drivers/misc/sgi-xp/xpc_main.c

   1 /*
   2  * This file is subject to the terms and conditions of the GNU General Public
   3  * License.  See the file "COPYING" in the main directory of this archive
   4  * for more details.
   5  *
   6  * Copyright (c) 2004-2008 Silicon Graphics, Inc.  All Rights Reserved.
   7  */
   8
   9 /*
  10  * Cross Partition Communication (XPC) support - standard version.
  11  *
  12  *      XPC provides a message passing capability that crosses partition
  13  *      boundaries. This module is made up of two parts:
  14  *
  15  *          partition   This part detects the presence/absence of other
  16  *                      partitions. It provides a heartbeat and monitors
  17  *                      the heartbeats of other partitions.
  18  *
  19  *          channel     This part manages the channels and sends/receives
  20  *                      messages across them to/from other partitions.
  21  *
  22  *      There are a couple of additional functions residing in XP, which
  23  *      provide an interface to XPC for its users.
  24  *
  25  *
  26  *      Caveats:
  27  *
  28  *        . We currently have no way to determine which nasid an IPI came
  29  *          from. Thus, xpc_IPI_send() does a remote AMO write followed by
  30  *          an IPI. The AMO indicates where data is to be pulled from, so
  31  *          after the IPI arrives, the remote partition checks the AMO word.
  32  *          The IPI can actually arrive before the AMO however, so other code
  33  *          must periodically check for this case. Also, remote AMO operations
  34  *          do not reliably time out. Thus we do a remote PIO read solely to
  35  *          know whether the remote partition is down and whether we should
  36  *          stop sending IPIs to it. This remote PIO read operation is set up
  37  *          in a special nofault region so SAL knows to ignore (and cleanup)
  38  *          any errors due to the remote AMO write, PIO read, and/or PIO
  39  *          write operations.
  40  *
  41  *          If/when new hardware solves this IPI problem, we should abandon
  42  *          the current approach.
  43  *
  44  */
  45
  46 #include <linux/kernel.h>
  47 #include <linux/module.h>
  48 #include <linux/init.h>
  49 #include <linux/cache.h>
  50 #include <linux/interrupt.h>
  51 #include <linux/delay.h>
  52 #include <linux/reboot.h>
  53 #include <linux/completion.h>
  54 #include <linux/kdebug.h>
  55 #include <linux/kthread.h>
  56 #include <linux/uaccess.h>
  57 #include <asm/sn/intr.h>
  58 #include <asm/sn/sn_sal.h>
  59 #include "xpc.h"
  60
  61 /* define two XPC debug device structures to be used with dev_dbg() et al */
  62
  63 struct device_driver xpc_dbg_name = {
  64         .name = "xpc"
  65 };
  66
  67 struct device xpc_part_dbg_subname = {
  68         .bus_id = {0},          /* set to "part" at xpc_init() time */
  69         .driver = &xpc_dbg_name
  70 };
  71
  72 struct device xpc_chan_dbg_subname = {
  73         .bus_id = {0},          /* set to "chan" at xpc_init() time */
  74         .driver = &xpc_dbg_name
  75 };
  76
  77 struct device *xpc_part = &xpc_part_dbg_subname;
  78 struct device *xpc_chan = &xpc_chan_dbg_subname;
  79
  80 static int xpc_kdebug_ignore;
  81
  82 /* systune related variables for /proc/sys directories */
  83
  84 static int xpc_hb_interval = XPC_HB_DEFAULT_INTERVAL;
  85 static int xpc_hb_min_interval = 1;
  86 static int xpc_hb_max_interval = 10;
  87
  88 static int xpc_hb_check_interval = XPC_HB_CHECK_DEFAULT_INTERVAL;
  89 static int xpc_hb_check_min_interval = 10;
  90 static int xpc_hb_check_max_interval = 120;
  91
  92 int xpc_disengage_request_timelimit = XPC_DISENGAGE_REQUEST_DEFAULT_TIMELIMIT;
  93 static int xpc_disengage_request_min_timelimit; /* = 0 */
  94 static int xpc_disengage_request_max_timelimit = 120;
  95
  96 static ctl_table xpc_sys_xpc_hb_dir[] = {
  97         {
  98          .ctl_name = CTL_UNNUMBERED,
  99          .procname = "hb_interval",
 100          .data = &xpc_hb_interval,
 101          .maxlen = sizeof(int),
 102          .mode = 0644,
 103          .proc_handler = &proc_dointvec_minmax,
 104          .strategy = &sysctl_intvec,
 105          .extra1 = &xpc_hb_min_interval,
 106          .extra2 = &xpc_hb_max_interval},
 107         {
 108          .ctl_name = CTL_UNNUMBERED,
 109          .procname = "hb_check_interval",
 110          .data = &xpc_hb_check_interval,
 111          .maxlen = sizeof(int),
 112          .mode = 0644,
 113          .proc_handler = &proc_dointvec_minmax,
 114          .strategy = &sysctl_intvec,
 115          .extra1 = &xpc_hb_check_min_interval,
 116          .extra2 = &xpc_hb_check_max_interval},
 117         {}
 118 };
 119 static ctl_table xpc_sys_xpc_dir[] = {
 120         {
 121          .ctl_name = CTL_UNNUMBERED,
 122          .procname = "hb",
 123          .mode = 0555,
 124          .child = xpc_sys_xpc_hb_dir},
 125         {
 126          .ctl_name = CTL_UNNUMBERED,
 127          .procname = "disengage_request_timelimit",
 128          .data = &xpc_disengage_request_timelimit,
 129          .maxlen = sizeof(int),
 130          .mode = 0644,
 131          .proc_handler = &proc_dointvec_minmax,
 132          .strategy = &sysctl_intvec,
 133          .extra1 = &xpc_disengage_request_min_timelimit,
 134          .extra2 = &xpc_disengage_request_max_timelimit},
 135         {}
 136 };
 137 static ctl_table xpc_sys_dir[] = {
 138         {
 139          .ctl_name = CTL_UNNUMBERED,
 140          .procname = "xpc",
 141          .mode = 0555,
 142          .child = xpc_sys_xpc_dir},
 143         {}
 144 };
 145 static struct ctl_table_header *xpc_sysctl;
 146
 147 /* non-zero if any remote partition disengage request was timed out */
 148 int xpc_disengage_request_timedout;
 149
 150 /* #of IRQs received */
 151 static atomic_t xpc_act_IRQ_rcvd;
 152
 153 /* IRQ handler notifies this wait queue on receipt of an IRQ */
 154 static DECLARE_WAIT_QUEUE_HEAD(xpc_act_IRQ_wq);
 155
 156 static unsigned long xpc_hb_check_timeout;
 157
 158 /* notification that the xpc_hb_checker thread has exited */
 159 static DECLARE_COMPLETION(xpc_hb_checker_exited);
 160
 161 /* notification that the xpc_discovery thread has exited */
 162 static DECLARE_COMPLETION(xpc_discovery_exited);
 163
 164 static struct timer_list xpc_hb_timer;
 165
 166 static void xpc_kthread_waitmsgs(struct xpc_partition *, struct xpc_channel *);
 167
 168 static int xpc_system_reboot(struct notifier_block *, unsigned long, void *);
 169 static struct notifier_block xpc_reboot_notifier = {
 170         .notifier_call = xpc_system_reboot,
 171 };
 172
 173 static int xpc_system_die(struct notifier_block *, unsigned long, void *);
 174 static struct notifier_block xpc_die_notifier = {
 175         .notifier_call = xpc_system_die,
 176 };
 177
 178 /*
 179  * Timer function to enforce the timelimit on the partition disengage request.
