static int llog_read_header(const struct lu_env *env,
struct llog_handle *handle,
struct obd_uuid *uuid)
{
struct llog_operations *lop;
int rc;
rc = llog_handle2ops(handle, &lop);
if (rc)
return rc;
if (lop->lop_read_header == NULL)
return -EOPNOTSUPP;
rc = lop->lop_read_header(env, handle);
if (rc == LLOG_EEMPTY) {
struct llog_log_hdr *llh = handle->lgh_hdr;
handle->lgh_last_idx = 0; /* header is record with index 0 */
llh->llh_count = 1; /* for the header record */
llh->llh_hdr.lrh_type = LLOG_HDR_MAGIC;
llh->llh_hdr.lrh_len = llh->llh_tail.lrt_len = LLOG_CHUNK_SIZE;
llh->llh_hdr.lrh_index = llh->llh_tail.lrt_index = 0;
llh->llh_timestamp = cfs_time_current_sec();
if (uuid)
memcpy(&llh->llh_tgtuuid, uuid,
sizeof(llh->llh_tgtuuid));
llh->llh_bitmap_offset = offsetof(typeof(*llh), llh_bitmap);
ext2_set_bit(0, llh->llh_bitmap);
rc = 0;
}
return rc;
}
/*
* bitmap_file_set_bit -- called before performing a write to the md device
* to set (and eventually sync) a particular bit in the bitmap file
*
* we set the bit immediately, then we record the page number so that
* when an unplug occurs, we can flush the dirty pages out to disk
*/
static void bitmap_file_set_bit(struct bitmap *bitmap, sector_t block)
{
unsigned long bit;
struct page *page;
void *kaddr;
unsigned long chunk = block >> CHUNK_BLOCK_SHIFT(bitmap);
if (!bitmap->filemap) {
return;
}
page = filemap_get_page(bitmap, chunk);
if (!page) return;
bit = file_page_offset(chunk);
/* set the bit */
kaddr = kmap_atomic(page, KM_USER0);
if (bitmap->flags & BITMAP_HOSTENDIAN)
set_bit(bit, kaddr);
else
ext2_set_bit(bit, kaddr);
kunmap_atomic(kaddr, KM_USER0);
PRINTK("set file bit %lu page %lu\n", bit, page->index);
/* record page number so it gets flushed to disk when unplug occurs */
set_page_attr(bitmap, page, BITMAP_PAGE_DIRTY);
}
/* alloc a new block */
int ext2_alloc_block ( unsigned int goal)
{
unsigned int block;
unsigned int block_group;
unsigned int bit;
struct ext2_group_desc *desc;
struct ext2_sb_info * sbi = EXT2_SBI();
void *bitmap;
block_group = ext2_get_group_num (goal, BLOCK);
bit = ext2_get_group_offset (goal, BLOCK);
bitmap = ext2_read_block_bitmap (block_group);
block = ext2_grab_block (bitmap, bit);
if ( !block)
ext2_error ("no free blocks any more");
desc = ext2_get_group_desc (block_group);
desc->bg_free_blocks_count --;
sbi->s_free_blocks_count --;
ext2_set_bit (bitmap, block);
return block;
}
/* returns negative on error; 0 if success; 1 if success & log destroyed */
int mlog_cancel_rec(struct mlog_handle *loghandle, int index)
{
struct mlog_log_hdr *mlh = loghandle->mgh_hdr;
int ret = 0;
MENTRY();
MDEBUG("Canceling %d in log %llx\n",
index, loghandle->mgh_id.mgl_oid);
if (index == 0) {
MERROR("Can't cancel index 0 which is header\n");
ret = -EINVAL;
goto out;
}
if (!ext2_clear_bit(index, mlh->mlh_bitmap)) {
MDEBUG("Catalog index %u already clear?\n", index);
ret = -ENOENT;
goto out;
}
mlh->mlh_count--;
if ((mlh->mlh_flags & MLOG_F_ZAP_WHEN_EMPTY) &&
(mlh->mlh_count == 1) &&
(loghandle->mgh_last_idx == (MLOG_BITMAP_BYTES * 8) - 1)) {
ret = mlog_destroy(loghandle);
if (ret) {
MERROR("Failure destroying log after last cancel: %d\n",
ret);
ext2_set_bit(index, mlh->mlh_bitmap);
mlh->mlh_count++;
} else {
ret = 1;
}
goto out;
}
ret = mlog_write_rec(loghandle, &mlh->mlh_hdr, NULL, 0, NULL, 0);
if (ret) {
MERROR("Failure re-writing header %d\n", ret);
ext2_set_bit(index, mlh->mlh_bitmap);
mlh->mlh_count++;
}
out:
MRETURN(ret);
}
/* returns negative on error; 0 if success; 1 if success & log destroyed */
int llog_cancel_rec(const struct lu_env *env, struct llog_handle *loghandle,
int index)
{
struct llog_log_hdr *llh = loghandle->lgh_hdr;
int rc = 0;
ENTRY;
CDEBUG(D_RPCTRACE, "Canceling %d in log "DOSTID"\n",
index, POSTID(&loghandle->lgh_id.lgl_oi));
if (index == 0) {
CERROR("Can't cancel index 0 which is header\n");
RETURN(-EINVAL);
}
spin_lock(&loghandle->lgh_hdr_lock);
if (!ext2_clear_bit(index, llh->llh_bitmap)) {
spin_unlock(&loghandle->lgh_hdr_lock);
CDEBUG(D_RPCTRACE, "Catalog index %u already clear?\n", index);
RETURN(-ENOENT);
}
llh->llh_count--;
if ((llh->llh_flags & LLOG_F_ZAP_WHEN_EMPTY) &&
(llh->llh_count == 1) &&
(loghandle->lgh_last_idx == (LLOG_BITMAP_BYTES * 8) - 1)) {
spin_unlock(&loghandle->lgh_hdr_lock);
rc = llog_destroy(env, loghandle);
if (rc < 0) {
CERROR("%s: can't destroy empty llog #"DOSTID
"#%08x: rc = %d\n",
loghandle->lgh_ctxt->loc_obd->obd_name,
POSTID(&loghandle->lgh_id.lgl_oi),
loghandle->lgh_id.lgl_ogen, rc);
GOTO(out_err, rc);
}
RETURN(1);
}
spin_unlock(&loghandle->lgh_hdr_lock);
rc = llog_write(env, loghandle, &llh->llh_hdr, NULL, 0, NULL, 0);
if (rc < 0) {
CERROR("%s: fail to write header for llog #"DOSTID
"#%08x: rc = %d\n",
loghandle->lgh_ctxt->loc_obd->obd_name,
POSTID(&loghandle->lgh_id.lgl_oi),
loghandle->lgh_id.lgl_ogen, rc);
GOTO(out_err, rc);
}
RETURN(0);
out_err:
spin_lock(&loghandle->lgh_hdr_lock);
ext2_set_bit(index, llh->llh_bitmap);
llh->llh_count++;
spin_unlock(&loghandle->lgh_hdr_lock);
return rc;
}
int llog_read_header(const struct lu_env *env, struct llog_handle *handle,
const struct obd_uuid *uuid)
{
struct llog_operations *lop;
int rc;
ENTRY;
rc = llog_handle2ops(handle, &lop);
if (rc)
RETURN(rc);
if (lop->lop_read_header == NULL)
RETURN(-EOPNOTSUPP);
rc = lop->lop_read_header(env, handle);
if (rc == LLOG_EEMPTY) {
