summaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorLuiz Otavio O Souza <luiz@netgate.com>2016-04-17 23:10:07 -0500
committerLuiz Otavio O Souza <luiz@netgate.com>2016-04-17 23:10:07 -0500
commitd83ceefd74a8cd372bb5f899a80bd540ceca8c8f (patch)
treeae4793ba52720ca9310b957817e256c9be09c514
parent0d7eb369f4ec7afd37b400407072943d34864188 (diff)
parentdab85cf430c20f64caedfab11029a1c75239181c (diff)
downloadFreeBSD-src-d83ceefd74a8cd372bb5f899a80bd540ceca8c8f.zip
FreeBSD-src-d83ceefd74a8cd372bb5f899a80bd540ceca8c8f.tar.gz
Merge remote-tracking branch 'origin/stable/10' into devel
-rw-r--r--sbin/devd/devd.conf.522
-rw-r--r--sbin/pfctl/parse.y147
-rw-r--r--sbin/pfctl/pfctl_altq.c264
-rw-r--r--sbin/pfctl/pfctl_parser.h16
-rw-r--r--sbin/pfctl/pfctl_qstats.c68
-rw-r--r--share/man/man4/altq.412
-rw-r--r--sys/cam/ata/ata_all.c31
-rw-r--r--sys/cam/ata/ata_all.h1
-rw-r--r--sys/cam/cam_periph.c105
-rw-r--r--sys/cam/scsi/scsi_all.c48
-rw-r--r--sys/cam/scsi/scsi_all.h1
-rw-r--r--sys/conf/NOTES2
-rw-r--r--sys/conf/files2
-rw-r--r--sys/conf/options2
-rw-r--r--sys/contrib/altq/altq/altq.h4
-rw-r--r--sys/contrib/altq/altq/altq_cbq.c4
-rw-r--r--sys/contrib/altq/altq/altq_cbq.h5
-rw-r--r--sys/contrib/altq/altq/altq_classq.h6
-rw-r--r--sys/contrib/altq/altq/altq_codel.c477
-rw-r--r--sys/contrib/altq/altq/altq_codel.h129
-rw-r--r--sys/contrib/altq/altq/altq_fairq.c909
-rw-r--r--sys/contrib/altq/altq/altq_fairq.h145
-rw-r--r--sys/contrib/altq/altq/altq_hfsc.c36
-rw-r--r--sys/contrib/altq/altq/altq_hfsc.h12
-rw-r--r--sys/contrib/altq/altq/altq_priq.c41
-rw-r--r--sys/contrib/altq/altq/altq_priq.h12
-rw-r--r--sys/contrib/altq/altq/altq_rmclass.c31
-rw-r--r--sys/contrib/altq/altq/altq_rmclass.h8
-rw-r--r--sys/contrib/altq/altq/altq_subr.c50
-rw-r--r--sys/contrib/altq/altq/altq_var.h12
-rw-r--r--sys/netpfil/pf/pf.c18
-rw-r--r--sys/netpfil/pf/pf_altq.h22
-rw-r--r--sys/netpfil/pf/pf_mtag.h1
33 files changed, 2593 insertions, 50 deletions
diff --git a/sbin/devd/devd.conf.5 b/sbin/devd/devd.conf.5
index a174de8..3df30aa 100644
--- a/sbin/devd/devd.conf.5
+++ b/sbin/devd/devd.conf.5
@@ -41,7 +41,7 @@
.\" ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS
.\" SOFTWARE.
.\"
-.Dd July 11, 2015
+.Dd April 14, 2016
.Dt DEVD.CONF 5
.Os
.Sh NAME
@@ -492,6 +492,23 @@ Information about the state of the system.
Notification that the system has woken from the suspended state.
.El
.El
+.Pp
+.It Li CAM
+Events related to the
+.Xr cam 4
+system.
+.Bl -tag -width ".Sy Subsystem" -compact
+.It Sy Subsystem
+.It Li periph
+Events related to peripheral devices.
+.Bl -tag -width ".li timeout" -compact
+.It Sy Type
+.It Li error
+Generic errors.
+.It Li timeout
+Command timeouts.
+.El
+.El
.El
.Pp
A link state change to UP on the interface
@@ -621,4 +638,5 @@ has many additional examples.
.Xr coretemp 4 ,
.Xr devfs 5 ,
.Xr re_format 7 ,
-.Xr devd 8
+.Xr devd 8 ,
+.Xr cam 4
diff --git a/sbin/pfctl/parse.y b/sbin/pfctl/parse.y
index 3edb634..b5577e2 100644
--- a/sbin/pfctl/parse.y
+++ b/sbin/pfctl/parse.y
@@ -48,8 +48,10 @@ __FBSDID("$FreeBSD$");
#include <arpa/inet.h>
#include <altq/altq.h>
#include <altq/altq_cbq.h>
+#include <altq/altq_codel.h>
#include <altq/altq_priq.h>
#include <altq/altq_hfsc.h>
+#include <altq/altq_fairq.h>
#include <stdio.h>
#include <unistd.h>
@@ -315,8 +317,9 @@ struct pool_opts {
} pool_opts;
-
+struct codel_opts codel_opts;
struct node_hfsc_opts hfsc_opts;
+struct node_fairq_opts fairq_opts;
struct node_state_opt *keep_state_defaults = NULL;
int disallow_table(struct node_host *, const char *);
@@ -440,6 +443,8 @@ typedef struct {
struct table_opts table_opts;
struct pool_opts pool_opts;
struct node_hfsc_opts hfsc_opts;
+ struct node_fairq_opts fairq_opts;
+ struct codel_opts codel_opts;
} v;
int lineno;
} YYSTYPE;
@@ -464,8 +469,8 @@ int parseport(char *, struct range *r, int);
%token REQUIREORDER SYNPROXY FINGERPRINTS NOSYNC DEBUG SKIP HOSTID
%token ANTISPOOF FOR INCLUDE
%token BITMASK RANDOM SOURCEHASH ROUNDROBIN STATICPORT PROBABILITY
-%token ALTQ CBQ PRIQ HFSC BANDWIDTH TBRSIZE LINKSHARE REALTIME UPPERLIMIT
-%token QUEUE PRIORITY QLIMIT RTABLE
+%token ALTQ CBQ CODEL PRIQ HFSC FAIRQ BANDWIDTH TBRSIZE LINKSHARE REALTIME
+%token UPPERLIMIT QUEUE PRIORITY QLIMIT HOGS BUCKETS RTABLE TARGET INTERVAL
%token DNPIPE DNQUEUE
%token LOAD RULESET_OPTIMIZATION
%token STICKYADDRESS MAXSRCSTATES MAXSRCNODES SOURCETRACK GLOBAL RULE
@@ -515,6 +520,8 @@ int parseport(char *, struct range *r, int);
%type <v.number> cbqflags_list cbqflags_item
%type <v.number> priqflags_list priqflags_item
%type <v.hfsc_opts> hfscopts_list hfscopts_item hfsc_opts
+%type <v.fairq_opts> fairqopts_list fairqopts_item fairq_opts
+%type <v.codel_opts> codelopts_list codelopts_item codel_opts
%type <v.queue_bwspec> bandwidth
%type <v.filter_opts> filter_opts filter_opt filter_opts_l
%type <v.antispoof_opts> antispoof_opts antispoof_opt antispoof_opts_l
@@ -1496,7 +1503,7 @@ altqif : ALTQ interface queue_opts QUEUE qassign {
a.scheduler = $3.scheduler.qtype;
a.qlimit = $3.qlimit;
a.tbrsize = $3.tbrsize;
- if ($5 == NULL) {
+ if ($5 == NULL && $3.scheduler.qtype != ALTQT_CODEL) {
yyerror("no child queues specified");
YYERROR;
}
@@ -1697,6 +1704,24 @@ scheduler : CBQ {
$$.qtype = ALTQT_HFSC;
$$.data.hfsc_opts = $3;
}
+ | FAIRQ {
+ $$.qtype = ALTQT_FAIRQ;
+ bzero(&$$.data.fairq_opts,
+ sizeof(struct node_fairq_opts));
+ }
+ | FAIRQ '(' fairq_opts ')' {
+ $$.qtype = ALTQT_FAIRQ;
+ $$.data.fairq_opts = $3;
+ }
+ | CODEL {
+ $$.qtype = ALTQT_CODEL;
+ bzero(&$$.data.codel_opts,
+ sizeof(struct codel_opts));
+ }
+ | CODEL '(' codel_opts ')' {
+ $$.qtype = ALTQT_CODEL;
+ $$.data.codel_opts = $3;
+ }
;
cbqflags_list : cbqflags_item { $$ |= $1; }
@@ -1714,6 +1739,8 @@ cbqflags_item : STRING {
$$ = CBQCLF_RED|CBQCLF_ECN;
else if (!strcmp($1, "rio"))
$$ = CBQCLF_RIO;
+ else if (!strcmp($1, "codel"))
+ $$ = CBQCLF_CODEL;
else {
yyerror("unknown cbq flag \"%s\"", $1);
free($1);
@@ -1736,6 +1763,8 @@ priqflags_item : STRING {
$$ = PRCF_RED|PRCF_ECN;
else if (!strcmp($1, "rio"))
$$ = PRCF_RIO;
+ else if (!strcmp($1, "codel"))
+ $$ = PRCF_CODEL;
else {
yyerror("unknown priq flag \"%s\"", $1);
free($1);
@@ -1836,6 +1865,8 @@ hfscopts_item : LINKSHARE bandwidth {
hfsc_opts.flags |= HFCF_RED|HFCF_ECN;
else if (!strcmp($1, "rio"))
hfsc_opts.flags |= HFCF_RIO;
+ else if (!strcmp($1, "codel"))
+ hfsc_opts.flags |= HFCF_CODEL;
else {
yyerror("unknown hfsc flag \"%s\"", $1);
free($1);
@@ -1845,6 +1876,102 @@ hfscopts_item : LINKSHARE bandwidth {
}
;
+fairq_opts : {
+ bzero(&fairq_opts,
+ sizeof(struct node_fairq_opts));
+ }
+ fairqopts_list {
+ $$ = fairq_opts;
+ }
+ ;
+
+fairqopts_list : fairqopts_item
+ | fairqopts_list comma fairqopts_item
+ ;
+
+fairqopts_item : LINKSHARE bandwidth {
+ if (fairq_opts.linkshare.used) {
+ yyerror("linkshare already specified");
+ YYERROR;
+ }
+ fairq_opts.linkshare.m2 = $2;
+ fairq_opts.linkshare.used = 1;
+ }
+ | LINKSHARE '(' bandwidth number bandwidth ')' {
+ if (fairq_opts.linkshare.used) {
+ yyerror("linkshare already specified");
+ YYERROR;
+ }
+ fairq_opts.linkshare.m1 = $3;
+ fairq_opts.linkshare.d = $4;
+ fairq_opts.linkshare.m2 = $5;
+ fairq_opts.linkshare.used = 1;
+ }
+ | HOGS bandwidth {
+ fairq_opts.hogs_bw = $2;
+ }
+ | BUCKETS number {
+ fairq_opts.nbuckets = $2;
+ }
+ | STRING {
+ if (!strcmp($1, "default"))
+ fairq_opts.flags |= FARF_DEFAULTCLASS;
+ else if (!strcmp($1, "red"))
+ fairq_opts.flags |= FARF_RED;
+ else if (!strcmp($1, "ecn"))
+ fairq_opts.flags |= FARF_RED|FARF_ECN;
+ else if (!strcmp($1, "rio"))
+ fairq_opts.flags |= FARF_RIO;
+ else if (!strcmp($1, "codel"))
+ fairq_opts.flags |= FARF_CODEL;
+ else {
+ yyerror("unknown fairq flag \"%s\"", $1);
+ free($1);
+ YYERROR;
+ }
+ free($1);
+ }
+ ;
+
+codel_opts : {
+ bzero(&codel_opts,
+ sizeof(struct codel_opts));
+ }
+ codelopts_list {
+ $$ = codel_opts;
+ }
+ ;
+
+codelopts_list : codelopts_item
+ | codelopts_list comma codelopts_item
+ ;
+
+codelopts_item : INTERVAL number {
+ if (codel_opts.interval) {
+ yyerror("interval already specified");
+ YYERROR;
+ }
+ codel_opts.interval = $2;
+ }
+ | TARGET number {
+ if (codel_opts.target) {
+ yyerror("target already specified");
+ YYERROR;
+ }
+ codel_opts.target = $2;
+ }
+ | STRING {
+ if (!strcmp($1, "ecn"))
+ codel_opts.ecn = 1;
+ else {
+ yyerror("unknown codel option \"%s\"", $1);
+ free($1);
+ YYERROR;
+ }
+ free($1);
+ }
+ ;
+
qassign : /* empty */ { $$ = NULL; }
| qassign_item { $$ = $1; }
| '{' optnl qassign_list '}' { $$ = $3; }
@@ -5014,7 +5141,8 @@ expand_altq(struct pf_altq *a, struct node_if *interfaces,
if ((pf->loadopt & PFCTL_FLAG_ALTQ) == 0) {
FREE_LIST(struct node_if, interfaces);
- FREE_LIST(struct node_queue, nqueues);
+ if (nqueues)
+ FREE_LIST(struct node_queue, nqueues);
return (0);
}
@@ -5105,7 +5233,8 @@ expand_altq(struct pf_altq *a, struct node_if *interfaces,
}
);
FREE_LIST(struct node_if, interfaces);
- FREE_LIST(struct node_queue, nqueues);
+ if (nqueues)
+ FREE_LIST(struct node_queue, nqueues);
return (errs);
}
@@ -5516,8 +5645,10 @@ lookup(char *s)
{ "bitmask", BITMASK},
{ "block", BLOCK},
{ "block-policy", BLOCKPOLICY},
+ { "buckets", BUCKETS},
{ "cbq", CBQ},
{ "code", CODE},
+ { "codelq", CODEL},
{ "crop", FRAGCROP},
{ "debug", DEBUG},
{ "divert-reply", DIVERTREPLY},
@@ -5528,6 +5659,7 @@ lookup(char *s)
{ "drop-ovl", FRAGDROP},
{ "dscp", DSCP},
{ "dup-to", DUPTO},
+ { "fairq", FAIRQ},
{ "fastroute", FASTROUTE},
{ "file", FILENAME},
{ "fingerprints", FINGERPRINTS},
@@ -5540,6 +5672,7 @@ lookup(char *s)
{ "global", GLOBAL},
{ "group", GROUP},
{ "hfsc", HFSC},
+ { "hogs", HOGS},
{ "hostid", HOSTID},
{ "icmp-type", ICMPTYPE},
{ "icmp6-type", ICMP6TYPE},
@@ -5550,6 +5683,7 @@ lookup(char *s)
{ "include", INCLUDE},
{ "inet", INET},
{ "inet6", INET6},
+ { "interval", INTERVAL},
{ "keep", KEEP},
{ "label", LABEL},
{ "limit", LIMIT},
@@ -5622,6 +5756,7 @@ lookup(char *s)
{ "table", TABLE},
{ "tag", TAG},
{ "tagged", TAGGED},
+ { "target", TARGET},
{ "tbrsize", TBRSIZE},
{ "timeout", TIMEOUT},
{ "to", TO},
diff --git a/sbin/pfctl/pfctl_altq.c b/sbin/pfctl/pfctl_altq.c
index a99e641..9a21754 100644
--- a/sbin/pfctl/pfctl_altq.c
+++ b/sbin/pfctl/pfctl_altq.c
@@ -41,8 +41,10 @@ __FBSDID("$FreeBSD$");
#include <altq/altq.h>
#include <altq/altq_cbq.h>
+#include <altq/altq_codel.h>
#include <altq/altq_priq.h>
#include <altq/altq_hfsc.h>
+#include <altq/altq_fairq.h>
#include "pfctl_parser.h"
#include "pfctl.h"
@@ -60,6 +62,9 @@ static int cbq_compute_idletime(struct pfctl *, struct pf_altq *);
static int check_commit_cbq(int, int, struct pf_altq *);
static int print_cbq_opts(const struct pf_altq *);
+static int print_codel_opts(const struct pf_altq *,
+ const struct node_queue_opt *);
+
static int eval_pfqueue_priq(struct pfctl *, struct pf_altq *);
static int check_commit_priq(int, int, struct pf_altq *);
static int print_priq_opts(const struct pf_altq *);
@@ -69,6 +74,11 @@ static int check_commit_hfsc(int, int, struct pf_altq *);
static int print_hfsc_opts(const struct pf_altq *,
const struct node_queue_opt *);
+static int eval_pfqueue_fairq(struct pfctl *, struct pf_altq *);
+static int print_fairq_opts(const struct pf_altq *,
+ const struct node_queue_opt *);
+static int check_commit_fairq(int, int, struct pf_altq *);
+
static void gsc_add_sc(struct gen_sc *, struct service_curve *);
static int is_gsc_under_sc(struct gen_sc *,
struct service_curve *);
@@ -89,6 +99,8 @@ int eval_queue_opts(struct pf_altq *, struct node_queue_opt *,
u_int32_t eval_bwspec(struct node_queue_bw *, u_int32_t);
void print_hfsc_sc(const char *, u_int, u_int, u_int,
const struct node_hfsc_sc *);
+void print_fairq_sc(const char *, u_int, u_int, u_int,
+ const struct node_fairq_sc *);
void
pfaltq_store(struct pf_altq *a)
@@ -174,6 +186,14 @@ print_altq(const struct pf_altq *a, unsigned int level,
if (!print_hfsc_opts(a, qopts))
printf("hfsc ");
break;
+ case ALTQT_FAIRQ:
+ if (!print_fairq_opts(a, qopts))
+ printf("fairq ");
+ break;
+ case ALTQT_CODEL:
+ if (!print_codel_opts(a, qopts))
+ printf("codel ");
+ break;
}
if (bw != NULL && bw->bw_percent > 0) {
@@ -204,7 +224,8 @@ print_queue(const struct pf_altq *a, unsigned int level,
printf("%s ", a->qname);
if (print_interface)
printf("on %s ", a->ifname);
- if (a->scheduler == ALTQT_CBQ || a->scheduler == ALTQT_HFSC) {
+ if (a->scheduler == ALTQT_CBQ || a->scheduler == ALTQT_HFSC ||
+ a->scheduler == ALTQT_FAIRQ) {
if (bw != NULL && bw->bw_percent > 0) {
if (bw->bw_percent < 100)
printf("bandwidth %u%% ", bw->bw_percent);
@@ -225,6 +246,9 @@ print_queue(const struct pf_altq *a, unsigned int level,
case ALTQT_HFSC:
print_hfsc_opts(a, qopts);
break;
+ case ALTQT_FAIRQ:
+ print_fairq_opts(a, qopts);
+ break;
}
}
@@ -291,6 +315,9 @@ check_commit_altq(int dev, int opts)
case ALTQT_HFSC:
error = check_commit_hfsc(dev, opts, altq);
break;
+ case ALTQT_FAIRQ:
+ error = check_commit_fairq(dev, opts, altq);
+ break;
default:
break;
}
@@ -339,7 +366,8 @@ eval_pfqueue(struct pfctl *pf, struct pf_altq *pa, struct node_queue_bw *bw,
if (pa->qlimit == 0)
pa->qlimit = DEFAULT_QLIMIT;
- if (pa->scheduler == ALTQT_CBQ || pa->scheduler == ALTQT_HFSC) {
+ if (pa->scheduler == ALTQT_CBQ || pa->scheduler == ALTQT_HFSC ||
+ pa->scheduler == ALTQT_FAIRQ) {
pa->bandwidth = eval_bwspec(bw,
parent == NULL ? 0 : parent->bandwidth);
@@ -385,6 +413,9 @@ eval_pfqueue(struct pfctl *pf, struct pf_altq *pa, struct node_queue_bw *bw,
case ALTQT_HFSC:
error = eval_pfqueue_hfsc(pf, pa);
break;
+ case ALTQT_FAIRQ:
+ error = eval_pfqueue_fairq(pf, pa);
+ break;
default:
break;
}
@@ -565,6 +596,8 @@ print_cbq_opts(const struct pf_altq *a)
printf(" ecn");
if (opts->flags & CBQCLF_RIO)
printf(" rio");
+ if (opts->flags & CBQCLF_CODEL)
+ printf(" codel");
if (opts->flags & CBQCLF_CLEARDSCP)
printf(" cleardscp");
if (opts->flags & CBQCLF_FLOWVALVE)
@@ -652,6 +685,8 @@ print_priq_opts(const struct pf_altq *a)
printf(" ecn");
if (opts->flags & PRCF_RIO)
printf(" rio");
+ if (opts->flags & PRCF_CODEL)
+ printf(" codel");
if (opts->flags & PRCF_CLEARDSCP)
printf(" cleardscp");
if (opts->flags & PRCF_DEFAULTCLASS)
@@ -797,6 +832,85 @@ err_ret:
return (-1);
}
+/*
+ * FAIRQ support functions
+ */
+static int
+eval_pfqueue_fairq(struct pfctl *pf __unused, struct pf_altq *pa)
+{
+ struct pf_altq *altq, *parent;
+ struct fairq_opts *opts;
+ struct service_curve sc;
+
+ opts = &pa->pq_u.fairq_opts;
+
+ if (pa->parent[0] == 0) {
+ /* root queue */
+ opts->lssc_m1 = pa->ifbandwidth;
+ opts->lssc_m2 = pa->ifbandwidth;
+ opts->lssc_d = 0;
+ return (0);
+ }
+
+ LIST_INIT(&lssc);
+
+ /* if link_share is not specified, use bandwidth */
+ if (opts->lssc_m2 == 0)
+ opts->lssc_m2 = pa->bandwidth;
+
+ /*
+ * admission control:
+ * for the real-time service curve, the sum of the service curves
+ * should not exceed 80% of the interface bandwidth. 20% is reserved
+ * not to over-commit the actual interface bandwidth.
