[for-next,V4,3/5] IB/uverbs: Enable device removal when there are active user space applications

Commit Message

Yishai Hadas June 11, 2015, 12:56 p.m. UTC

Enables the uverbs_remove_one to succeed despite the fact that there are
running IB applications working with the given ib device.  This functionality
enables a HW device to be unbind/reset despite the fact that there are running
user space applications using it.

It exposes a new IB kernel API named 'disassociate_ucontext' which lets a
driver detaching its HW resources from a given user context without
crashing/terminating the application. In case a driver implemented the above
API and registered with ib_uverb there will be no dependency between its device
to its uverbs_device. Upon calling remove_one of ib_uverbs the call should
return after disassociating the open HW resources without waiting to clients
disconnecting. In case driver didn't implement this API there will be no change
to current behaviour and uverbs_remove_one will return only when last client
has disconnected and reference count on uverbs device became 0.

In case the lower driver device was removed any application will continue
working over some zombie HCA, further calls will ended with an immediate error.

Signed-off-by: Yishai Hadas <[e-mail]>
Signed-off-by: Shachar Raindel <[e-mail]>
---
 drivers/infiniband/core/uverbs.h      |   14 ++-
 drivers/infiniband/core/uverbs_main.c |  363 ++++++++++++++++++++++++++-------
 include/rdma/ib_verbs.h               |    1 +
 3 files changed, 308 insertions(+), 70 deletions(-)

Comments

Jason Gunthorpe June 11, 2015, 5:38 p.m. UTC | #1

On Thu, Jun 11, 2015 at 03:56:52PM +0300, Yishai Hadas wrote:
> Enables the uverbs_remove_one to succeed despite the fact that there are
> running IB applications working with the given ib device.  This functionality
> enables a HW device to be unbind/reset despite the fact that there are running
> user space applications using it.

This looks much saner now, thanks. I only looked briefly today.

> -	struct ib_device		       *ib_dev;
> +	struct ib_device	__rcu	       *ib_dev;

Did you run a static checker to confirm the rcu annoation?

> @@ -332,9 +336,15 @@ static ssize_t ib_uverbs_event_read(struct file *filp, char __user *buf,
>  			return -EAGAIN;
>  
>  		if (wait_event_interruptible(file->poll_wait,
> -					     !list_empty(&file->event_list)))
> +					     (!list_empty(&file->event_list) ||
> +					     !file->uverbs_file->device->ib_dev)))

And it warned about this? Any place else?

The barriers built into wait_event_interruptible() and wake_up() make
this OK. That is worth a comment to explain why it is OK RCU protected
data is not using RCU here.

>  
> @@ -397,8 +407,11 @@ static int ib_uverbs_event_close(struct inode *inode, struct file *filp)
>  {
>  	struct ib_uverbs_event_file *file = filp->private_data;
>  	struct ib_uverbs_event *entry, *tmp;
> +	int closed_already = 0;
>  
> +	mutex_lock(&file->uverbs_file->device->lists_mutex);
>  	spin_lock_irq(&file->lock);
> +	closed_already = file->is_closed;
>  	file->is_closed = 1;

It doesn't loook like is_closed can be changed or read while
lists_mutex is held, so why this ordering and closed_already?

> +	struct ib_device *ib_dev = uverbs_dev->ib_dev;
> +	uverbs_dev->ib_dev = NULL;

These are the write side accesses - and we rely on the driver to
ensure there is only one call to ib_uverbs_free_hw_resources? A
comment would be good, it is unusual to see a RCU write side without
an explicit lock.

> +	/* Pending running commands to terminate */
> +	synchronize_srcu(&uverbs_dev->disassociate_srcu);
> +	event.event = IB_EVENT_DEVICE_FATAL;
> +	event.element.port_num = 0;
> +	event.device = ib_dev;
                     ^^^^^^^^^

How does this lifetime work? It doesn't look like
ib_uverbs_event_handler touches event.device? Is this a nop?

> +		/* We must release the mutex before going ahead and calling
> +		 * disassociate_ucontext. disassociate_ucontext might end up
> +		 * indirectly calling uverbs_close, for example due to freeing
> +		 * the resources (e.g mmput).
> +		 */

Can you provide a call trace for this deadlock possibility?

