[PATCH v2] wifi: brcmfmac: handle missing D3 ACK on BCM4377 T2 Macs
From: Andre Eikmeyer
Date: Thu Oct 01 2026 - 04:59:19 EST
BCM4377 on the affected T2 Macs does not acknowledge HOST_D3_INFORM.
The current PCIe suspend callback treats this as fatal and aborts system
suspend.
When this happens on BCM4377, mask the mailbox interrupt, preserve the
bus-down state, and let generic PCI/ACPI suspend proceed. On resume the
existing recovery path tears down and probes the device again instead of
relying on a D0 mailbox transaction.
Tested-on: MacBookAir9,1, MacBookPro15,4, MacBookPro16,3
Signed-off-by: Andre Eikmeyer <dev@deq.rocks>
---
Changes in v2:
- Drop the redundant DMI allowlist; BCM4377 is Apple-exclusive and only
appears in the three affected MacBook models.
- Keep the device check inside CONFIG_PM, fixing the CONFIG_PM=n unused
function warning reported by the kernel test robot.
drivers/net/wireless/broadcom/brcm80211/brcmfmac/pcie.c | 7 +++++++
1 file changed, 7 insertions(+)
diff --git a/drivers/net/wireless/broadcom/brcm80211/brcmfmac/pcie.c b/drivers/net/wireless/broadcom/brcm80211/brcmfmac/pcie.c
index 55f4d7b97..e2e050395 100644
--- a/drivers/net/wireless/broadcom/brcm80211/brcmfmac/pcie.c
+++ b/drivers/net/wireless/broadcom/brcm80211/brcmfmac/pcie.c
@@ -2652,11 +2652,18 @@ static int brcmf_pcie_pm_enter_D3(struct device *dev)
wait_event_timeout(devinfo->mbdata_resp_wait, devinfo->mbdata_completed,
BRCMF_PCIE_MBDATA_TIMEOUT);
if (!devinfo->mbdata_completed) {
+ if (devinfo->pdev->device == BRCM_PCIE_4377_DEVICE_ID) {
+ dev_info(dev, "D3 ACK missing; masking PCIe mailbox interrupt\n");
+ brcmf_pcie_intr_disable(devinfo);
+ goto done;
+ }
+
brcmf_err(bus, "Timeout on response for entering D3 substate\n");
brcmf_bus_change_state(bus, BRCMF_BUS_UP);
return -EIO;
}
+done:
devinfo->state = BRCMFMAC_PCIE_STATE_DOWN;
return 0;
base-commit: 1631d79ae57dce2c5f88ad278307028638a8b4d9
--
2.55.0