fs/bcachefs/movinggc.c

   1 // SPDX-License-Identifier: GPL-2.0
   2 /*
   3  * Moving/copying garbage collector
   4  *
   5  * Copyright 2012 Google, Inc.
   6  */
   7
   8 #include "bcachefs.h"
   9 #include "alloc_background.h"
  10 #include "alloc_foreground.h"
  11 #include "btree_iter.h"
  12 #include "btree_update.h"
  13 #include "btree_write_buffer.h"
  14 #include "buckets.h"
  15 #include "clock.h"
  16 #include "errcode.h"
  17 #include "error.h"
  18 #include "lru.h"
  19 #include "move.h"
  20 #include "movinggc.h"
  21 #include "trace.h"
  22
  23 #include <linux/freezer.h>
  24 #include <linux/kthread.h>
  25 #include <linux/math64.h>
  26 #include <linux/sched/task.h>
  27 #include <linux/wait.h>
  28
  29 struct buckets_in_flight {
  30         struct rhashtable               table;
  31         struct move_bucket_in_flight    *first;
  32         struct move_bucket_in_flight    *last;
  33         size_t                          nr;
  34         size_t                          sectors;
  35 };
  36
  37 static const struct rhashtable_params bch_move_bucket_params = {
  38         .head_offset    = offsetof(struct move_bucket_in_flight, hash),
  39         .key_offset     = offsetof(struct move_bucket_in_flight, bucket.k),
  40         .key_len        = sizeof(struct move_bucket_key),
  41 };
  42
  43 static struct move_bucket_in_flight *
  44 move_bucket_in_flight_add(struct buckets_in_flight *list, struct move_bucket b)
  45 {
  46         struct move_bucket_in_flight *new = kzalloc(sizeof(*new), GFP_KERNEL);
  47         int ret;
  48
  49         if (!new)
  50                 return ERR_PTR(-ENOMEM);
  51
  52         new->bucket = b;
  53
  54         ret = rhashtable_lookup_insert_fast(&list->table, &new->hash,
  55                                             bch_move_bucket_params);
  56         if (ret) {
  57                 kfree(new);
  58                 return ERR_PTR(ret);
  59         }
  60
  61         if (!list->first)
  62                 list->first = new;
  63         else
  64                 list->last->next = new;
  65
  66         list->last = new;
  67         list->nr++;
  68         list->sectors += b.sectors;
  69         return new;
  70 }
  71
  72 static int bch2_bucket_is_movable(struct btree_trans *trans,
  73                                   struct move_bucket *b, u64 time)
  74 {
  75         struct btree_iter iter;
  76         struct bkey_s_c k;
  77         struct bch_alloc_v4 _a;
  78         const struct bch_alloc_v4 *a;
  79         int ret;
  80
  81         if (bch2_bucket_is_open(trans->c,
  82                                 b->k.bucket.inode,
  83                                 b->k.bucket.offset))
  84                 return 0;
  85
  86         k = bch2_bkey_get_iter(trans, &iter, BTREE_ID_alloc,
  87                                b->k.bucket, BTREE_ITER_CACHED);
  88         ret = bkey_err(k);
  89         if (ret)
  90                 return ret;
  91
  92         a = bch2_alloc_to_v4(k, &_a);
  93         b->k.gen        = a->gen;
  94         b->sectors      = bch2_bucket_sectors_dirty(*a);
  95
  96         ret = data_type_movable(a->data_type) &&
  97                 a->fragmentation_lru &&
  98                 a->fragmentation_lru <= time;
  99
 100         bch2_trans_iter_exit(trans, &iter);
 101         return ret;
 102 }
 103
 104 static void move_buckets_wait(struct moving_context *ctxt,
 105                               struct buckets_in_flight *list,
 106                               bool flush)
 107 {
 108         struct move_bucket_in_flight *i;
 109         int ret;
 110
 111         while ((i = list->first)) {
 112                 if (flush)
 113                         move_ctxt_wait_event(ctxt, !atomic_read(&i->count));
 114
 115                 if (atomic_read(&i->count))
 116                         break;
 117
 118                 list->first = i->next;
 119                 if (!list->first)
 120                         list->last = NULL;
 121
 122                 list->nr--;
 123                 list->sectors -= i->bucket.sectors;
