Linux Audio

Check our new training course

Loading...
v4.10.11
 
   1/*
   2 * linux/fs/ocfs2/ioctl.c
   3 *
   4 * Copyright (C) 2006 Herbert Poetzl
   5 * adapted from Remy Card's ext2/ioctl.c
   6 */
   7
   8#include <linux/fs.h>
   9#include <linux/mount.h>
  10#include <linux/blkdev.h>
  11#include <linux/compat.h>
  12
  13#include <cluster/masklog.h>
  14
  15#include "ocfs2.h"
  16#include "alloc.h"
  17#include "dlmglue.h"
  18#include "file.h"
  19#include "inode.h"
  20#include "journal.h"
  21
  22#include "ocfs2_fs.h"
  23#include "ioctl.h"
  24#include "resize.h"
  25#include "refcounttree.h"
  26#include "sysfile.h"
  27#include "dir.h"
  28#include "buffer_head_io.h"
  29#include "suballoc.h"
  30#include "move_extents.h"
  31
  32#define o2info_from_user(a, b)	\
  33		copy_from_user(&(a), (b), sizeof(a))
  34#define o2info_to_user(a, b)	\
  35		copy_to_user((typeof(a) __user *)b, &(a), sizeof(a))
  36
  37/*
  38 * This is just a best-effort to tell userspace that this request
  39 * caused the error.
  40 */
  41static inline void o2info_set_request_error(struct ocfs2_info_request *kreq,
  42					struct ocfs2_info_request __user *req)
  43{
  44	kreq->ir_flags |= OCFS2_INFO_FL_ERROR;
  45	(void)put_user(kreq->ir_flags, (__u32 __user *)&(req->ir_flags));
  46}
  47
  48static inline void o2info_set_request_filled(struct ocfs2_info_request *req)
  49{
  50	req->ir_flags |= OCFS2_INFO_FL_FILLED;
  51}
  52
  53static inline void o2info_clear_request_filled(struct ocfs2_info_request *req)
  54{
  55	req->ir_flags &= ~OCFS2_INFO_FL_FILLED;
  56}
  57
  58static inline int o2info_coherent(struct ocfs2_info_request *req)
  59{
  60	return (!(req->ir_flags & OCFS2_INFO_FL_NON_COHERENT));
  61}
  62
  63static int ocfs2_get_inode_attr(struct inode *inode, unsigned *flags)
  64{
  65	int status;
  66
  67	status = ocfs2_inode_lock(inode, NULL, 0);
  68	if (status < 0) {
  69		mlog_errno(status);
  70		return status;
  71	}
  72	ocfs2_get_inode_flags(OCFS2_I(inode));
  73	*flags = OCFS2_I(inode)->ip_attr;
  74	ocfs2_inode_unlock(inode, 0);
  75
  76	return status;
  77}
  78
  79static int ocfs2_set_inode_attr(struct inode *inode, unsigned flags,
  80				unsigned mask)
  81{
  82	struct ocfs2_inode_info *ocfs2_inode = OCFS2_I(inode);
  83	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
  84	handle_t *handle = NULL;
  85	struct buffer_head *bh = NULL;
  86	unsigned oldflags;
  87	int status;
  88
  89	inode_lock(inode);
  90
  91	status = ocfs2_inode_lock(inode, &bh, 1);
  92	if (status < 0) {
  93		mlog_errno(status);
  94		goto bail;
  95	}
  96
  97	status = -EACCES;
  98	if (!inode_owner_or_capable(inode))
  99		goto bail_unlock;
 100
 101	if (!S_ISDIR(inode->i_mode))
 102		flags &= ~OCFS2_DIRSYNC_FL;
 103
 104	oldflags = ocfs2_inode->ip_attr;
 105	flags = flags & mask;
 106	flags |= oldflags & ~mask;
 107
 108	/*
 109	 * The IMMUTABLE and APPEND_ONLY flags can only be changed by
 110	 * the relevant capability.
 111	 */
 112	status = -EPERM;
 113	if ((oldflags & OCFS2_IMMUTABLE_FL) || ((flags ^ oldflags) &
 114		(OCFS2_APPEND_FL | OCFS2_IMMUTABLE_FL))) {
 115		if (!capable(CAP_LINUX_IMMUTABLE))
 116			goto bail_unlock;
 117	}
 118
 119	handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
 120	if (IS_ERR(handle)) {
 121		status = PTR_ERR(handle);
 122		mlog_errno(status);
 123		goto bail_unlock;
 124	}
 125
 126	ocfs2_inode->ip_attr = flags;
 127	ocfs2_set_inode_flags(inode);
 128
 129	status = ocfs2_mark_inode_dirty(handle, inode, bh);
 130	if (status < 0)
 131		mlog_errno(status);
 132
 133	ocfs2_commit_trans(osb, handle);
 134
 135bail_unlock:
 136	ocfs2_inode_unlock(inode, 1);
 137bail:
 138	inode_unlock(inode);
 139
 140	brelse(bh);
 141
 142	return status;
 143}
 144
 145static int ocfs2_info_handle_blocksize(struct inode *inode,
 146				       struct ocfs2_info_request __user *req)
 147{
 148	struct ocfs2_info_blocksize oib;
 149
 150	if (o2info_from_user(oib, req))
 151		return -EFAULT;
 152
 153	oib.ib_blocksize = inode->i_sb->s_blocksize;
 154
 155	o2info_set_request_filled(&oib.ib_req);
 156
 157	if (o2info_to_user(oib, req))
 158		return -EFAULT;
 159
 160	return 0;
 161}
 162
 163static int ocfs2_info_handle_clustersize(struct inode *inode,
 164					 struct ocfs2_info_request __user *req)
 165{
 166	struct ocfs2_info_clustersize oic;
 167	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
 168
 169	if (o2info_from_user(oic, req))
 170		return -EFAULT;
 171
 172	oic.ic_clustersize = osb->s_clustersize;
 173
 174	o2info_set_request_filled(&oic.ic_req);
 175
 176	if (o2info_to_user(oic, req))
 177		return -EFAULT;
 178
 179	return 0;
 180}
 181
 182static int ocfs2_info_handle_maxslots(struct inode *inode,
 183				      struct ocfs2_info_request __user *req)
 184{
 185	struct ocfs2_info_maxslots oim;
 186	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
 187
 188	if (o2info_from_user(oim, req))
 189		return -EFAULT;
 190
 191	oim.im_max_slots = osb->max_slots;
 192
 193	o2info_set_request_filled(&oim.im_req);
 194
 195	if (o2info_to_user(oim, req))
 196		return -EFAULT;
 197
 198	return 0;
 199}
 200
 201static int ocfs2_info_handle_label(struct inode *inode,
 202				   struct ocfs2_info_request __user *req)
 203{
 204	struct ocfs2_info_label oil;
 205	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
 206
 207	if (o2info_from_user(oil, req))
 208		return -EFAULT;
 209
 210	memcpy(oil.il_label, osb->vol_label, OCFS2_MAX_VOL_LABEL_LEN);
 211
 212	o2info_set_request_filled(&oil.il_req);
 213
 214	if (o2info_to_user(oil, req))
 215		return -EFAULT;
 216
 217	return 0;
 218}
 219
 220static int ocfs2_info_handle_uuid(struct inode *inode,
 221				  struct ocfs2_info_request __user *req)
 222{
 223	struct ocfs2_info_uuid oiu;
 224	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
 225
 226	if (o2info_from_user(oiu, req))
 227		return -EFAULT;
 228
 229	memcpy(oiu.iu_uuid_str, osb->uuid_str, OCFS2_TEXT_UUID_LEN + 1);
 230
 231	o2info_set_request_filled(&oiu.iu_req);
 232
 233	if (o2info_to_user(oiu, req))
 234		return -EFAULT;
 235
 236	return 0;
 237}
 238
 239static int ocfs2_info_handle_fs_features(struct inode *inode,
 240					 struct ocfs2_info_request __user *req)
 241{
 242	struct ocfs2_info_fs_features oif;
 243	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
 244
 245	if (o2info_from_user(oif, req))
 246		return -EFAULT;
 247
 248	oif.if_compat_features = osb->s_feature_compat;
 249	oif.if_incompat_features = osb->s_feature_incompat;
 250	oif.if_ro_compat_features = osb->s_feature_ro_compat;
 251
 252	o2info_set_request_filled(&oif.if_req);
 253
 254	if (o2info_to_user(oif, req))
