Skip to contentSkip to conversation

pull request (net): ipsec 2013-11-01

PATCH In review Steffen Klassert 6 participants 17 messages Nov 1, 2013 – Dec 16, 2013

Subsystems (3)
lore
Share & tools
lore

Steffen Klassert

author maintainer
#0 lore raw

1) Fix a possible race on ipcomp scratch buffers because of too early enabled siftirqs. From Michal Kubecek.

2) The current xfrm garbage collector threshold is too small for some workloads, resulting in bad performance on these workloads. Increase the threshold from 1024 to 32768.

3) Some codepaths might not have a dst_entry attached to the skb when calling xfrm_decode_session(). So add a check to prevent a null pointer dereference in this case.

Please pull or let me know if there are problems.

Thanks!

The following changes since commit e9e4ea74f06635f2ffc1dffe5ef40c854faa0a90:

net: smc91x: dont't use SMC_outw for fixing up halfword-aligned data (2013-10-11 17:50:59 -0400)

are available in the git repository at:

git://git.kernel.org/pub/scm/linux/kernel/git/klassert/ipsec.git master

for you to fetch changes up to 84502b5ef9849a9694673b15c31bd3ac693010ae:

xfrm: Fix null pointer dereference when decoding sessions (2013-11-01 07:08:46 +0100)

----------------------------------------------------------------
Michal Kubecek (1):
xfrm: prevent ipcomp scratch buffer race condition

Steffen Klassert (2):
xfrm: Increase the garbage collector threshold
xfrm: Fix null pointer dereference when decoding sessions

 net/ipv4/xfrm4_policy.c |    8 ++++++--
 net/ipv6/xfrm6_policy.c |    8 ++++++--
 net/xfrm/xfrm_ipcomp.c  |   12 ++++++------
 3 files changed, 18 insertions(+), 10 deletions(-)
author maintainer
#1 lore raw

PATCH 1/3 xfrm: prevent ipcomp scratch buffer race condition

From: Michal Kubecek <[email protected]>

In ipcomp_compress(), sortirq is enabled too early, allowing the per-cpu scratch buffer to be rewritten by ipcomp_decompress() (called on the same CPU in softirq context) between populating the buffer and copying the compressed data to the skb.

v2: as pointed out by Steffen Klassert, if we also move the local_bh_disable() before reading the per-cpu pointers, we can get rid of get_cpu()/put_cpu().

v3: removed ipcomp_decompress part (as explained by Herbert Xu, it cannot be called from process context), get rid of cpu variable (thanks to Eric Dumazet)

---
 net/xfrm/xfrm_ipcomp.c |   12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)
net/xfrm/xfrm_ipcomp.c +6 -6
index 2906d52..3be02b6 100644
@@ -141,14 +141,14 @@ static int ipcomp_compress(struct xfrm_state *x, struct sk_buff *skb)
 	const int plen = skb->len;
 	int dlen = IPCOMP_SCRATCH_SIZE;
 	u8 *start = skb->data;
-	const int cpu = get_cpu();
-	u8 *scratch = *per_cpu_ptr(ipcomp_scratches, cpu);
-	struct crypto_comp *tfm = *per_cpu_ptr(ipcd->tfms, cpu);
+	struct crypto_comp *tfm;
+	u8 *scratch;
 	int err;
 
 	local_bh_disable();
+	scratch = *this_cpu_ptr(ipcomp_scratches);
+	tfm = *this_cpu_ptr(ipcd->tfms);
 	err = crypto_comp_compress(tfm, start, plen, scratch, &dlen);
-	local_bh_enable();
 	if (err)
 		goto out;
 
@@ -158,13 +158,13 @@ static int ipcomp_compress(struct xfrm_state *x, struct sk_buff *skb)
 	}
 
 	memcpy(start + sizeof(struct ip_comp_hdr), scratch, dlen);
-	put_cpu();
+	local_bh_enable();
 
 	pskb_trim(skb, dlen + sizeof(struct ip_comp_hdr));
 	return 0;
 
 out:
-	put_cpu();
+	local_bh_enable();
 	return err;
 }
 
Signature · 1 line
1.7.9.5
author maintainer
#2 lore raw

PATCH 2/3 xfrm: Increase the garbage collector threshold

In reply to Steffen Klassert

With the removal of the routing cache, we lost the option to tweak the garbage collector threshold along with the maximum routing cache size. So git commit 703fb94ec ("xfrm: Fix the gc threshold value for ipv4") moved back to a static threshold.

