fs/bcachefs/chardev.c

   1 // SPDX-License-Identifier: GPL-2.0
   2 #ifndef NO_BCACHEFS_CHARDEV
   3
   4 #include "bcachefs.h"
   5 #include "bcachefs_ioctl.h"
   6 #include "buckets.h"
   7 #include "chardev.h"
   8 #include "disk_accounting.h"
   9 #include "journal.h"
  10 #include "move.h"
  11 #include "recovery_passes.h"
  12 #include "replicas.h"
  13 #include "super.h"
  14 #include "super-io.h"
  15 #include "thread_with_file.h"
  16
  17 #include <linux/cdev.h>
  18 #include <linux/device.h>
  19 #include <linux/fs.h>
  20 #include <linux/ioctl.h>
  21 #include <linux/major.h>
  22 #include <linux/sched/task.h>
  23 #include <linux/slab.h>
  24 #include <linux/uaccess.h>
  25
  26 /* returns with ref on ca->ref */
  27 static struct bch_dev *bch2_device_lookup(struct bch_fs *c, u64 dev,
  28                                           unsigned flags)
  29 {
  30         struct bch_dev *ca;
  31
  32         if (flags & BCH_BY_INDEX) {
  33                 if (dev >= c->sb.nr_devices)
  34                         return ERR_PTR(-EINVAL);
  35
  36                 ca = bch2_dev_tryget_noerror(c, dev);
  37                 if (!ca)
  38                         return ERR_PTR(-EINVAL);
  39         } else {
  40                 char *path;
  41
  42                 path = strndup_user((const char __user *)
  43                                     (unsigned long) dev, PATH_MAX);
  44                 if (IS_ERR(path))
  45                         return ERR_CAST(path);
  46
  47                 ca = bch2_dev_lookup(c, path);
  48                 kfree(path);
  49         }
  50
  51         return ca;
  52 }
  53
  54 #if 0
  55 static long bch2_ioctl_assemble(struct bch_ioctl_assemble __user *user_arg)
  56 {
  57         struct bch_ioctl_assemble arg;
  58         struct bch_fs *c;
  59         u64 *user_devs = NULL;
  60         char **devs = NULL;
  61         unsigned i;
  62         int ret = -EFAULT;
  63
  64         if (copy_from_user(&arg, user_arg, sizeof(arg)))
  65                 return -EFAULT;
  66
  67         if (arg.flags || arg.pad)
  68                 return -EINVAL;
  69
  70         user_devs = kmalloc_array(arg.nr_devs, sizeof(u64), GFP_KERNEL);
  71         if (!user_devs)
  72                 return -ENOMEM;
  73
  74         devs = kcalloc(arg.nr_devs, sizeof(char *), GFP_KERNEL);
  75
  76         if (copy_from_user(user_devs, user_arg->devs,
  77                            sizeof(u64) * arg.nr_devs))
  78                 goto err;
  79
  80         for (i = 0; i < arg.nr_devs; i++) {
  81                 devs[i] = strndup_user((const char __user *)(unsigned long)
  82                                        user_devs[i],
  83                                        PATH_MAX);
  84                 ret= PTR_ERR_OR_ZERO(devs[i]);
  85                 if (ret)
  86                         goto err;
  87         }
  88
  89         c = bch2_fs_open(devs, arg.nr_devs, bch2_opts_empty());
  90         ret = PTR_ERR_OR_ZERO(c);
  91         if (!ret)
  92                 closure_put(&c->cl);
  93 err:
  94         if (devs)
  95                 for (i = 0; i < arg.nr_devs; i++)
  96                         kfree(devs[i]);
  97         kfree(devs);
  98         return ret;
  99 }
 100
 101 static long bch2_ioctl_incremental(struct bch_ioctl_incremental __user *user_arg)
 102 {
 103         struct bch_ioctl_incremental arg;
 104         const char *err;
 105         char *path;
 106
 107         if (copy_from_user(&arg, user_arg, sizeof(arg)))
 108                 return -EFAULT;
 109
 110         if (arg.flags || arg.pad)
 111                 return -EINVAL;
 112
 113         path = strndup_user((const char __user *)(unsigned long) arg.dev, PATH_MAX);
 114         ret = PTR_ERR_OR_ZERO(path);
 115         if (ret)
 116                 return ret;
 117
 118         err = bch2_fs_open_incremental(path);
 119         kfree(path);
 120
 121         if (err) {
 122                 pr_err("Could not register bcachefs devices: %s", err);
 123                 return -EINVAL;
 124         }
 125
 126         return 0;
 127 }
 128 #endif
 129
 130 struct fsck_thread {
 131         struct thread_with_stdio thr;
 132         struct bch_fs           *c;
 133         struct bch_opts         opts;
 134 };
 135
 136 static void bch2_fsck_thread_exit(struct thread_with_stdio *_thr)
 137 {
 138         struct fsck_thread *thr = container_of(_thr, struct fsck_thread, thr);
