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, "inconsistency detected - 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_topology_error(struct bch_fs *c)
  29 {
  30         set_bit(BCH_FS_TOPOLOGY_ERROR, &c->flags);
  31         if (test_bit(BCH_FS_INITIAL_GC_DONE, &c->flags))
  32                 bch2_inconsistent_error(c);
  33 }
  34
  35 void bch2_fatal_error(struct bch_fs *c)
  36 {
  37         if (bch2_fs_emergency_read_only(c))
  38                 bch_err(c, "fatal error - emergency read only");
  39 }
  40
  41 void bch2_io_error_work(struct work_struct *work)
  42 {
  43         struct bch_dev *ca = container_of(work, struct bch_dev, io_error_work);
  44         struct bch_fs *c = ca->fs;
  45         bool dev;
  46
  47         down_write(&c->state_lock);
  48         dev = bch2_dev_state_allowed(c, ca, BCH_MEMBER_STATE_ro,
  49                                     BCH_FORCE_IF_DEGRADED);
  50         if (dev
  51             ? __bch2_dev_set_state(c, ca, BCH_MEMBER_STATE_ro,
  52                                   BCH_FORCE_IF_DEGRADED)
  53             : bch2_fs_emergency_read_only(c))
  54                 bch_err(ca,
  55                         "too many IO errors, setting %s RO",
  56                         dev ? "device" : "filesystem");
  57         up_write(&c->state_lock);
  58 }
  59
  60 void bch2_io_error(struct bch_dev *ca)
  61 {
  62         //queue_work(system_long_wq, &ca->io_error_work);
  63 }
  64
  65 #ifdef __KERNEL__
  66 #define ask_yn()        false
  67 #else
  68 #include "tools-util.h"
  69 #endif
  70
  71 int bch2_fsck_err(struct bch_fs *c, unsigned flags, const char *fmt, ...)
  72 {
  73         struct fsck_err_state *s = NULL;
  74         va_list args;
  75         bool fix = false, print = true, suppressing = false;
  76         char _buf[sizeof(s->buf)], *buf = _buf;
  77
  78         if (test_bit(BCH_FS_FSCK_DONE, &c->flags)) {
  79                 va_start(args, fmt);
  80                 vprintk(fmt, args);
  81                 va_end(args);
  82
  83                 if (c->opts.errors == BCH_ON_ERROR_continue) {
  84                         bch_err(c, "fixing");
  85                         return -BCH_ERR_fsck_fix;
  86                 } else {
  87                         bch2_inconsistent_error(c);
  88                         return -BCH_ERR_fsck_errors_not_fixed;
  89                 }
  90         }
  91
  92         mutex_lock(&c->fsck_error_lock);
  93
  94         list_for_each_entry(s, &c->fsck_errors, list)
  95                 if (s->fmt == fmt)
  96                         goto found;
  97
  98         s = kzalloc(sizeof(*s), GFP_NOFS);
  99         if (!s) {
 100                 if (!c->fsck_alloc_err)
 101                         bch_err(c, "kmalloc err, cannot ratelimit fsck errs");
 102                 c->fsck_alloc_err = true;
 103                 buf = _buf;
 104                 goto print;
 105         }
 106
 107         INIT_LIST_HEAD(&s->list);
 108         s->fmt = fmt;
 109 found:
 110         list_move(&s->list, &c->fsck_errors);
 111         s->nr++;
 112         if (c->opts.ratelimit_errors &&
 113             !(flags & FSCK_NO_RATELIMIT) &&
 114             s->nr >= FSCK_ERR_RATELIMIT_NR) {
 115                 if (s->nr == FSCK_ERR_RATELIMIT_NR)
 116                         suppressing = true;
 117                 else
 118                         print = false;
 119         }
 120         buf             = s->buf;
 121 print:
 122         va_start(args, fmt);
 123         vscnprintf(buf, sizeof(_buf), fmt, args);
 124         va_end(args);
 125
 126         if (c->opts.fix_errors == FSCK_OPT_EXIT) {
 127                 bch_err(c, "%s, exiting", buf);
 128         } else if (flags & FSCK_CAN_FIX) {
 129                 if (c->opts.fix_errors == FSCK_OPT_ASK) {
 130                         printk(KERN_ERR "%s: fix?", buf);
 131                         fix = ask_yn();
 132                 } else if (c->opts.fix_errors == FSCK_OPT_YES ||
 133                            (c->opts.nochanges &&
 134                             !(flags & FSCK_CAN_IGNORE))) {
 135                         if (print)
 136                                 bch_err(c, "%s, fixing", buf);
 137                         fix = true;
 138                 } else {
 139                         if (print)
 140                                 bch_err(c, "%s, not fixing", buf);
 141                         fix = false;
 142                 }
 143         } else if (flags & FSCK_NEED_FSCK) {
 144                 if (print)
 145                         bch_err(c, "%s (run fsck to correct)", buf);
 146         } else {
 147                 if (print)
 148                         bch_err(c, "%s (repair unimplemented)", buf);
 149         }
 150
 151         if (suppressing)
 152                 bch_err(c, "Ratelimiting new instances of previous error");
 153
 154         mutex_unlock(&c->fsck_error_lock);
 155
 156         if (fix) {
 157                 set_bit(BCH_FS_ERRORS_FIXED, &c->flags);
 158                 return -BCH_ERR_fsck_fix;
 159         } else {
 160                 set_bit(BCH_FS_ERRORS_NOT_FIXED, &c->flags);
 161                 set_bit(BCH_FS_ERROR, &c->flags);
 162                 return c->opts.fix_errors == FSCK_OPT_EXIT ||
 163                         !(flags & FSCK_CAN_IGNORE)
 164                         ? -BCH_ERR_fsck_errors_not_fixed
 165                         : -BCH_ERR_fsck_ignore;
 166         }
 167 }
 168
 169 void bch2_flush_fsck_errs(struct bch_fs *c)
 170 {
 171         struct fsck_err_state *s, *n;
 172
 173         mutex_lock(&c->fsck_error_lock);
 174
 175         list_for_each_entry_safe(s, n, &c->fsck_errors, list) {
 176                 if (s->ratelimited)
 177                         bch_err(c, "Saw %llu errors like:\n    %s", s->nr, s->buf);
 178
 179                 list_del(&s->list);
 180                 kfree(s);
 181         }
 182
 183         mutex_unlock(&c->fsck_error_lock);
 184 }