diff mbox series

[v2,3/4] arm64: mte: Enable async tag check fault

Message ID 20210107172908.42686-4-vincenzo.frascino@arm.com (mailing list archive)
State New, archived
Headers show
Series arm64: ARMv8.5-A: MTE: Add async mode support | expand

Commit Message

Vincenzo Frascino Jan. 7, 2021, 5:29 p.m. UTC
MTE provides a mode that asynchronously updates the TFSR_EL1 register
when a tag check exception is detected.

To take advantage of this mode the kernel has to verify the status of
the register at:
  1. Context switching
  2. Return to user/EL0 (Not required in entry from EL0 since the kernel
  did not run)
  3. Kernel entry from EL1
  4. Kernel exit to EL1

If the register is non-zero a trace is reported.

Add the required features for EL1 detection and reporting.

Note: ITFSB bit is set in the SCTLR_EL1 register hence it guaranties that
the indirect writes to TFSR_EL1 are synchronized at exception entry to
EL1. On the context switch path the synchronization is guarantied by the
dsb() in __switch_to().

Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Will Deacon <will@kernel.org>
Signed-off-by: Vincenzo Frascino <vincenzo.frascino@arm.com>
---
 arch/arm64/include/asm/mte.h     |  4 ++++
 arch/arm64/kernel/entry-common.c |  6 ++++++
 arch/arm64/kernel/mte.c          | 37 ++++++++++++++++++++++++++++++++
 3 files changed, 47 insertions(+)

Comments

Catalin Marinas Jan. 13, 2021, 6:11 p.m. UTC | #1
On Thu, Jan 07, 2021 at 05:29:07PM +0000, Vincenzo Frascino wrote:
> diff --git a/arch/arm64/include/asm/mte.h b/arch/arm64/include/asm/mte.h
> index d02aff9f493d..a60d3718baae 100644
> --- a/arch/arm64/include/asm/mte.h
> +++ b/arch/arm64/include/asm/mte.h
> @@ -39,6 +39,7 @@ void mte_free_tag_storage(char *storage);
>  /* track which pages have valid allocation tags */
>  #define PG_mte_tagged	PG_arch_2
>  
> +void mte_check_tfsr_el1(void);
>  void mte_sync_tags(pte_t *ptep, pte_t pte);
>  void mte_copy_page_tags(void *kto, const void *kfrom);
>  void flush_mte_state(void);
> @@ -56,6 +57,9 @@ void mte_assign_mem_tag_range(void *addr, size_t size);
>  /* unused if !CONFIG_ARM64_MTE, silence the compiler */
>  #define PG_mte_tagged	0
>  
> +static inline void mte_check_tfsr_el1(void)
> +{
> +}

I think we should enable this dummy function when !CONFIG_KASAN_HW_TAGS.
It saves us an unnecessary function call in a few places.

>  static inline void mte_sync_tags(pte_t *ptep, pte_t pte)
>  {
>  }
> diff --git a/arch/arm64/kernel/entry-common.c b/arch/arm64/kernel/entry-common.c
> index 5346953e4382..74b020ce72d7 100644
> --- a/arch/arm64/kernel/entry-common.c
> +++ b/arch/arm64/kernel/entry-common.c
> @@ -37,6 +37,8 @@ static void noinstr enter_from_kernel_mode(struct pt_regs *regs)
>  	lockdep_hardirqs_off(CALLER_ADDR0);
>  	rcu_irq_enter_check_tick();
>  	trace_hardirqs_off_finish();
> +
> +	mte_check_tfsr_el1();
>  }
>  
>  /*
> @@ -47,6 +49,8 @@ static void noinstr exit_to_kernel_mode(struct pt_regs *regs)
>  {
>  	lockdep_assert_irqs_disabled();
>  
> +	mte_check_tfsr_el1();
> +
>  	if (interrupts_enabled(regs)) {
>  		if (regs->exit_rcu) {
>  			trace_hardirqs_on_prepare();
> @@ -243,6 +247,8 @@ asmlinkage void noinstr enter_from_user_mode(void)
>  
>  asmlinkage void noinstr exit_to_user_mode(void)
>  {
> +	mte_check_tfsr_el1();

While for kernel entry the asynchronous faults are sync'ed automatically
with TFSR_EL1, we don't have this for exit, so we'd need an explicit
DSB. But rather than placing it here, it's better if we add a bool sync
argument to mte_check_tfsr_el1() which issues a dsb() before checking
the register. I think that's the only place where such argument would be
true (for now).

