xpc_channel.c 60.4 KB
Newer Older
1 2 3 4 5 6 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 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59
/*
 * This file is subject to the terms and conditions of the GNU General Public
 * License.  See the file "COPYING" in the main directory of this archive
 * for more details.
 *
 * Copyright (c) 2004-2005 Silicon Graphics, Inc.  All Rights Reserved.
 */


/*
 * Cross Partition Communication (XPC) channel support.
 *
 *	This is the part of XPC that manages the channels and
 *	sends/receives messages across them to/from other partitions.
 *
 */


#include <linux/kernel.h>
#include <linux/init.h>
#include <linux/sched.h>
#include <linux/cache.h>
#include <linux/interrupt.h>
#include <linux/slab.h>
#include <asm/sn/bte.h>
#include <asm/sn/sn_sal.h>
#include "xpc.h"


/*
 * Set up the initial values for the XPartition Communication channels.
 */
static void
xpc_initialize_channels(struct xpc_partition *part, partid_t partid)
{
	int ch_number;
	struct xpc_channel *ch;


	for (ch_number = 0; ch_number < part->nchannels; ch_number++) {
		ch = &part->channels[ch_number];

		ch->partid = partid;
		ch->number = ch_number;
		ch->flags = XPC_C_DISCONNECTED;

		ch->local_GP = &part->local_GPs[ch_number];
		ch->local_openclose_args =
					&part->local_openclose_args[ch_number];

		atomic_set(&ch->kthreads_assigned, 0);
		atomic_set(&ch->kthreads_idle, 0);
		atomic_set(&ch->kthreads_active, 0);

		atomic_set(&ch->references, 0);
		atomic_set(&ch->n_to_notify, 0);

		spin_lock_init(&ch->lock);
		sema_init(&ch->msg_to_pull_sema, 1);	/* mutex */
60
		sema_init(&ch->wdisconnect_sema, 0);	/* event wait */
61 62 63 64 65 66 67 68 69 70 71 72 73 74 75

		atomic_set(&ch->n_on_msg_allocate_wq, 0);
		init_waitqueue_head(&ch->msg_allocate_wq);
		init_waitqueue_head(&ch->idle_wq);
	}
}


/*
 * Setup the infrastructure necessary to support XPartition Communication
 * between the specified remote partition and the local one.
 */
enum xpc_retval
xpc_setup_infrastructure(struct xpc_partition *part)
{
76
	int ret, cpuid;
77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169
	struct timer_list *timer;
	partid_t partid = XPC_PARTID(part);


	/*
	 * Zero out MOST of the entry for this partition. Only the fields
	 * starting with `nchannels' will be zeroed. The preceding fields must
	 * remain `viable' across partition ups and downs, since they may be
	 * referenced during this memset() operation.
	 */
	memset(&part->nchannels, 0, sizeof(struct xpc_partition) -
				offsetof(struct xpc_partition, nchannels));

	/*
	 * Allocate all of the channel structures as a contiguous chunk of
	 * memory.
	 */
	part->channels = kmalloc(sizeof(struct xpc_channel) * XPC_NCHANNELS,
								GFP_KERNEL);
	if (part->channels == NULL) {
		dev_err(xpc_chan, "can't get memory for channels\n");
		return xpcNoMemory;
	}
	memset(part->channels, 0, sizeof(struct xpc_channel) * XPC_NCHANNELS);

	part->nchannels = XPC_NCHANNELS;


	/* allocate all the required GET/PUT values */

	part->local_GPs = xpc_kmalloc_cacheline_aligned(XPC_GP_SIZE,
					GFP_KERNEL, &part->local_GPs_base);
	if (part->local_GPs == NULL) {
		kfree(part->channels);
		part->channels = NULL;
		dev_err(xpc_chan, "can't get memory for local get/put "
			"values\n");
		return xpcNoMemory;
	}
	memset(part->local_GPs, 0, XPC_GP_SIZE);

	part->remote_GPs = xpc_kmalloc_cacheline_aligned(XPC_GP_SIZE,
					GFP_KERNEL, &part->remote_GPs_base);
	if (part->remote_GPs == NULL) {
		kfree(part->channels);
		part->channels = NULL;
		kfree(part->local_GPs_base);
		part->local_GPs = NULL;
		dev_err(xpc_chan, "can't get memory for remote get/put "
			"values\n");
		return xpcNoMemory;
	}
	memset(part->remote_GPs, 0, XPC_GP_SIZE);


	/* allocate all the required open and close args */

	part->local_openclose_args = xpc_kmalloc_cacheline_aligned(
					XPC_OPENCLOSE_ARGS_SIZE, GFP_KERNEL,
					&part->local_openclose_args_base);
	if (part->local_openclose_args == NULL) {
		kfree(part->channels);
		part->channels = NULL;
		kfree(part->local_GPs_base);
		part->local_GPs = NULL;
		kfree(part->remote_GPs_base);
		part->remote_GPs = NULL;
		dev_err(xpc_chan, "can't get memory for local connect args\n");
		return xpcNoMemory;
	}
	memset(part->local_openclose_args, 0, XPC_OPENCLOSE_ARGS_SIZE);

	part->remote_openclose_args = xpc_kmalloc_cacheline_aligned(
					XPC_OPENCLOSE_ARGS_SIZE, GFP_KERNEL,
					&part->remote_openclose_args_base);
	if (part->remote_openclose_args == NULL) {
		kfree(part->channels);
		part->channels = NULL;
		kfree(part->local_GPs_base);
		part->local_GPs = NULL;
		kfree(part->remote_GPs_base);
		part->remote_GPs = NULL;
		kfree(part->local_openclose_args_base);
		part->local_openclose_args = NULL;
		dev_err(xpc_chan, "can't get memory for remote connect args\n");
		return xpcNoMemory;
	}
	memset(part->remote_openclose_args, 0, XPC_OPENCLOSE_ARGS_SIZE);


	xpc_initialize_channels(part, partid);

	atomic_set(&part->nchannels_active, 0);
170
	atomic_set(&part->nchannels_engaged, 0);
171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213


	/* local_IPI_amo were set to 0 by an earlier memset() */

	/* Initialize this partitions AMO_t structure */
	part->local_IPI_amo_va = xpc_IPI_init(partid);

	spin_lock_init(&part->IPI_lock);

	atomic_set(&part->channel_mgr_requests, 1);
	init_waitqueue_head(&part->channel_mgr_wq);

	sprintf(part->IPI_owner, "xpc%02d", partid);
	ret = request_irq(SGI_XPC_NOTIFY, xpc_notify_IRQ_handler, SA_SHIRQ,
				part->IPI_owner, (void *) (u64) partid);
	if (ret != 0) {
		kfree(part->channels);
		part->channels = NULL;
		kfree(part->local_GPs_base);
		part->local_GPs = NULL;
		kfree(part->remote_GPs_base);
		part->remote_GPs = NULL;
		kfree(part->local_openclose_args_base);
		part->local_openclose_args = NULL;
		kfree(part->remote_openclose_args_base);
		part->remote_openclose_args = NULL;
		dev_err(xpc_chan, "can't register NOTIFY IRQ handler, "
			"errno=%d\n", -ret);
		return xpcLackOfResources;
	}

	/* Setup a timer to check for dropped IPIs */
	timer = &part->dropped_IPI_timer;
	init_timer(timer);
	timer->function = (void (*)(unsigned long)) xpc_dropped_IPI_check;
	timer->data = (unsigned long) part;
	timer->expires = jiffies + XPC_P_DROPPED_IPI_WAIT;
	add_timer(timer);

	/*
	 * With the setting of the partition setup_state to XPC_P_SETUP, we're
	 * declaring that this partition is ready to go.
	 */
214
	part->setup_state = XPC_P_SETUP;
215 216 217 218 219 220 221 222 223 224 225 226 227


	/*
	 * Setup the per partition specific variables required by the
	 * remote partition to establish channel connections with us.
	 *
	 * The setting of the magic # indicates that these per partition
	 * specific variables are ready to be used.
	 */
	xpc_vars_part[partid].GPs_pa = __pa(part->local_GPs);
	xpc_vars_part[partid].openclose_args_pa =
					__pa(part->local_openclose_args);
	xpc_vars_part[partid].IPI_amo_pa = __pa(part->local_IPI_amo_va);
228 229 230
	cpuid = raw_smp_processor_id();	/* any CPU in this partition will do */
	xpc_vars_part[partid].IPI_nasid = cpuid_to_nasid(cpuid);
	xpc_vars_part[partid].IPI_phys_cpuid = cpu_physical_id(cpuid);
231
	xpc_vars_part[partid].nchannels = part->nchannels;
232
	xpc_vars_part[partid].magic = XPC_VP_MAGIC1;
233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359

	return xpcSuccess;
}


/*
 * Create a wrapper that hides the underlying mechanism for pulling a cacheline
 * (or multiple cachelines) from a remote partition.
 *
 * src must be a cacheline aligned physical address on the remote partition.
 * dst must be a cacheline aligned virtual address on this partition.
 * cnt must be an cacheline sized
 */
static enum xpc_retval
xpc_pull_remote_cachelines(struct xpc_partition *part, void *dst,
				const void *src, size_t cnt)
{
	bte_result_t bte_ret;


	DBUG_ON((u64) src != L1_CACHE_ALIGN((u64) src));
	DBUG_ON((u64) dst != L1_CACHE_ALIGN((u64) dst));
	DBUG_ON(cnt != L1_CACHE_ALIGN(cnt));

	if (part->act_state == XPC_P_DEACTIVATING) {
		return part->reason;
	}

	bte_ret = xp_bte_copy((u64) src, (u64) ia64_tpa((u64) dst),
				(u64) cnt, (BTE_NORMAL | BTE_WACQUIRE), NULL);
	if (bte_ret == BTE_SUCCESS) {
		return xpcSuccess;
	}

	dev_dbg(xpc_chan, "xp_bte_copy() from partition %d failed, ret=%d\n",
		XPC_PARTID(part), bte_ret);

	return xpc_map_bte_errors(bte_ret);
}


/*
 * Pull the remote per partititon specific variables from the specified
 * partition.
 */
enum xpc_retval
xpc_pull_remote_vars_part(struct xpc_partition *part)
{
	u8 buffer[L1_CACHE_BYTES * 2];
	struct xpc_vars_part *pulled_entry_cacheline =
			(struct xpc_vars_part *) L1_CACHE_ALIGN((u64) buffer);
	struct xpc_vars_part *pulled_entry;
	u64 remote_entry_cacheline_pa, remote_entry_pa;
	partid_t partid = XPC_PARTID(part);
	enum xpc_retval ret;