struct llog_log_hdr *llh = handle->lgh_hdr;
/* lrh_len should be initialized in llog_init_handle */
handle->lgh_last_idx = 0; /* header is record with index 0 */
handle->lgh_write_offset = 0;
llh->llh_count = 1; /* for the header record */
llh->llh_hdr.lrh_type = LLOG_HDR_MAGIC;
LASSERT(handle->lgh_ctxt->loc_chunk_size >=
LLOG_MIN_CHUNK_SIZE);
llh->llh_hdr.lrh_len = handle->lgh_ctxt->loc_chunk_size;
llh->llh_hdr.lrh_index = 0;
llh->llh_timestamp = cfs_time_current_sec();
if (uuid)
memcpy(&llh->llh_tgtuuid, uuid,
sizeof(llh->llh_tgtuuid));
llh->llh_bitmap_offset = offsetof(typeof(*llh), llh_bitmap);
/* Since update llog header might also call this function,
* let's reset the bitmap to 0 here */
memset(LLOG_HDR_BITMAP(llh), 0, llh->llh_hdr.lrh_len -
llh->llh_bitmap_offset -
sizeof(llh->llh_tail));
ext2_set_bit(0, LLOG_HDR_BITMAP(llh));
LLOG_HDR_TAIL(llh)->lrt_len = llh->llh_hdr.lrh_len;
LLOG_HDR_TAIL(llh)->lrt_index = llh->llh_hdr.lrh_index;
rc = 0;
}
RETURN(rc);
}
static __inline__ void
store_block (unsigned block, uchar * src, unsigned len)
{
ulong offset = block * TftpBlkSize + TftpBlockWrapOffset;
ulong newsize = offset + len;
#ifdef CONFIG_SYS_DIRECT_FLASH_TFTP
int i, rc = 0;
for (i=0; i<CONFIG_SYS_MAX_FLASH_BANKS; i++) {
/* start address in flash? */
if (flash_info[i].flash_id == FLASH_UNKNOWN)
continue;
if (load_addr + offset >= flash_info[i].start[0]) {
rc = 1;
break;
}
}
if (rc) { /* Flash is destination for this packet */
rc = flash_write ((char *)src, (ulong)(load_addr+offset), len);
if (rc) {
flash_perror (rc);
NetState = NETLOOP_FAIL;
return;
}
}
else
#endif /* CONFIG_SYS_DIRECT_FLASH_TFTP */
{
(void)memcpy((void *)(load_addr + offset), src, len);
}
#ifdef CONFIG_MCAST_TFTP
if (Multicast)
ext2_set_bit(block, Bitmap);
#endif
if (NetBootFileXferSize < newsize)
NetBootFileXferSize = newsize;
}
struct inode * ext2_new_inode (const struct inode * dir, int mode)
{
struct super_block * sb;
struct buffer_head * bh;
struct buffer_head * bh2;
int group, i;
ino_t ino;
struct inode * inode;
struct ext2_group_desc * desc;
struct ext2_super_block * es;
int err;
sb = dir->i_sb;
inode = new_inode(sb);
if (!inode)
return ERR_PTR(-ENOMEM);
lock_super (sb);
es = sb->u.ext2_sb.s_es;
repeat:
if (S_ISDIR(mode))
group = find_group_dir(sb, dir->u.ext2_i.i_block_group);
else
group = find_group_other(sb, dir->u.ext2_i.i_block_group);
err = -ENOSPC;
if (group == -1)
goto fail;
err = -EIO;
bh = load_inode_bitmap (sb, group);
if (IS_ERR(bh))
goto fail2;
i = ext2_find_first_zero_bit ((unsigned long *) bh->b_data,
EXT2_INODES_PER_GROUP(sb));
if (i >= EXT2_INODES_PER_GROUP(sb))
goto bad_count;
ext2_set_bit (i, bh->b_data);
mark_buffer_dirty(bh);
if (sb->s_flags & MS_SYNCHRONOUS) {
ll_rw_block (WRITE, 1, &bh);
wait_on_buffer (bh);
}
ino = group * EXT2_INODES_PER_GROUP(sb) + i + 1;
if (ino < EXT2_FIRST_INO(sb) || ino > le32_to_cpu(es->s_inodes_count)) {
ext2_error (sb, "ext2_new_inode",
"reserved inode or inode > inodes count - "
"block_group = %d,inode=%ld", group, ino);
err = -EIO;
goto fail2;
}
es->s_free_inodes_count =
cpu_to_le32(le32_to_cpu(es->s_free_inodes_count) - 1);
mark_buffer_dirty(sb->u.ext2_sb.s_sbh);
sb->s_dirt = 1;
inode->i_uid = current->fsuid;
if (test_opt (sb, GRPID))
inode->i_gid = dir->i_gid;
else if (dir->i_mode & S_ISGID) {
inode->i_gid = dir->i_gid;
if (S_ISDIR(mode))
mode |= S_ISGID;
} else
inode->i_gid = current->fsgid;
inode->i_mode = mode;
inode->i_ino = ino;
inode->i_blksize = PAGE_SIZE; /* This is the optimal IO size (for stat), not the fs block size */
inode->i_blocks = 0;
inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
inode->u.ext2_i.i_new_inode = 1;
inode->u.ext2_i.i_flags = dir->u.ext2_i.i_flags & ~EXT2_BTREE_FL;
if (S_ISLNK(mode))
inode->u.ext2_i.i_flags &= ~(EXT2_IMMUTABLE_FL|EXT2_APPEND_FL);
inode->u.ext2_i.i_block_group = group;
if (inode->u.ext2_i.i_flags & EXT2_SYNC_FL)
inode->i_flags |= S_SYNC;
insert_inode_hash(inode);
inode->i_generation = event++;
mark_inode_dirty(inode);
unlock_super (sb);
if(DQUOT_ALLOC_INODE(inode)) {
DQUOT_DROP(inode);
inode->i_flags |= S_NOQUOTA;
inode->i_nlink = 0;
iput(inode);
return ERR_PTR(-EDQUOT);
}
ext2_debug ("allocating inode %lu\n", inode->i_ino);
return inode;
fail2:
desc = ext2_get_group_desc (sb, group, &bh2);
desc->bg_free_inodes_count =
cpu_to_le16(le16_to_cpu(desc->bg_free_inodes_count) + 1);
if (S_ISDIR(mode))
desc->bg_used_dirs_count =
cpu_to_le16(le16_to_cpu(desc->bg_used_dirs_count) - 1);
mark_buffer_dirty(bh2);
fail:
unlock_super(sb);
make_bad_inode(inode);
iput(inode);
return ERR_PTR(err);
bad_count:
ext2_error (sb, "ext2_new_inode",
"Free inodes count corrupted in group %d",
group);
/* Is it really ENOSPC? */
err = -ENOSPC;
if (sb->s_flags & MS_RDONLY)
goto fail;
desc = ext2_get_group_desc (sb, group, &bh2);
desc->bg_free_inodes_count = 0;
mark_buffer_dirty(bh2);
goto repeat;
}
/**
* Implementation of the llog_operations::lop_write
*
* This function writes the new record in the llog or modify the existed one.
*
* \param[in] env execution environment
* \param[in] loghandle llog handle of the current llog
* \param[in] rec llog record header. This is a real header of
* the full llog record to write. This is
* the beginning of buffer to write, the length
* of buffer is stored in \a rec::lrh_len
* \param[out] reccookie pointer to the cookie to return back if needed.
