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 "journal.h"
   9 #include "move.h"
  10 #include "recovery_passes.h"
  11 #include "replicas.h"
  12 #include "super.h"
  13 #include "super-io.h"
  14 #include "thread_with_file.h"
  15
  16 #include <linux/cdev.h>
  17 #include <linux/device.h>
  18 #include <linux/fs.h>
  19 #include <linux/ioctl.h>
  20 #include <linux/major.h>
  21 #include <linux/sched/task.h>
  22 #include <linux/slab.h>
  23 #include <linux/uaccess.h>
  24
  25 /* returns with ref on ca->ref */
  26 static struct bch_dev *bch2_device_lookup(struct bch_fs *c, u64 dev,
  27                                           unsigned flags)
  28 {
  29         struct bch_dev *ca;
  30
  31         if (flags & BCH_BY_INDEX) {
  32                 if (dev >= c->sb.nr_devices)
  33                         return ERR_PTR(-EINVAL);
  34
  35                 rcu_read_lock();
  36                 ca = rcu_dereference(c->devs[dev]);
  37                 if (ca)
  38                         percpu_ref_get(&ca->ref);
  39                 rcu_read_unlock();
  40
  41                 if (!ca)
  42                         return ERR_PTR(-EINVAL);
  43         } else {
  44                 char *path;
  45
  46                 path = strndup_user((const char __user *)
  47                                     (unsigned long) dev, PATH_MAX);
  48                 if (IS_ERR(path))
  49                         return ERR_CAST(path);
  50
  51                 ca = bch2_dev_lookup(c, path);
  52                 kfree(path);
  53         }
  54
  55         return ca;
  56 }
  57
  58 #if 0
  59 static long bch2_ioctl_assemble(struct bch_ioctl_assemble __user *user_arg)
  60 {
  61         struct bch_ioctl_assemble arg;
  62         struct bch_fs *c;
  63         u64 *user_devs = NULL;
  64         char **devs = NULL;
  65         unsigned i;
  66         int ret = -EFAULT;
  67
  68         if (copy_from_user(&arg, user_arg, sizeof(arg)))
  69                 return -EFAULT;
  70
  71         if (arg.flags || arg.pad)
  72                 return -EINVAL;
  73
  74         user_devs = kmalloc_array(arg.nr_devs, sizeof(u64), GFP_KERNEL);
  75         if (!user_devs)
  76                 return -ENOMEM;
  77
  78         devs = kcalloc(arg.nr_devs, sizeof(char *), GFP_KERNEL);
  79
  80         if (copy_from_user(user_devs, user_arg->devs,
  81                            sizeof(u64) * arg.nr_devs))
  82                 goto err;
  83
  84         for (i = 0; i < arg.nr_devs; i++) {
  85                 devs[i] = strndup_user((const char __user *)(unsigned long)
  86                                        user_devs[i],
  87                                        PATH_MAX);
  88                 ret= PTR_ERR_OR_ZERO(devs[i]);
  89                 if (ret)
  90                         goto err;
  91         }
  92
  93         c = bch2_fs_open(devs, arg.nr_devs, bch2_opts_empty());
  94         ret = PTR_ERR_OR_ZERO(c);
  95         if (!ret)
  96                 closure_put(&c->cl);
  97 err:
  98         if (devs)
  99                 for (i = 0; i < arg.nr_devs; i++)
 100                         kfree(devs[i]);
 101         kfree(devs);
 102         return ret;
 103 }
 104
 105 static long bch2_ioctl_incremental(struct bch_ioctl_incremental __user *user_arg)
 106 {
 107         struct bch_ioctl_incremental arg;
 108         const char *err;
 109         char *path;
 110
 111         if (copy_from_user(&arg, user_arg, sizeof(arg)))
 112                 return -EFAULT;
 113
 114         if (arg.flags || arg.pad)
 115                 return -EINVAL;
 116
 117         path = strndup_user((const char __user *)(unsigned long) arg.dev, PATH_MAX);
 118         ret = PTR_ERR_OR_ZERO(path);
 119         if (ret)
 120                 return ret;
 121
 122         err = bch2_fs_open_incremental(path);
 123         kfree(path);
 124
 125         if (err) {
 126                 pr_err("Could not register bcachefs devices: %s", err);
 127                 return -EINVAL;
 128         }
 129
 130         return 0;
 131 }
 132 #endif
 133
 134 struct fsck_thread {
 135         struct thread_with_stdio thr;
 136         struct bch_fs           *c;
 137         struct bch_opts         opts;
 138 };
 139
 140 static void bch2_fsck_thread_exit(struct thread_with_stdio *_thr)
 141 {
 142         struct fsck_thread *thr = container_of(_thr, struct fsck_thread, thr);
