path: root/fs/bcachefs/journal.c

// SPDX-License-Identifier: GPL-2.0
/*
 * bcachefs journalling code, for btree insertions
 *
 * Copyright 2012 Google, Inc.
 */

#include "bcachefs.h"
#include "alloc_foreground.h"
#include "bkey_methods.h"
#include "btree_gc.h"
#include "btree_update.h"
#include "btree_write_buffer.h"
#include "buckets.h"
#include "error.h"
#include "journal.h"
#include "journal_io.h"
#include "journal_reclaim.h"
#include "journal_sb.h"
#include "journal_seq_blacklist.h"
#include "trace.h"

static const char * const bch2_journal_errors[] = {
#define x(n)	#n,
	JOURNAL_ERRORS()
#undef x
	NULL
};

static inline bool journal_seq_unwritten(struct journal *j, u64 seq)
{
	return seq > j->seq_ondisk;
}

static bool __journal_entry_is_open(union journal_res_state state)
{
	return state.cur_entry_offset < JOURNAL_ENTRY_CLOSED_VAL;
}

static inline unsigned nr_unwritten_journal_entries(struct journal *j)
{
	return atomic64_read(&j->seq) - j->seq_ondisk;
}

static bool journal_entry_is_open(struct journal *j)
{
	return __journal_entry_is_open(j->reservations);
}

static void bch2_journal_buf_to_text(struct printbuf *out, struct journal *j, u64 seq)
{
	union journal_res_state s = READ_ONCE(j->reservations);
	unsigned i = seq & JOURNAL_BUF_MASK;
	struct journal_buf *buf = j->buf + i;

	prt_printf(out, "seq:\t%llu\n", seq);
	printbuf_indent_add(out, 2);

	prt_printf(out, "refcount:\t%u\n", journal_state_count(s, i));

	prt_printf(out, "size:\t");
	prt_human_readable_u64(out, vstruct_bytes(buf->data));
	prt_newline(out);

	prt_printf(out, "expires:\t");
	prt_printf(out, "%li jiffies\n", buf->expires - jiffies);

	prt_printf(out, "flags:\t");
	if (buf->noflush)
		prt_str(out, "noflush ");
	if (buf->must_flush)
		prt_str(out, "must_flush ");
	if (buf->separate_flush)
		prt_str(out, "separate_flush ");
	if (buf->need_flush_to_write_buffer)
		prt_str(out, "need_flush_to_write_buffer ");
	if (buf->write_started)
		prt_str(out, "write_started ");
	if (buf->write_allocated)
		prt_str(out, "write_allocated ");
	if (buf->write_done)
		prt_str(out, "write_done");
	prt_newline(out);

	printbuf_indent_sub(out, 2);
}

static void bch2_journal_bufs_to_text(struct printbuf *out, struct journal *j)
{
	if (!out->nr_tabstops)
		printbuf_tabstop_push(out, 24);

	for (u64 seq = journal_last_unwritten_seq(j);
	     seq <= journal_cur_seq(j);
	     seq++)
		bch2_journal_buf_to_text(out, j, seq);
	prt_printf(out, "last buf %s\n", journal_entry_is_open(j) ? "open" : "closed");
}

static inline struct journal_buf *
journal_seq_to_buf(struct journal *j, u64 seq)
{
	struct journal_buf *buf = NULL;

	EBUG_ON(seq > journal_cur_seq(j));

	if (journal_seq_unwritten(j, seq)) {
		buf = j->buf + (seq & JOURNAL_BUF_MASK);
		EBUG_ON(le64_to_cpu(buf->data->seq) != seq);
	}
	return buf;
}

static void journal_pin_list_init(struct journal_entry_pin_list *p, int count)
{
	unsigned i;

	for (i = 0; i < ARRAY_SIZE(p->list); i++)
		INIT_LIST_HEAD(&p->list[i]);
	INIT_LIST_HEAD(&p->flushed);
	atomic_set(&p->count, count);
	p->devs.nr = 0;
}

/*
 * Detect stuck journal conditions and trigger shutdown. Technically the journal
 * can end up stuck for a variety of reasons, such as a blocked I/O, journal
 * reservation lockup, etc. Since this is a fatal error with potentially
 * unpredictable characteristics, we want to be fairly conservative before we
 * decide to shut things down.
 *
 * Consider the journal stuck when it appears full with no ability to commit
 * btree transactions, to discard journal buckets, nor acquire priority
 * (reserved watermark) reservation.
 */
static inline bool
journal_error_check_stuck(struct journal *j, int error, unsigned flags)
{
	struct bch_fs *c = container_of(j, struct bch_fs, journal);
	bool stuck = false;
	struct printbuf buf = PRINTBUF;

	if (!(error == JOURNAL_ERR_journal_full ||
	      error == JOURNAL_ERR_journal_pin_full) ||
	    nr_unwritten_journal_entries(j) ||
	    (flags & BCH_WATERMARK_MASK) != BCH_WATERMARK_reclaim)
		return stuck;

	spin_lock(&j->lock);

	if (j->can_discard) {
		spin_unlock(&j->lock);
		return stuck;
	}

	stuck = true;

	/*
	 * The journal shutdown path will set ->err_seq, but do it here first to
	 * serialize against concurrent failures and avoid duplicate error
	 * reports.
	 */
	if (j->err_seq) {
		spin_unlock(&j->lock);
		return stuck;
	}
	j->err_seq = journal_cur_seq(j);
	spin_unlock(&j->lock);

	bch_err(c, "Journal stuck! Hava a pre-reservation but journal full (error %s)",
		bch2_journal_errors[error]);
	bch2_journal_debug_to_text(&buf, j);
	bch_err(c, "%s", buf.buf);

	printbuf_reset(&buf);
	bch2_journal_pins_to_text(&buf, j);
	bch_err(c, "Journal pins:\n%s", buf.buf);
	printbuf_exit(&buf);

	bch2_fatal_error(c);
	dump_stack();

	return stuck;
}

void bch2_journal_do_writes(struct journal *j)
{
	for (u64 seq = journal_last_unwritten_seq(j);
	     seq <= journal_cur_seq(j);
	     seq++) {
		unsigned idx = seq & JOURNAL_BUF_MASK;
		struct journal_buf *w = j->buf + idx;

		if (w->write_started && !w->write_allocated)
			break;
		if (w->write_started)
			continue;

		if (!journal_state_count(j->reservations, idx)) {
			w->write_started = true;
			closure_call(&w->io, bch2_journal_write, j->wq, NULL);
		}

		break;
	}
}

/*
 * Final processing when the last reference of a journal buffer has been
 * dropped. Drop the pin list reference acquired at journal entry open and write
 * the buffer, if requested.
 */
void bch2_journal_buf_put_final(struct journal *j, u64 seq)
{
	lockdep_assert_held(&j->lock);

	if (__bch2_journal_pin_put(j, seq))
		bch2_journal_reclaim_fast(j);
	bch2_journal_do_writes(j);
}

/*
 * Returns true if journal entry is now closed:
 *
 * We don't close a journal_buf until the next journal_buf is finished writing,
 * and can be opened again - this also initializes the next journal_buf:
 */
static void __journal_entry_close(struct journal *j, unsigned closed_val, bool trace)
{
	struct bch_fs *c =