[PATCH v5 0/2] Update mce_record tracepoint

[PATCH v5 0/2] Update mce_record tracepoint

[Avadhut Naik]

This patchset updates the mce_record tracepoint so that the recently added
fields of struct mce are exported through it to userspace.

The first patch adds PPIN (Protected Processor Inventory Number) field to
the tracepoint.

The second patch adds the microcode field (Microcode Revision) to the
tracepoint.

Changes in v2:
 - Export microcode field (Microcode Revision) through the tracepoiont in
   addition to PPIN.

Changes in v3:
 - Change format specifier for microcode revision from %u to %x
 - Fix tab alignments
 - Add Reviewed-by: Sohil Mehta <sohil.mehta@intel.com>

Changes in v4:
 - Update commit messages to reflect the reason for the fields being
   added to the tracepoint.
 - Add comment to explicitly state the type of information that should
   be added to the tracepoint.
 - Add Reviewed-by: Steven Rostedt (Google) <rostedt@goodmis.org>

Changes in v5:
 - Changed "MICROCODE REVISION" to just "MICROCODE".
 - Changed words which are not acronyms from ALL CAPS to no caps.
 - Added Reviewed-by: Tony Luck <tony.luck@intel.com>

[NOTE:

 - Since only caps of words which are not acronyms have been changed in
   this version and the word "REVISION" has been removed i.e. changes are
   very minor, have retained the the below tags received for previous
   versions:
    Reviewed-by: Sohil Mehta <sohil.mehta@intel.com>
    Reviewed-by: Steven Rostedt (Google) <rostedt@goodmis.org>
    Reviewed-by: Tony Luck <tony.luck@intel.com>]

Avadhut Naik (2):
  tracing: Include PPIN in mce_record tracepoint
  tracing: Include Microcode Revision in mce_record tracepoint

 include/trace/events/mce.h | 18 ++++++++++++++++--
 1 file changed, 16 insertions(+), 2 deletions(-)


base-commit: 9699768066a9b12db247d5ee9276cf298b67528d
-- 
2.34.1

[PATCH v5 1/2] tracing: Include PPIN in mce_record tracepoint

[Avadhut Naik]

Machine Check Error information from struct mce is exported to userspace
through the mce_record tracepoint.

Currently, however, the PPIN (Protected Processor Inventory Number) field
of struct mce is not exported through the tracepoint.

Export PPIN through the tracepoint as it provides a unique identifier for
the system (or socket in case of multi-socket systems) on which the MCE
has been received.

Additionally, also change words which are not acronyms from uppercase
(ALL CAPS) to lowercase in tracepoint's output.

Signed-off-by: Avadhut Naik <avadhut.naik@amd.com>
Reviewed-by: Sohil Mehta <sohil.mehta@intel.com>
Reviewed-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Reviewed-by: Tony Luck <tony.luck@intel.com>
---
 include/trace/events/mce.h | 13 ++++++++++++-
 1 file changed, 12 insertions(+), 1 deletion(-)

diff --git a/include/trace/events/mce.h b/include/trace/events/mce.h
index 1391ada0da3b..2e339ac9d706 100644
--- a/include/trace/events/mce.h
+++ b/include/trace/events/mce.h
@@ -9,6 +9,14 @@
 #include <linux/tracepoint.h>
 #include <asm/mce.h>
 
+/*
+ * MCE Event Record.
+ *
+ * Only very relevant and transient information which cannot be
+ * gathered from a system by any other means or which can only be
+ * acquired arduously should be added to this record.
+ */
+
 TRACE_EVENT(mce_record,
 
 	TP_PROTO(struct mce *m),
@@ -25,6 +33,7 @@ TRACE_EVENT(mce_record,
 		__field(	u64,		ipid		)
 		__field(	u64,		ip		)
 		__field(	u64,		tsc		)
+		__field(	u64,		ppin		)
 		__field(	u64,		walltime	)
 		__field(	u32,		cpu		)
 		__field(	u32,		cpuid		)
@@ -45,6 +54,7 @@ TRACE_EVENT(mce_record,
 		__entry->ipid		= m->ipid;
 		__entry->ip		= m->ip;
 		__entry->tsc		= m->tsc;
+		__entry->ppin		= m->ppin;
 		__entry->walltime	= m->time;
 		__entry->cpu		= m->extcpu;
 		__entry->cpuid		= m->cpuid;
@@ -55,7 +65,7 @@ TRACE_EVENT(mce_record,
 		__entry->cpuvendor	= m->cpuvendor;
 	),
 
