git.sesse.net Git - bcachefs-tools-debian/blob - libbcachefs/error.c

   1 // SPDX-License-Identifier: GPL-2.0
   2 #include "bcachefs.h"
   3 #include "error.h"
   4 #include "io.h"
   5 #include "super.h"
   6
   7 #define FSCK_ERR_RATELIMIT_NR   10
   8
   9 bool bch2_inconsistent_error(struct bch_fs *c)
  10 {
  11         set_bit(BCH_FS_ERROR, &c->flags);
  12
  13         switch (c->opts.errors) {
  14         case BCH_ON_ERROR_CONTINUE:
  15                 return false;
  16         case BCH_ON_ERROR_RO:
  17                 if (bch2_fs_emergency_read_only(c))
  18                         bch_err(c, "emergency read only");
  19                 return true;
  20         case BCH_ON_ERROR_PANIC:
  21                 panic(bch2_fmt(c, "panic after error"));
  22                 return true;
  23         default:
  24                 BUG();
  25         }
  26 }
  27
  28 void bch2_fatal_error(struct bch_fs *c)
  29 {
  30         if (bch2_fs_emergency_read_only(c))
  31                 bch_err(c, "emergency read only");
  32 }
  33
  34 void bch2_io_error_work(struct work_struct *work)
  35 {
  36         struct bch_dev *ca = container_of(work, struct bch_dev, io_error_work);
  37         struct bch_fs *c = ca->fs;
  38         bool dev;
  39
  40         down_write(&c->state_lock);
  41         dev = bch2_dev_state_allowed(c, ca, BCH_MEMBER_STATE_RO,
  42                                     BCH_FORCE_IF_DEGRADED);
  43         if (dev
  44             ? __bch2_dev_set_state(c, ca, BCH_MEMBER_STATE_RO,
  45                                   BCH_FORCE_IF_DEGRADED)
  46             : bch2_fs_emergency_read_only(c))
  47                 bch_err(ca,
  48                         "too many IO errors, setting %s RO",
  49                         dev ? "device" : "filesystem");
  50         up_write(&c->state_lock);
  51 }
  52
  53 void bch2_io_error(struct bch_dev *ca)
  54 {
  55         //queue_work(system_long_wq, &ca->io_error_work);
  56 }
  57
  58 #ifdef __KERNEL__
  59 #define ask_yn()        false
  60 #else
  61 #include "tools-util.h"
  62 #endif
  63
  64 enum fsck_err_ret bch2_fsck_err(struct bch_fs *c, unsigned flags,
  65                                 const char *fmt, ...)
  66 {
  67         struct fsck_err_state *s = NULL;
  68         va_list args;
  69         bool fix = false, print = true, suppressing = false;
  70         char _buf[sizeof(s->buf)], *buf = _buf;
  71
  72         if (test_bit(BCH_FS_FSCK_DONE, &c->flags)) {
  73                 va_start(args, fmt);
  74                 vprintk(fmt, args);
  75                 va_end(args);
  76
  77                 return bch2_inconsistent_error(c)
  78                         ? FSCK_ERR_EXIT
  79                         : FSCK_ERR_FIX;
  80         }
  81
  82         mutex_lock(&c->fsck_error_lock);
  83
  84         list_for_each_entry(s, &c->fsck_errors, list)
  85                 if (s->fmt == fmt)
  86                         goto found;
  87
  88         s = kzalloc(sizeof(*s), GFP_NOFS);
  89         if (!s) {
  90                 if (!c->fsck_alloc_err)
  91                         bch_err(c, "kmalloc err, cannot ratelimit fsck errs");
  92                 c->fsck_alloc_err = true;
  93                 buf = _buf;
  94                 goto print;
  95         }
  96
  97         INIT_LIST_HEAD(&s->list);
  98         s->fmt = fmt;
  99 found:
 100         list_move(&s->list, &c->fsck_errors);
 101         s->nr++;
 102         if (c->opts.ratelimit_errors &&
 103             s->nr >= FSCK_ERR_RATELIMIT_NR) {
 104                 if (s->nr == FSCK_ERR_RATELIMIT_NR)
 105                         suppressing = true;
 106                 else
 107                         print = false;
 108         }
 109         buf             = s->buf;
 110 print:
 111         va_start(args, fmt);
 112         vscnprintf(buf, sizeof(_buf), fmt, args);
 113         va_end(args);
 114
 115         if (c->opts.fix_errors == FSCK_OPT_EXIT) {
 116                 bch_err(c, "%s, exiting", buf);
 117         } else if (flags & FSCK_CAN_FIX) {
 118                 if (c->opts.fix_errors == FSCK_OPT_ASK) {
 119                         printk(KERN_ERR "%s: fix?", buf);
 120                         fix = ask_yn();
 121                 } else if (c->opts.fix_errors == FSCK_OPT_YES ||
 122                            (c->opts.nochanges &&
 123                             !(flags & FSCK_CAN_IGNORE))) {
 124                         if (print)
 125                                 bch_err(c, "%s, fixing", buf);
 126                         fix = true;
 127                 } else {
 128                         if (print)
 129                                 bch_err(c, "%s, not fixing", buf);
 130                         fix = false;
 131                 }
 132         } else if (flags & FSCK_NEED_FSCK) {
 133                 if (print)
 134                         bch_err(c, "%s (run fsck to correct)", buf);
 135         } else {
 136                 if (print)
 137                         bch_err(c, "%s (repair unimplemented)", buf);
 138         }
 139
 140         if (suppressing)
 141                 bch_err(c, "Ratelimiting new instances of previous error");
 142
 143         mutex_unlock(&c->fsck_error_lock);
 144
 145         if (fix) {
 146                 set_bit(BCH_FS_ERRORS_FIXED, &c->flags);
 147                 return FSCK_ERR_FIX;
 148         } else {
 149                 set_bit(BCH_FS_ERROR, &c->flags);
 150                 return c->opts.fix_errors == FSCK_OPT_EXIT ||
 151                         !(flags & FSCK_CAN_IGNORE)
 152                         ? FSCK_ERR_EXIT
 153                         : FSCK_ERR_IGNORE;
 154         }
 155 }
 156
 157 void bch2_flush_fsck_errs(struct bch_fs *c)
 158 {
 159         struct fsck_err_state *s, *n;
 160
 161         mutex_lock(&c->fsck_error_lock);
 162
 163         list_for_each_entry_safe(s, n, &c->fsck_errors, list) {
 164                 if (s->ratelimited)
 165                         bch_err(c, "Saw %llu errors like:\n    %s", s->nr, s->buf);
 166
 167                 list_del(&s->list);
 168                 kfree(s);
 169         }
 170
 171         mutex_unlock(&c->fsck_error_lock);
 172 }