Skip to contentSkip to conversation

auto-t: no hostapd instance graceful failure

PATCH In review James Prestwood 3 participants 23 messages Oct 20, 2020 Kbuild

lore
Share & tools
lore
#0 lore raw

PATCH 1/13 auto-t: no hostapd instance graceful failure

If a test does not need any hostapd instances but still loads hostapd.py for some reason we want to gracefully throw an exception rather than fail in some other manor.

---
 autotests/util/hostapd.py | 7 +++++++
 1 file changed, 7 insertions(+)
autotests/util/hostapd.py +7 -0
index 7c0d8385..cf82c010 100644
@@ -32,6 +32,10 @@ class HostapdCLI:
     def _init_hostapd(self, config=None):
         global ctrl_count
         interface = None
+        self.ctrl_sock = None
+
+        if not ctx.hostapd:
+            raise Exception("No hostapd instances are configured")
 
         if not config and len(ctx.hostapd.instances) > 1:
             raise Exception('config must be provided if more than one hostapd instance exists')
@@ -110,6 +114,9 @@ class HostapdCLI:
         raise Exception('timeout waiting for control response')
 
     def _del_hostapd(self, force=False):
+        if not self.ctrl_sock:
+            return
+
         self.ctrl_sock.close()
         os.remove(self.local_ctrl)
 
Signature · 1 line
2.26.2
#1 lore raw

PATCH 2/13 auto-t: add copy_to_ap utility

---
 autotests/util/iwd.py | 8 ++++++++
 1 file changed, 8 insertions(+)
autotests/util/iwd.py +8 -0
index dc4f93a2..fb22412d 100755
@@ -1027,6 +1027,7 @@ class IWD(AsyncOpAbstract):
     def clear_storage(storage_dir=IWD_STORAGE_DIR):
         os.system('rm -rf ' + storage_dir + '/*')
         os.system('rm -rf ' + storage_dir + '/hotspot/*')
+        os.system('rm -rf ' + storage_dir + '/ap/*')
 
     @staticmethod
     def create_in_storage(file_name, file_content):
@@ -1054,6 +1055,13 @@ class IWD(AsyncOpAbstract):
 
         shutil.copy(source, IWD_STORAGE_DIR + "/hotspot")
 
+    @staticmethod
+    def copy_to_ap(source):
+        if not os.path.exists(IWD_STORAGE_DIR + "/ap"):
+            os.mkdir(IWD_STORAGE_DIR + "/ap")
+
+        IWD.copy_to_storage(source, IWD_STORAGE_DIR + '/ap/')
+
     @staticmethod
     def remove_from_storage(file_name):
         os.system('rm -rf ' + IWD_STORAGE_DIR + '/\'' + file_name + '\'')
Signature · 1 line
2.26.2
#2 lore raw

PATCH 3/13 auto-t: simplify copy_to_hotspot

In reply to James Prestwood

Since copy_to_storage can take an arbitrary destination path now we can reuse that instead of duplicating code.

---
 autotests/util/iwd.py | 6 +-----
 1 file changed, 1 insertion(+), 5 deletions(-)
autotests/util/iwd.py +1 -5
index fb22412d..9bfc8c1a 100755
@@ -1046,14 +1046,10 @@ class IWD(AsyncOpAbstract):
 
     @staticmethod
     def copy_to_hotspot(source):
-        import shutil
-
-        assert not os.path.isabs(source)
-
         if not os.path.exists(IWD_STORAGE_DIR + "/hotspot"):
             os.mkdir(IWD_STORAGE_DIR + "/hotspot")
 
-        shutil.copy(source, IWD_STORAGE_DIR + "/hotspot")
+        IWD.copy_to_storage(source, IWD_STORAGE_DIR + "/hotspot")
 
     @staticmethod
     def copy_to_ap(source):
Signature · 1 line
2.26.2
#3 lore raw

PATCH 4/13 storage: allow NULL type on storage_network_ssid_from_path

In reply to James Prestwood

This is to allow this API to be used simply to get the file name without extension. AP will be storing provisioning files similarly to the network format but the extension will not map to existing security types. For this case it makes sense to allow a NULL type which will prevent the security type from being set/parsed.

---
 src/storage.c | 10 +++++++++-
 1 file changed, 9 insertions(+), 1 deletion(-)
src/storage.c +9 -1
index 00d93933..9ef84e32 100644
@@ -287,6 +287,11 @@ char *storage_get_network_file_path(enum security type, const char *ssid)
 	return path;
 }
 
+/*
+ * Get filename (without extension) which IWD treats as an SSID. Optionally
+ * parse the extension to get the security type (*.psk, *.open, *.8021x) by
+ * providing a 'type' pointer.
+ */
 const char *storage_network_ssid_from_path(const char *path,
 							enum security *type)
 {
@@ -302,7 +307,10 @@ const char *storage_network_ssid_from_path(const char *path,
 
 	end = strchr(filename, '.');
 
-	if (!end || !security_from_str(end + 1, type))
+	if (!end)
+		return NULL;
+
+	if (type && !security_from_str(end + 1, type))
 		return NULL;
 
 	if (filename[0] != '=') {
Signature · 1 line
2.26.2
#4 lore raw

Re: [PATCH 04/13] storage: allow NULL type on storage_network_ssid_from_path

Hi James,

Quoted 14 lines · James Prestwood

This is to allow this API to be used simply to get the file name without extension. AP will be storing provisioning files similarly to the network format but the extension will not map to existing security types. For this case it makes sense to allow a NULL type which will prevent the security type from being set/parsed.

---
  src/storage.c | 10 +++++++++-
  1 file changed, 9 insertions(+), 1 deletion(-)
src/storage.c +0 -0
index 00d93933..9ef84e32 100644

src/storage.c char *storage_get_network_file_path(enum security type, const char *ssid) · line 296

Quoted 10 lines
src/storage.c +5 -0
@@ -287,6 +287,11 @@ char *storage_get_network_file_path(enum security type, const char *ssid)
  	return path;
  }
 
+/*
+ * Get filename (without extension) which IWD treats as an SSID. Optionally
+ * parse the extension to get the security type (*.psk, *.open, *.8021x) by
+ * providing a 'type' pointer.
+ */
  const char *storage_network_ssid_from_path(const char *path,
  							enum security *type)

No, please don't do this. Make up a AP specific API function instead and commonize the implementation if needed.

Quoted 4 lines
src/storage.c +0 -0
  {
@@ -302,7 +307,10 @@ const char *storage_network_ssid_from_path(const char *path,
 

src/storage.c const char *storage_network_ssid_from_path(const char *path, · line 316

Quoted 10 lines
src/storage.c +4 -1
  	end = strchr(filename, '.');
 
-	if (!end || !security_from_str(end + 1, type))
+	if (!end)
+		return NULL;
+
+	if (type && !security_from_str(end + 1, type))
  		return NULL;
 
  	if (filename[0] != '=') {

Regards,
-Denis

#5 lore raw

PATCH 5/13 storage: add storage_get_ap_path

In reply to James Prestwood
---
 src/storage.c | 30 ++++++++++++++++++++++++++++++
 src/storage.h |  1 +
 2 files changed, 31 insertions(+)
src/storage.c +30 -0
index 9ef84e32..b509e161 100644
@@ -50,6 +50,7 @@
 
 static char *storage_path = NULL;
 static char *storage_hotspot_path = NULL;
+static char *storage_ap_path = NULL;
 
 static int create_dirs(const char *filename)
 {
@@ -221,6 +222,16 @@ bool storage_create_dirs(void)
 		return false;
 	}
 
+	storage_ap_path = l_strdup_printf("%s/ap/", storage_path);
+
+	if (create_dirs(storage_ap_path)) {
+		l_error("Failed to create %s", storage_ap_path);
+
+		l_free(storage_path);
+		l_free(storage_hotspot_path);
+		l_free(storage_ap_path);
+	}
+
 	return true;
 }
 
