TAHI testing fails for IPv6 Fragments in Kernel 4.9

TAHI testing fails for IPv6 Fragments in Kernel 4.9

[Captain Wiggum]

We are using the TAHI Self-test tools from IPv6 Ready Logo Program:
https://www.ipv6ready.org/?page=documents&tag=ipv6-core-protocols

The test passed up to 4.9.133, then fails ever since.

The are about 20 failing test cases regarding IPv6 fragments,
where the kernel is issuing an ICMPv6 parameter problem pointing
to the Fragmentation Header.

I see lots of commits regarding improving fragment processing.
Has anyone else run TAHI tests on kernel 4.9 or later?
Is there any interest in looking into this to improve
the IPv6 functionality?

Any tips appreciated.

--John Masinter

Re: TAHI testing fails for IPv6 Fragments in Kernel 4.9

[David Miller]

From: Captain Wiggum <captwiggum@gmail.com>
Date: Wed, 6 Mar 2019 15:26:43 -0700

> We are using the TAHI Self-test tools from IPv6 Ready Logo Program:
> https://www.ipv6ready.org/?page=documents&tag=ipv6-core-protocols
> 
> The test passed up to 4.9.133, then fails ever since.
> 
> The are about 20 failing test cases regarding IPv6 fragments,
> where the kernel is issuing an ICMPv6 parameter problem pointing
> to the Fragmentation Header.
> 
> I see lots of commits regarding improving fragment processing.
> Has anyone else run TAHI tests on kernel 4.9 or later?
> Is there any interest in looking into this to improve
> the IPv6 functionality?

It is intentionally failing those tests to fix a denial of service
issue with ipv6 fragmentation and this will therefore not be changed.

Re: TAHI testing fails for IPv6 Fragments in Kernel 4.9

[Eric Dumazet]

On 03/06/2019 02:28 PM, David Miller wrote:
> From: Captain Wiggum <captwiggum@gmail.com>
> Date: Wed, 6 Mar 2019 15:26:43 -0700
> 
>> We are using the TAHI Self-test tools from IPv6 Ready Logo Program:
>> https://www.ipv6ready.org/?page=documents&tag=ipv6-core-protocols
>>
>> The test passed up to 4.9.133, then fails ever since.
>>
>> The are about 20 failing test cases regarding IPv6 fragments,
>> where the kernel is issuing an ICMPv6 parameter problem pointing
>> to the Fragmentation Header.
>>
>> I see lots of commits regarding improving fragment processing.
>> Has anyone else run TAHI tests on kernel 4.9 or later?
>> Is there any interest in looking into this to improve
>> the IPv6 functionality?
> 
> It is intentionally failing those tests to fix a denial of service
> issue with ipv6 fragmentation and this will therefore not be changed.


That was Florian patch that was later reverted, right ?

0ed4229b08c13c84a3c301a08defdc9e7f4467e6 ipv6: defrag: drop non-last frags smaller than min mtu

-> reverted in 

d4289fcc9b16b89619ee1c54f829e05e56de8b9a net: IP6 defrag: use rbtrees for IPv6 defrag

Re: TAHI testing fails for IPv6 Fragments in Kernel 4.9

[David Miller]

From: Eric Dumazet <eric.dumazet@gmail.com>
Date: Wed, 6 Mar 2019 14:34:25 -0800

> That was Florian patch that was later reverted, right ?
> 
> 0ed4229b08c13c84a3c301a08defdc9e7f4467e6 ipv6: defrag: drop non-last frags smaller than min mtu
> 
> -> reverted in 
> 
> d4289fcc9b16b89619ee1c54f829e05e56de8b9a net: IP6 defrag: use rbtrees for IPv6 defrag

Oh yes, that's right.

Re: TAHI testing fails for IPv6 Fragments in Kernel 4.9

[Captain Wiggum]

Thank you all for the reply.
The problem started with 4.9.133, and persists to latest 4.9 today.
I will work on the bisect approach to find the bad commit(s).
It will take a few days. I will reply back when I have more info.

