adapter structure */
struct adapter {
if_t ifp;
struct ifnet *ifp;
struct e1000_hw hw;
if_softc_ctx_t shared;
if_ctx_t ctx;
#define tx_num_queues shared->isc_ntxqsets
#define rx_num_queues shared->isc_nrxqsets
#define intr_type shared->isc_intr
/* FreeBSD operating-system-specific structures. */
struct e1000_osdep osdep;
device_t dev;
struct device *dev;
struct cdev *led_dev;
struct em_tx_queue *tx_queues;
struct em_rx_queue *rx_queues;
struct if_irq irq;
struct resource *memory;
struct resource *flash;
struct resource *msix_mem;
struct resource *ioport;
int io_rid;
struct resource *res;
void *tag;
u32 linkvec;
u32 ivars;
struct ifmedia media;
struct callout timer;
struct ifmedia *media;
int msix;
int if_flags;
int max_frame_size;
int min_frame_size;
struct mtx core_mtx;
int em_insert_vlan_header;
u32 ims;
bool in_detach;
/* Task for FAST handling */
struct task link_task;
struct task que_task;
struct taskqueue *tq; /* private task queue */
struct grouptask link_task;
eventhandler_tag vlan_attach;
eventhandler_tag vlan_detach;
u16 num_vlans;
u8 num_queues;
/*
* Transmit rings:
* Allocated at run time, an array of rings.
*/
struct tx_ring *tx_rings;
int num_tx_desc;
u16 num_vlans;
u32 txd_cmd;
/*
* Receive rings:
* Allocated at run time, an array of rings.
*/
struct rx_ring *rx_rings;
int num_rx_desc;
u32 tx_process_limit;
u32 rx_process_limit;
u32 rx_mbuf_sz;
@ -467,7 +475,12 @@ struct adapter {
u16 link_speed;
u16 link_duplex;
u32 smartspeed;
u32 dmac;
int link_mask;
u64 que_mask;
struct em_int_delay_info tx_int_delay;
struct em_int_delay_info tx_abs_int_delay;
struct em_int_delay_info rx_int_delay;
@ -502,33 +515,9 @@ typedef struct _em_vendor_info_t {
} em_vendor_info_t;
struct em_txbuffer {
int next_eop; /* Index of the desc to watch */
struct mbuf *m_head;
bus_dmamap_t map; /* bus_dma map for packet */
int eop;
};
struct em_rxbuffer {
int next_eop; /* Index of the desc to watch */
struct mbuf *m_head;
bus_dmamap_t map; /* bus_dma map for packet */
bus_addr_t paddr;
};
/*
** Find the number of unrefreshed RX descriptors
*/
static inline u16
e1000_rx_unrefreshed(struct rx_ring *rxr)
{
struct adapter *adapter = rxr->adapter;
if (rxr->next_to_check > rxr->next_to_refresh)
return (rxr->next_to_check - rxr->next_to_refresh - 1);
else
return ((adapter->num_rx_desc + rxr->next_to_check) -
rxr->next_to_refresh - 1);
}
#define EM_CORE_LOCK_INIT(_sc, _name) \
mtx_init(&(_sc)->core_mtx, _name, "EM Core Lock", MTX_DEF)