> +	}
> +
> +	mutex_unlock(&uverbs_dev->lists_mutex);
> +
> +	mutex_lock(&uverbs_dev->lists_mutex);

???

> +	while (!list_empty(&uverbs_dev->uverbs_events_file_list)) {
> +		event_file = list_first_entry(&uverbs_dev->
> +					      uverbs_events_file_list,
> +					      struct ib_uverbs_event_file,
> +					      list);
> +		event_file->is_closed = 1;
> +
> +		list_del(&event_file->list);
> +		if (event_file->is_async)
> +			ib_unregister_event_handler(&event_file->uverbs_file->
> +						    event_handler);

Why do we need to do this in ib_uverbs_free_hw_resources ? How does
the event handler hold a ref on the ib_dev?

> +	if (uverbs_dev->flags & UVERBS_FLAG_DISASSOCIATE) {

I wonder if the flag is needed, isn't having a non-null
disassociate_ucontext function enough proof the driver supports this?

Jason
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to [e-mail]
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Patch

diff --git a/drivers/infiniband/core/uverbs.h b/drivers/infiniband/core/uverbs.h
index 1bf9d04..d6c3ef9 100644
--- a/drivers/infiniband/core/uverbs.h
+++ b/drivers/infiniband/core/uverbs.h
@@ -63,6 +63,10 @@ 
 		(udata)->outlen = (olen);					\
 	} while (0)
 
+enum uverbs_flags {
+	UVERBS_FLAG_DISASSOCIATE = 1
+};
+
 /*
  * Our lifetime rules for these structs are the following:
  *
@@ -89,11 +93,16 @@  struct ib_uverbs_device {
 	int					num_comp_vectors;
 	struct completion			comp;
 	struct device			       *dev;
-	struct ib_device		       *ib_dev;
+	struct ib_device	__rcu	       *ib_dev;
 	int					devnum;
 	struct cdev			        cdev;
 	struct rb_root				xrcd_tree;
 	struct mutex				xrcd_tree_mutex;
+	u32					flags;
+	struct srcu_struct			disassociate_srcu;
+	struct mutex				lists_mutex; /* protect lists */
+	struct list_head			uverbs_file_list;
+	struct list_head			uverbs_events_file_list;
 };
 
 struct ib_uverbs_event_file {
@@ -105,6 +114,7 @@  struct ib_uverbs_event_file {
 	wait_queue_head_t			poll_wait;
 	struct fasync_struct		       *async_queue;
 	struct list_head			event_list;
+	struct list_head			list;
 };
 
 struct ib_uverbs_file {
@@ -114,6 +124,8 @@  struct ib_uverbs_file {
 	struct ib_ucontext		       *ucontext;
 	struct ib_event_handler			event_handler;
 	struct ib_uverbs_event_file	       *async_file;
+	struct list_head			list;
+	int					is_closed;
 };
 
 struct ib_uverbs_event {
diff --git a/drivers/infiniband/core/uverbs_main.c b/drivers/infiniband/core/uverbs_main.c
index 8675290..ff62b5f 100644
--- a/drivers/infiniband/core/uverbs_main.c
+++ b/drivers/infiniband/core/uverbs_main.c
@@ -136,7 +136,12 @@  static void ib_uverbs_release_dev(struct kref *ref)
 	struct ib_uverbs_device *dev =
 		container_of(ref, struct ib_uverbs_device, ref);
 
-	complete(&dev->comp);
+	if (!dev->ib_dev) {
+		cleanup_srcu_struct(&dev->disassociate_srcu);
+		kfree(dev);
+	} else {
+		complete(&dev->comp);
+	}
 }
 
 static void ib_uverbs_release_event_file(struct kref *ref)
@@ -202,9 +207,6 @@  static int ib_uverbs_cleanup_ucontext(struct ib_uverbs_file *file,
 {
 	struct ib_uobject *uobj, *tmp;
 
