9 files changed, 232 insertions, 129 deletions

diff --git a/drivers/vfio/container.c b/drivers/vfio/container.c
index d74164abb..ab9d8e348 100644
--- a/drivers/vfio/container.c
+++ b/drivers/vfio/container.c
@@ -366,7 +366,7 @@ static int vfio_fops_open(struct inode *inode, struct file *filep)
 {
 	struct vfio_container *container;
 
-	container = kzalloc(sizeof(*container), GFP_KERNEL);
+	container = kzalloc(sizeof(*container), GFP_KERNEL_ACCOUNT);
 	if (!container)
 		return -ENOMEM;
 
diff --git a/drivers/vfio/fsl-mc/vfio_fsl_mc_intr.c b/drivers/vfio/fsl-mc/vfio_fsl_mc_intr.c
index 7b428eac3..b125b6edf 100644
--- a/drivers/vfio/fsl-mc/vfio_fsl_mc_intr.c
+++ b/drivers/vfio/fsl-mc/vfio_fsl_mc_intr.c
@@ -142,13 +142,14 @@ static int vfio_fsl_mc_set_irq_trigger(struct vfio_fsl_mc_device *vdev,
 	irq = &vdev->mc_irqs[index];
 
 	if (flags & VFIO_IRQ_SET_DATA_NONE) {
-		vfio_fsl_mc_irq_handler(hwirq, irq);
+		if (irq->trigger)
+			eventfd_signal(irq->trigger, 1);
 
 	} else if (flags & VFIO_IRQ_SET_DATA_BOOL) {
 		u8 trigger = *(u8 *)data;
 
-		if (trigger)
-			vfio_fsl_mc_irq_handler(hwirq, irq);
+		if (trigger && irq->trigger)
+			eventfd_signal(irq->trigger, 1);
 	}
 
 	return 0;
diff --git a/drivers/vfio/pci/vfio_pci_config.c b/drivers/vfio/pci/vfio_pci_config.c
index 4a350421c..523e0144c 100644
--- a/drivers/vfio/pci/vfio_pci_config.c
+++ b/drivers/vfio/pci/vfio_pci_config.c
@@ -1244,7 +1244,7 @@ static int vfio_msi_cap_len(struct vfio_pci_core_device *vdev, u8 pos)
 	if (vdev->msi_perm)
 		return len;
 
-	vdev->msi_perm = kmalloc(sizeof(struct perm_bits), GFP_KERNEL);
+	vdev->msi_perm = kmalloc(sizeof(struct perm_bits), GFP_KERNEL_ACCOUNT);
 	if (!vdev->msi_perm)
 		return -ENOMEM;
 
@@ -1731,11 +1731,11 @@ int vfio_config_init(struct vfio_pci_core_device *vdev)
 	 * no requirements on the length of a capability, so the gap between
 	 * capabilities needs byte granularity.
 	 */
-	map = kmalloc(pdev->cfg_size, GFP_KERNEL);
+	map = kmalloc(pdev->cfg_size, GFP_KERNEL_ACCOUNT);
 	if (!map)
 		return -ENOMEM;
 
-	vconfig = kmalloc(pdev->cfg_size, GFP_KERNEL);
+	vconfig = kmalloc(pdev->cfg_size, GFP_KERNEL_ACCOUNT);
 	if (!vconfig) {
 		kfree(map);
 		return -ENOMEM;
diff --git a/drivers/vfio/pci/vfio_pci_core.c b/drivers/vfio/pci/vfio_pci_core.c
index e030c2120..f357fd157 100644
--- a/drivers/vfio/pci/vfio_pci_core.c
+++ b/drivers/vfio/pci/vfio_pci_core.c
@@ -141,7 +141,8 @@ static void vfio_pci_probe_mmaps(struct vfio_pci_core_device *vdev)
 			 * of the exclusive page in case that hot-add
 			 * device's bar is assigned into it.
 			 */
-			dummy_res = kzalloc(sizeof(*dummy_res), GFP_KERNEL);
+			dummy_res =
+				kzalloc(sizeof(*dummy_res), GFP_KERNEL_ACCOUNT);
 			if (dummy_res == NULL)
 				goto no_mmap;
 
