callback_proc.c 14.2 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9
/*
 * linux/fs/nfs/callback_proc.c
 *
 * Copyright (C) 2004 Trond Myklebust
 *
 * NFSv4 callback procedures
 */
#include <linux/nfs4.h>
#include <linux/nfs_fs.h>
10
#include <linux/slab.h>
11
#include <linux/rcupdate.h>
12
#include "nfs4_fs.h"
L
Linus Torvalds 已提交
13 14
#include "callback.h"
#include "delegation.h"
15
#include "internal.h"
F
Fred Isaman 已提交
16
#include "pnfs.h"
17
#include "nfs4session.h"
18
#include "nfs4trace.h"
L
Linus Torvalds 已提交
19

20
#ifdef NFS_DEBUG
L
Linus Torvalds 已提交
21
#define NFSDBG_FACILITY NFSDBG_CALLBACK
22
#endif
23 24 25 26

__be32 nfs4_callback_getattr(struct cb_getattrargs *args,
			     struct cb_getattrres *res,
			     struct cb_process_state *cps)
L
Linus Torvalds 已提交
27 28 29 30
{
	struct nfs_delegation *delegation;
	struct nfs_inode *nfsi;
	struct inode *inode;
31

32 33 34 35
	res->status = htonl(NFS4ERR_OP_NOT_IN_SESSION);
	if (!cps->clp) /* Always set for v4.0. Set in cb_sequence for v4.1 */
		goto out;

L
Linus Torvalds 已提交
36 37
	res->bitmap[0] = res->bitmap[1] = 0;
	res->status = htonl(NFS4ERR_BADHANDLE);
38

39
	dprintk_rcu("NFS: GETATTR callback request from %s\n",
40
		rpc_peeraddr2str(cps->clp->cl_rpcclient, RPC_DISPLAY_ADDR));
41

42
	inode = nfs_delegation_find_inode(cps->clp, &args->fh);
L
Linus Torvalds 已提交
43
	if (inode == NULL)
44
		goto out;
L
Linus Torvalds 已提交
45
	nfsi = NFS_I(inode);
46 47
	rcu_read_lock();
	delegation = rcu_dereference(nfsi->delegation);
L
Linus Torvalds 已提交
48 49 50
	if (delegation == NULL || (delegation->type & FMODE_WRITE) == 0)
		goto out_iput;
	res->size = i_size_read(inode);
51 52 53
	res->change_attr = delegation->change_attr;
	if (nfsi->npages != 0)
		res->change_attr++;
L
Linus Torvalds 已提交
54 55 56 57 58 59 60 61
	res->ctime = inode->i_ctime;
	res->mtime = inode->i_mtime;
	res->bitmap[0] = (FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE) &
		args->bitmap[0];
	res->bitmap[1] = (FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY) &
		args->bitmap[1];
	res->status = 0;
out_iput:
62
	rcu_read_unlock();
L
Linus Torvalds 已提交
63 64
	iput(inode);
out:
65
	dprintk("%s: exit with status = %d\n", __func__, ntohl(res->status));
L
Linus Torvalds 已提交
66 67 68
	return res->status;
}

69 70
__be32 nfs4_callback_recall(struct cb_recallargs *args, void *dummy,
			    struct cb_process_state *cps)
L
Linus Torvalds 已提交
71 72
{
	struct inode *inode;
73
	__be32 res;
L
Linus Torvalds 已提交
74
	
75 76
	res = htonl(NFS4ERR_OP_NOT_IN_SESSION);
	if (!cps->clp) /* Always set for v4.0. Set in cb_sequence for v4.1 */
L
Linus Torvalds 已提交
77
		goto out;
78

79
	dprintk_rcu("NFS: RECALL callback request from %s\n",
80 81 82 83 84 85 86 87 88 89 90 91
		rpc_peeraddr2str(cps->clp->cl_rpcclient, RPC_DISPLAY_ADDR));

	res = htonl(NFS4ERR_BADHANDLE);
	inode = nfs_delegation_find_inode(cps->clp, &args->fh);
	if (inode == NULL)
		goto out;
	/* Set up a helper thread to actually return the delegation */
	switch (nfs_async_inode_return_delegation(inode, &args->stateid)) {
	case 0:
		res = 0;
		break;
	case -ENOENT:
92
		res = htonl(NFS4ERR_BAD_STATEID);
93 94 95 96
		break;
	default:
		res = htonl(NFS4ERR_RESOURCE);
	}
97
	trace_nfs4_recall_delegation(inode, -ntohl(res));
98
	iput(inode);
L
Linus Torvalds 已提交
99
out:
100
	dprintk("%s: exit with status = %d\n", __func__, ntohl(res));
L
Linus Torvalds 已提交
101 102
	return res;
}
103 104 105

