On 2016/11/04 at 16:23, Dave Young wrote:
> Hi Xunlei,
>
>>> If change the logic a bit, it will be clear:
>>>
>>> is_trace_prepared()
>>> {
>>> this function only check if tracing is enabled and events matched.
>>> }
>>>
>>> prepare_trace()
>>> {
>>> enable and prepare trace
>>> }
>>>
>>> parse_trace_data()
>>> {
>>>
>>> }
>>>
>>> if is_trace_prepared == true; then
>>> parse_trace_data
>>> else
>>> prepare_trace
>>> fi
>>>
>>> So in cmdline hook is_trace_prepared is false so the script only prepare
>>> trace and enable tracing, in latter hooks it will parse trace data.
>>>
>>> But there should be another function to disable trace before pivot root.
>> OK
>>
>> We should find a neat way to disable the trace function, what do you think the following way?
>> Change show_memstats() and make_trace_mem() as follows:
>> make_trace_mem "hook cmdline" '1+:mem' '1+:iomem' '3+:slab' '4+:komem'
>> ... ...
>> make_trace_mem "hook pre-pivot" '1+:mem' '1+:iomem' '3+:slab' '4+:komem'
>> make_trace_mem "hook pre-pivot" '4+:komemfinish' # cleanup trace used by showkomem after use.
>>
>> show_memstats()
>> ... ...
>>
>> + komem)
>> + showkomem
>> + ;;
>> + komemfinish)
>> + # disable trace after use.
>> + showkomem finish
>> + ;;
>> esac
>>
> Hmm, I'm not sure it is worth, maybe add a function like
> cleanup_trace_mem() is enough and we can add it just before cleanup
> hook which is designed for cleanup purpose, cleanup hook is sourced
> after pre-pivot so that it will still work for adding trace in any
> pre-pivot scripts.
OK
I'll send v2 later, thanks for all the valuable suggestions.
Regards,
Xunlei
--
To unsubscribe from this list: send the line "unsubscribe initramfs" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
