Re: ANNOUNCE: nfs-utils-1.2.2 released.

[Lukas Hejtmanek <xhejtman@ics.muni.cz>]

On Mon, Mar 08, 2010 at 02:48:25PM -0500, Jeff Layton wrote:
> What would probably help some is to do a bit more analysis with gdb.
> Maybe see if you can determine where it's falling down in libgssglue
> and why. That may help us to determine what the actual problem is.

using gdb and debug in libgssglue, it looks like there is a mess in
structures:
Core was generated by `/tmp/nfs-utils-1.2.2/utils/gssd/gssd -vvv'.
Program terminated with signal 11, Segmentation fault.
#0  0x00007f8a77c0a5ef in __gss_get_mechanism_cred (union_cred=3D0x1be3=
a00,
mech_type=3D0x6114f0) at g_glue.c:295
295             if (g_OID_equal(mech_type, &union_cred->mechs_array[i])=
)
(gdb) print *union_cred
$1 =3D {count =3D 29243904, mechs_array =3D 0x1, cred_array =3D 0x1be8b=
e0, auxinfo
=3D {name =3D {length =3D 29264896, value =3D 0x11},=20
    name_type =3D 0x1be8bc0, creation_time =3D 140232705185984, time_re=
c
=3D 1268127776, cred_usage =3D 86400}}

this seems to be because gss_acquire_cred is defined in both libgssglue=
=2Eso and
libgssapi_krb5.so

and these two libs do not agree on gss_union_cred_t:

libgssapi_krb5:
typedef struct gss_cred_id_struct {
        struct gss_cred_id_struct *loopback;
        int                     count;
        gss_OID                 mechs_array;
        gss_cred_id_t           *cred_array;
        gss_union_cred_auxinfo  auxinfo;
} gss_union_cred_desc, *gss_union_cred_t;

libgssglu:
typedef struct gss_union_cred_t {
        int                     count;
        gss_OID                 mechs_array;
        gss_cred_id_t *         cred_array;
        gss_union_cred_auxinfo  auxinfo;
} gss_union_cred_desc, *gss_union_cred_t;


--=20
Luk=E1=B9 Hejtm=E1nek