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/anon_inodes.h>
  16 #include <linux/cdev.h>
  17 #include <linux/device.h>
  18 #include <linux/file.h>
  19 #include <linux/fs.h>
  20 #include <linux/ioctl.h>
  21 #include <linux/kthread.h>
  22 #include <linux/major.h>
  23 #include <linux/poll.h>
  24 #include <linux/sched/task.h>
  25 #include <linux/slab.h>
  26 #include <linux/uaccess.h>
  27
  28 __must_check
  29 static int copy_to_user_errcode(void __user *to, const void *from, unsigned long n)
  30 {
  31         return copy_to_user(to, from, n) ? -EFAULT : 0;
  32 }
  33
  34 struct thread_with_file {
  35         struct task_struct      *task;
  36         int                     ret;
  37         bool                    done;
  38 };
  39
  40 static void thread_with_file_exit(struct thread_with_file *thr)
  41 {
  42         if (thr->task) {
  43                 kthread_stop(thr->task);
  44                 put_task_struct(thr->task);
  45         }
  46 }
  47
  48 __printf(4, 0)
  49 static int run_thread_with_file(struct thread_with_file *thr,
  50                                 const struct file_operations *fops,
  51                                 int (*fn)(void *), const char *fmt, ...)
  52 {
  53         va_list args;
  54         struct file *file = NULL;
  55         int ret, fd = -1;
  56         struct printbuf name = PRINTBUF;
  57         unsigned fd_flags = O_RDONLY|O_CLOEXEC|O_NONBLOCK;
  58
  59         va_start(args, fmt);
  60         prt_vprintf(&name, fmt, args);
  61         va_end(args);
  62
  63         thr->ret = 0;
  64         thr->task = kthread_create(fn, thr, name.buf);
  65         ret = PTR_ERR_OR_ZERO(thr->task);
  66         if (ret)
  67                 goto err;
  68
  69         ret = get_unused_fd_flags(fd_flags);
  70         if (ret < 0)
  71                 goto err_stop_task;
  72         fd = ret;
  73
  74         file = anon_inode_getfile(name.buf, fops, thr, fd_flags);
  75         ret = PTR_ERR_OR_ZERO(file);
  76         if (ret)
  77                 goto err_put_fd;
  78
  79         fd_install(fd, file);
  80         get_task_struct(thr->task);
  81         wake_up_process(thr->task);
  82         printbuf_exit(&name);
  83         return fd;
  84 err_put_fd:
  85         put_unused_fd(fd);
  86 err_stop_task:
  87         kthread_stop(thr->task);
  88 err:
  89         printbuf_exit(&name);
  90         return ret;
  91 }
  92
  93 /* returns with ref on ca->ref */
  94 static struct bch_dev *bch2_device_lookup(struct bch_fs *c, u64 dev,
  95                                           unsigned flags)
  96 {
  97         struct bch_dev *ca;
  98
  99         if (flags & BCH_BY_INDEX) {
 100                 if (dev >= c->sb.nr_devices)
 101                         return ERR_PTR(-EINVAL);
 102
 103                 rcu_read_lock();
 104                 ca = rcu_dereference(c->devs[dev]);
 105                 if (ca)
 106                         percpu_ref_get(&ca->ref);
 107                 rcu_read_unlock();
 108
 109                 if (!ca)
 110                         return ERR_PTR(-EINVAL);
 111         } else {
 112                 char *path;
 113
 114                 path = strndup_user((const char __user *)
 115                                     (unsigned long) dev, PATH_MAX);
 116                 if (IS_ERR(path))
 117                         return ERR_CAST(path);
 118
 119                 ca = bch2_dev_lookup(c, path);
 120                 kfree(path);
 121         }
 122
 123         return ca;
 124 }
 125
 126 #if 0
 127 static long bch2_ioctl_assemble(struct bch_ioctl_assemble __user *user_arg)
 128 {
 129         struct bch_ioctl_assemble arg;
 130         struct bch_fs *c;
 131         u64 *user_devs = NULL;
 132         char **devs = NULL;
 133         unsigned i;
 134         int ret = -EFAULT;
 135
 136         if (copy_from_user(&arg, user_arg, sizeof(arg)))
 137                 return -EFAULT;
 138
 139         if (arg.flags || arg.pad)
 140                 return -EINVAL;
 141
 142         user_devs = kmalloc_array(arg.nr_devs, sizeof(u64), GFP_KERNEL);
 143         if (!user_devs)
 144                 return -ENOMEM;
 145
 146         devs = kcalloc(arg.nr_devs, sizeof(char *), GFP_KERNEL);
 147
 148         if (copy_from_user(user_devs, user_arg->devs,
 149                            sizeof(u64) * arg.nr_devs))
 150                 goto err;
 151
 152         for (i = 0; i < arg.nr_devs; i++) {
 153                 devs[i] = strndup_user((const char __user *)(unsigned long)
 154                                        user_devs[i],
 155                                        PATH_MAX);
 156                 ret= PTR_ERR_OR_ZERO(devs[i]);
 157                 if (ret)
 158                         goto err;
 159         }
 160
 161         c = bch2_fs_open(devs, arg.nr_devs, bch2_opts_empty());
 162         ret = PTR_ERR_OR_ZERO(c);
 163         if (!ret)
 164                 closure_put(&c->cl);
 165 err:
 166         if (devs)
 167                 for (i = 0; i < arg.nr_devs; i++)
 168                         kfree(devs[i]);
 169         kfree(devs);
 170         return ret;
 171 }
 172