6450
sys/dev/e1000/if_igb.c
View File

File diff suppressed because it is too large Load Diff

634
sys/dev/e1000/if_igb.h
View File

@ -1,634 +0,0 @@
/******************************************************************************
Copyright (c) 2001-2015, Intel Corporation
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.
3. Neither the name of the Intel Corporation nor the names of its
contributors may be used to endorse or promote products derived from
this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS 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 COPYRIGHT OWNER 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.
******************************************************************************/
/*$FreeBSD$*/
#ifndef _IF_IGB_H_
#define _IF_IGB_H_
#ifdef ALTQ
#define IGB_LEGACY_TX
#endif
#include <sys/param.h>
#include <sys/systm.h>
#ifndef IGB_LEGACY_TX
#include <sys/buf_ring.h>
#endif
#include <sys/bus.h>
#include <sys/endian.h>
#include <sys/kernel.h>
#include <sys/kthread.h>
#include <sys/malloc.h>
#include <sys/mbuf.h>
#include <sys/module.h>
#include <sys/rman.h>
#include <sys/socket.h>
#include <sys/sockio.h>
#include <sys/sysctl.h>
#include <sys/taskqueue.h>
#include <sys/eventhandler.h>
#include <sys/pcpu.h>
#include <sys/smp.h>
#include <machine/smp.h>
#include <machine/bus.h>
#include <machine/resource.h>
#include <net/bpf.h>
#include <net/ethernet.h>
#include <net/if.h>
#include <net/if_var.h>
#include <net/if_arp.h>
#include <net/if_dl.h>
#include <net/if_media.h>
#ifdef RSS
#include <net/rss_config.h>
#include <netinet/in_rss.h>
#endif
#include <net/if_types.h>
#include <net/if_vlan_var.h>
#include <netinet/in_systm.h>
#include <netinet/in.h>
#include <netinet/if_ether.h>
#include <netinet/ip.h>
#include <netinet/ip6.h>
#include <netinet/tcp.h>
#include <netinet/tcp_lro.h>
#include <netinet/udp.h>
#include <machine/in_cksum.h>
#include <dev/led/led.h>
#include <dev/pci/pcivar.h>
#include <dev/pci/pcireg.h>
#include "e1000_api.h"
#include "e1000_82575.h"
/* Tunables */
/*
* IGB_TXD: Maximum number of Transmit Descriptors
*
* This value is the number of transmit descriptors allocated by the driver.
* Increasing this value allows the driver to queue more transmits. Each
* descriptor is 16 bytes.
* Since TDLEN should be multiple of 128bytes, the number of transmit
* desscriptors should meet the following condition.
* (num_tx_desc * sizeof(struct e1000_tx_desc)) % 128 == 0
*/
#define IGB_MIN_TXD 256
#define IGB_DEFAULT_TXD 1024
#define IGB_MAX_TXD 4096
/*
* IGB_RXD: Maximum number of Receive Descriptors
*
* This value is the number of receive descriptors allocated by the driver.
* Increasing this value allows the driver to buffer more incoming packets.
* Each descriptor is 16 bytes. A receive buffer is also allocated for each
* descriptor. The maximum MTU size is 16110.
* Since TDLEN should be multiple of 128bytes, the number of transmit
* desscriptors should meet the following condition.
* (num_tx_desc * sizeof(struct e1000_tx_desc)) % 128 == 0
*/
#define IGB_MIN_RXD 256
#define IGB_DEFAULT_RXD 1024
#define IGB_MAX_RXD 4096
/*
* IGB_TIDV - Transmit Interrupt Delay Value
* Valid Range: 0-65535 (0=off)
* Default Value: 64
* This value delays the generation of transmit interrupts in units of
* 1.024 microseconds. Transmit interrupt reduction can improve CPU
* efficiency if properly tuned for specific network traffic. If the
* system is reporting dropped transmits, this value may be set too high
* causing the driver to run out of available transmit descriptors.
*/
#define IGB_TIDV 64
/*
* IGB_TADV - Transmit Absolute Interrupt Delay Value
* Valid Range: 0-65535 (0=off)
* Default Value: 64
* This value, in units of 1.024 microseconds, limits the delay in which a
* transmit interrupt is generated. Useful only if IGB_TIDV is non-zero,
* this value ensures that an interrupt is generated after the initial
* packet is sent on the wire within the set amount of time. Proper tuning,
* along with IGB_TIDV, may improve traffic throughput in specific
* network conditions.
*/
#define IGB_TADV 64
/*
* IGB_RDTR - Receive Interrupt Delay Timer (Packet Timer)
* Valid Range: 0-65535 (0=off)
* Default Value: 0
* This value delays the generation of receive interrupts in units of 1.024
* microseconds. Receive interrupt reduction can improve CPU efficiency if
* properly tuned for specific network traffic. Increasing this value adds
* extra latency to frame reception and can end up decreasing the throughput
* of TCP traffic. If the system is reporting dropped receives, this value
* may be set too high, causing the driver to run out of available receive
* descriptors.
*
* CAUTION: When setting IGB_RDTR to a value other than 0, adapters
* may hang (stop transmitting) under certain network conditions.
* If this occurs a WATCHDOG message is logged in the system
* event log. In addition, the controller is automatically reset,
* restoring the network connection. To eliminate the potential
* for the hang ensure that IGB_RDTR is set to 0.
*/
#define IGB_RDTR 0
/*
* Receive Interrupt Absolute Delay Timer (Not valid for 82542/82543/82544)
* Valid Range: 0-65535 (0=off)
* Default Value: 64
* This value, in units of 1.024 microseconds, limits the delay in which a
* receive interrupt is generated. Useful only if IGB_RDTR is non-zero,
* this value ensures that an interrupt is generated after the initial
* packet is received within the set amount of time. Proper tuning,