It turned out that the current threshold before we start garbage collecting is much to small for some workloads, so increase it from 1024 to 32768. This means that we start the garbage collector if we have more than 32768 dst entries in the system and refuse new allocations if we are above 65536.

---
 net/ipv4/xfrm4_policy.c |    2 +-
 net/ipv6/xfrm6_policy.c |    2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)
net/ipv4/xfrm4_policy.c +1 -1
index ccde542..4764ee4 100644
@@ -236,7 +236,7 @@ static struct dst_ops xfrm4_dst_ops = {
 	.destroy =		xfrm4_dst_destroy,
 	.ifdown =		xfrm4_dst_ifdown,
 	.local_out =		__ip_local_out,
-	.gc_thresh =		1024,
+	.gc_thresh =		32768,
 };
 
 static struct xfrm_policy_afinfo xfrm4_policy_afinfo = {
net/ipv6/xfrm6_policy.c +1 -1
index 08ed277..dd503a3 100644
@@ -285,7 +285,7 @@ static struct dst_ops xfrm6_dst_ops = {
 	.destroy =		xfrm6_dst_destroy,
 	.ifdown =		xfrm6_dst_ifdown,
 	.local_out =		__ip6_local_out,
-	.gc_thresh =		1024,
+	.gc_thresh =		32768,
 };
 
 static struct xfrm_policy_afinfo xfrm6_policy_afinfo = {
Signature · 1 line
1.7.9.5
#3 lore raw

Re: [PATCH 2/3] xfrm: Increase the garbage collector threshold

On Fri, 2013-11-01 at 09:21 +0100, Steffen Klassert wrote:

Hello Steffen,

Quoted 6 lines

It turned out that the current threshold before we start garbage collecting is much to small for some workloads, so increase it from 1024 to 32768. This means that we start the garbage collector if we have more than 32768 dst entries in the system and refuse new allocations if we are above 65536.

sorry to dig an old thread, but could you please explain the lifetime of those dst entries ?

After basic tests it seems the ipv4 dst entry needs to be released first before the xfrm dst can be released by gc.

Signature · 1 line
Maxime
author maintainer
#4 lore raw

Re: [PATCH 2/3] xfrm: Increase the garbage collector threshold

Quoted 12 lines · Maxime Bizon

On Fri, 2013-11-01 at 09:21 +0100, Steffen Klassert wrote:

Hello Steffen,

Quoted 6 lines

It turned out that the current threshold before we start garbage collecting is much to small for some workloads, so increase it from 1024 to 32768. This means that we start the garbage collector if we have more than 32768 dst entries in the system and refuse new allocations if we are above 65536.

Quoted 6 lines

sorry to dig an old thread, but could you please explain the lifetime of those dst entries ?

After basic tests it seems the ipv4 dst entry needs to be released first before the xfrm dst can be released by gc.

Can you please be a bit more precise with your problem description?

This patch changes only the number of cache entries before we start garbage collecting. It does not change anything on the garbage collector itself.

#5 lore raw

Re: [PATCH 2/3] xfrm: Increase the garbage collector threshold

Quoted 1 line · Steffen Klassert

Can you please be a bit more precise with your problem description?

yes sorry I wasn't clear.

I have a problem with an even simpler workload that the one using apache bench in the original bug report.