@@ -856,7 +857,7 @@ int vfio_pci_core_register_dev_region(struct vfio_pci_core_device *vdev,
 
 	region = krealloc(vdev->region,
 			  (vdev->num_regions + 1) * sizeof(*region),
-			  GFP_KERNEL);
+			  GFP_KERNEL_ACCOUNT);
 	if (!region)
 		return -ENOMEM;
 
@@ -1637,7 +1638,7 @@ static int __vfio_pci_add_vma(struct vfio_pci_core_device *vdev,
 {
 	struct vfio_pci_mmap_vma *mmap_vma;
 
-	mmap_vma = kmalloc(sizeof(*mmap_vma), GFP_KERNEL);
+	mmap_vma = kmalloc(sizeof(*mmap_vma), GFP_KERNEL_ACCOUNT);
 	if (!mmap_vma)
 		return -ENOMEM;
 
diff --git a/drivers/vfio/pci/vfio_pci_igd.c b/drivers/vfio/pci/vfio_pci_igd.c
index 5e6ca5926..dd70e2431 100644
--- a/drivers/vfio/pci/vfio_pci_igd.c
+++ b/drivers/vfio/pci/vfio_pci_igd.c
@@ -180,7 +180,7 @@ static int vfio_pci_igd_opregion_init(struct vfio_pci_core_device *vdev)
 	if (!addr || !(~addr))
 		return -ENODEV;
 
-	opregionvbt = kzalloc(sizeof(*opregionvbt), GFP_KERNEL);
+	opregionvbt = kzalloc(sizeof(*opregionvbt), GFP_KERNEL_ACCOUNT);
 	if (!opregionvbt)
 		return -ENOMEM;
 
diff --git a/drivers/vfio/pci/vfio_pci_intrs.c b/drivers/vfio/pci/vfio_pci_intrs.c
index 40c3d7cf1..03246a59b 100644
--- a/drivers/vfio/pci/vfio_pci_intrs.c
+++ b/drivers/vfio/pci/vfio_pci_intrs.c
@@ -55,17 +55,24 @@ static void vfio_send_intx_eventfd(void *opaque, void *unused)
 {
 	struct vfio_pci_core_device *vdev = opaque;
 
-	if (likely(is_intx(vdev) && !vdev->virq_disabled))
-		eventfd_signal(vdev->ctx[0].trigger, 1);
+	if (likely(is_intx(vdev) && !vdev->virq_disabled)) {
+		struct eventfd_ctx *trigger;
+
+		trigger = READ_ONCE(vdev->ctx[0].trigger);
+		if (likely(trigger))
+			eventfd_signal(trigger, 1);
+	}
 }
 
 /* Returns true if the INTx vfio_pci_irq_ctx.masked value is changed. */
-bool vfio_pci_intx_mask(struct vfio_pci_core_device *vdev)
+static bool __vfio_pci_intx_mask(struct vfio_pci_core_device *vdev)
 {
 	struct pci_dev *pdev = vdev->pdev;
 	unsigned long flags;
 	bool masked_changed = false;
 
+	lockdep_assert_held(&vdev->igate);
+
 	spin_lock_irqsave(&vdev->irqlock, flags);
 
 	/*
@@ -95,6 +102,17 @@ bool vfio_pci_intx_mask(struct vfio_pci_core_device *vdev)
 	return masked_changed;
 }
 
+bool vfio_pci_intx_mask(struct vfio_pci_core_device *vdev)
+{
+	bool mask_changed;
+
+	mutex_lock(&vdev->igate);
+	mask_changed = __vfio_pci_intx_mask(vdev);
+	mutex_unlock(&vdev->igate);
+
+	return mask_changed;
+}
+
 /*
  * If this is triggered by an eventfd, we can't call eventfd_signal
  * or else we'll deadlock on the eventfd wait queue.  Return >0 when
@@ -137,12 +155,21 @@ static int vfio_pci_intx_unmask_handler(void *opaque, void *unused)
 	return ret;
 }
 
-void vfio_pci_intx_unmask(struct vfio_pci_core_device *vdev)
+static void __vfio_pci_intx_unmask(struct vfio_pci_core_device *vdev)
 {
+	lockdep_assert_held(&vdev->igate);
+
 	if (vfio_pci_intx_unmask_handler(vdev, NULL) > 0)
 		vfio_send_intx_eventfd(vdev, NULL);
 }
 
