Page MenuHomeFreeBSD
Differential D8247

Do not install xemacs elisp
ClosedPublic
Actions

Authored by jrm on Oct 14 2016, 2:09 AM.
Tags
None
Referenced Files
Unknown Object (File)
Mon, Mar 31, 6:22 PM
Unknown Object (File)
Sun, Mar 16, 10:08 AM
Unknown Object (File)
Mar 8 2025, 8:36 AM
Unknown Object (File)
Mar 8 2025, 6:00 AM
Unknown Object (File)
Mar 8 2025, 1:09 AM
Unknown Object (File)
Mar 7 2025, 8:40 PM
Unknown Object (File)
Mar 7 2025, 2:44 PM
Unknown Object (File)
Mar 7 2025, 11:32 AM
View All Files
Subscribers
mat

Details

Reviewers
swills
AMDmi3
garga
Commits
rP424738: devel/git: Do not install elisp files for XEmacs
Summary

Do not install xemacs elisp

Test Plan

portlint -AC: OK
testport: OK (poudriere: 11.0-RELEASE-p1, amd64, default options)

Diff Detail

Repository
rP FreeBSD ports repository
Lint
Lint Not Applicable
Unit
Tests Not Applicable

Event Timeline

jrm updated this revision to Diff 21382.Oct 14 2016, 2:09 AM
jrm retitled this revision from to Use relative paths when creating symbolic links and remove empty post-stage target.
jrm updated this object.
jrm edited the test plan for this revision. (Show Details)
jrm added reviewers: swills, AMDmi3, garga.
Herald added a subscriber: mat. · View Herald TranscriptOct 14 2016, 2:09 AM
mat added a comment.Oct 14 2016, 7:00 AM

We have not had xemacs in the ports tree for about two years, and xemacs.org is now no longer available. Maybe instead of fixing the symlinks (which is a great thing) removing all the xemacs thingies from the ports tree would be better :-)

jrm updated this revision to Diff 21386.Oct 14 2016, 12:06 PM

Do not install elisp for xemacs

Updating D8247: Use relative paths when creating symbolic links and remove empty post-stage

target

jrm retitled this revision from Use relative paths when creating symbolic links and remove empty post-stage target to Do not install xemacs elisp.Oct 14 2016, 12:08 PM
jrm updated this object.
jrm updated this revision to Diff 21387.Oct 14 2016, 12:09 PM

..and update pkg-plist

garga requested changes to this revision.Oct 25 2016, 10:40 PM
garga edited edge metadata.

Please bump PORTREVISION

This revision now requires changes to proceed.Oct 25 2016, 10:40 PM
jrm updated this revision to Diff 21682.Oct 25 2016, 11:07 PM
jrm edited edge metadata.

Bump PORTREVISION

garga accepted this revision.Oct 26 2016, 9:34 AM
garga edited edge metadata.

Looks good

This revision is now accepted and ready to land.Oct 26 2016, 9:34 AM
swills accepted this revision.Oct 26 2016, 9:08 PM
swills edited edge metadata.

Approved

Closed by commit rP424738: devel/git: Do not install elisp files for XEmacs (authored by jrm). · Explain WhyOct 26 2016, 9:32 PM
This revision was automatically updated to reflect the committed changes.