On Wed, Mar 6, 2019 at 3:34 PM Eric Dumazet <eric.dumazet@gmail.com> wrote:
>
>
>
> On 03/06/2019 02:28 PM, David Miller wrote:
> > From: Captain Wiggum <captwiggum@gmail.com>
> > Date: Wed, 6 Mar 2019 15:26:43 -0700
> >
> >> We are using the TAHI Self-test tools from IPv6 Ready Logo Program:
> >> https://www.ipv6ready.org/?page=documents&tag=ipv6-core-protocols
> >>
> >> The test passed up to 4.9.133, then fails ever since.
> >>
> >> The are about 20 failing test cases regarding IPv6 fragments,
> >> where the kernel is issuing an ICMPv6 parameter problem pointing
> >> to the Fragmentation Header.
> >>
> >> I see lots of commits regarding improving fragment processing.
> >> Has anyone else run TAHI tests on kernel 4.9 or later?
> >> Is there any interest in looking into this to improve
> >> the IPv6 functionality?
> >
> > It is intentionally failing those tests to fix a denial of service
> > issue with ipv6 fragmentation and this will therefore not be changed.
>
>
> That was Florian patch that was later reverted, right ?
>
> 0ed4229b08c13c84a3c301a08defdc9e7f4467e6 ipv6: defrag: drop non-last frags smaller than min mtu
>
> -> reverted in
>
> d4289fcc9b16b89619ee1c54f829e05e56de8b9a net: IP6 defrag: use rbtrees for IPv6 defrag
>
>
>

Re: TAHI testing fails for IPv6 Fragments in Kernel 4.9

[Captain Wiggum]

Hi All,

To summarize this thread, we test for IPv6 Ready Logo using Self-test
Tools (TAHI Project) here:
https://www.ipv6ready.org/?page=documents&tag=ipv6-core-protocols

4.9.133 and previous passed 100%. Beginning with 4.9.134 it fails 18
tests related to IPv6 fragment headers.
I have done a git bisect, built and tested 7 revisions between 4.9.133
- 4.9.134, this is the final bisect result:
------
$ git bisect bad
a8444b1ccb20339774af58e40ad42296074fb484 is the first bad commit
commit a8444b1ccb20339774af58e40ad42296074fb484
Author: Florian Westphal <fw@strlen.de>
Date:   Wed Oct 10 12:30:10 2018 -0700
ipv6: defrag: drop non-last frags smaller than min mtu
------
Also I believe it fails in all kernels since then, 4.19, stable, etc.
Would it be possible for someone to look at that?
Any tips appreciated.

--John Masinter

Re: TAHI testing fails for IPv6 Fragments in Kernel 4.9

[Captain Wiggum]

In case it is useful, this was the bisect test progress:

my-build-name     test   commit/release
----------------  -----  --------------
kernel-4.9.133-1  good   [deb3303f665b31c29210ef4b30b1e69cb06cc397]
Linux 4.9.133
kernel-4.9.134-1  bad    [1d326a94d1039e4543edd97cbaf0fc38d2cefbb5]
Linux 4.9.134
kernel-4.9.133-2  good   [7f10b7e254b3e30bcac471de29e562aff4e22c60]
net: dsa: bcm_sf2: Fix unbind ordering
kernel-4.9.133-3  good   [7f6170683223cb38cabaff21ecbb9a6375ad10f6]
inet: frags: break the 2GB limit for frags storage
kernel-4.9.133-4  good   [871695951ec6f6b0b1a258c9bb5336bfeffab409]
net: modify skb_rbtree_purge to return the truesize
kernel-4.9.133-5  bad    [10043954eadac2d8f8c1886190f7a7ee584ff939]
ip: use rb trees for IP frag queue.
kernel-4.9.133-6  bad    [791521e2e377f66ef5ee6e5002dec758234d8d32]
net: pskb_trim_rcsum() and CHECKSUM_COMPLETE are friends
kernel-4.9.133-7  bad    [a8444b1ccb20339774af58e40ad42296074fb484]
ipv6: defrag: drop non-last frags smaller than min mtu
$ git bisect bad
74fb484 is the first bad commit

On Mon, Mar 11, 2019 at 10:12 PM Captain Wiggum <captwiggum@gmail.com> wrote:
>
> Hi All,
>
> To summarize this thread, we test for IPv6 Ready Logo using Self-test
> Tools (TAHI Project) here:
> https://www.ipv6ready.org/?page=documents&tag=ipv6-core-protocols
>
> 4.9.133 and previous passed 100%. Beginning with 4.9.134 it fails 18
> tests related to IPv6 fragment headers.
> I have done a git bisect, built and tested 7 revisions between 4.9.133
> - 4.9.134, this is the final bisect result:
> ------
> $ git bisect bad
> a8444b1ccb20339774af58e40ad42296074fb484 is the first bad commit
> commit a8444b1ccb20339774af58e40ad42296074fb484
> Author: Florian Westphal <fw@strlen.de>
> Date:   Wed Oct 10 12:30:10 2018 -0700
> ipv6: defrag: drop non-last frags smaller than min mtu
> ------
> Also I believe it fails in all kernels since then, 4.19, stable, etc.
> Would it be possible for someone to look at that?
> Any tips appreciated.
>
> --John Masinter

