summaryrefslogtreecommitdiff
path: root/fs/orangefs/devpvfs2-req.c
blob: e37b6479a6a19116c8ddd842018e117b3a2ed467 (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
/*
 * (C) 2001 Clemson University and The University of Chicago
 *
 * Changes by Acxiom Corporation to add protocol version to kernel
 * communication, Copyright Acxiom Corporation, 2005.
 *
 * See COPYING in top-level directory.
 */

#include "protocol.h"
#include "pvfs2-kernel.h"
#include "pvfs2-dev-proto.h"
#include "pvfs2-bufmap.h"

#include <linux/debugfs.h>
#include <linux/slab.h>

/* this file implements the /dev/pvfs2-req device node */

static int open_access_count;

#define DUMP_DEVICE_ERROR()                                                   \
do {                                                                          \
	gossip_err("*****************************************************\n");\
	gossip_err("PVFS2 Device Error:  You cannot open the device file ");  \
	gossip_err("\n/dev/%s more than once.  Please make sure that\nthere " \
		   "are no ", PVFS2_REQDEVICE_NAME);                          \
	gossip_err("instances of a program using this device\ncurrently "     \
		   "running. (You must verify this!)\n");                     \
	gossip_err("For example, you can use the lsof program as follows:\n");\
	gossip_err("'lsof | grep %s' (run this as root)\n",                   \
		   PVFS2_REQDEVICE_NAME);                                     \
	gossip_err("  open_access_count = %d\n", open_access_count);          \
	gossip_err("*****************************************************\n");\
} while (0)

static int hash_func(__u64 tag, int table_size)
{
	return do_div(tag, (unsigned int)table_size);
}

static void pvfs2_devreq_add_op(struct pvfs2_kernel_op_s *op)
{
	int index = hash_func(op->tag, hash_table_size);

	spin_lock(&htable_ops_in_progress_lock);
	list_add_tail(&op->list, &htable_ops_in_progress[index]);
	spin_unlock(&htable_ops_in_progress_lock);
}

static struct pvfs2_kernel_op_s *pvfs2_devreq_remove_op(__u64 tag)
{
	struct pvfs2_kernel_op_s *op, *next;
	int index;

	index = hash_func(tag, hash_table_size);

	spin_lock(&htable_ops_in_progress_lock);
	list_for_each_entry_safe(op,
				 next,
				 &htable_ops_in_progress[index],
				 list) {
		if (op->tag == tag) {
			list_del(&op->list);
			spin_unlock(&htable_ops_in_progress_lock);
			return op;
		}
	}

	spin_unlock(&htable_ops_in_progress_lock);
	return NULL;
}

static int pvfs2_devreq_open(struct inode *inode, struct file *file)
{
	int ret = -EINVAL;

	if (!(file->f_flags & O_NONBLOCK)) {
		gossip_err("pvfs2: device cannot be opened in blocking mode\n");
		goto out;
	}
	ret = -EACCES;
	gossip_debug(GOSSIP_DEV_DEBUG, "pvfs2-client-core: opening device\n");
	mutex_lock(&devreq_mutex);

	if (open_access_count == 0) {
		ret = generic_file_open(inode, file);
		if (ret == 0)
			open_access_count++;
	} else {
		DUMP_DEVICE_ERROR();
	}
	mutex_unlock(&devreq_mutex);

out:

	gossip_debug(GOSSIP_DEV_DEBUG,
		     "pvfs2-client-core: open device complete (ret = %d)\n",
		     ret);
	return ret;
}

static ssize_t pvfs2_devreq_read(struct file *file,
				 char __user *buf,
				 size_t count, loff_t *offset)
{
	struct pvfs2_kernel_op_s *op, *temp;
	__s32 proto_ver = PVFS_KERNEL_PROTO_VERSION;
	static __s32 magic = PVFS2_DEVREQ_MAGIC;
	struct pvfs2_kernel_op_s *cur_op = NULL;
	unsigned long ret;

	/* We do not support blocking IO. */
	if (!(file->f_flags & O_NONBLOCK)) {
		gossip_err("orangefs: blocking reads are not supported! (pvfs2-client-core bug)\n");
		return -EINVAL;
	}