-	if (!context)
-		return 0;
-
 	context->closing = 1;
 
 	list_for_each_entry_safe(uobj, tmp, &context->ah_list, list) {
@@ -309,7 +311,9 @@  static 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);
+	if (!(file->device->flags & UVERBS_FLAG_DISASSOCIATE))
+		module_put(file->device->ib_dev->owner);
+
 	kref_put(&file->device->ref, ib_uverbs_release_dev);
 
 	kfree(file);
@@ -332,9 +336,15 @@  static ssize_t ib_uverbs_event_read(struct file *filp, char __user *buf,
 			return -EAGAIN;
 
 		if (wait_event_interruptible(file->poll_wait,
-					     !list_empty(&file->event_list)))
+					     (!list_empty(&file->event_list) ||
+					     !file->uverbs_file->device->ib_dev)))
 			return -ERESTARTSYS;
 
+		/* If device was disassociated and no event exists set an error */
+		if (list_empty(&file->event_list) &&
+		    !file->uverbs_file->device->ib_dev)
+			return -EIO;
+
 		spin_lock_irq(&file->lock);
 	}
 
@@ -397,8 +407,11 @@  static int ib_uverbs_event_close(struct inode *inode, struct file *filp)
 {
 	struct ib_uverbs_event_file *file = filp->private_data;
 	struct ib_uverbs_event *entry, *tmp;
+	int closed_already = 0;
 
+	mutex_lock(&file->uverbs_file->device->lists_mutex);
 	spin_lock_irq(&file->lock);
+	closed_already = file->is_closed;
 	file->is_closed = 1;
 	list_for_each_entry_safe(entry, tmp, &file->event_list, list) {
 		if (entry->counter)
@@ -406,9 +419,14 @@  static int ib_uverbs_event_close(struct inode *inode, struct file *filp)
 		kfree(entry);
 	}
 	spin_unlock_irq(&file->lock);
+	if (!closed_already) {
+		list_del(&file->list);
+		if (file->is_async)
+			ib_unregister_event_handler(&file->uverbs_file->
+				event_handler);
+	}
+	mutex_unlock(&file->uverbs_file->device->lists_mutex);
 
-	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);
 
@@ -549,7 +567,7 @@  struct file *ib_uverbs_alloc_event_file(struct ib_uverbs_file *uverbs_file,
 	struct file *filp;
 	int ret;
 
-	ev_file = kmalloc(sizeof *ev_file, GFP_KERNEL);
+	ev_file = kzalloc(sizeof(*ev_file), GFP_KERNEL);
 	if (!ev_file)
 		return ERR_PTR(-ENOMEM);
 
@@ -577,6 +595,11 @@  struct file *ib_uverbs_alloc_event_file(struct ib_uverbs_file *uverbs_file,
 			goto put_file;
 	}
 
+	mutex_lock(&uverbs_file->device->lists_mutex);
+	list_add_tail(&ev_file->list,
+		      &uverbs_file->device->uverbs_events_file_list);
+	mutex_unlock(&uverbs_file->device->lists_mutex);
+
 	kref_get(&uverbs_file->ref);
 	return filp;
 
@@ -621,12 +644,11 @@  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;
-	struct ib_device *ib_dev = file->device->ib_dev;
+	struct ib_device *ib_dev;
 	struct ib_uverbs_cmd_hdr hdr;
 	__u32 flags;
-
-	if (!ib_dev)
-		return -ENODEV;
+	int srcu_key;
+	ssize_t ret;
 
 	if (count < sizeof hdr)
 		return -EINVAL;
@@ -634,6 +656,14 @@  static ssize_t ib_uverbs_write(struct file *filp, const char __user *buf,
 	if (copy_from_user(&hdr, buf, sizeof hdr))
 		return -EFAULT;
 
+	srcu_key = srcu_read_lock(&file->device->disassociate_srcu);
+	ib_dev = srcu_dereference(file->device->ib_dev,
+				  &file->device->disassociate_srcu);
+	if (!ib_dev) {
+		ret = -EIO;
+		goto out;
+	}
+
 	flags = (hdr.command &
 		 IB_USER_VERBS_CMD_FLAGS_MASK) >> IB_USER_VERBS_CMD_FLAGS_SHIFT;
 