> +
>  	trace_hardirqs_on_prepare();
>  	lockdep_hardirqs_on_prepare(CALLER_ADDR0);
>  	user_enter_irqoff();
> diff --git a/arch/arm64/kernel/mte.c b/arch/arm64/kernel/mte.c
> index 5d992e16b420..26030f0b79fe 100644
> --- a/arch/arm64/kernel/mte.c
> +++ b/arch/arm64/kernel/mte.c
> @@ -185,6 +185,34 @@ void mte_enable_kernel(enum kasan_arg_mode mode)
>  	isb();
>  }
>  
> +void mte_check_tfsr_el1(void)
> +{
> +	u64 tfsr_el1;
> +
> +	if (!IS_ENABLED(CONFIG_KASAN_HW_TAGS))
> +		return;

If we define the static inline when !CONFIG_KASAN_HW_TAGS, we could add
the #ifdef here around the whole function.

> +	if (!system_supports_mte())
> +		return;
> +
> +	tfsr_el1 = read_sysreg_s(SYS_TFSR_EL1);
> +
> +	/*
> +	 * The kernel should never hit the condition TF0 == 1
> +	 * at this point because for the futex code we set
> +	 * PSTATE.TCO.
> +	 */
> +	WARN_ON(tfsr_el1 & SYS_TFSR_EL1_TF0);
> +
> +	if (tfsr_el1 & SYS_TFSR_EL1_TF1) {
> +		write_sysreg_s(0, SYS_TFSR_EL1);
> +		isb();
> +
> +		pr_err("MTE: Asynchronous tag exception detected!");
> +	}
> +}
> +NOKPROBE_SYMBOL(mte_check_tfsr_el1);

Do we need this to be NOKPROBE_SYMBOL? It's not that low level.

> +
>  static void update_sctlr_el1_tcf0(u64 tcf0)
>  {
>  	/* ISB required for the kernel uaccess routines */
> @@ -250,6 +278,15 @@ void mte_thread_switch(struct task_struct *next)
>  	/* avoid expensive SCTLR_EL1 accesses if no change */
>  	if (current->thread.sctlr_tcf0 != next->thread.sctlr_tcf0)
>  		update_sctlr_el1_tcf0(next->thread.sctlr_tcf0);
> +
> +	/*
> +	 * Check if an async tag exception occurred at EL1.
> +	 *
> +	 * Note: On the context switch patch we rely on the dsb() present

s/patch/path/

> +	 * in __switch_to() to guaranty that the indirect writes to TFSR_EL1

s/guaranty/guarantee/ (well, still valid though I think rarely used).

> +	 * are synchronized before this point.
> +	 */
> +	mte_check_tfsr_el1();
>  }
>  
>  void mte_suspend_exit(void)
> -- 
> 2.30.0
Vincenzo Frascino Jan. 14, 2021, 10:24 a.m. UTC | #2
On 1/13/21 6:11 PM, Catalin Marinas wrote:
> On Thu, Jan 07, 2021 at 05:29:07PM +0000, Vincenzo Frascino wrote:
>> diff --git a/arch/arm64/include/asm/mte.h b/arch/arm64/include/asm/mte.h
>> index d02aff9f493d..a60d3718baae 100644
>> --- a/arch/arm64/include/asm/mte.h
>> +++ b/arch/arm64/include/asm/mte.h
>> @@ -39,6 +39,7 @@ void mte_free_tag_storage(char *storage);
>>  /* track which pages have valid allocation tags */
>>  #define PG_mte_tagged	PG_arch_2
>>  
>> +void mte_check_tfsr_el1(void);
>>  void mte_sync_tags(pte_t *ptep, pte_t pte);
>>  void mte_copy_page_tags(void *kto, const void *kfrom);
>>  void flush_mte_state(void);
>> @@ -56,6 +57,9 @@ void mte_assign_mem_tag_range(void *addr, size_t size);
>>  /* unused if !CONFIG_ARM64_MTE, silence the compiler */
>>  #define PG_mte_tagged	0
>>  
>> +static inline void mte_check_tfsr_el1(void)
>> +{
>> +}
> 
> I think we should enable this dummy function when !CONFIG_KASAN_HW_TAGS.
> It saves us an unnecessary function call in a few places.
> 