+ * for the link-sharing service curve, the sum of the child service
+ * curve should not exceed the parent service curve.
+ * for the upper-limit service curve, the assigned bandwidth should
+ * be smaller than the interface bandwidth, and the upper-limit should
+ * be larger than the real-time service curve when both are defined.
+ */
+ parent = qname_to_pfaltq(pa->parent, pa->ifname);
+ if (parent == NULL)
+ errx(1, "parent %s not found for %s", pa->parent, pa->qname);
+
+ TAILQ_FOREACH(altq, &altqs, entries) {
+ if (strncmp(altq->ifname, pa->ifname, IFNAMSIZ) != 0)
+ continue;
+ if (altq->qname[0] == 0) /* this is for interface */
+ continue;
+
+ if (strncmp(altq->parent, pa->parent, PF_QNAME_SIZE) != 0)
+ continue;
+
+ /* if the class has a link-sharing service curve, add it. */
+ if (opts->lssc_m2 != 0 && altq->pq_u.fairq_opts.lssc_m2 != 0) {
+ sc.m1 = altq->pq_u.fairq_opts.lssc_m1;
+ sc.d = altq->pq_u.fairq_opts.lssc_d;
+ sc.m2 = altq->pq_u.fairq_opts.lssc_m2;
+ gsc_add_sc(&lssc, &sc);
+ }
+ }
+
+ /* check the link-sharing service curve. */
+ if (opts->lssc_m2 != 0) {
+ sc.m1 = parent->pq_u.fairq_opts.lssc_m1;
+ sc.d = parent->pq_u.fairq_opts.lssc_d;
+ sc.m2 = parent->pq_u.fairq_opts.lssc_m2;
+ if (!is_gsc_under_sc(&lssc, &sc)) {
+ warnx("link-sharing sc exceeds parent's sc");
+ goto err_ret;
+ }
+ }
+
+ gsc_destroy(&lssc);
+
+ return (0);
+
+err_ret:
+ gsc_destroy(&lssc);
+ return (-1);
+}
+
static int
check_commit_hfsc(int dev, int opts, struct pf_altq *pa)
{
@@ -837,6 +951,43 @@ check_commit_hfsc(int dev, int opts, struct pf_altq *pa)
}
static int
+check_commit_fairq(int dev __unused, int opts __unused, struct pf_altq *pa)
+{
+ struct pf_altq *altq, *def = NULL;
+ int default_class;
+ int error = 0;
+
+ /* check if fairq has one default queue for this interface */
+ default_class = 0;
+ TAILQ_FOREACH(altq, &altqs, entries) {
+ if (strncmp(altq->ifname, pa->ifname, IFNAMSIZ) != 0)
+ continue;
+ if (altq->qname[0] == 0) /* this is for interface */
+ continue;
+ if (altq->pq_u.fairq_opts.flags & FARF_DEFAULTCLASS) {
+ default_class++;
+ def = altq;
+ }
+ }
+ if (default_class != 1) {
+ warnx("should have one default queue on %s", pa->ifname);
+ return (1);
+ }
+ /* make sure the default queue is a leaf */
+ TAILQ_FOREACH(altq, &altqs, entries) {
+ if (strncmp(altq->ifname, pa->ifname, IFNAMSIZ) != 0)
+ continue;
+ if (altq->qname[0] == 0) /* this is for interface */
+ continue;
+ if (strncmp(altq->parent, def->qname, PF_QNAME_SIZE) == 0) {
+ warnx("default queue is not a leaf");
+ error++;
+ }
+ }
+ return (error);
+}
+
+static int
print_hfsc_opts(const struct pf_altq *a, const struct node_queue_opt *qopts)
{
const struct hfsc_opts *opts;
@@ -861,6 +1012,8 @@ print_hfsc_opts(const struct pf_altq *a, const struct node_queue_opt *qopts)
printf(" ecn");
if (opts->flags & HFCF_RIO)
printf(" rio");
+ if (opts->flags & HFCF_CODEL)
+ printf(" codel");
if (opts->flags & HFCF_CLEARDSCP)
printf(" cleardscp");
if (opts->flags & HFCF_DEFAULTCLASS)
@@ -882,6 +1035,67 @@ print_hfsc_opts(const struct pf_altq *a, const struct node_queue_opt *qopts)
return (0);
}
+static int
+print_codel_opts(const struct pf_altq *a, const struct node_queue_opt *qopts)
+{
+ const struct codel_opts *opts;
+
+ opts = &a->pq_u.codel_opts;
+ if (opts->target || opts->interval || opts->ecn) {
+ printf("codel(");
+ if (opts->target)
+ printf(" target %d", opts->target);
+ if (opts->interval)
+ printf(" interval %d", opts->interval);
+ if (opts->ecn)
+ printf("ecn");
+ printf(" ) ");
+
+ return (1);
+ }
+
+ return (0);
+}
+
+static int
+print_fairq_opts(const struct pf_altq *a, const struct node_queue_opt *qopts)
+{
+ const struct fairq_opts *opts;
+ const struct node_fairq_sc *loc_lssc;
+
+ opts = &a->pq_u.fairq_opts;
+ if (qopts == NULL)
+ loc_lssc = NULL;
+ else
+ loc_lssc = &qopts->data.fairq_opts.linkshare;
+
+ if (opts->flags ||
+ (opts->lssc_m2 != 0 && (opts->lssc_m2 != a->bandwidth ||
+ opts->lssc_d != 0))) {
+ printf("fairq(");
+ if (opts->flags & FARF_RED)
+ printf(" red");
+ if (opts->flags & FARF_ECN)
+ printf(" ecn");
+ if (opts->flags & FARF_RIO)
+ printf(" rio");
+ if (opts->flags & FARF_CODEL)
+ printf(" codel");
+ if (opts->flags & FARF_CLEARDSCP)
+ printf(" cleardscp");
+ if (opts->flags & FARF_DEFAULTCLASS)
+ printf(" default");
+ if (opts->lssc_m2 != 0 && (opts->lssc_m2 != a->bandwidth ||
+ opts->lssc_d != 0))
+ print_fairq_sc("linkshare", opts->lssc_m1, opts->lssc_d,
+ opts->lssc_m2, loc_lssc);
+ printf(" ) ");
+
+ return (1);
+ } else
+ return (0);
+}
+
/*
* admission control using generalized service curve
*/
@@ -1201,6 +1415,28 @@ eval_queue_opts(struct pf_altq *pa, struct node_queue_opt *opts,
opts->data.hfsc_opts.upperlimit.d;
}
break;
+ case ALTQT_FAIRQ:
+ pa->pq_u.fairq_opts.flags = opts->data.fairq_opts.flags;
+ pa->pq_u.fairq_opts.nbuckets = opts->data.fairq_opts.nbuckets;
+ pa->pq_u.fairq_opts.hogs_m1 =
+ eval_bwspec(&opts->data.fairq_opts.hogs_bw, ref_bw);
+
+ if (opts->data.fairq_opts.linkshare.used) {
+ pa->pq_u.fairq_opts.lssc_m1 =
+ eval_bwspec(&opts->data.fairq_opts.linkshare.m1,
+ ref_bw);
+ pa->pq_u.fairq_opts.lssc_m2 =
+ eval_bwspec(&opts->data.fairq_opts.linkshare.m2,
+ ref_bw);
+ pa->pq_u.fairq_opts.lssc_d =
+ opts->data.fairq_opts.linkshare.d;
+ }
+ break;
+ case ALTQT_CODEL:
+ pa->pq_u.codel_opts.target = opts->data.codel_opts.target;
+ pa->pq_u.codel_opts.interval = opts->data.codel_opts.interval;
+ pa->pq_u.codel_opts.ecn = opts->data.codel_opts.ecn;
+ break;
default:
warnx("eval_queue_opts: unknown scheduler type %u",
opts->qtype);
@@ -1246,3 +1482,27 @@ print_hfsc_sc(const char *scname, u_int m1, u_int d, u_int m2,
if (d != 0)
printf(")");
}
+
+void
+print_fairq_sc(const char *scname, u_int m1, u_int d, u_int m2,
+ const struct node_fairq_sc *sc)
+{
+ printf(" %s", scname);
+
+ if (d != 0) {
+ printf("(");
+ if (sc != NULL && sc->m1.bw_percent > 0)
+ printf("%u%%", sc->m1.bw_percent);
+ else
+ printf("%s", rate2str((double)m1));
+ printf(" %u", d);
+ }
+
+ if (sc != NULL && sc->m2.bw_percent > 0)
+ printf(" %u%%", sc->m2.bw_percent);
+ else
+ printf(" %s", rate2str((double)m2));
+
+ if (d != 0)
+ printf(")");
+}
diff --git a/sbin/pfctl/pfctl_parser.h b/sbin/pfctl/pfctl_parser.h
index 5c909e9..2b7fea7 100644
--- a/sbin/pfctl/pfctl_parser.h
+++ b/sbin/pfctl/pfctl_parser.h
@@ -150,12 +150,28 @@ struct node_hfsc_opts {
int flags;
};
+struct node_fairq_sc {
+ struct node_queue_bw m1; /* slope of 1st segment; bps */
+ u_int d; /* x-projection of m1; msec */
+ struct node_queue_bw m2; /* slope of 2nd segment; bps */
+ u_int8_t used;
+};
+
+struct node_fairq_opts {
+ struct node_fairq_sc linkshare;
+ struct node_queue_bw hogs_bw;
+ u_int nbuckets;
+ int flags;
+};
+
struct node_queue_opt {
int qtype;
union {
struct cbq_opts cbq_opts;
+ struct codel_opts codel_opts;
struct priq_opts priq_opts;
struct node_hfsc_opts hfsc_opts;
+ struct node_fairq_opts fairq_opts;
} data;
};
diff --git a/sbin/pfctl/pfctl_qstats.c b/sbin/pfctl/pfctl_qstats.c
index 95371e4..8ffd87d 100644
--- a/sbin/pfctl/pfctl_qstats.c
+++ b/sbin/pfctl/pfctl_qstats.c
@@ -36,8 +36,10 @@ __FBSDID("$FreeBSD$");
#include <altq/altq.h>
#include <altq/altq_cbq.h>
+#include <altq/altq_codel.h>
#include <altq/altq_priq.h>
#include <altq/altq_hfsc.h>
+#include <altq/altq_fairq.h>
#include "pfctl.h"
#include "pfctl_parser.h"
@@ -46,6 +48,8 @@ union class_stats {
class_stats_t cbq_stats;
struct priq_classstats priq_stats;
struct hfsc_classstats hfsc_stats;
+ struct fairq_classstats fairq_stats;
+ struct codel_ifstats codel_stats;
};
#define AVGN_MAX 8
@@ -75,8 +79,10 @@ struct pf_altq_node *pfctl_find_altq_node(struct pf_altq_node *,
void pfctl_print_altq_node(int, const struct pf_altq_node *,
unsigned, int);
void print_cbqstats(struct queue_stats);
+void print_codelstats(struct queue_stats);
void print_priqstats(struct queue_stats);
void print_hfscstats(struct queue_stats);
+void print_fairqstats(struct queue_stats);
void pfctl_free_altq_node(struct pf_altq_node *);
void pfctl_print_altq_nodestat(int,
const struct pf_altq_node *);
@@ -162,7 +168,7 @@ pfctl_update_qstats(int dev, struct pf_altq_node **root)
return (-1);
}
#ifdef __FreeBSD__
- if (pa.altq.qid > 0 &&
+ if ((pa.altq.qid > 0 || pa.altq.scheduler == ALTQT_CODEL) &&
!(pa.altq.local_flags & PFALTQ_FLAG_IF_REMOVED)) {
#else
if (pa.altq.qid > 0) {
@@ -300,7 +306,7 @@ pfctl_print_altq_node(int dev, const struct pf_altq_node *node,
void
pfctl_print_altq_nodestat(int dev, const struct pf_altq_node *a)
{
- if (a->altq.qid == 0)
+ if (a->altq.qid == 0 && a->altq.scheduler != ALTQT_CODEL)
return;
#ifdef __FreeBSD__
@@ -317,6 +323,12 @@ pfctl_print_altq_nodestat(int dev, const struct pf_altq_node *a)
case ALTQT_HFSC:
print_hfscstats(a->qstats);
break;
+ case ALTQT_FAIRQ:
+ print_fairqstats(a->qstats);
+ break;
+ case ALTQT_CODEL:
+ print_codelstats(a->qstats);
+ break;
}
}
@@ -342,6 +354,28 @@ print_cbqstats(struct queue_stats cur)
}
void
+print_codelstats(struct queue_stats cur)
+{
+ printf(" [ pkts: %10llu bytes: %10llu "
+ "dropped pkts: %6llu bytes: %6llu ]\n",
+ (unsigned long long)cur.data.codel_stats.cl_xmitcnt.packets,
+ (unsigned long long)cur.data.codel_stats.cl_xmitcnt.bytes,
+ (unsigned long long)cur.data.codel_stats.cl_dropcnt.packets +
+ cur.data.codel_stats.stats.drop_cnt.packets,
+ (unsigned long long)cur.data.codel_stats.cl_dropcnt.bytes +
+ cur.data.codel_stats.stats.drop_cnt.bytes);
+ printf(" [ qlength: %3d/%3d ]\n",
+ cur.data.codel_stats.qlength, cur.data.codel_stats.qlimit);
+
+ if (cur.avgn < 2)
+ return;
+
+ printf(" [ measured: %7.1f packets/s, %s/s ]\n",
+ cur.avg_packets / STAT_INTERVAL,
+ rate2str((8 * cur.avg_bytes) / STAT_INTERVAL));
+}
+
+void
print_priqstats(struct queue_stats cur)
{
printf(" [ pkts: %10llu bytes: %10llu "
@@ -382,6 +416,26 @@ print_hfscstats(struct queue_stats cur)
}
void
+print_fairqstats(struct queue_stats cur)
+{
+ printf(" [ pkts: %10llu bytes: %10llu "
+ "dropped pkts: %6llu bytes: %6llu ]\n",
+ (unsigned long long)cur.data.fairq_stats.xmit_cnt.packets,
+ (unsigned long long)cur.data.fairq_stats.xmit_cnt.bytes,
+ (unsigned long long)cur.data.fairq_stats.drop_cnt.packets,
+ (unsigned long long)cur.data.fairq_stats.drop_cnt.bytes);
+ printf(" [ qlength: %3d/%3d ]\n",
+ cur.data.fairq_stats.qlength, cur.data.fairq_stats.qlimit);
+
+ if (cur.avgn < 2)
+ return;
+
+ printf(" [ measured: %7.1f packets/s, %s/s ]\n",
+ cur.avg_packets / STAT_INTERVAL,
+ rate2str((8 * cur.avg_bytes) / STAT_INTERVAL));
+}
+
+void
pfctl_free_altq_node(struct pf_altq_node *node)
{
while (node != NULL) {
@@ -402,7 +456,7 @@ update_avg(struct pf_altq_node *a)
u_int64_t b, p;
int n;
- if (a->altq.qid == 0)
+ if (a->altq.qid == 0 && a->altq.scheduler != ALTQT_CODEL)
return;
qs = &a->qstats;
@@ -421,6 +475,14 @@ update_avg(struct pf_altq_node *a)
b = qs->data.hfsc_stats.xmit_cnt.bytes;
p = qs->data.hfsc_stats.xmit_cnt.packets;
break;
+ case ALTQT_FAIRQ:
+ b = qs->data.fairq_stats.xmit_cnt.bytes;
+ p = qs->data.fairq_stats.xmit_cnt.packets;
+ break;
+ case ALTQT_CODEL:
+ b = qs->data.codel_stats.cl_xmitcnt.bytes;
+ p = qs->data.codel_stats.cl_xmitcnt.packets;
+ break;
default:
b = 0;
p = 0;
diff --git a/share/man/man4/altq.4 b/share/man/man4/altq.4
index 1effdb2..771dc53 100644
--- a/share/man/man4/altq.4
+++ b/share/man/man4/altq.4
@@ -25,7 +25,7 @@
.\"
.\" $FreeBSD$
.\"
-.Dd December 9, 2011
+.Dd July 24, 2015
.Dt ALTQ 4
.Os
.Sh NAME
@@ -35,11 +35,13 @@
.Cd options ALTQ
.Pp
.Cd options ALTQ_CBQ
+.Cd options ALTQ_CODEL
.Cd options ALTQ_RED
.Cd options ALTQ_RIO
.Cd options ALTQ_HFSC
.Cd options ALTQ_CDNR
.Cd options ALTQ_PRIQ
+.Cd options ALTQ_FAIRQ
.Sh DESCRIPTION
The
.Nm
@@ -73,6 +75,10 @@ Enable
Build the
.Dq "Class Based Queuing"
discipline.
