nuttx/fs/mnemofs/mnemofs_journal.c
chenrun1 3f47fd767a fs/xxfs:Replace kmm with fs heap
Summary:
  1.Add configuration to allocate memory from the specified section
  2.Replace all memory operations (kmm_) in the vfs with
    fs_heap_. When FS_HEAPSIZE > 0, memory is requested for the file system by specifying a configured heap location. By default (i.e. FS_HEAPSIZE=0) fs_heap_ is equivalent to kmm_

Signed-off-by: chenrun1 <chenrun1@xiaomi.com>
2024-10-10 15:30:41 +02:00

887 lines
25 KiB
C

/****************************************************************************
* fs/mnemofs/mnemofs_journal.c
* Journal of mnemofs.
*
* Licensed to the Apache Software Foundation (ASF) under one or more
* contributor license agreements. See the NOTICE file distributed with
* this work for additional information regarding copyright ownership. The
* ASF licenses this file to you under the Apache License, Version 2.0 (the
* "License"); you may not use this file except in compliance with the
* License. You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
* License for the specific language governing permissions and limitations
* under the License.
*
* Alternatively, the contents of this file may be used under the terms of
* the BSD-3-Clause license:
*
* SPDX-License-Identifier: BSD-3-Clause
*
* Copyright (c) 2024 Saurav Pal
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. Neither the name of the author nor the names of its contributors may
* be used to endorse or promote products derived from this software
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS "AS IS" AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
****************************************************************************/
/****************************************************************************
* In mnemofs, the journal stores the path, depth and the new location of the
* CTZ file called logs, and also the location of the master block. The first
* n blocks of the journal store the logs, while the last two blocks contain
* master nodes, and the blocks are called as master blocks. The two master
* blocks are identical copies for backup.
*
* Due to LRU, and the structure of mnemofs, the first n blocks of the
* journal get filled up much faster than the master blocks, and move more.
* There will be certain point where the entire journal (the n+2 blocks)
* move, but mostly, its the first n blocks that move.
*
* The first block starts with an 8 byte magic sequence, a 2 bytes long
* number denoting number of blocks in the journal, and then follows up
* with an array containing the block numbers of all blocks in the journal
* including the first block. Then the logs start.
*
* The logs take up size in multiples of pages. There might be unitilzed
* space at the end of a log.
*
* All logs are followed by a byte-long hash of the log.
****************************************************************************/
/****************************************************************************
* Included Files
****************************************************************************/
#include <endian.h>
#include <nuttx/kmalloc.h>
#include <nuttx/list.h>
#include <sys/param.h>
#include "mnemofs.h"
#include "fs_heap.h"
/****************************************************************************
* Pre-processor Definitions
****************************************************************************/
#define MFS_JRNL_SUFFIXSZ (8 + 2) /* 8 byte magic sequence + no. of blks */
#define MFS_LOGSZ(depth) (sizeof(mfs_t) * 2 + sizeof(struct mfs_ctz_s) + \
sizeof(struct mfs_path_s) * depth + \
sizeof(struct timespec) * 3 + sizeof(uint16_t))
/****************************************************************************
* Private Types
****************************************************************************/
/* NOTE: Even if higher level functions use path as struct mfs_path_s,
* journal only uses struct mfs_ctz_s to avoid problems during write and
* read of logs. The offsets in struct mfs_path_s will be applied by the
* search methods of higher functions.
*/
/****************************************************************************
* Private Function Prototypes
****************************************************************************/
FAR static const char *deser_log(FAR const char * const in,
FAR struct mfs_jrnl_log_s * const x);
FAR static char *ser_log(FAR const struct mfs_jrnl_log_s * const x,
FAR char * const out);
/****************************************************************************
* Private Data
****************************************************************************/
/****************************************************************************
* Public Data
****************************************************************************/
/****************************************************************************
* Private Functions
****************************************************************************/
/****************************************************************************
* Name: mfs_jrnl_rdlog
*
* Description:
* Read a log to the journal from given location, and update location to
* point to next log location.
*
* Input Parameters:
* sb - Superblock instance of the device.
* blkidx - Journal Block Index of the current block.
* pg_in_blk - Page offset in the block.
* log - To populate with the log.
