[PATCH rdma-next V1 00/13] [PATCH V1 for-next 00/13] IB/core: SG IOCTL based RDMA ABI

[PATCH rdma-next V1 00/13] [PATCH V1 for-next 00/13] IB/core: SG IOCTL based RDMA ABI

[Matan Barak]

Hi Doug,

This patch series adds ioctl() interface to the existing write() interface and
provide an easy route to backport this change to legacy supported systems.
Analyzing the current uverbs role in dispatching and parsing commands, we find
that:
(a) uverbs validates the basic properties of the command.
(b) uverbs is responsible of doing all the IDR and uobject management and
    locking. It's also responsible for handling completion FDs.
(c) uverbs transforms the user<-->kernel ABI to kernel API.

(a) and (b) are valid for every kABI. Although the nature of commands could
change, they still have to be validated and transform to kernel pointers.
In order to avoid duplications between the various drivers, we would like to
keep (a) and (b) as shared code.

In addition, this is a good time to expand the ABI to be more scalable, so we
added a few goals:
(1) Command's attributes shall be extensible in an easy one. Either by allowing
    drivers to have their own extensible set of attributes or core code
    extensible attributes.
(2) Each driver may have specific type system (i.e QP, CQ, ....). It could extend
    this type system in the future. Try to avoid duplicating existing types or
    actions.

Thus, in order to allow this flexibility, we decide giving (a) and (b) as a
common infrastructure, but use per-driver guidelines in order to do that
parsing and uobject management. Handlers are also set by the drivers
themselves (though they can point to either shared common code) or
driver specific code.

We introduce a hierarchal object-method-attributes structure. Adding an
entity to this hierarchy doesn't affect the rest of the interface.
Such a hierarchy could be rooted in a specific device and describes both the
common features and features which are unique to this specific device.
This hierarchy is actually a per-device parsing tree, composed of three
layers - objects, actions and attributes. Each such layer contains two
namespaces - common entities and hardware specific entities. This way, a
device could add hardware specific actions to a common object, it could
add hardware specific objects, etc. Abstractions which really make sense,
should go to the common section. This means that we still need to be able to
pass optional parameters. In order to enable optional parameters, each command
is composed of a header and a bunch of TLVs to pass the attributes of this
command. The supported attribute classes are:
* PTR_IN (command) [in case of a small buffer, we could pass the data inlined]
* PTR_OUT (response)
* IDR_OBJECT
* FD_OBJECT
We differentiate between blobs and objects in order to allow a generic piece of
code in the kernel to do some syntactic validations and translate the given
user object id to a kernel structure. This could really help in sharing code
between different handlers.

Scatter gather was chosen in order to allow us not to recompile user space
drivers. By using pointers to driver specific data, we could just use it
without introduce copying data and without changing the user-space driver at
all.

We elevate the locking and IDR changes accepted to linux-rdma in this series.
Since types are no longer enforced by the common infrastructure, there is no
point of pre-allocating common IDR types in the common code. Instead, we
provide an API for driver to add new types. We use one IDR per context
for all its IDR types. The driver declared all its supported objects, their
free function and release order. After that, all uboject, exclusive access
and objects are handled automatically for the driver by the infrastructure.

When putting the pieces together, we have per-device parsing tree, that actually
describes all the objects, methods and attributes a device supports by using a
descriptive language. A command is given by the user-space, as a header plus an
array of Type-Length-Pointer/Object attributes. The ioctl callback executes a
generic code that shares as much logic between the various verbs handlers as
possible. This generic code gets the command input from the user-space and by
reading the device's parsing tree, it could syntactically validate it, grab all
required objects, lock them, call the right handler and then
commit/unlock/rollback the result, depending on the handler's result. Having
such a flexible extensible mechanism, that allows introducing new common and
hardware-specific to existing common attributes, but also allows adding new
hardware-specific entities, enhances the support for device diversity quite
vastly.

The developer of such verbs isn't aware of namespacing, but a shared
code transforms the hierarchy of objects-methods-attributes to a
hash, which its different buckets corresponds to namespaces and its
lower part of the id correspond to the entity within the bucket.

This series lays the foundations of such an infrastructure. It demonstrate the
CREATE_CQ and DESTROY_CQ handlers that use this new infrastructure with most of
its features. We still use a common parsing tree for all devices, but this will
be changed in the future, when we introduce a parse tree that is tailored for
devices. We still don't share code between the regular write() interface handlers
and the new ioctl(), but it should be done in the future too. An enhanced query
mechanism, based on the parse tree, is planned in the future as well.

After this infrastructure is merged, we could continue enhancing and build the
require functionality on top of it.

This is rebased over Doug's latest k.o/for-next branch.

The series (with some extra untested/un-reviewed extra handlers) is available
here:
https://github.com/matanb10/linux         branch: kabi_ioctl_v1

rdma-core code for testing:
https://github.com/matanb10/rdma-core     branch: kabi_v1_ioctl_testing

Regards,
Matan

Changes from V0:
1. Change names: group->namespace, type->object, action->method
2. Hide the namespace mechanism from the verb developer and
   initialize it at runtime.
3. Macros now support old compilers.

Matan Barak (13):
  IB/core: Add a generic way to execute an operation on a uobject
  IB/core: Add support to finalize objects in one transaction
  IB/core: Add new ioctl interface
  IB/core: Declare an object instead of declaring only type attributes
  IB/core: Add DEVICE object and root tree structure
  IB/core: Add uverbs merge trees functionality
  IB/core: Add macros for declaring methods and attributes
  IB/core: Explicitly destroy an object while keeping uobject
  IB/core: Export ioctl enum types to user-space
  IB/core: Add legacy driver's user-data
  IB/core: Add completion queue (cq) object actions
  IB/core: Assign root to all drivers
  IB/core: Expose ioctl interface through experimental Kconfig

 drivers/infiniband/Kconfig                   |   9 +
 drivers/infiniband/core/Makefile             |   3 +-
 drivers/infiniband/core/rdma_core.c          | 179 +++++++
 drivers/infiniband/core/rdma_core.h          |  42 ++
 drivers/infiniband/core/uverbs.h             |   3 +
 drivers/infiniband/core/uverbs_ioctl.c       | 364 +++++++++++++++
 drivers/infiniband/core/uverbs_ioctl_merge.c | 665 +++++++++++++++++++++++++++
 drivers/infiniband/core/uverbs_main.c        |  24 +
 drivers/infiniband/core/uverbs_std_types.c   | 281 ++++++++---
 include/rdma/ib_verbs.h                      |   2 +
 include/rdma/uverbs_ioctl.h                  | 438 ++++++++++++++++++
 include/rdma/uverbs_std_types.h              |  58 ++-
 include/rdma/uverbs_types.h                  |  39 +-
 include/uapi/rdma/ib_user_ioctl_verbs.h      |  84 ++++
 include/uapi/rdma/rdma_user_ioctl.h          |  33 ++
 15 files changed, 2135 insertions(+), 89 deletions(-)
 create mode 100644 drivers/infiniband/core/uverbs_ioctl.c
 create mode 100644 drivers/infiniband/core/uverbs_ioctl_merge.c
 create mode 100644 include/rdma/uverbs_ioctl.h
 create mode 100644 include/uapi/rdma/ib_user_ioctl_verbs.h

-- 
1.8.3.1

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH rdma-next V1 01/13] IB/core: Add a generic way to execute an operation on a uobject

[Matan Barak]

The ioctl infrastructure treats all user-objects in the same manner.
It gets objects ids from the user-space and by using the object type
and type attributes mentioned in the object specification, it executes
this required method. Passing an object id from the user-space as
an attribute is carried out in three stages. The first is carried out
before the actual handler and the last is carried out afterwards.

The different supported operations are read, write, destroy and create.
In the first stage, the former three actions just fetches the object
from the repository (by using its id) and locks it. The last action
allocates a new uobject. Afterwards, the second stage is carried out
when the handler itself carries out the required modification of the
object. The last stage is carried out after the handler finishes and
commits the result. The former two operations just unlock the object.
Destroy calls the "free object" operation, taking into account the
object's type and releases the uobject as well. Creation just adds the
new uobject to the repository, making the object visible to the
application.

In order to abstract these details from the ioctl infrastructure
layer, we add uverbs_get_uobject_from_context and
uverbs_finalize_object functions which corresponds to the first
and last stages respectively.

Signed-off-by: Matan Barak <matanb-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
Reviewed-by: Yishai Hadas <yishaih-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
---
 drivers/infiniband/core/rdma_core.c | 58 +++++++++++++++++++++++++++++++++++++
 drivers/infiniband/core/rdma_core.h | 17 +++++++++++
 include/rdma/uverbs_ioctl.h         | 52 +++++++++++++++++++++++++++++++++
 3 files changed, 127 insertions(+)
 create mode 100644 include/rdma/uverbs_ioctl.h

diff --git a/drivers/infiniband/core/rdma_core.c b/drivers/infiniband/core/rdma_core.c
index 41c31a2..2bd58ff 100644
--- a/drivers/infiniband/core/rdma_core.c
+++ b/drivers/infiniband/core/rdma_core.c
@@ -35,6 +35,7 @@
 #include <rdma/ib_verbs.h>
 #include <rdma/uverbs_types.h>
 #include <linux/rcupdate.h>
+#include <rdma/uverbs_ioctl.h>
 #include "uverbs.h"
 #include "core_priv.h"
 #include "rdma_core.h"
@@ -625,3 +626,60 @@ void uverbs_initialize_ucontext(struct ib_ucontext *ucontext)
 	.needs_kfree_rcu = false,
 };
 
+struct ib_uobject *uverbs_get_uobject_from_context(const struct uverbs_obj_type *type_attrs,
+						   struct ib_ucontext *ucontext,
+						   enum uverbs_obj_access access,
+						   int id)
+{
+	switch (access) {
+	case UVERBS_ACCESS_READ:
+		return rdma_lookup_get_uobject(type_attrs, ucontext, id, false);
+	case UVERBS_ACCESS_DESTROY:
+	case UVERBS_ACCESS_WRITE:
+		return rdma_lookup_get_uobject(type_attrs, ucontext, id, true);
+	case UVERBS_ACCESS_NEW:
+		return rdma_alloc_begin_uobject(type_attrs, ucontext);
+	default:
+		WARN_ON(true);
+		return ERR_PTR(-EOPNOTSUPP);
+	}
+}
+
+int uverbs_finalize_object(struct ib_uobject *uobj,
+			   enum uverbs_obj_access access,
+			   bool commit)
+{
+	int ret = 0;
+
+	/*
+	 * refcounts should be handled at the object level and not at the
+	 * uobject level. Refcounts of the objects themselves are done in
+	 * handlers.
+	 */
+
+	switch (access) {
+	case UVERBS_ACCESS_READ:
+		rdma_lookup_put_uobject(uobj, false);
+		break;
+	case UVERBS_ACCESS_WRITE:
+		rdma_lookup_put_uobject(uobj, true);
+		break;
+	case UVERBS_ACCESS_DESTROY:
+		if (commit)
+			ret = rdma_remove_commit_uobject(uobj);
+		else
+			rdma_lookup_put_uobject(uobj, true);
+		break;
+	case UVERBS_ACCESS_NEW:
+		if (commit)
+			ret = rdma_alloc_commit_uobject(uobj);
+		else
+			rdma_alloc_abort_uobject(uobj);
+		break;
+	default:
+		WARN_ON(true);
+		ret = -EOPNOTSUPP;
+	}
+
+	return ret;
+}
diff --git a/drivers/infiniband/core/rdma_core.h b/drivers/infiniband/core/rdma_core.h
index 1b82e7f..97483d1 100644
--- a/drivers/infiniband/core/rdma_core.h
+++ b/drivers/infiniband/core/rdma_core.h
@@ -39,6 +39,7 @@
 
 #include <linux/idr.h>
 #include <rdma/uverbs_types.h>
+#include <rdma/uverbs_ioctl.h>
 #include <rdma/ib_verbs.h>
 #include <linux/mutex.h>
 
@@ -75,4 +76,20 @@
  */
 void uverbs_close_fd(struct file *f);
 
+/*
+ * Get an ib_uobject that corresponds to the given id from ucontext, assuming
+ * the object is from the given type. Lock it to the required access when
+ * applicable.
+ * This function could create (access == NEW), destroy (access == DESTROY)
+ * or unlock (access == READ || access == WRITE) objects if required.
+ * The action will be finalized only when uverbs_finalize_object is called.
+ */
+struct ib_uobject *uverbs_get_uobject_from_context(const struct uverbs_obj_type *type_attrs,
+						   struct ib_ucontext *ucontext,
+						   enum uverbs_obj_access access,
+						   int id);
+int uverbs_finalize_object(struct ib_uobject *uobj,
+			   enum uverbs_obj_access access,
+			   bool commit);
+
 #endif /* RDMA_CORE_H */
diff --git a/include/rdma/uverbs_ioctl.h b/include/rdma/uverbs_ioctl.h
new file mode 100644
index 0000000..6885b92
--- /dev/null
+++ b/include/rdma/uverbs_ioctl.h
@@ -0,0 +1,52 @@
+/*
+ * Copyright (c) 2017, Mellanox Technologies inc.  All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#ifndef _UVERBS_IOCTL_
+#define _UVERBS_IOCTL_
+
+#include <rdma/uverbs_types.h>
+
+/*
+ * =======================================
+ *	Verbs action specifications
+ * =======================================
+ */
+
+enum uverbs_obj_access {
+	UVERBS_ACCESS_READ,
+	UVERBS_ACCESS_WRITE,
+	UVERBS_ACCESS_NEW,
+	UVERBS_ACCESS_DESTROY
+};
+
+#endif
+
-- 
1.8.3.1

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH rdma-next V1 02/13] IB/core: Add support to finalize objects in one transaction

[Matan Barak]

The new ioctl based infrastructure either commits or rollbacks
all objects of the method as one transaction. In order to do
that, we introduce a notion of dealing with a collection of
objects that are related to a specific method.

This also requires adding a notion of a method and attribute.
A method contains a hash of attributes, where each bucket
contains several attributes. The attributes are hashed according
to their namespace which resides in the four upper bits of the id.

For example, an object could be a CQ, which has an action of CREATE_CQ.
This action has multiple attributes. For example, the CQ's new handle
and the comp_channel. Each layer in this hierarchy - objects, methods
and attributes is split into namespaces. The basic example for that is
one namespace representing the default entities and another one
representing the driver specific entities.

When declaring these methods and attributes, we actually declare
their specifications. When a method is executed, we actually
allocates some space to hold auxiliary information. This auxiliary
information contains meta-data about the required objects, such
as pointers to their type information, pointers to the uobjects
themselves (if exist), etc.
The specification, along with the auxiliary information we allocated
and filled is given to the finalize_objects function.

Signed-off-by: Matan Barak <matanb-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
Reviewed-by: Yishai Hadas <yishaih-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
---
 drivers/infiniband/core/rdma_core.c | 40 ++++++++++++++++++++++++++++
 drivers/infiniband/core/rdma_core.h | 22 ++++++++++++++-
 include/rdma/uverbs_ioctl.h         | 53 +++++++++++++++++++++++++++++++++++++
 3 files changed, 114 insertions(+), 1 deletion(-)

diff --git a/drivers/infiniband/core/rdma_core.c b/drivers/infiniband/core/rdma_core.c
index 2bd58ff..0fe8ef9 100644
--- a/drivers/infiniband/core/rdma_core.c
+++ b/drivers/infiniband/core/rdma_core.c
@@ -683,3 +683,43 @@ int uverbs_finalize_object(struct ib_uobject *uobj,
 
 	return ret;
 }
+
+int uverbs_finalize_objects(struct uverbs_attr_bundle *attrs_bundle,
+			    struct uverbs_attr_spec_hash * const *spec_hash,
+			    size_t num,
+			    bool commit)
+{
+	unsigned int i;
+	int ret = 0;
+
+	for (i = 0; i < num; i++) {
+		struct uverbs_attr_bundle_hash *curr_bundle =
+			&attrs_bundle->hash[i];
+		const struct uverbs_attr_spec_hash *curr_spec_bucket =
+			spec_hash[i];
+		unsigned int j;
+
+		for (j = 0; j < curr_bundle->num_attrs; j++) {
+			struct uverbs_attr *attr;
+			const struct uverbs_attr_spec *spec;
+
+			if (!uverbs_attr_is_valid_in_hash(curr_bundle, j))
+				continue;
+
+			attr = &curr_bundle->attrs[j];
+			spec = &curr_spec_bucket->attrs[j];
+
+			if (spec->type == UVERBS_ATTR_TYPE_IDR ||
+			    spec->type == UVERBS_ATTR_TYPE_FD) {
+				int current_ret;
+
+				current_ret = uverbs_finalize_object(attr->obj_attr.uobject,
+								     spec->obj.access,
+								     commit);
+				if (!ret)
+					ret = current_ret;
+			}
+		}
+	}
+	return ret;
+}
diff --git a/drivers/infiniband/core/rdma_core.h b/drivers/infiniband/core/rdma_core.h
index 97483d1..9ed6ad0 100644
--- a/drivers/infiniband/core/rdma_core.h
+++ b/drivers/infiniband/core/rdma_core.h
@@ -82,7 +82,8 @@
  * applicable.
  * This function could create (access == NEW), destroy (access == DESTROY)
  * or unlock (access == READ || access == WRITE) objects if required.
- * The action will be finalized only when uverbs_finalize_object is called.
+ * The action will be finalized only when uverbs_finalize_object or
+ * uverbs_finalize_objects are called.
  */
 struct ib_uobject *uverbs_get_uobject_from_context(const struct uverbs_obj_type *type_attrs,
 						   struct ib_ucontext *ucontext,
@@ -91,5 +92,24 @@ struct ib_uobject *uverbs_get_uobject_from_context(const struct uverbs_obj_type
 int uverbs_finalize_object(struct ib_uobject *uobj,
 			   enum uverbs_obj_access access,
 			   bool commit);
+/*
+ * Note that certain finalize stages could return a status:
+ *   (a) alloc_commit could return a failure if the object is committed at the
+ *       same time when the context is destroyed.
+ *   (b) remove_commit could fail if the object wasn't destroyed successfully.
+ * Since multiple objects could be finalized in one transaction, it is very NOT
+ * recommended to have several finalize actions which have side effects.
+ * For example, it's NOT recommended to have a certain action which has both
+ * a commit action and a destroy action or two destroy objects in the same
+ * action. The rule of thumb is to have one destroy or commit action with
+ * multiple lookups.
+ * The first non zero return value of finalize_object is returned from this
+ * function. For example, this could happen when we couldn't destroy an
+ * object.
+ */
+int uverbs_finalize_objects(struct uverbs_attr_bundle *attrs_bundle,
+			    struct uverbs_attr_spec_hash * const *spec_hash,
+			    size_t num,
+			    bool commit);
 
 #endif /* RDMA_CORE_H */
diff --git a/include/rdma/uverbs_ioctl.h b/include/rdma/uverbs_ioctl.h
index 6885b92..d3ec02b 100644
--- a/include/rdma/uverbs_ioctl.h
+++ b/include/rdma/uverbs_ioctl.h
@@ -41,6 +41,12 @@
  * =======================================
  */
 
+enum uverbs_attr_type {
+	UVERBS_ATTR_TYPE_NA,
+	UVERBS_ATTR_TYPE_IDR,
+	UVERBS_ATTR_TYPE_FD,
+};
+
 enum uverbs_obj_access {
 	UVERBS_ACCESS_READ,
 	UVERBS_ACCESS_WRITE,
@@ -48,5 +54,52 @@ enum uverbs_obj_access {
 	UVERBS_ACCESS_DESTROY
 };
 
+struct uverbs_attr_spec {
+	enum uverbs_attr_type		type;
+	struct {
+		/*
+		 * higher bits mean the namespace and lower bits mean
+		 * the type id within the namespace.
+		 */
+		u16			obj_type;
+		u8			access;
+	} obj;
+};
+
+struct uverbs_attr_spec_hash {
+	size_t				num_attrs;
+	struct uverbs_attr_spec		attrs[0];
+};
+
+struct uverbs_obj_attr {
+	struct ib_uobject		*uobject;
+};
+
+struct uverbs_attr {
+	struct uverbs_obj_attr	obj_attr;
+};
+
+struct uverbs_attr_bundle_hash {
+	/* if bit i is set, it means attrs[i] contains valid information */
+	unsigned long *valid_bitmap;
+	size_t num_attrs;
+	/*
+	 * arrays of attributes, each element corresponds to the specification
+	 * of the attribute in the same index.
+	 */
+	struct uverbs_attr *attrs;
+};
+
+struct uverbs_attr_bundle {
+	size_t				num_buckets;
+	struct uverbs_attr_bundle_hash  hash[];
+};
+
+static inline bool uverbs_attr_is_valid_in_hash(const struct uverbs_attr_bundle_hash *attrs_hash,
+						unsigned int idx)
+{
+	return test_bit(idx, attrs_hash->valid_bitmap);
+}
+
 #endif
 
-- 
1.8.3.1

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH rdma-next V1 03/13] IB/core: Add new ioctl interface

[Matan Barak]

In this ioctl interface, processing the command starts from
properties of the command and fetching the appropriate user objects
before calling the handler.

