[PATCH] x86/hw_breakpoint: Prevent data breakpoints on cpu_entry_area

[PATCH] x86/hw_breakpoint: Prevent data breakpoints on cpu_entry_area

[Andy Lutomirski]

A data breakpoint near the top of an IST stack will cause unresoverable
recursion.  A data breakpoint on the GDT, IDT, or TSS is terrifying.
Prevent either of these from happening.

Co-developed-by: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Andy Lutomirski <luto@kernel.org>
---

The rest of my series is still in progress -- as we all know, idtentry
is a morass.  But this is self-contained and is an obvious fix.

arch/x86/include/asm/cpu_entry_area.h | 10 ++++++++++
 arch/x86/kernel/hw_breakpoint.c       | 17 +++++++++++++++++
 2 files changed, 27 insertions(+)

diff --git a/arch/x86/include/asm/cpu_entry_area.h b/arch/x86/include/asm/cpu_entry_area.h
index e23e2d9a92d7..3f50d4738487 100644
--- a/arch/x86/include/asm/cpu_entry_area.h
+++ b/arch/x86/include/asm/cpu_entry_area.h
@@ -126,6 +126,16 @@ static inline struct entry_stack *cpu_entry_stack(int cpu)
 	return &get_cpu_entry_area(cpu)->entry_stack_page.stack;
 }
 
+/*
+ * Checks whether the range from addr to end, inclusive, overlaps the CPU
+ * entry area range.
+ */
+static inline bool within_cpu_entry_area(unsigned long addr, unsigned long end)
+{
+	return end >= CPU_ENTRY_AREA_PER_CPU &&
+		addr < (CPU_ENTRY_AREA_PER_CPU + CPU_ENTRY_AREA_TOT_SIZE);
+}
+
 #define __this_cpu_ist_top_va(name)					\
 	CEA_ESTACK_TOP(__this_cpu_read(cea_exception_stacks), name)
 
diff --git a/arch/x86/kernel/hw_breakpoint.c b/arch/x86/kernel/hw_breakpoint.c
index 218c8917118e..dc4581fe4b4e 100644
--- a/arch/x86/kernel/hw_breakpoint.c
+++ b/arch/x86/kernel/hw_breakpoint.c
@@ -231,6 +231,23 @@ static int arch_build_bp_info(struct perf_event *bp,
 			      const struct perf_event_attr *attr,
 			      struct arch_hw_breakpoint *hw)
 {
+	unsigned long bp_end;
+
+	/* Ensure that bp_end does not oveflow. */
+	if (attr->bp_len >= ULONG_MAX - attr->bp_addr)
+		return -EINVAL;
+
+	bp_end = attr->bp_addr + attr->bp_len - 1;
+
+	/*
+	 * Prevent any breakpoint of any type that overlaps the
+	 * cpu_entry_area.  This protects the IST stacks and also
+	 * reduces the chance that we ever find out what happens if
+	 * there's a data breakpoint on the GDT, IDT, or TSS.
+	 */
+	if (within_cpu_entry_area(attr->bp_addr, bp_end))
+		return -EINVAL;
+
 	hw->address = attr->bp_addr;
 	hw->mask = 0;
 
-- 
2.21.0

Re: [PATCH] x86/hw_breakpoint: Prevent data breakpoints on cpu_entry_area

[Peter Zijlstra]

On Thu, Jul 25, 2019 at 09:37:15AM -0700, Andy Lutomirski wrote:
> A data breakpoint near the top of an IST stack will cause unresoverable
> recursion.  A data breakpoint on the GDT, IDT, or TSS is terrifying.
> Prevent either of these from happening.
> 
> Co-developed-by: Peter Zijlstra <peterz@infradead.org>
> Signed-off-by: Andy Lutomirski <luto@kernel.org>

Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>

One small nit beflow.

