summaryrefslogtreecommitdiffstats
path: root/arch/ia64/sn/kernel/xpc_main.c
blob: b617236524c690ea2eb95122e26b9ad133d30819 (plain)
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
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
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
170
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
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
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
360
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
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
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
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
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
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
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
1035
1036
1037
1038
1039
1040
1041
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
1095
1096
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
1269
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
1297
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
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
/*
 * 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) support - standard version.
 *
 *	XPC provides a message passing capability that crosses partition
 *	boundaries. This module is made up of two parts:
 *
 *	    partition	This part detects the presence/absence of other
 *			partitions. It provides a heartbeat and monitors
 *			the heartbeats of other partitions.
 *
 *	    channel	This part manages the channels and sends/receives
 *			messages across them to/from other partitions.
 *
 *	There are a couple of additional functions residing in XP, which
 *	provide an interface to XPC for its users.
 *
 *
 *	Caveats:
 *
 *	  . We currently have no way to determine which nasid an IPI came
 *	    from. Thus, xpc_IPI_send() does a remote AMO write followed by
 *	    an IPI. The AMO indicates where data is to be pulled from, so
 *	    after the IPI arrives, the remote partition checks the AMO word.
 *	    The IPI can actually arrive before the AMO however, so other code
 *	    must periodically check for this case. Also, remote AMO operations
 *	    do not reliably time out. Thus we do a remote PIO read solely to
 *	    know whether the remote partition is down and whether we should
 *	    stop sending IPIs to it. This remote PIO read operation is set up
 *	    in a special nofault region so SAL knows to ignore (and cleanup)
 *	    any errors due to the remote AMO write, PIO read, and/or PIO
 *	    write operations.
 *
 *	    If/when new hardware solves this IPI problem, we should abandon
 *	    the current approach.
 *
 */


#include <linux/kernel.h>
#include <linux/module.h>
#include <linux/init.h>
#include <linux/sched.h>
#include <linux/syscalls.h>
#include <linux/cache.h>
#include <linux/interrupt.h>
#include <linux/slab.h>
#include <linux/delay.h>
#include <linux/reboot.h>
#include <asm/sn/intr.h>
#include <asm/sn/sn_sal.h>
#include <asm/kdebug.h>
#include <asm/uaccess.h>
#include "xpc.h"


/* define two XPC debug device structures to be used with dev_dbg() et al */

struct device_driver xpc_dbg_name = {
	.name = "xpc"
};

struct device xpc_part_dbg_subname = {
	.bus_id = {0},		/* set to "part" at xpc_init() time */
	.driver = &xpc_dbg_name
};

struct device xpc_chan_dbg_subname = {
	.bus_id = {0},		/* set to "chan" at xpc_init() time */
	.driver = &xpc_dbg_name
};

struct device *xpc_part = &xpc_part_dbg_subname;
struct device *xpc_chan = &xpc_chan_dbg_subname;


/* systune related variables for /proc/sys directories */

static int xpc_hb_interval = XPC_HB_DEFAULT_INTERVAL;
static int xpc_hb_min_interval = 1;
static int xpc_hb_max_interval = 10;

static int xpc_hb_check_interval = XPC_HB_CHECK_DEFAULT_INTERVAL;
static int xpc_hb_check_min_interval = 10;
static int xpc_hb_check_max_interval = 120;

int xpc_disengage_request_timelimit = XPC_DISENGAGE_REQUEST_DEFAULT_TIMELIMIT;
static int xpc_disengage_request_min_timelimit = 0;
static int xpc_disengage_request_max_timelimit = 120;

static ctl_table xpc_sys_xpc_hb_dir[] = {
	{
		1,
		"hb_interval",
		&xpc_hb_interval,
		sizeof(int),
		0644,
		NULL,
		&proc_dointvec_minmax,
		&sysctl_intvec,
		NULL,
		&xpc_hb_min_interval,
		&xpc_hb_max_interval
	},
	{
		2,
		"hb_check_interval",
		&xpc_hb_check_interval,
		sizeof(int),
		0644,
		NULL,
		&proc_dointvec_minmax,
		&sysctl_intvec,
		NULL,
		&xpc_hb_check_min_interval,
		&xpc_hb_check_max_interval
	},
	{0}
};
static ctl_table xpc_sys_xpc_dir[] = {
	{
		1,
		"hb",
		NULL,
		0,
		0555,
		xpc_sys_xpc_hb_dir
	},
	{
		2,
		"disengage_request_timelimit",
		&xpc_disengage_request_timelimit,
		sizeof(int),
		0644,
		NULL,
		&proc_dointvec_minmax,
		&sysctl_intvec,
		NULL,
		&xpc_disengage_request_min_timelimit,
		&xpc_disengage_request_max_timelimit
	},
	{0}
};
static ctl_table xpc_sys_dir[] = {
	{
		1,
		"xpc",
		NULL,
		0,
		0555,
		xpc_sys_xpc_dir
	},
	{0}
};
static struct ctl_table_header *xpc_sysctl;


/* #of IRQs received */
static atomic_t xpc_act_IRQ_rcvd;

/* IRQ handler notifies this wait queue on receipt of an IRQ */
static DECLARE_WAIT_QUEUE_HEAD(xpc_act_IRQ_wq);

static unsigned long xpc_hb_check_timeout;

