vfio/pci: Use RCU for error/request triggers to avoid circular locking

Thanks to a device generating an ACS violation during bus reset,
lockdep reported the following circular locking issue:

CPU0: SET_IRQS (MSI/X): holds igate, acquires memory_lock
CPU1: HOT_RESET: holds memory_lock, acquires pci_bus_sem
CPU2: AER: holds pci_bus_sem, acquires igate

This results in a potential 3-way deadlock.

Remove the pci_bus_sem->igate leg of the triangle by using RCU
to peek at the eventfd rather than locking it with igate.

Fixes: 3be3a074cf ("vfio-pci: Don't use device_lock around AER interrupt setup")
Signed-off-by: Alex Williamson <alex.williamson@nvidia.com>
Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>
Link: https://lore.kernel.org/r/20251124223623.2770706-1-alex@shazbot.org
Signed-off-by: Alex Williamson <alex@shazbot.org>

drivers/vfio/pci/vfio_pci_core.c

@@ -42,6 +42,40 @@ static bool nointxmask;
 static bool disable_vga;
 static bool disable_idle_d3;
 
+static void vfio_pci_eventfd_rcu_free(struct rcu_head *rcu)
+{
+	struct vfio_pci_eventfd *eventfd =
+		container_of(rcu, struct vfio_pci_eventfd, rcu);
+
+	eventfd_ctx_put(eventfd->ctx);
+	kfree(eventfd);
+}
+
+int vfio_pci_eventfd_replace_locked(struct vfio_pci_core_device *vdev,
+				    struct vfio_pci_eventfd __rcu **peventfd,
+				    struct eventfd_ctx *ctx)
+{
+	struct vfio_pci_eventfd *new = NULL;
+	struct vfio_pci_eventfd *old;
+
+	lockdep_assert_held(&vdev->igate);
+
+	if (ctx) {
+		new = kzalloc(sizeof(*new), GFP_KERNEL_ACCOUNT);
+		if (!new)
+			return -ENOMEM;
+
+		new->ctx = ctx;
+	}
+
+	old = rcu_replace_pointer(*peventfd, new,
+				  lockdep_is_held(&vdev->igate));
+	if (old)
+		call_rcu(&old->rcu, vfio_pci_eventfd_rcu_free);
+
+	return 0;
+}
+
 /* List of PF's that vfio_pci_core_sriov_configure() has been called on */
 static DEFINE_MUTEX(vfio_pci_sriov_pfs_mutex);
 static LIST_HEAD(vfio_pci_sriov_pfs);
@@ -697,14 +731,8 @@ void vfio_pci_core_close_device(struct vfio_device *core_vdev)
 	vfio_pci_dma_buf_cleanup(vdev);
 
 	mutex_lock(&vdev->igate);
-	if (vdev->err_trigger) {
-		eventfd_ctx_put(vdev->err_trigger);
-		vdev->err_trigger = NULL;
-	}
-	if (vdev->req_trigger) {
-		eventfd_ctx_put(vdev->req_trigger);
-		vdev->req_trigger = NULL;
-	}
+	vfio_pci_eventfd_replace_locked(vdev, &vdev->err_trigger, NULL);
+	vfio_pci_eventfd_replace_locked(vdev, &vdev->req_trigger, NULL);
 	mutex_unlock(&vdev->igate);
 }
 EXPORT_SYMBOL_GPL(vfio_pci_core_close_device);
@@ -1784,21 +1812,21 @@ void vfio_pci_core_request(struct vfio_device *core_vdev, unsigned int count)
 	struct vfio_pci_core_device *vdev =
 		container_of(core_vdev, struct vfio_pci_core_device, vdev);
 	struct pci_dev *pdev = vdev->pdev;
+	struct vfio_pci_eventfd *eventfd;
 
