[Python-Dev] Need help to fix HTTP Header Injection vulnerability (original) (raw)

Gregory P. Smith greg at krypto.org
Tue Apr 9 20:41:27 EDT 2019

On Tue, Apr 9, 2019 at 4:45 PM Karthikeyan <tir.karthi at gmail.com> wrote:

I would recommend fixing it since it's potentially remote code execution on systems like Redis (latest versions of Redis have this mitigated) though I must admit I don't fully understand the complexity since there are multiple issues linked. Go was also assigned a CVE for linked issue and it seemed to be the same reporter by username : CVE-2019-9741 . I tried using go's approach in the commit but urlopen accepts more URLs like data URLs [0] that seemed to accept \n as a valid case and the patch broke some tests. Looking at the issue discussion complexity also involves backwards compatibility. golang also pushed an initial fix that seemed to broke their internal tests [0] to arrive at a more simpler fix.

[0] https://github.com/python/cpython/blob/a40681dd5db8deaf05a635eecb91498dac882aa4/Lib/test/testurllib.py#L482 [1] https://go-review.googlesource.com/c/go/+/159157/2#message-39c6be13a192bf760f6318ac641b432a6ab8fdc8 -- Regards, Karthikeyan S

useful references, thanks! limiting the checks to only http and https as those are the text based protocols with urls transmitted in text form makes sense and avoids the data: test failures.

proposed simple fix in https://github.com/python/cpython/pull/12755

but tests are needed as is an audit of the code to see where else we may potentially need to do such things.

-gps -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://mail.python.org/pipermail/python-dev/attachments/20190409/22802cd3/attachment.html>

More information about the Python-Dev mailing list