 180  */
 181 static void
 182 xpc_timeout_partition_disengage_request(unsigned long data)
 183 {
 184         struct xpc_partition *part = (struct xpc_partition *)data;
 185
 186         DBUG_ON(time_before(jiffies, part->disengage_request_timeout));
 187
 188         (void)xpc_partition_disengaged(part);
 189
 190         DBUG_ON(part->disengage_request_timeout != 0);
 191         DBUG_ON(xpc_partition_engaged(1UL << XPC_PARTID(part)) != 0);
 192 }
 193
 194 /*
 195  * Notify the heartbeat check thread that an IRQ has been received.
 196  */
 197 static irqreturn_t
 198 xpc_act_IRQ_handler(int irq, void *dev_id)
 199 {
 200         atomic_inc(&xpc_act_IRQ_rcvd);
 201         wake_up_interruptible(&xpc_act_IRQ_wq);
 202         return IRQ_HANDLED;
 203 }
 204
 205 /*
 206  * Timer to produce the heartbeat.  The timer structures function is
 207  * already set when this is initially called.  A tunable is used to
 208  * specify when the next timeout should occur.
 209  */
 210 static void
 211 xpc_hb_beater(unsigned long dummy)
 212 {
 213         xpc_vars->heartbeat++;
 214
 215         if (time_after_eq(jiffies, xpc_hb_check_timeout))
 216                 wake_up_interruptible(&xpc_act_IRQ_wq);
 217
 218         xpc_hb_timer.expires = jiffies + (xpc_hb_interval * HZ);
 219         add_timer(&xpc_hb_timer);
 220 }
 221
 222 /*
 223  * This thread is responsible for nearly all of the partition
 224  * activation/deactivation.
 225  */
 226 static int
 227 xpc_hb_checker(void *ignore)
 228 {
 229         int last_IRQ_count = 0;
 230         int new_IRQ_count;
 231         int force_IRQ = 0;
 232         cpumask_of_cpu_ptr(cpumask, XPC_HB_CHECK_CPU);
 233
 234         /* this thread was marked active by xpc_hb_init() */
 235
 236         set_cpus_allowed_ptr(current, cpumask);
 237
 238         /* set our heartbeating to other partitions into motion */
 239         xpc_hb_check_timeout = jiffies + (xpc_hb_check_interval * HZ);
 240         xpc_hb_beater(0);
 241
 242         while (!xpc_exiting) {
 243
 244                 dev_dbg(xpc_part, "woke up with %d ticks rem; %d IRQs have "
 245                         "been received\n",
 246                         (int)(xpc_hb_check_timeout - jiffies),
 247                         atomic_read(&xpc_act_IRQ_rcvd) - last_IRQ_count);
 248
 249                 /* checking of remote heartbeats is skewed by IRQ handling */
 250                 if (time_after_eq(jiffies, xpc_hb_check_timeout)) {
 251                         dev_dbg(xpc_part, "checking remote heartbeats\n");
 252                         xpc_check_remote_hb();
 253
 254                         /*
 255                          * We need to periodically recheck to ensure no
 256                          * IPI/AMO pairs have been missed.  That check
 257                          * must always reset xpc_hb_check_timeout.
 258                          */
 259                         force_IRQ = 1;
 260                 }
 261
 262                 /* check for outstanding IRQs */
 263                 new_IRQ_count = atomic_read(&xpc_act_IRQ_rcvd);
 264                 if (last_IRQ_count < new_IRQ_count || force_IRQ != 0) {
 265                         force_IRQ = 0;
 266
 267                         dev_dbg(xpc_part, "found an IRQ to process; will be "
 268                                 "resetting xpc_hb_check_timeout\n");
 269
 270                         last_IRQ_count += xpc_identify_act_IRQ_sender();
 271                         if (last_IRQ_count < new_IRQ_count) {
 272                                 /* retry once to help avoid missing AMO */
 273                                 (void)xpc_identify_act_IRQ_sender();
 274                         }
 275                         last_IRQ_count = new_IRQ_count;
 276
 277                         xpc_hb_check_timeout = jiffies +
 278                             (xpc_hb_check_interval * HZ);
 279                 }
 280
 281                 /* wait for IRQ or timeout */
 282                 (void)wait_event_interruptible(xpc_act_IRQ_wq,
 283                                                (last_IRQ_count <
 284                                                 atomic_read(&xpc_act_IRQ_rcvd)
 285                                                 || time_after_eq(jiffies,
 286                                                         xpc_hb_check_timeout) ||
 287                                                 xpc_exiting));
 288         }
 289
 290         dev_dbg(xpc_part, "heartbeat checker is exiting\n");
 291
 292         /* mark this thread as having exited */
 293         complete(&xpc_hb_checker_exited);
 294         return 0;
 295 }
 296
 297 /*
 298  * This thread will attempt to discover other partitions to activate
 299  * based on info provided by SAL. This new thread is short lived and
 300  * will exit once discovery is complete.
 301  */
 302 static int
 303 xpc_initiate_discovery(void *ignore)
 304 {
 305         xpc_discovery();
 306
 307         dev_dbg(xpc_part, "discovery thread is exiting\n");
 308
 309         /* mark this thread as having exited */
 310         complete(&xpc_discovery_exited);
 311         return 0;
 312 }
 313
 314 /*
 315  * Establish first contact with the remote partititon. This involves pulling
 316  * the XPC per partition variables from the remote partition and waiting for
 317  * the remote partition to pull ours.
