Re: [PATCH] ext2: strengthen value length check in ext2_xattr_set()

["cgxu519@zoho.com.cn" <cgxu519@zoho.com.cn>]

On Wed, 2019-05-22 at 11:50 +0200, Jan Kara wrote:
> On Wed 22-05-19 16:28:46, Chengguang Xu wrote:
> > Actually maximum length of a valid entry value is not
> > ->s_blocksize because header, last entry and entry
> > name will also occupy some spaces. This patch
> > strengthens the value length check and return -ERANGE
> > when the length is larger than allowed maximum length.
> > 
> > Signed-off-by: Chengguang Xu <cgxu519@zoho.com.cn>
> 
> Thanks for the patch! But what's the point of this change? We would return
> ERANGE instead of ENOSPC? I don't think that's serious enough to warrant
> changing existing behavior...

Hi Jan,

Thanks for the review. 

The motivation is seprating error situations of ENOSPC/ERANGE
because ENOSPC is giving a hint that we can save an EA entry
(name+value > allowed maximum length) by deleting some existing
entries. However, as you has pointed out, I also think the
difference is not so important because some EA entries
(like security index) is invisible for user...

Thanks,
Chengguang


> 
> > @@ -423,7 +423,10 @@ ext2_xattr_set(struct inode *inode, int name_index,
> > const char *name,
> >  	if (name == NULL)
> >  		return -EINVAL;
> >  	name_len = strlen(name);
> > -	if (name_len > 255 || value_len > sb->s_blocksize)
> > +	max_len = sb->s_blocksize - sizeof(struct ext2_xattr_header)
> > +			- sizeof(__u32);
> > +	if (name_len > 255 ||
> > +	    EXT2_XATTR_LEN(name_len) + EXT2_XATTR_SIZE(value_len) > max_len)
> >  		return -ERANGE;
> >  	down_write(&EXT2_I(inode)->xattr_sem);
> >  	if (EXT2_I(inode)->i_file_acl) {
> 
> 								Honza
>