 255		return -EFAULT;
 256
 257	return 0;
 258}
 259
 260static int ocfs2_info_handle_journal_size(struct inode *inode,
 261					  struct ocfs2_info_request __user *req)
 262{
 263	struct ocfs2_info_journal_size oij;
 264	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
 265
 266	if (o2info_from_user(oij, req))
 267		return -EFAULT;
 268
 269	oij.ij_journal_size = i_size_read(osb->journal->j_inode);
 270
 271	o2info_set_request_filled(&oij.ij_req);
 272
 273	if (o2info_to_user(oij, req))
 274		return -EFAULT;
 275
 276	return 0;
 277}
 278
 279static int ocfs2_info_scan_inode_alloc(struct ocfs2_super *osb,
 280				       struct inode *inode_alloc, u64 blkno,
 281				       struct ocfs2_info_freeinode *fi,
 282				       u32 slot)
 283{
 284	int status = 0, unlock = 0;
 285
 286	struct buffer_head *bh = NULL;
 287	struct ocfs2_dinode *dinode_alloc = NULL;
 288
 289	if (inode_alloc)
 290		inode_lock(inode_alloc);
 291
 292	if (o2info_coherent(&fi->ifi_req)) {
 293		status = ocfs2_inode_lock(inode_alloc, &bh, 0);
 294		if (status < 0) {
 295			mlog_errno(status);
 296			goto bail;
 297		}
 298		unlock = 1;
 299	} else {
 300		status = ocfs2_read_blocks_sync(osb, blkno, 1, &bh);
 301		if (status < 0) {
 302			mlog_errno(status);
 303			goto bail;
 304		}
 305	}
 306
 307	dinode_alloc = (struct ocfs2_dinode *)bh->b_data;
 308
 309	fi->ifi_stat[slot].lfi_total =
 310		le32_to_cpu(dinode_alloc->id1.bitmap1.i_total);
 311	fi->ifi_stat[slot].lfi_free =
 312		le32_to_cpu(dinode_alloc->id1.bitmap1.i_total) -
 313		le32_to_cpu(dinode_alloc->id1.bitmap1.i_used);
 314
 315bail:
 316	if (unlock)
 317		ocfs2_inode_unlock(inode_alloc, 0);
 318
 319	if (inode_alloc)
 320		inode_unlock(inode_alloc);
 321
 322	brelse(bh);
 323
 324	return status;
 325}
 326
 327static int ocfs2_info_handle_freeinode(struct inode *inode,
 328				       struct ocfs2_info_request __user *req)
 329{
 330	u32 i;
 331	u64 blkno = -1;
 332	char namebuf[40];
 333	int status, type = INODE_ALLOC_SYSTEM_INODE;
 334	struct ocfs2_info_freeinode *oifi = NULL;
 335	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
 336	struct inode *inode_alloc = NULL;
 337
 338	oifi = kzalloc(sizeof(struct ocfs2_info_freeinode), GFP_KERNEL);
 339	if (!oifi) {
 340		status = -ENOMEM;
 341		mlog_errno(status);
 342		goto out_err;
 343	}
 344
 345	if (o2info_from_user(*oifi, req)) {
 346		status = -EFAULT;
 347		goto out_free;
 348	}
 349
 350	oifi->ifi_slotnum = osb->max_slots;
 351
 352	for (i = 0; i < oifi->ifi_slotnum; i++) {
 353		if (o2info_coherent(&oifi->ifi_req)) {
 354			inode_alloc = ocfs2_get_system_file_inode(osb, type, i);
 355			if (!inode_alloc) {
 356				mlog(ML_ERROR, "unable to get alloc inode in "
 357				    "slot %u\n", i);
 358				status = -EIO;
 359				goto bail;
 360			}
 361		} else {
 362			ocfs2_sprintf_system_inode_name(namebuf,
 363							sizeof(namebuf),
 364							type, i);
 365			status = ocfs2_lookup_ino_from_name(osb->sys_root_inode,
 366							    namebuf,
 367							    strlen(namebuf),
 368							    &blkno);
 369			if (status < 0) {
 370				status = -ENOENT;
 371				goto bail;
 372			}
 373		}
 374
 375		status = ocfs2_info_scan_inode_alloc(osb, inode_alloc, blkno, oifi, i);
 376
 377		iput(inode_alloc);
 378		inode_alloc = NULL;
 379
 380		if (status < 0)
 381			goto bail;
 382	}
 383
 384	o2info_set_request_filled(&oifi->ifi_req);
 385
 386	if (o2info_to_user(*oifi, req)) {
 387		status = -EFAULT;
 388		goto out_free;
 389	}
 390
 391	status = 0;
 392bail:
 393	if (status)
 394		o2info_set_request_error(&oifi->ifi_req, req);
 395out_free:
 396	kfree(oifi);
 397out_err:
 398	return status;
 399}
 400
 401static void o2ffg_update_histogram(struct ocfs2_info_free_chunk_list *hist,
 402				   unsigned int chunksize)
 403{
 404	int index;
 405
 406	index = __ilog2_u32(chunksize);
 407	if (index >= OCFS2_INFO_MAX_HIST)
 408		index = OCFS2_INFO_MAX_HIST - 1;
 409
 410	hist->fc_chunks[index]++;
 411	hist->fc_clusters[index] += chunksize;
 412}
 413
 414static void o2ffg_update_stats(struct ocfs2_info_freefrag_stats *stats,
 415			       unsigned int chunksize)
 416{
 417	if (chunksize > stats->ffs_max)
 418		stats->ffs_max = chunksize;
 419
 420	if (chunksize < stats->ffs_min)
 421		stats->ffs_min = chunksize;
 422
 423	stats->ffs_avg += chunksize;
 424	stats->ffs_free_chunks_real++;
 425}
 426
 427static void ocfs2_info_update_ffg(struct ocfs2_info_freefrag *ffg,
 428				  unsigned int chunksize)
 429{
 430	o2ffg_update_histogram(&(ffg->iff_ffs.ffs_fc_hist), chunksize);
 431	o2ffg_update_stats(&(ffg->iff_ffs), chunksize);
 432}
 433
 434static int ocfs2_info_freefrag_scan_chain(struct ocfs2_super *osb,
 435					  struct inode *gb_inode,
 436					  struct ocfs2_dinode *gb_dinode,
 437					  struct ocfs2_chain_rec *rec,
 438					  struct ocfs2_info_freefrag *ffg,
 439					  u32 chunks_in_group)
 440{
 441	int status = 0, used;
 442	u64 blkno;
 443
 444	struct buffer_head *bh = NULL;
 445	struct ocfs2_group_desc *bg = NULL;
 446
 447	unsigned int max_bits, num_clusters;
 448	unsigned int offset = 0, cluster, chunk;
 449	unsigned int chunk_free, last_chunksize = 0;
 450
 451	if (!le32_to_cpu(rec->c_free))
 452		goto bail;
 453
 454	do {
 455		if (!bg)
 456			blkno = le64_to_cpu(rec->c_blkno);
 457		else
 458			blkno = le64_to_cpu(bg->bg_next_group);
 459
 460		if (bh) {
 461			brelse(bh);
 462			bh = NULL;
 463		}
 464
 465		if (o2info_coherent(&ffg->iff_req))
 466			status = ocfs2_read_group_descriptor(gb_inode,
 467							     gb_dinode,
 468							     blkno, &bh);
 469		else
 470			status = ocfs2_read_blocks_sync(osb, blkno, 1, &bh);
 471
 472		if (status < 0) {
 473			mlog(ML_ERROR, "Can't read the group descriptor # "
 474			     "%llu from device.", (unsigned long long)blkno);
 475			status = -EIO;
 476			goto bail;
 477		}
 478
 479		bg = (struct ocfs2_group_desc *)bh->b_data;
 480
 481		if (!le16_to_cpu(bg->bg_free_bits_count))
 482			continue;
 483
 484		max_bits = le16_to_cpu(bg->bg_bits);
 485		offset = 0;
 486
 487		for (chunk = 0; chunk < chunks_in_group; chunk++) {
 488			/*
 489			 * last chunk may be not an entire one.
 490			 */
 491			if ((offset + ffg->iff_chunksize) > max_bits)
 492				num_clusters = max_bits - offset;
 493			else
 494				num_clusters = ffg->iff_chunksize;
 495
 496			chunk_free = 0;
 497			for (cluster = 0; cluster < num_clusters; cluster++) {
 498				used = ocfs2_test_bit(offset,
 499						(unsigned long *)bg->bg_bitmap);
 500				/*
 501				 * - chunk_free counts free clusters in #N chunk.
 502				 * - last_chunksize records the size(in) clusters
 503				 *   for the last real free chunk being counted.
