netdev-linux: Use 64-bit rates in htb tc classes.

Currently, htb rates are capped at ~34Gbps because they are internally
expressed as 32-bit fields.

Move min and max rates to 64-bit fields and use TCA_HTB_RATE64 and
TCA_HTB_CEIL64 to configure HTC classes to break this barrier.

In order to test this, create a dummy tuntap device and set it's
speed to a very high value so we can try adding a QoS queue with big
rates.

Reported-at: https://bugzilla.redhat.com/show_bug.cgi?id=2137619
Acked-by: Eelco Chaudron <echaudro@redhat.com>
Signed-off-by: Adrian Moreno <amorenoz@redhat.com>
Signed-off-by: Ilya Maximets <i.maximets@ovn.org>

NEWS

@@ -35,6 +35,7 @@ Post-v3.1.0
        in order to create OVSDB sockets with access mode of 0770.
    - QoS:
      * Added new configuration option 'jitter' for a linux-netem QoS type.
+     * 'linux-htb' QoS type now supports rates higher than 34 Gbps.
    - DPDK:
      * ovs-vswitchd will keep the CAP_SYS_RAWIO capability when started
        with the --hw-rawio-access command line option.  This allows the

acinclude.m4

@@ -218,6 +218,16 @@ AC_DEFUN([OVS_CHECK_LINUX_TC], [
     ])],
     [AC_DEFINE([HAVE_TCA_STATS_PKT64], [1],
                [Define to 1 if TCA_STATS_PKT64 is available.])])
+
+  AC_COMPILE_IFELSE([
+    AC_LANG_PROGRAM([#include <linux/pkt_sched.h>], [
+        int x = TCA_HTB_RATE64;
+    ])],
+    [AC_SUBST(HAVE_TCA_HTB_RATE64,yes)
+     AC_DEFINE([HAVE_TCA_HTB_RATE64], [1],
+               [Define to 1 if TCA_HTB_RATE64 is available.])],
+    [AC_SUBST(HAVE_TCA_HTB_RATE64,no)]
+    )
 ])
 
 dnl OVS_CHECK_LINUX_SCTP_CT

lib/netdev-linux.c

@@ -4601,13 +4601,13 @@ static const struct tc_ops tc_ops_netem = {
 
 struct htb {
     struct tc tc;
-    unsigned int max_rate;      /* In bytes/s. */
+    uint64_t max_rate;          /* In bytes/s. */
 };
 
 struct htb_class {
     struct tc_queue tc_queue;
-    unsigned int min_rate;      /* In bytes/s. */
+    uint64_t min_rate;          /* In bytes/s. */
-    unsigned int max_rate;      /* In bytes/s. */
+    uint64_t max_rate;          /* In bytes/s. */
     unsigned int burst;         /* In bytes. */
     unsigned int priority;      /* Lower values are higher priorities. */
 };
@@ -4695,8 +4695,8 @@ htb_setup_class__(struct netdev *netdev, unsigned int handle,
     if ((class->min_rate / HTB_RATE2QUANTUM) < mtu) {
         opt.quantum = mtu;
     }
-    opt.buffer = tc_calc_buffer(opt.rate.rate, mtu, class->burst);
+    opt.buffer = tc_calc_buffer(class->min_rate, mtu, class->burst);
-    opt.cbuffer = tc_calc_buffer(opt.ceil.rate, mtu, class->burst);
+    opt.cbuffer = tc_calc_buffer(class->max_rate, mtu, class->burst);
     opt.prio = class->priority;
 
     tcmsg = netdev_linux_tc_make_request(netdev, RTM_NEWTCLASS, NLM_F_CREATE,
@@ -4709,15 +4709,26 @@ htb_setup_class__(struct netdev *netdev, unsigned int handle,
 
     nl_msg_put_string(&request, TCA_KIND, "htb");
     opt_offset = nl_msg_start_nested(&request, TCA_OPTIONS);
+
+#ifdef HAVE_TCA_HTB_RATE64
+    if (class->min_rate > UINT32_MAX) {
+        nl_msg_put_u64(&request, TCA_HTB_RATE64, class->min_rate);
+    }
+    if (class->max_rate > UINT32_MAX) {
+        nl_msg_put_u64(&request, TCA_HTB_CEIL64, class->max_rate);
+    }
+#endif
     nl_msg_put_unspec(&request, TCA_HTB_PARMS, &opt, sizeof opt);
-    tc_put_rtab(&request, TCA_HTB_RTAB, &opt.rate, 0);
+
-    tc_put_rtab(&request, TCA_HTB_CTAB, &opt.ceil, 0);
+    tc_put_rtab(&request, TCA_HTB_RTAB, &opt.rate, class->min_rate);
+    tc_put_rtab(&request, TCA_HTB_CTAB, &opt.ceil, class->max_rate);
     nl_msg_end_nested(&request, opt_offset);
 
