Skip to contentSkip to conversation

ipv6: Fix preferred_lft not updating in some cases

PATCH Applied Paul Marks 3 participants 5 messages Sep 25, 2013 – Sep 30, 2013

Subsystems (3)
lore
Share & tools
lore

Paul Marks

author
#0 lore raw

PATCH 1/1 ipv6: Fix preferred_lft not updating in some cases

Consider the scenario where an IPv6 router is advertising a fixed preferred_lft of 1800 seconds, while the valid_lft begins at 3600 seconds and counts down in realtime.

A client should reset its preferred_lft to 1800 every time the RA is received, but a bug is causing Linux to ignore the update.

The core problem is here:
if (prefered_lft != ifp->prefered_lft) {

Note that ifp->prefered_lft is an offset, so it doesn't decrease over time. Thus, the comparison is always (1800 != 1800), which fails to trigger an update.

The most direct solution would be to compute a "stored_prefered_lft", and use that value in the comparison. But I think that trying to filter out unnecessary updates here is a premature optimization. In order for the filter to apply, both of these would need to hold:

- The advertised valid_lft and preferred_lft are both declining in real time.
- No clock skew exists between the router & client.

So in this patch, I've set "update_lft = 1" unconditionally, which allows the surrounding code to be greatly simplified.

---
 net/ipv6/addrconf.c | 52 +++++++++++++++-------------------------------------
 1 file changed, 15 insertions(+), 37 deletions(-)
net/ipv6/addrconf.c +15 -37
index d6ff126..9a5052c 100644
@@ -2193,43 +2193,21 @@ ok:
 			else
 				stored_lft = 0;
 			if (!update_lft && !create && stored_lft) {
-				if (valid_lft > MIN_VALID_LIFETIME ||
-				    valid_lft > stored_lft)
-					update_lft = 1;
-				else if (stored_lft <= MIN_VALID_LIFETIME) {
-					/* valid_lft <= stored_lft is always true */
-					/*
-					 * RFC 4862 Section 5.5.3e:
-					 * "Note that the preferred lifetime of
-					 *  the corresponding address is always
-					 *  reset to the Preferred Lifetime in
-					 *  the received Prefix Information
-					 *  option, regardless of whether the
-					 *  valid lifetime is also reset or
-					 *  ignored."
-					 *
-					 *  So if the preferred lifetime in
-					 *  this advertisement is different
-					 *  than what we have stored, but the
-					 *  valid lifetime is invalid, just
-					 *  reset prefered_lft.
-					 *
-					 *  We must set the valid lifetime
-					 *  to the stored lifetime since we'll
-					 *  be updating the timestamp below,
-					 *  else we'll set it back to the
-					 *  minimum.
-					 */
-					if (prefered_lft != ifp->prefered_lft) {
-						valid_lft = stored_lft;
-						update_lft = 1;
-					}
-				} else {
-					valid_lft = MIN_VALID_LIFETIME;
-					if (valid_lft < prefered_lft)
-						prefered_lft = valid_lft;
-					update_lft = 1;
-				}
+				const u32 minimum_lft = min(
+					stored_lft, (u32)MIN_VALID_LIFETIME);
+				valid_lft = max(valid_lft, minimum_lft);
+
+				/* RFC4862 Section 5.5.3e:
+				 * "Note that the preferred lifetime of the
+				 *  corresponding address is always reset to
+				 *  the Preferred Lifetime in the received
+				 *  Prefix Information option, regardless of
+				 *  whether the valid lifetime is also reset or
+				 *  ignored."
+				 *
+				 * So we should always update prefered_lft here.
+				 */
+				update_lft = 1;
 			}
 
 			if (update_lft) {
Signature · 1 line
1.8.4
#1 lore raw
Quoted 19 lines · Paul Marks

Consider the scenario where an IPv6 router is advertising a fixed preferred_lft of 1800 seconds, while the valid_lft begins at 3600 seconds and counts down in realtime.

A client should reset its preferred_lft to 1800 every time the RA is received, but a bug is causing Linux to ignore the update.

The core problem is here:
if (prefered_lft != ifp->prefered_lft) {

Note that ifp->prefered_lft is an offset, so it doesn't decrease over time. Thus, the comparison is always (1800 != 1800), which fails to trigger an update.

The most direct solution would be to compute a "stored_prefered_lft", and use that value in the comparison. But I think that trying to filter out unnecessary updates here is a premature optimization. In order for the filter to apply, both of these would need to hold:

Quoted 6 lines

- The advertised valid_lft and preferred_lft are both declining in real time.
- No clock skew exists between the router & client.

So in this patch, I've set "update_lft = 1" unconditionally, which allows the surrounding code to be greatly simplified.

I actually find this much harder to follow when verifying against the RFC.