 124
 125                 ret = rhashtable_remove_fast(&list->table, &i->hash,
 126                                              bch_move_bucket_params);
 127                 BUG_ON(ret);
 128                 kfree(i);
 129         }
 130
 131         bch2_trans_unlock_long(ctxt->trans);
 132 }
 133
 134 static bool bucket_in_flight(struct buckets_in_flight *list,
 135                              struct move_bucket_key k)
 136 {
 137         return rhashtable_lookup_fast(&list->table, &k, bch_move_bucket_params);
 138 }
 139
 140 typedef DARRAY(struct move_bucket) move_buckets;
 141
 142 static int bch2_copygc_get_buckets(struct moving_context *ctxt,
 143                         struct buckets_in_flight *buckets_in_flight,
 144                         move_buckets *buckets)
 145 {
 146         struct btree_trans *trans = ctxt->trans;
 147         struct bch_fs *c = trans->c;
 148         size_t nr_to_get = max_t(size_t, 16U, buckets_in_flight->nr / 4);
 149         size_t saw = 0, in_flight = 0, not_movable = 0, sectors = 0;
 150         int ret;
 151
 152         move_buckets_wait(ctxt, buckets_in_flight, false);
 153
 154         ret = bch2_btree_write_buffer_tryflush(trans);
 155         if (bch2_err_matches(ret, EROFS))
 156                 return ret;
 157
 158         if (bch2_fs_fatal_err_on(ret, c, "%s: from bch2_btree_write_buffer_tryflush()", bch2_err_str(ret)))
 159                 return ret;
 160
 161         ret = for_each_btree_key_upto(trans, iter, BTREE_ID_lru,
 162                                   lru_pos(BCH_LRU_FRAGMENTATION_START, 0, 0),
 163                                   lru_pos(BCH_LRU_FRAGMENTATION_START, U64_MAX, LRU_TIME_MAX),
 164                                   0, k, ({
 165                 struct move_bucket b = { .k.bucket = u64_to_bucket(k.k->p.offset) };
 166                 int ret2 = 0;
 167
 168                 saw++;
 169
 170                 ret2 = bch2_bucket_is_movable(trans, &b, lru_pos_time(k.k->p));
 171                 if (ret2 < 0)
 172                         goto err;
 173
 174                 if (!ret2)
 175                         not_movable++;
 176                 else if (bucket_in_flight(buckets_in_flight, b.k))
 177                         in_flight++;
 178                 else {
 179                         ret2 = darray_push(buckets, b);
 180                         if (ret2)
 181                                 goto err;
 182                         sectors += b.sectors;
 183                 }
 184
 185                 ret2 = buckets->nr >= nr_to_get;
 186 err:
 187                 ret2;
 188         }));
 189
 190         pr_debug("have: %zu (%zu) saw %zu in flight %zu not movable %zu got %zu (%zu)/%zu buckets ret %i",
 191                  buckets_in_flight->nr, buckets_in_flight->sectors,
 192                  saw, in_flight, not_movable, buckets->nr, sectors, nr_to_get, ret);
 193
 194         return ret < 0 ? ret : 0;
 195 }
 196
 197 noinline
 198 static int bch2_copygc(struct moving_context *ctxt,
 199                        struct buckets_in_flight *buckets_in_flight,
 200                        bool *did_work)
 201 {
 202         struct btree_trans *trans = ctxt->trans;
 203         struct bch_fs *c = trans->c;
 204         struct data_update_opts data_opts = {
 205                 .btree_insert_flags = BCH_WATERMARK_copygc,
 206         };
 207         move_buckets buckets = { 0 };
 208         struct move_bucket_in_flight *f;
 209         u64 moved = atomic64_read(&ctxt->stats->sectors_moved);
 210         int ret = 0;
 211
 212         ret = bch2_copygc_get_buckets(ctxt, buckets_in_flight, &buckets);
 213         if (ret)
 214                 goto err;
 215
 216         darray_for_each(buckets, i) {
 217                 if (kthread_should_stop() || freezing(current))
 218                         break;
 219
 220                 f = move_bucket_in_flight_add(buckets_in_flight, *i);
 221                 ret = PTR_ERR_OR_ZERO(f);
 222                 if (ret == -EEXIST) { /* rare race: copygc_get_buckets returned same bucket more than once */
