In article <[email protected]>,
Patrick Welche  <[email protected]> wrote:
>I'm just reading if_iwn.c, and I don't see how iwn_iter_func can be legal:
>
>static void     
>iwn_iter_func(void *arg, struct ieee80211_node *ni)
>{
>        struct iwn_softc *sc = arg;
>        struct iwn_node *wn = (struct iwn_node *)ni;
>
>        ieee80211_amrr_choose(&sc->amrr, ni, &wn->amn);
>}
>
>iwn_node is bigger than ieee80211_node, as it starts with an ieee80211_node:
>
>struct iwn_node {       
>        struct  ieee80211_node          ni;     /* must be the first */
>        struct  ieee80211_amrr_node     amn;
>        uint16_t                        disable_tid;
>        uint8_t                         id;
>        uint8_t                         ridx[IEEE80211_RATE_MAXSIZE];
>};
>
>Then, the call to ieee80211_amrr_choose uses amn, which is after the
>struct the iwn_node was initalised with. ieee80211_amrr_choose then starts
>by dereferencing bits of the amn. Isn't that pointing at garbage?
>
>I haven't used iwn as an "infrastructure station", so have never run into
>trouble, but is the C analysis right?

It allocates enough space because the 80211 code uses iwn_node_alloc to
allocate ieee80211_nodes, so that is fine.

christos

Reply via email to