 143         kfree(thr);
 144 }
 145
 146 static int bch2_fsck_offline_thread_fn(struct thread_with_stdio *stdio)
 147 {
 148         struct fsck_thread *thr = container_of(stdio, struct fsck_thread, thr);
 149         struct bch_fs *c = thr->c;
 150
 151         int ret = PTR_ERR_OR_ZERO(c);
 152         if (ret)
 153                 return ret;
 154
 155         ret = bch2_fs_start(thr->c);
 156         if (ret)
 157                 goto err;
 158
 159         if (test_bit(BCH_FS_errors_fixed, &c->flags)) {
 160                 bch2_stdio_redirect_printf(&stdio->stdio, false, "%s: errors fixed\n", c->name);
 161                 ret |= 1;
 162         }
 163         if (test_bit(BCH_FS_error, &c->flags)) {
 164                 bch2_stdio_redirect_printf(&stdio->stdio, false, "%s: still has errors\n", c->name);
 165                 ret |= 4;
 166         }
 167 err:
 168         bch2_fs_stop(c);
 169         return ret;
 170 }
 171
 172 static const struct thread_with_stdio_ops bch2_offline_fsck_ops = {
 173         .exit           = bch2_fsck_thread_exit,
 174         .fn             = bch2_fsck_offline_thread_fn,
 175 };
 176
 177 static long bch2_ioctl_fsck_offline(struct bch_ioctl_fsck_offline __user *user_arg)
 178 {
 179         struct bch_ioctl_fsck_offline arg;
 180         struct fsck_thread *thr = NULL;
 181         darray_str(devs) = {};
 182         long ret = 0;
 183
 184         if (copy_from_user(&arg, user_arg, sizeof(arg)))
 185                 return -EFAULT;
 186
 187         if (arg.flags)
 188                 return -EINVAL;
 189
 190         if (!capable(CAP_SYS_ADMIN))
 191                 return -EPERM;
 192
 193         for (size_t i = 0; i < arg.nr_devs; i++) {
 194                 u64 dev_u64;
 195                 ret = copy_from_user_errcode(&dev_u64, &user_arg->devs[i], sizeof(u64));
 196                 if (ret)
 197                         goto err;
 198
 199                 char *dev_str = strndup_user((char __user *)(unsigned long) dev_u64, PATH_MAX);
 200                 ret = PTR_ERR_OR_ZERO(dev_str);
 201                 if (ret)
 202                         goto err;
 203
 204                 ret = darray_push(&devs, dev_str);
 205                 if (ret) {
 206                         kfree(dev_str);
 207                         goto err;
 208                 }
 209         }
 210
 211         thr = kzalloc(sizeof(*thr), GFP_KERNEL);
 212         if (!thr) {
 213                 ret = -ENOMEM;
 214                 goto err;
 215         }
 216
 217         thr->opts = bch2_opts_empty();
 218
 219         if (arg.opts) {
 220                 char *optstr = strndup_user((char __user *)(unsigned long) arg.opts, 1 << 16);
 221
 222                 ret =   PTR_ERR_OR_ZERO(optstr) ?:
 223                         bch2_parse_mount_opts(NULL, &thr->opts, optstr);
 224                 kfree(optstr);
 225
 226                 if (ret)
 227                         goto err;
 228         }
 229
 230         opt_set(thr->opts, stdio, (u64)(unsigned long)&thr->thr.stdio);
 231
 232         /* We need request_key() to be called before we punt to kthread: */
 233         opt_set(thr->opts, nostart, true);
 234
 235         bch2_thread_with_stdio_init(&thr->thr, &bch2_offline_fsck_ops);
 236
 237         thr->c = bch2_fs_open(devs.data, arg.nr_devs, thr->opts);
 238
 239         if (!IS_ERR(thr->c) &&
 240             thr->c->opts.errors == BCH_ON_ERROR_panic)
 241                 thr->c->opts.errors = BCH_ON_ERROR_ro;
 242
 243         ret = __bch2_run_thread_with_stdio(&thr->thr);
 244 out:
 245         darray_for_each(devs, i)
 246                 kfree(*i);
 247         darray_exit(&devs);
 248         return ret;
 249 err:
 250         if (thr)
 251                 bch2_fsck_thread_exit(&thr->thr);
 252         pr_err("ret %s", bch2_err_str(ret));
 253         goto out;
 254 }
 255
 256 static long bch2_global_ioctl(unsigned cmd, void __user *arg)
 257 {
 258         long ret;
 259
 260         switch (cmd) {
 261 #if 0
 262         case BCH_IOCTL_ASSEMBLE:
 263                 return bch2_ioctl_assemble(arg);
 264         case BCH_IOCTL_INCREMENTAL:
 265                 return bch2_ioctl_incremental(arg);
 266 #endif
 267         case BCH_IOCTL_FSCK_OFFLINE: {
 268                 ret = bch2_ioctl_fsck_offline(arg);
 269                 break;
 270         }
