file.c

ocfs1.4.1 oracle分布式文件系统

	return ret;
}

static int ocfs2_remove_inode_range(struct inode *inode,
				    struct buffer_head *di_bh, u64 byte_start,
				    u64 byte_len)
{
	int ret = 0;
	u32 trunc_start, trunc_len, cpos, phys_cpos, alloc_size;
	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
	struct ocfs2_cached_dealloc_ctxt dealloc;
	struct address_space *mapping = inode->i_mapping;

	ocfs2_init_dealloc_ctxt(&dealloc);

	if (byte_len == 0)
		return 0;

	if (OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) {
		ret = ocfs2_truncate_inline(inode, di_bh, byte_start,
					    byte_start + byte_len, 0);
		if (ret) {
			mlog_errno(ret);
			goto out;
		}
		/*
		 * There's no need to get fancy with the page cache
		 * truncate of an inline-data inode. We're talking
		 * about less than a page here, which will be cached
		 * in the dinode buffer anyway.
		 */
		unmap_mapping_range(mapping, 0, 0, 0);
		truncate_inode_pages(mapping, 0);
		goto out;
	}

	trunc_start = ocfs2_clusters_for_bytes(osb->sb, byte_start);
	trunc_len = (byte_start + byte_len) >> osb->s_clustersize_bits;
	if (trunc_len >= trunc_start)
		trunc_len -= trunc_start;
	else
		trunc_len = 0;

	mlog(0, "Inode: %llu, start: %llu, len: %llu, cstart: %u, clen: %u\n",
	     (unsigned long long)OCFS2_I(inode)->ip_blkno,
	     (unsigned long long)byte_start,
	     (unsigned long long)byte_len, trunc_start, trunc_len);

	ret = ocfs2_zero_partial_clusters(inode, byte_start, byte_len);
	if (ret) {
		mlog_errno(ret);
		goto out;
	}

	cpos = trunc_start;
	while (trunc_len) {
		ret = ocfs2_get_clusters(inode, cpos, &phys_cpos,
					 &alloc_size, NULL);
		if (ret) {
			mlog_errno(ret);
			goto out;
		}

		if (alloc_size > trunc_len)
			alloc_size = trunc_len;

		/* Only do work for non-holes */
		if (phys_cpos != 0) {
			ret = __ocfs2_remove_inode_range(inode, di_bh, cpos,
							 phys_cpos, alloc_size,
							 &dealloc);
			if (ret) {
				mlog_errno(ret);
				goto out;
			}
		}

		cpos += alloc_size;
		trunc_len -= alloc_size;
	}

	ocfs2_truncate_cluster_pages(inode, byte_start, byte_len);

out:
	ocfs2_schedule_truncate_log_flush(osb, 1);
	ocfs2_run_deallocs(osb, &dealloc);

	return ret;
}

/*
 * Parts of this function taken from xfs_change_file_space()
 */
static int __ocfs2_change_file_space(struct file *file, struct inode *inode,
				     loff_t f_pos, unsigned int cmd,
				     struct ocfs2_space_resv *sr,
				     int change_size)
{
	int ret;
	s64 llen;
	loff_t size;
	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
	struct buffer_head *di_bh = NULL;
	handle_t *handle;
	unsigned long long max_off = inode->i_sb->s_maxbytes;

	if (ocfs2_is_hard_readonly(osb) || ocfs2_is_soft_readonly(osb))
		return -EROFS;

	mutex_lock(&inode->i_mutex);

	/*
	 * This prevents concurrent writes on other nodes
	 */
	ret = ocfs2_rw_lock(inode, 1);
	if (ret) {
		mlog_errno(ret);
		goto out;
	}

	ret = ocfs2_inode_lock(inode, &di_bh, 1);
	if (ret) {
		mlog_errno(ret);
		goto out_rw_unlock;
	}

	if (inode->i_flags & (S_IMMUTABLE|S_APPEND)) {
		ret = -EPERM;
		goto out_inode_unlock;
	}

	switch (sr->l_whence) {
	case 0: /*SEEK_SET*/
		break;
	case 1: /*SEEK_CUR*/
		sr->l_start += f_pos;
		break;
	case 2: /*SEEK_END*/
		sr->l_start += i_size_read(inode);
		break;
	default:
		ret = -EINVAL;
		goto out_inode_unlock;
	}
	sr->l_whence = 0;

	llen = sr->l_len > 0 ? sr->l_len - 1 : sr->l_len;

	if (sr->l_start < 0
	    || sr->l_start > max_off
	    || (sr->l_start + llen) < 0
	    || (sr->l_start + llen) > max_off) {
		ret = -EINVAL;
		goto out_inode_unlock;
	}
	size = sr->l_start + sr->l_len;

	if (cmd == OCFS2_IOC_RESVSP || cmd == OCFS2_IOC_RESVSP64) {
		if (sr->l_len <= 0) {
			ret = -EINVAL;
			goto out_inode_unlock;
		}
	}

	if (file && should_remove_suid(filp_dentry(file))) {
		ret = __ocfs2_write_remove_suid(inode, di_bh);
		if (ret) {
			mlog_errno(ret);
			goto out_inode_unlock;
		}
	}