Parsing and validation is done according to a specifier declared by
the driver's code. In the driver, all supported objects are declared.
These objects are separated to different object namepsaces. Dividing
objects to namespaces is done at initialization by using the higher
bits of the object ids. This initialization can mix objects declared
in different places to one parsing tree using in this ioctl interface.

For each object we list all supported methods. Similarly to objects,
methods are separated to method namespaces too. Namespacing is done
similarly to the objects case. This could be used in order to add
methods to an existing object.

Each method has a specific handler, which could be either a default
handler or a driver specific handler.
Along with the handler, a bunch of attributes are specified as well.
Similarly to objects and method, attributes are namespaced and hashed
by their ids at initialization too. All supported attributes are
subject to automatic fetching and validation. These attributes include
the command, response and the method's related objects' ids.

When these entities (objects, methods and attributes) are used, the
high bits of the entities ids are used in order to calculate the hash
bucket index. Then, these high bits are masked out in order to have a
zero based index. Since we use these high bits for both bucketing and
namespacing, we get a compact representation and O(1) array access.
This is mandatory for efficient dispatching.

Each attribute has a type (PTR_IN, PTR_OUT, IDR and FD) and a length.
Attributes could be validated through some attributes, like:
(*) Minimum size / Exact size
(*) Fops for FD
(*) Object type for IDR

If an IDR/fd attribute is specified, the kernel also states the object
type and the required access (NEW, WRITE, READ or DESTROY).
All uobject/fd management is done automatically by the infrastructure,
meaning - the infrastructure will fail concurrent commands that at
least one of them requires concurrent access (WRITE/DESTROY),
synchronize actions with device removals (dissociate context events)
and take care of reference counting (increase/decrease) for concurrent
actions invocation. The reference counts on the actual kernel objects
shall be handled by the handlers.

 objects
+--------+
|        |
|        |   methods                                                                +--------+
|        |   ns         method      method_spec                           +-----+   |len     |
+--------+  +------+[d]+-------+   +----------------+[d]+------------+    |attr1+-> |type    |
| object +> |method+-> | spec  +-> +  attr_buckets  +-> |default_chain+--> +-----+   |idr_type|
+--------+  +------+   |handler|   |                |   +------------+    |attr2|   |access  |
|        |  |      |   +-------+   +----------------+   |driver chain|    +-----+   +--------+
|        |  |      |                                    +------------+
|        |  +------+
|        |
|        |
|        |
|        |
|        |
|        |
|        |
|        |
|        |
|        |
+--------+

[d] = Hash ids to groups using the high order bits

The right types table is also chosen by using the high bits from
the ids. Currently we have either default or driver specific groups.

Once validation and object fetching (or creation) completed, we call
the handler:
int (*handler)(struct ib_device *ib_dev, struct ib_uverbs_file *ufile,
               struct uverbs_attr_bundle *ctx);

ctx bundles attributes of different namespaces. Each element there
is an array of attributes which corresponds to one namespaces of
attributes. For example, in the usually used case:

 ctx                               core
+----------------------------+     +------------+
| core:                      +---> | valid      |
+----------------------------+     | cmd_attr   |
| driver:                    |     +------------+
|----------------------------+--+  | valid      |
                                |  | cmd_attr   |
                                |  +------------+
                                |  | valid      |
                                |  | obj_attr   |
                                |  +------------+
                                |
                                |  drivers
                                |  +------------+
                                +> | valid      |
                                   | cmd_attr   |
                                   +------------+
                                   | valid      |
                                   | cmd_attr   |
                                   +------------+
                                   | valid      |
                                   | obj_attr   |
                                   +------------+

Signed-off-by: Matan Barak <matanb-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
Reviewed-by: Yishai Hadas <yishaih-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
---
 drivers/infiniband/core/Makefile       |   2 +-
 drivers/infiniband/core/rdma_core.c    |  46 +++++
 drivers/infiniband/core/rdma_core.h    |   5 +
 drivers/infiniband/core/uverbs_ioctl.c | 364 +++++++++++++++++++++++++++++++++
 include/rdma/ib_verbs.h                |   2 +
 include/rdma/uverbs_ioctl.h            | 101 ++++++++-
 include/uapi/rdma/rdma_user_ioctl.h    |  33 +++
 7 files changed, 543 insertions(+), 10 deletions(-)
 create mode 100644 drivers/infiniband/core/uverbs_ioctl.c

diff --git a/drivers/infiniband/core/Makefile b/drivers/infiniband/core/Makefile
index e3cdaff..ca6d067 100644
--- a/drivers/infiniband/core/Makefile
+++ b/drivers/infiniband/core/Makefile
@@ -31,4 +31,4 @@ ib_umad-y :=			user_mad.o
 ib_ucm-y :=			ucm.o
 
 ib_uverbs-y :=			uverbs_main.o uverbs_cmd.o uverbs_marshall.o \
-				rdma_core.o uverbs_std_types.o
+				rdma_core.o uverbs_std_types.o uverbs_ioctl.o
diff --git a/drivers/infiniband/core/rdma_core.c b/drivers/infiniband/core/rdma_core.c
index 0fe8ef9..2a2f002 100644
--- a/drivers/infiniband/core/rdma_core.c
+++ b/drivers/infiniband/core/rdma_core.c
@@ -36,10 +36,56 @@
 #include <rdma/uverbs_types.h>
 #include <linux/rcupdate.h>
 #include <rdma/uverbs_ioctl.h>
+#include <rdma/rdma_user_ioctl.h>
 #include "uverbs.h"
 #include "core_priv.h"
 #include "rdma_core.h"
 
+int uverbs_ns_idx(u16 *id, unsigned int ns_count)
+{
+	int ret = (*id & UVERBS_ID_NS_MASK) >> UVERBS_ID_NS_SHIFT;
+
+	if (ret >= ns_count)
+		return -EINVAL;
+
+	*id &= ~UVERBS_ID_NS_MASK;
+	return ret;
+}
+
+const struct uverbs_object_spec *uverbs_get_object(const struct ib_device *ibdev,
+						   uint16_t object)
+{
+	const struct uverbs_root_spec *object_hash = ibdev->specs_root;
+	const struct uverbs_object_spec_hash *objects;
+	int ret = uverbs_ns_idx(&object, object_hash->num_buckets);
+
+	if (ret < 0)
+		return NULL;
+
+	objects = object_hash->object_buckets[ret];
+
+	if (object >= objects->num_objects)
+		return NULL;
+
+	return objects->objects[object];
+}
+
+const struct uverbs_method_spec *uverbs_get_method(const struct uverbs_object_spec *object,
+						   uint16_t method)
+{
+	const struct uverbs_method_spec_hash *methods;
+	int ret = uverbs_ns_idx(&method, object->num_buckets);
+
+	if (ret < 0)
+		return NULL;
+
+	methods = object->method_buckets[ret];
+	if (method >= methods->num_methods)
+		return NULL;
+
+	return methods->methods[method];
+}
+
 void uverbs_uobject_get(struct ib_uobject *uobject)
 {
 	kref_get(&uobject->ref);
diff --git a/drivers/infiniband/core/rdma_core.h b/drivers/infiniband/core/rdma_core.h
index 9ed6ad0..1efcf93 100644
--- a/drivers/infiniband/core/rdma_core.h
+++ b/drivers/infiniband/core/rdma_core.h
@@ -43,6 +43,11 @@
 #include <rdma/ib_verbs.h>
 #include <linux/mutex.h>
 
+int uverbs_ns_idx(u16 *id, unsigned int ns_count);
+const struct uverbs_object_spec *uverbs_get_object(const struct ib_device *ibdev,
+						   uint16_t object);
+const struct uverbs_method_spec *uverbs_get_method(const struct uverbs_object_spec *object,
+						   uint16_t method);
 /*
  * These functions initialize the context and cleanups its uobjects.
  * The context has a list of objects which is protected by a mutex
diff --git a/drivers/infiniband/core/uverbs_ioctl.c b/drivers/infiniband/core/uverbs_ioctl.c
new file mode 100644
index 0000000..5286ad5
--- /dev/null
+++ b/drivers/infiniband/core/uverbs_ioctl.c
@@ -0,0 +1,364 @@
+/*
+ * Copyright (c) 2017, Mellanox Technologies inc.  All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <rdma/rdma_user_ioctl.h>
+#include <rdma/uverbs_ioctl.h>
+#include "rdma_core.h"
+#include "uverbs.h"
+
+static int uverbs_process_attr(struct ib_device *ibdev,
+			       struct ib_ucontext *ucontext,
+			       const struct ib_uverbs_attr *uattr,
+			       u16 attr_id,
+			       const struct uverbs_attr_spec_hash *attr_spec_bucket,
+			       struct uverbs_attr_bundle_hash *attr_bundle_h,
+			       struct ib_uverbs_attr __user *uattr_ptr)
+{
+	const struct uverbs_attr_spec *spec;
+	struct uverbs_attr *e;
+	const struct uverbs_object_spec *object;
+	struct uverbs_obj_attr *o_attr;
+	struct uverbs_attr *elements = attr_bundle_h->attrs;
+
+	if (uattr->reserved)
+		return -EINVAL;
+
+	if (attr_id >= attr_spec_bucket->num_attrs) {
+		if (uattr->flags & UVERBS_ATTR_F_MANDATORY)
+			return -EINVAL;
+		else
+			return 0;
+	}
+
+	spec = &attr_spec_bucket->attrs[attr_id];
+	e = &elements[attr_id];
+	e->uattr = uattr_ptr;
+
+	switch (spec->type) {
+	case UVERBS_ATTR_TYPE_PTR_IN:
+	case UVERBS_ATTR_TYPE_PTR_OUT:
+		if (uattr->len < spec->len ||
+		    (!(spec->flags & UVERBS_ATTR_SPEC_F_MIN_SZ) &&
+		     uattr->len > spec->len))
+			return -EINVAL;
+
+		e->ptr_attr.data = uattr->data;
+		e->ptr_attr.len = uattr->len;
+		e->ptr_attr.flags = uattr->flags;
+		break;
+
+	case UVERBS_ATTR_TYPE_IDR:
+		if (uattr->data >> 32)
+			return -EINVAL;
+	/* fall through */
+	case UVERBS_ATTR_TYPE_FD:
+		if (uattr->len != 0 || !ucontext || uattr->data > INT_MAX)
+			return -EINVAL;
+
+		o_attr = &e->obj_attr;
+		object = uverbs_get_object(ibdev, spec->obj.obj_type);
+		if (!object)
+			return -EINVAL;
+		o_attr->type = object->type_attrs;
+
+		o_attr->id = (int)uattr->data;
+		o_attr->uobject = uverbs_get_uobject_from_context(
+					o_attr->type,
+					ucontext,
+					spec->obj.access,
+					o_attr->id);
+
+		if (IS_ERR(o_attr->uobject))
+			return PTR_ERR(o_attr->uobject);
+
+		if (spec->obj.access == UVERBS_ACCESS_NEW) {
+			u64 id = o_attr->uobject->id;
+
+			/* Copy the allocated id to the user-space */
+			if (put_user(id, &e->uattr->data)) {
+				uverbs_finalize_object(o_attr->uobject,
+						       UVERBS_ACCESS_NEW,
+						       false);
+				return -EFAULT;
+			}
+		}
+
+		break;
+	default:
+		return -EOPNOTSUPP;
+	}
+
+	set_bit(attr_id, attr_bundle_h->valid_bitmap);
+	return 0;
+}
+
+static int uverbs_uattrs_process(struct ib_device *ibdev,
+				 struct ib_ucontext *ucontext,
+				 const struct ib_uverbs_attr *uattrs,
+				 size_t num_uattrs,
+				 const struct uverbs_method_spec *method,
+				 struct uverbs_attr_bundle *attr_bundle,
+				 struct ib_uverbs_attr __user *uattr_ptr)
+{
+	size_t i;
+	int ret = 0;
+	int num_given_buckets = 0;
+
+	for (i = 0; i < num_uattrs; i++) {
+		const struct ib_uverbs_attr *uattr = &uattrs[i];
+		u16 attr_id = uattr->attr_id;
+		struct uverbs_attr_spec_hash *attr_spec_bucket;
+
+		ret = uverbs_ns_idx(&attr_id, method->num_buckets);
+		if (ret < 0) {
+			if (uattr->flags & UVERBS_ATTR_F_MANDATORY) {
+				uverbs_finalize_objects(attr_bundle,
+							method->attr_buckets,
+							num_given_buckets,
+							false);
+				return ret;
+			}
+			continue;
+		}
+
+		/*
+		 * ret is the found ns, so increase num_given_buckets if
+		 * necessary.
+		 */
+		if (ret >= num_given_buckets)
+			num_given_buckets = ret + 1;
+
+		attr_spec_bucket = method->attr_buckets[ret];
+		ret = uverbs_process_attr(ibdev, ucontext, uattr, attr_id,
+					  attr_spec_bucket, &attr_bundle->hash[ret],
+					  uattr_ptr++);
+		if (ret) {
+			uverbs_finalize_objects(attr_bundle,
+						method->attr_buckets,
+						num_given_buckets,
+						false);
+			return ret;
+		}
+	}
+
+	return num_given_buckets;
+}
+
+static int uverbs_validate_kernel_mandatory(const struct uverbs_method_spec *method_spec,
+					    struct uverbs_attr_bundle *attr_bundle)
+{
+	unsigned int i;
+
+	for (i = 0; i < attr_bundle->num_buckets; i++) {
+		struct uverbs_attr_spec_hash *attr_spec_bucket =
+			method_spec->attr_buckets[i];
+
+		if (!bitmap_subset(attr_spec_bucket->mandatory_attrs_bitmask,
+				   attr_bundle->hash[i].valid_bitmap,
+				   attr_spec_bucket->num_attrs))
+			return -EINVAL;
+	}
+
+	return 0;
+}
+
+static int uverbs_handle_method(struct ib_uverbs_attr __user *uattr_ptr,
+				const struct ib_uverbs_attr *uattrs,
+				size_t num_uattrs,
+				struct ib_device *ibdev,
+				struct ib_uverbs_file *ufile,
+				const struct uverbs_method_spec *method_spec,
+				struct uverbs_attr_bundle *attr_bundle)
+{
+	int ret;
+	int finalize_ret;
+	int num_given_buckets;
+
+	num_given_buckets = uverbs_uattrs_process(ibdev, ufile->ucontext, uattrs,
+						  num_uattrs, method_spec,
+						  attr_bundle, uattr_ptr);
+	if (num_given_buckets <= 0)
+		return -EINVAL;
+
+	attr_bundle->num_buckets = num_given_buckets;
+	ret = uverbs_validate_kernel_mandatory(method_spec, attr_bundle);
+	if (ret)
+		goto cleanup;
+
+	ret = method_spec->handler(ibdev, ufile, attr_bundle);
+cleanup:
+	finalize_ret = uverbs_finalize_objects(attr_bundle,
+					       method_spec->attr_buckets,
+					       attr_bundle->num_buckets,
+					       !ret);
+
+	return ret ? ret : finalize_ret;
+}
+
+#define UVERBS_OPTIMIZE_USING_STACK_SZ  256
+static long ib_uverbs_cmd_verbs(struct ib_device *ib_dev,
+				struct ib_uverbs_file *file,
+				struct ib_uverbs_ioctl_hdr *hdr,
+				void __user *buf)
+{
+	const struct uverbs_object_spec *object_spec;
+	const struct uverbs_method_spec *method_spec;
+	long err = 0;
+	unsigned int i;
+	struct {
+		struct ib_uverbs_attr		*uattrs;
+		struct uverbs_attr_bundle	*uverbs_attr_bundle;
+	} *ctx = NULL;
+	struct uverbs_attr *curr_attr;
+	unsigned long *curr_bitmap;
+	size_t ctx_size;
+#ifdef UVERBS_OPTIMIZE_USING_STACK_SZ
+	uintptr_t data[UVERBS_OPTIMIZE_USING_STACK_SZ / sizeof(uintptr_t)];
+#endif
+
+	if (hdr->reserved)
+		return -EINVAL;
+
+	object_spec = uverbs_get_object(ib_dev, hdr->object_id);
+	if (!object_spec)
+		return -EOPNOTSUPP;
+
+	method_spec = uverbs_get_method(object_spec, hdr->method_id);
+	if (!method_spec)
+		return -EOPNOTSUPP;
+
+	if ((method_spec->flags & UVERBS_ACTION_FLAG_CREATE_ROOT) ^ !file->ucontext)
+		return -EINVAL;
+
+	ctx_size = sizeof(*ctx) +
+		   sizeof(struct uverbs_attr_bundle) +
+		   sizeof(struct uverbs_attr_bundle_hash) * method_spec->num_buckets +
+		   sizeof(*ctx->uattrs) * hdr->num_attrs +
+		   sizeof(*ctx->uverbs_attr_bundle->hash[0].attrs) *
+		   method_spec->num_child_attrs +
+		   sizeof(*ctx->uverbs_attr_bundle->hash[0].valid_bitmap) *
+			(method_spec->num_child_attrs / BITS_PER_LONG +
+			 method_spec->num_buckets);
+
+#ifdef UVERBS_OPTIMIZE_USING_STACK_SZ
+	if (ctx_size <= UVERBS_OPTIMIZE_USING_STACK_SZ)
+		ctx = (void *)data;
+
+	if (!ctx)
+#endif
+	ctx = kmalloc(ctx_size, GFP_KERNEL);
+	if (!ctx)
+		return -ENOMEM;
+
+	ctx->uverbs_attr_bundle = (void *)ctx + sizeof(*ctx);
+	ctx->uattrs = (void *)(ctx->uverbs_attr_bundle + 1) +
+			      (sizeof(ctx->uverbs_attr_bundle->hash[0]) *
+			       method_spec->num_buckets);
+	curr_attr = (void *)(ctx->uattrs + hdr->num_attrs);
+	curr_bitmap = (void *)(curr_attr + method_spec->num_child_attrs);
+
+	/*
+	 * We just fill the pointers and num_attrs here. The data itself will be
+	 * filled at a later stage (uverbs_process_attr)
+	 */
+	for (i = 0; i < method_spec->num_buckets; i++) {
+		unsigned int curr_num_attrs = method_spec->attr_buckets[i]->num_attrs;
+
+		ctx->uverbs_attr_bundle->hash[i].attrs = curr_attr;
+		curr_attr += curr_num_attrs;
+		ctx->uverbs_attr_bundle->hash[i].num_attrs = curr_num_attrs;
+		ctx->uverbs_attr_bundle->hash[i].valid_bitmap = curr_bitmap;
+		bitmap_zero(curr_bitmap, curr_num_attrs);
+		curr_bitmap += BITS_TO_LONGS(curr_num_attrs);
+	}
+
+	err = copy_from_user(ctx->uattrs, buf,
+			     sizeof(*ctx->uattrs) * hdr->num_attrs);
+	if (err) {
+		err = -EFAULT;
+		goto out;
+	}
+
+	err = uverbs_handle_method(buf, ctx->uattrs, hdr->num_attrs, ib_dev,
+				   file, method_spec, ctx->uverbs_attr_bundle);
+out:
+#ifdef UVERBS_OPTIMIZE_USING_STACK_SZ
+	if (ctx_size > UVERBS_OPTIMIZE_USING_STACK_SZ)
+#endif
+	kfree(ctx);
+	return err;
+}
+
+#define IB_UVERBS_MAX_CMD_SZ 4096
+
+long ib_uverbs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
+{
+	struct ib_uverbs_file *file = filp->private_data;
+	struct ib_uverbs_ioctl_hdr __user *user_hdr =
+		(struct ib_uverbs_ioctl_hdr __user *)arg;
+	struct ib_uverbs_ioctl_hdr hdr;
+	struct ib_device *ib_dev;
+	int srcu_key;
+	long err;
+
+	srcu_key = srcu_read_lock(&file->device->disassociate_srcu);
+	ib_dev = srcu_dereference(file->device->ib_dev,
+				  &file->device->disassociate_srcu);
+	if (!ib_dev) {
+		err = -EIO;
+		goto out;
+	}
+
+	if (cmd == RDMA_VERBS_IOCTL) {
+		err = copy_from_user(&hdr, user_hdr, sizeof(hdr));
+
+		if (err || hdr.length > IB_UVERBS_MAX_CMD_SZ ||
+		    hdr.length != sizeof(hdr) + hdr.num_attrs * sizeof(struct ib_uverbs_attr)) {
+			err = -EINVAL;
+			goto out;
+		}
+
+		if (hdr.reserved) {
+			err = -EOPNOTSUPP;
+			goto out;
+		}
+
+		err = ib_uverbs_cmd_verbs(ib_dev, file, &hdr,
+					  (__user void *)arg + sizeof(hdr));
+	} else {
+		err = -ENOIOCTLCMD;
+	}
+out:
+	srcu_read_unlock(&file->device->disassociate_srcu, srcu_key);
+
+	return err;
+}
diff --git a/include/rdma/ib_verbs.h b/include/rdma/ib_verbs.h
index 1082b4c..7bf3e5c 100644
--- a/include/rdma/ib_verbs.h
+++ b/include/rdma/ib_verbs.h
@@ -2306,6 +2306,8 @@ struct ib_device {
 	 */
 	int (*get_port_immutable)(struct ib_device *, u8, struct ib_port_immutable *);
 	void (*get_dev_fw_str)(struct ib_device *, char *str, size_t str_len);
+
+	struct uverbs_root_spec	     *specs_root;
 };
 
