Merge pull request #14320 from yuwata/network-tc-fq_codel-more

network: tc: introduce more FQ-CoDel settings

man/systemd.network.xml

@@ -2450,6 +2450,64 @@
         </listitem>
       </varlistentry>
 
+      <varlistentry>
+        <term><varname>FairQueuingControlledDelayMemoryLimit=</varname></term>
+        <listitem>
+          <para>Specifies the limit on the total number of bytes that can be queued in this FQ-CoDel instance.
+          When suffixed with K, M, or G, the specified size is parsed as Kilobytes, Megabytes, or Gigabytes,
+          respectively, to the base of 1024. Defaults to unset and kernel's default is used.</para>
+        </listitem>
+      </varlistentry>
+
+      <varlistentry>
+        <term><varname>FairQueuingControlledDelayFlows=</varname></term>
+        <listitem>
+          <para>Specifies the number of flows into which the incoming packets are classified.
+          Defaults to unset and kernel's default is used.</para>
+        </listitem>
+      </varlistentry>
+
+      <varlistentry>
+        <term><varname>FairQueuingControlledDelayTargetSec=</varname></term>
+        <listitem>
+          <para>Takes a timespan. Specifies the acceptable minimum standing/persistent queue delay.
+          Defaults to unset and kernel's default is used.</para>
+        </listitem>
+      </varlistentry>
+
+      <varlistentry>
+        <term><varname>FairQueuingControlledDelayIntervalSec=</varname></term>
+        <listitem>
+          <para>Takes a timespan. This is used to ensure that the measured minimum delay does not
+          become too stale. Defaults to unset and kernel's default is used.</para>
+        </listitem>
+      </varlistentry>
+
+      <varlistentry>
+        <term><varname>FairQueuingControlledDelayQuantum=</varname></term>
+        <listitem>
+          <para>Specifies the number of bytes used as 'deficit' in the fair queuing algorithmtimespan.
+          When suffixed with K, M, or G, the specified size is parsed as Kilobytes, Megabytes, or Gigabytes,
+          respectively, to the base of 1024. Defaults to unset and kernel's default is used.</para>
+        </listitem>
+      </varlistentry>
+
+      <varlistentry>
+        <term><varname>FairQueuingControlledDelayECN=</varname></term>
+        <listitem>
+          <para>Takes a boolean. This can be used to mark packets instead of dropping them. Defaults to
+          unset and kernel's default is used.</para>
+        </listitem>
+      </varlistentry>
+
+      <varlistentry>
+        <term><varname>FairQueuingControlledDelayCEThresholdSec=</varname></term>
+        <listitem>
+          <para>Takes a timespan. This sets a threshold above which all packets are marked with ECN
+          Congestion Experienced (CE). Defaults to unset and kernel's default is used.</para>
+        </listitem>
+      </varlistentry>
+
       <varlistentry>
         <term><varname>FairQueueTrafficPolicingPacketLimit=</varname></term>
         <listitem>

src/network/networkd-network-gperf.gperf

@@ -258,7 +258,14 @@ 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
+TrafficControlQueueingDiscipline.FairQueuingControlledDelayPacketLimit,      config_parse_tc_fair_queuing_controlled_delay_u32,           0, 0
+TrafficControlQueueingDiscipline.FairQueuingControlledDelayMemoryLimit,      config_parse_tc_fair_queuing_controlled_delay_size,          0, 0
+TrafficControlQueueingDiscipline.FairQueuingControlledDelayFlows,            config_parse_tc_fair_queuing_controlled_delay_u32,           0, 0
+TrafficControlQueueingDiscipline.FairQueuingControlledDelayQuantum,          config_parse_tc_fair_queuing_controlled_delay_size,          0, 0
+TrafficControlQueueingDiscipline.FairQueuingControlledDelayTargetSec,        config_parse_tc_fair_queuing_controlled_delay_usec,          0, 0
+TrafficControlQueueingDiscipline.FairQueuingControlledDelayIntervalSec,      config_parse_tc_fair_queuing_controlled_delay_usec,          0, 0
+TrafficControlQueueingDiscipline.FairQueuingControlledDelayCEThresholdSec,   config_parse_tc_fair_queuing_controlled_delay_usec,          0, 0
+TrafficControlQueueingDiscipline.FairQueuingControlledDelayECN,              config_parse_tc_fair_queuing_controlled_delay_bool,          0, 0
 TrafficControlQueueingDiscipline.FairQueueTrafficPolicingPacketLimit,        config_parse_tc_fair_queue_traffic_policing_packet_limit,    0, 0
 /* backwards compatibility: do not add new entries to this section */
 Network.IPv4LL,                         config_parse_ipv4ll,                             0,                             offsetof(Network, link_local)

src/network/tc/fq-codel.c

@@ -10,6 +10,20 @@
 #include "qdisc.h"
 #include "string-util.h"
 
