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],
 364                            array_size(sizeof(user_arg->devs[0]), arg.nr_devs))) {
 365                 ret = -EINVAL;
 366                 goto err;
 367         }
 368
 369         for (size_t i = 0; i < arg.nr_devs; i++) {
 370                 thr->devs[i] = strndup_user((char __user *)(unsigned long) devs[i], PATH_MAX);
 371                 ret = PTR_ERR_OR_ZERO(thr->devs[i]);
 372                 if (ret)
 373                         goto err;
 374         }
 375
 376         if (arg.opts) {
 377                 char *optstr = strndup_user((char __user *)(unsigned long) arg.opts, 1 << 16);
 378
 379                 ret =   PTR_ERR_OR_ZERO(optstr) ?:
 380                         bch2_parse_mount_opts(NULL, &thr->opts, optstr);
 381                 kfree(optstr);
 382
 383                 if (ret)
 384                         goto err;
 385         }
 386
 387         opt_set(thr->opts, log_output, (u64)(unsigned long)&thr->output);
 388
 389         ret = run_thread_with_file(&thr->thr,
 390                                    &fsck_thread_ops,
 391                                    bch2_fsck_offline_thread_fn,
 392                                    "bch-fsck");
 393 err:
 394         if (ret < 0) {
 395                 if (thr)
 396                         bch2_fsck_thread_free(thr);
 397                 pr_err("ret %s", bch2_err_str(ret));
 398         }
 399         kfree(devs);
 400         return ret;
 401 }
 402
 403 static long bch2_global_ioctl(unsigned cmd, void __user *arg)
 404 {
 405         long ret;
 406
 407         switch (cmd) {
 408 #if 0
 409         case BCH_IOCTL_ASSEMBLE:
 410                 return bch2_ioctl_assemble(arg);
 411         case BCH_IOCTL_INCREMENTAL:
 412                 return bch2_ioctl_incremental(arg);
 413 #endif
 414         case BCH_IOCTL_FSCK_OFFLINE: {
 415                 ret = bch2_ioctl_fsck_offline(arg);
 416                 break;
 417         }
 418         default:
 419                 ret = -ENOTTY;
 420                 break;
 421         }
 422
 423         if (ret < 0)
 424                 ret = bch2_err_class(ret);
 425         return ret;
 426 }
 427
 428 static long bch2_ioctl_query_uuid(struct bch_fs *c,
 429                         struct bch_ioctl_query_uuid __user *user_arg)
 430 {
 431         return copy_to_user_errcode(&user_arg->uuid, &c->sb.user_uuid,
 432                                     sizeof(c->sb.user_uuid));
 433 }
 434
 435 #if 0
 436 static long bch2_ioctl_start(struct bch_fs *c, struct bch_ioctl_start arg)
 437 {
 438         if (!capable(CAP_SYS_ADMIN))
 439                 return -EPERM;
 440
 441         if (arg.flags || arg.pad)
 442                 return -EINVAL;
 443
 444         return bch2_fs_start(c);
 445 }
 446
 447 static long bch2_ioctl_stop(struct bch_fs *c)
 448 {
 449         if (!capable(CAP_SYS_ADMIN))
 450                 return -EPERM;
 451
 452         bch2_fs_stop(c);
 453         return 0;
 454 }
 455 #endif
 456
 457 static long bch2_ioctl_disk_add(struct bch_fs *c, struct bch_ioctl_disk arg)
 458 {
 459         char *path;
 460         int ret;
 461
 462         if (!capable(CAP_SYS_ADMIN))
 463                 return -EPERM;
 464
 465         if (arg.flags || arg.pad)
 466                 return -EINVAL;
 467
 468         path = strndup_user((const char __user *)(unsigned long) arg.dev, PATH_MAX);
 469         ret = PTR_ERR_OR_ZERO(path);
 470         if (ret)
