linux/fs/gfs2/rgrp.c
<<
>>
Prefs
   1/*
   2 * Copyright (C) Sistina Software, Inc.  1997-2003 All rights reserved.
   3 * Copyright (C) 2004-2008 Red Hat, Inc.  All rights reserved.
   4 *
   5 * This copyrighted material is made available to anyone wishing to use,
   6 * modify, copy, or redistribute it subject to the terms and conditions
   7 * of the GNU General Public License version 2.
   8 */
   9
  10#include <linux/slab.h>
  11#include <linux/spinlock.h>
  12#include <linux/completion.h>
  13#include <linux/buffer_head.h>
  14#include <linux/fs.h>
  15#include <linux/gfs2_ondisk.h>
  16#include <linux/prefetch.h>
  17#include <linux/blkdev.h>
  18#include <linux/rbtree.h>
  19
  20#include "gfs2.h"
  21#include "incore.h"
  22#include "glock.h"
  23#include "glops.h"
  24#include "lops.h"
  25#include "meta_io.h"
  26#include "quota.h"
  27#include "rgrp.h"
  28#include "super.h"
  29#include "trans.h"
  30#include "util.h"
  31#include "log.h"
  32#include "inode.h"
  33#include "trace_gfs2.h"
  34
  35#define BFITNOENT ((u32)~0)
  36#define NO_BLOCK ((u64)~0)
  37
  38#define RSRV_CONTENTION_FACTOR 4
  39#define RGRP_RSRV_MAX_CONTENDERS 2
  40
  41#if BITS_PER_LONG == 32
  42#define LBITMASK   (0x55555555UL)
  43#define LBITSKIP55 (0x55555555UL)
  44#define LBITSKIP00 (0x00000000UL)
  45#else
  46#define LBITMASK   (0x5555555555555555UL)
  47#define LBITSKIP55 (0x5555555555555555UL)
  48#define LBITSKIP00 (0x0000000000000000UL)
  49#endif
  50
  51/*
  52 * These routines are used by the resource group routines (rgrp.c)
  53 * to keep track of block allocation.  Each block is represented by two
  54 * bits.  So, each byte represents GFS2_NBBY (i.e. 4) blocks.
  55 *
  56 * 0 = Free
  57 * 1 = Used (not metadata)
  58 * 2 = Unlinked (still in use) inode
  59 * 3 = Used (metadata)
  60 */
  61
  62static const char valid_change[16] = {
  63                /* current */
  64        /* n */ 0, 1, 1, 1,
  65        /* e */ 1, 0, 0, 0,
  66        /* w */ 0, 0, 0, 1,
  67                1, 0, 0, 0
  68};
  69
  70static u32 rgblk_search(struct gfs2_rgrpd *rgd, u32 goal,
  71                        unsigned char old_state,
  72                        struct gfs2_bitmap **rbi);
  73
  74/**
  75 * gfs2_setbit - Set a bit in the bitmaps
  76 * @rgd: the resource group descriptor
  77 * @buf2: the clone buffer that holds the bitmaps
  78 * @bi: the bitmap structure
  79 * @block: the block to set
  80 * @new_state: the new state of the block
  81 *
  82 */
  83
  84static inline void gfs2_setbit(struct gfs2_rgrpd *rgd, unsigned char *buf2,
  85                               struct gfs2_bitmap *bi, u32 block,
  86                               unsigned char new_state)
  87{
  88        unsigned char *byte1, *byte2, *end, cur_state;
  89        unsigned int buflen = bi->bi_len;
  90        const unsigned int bit = (block % GFS2_NBBY) * GFS2_BIT_SIZE;
  91
  92        byte1 = bi->bi_bh->b_data + bi->bi_offset + (block / GFS2_NBBY);
  93        end = bi->bi_bh->b_data + bi->bi_offset + buflen;
  94
  95        BUG_ON(byte1 >= end);
  96
  97        cur_state = (*byte1 >> bit) & GFS2_BIT_MASK;
  98
  99        if (unlikely(!valid_change[new_state * 4 + cur_state])) {
 100                printk(KERN_WARNING "GFS2: buf_blk = 0x%llx old_state=%d, "
 101                       "new_state=%d\n",
 102                       (unsigned long long)block, cur_state, new_state);
 103                printk(KERN_WARNING "GFS2: rgrp=0x%llx bi_start=0x%lx\n",
 104                       (unsigned long long)rgd->rd_addr,
 105                       (unsigned long)bi->bi_start);
 106                printk(KERN_WARNING "GFS2: bi_offset=0x%lx bi_len=0x%lx\n",
 107                       (unsigned long)bi->bi_offset,
 108                       (unsigned long)bi->bi_len);
 109                dump_stack();
 110                gfs2_consist_rgrpd(rgd);
 111                return;
 112        }
 113        *byte1 ^= (cur_state ^ new_state) << bit;
 114
 115        if (buf2) {
 116                byte2 = buf2 + bi->bi_offset + (block / GFS2_NBBY);
 117                cur_state = (*byte2 >> bit) & GFS2_BIT_MASK;
 118                *byte2 ^= (cur_state ^ new_state) << bit;
 119        }
 120}
 121
 122/**
 123 * gfs2_testbit - test a bit in the bitmaps
 124 * @rgd: the resource group descriptor
 125 * @buffer: the buffer that holds the bitmaps
 126 * @buflen: the length (in bytes) of the buffer
 127 * @block: the block to read
 128 *
 129 */
 130
 131static inline unsigned char gfs2_testbit(struct gfs2_rgrpd *rgd,
 132                                         const unsigned char *buffer,
 133                                         unsigned int buflen, u32 block)
 134{
 135        const unsigned char *byte, *end;
 136        unsigned char cur_state;
 137        unsigned int bit;
 138
 139        byte = buffer + (block / GFS2_NBBY);
 140        bit = (block % GFS2_NBBY) * GFS2_BIT_SIZE;
 141        end = buffer + buflen;
 142
 143        gfs2_assert(rgd->rd_sbd, byte < end);
 144
 145        cur_state = (*byte >> bit) & GFS2_BIT_MASK;
 146
 147        return cur_state;
 148}
 149
 150/**
 151 * gfs2_bit_search
 152 * @ptr: Pointer to bitmap data
 153 * @mask: Mask to use (normally 0x55555.... but adjusted for search start)
 154 * @state: The state we are searching for
 155 *
 156 * We xor the bitmap data with a patter which is the bitwise opposite
 157 * of what we are looking for, this gives rise to a pattern of ones
 158 * wherever there is a match. Since we have two bits per entry, we
 159 * take this pattern, shift it down by one place and then and it with
 160 * the original. All the even bit positions (0,2,4, etc) then represent
 161 * successful matches, so we mask with 0x55555..... to remove the unwanted
 162 * odd bit positions.
 163 *
 164 * This allows searching of a whole u64 at once (32 blocks) with a
 165 * single test (on 64 bit arches).
 166 */
 167
 168static inline u64 gfs2_bit_search(const __le64 *ptr, u64 mask, u8 state)
 169{
 170        u64 tmp;
 171        static const u64 search[] = {
 172                [0] = 0xffffffffffffffffULL,
 173                [1] = 0xaaaaaaaaaaaaaaaaULL,
 174                [2] = 0x5555555555555555ULL,
 175                [3] = 0x0000000000000000ULL,
 176        };
 177        tmp = le64_to_cpu(*ptr) ^ search[state];
 178        tmp &= (tmp >> 1);
 179        tmp &= mask;
 180        return tmp;
 181}
 182
 183/**
 184 * rs_cmp - multi-block reservation range compare
 185 * @blk: absolute file system block number of the new reservation
 186 * @len: number of blocks in the new reservation
 187 * @rs: existing reservation to compare against
 188 *
 189 * returns: 1 if the block range is beyond the reach of the reservation
 190 *         -1 if the block range is before the start of the reservation
 191 *          0 if the block range overlaps with the reservation
 192 */
 193static inline int rs_cmp(u64 blk, u32 len, struct gfs2_blkreserv *rs)
 194{
 195        u64 startblk = gfs2_rs_startblk(rs);
 196
 197        if (blk >= startblk + rs->rs_free)
 198                return 1;
 199        if (blk + len - 1 < startblk)
 200                return -1;
 201        return 0;
 202}
 203
 204/**
 205 * rs_find - Find a rgrp multi-block reservation that contains a given block
 206 * @rgd: The rgrp
 207 * @rgblk: The block we're looking for, relative to the rgrp
 208 */
 209static struct gfs2_blkreserv *rs_find(struct gfs2_rgrpd *rgd, u32 rgblk)
 210{
 211        struct rb_node **newn;
 212        int rc;
 213        u64 fsblk = rgblk + rgd->rd_data0;
 214
 215        spin_lock(&rgd->rd_rsspin);
 216        newn = &rgd->rd_rstree.rb_node;
 217        while (*newn) {
 218                struct gfs2_blkreserv *cur =
 219                        rb_entry(*newn, struct gfs2_blkreserv, rs_node);
 220                rc = rs_cmp(fsblk, 1, cur);
 221                if (rc < 0)
 222                        newn = &((*newn)->rb_left);
 223                else if (rc > 0)
 224                        newn = &((*newn)->rb_right);
 225                else {
 226                        spin_unlock(&rgd->rd_rsspin);
 227                        return cur;
 228                }
 229        }
 230        spin_unlock(&rgd->rd_rsspin);
 231        return NULL;
 232}
 233
 234/**
 235 * gfs2_bitfit - Search an rgrp's bitmap buffer to find a bit-pair representing
 236 *       a block in a given allocation state.
 237 * @buf: the buffer that holds the bitmaps
 238 * @len: the length (in bytes) of the buffer
 239 * @goal: start search at this block's bit-pair (within @buffer)
 240 * @state: GFS2_BLKST_XXX the state of the block we're looking for.
 241 *
 242 * Scope of @goal and returned block number is only within this bitmap buffer,
 243 * not entire rgrp or filesystem.  @buffer will be offset from the actual
 244 * beginning of a bitmap block buffer, skipping any header structures, but
 245 * headers are always a multiple of 64 bits long so that the buffer is
 246 * always aligned to a 64 bit boundary.
 247 *
 248 * The size of the buffer is in bytes, but is it assumed that it is
 249 * always ok to read a complete multiple of 64 bits at the end
 250 * of the block in case the end is no aligned to a natural boundary.