 318  */
 319 static enum xp_retval
 320 xpc_make_first_contact(struct xpc_partition *part)
 321 {
 322         enum xp_retval ret;
 323
 324         while ((ret = xpc_pull_remote_vars_part(part)) != xpSuccess) {
 325                 if (ret != xpRetry) {
 326                         XPC_DEACTIVATE_PARTITION(part, ret);
 327                         return ret;
 328                 }
 329
 330                 dev_dbg(xpc_chan, "waiting to make first contact with "
 331                         "partition %d\n", XPC_PARTID(part));
 332
 333                 /* wait a 1/4 of a second or so */
 334                 (void)msleep_interruptible(250);
 335
 336                 if (part->act_state == XPC_P_DEACTIVATING)
 337                         return part->reason;
 338         }
 339
 340         return xpc_mark_partition_active(part);
 341 }
 342
 343 /*
 344  * The first kthread assigned to a newly activated partition is the one
 345  * created by XPC HB with which it calls xpc_partition_up(). XPC hangs on to
 346  * that kthread until the partition is brought down, at which time that kthread
 347  * returns back to XPC HB. (The return of that kthread will signify to XPC HB
 348  * that XPC has dismantled all communication infrastructure for the associated
 349  * partition.) This kthread becomes the channel manager for that partition.
 350  *
 351  * Each active partition has a channel manager, who, besides connecting and
 352  * disconnecting channels, will ensure that each of the partition's connected
 353  * channels has the required number of assigned kthreads to get the work done.
 354  */
 355 static void
 356 xpc_channel_mgr(struct xpc_partition *part)
 357 {
 358         while (part->act_state != XPC_P_DEACTIVATING ||
 359                atomic_read(&part->nchannels_active) > 0 ||
 360                !xpc_partition_disengaged(part)) {
 361
 362                 xpc_process_channel_activity(part);
 363
 364                 /*
 365                  * Wait until we've been requested to activate kthreads or
 366                  * all of the channel's message queues have been torn down or
 367                  * a signal is pending.
 368                  *
 369                  * The channel_mgr_requests is set to 1 after being awakened,
 370                  * This is done to prevent the channel mgr from making one pass
 371                  * through the loop for each request, since he will
 372                  * be servicing all the requests in one pass. The reason it's
 373                  * set to 1 instead of 0 is so that other kthreads will know
 374                  * that the channel mgr is running and won't bother trying to
 375                  * wake him up.
 376                  */
 377                 atomic_dec(&part->channel_mgr_requests);
 378                 (void)wait_event_interruptible(part->channel_mgr_wq,
 379                                 (atomic_read(&part->channel_mgr_requests) > 0 ||
 380                                  part->local_IPI_amo != 0 ||
 381                                  (part->act_state == XPC_P_DEACTIVATING &&
 382                                  atomic_read(&part->nchannels_active) == 0 &&
 383                                  xpc_partition_disengaged(part))));
 384                 atomic_set(&part->channel_mgr_requests, 1);
 385         }
 386 }
 387
 388 /*
 389  * When XPC HB determines that a partition has come up, it will create a new
 390  * kthread and that kthread will call this function to attempt to set up the
 391  * basic infrastructure used for Cross Partition Communication with the newly
 392  * upped partition.
 393  *
 394  * The kthread that was created by XPC HB and which setup the XPC
 395  * infrastructure will remain assigned to the partition until the partition
 396  * goes down. At which time the kthread will teardown the XPC infrastructure
 397  * and then exit.
 398  *
 399  * XPC HB will put the remote partition's XPC per partition specific variables
 400  * physical address into xpc_partitions[partid].remote_vars_part_pa prior to
 401  * calling xpc_partition_up().
 402  */
 403 static void
 404 xpc_partition_up(struct xpc_partition *part)
 405 {
 406         DBUG_ON(part->channels != NULL);
 407
 408         dev_dbg(xpc_chan, "activating partition %d\n", XPC_PARTID(part));
 409
 410         if (xpc_setup_infrastructure(part) != xpSuccess)
 411                 return;
 412
 413         /*
 414          * The kthread that XPC HB called us with will become the
 415          * channel manager for this partition. It will not return
 416          * back to XPC HB until the partition's XPC infrastructure
 417          * has been dismantled.
 418          */
 419
 420         (void)xpc_part_ref(part);       /* this will always succeed */
 421
 422         if (xpc_make_first_contact(part) == xpSuccess)
 423                 xpc_channel_mgr(part);
 424
 425         xpc_part_deref(part);
 426
 427         xpc_teardown_infrastructure(part);
 428 }
 429
 430 static int
 431 xpc_activating(void *__partid)
 432 {
 433         short partid = (u64)__partid;
 434         struct xpc_partition *part = &xpc_partitions[partid];
 435         unsigned long irq_flags;
 436
 437         DBUG_ON(partid <= 0 || partid >= XP_MAX_PARTITIONS);
 438
 439         spin_lock_irqsave(&part->act_lock, irq_flags);
 440
 441         if (part->act_state == XPC_P_DEACTIVATING) {
 442                 part->act_state = XPC_P_INACTIVE;
 443                 spin_unlock_irqrestore(&part->act_lock, irq_flags);
 444                 part->remote_rp_pa = 0;
 445                 return 0;
 446         }
 447
 448         /* indicate the thread is activating */
 449         DBUG_ON(part->act_state != XPC_P_ACTIVATION_REQ);
 450         part->act_state = XPC_P_ACTIVATING;
 451
 452         XPC_SET_REASON(part, 0, 0);
 453         spin_unlock_irqrestore(&part->act_lock, irq_flags);
 454
 455         dev_dbg(xpc_part, "bringing partition %d up\n", partid);
 456
 457         /*
 458          * Register the remote partition's AMOs with SAL so it can handle
 459          * and cleanup errors within that address range should the remote
 460          * partition go down. We don't unregister this range because it is
 461          * difficult to tell when outstanding writes to the remote partition
 462          * are finished and thus when it is safe to unregister. This should
 463          * not result in wasted space in the SAL xp_addr_region table because
 464          * we should get the same page for remote_amos_page_pa after module
 465          * reloads and system reboots.
 466          */
 467         if (sn_register_xp_addr_region(part->remote_amos_page_pa,
 468                                        PAGE_SIZE, 1) < 0) {
 469                 dev_warn(xpc_part, "xpc_partition_up(%d) failed to register "
 470                          "xp_addr region\n", partid);
 471
 472                 spin_lock_irqsave(&part->act_lock, irq_flags);
 473                 part->act_state = XPC_P_INACTIVE;
 474                 XPC_SET_REASON(part, xpPhysAddrRegFailed, __LINE__);
 475                 spin_unlock_irqrestore(&part->act_lock, irq_flags);
 476                 part->remote_rp_pa = 0;
 477                 return 0;
 478         }
 479
 480         xpc_allow_hb(partid, xpc_vars);
 481         xpc_IPI_send_activated(part);
 482
 483         /*
 484          * xpc_partition_up() holds this thread and marks this partition as
 485          * XPC_P_ACTIVE by calling xpc_hb_mark_active().
