Received-SPF: pass (google.com: best guess record for domain of linux-nfs-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Subject: Re: gssd question/patch
To:     Olga Kornievskaia <aglo@umich.edu>
Cc:     linux-nfs <linux-nfs@vger.kernel.org>
References: <CAN-5tyHJg4C5j72_CrCJhZ8hyzDe71Q9zw1USgmyxePg+3juZw@mail.gmail.com>
From:   Steve Dickson <SteveD@RedHat.com>
Message-ID: <c8b304f3-1146-e1f8-6c5e-2b2c558868c6@RedHat.com>
Date:   Wed, 11 Dec 2019 11:41:24 -0500
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:68.0) Gecko/20100101
 Thunderbird/68.2.2
MIME-Version: 1.0
In-Reply-To: <CAN-5tyHJg4C5j72_CrCJhZ8hyzDe71Q9zw1USgmyxePg+3juZw@mail.gmail.com>
Content-Type: text/plain; charset=utf-8
Content-Language: en-US
Content-Transfer-Encoding: 7bit
Sender: linux-nfs-owner@vger.kernel.org
Precedence: bulk


On 12/6/19 1:29 PM, Olga Kornievskaia wrote:
> Hi Steve,
> 
> Question: Is this an interesting failure scenario (bug) that should be
> fixed: client did a mount which acquired gss creds and stored in the
> credential cache. Then say it umounts at some point. Then for some
> reason the Kerberos cache is deleted (rm -f /tmp/krb5cc*). Now client
> mounts again. This currently fails. Because gssd uses internal cache
> to store creds lifetimes and thinks that tgt is still valid but then
> trying to acquire a service ticket it fails (since there is no tgt).
OK... turning off gssproxy... I know see the failure... 
> 
> Here's my proposed fix (I can send as a patch if this agreed upon).
> 
> diff --git a/utils/gssd/krb5_util.c b/utils/gssd/krb5_util.c
> index 0474783..3678524 100644
> --- a/utils/gssd/krb5_util.c
> +++ b/utils/gssd/krb5_util.c
> @@ -121,6 +121,9 @@
>  #include <krb5.h>
>  #include <rpc/auth_gss.h>
> 
> +#include <sys/types.h>
> +#include <fcntl.h>
> +
>  #include "nfslib.h"
>  #include "gssd.h"
>  #include "err_util.h"
> @@ -314,6 +317,25 @@ gssd_find_existing_krb5_ccache(uid_t uid, char *dirname,
>         return err;
>  }
> 
> +/* check if the ticket cache exists, if not set nocache=1 so that new
> + * tgt is gotten
> + */
> +static int
> +gssd_check_if_cc_exists(struct gssd_k5_kt_princ *ple)
> +{
> +       int fd;
> +       char cc_name[BUFSIZ];
> +
> +       snprintf(cc_name, sizeof(cc_name), "%s/%s%s_%s",
> +               ccachesearch[0], GSSD_DEFAULT_CRED_PREFIX,
> +               GSSD_DEFAULT_MACHINE_CRED_SUFFIX, ple->realm);
> +       fd = open(cc_name, O_RDONLY);
> +       if (fd < 0)
> +               return 1;
> +       close(fd);
> +       return 0;
> +}
> +
>  /*
>   * Obtain credentials via a key in the keytab given
>   * a keytab handle and a gssd_k5_kt_princ structure.
> @@ -348,6 +370,8 @@ gssd_get_single_krb5_cred(krb5_context context,
> 
>         memset(&my_creds, 0, sizeof(my_creds));
> 
> +       if (!nocache && !use_memcache)
> +               nocache = gssd_check_if_cc_exists(ple);
>         /*
>          * Workaround for clock skew among NFS server, NFS client and KDC
>          * 300 because clock skew must be within 300sec for kerberos
> 
I guess this seem reasonable... but it does apply on the latest upstream
version... So please do the official patch and we'll go from there..

thanks!

steved.