#if defined(CONFIG_NFS_V4_1)

106 107 108 109 110 111 112 113 114 115
/*
 * Lookup a layout by filehandle.
 *
 * Note: gets a refcount on the layout hdr and on its respective inode.
 * Caller must put the layout hdr and the inode.
 *
 * TODO: keep track of all layouts (and delegations) in a hash table
 * hashed by filehandle.
 */
static struct pnfs_layout_hdr * get_layout_by_fh_locked(struct nfs_client *clp, struct nfs_fh *fh)
F
Fred Isaman 已提交
116
{
117
	struct nfs_server *server;
F
Fred Isaman 已提交
118
	struct inode *ino;
119
	struct pnfs_layout_hdr *lo;
F
Fred Isaman 已提交
120

121 122
	list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
		list_for_each_entry(lo, &server->layouts, plh_layouts) {
123
			if (nfs_compare_fh(fh, &NFS_I(lo->plh_inode)->fh))
124 125 126 127
				continue;
			ino = igrab(lo->plh_inode);
			if (!ino)
				continue;
128 129 130 131 132 133 134
			spin_lock(&ino->i_lock);
			/* Is this layout in the process of being freed? */
			if (NFS_I(ino)->layout != lo) {
				spin_unlock(&ino->i_lock);
				iput(ino);
				continue;
			}
135
			pnfs_get_layout_hdr(lo);
136
			spin_unlock(&ino->i_lock);
137
			return lo;
138
		}
F
Fred Isaman 已提交
139
	}
140 141 142 143 144 145 146 147 148 149 150

	return NULL;
}

static struct pnfs_layout_hdr * get_layout_by_fh(struct nfs_client *clp, struct nfs_fh *fh)
{
	struct pnfs_layout_hdr *lo;

	spin_lock(&clp->cl_lock);
	rcu_read_lock();
	lo = get_layout_by_fh_locked(clp, fh);
151
	rcu_read_unlock();
F
Fred Isaman 已提交
152
	spin_unlock(&clp->cl_lock);
153

154 155 156 157 158 159 160 161 162 163 164 165 166
	return lo;
}

static u32 initiate_file_draining(struct nfs_client *clp,
				  struct cb_layoutrecallargs *args)
{
	struct inode *ino;
	struct pnfs_layout_hdr *lo;
	u32 rv = NFS4ERR_NOMATCHING_LAYOUT;
	LIST_HEAD(free_me_list);

	lo = get_layout_by_fh(clp, &args->cbl_fh);
	if (!lo)
F
Fred Isaman 已提交
167 168
		return NFS4ERR_NOMATCHING_LAYOUT;

169
	ino = lo->plh_inode;
F
Fred Isaman 已提交
170 171
	spin_lock(&ino->i_lock);
	if (test_bit(NFS_LAYOUT_BULK_RECALL, &lo->plh_flags) ||
172
	    pnfs_mark_matching_lsegs_invalid(lo, &free_me_list,
173
					&args->cbl_range))
F
Fred Isaman 已提交
174 175 176 177 178 179
		rv = NFS4ERR_DELAY;
	else
		rv = NFS4ERR_NOMATCHING_LAYOUT;
	pnfs_set_layout_stateid(lo, &args->cbl_stateid, true);
	spin_unlock(&ino->i_lock);
	pnfs_free_lseg_list(&free_me_list);
180
	pnfs_put_layout_hdr(lo);
F
Fred Isaman 已提交
181 182 183 184 185 186 187
	iput(ino);
	return rv;
}

static u32 initiate_bulk_draining(struct nfs_client *clp,
				  struct cb_layoutrecallargs *args)
{
188
	int stat;
189

190 191 192 193 194 195 196
	if (args->cbl_recall_type == RETURN_FSID)
		stat = pnfs_destroy_layouts_byfsid(clp, &args->cbl_fsid, true);
	else
		stat = pnfs_destroy_layouts_byclid(clp, true);
	if (stat != 0)
		return NFS4ERR_DELAY;
	return NFS4ERR_NOMATCHING_LAYOUT;
F
Fred Isaman 已提交
197 198 199 200 201
}

static u32 do_callback_layoutrecall(struct nfs_client *clp,
				    struct cb_layoutrecallargs *args)
{
202
	u32 res;
F
Fred Isaman 已提交
203 204 205 206 207 208 209 210 211 212 213

	dprintk("%s enter, type=%i\n", __func__, args->cbl_recall_type);
	if (args->cbl_recall_type == RETURN_FILE)
		res = initiate_file_draining(clp, args);
	else
		res = initiate_bulk_draining(clp, args);
	dprintk("%s returning %i\n", __func__, res);
	return res;

