Re: [PATCH] futex: Fix fault_in_user_writeable()

[Huacai Chen]

Hi, all,

On Tue, Aug 17, 2021 at 5:45 PM Peter Zijlstra <peterz@xxxxxxxxxxxxx> wrote:
>
> On Tue, Aug 17, 2021 at 11:05:15AM +0200, Thomas Gleixner wrote:
> > Huacai,
> >
> > On Tue, Aug 17 2021 at 15:38, Huacai Chen wrote:
> > > On Tue, Aug 17, 2021 at 3:07 PM Thomas Gleixner <tglx@xxxxxxxxxxxxx> wrote:
> > > On X86, it returns 0; on MIPS64 without patch, it hangs in kernel; on
> > > MIPS64 with this patch, it returns -1.
> >
> > As expected.
> >
> > > Then, I want to know, on "W implies R" archs (such as X86), should it
> > > return 0? Maybe return -1 is more reasonable? (because the VMA is
> > > marked as write-only). If this program should return -1, then I don't
> > > think this is a MIPS-specific problem.
> >
> > No. mmap(.., PROT_WRITE...) is simply impossible on x86 and implies
> > PROT_READ as documented in mmap(2).
> >
> > So why should this fail and only fail in the fault case, but succeed
> > when the PTE is already established?
>
> I wouldn't actually mind if it failed on fault -- it's the 'best' we can
> do on x86. Doing a RmW op on PROT_WRITE is silly and deserves all the
> wreckage it can get.
If we must fix it in arch code, there are two methods: 1, don't use
write-only map (modify protection_map as Liu Lichao did); 2, override
arch_vma_access_permitted() to do extra checks. Thomas, which is
better?

Huacai