/* notification that the xpc_hb_checker thread has exited */
static DECLARE_MUTEX_LOCKED(xpc_hb_checker_exited);

/* notification that the xpc_discovery thread has exited */
static DECLARE_MUTEX_LOCKED(xpc_discovery_exited);


static struct timer_list xpc_hb_timer;


static void xpc_kthread_waitmsgs(struct xpc_partition *, struct xpc_channel *);


static int xpc_system_reboot(struct notifier_block *, unsigned long, void *);
static struct notifier_block xpc_reboot_notifier = {
	.notifier_call = xpc_system_reboot,
};

static int xpc_system_die(struct notifier_block *, unsigned long, void *);
static struct notifier_block xpc_die_notifier = {
	.notifier_call = xpc_system_die,
};


/*
 * Timer function to enforce the timelimit on the partition disengage request.
 */
static void
xpc_timeout_partition_disengage_request(unsigned long data)
{
	struct xpc_partition *part = (struct xpc_partition *) data;


	DBUG_ON(jiffies < part->disengage_request_timeout);

	(void) xpc_partition_disengaged(part);

	DBUG_ON(part->disengage_request_timeout != 0);
	DBUG_ON(xpc_partition_engaged(1UL << XPC_PARTID(part)) != 0);
}


/*
 * Notify the heartbeat check thread that an IRQ has been received.
 */
static irqreturn_t
xpc_act_IRQ_handler(int irq, void *dev_id, struct pt_regs *regs)
{
	atomic_inc(&xpc_act_IRQ_rcvd);
	wake_up_interruptible(&xpc_act_IRQ_wq);
	return IRQ_HANDLED;
}


/*
 * Timer to produce the heartbeat.  The timer structures function is
 * already set when this is initially called.  A tunable is used to
 * specify when the next timeout should occur.
 */
static void
xpc_hb_beater(unsigned long dummy)
{
	xpc_vars->heartbeat++;

	if (jiffies >= xpc_hb_check_timeout) {
		wake_up_interruptible(&xpc_act_IRQ_wq);
	}

	xpc_hb_timer.expires = jiffies + (xpc_hb_interval * HZ);
	add_timer(&xpc_hb_timer);
}


/*
 * This thread is responsible for nearly all of the partition
 * activation/deactivation.
 */
static int
xpc_hb_checker(void *ignore)
{
	int last_IRQ_count = 0;
	int new_IRQ_count;
	int force_IRQ=0;


	/* this thread was marked active by xpc_hb_init() */

	daemonize(XPC_HB_CHECK_THREAD_NAME);

	set_cpus_allowed(current, cpumask_of_cpu(XPC_HB_CHECK_CPU));

	xpc_hb_check_timeout = jiffies + (xpc_hb_check_interval * HZ);

	while (!(volatile int) xpc_exiting) {

		dev_dbg(xpc_part, "woke up with %d ticks rem; %d IRQs have "
			"been received\n",
			(int) (xpc_hb_check_timeout - jiffies),
			atomic_read(&xpc_act_IRQ_rcvd) - last_IRQ_count);


		/* checking of remote heartbeats is skewed by IRQ handling */
		if (jiffies >= xpc_hb_check_timeout) {
			dev_dbg(xpc_part, "checking remote heartbeats\n");
			xpc_check_remote_hb();

			/*
			 * We need to periodically recheck to ensure no
			 * IPI/AMO pairs have been missed.  That check
			 * must always reset xpc_hb_check_timeout.
			 */
			force_IRQ = 1;
		}


		/* check for outstanding IRQs */
		new_IRQ_count = atomic_read(&xpc_act_IRQ_rcvd);
		if (last_IRQ_count < new_IRQ_count || force_IRQ != 0) {
			force_IRQ = 0;

			dev_dbg(xpc_part, "found an IRQ to process; will be "
				"resetting xpc_hb_check_timeout\n");

			last_IRQ_count += xpc_identify_act_IRQ_sender();
			if (last_IRQ_count < new_IRQ_count) {
				/* retry once to help avoid missing AMO */
				(void) xpc_identify_act_IRQ_sender();
			}
			last_IRQ_count = new_IRQ_count;

			xpc_hb_check_timeout = jiffies +
					   (xpc_hb_check_interval * HZ);
		}

		/* wait for IRQ or timeout */
		(void) wait_event_interruptible(xpc_act_IRQ_wq,
			    (last_IRQ_count < atomic_read(&xpc_act_IRQ_rcvd) ||
					jiffies >= xpc_hb_check_timeout ||
						(volatile int) xpc_exiting));
	}

	dev_dbg(xpc_part, "heartbeat checker is exiting\n");


	/* mark this thread as having exited */
	up(&xpc_hb_checker_exited);
	return 0;
}


/*
 * This thread will attempt to discover other partitions to activate
 * based on info provided by SAL. This new thread is short lived and
 * will exit once discovery is complete.
 */
static int
xpc_initiate_discovery(void *ignore)
{
	daemonize(XPC_DISCOVERY_THREAD_NAME);

	xpc_discovery();

	dev_dbg(xpc_part, "discovery thread is exiting\n");