 139         kfree(thr);
 140 }
 141
 142 static int bch2_fsck_offline_thread_fn(struct thread_with_stdio *stdio)
 143 {
 144         struct fsck_thread *thr = container_of(stdio, struct fsck_thread, thr);
 145         struct bch_fs *c = thr->c;
 146
 147         int ret = PTR_ERR_OR_ZERO(c);
 148         if (ret)
 149                 return ret;
 150
 151         ret = bch2_fs_start(thr->c);
 152         if (ret)
 153                 goto err;
 154
 155         if (test_bit(BCH_FS_errors_fixed, &c->flags)) {
 156                 bch2_stdio_redirect_printf(&stdio->stdio, false, "%s: errors fixed\n", c->name);
 157                 ret |= 1;
 158         }
 159         if (test_bit(BCH_FS_error, &c->flags)) {
 160                 bch2_stdio_redirect_printf(&stdio->stdio, false, "%s: still has errors\n", c->name);
 161                 ret |= 4;
 162         }
 163 err:
 164         bch2_fs_stop(c);
 165         return ret;
 166 }
 167
 168 static const struct thread_with_stdio_ops bch2_offline_fsck_ops = {
 169         .exit           = bch2_fsck_thread_exit,
 170         .fn             = bch2_fsck_offline_thread_fn,
 171 };
 172
 173 static long bch2_ioctl_fsck_offline(struct bch_ioctl_fsck_offline __user *user_arg)
 174 {
 175         struct bch_ioctl_fsck_offline arg;
 176         struct fsck_thread *thr = NULL;
 177         darray_str(devs) = {};
 178         long ret = 0;
 179
 180         if (copy_from_user(&arg, user_arg, sizeof(arg)))
 181                 return -EFAULT;
 182
 183         if (arg.flags)
 184                 return -EINVAL;
 185
 186         if (!capable(CAP_SYS_ADMIN))
 187                 return -EPERM;
 188
 189         for (size_t i = 0; i < arg.nr_devs; i++) {
 190                 u64 dev_u64;
 191                 ret = copy_from_user_errcode(&dev_u64, &user_arg->devs[i], sizeof(u64));
 192                 if (ret)
 193                         goto err;
 194
 195                 char *dev_str = strndup_user((char __user *)(unsigned long) dev_u64, PATH_MAX);
 196                 ret = PTR_ERR_OR_ZERO(dev_str);
 197                 if (ret)
 198                         goto err;
 199
 200                 ret = darray_push(&devs, dev_str);
 201                 if (ret) {
 202                         kfree(dev_str);
 203                         goto err;
 204                 }
 205         }
 206
 207         thr = kzalloc(sizeof(*thr), GFP_KERNEL);
 208         if (!thr) {
 209                 ret = -ENOMEM;
 210                 goto err;
 211         }
 212
 213         thr->opts = bch2_opts_empty();
 214
 215         if (arg.opts) {
 216                 char *optstr = strndup_user((char __user *)(unsigned long) arg.opts, 1 << 16);
 217                 ret =   PTR_ERR_OR_ZERO(optstr) ?:
 218                         bch2_parse_mount_opts(NULL, &thr->opts, NULL, optstr);
 219                 if (!IS_ERR(optstr))
 220                         kfree(optstr);
 221
 222                 if (ret)
 223                         goto err;
 224         }
 225
 226         opt_set(thr->opts, stdio, (u64)(unsigned long)&thr->thr.stdio);
 227         opt_set(thr->opts, read_only, 1);
 228         opt_set(thr->opts, ratelimit_errors, 0);
 229
 230         /* We need request_key() to be called before we punt to kthread: */
 231         opt_set(thr->opts, nostart, true);
 232
 233         bch2_thread_with_stdio_init(&thr->thr, &bch2_offline_fsck_ops);
 234
 235         thr->c = bch2_fs_open(devs.data, arg.nr_devs, thr->opts);
 236
 237         if (!IS_ERR(thr->c) &&
 238             thr->c->opts.errors == BCH_ON_ERROR_panic)
 239                 thr->c->opts.errors = BCH_ON_ERROR_ro;
 240
 241         ret = __bch2_run_thread_with_stdio(&thr->thr);
 242 out:
 243         darray_for_each(devs, i)
 244                 kfree(*i);
 245         darray_exit(&devs);
 246         return ret;
 247 err:
 248         if (thr)
 249                 bch2_fsck_thread_exit(&thr->thr);
 250         pr_err("ret %s", bch2_err_str(ret));
 251         goto out;
 252 }
 253
 254 static long bch2_global_ioctl(unsigned cmd, void __user *arg)
 255 {
 256         long ret;
 257
 258         switch (cmd) {
 259 #if 0
 260         case BCH_IOCTL_ASSEMBLE:
 261                 return bch2_ioctl_assemble(arg);
 262         case BCH_IOCTL_INCREMENTAL:
 263                 return bch2_ioctl_incremental(arg);
 264 #endif
 265         case BCH_IOCTL_FSCK_OFFLINE: {
 266                 ret = bch2_ioctl_fsck_offline(arg);
