1 files changed, 928 insertions, 499 deletions

diff --git a/drivers/infiniband/core/uverbs_main.c b/drivers/infiniband/core/uverbs_main.c
index 2c6f0f2ecd9d..973fe2c7ef53 100644
--- a/drivers/infiniband/core/uverbs_main.c
+++ b/drivers/infiniband/core/uverbs_main.c
@@ -45,10 +45,18 @@
 #include <linux/cdev.h>
 #include <linux/anon_inodes.h>
 #include <linux/slab.h>
+#include <linux/sched/mm.h>
 
-#include <asm/uaccess.h>
+#include <linux/uaccess.h>
+
+#include <rdma/ib.h>
+#include <rdma/uverbs_std_types.h>
+#include <rdma/rdma_netlink.h>
+#include <rdma/ib_ucaps.h>
 
 #include "uverbs.h"
+#include "core_priv.h"
+#include "rdma_core.h"
 
 MODULE_AUTHOR("Roland Dreier");
 MODULE_DESCRIPTION("InfiniBand userspace verbs access");
@@ -57,124 +65,120 @@ MODULE_LICENSE("Dual BSD/GPL");
 enum {
 	IB_UVERBS_MAJOR       = 231,
 	IB_UVERBS_BASE_MINOR  = 192,
-	IB_UVERBS_MAX_DEVICES = 32
+	IB_UVERBS_MAX_DEVICES = RDMA_MAX_PORTS,
+	IB_UVERBS_NUM_FIXED_MINOR = 32,
+	IB_UVERBS_NUM_DYNAMIC_MINOR = IB_UVERBS_MAX_DEVICES - IB_UVERBS_NUM_FIXED_MINOR,
 };
 
 #define IB_UVERBS_BASE_DEV	MKDEV(IB_UVERBS_MAJOR, IB_UVERBS_BASE_MINOR)
 
-static struct class *uverbs_class;
-
-DEFINE_SPINLOCK(ib_uverbs_idr_lock);
-DEFINE_IDR(ib_uverbs_pd_idr);
-DEFINE_IDR(ib_uverbs_mr_idr);
-DEFINE_IDR(ib_uverbs_mw_idr);
-DEFINE_IDR(ib_uverbs_ah_idr);
-DEFINE_IDR(ib_uverbs_cq_idr);
-DEFINE_IDR(ib_uverbs_qp_idr);
-DEFINE_IDR(ib_uverbs_srq_idr);
-DEFINE_IDR(ib_uverbs_xrcd_idr);
-
-static DEFINE_SPINLOCK(map_lock);
-static DECLARE_BITMAP(dev_map, IB_UVERBS_MAX_DEVICES);
-
-static ssize_t (*uverbs_cmd_table[])(struct ib_uverbs_file *file,
-				     const char __user *buf, int in_len,
-				     int out_len) = {
-	[IB_USER_VERBS_CMD_GET_CONTEXT]		= ib_uverbs_get_context,
-	[IB_USER_VERBS_CMD_QUERY_DEVICE]	= ib_uverbs_query_device,
-	[IB_USER_VERBS_CMD_QUERY_PORT]		= ib_uverbs_query_port,
-	[IB_USER_VERBS_CMD_ALLOC_PD]		= ib_uverbs_alloc_pd,
-	[IB_USER_VERBS_CMD_DEALLOC_PD]		= ib_uverbs_dealloc_pd,
-	[IB_USER_VERBS_CMD_REG_MR]		= ib_uverbs_reg_mr,
-	[IB_USER_VERBS_CMD_DEREG_MR]		= ib_uverbs_dereg_mr,
-	[IB_USER_VERBS_CMD_ALLOC_MW]		= ib_uverbs_alloc_mw,
-	[IB_USER_VERBS_CMD_DEALLOC_MW]		= ib_uverbs_dealloc_mw,
-	[IB_USER_VERBS_CMD_CREATE_COMP_CHANNEL] = ib_uverbs_create_comp_channel,
-	[IB_USER_VERBS_CMD_CREATE_CQ]		= ib_uverbs_create_cq,
-	[IB_USER_VERBS_CMD_RESIZE_CQ]		= ib_uverbs_resize_cq,
-	[IB_USER_VERBS_CMD_POLL_CQ]		= ib_uverbs_poll_cq,
-	[IB_USER_VERBS_CMD_REQ_NOTIFY_CQ]	= ib_uverbs_req_notify_cq,
-	[IB_USER_VERBS_CMD_DESTROY_CQ]		= ib_uverbs_destroy_cq,
-	[IB_USER_VERBS_CMD_CREATE_QP]		= ib_uverbs_create_qp,
-	[IB_USER_VERBS_CMD_QUERY_QP]		= ib_uverbs_query_qp,
-	[IB_USER_VERBS_CMD_MODIFY_QP]		= ib_uverbs_modify_qp,
-	[IB_USER_VERBS_CMD_DESTROY_QP]		= ib_uverbs_destroy_qp,
-	[IB_USER_VERBS_CMD_POST_SEND]		= ib_uverbs_post_send,
-	[IB_USER_VERBS_CMD_POST_RECV]		= ib_uverbs_post_recv,
-	[IB_USER_VERBS_CMD_POST_SRQ_RECV]	= ib_uverbs_post_srq_recv,
-	[IB_USER_VERBS_CMD_CREATE_AH]		= ib_uverbs_create_ah,
-	[IB_USER_VERBS_CMD_DESTROY_AH]		= ib_uverbs_destroy_ah,
-	[IB_USER_VERBS_CMD_ATTACH_MCAST]	= ib_uverbs_attach_mcast,
-	[IB_USER_VERBS_CMD_DETACH_MCAST]	= ib_uverbs_detach_mcast,
-	[IB_USER_VERBS_CMD_CREATE_SRQ]		= ib_uverbs_create_srq,
-	[IB_USER_VERBS_CMD_MODIFY_SRQ]		= ib_uverbs_modify_srq,
-	[IB_USER_VERBS_CMD_QUERY_SRQ]		= ib_uverbs_query_srq,
-	[IB_USER_VERBS_CMD_DESTROY_SRQ]		= ib_uverbs_destroy_srq,
-	[IB_USER_VERBS_CMD_OPEN_XRCD]		= ib_uverbs_open_xrcd,
-	[IB_USER_VERBS_CMD_CLOSE_XRCD]		= ib_uverbs_close_xrcd,
-	[IB_USER_VERBS_CMD_CREATE_XSRQ]		= ib_uverbs_create_xsrq,
-	[IB_USER_VERBS_CMD_OPEN_QP]		= ib_uverbs_open_qp
+static dev_t dynamic_uverbs_dev;
+
+static DEFINE_IDA(uverbs_ida);
+static int ib_uverbs_add_one(struct ib_device *device);
+static void ib_uverbs_remove_one(struct ib_device *device, void *client_data);
+static struct ib_client uverbs_client;
+
+static char *uverbs_devnode(const struct device *dev, umode_t *mode)
+{
+	if (mode)
+		*mode = 0666;
+	return kasprintf(GFP_KERNEL, "infiniband/%s", dev_name(dev));
+}
+
+static const struct class uverbs_class = {
+	.name = "infiniband_verbs",
+	.devnode = uverbs_devnode,
 };
 
-static void ib_uverbs_add_one(struct ib_device *device);
-static void ib_uverbs_remove_one(struct ib_device *device);
+/*
+ * Must be called with the ufile->device->disassociate_srcu held, and the lock
+ * must be held until use of the ucontext is finished.
+ */
+struct ib_ucontext *ib_uverbs_get_ucontext_file(struct ib_uverbs_file *ufile)
+{
+	/*
+	 * We do not hold the hw_destroy_rwsem lock for this flow, instead
+	 * srcu is used. It does not matter if someone races this with
+	 * get_context, we get NULL or valid ucontext.
+	 */
+	struct ib_ucontext *ucontext = smp_load_acquire(&ufile->ucontext);
+
+	if (!srcu_dereference(ufile->device->ib_dev,
+			      &ufile->device->disassociate_srcu))
+		return ERR_PTR(-EIO);
+
+	if (!ucontext)
+		return ERR_PTR(-EINVAL);
 
-static void ib_uverbs_release_dev(struct kref *ref)
+	return ucontext;
+}
+EXPORT_SYMBOL(ib_uverbs_get_ucontext_file);
+
+int uverbs_dealloc_mw(struct ib_mw *mw)
 {
-	struct ib_uverbs_device *dev =
-		container_of(ref, struct ib_uverbs_device, ref);
+	struct ib_pd *pd = mw->pd;
+	int ret;
 
-	complete(&dev->comp);
+	ret = mw->device->ops.dealloc_mw(mw);
+	if (ret)
+		return ret;
+
+	atomic_dec(&pd->usecnt);
+	kfree(mw);
+	return ret;
 }
 
-static void ib_uverbs_release_event_file(struct kref *ref)
+static void ib_uverbs_release_dev(struct device *device)
 {
-	struct ib_uverbs_event_file *file =
-		container_of(ref, struct ib_uverbs_event_file, ref);
+	struct ib_uverbs_device *dev =
+			container_of(device, struct ib_uverbs_device, dev);
 
-	kfree(file);
+	uverbs_destroy_api(dev->uapi);
+	cleanup_srcu_struct(&dev->disassociate_srcu);
+	mutex_destroy(&dev->lists_mutex);
+	mutex_destroy(&dev->xrcd_tree_mutex);
+	kfree(dev);
 }
 
