Re: [PATCH] mwl8k: Marvell TOPDOG wireless driver

[Johannes Berg <johannes@sipsolutions.net>]

[-- Attachment #1: Type: text/plain, Size: 1805 bytes --]

On Tue, 2009-03-10 at 10:13 +0100, Lennert Buytenhek wrote:

> +		/*
> +		 * Check for pending join operation. save a copy of
> +		 * the beacon and schedule a tasklet to send finalize
> +		 * join command to the firmware.
> +		 */
> +		if (mwl8k_capture_bssid(priv, wh))
> +			mwl8k_save_beacon(priv, skb);

This raises an interesting problem I'm debugging in iwlwifi right now --
when you suspend/resume mac80211 assumes, currently, that all it has to
do is completely reconfigure the hardware (cf. net/mac80211/pm.c,
__ieee80211_resume()), it is currently missing this piece of code
though:

        list_for_each_entry(sdata, &local->interfaces, list) {
                if (!netif_running(sdata->dev))
                        continue;
                if (sdata->vif.type == NL80211_IFTYPE_MONITOR)
                        continue;
                ieee80211_if_config(sdata, ~0);
                ieee80211_bss_info_change_notify(sdata, ~0);
        }

(Think about it for a second before I explain)




This means that after resume, the finalize join command will not be run,
although mac80211 assumes that it is connected to the AP, and the
firmware was reset completely.

We may be disconnected if we sleep for too long, but that is acceptable,
as soon as we send something the AP will then throw us out and we go
through the association procedure again. We may want to add some code to
detect that (maybe just send a nullfunc "going to sleep" before suspend,
and "awake now" after resume?) but it should work anyway.

I don't have a good solution right now. We could always go through the
assoc process again, but that seems wasteful. Or we could just have
mac80211 hand down the beacon/probe response data in the assoc setting
(bss config)

johannes

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 836 bytes --]