 271         default:
 272                 ret = -ENOTTY;
 273                 break;
 274         }
 275
 276         if (ret < 0)
 277                 ret = bch2_err_class(ret);
 278         return ret;
 279 }
 280
 281 static long bch2_ioctl_query_uuid(struct bch_fs *c,
 282                         struct bch_ioctl_query_uuid __user *user_arg)
 283 {
 284         return copy_to_user_errcode(&user_arg->uuid, &c->sb.user_uuid,
 285                                     sizeof(c->sb.user_uuid));
 286 }
 287
 288 #if 0
 289 static long bch2_ioctl_start(struct bch_fs *c, struct bch_ioctl_start arg)
 290 {
 291         if (!capable(CAP_SYS_ADMIN))
 292                 return -EPERM;
 293
 294         if (arg.flags || arg.pad)
 295                 return -EINVAL;
 296
 297         return bch2_fs_start(c);
 298 }
 299
 300 static long bch2_ioctl_stop(struct bch_fs *c)
 301 {
 302         if (!capable(CAP_SYS_ADMIN))
 303                 return -EPERM;
 304
 305         bch2_fs_stop(c);
 306         return 0;
 307 }
 308 #endif
 309
 310 static long bch2_ioctl_disk_add(struct bch_fs *c, struct bch_ioctl_disk arg)
 311 {
 312         char *path;
 313         int ret;
 314
 315         if (!capable(CAP_SYS_ADMIN))
 316                 return -EPERM;
 317
 318         if (arg.flags || arg.pad)
 319                 return -EINVAL;
 320
 321         path = strndup_user((const char __user *)(unsigned long) arg.dev, PATH_MAX);
 322         ret = PTR_ERR_OR_ZERO(path);
 323         if (ret)
 324                 return ret;
 325
 326         ret = bch2_dev_add(c, path);
 327         kfree(path);
 328
 329         return ret;
 330 }
 331
 332 static long bch2_ioctl_disk_remove(struct bch_fs *c, struct bch_ioctl_disk arg)
 333 {
 334         struct bch_dev *ca;
 335
 336         if (!capable(CAP_SYS_ADMIN))
 337                 return -EPERM;
 338
 339         if ((arg.flags & ~(BCH_FORCE_IF_DATA_LOST|
 340                            BCH_FORCE_IF_METADATA_LOST|
 341                            BCH_FORCE_IF_DEGRADED|
 342                            BCH_BY_INDEX)) ||
 343             arg.pad)
 344                 return -EINVAL;
 345
 346         ca = bch2_device_lookup(c, arg.dev, arg.flags);
 347         if (IS_ERR(ca))
 348                 return PTR_ERR(ca);
 349
 350         return bch2_dev_remove(c, ca, arg.flags);
 351 }
 352
 353 static long bch2_ioctl_disk_online(struct bch_fs *c, struct bch_ioctl_disk arg)
 354 {
 355         char *path;
 356         int ret;
 357
 358         if (!capable(CAP_SYS_ADMIN))
 359                 return -EPERM;
 360
 361         if (arg.flags || arg.pad)
 362                 return -EINVAL;
 363
 364         path = strndup_user((const char __user *)(unsigned long) arg.dev, PATH_MAX);
 365         ret = PTR_ERR_OR_ZERO(path);
 366         if (ret)
 367                 return ret;
 368
 369         ret = bch2_dev_online(c, path);
 370         kfree(path);
 371         return ret;
 372 }
 373
 374 static long bch2_ioctl_disk_offline(struct bch_fs *c, struct bch_ioctl_disk arg)
 375 {
 376         struct bch_dev *ca;
 377         int ret;
 378
 379         if (!capable(CAP_SYS_ADMIN))
 380                 return -EPERM;
 381
 382         if ((arg.flags & ~(BCH_FORCE_IF_DATA_LOST|
 383                            BCH_FORCE_IF_METADATA_LOST|
 384                            BCH_FORCE_IF_DEGRADED|
 385                            BCH_BY_INDEX)) ||
 386             arg.pad)
 387                 return -EINVAL;
 388
 389         ca = bch2_device_lookup(c, arg.dev, arg.flags);
 390         if (IS_ERR(ca))
 391                 return PTR_ERR(ca);
 392
 393         ret = bch2_dev_offline(c, ca, arg.flags);
 394         percpu_ref_put(&ca->ref);
 395         return ret;
 396 }
 397
 398 static long bch2_ioctl_disk_set_state(struct bch_fs *c,
 399                         struct bch_ioctl_disk_set_state arg)
 400 {
 401         struct bch_dev *ca;
 402         int ret;
 403
 404         if (!capable(CAP_SYS_ADMIN))
 405                 return -EPERM;
 406
 407         if ((arg.flags & ~(BCH_FORCE_IF_DATA_LOST|
 408                            BCH_FORCE_IF_METADATA_LOST|
 409                            BCH_FORCE_IF_DEGRADED|
 410                            BCH_BY_INDEX)) ||
 411             arg.pad[0] || arg.pad[1] || arg.pad[2] ||
 412             arg.new_state >= BCH_MEMBER_STATE_NR)
 413                 return -EINVAL;
