[linux.git] / fs / bcachefs / error.c

// SPDX-License-Identifier: GPL-2.0
#include "bcachefs.h"
#include "btree_iter.h"
#include "error.h"
#include "journal.h"
#include "recovery_passes.h"
#include "super.h"
#include "thread_with_file.h"

#define FSCK_ERR_RATELIMIT_NR	10

bool bch2_inconsistent_error(struct bch_fs *c)
{
	set_bit(BCH_FS_error, &c->flags);

	switch (c->opts.errors) {
	case BCH_ON_ERROR_continue:
		return false;
	case BCH_ON_ERROR_fix_safe:
	case BCH_ON_ERROR_ro:
		if (bch2_fs_emergency_read_only(c))
			bch_err(c, "inconsistency detected - emergency read only at journal seq %llu",
				journal_cur_seq(&c->journal));
		return true;
	case BCH_ON_ERROR_panic:
		panic(bch2_fmt(c, "panic after error"));
		return true;
	default:
		BUG();
	}
}

int bch2_topology_error(struct bch_fs *c)
{
	set_bit(BCH_FS_topology_error, &c->flags);
	if (!test_bit(BCH_FS_fsck_running, &c->flags)) {
		bch2_inconsistent_error(c);
		return -BCH_ERR_btree_need_topology_repair;
	} else {
		return bch2_run_explicit_recovery_pass(c, BCH_RECOVERY_PASS_check_topology) ?:
			-BCH_ERR_btree_node_read_validate_error;
	}
}

void bch2_fatal_error(struct bch_fs *c)
{
	if (bch2_fs_emergency_read_only(c))
		bch_err(c, "fatal error - emergency read only");
}

void bch2_io_error_work(struct work_struct *work)
{
	struct bch_dev *ca = container_of(work, struct bch_dev, io_error_work);
	struct bch_fs *c = ca->fs;
	bool dev;

	down_write(&c->state_lock);
	dev = bch2_dev_state_allowed(c, ca, BCH_MEMBER_STATE_ro,
				    BCH_FORCE_IF_DEGRADED);
	if (dev
	    ? __bch2_dev_set_state(c, ca, BCH_MEMBER_STATE_ro,
				  BCH_FORCE_IF_DEGRADED)
	    : bch2_fs_emergency_read_only(c))
		bch_err(ca,
			"too many IO errors, setting %s RO",
			dev ? "device" : "filesystem");
	up_write(&c->state_lock);
}

void bch2_io_error(struct bch_dev *ca, enum bch_member_error_type type)
{
	atomic64_inc(&ca->errors[type]);
	//queue_work(system_long_wq, &ca->io_error_work);
}

enum ask_yn {
	YN_NO,
	YN_YES,
	YN_ALLNO,
	YN_ALLYES,
};

static enum ask_yn parse_yn_response(char *buf)
{
	buf = strim(buf);

	if (strlen(buf) == 1)
		switch (buf[0]) {
		case 'n':
			return YN_NO;
		case 'y':
			return YN_YES;
		case 'N':
			return YN_ALLNO;
		case 'Y':
			return YN_ALLYES;
		}
	return -1;
}

#ifdef __KERNEL__
static enum ask_yn bch2_fsck_ask_yn(struct bch_fs *c, struct btree_trans *trans)
{
	struct stdio_redirect *stdio = c->stdio;

	if (c->stdio_filter && c->stdio_filter != current)
		stdio = NULL;

	if (!stdio)
		return YN_NO;

	if (trans)
		bch2_trans_unlock(trans);

	unsigned long unlock_long_at = trans ? jiffies + HZ * 2 : 0;
	darray_char line = {};
	int ret;

	do {
		unsigned long t;
		bch2_print(c, " (y,n, or Y,N for all errors of this type) ");
rewait:
		t = unlock_long_at
			? max_t(long, unlock_long_at - jiffies, 0)
			: MAX_SCHEDULE_TIMEOUT;

		int r = bch2_stdio_redirect_readline_timeout(stdio, &line, t);
		if (r == -ETIME) {
			bch2_trans_unlock_long(trans);
			unlock_long_at = 0;
			goto rewait;
		}

		if (r < 0) {
			ret = YN_NO;
			break;
		}

		darray_last(line) = '\0';
	} while ((ret = parse_yn_response(line.data)) < 0);

	darray_exit(&line);
	return ret;
}
#else

#include "tools-util.h"

static enum ask_yn bch2_fsck_ask_yn(struct bch_fs *c, struct btree_trans *trans)
{
	char *buf = NULL;
	size_t buflen = 0;
	int ret;

	do {
		fputs(" (y,n, or Y,N for all errors of this type) ", stdout);
		fflush(stdout);

