aboutsummaryrefslogtreecommitdiff
path: root/sys/net
diff options
context:
space:
mode:
Diffstat (limited to 'sys/net')
-rw-r--r--sys/net/altq/altq_cbq.c2
-rw-r--r--sys/net/altq/altq_fairq.c2
-rw-r--r--sys/net/altq/altq_priq.c2
-rw-r--r--sys/net/bpf.c135
-rw-r--r--sys/net/bpf.h1
-rw-r--r--sys/net/ieee8023ad_lacp.c2
-rw-r--r--sys/net/if.c34
-rw-r--r--sys/net/if_media.h6
-rw-r--r--sys/net/if_ovpn.c2
-rw-r--r--sys/net/if_tuntap.c65
-rw-r--r--sys/net/if_var.h8
-rw-r--r--sys/net/if_vxlan.c4
-rw-r--r--sys/net/iflib.c173
-rw-r--r--sys/net/iflib.h2
-rw-r--r--sys/net/pfvar.h9
-rw-r--r--sys/net/route.c2
-rw-r--r--sys/net/route/route_tables.c2
-rw-r--r--sys/net/rtsock.c2
-rw-r--r--sys/net/sff8436.h2
19 files changed, 263 insertions, 192 deletions
diff --git a/sys/net/altq/altq_cbq.c b/sys/net/altq/altq_cbq.c
index fdf39690160b..2333b9ea8678 100644
--- a/sys/net/altq/altq_cbq.c
+++ b/sys/net/altq/altq_cbq.c
@@ -173,6 +173,8 @@ cbq_request(struct ifaltq *ifq, int req, void *arg)
static void
get_class_stats(class_stats_t *statsp, struct rm_class *cl)
{
+ memset(statsp, 0, sizeof(*statsp));
+
statsp->xmit_cnt = cl->stats_.xmit_cnt;
statsp->drop_cnt = cl->stats_.drop_cnt;
statsp->over = cl->stats_.over;
diff --git a/sys/net/altq/altq_fairq.c b/sys/net/altq/altq_fairq.c
index 6069865101a0..0a00168e547e 100644
--- a/sys/net/altq/altq_fairq.c
+++ b/sys/net/altq/altq_fairq.c
@@ -857,6 +857,8 @@ get_class_stats(struct fairq_classstats *sp, struct fairq_class *cl)
{
fairq_bucket_t *b;
+ memset(sp, 0, sizeof(*sp));
+
sp->class_handle = cl->cl_handle;
sp->qlimit = cl->cl_qlimit;
sp->xmit_cnt = cl->cl_xmitcnt;
diff --git a/sys/net/altq/altq_priq.c b/sys/net/altq/altq_priq.c
index 026346639b2e..fec488418546 100644
--- a/sys/net/altq/altq_priq.c
+++ b/sys/net/altq/altq_priq.c
@@ -597,6 +597,8 @@ priq_purgeq(struct priq_class *cl)
static void
get_class_stats(struct priq_classstats *sp, struct priq_class *cl)
{
+ memset(sp, 0, sizeof(*sp));
+
sp->class_handle = cl->cl_handle;
sp->qlength = qlen(cl->cl_q);
sp->qlimit = qlimit(cl->cl_q);
diff --git a/sys/net/bpf.c b/sys/net/bpf.c
index a347dbe2eb73..23a23fbfe22e 100644
--- a/sys/net/bpf.c
+++ b/sys/net/bpf.c
@@ -37,7 +37,6 @@
#include <sys/cdefs.h>
#include "opt_bpf.h"
-#include "opt_ddb.h"
#include "opt_netgraph.h"
#include <sys/param.h>
@@ -68,10 +67,6 @@
#include <sys/socket.h>
-#ifdef DDB
-#include <ddb/ddb.h>
-#endif
-
#include <net/if.h>
#include <net/if_var.h>
#include <net/if_private.h>
@@ -192,8 +187,7 @@ static void bpfif_rele(struct bpf_if *);
static void bpfd_ref(struct bpf_d *);
static void bpfd_rele(struct bpf_d *);
static void bpf_attachd(struct bpf_d *, struct bpf_if *);
-static void bpf_detachd(struct bpf_d *);
-static void bpf_detachd_locked(struct bpf_d *, bool);
+static void bpf_detachd(struct bpf_d *, bool);
static void bpfd_free(epoch_context_t);
static int bpf_movein(struct uio *, int, struct ifnet *, struct mbuf **,
struct sockaddr *, int *, struct bpf_d *);
@@ -253,12 +247,14 @@ static const struct filterops bpfread_filtops = {
.f_isfd = 1,
.f_detach = filt_bpfdetach,
.f_event = filt_bpfread,
+ .f_copy = knote_triv_copy,
};
static const struct filterops bpfwrite_filtops = {
.f_isfd = 1,
.f_detach = filt_bpfdetach,
.f_event = filt_bpfwrite,
+ .f_copy = knote_triv_copy,
};
/*
@@ -731,7 +727,7 @@ bpf_attachd(struct bpf_d *d, struct bpf_if *bp)
op_w = V_bpf_optimize_writers || d->bd_writer;
if (d->bd_bif != NULL)
- bpf_detachd_locked(d, false);
+ bpf_detachd(d, false);
/*
* Point d at bp, and add d to the interface's list.
* Since there are many applications using BPF for
@@ -841,15 +837,7 @@ bpf_check_upgrade(u_long cmd, struct bpf_d *d, struct bpf_insn *fcode,
* Detach a file from its interface.
*/
static void
-bpf_detachd(struct bpf_d *d)
-{
- BPF_LOCK();
- bpf_detachd_locked(d, false);
- BPF_UNLOCK();
-}
-
-static void
-bpf_detachd_locked(struct bpf_d *d, bool detached_ifp)
+bpf_detachd(struct bpf_d *d, bool detached_ifp)
{
struct bpf_if *bp;
struct ifnet *ifp;
@@ -921,7 +909,9 @@ bpf_dtor(void *data)
d->bd_state = BPF_IDLE;
BPFD_UNLOCK(d);
funsetown(&d->bd_sigio);
- bpf_detachd(d);
+ BPF_LOCK();
+ bpf_detachd(d, false);
+ BPF_UNLOCK();
#ifdef MAC
mac_bpfdesc_destroy(d);
#endif /* MAC */
@@ -1607,33 +1597,28 @@ bpfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags,
* Set interface.
