Re: [patch 21/78] kasan: split out shadow.c from common.c

[Marco Elver <elver@google.com>]

[Ignore previous email without reply -- this time with actual reply]

On Sat, 19 Dec 2020 at 02:13, Andrew Morton <akpm@linux-foundation.org> wrote:
> On Sat, 19 Dec 2020 01:28:29 +0100 Marco Elver <elver@google.com> wrote:
> > [...]
> > > -/*
> > > - * Poisons the shadow memory for 'size' bytes starting from 'addr'.
> > > - * Memory addresses should be aligned to KASAN_GRANULE_SIZE.
> > > - */
> > > -void poison_range(const void *address, size_t size, u8 value)
> > > -{
> > > -   void *shadow_start, *shadow_end;
> > > -
> > > -   /*
> > > -    * Perform shadow offset calculation based on untagged address, as
> > > -    * some of the callers (e.g. kasan_poison_object_data) pass tagged
> > > -    * addresses to this function.
> > > -    */
> > > -   address = reset_tag(address);
> > > -
> >
> > The moved lines do not mention kfence...
> > (The same commit in -next does.)
>
> They shouldn't.
>
> > > -   shadow_start = kasan_mem_to_shadow(address);
> > > -   shadow_end = kasan_mem_to_shadow(address + size);
> > > -
> > > -   __memset(shadow_start, value, shadow_end - shadow_start);
> > > -}
> > [...]
> > > --- /dev/null
> > > +++ a/mm/kasan/shadow.c
> > > @@ -0,0 +1,518 @@
> > > +// SPDX-License-Identifier: GPL-2.0
> > > +/*
> > > + * This file contains KASAN runtime code that manages shadow memory for
> > > + * generic and software tag-based KASAN modes.
> > > + *
> > > + * Copyright (c) 2014 Samsung Electronics Co., Ltd.
> > > + * Author: Andrey Ryabinin <ryabinin.a.a@gmail.com>
> > > + *
> > > + * Some code borrowed from https://github.com/xairy/kasan-prototype by
> > > + *        Andrey Konovalov <andreyknvl@gmail.com>
> > > + */
> > > +
> > > +#include <linux/init.h>
> > > +#include <linux/kasan.h>
> > > +#include <linux/kernel.h>
> > > +#include <linux/kfence.h>
> >
> > This is the first time kfence is mentioned. Is this correct?
>
> Yes.
>
> > Is my assumption correct that the kasan changes and kfence changes are
> > to be swapped?
>
> Yes, kfence came in fairly late and seems a bit fresh.  I was planning
> on holding it off until next cycle.
>
> Sigh.  I don't have access to my capable-of-compiling-KASAN machine at
> present :(  We'll need this, yes?

Looks reasonable; any mention of kfence should be removed from any of
the kasan patches if the kasan series goes before kfence. And kfence's
"kfence, kasan: make KFENCE compatible with KASAN" should absorb any
of those reverted changes.

Because kfence was picked up earlier, and appeared in -next before the
kasan series, the kasan series was rebased to not conflict with those
changes from kfence. Sorry for the inconvenience, and thank you for
sorting it out.

Thanks,
-- Marco

> --- a/mm/kasan/kasan.h~a
> +++ a/mm/kasan/kasan.h
> @@ -3,7 +3,6 @@
>  #define __MM_KASAN_KASAN_H
>
>  #include <linux/kasan.h>
> -#include <linux/kfence.h>
>  #include <linux/stackdepot.h>
>
>  #ifdef CONFIG_KASAN_HW_TAGS
> @@ -305,20 +304,12 @@ static inline u8 random_tag(void) { retu
>
>  static inline void poison_range(const void *address, size_t size, u8 value)
>  {
> -       /* Skip KFENCE memory if called explicitly outside of sl*b. */
> -       if (is_kfence_address(address))
> -               return;
> -
>         hw_set_mem_tag_range(kasan_reset_tag(address),
>                         round_up(size, KASAN_GRANULE_SIZE), value);
>  }
>
>  static inline void unpoison_range(const void *address, size_t size)
>  {
> -       /* Skip KFENCE memory if called explicitly outside of sl*b. */
> -       if (is_kfence_address(address))
> -               return;
> -
>         hw_set_mem_tag_range(kasan_reset_tag(address),
>                         round_up(size, KASAN_GRANULE_SIZE), get_tag(address));
>  }
> --- a/mm/kasan/shadow.c~a
> +++ a/mm/kasan/shadow.c
> @@ -13,7 +13,6 @@
>  #include <linux/init.h>
>  #include <linux/kasan.h>
>  #include <linux/kernel.h>
> -#include <linux/kfence.h>
>  #include <linux/kmemleak.h>
>  #include <linux/memory.h>
>  #include <linux/mm.h>
> @@ -85,10 +84,6 @@ void poison_range(const void *address, s
>         address = kasan_reset_tag(address);
>         size = round_up(size, KASAN_GRANULE_SIZE);
>
> -       /* Skip KFENCE memory if called explicitly outside of sl*b. */
> -       if (is_kfence_address(address))
> -               return;
> -
>         shadow_start = kasan_mem_to_shadow(address);
>         shadow_end = kasan_mem_to_shadow(address + size);
>
> @@ -106,14 +101,6 @@ void unpoison_range(const void *address,
>          */
>         address = kasan_reset_tag(address);
>
> -       /*
> -        * Skip KFENCE memory if called explicitly outside of sl*b. Also note
> -        * that calls to ksize(), where size is not a multiple of machine-word
> -        * size, would otherwise poison the invalid portion of the word.
> -        */
> -       if (is_kfence_address(address))
> -               return;
> -
>         poison_range(address, size, tag);
>
>         if (size & KASAN_GRANULE_MASK) {
> _
>