uverbs_main.c 27.2 KB
Newer Older
1 2
/*
 * Copyright (c) 2005 Topspin Communications.  All rights reserved.
3
 * Copyright (c) 2005, 2006 Cisco Systems.  All rights reserved.
4 5
 * Copyright (c) 2005 Mellanox Technologies. All rights reserved.
 * Copyright (c) 2005 Voltaire, Inc. All rights reserved.
6
 * Copyright (c) 2005 PathScale, Inc. All rights reserved.
7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42
 *
 * 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/module.h>
#include <linux/init.h>
#include <linux/device.h>
#include <linux/err.h>
#include <linux/fs.h>
#include <linux/poll.h>
43
#include <linux/sched.h>
44
#include <linux/file.h>
45
#include <linux/cdev.h>
46
#include <linux/anon_inodes.h>
47
#include <linux/slab.h>
48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64

#include <asm/uaccess.h>

#include "uverbs.h"

MODULE_AUTHOR("Roland Dreier");
MODULE_DESCRIPTION("InfiniBand userspace verbs access");
MODULE_LICENSE("Dual BSD/GPL");

enum {
	IB_UVERBS_MAJOR       = 231,
	IB_UVERBS_BASE_MINOR  = 192,
	IB_UVERBS_MAX_DEVICES = 32
};

#define IB_UVERBS_BASE_DEV	MKDEV(IB_UVERBS_MAJOR, IB_UVERBS_BASE_MINOR)

65 66
static struct class *uverbs_class;

67
DEFINE_SPINLOCK(ib_uverbs_idr_lock);
68 69 70 71 72 73
DEFINE_IDR(ib_uverbs_pd_idr);
DEFINE_IDR(ib_uverbs_mr_idr);
DEFINE_IDR(ib_uverbs_mw_idr);
DEFINE_IDR(ib_uverbs_ah_idr);
DEFINE_IDR(ib_uverbs_cq_idr);
DEFINE_IDR(ib_uverbs_qp_idr);
74
DEFINE_IDR(ib_uverbs_srq_idr);
75
DEFINE_IDR(ib_uverbs_xrcd_idr);
76
DEFINE_IDR(ib_uverbs_rule_idr);
77

78
static DEFINE_SPINLOCK(map_lock);
79 80 81 82 83
static DECLARE_BITMAP(dev_map, IB_UVERBS_MAX_DEVICES);

static ssize_t (*uverbs_cmd_table[])(struct ib_uverbs_file *file,
				     const char __user *buf, int in_len,
				     int out_len) = {
A
Alexander Chiang 已提交
84 85 86 87 88 89
	[IB_USER_VERBS_CMD_GET_CONTEXT]		= ib_uverbs_get_context,
	[IB_USER_VERBS_CMD_QUERY_DEVICE]	= ib_uverbs_query_device,
	[IB_USER_VERBS_CMD_QUERY_PORT]		= ib_uverbs_query_port,
	[IB_USER_VERBS_CMD_ALLOC_PD]		= ib_uverbs_alloc_pd,
	[IB_USER_VERBS_CMD_DEALLOC_PD]		= ib_uverbs_dealloc_pd,
	[IB_USER_VERBS_CMD_REG_MR]		= ib_uverbs_reg_mr,
90
	[IB_USER_VERBS_CMD_REREG_MR]		= ib_uverbs_rereg_mr,
A
Alexander Chiang 已提交
91
	[IB_USER_VERBS_CMD_DEREG_MR]		= ib_uverbs_dereg_mr,
92 93
	[IB_USER_VERBS_CMD_ALLOC_MW]		= ib_uverbs_alloc_mw,
	[IB_USER_VERBS_CMD_DEALLOC_MW]		= ib_uverbs_dealloc_mw,
94
	[IB_USER_VERBS_CMD_CREATE_COMP_CHANNEL] = ib_uverbs_create_comp_channel,
A
Alexander Chiang 已提交
95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114
	[IB_USER_VERBS_CMD_CREATE_CQ]		= ib_uverbs_create_cq,
	[IB_USER_VERBS_CMD_RESIZE_CQ]		= ib_uverbs_resize_cq,
	[IB_USER_VERBS_CMD_POLL_CQ]		= ib_uverbs_poll_cq,
	[IB_USER_VERBS_CMD_REQ_NOTIFY_CQ]	= ib_uverbs_req_notify_cq,
	[IB_USER_VERBS_CMD_DESTROY_CQ]		= ib_uverbs_destroy_cq,
	[IB_USER_VERBS_CMD_CREATE_QP]		= ib_uverbs_create_qp,
	[IB_USER_VERBS_CMD_QUERY_QP]		= ib_uverbs_query_qp,
	[IB_USER_VERBS_CMD_MODIFY_QP]		= ib_uverbs_modify_qp,
	[IB_USER_VERBS_CMD_DESTROY_QP]		= ib_uverbs_destroy_qp,
	[IB_USER_VERBS_CMD_POST_SEND]		= ib_uverbs_post_send,
	[IB_USER_VERBS_CMD_POST_RECV]		= ib_uverbs_post_recv,
	[IB_USER_VERBS_CMD_POST_SRQ_RECV]	= ib_uverbs_post_srq_recv,
	[IB_USER_VERBS_CMD_CREATE_AH]		= ib_uverbs_create_ah,
	[IB_USER_VERBS_CMD_DESTROY_AH]		= ib_uverbs_destroy_ah,
	[IB_USER_VERBS_CMD_ATTACH_MCAST]	= ib_uverbs_attach_mcast,
	[IB_USER_VERBS_CMD_DETACH_MCAST]	= ib_uverbs_detach_mcast,
	[IB_USER_VERBS_CMD_CREATE_SRQ]		= ib_uverbs_create_srq,
	[IB_USER_VERBS_CMD_MODIFY_SRQ]		= ib_uverbs_modify_srq,
	[IB_USER_VERBS_CMD_QUERY_SRQ]		= ib_uverbs_query_srq,
	[IB_USER_VERBS_CMD_DESTROY_SRQ]		= ib_uverbs_destroy_srq,
115 116
	[IB_USER_VERBS_CMD_OPEN_XRCD]		= ib_uverbs_open_xrcd,
	[IB_USER_VERBS_CMD_CLOSE_XRCD]		= ib_uverbs_close_xrcd,
117
	[IB_USER_VERBS_CMD_CREATE_XSRQ]		= ib_uverbs_create_xsrq,
118
	[IB_USER_VERBS_CMD_OPEN_QP]		= ib_uverbs_open_qp,
119 120 121 122 123 124
};