 267                 break;
 268         }
 269         default:
 270                 ret = -ENOTTY;
 271                 break;
 272         }
 273
 274         if (ret < 0)
 275                 ret = bch2_err_class(ret);
 276         return ret;
 277 }
 278
 279 static long bch2_ioctl_query_uuid(struct bch_fs *c,
 280                         struct bch_ioctl_query_uuid __user *user_arg)
 281 {
 282         return copy_to_user_errcode(&user_arg->uuid, &c->sb.user_uuid,
 283                                     sizeof(c->sb.user_uuid));
 284 }
 285
 286 #if 0
 287 static long bch2_ioctl_start(struct bch_fs *c, struct bch_ioctl_start arg)
 288 {
 289         if (!capable(CAP_SYS_ADMIN))
 290                 return -EPERM;
 291
 292         if (arg.flags || arg.pad)
 293                 return -EINVAL;
 294
 295         return bch2_fs_start(c);
 296 }
 297
 298 static long bch2_ioctl_stop(struct bch_fs *c)
 299 {
 300         if (!capable(CAP_SYS_ADMIN))
 301                 return -EPERM;
 302
 303         bch2_fs_stop(c);
 304         return 0;
 305 }
 306 #endif
 307
 308 static long bch2_ioctl_disk_add(struct bch_fs *c, struct bch_ioctl_disk arg)
 309 {
 310         char *path;
 311         int ret;
 312
 313         if (!capable(CAP_SYS_ADMIN))
 314                 return -EPERM;
 315
 316         if (arg.flags || arg.pad)
 317                 return -EINVAL;
 318
 319         path = strndup_user((const char __user *)(unsigned long) arg.dev, PATH_MAX);
 320         ret = PTR_ERR_OR_ZERO(path);
 321         if (ret)
 322                 return ret;
 323
 324         ret = bch2_dev_add(c, path);
 325         if (!IS_ERR(path))
 326                 kfree(path);
 327
 328         return ret;
 329 }
 330
 331 static long bch2_ioctl_disk_remove(struct bch_fs *c, struct bch_ioctl_disk arg)
 332 {
 333         struct bch_dev *ca;
 334
 335         if (!capable(CAP_SYS_ADMIN))
 336                 return -EPERM;
 337
 338         if ((arg.flags & ~(BCH_FORCE_IF_DATA_LOST|
 339                            BCH_FORCE_IF_METADATA_LOST|
 340                            BCH_FORCE_IF_DEGRADED|
 341                            BCH_BY_INDEX)) ||
 342             arg.pad)
 343                 return -EINVAL;
 344
 345         ca = bch2_device_lookup(c, arg.dev, arg.flags);
 346         if (IS_ERR(ca))
 347                 return PTR_ERR(ca);
 348
 349         return bch2_dev_remove(c, ca, arg.flags);
 350 }
 351
 352 static long bch2_ioctl_disk_online(struct bch_fs *c, struct bch_ioctl_disk arg)
 353 {
 354         char *path;
 355         int ret;
 356
 357         if (!capable(CAP_SYS_ADMIN))
 358                 return -EPERM;
 359
 360         if (arg.flags || arg.pad)
 361                 return -EINVAL;
 362
 363         path = strndup_user((const char __user *)(unsigned long) arg.dev, PATH_MAX);
 364         ret = PTR_ERR_OR_ZERO(path);
 365         if (ret)
 366                 return ret;
 367
 368         ret = bch2_dev_online(c, path);
 369         kfree(path);
 370         return ret;
 371 }
 372
 373 static long bch2_ioctl_disk_offline(struct bch_fs *c, struct bch_ioctl_disk arg)
 374 {
 375         struct bch_dev *ca;
 376         int ret;
 377
 378         if (!capable(CAP_SYS_ADMIN))
 379                 return -EPERM;
 380
 381         if ((arg.flags & ~(BCH_FORCE_IF_DATA_LOST|
 382                            BCH_FORCE_IF_METADATA_LOST|
 383                            BCH_FORCE_IF_DEGRADED|
 384                            BCH_BY_INDEX)) ||
 385             arg.pad)
 386                 return -EINVAL;
 387
 388         ca = bch2_device_lookup(c, arg.dev, arg.flags);
 389         if (IS_ERR(ca))
 390                 return PTR_ERR(ca);
 391
 392         ret = bch2_dev_offline(c, ca, arg.flags);
 393         bch2_dev_put(ca);
 394         return ret;
 395 }
 396
 397 static long bch2_ioctl_disk_set_state(struct bch_fs *c,
 398                         struct bch_ioctl_disk_set_state arg)
 399 {
 400         struct bch_dev *ca;
 401         int ret;
 402
 403         if (!capable(CAP_SYS_ADMIN))
 404                 return -EPERM;
 405
 406         if ((arg.flags & ~(BCH_FORCE_IF_DATA_LOST|
 407                            BCH_FORCE_IF_METADATA_LOST|
 408                            BCH_FORCE_IF_DEGRADED|
 409                            BCH_BY_INDEX)) ||
 410             arg.pad[0] || arg.pad[1] || arg.pad[2] ||
 411             arg.new_state >= BCH_MEMBER_STATE_NR)