	/* pull the cacheline that contains the variables we're interested in */

	DBUG_ON(part->remote_vars_part_pa !=
				L1_CACHE_ALIGN(part->remote_vars_part_pa));
	DBUG_ON(sizeof(struct xpc_vars_part) != L1_CACHE_BYTES / 2);

	remote_entry_pa = part->remote_vars_part_pa +
			sn_partition_id * sizeof(struct xpc_vars_part);

	remote_entry_cacheline_pa = (remote_entry_pa & ~(L1_CACHE_BYTES - 1));

	pulled_entry = (struct xpc_vars_part *) ((u64) pulled_entry_cacheline +
				(remote_entry_pa & (L1_CACHE_BYTES - 1)));

	ret = xpc_pull_remote_cachelines(part, pulled_entry_cacheline,
					(void *) remote_entry_cacheline_pa,
					L1_CACHE_BYTES);
	if (ret != xpcSuccess) {
		dev_dbg(xpc_chan, "failed to pull XPC vars_part from "
			"partition %d, ret=%d\n", partid, ret);
		return ret;
	}


	/* see if they've been set up yet */

	if (pulled_entry->magic != XPC_VP_MAGIC1 &&
				pulled_entry->magic != XPC_VP_MAGIC2) {

		if (pulled_entry->magic != 0) {
			dev_dbg(xpc_chan, "partition %d's XPC vars_part for "
				"partition %d has bad magic value (=0x%lx)\n",
				partid, sn_partition_id, pulled_entry->magic);
			return xpcBadMagic;
		}

		/* they've not been initialized yet */
		return xpcRetry;
	}

	if (xpc_vars_part[partid].magic == XPC_VP_MAGIC1) {

		/* validate the variables */

		if (pulled_entry->GPs_pa == 0 ||
				pulled_entry->openclose_args_pa == 0 ||
					pulled_entry->IPI_amo_pa == 0) {

			dev_err(xpc_chan, "partition %d's XPC vars_part for "
				"partition %d are not valid\n", partid,
				sn_partition_id);
			return xpcInvalidAddress;
		}

		/* the variables we imported look to be valid */

		part->remote_GPs_pa = pulled_entry->GPs_pa;
		part->remote_openclose_args_pa =
					pulled_entry->openclose_args_pa;
		part->remote_IPI_amo_va =
				      (AMO_t *) __va(pulled_entry->IPI_amo_pa);
		part->remote_IPI_nasid = pulled_entry->IPI_nasid;
		part->remote_IPI_phys_cpuid = pulled_entry->IPI_phys_cpuid;

		if (part->nchannels > pulled_entry->nchannels) {
			part->nchannels = pulled_entry->nchannels;
		}

		/* let the other side know that we've pulled their variables */

360
		xpc_vars_part[partid].magic = XPC_VP_MAGIC2;
361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627
	}

	if (pulled_entry->magic == XPC_VP_MAGIC1) {
		return xpcRetry;
	}

	return xpcSuccess;
}


/*
 * Get the IPI flags and pull the openclose args and/or remote GPs as needed.
 */
static u64
xpc_get_IPI_flags(struct xpc_partition *part)
{
	unsigned long irq_flags;
	u64 IPI_amo;
	enum xpc_retval ret;


	/*
	 * See if there are any IPI flags to be handled.
	 */

	spin_lock_irqsave(&part->IPI_lock, irq_flags);
	if ((IPI_amo = part->local_IPI_amo) != 0) {
		part->local_IPI_amo = 0;
	}
	spin_unlock_irqrestore(&part->IPI_lock, irq_flags);


	if (XPC_ANY_OPENCLOSE_IPI_FLAGS_SET(IPI_amo)) {
		ret = xpc_pull_remote_cachelines(part,
					part->remote_openclose_args,
					(void *) part->remote_openclose_args_pa,
					XPC_OPENCLOSE_ARGS_SIZE);
		if (ret != xpcSuccess) {
			XPC_DEACTIVATE_PARTITION(part, ret);

			dev_dbg(xpc_chan, "failed to pull openclose args from "
				"partition %d, ret=%d\n", XPC_PARTID(part),
				ret);

			/* don't bother processing IPIs anymore */
			IPI_amo = 0;
		}
	}

	if (XPC_ANY_MSG_IPI_FLAGS_SET(IPI_amo)) {
		ret = xpc_pull_remote_cachelines(part, part->remote_GPs,
						(void *) part->remote_GPs_pa,
						XPC_GP_SIZE);
		if (ret != xpcSuccess) {
			XPC_DEACTIVATE_PARTITION(part, ret);

			dev_dbg(xpc_chan, "failed to pull GPs from partition "
				"%d, ret=%d\n", XPC_PARTID(part), ret);

			/* don't bother processing IPIs anymore */
			IPI_amo = 0;
		}
	}

	return IPI_amo;
}


/*
 * Allocate the local message queue and the notify queue.
 */
static enum xpc_retval
xpc_allocate_local_msgqueue(struct xpc_channel *ch)
{
	unsigned long irq_flags;
	int nentries;
	size_t nbytes;


	// >>> may want to check for ch->flags & XPC_C_DISCONNECTING between
	// >>> iterations of the for-loop, bail if set?

	// >>> should we impose a minumum #of entries? like 4 or 8?
	for (nentries = ch->local_nentries; nentries > 0; nentries--) {

		nbytes = nentries * ch->msg_size;
		ch->local_msgqueue = xpc_kmalloc_cacheline_aligned(nbytes,
						(GFP_KERNEL | GFP_DMA),
						&ch->local_msgqueue_base);
		if (ch->local_msgqueue == NULL) {
			continue;
		}
		memset(ch->local_msgqueue, 0, nbytes);

		nbytes = nentries * sizeof(struct xpc_notify);
		ch->notify_queue = kmalloc(nbytes, (GFP_KERNEL | GFP_DMA));
		if (ch->notify_queue == NULL) {
			kfree(ch->local_msgqueue_base);
			ch->local_msgqueue = NULL;
			continue;
		}
		memset(ch->notify_queue, 0, nbytes);

		spin_lock_irqsave(&ch->lock, irq_flags);
		if (nentries < ch->local_nentries) {
			dev_dbg(xpc_chan, "nentries=%d local_nentries=%d, "
				"partid=%d, channel=%d\n", nentries,
				ch->local_nentries, ch->partid, ch->number);

			ch->local_nentries = nentries;
		}
		spin_unlock_irqrestore(&ch->lock, irq_flags);
		return xpcSuccess;
	}

	dev_dbg(xpc_chan, "can't get memory for local message queue and notify "
		"queue, partid=%d, channel=%d\n", ch->partid, ch->number);
	return xpcNoMemory;
}


/*
 * Allocate the cached remote message queue.
 */
static enum xpc_retval
xpc_allocate_remote_msgqueue(struct xpc_channel *ch)
{
	unsigned long irq_flags;
	int nentries;
	size_t nbytes;


	DBUG_ON(ch->remote_nentries <= 0);

	// >>> may want to check for ch->flags & XPC_C_DISCONNECTING between
	// >>> iterations of the for-loop, bail if set?

	// >>> should we impose a minumum #of entries? like 4 or 8?
	for (nentries = ch->remote_nentries; nentries > 0; nentries--) {

		nbytes = nentries * ch->msg_size;
		ch->remote_msgqueue = xpc_kmalloc_cacheline_aligned(nbytes,
						(GFP_KERNEL | GFP_DMA),
						&ch->remote_msgqueue_base);
		if (ch->remote_msgqueue == NULL) {
			continue;
		}
		memset(ch->remote_msgqueue, 0, nbytes);

		spin_lock_irqsave(&ch->lock, irq_flags);
		if (nentries < ch->remote_nentries) {
			dev_dbg(xpc_chan, "nentries=%d remote_nentries=%d, "
				"partid=%d, channel=%d\n", nentries,
				ch->remote_nentries, ch->partid, ch->number);

			ch->remote_nentries = nentries;
		}
		spin_unlock_irqrestore(&ch->lock, irq_flags);
		return xpcSuccess;
	}

	dev_dbg(xpc_chan, "can't get memory for cached remote message queue, "
		"partid=%d, channel=%d\n", ch->partid, ch->number);
	return xpcNoMemory;
}


/*
 * Allocate message queues and other stuff associated with a channel.
 *
 * Note: Assumes all of the channel sizes are filled in.
 */
static enum xpc_retval
xpc_allocate_msgqueues(struct xpc_channel *ch)
{
	unsigned long irq_flags;
	int i;
	enum xpc_retval ret;


	DBUG_ON(ch->flags & XPC_C_SETUP);

	if ((ret = xpc_allocate_local_msgqueue(ch)) != xpcSuccess) {
		return ret;
	}

	if ((ret = xpc_allocate_remote_msgqueue(ch)) != xpcSuccess) {
		kfree(ch->local_msgqueue_base);
		ch->local_msgqueue = NULL;
		kfree(ch->notify_queue);
		ch->notify_queue = NULL;
		return ret;
	}

	for (i = 0; i < ch->local_nentries; i++) {
		/* use a semaphore as an event wait queue */
		sema_init(&ch->notify_queue[i].sema, 0);
	}

	spin_lock_irqsave(&ch->lock, irq_flags);
	ch->flags |= XPC_C_SETUP;
	spin_unlock_irqrestore(&ch->lock, irq_flags);

	return xpcSuccess;
}


/*
 * Process a connect message from a remote partition.
 *
 * Note: xpc_process_connect() is expecting to be called with the
 * spin_lock_irqsave held and will leave it locked upon return.
 */
static void
xpc_process_connect(struct xpc_channel *ch, unsigned long *irq_flags)
{
	enum xpc_retval ret;


