freebsd-nq/sys/powerpc/pseries/phyp_llan.c
Matt Macy d7c5a620e2 ifnet: Replace if_addr_lock rwlock with epoch + mutex
Run on LLNW canaries and tested by pho@

gallatin:
Using a 14-core, 28-HTT single socket E5-2697 v3 with a 40GbE MLX5
based ConnectX 4-LX NIC, I see an almost 12% improvement in received
packet rate, and a larger improvement in bytes delivered all the way
to userspace.

When the host receiving 64 streams of netperf -H $DUT -t UDP_STREAM -- -m 1,
I see, using nstat -I mce0 1 before the patch:

InMpps OMpps  InGbs  OGbs err TCP Est %CPU syscalls csw     irq GBfree
4.98   0.00   4.42   0.00 4235592     33   83.80 4720653 2149771   1235 247.32
4.73   0.00   4.20   0.00 4025260     33   82.99 4724900 2139833   1204 247.32
4.72   0.00   4.20   0.00 4035252     33   82.14 4719162 2132023   1264 247.32
4.71   0.00   4.21   0.00 4073206     33   83.68 4744973 2123317   1347 247.32
4.72   0.00   4.21   0.00 4061118     33   80.82 4713615 2188091   1490 247.32
4.72   0.00   4.21   0.00 4051675     33   85.29 4727399 2109011   1205 247.32
4.73   0.00   4.21   0.00 4039056     33   84.65 4724735 2102603   1053 247.32

After the patch

InMpps OMpps  InGbs  OGbs err TCP Est %CPU syscalls csw     irq GBfree
5.43   0.00   4.20   0.00 3313143     33   84.96 5434214 1900162   2656 245.51
5.43   0.00   4.20   0.00 3308527     33   85.24 5439695 1809382   2521 245.51
5.42   0.00   4.19   0.00 3316778     33   87.54 5416028 1805835   2256 245.51
5.42   0.00   4.19   0.00 3317673     33   90.44 5426044 1763056   2332 245.51
5.42   0.00   4.19   0.00 3314839     33   88.11 5435732 1792218   2499 245.52
5.44   0.00   4.19   0.00 3293228     33   91.84 5426301 1668597   2121 245.52

Similarly, netperf reports 230Mb/s before the patch, and 270Mb/s after the patch

Reviewed by:	gallatin
Sponsored by:	Limelight Networks
Differential Revision:	https://reviews.freebsd.org/D15366
2018-05-18 20:13:34 +00:00

