linux-kernel - Re: [PATCH v4 08/10] userfaultfd: add UFFDIO

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <20210210190612.GR103365@xz-x1>
Date:   Wed, 10 Feb 2021 14:06:12 -0500
From:   Peter Xu <peterx@...hat.com>
To:     Axel Rasmussen <axelrasmussen@...gle.com>
Cc:     Alexander Viro <viro@...iv.linux.org.uk>,
        Alexey Dobriyan <adobriyan@...il.com>,
        Andrea Arcangeli <aarcange@...hat.com>,
        Andrew Morton <akpm@...ux-foundation.org>,
        Anshuman Khandual <anshuman.khandual@....com>,
        Catalin Marinas <catalin.marinas@....com>,
        Chinwen Chang <chinwen.chang@...iatek.com>,
        Huang Ying <ying.huang@...el.com>,
        Ingo Molnar <mingo@...hat.com>, Jann Horn <jannh@...gle.com>,
        Jerome Glisse <jglisse@...hat.com>,
        Lokesh Gidra <lokeshgidra@...gle.com>,
        "Matthew Wilcox (Oracle)" <willy@...radead.org>,
        Michael Ellerman <mpe@...erman.id.au>,
        Michal Koutný <mkoutny@...e.com>,
        Michel Lespinasse <walken@...gle.com>,
        Mike Kravetz <mike.kravetz@...cle.com>,
        Mike Rapoport <rppt@...ux.vnet.ibm.com>,
        Nicholas Piggin <npiggin@...il.com>, Shaohua Li <shli@...com>,
        Shawn Anastasio <shawn@...stas.io>,
        Steven Rostedt <rostedt@...dmis.org>,
        Steven Price <steven.price@....com>,
        Vlastimil Babka <vbabka@...e.cz>,
        LKML <linux-kernel@...r.kernel.org>,
        linux-fsdevel@...r.kernel.org, Linux MM <linux-mm@...ck.org>,
        Adam Ruprecht <ruprecht@...gle.com>,
        Cannon Matthews <cannonmatthews@...gle.com>,
        "Dr . David Alan Gilbert" <dgilbert@...hat.com>,
        David Rientjes <rientjes@...gle.com>,
        Mina Almasry <almasrymina@...gle.com>,
        Oliver Upton <oupton@...gle.com>
Subject: Re: [PATCH v4 08/10] userfaultfd: add UFFDIO_CONTINUE ioctl

On Wed, Feb 10, 2021 at 10:00:21AM -0800, Axel Rasmussen wrote:
> > >  static __always_inline ssize_t mfill_atomic_pte(struct mm_struct *dst_mm,
> > > @@ -417,10 +416,14 @@ static __always_inline ssize_t mfill_atomic_pte(struct mm_struct *dst_mm,
> > >                                               unsigned long dst_addr,
> > >                                               unsigned long src_addr,
> > >                                               struct page **page,
> > > -                                             bool zeropage,
> > > +                                             enum mcopy_atomic_mode mode,
> > >                                               bool wp_copy)
> > >  {
> > >       ssize_t err;
> > > +     bool zeropage = (mode == MCOPY_ATOMIC_ZEROPAGE);
> > > +
> > > +     if (mode == MCOPY_ATOMIC_CONTINUE)
> > > +             return -EINVAL;
> >
> > So you still passed in the mode into mfill_atomic_pte() just to make sure
> > CONTINUE is not called there.  It's okay, but again I think it's not extremely
> > necessary: we should make sure to fail early at the entry of uffdio_continue()
> > by checking against the vma type to be hugetlb, rather than reaching here.
> 
> Hmm, it's not quite as simple as that. We don't have the dst_vma yet
> in uffdio_continue(), __mcopy_atomic looks it up.
> 
> I'd prefer not to look it up in uffdio_continue(), because I think
> that means changing the API so all the ioctls look up the vma, and
> then plumb it into __mcopy_atomic. (We don't want to look it up twice,
> since each lookup has to traverse the rbtree.) This is complicated too
> by the fact that the ioctl handlers would need to perform various
> validation / checks - e.g., acquiring mmap_lock, dealing with
> *mmap_changing, validating the range, ....

Sure.

> 
> We can move the enforcement up one more layer, into __mcopy_atomic,
> easily enough, though.

Right, that sounds good to me.  It should be right after the "if
(!vma_is_anonymous(dst_vma) && !vma_is_shmem(dst_vma))" check.

Thanks,

-- 
Peter Xu