Ok, I will add it in v3.

>>  static inline void mte_sync_tags(pte_t *ptep, pte_t pte)
>>  {
>>  }
>> diff --git a/arch/arm64/kernel/entry-common.c b/arch/arm64/kernel/entry-common.c
>> index 5346953e4382..74b020ce72d7 100644
>> --- a/arch/arm64/kernel/entry-common.c
>> +++ b/arch/arm64/kernel/entry-common.c
>> @@ -37,6 +37,8 @@ static void noinstr enter_from_kernel_mode(struct pt_regs *regs)
>>  	lockdep_hardirqs_off(CALLER_ADDR0);
>>  	rcu_irq_enter_check_tick();
>>  	trace_hardirqs_off_finish();
>> +
>> +	mte_check_tfsr_el1();
>>  }
>>  
>>  /*
>> @@ -47,6 +49,8 @@ static void noinstr exit_to_kernel_mode(struct pt_regs *regs)
>>  {
>>  	lockdep_assert_irqs_disabled();
>>  
>> +	mte_check_tfsr_el1();
>> +
>>  	if (interrupts_enabled(regs)) {
>>  		if (regs->exit_rcu) {
>>  			trace_hardirqs_on_prepare();
>> @@ -243,6 +247,8 @@ asmlinkage void noinstr enter_from_user_mode(void)
>>  
>>  asmlinkage void noinstr exit_to_user_mode(void)
>>  {
>> +	mte_check_tfsr_el1();
> 
> While for kernel entry the asynchronous faults are sync'ed automatically
> with TFSR_EL1, we don't have this for exit, so we'd need an explicit
> DSB. But rather than placing it here, it's better if we add a bool sync
> argument to mte_check_tfsr_el1() which issues a dsb() before checking
> the register. I think that's the only place where such argument would be
> true (for now).
> 

Good point, I will add the dsb() in mte_check_tfsr_el1() but instead of a bool
parameter I will add something more explicit.

>> +
>>  	trace_hardirqs_on_prepare();
>>  	lockdep_hardirqs_on_prepare(CALLER_ADDR0);
>>  	user_enter_irqoff();
>> diff --git a/arch/arm64/kernel/mte.c b/arch/arm64/kernel/mte.c
>> index 5d992e16b420..26030f0b79fe 100644
>> --- a/arch/arm64/kernel/mte.c
>> +++ b/arch/arm64/kernel/mte.c
>> @@ -185,6 +185,34 @@ void mte_enable_kernel(enum kasan_arg_mode mode)
>>  	isb();
>>  }
>>  
>> +void mte_check_tfsr_el1(void)
>> +{
>> +	u64 tfsr_el1;
>> +
>> +	if (!IS_ENABLED(CONFIG_KASAN_HW_TAGS))
>> +		return;
> 
> If we define the static inline when !CONFIG_KASAN_HW_TAGS, we could add
> the #ifdef here around the whole function.
>

Ok. I will add it in v3.