*/
case BIOCSETIF:
- {
- int alloc_buf, size;
+ /*
+ * Behavior here depends on the buffering model. If we're
+ * using kernel memory buffers, then we can allocate them here.
+ * If we're using zero-copy, then the user process must have
+ * registered buffers by the time we get here.
+ */
+ BPFD_LOCK(d);
+ if (d->bd_bufmode == BPF_BUFMODE_BUFFER &&
+ d->bd_sbuf == NULL) {
+ u_int size;
- /*
- * Behavior here depends on the buffering model. If
- * we're using kernel memory buffers, then we can
- * allocate them here. If we're using zero-copy,
- * then the user process must have registered buffers
- * by the time we get here.
- */
- alloc_buf = 0;
- BPFD_LOCK(d);
- if (d->bd_bufmode == BPF_BUFMODE_BUFFER &&
- d->bd_sbuf == NULL)
- alloc_buf = 1;
+ size = d->bd_bufsize;
BPFD_UNLOCK(d);
- if (alloc_buf) {
- size = d->bd_bufsize;
- error = bpf_buffer_ioctl_sblen(d, &size);
- if (error != 0)
- break;
- }
- BPF_LOCK();
- error = bpf_setif(d, (struct ifreq *)addr);
- BPF_UNLOCK();
- break;
- }
+ error = bpf_buffer_ioctl_sblen(d, &size);
+ if (error != 0)
+ break;
+ } else
+ BPFD_UNLOCK(d);
+ BPF_LOCK();
+ error = bpf_setif(d, (struct ifreq *)addr);
+ BPF_UNLOCK();
+ break;
/*
* Set read timeout.
@@ -2827,30 +2812,6 @@ bpfattach2(struct ifnet *ifp, u_int dlt, u_int hdrlen,
#ifdef VIMAGE
/*
- * When moving interfaces between vnet instances we need a way to
- * query the dlt and hdrlen before detach so we can re-attch the if_bpf
- * after the vmove. We unfortunately have no device driver infrastructure
- * to query the interface for these values after creation/attach, thus
- * add this as a workaround.
- */
-int
-bpf_get_bp_params(struct bpf_if *bp, u_int *bif_dlt, u_int *bif_hdrlen)
-{
-
- if (bp == NULL)
- return (ENXIO);
- if (bif_dlt == NULL && bif_hdrlen == NULL)
- return (0);
-
- if (bif_dlt != NULL)
- *bif_dlt = bp->bif_dlt;
- if (bif_hdrlen != NULL)
- *bif_hdrlen = bp->bif_hdrlen;
-
- return (0);
-}
-
-/*
* Detach descriptors on interface's vmove event.
*/
void
@@ -2866,12 +2827,12 @@ bpf_ifdetach(struct ifnet *ifp)
/* Detach common descriptors */
while ((d = CK_LIST_FIRST(&bp->bif_dlist)) != NULL) {
- bpf_detachd_locked(d, true);
+ bpf_detachd(d, true);
}
/* Detach writer-only descriptors */
while ((d = CK_LIST_FIRST(&bp->bif_wlist)) != NULL) {
- bpf_detachd_locked(d, true);
+ bpf_detachd(d, true);
}
}
BPF_UNLOCK();
@@ -2904,12 +2865,12 @@ bpfdetach(struct ifnet *ifp)
/* Detach common descriptors */
while ((d = CK_LIST_FIRST(&bp->bif_dlist)) != NULL) {
- bpf_detachd_locked(d, true);
+ bpf_detachd(d, true);
}
/* Detach writer-only descriptors */
while ((d = CK_LIST_FIRST(&bp->bif_wlist)) != NULL) {
- bpf_detachd_locked(d, true);
+ bpf_detachd(d, true);
}
bpfif_rele(bp);
}
@@ -3222,35 +3183,3 @@ bpf_validate(const struct bpf_insn *f, int len)
}
#endif /* !DEV_BPF && !NETGRAPH_BPF */
-
-#ifdef DDB
-static void
-bpf_show_bpf_if(struct bpf_if *bpf_if)
-{
-
- if (bpf_if == NULL)
- return;
- db_printf("%p:\n", bpf_if);
-#define BPF_DB_PRINTF(f, e) db_printf(" %s = " f "\n", #e, bpf_if->e);
-#define BPF_DB_PRINTF_RAW(f, e) db_printf(" %s = " f "\n", #e, e);
- /* bif_ext.bif_next */
- /* bif_ext.bif_dlist */
- BPF_DB_PRINTF("%#x", bif_dlt);
- BPF_DB_PRINTF("%u", bif_hdrlen);
- /* bif_wlist */
- BPF_DB_PRINTF("%p", bif_ifp);
- BPF_DB_PRINTF("%p", bif_bpf);
- BPF_DB_PRINTF_RAW("%u", refcount_load(&bpf_if->bif_refcnt));
-}
-
-DB_SHOW_COMMAND(bpf_if, db_show_bpf_if)
-{
-
- if (!have_addr) {
- db_printf("usage: show bpf_if <struct bpf_if *>\n");
- return;
- }
-
- bpf_show_bpf_if((struct bpf_if *)addr);
-}
-#endif
diff --git a/sys/net/bpf.h b/sys/net/bpf.h
index 486e3f59f7d5..dfb8373bb329 100644
--- a/sys/net/bpf.h
+++ b/sys/net/bpf.h
@@ -428,7 +428,6 @@ void bpfattach2(struct ifnet *, u_int, u_int, struct bpf_if **);
void bpfdetach(struct ifnet *);
bool bpf_peers_present_if(struct ifnet *);
