[RFC] x86/entry/64: randomize kernel stack offset upon syscall
diff mbox series

Message ID 20190329081358.30497-1-elena.reshetova@intel.com
State New
Headers show
Series
  • [RFC] x86/entry/64: randomize kernel stack offset upon syscall
Related show

Commit Message

Reshetova, Elena March 29, 2019, 8:13 a.m. UTC
If CONFIG_RANDOMIZE_KSTACK_OFFSET is selected,
the kernel stack offset is randomized upon each
entry to a system call after fixed location of pt_regs
struct.

This feature is based on the original idea from
the PaX's RANDKSTACK feature:
https://pax.grsecurity.net/docs/randkstack.txt
All the credits for the original idea goes to the PaX team.
However, the design and implementation of
RANDOMIZE_KSTACK_OFFSET differs greatly from the RANDKSTACK
feature (see below).

Reasoning for the feature:

This feature aims to make considerably harder various
stack-based attacks that rely on deterministic stack
structure.
We have had many of such attacks in past [1],[2],[3]
(just to name few), and as Linux kernel stack protections
have been constantly improving (vmap-based stack
allocation with guard pages, removal of thread_info,
STACKLEAK), attackers have to find new ways for their
exploits to work.

It is important to note that we currently cannot show
a concrete attack that would be stopped by this new
feature (given that other existing stack protections
are enabled), so this is an attempt to be on a proactive
side vs. catching up with existing successful exploits.

The main idea is that since the stack offset is
randomized upon each system call, it is very hard for
attacker to reliably land in any particular place on
the thread stack when attack is performed.
Also, since randomization is performed *after* pt_regs,
the ptrace-based approach to discover randomization
offset during a long-running syscall should not be
possible.

[1] jon.oberheide.org/files/infiltrate12-thestackisback.pdf
[2] jon.oberheide.org/files/stackjacking-infiltrate11.pdf
[3] googleprojectzero.blogspot.com/2016/06/exploiting-
recursion-in-linux-kernel_20.html

Design description:

During most of the kernel's execution, it runs on the "thread
stack", which is allocated at fork.c/dup_task_struct() and stored in
a per-task variable (tsk->stack). Since stack is growing downward,
the stack top can be always calculated using task_top_of_stack(tsk)
function, which essentially returns an address of tsk->stack + stack
size. When VMAP_STACK is enabled, the thread stack is allocated from
vmalloc space.

Thread stack is pretty deterministic on its structure - fixed in size,
and upon every entry from a userspace to kernel on a
syscall the thread stack is started to be constructed from an
address fetched from a per-cpu cpu_current_top_of_stack variable.
The first element to be pushed to the thread stack is the pt_regs struct
that stores all required CPU registers and sys call parameters.

The goal of RANDOMIZE_KSTACK_OFFSET feature is to add a random offset
after the pt_regs has been pushed to the stack and the rest of thread
stack (used during the syscall processing) every time a process issues
a syscall. The source of randomness can be taken either from prandom_u32()
pseudo random generator (not cryptographically secure). The offset is
added using alloca() call since it helps avoiding changes in assembly
syscall entry code and unwinder. I am not that greatly happy about the
generated assembly code (but I don't know if how to force gcc to produce
anything better):

...
	size_t offset = ((size_t)prandom_u32()) % 256;
	char * ptr = alloca(offset);
0xffffffff8100426d add    $0x16,%rax
0xffffffff81004271 and    $0x1f8,%eax
0xffffffff81004276 sub    %rax,%rsp
0xffffffff81004279 lea    0xf(%rsp),%rax
0xffffffff8100427e and    $0xfffffffffffffff0,%rax
	asm volatile("":"=m"(*ptr));
...

As a result of the above gcc-produce code this patch introduces 6 bits
of randomness (bits 3 - 8 are randomized, bits 0-2 are zero due to stack alignment)
after pt_regs location on the thread stack.
The amount of randomness can be adjusted based on how much of the
stack space we wish/can trade for security.

Performance:

1) lmbench: ./lat_syscall -N 1000000 null
    base:                                        Simple syscall: 0.1774 microseconds
    random_offset (prandom_u32() every syscall): Simple syscall: 0.1822 microseconds

2)  Andy's tests, misc-tests: ./timing_test_64 10M sys_enosys
    base:                                        10000000 loops in 1.62224s = 162.22 nsec / loop
    random_offset (prandom_u32() every syscall): 10000000 loops in 1.64660s = 166.26 nsec / loop

Comparison to grsecurity RANDKSTACK feature:

RANDKSTACK feature randomizes the location of the stack start
(cpu_current_top_of_stack), i.e. location of pt_regs structure
itself on the stack. Initially this patch followed the same approach,
but during the recent discussions [4], it has been determined
to be of a little value since, if ptrace functionality is available
for an attacker, he can use PTRACE_PEEKUSR/PTRACE_POKEUSR api to read/write
different offsets in the pt_regs struct, observe the cache
behavior of the pt_regs accesses, and figure out the random stack offset.