>> +	if (!system_supports_mte())
>> +		return;
>> +
>> +	tfsr_el1 = read_sysreg_s(SYS_TFSR_EL1);
>> +
>> +	/*
>> +	 * The kernel should never hit the condition TF0 == 1
>> +	 * at this point because for the futex code we set
>> +	 * PSTATE.TCO.
>> +	 */
>> +	WARN_ON(tfsr_el1 & SYS_TFSR_EL1_TF0);
>> +
>> +	if (tfsr_el1 & SYS_TFSR_EL1_TF1) {
>> +		write_sysreg_s(0, SYS_TFSR_EL1);
>> +		isb();
>> +
>> +		pr_err("MTE: Asynchronous tag exception detected!");
>> +	}
>> +}
>> +NOKPROBE_SYMBOL(mte_check_tfsr_el1);
> 
> Do we need this to be NOKPROBE_SYMBOL? It's not that low level.
>
It is an inheritance from when I had this code called very early. I will remove
it in the next version.

>> +
>>  static void update_sctlr_el1_tcf0(u64 tcf0)
>>  {
>>  	/* ISB required for the kernel uaccess routines */
>> @@ -250,6 +278,15 @@ void mte_thread_switch(struct task_struct *next)
>>  	/* avoid expensive SCTLR_EL1 accesses if no change */
>>  	if (current->thread.sctlr_tcf0 != next->thread.sctlr_tcf0)
>>  		update_sctlr_el1_tcf0(next->thread.sctlr_tcf0);
>> +
>> +	/*
>> +	 * Check if an async tag exception occurred at EL1.
>> +	 *
>> +	 * Note: On the context switch patch we rely on the dsb() present
> 
> s/patch/path/
> 
>> +	 * in __switch_to() to guaranty that the indirect writes to TFSR_EL1
> 
> s/guaranty/guarantee/ (well, still valid though I think rarely used).
> 
>> +	 * are synchronized before this point.
>> +	 */
>> +	mte_check_tfsr_el1();
>>  }
>>  
>>  void mte_suspend_exit(void)
>> -- 
>> 2.30.0
>
Catalin Marinas Jan. 14, 2021, 2:25 p.m. UTC | #3
On Thu, Jan 14, 2021 at 10:24:25AM +0000, Vincenzo Frascino wrote:
> On 1/13/21 6:11 PM, Catalin Marinas wrote:
> > On Thu, Jan 07, 2021 at 05:29:07PM +0000, Vincenzo Frascino wrote:
> >>  static inline void mte_sync_tags(pte_t *ptep, pte_t pte)
> >>  {
> >>  }
> >> diff --git a/arch/arm64/kernel/entry-common.c b/arch/arm64/kernel/entry-common.c
> >> index 5346953e4382..74b020ce72d7 100644
> >> --- a/arch/arm64/kernel/entry-common.c
> >> +++ b/arch/arm64/kernel/entry-common.c
> >> @@ -37,6 +37,8 @@ static void noinstr enter_from_kernel_mode(struct pt_regs *regs)
> >>  	lockdep_hardirqs_off(CALLER_ADDR0);
> >>  	rcu_irq_enter_check_tick();
> >>  	trace_hardirqs_off_finish();
> >> +
> >> +	mte_check_tfsr_el1();
> >>  }
> >>  
> >>  /*
> >> @@ -47,6 +49,8 @@ static void noinstr exit_to_kernel_mode(struct pt_regs *regs)
> >>  {
> >>  	lockdep_assert_irqs_disabled();
> >>  
> >> +	mte_check_tfsr_el1();
> >> +
> >>  	if (interrupts_enabled(regs)) {
> >>  		if (regs->exit_rcu) {
> >>  			trace_hardirqs_on_prepare();
> >> @@ -243,6 +247,8 @@ asmlinkage void noinstr enter_from_user_mode(void)
> >>  
> >>  asmlinkage void noinstr exit_to_user_mode(void)
> >>  {
> >> +	mte_check_tfsr_el1();
> > 
> > While for kernel entry the asynchronous faults are sync'ed automatically
> > with TFSR_EL1, we don't have this for exit, so we'd need an explicit
> > DSB. But rather than placing it here, it's better if we add a bool sync
> > argument to mte_check_tfsr_el1() which issues a dsb() before checking
> > the register. I think that's the only place where such argument would be
> > true (for now).
> 
> Good point, I will add the dsb() in mte_check_tfsr_el1() but instead of a bool
> parameter I will add something more explicit.