+void vfio_pci_intx_unmask(struct vfio_pci_core_device *vdev)
+{
+	mutex_lock(&vdev->igate);
+	__vfio_pci_intx_unmask(vdev);
+	mutex_unlock(&vdev->igate);
+}
+
 static irqreturn_t vfio_intx_handler(int irq, void *dev_id)
 {
 	struct vfio_pci_core_device *vdev = dev_id;
@@ -169,95 +196,104 @@ static irqreturn_t vfio_intx_handler(int irq, void *dev_id)
 	return ret;
 }
 
-static int vfio_intx_enable(struct vfio_pci_core_device *vdev)
+static int vfio_intx_enable(struct vfio_pci_core_device *vdev,
+			    struct eventfd_ctx *trigger)
 {
+	struct pci_dev *pdev = vdev->pdev;
+	unsigned long irqflags;
+	char *name;
+	int ret;
+
 	if (!is_irq_none(vdev))
 		return -EINVAL;
 
-	if (!vdev->pdev->irq)
+	if (!pdev->irq)
 		return -ENODEV;
 
-	vdev->ctx = kzalloc(sizeof(struct vfio_pci_irq_ctx), GFP_KERNEL);
+	name = kasprintf(GFP_KERNEL_ACCOUNT, "vfio-intx(%s)", pci_name(pdev));
+	if (!name)
+		return -ENOMEM;
+
+	vdev->ctx = kzalloc(sizeof(struct vfio_pci_irq_ctx), GFP_KERNEL_ACCOUNT);
 	if (!vdev->ctx)
 		return -ENOMEM;
 
 	vdev->num_ctx = 1;
 
+	vdev->ctx[0].name = name;
+	vdev->ctx[0].trigger = trigger;
+
 	/*
-	 * If the virtual interrupt is masked, restore it.  Devices
-	 * supporting DisINTx can be masked at the hardware level
-	 * here, non-PCI-2.3 devices will have to wait until the
-	 * interrupt is enabled.
+	 * Fill the initial masked state based on virq_disabled.  After
+	 * enable, changing the DisINTx bit in vconfig directly changes INTx
+	 * masking.  igate prevents races during setup, once running masked
+	 * is protected via irqlock.
+	 *
+	 * Devices supporting DisINTx also reflect the current mask state in
+	 * the physical DisINTx bit, which is not affected during IRQ setup.
+	 *
+	 * Devices without DisINTx support require an exclusive interrupt.
+	 * IRQ masking is performed at the IRQ chip.  Again, igate protects
+	 * against races during setup and IRQ handlers and irqfds are not
+	 * yet active, therefore masked is stable and can be used to
+	 * conditionally auto-enable the IRQ.
+	 *
+	 * irq_type must be stable while the IRQ handler is registered,
+	 * therefore it must be set before request_irq().
 	 */
 	vdev->ctx[0].masked = vdev->virq_disabled;
-	if (vdev->pci_2_3)
-		pci_intx(vdev->pdev, !vdev->ctx[0].masked);
+	if (vdev->pci_2_3) {
+		pci_intx(pdev, !vdev->ctx[0].masked);
+		irqflags = IRQF_SHARED;
+	} else {
+		irqflags = vdev->ctx[0].masked ? IRQF_NO_AUTOEN : 0;
+	}
 
 	vdev->irq_type = VFIO_PCI_INTX_IRQ_INDEX;
 
+	ret = request_irq(pdev->irq, vfio_intx_handler,
+			  irqflags, vdev->ctx[0].name, vdev);
+	if (ret) {
+		vdev->irq_type = VFIO_PCI_NUM_IRQS;
+		kfree(name);
+		vdev->num_ctx = 0;
+		kfree(vdev->ctx);
+		return ret;
+	}
+
 	return 0;
 }
 
-static int vfio_intx_set_signal(struct vfio_pci_core_device *vdev, int fd)
+static int vfio_intx_set_signal(struct vfio_pci_core_device *vdev,
+				struct eventfd_ctx *trigger)
 {
 	struct pci_dev *pdev = vdev->pdev;
-	unsigned long irqflags = IRQF_SHARED;
-	struct eventfd_ctx *trigger;
-	unsigned long flags;
-	int ret;
+	struct eventfd_ctx *old;
 
