alx: fix dead loop for hardware fatal error

when reset in alx_task/alx_reinit, alx_halt function calls
cancel_work_sync(task), which makes a dead loop. This patch fixes it by
adding parameter of in_task for alx_halt & alx_reinit.

Signed-off-by: xiong <xiong@qca.qualcomm.com>

3 files changed, 12 insertions, 11 deletions

diff --git a/src/alx.h b/src/alx.h
index 78bbe5d..bbf147f 100644
--- a/src/alx.h
+++ b/src/alx.h
@@ -205,7 +205,7 @@ extern int alx_alloc_rxring_buf(struct alx_adapter *adpt,
 				struct alx_rx_queue *rxq);
 extern void alx_init_intr(struct alx_adapter *adpt);
 extern void alx_disable_advanced_intr(struct alx_adapter *adpt);
-extern void alx_reinit(struct alx_adapter *adpt);
+extern void alx_reinit(struct alx_adapter *adpt, bool in_task);
 extern void alx_set_ethtool_ops(struct net_device *dev);
 extern char alx_drv_name[];
 extern char alx_drv_version[];
diff --git a/src/alx_ethtool.c b/src/alx_ethtool.c
index 998fe87..074c640 100644
--- a/src/alx_ethtool.c
+++ b/src/alx_ethtool.c
@@ -328,7 +328,7 @@ static int alx_nway_reset(struct net_device *netdev)
 	struct alx_adapter *adpt = netdev_priv(netdev);
 
 	if (netif_running(netdev))
-		alx_reinit(adpt);
+		alx_reinit(adpt, false);
 
 	return 0;
 }
diff --git a/src/alx_main.c b/src/alx_main.c
index 7686aec..02abef4 100644
--- a/src/alx_main.c
+++ b/src/alx_main.c
@@ -35,7 +35,7 @@
 
 #define DRV_MAJ		1
 #define DRV_MIN		2
-#define DRV_PATCH	2
+#define DRV_PATCH	3
 #define DRV_MODULE_VER	\
 	__stringify(DRV_MAJ) "." __stringify(DRV_MIN) "." \
 	__stringify(DRV_PATCH)
@@ -1197,7 +1197,7 @@ static int alx_change_mtu(struct net_device *netdev, int new_mtu)
 				   ALIGN(max_frame, 8) : ALX_DEF_RXBUF_SIZE;
 		netdev_update_features(netdev);
 		if (netif_running(netdev))
-			alx_reinit(adpt);
+			alx_reinit(adpt, false);
 	}
 
 	return 0;
@@ -1360,12 +1360,13 @@ err_out:
 	return err;
 }
 
-static void alx_halt(struct alx_adapter *adpt)
+static void alx_halt(struct alx_adapter *adpt, bool in_task)
 {
 	struct alx_hw *hw = &adpt->hw;
 
 	ALX_FLAG_SET(adpt, HALT);
-	alx_cancel_work(adpt);
+	if (!in_task)
+		alx_cancel_work(adpt);
 
 	alx_netif_stop(adpt);
 	hw->link_up = false;
@@ -1397,7 +1398,7 @@ static void alx_activate(struct alx_adapter *adpt)
 
 static void __alx_stop(struct alx_adapter *adpt)
 {
-	alx_halt(adpt);
+	alx_halt(adpt, false);
 
 	alx_free_irq(adpt);
 
@@ -1802,7 +1803,7 @@ static void alx_update_stats(struct alx_adapter *adpt)
 	spin_unlock(&adpt->smb_lock);
 }
 
-void alx_reinit(struct alx_adapter *adpt)
+void alx_reinit(struct alx_adapter *adpt, bool in_task)
 {
 	WARN_ON(in_interrupt());
 
@@ -1812,7 +1813,7 @@ void alx_reinit(struct alx_adapter *adpt)
 	if (ALX_FLAG(adpt, HALT))
 		return;
 
-	alx_halt(adpt);
+	alx_halt(adpt, in_task);
 	alx_activate(adpt);
 
 	ALX_FLAG_CLEAR(adpt, RESETING);
@@ -1833,7 +1834,7 @@ static void alx_task(struct work_struct *work)
 	if (test_and_clear_bit(ALX_FLAG_TASK_RESET, &adpt->flags)) {
 		netif_info(adpt, hw, adpt->netdev,
 			   "task:alx_reinit\n");
-		alx_reinit(adpt);
+		alx_reinit(adpt, true);
 	}
 
 	if (test_and_clear_bit(ALX_FLAG_TASK_UPDATE_SMB, &adpt->flags))
@@ -2720,7 +2721,7 @@ static pci_ers_result_t alx_pci_error_detected(struct pci_dev *pdev,
 
 	if (netif_running(netdev)) {
 		netif_device_detach(netdev);
-		alx_halt(adpt);
+		alx_halt(adpt, false);
 	}
 	if (state == pci_channel_io_perm_failure)
 		rc = PCI_ERS_RESULT_DISCONNECT;