 173 static long bch2_ioctl_incremental(struct bch_ioctl_incremental __user *user_arg)
 174 {
 175         struct bch_ioctl_incremental arg;
 176         const char *err;
 177         char *path;
 178
 179         if (copy_from_user(&arg, user_arg, sizeof(arg)))
 180                 return -EFAULT;
 181
 182         if (arg.flags || arg.pad)
 183                 return -EINVAL;
 184
 185         path = strndup_user((const char __user *)(unsigned long) arg.dev, PATH_MAX);
 186         ret = PTR_ERR_OR_ZERO(path);
 187         if (ret)
 188                 return ret;
 189
 190         err = bch2_fs_open_incremental(path);
 191         kfree(path);
 192
 193         if (err) {
 194                 pr_err("Could not register bcachefs devices: %s", err);
 195                 return -EINVAL;
 196         }
 197
 198         return 0;
 199 }
 200 #endif
 201
 202 struct fsck_thread {
 203         struct thread_with_file thr;
 204         struct printbuf         buf;
 205         struct bch_fs           *c;
 206         char                    **devs;
 207         size_t                  nr_devs;
 208         struct bch_opts         opts;
 209
 210         struct log_output       output;
 211         DARRAY(char)            output2;
 212 };
 213
 214 static void bch2_fsck_thread_free(struct fsck_thread *thr)
 215 {
 216         thread_with_file_exit(&thr->thr);
 217         if (thr->devs)
 218                 for (size_t i = 0; i < thr->nr_devs; i++)
 219                         kfree(thr->devs[i]);
 220         darray_exit(&thr->output2);
 221         printbuf_exit(&thr->output.buf);
 222         kfree(thr->devs);
 223         kfree(thr);
 224 }
 225
 226 static int bch2_fsck_thread_release(struct inode *inode, struct file *file)
 227 {
 228         struct fsck_thread *thr = container_of(file->private_data, struct fsck_thread, thr);
 229
 230         bch2_fsck_thread_free(thr);
 231         return 0;
 232 }
 233
 234 static bool fsck_thread_ready(struct fsck_thread *thr)
 235 {
 236         return thr->output.buf.pos ||
 237                 thr->output2.nr ||
 238                 thr->thr.done;
 239 }
 240
 241 static ssize_t bch2_fsck_thread_read(struct file *file, char __user *buf,
 242                                      size_t len, loff_t *ppos)
 243 {
 244         struct fsck_thread *thr = container_of(file->private_data, struct fsck_thread, thr);
 245         size_t copied = 0, b;
 246         int ret = 0;
 247
 248         if ((file->f_flags & O_NONBLOCK) &&
 249             !fsck_thread_ready(thr))
 250                 return -EAGAIN;
 251
 252         ret = wait_event_interruptible(thr->output.wait,
 253                         fsck_thread_ready(thr));
 254         if (ret)
 255                 return ret;
 256
 257         if (thr->thr.done)
 258                 return 0;
 259
 260         while (len) {
 261                 ret = darray_make_room(&thr->output2, thr->output.buf.pos);
 262                 if (ret)
 263                         break;
 264
 265                 spin_lock_irq(&thr->output.lock);
 266                 b = min_t(size_t, darray_room(thr->output2), thr->output.buf.pos);
 267
 268                 memcpy(&darray_top(thr->output2), thr->output.buf.buf, b);
 269                 memmove(thr->output.buf.buf,
 270                         thr->output.buf.buf + b,
 271                         thr->output.buf.pos - b);
 272
 273                 thr->output2.nr += b;
 274                 thr->output.buf.pos -= b;
 275                 spin_unlock_irq(&thr->output.lock);
 276
 277                 b = min(len, thr->output2.nr);
 278                 if (!b)
 279                         break;
 280
 281                 b -= copy_to_user(buf, thr->output2.data, b);
 282                 if (!b) {
 283                         ret = -EFAULT;
 284                         break;
 285                 }
 286
 287                 copied  += b;
 288                 buf     += b;
 289                 len     -= b;
 290
 291                 memmove(thr->output2.data,
 292                         thr->output2.data + b,
 293                         thr->output2.nr - b);
 294                 thr->output2.nr -= b;
 295         }
 296
 297         return copied ?: ret;
 298 }
 299
 300 static __poll_t bch2_fsck_thread_poll(struct file *file, struct poll_table_struct *wait)
 301 {
 302         struct fsck_thread *thr = container_of(file->private_data, struct fsck_thread, thr);
 303
 304         poll_wait(file, &thr->output.wait, wait);
 305
 306         return fsck_thread_ready(thr)
 307                 ? EPOLLIN|EPOLLHUP
 308                 : 0;
 309 }
 310
 311 static const struct file_operations fsck_thread_ops = {
 312         .release        = bch2_fsck_thread_release,
 313         .read           = bch2_fsck_thread_read,
 314         .poll           = bch2_fsck_thread_poll,
 315         .llseek         = no_llseek,
 316 };
 317
 318 static int bch2_fsck_offline_thread_fn(void *arg)
 319 {
 320         struct fsck_thread *thr = container_of(arg, struct fsck_thread, thr);
 321         struct bch_fs *c = bch2_fs_open(thr->devs, thr->nr_devs, thr->opts);
 322
 323         thr->thr.ret = PTR_ERR_OR_ZERO(c);
 324         if (!thr->thr.ret)
