[PATCHv3] toolchain: granular choice for stack protector

From: Steven Noonan <steven@uplinklabs.net>

Currently, we only support two levels of stach-smashing protection:
  - entirely disabled,
  - protect _all_ functions with -fstack-protector-all.

-fstack-protector-all tends to be far too aggressive and impacts
performance too much to be worth on a real product.

Add a choice that allows us to select between different levels of
stack-smashing protection:
  - none
  - basic   (NEW)
  - strong  (NEW)
  - all

The differences are documented in the GCC online documentation:
    https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Optimize-Options.html

Signed-off-by: Steven Noonan <steven@uplinklabs.net>
[yann.morin.1998@free.fr:
  - rebase
  - add legacy handling
  - SSP-strong depends on gcc >= 4.9
  - slightly simple ifeq-block in package/Makefile.in
  - keep the comment in the choice; add a comment shen strong is not
    available
  - drop the defaults (only keep the legacy)
  - update commit log
]
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

---
Changes v2 -> v3:
  - drop the new defaults, only keep legacy  (Thomas)

Changes v1 -> v2:
  - rebase
  - add legacy handling
  - SSP-strong depends on gcc >= 4.9
  - slightly simple ifeq-block in package/Makefile.in
  - keep the comment in the choice; add a comment shen strong is not
    available
  - update commit log

---
Note: I (Yann) have only slightly tested this patch. More testing is in
order before we can apply this. Steven, care to see if it still fits
your need? Thanks! :-)
---
 Config.in           | 48 +++++++++++++++++++++++++++++++++++++++++++-----
 Config.in.legacy    |  8 ++++++++
 package/Makefile.in |  8 +++++++-
 3 files changed, 58 insertions(+), 6 deletions(-)

On Sun, Dec 27, 2015 at 3:07 AM, Yann E. MORIN <yann.morin.1998@free.fr> wrote:
> From: Steven Noonan <steven@uplinklabs.net>
>
> Currently, we only support two levels of stach-smashing protection:
>   - entirely disabled,
>   - protect _all_ functions with -fstack-protector-all.
>
> -fstack-protector-all tends to be far too aggressive and impacts
> performance too much to be worth on a real product.
>
> Add a choice that allows us to select between different levels of
> stack-smashing protection:
>   - none
>   - basic   (NEW)
>   - strong  (NEW)
>   - all
>
> The differences are documented in the GCC online documentation:
>     https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Optimize-Options.html
>
> Signed-off-by: Steven Noonan <steven@uplinklabs.net>
> [yann.morin.1998@free.fr:
>   - rebase
>   - add legacy handling
>   - SSP-strong depends on gcc >= 4.9
>   - slightly simple ifeq-block in package/Makefile.in
>   - keep the comment in the choice; add a comment shen strong is not
>     available
>   - drop the defaults (only keep the legacy)
>   - update commit log
> ]
> Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
> Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
>
> ---
> Changes v2 -> v3:
>   - drop the new defaults, only keep legacy  (Thomas)
>
> Changes v1 -> v2:
>   - rebase
>   - add legacy handling
>   - SSP-strong depends on gcc >= 4.9
>   - slightly simple ifeq-block in package/Makefile.in
>   - keep the comment in the choice; add a comment shen strong is not
>     available
>   - update commit log
>
> ---
> Note: I (Yann) have only slightly tested this patch. More testing is in
> order before we can apply this. Steven, care to see if it still fits
> your need? Thanks! :-)
> ---
>  Config.in           | 48 +++++++++++++++++++++++++++++++++++++++++++-----
>  Config.in.legacy    |  8 ++++++++
>  package/Makefile.in |  8 +++++++-
>  3 files changed, 58 insertions(+), 6 deletions(-)
>
> diff --git a/Config.in b/Config.in
> index 0be44d9..e892d6d 100644
> --- a/Config.in
> +++ b/Config.in
> @@ -522,12 +522,12 @@ config BR2_GOOGLE_BREAKPAD_INCLUDE_FILES
>
>  endif
>
> -config BR2_ENABLE_SSP
> +choice
>         bool "build code with Stack Smashing Protection"
> -       depends on BR2_TOOLCHAIN_HAS_SSP
> +       default BR2_SSP_ALL if BR2_ENABLE_SSP # legacy

Oh, wait I think I misread this -- I take back my previous comments.
We weren't changing the default of SSP enabled or disabled, but rather
the default SSP type when it's enabled.

When SSP is *enabled* (BR2_ENABLE_SSP) the default should be
BR2_SSP_STRONG (if available). It's generates code that's
better-protected than BR2_SSP_REGULAR, but faster and smaller than
BR2_SSP_ALL.

