Merge: s390/pci: Fix immediate re-add of PCI function after remove

MR: https://gitlab.com/redhat/centos-stream/src/kernel/centos-stream-10/-/merge_requests/1001

JIRA: https://issues.redhat.com/browse/RHEL-94944

commit ids:
```
c4a585e
42420c5
05a2538
d76f963
47c3978
4b1815a
774a1fa
```

This fix is urgent!

Signed-off-by: Mete Durlu <mdurlu@redhat.com>

Approved-by: Steve Best <sbest@redhat.com>
Approved-by: Tony Camuso <tcamuso@redhat.com>
Approved-by: CKI KWF Bot <cki-ci-bot+kwf-gitlab-com@redhat.com>

Merged-by: CKI GitLab Kmaint Pipeline Bot <26919896-cki-kmaint-pipeline-bot@users.noreply.gitlab.com>

Author: CKI KWF Bot

arch/s390/pci/pci.c

@@ -44,6 +44,7 @@
 /* list of all detected zpci devices */
 static LIST_HEAD(zpci_list);
 static DEFINE_SPINLOCK(zpci_list_lock);
+static DEFINE_MUTEX(zpci_add_remove_lock);
 
 static DECLARE_BITMAP(zpci_domain, ZPCI_DOMAIN_BITMAP_SIZE);
 static DEFINE_SPINLOCK(zpci_domain_lock);
@@ -69,6 +70,15 @@ EXPORT_SYMBOL_GPL(zpci_aipb);
 struct airq_iv *zpci_aif_sbv;
 EXPORT_SYMBOL_GPL(zpci_aif_sbv);
 
+void zpci_zdev_put(struct zpci_dev *zdev)
+{
+	if (!zdev)
+		return;
+	mutex_lock(&zpci_add_remove_lock);
+	kref_put_lock(&zdev->kref, zpci_release_device, &zpci_list_lock);
+	mutex_unlock(&zpci_add_remove_lock);
+}
+
 struct zpci_dev *get_zdev_by_fid(u32 fid)
 {
 	struct zpci_dev *tmp, *zdev = NULL;
@@ -831,6 +841,7 @@ int zpci_add_device(struct zpci_dev *zdev)
 {
 	int rc;
 
+	mutex_lock(&zpci_add_remove_lock);
 	zpci_dbg(1, "add fid:%x, fh:%x, c:%d\n", zdev->fid, zdev->fh, zdev->state);
 	rc = zpci_init_iommu(zdev);
 	if (rc)
@@ -844,12 +855,14 @@ int zpci_add_device(struct zpci_dev *zdev)
 	spin_lock(&zpci_list_lock);
 	list_add_tail(&zdev->entry, &zpci_list);
 	spin_unlock(&zpci_list_lock);
+	mutex_unlock(&zpci_add_remove_lock);
 	return 0;
 
 error_destroy_iommu:
 	zpci_destroy_iommu(zdev);
 error:
 	zpci_dbg(0, "add fid:%x, rc:%d\n", zdev->fid, rc);
+	mutex_unlock(&zpci_add_remove_lock);
 	return rc;
 }
 
@@ -919,60 +932,44 @@ int zpci_deconfigure_device(struct zpci_dev *zdev)
  * @zdev: the zpci_dev that was reserved
  *
  * Handle the case that a given zPCI function was reserved by another system.
- * After a call to this function the zpci_dev can not be found via
- * get_zdev_by_fid() anymore but may still be accessible via existing
- * references though it will not be functional anymore.
  */
 void zpci_device_reserved(struct zpci_dev *zdev)
 {
-	/*
-	 * Remove device from zpci_list as it is going away. This also
-	 * makes sure we ignore subsequent zPCI events for this device.
-	 */
-	spin_lock(&zpci_list_lock);
-	list_del(&zdev->entry);
-	spin_unlock(&zpci_list_lock);
+	lockdep_assert_held(&zdev->state_lock);
+	/* We may declare the device reserved multiple times */
+	if (zdev->state == ZPCI_FN_STATE_RESERVED)
+		return;
 	zdev->state = ZPCI_FN_STATE_RESERVED;
 	zpci_dbg(3, "rsv fid:%x\n", zdev->fid);
+	/*
+	 * The underlying device is gone. Allow the zdev to be freed
+	 * as soon as all other references are gone by accounting for
+	 * the removal as a dropped reference.
+	 */
 	zpci_zdev_put(zdev);
 }
 
 void zpci_release_device(struct kref *kref)
 {
 	struct zpci_dev *zdev = container_of(kref, struct zpci_dev, kref);
-	int ret;
+
+	lockdep_assert_held(&zpci_add_remove_lock);
+	WARN_ON(zdev->state != ZPCI_FN_STATE_RESERVED);
+	/*
+	 * We already hold zpci_list_lock thanks to kref_put_lock().
+	 * This makes sure no new reference can be taken from the list.
+	 */
+	list_del(&zdev->entry);
+	spin_unlock(&zpci_list_lock);
 
 	if (zdev->has_hp_slot)
 		zpci_exit_slot(zdev);
 