+static int fair_queuing_controlled_delay_init(QDisc *qdisc) {
+        FairQueuingControlledDelay *fqcd;
+
+        assert(qdisc);
+
+        fqcd = FQ_CODEL(qdisc);
+
+        fqcd->memory_limit = UINT32_MAX;
+        fqcd->ce_threshold_usec = USEC_INFINITY;
+        fqcd->ecn = -1;
+
+        return 0;
+}
+
 static int fair_queuing_controlled_delay_fill_message(Link *link, QDisc *qdisc, sd_netlink_message *req) {
         FairQueuingControlledDelay *fqcd;
         int r;
@@ -24,9 +38,53 @@ static int fair_queuing_controlled_delay_fill_message(Link *link, QDisc *qdisc,
         if (r < 0)
                 return log_link_error_errno(link, r, "Could not open container TCA_OPTIONS: %m");
 
-        r = sd_netlink_message_append_u32(req, TCA_FQ_CODEL_LIMIT, fqcd->limit);
-        if (r < 0)
-                return log_link_error_errno(link, r, "Could not append TCA_FQ_CODEL_LIMIT attribute: %m");
+        if (fqcd->packet_limit > 0) {
+                r = sd_netlink_message_append_u32(req, TCA_FQ_CODEL_LIMIT, fqcd->packet_limit);
+                if (r < 0)
+                        return log_link_error_errno(link, r, "Could not append TCA_FQ_CODEL_LIMIT attribute: %m");
+        }
+
+        if (fqcd->flows > 0) {
+                r = sd_netlink_message_append_u32(req, TCA_FQ_CODEL_FLOWS, fqcd->flows);
+                if (r < 0)
+                        return log_link_error_errno(link, r, "Could not append TCA_FQ_CODEL_FLOWS attribute: %m");
+        }
+
+        if (fqcd->quantum > 0) {
+                r = sd_netlink_message_append_u32(req, TCA_FQ_CODEL_QUANTUM, fqcd->quantum);
+                if (r < 0)
+                        return log_link_error_errno(link, r, "Could not append TCA_FQ_CODEL_QUANTUM attribute: %m");
+        }
+
+        if (fqcd->interval_usec > 0) {
+                r = sd_netlink_message_append_u32(req, TCA_FQ_CODEL_INTERVAL, fqcd->interval_usec);
+                if (r < 0)
+                        return log_link_error_errno(link, r, "Could not append TCA_FQ_CODEL_INTERVAL attribute: %m");
+        }
+
+        if (fqcd->target_usec > 0) {
+                r = sd_netlink_message_append_u32(req, TCA_FQ_CODEL_TARGET, fqcd->target_usec);
+                if (r < 0)
+                        return log_link_error_errno(link, r, "Could not append TCA_FQ_CODEL_TARGET attribute: %m");
+        }
+
+        if (fqcd->ecn >= 0) {
+                r = sd_netlink_message_append_u32(req, TCA_FQ_CODEL_ECN, fqcd->ecn);
+                if (r < 0)
+                        return log_link_error_errno(link, r, "Could not append TCA_FQ_CODEL_ECN attribute: %m");
+        }
+
+        if (fqcd->ce_threshold_usec != USEC_INFINITY) {
+                r = sd_netlink_message_append_u32(req, TCA_FQ_CODEL_CE_THRESHOLD, fqcd->ce_threshold_usec);
+                if (r < 0)
+                        return log_link_error_errno(link, r, "Could not append TCA_FQ_CODEL_CE_THRESHOLD attribute: %m");
+        }
+
+        if (fqcd->memory_limit != UINT32_MAX) {
+                r = sd_netlink_message_append_u32(req, TCA_FQ_CODEL_MEMORY_LIMIT, fqcd->memory_limit);
+                if (r < 0)
+                        return log_link_error_errno(link, r, "Could not append TCA_FQ_CODEL_MEMORY_LIMIT attribute: %m");
+        }
 
         r = sd_netlink_message_close_container(req);
         if (r < 0)
@@ -35,7 +93,130 @@ static int fair_queuing_controlled_delay_fill_message(Link *link, QDisc *qdisc,
         return 0;
 }
 
