Merge pull request #14278 from ssahani/tc

network tc: qdisc parent add support to set ingress

10 files changed, 168 insertions, 24 deletions

diff --git a/man/systemd.network.xml b/man/systemd.network.xml
index 692a686020..acafa8a712 100644
--- a/man/systemd.network.xml
+++ b/man/systemd.network.xml
@@ -2327,8 +2327,8 @@
       <varlistentry>
         <term><varname>Parent=</varname></term>
         <listitem>
-          <para>Specifies the parent Queueing Discipline (qdisc). Takes one of <literal>root</literal>
-          or <literal>clsact</literal>. Defaults to <literal>root</literal>.</para>
+          <para>Specifies the parent Queueing Discipline (qdisc). Takes one of <literal>root</literal>,
+          <literal>clsact</literal> or <literal>ingress</literal>. Defaults to <literal>root</literal>.</para>
         </listitem>
       </varlistentry>
 
@@ -2442,6 +2442,14 @@
         </listitem>
       </varlistentry>
 
+      <varlistentry>
+        <term><varname>FairQueuingControlledDelayPacketLimit=</varname></term>
+        <listitem>
+          <para>Specifies the hard limit on the real queue size. When this limit is reached, incoming packets are
+          dropped. Defaults to unset and kernel's default is used.</para>
+        </listitem>
+      </varlistentry>
+
     </variablelist>
   </refsect1>
 
