freebsd-dev/sys/dev/cxgbe/tom/t4_tom_l2t.c
Navdeep Parhar 09fe63205c - Updated TOE support in the kernel.
- Stateful TCP offload drivers for Terminator 3 and 4 (T3 and T4) ASICs.
  These are available as t3_tom and t4_tom modules that augment cxgb(4)
  and cxgbe(4) respectively.  The cxgb/cxgbe drivers continue to work as
  usual with or without these extra features.

- iWARP driver for Terminator 3 ASIC (kernel verbs).  T4 iWARP in the
  works and will follow soon.

Build-tested with make universe.

30s overview
============
What interfaces support TCP offload?  Look for TOE4 and/or TOE6 in the
capabilities of an interface:
# ifconfig -m | grep TOE

Enable/disable TCP offload on an interface (just like any other ifnet
capability):
# ifconfig cxgbe0 toe
# ifconfig cxgbe0 -toe

Which connections are offloaded?  Look for toe4 and/or toe6 in the
output of netstat and sockstat:
# netstat -np tcp | grep toe
# sockstat -46c | grep toe

Reviewed by:	bz, gnn
Sponsored by:	Chelsio communications.
MFC after:	~3 months (after 9.1, and after ensuring MFC is feasible)
2012-06-19 07:34:13 +00:00

