Commit 31453804 authored by Martin Fuzzey's avatar Martin Fuzzey Committed by Kalle Valo

rsi: fix AP mode with WPA failure due to encrypted EAPOL

In AP mode WPA2-PSK connections were not established.

The reason was that the AP was sending the first message
of the 4 way handshake encrypted, even though no pairwise
key had (correctly) yet been set.

Encryption was enabled if the "security_enable" driver flag
was set and encryption was not explicitly disabled by
IEEE80211_TX_INTFL_DONT_ENCRYPT.

However security_enable was set when *any* key, including
the AP GTK key, had been set which was causing unwanted
encryption even if no key was avaialble for the unicast
packet to be sent.

Fix this by adding a check that we have a key and drop
the old security_enable driver flag which is insufficient
and redundant.

The Redpine downstream out of tree driver does it this way too.

Regarding the Fixes tag the actual code being modified was
introduced earlier, with the original driver submission, in
dad0d04f ("rsi: Add RS9113 wireless driver"), however
at that time AP mode was not yet supported so there was
no bug at that point.

So I have tagged the introduction of AP support instead
which was part of the patch set "rsi: support for AP mode" [1]

It is not clear whether AP WPA has ever worked, I can see nothing
on the kernel side that broke it afterwards yet the AP support
patch series says "Tests are performed to confirm aggregation,
connections in WEP and WPA/WPA2 security."

One possibility is that the initial tests were done with a modified
userspace (hostapd).

[1] https://www.spinics.net/lists/linux-wireless/msg165302.htmlSigned-off-by: default avatarMartin Fuzzey <martin.fuzzey@flowbird.group>
Fixes: 38ef6235 ("rsi: security enhancements for AP mode")
CC: stable@vger.kernel.org
Signed-off-by: default avatarKalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/1622564459-24430-1-git-send-email-martin.fuzzey@flowbird.group
parent 8667ab49
......@@ -203,7 +203,7 @@ int rsi_prepare_data_desc(struct rsi_common *common, struct sk_buff *skb)
wh->frame_control |= cpu_to_le16(RSI_SET_PS_ENABLE);
if ((!(info->flags & IEEE80211_TX_INTFL_DONT_ENCRYPT)) &&
(common->secinfo.security_enable)) {
info->control.hw_key) {
if (rsi_is_cipher_wep(common))
ieee80211_size += 4;
else
......
......@@ -1045,7 +1045,6 @@ static int rsi_mac80211_set_key(struct ieee80211_hw *hw,
mutex_lock(&common->mutex);
switch (cmd) {
case SET_KEY:
secinfo->security_enable = true;
status = rsi_hal_key_config(hw, vif, key, sta);
if (status) {
mutex_unlock(&common->mutex);
......@@ -1064,8 +1063,6 @@ static int rsi_mac80211_set_key(struct ieee80211_hw *hw,
break;
case DISABLE_KEY:
if (vif->type == NL80211_IFTYPE_STATION)
secinfo->security_enable = false;
rsi_dbg(ERR_ZONE, "%s: RSI del key\n", __func__);
memset(key, 0, sizeof(struct ieee80211_key_conf));
status = rsi_hal_key_config(hw, vif, key, sta);
......
......@@ -1803,8 +1803,7 @@ int rsi_send_wowlan_request(struct rsi_common *common, u16 flags,
RSI_WIFI_MGMT_Q);
cmd_frame->desc.desc_dword0.frame_type = WOWLAN_CONFIG_PARAMS;
cmd_frame->host_sleep_status = sleep_status;
if (common->secinfo.security_enable &&
common->secinfo.gtk_cipher)
if (common->secinfo.gtk_cipher)
flags |= RSI_WOW_GTK_REKEY;
if (sleep_status)
cmd_frame->wow_flags = flags;
......
......@@ -151,7 +151,6 @@ enum edca_queue {
};
struct security_info {
bool security_enable;
u32 ptk_cipher;
u32 gtk_cipher;
};
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment