Re: [PATCH 4.4 010/149] f2fs: check memory boundary by insane namelen

[Eric Biggers]

On Thu, Aug 20, 2020 at 11:21:27AM +0200, Greg Kroah-Hartman wrote:
> From: Jaegeuk Kim <jaegeuk@xxxxxxxxxx>
> 
> [ Upstream commit 4e240d1bab1ead280ddf5eb05058dba6bbd57d10 ]
> 
> If namelen is corrupted to have very long value, fill_dentries can copy
> wrong memory area.
> 
> Reviewed-by: Chao Yu <yuchao0@xxxxxxxxxx>
> Signed-off-by: Jaegeuk Kim <jaegeuk@xxxxxxxxxx>
> Signed-off-by: Sasha Levin <sashal@xxxxxxxxxx>
> ---
>  fs/f2fs/dir.c | 11 ++++++++++-
>  1 file changed, 10 insertions(+), 1 deletion(-)
> 
> diff --git a/fs/f2fs/dir.c b/fs/f2fs/dir.c
> index 92a240616f520..5411d6667781f 100644
> --- a/fs/f2fs/dir.c
> +++ b/fs/f2fs/dir.c
> @@ -805,6 +805,16 @@ bool f2fs_fill_dentries(struct dir_context *ctx, struct f2fs_dentry_ptr *d,
>  		de_name.name = d->filename[bit_pos];
>  		de_name.len = le16_to_cpu(de->name_len);
>  
> +		/* check memory boundary before moving forward */
> +		bit_pos += GET_DENTRY_SLOTS(le16_to_cpu(de->name_len));
> +		if (unlikely(bit_pos > d->max)) {
> +			f2fs_msg(F2FS_I_SB(d->inode)->sb, KERN_WARNING,
> +				"%s: corrupted namelen=%d, run fsck to fix.",
> +				__func__, le16_to_cpu(de->name_len));
> +			set_sbi_flag(F2FS_I_SB(d->inode)->sb->s_fs_info, SBI_NEED_FSCK);
> +			return -EINVAL;
> +		}
> +
>  		if (f2fs_encrypted_inode(d->inode)) {
>  			int save_len = fstr->len;
>  			int ret;
> @@ -829,7 +839,6 @@ bool f2fs_fill_dentries(struct dir_context *ctx, struct f2fs_dentry_ptr *d,
>  					le32_to_cpu(de->ino), d_type))
>  			return true;
>  
> -		bit_pos += GET_DENTRY_SLOTS(le16_to_cpu(de->name_len));
>  		ctx->pos = start_pos + bit_pos;

This buggy backport broke f2fs encryption in 4.4-stable, due to a missing
prerequisite commit
(https://lkml.kernel.org/stable/20201031195809.377983-1-ebiggers@xxxxxxxxxx/).
Why didn't this backport get Cc'ed to the subsystem mailing list?  It wasn't
even a clean cherry pick.

- Eric