Date:   Mon, 17 Apr 2017 11:18:43 -0600
From:   "Baicar, Tyler" <tbaicar@...eaurora.org>
To:     Xie XiuQi <xiexiuqi@...wei.com>, christoffer.dall@...aro.org,
        marc.zyngier@....com, catalin.marinas@....com, will.deacon@....com,
        james.morse@....com, fu.wei@...aro.org, rostedt@...dmis.org,
        hanjun.guo@...aro.org, shiju.jose@...wei.com
Cc:     linux-arm-kernel@...ts.infradead.org, kvmarm@...ts.cs.columbia.edu,
        kvm@...r.kernel.org, linux-kernel@...r.kernel.org,
        linux-acpi@...r.kernel.org, gengdongjiu@...wei.com,
        zhengqiang10@...wei.com, wuquanming@...wei.com,
        wangxiongfeng2@...wei.com
Subject: Re: [PATCH v3 1/8] trace: ras: add ARM processor error information
 trace event

On 4/16/2017 9:16 PM, Xie XiuQi wrote:
> On 2017/4/17 11:08, Xie XiuQi wrote:
>>> On 3/30/2017 4:31 AM, Xie XiuQi wrote:
>>>> Add a new trace event for ARM processor error information, so that
>>>> the user will know what error occurred. With this information the
>>>> user may take appropriate action.
>>>>
>>>> These trace events are consistent with the ARM processor error
>>>> information table which defined in UEFI 2.6 spec section N.2.4.4.1.
>>>>
>>>> ---
>>>> v2: add trace enabled condition as Steven's suggestion.
>>>>       fix a typo.
>>>> ---
>>>>
>>>> Cc: Steven Rostedt <rostedt@...dmis.org>
>>>> Cc: Tyler Baicar <tbaicar@...eaurora.org>
>>>> Signed-off-by: Xie XiuQi <xiexiuqi@...wei.com>
>>>> ---
>>>
...
>>>> +/*
>>>> + * First define the enums in MM_ACTION_RESULT to be exported to userspace
>>>> + * via TRACE_DEFINE_ENUM().
>>>> + */
>>>> +#undef EM
>>>> +#undef EMe
>>>> +#define EM(a, b) TRACE_DEFINE_ENUM(a);
>>>> +#define EMe(a, b)    TRACE_DEFINE_ENUM(a);
>>>> +
>>>> +ARM_PROC_ERR_TYPE
>>>> +ARM_PROC_ERR_FLAGS
>>> Are the above two lines supposed to be here?
>>>> +
>>>> +/*
>>>> + * Now redefine the EM() and EMe() macros to map the enums to the strings
>>>> + * that will be printed in the output.
>>>> + */
>>>> +#undef EM
>>>> +#undef EMe
>>>> +#define EM(a, b)        { a, b },
>>>> +#define EMe(a, b)    { a, b }
>>>> +
>>>> +TRACE_EVENT(arm_proc_err,
>>> I think it would be better to keep this similar to the naming of the current RAS trace events (right now we have mc_event, arm_event, aer_event, etc.). I would suggest using "arm_err_info_event" since this is handling the error information structures of the arm errors.
>>>> +
>>>> +    TP_PROTO(const struct cper_arm_err_info *err),
>>>> +
>>>> +    TP_ARGS(err),
>>>> +
>>>> +    TP_STRUCT__entry(
>>>> +        __field(u8, type)
>>>> +        __field(u16, multiple_error)
>>>> +        __field(u8, flags)
>>>> +        __field(u64, error_info)
>>>> +        __field(u64, virt_fault_addr)
>>>> +        __field(u64, physical_fault_addr)
>>> Validation bits should also be a part of this structure that way user space tools will know which of these fields are valid.
>> Could we use the default value to check the validation which we have checked in TP_fast_assign?
Yes, true...I guess we really don't need the validation bits then.
>>>> +    ),
>>>> +
>>>> +    TP_fast_assign(
>>>> +        __entry->type = err->type;
>>>> +
>>>> +        if (err->validation_bits & CPER_ARM_INFO_VALID_MULTI_ERR)
>>>> +            __entry->multiple_error = err->multiple_error;
>>>> +        else
>>>> +            __entry->multiple_error = ~0;
>>>> +
>>>> +        if (err->validation_bits & CPER_ARM_INFO_VALID_FLAGS)
>>>> +            __entry->flags = err->flags;
>>>> +        else
>>>> +            __entry->flags = ~0;
>>>> +
>>>> +        if (err->validation_bits & CPER_ARM_INFO_VALID_ERR_INFO)
>>>> +            __entry->error_info = err->error_info;
>>>> +        else
>>>> +            __entry->error_info = 0ULL;
>>>> +
>>>> +        if (err->validation_bits & CPER_ARM_INFO_VALID_VIRT_ADDR)
>>>> +            __entry->virt_fault_addr = err->virt_fault_addr;
>>>> +        else
>>>> +            __entry->virt_fault_addr = 0ULL;
>>>> +
>>>> +        if (err->validation_bits & CPER_ARM_INFO_VALID_PHYSICAL_ADDR)
>>>> +            __entry->physical_fault_addr = err->physical_fault_addr;
>>>> +        else
>>>> +            __entry->physical_fault_addr = 0ULL;
>>>> +    ),
>>>> +
>>>> +    TP_printk("ARM Processor Error: type %s; count: %u; flags: %s;"
>>> I think the "ARM Processor Error:" part of this should just be removed. Here's the output with this removed and the trace event renamed to arm_err_info_event. I think this looks much cleaner and matches the style used with the arm_event.
>>>
>>>            <idle>-0     [020] .ns.   366.592434: arm_event: affinity level: 2; MPIDR: 0000000000000000; MIDR: 00000000510f8000; running state: 1; PSCI state: 0
>>>            <idle>-0     [020] .ns.   366.592437: arm_err_info_event: type cache error; count: 0; flags: 0x3; error info: 0000000000c20058; virtual address: 0000000000000000; physical address: 0000000000000000
> As this section is ARM Processor Error Section, how about use arm_proc_err_event?
This is not for the ARM Processor Error Section, that is what the 
arm_event is handling. What you are adding this trace support for here 
is called the ARM Processor Error Information (UEFI 2.6 spec section 
N.2.4.4.1). So I think your trace event here should be called 
arm_err_info_event. This will also be consistent with the other two 
trace events that I'm planning on adding:

arm_ctx_info_event: ARM Processor Context Information (UEFI 2.6 section 
N.2.4.4.2)
arm_vendor_info_event: This is the "Vendor Specific Error Information" 
in the ARM Processor Error Section (Table 260). It's possible I may just 
add this into the arm_event trace event, but I haven't looked into it 
enough yet.

Thanks,
Tyler

-- 
Qualcomm Datacenter Technologies, Inc. as an affiliate of Qualcomm Technologies, Inc.
Qualcomm Technologies, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project.

Powered by blists - more mailing lists