553 lines
14 KiB
C

/*-
* SPDX-License-Identifier: BSD-2-Clause-FreeBSD
*
* Copyright 2013 Nathan Whitehorn
* 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.
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/sockio.h>
#include <sys/endian.h>
#include <sys/lock.h>
#include <sys/mbuf.h>
#include <sys/module.h>
#include <sys/malloc.h>
#include <sys/mutex.h>
#include <sys/kernel.h>
#include <sys/socket.h>
#include <net/bpf.h>
#include <net/if.h>
#include <net/if_var.h>
#include <net/ethernet.h>
#include <net/if_dl.h>
#include <net/if_media.h>
#include <net/if_types.h>
#include <dev/ofw/openfirm.h>
#include <dev/ofw/ofw_bus.h>
#include <dev/ofw/ofw_bus_subr.h>
#include <machine/bus.h>
#include <machine/resource.h>
#include <sys/bus.h>
#include <sys/rman.h>
#include <powerpc/pseries/phyp-hvcall.h>
#define LLAN_MAX_RX_PACKETS 100
#define LLAN_MAX_TX_PACKETS 100
#define LLAN_RX_BUF_LEN 8*PAGE_SIZE
#define LLAN_BUFDESC_VALID (1ULL << 63)
#define LLAN_ADD_MULTICAST 0x1
#define LLAN_DEL_MULTICAST 0x2
#define LLAN_CLEAR_MULTICAST 0x3
struct llan_xfer {
struct mbuf *rx_mbuf;
bus_dmamap_t rx_dmamap;
uint64_t rx_bufdesc;
};
struct llan_receive_queue_entry { /* PAPR page 539 */
uint8_t control;
uint8_t reserved;
uint16_t offset;
uint32_t length;
uint64_t handle;
} __packed;
struct llan_softc {
device_t dev;
struct mtx io_lock;
cell_t unit;
uint8_t mac_address[8];
struct ifmedia media;
int irqid;
struct resource *irq;
void *irq_cookie;
bus_dma_tag_t rx_dma_tag;
bus_dma_tag_t rxbuf_dma_tag;
bus_dma_tag_t tx_dma_tag;
bus_dmamap_t tx_dma_map;
struct llan_receive_queue_entry *rx_buf;
int rx_dma_slot;
int rx_valid_val;
bus_dmamap_t rx_buf_map;
bus_addr_t rx_buf_phys;
bus_size_t rx_buf_len;
bus_addr_t input_buf_phys;
bus_addr_t filter_buf_phys;
struct llan_xfer rx_xfer[LLAN_MAX_RX_PACKETS];
struct ifnet *ifp;
};
static int llan_probe(device_t);
static int llan_attach(device_t);
static void llan_intr(void *xsc);
static void llan_init(void *xsc);
static void llan_start(struct ifnet *ifp);
static int llan_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data);
static void llan_media_status(struct ifnet *ifp, struct ifmediareq *ifmr);
static int llan_media_change(struct ifnet *ifp);
static void llan_rx_load_cb(void *xsc, bus_dma_segment_t *segs, int nsegs,
int err);
static int llan_add_rxbuf(struct llan_softc *sc, struct llan_xfer *rx);
static int llan_set_multicast(struct llan_softc *sc);
static devclass_t llan_devclass;
static device_method_t llan_methods[] = {
DEVMETHOD(device_probe, llan_probe),
DEVMETHOD(device_attach, llan_attach),
DEVMETHOD_END
};
static driver_t llan_driver = {
"llan",
llan_methods,
sizeof(struct llan_softc)
};
DRIVER_MODULE(llan, vdevice, llan_driver, llan_devclass, 0, 0);
static int
llan_probe(device_t dev)
{
if (!ofw_bus_is_compatible(dev,"IBM,l-lan"))
return (ENXIO);
device_set_desc(dev, "POWER Hypervisor Virtual Ethernet");
return (0);
}
static int
llan_attach(device_t dev)
{
struct llan_softc *sc;
phandle_t node;
int error, i;
sc = device_get_softc(dev);
sc->dev = dev;
/* Get firmware properties */
node = ofw_bus_get_node(dev);
OF_getprop(node, "local-mac-address", sc->mac_address,