* It is used for further cancel of this llog
* record.
* \param[in] idx index of the llog record. If \a idx == -1 then
* this is append case, otherwise \a idx is
* the index of record to modify
* \param[in] th current transaction handle
*
* \retval 0 on successful write && \a reccookie == NULL
* 1 on successful write && \a reccookie != NULL
* \retval negative error if write failed
*/
static int llog_osd_write_rec(const struct lu_env *env,
struct llog_handle *loghandle,
struct llog_rec_hdr *rec,
struct llog_cookie *reccookie,
int idx, struct thandle *th)
{
struct llog_thread_info *lgi = llog_info(env);
struct llog_log_hdr *llh;
int reclen = rec->lrh_len;
int index, rc;
struct llog_rec_tail *lrt;
struct dt_object *o;
size_t left;
bool header_is_updated = false;
ENTRY;
LASSERT(env);
llh = loghandle->lgh_hdr;
LASSERT(llh);
o = loghandle->lgh_obj;
LASSERT(o);
LASSERT(th);
CDEBUG(D_OTHER, "new record %x to "DFID"\n",
rec->lrh_type, PFID(lu_object_fid(&o->do_lu)));
/* record length should not bigger than LLOG_CHUNK_SIZE */
if (reclen > LLOG_CHUNK_SIZE)
RETURN(-E2BIG);
rc = dt_attr_get(env, o, &lgi->lgi_attr, NULL);
if (rc)
RETURN(rc);
/**
* The modification case.
* If idx set then the record with that index must be modified.
* There are three cases possible:
* 1) the common case is the llog header update (idx == 0)
* 2) the llog record modification during llog process.
* This is indicated by the \a loghandle::lgh_cur_idx > 0.
* In that case the \a loghandle::lgh_cur_offset
* 3) otherwise this is assumed that llog consist of records of
* fixed size, i.e. catalog. The llog header must has llh_size
* field equal to record size. The record offset is calculated
* just by /a idx value
*
* During modification we don't need extra header update because
* the bitmap and record count are not changed. The record header
* and tail remains the same too.
*/
if (idx != LLOG_NEXT_IDX) {
/* llog can be empty only when first record is being written */
LASSERT(ergo(idx > 0, lgi->lgi_attr.la_size > 0));
if (!ext2_test_bit(idx, llh->llh_bitmap)) {
CERROR("%s: modify unset record %u\n",
o->do_lu.lo_dev->ld_obd->obd_name, idx);
RETURN(-ENOENT);
}
if (idx != rec->lrh_index) {
CERROR("%s: modify index mismatch %d %u\n",
o->do_lu.lo_dev->ld_obd->obd_name, idx,
rec->lrh_index);
RETURN(-EFAULT);
}
if (idx == LLOG_HEADER_IDX) {
/* llog header update */
LASSERT(reclen == sizeof(struct llog_log_hdr));
LASSERT(rec == &llh->llh_hdr);
lgi->lgi_off = 0;
lgi->lgi_buf.lb_len = reclen;
lgi->lgi_buf.lb_buf = rec;
rc = dt_record_write(env, o, &lgi->lgi_buf,
&lgi->lgi_off, th);
RETURN(rc);
} else if (loghandle->lgh_cur_idx > 0) {
/**
* The lgh_cur_offset can be used only if index is
* the same.
*/
if (idx != loghandle->lgh_cur_idx) {
CERROR("%s: modify index mismatch %d %d\n",
o->do_lu.lo_dev->ld_obd->obd_name, idx,
loghandle->lgh_cur_idx);
RETURN(-EFAULT);
}
lgi->lgi_off = loghandle->lgh_cur_offset;
CDEBUG(D_OTHER, "modify record "DOSTID": idx:%d, "
"len:%u offset %llu\n",
POSTID(&loghandle->lgh_id.lgl_oi), idx,
rec->lrh_len, (long long)lgi->lgi_off);
} else if (llh->llh_size > 0) {
if (llh->llh_size != rec->lrh_len) {
CERROR("%s: wrong record size, llh_size is %u"
" but record size is %u\n",
o->do_lu.lo_dev->ld_obd->obd_name,
llh->llh_size, rec->lrh_len);
RETURN(-EINVAL);
}
lgi->lgi_off = sizeof(*llh) + (idx - 1) * reclen;
} else {
/* This can be result of lgh_cur_idx is not set during
* llog processing or llh_size is not set to proper
* record size for fixed records llog. Therefore it is
* impossible to get record offset. */
CERROR("%s: can't get record offset, idx:%d, "
"len:%u.\n", o->do_lu.lo_dev->ld_obd->obd_name,
idx, rec->lrh_len);
RETURN(-EFAULT);
}
/* update only data, header and tail remain the same */
lgi->lgi_off += sizeof(struct llog_rec_hdr);
lgi->lgi_buf.lb_len = REC_DATA_LEN(rec);
lgi->lgi_buf.lb_buf = REC_DATA(rec);
rc = dt_record_write(env, o, &lgi->lgi_buf, &lgi->lgi_off, th);
if (rc == 0 && reccookie) {
reccookie->lgc_lgl = loghandle->lgh_id;
reccookie->lgc_index = idx;
rc = 1;
}
RETURN(rc);
}
/**
* The append case.
* The most common case of using llog. The new index is assigned to
* the new record, new bit is set in llog bitmap and llog count is
* incremented.
*
* Make sure that records don't cross a chunk boundary, so we can
* process them page-at-a-time if needed. If it will cross a chunk
* boundary, write in a fake (but referenced) entry to pad the chunk.
*/
LASSERT(lgi->lgi_attr.la_valid & LA_SIZE);
lgi->lgi_off = lgi->lgi_attr.la_size;
left = LLOG_CHUNK_SIZE - (lgi->lgi_off & (LLOG_CHUNK_SIZE - 1));
/* NOTE: padding is a record, but no bit is set */
if (left != 0 && left != reclen &&
left < (reclen + LLOG_MIN_REC_SIZE)) {
index = loghandle->lgh_last_idx + 1;
rc = llog_osd_pad(env, o, &lgi->lgi_off, left, index, th);
if (rc)
RETURN(rc);
loghandle->lgh_last_idx++; /* for pad rec */
}
/* if it's the last idx in log file, then return -ENOSPC */
if (loghandle->lgh_last_idx >= LLOG_BITMAP_SIZE(llh) - 1)
RETURN(-ENOSPC);
/* increment the last_idx along with llh_tail index, they should
* be equal for a llog lifetime */
loghandle->lgh_last_idx++;
index = loghandle->lgh_last_idx;
llh->llh_tail.lrt_index = index;
/**
* NB: the caller should make sure only 1 process access
* the lgh_last_idx, e.g. append should be exclusive.
* Otherwise it might hit the assert.