	DBUG_ON(!spin_is_locked(&ch->lock));

	if (!(ch->flags & XPC_C_OPENREQUEST) ||
				!(ch->flags & XPC_C_ROPENREQUEST)) {
		/* nothing more to do for now */
		return;
	}
	DBUG_ON(!(ch->flags & XPC_C_CONNECTING));

	if (!(ch->flags & XPC_C_SETUP)) {
		spin_unlock_irqrestore(&ch->lock, *irq_flags);
		ret = xpc_allocate_msgqueues(ch);
		spin_lock_irqsave(&ch->lock, *irq_flags);

		if (ret != xpcSuccess) {
			XPC_DISCONNECT_CHANNEL(ch, ret, irq_flags);
		}
		if (ch->flags & (XPC_C_CONNECTED | XPC_C_DISCONNECTING)) {
			return;
		}

		DBUG_ON(!(ch->flags & XPC_C_SETUP));
		DBUG_ON(ch->local_msgqueue == NULL);
		DBUG_ON(ch->remote_msgqueue == NULL);
	}

	if (!(ch->flags & XPC_C_OPENREPLY)) {
		ch->flags |= XPC_C_OPENREPLY;
		xpc_IPI_send_openreply(ch, irq_flags);
	}

	if (!(ch->flags & XPC_C_ROPENREPLY)) {
		return;
	}

	DBUG_ON(ch->remote_msgqueue_pa == 0);

	ch->flags = (XPC_C_CONNECTED | XPC_C_SETUP);	/* clear all else */

	dev_info(xpc_chan, "channel %d to partition %d connected\n",
		ch->number, ch->partid);

	spin_unlock_irqrestore(&ch->lock, *irq_flags);
	xpc_create_kthreads(ch, 1);
	spin_lock_irqsave(&ch->lock, *irq_flags);
}


628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676
/*
 * Notify those who wanted to be notified upon delivery of their message.
 */
static void
xpc_notify_senders(struct xpc_channel *ch, enum xpc_retval reason, s64 put)
{
	struct xpc_notify *notify;
	u8 notify_type;
	s64 get = ch->w_remote_GP.get - 1;


	while (++get < put && atomic_read(&ch->n_to_notify) > 0) {

		notify = &ch->notify_queue[get % ch->local_nentries];

		/*
		 * See if the notify entry indicates it was associated with
		 * a message who's sender wants to be notified. It is possible
		 * that it is, but someone else is doing or has done the
		 * notification.
		 */
		notify_type = notify->type;
		if (notify_type == 0 ||
				cmpxchg(&notify->type, notify_type, 0) !=
								notify_type) {
			continue;
		}

		DBUG_ON(notify_type != XPC_N_CALL);

		atomic_dec(&ch->n_to_notify);

		if (notify->func != NULL) {
			dev_dbg(xpc_chan, "notify->func() called, notify=0x%p, "
				"msg_number=%ld, partid=%d, channel=%d\n",
				(void *) notify, get, ch->partid, ch->number);

			notify->func(reason, ch->partid, ch->number,
								notify->key);

			dev_dbg(xpc_chan, "notify->func() returned, "
				"notify=0x%p, msg_number=%ld, partid=%d, "
				"channel=%d\n", (void *) notify, get,
				ch->partid, ch->number);
		}
	}
}


677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731
/*
 * Free up message queues and other stuff that were allocated for the specified
 * channel.
 *
 * Note: ch->reason and ch->reason_line are left set for debugging purposes,
 * they're cleared when XPC_C_DISCONNECTED is cleared.
 */
static void
xpc_free_msgqueues(struct xpc_channel *ch)
{
	DBUG_ON(!spin_is_locked(&ch->lock));
	DBUG_ON(atomic_read(&ch->n_to_notify) != 0);

	ch->remote_msgqueue_pa = 0;
	ch->func = NULL;
	ch->key = NULL;
	ch->msg_size = 0;
	ch->local_nentries = 0;
	ch->remote_nentries = 0;
	ch->kthreads_assigned_limit = 0;
	ch->kthreads_idle_limit = 0;

	ch->local_GP->get = 0;
	ch->local_GP->put = 0;
	ch->remote_GP.get = 0;
	ch->remote_GP.put = 0;
	ch->w_local_GP.get = 0;
	ch->w_local_GP.put = 0;
	ch->w_remote_GP.get = 0;
	ch->w_remote_GP.put = 0;
	ch->next_msg_to_pull = 0;

	if (ch->flags & XPC_C_SETUP) {
		ch->flags &= ~XPC_C_SETUP;

		dev_dbg(xpc_chan, "ch->flags=0x%x, partid=%d, channel=%d\n",
			ch->flags, ch->partid, ch->number);

		kfree(ch->local_msgqueue_base);
		ch->local_msgqueue = NULL;
		kfree(ch->remote_msgqueue_base);
		ch->remote_msgqueue = NULL;
		kfree(ch->notify_queue);
		ch->notify_queue = NULL;
	}
}


/*
 * spin_lock_irqsave() is expected to be held on entry.
 */
static void
xpc_process_disconnect(struct xpc_channel *ch, unsigned long *irq_flags)
{
	struct xpc_partition *part = &xpc_partitions[ch->partid];
732
	u32 channel_was_connected = (ch->flags & XPC_C_WASCONNECTED);
733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749


	DBUG_ON(!spin_is_locked(&ch->lock));

	if (!(ch->flags & XPC_C_DISCONNECTING)) {
		return;
	}

	DBUG_ON(!(ch->flags & XPC_C_CLOSEREQUEST));

	/* make sure all activity has settled down first */

	if (atomic_read(&ch->references) > 0) {
		return;
	}
	DBUG_ON(atomic_read(&ch->kthreads_assigned) != 0);

750 751 752 753 754
	if (part->act_state == XPC_P_DEACTIVATING) {
		/* can't proceed until the other side disengages from us */
		if (xpc_partition_engaged(1UL << ch->partid)) {
			return;
		}
755

756
	} else {
757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773

		/* as long as the other side is up do the full protocol */

		if (!(ch->flags & XPC_C_RCLOSEREQUEST)) {
			return;
		}

		if (!(ch->flags & XPC_C_CLOSEREPLY)) {
			ch->flags |= XPC_C_CLOSEREPLY;
			xpc_IPI_send_closereply(ch, irq_flags);
		}

		if (!(ch->flags & XPC_C_RCLOSEREPLY)) {
			return;
		}
	}

774 775 776 777 778 779
	/* wake those waiting for notify completion */
	if (atomic_read(&ch->n_to_notify) > 0) {
		/* >>> we do callout while holding ch->lock */
		xpc_notify_senders(ch, ch->reason, ch->w_local_GP.put);
	}

780 781
	/* both sides are disconnected now */

782 783 784 785 786 787
	if (ch->flags & XPC_C_CONNECTCALLOUT) {
		spin_unlock_irqrestore(&ch->lock, *irq_flags);
		xpc_disconnect_callout(ch, xpcDisconnected);
		spin_lock_irqsave(&ch->lock, *irq_flags);
	}

788 789 790 791 792
	/* it's now safe to free the channel's message queues */
	xpc_free_msgqueues(ch);

	/* mark disconnected, clear all other flags except XPC_C_WDISCONNECT */
	ch->flags = (XPC_C_DISCONNECTED | (ch->flags & XPC_C_WDISCONNECT));
793 794 795

	atomic_dec(&part->nchannels_active);

796
	if (channel_was_connected) {
797 798 799
		dev_info(xpc_chan, "channel %d to partition %d disconnected, "
			"reason=%d\n", ch->number, ch->partid, ch->reason);
	}
800 801 802 803 804

	if (ch->flags & XPC_C_WDISCONNECT) {
		spin_unlock_irqrestore(&ch->lock, *irq_flags);
		up(&ch->wdisconnect_sema);
		spin_lock_irqsave(&ch->lock, *irq_flags);
805 806 807 808 809 810 811 812 813 814

	} else if (ch->delayed_IPI_flags) {
		if (part->act_state != XPC_P_DEACTIVATING) {
			/* time to take action on any delayed IPI flags */
			spin_lock(&part->IPI_lock);
			XPC_SET_IPI_FLAGS(part->local_IPI_amo, ch->number,
							ch->delayed_IPI_flags);
			spin_unlock(&part->IPI_lock);
		}
		ch->delayed_IPI_flags = 0;
815
	}
816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835
}


/*
 * Process a change in the channel's remote connection state.
 */
static void
xpc_process_openclose_IPI(struct xpc_partition *part, int ch_number,
				u8 IPI_flags)
{
	unsigned long irq_flags;
	struct xpc_openclose_args *args =
				&part->remote_openclose_args[ch_number];
	struct xpc_channel *ch = &part->channels[ch_number];
	enum xpc_retval reason;



	spin_lock_irqsave(&ch->lock, irq_flags);

836 837 838 839 840 841 842 843 844 845 846 847 848
again:

	if ((ch->flags & XPC_C_DISCONNECTED) &&
					(ch->flags & XPC_C_WDISCONNECT)) {
		/*
		 * Delay processing IPI flags until thread waiting disconnect
		 * has had a chance to see that the channel is disconnected.
		 */
		ch->delayed_IPI_flags |= IPI_flags;
		spin_unlock_irqrestore(&ch->lock, irq_flags);
		return;
	}

849 850 851 852 853 854 855 856 857 858

	if (IPI_flags & XPC_IPI_CLOSEREQUEST) {

		dev_dbg(xpc_chan, "XPC_IPI_CLOSEREQUEST (reason=%d) received "
			"from partid=%d, channel=%d\n", args->reason,
			ch->partid, ch->number);

