On Sun, 2008-12-14 at 22:50 -0500, Bob Copeland wrote:
> This patch introduces suspend and resume callbacks to mac80211,
> allowing mac80211 to quiesce its state (bringing down interfaces,
> removing keys, etc) in preparation for suspend.  cfg80211 will call
> the suspend hook before the device suspend, and resume hook after
> the device resume.

Looks good to me.

> Signed-off-by: Bob Copeland <me@bobcopeland.com>

Acked-by: Johannes Berg <johannes@sipsolutions.net>

> +int __ieee80211_suspend(struct ieee80211_hw *hw)
> +{
> +	struct ieee80211_local *local = hw_to_local(hw);
> +	struct ieee80211_sub_if_data *sdata;
> +	struct ieee80211_if_init_conf conf;
> +	struct sta_info *sta;
> +
> +	flush_workqueue(local->hw.workqueue);
> +
> +	/* disable keys */
> +	list_for_each_entry(sdata, &local->interfaces, list)
> +		ieee80211_disable_keys(sdata);
> +
> +	/* remove STAs */
> +	list_for_each_entry(sta, &local->sta_list, list) {
> +
> +		if (local->ops->sta_notify) {

One could reorder this to:
	if (local->ops->sta_notify) {
		list_for_each_entry(sta, &local->sta_list, list) {

and the remaining code stays, but it really doesn't matter in this code
path.

Does it work now? Did you ever figure out why it failed occasionally
before?

johannes