Revision Contents
Changeset List

PathSize
head/
devel/
git/
8 lines
2 lines

Diff 21709

View Options

head/devel/git/Makefile

/* $OpenBSD: if_trunk.c,v 1.30 2007/01/31 06:20:19 reyk Exp $ */ /* $OpenBSD: if_trunk.c,v 1.30 2007/01/31 06:20:19 reyk Exp $ */
/* /*
* Copyright (c) 2005, 2006 Reyk Floeter <reyk@openbsd.org> * Copyright (c) 2005, 2006 Reyk Floeter <reyk@openbsd.org>
* Copyright (c) 2007 Andrew Thompson <thompsa@FreeBSD.org> * Copyright (c) 2007 Andrew Thompson <thompsa@FreeBSD.org>
* Copyright (c) 2014 Marcelo Araujo <araujo@FreeBSD.org> * Copyright (c) 2014, 2016 Marcelo Araujo <araujo@FreeBSD.org>
* *
* Permission to use, copy, modify, and distribute this software for any * Permission to use, copy, modify, and distribute this software for any
* purpose with or without fee is hereby granted, provided that the above * purpose with or without fee is hereby granted, provided that the above
* copyright notice and this permission notice appear in all copies. * copyright notice and this permission notice appear in all copies.
* *
* THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
* WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
* MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
▲ Show 20 Lines • Show All 172 Lines • ▼ Show 20 Lines .pr_num = LAGG_PROTO_NONE
}, },
{ {
.pr_num = LAGG_PROTO_ROUNDROBIN, .pr_num = LAGG_PROTO_ROUNDROBIN,
.pr_attach = lagg_rr_attach, .pr_attach = lagg_rr_attach,
.pr_start = lagg_rr_start, .pr_start = lagg_rr_start,
.pr_input = lagg_rr_input, .pr_input = lagg_rr_input,
}, },
{ {
.pr_num = LAGG_PROTO_FAILOVER, .pr_num = LAGG_PROTO_FAILOVER,
thompsaUnsubmitted
Not Done
Inline Actions

This may want to be a sysctl function to check the bounds and also lower reset sc_bkt to the new value. Otherwise there is no way to correct a large count.

thompsa: This may want to be a sysctl function to check the bounds and also lower reset sc_bkt to the…
araujoAuthorUnsubmitted
Not Done
Inline Actions

Yes, in this case I totally agree.
I'm gonna change the patch to cover this case, but I'm not sure yet if a sysctl function is needed.

araujo: Yes, in this case I totally agree. I'm gonna change the patch to cover this case, but I'm not…
.pr_start = lagg_fail_start, .pr_start = lagg_fail_start,
.pr_input = lagg_fail_input, .pr_input = lagg_fail_input,
}, },
{ {
.pr_num = LAGG_PROTO_LOADBALANCE, .pr_num = LAGG_PROTO_LOADBALANCE,
.pr_attach = lagg_lb_attach, .pr_attach = lagg_lb_attach,
.pr_detach = lagg_lb_detach, .pr_detach = lagg_lb_detach,
.pr_start = lagg_lb_start, .pr_start = lagg_lb_start,
▲ Show 20 Lines • Show All 1,082 Lines • ▼ Show 20 Lines if (sc->sc_proto == LAGG_PROTO_LACP) {
ro->ro_opts |= LAGG_OPT_LACP_TIMEOUT; ro->ro_opts |= LAGG_OPT_LACP_TIMEOUT;
ro->ro_active = sc->sc_active; ro->ro_active = sc->sc_active;
} else { } else {
ro->ro_active = 0; ro->ro_active = 0;
SLIST_FOREACH(lp, &sc->sc_ports, lp_entries) SLIST_FOREACH(lp, &sc->sc_ports, lp_entries)
ro->ro_active += LAGG_PORTACTIVE(lp); ro->ro_active += LAGG_PORTACTIVE(lp);
} }
ro->ro_bkt = sc->sc_bkt;
ro->ro_flapping = sc->sc_flapping; ro->ro_flapping = sc->sc_flapping;
ro->ro_flowid_shift = sc->flowid_shift; ro->ro_flowid_shift = sc->flowid_shift;
break; break;
case SIOCSLAGGOPTS: case SIOCSLAGGOPTS:
if (sc->sc_proto == LAGG_PROTO_ROUNDROBIN) {
if (ro->ro_bkt == 0)
sc->sc_bkt = 1; // Minimum 1 packet per iface.
else
sc->sc_bkt = ro->ro_bkt;
}
error = priv_check(td, PRIV_NET_LAGG); error = priv_check(td, PRIV_NET_LAGG);
if (error) if (error)
break; break;
if (ro->ro_opts == 0) if (ro->ro_opts == 0)
break; break;
/* /*
* Set options. LACP options are stored in sc->sc_psc, * Set options. LACP options are stored in sc->sc_psc,
* not in sc_opts. * not in sc_opts.
Show All 18 Lines case -LAGG_OPT_LACP_TIMEOUT:
valid = lacp = 1; valid = lacp = 1;
break; break;
default: default:
valid = lacp = 0; valid = lacp = 0;
break; break;
} }
LAGG_WLOCK(sc); LAGG_WLOCK(sc);
if (valid == 0 || if (valid == 0 ||
(lacp == 1 && sc->sc_proto != LAGG_PROTO_LACP)) { (lacp == 1 && sc->sc_proto != LAGG_PROTO_LACP)) {
/* Invalid combination of options specified. */ /* Invalid combination of options specified. */
error = EINVAL; error = EINVAL;
LAGG_WUNLOCK(sc); LAGG_WUNLOCK(sc);
break; /* Return from SIOCSLAGGOPTS. */ break; /* Return from SIOCSLAGGOPTS. */
} }
/* /*
▲ Show 20 Lines • Show All 342 Lines • ▼ Show 20 Lineslagg_transmit(struct ifnet *ifp, struct mbuf *m)
if (error != 0) if (error != 0)
if_inc_counter(ifp, IFCOUNTER_OERRORS, 1); if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
return (error); return (error);
} }
/* /*
* The ifp->if_qflush entry point for lagg(4) is no-op. * The ifp->if_qflush entry point for lagg(4) is no-op.
thompsaUnsubmitted
Not Done
Inline Actions