 412                 return -EINVAL;
 413
 414         ca = bch2_device_lookup(c, arg.dev, arg.flags);
 415         if (IS_ERR(ca))
 416                 return PTR_ERR(ca);
 417
 418         ret = bch2_dev_set_state(c, ca, arg.new_state, arg.flags);
 419         if (ret)
 420                 bch_err(c, "Error setting device state: %s", bch2_err_str(ret));
 421
 422         bch2_dev_put(ca);
 423         return ret;
 424 }
 425
 426 struct bch_data_ctx {
 427         struct thread_with_file         thr;
 428
 429         struct bch_fs                   *c;
 430         struct bch_ioctl_data           arg;
 431         struct bch_move_stats           stats;
 432 };
 433
 434 static int bch2_data_thread(void *arg)
 435 {
 436         struct bch_data_ctx *ctx = container_of(arg, struct bch_data_ctx, thr);
 437
 438         ctx->thr.ret = bch2_data_job(ctx->c, &ctx->stats, ctx->arg);
 439         ctx->stats.data_type = U8_MAX;
 440         return 0;
 441 }
 442
 443 static int bch2_data_job_release(struct inode *inode, struct file *file)
 444 {
 445         struct bch_data_ctx *ctx = container_of(file->private_data, struct bch_data_ctx, thr);
 446
 447         bch2_thread_with_file_exit(&ctx->thr);
 448         kfree(ctx);
 449         return 0;
 450 }
 451
 452 static ssize_t bch2_data_job_read(struct file *file, char __user *buf,
 453                                   size_t len, loff_t *ppos)
 454 {
 455         struct bch_data_ctx *ctx = container_of(file->private_data, struct bch_data_ctx, thr);
 456         struct bch_fs *c = ctx->c;
 457         struct bch_ioctl_data_event e = {
 458                 .type                   = BCH_DATA_EVENT_PROGRESS,
 459                 .p.data_type            = ctx->stats.data_type,
 460                 .p.btree_id             = ctx->stats.pos.btree,
 461                 .p.pos                  = ctx->stats.pos.pos,
 462                 .p.sectors_done         = atomic64_read(&ctx->stats.sectors_seen),
 463                 .p.sectors_total        = bch2_fs_usage_read_short(c).used,
 464         };
 465
 466         if (len < sizeof(e))
 467                 return -EINVAL;
 468
 469         return copy_to_user_errcode(buf, &e, sizeof(e)) ?: sizeof(e);
 470 }
 471
 472 static const struct file_operations bcachefs_data_ops = {
 473         .release        = bch2_data_job_release,
 474         .read           = bch2_data_job_read,
 475 };
 476
 477 static long bch2_ioctl_data(struct bch_fs *c,
 478                             struct bch_ioctl_data arg)
 479 {
 480         struct bch_data_ctx *ctx;
 481         int ret;
 482
 483         if (!capable(CAP_SYS_ADMIN))
 484                 return -EPERM;
 485
 486         if (arg.op >= BCH_DATA_OP_NR || arg.flags)
 487                 return -EINVAL;
 488
 489         ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
 490         if (!ctx)
 491                 return -ENOMEM;
 492
 493         ctx->c = c;
 494         ctx->arg = arg;
 495
 496         ret = bch2_run_thread_with_file(&ctx->thr,
 497                         &bcachefs_data_ops,
 498                         bch2_data_thread);
 499         if (ret < 0)
 500                 kfree(ctx);
 501         return ret;
 502 }
 503
 504 static long bch2_ioctl_fs_usage(struct bch_fs *c,
 505                                 struct bch_ioctl_fs_usage __user *user_arg)
 506 {
 507         struct bch_ioctl_fs_usage arg = {};
 508         darray_char replicas = {};
 509         u32 replica_entries_bytes;
 510         int ret = 0;
 511
 512         if (!test_bit(BCH_FS_started, &c->flags))
 513                 return -EINVAL;
 514
 515         if (get_user(replica_entries_bytes, &user_arg->replica_entries_bytes))
 516                 return -EFAULT;
 517
 518         ret   = bch2_fs_replicas_usage_read(c, &replicas) ?:
 519                 (replica_entries_bytes < replicas.nr ? -ERANGE : 0) ?:
 520                 copy_to_user_errcode(&user_arg->replicas, replicas.data, replicas.nr);
 521         if (ret)
 522                 goto err;
 523
 524         struct bch_fs_usage_short u = bch2_fs_usage_read_short(c);
 525         arg.capacity            = c->capacity;
 526         arg.used                = u.used;
 527         arg.online_reserved     = percpu_u64_get(c->online_reserved);
 528         arg.replica_entries_bytes = replicas.nr;
 529
 530         for (unsigned i = 0; i < BCH_REPLICAS_MAX; i++) {
 531                 struct disk_accounting_pos k = {