#ifdef VIMAGE
-int bpf_get_bp_params(struct bpf_if *, u_int *, u_int *);
void bpf_ifdetach(struct ifnet *);
#endif
diff --git a/sys/net/ieee8023ad_lacp.c b/sys/net/ieee8023ad_lacp.c
index 9ebdd11f70f3..77b5a5d53a67 100644
--- a/sys/net/ieee8023ad_lacp.c
+++ b/sys/net/ieee8023ad_lacp.c
@@ -1264,6 +1264,8 @@ lacp_compose_key(struct lacp_port *lp)
case IFM_400G_DR4:
case IFM_400G_AUI8_AC:
case IFM_400G_AUI8:
+ case IFM_400G_SR8:
+ case IFM_400G_CR8:
key = IFM_400G_FR8;
break;
default:
diff --git a/sys/net/if.c b/sys/net/if.c
index b6a798aa0fab..cb9c47c14c32 100644
--- a/sys/net/if.c
+++ b/sys/net/if.c
@@ -2842,15 +2842,20 @@ ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
break;
case SIOCAIFGROUP:
+ {
+ const char *groupname;
+
error = priv_check(td, PRIV_NET_ADDIFGROUP);
if (error)
return (error);
- error = if_addgroup(ifp,
- ((struct ifgroupreq *)data)->ifgr_group);
+ groupname = ((struct ifgroupreq *)data)->ifgr_group;
+ if (strnlen(groupname, IFNAMSIZ) == IFNAMSIZ)
+ return (EINVAL);
+ error = if_addgroup(ifp, groupname);
if (error != 0)
return (error);
break;
-
+ }
case SIOCGIFGROUP:
{
struct epoch_tracker et;
@@ -2862,15 +2867,20 @@ ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
}
case SIOCDIFGROUP:
+ {
+ const char *groupname;
+
error = priv_check(td, PRIV_NET_DELIFGROUP);
if (error)
return (error);
- error = if_delgroup(ifp,
- ((struct ifgroupreq *)data)->ifgr_group);
+ groupname = ((struct ifgroupreq *)data)->ifgr_group;
+ if (strnlen(groupname, IFNAMSIZ) == IFNAMSIZ)
+ return (EINVAL);
+ error = if_delgroup(ifp, groupname);
if (error != 0)
return (error);
break;
-
+ }
default:
error = ENOIOCTL;
break;
@@ -3014,9 +3024,17 @@ ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
goto out_noref;
case SIOCGIFGMEMB:
- error = if_getgroupmembers((struct ifgroupreq *)data);
- goto out_noref;
+ {
+ struct ifgroupreq *req;
+ req = (struct ifgroupreq *)data;
+ if (strnlen(req->ifgr_name, IFNAMSIZ) == IFNAMSIZ) {
+ error = EINVAL;
+ goto out_noref;
+ }
+ error = if_getgroupmembers(req);
+ goto out_noref;
+ }
#if defined(INET) || defined(INET6)
case SIOCSVH:
case SIOCGVH:
diff --git a/sys/net/if_media.h b/sys/net/if_media.h
index a2cac00550ef..892b7d1e3e52 100644
--- a/sys/net/if_media.h
+++ b/sys/net/if_media.h
@@ -260,6 +260,8 @@ uint64_t ifmedia_baudrate(int);
#define IFM_40G_LM4 IFM_X(119) /* 40GBase-LM4 */
#define IFM_100_BX IFM_X(120) /* 100Base-BX */
#define IFM_1000_BX IFM_X(121) /* 1000Base-BX */
+#define IFM_400G_SR8 IFM_X(122) /* 400GBase-SR8 */
+#define IFM_400G_CR8 IFM_X(123) /* 400GBase-CR8 */
/*
* Please update ieee8023ad_lacp.c:lacp_compose_key()
@@ -550,6 +552,8 @@ struct ifmedia_description {
{ IFM_400G_DR4, "400GBase-DR4" }, \
{ IFM_400G_AUI8_AC, "400G-AUI8-AC" }, \
{ IFM_400G_AUI8, "400G-AUI8" }, \
+ { IFM_400G_SR8, "400GBase-SR8" }, \
+ { IFM_400G_CR8, "400GBase-CR8" }, \
{ 0, NULL }, \
}
@@ -897,6 +901,8 @@ struct ifmedia_baudrate {
{ IFM_ETHER | IFM_400G_DR4, IF_Gbps(400ULL) }, \
{ IFM_ETHER | IFM_400G_AUI8_AC, IF_Gbps(400ULL) }, \
{ IFM_ETHER | IFM_400G_AUI8, IF_Gbps(400ULL) }, \
+ { IFM_ETHER | IFM_400G_SR8, IF_Gbps(400ULL) }, \
+ { IFM_ETHER | IFM_400G_CR8, IF_Gbps(400ULL) }, \
\
{ IFM_IEEE80211 | IFM_IEEE80211_FH1, IF_Mbps(1) }, \
{ IFM_IEEE80211 | IFM_IEEE80211_FH2, IF_Mbps(2) }, \
diff --git a/sys/net/if_ovpn.c b/sys/net/if_ovpn.c
index 1c18baac3417..674df4d17eb4 100644
--- a/sys/net/if_ovpn.c
+++ b/sys/net/if_ovpn.c
@@ -2691,7 +2691,7 @@ ovpn_clone_create(struct if_clone *ifc, char *name, size_t len,
return (EEXIST);
sc = malloc(sizeof(struct ovpn_softc), M_OVPN, M_WAITOK | M_ZERO);
- sc->ifp = if_alloc(IFT_ENC);
+ sc->ifp = if_alloc(IFT_TUNNEL);
rm_init_flags(&sc->lock, "if_ovpn_lock", RM_RECURSE);
sc->refcount = 0;
diff --git a/sys/net/if_tuntap.c b/sys/net/if_tuntap.c
index c8dbb6aa8893..0dc3a58f6ae6 100644
--- a/sys/net/if_tuntap.c