-void ib_uverbs_release_ucq(struct ib_uverbs_file *file,
-			  struct ib_uverbs_event_file *ev_file,
-			  struct ib_ucq_object *uobj)
+void ib_uverbs_release_ucq(struct ib_uverbs_completion_event_file *ev_file,
+			   struct ib_ucq_object *uobj)
 {
 	struct ib_uverbs_event *evt, *tmp;
 
 	if (ev_file) {
-		spin_lock_irq(&ev_file->lock);
+		spin_lock_irq(&ev_file->ev_queue.lock);
 		list_for_each_entry_safe(evt, tmp, &uobj->comp_list, obj_list) {
 			list_del(&evt->list);
 			kfree(evt);
 		}
-		spin_unlock_irq(&ev_file->lock);
+		spin_unlock_irq(&ev_file->ev_queue.lock);
 
-		kref_put(&ev_file->ref, ib_uverbs_release_event_file);
+		uverbs_uobject_put(&ev_file->uobj);
 	}
 
-	spin_lock_irq(&file->async_file->lock);
-	list_for_each_entry_safe(evt, tmp, &uobj->async_list, obj_list) {
-		list_del(&evt->list);
-		kfree(evt);
-	}
-	spin_unlock_irq(&file->async_file->lock);
+	ib_uverbs_release_uevent(&uobj->uevent);
 }
 
-void ib_uverbs_release_uevent(struct ib_uverbs_file *file,
-			      struct ib_uevent_object *uobj)
+void ib_uverbs_release_uevent(struct ib_uevent_object *uobj)
 {
+	struct ib_uverbs_async_event_file *async_file = uobj->event_file;
 	struct ib_uverbs_event *evt, *tmp;
 
-	spin_lock_irq(&file->async_file->lock);
+	if (!async_file)
+		return;
+
+	spin_lock_irq(&async_file->ev_queue.lock);
 	list_for_each_entry_safe(evt, tmp, &uobj->event_list, obj_list) {
 		list_del(&evt->list);
 		kfree(evt);
 	}
-	spin_unlock_irq(&file->async_file->lock);
+	spin_unlock_irq(&async_file->ev_queue.lock);
+	uverbs_uobject_put(&async_file->uobj);
 }
 
-static void ib_uverbs_detach_umcast(struct ib_qp *qp,
-				    struct ib_uqp_object *uobj)
+void ib_uverbs_detach_umcast(struct ib_qp *qp,
+			     struct ib_uqp_object *uobj)
 {
 	struct ib_uverbs_mcast_entry *mcast, *tmp;
 
@@ -185,156 +189,84 @@ static void ib_uverbs_detach_umcast(struct ib_qp *qp,
 	}
 }
 
-static int ib_uverbs_cleanup_ucontext(struct ib_uverbs_file *file,
-				      struct ib_ucontext *context)
+static void ib_uverbs_comp_dev(struct ib_uverbs_device *dev)
 {
-	struct ib_uobject *uobj, *tmp;
-
-	if (!context)
-		return 0;
-
-	context->closing = 1;
-
-	list_for_each_entry_safe(uobj, tmp, &context->ah_list, list) {
-		struct ib_ah *ah = uobj->object;
-
-		idr_remove_uobj(&ib_uverbs_ah_idr, uobj);
-		ib_destroy_ah(ah);
-		kfree(uobj);
-	}
-
-	/* Remove MWs before QPs, in order to support type 2A MWs. */
-	list_for_each_entry_safe(uobj, tmp, &context->mw_list, list) {
-		struct ib_mw *mw = uobj->object;
-
-		idr_remove_uobj(&ib_uverbs_mw_idr, uobj);
-		ib_dealloc_mw(mw);
-		kfree(uobj);
-	}
-
-	list_for_each_entry_safe(uobj, tmp, &context->qp_list, list) {
-		struct ib_qp *qp = uobj->object;
-		struct ib_uqp_object *uqp =
-			container_of(uobj, struct ib_uqp_object, uevent.uobject);
-
-		idr_remove_uobj(&ib_uverbs_qp_idr, uobj);
-		if (qp != qp->real_qp) {
-			ib_close_qp(qp);
-		} else {
-			ib_uverbs_detach_umcast(qp, uqp);
-			ib_destroy_qp(qp);
-		}
-		ib_uverbs_release_uevent(file, &uqp->uevent);
-		kfree(uqp);
-	}
-
-	list_for_each_entry_safe(uobj, tmp, &context->cq_list, list) {
-		struct ib_cq *cq = uobj->object;
-		struct ib_uverbs_event_file *ev_file = cq->cq_context;
-		struct ib_ucq_object *ucq =
-			container_of(uobj, struct ib_ucq_object, uobject);
-
-		idr_remove_uobj(&ib_uverbs_cq_idr, uobj);
-		ib_destroy_cq(cq);
-		ib_uverbs_release_ucq(file, ev_file, ucq);
-		kfree(ucq);
-	}
-
-	list_for_each_entry_safe(uobj, tmp, &context->srq_list, list) {
-		struct ib_srq *srq = uobj->object;
-		struct ib_uevent_object *uevent =
-			container_of(uobj, struct ib_uevent_object, uobject);
-
-		idr_remove_uobj(&ib_uverbs_srq_idr, uobj);
-		ib_destroy_srq(srq);
-		ib_uverbs_release_uevent(file, uevent);
-		kfree(uevent);
-	}
-
-	list_for_each_entry_safe(uobj, tmp, &context->mr_list, list) {
-		struct ib_mr *mr = uobj->object;
-
-		idr_remove_uobj(&ib_uverbs_mr_idr, uobj);
-		ib_dereg_mr(mr);
-		kfree(uobj);
-	}
-
-	mutex_lock(&file->device->xrcd_tree_mutex);
-	list_for_each_entry_safe(uobj, tmp, &context->xrcd_list, list) {
-		struct ib_xrcd *xrcd = uobj->object;
-		struct ib_uxrcd_object *uxrcd =
-			container_of(uobj, struct ib_uxrcd_object, uobject);
-
-		idr_remove_uobj(&ib_uverbs_xrcd_idr, uobj);
-		ib_uverbs_dealloc_xrcd(file->device, xrcd);
-		kfree(uxrcd);
-	}
-	mutex_unlock(&file->device->xrcd_tree_mutex);
-
-	list_for_each_entry_safe(uobj, tmp, &context->pd_list, list) {
-		struct ib_pd *pd = uobj->object;
-
-		idr_remove_uobj(&ib_uverbs_pd_idr, uobj);
-		ib_dealloc_pd(pd);
-		kfree(uobj);
-	}
-
-	return context->device->dealloc_ucontext(context);
+	complete(&dev->comp);
 }
 
-static void ib_uverbs_release_file(struct kref *ref)
+void ib_uverbs_release_file(struct kref *ref)
 {
 	struct ib_uverbs_file *file =
 		container_of(ref, struct ib_uverbs_file, ref);
-
-	module_put(file->device->ib_dev->owner);
-	kref_put(&file->device->ref, ib_uverbs_release_dev);
-
+	struct ib_device *ib_dev;
+	int srcu_key;
+
+	release_ufile_idr_uobject(file);
+
+	srcu_key = srcu_read_lock(&file->device->disassociate_srcu);
+	ib_dev = srcu_dereference(file->device->ib_dev,
+				  &file->device->disassociate_srcu);
+	if (ib_dev && !ib_dev->ops.disassociate_ucontext)
+		module_put(ib_dev->ops.owner);
+	srcu_read_unlock(&file->device->disassociate_srcu, srcu_key);
+
+	if (refcount_dec_and_test(&file->device->refcount))
+		ib_uverbs_comp_dev(file->device);
+
+	if (file->default_async_file)
+		uverbs_uobject_put(&file->default_async_file->uobj);
+	put_device(&file->device->dev);
+
+	if (file->disassociate_page)
+		__free_pages(file->disassociate_page, 0);
+	mutex_destroy(&file->disassociation_lock);
+	mutex_destroy(&file->umap_lock);
+	mutex_destroy(&file->ucontext_lock);
 	kfree(file);
 }
 
-static ssize_t ib_uverbs_event_read(struct file *filp, char __user *buf,
-				    size_t count, loff_t *pos)
+static ssize_t ib_uverbs_event_read(struct ib_uverbs_event_queue *ev_queue,
+				    struct file *filp, char __user *buf,
+				    size_t count, loff_t *pos,
+				    size_t eventsz)
 {
-	struct ib_uverbs_event_file *file = filp->private_data;
 	struct ib_uverbs_event *event;
-	int eventsz;
 	int ret = 0;
 
-	spin_lock_irq(&file->lock);
+	spin_lock_irq(&ev_queue->lock);
 
-	while (list_empty(&file->event_list)) {
-		spin_unlock_irq(&file->lock);
+	while (list_empty(&ev_queue->event_list)) {
+		if (ev_queue->is_closed) {
+			spin_unlock_irq(&ev_queue->lock);
+			return -EIO;
+		}
 
+		spin_unlock_irq(&ev_queue->lock);
 		if (filp->f_flags & O_NONBLOCK)
 			return -EAGAIN;
 