 486          */
 487         (void)xpc_partition_up(part);
 488
 489         xpc_disallow_hb(partid, xpc_vars);
 490         xpc_mark_partition_inactive(part);
 491
 492         if (part->reason == xpReactivating) {
 493                 /* interrupting ourselves results in activating partition */
 494                 xpc_IPI_send_reactivate(part);
 495         }
 496
 497         return 0;
 498 }
 499
 500 void
 501 xpc_activate_partition(struct xpc_partition *part)
 502 {
 503         short partid = XPC_PARTID(part);
 504         unsigned long irq_flags;
 505         struct task_struct *kthread;
 506
 507         spin_lock_irqsave(&part->act_lock, irq_flags);
 508
 509         DBUG_ON(part->act_state != XPC_P_INACTIVE);
 510
 511         part->act_state = XPC_P_ACTIVATION_REQ;
 512         XPC_SET_REASON(part, xpCloneKThread, __LINE__);
 513
 514         spin_unlock_irqrestore(&part->act_lock, irq_flags);
 515
 516         kthread = kthread_run(xpc_activating, (void *)((u64)partid), "xpc%02d",
 517                               partid);
 518         if (IS_ERR(kthread)) {
 519                 spin_lock_irqsave(&part->act_lock, irq_flags);
 520                 part->act_state = XPC_P_INACTIVE;
 521                 XPC_SET_REASON(part, xpCloneKThreadFailed, __LINE__);
 522                 spin_unlock_irqrestore(&part->act_lock, irq_flags);
 523         }
 524 }
 525
 526 /*
 527  * Handle the receipt of a SGI_XPC_NOTIFY IRQ by seeing whether the specified
 528  * partition actually sent it. Since SGI_XPC_NOTIFY IRQs may be shared by more
 529  * than one partition, we use an AMO_t structure per partition to indicate
 530  * whether a partition has sent an IPI or not.  If it has, then wake up the
 531  * associated kthread to handle it.
 532  *
 533  * All SGI_XPC_NOTIFY IRQs received by XPC are the result of IPIs sent by XPC
 534  * running on other partitions.
 535  *
 536  * Noteworthy Arguments:
 537  *
 538  *      irq - Interrupt ReQuest number. NOT USED.
 539  *
 540  *      dev_id - partid of IPI's potential sender.
 541  */
 542 irqreturn_t
 543 xpc_notify_IRQ_handler(int irq, void *dev_id)
 544 {
 545         short partid = (short)(u64)dev_id;
 546         struct xpc_partition *part = &xpc_partitions[partid];
 547
 548         DBUG_ON(partid <= 0 || partid >= XP_MAX_PARTITIONS);
 549
 550         if (xpc_part_ref(part)) {
 551                 xpc_check_for_channel_activity(part);
 552
 553                 xpc_part_deref(part);
 554         }
 555         return IRQ_HANDLED;
 556 }
 557
 558 /*
 559  * Check to see if xpc_notify_IRQ_handler() dropped any IPIs on the floor
 560  * because the write to their associated IPI amo completed after the IRQ/IPI
 561  * was received.
 562  */
 563 void
 564 xpc_dropped_IPI_check(struct xpc_partition *part)
 565 {
 566         if (xpc_part_ref(part)) {
 567                 xpc_check_for_channel_activity(part);
 568
 569                 part->dropped_IPI_timer.expires = jiffies +
 570                     XPC_P_DROPPED_IPI_WAIT;
 571                 add_timer(&part->dropped_IPI_timer);
 572                 xpc_part_deref(part);
 573         }
 574 }
 575
 576 void
 577 xpc_activate_kthreads(struct xpc_channel *ch, int needed)
 578 {
 579         int idle = atomic_read(&ch->kthreads_idle);
 580         int assigned = atomic_read(&ch->kthreads_assigned);
 581         int wakeup;
 582
 583         DBUG_ON(needed <= 0);
 584
 585         if (idle > 0) {
 586                 wakeup = (needed > idle) ? idle : needed;
 587                 needed -= wakeup;
 588
 589                 dev_dbg(xpc_chan, "wakeup %d idle kthreads, partid=%d, "
 590                         "channel=%d\n", wakeup, ch->partid, ch->number);
 591
 592                 /* only wakeup the requested number of kthreads */
 593                 wake_up_nr(&ch->idle_wq, wakeup);
 594         }
 595
 596         if (needed <= 0)
 597                 return;
 598
 599         if (needed + assigned > ch->kthreads_assigned_limit) {
 600                 needed = ch->kthreads_assigned_limit - assigned;
 601                 if (needed <= 0)
 602                         return;
 603         }
 604
 605         dev_dbg(xpc_chan, "create %d new kthreads, partid=%d, channel=%d\n",
 606                 needed, ch->partid, ch->number);
 607
 608         xpc_create_kthreads(ch, needed, 0);
 609 }
 610
 611 /*
 612  * This function is where XPC's kthreads wait for messages to deliver.
 613  */
 614 static void
 615 xpc_kthread_waitmsgs(struct xpc_partition *part, struct xpc_channel *ch)
 616 {
 617         do {
 618                 /* deliver messages to their intended recipients */
 619
 620                 while (ch->w_local_GP.get < ch->w_remote_GP.put &&
 621                        !(ch->flags & XPC_C_DISCONNECTING)) {
 622                         xpc_deliver_msg(ch);
 623                 }
 624
 625                 if (atomic_inc_return(&ch->kthreads_idle) >
 626                     ch->kthreads_idle_limit) {
 627                         /* too many idle kthreads on this channel */
 628                         atomic_dec(&ch->kthreads_idle);
 629                         break;
 630                 }
 631
 632                 dev_dbg(xpc_chan, "idle kthread calling "
 633                         "wait_event_interruptible_exclusive()\n");
 634
 635                 (void)wait_event_interruptible_exclusive(ch->idle_wq,
 636                                 (ch->w_local_GP.get < ch->w_remote_GP.put ||
 637                                  (ch->flags & XPC_C_DISCONNECTING)));
 638
 639                 atomic_dec(&ch->kthreads_idle);
 640
 641         } while (!(ch->flags & XPC_C_DISCONNECTING));
 642 }
 643
 644 static int
 645 xpc_kthread_start(void *args)
 646 {
 647         short partid = XPC_UNPACK_ARG1(args);
 648         u16 ch_number = XPC_UNPACK_ARG2(args);
 649         struct xpc_partition *part = &xpc_partitions[partid];
 650         struct xpc_channel *ch;
 651         int n_needed;
 652         unsigned long irq_flags;
 653
 654         dev_dbg(xpc_chan, "kthread starting, partid=%d, channel=%d\n",
 655                 partid, ch_number);
 656
 657         ch = &part->channels[ch_number];
 658
 659         if (!(ch->flags & XPC_C_DISCONNECTING)) {
 660
 661                 /* let registerer know that connection has been established */
 662
 663                 spin_lock_irqsave(&ch->lock, irq_flags);
 664                 if (!(ch->flags & XPC_C_CONNECTEDCALLOUT)) {
 665                         ch->flags |= XPC_C_CONNECTEDCALLOUT;
 666                         spin_unlock_irqrestore(&ch->lock, irq_flags);
 667
 668                         xpc_connected_callout(ch);
 669
 670                         spin_lock_irqsave(&ch->lock, irq_flags);
 671                         ch->flags |= XPC_C_CONNECTEDCALLOUT_MADE;
 672                         spin_unlock_irqrestore(&ch->lock, irq_flags);
 673
 674                         /*
 675                          * It is possible that while the callout was being
 676                          * made that the remote partition sent some messages.