@@ -641,26 +671,36 @@  static ssize_t ib_uverbs_write(struct file *filp, const char __user *buf,
 		__u32 command;
 
 		if (hdr.command & ~(__u32)(IB_USER_VERBS_CMD_FLAGS_MASK |
-					   IB_USER_VERBS_CMD_COMMAND_MASK))
-			return -EINVAL;
+					   IB_USER_VERBS_CMD_COMMAND_MASK)) {
+			ret = -EINVAL;
+			goto out;
+		}
 
 		command = hdr.command & IB_USER_VERBS_CMD_COMMAND_MASK;
 
 		if (command >= ARRAY_SIZE(uverbs_cmd_table) ||
-		    !uverbs_cmd_table[command])
-			return -EINVAL;
+		    !uverbs_cmd_table[command]) {
+			ret = -EINVAL;
+			goto out;
+		}
 
 		if (!file->ucontext &&
-		    command != IB_USER_VERBS_CMD_GET_CONTEXT)
-			return -EINVAL;
+		    command != IB_USER_VERBS_CMD_GET_CONTEXT) {
+			ret = -EINVAL;
+			goto out;
+		}
 
-		if (!(ib_dev->uverbs_cmd_mask & (1ull << command)))
-			return -ENOSYS;
+		if (!(ib_dev->uverbs_cmd_mask & (1ull << command))) {
+			ret = -ENOSYS;
+			goto out;
+		}
 
-		if (hdr.in_words * 4 != count)
-			return -EINVAL;
+		if (hdr.in_words * 4 != count) {
+			ret = -EINVAL;
+			goto out;
+		}
 
-		return uverbs_cmd_table[command](file, ib_dev,
+		ret = uverbs_cmd_table[command](file, ib_dev,
 						 buf + sizeof(hdr),
 						 hdr.in_words * 4,
 						 hdr.out_words * 4);
@@ -671,51 +711,72 @@  static ssize_t ib_uverbs_write(struct file *filp, const char __user *buf,
 		struct ib_uverbs_ex_cmd_hdr ex_hdr;
 		struct ib_udata ucore;
 		struct ib_udata uhw;
-		int err;
 		size_t written_count = count;
 
 		if (hdr.command & ~(__u32)(IB_USER_VERBS_CMD_FLAGS_MASK |
-					   IB_USER_VERBS_CMD_COMMAND_MASK))
-			return -EINVAL;
+					   IB_USER_VERBS_CMD_COMMAND_MASK)) {
+			ret = -EINVAL;
+			goto out;
+		}
 
 		command = hdr.command & IB_USER_VERBS_CMD_COMMAND_MASK;
 
 		if (command >= ARRAY_SIZE(uverbs_ex_cmd_table) ||
-		    !uverbs_ex_cmd_table[command])
-			return -ENOSYS;
+		    !uverbs_ex_cmd_table[command]) {
+			ret = -ENOSYS;
+			goto out;
+		}
 
-		if (!file->ucontext)
-			return -EINVAL;
+		if (!file->ucontext) {
+			ret = -EINVAL;
+			goto out;
+		}
 
-		if (!(ib_dev->uverbs_ex_cmd_mask & (1ull << command)))
-			return -ENOSYS;
+		if (!(ib_dev->uverbs_ex_cmd_mask & (1ull << command))) {
+			ret = -ENOSYS;
+			goto out;
+		}
 
-		if (count < (sizeof(hdr) + sizeof(ex_hdr)))
-			return -EINVAL;
+		if (count < (sizeof(hdr) + sizeof(ex_hdr))) {
+			ret = -EINVAL;
+			goto out;
+		}
 
-		if (copy_from_user(&ex_hdr, buf + sizeof(hdr), sizeof(ex_hdr)))
-			return -EFAULT;
+		if (copy_from_user(&ex_hdr, buf + sizeof(hdr), sizeof(ex_hdr))) {
+			ret = -EFAULT;
+			goto out;
+		}
 
 		count -= sizeof(hdr) + sizeof(ex_hdr);
 		buf += sizeof(hdr) + sizeof(ex_hdr);
 
-		if ((hdr.in_words + ex_hdr.provider_in_words) * 8 != count)
-			return -EINVAL;
+		if ((hdr.in_words + ex_hdr.provider_in_words) * 8 != count) {
+			ret = -EINVAL;
+			goto out;
+		}
 