-		if (wait_event_interruptible(file->poll_wait,
-					     !list_empty(&file->event_list)))
+		if (wait_event_interruptible(ev_queue->poll_wait,
+					     (!list_empty(&ev_queue->event_list) ||
+					      ev_queue->is_closed)))
 			return -ERESTARTSYS;
 
-		spin_lock_irq(&file->lock);
+		spin_lock_irq(&ev_queue->lock);
 	}
 
-	event = list_entry(file->event_list.next, struct ib_uverbs_event, list);
-
-	if (file->is_async)
-		eventsz = sizeof (struct ib_uverbs_async_event_desc);
-	else
-		eventsz = sizeof (struct ib_uverbs_comp_event_desc);
+	event = list_entry(ev_queue->event_list.next, struct ib_uverbs_event, list);
 
 	if (eventsz > count) {
 		ret   = -EINVAL;
 		event = NULL;
 	} else {
-		list_del(file->event_list.next);
+		list_del(ev_queue->event_list.next);
 		if (event->counter) {
 			++(*event->counter);
 			list_del(&event->obj_list);
 		}
 	}
 
-	spin_unlock_irq(&file->lock);
+	spin_unlock_irq(&ev_queue->lock);
 
 	if (event) {
 		if (copy_to_user(buf, event, eventsz))
@@ -348,269 +280,642 @@ static ssize_t ib_uverbs_event_read(struct file *filp, char __user *buf,
 	return ret;
 }
 
-static unsigned int ib_uverbs_event_poll(struct file *filp,
+static ssize_t ib_uverbs_async_event_read(struct file *filp, char __user *buf,
+					  size_t count, loff_t *pos)
+{
+	struct ib_uverbs_async_event_file *file = filp->private_data;
+
+	return ib_uverbs_event_read(&file->ev_queue, filp, buf, count, pos,
+				    sizeof(struct ib_uverbs_async_event_desc));
+}
+
+static ssize_t ib_uverbs_comp_event_read(struct file *filp, char __user *buf,
+					 size_t count, loff_t *pos)
+{
+	struct ib_uverbs_completion_event_file *comp_ev_file =
+		filp->private_data;
+
+	return ib_uverbs_event_read(&comp_ev_file->ev_queue, filp, buf, count,
+				    pos,
+				    sizeof(struct ib_uverbs_comp_event_desc));
+}
+
+static __poll_t ib_uverbs_event_poll(struct ib_uverbs_event_queue *ev_queue,
+					 struct file *filp,
 					 struct poll_table_struct *wait)
 {
-	unsigned int pollflags = 0;
-	struct ib_uverbs_event_file *file = filp->private_data;
+	__poll_t pollflags = 0;
 
-	poll_wait(filp, &file->poll_wait, wait);
+	poll_wait(filp, &ev_queue->poll_wait, wait);
 
-	spin_lock_irq(&file->lock);
-	if (!list_empty(&file->event_list))
-		pollflags = POLLIN | POLLRDNORM;
-	spin_unlock_irq(&file->lock);
+	spin_lock_irq(&ev_queue->lock);
+	if (!list_empty(&ev_queue->event_list))
+		pollflags = EPOLLIN | EPOLLRDNORM;
+	else if (ev_queue->is_closed)
+		pollflags = EPOLLERR;
+	spin_unlock_irq(&ev_queue->lock);
 
 	return pollflags;
 }
 
-static int ib_uverbs_event_fasync(int fd, struct file *filp, int on)
+static __poll_t ib_uverbs_async_event_poll(struct file *filp,
+					       struct poll_table_struct *wait)
 {
-	struct ib_uverbs_event_file *file = filp->private_data;
+	struct ib_uverbs_async_event_file *file = filp->private_data;
 
-	return fasync_helper(fd, filp, on, &file->async_queue);
+	return ib_uverbs_event_poll(&file->ev_queue, filp, wait);
 }
 
-static int ib_uverbs_event_close(struct inode *inode, struct file *filp)
+static __poll_t ib_uverbs_comp_event_poll(struct file *filp,
+					      struct poll_table_struct *wait)
 {
-	struct ib_uverbs_event_file *file = filp->private_data;
-	struct ib_uverbs_event *entry, *tmp;
+	struct ib_uverbs_completion_event_file *comp_ev_file =
+		filp->private_data;
 
-	spin_lock_irq(&file->lock);
-	file->is_closed = 1;
-	list_for_each_entry_safe(entry, tmp, &file->event_list, list) {
-		if (entry->counter)
-			list_del(&entry->obj_list);
-		kfree(entry);
-	}
-	spin_unlock_irq(&file->lock);
+	return ib_uverbs_event_poll(&comp_ev_file->ev_queue, filp, wait);
+}
 
-	if (file->is_async) {
-		ib_unregister_event_handler(&file->uverbs_file->event_handler);
-		kref_put(&file->uverbs_file->ref, ib_uverbs_release_file);
-	}
-	kref_put(&file->ref, ib_uverbs_release_event_file);
+static int ib_uverbs_async_event_fasync(int fd, struct file *filp, int on)
+{
+	struct ib_uverbs_async_event_file *file = filp->private_data;
 
-	return 0;
+	return fasync_helper(fd, filp, on, &file->ev_queue.async_queue);
+}
+
+static int ib_uverbs_comp_event_fasync(int fd, struct file *filp, int on)
+{
+	struct ib_uverbs_completion_event_file *comp_ev_file =
+		filp->private_data;
+
+	return fasync_helper(fd, filp, on, &comp_ev_file->ev_queue.async_queue);
 }
 
-static const struct file_operations uverbs_event_fops = {
+const struct file_operations uverbs_event_fops = {
+	.owner	 = THIS_MODULE,
+	.read	 = ib_uverbs_comp_event_read,
+	.poll    = ib_uverbs_comp_event_poll,
+	.release = uverbs_uobject_fd_release,
+	.fasync  = ib_uverbs_comp_event_fasync,
+};
+
+const struct file_operations uverbs_async_event_fops = {
 	.owner	 = THIS_MODULE,
-	.read	 = ib_uverbs_event_read,
-	.poll    = ib_uverbs_event_poll,
-	.release = ib_uverbs_event_close,
-	.fasync  = ib_uverbs_event_fasync,
-	.llseek	 = no_llseek,
+	.read	 = ib_uverbs_async_event_read,
+	.poll    = ib_uverbs_async_event_poll,
+	.release = uverbs_async_event_release,
+	.fasync  = ib_uverbs_async_event_fasync,
 };
 
 void ib_uverbs_comp_handler(struct ib_cq *cq, void *cq_context)
 {
-	struct ib_uverbs_event_file    *file = cq_context;
+	struct ib_uverbs_event_queue   *ev_queue = cq_context;
 	struct ib_ucq_object	       *uobj;
 	struct ib_uverbs_event	       *entry;
 	unsigned long			flags;
 
-	if (!file)
+	if (!ev_queue)
 		return;
 
-	spin_lock_irqsave(&file->lock, flags);
-	if (file->is_closed) {
-		spin_unlock_irqrestore(&file->lock, flags);
+	spin_lock_irqsave(&ev_queue->lock, flags);
+	if (ev_queue->is_closed) {
+		spin_unlock_irqrestore(&ev_queue->lock, flags);
 		return;
 	}
 
-	entry = kmalloc(sizeof *entry, GFP_ATOMIC);
+	entry = kmalloc(sizeof(*entry), GFP_ATOMIC);
 	if (!entry) {
-		spin_unlock_irqrestore(&file->lock, flags);
+		spin_unlock_irqrestore(&ev_queue->lock, flags);
 		return;
 	}
 
-	uobj = container_of(cq->uobject, struct ib_ucq_object, uobject);
+	uobj = cq->uobject;
 
-	entry->desc.comp.cq_handle = cq->uobject->user_handle;
+	entry->desc.comp.cq_handle = cq->uobject->uevent.uobject.user_handle;
 	entry->counter		   = &uobj->comp_events_reported;
 
-	list_add_tail(&entry->list, &file->event_list);
+	list_add_tail(&entry->list, &ev_queue->event_list);
 	list_add_tail(&entry->obj_list, &uobj->comp_list);
-	spin_unlock_irqrestore(&file->lock, flags);
+	spin_unlock_irqrestore(&ev_queue->lock, flags);
 
-	wake_up_interruptible(&file->poll_wait);
-	kill_fasync(&file->async_queue, SIGIO, POLL_IN);
+	wake_up_interruptible(&ev_queue->poll_wait);
+	kill_fasync(&ev_queue->async_queue, SIGIO, POLL_IN);
 }
 
-static void ib_uverbs_async_handler(struct ib_uverbs_file *file,
-				    __u64 element, __u64 event,
-				    struct list_head *obj_list,
-				    u32 *counter)
+void ib_uverbs_async_handler(struct ib_uverbs_async_event_file *async_file,
+			     __u64 element, __u64 event,
+			     struct list_head *obj_list, u32 *counter)
 {
 	struct ib_uverbs_event *entry;
 	unsigned long flags;
 
-	spin_lock_irqsave(&file->async_file->lock, flags);
-	if (file->async_file->is_closed) {
-		spin_unlock_irqrestore(&file->async_file->lock, flags);
+	if (!async_file)
+		return;
+
+	spin_lock_irqsave(&async_file->ev_queue.lock, flags);
+	if (async_file->ev_queue.is_closed) {
+		spin_unlock_irqrestore(&async_file->ev_queue.lock, flags);
 		return;
 	}
 