 223                         ret = 0;
 224                         continue;
 225                 }
 226                 if (ret == -ENOMEM) { /* flush IO, continue later */
 227                         ret = 0;
 228                         break;
 229                 }
 230
 231                 ret = bch2_evacuate_bucket(ctxt, f, f->bucket.k.bucket,
 232                                              f->bucket.k.gen, data_opts);
 233                 if (ret)
 234                         goto err;
 235
 236                 *did_work = true;
 237         }
 238 err:
 239         darray_exit(&buckets);
 240
 241         /* no entries in LRU btree found, or got to end: */
 242         if (bch2_err_matches(ret, ENOENT))
 243                 ret = 0;
 244
 245         if (ret < 0 && !bch2_err_matches(ret, EROFS))
 246                 bch_err_msg(c, ret, "from bch2_move_data()");
 247
 248         moved = atomic64_read(&ctxt->stats->sectors_moved) - moved;
 249         trace_and_count(c, copygc, c, moved, 0, 0, 0);
 250         return ret;
 251 }
 252
 253 /*
 254  * Copygc runs when the amount of fragmented data is above some arbitrary
 255  * threshold:
 256  *
 257  * The threshold at the limit - when the device is full - is the amount of space
 258  * we reserved in bch2_recalc_capacity; we can't have more than that amount of
 259  * disk space stranded due to fragmentation and store everything we have
 260  * promised to store.
 261  *
 262  * But we don't want to be running copygc unnecessarily when the device still
 263  * has plenty of free space - rather, we want copygc to smoothly run every so
 264  * often and continually reduce the amount of fragmented space as the device
 265  * fills up. So, we increase the threshold by half the current free space.
 266  */
 267 unsigned long bch2_copygc_wait_amount(struct bch_fs *c)
 268 {
 269         s64 wait = S64_MAX, fragmented_allowed, fragmented;
 270
 271         for_each_rw_member(c, ca) {
 272                 struct bch_dev_usage usage = bch2_dev_usage_read(ca);
 273
 274                 fragmented_allowed = ((__dev_buckets_available(ca, usage, BCH_WATERMARK_stripe) *
 275                                        ca->mi.bucket_size) >> 1);
 276                 fragmented = 0;
 277
 278                 for (unsigned i = 0; i < BCH_DATA_NR; i++)
 279                         if (data_type_movable(i))
 280                                 fragmented += usage.d[i].fragmented;
 281
 282                 wait = min(wait, max(0LL, fragmented_allowed - fragmented));
 283         }
 284
 285         return wait;
 286 }
 287
 288 void bch2_copygc_wait_to_text(struct printbuf *out, struct bch_fs *c)
 289 {
 290         prt_printf(out, "Currently waiting for:     ");
 291         prt_human_readable_u64(out, max(0LL, c->copygc_wait -
 292                                         atomic64_read(&c->io_clock[WRITE].now)) << 9);
 293         prt_newline(out);
 294
 295         prt_printf(out, "Currently waiting since:   ");
 296         prt_human_readable_u64(out, max(0LL,
 297                                         atomic64_read(&c->io_clock[WRITE].now) -
 298                                         c->copygc_wait_at) << 9);
 299         prt_newline(out);
 300
 301         prt_printf(out, "Currently calculated wait: ");
 302         prt_human_readable_u64(out, bch2_copygc_wait_amount(c));
 303         prt_newline(out);
 304 }
 305
 306 static int bch2_copygc_thread(void *arg)
 307 {
 308         struct bch_fs *c = arg;
 309         struct moving_context ctxt;
 310         struct bch_move_stats move_stats;
 311         struct io_clock *clock = &c->io_clock[WRITE];
 312         struct buckets_in_flight *buckets;
 313         u64 last, wait;
 314         int ret = 0;
 315
 316         buckets = kzalloc(sizeof(struct buckets_in_flight), GFP_KERNEL);
 317         if (!buckets)
 318                 return -ENOMEM;
 319         ret = rhashtable_init(&buckets->table, &bch_move_bucket_params);
 320         bch_err_msg(c, ret, "allocating copygc buckets in flight");
