OpenCloudOS-Kernel/drivers/infiniband/core/rdma_core.c

782 lines
21 KiB
C
Raw Normal View History

/*
* Copyright (c) 2016, 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 <linux/file.h>
#include <linux/anon_inodes.h>
#include <rdma/ib_verbs.h>
#include <rdma/uverbs_types.h>
#include <linux/rcupdate.h>
IB/core: Add a generic way to execute an operation on a uobject 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@mellanox.com> Reviewed-by: Yishai Hadas <yishaih@mellanox.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
2017-08-03 21:06:55 +08:00
#include <rdma/uverbs_ioctl.h>
IB/core: Add new ioctl interface 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@mellanox.com> Reviewed-by: Yishai Hadas <yishaih@mellanox.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
2017-08-03 21:06:57 +08:00
#include <rdma/rdma_user_ioctl.h>
#include "uverbs.h"
#include "core_priv.h"
#include "rdma_core.h"
IB/core: Add new ioctl interface 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@mellanox.com> Reviewed-by: Yishai Hadas <yishaih@mellanox.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
2017-08-03 21:06:57 +08:00
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);
}
static void uverbs_uobject_free(struct kref *ref)
{
struct ib_uobject *uobj =
container_of(ref, struct ib_uobject, ref);
if (uobj->type->type_class->needs_kfree_rcu)
kfree_rcu(uobj, rcu);
else
kfree(uobj);
}
void uverbs_uobject_put(struct ib_uobject *uobject)
{
kref_put(&uobject->ref, uverbs_uobject_free);
}
static int uverbs_try_lock_object(struct ib_uobject *uobj, bool exclusive)
{
/*
* When a shared access is required, we use a positive counter. Each
* shared access request checks that the value != -1 and increment it.
* Exclusive access is required for operations like write or destroy.
* In exclusive access mode, we check that the counter is zero (nobody
* claimed this object) and we set it to -1. Releasing a shared access
* lock is done simply by decreasing the counter. As for exclusive
* access locks, since only a single one of them is is allowed
* concurrently, setting the counter to zero is enough for releasing
* this lock.
*/
if (!exclusive)
return __atomic_add_unless(&uobj->usecnt, 1, -1) == -1 ?
-EBUSY : 0;
/* lock is either WRITE or DESTROY - should be exclusive */
return atomic_cmpxchg(&uobj->usecnt, 0, -1) == 0 ? 0 : -EBUSY;
}
static struct ib_uobject *alloc_uobj(struct ib_ucontext *context,
const struct uverbs_obj_type *type)
{
struct ib_uobject *uobj = kzalloc(type->obj_size, GFP_KERNEL);
if (!uobj)
return ERR_PTR(-ENOMEM);
/*
* user_handle should be filled by the handler,
* The object is added to the list in the commit stage.
*/
uobj->context = context;
uobj->type = type;
/*
* Allocated objects start out as write locked to deny any other
* syscalls from accessing them until they are committed. See
* rdma_alloc_commit_uobject
*/
atomic_set(&uobj->usecnt, -1);
kref_init(&uobj->ref);
return uobj;
}
static int idr_add_uobj(struct ib_uobject *uobj)
{
int ret;
idr_preload(GFP_KERNEL);
spin_lock(&uobj->context->ufile->idr_lock);
/*
* We start with allocating an idr pointing to NULL. This represents an
* object which isn't initialized yet. We'll replace it later on with
* the real object once we commit.
*/
ret = idr_alloc(&uobj->context->ufile->idr, NULL, 0,
min_t(unsigned long, U32_MAX - 1, INT_MAX), GFP_NOWAIT);
if (ret >= 0)
uobj->id = ret;
spin_unlock(&uobj->context->ufile->idr_lock);
idr_preload_end();
return ret < 0 ? ret : 0;
}
/*
* It only removes it from the uobjects list, uverbs_uobject_put() is still
* required.