 325                 bch2_fs_stop(c);
 326
 327         thr->thr.done = true;
 328         wake_up(&thr->output.wait);
 329         return 0;
 330 }
 331
 332 static long bch2_ioctl_fsck_offline(struct bch_ioctl_fsck_offline __user *user_arg)
 333 {
 334         struct bch_ioctl_fsck_offline arg;
 335         struct fsck_thread *thr = NULL;
 336         u64 *devs = NULL;
 337         long ret = 0;
 338
 339         if (copy_from_user(&arg, user_arg, sizeof(arg)))
 340                 return -EFAULT;
 341
 342         if (arg.flags)
 343                 return -EINVAL;
 344
 345         if (!capable(CAP_SYS_ADMIN))
 346                 return -EPERM;
 347
 348         if (!(devs = kcalloc(arg.nr_devs, sizeof(*devs), GFP_KERNEL)) ||
 349             !(thr = kzalloc(sizeof(*thr), GFP_KERNEL)) ||
 350             !(thr->devs = kcalloc(arg.nr_devs, sizeof(*thr->devs), GFP_KERNEL))) {
 351                 ret = -ENOMEM;
 352                 goto err;
 353         }
 354
 355         thr->opts = bch2_opts_empty();
 356         thr->nr_devs = arg.nr_devs;
 357         thr->output.buf = PRINTBUF;
 358         thr->output.buf.atomic++;
 359         spin_lock_init(&thr->output.lock);
 360         init_waitqueue_head(&thr->output.wait);
 361         darray_init(&thr->output2);
 362
 363         if (copy_from_user(devs, &user_arg->devs[0], sizeof(user_arg->devs[0]) * arg.nr_devs)) {
 364                 ret = -EINVAL;
 365                 goto err;
 366         }
 367
 368         for (size_t i = 0; i < arg.nr_devs; i++) {
 369                 thr->devs[i] = strndup_user((char __user *)(unsigned long) devs[i], PATH_MAX);
 370                 ret = PTR_ERR_OR_ZERO(thr->devs[i]);
 371                 if (ret)
 372                         goto err;
 373         }
 374
 375         if (arg.opts) {
 376                 char *optstr = strndup_user((char __user *)(unsigned long) arg.opts, 1 << 16);
 377
 378                 ret =   PTR_ERR_OR_ZERO(optstr) ?:
 379                         bch2_parse_mount_opts(NULL, &thr->opts, optstr);
 380                 kfree(optstr);
 381
 382                 if (ret)
 383                         goto err;
 384         }
 385
 386         opt_set(thr->opts, log_output, (u64)(unsigned long)&thr->output);
 387
 388         ret = run_thread_with_file(&thr->thr,
 389                                    &fsck_thread_ops,
 390                                    bch2_fsck_offline_thread_fn,
 391                                    "bch-fsck");
 392 err:
 393         if (ret < 0) {
 394                 if (thr)
 395                         bch2_fsck_thread_free(thr);
 396                 pr_err("ret %s", bch2_err_str(ret));
 397         }
 398         kfree(devs);
 399         return ret;
 400 }
 401
 402 static long bch2_global_ioctl(unsigned cmd, void __user *arg)
 403 {
 404         long ret;
 405
 406         switch (cmd) {
 407 #if 0
 408         case BCH_IOCTL_ASSEMBLE:
 409                 return bch2_ioctl_assemble(arg);
 410         case BCH_IOCTL_INCREMENTAL:
 411                 return bch2_ioctl_incremental(arg);
 412 #endif
 413         case BCH_IOCTL_FSCK_OFFLINE: {
 414                 ret = bch2_ioctl_fsck_offline(arg);
 415                 break;
 416         }
 417         default:
 418                 ret = -ENOTTY;
 419                 break;
 420         }
 421
 422         if (ret < 0)
 423                 ret = bch2_err_class(ret);
 424         return ret;
 425 }
 426
 427 static long bch2_ioctl_query_uuid(struct bch_fs *c,
 428                         struct bch_ioctl_query_uuid __user *user_arg)
 429 {
 430         return copy_to_user_errcode(&user_arg->uuid, &c->sb.user_uuid,
 431                                     sizeof(c->sb.user_uuid));
 432 }
 433
 434 #if 0
 435 static long bch2_ioctl_start(struct bch_fs *c, struct bch_ioctl_start arg)
 436 {
 437         if (!capable(CAP_SYS_ADMIN))
 438                 return -EPERM;
 439
 440         if (arg.flags || arg.pad)
 441                 return -EINVAL;
 442
 443         return bch2_fs_start(c);
 444 }
 445
 446 static long bch2_ioctl_stop(struct bch_fs *c)
 447 {
 448         if (!capable(CAP_SYS_ADMIN))
 449                 return -EPERM;
 450
 451         bch2_fs_stop(c);
 452         return 0;
 453 }
 454 #endif
 455
 456 static long bch2_ioctl_disk_add(struct bch_fs *c, struct bch_ioctl_disk arg)
 457 {
 458         char *path;
 459         int ret;
 460
 461         if (!capable(CAP_SYS_ADMIN))
 462                 return -EPERM;
 463
 464         if (arg.flags || arg.pad)
 465                 return -EINVAL;
 466
 467         path = strndup_user((const char __user *)(unsigned long) arg.dev, PATH_MAX);
 468         ret = PTR_ERR_OR_ZERO(path);
 469         if (ret)
 470                 return ret;
 471
 472         ret = bch2_dev_add(c, path);
 473         kfree(path);
 474
 475         return ret;
 476 }
 477
 478 static long bch2_ioctl_disk_remove(struct bch_fs *c, struct bch_ioctl_disk arg)
 479 {
 480         struct bch_dev *ca;
 481
 482         if (!capable(CAP_SYS_ADMIN))
 483                 return -EPERM;
 484
 485         if ((arg.flags & ~(BCH_FORCE_IF_DATA_LOST|
 486                            BCH_FORCE_IF_METADATA_LOST|
