Merging upstream version 4.19.289.

Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>

5 files changed, 54 insertions, 3 deletions

diff --git a/drivers/pci/controller/pci-hyperv.c b/drivers/pci/controller/pci-hyperv.c
index 63c79e140..f5f201bfc 100644
--- a/drivers/pci/controller/pci-hyperv.c
+++ b/drivers/pci/controller/pci-hyperv.c
@@ -2445,6 +2445,24 @@ static int hv_pci_query_relations(struct hv_device *hdev)
 	if (!ret)
 		ret = wait_for_response(hdev, &comp);
 
+	/*
+	 * In the case of fast device addition/removal, it's possible that
+	 * vmbus_sendpacket() or wait_for_response() returns -ENODEV but we
+	 * already got a PCI_BUS_RELATIONS* message from the host and the
+	 * channel callback already scheduled a work to hbus->wq, which can be
+	 * running pci_devices_present_work() -> survey_child_resources() ->
+	 * complete(&hbus->survey_event), even after hv_pci_query_relations()
+	 * exits and the stack variable 'comp' is no longer valid; as a result,
+	 * a hang or a page fault may happen when the complete() calls
+	 * raw_spin_lock_irqsave(). Flush hbus->wq before we exit from
+	 * hv_pci_query_relations() to avoid the issues. Note: if 'ret' is
+	 * -ENODEV, there can't be any more work item scheduled to hbus->wq
+	 * after the flush_workqueue(): see vmbus_onoffer_rescind() ->
+	 * vmbus_reset_channel_cb(), vmbus_rescind_cleanup() ->
+	 * channel->rescind = true.
+	 */
+	flush_workqueue(hbus->wq);
+
 	return ret;
 }
 
diff --git a/drivers/pci/hotplug/pciehp.h b/drivers/pci/hotplug/pciehp.h
index ef6071807..522719ca1 100644
--- a/drivers/pci/hotplug/pciehp.h
+++ b/drivers/pci/hotplug/pciehp.h
@@ -84,6 +84,8 @@ struct slot {
  * @reset_lock: prevents access to the Data Link Layer Link Active bit in the
  *	Link Status register and to the Presence Detect State bit in the Slot
  *	Status register during a slot reset which may cause them to flap
+ * @depth: Number of additional hotplug ports in the path to the root bus,
+ *	used as lock subclass for @reset_lock
  * @slot: pointer to the controller's slot structure
  * @queue: wait queue to wake up on reception of a Command Completed event,
  *	used for synchronous writes to the Slot Control register
@@ -115,6 +117,7 @@ struct controller {
 	struct mutex ctrl_lock;
 	struct pcie_device *pcie;
 	struct rw_semaphore reset_lock;
+	unsigned int depth;
 	struct slot *slot;
 	wait_queue_head_t queue;
 	u32 slot_cap;
diff --git a/drivers/pci/hotplug/pciehp_core.c b/drivers/pci/hotplug/pciehp_core.c
index 518c46f8e..5ebfff935 100644
--- a/drivers/pci/hotplug/pciehp_core.c
+++ b/drivers/pci/hotplug/pciehp_core.c
@@ -215,7 +215,7 @@ static void pciehp_check_presence(struct controller *ctrl)
 	struct slot *slot = ctrl->slot;
 	u8 occupied;
 
-	down_read(&ctrl->reset_lock);
+	down_read_nested(&ctrl->reset_lock, ctrl->depth);
 	mutex_lock(&slot->lock);
 
 	pciehp_get_adapter_status(slot, &occupied);
diff --git a/drivers/pci/hotplug/pciehp_hpc.c b/drivers/pci/hotplug/pciehp_hpc.c
index 279544523..7392b26e9 100644
--- a/drivers/pci/hotplug/pciehp_hpc.c
+++ b/drivers/pci/hotplug/pciehp_hpc.c
@@ -674,7 +674,7 @@ static irqreturn_t pciehp_ist(int irq, void *dev_id)
 	 * Disable requests have higher priority than Presence Detect Changed
 	 * or Data Link Layer State Changed events.
 	 */
-	down_read(&ctrl->reset_lock);
+	down_read_nested(&ctrl->reset_lock, ctrl->depth);
 	if (events & DISABLE_SLOT)
 		pciehp_handle_disable_request(slot);
 	else if (events & (PCI_EXP_SLTSTA_PDC | PCI_EXP_SLTSTA_DLLSC))
@@ -785,7 +785,7 @@ int pciehp_reset_slot(struct slot *slot, int probe)
 	if (probe)
 		return 0;
 
-	down_write(&ctrl->reset_lock);
+	down_write_nested(&ctrl->reset_lock, ctrl->depth);
 
 	if (!ATTN_BUTTN(ctrl)) {
 		ctrl_mask |= PCI_EXP_SLTCTL_PDCE;
@@ -872,6 +872,20 @@ static inline void dbg_ctrl(struct controller *ctrl)
 
 #define FLAG(x, y)	(((x) & (y)) ? '+' : '-')
 
+static inline int pcie_hotplug_depth(struct pci_dev *dev)
+{
+	struct pci_bus *bus = dev->bus;
+	int depth = 0;
+
+	while (bus->parent) {
+		bus = bus->parent;
+		if (bus->self && bus->self->is_hotplug_bridge)
+			depth++;
+	}
+
+	return depth;
+}
+
 struct controller *pcie_init(struct pcie_device *dev)
 {
 	struct controller *ctrl;
@@ -884,6 +898,7 @@ struct controller *pcie_init(struct pcie_device *dev)
 		goto abort;
 
 	ctrl->pcie = dev;
+	ctrl->depth = pcie_hotplug_depth(dev->port);
 	pcie_capability_read_dword(pdev, PCI_EXP_SLTCAP, &slot_cap);
 
 	if (pdev->hotplug_user_indicators)
diff --git a/drivers/pci/hotplug/pciehp_pci.c b/drivers/pci/hotplug/pciehp_pci.c
index 5c58c22e0..a32023afa 100644
--- a/drivers/pci/hotplug/pciehp_pci.c
+++ b/drivers/pci/hotplug/pciehp_pci.c
@@ -55,7 +55,14 @@ int pciehp_configure_device(struct slot *p_slot)
 
 	pci_assign_unassigned_bridge_resources(bridge);
 	pcie_bus_configure_settings(parent);
+
+	/*
+	 * Release reset_lock during driver binding
+	 * to avoid AB-BA deadlock with device_lock.
+	 */
+	up_read(&ctrl->reset_lock);
 	pci_bus_add_devices(parent);
+	down_read_nested(&ctrl->reset_lock, ctrl->depth);
 
  out:
 	pci_unlock_rescan_remove();
@@ -91,7 +98,15 @@ void pciehp_unconfigure_device(struct slot *p_slot)
 				pci_walk_bus(dev->subordinate,
 					     pci_dev_set_disconnected, NULL);
 		}
+
+		/*
+		 * Release reset_lock during driver unbinding
+		 * to avoid AB-BA deadlock with device_lock.
+		 */
+		up_read(&ctrl->reset_lock);
 		pci_stop_and_remove_bus_device(dev);
+		down_read_nested(&ctrl->reset_lock, ctrl->depth);
+
 		/*
 		 * Ensure that no new Requests will be generated from
 		 * the device.