[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CAG_fn=WmEFOPjPBB86o+7PCwuBD6qOic75D7TvKo+MrGozFtKg@mail.gmail.com>
Date: Fri, 8 Mar 2019 11:09:06 +0100
From: Alexander Potapenko <glider@...gle.com>
To: Kees Cook <keescook@...omium.org>
Cc: Masahiro Yamada <yamada.masahiro@...ionext.com>, James Morris <jmorris@...ei.org>,
"Serge E. Hallyn" <serge@...lyn.com>,
linux-security-module <linux-security-module@...r.kernel.org>,
linux-kbuild <linux-kbuild@...r.kernel.org>, Nick Desaulniers <ndesaulniers@...gle.com>,
Kostya Serebryany <kcc@...gle.com>, Dmitry Vyukov <dvyukov@...gle.com>, Sandeep Patil <sspatil@...roid.com>,
Kernel Hardening <kernel-hardening@...ts.openwall.com>
Subject: Re: [PATCH 1/1] RFC: initmem: introduce CONFIG_INIT_ALL_MEMORY and CONFIG_INIT_ALL_STACK
On Thu, Mar 7, 2019 at 6:12 PM Kees Cook <keescook@...omium.org> wrote:
>
> On Thu, Mar 7, 2019 at 5:35 AM Alexander Potapenko <glider@...gle.com> wrote:
> >
> > This patch is a part of a bigger initiative to allow initializing
> > heap/stack memory in the Linux kernels by default.
> > The rationale behind doing so is to reduce the severity of bugs caused
> > by using uninitialized memory.
> >
> > CONFIG_INIT_ALL_MEMORY is going to be an umbrella config for options
> > that force heap and stack initialization.
>
> We have some level of control over the heap portions already with the
> poisoning options, but they currently require command-line enabling.
>
> For the slab allocator:
> CONFIG_SLUB_DEBUG=y
> and boot with "slub_debug=P"
>
> For the page allocator:
> CONFIG_PAGE_POISONING=y
> CONFIG_PAGE_POISONING_NO_SANITY=y
> CONFIG_PAGE_POISONING_ZERO=y
> and boot with "page_poison=1"
>
> So I think it would make sense to wire up the boot defaults and
> continue some of the benchmarking work to improve the performance hit.
Should we care about the poison values being different for stack and
heap initialization (probably no)?
> > CONFIG_INIT_ALL_STACK turns on stack initialization based on the
> > -ftrivial-auto-var-init Clang flag.
> >
> > -ftrivial-auto-var-init is a Clang flag that provides trivial
> > initializers for uninitialized local variables, variable fields and
> > padding.
> >
> > It has three possible values:
> > pattern - uninitialized locals are filled with a fixed pattern
> > (mostly 0xAA on 64-bit platforms, see https://reviews.llvm.org/D54604
> > for more details) likely to cause crashes when uninitialized value is
> > used;
> > zero (it's still debated whether this flag makes it to the official
> > Clang release) - uninitialized locals are filled with zeroes;
> > uninitialized (default) - uninitialized locals are left intact.
>
> This should get wired up to CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF_ALL in
> the GCC case.
I'm not really familiar with CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF_ALL.
Does it initialize a variable if it's not explicitly address-taken,
but is used in a computation of another variable's value?
Also, are we ok with Clang and GCC producing different poison values
for the stack? (probably yes)
> > The proposed config builds the kernel with
> > -ftrivial-auto-var-init=pattern.
> >
> > Developers have the possibility to opt-out of this feature on a
> > per-file (by using the INIT_ALL_MEMORY_ Makefile prefix) or per-variable
> > (by using __attribute__((uninitialized))) basis.
>
> I would like a clean way to offer a paranoid "true init all memory"
> config that will ignore the attribute.
How paranoid should it be? E.g. current instrumentation in Clang (nor
in GCC, I guess) doesn't prevent the following situation:
char local[8];
int init_later;
copy_to_user(userp, local, 12);
init_later = 1;
Do we want to true init |init_later| despite there's a write to it later?
Regarding the attribute, right now it only makes sense to use it in
the cases the compilers fail to apply DSE to the unneeded
initializers.
For example, Clang fails to do so for |oldbit| in __test_and_set_bit()
here: https://elixir.bootlin.com/linux/latest/source/arch/x86/include/asm/bitops.h#L245
(We could fix this with a heuristic to not initialize scalars used as
assembly outputs, but that's walking on thin ice)
Do you know how GCC handles this situation?