sizeof(sc->mac_address));
OF_getencprop(node, "reg", &sc->unit, sizeof(sc->unit));
mtx_init(&sc->io_lock, "llan", NULL, MTX_DEF);
/* Setup interrupt */
sc->irqid = 0;
sc->irq = bus_alloc_resource_any(dev, SYS_RES_IRQ, &sc->irqid,
RF_ACTIVE);
if (!sc->irq) {
device_printf(dev, "Could not allocate IRQ\n");
mtx_destroy(&sc->io_lock);
return (ENXIO);
}
bus_setup_intr(dev, sc->irq, INTR_TYPE_MISC | INTR_MPSAFE |
INTR_ENTROPY, NULL, llan_intr, sc, &sc->irq_cookie);
/* Setup DMA */
error = bus_dma_tag_create(bus_get_dma_tag(dev), 16, 0,
BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL, NULL,
LLAN_RX_BUF_LEN, 1, BUS_SPACE_MAXSIZE_32BIT,
0, NULL, NULL, &sc->rx_dma_tag);
error = bus_dma_tag_create(bus_get_dma_tag(dev), 4, 0,
BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL, NULL,
BUS_SPACE_MAXSIZE, 1, BUS_SPACE_MAXSIZE_32BIT,
0, NULL, NULL, &sc->rxbuf_dma_tag);
error = bus_dma_tag_create(bus_get_dma_tag(dev), 1, 0,
BUS_SPACE_MAXADDR_32BIT, BUS_SPACE_MAXADDR, NULL, NULL,
BUS_SPACE_MAXSIZE, 6, BUS_SPACE_MAXSIZE_32BIT, 0,
busdma_lock_mutex, &sc->io_lock, &sc->tx_dma_tag);
error = bus_dmamem_alloc(sc->rx_dma_tag, (void **)&sc->rx_buf,
BUS_DMA_WAITOK | BUS_DMA_ZERO, &sc->rx_buf_map);
error = bus_dmamap_load(sc->rx_dma_tag, sc->rx_buf_map, sc->rx_buf,
LLAN_RX_BUF_LEN, llan_rx_load_cb, sc, 0);
/* TX DMA maps */
bus_dmamap_create(sc->tx_dma_tag, 0, &sc->tx_dma_map);
/* RX DMA */
for (i = 0; i < LLAN_MAX_RX_PACKETS; i++) {
error = bus_dmamap_create(sc->rxbuf_dma_tag, 0,
&sc->rx_xfer[i].rx_dmamap);
sc->rx_xfer[i].rx_mbuf = NULL;
}
/* Attach to network stack */
sc->ifp = if_alloc(IFT_ETHER);
sc->ifp->if_softc = sc;
if_initname(sc->ifp, device_get_name(dev), device_get_unit(dev));
sc->ifp->if_mtu = ETHERMTU; /* XXX max-frame-size from OF? */
sc->ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
sc->ifp->if_hwassist = 0; /* XXX: ibm,illan-options */
sc->ifp->if_capabilities = 0;
sc->ifp->if_capenable = 0;
sc->ifp->if_start = llan_start;
sc->ifp->if_ioctl = llan_ioctl;
sc->ifp->if_init = llan_init;
ifmedia_init(&sc->media, IFM_IMASK, llan_media_change,
llan_media_status);
ifmedia_add(&sc->media, IFM_ETHER | IFM_AUTO, 0, NULL);
ifmedia_set(&sc->media, IFM_ETHER | IFM_AUTO);
IFQ_SET_MAXLEN(&sc->ifp->if_snd, LLAN_MAX_TX_PACKETS);
sc->ifp->if_snd.ifq_drv_maxlen = LLAN_MAX_TX_PACKETS;
IFQ_SET_READY(&sc->ifp->if_snd);
ether_ifattach(sc->ifp, &sc->mac_address[2]);
/* We don't have link state reporting, so make it always up */
if_link_state_change(sc->ifp, LINK_STATE_UP);
return (0);
}
static int
llan_media_change(struct ifnet *ifp)
{
struct llan_softc *sc = ifp->if_softc;
if (IFM_TYPE(sc->media.ifm_media) != IFM_ETHER)
return (EINVAL);
if (IFM_SUBTYPE(sc->media.ifm_media) != IFM_AUTO)
return (EINVAL);
return (0);
}
static void
llan_media_status(struct ifnet *ifp, struct ifmediareq *ifmr)
{
ifmr->ifm_status = IFM_AVALID | IFM_ACTIVE | IFM_UNKNOWN | IFM_FDX;
ifmr->ifm_active = IFM_ETHER;
}
static void
llan_rx_load_cb(void *xsc, bus_dma_segment_t *segs, int nsegs, int err)
{
struct llan_softc *sc = xsc;
sc->rx_buf_phys = segs[0].ds_addr;
sc->rx_buf_len = segs[0].ds_len - 2*PAGE_SIZE;