 677                          * If that is the case, we may need to activate
 678                          * additional kthreads to help deliver them. We only
 679                          * need one less than total #of messages to deliver.
 680                          */
 681                         n_needed = ch->w_remote_GP.put - ch->w_local_GP.get - 1;
 682                         if (n_needed > 0 && !(ch->flags & XPC_C_DISCONNECTING))
 683                                 xpc_activate_kthreads(ch, n_needed);
 684
 685                 } else {
 686                         spin_unlock_irqrestore(&ch->lock, irq_flags);
 687                 }
 688
 689                 xpc_kthread_waitmsgs(part, ch);
 690         }
 691
 692         /* let registerer know that connection is disconnecting */
 693
 694         spin_lock_irqsave(&ch->lock, irq_flags);
 695         if ((ch->flags & XPC_C_CONNECTEDCALLOUT_MADE) &&
 696             !(ch->flags & XPC_C_DISCONNECTINGCALLOUT)) {
 697                 ch->flags |= XPC_C_DISCONNECTINGCALLOUT;
 698                 spin_unlock_irqrestore(&ch->lock, irq_flags);
 699
 700                 xpc_disconnect_callout(ch, xpDisconnecting);
 701
 702                 spin_lock_irqsave(&ch->lock, irq_flags);
 703                 ch->flags |= XPC_C_DISCONNECTINGCALLOUT_MADE;
 704         }
 705         spin_unlock_irqrestore(&ch->lock, irq_flags);
 706
 707         if (atomic_dec_return(&ch->kthreads_assigned) == 0) {
 708                 if (atomic_dec_return(&part->nchannels_engaged) == 0) {
 709                         xpc_mark_partition_disengaged(part);
 710                         xpc_IPI_send_disengage(part);
 711                 }
 712         }
 713
 714         xpc_msgqueue_deref(ch);
 715
 716         dev_dbg(xpc_chan, "kthread exiting, partid=%d, channel=%d\n",
 717                 partid, ch_number);
 718
 719         xpc_part_deref(part);
 720         return 0;
 721 }
 722
 723 /*
 724  * For each partition that XPC has established communications with, there is
 725  * a minimum of one kernel thread assigned to perform any operation that
 726  * may potentially sleep or block (basically the callouts to the asynchronous
 727  * functions registered via xpc_connect()).
 728  *
 729  * Additional kthreads are created and destroyed by XPC as the workload
 730  * demands.
 731  *
 732  * A kthread is assigned to one of the active channels that exists for a given
 733  * partition.
 734  */
 735 void
 736 xpc_create_kthreads(struct xpc_channel *ch, int needed,
 737                     int ignore_disconnecting)
 738 {
 739         unsigned long irq_flags;
 740         u64 args = XPC_PACK_ARGS(ch->partid, ch->number);
 741         struct xpc_partition *part = &xpc_partitions[ch->partid];
 742         struct task_struct *kthread;
 743
 744         while (needed-- > 0) {
 745
 746                 /*
 747                  * The following is done on behalf of the newly created
 748                  * kthread. That kthread is responsible for doing the
 749                  * counterpart to the following before it exits.
 750                  */
 751                 if (ignore_disconnecting) {
 752                         if (!atomic_inc_not_zero(&ch->kthreads_assigned)) {
 753                                 /* kthreads assigned had gone to zero */
 754                                 BUG_ON(!(ch->flags &
 755                                          XPC_C_DISCONNECTINGCALLOUT_MADE));
 756                                 break;
 757                         }
 758
 759                 } else if (ch->flags & XPC_C_DISCONNECTING) {
 760                         break;
 761
 762                 } else if (atomic_inc_return(&ch->kthreads_assigned) == 1) {
 763                         if (atomic_inc_return(&part->nchannels_engaged) == 1)
 764                                 xpc_mark_partition_engaged(part);
 765                 }
 766                 (void)xpc_part_ref(part);
 767                 xpc_msgqueue_ref(ch);
 768
 769                 kthread = kthread_run(xpc_kthread_start, (void *)args,
 770                                       "xpc%02dc%d", ch->partid, ch->number);
 771                 if (IS_ERR(kthread)) {
 772                         /* the fork failed */
 773
 774                         /*
 775                          * NOTE: if (ignore_disconnecting &&
 776                          * !(ch->flags & XPC_C_DISCONNECTINGCALLOUT)) is true,
 777                          * then we'll deadlock if all other kthreads assigned
 778                          * to this channel are blocked in the channel's
 779                          * registerer, because the only thing that will unblock
 780                          * them is the xpDisconnecting callout that this
 781                          * failed kthread_run() would have made.
 782                          */
 783
 784                         if (atomic_dec_return(&ch->kthreads_assigned) == 0 &&
 785                             atomic_dec_return(&part->nchannels_engaged) == 0) {
 786                                 xpc_mark_partition_disengaged(part);
 787                                 xpc_IPI_send_disengage(part);
 788                         }
 789                         xpc_msgqueue_deref(ch);
 790                         xpc_part_deref(part);
 791
 792                         if (atomic_read(&ch->kthreads_assigned) <
 793                             ch->kthreads_idle_limit) {
 794                                 /*
 795                                  * Flag this as an error only if we have an
 796                                  * insufficient #of kthreads for the channel
 797                                  * to function.