	/*
	 * The client will do an ioctl to find MAX_ALIGNED_DEV_REQ_UPSIZE, then
	 * always read with that size buffer.
	 */
	if (count != MAX_ALIGNED_DEV_REQ_UPSIZE) {
		gossip_err("orangefs: client-core tried to read wrong size\n");
		return -EINVAL;
	}

	/* Get next op (if any) from top of list. */
	spin_lock(&pvfs2_request_list_lock);
	list_for_each_entry_safe(op, temp, &pvfs2_request_list, list) {
		__s32 fsid;
		/* This lock is held past the end of the loop when we break. */
		spin_lock(&op->lock);

		fsid = fsid_of_op(op);
		if (fsid != PVFS_FS_ID_NULL) {
			int ret;
			/* Skip ops whose filesystem needs to be mounted. */
			ret = fs_mount_pending(fsid);
			if (ret == 1) {
				gossip_debug(GOSSIP_DEV_DEBUG,
				    "orangefs: skipping op tag %llu %s\n",
				    llu(op->tag), get_opname_string(op));
				spin_unlock(&op->lock);
				continue;
			/* Skip ops whose filesystem we don't know about unless
			 * it is being mounted. */
			/* XXX: is there a better way to detect this? */
			} else if (ret == -1 &&
				   !(op->upcall.type == PVFS2_VFS_OP_FS_MOUNT ||
				     op->upcall.type == PVFS2_VFS_OP_GETATTR)) {
				gossip_debug(GOSSIP_DEV_DEBUG,
				    "orangefs: skipping op tag %llu %s\n",
				    llu(op->tag), get_opname_string(op));
				gossip_err(
				    "orangefs: ERROR: fs_mount_pending %d\n",
				    fsid);
				spin_unlock(&op->lock);
				continue;
			}
		}
		/*
		 * Either this op does not pertain to a filesystem, is mounting
		 * a filesystem, or pertains to a mounted filesystem. Let it
		 * through.
		 */
		cur_op = op;
		break;
	}

	/*
	 * At this point we either have a valid op and can continue or have not
	 * found an op and must ask the client to try again later.
	 */
	if (!cur_op) {
		spin_unlock(&pvfs2_request_list_lock);
		return -EAGAIN;
	}

	gossip_debug(GOSSIP_DEV_DEBUG, "orangefs: reading op tag %llu %s\n",
		     llu(cur_op->tag), get_opname_string(cur_op));

	/*
	 * Such an op should never be on the list in the first place. If so, we
	 * will abort.
	 */
	if (op_state_in_progress(cur_op) || op_state_serviced(cur_op)) {
		gossip_err("orangefs: ERROR: Current op already queued.\n");
		list_del(&cur_op->list);
		spin_unlock(&cur_op->lock);
		spin_unlock(&pvfs2_request_list_lock);
		return -EAGAIN;
	}

	/*
	 * Set the operation to be in progress and move it between lists since
	 * it has been sent to the client.
	 */
	set_op_state_inprogress(cur_op);

	list_del(&cur_op->list);
	spin_unlock(&pvfs2_request_list_lock);
	pvfs2_devreq_add_op(cur_op);
	spin_unlock(&cur_op->lock);

	/* Push the upcall out. */
	ret = copy_to_user(buf, &proto_ver, sizeof(__s32));
	if (ret != 0)
		goto error;
	ret = copy_to_user(buf+sizeof(__s32), &magic, sizeof(__s32));
	if (ret != 0)
		goto error;
	ret = copy_to_user(buf+2 * sizeof(__s32), &cur_op->tag, sizeof(__u64));
	if (ret != 0)
		goto error;
	ret = copy_to_user(buf+2*sizeof(__s32)+sizeof(__u64), &cur_op->upcall,
			   sizeof(struct pvfs2_upcall_s));
	if (ret != 0)
		goto error;

	/* The client only asks to read one size buffer. */
	return MAX_ALIGNED_DEV_REQ_UPSIZE;
error:
	/*
	 * We were unable to copy the op data to the client. Put the op back in
	 * list. If client has crashed, the op will be purged later when the
	 * device is released.
	 */
	gossip_err("orangefs: Failed to copy data to user space\n");
	spin_lock(&pvfs2_request_list_lock);
	spin_lock(&cur_op->lock);
	set_op_state_waiting(cur_op);
	pvfs2_devreq_remove_op(cur_op->tag);
	list_add(&cur_op->list, &pvfs2_request_list);
	spin_unlock(&cur_op->lock);
	spin_unlock(&pvfs2_request_list_lock);
	return -EFAULT;
}