Quoted 31 lines
---
 net/ipv6/addrconf.c | 52 +++++++++++++++-------------------------------------
 1 file changed, 15 insertions(+), 37 deletions(-)
net/ipv6/addrconf.c +0 -18
index d6ff126..9a5052c 100644
@@ -2193,43 +2193,21 @@ ok:
 			else
 				stored_lft = 0;
 			if (!update_lft && !create && stored_lft) {
-				if (valid_lft > MIN_VALID_LIFETIME ||
-				    valid_lft > stored_lft)
-					update_lft = 1;
-				else if (stored_lft <= MIN_VALID_LIFETIME) {
-					/* valid_lft <= stored_lft is always true */
-					/*
-					 * RFC 4862 Section 5.5.3e:
-					 * "Note that the preferred lifetime of
-					 *  the corresponding address is always
-					 *  reset to the Preferred Lifetime in
-					 *  the received Prefix Information
-					 *  option, regardless of whether the
-					 *  valid lifetime is also reset or
-					 *  ignored."
-					 *
-					 *  So if the preferred lifetime in
-					 *  this advertisement is different
-					 *  than what we have stored, but the

net/ipv6/addrconf.c ok: · line 2223

Quoted 10 lines
net/ipv6/addrconf.c +0 -10
-					 *  valid lifetime is invalid, just
-					 *  reset prefered_lft.
-					 *
-					 *  We must set the valid lifetime
-					 *  to the stored lifetime since we'll
-					 *  be updating the timestamp below,
-					 *  else we'll set it back to the
-					 *  minimum.
-					 */
-					if (prefered_lft != ifp->prefered_lft) {

Wouldn't the easiest solution be to just drop this if and execute the two lines below unconditionally?

net/ipv6/addrconf.c ok: · line 2198

Quoted 12 lines
net/ipv6/addrconf.c +3 -9
-						valid_lft = stored_lft;
-						update_lft = 1;
-					}
-				} else {
-					valid_lft = MIN_VALID_LIFETIME;
-					if (valid_lft < prefered_lft)
-						prefered_lft = valid_lft;
-					update_lft = 1;
-				}
+				const u32 minimum_lft = min(
+					stored_lft, (u32)MIN_VALID_LIFETIME);
+				valid_lft = max(valid_lft, minimum_lft);

Quick question: Don't we need a prefered_lft = min(preferred_lft, valid_lft) here?

Quoted 5 lines
net/ipv6/addrconf.c +5 -0
+
+				/* RFC4862 Section 5.5.3e:
+				 * "Note that the preferred lifetime of the
+				 *  corresponding address is always reset to
+				 *  the Preferred Lifetime in the received

net/ipv6/addrconf.c ok: · line 2213

Quoted 10 lines
net/ipv6/addrconf.c +7 -0
+				 *  Prefix Information option, regardless of
+				 *  whether the valid lifetime is also reset or
+				 *  ignored."
+				 *
+				 * So we should always update prefered_lft here.
+				 */
+				update_lft = 1;
 			}
 
 			if (update_lft) {

Thanks,

Hannes

author
#2 lore raw
Quoted 5 lines · Hannes Frederic Sowa
Quoted 1 line · Paul Marks

- if (prefered_lft != ifp->prefered_lft) {

Wouldn't the easiest solution be to just drop this if and execute the two lines below unconditionally?

Yes, that's also correct. But is it not better to have simpler code than shorter diffs? Should we transliterate English to C, or think about what the algorithm is actually doing? The fact that this bug has gone unnoticed provides some evidence that the code may have been too complicated.

Quoted 6 lines

net/ipv6/addrconf.c ok: · line 2198

Quoted 3 lines
net/ipv6/addrconf.c +3 -0
+                             const u32 minimum_lft = min(
+                                     stored_lft, (u32)MIN_VALID_LIFETIME);
+                             valid_lft = max(valid_lft, minimum_lft);

Quick question: Don't we need a prefered_lft = min(preferred_lft, valid_lft) here?

The invariant is (preferred_lft <= valid_lft), and valid_lft can only get bigger, so I don't think there's a problem.

#3 lore raw
Quoted 5 lines · Paul Marks
Quoted 3 lines · Hannes Frederic Sowa
Quoted 1 line · Paul Marks

- if (prefered_lft != ifp->prefered_lft) {

Quoted 8 lines
Quoted 2 lines

Wouldn't the easiest solution be to just drop this if and execute the two lines below unconditionally?

Yes, that's also correct. But is it not better to have simpler code than shorter diffs? Should we transliterate English to C, or think about what the algorithm is actually doing? The fact that this bug has gone unnoticed provides some evidence that the code may have been too complicated.

I don't care about the length of diffs or shorter code. I would favour a transliteration here because it makes verification easier (at least for me). The algorithm is not that complex and I guess the bug has been unnoticed because nobody ran into problems and cared til now.

So, why not get rid of update_lft then?

Quoted 9 lines
Quoted 6 lines

net/ipv6/addrconf.c ok: · line 2198

Quoted 3 lines
net/ipv6/addrconf.c +3 -0
+                             const u32 minimum_lft = min(
+                                     stored_lft, (u32)MIN_VALID_LIFETIME);
+                             valid_lft = max(valid_lft, minimum_lft);

Quick question: Don't we need a prefered_lft = min(preferred_lft, valid_lft) here?

The invariant is (preferred_lft <= valid_lft), and valid_lft can only get bigger, so I don't think there's a problem.

Ah, I got confused. Missed in the last case that it got tested earlier in the function. Your code looks correct regarding every rule.

Thanks,

Hannes