static int (*uverbs_ex_cmd_table[])(struct ib_uverbs_file *file,
				    struct ib_udata *ucore,
				    struct ib_udata *uhw) = {
	[IB_USER_VERBS_EX_CMD_CREATE_FLOW]	= ib_uverbs_ex_create_flow,
125 126
	[IB_USER_VERBS_EX_CMD_DESTROY_FLOW]	= ib_uverbs_ex_destroy_flow,
	[IB_USER_VERBS_EX_CMD_QUERY_DEVICE]	= ib_uverbs_ex_query_device
127 128 129 130 131
};

static void ib_uverbs_add_one(struct ib_device *device);
static void ib_uverbs_remove_one(struct ib_device *device);

132 133 134 135 136
static void ib_uverbs_release_dev(struct kref *ref)
{
	struct ib_uverbs_device *dev =
		container_of(ref, struct ib_uverbs_device, ref);

137
	complete(&dev->comp);
138 139
}

140 141 142 143 144 145 146 147
static void ib_uverbs_release_event_file(struct kref *ref)
{
	struct ib_uverbs_event_file *file =
		container_of(ref, struct ib_uverbs_event_file, ref);

	kfree(file);
}

148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185
void ib_uverbs_release_ucq(struct ib_uverbs_file *file,
			  struct ib_uverbs_event_file *ev_file,
			  struct ib_ucq_object *uobj)
{
	struct ib_uverbs_event *evt, *tmp;

	if (ev_file) {
		spin_lock_irq(&ev_file->lock);
		list_for_each_entry_safe(evt, tmp, &uobj->comp_list, obj_list) {
			list_del(&evt->list);
			kfree(evt);
		}
		spin_unlock_irq(&ev_file->lock);

		kref_put(&ev_file->ref, ib_uverbs_release_event_file);
	}

	spin_lock_irq(&file->async_file->lock);
	list_for_each_entry_safe(evt, tmp, &uobj->async_list, obj_list) {
		list_del(&evt->list);
		kfree(evt);
	}
	spin_unlock_irq(&file->async_file->lock);
}

void ib_uverbs_release_uevent(struct ib_uverbs_file *file,
			      struct ib_uevent_object *uobj)
{
	struct ib_uverbs_event *evt, *tmp;

	spin_lock_irq(&file->async_file->lock);
	list_for_each_entry_safe(evt, tmp, &uobj->event_list, obj_list) {
		list_del(&evt->list);
		kfree(evt);
	}
	spin_unlock_irq(&file->async_file->lock);
}

186 187 188 189 190 191 192 193 194 195 196 197
static void ib_uverbs_detach_umcast(struct ib_qp *qp,
				    struct ib_uqp_object *uobj)
{
	struct ib_uverbs_mcast_entry *mcast, *tmp;

	list_for_each_entry_safe(mcast, tmp, &uobj->mcast_list, list) {
		ib_detach_mcast(qp, &mcast->gid, mcast->lid);
		list_del(&mcast->list);
		kfree(mcast);
	}
}

198 199
static int ib_uverbs_cleanup_ucontext(struct ib_uverbs_file *file,
				      struct ib_ucontext *context)
200 201 202 203 204 205
{
	struct ib_uobject *uobj, *tmp;

	if (!context)
		return 0;

206 207
	context->closing = 1;

208
	list_for_each_entry_safe(uobj, tmp, &context->ah_list, list) {
209 210 211
		struct ib_ah *ah = uobj->object;

		idr_remove_uobj(&ib_uverbs_ah_idr, uobj);
212 213 214
		ib_destroy_ah(ah);
		kfree(uobj);
	}
215