-	if (vdev->ctx[0].trigger) {
-		free_irq(pdev->irq, vdev);
-		kfree(vdev->ctx[0].name);
-		eventfd_ctx_put(vdev->ctx[0].trigger);
-		vdev->ctx[0].trigger = NULL;
-	}
-
-	if (fd < 0) /* Disable only */
-		return 0;
-
-	vdev->ctx[0].name = kasprintf(GFP_KERNEL, "vfio-intx(%s)",
-				      pci_name(pdev));
-	if (!vdev->ctx[0].name)
-		return -ENOMEM;
+	old = vdev->ctx[0].trigger;
 
-	trigger = eventfd_ctx_fdget(fd);
-	if (IS_ERR(trigger)) {
-		kfree(vdev->ctx[0].name);
-		return PTR_ERR(trigger);
-	}
+	WRITE_ONCE(vdev->ctx[0].trigger, trigger);
 
-	vdev->ctx[0].trigger = trigger;
-
-	if (!vdev->pci_2_3)
-		irqflags = 0;
-
-	ret = request_irq(pdev->irq, vfio_intx_handler,
-			  irqflags, vdev->ctx[0].name, vdev);
-	if (ret) {
-		vdev->ctx[0].trigger = NULL;
-		kfree(vdev->ctx[0].name);
-		eventfd_ctx_put(trigger);
-		return ret;
+	/* Releasing an old ctx requires synchronizing in-flight users */
+	if (old) {
+		synchronize_irq(pdev->irq);
+		vfio_virqfd_flush_thread(&vdev->ctx[0].unmask);
+		eventfd_ctx_put(old);
 	}
 
-	/*
-	 * INTx disable will stick across the new irq setup,
-	 * disable_irq won't.
-	 */
-	spin_lock_irqsave(&vdev->irqlock, flags);
-	if (!vdev->pci_2_3 && vdev->ctx[0].masked)
-		disable_irq_nosync(pdev->irq);
-	spin_unlock_irqrestore(&vdev->irqlock, flags);
-
 	return 0;
 }
 
 static void vfio_intx_disable(struct vfio_pci_core_device *vdev)
 {
+	struct pci_dev *pdev = vdev->pdev;
+
 	vfio_virqfd_disable(&vdev->ctx[0].unmask);
 	vfio_virqfd_disable(&vdev->ctx[0].mask);
-	vfio_intx_set_signal(vdev, -1);
+	free_irq(pdev->irq, vdev);
+	if (vdev->ctx[0].trigger)
+		eventfd_ctx_put(vdev->ctx[0].trigger);
+	kfree(vdev->ctx[0].name);
 	vdev->irq_type = VFIO_PCI_NUM_IRQS;
 	vdev->num_ctx = 0;
 	kfree(vdev->ctx);
@@ -284,7 +320,8 @@ static int vfio_msi_enable(struct vfio_pci_core_device *vdev, int nvec, bool msi
 	if (!is_irq_none(vdev))
 		return -EINVAL;
 
-	vdev->ctx = kcalloc(nvec, sizeof(struct vfio_pci_irq_ctx), GFP_KERNEL);
+	vdev->ctx = kcalloc(nvec, sizeof(struct vfio_pci_irq_ctx),
+			    GFP_KERNEL_ACCOUNT);
 	if (!vdev->ctx)
 		return -ENOMEM;
 
@@ -316,14 +353,14 @@ static int vfio_msi_enable(struct vfio_pci_core_device *vdev, int nvec, bool msi
 }
 
 static int vfio_msi_set_vector_signal(struct vfio_pci_core_device *vdev,
-				      int vector, int fd, bool msix)
+				      unsigned int vector, int fd, bool msix)
 {
 	struct pci_dev *pdev = vdev->pdev;
 	struct eventfd_ctx *trigger;
 	int irq, ret;
 	u16 cmd;
 
-	if (vector < 0 || vector >= vdev->num_ctx)
+	if (vector >= vdev->num_ctx)
 		return -EINVAL;
 
 	irq = pci_irq_vector(pdev, vector);
@@ -343,7 +380,8 @@ static int vfio_msi_set_vector_signal(struct vfio_pci_core_device *vdev,
 	if (fd < 0)
 		return 0;
 
