There is also code that disables locking for FIPS, that was the main part of 
TS-3576.  If you would like to submit a github pull request to create a 
configurable option that would enable FIPS and enable the locking that would be 
great.

I would also be in favor of having a configurable option to use SHA256 instead 
of MD5.  I don’t know of anyone working on these enhancements. 

-Bryan


> On Dec 23, 2015, at 5:33 AM, Craig Schomburg (craigs) <cra...@cisco.com> 
> wrote:
> 
> 
> I was looking through various Apache Traffic Server posts and noticed that 
> some FIPS related work was done with Apache Traffic Server (ATS).  Was 
> looking for someone with first hand knowledge of the ATS FIPS status that 
> might have some time for a few questions...
> 
> I am working with one of our product teams on FIPS enablement on a product 
> that is using Apache Traffic Server.  I just completed upgrading our product 
> to pull in ATS 6.0.0 code base and started working on enabling FIPS mode.
> 
> Had a few questions pertaining to FIPS support on ATS 6.0.0 as well as some 
> changes made via "TS-3576 Remove the need for FIPS locking for OpenSSL".
> 
> First question is basically how far has the support for FIPs mode progressed 
> with ATS?
> 
> Follow up question and observation...  I had to make local modifications to 
> the TS-3576 change that was mentioned in a thread regarding 
> SSL_CTX_add_extra_chain_cert_file() update of FIPS mode.  As was mentioned in 
> the separate e-mail thread the committed code really does nothing as the 
> FIPS_mode() call simply reads the current state and then the call to 
> FIPS_mode_set(mode) basically sets OpenSSL to the same state it is already in 
> (NO-OP).
> 
> +#ifdef OPENSSL_FIPS
> +    int mode = FIPS_mode();
> +    FIPS_mode_set(mode);
> +    Debug("ssl", "FIPS_mode: %d", mode);
> +#endif
> 
> I made a local modification in our repository to basically add a new config 
> option to records.config and then set the mode based on the config setting.
> 
> That got me by the first issue then I hit the next major issue which is that 
> the Apache Traffic Server code is pretty heavily entrenched in using MD5.  
> Since MD5 is not FIPS compliant the call to MD5_Init() in the Ink code then 
> causes a process to crash.  I am now looking into the possibility of 
> converting the existing MD5 references to SHA256 or making a model where it 
> could be switched between MD5/SHA256 based on the fips_mode setting.  Have 
> not  really started digging into this yet as I wanted to first probe the ATS 
> community to see if this work may have already been started, if there was any 
> position statement, plan, etc. on moving to a FIPS compliant hash, or if this 
> work was being avoided for other reasons.
> 
> Any input would be greatly appreciated.  Likewise if there is a better forum 
> for posting this question, please let me know.
> 
> Thanks,
> 
> Craig Schomburg
> 

Reply via email to