	down_write(&OCFS2_I(inode)->ip_alloc_sem);
	switch (cmd) {
	case OCFS2_IOC_RESVSP:
	case OCFS2_IOC_RESVSP64:
		/*
		 * This takes unsigned offsets, but the signed ones we
		 * pass have been checked against overflow above.
		 */
		ret = ocfs2_allocate_unwritten_extents(inode, sr->l_start,
						       sr->l_len);
		break;
	case OCFS2_IOC_UNRESVSP:
	case OCFS2_IOC_UNRESVSP64:
		ret = ocfs2_remove_inode_range(inode, di_bh, sr->l_start,
					       sr->l_len);
		break;
	default:
		ret = -EINVAL;
	}
	up_write(&OCFS2_I(inode)->ip_alloc_sem);
	if (ret) {
		mlog_errno(ret);
		goto out_inode_unlock;
	}

	/*
	 * We update c/mtime for these changes
	 */
	handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
	if (IS_ERR(handle)) {
		ret = PTR_ERR(handle);
		mlog_errno(ret);
		goto out_inode_unlock;
	}

	if (change_size && i_size_read(inode) < size)
		i_size_write(inode, size);

	inode->i_ctime = inode->i_mtime = CURRENT_TIME;
	ret = ocfs2_mark_inode_dirty(handle, inode, di_bh);
	if (ret < 0)
		mlog_errno(ret);

	ocfs2_commit_trans(osb, handle);

out_inode_unlock:
	brelse(di_bh);
	ocfs2_inode_unlock(inode, 1);
out_rw_unlock:
	ocfs2_rw_unlock(inode, 1);

out:
	mutex_unlock(&inode->i_mutex);
	return ret;
}

int ocfs2_change_file_space(struct file *file, unsigned int cmd,
			    struct ocfs2_space_resv *sr)
{
	struct inode *inode = filp_dentry(file)->d_inode;
	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);;

	if ((cmd == OCFS2_IOC_RESVSP || cmd == OCFS2_IOC_RESVSP64) &&
	    !ocfs2_writes_unwritten_extents(osb))
		return -ENOTTY;
	else if ((cmd == OCFS2_IOC_UNRESVSP || cmd == OCFS2_IOC_UNRESVSP64) &&
		 !ocfs2_sparse_alloc(osb))
		return -ENOTTY;

	if (!S_ISREG(inode->i_mode))
		return -EINVAL;

	if (!(file->f_mode & FMODE_WRITE))
		return -EBADF;

	return __ocfs2_change_file_space(file, inode, file->f_pos, cmd, sr, 0);
}

#ifndef NO_FALLOCATE
static long ocfs2_fallocate(struct inode *inode, int mode, loff_t offset,
			    loff_t len)
{
	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
	struct ocfs2_space_resv sr;
	int change_size = 1;

	if (!ocfs2_writes_unwritten_extents(osb))
		return -EOPNOTSUPP;

	if (S_ISDIR(inode->i_mode))
		return -ENODEV;

	if (mode & FALLOC_FL_KEEP_SIZE)
		change_size = 0;

	sr.l_whence = 0;
	sr.l_start = (s64)offset;
	sr.l_len = (s64)len;

	return __ocfs2_change_file_space(NULL, inode, offset,
					 OCFS2_IOC_RESVSP64, &sr, change_size);
}
#endif

static int ocfs2_prepare_inode_for_write(struct dentry *dentry,
					 loff_t *ppos,
					 size_t count,
					 int appending,
					 int *direct_io)
{
	int ret = 0, meta_level = 0;
	struct inode *inode = dentry->d_inode;
	loff_t saved_pos, end;

	/* 
	 * We start with a read level meta lock and only jump to an ex
	 * if we need to make modifications here.
	 */
	for(;;) {
		ret = ocfs2_inode_lock(inode, NULL, meta_level);
		if (ret < 0) {
			meta_level = -1;
			mlog_errno(ret);
			goto out;
		}

		/* Clear suid / sgid if necessary. We do this here
		 * instead of later in the write path because
		 * remove_suid() calls ->setattr without any hint that
		 * we may have already done our cluster locking. Since
		 * ocfs2_setattr() *must* take cluster locks to
		 * proceeed, this will lead us to recursively lock the
		 * inode. There's also the dinode i_size state which
		 * can be lost via setattr during extending writes (we
		 * set inode->i_size at the end of a write. */
		if (should_remove_suid(dentry)) {
			if (meta_level == 0) {
				ocfs2_inode_unlock(inode, meta_level);
				meta_level = 1;
				continue;
			}

			ret = ocfs2_write_remove_suid(inode);
			if (ret < 0) {
				mlog_errno(ret);
				goto out_unlock;
			}
		}

		/* work on a copy of ppos until we're sure that we won't have
		 * to recalculate it due to relocking. */
		if (appending) {
			saved_pos = i_size_read(inode);
			mlog(0, "O_APPEND: inode->i_size=%llu\n", saved_pos);
		} else {
			saved_pos = *ppos;
		}

		end = saved_pos + count;

