1 // SPDX-License-Identifier: GPL-2.0
6 #define FSCK_ERR_RATELIMIT_NR 10
8 bool bch2_inconsistent_error(struct bch_fs *c)
10 set_bit(BCH_FS_ERROR, &c->flags);
12 switch (c->opts.errors) {
13 case BCH_ON_ERROR_continue:
16 if (bch2_fs_emergency_read_only(c))
17 bch_err(c, "inconsistency detected - emergency read only");
19 case BCH_ON_ERROR_panic:
20 panic(bch2_fmt(c, "panic after error"));
27 void bch2_topology_error(struct bch_fs *c)
29 set_bit(BCH_FS_TOPOLOGY_ERROR, &c->flags);
30 if (test_bit(BCH_FS_FSCK_DONE, &c->flags))
31 bch2_inconsistent_error(c);
34 void bch2_fatal_error(struct bch_fs *c)
36 if (bch2_fs_emergency_read_only(c))
37 bch_err(c, "fatal error - emergency read only");
40 void bch2_io_error_work(struct work_struct *work)
42 struct bch_dev *ca = container_of(work, struct bch_dev, io_error_work);
43 struct bch_fs *c = ca->fs;
46 down_write(&c->state_lock);
47 dev = bch2_dev_state_allowed(c, ca, BCH_MEMBER_STATE_ro,
48 BCH_FORCE_IF_DEGRADED);
50 ? __bch2_dev_set_state(c, ca, BCH_MEMBER_STATE_ro,
51 BCH_FORCE_IF_DEGRADED)
52 : bch2_fs_emergency_read_only(c))
54 "too many IO errors, setting %s RO",
55 dev ? "device" : "filesystem");
56 up_write(&c->state_lock);
59 void bch2_io_error(struct bch_dev *ca)
61 //queue_work(system_long_wq, &ca->io_error_work);
72 #define bch2_fsck_ask_yn() YN_NO
75 #include "tools-util.h"
77 enum ask_yn bch2_fsck_ask_yn(void)
84 fputs(" (y,n, or Y,N for all errors of this type) ", stdout);
87 if (getline(&buf, &buflen, stdin) < 0)
88 die("error reading from standard input");
112 static struct fsck_err_state *fsck_err_get(struct bch_fs *c, const char *fmt)
114 struct fsck_err_state *s;
116 if (test_bit(BCH_FS_FSCK_DONE, &c->flags))
119 list_for_each_entry(s, &c->fsck_errors, list)
122 * move it to the head of the list: repeated fsck errors
125 list_move(&s->list, &c->fsck_errors);
129 s = kzalloc(sizeof(*s), GFP_NOFS);
131 if (!c->fsck_alloc_err)
132 bch_err(c, "kmalloc err, cannot ratelimit fsck errs");
133 c->fsck_alloc_err = true;
137 INIT_LIST_HEAD(&s->list);
139 list_add(&s->list, &c->fsck_errors);
143 int bch2_fsck_err(struct bch_fs *c, unsigned flags, const char *fmt, ...)
145 struct fsck_err_state *s = NULL;
147 bool print = true, suppressing = false, inconsistent = false;
148 struct printbuf buf = PRINTBUF, *out = &buf;
149 int ret = -BCH_ERR_fsck_ignore;
152 prt_vprintf(out, fmt, args);
155 mutex_lock(&c->fsck_error_lock);
156 s = fsck_err_get(c, fmt);
159 * We may be called multiple times for the same error on
160 * transaction restart - this memoizes instead of asking the user
161 * multiple times for the same error:
163 if (s->last_msg && !strcmp(buf.buf, s->last_msg)) {
165 mutex_unlock(&c->fsck_error_lock);
171 s->last_msg = kstrdup(buf.buf, GFP_KERNEL);
173 if (c->opts.ratelimit_errors &&
174 !(flags & FSCK_NO_RATELIMIT) &&
175 s->nr >= FSCK_ERR_RATELIMIT_NR) {
176 if (s->nr == FSCK_ERR_RATELIMIT_NR)
185 #ifdef BCACHEFS_LOG_PREFIX
186 if (!strncmp(fmt, "bcachefs:", 9))
187 prt_printf(out, bch2_log_msg(c, ""));
190 if (test_bit(BCH_FS_FSCK_DONE, &c->flags)) {
191 if (c->opts.errors != BCH_ON_ERROR_continue ||
192 !(flags & (FSCK_CAN_FIX|FSCK_CAN_IGNORE))) {
193 prt_str(out, ", shutting down");
195 ret = -BCH_ERR_fsck_errors_not_fixed;
196 } else if (flags & FSCK_CAN_FIX) {
197 prt_str(out, ", fixing");
198 ret = -BCH_ERR_fsck_fix;
200 prt_str(out, ", continuing");
201 ret = -BCH_ERR_fsck_ignore;
203 } else if (c->opts.fix_errors == FSCK_FIX_exit) {
204 prt_str(out, ", exiting");
205 ret = -BCH_ERR_fsck_errors_not_fixed;
206 } else if (flags & FSCK_CAN_FIX) {
207 int fix = s && s->fix
209 : c->opts.fix_errors;
211 if (fix == FSCK_FIX_ask) {
214 prt_str(out, ": fix?");
215 bch2_print_string_as_lines(KERN_ERR, out->buf);
218 ask = bch2_fsck_ask_yn();
220 if (ask >= YN_ALLNO && s)
221 s->fix = ask == YN_ALLNO
227 : -BCH_ERR_fsck_ignore;
228 } else if (fix == FSCK_FIX_yes ||
229 (c->opts.nochanges &&
230 !(flags & FSCK_CAN_IGNORE))) {
231 prt_str(out, ", fixing");
232 ret = -BCH_ERR_fsck_fix;
234 prt_str(out, ", not fixing");
236 } else if (flags & FSCK_NEED_FSCK) {
237 prt_str(out, " (run fsck to correct)");
239 prt_str(out, " (repair unimplemented)");
242 if (ret == -BCH_ERR_fsck_ignore &&
243 (c->opts.fix_errors == FSCK_FIX_exit ||
244 !(flags & FSCK_CAN_IGNORE)))
245 ret = -BCH_ERR_fsck_errors_not_fixed;
248 bch2_print_string_as_lines(KERN_ERR, out->buf);
250 if (!test_bit(BCH_FS_FSCK_DONE, &c->flags) &&
251 (ret != -BCH_ERR_fsck_fix &&
252 ret != -BCH_ERR_fsck_ignore))
253 bch_err(c, "Unable to continue, halting");
254 else if (suppressing)
255 bch_err(c, "Ratelimiting new instances of previous error");
260 mutex_unlock(&c->fsck_error_lock);
265 bch2_inconsistent_error(c);
267 if (ret == -BCH_ERR_fsck_fix) {
268 set_bit(BCH_FS_ERRORS_FIXED, &c->flags);
270 set_bit(BCH_FS_ERRORS_NOT_FIXED, &c->flags);
271 set_bit(BCH_FS_ERROR, &c->flags);
277 void bch2_flush_fsck_errs(struct bch_fs *c)
279 struct fsck_err_state *s, *n;
281 mutex_lock(&c->fsck_error_lock);
283 list_for_each_entry_safe(s, n, &c->fsck_errors, list) {
284 if (s->ratelimited && s->last_msg)
285 bch_err(c, "Saw %llu errors like:\n %s", s->nr, s->last_msg);
292 mutex_unlock(&c->fsck_error_lock);