@@ -228,6 +239,7 @@ void storage_cleanup_dirs(void)
 {
 	l_free(storage_path);
 	l_free(storage_hotspot_path);
+	l_free(storage_ap_path);
 }
 
 char *storage_get_path(const char *format, ...)
@@ -266,6 +278,24 @@ char *storage_get_hotspot_path(const char *format, ...)
 	return str;
 }
 
+char *storage_get_ap_path(const char *format, ...)
+{
+	va_list args;
+	char *fmt, *str;
+
+	if (!format)
+		return l_strdup(storage_ap_path);
+
+	fmt = l_strdup_printf("%s/%s", storage_ap_path, format);
+
+	va_start(args, format);
+	str = l_strdup_vprintf(fmt, args);
+	va_end(args);
+
+	l_free(fmt);
+	return str;
+}
+
 char *storage_get_network_file_path(enum security type, const char *ssid)
 {
 	char *path;
src/storage.h +1 -0
index 80b63c53..54c81d10 100644
@@ -36,6 +36,7 @@ bool storage_create_dirs(void);
 void storage_cleanup_dirs(void);
 char *storage_get_path(const char *format, ...);
 char *storage_get_hotspot_path(const char *format, ...);
+char *storage_get_ap_path(const char *format, ...);
 
 const char *storage_network_ssid_from_path(const char *path,
 							enum security *type);
Signature · 1 line
2.26.2
#6 lore raw

PATCH 6/13 ap: refactor AP to use provisioning files

In reply to James Prestwood

Currently the AP DBus APIs take an SSID and passphrase to start an access point. This works for basic access points but isn't very scaleable once other AP options are added.

Instead we can define AP networks similar to how we define station networks by providing a provisioning file. The provisioning files for AP will reside in $STATE_DIRECTORY/ap (/var/lib/iwd/ap by default). The name of the file indicates the SSID and, for now, the only value will be [Security].Passphrase to match the DBus API. The file extension does not matter at this point, but in tests and examples *.ap will be used.

AP provisioning files will be loaded in at startup and remain as static configs in a queue. When Start() is called the ssid will be looked up and that config will be used. If no config is found a "Not Found" error will be returned.

---
 src/ap.c | 106 ++++++++++++++++++++++++++++++++++++++++++++++++++-----
 src/ap.h |   1 +
 2 files changed, 99 insertions(+), 8 deletions(-)
src/ap.c +98 -8
index 3c4ae907..9df6e14e 100644
@@ -26,6 +26,7 @@
 
 #include <errno.h>
 #include <linux/if_ether.h>
+#include <dirent.h>
 
 #include <ell/ell.h>
 
@@ -49,6 +50,7 @@
 #include "src/wscutil.h"
 #include "src/eap-wsc.h"
 #include "src/ap.h"
+#include "src/storage.h"
 
 struct ap_state {
 	struct netdev *netdev;
@@ -105,6 +107,7 @@ struct ap_wsc_pbc_probe_record {
 };
 
 static uint32_t netdev_watch;
+static struct l_queue *ap_configs;
 
 void ap_config_free(struct ap_config *config)
 {
@@ -186,7 +189,10 @@ static void ap_reset(struct ap_state *ap)
 	if (ap->rates)
 		l_uintset_free(ap->rates);
 
-	ap_config_free(ap->config);
+
+	if (!ap->config->no_free)
+		ap_config_free(ap->config);
+
 	ap->config = NULL;
 
 	l_queue_destroy(ap->wsc_pbc_probes, l_free);
@@ -2376,6 +2382,14 @@ static const struct ap_ops ap_dbus_ops = {
 	.handle_event = ap_if_event_func,
 };
 
+static bool match_config_ssid(const void *data, const void *user_data)
+{
+	const struct ap_config *config = data;
+	const char *ssid = user_data;
+
+	return !strcmp(config->ssid, ssid);
+}
+
 static struct l_dbus_message *ap_dbus_start(struct l_dbus *dbus,
 		struct l_dbus_message *message, void *user_data)
 {
@@ -2393,16 +2407,13 @@ static struct l_dbus_message *ap_dbus_start(struct l_dbus *dbus,
 						&ssid, &wpa2_passphrase))
 		return dbus_error_invalid_args(message);
 
-	config = l_new(struct ap_config, 1);
-	config->ssid = l_strdup(ssid);
-	l_strlcpy(config->passphrase, wpa2_passphrase,
-			sizeof(config->passphrase));
+	config = l_queue_find(ap_configs, match_config_ssid, ssid);
+	if (!config)
+		return dbus_error_not_found(message);
 
 	ap_if->ap = ap_start(ap_if->netdev, config, &ap_dbus_ops, ap_if);
-	if (!ap_if->ap) {
-		ap_config_free(config);
+	if (!ap_if->ap)
 		return dbus_error_invalid_args(message);
-	}
 
 	ap_if->pending = l_dbus_message_ref(message);
 	return NULL;
@@ -2528,13 +2539,90 @@ static void ap_netdev_watch(struct netdev *netdev,
 	}
 }
 
+static struct ap_config *ap_config_from_settings(struct l_settings *settings,
+							const char *ssid)
+{
+	struct ap_config *config = l_new(struct ap_config, 1);
+	char *passphrase;
+
+	/*
+	 * Dont free when ap interface goes up/down. This is a static config
+	 * that we want to keep around until IWD stops.
+	 */
+	config->no_free = true;
+	config->ssid = l_strdup(ssid);
+
+	passphrase = l_settings_get_string(settings, "Security", "Passphrase");
+	if (passphrase) {
+		if (strlen(passphrase) > 63) {
+			l_error("[Security].Passphrase must not exceed "
+					"63 characters");
+			goto failed;
+		}
+
+		strcpy(config->passphrase, passphrase);
+		l_free(passphrase);
+	}
+
+done:
+	l_info("Loaded AP configuration %s", config->ssid);
+	return config;
+
+failed:
+	ap_config_free(config);
+	return NULL;
+}
+
 static int ap_init(void)
 {
+	DIR *dir;
+	struct dirent *dirent;
+
+	L_AUTO_FREE_VAR(char *, ap_dir) = storage_get_ap_path(NULL);
+
 	netdev_watch = netdev_watch_add(ap_netdev_watch, NULL, NULL);
 
 	l_dbus_register_interface(dbus_get_bus(), IWD_AP_INTERFACE,
 			ap_setup_interface, ap_destroy_interface, false);
 
+	ap_configs = l_queue_new();
+
+	dir = opendir(ap_dir);
+	if (!dir)
+		return -ENOENT;
+
+	while ((dirent = readdir(dir))) {
+		struct ap_config *config;
+		struct l_settings *s;
+		char *filename;
+		const char *ssid;
+
+		if (dirent->d_type != DT_REG && dirent->d_type != DT_LNK)
+			continue;
+
+		filename = l_strdup_printf("%s/%s", ap_dir, dirent->d_name);
+		s = l_settings_new();
+
+		if (!l_settings_load_from_file(s, filename))
+			goto next;
+
+		ssid = storage_network_ssid_from_path(filename, NULL);
+
+		config = ap_config_from_settings(s, ssid);
+		if (!config)
+			goto next;
+
+		l_debug("loaded ap config %s", filename);
+
+		l_queue_push_head(ap_configs, config);
+
+next:
+		l_free(filename);
+		l_settings_free(s);
+	}
+
+	closedir(dir);
+
 	return 0;
 }
 
@@ -2542,6 +2630,8 @@ static void ap_exit(void)
 {
 	netdev_watch_remove(netdev_watch);
 	l_dbus_unregister_interface(dbus_get_bus(), IWD_AP_INTERFACE);
+
+	l_queue_destroy(ap_configs, (l_queue_destroy_func_t) ap_config_free);
 }
 
 IWD_MODULE(ap, ap_init, ap_exit)
src/ap.h +1 -0
index a39797aa..90497008 100644
@@ -64,6 +64,7 @@ struct ap_config {
 	char *wsc_name;
 	struct wsc_primary_device_type wsc_primary_device_type;
 	bool no_cck_rates : 1;
+	bool no_free : 1;
 };
 
 struct ap_ops {
Signature · 1 line
2.26.2
#7 lore raw

PATCH 7/13 doc: update AP docs with new Start() arguments

In reply to James Prestwood

The 'psk' argument has now been removed and replaced with a dedicated provisioning file which should contain the psk.