 471                 return ret;
 472
 473         ret = bch2_dev_add(c, path);
 474         kfree(path);
 475
 476         return ret;
 477 }
 478
 479 static long bch2_ioctl_disk_remove(struct bch_fs *c, struct bch_ioctl_disk arg)
 480 {
 481         struct bch_dev *ca;
 482
 483         if (!capable(CAP_SYS_ADMIN))
 484                 return -EPERM;
 485
 486         if ((arg.flags & ~(BCH_FORCE_IF_DATA_LOST|
 487                            BCH_FORCE_IF_METADATA_LOST|
 488                            BCH_FORCE_IF_DEGRADED|
 489                            BCH_BY_INDEX)) ||
 490             arg.pad)
 491                 return -EINVAL;
 492
 493         ca = bch2_device_lookup(c, arg.dev, arg.flags);
 494         if (IS_ERR(ca))
 495                 return PTR_ERR(ca);
 496
 497         return bch2_dev_remove(c, ca, arg.flags);
 498 }
 499
 500 static long bch2_ioctl_disk_online(struct bch_fs *c, struct bch_ioctl_disk arg)
 501 {
 502         char *path;
 503         int ret;
 504
 505         if (!capable(CAP_SYS_ADMIN))
 506                 return -EPERM;
 507
 508         if (arg.flags || arg.pad)
 509                 return -EINVAL;
 510
 511         path = strndup_user((const char __user *)(unsigned long) arg.dev, PATH_MAX);
 512         ret = PTR_ERR_OR_ZERO(path);
 513         if (ret)
 514                 return ret;
 515
 516         ret = bch2_dev_online(c, path);
 517         kfree(path);
 518         return ret;
 519 }
 520
 521 static long bch2_ioctl_disk_offline(struct bch_fs *c, struct bch_ioctl_disk arg)
 522 {
 523         struct bch_dev *ca;
 524         int ret;
 525
 526         if (!capable(CAP_SYS_ADMIN))
 527                 return -EPERM;
 528
 529         if ((arg.flags & ~(BCH_FORCE_IF_DATA_LOST|
 530                            BCH_FORCE_IF_METADATA_LOST|
 531                            BCH_FORCE_IF_DEGRADED|
 532                            BCH_BY_INDEX)) ||
 533             arg.pad)
 534                 return -EINVAL;
 535
 536         ca = bch2_device_lookup(c, arg.dev, arg.flags);
 537         if (IS_ERR(ca))
 538                 return PTR_ERR(ca);
 539
 540         ret = bch2_dev_offline(c, ca, arg.flags);
 541         percpu_ref_put(&ca->ref);
 542         return ret;
 543 }
 544
 545 static long bch2_ioctl_disk_set_state(struct bch_fs *c,
 546                         struct bch_ioctl_disk_set_state arg)
 547 {
 548         struct bch_dev *ca;
 549         int ret;
 550
 551         if (!capable(CAP_SYS_ADMIN))
 552                 return -EPERM;
 553
 554         if ((arg.flags & ~(BCH_FORCE_IF_DATA_LOST|
 555                            BCH_FORCE_IF_METADATA_LOST|
 556                            BCH_FORCE_IF_DEGRADED|
 557                            BCH_BY_INDEX)) ||
 558             arg.pad[0] || arg.pad[1] || arg.pad[2] ||
 559             arg.new_state >= BCH_MEMBER_STATE_NR)
 560                 return -EINVAL;
 561
 562         ca = bch2_device_lookup(c, arg.dev, arg.flags);
 563         if (IS_ERR(ca))
 564                 return PTR_ERR(ca);
 565
 566         ret = bch2_dev_set_state(c, ca, arg.new_state, arg.flags);
 567         if (ret)
 568                 bch_err(c, "Error setting device state: %s", bch2_err_str(ret));
 569
 570         percpu_ref_put(&ca->ref);
 571         return ret;
 572 }
 573
 574 struct bch_data_ctx {
 575         struct thread_with_file         thr;
 576
 577         struct bch_fs                   *c;
 578         struct bch_ioctl_data           arg;