/* Function for writev() callers into the device */
static ssize_t pvfs2_devreq_writev(struct file *file,
				   const struct iovec *iov,
				   size_t count,
				   loff_t *offset)
{
	struct pvfs2_kernel_op_s *op = NULL;
	void *buffer = NULL;
	void *ptr = NULL;
	unsigned long i = 0;
	static int max_downsize = MAX_ALIGNED_DEV_REQ_DOWNSIZE;
	int ret = 0, num_remaining = max_downsize;
	int notrailer_count = 4; /* num elements in iovec without trailer */
	int payload_size = 0;
	__s32 magic = 0;
	__s32 proto_ver = 0;
	__u64 tag = 0;
	ssize_t total_returned_size = 0;

	/* Either there is a trailer or there isn't */
	if (count != notrailer_count && count != (notrailer_count + 1)) {
		gossip_err("Error: Number of iov vectors is (%zu) and notrailer count is %d\n",
			count,
			notrailer_count);
		return -EPROTO;
	}
	buffer = dev_req_alloc();
	if (!buffer)
		return -ENOMEM;
	ptr = buffer;

	for (i = 0; i < notrailer_count; i++) {
		if (iov[i].iov_len > num_remaining) {
			gossip_err
			    ("writev error: Freeing buffer and returning\n");
			dev_req_release(buffer);
			return -EMSGSIZE;
		}
		ret = copy_from_user(ptr, iov[i].iov_base, iov[i].iov_len);
		if (ret) {
			gossip_err("Failed to copy data from user space\n");
			dev_req_release(buffer);
			return -EIO;
		}
		num_remaining -= iov[i].iov_len;
		ptr += iov[i].iov_len;
		payload_size += iov[i].iov_len;
	}
	total_returned_size = payload_size;

	/* these elements are currently 8 byte aligned (8 bytes for (version +
	 * magic) 8 bytes for tag).  If you add another element, either
	 * make it 8 bytes big, or use get_unaligned when asigning.
	 */
	ptr = buffer;
	proto_ver = *((__s32 *) ptr);
	ptr += sizeof(__s32);

	magic = *((__s32 *) ptr);
	ptr += sizeof(__s32);

	tag = *((__u64 *) ptr);
	ptr += sizeof(__u64);

	if (magic != PVFS2_DEVREQ_MAGIC) {
		gossip_err("Error: Device magic number does not match.\n");
		dev_req_release(buffer);
		return -EPROTO;
	}

	/*
	 * proto_ver = 20902 for 2.9.2
	 */

	op = pvfs2_devreq_remove_op(tag);
	if (op) {
		/* Increase ref count! */
		get_op(op);
		/* cut off magic and tag from payload size */
		payload_size -= (2 * sizeof(__s32) + sizeof(__u64));
		if (payload_size <= sizeof(struct pvfs2_downcall_s))
			/* copy the passed in downcall into the op */
			memcpy(&op->downcall,
			       ptr,
			       sizeof(struct pvfs2_downcall_s));
		else
			gossip_debug(GOSSIP_DEV_DEBUG,
				     "writev: Ignoring %d bytes\n",
				     payload_size);