216 217 218 219 220 221 222 223 224
	/* Remove MWs before QPs, in order to support type 2A MWs. */
	list_for_each_entry_safe(uobj, tmp, &context->mw_list, list) {
		struct ib_mw *mw = uobj->object;

		idr_remove_uobj(&ib_uverbs_mw_idr, uobj);
		ib_dealloc_mw(mw);
		kfree(uobj);
	}

225 226 227 228 229 230 231 232
	list_for_each_entry_safe(uobj, tmp, &context->rule_list, list) {
		struct ib_flow *flow_id = uobj->object;

		idr_remove_uobj(&ib_uverbs_rule_idr, uobj);
		ib_destroy_flow(flow_id);
		kfree(uobj);
	}

233
	list_for_each_entry_safe(uobj, tmp, &context->qp_list, list) {
234
		struct ib_qp *qp = uobj->object;
235 236
		struct ib_uqp_object *uqp =
			container_of(uobj, struct ib_uqp_object, uevent.uobject);
237 238

		idr_remove_uobj(&ib_uverbs_qp_idr, uobj);
239 240
		if (qp != qp->real_qp) {
			ib_close_qp(qp);
241 242 243 244
		} else {
			ib_uverbs_detach_umcast(qp, uqp);
			ib_destroy_qp(qp);
		}
245 246
		ib_uverbs_release_uevent(file, &uqp->uevent);
		kfree(uqp);
247 248 249
	}

	list_for_each_entry_safe(uobj, tmp, &context->cq_list, list) {
250
		struct ib_cq *cq = uobj->object;
251 252 253
		struct ib_uverbs_event_file *ev_file = cq->cq_context;
		struct ib_ucq_object *ucq =
			container_of(uobj, struct ib_ucq_object, uobject);
254 255

		idr_remove_uobj(&ib_uverbs_cq_idr, uobj);
256
		ib_destroy_cq(cq);
257 258
		ib_uverbs_release_ucq(file, ev_file, ucq);
		kfree(ucq);
259 260
	}

261
	list_for_each_entry_safe(uobj, tmp, &context->srq_list, list) {
262
		struct ib_srq *srq = uobj->object;
263 264
		struct ib_uevent_object *uevent =
			container_of(uobj, struct ib_uevent_object, uobject);
265 266

		idr_remove_uobj(&ib_uverbs_srq_idr, uobj);
267
		ib_destroy_srq(srq);
268 269
		ib_uverbs_release_uevent(file, uevent);
		kfree(uevent);
270 271
	}

272
	list_for_each_entry_safe(uobj, tmp, &context->mr_list, list) {
273
		struct ib_mr *mr = uobj->object;
274

275
		idr_remove_uobj(&ib_uverbs_mr_idr, uobj);
276
		ib_dereg_mr(mr);
277
		kfree(uobj);
278 279
	}

280 281 282 283 284 285 286 287 288 289 290 291
	mutex_lock(&file->device->xrcd_tree_mutex);
	list_for_each_entry_safe(uobj, tmp, &context->xrcd_list, list) {
		struct ib_xrcd *xrcd = uobj->object;
		struct ib_uxrcd_object *uxrcd =
			container_of(uobj, struct ib_uxrcd_object, uobject);

		idr_remove_uobj(&ib_uverbs_xrcd_idr, uobj);
		ib_uverbs_dealloc_xrcd(file->device, xrcd);
		kfree(uxrcd);
	}
	mutex_unlock(&file->device->xrcd_tree_mutex);

292
	list_for_each_entry_safe(uobj, tmp, &context->pd_list, list) {
293 294 295
		struct ib_pd *pd = uobj->object;

		idr_remove_uobj(&ib_uverbs_pd_idr, uobj);
296 297 298 299 300 301 302 303 304 305 306 307 308
		ib_dealloc_pd(pd);
		kfree(uobj);
	}

	return context->device->dealloc_ucontext(context);
}

static void ib_uverbs_release_file(struct kref *ref)
{
	struct ib_uverbs_file *file =
		container_of(ref, struct ib_uverbs_file, ref);

	module_put(file->device->ib_dev->owner);
309 310
	kref_put(&file->device->ref, ib_uverbs_release_dev);

311 312 313 314 315 316 317
	kfree(file);
}

static ssize_t ib_uverbs_event_read(struct file *filp, char __user *buf,
				    size_t count, loff_t *pos)
{
	struct ib_uverbs_event_file *file = filp->private_data;
318
	struct ib_uverbs_event *event;
319 320 321 322 323
	int eventsz;
	int ret = 0;

	spin_lock_irq(&file->lock);

324
	while (list_empty(&file->event_list)) {
325 326 327 328 329 330
		spin_unlock_irq(&file->lock);

		if (filp->f_flags & O_NONBLOCK)
			return -EAGAIN;

		if (wait_event_interruptible(file->poll_wait,
331
					     !list_empty(&file->event_list)))
332 333 334 335 336
			return -ERESTARTSYS;

		spin_lock_irq(&file->lock);
	}

337 338 339
	event = list_entry(file->event_list.next, struct ib_uverbs_event, list);

	if (file->is_async)
340
		eventsz = sizeof (struct ib_uverbs_async_event_desc);
341
	else
342 343 344 345 346
		eventsz = sizeof (struct ib_uverbs_comp_event_desc);

	if (eventsz > count) {
		ret   = -EINVAL;
		event = NULL;
347
	} else {
348
		list_del(file->event_list.next);
349 350 351 352 353
		if (event->counter) {
			++(*event->counter);
			list_del(&event->obj_list);
		}
	}
354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377