 414
 415         ca = bch2_device_lookup(c, arg.dev, arg.flags);
 416         if (IS_ERR(ca))
 417                 return PTR_ERR(ca);
 418
 419         ret = bch2_dev_set_state(c, ca, arg.new_state, arg.flags);
 420         if (ret)
 421                 bch_err(c, "Error setting device state: %s", bch2_err_str(ret));
 422
 423         percpu_ref_put(&ca->ref);
 424         return ret;
 425 }
 426
 427 struct bch_data_ctx {
 428         struct thread_with_file         thr;
 429
 430         struct bch_fs                   *c;
 431         struct bch_ioctl_data           arg;
 432         struct bch_move_stats           stats;
 433 };
 434
 435 static int bch2_data_thread(void *arg)
 436 {
 437         struct bch_data_ctx *ctx = container_of(arg, struct bch_data_ctx, thr);
 438
 439         ctx->thr.ret = bch2_data_job(ctx->c, &ctx->stats, ctx->arg);
 440         ctx->stats.data_type = U8_MAX;
 441         return 0;
 442 }
 443
 444 static int bch2_data_job_release(struct inode *inode, struct file *file)
 445 {
 446         struct bch_data_ctx *ctx = container_of(file->private_data, struct bch_data_ctx, thr);
 447
 448         bch2_thread_with_file_exit(&ctx->thr);
 449         kfree(ctx);
 450         return 0;
 451 }
 452
 453 static ssize_t bch2_data_job_read(struct file *file, char __user *buf,
 454                                   size_t len, loff_t *ppos)
 455 {
 456         struct bch_data_ctx *ctx = container_of(file->private_data, struct bch_data_ctx, thr);
 457         struct bch_fs *c = ctx->c;
 458         struct bch_ioctl_data_event e = {
 459                 .type                   = BCH_DATA_EVENT_PROGRESS,
 460                 .p.data_type            = ctx->stats.data_type,
 461                 .p.btree_id             = ctx->stats.pos.btree,
 462                 .p.pos                  = ctx->stats.pos.pos,
 463                 .p.sectors_done         = atomic64_read(&ctx->stats.sectors_seen),
 464                 .p.sectors_total        = bch2_fs_usage_read_short(c).used,
 465         };
 466
 467         if (len < sizeof(e))
 468                 return -EINVAL;
 469
 470         return copy_to_user_errcode(buf, &e, sizeof(e)) ?: sizeof(e);
 471 }
 472
 473 static const struct file_operations bcachefs_data_ops = {
 474         .release        = bch2_data_job_release,
 475         .read           = bch2_data_job_read,
 476         .llseek         = no_llseek,
 477 };
 478
 479 static long bch2_ioctl_data(struct bch_fs *c,
 480                             struct bch_ioctl_data arg)
 481 {
 482         struct bch_data_ctx *ctx;
 483         int ret;
 484
 485         if (!capable(CAP_SYS_ADMIN))
 486                 return -EPERM;
 487
 488         if (arg.op >= BCH_DATA_OP_NR || arg.flags)
 489                 return -EINVAL;
 490
 491         ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
 492         if (!ctx)
 493                 return -ENOMEM;
 494
 495         ctx->c = c;
 496         ctx->arg = arg;
 497
 498         ret = bch2_run_thread_with_file(&ctx->thr,
 499                         &bcachefs_data_ops,
 500                         bch2_data_thread);
 501         if (ret < 0)
 502                 kfree(ctx);
 503         return ret;
 504 }
 505
 506 static long bch2_ioctl_fs_usage(struct bch_fs *c,
 507                                 struct bch_ioctl_fs_usage __user *user_arg)
 508 {
 509         struct bch_ioctl_fs_usage *arg = NULL;
 510         struct bch_replicas_usage *dst_e, *dst_end;
 511         struct bch_fs_usage_online *src;
 512         u32 replica_entries_bytes;
 513         unsigned i;
 514         int ret = 0;
 515
 516         if (!test_bit(BCH_FS_started, &c->flags))
 517                 return -EINVAL;
 518
 519         if (get_user(replica_entries_bytes, &user_arg->replica_entries_bytes))
 520                 return -EFAULT;
 521
 522         arg = kzalloc(size_add(sizeof(*arg), replica_entries_bytes), GFP_KERNEL);
 523         if (!arg)
 524                 return -ENOMEM;
 525
 526         src = bch2_fs_usage_read(c);
 527         if (!src) {
 528                 ret = -ENOMEM;
 529                 goto err;
 530         }
 531
 532         arg->capacity           = c->capacity;
 533         arg->used               = bch2_fs_sectors_used(c, src);
 534         arg->online_reserved    = src->online_reserved;
 535
 536         for (i = 0; i < BCH_REPLICAS_MAX; i++)
