Beniamino Galvani 2018-05-06 13:49:53 +02:00
parent 9fa973f15f
commit e4a9ce242d
2 changed files with 407 additions and 1 deletions

View File

@ -0,0 +1,401 @@
From cc598e315eb264b44bd4b99db4f0a0f8e95111de Mon Sep 17 00:00:00 2001
From: Beniamino Galvani <bgalvani@redhat.com>
Date: Wed, 7 Mar 2018 15:27:44 +0100
Subject: [PATCH] dhcp: handle expiry by letting the client continue for some
time
Previously we would kill the client when the lease expired and we
restarted it 3 times at 2 minutes intervals before failing the
connection. If the client is killed after it received a NACK from the
server, it doesn't have the chance to delete the lease file and the
next time it is started it will request the same lease again.
Also, the previous restart logic is a bit convoluted.
Since clients already know how to deal with NACKs, let them continue
for a grace period after the expiry. When the grace period ends, we
fail the method and this can either fail the whole connection or keep
it active depending on the may-fail configuration.
https://bugzilla.gnome.org/show_bug.cgi?id=783391
(cherry picked from commit 17009ed91da8b3e0b10ee7e94d220be9bd3fa84c)
(cherry picked from commit 7fbbe7ebee99785e38d39c37e515a64a28edef0f)
---
src/devices/nm-device.c | 220 ++++++++++++++++++++----------------------------
1 file changed, 92 insertions(+), 128 deletions(-)
diff --git a/src/devices/nm-device.c b/src/devices/nm-device.c
index c3bba5206..84f1e1dae 100644
--- a/src/devices/nm-device.c
+++ b/src/devices/nm-device.c
@@ -83,9 +83,8 @@ _LOG_DECLARE_SELF (NMDevice);
/*****************************************************************************/
-#define DHCP_RESTART_TIMEOUT 120
-#define DHCP_NUM_TRIES_MAX 3
#define DEFAULT_AUTOCONNECT TRUE
+#define DHCP_GRACE_PERIOD_SEC 480
#define CARRIER_WAIT_TIME_MS 6000
#define CARRIER_WAIT_TIME_AFTER_MTU_MS 10000
@@ -387,10 +386,9 @@ typedef struct _NMDevicePrivate {
NMDhcpClient * client;
gulong state_sigid;
NMDhcp4Config * config;
- guint restart_id;
- guint num_tries_left;
char * pac_url;
bool was_active;
+ guint grace_id;
} dhcp4;
struct {
@@ -461,10 +459,9 @@ typedef struct _NMDevicePrivate {
NMIP6Config * ip6_config;
/* Event ID of the current IP6 config from DHCP */
char * event_id;
- guint restart_id;
- guint num_tries_left;
guint needed_prefixes;
bool was_active;
+ guint grace_id;
} dhcp6;
gboolean needs_ip6_subnet;
@@ -556,7 +553,6 @@ static void realize_start_setup (NMDevice *self,
NMUnmanFlagOp unmanaged_user_explicit);
static void _set_mtu (NMDevice *self, guint32 mtu);
static void _commit_mtu (NMDevice *self, const NMIP4Config *config);
-static void dhcp_schedule_restart (NMDevice *self, int addr_family, const char *reason);
static void _cancel_activation (NMDevice *self);
/*****************************************************************************/
@@ -5898,7 +5894,7 @@ dhcp4_cleanup (NMDevice *self, CleanupType cleanup_type, gboolean release)
{
NMDevicePrivate *priv = NM_DEVICE_GET_PRIVATE (self);
- nm_clear_g_source (&priv->dhcp4.restart_id);
+ nm_clear_g_source (&priv->dhcp4.grace_id);