 504				 */
 505				if (!used) {
 506					last_chunksize++;
 507					chunk_free++;
 508				}
 509
 510				if (used && last_chunksize) {
 511					ocfs2_info_update_ffg(ffg,
 512							      last_chunksize);
 513					last_chunksize = 0;
 514				}
 515
 516				offset++;
 517			}
 518
 519			if (chunk_free == ffg->iff_chunksize)
 520				ffg->iff_ffs.ffs_free_chunks++;
 521		}
 522
 523		/*
 524		 * need to update the info for last free chunk.
 525		 */
 526		if (last_chunksize)
 527			ocfs2_info_update_ffg(ffg, last_chunksize);
 528
 529	} while (le64_to_cpu(bg->bg_next_group));
 530
 531bail:
 532	brelse(bh);
 533
 534	return status;
 535}
 536
 537static int ocfs2_info_freefrag_scan_bitmap(struct ocfs2_super *osb,
 538					   struct inode *gb_inode, u64 blkno,
 539					   struct ocfs2_info_freefrag *ffg)
 540{
 541	u32 chunks_in_group;
 542	int status = 0, unlock = 0, i;
 543
 544	struct buffer_head *bh = NULL;
 545	struct ocfs2_chain_list *cl = NULL;
 546	struct ocfs2_chain_rec *rec = NULL;
 547	struct ocfs2_dinode *gb_dinode = NULL;
 548
 549	if (gb_inode)
 550		inode_lock(gb_inode);
 551
 552	if (o2info_coherent(&ffg->iff_req)) {
 553		status = ocfs2_inode_lock(gb_inode, &bh, 0);
 554		if (status < 0) {
 555			mlog_errno(status);
 556			goto bail;
 557		}
 558		unlock = 1;
 559	} else {
 560		status = ocfs2_read_blocks_sync(osb, blkno, 1, &bh);
 561		if (status < 0) {
 562			mlog_errno(status);
 563			goto bail;
 564		}
 565	}
 566
 567	gb_dinode = (struct ocfs2_dinode *)bh->b_data;
 568	cl = &(gb_dinode->id2.i_chain);
 569
 570	/*
 571	 * Chunksize(in) clusters from userspace should be
 572	 * less than clusters in a group.
 573	 */
 574	if (ffg->iff_chunksize > le16_to_cpu(cl->cl_cpg)) {
 575		status = -EINVAL;
 576		goto bail;
 577	}
 578
 579	memset(&ffg->iff_ffs, 0, sizeof(struct ocfs2_info_freefrag_stats));
 580
 581	ffg->iff_ffs.ffs_min = ~0U;
 582	ffg->iff_ffs.ffs_clusters =
 583			le32_to_cpu(gb_dinode->id1.bitmap1.i_total);
 584	ffg->iff_ffs.ffs_free_clusters = ffg->iff_ffs.ffs_clusters -
 585			le32_to_cpu(gb_dinode->id1.bitmap1.i_used);
 586
 587	chunks_in_group = le16_to_cpu(cl->cl_cpg) / ffg->iff_chunksize + 1;
 588
 589	for (i = 0; i < le16_to_cpu(cl->cl_next_free_rec); i++) {
 590		rec = &(cl->cl_recs[i]);
 591		status = ocfs2_info_freefrag_scan_chain(osb, gb_inode,
 592							gb_dinode,
 593							rec, ffg,
 594							chunks_in_group);
 595		if (status)
 596			goto bail;
 597	}
 598
 599	if (ffg->iff_ffs.ffs_free_chunks_real)
 600		ffg->iff_ffs.ffs_avg = (ffg->iff_ffs.ffs_avg /
 601					ffg->iff_ffs.ffs_free_chunks_real);
 602bail:
 603	if (unlock)
 604		ocfs2_inode_unlock(gb_inode, 0);
 605
 606	if (gb_inode)
 607		inode_unlock(gb_inode);
 608
 609	iput(gb_inode);
 610	brelse(bh);
 611
 612	return status;
 613}
 614
 615static int ocfs2_info_handle_freefrag(struct inode *inode,
 616				      struct ocfs2_info_request __user *req)
 617{
 618	u64 blkno = -1;
 619	char namebuf[40];
 620	int status, type = GLOBAL_BITMAP_SYSTEM_INODE;
 621
 622	struct ocfs2_info_freefrag *oiff;
 623	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
 624	struct inode *gb_inode = NULL;
 625
 626	oiff = kzalloc(sizeof(struct ocfs2_info_freefrag), GFP_KERNEL);
 627	if (!oiff) {
 628		status = -ENOMEM;
 629		mlog_errno(status);
 630		goto out_err;
 631	}
 632
 633	if (o2info_from_user(*oiff, req)) {
 634		status = -EFAULT;
 635		goto out_free;
 636	}
 637	/*
 638	 * chunksize from userspace should be power of 2.
 639	 */
 640	if ((oiff->iff_chunksize & (oiff->iff_chunksize - 1)) ||
 641	    (!oiff->iff_chunksize)) {
 642		status = -EINVAL;
 643		goto bail;
 644	}
 645
 646	if (o2info_coherent(&oiff->iff_req)) {
 647		gb_inode = ocfs2_get_system_file_inode(osb, type,
 648						       OCFS2_INVALID_SLOT);
 649		if (!gb_inode) {
 650			mlog(ML_ERROR, "unable to get global_bitmap inode\n");
 651			status = -EIO;
 652			goto bail;
 653		}
 654	} else {
 655		ocfs2_sprintf_system_inode_name(namebuf, sizeof(namebuf), type,
 656						OCFS2_INVALID_SLOT);
 657		status = ocfs2_lookup_ino_from_name(osb->sys_root_inode,
 658						    namebuf,
 659						    strlen(namebuf),
 660						    &blkno);
 661		if (status < 0) {
 662			status = -ENOENT;
 663			goto bail;
 664		}
 665	}
 666
 667	status = ocfs2_info_freefrag_scan_bitmap(osb, gb_inode, blkno, oiff);
 668	if (status < 0)
 669		goto bail;
 670
 671	o2info_set_request_filled(&oiff->iff_req);
 672
 673	if (o2info_to_user(*oiff, req)) {
 674		status = -EFAULT;
 675		goto out_free;
 676	}
 677
 678	status = 0;
 679bail:
 680	if (status)
 681		o2info_set_request_error(&oiff->iff_req, req);
 682out_free:
 683	kfree(oiff);
 684out_err:
 685	return status;
 686}
 687
 688static int ocfs2_info_handle_unknown(struct inode *inode,
 689				     struct ocfs2_info_request __user *req)
 690{
 691	struct ocfs2_info_request oir;
 692
 693	if (o2info_from_user(oir, req))
 694		return -EFAULT;
 695
 696	o2info_clear_request_filled(&oir);
 697
 698	if (o2info_to_user(oir, req))
 699		return -EFAULT;
 700
 701	return 0;
 702}
 703
 704/*
 705 * Validate and distinguish OCFS2_IOC_INFO requests.
 706 *
 707 * - validate the magic number.
 708 * - distinguish different requests.
 709 * - validate size of different requests.