I am using ipsec transport mode between two hosts and just run this on one side:

while :; do wget -O /dev/null http://remote_host/; done

I was surprised to see it fails after only 1024 requests (ENOBUF on connect), and how long I had to wait to be able to do new requests.

After debugging I saw that the xfrm gc was called but was not able to release anything.

after running "ip route flush cache", which forces all ipv4 dst entries to be released, suddenly the xfrm gc had something to free, and xfrm entry count went to zero.

So if it is correct that once a ipv4 dst entry exists, the xfrm entry cannot be gc-ed, then we need to make sure we allow more xfrm entries to be allocated than ipv4 dst.

Quoted 3 lines

This patch changes only the number of cache entries before we start garbage collecting. It does not change anything on the garbage collector itself.

Yes my point was that it just hides another underlying problem.

Signature · 1 line
Maxime
author maintainer
#6 lore raw

Re: [PATCH 2/3] xfrm: Increase the garbage collector threshold

Quoted 18 lines · Maxime Bizon
Quoted 1 line · Steffen Klassert

Can you please be a bit more precise with your problem description?

yes sorry I wasn't clear.

I have a problem with an even simpler workload that the one using apache bench in the original bug report.

I am using ipsec transport mode between two hosts and just run this on one side:

while :; do wget -O /dev/null http://remote_host/; done

I was surprised to see it fails after only 1024 requests (ENOBUF on connect), and how long I had to wait to be able to do new requests.

Quoted 6 lines

After debugging I saw that the xfrm gc was called but was not able to release anything.

after running "ip route flush cache", which forces all ipv4 dst entries to be released, suddenly the xfrm gc had something to free, and xfrm entry count went to zero.

Well, "ip route flush cache" bumps the rt_genid what marks all routes as invalid. The xfrm garbage collector will notice this on the next run and deletes all invalid cached routes.

Garbage collecting is different from flushing, it only removes stale routes and keeps everything that was used recently. That's the whole point of this caching, we cache recently used IPsec routes to avoid a slow path lookup.

Quoted 4 lines

So if it is correct that once a ipv4 dst entry exists, the xfrm entry cannot be gc-ed, then we need to make sure we allow more xfrm entries to be allocated than ipv4 dst.

No, we don't cache plain ipv4 routes, we cache only IPsec routes. The original ipv4 dst_entry is cached together with the xfrm dst_entry as a xfrm bundle at the IPsec flow cache, so it does not make sense to ensure something like that.

#7 lore raw

Re: [PATCH 2/3] xfrm: Increase the garbage collector threshold

Quoted 12 lines · Steffen Klassert
Quoted 11 lines

while :; do wget -O /dev/null http://remote_host/; done

I was surprised to see it fails after only 1024 requests (ENOBUF on connect), and how long I had to wait to be able to do new requests.

After debugging I saw that the xfrm gc was called but was not able to release anything.

after running "ip route flush cache", which forces all ipv4 dst entries to be released, suddenly the xfrm gc had something to free, and xfrm entry count went to zero.

Quoted 8 lines

Well, "ip route flush cache" bumps the rt_genid what marks all routes as invalid. The xfrm garbage collector will notice this on the next run and deletes all invalid cached routes.

Garbage collecting is different from flushing, it only removes stale routes and keeps everything that was used recently. That's the whole point of this caching, we cache recently used IPsec routes to avoid a slow path lookup.

ok I don't know this part of the code at all.

but if it is only a *cache*, then it should not limit any new allocations for the reason that the cache is full, it should take slow path instead.

Signature · 1 line
Maxime
author maintainer
#8 lore raw

PATCH 3/3 xfrm: Fix null pointer dereference when decoding sessions

In reply to Steffen Klassert

On some codepaths the skb does not have a dst entry when xfrm_decode_session() is called. So check for a valid skb_dst() before dereferencing the device interface index. We use 0 as the device index if there is no valid skb_dst(), or at reverse decoding we use skb_iif as device interface index.

Bug was introduced with git commit bafd4bd4dc ("xfrm: Decode sessions with output interface.").

