Re: KASAN: use-after-free Read in mnt_warn_timestamp_expiry

[Eric Biggers]

On Thu, Oct 17, 2019 at 02:58:53AM +0100, Al Viro wrote:
> On Wed, Oct 16, 2019 at 06:47:55PM -0700, Eric Biggers wrote:
> > On Wed, Oct 16, 2019 at 06:42:11PM -0700, syzbot wrote:
> > > ==================================================================
> > > BUG: KASAN: use-after-free in mnt_warn_timestamp_expiry+0x4a/0x250
> > > fs/namespace.c:2471
> > > Read of size 8 at addr ffff888099937328 by task syz-executor.1/18510
> > > 
> > 
> > Looks like a duplicate of this:
> > 
> > #syz dup: KASAN: use-after-free Read in do_mount
> > 
> > See the existing thread and proposed fix here:
> > https://lkml.kernel.org/linux-fsdevel/000000000000805e5505945a234b@xxxxxxxxxx/T/#u
> 
> FWIW, I'd go with your "move mnt_warn_timestamp_expiry() up".  However,
> I'd probably turn the message into something like
> 	foofs filesystem getting mounted at /mnt/barf supports...
> And s/mounted/reconfigured/ if mnt_has_parent(mnt) is already true.
> 
> Objections?
> 

How about the following?

	pr_warn("%s filesystem being %s at %s supports timestamps until %04ld (0x%llx)\n",
		sb->s_type->name,
		is_mounted(mnt) ? "remounted" : "mounted",
		mntpath,
		tm.tm_year+1900, (unsigned long long)sb->s_time_max);

I think more people would understand "remounted" than "reconfigured".  Also,
is_mounted(mnt) seems like a better choice than mnt_has_parent(real_mount(mnt)).

- Eric