---
 doc/access-point-api.txt | 8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)
doc/access-point-api.txt +5 -3
index 9ee06baa..7d504b90 100644
@@ -5,15 +5,17 @@ Service		net.connman.iwd
 Interface	net.connman.iwd.AccessPoint [Experimental]
 Object path	/net/connman/iwd/{phy0,phy1,...}/{1,2,...}
 
-Methods		void Start(string ssid, string psk)
+Methods		void Start(string ssid)
 
-			Start an access point called ssid with a passphrase
-			of psk.
+			Start an access point called ssid. A network
+			provisioning file must exist in /var/lib/iwd/ap/ who's
+			file name matches the ssid in the DBus message.
 
 			Possible errors:	net.connman.iwd.Busy
 						net.connman.iwd.Failed
 						net.connman.iwd.InvalidArguments
 						net.connman.iwd.AlreadyExists
+						net.connman.iwd.NotFound
 
 		void Stop()
 
Signature · 1 line
2.26.2
#8 lore raw

PATCH 8/13 ap: remove 'psk' from Start()

In reply to James Prestwood

This has been replaced by dedicated provisioning files.

---
 src/ap.c | 7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)
src/ap.c +3 -4
index 9df6e14e..41de69a9 100644
@@ -2394,7 +2394,7 @@ static struct l_dbus_message *ap_dbus_start(struct l_dbus *dbus,
 		struct l_dbus_message *message, void *user_data)
 {
 	struct ap_if_data *ap_if = user_data;
-	const char *ssid, *wpa2_passphrase;
+	const char *ssid;
 	struct ap_config *config;
 
 	if (ap_if->ap && ap_if->ap->started)
@@ -2403,8 +2403,7 @@ static struct l_dbus_message *ap_dbus_start(struct l_dbus *dbus,
 	if (ap_if->ap || ap_if->pending)
 		return dbus_error_busy(message);
 
-	if (!l_dbus_message_get_arguments(message, "ss",
-						&ssid, &wpa2_passphrase))
+	if (!l_dbus_message_get_arguments(message, "s", &ssid))
 		return dbus_error_invalid_args(message);
 
 	config = l_queue_find(ap_configs, match_config_ssid, ssid);
@@ -2473,7 +2472,7 @@ static bool ap_dbus_property_get_started(struct l_dbus *dbus,
 static void ap_setup_interface(struct l_dbus_interface *interface)
 {
 	l_dbus_interface_method(interface, "Start", 0, ap_dbus_start, "",
-			"ss", "ssid", "wpa2_passphrase");
+			"s", "ssid");
 	l_dbus_interface_method(interface, "Stop", 0, ap_dbus_stop, "", "");
 
 	l_dbus_interface_property(interface, "Started", 0, "b",
Signature · 1 line
2.26.2
#9 lore raw

Re: [PATCH 08/13] ap: remove 'psk' from Start()

Hi James,

Quoted 2 lines · James Prestwood

This has been replaced by dedicated provisioning files.

This is going to break Network Manager and an additional problem is that clients have no way to tell which version of IWD they're talking to.

Currently clients can get a basic AP set up with just one method call and with this change they will have to use two different APIs:
filesystem + D-Bus. Can we make the config file optional? In my opinion this is a significant complication for the typical use-case (Android-style UI), while in most other scenarios hostapd is going to work better.

Best regards

#10 lore raw

Re: [PATCH 08/13] ap: remove 'psk' from Start()

Hi Andrew,

Quoted 10 lines · Andrew Zaborowski

Hi James,

Quoted 1 line · James Prestwood

This has been replaced by dedicated provisioning files.

This is going to break Network Manager and an additional problem is that clients have no way to tell which version of IWD they're talking to.

Quoted 6 lines

Currently clients can get a basic AP set up with just one method call and with this change they will have to use two different APIs:
filesystem + D-Bus. Can we make the config file optional? In my opinion this is a significant complication for the typical use-case (Android-style UI), while in most other scenarios hostapd is going to work better.

Yes Denis and I discussed this and came to the same conclusion. Better to have a quick and easy Start() API and probably add a sparate API for starting with a config file.

But if Network Manager is actually using this API we should probably remove the "experimental" label from the docs. I wasn't aware of this which is why I thought it was fine modifying it.

Thanks,
James

Quoted 2 lines

Best regards

#11 lore raw

PATCH 9/13 auto-t: update start_ap() to remove psk argument

In reply to James Prestwood
---
 autotests/util/iwd.py | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
autotests/util/iwd.py +2 -2
index 9bfc8c1a..c1349bbd 100755
@@ -467,7 +467,7 @@ class Device(IWDDBusAbstract):
                                         error_handler=self._failure)
         self._wait_for_async_op()
 
-    def start_ap(self, ssid, psk):
+    def start_ap(self, ssid):
         try:
             self._prop_proxy.Set(IWD_DEVICE_INTERFACE, 'Mode', 'ap')
         except Exception as e:
@@ -476,7 +476,7 @@ class Device(IWDDBusAbstract):
         self._ap_iface = dbus.Interface(self._bus.get_object(IWD_SERVICE,
                                             self.device_path),
                                             IWD_AP_INTERFACE)
-        self._ap_iface.Start(ssid, psk, reply_handler=self._success,
+        self._ap_iface.Start(ssid, reply_handler=self._success,
                                      error_handler=self._failure)
         self._wait_for_async_op()
 
Signature · 1 line
2.26.2
#12 lore raw

PATCH 10/13 auto-t: update AP tests with provisioning files

In reply to James Prestwood

Also removed some unused config values which are no longer needed after the test-runner re-write.

---
 autotests/testAP-no-support/TestAP2.ap         | 2 ++
 autotests/testAP-no-support/connection_test.py | 6 +++---
 autotests/testAP-no-support/hw.conf            | 2 +-
 autotests/testAP/TestAP2.ap                    | 2 ++
 autotests/testAP/connection_test.py            | 5 +++--
 autotests/testAP/failure_test.py               | 5 +++--
 autotests/testAP/hw.conf                       | 3 +--
 7 files changed, 15 insertions(+), 10 deletions(-)
 create mode 100644 autotests/testAP-no-support/TestAP2.ap
 create mode 100644 autotests/testAP/TestAP2.ap
autotests/testAP-no-support/TestAP2.ap new file +2 -0
new file mode 100644
index 00000000..27f086cb
@@ -0,0 +1,2 @@
+[Security]
+Passphrase=Password2
autotests/testAP-no-support/connection_test.py +3 -3
index 1bbc0bde..ff6fdb25 100644
@@ -9,16 +9,16 @@ from iwd import IWD
 
 class Test(unittest.TestCase):
     def test_connection_success(self):
-        wd = IWD()
+        wd = IWD(True)
 
         dev = wd.list_devices(1)[0]
 
         with self.assertRaises(iwd.NotSupportedEx):
-            dev.start_ap('TestAP2', 'Password2')
+            dev.start_ap('TestAP2')
 