 537                 arg->persistent_reserved[i] = src->u.persistent_reserved[i];
 538
 539         dst_e   = arg->replicas;
 540         dst_end = (void *) arg->replicas + replica_entries_bytes;
 541
 542         for (i = 0; i < c->replicas.nr; i++) {
 543                 struct bch_replicas_entry_v1 *src_e =
 544                         cpu_replicas_entry(&c->replicas, i);
 545
 546                 /* check that we have enough space for one replicas entry */
 547                 if (dst_e + 1 > dst_end) {
 548                         ret = -ERANGE;
 549                         break;
 550                 }
 551
 552                 dst_e->sectors          = src->u.replicas[i];
 553                 dst_e->r                = *src_e;
 554
 555                 /* recheck after setting nr_devs: */
 556                 if (replicas_usage_next(dst_e) > dst_end) {
 557                         ret = -ERANGE;
 558                         break;
 559                 }
 560
 561                 memcpy(dst_e->r.devs, src_e->devs, src_e->nr_devs);
 562
 563                 dst_e = replicas_usage_next(dst_e);
 564         }
 565
 566         arg->replica_entries_bytes = (void *) dst_e - (void *) arg->replicas;
 567
 568         percpu_up_read(&c->mark_lock);
 569         kfree(src);
 570
 571         if (ret)
 572                 goto err;
 573
 574         ret = copy_to_user_errcode(user_arg, arg,
 575                         sizeof(*arg) + arg->replica_entries_bytes);
 576 err:
 577         kfree(arg);
 578         return ret;
 579 }
 580
 581 /* obsolete, didn't allow for new data types: */
 582 static long bch2_ioctl_dev_usage(struct bch_fs *c,
 583                                  struct bch_ioctl_dev_usage __user *user_arg)
 584 {
 585         struct bch_ioctl_dev_usage arg;
 586         struct bch_dev_usage src;
 587         struct bch_dev *ca;
 588         unsigned i;
 589
 590         if (!test_bit(BCH_FS_started, &c->flags))
 591                 return -EINVAL;
 592
 593         if (copy_from_user(&arg, user_arg, sizeof(arg)))
 594                 return -EFAULT;
 595
 596         if ((arg.flags & ~BCH_BY_INDEX) ||
 597             arg.pad[0] ||
 598             arg.pad[1] ||
 599             arg.pad[2])
 600                 return -EINVAL;
 601
 602         ca = bch2_device_lookup(c, arg.dev, arg.flags);
 603         if (IS_ERR(ca))
 604                 return PTR_ERR(ca);
 605
 606         src = bch2_dev_usage_read(ca);
 607
 608         arg.state               = ca->mi.state;
 609         arg.bucket_size         = ca->mi.bucket_size;
 610         arg.nr_buckets          = ca->mi.nbuckets - ca->mi.first_bucket;
 611
 612         for (i = 0; i < BCH_DATA_NR; i++) {
 613                 arg.d[i].buckets        = src.d[i].buckets;
 614                 arg.d[i].sectors        = src.d[i].sectors;
 615                 arg.d[i].fragmented     = src.d[i].fragmented;
 616         }
 617
 618         percpu_ref_put(&ca->ref);
 619
 620         return copy_to_user_errcode(user_arg, &arg, sizeof(arg));
 621 }
 622
 623 static long bch2_ioctl_dev_usage_v2(struct bch_fs *c,
 624                                  struct bch_ioctl_dev_usage_v2 __user *user_arg)
 625 {
 626         struct bch_ioctl_dev_usage_v2 arg;
 627         struct bch_dev_usage src;
 628         struct bch_dev *ca;
 629         int ret = 0;
 630
 631         if (!test_bit(BCH_FS_started, &c->flags))
 632                 return -EINVAL;
 633
 634         if (copy_from_user(&arg, user_arg, sizeof(arg)))
 635                 return -EFAULT;
 636
 637         if ((arg.flags & ~BCH_BY_INDEX) ||
 638             arg.pad[0] ||
 639             arg.pad[1] ||
 640             arg.pad[2])
 641                 return -EINVAL;
 642
 643         ca = bch2_device_lookup(c, arg.dev, arg.flags);
 644         if (IS_ERR(ca))
 645                 return PTR_ERR(ca);
 646
 647         src = bch2_dev_usage_read(ca);
 648
 649         arg.state               = ca->mi.state;
 650         arg.bucket_size         = ca->mi.bucket_size;
 651         arg.nr_data_types       = min(arg.nr_data_types, BCH_DATA_NR);
 652         arg.nr_buckets          = ca->mi.nbuckets - ca->mi.first_bucket;
 653
 654         ret = copy_to_user_errcode(user_arg, &arg, sizeof(arg));
 655         if (ret)
 656                 goto err;
 657
 658         for (unsigned i = 0; i < arg.nr_data_types; i++) {
