MIT Kerberos using invalid in-memory credential cache

[Vipul Mehta]

I tried the fix but it didn't work. I observed there are some other issues
w.r.t. mechanism type from curl and would like to understand behavior
before further analysis.

1) How is default credential picked up when GSS_C_NO_CREDENTIAL is
specified in gss_init_sec_context() ? How is in-memory credential cache
used here ?
2) If the GSS credential passed to  gss_init_sec_context() does not match
the specified mechanism type then  GSS_C_NO_CREDENTIAL is the value
returned from gssint_get_mechanism_cred().  Why it does not simply fail
here as client didn't pass matching oid and cred ?



On Wed, Dec 30, 2020 at 10:43 PM Greg Hudson <ghudson at mit.edu> wrote:

> On 12/29/20 7:21 AM, Vipul Mehta wrote:
> > I see that some fix has been done in newer version :
> >
> https://github.com/krb5/krb5/commit/146dadec8fe7ccc4149eb2e3f577cc320aee6efb#diff-8f14845d698c6c1242bf1288e7bfec3db113dd57279601af016ec0df4a20949e
> >
> > Will it help ? How to debug this issue further in our service ?
>
> It might.  One of the bugs fixed in that commit is that two simultaneous
> references to the same memory cache would cause one of the references to
> become a dangling pointer when the other is destroyed.  I'm not sure how
> upgrading curl would lead to that scenario, though.
>
> If you choose to backport this commit, note that it contained a bug,
> described here:
>
>   https://krbdev.mit.edu/rt/Ticket/Display.html?id=8771
>


-- 
Regards,
Vipul