 579         struct bch_move_stats           stats;
 580 };
 581
 582 static int bch2_data_thread(void *arg)
 583 {
 584         struct bch_data_ctx *ctx = container_of(arg, struct bch_data_ctx, thr);
 585
 586         ctx->thr.ret = bch2_data_job(ctx->c, &ctx->stats, ctx->arg);
 587         ctx->stats.data_type = U8_MAX;
 588         return 0;
 589 }
 590
 591 static int bch2_data_job_release(struct inode *inode, struct file *file)
 592 {
 593         struct bch_data_ctx *ctx = container_of(file->private_data, struct bch_data_ctx, thr);
 594
 595         thread_with_file_exit(&ctx->thr);
 596         kfree(ctx);
 597         return 0;
 598 }
 599
 600 static ssize_t bch2_data_job_read(struct file *file, char __user *buf,
 601                                   size_t len, loff_t *ppos)
 602 {
 603         struct bch_data_ctx *ctx = container_of(file->private_data, struct bch_data_ctx, thr);
 604         struct bch_fs *c = ctx->c;
 605         struct bch_ioctl_data_event e = {
 606                 .type                   = BCH_DATA_EVENT_PROGRESS,
 607                 .p.data_type            = ctx->stats.data_type,
 608                 .p.btree_id             = ctx->stats.pos.btree,
 609                 .p.pos                  = ctx->stats.pos.pos,
 610                 .p.sectors_done         = atomic64_read(&ctx->stats.sectors_seen),
 611                 .p.sectors_total        = bch2_fs_usage_read_short(c).used,
 612         };
 613
 614         if (len < sizeof(e))
 615                 return -EINVAL;
 616
 617         return copy_to_user_errcode(buf, &e, sizeof(e)) ?: sizeof(e);
 618 }
 619
 620 static const struct file_operations bcachefs_data_ops = {
 621         .release        = bch2_data_job_release,
 622         .read           = bch2_data_job_read,
 623         .llseek         = no_llseek,
 624 };
 625
 626 static long bch2_ioctl_data(struct bch_fs *c,
 627                             struct bch_ioctl_data arg)
 628 {
 629         struct bch_data_ctx *ctx;
 630         int ret;
 631
 632         if (!capable(CAP_SYS_ADMIN))
 633                 return -EPERM;
 634
 635         if (arg.op >= BCH_DATA_OP_NR || arg.flags)
 636                 return -EINVAL;
 637
 638         ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
 639         if (!ctx)
 640                 return -ENOMEM;
 641
 642         ctx->c = c;
 643         ctx->arg = arg;
 644
 645         ret = run_thread_with_file(&ctx->thr,
 646                                    &bcachefs_data_ops,
 647                                    bch2_data_thread,
 648                                    "bch-data/%s", c->name);
 649         if (ret < 0)
 650                 kfree(ctx);
 651         return ret;
 652 }
 653
 654 static long bch2_ioctl_fs_usage(struct bch_fs *c,
 655                                 struct bch_ioctl_fs_usage __user *user_arg)
 656 {
 657         struct bch_ioctl_fs_usage *arg = NULL;
 658         struct bch_replicas_usage *dst_e, *dst_end;
 659         struct bch_fs_usage_online *src;
 660         u32 replica_entries_bytes;
 661         unsigned i;
 662         int ret = 0;
 663
 664         if (!test_bit(BCH_FS_started, &c->flags))
 665                 return -EINVAL;
 666
 667         if (get_user(replica_entries_bytes, &user_arg->replica_entries_bytes))
 668                 return -EFAULT;
 669
 670         arg = kzalloc(size_add(sizeof(*arg), replica_entries_bytes), GFP_KERNEL);
 671         if (!arg)
 672                 return -ENOMEM;
 673