+.It Dv ALTQ_CODEL
+Build the
+.Dq "Controlled Delay"
+discipline.
.It Dv ALTQ_RED
Build the
.Dq "Random Early Detection"
@@ -93,6 +99,10 @@ any of the available disciplines or consumers.
Build the
.Dq "Priority Queuing"
discipline.
+.It Dv ALTQ_FAIRQ
+Build the
+.Dq "Fair Queuing"
+discipline.
.It Dv ALTQ_NOPCC
Required if the TSC is unusable.
.It Dv ALTQ_DEBUG
diff --git a/sys/cam/ata/ata_all.c b/sys/cam/ata/ata_all.c
index d5220e8..f0d6202 100644
--- a/sys/cam/ata/ata_all.c
+++ b/sys/cam/ata/ata_all.c
@@ -211,15 +211,31 @@ ata_op_string(struct ata_cmd *cmd)
char *
ata_cmd_string(struct ata_cmd *cmd, char *cmd_string, size_t len)
{
+ struct sbuf sb;
+ int error;
- snprintf(cmd_string, len, "%02x %02x %02x %02x "
+ if (len == 0)
+ return ("");
+
+ sbuf_new(&sb, cmd_string, len, SBUF_FIXEDLEN);
+ ata_cmd_sbuf(cmd, &sb);
+
+ error = sbuf_finish(&sb);
+ if (error != 0 && error != ENOMEM)
+ return ("");
+
+ return(sbuf_data(&sb));
+}
+
+void
+ata_cmd_sbuf(struct ata_cmd *cmd, struct sbuf *sb)
+{
+ sbuf_printf(sb, "%02x %02x %02x %02x "
"%02x %02x %02x %02x %02x %02x %02x %02x",
cmd->command, cmd->features,
cmd->lba_low, cmd->lba_mid, cmd->lba_high, cmd->device,
cmd->lba_low_exp, cmd->lba_mid_exp, cmd->lba_high_exp,
cmd->features_exp, cmd->sector_count, cmd->sector_count_exp);
-
- return(cmd_string);
}
char *
@@ -233,7 +249,7 @@ ata_res_string(struct ata_res *res, char *res_string, size_t len)
res->lba_low_exp, res->lba_mid_exp, res->lba_high_exp,
res->sector_count, res->sector_count_exp);
- return(res_string);
+ return (res_string);
}
/*
@@ -242,11 +258,10 @@ ata_res_string(struct ata_res *res, char *res_string, size_t len)
int
ata_command_sbuf(struct ccb_ataio *ataio, struct sbuf *sb)
{
- char cmd_str[(12 * 3) + 1];
- sbuf_printf(sb, "%s. ACB: %s",
- ata_op_string(&ataio->cmd),
- ata_cmd_string(&ataio->cmd, cmd_str, sizeof(cmd_str)));
+ sbuf_printf(sb, "%s. ACB: ",
+ ata_op_string(&ataio->cmd));
+ ata_cmd_sbuf(&ataio->cmd, sb);
return(0);
}
diff --git a/sys/cam/ata/ata_all.h b/sys/cam/ata/ata_all.h
index 91e941c..849640f 100644
--- a/sys/cam/ata/ata_all.h
+++ b/sys/cam/ata/ata_all.h
@@ -103,6 +103,7 @@ int ata_version(int ver);
char * ata_op_string(struct ata_cmd *cmd);
char * ata_cmd_string(struct ata_cmd *cmd, char *cmd_string, size_t len);
+void ata_cmd_sbuf(struct ata_cmd *cmd, struct sbuf *sb);
char * ata_res_string(struct ata_res *res, char *res_string, size_t len);
int ata_command_sbuf(struct ccb_ataio *ataio, struct sbuf *sb);
int ata_status_sbuf(struct ccb_ataio *ataio, struct sbuf *sb);
diff --git a/sys/cam/cam_periph.c b/sys/cam/cam_periph.c
index 625be21..d8940a0 100644
--- a/sys/cam/cam_periph.c
+++ b/sys/cam/cam_periph.c
@@ -86,6 +86,7 @@ static int camperiphscsisenseerror(union ccb *ccb,
u_int32_t *timeout,
u_int32_t *action,
const char **action_string);
+static void cam_periph_devctl_notify(union ccb *ccb);
static int nperiph_drivers;
static int initialized = 0;
@@ -1615,7 +1616,7 @@ cam_periph_error(union ccb *ccb, cam_flags camflags,
struct cam_periph *periph;
const char *action_string;
cam_status status;
- int frozen, error, openings;
+ int frozen, error, openings, devctl_err;
u_int32_t action, relsim_flags, timeout;
action = SSQ_PRINT_SENSE;
@@ -1624,9 +1625,26 @@ cam_periph_error(union ccb *ccb, cam_flags camflags,
status = ccb->ccb_h.status;
frozen = (status & CAM_DEV_QFRZN) != 0;
status &= CAM_STATUS_MASK;
- openings = relsim_flags = timeout = 0;
+ devctl_err = openings = relsim_flags = timeout = 0;
orig_ccb = ccb;
+ /* Filter the errors that should be reported via devctl */
+ switch (ccb->ccb_h.status & CAM_STATUS_MASK) {
+ case CAM_CMD_TIMEOUT:
+ case CAM_REQ_ABORTED:
+ case CAM_REQ_CMP_ERR:
+ case CAM_REQ_TERMIO:
+ case CAM_UNREC_HBA_ERROR:
+ case CAM_DATA_RUN_ERR:
+ case CAM_SCSI_STATUS_ERROR:
+ case CAM_ATA_STATUS_ERROR:
+ case CAM_SMP_STATUS_ERROR:
+ devctl_err++;
+ break;
+ default:
+ break;
+ }
+
switch (status) {
case CAM_REQ_CMP:
error = 0;
@@ -1754,6 +1772,9 @@ cam_periph_error(union ccb *ccb, cam_flags camflags,
xpt_print(ccb->ccb_h.path, "Retrying command\n");
}
+ if (devctl_err)
+ cam_periph_devctl_notify(orig_ccb);
+
if ((action & SSQ_LOST) != 0) {
lun_id_t lun_id;
@@ -1824,3 +1845,83 @@ cam_periph_error(union ccb *ccb, cam_flags camflags,
return (error);
}
+
+#define CAM_PERIPH_DEVD_MSG_SIZE 256
+
+static void
+cam_periph_devctl_notify(union ccb *ccb)
+{
+ struct cam_periph *periph;
+ struct ccb_getdev *cgd;
+ struct sbuf sb;
+ int serr, sk, asc, ascq;
+ char *sbmsg, *type;
+
+ sbmsg = malloc(CAM_PERIPH_DEVD_MSG_SIZE, M_CAMPERIPH, M_NOWAIT);
+ if (sbmsg == NULL)
+ return;
+
+ sbuf_new(&sb, sbmsg, CAM_PERIPH_DEVD_MSG_SIZE, SBUF_FIXEDLEN);
+
+ periph = xpt_path_periph(ccb->ccb_h.path);
+ sbuf_printf(&sb, "device=%s%d ", periph->periph_name,
+ periph->unit_number);
+
+ sbuf_printf(&sb, "serial=\"");
+ if ((cgd = (struct ccb_getdev *)xpt_alloc_ccb_nowait()) != NULL) {
+ xpt_setup_ccb(&cgd->ccb_h, ccb->ccb_h.path,
+ CAM_PRIORITY_NORMAL);
+ cgd->ccb_h.func_code = XPT_GDEV_TYPE;
+ xpt_action((union ccb *)cgd);
+
+ if (cgd->ccb_h.status == CAM_REQ_CMP)
+ sbuf_bcat(&sb, cgd->serial_num, cgd->serial_num_len);
+ }
+ sbuf_printf(&sb, "\" ");
+ sbuf_printf(&sb, "cam_status=\"0x%x\" ", ccb->ccb_h.status);
+
+ switch (ccb->ccb_h.status & CAM_STATUS_MASK) {
+ case CAM_CMD_TIMEOUT:
+ sbuf_printf(&sb, "timeout=%d ", ccb->ccb_h.timeout);
+ type = "timeout";
+ break;
+ case CAM_SCSI_STATUS_ERROR:
+ sbuf_printf(&sb, "scsi_status=%d ", ccb->csio.scsi_status);
+ if (scsi_extract_sense_ccb(ccb, &serr, &sk, &asc, &ascq))
+ sbuf_printf(&sb, "scsi_sense=\"%02x %02x %02x %02x\" ",
+ serr, sk, asc, ascq);
+ type = "error";
+ break;
+ case CAM_ATA_STATUS_ERROR:
+ {
+ char res_str[(11 * 3) + 1];
+
+ sbuf_printf(&sb, "RES=\"%s\" ", ata_res_string(&ccb->ataio.res,
+ res_str, sizeof(res_str)));
+ type = "error";
+ break;
+ }
+ default:
+ type = "error";
+ break;
+ }
+
+ if (ccb->ccb_h.func_code == XPT_SCSI_IO) {
+ sbuf_printf(&sb, "CDB=\"");
+ if ((ccb->ccb_h.flags & CAM_CDB_POINTER) != 0)
+ scsi_cdb_sbuf(ccb->csio.cdb_io.cdb_ptr, &sb);
+ else
+ scsi_cdb_sbuf(ccb->csio.cdb_io.cdb_bytes, &sb);
+ sbuf_printf(&sb, "\" ");
+ } else if (ccb->ccb_h.func_code == XPT_ATA_IO) {
+ sbuf_printf(&sb, "ACB=\"");
+ ata_cmd_sbuf(&ccb->ataio.cmd, &sb);
+ sbuf_printf(&sb, "\" ");
+ }
+
+ if (sbuf_finish(&sb) == 0)
+ devctl_notify("CAM", "periph", type, sbuf_data(&sb));
+ sbuf_delete(&sb);
+ free(sbmsg, M_CAMPERIPH);
+}
+
diff --git a/sys/cam/scsi/scsi_all.c b/sys/cam/scsi/scsi_all.c
index 023103b..3a073ee 100644
--- a/sys/cam/scsi/scsi_all.c
+++ b/sys/cam/scsi/scsi_all.c
@@ -3464,14 +3464,32 @@ scsi_error_action(struct ccb_scsiio *csio, struct scsi_inquiry_data *inq_data,
char *
scsi_cdb_string(u_int8_t *cdb_ptr, char *cdb_string, size_t len)
{
+ struct sbuf sb;
+ int error;
+
+ if (len == 0)
+ return ("");
+
+ sbuf_new(&sb, cdb_string, len, SBUF_FIXEDLEN);
+
+ scsi_cdb_sbuf(cdb_ptr, &sb);
+
+ /* ENOMEM just means that the fixed buffer is full, OK to ignore */
+ error = sbuf_finish(&sb);
+ if (error != 0 && error != ENOMEM)
+ return ("");
+
+ return(sbuf_data(&sb));
+}
+
+void
+scsi_cdb_sbuf(u_int8_t *cdb_ptr, struct sbuf *sb)
+{
u_int8_t cdb_len;
int i;
if (cdb_ptr == NULL)
- return("");
-
- /* Silence warnings */
- cdb_len = 0;
+ return;
/*
* This is taken from the SCSI-3 draft spec.