*/
static void uverbs_idr_remove_uobj(struct ib_uobject *uobj)
{
spin_lock(&uobj->context->ufile->idr_lock);
idr_remove(&uobj->context->ufile->idr, uobj->id);
spin_unlock(&uobj->context->ufile->idr_lock);
}
/* Returns the ib_uobject or an error. The caller should check for IS_ERR. */
static struct ib_uobject *lookup_get_idr_uobject(const struct uverbs_obj_type *type,
struct ib_ucontext *ucontext,
int id, bool exclusive)
{
struct ib_uobject *uobj;
rcu_read_lock();
/* object won't be released as we're protected in rcu */
uobj = idr_find(&ucontext->ufile->idr, id);
if (!uobj) {
uobj = ERR_PTR(-ENOENT);
goto free;
}
RDMA/uverbs: Protect from races between lookup and destroy of uobjects The race is between lookup_get_idr_uobject and uverbs_idr_remove_uobj -> uverbs_uobject_put. We deliberately do not call sychronize_rcu after the idr_remove in uverbs_idr_remove_uobj for performance reasons, instead we call kfree_rcu() during uverbs_uobject_put. However, this means we can obtain pointers to uobj's that have already been released and must protect against krefing them using kref_get_unless_zero. ================================================================== BUG: KASAN: use-after-free in copy_ah_attr_from_uverbs.isra.2+0x860/0xa00 Read of size 4 at addr ffff88005fda1ac8 by task syz-executor2/441 CPU: 1 PID: 441 Comm: syz-executor2 Not tainted 4.15.0-rc2+ #56 Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.7.5-0-ge51488c-20140602_164612-nilsson.home.kraxel.org 04/01/2014 Call Trace: dump_stack+0x8d/0xd4 print_address_description+0x73/0x290 kasan_report+0x25c/0x370 ? copy_ah_attr_from_uverbs.isra.2+0x860/0xa00 copy_ah_attr_from_uverbs.isra.2+0x860/0xa00 ? uverbs_try_lock_object+0x68/0xc0 ? modify_qp.isra.7+0xdc4/0x10e0 modify_qp.isra.7+0xdc4/0x10e0 ib_uverbs_modify_qp+0xfe/0x170 ? ib_uverbs_query_qp+0x970/0x970 ? __lock_acquire+0xa11/0x1da0 ib_uverbs_write+0x55a/0xad0 ? ib_uverbs_query_qp+0x970/0x970 ? ib_uverbs_query_qp+0x970/0x970 ? ib_uverbs_open+0x760/0x760 ? futex_wake+0x147/0x410 ? sched_clock_cpu+0x18/0x180 ? check_prev_add+0x1680/0x1680 ? do_futex+0x3b6/0xa30 ? sched_clock_cpu+0x18/0x180 __vfs_write+0xf7/0x5c0 ? ib_uverbs_open+0x760/0x760 ? kernel_read+0x110/0x110 ? lock_acquire+0x370/0x370 ? __fget+0x264/0x3b0 vfs_write+0x18a/0x460 SyS_write+0xc7/0x1a0 ? SyS_read+0x1a0/0x1a0 ? trace_hardirqs_on_thunk+0x1a/0x1c entry_SYSCALL_64_fastpath+0x18/0x85 RIP: 0033:0x448e29 RSP: 002b:00007f443fee0c58 EFLAGS: 00000246 ORIG_RAX: 0000000000000001 RAX: ffffffffffffffda RBX: 00007f443fee16bc RCX: 0000000000448e29 RDX: 0000000000000078 RSI: 00000000209f8000 RDI: 0000000000000012 RBP: 000000000070bea0 R08: 0000000000000000 R09: 0000000000000000 R10: 0000000000000000 R11: 0000000000000246 R12: 00000000ffffffff R13: 0000000000008e98 R14: 00000000006ebf38 R15: 0000000000000000 Allocated by task 1: kmem_cache_alloc_trace+0x16c/0x2f0 mlx5_alloc_cmd_msg+0x12e/0x670 cmd_exec+0x419/0x1810 mlx5_cmd_exec+0x40/0x70 mlx5_core_mad_ifc+0x187/0x220 mlx5_MAD_IFC+0xd7/0x1b0 mlx5_query_mad_ifc_gids+0x1f3/0x650 mlx5_ib_query_gid+0xa4/0xc0 ib_query_gid+0x152/0x1a0 ib_query_port+0x21e/0x290 mlx5_port_immutable+0x30f/0x490 ib_register_device+0x5dd/0x1130 mlx5_ib_add+0x3e7/0x700 mlx5_add_device+0x124/0x510 mlx5_register_interface+0x11f/0x1c0 mlx5_ib_init+0x56/0x61 do_one_initcall+0xa3/0x250 kernel_init_freeable+0x309/0x3b8 kernel_init+0x14/0x180 ret_from_fork+0x24/0x30 Freed by task 1: kfree+0xeb/0x2f0 mlx5_free_cmd_msg+0xcd/0x140 cmd_exec+0xeba/0x1810 mlx5_cmd_exec+0x40/0x70 mlx5_core_mad_ifc+0x187/0x220 mlx5_MAD_IFC+0xd7/0x1b0 mlx5_query_mad_ifc_gids+0x1f3/0x650 mlx5_ib_query_gid+0xa4/0xc0 ib_query_gid+0x152/0x1a0 ib_query_port+0x21e/0x290 mlx5_port_immutable+0x30f/0x490 ib_register_device+0x5dd/0x1130 mlx5_ib_add+0x3e7/0x700 mlx5_add_device+0x124/0x510 mlx5_register_interface+0x11f/0x1c0 mlx5_ib_init+0x56/0x61 do_one_initcall+0xa3/0x250 kernel_init_freeable+0x309/0x3b8 kernel_init+0x14/0x180 ret_from_fork+0x24/0x30 The buggy address belongs to the object at ffff88005fda1ab0 which belongs to the cache kmalloc-32 of size 32 The buggy address is located 24 bytes inside of 32-byte region [ffff88005fda1ab0, ffff88005fda1ad0) The buggy address belongs to the page: page:00000000d5655c19 count:1 mapcount:0 mapping: (null) index:0xffff88005fda1fc0 flags: 0x4000000000000100(slab) raw: 4000000000000100 0000000000000000 ffff88005fda1fc0 0000000180550008 raw: ffffea00017f6780 0000000400000004 ffff88006c803980 0000000000000000 page dumped because: kasan: bad access detected Memory state around the buggy address: ffff88005fda1980: fc fc fb fb fb fb fc fc fb fb fb fb fc fc fb fb ffff88005fda1a00: fb fb fc fc fb fb fb fb fc fc 00 00 00 00 fc fc ffff88005fda1a80: fb fb fb fb fc fc fb fb fb fb fc fc fb fb fb fb ffff88005fda1b00: fc fc 00 00 00 00 fc fc fb fb fb fb fc fc fb fb ffff88005fda1b80: fb fb fc fc fb fb fb fb fc fc fb fb fb fb fc fc ==================================================================@ Cc: syzkaller <syzkaller@googlegroups.com> Cc: <stable@vger.kernel.org> # 4.11 Fixes: 3832125624b7 ("IB/core: Add support for idr types") Reported-by: Noa Osherovich <noaos@mellanox.com> Signed-off-by: Leon Romanovsky <leonro@mellanox.com> Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
2018-02-13 18:18:37 +08:00
/*
* The idr_find is guaranteed to return a pointer to something that
* isn't freed yet, or NULL, as the free after idr_remove goes through
* kfree_rcu(). However the object may still have been released and
* kfree() could be called at any time.
