Am Mittwoch, dem 31.07.2024 um 11:51 +0300 schrieb Kalle Valo: > Bert Karwatzki <spasswolf@xxxxxx> writes: > > > Am Mittwoch, dem 17.07.2024 um 17:25 +0200 schrieb Felix Fietkau: > > > > > On 17.07.24 16:38, Bert Karwatzki wrote: > > > > > > > So mvif->phy can be NULL at the start of mt7921_ipv6_addr_change. The early > > > > return in that case avoids the NULL pointer and mvif->phy has its usual value > > > > again on the next call to mt7921_ipv6_addr_change so Wifi is working again. I > > > > don't know how this could happen but perhaps you have an idea. > > > > > > This change should fix it: https://nbd.name/p/0747f54f > > > Please test. > > > > The BUG is still present in linux-6.11-rc1. > > I'm not sure what's the status with this. There's one mt76 patch going > to v6.11-rc2: > > https://git.kernel.org/pub/scm/linux/kernel/git/wireless/wireless.git/commit/?id=6557a28f3e3a54cff4f0dcdd1dfa649b26557ab3 > > But that looks to be a fix for a different problem, right? Felix, are > you planning to submit that 0747f54f as a proper patch? I could then > take it to wireless tree. > The Bug is still present in linux-6.11-rc2 and linux-next-20240806. Also the mvif->phy NULL check in the original patch is not neccessary (and feels a little out of place as mvif->phy is not needed anymore). This patch is sufficient to fix the NULL pointer dereference: diff --git a/drivers/net/wireless/mediatek/mt76/mt7921/main.c b/drivers/net/wireless/mediatek/mt76/mt7921/main.c index 1bab93d049df..23b228804289 100644 --- a/drivers/net/wireless/mediatek/mt76/mt7921/main.c +++ b/drivers/net/wireless/mediatek/mt76/mt7921/main.c @@ -1183,7 +1183,7 @@ static void mt7921_ipv6_addr_change(struct ieee80211_hw *hw, struct inet6_dev *idev) { struct mt792x_vif *mvif = (struct mt792x_vif *)vif->drv_priv; - struct mt792x_dev *dev = mvif->phy->dev; + struct mt792x_dev *dev = mt792x_hw_dev(hw); struct inet6_ifaddr *ifa; struct in6_addr ns_addrs[IEEE80211_BSS_ARP_ADDR_LIST_LEN]; struct sk_buff *skb; Bert Karwatzki