	spin_unlock_irq(&file->lock);

	if (event) {
		if (copy_to_user(buf, event, eventsz))
			ret = -EFAULT;
		else
			ret = eventsz;
	}

	kfree(event);

	return ret;
}

static unsigned int ib_uverbs_event_poll(struct file *filp,
					 struct poll_table_struct *wait)
{
	unsigned int pollflags = 0;
	struct ib_uverbs_event_file *file = filp->private_data;

	poll_wait(filp, &file->poll_wait, wait);

	spin_lock_irq(&file->lock);
378
	if (!list_empty(&file->event_list))
379 380 381 382 383 384
		pollflags = POLLIN | POLLRDNORM;
	spin_unlock_irq(&file->lock);

	return pollflags;
}

G
Gleb Natapov 已提交
385 386 387 388 389 390 391
static int ib_uverbs_event_fasync(int fd, struct file *filp, int on)
{
	struct ib_uverbs_event_file *file = filp->private_data;

	return fasync_helper(fd, filp, on, &file->async_queue);
}

392 393 394
static int ib_uverbs_event_close(struct inode *inode, struct file *filp)
{
	struct ib_uverbs_event_file *file = filp->private_data;
395 396 397
	struct ib_uverbs_event *entry, *tmp;

	spin_lock_irq(&file->lock);
398
	file->is_closed = 1;
399 400 401 402 403 404
	list_for_each_entry_safe(entry, tmp, &file->event_list, list) {
		if (entry->counter)
			list_del(&entry->obj_list);
		kfree(entry);
	}
	spin_unlock_irq(&file->lock);
405

406 407 408 409 410
	if (file->is_async) {
		ib_unregister_event_handler(&file->uverbs_file->event_handler);
		kref_put(&file->uverbs_file->ref, ib_uverbs_release_file);
	}
	kref_put(&file->ref, ib_uverbs_release_event_file);
411 412 413 414

	return 0;
}

415
static const struct file_operations uverbs_event_fops = {
416
	.owner	 = THIS_MODULE,
A
Alexander Chiang 已提交
417
	.read	 = ib_uverbs_event_read,
418
	.poll    = ib_uverbs_event_poll,
G
Gleb Natapov 已提交
419
	.release = ib_uverbs_event_close,
420 421
	.fasync  = ib_uverbs_event_fasync,
	.llseek	 = no_llseek,
422 423 424 425
};

void ib_uverbs_comp_handler(struct ib_cq *cq, void *cq_context)
{
426 427 428 429 430 431 432 433 434
	struct ib_uverbs_event_file    *file = cq_context;
	struct ib_ucq_object	       *uobj;
	struct ib_uverbs_event	       *entry;
	unsigned long			flags;

	if (!file)
		return;

	spin_lock_irqsave(&file->lock, flags);
435
	if (file->is_closed) {
436 437 438
		spin_unlock_irqrestore(&file->lock, flags);
		return;
	}
439 440

	entry = kmalloc(sizeof *entry, GFP_ATOMIC);
441 442
	if (!entry) {
		spin_unlock_irqrestore(&file->lock, flags);
443
		return;
444
	}
445

446 447 448 449
	uobj = container_of(cq->uobject, struct ib_ucq_object, uobject);

	entry->desc.comp.cq_handle = cq->uobject->user_handle;
	entry->counter		   = &uobj->comp_events_reported;
450

451
	list_add_tail(&entry->list, &file->event_list);
452
	list_add_tail(&entry->obj_list, &uobj->comp_list);
453
	spin_unlock_irqrestore(&file->lock, flags);
454

455 456
	wake_up_interruptible(&file->poll_wait);
	kill_fasync(&file->async_queue, SIGIO, POLL_IN);
457 458 459
}

static void ib_uverbs_async_handler(struct ib_uverbs_file *file,
460 461 462
				    __u64 element, __u64 event,
				    struct list_head *obj_list,
				    u32 *counter)
463
{
464
	struct ib_uverbs_event *entry;
465 466
	unsigned long flags;

467
	spin_lock_irqsave(&file->async_file->lock, flags);
468
	if (file->async_file->is_closed) {
469 470 471 472
		spin_unlock_irqrestore(&file->async_file->lock, flags);
		return;
	}

473
	entry = kmalloc(sizeof *entry, GFP_ATOMIC);
474 475
	if (!entry) {
		spin_unlock_irqrestore(&file->async_file->lock, flags);
476
		return;
477
	}
478

479 480
	entry->desc.async.element    = element;
	entry->desc.async.event_type = event;
481
	entry->desc.async.reserved   = 0;
482
	entry->counter               = counter;
483

484
	list_add_tail(&entry->list, &file->async_file->event_list);
485 486
	if (obj_list)
		list_add_tail(&entry->obj_list, obj_list);
487
	spin_unlock_irqrestore(&file->async_file->lock, flags);
488

489 490
	wake_up_interruptible(&file->async_file->poll_wait);
	kill_fasync(&file->async_file->async_queue, SIGIO, POLL_IN);
491 492 493 494
}

void ib_uverbs_cq_event_handler(struct ib_event *event, void *context_ptr)
{
495 496
	struct ib_ucq_object *uobj = container_of(event->element.cq->uobject,
						  struct ib_ucq_object, uobject);
497

