Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
MIME-Version: 1.0
References: <cover.1603372719.git.andreyknvl@google.com> <ce573435398f21d3e604f104c29ba65eca70d9e7.1603372719.git.andreyknvl@google.com>
 <CACT4Y+YF9bL8jRjVMfryr+LExYjH-rNdDEq2SvuQD+rGT4mVJQ@mail.gmail.com>
In-Reply-To: <CACT4Y+YF9bL8jRjVMfryr+LExYjH-rNdDEq2SvuQD+rGT4mVJQ@mail.gmail.com>
From:   Andrey Konovalov <andreyknvl@google.com>
Date:   Mon, 2 Nov 2020 16:17:04 +0100
Message-ID: <CAAeHK+zTPyX6h+8uJvjKwryA1U3L0ErufSoAmNvBu=QC5bomXw@mail.gmail.com>
Subject: Re: [PATCH RFC v2 16/21] kasan: optimize poisoning in kmalloc and krealloc
To:     Dmitry Vyukov <dvyukov@google.com>
Cc:     Catalin Marinas <catalin.marinas@arm.com>,
        Will Deacon <will.deacon@arm.com>,
        Vincenzo Frascino <vincenzo.frascino@arm.com>,
        Alexander Potapenko <glider@google.com>,
        Marco Elver <elver@google.com>,
        Evgenii Stepanov <eugenis@google.com>,
        Kostya Serebryany <kcc@google.com>,
        Peter Collingbourne <pcc@google.com>,
        Serban Constantinescu <serbanc@google.com>,
        Andrey Ryabinin <aryabinin@virtuozzo.com>,
        Elena Petrova <lenaptr@google.com>,
        Branislav Rankov <Branislav.Rankov@arm.com>,
        Kevin Brodsky <kevin.brodsky@arm.com>,
        Andrew Morton <akpm@linux-foundation.org>,
        kasan-dev <kasan-dev@googlegroups.com>,
        Linux ARM <linux-arm-kernel@lists.infradead.org>,
        Linux-MM <linux-mm@kvack.org>,
        LKML <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Wed, Oct 28, 2020 at 5:55 PM Dmitry Vyukov <dvyukov@google.com> wrote:
>
> On Thu, Oct 22, 2020 at 3:20 PM Andrey Konovalov <andreyknvl@google.com> wrote:
> >
> > Since kasan_kmalloc() always follows kasan_slab_alloc(), there's no need
> > to reunpoison the object data, only to poison the redzone.
> >
> > This requires changing kasan annotation for early SLUB cache to
> > kasan_slab_alloc(). Otherwise kasan_kmalloc() doesn't untag the object.
> > This doesn't do any functional changes, as kmem_cache_node->object_size
> > is equal to sizeof(struct kmem_cache_node).
> >
> > Similarly for kasan_krealloc(), as it's called after ksize(), which
> > already unpoisoned the object, there's no need to do it again.
>
> Have you considered doing this the other way around: make krealloc
> call __ksize and unpoison in kasan_krealloc?
> This has the advantage of more precise poisoning as ksize will
> unpoison the whole underlying object.
>
> But then maybe we will need to move first checks in ksize into __ksize
> as we may need them in krealloc as well.

This might be a good idea. I won't implement this for the next
version, but will look into this after that. Thanks!

>
>
>
>
>
> > Signed-off-by: Andrey Konovalov <andreyknvl@google.com>
> > Link: https://linux-review.googlesource.com/id/I4083d3b55605f70fef79bca9b90843c4390296f2
> > ---
> >  mm/kasan/common.c | 31 +++++++++++++++++++++----------
> >  mm/slub.c         |  3 +--
> >  2 files changed, 22 insertions(+), 12 deletions(-)
> >
> > diff --git a/mm/kasan/common.c b/mm/kasan/common.c
> > index c5ec60e1a4d2..a581937c2a44 100644
> > --- a/mm/kasan/common.c
> > +++ b/mm/kasan/common.c
> > @@ -360,8 +360,14 @@ static void *____kasan_kmalloc(struct kmem_cache *cache, const void *object,
> >         if (IS_ENABLED(CONFIG_KASAN_SW_TAGS) || IS_ENABLED(CONFIG_KASAN_HW_TAGS))
> >                 tag = assign_tag(cache, object, false, keep_tag);
> >
> > -       /* Tag is ignored in set_tag without CONFIG_KASAN_SW/HW_TAGS */
> > -       kasan_unpoison_memory(set_tag(object, tag), size);
> > +       /*
> > +        * Don't unpoison the object when keeping the tag. Tag is kept for:
> > +        * 1. krealloc(), and then the memory has already been unpoisoned via ksize();
> > +        * 2. kmalloc(), and then the memory has already been unpoisoned by kasan_kmalloc().