 251 *
 252 * Return: the block number (bitmap buffer scope) that was found
 253 */
 254
 255static u32 gfs2_bitfit(const u8 *buf, const unsigned int len,
 256                       u32 goal, u8 state)
 257{
 258        u32 spoint = (goal << 1) & ((8*sizeof(u64)) - 1);
 259        const __le64 *ptr = ((__le64 *)buf) + (goal >> 5);
 260        const __le64 *end = (__le64 *)(buf + ALIGN(len, sizeof(u64)));
 261        u64 tmp;
 262        u64 mask = 0x5555555555555555ULL;
 263        u32 bit;
 264
 265        BUG_ON(state > 3);
 266
 267        /* Mask off bits we don't care about at the start of the search */
 268        mask <<= spoint;
 269        tmp = gfs2_bit_search(ptr, mask, state);
 270        ptr++;
 271        while(tmp == 0 && ptr < end) {
 272                tmp = gfs2_bit_search(ptr, 0x5555555555555555ULL, state);
 273                ptr++;
 274        }
 275        /* Mask off any bits which are more than len bytes from the start */
 276        if (ptr == end && (len & (sizeof(u64) - 1)))
 277                tmp &= (((u64)~0) >> (64 - 8*(len & (sizeof(u64) - 1))));
 278        /* Didn't find anything, so return */
 279        if (tmp == 0)
 280                return BFITNOENT;
 281        ptr--;
 282        bit = __ffs64(tmp);
 283        bit /= 2;       /* two bits per entry in the bitmap */
 284        return (((const unsigned char *)ptr - buf) * GFS2_NBBY) + bit;
 285}
 286
 287/**
 288 * gfs2_bitcount - count the number of bits in a certain state
 289 * @rgd: the resource group descriptor
 290 * @buffer: the buffer that holds the bitmaps
 291 * @buflen: the length (in bytes) of the buffer
 292 * @state: the state of the block we're looking for
 293 *
 294 * Returns: The number of bits
 295 */
 296
 297static u32 gfs2_bitcount(struct gfs2_rgrpd *rgd, const u8 *buffer,
 298                         unsigned int buflen, u8 state)
 299{
 300        const u8 *byte = buffer;
 301        const u8 *end = buffer + buflen;
 302        const u8 state1 = state << 2;
 303        const u8 state2 = state << 4;
 304        const u8 state3 = state << 6;
 305        u32 count = 0;
 306
 307        for (; byte < end; byte++) {
 308                if (((*byte) & 0x03) == state)
 309                        count++;
 310                if (((*byte) & 0x0C) == state1)
 311                        count++;
 312                if (((*byte) & 0x30) == state2)
 313                        count++;
 314                if (((*byte) & 0xC0) == state3)
 315                        count++;
 316        }
 317
 318        return count;
 319}
 320
 321/**
 322 * gfs2_rgrp_verify - Verify that a resource group is consistent
 323 * @rgd: the rgrp
 324 *
 325 */
 326
 327void gfs2_rgrp_verify(struct gfs2_rgrpd *rgd)
 328{
 329        struct gfs2_sbd *sdp = rgd->rd_sbd;
 330        struct gfs2_bitmap *bi = NULL;
 331        u32 length = rgd->rd_length;
 332        u32 count[4], tmp;
 333        int buf, x;
 334
 335        memset(count, 0, 4 * sizeof(u32));
 336
 337        /* Count # blocks in each of 4 possible allocation states */
 338        for (buf = 0; buf < length; buf++) {
 339                bi = rgd->rd_bits + buf;
 340                for (x = 0; x < 4; x++)
 341                        count[x] += gfs2_bitcount(rgd,
 342                                                  bi->bi_bh->b_data +
 343                                                  bi->bi_offset,
 344                                                  bi->bi_len, x);
 345        }
 346
 347        if (count[0] != rgd->rd_free) {
 348                if (gfs2_consist_rgrpd(rgd))
 349                        fs_err(sdp, "free data mismatch:  %u != %u\n",
 350                               count[0], rgd->rd_free);
 351                return;
 352        }
 353
 354        tmp = rgd->rd_data - rgd->rd_free - rgd->rd_dinodes;
 355        if (count[1] != tmp) {
 356                if (gfs2_consist_rgrpd(rgd))
 357                        fs_err(sdp, "used data mismatch:  %u != %u\n",
 358                               count[1], tmp);
 359                return;
 360        }
 361
 362        if (count[2] + count[3] != rgd->rd_dinodes) {
 363                if (gfs2_consist_rgrpd(rgd))
 364                        fs_err(sdp, "used metadata mismatch:  %u != %u\n",
 365                               count[2] + count[3], rgd->rd_dinodes);
 366                return;
 367        }
 368}
 369
 370static inline int rgrp_contains_block(struct gfs2_rgrpd *rgd, u64 block)
 371{
 372        u64 first = rgd->rd_data0;
 373        u64 last = first + rgd->rd_data;
 374        return first <= block && block < last;
 375}
 376
 377/**
 378 * gfs2_blk2rgrpd - Find resource group for a given data/meta block number
 379 * @sdp: The GFS2 superblock
 380 * @blk: The data block number
 381 * @exact: True if this needs to be an exact match
 382 *
 383 * Returns: The resource group, or NULL if not found
 384 */
 385
 386struct gfs2_rgrpd *gfs2_blk2rgrpd(struct gfs2_sbd *sdp, u64 blk, bool exact)
 387{
 388        struct rb_node *n, *next;
 389        struct gfs2_rgrpd *cur;
 390
 391        spin_lock(&sdp->sd_rindex_spin);
 392        n = sdp->sd_rindex_tree.rb_node;
 393        while (n) {
 394                cur = rb_entry(n, struct gfs2_rgrpd, rd_node);
 395                next = NULL;
 396                if (blk < cur->rd_addr)
 397                        next = n->rb_left;
 398                else if (blk >= cur->rd_data0 + cur->rd_data)
 399                        next = n->rb_right;
 400                if (next == NULL) {
 401                        spin_unlock(&sdp->sd_rindex_spin);
 402                        if (exact) {
 403                                if (blk < cur->rd_addr)
 404                                        return NULL;
 405                                if (blk >= cur->rd_data0 + cur->rd_data)
 406                                        return NULL;
 407                        }
 408                        return cur;
 409                }
 410                n = next;
 411        }
 412        spin_unlock(&sdp->sd_rindex_spin);
 413
 414        return NULL;
 415}
 416
 417/**
 418 * gfs2_rgrpd_get_first - get the first Resource Group in the filesystem
 419 * @sdp: The GFS2 superblock
 420 *
 421 * Returns: The first rgrp in the filesystem
 422 */
 423
 424struct gfs2_rgrpd *gfs2_rgrpd_get_first(struct gfs2_sbd *sdp)
 425{
 426        const struct rb_node *n;
 427        struct gfs2_rgrpd *rgd;
 428
 429        spin_lock(&sdp->sd_rindex_spin);
 430        n = rb_first(&sdp->sd_rindex_tree);
 431        rgd = rb_entry(n, struct gfs2_rgrpd, rd_node);
 432        spin_unlock(&sdp->sd_rindex_spin);
 433
 434        return rgd;
 435}
 436
 437/**
 438 * gfs2_rgrpd_get_next - get the next RG
 439 * @rgd: the resource group descriptor
 440 *
 441 * Returns: The next rgrp
 442 */
 443
 444struct gfs2_rgrpd *gfs2_rgrpd_get_next(struct gfs2_rgrpd *rgd)
 445{
 446        struct gfs2_sbd *sdp = rgd->rd_sbd;
 447        const struct rb_node *n;
 448
 449        spin_lock(&sdp->sd_rindex_spin);
 450        n = rb_next(&rgd->rd_node);
 451        if (n == NULL)
 452                n = rb_first(&sdp->sd_rindex_tree);
 453
 454        if (unlikely(&rgd->rd_node == n)) {
 455                spin_unlock(&sdp->sd_rindex_spin);
 456                return NULL;
 457        }
 458        rgd = rb_entry(n, struct gfs2_rgrpd, rd_node);
 459        spin_unlock(&sdp->sd_rindex_spin);
 460        return rgd;
 461}
 462
 463void gfs2_free_clones(struct gfs2_rgrpd *rgd)
 464{
 465        int x;
 466
 467        for (x = 0; x < rgd->rd_length; x++) {
 468                struct gfs2_bitmap *bi = rgd->rd_bits + x;
 469                kfree(bi->bi_clone);
 470                bi->bi_clone = NULL;
 471        }
 472}
 473
 474/**
 475 * gfs2_rs_alloc - make sure we have a reservation assigned to the inode
 476 * @ip: the inode for this reservation
 477 */
 478int gfs2_rs_alloc(struct gfs2_inode *ip)
 479{
 480        int error = 0;
 481        struct gfs2_blkreserv *res;
 482
 483        if (ip->i_res)
 484                return 0;
 485
 486        res = kmem_cache_zalloc(gfs2_rsrv_cachep, GFP_NOFS);
 487        if (!res)
 488                error = -ENOMEM;
 489
 490        down_write(&ip->i_rw_mutex);
 491        if (ip->i_res)
 492                kmem_cache_free(gfs2_rsrv_cachep, res);
 493        else
 494                ip->i_res = res;
 495        up_write(&ip->i_rw_mutex);
 496        return error;
 497}
 498
 499static void dump_rs(struct seq_file *seq, struct gfs2_blkreserv *rs)
 500{
 501        gfs2_print_dbg(seq, "  r: %llu s:%llu b:%u f:%u\n",
 502                       rs->rs_rgd->rd_addr, gfs2_rs_startblk(rs), rs->rs_biblk,
 503                       rs->rs_free);
 504}
 505
 506/**
 507 * __rs_deltree - remove a multi-block reservation from the rgd tree
 508 * @rs: The reservation to remove
 509 *
 510 */
 511static void __rs_deltree(struct gfs2_blkreserv *rs)
 512{
 513        struct gfs2_rgrpd *rgd;
 514
 515        if (!gfs2_rs_active(rs))
 516                return;
 517
 518        rgd = rs->rs_rgd;
 519        /* We can't do this: The reason is that when the rgrp is invalidated,
 520           it's in the "middle" of acquiring the glock, but the HOLDER bit
 521           isn't set yet:
 522           BUG_ON(!gfs2_glock_is_locked_by_me(rs->rs_rgd->rd_gl));*/
 523        trace_gfs2_rs(NULL, rs, TRACE_RS_TREEDEL);
 524
 525        if (!RB_EMPTY_ROOT(&rgd->rd_rstree))
 526                rb_erase(&rs->rs_node, &rgd->rd_rstree);
 527        BUG_ON(!rgd->rd_rs_cnt);
 528        rgd->rd_rs_cnt--;
 529
 530        if (rs->rs_free) {
 531                /* return reserved blocks to the rgrp and the ip */
 532                BUG_ON(rs->rs_rgd->rd_reserved < rs->rs_free);
 533                rs->rs_rgd->rd_reserved -= rs->rs_free;
 534                rs->rs_free = 0;
 535                clear_bit(GBF_FULL, &rs->rs_bi->bi_flags);
 536                smp_mb__after_clear_bit();
 537        }
 538        /* We can't change any of the step 1 or step 2 components of the rs.
 539           E.g. We can't set rs_rgd to NULL because the rgd glock is held and
 540           dequeued through this pointer.
 541           Can't: atomic_set(&rs->rs_sizehint, 0);
 542           Can't: rs->rs_requested = 0;
 543           Can't: rs->rs_rgd = NULL;*/
 544        rs->rs_bi = NULL;
 545        rs->rs_biblk = 0;
 546}
 547
 548/**
 549 * gfs2_rs_deltree - remove a multi-block reservation from the rgd tree
 550 * @rs: The reservation to remove
 551 *
 552 */
 553void gfs2_rs_deltree(struct gfs2_blkreserv *rs)
 554{
 555        struct gfs2_rgrpd *rgd;
 556
 557        if (!gfs2_rs_active(rs))
 558                return;
 559
 560        rgd = rs->rs_rgd;
 561        spin_lock(&rgd->rd_rsspin);
 562        __rs_deltree(rs);
 563        spin_unlock(&rgd->rd_rsspin);
 564}
 565
 566/**
 567 * gfs2_rs_delete - delete a multi-block reservation
 568 * @ip: The inode for this reservation
 569 *
 570 */
 571void gfs2_rs_delete(struct gfs2_inode *ip)
 572{
 573        down_write(&ip->i_rw_mutex);
 574        if (ip->i_res) {
 575                gfs2_rs_deltree(ip->i_res);
 576                trace_gfs2_rs(ip, ip->i_res, TRACE_RS_DELETE);
 577                BUG_ON(ip->i_res->rs_free);
 578                kmem_cache_free(gfs2_rsrv_cachep, ip->i_res);
 579                ip->i_res = NULL;
 580        }
 581        up_write(&ip->i_rw_mutex);
 582}
 583
 584/**
 585 * return_all_reservations - return all reserved blocks back to the rgrp.
 586 * @rgd: the rgrp that needs its space back
 587 *
 588 * We previously reserved a bunch of blocks for allocation. Now we need to
 589 * give them back. This leave the reservation structures in tact, but removes
 590 * all of their corresponding "no-fly zones".