		/* Do not allocate needlessly if client-core forgets
		 * to reset trailer size on op errors.
		 */
		if (op->downcall.status == 0 && op->downcall.trailer_size > 0) {
			__u64 trailer_size = op->downcall.trailer_size;
			size_t size;
			gossip_debug(GOSSIP_DEV_DEBUG,
				     "writev: trailer size %ld\n",
				     (unsigned long)trailer_size);
			if (count != (notrailer_count + 1)) {
				gossip_err("Error: trailer size (%ld) is non-zero, no trailer elements though? (%zu)\n", (unsigned long)trailer_size, count);
				dev_req_release(buffer);
				put_op(op);
				return -EPROTO;
			}
			size = iov[notrailer_count].iov_len;
			if (size > trailer_size) {
				gossip_err("writev error: trailer size (%ld) != iov_len (%zd)\n", (unsigned long)trailer_size, size);
				dev_req_release(buffer);
				put_op(op);
				return -EMSGSIZE;
			}
			/* Allocate a buffer large enough to hold the
			 * trailer bytes.
			 */
			op->downcall.trailer_buf = vmalloc(trailer_size);
			if (op->downcall.trailer_buf != NULL) {
				gossip_debug(GOSSIP_DEV_DEBUG, "vmalloc: %p\n",
					     op->downcall.trailer_buf);
				ret = copy_from_user(op->downcall.trailer_buf,
						     iov[notrailer_count].
						     iov_base,
						     size);
				if (ret) {
					gossip_err("Failed to copy trailer data from user space\n");
					dev_req_release(buffer);
					gossip_debug(GOSSIP_DEV_DEBUG,
						     "vfree: %p\n",
						     op->downcall.trailer_buf);
					vfree(op->downcall.trailer_buf);
					op->downcall.trailer_buf = NULL;
					put_op(op);
					return -EIO;
				}
				memset(op->downcall.trailer_buf + size, 0,
				       trailer_size - size);
			} else {
				/* Change downcall status */
				op->downcall.status = -ENOMEM;
				gossip_err("writev: could not vmalloc for trailer!\n");
			}
		}

		/* if this operation is an I/O operation and if it was
		 * initiated on behalf of a *synchronous* VFS I/O operation,
		 * only then we need to wait
		 * for all data to be copied before we can return to avoid
		 * buffer corruption and races that can pull the buffers
		 * out from under us.
		 *
		 * Essentially we're synchronizing with other parts of the
		 * vfs implicitly by not allowing the user space
		 * application reading/writing this device to return until
		 * the buffers are done being used.
		 */
		if (op->upcall.type == PVFS2_VFS_OP_FILE_IO &&
		    op->upcall.req.io.async_vfs_io == PVFS_VFS_SYNC_IO) {
			int timed_out = 0;
			DECLARE_WAITQUEUE(wait_entry, current);

			/* tell the vfs op waiting on a waitqueue
			 * that this op is done
			 */
			spin_lock(&op->lock);
			set_op_state_serviced(op);
			spin_unlock(&op->lock);

			add_wait_queue_exclusive(&op->io_completion_waitq,
						 &wait_entry);
			wake_up_interruptible(&op->waitq);

			while (1) {
				set_current_state(TASK_INTERRUPTIBLE);

				spin_lock(&op->lock);
				if (op->io_completed) {
					spin_unlock(&op->lock);
					break;
				}
				spin_unlock(&op->lock);

				if (!signal_pending(current)) {
					int timeout =
					    MSECS_TO_JIFFIES(1000 *
							     op_timeout_secs);
					if (!schedule_timeout(timeout)) {
						gossip_debug(GOSSIP_DEV_DEBUG, "*** I/O wait time is up\n");
						timed_out = 1;
						break;
					}
					continue;
				}

				gossip_debug(GOSSIP_DEV_DEBUG, "*** signal on I/O wait -- aborting\n");
				break;
			}

			set_current_state(TASK_RUNNING);
			remove_wait_queue(&op->io_completion_waitq,
					  &wait_entry);

			/* NOTE: for I/O operations we handle releasing the op
			 * object except in the case of timeout.  the reason we
			 * can't free the op in timeout cases is that the op
			 * service logic in the vfs retries operations using
			 * the same op ptr, thus it can't be freed.
			 */
			if (!timed_out)
				op_release(op);
		} else {

			/*
			 * tell the vfs op waiting on a waitqueue that
			 * this op is done
			 */
			spin_lock(&op->lock);
			set_op_state_serviced(op);
			spin_unlock(&op->lock);
			/*
			 * for every other operation (i.e. non-I/O), we need to
			 * wake up the callers for downcall completion
			 * notification
			 */
			wake_up_interruptible(&op->waitq);
		}
	} else {
		/* ignore downcalls that we're not interested in */
		gossip_debug(GOSSIP_DEV_DEBUG,
			     "WARNING: No one's waiting for tag %llu\n",
			     llu(tag));
	}
	dev_req_release(buffer);

	return total_returned_size;
}

static ssize_t pvfs2_devreq_write_iter(struct kiocb *iocb,
				      struct iov_iter *iter)
{
	return pvfs2_devreq_writev(iocb->ki_filp,
				   iter->iov,
				   iter->nr_segs,
				   &iocb->ki_pos);
}