-	vdev->ctx[vector].name = kasprintf(GFP_KERNEL, "vfio-msi%s[%d](%s)",
+	vdev->ctx[vector].name = kasprintf(GFP_KERNEL_ACCOUNT,
+					   "vfio-msi%s[%d](%s)",
 					   msix ? "x" : "", vector,
 					   pci_name(pdev));
 	if (!vdev->ctx[vector].name)
@@ -397,7 +435,8 @@ static int vfio_msi_set_vector_signal(struct vfio_pci_core_device *vdev,
 static int vfio_msi_set_block(struct vfio_pci_core_device *vdev, unsigned start,
 			      unsigned count, int32_t *fds, bool msix)
 {
-	int i, j, ret = 0;
+	unsigned int i, j;
+	int ret = 0;
 
 	if (start >= vdev->num_ctx || start + count > vdev->num_ctx)
 		return -EINVAL;
@@ -408,8 +447,8 @@ static int vfio_msi_set_block(struct vfio_pci_core_device *vdev, unsigned start,
 	}
 
 	if (ret) {
-		for (--j; j >= (int)start; j--)
-			vfio_msi_set_vector_signal(vdev, j, -1, msix);
+		for (i = start; i < j; i++)
+			vfio_msi_set_vector_signal(vdev, i, -1, msix);
 	}
 
 	return ret;
@@ -418,16 +457,15 @@ static int vfio_msi_set_block(struct vfio_pci_core_device *vdev, unsigned start,
 static void vfio_msi_disable(struct vfio_pci_core_device *vdev, bool msix)
 {
 	struct pci_dev *pdev = vdev->pdev;
-	int i;
+	unsigned int i;
 	u16 cmd;
 
 	for (i = 0; i < vdev->num_ctx; i++) {
 		vfio_virqfd_disable(&vdev->ctx[i].unmask);
 		vfio_virqfd_disable(&vdev->ctx[i].mask);
+		vfio_msi_set_vector_signal(vdev, i, -1, msix);
 	}
 
-	vfio_msi_set_block(vdev, 0, vdev->num_ctx, NULL, msix);
-
 	cmd = vfio_pci_memory_lock_and_enable(vdev);
 	pci_free_irq_vectors(pdev);
 	vfio_pci_memory_unlock_and_restore(vdev, cmd);
@@ -455,11 +493,11 @@ static int vfio_pci_set_intx_unmask(struct vfio_pci_core_device *vdev,
 		return -EINVAL;
 
 	if (flags & VFIO_IRQ_SET_DATA_NONE) {
-		vfio_pci_intx_unmask(vdev);
+		__vfio_pci_intx_unmask(vdev);
 	} else if (flags & VFIO_IRQ_SET_DATA_BOOL) {
 		uint8_t unmask = *(uint8_t *)data;
 		if (unmask)
-			vfio_pci_intx_unmask(vdev);
+			__vfio_pci_intx_unmask(vdev);
 	} else if (flags & VFIO_IRQ_SET_DATA_EVENTFD) {
 		int32_t fd = *(int32_t *)data;
 		if (fd >= 0)
@@ -482,11 +520,11 @@ static int vfio_pci_set_intx_mask(struct vfio_pci_core_device *vdev,
 		return -EINVAL;
 
 	if (flags & VFIO_IRQ_SET_DATA_NONE) {
-		vfio_pci_intx_mask(vdev);
+		__vfio_pci_intx_mask(vdev);
 	} else if (flags & VFIO_IRQ_SET_DATA_BOOL) {
 		uint8_t mask = *(uint8_t *)data;
 		if (mask)
-			vfio_pci_intx_mask(vdev);
+			__vfio_pci_intx_mask(vdev);
 	} else if (flags & VFIO_IRQ_SET_DATA_EVENTFD) {
 		return -ENOTTY; /* XXX implement me */
 	}
@@ -507,19 +545,23 @@ static int vfio_pci_set_intx_trigger(struct vfio_pci_core_device *vdev,
 		return -EINVAL;
 
 	if (flags & VFIO_IRQ_SET_DATA_EVENTFD) {
+		struct eventfd_ctx *trigger = NULL;
 		int32_t fd = *(int32_t *)data;
 		int ret;
 
-		if (is_intx(vdev))
-			return vfio_intx_set_signal(vdev, fd);
+		if (fd >= 0) {
+			trigger = eventfd_ctx_fdget(fd);
+			if (IS_ERR(trigger))
+				return PTR_ERR(trigger);
+		}
 
-		ret = vfio_intx_enable(vdev);
-		if (ret)
-			return ret;
+		if (is_intx(vdev))
+			ret = vfio_intx_set_signal(vdev, trigger);
+		else
+			ret = vfio_intx_enable(vdev, trigger);
 