Another big difference is that randomization is done upon
syscall entry and not the exit, as with RANDKSTACK.

Also, as a result of the above two differences, the implementation
of RANDKSTACK and RANDOMIZE_KSTACK_OFFSET has nothing in common.

[4] https://www.openwall.com/lists/kernel-hardening/2019/02/08/6

Signed-off-by: Elena Reshetova <elena.reshetova@intel.com>
---
 arch/Kconfig            | 15 +++++++++++++++
 arch/x86/Kconfig        |  1 +
 arch/x86/entry/common.c | 16 ++++++++++++++++
 3 files changed, 32 insertions(+)

Comments

Kees Cook April 3, 2019, 9:17 p.m. UTC | #1
On Fri, Mar 29, 2019 at 1:14 AM Elena Reshetova
<elena.reshetova@intel.com> wrote:
> diff --git a/arch/x86/entry/common.c b/arch/x86/entry/common.c
> index 7bc105f47d21..28cb3687bf82 100644
> --- a/arch/x86/entry/common.c
> +++ b/arch/x86/entry/common.c
> @@ -32,6 +32,10 @@
>  #include <linux/uaccess.h>
>  #include <asm/cpufeature.h>
>
> +#ifdef CONFIG_RANDOMIZE_KSTACK_OFFSET
> +#include <linux/random.h>
> +#endif
> +
>  #define CREATE_TRACE_POINTS
>  #include <trace/events/syscalls.h>
>
> @@ -269,10 +273,22 @@ __visible inline void syscall_return_slowpath(struct pt_regs *regs)
>  }
>
>  #ifdef CONFIG_X86_64
> +
> +#ifdef CONFIG_RANDOMIZE_KSTACK_OFFSET
> +void *alloca(size_t size);
> +#endif
> +
>  __visible void do_syscall_64(unsigned long nr, struct pt_regs *regs)
>  {
>         struct thread_info *ti;
>
> +#ifdef CONFIG_RANDOMIZE_KSTACK_OFFSET
> +       size_t offset = ((size_t)prandom_u32()) % 256;
> +       char *ptr = alloca(offset);
> +
> +       asm volatile("":"=m"(*ptr));
> +#endif
> +
>         enter_from_user_mode();
>         local_irq_enable();
>         ti = current_thread_info();

Well this is delightfully short! The alloca() definition could even be
moved up after the #include of random.h, just to reduce the number of
#ifdef lines, too. I patched getpid() to report stack locations for a
given pid, just to get a sense of the entropy. On 10,000 getpid()
calls I see counts like:

    229  ffffa58240697dbc
    294  ffffa58240697dc4
    315  ffffa58240697dcc
    298  ffffa58240697dd4
    335  ffffa58240697ddc
    311  ffffa58240697de4
    295  ffffa58240697dec
    303  ffffa58240697df4
    334  ffffa58240697dfc
    331  ffffa58240697e04
    321  ffffa58240697e0c
    298  ffffa58240697e14
    290  ffffa58240697e1c
    306  ffffa58240697e24
    308  ffffa58240697e2c
    325  ffffa58240697e34
    301  ffffa58240697e3c
    336  ffffa58240697e44
    328  ffffa58240697e4c
    326  ffffa58240697e54
    314  ffffa58240697e5c
    305  ffffa58240697e64
    315  ffffa58240697e6c
    325  ffffa58240697e74
    287  ffffa58240697e7c
    319  ffffa58240697e84
    309  ffffa58240697e8c
    329  ffffa58240697e94
    311  ffffa58240697e9c
    306  ffffa58240697ea4
    313  ffffa58240697eac
    289  ffffa58240697eb4
     94  ffffa58240697ebc

So it looks more like 5 bits of entropy in practice (here are 33
unique stack locations), but that still looks good to me.

Can you send the next version with a CC to lkml too?