 591 */
 592static void return_all_reservations(struct gfs2_rgrpd *rgd)
 593{
 594        struct rb_node *n;
 595        struct gfs2_blkreserv *rs;
 596
 597        spin_lock(&rgd->rd_rsspin);
 598        while ((n = rb_first(&rgd->rd_rstree))) {
 599                rs = rb_entry(n, struct gfs2_blkreserv, rs_node);
 600                __rs_deltree(rs);
 601        }
 602        spin_unlock(&rgd->rd_rsspin);
 603}
 604
 605void gfs2_clear_rgrpd(struct gfs2_sbd *sdp)
 606{
 607        struct rb_node *n;
 608        struct gfs2_rgrpd *rgd;
 609        struct gfs2_glock *gl;
 610
 611        while ((n = rb_first(&sdp->sd_rindex_tree))) {
 612                rgd = rb_entry(n, struct gfs2_rgrpd, rd_node);
 613                gl = rgd->rd_gl;
 614
 615                rb_erase(n, &sdp->sd_rindex_tree);
 616
 617                if (gl) {
 618                        spin_lock(&gl->gl_spin);
 619                        gl->gl_object = NULL;
 620                        spin_unlock(&gl->gl_spin);
 621                        gfs2_glock_add_to_lru(gl);
 622                        gfs2_glock_put(gl);
 623                }
 624
 625                gfs2_free_clones(rgd);
 626                kfree(rgd->rd_bits);
 627                return_all_reservations(rgd);
 628                kmem_cache_free(gfs2_rgrpd_cachep, rgd);
 629        }
 630}
 631
 632static void gfs2_rindex_print(const struct gfs2_rgrpd *rgd)
 633{
 634        printk(KERN_INFO "  ri_addr = %llu\n", (unsigned long long)rgd->rd_addr);
 635        printk(KERN_INFO "  ri_length = %u\n", rgd->rd_length);
 636        printk(KERN_INFO "  ri_data0 = %llu\n", (unsigned long long)rgd->rd_data0);
 637        printk(KERN_INFO "  ri_data = %u\n", rgd->rd_data);
 638        printk(KERN_INFO "  ri_bitbytes = %u\n", rgd->rd_bitbytes);
 639}
 640
 641/**
 642 * gfs2_compute_bitstructs - Compute the bitmap sizes
 643 * @rgd: The resource group descriptor
 644 *
 645 * Calculates bitmap descriptors, one for each block that contains bitmap data
 646 *
 647 * Returns: errno
 648 */
 649
 650static int compute_bitstructs(struct gfs2_rgrpd *rgd)
 651{
 652        struct gfs2_sbd *sdp = rgd->rd_sbd;
 653        struct gfs2_bitmap *bi;
 654        u32 length = rgd->rd_length; /* # blocks in hdr & bitmap */
 655        u32 bytes_left, bytes;
 656        int x;
 657
 658        if (!length)
 659                return -EINVAL;
 660
 661        rgd->rd_bits = kcalloc(length, sizeof(struct gfs2_bitmap), GFP_NOFS);
 662        if (!rgd->rd_bits)
 663                return -ENOMEM;
 664
 665        bytes_left = rgd->rd_bitbytes;
 666
 667        for (x = 0; x < length; x++) {
 668                bi = rgd->rd_bits + x;
 669
 670                bi->bi_flags = 0;
 671                /* small rgrp; bitmap stored completely in header block */
 672                if (length == 1) {
 673                        bytes = bytes_left;
 674                        bi->bi_offset = sizeof(struct gfs2_rgrp);
 675                        bi->bi_start = 0;
 676                        bi->bi_len = bytes;
 677                /* header block */
 678                } else if (x == 0) {
 679                        bytes = sdp->sd_sb.sb_bsize - sizeof(struct gfs2_rgrp);
 680                        bi->bi_offset = sizeof(struct gfs2_rgrp);
 681                        bi->bi_start = 0;
 682                        bi->bi_len = bytes;
 683                /* last block */
 684                } else if (x + 1 == length) {
 685                        bytes = bytes_left;
 686                        bi->bi_offset = sizeof(struct gfs2_meta_header);
 687                        bi->bi_start = rgd->rd_bitbytes - bytes_left;
 688                        bi->bi_len = bytes;
 689                /* other blocks */
 690                } else {
 691                        bytes = sdp->sd_sb.sb_bsize -
 692                                sizeof(struct gfs2_meta_header);
 693                        bi->bi_offset = sizeof(struct gfs2_meta_header);
 694                        bi->bi_start = rgd->rd_bitbytes - bytes_left;
 695                        bi->bi_len = bytes;
 696                }
 697
 698                bytes_left -= bytes;
 699        }
 700
 701        if (bytes_left) {
 702                gfs2_consist_rgrpd(rgd);
 703                return -EIO;
 704        }
 705        bi = rgd->rd_bits + (length - 1);
 706        if ((bi->bi_start + bi->bi_len) * GFS2_NBBY != rgd->rd_data) {
 707                if (gfs2_consist_rgrpd(rgd)) {
 708                        gfs2_rindex_print(rgd);
 709                        fs_err(sdp, "start=%u len=%u offset=%u\n",
 710                               bi->bi_start, bi->bi_len, bi->bi_offset);
 711                }
 712                return -EIO;
 713        }
 714
 715        return 0;
 716}
 717
 718/**
 719 * gfs2_ri_total - Total up the file system space, according to the rindex.
 720 * @sdp: the filesystem
 721 *
 722 */
 723u64 gfs2_ri_total(struct gfs2_sbd *sdp)
 724{
 725        u64 total_data = 0;     
 726        struct inode *inode = sdp->sd_rindex;
 727        struct gfs2_inode *ip = GFS2_I(inode);
 728        char buf[sizeof(struct gfs2_rindex)];
 729        int error, rgrps;
 730
 731        for (rgrps = 0;; rgrps++) {
 732                loff_t pos = rgrps * sizeof(struct gfs2_rindex);
 733
 734                if (pos + sizeof(struct gfs2_rindex) > i_size_read(inode))
 735                        break;
 736                error = gfs2_internal_read(ip, buf, &pos,
 737                                           sizeof(struct gfs2_rindex));
 738                if (error != sizeof(struct gfs2_rindex))
 739                        break;
 740                total_data += be32_to_cpu(((struct gfs2_rindex *)buf)->ri_data);
 741        }
 742        return total_data;
 743}
 744
 745static int rgd_insert(struct gfs2_rgrpd *rgd)
 746{
 747        struct gfs2_sbd *sdp = rgd->rd_sbd;
 748        struct rb_node **newn = &sdp->sd_rindex_tree.rb_node, *parent = NULL;
 749
 750        /* Figure out where to put new node */
 751        while (*newn) {
 752                struct gfs2_rgrpd *cur = rb_entry(*newn, struct gfs2_rgrpd,
 753                                                  rd_node);
 754
 755                parent = *newn;
 756                if (rgd->rd_addr < cur->rd_addr)
 757                        newn = &((*newn)->rb_left);
 758                else if (rgd->rd_addr > cur->rd_addr)
 759                        newn = &((*newn)->rb_right);
 760                else
 761                        return -EEXIST;
 762        }
 763
 764        rb_link_node(&rgd->rd_node, parent, newn);
 765        rb_insert_color(&rgd->rd_node, &sdp->sd_rindex_tree);
 766        sdp->sd_rgrps++;
 767        return 0;
 768}
 769
 770/**
 771 * read_rindex_entry - Pull in a new resource index entry from the disk
 772 * @ip: Pointer to the rindex inode
 773 *
 774 * Returns: 0 on success, > 0 on EOF, error code otherwise
 775 */
 776
 777static int read_rindex_entry(struct gfs2_inode *ip)
 778{
 779        struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
 780        loff_t pos = sdp->sd_rgrps * sizeof(struct gfs2_rindex);
 781        struct gfs2_rindex buf;
 782        int error;
 783        struct gfs2_rgrpd *rgd;
 784
 785        if (pos >= i_size_read(&ip->i_inode))
 786                return 1;
 787
 788        error = gfs2_internal_read(ip, (char *)&buf, &pos,
 789                                   sizeof(struct gfs2_rindex));
 790
 791        if (error != sizeof(struct gfs2_rindex))
 792                return (error == 0) ? 1 : error;
 793
 794        rgd = kmem_cache_zalloc(gfs2_rgrpd_cachep, GFP_NOFS);
 795        error = -ENOMEM;
 796        if (!rgd)
 797                return error;
 798
 799        rgd->rd_sbd = sdp;
 800        rgd->rd_addr = be64_to_cpu(buf.ri_addr);
 801        rgd->rd_length = be32_to_cpu(buf.ri_length);
 802        rgd->rd_data0 = be64_to_cpu(buf.ri_data0);
 803        rgd->rd_data = be32_to_cpu(buf.ri_data);
 804        rgd->rd_bitbytes = be32_to_cpu(buf.ri_bitbytes);
 805        spin_lock_init(&rgd->rd_rsspin);
 806
 807        error = compute_bitstructs(rgd);
 808        if (error)
 809                goto fail;
 810
 811        error = gfs2_glock_get(sdp, rgd->rd_addr,
 812                               &gfs2_rgrp_glops, CREATE, &rgd->rd_gl);
 813        if (error)
 814                goto fail;
 815
 816        rgd->rd_gl->gl_object = rgd;
 817        rgd->rd_rgl = (struct gfs2_rgrp_lvb *)rgd->rd_gl->gl_lvb;
 818        rgd->rd_flags &= ~GFS2_RDF_UPTODATE;
 819        if (rgd->rd_data > sdp->sd_max_rg_data)
 820                sdp->sd_max_rg_data = rgd->rd_data;
 821        spin_lock(&sdp->sd_rindex_spin);
 822        error = rgd_insert(rgd);
 823        spin_unlock(&sdp->sd_rindex_spin);
 824        if (!error)
 825                return 0;
 826
 827        error = 0; /* someone else read in the rgrp; free it and ignore it */
 828        gfs2_glock_put(rgd->rd_gl);
 829
 830fail:
 831        kfree(rgd->rd_bits);
 832        kmem_cache_free(gfs2_rgrpd_cachep, rgd);
 833        return error;
 834}
 835
 836/**
 837 * gfs2_ri_update - Pull in a new resource index from the disk
 838 * @ip: pointer to the rindex inode
 839 *
 840 * Returns: 0 on successful update, error code otherwise
 841 */
 842
 843static int gfs2_ri_update(struct gfs2_inode *ip)
 844{
 845        struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
 846        int error;
 847
 848        do {
 849                error = read_rindex_entry(ip);
 850        } while (error == 0);
 851
 852        if (error < 0)
 853                return error;
 854
 855        sdp->sd_rindex_uptodate = 1;
 856        return 0;
 857}
 858
 859/**
 860 * gfs2_rindex_update - Update the rindex if required
 861 * @sdp: The GFS2 superblock
 862 *
 863 * We grab a lock on the rindex inode to make sure that it doesn't
 864 * change whilst we are performing an operation. We keep this lock
 865 * for quite long periods of time compared to other locks. This
 866 * doesn't matter, since it is shared and it is very, very rarely
 867 * accessed in the exclusive mode (i.e. only when expanding the filesystem).
 868 *
 869 * This makes sure that we're using the latest copy of the resource index
 870 * special file, which might have been updated if someone expanded the
 871 * filesystem (via gfs2_grow utility), which adds new resource groups.