-	mutex_lock(&vdev->igate);
-
-	if (vdev->req_trigger) {
+	rcu_read_lock();
+	eventfd = rcu_dereference(vdev->req_trigger);
+	if (eventfd) {
 		if (!(count % 10))
 			pci_notice_ratelimited(pdev,
 				"Relaying device request to user (#%u)\n",
 				count);
-		eventfd_signal(vdev->req_trigger);
+		eventfd_signal(eventfd->ctx);
 	} else if (count == 0) {
 		pci_warn(pdev,
 			"No device request channel registered, blocked until released by user\n");
 	}
-
-	mutex_unlock(&vdev->igate);
+	rcu_read_unlock();
 }
 EXPORT_SYMBOL_GPL(vfio_pci_core_request);
 
@@ -2216,13 +2244,13 @@ pci_ers_result_t vfio_pci_core_aer_err_detected(struct pci_dev *pdev,
 						pci_channel_state_t state)
 {
 	struct vfio_pci_core_device *vdev = dev_get_drvdata(&pdev->dev);
+	struct vfio_pci_eventfd *eventfd;
 
-	mutex_lock(&vdev->igate);
-
-	if (vdev->err_trigger)
-		eventfd_signal(vdev->err_trigger);
-
-	mutex_unlock(&vdev->igate);
+	rcu_read_lock();
+	eventfd = rcu_dereference(vdev->err_trigger);
+	if (eventfd)
+		eventfd_signal(eventfd->ctx);
+	rcu_read_unlock();
 
 	return PCI_ERS_RESULT_CAN_RECOVER;
 }

drivers/vfio/pci/vfio_pci_intrs.c

@@ -731,21 +731,27 @@ static int vfio_pci_set_msi_trigger(struct vfio_pci_core_device *vdev,
 	return 0;
 }
 
-static int vfio_pci_set_ctx_trigger_single(struct eventfd_ctx **ctx,
+static int vfio_pci_set_ctx_trigger_single(struct vfio_pci_core_device *vdev,
+					   struct vfio_pci_eventfd __rcu **peventfd,
 					   unsigned int count, uint32_t flags,
 					   void *data)
 {
 	/* DATA_NONE/DATA_BOOL enables loopback testing */
 	if (flags & VFIO_IRQ_SET_DATA_NONE) {
-		if (*ctx) {
-			if (count) {
-				eventfd_signal(*ctx);
-			} else {
-				eventfd_ctx_put(*ctx);
-				*ctx = NULL;
-			}
+		struct vfio_pci_eventfd *eventfd;
+
+		eventfd = rcu_dereference_protected(*peventfd,
+						lockdep_is_held(&vdev->igate));
+
+		if (!eventfd)
+			return -EINVAL;
+
+		if (count) {
+			eventfd_signal(eventfd->ctx);
 			return 0;
 		}
+
+		return vfio_pci_eventfd_replace_locked(vdev, peventfd, NULL);
 	} else if (flags & VFIO_IRQ_SET_DATA_BOOL) {
 		uint8_t trigger;
 
@@ -753,8 +759,15 @@ static int vfio_pci_set_ctx_trigger_single(struct eventfd_ctx **ctx,
 			return -EINVAL;
 
 		trigger = *(uint8_t *)data;
-		if (trigger && *ctx)
-			eventfd_signal(*ctx);
+
+		if (trigger) {
+			struct vfio_pci_eventfd *eventfd =
+					rcu_dereference_protected(*peventfd,
+					lockdep_is_held(&vdev->igate));
+
+			if (eventfd)
+				eventfd_signal(eventfd->ctx);
+		}
 
 		return 0;
 	} else if (flags & VFIO_IRQ_SET_DATA_EVENTFD) {
@@ -765,22 +778,23 @@ static int vfio_pci_set_ctx_trigger_single(struct eventfd_ctx **ctx,
 
 		fd = *(int32_t *)data;
 		if (fd == -1) {
-			if (*ctx)
-				eventfd_ctx_put(*ctx);
-			*ctx = NULL;
+			return vfio_pci_eventfd_replace_locked(vdev,
+							       peventfd, NULL);
 		} else if (fd >= 0) {
 			struct eventfd_ctx *efdctx;
+			int ret;
 
 			efdctx = eventfd_ctx_fdget(fd);
 			if (IS_ERR(efdctx))
 				return PTR_ERR(efdctx);
 