 487                            BCH_FORCE_IF_DEGRADED|
 488                            BCH_BY_INDEX)) ||
 489             arg.pad)
 490                 return -EINVAL;
 491
 492         ca = bch2_device_lookup(c, arg.dev, arg.flags);
 493         if (IS_ERR(ca))
 494                 return PTR_ERR(ca);
 495
 496         return bch2_dev_remove(c, ca, arg.flags);
 497 }
 498
 499 static long bch2_ioctl_disk_online(struct bch_fs *c, struct bch_ioctl_disk arg)
 500 {
 501         char *path;
 502         int ret;
 503
 504         if (!capable(CAP_SYS_ADMIN))
 505                 return -EPERM;
 506
 507         if (arg.flags || arg.pad)
 508                 return -EINVAL;
 509
 510         path = strndup_user((const char __user *)(unsigned long) arg.dev, PATH_MAX);
 511         ret = PTR_ERR_OR_ZERO(path);
 512         if (ret)
 513                 return ret;
 514
 515         ret = bch2_dev_online(c, path);
 516         kfree(path);
 517         return ret;
 518 }
 519
 520 static long bch2_ioctl_disk_offline(struct bch_fs *c, struct bch_ioctl_disk arg)
 521 {
 522         struct bch_dev *ca;
 523         int ret;
 524
 525         if (!capable(CAP_SYS_ADMIN))
 526                 return -EPERM;
 527
 528         if ((arg.flags & ~(BCH_FORCE_IF_DATA_LOST|
 529                            BCH_FORCE_IF_METADATA_LOST|
 530                            BCH_FORCE_IF_DEGRADED|
 531                            BCH_BY_INDEX)) ||
 532             arg.pad)
 533                 return -EINVAL;
 534
 535         ca = bch2_device_lookup(c, arg.dev, arg.flags);
 536         if (IS_ERR(ca))
 537                 return PTR_ERR(ca);
 538
 539         ret = bch2_dev_offline(c, ca, arg.flags);
 540         percpu_ref_put(&ca->ref);
 541         return ret;
 542 }
 543
 544 static long bch2_ioctl_disk_set_state(struct bch_fs *c,
 545                         struct bch_ioctl_disk_set_state arg)
 546 {
 547         struct bch_dev *ca;
 548         int ret;
 549
 550         if (!capable(CAP_SYS_ADMIN))
 551                 return -EPERM;
 552
 553         if ((arg.flags & ~(BCH_FORCE_IF_DATA_LOST|
 554                            BCH_FORCE_IF_METADATA_LOST|
 555                            BCH_FORCE_IF_DEGRADED|
 556                            BCH_BY_INDEX)) ||
 557             arg.pad[0] || arg.pad[1] || arg.pad[2] ||
 558             arg.new_state >= BCH_MEMBER_STATE_NR)
 559                 return -EINVAL;
 560
 561         ca = bch2_device_lookup(c, arg.dev, arg.flags);
 562         if (IS_ERR(ca))
 563                 return PTR_ERR(ca);
 564
 565         ret = bch2_dev_set_state(c, ca, arg.new_state, arg.flags);
 566         if (ret)
 567                 bch_err(c, "Error setting device state: %s", bch2_err_str(ret));
 568
 569         percpu_ref_put(&ca->ref);
 570         return ret;
 571 }
 572
 573 struct bch_data_ctx {
 574         struct thread_with_file         thr;
 575
 576         struct bch_fs                   *c;
 577         struct bch_ioctl_data           arg;
 578         struct bch_move_stats           stats;
 579 };
 580
 581 static int bch2_data_thread(void *arg)
 582 {
 583         struct bch_data_ctx *ctx = container_of(arg, struct bch_data_ctx, thr);
 584
 585         ctx->thr.ret = bch2_data_job(ctx->c, &ctx->stats, ctx->arg);
 586         ctx->stats.data_type = U8_MAX;
 587         return 0;
 588 }
 589
 590 static int bch2_data_job_release(struct inode *inode, struct file *file)
 591 {
 592         struct bch_data_ctx *ctx = container_of(file->private_data, struct bch_data_ctx, thr);
 593
 594         thread_with_file_exit(&ctx->thr);
 595         kfree(ctx);
 596         return 0;
 597 }
 598
 599 static ssize_t bch2_data_job_read(struct file *file, char __user *buf,
 600                                   size_t len, loff_t *ppos)
 601 {
 602         struct bch_data_ctx *ctx = container_of(file->private_data, struct bch_data_ctx, thr);
 603         struct bch_fs *c = ctx->c;
 604         struct bch_ioctl_data_event e = {
 605                 .type                   = BCH_DATA_EVENT_PROGRESS,
 606                 .p.data_type            = ctx->stats.data_type,
 607                 .p.btree_id             = ctx->stats.pos.btree,
 608                 .p.pos                  = ctx->stats.pos.pos,
 609                 .p.sectors_done         = atomic64_read(&ctx->stats.sectors_seen),
 610                 .p.sectors_total        = bch2_fs_usage_read_short(c).used,
 611         };
 612
 613         if (len < sizeof(e))
 614                 return -EINVAL;
 615
 616         return copy_to_user_errcode(buf, &e, sizeof(e)) ?: sizeof(e);
 617 }
 618
 619 static const struct file_operations bcachefs_data_ops = {
 620         .release        = bch2_data_job_release,
 621         .read           = bch2_data_job_read,
 622         .llseek         = no_llseek,
 623 };
 624
 625 static long bch2_ioctl_data(struct bch_fs *c,
 626                             struct bch_ioctl_data arg)
 627 {
 628         struct bch_data_ctx *ctx;
 629         int ret;
 630
 631         if (!capable(CAP_SYS_ADMIN))
 632                 return -EPERM;
 633
 634         if (arg.op >= BCH_DATA_OP_NR || arg.flags)
 635                 return -EINVAL;