 struct ib_client {
diff --git a/include/rdma/uverbs_ioctl.h b/include/rdma/uverbs_ioctl.h
index d3ec02b..f83f563 100644
--- a/include/rdma/uverbs_ioctl.h
+++ b/include/rdma/uverbs_ioctl.h
@@ -43,6 +43,8 @@
 
 enum uverbs_attr_type {
 	UVERBS_ATTR_TYPE_NA,
+	UVERBS_ATTR_TYPE_PTR_IN,
+	UVERBS_ATTR_TYPE_PTR_OUT,
 	UVERBS_ATTR_TYPE_IDR,
 	UVERBS_ATTR_TYPE_FD,
 };
@@ -54,29 +56,110 @@ enum uverbs_obj_access {
 	UVERBS_ACCESS_DESTROY
 };
 
+enum {
+	UVERBS_ATTR_SPEC_F_MANDATORY	= 1U << 0,
+	/* Support extending attributes by length */
+	UVERBS_ATTR_SPEC_F_MIN_SZ	= 1U << 1,
+};
+
 struct uverbs_attr_spec {
 	enum uverbs_attr_type		type;
-	struct {
-		/*
-		 * higher bits mean the namespace and lower bits mean
-		 * the type id within the namespace.
-		 */
-		u16			obj_type;
-		u8			access;
-	} obj;
+	union {
+		u16				len;
+		struct {
+			/*
+			 * higher bits mean the namespace and lower bits mean
+			 * the type id within the namespace.
+			 */
+			u16			obj_type;
+			u8			access;
+		} obj;
+	};
+	/* Combination of bits from enum UVERBS_ATTR_SPEC_F_XXXX */
+	u8				flags;
 };
 
 struct uverbs_attr_spec_hash {
 	size_t				num_attrs;
+	unsigned long			*mandatory_attrs_bitmask;
 	struct uverbs_attr_spec		attrs[0];
 };
 
+struct uverbs_attr_bundle;
+struct ib_uverbs_file;
+
+enum {
+	/*
+	 * Action marked with this flag creates a context (or root for all
+	 * objects).
+	 */
+	UVERBS_ACTION_FLAG_CREATE_ROOT = 1U << 0,
+};
+
+struct uverbs_method_spec {
+	/* Combination of bits from enum UVERBS_ACTION_FLAG_XXXX */
+	u32						flags;
+	size_t						num_buckets;
+	size_t						num_child_attrs;
+	int (*handler)(struct ib_device *ib_dev, struct ib_uverbs_file *ufile,
+		       struct uverbs_attr_bundle *ctx);
+	struct uverbs_attr_spec_hash		*attr_buckets[0];
+};
+
+struct uverbs_method_spec_hash {
+	size_t					num_methods;
+	struct uverbs_method_spec		*methods[0];
+};
+
+struct uverbs_object_spec {
+	const struct uverbs_obj_type		*type_attrs;
+	size_t					num_buckets;
+	struct uverbs_method_spec_hash		*method_buckets[0];
+};
+
+struct uverbs_object_spec_hash {
+	size_t					num_objects;
+	struct uverbs_object_spec		*objects[0];
+};
+
+struct uverbs_root_spec {
+	size_t					num_buckets;
+	struct uverbs_object_spec_hash		*object_buckets[0];
+};
+
+/* =================================================
+ *              Parsing infrastructure
+ * =================================================
+ */
+
+struct uverbs_ptr_attr {
+	union {
+		u64		data;
+		void	__user *ptr;
+	};
+	u16		len;
+	/* Combination of bits from enum UVERBS_ATTR_F_XXXX */
+	u16		flags;
+};
+
 struct uverbs_obj_attr {
+	/* pointer to the kernel descriptor -> type, access, etc */
+	const struct uverbs_obj_type	*type;
 	struct ib_uobject		*uobject;
+	/* fd or id in idr of this object */
+	int				id;
 };
 
 struct uverbs_attr {
-	struct uverbs_obj_attr	obj_attr;
+	/*
+	 * pointer to the user-space given attribute, in order to write the
+	 * new uobject's id or update flags.
+	 */
+	struct ib_uverbs_attr __user	*uattr;
+	union {
+		struct uverbs_ptr_attr	ptr_attr;
+		struct uverbs_obj_attr	obj_attr;
+	};
 };
 
 struct uverbs_attr_bundle_hash {
diff --git a/include/uapi/rdma/rdma_user_ioctl.h b/include/uapi/rdma/rdma_user_ioctl.h
index 9388125..165a27e 100644
--- a/include/uapi/rdma/rdma_user_ioctl.h
+++ b/include/uapi/rdma/rdma_user_ioctl.h
@@ -43,6 +43,39 @@
 /* Legacy name, for user space application which already use it */
 #define IB_IOCTL_MAGIC		RDMA_IOCTL_MAGIC
 
+#define RDMA_VERBS_IOCTL \
+	_IOWR(RDMA_IOCTL_MAGIC, 1, struct ib_uverbs_ioctl_hdr)
+
+#define UVERBS_ID_NS_MASK 0xF000
+#define UVERBS_ID_NS_SHIFT 12
+
+enum {
+	/* User input */
+	UVERBS_ATTR_F_MANDATORY = 1U << 0,
+	/*
+	 * Valid output bit should be ignored and considered set in
+	 * mandatory fields. This bit is kernel output.
+	 */
+	UVERBS_ATTR_F_VALID_OUTPUT = 1U << 1,
+};
+
+struct ib_uverbs_attr {
+	__u16 attr_id;		/* command specific type attribute */
+	__u16 len;		/* only for pointers */
+	__u16 flags;		/* combination of UVERBS_ATTR_F_XXXX */
+	__u16 reserved;
+	__u64 data;		/* ptr to command, inline data or idr/fd */
+};
+
+struct ib_uverbs_ioctl_hdr {
+	__u16 length;
+	__u16 object_id;
+	__u16 method_id;
+	__u16 num_attrs;
+	__u64 reserved;
+	struct ib_uverbs_attr  attrs[0];
+};
+
 /*
  * General blocks assignments
  * It is closed on purpose do not expose it it user space
-- 
1.8.3.1

[PATCH rdma-next V1 04/13] IB/core: Declare an object instead of declaring only type attributes

[Matan Barak]

Switch all uverbs_type_attrs_xxxx with DECLARE_UVERBS_OBJECT
macros. This will be later used in order to embed the object
specific methods in the objects as well.

Signed-off-by: Matan Barak <matanb-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
Reviewed-by: Yishai Hadas <yishaih-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
---
 drivers/infiniband/core/uverbs_std_types.c | 112 +++++++++++++----------------
 include/rdma/uverbs_ioctl.h                |  16 +++++
 include/rdma/uverbs_std_types.h            |  40 +++++------
 include/rdma/uverbs_types.h                |  38 ++++++----
 4 files changed, 107 insertions(+), 99 deletions(-)

diff --git a/drivers/infiniband/core/uverbs_std_types.c b/drivers/infiniband/core/uverbs_std_types.c
index ef29337..b75c7da 100644
--- a/drivers/infiniband/core/uverbs_std_types.c
+++ b/drivers/infiniband/core/uverbs_std_types.c
@@ -209,67 +209,51 @@ static int uverbs_hot_unplug_completion_event_file(struct ib_uobject_file *uobj_
 	return 0;
 };
 
-const struct uverbs_obj_fd_type uverbs_type_attrs_comp_channel = {
-	.type = UVERBS_TYPE_ALLOC_FD(sizeof(struct ib_uverbs_completion_event_file), 0),
-	.context_closed = uverbs_hot_unplug_completion_event_file,
-	.fops = &uverbs_event_fops,
-	.name = "[infinibandevent]",
-	.flags = O_RDONLY,
-};
-
-const struct uverbs_obj_idr_type uverbs_type_attrs_cq = {
-	.type = UVERBS_TYPE_ALLOC_IDR_SZ(sizeof(struct ib_ucq_object), 0),
-	.destroy_object = uverbs_free_cq,
-};
-
-const struct uverbs_obj_idr_type uverbs_type_attrs_qp = {
-	.type = UVERBS_TYPE_ALLOC_IDR_SZ(sizeof(struct ib_uqp_object), 0),
-	.destroy_object = uverbs_free_qp,
-};
-
-const struct uverbs_obj_idr_type uverbs_type_attrs_mw = {
-	.type = UVERBS_TYPE_ALLOC_IDR(0),
-	.destroy_object = uverbs_free_mw,
-};
-
-const struct uverbs_obj_idr_type uverbs_type_attrs_mr = {
-	/* 1 is used in order to free the MR after all the MWs */
-	.type = UVERBS_TYPE_ALLOC_IDR(1),
-	.destroy_object = uverbs_free_mr,
-};
-
-const struct uverbs_obj_idr_type uverbs_type_attrs_srq = {
-	.type = UVERBS_TYPE_ALLOC_IDR_SZ(sizeof(struct ib_usrq_object), 0),
-	.destroy_object = uverbs_free_srq,
-};
-
-const struct uverbs_obj_idr_type uverbs_type_attrs_ah = {
-	.type = UVERBS_TYPE_ALLOC_IDR(0),
-	.destroy_object = uverbs_free_ah,
-};
-
-const struct uverbs_obj_idr_type uverbs_type_attrs_flow = {
-	.type = UVERBS_TYPE_ALLOC_IDR(0),
-	.destroy_object = uverbs_free_flow,
-};
-
-const struct uverbs_obj_idr_type uverbs_type_attrs_wq = {
-	.type = UVERBS_TYPE_ALLOC_IDR_SZ(sizeof(struct ib_uwq_object), 0),
-	.destroy_object = uverbs_free_wq,
-};
-
-const struct uverbs_obj_idr_type uverbs_type_attrs_rwq_ind_table = {
-	.type = UVERBS_TYPE_ALLOC_IDR(0),
-	.destroy_object = uverbs_free_rwq_ind_tbl,
-};
-
-const struct uverbs_obj_idr_type uverbs_type_attrs_xrcd = {
-	.type = UVERBS_TYPE_ALLOC_IDR_SZ(sizeof(struct ib_uxrcd_object), 0),
-	.destroy_object = uverbs_free_xrcd,
-};
-
-const struct uverbs_obj_idr_type uverbs_type_attrs_pd = {
-	/* 2 is used in order to free the PD after MRs */
-	.type = UVERBS_TYPE_ALLOC_IDR(2),
-	.destroy_object = uverbs_free_pd,
-};
+DECLARE_UVERBS_OBJECT(uverbs_object_comp_channel,
+		      UVERBS_OBJECT_COMP_CHANNEL,
+		      &UVERBS_TYPE_ALLOC_FD(0,
+					      sizeof(struct ib_uverbs_completion_event_file),
+					      uverbs_hot_unplug_completion_event_file,
+					      &uverbs_event_fops,
+					      "[infinibandevent]", O_RDONLY));
+
+DECLARE_UVERBS_OBJECT(uverbs_object_cq, UVERBS_OBJECT_CQ,
+		      &UVERBS_TYPE_ALLOC_IDR_SZ(sizeof(struct ib_ucq_object), 0,
+						  uverbs_free_cq));
+
+DECLARE_UVERBS_OBJECT(uverbs_object_qp, UVERBS_OBJECT_QP,
+		      &UVERBS_TYPE_ALLOC_IDR_SZ(sizeof(struct ib_uqp_object), 0,
+						  uverbs_free_qp));
+
+DECLARE_UVERBS_OBJECT(uverbs_object_mw, UVERBS_OBJECT_MW,
+		      &UVERBS_TYPE_ALLOC_IDR(0, uverbs_free_mw));
+
+DECLARE_UVERBS_OBJECT(uverbs_object_mr, UVERBS_OBJECT_MR,
+		      /* 1 is used in order to free the MR after all the MWs */
+		      &UVERBS_TYPE_ALLOC_IDR(1, uverbs_free_mr));
+
+DECLARE_UVERBS_OBJECT(uverbs_object_srq, UVERBS_OBJECT_SRQ,
+		      &UVERBS_TYPE_ALLOC_IDR_SZ(sizeof(struct ib_usrq_object), 0,
+						  uverbs_free_srq));
+
+DECLARE_UVERBS_OBJECT(uverbs_object_ah, UVERBS_OBJECT_AH,
+		      &UVERBS_TYPE_ALLOC_IDR(0, uverbs_free_ah));
+
+DECLARE_UVERBS_OBJECT(uverbs_object_flow, UVERBS_OBJECT_FLOW,
+		      &UVERBS_TYPE_ALLOC_IDR(0, uverbs_free_flow));
+
+DECLARE_UVERBS_OBJECT(uverbs_object_wq, UVERBS_OBJECT_WQ,
+		      &UVERBS_TYPE_ALLOC_IDR_SZ(sizeof(struct ib_uwq_object), 0,
+						  uverbs_free_wq));
+
+DECLARE_UVERBS_OBJECT(uverbs_object_rwq_ind_table,
+		      UVERBS_OBJECT_RWQ_IND_TBL,
+		      &UVERBS_TYPE_ALLOC_IDR(0, uverbs_free_rwq_ind_tbl));
+
+DECLARE_UVERBS_OBJECT(uverbs_object_xrcd, UVERBS_OBJECT_XRCD,
+		      &UVERBS_TYPE_ALLOC_IDR_SZ(sizeof(struct ib_uxrcd_object), 0,
+						  uverbs_free_xrcd));
+
+DECLARE_UVERBS_OBJECT(uverbs_object_pd, UVERBS_OBJECT_PD,
+		      /* 2 is used in order to free the PD after MRs */
+		      &UVERBS_TYPE_ALLOC_IDR(2, uverbs_free_pd));
diff --git a/include/rdma/uverbs_ioctl.h b/include/rdma/uverbs_ioctl.h
index f83f563..9913008 100644
--- a/include/rdma/uverbs_ioctl.h
+++ b/include/rdma/uverbs_ioctl.h
@@ -127,6 +127,22 @@ struct uverbs_root_spec {
 	struct uverbs_object_spec_hash		*object_buckets[0];
 };
 