-	TP_printk("CPU: %d, MCGc/s: %llx/%llx, MC%d: %016Lx, IPID: %016Lx, ADDR/MISC/SYND: %016Lx/%016Lx/%016Lx, RIP: %02x:<%016Lx>, TSC: %llx, PROCESSOR: %u:%x, TIME: %llu, SOCKET: %u, APIC: %x",
+	TP_printk("CPU: %d, MCGc/s: %llx/%llx, MC%d: %016Lx, IPID: %016Lx, ADDR/MISC/SYND: %016Lx/%016Lx/%016Lx, RIP: %02x:<%016Lx>, TSC: %llx, PPIN: %llx, processor: %u:%x, time: %llu, socket: %u, APIC: %x",
 		__entry->cpu,
 		__entry->mcgcap, __entry->mcgstatus,
 		__entry->bank, __entry->status,
@@ -63,6 +73,7 @@ TRACE_EVENT(mce_record,
 		__entry->addr, __entry->misc, __entry->synd,
 		__entry->cs, __entry->ip,
 		__entry->tsc,
+		__entry->ppin,
 		__entry->cpuvendor, __entry->cpuid,
 		__entry->walltime,
 		__entry->socketid,
-- 
2.34.1

[PATCH v5 2/2] tracing: Include Microcode Revision in mce_record tracepoint

[Avadhut Naik]

Currently, the microcode field (Microcode Revision) of struct mce is not
exported to userspace through the mce_record tracepoint.

Knowing the microcode version on which the MCE was received is critical
information for debugging. If the version is not recorded, later attempts
to acquire the version might result in discrepancies since it can be
changed at runtime.

Export microcode version through the tracepoint to prevent ambiguity over
the active version on the system when the MCE was received.

Signed-off-by: Avadhut Naik <avadhut.naik@amd.com>
Reviewed-by: Sohil Mehta <sohil.mehta@intel.com>
Reviewed-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Reviewed-by: Tony Luck <tony.luck@intel.com>
---
 include/trace/events/mce.h | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/include/trace/events/mce.h b/include/trace/events/mce.h
index 2e339ac9d706..c7d0edc778c3 100644
--- a/include/trace/events/mce.h
+++ b/include/trace/events/mce.h
@@ -42,6 +42,7 @@ TRACE_EVENT(mce_record,
 		__field(	u8,		cs		)
 		__field(	u8,		bank		)
 		__field(	u8,		cpuvendor	)
+		__field(	u32,		microcode	)
 	),
 
 	TP_fast_assign(
@@ -63,9 +64,10 @@ TRACE_EVENT(mce_record,
 		__entry->cs		= m->cs;
 		__entry->bank		= m->bank;
 		__entry->cpuvendor	= m->cpuvendor;
+		__entry->microcode	= m->microcode;
 	),
 
-	TP_printk("CPU: %d, MCGc/s: %llx/%llx, MC%d: %016Lx, IPID: %016Lx, ADDR/MISC/SYND: %016Lx/%016Lx/%016Lx, RIP: %02x:<%016Lx>, TSC: %llx, PPIN: %llx, processor: %u:%x, time: %llu, socket: %u, APIC: %x",
+	TP_printk("CPU: %d, MCGc/s: %llx/%llx, MC%d: %016Lx, IPID: %016Lx, ADDR/MISC/SYND: %016Lx/%016Lx/%016Lx, RIP: %02x:<%016Lx>, TSC: %llx, PPIN: %llx, processor: %u:%x, time: %llu, socket: %u, APIC: %x, microcode: %x",
 		__entry->cpu,
 		__entry->mcgcap, __entry->mcgstatus,
 		__entry->bank, __entry->status,
@@ -77,7 +79,8 @@ TRACE_EVENT(mce_record,
 		__entry->cpuvendor, __entry->cpuid,
 		__entry->walltime,
 		__entry->socketid,
-		__entry->apicid)
+		__entry->apicid,
+		__entry->microcode)
 );
 