g_clear_pointer (&priv->dhcp4.pac_url, g_free);
if (priv->dhcp4.client) {
@@ -6043,20 +6039,17 @@ dhcp4_lease_change (NMDevice *self, NMIP4Config *config)
}
static gboolean
-dhcp4_restart_cb (gpointer user_data)
+dhcp4_grace_period_expired (gpointer user_data)
{
NMDevice *self = user_data;
- NMDevicePrivate *priv;
-
- g_return_val_if_fail (NM_IS_DEVICE (self), FALSE);
- priv = NM_DEVICE_GET_PRIVATE (self);
- priv->dhcp4.restart_id = 0;
+ _LOGI (LOGD_DHCP4, "DHCPv4: grace period expired");
- if (dhcp4_start (self) == NM_ACT_STAGE_RETURN_FAILURE)
- dhcp_schedule_restart (self, AF_INET, NULL);
+ nm_device_ip_method_failed (self, AF_INET,
+ NM_DEVICE_STATE_REASON_IP_CONFIG_EXPIRED);
+ /* If the device didn't fail, the DHCP client will continue */
- return FALSE;
+ return G_SOURCE_REMOVE;
}
static void
@@ -6064,44 +6057,48 @@ dhcp4_fail (NMDevice *self, gboolean timeout)
{
NMDevicePrivate *priv = NM_DEVICE_GET_PRIVATE (self);
- _LOGD (LOGD_DHCP4, "DHCPv4 failed: timeout %d, num tries left %u",
- timeout, priv->dhcp4.num_tries_left);
+ _LOGD (LOGD_DHCP4, "DHCPv4 failed%s", timeout ? " (timeout)" : "");
- dhcp4_cleanup (self, CLEANUP_TYPE_DECONFIGURE, FALSE);
-
- /* Don't fail if there are static addresses configured on
- * the device, instead retry after some time.
+ /* Keep client running if there are static addresses configured
+ * on the interface.
*/
if ( priv->ip4_state == IP_DONE
&& priv->con_ip4_config
- && nm_ip4_config_get_num_addresses (priv->con_ip4_config) > 0) {
- dhcp_schedule_restart (self, AF_INET, "device has IP addresses");
+ && nm_ip4_config_get_num_addresses (priv->con_ip4_config) > 0)
+ goto clear_config;
+
+ /* Fail the method in case of timeout or failure during initial
+ * configuration.
+ */
+ if ( !priv->dhcp4.was_active
+ && (timeout || priv->ip4_state == IP_CONF)) {
+ dhcp4_cleanup (self, CLEANUP_TYPE_DECONFIGURE, FALSE);
+ nm_device_activate_schedule_ip4_config_timeout (self);
return;
}
- if ( priv->dhcp4.num_tries_left == DHCP_NUM_TRIES_MAX
- && (timeout || (priv->ip4_state == IP_CONF))
- && !priv->dhcp4.was_active)
- nm_device_activate_schedule_ip4_config_timeout (self);
- else if ( priv->dhcp4.num_tries_left < DHCP_NUM_TRIES_MAX
- || priv->ip4_state == IP_DONE
- || priv->dhcp4.was_active) {
- /* Don't fail immediately when the lease expires but try to
- * restart DHCP for a predefined number of times.
- */
- if (priv->dhcp4.num_tries_left) {
- priv->dhcp4.num_tries_left--;
- dhcp_schedule_restart (self, AF_INET, "lease expired");
- } else {
- nm_device_ip_method_failed (self, AF_INET, NM_DEVICE_STATE_REASON_IP_CONFIG_EXPIRED);
- /* We failed the ipv4 method but schedule again the retries if the ipv6 method is
- * configured, keeping the connection up.
- */
- if (nm_device_get_state (self) != NM_DEVICE_STATE_FAILED)
- dhcp_schedule_restart (self, AF_INET, "renewal failed");
- }
- } else
- g_warn_if_reached ();
+ /* In any other case (expired lease, assumed connection, etc.),
+ * start a grace period in which we keep the client running,
+ * hoping that it will regain a lease.