+++ b/sys/net/if_tuntap.c
@@ -138,6 +138,7 @@ struct tuntap_softc {
#define TUN_READY (TUN_OPEN | TUN_INITED)
pid_t tun_pid; /* owning pid */
+ struct epoch_context tun_epoch_ctx;
struct ifnet *tun_ifp; /* the interface */
struct sigio *tun_sigio; /* async I/O info */
struct tuntap_driver *tun_drv; /* appropriate driver */
@@ -261,6 +262,7 @@ static const struct filterops tun_read_filterops = {
.f_attach = NULL,
.f_detach = tunkqdetach,
.f_event = tunkqread,
+ .f_copy = knote_triv_copy,
};
static const struct filterops tun_write_filterops = {
@@ -268,6 +270,7 @@ static const struct filterops tun_write_filterops = {
.f_attach = NULL,
.f_detach = tunkqdetach,
.f_event = tunkqwrite,
+ .f_copy = knote_triv_copy,
};
static struct tuntap_driver {
@@ -628,6 +631,18 @@ out:
CURVNET_RESTORE();
}
+static void
+tunfree(struct epoch_context *ctx)
+{
+ struct tuntap_softc *tp;
+
+ tp = __containerof(ctx, struct tuntap_softc, tun_epoch_ctx);
+
+ /* Any remaining resources that would be needed by a concurrent open. */
+ mtx_destroy(&tp->tun_mtx);
+ free(tp, M_TUN);
+}
+
static int
tun_destroy(struct tuntap_softc *tp, bool may_intr)
{
@@ -647,7 +662,7 @@ tun_destroy(struct tuntap_softc *tp, bool may_intr)
error = cv_wait_sig(&tp->tun_cv, &tp->tun_mtx);
else
cv_wait(&tp->tun_cv, &tp->tun_mtx);
- if (error != 0) {
+ if (error != 0 && tp->tun_busy != 0) {
tp->tun_flags &= ~TUN_DYING;
TUN_UNLOCK(tp);
return (error);
@@ -661,8 +676,18 @@ tun_destroy(struct tuntap_softc *tp, bool may_intr)
TAILQ_REMOVE(&tunhead, tp, tun_list);
mtx_unlock(&tunmtx);
- /* destroy_dev will take care of any alias. */
- destroy_dev(tp->tun_dev);
+ /*
+ * destroy_dev will take care of any alias. For transient tunnels,
+ * we're being called from close(2) so we can't destroy it ourselves
+ * without deadlocking, but we already know that we can cleanup
+ * everything else and just continue to prevent it from being reopened.
+ */
+ if ((tp->tun_flags & TUN_TRANSIENT) != 0) {
+ atomic_store_ptr(&tp->tun_dev->si_drv1, tp->tun_dev);
+ destroy_dev_sched(tp->tun_dev);
+ } else {
+ destroy_dev(tp->tun_dev);
+ }
seldrain(&tp->tun_rsel);
knlist_clear(&tp->tun_rsel.si_note, 0);
knlist_destroy(&tp->tun_rsel.si_note);
@@ -677,9 +702,8 @@ tun_destroy(struct tuntap_softc *tp, bool may_intr)
sx_xunlock(&tun_ioctl_sx);
free_unr(tp->tun_drv->unrhdr, TUN2IFP(tp)->if_dunit);
if_free(TUN2IFP(tp));
- mtx_destroy(&tp->tun_mtx);
cv_destroy(&tp->tun_cv);
- free(tp, M_TUN);
+ NET_EPOCH_CALL(tunfree, &tp->tun_epoch_ctx);
CURVNET_RESTORE();
return (0);
@@ -740,9 +764,11 @@ tun_uninit(const void *unused __unused)
mtx_unlock(&tunmtx);
for (i = 0; i < nitems(tuntap_drivers); ++i) {
drv = &tuntap_drivers[i];
+ destroy_dev_drain(&drv->cdevsw);
delete_unrhdr(drv->unrhdr);
clone_cleanup(&drv->clones);
}
+ NET_EPOCH_DRAIN_CALLBACKS();
mtx_destroy(&tunmtx);
}
SYSUNINIT(tun_uninit, SI_SUB_PROTO_IF, SI_ORDER_ANY, tun_uninit, NULL);
@@ -1102,19 +1128,43 @@ out:
static int
tunopen(struct cdev *dev, int flag, int mode, struct thread *td)
{
+ struct epoch_tracker et;
struct ifnet *ifp;
struct tuntap_softc *tp;
+ void *p;
int error __diagused, tunflags;
+ /*
+ * Transient tunnels do deferred destroy of the tun device but want
+ * to immediately cleanup state, so they clobber si_drv1 to avoid a
+ * use-after-free in case someone does happen to open it in the interim.
+ * We avoid using NULL to be able to distinguish from an uninitialized
+ * cdev.
+ *
+ * We use the net epoch here to let a concurrent tun_destroy() schedule
+ * freeing our tuntap_softc, in case we entered here and loaded si_drv1
+ * before it was swapped out. If we managed to load this while it was
+ * still a softc, then the concurrent tun_destroy() hasn't yet scheduled
+ * it to be free- that will take place sometime after the epoch we just
+ * entered, so we can safely use it.