Or rename the function to mte_check_tfsr_el1_no_sync() and have a static
inline mte_check_tfsr_el1() which issues a dsb() before calling the
*no_sync variant.

Adding an enum instead here is not worth it (if that's what you meant by
not using a bool).
Vincenzo Frascino Jan. 14, 2021, 2:57 p.m. UTC | #4
On 1/14/21 2:25 PM, Catalin Marinas wrote:
> On Thu, Jan 14, 2021 at 10:24:25AM +0000, Vincenzo Frascino wrote:
>> On 1/13/21 6:11 PM, Catalin Marinas wrote:
>>> On Thu, Jan 07, 2021 at 05:29:07PM +0000, Vincenzo Frascino wrote:
>>>>  static inline void mte_sync_tags(pte_t *ptep, pte_t pte)
>>>>  {
>>>>  }
>>>> diff --git a/arch/arm64/kernel/entry-common.c b/arch/arm64/kernel/entry-common.c
>>>> index 5346953e4382..74b020ce72d7 100644
>>>> --- a/arch/arm64/kernel/entry-common.c
>>>> +++ b/arch/arm64/kernel/entry-common.c
>>>> @@ -37,6 +37,8 @@ static void noinstr enter_from_kernel_mode(struct pt_regs *regs)
>>>>  	lockdep_hardirqs_off(CALLER_ADDR0);
>>>>  	rcu_irq_enter_check_tick();
>>>>  	trace_hardirqs_off_finish();
>>>> +
>>>> +	mte_check_tfsr_el1();
>>>>  }
>>>>  
>>>>  /*
>>>> @@ -47,6 +49,8 @@ static void noinstr exit_to_kernel_mode(struct pt_regs *regs)
>>>>  {
>>>>  	lockdep_assert_irqs_disabled();
>>>>  
>>>> +	mte_check_tfsr_el1();
>>>> +
>>>>  	if (interrupts_enabled(regs)) {
>>>>  		if (regs->exit_rcu) {
>>>>  			trace_hardirqs_on_prepare();
>>>> @@ -243,6 +247,8 @@ asmlinkage void noinstr enter_from_user_mode(void)
>>>>  
>>>>  asmlinkage void noinstr exit_to_user_mode(void)
>>>>  {
>>>> +	mte_check_tfsr_el1();
>>>
>>> While for kernel entry the asynchronous faults are sync'ed automatically
>>> with TFSR_EL1, we don't have this for exit, so we'd need an explicit
>>> DSB. But rather than placing it here, it's better if we add a bool sync
>>> argument to mte_check_tfsr_el1() which issues a dsb() before checking
>>> the register. I think that's the only place where such argument would be
>>> true (for now).
>>
>> Good point, I will add the dsb() in mte_check_tfsr_el1() but instead of a bool
>> parameter I will add something more explicit.
> 
> Or rename the function to mte_check_tfsr_el1_no_sync() and have a static
> inline mte_check_tfsr_el1() which issues a dsb() before calling the
> *no_sync variant.
> 
> Adding an enum instead here is not worth it (if that's what you meant by
> not using a bool).
> 

I like this option more, thanks for pointing it out.
diff mbox series

Patch

diff --git a/arch/arm64/include/asm/mte.h b/arch/arm64/include/asm/mte.h
index d02aff9f493d..a60d3718baae 100644
--- a/arch/arm64/include/asm/mte.h
+++ b/arch/arm64/include/asm/mte.h
@@ -39,6 +39,7 @@  void mte_free_tag_storage(char *storage);
 /* track which pages have valid allocation tags */
 #define PG_mte_tagged	PG_arch_2
 
