On HCI_OP_RESET command, firmware raises alive interrupt. Driver needs
to wait for this before sending other command. This patch fixes the potential
miss of alive interrupt due to which HCI_OP_RESET can timeout.
Expected flow:
If tx command is HCI_OP_RESET,
1. set data->gp0_received = false
2. send HCI_OP_RESET
3. wait for alive interrupt
Actual flow having potential race:
If tx command is HCI_OP_RESET,
1. send HCI_OP_RESET
1a. Firmware raises alive interrupt here and in ISR
data->gp0_received is set to true
2. set data->gp0_received = false
3. wait for alive interrupt
Signed-off-by: Kiran K <kiran.k@xxxxxxxxx>
Fixes: 05c200c8f029 ("Bluetooth: btintel_pcie: Add handshake between driver and firmware")
Reported-by: Bjorn Helgaas <helgaas@xxxxxxxxxx>
Closes: https://patchwork.kernel.org/project/bluetooth/patch/20241001104451.626964-1-kiran.k@xxxxxxxxx/
---
drivers/bluetooth/btintel_pcie.c | 5 ++++-
1 file changed, 4 insertions(+), 1 deletion(-)
diff --git a/drivers/bluetooth/btintel_pcie.c b/drivers/bluetooth/btintel_pcie.c
index 63eca52c0e0b..b8b241a92bf9 100644
--- a/drivers/bluetooth/btintel_pcie.c
+++ b/drivers/bluetooth/btintel_pcie.c
@@ -1321,6 +1321,10 @@ static int btintel_pcie_send_frame(struct hci_dev *hdev,
if (opcode == 0xfc01)
btintel_pcie_inject_cmd_complete(hdev, opcode);
}
+ /* Firmware raises alive interrupt on HCI_OP_RESET */
+ if (opcode == HCI_OP_RESET)
+ data->gp0_received = false;
+
hdev->stat.cmd_tx++;
break;
case HCI_ACLDATA_PKT:
@@ -1358,7 +1362,6 @@ static int btintel_pcie_send_frame(struct hci_dev *hdev,
opcode, btintel_pcie_alivectxt_state2str(old_ctxt),
btintel_pcie_alivectxt_state2str(data->alive_intr_ctxt));
if (opcode == HCI_OP_RESET) {
- data->gp0_received = false;
ret = wait_event_timeout(data->gp0_wait_q,
data->gp0_received,
msecs_to_jiffies(BTINTEL_DEFAULT_INTR_TIMEOUT_MS));
--
2.43.0