	/* mark this thread as having exited */
	up(&xpc_discovery_exited);
	return 0;
}


/*
 * Establish first contact with the remote partititon. This involves pulling
 * the XPC per partition variables from the remote partition and waiting for
 * the remote partition to pull ours.
 */
static enum xpc_retval
xpc_make_first_contact(struct xpc_partition *part)
{
	enum xpc_retval ret;


	while ((ret = xpc_pull_remote_vars_part(part)) != xpcSuccess) {
		if (ret != xpcRetry) {
			XPC_DEACTIVATE_PARTITION(part, ret);
			return ret;
		}

		dev_dbg(xpc_chan, "waiting to make first contact with "
			"partition %d\n", XPC_PARTID(part));

		/* wait a 1/4 of a second or so */
		(void) msleep_interruptible(250);

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

	return xpc_mark_partition_active(part);
}


/*
 * The first kthread assigned to a newly activated partition is the one
 * created by XPC HB with which it calls xpc_partition_up(). XPC hangs on to
 * that kthread until the partition is brought down, at which time that kthread
 * returns back to XPC HB. (The return of that kthread will signify to XPC HB
 * that XPC has dismantled all communication infrastructure for the associated
 * partition.) This kthread becomes the channel manager for that partition.
 *
 * Each active partition has a channel manager, who, besides connecting and
 * disconnecting channels, will ensure that each of the partition's connected
 * channels has the required number of assigned kthreads to get the work done.
 */
static void
xpc_channel_mgr(struct xpc_partition *part)
{
	while (part->act_state != XPC_P_DEACTIVATING ||
			atomic_read(&part->nchannels_active) > 0 ||
					!xpc_partition_disengaged(part)) {

		xpc_process_channel_activity(part);


		/*
		 * Wait until we've been requested to activate kthreads or
		 * all of the channel's message queues have been torn down or
		 * a signal is pending.
		 *
		 * The channel_mgr_requests is set to 1 after being awakened,
		 * This is done to prevent the channel mgr from making one pass
		 * through the loop for each request, since he will
		 * be servicing all the requests in one pass. The reason it's
		 * set to 1 instead of 0 is so that other kthreads will know
		 * that the channel mgr is running and won't bother trying to
		 * wake him up.
		 */
		atomic_dec(&part->channel_mgr_requests);
		(void) wait_event_interruptible(part->channel_mgr_wq,
				(atomic_read(&part->channel_mgr_requests) > 0 ||
				(volatile u64) part->local_IPI_amo != 0 ||
				((volatile u8) part->act_state ==
							XPC_P_DEACTIVATING &&
				atomic_read(&part->nchannels_active) == 0 &&
				xpc_partition_disengaged(part))));
		atomic_set(&part->channel_mgr_requests, 1);

		// >>> Does it need to wakeup periodically as well? In case we
		// >>> miscalculated the #of kthreads to wakeup or create?
	}
}


/*
 * When XPC HB determines that a partition has come up, it will create a new
 * kthread and that kthread will call this function to attempt to set up the
 * basic infrastructure used for Cross Partition Communication with the newly
 * upped partition.
 *
 * The kthread that was created by XPC HB and which setup the XPC
 * infrastructure will remain assigned to the partition until the partition
 * goes down. At which time the kthread will teardown the XPC infrastructure
 * and then exit.
 *
 * XPC HB will put the remote partition's XPC per partition specific variables
 * physical address into xpc_partitions[partid].remote_vars_part_pa prior to
 * calling xpc_partition_up().
 */
static void
xpc_partition_up(struct xpc_partition *part)
{
	DBUG_ON(part->channels != NULL);

	dev_dbg(xpc_chan, "activating partition %d\n", XPC_PARTID(part));

	if (xpc_setup_infrastructure(part) != xpcSuccess) {
		return;
	}

	/*
	 * The kthread that XPC HB called us with will become the
	 * channel manager for this partition. It will not return
	 * back to XPC HB until the partition's XPC infrastructure
	 * has been dismantled.
	 */

	(void) xpc_part_ref(part);	/* this will always succeed */

	if (xpc_make_first_contact(part) == xpcSuccess) {
		xpc_channel_mgr(part);
	}

	xpc_part_deref(part);

	xpc_teardown_infrastructure(part);
}


static int
xpc_activating(void *__partid)
{
	partid_t partid = (u64) __partid;
	struct xpc_partition *part = &xpc_partitions[partid];
	unsigned long irq_flags;
	struct sched_param param = { sched_priority: MAX_RT_PRIO - 1 };
	int ret;


	DBUG_ON(partid <= 0 || partid >= XP_MAX_PARTITIONS);

	spin_lock_irqsave(&part->act_lock, irq_flags);

	if (part->act_state == XPC_P_DEACTIVATING) {
		part->act_state = XPC_P_INACTIVE;
		spin_unlock_irqrestore(&part->act_lock, irq_flags);
		part->remote_rp_pa = 0;
		return 0;
	}

	/* indicate the thread is activating */
	DBUG_ON(part->act_state != XPC_P_ACTIVATION_REQ);
	part->act_state = XPC_P_ACTIVATING;

	XPC_SET_REASON(part, 0, 0);
	spin_unlock_irqrestore(&part->act_lock, irq_flags);

	dev_dbg(xpc_part, "bringing partition %d up\n", partid);

	daemonize("xpc%02d", partid);