498
	ib_uverbs_async_handler(uobj->uverbs_file, uobj->uobject.user_handle,
499 500
				event->event, &uobj->async_list,
				&uobj->async_events_reported);
501 502 503 504
}

void ib_uverbs_qp_event_handler(struct ib_event *event, void *context_ptr)
{
505 506
	struct ib_uevent_object *uobj;

507 508 509 510
	/* for XRC target qp's, check that qp is live */
	if (!event->element.qp->uobject || !event->element.qp->uobject->live)
		return;

511 512 513 514 515 516
	uobj = container_of(event->element.qp->uobject,
			    struct ib_uevent_object, uobject);

	ib_uverbs_async_handler(context_ptr, uobj->uobject.user_handle,
				event->event, &uobj->event_list,
				&uobj->events_reported);
517 518
}

519 520
void ib_uverbs_srq_event_handler(struct ib_event *event, void *context_ptr)
{
521 522 523 524 525 526 527 528
	struct ib_uevent_object *uobj;

	uobj = container_of(event->element.srq->uobject,
			    struct ib_uevent_object, uobject);

	ib_uverbs_async_handler(context_ptr, uobj->uobject.user_handle,
				event->event, &uobj->event_list,
				&uobj->events_reported);
529 530
}

531 532
void ib_uverbs_event_handler(struct ib_event_handler *handler,
			     struct ib_event *event)
533 534 535 536
{
	struct ib_uverbs_file *file =
		container_of(handler, struct ib_uverbs_file, event_handler);

537 538
	ib_uverbs_async_handler(file, event->element.port_num, event->event,
				NULL, NULL);
539 540
}

541
struct file *ib_uverbs_alloc_event_file(struct ib_uverbs_file *uverbs_file,
542
					int is_async)
543
{
544
	struct ib_uverbs_event_file *ev_file;
545 546
	struct file *filp;

547 548 549 550 551 552 553 554 555 556 557
	ev_file = kmalloc(sizeof *ev_file, GFP_KERNEL);
	if (!ev_file)
		return ERR_PTR(-ENOMEM);

	kref_init(&ev_file->ref);
	spin_lock_init(&ev_file->lock);
	INIT_LIST_HEAD(&ev_file->event_list);
	init_waitqueue_head(&ev_file->poll_wait);
	ev_file->uverbs_file = uverbs_file;
	ev_file->async_queue = NULL;
	ev_file->is_async    = is_async;
558
	ev_file->is_closed   = 0;
559

560
	filp = anon_inode_getfile("[infinibandevent]", &uverbs_event_fops,
561
				  ev_file, O_RDONLY);
562 563
	if (IS_ERR(filp))
		kfree(ev_file);
564

565 566 567 568 569 570 571 572 573 574 575
	return filp;
}

/*
 * Look up a completion event file by FD.  If lookup is successful,
 * takes a ref to the event file struct that it returns; if
 * unsuccessful, returns NULL.
 */
struct ib_uverbs_event_file *ib_uverbs_lookup_comp_file(int fd)
{
	struct ib_uverbs_event_file *ev_file = NULL;
576
	struct fd f = fdget(fd);
577

578
	if (!f.file)
579 580
		return NULL;

581
	if (f.file->f_op != &uverbs_event_fops)
582 583
		goto out;

584
	ev_file = f.file->private_data;
585 586 587 588 589 590 591 592
	if (ev_file->is_async) {
		ev_file = NULL;
		goto out;
	}

	kref_get(&ev_file->ref);

out:
593
	fdput(f);
594
	return ev_file;
595 596 597 598 599 600 601
}

static ssize_t ib_uverbs_write(struct file *filp, const char __user *buf,
			     size_t count, loff_t *pos)
{
	struct ib_uverbs_file *file = filp->private_data;
	struct ib_uverbs_cmd_hdr hdr;
602
	__u32 flags;
603 604 605 606 607 608 609

	if (count < sizeof hdr)
		return -EINVAL;

	if (copy_from_user(&hdr, buf, sizeof hdr))
		return -EFAULT;

610 611
	flags = (hdr.command &
		 IB_USER_VERBS_CMD_FLAGS_MASK) >> IB_USER_VERBS_CMD_FLAGS_SHIFT;
612

613 614
	if (!flags) {
		__u32 command;
615

616 617 618
		if (hdr.command & ~(__u32)(IB_USER_VERBS_CMD_FLAGS_MASK |
					   IB_USER_VERBS_CMD_COMMAND_MASK))
			return -EINVAL;
619

620
		command = hdr.command & IB_USER_VERBS_CMD_COMMAND_MASK;
621

622 623 624
		if (command >= ARRAY_SIZE(uverbs_cmd_table) ||
		    !uverbs_cmd_table[command])
			return -EINVAL;
625

626 627
		if (!file->ucontext &&
		    command != IB_USER_VERBS_CMD_GET_CONTEXT)
628 629
			return -EINVAL;

630 631 632
		if (!(file->device->ib_dev->uverbs_cmd_mask & (1ull << command)))
			return -ENOSYS;

