linux-kernel.diff

一个可以实现MPLS实验的linux源代码.你不仅可以实现单层标签的转发,还可以实现2个标签的实验,很好的.

+ */
+int shim_rta_blk_cmp(struct rtshim *a, struct shim_blk *b)
+{
+	int n = 0;
+	if (a && b) {
+		if (!(n = strncmp(a->name, b->shim->name, SHIMNAMSIZ)))
+		    n = memcmp(a->data, b->data, a->datalen);
+	} else {
+		if (a) n = 1;
+		if (b) n = -1;
+	}
+	return n;
+}
+
+/**
+ *	shim_blk_cmp - compare two active shim blks
+ *	@a: shim blk
+ *	@b: shim blk
+ *
+ *	Used for comparing two existing shim blks
+ */
+int shim_blk_cmp(struct shim_blk *a, struct shim_blk *b)
+{
+	int n = 0;
+	if (a && b) {
+		if (!(n = strncmp(a->shim->name, b->shim->name, SHIMNAMSIZ)))
+		    n = memcmp(a->data, b->data, a->datalen);
+	} else {
+		if (a) n = 1;
+		if (b) n = -1;
+	}
+	return n;
+}
+
+void __init shim_init(void)
+{
+	printk("NET: shim interface - <[e-mail]>\n");
+	INIT_LIST_HEAD(&shim_proto_list);
+}
+
+EXPORT_SYMBOL(shim_proto_add);
+EXPORT_SYMBOL(shim_proto_remove);
+EXPORT_SYMBOL(shim_build_blk);
+EXPORT_SYMBOL(shim_destroy_blk);
+EXPORT_SYMBOL(shim_proto_lock);
+EXPORT_SYMBOL(shim_proto_list);
diff -uNr linux-kernel/net/core/shim_procfs.c mpls-kernel-3/net/core/shim_procfs.c
--- linux-kernel/net/core/shim_procfs.c	1969-12-31 18:00:00.000000000 -0600
+++ mpls-kernel-3/net/core/shim_procfs.c	2005-11-11 12:19:10.000000000 -0600
@@ -0,0 +1,170 @@
+/*
+ *      Network shim interface for protocols that live below L3 and above L2
+ *
+ *		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.
+ *
+ *	Heavily borrowed from dev_remove_pack/dev_add_pack
+ *
+ *	Authors:	James R. Leu <[e-mail]>
+ */
+
+#include <linux/init.h>
+#include <linux/kernel.h>
+#include <linux/spinlock.h>
+#include <asm/byteorder.h>
+#include <linux/list.h>
+#include <net/shim.h>
+#include <linux/proc_fs.h>
+#include <linux/seq_file.h>
+
+extern spinlock_t shim_proto_lock;
+extern struct list_head shim_proto_list;
+
+/*
+ * MODULE Information and attributes
+ */
+
+MODULE_AUTHOR("James R. Leu <[e-mail]>");
+MODULE_DESCRIPTION("net-shim procfs module");
+MODULE_LICENSE("GPL");
+
+/*
+ * Proc filesystem directory entries.
+ */
+
+/*
+ *      /proc/net/shim
+ */
+										
+static struct proc_dir_entry *proc_net_shim;
+
+/*
+ *      Names of the proc directory entry
+ */
+										
+static const char name_shim[]    = "shim";
+
+/*
+ * The following few functions build the content of /proc/net/shim
+ */
+
+/* starting at shim, find the next registered protocol */
+static struct shim *shim_skip(struct shim *shim)
+{
+	struct shim *shim1;
+	int next = 0;
+
+	if (!shim)
+		next = 1;
