netdev
[Top] [All Lists]

Re: Dump of TCP hang in established state

To: niv@xxxxxxxxxx
Subject: Re: Dump of TCP hang in established state
From: "David S. Miller" <davem@xxxxxxxxxx>
Date: Thu, 29 May 2003 14:52:46 -0700 (PDT)
Cc: gandalf@xxxxxxxxxxxxxx, netdev@xxxxxxxxxxx
In-reply-to: <3ED68090.5040903@xxxxxxxxxx>
References: <1054217010.757.39.camel@xxxxxxxxxxxxxx> <1054238230.709.1.camel@xxxxxxxxxxxxxx> <3ED68090.5040903@xxxxxxxxxx>
Sender: netdev-bounce@xxxxxxxxxxx
   From: Nivedita Singhvi <niv@xxxxxxxxxx>
   Date: Thu, 29 May 2003 14:50:08 -0700

   Martin Josefsson wrote:
   
   > I've just tested 2.5-bk and I can still reproduce it.
   > The wording of my mail might have been a bit wrong, I don't see complete
   > hangs, just ~2 minute stalls in ESTABLISHED state and then it continues
   > like nothing happened.
   > 
   
   This actually addresses a situation of memory corruption
   and illegal reference panics, correct?

It prevents TCP input from "reading" bogus memory when comparing
time_wait buckets to the saddr/sport/daddr/dport of an incoming
packet.

By bogus memory I mean it's some large offset from the base of the
struct tcp_tw_bucket structure, past it's end.

<Prev in Thread] Current Thread [Next in Thread>