 710 */
 711static int ocfs2_info_handle_request(struct inode *inode,
 712				     struct ocfs2_info_request __user *req)
 713{
 714	int status = -EFAULT;
 715	struct ocfs2_info_request oir;
 716
 717	if (o2info_from_user(oir, req))
 718		goto bail;
 719
 720	status = -EINVAL;
 721	if (oir.ir_magic != OCFS2_INFO_MAGIC)
 722		goto bail;
 723
 724	switch (oir.ir_code) {
 725	case OCFS2_INFO_BLOCKSIZE:
 726		if (oir.ir_size == sizeof(struct ocfs2_info_blocksize))
 727			status = ocfs2_info_handle_blocksize(inode, req);
 728		break;
 729	case OCFS2_INFO_CLUSTERSIZE:
 730		if (oir.ir_size == sizeof(struct ocfs2_info_clustersize))
 731			status = ocfs2_info_handle_clustersize(inode, req);
 732		break;
 733	case OCFS2_INFO_MAXSLOTS:
 734		if (oir.ir_size == sizeof(struct ocfs2_info_maxslots))
 735			status = ocfs2_info_handle_maxslots(inode, req);
 736		break;
 737	case OCFS2_INFO_LABEL:
 738		if (oir.ir_size == sizeof(struct ocfs2_info_label))
 739			status = ocfs2_info_handle_label(inode, req);
 740		break;
 741	case OCFS2_INFO_UUID:
 742		if (oir.ir_size == sizeof(struct ocfs2_info_uuid))
 743			status = ocfs2_info_handle_uuid(inode, req);
 744		break;
 745	case OCFS2_INFO_FS_FEATURES:
 746		if (oir.ir_size == sizeof(struct ocfs2_info_fs_features))
 747			status = ocfs2_info_handle_fs_features(inode, req);
 748		break;
 749	case OCFS2_INFO_JOURNAL_SIZE:
 750		if (oir.ir_size == sizeof(struct ocfs2_info_journal_size))
 751			status = ocfs2_info_handle_journal_size(inode, req);
 752		break;
 753	case OCFS2_INFO_FREEINODE:
 754		if (oir.ir_size == sizeof(struct ocfs2_info_freeinode))
 755			status = ocfs2_info_handle_freeinode(inode, req);
 756		break;
 757	case OCFS2_INFO_FREEFRAG:
 758		if (oir.ir_size == sizeof(struct ocfs2_info_freefrag))
 759			status = ocfs2_info_handle_freefrag(inode, req);
 760		break;
 761	default:
 762		status = ocfs2_info_handle_unknown(inode, req);
 763		break;
 764	}
 765
 766bail:
 767	return status;
 768}
 769
 770static int ocfs2_get_request_ptr(struct ocfs2_info *info, int idx,
 771				 u64 *req_addr, int compat_flag)
 772{
 773	int status = -EFAULT;
 774	u64 __user *bp = NULL;
 775
 776	if (compat_flag) {
 777#ifdef CONFIG_COMPAT
 778		/*
 779		 * pointer bp stores the base address of a pointers array,
 780		 * which collects all addresses of separate request.
 781		 */
 782		bp = (u64 __user *)(unsigned long)compat_ptr(info->oi_requests);
 783#else
 784		BUG();
 785#endif
 786	} else
 787		bp = (u64 __user *)(unsigned long)(info->oi_requests);
 788
 789	if (o2info_from_user(*req_addr, bp + idx))
 790		goto bail;
 791
 792	status = 0;
 793bail:
 794	return status;
 795}
 796
 797/*
 798 * OCFS2_IOC_INFO handles an array of requests passed from userspace.
 799 *
 800 * ocfs2_info_handle() recevies a large info aggregation, grab and
 801 * validate the request count from header, then break it into small
 802 * pieces, later specific handlers can handle them one by one.
 803 *
 804 * Idea here is to make each separate request small enough to ensure
 805 * a better backward&forward compatibility, since a small piece of
 806 * request will be less likely to be broken if disk layout get changed.
 807 */
 808static int ocfs2_info_handle(struct inode *inode, struct ocfs2_info *info,
 809			     int compat_flag)
 810{
 811	int i, status = 0;
 812	u64 req_addr;
 813	struct ocfs2_info_request __user *reqp;
 814
 815	if ((info->oi_count > OCFS2_INFO_MAX_REQUEST) ||
 816	    (!info->oi_requests)) {
 817		status = -EINVAL;
 818		goto bail;
 819	}
 820
 821	for (i = 0; i < info->oi_count; i++) {
 822
 823		status = ocfs2_get_request_ptr(info, i, &req_addr, compat_flag);
 824		if (status)
 825			break;
 826
 827		reqp = (struct ocfs2_info_request __user *)(unsigned long)req_addr;
 828		if (!reqp) {
 829			status = -EINVAL;
 830			goto bail;
 831		}
 832
 833		status = ocfs2_info_handle_request(inode, reqp);
 834		if (status)
 835			break;
 836	}
 837
 838bail:
 839	return status;
 840}
 841
 842long ocfs2_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
 843{
 844	struct inode *inode = file_inode(filp);
 845	unsigned int flags;
 846	int new_clusters;
 847	int status;
 848	struct ocfs2_space_resv sr;
 849	struct ocfs2_new_group_input input;
 850	struct reflink_arguments args;
 851	const char __user *old_path;
 852	const char __user *new_path;
 853	bool preserve;
 854	struct ocfs2_info info;
 855	void __user *argp = (void __user *)arg;
 856
 857	switch (cmd) {
 858	case OCFS2_IOC_GETFLAGS:
 859		status = ocfs2_get_inode_attr(inode, &flags);
 860		if (status < 0)
 861			return status;
 862
 863		flags &= OCFS2_FL_VISIBLE;
 864		return put_user(flags, (int __user *) arg);
 865	case OCFS2_IOC_SETFLAGS:
 866		if (get_user(flags, (int __user *) arg))
 867			return -EFAULT;
 868
 869		status = mnt_want_write_file(filp);
 870		if (status)
 871			return status;
 872		status = ocfs2_set_inode_attr(inode, flags,
 873			OCFS2_FL_MODIFIABLE);
 874		mnt_drop_write_file(filp);
 875		return status;
 876	case OCFS2_IOC_RESVSP:
 877	case OCFS2_IOC_RESVSP64:
 878	case OCFS2_IOC_UNRESVSP:
 879	case OCFS2_IOC_UNRESVSP64:
 880		if (copy_from_user(&sr, (int __user *) arg, sizeof(sr)))
 881			return -EFAULT;
 882
 883		return ocfs2_change_file_space(filp, cmd, &sr);
 884	case OCFS2_IOC_GROUP_EXTEND:
 885		if (!capable(CAP_SYS_RESOURCE))
 886			return -EPERM;
 887
 888		if (get_user(new_clusters, (int __user *)arg))
 889			return -EFAULT;
 890
 891		status = mnt_want_write_file(filp);
 892		if (status)
 893			return status;
 894		status = ocfs2_group_extend(inode, new_clusters);
 895		mnt_drop_write_file(filp);
 896		return status;
 897	case OCFS2_IOC_GROUP_ADD:
 898	case OCFS2_IOC_GROUP_ADD64:
 899		if (!capable(CAP_SYS_RESOURCE))
 900			return -EPERM;
 901
 902		if (copy_from_user(&input, (int __user *) arg, sizeof(input)))
 903			return -EFAULT;
 904
 905		status = mnt_want_write_file(filp);
 906		if (status)
 907			return status;
 908		status = ocfs2_group_add(inode, &input);
 909		mnt_drop_write_file(filp);
 910		return status;
 911	case OCFS2_IOC_REFLINK:
 912		if (copy_from_user(&args, argp, sizeof(args)))
 913			return -EFAULT;
 914		old_path = (const char __user *)(unsigned long)args.old_path;
 915		new_path = (const char __user *)(unsigned long)args.new_path;
 916		preserve = (args.preserve != 0);
 917
 918		return ocfs2_reflink_ioctl(inode, old_path, new_path, preserve);
 919	case OCFS2_IOC_INFO:
 920		if (copy_from_user(&info, argp, sizeof(struct ocfs2_info)))
 921			return -EFAULT;
 922
 923		return ocfs2_info_handle(inode, &info, 0);
 924	case FITRIM:
 925	{
 926		struct super_block *sb = inode->i_sb;
 927		struct request_queue *q = bdev_get_queue(sb->s_bdev);
 928		struct fstrim_range range;
 929		int ret = 0;
 930
 931		if (!capable(CAP_SYS_ADMIN))
 932			return -EPERM;
 933
 934		if (!blk_queue_discard(q))
 935			return -EOPNOTSUPP;
 936
 937		if (copy_from_user(&range, argp, sizeof(range)))
 938			return -EFAULT;
 939
 940		range.minlen = max_t(u64, q->limits.discard_granularity,
 941				     range.minlen);
 942		ret = ocfs2_trim_fs(sb, &range);