-			if (*ctx)
-				eventfd_ctx_put(*ctx);
+			ret = vfio_pci_eventfd_replace_locked(vdev,
+							      peventfd, efdctx);
+			if (ret)
+				eventfd_ctx_put(efdctx);
 
-			*ctx = efdctx;
+			return ret;
 		}
-		return 0;
 	}
 
 	return -EINVAL;
@@ -793,7 +807,7 @@ static int vfio_pci_set_err_trigger(struct vfio_pci_core_device *vdev,
 	if (index != VFIO_PCI_ERR_IRQ_INDEX || start != 0 || count > 1)
 		return -EINVAL;
 
-	return vfio_pci_set_ctx_trigger_single(&vdev->err_trigger,
+	return vfio_pci_set_ctx_trigger_single(vdev, &vdev->err_trigger,
 					       count, flags, data);
 }
 
@@ -804,7 +818,7 @@ static int vfio_pci_set_req_trigger(struct vfio_pci_core_device *vdev,
 	if (index != VFIO_PCI_REQ_IRQ_INDEX || start != 0 || count > 1)
 		return -EINVAL;
 
-	return vfio_pci_set_ctx_trigger_single(&vdev->req_trigger,
+	return vfio_pci_set_ctx_trigger_single(vdev, &vdev->req_trigger,
 					       count, flags, data);
 }
 

drivers/vfio/pci/vfio_pci_priv.h

@@ -26,6 +26,10 @@ struct vfio_pci_ioeventfd {
 bool vfio_pci_intx_mask(struct vfio_pci_core_device *vdev);
 void vfio_pci_intx_unmask(struct vfio_pci_core_device *vdev);
 
+int vfio_pci_eventfd_replace_locked(struct vfio_pci_core_device *vdev,
+				    struct vfio_pci_eventfd __rcu **peventfd,
+				    struct eventfd_ctx *ctx);
+
 int vfio_pci_set_irqs_ioctl(struct vfio_pci_core_device *vdev, uint32_t flags,
 			    unsigned index, unsigned start, unsigned count,
 			    void *data);

include/linux/vfio_pci_core.h

@@ -12,6 +12,7 @@
 #include <linux/pci.h>
 #include <linux/vfio.h>
 #include <linux/irqbypass.h>
+#include <linux/rcupdate.h>
 #include <linux/types.h>
 #include <linux/uuid.h>
 #include <linux/notifier.h>
@@ -29,6 +30,11 @@ struct vfio_pci_region;
 struct p2pdma_provider;
 struct dma_buf_phys_vec;
 
+struct vfio_pci_eventfd {
+	struct eventfd_ctx	*ctx;
+	struct rcu_head		rcu;
+};
+
 struct vfio_pci_regops {
 	ssize_t (*rw)(struct vfio_pci_core_device *vdev, char __user *buf,
 		      size_t count, loff_t *ppos, bool iswrite);
@@ -124,8 +130,8 @@ struct vfio_pci_core_device {
 	struct pci_saved_state	*pci_saved_state;
 	struct pci_saved_state	*pm_save;
 	int			ioeventfds_nr;
-	struct eventfd_ctx	*err_trigger;
-	struct eventfd_ctx	*req_trigger;
+	struct vfio_pci_eventfd __rcu *err_trigger;
+	struct vfio_pci_eventfd __rcu *req_trigger;
 	struct eventfd_ctx	*pm_wake_eventfd_ctx;
 	struct list_head	dummy_resources_list;
 	struct mutex		ioeventfds_lock;