406 lines
9.9 KiB
C

/*-
* Copyright (c) 2012 Chelsio Communications, Inc.
* 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 "opt_inet.h"
#ifdef TCP_OFFLOAD
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/kernel.h>
#include <sys/module.h>
#include <sys/bus.h>
#include <sys/lock.h>
#include <sys/mutex.h>
#include <sys/rwlock.h>
#include <sys/socket.h>
#include <sys/sbuf.h>
#include <net/if.h>
#include <net/if_types.h>
#include <net/ethernet.h>
#include <net/if_vlan_var.h>
#include <net/route.h>
#include <netinet/in.h>
#include <netinet/toecore.h>
#include "common/common.h"
#include "common/jhash.h"
#include "common/t4_msg.h"
#include "tom/t4_tom_l2t.h"
#include "tom/t4_tom.h"
#define VLAN_NONE 0xfff
#define SA(x) ((struct sockaddr *)(x))
#define SIN(x) ((struct sockaddr_in *)(x))
#define SINADDR(x) (SIN(x)->sin_addr.s_addr)
static inline void
l2t_hold(struct l2t_data *d, struct l2t_entry *e)
{
if (atomic_fetchadd_int(&e->refcnt, 1) == 0) /* 0 -> 1 transition */
atomic_subtract_int(&d->nfree, 1);
}
static inline unsigned int
arp_hash(const uint32_t key, int ifindex)
{
return jhash_2words(key, ifindex, 0) & (L2T_SIZE - 1);
}
/*
* Add a WR to an L2T entry's queue of work requests awaiting resolution.
* Must be called with the entry's lock held.
*/
static inline void
arpq_enqueue(struct l2t_entry *e, struct wrqe *wr)
{
mtx_assert(&e->lock, MA_OWNED);
STAILQ_INSERT_TAIL(&e->wr_list, wr, link);
}
static inline void
send_pending(struct adapter *sc, struct l2t_entry *e)
{
struct wrqe *wr;
mtx_assert(&e->lock, MA_OWNED);
while ((wr = STAILQ_FIRST(&e->wr_list)) != NULL) {
STAILQ_REMOVE_HEAD(&e->wr_list, link);
t4_wrq_tx(sc, wr);
}
}
static void
resolution_failed_for_wr(struct wrqe *wr)
{
log(LOG_ERR, "%s: leaked work request %p, wr_len %d", __func__, wr,
wr->wr_len);
/* free(wr, M_CXGBE); */
}
static void
resolution_failed(struct l2t_entry *e)
{
struct wrqe *wr;
mtx_assert(&e->lock, MA_OWNED);
while ((wr = STAILQ_FIRST(&e->wr_list)) != NULL) {
STAILQ_REMOVE_HEAD(&e->wr_list, link);
resolution_failed_for_wr(wr);
}
}
static void
update_entry(struct adapter *sc, struct l2t_entry *e, uint8_t *lladdr,
uint16_t vtag)
{
mtx_assert(&e->lock, MA_OWNED);
/*
* The entry may be in active use (e->refcount > 0) or not. We update
* it even when it's not as this simplifies the case where we decide to
* reuse the entry later.
*/
if (lladdr == NULL &&
(e->state == L2T_STATE_RESOLVING || e->state == L2T_STATE_FAILED)) {
/*
* Never got a valid L2 address for this one. Just mark it as
* failed instead of removing it from the hash (for which we'd
* need to wlock the table).
*/
e->state = L2T_STATE_FAILED;
resolution_failed(e);
return;
} else if (lladdr == NULL) {
/* Valid or already-stale entry was deleted (or expired) */
KASSERT(e->state == L2T_STATE_VALID ||
e->state == L2T_STATE_STALE,
("%s: lladdr NULL, state %d", __func__, e->state));
e->state = L2T_STATE_STALE;
} else {
if (e->state == L2T_STATE_RESOLVING ||
e->state == L2T_STATE_FAILED ||
memcmp(e->dmac, lladdr, ETHER_ADDR_LEN)) {
/* unresolved -> resolved; or dmac changed */
memcpy(e->dmac, lladdr, ETHER_ADDR_LEN);
e->vlan = vtag;
t4_write_l2e(sc, e, 1);
}
e->state = L2T_STATE_VALID;
}
}
static int
resolve_entry(struct adapter *sc, struct l2t_entry *e)
{
struct tom_data *td = sc->tom_softc;
struct toedev *tod = &td->tod;
struct sockaddr_in sin = {0};
uint8_t dmac[ETHER_ADDR_LEN];
uint16_t vtag = VLAN_NONE;
int rc;
sin.sin_family = AF_INET;
sin.sin_len = sizeof(struct sockaddr_in);
SINADDR(&sin) = e->addr;
rc = toe_l2_resolve(tod, e->ifp, SA(&sin), dmac, &vtag);
if (rc == EWOULDBLOCK)
return (rc);
mtx_lock(&e->lock);
update_entry(sc, e, rc == 0 ? dmac : NULL, vtag);
mtx_unlock(&e->lock);
return (rc);
}
int
t4_l2t_send_slow(struct adapter *sc, struct wrqe *wr, struct l2t_entry *e)
{
again:
switch (e->state) {
case L2T_STATE_STALE: /* entry is stale, kick off revalidation */
if (resolve_entry(sc, e) != EWOULDBLOCK)
goto again; /* entry updated, re-examine state */
/* Fall through */
case L2T_STATE_VALID: /* fast-path, send the packet on */
t4_wrq_tx(sc, wr);
return (0);
case L2T_STATE_RESOLVING:
case L2T_STATE_SYNC_WRITE:
mtx_lock(&e->lock);
if (e->state != L2T_STATE_SYNC_WRITE &&