633 634 635
		if (hdr.in_words * 4 != count)
			return -EINVAL;

636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677
		return uverbs_cmd_table[command](file,
						 buf + sizeof(hdr),
						 hdr.in_words * 4,
						 hdr.out_words * 4);

	} else if (flags == IB_USER_VERBS_CMD_FLAG_EXTENDED) {
		__u32 command;

		struct ib_uverbs_ex_cmd_hdr ex_hdr;
		struct ib_udata ucore;
		struct ib_udata uhw;
		int err;
		size_t written_count = count;

		if (hdr.command & ~(__u32)(IB_USER_VERBS_CMD_FLAGS_MASK |
					   IB_USER_VERBS_CMD_COMMAND_MASK))
			return -EINVAL;

		command = hdr.command & IB_USER_VERBS_CMD_COMMAND_MASK;

		if (command >= ARRAY_SIZE(uverbs_ex_cmd_table) ||
		    !uverbs_ex_cmd_table[command])
			return -ENOSYS;

		if (!file->ucontext)
			return -EINVAL;

		if (!(file->device->ib_dev->uverbs_ex_cmd_mask & (1ull << command)))
			return -ENOSYS;

		if (count < (sizeof(hdr) + sizeof(ex_hdr)))
			return -EINVAL;

		if (copy_from_user(&ex_hdr, buf + sizeof(hdr), sizeof(ex_hdr)))
			return -EFAULT;

		count -= sizeof(hdr) + sizeof(ex_hdr);
		buf += sizeof(hdr) + sizeof(ex_hdr);

		if ((hdr.in_words + ex_hdr.provider_in_words) * 8 != count)
			return -EINVAL;

678 679 680
		if (ex_hdr.cmd_hdr_reserved)
			return -EINVAL;

681 682 683
		if (ex_hdr.response) {
			if (!hdr.out_words && !ex_hdr.provider_out_words)
				return -EINVAL;
684 685 686 687 688

			if (!access_ok(VERIFY_WRITE,
				       (void __user *) (unsigned long) ex_hdr.response,
				       (hdr.out_words + ex_hdr.provider_out_words) * 8))
				return -EFAULT;
689 690 691 692 693
		} else {
			if (hdr.out_words || ex_hdr.provider_out_words)
				return -EINVAL;
		}

694 695 696 697 698 699 700 701
		INIT_UDATA_BUF_OR_NULL(&ucore, buf, (unsigned long) ex_hdr.response,
				       hdr.in_words * 8, hdr.out_words * 8);

		INIT_UDATA_BUF_OR_NULL(&uhw,
				       buf + ucore.inlen,
				       (unsigned long) ex_hdr.response + ucore.outlen,
				       ex_hdr.provider_in_words * 8,
				       ex_hdr.provider_out_words * 8);
702 703 704 705 706 707 708 709 710

		err = uverbs_ex_cmd_table[command](file,
						   &ucore,
						   &uhw);

		if (err)
			return err;

		return written_count;
711
	}
712 713

	return -ENOSYS;
714 715 716 717 718 719 720 721 722 723 724 725
}

static int ib_uverbs_mmap(struct file *filp, struct vm_area_struct *vma)
{
	struct ib_uverbs_file *file = filp->private_data;

	if (!file->ucontext)
		return -ENODEV;
	else
		return file->device->ib_dev->mmap(file->ucontext, vma);
}

726 727 728
/*
 * ib_uverbs_open() does not need the BKL:
 *
A
Alexander Chiang 已提交
729
 *  - the ib_uverbs_device structures are properly reference counted and
730 731 732
 *    everything else is purely local to the file being created, so
 *    races against other open calls are not a problem;
 *  - there is no ioctl method to race against;
A
Alexander Chiang 已提交
733 734
 *  - the open method will either immediately run -ENXIO, or all
 *    required initialization will be done.
735
 */
736 737
static int ib_uverbs_open(struct inode *inode, struct file *filp)
{
738
	struct ib_uverbs_device *dev;
739
	struct ib_uverbs_file *file;
740
	int ret;
741

A
Alexander Chiang 已提交
742
	dev = container_of(inode->i_cdev, struct ib_uverbs_device, cdev);
743 744
	if (dev)
		kref_get(&dev->ref);
A
Alexander Chiang 已提交
745
	else
746 747 748 749 750 751
		return -ENXIO;

	if (!try_module_get(dev->ib_dev->owner)) {
		ret = -ENODEV;
		goto err;
	}
752

753
	file = kmalloc(sizeof *file, GFP_KERNEL);
754
	if (!file) {
755 756
		ret = -ENOMEM;
		goto err_module;
757
	}
758

759 760 761
	file->device	 = dev;
	file->ucontext	 = NULL;
	file->async_file = NULL;
762
	kref_init(&file->ref);
763
	mutex_init(&file->mutex);
764 765 766

	filp->private_data = file;

767
	return nonseekable_open(inode, filp);
768 769 770 771 772 773 774

err_module:
	module_put(dev->ib_dev->owner);

err:
	kref_put(&dev->ref, ib_uverbs_release_dev);
	return ret;
775 776 777 778 779 780
}

static int ib_uverbs_close(struct inode *inode, struct file *filp)
{
	struct ib_uverbs_file *file = filp->private_data;