     @classmethod
     def setUpClass(cls):
-        pass
+        IWD.copy_to_ap('TestAP2.ap')
 
     @classmethod
     def tearDownClass(cls):
autotests/testAP-no-support/hw.conf +1 -1
index 9df3b349..93177bb6 100644
@@ -1,6 +1,6 @@
 [SETUP]
 num_radios=1
-radio_confs=rad0
+start_iwd=0
 
 [rad0]
 iftype_disable=ap
autotests/testAP/TestAP2.ap new file +2 -0
new file mode 100644
index 00000000..27f086cb
@@ -0,0 +1,2 @@
+[Security]
+Passphrase=Password2
autotests/testAP/connection_test.py +3 -2
index c3744c55..814ed070 100644
@@ -38,13 +38,13 @@ class Test(unittest.TestCase):
         wd.wait_for_object_condition(ordered_network.network_object, condition)
 
     def test_connection_success(self):
-        wd = IWD()
+        wd = IWD(True)
 
         dev1, dev2 = wd.list_devices(2)
 
         self.client_connect(wd, dev1)
 
-        dev1.start_ap('TestAP2', 'Password2')
+        dev1.start_ap('TestAP2')
 
         try:
             condition = 'not obj.scanning'
@@ -96,6 +96,7 @@ class Test(unittest.TestCase):
     @classmethod
     def setUpClass(cls):
         IWD.copy_to_storage('TestAP1.psk')
+        IWD.copy_to_ap('TestAP2.ap')
 
     @classmethod
     def tearDownClass(cls):
autotests/testAP/failure_test.py +3 -2
index 4959802f..7c3290b0 100644
@@ -39,13 +39,13 @@ class Test(unittest.TestCase):
         wd.wait_for_object_condition(ordered_network.network_object, condition)
 
     def test_connection_failure(self):
-        wd = IWD()
+        wd = IWD(True)
 
         dev1, dev2 = wd.list_devices(2)
 
         self.client_connect(wd, dev1)
 
-        dev1.start_ap('TestAP2', 'Password2')
+        dev1.start_ap('TestAP2')
 
         try:
             condition = 'not obj.scanning'
@@ -77,6 +77,7 @@ class Test(unittest.TestCase):
     @classmethod
     def setUpClass(cls):
         IWD.copy_to_storage('TestAP1.psk')
+        IWD.copy_to_ap('TestAP2.ap')
 
     @classmethod
     def tearDownClass(cls):
autotests/testAP/hw.conf +1 -2
index 96a2bbb1..6b8b2aeb 100644
@@ -1,7 +1,6 @@
 [SETUP]
 num_radios=3
-max_test_exec_interval_sec=40
-tmpfs_extra_stuff=main.conf
+start_iwd=0
 
 [HOSTAPD]
 rad0=psk-ccmp.conf
Signature · 1 line
2.26.2
#13 lore raw

PATCH 11/13 build: add ELL dhcp-util.c to build

In reply to James Prestwood
---
 Makefile.am | 1 +
 1 file changed, 1 insertion(+)
Makefile.am +1 -0
index 1ee414cb..7ed8baa9 100644
@@ -121,6 +121,7 @@ ell_sources = ell/private.h \
 			ell/dhcp-transport.c \
 			ell/dhcp-lease.c \
 			ell/dhcp-util.c \
+			ell/dhcp-server.c \
 			ell/cert-private.h \
 			ell/cert.c \
 			ell/ecc-external.c \
Signature · 1 line
2.26.2
#14 lore raw

PATCH 12/13 ap: add support for DHCPv4 server

In reply to James Prestwood

The DHCP server can be enabled by including an [IPv4] group in the AP provisioning file and including at least a subnet mask.

---
 src/ap.c | 108 +++++++++++++++++++++++++++++++++++++++++++++++++++----
 src/ap.h |   8 +++++
 2 files changed, 110 insertions(+), 6 deletions(-)
src/ap.c +102 -6
index 41de69a9..e9f79e5a 100644
@@ -76,6 +76,8 @@ struct ap_state {
 	uint16_t last_aid;
 	struct l_queue *sta_states;
 
+	struct l_dhcp_server *server;
+
 	bool started : 1;
 	bool gtk_set : 1;
 };
@@ -120,6 +122,13 @@ void ap_config_free(struct ap_config *config)
 	explicit_bzero(config->psk, sizeof(config->psk));
 	l_free(config->authorized_macs);
 	l_free(config->wsc_name);
+
+	l_free(config->address);
+	l_free(config->gateway);
+	l_free(config->netmask);
+	l_strv_free(config->dns_list);
+	l_strv_free(config->ip_range);
+
 	l_free(config);
 }
 
@@ -198,6 +207,9 @@ static void ap_reset(struct ap_state *ap)
 	l_queue_destroy(ap->wsc_pbc_probes, l_free);
 
 	ap->started = false;
+
+	if (ap->server)
+		l_dhcp_server_stop(ap->server);
 }
 
 static void ap_del_station(struct sta_state *sta, uint16_t reason,
@@ -1906,6 +1918,55 @@ static void ap_deauth_cb(const struct mmpdu_header *hdr, const void *body,
 	ap_sta_free(sta);
 }
 
+static bool ap_start_dhcp(struct ap_state *ap)
+{
+	uint32_t ifindex = netdev_get_ifindex(ap->netdev);
+
+	/*
+	 * [IPv4].Netmask is the only required property since .Address and
+	 * .Gateway may inherit from the interfaces IP that was already set.
+	 * If this is not found we assume DHCP is not being used.
+	 */
+	if (!ap->config->netmask)
+		return true;
+
+	ap->server = l_dhcp_server_new(ifindex);
+	if (!ap->server) {
+		l_error("Failed to create DHCP server on %u", ifindex);
+		return false;
+	}
+
+	/* TODO: set address via rtnl */
+	if (ap->config->address)
+		if (!l_dhcp_server_set_ip_address(ap->server,
+							ap->config->address))
+			return false;
+
+	if (ap->config->gateway)
+		if (!l_dhcp_server_set_gateway(ap->server, ap->config->gateway))
+			return false;
+
+	if (ap->config->dns_list)
+		if (!l_dhcp_server_set_dns(ap->server, ap->config->dns_list))
+			return false;
+
+	if (ap->config->lease_time)
+		if (!l_dhcp_server_set_lease_time(ap->server,
+						ap->config->lease_time))
+			return false;
+
+	if (ap->config->ip_range)
+		if (!l_dhcp_server_set_ip_range(ap->server,
+						ap->config->ip_range[0],
+						ap->config->ip_range[1]))
+			return false;
+
+	if (!l_dhcp_server_set_netmask(ap->server, ap->config->netmask))
+		return false;
+
+	return l_dhcp_server_start(ap->server);
+}
+
 static void ap_start_cb(struct l_genl_msg *msg, void *user_data)
 {
 	struct ap_state *ap = user_data;
@@ -1915,16 +1976,24 @@ static void ap_start_cb(struct l_genl_msg *msg, void *user_data)
 	if (l_genl_msg_get_error(msg) < 0) {
 		l_error("START_AP failed: %i", l_genl_msg_get_error(msg));
 
-		ap->ops->handle_event(AP_EVENT_START_FAILED, NULL,
-					ap->user_data);
-		ap_reset(ap);
-		l_genl_family_free(ap->nl80211);
-		l_free(ap);
-		return;
+		goto failed;
+	}
+
+	if (!ap_start_dhcp(ap)) {
+		l_error("DHCP server failed to start");
+		goto failed;
 	}
 
 	ap->started = true;
 	ap->ops->handle_event(AP_EVENT_STARTED, NULL, ap->user_data);
+
+	return;
+
+failed:
+	ap->ops->handle_event(AP_EVENT_START_FAILED, NULL, ap->user_data);
+	ap_reset(ap);
+	l_genl_family_free(ap->nl80211);
+	l_free(ap);
 }
 
 static struct l_genl_msg *ap_build_cmd_start_ap(struct ap_state *ap)
@@ -2563,6 +2632,33 @@ static struct ap_config *ap_config_from_settings(struct l_settings *settings,
 		l_free(passphrase);
 	}
 