* along with IGB_RDTR, may improve traffic throughput in specific network
* conditions.
*/
#define IGB_RADV 64
/*
* This parameter controls the duration of transmit watchdog timer.
*/
#define IGB_WATCHDOG (10 * hz)
/*
* This parameter controls when the driver calls the routine to reclaim
* transmit descriptors. Cleaning earlier seems a win.
*/
#define IGB_TX_CLEANUP_THRESHOLD (adapter->num_tx_desc / 2)
/*
* This parameter controls whether or not autonegotation is enabled.
* 0 - Disable autonegotiation
* 1 - Enable autonegotiation
*/
#define DO_AUTO_NEG 1
/*
* This parameter control whether or not the driver will wait for
* autonegotiation to complete.
* 1 - Wait for autonegotiation to complete
* 0 - Don't wait for autonegotiation to complete
*/
#define WAIT_FOR_AUTO_NEG_DEFAULT 0
/* Tunables -- End */
#define AUTONEG_ADV_DEFAULT (ADVERTISE_10_HALF | ADVERTISE_10_FULL | \
ADVERTISE_100_HALF | ADVERTISE_100_FULL | \
ADVERTISE_1000_FULL)
#define AUTO_ALL_MODES 0
/* PHY master/slave setting */
#define IGB_MASTER_SLAVE e1000_ms_hw_default
/* Support AutoMediaDetect for Marvell M88 PHY in i354 */
#define IGB_MEDIA_RESET (1 << 0)
/*
* Micellaneous constants
*/
#define IGB_INTEL_VENDOR_ID 0x8086
#define IGB_JUMBO_PBA 0x00000028
#define IGB_DEFAULT_PBA 0x00000030
#define IGB_SMARTSPEED_DOWNSHIFT 3
#define IGB_SMARTSPEED_MAX 15
#define IGB_MAX_LOOP 10
#define IGB_RX_PTHRESH ((hw->mac.type == e1000_i354) ? 12 : \
((hw->mac.type <= e1000_82576) ? 16 : 8))
#define IGB_RX_HTHRESH 8
#define IGB_RX_WTHRESH ((hw->mac.type == e1000_82576 && \
adapter->msix_mem) ? 1 : 4)
#define IGB_TX_PTHRESH ((hw->mac.type == e1000_i354) ? 20 : 8)
#define IGB_TX_HTHRESH 1
#define IGB_TX_WTHRESH ((hw->mac.type != e1000_82575 && \
adapter->msix_mem) ? 1 : 16)
#define MAX_NUM_MULTICAST_ADDRESSES 128
#define PCI_ANY_ID (~0U)
#define ETHER_ALIGN 2
#define IGB_TX_BUFFER_SIZE ((uint32_t) 1514)
#define IGB_FC_PAUSE_TIME 0x0680
#define IGB_EEPROM_APME 0x400;
/* Queue minimum free for use */
#define IGB_QUEUE_THRESHOLD (adapter->num_tx_desc / 8)
/*
* TDBA/RDBA should be aligned on 16 byte boundary. But TDLEN/RDLEN should be
* multiple of 128 bytes. So we align TDBA/RDBA on 128 byte boundary. This will
* also optimize cache line size effect. H/W supports up to cache line size 128.
*/
#define IGB_DBA_ALIGN 128
#define SPEED_MODE_BIT (1<<21) /* On PCI-E MACs only */
/* PCI Config defines */
#define IGB_MSIX_BAR 3
/* Defines for printing debug information */
#define DEBUG_INIT 0
#define DEBUG_IOCTL 0
#define DEBUG_HW 0
#define INIT_DEBUGOUT(S) if (DEBUG_INIT) printf(S "\n")
#define INIT_DEBUGOUT1(S, A) if (DEBUG_INIT) printf(S "\n", A)
#define INIT_DEBUGOUT2(S, A, B) if (DEBUG_INIT) printf(S "\n", A, B)
#define IOCTL_DEBUGOUT(S) if (DEBUG_IOCTL) printf(S "\n")
#define IOCTL_DEBUGOUT1(S, A) if (DEBUG_IOCTL) printf(S "\n", A)
#define IOCTL_DEBUGOUT2(S, A, B) if (DEBUG_IOCTL) printf(S "\n", A, B)
#define HW_DEBUGOUT(S) if (DEBUG_HW) printf(S "\n")
#define HW_DEBUGOUT1(S, A) if (DEBUG_HW) printf(S "\n", A)
#define HW_DEBUGOUT2(S, A, B) if (DEBUG_HW) printf(S "\n", A, B)
#define IGB_MAX_SCATTER 40
#define IGB_VFTA_SIZE 128
#define IGB_BR_SIZE 4096 /* ring buf size */
#define IGB_TSO_SIZE (65535 + sizeof(struct ether_vlan_header))
#define IGB_TSO_SEG_SIZE 4096 /* Max dma segment size */
#define IGB_TXPBSIZE 20408
#define IGB_HDR_BUF 128
#define IGB_PKTTYPE_MASK 0x0000FFF0
#define IGB_DMCTLX_DCFLUSH_DIS 0x80000000 /* Disable DMA Coalesce Flush */
#define ETH_ZLEN 60
#define ETH_ADDR_LEN 6
/* Offload bits in mbuf flag */
#if __FreeBSD_version >= 1000000
#define CSUM_OFFLOAD_IPV4 (CSUM_IP|CSUM_IP_TCP|CSUM_IP_UDP|CSUM_IP_SCTP)
#define CSUM_OFFLOAD_IPV6 (CSUM_IP6_TCP|CSUM_IP6_UDP|CSUM_IP6_SCTP)
#define CSUM_OFFLOAD (CSUM_OFFLOAD_IPV4|CSUM_OFFLOAD_IPV6)
#elif __FreeBSD_version >= 800000
#define CSUM_OFFLOAD (CSUM_IP|CSUM_TCP|CSUM_UDP|CSUM_SCTP)
#else
#define CSUM_OFFLOAD (CSUM_IP|CSUM_TCP|CSUM_UDP)
#endif
/* Define the starting Interrupt rate per Queue */
#define IGB_INTS_PER_SEC 8000
#define IGB_DEFAULT_ITR ((1000000/IGB_INTS_PER_SEC) << 2)
#define IGB_LINK_ITR 2000
#define I210_LINK_DELAY 1000
/* Precision Time Sync (IEEE 1588) defines */
#define ETHERTYPE_IEEE1588 0x88F7
#define PICOSECS_PER_TICK 20833
#define TSYNC_PORT 319 /* UDP port for the protocol */
/*
* Bus dma allocation structure used by
* e1000_dma_malloc and e1000_dma_free.
*/
struct igb_dma_alloc {
bus_addr_t dma_paddr;
caddr_t dma_vaddr;
bus_dma_tag_t dma_tag;
bus_dmamap_t dma_map;
bus_dma_segment_t dma_seg;
int dma_nseg;
};
/*
** Driver queue struct: this is the interrupt container
** for the associated tx and rx ring.
*/
struct igb_queue {
struct adapter *adapter;
u32 msix; /* This queue's MSIX vector */
u32 eims; /* This queue's EIMS bit */
u32 eitr_setting;
struct resource *res;
void *tag;
struct tx_ring *txr;
struct rx_ring *rxr;
struct task que_task;
struct taskqueue *tq;
u64 irqs;
};
/*
* The transmit ring, one per queue
*/
struct tx_ring {
struct adapter *adapter;
struct mtx tx_mtx;
u32 me;
int watchdog_time;
union e1000_adv_tx_desc *tx_base;
struct igb_tx_buf *tx_buffers;
struct igb_dma_alloc txdma;
volatile u16 tx_avail;
u16 next_avail_desc;
u16 next_to_clean;
u16 num_desc;
enum {
IGB_QUEUE_IDLE = 1,
IGB_QUEUE_WORKING = 2,
IGB_QUEUE_HUNG = 4,
IGB_QUEUE_DEPLETED = 8,
} queue_status;
u32 txd_cmd;
bus_dma_tag_t txtag;
char mtx_name[16];
#ifndef IGB_LEGACY_TX
struct buf_ring *br;
struct task txq_task;
#endif
