[PATCH 0/3] [PATCHSET] fib_rules: goto/nop action, flag for detached rules

[PATCH 0/3] [PATCHSET] fib_rules: goto/nop action, flag for detached rules

[Thomas Graf]

Patches to the fib rules code, not directly connected, see
description of each patch.

[PATCH 1/3] [NET] fib_rules: goto rule action

[Thomas Graf]

[-- Attachment #1: rules_goto --]
[-- Type: text/plain, Size: 6819 bytes --]

This patch adds a new rule action FR_ACT_GOTO which allows
to skip a set of rules by jumping to another rule. The rule
to jump to is specified via the FRA_GOTO attribute which
carries a rule preference.

Referring to a rule which doesn't exists is explicitely allowed.
Such goto rules are marked with the flag FIB_RULE_UNRESOLVED
and will act like a rule with a non-matching selector. The rule
will become functional as soon as its target is present.

The goto action enables performance optimizations by reducing
the average number of rules that have to be passed per lookup.

Example:
0:      from all lookup local 
40:     not from all to 192.168.23.128 goto 32766
41:     from all fwmark 0xa blackhole
42:     from all fwmark 0xff blackhole
32766:  from all lookup main 

Signed-off-by: Thomas Graf <tgraf@suug.ch>

Index: net-2.6.22/include/linux/fib_rules.h
===================================================================
--- net-2.6.22.orig/include/linux/fib_rules.h	2007-03-26 13:07:41.000000000 +0200
+++ net-2.6.22/include/linux/fib_rules.h	2007-03-27 01:43:14.000000000 +0200
@@ -7,6 +7,7 @@
 /* rule is permanent, and cannot be deleted */
 #define FIB_RULE_PERMANENT	1
 #define FIB_RULE_INVERT		2
+#define FIB_RULE_UNRESOLVED	4
 
 struct fib_rule_hdr
 {
@@ -29,7 +30,7 @@ enum
 	FRA_DST,	/* destination address */
 	FRA_SRC,	/* source address */
 	FRA_IFNAME,	/* interface name */
-	FRA_UNUSED1,
+	FRA_GOTO,	/* target to jump to (FR_ACT_GOTO) */
 	FRA_UNUSED2,
 	FRA_PRIORITY,	/* priority/preference */
 	FRA_UNUSED3,
@@ -51,7 +52,7 @@ enum
 {
 	FR_ACT_UNSPEC,
 	FR_ACT_TO_TBL,		/* Pass to fixed table */
-	FR_ACT_RES1,
+	FR_ACT_GOTO,		/* Jump to another rule */
 	FR_ACT_RES2,
 	FR_ACT_RES3,
 	FR_ACT_RES4,
Index: net-2.6.22/include/net/fib_rules.h
===================================================================
--- net-2.6.22.orig/include/net/fib_rules.h	2007-03-26 13:07:41.000000000 +0200
+++ net-2.6.22/include/net/fib_rules.h	2007-03-26 13:40:32.000000000 +0200
@@ -19,6 +19,8 @@ struct fib_rule
 	u32			flags;
 	u32			table;
 	u8			action;
+	u32			target;
+	struct fib_rule *	ctarget;
 	struct rcu_head		rcu;
 };
 
@@ -35,6 +37,8 @@ struct fib_rules_ops
 	struct list_head	list;
 	int			rule_size;
 	int			addr_size;
+	int			unresolved_rules;
+	int			nr_goto_rules;
 
 	int			(*action)(struct fib_rule *,
 					  struct flowi *, int,
@@ -66,7 +70,8 @@ struct fib_rules_ops
 	[FRA_PRIORITY]	= { .type = NLA_U32 }, \
 	[FRA_FWMARK]	= { .type = NLA_U32 }, \
 	[FRA_FWMASK]	= { .type = NLA_U32 }, \
-	[FRA_TABLE]     = { .type = NLA_U32 }
+	[FRA_TABLE]     = { .type = NLA_U32 }, \
+	[FRA_GOTO]	= { .type = NLA_U32 }
 
 static inline void fib_rule_get(struct fib_rule *rule)
 {
Index: net-2.6.22/net/core/fib_rules.c
===================================================================
--- net-2.6.22.orig/net/core/fib_rules.c	2007-03-26 13:07:41.000000000 +0200
+++ net-2.6.22/net/core/fib_rules.c	2007-03-27 01:43:39.000000000 +0200
@@ -132,10 +132,23 @@ int fib_rules_lookup(struct fib_rules_op
 	rcu_read_lock();
 
 	list_for_each_entry_rcu(rule, ops->rules_list, list) {
+jumped:
 		if (!fib_rule_match(rule, ops, fl, flags))
 			continue;
 
