git.sesse.net Git - bcachefs-tools-debian/blob - libbcachefs/rebalance.c

   1 // SPDX-License-Identifier: GPL-2.0
   2
   3 #include "bcachefs.h"
   4 #include "alloc_foreground.h"
   5 #include "btree_iter.h"
   6 #include "buckets.h"
   7 #include "clock.h"
   8 #include "compress.h"
   9 #include "disk_groups.h"
  10 #include "errcode.h"
  11 #include "move.h"
  12 #include "rebalance.h"
  13 #include "super-io.h"
  14 #include "trace.h"
  15
  16 #include <linux/freezer.h>
  17 #include <linux/kthread.h>
  18 #include <linux/sched/cputime.h>
  19
  20 /*
  21  * Check if an extent should be moved:
  22  * returns -1 if it should not be moved, or
  23  * device of pointer that should be moved, if known, or INT_MAX if unknown
  24  */
  25 static bool rebalance_pred(struct bch_fs *c, void *arg,
  26                            struct bkey_s_c k,
  27                            struct bch_io_opts *io_opts,
  28                            struct data_update_opts *data_opts)
  29 {
  30         struct bkey_ptrs_c ptrs = bch2_bkey_ptrs_c(k);
  31         unsigned i;
  32
  33         data_opts->rewrite_ptrs         = 0;
  34         data_opts->target               = io_opts->background_target;
  35         data_opts->extra_replicas       = 0;
  36         data_opts->btree_insert_flags   = 0;
  37
  38         if (io_opts->background_compression &&
  39             !bch2_bkey_is_incompressible(k)) {
  40                 const union bch_extent_entry *entry;
  41                 struct extent_ptr_decoded p;
  42
  43                 i = 0;
  44                 bkey_for_each_ptr_decode(k.k, ptrs, p, entry) {
  45                         if (!p.ptr.cached &&
  46                             p.crc.compression_type !=
  47                             bch2_compression_opt_to_type(io_opts->background_compression))
  48                                 data_opts->rewrite_ptrs |= 1U << i;
  49                         i++;
  50                 }
  51         }
  52
  53         if (io_opts->background_target) {
  54                 const struct bch_extent_ptr *ptr;
  55
  56                 i = 0;
  57                 bkey_for_each_ptr(ptrs, ptr) {
  58                         if (!ptr->cached &&
  59                             !bch2_dev_in_target(c, ptr->dev, io_opts->background_target) &&
  60                             bch2_target_accepts_data(c, BCH_DATA_user, io_opts->background_target))
  61                                 data_opts->rewrite_ptrs |= 1U << i;
  62                         i++;
  63                 }
  64         }
  65
  66         return data_opts->rewrite_ptrs != 0;
  67 }
  68
  69 void bch2_rebalance_add_key(struct bch_fs *c,
  70                             struct bkey_s_c k,
  71                             struct bch_io_opts *io_opts)
  72 {
  73         struct data_update_opts update_opts = { 0 };
  74         struct bkey_ptrs_c ptrs;
  75         const struct bch_extent_ptr *ptr;
  76         unsigned i;
  77
  78         if (!rebalance_pred(c, NULL, k, io_opts, &update_opts))
  79                 return;
  80
  81         i = 0;
  82         ptrs = bch2_bkey_ptrs_c(k);
  83         bkey_for_each_ptr(ptrs, ptr) {
  84                 if ((1U << i) && update_opts.rewrite_ptrs)
  85                         if (atomic64_add_return(k.k->size,
  86                                         &bch_dev_bkey_exists(c, ptr->dev)->rebalance_work) ==
  87                             k.k->size)
  88                                 rebalance_wakeup(c);
  89                 i++;
  90         }
  91 }
  92
  93 void bch2_rebalance_add_work(struct bch_fs *c, u64 sectors)
  94 {
  95         if (atomic64_add_return(sectors, &c->rebalance.work_unknown_dev) ==
  96             sectors)
  97                 rebalance_wakeup(c);
  98 }
  99
 100 struct rebalance_work {
 101         int             dev_most_full_idx;
 102         unsigned        dev_most_full_percent;
 103         u64             dev_most_full_work;
 104         u64             dev_most_full_capacity;
 105         u64             total_work;
 106 };
 107
 108 static void rebalance_work_accumulate(struct rebalance_work *w,
 109                 u64 dev_work, u64 unknown_dev, u64 capacity, int idx)
 110 {
 111         unsigned percent_full;
 112         u64 work = dev_work + unknown_dev;
 113
 114         /* avoid divide by 0 */
 115         if (!capacity)
 116                 return;
 117
 118         if (work < dev_work || work < unknown_dev)
 119                 work = U64_MAX;
 120         work = min(work, capacity);
 121
 122         percent_full = div64_u64(work * 100, capacity);