+ */
+ NET_EPOCH_ENTER(et);
+ p = atomic_load_ptr(&dev->si_drv1);
+ if (p == dev) {
+ NET_EPOCH_EXIT(et);
+ return (ENXIO);
+ }
+
tunflags = 0;
CURVNET_SET(TD_TO_VNET(td));
error = tuntap_name2info(dev->si_name, NULL, &tunflags);
if (error != 0) {
CURVNET_RESTORE();
+ NET_EPOCH_EXIT(et);
return (error); /* Shouldn't happen */
}
- tp = dev->si_drv1;
+ tp = p;
KASSERT(tp != NULL,
("si_drv1 should have been initialized at creation"));
@@ -1122,14 +1172,17 @@ tunopen(struct cdev *dev, int flag, int mode, struct thread *td)
if ((tp->tun_flags & TUN_INITED) == 0) {
TUN_UNLOCK(tp);
CURVNET_RESTORE();
+ NET_EPOCH_EXIT(et);
return (ENXIO);
}
if ((tp->tun_flags & (TUN_OPEN | TUN_DYING)) != 0) {
TUN_UNLOCK(tp);
CURVNET_RESTORE();
+ NET_EPOCH_EXIT(et);
return (EBUSY);
}
+ NET_EPOCH_EXIT(et);
error = tun_busy_locked(tp);
KASSERT(error == 0, ("Must be able to busy an unopen tunnel"));
ifp = TUN2IFP(tp);
diff --git a/sys/net/if_var.h b/sys/net/if_var.h
index f2df612b19c1..961259bb0ca1 100644
--- a/sys/net/if_var.h
+++ b/sys/net/if_var.h
@@ -383,18 +383,18 @@ struct ifg_group {
char ifg_group[IFNAMSIZ];
u_int ifg_refcnt;
void *ifg_pf_kif;
- CK_STAILQ_HEAD(, ifg_member) ifg_members; /* (CK_) */
- CK_STAILQ_ENTRY(ifg_group) ifg_next; /* (CK_) */
+ CK_STAILQ_HEAD(, ifg_member) ifg_members;
+ CK_STAILQ_ENTRY(ifg_group) ifg_next;
};
struct ifg_member {
- CK_STAILQ_ENTRY(ifg_member) ifgm_next; /* (CK_) */
+ CK_STAILQ_ENTRY(ifg_member) ifgm_next;
if_t ifgm_ifp;
};
struct ifg_list {
struct ifg_group *ifgl_group;
- CK_STAILQ_ENTRY(ifg_list) ifgl_next; /* (CK_) */
+ CK_STAILQ_ENTRY(ifg_list) ifgl_next;
};
#ifdef _SYS_EVENTHANDLER_H_
diff --git a/sys/net/if_vxlan.c b/sys/net/if_vxlan.c
index 03184c1fb678..f3a8410a2258 100644
--- a/sys/net/if_vxlan.c
+++ b/sys/net/if_vxlan.c
@@ -2533,7 +2533,7 @@ vxlan_encap4(struct vxlan_softc *sc, const union vxlan_sockaddr *fvxlsa,
ifp = sc->vxl_ifp;
srcaddr = sc->vxl_src_addr.in4.sin_addr;
- srcport = vxlan_pick_source_port(sc, m);
+ srcport = htons(vxlan_pick_source_port(sc, m));
dstaddr = fvxlsa->in4.sin_addr;
dstport = fvxlsa->in4.sin_port;
@@ -2644,7 +2644,7 @@ vxlan_encap6(struct vxlan_softc *sc, const union vxlan_sockaddr *fvxlsa,
ifp = sc->vxl_ifp;
srcaddr = &sc->vxl_src_addr.in6.sin6_addr;
- srcport = vxlan_pick_source_port(sc, m);
+ srcport = htons(vxlan_pick_source_port(sc, m));
dstaddr = &fvxlsa->in6.sin6_addr;
dstport = fvxlsa->in6.sin6_port;
diff --git a/sys/net/iflib.c b/sys/net/iflib.c
index 1e6d98291c04..3181bdbcb849 100644
--- a/sys/net/iflib.c
+++ b/sys/net/iflib.c
@@ -142,7 +142,9 @@ struct iflib_ctx;
static void iru_init(if_rxd_update_t iru, iflib_rxq_t rxq, uint8_t flid);
static void iflib_timer(void *arg);
static void iflib_tqg_detach(if_ctx_t ctx);
+#ifndef ALTQ
static int iflib_simple_transmit(if_t ifp, struct mbuf *m);
+#endif
typedef struct iflib_filter_info {
driver_filter_t *ifi_filter;
@@ -200,6 +202,8 @@ struct iflib_ctx {
uint16_t ifc_sysctl_extra_msix_vectors;
bool ifc_cpus_are_physical_cores;
bool ifc_sysctl_simple_tx;
+ uint16_t ifc_sysctl_tx_reclaim_thresh;
+ uint16_t ifc_sysctl_tx_reclaim_ticks;
qidx_t ifc_sysctl_ntxds[8];
qidx_t ifc_sysctl_nrxds[8];
@@ -343,7 +347,9 @@ struct iflib_txq {
uint16_t ift_npending;
uint16_t ift_db_pending;
uint16_t ift_rs_pending;
- /* implicit pad */
+ uint32_t ift_last_reclaim;
+ uint16_t ift_reclaim_thresh;
+ uint16_t ift_reclaim_ticks;
uint8_t ift_txd_size[8];
uint64_t ift_processed;
uint64_t ift_cleaned;
@@ -727,7 +733,7 @@ static void iflib_free_intr_mem(if_ctx_t ctx);
#ifndef __NO_STRICT_ALIGNMENT
static struct mbuf *iflib_fixup_rx(struct mbuf *m);
#endif
-static __inline int iflib_completed_tx_reclaim(iflib_txq_t txq, int thresh);
+static __inline int iflib_completed_tx_reclaim(iflib_txq_t txq);
static SLIST_HEAD(cpu_offset_list, cpu_offset) cpu_offsets =
SLIST_HEAD_INITIALIZER(cpu_offsets);
@@ -3082,8 +3088,6 @@ txq_max_rs_deferred(iflib_txq_t txq)
#define QIDX(ctx, m) ((((m)->m_pkthdr.flowid & ctx->ifc_softc_ctx.isc_rss_table_mask) % NTXQSETS(ctx)) + FIRST_QSET(ctx))
#define DESC_RECLAIMABLE(q) ((int)((q)->ift_processed - (q)->ift_cleaned - (q)->ift_ctx->ifc_softc_ctx.isc_tx_nsegments))
-/* XXX we should be setting this to something other than zero */
-#define RECLAIM_THRESH(ctx) ((ctx)->ifc_sctx->isc_tx_reclaim_thresh)
#define MAX_TX_DESC(ctx) MAX((ctx)->ifc_softc_ctx.isc_tx_tso_segments_max, \
(ctx)->ifc_softc_ctx.isc_tx_nsegments)
@@ -3445,25 +3449,6 @@ iflib_remove_mbuf(iflib_txq_t txq)
return (m);
}
-static inline caddr_t
-calc_next_txd(iflib_txq_t txq, int cidx, uint8_t qid)
-{
- qidx_t size;
- int ntxd;
- caddr_t start, end, cur, next;
-
- ntxd = txq->ift_size;
- size = txq->ift_txd_size[qid];
- start = txq->ift_ifdi[qid].idi_vaddr;
-
- if (__predict_false(size == 0))
- return (start);
- cur = start + size * cidx;
- end = start + size * ntxd;
- next = CACHE_PTR_NEXT(cur);
- return (next < end ? next : start);
-}
-
/*
* Pad an mbuf to ensure a minimum ethernet frame size.
