[PATCH] arm64/traps: show fault address in exception trace

[PATCH] arm64/traps: show fault address in exception trace

[Baruch Siach]

The FAR (fault address) register provides useful information when
debugging an unhandled exception.

Signed-off-by: Baruch Siach <baruch@tkos.co.il>
---
 arch/arm64/kernel/traps.c | 11 ++++++-----
 1 file changed, 6 insertions(+), 5 deletions(-)

diff --git a/arch/arm64/kernel/traps.c b/arch/arm64/kernel/traps.c
index 4e26bd356a48..14eb598e33a7 100644
--- a/arch/arm64/kernel/traps.c
+++ b/arch/arm64/kernel/traps.c
@@ -239,7 +239,7 @@ void die(const char *str, struct pt_regs *regs, long err)
 		make_task_dead(SIGSEGV);
 }
 
-static void arm64_show_signal(int signo, const char *str)
+static void arm64_show_signal(int signo, const char *str, unsigned long far)
 {
 	static DEFINE_RATELIMIT_STATE(rs, DEFAULT_RATELIMIT_INTERVAL,
 				      DEFAULT_RATELIMIT_BURST);
@@ -256,6 +256,7 @@ static void arm64_show_signal(int signo, const char *str)
 	pr_info("%s[%d]: unhandled exception: ", tsk->comm, task_pid_nr(tsk));
 	if (esr)
 		pr_cont("%s, ESR 0x%016lx, ", esr_get_class_string(esr), esr);
+	pr_cont("FAR 0x%016lx, ", far);
 
 	pr_cont("%s", str);
 	print_vma_addr(KERN_CONT " in ", regs->pc);
@@ -266,7 +267,7 @@ static void arm64_show_signal(int signo, const char *str)
 void arm64_force_sig_fault(int signo, int code, unsigned long far,
 			   const char *str)
 {
-	arm64_show_signal(signo, str);
+	arm64_show_signal(signo, str, far);
 	if (signo == SIGKILL)
 		force_sig(SIGKILL);
 	else
@@ -275,21 +276,21 @@ void arm64_force_sig_fault(int signo, int code, unsigned long far,
 
 void arm64_force_sig_fault_pkey(unsigned long far, const char *str, int pkey)
 {
-	arm64_show_signal(SIGSEGV, str);
+	arm64_show_signal(SIGSEGV, str, far);
 	force_sig_pkuerr((void __user *)far, pkey);
 }
 
 void arm64_force_sig_mceerr(int code, unsigned long far, short lsb,
 			    const char *str)
 {
-	arm64_show_signal(SIGBUS, str);
+	arm64_show_signal(SIGBUS, str, far);
 	force_sig_mceerr(code, (void __user *)far, lsb);
 }
 
 void arm64_force_sig_ptrace_errno_trap(int errno, unsigned long far,
 				       const char *str)
 {
-	arm64_show_signal(SIGTRAP, str);
+	arm64_show_signal(SIGTRAP, str, far);
 	force_sig_ptrace_errno_trap(errno, (void __user *)far);
 }
 
-- 
2.47.2

Re: [PATCH] arm64/traps: show fault address in exception trace

[Will Deacon]

On Tue, Apr 29, 2025 at 10:15:24AM +0300, Baruch Siach wrote:
> The FAR (fault address) register provides useful information when
> debugging an unhandled exception.
> 
> Signed-off-by: Baruch Siach <baruch@tkos.co.il>
> ---
>  arch/arm64/kernel/traps.c | 11 ++++++-----
>  1 file changed, 6 insertions(+), 5 deletions(-)
> 
> diff --git a/arch/arm64/kernel/traps.c b/arch/arm64/kernel/traps.c
> index 4e26bd356a48..14eb598e33a7 100644
> --- a/arch/arm64/kernel/traps.c
> +++ b/arch/arm64/kernel/traps.c
> @@ -239,7 +239,7 @@ void die(const char *str, struct pt_regs *regs, long err)
>  		make_task_dead(SIGSEGV);
>  }
>  
> -static void arm64_show_signal(int signo, const char *str)
> +static void arm64_show_signal(int signo, const char *str, unsigned long far)
>  {
>  	static DEFINE_RATELIMIT_STATE(rs, DEFAULT_RATELIMIT_INTERVAL,
>  				      DEFAULT_RATELIMIT_BURST);
> @@ -256,6 +256,7 @@ static void arm64_show_signal(int signo, const char *str)
>  	pr_info("%s[%d]: unhandled exception: ", tsk->comm, task_pid_nr(tsk));
>  	if (esr)
>  		pr_cont("%s, ESR 0x%016lx, ", esr_get_class_string(esr), esr);
> +	pr_cont("FAR 0x%016lx, ", far);

Are you sure the FAR is always valid here? For example, if we're delivering
a SIGILL due to an undefined instruction.

Also, once you have the PC and the registers (both of which we print),
it's not too hard to dig the instruction out of the binary and figure out
the faulting address that way.

Will

Re: [PATCH] arm64/traps: show fault address in exception trace

[Baruch Siach]

Hi Will,

On Tue, Apr 29 2025, Will Deacon wrote:
> On Tue, Apr 29, 2025 at 10:15:24AM +0300, Baruch Siach wrote:
>> The FAR (fault address) register provides useful information when
>> debugging an unhandled exception.
>> 
>> Signed-off-by: Baruch Siach <baruch@tkos.co.il>
>> ---
>>  arch/arm64/kernel/traps.c | 11 ++++++-----
>>  1 file changed, 6 insertions(+), 5 deletions(-)
>> 
>> diff --git a/arch/arm64/kernel/traps.c b/arch/arm64/kernel/traps.c
>> index 4e26bd356a48..14eb598e33a7 100644
>> --- a/arch/arm64/kernel/traps.c
>> +++ b/arch/arm64/kernel/traps.c
>> @@ -239,7 +239,7 @@ void die(const char *str, struct pt_regs *regs, long err)
>>  		make_task_dead(SIGSEGV);
>>  }
>>  
>> -static void arm64_show_signal(int signo, const char *str)
>> +static void arm64_show_signal(int signo, const char *str, unsigned long far)
>>  {
>>  	static DEFINE_RATELIMIT_STATE(rs, DEFAULT_RATELIMIT_INTERVAL,
>>  				      DEFAULT_RATELIMIT_BURST);
>> @@ -256,6 +256,7 @@ static void arm64_show_signal(int signo, const char *str)
>>  	pr_info("%s[%d]: unhandled exception: ", tsk->comm, task_pid_nr(tsk));
>>  	if (esr)
>>  		pr_cont("%s, ESR 0x%016lx, ", esr_get_class_string(esr), esr);
>> +	pr_cont("FAR 0x%016lx, ", far);
>
> Are you sure the FAR is always valid here? For example, if we're delivering
> a SIGILL due to an undefined instruction.

A user interested in hardware exception information should be in a
position to know whether this information is valid. But I guess that FAR
can't be valid when ESR isn't valid either.

> Also, once you have the PC and the registers (both of which we print),
> it's not too hard to dig the instruction out of the binary and figure out
> the faulting address that way.

I find FAR helpful to confirm the analysis of the instruction being
executed and its operands. The cost for adding this bit of information
isn't huge, I think.

Thanks for your review,
baruch

-- 
                                                     ~. .~   Tk Open Systems
=}------------------------------------------------ooO--U--Ooo------------{=
   - baruch@tkos.co.il - tel: +972.52.368.4656, http://www.tkos.co.il -