+ */
+ if (!priv->dhcp4.grace_id) {
+ priv->dhcp4.grace_id = g_timeout_add_seconds (DHCP_GRACE_PERIOD_SEC,
+ dhcp4_grace_period_expired,
+ self);
+ _LOGI (LOGD_DHCP4,
+ "DHCPv4: %u seconds grace period started",
+ DHCP_GRACE_PERIOD_SEC);
+ goto clear_config;
+ }
+ return;
+
+clear_config:
+ /* The previous configuration is no longer valid */
+ if (priv->dhcp4.config) {
+ nm_exported_object_clear_and_unexport (&priv->dhcp4.config);
+ priv->dhcp4.config = nm_dhcp4_config_new ();
+ _notify (self, PROP_DHCP4_CONFIG);
+ }
}
static void
@@ -6141,6 +6138,8 @@ dhcp4_state_changed (NMDhcpClient *client,
break;
}
+ nm_clear_g_source (&priv->dhcp4.grace_id);
+
/* After some failures, we have been able to renew the lease:
* update the ip state
*/
@@ -6153,7 +6152,6 @@ dhcp4_state_changed (NMDhcpClient *client,
nm_dhcp4_config_set_options (priv->dhcp4.config, options);
_notify (self, PROP_DHCP4_CONFIG);
- priv->dhcp4.num_tries_left = DHCP_NUM_TRIES_MAX;
if (priv->ip4_state == IP_CONF) {
connection = nm_device_get_applied_connection (self);
@@ -6524,7 +6522,6 @@ act_stage3_ip4_config_start (NMDevice *self,
}
method = nm_utils_get_ip_config_method (connection, NM_TYPE_SETTING_IP4_CONFIG);
- priv->dhcp4.num_tries_left = DHCP_NUM_TRIES_MAX;
/* Start IPv4 addressing based on the method requested */
if (strcmp (method, NM_SETTING_IP4_CONFIG_METHOD_AUTO) == 0) {
@@ -6579,7 +6576,7 @@ dhcp6_cleanup (NMDevice *self, CleanupType cleanup_type, gboolean release)
priv->dhcp6.mode = NM_NDISC_DHCP_LEVEL_NONE;
g_clear_object (&priv->dhcp6.ip6_config);
g_clear_pointer (&priv->dhcp6.event_id, g_free);
- nm_clear_g_source (&priv->dhcp6.restart_id);
+ nm_clear_g_source (&priv->dhcp6.grace_id);
if (priv->dhcp6.client) {
nm_clear_g_signal_handler (priv->dhcp6.client, &priv->dhcp6.state_sigid);
@@ -6763,53 +6760,17 @@ dhcp6_lease_change (NMDevice *self)
}
static gboolean
-dhcp6_restart_cb (gpointer user_data)
+dhcp6_grace_period_expired (gpointer user_data)
{
NMDevice *self = user_data;
- NMDevicePrivate *priv;
- g_return_val_if_fail (NM_IS_DEVICE (self), FALSE);
-
- priv = NM_DEVICE_GET_PRIVATE (self);
- priv->dhcp6.restart_id = 0;
-
- if (!dhcp6_start (self, FALSE))
- dhcp_schedule_restart (self, AF_INET6, NULL);
+ _LOGI (LOGD_DHCP6, "DHCPv6: grace period expired");
- return FALSE;
-}
+ nm_device_ip_method_failed (self, AF_INET6,
+ NM_DEVICE_STATE_REASON_IP_CONFIG_EXPIRED);
+ /* If the device didn't fail, the DHCP client will continue */
-static void
-dhcp_schedule_restart (NMDevice *self,
- int addr_family,
- const char *reason)
-{
- NMDevicePrivate *priv = NM_DEVICE_GET_PRIVATE (self);
- guint tries_left;
- char tries_str[255];
-
- nm_assert_addr_family (addr_family);
-
- tries_left = (addr_family == AF_INET)
- ? priv->dhcp4.num_tries_left
- : priv->dhcp6.num_tries_left;
-
- _LOGI ((addr_family == AF_INET) ? LOGD_DHCP4 : LOGD_DHCP6,
- "scheduling DHCPv%c restart in %u seconds%s%s%s%s",
- nm_utils_addr_family_to_char (addr_family),