-		ret = vfio_intx_set_signal(vdev, fd);
-		if (ret)
-			vfio_intx_disable(vdev);
+		if (ret && trigger)
+			eventfd_ctx_put(trigger);
 
 		return ret;
 	}
@@ -541,7 +583,7 @@ static int vfio_pci_set_msi_trigger(struct vfio_pci_core_device *vdev,
 				    unsigned index, unsigned start,
 				    unsigned count, uint32_t flags, void *data)
 {
-	int i;
+	unsigned int i;
 	bool msix = (index == VFIO_PCI_MSIX_IRQ_INDEX) ? true : false;
 
 	if (irq_is(vdev, index) && !count && (flags & VFIO_IRQ_SET_DATA_NONE)) {
diff --git a/drivers/vfio/pci/vfio_pci_rdwr.c b/drivers/vfio/pci/vfio_pci_rdwr.c
index e352a033b..e27de61ac 100644
--- a/drivers/vfio/pci/vfio_pci_rdwr.c
+++ b/drivers/vfio/pci/vfio_pci_rdwr.c
@@ -470,7 +470,7 @@ int vfio_pci_ioeventfd(struct vfio_pci_core_device *vdev, loff_t offset,
 		goto out_unlock;
 	}
 
-	ioeventfd = kzalloc(sizeof(*ioeventfd), GFP_KERNEL);
+	ioeventfd = kzalloc(sizeof(*ioeventfd), GFP_KERNEL_ACCOUNT);
 	if (!ioeventfd) {
 		ret = -ENOMEM;
 		goto out_unlock;
diff --git a/drivers/vfio/platform/vfio_platform_irq.c b/drivers/vfio/platform/vfio_platform_irq.c
index c5b09ec0a..7f4341a8d 100644
--- a/drivers/vfio/platform/vfio_platform_irq.c
+++ b/drivers/vfio/platform/vfio_platform_irq.c
@@ -136,6 +136,16 @@ static int vfio_platform_set_irq_unmask(struct vfio_platform_device *vdev,
 	return 0;
 }
 
+/*
+ * The trigger eventfd is guaranteed valid in the interrupt path
+ * and protected by the igate mutex when triggered via ioctl.
+ */
+static void vfio_send_eventfd(struct vfio_platform_irq *irq_ctx)
+{
+	if (likely(irq_ctx->trigger))
+		eventfd_signal(irq_ctx->trigger, 1);
+}
+
 static irqreturn_t vfio_automasked_irq_handler(int irq, void *dev_id)
 {
 	struct vfio_platform_irq *irq_ctx = dev_id;
@@ -155,7 +165,7 @@ static irqreturn_t vfio_automasked_irq_handler(int irq, void *dev_id)
 	spin_unlock_irqrestore(&irq_ctx->lock, flags);
 
 	if (ret == IRQ_HANDLED)
-		eventfd_signal(irq_ctx->trigger, 1);
+		vfio_send_eventfd(irq_ctx);
 
 	return ret;
 }
@@ -164,22 +174,19 @@ static irqreturn_t vfio_irq_handler(int irq, void *dev_id)
 {
 	struct vfio_platform_irq *irq_ctx = dev_id;
 
-	eventfd_signal(irq_ctx->trigger, 1);
+	vfio_send_eventfd(irq_ctx);
 
 	return IRQ_HANDLED;
 }
 
 static int vfio_set_trigger(struct vfio_platform_device *vdev, int index,
-			    int fd, irq_handler_t handler)
+			    int fd)
 {
 	struct vfio_platform_irq *irq = &vdev->irqs[index];
 	struct eventfd_ctx *trigger;
-	int ret;
 
 	if (irq->trigger) {
-		irq_clear_status_flags(irq->hwirq, IRQ_NOAUTOEN);
-		free_irq(irq->hwirq, irq);
-		kfree(irq->name);
+		disable_irq(irq->hwirq);
 		eventfd_ctx_put(irq->trigger);
 		irq->trigger = NULL;
 	}
@@ -187,30 +194,20 @@ static int vfio_set_trigger(struct vfio_platform_device *vdev, int index,
 	if (fd < 0) /* Disable only */
 		return 0;
 