 798                                  */
 799                                 spin_lock_irqsave(&ch->lock, irq_flags);
 800                                 XPC_DISCONNECT_CHANNEL(ch, xpLackOfResources,
 801                                                        &irq_flags);
 802                                 spin_unlock_irqrestore(&ch->lock, irq_flags);
 803                         }
 804                         break;
 805                 }
 806         }
 807 }
 808
 809 void
 810 xpc_disconnect_wait(int ch_number)
 811 {
 812         unsigned long irq_flags;
 813         short partid;
 814         struct xpc_partition *part;
 815         struct xpc_channel *ch;
 816         int wakeup_channel_mgr;
 817
 818         /* now wait for all callouts to the caller's function to cease */
 819         for (partid = 1; partid < XP_MAX_PARTITIONS; partid++) {
 820                 part = &xpc_partitions[partid];
 821
 822                 if (!xpc_part_ref(part))
 823                         continue;
 824
 825                 ch = &part->channels[ch_number];
 826
 827                 if (!(ch->flags & XPC_C_WDISCONNECT)) {
 828                         xpc_part_deref(part);
 829                         continue;
 830                 }
 831
 832                 wait_for_completion(&ch->wdisconnect_wait);
 833
 834                 spin_lock_irqsave(&ch->lock, irq_flags);
 835                 DBUG_ON(!(ch->flags & XPC_C_DISCONNECTED));
 836                 wakeup_channel_mgr = 0;
 837
 838                 if (ch->delayed_IPI_flags) {
 839                         if (part->act_state != XPC_P_DEACTIVATING) {
 840                                 spin_lock(&part->IPI_lock);
 841                                 XPC_SET_IPI_FLAGS(part->local_IPI_amo,
 842                                                   ch->number,
 843                                                   ch->delayed_IPI_flags);
 844                                 spin_unlock(&part->IPI_lock);
 845                                 wakeup_channel_mgr = 1;
 846                         }
 847                         ch->delayed_IPI_flags = 0;
 848                 }
 849
 850                 ch->flags &= ~XPC_C_WDISCONNECT;
 851                 spin_unlock_irqrestore(&ch->lock, irq_flags);
 852
 853                 if (wakeup_channel_mgr)
 854                         xpc_wakeup_channel_mgr(part);
 855
 856                 xpc_part_deref(part);
 857         }
 858 }
 859
 860 static void
 861 xpc_do_exit(enum xp_retval reason)
 862 {
 863         short partid;
 864         int active_part_count, printed_waiting_msg = 0;
 865         struct xpc_partition *part;
 866         unsigned long printmsg_time, disengage_request_timeout = 0;
 867
 868         /* a 'rmmod XPC' and a 'reboot' cannot both end up here together */
 869         DBUG_ON(xpc_exiting == 1);
 870
 871         /*
 872          * Let the heartbeat checker thread and the discovery thread
 873          * (if one is running) know that they should exit. Also wake up
 874          * the heartbeat checker thread in case it's sleeping.
 875          */
 876         xpc_exiting = 1;
 877         wake_up_interruptible(&xpc_act_IRQ_wq);
 878
 879         /* ignore all incoming interrupts */
 880         free_irq(SGI_XPC_ACTIVATE, NULL);
 881
 882         /* wait for the discovery thread to exit */
 883         wait_for_completion(&xpc_discovery_exited);
 884
 885         /* wait for the heartbeat checker thread to exit */
 886         wait_for_completion(&xpc_hb_checker_exited);
 887
 888         /* sleep for a 1/3 of a second or so */
 889         (void)msleep_interruptible(300);
 890
 891         /* wait for all partitions to become inactive */
 892
 893         printmsg_time = jiffies + (XPC_DISENGAGE_PRINTMSG_INTERVAL * HZ);
 894         xpc_disengage_request_timedout = 0;
 895
 896         do {
 897                 active_part_count = 0;
 898
 899                 for (partid = 1; partid < XP_MAX_PARTITIONS; partid++) {
 900                         part = &xpc_partitions[partid];
 901
 902                         if (xpc_partition_disengaged(part) &&
 903                             part->act_state == XPC_P_INACTIVE) {
 904                                 continue;
 905                         }
 906
 907                         active_part_count++;
 908
 909                         XPC_DEACTIVATE_PARTITION(part, reason);
 910
 911                         if (part->disengage_request_timeout >
 912                             disengage_request_timeout) {
 913                                 disengage_request_timeout =
 914                                     part->disengage_request_timeout;
 915                         }
 916                 }
 917
 918                 if (xpc_partition_engaged(-1UL)) {
 919                         if (time_after(jiffies, printmsg_time)) {
 920                                 dev_info(xpc_part, "waiting for remote "
 921                                          "partitions to disengage, timeout in "
 922                                          "%ld seconds\n",
 923                                          (disengage_request_timeout - jiffies)
 924                                          / HZ);
 925                                 printmsg_time = jiffies +
 926                                     (XPC_DISENGAGE_PRINTMSG_INTERVAL * HZ);
 927                                 printed_waiting_msg = 1;
 928                         }
 929
 930                 } else if (active_part_count > 0) {
 931                         if (printed_waiting_msg) {
 932                                 dev_info(xpc_part, "waiting for local partition"
 933                                          " to disengage\n");
 934                                 printed_waiting_msg = 0;
 935                         }
 936
 937                 } else {
 938                         if (!xpc_disengage_request_timedout) {
 939                                 dev_info(xpc_part, "all partitions have "
 940                                          "disengaged\n");
 941                         }
 942                         break;
 943                 }
 944
 945                 /* sleep for a 1/3 of a second or so */
 946                 (void)msleep_interruptible(300);
 947
 948         } while (1);
 949
 950         DBUG_ON(xpc_partition_engaged(-1UL));
 951
 952         /* indicate to others that our reserved page is uninitialized */
 953         xpc_rsvd_page->vars_pa = 0;
 954
 955         /* now it's time to eliminate our heartbeat */
 956         del_timer_sync(&xpc_hb_timer);
 957         DBUG_ON(xpc_vars->heartbeating_to_mask != 0);
 958
 959         if (reason == xpUnloading) {
 960                 /* take ourselves off of the reboot_notifier_list */
 961                 (void)unregister_reboot_notifier(&xpc_reboot_notifier);
 962
 963                 /* take ourselves off of the die_notifier list */
 964                 (void)unregister_die_notifier(&xpc_die_notifier);
 965         }
 966
 967         /* close down protections for IPI operations */
 968         xpc_restrict_IPI_ops();
 969
 970         /* clear the interface to XPC's functions */
 971         xpc_clear_interface();
 972
 973         if (xpc_sysctl)
 974                 unregister_sysctl_table(xpc_sysctl);
 975
 976         kfree(xpc_remote_copy_buffer_base);
 977 }
 978
 979 /*
 980  * This function is called when the system is being rebooted.