		if (getline(&buf, &buflen, stdin) < 0)
			die("error reading from standard input");
	} while ((ret = parse_yn_response(buf)) < 0);

	free(buf);
	return ret;
}

#endif

static struct fsck_err_state *fsck_err_get(struct bch_fs *c, const char *fmt)
{
	struct fsck_err_state *s;

	if (!test_bit(BCH_FS_fsck_running, &c->flags))
		return NULL;

	list_for_each_entry(s, &c->fsck_error_msgs, list)
		if (s->fmt == fmt) {
			/*
			 * move it to the head of the list: repeated fsck errors
			 * are common
			 */
			list_move(&s->list, &c->fsck_error_msgs);
			return s;
		}

	s = kzalloc(sizeof(*s), GFP_NOFS);
	if (!s) {
		if (!c->fsck_alloc_msgs_err)
			bch_err(c, "kmalloc err, cannot ratelimit fsck errs");
		c->fsck_alloc_msgs_err = true;
		return NULL;
	}

	INIT_LIST_HEAD(&s->list);
	s->fmt = fmt;
	list_add(&s->list, &c->fsck_error_msgs);
	return s;
}

/* s/fix?/fixing/ s/recreate?/recreating/ */
static void prt_actioning(struct printbuf *out, const char *action)
{
	unsigned len = strlen(action);

	BUG_ON(action[len - 1] != '?');
	--len;

	if (action[len - 1] == 'e')
		--len;

	prt_bytes(out, action, len);
	prt_str(out, "ing");
}

static const u8 fsck_flags_extra[] = {
#define x(t, n, flags)		[BCH_FSCK_ERR_##t] = flags,
	BCH_SB_ERRS()
#undef x
};

int __bch2_fsck_err(struct bch_fs *c,
		  struct btree_trans *trans,
		  enum bch_fsck_flags flags,
		  enum bch_sb_error_id err,
		  const char *fmt, ...)
{
	struct fsck_err_state *s = NULL;
	va_list args;
	bool print = true, suppressing = false, inconsistent = false;
	struct printbuf buf = PRINTBUF, *out = &buf;
	int ret = -BCH_ERR_fsck_ignore;
	const char *action_orig = "fix?", *action = action_orig;

	might_sleep();

	if (!WARN_ON(err >= ARRAY_SIZE(fsck_flags_extra)))
		flags |= fsck_flags_extra[err];

	if (!c)
		c = trans->c;

	/*
	 * Ugly: if there's a transaction in the current task it has to be
	 * passed in to unlock if we prompt for user input.
	 *
	 * But, plumbing a transaction and transaction restarts into
	 * bkey_validate() is problematic.
	 *
	 * So:
	 * - make all bkey errors AUTOFIX, they're simple anyways (we just
	 *   delete the key)
	 * - and we don't need to warn if we're not prompting
	 */
	WARN_ON(!(flags & FSCK_AUTOFIX) && !trans && bch2_current_has_btree_trans(c));

	if ((flags & FSCK_CAN_FIX) &&
	    test_bit(err, c->sb.errors_silent))
		return -BCH_ERR_fsck_fix;

	bch2_sb_error_count(c, err);

	va_start(args, fmt);
	prt_vprintf(out, fmt, args);
	va_end(args);

	/* Custom fix/continue/recreate/etc.? */
	if (out->buf[out->pos - 1] == '?') {
		const char *p = strrchr(out->buf, ',');
		if (p) {
			out->pos = p - out->buf;
			action = kstrdup(p + 2, GFP_KERNEL);
			if (!action) {
				ret = -ENOMEM;
				goto err;
			}
		}
	}

	mutex_lock(&c->fsck_error_msgs_lock);
	s = fsck_err_get(c, fmt);
	if (s) {
		/*
		 * We may be called multiple times for the same error on
		 * transaction restart - this memoizes instead of asking the user
		 * multiple times for the same error:
		 */
		if (s->last_msg && !strcmp(buf.buf, s->last_msg)) {
			ret = s->ret;
			mutex_unlock(&c->fsck_error_msgs_lock);
			goto err;
		}

		kfree(s->last_msg);
		s->last_msg = kstrdup(buf.buf, GFP_KERNEL);
		if (!s->last_msg) {
			mutex_unlock(&c->fsck_error_msgs_lock);
			ret = -ENOMEM;
			goto err;
		}

		if (c->opts.ratelimit_errors &&
		    !(flags & FSCK_NO_RATELIMIT) &&
		    s->nr >= FSCK_ERR_RATELIMIT_NR) {
			if (s->nr == FSCK_ERR_RATELIMIT_NR)
				suppressing = true;
			else
				print = false;
		}

		s->nr++;
	}

#ifdef BCACHEFS_LOG_PREFIX
	if (!strncmp(fmt, "bcachefs:", 9))
		prt_printf(out, bch2_log_msg(c, ""));
#endif