-	entry = kmalloc(sizeof *entry, GFP_ATOMIC);
+	entry = kmalloc(sizeof(*entry), GFP_ATOMIC);
 	if (!entry) {
-		spin_unlock_irqrestore(&file->async_file->lock, flags);
+		spin_unlock_irqrestore(&async_file->ev_queue.lock, flags);
 		return;
 	}
 
-	entry->desc.async.element    = element;
+	entry->desc.async.element = element;
 	entry->desc.async.event_type = event;
-	entry->counter               = counter;
+	entry->desc.async.reserved = 0;
+	entry->counter = counter;
 
-	list_add_tail(&entry->list, &file->async_file->event_list);
+	list_add_tail(&entry->list, &async_file->ev_queue.event_list);
 	if (obj_list)
 		list_add_tail(&entry->obj_list, obj_list);
-	spin_unlock_irqrestore(&file->async_file->lock, flags);
+	spin_unlock_irqrestore(&async_file->ev_queue.lock, flags);
 
-	wake_up_interruptible(&file->async_file->poll_wait);
-	kill_fasync(&file->async_file->async_queue, SIGIO, POLL_IN);
+	wake_up_interruptible(&async_file->ev_queue.poll_wait);
+	kill_fasync(&async_file->ev_queue.async_queue, SIGIO, POLL_IN);
 }
 
-void ib_uverbs_cq_event_handler(struct ib_event *event, void *context_ptr)
+static void uverbs_uobj_event(struct ib_uevent_object *eobj,
+			      struct ib_event *event)
 {
-	struct ib_ucq_object *uobj = container_of(event->element.cq->uobject,
-						  struct ib_ucq_object, uobject);
+	ib_uverbs_async_handler(eobj->event_file,
+				eobj->uobject.user_handle, event->event,
+				&eobj->event_list, &eobj->events_reported);
+}
 
-	ib_uverbs_async_handler(uobj->uverbs_file, uobj->uobject.user_handle,
-				event->event, &uobj->async_list,
-				&uobj->async_events_reported);
+void ib_uverbs_cq_event_handler(struct ib_event *event, void *context_ptr)
+{
+	uverbs_uobj_event(&event->element.cq->uobject->uevent, event);
 }
 
 void ib_uverbs_qp_event_handler(struct ib_event *event, void *context_ptr)
 {
-	struct ib_uevent_object *uobj;
+	/* for XRC target qp's, check that qp is live */
+	if (!event->element.qp->uobject)
+		return;
 
-	uobj = container_of(event->element.qp->uobject,
-			    struct ib_uevent_object, uobject);
+	uverbs_uobj_event(&event->element.qp->uobject->uevent, event);
+}
 
-	ib_uverbs_async_handler(context_ptr, uobj->uobject.user_handle,
-				event->event, &uobj->event_list,
-				&uobj->events_reported);
+void ib_uverbs_wq_event_handler(struct ib_event *event, void *context_ptr)
+{
+	uverbs_uobj_event(&event->element.wq->uobject->uevent, event);
 }
 
 void ib_uverbs_srq_event_handler(struct ib_event *event, void *context_ptr)
 {
-	struct ib_uevent_object *uobj;
-
-	uobj = container_of(event->element.srq->uobject,
-			    struct ib_uevent_object, uobject);
-
-	ib_uverbs_async_handler(context_ptr, uobj->uobject.user_handle,
-				event->event, &uobj->event_list,
-				&uobj->events_reported);
+	uverbs_uobj_event(&event->element.srq->uobject->uevent, event);
 }
 
-void ib_uverbs_event_handler(struct ib_event_handler *handler,
-			     struct ib_event *event)
+static void ib_uverbs_event_handler(struct ib_event_handler *handler,
+				    struct ib_event *event)
 {
-	struct ib_uverbs_file *file =
-		container_of(handler, struct ib_uverbs_file, event_handler);
-
-	ib_uverbs_async_handler(file, event->element.port_num, event->event,
-				NULL, NULL);
+	ib_uverbs_async_handler(
+		container_of(handler, struct ib_uverbs_async_event_file,
+			     event_handler),
+		event->element.port_num, event->event, NULL, NULL);
 }
 
-struct file *ib_uverbs_alloc_event_file(struct ib_uverbs_file *uverbs_file,
-					int is_async)
+void ib_uverbs_init_event_queue(struct ib_uverbs_event_queue *ev_queue)
 {
-	struct ib_uverbs_event_file *ev_file;
-	struct file *filp;
+	spin_lock_init(&ev_queue->lock);
+	INIT_LIST_HEAD(&ev_queue->event_list);
+	init_waitqueue_head(&ev_queue->poll_wait);
+	ev_queue->is_closed   = 0;
+	ev_queue->async_queue = NULL;
+}
 
-	ev_file = kmalloc(sizeof *ev_file, GFP_KERNEL);
-	if (!ev_file)
-		return ERR_PTR(-ENOMEM);
+void ib_uverbs_init_async_event_file(
+	struct ib_uverbs_async_event_file *async_file)
+{
+	struct ib_uverbs_file *uverbs_file = async_file->uobj.ufile;
+	struct ib_device *ib_dev = async_file->uobj.context->device;
 
-	kref_init(&ev_file->ref);
-	spin_lock_init(&ev_file->lock);
-	INIT_LIST_HEAD(&ev_file->event_list);
-	init_waitqueue_head(&ev_file->poll_wait);
-	ev_file->uverbs_file = uverbs_file;
-	ev_file->async_queue = NULL;
-	ev_file->is_async    = is_async;
-	ev_file->is_closed   = 0;
+	ib_uverbs_init_event_queue(&async_file->ev_queue);
 
-	filp = anon_inode_getfile("[infinibandevent]", &uverbs_event_fops,
-				  ev_file, O_RDONLY);
-	if (IS_ERR(filp))
-		kfree(ev_file);
+	/* The first async_event_file becomes the default one for the file. */
+	mutex_lock(&uverbs_file->ucontext_lock);
+	if (!uverbs_file->default_async_file) {
+		/* Pairs with the put in ib_uverbs_release_file */
+		uverbs_uobject_get(&async_file->uobj);
+		smp_store_release(&uverbs_file->default_async_file, async_file);
+	}
+	mutex_unlock(&uverbs_file->ucontext_lock);
 
-	return filp;
+	INIT_IB_EVENT_HANDLER(&async_file->event_handler, ib_dev,
+			      ib_uverbs_event_handler);
+	ib_register_event_handler(&async_file->event_handler);
 }
 
-/*
- * Look up a completion event file by FD.  If lookup is successful,
- * takes a ref to the event file struct that it returns; if
- * unsuccessful, returns NULL.
- */
-struct ib_uverbs_event_file *ib_uverbs_lookup_comp_file(int fd)
+static ssize_t verify_hdr(struct ib_uverbs_cmd_hdr *hdr,
+			  struct ib_uverbs_ex_cmd_hdr *ex_hdr, size_t count,
+			  const struct uverbs_api_write_method *method_elm)
 {
-	struct ib_uverbs_event_file *ev_file = NULL;
-	struct fd f = fdget(fd);
+	if (method_elm->is_ex) {
+		count -= sizeof(*hdr) + sizeof(*ex_hdr);
 
-	if (!f.file)
-		return NULL;
+		if ((hdr->in_words + ex_hdr->provider_in_words) * 8 != count)
+			return -EINVAL;
 
-	if (f.file->f_op != &uverbs_event_fops)
-		goto out;
+		if (hdr->in_words * 8 < method_elm->req_size)
+			return -ENOSPC;
 
-	ev_file = f.file->private_data;
-	if (ev_file->is_async) {
-		ev_file = NULL;
-		goto out;
+		if (ex_hdr->cmd_hdr_reserved)
+			return -EINVAL;
+
+		if (ex_hdr->response) {
+			if (!hdr->out_words && !ex_hdr->provider_out_words)
+				return -EINVAL;
+
+			if (hdr->out_words * 8 < method_elm->resp_size)
+				return -ENOSPC;
+
+			if (!access_ok(u64_to_user_ptr(ex_hdr->response),
+				       (hdr->out_words + ex_hdr->provider_out_words) * 8))
+				return -EFAULT;
+		} else {
+			if (hdr->out_words || ex_hdr->provider_out_words)
+				return -EINVAL;
+		}
+
+		return 0;
 	}
 
-	kref_get(&ev_file->ref);
+	/* not extended command */
+	if (hdr->in_words * 4 != count)
+		return -EINVAL;
 
-out:
-	fdput(f);
-	return ev_file;
+	if (count < method_elm->req_size + sizeof(*hdr)) {
+		/*
+		 * rdma-core v18 and v19 have a bug where they send DESTROY_CQ
+		 * with a 16 byte write instead of 24. Old kernels didn't
+		 * check the size so they allowed this. Now that the size is
+		 * checked provide a compatibility work around to not break
+		 * those userspaces.
+		 */
+		if (hdr->command == IB_USER_VERBS_CMD_DESTROY_CQ &&
+		    count == 16) {
+			hdr->in_words = 6;
+			return 0;
+		}
+		return -ENOSPC;
+	}
+	if (hdr->out_words * 4 < method_elm->resp_size)
+		return -ENOSPC;
+
+	return 0;
 }
 
 static ssize_t ib_uverbs_write(struct file *filp, const char __user *buf,
 			     size_t count, loff_t *pos)
 {
 	struct ib_uverbs_file *file = filp->private_data;
+	const struct uverbs_api_write_method *method_elm;
+	struct uverbs_api *uapi = file->device->uapi;
+	struct ib_uverbs_ex_cmd_hdr ex_hdr;
 	struct ib_uverbs_cmd_hdr hdr;
+	struct uverbs_attr_bundle bundle;
+	int srcu_key;
+	ssize_t ret;
+
+	if (!ib_safe_file_access(filp)) {
+		pr_err_once("uverbs_write: process %d (%s) changed security contexts after opening file descriptor, this is not allowed.\n",
+			    task_tgid_vnr(current), current->comm);
+		return -EACCES;
+	}
 