 636
 637         ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
 638         if (!ctx)
 639                 return -ENOMEM;
 640
 641         ctx->c = c;
 642         ctx->arg = arg;
 643
 644         ret = run_thread_with_file(&ctx->thr,
 645                                    &bcachefs_data_ops,
 646                                    bch2_data_thread,
 647                                    "bch-data/%s", c->name);
 648         if (ret < 0)
 649                 kfree(ctx);
 650         return ret;
 651 }
 652
 653 static long bch2_ioctl_fs_usage(struct bch_fs *c,
 654                                 struct bch_ioctl_fs_usage __user *user_arg)
 655 {
 656         struct bch_ioctl_fs_usage *arg = NULL;
 657         struct bch_replicas_usage *dst_e, *dst_end;
 658         struct bch_fs_usage_online *src;
 659         u32 replica_entries_bytes;
 660         unsigned i;
 661         int ret = 0;
 662
 663         if (!test_bit(BCH_FS_started, &c->flags))
 664                 return -EINVAL;
 665
 666         if (get_user(replica_entries_bytes, &user_arg->replica_entries_bytes))
 667                 return -EFAULT;
 668
 669         arg = kzalloc(size_add(sizeof(*arg), replica_entries_bytes), GFP_KERNEL);
 670         if (!arg)
 671                 return -ENOMEM;
 672
 673         src = bch2_fs_usage_read(c);
 674         if (!src) {
 675                 ret = -ENOMEM;
 676                 goto err;
 677         }
 678
 679         arg->capacity           = c->capacity;
 680         arg->used               = bch2_fs_sectors_used(c, src);
 681         arg->online_reserved    = src->online_reserved;
 682
 683         for (i = 0; i < BCH_REPLICAS_MAX; i++)
 684                 arg->persistent_reserved[i] = src->u.persistent_reserved[i];
 685
 686         dst_e   = arg->replicas;
 687         dst_end = (void *) arg->replicas + replica_entries_bytes;
 688
 689         for (i = 0; i < c->replicas.nr; i++) {
 690                 struct bch_replicas_entry_v1 *src_e =
 691                         cpu_replicas_entry(&c->replicas, i);
 692
 693                 /* check that we have enough space for one replicas entry */
 694                 if (dst_e + 1 > dst_end) {
 695                         ret = -ERANGE;
 696                         break;
 697                 }
 698
 699                 dst_e->sectors          = src->u.replicas[i];
 700                 dst_e->r                = *src_e;
 701
 702                 /* recheck after setting nr_devs: */
 703                 if (replicas_usage_next(dst_e) > dst_end) {
 704                         ret = -ERANGE;
 705                         break;
 706                 }
 707
 708                 memcpy(dst_e->r.devs, src_e->devs, src_e->nr_devs);
 709
 710                 dst_e = replicas_usage_next(dst_e);
 711         }
 712
 713         arg->replica_entries_bytes = (void *) dst_e - (void *) arg->replicas;
 714
 715         percpu_up_read(&c->mark_lock);
 716         kfree(src);
 717
 718         if (ret)
 719                 goto err;
 720
 721         ret = copy_to_user_errcode(user_arg, arg,
 722                         sizeof(*arg) + arg->replica_entries_bytes);
 723 err:
 724         kfree(arg);
 725         return ret;
 726 }
 727
 728 /* obsolete, didn't allow for new data types: */
 729 static long bch2_ioctl_dev_usage(struct bch_fs *c,
 730                                  struct bch_ioctl_dev_usage __user *user_arg)
 731 {
 732         struct bch_ioctl_dev_usage arg;
 733         struct bch_dev_usage src;
 734         struct bch_dev *ca;
 735         unsigned i;
 736
 737         if (!test_bit(BCH_FS_started, &c->flags))
 738                 return -EINVAL;
 739
 740         if (copy_from_user(&arg, user_arg, sizeof(arg)))
 741                 return -EFAULT;
 742
 743         if ((arg.flags & ~BCH_BY_INDEX) ||
 744             arg.pad[0] ||
 745             arg.pad[1] ||
 746             arg.pad[2])
 747                 return -EINVAL;
 748
 749         ca = bch2_device_lookup(c, arg.dev, arg.flags);
 750         if (IS_ERR(ca))
 751                 return PTR_ERR(ca);
 752
 753         src = bch2_dev_usage_read(ca);
 754
 755         arg.state               = ca->mi.state;
 756         arg.bucket_size         = ca->mi.bucket_size;
 757         arg.nr_buckets          = ca->mi.nbuckets - ca->mi.first_bucket;
 758
 759         for (i = 0; i < BCH_DATA_NR; i++) {
 760                 arg.d[i].buckets        = src.d[i].buckets;
 761                 arg.d[i].sectors        = src.d[i].sectors;
 762                 arg.d[i].fragmented     = src.d[i].fragmented;
 763         }
 764
 765         percpu_ref_put(&ca->ref);
 766
 767         return copy_to_user_errcode(user_arg, &arg, sizeof(arg));
 768 }
 769
 770 static long bch2_ioctl_dev_usage_v2(struct bch_fs *c,
 771                                  struct bch_ioctl_dev_usage_v2 __user *user_arg)
 772 {
 773         struct bch_ioctl_dev_usage_v2 arg;
 774         struct bch_dev_usage src;
 775         struct bch_dev *ca;
 776         int ret = 0;
 777
 778         if (!test_bit(BCH_FS_started, &c->flags))
 779                 return -EINVAL;
 780
 781         if (copy_from_user(&arg, user_arg, sizeof(arg)))