	if ((flags & FSCK_CAN_FIX) &&
	    (flags & FSCK_AUTOFIX) &&
	    (c->opts.errors == BCH_ON_ERROR_continue ||
	     c->opts.errors == BCH_ON_ERROR_fix_safe)) {
		prt_str(out, ", ");
		prt_actioning(out, action);
		ret = -BCH_ERR_fsck_fix;
	} else if (!test_bit(BCH_FS_fsck_running, &c->flags)) {
		if (c->opts.errors != BCH_ON_ERROR_continue ||
		    !(flags & (FSCK_CAN_FIX|FSCK_CAN_IGNORE))) {
			prt_str(out, ", shutting down");
			inconsistent = true;
			ret = -BCH_ERR_fsck_errors_not_fixed;
		} else if (flags & FSCK_CAN_FIX) {
			prt_str(out, ", ");
			prt_actioning(out, action);
			ret = -BCH_ERR_fsck_fix;
		} else {
			prt_str(out, ", continuing");
			ret = -BCH_ERR_fsck_ignore;
		}
	} else if (c->opts.fix_errors == FSCK_FIX_exit) {
		prt_str(out, ", exiting");
		ret = -BCH_ERR_fsck_errors_not_fixed;
	} else if (flags & FSCK_CAN_FIX) {
		int fix = s && s->fix
			? s->fix
			: c->opts.fix_errors;

		if (fix == FSCK_FIX_ask) {
			prt_str(out, ", ");
			prt_str(out, action);

			if (bch2_fs_stdio_redirect(c))
				bch2_print(c, "%s", out->buf);
			else
				bch2_print_string_as_lines(KERN_ERR, out->buf);
			print = false;

			int ask = bch2_fsck_ask_yn(c, trans);

			if (trans) {
				ret = bch2_trans_relock(trans);
				if (ret) {
					mutex_unlock(&c->fsck_error_msgs_lock);
					goto err;
				}
			}

			if (ask >= YN_ALLNO && s)
				s->fix = ask == YN_ALLNO
					? FSCK_FIX_no
					: FSCK_FIX_yes;

			ret = ask & 1
				? -BCH_ERR_fsck_fix
				: -BCH_ERR_fsck_ignore;
		} else if (fix == FSCK_FIX_yes ||
			   (c->opts.nochanges &&
			    !(flags & FSCK_CAN_IGNORE))) {
			prt_str(out, ", ");
			prt_actioning(out, action);
			ret = -BCH_ERR_fsck_fix;
		} else {
			prt_str(out, ", not ");
			prt_actioning(out, action);
		}
	} else if (flags & FSCK_NEED_FSCK) {
		prt_str(out, " (run fsck to correct)");
	} else {
		prt_str(out, " (repair unimplemented)");
	}

	if (ret == -BCH_ERR_fsck_ignore &&
	    (c->opts.fix_errors == FSCK_FIX_exit ||
	     !(flags & FSCK_CAN_IGNORE)))
		ret = -BCH_ERR_fsck_errors_not_fixed;

	bool exiting =
		test_bit(BCH_FS_fsck_running, &c->flags) &&
		(ret != -BCH_ERR_fsck_fix &&
		 ret != -BCH_ERR_fsck_ignore);

	if (exiting)
		print = true;

	if (print) {
		if (bch2_fs_stdio_redirect(c))
			bch2_print(c, "%s\n", out->buf);
		else
			bch2_print_string_as_lines(KERN_ERR, out->buf);
	}

	if (exiting)
		bch_err(c, "Unable to continue, halting");
	else if (suppressing)
		bch_err(c, "Ratelimiting new instances of previous error");

	if (s)
		s->ret = ret;

	mutex_unlock(&c->fsck_error_msgs_lock);

	if (inconsistent)
		bch2_inconsistent_error(c);

	if (ret == -BCH_ERR_fsck_fix) {
		set_bit(BCH_FS_errors_fixed, &c->flags);
	} else {
		set_bit(BCH_FS_errors_not_fixed, &c->flags);
		set_bit(BCH_FS_error, &c->flags);
	}
err:
	if (action != action_orig)
		kfree(action);
	printbuf_exit(&buf);
	return ret;
}

int __bch2_bkey_fsck_err(struct bch_fs *c,
			 struct bkey_s_c k,
			 enum bch_validate_flags validate_flags,
			 enum bch_sb_error_id err,
			 const char *fmt, ...)
{
	if (validate_flags & BCH_VALIDATE_silent)
		return -BCH_ERR_fsck_delete_bkey;

	unsigned fsck_flags = 0;
	if (!(validate_flags & (BCH_VALIDATE_write|BCH_VALIDATE_commit)))
		fsck_flags |= FSCK_AUTOFIX|FSCK_CAN_FIX;