-int config_parse_tc_fair_queuing_controlled_delay_limit(
+int config_parse_tc_fair_queuing_controlled_delay_u32(
+                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;
+        FairQueuingControlledDelay *fqcd;
+        Network *network = data;
+        uint32_t *p;
+        int r;
+
+        assert(filename);
+        assert(lvalue);
+        assert(rvalue);
+        assert(data);
+
+        r = qdisc_new_static(QDISC_KIND_FQ_CODEL, network, filename, section_line, &qdisc);
+        if (r == -ENOMEM)
+                return log_oom();
+        if (r < 0)
+                return log_syntax(unit, LOG_ERR, filename, line, r,
+                                  "More than one kind of queueing discipline, ignoring assignment: %m");
+
+        fqcd = FQ_CODEL(qdisc);
+
+        if (streq(lvalue, "FairQueuingControlledDelayPacketLimit"))
+                p = &fqcd->packet_limit;
+        else if (streq(lvalue, "FairQueuingControlledDelayFlows"))
+                p = &fqcd->flows;
+        else
+                assert_not_reached("Invalid lvalue.");
+
+        if (isempty(rvalue)) {
+                *p = 0;
+
+                qdisc = NULL;
+                return 0;
+        }
+
+        r = safe_atou32(rvalue, p);
+        if (r < 0) {
+                log_syntax(unit, LOG_ERR, filename, line, r,
+                           "Failed to parse '%s=', ignoring assignment: %s",
+                           lvalue, rvalue);
+                return 0;
+        }
+
+        qdisc = NULL;
+
+        return 0;
+}
+
+int config_parse_tc_fair_queuing_controlled_delay_usec(
+                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;
+        FairQueuingControlledDelay *fqcd;
+        Network *network = data;
+        usec_t *p;
+        int r;
+
+        assert(filename);
+        assert(lvalue);
+        assert(rvalue);
+        assert(data);
+
+        r = qdisc_new_static(QDISC_KIND_FQ_CODEL, network, filename, section_line, &qdisc);
+        if (r == -ENOMEM)
+                return log_oom();
+        if (r < 0)
+                return log_syntax(unit, LOG_ERR, filename, line, r,
+                                  "More than one kind of queueing discipline, ignoring assignment: %m");
+
+        fqcd = FQ_CODEL(qdisc);
+
+        if (streq(lvalue, "FairQueuingControlledDelayTargetSec"))
+                p = &fqcd->target_usec;
+        else if (streq(lvalue, "FairQueuingControlledDelayIntervalSec"))
+                p = &fqcd->interval_usec;
+        else if (streq(lvalue, "FairQueuingControlledDelayCEThresholdSec"))
+                p = &fqcd->ce_threshold_usec;
+        else
+                assert_not_reached("Invalid lvalue.");
+
+        if (isempty(rvalue)) {
+                if (streq(lvalue, "FairQueuingControlledDelayCEThresholdSec"))
+                        *p = USEC_INFINITY;
+                else
+                        *p = 0;
+
+                qdisc = NULL;
+                return 0;
+        }
+
+        r = parse_sec(rvalue, p);
+        if (r < 0) {
+                log_syntax(unit, LOG_ERR, filename, line, r,
+                           "Failed to parse '%s=', ignoring assignment: %s",
+                           lvalue, rvalue);
+                return 0;
+        }
+
+        qdisc = NULL;
+
+        return 0;
+}
+
+int config_parse_tc_fair_queuing_controlled_delay_bool(
                 const char *unit,
                 const char *filename,
                 unsigned line,
@@ -67,13 +248,13 @@ int config_parse_tc_fair_queuing_controlled_delay_limit(
         fqcd = FQ_CODEL(qdisc);
 
         if (isempty(rvalue)) {
-                fqcd->limit = 0;
+                fqcd->ecn = -1;
 
                 qdisc = NULL;
                 return 0;
         }
 
-        r = safe_atou32(rvalue, &fqcd->limit);
+        r = parse_boolean(rvalue);
         if (r < 0) {
                 log_syntax(unit, LOG_ERR, filename, line, r,
                            "Failed to parse '%s=', ignoring assignment: %s",
@@ -81,6 +262,77 @@ int config_parse_tc_fair_queuing_controlled_delay_limit(
                 return 0;
         }
 