*
* Returned Value:
* 0 - OK
* < 0 - Error
*
* Assumptions/Limitations:
* This function does NOT care about start of the journal, or even, if
* the initial requested area is inside the journal. It will malfunction
* if not used properly. Usually this is used in an iterative manner, and
* hence the first time blkidx and pg_in_blk are initialized, they should
* be derived from the values in MFS_JRNL(sb) respectively.
*
* This updates the blkidx and pg_in_blk to point to the next log, and
* returns an -ENOSPC when end of journal is reached in traversal.
*
* Free the log after use.
*
****************************************************************************/
int mfs_jrnl_rdlog(FAR const struct mfs_sb_s *const sb,
FAR mfs_t *blkidx, FAR mfs_t *pg_in_blk,
FAR struct mfs_jrnl_log_s *log)
{
int ret = OK;
char tmp[4];
mfs_t log_sz;
mfs_t jrnl_pg;
mfs_t jrnl_blk;
FAR char *buf = NULL;
DEBUGASSERT(*pg_in_blk % MFS_PGSZ(sb) == 0);
jrnl_blk = mfs_jrnl_blkidx2blk(sb, *blkidx);
jrnl_pg = MFS_BLK2PG(sb, jrnl_blk) + *pg_in_blk;
/* First 4 bytes contain the size of the entire log. */
ret = mfs_read_page(sb, tmp, 4, jrnl_pg, 0);
if (predict_false(ret < 0))
{
goto errout;
}
mfs_deser_mfs(tmp, &log_sz);
if (log_sz == 0)
{
ret = -ENOSPC;
goto errout;
}
buf = fs_heap_zalloc(log_sz);
if (predict_false(buf == NULL))
{
ret = -ENOMEM;
goto errout;
}
ret = mfs_read_page(sb, buf, log_sz, jrnl_pg, 4);
if (predict_false(ret < 0))
{
goto errout_with_buf;
}
ret = OK;
if (predict_false(deser_log(buf, log) == NULL))
{
ret = -ENOMEM;
goto errout_with_buf;
}
(*pg_in_blk)++;
if (*pg_in_blk >= MFS_PGINBLK(sb))
{
*pg_in_blk = 0;
(*blkidx)++;
}
errout_with_buf:
fs_heap_free(buf);
errout:
return ret;
}
/****************************************************************************
* Name: ser_log
*
* Description:
* Serialize a log.
*
* Input Parameters:
* n - Log to serialize
* out - Output array where to serialize.
*
* Returned Value:
* Pointer to byte after the end of serialized value.
*
* Assumptions/Limitations:
* This assumes the out buffer has enough space to hold the inline path.
*
* This doesn't require the hash to be pre-calculated.
*
****************************************************************************/
FAR static char *ser_log(FAR const struct mfs_jrnl_log_s * const x,
FAR char * const out)
{
FAR char *o = out;
mfs_t i;
o = mfs_ser_mfs(x->depth, o);
o = mfs_ser_mfs(x->sz_new, o);
o = mfs_ser_ctz(&x->loc_new, o);
o = mfs_ser_timespec(&x->st_mtim_new, o);
o = mfs_ser_timespec(&x->st_atim_new, o);
o = mfs_ser_timespec(&x->st_ctim_new, o);
for (i = 0; i < x->depth; i++)
{
o = mfs_ser_path(&x->path[i], o);
}
o = mfs_ser_16(mfs_hash(out, o - out), o);
return o;
}
/****************************************************************************
* Name: deser_log
*
* Description:
* Deserialize a log.
*
* Input Parameters:
* in - Input array from where to deserialize.
* n - Log to deserialize
*
* Returned Value:
* NULL - Error.
* Pointer to byte after the end of serialized value.
*
* Assumptions/Limitations:
* This allocates space for the path, and the log should freed after use.
*
****************************************************************************/
FAR static const char *deser_log(FAR const char * const in,
FAR struct mfs_jrnl_log_s * const x)
{
FAR const char *i = in;
mfs_t k;
i = mfs_deser_mfs(i, &x->depth);
i = mfs_deser_mfs(i, &x->sz_new);
i = mfs_deser_ctz(i, &x->loc_new);
i = mfs_deser_timespec(i, &x->st_mtim_new);
i = mfs_deser_timespec(i, &x->st_atim_new);
i = mfs_deser_timespec(i, &x->st_ctim_new);
/* Allocates path. Deallocate using mfs_jrnl_log_free. */
x->path = fs_heap_zalloc(sizeof(struct mfs_jrnl_log_s) * x->depth);
if (predict_false(x->path == NULL))
{
return NULL;
}
for (k = 0; k < x->depth; k++)
{
i = mfs_deser_path(i, &x->path[k]);
}
i = mfs_deser_16(i, &x->hash);
return i;
}
/****************************************************************************
* Name: mfs_jrnl_log_free
*
* Description:
* Free the log.