- DHCP_RESTART_TIMEOUT,
- (tries_left != DHCP_NUM_TRIES_MAX)
- ? nm_sprintf_buf (tries_str, ", %u tries left", tries_left + 1)
- : "",
- NM_PRINT_FMT_QUOTED (reason, " (reason: ", reason, ")", ""));
-
- if (addr_family == AF_INET) {
- priv->dhcp4.restart_id = g_timeout_add_seconds (DHCP_RESTART_TIMEOUT,
- dhcp4_restart_cb, self);
- } else {
- priv->dhcp6.restart_id = g_timeout_add_seconds (DHCP_RESTART_TIMEOUT,
- dhcp6_restart_cb, self);
- }
+ return G_SOURCE_REMOVE;
}
static void
@@ -6818,51 +6779,57 @@ dhcp6_fail (NMDevice *self, gboolean timeout)
NMDevicePrivate *priv = NM_DEVICE_GET_PRIVATE (self);
gboolean is_dhcp_managed;
- _LOGD (LOGD_DHCP6, "DHCPv6 failed: timeout %d, num tries left %u",
- timeout, priv->dhcp6.num_tries_left);
+ _LOGD (LOGD_DHCP6, "DHCPv6 failed%s", timeout ? " (timeout)" : "");
is_dhcp_managed = (priv->dhcp6.mode == NM_NDISC_DHCP_LEVEL_MANAGED);
- dhcp6_cleanup (self, CLEANUP_TYPE_DECONFIGURE, FALSE);
- if (is_dhcp_managed || priv->dhcp6.num_tries_left < DHCP_NUM_TRIES_MAX) {
- /* Don't fail if there are static addresses configured on
- * the device, instead retry after some time.
+ if (is_dhcp_managed) {
+ /* Keep client running if there are static addresses configured
+ * on the interface.
*/
if ( priv->ip6_state == IP_DONE
&& priv->con_ip6_config
- && nm_ip6_config_get_num_addresses (priv->con_ip6_config)) {
- dhcp_schedule_restart (self, AF_INET6, "device has IP addresses");
+ && nm_ip6_config_get_num_addresses (priv->con_ip6_config))
+ goto clear_config;
+
+ /* Fail the method in case of timeout or failure during initial
+ * configuration.
+ */
+ if ( !priv->dhcp6.was_active
+ && (timeout || priv->ip6_state == IP_CONF)) {
+ dhcp6_cleanup (self, CLEANUP_TYPE_DECONFIGURE, FALSE);
+ nm_device_activate_schedule_ip6_config_timeout (self);
return;
}
- if ( priv->dhcp6.num_tries_left == DHCP_NUM_TRIES_MAX
- && (timeout || (priv->ip6_state == IP_CONF))
- && !priv->dhcp6.was_active)
- nm_device_activate_schedule_ip6_config_timeout (self);
- else if ( priv->dhcp6.num_tries_left < DHCP_NUM_TRIES_MAX
- || priv->ip6_state == IP_DONE
- || priv->dhcp6.was_active) {
- /* Don't fail immediately when the lease expires but try to
- * restart DHCP for a predefined number of times.
- */
- if (priv->dhcp6.num_tries_left) {
- priv->dhcp6.num_tries_left--;
- dhcp_schedule_restart (self, AF_INET6, "lease expired");
- } else {
- nm_device_ip_method_failed (self, AF_INET6, NM_DEVICE_STATE_REASON_IP_CONFIG_EXPIRED);
- /* We failed the ipv6 method but schedule again the retries if the ipv4 method is
- * configured, keeping the connection up.
- */
- if (nm_device_get_state (self) != NM_DEVICE_STATE_FAILED)
- dhcp_schedule_restart (self, AF_INET6, "renewal failed");
- }
- } else
- g_warn_if_reached ();
+ /* In any other case (expired lease, assumed connection, etc.),
+ * start a grace period in which we keep the client running,
+ * hoping that it will regain a lease.