+/*
+ * =======================================
+ *	Verbs definitions
+ * =======================================
+ */
+
+struct uverbs_object_def {
+	const struct uverbs_obj_type	        *type_attrs;
+};
+
+#define _UVERBS_OBJECT(_id, _type_attrs, ...)				\
+	((const struct uverbs_object_def) {				\
+	 .type_attrs = _type_attrs})
+#define DECLARE_UVERBS_OBJECT(_name, _id, _type_attrs, ...)		\
+	const struct uverbs_object_def _name =				\
+		_UVERBS_OBJECT(_id, _type_attrs, ##__VA_ARGS__)
 /* =================================================
  *              Parsing infrastructure
  * =================================================
diff --git a/include/rdma/uverbs_std_types.h b/include/rdma/uverbs_std_types.h
index 7771ce9..eda271b 100644
--- a/include/rdma/uverbs_std_types.h
+++ b/include/rdma/uverbs_std_types.h
@@ -35,18 +35,18 @@
 
 #include <rdma/uverbs_types.h>
 
-extern const struct uverbs_obj_fd_type uverbs_type_attrs_comp_channel;
-extern const struct uverbs_obj_idr_type uverbs_type_attrs_cq;
-extern const struct uverbs_obj_idr_type uverbs_type_attrs_qp;
-extern const struct uverbs_obj_idr_type uverbs_type_attrs_rwq_ind_table;
-extern const struct uverbs_obj_idr_type uverbs_type_attrs_wq;
-extern const struct uverbs_obj_idr_type uverbs_type_attrs_srq;
-extern const struct uverbs_obj_idr_type uverbs_type_attrs_ah;
-extern const struct uverbs_obj_idr_type uverbs_type_attrs_flow;
-extern const struct uverbs_obj_idr_type uverbs_type_attrs_mr;
-extern const struct uverbs_obj_idr_type uverbs_type_attrs_mw;
-extern const struct uverbs_obj_idr_type uverbs_type_attrs_pd;
-extern const struct uverbs_obj_idr_type uverbs_type_attrs_xrcd;
+extern const struct uverbs_object_def uverbs_object_comp_channel;
+extern const struct uverbs_object_def uverbs_object_cq;
+extern const struct uverbs_object_def uverbs_object_qp;
+extern const struct uverbs_object_def uverbs_object_rwq_ind_table;
+extern const struct uverbs_object_def uverbs_object_wq;
+extern const struct uverbs_object_def uverbs_object_srq;
+extern const struct uverbs_object_def uverbs_object_ah;
+extern const struct uverbs_object_def uverbs_object_flow;
+extern const struct uverbs_object_def uverbs_object_mr;
+extern const struct uverbs_object_def uverbs_object_mw;
+extern const struct uverbs_object_def uverbs_object_pd;
+extern const struct uverbs_object_def uverbs_object_xrcd;
 
 static inline struct ib_uobject *__uobj_get(const struct uverbs_obj_type *type,
 					    bool write,
@@ -56,22 +56,22 @@ static inline struct ib_uobject *__uobj_get(const struct uverbs_obj_type *type,
 	return rdma_lookup_get_uobject(type, ucontext, id, write);
 }
 
-#define uobj_get_type(_type) uverbs_type_attrs_##_type.type
+#define uobj_get_type(_object) uverbs_object_##_object.type_attrs
 
 #define uobj_get_read(_type, _id, _ucontext)				\
-	 __uobj_get(&(_type), false, _ucontext, _id)
+	 __uobj_get(_type, false, _ucontext, _id)
 
-#define uobj_get_obj_read(_type, _id, _ucontext)			\
+#define uobj_get_obj_read(_object, _id, _ucontext)			\
 ({									\
-	struct ib_uobject *uobj =					\
-		__uobj_get(&uobj_get_type(_type),			\
+	struct ib_uobject *__uobj =					\
+		__uobj_get(uverbs_object_##_object.type_attrs,		\
 			   false, _ucontext, _id);			\
 									\
-	(struct ib_##_type *)(IS_ERR(uobj) ? NULL : uobj->object);	\
+	(struct ib_##_object *)(IS_ERR(__uobj) ? NULL : __uobj->object);\
 })
 
 #define uobj_get_write(_type, _id, _ucontext)				\
-	 __uobj_get(&(_type), true, _ucontext, _id)
+	 __uobj_get(_type, true, _ucontext, _id)
 
 static inline void uobj_put_read(struct ib_uobject *uobj)
 {
@@ -108,7 +108,7 @@ static inline struct ib_uobject *__uobj_alloc(const struct uverbs_obj_type *type
 }
 
 #define uobj_alloc(_type, ucontext)	\
-	__uobj_alloc(&(_type), ucontext)
+	__uobj_alloc(_type, ucontext)
 
 #endif
 
diff --git a/include/rdma/uverbs_types.h b/include/rdma/uverbs_types.h
index 351ea18..9760b6d 100644
--- a/include/rdma/uverbs_types.h
+++ b/include/rdma/uverbs_types.h
@@ -151,22 +151,30 @@ struct uverbs_obj_fd_type {
 
 #define UVERBS_BUILD_BUG_ON(cond) (sizeof(char[1 - 2 * !!(cond)]) -	\
 				   sizeof(char))
-#define UVERBS_TYPE_ALLOC_FD(_size, _order)				 \
-	{								 \
-		.destroy_order = _order,				 \
-		.type_class = &uverbs_fd_class,				 \
-		.obj_size = (_size) +					 \
-			  UVERBS_BUILD_BUG_ON((_size) <			 \
-					      sizeof(struct ib_uobject_file)),\
-	}
-#define UVERBS_TYPE_ALLOC_IDR_SZ(_size, _order)				\
-	{								\
+#define UVERBS_TYPE_ALLOC_FD(_order, _obj_size, _context_closed, _fops, _name, _flags)\
+	((&((const struct uverbs_obj_fd_type)				\
+	 {.type = {							\
+		.destroy_order = _order,				\
+		.type_class = &uverbs_fd_class,				\
+		.obj_size = (_obj_size) +				\
+			UVERBS_BUILD_BUG_ON((_obj_size) < sizeof(struct ib_uobject_file)), \
+	 },								\
+	 .context_closed = _context_closed,				\
+	 .fops = _fops,							\
+	 .name = _name,							\
+	 .flags = _flags}))->type)
+#define UVERBS_TYPE_ALLOC_IDR_SZ(_size, _order, _destroy_object)	\
+	((&((const struct uverbs_obj_idr_type)				\
+	 {.type = {							\
 		.destroy_order = _order,				\
 		.type_class = &uverbs_idr_class,			\
 		.obj_size = (_size) +					\
-			  UVERBS_BUILD_BUG_ON((_size) <			\
-					      sizeof(struct ib_uobject)), \
-	}
-#define UVERBS_TYPE_ALLOC_IDR(_order)					\
-	 UVERBS_TYPE_ALLOC_IDR_SZ(sizeof(struct ib_uobject), _order)
+			UVERBS_BUILD_BUG_ON((_size) <			\
+					    sizeof(struct ib_uobject))	\
+	 },								\
+	 .destroy_object = _destroy_object,}))->type)
+#define UVERBS_TYPE_ALLOC_IDR(_order, _destroy_object)			\
+	 UVERBS_TYPE_ALLOC_IDR_SZ(sizeof(struct ib_uobject), _order,	\
+				  _destroy_object)
+
 #endif
-- 
1.8.3.1

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH rdma-next V1 05/13] IB/core: Add DEVICE object and root tree structure

[Matan Barak]

This adds the DEVICE object. This object supports creating the context
that all objects are created from. Moreover, it supports executing
methods which are related to the device itself, such as QUERY_DEVICE.
This is a singleton object (per file instance).

All standard objects are put in the root structure. This root will later
on be used in drivers as the source for their whole parsing tree.
Later on, when new features are added, these drivers could mix this root
with other customized objects.

Signed-off-by: Matan Barak <matanb-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
Reviewed-by: Yishai Hadas <yishaih-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
---
 drivers/infiniband/core/uverbs_std_types.c | 17 +++++++++++++++
 include/rdma/uverbs_ioctl.h                | 35 ++++++++++++++++++++++++++++++
 include/rdma/uverbs_std_types.h            | 18 +++++++++++++++
 3 files changed, 70 insertions(+)

diff --git a/drivers/infiniband/core/uverbs_std_types.c b/drivers/infiniband/core/uverbs_std_types.c
index b75c7da..5f90978 100644
--- a/drivers/infiniband/core/uverbs_std_types.c
+++ b/drivers/infiniband/core/uverbs_std_types.c
@@ -257,3 +257,20 @@ static int uverbs_hot_unplug_completion_event_file(struct ib_uobject_file *uobj_
 DECLARE_UVERBS_OBJECT(uverbs_object_pd, UVERBS_OBJECT_PD,
 		      /* 2 is used in order to free the PD after MRs */
 		      &UVERBS_TYPE_ALLOC_IDR(2, uverbs_free_pd));
+
+DECLARE_UVERBS_OBJECT(uverbs_object_device, UVERBS_OBJECT_DEVICE, NULL);
+
+DECLARE_UVERBS_OBJECT_TREE(uverbs_default_objects,
+			   &uverbs_object_device,
+			   &uverbs_object_pd,
+			   &uverbs_object_mr,
+			   &uverbs_object_comp_channel,
+			   &uverbs_object_cq,
+			   &uverbs_object_qp,
+			   &uverbs_object_ah,
+			   &uverbs_object_mw,
+			   &uverbs_object_srq,
+			   &uverbs_object_flow,
+			   &uverbs_object_wq,
+			   &uverbs_object_rwq_ind_table,
+			   &uverbs_object_xrcd);
diff --git a/include/rdma/uverbs_ioctl.h b/include/rdma/uverbs_ioctl.h
index 9913008..2e89254 100644
--- a/include/rdma/uverbs_ioctl.h
+++ b/include/rdma/uverbs_ioctl.h
@@ -133,16 +133,51 @@ struct uverbs_root_spec {
  * =======================================
  */
 
+struct uverbs_attr_def {
+	u16                           id;
+	struct uverbs_attr_spec       attr;
+};
+
+struct uverbs_method_def {
+	u16                                  id;
+	/* Combination of bits from enum UVERBS_ACTION_FLAG_XXXX */
+	u32				     flags;
+	size_t				     num_attrs;
+	const struct uverbs_attr_def * const (*attrs)[];
+	int (*handler)(struct ib_device *ib_dev, struct ib_uverbs_file *ufile,
+		       struct uverbs_attr_bundle *ctx);
+};
+
 struct uverbs_object_def {
+	u16					 id;
 	const struct uverbs_obj_type	        *type_attrs;
+	size_t				         num_methods;
+	const struct uverbs_method_def * const (*methods)[];
+};
+
+struct uverbs_object_tree_def {
+	size_t					 num_objects;
+	const struct uverbs_object_def * const (*objects)[];
 };
 
 #define _UVERBS_OBJECT(_id, _type_attrs, ...)				\
 	((const struct uverbs_object_def) {				\
+	 .id = _id,							\
 	 .type_attrs = _type_attrs})
 #define DECLARE_UVERBS_OBJECT(_name, _id, _type_attrs, ...)		\
 	const struct uverbs_object_def _name =				\
 		_UVERBS_OBJECT(_id, _type_attrs, ##__VA_ARGS__)
+#define _UVERBS_TREE_OBJECTS_SZ(...)					\
+	(sizeof((const struct uverbs_object_def * const []){__VA_ARGS__}) / \
+	 sizeof(const struct uverbs_object_def *))
+#define _UVERBS_OBJECT_TREE(...)					\
+	((const struct uverbs_object_tree_def) {			\
+	 .num_objects = _UVERBS_TREE_OBJECTS_SZ(__VA_ARGS__),		\
+	 .objects = &(const struct uverbs_object_def * const []){__VA_ARGS__} })
+#define DECLARE_UVERBS_OBJECT_TREE(_name, ...)				\
+	const struct uverbs_object_tree_def _name =			\
+		_UVERBS_OBJECT_TREE(__VA_ARGS__)
+
 /* =================================================
  *              Parsing infrastructure
  * =================================================
diff --git a/include/rdma/uverbs_std_types.h b/include/rdma/uverbs_std_types.h
index eda271b..bef7409 100644
--- a/include/rdma/uverbs_std_types.h
+++ b/include/rdma/uverbs_std_types.h
@@ -35,6 +35,23 @@
 
 #include <rdma/uverbs_types.h>
 
+enum uverbs_default_objects {
+	UVERBS_OBJECT_DEVICE, /* No instances of DEVICE are allowed */
+	UVERBS_OBJECT_PD,
+	UVERBS_OBJECT_COMP_CHANNEL,
+	UVERBS_OBJECT_CQ,
+	UVERBS_OBJECT_QP,
+	UVERBS_OBJECT_SRQ,
+	UVERBS_OBJECT_AH,
+	UVERBS_OBJECT_MR,
+	UVERBS_OBJECT_MW,
+	UVERBS_OBJECT_FLOW,
+	UVERBS_OBJECT_XRCD,
+	UVERBS_OBJECT_RWQ_IND_TBL,
+	UVERBS_OBJECT_WQ,
+	UVERBS_OBJECT_LAST,
+};
+
 extern const struct uverbs_object_def uverbs_object_comp_channel;
 extern const struct uverbs_object_def uverbs_object_cq;
 extern const struct uverbs_object_def uverbs_object_qp;
@@ -47,6 +64,7 @@
 extern const struct uverbs_object_def uverbs_object_mw;
 extern const struct uverbs_object_def uverbs_object_pd;
 extern const struct uverbs_object_def uverbs_object_xrcd;
+extern const struct uverbs_object_def uverbs_object_device;
 
 static inline struct ib_uobject *__uobj_get(const struct uverbs_obj_type *type,
 					    bool write,
-- 
1.8.3.1

[PATCH rdma-next V1 06/13] IB/core: Add uverbs merge trees functionality

[Matan Barak]

Different drivers support different features and even subset of the
common uverbs implementation. Currently, this is handled as bitmask
in every driver that represents which kind of methods it supports, but
doesn't go down to attributes granularity. Moreover, drivers might
want to add their specific types, methods and attributes to let
their user-space counter-parts be exposed to some more efficient
abstractions. It means that existence of different features is
validated syntactically via the parsing infrastructure rather than
using a complex in-handler logic.

In order to do that, we allow defining features and abstractions
as parsing trees. These per-feature parsing tree could be merged
to an efficient (perfect-hash based) parsing tree, which is later
used by the parsing infrastructure.

To sum it up, this makes a parse tree unique for a device and
represents only the features this particular device supports.
This is done by having a root specification tree per feature.
Before a device registers itself as an IB device, it merges
all these trees into one parsing tree. This parsing tree
is used to parse all user-space commands.

A future user-space application could read this parse tree. This
tree represents which objects, methods and attributes are
supported by this device.

This is based on the idea of
Jason Gunthorpe <jgunthorpe-ePGOBjL8dl3ta4EC/59zMFaTQe2KTcn/@public.gmane.org>