 943		if (ret < 0)
 944			return ret;
 945
 946		if (copy_to_user(argp, &range, sizeof(range)))
 947			return -EFAULT;
 948
 949		return 0;
 950	}
 951	case OCFS2_IOC_MOVE_EXT:
 952		return ocfs2_ioctl_move_extents(filp, argp);
 953	default:
 954		return -ENOTTY;
 955	}
 956}
 957
 958#ifdef CONFIG_COMPAT
 959long ocfs2_compat_ioctl(struct file *file, unsigned cmd, unsigned long arg)
 960{
 961	bool preserve;
 962	struct reflink_arguments args;
 963	struct inode *inode = file_inode(file);
 964	struct ocfs2_info info;
 965	void __user *argp = (void __user *)arg;
 966
 967	switch (cmd) {
 968	case OCFS2_IOC32_GETFLAGS:
 969		cmd = OCFS2_IOC_GETFLAGS;
 970		break;
 971	case OCFS2_IOC32_SETFLAGS:
 972		cmd = OCFS2_IOC_SETFLAGS;
 973		break;
 974	case OCFS2_IOC_RESVSP:
 975	case OCFS2_IOC_RESVSP64:
 976	case OCFS2_IOC_UNRESVSP:
 977	case OCFS2_IOC_UNRESVSP64:
 978	case OCFS2_IOC_GROUP_EXTEND:
 979	case OCFS2_IOC_GROUP_ADD:
 980	case OCFS2_IOC_GROUP_ADD64:
 981		break;
 982	case OCFS2_IOC_REFLINK:
 983		if (copy_from_user(&args, argp, sizeof(args)))
 984			return -EFAULT;
 985		preserve = (args.preserve != 0);
 986
 987		return ocfs2_reflink_ioctl(inode, compat_ptr(args.old_path),
 988					   compat_ptr(args.new_path), preserve);
 989	case OCFS2_IOC_INFO:
 990		if (copy_from_user(&info, argp, sizeof(struct ocfs2_info)))
 991			return -EFAULT;
 992
 993		return ocfs2_info_handle(inode, &info, 1);
 
 994	case OCFS2_IOC_MOVE_EXT:
 995		break;
 996	default:
 997		return -ENOIOCTLCMD;
 998	}
 999
1000	return ocfs2_ioctl(file, cmd, arg);
1001}
1002#endif
v5.9
  1// SPDX-License-Identifier: GPL-2.0
  2/*
  3 * linux/fs/ocfs2/ioctl.c
  4 *
  5 * Copyright (C) 2006 Herbert Poetzl
  6 * adapted from Remy Card's ext2/ioctl.c
  7 */
  8
  9#include <linux/fs.h>
 10#include <linux/mount.h>
 11#include <linux/blkdev.h>
 12#include <linux/compat.h>
 13
 14#include <cluster/masklog.h>
 15
 16#include "ocfs2.h"
 17#include "alloc.h"
 18#include "dlmglue.h"
 19#include "file.h"
 20#include "inode.h"
 21#include "journal.h"
 22
 23#include "ocfs2_fs.h"
 24#include "ioctl.h"
 25#include "resize.h"
 26#include "refcounttree.h"
 27#include "sysfile.h"
 28#include "dir.h"
 29#include "buffer_head_io.h"
 30#include "suballoc.h"
 31#include "move_extents.h"
 32
 33#define o2info_from_user(a, b)	\
 34		copy_from_user(&(a), (b), sizeof(a))
 35#define o2info_to_user(a, b)	\
 36		copy_to_user((typeof(a) __user *)b, &(a), sizeof(a))
 37
 38/*
 39 * This is just a best-effort to tell userspace that this request
 40 * caused the error.
 41 */
 42static inline void o2info_set_request_error(struct ocfs2_info_request *kreq,
 43					struct ocfs2_info_request __user *req)
 44{
 45	kreq->ir_flags |= OCFS2_INFO_FL_ERROR;
 46	(void)put_user(kreq->ir_flags, (__u32 __user *)&(req->ir_flags));
 47}
 48
 49static inline void o2info_set_request_filled(struct ocfs2_info_request *req)
 50{
 51	req->ir_flags |= OCFS2_INFO_FL_FILLED;
 52}
 53
 54static inline void o2info_clear_request_filled(struct ocfs2_info_request *req)
 55{
 56	req->ir_flags &= ~OCFS2_INFO_FL_FILLED;
 57}
 58
 59static inline int o2info_coherent(struct ocfs2_info_request *req)
 60{
 61	return (!(req->ir_flags & OCFS2_INFO_FL_NON_COHERENT));
 62}
 63
 64static int ocfs2_get_inode_attr(struct inode *inode, unsigned *flags)
 65{
 66	int status;
 67
 68	status = ocfs2_inode_lock(inode, NULL, 0);
 69	if (status < 0) {
 70		mlog_errno(status);
 71		return status;
 72	}
 73	ocfs2_get_inode_flags(OCFS2_I(inode));
 74	*flags = OCFS2_I(inode)->ip_attr;
 75	ocfs2_inode_unlock(inode, 0);
 76
 77	return status;
 78}
 79
 80static int ocfs2_set_inode_attr(struct inode *inode, unsigned flags,
 81				unsigned mask)
 82{
 83	struct ocfs2_inode_info *ocfs2_inode = OCFS2_I(inode);
 84	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
 85	handle_t *handle = NULL;
 86	struct buffer_head *bh = NULL;
 87	unsigned oldflags;
 88	int status;
 89
 90	inode_lock(inode);
 91
 92	status = ocfs2_inode_lock(inode, &bh, 1);
 93	if (status < 0) {
 94		mlog_errno(status);
 95		goto bail;
 96	}
 97
 98	status = -EACCES;
 99	if (!inode_owner_or_capable(inode))
100		goto bail_unlock;
101
102	if (!S_ISDIR(inode->i_mode))
103		flags &= ~OCFS2_DIRSYNC_FL;
104
105	oldflags = ocfs2_inode->ip_attr;
106	flags = flags & mask;
107	flags |= oldflags & ~mask;
108
109	status = vfs_ioc_setflags_prepare(inode, oldflags, flags);
110	if (status)
111		goto bail_unlock;
 
 
 
 
 
 
 
112
113	handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
114	if (IS_ERR(handle)) {
115		status = PTR_ERR(handle);
116		mlog_errno(status);
117		goto bail_unlock;
118	}
119
120	ocfs2_inode->ip_attr = flags;
121	ocfs2_set_inode_flags(inode);
122
123	status = ocfs2_mark_inode_dirty(handle, inode, bh);
124	if (status < 0)
125		mlog_errno(status);
126
127	ocfs2_commit_trans(osb, handle);
128
129bail_unlock:
130	ocfs2_inode_unlock(inode, 1);
131bail:
132	inode_unlock(inode);
133
134	brelse(bh);
135
136	return status;
137}
138
139static int ocfs2_info_handle_blocksize(struct inode *inode,
140				       struct ocfs2_info_request __user *req)
141{
142	struct ocfs2_info_blocksize oib;
143
144	if (o2info_from_user(oib, req))
145		return -EFAULT;
146
147	oib.ib_blocksize = inode->i_sb->s_blocksize;
148
149	o2info_set_request_filled(&oib.ib_req);
150
151	if (o2info_to_user(oib, req))
152		return -EFAULT;
153
154	return 0;
155}
156
157static int ocfs2_info_handle_clustersize(struct inode *inode,
158					 struct ocfs2_info_request __user *req)
159{
160	struct ocfs2_info_clustersize oic;
161	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
162
163	if (o2info_from_user(oic, req))
164		return -EFAULT;
165
166	oic.ic_clustersize = osb->s_clustersize;
167
168	o2info_set_request_filled(&oic.ic_req);
169
170	if (o2info_to_user(oic, req))
171		return -EFAULT;
172
173	return 0;
174}
175
176static int ocfs2_info_handle_maxslots(struct inode *inode,
177				      struct ocfs2_info_request __user *req)
178{
179	struct ocfs2_info_maxslots oim;
180	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
181
182	if (o2info_from_user(oim, req))
183		return -EFAULT;
184
185	oim.im_max_slots = osb->max_slots;
186
187	o2info_set_request_filled(&oim.im_req);
188
189	if (o2info_to_user(oim, req))
190		return -EFAULT;
191
192	return 0;
193}
194
195static int ocfs2_info_handle_label(struct inode *inode,
196				   struct ocfs2_info_request __user *req)
197{
198	struct ocfs2_info_label oil;
199	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
200
201	if (o2info_from_user(oil, req))
202		return -EFAULT;
203
204	memcpy(oil.il_label, osb->vol_label, OCFS2_MAX_VOL_LABEL_LEN);
205
206	o2info_set_request_filled(&oil.il_req);
207
208	if (o2info_to_user(oil, req))
209		return -EFAULT;
210
211	return 0;
212}
213
214static int ocfs2_info_handle_uuid(struct inode *inode,
215				  struct ocfs2_info_request __user *req)
216{
217	struct ocfs2_info_uuid oiu;
218	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
219
220	if (o2info_from_user(oiu, req))
221		return -EFAULT;
222
223	memcpy(oiu.iu_uuid_str, osb->uuid_str, OCFS2_TEXT_UUID_LEN + 1);