		/*
		 * If RCLOSEREQUEST is set, we're probably waiting for
		 * RCLOSEREPLY. We should find it and a ROPENREQUEST packed
859
		 * with this RCLOSEREQUEST in the IPI_flags.
860 861 862 863 864 865 866 867 868 869 870 871 872 873
		 */

		if (ch->flags & XPC_C_RCLOSEREQUEST) {
			DBUG_ON(!(ch->flags & XPC_C_DISCONNECTING));
			DBUG_ON(!(ch->flags & XPC_C_CLOSEREQUEST));
			DBUG_ON(!(ch->flags & XPC_C_CLOSEREPLY));
			DBUG_ON(ch->flags & XPC_C_RCLOSEREPLY);

			DBUG_ON(!(IPI_flags & XPC_IPI_CLOSEREPLY));
			IPI_flags &= ~XPC_IPI_CLOSEREPLY;
			ch->flags |= XPC_C_RCLOSEREPLY;

			/* both sides have finished disconnecting */
			xpc_process_disconnect(ch, &irq_flags);
874 875
			DBUG_ON(!(ch->flags & XPC_C_DISCONNECTED));
			goto again;
876 877 878 879
		}

		if (ch->flags & XPC_C_DISCONNECTED) {
			if (!(IPI_flags & XPC_IPI_OPENREQUEST)) {
880 881 882 883 884 885 886 887 888 889
				if ((XPC_GET_IPI_FLAGS(part->local_IPI_amo,
					 ch_number) & XPC_IPI_OPENREQUEST)) {

					DBUG_ON(ch->delayed_IPI_flags != 0);
					spin_lock(&part->IPI_lock);
					XPC_SET_IPI_FLAGS(part->local_IPI_amo,
							ch_number,
							XPC_IPI_CLOSEREQUEST);
					spin_unlock(&part->IPI_lock);
				}
890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918
				spin_unlock_irqrestore(&ch->lock, irq_flags);
				return;
			}

			XPC_SET_REASON(ch, 0, 0);
			ch->flags &= ~XPC_C_DISCONNECTED;

			atomic_inc(&part->nchannels_active);
			ch->flags |= (XPC_C_CONNECTING | XPC_C_ROPENREQUEST);
		}

		IPI_flags &= ~(XPC_IPI_OPENREQUEST | XPC_IPI_OPENREPLY);

		/*
		 * The meaningful CLOSEREQUEST connection state fields are:
		 *      reason = reason connection is to be closed
		 */

		ch->flags |= XPC_C_RCLOSEREQUEST;

		if (!(ch->flags & XPC_C_DISCONNECTING)) {
			reason = args->reason;
			if (reason <= xpcSuccess || reason > xpcUnknownReason) {
				reason = xpcUnknownReason;
			} else if (reason == xpcUnregistering) {
				reason = xpcOtherUnregistering;
			}

			XPC_DISCONNECT_CHANNEL(ch, reason, &irq_flags);
919 920 921 922

			DBUG_ON(IPI_flags & XPC_IPI_CLOSEREPLY);
			spin_unlock_irqrestore(&ch->lock, irq_flags);
			return;
923
		}
924 925

		xpc_process_disconnect(ch, &irq_flags);
926 927 928 929 930 931 932 933 934 935 936 937 938 939 940
	}


	if (IPI_flags & XPC_IPI_CLOSEREPLY) {

		dev_dbg(xpc_chan, "XPC_IPI_CLOSEREPLY received from partid=%d,"
			" channel=%d\n", ch->partid, ch->number);

		if (ch->flags & XPC_C_DISCONNECTED) {
			DBUG_ON(part->act_state != XPC_P_DEACTIVATING);
			spin_unlock_irqrestore(&ch->lock, irq_flags);
			return;
		}

		DBUG_ON(!(ch->flags & XPC_C_CLOSEREQUEST));
941 942 943 944 945 946 947 948 949 950 951 952 953 954

		if (!(ch->flags & XPC_C_RCLOSEREQUEST)) {
			if ((XPC_GET_IPI_FLAGS(part->local_IPI_amo, ch_number)
						& XPC_IPI_CLOSEREQUEST)) {

				DBUG_ON(ch->delayed_IPI_flags != 0);
				spin_lock(&part->IPI_lock);
				XPC_SET_IPI_FLAGS(part->local_IPI_amo,
						ch_number, XPC_IPI_CLOSEREPLY);
				spin_unlock(&part->IPI_lock);
			}
			spin_unlock_irqrestore(&ch->lock, irq_flags);
			return;
		}
955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971

		ch->flags |= XPC_C_RCLOSEREPLY;

		if (ch->flags & XPC_C_CLOSEREPLY) {
			/* both sides have finished disconnecting */
			xpc_process_disconnect(ch, &irq_flags);
		}
	}


	if (IPI_flags & XPC_IPI_OPENREQUEST) {

		dev_dbg(xpc_chan, "XPC_IPI_OPENREQUEST (msg_size=%d, "
			"local_nentries=%d) received from partid=%d, "
			"channel=%d\n", args->msg_size, args->local_nentries,
			ch->partid, ch->number);

972 973 974 975 976 977 978 979
		if (part->act_state == XPC_P_DEACTIVATING ||
					(ch->flags & XPC_C_ROPENREQUEST)) {
			spin_unlock_irqrestore(&ch->lock, irq_flags);
			return;
		}

		if (ch->flags & (XPC_C_DISCONNECTING | XPC_C_WDISCONNECT)) {
			ch->delayed_IPI_flags |= XPC_IPI_OPENREQUEST;
980 981 982 983 984 985 986 987 988 989 990 991 992
			spin_unlock_irqrestore(&ch->lock, irq_flags);
			return;
		}
		DBUG_ON(!(ch->flags & (XPC_C_DISCONNECTED |
							XPC_C_OPENREQUEST)));
		DBUG_ON(ch->flags & (XPC_C_ROPENREQUEST | XPC_C_ROPENREPLY |
					XPC_C_OPENREPLY | XPC_C_CONNECTED));

		/*
		 * The meaningful OPENREQUEST connection state fields are:
		 *      msg_size = size of channel's messages in bytes
		 *      local_nentries = remote partition's local_nentries
		 */
993 994 995 996 997
		if (args->msg_size == 0 || args->local_nentries == 0) {
			/* assume OPENREQUEST was delayed by mistake */
			spin_unlock_irqrestore(&ch->lock, irq_flags);
			return;
		}
998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034

		ch->flags |= (XPC_C_ROPENREQUEST | XPC_C_CONNECTING);
		ch->remote_nentries = args->local_nentries;


		if (ch->flags & XPC_C_OPENREQUEST) {
			if (args->msg_size != ch->msg_size) {
				XPC_DISCONNECT_CHANNEL(ch, xpcUnequalMsgSizes,
								&irq_flags);
				spin_unlock_irqrestore(&ch->lock, irq_flags);
				return;
			}
		} else {
			ch->msg_size = args->msg_size;

			XPC_SET_REASON(ch, 0, 0);
			ch->flags &= ~XPC_C_DISCONNECTED;

			atomic_inc(&part->nchannels_active);
		}

		xpc_process_connect(ch, &irq_flags);
	}


	if (IPI_flags & XPC_IPI_OPENREPLY) {

		dev_dbg(xpc_chan, "XPC_IPI_OPENREPLY (local_msgqueue_pa=0x%lx, "
			"local_nentries=%d, remote_nentries=%d) received from "
			"partid=%d, channel=%d\n", args->local_msgqueue_pa,
			args->local_nentries, args->remote_nentries,
			ch->partid, ch->number);

		if (ch->flags & (XPC_C_DISCONNECTING | XPC_C_DISCONNECTED)) {
			spin_unlock_irqrestore(&ch->lock, irq_flags);
			return;
		}
1035 1036 1037 1038 1039 1040 1041
		if (!(ch->flags & XPC_C_OPENREQUEST)) {
			XPC_DISCONNECT_CHANNEL(ch, xpcOpenCloseError,
								&irq_flags);
			spin_unlock_irqrestore(&ch->lock, irq_flags);
			return;
		}

1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094
		DBUG_ON(!(ch->flags & XPC_C_ROPENREQUEST));
		DBUG_ON(ch->flags & XPC_C_CONNECTED);

		/*
		 * The meaningful OPENREPLY connection state fields are:
		 *      local_msgqueue_pa = physical address of remote
		 *			    partition's local_msgqueue
		 *      local_nentries = remote partition's local_nentries
		 *      remote_nentries = remote partition's remote_nentries
		 */
		DBUG_ON(args->local_msgqueue_pa == 0);
		DBUG_ON(args->local_nentries == 0);
		DBUG_ON(args->remote_nentries == 0);

		ch->flags |= XPC_C_ROPENREPLY;
		ch->remote_msgqueue_pa = args->local_msgqueue_pa;

		if (args->local_nentries < ch->remote_nentries) {
			dev_dbg(xpc_chan, "XPC_IPI_OPENREPLY: new "
				"remote_nentries=%d, old remote_nentries=%d, "
				"partid=%d, channel=%d\n",
				args->local_nentries, ch->remote_nentries,
				ch->partid, ch->number);

			ch->remote_nentries = args->local_nentries;
		}
		if (args->remote_nentries < ch->local_nentries) {
			dev_dbg(xpc_chan, "XPC_IPI_OPENREPLY: new "
				"local_nentries=%d, old local_nentries=%d, "
				"partid=%d, channel=%d\n",
				args->remote_nentries, ch->local_nentries,
				ch->partid, ch->number);

			ch->local_nentries = args->remote_nentries;
		}

		xpc_process_connect(ch, &irq_flags);
	}

	spin_unlock_irqrestore(&ch->lock, irq_flags);
}


/*
 * Attempt to establish a channel connection to a remote partition.
 */
static enum xpc_retval
xpc_connect_channel(struct xpc_channel *ch)
{
	unsigned long irq_flags;
	struct xpc_registration *registration = &xpc_registrations[ch->number];