 123
 124         if (percent_full >= w->dev_most_full_percent) {
 125                 w->dev_most_full_idx            = idx;
 126                 w->dev_most_full_percent        = percent_full;
 127                 w->dev_most_full_work           = work;
 128                 w->dev_most_full_capacity       = capacity;
 129         }
 130
 131         if (w->total_work + dev_work >= w->total_work &&
 132             w->total_work + dev_work >= dev_work)
 133                 w->total_work += dev_work;
 134 }
 135
 136 static struct rebalance_work rebalance_work(struct bch_fs *c)
 137 {
 138         struct bch_dev *ca;
 139         struct rebalance_work ret = { .dev_most_full_idx = -1 };
 140         u64 unknown_dev = atomic64_read(&c->rebalance.work_unknown_dev);
 141         unsigned i;
 142
 143         for_each_online_member(ca, c, i)
 144                 rebalance_work_accumulate(&ret,
 145                         atomic64_read(&ca->rebalance_work),
 146                         unknown_dev,
 147                         bucket_to_sector(ca, ca->mi.nbuckets -
 148                                          ca->mi.first_bucket),
 149                         i);
 150
 151         rebalance_work_accumulate(&ret,
 152                 unknown_dev, 0, c->capacity, -1);
 153
 154         return ret;
 155 }
 156
 157 static void rebalance_work_reset(struct bch_fs *c)
 158 {
 159         struct bch_dev *ca;
 160         unsigned i;
 161
 162         for_each_online_member(ca, c, i)
 163                 atomic64_set(&ca->rebalance_work, 0);
 164
 165         atomic64_set(&c->rebalance.work_unknown_dev, 0);
 166 }
 167
 168 static unsigned long curr_cputime(void)
 169 {
 170         u64 utime, stime;
 171
 172         task_cputime_adjusted(current, &utime, &stime);
 173         return nsecs_to_jiffies(utime + stime);
 174 }
 175
 176 static int bch2_rebalance_thread(void *arg)
 177 {
 178         struct bch_fs *c = arg;
 179         struct bch_fs_rebalance *r = &c->rebalance;
 180         struct io_clock *clock = &c->io_clock[WRITE];
 181         struct rebalance_work w, p;
 182         struct bch_move_stats move_stats;
 183         unsigned long start, prev_start;
 184         unsigned long prev_run_time, prev_run_cputime;
 185         unsigned long cputime, prev_cputime;
 186         u64 io_start;
 187         long throttle;
 188
 189         set_freezable();
 190
 191         io_start        = atomic64_read(&clock->now);
 192         p               = rebalance_work(c);
 193         prev_start      = jiffies;
 194         prev_cputime    = curr_cputime();
 195
 196         bch2_move_stats_init(&move_stats, "rebalance");
 197         while (!kthread_wait_freezable(r->enabled)) {
 198                 cond_resched();
 199
 200                 start                   = jiffies;
 201                 cputime                 = curr_cputime();
 202
 203                 prev_run_time           = start - prev_start;
 204                 prev_run_cputime        = cputime - prev_cputime;
 205
 206                 w                       = rebalance_work(c);
 207                 BUG_ON(!w.dev_most_full_capacity);
 208
 209                 if (!w.total_work) {
 210                         r->state = REBALANCE_WAITING;
 211                         kthread_wait_freezable(rebalance_work(c).total_work);
 212                         continue;
 213                 }
 214
 215                 /*
 216                  * If there isn't much work to do, throttle cpu usage:
 217                  */
 218                 throttle = prev_run_cputime * 100 /
 219                         max(1U, w.dev_most_full_percent) -
 220                         prev_run_time;
 221
 222                 if (w.dev_most_full_percent < 20 && throttle > 0) {
 223                         r->throttled_until_iotime = io_start +
 224                                 div_u64(w.dev_most_full_capacity *
 225                                         (20 - w.dev_most_full_percent),
 226                                         50);
 227
 228                         if (atomic64_read(&clock->now) + clock->max_slop <
 229                             r->throttled_until_iotime) {
 230                                 r->throttled_until_cputime = start + throttle;
 231                                 r->state = REBALANCE_THROTTLED;
 232
 233                                 bch2_kthread_io_clock_wait(clock,
 234                                         r->throttled_until_iotime,
 235                                         throttle);
 236                                 continue;