 321         if (ret) {
 322                 kfree(buckets);
 323                 return ret;
 324         }
 325
 326         set_freezable();
 327
 328         bch2_move_stats_init(&move_stats, "copygc");
 329         bch2_moving_ctxt_init(&ctxt, c, NULL, &move_stats,
 330                               writepoint_ptr(&c->copygc_write_point),
 331                               false);
 332
 333         while (!ret && !kthread_should_stop()) {
 334                 bool did_work = false;
 335
 336                 bch2_trans_unlock_long(ctxt.trans);
 337                 cond_resched();
 338
 339                 if (!c->copy_gc_enabled) {
 340                         move_buckets_wait(&ctxt, buckets, true);
 341                         kthread_wait_freezable(c->copy_gc_enabled ||
 342                                                kthread_should_stop());
 343                 }
 344
 345                 if (unlikely(freezing(current))) {
 346                         move_buckets_wait(&ctxt, buckets, true);
 347                         __refrigerator(false);
 348                         continue;
 349                 }
 350
 351                 last = atomic64_read(&clock->now);
 352                 wait = bch2_copygc_wait_amount(c);
 353
 354                 if (wait > clock->max_slop) {
 355                         c->copygc_wait_at = last;
 356                         c->copygc_wait = last + wait;
 357                         move_buckets_wait(&ctxt, buckets, true);
 358                         trace_and_count(c, copygc_wait, c, wait, last + wait);
 359                         bch2_kthread_io_clock_wait(clock, last + wait,
 360                                         MAX_SCHEDULE_TIMEOUT);
 361                         continue;
 362                 }
 363
 364                 c->copygc_wait = 0;
 365
 366                 c->copygc_running = true;
 367                 ret = bch2_copygc(&ctxt, buckets, &did_work);
 368                 c->copygc_running = false;
 369
 370                 wake_up(&c->copygc_running_wq);
 371
 372                 if (!wait && !did_work) {
 373                         u64 min_member_capacity = bch2_min_rw_member_capacity(c);
 374
 375                         if (min_member_capacity == U64_MAX)
 376                                 min_member_capacity = 128 * 2048;
 377
 378                         bch2_trans_unlock_long(ctxt.trans);
 379                         bch2_kthread_io_clock_wait(clock, last + (min_member_capacity >> 6),
 380                                         MAX_SCHEDULE_TIMEOUT);
 381                 }
 382         }
 383
 384         move_buckets_wait(&ctxt, buckets, true);
 385
 386         rhashtable_destroy(&buckets->table);
 387         kfree(buckets);
 388         bch2_moving_ctxt_exit(&ctxt);
 389         bch2_move_stats_exit(&move_stats, c);
 390
 391         return 0;
 392 }
 393
 394 void bch2_copygc_stop(struct bch_fs *c)
 395 {
 396         if (c->copygc_thread) {
 397                 kthread_stop(c->copygc_thread);
 398                 put_task_struct(c->copygc_thread);
 399         }
 400         c->copygc_thread = NULL;
 401 }
 402
 403 int bch2_copygc_start(struct bch_fs *c)
 404 {
 405         struct task_struct *t;
 406         int ret;
 407
 408         if (c->copygc_thread)
 409                 return 0;
 410
 411         if (c->opts.nochanges)
 412                 return 0;
 413
 414         if (bch2_fs_init_fault("copygc_start"))
 415                 return -ENOMEM;
 416
 417         t = kthread_create(bch2_copygc_thread, c, "bch-copygc/%s", c->name);
 418         ret = PTR_ERR_OR_ZERO(t);
 419         bch_err_msg(c, ret, "creating copygc thread");
 420         if (ret)
 421                 return ret;
 422
 423         get_task_struct(t);
 424
 425         c->copygc_thread = t;
 426         wake_up_process(c->copygc_thread);
 427
 428         return 0;
 429 }
 430
 431 void bch2_fs_copygc_init(struct bch_fs *c)
 432 {
 433         init_waitqueue_head(&c->copygc_running_wq);
 434         c->copygc_running = false;
 435 }