Re: [PATCH 3/4] mm: Support only one page_type per page

[Matthew Wilcox <willy@infradead.org>]

On Fri, Aug 01, 2025 at 04:13:33PM +0800, Kefeng Wang wrote:
> On 2025/8/1 10:43, Matthew Wilcox wrote:
> > On Wed, Aug 28, 2024 at 11:35:28AM +0800, Kefeng Wang wrote:
> > > There are some UBSAN warning about  __folio_set_##fname/__SetPage##uname,
> > > 
> > > UBSAN: shift-out-of-bounds in ../include/linux/page-flags.h:998:1
> > > left shift of 240 by 24 places cannot be represented in type 'int'
> > 
> > I can't reproduce this.  I know Andrew merged this patch in, but I
> > tried backing it out and enabling UBSAN and it doesn't show up for me.
> > Relevant part of .config:
> > 
> > CONFIG_ARCH_HAS_UBSAN=y
> > CONFIG_UBSAN=y
> > CONFIG_UBSAN_TRAP=y
> > CONFIG_CC_HAS_UBSAN_BOUNDS_STRICT=y
> > CONFIG_UBSAN_BOUNDS=y
> > CONFIG_UBSAN_BOUNDS_STRICT=y
> > CONFIG_UBSAN_SHIFT=y
> > # CONFIG_UBSAN_DIV_ZERO is not set
> > CONFIG_UBSAN_BOOL=y
> > CONFIG_UBSAN_ENUM=y
> > # CONFIG_TEST_UBSAN is not set
> > 
> > (I tried CONFIG_UBSAN_TRAP both on and off)
> 
> Hi Matthrew, I work on arm64, so the above OOB should occur on arm64,
> I remember the OOB appears on linux-next, but I tried last kernel and
> some old kernel version, but can't reproduce this too, it was too long ago,
> maybe gcc version related or some other changes...

Yes, I can't reproduce it either with an arm64 build.  I'll ignore this
issue in my refactoring, and if it pops up again, we'll deal with it.