> > +        * Tag is ignored in set_tag() without CONFIG_KASAN_SW/HW_TAGS.
> > +        */
> > +       if (!keep_tag)
> > +               kasan_unpoison_memory(set_tag(object, tag), size);
> >         kasan_poison_memory((void *)redzone_start, redzone_end - redzone_start,
> >                 KASAN_KMALLOC_REDZONE);
> >
> > @@ -384,10 +390,9 @@ void * __must_check __kasan_kmalloc(struct kmem_cache *cache, const void *object
> >  }
> >  EXPORT_SYMBOL(__kasan_kmalloc);
> >
> > -void * __must_check __kasan_kmalloc_large(const void *ptr, size_t size,
> > -                                               gfp_t flags)
> > +static void * __must_check ____kasan_kmalloc_large(struct page *page, const void *ptr,
> > +                                               size_t size, gfp_t flags, bool realloc)
> >  {
> > -       struct page *page;
> >         unsigned long redzone_start;
> >         unsigned long redzone_end;
> >
> > @@ -397,18 +402,24 @@ void * __must_check __kasan_kmalloc_large(const void *ptr, size_t size,
> >         if (unlikely(ptr == NULL))
> >                 return NULL;
> >
> > -       page = virt_to_page(ptr);
> > -       redzone_start = round_up((unsigned long)(ptr + size),
> > -                               KASAN_GRANULE_SIZE);
> > +       redzone_start = round_up((unsigned long)(ptr + size), KASAN_GRANULE_SIZE);
> >         redzone_end = (unsigned long)ptr + page_size(page);
> >
> > -       kasan_unpoison_memory(ptr, size);
> > +       /* ksize() in __do_krealloc() already unpoisoned the memory. */
> > +       if (!realloc)
> > +               kasan_unpoison_memory(ptr, size);
> >         kasan_poison_memory((void *)redzone_start, redzone_end - redzone_start,
> >                 KASAN_PAGE_REDZONE);
> >
> >         return (void *)ptr;
> >  }
> >
> > +void * __must_check __kasan_kmalloc_large(const void *ptr, size_t size,
> > +                                               gfp_t flags)
> > +{
> > +       return ____kasan_kmalloc_large(virt_to_page(ptr), ptr, size, flags, false);
> > +}
> > +
> >  void * __must_check __kasan_krealloc(const void *object, size_t size, gfp_t flags)
> >  {
> >         struct page *page;
> > @@ -419,7 +430,7 @@ void * __must_check __kasan_krealloc(const void *object, size_t size, gfp_t flag
> >         page = virt_to_head_page(object);
> >
> >         if (unlikely(!PageSlab(page)))
> > -               return __kasan_kmalloc_large(object, size, flags);
> > +               return ____kasan_kmalloc_large(page, object, size, flags, true);
> >         else
> >                 return ____kasan_kmalloc(page->slab_cache, object, size,
> >                                                 flags, true);
> > diff --git a/mm/slub.c b/mm/slub.c
> > index 1d3f2355df3b..afb035b0bf2d 100644
> > --- a/mm/slub.c
> > +++ b/mm/slub.c
> > @@ -3535,8 +3535,7 @@ static void early_kmem_cache_node_alloc(int node)
> >         init_object(kmem_cache_node, n, SLUB_RED_ACTIVE);
> >         init_tracking(kmem_cache_node, n);
> >  #endif
> > -       n = kasan_kmalloc(kmem_cache_node, n, sizeof(struct kmem_cache_node),
> > -                     GFP_KERNEL);
> > +       n = kasan_slab_alloc(kmem_cache_node, n, GFP_KERNEL);
> >         page->freelist = get_freepointer(kmem_cache_node, n);
> >         page->inuse = 1;
> >         page->frozen = 0;
> > --
> > 2.29.0.rc1.297.gfa9743e501-goog
> >