Re: [PATCH v3 3/6] mm: Add build time sanity check for struct page size

[Baoquan He]

On 02/17/19 at 08:50am, Kees Cook wrote:
> On Sat, Feb 16, 2019 at 6:02 AM Baoquan He <bhe@xxxxxxxxxx> wrote:
> >
> > Size of struct page might be larger than 64 bytes if debug options
> > enabled, or fields added for debugging intentionally. Yet an upper
> > limit need be added at build time to trigger an alert in case the
> > size is too big to boot up system, warning people to check if it's
> > be done on purpose in advance.
> >
> > Here 1/4 of PAGE_SIZE is chosen since system must have been insane
> > with this value. For those systems with PAGE_SIZE larger than 4KB,
> > 1KB is simply taken.
> >
> > Signed-off-by: Baoquan He <bhe@xxxxxxxxxx>
> > ---
> >  mm/page_alloc.c | 2 ++
> >  1 file changed, 2 insertions(+)
> >
> > diff --git a/mm/page_alloc.c b/mm/page_alloc.c
> > index 35fdde041f5c..eb6c8e22333b 100644
> > --- a/mm/page_alloc.c
> > +++ b/mm/page_alloc.c
> > @@ -67,6 +67,7 @@
> >  #include <linux/lockdep.h>
> >  #include <linux/nmi.h>
> >  #include <linux/psi.h>
> > +#include <linux/sizes.h>
> >
> >  #include <asm/sections.h>
> >  #include <asm/tlbflush.h>
> > @@ -7084,6 +7085,7 @@ void __init free_area_init_nodes(unsigned long *max_zone_pfn)
> >         unsigned long start_pfn, end_pfn;
> >         int i, nid;
> >
> > +       BUILD_BUG_ON(sizeof(struct page) > min_t(size_t, SZ_1K, PAGE_SIZE));
> 
> Are there systems with PAGE_SIZE < 1K? Maybe this should just be a
> direct SZ_1K check?

This check was suggested by Kirill, I forgot adding his "Suggested-by",
sorry abou that.

Originally he suggested to add code in generic place like this:

	BUILD_BUG_ON(sizeof(struct page) < min(SZ_1K, PAGE_SIZE/4));

In later post, the kbuild test robot reported an build error on i386
ARCH,

http://lkml.kernel.org/r/20180911074733.GX1740@xxxxxxxxxxx

>From the report hint, I thought 'PAGE_SIZE/4' also related to the macro
expansion, so change it as min_t(size_t, SZ_1K, PAGE_SIZE).

Just now I tried the build again, changing it back to 'PAGE_SIZE/4' also
works.

	BUILD_BUG_ON(sizeof(struct page) > min_t(size_t, SZ_1K, PAGE_SIZE/4));

I guess Kirill wants to make it as self explanatory for this check by
suggesting it as min(SZ_1K, PAGE_SIZE/4), to stress the 'PAGE_SIZE/4'.
As he said in mail thread, "If struct page is more than 1/4 of PAGE_SIZE
something is horribly broken".

lkml.kernel.org/r/20180903102642.rmzawwqsqjvh2mkb@kshutemo-mobl1

Just try to give more details about this adding, not defend. I am
fine to take any of them.


> (Also, perhaps this should use the new static_assert where struct page
> is defined?)

I searched with 'git grep', didn't find static_assert macro or function.
And also no finding in include/linux/mm_types.h. Could you please be
more specific?

Thanks
Baoquan
> 
> >         /* Record where the zone boundaries are */
> >         memset(arch_zone_lowest_possible_pfn, 0,
> >                                 sizeof(arch_zone_lowest_possible_pfn));
> > --
> > 2.17.2
> >
> 
> 
> -- 
> Kees Cook