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