 532                         .type = BCH_DISK_ACCOUNTING_persistent_reserved,
 533                         .persistent_reserved.nr_replicas = i,
 534                 };
 535
 536                 bch2_accounting_mem_read(c,
 537                                          disk_accounting_pos_to_bpos(&k),
 538                                          &arg.persistent_reserved[i], 1);
 539         }
 540
 541         ret = copy_to_user_errcode(user_arg, &arg, sizeof(arg));
 542 err:
 543         darray_exit(&replicas);
 544         return ret;
 545 }
 546
 547 static long bch2_ioctl_query_accounting(struct bch_fs *c,
 548                         struct bch_ioctl_query_accounting __user *user_arg)
 549 {
 550         struct bch_ioctl_query_accounting arg;
 551         darray_char accounting = {};
 552         int ret = 0;
 553
 554         if (!test_bit(BCH_FS_started, &c->flags))
 555                 return -EINVAL;
 556
 557         ret   = copy_from_user_errcode(&arg, user_arg, sizeof(arg)) ?:
 558                 bch2_fs_accounting_read(c, &accounting, arg.accounting_types_mask) ?:
 559                 (arg.accounting_u64s * sizeof(u64) < accounting.nr ? -ERANGE : 0) ?:
 560                 copy_to_user_errcode(&user_arg->accounting, accounting.data, accounting.nr);
 561         if (ret)
 562                 goto err;
 563
 564         arg.capacity            = c->capacity;
 565         arg.used                = bch2_fs_usage_read_short(c).used;
 566         arg.online_reserved     = percpu_u64_get(c->online_reserved);
 567         arg.accounting_u64s     = accounting.nr / sizeof(u64);
 568
 569         ret = copy_to_user_errcode(user_arg, &arg, sizeof(arg));
 570 err:
 571         darray_exit(&accounting);
 572         return ret;
 573 }
 574
 575 /* obsolete, didn't allow for new data types: */
 576 static long bch2_ioctl_dev_usage(struct bch_fs *c,
 577                                  struct bch_ioctl_dev_usage __user *user_arg)
 578 {
 579         struct bch_ioctl_dev_usage arg;
 580         struct bch_dev_usage src;
 581         struct bch_dev *ca;
 582         unsigned i;
 583
 584         if (!test_bit(BCH_FS_started, &c->flags))
 585                 return -EINVAL;
 586
 587         if (copy_from_user(&arg, user_arg, sizeof(arg)))
 588                 return -EFAULT;
 589
 590         if ((arg.flags & ~BCH_BY_INDEX) ||
 591             arg.pad[0] ||
 592             arg.pad[1] ||
 593             arg.pad[2])
 594                 return -EINVAL;
 595
 596         ca = bch2_device_lookup(c, arg.dev, arg.flags);
 597         if (IS_ERR(ca))
 598                 return PTR_ERR(ca);
 599
 600         src = bch2_dev_usage_read(ca);
 601
 602         arg.state               = ca->mi.state;
 603         arg.bucket_size         = ca->mi.bucket_size;
 604         arg.nr_buckets          = ca->mi.nbuckets - ca->mi.first_bucket;
 605
 606         for (i = 0; i < ARRAY_SIZE(arg.d); i++) {
 607                 arg.d[i].buckets        = src.d[i].buckets;
 608                 arg.d[i].sectors        = src.d[i].sectors;
 609                 arg.d[i].fragmented     = src.d[i].fragmented;
 610         }
 611
 612         bch2_dev_put(ca);
 613
 614         return copy_to_user_errcode(user_arg, &arg, sizeof(arg));
 615 }
 616
 617 static long bch2_ioctl_dev_usage_v2(struct bch_fs *c,
 618                                  struct bch_ioctl_dev_usage_v2 __user *user_arg)
 619 {
 620         struct bch_ioctl_dev_usage_v2 arg;
 621         struct bch_dev_usage src;
 622         struct bch_dev *ca;
 623         int ret = 0;
 624
 625         if (!test_bit(BCH_FS_started, &c->flags))
 626                 return -EINVAL;
 627
 628         if (copy_from_user(&arg, user_arg, sizeof(arg)))
 629                 return -EFAULT;
 630
 631         if ((arg.flags & ~BCH_BY_INDEX) ||
 632             arg.pad[0] ||
 633             arg.pad[1] ||
 634             arg.pad[2])
 635                 return -EINVAL;
 636
 637         ca = bch2_device_lookup(c, arg.dev, arg.flags);
 638         if (IS_ERR(ca))
 639                 return PTR_ERR(ca);
 640
 641         src = bch2_dev_usage_read(ca);
 642
 643         arg.state               = ca->mi.state;
 644         arg.bucket_size         = ca->mi.bucket_size;
 645         arg.nr_data_types       = min(arg.nr_data_types, BCH_DATA_NR);
 646         arg.nr_buckets          = ca->mi.nbuckets - ca->mi.first_bucket;
 647