	/*
	 * This thread needs to run at a realtime priority to prevent a
	 * significant performance degradation.
	 */
	ret = sched_setscheduler(current, SCHED_FIFO, &param);
	if (ret != 0) {
		dev_warn(xpc_part, "unable to set pid %d to a realtime "
			"priority, ret=%d\n", current->pid, ret);
	}

	/* allow this thread and its children to run on any CPU */
	set_cpus_allowed(current, CPU_MASK_ALL);

	/*
	 * Register the remote partition's AMOs with SAL so it can handle
	 * and cleanup errors within that address range should the remote
	 * partition go down. We don't unregister this range because it is
	 * difficult to tell when outstanding writes to the remote partition
	 * are finished and thus when it is safe to unregister. This should
	 * not result in wasted space in the SAL xp_addr_region table because
	 * we should get the same page for remote_amos_page_pa after module
	 * reloads and system reboots.
	 */
	if (sn_register_xp_addr_region(part->remote_amos_page_pa,
							PAGE_SIZE, 1) < 0) {
		dev_warn(xpc_part, "xpc_partition_up(%d) failed to register "
			"xp_addr region\n", partid);

		spin_lock_irqsave(&part->act_lock, irq_flags);
		part->act_state = XPC_P_INACTIVE;
		XPC_SET_REASON(part, xpcPhysAddrRegFailed, __LINE__);
		spin_unlock_irqrestore(&part->act_lock, irq_flags);
		part->remote_rp_pa = 0;
		return 0;
	}

	xpc_allow_hb(partid, xpc_vars);
	xpc_IPI_send_activated(part);


	/*
	 * xpc_partition_up() holds this thread and marks this partition as
	 * XPC_P_ACTIVE by calling xpc_hb_mark_active().
	 */
	(void) xpc_partition_up(part);

	xpc_disallow_hb(partid, xpc_vars);
	xpc_mark_partition_inactive(part);

	if (part->reason == xpcReactivating) {
		/* interrupting ourselves results in activating partition */
		xpc_IPI_send_reactivate(part);
	}

	return 0;
}


void
xpc_activate_partition(struct xpc_partition *part)
{
	partid_t partid = XPC_PARTID(part);
	unsigned long irq_flags;
	pid_t pid;


	spin_lock_irqsave(&part->act_lock, irq_flags);

	pid = kernel_thread(xpc_activating, (void *) ((u64) partid), 0);

	DBUG_ON(part->act_state != XPC_P_INACTIVE);

	if (pid > 0) {
		part->act_state = XPC_P_ACTIVATION_REQ;
		XPC_SET_REASON(part, xpcCloneKThread, __LINE__);
	} else {
		XPC_SET_REASON(part, xpcCloneKThreadFailed, __LINE__);
	}

	spin_unlock_irqrestore(&part->act_lock, irq_flags);
}


/*
 * Handle the receipt of a SGI_XPC_NOTIFY IRQ by seeing whether the specified
 * partition actually sent it. Since SGI_XPC_NOTIFY IRQs may be shared by more
 * than one partition, we use an AMO_t structure per partition to indicate
 * whether a partition has sent an IPI or not.  >>> If it has, then wake up the
 * associated kthread to handle it.
 *
 * All SGI_XPC_NOTIFY IRQs received by XPC are the result of IPIs sent by XPC
 * running on other partitions.
 *
 * Noteworthy Arguments:
 *
 *	irq - Interrupt ReQuest number. NOT USED.
 *
 *	dev_id - partid of IPI's potential sender.
 *
 *	regs - processor's context before the processor entered
 *	       interrupt code. NOT USED.
 */
irqreturn_t
xpc_notify_IRQ_handler(int irq, void *dev_id, struct pt_regs *regs)
{
	partid_t partid = (partid_t) (u64) dev_id;
	struct xpc_partition *part = &xpc_partitions[partid];


	DBUG_ON(partid <= 0 || partid >= XP_MAX_PARTITIONS);

	if (xpc_part_ref(part)) {
		xpc_check_for_channel_activity(part);

		xpc_part_deref(part);
	}
	return IRQ_HANDLED;
}


/*
 * Check to see if xpc_notify_IRQ_handler() dropped any IPIs on the floor
 * because the write to their associated IPI amo completed after the IRQ/IPI
 * was received.
 */
void
xpc_dropped_IPI_check(struct xpc_partition *part)
{
	if (xpc_part_ref(part)) {
		xpc_check_for_channel_activity(part);

		part->dropped_IPI_timer.expires = jiffies +
							XPC_P_DROPPED_IPI_WAIT;
		add_timer(&part->dropped_IPI_timer);
		xpc_part_deref(part);
	}
}


void
xpc_activate_kthreads(struct xpc_channel *ch, int needed)
{
	int idle = atomic_read(&ch->kthreads_idle);
	int assigned = atomic_read(&ch->kthreads_assigned);
	int wakeup;


	DBUG_ON(needed <= 0);

	if (idle > 0) {
		wakeup = (needed > idle) ? idle : needed;
		needed -= wakeup;

		dev_dbg(xpc_chan, "wakeup %d idle kthreads, partid=%d, "
			"channel=%d\n", wakeup, ch->partid, ch->number);