 782                 return -EFAULT;
 783
 784         if ((arg.flags & ~BCH_BY_INDEX) ||
 785             arg.pad[0] ||
 786             arg.pad[1] ||
 787             arg.pad[2])
 788                 return -EINVAL;
 789
 790         ca = bch2_device_lookup(c, arg.dev, arg.flags);
 791         if (IS_ERR(ca))
 792                 return PTR_ERR(ca);
 793
 794         src = bch2_dev_usage_read(ca);
 795
 796         arg.state               = ca->mi.state;
 797         arg.bucket_size         = ca->mi.bucket_size;
 798         arg.nr_data_types       = min(arg.nr_data_types, BCH_DATA_NR);
 799         arg.nr_buckets          = ca->mi.nbuckets - ca->mi.first_bucket;
 800
 801         ret = copy_to_user_errcode(user_arg, &arg, sizeof(arg));
 802         if (ret)
 803                 goto err;
 804
 805         for (unsigned i = 0; i < arg.nr_data_types; i++) {
 806                 struct bch_ioctl_dev_usage_type t = {
 807                         .buckets        = src.d[i].buckets,
 808                         .sectors        = src.d[i].sectors,
 809                         .fragmented     = src.d[i].fragmented,
 810                 };
 811
 812                 ret = copy_to_user_errcode(&user_arg->d[i], &t, sizeof(t));
 813                 if (ret)
 814                         goto err;
 815         }
 816 err:
 817         percpu_ref_put(&ca->ref);
 818         return ret;
 819 }
 820
 821 static long bch2_ioctl_read_super(struct bch_fs *c,
 822                                   struct bch_ioctl_read_super arg)
 823 {
 824         struct bch_dev *ca = NULL;
 825         struct bch_sb *sb;
 826         int ret = 0;
 827
 828         if (!capable(CAP_SYS_ADMIN))
 829                 return -EPERM;
 830
 831         if ((arg.flags & ~(BCH_BY_INDEX|BCH_READ_DEV)) ||
 832             arg.pad)
 833                 return -EINVAL;
 834
 835         mutex_lock(&c->sb_lock);
 836
 837         if (arg.flags & BCH_READ_DEV) {
 838                 ca = bch2_device_lookup(c, arg.dev, arg.flags);
 839
 840                 if (IS_ERR(ca)) {
 841                         ret = PTR_ERR(ca);
 842                         goto err;
 843                 }
 844
 845                 sb = ca->disk_sb.sb;
 846         } else {
 847                 sb = c->disk_sb.sb;
 848         }
 849
 850         if (vstruct_bytes(sb) > arg.size) {
 851                 ret = -ERANGE;
 852                 goto err;
 853         }
 854
 855         ret = copy_to_user_errcode((void __user *)(unsigned long)arg.sb, sb,
 856                                    vstruct_bytes(sb));
 857 err:
 858         if (!IS_ERR_OR_NULL(ca))
 859                 percpu_ref_put(&ca->ref);
 860         mutex_unlock(&c->sb_lock);
 861         return ret;
 862 }
 863
 864 static long bch2_ioctl_disk_get_idx(struct bch_fs *c,
 865                                     struct bch_ioctl_disk_get_idx arg)
 866 {
 867         dev_t dev = huge_decode_dev(arg.dev);
 868
 869         if (!capable(CAP_SYS_ADMIN))
 870                 return -EPERM;
 871
 872         if (!dev)
 873                 return -EINVAL;
 874
 875         for_each_online_member(c, ca)
 876                 if (ca->dev == dev) {
 877                         percpu_ref_put(&ca->io_ref);
 878                         return ca->dev_idx;
 879                 }
 880
 881         return -BCH_ERR_ENOENT_dev_idx_not_found;
 882 }
 883
 884 static long bch2_ioctl_disk_resize(struct bch_fs *c,
 885                                    struct bch_ioctl_disk_resize arg)
 886 {
 887         struct bch_dev *ca;
 888         int ret;
 889
 890         if (!capable(CAP_SYS_ADMIN))
 891                 return -EPERM;
 892
 893         if ((arg.flags & ~BCH_BY_INDEX) ||
 894             arg.pad)
 895                 return -EINVAL;
 896
 897         ca = bch2_device_lookup(c, arg.dev, arg.flags);
 898         if (IS_ERR(ca))
 899                 return PTR_ERR(ca);
 900
 901         ret = bch2_dev_resize(c, ca, arg.nbuckets);
 902
 903         percpu_ref_put(&ca->ref);
 904         return ret;
 905 }
 906
 907 static long bch2_ioctl_disk_resize_journal(struct bch_fs *c,
 908                                    struct bch_ioctl_disk_resize_journal arg)
 909 {
 910         struct bch_dev *ca;
 911         int ret;
 912
 913         if (!capable(CAP_SYS_ADMIN))
 914                 return -EPERM;
 915
 916         if ((arg.flags & ~BCH_BY_INDEX) ||
 917             arg.pad)
 918                 return -EINVAL;
 919
 920         if (arg.nbuckets > U32_MAX)
 921                 return -EINVAL;
 922
 923         ca = bch2_device_lookup(c, arg.dev, arg.flags);
 924         if (IS_ERR(ca))
 925                 return PTR_ERR(ca);
 926
 927         ret = bch2_set_nr_journal_buckets(c, ca, arg.nbuckets);
 928
 929         percpu_ref_put(&ca->ref);
 930         return ret;
 931 }
 932
 933 static int bch2_fsck_online_thread_fn(void *arg)
 934 {
 935         struct fsck_thread *thr = container_of(arg, struct fsck_thread, thr);
 936         struct bch_fs *c = thr->c;
 937
 938         c->output_filter = current;
 939         c->output = &thr->output;
 940
 941         /*
 942          * XXX: can we figure out a way to do this without mucking with c->opts?