/* Returns whether any FS are still pending remounted */
static int mark_all_pending_mounts(void)
{
	int unmounted = 1;
	struct pvfs2_sb_info_s *pvfs2_sb = NULL;

	spin_lock(&pvfs2_superblocks_lock);
	list_for_each_entry(pvfs2_sb, &pvfs2_superblocks, list) {
		/* All of these file system require a remount */
		pvfs2_sb->mount_pending = 1;
		unmounted = 0;
	}
	spin_unlock(&pvfs2_superblocks_lock);
	return unmounted;
}

/*
 * Determine if a given file system needs to be remounted or not
 *  Returns -1 on error
 *           0 if already mounted
 *           1 if needs remount
 */
int fs_mount_pending(__s32 fsid)
{
	int mount_pending = -1;
	struct pvfs2_sb_info_s *pvfs2_sb = NULL;

	spin_lock(&pvfs2_superblocks_lock);
	list_for_each_entry(pvfs2_sb, &pvfs2_superblocks, list) {
		if (pvfs2_sb->fs_id == fsid) {
			mount_pending = pvfs2_sb->mount_pending;
			break;
		}
	}
	spin_unlock(&pvfs2_superblocks_lock);
	return mount_pending;
}

/*
 * NOTE: gets called when the last reference to this device is dropped.
 * Using the open_access_count variable, we enforce a reference count
 * on this file so that it can be opened by only one process at a time.
 * the devreq_mutex is used to make sure all i/o has completed
 * before we call pvfs_bufmap_finalize, and similar such tricky
 * situations
 */
static int pvfs2_devreq_release(struct inode *inode, struct file *file)
{
	int unmounted = 0;

	gossip_debug(GOSSIP_DEV_DEBUG,
		     "%s:pvfs2-client-core: exiting, closing device\n",
		     __func__);

	mutex_lock(&devreq_mutex);
	pvfs_bufmap_finalize();

	open_access_count--;

	unmounted = mark_all_pending_mounts();
	gossip_debug(GOSSIP_DEV_DEBUG, "PVFS2 Device Close: Filesystem(s) %s\n",
		     (unmounted ? "UNMOUNTED" : "MOUNTED"));
	mutex_unlock(&devreq_mutex);

	/*
	 * Walk through the list of ops in the request list, mark them
	 * as purged and wake them up.
	 */
	purge_waiting_ops();
	/*
	 * Walk through the hash table of in progress operations; mark
	 * them as purged and wake them up
	 */
	purge_inprogress_ops();
	gossip_debug(GOSSIP_DEV_DEBUG,
		     "pvfs2-client-core: device close complete\n");
	return 0;
}

int is_daemon_in_service(void)
{
	int in_service;

	/*
	 * What this function does is checks if client-core is alive
	 * based on the access count we maintain on the device.
	 */
	mutex_lock(&devreq_mutex);
	in_service = open_access_count == 1 ? 0 : -EIO;
	mutex_unlock(&devreq_mutex);
	return in_service;
}

static inline long check_ioctl_command(unsigned int command)
{
	/* Check for valid ioctl codes */
	if (_IOC_TYPE(command) != PVFS_DEV_MAGIC) {
		gossip_err("device ioctl magic numbers don't match! Did you rebuild pvfs2-client-core/libpvfs2? [cmd %x, magic %x != %x]\n",
			command,
			_IOC_TYPE(command),
			PVFS_DEV_MAGIC);
		return -EINVAL;
	}
	/* and valid ioctl commands */
	if (_IOC_NR(command) >= PVFS_DEV_MAXNR || _IOC_NR(command) <= 0) {
		gossip_err("Invalid ioctl command number [%d >= %d]\n",
			   _IOC_NR(command), PVFS_DEV_MAXNR);
		return -ENOIOCTLCMD;
	}
	return 0;
}

static long dispatch_ioctl_command(unsigned int command, unsigned long arg)
{
	static __s32 magic = PVFS2_DEVREQ_MAGIC;
	static __s32 max_up_size = MAX_ALIGNED_DEV_REQ_UPSIZE;
	static __s32 max_down_size = MAX_ALIGNED_DEV_REQ_DOWNSIZE;
	struct PVFS_dev_map_desc user_desc;
	int ret = 0;
	struct dev_mask_info_s mask_info = { 0 };
	struct dev_mask2_info_s mask2_info = { 0, 0 };
	int upstream_kmod = 1;
	struct list_head *tmp = NULL;
	struct pvfs2_sb_info_s *pvfs2_sb = NULL;

