Re: Calling systemtap function from pure code

[Arkady <arkady dot miasnikov at gmail dot com>]

I read your e-mail a couple of times and probably I got it. You
want to know inside the systemtap script some details about the kernel
module the script is part of.

The easiest way to do this is to load a non-STAP kernel module which
contains a data structure with all the data. The steps:
* Load the stap driver, make sure that probe begin blocks until there
is an exported variable "Daniel" in the kernel and the value of the
variable is True.
* Collect the data you need from the user space, insmod a small kernel
module with the data, collect more data from the kernel if needed, set
the exported variable to True.
* At this point probe begin discovers an exported variable "Daniel"
and can continue.


If the kernel keeps the relevant data somewhere "probe begin" can try
to fetch the data from the kernel.

On Thu, Sep 14, 2017 at 10:20 AM, Daniel Doron
<danielmeirdoron@gmail.com> wrote:
> I will explain....
> i wanted to get as much info on kernel module insertion/removal but
> the init_module/delete_module probe points do not provide enough
> information. So , I thought about load_module() which does provide
> adequate info via struct module (local var),  but this gets filled
> along the way and I can;t really have access to it via systemtap. So I
> thought the simplest way to access that struct is though notification
> chain, i.e. use register_module_notifier() and extract the info from
> the struct there and then print it via systemtap.
> So I guess the right question would be how to access the STAP print
> function/mechanism/buffer...from standard kernel code.
>
> Thanks.
>
> On Thu, Sep 14, 2017 at 12:22 AM, David Smith <dsmith@redhat.com> wrote:
>> On Wed, Sep 13, 2017 at 9:41 AM, Daniel Doron <danielmeirdoron@gmail.com> wrote:
>>> Hi,
>>>
>>> Is it possible to call a stap function from kernel code (guru mode)?
>>
>> No, not really.
>>
>> Yes, all stap language functions get translated to a C function.
>> However, you can't call that function easily, for several reasons:
>>
>> 1) The stap language function name gets mangled when it gets
>> translated to C. That mangling isn't documented and changes from time
>> to time. The last time it changed was when we implemented function
>> overloading. We have no plans to document the current mangling scheme
>> (you could certainly figure it out looking at the generated output)
>> and we reserve the right to change the mangling scheme in the future.
>>
>> 2) Even if you know the correct C function name for a stap language
>> function, you can't easily call it. Function parameters aren't passed
>> on the stack, they are passed in a context structure, which once again
>> is undocumented and subject to change.
>>
>> I'm not really sure what you are really trying to do here, but if you
>> need to call common code from a stap language function and a C
>> function, I'd put that functionality in a C function, then write a
>> stap language wrapper for it. Something like the following (untested):
>>
>> ====
>> %{
>>
>> int internal_log_event(char *str)
>> {
>>     printk(KERN_INFO "name: %s\n", str);
>>     return 0;
>> }
>>
>> %}
>>
>> function log_event:long(str:string)
>> %{
>>     STAP_RETVALUE = internal_log_event(STAP_ARG_str);
>> %}
>>
>> function call_me()
>> {
>>     log_event("hello world")
>> }
>>
>> probe begin {
>>     call_me()
>> }
>> ====
>>
>> Note that the STAP_ARG_ function argument prefix and STAP_RETVALUE
>> item are documented and are supported.
>>
>> If this doesn't answer your question, you'll need to let us know what
>> you are really trying to do.
>>
>> --
>> David Smith
>> Principal Software Engineer
>> Red Hat