@@ -3508,12 +3526,11 @@ scsi_cdb_string(u_int8_t *cdb_ptr, char *cdb_string, size_t len)
cdb_len = 12;
break;
}
- *cdb_string = '\0';
+
for (i = 0; i < cdb_len; i++)
- snprintf(cdb_string + strlen(cdb_string),
- len - strlen(cdb_string), "%02hhx ", cdb_ptr[i]);
+ sbuf_printf(sb, "%02hhx ", cdb_ptr[i]);
- return(cdb_string);
+ return;
}
const char *
@@ -3562,7 +3579,6 @@ scsi_command_string(struct cam_device *device, struct ccb_scsiio *csio,
#endif /* _KERNEL/!_KERNEL */
{
struct scsi_inquiry_data *inq_data;
- char cdb_str[(SCSI_MAX_CDBLEN * 3) + 1];
#ifdef _KERNEL
struct ccb_getdev *cgd;
#endif /* _KERNEL */
@@ -3595,15 +3611,13 @@ scsi_command_string(struct cam_device *device, struct ccb_scsiio *csio,
#endif /* _KERNEL/!_KERNEL */
if ((csio->ccb_h.flags & CAM_CDB_POINTER) != 0) {
- sbuf_printf(sb, "%s. CDB: %s",
- scsi_op_desc(csio->cdb_io.cdb_ptr[0], inq_data),
- scsi_cdb_string(csio->cdb_io.cdb_ptr, cdb_str,
- sizeof(cdb_str)));
+ sbuf_printf(sb, "%s. CDB: ",
+ scsi_op_desc(csio->cdb_io.cdb_ptr[0], inq_data));
+ scsi_cdb_sbuf(csio->cdb_io.cdb_ptr, sb);
} else {
- sbuf_printf(sb, "%s. CDB: %s",
- scsi_op_desc(csio->cdb_io.cdb_bytes[0], inq_data),
- scsi_cdb_string(csio->cdb_io.cdb_bytes, cdb_str,
- sizeof(cdb_str)));
+ sbuf_printf(sb, "%s. CDB: ",
+ scsi_op_desc(csio->cdb_io.cdb_bytes[0], inq_data));
+ scsi_cdb_sbuf(csio->cdb_io.cdb_bytes, sb);
}
#ifdef _KERNEL
diff --git a/sys/cam/scsi/scsi_all.h b/sys/cam/scsi/scsi_all.h
index f2b4b21..615a42a 100644
--- a/sys/cam/scsi/scsi_all.h
+++ b/sys/cam/scsi/scsi_all.h
@@ -3644,6 +3644,7 @@ const char * scsi_op_desc(u_int16_t opcode,
struct scsi_inquiry_data *inq_data);
char * scsi_cdb_string(u_int8_t *cdb_ptr, char *cdb_string,
size_t len);
+void scsi_cdb_sbuf(u_int8_t *cdb_ptr, struct sbuf *sb);
void scsi_print_inquiry(struct scsi_inquiry_data *inq_data);
void scsi_print_inquiry_short(struct scsi_inquiry_data *inq_data);
diff --git a/sys/conf/NOTES b/sys/conf/NOTES
index a977898..7b5c725 100644
--- a/sys/conf/NOTES
+++ b/sys/conf/NOTES
@@ -699,7 +699,9 @@ options ALTQ
options ALTQ_CBQ # Class Based Queueing
options ALTQ_RED # Random Early Detection
options ALTQ_RIO # RED In/Out
+options ALTQ_CODEL # CoDel Active Queueing
options ALTQ_HFSC # Hierarchical Packet Scheduler
+options ALTQ_FAIRQ # Fair Packet Scheduler
options ALTQ_CDNR # Traffic conditioner
options ALTQ_PRIQ # Priority Queueing
options ALTQ_NOPCC # Required if the TSC is unusable
diff --git a/sys/conf/files b/sys/conf/files
index 23e0353..945bfe9 100644
--- a/sys/conf/files
+++ b/sys/conf/files
@@ -252,6 +252,8 @@ compat/freebsd32/freebsd32_syscalls.c optional compat_freebsd32
compat/freebsd32/freebsd32_sysent.c optional compat_freebsd32
contrib/altq/altq/altq_cbq.c optional altq
contrib/altq/altq/altq_cdnr.c optional altq
+contrib/altq/altq/altq_codel.c optional altq
+contrib/altq/altq/altq_fairq.c optional altq
contrib/altq/altq/altq_hfsc.c optional altq
contrib/altq/altq/altq_priq.c optional altq
contrib/altq/altq/altq_red.c optional altq
diff --git a/sys/conf/options b/sys/conf/options
index 0699c2c..4f1ab3a 100644
--- a/sys/conf/options
+++ b/sys/conf/options
@@ -384,7 +384,9 @@ ACCEPT_FILTER_HTTP
ALTQ opt_global.h
ALTQ_CBQ opt_altq.h
ALTQ_CDNR opt_altq.h
+ALTQ_CODEL opt_altq.h
ALTQ_DEBUG opt_altq.h
+ALTQ_FAIRQ opt_altq.h
ALTQ_HFSC opt_altq.h
ALTQ_NOPCC opt_altq.h
ALTQ_PRIQ opt_altq.h
diff --git a/sys/contrib/altq/altq/altq.h b/sys/contrib/altq/altq/altq.h
index c740ed3..4ef6cd2 100644
--- a/sys/contrib/altq/altq/altq.h
+++ b/sys/contrib/altq/altq/altq.h
@@ -63,7 +63,9 @@
#define ALTQT_BLUE 10 /* blue */
#define ALTQT_PRIQ 11 /* priority queue */
#define ALTQT_JOBS 12 /* JoBS */
-#define ALTQT_MAX 13 /* should be max discipline type + 1 */
+#define ALTQT_FAIRQ 13 /* fairq */
+#define ALTQT_CODEL 14 /* CoDel */
+#define ALTQT_MAX 15 /* should be max discipline type + 1 */
#ifdef ALTQ3_COMPAT
struct altqreq {
diff --git a/sys/contrib/altq/altq/altq_cbq.c b/sys/contrib/altq/altq/altq_cbq.c
index 3991d1d..13f9721 100644
--- a/sys/contrib/altq/altq/altq_cbq.c
+++ b/sys/contrib/altq/altq/altq_cbq.c
@@ -241,6 +241,10 @@ get_class_stats(class_stats_t *statsp, struct rm_class *cl)
if (q_is_rio(cl->q_))
rio_getstats((rio_t *)cl->red_, &statsp->red[0]);
#endif
+#ifdef ALTQ_CODEL
+ if (q_is_codel(cl->q_))
+ codel_getstats(cl->codel_, &statsp->codel);
+#endif
}
int
diff --git a/sys/contrib/altq/altq/altq_cbq.h b/sys/contrib/altq/altq/altq_cbq.h
index 76096af..f0d9172 100644
--- a/sys/contrib/altq/altq/altq_cbq.h
+++ b/sys/contrib/altq/altq/altq_cbq.h
@@ -34,6 +34,7 @@
#define _ALTQ_ALTQ_CBQ_H_
#include <altq/altq.h>
+#include <altq/altq_codel.h>
#include <altq/altq_rmclass.h>
#include <altq/altq_red.h>
#include <altq/altq_rio.h>
@@ -51,6 +52,7 @@ extern "C" {
#define CBQCLF_FLOWVALVE 0x0008 /* use flowvalve (aka penalty-box) */
#define CBQCLF_CLEARDSCP 0x0010 /* clear diffserv codepoint */
#define CBQCLF_BORROW 0x0020 /* borrow from parent */
+#define CBQCLF_CODEL 0x0040 /* use CoDel */
/* class flags only for root class */
#define CBQCLF_WRR 0x0100 /* weighted-round robin */
@@ -90,9 +92,10 @@ typedef struct _cbq_class_stats_ {
int qcnt; /* # packets in queue */
int avgidle;
- /* red and rio related info */
+ /* codel, red and rio related info */
int qtype;
struct redstats red[3];
+ struct codel_stats codel;
} class_stats_t;
#ifdef ALTQ3_COMPAT
diff --git a/sys/contrib/altq/altq/altq_classq.h b/sys/contrib/altq/altq/altq_classq.h
index dc5c646..daec994 100644
--- a/sys/contrib/altq/altq/altq_classq.h
+++ b/sys/contrib/altq/altq/altq_classq.h
@@ -49,6 +49,7 @@ extern "C" {
#define Q_RED 0x01
#define Q_RIO 0x02
#define Q_DROPTAIL 0x03
+#define Q_CODEL 0x04
#ifdef _KERNEL
@@ -59,6 +60,7 @@ struct _class_queue_ {
struct mbuf *tail_; /* Tail of packet queue */
int qlen_; /* Queue length (in number of packets) */
int qlim_; /* Queue limit (in number of packets*) */
+ int qsize_; /* Queue size (in number of bytes*) */
int qtype_; /* Queue type */
};
@@ -67,10 +69,12 @@ typedef struct _class_queue_ class_queue_t;
#define qtype(q) (q)->qtype_ /* Get queue type */
#define qlimit(q) (q)->qlim_ /* Max packets to be queued */
#define qlen(q) (q)->qlen_ /* Current queue length. */
+#define qsize(q) (q)->qsize_ /* Current queue size. */
#define qtail(q) (q)->tail_ /* Tail of the queue */
#define qhead(q) ((q)->tail_ ? (q)->tail_->m_nextpkt : NULL)
#define qempty(q) ((q)->qlen_ == 0) /* Is the queue empty?? */
+#define q_is_codel(q) ((q)->qtype_ == Q_CODEL) /* Is the queue a codel queue */
#define q_is_red(q) ((q)->qtype_ == Q_RED) /* Is the queue a red queue */
#define q_is_rio(q) ((q)->qtype_ == Q_RIO) /* Is the queue a rio queue */
#define q_is_red_or_rio(q) ((q)->qtype_ == Q_RED || (q)->qtype_ == Q_RIO)
@@ -100,6 +104,7 @@ _addq(class_queue_t *q, struct mbuf *m)
m0->m_nextpkt = m;
qtail(q) = m;
qlen(q)++;
+ qsize(q) += m_pktlen(m);
}
static __inline struct mbuf *
@@ -114,6 +119,7 @@ _getq(class_queue_t *q)
else
qtail(q) = NULL;
qlen(q)--;
+ qsize(q) -= m_pktlen(m0);
m0->m_nextpkt = NULL;
return (m0);
}
diff --git a/sys/contrib/altq/altq/altq_codel.c b/sys/contrib/altq/altq/altq_codel.c
new file mode 100644
index 0000000..45c5ca6
--- /dev/null
+++ b/sys/contrib/altq/altq/altq_codel.c
@@ -0,0 +1,477 @@
+/*
+ * CoDel - The Controlled-Delay Active Queue Management algorithm
+ *
+ * Copyright (C) 2013 Ermal Luçi <eri@FreeBSD.org>
+ * Copyright (C) 2011-2012 Kathleen Nichols <nichols@pollere.com>
+ * Copyright (C) 2011-2012 Van Jacobson <van@pollere.net>
+ * Copyright (C) 2012 Michael D. Taht <dave.taht@bufferbloat.net>
+ * Copyright (C) 2012 Eric Dumazet <edumazet@google.com>
+ *
+ * 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,
+ * without modification.
+ * 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. The names of the authors may not be used to endorse or promote products
+ * derived from this software without specific prior written permission.
+ *
+ * Alternatively, provided that this notice is retained in full, this
+ * software may be distributed under the terms of the GNU General
+ * Public License ("GPL") version 2, in which case the provisions of the
+ * GPL apply INSTEAD OF those given above.
+ *
+ * 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_altq.h"
+#include "opt_inet.h"
+#include "opt_inet6.h"
+
+#ifdef ALTQ_CODEL /* CoDel is enabled by ALTQ_CODEL option in opt_altq.h */
+
+#include <sys/param.h>
+#include <sys/malloc.h>
+#include <sys/mbuf.h>
+#include <sys/socket.h>
+#include <sys/systm.h>
+
+#include <net/if.h>
+#include <net/if_var.h>
+#include <netinet/in.h>
+
+#include <netpfil/pf/pf.h>
+#include <netpfil/pf/pf_altq.h>
+#include <altq/if_altq.h>
+#include <altq/altq.h>
+#include <altq/altq_codel.h>
+
+static int codel_should_drop(struct codel *, class_queue_t *,
+ struct mbuf *, u_int64_t);
+static void codel_Newton_step(struct codel_vars *);
+static u_int64_t codel_control_law(u_int64_t t, u_int64_t, u_int32_t);
+
+#define codel_time_after(a, b) ((int64_t)(a) - (int64_t)(b) > 0)
+#define codel_time_after_eq(a, b) ((int64_t)(a) - (int64_t)(b) >= 0)
+#define codel_time_before(a, b) ((int64_t)(a) - (int64_t)(b) < 0)
+#define codel_time_before_eq(a, b) ((int64_t)(a) - (int64_t)(b) <= 0)
+
+static int codel_request(struct ifaltq *, int, void *);
+
+static int codel_enqueue(struct ifaltq *, struct mbuf *, struct altq_pktattr *);
+static struct mbuf *codel_dequeue(struct ifaltq *, int);
+
+int
+codel_pfattach(struct pf_altq *a)
+{
+ struct ifnet *ifp;
+
+ if ((ifp = ifunit(a->ifname)) == NULL || a->altq_disc == NULL)
+ return (EINVAL);
+
+ return (altq_attach(&ifp->if_snd, ALTQT_CODEL, a->altq_disc,
+ codel_enqueue, codel_dequeue, codel_request, NULL, NULL));
+}
+
+int
+codel_add_altq(struct pf_altq *a)
+{
+ struct codel_if *cif;
+ struct ifnet *ifp;
+ struct codel_opts *opts;
+
+ if ((ifp = ifunit(a->ifname)) == NULL)
+ return (EINVAL);
+ if (!ALTQ_IS_READY(&ifp->if_snd))
+ return (ENODEV);
+
+ opts = &a->pq_u.codel_opts;
+
+ cif = malloc(sizeof(struct codel_if), M_DEVBUF, M_NOWAIT | M_ZERO);
+ if (cif == NULL)
+ return (ENOMEM);
+ cif->cif_bandwidth = a->ifbandwidth;
+ cif->cif_ifq = &ifp->if_snd;
+
+ cif->cl_q = malloc(sizeof(class_queue_t), M_DEVBUF, M_NOWAIT | M_ZERO);
+ if (cif->cl_q == NULL) {
+ free(cif, M_DEVBUF);
+ return (ENOMEM);
+ }
+
+ if (a->qlimit == 0)
+ a->qlimit = 50; /* use default. */
+ qlimit(cif->cl_q) = a->qlimit;
+ qtype(cif->cl_q) = Q_CODEL;
+ qlen(cif->cl_q) = 0;
+ qsize(cif->cl_q) = 0;
+
+ if (opts->target == 0)
+ opts->target = 5;
+ if (opts->interval == 0)
+ opts->interval = 100;
+ cif->codel.params.target = machclk_freq * opts->target / 1000;
+ cif->codel.params.interval = machclk_freq * opts->interval / 1000;
+ cif->codel.params.ecn = opts->ecn;
+ cif->codel.stats.maxpacket = 256;
+
+ cif->cl_stats.qlength = qlen(cif->cl_q);
+ cif->cl_stats.qlimit = qlimit(cif->cl_q);
+
+ /* keep the state in pf_altq */
+ a->altq_disc = cif;
+
+ return (0);
+}
+
+int
+codel_remove_altq(struct pf_altq *a)
+{
+ struct codel_if *cif;
+
+ if ((cif = a->altq_disc) == NULL)
+ return (EINVAL);
+ a->altq_disc = NULL;
+
+ if (cif->cl_q)
+ free(cif->cl_q, M_DEVBUF);
+ free(cif, M_DEVBUF);
+
+ return (0);
+}
+
+int
+codel_getqstats(struct pf_altq *a, void *ubuf, int *nbytes)
+{
+ struct codel_if *cif;
+ struct codel_ifstats stats;
+ int error = 0;
+
+ if ((cif = altq_lookup(a->ifname, ALTQT_CODEL)) == NULL)
+ return (EBADF);
+
+ if (*nbytes < sizeof(stats))
+ return (EINVAL);
+
+ stats = cif->cl_stats;
+ stats.stats = cif->codel.stats;
+
+ if ((error = copyout((caddr_t)&stats, ubuf, sizeof(stats))) != 0)
+ return (error);
+ *nbytes = sizeof(stats);
+
+ return (0);
+}
+
+static int
+codel_request(struct ifaltq *ifq, int req, void *arg)
+{
+ struct codel_if *cif = (struct codel_if *)ifq->altq_disc;
+ struct mbuf *m;
+
+ IFQ_LOCK_ASSERT(ifq);
+
+ switch (req) {
+ case ALTRQ_PURGE:
+ if (!ALTQ_IS_ENABLED(cif->cif_ifq))
+ break;
+
+ if (qempty(cif->cl_q))
+ break;
+
+ while ((m = _getq(cif->cl_q)) != NULL) {
+ PKTCNTR_ADD(&cif->cl_stats.cl_dropcnt, m_pktlen(m));
+ m_freem(m);
+ IFQ_DEC_LEN(cif->cif_ifq);
+ }
+ cif->cif_ifq->ifq_len = 0;
+ break;
+ }
+
+ return (0);
+}
+
+static int
+codel_enqueue(struct ifaltq *ifq, struct mbuf *m, struct altq_pktattr *pktattr)
+{
+
+ struct codel_if *cif = (struct codel_if *) ifq->altq_disc;
+
+ IFQ_LOCK_ASSERT(ifq);
+
+ /* grab class set by classifier */
+ if ((m->m_flags & M_PKTHDR) == 0) {
+ /* should not happen */
+ printf("altq: packet for %s does not have pkthdr\n",
+ ifq->altq_ifp->if_xname);
+ m_freem(m);
+ PKTCNTR_ADD(&cif->cl_stats.cl_dropcnt, m_pktlen(m));
+ return (ENOBUFS);
+ }
+
+ if (codel_addq(&cif->codel, cif->cl_q, m)) {
+ PKTCNTR_ADD(&cif->cl_stats.cl_dropcnt, m_pktlen(m));
+ return (ENOBUFS);
+ }
+ IFQ_INC_LEN(ifq);
+
+ return (0);
+}
+
+static struct mbuf *
+codel_dequeue(struct ifaltq *ifq, int op)
+{
+ struct codel_if *cif = (struct codel_if *)ifq->altq_disc;
+ struct mbuf *m;
+
+ IFQ_LOCK_ASSERT(ifq);
+
+ if (IFQ_IS_EMPTY(ifq))
+ return (NULL);
+
+ if (op == ALTDQ_POLL)
+ return (qhead(cif->cl_q));
+
+
+ m = codel_getq(&cif->codel, cif->cl_q);
+ if (m != NULL) {
+ IFQ_DEC_LEN(ifq);
+ PKTCNTR_ADD(&cif->cl_stats.cl_xmitcnt, m_pktlen(m));
+ return (m);
+ }
+
+ return (NULL);
+}
+
+struct codel *
+codel_alloc(int target, int interval, int ecn)
+{
+ struct codel *c;
+
+ c = malloc(sizeof(*c), M_DEVBUF, M_NOWAIT | M_ZERO);
+ if (c != NULL) {
+ c->params.target = machclk_freq * target / 1000;
+ c->params.interval = machclk_freq * interval / 1000;
+ c->params.ecn = ecn;
+ c->stats.maxpacket = 256;
+ }
+
+ return (c);
+}
+
+void
+codel_destroy(struct codel *c)
+{
+
+ free(c, M_DEVBUF);
+}
+
+#define MTAG_CODEL 1438031249
+int
+codel_addq(struct codel *c, class_queue_t *q, struct mbuf *m)
+{
+ struct m_tag *mtag;
+ uint64_t *enqueue_time;
+
+ if (qlen(q) < qlimit(q)) {
+ mtag = m_tag_locate(m, MTAG_CODEL, 0, NULL);
+ if (mtag == NULL)
+ mtag = m_tag_alloc(MTAG_CODEL, 0, sizeof(uint64_t),
+ M_NOWAIT);
+ if (mtag == NULL) {
+ m_freem(m);
+ return (-1);
+ }
+ enqueue_time = (uint64_t *)(mtag + 1);
+ *enqueue_time = read_machclk();
+ m_tag_prepend(m, mtag);
+ _addq(q, m);
+ return (0);
+ }
+ c->drop_overlimit++;
+ m_freem(m);
+
+ return (-1);
+}
+
+static int
+codel_should_drop(struct codel *c, class_queue_t *q, struct mbuf *m,
+ u_int64_t now)
+{
+ struct m_tag *mtag;
+ uint64_t *enqueue_time;
+
+ if (m == NULL) {
+ c->vars.first_above_time = 0;
+ return (0);
+ }
+
+ mtag = m_tag_locate(m, MTAG_CODEL, 0, NULL);
+ if (mtag == NULL) {
+ /* Only one warning per second. */
+ if (ppsratecheck(&c->last_log, &c->last_pps, 1))
+ printf("%s: could not found the packet mtag!\n",
+ __func__);
+ c->vars.first_above_time = 0;
+ return (0);
+ }
+ enqueue_time = (uint64_t *)(mtag + 1);
+ c->vars.ldelay = now - *enqueue_time;
+ c->stats.maxpacket = MAX(c->stats.maxpacket, m_pktlen(m));
+
+ if (codel_time_before(c->vars.ldelay, c->params.target) ||
+ qsize(q) <= c->stats.maxpacket) {
+ /* went below - stay below for at least interval */
+ c->vars.first_above_time = 0;
+ return (0);
+ }
+ if (c->vars.first_above_time == 0) {
+ /* just went above from below. If we stay above
+ * for at least interval we'll say it's ok to drop
+ */
+ c->vars.first_above_time = now + c->params.interval;
+ return (0);
+ }
+ if (codel_time_after(now, c->vars.first_above_time))
+ return (1);
+
+ return (0);
+}
+
+/*
+ * Run a Newton method step:
+ * new_invsqrt = (invsqrt / 2) * (3 - count * invsqrt^2)
+ *
+ * Here, invsqrt is a fixed point number (< 1.0), 32bit mantissa, aka Q0.32
+ */
+static void
+codel_Newton_step(struct codel_vars *vars)
+{
+ uint32_t invsqrt, invsqrt2;
+ uint64_t val;
+
+/* sizeof_in_bits(rec_inv_sqrt) */
+#define REC_INV_SQRT_BITS (8 * sizeof(u_int16_t))
+/* needed shift to get a Q0.32 number from rec_inv_sqrt */
+#define REC_INV_SQRT_SHIFT (32 - REC_INV_SQRT_BITS)
+
+ invsqrt = ((u_int32_t)vars->rec_inv_sqrt) << REC_INV_SQRT_SHIFT;
+ invsqrt2 = ((u_int64_t)invsqrt * invsqrt) >> 32;
+ val = (3LL << 32) - ((u_int64_t)vars->count * invsqrt2);
+ val >>= 2; /* avoid overflow in following multiply */
+ val = (val * invsqrt) >> (32 - 2 + 1);
+
+ vars->rec_inv_sqrt = val >> REC_INV_SQRT_SHIFT;
+}
+
+static u_int64_t
+codel_control_law(u_int64_t t, u_int64_t interval, u_int32_t rec_inv_sqrt)
+{
+
+ return (t + (u_int32_t)(((u_int64_t)interval *
+ (rec_inv_sqrt << REC_INV_SQRT_SHIFT)) >> 32));
+}
+
+struct mbuf *
+codel_getq(struct codel *c, class_queue_t *q)
+{
+ struct mbuf *m;
+ u_int64_t now;
+ int drop;
+
+ if ((m = _getq(q)) == NULL) {
+ c->vars.dropping = 0;
+ return (m);
+ }
+
+ now = read_machclk();
+ drop = codel_should_drop(c, q, m, now);
+ if (c->vars.dropping) {
+ if (!drop) {
+ /* sojourn time below target - leave dropping state */
+ c->vars.dropping = 0;
+ } else if (codel_time_after_eq(now, c->vars.drop_next)) {
+ /* It's time for the next drop. Drop the current
+ * packet and dequeue the next. The dequeue might
+ * take us out of dropping state.