	struct printbuf buf = PRINTBUF;
	va_list args;

	prt_str(&buf, "invalid bkey ");
	bch2_bkey_val_to_text(&buf, c, k);
	prt_str(&buf, "\n  ");
	va_start(args, fmt);
	prt_vprintf(&buf, fmt, args);
	va_end(args);
	prt_str(&buf, ": delete?");

	int ret = __bch2_fsck_err(c, NULL, fsck_flags, err, "%s", buf.buf);
	printbuf_exit(&buf);
	return ret;
}

void bch2_flush_fsck_errs(struct bch_fs *c)
{
	struct fsck_err_state *s, *n;

	mutex_lock(&c->fsck_error_msgs_lock);

	list_for_each_entry_safe(s, n, &c->fsck_error_msgs, list) {
		if (s->ratelimited && s->last_msg)
			bch_err(c, "Saw %llu errors like:\n    %s", s->nr, s->last_msg);

		list_del(&s->list);
		kfree(s->last_msg);
		kfree(s);
	}

	mutex_unlock(&c->fsck_error_msgs_lock);
}
Commit	Line	Data
1c6fdbd8 KO	1	// SPDX-License-Identifier: GPL-2.0
1c6fdbd8 KO	2	#include "bcachefs.h"
a850bde6	3	#include "btree_iter.h"
1c6fdbd8	4	#include "error.h"
b3c7fd35	5	#include "journal.h"
d2554263	6	#include "recovery_passes.h"
1c6fdbd8	7	#include "super.h"
96f37eab	8	#include "thread_with_file.h"
1c6fdbd8	9
89b05118 KO	10	#define FSCK_ERR_RATELIMIT_NR 10
89b05118 KO	11
1c6fdbd8 KO	12	bool bch2_inconsistent_error(struct bch_fs *c)
1c6fdbd8 KO	13	{
3c471b65	14	set_bit(BCH_FS_error, &c->flags);
1c6fdbd8 KO	15
1c6fdbd8 KO	16	switch (c->opts.errors) {
2436cb9f	17	case BCH_ON_ERROR_continue:
1c6fdbd8	18	return false;
33dfafa9	19	case BCH_ON_ERROR_fix_safe:
2436cb9f	20	case BCH_ON_ERROR_ro:
1c6fdbd8	21	if (bch2_fs_emergency_read_only(c))
b3c7fd35 KO	22	bch_err(c, "inconsistency detected - emergency read only at journal seq %llu",
b3c7fd35 KO	23	journal_cur_seq(&c->journal));
1c6fdbd8	24	return true;
2436cb9f	25	case BCH_ON_ERROR_panic:
1c6fdbd8 KO	26	panic(bch2_fmt(c, "panic after error"));
	27	return true;
	28	default:
	29	BUG();
	30	}
	31	}
	32
52946d82	33	int bch2_topology_error(struct bch_fs *c)
aae15aaf	34	{
3c471b65	35	set_bit(BCH_FS_topology_error, &c->flags);
52946d82	36	if (!test_bit(BCH_FS_fsck_running, &c->flags)) {
aae15aaf	37	bch2_inconsistent_error(c);
52946d82 KO	38	return -BCH_ERR_btree_need_topology_repair;
	39	} else {
	40	return bch2_run_explicit_recovery_pass(c, BCH_RECOVERY_PASS_check_topology) ?:
	41	-BCH_ERR_btree_node_read_validate_error;
	42	}
aae15aaf KO	43	}
aae15aaf KO	44
1c6fdbd8 KO	45	void bch2_fatal_error(struct bch_fs *c)
	46	{
	47	if (bch2_fs_emergency_read_only(c))
b74b147d	48	bch_err(c, "fatal error - emergency read only");
1c6fdbd8 KO	49	}
	50
	51	void bch2_io_error_work(struct work_struct *work)
	52	{
	53	struct bch_dev *ca = container_of(work, struct bch_dev, io_error_work);
	54	struct bch_fs *c = ca->fs;
	55	bool dev;
	56
1ada1606	57	down_write(&c->state_lock);
2436cb9f	58	dev = bch2_dev_state_allowed(c, ca, BCH_MEMBER_STATE_ro,
1c6fdbd8 KO	59	BCH_FORCE_IF_DEGRADED);
1c6fdbd8 KO	60	if (dev
2436cb9f	61	? __bch2_dev_set_state(c, ca, BCH_MEMBER_STATE_ro,
1c6fdbd8 KO	62	BCH_FORCE_IF_DEGRADED)
	63	: bch2_fs_emergency_read_only(c))
	64	bch_err(ca,
	65	"too many IO errors, setting %s RO",
	66	dev ? "device" : "filesystem");
1ada1606	67	up_write(&c->state_lock);
1c6fdbd8 KO	68	}
1c6fdbd8 KO	69