	/* mtmoore: add locking here */

	switch (command) {
	case PVFS_DEV_GET_MAGIC:
		return ((put_user(magic, (__s32 __user *) arg) == -EFAULT) ?
			-EIO :
			0);
	case PVFS_DEV_GET_MAX_UPSIZE:
		return ((put_user(max_up_size,
				  (__s32 __user *) arg) == -EFAULT) ?
					-EIO :
					0);
	case PVFS_DEV_GET_MAX_DOWNSIZE:
		return ((put_user(max_down_size,
				  (__s32 __user *) arg) == -EFAULT) ?
					-EIO :
					0);
	case PVFS_DEV_MAP:
		ret = copy_from_user(&user_desc,
				     (struct PVFS_dev_map_desc __user *)
				     arg,
				     sizeof(struct PVFS_dev_map_desc));
		return ret ? -EIO : pvfs_bufmap_initialize(&user_desc);
	case PVFS_DEV_REMOUNT_ALL:
		gossip_debug(GOSSIP_DEV_DEBUG,
			     "pvfs2_devreq_ioctl: got PVFS_DEV_REMOUNT_ALL\n");

		/*
		 * remount all mounted pvfs2 volumes to regain the lost
		 * dynamic mount tables (if any) -- NOTE: this is done
		 * without keeping the superblock list locked due to the
		 * upcall/downcall waiting.  also, the request semaphore is
		 * used to ensure that no operations will be serviced until
		 * all of the remounts are serviced (to avoid ops between
		 * mounts to fail)
		 */
		ret = mutex_lock_interruptible(&request_mutex);
		if (ret < 0)
			return ret;
		gossip_debug(GOSSIP_DEV_DEBUG,
			     "pvfs2_devreq_ioctl: priority remount in progress\n");
		list_for_each(tmp, &pvfs2_superblocks) {
			pvfs2_sb =
				list_entry(tmp, struct pvfs2_sb_info_s, list);
			if (pvfs2_sb && (pvfs2_sb->sb)) {
				gossip_debug(GOSSIP_DEV_DEBUG,
					     "Remounting SB %p\n",
					     pvfs2_sb);

				ret = pvfs2_remount(pvfs2_sb->sb);
				if (ret) {
					gossip_debug(GOSSIP_DEV_DEBUG,
						     "SB %p remount failed\n",
						     pvfs2_sb);
						break;
				}
			}
		}
		gossip_debug(GOSSIP_DEV_DEBUG,
			     "pvfs2_devreq_ioctl: priority remount complete\n");
		mutex_unlock(&request_mutex);
		return ret;

	case PVFS_DEV_UPSTREAM:
		ret = copy_to_user((void __user *)arg,
				    &upstream_kmod,
				    sizeof(upstream_kmod));

		if (ret != 0)
			return -EIO;
		else
			return ret;

	case PVFS_DEV_CLIENT_MASK:
		ret = copy_from_user(&mask2_info,
				     (void __user *)arg,
				     sizeof(struct dev_mask2_info_s));

		if (ret != 0)
			return -EIO;

		client_debug_mask.mask1 = mask2_info.mask1_value;
		client_debug_mask.mask2 = mask2_info.mask2_value;

		pr_info("%s: client debug mask has been been received "
			":%llx: :%llx:\n",
			__func__,
			(unsigned long long)client_debug_mask.mask1,
			(unsigned long long)client_debug_mask.mask2);

		return ret;

	case PVFS_DEV_CLIENT_STRING:
		ret = copy_from_user(&client_debug_array_string,
				     (void __user *)arg,
				     PVFS2_MAX_DEBUG_STRING_LEN);
		if (ret != 0) {
			pr_info("%s: "
				"PVFS_DEV_CLIENT_STRING: copy_from_user failed"
				"\n",
				__func__);
			return -EIO;
		}

		pr_info("%s: client debug array string has been been received."
			"\n",
			__func__);

		if (!help_string_initialized) {

			/* Free the "we don't know yet" default string... */
			kfree(debug_help_string);