 648         ret = copy_to_user_errcode(user_arg, &arg, sizeof(arg));
 649         if (ret)
 650                 goto err;
 651
 652         for (unsigned i = 0; i < arg.nr_data_types; i++) {
 653                 struct bch_ioctl_dev_usage_type t = {
 654                         .buckets        = src.d[i].buckets,
 655                         .sectors        = src.d[i].sectors,
 656                         .fragmented     = src.d[i].fragmented,
 657                 };
 658
 659                 ret = copy_to_user_errcode(&user_arg->d[i], &t, sizeof(t));
 660                 if (ret)
 661                         goto err;
 662         }
 663 err:
 664         bch2_dev_put(ca);
 665         return ret;
 666 }
 667
 668 static long bch2_ioctl_read_super(struct bch_fs *c,
 669                                   struct bch_ioctl_read_super arg)
 670 {
 671         struct bch_dev *ca = NULL;
 672         struct bch_sb *sb;
 673         int ret = 0;
 674
 675         if (!capable(CAP_SYS_ADMIN))
 676                 return -EPERM;
 677
 678         if ((arg.flags & ~(BCH_BY_INDEX|BCH_READ_DEV)) ||
 679             arg.pad)
 680                 return -EINVAL;
 681
 682         mutex_lock(&c->sb_lock);
 683
 684         if (arg.flags & BCH_READ_DEV) {
 685                 ca = bch2_device_lookup(c, arg.dev, arg.flags);
 686                 ret = PTR_ERR_OR_ZERO(ca);
 687                 if (ret)
 688                         goto err_unlock;
 689
 690                 sb = ca->disk_sb.sb;
 691         } else {
 692                 sb = c->disk_sb.sb;
 693         }
 694
 695         if (vstruct_bytes(sb) > arg.size) {
 696                 ret = -ERANGE;
 697                 goto err;
 698         }
 699
 700         ret = copy_to_user_errcode((void __user *)(unsigned long)arg.sb, sb,
 701                                    vstruct_bytes(sb));
 702 err:
 703         bch2_dev_put(ca);
 704 err_unlock:
 705         mutex_unlock(&c->sb_lock);
 706         return ret;
 707 }
 708
 709 static long bch2_ioctl_disk_get_idx(struct bch_fs *c,
 710                                     struct bch_ioctl_disk_get_idx arg)
 711 {
 712         dev_t dev = huge_decode_dev(arg.dev);
 713
 714         if (!capable(CAP_SYS_ADMIN))
 715                 return -EPERM;
 716
 717         if (!dev)
 718                 return -EINVAL;
 719
 720         for_each_online_member(c, ca)
 721                 if (ca->dev == dev) {
 722                         percpu_ref_put(&ca->io_ref);
 723                         return ca->dev_idx;
 724                 }
 725
 726         return -BCH_ERR_ENOENT_dev_idx_not_found;
 727 }
 728
 729 static long bch2_ioctl_disk_resize(struct bch_fs *c,
 730                                    struct bch_ioctl_disk_resize arg)
 731 {
 732         struct bch_dev *ca;
 733         int ret;
 734
 735         if (!capable(CAP_SYS_ADMIN))
 736                 return -EPERM;
 737
 738         if ((arg.flags & ~BCH_BY_INDEX) ||
 739             arg.pad)
 740                 return -EINVAL;
 741
 742         ca = bch2_device_lookup(c, arg.dev, arg.flags);
 743         if (IS_ERR(ca))
 744                 return PTR_ERR(ca);
 745
 746         ret = bch2_dev_resize(c, ca, arg.nbuckets);
 747
 748         bch2_dev_put(ca);
 749         return ret;
 750 }
 751
 752 static long bch2_ioctl_disk_resize_journal(struct bch_fs *c,
 753                                    struct bch_ioctl_disk_resize_journal arg)
 754 {
 755         struct bch_dev *ca;
 756         int ret;
 757
 758         if (!capable(CAP_SYS_ADMIN))
 759                 return -EPERM;
 760
 761         if ((arg.flags & ~BCH_BY_INDEX) ||
 762             arg.pad)
 763                 return -EINVAL;
 764
 765         if (arg.nbuckets > U32_MAX)
 766                 return -EINVAL;
 767
 768         ca = bch2_device_lookup(c, arg.dev, arg.flags);
 769         if (IS_ERR(ca))
 770                 return PTR_ERR(ca);
 771
 772         ret = bch2_set_nr_journal_buckets(c, ca, arg.nbuckets);
 773
 774         bch2_dev_put(ca);
 775         return ret;
 776 }
 777
 778 static int bch2_fsck_online_thread_fn(struct thread_with_stdio *stdio)
 779 {
 780         struct fsck_thread *thr = container_of(stdio, struct fsck_thread, thr);
 781         struct bch_fs *c = thr->c;
 782
 783         c->stdio_filter = current;
 784         c->stdio = &thr->thr.stdio;
 785
 786         /*
 787          * XXX: can we figure out a way to do this without mucking with c->opts?