1095 1096
	if (down_trylock(&registration->sema) != 0) {
		return xpcRetry;
1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268
	}

	if (!XPC_CHANNEL_REGISTERED(ch->number)) {
		up(&registration->sema);
		return xpcUnregistered;
	}

	spin_lock_irqsave(&ch->lock, irq_flags);

	DBUG_ON(ch->flags & XPC_C_CONNECTED);
	DBUG_ON(ch->flags & XPC_C_OPENREQUEST);

	if (ch->flags & XPC_C_DISCONNECTING) {
		spin_unlock_irqrestore(&ch->lock, irq_flags);
		up(&registration->sema);
		return ch->reason;
	}


	/* add info from the channel connect registration to the channel */

	ch->kthreads_assigned_limit = registration->assigned_limit;
	ch->kthreads_idle_limit = registration->idle_limit;
	DBUG_ON(atomic_read(&ch->kthreads_assigned) != 0);
	DBUG_ON(atomic_read(&ch->kthreads_idle) != 0);
	DBUG_ON(atomic_read(&ch->kthreads_active) != 0);

	ch->func = registration->func;
	DBUG_ON(registration->func == NULL);
	ch->key = registration->key;

	ch->local_nentries = registration->nentries;

	if (ch->flags & XPC_C_ROPENREQUEST) {
		if (registration->msg_size != ch->msg_size) {
			/* the local and remote sides aren't the same */

			/*
			 * Because XPC_DISCONNECT_CHANNEL() can block we're
			 * forced to up the registration sema before we unlock
			 * the channel lock. But that's okay here because we're
			 * done with the part that required the registration
			 * sema. XPC_DISCONNECT_CHANNEL() requires that the
			 * channel lock be locked and will unlock and relock
			 * the channel lock as needed.
			 */
			up(&registration->sema);
			XPC_DISCONNECT_CHANNEL(ch, xpcUnequalMsgSizes,
								&irq_flags);
			spin_unlock_irqrestore(&ch->lock, irq_flags);
			return xpcUnequalMsgSizes;
		}
	} else {
		ch->msg_size = registration->msg_size;

		XPC_SET_REASON(ch, 0, 0);
		ch->flags &= ~XPC_C_DISCONNECTED;

		atomic_inc(&xpc_partitions[ch->partid].nchannels_active);
	}

	up(&registration->sema);


	/* initiate the connection */

	ch->flags |= (XPC_C_OPENREQUEST | XPC_C_CONNECTING);
	xpc_IPI_send_openrequest(ch, &irq_flags);

	xpc_process_connect(ch, &irq_flags);

	spin_unlock_irqrestore(&ch->lock, irq_flags);

	return xpcSuccess;
}


/*
 * Clear some of the msg flags in the local message queue.
 */
static inline void
xpc_clear_local_msgqueue_flags(struct xpc_channel *ch)
{
	struct xpc_msg *msg;
	s64 get;


	get = ch->w_remote_GP.get;
	do {
		msg = (struct xpc_msg *) ((u64) ch->local_msgqueue +
				(get % ch->local_nentries) * ch->msg_size);
		msg->flags = 0;
	} while (++get < (volatile s64) ch->remote_GP.get);
}


/*
 * Clear some of the msg flags in the remote message queue.
 */
static inline void
xpc_clear_remote_msgqueue_flags(struct xpc_channel *ch)
{
	struct xpc_msg *msg;
	s64 put;


	put = ch->w_remote_GP.put;
	do {
		msg = (struct xpc_msg *) ((u64) ch->remote_msgqueue +
				(put % ch->remote_nentries) * ch->msg_size);
		msg->flags = 0;
	} while (++put < (volatile s64) ch->remote_GP.put);
}


static void
xpc_process_msg_IPI(struct xpc_partition *part, int ch_number)
{
	struct xpc_channel *ch = &part->channels[ch_number];
	int nmsgs_sent;


	ch->remote_GP = part->remote_GPs[ch_number];


	/* See what, if anything, has changed for each connected channel */

	xpc_msgqueue_ref(ch);

	if (ch->w_remote_GP.get == ch->remote_GP.get &&
				ch->w_remote_GP.put == ch->remote_GP.put) {
		/* nothing changed since GPs were last pulled */
		xpc_msgqueue_deref(ch);
		return;
	}

	if (!(ch->flags & XPC_C_CONNECTED)){
		xpc_msgqueue_deref(ch);
		return;
	}


	/*
	 * First check to see if messages recently sent by us have been
	 * received by the other side. (The remote GET value will have
	 * changed since we last looked at it.)
	 */

	if (ch->w_remote_GP.get != ch->remote_GP.get) {

		/*
		 * We need to notify any senders that want to be notified
		 * that their sent messages have been received by their
		 * intended recipients. We need to do this before updating
		 * w_remote_GP.get so that we don't allocate the same message
		 * queue entries prematurely (see xpc_allocate_msg()).
		 */
		if (atomic_read(&ch->n_to_notify) > 0) {
			/*
			 * Notify senders that messages sent have been
			 * received and delivered by the other side.
			 */
			xpc_notify_senders(ch, xpcMsgDelivered,
							ch->remote_GP.get);
		}

		/*
		 * Clear msg->flags in previously sent messages, so that
		 * they're ready for xpc_allocate_msg().
		 */
		xpc_clear_local_msgqueue_flags(ch);

1269
		ch->w_remote_GP.get = ch->remote_GP.get;
1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296

		dev_dbg(xpc_chan, "w_remote_GP.get changed to %ld, partid=%d, "
			"channel=%d\n", ch->w_remote_GP.get, ch->partid,
			ch->number);

		/*
		 * If anyone was waiting for message queue entries to become
		 * available, wake them up.
		 */
		if (atomic_read(&ch->n_on_msg_allocate_wq) > 0) {
			wake_up(&ch->msg_allocate_wq);
		}
	}


	/*
	 * Now check for newly sent messages by the other side. (The remote
	 * PUT value will have changed since we last looked at it.)
	 */

	if (ch->w_remote_GP.put != ch->remote_GP.put) {
		/*
		 * Clear msg->flags in previously received messages, so that
		 * they're ready for xpc_get_deliverable_msg().
		 */
		xpc_clear_remote_msgqueue_flags(ch);

1297
		ch->w_remote_GP.put = ch->remote_GP.put;
1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325

		dev_dbg(xpc_chan, "w_remote_GP.put changed to %ld, partid=%d, "
			"channel=%d\n", ch->w_remote_GP.put, ch->partid,
			ch->number);

		nmsgs_sent = ch->w_remote_GP.put - ch->w_local_GP.get;
		if (nmsgs_sent > 0) {
			dev_dbg(xpc_chan, "msgs waiting to be copied and "
				"delivered=%d, partid=%d, channel=%d\n",
				nmsgs_sent, ch->partid, ch->number);

			if (ch->flags & XPC_C_CONNECTCALLOUT) {
				xpc_activate_kthreads(ch, nmsgs_sent);
			}
		}
	}

	xpc_msgqueue_deref(ch);
}


void
xpc_process_channel_activity(struct xpc_partition *part)
{
	unsigned long irq_flags;
	u64 IPI_amo, IPI_flags;
	struct xpc_channel *ch;
	int ch_number;
1326
	u32 ch_flags;
1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352


	IPI_amo = xpc_get_IPI_flags(part);

	/*
	 * Initiate channel connections for registered channels.
	 *
	 * For each connected channel that has pending messages activate idle
	 * kthreads and/or create new kthreads as needed.
	 */

	for (ch_number = 0; ch_number < part->nchannels; ch_number++) {
		ch = &part->channels[ch_number];


		/*
		 * Process any open or close related IPI flags, and then deal
		 * with connecting or disconnecting the channel as required.
		 */

		IPI_flags = XPC_GET_IPI_FLAGS(IPI_amo, ch_number);

		if (XPC_ANY_OPENCLOSE_IPI_FLAGS_SET(IPI_flags)) {
			xpc_process_openclose_IPI(part, ch_number, IPI_flags);
		}

1353
		ch_flags = ch->flags;	/* need an atomic snapshot of flags */
1354

1355
		if (ch_flags & XPC_C_DISCONNECTING) {
1356 1357 1358 1359 1360 1361 1362 1363 1364 1365
			spin_lock_irqsave(&ch->lock, irq_flags);
			xpc_process_disconnect(ch, &irq_flags);
			spin_unlock_irqrestore(&ch->lock, irq_flags);
			continue;
		}

		if (part->act_state == XPC_P_DEACTIVATING) {
			continue;
		}

1366 1367 1368
		if (!(ch_flags & XPC_C_CONNECTED)) {
			if (!(ch_flags & XPC_C_OPENREQUEST)) {
				DBUG_ON(ch_flags & XPC_C_SETUP);
1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392
				(void) xpc_connect_channel(ch);
			} else {
				spin_lock_irqsave(&ch->lock, irq_flags);
				xpc_process_connect(ch, &irq_flags);
				spin_unlock_irqrestore(&ch->lock, irq_flags);
			}
			continue;
		}


		/*
		 * Process any message related IPI flags, this may involve the
		 * activation of kthreads to deliver any pending messages sent
		 * from the other partition.
		 */

		if (XPC_ANY_MSG_IPI_FLAGS_SET(IPI_flags)) {
			xpc_process_msg_IPI(part, ch_number);
		}
	}
}


/*
1393 1394
 * XPC's heartbeat code calls this function to inform XPC that a partition is
 * going down.  XPC responds by tearing down the XPartition Communication
1395 1396 1397 1398 1399 1400 1401
 * infrastructure used for the just downed partition.
 *
 * XPC's heartbeat code will never call this function and xpc_partition_up()
 * at the same time. Nor will it ever make multiple calls to either function
 * at the same time.
 */
void
1402
xpc_partition_going_down(struct xpc_partition *part, enum xpc_retval reason)
1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417
{
	unsigned long irq_flags;
	int ch_number;
	struct xpc_channel *ch;


	dev_dbg(xpc_chan, "deactivating partition %d, reason=%d\n",
		XPC_PARTID(part), reason);

	if (!xpc_part_ref(part)) {
		/* infrastructure for this partition isn't currently set up */
		return;
	}


