tc/htb: Hierarchical QoS hardware offload

This commit adds support for configuring HTB in offload mode. HTB
offload eliminates the single qdisc lock in the datapath and offloads
the algorithm to the NIC. The new 'offload' parameter is added to
enable this mode:

    # tc qdisc replace dev eth0 root handle 1: htb offload

Classes are created as usual, but filters should be moved to clsact for
lock-free classification (filters attached to HTB itself are not
supported in the offload mode):

    # tc filter add dev eth0 egress protocol ip flower dst_port 80
    action skbedit priority 1:10

tc qdisc show and tc class show will indicate whether the offload is
enabled. Example output:

$ tc qdisc show dev eth1
qdisc htb 1: root offloaded r2q 10 default 0 direct_packets_stat 0 direct_qlen 1000 offload
qdisc pfifo 0: parent 1: limit 1000p
qdisc pfifo 0: parent 1: limit 1000p
qdisc pfifo 0: parent 1: limit 1000p
qdisc pfifo 0: parent 1: limit 1000p
qdisc pfifo 0: parent 1: limit 1000p
qdisc pfifo 0: parent 1: limit 1000p
qdisc pfifo 0: parent 1: limit 1000p
qdisc pfifo 0: parent 1: limit 1000p
$ tc class show dev eth1
class htb 1:101 parent 1:1 prio 0 rate 4Gbit ceil 4Gbit burst 1000b cburst 1000b  offload
class htb 1:1 root rate 100Gbit ceil 100Gbit burst 0b cburst 0b  offload
class htb 1:103 parent 1:1 prio 0 rate 4Gbit ceil 4Gbit burst 1000b cburst 1000b  offload
class htb 1:102 parent 1:1 prio 0 rate 4Gbit ceil 4Gbit burst 1000b cburst 1000b  offload
class htb 1:105 parent 1:1 prio 0 rate 4Gbit ceil 4Gbit burst 1000b cburst 1000b  offload
class htb 1:104 parent 1:1 prio 0 rate 4Gbit ceil 4Gbit burst 1000b cburst 1000b  offload
class htb 1:107 parent 1:1 prio 0 rate 4Gbit ceil 4Gbit burst 1000b cburst 1000b  offload
class htb 1:106 parent 1:1 prio 0 rate 4Gbit ceil 4Gbit burst 1000b cburst 1000b  offload
class htb 1:108 parent 1:1 prio 0 rate 4Gbit ceil 4Gbit burst 1000b cburst 1000b  offload
$ tc -j qdisc show dev eth1
[{"kind":"htb","handle":"1:","root":true,"offloaded":true,"options":{"r2q":10,"default":"0","direct_packets_stat":0,"direct_qlen":1000,"offload":null}},{"kind":"pfifo","handle":"0:","parent":"1:","options":{"limit":1000}},{"kind":"pfifo","handle":"0:","parent":"1:","options":{"limit":1000}},{"kind":"pfifo","handle":"0:","parent":"1:","options":{"limit":1000}},{"kind":"pfifo","handle":"0:","parent":"1:","options":{"limit":1000}},{"kind":"pfifo","handle":"0:","parent":"1:","options":{"limit":1000}},{"kind":"pfifo","handle":"0:","parent":"1:","options":{"limit":1000}},{"kind":"pfifo","handle":"0:","parent":"1:","options":{"limit":1000}},{"kind":"pfifo","handle":"0:","parent":"1:","options":{"limit":1000}}]

Signed-off-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: David Ahern <dsahern@kernel.org>

diff --git a/man/man8/tc-htb.8 b/man/man8/tc-htb.8
index a41623427b807856ed43e6b42136f55b7fbe1bf3..031b73ac7794a299c1d54404b6e4245900cde4c1 100644 (file)
--- a/man/man8/tc-htb.8
+++ b/man/man8/tc-htb.8
@@ -12,7 +12,7 @@ major:
 minor-id
 .B ] [ r2q
 divisor
-.B ]
+.B ] [ offload ]
 
 .B tc class ... dev
 dev
@@ -104,6 +104,9 @@ Divisor used to calculate
 values for classes.  Classes divide
 .B rate
 by this number.  Default value is 10.
+.TP
+offload
+Offload the HTB algorithm to hardware (requires driver and device support).
 
 .SH CLASSES
 Classes have a host of parameters to configure their operation.

diff --git a/tc/q_htb.c b/tc/q_htb.c
index c609e9749c2c7778f67c2d7d21f947561ca42c2c..425663557430c838c821589fdd9385af1f6d1f1b 100644 (file)
--- a/tc/q_htb.c
+++ b/tc/q_htb.c
@@ -30,11 +30,12 @@
 static void explain(void)
 {
        fprintf(stderr, "Usage: ... qdisc add ... htb [default N] [r2q N]\n"
-               "                      [direct_qlen P]\n"
+               "                      [direct_qlen P] [offload]\n"
                " default  minor id of class to which unclassified packets are sent {0}\n"
                " r2q      DRR quantums are computed as rate in Bps/r2q {10}\n"
                " debug    string of 16 numbers each 0-3 {0}\n\n"
                " direct_qlen  Limit of the direct queue {in packets}\n"
+               " offload  enable hardware offload\n"
                "... class add ... htb rate R1 [burst B1] [mpu B] [overhead O]\n"
                "                      [prio P] [slot S] [pslot PS]\n"
                "                      [ceil R2] [cburst B2] [mtu MTU] [quantum Q]\n"
@@ -68,6 +69,7 @@ static int htb_parse_opt(struct qdisc_util *qu, int argc,
        };
        struct rtattr *tail;
        unsigned int i; char *p;
+       bool offload = false;
 
        while (argc > 0) {
                if (matches(*argv, "r2q") == 0) {
@@ -91,6 +93,8 @@ static int htb_parse_opt(struct qdisc_util *qu, int argc,
                        if (get_u32(&direct_qlen, *argv, 10)) {
                                explain1("direct_qlen"); return -1;
                        }
+               } else if (matches(*argv, "offload") == 0) {
+                       offload = true;
                } else {
                        fprintf(stderr, "What is \"%s\"?\n", *argv);
                        explain();
@@ -103,6 +107,8 @@ static int htb_parse_opt(struct qdisc_util *qu, int argc,
        if (direct_qlen != ~0U)
                addattr_l(n, 2024, TCA_HTB_DIRECT_QLEN,
                          &direct_qlen, sizeof(direct_qlen));
+       if (offload)
+               addattr(n, 2024, TCA_HTB_OFFLOAD);
        addattr_nest_end(n, tail);
        return 0;
 }
@@ -344,6 +350,8 @@ static int htb_print_opt(struct qdisc_util *qu, FILE *f, struct rtattr *opt)
                print_uint(PRINT_ANY, "direct_qlen", " direct_qlen %u",
                           direct_qlen);
        }
+       if (tb[TCA_HTB_OFFLOAD])
+               print_null(PRINT_ANY, "offload", " offload", NULL);
        return 0;
 }