     error = tc_transact(&request, NULL);
     if (error) {
         VLOG_WARN_RL(&rl, "failed to replace %s class %u:%u, parent %u:%u, "
-                     "min_rate=%u max_rate=%u burst=%u prio=%u (%s)",
+                     "min_rate=%"PRIu64" max_rate=%"PRIu64" burst=%u prio=%u "
+                     "(%s)",
                      netdev_get_name(netdev),
                      tc_get_major(handle), tc_get_minor(handle),
                      tc_get_major(parent), tc_get_minor(parent),
@@ -4737,6 +4748,10 @@ htb_parse_tca_options__(struct nlattr *nl_options, struct htb_class *class)
     static const struct nl_policy tca_htb_policy[] = {
         [TCA_HTB_PARMS] = { .type = NL_A_UNSPEC, .optional = false,
                             .min_len = sizeof(struct tc_htb_opt) },
+#ifdef HAVE_TCA_HTB_RATE64
+        [TCA_HTB_RATE64] = { .type = NL_A_U64, .optional = true },
+        [TCA_HTB_CEIL64] = { .type = NL_A_U64, .optional = true },
+#endif
     };
 
     struct nlattr *attrs[ARRAY_SIZE(tca_htb_policy)];
@@ -4751,7 +4766,15 @@ htb_parse_tca_options__(struct nlattr *nl_options, struct htb_class *class)
     htb = nl_attr_get(attrs[TCA_HTB_PARMS]);
     class->min_rate = htb->rate.rate;
     class->max_rate = htb->ceil.rate;
-    class->burst = tc_ticks_to_bytes(htb->rate.rate, htb->buffer);
+#ifdef HAVE_TCA_HTB_RATE64
+    if (attrs[TCA_HTB_RATE64]) {
+        class->min_rate = nl_attr_get_u64(attrs[TCA_HTB_RATE64]);
+    }
+    if (attrs[TCA_HTB_CEIL64]) {
+        class->max_rate = nl_attr_get_u64(attrs[TCA_HTB_CEIL64]);
+    }
+#endif
+    class->burst = tc_ticks_to_bytes(class->min_rate, htb->buffer);
     class->priority = htb->prio;
     return 0;
 }

tests/atlocal.in

@@ -7,6 +7,7 @@ HAVE_UNWIND='@HAVE_UNWIND@'
 EGREP='@EGREP@'
 PYTHON3='@PYTHON3@'
 CFLAGS='@CFLAGS@'
+HAVE_TCA_HTB_RATE64='@HAVE_TCA_HTB_RATE64@'
 
 # PYTHONCOERCECLOCALE=0 disables the Unicode compatibility warning on
 # stderr that breaks almost any Python3 test (PEP 0538)

tests/system-traffic.at

@@ -2354,6 +2354,34 @@ AT_CHECK([tc class show dev ovs-p1 | grep -q 'class htb .* HTB_CONF'])
 OVS_TRAFFIC_VSWITCHD_STOP
 AT_CLEANUP
 
+AT_SETUP([QoS - 64bit])
+AT_SKIP_IF([test $HAVE_TC = no])
+AT_SKIP_IF([test $HAVE_TCA_HTB_RATE64 = no])
+OVS_TRAFFIC_VSWITCHD_START()
+
+ADD_NAMESPACES(at_ns0, at_ns1)
+ADD_VETH(p0, at_ns0, br0, "10.1.1.1/24")
+ADD_VETH(p1, at_ns1, br0, "10.1.1.2/24")
+
+dnl Configure the QoS with rates that require 64bits, i.e: > 34Gbps.
+AT_CHECK([ovs-vsctl set port ovs-p0 qos=@qos -- set port ovs-p1 qos=@qos dnl
+            -- --id=@qos create qos dnl
+               type=linux-htb other-config:max-rate=50000000000 queues:0=@queue dnl
+            -- --id=@queue create queue dnl
+               other_config:min-rate=40000000000 other_config:max-rate=50000000000 dnl
+               other_config:burst=5000000],
+         [ignore], [ignore])
+
+OVS_WAIT_UNTIL([tc qdisc show dev ovs-p0 | grep -q htb])
+OVS_WAIT_UNTIL([tc qdisc show dev ovs-p1 | grep -q htb])
+
+m4_define([HTB_CONF], [rate 40Gbit ceil 50Gbit burst 620000b cburst 618750b])
+AT_CHECK([tc class show dev ovs-p0 | grep -q 'class htb .* HTB_CONF'])
+AT_CHECK([tc class show dev ovs-p1 | grep -q 'class htb .* HTB_CONF'])
+
+OVS_TRAFFIC_VSWITCHD_STOP
+AT_CLEANUP
+
 AT_BANNER([conntrack])
 
 AT_SETUP([conntrack - controller])