*
* Input Parameters:
* log - Log
*
* Assumptions/Limitations:
* This allocates space for the path, and the log should freed after use.
*
****************************************************************************/
void mfs_jrnl_log_free(FAR const struct mfs_jrnl_log_s * const log)
{
fs_heap_free(log->path);
}
/****************************************************************************
* Public Functions
****************************************************************************/
int mfs_jrnl_init(FAR struct mfs_sb_s * const sb, mfs_t blk)
{
char buftmp[2];
int ret = OK;
mfs_t sz;
mfs_t blkidx;
mfs_t pg_in_blk;
struct mfs_jrnl_log_s log;
/* Magic sequence is already used to find the block, so not required. */
mfs_read_page(sb, buftmp, 2, MFS_BLK2PG(sb, blk), 8);
mfs_deser_16(buftmp, &MFS_JRNL(sb).n_blks);
if (MFS_JRNL(sb).n_blks == 0)
{
ret = -EINVAL;
goto errout;
}
sz = MFS_JRNL_SUFFIXSZ + ((CONFIG_MNEMOFS_JOURNAL_NBLKS + 2) * 4);
MFS_JRNL(sb).jrnl_start = blk;
MFS_JRNL(sb).log_cpg = (sz + (MFS_PGSZ(sb) - 1)) / MFS_PGSZ(sb);
MFS_JRNL(sb).log_spg = MFS_JRNL(sb).log_cpg;
MFS_JRNL(sb).log_cblkidx = 0;
MFS_JRNL(sb).log_sblkidx = MFS_JRNL(sb).log_cblkidx;
MFS_JRNL(sb).jrnlarr_pg = MFS_BLK2PG(sb, blk); /* Assuming pgsz > 10 */
MFS_JRNL(sb).jrnlarr_pgoff = MFS_JRNL_SUFFIXSZ;
/* Number of logs */
MFS_JRNL(sb).n_logs = 0;
blkidx = MFS_JRNL(sb).log_sblkidx;
pg_in_blk = MFS_JRNL(sb).log_spg % MFS_PGINBLK(sb);
while (true)
{
ret = mfs_jrnl_rdlog(sb, &blkidx, &pg_in_blk, &log);
if (predict_false(ret < 0 && ret != -ENOSPC))
{
goto errout;
}
else if (ret == -ENOSPC)
{
ret = OK;
break;
}
/* Assumes checking the depth is enough to check if it's empty, as
* theoretically there are no blocks with depth 0, as root has a
* depth of 1.
*/
if (log.depth == 0)
{
DEBUGASSERT(log.path == NULL);
break;
}
MFS_JRNL(sb).n_logs++;
mfs_jrnl_log_free(&log);
}
/* Master node */
MFS_JRNL(sb).mblk1 = mfs_jrnl_blkidx2blk(sb, MFS_JRNL(sb).n_blks);
MFS_JRNL(sb).mblk2 = mfs_jrnl_blkidx2blk(sb, MFS_JRNL(sb).n_blks + 1);
/* TODO: Read all pages in master blocks to find the last master node
* update.