+ */
+ if (!priv->dhcp6.grace_id) {
+ priv->dhcp6.grace_id = g_timeout_add_seconds (DHCP_GRACE_PERIOD_SEC,
+ dhcp6_grace_period_expired,
+ self);
+ _LOGI (LOGD_DHCP6,
+ "DHCPv6: %u seconds grace period started",
+ DHCP_GRACE_PERIOD_SEC);
+ goto clear_config;
+ }
} else {
/* not a hard failure; just live with the RA info */
+ dhcp6_cleanup (self, CLEANUP_TYPE_DECONFIGURE, FALSE);
if (priv->ip6_state == IP_CONF)
nm_device_activate_schedule_ip6_config_result (self);
}
+ return;
+
+clear_config:
+ /* The previous configuration is no longer valid */
+ if (priv->dhcp6.config) {
+ nm_exported_object_clear_and_unexport (&priv->dhcp6.config);
+ priv->dhcp6.config = nm_dhcp6_config_new ();
+ _notify (self, PROP_DHCP6_CONFIG);
+ }
}
static void
@@ -6898,6 +6865,7 @@ dhcp6_state_changed (NMDhcpClient *client,
switch (state) {
case NM_DHCP_STATE_BOUND:
+ nm_clear_g_source (&priv->dhcp6.grace_id);
/* If the server sends multiple IPv6 addresses, we receive a state
* changed event for each of them. Use the event ID to merge IPv6
* addresses from the same transaction into a single configuration.
@@ -6928,8 +6896,6 @@ dhcp6_state_changed (NMDhcpClient *client,
if (priv->ip6_state == IP_FAIL)
_set_ip_state (self, AF_INET6, IP_CONF);
- priv->dhcp6.num_tries_left = DHCP_NUM_TRIES_MAX;
-
if (priv->ip6_state == IP_CONF) {
if (priv->dhcp6.ip6_config == NULL) {
nm_device_ip_method_failed (self, AF_INET6, NM_DEVICE_STATE_REASON_DHCP_FAILED);
@@ -8107,8 +8073,6 @@ act_stage3_ip6_config_start (NMDevice *self,
}
priv->dhcp6.mode = NM_NDISC_DHCP_LEVEL_NONE;
- priv->dhcp6.num_tries_left = DHCP_NUM_TRIES_MAX;
-
method = nm_utils_get_ip_config_method (connection, NM_TYPE_SETTING_IP6_CONFIG);
if (strcmp (method, NM_SETTING_IP6_CONFIG_METHOD_IGNORE) == 0) {
--
2.14.3

View File

@ -9,7 +9,7 @@
%global epoch_version 1 %global epoch_version 1
%global rpm_version 1.10.6 %global rpm_version 1.10.6
%global real_version 1.10.6 %global real_version 1.10.6
%global release_version 2 %global release_version 3
%global snapshot %{nil} %global snapshot %{nil}
%global git_sha %{nil} %global git_sha %{nil}
@ -92,6 +92,7 @@ Source3: 20-connectivity-fedora.conf
Patch1: 0001-fix-build-with-gcc8.patch Patch1: 0001-fix-build-with-gcc8.patch
Patch2: 0002-device-check-rp_filter-all-rh1565529.patch Patch2: 0002-device-check-rp_filter-all-rh1565529.patch
Patch3: 0003-dhcp-let-client-continue-on-expiry-rh1575370.patch
Requires(post): systemd Requires(post): systemd
Requires(preun): systemd Requires(preun): systemd
@ -361,6 +362,7 @@ by nm-connection-editor and nm-applet in a non-graphical environment.
%patch1 -p1 %patch1 -p1
%patch2 -p1 %patch2 -p1
%patch3 -p1
%build %build
%if %{with regen_docs} %if %{with regen_docs}
@ -712,6 +714,9 @@ fi
%endif %endif
%changelog %changelog
* Sun May 6 2018 Beniamino Galvani <bgalvani@redhat.com> - 1:1.10.6-3
- dhcp: better handle expiry and nacks (rh #1575370)
* Tue Apr 17 2018 Beniamino Galvani <bgalvani@redhat.com> - 1:1.10.6-2 * Tue Apr 17 2018 Beniamino Galvani <bgalvani@redhat.com> - 1:1.10.6-2
- device: fix setting 'rp_filter' value (rh #1565529) - device: fix setting 'rp_filter' value (rh #1565529)