}

F
Fred Isaman 已提交
214 215 216
__be32 nfs4_callback_layoutrecall(struct cb_layoutrecallargs *args,
				  void *dummy, struct cb_process_state *cps)
{
F
Fred Isaman 已提交
217 218 219 220 221 222 223 224 225 226 227
	u32 res;

	dprintk("%s: -->\n", __func__);

	if (cps->clp)
		res = do_callback_layoutrecall(cps->clp, args);
	else
		res = NFS4ERR_OP_NOT_IN_SESSION;

	dprintk("%s: exit with status = %d\n", __func__, res);
	return cpu_to_be32(res);
F
Fred Isaman 已提交
228 229
}

230 231 232 233 234 235 236 237 238 239 240
static void pnfs_recall_all_layouts(struct nfs_client *clp)
{
	struct cb_layoutrecallargs args;

	/* Pretend we got a CB_LAYOUTRECALL(ALL) */
	memset(&args, 0, sizeof(args));
	args.cbl_recall_type = RETURN_ALL;
	/* FIXME we ignore errors, what should we do? */
	do_callback_layoutrecall(clp, &args);
}

M
Marc Eshel 已提交
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
__be32 nfs4_callback_devicenotify(struct cb_devicenotifyargs *args,
				  void *dummy, struct cb_process_state *cps)
{
	int i;
	__be32 res = 0;
	struct nfs_client *clp = cps->clp;
	struct nfs_server *server = NULL;

	dprintk("%s: -->\n", __func__);

	if (!clp) {
		res = cpu_to_be32(NFS4ERR_OP_NOT_IN_SESSION);
		goto out;
	}

	for (i = 0; i < args->ndevs; i++) {
		struct cb_devicenotifyitem *dev = &args->devs[i];

		if (!server ||
		    server->pnfs_curr_ld->id != dev->cbd_layout_type) {
			rcu_read_lock();
			list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link)
				if (server->pnfs_curr_ld &&
				    server->pnfs_curr_ld->id == dev->cbd_layout_type) {
					rcu_read_unlock();
					goto found;
				}
			rcu_read_unlock();
			dprintk("%s: layout type %u not found\n",
				__func__, dev->cbd_layout_type);
			continue;
		}

	found:
		if (dev->cbd_notify_type == NOTIFY_DEVICEID4_CHANGE)
			dprintk("%s: NOTIFY_DEVICEID4_CHANGE not supported, "
				"deleting instead\n", __func__);
278
		nfs4_delete_deviceid(server->pnfs_curr_ld, clp, &dev->cbd_dev_id);
M
Marc Eshel 已提交
279 280 281 282 283 284 285 286 287
	}

out:
	kfree(args->devs);
	dprintk("%s: exit with status = %u\n",
		__func__, be32_to_cpu(res));
	return res;
}

288 289 290 291 292
/*
 * Validate the sequenceID sent by the server.
 * Return success if the sequenceID is one more than what we last saw on
 * this slot, accounting for wraparound.  Increments the slot's sequence.
 *
293 294
 * We don't yet implement a duplicate request cache, instead we set the
 * back channel ca_maxresponsesize_cached to zero. This is OK for now
295 296 297 298 299 300
 * since we only currently implement idempotent callbacks anyway.
 *
 * We have a single slot backchannel at this time, so we don't bother
 * checking the used_slots bit array on the table.  The lower layer guarantees
 * a single outstanding callback request at a time.
 */
301
static __be32
302
validate_seqid(struct nfs4_slot_table *tbl, struct cb_sequenceargs * args)
303 304 305 306
{
	struct nfs4_slot *slot;

	dprintk("%s enter. slotid %d seqid %d\n",
307
		__func__, args->csa_slotid, args->csa_sequenceid);
308

309
	if (args->csa_slotid >= NFS41_BC_MAX_CALLBACKS)
310 311
		return htonl(NFS4ERR_BADSLOT);

312
	slot = tbl->slots + args->csa_slotid;
313 314 315
	dprintk("%s slot table seqid: %d\n", __func__, slot->seq_nr);

	/* Normal */
316
	if (likely(args->csa_sequenceid == slot->seq_nr + 1)) {
317
		slot->seq_nr++;
318
		goto out_ok;
319 320 321
	}