---
 net/ipv4/xfrm4_policy.c |    6 +++++-
 net/ipv6/xfrm6_policy.c |    6 +++++-
 2 files changed, 10 insertions(+), 2 deletions(-)
net/ipv4/xfrm4_policy.c +5 -1
index 4764ee4..e1a6393 100644
@@ -104,10 +104,14 @@ _decode_session4(struct sk_buff *skb, struct flowi *fl, int reverse)
 	const struct iphdr *iph = ip_hdr(skb);
 	u8 *xprth = skb_network_header(skb) + iph->ihl * 4;
 	struct flowi4 *fl4 = &fl->u.ip4;
+	int oif = 0;
+
+	if (skb_dst(skb))
+		oif = skb_dst(skb)->dev->ifindex;
 
 	memset(fl4, 0, sizeof(struct flowi4));
 	fl4->flowi4_mark = skb->mark;
-	fl4->flowi4_oif = skb_dst(skb)->dev->ifindex;
+	fl4->flowi4_oif = reverse ? skb->skb_iif : oif;
 
 	if (!ip_is_fragment(iph)) {
 		switch (iph->protocol) {
net/ipv6/xfrm6_policy.c +5 -1
index dd503a3..5f8e128 100644
@@ -135,10 +135,14 @@ _decode_session6(struct sk_buff *skb, struct flowi *fl, int reverse)
 	struct ipv6_opt_hdr *exthdr;
 	const unsigned char *nh = skb_network_header(skb);
 	u8 nexthdr = nh[IP6CB(skb)->nhoff];
+	int oif = 0;
+
+	if (skb_dst(skb))
+		oif = skb_dst(skb)->dev->ifindex;
 
 	memset(fl6, 0, sizeof(struct flowi6));
 	fl6->flowi6_mark = skb->mark;
-	fl6->flowi6_oif = skb_dst(skb)->dev->ifindex;
+	fl6->flowi6_oif = reverse ? skb->skb_iif : oif;
 
 	fl6->daddr = reverse ? hdr->saddr : hdr->daddr;
 	fl6->saddr = reverse ? hdr->daddr : hdr->saddr;
Signature · 1 line
1.7.9.5
#9 lore raw

Re: [PATCH 3/3] xfrm: Fix null pointer dereference when decoding sessions

Quoted 7 lines · Steffen Klassert

On some codepaths the skb does not have a dst entry when xfrm_decode_session() is called. So check for a valid skb_dst() before dereferencing the device interface index. We use 0 as the device index if there is no valid skb_dst(), or at reverse decoding we use skb_iif as device interface index.

Quoted 6 lines

Bug was introduced with git commit bafd4bd4dc ("xfrm: Decode sessions with output interface.").

Please queue this for 3.12-stable, I've seen this bug trigger a panic when heavily using Bittorrent (screenshot:
http://i.imgur.com/OIKVccM.jpg )

Thanks,
Matthias

#10 lore raw

Re: [PATCH 3/3] xfrm: Fix null pointer dereference when decoding sessions

Quoted 15 lines · Matthias Schiffer
Quoted 7 lines · Steffen Klassert

On some codepaths the skb does not have a dst entry when xfrm_decode_session() is called. So check for a valid skb_dst() before dereferencing the device interface index. We use 0 as the device index if there is no valid skb_dst(), or at reverse decoding we use skb_iif as device interface index.

Quoted 6 lines

Bug was introduced with git commit bafd4bd4dc ("xfrm: Decode sessions with output interface.").