 872 *
 873 * Returns: 0 on succeess, error code otherwise
 874 */
 875
 876int gfs2_rindex_update(struct gfs2_sbd *sdp)
 877{
 878        struct gfs2_inode *ip = GFS2_I(sdp->sd_rindex);
 879        struct gfs2_glock *gl = ip->i_gl;
 880        struct gfs2_holder ri_gh;
 881        int error = 0;
 882        int unlock_required = 0;
 883
 884        /* Read new copy from disk if we don't have the latest */
 885        if (!sdp->sd_rindex_uptodate) {
 886                if (!gfs2_glock_is_locked_by_me(gl)) {
 887                        error = gfs2_glock_nq_init(gl, LM_ST_SHARED, 0, &ri_gh);
 888                        if (error)
 889                                return error;
 890                        unlock_required = 1;
 891                }
 892                if (!sdp->sd_rindex_uptodate)
 893                        error = gfs2_ri_update(ip);
 894                if (unlock_required)
 895                        gfs2_glock_dq_uninit(&ri_gh);
 896        }
 897
 898        return error;
 899}
 900
 901static void gfs2_rgrp_in(struct gfs2_rgrpd *rgd, const void *buf)
 902{
 903        const struct gfs2_rgrp *str = buf;
 904        u32 rg_flags;
 905
 906        rg_flags = be32_to_cpu(str->rg_flags);
 907        rg_flags &= ~GFS2_RDF_MASK;
 908        rgd->rd_flags &= GFS2_RDF_MASK;
 909        rgd->rd_flags |= rg_flags;
 910        rgd->rd_free = be32_to_cpu(str->rg_free);
 911        rgd->rd_dinodes = be32_to_cpu(str->rg_dinodes);
 912        rgd->rd_igeneration = be64_to_cpu(str->rg_igeneration);
 913}
 914
 915static void gfs2_rgrp_out(struct gfs2_rgrpd *rgd, void *buf)
 916{
 917        struct gfs2_rgrp *str = buf;
 918
 919        str->rg_flags = cpu_to_be32(rgd->rd_flags & ~GFS2_RDF_MASK);
 920        str->rg_free = cpu_to_be32(rgd->rd_free);
 921        str->rg_dinodes = cpu_to_be32(rgd->rd_dinodes);
 922        str->__pad = cpu_to_be32(0);
 923        str->rg_igeneration = cpu_to_be64(rgd->rd_igeneration);
 924        memset(&str->rg_reserved, 0, sizeof(str->rg_reserved));
 925}
 926
 927static int gfs2_rgrp_lvb_valid(struct gfs2_rgrpd *rgd)
 928{
 929        struct gfs2_rgrp_lvb *rgl = rgd->rd_rgl;
 930        struct gfs2_rgrp *str = (struct gfs2_rgrp *)rgd->rd_bits[0].bi_bh->b_data;
 931
 932        if (rgl->rl_flags != str->rg_flags || rgl->rl_free != str->rg_free ||
 933            rgl->rl_dinodes != str->rg_dinodes ||
 934            rgl->rl_igeneration != str->rg_igeneration)
 935                return 0;
 936        return 1;
 937}
 938
 939static void gfs2_rgrp_ondisk2lvb(struct gfs2_rgrp_lvb *rgl, const void *buf)
 940{
 941        const struct gfs2_rgrp *str = buf;
 942
 943        rgl->rl_magic = cpu_to_be32(GFS2_MAGIC);
 944        rgl->rl_flags = str->rg_flags;
 945        rgl->rl_free = str->rg_free;
 946        rgl->rl_dinodes = str->rg_dinodes;
 947        rgl->rl_igeneration = str->rg_igeneration;
 948        rgl->__pad = 0UL;
 949}
 950
 951static void update_rgrp_lvb_unlinked(struct gfs2_rgrpd *rgd, u32 change)
 952{
 953        struct gfs2_rgrp_lvb *rgl = rgd->rd_rgl;
 954        u32 unlinked = be32_to_cpu(rgl->rl_unlinked) + change;
 955        rgl->rl_unlinked = cpu_to_be32(unlinked);
 956}
 957
 958static u32 count_unlinked(struct gfs2_rgrpd *rgd)
 959{
 960        struct gfs2_bitmap *bi;
 961        const u32 length = rgd->rd_length;
 962        const u8 *buffer = NULL;
 963        u32 i, goal, count = 0;
 964
 965        for (i = 0, bi = rgd->rd_bits; i < length; i++, bi++) {
 966                goal = 0;
 967                buffer = bi->bi_bh->b_data + bi->bi_offset;
 968                WARN_ON(!buffer_uptodate(bi->bi_bh));
 969                while (goal < bi->bi_len * GFS2_NBBY) {
 970                        goal = gfs2_bitfit(buffer, bi->bi_len, goal,
 971                                           GFS2_BLKST_UNLINKED);
 972                        if (goal == BFITNOENT)
 973                                break;
 974                        count++;
 975                        goal++;
 976                }
 977        }
 978
 979        return count;
 980}
 981
 982
 983/**
 984 * gfs2_rgrp_bh_get - Read in a RG's header and bitmaps
 985 * @rgd: the struct gfs2_rgrpd describing the RG to read in
 986 *
 987 * Read in all of a Resource Group's header and bitmap blocks.
 988 * Caller must eventually call gfs2_rgrp_relse() to free the bitmaps.
 989 *
 990 * Returns: errno
 991 */
 992
 993int gfs2_rgrp_bh_get(struct gfs2_rgrpd *rgd)
 994{
 995        struct gfs2_sbd *sdp = rgd->rd_sbd;
 996        struct gfs2_glock *gl = rgd->rd_gl;
 997        unsigned int length = rgd->rd_length;
 998        struct gfs2_bitmap *bi;
 999        unsigned int x, y;
1000        int error;
1001
1002        if (rgd->rd_bits[0].bi_bh != NULL)
1003                return 0;
1004
1005        for (x = 0; x < length; x++) {
1006                bi = rgd->rd_bits + x;
1007                error = gfs2_meta_read(gl, rgd->rd_addr + x, 0, &bi->bi_bh);
1008                if (error)
1009                        goto fail;
1010        }
1011
1012        for (y = length; y--;) {
1013                bi = rgd->rd_bits + y;
1014                error = gfs2_meta_wait(sdp, bi->bi_bh);
1015                if (error)
1016                        goto fail;
1017                if (gfs2_metatype_check(sdp, bi->bi_bh, y ? GFS2_METATYPE_RB :
1018                                              GFS2_METATYPE_RG)) {
1019                        error = -EIO;
1020                        goto fail;
1021                }
1022        }
1023
1024        if (!(rgd->rd_flags & GFS2_RDF_UPTODATE)) {
1025                for (x = 0; x < length; x++)
1026                        clear_bit(GBF_FULL, &rgd->rd_bits[x].bi_flags);
1027                gfs2_rgrp_in(rgd, (rgd->rd_bits[0].bi_bh)->b_data);
1028                rgd->rd_flags |= (GFS2_RDF_UPTODATE | GFS2_RDF_CHECK);
1029                rgd->rd_free_clone = rgd->rd_free;
1030        }
1031        if (be32_to_cpu(GFS2_MAGIC) != rgd->rd_rgl->rl_magic) {
1032                rgd->rd_rgl->rl_unlinked = cpu_to_be32(count_unlinked(rgd));
1033                gfs2_rgrp_ondisk2lvb(rgd->rd_rgl,
1034                                     rgd->rd_bits[0].bi_bh->b_data);
1035        }
1036        else if (sdp->sd_args.ar_rgrplvb) {
1037                if (!gfs2_rgrp_lvb_valid(rgd)){
1038                        gfs2_consist_rgrpd(rgd);
1039                        error = -EIO;
1040                        goto fail;
1041                }
1042                if (rgd->rd_rgl->rl_unlinked == 0)
1043                        rgd->rd_flags &= ~GFS2_RDF_CHECK;
1044        }
1045        return 0;
1046
1047fail:
1048        while (x--) {
1049                bi = rgd->rd_bits + x;
1050                brelse(bi->bi_bh);
1051                bi->bi_bh = NULL;
1052                gfs2_assert_warn(sdp, !bi->bi_clone);
1053        }
1054
1055        return error;
1056}
1057
1058int update_rgrp_lvb(struct gfs2_rgrpd *rgd)
1059{
1060        u32 rl_flags;
1061
1062        if (rgd->rd_flags & GFS2_RDF_UPTODATE)
1063                return 0;
1064
1065        if (be32_to_cpu(GFS2_MAGIC) != rgd->rd_rgl->rl_magic)
1066                return gfs2_rgrp_bh_get(rgd);
1067
1068        rl_flags = be32_to_cpu(rgd->rd_rgl->rl_flags);
1069        rl_flags &= ~GFS2_RDF_MASK;
1070        rgd->rd_flags &= GFS2_RDF_MASK;
1071        rgd->rd_flags |= (rl_flags | GFS2_RDF_UPTODATE | GFS2_RDF_CHECK);
1072        if (rgd->rd_rgl->rl_unlinked == 0)
1073                rgd->rd_flags &= ~GFS2_RDF_CHECK;
1074        rgd->rd_free = be32_to_cpu(rgd->rd_rgl->rl_free);
1075        rgd->rd_free_clone = rgd->rd_free;
1076        rgd->rd_dinodes = be32_to_cpu(rgd->rd_rgl->rl_dinodes);
1077        rgd->rd_igeneration = be64_to_cpu(rgd->rd_rgl->rl_igeneration);
1078        return 0;
1079}
1080
1081int gfs2_rgrp_go_lock(struct gfs2_holder *gh)
1082{
1083        struct gfs2_rgrpd *rgd = gh->gh_gl->gl_object;
1084        struct gfs2_sbd *sdp = rgd->rd_sbd;
1085
1086        if (gh->gh_flags & GL_SKIP && sdp->sd_args.ar_rgrplvb)
1087                return 0;
1088        return gfs2_rgrp_bh_get((struct gfs2_rgrpd *)gh->gh_gl->gl_object);
1089}
1090
1091/**
1092 * gfs2_rgrp_go_unlock - Release RG bitmaps read in with gfs2_rgrp_bh_get()
1093 * @gh: The glock holder for the resource group
1094 *
1095 */
1096
1097void gfs2_rgrp_go_unlock(struct gfs2_holder *gh)
1098{
1099        struct gfs2_rgrpd *rgd = gh->gh_gl->gl_object;
1100        int x, length = rgd->rd_length;
1101
1102        for (x = 0; x < length; x++) {
1103                struct gfs2_bitmap *bi = rgd->rd_bits + x;
1104                if (bi->bi_bh) {
1105                        brelse(bi->bi_bh);
1106                        bi->bi_bh = NULL;
1107                }
1108        }
1109
1110}
1111
1112int gfs2_rgrp_send_discards(struct gfs2_sbd *sdp, u64 offset,
1113                             struct buffer_head *bh,
1114                             const struct gfs2_bitmap *bi, unsigned minlen, u64 *ptrimmed)
1115{
1116        struct super_block *sb = sdp->sd_vfs;
1117        struct block_device *bdev = sb->s_bdev;
1118        const unsigned int sects_per_blk = sdp->sd_sb.sb_bsize /
1119                                           bdev_logical_block_size(sb->s_bdev);
1120        u64 blk;
1121        sector_t start = 0;
1122        sector_t nr_sects = 0;
1123        int rv;
1124        unsigned int x;
1125        u32 trimmed = 0;
1126        u8 diff;
1127
1128        for (x = 0; x < bi->bi_len; x++) {
1129                const u8 *clone = bi->bi_clone ? bi->bi_clone : bi->bi_bh->b_data;
1130                clone += bi->bi_offset;
1131                clone += x;
1132                if (bh) {
1133                        const u8 *orig = bh->b_data + bi->bi_offset + x;
1134                        diff = ~(*orig | (*orig >> 1)) & (*clone | (*clone >> 1));
1135                } else {
1136                        diff = ~(*clone | (*clone >> 1));
1137                }
1138                diff &= 0x55;
1139                if (diff == 0)
1140                        continue;
1141                blk = offset + ((bi->bi_start + x) * GFS2_NBBY);
1142                blk *= sects_per_blk; /* convert to sectors */
1143                while(diff) {
1144                        if (diff & 1) {
1145                                if (nr_sects == 0)
1146                                        goto start_new_extent;
1147                                if ((start + nr_sects) != blk) {
1148                                        if (nr_sects >= minlen) {
1149                                                rv = blkdev_issue_discard(bdev,
1150                                                        start, nr_sects,
1151                                                        GFP_NOFS, 0);
1152                                                if (rv)
1153                                                        goto fail;
1154                                                trimmed += nr_sects;
1155                                        }
1156                                        nr_sects = 0;
1157start_new_extent:
1158                                        start = blk;
1159                                }
1160                                nr_sects += sects_per_blk;
1161                        }
1162                        diff >>= 2;
1163                        blk += sects_per_blk;
1164                }
1165        }
1166        if (nr_sects >= minlen) {
1167                rv = blkdev_issue_discard(bdev, start, nr_sects, GFP_NOFS, 0);
1168                if (rv)
1169                        goto fail;
1170                trimmed += nr_sects;
1171        }
1172        if (ptrimmed)
1173                *ptrimmed = trimmed;
1174        return 0;
1175
1176fail:
1177        if (sdp->sd_args.ar_discard)
1178                fs_warn(sdp, "error %d on discard request, turning discards off for this filesystem", rv);
1179        sdp->sd_args.ar_discard = 0;
1180        return -EIO;
1181}
1182
1183/**
1184 * gfs2_fitrim - Generate discard requests for unused bits of the filesystem
1185 * @filp: Any file on the filesystem
1186 * @argp: Pointer to the arguments (also used to pass result)
1187 *
1188 * Returns: 0 on success, otherwise error code
