[SSSD] [PATCH] Clean up kdcinfo and kpasswdinfo files when exiting

Jakub Hrozek jhrozek at redhat.com
Wed May 5 13:29:06 UTC 2010


-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 05/05/2010 02:18 PM, Stephen Gallagher wrote:
> On 05/05/2010 03:40 AM, Jakub Hrozek wrote:
>> -----BEGIN PGP SIGNED MESSAGE-----
>> Hash: SHA1
>>
>> On 05/04/2010 11:44 PM, Jakub Hrozek wrote:
>>> On 05/04/2010 10:54 PM, Stephen Gallagher wrote:
>>>> https://fedorahosted.org/sssd/ticket/452
>>>
>>>
>>>> This patch adds a signal handler for SIGTERM that will remove the
>>>> kdcinfo.REALM and kpasswd.REALM files before exiting. I had to move the
>>>> definition of krb5_ctx into krb5_common.h so I could dereference it to
>>>> get the realm name.
>>>
>>>> I also made the sig_term() function from server.c public, so I could
>>>> chain into it once my exit handler is done (to make sure it cleans up
>>>> any krb5_child processes that might still be in-progress).
>>>
>>>
>>> Nack, if using kdc for kpasswd operations, ctx->kpasswd_service is NULL,
>>> so you would get a segfault.
>>>
>>> I think the realms for kpasswd and kdc will always be the same, so maybe
>>> you can pass the realm as the private_data and avoid exposing the krb
>>> context at all?
>>
>> Also please define sige in sssm_krb5_auth_init()
> 
> New patch attached just passes in the realm as the private data of the
> callback. Also, I rebased this patch atop my "Properly set up SIGCHLD
> handlers" patch, as they were in conflict.
> 
> In case you're wondering why I didn't use talloc_get_name() in the
> signal callback as is our style, it's because the talloc_strdup() set's
> the context's name equal to the string's value. So I had to just use a
> traditional cast here instead.
> 
> 

ACK
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.10 (GNU/Linux)
Comment: Using GnuPG with Fedora - http://enigmail.mozdev.org/

iEYEARECAAYFAkvhcqIACgkQHsardTLnvCWCOgCeMAhgXMveEyljyy7Dx1VjEQVV
5zsAn255Cz29jNSULv8DAkKB3LDweybc
=GmhF
-----END PGP SIGNATURE-----



More information about the sssd-devel mailing list