gnutls-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: address@hidden: Bug#183176: libgnutls5: Crypts wrong on alpha]


From: Nikos Mavroyanopoulos
Subject: Re: address@hidden: Bug#183176: libgnutls5: Crypts wrong on alpha]
Date: Mon Mar 3 22:55:02 2003
User-agent: Mutt/1.5.3i

On Mon, Mar 03, 2003 at 06:51:48PM +0100, Ivo Timmermans wrote:

Does the attached patch fix the problem? (it should)


> ----- Forwarded message from "Andreas U. Trottmann" <address@hidden> -----
> 
> Subject: Bug#183176: libgnutls5: Crypts wrong on alpha
> Reply-To: "Andreas U. Trottmann" <address@hidden>,
>       address@hidden
> From: "Andreas U. Trottmann" <address@hidden>
> To: Debian Bug Tracking System <address@hidden>
> Date: Mon, 03 Mar 2003 01:40:50 +0100
> X-Spam-Status: No, hits=-6.5 required=5.0 
> tests=SENT_BY_BTS,FORGED_RCVD_FOUND,AWL version=2.20
> 
> Package: libgnutls5
> Version: 0.8.1-0mywoody1
> Severity: normal
> 
> On (at least) alpha, gnutls seems to be broken. While it generally can
> communicate fine for short transactions, after a couple of kilobytes of
> data transferred it either generates something the other side can't
> decode, or it can't decode something received by the other side.
> 
> I'm reporting the bug against a self-compiled backport of libgnutls5
> 0.8.1-1 to woody, but it also is present in (at least) the libgnutls3
> shipped with woody, and presumably also with the "official" sid 0.8.1-1.
> I can't test this for lack of a sid alpha system, however.
> 
> 
> The bug can be reproduced easily, for example using one of the following 
> methods:
> 
>  * read your mail on an alpha machine with mutt on an IMAP server over ssl. 
>    After some succesful reading you *will* get 
>    "tls_socket_read (Decryption of the TLS record packet has failed.)"
>    and your IMAP connection will be aborted
> 
> - or -
> 
>  * create a text file of some MB (for example uuencode your linux
>    kernel > bigfile). Then, on an i386 machine, run "gnutls-serv". 
>    On an alpha machine, run "gnutls-cli -p 5556 < bigfile i386.host.name"
>    You will get, after some successful data transmission, on the server:
>    "*** gnutls error[-24]: Decryption of the TLS record packet has failed. 
>    (recv)"
>    and on the client:
>    "*** Received corrupted data(-10) - server has terminated the connection 
>    abnormally"
> 
> 
> - or -
> 
>  * on any machine (tested: i386 and alpha): create a example certificate,
>    put it in a file "server.crt", then run "openssl s_server".
>    Then, on your alpha machine, run "gnutls-cli -p 4433 < bigfile
>    server.host.name"
>    On the server you will soon get
>    "21579:error:1408F455:SSL routines:SSL3_GET_RECORD:decryption failed
>    or bad record mac:s3_pkt.c:457:"
>    and on the client you will again get
>    "*** Received corrupted data(-9) - server has terminated the
>    connection abnormally"
>    
> 
> 
> To me, the facts that gnutls(alpha) to gnutls(i386) fails as well as
> gnutls(alpha) to openssl(alpha) makes it look like gnutls has some 
> issues on alpha, maybe regarding some effects of the 64 bit architecture.
> 
> Interestingly, gnutls(alpha) to gnutls(alpha) seems to work fine. So,
> apparently, the bug seems to affect encoding and decoding equally.
> 

-- 
Nikos Mavroyanopoulos

Attachment: koko
Description: Text document


reply via email to

[Prev in Thread] Current Thread [Next in Thread]