1418
	/* disconnect channels associated with the partition going down */
1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456

	for (ch_number = 0; ch_number < part->nchannels; ch_number++) {
		ch = &part->channels[ch_number];

		xpc_msgqueue_ref(ch);
		spin_lock_irqsave(&ch->lock, irq_flags);

		XPC_DISCONNECT_CHANNEL(ch, reason, &irq_flags);

		spin_unlock_irqrestore(&ch->lock, irq_flags);
		xpc_msgqueue_deref(ch);
	}

	xpc_wakeup_channel_mgr(part);

	xpc_part_deref(part);
}


/*
 * Teardown the infrastructure necessary to support XPartition Communication
 * between the specified remote partition and the local one.
 */
void
xpc_teardown_infrastructure(struct xpc_partition *part)
{
	partid_t partid = XPC_PARTID(part);


	/*
	 * We start off by making this partition inaccessible to local
	 * processes by marking it as no longer setup. Then we make it
	 * inaccessible to remote processes by clearing the XPC per partition
	 * specific variable's magic # (which indicates that these variables
	 * are no longer valid) and by ignoring all XPC notify IPIs sent to
	 * this partition.
	 */

1457
	DBUG_ON(atomic_read(&part->nchannels_engaged) != 0);
1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515
	DBUG_ON(atomic_read(&part->nchannels_active) != 0);
	DBUG_ON(part->setup_state != XPC_P_SETUP);
	part->setup_state = XPC_P_WTEARDOWN;

	xpc_vars_part[partid].magic = 0;


	free_irq(SGI_XPC_NOTIFY, (void *) (u64) partid);


	/*
	 * Before proceding with the teardown we have to wait until all
	 * existing references cease.
	 */
	wait_event(part->teardown_wq, (atomic_read(&part->references) == 0));


	/* now we can begin tearing down the infrastructure */

	part->setup_state = XPC_P_TORNDOWN;

	/* in case we've still got outstanding timers registered... */
	del_timer_sync(&part->dropped_IPI_timer);

	kfree(part->remote_openclose_args_base);
	part->remote_openclose_args = NULL;
	kfree(part->local_openclose_args_base);
	part->local_openclose_args = NULL;
	kfree(part->remote_GPs_base);
	part->remote_GPs = NULL;
	kfree(part->local_GPs_base);
	part->local_GPs = NULL;
	kfree(part->channels);
	part->channels = NULL;
	part->local_IPI_amo_va = NULL;
}


/*
 * Called by XP at the time of channel connection registration to cause
 * XPC to establish connections to all currently active partitions.
 */
void
xpc_initiate_connect(int ch_number)
{
	partid_t partid;
	struct xpc_partition *part;
	struct xpc_channel *ch;


	DBUG_ON(ch_number < 0 || ch_number >= XPC_NCHANNELS);

	for (partid = 1; partid < XP_MAX_PARTITIONS; partid++) {
		part = &xpc_partitions[partid];

		if (xpc_part_ref(part)) {
			ch = &part->channels[ch_number];

1516 1517 1518 1519 1520
			/*
			 * Initiate the establishment of a connection on the
			 * newly registered channel to the remote partition.
			 */
			xpc_wakeup_channel_mgr(part);
1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578
			xpc_part_deref(part);
		}
	}
}


void
xpc_connected_callout(struct xpc_channel *ch)
{
	/* let the registerer know that a connection has been established */

	if (ch->func != NULL) {
		dev_dbg(xpc_chan, "ch->func() called, reason=xpcConnected, "
			"partid=%d, channel=%d\n", ch->partid, ch->number);

		ch->func(xpcConnected, ch->partid, ch->number,
				(void *) (u64) ch->local_nentries, ch->key);

		dev_dbg(xpc_chan, "ch->func() returned, reason=xpcConnected, "
			"partid=%d, channel=%d\n", ch->partid, ch->number);
	}
}


/*
 * Called by XP at the time of channel connection unregistration to cause
 * XPC to teardown all current connections for the specified channel.
 *
 * Before returning xpc_initiate_disconnect() will wait until all connections
 * on the specified channel have been closed/torndown. So the caller can be
 * assured that they will not be receiving any more callouts from XPC to the
 * function they registered via xpc_connect().
 *
 * Arguments:
 *
 *	ch_number - channel # to unregister.
 */
void
xpc_initiate_disconnect(int ch_number)
{
	unsigned long irq_flags;
	partid_t partid;
	struct xpc_partition *part;
	struct xpc_channel *ch;


	DBUG_ON(ch_number < 0 || ch_number >= XPC_NCHANNELS);

	/* initiate the channel disconnect for every active partition */
	for (partid = 1; partid < XP_MAX_PARTITIONS; partid++) {
		part = &xpc_partitions[partid];

		if (xpc_part_ref(part)) {
			ch = &part->channels[ch_number];
			xpc_msgqueue_ref(ch);

			spin_lock_irqsave(&ch->lock, irq_flags);

1579 1580 1581 1582
			if (!(ch->flags & XPC_C_DISCONNECTED)) {
				ch->flags |= XPC_C_WDISCONNECT;

				XPC_DISCONNECT_CHANNEL(ch, xpcUnregistering,
1583
								&irq_flags);
1584
			}
1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599

			spin_unlock_irqrestore(&ch->lock, irq_flags);

			xpc_msgqueue_deref(ch);
			xpc_part_deref(part);
		}
	}

	xpc_disconnect_wait(ch_number);
}


/*
 * To disconnect a channel, and reflect it back to all who may be waiting.
 *
1600 1601 1602
 * An OPEN is not allowed until XPC_C_DISCONNECTING is cleared by
 * xpc_process_disconnect(), and if set, XPC_C_WDISCONNECT is cleared by
 * xpc_disconnect_wait().
1603 1604 1605 1606 1607 1608 1609
 *
 * THE CHANNEL IS TO BE LOCKED BY THE CALLER AND WILL REMAIN LOCKED UPON RETURN.
 */
void
xpc_disconnect_channel(const int line, struct xpc_channel *ch,
			enum xpc_retval reason, unsigned long *irq_flags)
{
1610
	u32 channel_was_connected = (ch->flags & XPC_C_CONNECTED);
1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624


	DBUG_ON(!spin_is_locked(&ch->lock));

	if (ch->flags & (XPC_C_DISCONNECTING | XPC_C_DISCONNECTED)) {
		return;
	}
	DBUG_ON(!(ch->flags & (XPC_C_CONNECTING | XPC_C_CONNECTED)));

	dev_dbg(xpc_chan, "reason=%d, line=%d, partid=%d, channel=%d\n",
		reason, line, ch->partid, ch->number);

	XPC_SET_REASON(ch, reason, line);

1625
	ch->flags |= (XPC_C_CLOSEREQUEST | XPC_C_DISCONNECTING);
1626 1627 1628 1629 1630 1631 1632
	/* some of these may not have been set */
	ch->flags &= ~(XPC_C_OPENREQUEST | XPC_C_OPENREPLY |
			XPC_C_ROPENREQUEST | XPC_C_ROPENREPLY |
			XPC_C_CONNECTING | XPC_C_CONNECTED);

	xpc_IPI_send_closerequest(ch, irq_flags);

1633
	if (channel_was_connected) {
1634 1635 1636
		ch->flags |= XPC_C_WASCONNECTED;
	}

1637 1638 1639
	spin_unlock_irqrestore(&ch->lock, *irq_flags);

	/* wake all idle kthreads so they can exit */
1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653
	if (atomic_read(&ch->kthreads_idle) > 0) {
		wake_up_all(&ch->idle_wq);
	}

	/* wake those waiting to allocate an entry from the local msg queue */
	if (atomic_read(&ch->n_on_msg_allocate_wq) > 0) {
		wake_up(&ch->msg_allocate_wq);
	}

	spin_lock_irqsave(&ch->lock, *irq_flags);
}


void
1654
xpc_disconnect_callout(struct xpc_channel *ch, enum xpc_retval reason)
1655 1656
{
	/*
1657
	 * Let the channel's registerer know that the channel is being
1658
	 * disconnected. We don't want to do this if the registerer was never
1659
	 * informed of a connection being made.
1660 1661 1662
	 */

	if (ch->func != NULL) {
1663 1664
		dev_dbg(xpc_chan, "ch->func() called, reason=%d, partid=%d, "
			"channel=%d\n", reason, ch->partid, ch->number);
1665

1666
		ch->func(reason, ch->partid, ch->number, NULL, ch->key);
1667

1668 1669
		dev_dbg(xpc_chan, "ch->func() returned, reason=%d, partid=%d, "
			"channel=%d\n", reason, ch->partid, ch->number);
1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915
	}
}


/*
 * Wait for a message entry to become available for the specified channel,
 * but don't wait any longer than 1 jiffy.
 */
static enum xpc_retval
xpc_allocate_msg_wait(struct xpc_channel *ch)
{
	enum xpc_retval ret;


	if (ch->flags & XPC_C_DISCONNECTING) {
		DBUG_ON(ch->reason == xpcInterrupted);  // >>> Is this true?
		return ch->reason;
	}

	atomic_inc(&ch->n_on_msg_allocate_wq);
	ret = interruptible_sleep_on_timeout(&ch->msg_allocate_wq, 1);
	atomic_dec(&ch->n_on_msg_allocate_wq);

	if (ch->flags & XPC_C_DISCONNECTING) {
		ret = ch->reason;
		DBUG_ON(ch->reason == xpcInterrupted);  // >>> Is this true?
	} else if (ret == 0) {
		ret = xpcTimeout;
	} else {
		ret = xpcInterrupted;
	}

	return ret;
}


/*
 * Allocate an entry for a message from the message queue associated with the
 * specified channel.
 */
static enum xpc_retval
xpc_allocate_msg(struct xpc_channel *ch, u32 flags,
			struct xpc_msg **address_of_msg)
{
	struct xpc_msg *msg;
	enum xpc_retval ret;
	s64 put;