Signed-off-by: Matan Barak <matanb-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
Reviewed-by: Yishai Hadas <yishaih-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
---
 drivers/infiniband/core/Makefile             |   3 +-
 drivers/infiniband/core/uverbs_ioctl_merge.c | 665 +++++++++++++++++++++++++++
 include/rdma/uverbs_ioctl.h                  |  40 +-
 3 files changed, 706 insertions(+), 2 deletions(-)
 create mode 100644 drivers/infiniband/core/uverbs_ioctl_merge.c

diff --git a/drivers/infiniband/core/Makefile b/drivers/infiniband/core/Makefile
index ca6d067..0516681 100644
--- a/drivers/infiniband/core/Makefile
+++ b/drivers/infiniband/core/Makefile
@@ -31,4 +31,5 @@ ib_umad-y :=			user_mad.o
 ib_ucm-y :=			ucm.o
 
 ib_uverbs-y :=			uverbs_main.o uverbs_cmd.o uverbs_marshall.o \
-				rdma_core.o uverbs_std_types.o uverbs_ioctl.o
+				rdma_core.o uverbs_std_types.o uverbs_ioctl.o \
+				uverbs_ioctl_merge.o
diff --git a/drivers/infiniband/core/uverbs_ioctl_merge.c b/drivers/infiniband/core/uverbs_ioctl_merge.c
new file mode 100644
index 0000000..76ddb65
--- /dev/null
+++ b/drivers/infiniband/core/uverbs_ioctl_merge.c
@@ -0,0 +1,665 @@
+/*
+ * Copyright (c) 2017, Mellanox Technologies inc.  All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <rdma/uverbs_ioctl.h>
+#include <rdma/rdma_user_ioctl.h>
+#include <linux/bitops.h>
+#include "uverbs.h"
+
+#define UVERBS_NUM_NS (UVERBS_ID_NS_MASK >> UVERBS_ID_NS_SHIFT)
+#define GET_NS_ID(idx) (((idx) & UVERBS_ID_NS_MASK) >> UVERBS_ID_NS_SHIFT)
+#define GET_ID(idx) ((idx) & ~UVERBS_ID_NS_MASK)
+
+#define _for_each_element(elem, tmpi, tmpj, hashes, num_buckets_offset,	       \
+			  buckets_offset)				       \
+	for (tmpj = 0,							       \
+	     elem = (*(const void ***)((hashes)[tmpi] +			       \
+				       (buckets_offset)))[0];	               \
+	     tmpj < *(size_t *)((hashes)[tmpi] + (num_buckets_offset));        \
+	     tmpj++)						               \
+		if ((elem = ((*(const void ***)(hashes[tmpi] +		       \
+						(buckets_offset)))[tmpj])))
+
+/*
+ * Iterate all elements of a few @hashes. The number of given hashes is
+ * indicated by @num_hashes. The offset of the number of buckets in the hash is
+ * represented by @num_buckets_offset, while the offset of the buckets array in
+ * the hash structure is represented by @buckets_offset. tmpi and tmpj are two
+ * short (or int) based indices that are given by the user. tmpi iterates over
+ * the different hashes. @elem points the current element in the hashes[tmpi]
+ * bucket we are looping on. To be honest, @hashes representation isn't exactly
+ * a hash, but more a collection of elements. These elements' ids are treated
+ * in a hash like manner, where the first upper bits are the bucket number.
+ * These elements are later mapped into a perfect-hash.
+ */
+#define for_each_element(elem, tmpi, tmpj, hashes, num_hashes,                 \
+			 num_buckets_offset, buckets_offset)		       \
+	for (tmpi = 0; tmpi < (num_hashes); tmpi++)		               \
+		_for_each_element(elem, tmpi, tmpj, hashes, num_buckets_offset,\
+				  buckets_offset)
+
+#define get_elements_iterators_entry_above(iters, num_elements, elements,     \
+					  num_objects_fld, objects_fld, bucket,\
+					  min_id)			       \
+	get_elements_above_id((const void **)iters, num_elements,       \
+				     (const void **)(elements),		       \
+				     offsetof(typeof(**elements),	       \
+					      num_objects_fld),		       \
+				     offsetof(typeof(**elements), objects_fld),\
+				     offsetof(typeof(***(*elements)->objects_fld), id),\
+				     bucket, min_id)
+
+#define get_objects_above_id(iters, num_trees, trees, bucket, min_id)	       \
+	get_elements_iterators_entry_above(iters, num_trees, trees,	       \
+					   num_objects, objects, bucket, min_id)
+
+#define get_methods_above_id(method_iters, num_iters, iters, bucket, min_id)\
+	get_elements_iterators_entry_above(method_iters, num_iters, iters,     \
+					   num_methods, methods, bucket, min_id)
+
+#define get_attrs_above_id(attrs_iters, num_iters, iters, bucket, min_id)\
+	get_elements_iterators_entry_above(attrs_iters, num_iters, iters,      \
+					   num_attrs, attrs, bucket, min_id)
+
+/*
+ * get_elements_above_id get a few hashes represented by @elements and
+ * @num_elements. The hashes fields are described by @num_offset, @data_offset
+ * and @id_offset in the same way as required by for_each_element. The function
+ * returns an array of @iters, represents an array of elements in the hashes
+ * buckets, which their ids are the smallest ids in all hashes but are all
+ * larger than the id given by min_id. Elements are only added to the iters
+ * array if their id belongs to the bucket @bucket. The number of elements in
+ * the returned array is returned by the function. @min_id is also updated to
+ * reflect the new min_id of all elements in iters.
+ */
+static size_t get_elements_above_id(const void **iters,
+				    unsigned int num_elements,
+				    const void **elements,
+				    size_t num_offset,
+				    size_t data_offset,
+				    size_t id_offset,
+				    u16 bucket,
+				    short *min_id)
+{
+	size_t num_iters = 0;
+	short min = SHRT_MAX;
+	const void *elem;
+	int i, j, last_stored = -1;
+
+	for_each_element(elem, i, j, elements, num_elements, num_offset,
+			 data_offset) {
+		u16 id = *(u16 *)(elem + id_offset);
+
+		if (GET_NS_ID(id) != bucket)
+			continue;
+
+		if (GET_ID(id) < *min_id ||
+		    (min != SHRT_MAX && GET_ID(id) > min))
+			continue;
+
+		/*
+		 * We first iterate all hashes represented by @elements. When
+		 * we do, we try to find an element @elem in the bucket @bucket
+		 * which its id is min. Since we can't ensure the user sorted
+		 * the elements in increasing order, we override this hash's
+		 * minimal id element we found, if a new element with a smaller
+		 * id was just found.
+		 */
+		iters[last_stored == i ? num_iters - 1 : num_iters++] = elem;
+		last_stored = i;
+		min = GET_ID(id);
+	}
+
+	/*
+	 * We only insert to our iters array an element, if its id is smaller
+	 * than all previous ids. Therefore, the final iters array is sorted so
+	 * that smaller ids are in the end of the array.
+	 * Therefore, we need to clean the beginning of the array to make sure
+	 * all ids of final elements are equal to min.
+	 */
+	for (i = num_iters - 1; i >= 0 &&
+	     GET_ID(*(u16 *)(iters[i] + id_offset)) == min; i--)
+		;
+
+	num_iters -= i + 1;
+	memmove(iters, iters + i + 1, sizeof(*iters) * num_iters);
+
+	*min_id = min;
+	return num_iters;
+}
+
+#define find_max_element_entry_id(num_elements, elements, num_objects_fld, \
+				  objects_fld, bucket)			   \
+	find_max_element_id(num_elements, (const void **)(elements),	   \
+			    offsetof(typeof(**elements), num_objects_fld),    \
+			    offsetof(typeof(**elements), objects_fld),	      \
+			    offsetof(typeof(***(*elements)->objects_fld), id),\
+			    bucket)
+
+static short find_max_element_ns_id(unsigned int num_elements,
+				    const void **elements,
+				    size_t num_offset,
+				    size_t data_offset,
+				    size_t id_offset)
+{
+	short max_ns = SHRT_MIN;
+	const void *elem;
+	int i, j;
+
+	for_each_element(elem, i, j, elements, num_elements, num_offset,
+			 data_offset) {
+		u16 id = *(u16 *)(elem + id_offset);
+
+		if (GET_NS_ID(id) > max_ns)
+			max_ns = GET_NS_ID(id);
+	}
+
+	return max_ns;
+}
+
+static short find_max_element_id(unsigned int num_elements,
+				 const void **elements,
+				 size_t num_offset,
+				 size_t data_offset,
+				 size_t id_offset,
+				 u16 bucket)
+{
+	short max_id = SHRT_MIN;
+	const void *elem;
+	int i, j;
+
+	for_each_element(elem, i, j, elements, num_elements, num_offset,
+			 data_offset) {
+		u16 id = *(u16 *)(elem + id_offset);
+
+		if (GET_NS_ID(id) == bucket &&
+		    GET_ID(id) > max_id)
+			max_id = GET_ID(id);
+	}
+	return max_id;
+}
+
+#define find_max_element_entry_id(num_elements, elements, num_objects_fld,   \
+				  objects_fld, bucket)			      \
+	find_max_element_id(num_elements, (const void **)(elements),	      \
+			    offsetof(typeof(**elements), num_objects_fld),    \
+			    offsetof(typeof(**elements), objects_fld),	      \
+			    offsetof(typeof(***(*elements)->objects_fld), id),\
+			    bucket)
+
+#define find_max_element_ns_entry_id(num_elements, elements,		    \
+				     num_objects_fld, objects_fld)	    \
+	find_max_element_ns_id(num_elements, (const void **)(elements),	    \
+			      offsetof(typeof(**elements), num_objects_fld),\
+			      offsetof(typeof(**elements), objects_fld),    \
+			      offsetof(typeof(***(*elements)->objects_fld), id))
+
+/*
+ * find_max_xxxx_ns_id gets a few elements. Each element is described by an id
+ * which its upper bits represents a namespace. It finds the max namespace. This
+ * could be used in order to know how many buckets do we need to allocate. If no
+ * elements exist, SHRT_MIN is returned. Namespace represents here different
+ * buckets. The common example is "common bucket" and "driver bucket".
+ *
+ * find_max_xxxx_id gets a few elements and a bucket. Each element is described
+ * by an id which its upper bits represent a namespace. It returns the max id
+ * which is contained in the same namespace defined in @bucket. This could be
+ * used in order to know how many elements do we need to allocate in the bucket.
+ * If no elements exist, SHRT_MIN is returned.
+ */
+
+#define find_max_object_id(num_trees, trees, bucket)			\
+		find_max_element_entry_id(num_trees, trees, num_objects,\
+					  objects, bucket)
+#define find_max_object_ns_id(num_trees, trees)			\
+		find_max_element_ns_entry_id(num_trees, trees,		\
+					     num_objects, objects)
+
+#define find_max_method_id(num_iters, iters, bucket)			\
+		find_max_element_entry_id(num_iters, iters, num_methods,\
+					  methods, bucket)
+#define find_max_method_ns_id(num_iters, iters)			\
+		find_max_element_ns_entry_id(num_iters, iters,		\
+					     num_methods, methods)
+
+#define find_max_attr_id(num_iters, iters, bucket)			\
+		find_max_element_entry_id(num_iters, iters, num_attrs,  \
+					  attrs, bucket)
+#define find_max_attr_ns_id(num_iters, iters)				\
+		find_max_element_ns_entry_id(num_iters, iters,		\
+					     num_attrs, attrs)
+
+static void free_method(struct uverbs_method_spec *method)
+{
+	unsigned int i;
+
+	if (!method)
+		return;
+
+	for (i = 0; i < method->num_buckets; i++)
+		kfree(method->attr_buckets[i]);
+
+	kfree(method);
+}
+
+#define IS_ATTR_OBJECT(attr) ((attr)->type == UVERBS_ATTR_TYPE_IDR || \
+			      (attr)->type == UVERBS_ATTR_TYPE_FD)
+
+/*
+ * This function gets array of size @num_method_defs which contains pointers to
+ * method definitions @method_defs. The function allocates an
+ * uverbs_method_spec structure and initializes its number of buckets and the
+ * elements in buckets to the correct attributes. While doing that, it
+ * validates that there aren't conflicts between attributes of different
+ * method_defs.
+ */
+static struct uverbs_method_spec *build_method_with_attrs(const struct uverbs_method_def **method_defs,
+							  size_t num_method_defs)
+{
+	int bucket_idx;
+	int max_attr_buckets = 0;
+	size_t num_attr_buckets = 0;
+	int res = 0;
+	struct uverbs_method_spec *method = NULL;
+	const struct uverbs_attr_def **attr_defs;
+	unsigned int num_of_singularities = 0;
+
+	max_attr_buckets = find_max_attr_ns_id(num_method_defs, method_defs);
+	if (max_attr_buckets >= 0)
+		num_attr_buckets = max_attr_buckets + 1;
+
+	method = kzalloc(sizeof(*method) +
+			 num_attr_buckets * sizeof(*method->attr_buckets),
+			 GFP_KERNEL);
+	if (!method)
+		return ERR_PTR(-ENOMEM);
+
+	method->num_buckets = num_attr_buckets;
+	attr_defs = kcalloc(num_method_defs, sizeof(*attr_defs), GFP_KERNEL);
+	if (!attr_defs) {
+		res = -ENOMEM;
+		goto free_method;
+	}
+	for (bucket_idx = 0; bucket_idx < method->num_buckets; bucket_idx++) {
+		short min_id = SHRT_MIN;
+		int attr_max_bucket = 0;
+		struct uverbs_attr_spec_hash *hash = NULL;
+
+		attr_max_bucket = find_max_attr_id(num_method_defs, method_defs,
+						   bucket_idx);
+		if (attr_max_bucket < 0)
+			continue;
+
+		hash = kzalloc(sizeof(*hash) +
+			       ALIGN(sizeof(*hash->attrs) * (attr_max_bucket + 1),
+				     sizeof(long)) +
+			       BITS_TO_LONGS(attr_max_bucket) * sizeof(long),
+			       GFP_KERNEL);
+		if (!hash) {
+			res = -ENOMEM;
+			goto free;
+		}
+		hash->num_attrs = attr_max_bucket + 1;
+		method->num_child_attrs += hash->num_attrs;
+		hash->mandatory_attrs_bitmask = (void *)(hash + 1) +
+						 ALIGN(sizeof(*hash->attrs) *
+						       (attr_max_bucket + 1),
+						       sizeof(long));
+
+		method->attr_buckets[bucket_idx] = hash;
+
+		do {
+			size_t			 num_attr_defs;
+			struct uverbs_attr_spec	*attr;
+			bool attr_obj_with_special_access;
+
+			num_attr_defs =
+				get_attrs_above_id(attr_defs,
+						   num_method_defs,
+						   method_defs,
+						   bucket_idx,
+						   &min_id);
+			/* Last attr in bucket */
+			if (!num_attr_defs)
+				break;
+
+			if (num_attr_defs > 1) {
+				/*
+				 * We don't allow two attribute definitions for
+				 * the same attribute. This is usually a
+				 * programmer error. If required, it's better to
+				 * just add a new attribute to capture the new
+				 * semantics.
+				 */
+				res = -EEXIST;
+				goto free;
+			}
+
+			attr = &hash->attrs[min_id];
+			memcpy(attr, &attr_defs[0]->attr, sizeof(*attr));
+
+			attr_obj_with_special_access = IS_ATTR_OBJECT(attr) &&
+				   (attr->obj.access == UVERBS_ACCESS_NEW ||
+				    attr->obj.access == UVERBS_ACCESS_DESTROY);
+			num_of_singularities +=  !!attr_obj_with_special_access;
+			if (WARN(num_of_singularities > 1,
+				 "ib_uverbs: Method contains more than one object attr (%d) with new/destroy access\n",
+				 min_id) ||
+			    WARN(attr_obj_with_special_access &&
+				 !(attr->flags & UVERBS_ATTR_SPEC_F_MANDATORY),
+				 "ib_uverbs: Tried to merge attr (%d) but it's an object with new/destroy aceess but isn't mandatory\n",
+				 min_id) ||
+			    WARN(IS_ATTR_OBJECT(attr) &&
+				 attr->flags & UVERBS_ATTR_SPEC_F_MIN_SZ,
+				 "ib_uverbs: Tried to merge attr (%d) but it's an object with min_sz flag\n",
+				 min_id)) {
+				res = -EINVAL;
+				goto free;
+			}
+
+			if (attr->flags & UVERBS_ATTR_SPEC_F_MANDATORY)
+				set_bit(min_id, hash->mandatory_attrs_bitmask);
+			min_id++;
+
+		} while (1);
+	}
+	kfree(attr_defs);
+	return method;
+
+free:
+	kfree(attr_defs);
+free_method:
+	free_method(method);
+	return ERR_PTR(res);
+}
+
+static void free_object(struct uverbs_object_spec *object)
+{
+	unsigned int i, j;
+
+	if (!object)
+		return;
+
+	for (i = 0; i < object->num_buckets; i++) {
+		struct uverbs_method_spec_hash	*method_buckets =
+			object->method_buckets[i];
+
+		if (!method_buckets)
+			continue;
+
+		for (j = 0; j < method_buckets->num_methods; j++)
+			free_method(method_buckets->methods[j]);
+
+		kfree(method_buckets);
+	}
+
+	kfree(object);
+}
+
+/*
+ * This function gets array of size @num_object_defs which contains pointers to
+ * object definitions @object_defs. The function allocated an
+ * uverbs_object_spec structure and initialize its number of buckets and the
+ * elements in buckets to the correct methods. While doing that, it
+ * sorts out the correct relationship between conflicts in the same method.
+ */
+static struct uverbs_object_spec *build_object_with_methods(const struct uverbs_object_def **object_defs,
+							    size_t num_object_defs)
+{
+	u16 bucket_idx;
+	int max_method_buckets = 0;
+	u16 num_method_buckets = 0;
+	int res = 0;
+	struct uverbs_object_spec *object = NULL;
+	const struct uverbs_method_def **method_defs;
+
+	max_method_buckets = find_max_method_ns_id(num_object_defs, object_defs);
+	if (max_method_buckets >= 0)
+		num_method_buckets = max_method_buckets + 1;
+
+	object = kzalloc(sizeof(*object) +
+			 num_method_buckets *
+			 sizeof(*object->method_buckets), GFP_KERNEL);
+	if (!object)
+		return ERR_PTR(-ENOMEM);
+
+	object->num_buckets = num_method_buckets;
+	method_defs = kcalloc(num_object_defs, sizeof(*method_defs), GFP_KERNEL);
+	if (!method_defs) {
+		res = -ENOMEM;
+		goto free_object;
+	}
+
+	for (bucket_idx = 0; bucket_idx < object->num_buckets; bucket_idx++) {
+		short min_id = SHRT_MIN;
+		int methods_max_bucket = 0;
+		struct uverbs_method_spec_hash *hash = NULL;
+
+		methods_max_bucket = find_max_method_id(num_object_defs, object_defs,
+							bucket_idx);
+		if (methods_max_bucket < 0)
+			continue;
+
+		hash = kzalloc(sizeof(*hash) +
+			       sizeof(*hash->methods) * (methods_max_bucket + 1),
+			       GFP_KERNEL);
+		if (!hash) {
+			res = -ENOMEM;
+			goto free;
+		}
+
+		hash->num_methods = methods_max_bucket + 1;
+		object->method_buckets[bucket_idx] = hash;
+
+		do {
+			size_t				num_method_defs;
+			struct uverbs_method_spec	*method;
+			int i;
+
+			num_method_defs =
+				get_methods_above_id(method_defs,
+						     num_object_defs,
+						     object_defs,
+						     bucket_idx,
+						     &min_id);
+			/* Last method in bucket */
+			if (!num_method_defs)
+				break;
+
+			method = build_method_with_attrs(method_defs,
+							 num_method_defs);
+			if (IS_ERR(method)) {
+				res = PTR_ERR(method);
+				goto free;
+			}
+
+			/*
+			 * The last tree which is given as an argument to the
+			 * merge overrides previous method handler.
+			 * Therefore, we iterate backwards and search for the
+			 * first handler which != NULL. This also defines the
+			 * set of flags used for this handler.
+			 */
+			for (i = num_object_defs - 1;
+			     i >= 0 && !method_defs[i]->handler; i--)
+				;
+			hash->methods[min_id++] = method;
+			/* NULL handler isn't allowed */
+			if (WARN(i < 0,
+				 "ib_uverbs: tried to merge function id %d, but all handlers are NULL\n",
+				 min_id)) {
+				res = -EINVAL;
+				goto free;
+			}
+			method->handler = method_defs[i]->handler;
+			method->flags = method_defs[i]->flags;
+
+		} while (1);
+	}
+	kfree(method_defs);
+	return object;
+
+free:
+	kfree(method_defs);
+free_object:
+	free_object(object);
+	return ERR_PTR(res);
+}
+
+void uverbs_free_spec_tree(struct uverbs_root_spec *root)
+{
+	unsigned int i, j;
+
+	if (!root)
+		return;
+
+	for (i = 0; i < root->num_buckets; i++) {
+		struct uverbs_object_spec_hash *object_hash =
+			root->object_buckets[i];
+
+		if (!object_hash)
+			continue;
+
+		for (j = 0; j < object_hash->num_objects; j++)
+			free_object(object_hash->objects[j]);
+
+		kfree(object_hash);
+	}
+
+	kfree(root);
+}
+EXPORT_SYMBOL(uverbs_free_spec_tree);
+
+struct uverbs_root_spec *uverbs_alloc_spec_tree(unsigned int num_trees,
+						const struct uverbs_object_tree_def **trees)
+{
+	u16 bucket_idx;
+	short max_object_buckets = 0;
+	size_t num_objects_buckets = 0;
+	struct uverbs_root_spec *root_spec = NULL;
+	const struct uverbs_object_def **object_defs;
+	int i;
+	int res = 0;
+
+	max_object_buckets = find_max_object_ns_id(num_trees, trees);
+	/*
+	 * Devices which don't want to support ib_uverbs, should just allocate
+	 * an empty parsing tree. Every user-space command won't hit any valid
+	 * entry in the parsing tree and thus will fail.
+	 */
+	if (max_object_buckets >= 0)
+		num_objects_buckets = max_object_buckets + 1;
+
+	root_spec = kzalloc(sizeof(*root_spec) +
+			    num_objects_buckets * sizeof(*root_spec->object_buckets),
+			    GFP_KERNEL);
+	if (!root_spec)
+		return ERR_PTR(-ENOMEM);
+	root_spec->num_buckets = num_objects_buckets;
+
+	object_defs = kcalloc(num_trees, sizeof(*object_defs),
+			      GFP_KERNEL);
+	if (!object_defs) {
+		res = -ENOMEM;
+		goto free_root;
+	}
+
+	for (bucket_idx = 0; bucket_idx < root_spec->num_buckets; bucket_idx++) {
+		short min_id = SHRT_MIN;
+		short objects_max_bucket;
+		struct uverbs_object_spec_hash *hash = NULL;
+
+		objects_max_bucket = find_max_object_id(num_trees, trees,
+							bucket_idx);
+		if (objects_max_bucket < 0)
+			continue;
+
+		hash = kzalloc(sizeof(*hash) +
+			       sizeof(*hash->objects) * (objects_max_bucket + 1),
+			       GFP_KERNEL);
+		if (!hash) {
+			res = -ENOMEM;
+			goto free;
+		}
+		hash->num_objects = objects_max_bucket + 1;
+		root_spec->object_buckets[bucket_idx] = hash;
+
+		do {
+			size_t				num_object_defs;
+			struct uverbs_object_spec	*object;
+
+			num_object_defs = get_objects_above_id(object_defs,
+							       num_trees,
+							       trees,
+							       bucket_idx,
+							       &min_id);
+			/* Last object in bucket */
+			if (!num_object_defs)
+				break;
+
+			object = build_object_with_methods(object_defs,
+							   num_object_defs);
+			if (IS_ERR(object)) {
+				res = PTR_ERR(object);
+				goto free;
+			}
+
+			/*
+			 * The last tree which is given as an argument to the
+			 * merge overrides previous object's type_attrs.
+			 * Therefore, we iterate backwards and search for the
+			 * first type_attrs which != NULL.
+			 */
+			for (i = num_object_defs - 1;
+			     i >= 0 && !object_defs[i]->type_attrs; i--)
+				;
+			/*
+			 * NULL is a valid type_attrs. It means an object we
+			 * can't instantiate (like DEVICE).
+			 */
+			object->type_attrs = i < 0 ? NULL :
+				object_defs[i]->type_attrs;
+
+			hash->objects[min_id++] = object;
+		} while (1);
+	}
+
+	kfree(object_defs);
+	return root_spec;
+
+free:
+	kfree(object_defs);
+free_root:
+	uverbs_free_spec_tree(root_spec);
+	return ERR_PTR(res);
+}
+EXPORT_SYMBOL(uverbs_alloc_spec_tree);
diff --git a/include/rdma/uverbs_ioctl.h b/include/rdma/uverbs_ioctl.h
index 2e89254..cf5b238 100644
--- a/include/rdma/uverbs_ioctl.h
+++ b/include/rdma/uverbs_ioctl.h
@@ -235,5 +235,43 @@ static inline bool uverbs_attr_is_valid_in_hash(const struct uverbs_attr_bundle_
 	return test_bit(idx, attrs_hash->valid_bitmap);
 }
 