-	irq->name = kasprintf(GFP_KERNEL, "vfio-irq[%d](%s)",
-						irq->hwirq, vdev->name);
-	if (!irq->name)
-		return -ENOMEM;
-
 	trigger = eventfd_ctx_fdget(fd);
-	if (IS_ERR(trigger)) {
-		kfree(irq->name);
+	if (IS_ERR(trigger))
 		return PTR_ERR(trigger);
-	}
 
 	irq->trigger = trigger;
 
-	irq_set_status_flags(irq->hwirq, IRQ_NOAUTOEN);
-	ret = request_irq(irq->hwirq, handler, 0, irq->name, irq);
-	if (ret) {
-		kfree(irq->name);
-		eventfd_ctx_put(trigger);
-		irq->trigger = NULL;
-		return ret;
-	}
-
-	if (!irq->masked)
-		enable_irq(irq->hwirq);
+	/*
+	 * irq->masked effectively provides nested disables within the overall
+	 * enable relative to trigger.  Specifically request_irq() is called
+	 * with NO_AUTOEN, therefore the IRQ is initially disabled.  The user
+	 * may only further disable the IRQ with a MASK operations because
+	 * irq->masked is initially false.
+	 */
+	enable_irq(irq->hwirq);
 
 	return 0;
 }
@@ -229,7 +226,7 @@ static int vfio_platform_set_irq_trigger(struct vfio_platform_device *vdev,
 		handler = vfio_irq_handler;
 
 	if (!count && (flags & VFIO_IRQ_SET_DATA_NONE))
-		return vfio_set_trigger(vdev, index, -1, handler);
+		return vfio_set_trigger(vdev, index, -1);
 
 	if (start != 0 || count != 1)
 		return -EINVAL;
@@ -237,7 +234,7 @@ static int vfio_platform_set_irq_trigger(struct vfio_platform_device *vdev,
 	if (flags & VFIO_IRQ_SET_DATA_EVENTFD) {
 		int32_t fd = *(int32_t *)data;
 
-		return vfio_set_trigger(vdev, index, fd, handler);
+		return vfio_set_trigger(vdev, index, fd);
 	}
 
 	if (flags & VFIO_IRQ_SET_DATA_NONE) {
@@ -261,6 +258,14 @@ int vfio_platform_set_irqs_ioctl(struct vfio_platform_device *vdev,
 		    unsigned start, unsigned count, uint32_t flags,
 		    void *data) = NULL;
 
+	/*
+	 * For compatibility, errors from request_irq() are local to the
+	 * SET_IRQS path and reflected in the name pointer.  This allows,
+	 * for example, polling mode fallback for an exclusive IRQ failure.
+	 */
+	if (IS_ERR(vdev->irqs[index].name))
+		return PTR_ERR(vdev->irqs[index].name);
+
 	switch (flags & VFIO_IRQ_SET_ACTION_TYPE_MASK) {
 	case VFIO_IRQ_SET_ACTION_MASK:
 		func = vfio_platform_set_irq_mask;
@@ -281,7 +286,7 @@ int vfio_platform_set_irqs_ioctl(struct vfio_platform_device *vdev,
 
 int vfio_platform_irq_init(struct vfio_platform_device *vdev)
 {
-	int cnt = 0, i;
+	int cnt = 0, i, ret = 0;
 
 	while (vdev->get_irq(vdev, cnt) >= 0)
 		cnt++;
@@ -292,37 +297,70 @@ int vfio_platform_irq_init(struct vfio_platform_device *vdev)
 
 	for (i = 0; i < cnt; i++) {
 		int hwirq = vdev->get_irq(vdev, i);
+		irq_handler_t handler = vfio_irq_handler;
 
-		if (hwirq < 0)
+		if (hwirq < 0) {
+			ret = -EINVAL;
 			goto err;
+		}
 
 		spin_lock_init(&vdev->irqs[i].lock);
 
 		vdev->irqs[i].flags = VFIO_IRQ_INFO_EVENTFD;
 