u32 bytes; /* used for AIM */
u32 packets;
/* Soft Stats */
unsigned long tso_tx;
unsigned long no_tx_map_avail;
unsigned long no_tx_dma_setup;
u64 no_desc_avail;
u64 total_packets;
};
/*
* Receive ring: one per queue
*/
struct rx_ring {
struct adapter *adapter;
u32 me;
struct igb_dma_alloc rxdma;
union e1000_adv_rx_desc *rx_base;
struct lro_ctrl lro;
bool lro_enabled;
bool hdr_split;
struct mtx rx_mtx;
char mtx_name[16];
u32 next_to_refresh;
u32 next_to_check;
struct igb_rx_buf *rx_buffers;
bus_dma_tag_t htag; /* dma tag for rx head */
bus_dma_tag_t ptag; /* dma tag for rx packet */
/*
* First/last mbuf pointers, for
* collecting multisegment RX packets.
*/
struct mbuf *fmp;
struct mbuf *lmp;
u32 bytes;
u32 packets;
int rdt;
int rdh;
/* Soft stats */
u64 rx_split_packets;
u64 rx_discarded;
u64 rx_packets;
u64 rx_bytes;
};
struct adapter {
struct ifnet *ifp;
struct e1000_hw hw;
struct e1000_osdep osdep;
device_t dev;
struct cdev *led_dev;
struct resource *pci_mem;
struct resource *msix_mem;
int memrid;
/*
* Interrupt resources: this set is
* either used for legacy, or for Link
* when doing MSIX
*/
void *tag;
struct resource *res;
struct ifmedia media;
struct callout timer;
int msix;
int if_flags;
int pause_frames;
struct mtx core_mtx;
eventhandler_tag vlan_attach;
eventhandler_tag vlan_detach;
u16 num_vlans;
u16 num_queues;
/*
** Shadow VFTA table, this is needed because
** the real vlan filter table gets cleared during
** a soft reset and the driver needs to be able
** to repopulate it.
*/
u32 shadow_vfta[IGB_VFTA_SIZE];
/* Info about the interface */
u32 optics;
u32 fc; /* local flow ctrl setting */
int advertise; /* link speeds */
bool link_active;
u16 max_frame_size;
u16 num_segs;
u16 link_speed;
bool link_up;
u32 linkvec;
u16 link_duplex;
u32 dmac;
int link_mask;
/* Flags */
u32 flags;
/* Mbuf cluster size */
u32 rx_mbuf_sz;
/* Support for pluggable optics */
bool sfp_probe;
struct task link_task; /* Link tasklet */
struct task mod_task; /* SFP tasklet */
struct task msf_task; /* Multispeed Fiber */
struct taskqueue *tq;
/*
** Queues:
** This is the irq holder, it has
** and RX/TX pair or rings associated
** with it.
*/
struct igb_queue *queues;
/*
* Transmit rings:
* Allocated at run time, an array of rings.
*/
struct tx_ring *tx_rings;
u32 num_tx_desc;
/*
* Receive rings:
* Allocated at run time, an array of rings.
*/
struct rx_ring *rx_rings;
u64 que_mask;
u32 num_rx_desc;
/* Multicast array memory */
u8 *mta;
/* Misc stats maintained by the driver */
unsigned long device_control;
unsigned long dropped_pkts;
unsigned long eint_mask;
unsigned long int_mask;
unsigned long link_irq;
unsigned long mbuf_defrag_failed;
unsigned long no_tx_dma_setup;
unsigned long packet_buf_alloc_rx;
unsigned long packet_buf_alloc_tx;
unsigned long rx_control;
unsigned long rx_overruns;
unsigned long watchdog_events;
/* Used in pf and vf */
void *stats;
int enable_aim;
int has_manage;
int wol;
int rx_process_limit;
int tx_process_limit;
u16 vf_ifp; /* a VF interface */
bool in_detach; /* Used only in igb_ioctl */
};
/* ******************************************************************************
* vendor_info_array
*
* This array contains the list of Subvendor/Subdevice IDs on which the driver
* should load.
*
* ******************************************************************************/
typedef struct _igb_vendor_info_t {
unsigned int vendor_id;
unsigned int device_id;
unsigned int subvendor_id;
unsigned int subdevice_id;
unsigned int index;
} igb_vendor_info_t;
struct igb_tx_buf {
union e1000_adv_tx_desc *eop;
struct mbuf *m_head;
bus_dmamap_t map;
};
struct igb_rx_buf {
struct mbuf *m_head;
struct mbuf *m_pack;
bus_dmamap_t hmap; /* bus_dma map for header */
bus_dmamap_t pmap; /* bus_dma map for packet */
};
/*
** Find the number of unrefreshed RX descriptors
*/
static inline u16
igb_rx_unrefreshed(struct rx_ring *rxr)
{
struct adapter *adapter = rxr->adapter;
if (rxr->next_to_check > rxr->next_to_refresh)
return (rxr->next_to_check - rxr->next_to_refresh - 1);
else
return ((adapter->num_rx_desc + rxr->next_to_check) -
rxr->next_to_refresh - 1);
}
#define IGB_CORE_LOCK_INIT(_sc, _name) \
mtx_init(&(_sc)->core_mtx, _name, "IGB Core Lock", MTX_DEF)
#define IGB_CORE_LOCK_DESTROY(_sc) mtx_destroy(&(_sc)->core_mtx)
#define IGB_CORE_LOCK(_sc) mtx_lock(&(_sc)->core_mtx)
#define IGB_CORE_UNLOCK(_sc) mtx_unlock(&(_sc)->core_mtx)
#define IGB_CORE_LOCK_ASSERT(_sc) mtx_assert(&(_sc)->core_mtx, MA_OWNED)
#define IGB_TX_LOCK_DESTROY(_sc) mtx_destroy(&(_sc)->tx_mtx)
#define IGB_TX_LOCK(_sc) mtx_lock(&(_sc)->tx_mtx)
#define IGB_TX_UNLOCK(_sc) mtx_unlock(&(_sc)->tx_mtx)
#define IGB_TX_TRYLOCK(_sc) mtx_trylock(&(_sc)->tx_mtx)
#define IGB_TX_LOCK_ASSERT(_sc) mtx_assert(&(_sc)->tx_mtx, MA_OWNED)
#define IGB_RX_LOCK_DESTROY(_sc) mtx_destroy(&(_sc)->rx_mtx)
#define IGB_RX_LOCK(_sc) mtx_lock(&(_sc)->rx_mtx)
#define IGB_RX_UNLOCK(_sc) mtx_unlock(&(_sc)->rx_mtx)
#define IGB_RX_LOCK_ASSERT(_sc) mtx_assert(&(_sc)->rx_mtx, MA_OWNED)
#define UPDATE_VF_REG(reg, last, cur) \
{ \
u32 new = E1000_READ_REG(hw, reg); \
if (new < last) \
cur += 0x100000000LL; \
last = new; \
cur &= 0xFFFFFFFF00000000LL; \
cur |= new; \
}
#if __FreeBSD_version >= 800000 && __FreeBSD_version < 800504
static __inline int
drbr_needs_enqueue(struct ifnet *ifp, struct buf_ring *br)
{
#ifdef ALTQ
if (ALTQ_IS_ENABLED(&ifp->if_snd))
return (1);
#endif
return (!buf_ring_empty(br));
}
#endif
#endif /* _IF_IGB_H_ */