1189 */
1190
1191int gfs2_fitrim(struct file *filp, void __user *argp)
1192{
1193        struct inode *inode = filp->f_dentry->d_inode;
1194        struct gfs2_sbd *sdp = GFS2_SB(inode);
1195        struct request_queue *q = bdev_get_queue(sdp->sd_vfs->s_bdev);
1196        struct buffer_head *bh;
1197        struct gfs2_rgrpd *rgd;
1198        struct gfs2_rgrpd *rgd_end;
1199        struct gfs2_holder gh;
1200        struct fstrim_range r;
1201        int ret = 0;
1202        u64 amt;
1203        u64 trimmed = 0;
1204        unsigned int x;
1205
1206        if (!capable(CAP_SYS_ADMIN))
1207                return -EPERM;
1208
1209        if (!blk_queue_discard(q))
1210                return -EOPNOTSUPP;
1211
1212        if (argp == NULL) {
1213                r.start = 0;
1214                r.len = ULLONG_MAX;
1215                r.minlen = 0;
1216        } else if (copy_from_user(&r, argp, sizeof(r)))
1217                return -EFAULT;
1218
1219        ret = gfs2_rindex_update(sdp);
1220        if (ret)
1221                return ret;
1222
1223        rgd = gfs2_blk2rgrpd(sdp, r.start, 0);
1224        rgd_end = gfs2_blk2rgrpd(sdp, r.start + r.len, 0);
1225
1226        while (1) {
1227
1228                ret = gfs2_glock_nq_init(rgd->rd_gl, LM_ST_EXCLUSIVE, 0, &gh);
1229                if (ret)
1230                        goto out;
1231
1232                if (!(rgd->rd_flags & GFS2_RGF_TRIMMED)) {
1233                        /* Trim each bitmap in the rgrp */
1234                        for (x = 0; x < rgd->rd_length; x++) {
1235                                struct gfs2_bitmap *bi = rgd->rd_bits + x;
1236                                ret = gfs2_rgrp_send_discards(sdp, rgd->rd_data0, NULL, bi, r.minlen, &amt);
1237                                if (ret) {
1238                                        gfs2_glock_dq_uninit(&gh);
1239                                        goto out;
1240                                }
1241                                trimmed += amt;
1242                        }
1243
1244                        /* Mark rgrp as having been trimmed */
1245                        ret = gfs2_trans_begin(sdp, RES_RG_HDR, 0);
1246                        if (ret == 0) {
1247                                bh = rgd->rd_bits[0].bi_bh;
1248                                rgd->rd_flags |= GFS2_RGF_TRIMMED;
1249                                gfs2_trans_add_bh(rgd->rd_gl, bh, 1);
1250                                gfs2_rgrp_out(rgd, bh->b_data);
1251                                gfs2_rgrp_ondisk2lvb(rgd->rd_rgl, bh->b_data);
1252                                gfs2_trans_end(sdp);
1253                        }
1254                }
1255                gfs2_glock_dq_uninit(&gh);
1256
1257                if (rgd == rgd_end)
1258                        break;
1259
1260                rgd = gfs2_rgrpd_get_next(rgd);
1261        }
1262
1263out:
1264        r.len = trimmed << 9;
1265        if (argp && copy_to_user(argp, &r, sizeof(r)))
1266                return -EFAULT;
1267
1268        return ret;
1269}
1270
1271/**
1272 * rs_insert - insert a new multi-block reservation into the rgrp's rb_tree
1273 * @bi: the bitmap with the blocks
1274 * @ip: the inode structure
1275 * @biblk: the 32-bit block number relative to the start of the bitmap
1276 * @amount: the number of blocks to reserve
1277 *
1278 * Returns: NULL - reservation was already taken, so not inserted
1279 *          pointer to the inserted reservation
1280 */
1281static struct gfs2_blkreserv *rs_insert(struct gfs2_bitmap *bi,
1282                                       struct gfs2_inode *ip, u32 biblk,
1283                                       int amount)
1284{
1285        struct rb_node **newn, *parent = NULL;
1286        int rc;
1287        struct gfs2_blkreserv *rs = ip->i_res;
1288        struct gfs2_rgrpd *rgd = rs->rs_rgd;
1289        u64 fsblock = gfs2_bi2rgd_blk(bi, biblk) + rgd->rd_data0;
1290
1291        spin_lock(&rgd->rd_rsspin);
1292        newn = &rgd->rd_rstree.rb_node;
1293        BUG_ON(!ip->i_res);
1294        BUG_ON(gfs2_rs_active(rs));
1295        /* Figure out where to put new node */
1296        /*BUG_ON(!gfs2_glock_is_locked_by_me(rgd->rd_gl));*/
1297        while (*newn) {
1298                struct gfs2_blkreserv *cur =
1299                        rb_entry(*newn, struct gfs2_blkreserv, rs_node);
1300
1301                parent = *newn;
1302                rc = rs_cmp(fsblock, amount, cur);
1303                if (rc > 0)
1304                        newn = &((*newn)->rb_right);
1305                else if (rc < 0)
1306                        newn = &((*newn)->rb_left);
1307                else {
1308                        spin_unlock(&rgd->rd_rsspin);
1309                        return NULL; /* reservation already in use */
1310                }
1311        }
1312
1313        /* Do our reservation work */
1314        rs = ip->i_res;
1315        rs->rs_free = amount;
1316        rs->rs_biblk = biblk;
1317        rs->rs_bi = bi;
1318        rb_link_node(&rs->rs_node, parent, newn);
1319        rb_insert_color(&rs->rs_node, &rgd->rd_rstree);
1320
1321        /* Do our inode accounting for the reservation */
1322        /*BUG_ON(!gfs2_glock_is_locked_by_me(ip->i_gl));*/
1323
1324        /* Do our rgrp accounting for the reservation */
1325        rgd->rd_reserved += amount; /* blocks reserved */
1326        rgd->rd_rs_cnt++; /* number of in-tree reservations */
1327        spin_unlock(&rgd->rd_rsspin);
1328        trace_gfs2_rs(ip, rs, TRACE_RS_INSERT);
1329        return rs;
1330}
1331
1332/**
1333 * unclaimed_blocks - return number of blocks that aren't spoken for
1334 */
1335static u32 unclaimed_blocks(struct gfs2_rgrpd *rgd)
1336{
1337        return rgd->rd_free_clone - rgd->rd_reserved;
1338}
1339
1340/**
1341 * rg_mblk_search - find a group of multiple free blocks
1342 * @rgd: the resource group descriptor
1343 * @rs: the block reservation
1344 * @ip: pointer to the inode for which we're reserving blocks
1345 *
1346 * This is very similar to rgblk_search, except we're looking for whole
1347 * 64-bit words that represent a chunk of 32 free blocks. I'm only focusing
1348 * on aligned dwords for speed's sake.
1349 *
1350 * Returns: 0 if successful or BFITNOENT if there isn't enough free space
1351 */
1352
1353static int rg_mblk_search(struct gfs2_rgrpd *rgd, struct gfs2_inode *ip)
1354{
1355        struct gfs2_bitmap *bi = rgd->rd_bits;
1356        const u32 length = rgd->rd_length;
1357        u32 blk;
1358        unsigned int buf, x, search_bytes;
1359        u8 *buffer = NULL;
1360        u8 *ptr, *end, *nonzero;
1361        u32 goal, rsv_bytes;
1362        struct gfs2_blkreserv *rs;
1363        u32 best_rs_bytes, unclaimed;
1364        int best_rs_blocks;
1365
1366        /* Find bitmap block that contains bits for goal block */
1367        if (rgrp_contains_block(rgd, ip->i_goal))
1368                goal = ip->i_goal - rgd->rd_data0;
1369        else
1370                goal = rgd->rd_last_alloc;
1371        for (buf = 0; buf < length; buf++) {
1372                bi = rgd->rd_bits + buf;
1373                /* Convert scope of "goal" from rgrp-wide to within
1374                   found bit block */
1375                if (goal < (bi->bi_start + bi->bi_len) * GFS2_NBBY) {
1376                        goal -= bi->bi_start * GFS2_NBBY;
1377                        goto do_search;
1378                }
1379        }
1380        buf = 0;
1381        goal = 0;
1382
1383do_search:
1384        best_rs_blocks = max_t(int, atomic_read(&ip->i_res->rs_sizehint),
1385                               (RGRP_RSRV_MINBLKS * rgd->rd_length));
1386        best_rs_bytes = (best_rs_blocks *
1387                         (1 + (RSRV_CONTENTION_FACTOR * rgd->rd_rs_cnt))) /
1388                GFS2_NBBY; /* 1 + is for our not-yet-created reservation */
1389        best_rs_bytes = ALIGN(best_rs_bytes, sizeof(u64));
1390        unclaimed = unclaimed_blocks(rgd);
1391        if (best_rs_bytes * GFS2_NBBY > unclaimed)
1392                best_rs_bytes = unclaimed >> GFS2_BIT_SIZE;
1393
1394        for (x = 0; x <= length; x++) {
1395                bi = rgd->rd_bits + buf;
1396
1397                if (test_bit(GBF_FULL, &bi->bi_flags))
1398                        goto skip;
1399
1400                WARN_ON(!buffer_uptodate(bi->bi_bh));
1401                if (bi->bi_clone)
1402                        buffer = bi->bi_clone + bi->bi_offset;
1403                else
1404                        buffer = bi->bi_bh->b_data + bi->bi_offset;
1405
1406                /* We have to keep the reservations aligned on u64 boundaries
1407                   otherwise we could get situations where a byte can't be
1408                   used because it's after a reservation, but a free bit still
1409                   is within the reservation's area. */
1410                ptr = buffer + ALIGN(goal >> GFS2_BIT_SIZE, sizeof(u64));
1411                end = (buffer + bi->bi_len);
1412                while (ptr < end) {
1413                        rsv_bytes = 0;
1414                        if ((ptr + best_rs_bytes) <= end)
1415                                search_bytes = best_rs_bytes;
1416                        else
1417                                search_bytes = end - ptr;
1418                        BUG_ON(!search_bytes);
1419                        nonzero = memchr_inv(ptr, 0, search_bytes);
1420                        /* If the lot is all zeroes, reserve the whole size. If
1421                           there's enough zeroes to satisfy the request, use
1422                           what we can. If there's not enough, keep looking. */
1423                        if (nonzero == NULL)
1424                                rsv_bytes = search_bytes;
1425                        else if ((nonzero - ptr) * GFS2_NBBY >=
1426                                 ip->i_res->rs_requested)
1427                                rsv_bytes = (nonzero - ptr);
1428
1429                        if (rsv_bytes) {
1430                                blk = ((ptr - buffer) * GFS2_NBBY);
1431                                BUG_ON(blk >= bi->bi_len * GFS2_NBBY);
1432                                rs = rs_insert(bi, ip, blk,
1433                                               rsv_bytes * GFS2_NBBY);
1434                                if (IS_ERR(rs))
1435                                        return PTR_ERR(rs);
1436                                if (rs)
1437                                        return 0;
1438                        }
1439                        ptr += ALIGN(search_bytes, sizeof(u64));
1440                }
1441skip:
1442                /* Try next bitmap block (wrap back to rgrp header
1443                   if at end) */
1444                buf++;
1445                buf %= length;
1446                goal = 0;
1447        }
1448
1449        return BFITNOENT;
1450}
1451
1452/**
1453 * try_rgrp_fit - See if a given reservation will fit in a given RG
1454 * @rgd: the RG data
1455 * @ip: the inode
1456 *
1457 * If there's room for the requested blocks to be allocated from the RG:
1458 * This will try to get a multi-block reservation first, and if that doesn't
1459 * fit, it will take what it can.
1460 *
1461 * Returns: 1 on success (it fits), 0 on failure (it doesn't fit)
1462 */
1463
1464static int try_rgrp_fit(struct gfs2_rgrpd *rgd, struct gfs2_inode *ip)
1465{
1466        struct gfs2_blkreserv *rs = ip->i_res;
1467
1468        if (rgd->rd_flags & (GFS2_RGF_NOALLOC | GFS2_RDF_ERROR))
1469                return 0;
1470        /* Look for a multi-block reservation. */
1471        if (unclaimed_blocks(rgd) >= RGRP_RSRV_MINBLKS &&
1472            rg_mblk_search(rgd, ip) != BFITNOENT)
1473                return 1;
1474        if (unclaimed_blocks(rgd) >= rs->rs_requested)
1475                return 1;
1476
1477        return 0;
1478}
1479
1480/**
1481 * try_rgrp_unlink - Look for any unlinked, allocated, but unused inodes
1482 * @rgd: The rgrp
1483 * @last_unlinked: block address of the last dinode we unlinked
1484 * @skip: block address we should explicitly not unlink
1485 *
1486 * Returns: 0 if no error
1487 *          The inode, if one has been found, in inode.