> Also, BYREF_ALL needs to gain
> the attribute knowledge.
>
> > Signed-off-by: Alexander Potapenko <glider@...gle.com>
> > Cc: Masahiro Yamada <yamada.masahiro@...ionext.com>
> > Cc: James Morris <jmorris@...ei.org>
> > Cc: "Serge E. Hallyn" <serge@...lyn.com>
> > Cc: Nick Desaulniers <ndesaulniers@...gle.com>
> > Cc: Kostya Serebryany <kcc@...gle.com>
> > Cc: Dmitry Vyukov <dvyukov@...gle.com>
> > Cc: Kees Cook <keescook@...omium.org>
> > Cc: Sandeep Patil <sspatil@...roid.com>
> > Cc: linux-security-module@...r.kernel.org
> > Cc: linux-kbuild@...r.kernel.org
> >
> > ---
> > Makefile | 3 ++-
> > scripts/Makefile.initmem | 17 +++++++++++++++++
> > scripts/Makefile.lib | 6 ++++++
> > security/Kconfig | 1 +
> > security/Kconfig.initmem | 22 ++++++++++++++++++++++
> > 5 files changed, 48 insertions(+), 1 deletion(-)
> > create mode 100644 scripts/Makefile.initmem
> > create mode 100644 security/Kconfig.initmem
> >
> > diff --git a/Makefile b/Makefile
> > index f070e0d65186..028ca37878fd 100644
> > --- a/Makefile
> > +++ b/Makefile
> > @@ -448,7 +448,7 @@ export HOSTCXX KBUILD_HOSTCXXFLAGS LDFLAGS_MODULE CHECK CHECKFLAGS
> >
> > export KBUILD_CPPFLAGS NOSTDINC_FLAGS LINUXINCLUDE OBJCOPYFLAGS KBUILD_LDFLAGS
> > export KBUILD_CFLAGS CFLAGS_KERNEL CFLAGS_MODULE
> > -export CFLAGS_KASAN CFLAGS_KASAN_NOSANITIZE CFLAGS_UBSAN
> > +export CFLAGS_KASAN CFLAGS_KASAN_NOSANITIZE CFLAGS_UBSAN CFLAGS_INITMEM
> > export KBUILD_AFLAGS AFLAGS_KERNEL AFLAGS_MODULE
> > export KBUILD_AFLAGS_MODULE KBUILD_CFLAGS_MODULE KBUILD_LDFLAGS_MODULE
> > export KBUILD_AFLAGS_KERNEL KBUILD_CFLAGS_KERNEL
> > @@ -840,6 +840,7 @@ KBUILD_ARFLAGS := $(call ar-option,D)
> > include scripts/Makefile.kasan
> > include scripts/Makefile.extrawarn
> > include scripts/Makefile.ubsan
> > +include scripts/Makefile.initmem
> >
> > # Add any arch overrides and user supplied CPPFLAGS, AFLAGS and CFLAGS as the
> > # last assignments
> > diff --git a/scripts/Makefile.initmem b/scripts/Makefile.initmem
> > new file mode 100644
> > index 000000000000..f49be398f2c1
> > --- /dev/null
> > +++ b/scripts/Makefile.initmem
> > @@ -0,0 +1,17 @@
> > +ifdef CONFIG_INIT_ALL_MEMORY
> > +
> > +# Clang's -ftrivial-auto-var-init=pattern flag initializes the
> > +# uninitialized parts of local variables (including fields and padding)
> > +# with a fixed pattern (0xAA in most cases).
> > +ifdef CONFIG_INIT_ALL_STACK
> > + CFLAGS_INITMEM := -ftrivial-auto-var-init=pattern
> > +endif
> > +
> > +ifeq ($(call cc-option, $(CFLAGS_INITMEM) -Werror),)
> > + ifneq ($(CONFIG_COMPILE_TEST),y)
> > + $(warning Cannot use CONFIG_INIT_ALL_MEMORY: \
> > + -ftrivial-auto-var-init is not supported by compiler)
> > + endif
> > +endif
>
> This should be done in the Kconfig (see how stack protector is
> handled), rather than the Makefile. See below.