4730
sys/dev/e1000/if_lem.c
View File

File diff suppressed because it is too large Load Diff

519
sys/dev/e1000/if_lem.h
View File

@ -1,519 +0,0 @@
/******************************************************************************
Copyright (c) 2001-2015, Intel Corporation
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.
3. Neither the name of the Intel Corporation nor the names of its
contributors may be used to endorse or promote products derived from
this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS 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 COPYRIGHT OWNER 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.
******************************************************************************/
/*$FreeBSD$*/
#ifndef _LEM_H_DEFINED_
#define _LEM_H_DEFINED_
/* Tunables */
/*
* EM_TXD: Maximum number of Transmit Descriptors
* Valid Range: 80-256 for 82542 and 82543-based adapters
* 80-4096 for others
* Default Value: 256
* This value is the number of transmit descriptors allocated by the driver.
* Increasing this value allows the driver to queue more transmits. Each
* descriptor is 16 bytes.
* Since TDLEN should be multiple of 128bytes, the number of transmit
* desscriptors should meet the following condition.
* (num_tx_desc * sizeof(struct e1000_tx_desc)) % 128 == 0
*/
#define EM_MIN_TXD 80
#define EM_MAX_TXD_82543 256
#define EM_MAX_TXD 4096
#define EM_DEFAULT_TXD EM_MAX_TXD_82543
/*
* EM_RXD - Maximum number of receive Descriptors
* Valid Range: 80-256 for 82542 and 82543-based adapters
* 80-4096 for others
* Default Value: 256
* This value is the number of receive descriptors allocated by the driver.
* Increasing this value allows the driver to buffer more incoming packets.
* Each descriptor is 16 bytes. A receive buffer is also allocated for each
* descriptor. The maximum MTU size is 16110.
* Since TDLEN should be multiple of 128bytes, the number of transmit
* desscriptors should meet the following condition.
* (num_tx_desc * sizeof(struct e1000_tx_desc)) % 128 == 0
*/
#define EM_MIN_RXD 80
#define EM_MAX_RXD_82543 256
#define EM_MAX_RXD 4096
#define EM_DEFAULT_RXD EM_MAX_RXD_82543
/*
* EM_TIDV - Transmit Interrupt Delay Value
* Valid Range: 0-65535 (0=off)
* Default Value: 64
* This value delays the generation of transmit interrupts in units of
* 1.024 microseconds. Transmit interrupt reduction can improve CPU
* efficiency if properly tuned for specific network traffic. If the
* system is reporting dropped transmits, this value may be set too high
* causing the driver to run out of available transmit descriptors.
*/
#define EM_TIDV 64
/*
* EM_TADV - Transmit Absolute Interrupt Delay Value
* (Not valid for 82542/82543/82544)
* Valid Range: 0-65535 (0=off)
* Default Value: 64
* This value, in units of 1.024 microseconds, limits the delay in which a
* transmit interrupt is generated. Useful only if EM_TIDV is non-zero,
* this value ensures that an interrupt is generated after the initial
* packet is sent on the wire within the set amount of time. Proper tuning,
* along with EM_TIDV, may improve traffic throughput in specific
* network conditions.
*/
#define EM_TADV 64
/*
* EM_RDTR - Receive Interrupt Delay Timer (Packet Timer)
* Valid Range: 0-65535 (0=off)
* Default Value: 0
* This value delays the generation of receive interrupts in units of 1.024
* microseconds. Receive interrupt reduction can improve CPU efficiency if
* properly tuned for specific network traffic. Increasing this value adds
* extra latency to frame reception and can end up decreasing the throughput
* of TCP traffic. If the system is reporting dropped receives, this value
* may be set too high, causing the driver to run out of available receive
* descriptors.
*
* CAUTION: When setting EM_RDTR to a value other than 0, adapters
* may hang (stop transmitting) under certain network conditions.
* If this occurs a WATCHDOG message is logged in the system
* event log. In addition, the controller is automatically reset,
* restoring the network connection. To eliminate the potential
* for the hang ensure that EM_RDTR is set to 0.
*/
#define EM_RDTR 0
/*
* Receive Interrupt Absolute Delay Timer (Not valid for 82542/82543/82544)
* Valid Range: 0-65535 (0=off)
* Default Value: 64
* This value, in units of 1.024 microseconds, limits the delay in which a
* receive interrupt is generated. Useful only if EM_RDTR is non-zero,
* this value ensures that an interrupt is generated after the initial
* packet is received within the set amount of time. Proper tuning,
* along with EM_RDTR, may improve traffic throughput in specific network
* conditions.
*/
#define EM_RADV 64
/*
* This parameter controls the max duration of transmit watchdog.
*/
#define EM_WATCHDOG (10 * hz)
/*
* This parameter controls when the driver calls the routine to reclaim
* transmit descriptors.
*/
#define EM_TX_CLEANUP_THRESHOLD (adapter->num_tx_desc / 8)
#define EM_TX_OP_THRESHOLD (adapter->num_tx_desc / 32)
/*
* This parameter controls whether or not autonegotation is enabled.
* 0 - Disable autonegotiation
* 1 - Enable autonegotiation
*/
#define DO_AUTO_NEG 1
/*
* This parameter control whether or not the driver will wait for
* autonegotiation to complete.
* 1 - Wait for autonegotiation to complete
* 0 - Don't wait for autonegotiation to complete
*/
#define WAIT_FOR_AUTO_NEG_DEFAULT 0
/* Tunables -- End */
#define AUTONEG_ADV_DEFAULT (ADVERTISE_10_HALF | ADVERTISE_10_FULL | \
ADVERTISE_100_HALF | ADVERTISE_100_FULL | \
ADVERTISE_1000_FULL)
#define AUTO_ALL_MODES 0
/* PHY master/slave setting */
#define EM_MASTER_SLAVE e1000_ms_hw_default
/*
* Micellaneous constants
*/
#define EM_VENDOR_ID 0x8086
#define EM_FLASH 0x0014
#define EM_JUMBO_PBA 0x00000028
#define EM_DEFAULT_PBA 0x00000030
#define EM_SMARTSPEED_DOWNSHIFT 3
#define EM_SMARTSPEED_MAX 15
#define EM_MAX_LOOP 10
#define MAX_NUM_MULTICAST_ADDRESSES 128
#define PCI_ANY_ID (~0U)
#define ETHER_ALIGN 2
#define EM_FC_PAUSE_TIME 0x0680
#define EM_EEPROM_APME 0x400;
#define EM_82544_APME 0x0004;
/* Code compatilbility between 6 and 7 */
#ifndef ETHER_BPF_MTAP
#define ETHER_BPF_MTAP BPF_MTAP
#endif
/*
* TDBA/RDBA should be aligned on 16 byte boundary. But TDLEN/RDLEN should be
* multiple of 128 bytes. So we align TDBA/RDBA on 128 byte boundary. This will
* also optimize cache line size effect. H/W supports up to cache line size 128.
*/
#define EM_DBA_ALIGN 128
#define SPEED_MODE_BIT (1<<21) /* On PCI-E MACs only */
/* PCI Config defines */
#define EM_BAR_TYPE(v) ((v) & EM_BAR_TYPE_MASK)
#define EM_BAR_TYPE_MASK 0x00000001
#define EM_BAR_TYPE_MMEM 0x00000000
#define EM_BAR_TYPE_IO 0x00000001
#define EM_BAR_TYPE_FLASH 0x0014
#define EM_BAR_MEM_TYPE(v) ((v) & EM_BAR_MEM_TYPE_MASK)
#define EM_BAR_MEM_TYPE_MASK 0x00000006
#define EM_BAR_MEM_TYPE_32BIT 0x00000000
#define EM_BAR_MEM_TYPE_64BIT 0x00000004
#define EM_MSIX_BAR 3 /* On 82575 */
#if __FreeBSD_version < 900000
#define SYSCTL_ADD_UQUAD SYSCTL_ADD_QUAD
#endif
/* Defines for printing debug information */
#define DEBUG_INIT 0
#define DEBUG_IOCTL 0
#define DEBUG_HW 0
#define INIT_DEBUGOUT(S) if (DEBUG_INIT) printf(S "\n")
#define INIT_DEBUGOUT1(S, A) if (DEBUG_INIT) printf(S "\n", A)
#define INIT_DEBUGOUT2(S, A, B) if (DEBUG_INIT) printf(S "\n", A, B)
#define IOCTL_DEBUGOUT(S) if (DEBUG_IOCTL) printf(S "\n")
#define IOCTL_DEBUGOUT1(S, A) if (DEBUG_IOCTL) printf(S "\n", A)
#define IOCTL_DEBUGOUT2(S, A, B) if (DEBUG_IOCTL) printf(S "\n", A, B)
#define HW_DEBUGOUT(S) if (DEBUG_HW) printf(S "\n")
#define HW_DEBUGOUT1(S, A) if (DEBUG_HW) printf(S "\n", A)
