[PATCH] mips/ftrace: Fix function tracing return address to match

[PATCH] mips/ftrace: Fix function tracing return address to match

[David Daney]

From: Corey Minyard <cminyard@mvista.com>

Dynamic function tracing was not working on MIPS.  When doing dynamic
tracing, the tracer attempts to match up the passed in address with
the one the compiler creates in the mcount tables.  The MIPS code was
passing in the return address from the tracing function call, but the
compiler tables were the address of the function call.  So they
wouldn't match.

Just subtracting 8 from the return address will give the address of
the function call.  Easy enough.

Signed-off-by: Corey Minyard <cminyard@mvista.com>
[david.daney@cavium.com: Adjusted code comment and patch Subject.]
Signed-off-by: David Daney <david.daney@cavium.com>
---
 arch/mips/kernel/mcount.S | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/arch/mips/kernel/mcount.S b/arch/mips/kernel/mcount.S
index a03e93c..539b629 100644
--- a/arch/mips/kernel/mcount.S
+++ b/arch/mips/kernel/mcount.S
@@ -83,7 +83,7 @@ _mcount:
 	PTR_S	MCOUNT_RA_ADDRESS_REG, PT_R12(sp)
 #endif
 
-	move	a0, ra		/* arg1: self return address */
+	PTR_SUBU a0, ra, 8	/* arg1: self address */
 	.globl ftrace_call
 ftrace_call:
 	nop	/* a placeholder for the call to a real tracing function */
-- 
1.7.11.7

Re: [PATCH] mips/ftrace: Fix function tracing return address to match

[Steven Rostedt]

On Mon, 2013-07-15 at 15:17 -0700, David Daney wrote:
> From: Corey Minyard <cminyard@mvista.com>
> 
> Dynamic function tracing was not working on MIPS.  When doing dynamic
> tracing, the tracer attempts to match up the passed in address with
> the one the compiler creates in the mcount tables.  The MIPS code was
> passing in the return address from the tracing function call, but the
> compiler tables were the address of the function call.  So they
> wouldn't match.
> 
> Just subtracting 8 from the return address will give the address of
> the function call.  Easy enough.
> 
> Signed-off-by: Corey Minyard <cminyard@mvista.com>
> [david.daney@cavium.com: Adjusted code comment and patch Subject.]
> Signed-off-by: David Daney <david.daney@cavium.com>
> ---
>  arch/mips/kernel/mcount.S | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/arch/mips/kernel/mcount.S b/arch/mips/kernel/mcount.S
> index a03e93c..539b629 100644
> --- a/arch/mips/kernel/mcount.S
> +++ b/arch/mips/kernel/mcount.S
> @@ -83,7 +83,7 @@ _mcount:
>  	PTR_S	MCOUNT_RA_ADDRESS_REG, PT_R12(sp)
>  #endif
>  
> -	move	a0, ra		/* arg1: self return address */
> +	PTR_SUBU a0, ra, 8	/* arg1: self address */
>  	.globl ftrace_call
>  ftrace_call:
>  	nop	/* a placeholder for the call to a real tracing function */

I applied this patch to my Yeeloong Lemote laptop and it causes the
system to crash. Not sure why. I'll try to investigate.

-- Steve

Re: [PATCH] mips/ftrace: Fix function tracing return address to match

[David Daney]

On 07/18/2013 01:11 PM, Steven Rostedt wrote:
> On Mon, 2013-07-15 at 15:17 -0700, David Daney wrote:
>> From: Corey Minyard <cminyard@mvista.com>
>>
>> Dynamic function tracing was not working on MIPS.  When doing dynamic
>> tracing, the tracer attempts to match up the passed in address with
>> the one the compiler creates in the mcount tables.  The MIPS code was
>> passing in the return address from the tracing function call, but the
>> compiler tables were the address of the function call.  So they
>> wouldn't match.
>>
>> Just subtracting 8 from the return address will give the address of
>> the function call.  Easy enough.
>>
>> Signed-off-by: Corey Minyard <cminyard@mvista.com>
>> [david.daney@cavium.com: Adjusted code comment and patch Subject.]
>> Signed-off-by: David Daney <david.daney@cavium.com>
>> ---
>>   arch/mips/kernel/mcount.S | 2 +-
>>   1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/arch/mips/kernel/mcount.S b/arch/mips/kernel/mcount.S
>> index a03e93c..539b629 100644
>> --- a/arch/mips/kernel/mcount.S
>> +++ b/arch/mips/kernel/mcount.S
>> @@ -83,7 +83,7 @@ _mcount:
>>   	PTR_S	MCOUNT_RA_ADDRESS_REG, PT_R12(sp)
>>   #endif
>>
>> -	move	a0, ra		/* arg1: self return address */
>> +	PTR_SUBU a0, ra, 8	/* arg1: self address */
>>   	.globl ftrace_call
>>   ftrace_call:
>>   	nop	/* a placeholder for the call to a real tracing function */
>
> I applied this patch to my Yeeloong Lemote laptop and it causes the
> system to crash. Not sure why. I'll try to investigate.
>

There is an mcount ABI difference based on which GCC version you are 
using, although I wouldn't think it would effect this bit.