 659                 struct bch_ioctl_dev_usage_type t = {
 660                         .buckets        = src.d[i].buckets,
 661                         .sectors        = src.d[i].sectors,
 662                         .fragmented     = src.d[i].fragmented,
 663                 };
 664
 665                 ret = copy_to_user_errcode(&user_arg->d[i], &t, sizeof(t));
 666                 if (ret)
 667                         goto err;
 668         }
 669 err:
 670         percpu_ref_put(&ca->ref);
 671         return ret;
 672 }
 673
 674 static long bch2_ioctl_read_super(struct bch_fs *c,
 675                                   struct bch_ioctl_read_super arg)
 676 {
 677         struct bch_dev *ca = NULL;
 678         struct bch_sb *sb;
 679         int ret = 0;
 680
 681         if (!capable(CAP_SYS_ADMIN))
 682                 return -EPERM;
 683
 684         if ((arg.flags & ~(BCH_BY_INDEX|BCH_READ_DEV)) ||
 685             arg.pad)
 686                 return -EINVAL;
 687
 688         mutex_lock(&c->sb_lock);
 689
 690         if (arg.flags & BCH_READ_DEV) {
 691                 ca = bch2_device_lookup(c, arg.dev, arg.flags);
 692
 693                 if (IS_ERR(ca)) {
 694                         ret = PTR_ERR(ca);
 695                         goto err;
 696                 }
 697
 698                 sb = ca->disk_sb.sb;
 699         } else {
 700                 sb = c->disk_sb.sb;
 701         }
 702
 703         if (vstruct_bytes(sb) > arg.size) {
 704                 ret = -ERANGE;
 705                 goto err;
 706         }
 707
 708         ret = copy_to_user_errcode((void __user *)(unsigned long)arg.sb, sb,
 709                                    vstruct_bytes(sb));
 710 err:
 711         if (!IS_ERR_OR_NULL(ca))
 712                 percpu_ref_put(&ca->ref);
 713         mutex_unlock(&c->sb_lock);
 714         return ret;
 715 }
 716
 717 static long bch2_ioctl_disk_get_idx(struct bch_fs *c,
 718                                     struct bch_ioctl_disk_get_idx arg)
 719 {
 720         dev_t dev = huge_decode_dev(arg.dev);
 721
 722         if (!capable(CAP_SYS_ADMIN))
 723                 return -EPERM;
 724
 725         if (!dev)
 726                 return -EINVAL;
 727
 728         for_each_online_member(c, ca)
 729                 if (ca->dev == dev) {
 730                         percpu_ref_put(&ca->io_ref);
 731                         return ca->dev_idx;
 732                 }
 733
 734         return -BCH_ERR_ENOENT_dev_idx_not_found;
 735 }
 736
 737 static long bch2_ioctl_disk_resize(struct bch_fs *c,
 738                                    struct bch_ioctl_disk_resize arg)
 739 {
 740         struct bch_dev *ca;
 741         int ret;
 742
 743         if (!capable(CAP_SYS_ADMIN))
 744                 return -EPERM;
 745
 746         if ((arg.flags & ~BCH_BY_INDEX) ||
 747             arg.pad)
 748                 return -EINVAL;
 749
 750         ca = bch2_device_lookup(c, arg.dev, arg.flags);
 751         if (IS_ERR(ca))
 752                 return PTR_ERR(ca);
 753
 754         ret = bch2_dev_resize(c, ca, arg.nbuckets);
 755
 756         percpu_ref_put(&ca->ref);
 757         return ret;
 758 }
 759
 760 static long bch2_ioctl_disk_resize_journal(struct bch_fs *c,
 761                                    struct bch_ioctl_disk_resize_journal arg)
 762 {
 763         struct bch_dev *ca;
 764         int ret;
 765
 766         if (!capable(CAP_SYS_ADMIN))
 767                 return -EPERM;
 768
 769         if ((arg.flags & ~BCH_BY_INDEX) ||
 770             arg.pad)
 771                 return -EINVAL;
 772
 773         if (arg.nbuckets > U32_MAX)
 774                 return -EINVAL;
 775
 776         ca = bch2_device_lookup(c, arg.dev, arg.flags);
 777         if (IS_ERR(ca))
 778                 return PTR_ERR(ca);
 779
 780         ret = bch2_set_nr_journal_buckets(c, ca, arg.nbuckets);
 781
 782         percpu_ref_put(&ca->ref);
 783         return ret;
 784 }
 785
 786 static int bch2_fsck_online_thread_fn(struct thread_with_stdio *stdio)
 787 {
 788         struct fsck_thread *thr = container_of(stdio, struct fsck_thread, thr);
 789         struct bch_fs *c = thr->c;
 790
 791         c->stdio_filter = current;
 792         c->stdio = &thr->thr.stdio;
 793
 794         /*
 795          * XXX: can we figure out a way to do this without mucking with c->opts?
