Hi David,
I've updated my RFC2863 operstatus and userspace forwarding patch to
2.5.49. As discussion on netdev seems to be complete, I need some
comment from your side.
Stefan diff -uprNX dontdiff linux-2.5.49/include/linux/netdevice.h
linux-2.5.49-lw/include/linux/netdevice.h
--- linux-2.5.49/include/linux/netdevice.h 2002-11-22 22:41:08.000000000
+0100
+++ linux-2.5.49-lw/include/linux/netdevice.h 2002-11-25 22:52:29.000000000
+0100
@@ -204,10 +204,23 @@ enum netdev_state_t
{
__LINK_STATE_XOFF=0,
__LINK_STATE_START,
- __LINK_STATE_PRESENT,
+ __LINK_STATE_PRESENT_OBSOLETE,
__LINK_STATE_SCHED,
- __LINK_STATE_NOCARRIER,
- __LINK_STATE_RX_SCHED
+ __LINK_STATE_NOCARRIER_OBSOLETE,
+ __LINK_STATE_RX_SCHED,
+ __LINK_STATE_LINKWATCH_PENDING
+};
+
+
+/* Device operative state as per RFC2863 */
+enum netdev_operstate_t {
+ NETDEV_OPER_UP = 1,
+ NETDEV_OPER_DOWN, /* Obsoletes LINK_STATE_NOCARRIER */
+ NETDEV_OPER_TESTING,
+ NETDEV_OPER_UNKNOWN,
+ NETDEV_OPER_DORMANT,
+ NETDEV_OPER_NOTPRESENT, /* Obsoletes !LINK_STATE_PRESENT */
+ NETDEV_OPER_LOWERDOWN
};
@@ -308,6 +321,10 @@ struct net_device
* which this device is member of.
*/
+ /* Operative state, access semaphore */
+ rwlock_t operstate_lock;
+ unsigned char operstate;
+
/* Interface address info. */
unsigned char broadcast[MAX_ADDR_LEN]; /* hw bcast add
*/
unsigned char dev_addr[MAX_ADDR_LEN]; /* hw address */
@@ -631,34 +648,76 @@ static inline void dev_put(struct net_de
* who is responsible for serialization of these calls.
*/
+#ifdef CONFIG_LINKWATCH
+extern void linkwatch_fire_event(struct net_device *dev);
+#endif
+
+static inline unsigned char netif_set_operstate(struct net_device *dev,
unsigned char newstate)
+{
+ unsigned long flags;
+ unsigned char oldstate;
+
+ write_lock_irqsave(&dev->operstate_lock, flags);
+ oldstate = dev->operstate;
+ dev->operstate = newstate;
+ write_unlock_irqrestore(&dev->operstate_lock, flags);
+
+#ifdef CONFIG_LINKWATCH
+ if (oldstate != newstate) linkwatch_fire_event(dev);
+#endif
+
+ return oldstate;
+}
+
+static inline unsigned char netif_get_operstate(struct net_device *dev)
+{
+ unsigned long flags;
+ unsigned char state;
+
+ read_lock_irqsave(&dev->operstate_lock, flags);
+ state = dev->operstate;
+ read_unlock_irqrestore(&dev->operstate_lock, flags);
+
+ return state;
+}
+
static inline int netif_carrier_ok(struct net_device *dev)
{
- return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
+ return netif_get_operstate(dev) != NETDEV_OPER_UP;
+}
+
+static inline int netif_operstate_to_iff_running(struct net_device *dev)
+{
+ unsigned char state = netif_get_operstate(dev);
+
+ return((1 << state) &
+ (1 << NETDEV_OPER_UP | 1 << NETDEV_OPER_UNKNOWN));
}
extern void __netdev_watchdog_up(struct net_device *dev);
+
static inline void netif_carrier_on(struct net_device *dev)
{
- clear_bit(__LINK_STATE_NOCARRIER, &dev->state);
+ netif_set_operstate(dev, NETDEV_OPER_UP);
if (netif_running(dev))
__netdev_watchdog_up(dev);
}
static inline void netif_carrier_off(struct net_device *dev)
{
- set_bit(__LINK_STATE_NOCARRIER, &dev->state);
+ netif_set_operstate(dev, NETDEV_OPER_DOWN);
}
/* Hot-plugging. */
static inline int netif_device_present(struct net_device *dev)
{
- return test_bit(__LINK_STATE_PRESENT, &dev->state);
+ return netif_get_operstate(dev) != NETDEV_OPER_NOTPRESENT;
}