1488 */
1489
1490static void try_rgrp_unlink(struct gfs2_rgrpd *rgd, u64 *last_unlinked, u64 skip)
1491{
1492        u32 goal = 0, block;
1493        u64 no_addr;
1494        struct gfs2_sbd *sdp = rgd->rd_sbd;
1495        struct gfs2_glock *gl;
1496        struct gfs2_inode *ip;
1497        int error;
1498        int found = 0;
1499        struct gfs2_bitmap *bi;
1500
1501        while (goal < rgd->rd_data) {
1502                down_write(&sdp->sd_log_flush_lock);
1503                block = rgblk_search(rgd, goal, GFS2_BLKST_UNLINKED, &bi);
1504                up_write(&sdp->sd_log_flush_lock);
1505                if (block == BFITNOENT)
1506                        break;
1507
1508                block = gfs2_bi2rgd_blk(bi, block);
1509                /* rgblk_search can return a block < goal, so we need to
1510                   keep it marching forward. */
1511                no_addr = block + rgd->rd_data0;
1512                goal = max(block + 1, goal + 1);
1513                if (*last_unlinked != NO_BLOCK && no_addr <= *last_unlinked)
1514                        continue;
1515                if (no_addr == skip)
1516                        continue;
1517                *last_unlinked = no_addr;
1518
1519                error = gfs2_glock_get(sdp, no_addr, &gfs2_inode_glops, CREATE, &gl);
1520                if (error)
1521                        continue;
1522
1523                /* If the inode is already in cache, we can ignore it here
1524                 * because the existing inode disposal code will deal with
1525                 * it when all refs have gone away. Accessing gl_object like
1526                 * this is not safe in general. Here it is ok because we do
1527                 * not dereference the pointer, and we only need an approx
1528                 * answer to whether it is NULL or not.
1529                 */
1530                ip = gl->gl_object;
1531
1532                if (ip || queue_work(gfs2_delete_workqueue, &gl->gl_delete) == 0)
1533                        gfs2_glock_put(gl);
1534                else
1535                        found++;
1536
1537                /* Limit reclaim to sensible number of tasks */
1538                if (found > NR_CPUS)
1539                        return;
1540        }
1541
1542        rgd->rd_flags &= ~GFS2_RDF_CHECK;
1543        return;
1544}
1545
1546/**
1547 * gfs2_inplace_reserve - Reserve space in the filesystem
1548 * @ip: the inode to reserve space for
1549 * @requested: the number of blocks to be reserved
1550 *
1551 * Returns: errno
1552 */
1553
1554int gfs2_inplace_reserve(struct gfs2_inode *ip, u32 requested)
1555{
1556        struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
1557        struct gfs2_rgrpd *begin = NULL;
1558        struct gfs2_blkreserv *rs = ip->i_res;
1559        int error = 0, rg_locked, flags = LM_FLAG_TRY;
1560        u64 last_unlinked = NO_BLOCK;
1561        int loops = 0;
1562
1563        if (sdp->sd_args.ar_rgrplvb)
1564                flags |= GL_SKIP;
1565        rs->rs_requested = requested;
1566        if (gfs2_assert_warn(sdp, requested)) {
1567                error = -EINVAL;
1568                goto out;
1569        }
1570        if (gfs2_rs_active(rs)) {
1571                begin = rs->rs_rgd;
1572                flags = 0; /* Yoda: Do or do not. There is no try */
1573        } else if (ip->i_rgd && rgrp_contains_block(ip->i_rgd, ip->i_goal)) {
1574                rs->rs_rgd = begin = ip->i_rgd;
1575        } else {
1576                rs->rs_rgd = begin = gfs2_blk2rgrpd(sdp, ip->i_goal, 1);
1577        }
1578        if (rs->rs_rgd == NULL)
1579                return -EBADSLT;
1580
1581        while (loops < 3) {
1582                rg_locked = 0;
1583
1584                if (gfs2_glock_is_locked_by_me(rs->rs_rgd->rd_gl)) {
1585                        rg_locked = 1;
1586                        error = 0;
1587                } else if (!loops && !gfs2_rs_active(rs) &&
1588                           rs->rs_rgd->rd_rs_cnt > RGRP_RSRV_MAX_CONTENDERS) {
1589                        /* If the rgrp already is maxed out for contenders,
1590                           we can eliminate it as a "first pass" without even
1591                           requesting the rgrp glock. */
1592                        error = GLR_TRYFAILED;
1593                } else {
1594                        error = gfs2_glock_nq_init(rs->rs_rgd->rd_gl,
1595                                                   LM_ST_EXCLUSIVE, flags,
1596                                                   &rs->rs_rgd_gh);
1597                        if (!error && sdp->sd_args.ar_rgrplvb) {
1598                                error = update_rgrp_lvb(rs->rs_rgd);
1599                                if (error) {
1600                                        gfs2_glock_dq_uninit(&rs->rs_rgd_gh);
1601                                        return error;
1602                                }
1603                        }
1604                }
1605                switch (error) {
1606                case 0:
1607                        if (gfs2_rs_active(rs)) {
1608                                if (unclaimed_blocks(rs->rs_rgd) +
1609                                    rs->rs_free >= rs->rs_requested) {
1610                                        ip->i_rgd = rs->rs_rgd;
1611                                        return 0;
1612                                }
1613                                /* We have a multi-block reservation, but the
1614                                   rgrp doesn't have enough free blocks to
1615                                   satisfy the request. Free the reservation
1616                                   and look for a suitable rgrp. */
1617                                gfs2_rs_deltree(rs);
1618                        }
1619                        if (try_rgrp_fit(rs->rs_rgd, ip)) {
1620                                if (sdp->sd_args.ar_rgrplvb)
1621                                        gfs2_rgrp_bh_get(rs->rs_rgd);
1622                                ip->i_rgd = rs->rs_rgd;
1623                                return 0;
1624                        }
1625                        if (rs->rs_rgd->rd_flags & GFS2_RDF_CHECK) {
1626                                if (sdp->sd_args.ar_rgrplvb)
1627                                        gfs2_rgrp_bh_get(rs->rs_rgd);
1628                                try_rgrp_unlink(rs->rs_rgd, &last_unlinked,
1629                                                ip->i_no_addr);
1630                        }
1631                        if (!rg_locked)
1632                                gfs2_glock_dq_uninit(&rs->rs_rgd_gh);
1633                        /* fall through */
1634                case GLR_TRYFAILED:
1635                        rs->rs_rgd = gfs2_rgrpd_get_next(rs->rs_rgd);
1636                        rs->rs_rgd = rs->rs_rgd ? : begin; /* if NULL, wrap */
1637                        if (rs->rs_rgd != begin) /* If we didn't wrap */
1638                                break;
1639
1640                        flags &= ~LM_FLAG_TRY;
1641                        loops++;
1642                        /* Check that fs hasn't grown if writing to rindex */
1643                        if (ip == GFS2_I(sdp->sd_rindex) &&
1644                            !sdp->sd_rindex_uptodate) {
1645                                error = gfs2_ri_update(ip);
1646                                if (error)
1647                                        goto out;
1648                        } else if (loops == 2)
1649                                /* Flushing the log may release space */
1650                                gfs2_log_flush(sdp, NULL);
1651                        break;
1652                default:
1653                        goto out;
1654                }
1655        }
1656        error = -ENOSPC;
1657
1658out:
1659        if (error)
1660                rs->rs_requested = 0;
1661        return error;
1662}
1663
1664/**
1665 * gfs2_inplace_release - release an inplace reservation
1666 * @ip: the inode the reservation was taken out on
1667 *
1668 * Release a reservation made by gfs2_inplace_reserve().
1669 */
1670
1671void gfs2_inplace_release(struct gfs2_inode *ip)
1672{
1673        struct gfs2_blkreserv *rs = ip->i_res;
1674
1675        if (!rs)
1676                return;
1677
1678        if (!rs->rs_free)
1679                gfs2_rs_deltree(rs);
1680
1681        if (rs->rs_rgd_gh.gh_gl)
1682                gfs2_glock_dq_uninit(&rs->rs_rgd_gh);
1683        rs->rs_requested = 0;
1684}
1685
1686/**
1687 * gfs2_get_block_type - Check a block in a RG is of given type
1688 * @rgd: the resource group holding the block
1689 * @block: the block number
1690 *
1691 * Returns: The block type (GFS2_BLKST_*)
1692 */
1693
1694static unsigned char gfs2_get_block_type(struct gfs2_rgrpd *rgd, u64 block)
1695{
1696        struct gfs2_bitmap *bi = NULL;
1697        u32 length, rgrp_block, buf_block;
1698        unsigned int buf;
1699        unsigned char type;
1700
1701        length = rgd->rd_length;
1702        rgrp_block = block - rgd->rd_data0;
1703
1704        for (buf = 0; buf < length; buf++) {
1705                bi = rgd->rd_bits + buf;
1706                if (rgrp_block < (bi->bi_start + bi->bi_len) * GFS2_NBBY)
1707                        break;
1708        }
1709
1710        gfs2_assert(rgd->rd_sbd, buf < length);
1711        buf_block = rgrp_block - bi->bi_start * GFS2_NBBY;
1712
1713        type = gfs2_testbit(rgd, bi->bi_bh->b_data + bi->bi_offset,
1714                           bi->bi_len, buf_block);
1715
1716        return type;
1717}
1718
1719/**
1720 * rgblk_search - find a block in @state
1721 * @rgd: the resource group descriptor
1722 * @goal: the goal block within the RG (start here to search for avail block)
1723 * @state: GFS2_BLKST_XXX the before-allocation state to find
1724 * @rbi: address of the pointer to the bitmap containing the block found
1725 *
1726 * Walk rgrp's bitmap to find bits that represent a block in @state.
1727 *
1728 * This function never fails, because we wouldn't call it unless we
1729 * know (from reservation results, etc.) that a block is available.
1730 *
1731 * Scope of @goal is just within rgrp, not the whole filesystem.
1732 * Scope of @returned block is just within bitmap, not the whole filesystem.
1733 *
1734 * Returns: the block number found relative to the bitmap rbi
1735 */
1736
1737static u32 rgblk_search(struct gfs2_rgrpd *rgd, u32 goal, unsigned char state,
1738                        struct gfs2_bitmap **rbi)
1739{
1740        struct gfs2_bitmap *bi = NULL;
1741        const u32 length = rgd->rd_length;
1742        u32 biblk = BFITNOENT;
1743        unsigned int buf, x;
1744        const u8 *buffer = NULL;
1745
1746        *rbi = NULL;
1747        /* Find bitmap block that contains bits for goal block */
1748        for (buf = 0; buf < length; buf++) {
1749                bi = rgd->rd_bits + buf;
1750                /* Convert scope of "goal" from rgrp-wide to within found bit block */
1751                if (goal < (bi->bi_start + bi->bi_len) * GFS2_NBBY) {
1752                        goal -= bi->bi_start * GFS2_NBBY;
1753                        goto do_search;
1754                }
1755        }
1756        buf = 0;
1757        goal = 0;
1758
1759do_search:
1760        /* Search (up to entire) bitmap in this rgrp for allocatable block.