 981  */
 982 static int
 983 xpc_system_reboot(struct notifier_block *nb, unsigned long event, void *unused)
 984 {
 985         enum xp_retval reason;
 986
 987         switch (event) {
 988         case SYS_RESTART:
 989                 reason = xpSystemReboot;
 990                 break;
 991         case SYS_HALT:
 992                 reason = xpSystemHalt;
 993                 break;
 994         case SYS_POWER_OFF:
 995                 reason = xpSystemPoweroff;
 996                 break;
 997         default:
 998                 reason = xpSystemGoingDown;
 999         }
1000
1001         xpc_do_exit(reason);
1002         return NOTIFY_DONE;
1003 }
1004
1005 /*
1006  * Notify other partitions to disengage from all references to our memory.
1007  */
1008 static void
1009 xpc_die_disengage(void)
1010 {
1011         struct xpc_partition *part;
1012         short partid;
1013         unsigned long engaged;
1014         long time, printmsg_time, disengage_request_timeout;
1015
1016         /* keep xpc_hb_checker thread from doing anything (just in case) */
1017         xpc_exiting = 1;
1018
1019         xpc_vars->heartbeating_to_mask = 0;     /* indicate we're deactivated */
1020
1021         for (partid = 1; partid < XP_MAX_PARTITIONS; partid++) {
1022                 part = &xpc_partitions[partid];
1023
1024                 if (!XPC_SUPPORTS_DISENGAGE_REQUEST(part->
1025                     remote_vars_version)) {
1026
1027                         /* just in case it was left set by an earlier XPC */
1028                         xpc_clear_partition_engaged(1UL << partid);
1029                         continue;
1030                 }
1031
1032                 if (xpc_partition_engaged(1UL << partid) ||
1033                     part->act_state != XPC_P_INACTIVE) {
1034                         xpc_request_partition_disengage(part);
1035                         xpc_mark_partition_disengaged(part);
1036                         xpc_IPI_send_disengage(part);
1037                 }
1038         }
1039
1040         time = rtc_time();
1041         printmsg_time = time +
1042             (XPC_DISENGAGE_PRINTMSG_INTERVAL * sn_rtc_cycles_per_second);
1043         disengage_request_timeout = time +
1044             (xpc_disengage_request_timelimit * sn_rtc_cycles_per_second);
1045
1046         /* wait for all other partitions to disengage from us */
1047
1048         while (1) {
1049                 engaged = xpc_partition_engaged(-1UL);
1050                 if (!engaged) {
1051                         dev_info(xpc_part, "all partitions have disengaged\n");
1052                         break;
1053                 }
1054
1055                 time = rtc_time();
1056                 if (time >= disengage_request_timeout) {
1057                         for (partid = 1; partid < XP_MAX_PARTITIONS; partid++) {
1058                                 if (engaged & (1UL << partid)) {
1059                                         dev_info(xpc_part, "disengage from "
1060                                                  "remote partition %d timed "
1061                                                  "out\n", partid);
1062                                 }
1063                         }
1064                         break;
1065                 }
1066
1067                 if (time >= printmsg_time) {
1068                         dev_info(xpc_part, "waiting for remote partitions to "
1069                                  "disengage, timeout in %ld seconds\n",
1070                                  (disengage_request_timeout - time) /
1071                                  sn_rtc_cycles_per_second);
1072                         printmsg_time = time +
1073                             (XPC_DISENGAGE_PRINTMSG_INTERVAL *
1074                              sn_rtc_cycles_per_second);
1075                 }
1076         }
1077 }
1078
1079 /*
1080  * This function is called when the system is being restarted or halted due
1081  * to some sort of system failure. If this is the case we need to notify the
1082  * other partitions to disengage from all references to our memory.
1083  * This function can also be called when our heartbeater could be offlined
1084  * for a time. In this case we need to notify other partitions to not worry
1085  * about the lack of a heartbeat.
1086  */
1087 static int
1088 xpc_system_die(struct notifier_block *nb, unsigned long event, void *unused)
1089 {
1090         switch (event) {
1091         case DIE_MACHINE_RESTART:
1092         case DIE_MACHINE_HALT:
1093                 xpc_die_disengage();
1094                 break;
1095
1096         case DIE_KDEBUG_ENTER:
1097                 /* Should lack of heartbeat be ignored by other partitions? */
1098                 if (!xpc_kdebug_ignore)
1099                         break;
1100
1101                 /* fall through */
1102         case DIE_MCA_MONARCH_ENTER:
1103         case DIE_INIT_MONARCH_ENTER:
1104                 xpc_vars->heartbeat++;
1105                 xpc_vars->heartbeat_offline = 1;
1106                 break;
1107
1108         case DIE_KDEBUG_LEAVE:
1109                 /* Is lack of heartbeat being ignored by other partitions? */
1110                 if (!xpc_kdebug_ignore)
1111                         break;
1112
1113                 /* fall through */
1114         case DIE_MCA_MONARCH_LEAVE:
1115         case DIE_INIT_MONARCH_LEAVE:
1116                 xpc_vars->heartbeat++;
1117                 xpc_vars->heartbeat_offline = 0;
1118                 break;
1119         }
1120
1121         return NOTIFY_DONE;
1122 }
1123
1124 int __init
1125 xpc_init(void)
1126 {
1127         int ret;
1128         short partid;
1129         struct xpc_partition *part;
1130         struct task_struct *kthread;
1131         size_t buf_size;
1132
1133         if (!ia64_platform_is("sn2"))
1134                 return -ENODEV;
1135
1136         buf_size = max(XPC_RP_VARS_SIZE,
1137                        XPC_RP_HEADER_SIZE + XP_NASID_MASK_BYTES);
1138         xpc_remote_copy_buffer = xpc_kmalloc_cacheline_aligned(buf_size,
1139                                                                GFP_KERNEL,
1140                                                   &xpc_remote_copy_buffer_base);
1141         if (xpc_remote_copy_buffer == NULL)
1142                 return -ENOMEM;
1143
1144         snprintf(xpc_part->bus_id, BUS_ID_SIZE, "part");
1145         snprintf(xpc_chan->bus_id, BUS_ID_SIZE, "chan");
1146
1147         xpc_sysctl = register_sysctl_table(xpc_sys_dir);
1148
1149         /*
1150          * The first few fields of each entry of xpc_partitions[] need to
1151          * be initialized now so that calls to xpc_connect() and
1152          * xpc_disconnect() can be made prior to the activation of any remote
1153          * partition. NOTE THAT NONE OF THE OTHER FIELDS BELONGING TO THESE
1154          * ENTRIES ARE MEANINGFUL UNTIL AFTER AN ENTRY'S CORRESPONDING
1155          * PARTITION HAS BEEN ACTIVATED.