-	if (count < sizeof hdr)
+	if (count < sizeof(hdr))
 		return -EINVAL;
 
-	if (copy_from_user(&hdr, buf, sizeof hdr))
+	if (copy_from_user(&hdr, buf, sizeof(hdr)))
 		return -EFAULT;
 
-	if (hdr.in_words * 4 != count)
-		return -EINVAL;
+	method_elm = uapi_get_method(uapi, hdr.command);
+	if (IS_ERR(method_elm))
+		return PTR_ERR(method_elm);
 
-	if (hdr.command >= ARRAY_SIZE(uverbs_cmd_table) ||
-	    !uverbs_cmd_table[hdr.command])
-		return -EINVAL;
+	if (method_elm->is_ex) {
+		if (count < (sizeof(hdr) + sizeof(ex_hdr)))
+			return -EINVAL;
+		if (copy_from_user(&ex_hdr, buf + sizeof(hdr), sizeof(ex_hdr)))
+			return -EFAULT;
+	}
 
-	if (!file->ucontext &&
-	    hdr.command != IB_USER_VERBS_CMD_GET_CONTEXT)
-		return -EINVAL;
+	ret = verify_hdr(&hdr, &ex_hdr, count, method_elm);
+	if (ret)
+		return ret;
+
+	srcu_key = srcu_read_lock(&file->device->disassociate_srcu);
+
+	buf += sizeof(hdr);
+
+	memset(bundle.attr_present, 0, sizeof(bundle.attr_present));
+	bundle.ufile = file;
+	bundle.context = NULL; /* only valid if bundle has uobject */
+	bundle.uobject = NULL;
+	if (!method_elm->is_ex) {
+		size_t in_len = hdr.in_words * 4 - sizeof(hdr);
+		size_t out_len = hdr.out_words * 4;
+		u64 response = 0;
+
+		if (method_elm->has_udata) {
+			bundle.driver_udata.inlen =
+				in_len - method_elm->req_size;
+			in_len = method_elm->req_size;
+			if (bundle.driver_udata.inlen)
+				bundle.driver_udata.inbuf = buf + in_len;
+			else
+				bundle.driver_udata.inbuf = NULL;
+		} else {
+			memset(&bundle.driver_udata, 0,
+			       sizeof(bundle.driver_udata));
+		}
 
-	if (!(file->device->ib_dev->uverbs_cmd_mask & (1ull << hdr.command)))
-		return -ENOSYS;
+		if (method_elm->has_resp) {
+			/*
+			 * The macros check that if has_resp is set
+			 * then the command request structure starts
+			 * with a '__aligned u64 response' member.
+			 */
+			ret = get_user(response, (const u64 __user *)buf);
+			if (ret)
+				goto out_unlock;
+
+			if (method_elm->has_udata) {
+				bundle.driver_udata.outlen =
+					out_len - method_elm->resp_size;
+				out_len = method_elm->resp_size;
+				if (bundle.driver_udata.outlen)
+					bundle.driver_udata.outbuf =
+						u64_to_user_ptr(response +
+								out_len);
+				else
+					bundle.driver_udata.outbuf = NULL;
+			}
+		} else {
+			bundle.driver_udata.outlen = 0;
+			bundle.driver_udata.outbuf = NULL;
+		}
+
+		ib_uverbs_init_udata_buf_or_null(
+			&bundle.ucore, buf, u64_to_user_ptr(response),
+			in_len, out_len);
+	} else {
+		buf += sizeof(ex_hdr);
+
+		ib_uverbs_init_udata_buf_or_null(&bundle.ucore, buf,
+					u64_to_user_ptr(ex_hdr.response),
+					hdr.in_words * 8, hdr.out_words * 8);
+
+		ib_uverbs_init_udata_buf_or_null(
+			&bundle.driver_udata, buf + bundle.ucore.inlen,
+			u64_to_user_ptr(ex_hdr.response) + bundle.ucore.outlen,
+			ex_hdr.provider_in_words * 8,
+			ex_hdr.provider_out_words * 8);
+
+	}
 
-	return uverbs_cmd_table[hdr.command](file, buf + sizeof hdr,
-					     hdr.in_words * 4, hdr.out_words * 4);
+	ret = method_elm->handler(&bundle);
+	if (bundle.uobject)
+		uverbs_finalize_object(bundle.uobject, UVERBS_ACCESS_NEW, true,
+				       !ret, &bundle);
+out_unlock:
+	srcu_read_unlock(&file->device->disassociate_srcu, srcu_key);
+	return (ret) ? : count;
 }
 
+static const struct vm_operations_struct rdma_umap_ops;
+
 static int ib_uverbs_mmap(struct file *filp, struct vm_area_struct *vma)
 {
 	struct ib_uverbs_file *file = filp->private_data;
+	struct ib_ucontext *ucontext;
+	int ret = 0;
+	int srcu_key;
 
-	if (!file->ucontext)
-		return -ENODEV;
-	else
-		return file->device->ib_dev->mmap(file->ucontext, vma);
+	srcu_key = srcu_read_lock(&file->device->disassociate_srcu);
+	ucontext = ib_uverbs_get_ucontext_file(file);
+	if (IS_ERR(ucontext)) {
+		ret = PTR_ERR(ucontext);
+		goto out;
+	}
+
+	mutex_lock(&file->disassociation_lock);
+
+	vma->vm_ops = &rdma_umap_ops;
+	ret = ucontext->device->ops.mmap(ucontext, vma);
+
+	mutex_unlock(&file->disassociation_lock);
+out:
+	srcu_read_unlock(&file->device->disassociate_srcu, srcu_key);
+	return ret;
+}
+
+/*
+ * The VMA has been dup'd, initialize the vm_private_data with a new tracking
+ * struct
+ */
+static void rdma_umap_open(struct vm_area_struct *vma)
+{
+	struct ib_uverbs_file *ufile = vma->vm_file->private_data;
+	struct rdma_umap_priv *opriv = vma->vm_private_data;
+	struct rdma_umap_priv *priv;
+
+	if (!opriv)
+		return;
+
+	/* We are racing with disassociation */
+	if (!down_read_trylock(&ufile->hw_destroy_rwsem))
+		goto out_zap;
+	mutex_lock(&ufile->disassociation_lock);
+
+	/*
+	 * Disassociation already completed, the VMA should already be zapped.
+	 */
+	if (!ufile->ucontext)
+		goto out_unlock;
+
+	priv = kzalloc(sizeof(*priv), GFP_KERNEL);
+	if (!priv)
+		goto out_unlock;
+	rdma_umap_priv_init(priv, vma, opriv->entry);
+
+	mutex_unlock(&ufile->disassociation_lock);
+	up_read(&ufile->hw_destroy_rwsem);
+	return;
+
+out_unlock:
+	mutex_unlock(&ufile->disassociation_lock);
+	up_read(&ufile->hw_destroy_rwsem);
+out_zap:
+	/*
+	 * We can't allow the VMA to be created with the actual IO pages, that
+	 * would break our API contract, and it can't be stopped at this
+	 * point, so zap it.
+	 */
+	vma->vm_private_data = NULL;
+	zap_vma_ptes(vma, vma->vm_start, vma->vm_end - vma->vm_start);
+}
+
+static void rdma_umap_close(struct vm_area_struct *vma)
+{
+	struct ib_uverbs_file *ufile = vma->vm_file->private_data;
+	struct rdma_umap_priv *priv = vma->vm_private_data;
+
+	if (!priv)
+		return;
+
+	/*
+	 * The vma holds a reference on the struct file that created it, which
+	 * in turn means that the ib_uverbs_file is guaranteed to exist at
+	 * this point.
+	 */
+	mutex_lock(&ufile->umap_lock);
+	if (priv->entry)
+		rdma_user_mmap_entry_put(priv->entry);
+
+	list_del(&priv->list);
+	mutex_unlock(&ufile->umap_lock);
+	kfree(priv);
+}
+
+/*
+ * Once the zap_vma_ptes has been called touches to the VMA will come here and
+ * we return a dummy writable zero page for all the pfns.
+ */
+static vm_fault_t rdma_umap_fault(struct vm_fault *vmf)
+{
+	struct ib_uverbs_file *ufile = vmf->vma->vm_file->private_data;
+	struct rdma_umap_priv *priv = vmf->vma->vm_private_data;
+	vm_fault_t ret = 0;
+
+	if (!priv)
+		return VM_FAULT_SIGBUS;
+
+	/* Read only pages can just use the system zero page. */
+	if (!(vmf->vma->vm_flags & (VM_WRITE | VM_MAYWRITE))) {
+		vmf->page = ZERO_PAGE(vmf->address);
+		get_page(vmf->page);
+		return 0;
+	}
+
+	mutex_lock(&ufile->umap_lock);
+	if (!ufile->disassociate_page)
+		ufile->disassociate_page =
+			alloc_pages(vmf->gfp_mask | __GFP_ZERO, 0);
+
+	if (ufile->disassociate_page) {
+		/*
+		 * This VMA is forced to always be shared so this doesn't have
+		 * to worry about COW.
+		 */
+		vmf->page = ufile->disassociate_page;
+		get_page(vmf->page);
+	} else {
+		ret = VM_FAULT_SIGBUS;
+	}
+	mutex_unlock(&ufile->umap_lock);
+
+	return ret;
 }
 