+ * If not, schedule the next drop.
+ * A large backlog might result in drop rates so high
+ * that the next drop should happen now,
+ * hence the while loop.
+ */
+ while (c->vars.dropping &&
+ codel_time_after_eq(now, c->vars.drop_next)) {
+ c->vars.count++; /* don't care of possible wrap
+ * since there is no more
+ * divide */
+ codel_Newton_step(&c->vars);
+ /* TODO ECN */
+ PKTCNTR_ADD(&c->stats.drop_cnt, m_pktlen(m));
+ m_freem(m);
+ m = _getq(q);
+ if (!codel_should_drop(c, q, m, now))
+ /* leave dropping state */
+ c->vars.dropping = 0;
+ else
+ /* and schedule the next drop */
+ c->vars.drop_next =
+ codel_control_law(c->vars.drop_next,
+ c->params.interval,
+ c->vars.rec_inv_sqrt);
+ }
+ }
+ } else if (drop) {
+ /* TODO ECN */
+ PKTCNTR_ADD(&c->stats.drop_cnt, m_pktlen(m));
+ m_freem(m);
+
+ m = _getq(q);
+ drop = codel_should_drop(c, q, m, now);
+
+ c->vars.dropping = 1;
+ /* if min went above target close to when we last went below it
+ * assume that the drop rate that controlled the queue on the
+ * last cycle is a good starting point to control it now.
+ */
+ if (codel_time_before(now - c->vars.drop_next,
+ 16 * c->params.interval)) {
+ c->vars.count = (c->vars.count - c->vars.lastcount) | 1;
+ /* we dont care if rec_inv_sqrt approximation
+ * is not very precise :
+ * Next Newton steps will correct it quadratically.
+ */
+ codel_Newton_step(&c->vars);
+ } else {
+ c->vars.count = 1;
+ c->vars.rec_inv_sqrt = ~0U >> REC_INV_SQRT_SHIFT;
+ }
+ c->vars.lastcount = c->vars.count;
+ c->vars.drop_next = codel_control_law(now, c->params.interval,
+ c->vars.rec_inv_sqrt);
+ }
+
+ return (m);
+}
+
+void
+codel_getstats(struct codel *c, struct codel_stats *s)
+{
+ *s = c->stats;
+}
+
+#endif /* ALTQ_CODEL */
diff --git a/sys/contrib/altq/altq/altq_codel.h b/sys/contrib/altq/altq/altq_codel.h
new file mode 100644
index 0000000..991af42
--- /dev/null
+++ b/sys/contrib/altq/altq/altq_codel.h
@@ -0,0 +1,129 @@
+/*
+ * CoDel - The Controlled-Delay Active Queue Management algorithm
+ *
+ * Copyright (C) 2013 Ermal Luçi <eri@FreeBSD.org>
+ * Copyright (C) 2011-2012 Kathleen Nichols <nichols@pollere.com>
+ * Copyright (C) 2011-2012 Van Jacobson <van@pollere.net>
+ * Copyright (C) 2012 Michael D. Taht <dave.taht@bufferbloat.net>
+ * Copyright (C) 2012 Eric Dumazet <edumazet@google.com>
+ *
+ * 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,
+ * without modification.
+ * 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. The names of the authors may not be used to endorse or promote products
+ * derived from this software without specific prior written permission.
+ *
+ * Alternatively, provided that this notice is retained in full, this
+ * software may be distributed under the terms of the GNU General
+ * Public License ("GPL") version 2, in which case the provisions of the
+ * GPL apply INSTEAD OF those given above.
+ *
+ * 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 _ALTQ_ALTQ_CODEL_H_
+#define _ALTQ_ALTQ_CODEL_H_
+
+struct codel_stats {
+ u_int32_t maxpacket;
+ struct pktcntr drop_cnt;
+ u_int marked_packets;
+};
+
+struct codel_ifstats {
+ u_int qlength;
+ u_int qlimit;
+ struct codel_stats stats;
+ struct pktcntr cl_xmitcnt; /* transmitted packet counter */
+ struct pktcntr cl_dropcnt; /* dropped packet counter */
+};
+
+#ifdef _KERNEL
+#include <altq/altq_classq.h>
+
+/**
+ * struct codel_params - contains codel parameters
+ * <at> target: target queue size (in time units)
+ * <at> interval: width of moving time window
+ * <at> ecn: is Explicit Congestion Notification enabled
+ */
+struct codel_params {
+ u_int64_t target;
+ u_int64_t interval;
+ int ecn;
+};
+
+/**
+ * struct codel_vars - contains codel variables
+ * <at> count: how many drops we've done since the last time we
+ * entered dropping state
+ * <at> lastcount: count at entry to dropping state
+ * <at> dropping: set to true if in dropping state
+ * <at> rec_inv_sqrt: reciprocal value of sqrt(count) >> 1
+ * <at> first_above_time: when we went (or will go) continuously above
+ * target for interval
+ * <at> drop_next: time to drop next packet, or when we dropped last
+ * <at> ldelay: sojourn time of last dequeued packet
+ */
+struct codel_vars {
+ u_int32_t count;
+ u_int32_t lastcount;
+ int dropping;
+ u_int16_t rec_inv_sqrt;
+ u_int64_t first_above_time;
+ u_int64_t drop_next;
+ u_int64_t ldelay;
+};
+
+struct codel {
+ int last_pps;
+ struct codel_params params;
+ struct codel_vars vars;
+ struct codel_stats stats;
+ struct timeval last_log;
+ u_int32_t drop_overlimit;
+};
+
+/*
+ * codel interface state
+ */
+struct codel_if {
+ struct codel_if *cif_next; /* interface state list */
+ struct ifaltq *cif_ifq; /* backpointer to ifaltq */
+ u_int cif_bandwidth; /* link bandwidth in bps */
+
+ class_queue_t *cl_q; /* class queue structure */
+ struct codel codel;
+
+ /* statistics */
+ struct codel_ifstats cl_stats;
+};
+
+struct codel *codel_alloc(int, int, int);
+void codel_destroy(struct codel *);
+int codel_addq(struct codel *, class_queue_t *, struct mbuf *);
+struct mbuf *codel_getq(struct codel *, class_queue_t *);
+void codel_getstats(struct codel *, struct codel_stats *);
+
+#endif /* _KERNEL */
+
+#endif /* _ALTQ_ALTQ_CODEL_H_ */
diff --git a/sys/contrib/altq/altq/altq_fairq.c b/sys/contrib/altq/altq/altq_fairq.c
new file mode 100644
index 0000000..2189996
--- /dev/null
+++ b/sys/contrib/altq/altq/altq_fairq.c
@@ -0,0 +1,909 @@
+/*
+ * Copyright (c) 2008 The DragonFly Project. All rights reserved.
+ *
+ * This code is derived from software contributed to The DragonFly Project
+ * by Matthew Dillon <dillon@backplane.com>
+ *
+ * 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 DragonFly Project 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 HOLDERS 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.
+ *
+ * $DragonFly: src/sys/net/altq/altq_fairq.c,v 1.1 2008/04/06 18:58:15 dillon Exp $
+ * $FreeBSD$
+ */
+/*
+ * Matt: I gutted altq_priq.c and used it as a skeleton on which to build
+ * fairq. The fairq algorithm is completely different then priq, of course,
+ * but because I used priq's skeleton I believe I should include priq's
+ * copyright.
+ *
+ * Copyright (C) 2000-2003
+ * Sony Computer Science Laboratories 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 SONY CSL 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 SONY CSL 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.
+ */
+
+/*
+ * FAIRQ - take traffic classified by keep state (hashed into
+ * mbuf->m_pkthdr.altq_state_hash) and bucketize it. Fairly extract
+ * the first packet from each bucket in a round-robin fashion.
+ *
+ * TODO - better overall qlimit support (right now it is per-bucket).
+ * - NOTE: red etc is per bucket, not overall.
+ * - better service curve support.
+ *
+ * EXAMPLE:
+ *
+ * altq on em0 fairq bandwidth 650Kb queue { std, bulk }
+ * queue std priority 3 bandwidth 400Kb \
+ * fairq (buckets 64, default, hogs 1Kb) qlimit 50
+ * queue bulk priority 2 bandwidth 100Kb \
+ * fairq (buckets 64, hogs 1Kb) qlimit 50
+ *
+ * pass out on em0 from any to any keep state queue std
+ * pass out on em0 inet proto tcp ..... port ... keep state queue bulk
+ */
+#include "opt_altq.h"
+#include "opt_inet.h"
+#include "opt_inet6.h"
+
+#ifdef ALTQ_FAIRQ /* fairq is enabled in the kernel conf */
+
+#include <sys/param.h>
+#include <sys/malloc.h>
+#include <sys/mbuf.h>
+#include <sys/socket.h>
+#include <sys/sockio.h>
+#include <sys/systm.h>
+#include <sys/proc.h>
+#include <sys/errno.h>
+#include <sys/kernel.h>
+#include <sys/queue.h>
+
+#include <net/if.h>
+#include <net/if_var.h>
+#include <netinet/in.h>
+
+#include <netpfil/pf/pf.h>
+#include <netpfil/pf/pf_altq.h>
+#include <netpfil/pf/pf_mtag.h>
+#include <altq/altq.h>
+#include <altq/altq_fairq.h>
+
+/*
+ * function prototypes
+ */
+static int fairq_clear_interface(struct fairq_if *);
+static int fairq_request(struct ifaltq *, int, void *);
+static void fairq_purge(struct fairq_if *);
+static struct fairq_class *fairq_class_create(struct fairq_if *, int, int, u_int, struct fairq_opts *, int);
+static int fairq_class_destroy(struct fairq_class *);
+static int fairq_enqueue(struct ifaltq *, struct mbuf *, struct altq_pktattr *);
+static struct mbuf *fairq_dequeue(struct ifaltq *, int);
+
+static int fairq_addq(struct fairq_class *, struct mbuf *, u_int32_t);
+static struct mbuf *fairq_getq(struct fairq_class *, uint64_t);
+static struct mbuf *fairq_pollq(struct fairq_class *, uint64_t, int *);
+static fairq_bucket_t *fairq_selectq(struct fairq_class *, int);
+static void fairq_purgeq(struct fairq_class *);
+
+static void get_class_stats(struct fairq_classstats *, struct fairq_class *);
+static struct fairq_class *clh_to_clp(struct fairq_if *, uint32_t);
+
+int
+fairq_pfattach(struct pf_altq *a)
+{
+ struct ifnet *ifp;
+ int error;
+
+ if ((ifp = ifunit(a->ifname)) == NULL || a->altq_disc == NULL)
+ return (EINVAL);
+
+ error = altq_attach(&ifp->if_snd, ALTQT_FAIRQ, a->altq_disc,
+ fairq_enqueue, fairq_dequeue, fairq_request, NULL, NULL);
+
+ return (error);
+}
+
+int
+fairq_add_altq(struct pf_altq *a)
+{
+ struct fairq_if *pif;
+ struct ifnet *ifp;
+
+ if ((ifp = ifunit(a->ifname)) == NULL)
+ return (EINVAL);
+ if (!ALTQ_IS_READY(&ifp->if_snd))
+ return (ENODEV);
+
+
+ pif = malloc(sizeof(struct fairq_if),
+ M_DEVBUF, M_WAITOK | M_ZERO);
+ pif->pif_bandwidth = a->ifbandwidth;
+ pif->pif_maxpri = -1;
+ pif->pif_ifq = &ifp->if_snd;
+
+ /* keep the state in pf_altq */
+ a->altq_disc = pif;
+
+ return (0);
+}
+
+int
+fairq_remove_altq(struct pf_altq *a)
+{
+ struct fairq_if *pif;
+
+ if ((pif = a->altq_disc) == NULL)
+ return (EINVAL);
+ a->altq_disc = NULL;
+
+ fairq_clear_interface(pif);
+
+ free(pif, M_DEVBUF);
+ return (0);
+}
+
+int
+fairq_add_queue(struct pf_altq *a)
+{
+ struct fairq_if *pif;
+ struct fairq_class *cl;
+
+ if ((pif = a->altq_disc) == NULL)
+ return (EINVAL);
+
+ /* check parameters */
+ if (a->priority >= FAIRQ_MAXPRI)
+ return (EINVAL);
+ if (a->qid == 0)
+ return (EINVAL);
+ if (pif->pif_classes[a->priority] != NULL)
+ return (EBUSY);
+ if (clh_to_clp(pif, a->qid) != NULL)
+ return (EBUSY);
+
+ cl = fairq_class_create(pif, a->priority, a->qlimit, a->bandwidth,
+ &a->pq_u.fairq_opts, a->qid);
+ if (cl == NULL)
+ return (ENOMEM);
+
+ return (0);
+}
+
+int
+fairq_remove_queue(struct pf_altq *a)
+{
+ struct fairq_if *pif;
+ struct fairq_class *cl;
+
+ if ((pif = a->altq_disc) == NULL)
+ return (EINVAL);
+
+ if ((cl = clh_to_clp(pif, a->qid)) == NULL)
+ return (EINVAL);
+
+ return (fairq_class_destroy(cl));
+}
+
+int
+fairq_getqstats(struct pf_altq *a, void *ubuf, int *nbytes)
+{
+ struct fairq_if *pif;
+ struct fairq_class *cl;
+ struct fairq_classstats stats;
+ int error = 0;
+
+ if ((pif = altq_lookup(a->ifname, ALTQT_FAIRQ)) == NULL)
+ return (EBADF);
+
+ if ((cl = clh_to_clp(pif, a->qid)) == NULL)
+ return (EINVAL);
+
+ if (*nbytes < sizeof(stats))
+ return (EINVAL);
+
+ get_class_stats(&stats, cl);
+
+ if ((error = copyout((caddr_t)&stats, ubuf, sizeof(stats))) != 0)
+ return (error);
+ *nbytes = sizeof(stats);
+ return (0);
+}
+
+/*
+ * bring the interface back to the initial state by discarding
+ * all the filters and classes.