+        fqcd->ecn = r;
+        qdisc = NULL;
+
+        return 0;
+}
+
+int config_parse_tc_fair_queuing_controlled_delay_size(
+                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;
+        FairQueuingControlledDelay *fqcd;
+        Network *network = data;
+        uint64_t sz;
+        uint32_t *p;
+        int r;
+
+        assert(filename);
+        assert(lvalue);
+        assert(rvalue);
+        assert(data);
+
+        r = qdisc_new_static(QDISC_KIND_FQ_CODEL, network, filename, section_line, &qdisc);
+        if (r == -ENOMEM)
+                return log_oom();
+        if (r < 0)
+                return log_syntax(unit, LOG_ERR, filename, line, r,
+                                  "More than one kind of queueing discipline, ignoring assignment: %m");
+
+        fqcd = FQ_CODEL(qdisc);
+
+        if (streq(lvalue, "FairQueuingControlledDelayMemoryLimit"))
+                p = &fqcd->memory_limit;
+        else if (streq(lvalue, "FairQueuingControlledDelayQuantum"))
+                p = &fqcd->quantum;
+        else
+                assert_not_reached("Invalid lvalue.");
+
+        if (isempty(rvalue)) {
+                if (streq(lvalue, "FairQueuingControlledMemoryLimit"))
+                        *p = UINT32_MAX;
+                else
+                        *p = 0;
+
+                qdisc = NULL;
+                return 0;
+        }
+
+        r = parse_size(rvalue, 1024, &sz);
+        if (r < 0) {
+                log_syntax(unit, LOG_ERR, filename, line, r,
+                           "Failed to parse '%s=', ignoring assignment: %s",
+                           lvalue, rvalue);
+                return 0;
+        }
+        if (sz >= UINT32_MAX) {
+                log_syntax(unit, LOG_ERR, filename, line, r,
+                           "Specified '%s=' is too large, ignoring assignment: %s",
+                           lvalue, rvalue);
+                return 0;
+        }
+
+        *p = sz;
         qdisc = NULL;
 
         return 0;
@@ -89,5 +341,6 @@ int config_parse_tc_fair_queuing_controlled_delay_limit(
 const QDiscVTable fq_codel_vtable = {
         .object_size = sizeof(FairQueuingControlledDelay),
         .tca_kind = "fq_codel",
+        .init = fair_queuing_controlled_delay_init,
         .fill_message = fair_queuing_controlled_delay_fill_message,
 };

src/network/tc/fq-codel.h

@@ -4,13 +4,25 @@
 
 #include "conf-parser.h"
 #include "qdisc.h"
+#include "time-util.h"
 
 typedef struct FairQueuingControlledDelay {
         QDisc meta;
-        uint32_t limit;
+
+        uint32_t packet_limit;
+        uint32_t flows;
+        uint32_t quantum;
+        uint32_t memory_limit;
+        usec_t target_usec;
+        usec_t interval_usec;
+        usec_t ce_threshold_usec;
+        int ecn;
 } FairQueuingControlledDelay;
 
 DEFINE_QDISC_CAST(FQ_CODEL, FairQueuingControlledDelay);
 extern const QDiscVTable fq_codel_vtable;
 
-CONFIG_PARSER_PROTOTYPE(config_parse_tc_fair_queuing_controlled_delay_limit);
+CONFIG_PARSER_PROTOTYPE(config_parse_tc_fair_queuing_controlled_delay_u32);
+CONFIG_PARSER_PROTOTYPE(config_parse_tc_fair_queuing_controlled_delay_usec);
+CONFIG_PARSER_PROTOTYPE(config_parse_tc_fair_queuing_controlled_delay_bool);
+CONFIG_PARSER_PROTOTYPE(config_parse_tc_fair_queuing_controlled_delay_size);

src/network/tc/qdisc.c

@@ -23,6 +23,7 @@ const QDiscVTable * const qdisc_vtable[_QDISC_KIND_MAX] = {
 
 static int qdisc_new(QDiscKind kind, QDisc **ret) {
         QDisc *qdisc;
+        int r;
 
         if (kind == _QDISC_KIND_INVALID) {
                 qdisc = new(QDisc, 1);
@@ -42,6 +43,12 @@ static int qdisc_new(QDiscKind kind, QDisc **ret) {
                 qdisc->family = AF_UNSPEC;
                 qdisc->parent = TC_H_ROOT;
                 qdisc->kind = kind;
+
+                if (QDISC_VTABLE(qdisc)->init) {
+                        r = QDISC_VTABLE(qdisc)->init(qdisc);
+                        if (r < 0)
+                                return r;
+                }
         }
 