static inline void netif_device_detach(struct net_device *dev)
{
- if (test_and_clear_bit(__LINK_STATE_PRESENT, &dev->state) &&
+ if (netif_set_operstate(dev, NETDEV_OPER_NOTPRESENT) !=
NETDEV_OPER_NOTPRESENT &&
netif_running(dev)) {
netif_stop_queue(dev);
}
@@ -666,7 +725,7 @@ static inline void netif_device_detach(s
static inline void netif_device_attach(struct net_device *dev)
{
- if (!test_and_set_bit(__LINK_STATE_PRESENT, &dev->state) &&
+ if (netif_set_operstate(dev, NETDEV_OPER_UNKNOWN) ==
NETDEV_OPER_NOTPRESENT &&
netif_running(dev)) {
netif_wake_queue(dev);
__netdev_watchdog_up(dev);
diff -uprNX dontdiff linux-2.5.49/net/Kconfig linux-2.5.49-lw/net/Kconfig
--- linux-2.5.49/net/Kconfig 2002-11-22 22:40:16.000000000 +0100
+++ linux-2.5.49-lw/net/Kconfig 2002-11-25 22:58:39.000000000 +0100
@@ -265,6 +265,20 @@ config ATM_MPOA
config VLAN_8021Q
tristate "802.1Q VLAN Support"
+config LINKWATCH
+ bool "Device link state notification (EXPERIMENTAL)"
+ depends on EXPERIMENTAL
+ help
+ When this option is enabled, the kernel will forward changes in the
+ operative ("RUNNING") state of an interface via the netlink socket.
+ This is most useful when running linux as a router.
+
+ Note that currently not many drivers support this, compliant ones
+ can be found by watching the the RUNNING flag in ifconfig output
+ that should follow operative state.
+
+ If unsure, say 'N'.
+
config LLC
tristate "ANSI/IEEE 802.2 - aka LLC (IPX, Appletalk, Token Ring)"
help
diff -uprNX dontdiff linux-2.5.49/net/core/Makefile
linux-2.5.49-lw/net/core/Makefile
--- linux-2.5.49/net/core/Makefile 2002-11-22 22:41:08.000000000 +0100
+++ linux-2.5.49-lw/net/core/Makefile 2002-11-25 22:52:33.000000000 +0100
@@ -22,4 +22,6 @@ obj-$(CONFIG_NET_RADIO) += wireless.o
# Ugly. I wish all wireless drivers were moved in drivers/net/wireless
obj-$(CONFIG_NET_PCMCIA_RADIO) += wireless.o
+obj-$(CONFIG_LINKWATCH) += link_watch.o
+
include $(TOPDIR)/Rules.make
diff -uprNX dontdiff linux-2.5.49/net/core/dev.c linux-2.5.49-lw/net/core/dev.c
--- linux-2.5.49/net/core/dev.c 2002-11-22 22:40:43.000000000 +0100
+++ linux-2.5.49-lw/net/core/dev.c 2002-11-25 22:52:33.000000000 +0100
@@ -199,7 +199,6 @@ int netdev_fastroute;
int netdev_fastroute_obstacles;
#endif
-
/*******************************************************************************
Protocol management and registration routines
@@ -262,6 +261,9 @@ void dev_add_pack(struct packet_type *pt
br_write_unlock_bh(BR_NETPROTO_LOCK);
}
+#ifdef CONFIG_LINKWATCH
+void linkwatch_run_queue(void);
+#endif
/**
* dev_remove_pack - remove packet handler
@@ -2018,7 +2020,7 @@ static int dev_ifsioc(struct ifreq *ifr,
IFF_RUNNING)) |
(dev->gflags & (IFF_PROMISC |
IFF_ALLMULTI));
- if (netif_running(dev) && netif_carrier_ok(dev))
+ if (netif_running(dev) &&
netif_operstate_to_iff_running(dev))
ifr->ifr_flags |= IFF_RUNNING;
return 0;
@@ -2433,6 +2435,10 @@ int register_netdevice(struct net_device
goto out;
#endif /* CONFIG_NET_DIVERT */
+ /* Initial operstate */
+ dev->operstate_lock = RW_LOCK_UNLOCKED;
+ dev->operstate = NETDEV_OPER_UNKNOWN;
+
dev->iflink = -1;
/* Init, if this function is available */
@@ -2458,13 +2464,6 @@ int register_netdevice(struct net_device
if (!dev->rebuild_header)
dev->rebuild_header = default_rebuild_header;
- /*
- * Default initial state at registry is that the
- * device is present.