		/*
		 * Skip the O_DIRECT checks if we don't need
		 * them.
		 */
		if (!direct_io || !(*direct_io))
			break;

		/*
		 * There's no sane way to do direct writes to an inode
		 * with inline data.
		 */
		if (OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) {
			*direct_io = 0;
			break;
		}

		/*
		 * Allowing concurrent direct writes means
		 * i_size changes wouldn't be synchronized, so
		 * one node could wind up truncating another
		 * nodes writes.
		 */
		if (end > i_size_read(inode)) {
			*direct_io = 0;
			break;
		}

		/*
		 * We don't fill holes during direct io, so
		 * check for them here. If any are found, the
		 * caller will have to retake some cluster
		 * locks and initiate the io as buffered.
		 */
		ret = ocfs2_check_range_for_holes(inode, saved_pos, count);
		if (ret == 1) {
			*direct_io = 0;
			ret = 0;
		} else if (ret < 0)
			mlog_errno(ret);
		break;
	}

	if (appending)
		*ppos = saved_pos;

out_unlock:
	ocfs2_inode_unlock(inode, meta_level);

out:
	return ret;
}

static inline void
ocfs2_set_next_iovec(const struct iovec **iovp, size_t *basep, size_t bytes)
{
	const struct iovec *iov = *iovp;
	size_t base = *basep;

	do {
		int copy = min(bytes, iov->iov_len - base);

		bytes -= copy;
		base += copy;
		if (iov->iov_len == base) {
			iov++;
			base = 0;
		}
	} while (bytes);
	*iovp = iov;
	*basep = base;
}

static struct page * ocfs2_get_write_source(char **ret_src_buf,
					    const struct iovec *cur_iov,
					    size_t iov_offset)
{
	int ret;
	char *buf = cur_iov->iov_base + iov_offset;
	struct page *src_page = NULL;
	unsigned long off;

	off = (unsigned long)(buf) & ~PAGE_CACHE_MASK;

	if (!segment_eq(get_fs(), KERNEL_DS)) {
		/*
		 * Pull in the user page. We want to do this outside
		 * of the meta data locks in order to preserve locking
		 * order in case of page fault.
		 */
		ret = get_user_pages(current, current->mm,
				     (unsigned long)buf & PAGE_CACHE_MASK, 1,
				     0, 0, &src_page, NULL);
		if (ret == 1)
			*ret_src_buf = kmap(src_page) + off;
		else
			src_page = ERR_PTR(-EFAULT);
	} else {
		*ret_src_buf = buf;
	}

	return src_page;
}

static void ocfs2_put_write_source(struct page *page)
{
	if (page) {
		kunmap(page);
		page_cache_release(page);
	}
}

static ssize_t ocfs2_file_buffered_write(struct file *file, loff_t *ppos,
					 const struct iovec *iov,
					 unsigned long nr_segs,
					 size_t count,
					 ssize_t o_direct_written)
{
	int ret = 0;
	ssize_t copied, total = 0;
	size_t iov_offset = 0, bytes;
	loff_t pos;
	const struct iovec *cur_iov = iov;
	struct page *user_page, *page;
	char * uninitialized_var(buf);
	char *dst;
	void *fsdata;

	/*
	 * handle partial DIO write.  Adjust cur_iov if needed.
	 */
	ocfs2_set_next_iovec(&cur_iov, &iov_offset, o_direct_written);

	do {
		pos = *ppos;

		user_page = ocfs2_get_write_source(&buf, cur_iov, iov_offset);
		if (IS_ERR(user_page)) {
			ret = PTR_ERR(user_page);
			goto out;
		}

		/* Stay within our page boundaries */
		bytes = min((PAGE_CACHE_SIZE - ((unsigned long)pos & ~PAGE_CACHE_MASK)),
			    (PAGE_CACHE_SIZE - ((unsigned long)buf & ~PAGE_CACHE_MASK)));
		/* Stay within the vector boundary */
		bytes = min_t(size_t, bytes, cur_iov->iov_len - iov_offset);
		/* Stay within count */
		bytes = min(bytes, count);

		page = NULL;
		ret = ocfs2_write_begin(file, file->f_mapping, pos, bytes, 0,
					&page, &fsdata);
		if (ret) {
			mlog_errno(ret);
			goto out;
		}

		dst = kmap_atomic(page, KM_USER0);
		memcpy(dst + (pos & (loff_t)(PAGE_CACHE_SIZE - 1)), buf, bytes);
		kunmap_atomic(dst, KM_USER0);
		flush_dcache_page(page);
		ocfs2_put_write_source(user_page);

		copied = ocfs2_write_end(file, file->f_mapping, pos, bytes,
					 bytes, page, fsdata);
		if (copied < 0) {
			mlog_errno(copied);
			ret = copied;
			goto out;
		}

		total += copied;
		*ppos = pos + copied;
		count -= copied;

		ocfs2_set_next_iovec(&cur_iov, &iov_offset, copied);
	} while(count);

out:
	return total ? total : ret;
}

static ssize_t __ocfs2_file_aio_write(struct kiocb *iocb,
				    const struct iovec *iov,
				    unsigned long nr_segs,