*/
if (!kref_get_unless_zero(&uobj->ref))
uobj = ERR_PTR(-ENOENT);
free:
rcu_read_unlock();
return uobj;
}
static struct ib_uobject *lookup_get_fd_uobject(const struct uverbs_obj_type *type,
struct ib_ucontext *ucontext,
int id, bool exclusive)
{
struct file *f;
struct ib_uobject *uobject;
const struct uverbs_obj_fd_type *fd_type =
container_of(type, struct uverbs_obj_fd_type, type);
if (exclusive)
return ERR_PTR(-EOPNOTSUPP);
f = fget(id);
if (!f)
return ERR_PTR(-EBADF);
uobject = f->private_data;
/*
* fget(id) ensures we are not currently running uverbs_close_fd,
* and the caller is expected to ensure that uverbs_close_fd is never
* done while a call top lookup is possible.
*/
if (f->f_op != fd_type->fops) {
fput(f);
return ERR_PTR(-EBADF);
}
uverbs_uobject_get(uobject);
return uobject;
}
struct ib_uobject *rdma_lookup_get_uobject(const struct uverbs_obj_type *type,
struct ib_ucontext *ucontext,
int id, bool exclusive)
{
struct ib_uobject *uobj;
int ret;
uobj = type->type_class->lookup_get(type, ucontext, id, exclusive);
if (IS_ERR(uobj))
return uobj;
if (uobj->type != type) {
ret = -EINVAL;
goto free;
}
ret = uverbs_try_lock_object(uobj, exclusive);
if (ret) {
WARN(ucontext->cleanup_reason,
"ib_uverbs: Trying to lookup_get while cleanup context\n");
goto free;
}
return uobj;
free:
uobj->type->type_class->lookup_put(uobj, exclusive);
uverbs_uobject_put(uobj);
return ERR_PTR(ret);
}
static struct ib_uobject *alloc_begin_idr_uobject(const struct uverbs_obj_type *type,
struct ib_ucontext *ucontext)
{
int ret;
struct ib_uobject *uobj;
uobj = alloc_uobj(ucontext, type);
if (IS_ERR(uobj))
return uobj;
ret = idr_add_uobj(uobj);
if (ret)
goto uobj_put;
ret = ib_rdmacg_try_charge(&uobj->cg_obj, ucontext->device,
RDMACG_RESOURCE_HCA_OBJECT);
if (ret)
goto idr_remove;
return uobj;
idr_remove:
uverbs_idr_remove_uobj(uobj);
uobj_put:
uverbs_uobject_put(uobj);
return ERR_PTR(ret);
}
static struct ib_uobject *alloc_begin_fd_uobject(const struct uverbs_obj_type *type,
struct ib_ucontext *ucontext)
{
const struct uverbs_obj_fd_type *fd_type =
container_of(type, struct uverbs_obj_fd_type, type);
int new_fd;
struct ib_uobject *uobj;
struct ib_uobject_file *uobj_file;
struct file *filp;
new_fd = get_unused_fd_flags(O_CLOEXEC);
if (new_fd < 0)
return ERR_PTR(new_fd);
uobj = alloc_uobj(ucontext, type);
if (IS_ERR(uobj)) {
put_unused_fd(new_fd);
return uobj;
}
uobj_file = container_of(uobj, struct ib_uobject_file, uobj);
filp = anon_inode_getfile(fd_type->name,
fd_type->fops,
uobj_file,
fd_type->flags);
if (IS_ERR(filp)) {
put_unused_fd(new_fd);
uverbs_uobject_put(uobj);
return (void *)filp;
}
uobj_file->uobj.id = new_fd;
uobj_file->uobj.object = filp;
uobj_file->ufile = ucontext->ufile;
INIT_LIST_HEAD(&uobj->list);
kref_get(&uobj_file->ufile->ref);
return uobj;
}
struct ib_uobject *rdma_alloc_begin_uobject(const struct uverbs_obj_type *type,
struct ib_ucontext *ucontext)
{
return type->type_class->alloc_begin(type, ucontext);
}
static int __must_check remove_commit_idr_uobject(struct ib_uobject *uobj,
enum rdma_remove_reason why)
{
const struct uverbs_obj_idr_type *idr_type =
container_of(uobj->type, struct uverbs_obj_idr_type,
type);
int ret = idr_type->destroy_object(uobj, why);
/*
* We can only fail gracefully if the user requested to destroy the
* object. In the rest of the cases, just remove whatever you can.