+	if (!l_settings_has_group(settings, "IPv4"))
+		goto done;
+
+	config->address = l_settings_get_string(settings, "IPv4", "Address");
+	config->gateway = l_settings_get_string(settings, "IPv4", "Gateway");
+	config->netmask = l_settings_get_string(settings, "IPv4", "Netmask");
+	config->dns_list = l_settings_get_string_list(settings, "IPv4",
+							"DNSList", ',');
+	config->ip_range = l_settings_get_string_list(settings, "IPv4",
+							"IPRange", ',');
+
+	if (!l_settings_get_uint(settings, "IPv4", "LeaseTime",
+					&config->lease_time))
+		config->lease_time = 0;
+
+
+	if (!config->netmask) {
+		l_error("[IPv4].Netmask is a required value for DHCP");
+		goto failed;
+	}
+
+	if (config->ip_range && l_strv_length(config->ip_range) != 2) {
+		l_error("IP range must be 2 values");
+		goto failed;
+	}
+
+
 done:
 	l_info("Loaded AP configuration %s", config->ssid);
 	return config;
src/ap.h +8 -0
index 90497008..277a5a2f 100644
@@ -63,6 +63,14 @@ struct ap_config {
 	unsigned int authorized_macs_num;
 	char *wsc_name;
 	struct wsc_primary_device_type wsc_primary_device_type;
+
+	char *address;
+	char *gateway;
+	char *netmask;
+	char **dns_list;
+	char **ip_range;
+	uint32_t lease_time;
+
 	bool no_cck_rates : 1;
 	bool no_free : 1;
 };
Signature · 1 line
2.26.2
#15 lore raw

Re: [PATCH 12/13] ap: add support for DHCPv4 server

Hi James,

Quoted 3 lines · James Prestwood

The DHCP server can be enabled by including an [IPv4] group in the AP provisioning file and including at least a subnet mask.

Any chance we can add this code to the current implementation without changing the API just yet? I think it is safe to assume that if the address is set prior to AP starting on a given interface, then we should not start our own DHCP, and if the address isn't set, then start DHCP server. If the address is provided by provisioning, then start our own DHCP as well, overriding whatever we picked.

Quoted 44 lines
---
  src/ap.c | 108 +++++++++++++++++++++++++++++++++++++++++++++++++++----
  src/ap.h |   8 +++++
  2 files changed, 110 insertions(+), 6 deletions(-)
src/ap.c +12 -0
index 41de69a9..e9f79e5a 100644
@@ -76,6 +76,8 @@ struct ap_state {
  	uint16_t last_aid;
  	struct l_queue *sta_states;
 
+	struct l_dhcp_server *server;
+
  	bool started : 1;
  	bool gtk_set : 1;
  };
@@ -120,6 +122,13 @@ void ap_config_free(struct ap_config *config)
  	explicit_bzero(config->psk, sizeof(config->psk));
  	l_free(config->authorized_macs);
  	l_free(config->wsc_name);
+
+	l_free(config->address);
+	l_free(config->gateway);
+	l_free(config->netmask);
+	l_strv_free(config->dns_list);
+	l_strv_free(config->ip_range);
+
  	l_free(config);
  }
 
@@ -198,6 +207,9 @@ static void ap_reset(struct ap_state *ap)
  	l_queue_destroy(ap->wsc_pbc_probes, l_free);
 
  	ap->started = false;
+
+	if (ap->server)
+		l_dhcp_server_stop(ap->server);
  }
 
  static void ap_del_station(struct sta_state *sta, uint16_t reason,
@@ -1906,6 +1918,55 @@ static void ap_deauth_cb(const struct mmpdu_header *hdr, const void *body,
  	ap_sta_free(sta);

src/ap.c static void ap_deauth_cb(const struct mmpdu_header *hdr, const void *body, · line 1928

Quoted 10 lines
src/ap.c +8 -0
  }
 
+static bool ap_start_dhcp(struct ap_state *ap)
+{
+	uint32_t ifindex = netdev_get_ifindex(ap->netdev);
+
+	/*
+	 * [IPv4].Netmask is the only required property since .Address and
+	 * .Gateway may inherit from the interfaces IP that was already set.
+	 * If this is not found we assume DHCP is not being used.

Netmask is also inherited actually. Its the prefix len...

denkenz(a)localhost ~/iwd-master $ ip addr
...
inet 192.168.1.249/24 brd 192.168.1.255 scope global dynamic noprefixroute
...

/24 = 255.255.255.0

Quoted 115 lines
src/ap.c +82 -6
+	 */
+	if (!ap->config->netmask)
+		return true;
+
+	ap->server = l_dhcp_server_new(ifindex);
+	if (!ap->server) {
+		l_error("Failed to create DHCP server on %u", ifindex);
+		return false;
+	}
+
+	/* TODO: set address via rtnl */
+	if (ap->config->address)
+		if (!l_dhcp_server_set_ip_address(ap->server,
+							ap->config->address))
+			return false;
+
+	if (ap->config->gateway)
+		if (!l_dhcp_server_set_gateway(ap->server, ap->config->gateway))
+			return false;
+
+	if (ap->config->dns_list)
+		if (!l_dhcp_server_set_dns(ap->server, ap->config->dns_list))
+			return false;
+
+	if (ap->config->lease_time)
+		if (!l_dhcp_server_set_lease_time(ap->server,
+						ap->config->lease_time))
+			return false;
+
+	if (ap->config->ip_range)
+		if (!l_dhcp_server_set_ip_range(ap->server,
+						ap->config->ip_range[0],
+						ap->config->ip_range[1]))
+			return false;
+
+	if (!l_dhcp_server_set_netmask(ap->server, ap->config->netmask))
+		return false;
+
+	return l_dhcp_server_start(ap->server);
+}
+
  static void ap_start_cb(struct l_genl_msg *msg, void *user_data)
  {
  	struct ap_state *ap = user_data;
@@ -1915,16 +1976,24 @@ static void ap_start_cb(struct l_genl_msg *msg, void *user_data)
  	if (l_genl_msg_get_error(msg) < 0) {
  		l_error("START_AP failed: %i", l_genl_msg_get_error(msg));
 
-		ap->ops->handle_event(AP_EVENT_START_FAILED, NULL,
-					ap->user_data);
-		ap_reset(ap);
-		l_genl_family_free(ap->nl80211);
-		l_free(ap);
-		return;
+		goto failed;
+	}
+
+	if (!ap_start_dhcp(ap)) {
+		l_error("DHCP server failed to start");
+		goto failed;
  	}
 
  	ap->started = true;
  	ap->ops->handle_event(AP_EVENT_STARTED, NULL, ap->user_data);
+
+	return;
+
+failed:
+	ap->ops->handle_event(AP_EVENT_START_FAILED, NULL, ap->user_data);
+	ap_reset(ap);
+	l_genl_family_free(ap->nl80211);
+	l_free(ap);
  }
 
  static struct l_genl_msg *ap_build_cmd_start_ap(struct ap_state *ap)
@@ -2563,6 +2632,33 @@ static struct ap_config *ap_config_from_settings(struct l_settings *settings,
  		l_free(passphrase);
  	}
 