 674         src = bch2_fs_usage_read(c);
 675         if (!src) {
 676                 ret = -ENOMEM;
 677                 goto err;
 678         }
 679
 680         arg->capacity           = c->capacity;
 681         arg->used               = bch2_fs_sectors_used(c, src);
 682         arg->online_reserved    = src->online_reserved;
 683
 684         for (i = 0; i < BCH_REPLICAS_MAX; i++)
 685                 arg->persistent_reserved[i] = src->u.persistent_reserved[i];
 686
 687         dst_e   = arg->replicas;
 688         dst_end = (void *) arg->replicas + replica_entries_bytes;
 689
 690         for (i = 0; i < c->replicas.nr; i++) {
 691                 struct bch_replicas_entry_v1 *src_e =
 692                         cpu_replicas_entry(&c->replicas, i);
 693
 694                 /* check that we have enough space for one replicas entry */
 695                 if (dst_e + 1 > dst_end) {
 696                         ret = -ERANGE;
 697                         break;
 698                 }
 699
 700                 dst_e->sectors          = src->u.replicas[i];
 701                 dst_e->r                = *src_e;
 702
 703                 /* recheck after setting nr_devs: */
 704                 if (replicas_usage_next(dst_e) > dst_end) {
 705                         ret = -ERANGE;
 706                         break;
 707                 }
 708
 709                 memcpy(dst_e->r.devs, src_e->devs, src_e->nr_devs);
 710
 711                 dst_e = replicas_usage_next(dst_e);
 712         }
 713
 714         arg->replica_entries_bytes = (void *) dst_e - (void *) arg->replicas;
 715
 716         percpu_up_read(&c->mark_lock);
 717         kfree(src);
 718
 719         if (ret)
 720                 goto err;
 721
 722         ret = copy_to_user_errcode(user_arg, arg,
 723                         sizeof(*arg) + arg->replica_entries_bytes);
 724 err:
 725         kfree(arg);
 726         return ret;
 727 }
 728
 729 /* obsolete, didn't allow for new data types: */
 730 static long bch2_ioctl_dev_usage(struct bch_fs *c,
 731                                  struct bch_ioctl_dev_usage __user *user_arg)
 732 {
 733         struct bch_ioctl_dev_usage arg;
 734         struct bch_dev_usage src;
 735         struct bch_dev *ca;
 736         unsigned i;
 737
 738         if (!test_bit(BCH_FS_started, &c->flags))
 739                 return -EINVAL;
 740
 741         if (copy_from_user(&arg, user_arg, sizeof(arg)))
 742                 return -EFAULT;
 743
 744         if ((arg.flags & ~BCH_BY_INDEX) ||
 745             arg.pad[0] ||
 746             arg.pad[1] ||
 747             arg.pad[2])
 748                 return -EINVAL;
 749
 750         ca = bch2_device_lookup(c, arg.dev, arg.flags);
 751         if (IS_ERR(ca))
 752                 return PTR_ERR(ca);
 753
 754         src = bch2_dev_usage_read(ca);
 755
 756         arg.state               = ca->mi.state;
 757         arg.bucket_size         = ca->mi.bucket_size;
 758         arg.nr_buckets          = ca->mi.nbuckets - ca->mi.first_bucket;
 759
 760         for (i = 0; i < BCH_DATA_NR; i++) {
 761                 arg.d[i].buckets        = src.d[i].buckets;
 762                 arg.d[i].sectors        = src.d[i].sectors;
 763                 arg.d[i].fragmented     = src.d[i].fragmented;
 764         }
 765
 766         percpu_ref_put(&ca->ref);
 767
 768         return copy_to_user_errcode(user_arg, &arg, sizeof(arg));
 769 }
 770
 771 static long bch2_ioctl_dev_usage_v2(struct bch_fs *c,