-		err = ops->action(rule, fl, flags, arg);
+		if (rule->action == FR_ACT_GOTO) {
+			struct fib_rule *target;
+
+			target = rcu_dereference(rule->ctarget);
+			if (target == NULL) {
+				continue;
+			} else {
+				rule = target;
+				goto jumped;
+			}
+		} else
+			err = ops->action(rule, fl, flags, arg);
+
 		if (err != -EAGAIN) {
 			fib_rule_get(rule);
 			arg->rule = rule;
@@ -180,7 +193,7 @@ static int fib_nl_newrule(struct sk_buff
 	struct fib_rules_ops *ops = NULL;
 	struct fib_rule *rule, *r, *last = NULL;
 	struct nlattr *tb[FRA_MAX+1];
-	int err = -EINVAL;
+	int err = -EINVAL, unresolved = 0;
 
 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*frh)))
 		goto errout;
@@ -237,6 +250,28 @@ static int fib_nl_newrule(struct sk_buff
 	if (!rule->pref && ops->default_pref)
 		rule->pref = ops->default_pref();
 
+	err = -EINVAL;
+	if (tb[FRA_GOTO]) {
+		if (rule->action != FR_ACT_GOTO)
+			goto errout_free;
+
+		rule->target = nla_get_u32(tb[FRA_GOTO]);
+		/* Backward jumps are prohibited to avoid endless loops */
+		if (rule->target <= rule->pref)
+			goto errout_free;
+
+		list_for_each_entry(r, ops->rules_list, list) {
+			if (r->pref == rule->target) {
+				rule->ctarget = r;
+				break;
+			}
+		}
+
+		if (rule->ctarget == NULL)
+			unresolved = 1;
+	} else if (rule->action == FR_ACT_GOTO)
+		goto errout_free;
+
 	err = ops->configure(rule, skb, nlh, frh, tb);
 	if (err < 0)
 		goto errout_free;
@@ -249,6 +284,28 @@ static int fib_nl_newrule(struct sk_buff
 
 	fib_rule_get(rule);
 
+	if (ops->unresolved_rules) {
+		/*
+		 * There are unresolved goto rules in the list, check if
+		 * any of them are pointing to this new rule.
+		 */
+		list_for_each_entry(r, ops->rules_list, list) {
+			if (r->action == FR_ACT_GOTO &&
+			    r->target == rule->pref) {
+				BUG_ON(r->ctarget != NULL);
+				rcu_assign_pointer(r->ctarget, rule);
+				if (--ops->unresolved_rules == 0)
+					break;
+			}
+		}
+	}
+
+	if (rule->action == FR_ACT_GOTO)
+		ops->nr_goto_rules++;
+
+	if (unresolved)
+		ops->unresolved_rules++;
+
 	if (last)
 		list_add_rcu(&rule->list, &last->list);
 	else
@@ -269,7 +326,7 @@ static int fib_nl_delrule(struct sk_buff
 {
 	struct fib_rule_hdr *frh = nlmsg_data(nlh);
 	struct fib_rules_ops *ops = NULL;
-	struct fib_rule *rule;
+	struct fib_rule *rule, *tmp;
 	struct nlattr *tb[FRA_MAX+1];
 	int err = -EINVAL;
 
@@ -322,6 +379,25 @@ static int fib_nl_delrule(struct sk_buff
 		}
 
 		list_del_rcu(&rule->list);
+
+		if (rule->action == FR_ACT_GOTO)
+			ops->nr_goto_rules--;
+
+		/*
+		 * Check if this rule is a target to any of them. If so,
+		 * disable them. As this operation is eventually very
+		 * expensive, it is only performed if goto rules have
+		 * actually been added.
+		 */
+		if (ops->nr_goto_rules > 0) {
+			list_for_each_entry(tmp, ops->rules_list, list) {
+				if (tmp->ctarget == rule) {
+					rcu_assign_pointer(tmp->ctarget, NULL);
+					ops->unresolved_rules++;
+				}
+			}
+		}
+
 		synchronize_rcu();
 		notify_rule_change(RTM_DELRULE, rule, ops, nlh,
 				   NETLINK_CB(skb).pid);
@@ -371,6 +447,9 @@ static int fib_nl_fill_rule(struct sk_bu
 	frh->action = rule->action;
 	frh->flags = rule->flags;
 