#define HW_DEBUGOUT2(S, A, B) if (DEBUG_HW) printf(S "\n", A, B)
#define EM_MAX_SCATTER 40
#define EM_VFTA_SIZE 128
#define EM_MSIX_MASK 0x01F00000 /* For 82574 use */
#define ETH_ZLEN 60
#define ETH_ADDR_LEN 6
#define CSUM_OFFLOAD 7 /* Offload bits in mbuf flag */
/*
* 82574 has a nonstandard address for EIAC
* and since its only used in MSIX, and in
* the em driver only 82574 uses MSIX we can
* solve it just using this define.
*/
#define EM_EIAC 0x000DC
/* Used in for 82547 10Mb Half workaround */
#define EM_PBA_BYTES_SHIFT 0xA
#define EM_TX_HEAD_ADDR_SHIFT 7
#define EM_PBA_TX_MASK 0xFFFF0000
#define EM_FIFO_HDR 0x10
#define EM_82547_PKT_THRESH 0x3e0
/* Precision Time Sync (IEEE 1588) defines */
#define ETHERTYPE_IEEE1588 0x88F7
#define PICOSECS_PER_TICK 20833
#define TSYNC_PORT 319 /* UDP port for the protocol */
#ifdef NIC_PARAVIRT
#define E1000_PARA_SUBDEV 0x1101 /* special id */
#define E1000_CSBAL 0x02830 /* csb phys. addr. low */
#define E1000_CSBAH 0x02834 /* csb phys. addr. hi */
#include <net/paravirt.h>
#endif /* NIC_PARAVIRT */
/*
* Bus dma allocation structure used by
* e1000_dma_malloc and e1000_dma_free.
*/
struct em_dma_alloc {
bus_addr_t dma_paddr;
caddr_t dma_vaddr;
bus_dma_tag_t dma_tag;
bus_dmamap_t dma_map;
bus_dma_segment_t dma_seg;
int dma_nseg;
};
struct adapter;
struct em_int_delay_info {
struct adapter *adapter; /* Back-pointer to the adapter struct */
int offset; /* Register offset to read/write */
int value; /* Current value in usecs */
};
/* Our adapter structure */
struct adapter {
if_t ifp;
struct e1000_hw hw;
/* FreeBSD operating-system-specific structures. */
struct e1000_osdep osdep;
device_t dev;
struct cdev *led_dev;
struct resource *memory;
struct resource *flash;
struct resource *msix;
struct resource *ioport;
int io_rid;
/* 82574 may use 3 int vectors */
struct resource *res[3];
void *tag[3];
int rid[3];
struct ifmedia media;
struct callout timer;
struct callout tx_fifo_timer;
bool watchdog_check;
int watchdog_time;
int msi;
int if_flags;
int max_frame_size;
int min_frame_size;
struct mtx core_mtx;
struct mtx tx_mtx;
struct mtx rx_mtx;
int em_insert_vlan_header;
/* Task for FAST handling */
struct task link_task;
struct task rxtx_task;
struct task rx_task;
struct task tx_task;
struct taskqueue *tq; /* private task queue */
eventhandler_tag vlan_attach;
eventhandler_tag vlan_detach;
u32 num_vlans;
/* Management and WOL features */
u32 wol;
bool has_manage;
bool has_amt;
/* Multicast array memory */
u8 *mta;
/*
** Shadow VFTA table, this is needed because
** the real vlan filter table gets cleared during
** a soft reset and the driver needs to be able
** to repopulate it.
*/
u32 shadow_vfta[EM_VFTA_SIZE];
/* Info about the interface */
uint8_t link_active;
uint16_t link_speed;
uint16_t link_duplex;
uint32_t smartspeed;
uint32_t fc_setting;
struct em_int_delay_info tx_int_delay;
struct em_int_delay_info tx_abs_int_delay;
struct em_int_delay_info rx_int_delay;
struct em_int_delay_info rx_abs_int_delay;
struct em_int_delay_info tx_itr;
/*
* Transmit definitions
*
* We have an array of num_tx_desc descriptors (handled
* by the controller) paired with an array of tx_buffers
* (at tx_buffer_area).
* The index of the next available descriptor is next_avail_tx_desc.
* The number of remaining tx_desc is num_tx_desc_avail.
*/
struct em_dma_alloc txdma; /* bus_dma glue for tx desc */
struct e1000_tx_desc *tx_desc_base;
uint32_t next_avail_tx_desc;
uint32_t next_tx_to_clean;
volatile uint16_t num_tx_desc_avail;
uint16_t num_tx_desc;
uint16_t last_hw_offload;
uint32_t txd_cmd;
struct em_buffer *tx_buffer_area;
bus_dma_tag_t txtag; /* dma tag for tx */
uint32_t tx_tso; /* last tx was tso */
/*
* Receive definitions
*
* we have an array of num_rx_desc rx_desc (handled by the
* controller), and paired with an array of rx_buffers
* (at rx_buffer_area).
* The next pair to check on receive is at offset next_rx_desc_to_check
*/
struct em_dma_alloc rxdma; /* bus_dma glue for rx desc */
struct e1000_rx_desc *rx_desc_base;
uint32_t next_rx_desc_to_check;
uint32_t rx_buffer_len;
uint16_t num_rx_desc;
int rx_process_limit;
struct em_buffer *rx_buffer_area;
bus_dma_tag_t rxtag;
bus_dmamap_t rx_sparemap;
/*
* First/last mbuf pointers, for
* collecting multisegment RX packets.
*/
struct mbuf *fmp;
struct mbuf *lmp;
/* Misc stats maintained by the driver */
unsigned long dropped_pkts;
unsigned long link_irq;
unsigned long mbuf_cluster_failed;
unsigned long mbuf_defrag_failed;
unsigned long no_tx_desc_avail1;
unsigned long no_tx_desc_avail2;
unsigned long no_tx_dma_setup;
unsigned long no_tx_map_avail;
unsigned long watchdog_events;
unsigned long rx_irq;
unsigned long rx_overruns;
unsigned long tx_irq;
/* 82547 workaround */
uint32_t tx_fifo_size;
uint32_t tx_fifo_head;
uint32_t tx_fifo_head_addr;
uint64_t tx_fifo_reset_cnt;
uint64_t tx_fifo_wrk_cnt;
uint32_t tx_head_addr;
/* For 82544 PCIX Workaround */
boolean_t pcix_82544;
boolean_t in_detach;
#ifdef NIC_SEND_COMBINING
/* 0 = idle; 1xxxx int-pending; 3xxxx int + d pending + tdt */
#define MIT_PENDING_INT 0x10000 /* pending interrupt */
#define MIT_PENDING_TDT 0x30000 /* both intr and tdt write are pending */
uint32_t shadow_tdt;
uint32_t sc_enable;
#endif /* NIC_SEND_COMBINING */
#ifdef BATCH_DISPATCH
uint32_t batch_enable;
#endif /* BATCH_DISPATCH */
#ifdef NIC_PARAVIRT
struct em_dma_alloc csb_mem; /* phys address */
struct paravirt_csb *csb; /* virtual addr */
uint32_t rx_retries; /* optimize rx loop */
uint32_t tdt_csb_count;// XXX stat
uint32_t tdt_reg_count;// XXX stat
uint32_t tdt_int_count;// XXX stat
uint32_t guest_need_kick_count;// XXX stat
#endif /* NIC_PARAVIRT */
struct e1000_hw_stats stats;
};
/* ******************************************************************************
* vendor_info_array
*
* This array contains the list of Subvendor/Subdevice IDs on which the driver
* should load.
*
* ******************************************************************************/
typedef struct _em_vendor_info_t {
unsigned int vendor_id;
unsigned int device_id;
unsigned int subvendor_id;
unsigned int subdevice_id;
unsigned int index;
} em_vendor_info_t;
struct em_buffer {
int next_eop; /* Index of the desc to watch */
struct mbuf *m_head;
bus_dmamap_t map; /* bus_dma map for packet */
};
/* For 82544 PCIX Workaround */
typedef struct _ADDRESS_LENGTH_PAIR
{
uint64_t address;
uint32_t length;
} ADDRESS_LENGTH_PAIR, *PADDRESS_LENGTH_PAIR;
typedef struct _DESCRIPTOR_PAIR
{
ADDRESS_LENGTH_PAIR descriptor[4];
uint32_t elements;
} DESC_ARRAY, *PDESC_ARRAY;
#define EM_CORE_LOCK_INIT(_sc, _name) \
mtx_init(&(_sc)->core_mtx, _name, "EM Core Lock", MTX_DEF)
#define EM_TX_LOCK_INIT(_sc, _name) \
mtx_init(&(_sc)->tx_mtx, _name, "EM TX Lock", MTX_DEF)
#define EM_RX_LOCK_INIT(_sc, _name) \
mtx_init(&(_sc)->rx_mtx, _name, "EM RX Lock", MTX_DEF)
#define EM_CORE_LOCK_DESTROY(_sc) mtx_destroy(&(_sc)->core_mtx)
#define EM_TX_LOCK_DESTROY(_sc) mtx_destroy(&(_sc)->tx_mtx)
#define EM_RX_LOCK_DESTROY(_sc) mtx_destroy(&(_sc)->rx_mtx)
#define EM_CORE_LOCK(_sc) mtx_lock(&(_sc)->core_mtx)
#define EM_TX_LOCK(_sc) mtx_lock(&(_sc)->tx_mtx)
#define EM_TX_TRYLOCK(_sc) mtx_trylock(&(_sc)->tx_mtx)
#define EM_RX_LOCK(_sc) mtx_lock(&(_sc)->rx_mtx)
#define EM_CORE_UNLOCK(_sc) mtx_unlock(&(_sc)->core_mtx)
#define EM_TX_UNLOCK(_sc) mtx_unlock(&(_sc)->tx_mtx)
#define EM_RX_UNLOCK(_sc) mtx_unlock(&(_sc)->rx_mtx)
#define EM_CORE_LOCK_ASSERT(_sc) mtx_assert(&(_sc)->core_mtx, MA_OWNED)
#define EM_TX_LOCK_ASSERT(_sc) mtx_assert(&(_sc)->tx_mtx, MA_OWNED)
#endif /* _LEM_H_DEFINED_ */