 #endif /* _TRACE_MCE_H */
-- 
2.34.1

Re: [PATCH v5 0/2] Update mce_record tracepoint

[Sohil Mehta]

On 3/28/2024 11:04 AM, Avadhut Naik wrote:

>  - Since only caps of words which are not acronyms have been changed in
>    this version and the word "REVISION" has been removed i.e. changes are
>    very minor, have retained the the below tags received for previous
>    versions:
>     Reviewed-by: Sohil Mehta <sohil.mehta@intel.com>

> Avadhut Naik (2):
>   tracing: Include PPIN in mce_record tracepoint
>   tracing: Include Microcode Revision in mce_record tracepoint


The patches look good to me.

Re: [PATCH v5 0/2] Update mce_record tracepoint

[Naik, Avadhut]

On 3/28/2024 13:14, Sohil Mehta wrote:
> On 3/28/2024 11:04 AM, Avadhut Naik wrote:
> 
>>  - Since only caps of words which are not acronyms have been changed in
>>    this version and the word "REVISION" has been removed i.e. changes are
>>    very minor, have retained the the below tags received for previous
>>    versions:
>>     Reviewed-by: Sohil Mehta <sohil.mehta@intel.com>
> 
>> Avadhut Naik (2):
>>   tracing: Include PPIN in mce_record tracepoint
>>   tracing: Include Microcode Revision in mce_record tracepoint
> 
> 
> The patches look good to me.

Kindly ignore this set.

As pointed out by Sohit, have missed adding linux-kernel@vger.kernel.org

Will resend the set.

-- 
Thanks,
Avadhut Naik

[PATCH] x86/mce: Clean up TP_printk() output line of the mce_record tracepoint

[Ingo Molnar]

* Avadhut Naik <avadhut.naik@amd.com> wrote:

>  
> +/*
> + * MCE Event Record.
> + *
> + * Only very relevant and transient information which cannot be
> + * gathered from a system by any other means or which can only be
> + * acquired arduously should be added to this record.
> + */
> +
>  TRACE_EVENT(mce_record,
>  
>  	TP_PROTO(struct mce *m),
> @@ -25,6 +33,7 @@ TRACE_EVENT(mce_record,
>  		__field(	u64,		ipid		)
>  		__field(	u64,		ip		)
>  		__field(	u64,		tsc		)
> +		__field(	u64,		ppin		)
>  		__field(	u64,		walltime	)
>  		__field(	u32,		cpu		)
>  		__field(	u32,		cpuid		)
> @@ -45,6 +54,7 @@ TRACE_EVENT(mce_record,
>  		__entry->ipid		= m->ipid;
>  		__entry->ip		= m->ip;
>  		__entry->tsc		= m->tsc;
> +		__entry->ppin		= m->ppin;
>  		__entry->walltime	= m->time;
>  		__entry->cpu		= m->extcpu;
>  		__entry->cpuid		= m->cpuid;
> @@ -55,7 +65,7 @@ TRACE_EVENT(mce_record,
>  		__entry->cpuvendor	= m->cpuvendor;
>  	),
>  
> -	TP_printk("CPU: %d, MCGc/s: %llx/%llx, MC%d: %016Lx, IPID: %016Lx, ADDR/MISC/SYND: %016Lx/%016Lx/%016Lx, RIP: %02x:<%016Lx>, TSC: %llx, PROCESSOR: %u:%x, TIME: %llu, SOCKET: %u, APIC: %x",
> +	TP_printk("CPU: %d, MCGc/s: %llx/%llx, MC%d: %016Lx, IPID: %016Lx, ADDR/MISC/SYND: %016Lx/%016Lx/%016Lx, RIP: %02x:<%016Lx>, TSC: %llx, PPIN: %llx, processor: %u:%x, time: %llu, socket: %u, APIC: %x",

Please split out the other (capitalization) changes to the output into 
a separate patch.

- While at it, don't forget to:

   s/ADDR/MISC/SYND
    /addr/misc/synd

- Also, it's a bit weird that we have 'CPU' but also 'processor' 
  fields, why isn't it 'vendor' and 'CPUID'?