*/
LASSERT(index < LLOG_BITMAP_SIZE(llh));
rec->lrh_index = index;
lrt = rec_tail(rec);
lrt->lrt_len = rec->lrh_len;
lrt->lrt_index = rec->lrh_index;
/* the lgh_hdr_lock protects llog header data from concurrent
* update/cancel, the llh_count and llh_bitmap are protected */
spin_lock(&loghandle->lgh_hdr_lock);
if (ext2_set_bit(index, llh->llh_bitmap)) {
CERROR("%s: index %u already set in log bitmap\n",
o->do_lu.lo_dev->ld_obd->obd_name, index);
spin_unlock(&loghandle->lgh_hdr_lock);
LBUG(); /* should never happen */
}
llh->llh_count++;
spin_unlock(&loghandle->lgh_hdr_lock);
lgi->lgi_off = 0;
lgi->lgi_buf.lb_len = llh->llh_hdr.lrh_len;
lgi->lgi_buf.lb_buf = &llh->llh_hdr;
rc = dt_record_write(env, o, &lgi->lgi_buf, &lgi->lgi_off, th);
if (rc)
GOTO(out, rc);
header_is_updated = true;
rc = dt_attr_get(env, o, &lgi->lgi_attr, NULL);
if (rc)
GOTO(out, rc);
LASSERT(lgi->lgi_attr.la_valid & LA_SIZE);
lgi->lgi_off = lgi->lgi_attr.la_size;
lgi->lgi_buf.lb_len = reclen;
lgi->lgi_buf.lb_buf = rec;
rc = dt_record_write(env, o, &lgi->lgi_buf, &lgi->lgi_off, th);
if (rc < 0)
GOTO(out, rc);
CDEBUG(D_OTHER, "added record "DOSTID": idx: %u, %u\n",
POSTID(&loghandle->lgh_id.lgl_oi), index, rec->lrh_len);
if (reccookie != NULL) {
reccookie->lgc_lgl = loghandle->lgh_id;
reccookie->lgc_index = index;
if ((rec->lrh_type == MDS_UNLINK_REC) ||
(rec->lrh_type == MDS_SETATTR64_REC))
reccookie->lgc_subsys = LLOG_MDS_OST_ORIG_CTXT;
else if (rec->lrh_type == OST_SZ_REC)
reccookie->lgc_subsys = LLOG_SIZE_ORIG_CTXT;
else
reccookie->lgc_subsys = -1;
rc = 1;
}
RETURN(rc);
out:
/* cleanup llog for error case */
spin_lock(&loghandle->lgh_hdr_lock);
ext2_clear_bit(index, llh->llh_bitmap);
llh->llh_count--;
spin_unlock(&loghandle->lgh_hdr_lock);
/* restore llog last_idx */
loghandle->lgh_last_idx--;
llh->llh_tail.lrt_index = loghandle->lgh_last_idx;
/* restore the header on disk if it was written */
if (header_is_updated) {
lgi->lgi_off = 0;
lgi->lgi_buf.lb_len = llh->llh_hdr.lrh_len;
lgi->lgi_buf.lb_buf = &llh->llh_hdr;
dt_record_write(env, o, &lgi->lgi_buf, &lgi->lgi_off, th);
}
RETURN(rc);
}
/* appends if idx == -1, otherwise overwrites record idx. */
static int llog_osd_write_rec(const struct lu_env *env,
struct llog_handle *loghandle,
struct llog_rec_hdr *rec,
struct llog_cookie *reccookie, int cookiecount,
void *buf, int idx, struct thandle *th)
{
struct llog_thread_info *lgi = llog_info(env);
struct llog_log_hdr *llh;
int reclen = rec->lrh_len;
int index, rc;
struct llog_rec_tail *lrt;
struct dt_object *o;
size_t left;
ENTRY;
LASSERT(env);
llh = loghandle->lgh_hdr;
LASSERT(llh);
o = loghandle->lgh_obj;
LASSERT(o);
LASSERT(th);
CDEBUG(D_OTHER, "new record %x to "DFID"\n",
rec->lrh_type, PFID(lu_object_fid(&o->do_lu)));
/* record length should not bigger than LLOG_CHUNK_SIZE */
if (buf)
rc = (reclen > LLOG_CHUNK_SIZE - sizeof(struct llog_rec_hdr) -
sizeof(struct llog_rec_tail)) ? -E2BIG : 0;
else
rc = (reclen > LLOG_CHUNK_SIZE) ? -E2BIG : 0;
if (rc)
RETURN(rc);
rc = dt_attr_get(env, o, &lgi->lgi_attr, NULL);
if (rc)
RETURN(rc);
if (buf)
/* write_blob adds header and tail to lrh_len. */
reclen = sizeof(*rec) + rec->lrh_len +
sizeof(struct llog_rec_tail);
if (idx != -1) {
/* no header: only allowed to insert record 1 */
if (idx != 1 && lgi->lgi_attr.la_size == 0)
LBUG();
if (idx && llh->llh_size && llh->llh_size != rec->lrh_len)
RETURN(-EINVAL);
if (!ext2_test_bit(idx, llh->llh_bitmap))
CERROR("%s: modify unset record %u\n",
o->do_lu.lo_dev->ld_obd->obd_name, idx);
if (idx != rec->lrh_index)
CERROR("%s: index mismatch %d %u\n",
o->do_lu.lo_dev->ld_obd->obd_name, idx,
rec->lrh_index);
lgi->lgi_off = 0;
rc = llog_osd_write_blob(env, o, &llh->llh_hdr, NULL,
&lgi->lgi_off, th);
/* we are done if we only write the header or on error */
if (rc || idx == 0)
RETURN(rc);
if (buf) {
/* We assume that caller has set lgh_cur_* */
lgi->lgi_off = loghandle->lgh_cur_offset;
CDEBUG(D_OTHER,
"modify record "LPX64": idx:%d/%u/%d, len:%u "
"offset %llu\n",
loghandle->lgh_id.lgl_oid, idx, rec->lrh_index,
loghandle->lgh_cur_idx, rec->lrh_len,
(long long)(lgi->lgi_off - sizeof(*llh)));
if (rec->lrh_index != loghandle->lgh_cur_idx) {
CERROR("%s: modify idx mismatch %u/%d\n",
o->do_lu.lo_dev->ld_obd->obd_name, idx,
loghandle->lgh_cur_idx);
RETURN(-EFAULT);
}
} else {
/* Assumes constant lrh_len */
lgi->lgi_off = sizeof(*llh) + (idx - 1) * reclen;
}
rc = llog_osd_write_blob(env, o, rec, buf, &lgi->lgi_off, th);
if (rc == 0 && reccookie) {
reccookie->lgc_lgl = loghandle->lgh_id;
reccookie->lgc_index = idx;
rc = 1;
}
RETURN(rc);
}
/* Make sure that records don't cross a chunk boundary, so we can
* process them page-at-a-time if needed. If it will cross a chunk
* boundary, write in a fake (but referenced) entry to pad the chunk.
*
* We know that llog_current_log() will return a loghandle that is
* big enough to hold reclen, so all we care about is padding here.