		/* only wakeup the requested number of kthreads */
		wake_up_nr(&ch->idle_wq, wakeup);
	}

	if (needed <= 0) {
		return;
	}

	if (needed + assigned > ch->kthreads_assigned_limit) {
		needed = ch->kthreads_assigned_limit - assigned;
		// >>>should never be less than 0
		if (needed <= 0) {
			return;
		}
	}

	dev_dbg(xpc_chan, "create %d new kthreads, partid=%d, channel=%d\n",
		needed, ch->partid, ch->number);

	xpc_create_kthreads(ch, needed);
}


/*
 * This function is where XPC's kthreads wait for messages to deliver.
 */
static void
xpc_kthread_waitmsgs(struct xpc_partition *part, struct xpc_channel *ch)
{
	do {
		/* deliver messages to their intended recipients */

		while ((volatile s64) ch->w_local_GP.get <
				(volatile s64) ch->w_remote_GP.put &&
					!((volatile u32) ch->flags &
						XPC_C_DISCONNECTING)) {
			xpc_deliver_msg(ch);
		}

		if (atomic_inc_return(&ch->kthreads_idle) >
						ch->kthreads_idle_limit) {
			/* too many idle kthreads on this channel */
			atomic_dec(&ch->kthreads_idle);
			break;
		}

		dev_dbg(xpc_chan, "idle kthread calling "
			"wait_event_interruptible_exclusive()\n");

		(void) wait_event_interruptible_exclusive(ch->idle_wq,
				((volatile s64) ch->w_local_GP.get <
					(volatile s64) ch->w_remote_GP.put ||
				((volatile u32) ch->flags &
						XPC_C_DISCONNECTING)));

		atomic_dec(&ch->kthreads_idle);

	} while (!((volatile u32) ch->flags & XPC_C_DISCONNECTING));
}


static int
xpc_daemonize_kthread(void *args)
{
	partid_t partid = XPC_UNPACK_ARG1(args);
	u16 ch_number = XPC_UNPACK_ARG2(args);
	struct xpc_partition *part = &xpc_partitions[partid];
	struct xpc_channel *ch;
	int n_needed;
	unsigned long irq_flags;


	daemonize("xpc%02dc%d", partid, ch_number);

	dev_dbg(xpc_chan, "kthread starting, partid=%d, channel=%d\n",
		partid, ch_number);

	ch = &part->channels[ch_number];

	if (!(ch->flags & XPC_C_DISCONNECTING)) {

		/* let registerer know that connection has been established */

		spin_lock_irqsave(&ch->lock, irq_flags);
		if (!(ch->flags & XPC_C_CONNECTCALLOUT)) {
			ch->flags |= XPC_C_CONNECTCALLOUT;
			spin_unlock_irqrestore(&ch->lock, irq_flags);

			xpc_connected_callout(ch);

			/*
			 * It is possible that while the callout was being
			 * made that the remote partition sent some messages.
			 * If that is the case, we may need to activate
			 * additional kthreads to help deliver them. We only
			 * need one less than total #of messages to deliver.
			 */
			n_needed = ch->w_remote_GP.put - ch->w_local_GP.get - 1;
			if (n_needed > 0 &&
					!(ch->flags & XPC_C_DISCONNECTING)) {
				xpc_activate_kthreads(ch, n_needed);
			}
		} else {
			spin_unlock_irqrestore(&ch->lock, irq_flags);
		}

		xpc_kthread_waitmsgs(part, ch);
	}

	if (atomic_dec_return(&ch->kthreads_assigned) == 0) {
		spin_lock_irqsave(&ch->lock, irq_flags);
		if ((ch->flags & XPC_C_CONNECTCALLOUT) &&
				!(ch->flags & XPC_C_DISCONNECTCALLOUT)) {
			ch->flags |= XPC_C_DISCONNECTCALLOUT;
			spin_unlock_irqrestore(&ch->lock, irq_flags);

			xpc_disconnecting_callout(ch);
		} else {
			spin_unlock_irqrestore(&ch->lock, irq_flags);
		}
		if (atomic_dec_return(&part->nchannels_engaged) == 0) {
			xpc_mark_partition_disengaged(part);
			xpc_IPI_send_disengage(part);
		}
	}


	xpc_msgqueue_deref(ch);

	dev_dbg(xpc_chan, "kthread exiting, partid=%d, channel=%d\n",
		partid, ch_number);

	xpc_part_deref(part);
	return 0;
}


/*
 * For each partition that XPC has established communications with, there is
 * a minimum of one kernel thread assigned to perform any operation that
 * may potentially sleep or block (basically the callouts to the asynchronous
 * functions registered via xpc_connect()).
 *
 * Additional kthreads are created and destroyed by XPC as the workload
 * demands.
 *
 * A kthread is assigned to one of the active channels that exists for a given
 * partition.
 */
void
xpc_create_kthreads(struct xpc_channel *ch, int needed)
{
	unsigned long irq_flags;
	pid_t pid;
	u64 args = XPC_PACK_ARGS(ch->partid, ch->number);
	struct xpc_partition *part = &xpc_partitions[ch->partid];


	while (needed-- > 0) {

		/*
		 * The following is done on behalf of the newly created
		 * kthread. That kthread is responsible for doing the
		 * counterpart to the following before it exits.
		 */
		(void) xpc_part_ref(part);
		xpc_msgqueue_ref(ch);
		if (atomic_inc_return(&ch->kthreads_assigned) == 1 &&
		    atomic_inc_return(&part->nchannels_engaged) == 1) {
			xpc_mark_partition_engaged(part);
		}