diff --git a/src/network/meson.build b/src/network/meson.build
index 8a09078315..41e482e7a1 100644
--- a/src/network/meson.build
+++ b/src/network/meson.build
@@ -107,6 +107,8 @@ sources = files('''
         networkd-util.h
         networkd-wifi.c
         networkd-wifi.h
+        tc/fq-codel.c
+        tc/fq-codel.h
         tc/netem.c
         tc/netem.h
         tc/qdisc.c
diff --git a/src/network/networkd-network-gperf.gperf b/src/network/networkd-network-gperf.gperf
index ea4ba31b80..e80f0afd8f 100644
--- a/src/network/networkd-network-gperf.gperf
+++ b/src/network/networkd-network-gperf.gperf
@@ -258,6 +258,7 @@ TrafficControlQueueingDiscipline.TokenBufferFilterMPUBytes,                  con
 TrafficControlQueueingDiscipline.TokenBufferFilterPeakRate,                  config_parse_tc_token_buffer_filter_size,                    0, 0
 TrafficControlQueueingDiscipline.TokenBufferFilterLatencySec,                config_parse_tc_token_buffer_filter_latency,                 0, 0
 TrafficControlQueueingDiscipline.StochasticFairnessQueueingPerturbPeriodSec, config_parse_tc_stochastic_fairness_queueing_perturb_period, 0, 0
+TrafficControlQueueingDiscipline.FairQueuingControlledDelayPacketLimit,      config_parse_tc_fair_queuing_controlled_delay_limit,         0, 0
 /* backwards compatibility: do not add new entries to this section */
 Network.IPv4LL,                         config_parse_ipv4ll,                             0,                             offsetof(Network, link_local)
 DHCP.ClientIdentifier,                  config_parse_dhcp_client_identifier,             0,                             offsetof(Network, dhcp_client_identifier)
diff --git a/src/network/tc/fq-codel.c b/src/network/tc/fq-codel.c
new file mode 100644
index 0000000000..174d70e1f7
--- /dev/null
+++ b/src/network/tc/fq-codel.c
@@ -0,0 +1,91 @@
+/* SPDX-License-Identifier: LGPL-2.1+
+ * Copyright © 2019 VMware, Inc. */
+
+#include <linux/pkt_sched.h>
+
+#include "alloc-util.h"
+#include "conf-parser.h"
+#include "netlink-util.h"
+#include "parse-util.h"
+#include "qdisc.h"
+#include "string-util.h"
+
+int fair_queuing_controlled_delay_new(FairQueuingControlledDelay **ret) {
+        FairQueuingControlledDelay *fqcd = NULL;
+
+        fqcd = new0(FairQueuingControlledDelay, 1);
+        if (!fqcd)
+                return -ENOMEM;
+
+        *ret = TAKE_PTR(fqcd);
+
+        return 0;
+}
+
+int fair_queuing_controlled_delay_fill_message(Link *link, const FairQueuingControlledDelay *fqcd, sd_netlink_message *req) {
+        int r;
+
+        assert(link);
+        assert(fqcd);
+        assert(req);
+
+        r = sd_netlink_message_open_array(req, TCA_OPTIONS);
+        if (r < 0)
+                return log_link_error_errno(link, r, "Could not open container TCA_OPTIONS: %m");
+
+        r = sd_netlink_message_append_data(req, TCA_FQ_CODEL_LIMIT, &fqcd->limit, sizeof(fqcd->limit));
+        if (r < 0)
+                return log_link_error_errno(link, r, "Could not append TCA_FQ_CODEL_LIMIT attribute: %m");
+
+        r = sd_netlink_message_close_container(req);
+        if (r < 0)
+                return log_link_error_errno(link, r, "Could not close container TCA_OPTIONS: %m");
+
+        return 0;
+}
+
+int config_parse_tc_fair_queuing_controlled_delay_limit(
+                const char *unit,
+                const char *filename,
+                unsigned line,
+                const char *section,
+                unsigned section_line,
+                const char *lvalue,
+                int ltype,
+                const char *rvalue,
+                void *data,
+                void *userdata) {
+
+        _cleanup_(qdisc_free_or_set_invalidp) QDisc *qdisc = NULL;
+        Network *network = data;
+        int r;
+
+        assert(filename);
+        assert(lvalue);
+        assert(rvalue);
+        assert(data);
+
+        r = qdisc_new_static(network, filename, section_line, &qdisc);
+        if (r < 0)
+                return r;
+
+        if (isempty(rvalue)) {
+                qdisc->fq_codel.limit = 0;
+
+                qdisc = NULL;
+                return 0;
+        }
+
+        r = safe_atou32(rvalue, &qdisc->fq_codel.limit);
+        if (r < 0) {
+                log_syntax(unit, LOG_ERR, filename, line, r,
+                           "Failed to parse '%s=', ignoring assignment: %s",
+                           lvalue, rvalue);
+                return 0;
+        }
+
+        qdisc->has_fair_queuing_controlled_delay = true;
+        qdisc = NULL;
+
+        return 0;
+}
diff --git a/src/network/tc/fq-codel.h b/src/network/tc/fq-codel.h
new file mode 100644
index 0000000000..0ddb5099a8
--- /dev/null
+++ b/src/network/tc/fq-codel.h
@@ -0,0 +1,17 @@
+/* SPDX-License-Identifier: LGPL-2.1+
+ * Copyright © 2019 VMware, Inc. */
+#pragma once
+
+#include "sd-netlink.h"
+
+#include "conf-parser.h"
+#include "networkd-link.h"
+
+typedef struct FairQueuingControlledDelay {
+        uint32_t limit;
+} FairQueuingControlledDelay;
+
+int fair_queuing_controlled_delay_new(FairQueuingControlledDelay **ret);
+int fair_queuing_controlled_delay_fill_message(Link *link, const FairQueuingControlledDelay *sfq, sd_netlink_message *req);
+
+CONFIG_PARSER_PROTOTYPE(config_parse_tc_fair_queuing_controlled_delay_limit);
diff --git a/src/network/tc/qdisc.c b/src/network/tc/qdisc.c
index 74b2b7a2c2..d066e0713f 100644
--- a/src/network/tc/qdisc.c
+++ b/src/network/tc/qdisc.c
@@ -84,6 +84,7 @@ void qdisc_free(QDisc *qdisc) {
 
         network_config_section_free(qdisc->section);
 
+        free(qdisc->tca_kind);
         free(qdisc);
 }
 
@@ -116,6 +117,7 @@ static int qdisc_handler(sd_netlink *rtnl, sd_netlink_message *m, Link *link) {
 int qdisc_configure(Link *link, QDisc *qdisc) {
         _cleanup_(sd_netlink_message_unrefp) sd_netlink_message *req = NULL;
         _cleanup_free_ char *tca_kind = NULL;
+        char *p;
         int r;
 
         assert(link);
@@ -131,12 +133,8 @@ int qdisc_configure(Link *link, QDisc *qdisc) {
         if (r < 0)
                 return log_link_error_errno(link, r, "Could not create tcm_parent message: %m");
 