+		
+	list_for_each_entry(shim1, &shim_proto_list, list) {
+		if (next)
+			return shim1;
+
+		if (shim1 == shim)
+			next = 1;
+	}
+
+	return NULL;
+}
+
+										
+/* start read of /proc/net/shim */
+static void *shim_seq_start(struct seq_file *seq, loff_t *pos)
+{
+	struct shim *shim;
+	loff_t i = 1;
+
+	spin_lock_bh(&shim_proto_lock);
+
+	if (*pos == 0)
+		return SEQ_START_TOKEN;
+
+	for (shim = shim_skip(NULL); shim && i < *pos;
+		shim = shim_skip(shim), ++i);
+										
+	return (i == *pos) ? shim : NULL;
+}
+
+static void *shim_seq_next(struct seq_file *seq, void *v, loff_t *pos)
+{
+	++*pos;
+										
+	return shim_skip((v == SEQ_START_TOKEN)
+			    ? NULL
+			    : (struct shim *)v);
+}
+										
+static void shim_seq_stop(struct seq_file *seq, void *v)
+{
+	spin_unlock_bh(&shim_proto_lock);
+}
+
+static int shim_seq_show(struct seq_file *seq, void *v)
+{
+	struct shim* shim = (struct shim*)v;
+	if (v != SEQ_START_TOKEN)
+		seq_printf(seq, "%s\t%d\n",
+		    shim->name ? shim->name : "(none)",
+		    atomic_read(&shim->__ref));
+	return 0;
+}
+
+/*
+ *      Generic /proc/net/shim file and inode operations
+ */
+										
+static struct seq_operations shim_seq_ops = {
+	.start = shim_seq_start,
+	.next = shim_seq_next,
+	.stop = shim_seq_stop,
+	.show = shim_seq_show,
+};
+										
+static int shim_seq_open(struct inode *inode, struct file *file)
+{
+	return seq_open(file, &shim_seq_ops);
+}
+										
+static struct file_operations shim_fops = {
+	.owner   = THIS_MODULE,
+	.open    = shim_seq_open,
+	.read    = seq_read,
+	.llseek  = seq_lseek,
+	.release = seq_release,
+};
+
+/*
+ *      Clean up /proc/net/shim
+ */
+										
+static void __exit shim_proc_cleanup(void)
+{
+	if (proc_net)
+		remove_proc_entry(name_shim, proc_net);
+}
+
+/*
+ *      Create /proc/net/shim
+ */
+
+static int __init shim_proc_init(void)
+{
+	if (proc_net) {
+		proc_net_shim = create_proc_entry(name_shim,
+						  S_IFREG|S_IRUSR|S_IWUSR,
+						  proc_net);
+		if (proc_net_shim) {
+			proc_net_shim->proc_fops = &shim_fops;
+			return 0;
+		}
+		return -ENOBUFS;
+	}
+	return -EINVAL;
+}
+
+module_init(shim_proc_init);
+module_exit(shim_proc_cleanup);
diff -uNr linux-kernel/net/ipv4/fib_semantics.c mpls-kernel-3/net/ipv4/fib_semantics.c
--- linux-kernel/net/ipv4/fib_semantics.c	2005-12-05 22:01:03.000000000 -0600
+++ mpls-kernel-3/net/ipv4/fib_semantics.c	2005-12-12 21:30:26.000000000 -0600
@@ -43,6 +43,7 @@
 #include <net/sock.h>
 #include <net/ip_fib.h>
 #include <net/ip_mp_alg.h>
+#include <net/shim.h>
 
