Mailing List Archive: Header reading problem

Hi,

I have another problem with a msg that just came in.
Some idiot sent a huge email, and amongst the half-dozen
attachments, was a video. Unfortunately, this attachment
was corrupted, so that the headers were elsewhere, resulting
in the attachment looking like this:

------=____1102359685177_mjnrnPN17R
CMCaydY5mzxGm2QCqAGLObOMCAAAAAAAABgAAAAECodyrjEepzxGO5ADADCBTZWgAAAAAAAAA
ki82QAAAxkEAABZ+QzAAAMEhIgAAAAAAAMeRDcpDlQB4AAAAAAAAABCuqhQAAAAAkOrgEgAA
AAC4CwAAAAAAAAIAAADISAAAyEgAAIC+FgCRB9y3t6nPEY7mAMAMIFNlcgAAAAAAAABAnmn4
TVvPEaj9AIBfXEQrUM3Dv49hzxGLsgCqALTiIAAAAAAAAAAAHAAAAAgAAAABAAAAAABhAQIA
gD4AANAHAACAAhAACgAAJAAADgCAAgAAAYACgAIBAACRB9y3t6nPEY7mAMAMIFNlgQAAAAAA
AADA7xm8TVvPEaj9AIBfXEQrAFf7IFVbzxGo/QCAX1xEKwAAAAAAAAAAMwAAAAAAAAACAAAA
AABAAQAA8AAAAAIoACgAAABAAQAA8AAAAAEAGABXTVYxAIQDAAAAAAAAAAAAAAAAAAAAAAC1

and so on, and so on. The problem is that when ripmime gets to
MIMEH_read_headers(), it keeps reading lines in and reallocing
space for it. So when MIMEH_are_headers_RFC822() is called, it
tries to do the strstr() looking for the headers on a *3 Meg* buffer.
Needless to say this takes a very long time, chewing up CPU. I have
code to check for hung processes, and it kills it after using 2 minutes
of CPU time.

Can the loop in MIMEH_read_headers() be modified to not add any
more text after a certain point? Shouldn't you be able to find
any headers in the first however-many KB, and fail if you don't?
I realize corrupted messages don't happen very often, but we still
need to deal with it gracefully when it does.

Brian Bebeau
Mycom Group, Inc.
bbebeau@mycom.com

----------------------------------------------------------------------------------------------
This message has been scanned by mycomPRO mailMAX II.
For further information, visit http://www.mycompro.com

_______________________________________________
Ripmime-general mailing list
Ripmime-general@pldaniels.com
http://www.pldaniels.com/mailman/listinfo/ripmime-general

Brian,

> Can the loop in MIMEH_read_headers() be modified to not add any
> more text after a certain point? Shouldn't you be able to find

One certainly could add such limit detection sysetms (After all, we have a ripMIME timeout already).

The problem with adding things like a Kb or count limit to header reading is that it then provides a possible vector for
people wanting to send malicious emails through, ie, buffer up the headers with XKb of junk, enough for ripMIME to give
up then put the real headers in. That said, yes, I will add some features for enabling a limits detection.

Paul.

--
PLDaniels - Software - Xamime
Unix systems Internet Development A.B.N. 19 500 721 806
PGP Public Key at http://www.pldaniels.com/gpg-keys.pld