224
225	o2info_set_request_filled(&oiu.iu_req);
226
227	if (o2info_to_user(oiu, req))
228		return -EFAULT;
229
230	return 0;
231}
232
233static int ocfs2_info_handle_fs_features(struct inode *inode,
234					 struct ocfs2_info_request __user *req)
235{
236	struct ocfs2_info_fs_features oif;
237	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
238
239	if (o2info_from_user(oif, req))
240		return -EFAULT;
241
242	oif.if_compat_features = osb->s_feature_compat;
243	oif.if_incompat_features = osb->s_feature_incompat;
244	oif.if_ro_compat_features = osb->s_feature_ro_compat;
245
246	o2info_set_request_filled(&oif.if_req);
247
248	if (o2info_to_user(oif, req))
249		return -EFAULT;
250
251	return 0;
252}
253
254static int ocfs2_info_handle_journal_size(struct inode *inode,
255					  struct ocfs2_info_request __user *req)
256{
257	struct ocfs2_info_journal_size oij;
258	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
259
260	if (o2info_from_user(oij, req))
261		return -EFAULT;
262
263	oij.ij_journal_size = i_size_read(osb->journal->j_inode);
264
265	o2info_set_request_filled(&oij.ij_req);
266
267	if (o2info_to_user(oij, req))
268		return -EFAULT;
269
270	return 0;
271}
272
273static int ocfs2_info_scan_inode_alloc(struct ocfs2_super *osb,
274				       struct inode *inode_alloc, u64 blkno,
275				       struct ocfs2_info_freeinode *fi,
276				       u32 slot)
277{
278	int status = 0, unlock = 0;
279
280	struct buffer_head *bh = NULL;
281	struct ocfs2_dinode *dinode_alloc = NULL;
282
283	if (inode_alloc)
284		inode_lock(inode_alloc);
285
286	if (inode_alloc && o2info_coherent(&fi->ifi_req)) {
287		status = ocfs2_inode_lock(inode_alloc, &bh, 0);
288		if (status < 0) {
289			mlog_errno(status);
290			goto bail;
291		}
292		unlock = 1;
293	} else {
294		status = ocfs2_read_blocks_sync(osb, blkno, 1, &bh);
295		if (status < 0) {
296			mlog_errno(status);
297			goto bail;
298		}
299	}
300
301	dinode_alloc = (struct ocfs2_dinode *)bh->b_data;
302
303	fi->ifi_stat[slot].lfi_total =
304		le32_to_cpu(dinode_alloc->id1.bitmap1.i_total);
305	fi->ifi_stat[slot].lfi_free =
306		le32_to_cpu(dinode_alloc->id1.bitmap1.i_total) -
307		le32_to_cpu(dinode_alloc->id1.bitmap1.i_used);
308
309bail:
310	if (unlock)
311		ocfs2_inode_unlock(inode_alloc, 0);
312
313	if (inode_alloc)
314		inode_unlock(inode_alloc);
315
316	brelse(bh);
317
318	return status;
319}
320
321static int ocfs2_info_handle_freeinode(struct inode *inode,
322				       struct ocfs2_info_request __user *req)
323{
324	u32 i;
325	u64 blkno = -1;
326	char namebuf[40];
327	int status, type = INODE_ALLOC_SYSTEM_INODE;
328	struct ocfs2_info_freeinode *oifi = NULL;
329	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
330	struct inode *inode_alloc = NULL;
331
332	oifi = kzalloc(sizeof(struct ocfs2_info_freeinode), GFP_KERNEL);
333	if (!oifi) {
334		status = -ENOMEM;
335		mlog_errno(status);
336		goto out_err;
337	}
338
339	if (o2info_from_user(*oifi, req)) {
340		status = -EFAULT;
341		goto out_free;
342	}
343
344	oifi->ifi_slotnum = osb->max_slots;
345
346	for (i = 0; i < oifi->ifi_slotnum; i++) {
347		if (o2info_coherent(&oifi->ifi_req)) {
348			inode_alloc = ocfs2_get_system_file_inode(osb, type, i);
349			if (!inode_alloc) {
350				mlog(ML_ERROR, "unable to get alloc inode in "
351				    "slot %u\n", i);
352				status = -EIO;
353				goto bail;
354			}
355		} else {
356			ocfs2_sprintf_system_inode_name(namebuf,
357							sizeof(namebuf),
358							type, i);
359			status = ocfs2_lookup_ino_from_name(osb->sys_root_inode,
360							    namebuf,
361							    strlen(namebuf),
362							    &blkno);
363			if (status < 0) {
364				status = -ENOENT;
365				goto bail;
366			}
367		}
368
369		status = ocfs2_info_scan_inode_alloc(osb, inode_alloc, blkno, oifi, i);
370
371		iput(inode_alloc);
372		inode_alloc = NULL;
373
374		if (status < 0)
375			goto bail;
376	}
377
378	o2info_set_request_filled(&oifi->ifi_req);
379
380	if (o2info_to_user(*oifi, req)) {
381		status = -EFAULT;
382		goto out_free;
383	}
384
385	status = 0;
386bail:
387	if (status)
388		o2info_set_request_error(&oifi->ifi_req, req);
389out_free:
390	kfree(oifi);
391out_err:
392	return status;
393}
394
395static void o2ffg_update_histogram(struct ocfs2_info_free_chunk_list *hist,
396				   unsigned int chunksize)
397{
398	u32 index;
399
400	index = __ilog2_u32(chunksize);
401	if (index >= OCFS2_INFO_MAX_HIST)
402		index = OCFS2_INFO_MAX_HIST - 1;
403
404	hist->fc_chunks[index]++;
405	hist->fc_clusters[index] += chunksize;
406}
407
408static void o2ffg_update_stats(struct ocfs2_info_freefrag_stats *stats,
409			       unsigned int chunksize)
410{
411	if (chunksize > stats->ffs_max)
412		stats->ffs_max = chunksize;
413
414	if (chunksize < stats->ffs_min)
415		stats->ffs_min = chunksize;
416
417	stats->ffs_avg += chunksize;
418	stats->ffs_free_chunks_real++;
419}
420
421static void ocfs2_info_update_ffg(struct ocfs2_info_freefrag *ffg,
422				  unsigned int chunksize)
423{
424	o2ffg_update_histogram(&(ffg->iff_ffs.ffs_fc_hist), chunksize);
425	o2ffg_update_stats(&(ffg->iff_ffs), chunksize);
426}
427
428static int ocfs2_info_freefrag_scan_chain(struct ocfs2_super *osb,
429					  struct inode *gb_inode,
430					  struct ocfs2_dinode *gb_dinode,
431					  struct ocfs2_chain_rec *rec,
432					  struct ocfs2_info_freefrag *ffg,
433					  u32 chunks_in_group)
434{
435	int status = 0, used;
436	u64 blkno;
437
438	struct buffer_head *bh = NULL;
439	struct ocfs2_group_desc *bg = NULL;
440
441	unsigned int max_bits, num_clusters;
442	unsigned int offset = 0, cluster, chunk;
443	unsigned int chunk_free, last_chunksize = 0;
444
445	if (!le32_to_cpu(rec->c_free))
446		goto bail;
447
448	do {
449		if (!bg)
450			blkno = le64_to_cpu(rec->c_blkno);
451		else
452			blkno = le64_to_cpu(bg->bg_next_group);
453
454		if (bh) {
455			brelse(bh);
456			bh = NULL;
457		}
458
459		if (o2info_coherent(&ffg->iff_req))
460			status = ocfs2_read_group_descriptor(gb_inode,
461							     gb_dinode,
462							     blkno, &bh);
463		else
464			status = ocfs2_read_blocks_sync(osb, blkno, 1, &bh);
465
466		if (status < 0) {
467			mlog(ML_ERROR, "Can't read the group descriptor # "
468			     "%llu from device.", (unsigned long long)blkno);
469			status = -EIO;
470			goto bail;
471		}
472
473		bg = (struct ocfs2_group_desc *)bh->b_data;
474
475		if (!le16_to_cpu(bg->bg_free_bits_count))
476			continue;
477
478		max_bits = le16_to_cpu(bg->bg_bits);
479		offset = 0;
480
481		for (chunk = 0; chunk < chunks_in_group; chunk++) {
482			/*
483			 * last chunk may be not an entire one.
484			 */
485			if ((offset + ffg->iff_chunksize) > max_bits)
486				num_clusters = max_bits - offset;
487			else
488				num_clusters = ffg->iff_chunksize;
489
490			chunk_free = 0;
491			for (cluster = 0; cluster < num_clusters; cluster++) {
492				used = ocfs2_test_bit(offset,
493						(unsigned long *)bg->bg_bitmap);
494				/*
495				 * - chunk_free counts free clusters in #N chunk.
496				 * - last_chunksize records the size(in) clusters
497				 *   for the last real free chunk being counted.
