[PATCH v2 03/23] ath6kl: add cfg80211.c

[Kalle Valo]

On 07/17/2011 02:12 PM, Johannes Berg wrote:
> On Sun, 2011-07-17 at 13:21 +0300, Kalle Valo wrote:
> 
>> +static int ath6kl_set_auth_type(struct ath6kl *ar,
>> +				enum nl80211_auth_type auth_type)
>> +{
> 
>> +	default:
>> +		ar->dot11_auth_mode = OPEN_AUTH;
> 
> are you sure you want that?

I don't know :)

That case shouldn't happen and we return an error code. But then again
callers don't check the return code and continue the normal execution
flow even when an error happens. I'm starting to think that fixing
callers to check for errors would be the best solution. What do you think?

>> +	if (!sme->ssid_len || sme->ssid_len > IEEE80211_MAX_SSID_LEN) {
>> +		ath6kl_err("%s: ssid invalid\n", __func__);
>> +		return -EINVAL;
>> +	}
> 
> I don't mind the code here -- but it's not going to happen anyway :)

I'll remove the check.

> 
>> +	if (down_interruptible(&ar->sem)) {
>> +		ath6kl_err("%s: busy, couldn't get access\n", __func__);
>> +		return -ERESTARTSYS;
>> +	}
> 
> Hmm, will that really work? I don't think netlink has any -ERESTARTSYS
> handling so this would go to userspace which isn't supposed to happen,
> better make it -EBUSY. Besides, no point in restarting something that
> continually fails right?

Agreed. I don't like this either and I'm planning to remove all that
once we switch to a mutex.

>> +	if (down_interruptible(&ar->sem)) {
>> +		ath6kl_err("%s: busy, couldn't get access\n", __func__);
>> +		return -ERESTARTSYS;
>> +	}
> 
> same here (disconnect) and maybe other places

Yes, I think all ar->sem uses do it similarly as here.

>> +	/*
>> +	 * TODO: Update target to include 802.11 mac header while sending
>> +	 * bss info. Target removes 802.11 mac header while sending the bss
>> +	 * info to host, cfg80211 needs it, for time being just filling the
>> +	 * da, sa and bssid fields alone.
>> +	 */
>> +	mgmt = (struct ieee80211_mgmt *)ieeemgmtbuf;
>> +	memset(mgmt->da, 0xff, ETH_ALEN);	/*broadcast addr */
>> +	memcpy(mgmt->sa, ni->ni_macaddr, ETH_ALEN);
>> +	memcpy(mgmt->bssid, ni->ni_macaddr, ETH_ALEN);
>> +	memcpy(ieeemgmtbuf + offsetof(struct ieee80211_mgmt, u),
>> +	       ni->ni_buf, ni->ni_framelen);
>> +
>> +	freq = cie->ie_chan;
>> +	channel = ieee80211_get_channel(wiphy, freq);
>> +	signal = ni->ni_snr * 100;
>> +
>> +	ath6kl_dbg(ATH6KL_DBG_WLAN_CFG,
>> +		   "%s: bssid %pM ch %d freq %d size %d\n", __func__,
>> +		   mgmt->bssid, channel->hw_value, freq, size);
>> +	cfg80211_inform_bss_frame(wiphy, channel, mgmt,
>> +				  size, signal, GFP_KERNEL);
> 
> What's with the comment and all the frame creation code; what's wrong
> with cfg80211_inform_bss() instead of _frame()?

I have no idea, it was there when I came here ;) I have been wondering
the same. I added a task to the todo list to change this.

http://wireless.kernel.org/en/users/Drivers/ath6kl/todo

>> +static int ath6kl_cfg80211_add_key(struct wiphy *wiphy, struct net_device *ndev,
>> +				   u8 key_index, bool pairwise,
>> +				   const u8 *mac_addr,
>> +				   struct key_params *params)
>> +{
> 
> ...
> 
>> +	if (!mac_addr || is_broadcast_ether_addr(mac_addr))
>> +		key_usage = GROUP_USAGE;
>> +	else
>> +		key_usage = PAIRWISE_USAGE;
> 
> Isn't that covered by the pairwise argument?