Andy, Thomas, how does this look to you?
Reshetova, Elena April 4, 2019, 11:41 a.m. UTC | #2
On Fri, Mar 29, 2019 at 1:14 AM Elena Reshetova
> <elena.reshetova@intel.com> wrote:
> > diff --git a/arch/x86/entry/common.c b/arch/x86/entry/common.c
> > index 7bc105f47d21..28cb3687bf82 100644
> > --- a/arch/x86/entry/common.c
> > +++ b/arch/x86/entry/common.c
> > @@ -32,6 +32,10 @@
> >  #include <linux/uaccess.h>
> >  #include <asm/cpufeature.h>
> >
> > +#ifdef CONFIG_RANDOMIZE_KSTACK_OFFSET
> > +#include <linux/random.h>
> > +#endif
> > +
> >  #define CREATE_TRACE_POINTS
> >  #include <trace/events/syscalls.h>
> >
> > @@ -269,10 +273,22 @@ __visible inline void syscall_return_slowpath(struct
> pt_regs *regs)
> >  }
> >
> >  #ifdef CONFIG_X86_64
> > +
> > +#ifdef CONFIG_RANDOMIZE_KSTACK_OFFSET
> > +void *alloca(size_t size);
> > +#endif
> > +
> >  __visible void do_syscall_64(unsigned long nr, struct pt_regs *regs)
> >  {
> >         struct thread_info *ti;
> >
> > +#ifdef CONFIG_RANDOMIZE_KSTACK_OFFSET
> > +       size_t offset = ((size_t)prandom_u32()) % 256;
> > +       char *ptr = alloca(offset);
> > +
> > +       asm volatile("":"=m"(*ptr));
> > +#endif
> > +
> >         enter_from_user_mode();
> >         local_irq_enable();
> >         ti = current_thread_info();
> 
> Well this is delightfully short! 

Yes :) Looks like when you are allowed to use forbidden APIs, life might be 
suddenly much easier :) 

The alloca() definition could even be
> moved up after the #include of random.h, just to reduce the number of
> #ifdef lines, too.

Sure, can do this. 

 I patched getpid() to report stack locations for a
> given pid, just to get a sense of the entropy. On 10,000 getpid()
> calls I see counts like:
> 
>     229  ffffa58240697dbc
>     294  ffffa58240697dc4
>     315  ffffa58240697dcc
>     298  ffffa58240697dd4
>     335  ffffa58240697ddc
>     311  ffffa58240697de4
>     295  ffffa58240697dec
>     303  ffffa58240697df4
>     334  ffffa58240697dfc
>     331  ffffa58240697e04
>     321  ffffa58240697e0c
>     298  ffffa58240697e14
>     290  ffffa58240697e1c
>     306  ffffa58240697e24
>     308  ffffa58240697e2c
>     325  ffffa58240697e34
>     301  ffffa58240697e3c
>     336  ffffa58240697e44
>     328  ffffa58240697e4c
>     326  ffffa58240697e54
>     314  ffffa58240697e5c
>     305  ffffa58240697e64
>     315  ffffa58240697e6c
>     325  ffffa58240697e74
>     287  ffffa58240697e7c
>     319  ffffa58240697e84
>     309  ffffa58240697e8c
>     329  ffffa58240697e94
>     311  ffffa58240697e9c
>     306  ffffa58240697ea4
>     313  ffffa58240697eac
>     289  ffffa58240697eb4
>      94  ffffa58240697ebc
> 
> So it looks more like 5 bits of entropy in practice (here are 33
> unique stack locations), but that still looks good to me.

What I still don't fully understand here (due to my little knowledge of
compilers) and afraid of is that the asm code that alloca generates (see my version)
 and the alignment might differ on the different targets, etc. 
If you tried it on yours, can you send me the asm code that it produced for you?
Is it different from mine? 

> 
> Can you send the next version with a CC to lkml too?

I was thinking on not spamming lkml before we get some agreement here, but
I can do it if people believe this is the right way. 

Getting Andy's feedback on this version first would be great! 

Best Regards,
Elena.
Kees Cook April 4, 2019, 5:03 p.m. UTC | #3
On Thu, Apr 4, 2019 at 4:41 AM Reshetova, Elena
<elena.reshetova@intel.com> wrote:
> What I still don't fully understand here (due to my little knowledge of
> compilers) and afraid of is that the asm code that alloca generates (see my version)
> and the alignment might differ on the different targets, etc.

I guess it's possible, but for x86_64, since appears to be consistent.

> If you tried it on yours, can you send me the asm code that it produced for you?
> Is it different from mine?

You can compare compiler outputs here. Here's gcc vs clang for this code:
https://godbolt.org/z/WJSbN8
You can adjust compiler versions, etc.
Reshetova, Elena April 5, 2019, 10:14 a.m. UTC | #4
> On Thu, Apr 4, 2019 at 4:41 AM Reshetova, Elena
> <elena.reshetova@intel.com> wrote:
> > What I still don't fully understand here (due to my little knowledge of
> > compilers) and afraid of is that the asm code that alloca generates (see my version)
> > and the alignment might differ on the different targets, etc.
> 
> I guess it's possible, but for x86_64, since appears to be consistent.

So, yes, I double checked this now with just printing all possible offsets I get for rsp
from do_syscall_64, it is indeed 33 different offsets, so it is indeed more like 5 bits of entropy. 
We can increase it, if we want and people are ok with losing a bit more stack space. 
 