sc->input_buf_phys = segs[0].ds_addr + segs[0].ds_len - PAGE_SIZE;
sc->filter_buf_phys = segs[0].ds_addr + segs[0].ds_len - 2*PAGE_SIZE;
}
static void
llan_init(void *xsc)
{
struct llan_softc *sc = xsc;
uint64_t rx_buf_desc;
uint64_t macaddr;
int err, i;
mtx_lock(&sc->io_lock);
phyp_hcall(H_FREE_LOGICAL_LAN, sc->unit);
/* Create buffers (page 539) */
sc->rx_dma_slot = 0;
sc->rx_valid_val = 1;
rx_buf_desc = LLAN_BUFDESC_VALID;
rx_buf_desc |= (sc->rx_buf_len << 32);
rx_buf_desc |= sc->rx_buf_phys;
memcpy(&macaddr, sc->mac_address, 8);
err = phyp_hcall(H_REGISTER_LOGICAL_LAN, sc->unit, sc->input_buf_phys,
rx_buf_desc, sc->filter_buf_phys, macaddr);
for (i = 0; i < LLAN_MAX_RX_PACKETS; i++)
llan_add_rxbuf(sc, &sc->rx_xfer[i]);
phyp_hcall(H_VIO_SIGNAL, sc->unit, 1); /* Enable interrupts */
/* Tell stack we're up */
sc->ifp->if_drv_flags |= IFF_DRV_RUNNING;
sc->ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
mtx_unlock(&sc->io_lock);
/* Check for pending receives scheduled before interrupt enable */
llan_intr(sc);
}
static int
llan_add_rxbuf(struct llan_softc *sc, struct llan_xfer *rx)
{
struct mbuf *m;
bus_dma_segment_t segs[1];
int error, nsegs;
mtx_assert(&sc->io_lock, MA_OWNED);
m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
if (m == NULL)
return (ENOBUFS);
m->m_len = m->m_pkthdr.len = m->m_ext.ext_size;
if (rx->rx_mbuf != NULL) {
bus_dmamap_sync(sc->rxbuf_dma_tag, rx->rx_dmamap,
BUS_DMASYNC_POSTREAD);
bus_dmamap_unload(sc->rxbuf_dma_tag, rx->rx_dmamap);
}
/* Save pointer to buffer structure */
m_copyback(m, 0, 8, (void *)&rx);
error = bus_dmamap_load_mbuf_sg(sc->rxbuf_dma_tag, rx->rx_dmamap, m,
segs, &nsegs, BUS_DMA_NOWAIT);
if (error != 0) {
device_printf(sc->dev,
"cannot load RX DMA map %p, error = %d\n", rx, error);
m_freem(m);
return (error);
}
/* If nsegs is wrong then the stack is corrupt. */
KASSERT(nsegs == 1,
("%s: too many DMA segments (%d)", __func__, nsegs));
rx->rx_mbuf = m;
bus_dmamap_sync(sc->rxbuf_dma_tag, rx->rx_dmamap, BUS_DMASYNC_PREREAD);
rx->rx_bufdesc = LLAN_BUFDESC_VALID;
rx->rx_bufdesc |= (((uint64_t)segs[0].ds_len) << 32);
rx->rx_bufdesc |= segs[0].ds_addr;
error = phyp_hcall(H_ADD_LOGICAL_LAN_BUFFER, sc->unit, rx->rx_bufdesc);
if (error != 0) {
m_freem(m);
rx->rx_mbuf = NULL;
return (ENOBUFS);
}
return (0);
}
static void
llan_intr(void *xsc)
{
struct llan_softc *sc = xsc;
struct llan_xfer *rx;
struct mbuf *m;
mtx_lock(&sc->io_lock);
restart:
phyp_hcall(H_VIO_SIGNAL, sc->unit, 0);
while ((sc->rx_buf[sc->rx_dma_slot].control >> 7) == sc->rx_valid_val) {
rx = (struct llan_xfer *)sc->rx_buf[sc->rx_dma_slot].handle;
m = rx->rx_mbuf;
m_adj(m, sc->rx_buf[sc->rx_dma_slot].offset - 8);
m->m_len = sc->rx_buf[sc->rx_dma_slot].length;
/* llan_add_rxbuf does DMA sync and unload as well as requeue */
if (llan_add_rxbuf(sc, rx) != 0) {
if_inc_counter(sc->ifp, IFCOUNTER_IERRORS, 1);
phyp_hcall(H_ADD_LOGICAL_LAN_BUFFER, sc->unit,
rx->rx_bufdesc);
continue;
}
if_inc_counter(sc->ifp, IFCOUNTER_IPACKETS, 1);
m_adj(m, sc->rx_buf[sc->rx_dma_slot].offset);
m->m_len = sc->rx_buf[sc->rx_dma_slot].length;
m->m_pkthdr.rcvif = sc->ifp;
m->m_pkthdr.len = m->m_len;
sc->rx_dma_slot++;
if (sc->rx_dma_slot >= sc->rx_buf_len/sizeof(sc->rx_buf[0])) {