This shouldnt be needed and may mess up the reference counting. Maybe add +1 in lagg_rr_init.

thompsa: This shouldnt be needed and may mess up the reference counting. Maybe add +1 in lagg_rr_init.
araujoAuthorUnsubmitted
Not Done
Inline Actions

Actually it is needed. Because lagg_rr_attach() is called only when we create the interface! If we create the interface and after set the lagg_rr_packets, at lagg_rr_start() it will never read the lagg_rr_packets again.

I can't see where it can mess up with the counter, at least in my tests it doesn't mess.

araujo: Actually it is needed. Because lagg_rr_attach() is called only when we create the interface! If…
*/ */
static void static void
lagg_qflush(struct ifnet *ifp __unused) lagg_qflush(struct ifnet *ifp __unused)
{ {
} }
static struct mbuf * static struct mbuf *
lagg_input(struct ifnet *ifp, struct mbuf *m) lagg_input(struct ifnet *ifp, struct mbuf *m)
{ {
struct lagg_port *lp = ifp->if_lagg; struct lagg_port *lp = ifp->if_lagg;
thompsaUnsubmitted
Not Done
Inline Actions

This is still not quite right. atomics are hard, someone else to review is needed.

if (lagg_rr_packets > 1) {
    if (atomic_fetchadd_32(&sc->sc_bkt, -1) == 1) {
      atomic_add_int(&sc->sc_bkt, lagg_rr_packets);
      p = atomic_fetchadd_32(&sc->sc_seq, 1);
    } else {
      p = sc->sc_seq;
   }
  } else
    p = atomic_fetchadd_32(&sc->sc_seq, 1);
thompsa: This is still not quite right. atomics are hard, someone else to review is needed. if…
adrianUnsubmitted
Not Done
Inline Actions

Yes, this also looks not quite right to me.