> ---
> 
>  arch/x86/include/asm/cpu_entry_area.h | 10 ++++++++++
>  arch/x86/kernel/hw_breakpoint.c       | 17 +++++++++++++++++
>  2 files changed, 27 insertions(+)
> 
> diff --git a/arch/x86/include/asm/cpu_entry_area.h b/arch/x86/include/asm/cpu_entry_area.h
> index e23e2d9a92d7..3f50d4738487 100644
> --- a/arch/x86/include/asm/cpu_entry_area.h
> +++ b/arch/x86/include/asm/cpu_entry_area.h
> @@ -126,6 +126,16 @@ static inline struct entry_stack *cpu_entry_stack(int cpu)
>  	return &get_cpu_entry_area(cpu)->entry_stack_page.stack;
>  }
>  
> +/*
> + * Checks whether the range from addr to end, inclusive, overlaps the CPU
> + * entry area range.
> + */
> +static inline bool within_cpu_entry_area(unsigned long addr, unsigned long end)
> +{
> +	return end >= CPU_ENTRY_AREA_PER_CPU &&
> +		addr < (CPU_ENTRY_AREA_PER_CPU + CPU_ENTRY_AREA_TOT_SIZE);
> +}
> +
>  #define __this_cpu_ist_top_va(name)					\
>  	CEA_ESTACK_TOP(__this_cpu_read(cea_exception_stacks), name)
>  
> diff --git a/arch/x86/kernel/hw_breakpoint.c b/arch/x86/kernel/hw_breakpoint.c
> index 218c8917118e..dc4581fe4b4e 100644
> --- a/arch/x86/kernel/hw_breakpoint.c
> +++ b/arch/x86/kernel/hw_breakpoint.c
> @@ -231,6 +231,23 @@ static int arch_build_bp_info(struct perf_event *bp,
>  			      const struct perf_event_attr *attr,
>  			      struct arch_hw_breakpoint *hw)
>  {
> +	unsigned long bp_end;
> +
> +	/* Ensure that bp_end does not oveflow. */
> +	if (attr->bp_len >= ULONG_MAX - attr->bp_addr)
> +		return -EINVAL;
> +
> +	bp_end = attr->bp_addr + attr->bp_len - 1;

The alternative (and possibly more conventional) overflow test would be:

	if (bp_end < attr->bp_addr)
		return -EINVAL;

> +
> +	/*
> +	 * Prevent any breakpoint of any type that overlaps the
> +	 * cpu_entry_area.  This protects the IST stacks and also
> +	 * reduces the chance that we ever find out what happens if
> +	 * there's a data breakpoint on the GDT, IDT, or TSS.
> +	 */
> +	if (within_cpu_entry_area(attr->bp_addr, bp_end))
> +		return -EINVAL;
> +
>  	hw->address = attr->bp_addr;
>  	hw->mask = 0;
>  

Re: [PATCH] x86/hw_breakpoint: Prevent data breakpoints on cpu_entry_area

[Thomas Gleixner]

On Thu, 25 Jul 2019, Peter Zijlstra wrote:

> On Thu, Jul 25, 2019 at 09:37:15AM -0700, Andy Lutomirski wrote:
> > A data breakpoint near the top of an IST stack will cause unresoverable

unresoverable?

> > recursion.  A data breakpoint on the GDT, IDT, or TSS is terrifying.
> > Prevent either of these from happening.
> > 
> > Co-developed-by: Peter Zijlstra <peterz@infradead.org>

Co-developed-by want's a Signed-off-by of the co-developer

> > Signed-off-by: Andy Lutomirski <luto@kernel.org>
> > diff --git a/arch/x86/kernel/hw_breakpoint.c b/arch/x86/kernel/hw_breakpoint.c
> > index 218c8917118e..dc4581fe4b4e 100644
> > --- a/arch/x86/kernel/hw_breakpoint.c
> > +++ b/arch/x86/kernel/hw_breakpoint.c
> > @@ -231,6 +231,23 @@ static int arch_build_bp_info(struct perf_event *bp,
> >  			      const struct perf_event_attr *attr,
> >  			      struct arch_hw_breakpoint *hw)
> >  {
> > +	unsigned long bp_end;
> > +
> > +	/* Ensure that bp_end does not oveflow. */

oveflow?

> > +	if (attr->bp_len >= ULONG_MAX - attr->bp_addr)
> > +		return -EINVAL;
> > +
> > +	bp_end = attr->bp_addr + attr->bp_len - 1;
> 
> The alternative (and possibly more conventional) overflow test would be:
> 
> 	if (bp_end < attr->bp_addr)
> 		return -EINVAL;

Yes please.

> > +
> > +	/*
> > +	 * Prevent any breakpoint of any type that overlaps the
> > +	 * cpu_entry_area.  This protects the IST stacks and also
> > +	 * reduces the chance that we ever find out what happens if

I surely hope that the chance is reduced to 0 ...

I know this is all an annoyance brought to us by hardware and I surely
enjoy the hidden sarcasm but please make this information as technically
accurate as possible. Put the rant into an extra line of the comment :)

> > +	 * there's a data breakpoint on the GDT, IDT, or TSS.
> > +	 */
> > +	if (within_cpu_entry_area(attr->bp_addr, bp_end))
> > +		return -EINVAL;

Thanks,

	tglx