 #include "fib_lookup.h"
 
@@ -148,9 +149,12 @@
 		return;
 	}
 	change_nexthops(fi) {
+		if (nh->nh_shim)
+			shim_destroy_blk(nh->nh_shim);
 		if (nh->nh_dev)
 			dev_put(nh->nh_dev);
 		nh->nh_dev = NULL;
+		nh->nh_shim = NULL;
 	} endfor_nexthops(fi);
 	fib_info_cnt--;
 	kfree(fi);
@@ -188,6 +192,7 @@
 #ifdef CONFIG_NET_CLS_ROUTE
 		    nh->nh_tclassid != onh->nh_tclassid ||
 #endif
+		    shim_blk_cmp(nh->nh_shim, onh->nh_shim) ||
 		    ((nh->nh_flags^onh->nh_flags)&~RTNH_F_DEAD))
 			return -1;
 		onh++;
@@ -341,14 +346,20 @@
 
 #ifdef CONFIG_IP_ROUTE_MULTIPATH
 
-static u32 fib_get_attr32(struct rtattr *attr, int attrlen, int type)
+static void* fib_get_attr(struct rtattr *attr, int attrlen, int type)
 {
 	while (RTA_OK(attr,attrlen)) {
 		if (attr->rta_type == type)
-			return *(u32*)RTA_DATA(attr);
+			return RTA_DATA(attr);
 		attr = RTA_NEXT(attr, attrlen);
 	}
-	return 0;
+	return NULL;
+}
+
+static u32 fib_get_attr32(struct rtattr *attr, int attrlen, int type)
+{
+	void *data = fib_get_attr(attr, attrlen, type);
+	return data ? *(u32*)data : 0;
 }
 
 static int
@@ -385,6 +396,9 @@
 #ifdef CONFIG_NET_CLS_ROUTE
 			nh->nh_tclassid = fib_get_attr32(RTNH_DATA(nhp), attrlen, RTA_FLOW);
 #endif
+			nh->nh_shim =
+				shim_build_blk(fib_get_attr(RTNH_DATA(nhp),
+				attrlen, RTA_SHIM));
 		}
 		nhp = RTNH_NEXT(nhp);
 	} endfor_nexthops(fi);
@@ -407,6 +421,7 @@
 
 	if (rta->rta_oif || rta->rta_gw) {
 		if ((!rta->rta_oif || *rta->rta_oif == fi->fib_nh->nh_oif) &&
+		    (!rta->rta_shim || shim_rta_blk_cmp(rta->rta_shim, fi->fib_nh->nh_shim) == 0) &&
 		    (!rta->rta_gw  || memcmp(rta->rta_gw, &fi->fib_nh->nh_gw, 4) == 0))
 			return 0;
 		return 1;
@@ -435,6 +450,9 @@
 			if (gw && gw != nh->nh_tclassid)
 				return 1;
 #endif
+			if (shim_rta_blk_cmp(fib_get_attr(RTNH_DATA(nhp),
+				attrlen, RTA_SHIM), nh->nh_shim))
+				return 1;
 		}
 		nhp = RTNH_NEXT(nhp);
 	} endfor_nexthops(fi);
@@ -749,6 +767,8 @@
 			goto err_inval;
 		if (rta->rta_gw && memcmp(&fi->fib_nh->nh_gw, rta->rta_gw, 4))
 			goto err_inval;
+		if (shim_rta_blk_cmp(rta->rta_shim, fi->fib_nh->nh_shim))
+			goto err_inval;
 #ifdef CONFIG_NET_CLS_ROUTE
 		if (rta->rta_flow && memcmp(&fi->fib_nh->nh_tclassid, rta->rta_flow, 4))
 			goto err_inval;
@@ -770,6 +790,11 @@
 #ifdef CONFIG_IP_ROUTE_MULTIPATH
 		nh->nh_weight = 1;
 #endif
+		if (rta->rta_shim) {
+			nh->nh_shim = shim_build_blk(rta->rta_shim);
+			if (!nh->nh_shim)
+				goto err_inval;
+		}
 	}
 
 #ifdef CONFIG_IP_ROUTE_MULTIPATH_CACHED
@@ -777,7 +802,7 @@
 #endif
 
 	if (fib_props[r->rtm_type].error) {
-		if (rta->rta_gw || rta->rta_oif || rta->rta_mp)
+		if (rta->rta_gw || rta->rta_oif || rta->rta_mp || rta->rta_shim)
 			goto err_inval;
 		goto link_it;
 	}
@@ -974,6 +999,12 @@
 			RTA_PUT(skb, RTA_GATEWAY, 4, &fi->fib_nh->nh_gw);
 		if (fi->fib_nh->nh_oif)
 			RTA_PUT(skb, RTA_OIF, sizeof(int), &fi->fib_nh->nh_oif);
+		if (fi->fib_nh->nh_shim) {
+			struct rtshim *shim = RTA_DATA(__RTA_PUT(skb, RTA_SHIM,
+				sizeof(struct rtshim) +
+				fi->fib_nh->nh_shim->datalen));
+			shim_unbuild_blk(shim, fi->fib_nh->nh_shim);
+		}
 	}
 #ifdef CONFIG_IP_ROUTE_MULTIPATH
 	if (fi->fib_nhs > 1) {
@@ -992,6 +1023,13 @@
 			nhp->rtnh_ifindex = nh->nh_oif;
 			if (nh->nh_gw)
 				RTA_PUT(skb, RTA_GATEWAY, 4, &nh->nh_gw);
+			if (nh->nh_shim) {
+				struct rtattr *rta;
+				rta = __RTA_PUT(skb, RTA_SHIM,
+					sizeof(struct rtshim) +
+					nh->nh_shim->datalen);
+				shim_unbuild_blk(RTA_DATA(rta), nh->nh_shim);
+			}
 			nhp->rtnh_len = skb->tail - (unsigned char*)nhp;
 		} endfor_nexthops(fi);
 		mp_head->rta_type = RTA_MULTIPATH;
diff -uNr linux-kernel/net/ipv4/ip_input.c mpls-kernel-3/net/ipv4/ip_input.c
--- linux-kernel/net/ipv4/ip_input.c	2005-11-15 21:43:54.000000000 -0600
+++ mpls-kernel-3/net/ipv4/ip_input.c	2005-12-06 22:36:05.000000000 -0600
@@ -443,3 +443,4 @@
 }
 
