Re: [PATCH] autofs - fix autofs_sbi() does not check super block type

[Ian Kent <raven@themaw.net>]

On Mon, 2018-08-27 at 02:03 +0100, Al Viro wrote:
> On Mon, Aug 20, 2018 at 04:37:09PM +0800, Ian Kent wrote:
> > The autofs_sbi() inline function does not check the super block
> > magic number to verify it has been given an autofs super block.
> 
> IMO it's the wrong way to fix it.  The one and only caller where that
> check might trigger is
> 
>                 if (!fp) {
>                         if (cmd == AUTOFS_DEV_IOCTL_ISMOUNTPOINT_CMD)
>                                 goto cont;
>                         err = -EBADF;
>                         goto out;
>                 }
> 
>                 sbi = autofs_dev_ioctl_sbi(fp);
>                 if (!sbi || sbi->magic != AUTOFS_SBI_MAGIC) {
>                         err = -EINVAL;
>                         fput(fp);
>                         goto out;
>                 }
> with
> static struct autofs_sb_info *autofs_dev_ioctl_sbi(struct file *f)
> {
>         struct autofs_sb_info *sbi = NULL;
>         struct inode *inode;
> 
>         if (f) { 
>                 inode = file_inode(f);
>                 sbi = autofs_sbi(inode->i_sb);
>         }
>         return sbi;
> }
> 
> First of all, what is that `if (f)' doing in there?  We have just checked
> that in the only caller.
> 
> Next, dereferencing the result of autofs_sbi() does need to be preceded
> by making sure that superblock is autofs one, all right... and what are
> we doing in that first dereferencing, again?
> 
> IOW, turn that into
> 
> 	if (!fp) {
> 		....
> 		goto out;
> 	}
> 	sb = file_inode(fp)->i_sb;
> 	if (sb->s_type != &autofs_fs_type)
> 		bugger off
> 	sbi = autofs_sbi(sb);
> 	....
> 
> and be done with that.  Other callers of autofs_sbi() really shouldn't
> happen to other filesystem's superblocks...

Yes, adding it to the inline does add a little extra for other
callers that won't get a non-autofs super block.

I was tempted to just change autofs_dev_ioctl_sbi() in case other
callers were added but your suggestion is somewhat simpler and
really only requires due attention if changes are made.

I'll send a patch to Andrew based on what you recommend.

Thanks
Ian