-#endif
+/* =================================================
+ *	 Definitions -> Specs infrastructure
+ * =================================================
+ */
+
+/*
+ * uverbs_alloc_spec_tree - Merges different common and driver specific feature
+ *	into one parsing tree that every uverbs command will be parsed upon.
+ *
+ * @num_trees: Number of trees in the array @trees.
+ * @trees: Array of pointers to tree root definitions to merge. Each such tree
+ *	   possibly contains objects, methods and attributes definitions.
+ *
+ * Returns:
+ *	uverbs_root_spec *: The root of the merged parsing tree.
+ *	On error, we return an error code. Error is checked via IS_ERR.
+ *
+ * The following merges could take place:
+ * a. Two trees representing the same method with different handler
+ *	-> We take the handler of the tree that its handler != NULL
+ *	   and its index in the trees array is greater. The incentive for that
+ *	   is that developers are expected to first merge common trees and then
+ *	   merge trees that gives specialized the behaviour.
+ * b. Two trees representing the same object with different
+ *    type_attrs (struct uverbs_obj_type):
+ *	-> We take the type_attrs of the tree that its type_attr != NULL
+ *	   and its index in the trees array is greater. This could be used
+ *	   in order to override the free function, allocation size, etc.
+ * c. Two trees representing the same method attribute (same id but possibly
+ *    different attributes):
+ *	-> ERROR (-ENOENT), we believe that's not the programmer's intent.
+ *
+ * An object without any methods is considered invalid and will abort the
+ * function with -ENOENT error.
+ */
+struct uverbs_root_spec *uverbs_alloc_spec_tree(unsigned int num_trees,
+						const struct uverbs_object_tree_def **trees);
+void uverbs_free_spec_tree(struct uverbs_root_spec *root);
 
+#endif
-- 
1.8.3.1

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH rdma-next V1 07/13] IB/core: Add macros for declaring methods and attributes

[Matan Barak]

This patch adds macros for declaring objects, methods and
attributes. These definitions are later used by downstream patches
to declare some of the default types.

Signed-off-by: Matan Barak <matanb-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
Reviewed-by: Yishai Hadas <yishaih-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
---
 include/rdma/uverbs_ioctl.h | 105 +++++++++++++++++++++++++++++++++++++++++++-
 1 file changed, 104 insertions(+), 1 deletion(-)

diff --git a/include/rdma/uverbs_ioctl.h b/include/rdma/uverbs_ioctl.h
index cf5b238..9a8d217 100644
--- a/include/rdma/uverbs_ioctl.h
+++ b/include/rdma/uverbs_ioctl.h
@@ -34,6 +34,8 @@
 #define _UVERBS_IOCTL_
 
 #include <rdma/uverbs_types.h>
+#include <linux/uaccess.h>
+#include <rdma/rdma_user_ioctl.h>
 
 /*
  * =======================================
@@ -160,10 +162,99 @@ struct uverbs_object_tree_def {
 	const struct uverbs_object_def * const (*objects)[];
 };
 
+#define UA_FLAGS(_flags)  .flags = _flags
+#define __UVERBS_ATTR0(_id, _len, _type, ...)                           \
+	((const struct uverbs_attr_def)				  \
+	 {.id = _id, .attr = {.type = _type, {.len = _len}, .flags = 0, } })
+#define __UVERBS_ATTR1(_id, _len, _type, _flags)                        \
+	((const struct uverbs_attr_def)				  \
+	 {.id = _id, .attr = {.type = _type, {.len = _len}, _flags, } })
+#define __UVERBS_ATTR(_id, _len, _type, _flags, _n, ...)		\
+	__UVERBS_ATTR##_n(_id, _len, _type, _flags)
+/*
+ * In new compiler, UVERBS_ATTR could be simplified by declaring it as
+ * [_id] = {.type = _type, .len = _len, ##__VA_ARGS__}
+ * But since we support older compilers too, we need the more complex code.
+ */
+#define UVERBS_ATTR(_id, _len, _type, ...)				\
+	__UVERBS_ATTR(_id, _len, _type, ##__VA_ARGS__, 1, 0)
+#define UVERBS_ATTR_PTR_IN_SZ(_id, _len, ...)				\
+	UVERBS_ATTR(_id, _len, UVERBS_ATTR_TYPE_PTR_IN, ##__VA_ARGS__)
+/* If sizeof(_type) <= sizeof(u64), this will be inlined rather than a pointer */
+#define UVERBS_ATTR_PTR_IN(_id, _type, ...)				\
+	UVERBS_ATTR_PTR_IN_SZ(_id, sizeof(_type), ##__VA_ARGS__)
+#define UVERBS_ATTR_PTR_OUT_SZ(_id, _len, ...)				\
+	UVERBS_ATTR(_id, _len, UVERBS_ATTR_TYPE_PTR_OUT, ##__VA_ARGS__)
+#define UVERBS_ATTR_PTR_OUT(_id, _type, ...)				\
+	UVERBS_ATTR_PTR_OUT_SZ(_id, sizeof(_type), ##__VA_ARGS__)
+
+/*
+ * In new compiler, UVERBS_ATTR_IDR (and FD) could be simplified by declaring
+ * it as
+ * {.id = _id,								\
+ *  .attr {.type = __obj_class,						\
+ *         .obj = {.obj_type = _idr_type,				\
+ *                       .access = _access                              \
+ *                }, ##__VA_ARGS__ } }
+ * But since we support older compilers too, we need the more complex code.
+ */
+#define ___UVERBS_ATTR_OBJ0(_id, _obj_class, _obj_type, _access, ...)\
+	((const struct uverbs_attr_def)					\
+	{.id = _id,							\
+	 .attr = {.type = _obj_class,					\
+		  {.obj = {.obj_type = _obj_type, .access = _access } },\
+		  .flags = 0} })
+#define ___UVERBS_ATTR_OBJ1(_id, _obj_class, _obj_type, _access, _flags)\
+	((const struct uverbs_attr_def)					\
+	{.id = _id,							\
+	.attr = {.type = _obj_class,					\
+		 {.obj = {.obj_type = _obj_type, .access = _access} },	\
+		  _flags} })
+#define ___UVERBS_ATTR_OBJ(_id, _obj_class, _obj_type, _access, _flags, \
+			   _n, ...)					\
+	___UVERBS_ATTR_OBJ##_n(_id, _obj_class, _obj_type, _access, _flags)
+#define __UVERBS_ATTR_OBJ(_id, _obj_class, _obj_type, _access, ...)	\
+	___UVERBS_ATTR_OBJ(_id, _obj_class, _obj_type, _access,		\
+			   ##__VA_ARGS__, 1, 0)
+#define UVERBS_ATTR_IDR(_id, _idr_type, _access, ...)			 \
+	__UVERBS_ATTR_OBJ(_id, UVERBS_ATTR_TYPE_IDR, _idr_type, _access,\
+			  ##__VA_ARGS__)
+#define UVERBS_ATTR_FD(_id, _fd_type, _access, ...)			\
+	__UVERBS_ATTR_OBJ(_id, UVERBS_ATTR_TYPE_FD, _fd_type,		\
+			  (_access) + BUILD_BUG_ON_ZERO(		\
+				(_access) != UVERBS_ACCESS_NEW &&	\
+				(_access) != UVERBS_ACCESS_READ),	\
+			  ##__VA_ARGS__)
+#define DECLARE_UVERBS_ATTR_SPEC(_name, ...)				\
+	const struct uverbs_attr_def _name = __VA_ARGS__
+
+#define _UVERBS_METHOD_ATTRS_SZ(...)					\
+	(sizeof((const struct uverbs_attr_def * const []){__VA_ARGS__}) /\
+	 sizeof(const struct uverbs_attr_def *))
+#define _UVERBS_METHOD(_id, _handler, _flags, ...)			\
+	((const struct uverbs_method_def) {				\
+	 .id = _id,							\
+	 .flags = _flags,						\
+	 .handler = _handler,						\
+	 .num_attrs = _UVERBS_METHOD_ATTRS_SZ(__VA_ARGS__),		\
+	 .attrs = &(const struct uverbs_attr_def * const []){__VA_ARGS__} })
+#define DECLARE_UVERBS_METHOD(_name, _id, _handler, ...)		\
+	const struct uverbs_method_def _name =				\
+		_UVERBS_METHOD(_id, _handler, 0, ##__VA_ARGS__)
+#define DECLARE_UVERBS_CTX_METHOD(_name, _id, _handler, _flags, ...)	\
+	const struct uverbs_method_def _name =				\
+		_UVERBS_METHOD(_id, _handler,				\
+			       UVERBS_ACTION_FLAG_CREATE_ROOT,		\
+			       ##__VA_ARGS__)
+#define _UVERBS_OBJECT_METHODS_SZ(...)					\
+	(sizeof((const struct uverbs_method_def * const []){__VA_ARGS__}) / \
+	 sizeof(const struct uverbs_method_def *))
 #define _UVERBS_OBJECT(_id, _type_attrs, ...)				\
 	((const struct uverbs_object_def) {				\
 	 .id = _id,							\
-	 .type_attrs = _type_attrs})
+	 .type_attrs = _type_attrs,					\
+	 .num_methods = _UVERBS_OBJECT_METHODS_SZ(__VA_ARGS__),		\
+	 .methods = &(const struct uverbs_method_def * const []){__VA_ARGS__} })
 #define DECLARE_UVERBS_OBJECT(_name, _id, _type_attrs, ...)		\
 	const struct uverbs_object_def _name =				\
 		_UVERBS_OBJECT(_id, _type_attrs, ##__VA_ARGS__)
@@ -235,6 +326,18 @@ static inline bool uverbs_attr_is_valid_in_hash(const struct uverbs_attr_bundle_
 	return test_bit(idx, attrs_hash->valid_bitmap);
 }
 
+static inline bool uverbs_attr_is_valid(const struct uverbs_attr_bundle *attrs_bundle,
+					unsigned int idx)
+{
+	u16 idx_bucket = idx >>	UVERBS_ID_NS_SHIFT;
+
+	if (attrs_bundle->num_buckets <= idx_bucket)
+		return false;
+
+	return uverbs_attr_is_valid_in_hash(&attrs_bundle->hash[idx_bucket],
+					    idx & ~UVERBS_ID_NS_MASK);
+}
+
 /* =================================================
  *	 Definitions -> Specs infrastructure
  * =================================================
-- 
1.8.3.1

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH rdma-next V1 08/13] IB/core: Explicitly destroy an object while keeping uobject

[Matan Barak]

When some objects are destroyed, we need to extract their status at
destruction. After object's destruction, this status
(e.g. events_reported) relies in the uobject. In order to have the
latest and correct status, the underlying object should be destroyed,
but we should keep the uobject alive and read this information off the
uobject. We introduce a rdma_explicit_destroy function. This function
destroys the class type object (for example, the IDR class type which
destroys the underlying object as well) and then convert the uobject
to be of a null class type. This uobject will then be destroyed as any
other uobject once uverbs_finalize_object[s] is called.

Signed-off-by: Matan Barak <matanb-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
Reviewed-by: Yishai Hadas <yishaih-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
---
 drivers/infiniband/core/rdma_core.c | 35 +++++++++++++++++++++++++++++++++++
 include/rdma/uverbs_types.h         |  1 +
 2 files changed, 36 insertions(+)

diff --git a/drivers/infiniband/core/rdma_core.c b/drivers/infiniband/core/rdma_core.c
index 2a2f002..85b5ee4 100644
--- a/drivers/infiniband/core/rdma_core.c
+++ b/drivers/infiniband/core/rdma_core.c
@@ -451,6 +451,41 @@ int __must_check rdma_remove_commit_uobject(struct ib_uobject *uobj)
 	return ret;
 }
 
+static int null_obj_type_class_remove_commit(struct ib_uobject *uobj,
+					     enum rdma_remove_reason why)
+{
+	return 0;
+}
+
+static const struct uverbs_obj_type null_obj_type = {
+	.type_class = &((const struct uverbs_obj_type_class){
+			.remove_commit = null_obj_type_class_remove_commit,
+			/* be cautious */
+			.needs_kfree_rcu = true}),
+};
+
+int rdma_explicit_destroy(struct ib_uobject *uobject)
+{
+	int ret;
+	struct ib_ucontext *ucontext = uobject->context;
+
+	/* Cleanup is running. Calling this should have been impossible */
+	if (!down_read_trylock(&ucontext->cleanup_rwsem)) {
+		WARN(true, "ib_uverbs: Cleanup is running while removing an uobject\n");
+		return 0;
+	}
+	lockdep_check(uobject, true);
+	ret = uobject->type->type_class->remove_commit(uobject,
+						       RDMA_REMOVE_DESTROY);
+	if (ret)
+		return ret;
+
+	uobject->type = &null_obj_type;
+
+	up_read(&ucontext->cleanup_rwsem);
+	return 0;
+}
+
 static void alloc_commit_idr_uobject(struct ib_uobject *uobj)
 {
 	uverbs_uobject_add(uobj);
diff --git a/include/rdma/uverbs_types.h b/include/rdma/uverbs_types.h
index 9760b6d..cc04ec6 100644
--- a/include/rdma/uverbs_types.h
+++ b/include/rdma/uverbs_types.h
@@ -129,6 +129,7 @@ struct ib_uobject *rdma_alloc_begin_uobject(const struct uverbs_obj_type *type,
 void rdma_alloc_abort_uobject(struct ib_uobject *uobj);
 int __must_check rdma_remove_commit_uobject(struct ib_uobject *uobj);
 int rdma_alloc_commit_uobject(struct ib_uobject *uobj);
+int rdma_explicit_destroy(struct ib_uobject *uobject);
 
 struct uverbs_obj_fd_type {
 	/*
-- 
1.8.3.1

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH rdma-next V1 09/13] IB/core: Export ioctl enum types to user-space

[Matan Barak]

Add a new ib_user_ioctl_verbs.h which exports all required ABI
enums and structs to the user-space.
Export the default types to user-space through this file.