781 782 783 784
	ib_uverbs_cleanup_ucontext(file, file->ucontext);

	if (file->async_file)
		kref_put(&file->async_file->ref, ib_uverbs_release_event_file);
785 786 787 788 789 790

	kref_put(&file->ref, ib_uverbs_release_file);

	return 0;
}

791
static const struct file_operations uverbs_fops = {
A
Alexander Chiang 已提交
792 793 794
	.owner	 = THIS_MODULE,
	.write	 = ib_uverbs_write,
	.open	 = ib_uverbs_open,
795 796
	.release = ib_uverbs_close,
	.llseek	 = no_llseek,
797 798
};

799
static const struct file_operations uverbs_mmap_fops = {
A
Alexander Chiang 已提交
800 801
	.owner	 = THIS_MODULE,
	.write	 = ib_uverbs_write,
802
	.mmap    = ib_uverbs_mmap,
A
Alexander Chiang 已提交
803
	.open	 = ib_uverbs_open,
804 805
	.release = ib_uverbs_close,
	.llseek	 = no_llseek,
806 807 808 809 810 811 812 813
};

static struct ib_client uverbs_client = {
	.name   = "uverbs",
	.add    = ib_uverbs_add_one,
	.remove = ib_uverbs_remove_one
};

814 815
static ssize_t show_ibdev(struct device *device, struct device_attribute *attr,
			  char *buf)
816
{
817
	struct ib_uverbs_device *dev = dev_get_drvdata(device);
818 819 820

	if (!dev)
		return -ENODEV;
821 822 823

	return sprintf(buf, "%s\n", dev->ib_dev->name);
}
824
static DEVICE_ATTR(ibdev, S_IRUGO, show_ibdev, NULL);
825

826 827
static ssize_t show_dev_abi_version(struct device *device,
				    struct device_attribute *attr, char *buf)
828
{
829
	struct ib_uverbs_device *dev = dev_get_drvdata(device);
830 831 832

	if (!dev)
		return -ENODEV;
833 834 835

	return sprintf(buf, "%d\n", dev->ib_dev->uverbs_abi_ver);
}
836
static DEVICE_ATTR(abi_version, S_IRUGO, show_dev_abi_version, NULL);
837

838 839
static CLASS_ATTR_STRING(abi_version, S_IRUGO,
			 __stringify(IB_USER_VERBS_ABI_VERSION));
840

841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868
static dev_t overflow_maj;
static DECLARE_BITMAP(overflow_map, IB_UVERBS_MAX_DEVICES);

/*
 * If we have more than IB_UVERBS_MAX_DEVICES, dynamically overflow by
 * requesting a new major number and doubling the number of max devices we
 * support. It's stupid, but simple.
 */
static int find_overflow_devnum(void)
{
	int ret;

	if (!overflow_maj) {
		ret = alloc_chrdev_region(&overflow_maj, 0, IB_UVERBS_MAX_DEVICES,
					  "infiniband_verbs");
		if (ret) {
			printk(KERN_ERR "user_verbs: couldn't register dynamic device number\n");
			return ret;
		}
	}

	ret = find_first_zero_bit(overflow_map, IB_UVERBS_MAX_DEVICES);
	if (ret >= IB_UVERBS_MAX_DEVICES)
		return -1;

	return ret;
}

869 870
static void ib_uverbs_add_one(struct ib_device *device)
{
871
	int devnum;
872
	dev_t base;
873 874 875 876 877
	struct ib_uverbs_device *uverbs_dev;

	if (!device->alloc_ucontext)
		return;

R
Roland Dreier 已提交
878
	uverbs_dev = kzalloc(sizeof *uverbs_dev, GFP_KERNEL);
879 880 881
	if (!uverbs_dev)
		return;

882
	kref_init(&uverbs_dev->ref);
883
	init_completion(&uverbs_dev->comp);
884 885
	uverbs_dev->xrcd_tree = RB_ROOT;
	mutex_init(&uverbs_dev->xrcd_tree_mutex);
886

887
	spin_lock(&map_lock);
888 889
	devnum = find_first_zero_bit(dev_map, IB_UVERBS_MAX_DEVICES);
	if (devnum >= IB_UVERBS_MAX_DEVICES) {
890
		spin_unlock(&map_lock);
891 892 893 894 895 896 897 898 899 900 901 902
		devnum = find_overflow_devnum();
		if (devnum < 0)
			goto err;

		spin_lock(&map_lock);
		uverbs_dev->devnum = devnum + IB_UVERBS_MAX_DEVICES;
		base = devnum + overflow_maj;
		set_bit(devnum, overflow_map);
	} else {
		uverbs_dev->devnum = devnum;
		base = devnum + IB_UVERBS_BASE_DEV;
		set_bit(devnum, dev_map);
903 904 905
	}
	spin_unlock(&map_lock);

906
	uverbs_dev->ib_dev           = device;
907
	uverbs_dev->num_comp_vectors = device->num_comp_vectors;
908

909 910 911 912
	cdev_init(&uverbs_dev->cdev, NULL);
	uverbs_dev->cdev.owner = THIS_MODULE;
	uverbs_dev->cdev.ops = device->mmap ? &uverbs_mmap_fops : &uverbs_fops;
	kobject_set_name(&uverbs_dev->cdev.kobj, "uverbs%d", uverbs_dev->devnum);
913
	if (cdev_add(&uverbs_dev->cdev, base, 1))
914
		goto err_cdev;
915