adrian: Yes, this also looks not quite right to me.
struct lagg_softc *sc = lp->lp_softc; struct lagg_softc *sc = lp->lp_softc;
struct ifnet *scifp = sc->sc_ifp; struct ifnet *scifp = sc->sc_ifp;
struct rm_priotracker tracker; struct rm_priotracker tracker;
LAGG_RLOCK(sc, &tracker); LAGG_RLOCK(sc, &tracker);
if ((scifp->if_drv_flags & IFF_DRV_RUNNING) == 0 || if ((scifp->if_drv_flags & IFF_DRV_RUNNING) == 0 ||
(lp->lp_flags & LAGG_PORT_DISABLED) || (lp->lp_flags & LAGG_PORT_DISABLED) ||
sc->sc_proto == LAGG_PROTO_NONE) { sc->sc_proto == LAGG_PROTO_NONE) {
▲ Show 20 Lines • Show All 165 Lines • ▼ Show 20 Lines
/* /*
* Simple round robin aggregation * Simple round robin aggregation
*/ */
static void static void
lagg_rr_attach(struct lagg_softc *sc) lagg_rr_attach(struct lagg_softc *sc)
{ {
sc->sc_capabilities = IFCAP_LAGG_FULLDUPLEX; sc->sc_capabilities = IFCAP_LAGG_FULLDUPLEX;
sc->sc_seq = 0; sc->sc_seq = 0;
sc->sc_bkt_count = sc->sc_bkt;
} }
static int static int
lagg_rr_start(struct lagg_softc *sc, struct mbuf *m) lagg_rr_start(struct lagg_softc *sc, struct mbuf *m)
{ {
struct lagg_port *lp; struct lagg_port *lp;
uint32_t p; uint32_t p;
if (sc->sc_bkt_count == 0 && sc->sc_bkt > 0)
sc->sc_bkt_count = sc->sc_bkt;
if (sc->sc_bkt > 0) {
atomic_subtract_int(&sc->sc_bkt_count, 1);
if (atomic_cmpset_int(&sc->sc_bkt_count, 0, sc->sc_bkt))
p = atomic_fetchadd_32(&sc->sc_seq, 1); p = atomic_fetchadd_32(&sc->sc_seq, 1);
else
p = sc->sc_seq;
} else
p = atomic_fetchadd_32(&sc->sc_seq, 1);
p %= sc->sc_count; p %= sc->sc_count;
lp = SLIST_FIRST(&sc->sc_ports); lp = SLIST_FIRST(&sc->sc_ports);
while (p--) while (p--)
lp = SLIST_NEXT(lp, lp_entries); lp = SLIST_NEXT(lp, lp_entries);
/* /*
* Check the port's link state. This will return the next active * Check the port's link state. This will return the next active
* port if the link is down or the port is NULL. * port if the link is down or the port is NULL.
*/ */
if ((lp = lagg_link_active(sc, lp)) == NULL) { if ((lp = lagg_link_active(sc, lp)) == NULL) {
Show All 32 Lines SLIST_FOREACH(lp, &sc->sc_ports, lp_entries) {
if (!LAGG_PORTACTIVE(lp)) if (!LAGG_PORTACTIVE(lp))
continue; continue;
active_ports++; active_ports++;
if (last != NULL) { if (last != NULL) {
m0 = m_copym(m, 0, M_COPYALL, M_NOWAIT); m0 = m_copym(m, 0, M_COPYALL, M_NOWAIT);
if (m0 == NULL) { if (m0 == NULL) {
ret = ENOBUFS; ret = ENOBUFS;
thompsaUnsubmitted
Not Done
Inline Actions

This is overly complicated and also is doing string operations in the fast path. This could all be avoided by a bucket count and only incrementing sc_seq when the bucket is empty.

thompsa: This is overly complicated and also is doing string operations in the fast path. This could all…
errors++; errors++;
break; break;
} }
ret = lagg_enqueue(last->lp_ifp, m0); ret = lagg_enqueue(last->lp_ifp, m0);
if (ret != 0) if (ret != 0)
errors++; errors++;
} }
▲ Show 20 Lines • Show All 270 Lines • Show Last 20 Lines
View Options