         *ret = TAKE_PTR(qdisc);

src/network/tc/qdisc.h

@@ -32,6 +32,8 @@ typedef struct QDisc {
 typedef struct QDiscVTable {
         size_t object_size;
         const char *tca_kind;
+        /* called in qdisc_new() */
+        int (*init)(QDisc *qdisc);
         int (*fill_message)(Link *link, QDisc *qdisc, sd_netlink_message *m);
         int (*verify)(QDisc *qdisc);
 } QDiscVTable;

src/nspawn/nspawn-mount.c

@@ -222,8 +222,6 @@ int bind_mount_parse(CustomMount **l, size_t *n, const char *s, bool read_only)
 
         if (!path_is_absolute(destination))
                 return -EINVAL;
-        if (empty_or_root(destination))
-                return -EINVAL;
 
         m = custom_mount_add(l, n, CUSTOM_MOUNT_BIND);
         if (!m)
@@ -262,8 +260,6 @@ int tmpfs_mount_parse(CustomMount **l, size_t *n, const char *s) {
 
         if (!path_is_absolute(path))
                 return -EINVAL;
-        if (empty_or_root(path))
-                return -EINVAL;
 
         m = custom_mount_add(l, n, CUSTOM_MOUNT_TMPFS);
         if (!m)
@@ -323,9 +319,6 @@ int overlay_mount_parse(CustomMount **l, size_t *n, const char *s, bool read_onl
                         return -EINVAL;
         }
 
-        if (empty_or_root(destination))
-                return -EINVAL;
-
         m = custom_mount_add(l, n, CUSTOM_MOUNT_OVERLAY);
         if (!m)
                 return -ENOMEM;
@@ -422,7 +415,7 @@ int mount_sysfs(const char *dest, MountSettingsMask mount_settings) {
 
         (void) mkdir(full, 0755);
 
-        if (mount_settings & MOUNT_APPLY_APIVFS_RO)
+        if (FLAGS_SET(mount_settings, MOUNT_APPLY_APIVFS_RO))
                 extra_flags |= MS_RDONLY;
 
         r = mount_verbose(LOG_ERR, "sysfs", full, "sysfs",
@@ -608,29 +601,29 @@ int mount_all(const char *dest,
 #endif
         };
 
-        bool use_userns = (mount_settings & MOUNT_USE_USERNS);
-        bool netns = (mount_settings & MOUNT_APPLY_APIVFS_NETNS);
-        bool ro = (mount_settings & MOUNT_APPLY_APIVFS_RO);
-        bool in_userns = (mount_settings & MOUNT_IN_USERNS);
-        bool tmpfs_tmp = (mount_settings & MOUNT_APPLY_TMPFS_TMP);
+        bool use_userns = FLAGS_SET(mount_settings, MOUNT_USE_USERNS);
+        bool netns = FLAGS_SET(mount_settings, MOUNT_APPLY_APIVFS_NETNS);
+        bool ro = FLAGS_SET(mount_settings, MOUNT_APPLY_APIVFS_RO);
+        bool in_userns = FLAGS_SET(mount_settings, MOUNT_IN_USERNS);
+        bool tmpfs_tmp = FLAGS_SET(mount_settings, MOUNT_APPLY_TMPFS_TMP);
         size_t k;
         int r;
 
         for (k = 0; k < ELEMENTSOF(mount_table); k++) {
                 _cleanup_free_ char *where = NULL, *options = NULL;
                 const char *o;
-                bool fatal = (mount_table[k].mount_settings & MOUNT_FATAL);
+                bool fatal = FLAGS_SET(mount_table[k].mount_settings, MOUNT_FATAL);
 
-                if (in_userns != (bool)(mount_table[k].mount_settings & MOUNT_IN_USERNS))
+                if (in_userns != FLAGS_SET(mount_table[k].mount_settings, MOUNT_IN_USERNS))
                         continue;
 