*/
LASSERT(lgi->lgi_attr.la_valid & LA_SIZE);
lgi->lgi_off = lgi->lgi_attr.la_size;
left = LLOG_CHUNK_SIZE - (lgi->lgi_off & (LLOG_CHUNK_SIZE - 1));
/* NOTE: padding is a record, but no bit is set */
if (left != 0 && left != reclen &&
left < (reclen + LLOG_MIN_REC_SIZE)) {
index = loghandle->lgh_last_idx + 1;
rc = llog_osd_pad(env, o, &lgi->lgi_off, left, index, th);
if (rc)
RETURN(rc);
loghandle->lgh_last_idx++; /*for pad rec*/
}
/* if it's the last idx in log file, then return -ENOSPC */
if (loghandle->lgh_last_idx >= LLOG_BITMAP_SIZE(llh) - 1)
RETURN(-ENOSPC);
loghandle->lgh_last_idx++;
index = loghandle->lgh_last_idx;
LASSERT(index < LLOG_BITMAP_SIZE(llh));
rec->lrh_index = index;
if (buf == NULL) {
lrt = (struct llog_rec_tail *)((char *)rec + rec->lrh_len -
sizeof(*lrt));
lrt->lrt_len = rec->lrh_len;
lrt->lrt_index = rec->lrh_index;
}
/* The caller should make sure only 1 process access the lgh_last_idx,
* Otherwise it might hit the assert.*/
LASSERT(index < LLOG_BITMAP_SIZE(llh));
spin_lock(&loghandle->lgh_hdr_lock);
if (ext2_set_bit(index, llh->llh_bitmap)) {
CERROR("%s: index %u already set in log bitmap\n",
o->do_lu.lo_dev->ld_obd->obd_name, index);
spin_unlock(&loghandle->lgh_hdr_lock);
LBUG(); /* should never happen */
}
llh->llh_count++;
spin_unlock(&loghandle->lgh_hdr_lock);
llh->llh_tail.lrt_index = index;
lgi->lgi_off = 0;
rc = llog_osd_write_blob(env, o, &llh->llh_hdr, NULL, &lgi->lgi_off,
th);
if (rc)
RETURN(rc);
rc = dt_attr_get(env, o, &lgi->lgi_attr, NULL);
if (rc)
RETURN(rc);
LASSERT(lgi->lgi_attr.la_valid & LA_SIZE);
lgi->lgi_off = lgi->lgi_attr.la_size;
rc = llog_osd_write_blob(env, o, rec, buf, &lgi->lgi_off, th);
if (rc)
RETURN(rc);
CDEBUG(D_RPCTRACE, "added record "LPX64": idx: %u, %u\n",
loghandle->lgh_id.lgl_oid, index, rec->lrh_len);
if (rc == 0 && reccookie) {
reccookie->lgc_lgl = loghandle->lgh_id;
reccookie->lgc_index = index;
if ((rec->lrh_type == MDS_UNLINK_REC) ||
(rec->lrh_type == MDS_SETATTR64_REC))
reccookie->lgc_subsys = LLOG_MDS_OST_ORIG_CTXT;
else if (rec->lrh_type == OST_SZ_REC)
reccookie->lgc_subsys = LLOG_SIZE_ORIG_CTXT;
else
reccookie->lgc_subsys = -1;
rc = 1;
}
RETURN(rc);
}
static inline void log_set_bit(struct log_c *l,
uint32_t *bs, unsigned bit)
{
ext2_set_bit(bit, (unsigned long *) bs);
l->touched_cleaned = 1;
}
/*
* There are two policies for allocating an inode. If the new inode is
* a directory, then a forward search is made for a block group with both
* free space and a low directory-to-inode ratio; if that fails, then of
* the groups with above-average free space, that group with the fewest
* directories already is chosen.
*
* For other inodes, search forward from the parent directory\'s block
* group to find a free inode.
*/
struct inode * ext2_new_inode (const struct inode * dir, int mode, int * err)
{
struct super_block * sb;
struct buffer_head * bh;
struct buffer_head * bh2;
int i, j, avefreei;
struct inode * inode;
int bitmap_nr;
struct ext2_group_desc * gdp;
struct ext2_group_desc * tmp;
struct ext2_super_block * es;
/* Cannot create files in a deleted directory */
if (!dir || !dir->i_nlink) {
*err = -EPERM;
return NULL;
}
inode = get_empty_inode ();
if (!inode) {
*err = -ENOMEM;
return NULL;
}
sb = dir->i_sb;
inode->i_sb = sb;
inode->i_flags = 0;
lock_super (sb);
es = sb->u.ext2_sb.s_es;
repeat:
gdp = NULL; i=0;
*err = -ENOSPC;
if (S_ISDIR(mode)) {
avefreei = le32_to_cpu(es->s_free_inodes_count) /
sb->u.ext2_sb.s_groups_count;
/* I am not yet convinced that this next bit is necessary.
i = dir->u.ext2_i.i_block_group;
for (j = 0; j < sb->u.ext2_sb.s_groups_count; j++) {
tmp = ext2_get_group_desc (sb, i, &bh2);
if (tmp &&
(le16_to_cpu(tmp->bg_used_dirs_count) << 8) <
le16_to_cpu(tmp->bg_free_inodes_count)) {
gdp = tmp;
break;
}
else
i = ++i % sb->u.ext2_sb.s_groups_count;
}
*/
if (!gdp) {
for (j = 0; j < sb->u.ext2_sb.s_groups_count; j++) {
tmp = ext2_get_group_desc (sb, j, &bh2);
if (tmp &&
le16_to_cpu(tmp->bg_free_inodes_count) &&
le16_to_cpu(tmp->bg_free_inodes_count) >= avefreei) {
if (!gdp ||
(le16_to_cpu(tmp->bg_free_blocks_count) >
le16_to_cpu(gdp->bg_free_blocks_count))) {
i = j;
gdp = tmp;
}
}
}
}
}
else
{
/*
* Try to place the inode in its parent directory
*/
i = dir->u.ext2_i.i_block_group;
tmp = ext2_get_group_desc (sb, i, &bh2);
if (tmp && le16_to_cpu(tmp->bg_free_inodes_count))
gdp = tmp;
else
{
/*
* Use a quadratic hash to find a group with a
* free inode
*/
for (j = 1; j < sb->u.ext2_sb.s_groups_count; j <<= 1) {
i += j;
if (i >= sb->u.ext2_sb.s_groups_count)
i -= sb->u.ext2_sb.s_groups_count;
tmp = ext2_get_group_desc (sb, i, &bh2);
if (tmp &&
le16_to_cpu(tmp->bg_free_inodes_count)) {
gdp = tmp;
break;
}
}
}
if (!gdp) {
/*
* That failed: try linear search for a free inode
*/
i = dir->u.ext2_i.i_block_group + 1;
for (j = 2; j < sb->u.ext2_sb.s_groups_count; j++) {
if (++i >= sb->u.ext2_sb.s_groups_count)
i = 0;
tmp = ext2_get_group_desc (sb, i, &bh2);
if (tmp &&
le16_to_cpu(tmp->bg_free_inodes_count)) {
gdp = tmp;
break;
}
}
}
}
if (!gdp) {
unlock_super (sb);
iput(inode);
return NULL;
}
bitmap_nr = load_inode_bitmap (sb, i);
if (bitmap_nr < 0) {
unlock_super (sb);
iput(inode);
*err = -EIO;
return NULL;
}
bh = sb->u.ext2_sb.s_inode_bitmap[bitmap_nr];
if ((j = ext2_find_first_zero_bit ((unsigned long *) bh->b_data,
EXT2_INODES_PER_GROUP(sb))) <
EXT2_INODES_PER_GROUP(sb)) {