 943          */
 944         if (opt_defined(thr->opts, fix_errors))
 945                 c->opts.fix_errors = thr->opts.fix_errors;
 946         c->opts.fsck = true;
 947
 948         c->curr_recovery_pass = BCH_RECOVERY_PASS_check_alloc_info;
 949         bch2_run_online_recovery_passes(c);
 950
 951         c->output = NULL;
 952         c->output_filter = NULL;
 953
 954         thr->thr.done = true;
 955         wake_up(&thr->output.wait);
 956
 957         up(&c->online_fsck_mutex);
 958         bch2_ro_ref_put(c);
 959         return 0;
 960 }
 961
 962 static long bch2_ioctl_fsck_online(struct bch_fs *c,
 963                                    struct bch_ioctl_fsck_online arg)
 964 {
 965         struct fsck_thread *thr = NULL;
 966         long ret = 0;
 967
 968         if (arg.flags)
 969                 return -EINVAL;
 970
 971         if (!capable(CAP_SYS_ADMIN))
 972                 return -EPERM;
 973
 974         if (!bch2_ro_ref_tryget(c))
 975                 return -EROFS;
 976
 977         if (down_trylock(&c->online_fsck_mutex)) {
 978                 bch2_ro_ref_put(c);
 979                 return -EAGAIN;
 980         }
 981
 982         thr = kzalloc(sizeof(*thr), GFP_KERNEL);
 983         if (!thr) {
 984                 ret = -ENOMEM;
 985                 goto err;
 986         }
 987
 988         thr->c = c;
 989         thr->opts = bch2_opts_empty();
 990         thr->output.buf = PRINTBUF;
 991         thr->output.buf.atomic++;
 992         spin_lock_init(&thr->output.lock);
 993         init_waitqueue_head(&thr->output.wait);
 994         darray_init(&thr->output2);
 995
 996         if (arg.opts) {
 997                 char *optstr = strndup_user((char __user *)(unsigned long) arg.opts, 1 << 16);
 998
 999                 ret =   PTR_ERR_OR_ZERO(optstr) ?:
1000                         bch2_parse_mount_opts(c, &thr->opts, optstr);
1001                 kfree(optstr);
1002
1003                 if (ret)
1004                         goto err;
1005         }
1006
1007         ret = run_thread_with_file(&thr->thr,
1008                                    &fsck_thread_ops,
1009                                    bch2_fsck_online_thread_fn,
1010                                    "bch-fsck");
1011 err:
1012         if (ret < 0) {
1013                 bch_err_fn(c, ret);
1014                 if (thr)
1015                         bch2_fsck_thread_free(thr);
1016                 up(&c->online_fsck_mutex);
1017                 bch2_ro_ref_put(c);
1018         }
1019         return ret;
1020 }
1021
1022 #define BCH_IOCTL(_name, _argtype)                                      \
1023 do {                                                                    \
1024         _argtype i;                                                     \
1025                                                                         \
1026         if (copy_from_user(&i, arg, sizeof(i)))                         \
1027                 return -EFAULT;                                         \
1028         ret = bch2_ioctl_##_name(c, i);                                 \
1029         goto out;                                                       \
1030 } while (0)
1031
1032 long bch2_fs_ioctl(struct bch_fs *c, unsigned cmd, void __user *arg)
1033 {
1034         long ret;
1035
1036         switch (cmd) {
1037         case BCH_IOCTL_QUERY_UUID:
1038                 return bch2_ioctl_query_uuid(c, arg);
1039         case BCH_IOCTL_FS_USAGE:
1040                 return bch2_ioctl_fs_usage(c, arg);
1041         case BCH_IOCTL_DEV_USAGE:
1042                 return bch2_ioctl_dev_usage(c, arg);
1043         case BCH_IOCTL_DEV_USAGE_V2:
1044                 return bch2_ioctl_dev_usage_v2(c, arg);
1045 #if 0
1046         case BCH_IOCTL_START:
1047                 BCH_IOCTL(start, struct bch_ioctl_start);
1048         case BCH_IOCTL_STOP:
1049                 return bch2_ioctl_stop(c);
1050 #endif
1051         case BCH_IOCTL_READ_SUPER:
1052                 BCH_IOCTL(read_super, struct bch_ioctl_read_super);
1053         case BCH_IOCTL_DISK_GET_IDX:
1054                 BCH_IOCTL(disk_get_idx, struct bch_ioctl_disk_get_idx);
1055         }