+	if (rule->action == FR_ACT_GOTO && rule->ctarget == NULL)
+		frh->flags |= FIB_RULE_UNRESOLVED;
+
 	if (rule->ifname[0])
 		NLA_PUT_STRING(skb, FRA_IFNAME, rule->ifname);
 
@@ -383,6 +462,9 @@ static int fib_nl_fill_rule(struct sk_bu
 	if (rule->mark_mask || rule->mark)
 		NLA_PUT_U32(skb, FRA_FWMASK, rule->mark_mask);
 
+	if (rule->target)
+		NLA_PUT_U32(skb, FRA_GOTO, rule->target);
+
 	if (ops->fill(rule, skb, nlh, frh) < 0)
 		goto nla_put_failure;
 

--

[PATCH 2/3] [NET] fib_rules: Mark rules detached from the device

[Thomas Graf]

[-- Attachment #1: rule_detached_flag --]
[-- Type: text/plain, Size: 1694 bytes --]

Rules which match against device names in their selector can
remain while the device itself disappears, in fact the device
doesn't have to present when the rule is added in the first
place. The device name is resolved by trying when the rule is
added and later by listening to NETDEV_REGISTER/UNREGISTER
notifications.

This patch adds the flag FIB_RULE_DEV_DETACHED which is set
towards userspace when a rule contains a device match which
is unresolved at the moment. This eases spotting the reason
why certain rules seem not to function properly.

Signed-off-by: Thomas Graf <tgraf@suug.ch>

Index: net-2.6.22/include/linux/fib_rules.h
===================================================================
--- net-2.6.22.orig/include/linux/fib_rules.h	2007-03-27 01:43:14.000000000 +0200
+++ net-2.6.22/include/linux/fib_rules.h	2007-03-27 01:45:56.000000000 +0200
@@ -8,6 +8,7 @@
 #define FIB_RULE_PERMANENT	1
 #define FIB_RULE_INVERT		2
 #define FIB_RULE_UNRESOLVED	4
+#define FIB_RULE_DEV_DETACHED	8
 
 struct fib_rule_hdr
 {
Index: net-2.6.22/net/core/fib_rules.c
===================================================================
--- net-2.6.22.orig/net/core/fib_rules.c	2007-03-27 01:43:39.000000000 +0200
+++ net-2.6.22/net/core/fib_rules.c	2007-03-27 01:46:22.000000000 +0200
@@ -450,9 +450,13 @@ static int fib_nl_fill_rule(struct sk_bu
 	if (rule->action == FR_ACT_GOTO && rule->ctarget == NULL)
 		frh->flags |= FIB_RULE_UNRESOLVED;
 
-	if (rule->ifname[0])
+	if (rule->ifname[0]) {
 		NLA_PUT_STRING(skb, FRA_IFNAME, rule->ifname);
 
+		if (rule->ifindex == -1)
+			frh->flags |= FIB_RULE_DEV_DETACHED;
+	}
+
 	if (rule->pref)
 		NLA_PUT_U32(skb, FRA_PRIORITY, rule->pref);
 

--

[PATCH 3/3] [NET] fib_rules: Add no-operation action

[Thomas Graf]

[-- Attachment #1: rule_nop --]
[-- Type: text/plain, Size: 1226 bytes --]

The use of nop rules simplifies the usage of goto rules
and adds more flexibility as they allow targets to remain
while the actual content of the branches can change easly.

Signed-off-by: Thomas Graf <tgraf@suug.ch>

Index: net-2.6.22/include/linux/fib_rules.h
===================================================================
--- net-2.6.22.orig/include/linux/fib_rules.h	2007-03-27 01:45:56.000000000 +0200
+++ net-2.6.22/include/linux/fib_rules.h	2007-03-27 01:46:33.000000000 +0200
@@ -54,7 +54,7 @@ enum
 	FR_ACT_UNSPEC,
 	FR_ACT_TO_TBL,		/* Pass to fixed table */
 	FR_ACT_GOTO,		/* Jump to another rule */
-	FR_ACT_RES2,
+	FR_ACT_NOP,		/* No operation */
 	FR_ACT_RES3,
 	FR_ACT_RES4,
 	FR_ACT_BLACKHOLE,	/* Drop without notification */
Index: net-2.6.22/net/core/fib_rules.c
===================================================================
--- net-2.6.22.orig/net/core/fib_rules.c	2007-03-27 01:46:22.000000000 +0200
+++ net-2.6.22/net/core/fib_rules.c	2007-03-27 01:46:33.000000000 +0200
@@ -146,7 +146,9 @@ jumped:
 				rule = target;
 				goto jumped;
 			}
-		} else
+		} else if (rule->action == FR_ACT_NOP)
+			continue;
+		else
 			err = ops->action(rule, fl, flags, arg);
 
 		if (err != -EAGAIN) {

--

Re: [PATCH 1/3] [NET] fib_rules: goto rule action

[David Miller]

