bpf-firewall: custom BPF programs through IP(Ingress|Egress)FilterPath=

Takes a single /sys/fs/bpf/pinned_prog string as argument, but may be
specified multiple times. An empty assignment resets all previous filters.

Closes #10227

man/systemd.resource-control.xml

@@ -618,6 +618,39 @@
         </listitem>
       </varlistentry>
 
+      <varlistentry>
+        <term><varname>IPIngressFilterPath=<replaceable>BPF_FS_PROGRAMM_PATH</replaceable></varname></term>
+        <term><varname>IPEgressFilterPath=<replaceable>BPF_FS_PROGRAMM_PATH</replaceable></varname></term>
+
+        <listitem>
+          <para>Add custom network traffic filters implemented as BPF programs, applying to all IP packets
+          sent and received over <constant>AF_INET</constant> and <constant>AF_INET6</constant> sockets.
+          Takes an absolute path to a pinned BPF program in the BPF virtual filesystem (<filename>/sys/fs/bpf/</filename>).
+          </para>
+
+          <para>The filters configured with this option are applied to all sockets created by processes
+          of this unit (or in the case of socket units, associated with it). The filters are loaded in addition
+          to filters any of the parent slice units this unit might be a member of as well as any
+          <varname>IPAddressAllow=</varname> and <varname>IPAddressDeny=</varname> filters in any of these units.
+          By default there are no filters specified.</para>
+
+          <para>If these settings are used multiple times in the same unit all the specified programs are attached. If an
+          empty string is assigned to these settings the program list is reset and all previous specified programs ignored.</para>
+
+          <para>Note that for socket-activated services, the IP filter programs configured on the socket unit apply to
+          all sockets associated with it directly, but not to any sockets created by the ultimately activated services
+          for it. Conversely, the IP filter programs configured for the service are not applied to any sockets passed into
+          the service via socket activation. Thus, it is usually a good idea, to replicate the IP filter programs on both
+          the socket and the service unit, however it often makes sense to maintain one configuration more open and the other
+          one more restricted, depending on the usecase.</para>
+
+          <para>Note that these settings might not be supported on some systems (for example if eBPF control group
+          support is not enabled in the underlying kernel or container manager). These settings will fail the service in
+          that case. If compatibility with such systems is desired it is hence recommended to attach your filter manually
+          (requires <varname>Delegate=</varname><constant>yes</constant>) instead of using this setting.</para>
+        </listitem>
+      </varlistentry>
+
       <varlistentry>
         <term><varname>DeviceAllow=</varname></term>
 

src/analyze/analyze-security.c

@@ -45,6 +45,9 @@ struct security_info {
         bool ip_address_allow_localhost;
         bool ip_address_allow_other;
 
+        bool ip_filters_custom_ingress;
+        bool ip_filters_custom_egress;
+
         char *keyring_mode;
         bool lock_personality;
         bool memory_deny_write_execute;
@@ -590,7 +593,10 @@ static int assess_ip_address_allow(
         assert(ret_badness);
         assert(ret_description);
 
-        if (!info->ip_address_deny_all) {
+        if (info->ip_filters_custom_ingress || info->ip_filters_custom_egress) {
+                d = strdup("Service defines custom ingress/egress IP filters with BPF programs");
+                b = 0;
+        } else if (!info->ip_address_deny_all) {
                 d = strdup("Service does not define an IP address whitelist");
                 b = 10;
         } else if (info->ip_address_allow_other) {
@@ -1824,6 +1830,33 @@ static int property_read_ip_address_allow(
         return sd_bus_message_exit_container(m);
 }
 