sc->rx_dma_slot = 0;
sc->rx_valid_val = !sc->rx_valid_val;
}
mtx_unlock(&sc->io_lock);
(*sc->ifp->if_input)(sc->ifp, m);
mtx_lock(&sc->io_lock);
}
phyp_hcall(H_VIO_SIGNAL, sc->unit, 1);
/*
* H_VIO_SIGNAL enables interrupts for future packets only.
* Make sure none were queued between the end of the loop and the
* enable interrupts call.
*/
if ((sc->rx_buf[sc->rx_dma_slot].control >> 7) == sc->rx_valid_val)
goto restart;
mtx_unlock(&sc->io_lock);
}
static void
llan_send_packet(void *xsc, bus_dma_segment_t *segs, int nsegs,
bus_size_t mapsize, int error)
{
struct llan_softc *sc = xsc;
uint64_t bufdescs[6];
int i;
bzero(bufdescs, sizeof(bufdescs));
for (i = 0; i < nsegs; i++) {
bufdescs[i] = LLAN_BUFDESC_VALID;
bufdescs[i] |= (((uint64_t)segs[i].ds_len) << 32);
bufdescs[i] |= segs[i].ds_addr;
}
phyp_hcall(H_SEND_LOGICAL_LAN, sc->unit, bufdescs[0],
bufdescs[1], bufdescs[2], bufdescs[3], bufdescs[4], bufdescs[5], 0);
/*
* The hypercall returning implies completion -- or that the call will
* not complete. In principle, we should try a few times if we get back
* H_BUSY based on the continuation token in R4. For now, just drop
* the packet in such cases.
*/
}
static void
llan_start_locked(struct ifnet *ifp)
{
struct llan_softc *sc = ifp->if_softc;
bus_addr_t first;
int nsegs;
struct mbuf *mb_head, *m;
mtx_assert(&sc->io_lock, MA_OWNED);
first = 0;
if ((ifp->if_drv_flags & (IFF_DRV_RUNNING | IFF_DRV_OACTIVE)) !=
IFF_DRV_RUNNING)
return;
while (!IFQ_DRV_IS_EMPTY(&ifp->if_snd)) {
IFQ_DRV_DEQUEUE(&ifp->if_snd, mb_head);
if (mb_head == NULL)
break;
BPF_MTAP(ifp, mb_head);
for (m = mb_head, nsegs = 0; m != NULL; m = m->m_next)
nsegs++;
if (nsegs > 6) {
m = m_collapse(mb_head, M_NOWAIT, 6);
if (m == NULL) {
m_freem(mb_head);
continue;
}
}
bus_dmamap_load_mbuf(sc->tx_dma_tag, sc->tx_dma_map,
mb_head, llan_send_packet, sc, 0);
bus_dmamap_unload(sc->tx_dma_tag, sc->tx_dma_map);
m_freem(mb_head);
}
}
static void
llan_start(struct ifnet *ifp)
{
struct llan_softc *sc = ifp->if_softc;
mtx_lock(&sc->io_lock);
llan_start_locked(ifp);
mtx_unlock(&sc->io_lock);
}
static int
llan_set_multicast(struct llan_softc *sc)
{
struct ifnet *ifp = sc->ifp;
struct ifmultiaddr *inm;
uint64_t macaddr;
mtx_assert(&sc->io_lock, MA_OWNED);
phyp_hcall(H_MULTICAST_CTRL, sc->unit, LLAN_CLEAR_MULTICAST, 0);
if_maddr_rlock(ifp);
CK_STAILQ_FOREACH(inm, &ifp->if_multiaddrs, ifma_link) {
if (inm->ifma_addr->sa_family != AF_LINK)
continue;
memcpy((uint8_t *)&macaddr + 2,
LLADDR((struct sockaddr_dl *)inm->ifma_addr), 6);
phyp_hcall(H_MULTICAST_CTRL, sc->unit, LLAN_ADD_MULTICAST,
macaddr);
}
if_maddr_runlock(ifp);
return (0);
}
static int
llan_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
{
int err = 0;
struct llan_softc *sc = ifp->if_softc;
switch (cmd) {
case SIOCADDMULTI:
case SIOCDELMULTI:
mtx_lock(&sc->io_lock);
if ((sc->ifp->if_drv_flags & IFF_DRV_RUNNING) != 0)
llan_set_multicast(sc);
mtx_unlock(&sc->io_lock);
break;
case SIOCGIFMEDIA:
case SIOCSIFMEDIA:
err = ifmedia_ioctl(ifp, (struct ifreq *)data, &sc->media, cmd);
break;
case SIOCSIFFLAGS:
default:
err = ether_ioctl(ifp, cmd, data);
break;
}
return (err);
}