916
	uverbs_dev->dev = device_create(uverbs_class, device->dma_device,
917
					uverbs_dev->cdev.dev, uverbs_dev,
918
					"uverbs%d", uverbs_dev->devnum);
919
	if (IS_ERR(uverbs_dev->dev))
920 921
		goto err_cdev;

922
	if (device_create_file(uverbs_dev->dev, &dev_attr_ibdev))
923
		goto err_class;
924
	if (device_create_file(uverbs_dev->dev, &dev_attr_abi_version))
925
		goto err_class;
926 927 928 929 930 931

	ib_set_client_data(device, &uverbs_client, uverbs_dev);

	return;

err_class:
932
	device_destroy(uverbs_class, uverbs_dev->cdev.dev);
933 934

err_cdev:
935
	cdev_del(&uverbs_dev->cdev);
936 937 938 939
	if (uverbs_dev->devnum < IB_UVERBS_MAX_DEVICES)
		clear_bit(devnum, dev_map);
	else
		clear_bit(devnum, overflow_map);
940 941

err:
942
	kref_put(&uverbs_dev->ref, ib_uverbs_release_dev);
943 944
	wait_for_completion(&uverbs_dev->comp);
	kfree(uverbs_dev);
945 946 947 948 949 950 951 952 953 954
	return;
}

static void ib_uverbs_remove_one(struct ib_device *device)
{
	struct ib_uverbs_device *uverbs_dev = ib_get_client_data(device, &uverbs_client);

	if (!uverbs_dev)
		return;

955
	dev_set_drvdata(uverbs_dev->dev, NULL);
956 957
	device_destroy(uverbs_class, uverbs_dev->cdev.dev);
	cdev_del(&uverbs_dev->cdev);
958

959 960 961 962
	if (uverbs_dev->devnum < IB_UVERBS_MAX_DEVICES)
		clear_bit(uverbs_dev->devnum, dev_map);
	else
		clear_bit(uverbs_dev->devnum - IB_UVERBS_MAX_DEVICES, overflow_map);
963

964
	kref_put(&uverbs_dev->ref, ib_uverbs_release_dev);
965 966
	wait_for_completion(&uverbs_dev->comp);
	kfree(uverbs_dev);
967 968
}

969
static char *uverbs_devnode(struct device *dev, umode_t *mode)
970
{
971 972
	if (mode)
		*mode = 0666;
973 974 975
	return kasprintf(GFP_KERNEL, "infiniband/%s", dev_name(dev));
}

976 977 978 979 980 981 982 983 984 985 986
static int __init ib_uverbs_init(void)
{
	int ret;

	ret = register_chrdev_region(IB_UVERBS_BASE_DEV, IB_UVERBS_MAX_DEVICES,
				     "infiniband_verbs");
	if (ret) {
		printk(KERN_ERR "user_verbs: couldn't register device number\n");
		goto out;
	}

987 988 989
	uverbs_class = class_create(THIS_MODULE, "infiniband_verbs");
	if (IS_ERR(uverbs_class)) {
		ret = PTR_ERR(uverbs_class);
990 991 992 993
		printk(KERN_ERR "user_verbs: couldn't create class infiniband_verbs\n");
		goto out_chrdev;
	}

994 995
	uverbs_class->devnode = uverbs_devnode;

996
	ret = class_create_file(uverbs_class, &class_attr_abi_version.attr);
997 998 999 1000 1001 1002 1003 1004
	if (ret) {
		printk(KERN_ERR "user_verbs: couldn't create abi_version attribute\n");
		goto out_class;
	}

	ret = ib_register_client(&uverbs_client);
	if (ret) {
		printk(KERN_ERR "user_verbs: couldn't register client\n");
1005
		goto out_class;
1006 1007 1008 1009 1010
	}

	return 0;

out_class:
1011
	class_destroy(uverbs_class);
1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022

out_chrdev:
	unregister_chrdev_region(IB_UVERBS_BASE_DEV, IB_UVERBS_MAX_DEVICES);

out:
	return ret;
}

static void __exit ib_uverbs_cleanup(void)
{
	ib_unregister_client(&uverbs_client);
1023
	class_destroy(uverbs_class);
1024
	unregister_chrdev_region(IB_UVERBS_BASE_DEV, IB_UVERBS_MAX_DEVICES);
1025 1026
	if (overflow_maj)
		unregister_chrdev_region(overflow_maj, IB_UVERBS_MAX_DEVICES);
1027 1028 1029 1030 1031 1032 1033
	idr_destroy(&ib_uverbs_pd_idr);
	idr_destroy(&ib_uverbs_mr_idr);
	idr_destroy(&ib_uverbs_mw_idr);
	idr_destroy(&ib_uverbs_ah_idr);
	idr_destroy(&ib_uverbs_cq_idr);
	idr_destroy(&ib_uverbs_qp_idr);
	idr_destroy(&ib_uverbs_srq_idr);
1034 1035 1036 1037
}

module_init(ib_uverbs_init);
module_exit(ib_uverbs_cleanup);