* min_frame_size is the frame size (less CRC) to pad the mbuf to
@@ -3517,37 +3502,22 @@ iflib_encap(iflib_txq_t txq, struct mbuf **m_headp)
bus_dma_tag_t buf_tag;
bus_dma_segment_t *segs;
struct mbuf *m_head, **ifsd_m;
- void *next_txd;
bus_dmamap_t map;
struct if_pkt_info pi;
int remap = 0;
- int err, nsegs, ndesc, max_segs, pidx, cidx, next, ntxd;
+ int err, nsegs, ndesc, max_segs, pidx;
ctx = txq->ift_ctx;
sctx = ctx->ifc_sctx;
scctx = &ctx->ifc_softc_ctx;
segs = txq->ift_segs;
- ntxd = txq->ift_size;
m_head = *m_headp;
map = NULL;
/*
* If we're doing TSO the next descriptor to clean may be quite far ahead
*/
- cidx = txq->ift_cidx;
pidx = txq->ift_pidx;
- if (ctx->ifc_flags & IFC_PREFETCH) {
- next = (cidx + CACHE_PTR_INCREMENT) & (ntxd - 1);
- if (!(ctx->ifc_flags & IFLIB_HAS_TXCQ)) {
- next_txd = calc_next_txd(txq, cidx, 0);
- prefetch(next_txd);
- }
-
- /* prefetch the next cache line of mbuf pointers and flags */
- prefetch(&txq->ift_sds.ifsd_m[next]);
- prefetch(&txq->ift_sds.ifsd_map[next]);
- next = (cidx + CACHE_LINE_SIZE) & (ntxd - 1);
- }
map = txq->ift_sds.ifsd_map[pidx];
ifsd_m = txq->ift_sds.ifsd_m;
@@ -3640,7 +3610,7 @@ defrag:
* cxgb
*/
if (__predict_false(nsegs + 2 > TXQ_AVAIL(txq))) {
- (void)iflib_completed_tx_reclaim(txq, RECLAIM_THRESH(ctx));
+ (void)iflib_completed_tx_reclaim(txq);
if (__predict_false(nsegs + 2 > TXQ_AVAIL(txq))) {
txq->ift_no_desc_avail++;
bus_dmamap_unload(buf_tag, map);
@@ -3733,24 +3703,16 @@ defrag_failed:
static void
iflib_tx_desc_free(iflib_txq_t txq, int n)
{
- uint32_t qsize, cidx, mask, gen;
+ uint32_t qsize, cidx, gen;
struct mbuf *m, **ifsd_m;
- bool do_prefetch;
cidx = txq->ift_cidx;
gen = txq->ift_gen;
qsize = txq->ift_size;
- mask = qsize - 1;
ifsd_m = txq->ift_sds.ifsd_m;
- do_prefetch = (txq->ift_ctx->ifc_flags & IFC_PREFETCH);
while (n-- > 0) {
- if (do_prefetch) {
- prefetch(ifsd_m[(cidx + 3) & mask]);
- prefetch(ifsd_m[(cidx + 4) & mask]);
- }
if ((m = ifsd_m[cidx]) != NULL) {
- prefetch(&ifsd_m[(cidx + CACHE_PTR_INCREMENT) & mask]);
if (m->m_pkthdr.csum_flags & CSUM_TSO) {
bus_dmamap_sync(txq->ift_tso_buf_tag,
txq->ift_sds.ifsd_tso_map[cidx],
@@ -3783,14 +3745,21 @@ iflib_tx_desc_free(iflib_txq_t txq, int n)
}
static __inline int
-iflib_completed_tx_reclaim(iflib_txq_t txq, int thresh)
+iflib_completed_tx_reclaim(iflib_txq_t txq)
{
- int reclaim;
+ int reclaim, thresh;
+ uint32_t now;
if_ctx_t ctx = txq->ift_ctx;
+ thresh = txq->ift_reclaim_thresh;
KASSERT(thresh >= 0, ("invalid threshold to reclaim"));
MPASS(thresh /*+ MAX_TX_DESC(txq->ift_ctx) */ < txq->ift_size);
+ now = ticks;
+ if (now <= (txq->ift_last_reclaim + txq->ift_reclaim_ticks) &&
+ txq->ift_in_use < thresh)
+ return (0);
+ txq->ift_last_reclaim = now;
/*
* Need a rate-limiting check so that this isn't called every time
*/
@@ -3871,7 +3840,7 @@ iflib_txq_drain(struct ifmp_ring *r, uint32_t cidx, uint32_t pidx)
DBG_COUNTER_INC(txq_drain_notready);
return (0);
}
- reclaimed = iflib_completed_tx_reclaim(txq, RECLAIM_THRESH(ctx));
+ reclaimed = iflib_completed_tx_reclaim(txq);
rang = iflib_txd_db_check(txq, reclaimed && txq->ift_db_pending);
avail = IDXDIFF(pidx, cidx, r->size);
@@ -3950,7 +3919,7 @@ iflib_txq_drain(struct ifmp_ring *r, uint32_t cidx, uint32_t pidx)
}
/* deliberate use of bitwise or to avoid gratuitous short-circuit */
- ring = rang ? false : (iflib_min_tx_latency | err);
+ ring = rang ? false : (iflib_min_tx_latency | err | (!!txq->ift_reclaim_thresh));
iflib_txd_db_check(txq, ring);
if_inc_counter(ifp, IFCOUNTER_OBYTES, bytes_sent);
if_inc_counter(ifp, IFCOUNTER_OPACKETS, pkt_sent);
@@ -4030,7 +3999,7 @@ _task_fn_tx(void *context)
#endif
if (ctx->ifc_sysctl_simple_tx) {
mtx_lock(&txq->ift_mtx);
- (void)iflib_completed_tx_reclaim(txq, RECLAIM_THRESH(ctx));