>
> > +
> > +endif
> > diff --git a/scripts/Makefile.lib b/scripts/Makefile.lib
> > index 12b88d09c3a4..53d18fd15c79 100644
> > --- a/scripts/Makefile.lib
> > +++ b/scripts/Makefile.lib
> > @@ -131,6 +131,12 @@ _c_flags += $(if $(patsubst n%,, \
> > $(CFLAGS_UBSAN))
> > endif
> >
> > +ifeq ($(CONFIG_INIT_ALL_MEMORY),y)
> > +_c_flags += $(if $(patsubst n%,, \
> > + $(INIT_ALL_MEMORY_$(basetarget).o)$(INIT_ALL_MEMORY)y), \
> > + $(CFLAGS_INITMEM))
> > +endif
> > +
> > ifeq ($(CONFIG_KCOV),y)
> > _c_flags += $(if $(patsubst n%,, \
> > $(KCOV_INSTRUMENT_$(basetarget).o)$(KCOV_INSTRUMENT)$(CONFIG_KCOV_INSTRUMENT_ALL)), \
> > diff --git a/security/Kconfig b/security/Kconfig
> > index e4fe2f3c2c65..cc12a39424dd 100644
> > --- a/security/Kconfig
> > +++ b/security/Kconfig
> > @@ -230,6 +230,7 @@ config STATIC_USERMODEHELPER_PATH
> > If you wish for all usermode helper programs to be disabled,
> > specify an empty string here (i.e. "").
> >
> > +source "security/Kconfig.initmem"
> > source "security/selinux/Kconfig"
> > source "security/smack/Kconfig"
> > source "security/tomoyo/Kconfig"
> > diff --git a/security/Kconfig.initmem b/security/Kconfig.initmem
> > new file mode 100644
> > index 000000000000..5ac3cf3e7f88
> > --- /dev/null
> > +++ b/security/Kconfig.initmem
> > @@ -0,0 +1,22 @@
> > +menu "Initialize all memory"
> > +
> > +config INIT_ALL_MEMORY
> > + bool "Initialize all memory"
> > + default n
> > + help
> > + Enforce memory initialization to mitigate infoleaks and make
> > + the control-flow bugs depending on uninitialized values more
> > + deterministic.
>
> See my notes about enabling page/slab poisoning via this option (which
> should probably be a separate HEAP option).
>
> > +
> > +if INIT_ALL_MEMORY
> > +
> > +config INIT_ALL_STACK
> > + bool "Initialize all stack"
> > + depends on INIT_ALL_MEMORY
>
> For example of doing this in Kconfig:
>
> config CC_HAS_AUTO_VAR_INIT
> def_bool $(cc-option,-ftrivial-auto-var-init=pattern)
>
> config INIT_ALL_STACK
> bool "Initialize all stack"
> depends on CC_HAS_AUTO_VAR_INIT || HAVE_GCC_PLUGINS
> select GCC_PLUGINS if !CC_HAS_AUTO_VAR_INIT
> select GCC_PLUGIN_STRUCTLEAK if !CC_HAS_AUTO_VAR_INIT
> select GCC_PLUGIN_STRUCTLEAK_BYREF_ALL if !CC_HAS_AUTO_VAR_INIT
>
> probably GCC_PLUGINS needs to be "default y" so this select mess is
> less crazy. I already started work on reorganizing the Kconfig there:
> https://git.kernel.org/pub/scm/linux/kernel/git/kees/linux.git/commit/?h=kspp/gcc-plugin/structleak&id=81a56f6dcd20325607d6008f4bb560c96f4c821a
>
> > + default y
> > + help
> > + Initialize uninitialized stack data with a 0xAA pattern.
> > + This config option only supports Clang builds at the moment.
> > +
> > +endif # INIT_ALL_MEMORY
> > +endmenu
> > --
> > 2.21.0.352.gf09ad66450-goog
> >
>
> --
> Kees Cook
--
Alexander Potapenko
Software Engineer
Google Germany GmbH
Erika-Mann-Straße, 33
80636 München
Geschäftsführer: Paul Manicle, Halimah DeLaine Prado
Registergericht und -nummer: Hamburg, HRB 86891
Sitz der Gesellschaft: Hamburg
Powered by blists - more mailing lists
Confused about mailing lists and their use?
Read about mailing lists on Wikipedia
and check out these
guidelines on proper formatting of your messages.