94119eeb	70	void bch2_io_error(struct bch_dev *ca, enum bch_member_error_type type)
1c6fdbd8	71	{
94119eeb	72	atomic64_inc(&ca->errors[type]);
1c6fdbd8 KO	73	//queue_work(system_long_wq, &ca->io_error_work);
	74	}
	75
853b7393 KO	76	enum ask_yn {
	77	YN_NO,
	78	YN_YES,
	79	YN_ALLNO,
	80	YN_ALLYES,
	81	};
	82
96f37eab KO	83	static enum ask_yn parse_yn_response(char *buf)
	84	{
	85	buf = strim(buf);
	86
	87	if (strlen(buf) == 1)
	88	switch (buf[0]) {
	89	case 'n':
	90	return YN_NO;
	91	case 'y':
	92	return YN_YES;
	93	case 'N':
	94	return YN_ALLNO;
	95	case 'Y':
	96	return YN_ALLYES;
	97	}
	98	return -1;
	99	}
	100
1c6fdbd8	101	#ifdef __KERNEL__
889fb3dc	102	static enum ask_yn bch2_fsck_ask_yn(struct bch_fs c, struct btree_trans trans)
96f37eab KO	103	{
	104	struct stdio_redirect *stdio = c->stdio;
	105
	106	if (c->stdio_filter && c->stdio_filter != current)
	107	stdio = NULL;
	108
	109	if (!stdio)
	110	return YN_NO;
	111
889fb3dc KO	112	if (trans)
	113	bch2_trans_unlock(trans);
	114
	115	unsigned long unlock_long_at = trans ? jiffies + HZ * 2 : 0;
0c97c437	116	darray_char line = {};
96f37eab KO	117	int ret;
	118
	119	do {
889fb3dc	120	unsigned long t;
96f37eab	121	bch2_print(c, " (y,n, or Y,N for all errors of this type) ");
889fb3dc KO	122	rewait:
	123	t = unlock_long_at
	124	? max_t(long, unlock_long_at - jiffies, 0)
	125	: MAX_SCHEDULE_TIMEOUT;
	126
	127	int r = bch2_stdio_redirect_readline_timeout(stdio, &line, t);
	128	if (r == -ETIME) {
	129	bch2_trans_unlock_long(trans);
	130	unlock_long_at = 0;
	131	goto rewait;
	132	}
96f37eab	133
0c97c437 KO	134	if (r < 0) {
	135	ret = YN_NO;
	136	break;
	137	}
889fb3dc	138
0c97c437 KO	139	darray_last(line) = '\0';
0c97c437 KO	140	} while ((ret = parse_yn_response(line.data)) < 0);
96f37eab	141
0c97c437	142	darray_exit(&line);
96f37eab KO	143	return ret;
96f37eab KO	144	}
1c6fdbd8	145	#else
853b7393	146
1c6fdbd8	147	#include "tools-util.h"
853b7393	148
889fb3dc	149	static enum ask_yn bch2_fsck_ask_yn(struct bch_fs c, struct btree_trans trans)
853b7393 KO	150	{
	151	char *buf = NULL;
	152	size_t buflen = 0;
96f37eab	153	int ret;
853b7393	154
96f37eab	155	do {
4f2c166e	156	fputs(" (y,n, or Y,N for all errors of this type) ", stdout);
853b7393 KO	157	fflush(stdout);
	158
	159	if (getline(&buf, &buflen, stdin) < 0)
	160	die("error reading from standard input");
96f37eab	161	} while ((ret = parse_yn_response(buf)) < 0);
853b7393 KO	162
	163	free(buf);
	164	return ret;
	165	}
	166
1c6fdbd8 KO	167	#endif
1c6fdbd8 KO	168
dbb9936b	169	static struct fsck_err_state fsck_err_get(struct bch_fs c, const char *fmt)
1c6fdbd8	170	{
dbb9936b	171	struct fsck_err_state *s;
1c6fdbd8	172
d55ddf6e	173	if (!test_bit(BCH_FS_fsck_running, &c->flags))
dbb9936b	174	return NULL;
1c6fdbd8	175
f5d26fa3	176	list_for_each_entry(s, &c->fsck_error_msgs, list)
dbb9936b KO	177	if (s->fmt == fmt) {
	178	/*
	179	* move it to the head of the list: repeated fsck errors
	180	* are common
	181	*/
f5d26fa3	182	list_move(&s->list, &c->fsck_error_msgs);
dbb9936b KO	183	return s;
dbb9936b KO	184	}
1c6fdbd8	185
beb6db68	186	s = kzalloc(sizeof(*s), GFP_NOFS);