+ (void)iflib_completed_tx_reclaim(txq);
mtx_unlock(&txq->ift_mtx);
goto skip_ifmp;
}
@@ -5881,6 +5850,7 @@ iflib_queues_alloc(if_ctx_t ctx)
device_printf(dev, "Unable to allocate buf_ring\n");
goto err_tx_desc;
}
+ txq->ift_reclaim_thresh = ctx->ifc_sysctl_tx_reclaim_thresh;
}
for (rxconf = i = 0; i < nrxqsets; i++, rxconf++, rxq++) {
@@ -6772,6 +6742,74 @@ mp_ndesc_handler(SYSCTL_HANDLER_ARGS)
return (rc);
}
+static int
+iflib_handle_tx_reclaim_thresh(SYSCTL_HANDLER_ARGS)
+{
+ if_ctx_t ctx = (void *)arg1;
+ iflib_txq_t txq;
+ int i, err;
+ int thresh;
+
+ thresh = ctx->ifc_sysctl_tx_reclaim_thresh;
+ err = sysctl_handle_int(oidp, &thresh, arg2, req);
+ if (err != 0) {
+ return err;
+ }
+
+ if (thresh == ctx->ifc_sysctl_tx_reclaim_thresh)
+ return 0;
+
+ if (thresh > ctx->ifc_softc_ctx.isc_ntxd[0] / 2) {
+ device_printf(ctx->ifc_dev, "TX Reclaim thresh must be <= %d\n",
+ ctx->ifc_softc_ctx.isc_ntxd[0] / 2);
+ return (EINVAL);
+ }
+
+ ctx->ifc_sysctl_tx_reclaim_thresh = thresh;
+ if (ctx->ifc_txqs == NULL)
+ return (err);
+
+ txq = &ctx->ifc_txqs[0];
+ for (i = 0; i < NTXQSETS(ctx); i++, txq++) {
+ txq->ift_reclaim_thresh = thresh;
+ }
+ return (err);
+}
+
+static int
+iflib_handle_tx_reclaim_ticks(SYSCTL_HANDLER_ARGS)
+{
+ if_ctx_t ctx = (void *)arg1;
+ iflib_txq_t txq;
+ int i, err;
+ int ticks;
+
+ ticks = ctx->ifc_sysctl_tx_reclaim_ticks;
+ err = sysctl_handle_int(oidp, &ticks, arg2, req);
+ if (err != 0) {
+ return err;
+ }
+
+ if (ticks == ctx->ifc_sysctl_tx_reclaim_ticks)
+ return 0;
+
+ if (ticks > hz) {
+ device_printf(ctx->ifc_dev,
+ "TX Reclaim ticks must be <= hz (%d)\n", hz);
+ return (EINVAL);
+ }
+
+ ctx->ifc_sysctl_tx_reclaim_ticks = ticks;
+ if (ctx->ifc_txqs == NULL)
+ return (err);
+
+ txq = &ctx->ifc_txqs[0];
+ for (i = 0; i < NTXQSETS(ctx); i++, txq++) {
+ txq->ift_reclaim_ticks = ticks;
+ }
+ return (err);
+}
+
#define NAME_BUFLEN 32
static void
iflib_add_device_sysctl_pre(if_ctx_t ctx)
@@ -6860,6 +6898,16 @@ iflib_add_device_sysctl_post(if_ctx_t ctx)
node = ctx->ifc_sysctl_node;
child = SYSCTL_CHILDREN(node);
+ SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "tx_reclaim_thresh",
+ CTLTYPE_INT | CTLFLAG_RWTUN, ctx,
+ 0, iflib_handle_tx_reclaim_thresh, "I",
+ "Number of TX descs outstanding before reclaim is called");
+
+ SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "tx_reclaim_ticks",
+ CTLTYPE_INT | CTLFLAG_RWTUN, ctx,
+ 0, iflib_handle_tx_reclaim_ticks, "I",
+ "Number of ticks before a TX reclaim is forced");
+
if (scctx->isc_ntxqsets > 100)
qfmt = "txq%03d";
else if (scctx->isc_ntxqsets > 10)
@@ -7107,7 +7155,7 @@ iflib_debugnet_poll(if_t ifp, int count)
return (EBUSY);
txq = &ctx->ifc_txqs[0];
- (void)iflib_completed_tx_reclaim(txq, RECLAIM_THRESH(ctx));
+ (void)iflib_completed_tx_reclaim(txq);
NET_EPOCH_ENTER(et);
for (i = 0; i < scctx->isc_nrxqsets; i++)
@@ -7117,7 +7165,7 @@ iflib_debugnet_poll(if_t ifp, int count)
}
#endif /* DEBUGNET */
-
+#ifndef ALTQ
static inline iflib_txq_t
iflib_simple_select_queue(if_ctx_t ctx, struct mbuf *m)
{
@@ -7140,9 +7188,13 @@ iflib_simple_transmit(if_t ifp, struct mbuf *m)
ctx = if_getsoftc(ifp);
- if ((if_getdrvflags(ifp) & (IFF_DRV_RUNNING | IFF_DRV_OACTIVE)) !=
- IFF_DRV_RUNNING)
- return (EBUSY);
+ if (__predict_false((if_getdrvflags(ifp) & IFF_DRV_RUNNING) == 0
+ || !LINK_ACTIVE(ctx))) {
+ DBG_COUNTER_INC(tx_frees);
+ m_freem(m);
+ return (ENETDOWN);
+ }
+
txq = iflib_simple_select_queue(ctx, m);
mtx_lock(&txq->ift_mtx);
error = iflib_encap(txq, &m);
@@ -7157,7 +7209,7 @@ iflib_simple_transmit(if_t ifp, struct mbuf *m)
else
if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
}
- (void)iflib_completed_tx_reclaim(txq, RECLAIM_THRESH(ctx));
+ (void)iflib_completed_tx_reclaim(txq);
mtx_unlock(&txq->ift_mtx);