Only crazy folks would use BR2_SSP_ALL if BR2_SSP_STRONG is an option. ;)

>         help
> -         Enable stack smashing protection support using GCCs
> -         -fstack-protector-all option.
> +         Enable stack smashing protection support using GCC's
> +         -fstack-protector option family.
>
>           See http://www.linuxfromscratch.org/hints/downloads/files/ssp.txt
>           for details.
> @@ -536,9 +536,47 @@ config BR2_ENABLE_SSP
>           support. This is always the case for glibc and eglibc
>           toolchain, but is optional in uClibc toolchains.
>
> -comment "enabling Stack Smashing Protection requires support in the toolchain"
> +config BR2_SSP_NONE
> +       bool "None"
> +       help
> +         Disable stack-smashing protection.
> +
> +comment "Stack Smashing Protection needs a toolchain w/ SSP"
>         depends on !BR2_TOOLCHAIN_HAS_SSP
>
> +config BR2_SSP_REGULAR
> +       bool "-fstack-protector"
> +       depends on BR2_TOOLCHAIN_HAS_SSP
> +       help
> +         Emit extra code to check for buffer overflows, such as stack
> +         smashing attacks. This is done by adding a guard variable to
> +         functions with vulnerable objects. This includes functions
> +         that call alloca, and functions with buffers larger than 8
> +         bytes. The guards are initialized when a function is entered
> +         and then checked when the function exits. If a guard check
> +         fails, an error message is printed and the program exits.
> +
> +config BR2_SSP_STRONG
> +       bool "-fstack-protector-strong"
> +       depends on BR2_TOOLCHAIN_HAS_SSP
> +       depends on BR2_TOOLCHAIN_GCC_AT_LEAST_4_9
> +       help
> +         Like -fstack-protector but includes additional functions to be
> +         protected - those that have local array definitions, or have
> +         references to local frame addresses.
> +
> +comment "Stack Smashing Protection strong needs a toolchain w/ gcc >= 4.9"
> +       depends on BR2_TOOLCHAIN_HAS_SSP
> +       depends on !BR2_TOOLCHAIN_GCC_AT_LEAST_4_9
> +
> +config BR2_SSP_ALL
> +       bool "-fstack-protector-all"
> +       depends on BR2_TOOLCHAIN_HAS_SSP
> +       help
> +         Like -fstack-protector except that all functions are protected.
> +
> +endchoice
> +
>  choice
>         bool "libraries"
>         default BR2_SHARED_LIBS if BR2_BINFMT_SUPPORTS_SHARED
> diff --git a/Config.in.legacy b/Config.in.legacy
> index 2628796..5d45d04 100644
> --- a/Config.in.legacy
> +++ b/Config.in.legacy
> @@ -145,6 +145,14 @@ endif
>  ###############################################################################
>  comment "Legacy options removed in 2016.02"
>
> +# BR2_ENABLE_SSP is still referenced in Config.in (default in choice)
> +config BR2_ENABLE_SSP
> +       bool "Stack Smashing protection now has different levels"
> +       help
> +         The protection offered by SSP can now be selected from different
> +         protection levels. Be sure to review the SSP level in the build
> +         options menu.
> +
>  config BR2_PACKAGE_DIRECTFB_CLE266
>         bool "cle266 driver for directfb removed"
>         select BR2_LEGACY
> diff --git a/package/Makefile.in b/package/Makefile.in
> index 82a66c2..c5652af 100644
> --- a/package/Makefile.in
> +++ b/package/Makefile.in
> @@ -159,7 +159,13 @@ TARGET_CFLAGS += -msep-data
>  TARGET_CXXFLAGS += -msep-data
>  endif
>
> -ifeq ($(BR2_ENABLE_SSP),y)
> +ifeq ($(BR2_SSP_REGULAR),y)
> +TARGET_CFLAGS += -fstack-protector
> +TARGET_CXXFLAGS += -fstack-protector
> +else ifeq ($(BR2_SSP_STRONG),y)
> +TARGET_CFLAGS += -fstack-protector-strong
> +TARGET_CXXFLAGS += -fstack-protector-strong
> +else ifeq ($(BR2_SSP_ALL),y)
>  TARGET_CFLAGS += -fstack-protector-all
>  TARGET_CXXFLAGS += -fstack-protector-all
>  endif
> --
> 1.9.1
>