		pid = kernel_thread(xpc_daemonize_kthread, (void *) args, 0);
		if (pid < 0) {
			/* the fork failed */
			if (atomic_dec_return(&ch->kthreads_assigned) == 0 &&
			    atomic_dec_return(&part->nchannels_engaged) == 0) {
				xpc_mark_partition_disengaged(part);
				xpc_IPI_send_disengage(part);
			}
			xpc_msgqueue_deref(ch);
			xpc_part_deref(part);

			if (atomic_read(&ch->kthreads_assigned) <
						ch->kthreads_idle_limit) {
				/*
				 * Flag this as an error only if we have an
				 * insufficient #of kthreads for the channel
				 * to function.
				 *
				 * No xpc_msgqueue_ref() is needed here since
				 * the channel mgr is doing this.
				 */
				spin_lock_irqsave(&ch->lock, irq_flags);
				XPC_DISCONNECT_CHANNEL(ch, xpcLackOfResources,
								&irq_flags);
				spin_unlock_irqrestore(&ch->lock, irq_flags);
			}
			break;
		}

		ch->kthreads_created++;	// >>> temporary debug only!!!
	}
}


void
xpc_disconnect_wait(int ch_number)
{
	unsigned long irq_flags;
	partid_t partid;
	struct xpc_partition *part;
	struct xpc_channel *ch;
	int wakeup_channel_mgr;


	/* now wait for all callouts to the caller's function to cease */
	for (partid = 1; partid < XP_MAX_PARTITIONS; partid++) {
		part = &xpc_partitions[partid];

		if (!xpc_part_ref(part)) {
			continue;
		}

		ch = &part->channels[ch_number];

		if (!(ch->flags & XPC_C_WDISCONNECT)) {
			xpc_part_deref(part);
			continue;
		}

		(void) down(&ch->wdisconnect_sema);

		spin_lock_irqsave(&ch->lock, irq_flags);
		DBUG_ON(!(ch->flags & XPC_C_DISCONNECTED));
		wakeup_channel_mgr = 0;

		if (ch->delayed_IPI_flags) {
			if (part->act_state != XPC_P_DEACTIVATING) {
				spin_lock(&part->IPI_lock);
				XPC_SET_IPI_FLAGS(part->local_IPI_amo,
					ch->number, ch->delayed_IPI_flags);
				spin_unlock(&part->IPI_lock);
				wakeup_channel_mgr = 1;
			}
			ch->delayed_IPI_flags = 0;
		}

		ch->flags &= ~XPC_C_WDISCONNECT;
		spin_unlock_irqrestore(&ch->lock, irq_flags);

		if (wakeup_channel_mgr) {
			xpc_wakeup_channel_mgr(part);
		}

		xpc_part_deref(part);
	}
}


static void
xpc_do_exit(enum xpc_retval reason)
{
	partid_t partid;
	int active_part_count;
	struct xpc_partition *part;
	unsigned long printmsg_time;


	/* a 'rmmod XPC' and a 'reboot' cannot both end up here together */
	DBUG_ON(xpc_exiting == 1);

	/*
	 * Let the heartbeat checker thread and the discovery thread
	 * (if one is running) know that they should exit. Also wake up
	 * the heartbeat checker thread in case it's sleeping.
	 */
	xpc_exiting = 1;
	wake_up_interruptible(&xpc_act_IRQ_wq);

	/* ignore all incoming interrupts */
	free_irq(SGI_XPC_ACTIVATE, NULL);

	/* wait for the discovery thread to exit */
	down(&xpc_discovery_exited);

	/* wait for the heartbeat checker thread to exit */
	down(&xpc_hb_checker_exited);


	/* sleep for a 1/3 of a second or so */
	(void) msleep_interruptible(300);


	/* wait for all partitions to become inactive */

	printmsg_time = jiffies;

	do {
		active_part_count = 0;

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

			if (xpc_partition_disengaged(part) &&
					part->act_state == XPC_P_INACTIVE) {
				continue;
			}

			active_part_count++;

			XPC_DEACTIVATE_PARTITION(part, reason);
		}

		if (active_part_count == 0) {
			break;
		}

		if (jiffies >= printmsg_time) {
			dev_info(xpc_part, "waiting for partitions to "
				"deactivate/disengage, active count=%d, remote "
				"engaged=0x%lx\n", active_part_count,
				xpc_partition_engaged(1UL << partid));

			printmsg_time = jiffies +
					(XPC_DISENGAGE_PRINTMSG_INTERVAL * HZ);
		}

		/* sleep for a 1/3 of a second or so */
		(void) msleep_interruptible(300);

	} while (1);

	DBUG_ON(xpc_partition_engaged(-1UL));


	/* indicate to others that our reserved page is uninitialized */
	xpc_rsvd_page->vars_pa = 0;

	/* now it's time to eliminate our heartbeat */
	del_timer_sync(&xpc_hb_timer);
	DBUG_ON(xpc_vars->heartbeating_to_mask != 0);

	/* take ourselves off of the reboot_notifier_list */
	(void) unregister_reboot_notifier(&xpc_reboot_notifier);

	/* take ourselves off of the die_notifier list */
	(void) unregister_die_notifier(&xpc_die_notifier);

	/* close down protections for IPI operations */
	xpc_restrict_IPI_ops();