+static int property_read_ip_filters(
+                sd_bus *bus,
+                const char *member,
+                sd_bus_message *m,
+                sd_bus_error *error,
+                void *userdata) {
+
+        struct security_info *info = userdata;
+        _cleanup_(strv_freep) char **l = NULL;
+        int r;
+
+        assert(bus);
+        assert(member);
+        assert(m);
+
+        r = sd_bus_message_read_strv(m, &l);
+        if (r < 0)
+                return r;
+
+        if (streq(member, "IPIngressFilterPath"))
+                info->ip_filters_custom_ingress = !strv_isempty(l);
+        else if (streq(member, "IPEgressFilterPath"))
+                info->ip_filters_custom_ingress = !strv_isempty(l);
+
+        return 0;
+}
+
 static int property_read_device_allow(
                 sd_bus *bus,
                 const char *member,
@@ -1873,6 +1906,8 @@ static int acquire_security_info(sd_bus *bus, const char *name, struct security_
                 { "FragmentPath",            "s",       NULL,                                    offsetof(struct security_info, fragment_path)             },
                 { "IPAddressAllow",          "a(iayu)", property_read_ip_address_allow,          0                                                         },
                 { "IPAddressDeny",           "a(iayu)", property_read_ip_address_allow,          0                                                         },
+                { "IPIngressFilterPath",     "as",      property_read_ip_filters,                0                                                         },
+                { "IPEgressFilterPath",      "as",      property_read_ip_filters,                0                                                         },
                 { "Id",                      "s",       NULL,                                    offsetof(struct security_info, id)                        },
                 { "KeyringMode",             "s",       NULL,                                    offsetof(struct security_info, keyring_mode)              },
                 { "LoadState",               "s",       NULL,                                    offsetof(struct security_info, load_state)                },

src/core/bpf-firewall.c

@@ -587,6 +587,95 @@ int bpf_firewall_compile(Unit *u) {
         return 0;
 }
 
+DEFINE_PRIVATE_HASH_OPS_WITH_VALUE_DESTRUCTOR(filter_prog_hash_ops, void, trivial_hash_func, trivial_compare_func, BPFProgram, bpf_program_unref);
+
+static int load_bpf_progs_from_fs_to_set(Unit *u, char **filter_paths, Set **set) {
+        char **bpf_fs_path;
+
+        set_clear(*set);
+
+        STRV_FOREACH(bpf_fs_path, filter_paths) {
+                _cleanup_free_ BPFProgram *prog = NULL;
+                int r;
+
+                r = bpf_program_new(BPF_PROG_TYPE_CGROUP_SKB, &prog);
+                if (r < 0)
+                        return log_unit_error_errno(u, r, "Can't allocate CGROUP SKB BPF program: %m");
+
+                r = bpf_program_load_from_bpf_fs(prog, *bpf_fs_path);
+                if (r < 0)
+                        return log_unit_error_errno(u, r, "Loading of ingress BPF program %s failed: %m", *bpf_fs_path);
+
+                r = set_ensure_allocated(set, &filter_prog_hash_ops);
+                if (r < 0)
+                        return log_unit_error_errno(u, r, "Can't allocate BPF program set: %m");
+
+                r = set_put(*set, prog);
+                if (r < 0)
+                        return log_unit_error_errno(u, r, "Can't add program to BPF program set: %m");
+                TAKE_PTR(prog);
+        }
+
+        return 0;
+}
+
+int bpf_firewall_load_custom(Unit *u) {
+        CGroupContext *cc;
+        int r, supported;
+
+        assert(u);
+
+        cc = unit_get_cgroup_context(u);
+        if (!cc)
+                return 0;
+
+        if (!(cc->ip_filters_ingress || cc->ip_filters_egress))
+                return 0;
+
+        supported = bpf_firewall_supported();
+        if (supported < 0)
+                return supported;
+
+        if (supported != BPF_FIREWALL_SUPPORTED_WITH_MULTI)
+                return log_unit_debug_errno(u, SYNTHETIC_ERRNO(EOPNOTSUPP), "BPF_F_ALLOW_MULTI not supported on this manager, cannot attach custom BPF programs.");
+
+        r = load_bpf_progs_from_fs_to_set(u, cc->ip_filters_ingress, &u->ip_bpf_custom_ingress);
+        if (r < 0)
+                return r;
+        r = load_bpf_progs_from_fs_to_set(u, cc->ip_filters_egress, &u->ip_bpf_custom_egress);
+        if (r < 0)
+                return r;
+
+        return 0;
+}
+
+static int attach_custom_bpf_progs(Unit *u, const char *path, int attach_type, Set **set, Set **set_installed) {
+        BPFProgram *prog;
+        Iterator i;
+        int r;
+
+        assert(u);
+
+        set_clear(*set_installed);
+
+        SET_FOREACH(prog, *set, i) {
+                r = bpf_program_cgroup_attach(prog, attach_type, path, BPF_F_ALLOW_MULTI);
+                if (r < 0)
+                        return log_unit_error_errno(u, r, "Attaching custom egress BPF program to cgroup %s failed: %m", path);
+                /* Remember that these BPF programs are installed now. */
+                r = set_ensure_allocated(set_installed, &filter_prog_hash_ops);
+                if (r < 0)
+                        return log_unit_error_errno(u, r, "Can't allocate BPF program set: %m");
+
+                r = set_put(*set_installed, prog);
+                if (r < 0)
+                        return log_unit_error_errno(u, r, "Can't add program to BPF program set: %m");
+                bpf_program_ref(prog);
+        }
+
+        return 0;
+}
+
 int bpf_firewall_install(Unit *u) {
         _cleanup_free_ char *path = NULL;
         CGroupContext *cc;
@@ -614,6 +703,9 @@ int bpf_firewall_install(Unit *u) {
                 log_unit_debug(u, "BPF_F_ALLOW_MULTI is not supported on this manager, not doing BPF firewall on slice units.");
                 return -EOPNOTSUPP;
         }
+        if (supported != BPF_FIREWALL_SUPPORTED_WITH_MULTI &&
+            (!set_isempty(u->ip_bpf_custom_ingress) || !set_isempty(u->ip_bpf_custom_egress)))
+                return log_unit_debug_errno(u, SYNTHETIC_ERRNO(EOPNOTSUPP), "BPF_F_ALLOW_MULTI not supported on this manager, cannot attach custom BPF programs.");
 