*/
errout:
return ret;
}
int mfs_jrnl_fmt(FAR struct mfs_sb_s * const sb, FAR mfs_t *blk1,
FAR mfs_t *blk2, FAR mfs_t *jrnl_blk)
{
int i;
int ret = OK;
mfs_t sz;
mfs_t pg;
mfs_t blk;
mfs_t alloc_blk;
FAR char *tmp;
FAR char *buf = NULL;
/* TODO: Replace jrnl_blk with MFS_JRNL(sb).jrnl_start if possible. */
/* Write magic sequence, size of jrnlarr, and then the jrnlarr. */
sz = MFS_JRNL_SUFFIXSZ + ((CONFIG_MNEMOFS_JOURNAL_NBLKS + 2) * 4);
buf = fs_heap_zalloc(sz);
if (predict_false(buf == NULL))
{
ret = -ENOMEM;
goto errout;
}
if (*blk1 == 0 && *blk2 == 0)
{
*blk1 = mfs_ba_getblk(sb);
if (predict_false(blk1 == 0))
{
ret = -ENOSPC;
goto errout_with_buf;
}
finfo("Allocated Master Block 1: %d.", *blk1);
*blk2 = mfs_ba_getblk(sb);
if (predict_false(blk2 == 0))
{
ret = -ENOSPC;
goto errout_with_buf;
}
finfo("Allocated Master Block 1: %d.", *blk2);
finfo("New locations for Master Blocks %d & %d.", *blk1, *blk2);
}
tmp = buf;
tmp = mfs_ser_str(MFS_JRNL_MAGIC, 8, tmp);
tmp = mfs_ser_16(CONFIG_MNEMOFS_JOURNAL_NBLKS, tmp);
for (i = 0; i < CONFIG_MNEMOFS_JOURNAL_NBLKS; i++)
{
alloc_blk = mfs_ba_getblk(sb);
tmp = mfs_ser_mfs(alloc_blk, tmp);
if (predict_false(i == 0))
{
blk = alloc_blk;
*jrnl_blk = alloc_blk;
MFS_JRNL(sb).jrnl_start = alloc_blk;
}
finfo("Allocated Journal Block %d at Block %d.", i, alloc_blk);
}
tmp = mfs_ser_mfs(*blk1, tmp);
tmp = mfs_ser_mfs(*blk2, tmp);
finfo("All Journal Blocks allocated.");
pg = MFS_BLK2PG(sb, blk);
ret = mfs_write_page(sb, buf, sz, pg, 0); /* Assuming array fits in a
* single page.
*/
if (predict_false(ret < 0))
{
goto errout_with_buf;
}
ret = OK; /* We reach here, we OK. */
finfo("Written magic sequence, size and journal array into the journal.");
MFS_JRNL(sb).n_logs = 0;
MFS_JRNL(sb).n_blks = CONFIG_MNEMOFS_JOURNAL_NBLKS;
MFS_JRNL(sb).log_cpg = pg + 1; /* Assumes 1 page for jrnl_arr. */
MFS_JRNL(sb).log_cblkidx = 0;
MFS_JRNL(sb).log_spg = MFS_JRNL(sb).log_cpg;
MFS_JRNL(sb).log_sblkidx = MFS_JRNL(sb).log_cblkidx;
MFS_JRNL(sb).jrnlarr_pg = MFS_BLK2PG(sb, blk);
MFS_JRNL(sb).jrnlarr_pgoff = MFS_JRNL_SUFFIXSZ;
MFS_JRNL(sb).mblk1 = *blk1;
MFS_JRNL(sb).mblk2 = *blk2;
errout_with_buf:
fs_heap_free(buf);
errout:
return ret;
}
void mfs_jrnl_free(FAR struct mfs_sb_s * const sb)
{
if (!mfs_jrnl_isempty(sb) &&
MFS_JRNL(sb).log_cblkidx >= MFS_JRNL_LIM(sb))
{
mfs_jrnl_flush(sb);
}
finfo("Journal Freed.");
}
mfs_t mfs_jrnl_blkidx2blk(FAR const struct mfs_sb_s * const sb,
const mfs_t blk_idx)
{
int ret = OK;
mfs_t pg;
mfs_t idx;
mfs_t blk;
mfs_t pgoff;
char buf[4];
pg = MFS_JRNL(sb).jrnlarr_pg;
pgoff = MFS_JRNL(sb).jrnlarr_pgoff;
blk = MFS_PG2BLK(sb, pg);
pgoff += blk_idx * 4;
if (pgoff > MFS_PGSZ(sb))
{
pg += pgoff / MFS_PGSZ(sb);
pgoff %= MFS_PGSZ(sb);
}
/* No pg overflow. The blocks have to be big enough. */
DEBUGASSERT(pg < (MFS_BLK2PG(sb, blk) + MFS_PGINBLK(sb)));
ret = mfs_read_page(sb, buf, 4, pg, pgoff);
if (predict_false(ret < 0))
{
return 0;
}
mfs_deser_mfs(buf, &idx);
/* FUTURE TODO: Make it such that the entire jrnlarr doesn't need to be in
* a single block.