We are using GCC-4.7 FWIW.

David Daney


> -- Steve
>
>
>
>

Re: [PATCH] mips/ftrace: Fix function tracing return address to match

[Steven Rostedt]

On Thu, 2013-07-18 at 13:14 -0700, David Daney wrote:

> There is an mcount ABI difference based on which GCC version you are 
> using, although I wouldn't think it would effect this bit.
> 
> We are using GCC-4.7 FWIW.
> 
> David Daney

I'm using 4.6.3 which I downloaded from
https://www.kernel.org/pub/tools/crosstool/files/bin/x86_64/4.6.3/

I can down load the 4.7 version and see if that makes a difference.

-- Steve

Re: [PATCH] mips/ftrace: Fix function tracing return address to match

[Steven Rostedt]

On Thu, 2013-07-18 at 16:26 -0400, Steven Rostedt wrote:
> I can down load the 4.7 version and see if that makes a difference.

I guess I can't. There is no 4.7 version for MIPS :-p

https://www.kernel.org/pub/tools/crosstool/files/bin/x86_64/4.7.3/

-- Steve

Re: [PATCH] mips/ftrace: Fix function tracing return address to match

[David Daney]

On 07/18/2013 01:26 PM, Steven Rostedt wrote:
> On Thu, 2013-07-18 at 13:14 -0700, David Daney wrote:
>
>> There is an mcount ABI difference based on which GCC version you are
>> using, although I wouldn't think it would effect this bit.
>>
>> We are using GCC-4.7 FWIW.
>>
>> David Daney
>
> I'm using 4.6.3 which I downloaded from
> https://www.kernel.org/pub/tools/crosstool/files/bin/x86_64/4.6.3/

I would expect that version to be fine too.  The option in question is 
-mmcount-ra-address, which I added to GCC-4.5

>
> I can down load the 4.7 version and see if that makes a difference.
>
> -- Steve
>
>
>

Re: [PATCH] mips/ftrace: Fix function tracing return address to match

[David Daney]

On 07/18/2013 01:26 PM, Steven Rostedt wrote:
> On Thu, 2013-07-18 at 13:14 -0700, David Daney wrote:
>
>> There is an mcount ABI difference based on which GCC version you are
>> using, although I wouldn't think it would effect this bit.
>>
>> We are using GCC-4.7 FWIW.
>>
>> David Daney
>
> I'm using 4.6.3 which I downloaded from
> https://www.kernel.org/pub/tools/crosstool/files/bin/x86_64/4.6.3/

I would expect that version to be fine too.  The option in question is 
-mmcount-ra-address, which I added to GCC-4.5

>
> I can down load the 4.7 version and see if that makes a difference.
>
> -- Steve
>
>
>

Re: [PATCH] mips/ftrace: Fix function tracing return address to match

[Corey Minyard]

On 07/18/2013 03:11 PM, Steven Rostedt wrote:
> On Mon, 2013-07-15 at 15:17 -0700, David Daney wrote:
>> From: Corey Minyard <cminyard@mvista.com>
>>
>> Dynamic function tracing was not working on MIPS.  When doing dynamic
>> tracing, the tracer attempts to match up the passed in address with
>> the one the compiler creates in the mcount tables.  The MIPS code was
>> passing in the return address from the tracing function call, but the
>> compiler tables were the address of the function call.  So they
>> wouldn't match.
>>
>> Just subtracting 8 from the return address will give the address of
>> the function call.  Easy enough.
>>
>> Signed-off-by: Corey Minyard <cminyard@mvista.com>
>> [david.daney@cavium.com: Adjusted code comment and patch Subject.]
>> Signed-off-by: David Daney <david.daney@cavium.com>
>> ---
>>   arch/mips/kernel/mcount.S | 2 +-
>>   1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/arch/mips/kernel/mcount.S b/arch/mips/kernel/mcount.S
>> index a03e93c..539b629 100644
>> --- a/arch/mips/kernel/mcount.S
>> +++ b/arch/mips/kernel/mcount.S
>> @@ -83,7 +83,7 @@ _mcount:
>>   	PTR_S	MCOUNT_RA_ADDRESS_REG, PT_R12(sp)
>>   #endif
>>   
>> -	move	a0, ra		/* arg1: self return address */
>> +	PTR_SUBU a0, ra, 8	/* arg1: self address */
>>   	.globl ftrace_call
>>   ftrace_call:
>>   	nop	/* a placeholder for the call to a real tracing function */
> I applied this patch to my Yeeloong Lemote laptop and it causes the
> system to crash. Not sure why. I'll try to investigate.
>
> -- Steve
>
>
That is bizarre.  That value should just be used for comparisons, it 
isn't dereferenced or anything like that.  What tracing do you have enabled?

-corey