 EXPORT_SYMBOL(ip_statistics);
+EXPORT_SYMBOL(ip_rcv);
diff -uNr linux-kernel/net/ipv4/ip_output.c mpls-kernel-3/net/ipv4/ip_output.c
--- linux-kernel/net/ipv4/ip_output.c	2005-12-05 22:01:03.000000000 -0600
+++ mpls-kernel-3/net/ipv4/ip_output.c	2005-12-12 21:30:43.000000000 -0600
@@ -167,6 +167,11 @@
 	struct net_device *dev = dst->dev;
 	int hh_len = LL_RESERVED_SPACE(dev);
 
+	if (dst->child) {
+		skb->dst = dst_pop(skb->dst);
+		return dst_output(skb);
+	}
+
 	/* Be paranoid, rather than too clever. */
 	if (unlikely(skb_headroom(skb) < hh_len && dev->hard_header)) {
 		struct sk_buff *skb2;
diff -uNr linux-kernel/net/ipv4/Kconfig mpls-kernel-3/net/ipv4/Kconfig
--- linux-kernel/net/ipv4/Kconfig	2005-11-15 21:43:55.000000000 -0600
+++ mpls-kernel-3/net/ipv4/Kconfig	2005-11-15 23:18:10.000000000 -0600
@@ -231,6 +231,15 @@
 	  operating on your network. Read <file:Documentation/nfsroot.txt> for
 	  details.
 
+config IP_MPLS
+	tristate "IP: MPLS support"
+	depends on INET && MPLS
+	---help---
+	  If you say Y here, the kernel will support being an ingress and
+	  egress LER for IPv4 packets
+
+	  If unsure, say N.
+
 # not yet ready..
 #   bool '    IP: ARP support' CONFIG_IP_PNP_ARP		
 config NET_IPIP
diff -uNr linux-kernel/net/ipv4/Makefile mpls-kernel-3/net/ipv4/Makefile
--- linux-kernel/net/ipv4/Makefile	2005-11-15 21:43:55.000000000 -0600
+++ mpls-kernel-3/net/ipv4/Makefile	2005-11-15 23:18:10.000000000 -0600
@@ -28,6 +28,7 @@
 obj-$(CONFIG_IP_ROUTE_MULTIPATH_RANDOM) += multipath_random.o
 obj-$(CONFIG_IP_ROUTE_MULTIPATH_WRANDOM) += multipath_wrandom.o
 obj-$(CONFIG_IP_ROUTE_MULTIPATH_DRR) += multipath_drr.o
+obj-$(CONFIG_IP_MPLS) += mpls4.o
 obj-$(CONFIG_NETFILTER)	+= netfilter/
 obj-$(CONFIG_IP_VS) += ipvs/
 obj-$(CONFIG_INET_DIAG) += inet_diag.o 
diff -uNr linux-kernel/net/ipv4/mpls4.c mpls-kernel-3/net/ipv4/mpls4.c
--- linux-kernel/net/ipv4/mpls4.c	1969-12-31 18:00:00.000000000 -0600
+++ mpls-kernel-3/net/ipv4/mpls4.c	2005-12-18 09:51:01.000000000 -0600
@@ -0,0 +1,394 @@
+/* mpls4.c: IPv4 MPLS protocol driver.
+ *
+ * Copyright (C) 2003 David S. Miller ([e-mail])
+ *
+ * Changes:
+ *	JLEU: 	Add ICMP handling
+ *		Add nexthop printing
+ *		Change nexthop resolve signature
+ *	JLEU:	Added mpls4_cache_flush()
+ *	JLEU:	un/register reserved labels in fini/init
+ *	JLEU:	removed sysfs print routin
+ */
+
+#include <linux/mo