	/* Replay */
322
	if (args->csa_sequenceid == slot->seq_nr) {
323
		dprintk("%s seqid %d is a replay\n",
324
			__func__, args->csa_sequenceid);
325 326 327 328 329 330 331
		/* Signal process_op to set this error on next op */
		if (args->csa_cachethis == 0)
			return htonl(NFS4ERR_RETRY_UNCACHED_REP);

		/* The ca_maxresponsesize_cached is 0 with no DRC */
		else if (args->csa_cachethis == 1)
			return htonl(NFS4ERR_REP_TOO_BIG_TO_CACHE);
332 333 334
	}

	/* Wraparound */
335
	if (args->csa_sequenceid == 1 && (slot->seq_nr + 1) == 0) {
336
		slot->seq_nr = 1;
337
		goto out_ok;
338 339 340 341
	}

	/* Misordered request */
	return htonl(NFS4ERR_SEQ_MISORDERED);
342 343 344
out_ok:
	tbl->highest_used_slotid = args->csa_slotid;
	return htonl(NFS4_OK);
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
/*
 * For each referring call triple, check the session's slot table for
 * a match.  If the slot is in use and the sequence numbers match, the
 * client is still waiting for a response to the original request.
 */
static bool referring_call_exists(struct nfs_client *clp,
				  uint32_t nrclists,
				  struct referring_call_list *rclists)
{
	bool status = 0;
	int i, j;
	struct nfs4_session *session;
	struct nfs4_slot_table *tbl;
	struct referring_call_list *rclist;
	struct referring_call *ref;

	/*
	 * XXX When client trunking is implemented, this becomes
	 * a session lookup from within the loop
	 */
	session = clp->cl_session;
	tbl = &session->fc_slot_table;

	for (i = 0; i < nrclists; i++) {
		rclist = &rclists[i];
		if (memcmp(session->sess_id.data,
			   rclist->rcl_sessionid.data,
			   NFS4_MAX_SESSIONID_LEN) != 0)
			continue;

		for (j = 0; j < rclist->rcl_nrefcalls; j++) {
			ref = &rclist->rcl_refcalls[j];

			dprintk("%s: sessionid %x:%x:%x:%x sequenceid %u "
				"slotid %u\n", __func__,
				((u32 *)&rclist->rcl_sessionid.data)[0],
				((u32 *)&rclist->rcl_sessionid.data)[1],
				((u32 *)&rclist->rcl_sessionid.data)[2],
				((u32 *)&rclist->rcl_sessionid.data)[3],
				ref->rc_sequenceid, ref->rc_slotid);

			spin_lock(&tbl->slot_tbl_lock);
			status = (test_bit(ref->rc_slotid, tbl->used_slots) &&
				  tbl->slots[ref->rc_slotid].seq_nr ==
					ref->rc_sequenceid);
			spin_unlock(&tbl->slot_tbl_lock);
			if (status)
				goto out;
		}
	}

out:
	return status;
}

402
__be32 nfs4_callback_sequence(struct cb_sequenceargs *args,
403 404
			      struct cb_sequenceres *res,
			      struct cb_process_state *cps)
405
{
406
	struct nfs4_slot_table *tbl;
407
	struct nfs_client *clp;
408
	int i;
409
	__be32 status = htonl(NFS4ERR_BADSESSION);
410

411 412
	clp = nfs4_find_client_sessionid(cps->net, args->csa_addr,
					 &args->csa_sessionid, cps->minorversion);
413 414 415
	if (clp == NULL)
		goto out;

416 417 418
	tbl = &clp->cl_session->bc_slot_table;

	spin_lock(&tbl->slot_tbl_lock);
419
	/* state manager is resetting the session */
420
	if (test_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state)) {
421 422
		spin_unlock(&tbl->slot_tbl_lock);
		status = htonl(NFS4ERR_DELAY);
423 424 425 426 427
		/* Return NFS4ERR_BADSESSION if we're draining the session
		 * in order to reset it.
		 */
		if (test_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state))
			status = htonl(NFS4ERR_BADSESSION);
428 429 430
		goto out;
	}

431
	status = validate_seqid(&clp->cl_session->bc_slot_table, args);
432
	spin_unlock(&tbl->slot_tbl_lock);
433
	if (status)
434
		goto out;
435

436 437
	cps->slotid = args->csa_slotid;

438 439 440 441 442 443 444
	/*
	 * Check for pending referring calls.  If a match is found, a
	 * related callback was received before the response to the original
	 * call.
	 */
	if (referring_call_exists(clp, args->csa_nrclists, args->csa_rclists)) {
		status = htonl(NFS4ERR_DELAY);
445
		goto out;
446 447
	}