498				 */
499				if (!used) {
500					last_chunksize++;
501					chunk_free++;
502				}
503
504				if (used && last_chunksize) {
505					ocfs2_info_update_ffg(ffg,
506							      last_chunksize);
507					last_chunksize = 0;
508				}
509
510				offset++;
511			}
512
513			if (chunk_free == ffg->iff_chunksize)
514				ffg->iff_ffs.ffs_free_chunks++;
515		}
516
517		/*
518		 * need to update the info for last free chunk.
519		 */
520		if (last_chunksize)
521			ocfs2_info_update_ffg(ffg, last_chunksize);
522
523	} while (le64_to_cpu(bg->bg_next_group));
524
525bail:
526	brelse(bh);
527
528	return status;
529}
530
531static int ocfs2_info_freefrag_scan_bitmap(struct ocfs2_super *osb,
532					   struct inode *gb_inode, u64 blkno,
533					   struct ocfs2_info_freefrag *ffg)
534{
535	u32 chunks_in_group;
536	int status = 0, unlock = 0, i;
537
538	struct buffer_head *bh = NULL;
539	struct ocfs2_chain_list *cl = NULL;
540	struct ocfs2_chain_rec *rec = NULL;
541	struct ocfs2_dinode *gb_dinode = NULL;
542
543	if (gb_inode)
544		inode_lock(gb_inode);
545
546	if (o2info_coherent(&ffg->iff_req)) {
547		status = ocfs2_inode_lock(gb_inode, &bh, 0);
548		if (status < 0) {
549			mlog_errno(status);
550			goto bail;
551		}
552		unlock = 1;
553	} else {
554		status = ocfs2_read_blocks_sync(osb, blkno, 1, &bh);
555		if (status < 0) {
556			mlog_errno(status);
557			goto bail;
558		}
559	}
560
561	gb_dinode = (struct ocfs2_dinode *)bh->b_data;
562	cl = &(gb_dinode->id2.i_chain);
563
564	/*
565	 * Chunksize(in) clusters from userspace should be
566	 * less than clusters in a group.
567	 */
568	if (ffg->iff_chunksize > le16_to_cpu(cl->cl_cpg)) {
569		status = -EINVAL;
570		goto bail;
571	}
572
573	memset(&ffg->iff_ffs, 0, sizeof(struct ocfs2_info_freefrag_stats));
574
575	ffg->iff_ffs.ffs_min = ~0U;
576	ffg->iff_ffs.ffs_clusters =
577			le32_to_cpu(gb_dinode->id1.bitmap1.i_total);
578	ffg->iff_ffs.ffs_free_clusters = ffg->iff_ffs.ffs_clusters -
579			le32_to_cpu(gb_dinode->id1.bitmap1.i_used);
580
581	chunks_in_group = le16_to_cpu(cl->cl_cpg) / ffg->iff_chunksize + 1;
582
583	for (i = 0; i < le16_to_cpu(cl->cl_next_free_rec); i++) {
584		rec = &(cl->cl_recs[i]);
585		status = ocfs2_info_freefrag_scan_chain(osb, gb_inode,
586							gb_dinode,
587							rec, ffg,
588							chunks_in_group);
589		if (status)
590			goto bail;
591	}
592
593	if (ffg->iff_ffs.ffs_free_chunks_real)
594		ffg->iff_ffs.ffs_avg = (ffg->iff_ffs.ffs_avg /
595					ffg->iff_ffs.ffs_free_chunks_real);
596bail:
597	if (unlock)
598		ocfs2_inode_unlock(gb_inode, 0);
599
600	if (gb_inode)
601		inode_unlock(gb_inode);
602
603	iput(gb_inode);
604	brelse(bh);
605
606	return status;
607}
608
609static int ocfs2_info_handle_freefrag(struct inode *inode,
610				      struct ocfs2_info_request __user *req)
611{
612	u64 blkno = -1;
613	char namebuf[40];
614	int status, type = GLOBAL_BITMAP_SYSTEM_INODE;
615
616	struct ocfs2_info_freefrag *oiff;
617	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
618	struct inode *gb_inode = NULL;
619
620	oiff = kzalloc(sizeof(struct ocfs2_info_freefrag), GFP_KERNEL);
621	if (!oiff) {
622		status = -ENOMEM;
623		mlog_errno(status);
624		goto out_err;
625	}
626
627	if (o2info_from_user(*oiff, req)) {
628		status = -EFAULT;
629		goto out_free;
630	}
631	/*
632	 * chunksize from userspace should be power of 2.
633	 */
634	if ((oiff->iff_chunksize & (oiff->iff_chunksize - 1)) ||
635	    (!oiff->iff_chunksize)) {
636		status = -EINVAL;
637		goto bail;
638	}
639
640	if (o2info_coherent(&oiff->iff_req)) {
641		gb_inode = ocfs2_get_system_file_inode(osb, type,
642						       OCFS2_INVALID_SLOT);
643		if (!gb_inode) {
644			mlog(ML_ERROR, "unable to get global_bitmap inode\n");
645			status = -EIO;
646			goto bail;
647		}
648	} else {
649		ocfs2_sprintf_system_inode_name(namebuf, sizeof(namebuf), type,
650						OCFS2_INVALID_SLOT);
651		status = ocfs2_lookup_ino_from_name(osb->sys_root_inode,
652						    namebuf,
653						    strlen(namebuf),
654						    &blkno);
655		if (status < 0) {
656			status = -ENOENT;
657			goto bail;
658		}
659	}
660
661	status = ocfs2_info_freefrag_scan_bitmap(osb, gb_inode, blkno, oiff);
662	if (status < 0)
663		goto bail;
664
665	o2info_set_request_filled(&oiff->iff_req);
666
667	if (o2info_to_user(*oiff, req)) {
668		status = -EFAULT;
669		goto out_free;
670	}
671
672	status = 0;
673bail:
674	if (status)
675		o2info_set_request_error(&oiff->iff_req, req);
676out_free:
677	kfree(oiff);
678out_err:
679	return status;
680}
681
682static int ocfs2_info_handle_unknown(struct inode *inode,
683				     struct ocfs2_info_request __user *req)
684{
685	struct ocfs2_info_request oir;
686
687	if (o2info_from_user(oir, req))
688		return -EFAULT;
689
690	o2info_clear_request_filled(&oir);
691
692	if (o2info_to_user(oir, req))
693		return -EFAULT;
694
695	return 0;
696}
697
698/*
699 * Validate and distinguish OCFS2_IOC_INFO requests.
700 *
701 * - validate the magic number.
702 * - distinguish different requests.
703 * - validate size of different requests.
704 */
705static int ocfs2_info_handle_request(struct inode *inode,
706				     struct ocfs2_info_request __user *req)
707{
708	int status = -EFAULT;
709	struct ocfs2_info_request oir;
710
711	if (o2info_from_user(oir, req))
712		goto bail;
713
714	status = -EINVAL;
715	if (oir.ir_magic != OCFS2_INFO_MAGIC)
716		goto bail;
717
718	switch (oir.ir_code) {
719	case OCFS2_INFO_BLOCKSIZE:
720		if (oir.ir_size == sizeof(struct ocfs2_info_blocksize))
721			status = ocfs2_info_handle_blocksize(inode, req);
722		break;
723	case OCFS2_INFO_CLUSTERSIZE:
724		if (oir.ir_size == sizeof(struct ocfs2_info_clustersize))
725			status = ocfs2_info_handle_clustersize(inode, req);
726		break;
727	case OCFS2_INFO_MAXSLOTS:
728		if (oir.ir_size == sizeof(struct ocfs2_info_maxslots))
729			status = ocfs2_info_handle_maxslots(inode, req);
730		break;
731	case OCFS2_INFO_LABEL:
732		if (oir.ir_size == sizeof(struct ocfs2_info_label))
733			status = ocfs2_info_handle_label(inode, req);
734		break;
735	case OCFS2_INFO_UUID:
736		if (oir.ir_size == sizeof(struct ocfs2_info_uuid))
737			status = ocfs2_info_handle_uuid(inode, req);
738		break;
739	case OCFS2_INFO_FS_FEATURES:
740		if (oir.ir_size == sizeof(struct ocfs2_info_fs_features))
741			status = ocfs2_info_handle_fs_features(inode, req);
742		break;
743	case OCFS2_INFO_JOURNAL_SIZE:
744		if (oir.ir_size == sizeof(struct ocfs2_info_journal_size))
745			status = ocfs2_info_handle_journal_size(inode, req);
746		break;
747	case OCFS2_INFO_FREEINODE:
748		if (oir.ir_size == sizeof(struct ocfs2_info_freeinode))
749			status = ocfs2_info_handle_freeinode(inode, req);
750		break;
751	case OCFS2_INFO_FREEFRAG:
752		if (oir.ir_size == sizeof(struct ocfs2_info_freefrag))
753			status = ocfs2_info_handle_freefrag(inode, req);
754		break;
755	default:
756		status = ocfs2_info_handle_unknown(inode, req);
757		break;
758	}
759
760bail:
761	return status;
762}
763
764static int ocfs2_get_request_ptr(struct ocfs2_info *info, int idx,
765				 u64 *req_addr, int compat_flag)
766{
767	int status = -EFAULT;
768	u64 __user *bp = NULL;
769
770	if (compat_flag) {
771#ifdef CONFIG_COMPAT
772		/*
773		 * pointer bp stores the base address of a pointers array,
774		 * which collects all addresses of separate request.