594
sys/dev/e1000/igb_txrx.c Normal file
View File

@ -0,0 +1,594 @@
/* $FreeBSD$ */
#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, uint32_t pidx);
static int igb_isc_txd_credits_update(void *arg, uint16_t txqid, uint32_t cidx, bool clear);
static void igb_isc_rxd_refill(void *arg, uint16_t rxqid, uint8_t flid __unused,
uint32_t pidx, uint64_t *paddrs, caddr_t *vaddrs __unused, uint16_t count, uint16_t buf_len __unused);
static void igb_isc_rxd_flush(void *arg, uint16_t rxqid, uint8_t flid __unused, uint32_t pidx);
static int igb_isc_rxd_available(void *arg, uint16_t rxqid, uint32_t idx,
int 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 & 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 __FreeBSD_version >= 1000000
if (pi->ipi_csum_flags & (CSUM_IP_TCP | CSUM_IP6_TCP))
#else
if (pi->ipi_csum_flags & CSUM_TCP)
#endif
type_tucmd_mlhl |= E1000_ADVTXD_TUCMD_L4T_TCP;
break;
case IPPROTO_UDP:
#if __FreeBSD_version >= 1000000
if (pi->ipi_csum_flags & (CSUM_IP_UDP | CSUM_IP6_UDP))
#else
if (pi->ipi_csum_flags & CSUM_UDP)
#endif
type_tucmd_mlhl |= E1000_ADVTXD_TUCMD_L4T_UDP;
break;
#if __FreeBSD_version >= 800000
case IPPROTO_SCTP:
#if __FreeBSD_version >= 1000000
if (pi->ipi_csum_flags & (CSUM_IP_SCTP | CSUM_IP6_SCTP))
#else
if (pi->ipi_csum_flags & CSUM_SCTP)
#endif
type_tucmd_mlhl |= E1000_ADVTXD_TUCMD_L4T_SCTP;
break;
#endif
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;
struct em_txbuffer *txbuf;
union e1000_adv_tx_desc *txd = NULL;
int i, j, first, pidx_last;
u32 olinfo_status, cmd_type_len;
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;
/* 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;
txbuf = &txr->tx_buffers[i];
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;
}
}
txd->read.cmd_type_len |=
htole32(E1000_TXD_CMD_EOP | E1000_TXD_CMD_RS);
/* Set the EOP descriptor that will be marked done */
txbuf = &txr->tx_buffers[first];
txbuf->eop = pidx_last;
pi->ipi_new_pidx = i;
return (0);
}
static void
igb_isc_txd_flush(void *arg, uint16_t txqid, uint32_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, uint32_t cidx_init, 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;
u32 cidx, ntxd, processed = 0;
struct em_txbuffer *buf;
union e1000_adv_tx_desc *txd, *eop;
int limit;
cidx = cidx_init;
buf = &txr->tx_buffers[cidx];
txd = (union e1000_adv_tx_desc *)&txr->tx_base[cidx];
ntxd = scctx->isc_ntxd[0];
limit = adapter->tx_process_limit;
do {
if (buf->eop == -1) /* No work */
break;
eop = (union e1000_adv_tx_desc *)&txr->tx_base[buf->eop];
if ((eop->wb.status & E1000_TXD_STAT_DD) == 0)
break; /* I/O not complete */
if (clear)
buf->eop = -1; /* clear indicate processed */
/* We clean the range if multi segment */
while (txd != eop) {
++txd;
++buf;
/* wrap the ring? */
if (++cidx == scctx->isc_ntxd[0]) {
cidx = 0;
buf = txr->tx_buffers;
txd = (union e1000_adv_tx_desc *)txr->tx_base;
}
buf = &txr->tx_buffers[cidx];
if (clear)
buf->eop = -1;
processed++;
}
processed++;
/* Try the next packet */
txd++;
buf++;
/* reset with a wrap */
if (++cidx == scctx->isc_ntxd[0]) {
cidx = 0;
buf = txr->tx_buffers;
txd = (union e1000_adv_tx_desc *)txr->tx_base;
}
prefetch(txd);
prefetch(txd+1);
} while (__predict_true(--limit) && cidx != cidx_init);
return (processed);
}
static void
igb_isc_rxd_refill(void *arg, uint16_t rxqid, uint8_t flid __unused,
uint32_t pidx, uint64_t *paddrs, caddr_t *vaddrs __unused,
uint16_t count, uint16_t buf_len __unused)
{
struct adapter *sc = arg;
if_softc_ctx_t scctx = sc->shared;
struct em_rx_queue *que = &sc->rx_queues[rxqid];
union e1000_adv_rx_desc *rxd;
struct rx_ring *rxr = &que->rxr;
int i;
uint32_t next_pidx;
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, uint32_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, uint32_t idx, int 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;
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++;
}
{
struct e1000_hw *hw = &sc->hw;
int rdt, rdh;
rdt = E1000_READ_REG(hw, E1000_RDT(rxr->me));
rdh = E1000_READ_REG(hw, E1000_RDH(rxr->me));
DPRINTF(iflib_get_dev(sc->ctx), "sidx:%d eidx:%d iter=%d pktcnt=%d RDT=%d RDH=%d\n", idx, i, iter, cnt, rdt, rdh);
}
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 __FreeBSD_version >= 800000
if (sctp) /* reassign */
type = CSUM_SCTP_VALID;
#endif
/* 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;
}
}

