Skip to contentSkip to conversation

Adjust RTATTR_MAX to IFLA_* changes

PATCH In review Thomas Graf 2 participants 4 messages Sep 9, 2004

Subsystems (2)
Networking Network drivers
lore
Share & tools
lore

Thomas Graf

author
#0 lore raw

PATCH 1/1 Adjust RTATTR_MAX to IFLA_* changes

IFLA_MAX is now bigger than RTA_MAX with the latest changes in IFLA_*. Adjust RTATTR to point to IFLA_MAX to have big enough buffer in rtnetlink_rcv_msg.

Signed-off-by Thomas Graf <[email protected]>
include/linux/rtnetlink.h +1 -1
@@ -662,7 +662,7 @@
 
 /* SUMMARY: maximal rtattr understood by kernel */
 
-#define RTATTR_MAX		RTA_MAX
+#define RTATTR_MAX		IFLA_MAX
 
 /* RTnetlink multicast groups */
maintainer
#1 lore raw
Quoted 3 lines · On Thu, 9 Sep 2004 18:31:31 +0200 Thomas Graf <[email protected]> wrote:

IFLA_MAX is now bigger than RTA_MAX with the latest changes in IFLA_*. Adjust RTATTR to point to IFLA_MAX to have big enough buffer in rtnetlink_rcv_msg.

Ok, this is going to get silly if some other attribute gets larger than IFLA_* next.

We should just compute this thing at run time. Else, if the macro can't be deleted because there are non-trivial other uses, use some expression involving max().

I just checked iproute2 sources and it does not reference this thing, so probably best to kill it off, like so.

===== include/linux/rtnetlink.h 1.42 vs edited =====

include/linux/rtnetlink.h +0 -4
@@ -660,10 +660,6 @@
 #define TCA_PAYLOAD(n) NLMSG_PAYLOAD(n,sizeof(struct tcmsg))
 
 
-/* SUMMARY: maximal rtattr understood by kernel */
-
-#define RTATTR_MAX		RTA_MAX
-
 /* RTnetlink multicast groups */
 
 #define RTMGRP_LINK		1

===== net/core/rtnetlink.c 1.24 vs edited =====

net/core/rtnetlink.c +17 -5
@@ -401,6 +401,10 @@
 	return 0;
 }
 
+/* Protected by RTNL sempahore.  */
+static struct rtattr **rta_buf;
+static int rtattr_max;
+
 /* Process one rtnetlink message. */
 
 static __inline__ int

@@ -408,8 +412,6 @@
 {
 	struct rtnetlink_link *link;
 	struct rtnetlink_link *link_tab;
-	struct rtattr	*rta[RTATTR_MAX];
-
 	int sz_idx, kind;
 	int min_len;
 	int family;

@@ -476,7 +478,7 @@
 		return -1;
 	}
 
-	memset(&rta, 0, sizeof(rta));
+	memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
 
 	min_len = rtm_min[sz_idx];
 	if (nlh->nlmsg_len < min_len)

@@ -491,7 +493,7 @@
 			if (flavor) {
 				if (flavor > rta_max[sz_idx])
 					goto err_inval;
-				rta[flavor-1] = attr;
+				rta_buf[flavor-1] = attr;
 			}
 			attr = RTA_NEXT(attr, attrlen);
 		}

@@ -501,7 +503,7 @@
 		link = &(rtnetlink_links[PF_UNSPEC][type]);
 	if (link->doit == NULL)
 		goto err_inval;
-	err = link->doit(skb, nlh, (void *)&rta);
+	err = link->doit(skb, nlh, (void *)&rta_buf[0]);
 
 	*errp = err;
 	return err;

@@ -621,6 +623,16 @@
 
 void __init rtnetlink_init(void)
 {
+	int i;
+
+	rtattr_max = 0;
+	for (i = 0; i < ARRAY_SIZE(rta_max); i++)
+		if (rta_max[i] > rtattr_max)
+			rtattr_max = rta_max[i];
+	rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
+	if (!rta_buf)
+		panic("rtnetlink_init: cannot allocate rta_buf\n");
+
 	rtnl = netlink_kernel_create(NETLINK_ROUTE, rtnetlink_rcv);
 	if (rtnl == NULL)
 		panic("rtnetlink_init: cannot initialize rtnetlink\n");
author
#2 lore raw

* David S. Miller <[email protected]> 2004-09-09 10:31

Quoted 10 lines
Quoted 3 lines · On Thu, 9 Sep 2004 18:31:31 +0200 Thomas Graf <[email protected]> wrote:

IFLA_MAX is now bigger than RTA_MAX with the latest changes in IFLA_*. Adjust RTATTR to point to IFLA_MAX to have big enough buffer in rtnetlink_rcv_msg.

Ok, this is going to get silly if some other attribute gets larger than IFLA_* next.

Quoted 7 lines

We should just compute this thing at run time. Else, if the macro can't be deleted because there are non-trivial other uses, use some expression involving max().

I just checked iproute2 sources and it does not reference this thing, so probably best to kill it off, like so.

This is even better, works fine for me.

maintainer
#3 lore raw
Quoted 5 lines · On Thu, 9 Sep 2004 19:49:43 +0200 Thomas Graf <[email protected]> wrote:

* David S. Miller <[email protected]> 2004-09-09 10:31

Quoted 2 lines

I just checked iproute2 sources and it does not reference this thing, so probably best to kill it off, like so.

This is even better, works fine for me.

Great, this is what I'll push upstream.