+	if (!l_settings_has_group(settings, "IPv4"))
+		goto done;
+
+	config->address = l_settings_get_string(settings, "IPv4", "Address");
+	config->gateway = l_settings_get_string(settings, "IPv4", "Gateway");
+	config->netmask = l_settings_get_string(settings, "IPv4", "Netmask");
+	config->dns_list = l_settings_get_string_list(settings, "IPv4",
+							"DNSList", ',');
+	config->ip_range = l_settings_get_string_list(settings, "IPv4",
+							"IPRange", ',');
+
+	if (!l_settings_get_uint(settings, "IPv4", "LeaseTime",
+					&config->lease_time))
+		config->lease_time = 0;
+
+
+	if (!config->netmask) {
+		l_error("[IPv4].Netmask is a required value for DHCP");
+		goto failed;
+	}
+
+	if (config->ip_range && l_strv_length(config->ip_range) != 2) {
+		l_error("IP range must be 2 values");
+		goto failed;
+	}
+
+
  done:
  	l_info("Loaded AP configuration %s", config->ssid);
  	return config;
src/ap.h +0 -0
index 90497008..277a5a2f 100644
@@ -63,6 +63,14 @@ struct ap_config {
  	unsigned int authorized_macs_num;

src/ap.h struct ap_config { · line 73

Quoted 10 lines
src/ap.h +8 -0
  	char *wsc_name;
  	struct wsc_primary_device_type wsc_primary_device_type;
+
+	char *address;
+	char *gateway;
+	char *netmask;
+	char **dns_list;
+	char **ip_range;
+	uint32_t lease_time;
+

I think we may need Andrew's opinion on whether these should be part of ap_config. P2P might not really care about setting up these variables and would prefer for ap.c to just figure things out.

src/ap.h struct ap_config { · line 76

Quoted 4 lines
src/ap.h +0 -0
  	bool no_cck_rates : 1;
  	bool no_free : 1;
  };

Regards,
-Denis

#16 lore raw

Re: [PATCH 12/13] ap: add support for DHCPv4 server

Quoted 7 lines · Denis Kenzior

Hi James,

Quoted 3 lines · James Prestwood

The DHCP server can be enabled by including an [IPv4] group in the AP provisioning file and including at least a subnet mask.

Quoted 10 lines

Any chance we can add this code to the current implementation without changing
the API just yet? I think it is safe to assume that if the address is set prior
to AP starting on a given interface, then we should not start our own DHCP, and
if the address isn't set, then start DHCP server. If the address is provided by
provisioning, then start our own DHCP as well, overriding whatever we picked.

Sure, so keep the 'psk' DBus argument and remove [Security].Passphrase? It seems weird to have both since we would be ignoring one.

But do we really want to make the decision to use DHCP based on if the address is set? It really seems like the user would want to decide that explicitly with a config file. OTOH if all no DHCP options are required (using defaults) we would need some way of knowing to start DHCP or not. I just don't see a correlation between the address being set and the user wanting/not wanting a DHCP server.

Quoted 63 lines
Quoted 59 lines
---
  src/ap.c | 108
+++++++++++++++++++++++++++++++++++++++++++++++++++----
  src/ap.h |   8 +++++
  2 files changed, 110 insertions(+), 6 deletions(-)
src/ap.c +12 -0
index 41de69a9..e9f79e5a 100644
@@ -76,6 +76,8 @@ struct ap_state {
  	uint16_t last_aid;
  	struct l_queue *sta_states;
 
+	struct l_dhcp_server *server;
+
  	bool started : 1;
  	bool gtk_set : 1;
  };
@@ -120,6 +122,13 @@ void ap_config_free(struct ap_config *config)
  	explicit_bzero(config->psk, sizeof(config->psk));
  	l_free(config->authorized_macs);
  	l_free(config->wsc_name);
+
+	l_free(config->address);
+	l_free(config->gateway);
+	l_free(config->netmask);
+	l_strv_free(config->dns_list);
+	l_strv_free(config->ip_range);
+
  	l_free(config);
  }
 
@@ -198,6 +207,9 @@ static void ap_reset(struct ap_state *ap)
  	l_queue_destroy(ap->wsc_pbc_probes, l_free);
 
  	ap->started = false;
+
+	if (ap->server)
+		l_dhcp_server_stop(ap->server);
  }
 
  static void ap_del_station(struct sta_state *sta, uint16_t

reason,

src/ap.c +6 -0
@@ -1906,6 +1918,55 @@ static void ap_deauth_cb(const struct
 mmpdu_header *hdr, const void *body,
  	ap_sta_free(sta);
  }
 
+static bool ap_start_dhcp(struct ap_state *ap)
+{
+	uint32_t ifindex = netdev_get_ifindex(ap->netdev);
+
+	/*
+	 * [IPv4].Netmask is the only required property since .Address

and

src/ap.c +1 -0
+	 * .Gateway may inherit from the interfaces IP that was already

set.

src/ap.c static void ap_deauth_cb(const struct mmpdu_header *hdr, const void *body, · line 1928

Quoted 1 line
src/ap.c +1 -0
+	 * If this is not found we assume DHCP is not being used.

Netmask is also inherited actually. Its the prefix len...

Quoted 7 lines

denkenz(a)localhost ~/iwd-master $ ip addr
...
inet 192.168.1.249/24 brd 192.168.1.255 scope global dynamic noprefixroute
...

/24 = 255.255.255.0

Ok I'll make this optional as well.

Quoted 141 lines
Quoted 141 lines
src/ap.c +14 -0
+	 */
+	if (!ap->config->netmask)
+		return true;
+
+	ap->server = l_dhcp_server_new(ifindex);
+	if (!ap->server) {
+		l_error("Failed to create DHCP server on %u", ifindex);
+		return false;
+	}
+
+	/* TODO: set address via rtnl */
+	if (ap->config->address)
+		if (!l_dhcp_server_set_ip_address(ap->server,
+							ap->config-
Quoted 1 line

address))

src/ap.c +4 -0
+			return false;
+
+	if (ap->config->gateway)
+		if (!l_dhcp_server_set_gateway(ap->server, ap->config-
Quoted 1 line

gateway))

src/ap.c +4 -0
+			return false;
+
+	if (ap->config->dns_list)
+		if (!l_dhcp_server_set_dns(ap->server, ap->config-
Quoted 1 line

dns_list))

src/ap.c +5 -0
+			return false;
+
+	if (ap->config->lease_time)
+		if (!l_dhcp_server_set_lease_time(ap->server,
+						ap->config-
Quoted 1 line

lease_time))

src/ap.c +5 -0
+			return false;
+
+	if (ap->config->ip_range)
+		if (!l_dhcp_server_set_ip_range(ap->server,
+						ap->config-
Quoted 1 line

ip_range[0],

src/ap.c +1 -0
+						ap->config-
Quoted 1 line

ip_range[1]))

src/ap.c +3 -0
+			return false;
+
+	if (!l_dhcp_server_set_netmask(ap->server, ap->config-
Quoted 1 line

netmask))

src/ap.c +19 -6
+		return false;
+
+	return l_dhcp_server_start(ap->server);
+}
+
  static void ap_start_cb(struct l_genl_msg *msg, void *user_data)
  {
  	struct ap_state *ap = user_data;
@@ -1915,16 +1976,24 @@ static void ap_start_cb(struct l_genl_msg
 *msg, void *user_data)
  	if (l_genl_msg_get_error(msg) < 0) {
  		l_error("START_AP failed: %i",
 l_genl_msg_get_error(msg));
 