 796          */
 797         unsigned old_fix_errors = c->opts.fix_errors;
 798         if (opt_defined(thr->opts, fix_errors))
 799                 c->opts.fix_errors = thr->opts.fix_errors;
 800         else
 801                 c->opts.fix_errors = FSCK_FIX_ask;
 802
 803         c->opts.fsck = true;
 804         set_bit(BCH_FS_fsck_running, &c->flags);
 805
 806         c->curr_recovery_pass = BCH_RECOVERY_PASS_check_alloc_info;
 807         int ret = bch2_run_online_recovery_passes(c);
 808
 809         clear_bit(BCH_FS_fsck_running, &c->flags);
 810         bch_err_fn(c, ret);
 811
 812         c->stdio = NULL;
 813         c->stdio_filter = NULL;
 814         c->opts.fix_errors = old_fix_errors;
 815
 816         up(&c->online_fsck_mutex);
 817         bch2_ro_ref_put(c);
 818         return ret;
 819 }
 820
 821 static const struct thread_with_stdio_ops bch2_online_fsck_ops = {
 822         .exit           = bch2_fsck_thread_exit,
 823         .fn             = bch2_fsck_online_thread_fn,
 824 };
 825
 826 static long bch2_ioctl_fsck_online(struct bch_fs *c,
 827                                    struct bch_ioctl_fsck_online arg)
 828 {
 829         struct fsck_thread *thr = NULL;
 830         long ret = 0;
 831
 832         if (arg.flags)
 833                 return -EINVAL;
 834
 835         if (!capable(CAP_SYS_ADMIN))
 836                 return -EPERM;
 837
 838         if (!bch2_ro_ref_tryget(c))
 839                 return -EROFS;
 840
 841         if (down_trylock(&c->online_fsck_mutex)) {
 842                 bch2_ro_ref_put(c);
 843                 return -EAGAIN;
 844         }
 845
 846         thr = kzalloc(sizeof(*thr), GFP_KERNEL);
 847         if (!thr) {
 848                 ret = -ENOMEM;
 849                 goto err;
 850         }
 851
 852         thr->c = c;
 853         thr->opts = bch2_opts_empty();
 854
 855         if (arg.opts) {
 856                 char *optstr = strndup_user((char __user *)(unsigned long) arg.opts, 1 << 16);
 857
 858                 ret =   PTR_ERR_OR_ZERO(optstr) ?:
 859                         bch2_parse_mount_opts(c, &thr->opts, optstr);
 860                 kfree(optstr);
 861
 862                 if (ret)
 863                         goto err;
 864         }
 865
 866         ret = bch2_run_thread_with_stdio(&thr->thr, &bch2_online_fsck_ops);
 867 err:
 868         if (ret < 0) {
 869                 bch_err_fn(c, ret);
 870                 if (thr)
 871                         bch2_fsck_thread_exit(&thr->thr);
 872                 up(&c->online_fsck_mutex);
 873                 bch2_ro_ref_put(c);
 874         }
 875         return ret;
 876 }
 877
 878 #define BCH_IOCTL(_name, _argtype)                                      \
 879 do {                                                                    \
 880         _argtype i;                                                     \
 881                                                                         \
 882         if (copy_from_user(&i, arg, sizeof(i)))                         \
 883                 return -EFAULT;                                         \
 884         ret = bch2_ioctl_##_name(c, i);                                 \
 885         goto out;                                                       \
 886 } while (0)
 887
 888 long bch2_fs_ioctl(struct bch_fs *c, unsigned cmd, void __user *arg)
 889 {
 890         long ret;
 891
 892         switch (cmd) {
 893         case BCH_IOCTL_QUERY_UUID:
 894                 return bch2_ioctl_query_uuid(c, arg);
 895         case BCH_IOCTL_FS_USAGE:
 896                 return bch2_ioctl_fs_usage(c, arg);
 897         case BCH_IOCTL_DEV_USAGE:
 898                 return bch2_ioctl_dev_usage(c, arg);
 899         case BCH_IOCTL_DEV_USAGE_V2:
 900                 return bch2_ioctl_dev_usage_v2(c, arg);
 901 #if 0
 902         case BCH_IOCTL_START:
 903                 BCH_IOCTL(start, struct bch_ioctl_start);
 904         case BCH_IOCTL_STOP:
 905                 return bch2_ioctl_stop(c);
 906 #endif
 907         case BCH_IOCTL_READ_SUPER:
 908                 BCH_IOCTL(read_super, struct bch_ioctl_read_super);
 909         case BCH_IOCTL_DISK_GET_IDX:
 910                 BCH_IOCTL(disk_get_idx, struct bch_ioctl_disk_get_idx);