1c6fdbd8	187	if (!s) {
f5d26fa3	188	if (!c->fsck_alloc_msgs_err)
1c6fdbd8	189	bch_err(c, "kmalloc err, cannot ratelimit fsck errs");
f5d26fa3	190	c->fsck_alloc_msgs_err = true;
dbb9936b	191	return NULL;
1c6fdbd8 KO	192	}
	193
	194	INIT_LIST_HEAD(&s->list);
	195	s->fmt = fmt;
f5d26fa3	196	list_add(&s->list, &c->fsck_error_msgs);
dbb9936b KO	197	return s;
	198	}
	199
19391b92 KO	200	/* s/fix?/fixing/ s/recreate?/recreating/ */
	201	static void prt_actioning(struct printbuf out, const char action)
	202	{
	203	unsigned len = strlen(action);
	204
	205	BUG_ON(action[len - 1] != '?');
	206	--len;
	207
	208	if (action[len - 1] == 'e')
	209	--len;
	210
	211	prt_bytes(out, action, len);
	212	prt_str(out, "ing");
	213	}
	214
33dfafa9 KO	215	static const u8 fsck_flags_extra[] = {
	216	#define x(t, n, flags) [BCH_FSCK_ERR_##t] = flags,
	217	BCH_SB_ERRS()
	218	#undef x
	219	};
	220
a850bde6 KO	221	int __bch2_fsck_err(struct bch_fs *c,
a850bde6 KO	222	struct btree_trans *trans,
b65db750 KO	223	enum bch_fsck_flags flags,
	224	enum bch_sb_error_id err,
	225	const char *fmt, ...)
dbb9936b KO	226	{
	227	struct fsck_err_state *s = NULL;
	228	va_list args;
22f51621	229	bool print = true, suppressing = false, inconsistent = false;
dbb9936b KO	230	struct printbuf buf = PRINTBUF, *out = &buf;
dbb9936b KO	231	int ret = -BCH_ERR_fsck_ignore;
19391b92	232	const char action_orig = "fix?", action = action_orig;
dbb9936b	233
e76a2b65 KO	234	might_sleep();
e76a2b65 KO	235
33dfafa9 KO	236	if (!WARN_ON(err >= ARRAY_SIZE(fsck_flags_extra)))
	237	flags \|= fsck_flags_extra[err];
	238
a850bde6 KO	239	if (!c)
	240	c = trans->c;
	241
5612daaf KO	242	/*
	243	* Ugly: if there's a transaction in the current task it has to be
	244	* passed in to unlock if we prompt for user input.
	245	*
	246	* But, plumbing a transaction and transaction restarts into
	247	* bkey_validate() is problematic.
	248	*
	249	* So:
	250	* - make all bkey errors AUTOFIX, they're simple anyways (we just
	251	* delete the key)
	252	* - and we don't need to warn if we're not prompting
	253	*/
	254	WARN_ON(!(flags & FSCK_AUTOFIX) && !trans && bch2_current_has_btree_trans(c));
a850bde6	255
a64a3733 KO	256	if ((flags & FSCK_CAN_FIX) &&
a64a3733 KO	257	test_bit(err, c->sb.errors_silent))
8b16413c KO	258	return -BCH_ERR_fsck_fix;
8b16413c KO	259
b65db750 KO	260	bch2_sb_error_count(c, err);
b65db750 KO	261
9c5d38bb KO	262	va_start(args, fmt);
	263	prt_vprintf(out, fmt, args);
	264	va_end(args);
	265
19391b92 KO	266	/* Custom fix/continue/recreate/etc.? */
	267	if (out->buf[out->pos - 1] == '?') {
	268	const char *p = strrchr(out->buf, ',');
	269	if (p) {
	270	out->pos = p - out->buf;
	271	action = kstrdup(p + 2, GFP_KERNEL);
	272	if (!action) {
	273	ret = -ENOMEM;
	274	goto err;
	275	}
	276	}
	277	}
	278
f5d26fa3	279	mutex_lock(&c->fsck_error_msgs_lock);
dbb9936b KO	280	s = fsck_err_get(c, fmt);
dbb9936b KO	281	if (s) {
c8d5b714 KO	282	/*
	283	* We may be called multiple times for the same error on
	284	* transaction restart - this memoizes instead of asking the user
	285	* multiple times for the same error:
	286	*/