         r = cg_get_path(SYSTEMD_CGROUP_CONTROLLER, u->cgroup_path, NULL, &path);
         if (r < 0)
@@ -628,7 +720,8 @@ int bpf_firewall_install(Unit *u) {
         u->ip_bpf_ingress_installed = bpf_program_unref(u->ip_bpf_ingress_installed);
 
         if (u->ip_bpf_egress) {
-                r = bpf_program_cgroup_attach(u->ip_bpf_egress, BPF_CGROUP_INET_EGRESS, path, flags);
+                r = bpf_program_cgroup_attach(u->ip_bpf_egress, BPF_CGROUP_INET_EGRESS, path,
+                                              flags | (set_isempty(u->ip_bpf_custom_egress) ? 0 : BPF_F_ALLOW_MULTI));
                 if (r < 0)
                         return log_unit_error_errno(u, r, "Attaching egress BPF program to cgroup %s failed: %m", path);
 
@@ -637,13 +730,22 @@ int bpf_firewall_install(Unit *u) {
         }
 
         if (u->ip_bpf_ingress) {
-                r = bpf_program_cgroup_attach(u->ip_bpf_ingress, BPF_CGROUP_INET_INGRESS, path, flags);
+                r = bpf_program_cgroup_attach(u->ip_bpf_ingress, BPF_CGROUP_INET_INGRESS, path,
+                                              flags | (set_isempty(u->ip_bpf_custom_ingress) ? 0 : BPF_F_ALLOW_MULTI));
                 if (r < 0)
                         return log_unit_error_errno(u, r, "Attaching ingress BPF program to cgroup %s failed: %m", path);
 
                 u->ip_bpf_ingress_installed = bpf_program_ref(u->ip_bpf_ingress);
         }
 
+        r = attach_custom_bpf_progs(u, path, BPF_CGROUP_INET_EGRESS, &u->ip_bpf_custom_egress, &u->ip_bpf_custom_egress_installed);
+        if (r < 0)
+                return r;
+
+        r = attach_custom_bpf_progs(u, path, BPF_CGROUP_INET_INGRESS, &u->ip_bpf_custom_ingress, &u->ip_bpf_custom_ingress_installed);
+        if (r < 0)
+                return r;
+
         return 0;
 }
 

