The ath9k driver was so far only able to rekey PTK0 keys correctly due
to the best effort queue flush added with commit 62872a9b9a10
("mac80211: Fix PTK rekey freezes and clear text leak").
Add the needed queue flush and set NL80211_EXT_FEATURE_CAN_REPLACE_PTK0
to tell mac80211 that the driver can now rekey PTK0 keys correctly and
no longer needs the best effort flush.
Effectively this prevents mac80211 to warn when rekeying a PTK0 key
only.
Signed-off-by: Alexander Wetzel <[email protected]>
---
To keep the patch readable I did not move functions around.
By moving functions we can avoid the extra declaration of ath9k_flush().
Ath9k was totally unable to rekey PTK0 under even light load till commit
62872a9b9a10 ("mac80211: Fix PTK rekey freezes and clear text leak")
was merged in linux 4.20.
This is just moving the needed queue flush into the driver and sets
NL80211_EXT_FEATURE_CAN_REPLACE_PTK0 to suppress the for ath9k known to
be bogus warning the kernel throws currently on rekey.
From usability and security perspective at least commit 62872a9b9a10
("mac80211: Fix PTK rekey freezes and clear text leak")
should be back ported to all stable versions.
Chances are the same patch works for the all ath cards < ath10k.
I can prepare corresponding patches but can't test those.
drivers/net/wireless/ath/ath9k/init.c | 1 +
drivers/net/wireless/ath/ath9k/main.c | 12 ++++++++++++
2 files changed, 13 insertions(+)
diff --git a/drivers/net/wireless/ath/ath9k/init.c b/drivers/net/wireless/ath/ath9k/init.c
index 289a2444d534..ea1d17a265a7 100644
--- a/drivers/net/wireless/ath/ath9k/init.c
+++ b/drivers/net/wireless/ath/ath9k/init.c
@@ -1014,6 +1014,7 @@ static void ath9k_set_hw_capab(struct ath_softc *sc, struct ieee80211_hw *hw)
wiphy_ext_feature_set(hw->wiphy, NL80211_EXT_FEATURE_AIRTIME_FAIRNESS);
wiphy_ext_feature_set(hw->wiphy,
NL80211_EXT_FEATURE_MULTICAST_REGISTRATIONS);
+ wiphy_ext_feature_set(hw->wiphy, NL80211_EXT_FEATURE_CAN_REPLACE_PTK0);
}
int ath9k_init_device(u16 devid, struct ath_softc *sc,
diff --git a/drivers/net/wireless/ath/ath9k/main.c b/drivers/net/wireless/ath/ath9k/main.c
index a47f6e978095..7d96414741da 100644
--- a/drivers/net/wireless/ath/ath9k/main.c
+++ b/drivers/net/wireless/ath/ath9k/main.c
@@ -19,6 +19,9 @@
#include "ath9k.h"
#include "btcoex.h"
+static void ath9k_flush(struct ieee80211_hw *hw, struct ieee80211_vif *vif,
+ u32 queues, bool drop);
+
u8 ath9k_parse_mpdudensity(u8 mpdudensity)
{
/*
@@ -1701,6 +1704,15 @@ static int ath9k_set_key(struct ieee80211_hw *hw,
return -EOPNOTSUPP;
}
+ /* There may be MPDUs queued for the outgoing PTK key. Flush queues to
+ * make sure these are not send unencrypted or with a wrong (new) key
+ */
+ if (cmd == DISABLE_KEY && key->flags & IEEE80211_KEY_FLAG_PAIRWISE) {
+ ieee80211_stop_queues(hw);
+ ath9k_flush(hw, vif, 0, true);
+ ieee80211_wake_queues(hw);
+ }
+
mutex_lock(&sc->mutex);
ath9k_ps_wakeup(sc);
ath_dbg(common, CONFIG, "Set HW Key %d\n", cmd);
--
2.28.0
Alexander Wetzel <[email protected]> wrote:
> The ath9k driver was so far only able to rekey PTK0 keys correctly due
> to the best effort queue flush added with commit 62872a9b9a10
> ("mac80211: Fix PTK rekey freezes and clear text leak").
>
> Add the needed queue flush and set NL80211_EXT_FEATURE_CAN_REPLACE_PTK0
> to tell mac80211 that the driver can now rekey PTK0 keys correctly and
> no longer needs the best effort flush.
>
> Effectively this prevents mac80211 to warn when rekeying a PTK0 key
> only.
>
> Signed-off-by: Alexander Wetzel <[email protected]>
> Signed-off-by: Kalle Valo <[email protected]>
Patch applied to ath-next branch of ath.git, thanks.
4bad3a2041b5 ath9k: add NL80211_EXT_FEATURE_CAN_REPLACE_PTK0 support
--
https://patchwork.kernel.org/patch/11700677/
https://wireless.wiki.kernel.org/en/developers/documentation/submittingpatches