1761           "x <= length", instead of "x < length", because we typically start
1762           the search in the middle of a bit block, but if we can't find an
1763           allocatable block anywhere else, we want to be able wrap around and
1764           search in the first part of our first-searched bit block.  */
1765        for (x = 0; x <= length; x++) {
1766                bi = rgd->rd_bits + buf;
1767
1768                if (test_bit(GBF_FULL, &bi->bi_flags) &&
1769                    (state == GFS2_BLKST_FREE))
1770                        goto skip;
1771
1772                /* The GFS2_BLKST_UNLINKED state doesn't apply to the clone
1773                   bitmaps, so we must search the originals for that. */
1774                buffer = bi->bi_bh->b_data + bi->bi_offset;
1775                WARN_ON(!buffer_uptodate(bi->bi_bh));
1776                if (state != GFS2_BLKST_UNLINKED && bi->bi_clone)
1777                        buffer = bi->bi_clone + bi->bi_offset;
1778
1779                while (1) {
1780                        struct gfs2_blkreserv *rs;
1781                        u32 rgblk;
1782
1783                        biblk = gfs2_bitfit(buffer, bi->bi_len, goal, state);
1784                        if (biblk == BFITNOENT)
1785                                break;
1786                        /* Check if this block is reserved() */
1787                        rgblk = gfs2_bi2rgd_blk(bi, biblk);
1788                        rs = rs_find(rgd, rgblk);
1789                        if (rs == NULL)
1790                                break;
1791
1792                        BUG_ON(rs->rs_bi != bi);
1793                        biblk = BFITNOENT;
1794                        /* This should jump to the first block after the
1795                           reservation. */
1796                        goal = rs->rs_biblk + rs->rs_free;
1797                        if (goal >= bi->bi_len * GFS2_NBBY)
1798                                break;
1799                }
1800                if (biblk != BFITNOENT)
1801                        break;
1802
1803                if ((goal == 0) && (state == GFS2_BLKST_FREE))
1804                        set_bit(GBF_FULL, &bi->bi_flags);
1805
1806                /* Try next bitmap block (wrap back to rgrp header if at end) */
1807skip:
1808                buf++;
1809                buf %= length;
1810                goal = 0;
1811        }
1812
1813        if (biblk != BFITNOENT)
1814                *rbi = bi;
1815
1816        return biblk;
1817}
1818
1819/**
1820 * gfs2_alloc_extent - allocate an extent from a given bitmap
1821 * @rgd: the resource group descriptor
1822 * @bi: the bitmap within the rgrp
1823 * @blk: the block within the bitmap
1824 * @dinode: TRUE if the first block we allocate is for a dinode
1825 * @n: The extent length
1826 *
1827 * Add the found bitmap buffer to the transaction.
1828 * Set the found bits to @new_state to change block's allocation state.
1829 * Returns: starting block number of the extent (fs scope)
1830 */
1831static u64 gfs2_alloc_extent(struct gfs2_rgrpd *rgd, struct gfs2_bitmap *bi,
1832                             u32 blk, bool dinode, unsigned int *n)
1833{
1834        const unsigned int elen = *n;
1835        u32 goal, rgblk;
1836        const u8 *buffer = NULL;
1837        struct gfs2_blkreserv *rs;
1838
1839        *n = 0;
1840        buffer = bi->bi_bh->b_data + bi->bi_offset;
1841        gfs2_trans_add_bh(rgd->rd_gl, bi->bi_bh, 1);
1842        gfs2_setbit(rgd, bi->bi_clone, bi, blk,
1843                    dinode ? GFS2_BLKST_DINODE : GFS2_BLKST_USED);
1844        (*n)++;
1845        goal = blk;
1846        while (*n < elen) {
1847                goal++;
1848                if (goal >= (bi->bi_len * GFS2_NBBY))
1849                        break;
1850                rgblk = gfs2_bi2rgd_blk(bi, goal);
1851                rs = rs_find(rgd, rgblk);
1852                if (rs) /* Oops, we bumped into someone's reservation */
1853                        break;
1854                if (gfs2_testbit(rgd, buffer, bi->bi_len, goal) !=
1855                    GFS2_BLKST_FREE)
1856                        break;
1857                gfs2_setbit(rgd, bi->bi_clone, bi, goal, GFS2_BLKST_USED);
1858                (*n)++;
1859        }
1860        blk = gfs2_bi2rgd_blk(bi, blk);
1861        rgd->rd_last_alloc = blk + *n - 1;
1862        return rgd->rd_data0 + blk;
1863}
1864
1865/**
1866 * rgblk_free - Change alloc state of given block(s)
1867 * @sdp: the filesystem
1868 * @bstart: the start of a run of blocks to free
1869 * @blen: the length of the block run (all must lie within ONE RG!)
1870 * @new_state: GFS2_BLKST_XXX the after-allocation block state
1871 *
1872 * Returns:  Resource group containing the block(s)
1873 */
1874
1875static struct gfs2_rgrpd *rgblk_free(struct gfs2_sbd *sdp, u64 bstart,
1876                                     u32 blen, unsigned char new_state)
1877{
1878        struct gfs2_rgrpd *rgd;
1879        struct gfs2_bitmap *bi = NULL;
1880        u32 length, rgrp_blk, buf_blk;
1881        unsigned int buf;
1882
1883        rgd = gfs2_blk2rgrpd(sdp, bstart, 1);
1884        if (!rgd) {
1885                if (gfs2_consist(sdp))
1886                        fs_err(sdp, "block = %llu\n", (unsigned long long)bstart);
1887                return NULL;
1888        }
1889
1890        length = rgd->rd_length;
1891
1892        rgrp_blk = bstart - rgd->rd_data0;
1893
1894        while (blen--) {
1895                for (buf = 0; buf < length; buf++) {
1896                        bi = rgd->rd_bits + buf;
1897                        if (rgrp_blk < (bi->bi_start + bi->bi_len) * GFS2_NBBY)
1898                                break;
1899                }
1900
1901                gfs2_assert(rgd->rd_sbd, buf < length);
1902
1903                buf_blk = rgrp_blk - bi->bi_start * GFS2_NBBY;
1904                rgrp_blk++;
1905
1906                if (!bi->bi_clone) {
1907                        bi->bi_clone = kmalloc(bi->bi_bh->b_size,
1908                                               GFP_NOFS | __GFP_NOFAIL);
1909                        memcpy(bi->bi_clone + bi->bi_offset,
1910                               bi->bi_bh->b_data + bi->bi_offset,
1911                               bi->bi_len);
1912                }
1913                gfs2_trans_add_bh(rgd->rd_gl, bi->bi_bh, 1);
1914                gfs2_setbit(rgd, NULL, bi, buf_blk, new_state);
1915        }
1916
1917        return rgd;
1918}
1919
1920/**
1921 * gfs2_rgrp_dump - print out an rgrp
1922 * @seq: The iterator
1923 * @gl: The glock in question
1924 *
1925 */
1926
1927int gfs2_rgrp_dump(struct seq_file *seq, const struct gfs2_glock *gl)
1928{
1929        struct gfs2_rgrpd *rgd = gl->gl_object;
1930        struct gfs2_blkreserv *trs;
1931        const struct rb_node *n;
1932
1933        if (rgd == NULL)
1934                return 0;
1935        gfs2_print_dbg(seq, " R: n:%llu f:%02x b:%u/%u i:%u r:%u\n",
1936                       (unsigned long long)rgd->rd_addr, rgd->rd_flags,
1937                       rgd->rd_free, rgd->rd_free_clone, rgd->rd_dinodes,
1938                       rgd->rd_reserved);
1939        spin_lock(&rgd->rd_rsspin);
1940        for (n = rb_first(&rgd->rd_rstree); n; n = rb_next(&trs->rs_node)) {
1941                trs = rb_entry(n, struct gfs2_blkreserv, rs_node);
1942                dump_rs(seq, trs);
1943        }
1944        spin_unlock(&rgd->rd_rsspin);
1945        return 0;
1946}
1947
1948static void gfs2_rgrp_error(struct gfs2_rgrpd *rgd)
1949{
1950        struct gfs2_sbd *sdp = rgd->rd_sbd;
1951        fs_warn(sdp, "rgrp %llu has an error, marking it readonly until umount\n",
1952                (unsigned long long)rgd->rd_addr);
1953        fs_warn(sdp, "umount on all nodes and run fsck.gfs2 to fix the error\n");
1954        gfs2_rgrp_dump(NULL, rgd->rd_gl);
1955        rgd->rd_flags |= GFS2_RDF_ERROR;
1956}
1957
1958/**
1959 * claim_reserved_blks - Claim previously reserved blocks
1960 * @ip: the inode that's claiming the reservation
1961 * @dinode: 1 if this block is a dinode block, otherwise data block
1962 * @nblocks: desired extent length
1963 *
1964 * Lay claim to previously reserved blocks.
1965 * Returns: Starting block number of the blocks claimed.
1966 * Sets *nblocks to the actual extent length allocated.
1967 */
1968static u64 claim_reserved_blks(struct gfs2_inode *ip, bool dinode,
1969                               unsigned int *nblocks)
1970{
1971        struct gfs2_blkreserv *rs = ip->i_res;
1972        struct gfs2_rgrpd *rgd = rs->rs_rgd;
1973        struct gfs2_bitmap *bi;
1974        u64 start_block = gfs2_rs_startblk(rs);
1975        const unsigned int elen = *nblocks;
1976
1977        bi = rs->rs_bi;
1978        gfs2_trans_add_bh(rgd->rd_gl, bi->bi_bh, 1);
1979
1980        for (*nblocks = 0; *nblocks < elen && rs->rs_free; (*nblocks)++) {
1981                if (gfs2_testbit(rgd, bi->bi_bh->b_data + bi->bi_offset,
1982                                 bi->bi_len, rs->rs_biblk) != GFS2_BLKST_FREE)
1983                        break;
1984                gfs2_setbit(rgd, bi->bi_clone, bi, rs->rs_biblk,
1985                            dinode ? GFS2_BLKST_DINODE : GFS2_BLKST_USED);
1986                rs->rs_biblk++;
1987                rs->rs_free--;
1988
1989                BUG_ON(!rgd->rd_reserved);
1990                rgd->rd_reserved--;
1991                dinode = false;
1992        }
1993
1994        trace_gfs2_rs(ip, rs, TRACE_RS_CLAIM);
1995        if (!rs->rs_free || *nblocks != elen)
1996                gfs2_rs_deltree(rs);
1997
1998        return start_block;
1999}
2000
2001/**
2002 * gfs2_alloc_blocks - Allocate one or more blocks of data and/or a dinode
2003 * @ip: the inode to allocate the block for
2004 * @bn: Used to return the starting block number
2005 * @nblocks: requested number of blocks/extent length (value/result)
2006 * @dinode: 1 if we're allocating a dinode block, else 0
2007 * @generation: the generation number of the inode
2008 *
2009 * Returns: 0 or error
2010 */
2011
2012int gfs2_alloc_blocks(struct gfs2_inode *ip, u64 *bn, unsigned int *nblocks,
2013                      bool dinode, u64 *generation)
2014{
2015        struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
2016        struct buffer_head *dibh;
2017        struct gfs2_rgrpd *rgd;
2018        unsigned int ndata;
2019        u32 goal, blk; /* block, within the rgrp scope */
2020        u64 block; /* block, within the file system scope */
2021        int error;
2022        struct gfs2_bitmap *bi;
2023
2024        /* Only happens if there is a bug in gfs2, return something distinctive
2025         * to ensure that it is noticed.