+static const struct vm_operations_struct rdma_umap_ops = {
+	.open = rdma_umap_open,
+	.close = rdma_umap_close,
+	.fault = rdma_umap_fault,
+};
+
+void uverbs_user_mmap_disassociate(struct ib_uverbs_file *ufile)
+{
+	struct rdma_umap_priv *priv, *next_priv;
+
+	mutex_lock(&ufile->disassociation_lock);
+
+	while (1) {
+		struct mm_struct *mm = NULL;
+
+		/* Get an arbitrary mm pointer that hasn't been cleaned yet */
+		mutex_lock(&ufile->umap_lock);
+		while (!list_empty(&ufile->umaps)) {
+			int ret;
+
+			priv = list_first_entry(&ufile->umaps,
+						struct rdma_umap_priv, list);
+			mm = priv->vma->vm_mm;
+			ret = mmget_not_zero(mm);
+			if (!ret) {
+				list_del_init(&priv->list);
+				if (priv->entry) {
+					rdma_user_mmap_entry_put(priv->entry);
+					priv->entry = NULL;
+				}
+				mm = NULL;
+				continue;
+			}
+			break;
+		}
+		mutex_unlock(&ufile->umap_lock);
+		if (!mm) {
+			mutex_unlock(&ufile->disassociation_lock);
+			return;
+		}
+
+		/*
+		 * The umap_lock is nested under mmap_lock since it used within
+		 * the vma_ops callbacks, so we have to clean the list one mm
+		 * at a time to get the lock ordering right. Typically there
+		 * will only be one mm, so no big deal.
+		 */
+		mmap_read_lock(mm);
+		mutex_lock(&ufile->umap_lock);
+		list_for_each_entry_safe (priv, next_priv, &ufile->umaps,
+					  list) {
+			struct vm_area_struct *vma = priv->vma;
+
+			if (vma->vm_mm != mm)
+				continue;
+			list_del_init(&priv->list);
+
+			zap_vma_ptes(vma, vma->vm_start,
+				     vma->vm_end - vma->vm_start);
+
+			if (priv->entry) {
+				rdma_user_mmap_entry_put(priv->entry);
+				priv->entry = NULL;
+			}
+		}
+		mutex_unlock(&ufile->umap_lock);
+		mmap_read_unlock(mm);
+		mmput(mm);
+	}
+
+	mutex_unlock(&ufile->disassociation_lock);
+}
+
+/**
+ * rdma_user_mmap_disassociate() - Revoke mmaps for a device
+ * @device: device to revoke
+ *
+ * This function should be called by drivers that need to disable mmaps for the
+ * device, for instance because it is going to be reset.
+ */
+void rdma_user_mmap_disassociate(struct ib_device *device)
+{
+	struct ib_uverbs_device *uverbs_dev =
+		ib_get_client_data(device, &uverbs_client);
+	struct ib_uverbs_file *ufile;
+
+	mutex_lock(&uverbs_dev->lists_mutex);
+	list_for_each_entry(ufile, &uverbs_dev->uverbs_file_list, list) {
+		if (ufile->ucontext)
+			uverbs_user_mmap_disassociate(ufile);
+	}
+	mutex_unlock(&uverbs_dev->lists_mutex);
+}
+EXPORT_SYMBOL(rdma_user_mmap_disassociate);
+
 /*
  * ib_uverbs_open() does not need the BKL:
  *
@@ -625,40 +930,82 @@ static int ib_uverbs_open(struct inode *inode, struct file *filp)
 {
 	struct ib_uverbs_device *dev;
 	struct ib_uverbs_file *file;
+	struct ib_device *ib_dev;
 	int ret;
+	int module_dependent;
+	int srcu_key;
 
 	dev = container_of(inode->i_cdev, struct ib_uverbs_device, cdev);
-	if (dev)
-		kref_get(&dev->ref);
-	else
+	if (!refcount_inc_not_zero(&dev->refcount))
 		return -ENXIO;
 
-	if (!try_module_get(dev->ib_dev->owner)) {
-		ret = -ENODEV;
+	get_device(&dev->dev);
+	srcu_key = srcu_read_lock(&dev->disassociate_srcu);
+	mutex_lock(&dev->lists_mutex);
+	ib_dev = srcu_dereference(dev->ib_dev,
+				  &dev->disassociate_srcu);
+	if (!ib_dev) {
+		ret = -EIO;
 		goto err;
 	}
 
-	file = kmalloc(sizeof *file, GFP_KERNEL);
+	if (!rdma_dev_access_netns(ib_dev, current->nsproxy->net_ns)) {
+		ret = -EPERM;
+		goto err;
+	}
+
+	/* In case IB device supports disassociate ucontext, there is no hard
+	 * dependency between uverbs device and its low level device.
+	 */
+	module_dependent = !(ib_dev->ops.disassociate_ucontext);
+
+	if (module_dependent) {
+		if (!try_module_get(ib_dev->ops.owner)) {
+			ret = -ENODEV;
+			goto err;
+		}
+	}
+
+	file = kzalloc(sizeof(*file), GFP_KERNEL);
 	if (!file) {
 		ret = -ENOMEM;
-		goto err_module;
+		if (module_dependent)
+			goto err_module;
+
+		goto err;
 	}
 
 	file->device	 = dev;
-	file->ucontext	 = NULL;
-	file->async_file = NULL;
 	kref_init(&file->ref);
-	mutex_init(&file->mutex);
+	mutex_init(&file->ucontext_lock);
+
+	spin_lock_init(&file->uobjects_lock);
+	INIT_LIST_HEAD(&file->uobjects);
+	init_rwsem(&file->hw_destroy_rwsem);
+	mutex_init(&file->umap_lock);
+	INIT_LIST_HEAD(&file->umaps);
+
+	mutex_init(&file->disassociation_lock);
 
 	filp->private_data = file;
+	list_add_tail(&file->list, &dev->uverbs_file_list);
+	mutex_unlock(&dev->lists_mutex);
+	srcu_read_unlock(&dev->disassociate_srcu, srcu_key);
+
+	setup_ufile_idr_uobject(file);
 
-	return nonseekable_open(inode, filp);
+	return stream_open(inode, filp);
 
 err_module:
-	module_put(dev->ib_dev->owner);
+	module_put(ib_dev->ops.owner);
 
 err:
-	kref_put(&dev->ref, ib_uverbs_release_dev);
+	mutex_unlock(&dev->lists_mutex);
+	srcu_read_unlock(&dev->disassociate_srcu, srcu_key);
+	if (refcount_dec_and_test(&dev->refcount))
+		ib_uverbs_comp_dev(dev);
+
+	put_device(&dev->dev);
 	return ret;
 }
 
@@ -666,10 +1013,11 @@ static int ib_uverbs_close(struct inode *inode, struct file *filp)
 {
 	struct ib_uverbs_file *file = filp->private_data;
 
-	ib_uverbs_cleanup_ucontext(file, file->ucontext);
+	uverbs_destroy_ufile_hw(file, RDMA_REMOVE_CLOSE);
 
-	if (file->async_file)
-		kref_put(&file->async_file->ref, ib_uverbs_release_event_file);
+	mutex_lock(&file->device->lists_mutex);
+	list_del_init(&file->list);
+	mutex_unlock(&file->device->lists_mutex);
 
 	kref_put(&file->ref, ib_uverbs_release_file);
 
@@ -681,7 +1029,8 @@ static const struct file_operations uverbs_fops = {
 	.write	 = ib_uverbs_write,
 	.open	 = ib_uverbs_open,
 	.release = ib_uverbs_close,
-	.llseek	 = no_llseek,
+	.unlocked_ioctl = ib_uverbs_ioctl,
+	.compat_ioctl = compat_ptr_ioctl,
 };
 
 static const struct file_operations uverbs_mmap_fops = {
@@ -690,216 +1039,300 @@ static const struct file_operations uverbs_mmap_fops = {
 	.mmap    = ib_uverbs_mmap,
 	.open	 = ib_uverbs_open,
 	.release = ib_uverbs_close,
-	.llseek	 = no_llseek,
+	.unlocked_ioctl = ib_uverbs_ioctl,
+	.compat_ioctl = compat_ptr_ioctl,
 };
 
