Re: Revisiting DTrace syscall provider

[Ryota Ozaki <ozaki-r%netbsd.org@localhost>]

On Wed, Feb 25, 2015 at 1:42 PM, Ryota Ozaki <ozaki-r%netbsd.org@localhost> wrote:
> On Wed, Feb 25, 2015 at 12:35 PM, Ryota Ozaki <ozaki-r%netbsd.org@localhost> wrote:
>> On Wed, Feb 25, 2015 at 4:10 AM, bch <brad.harder%gmail.com@localhost> wrote:
>>> w/ latest -current:
>>>
>>> % dtrace -l
>>> BEGIN
>>> END
>>> ERROR
>>> % dtrace -n BEGIN
>>> <fault/reboot>
>>>
>>>
>>> This is w/o loading any kernel modules, but the probes are still
>>> listed as available, but then fault.
>>
>> Autoloading modules is a feature of NetBSD. You can see solaris.kmod
>> and dtrace.kmod are loaded after dtrace -l. Note that BEGIN, END,
>> ERROR are 'dtrace' provider that is included in dtrace.kmod.
>>
>> OTOH, the fault should be a bug. I'm investigating it.
>
> Got it. NetBSD has also auto-unloading feature and it unloads
> dtarce.kmod after kern.module.autotime sec since auto-loading.
> Currently dtrace.kmod allows being unloaded without caring of users.
>
> A quick fix is:
>
> diff --git a/external/cddl/osnet/dev/dtrace/dtrace_modevent.c
> b/external/cddl/osnet/dev/dtrace/dtrace_modevent.c
> index 9df0cd1..40c8e7f 100644
> --- a/external/cddl/osnet/dev/dtrace/dtrace_modevent.c
> +++ b/external/cddl/osnet/dev/dtrace/dtrace_modevent.c
> @@ -38,6 +38,8 @@ dtrace_modcmd(modcmd_t cmd, void *data)
>         case MODULE_CMD_FINI:
>                 dtrace_unload();
>                 return devsw_detach(NULL, &dtrace_cdevsw);
> +       case MODULE_CMD_AUTOUNLOAD:
> +               return EBUSY;
>         default:
>                 return ENOTTY;
>         }
>
> It prevents auto-unloading. It should fix the problem.

...note that unloading dtrace.kmod by modunload with dtrace command
running also causes panic. We need to fix dtrace unload procedures
somehow :-/

  ozaki-r