 788          */
 789         unsigned old_fix_errors = c->opts.fix_errors;
 790         if (opt_defined(thr->opts, fix_errors))
 791                 c->opts.fix_errors = thr->opts.fix_errors;
 792         else
 793                 c->opts.fix_errors = FSCK_FIX_ask;
 794
 795         c->opts.fsck = true;
 796         set_bit(BCH_FS_fsck_running, &c->flags);
 797
 798         c->curr_recovery_pass = BCH_RECOVERY_PASS_check_alloc_info;
 799         int ret = bch2_run_online_recovery_passes(c);
 800
 801         clear_bit(BCH_FS_fsck_running, &c->flags);
 802         bch_err_fn(c, ret);
 803
 804         c->stdio = NULL;
 805         c->stdio_filter = NULL;
 806         c->opts.fix_errors = old_fix_errors;
 807
 808         up(&c->online_fsck_mutex);
 809         bch2_ro_ref_put(c);
 810         return ret;
 811 }
 812
 813 static const struct thread_with_stdio_ops bch2_online_fsck_ops = {
 814         .exit           = bch2_fsck_thread_exit,
 815         .fn             = bch2_fsck_online_thread_fn,
 816 };
 817
 818 static long bch2_ioctl_fsck_online(struct bch_fs *c,
 819                                    struct bch_ioctl_fsck_online arg)
 820 {
 821         struct fsck_thread *thr = NULL;
 822         long ret = 0;
 823
 824         if (arg.flags)
 825                 return -EINVAL;
 826
 827         if (!capable(CAP_SYS_ADMIN))
 828                 return -EPERM;
 829
 830         if (!bch2_ro_ref_tryget(c))
 831                 return -EROFS;
 832
 833         if (down_trylock(&c->online_fsck_mutex)) {
 834                 bch2_ro_ref_put(c);
 835                 return -EAGAIN;
 836         }
 837
 838         thr = kzalloc(sizeof(*thr), GFP_KERNEL);
 839         if (!thr) {
 840                 ret = -ENOMEM;
 841                 goto err;
 842         }
 843
 844         thr->c = c;
 845         thr->opts = bch2_opts_empty();
 846
 847         if (arg.opts) {
 848                 char *optstr = strndup_user((char __user *)(unsigned long) arg.opts, 1 << 16);
 849
 850                 ret =   PTR_ERR_OR_ZERO(optstr) ?:
 851                         bch2_parse_mount_opts(c, &thr->opts, NULL, optstr);
 852                 if (!IS_ERR(optstr))
 853                         kfree(optstr);
 854
 855                 if (ret)
 856                         goto err;
 857         }
 858
 859         ret = bch2_run_thread_with_stdio(&thr->thr, &bch2_online_fsck_ops);
 860 err:
 861         if (ret < 0) {
 862                 bch_err_fn(c, ret);
 863                 if (thr)
 864                         bch2_fsck_thread_exit(&thr->thr);
 865                 up(&c->online_fsck_mutex);
 866                 bch2_ro_ref_put(c);
 867         }
 868         return ret;
 869 }
 870
 871 #define BCH_IOCTL(_name, _argtype)                                      \
 872 do {                                                                    \
 873         _argtype i;                                                     \
 874                                                                         \
 875         if (copy_from_user(&i, arg, sizeof(i)))                         \
 876                 return -EFAULT;                                         \
 877         ret = bch2_ioctl_##_name(c, i);                                 \
 878         goto out;                                                       \
 879 } while (0)
 880
 881 long bch2_fs_ioctl(struct bch_fs *c, unsigned cmd, void __user *arg)
 882 {
 883         long ret;
 884
 885         switch (cmd) {
 886         case BCH_IOCTL_QUERY_UUID:
 887                 return bch2_ioctl_query_uuid(c, arg);
 888         case BCH_IOCTL_FS_USAGE:
 889                 return bch2_ioctl_fs_usage(c, arg);
 890         case BCH_IOCTL_DEV_USAGE:
 891                 return bch2_ioctl_dev_usage(c, arg);
 892         case BCH_IOCTL_DEV_USAGE_V2:
 893                 return bch2_ioctl_dev_usage_v2(c, arg);
 894 #if 0
 895         case BCH_IOCTL_START:
 896                 BCH_IOCTL(start, struct bch_ioctl_start);
 897         case BCH_IOCTL_STOP:
 898                 return bch2_ioctl_stop(c);
 899 #endif
 900         case BCH_IOCTL_READ_SUPER:
 901                 BCH_IOCTL(read_super, struct bch_ioctl_read_super);
 902         case BCH_IOCTL_DISK_GET_IDX:
 903                 BCH_IOCTL(disk_get_idx, struct bch_ioctl_disk_get_idx);
 904         }
 905