*/
if (why == RDMA_REMOVE_DESTROY && ret)
return ret;
ib_rdmacg_uncharge(&uobj->cg_obj, uobj->context->device,
RDMACG_RESOURCE_HCA_OBJECT);
uverbs_idr_remove_uobj(uobj);
return ret;
}
static void alloc_abort_fd_uobject(struct ib_uobject *uobj)
{
struct ib_uobject_file *uobj_file =
container_of(uobj, struct ib_uobject_file, uobj);
struct file *filp = uobj->object;
int id = uobj_file->uobj.id;
/* Unsuccessful NEW */
fput(filp);
put_unused_fd(id);
}
static int __must_check remove_commit_fd_uobject(struct ib_uobject *uobj,
enum rdma_remove_reason why)
{
const struct uverbs_obj_fd_type *fd_type =
container_of(uobj->type, struct uverbs_obj_fd_type, type);
struct ib_uobject_file *uobj_file =
container_of(uobj, struct ib_uobject_file, uobj);
int ret = fd_type->context_closed(uobj_file, why);
if (why == RDMA_REMOVE_DESTROY && ret)
return ret;
if (why == RDMA_REMOVE_DURING_CLEANUP) {
alloc_abort_fd_uobject(uobj);
return ret;
}
uobj_file->uobj.context = NULL;
return ret;
}
static void assert_uverbs_usecnt(struct ib_uobject *uobj, bool exclusive)
{
#ifdef CONFIG_LOCKDEP
if (exclusive)
WARN_ON(atomic_read(&uobj->usecnt) != -1);
else
WARN_ON(atomic_read(&uobj->usecnt) <= 0);
#endif
}
static int __must_check _rdma_remove_commit_uobject(struct ib_uobject *uobj,
enum rdma_remove_reason why)
{
int ret;
struct ib_ucontext *ucontext = uobj->context;
ret = uobj->type->type_class->remove_commit(uobj, why);
if (ret && why == RDMA_REMOVE_DESTROY) {
/* We couldn't remove the object, so just unlock the uobject */
atomic_set(&uobj->usecnt, 0);
uobj->type->type_class->lookup_put(uobj, true);
} else {
mutex_lock(&ucontext->uobjects_lock);
list_del(&uobj->list);
mutex_unlock(&ucontext->uobjects_lock);
/* put the ref we took when we created the object */
uverbs_uobject_put(uobj);
}
return ret;
}
/* This is called only for user requested DESTROY reasons */
int __must_check rdma_remove_commit_uobject(struct ib_uobject *uobj)
{
int ret;
struct ib_ucontext *ucontext = uobj->context;
/* put the ref count we took at lookup_get */
uverbs_uobject_put(uobj);
/* 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;
}
assert_uverbs_usecnt(uobj, true);
ret = _rdma_remove_commit_uobject(uobj, RDMA_REMOVE_DESTROY);
up_read(&ucontext->cleanup_rwsem);
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;
}
assert_uverbs_usecnt(uobject, true);
ret = uobject->type->type_class->remove_commit(uobject,
RDMA_REMOVE_DESTROY);
if (ret)
goto out;
uobject->type = &null_obj_type;
out:
up_read(&ucontext->cleanup_rwsem);
return ret;
}
static void alloc_commit_idr_uobject(struct ib_uobject *uobj)
{
spin_lock(&uobj->context->ufile->idr_lock);
/*
* We already allocated this IDR with a NULL object, so
* this shouldn't fail.