Re: TAHI testing fails for IPv6 Fragments in Kernel 4.9

[Eric Dumazet]

On 03/11/2019 09:12 PM, Captain Wiggum wrote:
> Hi All,
> 
> To summarize this thread, we test for IPv6 Ready Logo using Self-test
> Tools (TAHI Project) here:
> https://www.ipv6ready.org/?page=documents&tag=ipv6-core-protocols
> 
> 4.9.133 and previous passed 100%. Beginning with 4.9.134 it fails 18
> tests related to IPv6 fragment headers.
> I have done a git bisect, built and tested 7 revisions between 4.9.133
> - 4.9.134, this is the final bisect result:
> ------
> $ git bisect bad
> a8444b1ccb20339774af58e40ad42296074fb484 is the first bad commit
> commit a8444b1ccb20339774af58e40ad42296074fb484
> Author: Florian Westphal <fw@strlen.de>
> Date:   Wed Oct 10 12:30:10 2018 -0700
> ipv6: defrag: drop non-last frags smaller than min mtu
> ------
> Also I believe it fails in all kernels since then, 4.19, stable, etc.
> Would it be possible for someone to look at that?
> Any tips appreciated.
> 
> --John Masinter
> 

Hi John

Yes, this is basically what I answered last week about this problem,
there was no need for a bisection :/

This has been solved in upstream kernels with this patch, not yet backported to v4.9.xxx

d4289fcc9b16b89619ee1c54f829e05e56de8b9a net: IP6 defrag: use rbtrees for IPv6 defrag

Re: TAHI testing fails for IPv6 Fragments in Kernel 4.9

[Captain Wiggum]

Okay, then I misunderstood your earlier comment.
And guess it is not backported to latest 4.19 as that also fails.
Just to sanity check, I'll test mainline to satisfy my curiosity.
Thank you all for your time and replies!

On Mon, Mar 11, 2019 at 10:20 PM Eric Dumazet <eric.dumazet@gmail.com> wrote:
>
>
>
> On 03/11/2019 09:12 PM, Captain Wiggum wrote:
> > Hi All,
> >
> > To summarize this thread, we test for IPv6 Ready Logo using Self-test
> > Tools (TAHI Project) here:
> > https://www.ipv6ready.org/?page=documents&tag=ipv6-core-protocols
> >
> > 4.9.133 and previous passed 100%. Beginning with 4.9.134 it fails 18
> > tests related to IPv6 fragment headers.
> > I have done a git bisect, built and tested 7 revisions between 4.9.133
> > - 4.9.134, this is the final bisect result:
> > ------
> > $ git bisect bad
> > a8444b1ccb20339774af58e40ad42296074fb484 is the first bad commit
> > commit a8444b1ccb20339774af58e40ad42296074fb484
> > Author: Florian Westphal <fw@strlen.de>
> > Date:   Wed Oct 10 12:30:10 2018 -0700
> > ipv6: defrag: drop non-last frags smaller than min mtu
> > ------
> > Also I believe it fails in all kernels since then, 4.19, stable, etc.
> > Would it be possible for someone to look at that?
> > Any tips appreciated.
> >
> > --John Masinter
> >
>
> Hi John
>
> Yes, this is basically what I answered last week about this problem,
> there was no need for a bisection :/
>
> This has been solved in upstream kernels with this patch, not yet backported to v4.9.xxx
>
> d4289fcc9b16b89619ee1c54f829e05e56de8b9a net: IP6 defrag: use rbtrees for IPv6 defrag
>

Re: TAHI testing fails for IPv6 Fragments in Kernel 4.9

[Florian Fainelli]

On 3/6/19 2:26 PM, Captain Wiggum wrote:
> We are using the TAHI Self-test tools from IPv6 Ready Logo Program:
> https://www.ipv6ready.org/?page=documents&tag=ipv6-core-protocols
> 
> The test passed up to 4.9.133, then fails ever since.
> 
> The are about 20 failing test cases regarding IPv6 fragments,
> where the kernel is issuing an ICMPv6 parameter problem pointing
> to the Fragmentation Header.
> 
> I see lots of commits regarding improving fragment processing.
> Has anyone else run TAHI tests on kernel 4.9 or later?
> Is there any interest in looking into this to improve
> the IPv6 functionality?

Can you run a git bisection to determine when things started to fail
and/if a newer 4.9 release does improve the situation?
-- 
Florian