Re: [f2fs-dev] [PATCH v2 2/2] f2fs: add F2FS_IOC_SET_COMPRESS_OPTION ioctl

[Eric Biggers <ebiggers@kernel.org>]

On Thu, Oct 22, 2020 at 12:58:48PM +0900, Daeho Jeong wrote:
> diff --git a/fs/f2fs/compress.c b/fs/f2fs/compress.c
> index 7895186cc765..3b58a41223f8 100644
> --- a/fs/f2fs/compress.c
> +++ b/fs/f2fs/compress.c
> @@ -514,6 +514,11 @@ bool f2fs_is_compress_backend_ready(struct inode *inode)
>  	return f2fs_cops[F2FS_I(inode)->i_compress_algorithm];
>  }
>  
> +bool f2fs_is_compress_algorithm_ready(unsigned char algorithm)
> +{
> +	return algorithm >= COMPRESS_MAX ? false : f2fs_cops[algorithm];
> +}

The use of ?: here is a bit strange.  How about:

	return algorithm < COMPRESS_MAX && f2fs_cops[algorithm] != NULL;

> +	if (option.log_cluster_size < MIN_COMPRESS_LOG_SIZE ||
> +			option.log_cluster_size > MAX_COMPRESS_LOG_SIZE ||
> +			!f2fs_is_compress_algorithm_ready(option.algorithm))
> +		return -EINVAL;

Likewise, EINVAL tends to be over-used, which makes it ambiguous.  Maybe use
ENOPKG for the case where algorithm < COMPRESS_MAX but the algorithm wasn't
compiled into the kernel?  That would be similar to how opening an encrypted
file fails with ENOPKG when the encryption algorithm isn't available.

> +	if (f2fs_is_mmap_file(inode) ||
> +			get_dirty_pages(inode) || inode->i_size) {
> +		ret = -EINVAL;
> +		goto out;
> +	}

How about EBUSY for f2fs_is_mmap_file(inode) || get_dirty_pages(inode),
and EFBIG for inode->i_size != 0?

- Eric