-        if (qdisc->parent == TC_H_CLSACT) {
-                tca_kind = strdup("clsact");
-                if (!tca_kind)
-                        return log_oom();
-
-                r = sd_rtnl_message_set_qdisc_handle(req, TC_H_MAKE(TC_H_CLSACT, 0));
+        if (qdisc->handle != TC_H_UNSPEC) {
+                r = sd_rtnl_message_set_qdisc_handle(req, qdisc->handle);
                 if (r < 0)
                         return log_link_error_errno(link, r, "Could not set tcm_handle message: %m");
         }
@@ -171,8 +169,19 @@ int qdisc_configure(Link *link, QDisc *qdisc) {
                         return r;
         }
 
-        if (tca_kind) {
-                r = sd_netlink_message_append_string(req, TCA_KIND, tca_kind);
+        if (qdisc->has_fair_queuing_controlled_delay) {
+                r = free_and_strdup(&tca_kind, "fq_codel");
+                if (r < 0)
+                        return log_oom();
+
+                r = fair_queuing_controlled_delay_fill_message(link, &qdisc->fq_codel, req);
+                if (r < 0)
+                        return r;
+        }
+
+        p = tca_kind ?:qdisc->tca_kind;
+        if (p) {
+                r = sd_netlink_message_append_string(req, TCA_KIND, p);
                 if (r < 0)
                         return log_link_error_errno(link, r, "Could not append TCA_KIND attribute: %m");
         }
@@ -218,10 +227,10 @@ int qdisc_section_verify(QDisc *qdisc, bool *has_root, bool *has_clsact) {
                                                  "Ignoring [TrafficControlQueueingDiscipline] section from line %u.",
                                                  qdisc->section->filename, qdisc->section->line);
                 *has_root = true;
-        } else if (qdisc->parent == TC_H_CLSACT) {
+        } else if (qdisc->parent == TC_H_CLSACT) { /* TC_H_CLSACT == TC_H_INGRESS */
                 if (*has_clsact)
                         return log_warning_errno(SYNTHETIC_ERRNO(EINVAL),
-                                                 "%s: More than one clsact TrafficControlQueueingDiscipline sections are defined. "
+                                                 "%s: More than one clsact or ingress TrafficControlQueueingDiscipline sections are defined. "
                                                  "Ignoring [TrafficControlQueueingDiscipline] section from line %u.",
                                                  qdisc->section->filename, qdisc->section->line);
                 *has_clsact = true;
@@ -255,17 +264,30 @@ int config_parse_tc_qdiscs_parent(
         if (r < 0)
                 return r;
 
-        if (streq(rvalue, "root"))
+        if (streq(rvalue, "root")) {
                 qdisc->parent = TC_H_ROOT;
-        else if (streq(rvalue, "clsact"))
+                qdisc->handle = TC_H_UNSPEC;
+        } else if (streq(rvalue, "clsact")) {
                 qdisc->parent = TC_H_CLSACT;
-        else {
+                qdisc->handle = TC_H_MAKE(TC_H_CLSACT, 0);
+        } else if (streq(rvalue, "ingress")) {
+                qdisc->parent = TC_H_INGRESS;
+                qdisc->handle = TC_H_MAKE(TC_H_INGRESS, 0);
+        } else {
                 log_syntax(unit, LOG_ERR, filename, line, r,
-                           "Failed to parse [QueueDiscs] 'Parent=', ignoring assignment: %s",
+                           "Failed to parse 'Parent=', ignoring assignment: %s",
                            rvalue);
                 return 0;
         }
 
+        if (streq(rvalue, "root"))
+                qdisc->tca_kind = mfree(qdisc->tca_kind);
+        else {
+                r = free_and_strdup(&qdisc->tca_kind, rvalue);
+                if (r < 0)
+                        return log_oom();
+        }
+
         qdisc = NULL;
 
         return 0;
diff --git a/src/network/tc/qdisc.h b/src/network/tc/qdisc.h
index 1d06dc53f4..53968bcf1f 100644
--- a/src/network/tc/qdisc.h
+++ b/src/network/tc/qdisc.h
@@ -3,6 +3,7 @@
 #pragma once
 