 911         }
 912
 913         if (!test_bit(BCH_FS_started, &c->flags))
 914                 return -EINVAL;
 915
 916         switch (cmd) {
 917         case BCH_IOCTL_DISK_ADD:
 918                 BCH_IOCTL(disk_add, struct bch_ioctl_disk);
 919         case BCH_IOCTL_DISK_REMOVE:
 920                 BCH_IOCTL(disk_remove, struct bch_ioctl_disk);
 921         case BCH_IOCTL_DISK_ONLINE:
 922                 BCH_IOCTL(disk_online, struct bch_ioctl_disk);
 923         case BCH_IOCTL_DISK_OFFLINE:
 924                 BCH_IOCTL(disk_offline, struct bch_ioctl_disk);
 925         case BCH_IOCTL_DISK_SET_STATE:
 926                 BCH_IOCTL(disk_set_state, struct bch_ioctl_disk_set_state);
 927         case BCH_IOCTL_DATA:
 928                 BCH_IOCTL(data, struct bch_ioctl_data);
 929         case BCH_IOCTL_DISK_RESIZE:
 930                 BCH_IOCTL(disk_resize, struct bch_ioctl_disk_resize);
 931         case BCH_IOCTL_DISK_RESIZE_JOURNAL:
 932                 BCH_IOCTL(disk_resize_journal, struct bch_ioctl_disk_resize_journal);
 933         case BCH_IOCTL_FSCK_ONLINE:
 934                 BCH_IOCTL(fsck_online, struct bch_ioctl_fsck_online);
 935         default:
 936                 return -ENOTTY;
 937         }
 938 out:
 939         if (ret < 0)
 940                 ret = bch2_err_class(ret);
 941         return ret;
 942 }
 943
 944 static DEFINE_IDR(bch_chardev_minor);
 945
 946 static long bch2_chardev_ioctl(struct file *filp, unsigned cmd, unsigned long v)
 947 {
 948         unsigned minor = iminor(file_inode(filp));
 949         struct bch_fs *c = minor < U8_MAX ? idr_find(&bch_chardev_minor, minor) : NULL;
 950         void __user *arg = (void __user *) v;
 951
 952         return c
 953                 ? bch2_fs_ioctl(c, cmd, arg)
 954                 : bch2_global_ioctl(cmd, arg);
 955 }
 956
 957 static const struct file_operations bch_chardev_fops = {
 958         .owner          = THIS_MODULE,
 959         .unlocked_ioctl = bch2_chardev_ioctl,
 960         .open           = nonseekable_open,
 961 };
 962
 963 static int bch_chardev_major;
 964 static struct class *bch_chardev_class;
 965 static struct device *bch_chardev;
 966
 967 void bch2_fs_chardev_exit(struct bch_fs *c)
 968 {
 969         if (!IS_ERR_OR_NULL(c->chardev))
 970                 device_unregister(c->chardev);
 971         if (c->minor >= 0)
 972                 idr_remove(&bch_chardev_minor, c->minor);
 973 }
 974
 975 int bch2_fs_chardev_init(struct bch_fs *c)
 976 {
 977         c->minor = idr_alloc(&bch_chardev_minor, c, 0, 0, GFP_KERNEL);
 978         if (c->minor < 0)
 979                 return c->minor;
 980
 981         c->chardev = device_create(bch_chardev_class, NULL,
 982                                    MKDEV(bch_chardev_major, c->minor), c,
 983                                    "bcachefs%u-ctl", c->minor);
 984         if (IS_ERR(c->chardev))
 985                 return PTR_ERR(c->chardev);
 986
 987         return 0;
 988 }
 989
 990 void bch2_chardev_exit(void)
 991 {
 992         if (!IS_ERR_OR_NULL(bch_chardev_class))
 993                 device_destroy(bch_chardev_class,
 994                                MKDEV(bch_chardev_major, U8_MAX));
 995         if (!IS_ERR_OR_NULL(bch_chardev_class))
 996                 class_destroy(bch_chardev_class);
 997         if (bch_chardev_major > 0)
 998                 unregister_chrdev(bch_chardev_major, "bcachefs");
 999 }
1000
1001 int __init bch2_chardev_init(void)
1002 {
1003         bch_chardev_major = register_chrdev(0, "bcachefs-ctl", &bch_chardev_fops);
1004         if (bch_chardev_major < 0)
1005                 return bch_chardev_major;
1006
1007         bch_chardev_class = class_create("bcachefs");
1008         if (IS_ERR(bch_chardev_class))
1009                 return PTR_ERR(bch_chardev_class);
1010
1011         bch_chardev = device_create(bch_chardev_class, NULL,
1012                                     MKDEV(bch_chardev_major, U8_MAX),
1013                                     NULL, "bcachefs-ctl");
1014         if (IS_ERR(bch_chardev))
1015                 return PTR_ERR(bch_chardev);
1016
1017         return 0;
1018 }
1019
1020 #endif /* NO_BCACHEFS_CHARDEV */