1056
1057         if (!test_bit(BCH_FS_started, &c->flags))
1058                 return -EINVAL;
1059
1060         switch (cmd) {
1061         case BCH_IOCTL_DISK_ADD:
1062                 BCH_IOCTL(disk_add, struct bch_ioctl_disk);
1063         case BCH_IOCTL_DISK_REMOVE:
1064                 BCH_IOCTL(disk_remove, struct bch_ioctl_disk);
1065         case BCH_IOCTL_DISK_ONLINE:
1066                 BCH_IOCTL(disk_online, struct bch_ioctl_disk);
1067         case BCH_IOCTL_DISK_OFFLINE:
1068                 BCH_IOCTL(disk_offline, struct bch_ioctl_disk);
1069         case BCH_IOCTL_DISK_SET_STATE:
1070                 BCH_IOCTL(disk_set_state, struct bch_ioctl_disk_set_state);
1071         case BCH_IOCTL_DATA:
1072                 BCH_IOCTL(data, struct bch_ioctl_data);
1073         case BCH_IOCTL_DISK_RESIZE:
1074                 BCH_IOCTL(disk_resize, struct bch_ioctl_disk_resize);
1075         case BCH_IOCTL_DISK_RESIZE_JOURNAL:
1076                 BCH_IOCTL(disk_resize_journal, struct bch_ioctl_disk_resize_journal);
1077         case BCH_IOCTL_FSCK_ONLINE:
1078                 BCH_IOCTL(fsck_online, struct bch_ioctl_fsck_online);
1079         default:
1080                 return -ENOTTY;
1081         }
1082 out:
1083         if (ret < 0)
1084                 ret = bch2_err_class(ret);
1085         return ret;
1086 }
1087
1088 static DEFINE_IDR(bch_chardev_minor);
1089
1090 static long bch2_chardev_ioctl(struct file *filp, unsigned cmd, unsigned long v)
1091 {
1092         unsigned minor = iminor(file_inode(filp));
1093         struct bch_fs *c = minor < U8_MAX ? idr_find(&bch_chardev_minor, minor) : NULL;
1094         void __user *arg = (void __user *) v;
1095
1096         return c
1097                 ? bch2_fs_ioctl(c, cmd, arg)
1098                 : bch2_global_ioctl(cmd, arg);
1099 }
1100
1101 static const struct file_operations bch_chardev_fops = {
1102         .owner          = THIS_MODULE,
1103         .unlocked_ioctl = bch2_chardev_ioctl,
1104         .open           = nonseekable_open,
1105 };
1106
1107 static int bch_chardev_major;
1108 static struct class *bch_chardev_class;
1109 static struct device *bch_chardev;
1110
1111 void bch2_fs_chardev_exit(struct bch_fs *c)
1112 {
1113         if (!IS_ERR_OR_NULL(c->chardev))
1114                 device_unregister(c->chardev);
1115         if (c->minor >= 0)
1116                 idr_remove(&bch_chardev_minor, c->minor);
1117 }
1118
1119 int bch2_fs_chardev_init(struct bch_fs *c)
1120 {
1121         c->minor = idr_alloc(&bch_chardev_minor, c, 0, 0, GFP_KERNEL);
1122         if (c->minor < 0)
1123                 return c->minor;
1124
1125         c->chardev = device_create(bch_chardev_class, NULL,
1126                                    MKDEV(bch_chardev_major, c->minor), c,
1127                                    "bcachefs%u-ctl", c->minor);
1128         if (IS_ERR(c->chardev))
1129                 return PTR_ERR(c->chardev);
1130
1131         return 0;
1132 }
1133
1134 void bch2_chardev_exit(void)
1135 {
1136         if (!IS_ERR_OR_NULL(bch_chardev_class))
1137                 device_destroy(bch_chardev_class,
1138                                MKDEV(bch_chardev_major, U8_MAX));
1139         if (!IS_ERR_OR_NULL(bch_chardev_class))
1140                 class_destroy(bch_chardev_class);
1141         if (bch_chardev_major > 0)
1142                 unregister_chrdev(bch_chardev_major, "bcachefs");
1143 }
1144
1145 int __init bch2_chardev_init(void)
1146 {
1147         bch_chardev_major = register_chrdev(0, "bcachefs-ctl", &bch_chardev_fops);
1148         if (bch_chardev_major < 0)
1149                 return bch_chardev_major;
1150
1151         bch_chardev_class = class_create("bcachefs");
1152         if (IS_ERR(bch_chardev_class))
1153                 return PTR_ERR(bch_chardev_class);
1154
1155         bch_chardev = device_create(bch_chardev_class, NULL,
1156                                     MKDEV(bch_chardev_major, U8_MAX),
1157                                     NULL, "bcachefs-ctl");
1158         if (IS_ERR(bch_chardev))
1159                 return PTR_ERR(bch_chardev);
1160
1161         return 0;
1162 }
1163
1164 #endif /* NO_BCACHEFS_CHARDEV */