-	if (zdev->zbus->bus)
-		zpci_bus_remove_device(zdev, false);
-
-	if (zdev_enabled(zdev))
-		zpci_disable_device(zdev);
+	if (zdev->has_resources)
+		zpci_cleanup_bus_resources(zdev);
 
-	switch (zdev->state) {
-	case ZPCI_FN_STATE_CONFIGURED:
-		ret = sclp_pci_deconfigure(zdev->fid);
-		zpci_dbg(3, "deconf fid:%x, rc:%d\n", zdev->fid, ret);
-		fallthrough;
-	case ZPCI_FN_STATE_STANDBY:
-		if (zdev->has_hp_slot)
-			zpci_exit_slot(zdev);
-		spin_lock(&zpci_list_lock);
-		list_del(&zdev->entry);
-		spin_unlock(&zpci_list_lock);
-		zpci_dbg(3, "rsv fid:%x\n", zdev->fid);
-		fallthrough;
-	case ZPCI_FN_STATE_RESERVED:
-		if (zdev->has_resources)
-			zpci_cleanup_bus_resources(zdev);
-		zpci_bus_device_unregister(zdev);
-		zpci_destroy_iommu(zdev);
-		fallthrough;
-	default:
-		break;
-	}
+	zpci_bus_device_unregister(zdev);
+	zpci_destroy_iommu(zdev);
 	zpci_dbg(3, "rem fid:%x\n", zdev->fid);
 	kfree_rcu(zdev, rcu);
 }

arch/s390/pci/pci_bus.h

@@ -17,11 +17,8 @@ int zpci_bus_scan_device(struct zpci_dev *zdev);
 void zpci_bus_remove_device(struct zpci_dev *zdev, bool set_error);
 
 void zpci_release_device(struct kref *kref);
-static inline void zpci_zdev_put(struct zpci_dev *zdev)
-{
-	if (zdev)
-		kref_put(&zdev->kref, zpci_release_device);
-}
+
+void zpci_zdev_put(struct zpci_dev *zdev);
 
 static inline void zpci_zdev_get(struct zpci_dev *zdev)
 {

arch/s390/pci/pci_clp.c

@@ -427,6 +427,8 @@ static void __clp_add(struct clp_fh_list_entry *entry, void *data)
 		return;
 	}
 	zdev = zpci_create_device(entry->fid, entry->fh, entry->config_state);
+	if (IS_ERR(zdev))
+		return;
 	list_add_tail(&zdev->entry, scan_list);
 }
 

arch/s390/pci/pci_event.c

@@ -335,6 +335,22 @@ static void zpci_event_hard_deconfigured(struct zpci_dev *zdev, u32 fh)
 	zdev->state = ZPCI_FN_STATE_STANDBY;
 }
 
+static void zpci_event_reappear(struct zpci_dev *zdev)
+{
+	lockdep_assert_held(&zdev->state_lock);
+	/*
+	 * The zdev is in the reserved state. This means that it was presumed to
+	 * go away but there are still undropped references. Now, the platform
+	 * announced its availability again. Bring back the lingering zdev
+	 * to standby. This is safe because we hold a temporary reference
+	 * now so that it won't go away. Account for the re-appearance of the
+	 * underlying device by incrementing the reference count.
+	 */
+	zdev->state = ZPCI_FN_STATE_STANDBY;
+	zpci_zdev_get(zdev);
+	zpci_dbg(1, "rea fid:%x, fh:%x\n", zdev->fid, zdev->fh);
+}
+
 static void __zpci_event_availability(struct zpci_ccdf_avail *ccdf)
 {
 	struct zpci_dev *zdev = get_zdev_by_fid(ccdf->fid);
@@ -358,8 +374,10 @@ static void __zpci_event_availability(struct zpci_ccdf_avail *ccdf)
 				break;
 			}
 		} else {
+			if (zdev->state == ZPCI_FN_STATE_RESERVED)
+				zpci_event_reappear(zdev);
 			/* the configuration request may be stale */
-			if (zdev->state != ZPCI_FN_STATE_STANDBY)
+			else if (zdev->state != ZPCI_FN_STATE_STANDBY)
 				break;
 			zdev->state = ZPCI_FN_STATE_CONFIGURED;
 		}
@@ -375,6 +393,8 @@ static void __zpci_event_availability(struct zpci_ccdf_avail *ccdf)
 				break;
 			}
 		} else {
+			if (zdev->state == ZPCI_FN_STATE_RESERVED)
+				zpci_event_reappear(zdev);
 			zpci_update_fh(zdev, ccdf->fh);
 		}
 		break;

drivers/pci/hotplug/s390_pci_hpc.c

@@ -59,16 +59,15 @@ static int disable_slot(struct hotplug_slot *hotplug_slot)
 
 	pdev = pci_get_slot(zdev->zbus->bus, zdev->devfn);
 	if (pdev && pci_num_vf(pdev)) {
-		pci_dev_put(pdev);
 		rc = -EBUSY;
 		goto out;
 	}
 
 	rc = zpci_deconfigure_device(zdev);
 out:
-	mutex_unlock(&zdev->state_lock);
 	if (pdev)
 		pci_dev_put(pdev);
+	mutex_unlock(&zdev->state_lock);
 	return rc;
 }