 237                         }
 238                 }
 239
 240                 /* minimum 1 mb/sec: */
 241                 r->pd.rate.rate =
 242                         max_t(u64, 1 << 11,
 243                               r->pd.rate.rate *
 244                               max(p.dev_most_full_percent, 1U) /
 245                               max(w.dev_most_full_percent, 1U));
 246
 247                 io_start        = atomic64_read(&clock->now);
 248                 p               = w;
 249                 prev_start      = start;
 250                 prev_cputime    = cputime;
 251
 252                 r->state = REBALANCE_RUNNING;
 253                 memset(&move_stats, 0, sizeof(move_stats));
 254                 rebalance_work_reset(c);
 255
 256                 bch2_move_data(c,
 257                                0,               POS_MIN,
 258                                BTREE_ID_NR,     POS_MAX,
 259                                /* ratelimiting disabled for now */
 260                                NULL, /*  &r->pd.rate, */
 261                                &move_stats,
 262                                writepoint_ptr(&c->rebalance_write_point),
 263                                true,
 264                                rebalance_pred, NULL);
 265         }
 266
 267         return 0;
 268 }
 269
 270 void bch2_rebalance_work_to_text(struct printbuf *out, struct bch_fs *c)
 271 {
 272         struct bch_fs_rebalance *r = &c->rebalance;
 273         struct rebalance_work w = rebalance_work(c);
 274
 275         if (!out->nr_tabstops)
 276                 printbuf_tabstop_push(out, 20);
 277
 278         prt_printf(out, "fullest_dev (%i):", w.dev_most_full_idx);
 279         prt_tab(out);
 280
 281         prt_human_readable_u64(out, w.dev_most_full_work << 9);
 282         prt_printf(out, "/");
 283         prt_human_readable_u64(out, w.dev_most_full_capacity << 9);
 284         prt_newline(out);
 285
 286         prt_printf(out, "total work:");
 287         prt_tab(out);
 288
 289         prt_human_readable_u64(out, w.total_work << 9);
 290         prt_printf(out, "/");
 291         prt_human_readable_u64(out, c->capacity << 9);
 292         prt_newline(out);
 293
 294         prt_printf(out, "rate:");
 295         prt_tab(out);
 296         prt_printf(out, "%u", r->pd.rate.rate);
 297         prt_newline(out);
 298
 299         switch (r->state) {
 300         case REBALANCE_WAITING:
 301                 prt_printf(out, "waiting");
 302                 break;
 303         case REBALANCE_THROTTLED:
 304                 prt_printf(out, "throttled for %lu sec or ",
 305                        (r->throttled_until_cputime - jiffies) / HZ);
 306                 prt_human_readable_u64(out,
 307                             (r->throttled_until_iotime -
 308                              atomic64_read(&c->io_clock[WRITE].now)) << 9);
 309                 prt_printf(out, " io");
 310                 break;
 311         case REBALANCE_RUNNING:
 312                 prt_printf(out, "running");
 313                 break;
 314         }
 315         prt_newline(out);
 316 }
 317
 318 void bch2_rebalance_stop(struct bch_fs *c)
 319 {
 320         struct task_struct *p;
 321
 322         c->rebalance.pd.rate.rate = UINT_MAX;
 323         bch2_ratelimit_reset(&c->rebalance.pd.rate);
 324
 325         p = rcu_dereference_protected(c->rebalance.thread, 1);
 326         c->rebalance.thread = NULL;
 327
 328         if (p) {
 329                 /* for sychronizing with rebalance_wakeup() */
 330                 synchronize_rcu();
 331
 332                 kthread_stop(p);
 333                 put_task_struct(p);
 334         }
 335 }
 336
 337 int bch2_rebalance_start(struct bch_fs *c)
 338 {
 339         struct task_struct *p;
 340         int ret;
 341
 342         if (c->rebalance.thread)
 343                 return 0;
 344
 345         if (c->opts.nochanges)
 346                 return 0;
 347
 348         p = kthread_create(bch2_rebalance_thread, c, "bch-rebalance/%s", c->name);
 349         ret = PTR_ERR_OR_ZERO(p);
 350         if (ret) {
 351                 bch_err_msg(c, ret, "creating rebalance thread");
 352                 return ret;
 353         }
 354
 355         get_task_struct(p);
 356         rcu_assign_pointer(c->rebalance.thread, p);
 357         wake_up_process(p);
 358         return 0;
 359 }
 360
 361 void bch2_fs_rebalance_init(struct bch_fs *c)
 362 {
 363         bch2_pd_controller_init(&c->rebalance.pd);
 364
 365         atomic64_set(&c->rebalance.work_unknown_dev, S64_MAX);
 366 }