head/devel/git/pkg-plist

/* $OpenBSD: if_trunk.c,v 1.30 2007/01/31 06:20:19 reyk Exp $ */ /* $OpenBSD: if_trunk.c,v 1.30 2007/01/31 06:20:19 reyk Exp $ */
/* /*
* Copyright (c) 2005, 2006 Reyk Floeter <reyk@openbsd.org> * Copyright (c) 2005, 2006 Reyk Floeter <reyk@openbsd.org>
* Copyright (c) 2007 Andrew Thompson <thompsa@FreeBSD.org> * Copyright (c) 2007 Andrew Thompson <thompsa@FreeBSD.org>
* Copyright (c) 2014 Marcelo Araujo <araujo@FreeBSD.org> * Copyright (c) 2014, 2016 Marcelo Araujo <araujo@FreeBSD.org>
* *
* Permission to use, copy, modify, and distribute this software for any * Permission to use, copy, modify, and distribute this software for any
* purpose with or without fee is hereby granted, provided that the above * purpose with or without fee is hereby granted, provided that the above
* copyright notice and this permission notice appear in all copies. * copyright notice and this permission notice appear in all copies.
* *
* THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
* WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
* MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
▲ Show 20 Lines • Show All 172 Lines • ▼ Show 20 Lines .pr_num = LAGG_PROTO_NONE
}, },
{ {
.pr_num = LAGG_PROTO_ROUNDROBIN, .pr_num = LAGG_PROTO_ROUNDROBIN,
.pr_attach = lagg_rr_attach, .pr_attach = lagg_rr_attach,
.pr_start = lagg_rr_start, .pr_start = lagg_rr_start,
.pr_input = lagg_rr_input, .pr_input = lagg_rr_input,
}, },
{ {
.pr_num = LAGG_PROTO_FAILOVER, .pr_num = LAGG_PROTO_FAILOVER,
thompsaUnsubmitted
Not Done
Inline Actions

This may want to be a sysctl function to check the bounds and also lower reset sc_bkt to the new value. Otherwise there is no way to correct a large count.

thompsa: This may want to be a sysctl function to check the bounds and also lower reset sc_bkt to the…
araujoAuthorUnsubmitted
Not Done
Inline Actions

Yes, in this case I totally agree.
I'm gonna change the patch to cover this case, but I'm not sure yet if a sysctl function is needed.