 906         if (!test_bit(BCH_FS_started, &c->flags))
 907                 return -EINVAL;
 908
 909         switch (cmd) {
 910         case BCH_IOCTL_DISK_ADD:
 911                 BCH_IOCTL(disk_add, struct bch_ioctl_disk);
 912         case BCH_IOCTL_DISK_REMOVE:
 913                 BCH_IOCTL(disk_remove, struct bch_ioctl_disk);
 914         case BCH_IOCTL_DISK_ONLINE:
 915                 BCH_IOCTL(disk_online, struct bch_ioctl_disk);
 916         case BCH_IOCTL_DISK_OFFLINE:
 917                 BCH_IOCTL(disk_offline, struct bch_ioctl_disk);
 918         case BCH_IOCTL_DISK_SET_STATE:
 919                 BCH_IOCTL(disk_set_state, struct bch_ioctl_disk_set_state);
 920         case BCH_IOCTL_DATA:
 921                 BCH_IOCTL(data, struct bch_ioctl_data);
 922         case BCH_IOCTL_DISK_RESIZE:
 923                 BCH_IOCTL(disk_resize, struct bch_ioctl_disk_resize);
 924         case BCH_IOCTL_DISK_RESIZE_JOURNAL:
 925                 BCH_IOCTL(disk_resize_journal, struct bch_ioctl_disk_resize_journal);
 926         case BCH_IOCTL_FSCK_ONLINE:
 927                 BCH_IOCTL(fsck_online, struct bch_ioctl_fsck_online);
 928         case BCH_IOCTL_QUERY_ACCOUNTING:
 929                 return bch2_ioctl_query_accounting(c, arg);
 930         default:
 931                 return -ENOTTY;
 932         }
 933 out:
 934         if (ret < 0)
 935                 ret = bch2_err_class(ret);
 936         return ret;
 937 }
 938
 939 static DEFINE_IDR(bch_chardev_minor);
 940
 941 static long bch2_chardev_ioctl(struct file *filp, unsigned cmd, unsigned long v)
 942 {
 943         unsigned minor = iminor(file_inode(filp));
 944         struct bch_fs *c = minor < U8_MAX ? idr_find(&bch_chardev_minor, minor) : NULL;
 945         void __user *arg = (void __user *) v;
 946
 947         return c
 948                 ? bch2_fs_ioctl(c, cmd, arg)
 949                 : bch2_global_ioctl(cmd, arg);
 950 }
 951
 952 static const struct file_operations bch_chardev_fops = {
 953         .owner          = THIS_MODULE,
 954         .unlocked_ioctl = bch2_chardev_ioctl,
 955         .open           = nonseekable_open,
 956 };
 957
 958 static int bch_chardev_major;
 959 static const struct class bch_chardev_class = {
 960         .name = "bcachefs",
 961 };
 962 static struct device *bch_chardev;
 963
 964 void bch2_fs_chardev_exit(struct bch_fs *c)
 965 {
 966         if (!IS_ERR_OR_NULL(c->chardev))
 967                 device_unregister(c->chardev);
 968         if (c->minor >= 0)
 969                 idr_remove(&bch_chardev_minor, c->minor);
 970 }
 971
 972 int bch2_fs_chardev_init(struct bch_fs *c)
 973 {
 974         c->minor = idr_alloc(&bch_chardev_minor, c, 0, 0, GFP_KERNEL);
 975         if (c->minor < 0)
 976                 return c->minor;
 977
 978         c->chardev = device_create(&bch_chardev_class, NULL,
 979                                    MKDEV(bch_chardev_major, c->minor), c,
 980                                    "bcachefs%u-ctl", c->minor);
 981         if (IS_ERR(c->chardev))
 982                 return PTR_ERR(c->chardev);
 983
 984         return 0;
 985 }
 986
 987 void bch2_chardev_exit(void)
 988 {
 989         device_destroy(&bch_chardev_class, MKDEV(bch_chardev_major, U8_MAX));
 990         class_unregister(&bch_chardev_class);
 991         if (bch_chardev_major > 0)
 992                 unregister_chrdev(bch_chardev_major, "bcachefs");
 993 }
 994
 995 int __init bch2_chardev_init(void)
 996 {
 997         int ret;
 998
 999         bch_chardev_major = register_chrdev(0, "bcachefs-ctl", &bch_chardev_fops);
1000         if (bch_chardev_major < 0)
1001                 return bch_chardev_major;
1002
1003         ret = class_register(&bch_chardev_class);
1004         if (ret)
1005                 goto major_out;
1006
1007         bch_chardev = device_create(&bch_chardev_class, NULL,
1008                                     MKDEV(bch_chardev_major, U8_MAX),
1009                                     NULL, "bcachefs-ctl");
1010         if (IS_ERR(bch_chardev)) {
1011                 ret = PTR_ERR(bch_chardev);
1012                 goto class_out;
1013         }
1014
1015         return 0;
1016
1017 class_out:
1018         class_unregister(&bch_chardev_class);
1019 major_out:
1020         unregister_chrdev(bch_chardev_major, "bcachefs-ctl");
1021         return ret;
1022 }
1023
1024 #endif /* NO_BCACHEFS_CHARDEV */