Re: [PATCH] hugetlbfs: fix a truncation issue in hugepages parameter

[liuyuntao]

On 2022-02-10  0:43 UTC, Mike Kravetz wrote:
> On 2/9/22 05:40, liuyuntao wrote:
> > From: Liu Yuntao <liuyuntao10@xxxxxxxxxx>
> > 
> > When we specify a large number for node in hugepages parameter,
> > it may be parsed to another number due to truncation in this statement:
> > 	node = tmp;
> > 
> > For example, add following parameter in command line:
> > 	hugepagesz=1G hugepages=4294967297:5
> > and kernel will allocate 5 hugepages for node 1 instead of ignoring it.
> > 
> > I move the validation check earlier to fix this issue, and slightly
> > simplifies the condition here.
> > 
> > Fixes: b5389086ad7be0 ("hugetlbfs: extend the definition of hugepages parameter to support node allocation")
> > Signed-off-by: Liu Yuntao <liuyuntao10@xxxxxxxxxx>
> > ---
> >  mm/hugetlb.c | 4 ++--
> >  1 file changed, 2 insertions(+), 2 deletions(-)
> > 
> > diff --git a/mm/hugetlb.c b/mm/hugetlb.c
> > index 61895cc01d09..0929547f6ad6 100644
> > --- a/mm/hugetlb.c
> > +++ b/mm/hugetlb.c
> > @@ -4159,10 +4159,10 @@ static int __init hugepages_setup(char *s)
> >  				pr_warn("HugeTLB: architecture can't support node specific alloc, ignoring!\n");
> >  				return 0;
> >  			}
> > +			if (tmp >= nr_online_nodes)
> > +				goto invalid;
> >  			node = tmp;
> 
> I am surprised none of the automated checking complained about that
> assignment.

I think such assignments may be very common in kernel, and thus automated
checks just ignore them.

> 
> >  			p += count + 1;
> > -			if (node < 0 || node >= nr_online_nodes)
> 
> I can't remember, but I think that check for node < 0 was added to handle
> overflow during the above assignment.  Do you remember Zhenguo Yao?

No, I don't. I took a look and found that the check for node < 0 has been
there since his first version of patch.

>    
> > -				goto invalid;
> >  			/* Parse hugepages */
> >  			if (sscanf(p, "%lu%n", &tmp, &count) != 1)
> >  				goto invalid;
> 
> Thanks,
> 
> Reviewed-by: Mike Kravetz <mike.kravetz@xxxxxxxxxx>
> 
> -- 
> Mike Kravetz

--
Liu Yuntao