> 
> > If you tried it on yours, can you send me the asm code that it produced for you?
> > Is it different from mine?
> 
> You can compare compiler outputs here. Here's gcc vs clang for this code:
> https://godbolt.org/z/WJSbN8
> You can adjust compiler versions, etc.

Oh, this is handy! Thank you for the link! 


So, should I resend to lkml (with some cosmetic fixes) or how to proceed with this?
I will also update the randomness bit info. 

Best Regards,
Elena.
Andy Lutomirski April 5, 2019, 1:14 p.m. UTC | #5
On Apr 5, 2019, at 4:14 AM, Reshetova, Elena <elena.reshetova@intel.com> wrote:

>> On Thu, Apr 4, 2019 at 4:41 AM Reshetova, Elena
>> <elena.reshetova@intel.com> wrote:
>>> What I still don't fully understand here (due to my little knowledge of
>>> compilers) and afraid of is that the asm code that alloca generates (see my version)
>>> and the alignment might differ on the different targets, etc.
>> 
>> I guess it's possible, but for x86_64, since appears to be consistent.
> 
> So, yes, I double checked this now with just printing all possible offsets I get for rsp
> from do_syscall_64, it is indeed 33 different offsets, so it is indeed more like 5 bits of entropy. 
> We can increase it, if we want and people are ok with losing a bit more stack space. 
> 
>> 
>>> If you tried it on yours, can you send me the asm code that it produced for you?
>>> Is it different from mine?
>> 
>> You can compare compiler outputs here. Here's gcc vs clang for this code:
>> https://godbolt.org/z/WJSbN8
>> You can adjust compiler versions, etc.
> 
> Oh, this is handy! Thank you for the link! 
> 
> 
> So, should I resend to lkml (with some cosmetic fixes) or how to proceed with this?
> I will also update the randomness bit info. 
> 
> 

Go ahead and send a new version, please.

Patch
diff mbox series

diff --git a/arch/Kconfig b/arch/Kconfig
index 4cfb6de48f79..9a2557b0cfce 100644
--- a/arch/Kconfig
+++ b/arch/Kconfig
@@ -808,6 +808,21 @@  config VMAP_STACK
 	  the stack to map directly to the KASAN shadow map using a formula
 	  that is incorrect if the stack is in vmalloc space.
 
+config HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET
+	def_bool n
+	help
+	  An arch should select this symbol if it can support kernel stack
+	  offset randomization.
+
+config RANDOMIZE_KSTACK_OFFSET
+	default n
+	bool "Randomize kernel stack offset on syscall entry"
+	depends on HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET
+	help
+	  Enable this if you want the randomize kernel stack offset upon
+	  each syscall entry. This causes kernel stack (after pt_regs) to
+	  have a randomized offset upon executing each system call.
+
 config ARCH_OPTIONAL_KERNEL_RWX
 	def_bool n
 
diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
index ade12ec4224b..5edcae945b73 100644
--- a/arch/x86/Kconfig
+++ b/arch/x86/Kconfig
@@ -131,6 +131,7 @@  config X86
 	select HAVE_ARCH_TRANSPARENT_HUGEPAGE
 	select HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD if X86_64
 	select HAVE_ARCH_VMAP_STACK		if X86_64
+	select HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET  if X86_64
 	select HAVE_ARCH_WITHIN_STACK_FRAMES
 	select HAVE_CMPXCHG_DOUBLE
 	select HAVE_CMPXCHG_LOCAL
diff --git a/arch/x86/entry/common.c b/arch/x86/entry/common.c
index 7bc105f47d21..28cb3687bf82 100644
--- a/arch/x86/entry/common.c
+++ b/arch/x86/entry/common.c
@@ -32,6 +32,10 @@ 
 #include <linux/uaccess.h>
 #include <asm/cpufeature.h>
 
+#ifdef CONFIG_RANDOMIZE_KSTACK_OFFSET
+#include <linux/random.h>
+#endif
+
 #define CREATE_TRACE_POINTS
 #include <trace/events/syscalls.h>
 
@@ -269,10 +273,22 @@  __visible inline void syscall_return_slowpath(struct pt_regs *regs)
 }
 
 #ifdef CONFIG_X86_64
+
+#ifdef CONFIG_RANDOMIZE_KSTACK_OFFSET
+void *alloca(size_t size);
+#endif
+
 __visible void do_syscall_64(unsigned long nr, struct pt_regs *regs)
 {
 	struct thread_info *ti;
 
+#ifdef CONFIG_RANDOMIZE_KSTACK_OFFSET
+	size_t offset = ((size_t)prandom_u32()) % 256;
+	char *ptr = alloca(offset);
+
+	asm volatile("":"=m"(*ptr));
+#endif
+
 	enter_from_user_mode();
 	local_irq_enable();
 	ti = current_thread_info();