-                if (!netns && (bool)(mount_table[k].mount_settings & MOUNT_APPLY_APIVFS_NETNS))
+                if (!netns && FLAGS_SET(mount_table[k].mount_settings, MOUNT_APPLY_APIVFS_NETNS))
                         continue;
 
-                if (!ro && (bool)(mount_table[k].mount_settings & MOUNT_APPLY_APIVFS_RO))
+                if (!ro && FLAGS_SET(mount_table[k].mount_settings, MOUNT_APPLY_APIVFS_RO))
                         continue;
 
-                if (!tmpfs_tmp && (bool)(mount_table[k].mount_settings & MOUNT_APPLY_TMPFS_TMP))
+                if (!tmpfs_tmp && FLAGS_SET(mount_table[k].mount_settings, MOUNT_APPLY_TMPFS_TMP))
                         continue;
 
                 r = chase_symlinks(mount_table[k].where, dest, CHASE_NONEXISTENT|CHASE_PREFIX_ROOT, &where, NULL);
@@ -782,11 +775,7 @@ static int mount_bind(const char *dest, CustomMount *m) {
         return 0;
 }
 
-static int mount_tmpfs(
-                const char *dest,
-                CustomMount *m,
-                bool userns, uid_t uid_shift, uid_t uid_range,
-                const char *selinux_apifs_context) {
+static int mount_tmpfs(const char *dest, CustomMount *m, uid_t uid_shift, const char *selinux_apifs_context) {
 
         const char *options;
         _cleanup_free_ char *buf = NULL, *where = NULL;
@@ -921,9 +910,9 @@ static int mount_arbitrary(const char *dest, CustomMount *m) {
 int mount_custom(
                 const char *dest,
                 CustomMount *mounts, size_t n,
-                bool userns, uid_t uid_shift, uid_t uid_range,
+                uid_t uid_shift,
                 const char *selinux_apifs_context,
-                bool in_userns) {
+                MountSettingsMask mount_settings) {
 
         size_t i;
         int r;
@@ -933,7 +922,13 @@ int mount_custom(
         for (i = 0; i < n; i++) {
                 CustomMount *m = mounts + i;
 
-                if (m->in_userns != in_userns)
+                if (FLAGS_SET(mount_settings, MOUNT_IN_USERNS) != m->in_userns)
+                        continue;
+
+                if (FLAGS_SET(mount_settings, MOUNT_ROOT_ONLY) && !path_equal(m->destination, "/"))
+                        continue;
+
+                if (FLAGS_SET(mount_settings, MOUNT_NON_ROOT_ONLY) && path_equal(m->destination, "/"))
                         continue;
 
                 switch (m->type) {
@@ -943,7 +938,7 @@ int mount_custom(
                         break;
 
                 case CUSTOM_MOUNT_TMPFS:
-                        r = mount_tmpfs(dest, m, userns, uid_shift, uid_range, selinux_apifs_context);
+                        r = mount_tmpfs(dest, m, uid_shift, selinux_apifs_context);
                         break;
 
                 case CUSTOM_MOUNT_OVERLAY:
@@ -969,10 +964,7 @@ int mount_custom(
         return 0;
 }
 
-static int setup_volatile_state(
-                const char *directory,
-                bool userns, uid_t uid_shift, uid_t uid_range,
-                const char *selinux_apifs_context) {
+static int setup_volatile_state(const char *directory, uid_t uid_shift, const char *selinux_apifs_context) {
 
         _cleanup_free_ char *buf = NULL;
         const char *p, *options;
@@ -1001,10 +993,7 @@ static int setup_volatile_state(
         return mount_verbose(LOG_ERR, "tmpfs", p, "tmpfs", MS_STRICTATIME, options);
 }
 
-static int setup_volatile_yes(
-                const char *directory,
-                bool userns, uid_t uid_shift, uid_t uid_range,
-                const char *selinux_apifs_context) {
+static int setup_volatile_yes(const char *directory, uid_t uid_shift, const char *selinux_apifs_context) {
 
         bool tmpfs_mounted = false, bind_mounted = false;
         char template[] = "/tmp/nspawn-volatile-XXXXXX";
@@ -1091,10 +1080,7 @@ fail:
         return r;
 }
 