e->state != L2T_STATE_RESOLVING) {
/* state changed by the time we got here */
mtx_unlock(&e->lock);
goto again;
}
arpq_enqueue(e, wr);
mtx_unlock(&e->lock);
if (resolve_entry(sc, e) == EWOULDBLOCK)
break;
mtx_lock(&e->lock);
if (e->state == L2T_STATE_VALID && !STAILQ_EMPTY(&e->wr_list))
send_pending(sc, e);
if (e->state == L2T_STATE_FAILED)
resolution_failed(e);
mtx_unlock(&e->lock);
break;
case L2T_STATE_FAILED:
resolution_failed_for_wr(wr);
return (EHOSTUNREACH);
}
return (0);
}
/*
* Called when an L2T entry has no more users. The entry is left in the hash
* table since it is likely to be reused but we also bump nfree to indicate
* that the entry can be reallocated for a different neighbor. We also drop
* the existing neighbor reference in case the neighbor is going away and is
* waiting on our reference.
*
* Because entries can be reallocated to other neighbors once their ref count
* drops to 0 we need to take the entry's lock to avoid races with a new
* incarnation.
*/
static int
do_l2t_write_rpl2(struct sge_iq *iq, const struct rss_header *rss,
struct mbuf *m)
{
struct adapter *sc = iq->adapter;
const struct cpl_l2t_write_rpl *rpl = (const void *)(rss + 1);
unsigned int tid = GET_TID(rpl);
unsigned int idx = tid & (L2T_SIZE - 1);
int rc;
rc = do_l2t_write_rpl(iq, rss, m);
if (rc != 0)
return (rc);
if (tid & F_SYNC_WR) {
struct l2t_entry *e = &sc->l2t->l2tab[idx];
mtx_lock(&e->lock);
if (e->state != L2T_STATE_SWITCHING) {
send_pending(sc, e);
e->state = L2T_STATE_VALID;
}
mtx_unlock(&e->lock);
}
return (0);
}
void
t4_init_l2t_cpl_handlers(struct adapter *sc)
{
t4_register_cpl_handler(sc, CPL_L2T_WRITE_RPL, do_l2t_write_rpl2);
}
void
t4_uninit_l2t_cpl_handlers(struct adapter *sc)
{
t4_register_cpl_handler(sc, CPL_L2T_WRITE_RPL, do_l2t_write_rpl);
}
/*
* The TOE wants an L2 table entry that it can use to reach the next hop over
* the specified port. Produce such an entry - create one if needed.
*
* Note that the ifnet could be a pseudo-device like if_vlan, if_lagg, etc. on
* top of the real cxgbe interface.
*/
struct l2t_entry *
t4_l2t_get(struct port_info *pi, struct ifnet *ifp, struct sockaddr *sa)
{
struct l2t_entry *e;
struct l2t_data *d = pi->adapter->l2t;
uint32_t addr = SINADDR(sa);
int hash = arp_hash(addr, ifp->if_index);
unsigned int smt_idx = pi->port_id;
if (sa->sa_family != AF_INET)
return (NULL); /* XXX: no IPv6 support right now */
#ifndef VLAN_TAG
if (ifp->if_type == IFT_L2VLAN)
return (NULL);
#endif
rw_wlock(&d->lock);
for (e = d->l2tab[hash].first; e; e = e->next) {
if (e->addr == addr && e->ifp == ifp && e->smt_idx == smt_idx) {
l2t_hold(d, e);
goto done;
}
}
/* Need to allocate a new entry */
e = t4_alloc_l2e(d);
if (e) {
mtx_lock(&e->lock); /* avoid race with t4_l2t_free */
e->next = d->l2tab[hash].first;
d->l2tab[hash].first = e;
e->state = L2T_STATE_RESOLVING;
e->addr = addr;
e->ifp = ifp;
e->smt_idx = smt_idx;
e->hash = hash;
e->lport = pi->lport;
atomic_store_rel_int(&e->refcnt, 1);
#ifdef VLAN_TAG
if (ifp->if_type == IFT_L2VLAN)
VLAN_TAG(ifp, &e->vlan);
else
e->vlan = VLAN_NONE;
#endif
mtx_unlock(&e->lock);
}
done:
rw_wunlock(&d->lock);
return e;
}
/*
* Called when the host's ARP layer makes a change to some entry that is loaded
* into the HW L2 table.
*/
void
t4_l2_update(struct toedev *tod, struct ifnet *ifp, struct sockaddr *sa,
uint8_t *lladdr, uint16_t vtag)
{
struct adapter *sc = tod->tod_softc;
struct l2t_entry *e;
struct l2t_data *d = sc->l2t;
uint32_t addr = SINADDR(sa);
int hash = arp_hash(addr, ifp->if_index);
KASSERT(d != NULL, ("%s: no L2 table", __func__));
rw_rlock(&d->lock);
for (e = d->l2tab[hash].first; e; e = e->next) {
if (e->addr == addr && e->ifp == ifp) {
mtx_lock(&e->lock);
if (atomic_load_acq_int(&e->refcnt))
goto found;
e->state = L2T_STATE_STALE;
mtx_unlock(&e->lock);
break;
}
}
rw_runlock(&d->lock);
/*
* This is of no interest to us. We've never had an offloaded
* connection to this destination, and we aren't attempting one right
* now.
*/
return;
found:
rw_runlock(&d->lock);
KASSERT(e->state != L2T_STATE_UNUSED,
("%s: unused entry in the hash.", __func__));
update_entry(sc, e, lladdr, vtag);
mtx_unlock(&e->lock);
}
#endif