Steven, All,

On 2015-12-27 03:39 -0800, Steven Noonan spake thusly:
> On Sun, Dec 27, 2015 at 3:07 AM, Yann E. MORIN <yann.morin.1998@free.fr> wrote:
> > From: Steven Noonan <steven@uplinklabs.net>
> >
> > Currently, we only support two levels of stach-smashing protection:
> >   - entirely disabled,
> >   - protect _all_ functions with -fstack-protector-all.
> >
> > -fstack-protector-all tends to be far too aggressive and impacts
> > performance too much to be worth on a real product.
> >
> > Add a choice that allows us to select between different levels of
> > stack-smashing protection:
> >   - none
> >   - basic   (NEW)
> >   - strong  (NEW)
> >   - all
> >
> > The differences are documented in the GCC online documentation:
> >     https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Optimize-Options.html
[--SNIP--]
> > -config BR2_ENABLE_SSP
> > +choice
> >         bool "build code with Stack Smashing Protection"
> > -       depends on BR2_TOOLCHAIN_HAS_SSP
> > +       default BR2_SSP_ALL if BR2_ENABLE_SSP # legacy
> 
> Oh, wait I think I misread this -- I take back my previous comments.
> We weren't changing the default of SSP enabled or disabled, but rather
> the default SSP type when it's enabled.
> 
> When SSP is *enabled* (BR2_ENABLE_SSP) the default should be
> BR2_SSP_STRONG (if available). It's generates code that's
> better-protected than BR2_SSP_REGULAR, but faster and smaller than
> BR2_SSP_ALL.
> 
> Only crazy folks would use BR2_SSP_ALL if BR2_SSP_STRONG is an option. ;)

No, we want the legacy BR2_ENABLE_SSP symbol to set the same default as
it previously represented.

Currently, BR2_ENABLE_SSP meant 'ssp-all' so we want to keep that
behaviour.

Regards,
Yann E. MORIN.

> >         help
> > -         Enable stack smashing protection support using GCCs
> > -         -fstack-protector-all option.
> > +         Enable stack smashing protection support using GCC's
> > +         -fstack-protector option family.
> >
> >           See http://www.linuxfromscratch.org/hints/downloads/files/ssp.txt
> >           for details.
> > @@ -536,9 +536,47 @@ config BR2_ENABLE_SSP
> >           support. This is always the case for glibc and eglibc
> >           toolchain, but is optional in uClibc toolchains.
> >
> > -comment "enabling Stack Smashing Protection requires support in the toolchain"
> > +config BR2_SSP_NONE
> > +       bool "None"
> > +       help
> > +         Disable stack-smashing protection.
> > +
> > +comment "Stack Smashing Protection needs a toolchain w/ SSP"
> >         depends on !BR2_TOOLCHAIN_HAS_SSP
> >
> > +config BR2_SSP_REGULAR
> > +       bool "-fstack-protector"
> > +       depends on BR2_TOOLCHAIN_HAS_SSP
> > +       help
> > +         Emit extra code to check for buffer overflows, such as stack
> > +         smashing attacks. This is done by adding a guard variable to
> > +         functions with vulnerable objects. This includes functions
> > +         that call alloca, and functions with buffers larger than 8
> > +         bytes. The guards are initialized when a function is entered
> > +         and then checked when the function exits. If a guard check
> > +         fails, an error message is printed and the program exits.
> > +
> > +config BR2_SSP_STRONG
> > +       bool "-fstack-protector-strong"
> > +       depends on BR2_TOOLCHAIN_HAS_SSP
> > +       depends on BR2_TOOLCHAIN_GCC_AT_LEAST_4_9
> > +       help
> > +         Like -fstack-protector but includes additional functions to be
> > +         protected - those that have local array definitions, or have
> > +         references to local frame addresses.
> > +
> > +comment "Stack Smashing Protection strong needs a toolchain w/ gcc >= 4.9"
> > +       depends on BR2_TOOLCHAIN_HAS_SSP
> > +       depends on !BR2_TOOLCHAIN_GCC_AT_LEAST_4_9
> > +
> > +config BR2_SSP_ALL
> > +       bool "-fstack-protector-all"
> > +       depends on BR2_TOOLCHAIN_HAS_SSP
> > +       help
> > +         Like -fstack-protector except that all functions are protected.
> > +
> > +endchoice
> > +
> >  choice
> >         bool "libraries"
> >         default BR2_SHARED_LIBS if BR2_BINFMT_SUPPORTS_SHARED
> > diff --git a/Config.in.legacy b/Config.in.legacy
> > index 2628796..5d45d04 100644
> > --- a/Config.in.legacy
> > +++ b/Config.in.legacy
> > @@ -145,6 +145,14 @@ endif
> >  ###############################################################################
> >  comment "Legacy options removed in 2016.02"
> >
> > +# BR2_ENABLE_SSP is still referenced in Config.in (default in choice)
> > +config BR2_ENABLE_SSP
> > +       bool "Stack Smashing protection now has different levels"
> > +       help
> > +         The protection offered by SSP can now be selected from different
> > +         protection levels. Be sure to review the SSP level in the build
> > +         options menu.
> > +
> >  config BR2_PACKAGE_DIRECTFB_CLE266
> >         bool "cle266 driver for directfb removed"
> >         select BR2_LEGACY
> > diff --git a/package/Makefile.in b/package/Makefile.in
> > index 82a66c2..c5652af 100644
> > --- a/package/Makefile.in
> > +++ b/package/Makefile.in
> > @@ -159,7 +159,13 @@ TARGET_CFLAGS += -msep-data
> >  TARGET_CXXFLAGS += -msep-data
> >  endif
> >
> > -ifeq ($(BR2_ENABLE_SSP),y)
> > +ifeq ($(BR2_SSP_REGULAR),y)
> > +TARGET_CFLAGS += -fstack-protector
> > +TARGET_CXXFLAGS += -fstack-protector
> > +else ifeq ($(BR2_SSP_STRONG),y)
> > +TARGET_CFLAGS += -fstack-protector-strong
> > +TARGET_CXXFLAGS += -fstack-protector-strong
> > +else ifeq ($(BR2_SSP_ALL),y)
> >  TARGET_CFLAGS += -fstack-protector-all
> >  TARGET_CXXFLAGS += -fstack-protector-all
> >  endif
> > --
> > 1.9.1
> >

