Fwd: TCP fragmentation broken in 3.6-rc ?

Hi,

It seems somehow something was broken in 3.6-rc.
I'm not sure exactly what though. The symptoms are:

 - TCP connection get "stuck" with data in send queue when doing
"large" transfers ( like typing 'ps ax' on a ssh connection )
 - Only happens on path where the PMTU is lower than the MTU of the interface
 - Is not present right after boot, it only appears 10-20min after
boot or so. (and that's inside the _same_ TCP connection, it works
fine at first and then in the same ssh session, it'll get stuck)
 - Definitely seems related to fragments somehow since I see a router
sending ICMP message saying fragmentation is needed.
 - Exact same setup works fine with kernel 3.5.1

Cheers,

    Sylvain

On Mon, 2012-08-20 at 16:56 +0200, Sylvain Munaut wrote:

Hi,

It seems somehow something was broken in 3.6-rc.
I'm not sure exactly what though. The symptoms are:

 - TCP connection get "stuck" with data in send queue when doing
"large" transfers ( like typing 'ps ax' on a ssh connection )
 - Only happens on path where the PMTU is lower than the MTU of the interface
 - Is not present right after boot, it only appears 10-20min after
boot or so. (and that's inside the _same_ TCP connection, it works
fine at first and then in the same ssh session, it'll get stuck)
 - Definitely seems related to fragments somehow since I see a router
sending ICMP message saying fragmentation is needed.
 - Exact same setup works fine with kernel 3.5.1

Cheers,

Could you make sure its not a pfmemalloc problem, by applying following
patch :

https://lkml.org/lkml/2012/8/20/182

Thanks

Hi,

Could you make sure its not a pfmemalloc problem, by applying following
patch :

https://lkml.org/lkml/2012/8/20/182

I just tried it, but unfortunately it doesn't fix the problem :(

It may have improved a bit (seems like it takes 20/25 min to break now
... but it might just be random luck)


Cheers,

    Sylvain

On Mon, 2012-08-20 at 17:52 +0200, Sylvain Munaut wrote:

Hi,

quoted

Could you make sure its not a pfmemalloc problem, by applying following
patch :

https://lkml.org/lkml/2012/8/20/182

I just tried it, but unfortunately it doesn't fix the problem :(

It may have improved a bit (seems like it takes 20/25 min to break now
... but it might just be random luck)

OK thanks

Any chance you could post a tcpdump demonstrating the problem ? 
(including the tcp session plus icmp messages)

Hi,

Any chance you could post a tcpdump demonstrating the problem ?
(including the tcp session plus icmp messages)

Here it is :

http://ge.tt/api/1/files/76YVbLM/0/blob?download


Cheers,

    Sylvain