- */
-
- set_bit(__LINK_STATE_PRESENT, &dev->state);
-
dev->next = NULL;
dev_init_scheduler(dev);
write_lock_bh(&dev_base_lock);
@@ -2642,6 +2641,17 @@ int unregister_netdevice(struct net_devi
/* Rebroadcast unregister notification */
notifier_call_chain(&netdev_chain,
NETDEV_UNREGISTER, dev);
+
+#ifdef CONFIG_LINKWATCH
+ if (test_bit(__LINK_STATE_LINKWATCH_PENDING,
&dev->state)) {
+ /* We must not have linkwatch events pending
+ * on unregister. If this happens, we simply
+ * run the queue unscheduled, resulting in a
+ * noop for this device
+ */
+ linkwatch_run_queue();
+ }
+#endif
}
current->state = TASK_INTERRUPTIBLE;
schedule_timeout(HZ / 4);
@@ -2736,6 +2746,8 @@ static int __init net_dev_init(void)
#ifdef CONFIG_NET_FASTROUTE
dev->fastpath_lock = RW_LOCK_UNLOCKED;
#endif
+ dev->operstate_lock = RW_LOCK_UNLOCKED;
+ dev->operstate = NETDEV_OPER_UNKNOWN;
dev->xmit_lock_owner = -1;
dev->iflink = -1;
dev_hold(dev);
@@ -2768,7 +2780,6 @@ static int __init net_dev_init(void)
if (!dev->rebuild_header)
dev->rebuild_header = default_rebuild_header;
dev_init_scheduler(dev);
- set_bit(__LINK_STATE_PRESENT, &dev->state);
}
}
@@ -2849,3 +2860,5 @@ static int net_run_sbin_hotplug(struct n
return call_usermodehelper(argv [0], argv, envp);
}
#endif
+
+
diff -uprNX dontdiff linux-2.5.49/net/core/link_watch.c
linux-2.5.49-lw/net/core/link_watch.c
--- linux-2.5.49/net/core/link_watch.c 1970-01-01 01:00:00.000000000 +0100
+++ linux-2.5.49-lw/net/core/link_watch.c 2002-11-25 22:52:33.000000000
+0100
@@ -0,0 +1,134 @@
+/*
+ * Linux network device link state notifaction
+ *
+ * Author:
+ * Stefan Rompf <sux@xxxxxx>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version
+ * 2 of the License, or (at your option) any later version.