At least to me it looks like it is. I'll change it.

>> +static int ath6kl_cfg80211_set_default_mgmt_key(struct wiphy *wiphy,
>> +						struct net_device *ndev,
>> +						u8 key_index)
>> +{
>> +	struct ath6kl *ar = (struct ath6kl *)ath6kl_priv(ndev);
>> +
>> +	ath6kl_dbg(ATH6KL_DBG_WLAN_CFG, "%s: index %d\n", __func__, key_index);
>> +
>> +	if (!test_bit(WMI_READY, &ar->flag)) {
>> +		ath6kl_err("%s: wmi is not ready\n", __func__);
>> +		return -EIO;
>> +	}
>> +
>> +	if (ar->wlan_state == WLAN_DISABLED) {
>> +		ath6kl_err("%s: wlan disabled\n", __func__);
>> +		return -EIO;
>> +	}
>> +
>> +	ath6kl_dbg(ATH6KL_DBG_WLAN_CFG, "%s: not supported\n", __func__);
>> +	return -ENOTSUPP;
>> +}
> 
> This is completely pointless -- remove the entire function.

Will do.

>> +/*
>> + * The type nl80211_tx_power_setting replaces the following
>> + * data type from 2.6.36 onwards
>> +*/
> 
> ??

Don't ask :) Most likely some backwards compatibility code we removed
but forgot to remove the comment.

>> +static int ath6kl_cfg80211_change_iface(struct wiphy *wiphy,
>> +					struct net_device *ndev,
>> +					enum nl80211_iftype type, u32 *flags,
>> +					struct vif_params *params)
>> +{
>> +	struct ath6kl *ar = ath6kl_priv(ndev);
>> +	struct wireless_dev *wdev = ar->wdev;
>> +
>> +	ath6kl_dbg(ATH6KL_DBG_WLAN_CFG, "%s: type %u\n", __func__, type);
>> +
>> +	if (!test_bit(WMI_READY, &ar->flag)) {
>> +		ath6kl_err("%s: wmi is not ready\n", __func__);
>> +		return -EIO;
>> +	}
>> +
>> +	if (ar->wlan_state == WLAN_DISABLED) {
>> +		ath6kl_err("%s: wlan disabled\n", __func__);
>> +		return -EIO;
>> +	}
>> +
>> +	switch (type) {
>> +	case NL80211_IFTYPE_STATION:
>> +		ar->next_mode = INFRA_NETWORK;
>> +		break;
>> +	case NL80211_IFTYPE_ADHOC:
>> +		ar->next_mode = ADHOC_NETWORK;
>> +		break;
>> +	default:
>> +		ath6kl_err("%s: invalid type %u\n", __func__, type);
>> +		return -EOPNOTSUPP;
>> +	}
>> +
>> +	wdev->iftype = type;
> 
> We kinda expect it to have taken effect by the time you return from this
> function -- it doesn't quite look like it will?

No, it doesn't. This needs some thinking. Most probably I'll add a
comment to the code and revisit this later.

>> +	if (!ibss_param->ssid_len
>> +	    || IEEE80211_MAX_SSID_LEN < ibss_param->ssid_len) {
>> +		ath6kl_err("%s: ssid invalid\n", __func__);
>> +		return -EINVAL;
>> +	}
> 
> Like above -- won't happen. You can trust cfg80211 ;-)

I will always trust cfg80211 :D

So the test will be gone in v3.

>> +	ar->ssid_len = ibss_param->ssid_len;
>> +	memcpy(ar->ssid, ibss_param->ssid, ar->ssid_len);
>> +
>> +	if (ibss_param->channel)
>> +		ar->ch_hint = ibss_param->channel->center_freq;
>> +
>> +	if (ibss_param->channel_fixed) {
>> +		/*
>> +		 * TODO: channel_fixed: The channel should be fixed, do not
>> +		 * search for IBSSs to join on other channels. Target
>> +		 * firmware does not support this feature, needs to be
>> +		 * updated.
>> +		 */
>> +	}
> 
> return an error?

Oh yeah. I'll add that.

Johannes, thanks a lot for a very thorough review of our cfg80211 handling.

Kalle