 772                                  struct bch_ioctl_dev_usage_v2 __user *user_arg)
 773 {
 774         struct bch_ioctl_dev_usage_v2 arg;
 775         struct bch_dev_usage src;
 776         struct bch_dev *ca;
 777         int ret = 0;
 778
 779         if (!test_bit(BCH_FS_started, &c->flags))
 780                 return -EINVAL;
 781
 782         if (copy_from_user(&arg, user_arg, sizeof(arg)))
 783                 return -EFAULT;
 784
 785         if ((arg.flags & ~BCH_BY_INDEX) ||
 786             arg.pad[0] ||
 787             arg.pad[1] ||
 788             arg.pad[2])
 789                 return -EINVAL;
 790
 791         ca = bch2_device_lookup(c, arg.dev, arg.flags);
 792         if (IS_ERR(ca))
 793                 return PTR_ERR(ca);
 794
 795         src = bch2_dev_usage_read(ca);
 796
 797         arg.state               = ca->mi.state;
 798         arg.bucket_size         = ca->mi.bucket_size;
 799         arg.nr_data_types       = min(arg.nr_data_types, BCH_DATA_NR);
 800         arg.nr_buckets          = ca->mi.nbuckets - ca->mi.first_bucket;
 801
 802         ret = copy_to_user_errcode(user_arg, &arg, sizeof(arg));
 803         if (ret)
 804                 goto err;
 805
 806         for (unsigned i = 0; i < arg.nr_data_types; i++) {
 807                 struct bch_ioctl_dev_usage_type t = {
 808                         .buckets        = src.d[i].buckets,
 809                         .sectors        = src.d[i].sectors,
 810                         .fragmented     = src.d[i].fragmented,
 811                 };
 812
 813                 ret = copy_to_user_errcode(&user_arg->d[i], &t, sizeof(t));
 814                 if (ret)
 815                         goto err;
 816         }
 817 err:
 818         percpu_ref_put(&ca->ref);
 819         return ret;
 820 }
 821
 822 static long bch2_ioctl_read_super(struct bch_fs *c,
 823                                   struct bch_ioctl_read_super arg)
 824 {
 825         struct bch_dev *ca = NULL;
 826         struct bch_sb *sb;
 827         int ret = 0;
 828
 829         if (!capable(CAP_SYS_ADMIN))
 830                 return -EPERM;
 831
 832         if ((arg.flags & ~(BCH_BY_INDEX|BCH_READ_DEV)) ||
 833             arg.pad)
 834                 return -EINVAL;
 835
 836         mutex_lock(&c->sb_lock);
 837
 838         if (arg.flags & BCH_READ_DEV) {
 839                 ca = bch2_device_lookup(c, arg.dev, arg.flags);
 840
 841                 if (IS_ERR(ca)) {
 842                         ret = PTR_ERR(ca);
 843                         goto err;
 844                 }
 845
 846                 sb = ca->disk_sb.sb;
 847         } else {
 848                 sb = c->disk_sb.sb;
 849         }
 850
 851         if (vstruct_bytes(sb) > arg.size) {
 852                 ret = -ERANGE;
 853                 goto err;
 854         }
 855
 856         ret = copy_to_user_errcode((void __user *)(unsigned long)arg.sb, sb,
 857                                    vstruct_bytes(sb));
 858 err:
 859         if (!IS_ERR_OR_NULL(ca))
 860                 percpu_ref_put(&ca->ref);
 861         mutex_unlock(&c->sb_lock);
 862         return ret;
 863 }
 864
 865 static long bch2_ioctl_disk_get_idx(struct bch_fs *c,
 866                                     struct bch_ioctl_disk_get_idx arg)
 867 {
 868         dev_t dev = huge_decode_dev(arg.dev);
 869
 870         if (!capable(CAP_SYS_ADMIN))
 871                 return -EPERM;
 872
 873         if (!dev)
 874                 return -EINVAL;
 875
