Bug detected in buffer reuse of the eepro100b driver for linux

Pedro Ribeiro pribeiro@cc.isel.pt
Tue Aug 18 09:24:47 1998


Some days ago a detected a problem in two of my machines running Linux2.0.35
with EtherExpressPro100b netcards that i thinked that was kernel related,
then i sent this e-mail to the kernel mantainer (Alan Cox).

>While debugging a program with a low level network monitor i'v noticed that
>the implementation of (at least) TCP of (at least) Linux 2.0.35 uses as
>padding of some frames (usually ACK frames) some of the contents of the
last
>sent frames, shouldn't that be considered a unsecurity ?? Are linux reusing
>buffers as padding without the proper cleanup ?, At least i hope that this
>reuse is only within the same socket.
>
>I'm i saying something that worth read and fix in 2.0.36 ?
>
>Thanks for your time reading this. 8)

He answered with the message:

>From: Alan Cox <alan@cymru.net>
>Subject: Re: Possible BUG in Linux2.0.35
>To: pribeiro@cc.isel.pt
>Date: Fri, 7 Aug 1998 12:57:28 +0100 (BST)
>
>The Linux tcp/ip stack doesnt generate such padding at all. Some of the
>low level network drivers may be generating unclean trailing bytes however

Giving this direction, i'm reporting this problem to the mantainer of the
eepro100b Linux driver, are the packet padding beeing made with the reuse of
non cleaned buffers ??

Thanks, i hope this should help to fix one more detail/bug of the driver.

[]---------------------------------------------------------------[]
  Pedro Ribeiro
  Online: http://www.cc.isel.pt/Pessoais/PedroRibeiro
  IRC(PTnet) Nick: PAntMaR
  e-Mail: Personal:  pribeiro@cc.isel.pt
          Admin:     admin@isel.pt
  Tel:    +351-1-8317128 / Fax: +351-1-8317171
[]---------------------------------------------------------------[]