-		if (ex_hdr.cmd_hdr_reserved)
-			return -EINVAL;
+		if (ex_hdr.cmd_hdr_reserved) {
+			ret = -EINVAL;
+			goto out;
+		}
 
 		if (ex_hdr.response) {
-			if (!hdr.out_words && !ex_hdr.provider_out_words)
-				return -EINVAL;
+			if (!hdr.out_words && !ex_hdr.provider_out_words) {
+				ret = -EINVAL;
+				goto out;
+			}
 
 			if (!access_ok(VERIFY_WRITE,
 				       (void __user *) (unsigned long) ex_hdr.response,
-				       (hdr.out_words + ex_hdr.provider_out_words) * 8))
-				return -EFAULT;
+				       (hdr.out_words + ex_hdr.provider_out_words) * 8)) {
+				ret = -EFAULT;
+				goto out;
+			}
 		} else {
-			if (hdr.out_words || ex_hdr.provider_out_words)
-				return -EINVAL;
+			if (hdr.out_words || ex_hdr.provider_out_words) {
+				ret = -EINVAL;
+				goto out;
+			}
 		}
 
 		INIT_UDATA_BUF_OR_NULL(&ucore, buf, (unsigned long) ex_hdr.response,
@@ -727,29 +788,43 @@  static ssize_t ib_uverbs_write(struct file *filp, const char __user *buf,
 				       ex_hdr.provider_in_words * 8,
 				       ex_hdr.provider_out_words * 8);
 
-		err = uverbs_ex_cmd_table[command](file,
+		ret = uverbs_ex_cmd_table[command](file,
 						   ib_dev,
 						   &ucore,
 						   &uhw);
-
-		if (err)
-			return err;
-
-		return written_count;
+		if (!ret)
+			ret = written_count;
+	} else {
+		ret = -ENOSYS;
 	}
 
-	return -ENOSYS;
+out:
+	srcu_read_unlock(&file->device->disassociate_srcu, srcu_key);
+	return ret;
 }
 
 static int ib_uverbs_mmap(struct file *filp, struct vm_area_struct *vma)
 {
 	struct ib_uverbs_file *file = filp->private_data;
-	struct ib_device *ib_dev = file->device->ib_dev;
+	struct ib_device *ib_dev;
+	int ret = 0;
+	int srcu_key;
 
-	if (!ib_dev || !file->ucontext)
-		return -ENODEV;
+	srcu_key = srcu_read_lock(&file->device->disassociate_srcu);
+	ib_dev = srcu_dereference(file->device->ib_dev,
+				  &file->device->disassociate_srcu);
+	if (!ib_dev) {
+		ret = -EIO;
+		goto out;
+	}
+
+	if (!file->ucontext)
+		ret = -ENODEV;
 	else
-		return ib_dev->mmap(file->ucontext, vma);
+		ret = ib_dev->mmap(file->ucontext, vma);
+out:
+	srcu_read_unlock(&file->device->disassociate_srcu, srcu_key);
+	return ret;
 }
 
 /*
@@ -766,7 +841,10 @@  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)
@@ -774,15 +852,34 @@  static int ib_uverbs_open(struct inode *inode, struct file *filp)
 	else
 		return -ENXIO;
 
-	if (!try_module_get(dev->ib_dev->owner)) {
-		ret = -ENODEV;
+	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);
+	/* In case IB device supports disassociate ucontext, there is no hard
+	 * dependency between uverbs device and its low level device.
+	 */
+	module_dependent = !(dev->flags & UVERBS_FLAG_DISASSOCIATE);
+
+	if (module_dependent) {
+		if (!try_module_get(ib_dev->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;
@@ -792,13 +889,18 @@  static int ib_uverbs_open(struct inode *inode, struct file *filp)
 	mutex_init(&file->mutex);
 