Steven,

On Sun, 27 Dec 2015 03:39:19 -0800, Steven Noonan wrote:

> When SSP is *enabled* (BR2_ENABLE_SSP) the default should be
> BR2_SSP_STRONG (if available). It's generates code that's
> better-protected than BR2_SSP_REGULAR, but faster and smaller than
> BR2_SSP_ALL.
> 
> Only crazy folks would use BR2_SSP_ALL if BR2_SSP_STRONG is an option. ;)

We want to preserve existing behavior as much as possible. So people
who enabled BR2_ENABLE_SSP were paying the price of BR2_SSP_ALL, and we
should therefore keep using BR2_SSP_ALL for such users. That's the
point of legacy handling: minimizing the amount of "surprise" /changes
for users upgrading Buildroot.

Best regards,

Thomas

On Sun, Dec 27, 2015 at 3:45 AM, Thomas Petazzoni
<thomas.petazzoni@free-electrons.com> wrote:
> Steven,
>
> On Sun, 27 Dec 2015 03:39:19 -0800, Steven Noonan wrote:
>
>> When SSP is *enabled* (BR2_ENABLE_SSP) the default should be
>> BR2_SSP_STRONG (if available). It's generates code that's
>> better-protected than BR2_SSP_REGULAR, but faster and smaller than
>> BR2_SSP_ALL.
>>
>> Only crazy folks would use BR2_SSP_ALL if BR2_SSP_STRONG is an option. ;)
>
> We want to preserve existing behavior as much as possible. So people
> who enabled BR2_ENABLE_SSP were paying the price of BR2_SSP_ALL, and we
> should therefore keep using BR2_SSP_ALL for such users. That's the
> point of legacy handling: minimizing the amount of "surprise" /changes
> for users upgrading Buildroot.
>

I get the argument, but it -is- a solvable problem: you could make
BR2_ENABLE_SSP "legacy", force them to explicitly choose an SSP
variant. But that would be a separate patch.

Dear Yann E. MORIN,

On Sun, 27 Dec 2015 12:07:31 +0100, Yann E. MORIN wrote:
> From: Steven Noonan <steven@uplinklabs.net>
> 
> Currently, we only support two levels of stach-smashing protection:
>   - entirely disabled,
>   - protect _all_ functions with -fstack-protector-all.
> 
> -fstack-protector-all tends to be far too aggressive and impacts
> performance too much to be worth on a real product.
> 
> Add a choice that allows us to select between different levels of
> stack-smashing protection:
>   - none
>   - basic   (NEW)
>   - strong  (NEW)
>   - all
> 
> The differences are documented in the GCC online documentation:
>     https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Optimize-Options.html
> 
> Signed-off-by: Steven Noonan <steven@uplinklabs.net>
> [yann.morin.1998@free.fr:
>   - rebase
>   - add legacy handling
>   - SSP-strong depends on gcc >= 4.9
>   - slightly simple ifeq-block in package/Makefile.in
>   - keep the comment in the choice; add a comment shen strong is not
>     available
>   - drop the defaults (only keep the legacy)
>   - update commit log
> ]
> Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
> Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
> 
> ---
> Changes v2 -> v3:
>   - drop the new defaults, only keep legacy  (Thomas)

Applied with the following changes:

    [Thomas:
     - only show the choice if the toolchain has SSP support
     - add details for the BR2_SSP_ALL option that it has a significant
       performance impact.]

Thanks!

Thomas