 876         for_each_online_member(c, ca)
 877                 if (ca->dev == dev) {
 878                         percpu_ref_put(&ca->io_ref);
 879                         return ca->dev_idx;
 880                 }
 881
 882         return -BCH_ERR_ENOENT_dev_idx_not_found;
 883 }
 884
 885 static long bch2_ioctl_disk_resize(struct bch_fs *c,
 886                                    struct bch_ioctl_disk_resize arg)
 887 {
 888         struct bch_dev *ca;
 889         int ret;
 890
 891         if (!capable(CAP_SYS_ADMIN))
 892                 return -EPERM;
 893
 894         if ((arg.flags & ~BCH_BY_INDEX) ||
 895             arg.pad)
 896                 return -EINVAL;
 897
 898         ca = bch2_device_lookup(c, arg.dev, arg.flags);
 899         if (IS_ERR(ca))
 900                 return PTR_ERR(ca);
 901
 902         ret = bch2_dev_resize(c, ca, arg.nbuckets);
 903
 904         percpu_ref_put(&ca->ref);
 905         return ret;
 906 }
 907
 908 static long bch2_ioctl_disk_resize_journal(struct bch_fs *c,
 909                                    struct bch_ioctl_disk_resize_journal arg)
 910 {
 911         struct bch_dev *ca;
 912         int ret;
 913
 914         if (!capable(CAP_SYS_ADMIN))
 915                 return -EPERM;
 916
 917         if ((arg.flags & ~BCH_BY_INDEX) ||
 918             arg.pad)
 919                 return -EINVAL;
 920
 921         if (arg.nbuckets > U32_MAX)
 922                 return -EINVAL;
 923
 924         ca = bch2_device_lookup(c, arg.dev, arg.flags);
 925         if (IS_ERR(ca))
 926                 return PTR_ERR(ca);
 927
 928         ret = bch2_set_nr_journal_buckets(c, ca, arg.nbuckets);
 929
 930         percpu_ref_put(&ca->ref);
 931         return ret;
 932 }
 933
 934 static int bch2_fsck_online_thread_fn(void *arg)
 935 {
 936         struct fsck_thread *thr = container_of(arg, struct fsck_thread, thr);
 937         struct bch_fs *c = thr->c;
 938
 939         c->output_filter = current;
 940         c->output = &thr->output;
 941
 942         /*
 943          * XXX: can we figure out a way to do this without mucking with c->opts?
 944          */
 945         if (opt_defined(thr->opts, fix_errors))
 946                 c->opts.fix_errors = thr->opts.fix_errors;
 947         c->opts.fsck = true;
 948
 949         c->curr_recovery_pass = BCH_RECOVERY_PASS_check_alloc_info;
 950         bch2_run_online_recovery_passes(c);
 951
 952         c->output = NULL;
 953         c->output_filter = NULL;
 954
 955         thr->thr.done = true;
 956         wake_up(&thr->output.wait);
 957
 958         up(&c->online_fsck_mutex);
 959         bch2_ro_ref_put(c);
 960         return 0;
 961 }
 962
 963 static long bch2_ioctl_fsck_online(struct bch_fs *c,
 964                                    struct bch_ioctl_fsck_online arg)
 965 {
 966         struct fsck_thread *thr = NULL;
 967         long ret = 0;
 968
 969         if (arg.flags)
 970                 return -EINVAL;
 971
 972         if (!capable(CAP_SYS_ADMIN))
 973                 return -EPERM;
 974
 975         if (!bch2_ro_ref_tryget(c))
 976                 return -EROFS;
 977
 978         if (down_trylock(&c->online_fsck_mutex)) {
 979                 bch2_ro_ref_put(c);
 980                 return -EAGAIN;
 981         }
 982
 983         thr = kzalloc(sizeof(*thr), GFP_KERNEL);
 984         if (!thr) {
 985                 ret = -ENOMEM;
 986                 goto err;
 987         }
 988
 989         thr->c = c;
 990         thr->opts = bch2_opts_empty();