-		if (irq_get_trigger_type(hwirq) & IRQ_TYPE_LEVEL_MASK)
+		if (irq_get_trigger_type(hwirq) & IRQ_TYPE_LEVEL_MASK) {
 			vdev->irqs[i].flags |= VFIO_IRQ_INFO_MASKABLE
 						| VFIO_IRQ_INFO_AUTOMASKED;
+			handler = vfio_automasked_irq_handler;
+		}
 
 		vdev->irqs[i].count = 1;
 		vdev->irqs[i].hwirq = hwirq;
 		vdev->irqs[i].masked = false;
+		vdev->irqs[i].name = kasprintf(GFP_KERNEL,
+					       "vfio-irq[%d](%s)", hwirq,
+					       vdev->name);
+		if (!vdev->irqs[i].name) {
+			ret = -ENOMEM;
+			goto err;
+		}
+
+		ret = request_irq(hwirq, handler, IRQF_NO_AUTOEN,
+				  vdev->irqs[i].name, &vdev->irqs[i]);
+		if (ret) {
+			kfree(vdev->irqs[i].name);
+			vdev->irqs[i].name = ERR_PTR(ret);
+		}
 	}
 
 	vdev->num_irqs = cnt;
 
 	return 0;
 err:
+	for (--i; i >= 0; i--) {
+		if (!IS_ERR(vdev->irqs[i].name)) {
+			free_irq(vdev->irqs[i].hwirq, &vdev->irqs[i]);
+			kfree(vdev->irqs[i].name);
+		}
+	}
 	kfree(vdev->irqs);
-	return -EINVAL;
+	return ret;
 }
 
 void vfio_platform_irq_cleanup(struct vfio_platform_device *vdev)
 {
 	int i;
 
-	for (i = 0; i < vdev->num_irqs; i++)
-		vfio_set_trigger(vdev, i, -1, NULL);
+	for (i = 0; i < vdev->num_irqs; i++) {
+		vfio_virqfd_disable(&vdev->irqs[i].mask);
+		vfio_virqfd_disable(&vdev->irqs[i].unmask);
+		if (!IS_ERR(vdev->irqs[i].name)) {
+			free_irq(vdev->irqs[i].hwirq, &vdev->irqs[i]);
+			if (vdev->irqs[i].trigger)
+				eventfd_ctx_put(vdev->irqs[i].trigger);
+			kfree(vdev->irqs[i].name);
+		}
+	}
 
 	vdev->num_irqs = 0;
 	kfree(vdev->irqs);
diff --git a/drivers/vfio/virqfd.c b/drivers/vfio/virqfd.c
index 414e98d82..e06b32dde 100644
--- a/drivers/vfio/virqfd.c
+++ b/drivers/vfio/virqfd.c
@@ -104,6 +104,13 @@ static void virqfd_inject(struct work_struct *work)
 		virqfd->thread(virqfd->opaque, virqfd->data);
 }
 
+static void virqfd_flush_inject(struct work_struct *work)
+{
+	struct virqfd *virqfd = container_of(work, struct virqfd, flush_inject);
+
+	flush_work(&virqfd->inject);
+}
+
 int vfio_virqfd_enable(void *opaque,
 		       int (*handler)(void *, void *),
 		       void (*thread)(void *, void *),
@@ -115,7 +122,7 @@ int vfio_virqfd_enable(void *opaque,
 	int ret = 0;
 	__poll_t events;
 
-	virqfd = kzalloc(sizeof(*virqfd), GFP_KERNEL);
+	virqfd = kzalloc(sizeof(*virqfd), GFP_KERNEL_ACCOUNT);
 	if (!virqfd)
 		return -ENOMEM;
 
@@ -127,6 +134,7 @@ int vfio_virqfd_enable(void *opaque,
 
 	INIT_WORK(&virqfd->shutdown, virqfd_shutdown);
 	INIT_WORK(&virqfd->inject, virqfd_inject);
+	INIT_WORK(&virqfd->flush_inject, virqfd_flush_inject);
 
 	irqfd = fdget(fd);
 	if (!irqfd.file) {
@@ -217,6 +225,19 @@ void vfio_virqfd_disable(struct virqfd **pvirqfd)
 }
 EXPORT_SYMBOL_GPL(vfio_virqfd_disable);
 
+void vfio_virqfd_flush_thread(struct virqfd **pvirqfd)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&virqfd_lock, flags);
+	if (*pvirqfd && (*pvirqfd)->thread)
+		queue_work(vfio_irqfd_cleanup_wq, &(*pvirqfd)->flush_inject);
+	spin_unlock_irqrestore(&virqfd_lock, flags);
+
+	flush_workqueue(vfio_irqfd_cleanup_wq);
+}
+EXPORT_SYMBOL_GPL(vfio_virqfd_flush_thread);
+
 module_init(vfio_virqfd_init);
 module_exit(vfio_virqfd_exit);