1156          */
1157         for (partid = 1; partid < XP_MAX_PARTITIONS; partid++) {
1158                 part = &xpc_partitions[partid];
1159
1160                 DBUG_ON((u64)part != L1_CACHE_ALIGN((u64)part));
1161
1162                 part->act_IRQ_rcvd = 0;
1163                 spin_lock_init(&part->act_lock);
1164                 part->act_state = XPC_P_INACTIVE;
1165                 XPC_SET_REASON(part, 0, 0);
1166
1167                 init_timer(&part->disengage_request_timer);
1168                 part->disengage_request_timer.function =
1169                     xpc_timeout_partition_disengage_request;
1170                 part->disengage_request_timer.data = (unsigned long)part;
1171
1172                 part->setup_state = XPC_P_UNSET;
1173                 init_waitqueue_head(&part->teardown_wq);
1174                 atomic_set(&part->references, 0);
1175         }
1176
1177         /*
1178          * Open up protections for IPI operations (and AMO operations on
1179          * Shub 1.1 systems).
1180          */
1181         xpc_allow_IPI_ops();
1182
1183         /*
1184          * Interrupts being processed will increment this atomic variable and
1185          * awaken the heartbeat thread which will process the interrupts.
1186          */
1187         atomic_set(&xpc_act_IRQ_rcvd, 0);
1188
1189         /*
1190          * This is safe to do before the xpc_hb_checker thread has started
1191          * because the handler releases a wait queue.  If an interrupt is
1192          * received before the thread is waiting, it will not go to sleep,
1193          * but rather immediately process the interrupt.
1194          */
1195         ret = request_irq(SGI_XPC_ACTIVATE, xpc_act_IRQ_handler, 0,
1196                           "xpc hb", NULL);
1197         if (ret != 0) {
1198                 dev_err(xpc_part, "can't register ACTIVATE IRQ handler, "
1199                         "errno=%d\n", -ret);
1200
1201                 xpc_restrict_IPI_ops();
1202
1203                 if (xpc_sysctl)
1204                         unregister_sysctl_table(xpc_sysctl);
1205
1206                 kfree(xpc_remote_copy_buffer_base);
1207                 return -EBUSY;
1208         }
1209
1210         /*
1211          * Fill the partition reserved page with the information needed by
1212          * other partitions to discover we are alive and establish initial
1213          * communications.
1214          */
1215         xpc_rsvd_page = xpc_rsvd_page_init();
1216         if (xpc_rsvd_page == NULL) {
1217                 dev_err(xpc_part, "could not setup our reserved page\n");
1218
1219                 free_irq(SGI_XPC_ACTIVATE, NULL);
1220                 xpc_restrict_IPI_ops();
1221
1222                 if (xpc_sysctl)
1223                         unregister_sysctl_table(xpc_sysctl);
1224
1225                 kfree(xpc_remote_copy_buffer_base);
1226                 return -EBUSY;
1227         }
1228
1229         /* add ourselves to the reboot_notifier_list */
1230         ret = register_reboot_notifier(&xpc_reboot_notifier);
1231         if (ret != 0)
1232                 dev_warn(xpc_part, "can't register reboot notifier\n");
1233
1234         /* add ourselves to the die_notifier list */
1235         ret = register_die_notifier(&xpc_die_notifier);
1236         if (ret != 0)
1237                 dev_warn(xpc_part, "can't register die notifier\n");
1238
1239         init_timer(&xpc_hb_timer);
1240         xpc_hb_timer.function = xpc_hb_beater;
1241
1242         /*
1243          * The real work-horse behind xpc.  This processes incoming
1244          * interrupts and monitors remote heartbeats.
1245          */
1246         kthread = kthread_run(xpc_hb_checker, NULL, XPC_HB_CHECK_THREAD_NAME);
1247         if (IS_ERR(kthread)) {
1248                 dev_err(xpc_part, "failed while forking hb check thread\n");
1249
1250                 /* indicate to others that our reserved page is uninitialized */
1251                 xpc_rsvd_page->vars_pa = 0;
1252
1253                 /* take ourselves off of the reboot_notifier_list */
1254                 (void)unregister_reboot_notifier(&xpc_reboot_notifier);
1255
1256                 /* take ourselves off of the die_notifier list */
1257                 (void)unregister_die_notifier(&xpc_die_notifier);
1258
1259                 del_timer_sync(&xpc_hb_timer);
1260                 free_irq(SGI_XPC_ACTIVATE, NULL);
1261                 xpc_restrict_IPI_ops();
1262
1263                 if (xpc_sysctl)
1264                         unregister_sysctl_table(xpc_sysctl);
1265
1266                 kfree(xpc_remote_copy_buffer_base);
1267                 return -EBUSY;
1268         }
1269
1270         /*
1271          * Startup a thread that will attempt to discover other partitions to
1272          * activate based on info provided by SAL. This new thread is short
1273          * lived and will exit once discovery is complete.
1274          */
1275         kthread = kthread_run(xpc_initiate_discovery, NULL,
1276                               XPC_DISCOVERY_THREAD_NAME);
1277         if (IS_ERR(kthread)) {
1278                 dev_err(xpc_part, "failed while forking discovery thread\n");
1279
1280                 /* mark this new thread as a non-starter */
1281                 complete(&xpc_discovery_exited);
1282
1283                 xpc_do_exit(xpUnloading);
1284                 return -EBUSY;
1285         }
1286
1287         /* set the interface to point at XPC's functions */
1288         xpc_set_interface(xpc_initiate_connect, xpc_initiate_disconnect,
1289                           xpc_initiate_allocate, xpc_initiate_send,
1290                           xpc_initiate_send_notify, xpc_initiate_received,
1291                           xpc_initiate_partid_to_nasids);
1292
1293         return 0;
1294 }
1295
1296 module_init(xpc_init);
1297
1298 void __exit
1299 xpc_exit(void)
1300 {
1301         xpc_do_exit(xpUnloading);
1302 }
1303
1304 module_exit(xpc_exit);
1305
1306 MODULE_AUTHOR("Silicon Graphics, Inc.");
1307 MODULE_DESCRIPTION("Cross Partition Communication (XPC) support");
1308 MODULE_LICENSE("GPL");
1309
1310 module_param(xpc_hb_interval, int, 0);
1311 MODULE_PARM_DESC(xpc_hb_interval, "Number of seconds between "
1312                  "heartbeat increments.");
1313
1314 module_param(xpc_hb_check_interval, int, 0);
1315 MODULE_PARM_DESC(xpc_hb_check_interval, "Number of seconds between "
1316                  "heartbeat checks.");
1317
1318 module_param(xpc_disengage_request_timelimit, int, 0);
1319 MODULE_PARM_DESC(xpc_disengage_request_timelimit, "Number of seconds to wait "
1320                  "for disengage request to complete.");
1321
1322 module_param(xpc_kdebug_ignore, int, 0);
1323 MODULE_PARM_DESC(xpc_kdebug_ignore, "Should lack of heartbeat be ignored by "
1324                  "other partitions when dropping into kdebug.");