*/
WARN_ON(idr_replace(&uobj->context->ufile->idr,
uobj, uobj->id));
spin_unlock(&uobj->context->ufile->idr_lock);
}
static void alloc_commit_fd_uobject(struct ib_uobject *uobj)
{
struct ib_uobject_file *uobj_file =
container_of(uobj, struct ib_uobject_file, uobj);
fd_install(uobj_file->uobj.id, uobj->object);
/* This shouldn't be used anymore. Use the file object instead */
uobj_file->uobj.id = 0;
/* Get another reference as we export this to the fops */
uverbs_uobject_get(&uobj_file->uobj);
}
int rdma_alloc_commit_uobject(struct ib_uobject *uobj)
{
/* Cleanup is running. Calling this should have been impossible */
if (!down_read_trylock(&uobj->context->cleanup_rwsem)) {
int ret;
WARN(true, "ib_uverbs: Cleanup is running while allocating an uobject\n");
ret = uobj->type->type_class->remove_commit(uobj,
RDMA_REMOVE_DURING_CLEANUP);
if (ret)
pr_warn("ib_uverbs: cleanup of idr object %d failed\n",
uobj->id);
return ret;
}
/* matches atomic_set(-1) in alloc_uobj */
assert_uverbs_usecnt(uobj, true);
atomic_set(&uobj->usecnt, 0);
mutex_lock(&uobj->context->uobjects_lock);
list_add(&uobj->list, &uobj->context->uobjects);
mutex_unlock(&uobj->context->uobjects_lock);
uobj->type->type_class->alloc_commit(uobj);
up_read(&uobj->context->cleanup_rwsem);
return 0;
}
static void alloc_abort_idr_uobject(struct ib_uobject *uobj)
{
uverbs_idr_remove_uobj(uobj);
ib_rdmacg_uncharge(&uobj->cg_obj, uobj->context->device,
RDMACG_RESOURCE_HCA_OBJECT);
uverbs_uobject_put(uobj);
}
void rdma_alloc_abort_uobject(struct ib_uobject *uobj)
{
uobj->type->type_class->alloc_abort(uobj);
}
static void lookup_put_idr_uobject(struct ib_uobject *uobj, bool exclusive)
{
}
static void lookup_put_fd_uobject(struct ib_uobject *uobj, bool exclusive)
{
struct file *filp = uobj->object;
WARN_ON(exclusive);
/* This indirectly calls uverbs_close_fd and free the object */
fput(filp);
}
void rdma_lookup_put_uobject(struct ib_uobject *uobj, bool exclusive)
{
assert_uverbs_usecnt(uobj, exclusive);
uobj->type->type_class->lookup_put(uobj, exclusive);
/*
* In order to unlock an object, either decrease its usecnt for
* read access or zero it in case of exclusive access. See
* uverbs_try_lock_object for locking schema information.
*/
if (!exclusive)
atomic_dec(&uobj->usecnt);
else
atomic_set(&uobj->usecnt, 0);
uverbs_uobject_put(uobj);
}
const struct uverbs_obj_type_class uverbs_idr_class = {
.alloc_begin = alloc_begin_idr_uobject,
.lookup_get = lookup_get_idr_uobject,
.alloc_commit = alloc_commit_idr_uobject,
.alloc_abort = alloc_abort_idr_uobject,
.lookup_put = lookup_put_idr_uobject,
.remove_commit = remove_commit_idr_uobject,
/*
* When we destroy an object, we first just lock it for WRITE and
* actually DESTROY it in the finalize stage. So, the problematic
* scenario is when we just started the finalize stage of the
* destruction (nothing was executed yet). Now, the other thread
* fetched the object for READ access, but it didn't lock it yet.
* The DESTROY thread continues and starts destroying the object.
* When the other thread continue - without the RCU, it would
* access freed memory. However, the rcu_read_lock delays the free
* until the rcu_read_lock of the READ operation quits. Since the
* exclusive lock of the object is still taken by the DESTROY flow, the
* READ operation will get -EBUSY and it'll just bail out.
