Re: [PATCH 0/3] [RFC] init, tracing: Add initcall trace events
From: Joel Fernandes (Google)
Date: Mon Mar 26 2018 - 21:50:40 EST
Hi Steve,
On Fri, Mar 23, 2018 at 8:02 AM, Steven Rostedt <rostedt@xxxxxxxxxxx> wrote:
> A while ago we had a boot tracer. But it was eventually removed:
> commit 30dbb20e68e6f ("tracing: Remove boot tracer").
>
> The rational was because there is already a initcall_debug boot option
> that causes printk()s of all the initcall functions.
>
> The problem with the initcall_debug option is that printk() is awfully slow,
> and makes it difficult to see the real impact of initcalls. Mainly because
> a single printk() is usually slower than most initcall functions.
>
> Instead of bringing back the boot tracer, adding trace events around the
> initcall functions, and even one to denote which level the initcall
> functions are being called from, adds the necessary information to
> analyze the initcalls without the high overhead of printk()s, that
> can substantially slow down the boot process.
>
> Another positive, is that the console initcall functions themselves
> can also be traced. The timestamps are not operational at that time
> but you can see which consoles are being registered. I saw this on
> one of my test boxes:
>
> <idle>-0 [000] ...1 0.000000: initcall_level: level=console
> <idle>-0 [000] ...1 0.000000: initcall_start: func=con_init+0x0/0x224
> <idle>-0 [000] ...1 0.000000: initcall_finish: func=con_init+0x0/0x224 ret=0
> <idle>-0 [000] ...1 0.000000: initcall_start: func=hvc_console_init+0x0/0x19
> <idle>-0 [000] ...1 0.000000: initcall_finish: func=hvc_console_init+0x0/0x19 ret=0
> <idle>-0 [000] ...1 0.000000: initcall_start: func=xen_cons_init+0x0/0x60
> <idle>-0 [000] ...1 0.000000: initcall_finish: func=xen_cons_init+0x0/0x60 ret=0
> <idle>-0 [000] ...1 0.000000: initcall_start: func=univ8250_console_init+0x0/0x2d
> <idle>-0 [000] ...1 0.000000: initcall_finish: func=univ8250_console_init+0x0/0x2d ret=0
Will this make initcall_debug not work if CONFIG_TRACEPOINTS is turned
off? Although it builds but I think this initcall_debug feature will
fail, maybe CONFIG_TRACEPOINTS should be selected somewhere?
I recently ran into some issues like this for my preemptirq
tracepoints patch (which I will post again soon :D) where lockdep
needed the tracepoints and I had to select it.
thanks,
- Joel