+ */
+static int
+fairq_clear_interface(struct fairq_if *pif)
+{
+ struct fairq_class *cl;
+ int pri;
+
+ /* clear out the classes */
+ for (pri = 0; pri <= pif->pif_maxpri; pri++) {
+ if ((cl = pif->pif_classes[pri]) != NULL)
+ fairq_class_destroy(cl);
+ }
+
+ return (0);
+}
+
+static int
+fairq_request(struct ifaltq *ifq, int req, void *arg)
+{
+ struct fairq_if *pif = (struct fairq_if *)ifq->altq_disc;
+
+ IFQ_LOCK_ASSERT(ifq);
+
+ switch (req) {
+ case ALTRQ_PURGE:
+ fairq_purge(pif);
+ break;
+ }
+ return (0);
+}
+
+/* discard all the queued packets on the interface */
+static void
+fairq_purge(struct fairq_if *pif)
+{
+ struct fairq_class *cl;
+ int pri;
+
+ for (pri = 0; pri <= pif->pif_maxpri; pri++) {
+ if ((cl = pif->pif_classes[pri]) != NULL && cl->cl_head)
+ fairq_purgeq(cl);
+ }
+ if (ALTQ_IS_ENABLED(pif->pif_ifq))
+ pif->pif_ifq->ifq_len = 0;
+}
+
+static struct fairq_class *
+fairq_class_create(struct fairq_if *pif, int pri, int qlimit,
+ u_int bandwidth, struct fairq_opts *opts, int qid)
+{
+ struct fairq_class *cl;
+ int flags = opts->flags;
+ u_int nbuckets = opts->nbuckets;
+ int i;
+
+#ifndef ALTQ_RED
+ if (flags & FARF_RED) {
+#ifdef ALTQ_DEBUG
+ printf("fairq_class_create: RED not configured for FAIRQ!\n");
+#endif
+ return (NULL);
+ }
+#endif
+#ifndef ALTQ_CODEL
+ if (flags & FARF_CODEL) {
+#ifdef ALTQ_DEBUG
+ printf("fairq_class_create: CODEL not configured for FAIRQ!\n");
+#endif
+ return (NULL);
+ }
+#endif
+ if (nbuckets == 0)
+ nbuckets = 256;
+ if (nbuckets > FAIRQ_MAX_BUCKETS)
+ nbuckets = FAIRQ_MAX_BUCKETS;
+ /* enforce power-of-2 size */
+ while ((nbuckets ^ (nbuckets - 1)) != ((nbuckets << 1) - 1))
+ ++nbuckets;
+
+ if ((cl = pif->pif_classes[pri]) != NULL) {
+ /* modify the class instead of creating a new one */
+ IFQ_LOCK(cl->cl_pif->pif_ifq);
+ if (cl->cl_head)
+ fairq_purgeq(cl);
+ IFQ_UNLOCK(cl->cl_pif->pif_ifq);
+#ifdef ALTQ_RIO
+ if (cl->cl_qtype == Q_RIO)
+ rio_destroy((rio_t *)cl->cl_red);
+#endif
+#ifdef ALTQ_RED
+ if (cl->cl_qtype == Q_RED)
+ red_destroy(cl->cl_red);
+#endif
+#ifdef ALTQ_CODEL
+ if (cl->cl_qtype == Q_CODEL)
+ codel_destroy(cl->cl_codel);
+#endif
+ } else {
+ cl = malloc(sizeof(struct fairq_class),
+ M_DEVBUF, M_WAITOK | M_ZERO);
+ cl->cl_nbuckets = nbuckets;
+ cl->cl_nbucket_mask = nbuckets - 1;
+
+ cl->cl_buckets = malloc(
+ sizeof(struct fairq_bucket) * cl->cl_nbuckets,
+ M_DEVBUF, M_WAITOK | M_ZERO);
+ cl->cl_head = NULL;
+ }
+
+ pif->pif_classes[pri] = cl;
+ if (flags & FARF_DEFAULTCLASS)
+ pif->pif_default = cl;
+ if (qlimit == 0)
+ qlimit = 50; /* use default */
+ cl->cl_qlimit = qlimit;
+ for (i = 0; i < cl->cl_nbuckets; ++i) {
+ qlimit(&cl->cl_buckets[i].queue) = qlimit;
+ }
+ cl->cl_bandwidth = bandwidth / 8;
+ cl->cl_qtype = Q_DROPTAIL;
+ cl->cl_flags = flags & FARF_USERFLAGS;
+ cl->cl_pri = pri;
+ if (pri > pif->pif_maxpri)
+ pif->pif_maxpri = pri;
+ cl->cl_pif = pif;
+ cl->cl_handle = qid;
+ cl->cl_hogs_m1 = opts->hogs_m1 / 8;
+ cl->cl_lssc_m1 = opts->lssc_m1 / 8; /* NOT YET USED */
+
+#ifdef ALTQ_RED
+ if (flags & (FARF_RED|FARF_RIO)) {
+ int red_flags, red_pkttime;
+
+ red_flags = 0;
+ if (flags & FARF_ECN)
+ red_flags |= REDF_ECN;
+#ifdef ALTQ_RIO
+ if (flags & FARF_CLEARDSCP)
+ red_flags |= RIOF_CLEARDSCP;
+#endif
+ if (pif->pif_bandwidth < 8)
+ red_pkttime = 1000 * 1000 * 1000; /* 1 sec */
+ else
+ red_pkttime = (int64_t)pif->pif_ifq->altq_ifp->if_mtu
+ * 1000 * 1000 * 1000 / (pif->pif_bandwidth / 8);
+#ifdef ALTQ_RIO
+ if (flags & FARF_RIO) {
+ cl->cl_red = (red_t *)rio_alloc(0, NULL,
+ red_flags, red_pkttime);
+ if (cl->cl_red != NULL)
+ cl->cl_qtype = Q_RIO;
+ } else
+#endif
+ if (flags & FARF_RED) {
+ cl->cl_red = red_alloc(0, 0,
+ cl->cl_qlimit * 10/100,
+ cl->cl_qlimit * 30/100,
+ red_flags, red_pkttime);
+ if (cl->cl_red != NULL)
+ cl->cl_qtype = Q_RED;
+ }
+ }
+#endif /* ALTQ_RED */
+#ifdef ALTQ_CODEL
+ if (flags & FARF_CODEL) {
+ cl->cl_codel = codel_alloc(5, 100, 0);
+ if (cl->cl_codel != NULL)
+ cl->cl_qtype = Q_CODEL;
+ }
+#endif
+
+ return (cl);
+}
+
+static int
+fairq_class_destroy(struct fairq_class *cl)
+{
+ struct fairq_if *pif;
+ int pri;
+
+ IFQ_LOCK(cl->cl_pif->pif_ifq);
+
+ if (cl->cl_head)
+ fairq_purgeq(cl);
+
+ pif = cl->cl_pif;
+ pif->pif_classes[cl->cl_pri] = NULL;
+ if (pif->pif_poll_cache == cl)
+ pif->pif_poll_cache = NULL;
+ if (pif->pif_maxpri == cl->cl_pri) {
+ for (pri = cl->cl_pri; pri >= 0; pri--)
+ if (pif->pif_classes[pri] != NULL) {
+ pif->pif_maxpri = pri;
+ break;
+ }
+ if (pri < 0)
+ pif->pif_maxpri = -1;
+ }
+ IFQ_UNLOCK(cl->cl_pif->pif_ifq);
+
+ if (cl->cl_red != NULL) {
+#ifdef ALTQ_RIO
+ if (cl->cl_qtype == Q_RIO)
+ rio_destroy((rio_t *)cl->cl_red);
+#endif
+#ifdef ALTQ_RED
+ if (cl->cl_qtype == Q_RED)
+ red_destroy(cl->cl_red);
+#endif
+#ifdef ALTQ_CODEL
+ if (cl->cl_qtype == Q_CODEL)
+ codel_destroy(cl->cl_codel);
+#endif
+ }
+ free(cl->cl_buckets, M_DEVBUF);
+ free(cl, M_DEVBUF);
+
+ return (0);
+}
+
+/*
+ * fairq_enqueue is an enqueue function to be registered to
+ * (*altq_enqueue) in struct ifaltq.
+ */
+static int
+fairq_enqueue(struct ifaltq *ifq, struct mbuf *m, struct altq_pktattr *pktattr)
+{
+ struct fairq_if *pif = (struct fairq_if *)ifq->altq_disc;
+ struct fairq_class *cl = NULL; /* Make compiler happy */
+ struct pf_mtag *t;
+ u_int32_t qid_hash = 0;
+ int len;
+
+ IFQ_LOCK_ASSERT(ifq);
+
+ /* grab class set by classifier */
+ if ((m->m_flags & M_PKTHDR) == 0) {
+ /* should not happen */
+ printf("altq: packet for %s does not have pkthdr\n",
+ ifq->altq_ifp->if_xname);
+ m_freem(m);
+ return (ENOBUFS);
+ }
+
+ if ((t = pf_find_mtag(m)) != NULL) {
+ cl = clh_to_clp(pif, t->qid);
+ qid_hash = t->qid_hash;
+ }
+ if (cl == NULL) {
+ cl = pif->pif_default;
+ if (cl == NULL) {
+ m_freem(m);
+ return (ENOBUFS);
+ }
+ }
+ cl->cl_flags |= FARF_HAS_PACKETS;
+ cl->cl_pktattr = NULL;
+ len = m_pktlen(m);
+ if (fairq_addq(cl, m, qid_hash) != 0) {
+ /* drop occurred. mbuf was freed in fairq_addq. */
+ PKTCNTR_ADD(&cl->cl_dropcnt, len);
+ return (ENOBUFS);
+ }
+ IFQ_INC_LEN(ifq);
+
+ return (0);
+}
+
+/*
+ * fairq_dequeue is a dequeue function to be registered to
+ * (*altq_dequeue) in struct ifaltq.
+ *
+ * note: ALTDQ_POLL returns the next packet without removing the packet
+ * from the queue. ALTDQ_REMOVE is a normal dequeue operation.
+ * ALTDQ_REMOVE must return the same packet if called immediately
+ * after ALTDQ_POLL.
+ */
+static struct mbuf *
+fairq_dequeue(struct ifaltq *ifq, int op)
+{
+ struct fairq_if *pif = (struct fairq_if *)ifq->altq_disc;
+ struct fairq_class *cl;
+ struct fairq_class *best_cl;
+ struct mbuf *best_m;
+ struct mbuf *m = NULL;
+ uint64_t cur_time = read_machclk();
+ int pri;
+ int hit_limit;
+
+ IFQ_LOCK_ASSERT(ifq);
+
+ if (IFQ_IS_EMPTY(ifq)) {
+ return (NULL);
+ }
+
+ if (pif->pif_poll_cache && op == ALTDQ_REMOVE) {
+ best_cl = pif->pif_poll_cache;
+ m = fairq_getq(best_cl, cur_time);
+ pif->pif_poll_cache = NULL;
+ if (m) {
+ IFQ_DEC_LEN(ifq);
+ PKTCNTR_ADD(&best_cl->cl_xmitcnt, m_pktlen(m));
+ return (m);
+ }
+ } else {
+ best_cl = NULL;
+ best_m = NULL;
+
+ for (pri = pif->pif_maxpri; pri >= 0; pri--) {
+ if ((cl = pif->pif_classes[pri]) == NULL)
+ continue;
+ if ((cl->cl_flags & FARF_HAS_PACKETS) == 0)
+ continue;
+ m = fairq_pollq(cl, cur_time, &hit_limit);
+ if (m == NULL) {
+ cl->cl_flags &= ~FARF_HAS_PACKETS;
+ continue;
+ }
+
+ /*
+ * Only override the best choice if we are under
+ * the BW limit.
+ */
+ if (hit_limit == 0 || best_cl == NULL) {
+ best_cl = cl;
+ best_m = m;
+ }
+
+ /*
+ * Remember the highest priority mbuf in case we
+ * do not find any lower priority mbufs.
+ */
+ if (hit_limit)
+ continue;
+ break;
+ }
+ if (op == ALTDQ_POLL) {
+ pif->pif_poll_cache = best_cl;
+ m = best_m;
+ } else if (best_cl) {
+ m = fairq_getq(best_cl, cur_time);
+ if (m != NULL) {
+ IFQ_DEC_LEN(ifq);
+ PKTCNTR_ADD(&best_cl->cl_xmitcnt, m_pktlen(m));
+ }
+ }
+ return (m);
+ }
+ return (NULL);
+}
+
+static int
+fairq_addq(struct fairq_class *cl, struct mbuf *m, u_int32_t bucketid)
+{
+ fairq_bucket_t *b;
+ u_int hindex;
+ uint64_t bw;
+
+ /*
+ * If the packet doesn't have any keep state put it on the end of
+ * our queue. XXX this can result in out of order delivery.
+ */
+ if (bucketid == 0) {
+ if (cl->cl_head)
+ b = cl->cl_head->prev;
+ else
+ b = &cl->cl_buckets[0];
+ } else {
+ hindex = bucketid & cl->cl_nbucket_mask;
+ b = &cl->cl_buckets[hindex];
+ }
+
+ /*
+ * Add the bucket to the end of the circular list of active buckets.
+ *
+ * As a special case we add the bucket to the beginning of the list
+ * instead of the end if it was not previously on the list and if
+ * its traffic is less then the hog level.
+ */
+ if (b->in_use == 0) {
+ b->in_use = 1;
+ if (cl->cl_head == NULL) {
+ cl->cl_head = b;
+ b->next = b;
+ b->prev = b;
+ } else {
+ b->next = cl->cl_head;
+ b->prev = cl->cl_head->prev;
+ b->prev->next = b;
+ b->next->prev = b;
+
+ if (b->bw_delta && cl->cl_hogs_m1) {
+ bw = b->bw_bytes * machclk_freq / b->bw_delta;
+ if (bw < cl->cl_hogs_m1)
+ cl->cl_head = b;
+ }
+ }
+ }
+
+#ifdef ALTQ_RIO
+ if (cl->cl_qtype == Q_RIO)
+ return rio_addq((rio_t *)cl->cl_red, &b->queue, m, cl->cl_pktattr);
+#endif
+#ifdef ALTQ_RED
+ if (cl->cl_qtype == Q_RED)
+ return red_addq(cl->cl_red, &b->queue, m, cl->cl_pktattr);
+#endif
+#ifdef ALTQ_CODEL
+ if (cl->cl_qtype == Q_CODEL)
+ return codel_addq(cl->cl_codel, &b->queue, m);
+#endif
+ if (qlen(&b->queue) >= qlimit(&b->queue)) {
+ m_freem(m);
+ return (-1);
+ }
+
+ if (cl->cl_flags & FARF_CLEARDSCP)
+ write_dsfield(m, cl->cl_pktattr, 0);
+
+ _addq(&b->queue, m);
+
+ return (0);
+}
+
+static struct mbuf *
+fairq_getq(struct fairq_class *cl, uint64_t cur_time)
+{
+ fairq_bucket_t *b;
+ struct mbuf *m;
+
+ b = fairq_selectq(cl, 0);
+ if (b == NULL)
+ m = NULL;
+#ifdef ALTQ_RIO
+ else if (cl->cl_qtype == Q_RIO)
+ m = rio_getq((rio_t *)cl->cl_red, &b->queue);
+#endif
+#ifdef ALTQ_RED
+ else if (cl->cl_qtype == Q_RED)
+ m = red_getq(cl->cl_red, &b->queue);
+#endif
+#ifdef ALTQ_CODEL
+ else if (cl->cl_qtype == Q_CODEL)
+ m = codel_getq(cl->cl_codel, &b->queue);
+#endif
+ else
+ m = _getq(&b->queue);
+
+ /*
+ * Calculate the BW change
+ */
+ if (m != NULL) {
+ uint64_t delta;
+
+ /*
+ * Per-class bandwidth calculation
+ */
+ delta = (cur_time - cl->cl_last_time);
+ if (delta > machclk_freq * 8)
+ delta = machclk_freq * 8;
+ cl->cl_bw_delta += delta;
+ cl->cl_bw_bytes += m->m_pkthdr.len;
+ cl->cl_last_time = cur_time;
+ cl->cl_bw_delta -= cl->cl_bw_delta >> 3;
+ cl->cl_bw_bytes -= cl->cl_bw_bytes >> 3;
+
+ /*
+ * Per-bucket bandwidth calculation
+ */
+ delta = (cur_time - b->last_time);
+ if (delta > machclk_freq * 8)
+ delta = machclk_freq * 8;
+ b->bw_delta += delta;
+ b->bw_bytes += m->m_pkthdr.len;
+ b->last_time = cur_time;
+ b->bw_delta -= b->bw_delta >> 3;
+ b->bw_bytes -= b->bw_bytes >> 3;
+ }
+ return(m);
+}
+
+/*
+ * Figure out what the next packet would be if there were no limits. If
+ * this class hits its bandwidth limit *hit_limit is set to no-zero, otherwise
+ * it is set to 0. A non-NULL mbuf is returned either way.
+ */
+static struct mbuf *
+fairq_pollq(struct fairq_class *cl, uint64_t cur_time, int *hit_limit)
+{
+ fairq_bucket_t *b;
+ struct mbuf *m;
+ uint64_t delta;
+ uint64_t bw;
+
+ *hit_limit = 0;
+ b = fairq_selectq(cl, 1);
+ if (b == NULL)
+ return(NULL);
+ m = qhead(&b->queue);
+
+ /*
+ * Did this packet exceed the class bandwidth? Calculate the
+ * bandwidth component of the packet.