-static int setup_volatile_overlay(
-                const char *directory,
-                bool userns, uid_t uid_shift, uid_t uid_range,
-                const char *selinux_apifs_context) {
+static int setup_volatile_overlay(const char *directory, uid_t uid_shift, const char *selinux_apifs_context) {
 
         _cleanup_free_ char *buf = NULL, *escaped_directory = NULL, *escaped_upper = NULL, *escaped_work = NULL;
         char template[] = "/tmp/nspawn-volatile-XXXXXX";
@@ -1159,19 +1145,19 @@ finish:
 int setup_volatile_mode(
                 const char *directory,
                 VolatileMode mode,
-                bool userns, uid_t uid_shift, uid_t uid_range,
+                uid_t uid_shift,
                 const char *selinux_apifs_context) {
 
         switch (mode) {
 
         case VOLATILE_YES:
-                return setup_volatile_yes(directory, userns, uid_shift, uid_range, selinux_apifs_context);
+                return setup_volatile_yes(directory, uid_shift, selinux_apifs_context);
 
         case VOLATILE_STATE:
-                return setup_volatile_state(directory, userns, uid_shift, uid_range, selinux_apifs_context);
+                return setup_volatile_state(directory, uid_shift, selinux_apifs_context);
 
         case VOLATILE_OVERLAY:
-                return setup_volatile_overlay(directory, userns, uid_shift, uid_range, selinux_apifs_context);
+                return setup_volatile_overlay(directory, uid_shift, selinux_apifs_context);
 
         default:
                 return 0;

src/nspawn/nspawn-mount.h

@@ -14,6 +14,8 @@ typedef enum MountSettingsMask {
         MOUNT_APPLY_APIVFS_NETNS = 1 << 4, /* if set, /proc/sys/net will be mounted read-write.
                                                Works only if MOUNT_APPLY_APIVFS_RO is also set. */
         MOUNT_APPLY_TMPFS_TMP    = 1 << 5, /* if set, /tmp will be mounted as tmpfs */
+        MOUNT_ROOT_ONLY          = 1 << 6, /* if set, only root mounts are mounted */
+        MOUNT_NON_ROOT_ONLY      = 1 << 7, /* if set, only non-root mounts are mounted */
 } MountSettingsMask;
 
 typedef enum CustomMountType {
@@ -52,9 +54,9 @@ int inaccessible_mount_parse(CustomMount **l, size_t *n, const char *s);
 int mount_all(const char *dest, MountSettingsMask mount_settings, uid_t uid_shift, const char *selinux_apifs_context);
 int mount_sysfs(const char *dest, MountSettingsMask mount_settings);
 
-int mount_custom(const char *dest, CustomMount *mounts, size_t n, bool userns, uid_t uid_shift, uid_t uid_range, const char *selinux_apifs_context, bool in_userns);
+int mount_custom(const char *dest, CustomMount *mounts, size_t n, uid_t uid_shift, const char *selinux_apifs_context, MountSettingsMask mount_settings);
 
-int setup_volatile_mode(const char *directory, VolatileMode mode, bool userns, uid_t uid_shift, uid_t uid_range, const char *selinux_apifs_context);
+int setup_volatile_mode(const char *directory, VolatileMode mode, uid_t uid_shift, const char *selinux_apifs_context);
 
 int pivot_root_parse(char **pivot_root_new, char **pivot_root_old, const char *s);
 int setup_pivot_root(const char *directory, const char *pivot_root_new, const char *pivot_root_old);

src/nspawn/nspawn.c

@@ -2975,11 +2975,9 @@ static int inner_child(
                         "/",
                         arg_custom_mounts,
                         arg_n_custom_mounts,
-                        false,
-                        0,
                         0,
                         arg_selinux_apifs_context,
-                        true);
+                        MOUNT_NON_ROOT_ONLY | MOUNT_IN_USERNS);
         if (r < 0)
                 return r;
 
@@ -3345,13 +3343,6 @@ static int outer_child(
                         return r;
 
                 directory = "/run/systemd/nspawn-root";
-
-        } else if (!dissected_image) {
-                /* Turn directory into bind mount (we need that so that we can move the bind mount to root
-                 * later on). */
-                r = mount_verbose(LOG_ERR, directory, directory, NULL, MS_BIND|MS_REC, NULL);
-                if (r < 0)
-                        return r;
         }
 
         r = setup_pivot_root(
@@ -3364,13 +3355,28 @@ static int outer_child(
         r = setup_volatile_mode(
                         directory,
                         arg_volatile_mode,
-                        arg_userns_mode != USER_NAMESPACE_NO,
                         arg_uid_shift,
-                        arg_uid_range,
                         arg_selinux_apifs_context);
         if (r < 0)
                 return r;
 