*/
return idx;
}
int mfs_jrnl_updatedinfo(FAR const struct mfs_sb_s * const sb,
FAR struct mfs_path_s * const path,
const mfs_t depth)
{
int ret = OK;
mfs_t blkidx;
mfs_t counter = 0;
mfs_t pg_in_block;
struct mfs_jrnl_log_s tmplog;
/* TODO: Allow optional filling of updated timestamps, etc. */
DEBUGASSERT(depth > 0);
blkidx = MFS_JRNL(sb).log_sblkidx;
pg_in_block = MFS_JRNL(sb).log_spg % MFS_PGINBLK(sb);
while (blkidx < MFS_JRNL(sb).n_blks && counter < MFS_JRNL(sb).n_logs)
{
ret = mfs_jrnl_rdlog(sb, &blkidx, &pg_in_block, &tmplog);
if (predict_false(ret < 0 && ret != -ENOSPC))
{
goto errout;
}
else if (ret == -ENOSPC)
{
break;
}
DEBUGASSERT(tmplog.depth > 0);
if (tmplog.depth > depth)
{
/* Not suitable. */
}
else
{
DEBUGASSERT(tmplog.depth > 0);
if (mfs_path_eq(&tmplog.path[tmplog.depth - 1],
&path[tmplog.depth - 1]))
{
path[tmplog.depth - 1].ctz = tmplog.loc_new;
path[tmplog.depth - 1].sz = tmplog.sz_new;
}
}
mfs_jrnl_log_free(&tmplog);
counter++;
}
errout:
return ret;
}
int mfs_jrnl_wrlog(FAR struct mfs_sb_s * const sb,
FAR const struct mfs_node_s *node,
const struct mfs_ctz_s loc_new, const mfs_t sz_new)
{
int ret = OK;
mfs_t jrnl_pg;
FAR char *buf = NULL;
FAR char *tmp = NULL;
const mfs_t log_sz = sizeof(mfs_t) + MFS_LOGSZ(node->depth);
struct mfs_jrnl_log_s log;
buf = fs_heap_zalloc(log_sz); /* For size before log. */
if (predict_false(buf == NULL))
{
ret = -ENOMEM;
goto errout;
}
/* Serialize */
log.depth = node->depth;
log.sz_new = sz_new;
log.loc_new = loc_new;
log.st_mtim_new = node->st_mtim;
log.st_atim_new = node->st_atim;
log.st_ctim_new = node->st_ctim;
log.path = node->path; /* Fine as temporarily usage. */
tmp = buf;
tmp = mfs_ser_mfs(log_sz - sizeof(mfs_t), tmp); /* First 4 bytes have sz */
tmp = ser_log(&log, tmp);
/* Store */
jrnl_pg = MFS_JRNL(sb).log_cpg;
/* TODO: It assumes it takes only one page per log. */
ret = mfs_write_page(sb, buf, log_sz, jrnl_pg, 0);
if (predict_false(ret < 0))
{
goto errout_with_buf;
}
ret = OK;
jrnl_pg++;
if (jrnl_pg % MFS_PGINBLK(sb) == 0)
{
MFS_JRNL(sb).log_cblkidx++;
}
MFS_JRNL(sb).log_cpg = jrnl_pg;
MFS_JRNL(sb).n_logs++;
errout_with_buf:
fs_heap_free(buf);
errout:
return ret;
}
int mfs_jrnl_flush(FAR struct mfs_sb_s * const sb)
{
/* When a file or a directory is deleted.
*
* It will be modified to an entry in the LRU which details the deletion
* of all bytes from the child... as in, offset 0, deleted bytes is the
* size of the file.
*
* The new "location" can be used as (0, 0) to signify a deletion, even in
* its journal log.
*
* Also ensure if the size gets updated to 0.
*
* Then the flush operation problem will be solved for removal of files or
* directories.
*
* Move operation will not empty the child, but only the parent from the
* old parent.
*/
/* Time complexity is going to be horrendous. Hint: O(n^2). HOWEVER, as
* littlefs points out....if n is constant, it's essentially a O(k), or
* O(1) :D
*/
/* TODO: Need to consider how the LRU and Journal interact with each other
* for newly created fs object's entries.
*/
/* We're using updatectz to update the LRU inside the journal. Think
* about how that might affect the iteration attempts.