+ *
+ * - Calculate bytes per second
+ */
+ delta = cur_time - cl->cl_last_time;
+ if (delta > machclk_freq * 8)
+ delta = machclk_freq * 8;
+ cl->cl_bw_delta += delta;
+ cl->cl_last_time = cur_time;
+ if (cl->cl_bw_delta) {
+ bw = cl->cl_bw_bytes * machclk_freq / cl->cl_bw_delta;
+
+ if (bw > cl->cl_bandwidth)
+ *hit_limit = 1;
+#ifdef ALTQ_DEBUG
+ printf("BW %6ju relative to %6u %d queue %p\n",
+ (uintmax_t)bw, cl->cl_bandwidth, *hit_limit, b);
+#endif
+ }
+ return(m);
+}
+
+/*
+ * Locate the next queue we want to pull a packet out of. This code
+ * is also responsible for removing empty buckets from the circular list.
+ */
+static
+fairq_bucket_t *
+fairq_selectq(struct fairq_class *cl, int ispoll)
+{
+ fairq_bucket_t *b;
+ uint64_t bw;
+
+ if (ispoll == 0 && cl->cl_polled) {
+ b = cl->cl_polled;
+ cl->cl_polled = NULL;
+ return(b);
+ }
+
+ while ((b = cl->cl_head) != NULL) {
+ /*
+ * Remove empty queues from consideration
+ */
+ if (qempty(&b->queue)) {
+ b->in_use = 0;
+ cl->cl_head = b->next;
+ if (cl->cl_head == b) {
+ cl->cl_head = NULL;
+ } else {
+ b->next->prev = b->prev;
+ b->prev->next = b->next;
+ }
+ continue;
+ }
+
+ /*
+ * Advance the round robin. Queues with bandwidths less
+ * then the hog bandwidth are allowed to burst.
+ */
+ if (cl->cl_hogs_m1 == 0) {
+ cl->cl_head = b->next;
+ } else if (b->bw_delta) {
+ bw = b->bw_bytes * machclk_freq / b->bw_delta;
+ if (bw >= cl->cl_hogs_m1) {
+ cl->cl_head = b->next;
+ }
+ /*
+ * XXX TODO -
+ */
+ }
+
+ /*
+ * Return bucket b.
+ */
+ break;
+ }
+ if (ispoll)
+ cl->cl_polled = b;
+ return(b);
+}
+
+static void
+fairq_purgeq(struct fairq_class *cl)
+{
+ fairq_bucket_t *b;
+ struct mbuf *m;
+
+ while ((b = fairq_selectq(cl, 0)) != NULL) {
+ while ((m = _getq(&b->queue)) != NULL) {
+ PKTCNTR_ADD(&cl->cl_dropcnt, m_pktlen(m));
+ m_freem(m);
+ }
+ ASSERT(qlen(&b->queue) == 0);
+ }
+}
+
+static void
+get_class_stats(struct fairq_classstats *sp, struct fairq_class *cl)
+{
+ fairq_bucket_t *b;
+
+ sp->class_handle = cl->cl_handle;
+ sp->qlimit = cl->cl_qlimit;
+ sp->xmit_cnt = cl->cl_xmitcnt;
+ sp->drop_cnt = cl->cl_dropcnt;
+ sp->qtype = cl->cl_qtype;
+ sp->qlength = 0;
+
+ if (cl->cl_head) {
+ b = cl->cl_head;
+ do {
+ sp->qlength += qlen(&b->queue);
+ b = b->next;
+ } while (b != cl->cl_head);
+ }
+
+#ifdef ALTQ_RED
+ if (cl->cl_qtype == Q_RED)
+ red_getstats(cl->cl_red, &sp->red[0]);
+#endif
+#ifdef ALTQ_RIO
+ if (cl->cl_qtype == Q_RIO)
+ rio_getstats((rio_t *)cl->cl_red, &sp->red[0]);
+#endif
+#ifdef ALTQ_CODEL
+ if (cl->cl_qtype == Q_CODEL)
+ codel_getstats(cl->cl_codel, &sp->codel);
+#endif
+}
+
+/* convert a class handle to the corresponding class pointer */
+static struct fairq_class *
+clh_to_clp(struct fairq_if *pif, uint32_t chandle)
+{
+ struct fairq_class *cl;
+ int idx;
+
+ if (chandle == 0)
+ return (NULL);
+
+ for (idx = pif->pif_maxpri; idx >= 0; idx--)
+ if ((cl = pif->pif_classes[idx]) != NULL &&
+ cl->cl_handle == chandle)
+ return (cl);
+
+ return (NULL);
+}
+
+#endif /* ALTQ_FAIRQ */
diff --git a/sys/contrib/altq/altq/altq_fairq.h b/sys/contrib/altq/altq/altq_fairq.h
new file mode 100644
index 0000000..8fede5b
--- /dev/null
+++ b/sys/contrib/altq/altq/altq_fairq.h
@@ -0,0 +1,145 @@
+/*
+ * Copyright (c) 2008 The DragonFly Project. All rights reserved.
+ *
+ * This code is derived from software contributed to The DragonFly Project
+ * by Matthew Dillon <dillon@backplane.com>
+ *
+ * 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 DragonFly Project 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 HOLDERS 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.
+ *
+ * $DragonFly: src/sys/net/altq/altq_fairq.h,v 1.1 2008/04/06 18:58:15 dillon Exp $
+ * $FreeBSD$
+ */
+
+#ifndef _ALTQ_ALTQ_FAIRQ_H_
+#define _ALTQ_ALTQ_FAIRQ_H_
+
+#include <altq/altq.h>
+#include <altq/altq_classq.h>
+#include <altq/altq_codel.h>
+#include <altq/altq_red.h>
+#include <altq/altq_rio.h>
+#include <altq/altq_rmclass.h>
+
+#define FAIRQ_MAX_BUCKETS 2048 /* maximum number of sorting buckets */
+#define FAIRQ_MAXPRI RM_MAXPRIO
+#define FAIRQ_BITMAP_WIDTH (sizeof(fairq_bitmap_t)*8)
+#define FAIRQ_BITMAP_MASK (FAIRQ_BITMAP_WIDTH - 1)
+
+/* fairq class flags */
+#define FARF_RED 0x0001 /* use RED */
+#define FARF_ECN 0x0002 /* use RED/ECN */
+#define FARF_RIO 0x0004 /* use RIO */
+#define FARF_CODEL 0x0008 /* use CoDel */
+#define FARF_CLEARDSCP 0x0010 /* clear diffserv codepoint */
+#define FARF_DEFAULTCLASS 0x1000 /* default class */
+
+#define FARF_HAS_PACKETS 0x2000 /* might have queued packets */
+
+#define FARF_USERFLAGS (FARF_RED|FARF_ECN|FARF_RIO|FARF_CLEARDSCP| \
+ FARF_DEFAULTCLASS)
+
+/* special class handles */
+#define FAIRQ_NULLCLASS_HANDLE 0
+
+typedef u_int fairq_bitmap_t;
+
+struct fairq_classstats {
+ uint32_t class_handle;
+
+ u_int qlength;
+ u_int qlimit;
+ struct pktcntr xmit_cnt; /* transmitted packet counter */
+ struct pktcntr drop_cnt; /* dropped packet counter */
+
+ /* codel, red and rio related info */
+ int qtype;
+ struct redstats red[3]; /* rio has 3 red stats */
+ struct codel_stats codel;
+};
+
+#ifdef _KERNEL
+
+typedef struct fairq_bucket {
+ struct fairq_bucket *next; /* circular list */
+ struct fairq_bucket *prev; /* circular list */
+ class_queue_t queue; /* the actual queue */
+ uint64_t bw_bytes; /* statistics used to calculate bw */
+ uint64_t bw_delta; /* statistics used to calculate bw */
+ uint64_t last_time;
+ int in_use;
+} fairq_bucket_t;
+
+struct fairq_class {
+ uint32_t cl_handle; /* class handle */
+ u_int cl_nbuckets; /* (power of 2) */
+ u_int cl_nbucket_mask; /* bucket mask */
+ fairq_bucket_t *cl_buckets;
+ fairq_bucket_t *cl_head; /* head of circular bucket list */
+ fairq_bucket_t *cl_polled;
+ union {
+ struct red *cl_red; /* RED state */
+ struct codel *cl_codel; /* CoDel state */
+ } cl_aqm;
+#define cl_red cl_aqm.cl_red
+#define cl_codel cl_aqm.cl_codel
+ u_int cl_hogs_m1;
+ u_int cl_lssc_m1;
+ u_int cl_bandwidth;
+ uint64_t cl_bw_bytes;
+ uint64_t cl_bw_delta;
+ uint64_t cl_last_time;
+ int cl_qtype; /* rollup */
+ int cl_qlimit;
+ int cl_pri; /* priority */
+ int cl_flags; /* class flags */
+ struct fairq_if *cl_pif; /* back pointer to pif */
+ struct altq_pktattr *cl_pktattr; /* saved header used by ECN */
+
+ /* round robin index */
+
+ /* statistics */
+ struct pktcntr cl_xmitcnt; /* transmitted packet counter */
+ struct pktcntr cl_dropcnt; /* dropped packet counter */
+};
+
+/*
+ * fairq interface state
+ */
+struct fairq_if {
+ struct fairq_if *pif_next; /* interface state list */
+ struct ifaltq *pif_ifq; /* backpointer to ifaltq */
+ u_int pif_bandwidth; /* link bandwidth in bps */
+ int pif_maxpri; /* max priority in use */
+ struct fairq_class *pif_poll_cache;/* cached poll */
+ struct fairq_class *pif_default; /* default class */
+ struct fairq_class *pif_classes[FAIRQ_MAXPRI]; /* classes */
+};
+
+#endif /* _KERNEL */
+
+#endif /* _ALTQ_ALTQ_FAIRQ_H_ */
diff --git a/sys/contrib/altq/altq/altq_hfsc.c b/sys/contrib/altq/altq/altq_hfsc.c
index 0363016..d4f7ab9 100644
--- a/sys/contrib/altq/altq/altq_hfsc.c
+++ b/sys/contrib/altq/altq/altq_hfsc.c
@@ -391,6 +391,14 @@ hfsc_class_create(struct hfsc_if *hif, struct service_curve *rsc,
return (NULL);
}
#endif
+#ifndef ALTQ_CODEL
+ if (flags & HFCF_CODEL) {
+#ifdef ALTQ_DEBUG
+ printf("hfsc_class_create: CODEL not configured for HFSC!\n");
+#endif
+ return (NULL);
+ }
+#endif
cl = malloc(sizeof(struct hfsc_class), M_DEVBUF, M_NOWAIT | M_ZERO);
if (cl == NULL)
@@ -407,6 +415,7 @@ hfsc_class_create(struct hfsc_if *hif, struct service_curve *rsc,
qlimit(cl->cl_q) = qlimit;
qtype(cl->cl_q) = Q_DROPTAIL;
qlen(cl->cl_q) = 0;
+ qsize(cl->cl_q) = 0;
cl->cl_flags = flags;
#ifdef ALTQ_RED
if (flags & (HFCF_RED|HFCF_RIO)) {
@@ -451,6 +460,13 @@ hfsc_class_create(struct hfsc_if *hif, struct service_curve *rsc,
#endif
}
#endif /* ALTQ_RED */
+#ifdef ALTQ_CODEL
+ if (flags & HFCF_CODEL) {
+ cl->cl_codel = codel_alloc(5, 100, 0);
+ if (cl->cl_codel != NULL)
+ qtype(cl->cl_q) = Q_CODEL;
+ }
+#endif
if (rsc != NULL && (rsc->m1 != 0 || rsc->m2 != 0)) {
cl->cl_rsc = malloc(sizeof(struct internal_sc),
@@ -543,6 +559,10 @@ hfsc_class_create(struct hfsc_if *hif, struct service_curve *rsc,
if (q_is_red(cl->cl_q))
red_destroy(cl->cl_red);
#endif
+#ifdef ALTQ_CODEL
+ if (q_is_codel(cl->cl_q))
+ codel_destroy(cl->cl_codel);
+#endif
}
if (cl->cl_fsc != NULL)
free(cl->cl_fsc, M_DEVBUF);
@@ -617,6 +637,10 @@ hfsc_class_destroy(struct hfsc_class *cl)
if (q_is_red(cl->cl_q))
red_destroy(cl->cl_red);
#endif
+#ifdef ALTQ_CODEL
+ if (q_is_codel(cl->cl_q))
+ codel_destroy(cl->cl_codel);
+#endif
}
IFQ_LOCK(cl->cl_hif->hif_ifq);
@@ -844,6 +868,10 @@ hfsc_addq(struct hfsc_class *cl, struct mbuf *m)
if (q_is_red(cl->cl_q))
return red_addq(cl->cl_red, cl->cl_q, m, cl->cl_pktattr);
#endif
+#ifdef ALTQ_CODEL
+ if (q_is_codel(cl->cl_q))
+ return codel_addq(cl->cl_codel, cl->cl_q, m);
+#endif
if (qlen(cl->cl_q) >= qlimit(cl->cl_q)) {
m_freem(m);
return (-1);
@@ -868,6 +896,10 @@ hfsc_getq(struct hfsc_class *cl)
if (q_is_red(cl->cl_q))
return red_getq(cl->cl_red, cl->cl_q);
#endif
+#ifdef ALTQ_CODEL
+ if (q_is_codel(cl->cl_q))
+ return codel_getq(cl->cl_codel, cl->cl_q);
+#endif
return _getq(cl->cl_q);
}
@@ -1652,6 +1684,10 @@ get_class_stats(struct hfsc_classstats *sp, struct hfsc_class *cl)
if (q_is_rio(cl->cl_q))
rio_getstats((rio_t *)cl->cl_red, &sp->red[0]);
#endif
+#ifdef ALTQ_CODEL
+ if (q_is_codel(cl->cl_q))
+ codel_getstats(cl->cl_codel, &sp->codel);
+#endif
}
/* convert a class handle to the corresponding class pointer */
diff --git a/sys/contrib/altq/altq/altq_hfsc.h b/sys/contrib/altq/altq/altq_hfsc.h
index e5595cb..b0228c5 100644
--- a/sys/contrib/altq/altq/altq_hfsc.h
+++ b/sys/contrib/altq/altq/altq_hfsc.h
@@ -34,6 +34,7 @@
#include <altq/altq.h>
#include <altq/altq_classq.h>
+#include <altq/altq_codel.h>
#include <altq/altq_red.h>
#include <altq/altq_rio.h>
@@ -55,6 +56,7 @@ struct service_curve {
#define HFCF_RED 0x0001 /* use RED */
#define HFCF_ECN 0x0002 /* use RED/ECN */
#define HFCF_RIO 0x0004 /* use RIO */
+#define HFCF_CODEL 0x0008 /* use CoDel */
#define HFCF_CLEARDSCP 0x0010 /* clear diffserv codepoint */
#define HFCF_DEFAULTCLASS 0x1000 /* default class */
@@ -101,9 +103,10 @@ struct hfsc_classstats {
u_int parentperiod; /* parent's vt period seqno */
int nactive; /* number of active children */
- /* red and rio related info */
+ /* codel, red and rio related info */
int qtype;
struct redstats red[3];
+ struct codel_stats codel;
};
#ifdef ALTQ3_COMPAT
@@ -229,7 +232,12 @@ struct hfsc_class {
struct hfsc_class *cl_children; /* child classes */
class_queue_t *cl_q; /* class queue structure */
- struct red *cl_red; /* RED state */
+ union {
+ struct red *cl_red; /* RED state */
+ struct codel *cl_codel; /* CoDel state */
+ } cl_aqm;
+#define cl_red cl_aqm.cl_red
+#define cl_codel cl_aqm.cl_codel
struct altq_pktattr *cl_pktattr; /* saved header used by ECN */
u_int64_t cl_total; /* total work in bytes */
diff --git a/sys/contrib/altq/altq/altq_priq.c b/sys/contrib/altq/altq/altq_priq.c
index 3ce65dc..c77ba40 100644
--- a/sys/contrib/altq/altq/altq_priq.c
+++ b/sys/contrib/altq/altq/altq_priq.c
@@ -297,6 +297,14 @@ priq_class_create(struct priq_if *pif, int pri, int qlimit, int flags, int qid)
return (NULL);
}
#endif
+#ifndef ALTQ_CODEL
+ if (flags & PRCF_CODEL) {
+#ifdef ALTQ_DEBUG
+ printf("priq_class_create: CODEL not configured for PRIQ!\n");
+#endif
+ return (NULL);
+ }
+#endif
if ((cl = pif->pif_classes[pri]) != NULL) {
/* modify the class instead of creating a new one */
@@ -318,6 +326,10 @@ priq_class_create(struct priq_if *pif, int pri, int qlimit, int flags, int qid)
if (q_is_red(cl->cl_q))
red_destroy(cl->cl_red);
#endif
+#ifdef ALTQ_CODEL
+ if (q_is_codel(cl->cl_q))
+ codel_destroy(cl->cl_codel);
+#endif
} else {
cl = malloc(sizeof(struct priq_class), M_DEVBUF,
M_NOWAIT | M_ZERO);