*/
.needs_kfree_rcu = true,
};
EXPORT_SYMBOL(uverbs_idr_class);
static void _uverbs_close_fd(struct ib_uobject_file *uobj_file)
{
struct ib_ucontext *ucontext;
struct ib_uverbs_file *ufile = uobj_file->ufile;
int ret;
mutex_lock(&uobj_file->ufile->cleanup_mutex);
/* uobject was either already cleaned up or is cleaned up right now anyway */
if (!uobj_file->uobj.context ||
!down_read_trylock(&uobj_file->uobj.context->cleanup_rwsem))
goto unlock;
ucontext = uobj_file->uobj.context;
ret = _rdma_remove_commit_uobject(&uobj_file->uobj, RDMA_REMOVE_CLOSE);
up_read(&ucontext->cleanup_rwsem);
if (ret)
pr_warn("uverbs: unable to clean up uobject file in uverbs_close_fd.\n");
unlock:
mutex_unlock(&ufile->cleanup_mutex);
}
void uverbs_close_fd(struct file *f)
{
struct ib_uobject_file *uobj_file = f->private_data;
struct kref *uverbs_file_ref = &uobj_file->ufile->ref;
_uverbs_close_fd(uobj_file);
uverbs_uobject_put(&uobj_file->uobj);
kref_put(uverbs_file_ref, ib_uverbs_release_file);
}
void uverbs_cleanup_ucontext(struct ib_ucontext *ucontext, bool device_removed)
{
enum rdma_remove_reason reason = device_removed ?
RDMA_REMOVE_DRIVER_REMOVE : RDMA_REMOVE_CLOSE;
unsigned int cur_order = 0;
ucontext->cleanup_reason = reason;
/*
* Waits for all remove_commit and alloc_commit to finish. Logically, We
* want to hold this forever as the context is going to be destroyed,
* but we'll release it since it causes a "held lock freed" BUG message.
*/
down_write(&ucontext->cleanup_rwsem);
while (!list_empty(&ucontext->uobjects)) {
struct ib_uobject *obj, *next_obj;
unsigned int next_order = UINT_MAX;
/*
* This shouldn't run while executing other commands on this
* context. Thus, the only thing we should take care of is
* releasing a FD while traversing this list. The FD could be
* closed and released from the _release fop of this FD.
* In order to mitigate this, we add a lock.
* We take and release the lock per order traversal in order
* to let other threads (which might still use the FDs) chance
* to run.
*/
mutex_lock(&ucontext->uobjects_lock);
list_for_each_entry_safe(obj, next_obj, &ucontext->uobjects,
list) {
if (obj->type->destroy_order == cur_order) {
int ret;
/*
* if we hit this WARN_ON, that means we are
* racing with a lookup_get.
*/
WARN_ON(uverbs_try_lock_object(obj, true));
ret = obj->type->type_class->remove_commit(obj,
reason);
list_del(&obj->list);
if (ret)
pr_warn("ib_uverbs: failed to remove uobject id %d order %u\n",
obj->id, cur_order);
/* put the ref we took when we created the object */
uverbs_uobject_put(obj);
} else {
next_order = min(next_order,
obj->type->destroy_order);
}
}
mutex_unlock(&ucontext->uobjects_lock);
cur_order = next_order;
}
up_write(&ucontext->cleanup_rwsem);
}
void uverbs_initialize_ucontext(struct ib_ucontext *ucontext)
{
ucontext->cleanup_reason = 0;
mutex_init(&ucontext->uobjects_lock);
INIT_LIST_HEAD(&ucontext->uobjects);
init_rwsem(&ucontext->cleanup_rwsem);
}
const struct uverbs_obj_type_class uverbs_fd_class = {
.alloc_begin = alloc_begin_fd_uobject,
.lookup_get = lookup_get_fd_uobject,
.alloc_commit = alloc_commit_fd_uobject,
.alloc_abort = alloc_abort_fd_uobject,
.lookup_put = lookup_put_fd_uobject,
.remove_commit = remove_commit_fd_uobject,
.needs_kfree_rcu = false,
};
EXPORT_SYMBOL(uverbs_fd_class);
IB/core: Add a generic way to execute an operation on a uobject 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@mellanox.com> Reviewed-by: Yishai Hadas <yishaih@mellanox.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
2017-08-03 21:06:55 +08:00
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;
}