 991         thr->output.buf = PRINTBUF;
 992         thr->output.buf.atomic++;
 993         spin_lock_init(&thr->output.lock);
 994         init_waitqueue_head(&thr->output.wait);
 995         darray_init(&thr->output2);
 996
 997         if (arg.opts) {
 998                 char *optstr = strndup_user((char __user *)(unsigned long) arg.opts, 1 << 16);
 999
1000                 ret =   PTR_ERR_OR_ZERO(optstr) ?:
1001                         bch2_parse_mount_opts(c, &thr->opts, optstr);
1002                 kfree(optstr);
1003
1004                 if (ret)
1005                         goto err;
1006         }
1007
1008         ret = run_thread_with_file(&thr->thr,
1009                                    &fsck_thread_ops,
1010                                    bch2_fsck_online_thread_fn,
1011                                    "bch-fsck");
1012 err:
1013         if (ret < 0) {
1014                 bch_err_fn(c, ret);
1015                 if (thr)
1016                         bch2_fsck_thread_free(thr);
1017                 up(&c->online_fsck_mutex);
1018                 bch2_ro_ref_put(c);
1019         }
1020         return ret;
1021 }
1022
1023 #define BCH_IOCTL(_name, _argtype)                                      \
1024 do {                                                                    \
1025         _argtype i;                                                     \
1026                                                                         \
1027         if (copy_from_user(&i, arg, sizeof(i)))                         \
1028                 return -EFAULT;                                         \
1029         ret = bch2_ioctl_##_name(c, i);                                 \
1030         goto out;                                                       \
1031 } while (0)
1032
1033 long bch2_fs_ioctl(struct bch_fs *c, unsigned cmd, void __user *arg)
1034 {
1035         long ret;
1036
1037         switch (cmd) {
1038         case BCH_IOCTL_QUERY_UUID:
1039                 return bch2_ioctl_query_uuid(c, arg);
1040         case BCH_IOCTL_FS_USAGE:
1041                 return bch2_ioctl_fs_usage(c, arg);
1042         case BCH_IOCTL_DEV_USAGE:
1043                 return bch2_ioctl_dev_usage(c, arg);
1044         case BCH_IOCTL_DEV_USAGE_V2:
1045                 return bch2_ioctl_dev_usage_v2(c, arg);
1046 #if 0
1047         case BCH_IOCTL_START:
1048                 BCH_IOCTL(start, struct bch_ioctl_start);
1049         case BCH_IOCTL_STOP:
1050                 return bch2_ioctl_stop(c);
1051 #endif
1052         case BCH_IOCTL_READ_SUPER:
1053                 BCH_IOCTL(read_super, struct bch_ioctl_read_super);
1054         case BCH_IOCTL_DISK_GET_IDX:
1055                 BCH_IOCTL(disk_get_idx, struct bch_ioctl_disk_get_idx);
1056         }
1057
1058         if (!test_bit(BCH_FS_started, &c->flags))
1059                 return -EINVAL;
1060
1061         switch (cmd) {
1062         case BCH_IOCTL_DISK_ADD:
1063                 BCH_IOCTL(disk_add, struct bch_ioctl_disk);
1064         case BCH_IOCTL_DISK_REMOVE:
1065                 BCH_IOCTL(disk_remove, struct bch_ioctl_disk);
1066         case BCH_IOCTL_DISK_ONLINE:
1067                 BCH_IOCTL(disk_online, struct bch_ioctl_disk);
1068         case BCH_IOCTL_DISK_OFFLINE:
1069                 BCH_IOCTL(disk_offline, struct bch_ioctl_disk);
1070         case BCH_IOCTL_DISK_SET_STATE:
1071                 BCH_IOCTL(disk_set_state, struct bch_ioctl_disk_set_state);