From: Thomas Graf <tgraf@suug.ch>
Date: Tue, 27 Mar 2007 01:54:51 +0200

> This patch adds a new rule action FR_ACT_GOTO which allows
> to skip a set of rules by jumping to another rule. The rule
> to jump to is specified via the FRA_GOTO attribute which
> carries a rule preference.
> 
> Referring to a rule which doesn't exists is explicitely allowed.
> Such goto rules are marked with the flag FIB_RULE_UNRESOLVED
> and will act like a rule with a non-matching selector. The rule
> will become functional as soon as its target is present.
> 
> The goto action enables performance optimizations by reducing
> the average number of rules that have to be passed per lookup.
> 
> Example:
> 0:      from all lookup local 
> 40:     not from all to 192.168.23.128 goto 32766
> 41:     from all fwmark 0xa blackhole
> 42:     from all fwmark 0xff blackhole
> 32766:  from all lookup main 
> 
> Signed-off-by: Thomas Graf <tgraf@suug.ch>

This looks excellent, applied to net-2.6.22, thanks Thomas.

Re: [PATCH 2/3] [NET] fib_rules: Mark rules detached from the device

[David Miller]

From: Thomas Graf <tgraf@suug.ch>
Date: Tue, 27 Mar 2007 01:54:52 +0200

> Rules which match against device names in their selector can
> remain while the device itself disappears, in fact the device
> doesn't have to present when the rule is added in the first
> place. The device name is resolved by trying when the rule is
> added and later by listening to NETDEV_REGISTER/UNREGISTER
> notifications.
> 
> This patch adds the flag FIB_RULE_DEV_DETACHED which is set
> towards userspace when a rule contains a device match which
> is unresolved at the moment. This eases spotting the reason
> why certain rules seem not to function properly.
> 
> Signed-off-by: Thomas Graf <tgraf@suug.ch>

Yep, this should make failure diagnosis easier.

Applied to net-2.6.22, thanks.

Re: [PATCH 3/3] [NET] fib_rules: Add no-operation action

[David Miller]

From: Thomas Graf <tgraf@suug.ch>
Date: Tue, 27 Mar 2007 01:54:53 +0200

> The use of nop rules simplifies the usage of goto rules
> and adds more flexibility as they allow targets to remain
> while the actual content of the branches can change easly.
> 
> Signed-off-by: Thomas Graf <tgraf@suug.ch>

This also looks good, applied to net-2.6.22

Re: [PATCH 1/3] [NET] fib_rules: goto rule action

[Stephen Hemminger]

David Miller wrote:
> From: Thomas Graf <tgraf@suug.ch>
> Date: Tue, 27 Mar 2007 01:54:51 +0200
>
>   
>> This patch adds a new rule action FR_ACT_GOTO which allows
>> to skip a set of rules by jumping to another rule. The rule
>> to jump to is specified via the FRA_GOTO attribute which
>> carries a rule preference.
>>
>> Referring to a rule which doesn't exists is explicitely allowed.
>> Such goto rules are marked with the flag FIB_RULE_UNRESOLVED
>> and will act like a rule with a non-matching selector. The rule
>> will become functional as soon as its target is present.
>>
>> The goto action enables performance optimizations by reducing
>> the average number of rules that have to be passed per lookup.
>>
>> Example:
>> 0:      from all lookup local 
>> 40:     not from all to 192.168.23.128 goto 32766
>> 41:     from all fwmark 0xa blackhole
>> 42:     from all fwmark 0xff blackhole
>> 32766:  from all lookup main 
>>
>> Signed-off-by: Thomas Graf <tgraf@suug.ch>
>>     
>
> This looks excellent, applied to net-2.6.22, thanks Thomas.
> -
> To unsubscribe from this list: send the line "unsubscribe netdev" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>   

Do we have to worry about self inflicted infinite loops?
Shouldn't there be some way to some basic validation?

Re: [PATCH 1/3] [NET] fib_rules: goto rule action

[David Miller]

From: Stephen Hemminger <shemminger@osdl.org>
Date: Mon, 26 Mar 2007 21:13:07 -0700

> Do we have to worry about self inflicted infinite loops?
> Shouldn't there be some way to some basic validation?

We do, and there is.  You obviously didn't read Thomas's patch.

He prevents loops by only allowing forward gotos.