if (ext2_set_bit (j, bh->b_data)) {
ext2_warning (sb, "ext2_new_inode",
"bit already set for inode %d", j);
goto repeat;
}
mark_buffer_dirty(bh, 1);
if (sb->s_flags & MS_SYNCHRONOUS) {
ll_rw_block (WRITE, 1, &bh);
wait_on_buffer (bh);
}
} else {
if (le16_to_cpu(gdp->bg_free_inodes_count) != 0) {
ext2_error (sb, "ext2_new_inode",
"Free inodes count corrupted in group %d",
i);
unlock_super (sb);
iput (inode);
return NULL;
}
goto repeat;
}
j += i * EXT2_INODES_PER_GROUP(sb) + 1;
if (j < EXT2_FIRST_INO(sb) || j > le32_to_cpu(es->s_inodes_count)) {
ext2_error (sb, "ext2_new_inode",
"reserved inode or inode > inodes count - "
"block_group = %d,inode=%d", i, j);
unlock_super (sb);
iput (inode);
return NULL;
}
gdp->bg_free_inodes_count =
cpu_to_le16(le16_to_cpu(gdp->bg_free_inodes_count) - 1);
if (S_ISDIR(mode))
gdp->bg_used_dirs_count =
cpu_to_le16(le16_to_cpu(gdp->bg_used_dirs_count) + 1);
mark_buffer_dirty(bh2, 1);
es->s_free_inodes_count =
cpu_to_le32(le32_to_cpu(es->s_free_inodes_count) - 1);
mark_buffer_dirty(sb->u.ext2_sb.s_sbh, 1);
sb->s_dirt = 1;
inode->i_mode = mode;
inode->i_sb = sb;
inode->i_nlink = 1;
inode->i_dev = sb->s_dev;
inode->i_uid = current->fsuid;
if (test_opt (sb, GRPID))
inode->i_gid = dir->i_gid;
else if (dir->i_mode & S_ISGID) {
inode->i_gid = dir->i_gid;
if (S_ISDIR(mode))
mode |= S_ISGID;
} else
inode->i_gid = current->fsgid;
inode->i_ino = j;
inode->i_blksize = PAGE_SIZE; /* This is the optimal IO size (for stat), not the fs block size */
inode->i_blocks = 0;
inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
inode->u.ext2_i.i_new_inode = 1;
inode->u.ext2_i.i_flags = dir->u.ext2_i.i_flags;
if (S_ISLNK(mode))
inode->u.ext2_i.i_flags &= ~(EXT2_IMMUTABLE_FL | EXT2_APPEND_FL);
inode->u.ext2_i.i_faddr = 0;
inode->u.ext2_i.i_frag_no = 0;
inode->u.ext2_i.i_frag_size = 0;
inode->u.ext2_i.i_file_acl = 0;
inode->u.ext2_i.i_dir_acl = 0;
inode->u.ext2_i.i_dtime = 0;
inode->u.ext2_i.i_block_group = i;
inode->i_op = NULL;
if (inode->u.ext2_i.i_flags & EXT2_SYNC_FL)
inode->i_flags |= MS_SYNCHRONOUS;
insert_inode_hash(inode);
mark_inode_dirty(inode);
inc_inode_version (inode, gdp, mode);
unlock_super (sb);
if(DQUOT_ALLOC_INODE(sb, inode)) {
sb->dq_op->drop(inode);
inode->i_nlink = 0;
iput(inode);
*err = -EDQUOT;
return NULL;
}
ext2_debug ("allocating inode %lu\n", inode->i_ino);
*err = 0;
return inode;
}
/* returns negative on error; 0 if success; 1 if success & log destroyed */
int llog_cancel_rec(const struct lu_env *env, struct llog_handle *loghandle,
int index)
{
struct llog_thread_info *lgi = llog_info(env);
struct dt_device *dt;
struct llog_log_hdr *llh = loghandle->lgh_hdr;
struct thandle *th;
int rc;
int rc1;
bool subtract_count = false;
ENTRY;
CDEBUG(D_RPCTRACE, "Canceling %d in log "DOSTID"\n", index,
POSTID(&loghandle->lgh_id.lgl_oi));
if (index == 0) {
CERROR("Can't cancel index 0 which is header\n");
RETURN(-EINVAL);
}
LASSERT(loghandle != NULL);
LASSERT(loghandle->lgh_ctxt != NULL);
LASSERT(loghandle->lgh_obj != NULL);
dt = lu2dt_dev(loghandle->lgh_obj->do_lu.lo_dev);
th = dt_trans_create(env, dt);
if (IS_ERR(th))
RETURN(PTR_ERR(th));
rc = llog_declare_write_rec(env, loghandle, &llh->llh_hdr, index, th);
if (rc < 0)
GOTO(out_trans, rc);
if ((llh->llh_flags & LLOG_F_ZAP_WHEN_EMPTY))
rc = llog_declare_destroy(env, loghandle, th);
th->th_wait_submit = 1;
rc = dt_trans_start_local(env, dt, th);
if (rc < 0)
GOTO(out_trans, rc);
down_write(&loghandle->lgh_lock);
/* clear bitmap */
mutex_lock(&loghandle->lgh_hdr_mutex);
if (!ext2_clear_bit(index, LLOG_HDR_BITMAP(llh))) {
CDEBUG(D_RPCTRACE, "Catalog index %u already clear?\n", index);
GOTO(out_unlock, rc);
}
loghandle->lgh_hdr->llh_count--;
subtract_count = true;
/* Pass this index to llog_osd_write_rec(), which will use the index
* to only update the necesary bitmap. */
lgi->lgi_cookie.lgc_index = index;
/* update header */
rc = llog_write_rec(env, loghandle, &llh->llh_hdr, &lgi->lgi_cookie,
LLOG_HEADER_IDX, th);
if (rc != 0)
GOTO(out_unlock, rc);
if ((llh->llh_flags & LLOG_F_ZAP_WHEN_EMPTY) &&
(llh->llh_count == 1) &&
((loghandle->lgh_last_idx == LLOG_HDR_BITMAP_SIZE(llh) - 1) ||
(loghandle->u.phd.phd_cat_handle != NULL &&
loghandle->u.phd.phd_cat_handle->u.chd.chd_current_log !=
loghandle))) {
/* never try to destroy it again */
llh->llh_flags &= ~LLOG_F_ZAP_WHEN_EMPTY;
rc = llog_trans_destroy(env, loghandle, th);
if (rc < 0) {
/* Sigh, can not destroy the final plain llog, but
* the bitmap has been clearly, so the record can not
* be accessed anymore, let's return 0 for now, and
* the orphan will be handled by LFSCK. */
CERROR("%s: can't destroy empty llog #"DOSTID
"#%08x: rc = %d\n",
loghandle->lgh_ctxt->loc_obd->obd_name,
POSTID(&loghandle->lgh_id.lgl_oi),
loghandle->lgh_id.lgl_ogen, rc);
GOTO(out_unlock, rc);
}
rc = LLOG_DEL_PLAIN;
}
out_unlock:
mutex_unlock(&loghandle->lgh_hdr_mutex);
up_write(&loghandle->lgh_lock);
out_trans:
rc1 = dt_trans_stop(env, dt, th);
if (rc == 0)
rc = rc1;
if (rc < 0 && subtract_count) {
mutex_lock(&loghandle->lgh_hdr_mutex);
loghandle->lgh_hdr->llh_count++;
ext2_set_bit(index, LLOG_HDR_BITMAP(llh));
mutex_unlock(&loghandle->lgh_hdr_mutex);
}
RETURN(rc);
}
/* appends if idx == -1, otherwise overwrites record idx. */
static int llog_lvfs_write_rec(struct llog_handle *loghandle,
struct llog_rec_hdr *rec,
struct llog_cookie *reccookie, int cookiecount,