 	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);
 
 	return nonseekable_open(inode, filp);
 
 err_module:
-	module_put(dev->ib_dev->owner);
+	module_put(ib_dev->owner);
 
 err:
+	mutex_unlock(&dev->lists_mutex);
+	srcu_read_unlock(&dev->disassociate_srcu, srcu_key);
 	kref_put(&dev->ref, ib_uverbs_release_dev);
 	return ret;
 }
@@ -806,8 +908,18 @@  err:
 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);
+	struct ib_ucontext *ucontext = NULL;
+
+	mutex_lock(&file->device->lists_mutex);
+	ucontext = file->ucontext;
+	file->ucontext = NULL;
+	if (!file->is_closed) {
+		list_del(&file->list);
+		file->is_closed = 1;
+	}
+	mutex_unlock(&file->device->lists_mutex);
+	if (ucontext)
+		ib_uverbs_cleanup_ucontext(file, ucontext);
 
 	if (file->async_file)
 		kref_put(&file->async_file->ref, ib_uverbs_release_event_file);
@@ -843,12 +955,21 @@  static struct ib_client uverbs_client = {
 static ssize_t show_ibdev(struct device *device, struct device_attribute *attr,
 			  char *buf)
 {
+	int ret = -ENODEV;
+	int srcu_key;
 	struct ib_uverbs_device *dev = dev_get_drvdata(device);
+	struct ib_device *ib_dev;
 
 	if (!dev)
 		return -ENODEV;
 
-	return sprintf(buf, "%s\n", dev->ib_dev->name);
+	srcu_key = srcu_read_lock(&dev->disassociate_srcu);
+	ib_dev = srcu_dereference(dev->ib_dev, &dev->disassociate_srcu);
+	if (ib_dev)
+		ret = sprintf(buf, "%s\n", ib_dev->name);
+	srcu_read_unlock(&dev->disassociate_srcu, srcu_key);
+
+	return ret;
 }
 static DEVICE_ATTR(ibdev, S_IRUGO, show_ibdev, NULL);
 
@@ -856,11 +977,19 @@  static ssize_t show_dev_abi_version(struct device *device,
 				    struct device_attribute *attr, char *buf)
 {
 	struct ib_uverbs_device *dev = dev_get_drvdata(device);
+	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 = sprintf(buf, "%d\n", ib_dev->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);
 
@@ -900,6 +1029,7 @@  static void 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;
@@ -912,6 +1042,13 @@  static void ib_uverbs_add_one(struct ib_device *device)
 	init_completion(&uverbs_dev->comp);
 	uverbs_dev->xrcd_tree = RB_ROOT;
 	mutex_init(&uverbs_dev->xrcd_tree_mutex);
+	mutex_init(&uverbs_dev->lists_mutex);
+	ret = init_srcu_struct(&uverbs_dev->disassociate_srcu);
+	if (ret)
+		goto err_init;
+
+	INIT_LIST_HEAD(&uverbs_dev->uverbs_file_list);
+	INIT_LIST_HEAD(&uverbs_dev->uverbs_events_file_list);
 
 	spin_lock(&map_lock);
 	devnum = find_first_zero_bit(dev_map, IB_UVERBS_MAX_DEVICES);
@@ -953,6 +1090,9 @@  static void ib_uverbs_add_one(struct ib_device *device)
 	if (device_create_file(uverbs_dev->dev, &dev_attr_abi_version))
 		goto err_class;
 
+	if (device->disassociate_ucontext)
+		uverbs_dev->flags |= UVERBS_FLAG_DISASSOCIATE;
+
 	ib_set_client_data(device, &uverbs_client, uverbs_dev);
 
 	return;
@@ -968,15 +1108,82 @@  err_cdev:
 		clear_bit(devnum, overflow_map);
 
 err:
+	cleanup_srcu_struct(&uverbs_dev->disassociate_srcu);
+
+err_init:
 	kref_put(&uverbs_dev->ref, ib_uverbs_release_dev);
 	wait_for_completion(&uverbs_dev->comp);
 	kfree(uverbs_dev);
 	return;
 }
 