+static int ib_uverbs_get_nl_info(struct ib_device *ibdev, void *client_data,
+				 struct ib_client_nl_info *res)
+{
+	struct ib_uverbs_device *uverbs_dev = client_data;
+	int ret;
+
+	if (res->port != -1)
+		return -EINVAL;
+
+	res->abi = ibdev->ops.uverbs_abi_ver;
+	res->cdev = &uverbs_dev->dev;
+
+	/*
+	 * To support DRIVER_ID binding in userspace some of the driver need
+	 * upgrading to expose their PCI dependent revision information
+	 * through get_context instead of relying on modalias matching. When
+	 * the drivers are fixed they can drop this flag.
+	 */
+	if (!ibdev->ops.uverbs_no_driver_id_binding) {
+		ret = nla_put_u32(res->nl_msg, RDMA_NLDEV_ATTR_UVERBS_DRIVER_ID,
+				  ibdev->ops.driver_id);
+		if (ret)
+			return ret;
+	}
+	return 0;
+}
+
 static struct ib_client uverbs_client = {
 	.name   = "uverbs",
+	.no_kverbs_req = true,
 	.add    = ib_uverbs_add_one,
-	.remove = ib_uverbs_remove_one
+	.remove = ib_uverbs_remove_one,
+	.get_nl_info = ib_uverbs_get_nl_info,
 };
+MODULE_ALIAS_RDMA_CLIENT("uverbs");
 
-static ssize_t show_ibdev(struct device *device, struct device_attribute *attr,
+static ssize_t ibdev_show(struct device *device, struct device_attribute *attr,
 			  char *buf)
 {
-	struct ib_uverbs_device *dev = dev_get_drvdata(device);
+	struct ib_uverbs_device *dev =
+			container_of(device, struct ib_uverbs_device, dev);
+	int ret = -ENODEV;
+	int srcu_key;
+	struct ib_device *ib_dev;
 
-	if (!dev)
-		return -ENODEV;
+	srcu_key = srcu_read_lock(&dev->disassociate_srcu);
+	ib_dev = srcu_dereference(dev->ib_dev, &dev->disassociate_srcu);
+	if (ib_dev)
+		ret = sysfs_emit(buf, "%s\n", dev_name(&ib_dev->dev));
+	srcu_read_unlock(&dev->disassociate_srcu, srcu_key);
 
-	return sprintf(buf, "%s\n", dev->ib_dev->name);
+	return ret;
 }
-static DEVICE_ATTR(ibdev, S_IRUGO, show_ibdev, NULL);
+static DEVICE_ATTR_RO(ibdev);
 
-static ssize_t show_dev_abi_version(struct device *device,
-				    struct device_attribute *attr, char *buf)
+static ssize_t abi_version_show(struct device *device,
+				struct device_attribute *attr, char *buf)
 {
-	struct ib_uverbs_device *dev = dev_get_drvdata(device);
+	struct ib_uverbs_device *dev =
+			container_of(device, struct ib_uverbs_device, dev);
+	int ret = -ENODEV;
+	int srcu_key;
+	struct ib_device *ib_dev;
 
-	if (!dev)
-		return -ENODEV;
+	srcu_key = srcu_read_lock(&dev->disassociate_srcu);
+	ib_dev = srcu_dereference(dev->ib_dev, &dev->disassociate_srcu);
+	if (ib_dev)
+		ret = sysfs_emit(buf, "%u\n", ib_dev->ops.uverbs_abi_ver);
+	srcu_read_unlock(&dev->disassociate_srcu, srcu_key);
 
-	return sprintf(buf, "%d\n", dev->ib_dev->uverbs_abi_ver);
+	return ret;
 }
-static DEVICE_ATTR(abi_version, S_IRUGO, show_dev_abi_version, NULL);
+static DEVICE_ATTR_RO(abi_version);
+
+static struct attribute *ib_dev_attrs[] = {
+	&dev_attr_abi_version.attr,
+	&dev_attr_ibdev.attr,
+	NULL,
+};
+
+static const struct attribute_group dev_attr_group = {
+	.attrs = ib_dev_attrs,
+};
 
 static CLASS_ATTR_STRING(abi_version, S_IRUGO,
 			 __stringify(IB_USER_VERBS_ABI_VERSION));
 
-static dev_t overflow_maj;
-static DECLARE_BITMAP(overflow_map, IB_UVERBS_MAX_DEVICES);
-
-/*
- * If we have more than IB_UVERBS_MAX_DEVICES, dynamically overflow by
- * requesting a new major number and doubling the number of max devices we
- * support. It's stupid, but simple.
- */
-static int find_overflow_devnum(void)
+static int ib_uverbs_create_uapi(struct ib_device *device,
+				 struct ib_uverbs_device *uverbs_dev)
 {
-	int ret;
-
-	if (!overflow_maj) {
-		ret = alloc_chrdev_region(&overflow_maj, 0, IB_UVERBS_MAX_DEVICES,
-					  "infiniband_verbs");
-		if (ret) {
-			printk(KERN_ERR "user_verbs: couldn't register dynamic device number\n");
-			return ret;
-		}
-	}
+	struct uverbs_api *uapi;
 
-	ret = find_first_zero_bit(overflow_map, IB_UVERBS_MAX_DEVICES);
-	if (ret >= IB_UVERBS_MAX_DEVICES)
-		return -1;
+	uapi = uverbs_alloc_api(device);
+	if (IS_ERR(uapi))
+		return PTR_ERR(uapi);
 
-	return ret;
+	uverbs_dev->uapi = uapi;
+	return 0;
 }
 
-static void ib_uverbs_add_one(struct ib_device *device)
+static int ib_uverbs_add_one(struct ib_device *device)
 {
 	int devnum;
 	dev_t base;
 	struct ib_uverbs_device *uverbs_dev;
+	int ret;
 
-	if (!device->alloc_ucontext)
-		return;
+	if (!device->ops.alloc_ucontext ||
+	    device->type == RDMA_DEVICE_TYPE_SMI)
+		return -EOPNOTSUPP;
 
-	uverbs_dev = kzalloc(sizeof *uverbs_dev, GFP_KERNEL);
+	uverbs_dev = kzalloc(sizeof(*uverbs_dev), GFP_KERNEL);
 	if (!uverbs_dev)
-		return;
+		return -ENOMEM;
+
+	ret = init_srcu_struct(&uverbs_dev->disassociate_srcu);
+	if (ret) {
+		kfree(uverbs_dev);
+		return -ENOMEM;
+	}
 
-	kref_init(&uverbs_dev->ref);
+	device_initialize(&uverbs_dev->dev);
+	uverbs_dev->dev.class = &uverbs_class;
+	uverbs_dev->dev.parent = device->dev.parent;
+	uverbs_dev->dev.release = ib_uverbs_release_dev;
+	uverbs_dev->groups[0] = &dev_attr_group;
+	uverbs_dev->dev.groups = uverbs_dev->groups;
+	refcount_set(&uverbs_dev->refcount, 1);
 	init_completion(&uverbs_dev->comp);
 	uverbs_dev->xrcd_tree = RB_ROOT;
 	mutex_init(&uverbs_dev->xrcd_tree_mutex);
+	mutex_init(&uverbs_dev->lists_mutex);
+	INIT_LIST_HEAD(&uverbs_dev->uverbs_file_list);
+	rcu_assign_pointer(uverbs_dev->ib_dev, device);
+	uverbs_dev->num_comp_vectors = device->num_comp_vectors;
 
-	spin_lock(&map_lock);
-	devnum = find_first_zero_bit(dev_map, IB_UVERBS_MAX_DEVICES);
-	if (devnum >= IB_UVERBS_MAX_DEVICES) {
-		spin_unlock(&map_lock);
-		devnum = find_overflow_devnum();
-		if (devnum < 0)
-			goto err;
-
-		spin_lock(&map_lock);
-		uverbs_dev->devnum = devnum + IB_UVERBS_MAX_DEVICES;
-		base = devnum + overflow_maj;
-		set_bit(devnum, overflow_map);
-	} else {
-		uverbs_dev->devnum = devnum;
-		base = devnum + IB_UVERBS_BASE_DEV;
-		set_bit(devnum, dev_map);
+	devnum = ida_alloc_max(&uverbs_ida, IB_UVERBS_MAX_DEVICES - 1,
+			       GFP_KERNEL);
+	if (devnum < 0) {
+		ret = -ENOMEM;
+		goto err;
 	}
-	spin_unlock(&map_lock);
+	uverbs_dev->devnum = devnum;
+	if (devnum >= IB_UVERBS_NUM_FIXED_MINOR)
+		base = dynamic_uverbs_dev + devnum - IB_UVERBS_NUM_FIXED_MINOR;
+	else
+		base = IB_UVERBS_BASE_DEV + devnum;
 
-	uverbs_dev->ib_dev           = device;
-	uverbs_dev->num_comp_vectors = device->num_comp_vectors;
+	ret = ib_uverbs_create_uapi(device, uverbs_dev);
+	if (ret)
+		goto err_uapi;
 
-	cdev_init(&uverbs_dev->cdev, NULL);
-	uverbs_dev->cdev.owner = THIS_MODULE;
-	uverbs_dev->cdev.ops = device->mmap ? &uverbs_mmap_fops : &uverbs_fops;
-	kobject_set_name(&uverbs_dev->cdev.kobj, "uverbs%d", uverbs_dev->devnum);
-	if (cdev_add(&uverbs_dev->cdev, base, 1))
-		goto err_cdev;
-
-	uverbs_dev->dev = device_create(uverbs_class, device->dma_device,
-					uverbs_dev->cdev.dev, uverbs_dev,
-					"uverbs%d", uverbs_dev->devnum);
-	if (IS_ERR(uverbs_dev->dev))
-		goto err_cdev;
-
-	if (device_create_file(uverbs_dev->dev, &dev_attr_ibdev))
-		goto err_class;
-	if (device_create_file(uverbs_dev->dev, &dev_attr_abi_version))
-		goto err_class;
+	uverbs_dev->dev.devt = base;
+	dev_set_name(&uverbs_dev->dev, "uverbs%d", uverbs_dev->devnum);
 