Signed-off-by: Matan Barak <matanb-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
Reviewed-by: Yishai Hadas <yishaih-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
---
 include/rdma/uverbs_std_types.h         | 18 +----------
 include/uapi/rdma/ib_user_ioctl_verbs.h | 54 +++++++++++++++++++++++++++++++++
 2 files changed, 55 insertions(+), 17 deletions(-)
 create mode 100644 include/uapi/rdma/ib_user_ioctl_verbs.h

diff --git a/include/rdma/uverbs_std_types.h b/include/rdma/uverbs_std_types.h
index bef7409..400efe2 100644
--- a/include/rdma/uverbs_std_types.h
+++ b/include/rdma/uverbs_std_types.h
@@ -34,23 +34,7 @@
 #define _UVERBS_STD_TYPES__
 
 #include <rdma/uverbs_types.h>
-
-enum uverbs_default_objects {
-	UVERBS_OBJECT_DEVICE, /* No instances of DEVICE are allowed */
-	UVERBS_OBJECT_PD,
-	UVERBS_OBJECT_COMP_CHANNEL,
-	UVERBS_OBJECT_CQ,
-	UVERBS_OBJECT_QP,
-	UVERBS_OBJECT_SRQ,
-	UVERBS_OBJECT_AH,
-	UVERBS_OBJECT_MR,
-	UVERBS_OBJECT_MW,
-	UVERBS_OBJECT_FLOW,
-	UVERBS_OBJECT_XRCD,
-	UVERBS_OBJECT_RWQ_IND_TBL,
-	UVERBS_OBJECT_WQ,
-	UVERBS_OBJECT_LAST,
-};
+#include <rdma/ib_user_ioctl_verbs.h>
 
 extern const struct uverbs_object_def uverbs_object_comp_channel;
 extern const struct uverbs_object_def uverbs_object_cq;
diff --git a/include/uapi/rdma/ib_user_ioctl_verbs.h b/include/uapi/rdma/ib_user_ioctl_verbs.h
new file mode 100644
index 0000000..78a2e5b
--- /dev/null
+++ b/include/uapi/rdma/ib_user_ioctl_verbs.h
@@ -0,0 +1,54 @@
+/*
+ * Copyright (c) 2017, Mellanox Technologies inc.  All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#ifndef IB_USER_IOCTL_VERBS_H
+#define IB_USER_IOCTL_VERBS_H
+
+enum uverbs_default_objects {
+	UVERBS_OBJECT_DEVICE, /* No instances of DEVICE are allowed */
+	UVERBS_OBJECT_PD,
+	UVERBS_OBJECT_COMP_CHANNEL,
+	UVERBS_OBJECT_CQ,
+	UVERBS_OBJECT_QP,
+	UVERBS_OBJECT_SRQ,
+	UVERBS_OBJECT_AH,
+	UVERBS_OBJECT_MR,
+	UVERBS_OBJECT_MW,
+	UVERBS_OBJECT_FLOW,
+	UVERBS_OBJECT_XRCD,
+	UVERBS_OBJECT_RWQ_IND_TBL,
+	UVERBS_OBJECT_WQ,
+	UVERBS_OBJECT_LAST,
+};
+
+#endif
+
-- 
1.8.3.1

[PATCH rdma-next V1 10/13] IB/core: Add legacy driver's user-data

[Matan Barak]

In this phase, we don't want to change all the drivers to use
flexible driver's specific attributes. Therefore, we add two default
attributes: UHW_IN and UHW_OUT. These attributes are optional in some
methods and they encode the driver specific command data. We add
a function that extract this data and creates the legacy udata over
it.

Driver's data should start from UVERBS_UDATA_DRIVER_DATA_FLAG. This
turns on the first bit of the namespace, indicating this attribute
belongs to the driver's namespace.

Signed-off-by: Matan Barak <matanb-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
Reviewed-by: Yishai Hadas <yishaih-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
---
 drivers/infiniband/core/uverbs_std_types.c | 40 ++++++++++++++++++++++++++
 include/rdma/uverbs_ioctl.h                | 46 ++++++++++++++++++++++++++++++
 include/uapi/rdma/ib_user_ioctl_verbs.h    | 10 +++++++
 3 files changed, 96 insertions(+)

diff --git a/drivers/infiniband/core/uverbs_std_types.c b/drivers/infiniband/core/uverbs_std_types.c
index 5f90978..db66c18 100644
--- a/drivers/infiniband/core/uverbs_std_types.c
+++ b/drivers/infiniband/core/uverbs_std_types.c
@@ -209,6 +209,46 @@ static int uverbs_hot_unplug_completion_event_file(struct ib_uobject_file *uobj_
 	return 0;
 };
 
+/*
+ * This spec is used in order to pass information to the hardware driver in a
+ * legacy way. Every verb that could get driver specific data should get this
+ * spec.
+ */
+static const struct uverbs_attr_def uverbs_uhw_compat_in =
+	UVERBS_ATTR_PTR_IN_SZ(UVERBS_UHW_IN, 0, UA_FLAGS(UVERBS_ATTR_SPEC_F_MIN_SZ));
+static const struct uverbs_attr_def uverbs_uhw_compat_out =
+	UVERBS_ATTR_PTR_OUT_SZ(UVERBS_UHW_OUT, 0, UA_FLAGS(UVERBS_ATTR_SPEC_F_MIN_SZ));
+
+static void create_udata(struct uverbs_attr_bundle *ctx,
+			 struct ib_udata *udata)
+{
+	/*
+	 * This is for ease of conversion. The purpose is to convert all drivers
+	 * to use uverbs_attr_bundle instead of ib_udata.
+	 * Assume attr == 0 is input and attr == 1 is output.
+	 */
+	void __user *inbuf;
+	size_t inbuf_len = 0;
+	void __user *outbuf;
+	size_t outbuf_len = 0;
+	const struct uverbs_attr *uhw_in =
+		uverbs_attr_get(ctx, UVERBS_UHW_IN);
+	const struct uverbs_attr *uhw_out =
+		uverbs_attr_get(ctx, UVERBS_UHW_OUT);
+
+	if (!IS_ERR(uhw_in)) {
+		inbuf = uhw_in->ptr_attr.ptr;
+		inbuf_len = uhw_in->ptr_attr.len;
+	}
+
+	if (!IS_ERR(uhw_out)) {
+		outbuf = uhw_out->ptr_attr.ptr;
+		outbuf_len = uhw_out->ptr_attr.len;
+	}
+
+	INIT_UDATA_BUF_OR_NULL(udata, inbuf, outbuf, inbuf_len, outbuf_len);
+}
+
 DECLARE_UVERBS_OBJECT(uverbs_object_comp_channel,
 		      UVERBS_OBJECT_COMP_CHANNEL,
 		      &UVERBS_TYPE_ALLOC_FD(0,
diff --git a/include/rdma/uverbs_ioctl.h b/include/rdma/uverbs_ioctl.h
index 9a8d217..759afa0 100644
--- a/include/rdma/uverbs_ioctl.h
+++ b/include/rdma/uverbs_ioctl.h
@@ -36,6 +36,7 @@
 #include <rdma/uverbs_types.h>
 #include <linux/uaccess.h>
 #include <rdma/rdma_user_ioctl.h>
+#include <rdma/ib_user_ioctl_verbs.h>
 
 /*
  * =======================================
@@ -338,6 +339,51 @@ static inline bool uverbs_attr_is_valid(const struct uverbs_attr_bundle *attrs_b
 					    idx & ~UVERBS_ID_NS_MASK);
 }
 
+static inline const struct uverbs_attr *uverbs_attr_get(const struct uverbs_attr_bundle *attrs_bundle,
+							u16 idx)
+{
+	u16 idx_bucket = idx >>	UVERBS_ID_NS_SHIFT;
+
+	if (!uverbs_attr_is_valid(attrs_bundle, idx))
+		return ERR_PTR(-ENOENT);
+
+	return &attrs_bundle->hash[idx_bucket].attrs[idx & ~UVERBS_ID_NS_MASK];
+}
+
+static inline int uverbs_copy_to(const struct uverbs_attr_bundle *attrs_bundle,
+				 size_t idx, const void *from)
+{
+	const struct uverbs_attr *attr = uverbs_attr_get(attrs_bundle, idx);
+	u16 flags;
+
+	if (IS_ERR(attr))
+		return PTR_ERR(attr);
+
+	flags = attr->ptr_attr.flags | UVERBS_ATTR_F_VALID_OUTPUT;
+	return (!copy_to_user(attr->ptr_attr.ptr, from, attr->ptr_attr.len) &&
+		!put_user(flags, &attr->uattr->flags)) ? 0 : -EFAULT;
+}
+
+static inline int _uverbs_copy_from(void *to, size_t to_size,
+				    const struct uverbs_attr_bundle *attrs_bundle,
+				    size_t idx)
+{
+	const struct uverbs_attr *attr = uverbs_attr_get(attrs_bundle, idx);
+
+	if (IS_ERR(attr))
+		return PTR_ERR(attr);
+
+	if (to_size <= sizeof(((struct ib_uverbs_attr *)0)->data))
+		memcpy(to, &attr->ptr_attr.data, attr->ptr_attr.len);
+	else if (copy_from_user(to, attr->ptr_attr.ptr, attr->ptr_attr.len))
+		return -EFAULT;
+
+	return 0;
+}
+
+#define uverbs_copy_from(to, attrs_bundle, idx)				      \
+	_uverbs_copy_from(to, sizeof(*(to)), attrs_bundle, idx)
+
 /* =================================================
  *	 Definitions -> Specs infrastructure
  * =================================================
diff --git a/include/uapi/rdma/ib_user_ioctl_verbs.h b/include/uapi/rdma/ib_user_ioctl_verbs.h
index 78a2e5b..90f81ee 100644
--- a/include/uapi/rdma/ib_user_ioctl_verbs.h
+++ b/include/uapi/rdma/ib_user_ioctl_verbs.h
@@ -33,6 +33,11 @@
 #ifndef IB_USER_IOCTL_VERBS_H
 #define IB_USER_IOCTL_VERBS_H
 
+#include <rdma/rdma_user_ioctl.h>
+
+#define UVERBS_UDATA_DRIVER_DATA_NS	1
+#define UVERBS_UDATA_DRIVER_DATA_FLAG	(1UL << UVERBS_ID_NS_SHIFT)
+
 enum uverbs_default_objects {
 	UVERBS_OBJECT_DEVICE, /* No instances of DEVICE are allowed */
 	UVERBS_OBJECT_PD,
@@ -50,5 +55,10 @@ enum uverbs_default_objects {
 	UVERBS_OBJECT_LAST,
 };
 
+enum {
+	UVERBS_UHW_IN = UVERBS_UDATA_DRIVER_DATA_FLAG,
+	UVERBS_UHW_OUT,
+};
+
 #endif
 
-- 
1.8.3.1

[PATCH rdma-next V1 11/13] IB/core: Add completion queue (cq) object actions

[Matan Barak]

Adding CQ ioctl actions:
1. create_cq
2. destroy_cq

This requires adding the following:
1. A specification describing the method
	a. Handler
	b. Attributes specification
		Each attribute is one of the following:
		a. PTR_IN - input data
			    Note: This could be encoded inlined for
				  data < 64bit
		b. PTR_OUT - response data
		c. IDR - idr based object
		d. FD - fd based object
                Blobs attributes (clauses a and b) contain their type,
	        while objects specifications (clauses c and d)
                contains the expected object type (for example, the
                given id should be UVERBS_TYPE_PD) and the required
                access (READ, WRITE, NEW or DESTROY). If a NEW is
                required, the new object's id will be assigned to this
                attribute. All attributes could get UA_FLAGS
                attribute. Currently we support stating that an
		attribute is mandatory or that the specification size
                corresponds to a lower bound (and that this attribute
		could be extended).
		We currently add both default attributes and the two
		generic UHW_IN and UHW_OUT driver specific attributes.
2. Handler
   A handler gets a uverbs_attr_bundle. The handler developer uses
   uverbs_attr_get to fetch an attribute of a given id.
   Each of these attribute groups correspond to the specification
   group defined in the action (clauses 1.b and 1.c respectively).
   The indices of these arrays corresponds to the attribute ids
   declared in the specifications (clause 2).

   The handler is quite simple. It assumes the infrastructure fetched
   all objects and locked, created or destroyed them as required by
   the specification. Pointer (or blob) attributes were validated to
   match their required sizes. After the handler finished, the
   infrastructure commits or rollbacks the objects.

Signed-off-by: Matan Barak <matanb-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
Reviewed-by: Yishai Hadas <yishaih-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
---
 drivers/infiniband/core/uverbs_std_types.c | 138 ++++++++++++++++++++++++++++-
 include/uapi/rdma/ib_user_ioctl_verbs.h    |  20 +++++
 2 files changed, 157 insertions(+), 1 deletion(-)

diff --git a/drivers/infiniband/core/uverbs_std_types.c b/drivers/infiniband/core/uverbs_std_types.c
index db66c18..0a98579 100644
--- a/drivers/infiniband/core/uverbs_std_types.c
+++ b/drivers/infiniband/core/uverbs_std_types.c
@@ -249,6 +249,140 @@ static void create_udata(struct uverbs_attr_bundle *ctx,
 	INIT_UDATA_BUF_OR_NULL(udata, inbuf, outbuf, inbuf_len, outbuf_len);
 }
 
+static int uverbs_create_cq_handler(struct ib_device *ib_dev,
+				    struct ib_uverbs_file *file,
+				    struct uverbs_attr_bundle *attrs)
+{
+	struct ib_ucontext *ucontext = file->ucontext;
+	struct ib_ucq_object           *obj;
+	struct ib_udata uhw;
+	int ret;
+	u64 user_handle;
+	struct ib_cq_init_attr attr = {};
+	struct ib_cq                   *cq;
+	struct ib_uverbs_completion_event_file    *ev_file = NULL;
+	const struct uverbs_attr *ev_file_attr;
+	struct ib_uobject *ev_file_uobj;
+
+	if (!(ib_dev->uverbs_cmd_mask & 1ULL << IB_USER_VERBS_CMD_CREATE_CQ))
+		return -EOPNOTSUPP;
+
+	ret = uverbs_copy_from(&attr.comp_vector, attrs, CREATE_CQ_COMP_VECTOR);
+	if (!ret)
+		ret = uverbs_copy_from(&attr.cqe, attrs, CREATE_CQ_CQE);
+	if (!ret)
+		ret = uverbs_copy_from(&user_handle, attrs, CREATE_CQ_USER_HANDLE);
+	if (ret)
+		return ret;
+
+	/* Optional param, if it doesn't exist, we get -ENOENT and skip it */
+	if (uverbs_copy_from(&attr.flags, attrs, CREATE_CQ_FLAGS) == -EFAULT)
+		return -EFAULT;
+
+	ev_file_attr = uverbs_attr_get(attrs, CREATE_CQ_COMP_CHANNEL);
+	if (!IS_ERR(ev_file_attr)) {
+		ev_file_uobj = ev_file_attr->obj_attr.uobject;
+
+		ev_file = container_of(ev_file_uobj,
+				       struct ib_uverbs_completion_event_file,
+				       uobj_file.uobj);
+		uverbs_uobject_get(ev_file_uobj);
+	}
+
+	if (attr.comp_vector >= ucontext->ufile->device->num_comp_vectors) {
+		ret = -EINVAL;
+		goto err_event_file;
+	}
+
+	obj = container_of(uverbs_attr_get(attrs, CREATE_CQ_HANDLE)->obj_attr.uobject,
+			   typeof(*obj), uobject);
+	obj->uverbs_file	   = ucontext->ufile;
+	obj->comp_events_reported  = 0;
+	obj->async_events_reported = 0;
+	INIT_LIST_HEAD(&obj->comp_list);
+	INIT_LIST_HEAD(&obj->async_list);
+
+	/* Temporary, only until drivers get the new uverbs_attr_bundle */
+	create_udata(attrs, &uhw);
+
+	cq = ib_dev->create_cq(ib_dev, &attr, ucontext, &uhw);
+	if (IS_ERR(cq)) {
+		ret = PTR_ERR(cq);
+		goto err_event_file;
+	}
+
+	cq->device        = ib_dev;
+	cq->uobject       = &obj->uobject;
+	cq->comp_handler  = ib_uverbs_comp_handler;
+	cq->event_handler = ib_uverbs_cq_event_handler;
+	cq->cq_context    = &ev_file->ev_queue;
+	obj->uobject.object = cq;
+	obj->uobject.user_handle = user_handle;
+	atomic_set(&cq->usecnt, 0);
+
+	ret = uverbs_copy_to(attrs, CREATE_CQ_RESP_CQE, &cq->cqe);
+	if (ret)
+		goto err_cq;
+
+	return 0;
+err_cq:
+	ib_destroy_cq(cq);
+
+err_event_file:
+	if (ev_file)
+		uverbs_uobject_put(ev_file_uobj);
+	return ret;
+};
+
+static DECLARE_UVERBS_METHOD(
+	uverbs_method_cq_create, UVERBS_CQ_CREATE, uverbs_create_cq_handler,
+	&UVERBS_ATTR_IDR(CREATE_CQ_HANDLE, UVERBS_OBJECT_CQ, UVERBS_ACCESS_NEW,
+			 UA_FLAGS(UVERBS_ATTR_SPEC_F_MANDATORY)),
+	&UVERBS_ATTR_PTR_IN(CREATE_CQ_CQE, u32,
+			    UA_FLAGS(UVERBS_ATTR_SPEC_F_MANDATORY)),
+	&UVERBS_ATTR_PTR_IN(CREATE_CQ_USER_HANDLE, u64,
+			    UA_FLAGS(UVERBS_ATTR_SPEC_F_MANDATORY)),
+	&UVERBS_ATTR_FD(CREATE_CQ_COMP_CHANNEL, UVERBS_OBJECT_COMP_CHANNEL,
+			UVERBS_ACCESS_READ),
+	&UVERBS_ATTR_PTR_IN(CREATE_CQ_COMP_VECTOR, u32,
+			    UA_FLAGS(UVERBS_ATTR_SPEC_F_MANDATORY)),
+	&UVERBS_ATTR_PTR_IN(CREATE_CQ_FLAGS, u32),
+	&UVERBS_ATTR_PTR_OUT(CREATE_CQ_RESP_CQE, u32,
+			     UA_FLAGS(UVERBS_ATTR_SPEC_F_MANDATORY)),
+	&uverbs_uhw_compat_in, &uverbs_uhw_compat_out);
+
+static int uverbs_destroy_cq_handler(struct ib_device *ib_dev,
+				     struct ib_uverbs_file *file,
+				     struct uverbs_attr_bundle *attrs)
+{
+	struct ib_uverbs_destroy_cq_resp resp;
+	struct ib_uobject *uobj =
+		uverbs_attr_get(attrs, DESTROY_CQ_HANDLE)->obj_attr.uobject;
+	struct ib_ucq_object *obj = container_of(uobj, struct ib_ucq_object,
+						 uobject);
+	int ret;
+
+	if (!(ib_dev->uverbs_cmd_mask & 1ULL << IB_USER_VERBS_CMD_DESTROY_CQ))
+		return -EOPNOTSUPP;
+
+	ret = rdma_explicit_destroy(uobj);
+	if (ret)
+		return ret;
+
+	resp.comp_events_reported  = obj->comp_events_reported;
+	resp.async_events_reported = obj->async_events_reported;
+
+	return uverbs_copy_to(attrs, DESTROY_CQ_RESP, &resp);
+}
+
+static DECLARE_UVERBS_METHOD(
+	uverbs_method_cq_destroy, UVERBS_CQ_DESTROY, uverbs_destroy_cq_handler,
+	&UVERBS_ATTR_IDR(DESTROY_CQ_HANDLE, UVERBS_OBJECT_CQ,
+			 UVERBS_ACCESS_DESTROY,
+			 UA_FLAGS(UVERBS_ATTR_SPEC_F_MANDATORY)),
+	&UVERBS_ATTR_PTR_OUT(DESTROY_CQ_RESP, struct ib_uverbs_destroy_cq_resp,
+			     UA_FLAGS(UVERBS_ATTR_SPEC_F_MANDATORY)));
+
 DECLARE_UVERBS_OBJECT(uverbs_object_comp_channel,
 		      UVERBS_OBJECT_COMP_CHANNEL,
 		      &UVERBS_TYPE_ALLOC_FD(0,
@@ -259,7 +393,9 @@ static void create_udata(struct uverbs_attr_bundle *ctx,
 