void *buf, int idx)
{
struct llog_log_hdr *llh;
int reclen = rec->lrh_len, index, rc;
struct llog_rec_tail *lrt;
struct obd_device *obd;
struct file *file;
size_t left;
ENTRY;
llh = loghandle->lgh_hdr;
file = loghandle->lgh_file;
obd = loghandle->lgh_ctxt->loc_exp->exp_obd;
/* record length should not bigger than LLOG_CHUNK_SIZE */
if (buf)
rc = (reclen > LLOG_CHUNK_SIZE - sizeof(struct llog_rec_hdr) -
sizeof(struct llog_rec_tail)) ? -E2BIG : 0;
else
rc = (reclen > LLOG_CHUNK_SIZE) ? -E2BIG : 0;
if (rc)
RETURN(rc);
if (buf)
/* write_blob adds header and tail to lrh_len. */
reclen = sizeof(*rec) + rec->lrh_len +
sizeof(struct llog_rec_tail);
if (idx != -1) {
loff_t saved_offset;
/* no header: only allowed to insert record 1 */
if (idx != 1 && !i_size_read(file->f_dentry->d_inode)) {
CERROR("idx != -1 in empty log\n");
LBUG();
}
if (idx && llh->llh_size && llh->llh_size != rec->lrh_len)
RETURN(-EINVAL);
if (!ext2_test_bit(idx, llh->llh_bitmap))
CERROR("Modify unset record %u\n", idx);
if (idx != rec->lrh_index)
CERROR("Index mismatch %d %u\n", idx, rec->lrh_index);
rc = llog_lvfs_write_blob(obd, file, &llh->llh_hdr, NULL, 0);
/* we are done if we only write the header or on error */
if (rc || idx == 0)
RETURN(rc);
/* Assumes constant lrh_len */
saved_offset = sizeof(*llh) + (idx - 1) * reclen;
if (buf) {
struct llog_rec_hdr check;
/* We assume that caller has set lgh_cur_* */
saved_offset = loghandle->lgh_cur_offset;
CDEBUG(D_OTHER,
"modify record "LPX64": idx:%d/%u/%d, len:%u "
"offset %llu\n",
loghandle->lgh_id.lgl_oid, idx, rec->lrh_index,
loghandle->lgh_cur_idx, rec->lrh_len,
(long long)(saved_offset - sizeof(*llh)));
if (rec->lrh_index != loghandle->lgh_cur_idx) {
CERROR("modify idx mismatch %u/%d\n",
idx, loghandle->lgh_cur_idx);
RETURN(-EFAULT);
}
#if 1 /* FIXME remove this safety check at some point */
/* Verify that the record we're modifying is the
right one. */
rc = llog_lvfs_read_blob(obd, file, &check,
sizeof(check), saved_offset);
if (check.lrh_index != idx || check.lrh_len != reclen) {
CERROR("Bad modify idx %u/%u size %u/%u (%d)\n",
idx, check.lrh_index, reclen,
check.lrh_len, rc);
RETURN(-EFAULT);
}
#endif
}
rc = llog_lvfs_write_blob(obd, file, rec, buf, saved_offset);
if (rc == 0 && reccookie) {
reccookie->lgc_lgl = loghandle->lgh_id;
reccookie->lgc_index = idx;
rc = 1;
}
RETURN(rc);
}
/* Make sure that records don't cross a chunk boundary, so we can
* process them page-at-a-time if needed. If it will cross a chunk
* boundary, write in a fake (but referenced) entry to pad the chunk.
*
* We know that llog_current_log() will return a loghandle that is
* big enough to hold reclen, so all we care about is padding here.
*/
left = LLOG_CHUNK_SIZE - (file->f_pos & (LLOG_CHUNK_SIZE - 1));
/* NOTE: padding is a record, but no bit is set */
if (left != 0 && left != reclen &&
left < (reclen + LLOG_MIN_REC_SIZE)) {
index = loghandle->lgh_last_idx + 1;
rc = llog_lvfs_pad(obd, file, left, index);
if (rc)
RETURN(rc);
loghandle->lgh_last_idx++; /*for pad rec*/
}
/* if it's the last idx in log file, then return -ENOSPC */
if (loghandle->lgh_last_idx >= LLOG_BITMAP_SIZE(llh) - 1)
RETURN(-ENOSPC);
loghandle->lgh_last_idx++;
index = loghandle->lgh_last_idx;
LASSERT(index < LLOG_BITMAP_SIZE(llh));
rec->lrh_index = index;
if (buf == NULL) {
lrt = (struct llog_rec_tail *)
((char *)rec + rec->lrh_len - sizeof(*lrt));
lrt->lrt_len = rec->lrh_len;
lrt->lrt_index = rec->lrh_index;
}
/*The caller should make sure only 1 process access the lgh_last_idx,
*Otherwise it might hit the assert.*/
LASSERT(index < LLOG_BITMAP_SIZE(llh));
if (ext2_set_bit(index, llh->llh_bitmap)) {
CERROR("argh, index %u already set in log bitmap?\n", index);
LBUG(); /* should never happen */
}
llh->llh_count++;
llh->llh_tail.lrt_index = index;
rc = llog_lvfs_write_blob(obd, file, &llh->llh_hdr, NULL, 0);
if (rc)
RETURN(rc);
rc = llog_lvfs_write_blob(obd, file, rec, buf, file->f_pos);
if (rc)
RETURN(rc);
CDEBUG(D_RPCTRACE, "added record "LPX64": idx: %u, %u \n",
loghandle->lgh_id.lgl_oid, index, rec->lrh_len);
if (rc == 0 && reccookie) {
reccookie->lgc_lgl = loghandle->lgh_id;
reccookie->lgc_index = index;
if ((rec->lrh_type == MDS_UNLINK_REC) ||
(rec->lrh_type == MDS_SETATTR_REC) ||
(rec->lrh_type == MDS_SETATTR64_REC))
reccookie->lgc_subsys = LLOG_MDS_OST_ORIG_CTXT;
else if (rec->lrh_type == OST_SZ_REC)
reccookie->lgc_subsys = LLOG_SIZE_ORIG_CTXT;
else if (rec->lrh_type == OST_RAID1_REC)
reccookie->lgc_subsys = LLOG_RD1_ORIG_CTXT;
else
reccookie->lgc_subsys = -1;
rc = 1;
}
if (rc == 0 && rec->lrh_type == LLOG_GEN_REC)
rc = 1;
RETURN(rc);
}
/* appends if idx == -1, otherwise overwrites record idx. */
static int mlog_vfs_write_rec(struct mlog_handle *loghandle,
struct mlog_rec_hdr *rec,
struct mlog_cookie *reccookie,
int cookiecount,
void *buf, int idx)
{
struct mlog_log_hdr *mlh;
int reclen = rec->mrh_len;
int index = 0;
int ret = 0;
struct mlog_rec_tail *mrt;
struct file *file;
size_t left;
struct mtfs_lowerfs *lowerfs = NULL;
MENTRY();
mlh = loghandle->mgh_hdr;
file = loghandle->mgh_file;
lowerfs = loghandle->mgh_ctxt->moc_lowerfs;
/* record length should not bigger than MLOG_CHUNK_SIZE */
if (buf){
ret = (reclen > MLOG_CHUNK_SIZE - sizeof(struct mlog_rec_hdr) -
sizeof(struct mlog_rec_tail)) ? -E2BIG : 0;
} else {
ret = (reclen > MLOG_CHUNK_SIZE) ? -E2BIG : 0;
}
if (ret) {
goto out;