if_inc_counter(ifp, IFCOUNTER_OBYTES, bytes_sent);
if_inc_counter(ifp, IFCOUNTER_OPACKETS, pkt_sent);
@@ -7166,3 +7218,4 @@ iflib_simple_transmit(if_t ifp, struct mbuf *m)
return (error);
}
+#endif
diff --git a/sys/net/iflib.h b/sys/net/iflib.h
index 3817445228d0..e65c936fc4b4 100644
--- a/sys/net/iflib.h
+++ b/sys/net/iflib.h
@@ -272,7 +272,7 @@ struct if_shared_ctx {
int isc_ntxqs; /* # of tx queues per tx qset - usually 1 */
int isc_nrxqs; /* # of rx queues per rx qset - intel 1, chelsio 2, broadcom 3 */
int __spare0__;
- int isc_tx_reclaim_thresh;
+ int __spare1__;
int isc_flags;
};
diff --git a/sys/net/pfvar.h b/sys/net/pfvar.h
index 8aefe514946e..ce266a267f3c 100644
--- a/sys/net/pfvar.h
+++ b/sys/net/pfvar.h
@@ -2038,14 +2038,15 @@ struct pfioc_trans {
} *array;
};
-#define PFR_FLAG_ATOMIC 0x00000001 /* unused */
+#define PFR_FLAG_START 0x00000001
#define PFR_FLAG_DUMMY 0x00000002
#define PFR_FLAG_FEEDBACK 0x00000004
#define PFR_FLAG_CLSTATS 0x00000008
#define PFR_FLAG_ADDRSTOO 0x00000010
#define PFR_FLAG_REPLACE 0x00000020
#define PFR_FLAG_ALLRSETS 0x00000040
-#define PFR_FLAG_ALLMASK 0x0000007F
+#define PFR_FLAG_DONE 0x00000080
+#define PFR_FLAG_ALLMASK 0x000000FF
#ifdef _KERNEL
#define PFR_FLAG_USERIOCTL 0x10000000
#endif
@@ -2435,6 +2436,7 @@ extern struct pf_ksrc_node *pf_find_src_node(struct pf_addr *,
struct pf_srchash **, pf_sn_types_t, bool);
extern void pf_unlink_src_node(struct pf_ksrc_node *);
extern u_int pf_free_src_nodes(struct pf_ksrc_node_list *);
+extern void pf_free_src_node(struct pf_ksrc_node *);
extern void pf_print_state(struct pf_kstate *);
extern void pf_print_flags(uint16_t);
extern int pf_addr_wrap_neq(struct pf_addr_wrap *,
@@ -2521,6 +2523,9 @@ uint16_t pf_qname2qid(const char *, bool);
void pfr_initialize(void);
void pfr_cleanup(void);
+struct pfr_kentry *
+ pfr_kentry_byaddr(struct pfr_ktable *, struct pf_addr *, sa_family_t,
+ int);
int pfr_match_addr(struct pfr_ktable *, struct pf_addr *, sa_family_t);
void pfr_update_stats(struct pfr_ktable *, struct pf_addr *, sa_family_t,
u_int64_t, int, int, int);
diff --git a/sys/net/route.c b/sys/net/route.c
index 7a50bcc43e06..d2c9f3e39c17 100644
--- a/sys/net/route.c
+++ b/sys/net/route.c
@@ -89,7 +89,7 @@ static int rt_ifdelroute(const struct rtentry *rt, const struct nhop_object *,
* SI_ORDER_MIDDLE.
*/
static void
-route_init(void)
+route_init(void *dummy __unused)
{
nhops_init();
diff --git a/sys/net/route/route_tables.c b/sys/net/route/route_tables.c
index 176ca43fa1c5..3b7bb1385d0e 100644
--- a/sys/net/route/route_tables.c
+++ b/sys/net/route/route_tables.c
@@ -186,7 +186,7 @@ rtables_prison_destructor(void *data)
}
static void
-rtables_init(void)
+rtables_init(void *dummy __unused)
{
osd_method_t methods[PR_MAXMETHOD] = {
[PR_METHOD_ATTACH] = rtables_check_proc_fib,
diff --git a/sys/net/rtsock.c b/sys/net/rtsock.c
index f0dcc973ca7c..be858428bb3e 100644
--- a/sys/net/rtsock.c
+++ b/sys/net/rtsock.c
@@ -309,7 +309,7 @@ rtsock_notify_event(uint32_t fibnum, const struct rib_cmd_info *rc)
}
static void
-rtsock_init(void)
+rtsock_init(void *dummy __unused)
{
rtsbridge_orig_p = rtsock_callback_p;
rtsock_callback_p = &rtsbridge;
diff --git a/sys/net/sff8436.h b/sys/net/sff8436.h
index deed74c7cdb4..dbf5c69df832 100644
--- a/sys/net/sff8436.h
+++ b/sys/net/sff8436.h
@@ -151,7 +151,7 @@ enum {
* OM2 fiber, units of 1 m */
SFF_8436_LEN_OM1 = 145, /* Link length supported for 1310 nm
* 50um multi-mode fiber, units of 1m*/
- SFF_8436_LEN_ASM = 144, /* Link length of passive cable assembly
+ SFF_8436_LEN_ASM = 146, /* Link length of passive cable assembly
* Length is specified as in the INF
* 8074, units of 1m. 0 means this is
* not value assembly. Value of 255
generated by cgit v1.2.3 (git 2.25.1) at 2025-11-22 12:40:25 +0000