Re: [RFC][PATCH] convert block trace points to TRACE_EVENT()

[Li Zefan <lizf@cn.fujitsu.com>]

Frederic Weisbecker wrote:
> On Thu, May 14, 2009 at 02:20:03PM +0800, Li Zefan wrote:
>> TRACE_EVENT is a more generic way to define tracepoints. Doing so adds
>> these new capabilities to this tracepoint:
>>
>>   - zero-copy and per-cpu splice() tracing
>>   - binary tracing without printf overhead
>>   - structured logging records exposed under /debug/tracing/events
>>   - trace events embedded in function tracer output and other plugins
>>   - user-defined, per tracepoint filter expressions
>>   ...
>>
>> Cons and problems:
>>
>>   - no dev_t info for the output of plug, unplug_timer and unplug_io events.
>>     no dev_t info for getrq and sleeprq events if bio == NULL.
>>     no dev_t info for rq_abort,...,rq_requeue events if rq->rq_disk == NULL.
>>
>>   - for large packet commands, only 16 bytes of the command will be output.
>>     Because TRACE_EVENT doesn't support dynamic-sized arrays, though it
>>     supports dynamic-sized strings.
>>
>>   - a packet command is converted to a string in TP_assign, not TP_print.
>>     While blktrace do the convertion just before output.
>>
>>   - in blktrace, an event can have 2 different print formats, but a TRACE_EVENT
>>     has a unique format. (see the output of getrq and rq_insert)
> 
> I'm starting to think it would be nice to choose between several outputs
> in a trace event.
> Ie: perhaps we need a kind of per event flag, something simple, just to
> choose between several TP_printk output. Not sure how much it would
> (non) trivial to implement though...
> 

If a trace event wants several TP_printk output, probably it wants different
structs of trace entry, otherwise we'll be wasting memory.