448 449 450 451 452 453 454
	memcpy(&res->csr_sessionid, &args->csa_sessionid,
	       sizeof(res->csr_sessionid));
	res->csr_sequenceid = args->csa_sequenceid;
	res->csr_slotid = args->csa_slotid;
	res->csr_highestslotid = NFS41_BC_MAX_CALLBACKS - 1;
	res->csr_target_highestslotid = NFS41_BC_MAX_CALLBACKS - 1;

455
out:
456
	cps->clp = clp; /* put in nfs4_callback_compound */
457 458 459 460
	for (i = 0; i < args->csa_nrclists; i++)
		kfree(args->csa_rclists[i].rcl_refcalls);
	kfree(args->csa_rclists);

461 462 463 464
	if (status == htonl(NFS4ERR_RETRY_UNCACHED_REP)) {
		cps->drc_status = status;
		status = 0;
	} else
465
		res->csr_status = status;
466

467 468 469
	dprintk("%s: exit with status = %d res->csr_status %d\n", __func__,
		ntohl(status), ntohl(res->csr_status));
	return status;
470 471
}

472 473 474 475 476 477
static bool
validate_bitmap_values(unsigned long mask)
{
	return (mask & ~RCA4_TYPE_MASK_ALL) == 0;
}

478 479
__be32 nfs4_callback_recallany(struct cb_recallanyargs *args, void *dummy,
			       struct cb_process_state *cps)
480
{
481
	__be32 status;
482 483
	fmode_t flags = 0;

484
	status = cpu_to_be32(NFS4ERR_OP_NOT_IN_SESSION);
485
	if (!cps->clp) /* set in cb_sequence */
486 487
		goto out;

488
	dprintk_rcu("NFS: RECALL_ANY callback request from %s\n",
489
		rpc_peeraddr2str(cps->clp->cl_rpcclient, RPC_DISPLAY_ADDR));
490

491 492 493 494 495
	status = cpu_to_be32(NFS4ERR_INVAL);
	if (!validate_bitmap_values(args->craa_type_mask))
		goto out;

	status = cpu_to_be32(NFS4_OK);
496 497 498 499 500 501
	if (test_bit(RCA4_TYPE_MASK_RDATA_DLG, (const unsigned long *)
		     &args->craa_type_mask))
		flags = FMODE_READ;
	if (test_bit(RCA4_TYPE_MASK_WDATA_DLG, (const unsigned long *)
		     &args->craa_type_mask))
		flags |= FMODE_WRITE;
502 503 504
	if (test_bit(RCA4_TYPE_MASK_FILE_LAYOUT, (const unsigned long *)
		     &args->craa_type_mask))
		pnfs_recall_all_layouts(cps->clp);
505
	if (flags)
506
		nfs_expire_unused_delegation_types(cps->clp, flags);
507 508 509 510
out:
	dprintk("%s: exit with status = %d\n", __func__, ntohl(status));
	return status;
}
A
Andy Adamson 已提交
511 512

/* Reduce the fore channel's max_slots to the target value */
513 514
__be32 nfs4_callback_recallslot(struct cb_recallslotargs *args, void *dummy,
				struct cb_process_state *cps)
A
Andy Adamson 已提交
515 516
{
	struct nfs4_slot_table *fc_tbl;
517
	__be32 status;
A
Andy Adamson 已提交
518 519

	status = htonl(NFS4ERR_OP_NOT_IN_SESSION);
520
	if (!cps->clp) /* set in cb_sequence */
A
Andy Adamson 已提交
521 522
		goto out;

523
	dprintk_rcu("NFS: CB_RECALL_SLOT request from %s target highest slotid %d\n",
524
		rpc_peeraddr2str(cps->clp->cl_rpcclient, RPC_DISPLAY_ADDR),
525
		args->crsa_target_highest_slotid);
A
Andy Adamson 已提交
526

527
	fc_tbl = &cps->clp->cl_session->fc_slot_table;
A
Andy Adamson 已提交
528

529
	status = htonl(NFS4_OK);
A
Andy Adamson 已提交
530

531
	nfs41_set_target_slotid(fc_tbl, args->crsa_target_highest_slotid);
532
	nfs41_server_notify_target_slotid_update(cps->clp);
A
Andy Adamson 已提交
533 534 535 536
out:
	dprintk("%s: exit with status = %d\n", __func__, ntohl(status));
	return status;
}
537
#endif /* CONFIG_NFS_V4_1 */