9c5d38bb KO	287	if (s->last_msg && !strcmp(buf.buf, s->last_msg)) {
9c5d38bb KO	288	ret = s->ret;
f5d26fa3	289	mutex_unlock(&c->fsck_error_msgs_lock);
19391b92	290	goto err;
9c5d38bb KO	291	}
	292
	293	kfree(s->last_msg);
	294	s->last_msg = kstrdup(buf.buf, GFP_KERNEL);
19391b92 KO	295	if (!s->last_msg) {
	296	mutex_unlock(&c->fsck_error_msgs_lock);
	297	ret = -ENOMEM;
	298	goto err;
	299	}
9c5d38bb	300
dbb9936b KO	301	if (c->opts.ratelimit_errors &&
	302	!(flags & FSCK_NO_RATELIMIT) &&
	303	s->nr >= FSCK_ERR_RATELIMIT_NR) {
	304	if (s->nr == FSCK_ERR_RATELIMIT_NR)
	305	suppressing = true;
	306	else
	307	print = false;
	308	}
	309
dbb9936b	310	s->nr++;
e2ee3eaa	311	}
dbb9936b	312
b2d1d56b	313	#ifdef BCACHEFS_LOG_PREFIX
dbb9936b KO	314	if (!strncmp(fmt, "bcachefs:", 9))
dbb9936b KO	315	prt_printf(out, bch2_log_msg(c, ""));
b2d1d56b	316	#endif
dbb9936b	317
33dfafa9 KO	318	if ((flags & FSCK_CAN_FIX) &&
	319	(flags & FSCK_AUTOFIX) &&
	320	(c->opts.errors == BCH_ON_ERROR_continue \|\|
	321	c->opts.errors == BCH_ON_ERROR_fix_safe)) {
	322	prt_str(out, ", ");
	323	prt_actioning(out, action);
	324	ret = -BCH_ERR_fsck_fix;
	325	} else if (!test_bit(BCH_FS_fsck_running, &c->flags)) {
dbb9936b KO	326	if (c->opts.errors != BCH_ON_ERROR_continue \|\|
	327	!(flags & (FSCK_CAN_FIX\|FSCK_CAN_IGNORE))) {
	328	prt_str(out, ", shutting down");
22f51621	329	inconsistent = true;
dbb9936b KO	330	ret = -BCH_ERR_fsck_errors_not_fixed;
dbb9936b KO	331	} else if (flags & FSCK_CAN_FIX) {
19391b92 KO	332	prt_str(out, ", ");
19391b92 KO	333	prt_actioning(out, action);
dbb9936b KO	334	ret = -BCH_ERR_fsck_fix;
	335	} else {
	336	prt_str(out, ", continuing");
	337	ret = -BCH_ERR_fsck_ignore;
	338	}
a0f8faea	339	} else if (c->opts.fix_errors == FSCK_FIX_exit) {
dbb9936b KO	340	prt_str(out, ", exiting");
dbb9936b KO	341	ret = -BCH_ERR_fsck_errors_not_fixed;
0bc166ff	342	} else if (flags & FSCK_CAN_FIX) {
853b7393 KO	343	int fix = s && s->fix
	344	? s->fix
	345	: c->opts.fix_errors;
	346
a0f8faea	347	if (fix == FSCK_FIX_ask) {
19391b92 KO	348	prt_str(out, ", ");
19391b92 KO	349	prt_str(out, action);
853b7393	350
96f37eab KO	351	if (bch2_fs_stdio_redirect(c))
	352	bch2_print(c, "%s", out->buf);
	353	else
	354	bch2_print_string_as_lines(KERN_ERR, out->buf);
dbb9936b	355	print = false;
853b7393	356
889fb3dc KO	357	int ask = bch2_fsck_ask_yn(c, trans);
	358
	359	if (trans) {
	360	ret = bch2_trans_relock(trans);
	361	if (ret) {
	362	mutex_unlock(&c->fsck_error_msgs_lock);
	363	goto err;
	364	}
	365	}
853b7393 KO	366
	367	if (ask >= YN_ALLNO && s)
	368	s->fix = ask == YN_ALLNO
a0f8faea KO	369	? FSCK_FIX_no
a0f8faea KO	370	: FSCK_FIX_yes;
853b7393 KO	371
853b7393 KO	372	ret = ask & 1
dbb9936b KO	373	? -BCH_ERR_fsck_fix
dbb9936b KO	374	: -BCH_ERR_fsck_ignore;
a0f8faea	375	} else if (fix == FSCK_FIX_yes \|\|
1c6fdbd8 KO	376	(c->opts.nochanges &&
1c6fdbd8 KO	377	!(flags & FSCK_CAN_IGNORE))) {
19391b92 KO	378	prt_str(out, ", ");
19391b92 KO	379	prt_actioning(out, action);
dbb9936b	380	ret = -BCH_ERR_fsck_fix;
1c6fdbd8	381	} else {
19391b92 KO	382	prt_str(out, ", not ");
19391b92 KO	383	prt_actioning(out, action);