 DECLARE_UVERBS_OBJECT(uverbs_object_cq, UVERBS_OBJECT_CQ,
 		      &UVERBS_TYPE_ALLOC_IDR_SZ(sizeof(struct ib_ucq_object), 0,
-						  uverbs_free_cq));
+						  uverbs_free_cq),
+		      &uverbs_method_cq_create,
+		      &uverbs_method_cq_destroy);
 
 DECLARE_UVERBS_OBJECT(uverbs_object_qp, UVERBS_OBJECT_QP,
 		      &UVERBS_TYPE_ALLOC_IDR_SZ(sizeof(struct ib_uqp_object), 0,
diff --git a/include/uapi/rdma/ib_user_ioctl_verbs.h b/include/uapi/rdma/ib_user_ioctl_verbs.h
index 90f81ee..842792e 100644
--- a/include/uapi/rdma/ib_user_ioctl_verbs.h
+++ b/include/uapi/rdma/ib_user_ioctl_verbs.h
@@ -60,5 +60,25 @@ enum {
 	UVERBS_UHW_OUT,
 };
 
+enum uverbs_create_cq_cmd_attr_ids {
+	CREATE_CQ_HANDLE,
+	CREATE_CQ_CQE,
+	CREATE_CQ_USER_HANDLE,
+	CREATE_CQ_COMP_CHANNEL,
+	CREATE_CQ_COMP_VECTOR,
+	CREATE_CQ_FLAGS,
+	CREATE_CQ_RESP_CQE,
+};
+
+enum uverbs_destroy_cq_cmd_attr_ids {
+	DESTROY_CQ_HANDLE,
+	DESTROY_CQ_RESP,
+};
+
+enum uverbs_actions_cq_ops {
+	UVERBS_CQ_CREATE,
+	UVERBS_CQ_DESTROY,
+};
+
 #endif
 
-- 
1.8.3.1

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH rdma-next V1 12/13] IB/core: Assign root to all drivers

[Matan Barak]

In order to use the parsing tree, we need to assign the root
to all drivers. Currently, we just assign the default parsing
tree via ib_uverbs_add_one. The driver could override this by
assigning a parsing tree prior to registering the device.

Signed-off-by: Matan Barak <matanb-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
Reviewed-by: Yishai Hadas <yishaih-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
---
 drivers/infiniband/core/uverbs.h      |  1 +
 drivers/infiniband/core/uverbs_main.c | 18 ++++++++++++++++++
 include/rdma/uverbs_ioctl.h           | 12 ++++++++++++
 include/rdma/uverbs_std_types.h       | 14 ++++++++++++++
 4 files changed, 45 insertions(+)

diff --git a/drivers/infiniband/core/uverbs.h b/drivers/infiniband/core/uverbs.h
index 64d494a..0f6f768 100644
--- a/drivers/infiniband/core/uverbs.h
+++ b/drivers/infiniband/core/uverbs.h
@@ -100,6 +100,7 @@ struct ib_uverbs_device {
 	struct mutex				lists_mutex; /* protect lists */
 	struct list_head			uverbs_file_list;
 	struct list_head			uverbs_events_file_list;
+	struct uverbs_root_spec			*specs_root;
 };
 
 struct ib_uverbs_event_queue {
diff --git a/drivers/infiniband/core/uverbs_main.c b/drivers/infiniband/core/uverbs_main.c
index 3d26096..0fcea1b 100644
--- a/drivers/infiniband/core/uverbs_main.c
+++ b/drivers/infiniband/core/uverbs_main.c
@@ -49,6 +49,7 @@
 #include <linux/uaccess.h>
 
 #include <rdma/ib.h>
+#include <rdma/uverbs_std_types.h>
 
 #include "uverbs.h"
 #include "core_priv.h"
@@ -1109,6 +1110,18 @@ 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->specs_root) {
+		const struct uverbs_object_tree_def *default_root[] = {
+			uverbs_default_get_objects()};
+
+		uverbs_dev->specs_root = uverbs_alloc_spec_tree(1,
+								default_root);
+		if (IS_ERR(uverbs_dev->specs_root))
+			goto err_class;
+
+		device->specs_root = uverbs_dev->specs_root;
+	}
+
 	ib_set_client_data(device, &uverbs_client, uverbs_dev);
 
 	return;
@@ -1240,6 +1253,11 @@ static void ib_uverbs_remove_one(struct ib_device *device, void *client_data)
 		ib_uverbs_comp_dev(uverbs_dev);
 	if (wait_clients)
 		wait_for_completion(&uverbs_dev->comp);
+	if (uverbs_dev->specs_root) {
+		uverbs_free_spec_tree(uverbs_dev->specs_root);
+		device->specs_root = NULL;
+	}
+
 	kobject_put(&uverbs_dev->kobj);
 }
 
diff --git a/include/rdma/uverbs_ioctl.h b/include/rdma/uverbs_ioctl.h
index 759afa0..6da4407 100644
--- a/include/rdma/uverbs_ioctl.h
+++ b/include/rdma/uverbs_ioctl.h
@@ -419,8 +419,20 @@ static inline int _uverbs_copy_from(void *to, size_t to_size,
  * An object without any methods is considered invalid and will abort the
  * function with -ENOENT error.
  */
+#if IS_ENABLED(CONFIG_INFINIBAND_USER_ACCESS)
 struct uverbs_root_spec *uverbs_alloc_spec_tree(unsigned int num_trees,
 						const struct uverbs_object_tree_def **trees);
 void uverbs_free_spec_tree(struct uverbs_root_spec *root);
+#else
+static inline struct uverbs_root_spec *uverbs_alloc_spec_tree(unsigned int num_trees,
+							      const struct uverbs_object_tree_def **trees)
+{
+	return NULL;
+}
+
+static inline void uverbs_free_spec_tree(struct uverbs_root_spec *root)
+{
+}
+#endif
 
 #endif
diff --git a/include/rdma/uverbs_std_types.h b/include/rdma/uverbs_std_types.h
index 400efe2..5f8e20b 100644
--- a/include/rdma/uverbs_std_types.h
+++ b/include/rdma/uverbs_std_types.h
@@ -34,8 +34,10 @@
 #define _UVERBS_STD_TYPES__
 
 #include <rdma/uverbs_types.h>
+#include <rdma/uverbs_ioctl.h>
 #include <rdma/ib_user_ioctl_verbs.h>
 
+#if IS_ENABLED(CONFIG_INFINIBAND_USER_ACCESS)
 extern const struct uverbs_object_def uverbs_object_comp_channel;
 extern const struct uverbs_object_def uverbs_object_cq;
 extern const struct uverbs_object_def uverbs_object_qp;
@@ -50,6 +52,18 @@
 extern const struct uverbs_object_def uverbs_object_xrcd;
 extern const struct uverbs_object_def uverbs_object_device;
 
+extern const struct uverbs_object_tree_def uverbs_default_objects;
+static inline const struct uverbs_object_tree_def *uverbs_default_get_objects(void)
+{
+	return &uverbs_default_objects;
+}
+#else
+static inline const struct uverbs_object_tree_def *uverbs_default_get_objects(void)
+{
+	return NULL;
+}
+#endif
+
 static inline struct ib_uobject *__uobj_get(const struct uverbs_obj_type *type,
 					    bool write,
 					    struct ib_ucontext *ucontext,
-- 
1.8.3.1

[PATCH rdma-next V1 13/13] IB/core: Expose ioctl interface through experimental Kconfig

[Matan Barak]

Add CONFIG_INFINIBAND_EXP_USER_ACCESS that enables the ioctl
interface. This interface is experimental and is subject to change.

Signed-off-by: Matan Barak <matanb-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
Reviewed-by: Yishai Hadas <yishaih-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
---
 drivers/infiniband/Kconfig            | 9 +++++++++
 drivers/infiniband/core/uverbs.h      | 2 ++
 drivers/infiniband/core/uverbs_main.c | 6 ++++++
 3 files changed, 17 insertions(+)

diff --git a/drivers/infiniband/Kconfig b/drivers/infiniband/Kconfig
index 234fe01..3726205 100644
--- a/drivers/infiniband/Kconfig
+++ b/drivers/infiniband/Kconfig
@@ -34,6 +34,15 @@ config INFINIBAND_USER_ACCESS
 	  libibverbs, libibcm and a hardware driver library from
 	  <http://www.openfabrics.org/git/>.
 
+config INFINIBAND_EXP_USER_ACCESS
+	bool "Allow experimental support for Infiniband ABI"
+	depends on INFINIBAND_USER_ACCESS
+	---help---
+	  IOCTL based ABI support for Infiniband. This allows userspace
+	  to invoke the experimental IOCTL based ABI.
+	  These commands are parsed via per-device parsing tree and
+	  enables per-device features.
+
 config INFINIBAND_USER_MEM
 	bool
 	depends on INFINIBAND_USER_ACCESS != n
diff --git a/drivers/infiniband/core/uverbs.h b/drivers/infiniband/core/uverbs.h
index 0f6f768..37c8903 100644
--- a/drivers/infiniband/core/uverbs.h
+++ b/drivers/infiniband/core/uverbs.h
@@ -219,6 +219,8 @@ int ib_uverbs_dealloc_xrcd(struct ib_uverbs_device *dev, struct ib_xrcd *xrcd,
 void ib_uverbs_detach_umcast(struct ib_qp *qp,
 			     struct ib_uqp_object *uobj);
 
+long ib_uverbs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg);
+
 struct ib_uverbs_flow_spec {
 	union {
 		union {
diff --git a/drivers/infiniband/core/uverbs_main.c b/drivers/infiniband/core/uverbs_main.c
index 0fcea1b..a88d223 100644
--- a/drivers/infiniband/core/uverbs_main.c
+++ b/drivers/infiniband/core/uverbs_main.c
@@ -951,6 +951,9 @@ static int ib_uverbs_close(struct inode *inode, struct file *filp)
 	.open	 = ib_uverbs_open,
 	.release = ib_uverbs_close,
 	.llseek	 = no_llseek,
+#if IS_ENABLED(CONFIG_INFINIBAND_EXP_USER_ACCESS)
+	.unlocked_ioctl = ib_uverbs_ioctl,
+#endif
 };
 
 static const struct file_operations uverbs_mmap_fops = {
@@ -960,6 +963,9 @@ static int ib_uverbs_close(struct inode *inode, struct file *filp)
 	.open	 = ib_uverbs_open,
 	.release = ib_uverbs_close,
 	.llseek	 = no_llseek,
+#if IS_ENABLED(CONFIG_INFINIBAND_EXP_USER_ACCESS)
+	.unlocked_ioctl = ib_uverbs_ioctl,
+#endif
 };
 
 static struct ib_client uverbs_client = {
-- 
1.8.3.1

Re: [PATCH rdma-next V1 00/13] [PATCH V1 for-next 00/13] IB/core: SG IOCTL based RDMA ABI

[Doug Ledford]

On Thu, 2017-08-03 at 16:06 +0300, Matan Barak wrote:
> Hi Doug,
> 
> This patch series adds ioctl() interface to the existing write()
> interface and
> provide an easy route to backport this change to legacy supported
> systems.

[ snip ]

Hi Matan,

Thanks for getting this to the linux-api mailing list, I was about to
suggest the same.

-- 
Doug Ledford <dledford-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>
    GPG KeyID: B826A3330E572FDD
    Key fingerprint = AE6B 1BDA 122B 23B4 265B  1274 B826 A333 0E57 2FDD

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH rdma-next V1 00/13] [PATCH V1 for-next 00/13] IB/core: SG IOCTL based RDMA ABI

[Doug Ledford]

On Thu, 2017-08-03 at 11:17 -0400, Doug Ledford wrote:
> On Thu, 2017-08-03 at 16:06 +0300, Matan Barak wrote:
> > Hi Doug,
> > 
> > This patch series adds ioctl() interface to the existing write()
> > interface and
> > provide an easy route to backport this change to legacy supported
> > systems.
> 
> [ snip ]
> 
> Hi Matan,
> 
> Thanks for getting this to the linux-api mailing list, I was about to
> suggest the same.

This has now been on the linux-api mailing list for 21 days with no
objections.  I very well may take this into my for-next branch, so
should anyone want to look at this before hand, now is the time. 
Thanks.

-- 
Doug Ledford <dledford-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>
    GPG KeyID: B826A3330E572FDD
    Key fingerprint = AE6B 1BDA 122B 23B4 265B  1274 B826 A333 0E57 2FDD

Re: [PATCH rdma-next V1 00/13] [PATCH V1 for-next 00/13] IB/core: SG IOCTL based RDMA ABI

[Jason Gunthorpe]

On Thu, Aug 24, 2017 at 06:06:08PM -0400, Doug Ledford wrote:
> On Thu, 2017-08-03 at 11:17 -0400, Doug Ledford wrote:
> > On Thu, 2017-08-03 at 16:06 +0300, Matan Barak wrote:
> > > Hi Doug,
> > > 
> > > This patch series adds ioctl() interface to the existing write()
> > > interface and
> > > provide an easy route to backport this change to legacy supported
> > > systems.
> > 
> > [ snip ]
> > 
> > Hi Matan,
> > 
> > Thanks for getting this to the linux-api mailing list, I was about to
> > suggest the same.
> 
> This has now been on the linux-api mailing list for 21 days with no
> objections.  I very well may take this into my for-next branch, so
> should anyone want to look at this before hand, now is the time. 
> Thanks.

I think you should wait until after plumbers.

Jason
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

RE: [PATCH rdma-next V1 00/13] [PATCH V1 for-next 00/13] IB/core: SG IOCTL based RDMA ABI

[Hefty, Sean]

> > This has now been on the linux-api mailing list for 21 days with no
> > objections.  I very well may take this into my for-next branch, so
> > should anyone want to look at this before hand, now is the time.
> > Thanks.
> 
> I think you should wait until after plumbers.

I tried to review the code and gave up.  I still find it overly complex for implementing what is conceptually a function dispatch table.

RE: [PATCH rdma-next V1 00/13] [PATCH V1 for-next 00/13] IB/core: SG IOCTL based RDMA ABI

[Christopher Lameter]

On Fri, 25 Aug 2017, Hefty, Sean wrote:

> > I think you should wait until after plumbers.
>
> I tried to review the code and gave up.  I still find it overly complex for implementing what is conceptually a function dispatch table.

Welll this fixes a long standing major issue in the ABI. No one else is
working on it so its the only way to go. I think this ought to be merged
ASAP. If it breaks something then at least everyone is affected and will
be fixing this thing up so we can finally move beyond the rpc via file ops
issue.


--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH rdma-next V1 00/13] [PATCH V1 for-next 00/13] IB/core: SG IOCTL based RDMA ABI

[Matan Barak]

On 25/08/2017 21:27, Christopher Lameter wrote:
> On Fri, 25 Aug 2017, Hefty, Sean wrote:
>
>>> I think you should wait until after plumbers.
>>
>> I tried to review the code and gave up.  I still find it overly complex for implementing what is conceptually a function dispatch table.
>

Every such an API could be defined as a way to trigger operations 
(functions) from user-space to kernel....
It doesn't necessarily mean that when you take all requirements into 
account it falls into a simple "function dispatch table" (i.e. netlink).

We all know the current uverbs design isn't flexible enough - hfi1 went 
and implemented its own ABI and didn't use the RDMA subsystem default 
one.... The goal of all of this is to have a suitable interface between 
user-space and kernel that fits all the required RDMA consumers and 
user-space drivers. This is mandatory giving the new fabrics and 
features added lately to this subsystem.

Anyway, we established the concepts in the RDMA mailing list, linux 
plumbers, OFVWG and OFA. There were a lot of RFC rounds. V0 of the 
patches got comments, which are fixed here.

> Welll this fixes a long standing major issue in the ABI. No one else is
> working on it so its the only way to go. I think this ought to be merged
> ASAP. If it breaks something then at least everyone is affected and will
> be fixing this thing up so we can finally move beyond the rpc via file ops
> issue.
>
>

I totally agree. Most of the involved people attended at least some of 
the conferences/meetings. The patches and RFCs exist in the ML for a 
long time. It shouldn't break anything as currently it's just an 
experimental interface along what we have now, but it'll allow people to 
start test this infrastructure, port verbs and write new features.

Re: [PATCH rdma-next V1 00/13] [PATCH V1 for-next 00/13] IB/core: SG IOCTL based RDMA ABI

[Doug Ledford]

On Fri, 2017-08-25 at 17:20 +0000, Hefty, Sean wrote:
> > > This has now been on the linux-api mailing list for 21 days with
> > > no
> > > objections.  I very well may take this into my for-next branch,
> > > so
> > > should anyone want to look at this before hand, now is the time.
> > > Thanks.
> > 
> > I think you should wait until after plumbers.
> 
> I tried to review the code and gave up.  I still find it overly
> complex for implementing what is conceptually a function dispatch
> table.

Maybe, but it's encased in a Kconfig item that marks it experimental,
so including it doesn't freeze it in stone.  As such, I took it because
we really need to make progress on this and not just keep pushing it on
out.  Now is the time to focus and raise objections, send fixes, etc. 
Otherwise it'll be what it'll be.

-- 
Doug Ledford <dledford-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>
    GPG KeyID: B826A3330E572FDD
    Key fingerprint = AE6B 1BDA 122B 23B4 265B  1274 B826 A333 0E57 2FDD