2026         */
2027        if (ip->i_res->rs_requested == 0)
2028                return -ECANCELED;
2029
2030        /* If we have a reservation, claim blocks from it. */
2031        if (gfs2_rs_active(ip->i_res)) {
2032                BUG_ON(!ip->i_res->rs_free);
2033                rgd = ip->i_res->rs_rgd;
2034                block = claim_reserved_blks(ip, dinode, nblocks);
2035                if (*nblocks)
2036                        goto found_blocks;
2037        }
2038
2039        rgd = ip->i_rgd;
2040
2041        if (!dinode && rgrp_contains_block(rgd, ip->i_goal))
2042                goal = ip->i_goal - rgd->rd_data0;
2043        else
2044                goal = rgd->rd_last_alloc;
2045
2046        blk = rgblk_search(rgd, goal, GFS2_BLKST_FREE, &bi);
2047
2048        /* Since all blocks are reserved in advance, this shouldn't happen */
2049        if (blk == BFITNOENT) {
2050                printk(KERN_WARNING "BFITNOENT, nblocks=%u\n", *nblocks);
2051                printk(KERN_WARNING "FULL=%d\n",
2052                       test_bit(GBF_FULL, &rgd->rd_bits->bi_flags));
2053                goto rgrp_error;
2054        }
2055
2056        block = gfs2_alloc_extent(rgd, bi, blk, dinode, nblocks);
2057found_blocks:
2058        ndata = *nblocks;
2059        if (dinode)
2060                ndata--;
2061
2062        if (!dinode) {
2063                ip->i_goal = block + ndata - 1;
2064                error = gfs2_meta_inode_buffer(ip, &dibh);
2065                if (error == 0) {
2066                        struct gfs2_dinode *di =
2067                                (struct gfs2_dinode *)dibh->b_data;
2068                        gfs2_trans_add_bh(ip->i_gl, dibh, 1);
2069                        di->di_goal_meta = di->di_goal_data =
2070                                cpu_to_be64(ip->i_goal);
2071                        brelse(dibh);
2072                }
2073        }
2074        if (rgd->rd_free < *nblocks) {
2075                printk(KERN_WARNING "nblocks=%u\n", *nblocks);
2076                goto rgrp_error;
2077        }
2078
2079        rgd->rd_free -= *nblocks;
2080        if (dinode) {
2081                rgd->rd_dinodes++;
2082                *generation = rgd->rd_igeneration++;
2083                if (*generation == 0)
2084                        *generation = rgd->rd_igeneration++;
2085        }
2086
2087        gfs2_trans_add_bh(rgd->rd_gl, rgd->rd_bits[0].bi_bh, 1);
2088        gfs2_rgrp_out(rgd, rgd->rd_bits[0].bi_bh->b_data);
2089        gfs2_rgrp_ondisk2lvb(rgd->rd_rgl, rgd->rd_bits[0].bi_bh->b_data);
2090
2091        gfs2_statfs_change(sdp, 0, -(s64)*nblocks, dinode ? 1 : 0);
2092        if (dinode)
2093                gfs2_trans_add_unrevoke(sdp, block, 1);
2094
2095        /*
2096         * This needs reviewing to see why we cannot do the quota change
2097         * at this point in the dinode case.
2098         */
2099        if (ndata)
2100                gfs2_quota_change(ip, ndata, ip->i_inode.i_uid,
2101                                  ip->i_inode.i_gid);
2102
2103        rgd->rd_free_clone -= *nblocks;
2104        trace_gfs2_block_alloc(ip, rgd, block, *nblocks,
2105                               dinode ? GFS2_BLKST_DINODE : GFS2_BLKST_USED);
2106        *bn = block;
2107        return 0;
2108
2109rgrp_error:
2110        gfs2_rgrp_error(rgd);
2111        return -EIO;
2112}
2113
2114/**
2115 * __gfs2_free_blocks - free a contiguous run of block(s)
2116 * @ip: the inode these blocks are being freed from
2117 * @bstart: first block of a run of contiguous blocks
2118 * @blen: the length of the block run
2119 * @meta: 1 if the blocks represent metadata
2120 *
2121 */
2122
2123void __gfs2_free_blocks(struct gfs2_inode *ip, u64 bstart, u32 blen, int meta)
2124{
2125        struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
2126        struct gfs2_rgrpd *rgd;
2127
2128        rgd = rgblk_free(sdp, bstart, blen, GFS2_BLKST_FREE);
2129        if (!rgd)
2130                return;
2131        trace_gfs2_block_alloc(ip, rgd, bstart, blen, GFS2_BLKST_FREE);
2132        rgd->rd_free += blen;
2133        rgd->rd_flags &= ~GFS2_RGF_TRIMMED;
2134        gfs2_trans_add_bh(rgd->rd_gl, rgd->rd_bits[0].bi_bh, 1);
2135        gfs2_rgrp_out(rgd, rgd->rd_bits[0].bi_bh->b_data);
2136        gfs2_rgrp_ondisk2lvb(rgd->rd_rgl, rgd->rd_bits[0].bi_bh->b_data);
2137
2138        /* Directories keep their data in the metadata address space */
2139        if (meta || ip->i_depth)
2140                gfs2_meta_wipe(ip, bstart, blen);
2141}
2142
2143/**
2144 * gfs2_free_meta - free a contiguous run of data block(s)
2145 * @ip: the inode these blocks are being freed from
2146 * @bstart: first block of a run of contiguous blocks
2147 * @blen: the length of the block run
2148 *
2149 */
2150
2151void gfs2_free_meta(struct gfs2_inode *ip, u64 bstart, u32 blen)
2152{
2153        struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
2154
2155        __gfs2_free_blocks(ip, bstart, blen, 1);
2156        gfs2_statfs_change(sdp, 0, +blen, 0);
2157        gfs2_quota_change(ip, -(s64)blen, ip->i_inode.i_uid, ip->i_inode.i_gid);
2158}
2159
2160void gfs2_unlink_di(struct inode *inode)
2161{
2162        struct gfs2_inode *ip = GFS2_I(inode);
2163        struct gfs2_sbd *sdp = GFS2_SB(inode);
2164        struct gfs2_rgrpd *rgd;
2165        u64 blkno = ip->i_no_addr;
2166
2167        rgd = rgblk_free(sdp, blkno, 1, GFS2_BLKST_UNLINKED);
2168        if (!rgd)
2169                return;
2170        trace_gfs2_block_alloc(ip, rgd, blkno, 1, GFS2_BLKST_UNLINKED);
2171        gfs2_trans_add_bh(rgd->rd_gl, rgd->rd_bits[0].bi_bh, 1);
2172        gfs2_rgrp_out(rgd, rgd->rd_bits[0].bi_bh->b_data);
2173        gfs2_rgrp_ondisk2lvb(rgd->rd_rgl, rgd->rd_bits[0].bi_bh->b_data);
2174        update_rgrp_lvb_unlinked(rgd, 1);
2175}
2176
2177static void gfs2_free_uninit_di(struct gfs2_rgrpd *rgd, u64 blkno)
2178{
2179        struct gfs2_sbd *sdp = rgd->rd_sbd;
2180        struct gfs2_rgrpd *tmp_rgd;
2181
2182        tmp_rgd = rgblk_free(sdp, blkno, 1, GFS2_BLKST_FREE);
2183        if (!tmp_rgd)
2184                return;
2185        gfs2_assert_withdraw(sdp, rgd == tmp_rgd);
2186
2187        if (!rgd->rd_dinodes)
2188                gfs2_consist_rgrpd(rgd);
2189        rgd->rd_dinodes--;
2190        rgd->rd_free++;
2191
2192        gfs2_trans_add_bh(rgd->rd_gl, rgd->rd_bits[0].bi_bh, 1);
2193        gfs2_rgrp_out(rgd, rgd->rd_bits[0].bi_bh->b_data);
2194        gfs2_rgrp_ondisk2lvb(rgd->rd_rgl, rgd->rd_bits[0].bi_bh->b_data);
2195        update_rgrp_lvb_unlinked(rgd, -1);
2196
2197        gfs2_statfs_change(sdp, 0, +1, -1);
2198}
2199
2200
2201void gfs2_free_di(struct gfs2_rgrpd *rgd, struct gfs2_inode *ip)
2202{
2203        gfs2_free_uninit_di(rgd, ip->i_no_addr);
2204        trace_gfs2_block_alloc(ip, rgd, ip->i_no_addr, 1, GFS2_BLKST_FREE);
2205        gfs2_quota_change(ip, -1, ip->i_inode.i_uid, ip->i_inode.i_gid);
2206        gfs2_meta_wipe(ip, ip->i_no_addr, 1);
2207}
2208
2209/**
2210 * gfs2_check_blk_type - Check the type of a block
2211 * @sdp: The superblock
2212 * @no_addr: The block number to check
2213 * @type: The block type we are looking for
2214 *
2215 * Returns: 0 if the block type matches the expected type
2216 *          -ESTALE if it doesn't match
2217 *          or -ve errno if something went wrong while checking
2218 */
2219
2220int gfs2_check_blk_type(struct gfs2_sbd *sdp, u64 no_addr, unsigned int type)
2221{
2222        struct gfs2_rgrpd *rgd;
2223        struct gfs2_holder rgd_gh;
2224        int error = -EINVAL;
2225
2226        rgd = gfs2_blk2rgrpd(sdp, no_addr, 1);
2227        if (!rgd)
2228                goto fail;
2229
2230        error = gfs2_glock_nq_init(rgd->rd_gl, LM_ST_SHARED, 0, &rgd_gh);
2231        if (error)
2232                goto fail;
2233
2234        if (gfs2_get_block_type(rgd, no_addr) != type)
2235                error = -ESTALE;
2236
2237        gfs2_glock_dq_uninit(&rgd_gh);
2238fail:
2239        return error;
2240}
2241
2242/**
2243 * gfs2_rlist_add - add a RG to a list of RGs
2244 * @ip: the inode
2245 * @rlist: the list of resource groups
2246 * @block: the block
2247 *
2248 * Figure out what RG a block belongs to and add that RG to the list
2249 *
2250 * FIXME: Don't use NOFAIL
2251 *
2252 */
2253
2254void gfs2_rlist_add(struct gfs2_inode *ip, struct gfs2_rgrp_list *rlist,
2255                    u64 block)
2256{
2257        struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
2258        struct gfs2_rgrpd *rgd;
2259        struct gfs2_rgrpd **tmp;
2260        unsigned int new_space;
2261        unsigned int x;
2262
2263        if (gfs2_assert_warn(sdp, !rlist->rl_ghs))
2264                return;
2265
2266        if (ip->i_rgd && rgrp_contains_block(ip->i_rgd, block))
2267                rgd = ip->i_rgd;
2268        else
2269                rgd = gfs2_blk2rgrpd(sdp, block, 1);
2270        if (!rgd) {
2271                fs_err(sdp, "rlist_add: no rgrp for block %llu\n", (unsigned long long)block);
2272                return;
2273        }
2274        ip->i_rgd = rgd;
2275
2276        for (x = 0; x < rlist->rl_rgrps; x++)
2277                if (rlist->rl_rgd[x] == rgd)
2278                        return;
2279
2280        if (rlist->rl_rgrps == rlist->rl_space) {
2281                new_space = rlist->rl_space + 10;
2282
2283                tmp = kcalloc(new_space, sizeof(struct gfs2_rgrpd *),
2284                              GFP_NOFS | __GFP_NOFAIL);
2285
2286                if (rlist->rl_rgd) {
2287                        memcpy(tmp, rlist->rl_rgd,
2288                               rlist->rl_space * sizeof(struct gfs2_rgrpd *));
2289                        kfree(rlist->rl_rgd);
2290                }
2291
2292                rlist->rl_space = new_space;
2293                rlist->rl_rgd = tmp;
2294        }
2295
2296        rlist->rl_rgd[rlist->rl_rgrps++] = rgd;
2297}
2298
2299/**
2300 * gfs2_rlist_alloc - all RGs have been added to the rlist, now allocate
2301 *      and initialize an array of glock holders for them
2302 * @rlist: the list of resource groups
2303 * @state: the lock state to acquire the RG lock in
2304 *
2305 * FIXME: Don't use NOFAIL
2306 *
2307 */
2308
2309void gfs2_rlist_alloc(struct gfs2_rgrp_list *rlist, unsigned int state)
2310{
2311        unsigned int x;
2312
2313        rlist->rl_ghs = kcalloc(rlist->rl_rgrps, sizeof(struct gfs2_holder),
2314                                GFP_NOFS | __GFP_NOFAIL);
2315        for (x = 0; x < rlist->rl_rgrps; x++)
2316                gfs2_holder_init(rlist->rl_rgd[x]->rd_gl,
2317                                state, 0,
2318                                &rlist->rl_ghs[x]);
2319}
2320
2321/**
2322 * gfs2_rlist_free - free a resource group list
2323 * @list: the list of resource groups
2324 *
2325 */
2326
2327void gfs2_rlist_free(struct gfs2_rgrp_list *rlist)
2328{
2329        unsigned int x;
2330
2331        kfree(rlist->rl_rgd);
2332
2333        if (rlist->rl_ghs) {
2334                for (x = 0; x < rlist->rl_rgrps; x++)
2335                        gfs2_holder_uninit(&rlist->rl_ghs[x]);
2336                kfree(rlist->rl_ghs);
2337                rlist->rl_ghs = NULL;
2338        }
2339}
2340
2341
lxr.linux.no kindly hosted by Redpill Linpro AS, provider of Linux consulting and operations services since 1995.