*/
int ret = OK;
mfs_t blkidx = MFS_JRNL(sb).log_sblkidx;
mfs_t log_itr = 0;
mfs_t pg_in_blk = MFS_JRNL(sb).log_spg \
% MFS_PGINBLK(sb);
mfs_t tmp_blkidx;
mfs_t tmp_pg_in_blk;
mfs_t mn_blk1;
mfs_t mn_blk2;
mfs_t i;
mfs_t jrnl_blk;
mfs_t blk;
struct mfs_jrnl_log_s log;
struct mfs_jrnl_log_s tmp_log;
FAR struct mfs_path_s *path = NULL;
struct mfs_jrnl_state_s j_state;
struct mfs_mn_s mn_state;
while (log_itr < MFS_JRNL(sb).n_logs)
{
ret = mfs_jrnl_rdlog(sb, &blkidx, &pg_in_blk, &log);
if (predict_false(ret < 0))
{
DEBUGASSERT(ret != -ENOSPC); /* While condition is sufficient. */
goto errout;
}
if (log.loc_new.pg_e == 0 && log.loc_new.idx_e == 0)
{
/* Entry is deleted, do not bother with it. */
break;
}
tmp_blkidx = blkidx;
tmp_pg_in_blk = pg_in_blk;
path = fs_heap_zalloc(log.depth * sizeof(struct mfs_path_s));
if (predict_false(path == NULL))
{
goto errout;
}
memcpy(path, log.path, log.depth * sizeof(struct mfs_path_s));
path[log.depth - 1].ctz = log.loc_new;
for (; ; )
{
ret = mfs_jrnl_rdlog(sb, &tmp_blkidx, &tmp_pg_in_blk, &tmp_log);
if (ret == -ENOSPC)
{
break;
}
else if (predict_false(ret < 0))
{
mfs_jrnl_log_free(&log);
mfs_free_patharr(path);
goto errout;
}
if (tmp_log.depth > log.depth)
{
mfs_jrnl_log_free(&tmp_log);
continue;
}
if (!mfs_path_eq(&path[tmp_log.depth - 1],
&tmp_log.path[tmp_log.depth - 1]))
{
mfs_jrnl_log_free(&tmp_log);
continue;
}
path[tmp_log.depth - 1] = tmp_log.path[tmp_log.depth - 1];
if (tmp_log.loc_new.pg_e == 0 && tmp_log.loc_new.idx_e == 0)
{
/* Entry is deleted, do not bother with it. */
break;
}
}
if (log.depth == 1)
{
MFS_MN(sb).root_ctz = path[log.depth - 1].ctz;
MFS_MN(sb).root_sz = path[log.depth - 1].sz;
/* TODO: Other parameters. */
}
else
{
ret = mfs_lru_updatectz(sb, path, log.depth,
path[log.depth - 1].ctz,
path[log.depth - 1].sz);
if (predict_false(ret < 0))
{
mfs_free_patharr(path);
mfs_jrnl_log_free(&log);
goto errout;
}
}
mfs_free_patharr(path);
mfs_jrnl_log_free(&log);
}
if (MFS_MN(sb).mblk_idx == MFS_PGINBLK(sb))
{
mn_blk1 = 0;
mn_blk2 = 0;
}
else
{
/* FUTURE TODO: Save the two block numbers in master node structure to
* be faster.
*/
mn_blk1 = mfs_jrnl_blkidx2blk(sb, MFS_JRNL(sb).n_blks);
mn_blk2 = mfs_jrnl_blkidx2blk(sb, MFS_JRNL(sb).n_blks + 1);
}
/* Reallocate journal. */
j_state = MFS_JRNL(sb);
mn_state = MFS_MN(sb);
ret = mfs_jrnl_fmt(sb, &mn_blk1, &mn_blk2, &jrnl_blk);
if (predict_false(ret < 0))
{
MFS_JRNL(sb) = j_state;
goto errout;
}
/* Write master node entry. */
ret = mfs_mn_sync(sb, &path[0], mn_blk1, mn_blk2, jrnl_blk);
if (predict_false(ret < 0))
{
MFS_MN(sb) = mn_state;
goto errout;
}
/* Mark all old blocks of journal (and master blocks) as deletable. */
for (i = 0; i < MFS_JRNL(sb).n_blks + 2; i++)
{
blk = mfs_jrnl_blkidx2blk(sb, i);
mfs_ba_blkmarkdel(sb, blk);
}
/* Delete outdated blocks. */
ret = mfs_ba_delmarked(sb);
if (predict_false(ret < 0))
{
goto errout;
}
errout:
return ret;
}
bool mfs_jrnl_isempty(FAR const struct mfs_sb_s * const sb)
{
return MFS_JRNL(sb).n_logs == 0;
}