	/* this reference will be dropped in xpc_send_msg() */
	xpc_msgqueue_ref(ch);

	if (ch->flags & XPC_C_DISCONNECTING) {
		xpc_msgqueue_deref(ch);
		return ch->reason;
	}
	if (!(ch->flags & XPC_C_CONNECTED)) {
		xpc_msgqueue_deref(ch);
		return xpcNotConnected;
	}


	/*
	 * Get the next available message entry from the local message queue.
	 * If none are available, we'll make sure that we grab the latest
	 * GP values.
	 */
	ret = xpcTimeout;

	while (1) {

		put = (volatile s64) ch->w_local_GP.put;
		if (put - (volatile s64) ch->w_remote_GP.get <
							ch->local_nentries) {

			/* There are available message entries. We need to try
			 * to secure one for ourselves. We'll do this by trying
			 * to increment w_local_GP.put as long as someone else
			 * doesn't beat us to it. If they do, we'll have to
			 * try again.
		 	 */
			if (cmpxchg(&ch->w_local_GP.put, put, put + 1) ==
									put) {
				/* we got the entry referenced by put */
				break;
			}
			continue;	/* try again */
		}


		/*
		 * There aren't any available msg entries at this time.
		 *
		 * In waiting for a message entry to become available,
		 * we set a timeout in case the other side is not
		 * sending completion IPIs. This lets us fake an IPI
		 * that will cause the IPI handler to fetch the latest
		 * GP values as if an IPI was sent by the other side.
		 */
		if (ret == xpcTimeout) {
			xpc_IPI_send_local_msgrequest(ch);
		}

		if (flags & XPC_NOWAIT) {
			xpc_msgqueue_deref(ch);
			return xpcNoWait;
		}

		ret = xpc_allocate_msg_wait(ch);
		if (ret != xpcInterrupted && ret != xpcTimeout) {
			xpc_msgqueue_deref(ch);
			return ret;
		}
	}


	/* get the message's address and initialize it */
	msg = (struct xpc_msg *) ((u64) ch->local_msgqueue +
				(put % ch->local_nentries) * ch->msg_size);


	DBUG_ON(msg->flags != 0);
	msg->number = put;

	dev_dbg(xpc_chan, "w_local_GP.put changed to %ld; msg=0x%p, "
		"msg_number=%ld, partid=%d, channel=%d\n", put + 1,
		(void *) msg, msg->number, ch->partid, ch->number);

	*address_of_msg = msg;

	return xpcSuccess;
}


/*
 * Allocate an entry for a message from the message queue associated with the
 * specified channel. NOTE that this routine can sleep waiting for a message
 * entry to become available. To not sleep, pass in the XPC_NOWAIT flag.
 *
 * Arguments:
 *
 *	partid - ID of partition to which the channel is connected.
 *	ch_number - channel #.
 *	flags - see xpc.h for valid flags.
 *	payload - address of the allocated payload area pointer (filled in on
 * 	          return) in which the user-defined message is constructed.
 */
enum xpc_retval
xpc_initiate_allocate(partid_t partid, int ch_number, u32 flags, void **payload)
{
	struct xpc_partition *part = &xpc_partitions[partid];
	enum xpc_retval ret = xpcUnknownReason;
	struct xpc_msg *msg;


	DBUG_ON(partid <= 0 || partid >= XP_MAX_PARTITIONS);
	DBUG_ON(ch_number < 0 || ch_number >= part->nchannels);

	*payload = NULL;

	if (xpc_part_ref(part)) {
		ret = xpc_allocate_msg(&part->channels[ch_number], flags, &msg);
		xpc_part_deref(part);

		if (msg != NULL) {
			*payload = &msg->payload;
		}
	}

	return ret;
}


/*
 * Now we actually send the messages that are ready to be sent by advancing
 * the local message queue's Put value and then send an IPI to the recipient
 * partition.
 */
static void
xpc_send_msgs(struct xpc_channel *ch, s64 initial_put)
{
	struct xpc_msg *msg;
	s64 put = initial_put + 1;
	int send_IPI = 0;


	while (1) {

		while (1) {
			if (put == (volatile s64) ch->w_local_GP.put) {
				break;
			}

			msg = (struct xpc_msg *) ((u64) ch->local_msgqueue +
			       (put % ch->local_nentries) * ch->msg_size);

			if (!(msg->flags & XPC_M_READY)) {
				break;
			}

			put++;
		}

		if (put == initial_put) {
			/* nothing's changed */
			break;
		}

		if (cmpxchg_rel(&ch->local_GP->put, initial_put, put) !=
								initial_put) {
			/* someone else beat us to it */
			DBUG_ON((volatile s64) ch->local_GP->put < initial_put);
			break;
		}

		/* we just set the new value of local_GP->put */

		dev_dbg(xpc_chan, "local_GP->put changed to %ld, partid=%d, "
			"channel=%d\n", put, ch->partid, ch->number);

		send_IPI = 1;

		/*
		 * We need to ensure that the message referenced by
		 * local_GP->put is not XPC_M_READY or that local_GP->put
		 * equals w_local_GP.put, so we'll go have a look.
		 */
		initial_put = put;
	}

	if (send_IPI) {
		xpc_IPI_send_msgrequest(ch);
	}
}


/*
 * Common code that does the actual sending of the message by advancing the
 * local message queue's Put value and sends an IPI to the partition the
 * message is being sent to.
 */
static enum xpc_retval
xpc_send_msg(struct xpc_channel *ch, struct xpc_msg *msg, u8 notify_type,
			xpc_notify_func func, void *key)
{
	enum xpc_retval ret = xpcSuccess;
1916
	struct xpc_notify *notify = notify;
1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942
	s64 put, msg_number = msg->number;


	DBUG_ON(notify_type == XPC_N_CALL && func == NULL);
	DBUG_ON((((u64) msg - (u64) ch->local_msgqueue) / ch->msg_size) !=
					msg_number % ch->local_nentries);
	DBUG_ON(msg->flags & XPC_M_READY);

	if (ch->flags & XPC_C_DISCONNECTING) {
		/* drop the reference grabbed in xpc_allocate_msg() */
		xpc_msgqueue_deref(ch);
		return ch->reason;
	}

	if (notify_type != 0) {
		/*
		 * Tell the remote side to send an ACK interrupt when the
		 * message has been delivered.
		 */
		msg->flags |= XPC_M_INTERRUPT;

		atomic_inc(&ch->n_to_notify);

		notify = &ch->notify_queue[msg_number % ch->local_nentries];
		notify->func = func;
		notify->key = key;
1943
		notify->type = notify_type;
1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360 2361 2362

		// >>> is a mb() needed here?

		if (ch->flags & XPC_C_DISCONNECTING) {
			/*
			 * An error occurred between our last error check and
			 * this one. We will try to clear the type field from
			 * the notify entry. If we succeed then
			 * xpc_disconnect_channel() didn't already process
			 * the notify entry.
			 */
			if (cmpxchg(&notify->type, notify_type, 0) ==
								notify_type) {
				atomic_dec(&ch->n_to_notify);
				ret = ch->reason;
			}

			/* drop the reference grabbed in xpc_allocate_msg() */
			xpc_msgqueue_deref(ch);
			return ret;
		}
	}

	msg->flags |= XPC_M_READY;

	/*
	 * The preceding store of msg->flags must occur before the following
	 * load of ch->local_GP->put.
	 */
	mb();

	/* see if the message is next in line to be sent, if so send it */

	put = ch->local_GP->put;
	if (put == msg_number) {
		xpc_send_msgs(ch, put);
	}

	/* drop the reference grabbed in xpc_allocate_msg() */
	xpc_msgqueue_deref(ch);
	return ret;
}


/*
 * Send a message previously allocated using xpc_initiate_allocate() on the
 * specified channel connected to the specified partition.
 *
 * This routine will not wait for the message to be received, nor will
 * notification be given when it does happen. Once this routine has returned
 * the message entry allocated via xpc_initiate_allocate() is no longer
 * accessable to the caller.
 *
 * This routine, although called by users, does not call xpc_part_ref() to
 * ensure that the partition infrastructure is in place. It relies on the
 * fact that we called xpc_msgqueue_ref() in xpc_allocate_msg().
 *
 * Arguments:
 *
 *	partid - ID of partition to which the channel is connected.
 *	ch_number - channel # to send message on.
 *	payload - pointer to the payload area allocated via
 *			xpc_initiate_allocate().
 */
enum xpc_retval
xpc_initiate_send(partid_t partid, int ch_number, void *payload)
{
	struct xpc_partition *part = &xpc_partitions[partid];
	struct xpc_msg *msg = XPC_MSG_ADDRESS(payload);
	enum xpc_retval ret;


	dev_dbg(xpc_chan, "msg=0x%p, partid=%d, channel=%d\n", (void *) msg,
		partid, ch_number);