+static void ib_uverbs_free_hw_resources(struct ib_uverbs_device *uverbs_dev)
+{
+	struct ib_uverbs_file *file;
+	struct ib_uverbs_event_file *event_file;
+	struct ib_device *ib_dev = uverbs_dev->ib_dev;
+	struct ib_event event;
+
+	uverbs_dev->ib_dev = NULL;
+
+	/* Pending running commands to terminate */
+	synchronize_srcu(&uverbs_dev->disassociate_srcu);
+	event.event = IB_EVENT_DEVICE_FATAL;
+	event.element.port_num = 0;
+	event.device = ib_dev;
+
+	mutex_lock(&uverbs_dev->lists_mutex);
+	while (!list_empty(&uverbs_dev->uverbs_file_list)) {
+		struct ib_ucontext *ucontext;
+
+		file = list_first_entry(&uverbs_dev->uverbs_file_list,
+					struct ib_uverbs_file, list);
+		file->is_closed = 1;
+		ucontext = file->ucontext;
+		list_del(&file->list);
+		file->ucontext = NULL;
+		kref_get(&file->ref);
+		mutex_unlock(&uverbs_dev->lists_mutex);
+		/* We must release the mutex before going ahead and calling
+		 * disassociate_ucontext. disassociate_ucontext might end up
+		 * indirectly calling uverbs_close, for example due to freeing
+		 * the resources (e.g mmput).
+		 */
+		ib_uverbs_event_handler(&file->event_handler, &event);
+		if (ucontext) {
+			ib_dev->disassociate_ucontext(ucontext);
+			ib_uverbs_cleanup_ucontext(file, ucontext);
+		}
+
+		mutex_lock(&uverbs_dev->lists_mutex);
+		kref_put(&file->ref, ib_uverbs_release_file);
+	}
+
+	mutex_unlock(&uverbs_dev->lists_mutex);
+
+	mutex_lock(&uverbs_dev->lists_mutex);
+	while (!list_empty(&uverbs_dev->uverbs_events_file_list)) {
+		event_file = list_first_entry(&uverbs_dev->
+					      uverbs_events_file_list,
+					      struct ib_uverbs_event_file,
+					      list);
+		event_file->is_closed = 1;
+
+		list_del(&event_file->list);
+		if (event_file->is_async)
+			ib_unregister_event_handler(&event_file->uverbs_file->
+						    event_handler);
+
+		wake_up_interruptible(&event_file->poll_wait);
+		kill_fasync(&event_file->async_queue, SIGIO, POLL_IN);
+	}
+	mutex_unlock(&uverbs_dev->lists_mutex);
+}
+
 static void ib_uverbs_remove_one(struct ib_device *device)
 {
 	struct ib_uverbs_device *uverbs_dev = ib_get_client_data(device, &uverbs_client);
+	int wait_clients = 1;
 
 	if (!uverbs_dev)
 		return;
@@ -990,9 +1197,27 @@  static void ib_uverbs_remove_one(struct ib_device *device)
 	else
 		clear_bit(uverbs_dev->devnum - IB_UVERBS_MAX_DEVICES, overflow_map);
 
+	if (uverbs_dev->flags & UVERBS_FLAG_DISASSOCIATE) {
+		/* We disassociate HW resources and immediately returning, not
+		 * pending to active userspace clients. Upon returning ib_device
+		 * may be freed internally and is not valid any more.
+		 * uverbs_device is still available, when all clients close
+		 * their files, the uverbs device ref count will be zero and its
+		 * resources will be freed.
+		 * Note: At that step no more files can be opened on that cdev
+		 * as it was deleted, however active clients can still issue
+		 * commands and close their open files.
+		 */
+		ib_uverbs_free_hw_resources(uverbs_dev);
+		wait_clients = 0;
+	}
+	/* ref count taken as part of add one is put back in both modes. */
 	kref_put(&uverbs_dev->ref, ib_uverbs_release_dev);
-	wait_for_completion(&uverbs_dev->comp);
-	kfree(uverbs_dev);
+	if (wait_clients) {
+		wait_for_completion(&uverbs_dev->comp);
+		cleanup_srcu_struct(&uverbs_dev->disassociate_srcu);
+		kfree(uverbs_dev);
+	}
 }
 
 static char *uverbs_devnode(struct device *dev, umode_t *mode)
diff --git a/include/rdma/ib_verbs.h b/include/rdma/ib_verbs.h
index 7d78794..4645746 100644
--- a/include/rdma/ib_verbs.h
+++ b/include/rdma/ib_verbs.h
@@ -1705,6 +1705,7 @@  struct ib_device {
 	int			   (*destroy_flow)(struct ib_flow *flow_id);
 	int			   (*check_mr_status)(struct ib_mr *mr, u32 check_mask,
 						      struct ib_mr_status *mr_status);
+	void			   (*disassociate_ucontext)(struct ib_ucontext *ibcontext);
 
 	struct ib_dma_mapping_ops   *dma_ops;