Quoted 7 lines

Please queue this for 3.12-stable, I've seen this bug trigger a panic when heavily using Bittorrent (screenshot:
http://i.imgur.com/OIKVccM.jpg )

Thanks,
Matthias

Ping? Several people I know have hit this panic with 3.12.x at least once or twice by now during heavy IPv6 usage.

maintainer
#11 lore raw

Re: [PATCH 3/3] xfrm: Fix null pointer dereference when decoding sessions

From: Matthias Schiffer <[email protected]>
Date: Sun, 01 Dec 2013 05:19:53 +0100

Quoted 20 lines · Matthias Schiffer
Quoted 19 lines · Steffen Klassert
Quoted 7 lines

On some codepaths the skb does not have a dst entry when xfrm_decode_session() is called. So check for a valid skb_dst() before dereferencing the device interface index. We use 0 as the device index if there is no valid skb_dst(), or at reverse decoding we use skb_iif as device interface index.

Quoted 6 lines

Bug was introduced with git commit bafd4bd4dc ("xfrm: Decode sessions with output interface.").

Please queue this for 3.12-stable, I've seen this bug trigger a panic when heavily using Bittorrent (screenshot:
http://i.imgur.com/OIKVccM.jpg )

Quoted 6 lines
Quoted 3 lines

Thanks,
Matthias

Ping? Several people I know have hit this panic with 3.12.x at least once or twice by now during heavy IPv6 usage.

Steffen submitted it to [email protected] for inclusion on Monday November 25th.

author maintainer
#12 lore raw

Re: [PATCH 3/3] xfrm: Fix null pointer dereference when decoding sessions

Ccing stable.

Quoted 27 lines · David Miller

From: Matthias Schiffer <[email protected]>
Date: Sun, 01 Dec 2013 05:19:53 +0100

Quoted 24 lines · Matthias Schiffer
Quoted 22 lines · Steffen Klassert
Quoted 7 lines

On some codepaths the skb does not have a dst entry when xfrm_decode_session() is called. So check for a valid skb_dst() before dereferencing the device interface index. We use 0 as the device index if there is no valid skb_dst(), or at reverse decoding we use skb_iif as device interface index.

Quoted 6 lines

Bug was introduced with git commit bafd4bd4dc ("xfrm: Decode sessions with output interface.").

Please queue this for 3.12-stable, I've seen this bug trigger a panic when heavily using Bittorrent (screenshot:
http://i.imgur.com/OIKVccM.jpg )

Thanks,
Matthias

Quoted 5 lines
Quoted 2 lines

Ping? Several people I know have hit this panic with 3.12.x at least once or twice by now during heavy IPv6 usage.

Steffen submitted it to [email protected] for inclusion on Monday November 25th.

It apparently did not make it into v3.12.2 and it is not in the v3.12.3 review included. So I wonder if I need to do anything additional to get it included. It is the first time that I did a stable submission myself.

I tried to follow the instructions at Documentation/stable_kernel_rules.txt but maybe I've missed something.

#13 lore raw

Re: [PATCH 3/3] xfrm: Fix null pointer dereference when decoding sessions

Am Dienstag, 3. Dezember 2013, 10:56:45 schrieb Steffen Klassert:

Quoted 35 lines

Ccing stable.

Quoted 23 lines · David Miller

From: Matthias Schiffer <[email protected]>
Date: Sun, 01 Dec 2013 05:19:53 +0100

Quoted 20 lines · Matthias Schiffer
Quoted 19 lines · Steffen Klassert
Quoted 7 lines

On some codepaths the skb does not have a dst entry when xfrm_decode_session() is called. So check for a valid skb_dst() before dereferencing the device interface index. We use 0 as the device index if there is no valid skb_dst(), or at reverse decoding we use skb_iif as device interface index.

Quoted 6 lines

Bug was introduced with git commit bafd4bd4dc ("xfrm: Decode sessions with output interface.").

Please queue this for 3.12-stable, I've seen this bug trigger a panic when heavily using Bittorrent (screenshot:
http://i.imgur.com/OIKVccM.jpg )

Quoted 8 lines
Quoted 5 lines
Quoted 2 lines

Thanks,
Matthias

Ping? Several people I know have hit this panic with 3.12.x at least once or twice by now during heavy IPv6 usage.

Steffen submitted it to [email protected] for inclusion on Monday November 25th.

