Re: [PATCH] wifi: brcmfmac: leave the AP 4-way handshake to the host on Apple chips

From: Arend van Spriel

Date: Tue Nov 17 2026 - 14:44:31 EST


On 17/11/2026 18:49, Aditya Garg wrote:


On 17 Nov 2026, at 2:07 AM, Arend van Spriel <arend.vanspriel@xxxxxxxxxxxx> wrote:

On Wed, 4 Nov 2026 16:16:29 +0100, Andre Eikmeyer <dev@deq.rocks> wrote:
The Apple firmware for the BCM4364 and BCM4377 lists "idauth" in its
capabilities, so brcmfmac enables BRCMF_FEAT_FWAUTH and advertises
NL80211_EXT_FEATURE_4WAY_HANDSHAKE_AP_PSK. hostapd and wpa_supplicant
then pass the PSK to the driver and leave the AP side of the 4-way
handshake to the firmware. On the BCM4364 that handshake never
completes.

[...]

Clear BRCMF_FEAT_FWAUTH on the chips in Apple's T2 Macs, so the host
runs the handshake.

[...]

---
.../wireless/broadcom/brcm80211/brcmfmac/feature.c | 14 ++++++++++++++
1 file changed, 14 insertions(+)

@@ -318,6 +318,20 @@ void brcmf_feat_attach(struct brcmf_pub *drvr)
BIT(BRCMF_FEAT_WOWL_GTK);
}
}
+ /* The Apple firmware for these chips lists "idauth", but in AP mode
+ * its authenticator never completes the 4-way handshake and the
+ * station is dropped after about 8 s. Leave the handshake to the host.
+ */
+ switch (drvr->bus_if->chip) {
+ case BRCM_CC_4355_CHIP_ID:
+ case BRCM_CC_4364_CHIP_ID:
+ case BRCM_CC_4377_CHIP_ID:
+ ifp->drvr->feat_flags &= ~BIT(BRCMF_FEAT_FWAUTH);
+ break;
+ default:
+ break;
+ }
+
/* MBSS does not work for all chips */
switch (drvr->bus_if->chip) {

The commit message and the test report only cover BCM4364 and BCM4377,
but the code also clears FWAUTH for BCM4355. Where does that come from?
Was it tested, and is BCM4355 found in T2 Macs? If not, please drop it
or justify it.

4355 is also a Mac specific chip found in MacBookAir8,1 and 8,2. I'm actually surprised you forgot about that especially when there was a whole issue when Cypress named their chip firmware wrongly as 4355.

Getting older and a bit senile I guess :-p Still it pops up here in the patch without mentioning it in the commit message so my question still stands. Also are there any logs made while debugging this issue? Having firmware console logging enabled might be useful to get a clue what is holding the firmware from completing the handshake.

Regards,
Arend