1c6fdbd8 KO	384	}
1c6fdbd8 KO	385	} else if (flags & FSCK_NEED_FSCK) {
dbb9936b	386	prt_str(out, " (run fsck to correct)");
1c6fdbd8	387	} else {
dbb9936b	388	prt_str(out, " (repair unimplemented)");
1c6fdbd8 KO	389	}
1c6fdbd8 KO	390
dbb9936b	391	if (ret == -BCH_ERR_fsck_ignore &&
a0f8faea	392	(c->opts.fix_errors == FSCK_FIX_exit \|\|
dbb9936b KO	393	!(flags & FSCK_CAN_IGNORE)))
	394	ret = -BCH_ERR_fsck_errors_not_fixed;
	395
492e24d7 KO	396	bool exiting =
	397	test_bit(BCH_FS_fsck_running, &c->flags) &&
	398	(ret != -BCH_ERR_fsck_fix &&
	399	ret != -BCH_ERR_fsck_ignore);
	400
	401	if (exiting)
	402	print = true;
	403
96f37eab KO	404	if (print) {
	405	if (bch2_fs_stdio_redirect(c))
	406	bch2_print(c, "%s\n", out->buf);
	407	else
	408	bch2_print_string_as_lines(KERN_ERR, out->buf);
	409	}
dbb9936b	410
492e24d7	411	if (exiting)
dbb9936b KO	412	bch_err(c, "Unable to continue, halting");
dbb9936b KO	413	else if (suppressing)
1c6fdbd8 KO	414	bch_err(c, "Ratelimiting new instances of previous error");
1c6fdbd8 KO	415
9c5d38bb KO	416	if (s)
	417	s->ret = ret;
	418
f5d26fa3	419	mutex_unlock(&c->fsck_error_msgs_lock);
1c6fdbd8	420
22f51621 KO	421	if (inconsistent)
	422	bch2_inconsistent_error(c);
	423
dbb9936b	424	if (ret == -BCH_ERR_fsck_fix) {
3c471b65	425	set_bit(BCH_FS_errors_fixed, &c->flags);
0bc166ff	426	} else {
3c471b65 KO	427	set_bit(BCH_FS_errors_not_fixed, &c->flags);
3c471b65 KO	428	set_bit(BCH_FS_error, &c->flags);
0bc166ff	429	}
19391b92 KO	430	err:
	431	if (action != action_orig)
	432	kfree(action);
	433	printbuf_exit(&buf);
dbb9936b	434	return ret;
1c6fdbd8 KO	435	}
1c6fdbd8 KO	436
d97de0d0 KO	437	int __bch2_bkey_fsck_err(struct bch_fs *c,
d97de0d0 KO	438	struct bkey_s_c k,
658c82f4	439	enum bch_validate_flags validate_flags,
d97de0d0 KO	440	enum bch_sb_error_id err,
	441	const char *fmt, ...)
	442	{
658c82f4 KO	443	if (validate_flags & BCH_VALIDATE_silent)
	444	return -BCH_ERR_fsck_delete_bkey;
	445
	446	unsigned fsck_flags = 0;
	447	if (!(validate_flags & (BCH_VALIDATE_write\|BCH_VALIDATE_commit)))
	448	fsck_flags \|= FSCK_AUTOFIX\|FSCK_CAN_FIX;
	449
d97de0d0 KO	450	struct printbuf buf = PRINTBUF;
	451	va_list args;
	452
	453	prt_str(&buf, "invalid bkey ");
	454	bch2_bkey_val_to_text(&buf, c, k);
	455	prt_str(&buf, "\n ");
	456	va_start(args, fmt);
	457	prt_vprintf(&buf, fmt, args);
	458	va_end(args);
	459	prt_str(&buf, ": delete?");
	460
658c82f4	461	int ret = __bch2_fsck_err(c, NULL, fsck_flags, err, "%s", buf.buf);
d97de0d0 KO	462	printbuf_exit(&buf);
	463	return ret;
	464	}
	465
1c6fdbd8 KO	466	void bch2_flush_fsck_errs(struct bch_fs *c)
	467	{
	468	struct fsck_err_state s, n;
	469
f5d26fa3	470	mutex_lock(&c->fsck_error_msgs_lock);
1c6fdbd8	471
f5d26fa3	472	list_for_each_entry_safe(s, n, &c->fsck_error_msgs, list) {
9c5d38bb KO	473	if (s->ratelimited && s->last_msg)
9c5d38bb KO	474	bch_err(c, "Saw %llu errors like:\n %s", s->nr, s->last_msg);
1c6fdbd8 KO	475
1c6fdbd8 KO	476	list_del(&s->list);
9c5d38bb	477	kfree(s->last_msg);
1c6fdbd8 KO	478	kfree(s);
	479	}
	480
f5d26fa3	481	mutex_unlock(&c->fsck_error_msgs_lock);
1c6fdbd8	482	}