src/core/bpf-firewall.h

@@ -15,6 +15,7 @@ int bpf_firewall_supported(void);
 
 int bpf_firewall_compile(Unit *u);
 int bpf_firewall_install(Unit *u);
+int bpf_firewall_load_custom(Unit *u);
 
 int bpf_firewall_read_accounting(int map_fd, uint64_t *ret_bytes, uint64_t *ret_packets);
 int bpf_firewall_reset_accounting(int map_fd);

src/core/cgroup.c

@@ -199,6 +199,9 @@ void cgroup_context_done(CGroupContext *c) {
 
         c->ip_address_allow = ip_address_access_free_all(c->ip_address_allow);
         c->ip_address_deny = ip_address_access_free_all(c->ip_address_deny);
+
+        c->ip_filters_ingress = strv_free(c->ip_filters_ingress);
+        c->ip_filters_egress = strv_free(c->ip_filters_egress);
 }
 
 void cgroup_context_dump(CGroupContext *c, FILE* f, const char *prefix) {
@@ -210,6 +213,7 @@ void cgroup_context_dump(CGroupContext *c, FILE* f, const char *prefix) {
         CGroupBlockIODeviceWeight *w;
         CGroupDeviceAllow *a;
         IPAddressAccessItem *iaai;
+        char **path;
         char u[FORMAT_TIMESPAN_MAX];
         char v[FORMAT_TIMESPAN_MAX];
 
@@ -360,6 +364,12 @@ void cgroup_context_dump(CGroupContext *c, FILE* f, const char *prefix) {
                 (void) in_addr_to_string(iaai->family, &iaai->address, &k);
                 fprintf(f, "%sIPAddressDeny=%s/%u\n", prefix, strnull(k), iaai->prefixlen);
         }
+
+        STRV_FOREACH(path, c->ip_filters_ingress)
+                fprintf(f, "%sIPIngressFilterPath=%s\n", prefix, *path);
+
+        STRV_FOREACH(path, c->ip_filters_egress)
+                fprintf(f, "%sIPEgressFilterPath=%s\n", prefix, *path);
 }
 
 int cgroup_add_device_allow(CGroupContext *c, const char *dev, const char *mode) {
@@ -945,6 +955,7 @@ static void cgroup_apply_firewall(Unit *u) {
         if (bpf_firewall_compile(u) < 0)
                 return;
 
+        (void) bpf_firewall_load_custom(u);
         (void) bpf_firewall_install(u);
 }
 
@@ -1353,7 +1364,9 @@ static bool unit_get_needs_bpf_firewall(Unit *u) {
 
         if (c->ip_accounting ||
             c->ip_address_allow ||
-            c->ip_address_deny)
+            c->ip_address_deny ||
+            c->ip_filters_ingress ||
+            c->ip_filters_egress)
                 return true;
 
         /* If any parent slice has an IP access list defined, it applies too */
@@ -1919,6 +1932,12 @@ int unit_attach_pids_to_cgroup(Unit *u, Set *pids, const char *suffix_path) {
         if (set_isempty(pids))
                 return 0;
 
+        /* Load any custom firewall BPF programs here once to test if they are existing and actually loadable.
+         * Fail here early since later errors in the call chain unit_realize_cgroup to cgroup_context_apply are ignored. */
+        r = bpf_firewall_load_custom(u);
+        if (r < 0)
+                return r;
+
         r = unit_realize_cgroup(u);
         if (r < 0)
                 return r;

src/core/cgroup.h

@@ -114,6 +114,9 @@ struct CGroupContext {
         LIST_HEAD(IPAddressAccessItem, ip_address_allow);
         LIST_HEAD(IPAddressAccessItem, ip_address_deny);
 
+        char **ip_filters_ingress;
+        char **ip_filters_egress;
+
         /* For legacy hierarchies */
         uint64_t cpu_shares;
         uint64_t startup_cpu_shares;