}
if (buf) {
/* write_blob adds header and tail to mrh_len. */
reclen = sizeof(*rec) + rec->mrh_len +
sizeof(struct mlog_rec_tail);
}
if (idx != -1) {
loff_t saved_offset;
/* no header: only allowed to insert record 1 */
if (idx != 1 && !i_size_read(file->f_dentry->d_inode)) {
MERROR("idx != -1 in empty log\n");
MBUG();
}
if (idx && mlh->mlh_size && mlh->mlh_size != rec->mrh_len) {
ret = -EINVAL;
goto out;
}
if (!ext2_test_bit(idx, mlh->mlh_bitmap)) {
MERROR("Modify unset record %u\n", idx);
}
if (idx != rec->mrh_index) {
MERROR("Index mismatch %d %u\n", idx, rec->mrh_index);
}
ret = mlog_vfs_write_blob(lowerfs, file, &mlh->mlh_hdr, NULL, 0);
/* we are done if we only write the header or on error */
if (ret || idx == 0) {
goto out;
}
/* Assumes constant mrh_len */
saved_offset = sizeof(*mlh) + (idx - 1) * reclen;
if (buf) {
struct mlog_rec_hdr check;
/* We assume that caller has set mgh_cur_* */
saved_offset = loghandle->mgh_cur_offset;
MDEBUG("modify record %I64x: idx:%d/%u/%d, len:%u "
"offset %llu\n",
loghandle->mgh_id.mgl_oid, idx, rec->mrh_index,
loghandle->mgh_cur_idx, rec->mrh_len,
(long long)(saved_offset - sizeof(*mlh)));
if (rec->mrh_index != loghandle->mgh_cur_idx) {
MERROR("modify idx mismatch %u/%d\n",
idx, loghandle->mgh_cur_idx);
ret = -EFAULT;
goto out;
}
#if 1 /* FIXME remove this safety check at some point */
/* Verify that the record we're modifying is the
right one. */
ret = mlog_vfs_read_blob(lowerfs, file, &check,
sizeof(check), saved_offset);
if (check.mrh_index != idx || check.mrh_len != reclen) {
MERROR("bad modify idx %u/%u size %u/%u (%d)\n",
idx, check.mrh_index, reclen,
check.mrh_len, ret);
ret = -EFAULT;
goto out;
}
#endif
}
ret = mlog_vfs_write_blob(lowerfs, file, rec, buf, saved_offset);
if (ret == 0 && reccookie) {
reccookie->mgc_mgl = loghandle->mgh_id;
reccookie->mgc_index = idx;
ret = 1;
}
goto out;
}
/* Make sure that records don't cross a chunk boundary, so we can
* process them page-at-a-time if needed. If it will cross a chunk
* boundary, write in a fake (but referenced) entry to pad the chunk.
*
* We know that mlog_current_log() will return a loghandle that is
* big enough to hold reclen, so all we care about is padding here.
*/
left = MLOG_CHUNK_SIZE - (file->f_pos & (MLOG_CHUNK_SIZE - 1));
/* NOTE: padding is a record, but no bit is set */
if (left != 0 && left != reclen &&
left < (reclen + MLOG_MIN_REC_SIZE)) {
index = loghandle->mgh_last_idx + 1;
ret = mlog_vfs_pad(lowerfs, file, left, index);
if (ret) {
goto out;
}
loghandle->mgh_last_idx++; /*for pad rec*/
}
/* if it's the last idx in log file, then return -ENOSPC */
if (loghandle->mgh_last_idx >= MLOG_BITMAP_SIZE(mlh) - 1) {
ret = -ENOSPC;
goto out;
}
index = ++loghandle->mgh_last_idx;
rec->mrh_index = index;
if (buf == NULL) {
mrt = (struct mlog_rec_tail *)
((char *)rec + rec->mrh_len - sizeof(*mrt));
mrt->mrt_len = rec->mrh_len;
mrt->mrt_index = rec->mrh_index;
}
/*The caller should make sure only 1 process access the mgh_last_idx,
*Otherwise it might hit the assert.*/
MASSERT(index < MLOG_BITMAP_SIZE(mlh));
if (ext2_set_bit(index, mlh->mlh_bitmap)) {
MERROR("argh, index %u already set in log bitmap?\n", index);
MBUG(); /* should never happen */
}
mlh->mlh_count++;
mlh->mlh_tail.mrt_index = index;
ret = mlog_vfs_write_blob(lowerfs, file, &mlh->mlh_hdr, NULL, 0);
if (ret) {
goto out;
}
ret = mlog_vfs_write_blob(lowerfs, file, rec, buf, file->f_pos);
if (ret) {
goto out;
}
MDEBUG("added record %I64x: idx: %u, %u bytes\n",
loghandle->mgh_id.mgl_oid, index, rec->mrh_len);
if (ret == 0 && reccookie) {
reccookie->mgc_mgl = loghandle->mgh_id;
reccookie->mgc_index = index;
ret = 1;
}
if (ret == 0 && rec->mrh_type == MLOG_GEN_REC) {
ret = 1;
}
out:
MRETURN(ret);
}
int mlog_init_handle(struct mlog_handle *handle, int flags,
struct mlog_uuid *uuid)
{
int ret = 0;
struct mlog_log_hdr *mlh = NULL;
MENTRY();
MASSERT(handle->mgh_hdr == NULL);
MTFS_ALLOC(mlh, sizeof(*mlh));
if (mlh == NULL) {
MERROR("not enough memory\n");
ret = -ENOMEM;
goto out;
}
handle->mgh_hdr = mlh;
/* first assign flags to use mlog_client_ops */
mlh->mlh_flags = flags;
ret = mlog_read_header(handle);
if (ret == 0) {
flags = mlh->mlh_flags;
if (uuid && !mlog_uuid_equals(uuid, &mlh->mlh_tgtuuid)) {
MERROR("uuid mismatch: %s/%s\n", (char *)uuid->uuid,
(char *)mlh->mlh_tgtuuid.uuid);
ret = -EEXIST;
}
goto out;
} else if (ret != MLOG_EEMPTY || !flags) {
/* set a pesudo flag for initialization */
flags = MLOG_F_IS_CAT;
goto out;
}
ret = 0;
handle->mgh_last_idx = 0; /* header is record with index 0 */
mlh->mlh_count = 1; /* for the header record */
mlh->mlh_hdr.mrh_type = MLOG_HDR_MAGIC;
mlh->mlh_hdr.mrh_len = mlh->mlh_tail.mrt_len = MLOG_CHUNK_SIZE;
mlh->mlh_hdr.mrh_index = mlh->mlh_tail.mrt_index = 0;
mlh->mlh_timestamp = get_seconds();
if (uuid)
memcpy(&mlh->mlh_tgtuuid, uuid, sizeof(mlh->mlh_tgtuuid));
mlh->mlh_bitmap_offset = offsetof(typeof(*mlh), mlh_bitmap);
ext2_set_bit(0, mlh->mlh_bitmap);
out:
if (flags & MLOG_F_IS_CAT) {
MTFS_INIT_LIST_HEAD(&handle->u.chd.chd_head);
mlh->mlh_size = sizeof(struct mlog_logid_rec);
} else if (flags & MLOG_F_IS_PLAIN) {
MTFS_INIT_LIST_HEAD(&handle->u.phd.phd_entry);
} else {
MERROR("Unknown flags: %#x (Expected %#x or %#x\n",
flags, MLOG_F_IS_CAT, MLOG_F_IS_PLAIN);
MBUG();
}
if (ret) {
MTFS_FREE(mlh, sizeof(*mlh));
handle->mgh_hdr = NULL;
}
MRETURN(ret);
}