+ *
+ */
+
+#include <linux/workqueue.h>
+#include <linux/config.h>
+#include <linux/netdevice.h>
+#include <linux/if.h>
+#include <linux/rtnetlink.h>
+#include <linux/jiffies.h>
+#include <linux/spinlock.h>
+#include <linux/list.h>
+#include <linux/slab.h>
+#include <linux/workqueue.h>
+#include <asm/bitops.h>
+#include <asm/types.h>
+
+
+enum lw_bits {
+ LW_RUNNING = 0,
+ LW_SE_USED
+};
+
+static unsigned long linkwatch_flags = 0;
+static unsigned long linkwatch_nextevent = 0;
+
+static void linkwatch_event(void *dummy);
+static DECLARE_WORK(linkwatch_work, linkwatch_event, NULL);
+
+static LIST_HEAD(lweventlist);
+static spinlock_t lweventlist_lock = SPIN_LOCK_UNLOCKED;
+
+struct lw_event {
+ struct list_head list;
+ struct net_device *dev;
+};
+
+/* Avoid kmalloc() for most systems */
+struct lw_event singleevent;
+
+/* Must be called with the rtnl semaphore held */
+void linkwatch_run_queue(void) {
+ LIST_HEAD(head);
+ struct list_head *n, *next;
+
+ spin_lock_irq(&lweventlist_lock);
+ list_splice_init(&lweventlist, &head);
+ spin_unlock_irq(&lweventlist_lock);
+
+ list_for_each_safe(n, next, &head) {
+ struct lw_event *event = list_entry(n, struct lw_event, list);
+ struct net_device *dev = event->dev;
+
+ if (event == &singleevent) {
+ clear_bit(LW_SE_USED, &linkwatch_flags);
+ } else {
+ kfree(event);
+ }
+
+ /* We are about to handle this device,
+ * so new events can be accepted
+ */
+ clear_bit(__LINK_STATE_LINKWATCH_PENDING, &dev->state);
+
+ if (dev->flags & IFF_UP) {
+ netdev_state_change(dev);
+ }
+
+ dev_put(dev);
+ }
+}
+
+
+static void linkwatch_event(void *dummy)
+{
+ /* Limit the number of linkwatch events to one
+ * per second so that a runaway driver does not
+ * cause a storm of messages on the netlink
+ * socket
+ */
+ linkwatch_nextevent = jiffies + HZ;
+ clear_bit(LW_RUNNING, &linkwatch_flags);
+
+ rtnl_lock();
+ linkwatch_run_queue();
+ rtnl_unlock();
+}
+
+
+void linkwatch_fire_event(struct net_device *dev)
+{
+ if (!test_and_set_bit(__LINK_STATE_LINKWATCH_PENDING, &dev->state)) {
+ unsigned long flags;
+ struct lw_event *event;
+
+ if (test_and_set_bit(LW_SE_USED, &linkwatch_flags)) {
+ event = kmalloc(sizeof(struct lw_event), GFP_ATOMIC);
+
+ if (unlikely(event == NULL)) {
+ clear_bit(__LINK_STATE_LINKWATCH_PENDING,
&dev->state);
+ return;
+ }
+ } else {
+ event = &singleevent;
+ }
+
+ dev_hold(dev);
+ event->dev = dev;
+
+ spin_lock_irqsave(&lweventlist_lock, flags);
+ list_add_tail(&event->list, &lweventlist);
+ spin_unlock_irqrestore(&lweventlist_lock, flags);
+
+ if (!test_and_set_bit(LW_RUNNING, &linkwatch_flags)) {
+ unsigned long thisevent = jiffies;
+
+ if (thisevent >= linkwatch_nextevent) {
+ schedule_work(&linkwatch_work);
+ } else {
+ schedule_delayed_work(&linkwatch_work,
linkwatch_nextevent - thisevent);
+ }
+ }
+ }
+}
+
diff -uprNX dontdiff linux-2.5.49/net/core/rtnetlink.c
linux-2.5.49-lw/net/core/rtnetlink.c
--- linux-2.5.49/net/core/rtnetlink.c 2002-11-22 22:41:13.000000000 +0100
+++ linux-2.5.49-lw/net/core/rtnetlink.c 2002-11-25 22:52:33.000000000
+0100
@@ -165,7 +165,7 @@ static int rtnetlink_fill_ifinfo(struct
r->ifi_flags = dev->flags;
r->ifi_change = change;
- if (!netif_running(dev) || !netif_carrier_ok(dev))
+ if (!netif_running(dev) || !netif_operstate_to_iff_running(dev))
r->ifi_flags &= ~IFF_RUNNING;
else
r->ifi_flags |= IFF_RUNNING;
diff -uprNX dontdiff linux-2.5.49/net/netsyms.c linux-2.5.49-lw/net/netsyms.c
--- linux-2.5.49/net/netsyms.c 2002-11-22 22:40:39.000000000 +0100
+++ linux-2.5.49-lw/net/netsyms.c 2002-11-25 22:52:33.000000000 +0100
@@ -632,4 +632,8 @@ extern void wireless_send_event(struct n
EXPORT_SYMBOL(wireless_send_event);
#endif /* CONFIG_NET_RADIO || CONFIG_NET_PCMCIA_RADIO */
+#ifdef CONFIG_LINKWATCH
+EXPORT_SYMBOL(linkwatch_fire_event);
+#endif
+
#endif /* CONFIG_NET */
|