+void mte_check_tfsr_el1(void);
 void mte_sync_tags(pte_t *ptep, pte_t pte);
 void mte_copy_page_tags(void *kto, const void *kfrom);
 void flush_mte_state(void);
@@ -56,6 +57,9 @@  void mte_assign_mem_tag_range(void *addr, size_t size);
 /* unused if !CONFIG_ARM64_MTE, silence the compiler */
 #define PG_mte_tagged	0
 
+static inline void mte_check_tfsr_el1(void)
+{
+}
 static inline void mte_sync_tags(pte_t *ptep, pte_t pte)
 {
 }
diff --git a/arch/arm64/kernel/entry-common.c b/arch/arm64/kernel/entry-common.c
index 5346953e4382..74b020ce72d7 100644
--- a/arch/arm64/kernel/entry-common.c
+++ b/arch/arm64/kernel/entry-common.c
@@ -37,6 +37,8 @@  static void noinstr enter_from_kernel_mode(struct pt_regs *regs)
 	lockdep_hardirqs_off(CALLER_ADDR0);
 	rcu_irq_enter_check_tick();
 	trace_hardirqs_off_finish();
+
+	mte_check_tfsr_el1();
 }
 
 /*
@@ -47,6 +49,8 @@  static void noinstr exit_to_kernel_mode(struct pt_regs *regs)
 {
 	lockdep_assert_irqs_disabled();
 
+	mte_check_tfsr_el1();
+
 	if (interrupts_enabled(regs)) {
 		if (regs->exit_rcu) {
 			trace_hardirqs_on_prepare();
@@ -243,6 +247,8 @@  asmlinkage void noinstr enter_from_user_mode(void)
 
 asmlinkage void noinstr exit_to_user_mode(void)
 {
+	mte_check_tfsr_el1();
+
 	trace_hardirqs_on_prepare();
 	lockdep_hardirqs_on_prepare(CALLER_ADDR0);
 	user_enter_irqoff();
diff --git a/arch/arm64/kernel/mte.c b/arch/arm64/kernel/mte.c
index 5d992e16b420..26030f0b79fe 100644
--- a/arch/arm64/kernel/mte.c
+++ b/arch/arm64/kernel/mte.c
@@ -185,6 +185,34 @@  void mte_enable_kernel(enum kasan_arg_mode mode)
 	isb();
 }
 
+void mte_check_tfsr_el1(void)
+{
+	u64 tfsr_el1;
+
+	if (!IS_ENABLED(CONFIG_KASAN_HW_TAGS))
+		return;
+
+	if (!system_supports_mte())
+		return;
+
+	tfsr_el1 = read_sysreg_s(SYS_TFSR_EL1);
+
+	/*
+	 * The kernel should never hit the condition TF0 == 1
+	 * at this point because for the futex code we set
+	 * PSTATE.TCO.
+	 */
+	WARN_ON(tfsr_el1 & SYS_TFSR_EL1_TF0);
+
+	if (tfsr_el1 & SYS_TFSR_EL1_TF1) {
+		write_sysreg_s(0, SYS_TFSR_EL1);
+		isb();
+
+		pr_err("MTE: Asynchronous tag exception detected!");
+	}
+}
+NOKPROBE_SYMBOL(mte_check_tfsr_el1);
+
 static void update_sctlr_el1_tcf0(u64 tcf0)
 {
 	/* ISB required for the kernel uaccess routines */
@@ -250,6 +278,15 @@  void mte_thread_switch(struct task_struct *next)
 	/* avoid expensive SCTLR_EL1 accesses if no change */
 	if (current->thread.sctlr_tcf0 != next->thread.sctlr_tcf0)
 		update_sctlr_el1_tcf0(next->thread.sctlr_tcf0);
+
+	/*
+	 * Check if an async tag exception occurred at EL1.
+	 *
+	 * Note: On the context switch patch we rely on the dsb() present
+	 * in __switch_to() to guaranty that the indirect writes to TFSR_EL1
+	 * are synchronized before this point.
+	 */
+	mte_check_tfsr_el1();
 }
 
 void mte_suspend_exit(void)