@@ -338,6 +350,7 @@ priq_class_create(struct priq_if *pif, int pri, int qlimit, int flags, int qid)
qlimit(cl->cl_q) = qlimit;
qtype(cl->cl_q) = Q_DROPTAIL;
qlen(cl->cl_q) = 0;
+ qsize(cl->cl_q) = 0;
cl->cl_flags = flags;
cl->cl_pri = pri;
if (pri > pif->pif_maxpri)
@@ -381,6 +394,13 @@ priq_class_create(struct priq_if *pif, int pri, int qlimit, int flags, int qid)
}
}
#endif /* ALTQ_RED */
+#ifdef ALTQ_CODEL
+ if (flags & PRCF_CODEL) {
+ cl->cl_codel = codel_alloc(5, 100, 0);
+ if (cl->cl_codel != NULL)
+ qtype(cl->cl_q) = Q_CODEL;
+ }
+#endif
return (cl);
@@ -394,6 +414,10 @@ priq_class_create(struct priq_if *pif, int pri, int qlimit, int flags, int qid)
if (q_is_red(cl->cl_q))
red_destroy(cl->cl_red);
#endif
+#ifdef ALTQ_CODEL
+ if (q_is_codel(cl->cl_q))
+ codel_destroy(cl->cl_codel);
+#endif
}
if (cl->cl_q != NULL)
free(cl->cl_q, M_DEVBUF);
@@ -445,6 +469,10 @@ priq_class_destroy(struct priq_class *cl)
if (q_is_red(cl->cl_q))
red_destroy(cl->cl_red);
#endif
+#ifdef ALTQ_CODEL
+ if (q_is_codel(cl->cl_q))
+ codel_destroy(cl->cl_codel);
+#endif
}
free(cl->cl_q, M_DEVBUF);
free(cl, M_DEVBUF);
@@ -560,6 +588,10 @@ priq_addq(struct priq_class *cl, struct mbuf *m)
if (q_is_red(cl->cl_q))
return red_addq(cl->cl_red, cl->cl_q, m, cl->cl_pktattr);
#endif
+#ifdef ALTQ_CODEL
+ if (q_is_codel(cl->cl_q))
+ return codel_addq(cl->cl_codel, cl->cl_q, m);
+#endif
if (qlen(cl->cl_q) >= qlimit(cl->cl_q)) {
m_freem(m);
return (-1);
@@ -584,6 +616,10 @@ priq_getq(struct priq_class *cl)
if (q_is_red(cl->cl_q))
return red_getq(cl->cl_red, cl->cl_q);
#endif
+#ifdef ALTQ_CODEL
+ if (q_is_codel(cl->cl_q))
+ return codel_getq(cl->cl_codel, cl->cl_q);
+#endif
return _getq(cl->cl_q);
}
@@ -628,7 +664,10 @@ get_class_stats(struct priq_classstats *sp, struct priq_class *cl)
if (q_is_rio(cl->cl_q))
rio_getstats((rio_t *)cl->cl_red, &sp->red[0]);
#endif
-
+#ifdef ALTQ_CODEL
+ if (q_is_codel(cl->cl_q))
+ codel_getstats(cl->cl_codel, &sp->codel);
+#endif
}
/* convert a class handle to the corresponding class pointer */
diff --git a/sys/contrib/altq/altq/altq_priq.h b/sys/contrib/altq/altq/altq_priq.h
index 481d31b..110847d 100644
--- a/sys/contrib/altq/altq/altq_priq.h
+++ b/sys/contrib/altq/altq/altq_priq.h
@@ -30,6 +30,7 @@
#include <altq/altq.h>
#include <altq/altq_classq.h>
+#include <altq/altq_codel.h>
#include <altq/altq_red.h>
#include <altq/altq_rio.h>
@@ -59,6 +60,7 @@ struct priq_add_class {
#define PRCF_RED 0x0001 /* use RED */
#define PRCF_ECN 0x0002 /* use RED/ECN */
#define PRCF_RIO 0x0004 /* use RIO */
+#define PRCF_CODEL 0x0008 /* use CoDel */
#define PRCF_CLEARDSCP 0x0010 /* clear diffserv codepoint */
#define PRCF_DEFAULTCLASS 0x1000 /* default class */
@@ -102,9 +104,10 @@ struct priq_classstats {
struct pktcntr xmitcnt; /* transmitted packet counter */
struct pktcntr dropcnt; /* dropped packet counter */
- /* red and rio related info */
+ /* codel, red and rio related info */
int qtype;
struct redstats red[3]; /* rio has 3 red stats */
+ struct codel_stats codel;
};
#ifdef ALTQ3_COMPAT
@@ -134,7 +137,12 @@ struct priq_class_stats {
struct priq_class {
u_int32_t cl_handle; /* class handle */
class_queue_t *cl_q; /* class queue structure */
- struct red *cl_red; /* RED state */
+ union {
+ struct red *cl_red; /* RED state */
+ struct codel *cl_codel; /* CoDel state */
+ } cl_aqm;
+#define cl_red cl_aqm.cl_red
+#define cl_codel cl_aqm.cl_codel
int cl_pri; /* priority */
int cl_flags; /* class flags */
struct priq_if *cl_pif; /* back pointer to pif */
diff --git a/sys/contrib/altq/altq/altq_rmclass.c b/sys/contrib/altq/altq/altq_rmclass.c
index c433024..656d0ee 100644
--- a/sys/contrib/altq/altq/altq_rmclass.c
+++ b/sys/contrib/altq/altq/altq_rmclass.c
@@ -68,6 +68,7 @@
#include <altq/if_altq.h>
#include <altq/altq.h>
+#include <altq/altq_codel.h>
#include <altq/altq_rmclass.h>
#include <altq/altq_rmclass_debug.h>
#include <altq/altq_red.h>
@@ -218,6 +219,14 @@ rmc_newclass(int pri, struct rm_ifdat *ifd, u_int nsecPerByte,
return (NULL);
}
#endif
+#ifndef ALTQ_CODEL
+ if (flags & RMCF_CODEL) {
+#ifdef ALTQ_DEBUG
+ printf("rmc_newclass: CODEL not configured for CBQ!\n");
+#endif
+ return (NULL);
+ }
+#endif
cl = malloc(sizeof(struct rm_class), M_DEVBUF, M_NOWAIT | M_ZERO);
if (cl == NULL)
@@ -302,6 +311,13 @@ rmc_newclass(int pri, struct rm_ifdat *ifd, u_int nsecPerByte,
#endif
}
#endif /* ALTQ_RED */
+#ifdef ALTQ_CODEL
+ if (flags & RMCF_CODEL) {
+ cl->codel_ = codel_alloc(5, 100, 0);
+ if (cl->codel_ != NULL)
+ qtype(cl->q_) = Q_CODEL;
+ }
+#endif
/*
* put the class into the class tree
@@ -652,6 +668,10 @@ rmc_delete_class(struct rm_ifdat *ifd, struct rm_class *cl)
if (q_is_red(cl->q_))
red_destroy(cl->red_);
#endif
+#ifdef ALTQ_CODEL
+ if (q_is_codel(cl->q_))
+ codel_destroy(cl->codel_);
+#endif
}
free(cl->q_, M_DEVBUF);
free(cl, M_DEVBUF);
@@ -1618,6 +1638,10 @@ _rmc_addq(rm_class_t *cl, mbuf_t *m)
if (q_is_red(cl->q_))
return red_addq(cl->red_, cl->q_, m, cl->pktattr_);
#endif /* ALTQ_RED */
+#ifdef ALTQ_CODEL
+ if (q_is_codel(cl->q_))
+ return codel_addq(cl->codel_, cl->q_, m);
+#endif
if (cl->flags_ & RMCF_CLEARDSCP)
write_dsfield(m, cl->pktattr_, 0);
@@ -1647,6 +1671,10 @@ _rmc_getq(rm_class_t *cl)
if (q_is_red(cl->q_))
return red_getq(cl->red_, cl->q_);
#endif
+#ifdef ALTQ_CODEL
+ if (q_is_codel(cl->q_))
+ return codel_getq(cl->codel_, cl->q_);
+#endif
return _getq(cl->q_);
}
@@ -1717,7 +1745,8 @@ void cbqtrace_dump(int counter)
#endif /* CBQ_TRACE */
#endif /* ALTQ_CBQ */
-#if defined(ALTQ_CBQ) || defined(ALTQ_RED) || defined(ALTQ_RIO) || defined(ALTQ_HFSC) || defined(ALTQ_PRIQ)
+#if defined(ALTQ_CBQ) || defined(ALTQ_RED) || defined(ALTQ_RIO) || \
+ defined(ALTQ_HFSC) || defined(ALTQ_PRIQ) || defined(ALTQ_CODEL)
#if !defined(__GNUC__) || defined(ALTQ_DEBUG)
void
diff --git a/sys/contrib/altq/altq/altq_rmclass.h b/sys/contrib/altq/altq/altq_rmclass.h
index cf0ddf4..782bc13 100644
--- a/sys/contrib/altq/altq/altq_rmclass.h
+++ b/sys/contrib/altq/altq/altq_rmclass.h
@@ -164,7 +164,12 @@ struct rm_class {
void (*overlimit)(struct rm_class *, struct rm_class *);
void (*drop)(struct rm_class *); /* Class drop action. */
- struct red *red_; /* RED state pointer */
+ union {
+ struct red *red_; /* RED state pointer */
+ struct codel *codel_; /* codel state pointer */
+ } cl_aqm_;
+#define red_ cl_aqm_.red_
+#define codel_ cl_aqm_.codel_
struct altq_pktattr *pktattr_; /* saved hdr used by RED/ECN */
int flags_;
@@ -233,6 +238,7 @@ struct rm_ifdat {
#define RMCF_RIO 0x0004
#define RMCF_FLOWVALVE 0x0008 /* use flowvalve (aka penalty-box) */
#define RMCF_CLEARDSCP 0x0010 /* clear diffserv codepoint */
+#define RMCF_CODEL 0x0020
/* flags for rmc_init */
#define RMCF_WRR 0x0100
diff --git a/sys/contrib/altq/altq/altq_subr.c b/sys/contrib/altq/altq/altq_subr.c
index 16b796a..bddf73f 100644
--- a/sys/contrib/altq/altq/altq_subr.c
+++ b/sys/contrib/altq/altq/altq_subr.c
@@ -537,6 +537,16 @@ altq_pfattach(struct pf_altq *a)
error = hfsc_pfattach(a);
break;
#endif
+#ifdef ALTQ_FAIRQ
+ case ALTQT_FAIRQ:
+ error = fairq_pfattach(a);
+ break;
+#endif
+#ifdef ALTQ_CODEL
+ case ALTQT_CODEL:
+ error = codel_pfattach(a);
+ break;
+#endif
default:
error = ENXIO;
}
@@ -612,6 +622,16 @@ altq_add(struct pf_altq *a)
error = hfsc_add_altq(a);
break;
#endif
+#ifdef ALTQ_FAIRQ
+ case ALTQT_FAIRQ:
+ error = fairq_add_altq(a);
+ break;
+#endif
+#ifdef ALTQ_CODEL
+ case ALTQT_CODEL:
+ error = codel_add_altq(a);
+ break;
+#endif
default:
error = ENXIO;
}
@@ -648,6 +668,16 @@ altq_remove(struct pf_altq *a)
error = hfsc_remove_altq(a);
break;
#endif
+#ifdef ALTQ_FAIRQ
+ case ALTQT_FAIRQ:
+ error = fairq_remove_altq(a);
+ break;
+#endif
+#ifdef ALTQ_CODEL
+ case ALTQT_CODEL:
+ error = codel_remove_altq(a);
+ break;
+#endif
default:
error = ENXIO;
}
@@ -681,6 +711,11 @@ altq_add_queue(struct pf_altq *a)
error = hfsc_add_queue(a);
break;
#endif
+#ifdef ALTQ_FAIRQ
+ case ALTQT_FAIRQ:
+ error = fairq_add_queue(a);
+ break;
+#endif
default:
error = ENXIO;
}
@@ -714,6 +749,11 @@ altq_remove_queue(struct pf_altq *a)
error = hfsc_remove_queue(a);
break;
#endif
+#ifdef ALTQ_FAIRQ
+ case ALTQT_FAIRQ:
+ error = fairq_remove_queue(a);
+ break;
+#endif
default:
error = ENXIO;
}
@@ -747,6 +787,16 @@ altq_getqstats(struct pf_altq *a, void *ubuf, int *nbytes)
error = hfsc_getqstats(a, ubuf, nbytes);
break;
#endif
+#ifdef ALTQ_FAIRQ
+ case ALTQT_FAIRQ:
+ error = fairq_getqstats(a, ubuf, nbytes);
+ break;
+#endif
+#ifdef ALTQ_CODEL
+ case ALTQT_CODEL:
+ error = codel_getqstats(a, ubuf, nbytes);
+ break;
+#endif
default:
error = ENXIO;
}
diff --git a/sys/contrib/altq/altq/altq_var.h b/sys/contrib/altq/altq/altq_var.h
index 956ee16..355c94c 100644
--- a/sys/contrib/altq/altq/altq_var.h
+++ b/sys/contrib/altq/altq/altq_var.h
@@ -243,6 +243,11 @@ int cbq_add_queue(struct pf_altq *);
int cbq_remove_queue(struct pf_altq *);
int cbq_getqstats(struct pf_altq *, void *, int *);
+int codel_pfattach(struct pf_altq *);
+int codel_add_altq(struct pf_altq *);
+int codel_remove_altq(struct pf_altq *);
+int codel_getqstats(struct pf_altq *, void *, int *);
+
int priq_pfattach(struct pf_altq *);
int priq_add_altq(struct pf_altq *);
int priq_remove_altq(struct pf_altq *);
@@ -257,5 +262,12 @@ int hfsc_add_queue(struct pf_altq *);
int hfsc_remove_queue(struct pf_altq *);
int hfsc_getqstats(struct pf_altq *, void *, int *);
+int fairq_pfattach(struct pf_altq *);
+int fairq_add_altq(struct pf_altq *);
+int fairq_remove_altq(struct pf_altq *);
+int fairq_add_queue(struct pf_altq *);
+int fairq_remove_queue(struct pf_altq *);
+int fairq_getqstats(struct pf_altq *, void *, int *);
+
#endif /* _KERNEL */
#endif /* _ALTQ_ALTQ_VAR_H_ */
diff --git a/sys/netpfil/pf/pf.c b/sys/netpfil/pf/pf.c
index 4563ee4..ebda220 100644
--- a/sys/netpfil/pf/pf.c
+++ b/sys/netpfil/pf/pf.c
@@ -614,6 +614,20 @@ pf_hashsrc(struct pf_addr *addr, sa_family_t af)
return (h & pf_srchashmask);
}
+#ifdef ALTQ
+static int
+pf_state_hash(struct pf_state *s)
+{
+ u_int32_t hv = (intptr_t)s / sizeof(*s);
+
+ hv ^= crc32(&s->src, sizeof(s->src));
+ hv ^= crc32(&s->dst, sizeof(s->dst));
+ if (hv == 0)
+ hv = 1;
+ return (hv);
+}
+#endif
+
#ifdef INET6
void
pf_addrcpy(struct pf_addr *dst, struct pf_addr *src, sa_family_t af)
@@ -6439,6 +6453,8 @@ done:
pd.act.qid = r->qid;
}
if (action == PF_PASS && pd.act.qid) {
+ if (s != NULL)
+ pd.pf_mtag->qid_hash = pf_state_hash(s);
if (pqid || (pd.tos & IPTOS_LOWDELAY))
pd.pf_mtag->qid = pd.act.pqid;
else
@@ -7006,6 +7022,8 @@ done:
pd.act.qid = r->qid;
}
if (action == PF_PASS && pd.act.qid) {
+ if (s != NULL)
+ pd.pf_mtag->qid_hash = pf_state_hash(s);
if (pd.tos & IPTOS_LOWDELAY)
pd.pf_mtag->qid = pd.act.pqid;
else
diff --git a/sys/netpfil/pf/pf_altq.h b/sys/netpfil/pf/pf_altq.h
index eda0965..3efd4ff 100644
--- a/sys/netpfil/pf/pf_altq.h
+++ b/sys/netpfil/pf/pf_altq.h
@@ -45,6 +45,12 @@ struct cbq_opts {
int flags;
};
+struct codel_opts {
+ u_int target;
+ u_int interval;
+ int ecn;
+};
+
struct priq_opts {
int flags;
};
@@ -65,6 +71,20 @@ struct hfsc_opts {
int flags;
};
+/*
+ * XXX this needs some work
+ */
+struct fairq_opts {
+ u_int nbuckets;
+ u_int hogs_m1;
+ int flags;
+
+ /* link sharing service curve */
+ u_int lssc_m1;
+ u_int lssc_d;
+ u_int lssc_m2;
+};
+
struct pf_altq {
char ifname[IFNAMSIZ];
@@ -89,8 +109,10 @@ struct pf_altq {
uint16_t flags; /* misc flags */
union {
struct cbq_opts cbq_opts;
+ struct codel_opts codel_opts;
struct priq_opts priq_opts;
struct hfsc_opts hfsc_opts;
+ struct fairq_opts fairq_opts;
} pq_u;
uint32_t qid; /* return value */
diff --git a/sys/netpfil/pf/pf_mtag.h b/sys/netpfil/pf/pf_mtag.h
index 3aacb2e..fd8554a 100644
--- a/sys/netpfil/pf/pf_mtag.h
+++ b/sys/netpfil/pf/pf_mtag.h
@@ -44,6 +44,7 @@
struct pf_mtag {
void *hdr; /* saved hdr pos in mbuf, for ECN */
u_int32_t qid; /* queue id */
+ u_int32_t qid_hash; /* queue hashid used by WFQ like algos */
u_int16_t tag; /* tag id */
u_int8_t flags;
u_int8_t routed;
OpenPOWER on IntegriCloud