 #include "conf-parser.h"
+#include "fq-codel.h"
 #include "netem.h"
 #include "networkd-link.h"
 #include "networkd-network.h"
@@ -21,13 +22,16 @@ typedef struct QDisc {
         uint32_t handle;
         uint32_t parent;
 
+        char *tca_kind;
         bool has_network_emulator:1;
         bool has_token_buffer_filter:1;
         bool has_stochastic_fairness_queueing:1;
+        bool has_fair_queuing_controlled_delay:1;
 
         NetworkEmulator ne;
         TokenBufferFilter tbf;
         StochasticFairnessQueueing sfq;
+        FairQueuingControlledDelay fq_codel;
 } QDisc;
 
 void qdisc_free(QDisc *qdisc);
diff --git a/test/fuzz/fuzz-network-parser/directives.network b/test/fuzz/fuzz-network-parser/directives.network
index c3264522b4..c71fe5ce77 100644
--- a/test/fuzz/fuzz-network-parser/directives.network
+++ b/test/fuzz/fuzz-network-parser/directives.network
@@ -278,3 +278,4 @@ TokenBufferFilterMPUBytes=
 TokenBufferFilterPeakRate=
 TokenBufferFilterLatencySec=
 StochasticFairnessQueueingPerturbPeriodSec=
+FairQueuingControlledDelayPacketLimit=
diff --git a/test/test-network/conf/25-qdisc-netem.network b/test/test-network/conf/25-qdisc-netem-and-fqcodel.network
index de8f7243ce..b7defa3313 100644
--- a/test/test-network/conf/25-qdisc-netem.network
+++ b/test/test-network/conf/25-qdisc-netem-and-fqcodel.network
@@ -13,8 +13,5 @@ NetworkEmulatorLossRate=20%
 NetworkEmulatorPacketLimit=100
 
 [TrafficControlQueueingDiscipline]
-Parent=clsact
-NetworkEmulatorDelaySec=100ms
-NetworkEmulatorDelayJitterSec=13ms
-NetworkEmulatorLossRate=20.5%
-NetworkEmulatorPacketLimit=200
+Parent=ingress
+FairQueuingControlledDelayPacketLimit=20480
diff --git a/test/test-network/systemd-networkd-tests.py b/test/test-network/systemd-networkd-tests.py
index 9d22d788dd..a72b9abf76 100755
--- a/test/test-network/systemd-networkd-tests.py
+++ b/test/test-network/systemd-networkd-tests.py
@@ -1509,7 +1509,7 @@ class NetworkdNetworkTests(unittest.TestCase, Utilities):
         '25-neighbor-ip-dummy.network',
         '25-neighbor-ip.network',
         '25-nexthop.network',
-        '25-qdisc-netem.network',
+        '25-qdisc-netem-and-fqcodel.network',
         '25-qdisc-tbf-and-sfq.network',
         '25-route-ipv6-src.network',
         '25-route-static.network',
@@ -2083,7 +2083,7 @@ class NetworkdNetworkTests(unittest.TestCase, Utilities):
         self.assertRegex(output, '192.168.5.1')
 
     def test_qdisc(self):
-        copy_unit_to_networkd_unit_path('25-qdisc-netem.network', '12-dummy.netdev',
+        copy_unit_to_networkd_unit_path('25-qdisc-netem-and-fqcodel.network', '12-dummy.netdev',
                                         '25-qdisc-tbf-and-sfq.network', '11-dummy.netdev')
         start_networkd()
 
@@ -2093,7 +2093,8 @@ class NetworkdNetworkTests(unittest.TestCase, Utilities):
         print(output)
         self.assertRegex(output, 'qdisc netem')
         self.assertRegex(output, 'limit 100 delay 50.0ms  10.0ms loss 20%')
-        self.assertRegex(output, 'limit 200 delay 100.0ms  13.0ms loss 20.5%')
+        self.assertRegex(output, 'qdisc fq_codel')
+        self.assertRegex(output, 'limit 20480p')
         output = check_output('tc qdisc show dev test1')
         print(output)
         self.assertRegex(output, 'qdisc tbf')