Re: [PATCH 5/5] vfs: don't allow writes to swap files

["Darrick J. Wong" <darrick.wong@oracle.com>]

On Wed, Jun 26, 2019 at 04:51:51AM +0100, Al Viro wrote:
> On Tue, Jun 25, 2019 at 07:33:31PM -0700, Darrick J. Wong wrote:
> > --- a/fs/attr.c
> > +++ b/fs/attr.c
> > @@ -236,6 +236,9 @@ int notify_change(struct dentry * dentry, struct iattr * attr, struct inode **de
> >  	if (IS_IMMUTABLE(inode))
> >  		return -EPERM;
> >  
> > +	if (IS_SWAPFILE(inode))
> > +		return -ETXTBSY;
> > +
> >  	if ((ia_valid & (ATTR_MODE | ATTR_UID | ATTR_GID | ATTR_TIMES_SET)) &&
> >  	    IS_APPEND(inode))
> >  		return -EPERM;
> 
> Er...  So why exactly is e.g. chmod(2) forbidden for swapfiles?  Or touch(1),
> for that matter...

Oops, that check is overly broad; I think the only attribute change we
need to filter here is ATTR_SIZE.... which we could do unconditionally
in inode_newsize_ok.

What's the use case for allowing userspace to increase the size of an
active swapfile?  I don't see any; the kernel has a permanent lease on
the file space mapping (at least until swapoff)...

> > diff --git a/mm/swapfile.c b/mm/swapfile.c
> > index 596ac98051c5..1ca4ee8c2d60 100644
> > --- a/mm/swapfile.c
> > +++ b/mm/swapfile.c
> > @@ -3165,6 +3165,19 @@ SYSCALL_DEFINE2(swapon, const char __user *, specialfile, int, swap_flags)
> >  	if (error)
> >  		goto bad_swap;
> >  
> > +	/*
> > +	 * Flush any pending IO and dirty mappings before we start using this
> > +	 * swap file.
> > +	 */
> > +	if (S_ISREG(inode->i_mode)) {
> > +		inode->i_flags |= S_SWAPFILE;
> > +		error = inode_drain_writes(inode);
> > +		if (error) {
> > +			inode->i_flags &= ~S_SWAPFILE;
> > +			goto bad_swap;
> > +		}
> > +	}
> 
> Why are swap partitions any less worthy of protection?

Hmm, yeah, S_SWAPFILE should apply to block devices too.  I figured that
the mantra of "sane tools will open block devices with O_EXCL" should
have sufficed, but there's really no reason to allow that either.

--D