[PATCH V2] powerpc/ptrace: Mitigate potential Spectre v1

[PATCH V2] powerpc/ptrace: Mitigate potential Spectre v1

[Breno Leitao]

'regno' is directly controlled by user space, hence leading to a potential
exploitation of the Spectre variant 1 vulnerability.

On PTRACE_SETREGS and PTRACE_GETREGS requests, user space passes the
register number that would be read or written. This register number is
called 'regno' which is part of the 'addr' syscall parameter.

This 'regno' value is checked against the maximum pt_regs structure size,
and then used to dereference it, which matches the initial part of a
Spectre v1 (and Spectre v1.1) attack. The dereferenced value, then,
is returned to userspace in the GETREGS case.

This patch sanitizes 'regno' before using it to dereference pt_reg.

Notice that given that speculation windows are large, the policy is
to kill the speculation on the first load and not worry if it can be
completed with a dependent load/store [1].

[1] https://marc.info/?l=linux-kernel&m=152449131114778&w=2

Signed-off-by: Breno Leitao <leitao@debian.org>
---
 arch/powerpc/kernel/ptrace.c | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)

diff --git a/arch/powerpc/kernel/ptrace.c b/arch/powerpc/kernel/ptrace.c
index cdd5d1d3ae41..7535f89e08cd 100644
--- a/arch/powerpc/kernel/ptrace.c
+++ b/arch/powerpc/kernel/ptrace.c
@@ -33,6 +33,7 @@
 #include <linux/hw_breakpoint.h>
 #include <linux/perf_event.h>
 #include <linux/context_tracking.h>
+#include <linux/nospec.h>
 
 #include <linux/uaccess.h>
 #include <linux/pkeys.h>
@@ -274,6 +275,8 @@ static int set_user_trap(struct task_struct *task, unsigned long trap)
  */
 int ptrace_get_reg(struct task_struct *task, int regno, unsigned long *data)
 {
+	unsigned int regs_max;
+
 	if ((task->thread.regs == NULL) || !data)
 		return -EIO;
 
@@ -297,7 +300,9 @@ int ptrace_get_reg(struct task_struct *task, int regno, unsigned long *data)
 	}
 #endif
 
-	if (regno < (sizeof(struct user_pt_regs) / sizeof(unsigned long))) {
+	regs_max = sizeof(struct user_pt_regs) / sizeof(unsigned long);
+	if (regno < regs_max) {
+		regno = array_index_nospec(regno, regs_max);
 		*data = ((unsigned long *)task->thread.regs)[regno];
 		return 0;
 	}
@@ -321,6 +326,7 @@ int ptrace_put_reg(struct task_struct *task, int regno, unsigned long data)
 		return set_user_dscr(task, data);
 
 	if (regno <= PT_MAX_PUT_REG) {
+		regno = array_index_nospec(regno, PT_MAX_PUT_REG + 1);
 		((unsigned long *)task->thread.regs)[regno] = data;
 		return 0;
 	}
-- 
2.19.0

Re: [PATCH V2] powerpc/ptrace: Mitigate potential Spectre v1

[Gustavo A. R. Silva]

On 1/30/19 6:46 AM, Breno Leitao wrote:
> 'regno' is directly controlled by user space, hence leading to a potential
> exploitation of the Spectre variant 1 vulnerability.
> 
> On PTRACE_SETREGS and PTRACE_GETREGS requests, user space passes the
> register number that would be read or written. This register number is
> called 'regno' which is part of the 'addr' syscall parameter.
> 
> This 'regno' value is checked against the maximum pt_regs structure size,
> and then used to dereference it, which matches the initial part of a
> Spectre v1 (and Spectre v1.1) attack. The dereferenced value, then,
> is returned to userspace in the GETREGS case.
> 
> This patch sanitizes 'regno' before using it to dereference pt_reg.
> 
> Notice that given that speculation windows are large, the policy is
> to kill the speculation on the first load and not worry if it can be
> completed with a dependent load/store [1].
> 
> [1] https://marc.info/?l=linux-kernel&m=152449131114778&w=2
> 
> Signed-off-by: Breno Leitao <leitao@debian.org>

Acked-by: Gustavo A. R. Silva <gustavo@embeddedor.com>

> ---
>  arch/powerpc/kernel/ptrace.c | 8 +++++++-
>  1 file changed, 7 insertions(+), 1 deletion(-)
> 
> diff --git a/arch/powerpc/kernel/ptrace.c b/arch/powerpc/kernel/ptrace.c
> index cdd5d1d3ae41..7535f89e08cd 100644
> --- a/arch/powerpc/kernel/ptrace.c
> +++ b/arch/powerpc/kernel/ptrace.c
> @@ -33,6 +33,7 @@
>  #include <linux/hw_breakpoint.h>
>  #include <linux/perf_event.h>
>  #include <linux/context_tracking.h>
> +#include <linux/nospec.h>
>  
>  #include <linux/uaccess.h>
>  #include <linux/pkeys.h>
> @@ -274,6 +275,8 @@ static int set_user_trap(struct task_struct *task, unsigned long trap)
>   */
>  int ptrace_get_reg(struct task_struct *task, int regno, unsigned long *data)
>  {
> +	unsigned int regs_max;
> +
>  	if ((task->thread.regs == NULL) || !data)
>  		return -EIO;
>  
> @@ -297,7 +300,9 @@ int ptrace_get_reg(struct task_struct *task, int regno, unsigned long *data)
>  	}
>  #endif
>  
> -	if (regno < (sizeof(struct user_pt_regs) / sizeof(unsigned long))) {
> +	regs_max = sizeof(struct user_pt_regs) / sizeof(unsigned long);
> +	if (regno < regs_max) {
> +		regno = array_index_nospec(regno, regs_max);
>  		*data = ((unsigned long *)task->thread.regs)[regno];
>  		return 0;
>  	}
> @@ -321,6 +326,7 @@ int ptrace_put_reg(struct task_struct *task, int regno, unsigned long data)
>  		return set_user_dscr(task, data);
>  
>  	if (regno <= PT_MAX_PUT_REG) {
> +		regno = array_index_nospec(regno, PT_MAX_PUT_REG + 1);
>  		((unsigned long *)task->thread.regs)[regno] = data;
>  		return 0;
>  	}
> 

Re: [V2] powerpc/ptrace: Mitigate potential Spectre v1

[Michael Ellerman]

On Wed, 2019-01-30 at 12:46:00 UTC, Breno Leitao wrote:
> 'regno' is directly controlled by user space, hence leading to a potential
> exploitation of the Spectre variant 1 vulnerability.
> 
> On PTRACE_SETREGS and PTRACE_GETREGS requests, user space passes the
> register number that would be read or written. This register number is
> called 'regno' which is part of the 'addr' syscall parameter.
> 
> This 'regno' value is checked against the maximum pt_regs structure size,
> and then used to dereference it, which matches the initial part of a
> Spectre v1 (and Spectre v1.1) attack. The dereferenced value, then,
> is returned to userspace in the GETREGS case.
> 
> This patch sanitizes 'regno' before using it to dereference pt_reg.
> 
> Notice that given that speculation windows are large, the policy is
> to kill the speculation on the first load and not worry if it can be
> completed with a dependent load/store [1].
> 
> [1] https://marc.info/?l=linux-kernel&m=152449131114778&w=2
> 
> Signed-off-by: Breno Leitao <leitao@debian.org>
> Acked-by: Gustavo A. R. Silva <gustavo@embeddedor.com>

Applied to powerpc next, thanks.

https://git.kernel.org/powerpc/c/ebb0e13ead2ddc186a80b1b0235deeef

cheers