araujo: Yes, in this case I totally agree. I'm gonna change the patch to cover this case, but I'm not…
.pr_start = lagg_fail_start, .pr_start = lagg_fail_start,
.pr_input = lagg_fail_input, .pr_input = lagg_fail_input,
}, },
{ {
.pr_num = LAGG_PROTO_LOADBALANCE, .pr_num = LAGG_PROTO_LOADBALANCE,
.pr_attach = lagg_lb_attach, .pr_attach = lagg_lb_attach,
.pr_detach = lagg_lb_detach, .pr_detach = lagg_lb_detach,
.pr_start = lagg_lb_start, .pr_start = lagg_lb_start,
▲ Show 20 Lines • Show All 1,082 Lines • ▼ Show 20 Lines if (sc->sc_proto == LAGG_PROTO_LACP) {
ro->ro_opts |= LAGG_OPT_LACP_TIMEOUT; ro->ro_opts |= LAGG_OPT_LACP_TIMEOUT;
ro->ro_active = sc->sc_active; ro->ro_active = sc->sc_active;
} else { } else {
ro->ro_active = 0; ro->ro_active = 0;
SLIST_FOREACH(lp, &sc->sc_ports, lp_entries) SLIST_FOREACH(lp, &sc->sc_ports, lp_entries)
ro->ro_active += LAGG_PORTACTIVE(lp); ro->ro_active += LAGG_PORTACTIVE(lp);
} }
ro->ro_bkt = sc->sc_bkt;
ro->ro_flapping = sc->sc_flapping; ro->ro_flapping = sc->sc_flapping;
ro->ro_flowid_shift = sc->flowid_shift; ro->ro_flowid_shift = sc->flowid_shift;
break; break;
case SIOCSLAGGOPTS: case SIOCSLAGGOPTS:
if (sc->sc_proto == LAGG_PROTO_ROUNDROBIN) {
if (ro->ro_bkt == 0)
sc->sc_bkt = 1; // Minimum 1 packet per iface.
else
sc->sc_bkt = ro->ro_bkt;
}
error = priv_check(td, PRIV_NET_LAGG); error = priv_check(td, PRIV_NET_LAGG);
if (error) if (error)
break; break;
if (ro->ro_opts == 0) if (ro->ro_opts == 0)
break; break;
/* /*
* Set options. LACP options are stored in sc->sc_psc, * Set options. LACP options are stored in sc->sc_psc,
* not in sc_opts. * not in sc_opts.
Show All 18 Lines case -LAGG_OPT_LACP_TIMEOUT:
valid = lacp = 1; valid = lacp = 1;
break; break;
default: default:
valid = lacp = 0; valid = lacp = 0;
break; break;
} }
LAGG_WLOCK(sc); LAGG_WLOCK(sc);
if (valid == 0 || if (valid == 0 ||
(lacp == 1 && sc->sc_proto != LAGG_PROTO_LACP)) { (lacp == 1 && sc->sc_proto != LAGG_PROTO_LACP)) {
/* Invalid combination of options specified. */ /* Invalid combination of options specified. */
error = EINVAL; error = EINVAL;
LAGG_WUNLOCK(sc); LAGG_WUNLOCK(sc);
break; /* Return from SIOCSLAGGOPTS. */ break; /* Return from SIOCSLAGGOPTS. */
} }
/* /*
▲ Show 20 Lines • Show All 342 Lines • ▼ Show 20 Lineslagg_transmit(struct ifnet *ifp, struct mbuf *m)
if (error != 0) if (error != 0)
if_inc_counter(ifp, IFCOUNTER_OERRORS, 1); if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
return (error); return (error);
} }
/* /*
* The ifp->if_qflush entry point for lagg(4) is no-op. * The ifp->if_qflush entry point for lagg(4) is no-op.
thompsaUnsubmitted
Not Done
Inline Actions

This shouldnt be needed and may mess up the reference counting. Maybe add +1 in lagg_rr_init.

thompsa: This shouldnt be needed and may mess up the reference counting. Maybe add +1 in lagg_rr_init.
araujoAuthorUnsubmitted
Not Done
Inline Actions

Actually it is needed. Because lagg_rr_attach() is called only when we create the interface! If we create the interface and after set the lagg_rr_packets, at lagg_rr_start() it will never read the lagg_rr_packets again.

I can't see where it can mess up with the counter, at least in my tests it doesn't mess.

araujo: Actually it is needed. Because lagg_rr_attach() is called only when we create the interface! If…
*/ */
static void static void
lagg_qflush(struct ifnet *ifp __unused) lagg_qflush(struct ifnet *ifp __unused)
{ {
} }
static struct mbuf * static struct mbuf *
lagg_input(struct ifnet *ifp, struct mbuf *m) lagg_input(struct ifnet *ifp, struct mbuf *m)
{ {
struct lagg_port *lp = ifp->if_lagg; struct lagg_port *lp = ifp->if_lagg;
thompsaUnsubmitted
Not Done
Inline Actions

This is still not quite right. atomics are hard, someone else to review is needed.

if (lagg_rr_packets > 1) {
    if (atomic_fetchadd_32(&sc->sc_bkt, -1) == 1) {
      atomic_add_int(&sc->sc_bkt, lagg_rr_packets);
      p = atomic_fetchadd_32(&sc->sc_seq, 1);
    } else {
      p = sc->sc_seq;
   }
  } else
    p = atomic_fetchadd_32(&sc->sc_seq, 1);
thompsa: This is still not quite right. atomics are hard, someone else to review is needed. if…
adrianUnsubmitted
Not Done
Inline Actions

Yes, this also looks not quite right to me.

