Re: [PATCH] arch:arm:mm:Correction in the boundary check for module end address.

["Hillf Danton" <hillf.zj@alibaba-inc.com>]

> On Tue, Nov 10, 2015 at 12:35:50AM +0900, Jungseung Lee wrote:
> > Hi,
> >
> > 2015-11-09 16:57 GMT+09:00 Shailendra Verma <shailendra.v@samung.com>:
> > > From: Shailendra Verma <Shailendra.v@samsung.com>
> > >
> > > The module end boundary check is not proper.The out of bound value
> > > of module end can produce undesired results.
> > >
> > > Signed-off-by: Shailendra Verma <Shailendra.v@samsung.com>
> > > Reviewed-by: Ravikant Bijendra Sharma <ravikant.s2@samsung.com>
> > > ---
> > >  linux-4.3-rc6/arch/arm/mm/pageattr.c |    2 +-
> > >  1 file changed, 1 insertion(+), 1 deletion(-)
> > >
> > > diff --git a/linux-4.3-rc6/arch/arm/mm/pageattr.c b/linux-4.3-rc6/arch/arm/mm/pageattr.c
> > > index cf30daf..be7fe4b 100644
> > > --- a/linux-4.3-rc6/arch/arm/mm/pageattr.c
> > > +++ b/linux-4.3-rc6/arch/arm/mm/pageattr.c
> > > @@ -52,7 +52,7 @@ static int change_memory_common(unsigned long addr, int numpages,
> > >         if (start < MODULES_VADDR || start >= MODULES_END)
> > >                 return -EINVAL;
> > >
> > > -       if (end < MODULES_VADDR || start >= MODULES_END)
> > > +       if (end < MODULES_VADDR || end >= MODULES_END)
> > >                 return -EINVAL;
> > >
> > >         data.set_mask = set_mask;
> > > --
> > > 1.7.9.5
> > >
> > Same patch with proper format is already submitted by Hillf.
> > https://lkml.org/lkml/2015/5/3/202
> 
> What happened to that patch?  It should at least have had a:
> 
> Fixes: f2ca09f381a5 ("ARM: 8311/1: Don't use is_module_addr in setting page attributes")
> 
> tag on it, and it needs to find its way into the patch system.  As
> no one has reported a crash (I don't think it's crash-causing, but is
> merely a correctness issue) I don't see any reason to backport it to
> stable trees.
> 
> Other changes are needed here as well - the original commit creating
> this contains:
> 
> +       if (!IS_ALIGNED(addr, PAGE_SIZE)) {
> +               start &= PAGE_MASK;
> +               end = start + size;
> +               WARN_ON_ONCE(1);
> +       }
> 
> which is truely amazing.  Fine, it may not be intended to work with
> non-aligned addresses, but if we're going to round the start down,
> then rounding the end down as well like that is also buggy.
> 
> 	unsigned long start = addr;
> 	unsigned long size = PAGE_SIZE * numpages;
> 	unsigned long end = start + size;
> 
> 	if (WARN_ON_ONCE(!IS_ALIGNED(addr, PAGE_SIZE)) {
> 		start &= PAGE_MASK;
> 		end = PAGE_ALIGN(end);
> 	}
> 
> would be much better - this will round 'end' up, so we encompass the
> entire requested region.
> 
Would you please respin, Shailendra, with Russell's comments
addressed, and with
Reported-by: Hillf Danton <hillf.zj@alibaba-inc.com> ?

thanks
Hillf