			/* build a proper debug help string */
			if (orangefs_prepare_debugfs_help_string(0)) {
				gossip_err("%s: "
					   "prepare_debugfs_help_string failed"
					   "\n",
					   __func__);
				return -EIO;
			}

			/* Replace the boilerplate boot-time debug-help file. */
			debugfs_remove(help_file_dentry);

			help_file_dentry =
				debugfs_create_file(
					ORANGEFS_KMOD_DEBUG_HELP_FILE,
					0444,
					debug_dir,
					debug_help_string,
					&debug_help_fops);

			if (!help_file_dentry) {
				gossip_err("%s: debugfs_create_file failed for"
					   " :%s:!\n",
					   __func__,
					   ORANGEFS_KMOD_DEBUG_HELP_FILE);
				return -EIO;
			}
		}

		debug_mask_to_string(&client_debug_mask, 1);

		debugfs_remove(client_debug_dentry);

		pvfs2_client_debug_init();

		help_string_initialized++;

		return ret;

	case PVFS_DEV_DEBUG:
		ret = copy_from_user(&mask_info,
				     (void __user *)arg,
				     sizeof(mask_info));

		if (ret != 0)
			return -EIO;

		if (mask_info.mask_type == KERNEL_MASK) {
			if ((mask_info.mask_value == 0)
			    && (kernel_mask_set_mod_init)) {
				/*
				 * the kernel debug mask was set when the
				 * kernel module was loaded; don't override
				 * it if the client-core was started without
				 * a value for PVFS2_KMODMASK.
				 */
				return 0;
			}
			debug_mask_to_string(&mask_info.mask_value,
					     mask_info.mask_type);
			gossip_debug_mask = mask_info.mask_value;
			pr_info("PVFS: kernel debug mask has been modified to "
				":%s: :%llx:\n",
				kernel_debug_string,
				(unsigned long long)gossip_debug_mask);
		} else if (mask_info.mask_type == CLIENT_MASK) {
			debug_mask_to_string(&mask_info.mask_value,
					     mask_info.mask_type);
			pr_info("PVFS: client debug mask has been modified to"
				":%s: :%llx:\n",
				client_debug_string,
				llu(mask_info.mask_value));
		} else {
			gossip_lerr("Invalid mask type....\n");
			return -EINVAL;
		}

		return ret;

	default:
		return -ENOIOCTLCMD;
	}
	return -ENOIOCTLCMD;
}

static long pvfs2_devreq_ioctl(struct file *file,
			       unsigned int command, unsigned long arg)
{
	long ret;

	/* Check for properly constructed commands */
	ret = check_ioctl_command(command);
	if (ret < 0)
		return (int)ret;

	return (int)dispatch_ioctl_command(command, arg);
}

#ifdef CONFIG_COMPAT		/* CONFIG_COMPAT is in .config */

/*  Compat structure for the PVFS_DEV_MAP ioctl */
struct PVFS_dev_map_desc32 {
	compat_uptr_t ptr;
	__s32 total_size;
	__s32 size;
	__s32 count;
};

static unsigned long translate_dev_map26(unsigned long args, long *error)
{
	struct PVFS_dev_map_desc32 __user *p32 = (void __user *)args;
	/*
	 * Depending on the architecture, allocate some space on the
	 * user-call-stack based on our expected layout.
	 */
	struct PVFS_dev_map_desc __user *p =
	    compat_alloc_user_space(sizeof(*p));
	compat_uptr_t addr;

	*error = 0;
	/* get the ptr from the 32 bit user-space */
	if (get_user(addr, &p32->ptr))
		goto err;
	/* try to put that into a 64-bit layout */
	if (put_user(compat_ptr(addr), &p->ptr))
		goto err;
	/* copy the remaining fields */
	if (copy_in_user(&p->total_size, &p32->total_size, sizeof(__s32)))
		goto err;
	if (copy_in_user(&p->size, &p32->size, sizeof(__s32)))
		goto err;
	if (copy_in_user(&p->count, &p32->count, sizeof(__s32)))
		goto err;
	return (unsigned long)p;
err:
	*error = -EFAULT;
	return 0;
}