-	ib_set_client_data(device, &uverbs_client, uverbs_dev);
-
-	return;
+	cdev_init(&uverbs_dev->cdev,
+		  device->ops.mmap ? &uverbs_mmap_fops : &uverbs_fops);
+	uverbs_dev->cdev.owner = THIS_MODULE;
 
-err_class:
-	device_destroy(uverbs_class, uverbs_dev->cdev.dev);
+	ret = cdev_device_add(&uverbs_dev->cdev, &uverbs_dev->dev);
+	if (ret)
+		goto err_uapi;
 
-err_cdev:
-	cdev_del(&uverbs_dev->cdev);
-	if (uverbs_dev->devnum < IB_UVERBS_MAX_DEVICES)
-		clear_bit(devnum, dev_map);
-	else
-		clear_bit(devnum, overflow_map);
+	ib_set_client_data(device, &uverbs_client, uverbs_dev);
+	return 0;
 
+err_uapi:
+	ida_free(&uverbs_ida, devnum);
 err:
-	kref_put(&uverbs_dev->ref, ib_uverbs_release_dev);
+	if (refcount_dec_and_test(&uverbs_dev->refcount))
+		ib_uverbs_comp_dev(uverbs_dev);
 	wait_for_completion(&uverbs_dev->comp);
-	kfree(uverbs_dev);
-	return;
+	put_device(&uverbs_dev->dev);
+	return ret;
 }
 
-static void ib_uverbs_remove_one(struct ib_device *device)
+static void ib_uverbs_free_hw_resources(struct ib_uverbs_device *uverbs_dev,
+					struct ib_device *ib_dev)
 {
-	struct ib_uverbs_device *uverbs_dev = ib_get_client_data(device, &uverbs_client);
+	struct ib_uverbs_file *file;
 
-	if (!uverbs_dev)
-		return;
+	/* Pending running commands to terminate */
+	uverbs_disassociate_api_pre(uverbs_dev);
 
-	dev_set_drvdata(uverbs_dev->dev, NULL);
-	device_destroy(uverbs_class, uverbs_dev->cdev.dev);
-	cdev_del(&uverbs_dev->cdev);
+	mutex_lock(&uverbs_dev->lists_mutex);
+	while (!list_empty(&uverbs_dev->uverbs_file_list)) {
+		file = list_first_entry(&uverbs_dev->uverbs_file_list,
+					struct ib_uverbs_file, list);
+		list_del_init(&file->list);
+		kref_get(&file->ref);
 
-	if (uverbs_dev->devnum < IB_UVERBS_MAX_DEVICES)
-		clear_bit(uverbs_dev->devnum, dev_map);
-	else
-		clear_bit(uverbs_dev->devnum - IB_UVERBS_MAX_DEVICES, overflow_map);
+		/* We must release the mutex before going ahead and calling
+		 * uverbs_cleanup_ufile, as it might end up indirectly calling
+		 * uverbs_close, for example due to freeing the resources (e.g
+		 * mmput).
+		 */
+		mutex_unlock(&uverbs_dev->lists_mutex);
 
-	kref_put(&uverbs_dev->ref, ib_uverbs_release_dev);
-	wait_for_completion(&uverbs_dev->comp);
-	kfree(uverbs_dev);
+		uverbs_destroy_ufile_hw(file, RDMA_REMOVE_DRIVER_REMOVE);
+		kref_put(&file->ref, ib_uverbs_release_file);
+
+		mutex_lock(&uverbs_dev->lists_mutex);
+	}
+	mutex_unlock(&uverbs_dev->lists_mutex);
+
+	uverbs_disassociate_api(uverbs_dev->uapi);
 }
 
-static char *uverbs_devnode(struct device *dev, umode_t *mode)
+static void ib_uverbs_remove_one(struct ib_device *device, void *client_data)
 {
-	if (mode)
-		*mode = 0666;
-	return kasprintf(GFP_KERNEL, "infiniband/%s", dev_name(dev));
+	struct ib_uverbs_device *uverbs_dev = client_data;
+	int wait_clients = 1;
+
+	cdev_device_del(&uverbs_dev->cdev, &uverbs_dev->dev);
+	ida_free(&uverbs_ida, uverbs_dev->devnum);
+
+	if (device->ops.disassociate_ucontext) {
+		/* We disassociate HW resources and immediately return.
+		 * Userspace will see a EIO errno for all future access.
+		 * Upon returning, ib_device may be freed internally and is not
+		 * valid any more.
+		 * uverbs_device is still available until all clients close
+		 * their files, then the uverbs device ref count will be zero
+		 * and its resources will be freed.
+		 * Note: At this point no more files can be opened since the
+		 * cdev was deleted, however active clients can still issue
+		 * commands and close their open files.
+		 */
+		ib_uverbs_free_hw_resources(uverbs_dev, device);
+		wait_clients = 0;
+	}
+
+	if (refcount_dec_and_test(&uverbs_dev->refcount))
+		ib_uverbs_comp_dev(uverbs_dev);
+	if (wait_clients)
+		wait_for_completion(&uverbs_dev->comp);
+
+	put_device(&uverbs_dev->dev);
 }
 
 static int __init ib_uverbs_init(void)
 {
 	int ret;
 
-	ret = register_chrdev_region(IB_UVERBS_BASE_DEV, IB_UVERBS_MAX_DEVICES,
+	ret = register_chrdev_region(IB_UVERBS_BASE_DEV,
+				     IB_UVERBS_NUM_FIXED_MINOR,
 				     "infiniband_verbs");
 	if (ret) {
-		printk(KERN_ERR "user_verbs: couldn't register device number\n");
+		pr_err("user_verbs: couldn't register device number\n");
 		goto out;
 	}
 
-	uverbs_class = class_create(THIS_MODULE, "infiniband_verbs");
-	if (IS_ERR(uverbs_class)) {
-		ret = PTR_ERR(uverbs_class);
-		printk(KERN_ERR "user_verbs: couldn't create class infiniband_verbs\n");
-		goto out_chrdev;
+	ret = alloc_chrdev_region(&dynamic_uverbs_dev, 0,
+				  IB_UVERBS_NUM_DYNAMIC_MINOR,
+				  "infiniband_verbs");
+	if (ret) {
+		pr_err("couldn't register dynamic device number\n");
+		goto out_alloc;
 	}
 
-	uverbs_class->devnode = uverbs_devnode;
+	ret = class_register(&uverbs_class);
+	if (ret) {
+		pr_err("user_verbs: couldn't create class infiniband_verbs\n");
+		goto out_chrdev;
+	}
 
-	ret = class_create_file(uverbs_class, &class_attr_abi_version.attr);
+	ret = class_create_file(&uverbs_class, &class_attr_abi_version.attr);
 	if (ret) {
-		printk(KERN_ERR "user_verbs: couldn't create abi_version attribute\n");
+		pr_err("user_verbs: couldn't create abi_version attribute\n");
 		goto out_class;
 	}
 
 	ret = ib_register_client(&uverbs_client);
 	if (ret) {
-		printk(KERN_ERR "user_verbs: couldn't register client\n");
+		pr_err("user_verbs: couldn't register client\n");
 		goto out_class;
 	}
 
 	return 0;
 
 out_class:
-	class_destroy(uverbs_class);
+	class_unregister(&uverbs_class);
 
 out_chrdev:
-	unregister_chrdev_region(IB_UVERBS_BASE_DEV, IB_UVERBS_MAX_DEVICES);
+	unregister_chrdev_region(dynamic_uverbs_dev,
+				 IB_UVERBS_NUM_DYNAMIC_MINOR);
+
+out_alloc:
+	unregister_chrdev_region(IB_UVERBS_BASE_DEV,
+				 IB_UVERBS_NUM_FIXED_MINOR);
 
 out:
 	return ret;
@@ -908,17 +1341,13 @@ out:
 static void __exit ib_uverbs_cleanup(void)
 {
 	ib_unregister_client(&uverbs_client);
-	class_destroy(uverbs_class);
-	unregister_chrdev_region(IB_UVERBS_BASE_DEV, IB_UVERBS_MAX_DEVICES);
-	if (overflow_maj)
-		unregister_chrdev_region(overflow_maj, IB_UVERBS_MAX_DEVICES);
-	idr_destroy(&ib_uverbs_pd_idr);
-	idr_destroy(&ib_uverbs_mr_idr);
-	idr_destroy(&ib_uverbs_mw_idr);
-	idr_destroy(&ib_uverbs_ah_idr);
-	idr_destroy(&ib_uverbs_cq_idr);
-	idr_destroy(&ib_uverbs_qp_idr);
-	idr_destroy(&ib_uverbs_srq_idr);
+	class_unregister(&uverbs_class);
+	unregister_chrdev_region(IB_UVERBS_BASE_DEV,
+				 IB_UVERBS_NUM_FIXED_MINOR);
+	unregister_chrdev_region(dynamic_uverbs_dev,
+				 IB_UVERBS_NUM_DYNAMIC_MINOR);
+	ib_cleanup_ucaps();
+	mmu_notifier_synchronize();
 }
 
 module_init(ib_uverbs_init);