-		ap->ops->handle_event(AP_EVENT_START_FAILED, NULL,
-					ap->user_data);
-		ap_reset(ap);
-		l_genl_family_free(ap->nl80211);
-		l_free(ap);
-		return;
+		goto failed;
+	}
+
+	if (!ap_start_dhcp(ap)) {
+		l_error("DHCP server failed to start");
+		goto failed;
  	}
 
  	ap->started = true;
  	ap->ops->handle_event(AP_EVENT_STARTED, NULL, ap->user_data);
+
+	return;
+
+failed:
+	ap->ops->handle_event(AP_EVENT_START_FAILED, NULL, ap-
 >user_data);
+	ap_reset(ap);
+	l_genl_family_free(ap->nl80211);
+	l_free(ap);
  }

static struct l_genl_msg *ap_build_cmd_start_ap(struct ap_state
*ap)

src/ap.c +6 -0
@@ -2563,6 +2632,33 @@ static struct ap_config
 *ap_config_from_settings(struct l_settings *settings,
  		l_free(passphrase);
  	}
 
+	if (!l_settings_has_group(settings, "IPv4"))
+		goto done;
+
+	config->address = l_settings_get_string(settings, "IPv4",
 "Address");
+	config->gateway = l_settings_get_string(settings, "IPv4",
 "Gateway");
+	config->netmask = l_settings_get_string(settings, "IPv4",

"Netmask");

src/ap.c +2 -0
+	config->dns_list = l_settings_get_string_list(settings, "IPv4",
+							"DNSList",

',');

src/ap.c +2 -0
+	config->ip_range = l_settings_get_string_list(settings, "IPv4",
+							"IPRange",

',');

src/ap.c +17 -0
+
+	if (!l_settings_get_uint(settings, "IPv4", "LeaseTime",
+					&config->lease_time))
+		config->lease_time = 0;
+
+
+	if (!config->netmask) {
+		l_error("[IPv4].Netmask is a required value for DHCP");
+		goto failed;
+	}
+
+	if (config->ip_range && l_strv_length(config->ip_range) != 2) {
+		l_error("IP range must be 2 values");
+		goto failed;
+	}
+
+
  done:
  	l_info("Loaded AP configuration %s", config->ssid);
  	return config;
src/ap.h +6 -0
index 90497008..277a5a2f 100644
@@ -63,6 +63,14 @@ struct ap_config {
  	unsigned int authorized_macs_num;
  	char *wsc_name;
  	struct wsc_primary_device_type wsc_primary_device_type;
+
+	char *address;
+	char *gateway;
+	char *netmask;
+	char **dns_list;
+	char **ip_range;
Quoted 8 lines

src/ap.h struct ap_config { · line 73

Quoted 2 lines
src/ap.h +2 -0
+	uint32_t lease_time;
+

I think we may need Andrew's opinion on whether these should be part of
ap_config. P2P might not really care about setting up these variables and would
prefer for ap.c to just figure things out.

Sure, but they aren't required anyways so p2p doesn't have to include them and defaults will be used (once netmask is fixed). The ap_config structure was just a convenient storage object for these but we could define something else too.

Quoted 8 lines

src/ap.h struct ap_config { · line 76

Quoted 5 lines
src/ap.h +0 -0
  	bool no_cck_rates : 1;
  	bool no_free : 1;
  };

Regards,
-Denis

Thanks,
James

#17 lore raw

Re: [PATCH 12/13] ap: add support for DHCPv4 server

Hi James,

Quoted 19 lines · James Prestwood
Quoted 7 lines · Denis Kenzior

Hi James,

Quoted 3 lines · James Prestwood

The DHCP server can be enabled by including an [IPv4] group in the AP provisioning file and including at least a subnet mask.

Quoted 10 lines

Any chance we can add this code to the current implementation without changing
the API just yet? I think it is safe to assume that if the address is set prior
to AP starting on a given interface, then we should not start our own DHCP, and
if the address isn't set, then start DHCP server. If the address is provided by
provisioning, then start our own DHCP as well, overriding whatever we picked.

Quoted 2 lines

Sure, so keep the 'psk' DBus argument and remove [Security].Passphrase? It seems weird to have both since we would be ignoring one.

I don't know yet. I don't think removing the simple 'Start' method which just provides an SSID and PSK is a good idea. It is useful for quick and dirty setups. Provisioning files might be for more advanced users and might need their own dedicated Start method. So something like KnownNetwork(s) but for AccessPointConfiguration(s).

Quoted 7 lines

But do we really want to make the decision to use DHCP based on if the address is set? It really seems like the user would want to decide that explicitly with a config file. OTOH if all no DHCP options are required (using defaults) we would need some way of knowing to start DHCP or not. I just don't see a correlation between the address being set and the user wanting/not wanting a DHCP server.

Well, the user has to first switch the device into AP mode or create the required interface first somehow. So before they actually hit .Start, they have a chance to configure the IP or not. If the IP is configured, then we assume the interface is managed by someone else. If it isn't, we take ownership.

I suppose we can add an explicit setting in main.conf somewhere, or in the profile itself. But I'd keep things simple for now.

Quoted 11 lines
Quoted 6 lines

I think we may need Andrew's opinion on whether these should be part of
ap_config. P2P might not really care about setting up these variables and would
prefer for ap.c to just figure things out.

Sure, but they aren't required anyways so p2p doesn't have to include them and defaults will be used (once netmask is fixed). The ap_config structure was just a convenient storage object for these but we could define something else too.

Yes, true. But not much sense in actually storing them long-term if that's the case. Generate them / load from settings, give them to dhcp-server and forget.

Regards,
-Denis

#18 lore raw

Re: [PATCH 12/13] ap: add support for DHCPv4 server

Hi,

Quoted 23 lines · Denis Kenzior

Hi James,

Quoted 19 lines · James Prestwood
Quoted 17 lines · Denis Kenzior

Hi James,

Quoted 3 lines · James Prestwood

The DHCP server can be enabled by including an [IPv4] group in the AP provisioning file and including at least a subnet mask.

Any chance we can add this code to the current implementation without changing
the API just yet? I think it is safe to assume that if the address is set prior
to AP starting on a given interface, then we should not start our own DHCP, and
if the address isn't set, then start DHCP server. If the address is provided by
provisioning, then start our own DHCP as well, overriding whatever we picked.

Quoted 8 lines
Quoted 2 lines

Sure, so keep the 'psk' DBus argument and remove [Security].Passphrase? It seems weird to have both since we would be ignoring one.

I don't know yet. I don't think removing the simple 'Start' method which just provides an SSID and PSK is a good idea. It is useful for quick and dirty setups. Provisioning files might be for more advanced users and might need their own dedicated Start method. So something like KnownNetwork(s) but for AccessPointConfiguration(s).

I replied to a different patch in this series without seeing this thread, but basically I agree with what Denis said, I prefer if we keep the DBus arguments unchanged.

Quoted 9 lines
Quoted 8 lines

But do we really want to make the decision to use DHCP based on if the address is set? It really seems like the user would want to decide that explicitly with a config file. OTOH if all no DHCP options are required (using defaults) we would need some way of knowing to start DHCP or not. I just don't see a correlation between the address being set and the user wanting/not wanting a DHCP server.

Quoted 7 lines

Well, the user has to first switch the device into AP mode or create the required interface first somehow. So before they actually hit .Start, they have a chance to configure the IP or not. If the IP is configured, then we assume the interface is managed by someone else. If it isn't, we take ownership.

I suppose we can add an explicit setting in main.conf somewhere, or in the profile itself. But I'd keep things simple for now.

Maybe we can re-use [General].EnableNetworkConfiguration for the basic on/off switch and possibly allow individual access point config files to override this?

In the case of Network Manager users, if they're interfacing with IWD through NM they will currently not want DHCP as either client or server, and if they use iwctl they might want DHCP everywhere.

Also for the record, NM only sets the local IP after the AP is started.

Quoted 8 lines
Quoted 8 lines
Quoted 7 lines

I think we may need Andrew's opinion on whether these should be part of
ap_config. P2P might not really care about setting up these variables and would
prefer for ap.c to just figure things out.

