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 } 436