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