Quoted 7 lines
Quoted 4 lines

Sure, but they aren't required anyways so p2p doesn't have to include them and defaults will be used (once netmask is fixed). The ap_config structure was just a convenient storage object for these but we could define something else too.

Yes, true. But not much sense in actually storing them long-term if that's the case. Generate them / load from settings, give them to dhcp-server and forget.

It does seem logical to put those settings in struct ap_config. P2P wouldn't normally need to use them, it should be fine with the defaults although I'm thinking in the future there may be exceptions where we need to work around client's bugs (hopefully not).

Note that the P2P Group Owner will need to decide on an IP for itself and the client before the 4-Way Handshake starts to implement the optional IP assignment during the handshake.

Best regards

#19 lore raw

Re: [PATCH 12/13] ap: add support for DHCPv4 server

Hi Andrew,

Quoted 12 lines
Quoted 7 lines

Well, the user has to first switch the device into AP mode or create the required interface first somehow. So before they actually hit .Start, they have a chance to configure the IP or not. If the IP is configured, then we assume the interface is managed by someone else. If it isn't, we take ownership.

I suppose we can add an explicit setting in main.conf somewhere, or in the profile itself. But I'd keep things simple for now.

Maybe we can re-use [General].EnableNetworkConfiguration for the basic on/off switch and possibly allow individual access point config files to override this?

Quoted 6 lines

In the case of Network Manager users, if they're interfacing with IWD through NM they will currently not want DHCP as either client or server, and if they use iwctl they might want DHCP everywhere.

Also for the record, NM only sets the local IP after the AP is started.

Good points. I'm fine re-using EnableNetworkConfiguration for this.

Quoted 16 lines
Quoted 8 lines
Quoted 8 lines
Quoted 7 lines

I think we may need Andrew's opinion on whether these should be part of
ap_config. P2P might not really care about setting up these variables and would
prefer for ap.c to just figure things out.

Quoted 7 lines
Quoted 4 lines

Sure, but they aren't required anyways so p2p doesn't have to include them and defaults will be used (once netmask is fixed). The ap_config structure was just a convenient storage object for these but we could define something else too.

Yes, true. But not much sense in actually storing them long-term if that's the case. Generate them / load from settings, give them to dhcp-server and forget.

Quoted 8 lines

It does seem logical to put those settings in struct ap_config. P2P wouldn't normally need to use them, it should be fine with the defaults although I'm thinking in the future there may be exceptions where we need to work around client's bugs (hopefully not).

Note that the P2P Group Owner will need to decide on an IP for itself and the client before the 4-Way Handshake starts to implement the optional IP assignment during the handshake.

Hmm, wouldn't it make sense for AP to figure out the setup and for P2P to just use that? Especially if we start supporting persistent groups eventually.

You do bring up a good point that we'd need a 'reserve' API of sorts on the DHCP server side. That way IP details for the handshaking client can be pre-populated if IP configuration through the handshake is supported.

Regards,
-Denis

#20 lore raw

Re: [PATCH 12/13] ap: add support for DHCPv4 server

Quoted 19 lines · Denis Kenzior
Quoted 19 lines
Quoted 6 lines
Quoted 6 lines
Quoted 5 lines

I think we may need Andrew's opinion on whether these should be part of
ap_config. P2P might not really care about setting up these variables and would
prefer for ap.c to just figure things out.

Quoted 7 lines
Quoted 4 lines

Sure, but they aren't required anyways so p2p doesn't have to include them and defaults will be used (once netmask is fixed). The ap_config structure was just a convenient storage object for these but we could define something else too.

Yes, true. But not much sense in actually storing them long-term if that's the case. Generate them / load from settings, give them to dhcp-server and forget.

It does seem logical to put those settings in struct ap_config. P2P wouldn't normally need to use them, it should be fine with the defaults although I'm thinking in the future there may be exceptions where we need to work around client's bugs (hopefully not).

Quoted 6 lines
Quoted 3 lines

Note that the P2P Group Owner will need to decide on an IP for itself and the client before the 4-Way Handshake starts to implement the optional IP assignment during the handshake.

Hmm, wouldn't it make sense for AP to figure out the setup and for P2P to just use that? Especially if we start supporting persistent groups eventually.

Yeah, normally P2P doesn't need to fix specific IP vaues, it can let AP use its defaults.

Even with persistent groups afaik mainly the PSK is persisted, IPs can differ between each session.

Quoted 4 lines

You do bring up a good point that we'd need a 'reserve' API of sorts on the DHCP server side. That way IP details for the handshaking client can be pre-populated if IP configuration through the handshake is supported.

Usually we'll just have one client. Even with multiple clients we can use disjoint ranges of IPs for DHCP and for the 4-Way handshake IP setup... but we could instead 'reserve' an IP for the client and offer it through both methods.

Many standalone APs let you manually assign IPs to MACs, including IPs in the DHCP range, so it may be useful to have this option.

Best regards

#21 lore raw

PATCH 13/13 auto-t: add AP test with DHCP server

In reply to James Prestwood
---
 autotests/testAP-DHCP/TestAP2.ap         | 31 ++++++++++
 autotests/testAP-DHCP/connection_test.py | 76 ++++++++++++++++++++++++
 autotests/testAP-DHCP/failure_test.py    | 53 +++++++++++++++++
 autotests/testAP-DHCP/hw.conf            |  3 +
 autotests/testAP-DHCP/main.conf          |  5 ++
 5 files changed, 168 insertions(+)
 create mode 100644 autotests/testAP-DHCP/TestAP2.ap
 create mode 100644 autotests/testAP-DHCP/connection_test.py
 create mode 100644 autotests/testAP-DHCP/failure_test.py
 create mode 100644 autotests/testAP-DHCP/hw.conf
 create mode 100644 autotests/testAP-DHCP/main.conf
autotests/testAP-DHCP/TestAP2.ap new file +31 -0
new file mode 100644
index 00000000..dc1a3579
@@ -0,0 +1,31 @@
+[Security]
+Passphrase=Password2
+
+[IPv4]
+#
+# IP address of access point interface. If omitted the current interface IP
+# address will be used. If provided the IP will be changed to this address.
+#
+Address=192.168.1.1
+Netmask=255.255.255.0
+
+#
+# Gateway address for AP. If omitted [IPv4].Address will be used.
+#
+Gateway=192.168.1.1
+
+#
+# Optional list of DNS servers
+#
+DNSList=192.168.1.1,192.168.1.254
+
+#
+# Required IP range
+#
+IPRange=192.168.1.2,192.168.1.100
+
+#
+# Optional lease time. If omitted an infinite lease will be used.
+#
+LeaseTime=10
+
autotests/testAP-DHCP/connection_test.py new file +76 -0

76 lines not on this page · raw message

autotests/testAP-DHCP/failure_test.py new file +53 -0

53 lines not on this page · raw message

autotests/testAP-DHCP/hw.conf new file +3 -0
new file mode 100644
index 00000000..c6553537
@@ -0,0 +1,3 @@
+[SETUP]
+num_radios=2
+start_iwd=0
autotests/testAP-DHCP/main.conf new file +5 -0
new file mode 100644
index 00000000..25ebe4ff
@@ -0,0 +1,5 @@
+[Scan]
+DisableMacAddressRandomization=true
+
+[General]
+EnableNetworkConfiguration=true
Signature · 1 line
2.26.2
#22 lore raw
In reply to James Prestwood

Hi James,

Quoted 7 lines · James Prestwood

If a test does not need any hostapd instances but still loads hostapd.py for some reason we want to gracefully throw an exception rather than fail in some other manor.

---
  autotests/util/hostapd.py | 7 +++++++
  1 file changed, 7 insertions(+)

Patch 1 & 11 applied, thanks.

Regards,
-Denis