/*
 * 32 bit user-space apps' ioctl handlers when kernel modules
 * is compiled as a 64 bit one
 */
static long pvfs2_devreq_compat_ioctl(struct file *filp, unsigned int cmd,
				      unsigned long args)
{
	long ret;
	unsigned long arg = args;

	/* Check for properly constructed commands */
	ret = check_ioctl_command(cmd);
	if (ret < 0)
		return ret;
	if (cmd == PVFS_DEV_MAP) {
		/*
		 * convert the arguments to what we expect internally
		 * in kernel space
		 */
		arg = translate_dev_map26(args, &ret);
		if (ret < 0) {
			gossip_err("Could not translate dev map\n");
			return ret;
		}
	}
	/* no other ioctl requires translation */
	return dispatch_ioctl_command(cmd, arg);
}

#endif /* CONFIG_COMPAT is in .config */

/*
 * The following two ioctl32 functions had been refactored into the above
 * CONFIG_COMPAT ifdef, but that was an over simplification that was
 * not noticed until we tried to compile on power pc...
 */
#if (defined(CONFIG_COMPAT) && !defined(HAVE_REGISTER_IOCTL32_CONVERSION)) || !defined(CONFIG_COMPAT)
static int pvfs2_ioctl32_init(void)
{
	return 0;
}

static void pvfs2_ioctl32_cleanup(void)
{
	return;
}
#endif

/* the assigned character device major number */
static int pvfs2_dev_major;

/*
 * Initialize pvfs2 device specific state:
 * Must be called at module load time only
 */
int pvfs2_dev_init(void)
{
	int ret;

	/* register the ioctl32 sub-system */
	ret = pvfs2_ioctl32_init();
	if (ret < 0)
		return ret;

	/* register pvfs2-req device  */
	pvfs2_dev_major = register_chrdev(0,
					  PVFS2_REQDEVICE_NAME,
					  &pvfs2_devreq_file_operations);
	if (pvfs2_dev_major < 0) {
		gossip_debug(GOSSIP_DEV_DEBUG,
			     "Failed to register /dev/%s (error %d)\n",
			     PVFS2_REQDEVICE_NAME, pvfs2_dev_major);
		pvfs2_ioctl32_cleanup();
		return pvfs2_dev_major;
	}

	gossip_debug(GOSSIP_DEV_DEBUG,
		     "*** /dev/%s character device registered ***\n",
		     PVFS2_REQDEVICE_NAME);
	gossip_debug(GOSSIP_DEV_DEBUG, "'mknod /dev/%s c %d 0'.\n",
		     PVFS2_REQDEVICE_NAME, pvfs2_dev_major);
	return 0;
}

void pvfs2_dev_cleanup(void)
{
	unregister_chrdev(pvfs2_dev_major, PVFS2_REQDEVICE_NAME);
	gossip_debug(GOSSIP_DEV_DEBUG,
		     "*** /dev/%s character device unregistered ***\n",
		     PVFS2_REQDEVICE_NAME);
	/* unregister the ioctl32 sub-system */
	pvfs2_ioctl32_cleanup();
}

static unsigned int pvfs2_devreq_poll(struct file *file,
				      struct poll_table_struct *poll_table)
{
	int poll_revent_mask = 0;

	if (open_access_count == 1) {
		poll_wait(file, &pvfs2_request_list_waitq, poll_table);

		spin_lock(&pvfs2_request_list_lock);
		if (!list_empty(&pvfs2_request_list))
			poll_revent_mask |= POLL_IN;
		spin_unlock(&pvfs2_request_list_lock);
	}
	return poll_revent_mask;
}

const struct file_operations pvfs2_devreq_file_operations = {
	.owner = THIS_MODULE,
	.read = pvfs2_devreq_read,
	.write_iter = pvfs2_devreq_write_iter,
	.open = pvfs2_devreq_open,
	.release = pvfs2_devreq_release,
	.unlocked_ioctl = pvfs2_devreq_ioctl,

#ifdef CONFIG_COMPAT		/* CONFIG_COMPAT is in .config */
	.compat_ioctl = pvfs2_devreq_compat_ioctl,
#endif
	.poll = pvfs2_devreq_poll
};