Quoted 7 lines

It apparently did not make it into v3.12.2 and it is not in the v3.12.3 review included. So I wonder if I need to do anything additional to get it included. It is the first time that I did a stable submission myself.

I tried to follow the instructions at Documentation/stable_kernel_rules.txt but maybe I've missed something.

There is another weakness with ipv6 in 3.12.x which - under certain circumstances - can be exploited remotely:

ipv6 fragmentation-related panic in netfilter

I would be very glad if this would make it into 3.12 stable soon.

Regards,

Signature · 3 lines
Wolfgang Walter
Studentenwerk München
Anstalt des öffentlichen Rechts
#14 lore raw

Re: [PATCH 3/3] xfrm: Fix null pointer dereference when decoding sessions

Quoted 44 lines · Wolfgang Walter

Am Dienstag, 3. Dezember 2013, 10:56:45 schrieb Steffen Klassert:

Quoted 35 lines

Ccing stable.

Quoted 23 lines · David Miller

From: Matthias Schiffer <[email protected]>
Date: Sun, 01 Dec 2013 05:19:53 +0100

Quoted 20 lines · Matthias Schiffer
Quoted 19 lines · Steffen Klassert
Quoted 7 lines

On some codepaths the skb does not have a dst entry when xfrm_decode_session() is called. So check for a valid skb_dst() before dereferencing the device interface index. We use 0 as the device index if there is no valid skb_dst(), or at reverse decoding we use skb_iif as device interface index.

Quoted 6 lines

Bug was introduced with git commit bafd4bd4dc ("xfrm: Decode sessions with output interface.").

Please queue this for 3.12-stable, I've seen this bug trigger a panic when heavily using Bittorrent (screenshot:
http://i.imgur.com/OIKVccM.jpg )

Quoted 8 lines
Quoted 5 lines
Quoted 2 lines

Thanks,
Matthias

Ping? Several people I know have hit this panic with 3.12.x at least once or twice by now during heavy IPv6 usage.

Steffen submitted it to [email protected] for inclusion on Monday November 25th.

Quoted 7 lines

It apparently did not make it into v3.12.2 and it is not in the v3.12.3 review included. So I wonder if I need to do anything additional to get it included. It is the first time that I did a stable submission myself.

I tried to follow the instructions at Documentation/stable_kernel_rules.txt but maybe I've missed something.

Quoted 6 lines

There is another weakness with ipv6 in 3.12.x which - under certain circumstances - can be exploited remotely:

ipv6 fragmentation-related panic in netfilter

I would be very glad if this would make it into 3.12 stable soon.

These patches are already queued up for stable by David:
http://patchwork.ozlabs.org/bundle/davem/stable/?state=*

Greetings,

Hannes

maintainer
#15 lore raw

Re: [PATCH 3/3] xfrm: Fix null pointer dereference when decoding sessions

In reply to Steffen Klassert

From: Steffen Klassert <[email protected]>
Date: Tue, 3 Dec 2013 10:56:45 +0100

Quoted 4 lines

It apparently did not make it into v3.12.2 and it is not in the v3.12.3 review included. So I wonder if I need to do anything additional to get it included. It is the first time that I did a stable submission myself.

I'm taking care of this myself now, don't worry any more about it.

Thanks.

maintainer
#16 lore raw
In reply to Steffen Klassert

From: Steffen Klassert <[email protected]>
Date: Fri, 1 Nov 2013 09:21:44 +0100

Quoted 7 lines

1) Fix a possible race on ipcomp scratch buffers because of too early enabled siftirqs. From Michal Kubecek.

2) The current xfrm garbage collector threshold is too small for some workloads, resulting in bad performance on these workloads. Increase the threshold from 1024 to 32768.

Quoted 5 lines

3) Some codepaths might not have a dst_entry attached to the skb when calling xfrm_decode_session(). So add a check to prevent a null pointer dereference in this case.

Please pull or let me know if there are problems.

Pulled, thanks a lot Steffen!