1072         case BCH_IOCTL_DATA:
1073                 BCH_IOCTL(data, struct bch_ioctl_data);
1074         case BCH_IOCTL_DISK_RESIZE:
1075                 BCH_IOCTL(disk_resize, struct bch_ioctl_disk_resize);
1076         case BCH_IOCTL_DISK_RESIZE_JOURNAL:
1077                 BCH_IOCTL(disk_resize_journal, struct bch_ioctl_disk_resize_journal);
1078         case BCH_IOCTL_FSCK_ONLINE:
1079                 BCH_IOCTL(fsck_online, struct bch_ioctl_fsck_online);
1080         default:
1081                 return -ENOTTY;
1082         }
1083 out:
1084         if (ret < 0)
1085                 ret = bch2_err_class(ret);
1086         return ret;
1087 }
1088
1089 static DEFINE_IDR(bch_chardev_minor);
1090
1091 static long bch2_chardev_ioctl(struct file *filp, unsigned cmd, unsigned long v)
1092 {
1093         unsigned minor = iminor(file_inode(filp));
1094         struct bch_fs *c = minor < U8_MAX ? idr_find(&bch_chardev_minor, minor) : NULL;
1095         void __user *arg = (void __user *) v;
1096
1097         return c
1098                 ? bch2_fs_ioctl(c, cmd, arg)
1099                 : bch2_global_ioctl(cmd, arg);
1100 }
1101
1102 static const struct file_operations bch_chardev_fops = {
1103         .owner          = THIS_MODULE,
1104         .unlocked_ioctl = bch2_chardev_ioctl,
1105         .open           = nonseekable_open,
1106 };
1107
1108 static int bch_chardev_major;
1109 static struct class *bch_chardev_class;
1110 static struct device *bch_chardev;
1111
1112 void bch2_fs_chardev_exit(struct bch_fs *c)
1113 {
1114         if (!IS_ERR_OR_NULL(c->chardev))
1115                 device_unregister(c->chardev);
1116         if (c->minor >= 0)
1117                 idr_remove(&bch_chardev_minor, c->minor);
1118 }
1119
1120 int bch2_fs_chardev_init(struct bch_fs *c)
1121 {
1122         c->minor = idr_alloc(&bch_chardev_minor, c, 0, 0, GFP_KERNEL);
1123         if (c->minor < 0)
1124                 return c->minor;
1125
1126         c->chardev = device_create(bch_chardev_class, NULL,
1127                                    MKDEV(bch_chardev_major, c->minor), c,
1128                                    "bcachefs%u-ctl", c->minor);
1129         if (IS_ERR(c->chardev))
1130                 return PTR_ERR(c->chardev);
1131
1132         return 0;
1133 }
1134
1135 void bch2_chardev_exit(void)
1136 {
1137         if (!IS_ERR_OR_NULL(bch_chardev_class))
1138                 device_destroy(bch_chardev_class,
1139                                MKDEV(bch_chardev_major, U8_MAX));
1140         if (!IS_ERR_OR_NULL(bch_chardev_class))
1141                 class_destroy(bch_chardev_class);
1142         if (bch_chardev_major > 0)
1143                 unregister_chrdev(bch_chardev_major, "bcachefs");
1144 }
1145
1146 int __init bch2_chardev_init(void)
1147 {
1148         bch_chardev_major = register_chrdev(0, "bcachefs-ctl", &bch_chardev_fops);
1149         if (bch_chardev_major < 0)
1150                 return bch_chardev_major;
1151
1152         bch_chardev_class = class_create("bcachefs");
1153         if (IS_ERR(bch_chardev_class))
1154                 return PTR_ERR(bch_chardev_class);
1155
1156         bch_chardev = device_create(bch_chardev_class, NULL,
1157                                     MKDEV(bch_chardev_major, U8_MAX),
1158                                     NULL, "bcachefs-ctl");
1159         if (IS_ERR(bch_chardev))
1160                 return PTR_ERR(bch_chardev);
1161
1162         return 0;
1163 }
1164
1165 #endif /* NO_BCACHEFS_CHARDEV */