- Finally, why are some fields 'merged' as per ADDR/MISC/SYND, while 
  others are listed separately? All that have separate names should be 
  listed separately.

Ie. something like the patch below?

Thanks,

	Ingo

============>
From: Ingo Molnar <mingo@kernel.org>
Date: Fri, 29 Mar 2024 08:09:23 +0100
Subject: [PATCH] x86/mce: Clean up TP_printk() output line of the mce_record tracepoint

 - Only capitalize entries where that makes sense
 - Print separate values separately
 - Rename 'PROCESSOR' to vendor & CPUID

Signed-off-by: Ingo Molnar <mingo@kernel.org>
---
 include/trace/events/mce.h | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)

diff --git a/include/trace/events/mce.h b/include/trace/events/mce.h
index 1391ada0da3b..c5b0523f25ee 100644
--- a/include/trace/events/mce.h
+++ b/include/trace/events/mce.h
@@ -55,15 +55,18 @@ TRACE_EVENT(mce_record,
 		__entry->cpuvendor	= m->cpuvendor;
 	),
 
-	TP_printk("CPU: %d, MCGc/s: %llx/%llx, MC%d: %016Lx, IPID: %016Lx, ADDR/MISC/SYND: %016Lx/%016Lx/%016Lx, RIP: %02x:<%016Lx>, TSC: %llx, PROCESSOR: %u:%x, TIME: %llu, SOCKET: %u, APIC: %x",
+	TP_printk("CPU: %d, MCGc/s: %llx/%llx, MC%d: %016Lx, IPID: %016Lx, addr: %016Lx, misc: %016Lx, SYND: %016Lx, RIP: %02x:<%016Lx>, TSC: %llx, vendor: %u, CPUID: %x, time: %llu, socket: %u, APIC: %x",
 		__entry->cpu,
 		__entry->mcgcap, __entry->mcgstatus,
 		__entry->bank, __entry->status,
 		__entry->ipid,
-		__entry->addr, __entry->misc, __entry->synd,
+		__entry->addr,
+		__entry->misc,
+		__entry->synd,
 		__entry->cs, __entry->ip,
 		__entry->tsc,
-		__entry->cpuvendor, __entry->cpuid,
+		__entry->cpuvendor,
+		__entry->cpuid,
 		__entry->walltime,
 		__entry->socketid,
 		__entry->apicid)

[PATCH] x86/mce: Clean up TP_printk() output line of the mce_record tracepoint

[Naik, Avadhut]

On 3/29/2024 02:11, Ingo Molnar wrote:
> 
> Please split out the other (capitalization) changes to the output into 
> a separate patch.
> 
Okay. Will put the capitalization changes into a separate patch.

> - While at it, don't forget to:
> 
>    s/ADDR/MISC/SYND
>     /addr/misc/synd
>
These are actually acronyms for Address, Miscellaneous and Syndrome registers.

It was recommended to keep the acronyms in ALL CAPS. Hence, didn't change them.

https://lore.kernel.org/linux-edac/20240327205435.3667588-1-avadhut.naik@amd.com/T/#m0c04f1c0deaa0347af66653a5950aad5f6b320e7
 
> - Also, it's a bit weird that we have 'CPU' but also 'processor' 
>   fields, why isn't it 'vendor' and 'CPUID'?
> 
Think it has been this way since the tracepoint was created back
in 2009. Will modify the field per your suggestion.