+        r = mount_custom(
+                        directory,
+                        arg_custom_mounts,
+                        arg_n_custom_mounts,
+                        arg_uid_shift,
+                        arg_selinux_apifs_context,
+                        MOUNT_ROOT_ONLY);
+        if (r < 0)
+                return r;
+
+        /* Make sure we always have a mount that we can move to root later on. */
+        if (!path_is_mount_point(directory, NULL, 0)) {
+                r = mount_verbose(LOG_ERR, directory, directory, NULL, MS_BIND|MS_REC, NULL);
+                if (r < 0)
+                        return r;
+        }
+
         if (dissected_image) {
                 /* Now we know the uid shift, let's now mount everything else that might be in the image. */
                 r = dissected_image_mount(dissected_image, directory, arg_uid_shift,
@@ -3401,7 +3407,12 @@ static int outer_child(
          * inside the container that create a new mount namespace.
          * See https://github.com/systemd/systemd/issues/3860
          * Further submounts (such as /dev) done after this will inherit the
-         * shared propagation mode. */
+         * shared propagation mode.
+         *
+         * IMPORTANT: Do not overmount the root directory anymore from now on to
+         * enable moving the root directory mount to root later on.
+         * https://github.com/systemd/systemd/issues/3847#issuecomment-562735251
+         */
         r = mount_verbose(LOG_ERR, NULL, directory, NULL, MS_SHARED|MS_REC, NULL);
         if (r < 0)
                 return r;
@@ -3470,11 +3481,9 @@ static int outer_child(
                         directory,
                         arg_custom_mounts,
                         arg_n_custom_mounts,
-                        arg_userns_mode != USER_NAMESPACE_NO,
                         arg_uid_shift,
-                        arg_uid_range,
                         arg_selinux_apifs_context,
-                        false);
+                        MOUNT_NON_ROOT_ONLY);
         if (r < 0)
                 return r;
 

test/fuzz/fuzz-network-parser/directives.network

@@ -279,4 +279,11 @@ TokenBufferFilterPeakRate=
 TokenBufferFilterLatencySec=
 StochasticFairnessQueueingPerturbPeriodSec=
 FairQueuingControlledDelayPacketLimit=
+FairQueuingControlledDelayMemoryLimit=
+FairQueuingControlledDelayFlows=
+FairQueuingControlledDelayQuantum=
+FairQueuingControlledDelayTargetSec=
+FairQueuingControlledDelayIntervalSec=
+FairQueuingControlledDelayCEThresholdSec=
+FairQueuingControlledDelayECN=
 FairQueueTrafficPolicingPacketLimit=

test/test-network/conf/25-qdisc-netem-and-fqcodel.network

@@ -15,3 +15,10 @@ NetworkEmulatorPacketLimit=100
 [TrafficControlQueueingDiscipline]
 Parent=ingress
 FairQueuingControlledDelayPacketLimit=20480
+FairQueuingControlledDelayMemoryLimit=64M
+FairQueuingControlledDelayFlows=2048
+FairQueuingControlledDelayTargetSec=10ms
+FairQueuingControlledDelayIntervalSec=200ms
+FairQueuingControlledDelayQuantum=1400
+FairQueuingControlledDelayECN=yes
+FairQueuingControlledDelayCEThresholdSec=100ms

test/test-network/systemd-networkd-tests.py

@@ -2095,7 +2095,7 @@ class NetworkdNetworkTests(unittest.TestCase, Utilities):
         self.assertRegex(output, 'qdisc netem')
         self.assertRegex(output, 'limit 100 delay 50.0ms  10.0ms loss 20%')
         self.assertRegex(output, 'qdisc fq_codel')
-        self.assertRegex(output, 'limit 20480p')
+        self.assertRegex(output, 'limit 20480p flows 2048 quantum 1400 target 10.0ms ce_threshold 100.0ms interval 200.0ms memory_limit 64Mb ecn')
         output = check_output('tc qdisc show dev test1')
         print(output)
         self.assertRegex(output, 'qdisc tbf')