	/* clear the interface to XPC's functions */
	xpc_clear_interface();

	if (xpc_sysctl) {
		unregister_sysctl_table(xpc_sysctl);
	}
}


/*
 * Called when the system is about to be either restarted or halted.
 */
static void
xpc_die_disengage(void)
{
	struct xpc_partition *part;
	partid_t partid;
	unsigned long engaged;
	long time, print_time, disengage_request_timeout;


	/* keep xpc_hb_checker thread from doing anything (just in case) */
	xpc_exiting = 1;

	xpc_vars->heartbeating_to_mask = 0;  /* indicate we're deactivated */

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

		if (!XPC_SUPPORTS_DISENGAGE_REQUEST(part->
							remote_vars_version)) {

			/* just in case it was left set by an earlier XPC */
			xpc_clear_partition_engaged(1UL << partid);
			continue;
		}

		if (xpc_partition_engaged(1UL << partid) ||
					part->act_state != XPC_P_INACTIVE) {
			xpc_request_partition_disengage(part);
			xpc_mark_partition_disengaged(part);
			xpc_IPI_send_disengage(part);
		}
	}

	print_time = rtc_time();
	disengage_request_timeout = print_time +
		(xpc_disengage_request_timelimit * sn_rtc_cycles_per_second);

	/* wait for all other partitions to disengage from us */

	while ((engaged = xpc_partition_engaged(-1UL)) &&
			(time = rtc_time()) < disengage_request_timeout) {

		if (time >= print_time) {
			dev_info(xpc_part, "waiting for remote partitions to "
				"disengage, engaged=0x%lx\n", engaged);
			print_time = time + (XPC_DISENGAGE_PRINTMSG_INTERVAL *
						sn_rtc_cycles_per_second);
		}
	}
	dev_info(xpc_part, "finished waiting for remote partitions to "
				"disengage, engaged=0x%lx\n", engaged);
}


/*
 * This function is called when the system is being rebooted.
 */
static int
xpc_system_reboot(struct notifier_block *nb, unsigned long event, void *unused)
{
	enum xpc_retval reason;


	switch (event) {
	case SYS_RESTART:
		reason = xpcSystemReboot;
		break;
	case SYS_HALT:
		reason = xpcSystemHalt;
		break;
	case SYS_POWER_OFF:
		reason = xpcSystemPoweroff;
		break;
	default:
		reason = xpcSystemGoingDown;
	}

	xpc_do_exit(reason);
	return NOTIFY_DONE;
}


/*
 * This function is called when the system is being rebooted.
 */
static int
xpc_system_die(struct notifier_block *nb, unsigned long event, void *unused)
{
	switch (event) {
	case DIE_MACHINE_RESTART:
	case DIE_MACHINE_HALT:
		xpc_die_disengage();
		break;
	case DIE_MCA_MONARCH_ENTER:
	case DIE_INIT_MONARCH_ENTER:
		xpc_vars->heartbeat++;
		xpc_vars->heartbeat_offline = 1;
		break;
	case DIE_MCA_MONARCH_LEAVE:
	case DIE_INIT_MONARCH_LEAVE:
		xpc_vars->heartbeat++;
		xpc_vars->heartbeat_offline = 0;
		break;
	}

	return NOTIFY_DONE;
}


int __init
xpc_init(void)
{
	int ret;
	partid_t partid;
	struct xpc_partition *part;
	pid_t pid;


	if (!ia64_platform_is("sn2")) {
		return -ENODEV;
	}

	/*
	 * xpc_remote_copy_buffer is used as a temporary buffer for bte_copy'ng
	 * various portions of a partition's reserved page. Its size is based
	 * on the size of the reserved page header and part_nasids mask. So we
	 * need to ensure that the other items will fit as well.
	 */
	if (XPC_RP_VARS_SIZE > XPC_RP_HEADER_SIZE + XP_NASID_MASK_BYTES) {
		dev_err(xpc_part, "xpc_remote_copy_buffer is not big enough\n");
		return -EPERM;
	}
	DBUG_ON((u64) xpc_remote_copy_buffer !=
				L1_CACHE_ALIGN((u64) xpc_remote_copy_buffer));

	snprintf(xpc_part->bus_id, BUS_ID_SIZE, "part");
	snprintf(xpc_chan->bus_id, BUS_ID_SIZE, "chan");

	xpc_sysctl = register_sysctl_table(xpc_sys_dir, 1);

	/*
	 * The first few fields of each entry of xpc_partitions[] need to
	 * be initialized now so that calls to xpc_connect() and
	 * xpc_disconnect() can be made prior to the activation of any remote
	 * partition. NOTE THAT NONE OF THE OTHER FIELDS BELONGING TO THESE
	 * ENTRIES ARE MEANINGFUL UNTIL AFTER AN ENTRY'S CORRESPONDING
	 * PARTITION HAS BEEN ACTIVATED.
	 */
	for (partid = 1; partid < XP_MAX_PARTITIONS; partid++) {
		part = &xpc_partitions[partid];

		DBUG_ON((u64) part != L1_CACHE_ALIGN((u64) part));

		part->act_IRQ_rcvd = 0;
		spin_lock_init(&part->act_lock);
		part->act_state = XPC_P_INACTIVE;
		XPC_SET_REASON(part, 0, 0);

		init_timer(&part->disengage_request_timer);
		part->disengage_request_timer.function =
				xpc_timeout_partition_disengage_request;
		part->disengage_request_timer.data = (unsigned long) part;

		part->setup_state = XPC_P_UNSET;
		init_waitqueue_head(&part->teardown_wq);
		atomic_set(&part->references, 0);
	}