> - Finally, why are some fields 'merged' as per ADDR/MISC/SYND, while 
>   others are listed separately? All that have separate names should be 
>   listed separately.
> 
Will separate the fields so that each is listed individually.
> Ie. something like the patch below?
> 
> Thanks,
> 
> 	Ingo
> 
> ============>
> From: Ingo Molnar <mingo@kernel.org>
> Date: Fri, 29 Mar 2024 08:09:23 +0100
> Subject: [PATCH] x86/mce: Clean up TP_printk() output line of the mce_record tracepoint
> 
>  - Only capitalize entries where that makes sense
>  - Print separate values separately
>  - Rename 'PROCESSOR' to vendor & CPUID
> 
> Signed-off-by: Ingo Molnar <mingo@kernel.org>
> ---
>  include/trace/events/mce.h | 9 ++++++---
>  1 file changed, 6 insertions(+), 3 deletions(-)
> 
> diff --git a/include/trace/events/mce.h b/include/trace/events/mce.h
> index 1391ada0da3b..c5b0523f25ee 100644
> --- a/include/trace/events/mce.h
> +++ b/include/trace/events/mce.h
> @@ -55,15 +55,18 @@ TRACE_EVENT(mce_record,
>  		__entry->cpuvendor	= m->cpuvendor;
>  	),
>  
> -	TP_printk("CPU: %d, MCGc/s: %llx/%llx, MC%d: %016Lx, IPID: %016Lx, ADDR/MISC/SYND: %016Lx/%016Lx/%016Lx, RIP: %02x:<%016Lx>, TSC: %llx, PROCESSOR: %u:%x, TIME: %llu, SOCKET: %u, APIC: %x",
> +	TP_printk("CPU: %d, MCGc/s: %llx/%llx, MC%d: %016Lx, IPID: %016Lx, addr: %016Lx, misc: %016Lx, SYND: %016Lx, RIP: %02x:<%016Lx>, TSC: %llx, vendor: %u, CPUID: %x, time: %llu, socket: %u, APIC: %x",
>  		__entry->cpu,
>  		__entry->mcgcap, __entry->mcgstatus,
>  		__entry->bank, __entry->status,
>  		__entry->ipid,
> -		__entry->addr, __entry->misc, __entry->synd,
> +		__entry->addr,
> +		__entry->misc,
> +		__entry->synd,
>  		__entry->cs, __entry->ip,
>  		__entry->tsc,
> -		__entry->cpuvendor, __entry->cpuid,
> +		__entry->cpuvendor,
> +		__entry->cpuid,
>  		__entry->walltime,
>  		__entry->socketid,
>  		__entry->apicid)

-- 
Thanks,
Avadhut Naik

RE: [PATCH] x86/mce: Clean up TP_printk() output line of the mce_record tracepoint

[Luck, Tony]

>> - While at it, don't forget to:
>> 
>>    s/ADDR/MISC/SYND
>>     /addr/misc/synd
>>
> These are actually acronyms for Address, Miscellaneous and Syndrome registers.

They look like abbreviations, not acronyms to me.

-Tony

[PATCH] x86/mce: Clean up TP_printk() output line of the mce_record tracepoint

[Naik, Avadhut]

On 3/29/2024 11:50, Luck, Tony wrote:
>>> - While at it, don't forget to:
>>>
>>>    s/ADDR/MISC/SYND
>>>     /addr/misc/synd
>>>
>> These are actually acronyms for Address, Miscellaneous and Syndrome registers.
> 
> They look like abbreviations, not acronyms to me.
> 
> -Tony

Yes, they are actually abbreviations. Wrong choice of words on my part.
Was under the impression that Boris' recommendation also applied to
abbreviations. Will change them though and resubmit.

-- 
Thanks,
Avadhut Naik

Re: [PATCH] x86/mce: Clean up TP_printk() output line of the mce_record tracepoint

[Ingo Molnar]

* Naik, Avadhut <avadnaik@amd.com> wrote:

> On 3/29/2024 02:11, Ingo Molnar wrote:
> > 
> > Please split out the other (capitalization) changes to the output into 
> > a separate patch.
> > 
> Okay. Will put the capitalization changes into a separate patch.
> 
> > - While at it, don't forget to:
> > 
> >    s/ADDR/MISC/SYND
> >     /addr/misc/synd
> >
> These are actually acronyms for Address, Miscellaneous and Syndrome registers.

I kept SYND capitalized in the patch:

> > +	TP_printk("CPU: %d, MCGc/s: %llx/%llx, MC%d: %016Lx, IPID: %016Lx, addr: %016Lx, misc: %016Lx, SYND: %016Lx, RIP: %02x:<%016Lx>, TSC: %llx, vendor: %u, CPUID: %x, time: %llu, socket: %u, APIC: %x",

But I guess ADDR and MISC are fine too.

To move forward with this I've committed the cleanup in tip:x86/cpu:

  ac5e80e94f5c x86/mce: Clean up TP_printk() output line of the 'mce_record' tracepoint

... and please post extensions of the tracepoint on top of that sane(er) code.

Thanks,

	Ingo