	DBUG_ON(partid <= 0 || partid >= XP_MAX_PARTITIONS);
	DBUG_ON(ch_number < 0 || ch_number >= part->nchannels);
	DBUG_ON(msg == NULL);

	ret = xpc_send_msg(&part->channels[ch_number], msg, 0, NULL, NULL);

	return ret;
}


/*
 * Send a message previously allocated using xpc_initiate_allocate on the
 * specified channel connected to the specified partition.
 *
 * This routine will not wait for the message to be sent. Once this routine
 * has returned the message entry allocated via xpc_initiate_allocate() is no
 * longer accessable to the caller.
 *
 * Once the remote end of the channel has received the message, the function
 * passed as an argument to xpc_initiate_send_notify() will be called. This
 * allows the sender to free up or re-use any buffers referenced by the
 * message, but does NOT mean the message has been processed at the remote
 * end by a receiver.
 *
 * If this routine returns an error, the caller's function will NOT be called.
 *
 * This routine, although called by users, does not call xpc_part_ref() to
 * ensure that the partition infrastructure is in place. It relies on the
 * fact that we called xpc_msgqueue_ref() in xpc_allocate_msg().
 *
 * Arguments:
 *
 *	partid - ID of partition to which the channel is connected.
 *	ch_number - channel # to send message on.
 *	payload - pointer to the payload area allocated via
 *			xpc_initiate_allocate().
 *	func - function to call with asynchronous notification of message
 *		  receipt. THIS FUNCTION MUST BE NON-BLOCKING.
 *	key - user-defined key to be passed to the function when it's called.
 */
enum xpc_retval
xpc_initiate_send_notify(partid_t partid, int ch_number, void *payload,
				xpc_notify_func func, void *key)
{
	struct xpc_partition *part = &xpc_partitions[partid];
	struct xpc_msg *msg = XPC_MSG_ADDRESS(payload);
	enum xpc_retval ret;


	dev_dbg(xpc_chan, "msg=0x%p, partid=%d, channel=%d\n", (void *) msg,
		partid, ch_number);

	DBUG_ON(partid <= 0 || partid >= XP_MAX_PARTITIONS);
	DBUG_ON(ch_number < 0 || ch_number >= part->nchannels);
	DBUG_ON(msg == NULL);
	DBUG_ON(func == NULL);

	ret = xpc_send_msg(&part->channels[ch_number], msg, XPC_N_CALL,
								func, key);
	return ret;
}


static struct xpc_msg *
xpc_pull_remote_msg(struct xpc_channel *ch, s64 get)
{
	struct xpc_partition *part = &xpc_partitions[ch->partid];
	struct xpc_msg *remote_msg, *msg;
	u32 msg_index, nmsgs;
	u64 msg_offset;
	enum xpc_retval ret;


	if (down_interruptible(&ch->msg_to_pull_sema) != 0) {
		/* we were interrupted by a signal */
		return NULL;
	}

	while (get >= ch->next_msg_to_pull) {

		/* pull as many messages as are ready and able to be pulled */

		msg_index = ch->next_msg_to_pull % ch->remote_nentries;

		DBUG_ON(ch->next_msg_to_pull >=
					(volatile s64) ch->w_remote_GP.put);
		nmsgs =  (volatile s64) ch->w_remote_GP.put -
						ch->next_msg_to_pull;
		if (msg_index + nmsgs > ch->remote_nentries) {
			/* ignore the ones that wrap the msg queue for now */
			nmsgs = ch->remote_nentries - msg_index;
		}

		msg_offset = msg_index * ch->msg_size;
		msg = (struct xpc_msg *) ((u64) ch->remote_msgqueue +
								msg_offset);
		remote_msg = (struct xpc_msg *) (ch->remote_msgqueue_pa +
								msg_offset);

		if ((ret = xpc_pull_remote_cachelines(part, msg, remote_msg,
				nmsgs * ch->msg_size)) != xpcSuccess) {

			dev_dbg(xpc_chan, "failed to pull %d msgs starting with"
				" msg %ld from partition %d, channel=%d, "
				"ret=%d\n", nmsgs, ch->next_msg_to_pull,
				ch->partid, ch->number, ret);

			XPC_DEACTIVATE_PARTITION(part, ret);

			up(&ch->msg_to_pull_sema);
			return NULL;
		}

		mb();	/* >>> this may not be needed, we're not sure */

		ch->next_msg_to_pull += nmsgs;
	}

	up(&ch->msg_to_pull_sema);

	/* return the message we were looking for */
	msg_offset = (get % ch->remote_nentries) * ch->msg_size;
	msg = (struct xpc_msg *) ((u64) ch->remote_msgqueue + msg_offset);

	return msg;
}


/*
 * Get a message to be delivered.
 */
static struct xpc_msg *
xpc_get_deliverable_msg(struct xpc_channel *ch)
{
	struct xpc_msg *msg = NULL;
	s64 get;


	do {
		if ((volatile u32) ch->flags & XPC_C_DISCONNECTING) {
			break;
		}

		get = (volatile s64) ch->w_local_GP.get;
		if (get == (volatile s64) ch->w_remote_GP.put) {
			break;
		}

		/* There are messages waiting to be pulled and delivered.
		 * We need to try to secure one for ourselves. We'll do this
		 * by trying to increment w_local_GP.get and hope that no one
		 * else beats us to it. If they do, we'll we'll simply have
		 * to try again for the next one.
	 	 */

		if (cmpxchg(&ch->w_local_GP.get, get, get + 1) == get) {
			/* we got the entry referenced by get */

			dev_dbg(xpc_chan, "w_local_GP.get changed to %ld, "
				"partid=%d, channel=%d\n", get + 1,
				ch->partid, ch->number);

			/* pull the message from the remote partition */

			msg = xpc_pull_remote_msg(ch, get);

			DBUG_ON(msg != NULL && msg->number != get);
			DBUG_ON(msg != NULL && (msg->flags & XPC_M_DONE));
			DBUG_ON(msg != NULL && !(msg->flags & XPC_M_READY));

			break;
		}

	} while (1);

	return msg;
}


/*
 * Deliver a message to its intended recipient.
 */
void
xpc_deliver_msg(struct xpc_channel *ch)
{
	struct xpc_msg *msg;


	if ((msg = xpc_get_deliverable_msg(ch)) != NULL) {

		/*
		 * This ref is taken to protect the payload itself from being
		 * freed before the user is finished with it, which the user
		 * indicates by calling xpc_initiate_received().
		 */
		xpc_msgqueue_ref(ch);

		atomic_inc(&ch->kthreads_active);

		if (ch->func != NULL) {
			dev_dbg(xpc_chan, "ch->func() called, msg=0x%p, "
				"msg_number=%ld, partid=%d, channel=%d\n",
				(void *) msg, msg->number, ch->partid,
				ch->number);

			/* deliver the message to its intended recipient */
			ch->func(xpcMsgReceived, ch->partid, ch->number,
					&msg->payload, ch->key);

			dev_dbg(xpc_chan, "ch->func() returned, msg=0x%p, "
				"msg_number=%ld, partid=%d, channel=%d\n",
				(void *) msg, msg->number, ch->partid,
				ch->number);
		}

		atomic_dec(&ch->kthreads_active);
	}
}


/*
 * Now we actually acknowledge the messages that have been delivered and ack'd
 * by advancing the cached remote message queue's Get value and if requested
 * send an IPI to the message sender's partition.
 */
static void
xpc_acknowledge_msgs(struct xpc_channel *ch, s64 initial_get, u8 msg_flags)
{
	struct xpc_msg *msg;
	s64 get = initial_get + 1;
	int send_IPI = 0;


	while (1) {

		while (1) {
			if (get == (volatile s64) ch->w_local_GP.get) {
				break;
			}

			msg = (struct xpc_msg *) ((u64) ch->remote_msgqueue +
			       (get % ch->remote_nentries) * ch->msg_size);

			if (!(msg->flags & XPC_M_DONE)) {
				break;
			}

			msg_flags |= msg->flags;
			get++;
		}

		if (get == initial_get) {
			/* nothing's changed */
			break;
		}

		if (cmpxchg_rel(&ch->local_GP->get, initial_get, get) !=
								initial_get) {
			/* someone else beat us to it */
			DBUG_ON((volatile s64) ch->local_GP->get <=
								initial_get);
			break;
		}

		/* we just set the new value of local_GP->get */

		dev_dbg(xpc_chan, "local_GP->get changed to %ld, partid=%d, "
			"channel=%d\n", get, ch->partid, ch->number);

		send_IPI = (msg_flags & XPC_M_INTERRUPT);

		/*
		 * We need to ensure that the message referenced by
		 * local_GP->get is not XPC_M_DONE or that local_GP->get
		 * equals w_local_GP.get, so we'll go have a look.
		 */
		initial_get = get;
	}

	if (send_IPI) {
		xpc_IPI_send_msgrequest(ch);
	}
}


/*
 * Acknowledge receipt of a delivered message.
 *
 * If a message has XPC_M_INTERRUPT set, send an interrupt to the partition
 * that sent the message.
 *
 * This function, although called by users, does not call xpc_part_ref() to
 * ensure that the partition infrastructure is in place. It relies on the
 * fact that we called xpc_msgqueue_ref() in xpc_deliver_msg().
 *
 * Arguments:
 *
 *	partid - ID of partition to which the channel is connected.
 *	ch_number - channel # message received on.
 *	payload - pointer to the payload area allocated via
 *			xpc_initiate_allocate().
 */
void
xpc_initiate_received(partid_t partid, int ch_number, void *payload)
{
	struct xpc_partition *part = &xpc_partitions[partid];
	struct xpc_channel *ch;
	struct xpc_msg *msg = XPC_MSG_ADDRESS(payload);
	s64 get, msg_number = msg->number;


	DBUG_ON(partid <= 0 || partid >= XP_MAX_PARTITIONS);
	DBUG_ON(ch_number < 0 || ch_number >= part->nchannels);

	ch = &part->channels[ch_number];

	dev_dbg(xpc_chan, "msg=0x%p, msg_number=%ld, partid=%d, channel=%d\n",
		(void *) msg, msg_number, ch->partid, ch->number);

	DBUG_ON((((u64) msg - (u64) ch->remote_msgqueue) / ch->msg_size) !=
					msg_number % ch->remote_nentries);
	DBUG_ON(msg->flags & XPC_M_DONE);

	msg->flags |= XPC_M_DONE;

	/*
	 * The preceding store of msg->flags must occur before the following
	 * load of ch->local_GP->get.
	 */
	mb();

	/*
	 * See if this message is next in line to be acknowledged as having
	 * been delivered.
	 */
	get = ch->local_GP->get;
	if (get == msg_number) {
		xpc_acknowledge_msgs(ch, get, msg->flags);
	}

	/* the call to xpc_msgqueue_ref() was done by xpc_deliver_msg()  */
	xpc_msgqueue_deref(ch);
}