775		 */
776		bp = (u64 __user *)(unsigned long)compat_ptr(info->oi_requests);
777#else
778		BUG();
779#endif
780	} else
781		bp = (u64 __user *)(unsigned long)(info->oi_requests);
782
783	if (o2info_from_user(*req_addr, bp + idx))
784		goto bail;
785
786	status = 0;
787bail:
788	return status;
789}
790
791/*
792 * OCFS2_IOC_INFO handles an array of requests passed from userspace.
793 *
794 * ocfs2_info_handle() recevies a large info aggregation, grab and
795 * validate the request count from header, then break it into small
796 * pieces, later specific handlers can handle them one by one.
797 *
798 * Idea here is to make each separate request small enough to ensure
799 * a better backward&forward compatibility, since a small piece of
800 * request will be less likely to be broken if disk layout get changed.
801 */
802static int ocfs2_info_handle(struct inode *inode, struct ocfs2_info *info,
803			     int compat_flag)
804{
805	int i, status = 0;
806	u64 req_addr;
807	struct ocfs2_info_request __user *reqp;
808
809	if ((info->oi_count > OCFS2_INFO_MAX_REQUEST) ||
810	    (!info->oi_requests)) {
811		status = -EINVAL;
812		goto bail;
813	}
814
815	for (i = 0; i < info->oi_count; i++) {
816
817		status = ocfs2_get_request_ptr(info, i, &req_addr, compat_flag);
818		if (status)
819			break;
820
821		reqp = (struct ocfs2_info_request __user *)(unsigned long)req_addr;
822		if (!reqp) {
823			status = -EINVAL;
824			goto bail;
825		}
826
827		status = ocfs2_info_handle_request(inode, reqp);
828		if (status)
829			break;
830	}
831
832bail:
833	return status;
834}
835
836long ocfs2_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
837{
838	struct inode *inode = file_inode(filp);
839	unsigned int flags;
840	int new_clusters;
841	int status;
842	struct ocfs2_space_resv sr;
843	struct ocfs2_new_group_input input;
844	struct reflink_arguments args;
845	const char __user *old_path;
846	const char __user *new_path;
847	bool preserve;
848	struct ocfs2_info info;
849	void __user *argp = (void __user *)arg;
850
851	switch (cmd) {
852	case OCFS2_IOC_GETFLAGS:
853		status = ocfs2_get_inode_attr(inode, &flags);
854		if (status < 0)
855			return status;
856
857		flags &= OCFS2_FL_VISIBLE;
858		return put_user(flags, (int __user *) arg);
859	case OCFS2_IOC_SETFLAGS:
860		if (get_user(flags, (int __user *) arg))
861			return -EFAULT;
862
863		status = mnt_want_write_file(filp);
864		if (status)
865			return status;
866		status = ocfs2_set_inode_attr(inode, flags,
867			OCFS2_FL_MODIFIABLE);
868		mnt_drop_write_file(filp);
869		return status;
870	case OCFS2_IOC_RESVSP:
871	case OCFS2_IOC_RESVSP64:
872	case OCFS2_IOC_UNRESVSP:
873	case OCFS2_IOC_UNRESVSP64:
874		if (copy_from_user(&sr, (int __user *) arg, sizeof(sr)))
875			return -EFAULT;
876
877		return ocfs2_change_file_space(filp, cmd, &sr);
878	case OCFS2_IOC_GROUP_EXTEND:
879		if (!capable(CAP_SYS_RESOURCE))
880			return -EPERM;
881
882		if (get_user(new_clusters, (int __user *)arg))
883			return -EFAULT;
884
885		status = mnt_want_write_file(filp);
886		if (status)
887			return status;
888		status = ocfs2_group_extend(inode, new_clusters);
889		mnt_drop_write_file(filp);
890		return status;
891	case OCFS2_IOC_GROUP_ADD:
892	case OCFS2_IOC_GROUP_ADD64:
893		if (!capable(CAP_SYS_RESOURCE))
894			return -EPERM;
895
896		if (copy_from_user(&input, (int __user *) arg, sizeof(input)))
897			return -EFAULT;
898
899		status = mnt_want_write_file(filp);
900		if (status)
901			return status;
902		status = ocfs2_group_add(inode, &input);
903		mnt_drop_write_file(filp);
904		return status;
905	case OCFS2_IOC_REFLINK:
906		if (copy_from_user(&args, argp, sizeof(args)))
907			return -EFAULT;
908		old_path = (const char __user *)(unsigned long)args.old_path;
909		new_path = (const char __user *)(unsigned long)args.new_path;
910		preserve = (args.preserve != 0);
911
912		return ocfs2_reflink_ioctl(inode, old_path, new_path, preserve);
913	case OCFS2_IOC_INFO:
914		if (copy_from_user(&info, argp, sizeof(struct ocfs2_info)))
915			return -EFAULT;
916
917		return ocfs2_info_handle(inode, &info, 0);
918	case FITRIM:
919	{
920		struct super_block *sb = inode->i_sb;
921		struct request_queue *q = bdev_get_queue(sb->s_bdev);
922		struct fstrim_range range;
923		int ret = 0;
924
925		if (!capable(CAP_SYS_ADMIN))
926			return -EPERM;
927
928		if (!blk_queue_discard(q))
929			return -EOPNOTSUPP;
930
931		if (copy_from_user(&range, argp, sizeof(range)))
932			return -EFAULT;
933
934		range.minlen = max_t(u64, q->limits.discard_granularity,
935				     range.minlen);
936		ret = ocfs2_trim_fs(sb, &range);
937		if (ret < 0)
938			return ret;
939
940		if (copy_to_user(argp, &range, sizeof(range)))
941			return -EFAULT;
942
943		return 0;
944	}
945	case OCFS2_IOC_MOVE_EXT:
946		return ocfs2_ioctl_move_extents(filp, argp);
947	default:
948		return -ENOTTY;
949	}
950}
951
952#ifdef CONFIG_COMPAT
953long ocfs2_compat_ioctl(struct file *file, unsigned cmd, unsigned long arg)
954{
955	bool preserve;
956	struct reflink_arguments args;
957	struct inode *inode = file_inode(file);
958	struct ocfs2_info info;
959	void __user *argp = (void __user *)arg;
960
961	switch (cmd) {
962	case OCFS2_IOC32_GETFLAGS:
963		cmd = OCFS2_IOC_GETFLAGS;
964		break;
965	case OCFS2_IOC32_SETFLAGS:
966		cmd = OCFS2_IOC_SETFLAGS;
967		break;
968	case OCFS2_IOC_RESVSP:
969	case OCFS2_IOC_RESVSP64:
970	case OCFS2_IOC_UNRESVSP:
971	case OCFS2_IOC_UNRESVSP64:
972	case OCFS2_IOC_GROUP_EXTEND:
973	case OCFS2_IOC_GROUP_ADD:
974	case OCFS2_IOC_GROUP_ADD64:
975		break;
976	case OCFS2_IOC_REFLINK:
977		if (copy_from_user(&args, argp, sizeof(args)))
978			return -EFAULT;
979		preserve = (args.preserve != 0);
980
981		return ocfs2_reflink_ioctl(inode, compat_ptr(args.old_path),
982					   compat_ptr(args.new_path), preserve);
983	case OCFS2_IOC_INFO:
984		if (copy_from_user(&info, argp, sizeof(struct ocfs2_info)))
985			return -EFAULT;
986
987		return ocfs2_info_handle(inode, &info, 1);
988	case FITRIM:
989	case OCFS2_IOC_MOVE_EXT:
990		break;
991	default:
992		return -ENOIOCTLCMD;
993	}
994
995	return ocfs2_ioctl(file, cmd, arg);
996}
997#endif