	/*
	 * Open up protections for IPI operations (and AMO operations on
	 * Shub 1.1 systems).
	 */
	xpc_allow_IPI_ops();

	/*
	 * Interrupts being processed will increment this atomic variable and
	 * awaken the heartbeat thread which will process the interrupts.
	 */
	atomic_set(&xpc_act_IRQ_rcvd, 0);

	/*
	 * This is safe to do before the xpc_hb_checker thread has started
	 * because the handler releases a wait queue.  If an interrupt is
	 * received before the thread is waiting, it will not go to sleep,
	 * but rather immediately process the interrupt.
	 */
	ret = request_irq(SGI_XPC_ACTIVATE, xpc_act_IRQ_handler, 0,
							"xpc hb", NULL);
	if (ret != 0) {
		dev_err(xpc_part, "can't register ACTIVATE IRQ handler, "
			"errno=%d\n", -ret);

		xpc_restrict_IPI_ops();

		if (xpc_sysctl) {
			unregister_sysctl_table(xpc_sysctl);
		}
		return -EBUSY;
	}

	/*
	 * Fill the partition reserved page with the information needed by
	 * other partitions to discover we are alive and establish initial
	 * communications.
	 */
	xpc_rsvd_page = xpc_rsvd_page_init();
	if (xpc_rsvd_page == NULL) {
		dev_err(xpc_part, "could not setup our reserved page\n");

		free_irq(SGI_XPC_ACTIVATE, NULL);
		xpc_restrict_IPI_ops();

		if (xpc_sysctl) {
			unregister_sysctl_table(xpc_sysctl);
		}
		return -EBUSY;
	}


	/* add ourselves to the reboot_notifier_list */
	ret = register_reboot_notifier(&xpc_reboot_notifier);
	if (ret != 0) {
		dev_warn(xpc_part, "can't register reboot notifier\n");
	}

	/* add ourselves to the die_notifier list (i.e., ia64die_chain) */
	ret = register_die_notifier(&xpc_die_notifier);
	if (ret != 0) {
		dev_warn(xpc_part, "can't register die notifier\n");
	}


	/*
	 * Set the beating to other partitions into motion.  This is
	 * the last requirement for other partitions' discovery to
	 * initiate communications with us.
	 */
	init_timer(&xpc_hb_timer);
	xpc_hb_timer.function = xpc_hb_beater;
	xpc_hb_beater(0);


	/*
	 * The real work-horse behind xpc.  This processes incoming
	 * interrupts and monitors remote heartbeats.
	 */
	pid = kernel_thread(xpc_hb_checker, NULL, 0);
	if (pid < 0) {
		dev_err(xpc_part, "failed while forking hb check thread\n");

		/* indicate to others that our reserved page is uninitialized */
		xpc_rsvd_page->vars_pa = 0;

		/* take ourselves off of the reboot_notifier_list */
		(void) unregister_reboot_notifier(&xpc_reboot_notifier);

		/* take ourselves off of the die_notifier list */
		(void) unregister_die_notifier(&xpc_die_notifier);

		del_timer_sync(&xpc_hb_timer);
		free_irq(SGI_XPC_ACTIVATE, NULL);
		xpc_restrict_IPI_ops();

		if (xpc_sysctl) {
			unregister_sysctl_table(xpc_sysctl);
		}
		return -EBUSY;
	}


	/*
	 * Startup a thread that will attempt to discover other partitions to
	 * activate based on info provided by SAL. This new thread is short
	 * lived and will exit once discovery is complete.
	 */
	pid = kernel_thread(xpc_initiate_discovery, NULL, 0);
	if (pid < 0) {
		dev_err(xpc_part, "failed while forking discovery thread\n");

		/* mark this new thread as a non-starter */
		up(&xpc_discovery_exited);

		xpc_do_exit(xpcUnloading);
		return -EBUSY;
	}


	/* set the interface to point at XPC's functions */
	xpc_set_interface(xpc_initiate_connect, xpc_initiate_disconnect,
			  xpc_initiate_allocate, xpc_initiate_send,
			  xpc_initiate_send_notify, xpc_initiate_received,
			  xpc_initiate_partid_to_nasids);

	return 0;
}
module_init(xpc_init);


void __exit
xpc_exit(void)
{
	xpc_do_exit(xpcUnloading);
}
module_exit(xpc_exit);


MODULE_AUTHOR("Silicon Graphics, Inc.");
MODULE_DESCRIPTION("Cross Partition Communication (XPC) support");
MODULE_LICENSE("GPL");

module_param(xpc_hb_interval, int, 0);
MODULE_PARM_DESC(xpc_hb_interval, "Number of seconds between "
		"heartbeat increments.");

module_param(xpc_hb_check_interval, int, 0);
MODULE_PARM_DESC(xpc_hb_check_interval, "Number of seconds between "
		"heartbeat checks.");

module_param(xpc_disengage_request_timelimit, int, 0);
MODULE_PARM_DESC(xpc_disengage_request_timelimit, "Number of seconds to wait "
		"for disengage request to complete.");