Re: [RFC PATCH 2/5] tracing/events: nicer print format for parsing

[Ingo Molnar <mingo@elte.hu>]

* Steven Rostedt <rostedt@goodmis.org> wrote:

> 
> On Wed, 10 Jun 2009, Ingo Molnar wrote:
> > 
> > And i kind of like the whole notion on a design level as weell: the 
> > kernel exporting C source code for tools :-)
> > 
> > 	Ingo
> > 
> > ------------------>
> > 
> > struct record {
> > 	unsigned short common_type;
> > 	unsigned char common_flags;
> > 	unsigned char common_preempt;
> > 	int common_pid;
> > 	int common_tgid;
> > 	int dev;
> > 	unsigned long long sector;
> > 	unsigned int nr_sector;
> > 	char rwbs[6];
> > 	char comm[16];
> > } this_record = { 1, 2, 3, 4, 5, 6, 7, 8, { 'a', }, "abc" };
> > 
> > void main(void)
> > {
> > 	struct record *REC = &this_record;
> > 
> > 	printf("%d,%d %s %llu + %u [%s]", ((unsigned int) ((REC->dev) >> 20)), ((unsigned int) ((REC->dev) & ((1U << 20) - 1))), REC->rwbs, (unsigned long long)REC->sector, REC->nr_sector, REC->comm);
> > }
> 
> I actually tried this first. But it breaks once we start getting types 
> into the code:
> 
> print fmt: "call_site=%lx ptr=%p bytes_req=%zu bytes_alloc=%zu gfp_flags=%s",
>           REC->call_site, REC->ptr, REC->bytes_req, REC->bytes_alloc, 
>           (REC->gfp_flags) ? ({ static const struct trace_print_flags 
>           flags[] = { {(unsigned long)(((gfp_t)0x10u) | ((gfp_t)0x40u) |
>                       ((gfp_t)0x80u) | ((gfp_t)0x20000u) | ((gfp_t)0x02u) | 
>                        ((gfp_t)0x100000u)), "GFP_HIGHUSER_MOVABLE"}, [...]
> 
> Will break on gfp_t. [...]

It wont break if we introduce a couple of common-sense types into 
the parsing/translation code. gfp_t is well-known.

Modules wont be able to generate new dynamic types - but that's OK i 
think, existing C types and common kernel types (and anything else 
we add) ought to be plenty enough.

> [...] We also have cases where the enum name slips in too:
> 
> print fmt: "softirq=%d action=%s", REC->vec, ({ static const struct trace_print_flags symbols[] =
>             { { HI_SOFTIRQ, "HI" }, { TIMER_SOFTIRQ, "TIMER" }, { NET_TX_SOFTIRQ, "NET_TX" },
>               { NET_RX_SOFTIRQ, "NET_RX" }, { BLOCK_SOFTIRQ, "BLOCK" },
>               { TASKLET_SOFTIRQ, "TASKLET" }, { SCHED_SOFTIRQ, "SCHED" },
>               { HRTIMER_SOFTIRQ, "HRTIMER" }, { RCU_SOFTIRQ, "RCU" },
>               { -1, ((void *)0) }}; ftrace_print_symbols_seq(p, REC->vec, symbols); })
> 
> Yes we can add special types for things like gfp_t, but as we get 
> more and more users of TRACE_EVENT, the tools will never be able 
> to keep up.

i still disagree. The tool will have to know about gfp_t in the tag 
language too. So there's always going to be a constant expansion of 
the type space.

The point is that the number of new types is an order of magnitude 
less than the number of new tracepoints.

Also, with tools like perf in the kernel repo under tools/perf/, 
we'll be able to keep up with mainline types very flexibly and very 
accurately.

	Ingo