1
sys/i386/conf/GENERIC
View File

@ -235,7 +235,6 @@ device puc # Multi I/O cards and multi-channel UARTs
device bxe # Broadcom NetXtreme II BCM5771X/BCM578XX 10GbE
device de # DEC/Intel DC21x4x (``Tulip'')
device em # Intel PRO/1000 Gigabit Ethernet Family
device igb # Intel PRO/1000 PCIE Server Gigabit Family
device ixgb # Intel PRO/10GbE Ethernet Card
device le # AMD Am7900 LANCE and Am79C9xx PCnet
device ti # Alteon Networks Tigon I/II gigabit Ethernet

1
sys/mips/conf/OCTEON1
View File

@ -195,7 +195,6 @@ device octm
# PCI Ethernet NICs.
device de # DEC/Intel DC21x4x (``Tulip'')
device em # Intel PRO/1000 Gigabit Ethernet Family
device igb # Intel PRO/1000 PCIE Server Gigabit Family
device ix # Intel PRO/10GbE PF PCIE Ethernet Family
device ixv # Intel PRO/10GbE VF PCIE Ethernet Family
device le # AMD Am7900 LANCE and Am79C9xx PCnet

4
sys/modules/Makefile
View File

@ -163,7 +163,6 @@ SUBDIR= \
if_tun \
if_vlan \
if_vxlan \
${_igb} \
${_iir} \
imgact_binmisc \
${_intelspi} \
@ -544,7 +543,6 @@ _cxgb= cxgb
.if ${MACHINE_CPUARCH} == "aarch64"
_armv8crypto= armv8crypto
_em= em
_igb= igb
.endif
.if ${MACHINE_CPUARCH} == "i386" || ${MACHINE_CPUARCH} == "amd64"
@ -572,7 +570,6 @@ _fe= fe
_ibcore= ibcore
.endif
_if_ndis= if_ndis
_igb= igb
_io= io
.if ${MK_OFED} != "no" || defined(ALL_MODULES)
_ipoib= ipoib
@ -779,7 +776,6 @@ _nvram= powermac_nvram
_auxio= auxio
_em= em
_epic= epic
_igb= igb
.endif
.if (${MACHINE_CPUARCH} == "amd64" || ${MACHINE_ARCH} == "armv6" || \

13
sys/modules/em/Makefile
View File

@ -1,23 +1,28 @@
# $FreeBSD$
.PATH: ${.CURDIR}/../../dev/e1000
KMOD = if_em
SRCS = device_if.h bus_if.h pci_if.h opt_ddb.h opt_em.h opt_inet.h \
opt_inet6.h
opt_inet6.h ifdi_if.h
SRCS += $(CORE_SRC) $(LEGACY_SRC)
SRCS += $(COMMON_SHARED) $(LEGACY_SHARED) $(PCIE_SHARED)
CORE_SRC = if_em.c e1000_osdep.c
CORE_SRC = if_em.c em_txrx.c e1000_osdep.c
CORE_SRC += igb_txrx.c
# This is the Legacy, pre-PCIE source, it can be
# undefined when using modular driver if not needed
LEGACY_SRC += if_lem.c
COMMON_SHARED = e1000_api.c e1000_phy.c e1000_nvm.c e1000_mac.c \
e1000_manage.c e1000_vf.c e1000_mbx.c e1000_i210.c
PCIE_SHARED = e1000_80003es2lan.c e1000_ich8lan.c e1000_82571.c e1000_82575.c
LEGACY_SHARED = e1000_82540.c e1000_82542.c e1000_82541.c e1000_82543.c
CFLAGS += -I${.CURDIR}/../../dev/e1000
CFLAGS += -I${.CURDIR}/../../../dev/e1000
# DEVICE_POLLING for a non-interrupt-driven method
#CFLAGS += -DDEVICE_POLLING
afterinstall:
ln -sf ${DESTDIR}${KMODDIR}/${KMOD}.ko ${DESTDIR}${KMODDIR}/if_igb.ko
.include <bsd.kmod.mk>

24
sys/modules/igb/Makefile
View File

@ -1,24 +0,0 @@
#$FreeBSD$
.PATH: ${.CURDIR}/../../dev/e1000
KMOD = if_igb
SRCS = device_if.h bus_if.h pci_if.h opt_inet.h opt_inet6.h opt_rss.h
SRCS += if_igb.c $(SHARED_SRCS)
SHARED_SRCS = e1000_api.c e1000_phy.c e1000_nvm.c e1000_mac.c e1000_manage.c
SHARED_SRCS += e1000_80003es2lan.c e1000_82542.c e1000_82541.c e1000_82543.c
SHARED_SRCS += e1000_82540.c e1000_ich8lan.c e1000_82571.c e1000_osdep.c
SHARED_SRCS += e1000_82575.c e1000_vf.c e1000_mbx.c e1000_i210.c
CFLAGS += -I${.CURDIR}/../../dev/e1000 -DSMP
# DEVICE_POLLING gives you non-interrupt handling
# not advisable since MSIX gives better results
#CFLAGS += -DDEVICE_POLLING
# IGB_LEGACY_TX will override the stack if_transmit path and
# instead use the older if_start non-multiqueue capable interface.
# This might be desirable for testing, or to enable the use of
# ALTQ.
#CFLAGS += -DIGB_LEGACY_TX
.include <bsd.kmod.mk>

1
sys/powerpc/conf/GENERIC64
View File

@ -139,7 +139,6 @@ device uart_z8530
# Ethernet hardware
device em # Intel PRO/1000 Gigabit Ethernet Family
device igb # Intel PRO/1000 PCIE Server Gigabit Family
device ix # Intel PRO/10GbE PCIE PF Ethernet Family
device ixv # Intel PRO/10GbE PCIE VF Ethernet Family
device glc # Sony Playstation 3 Ethernet