adrian: Yes, this also looks not quite right to me.
struct lagg_softc *sc = lp->lp_softc; struct lagg_softc *sc = lp->lp_softc;
struct ifnet *scifp = sc->sc_ifp; struct ifnet *scifp = sc->sc_ifp;
struct rm_priotracker tracker; struct rm_priotracker tracker;
LAGG_RLOCK(sc, &tracker); LAGG_RLOCK(sc, &tracker);
if ((scifp->if_drv_flags & IFF_DRV_RUNNING) == 0 || if ((scifp->if_drv_flags & IFF_DRV_RUNNING) == 0 ||
(lp->lp_flags & LAGG_PORT_DISABLED) || (lp->lp_flags & LAGG_PORT_DISABLED) ||
sc->sc_proto == LAGG_PROTO_NONE) { sc->sc_proto == LAGG_PROTO_NONE) {
▲ Show 20 Lines • Show All 165 Lines • ▼ Show 20 Lines
/* /*
* Simple round robin aggregation * Simple round robin aggregation
*/ */
static void static void
lagg_rr_attach(struct lagg_softc *sc) lagg_rr_attach(struct lagg_softc *sc)
{ {
sc->sc_capabilities = IFCAP_LAGG_FULLDUPLEX; sc->sc_capabilities = IFCAP_LAGG_FULLDUPLEX;
sc->sc_seq = 0; sc->sc_seq = 0;
sc->sc_bkt_count = sc->sc_bkt;
} }
static int static int
lagg_rr_start(struct lagg_softc *sc, struct mbuf *m) lagg_rr_start(struct lagg_softc *sc, struct mbuf *m)
{ {
struct lagg_port *lp; struct lagg_port *lp;
uint32_t p; uint32_t p;
if (sc->sc_bkt_count == 0 && sc->sc_bkt > 0)
sc->sc_bkt_count = sc->sc_bkt;
if (sc->sc_bkt > 0) {
atomic_subtract_int(&sc->sc_bkt_count, 1);
if (atomic_cmpset_int(&sc->sc_bkt_count, 0, sc->sc_bkt))
p = atomic_fetchadd_32(&sc->sc_seq, 1); p = atomic_fetchadd_32(&sc->sc_seq, 1);
else
p = sc->sc_seq;
} else
p = atomic_fetchadd_32(&sc->sc_seq, 1);
p %= sc->sc_count; p %= sc->sc_count;
lp = SLIST_FIRST(&sc->sc_ports); lp = SLIST_FIRST(&sc->sc_ports);
while (p--) while (p--)
lp = SLIST_NEXT(lp, lp_entries); lp = SLIST_NEXT(lp, lp_entries);
/* /*
* Check the port's link state. This will return the next active * Check the port's link state. This will return the next active
* port if the link is down or the port is NULL. * port if the link is down or the port is NULL.
*/ */
if ((lp = lagg_link_active(sc, lp)) == NULL) { if ((lp = lagg_link_active(sc, lp)) == NULL) {
Show All 32 Lines SLIST_FOREACH(lp, &sc->sc_ports, lp_entries) {
if (!LAGG_PORTACTIVE(lp)) if (!LAGG_PORTACTIVE(lp))
continue; continue;
active_ports++; active_ports++;
if (last != NULL) { if (last != NULL) {
m0 = m_copym(m, 0, M_COPYALL, M_NOWAIT); m0 = m_copym(m, 0, M_COPYALL, M_NOWAIT);
if (m0 == NULL) { if (m0 == NULL) {
ret = ENOBUFS; ret = ENOBUFS;
thompsaUnsubmitted
Not Done
Inline Actions

This is overly complicated and also is doing string operations in the fast path. This could all be avoided by a bucket count and only incrementing sc_seq when the bucket is empty.

thompsa: This is overly complicated and also is doing string operations in the fast path. This could all…
errors++; errors++;
break; break;
} }
ret = lagg_enqueue(last->lp_ifp, m0); ret = lagg_enqueue(last->lp_ifp, m0);
if (ret != 0) if (ret != 0)
errors++; errors++;
} }
▲ Show 20 Lines • Show All 270 Lines • Show Last 20 Lines