Re: [PATCH v10 2/4] trace/objtrace: Get the value of the object

[Jeff Xie <xiehuan09@gmail.com>]

Hi Masami,

On Wed, Jun 1, 2022 at 11:13 PM Masami Hiramatsu <mhiramat@kernel.org> wrote:
>
> On Tue, 31 May 2022 15:13:24 +0800
> Jeff Xie <xiehuan09@gmail.com> wrote:
>
> > Hi Masami and steve,
> >
> > On Sun, May 22, 2022 at 10:22 PM Masami Hiramatsu <mhiramat@kernel.org> wrote:
> > >
> > > Hi Jeff,
> > >
> > > On Fri, 13 May 2022 01:00:06 +0800
> > > Jeff Xie <xiehuan09@gmail.com> wrote:
> > >
> > > [...]
> > > > @@ -175,9 +271,27 @@ trace_object_trigger(struct event_trigger_data *data,
> > > >
> > > >       field = obj_data->field;
> > > >       memcpy(&obj, rec + field->offset, sizeof(obj));
> > > > -     set_trace_object(obj, tr);
> > > > +     /* set the offset from the special object and the type size of the value*/
> > > > +     set_trace_object(obj, obj_data->obj_offset,
> > > > +                     obj_data->obj_value_type_size, tr);
> > > >  }
> > > >
> > > > +static const struct objtrace_fetch_type objtrace_fetch_types[] = {
> > > > +     {"u8", 1},
> > > > +     {"s8", 1},
> > > > +     {"x8", 1},
> > > > +     {"u16", 2},
> > > > +     {"s16", 2},
> > > > +     {"x16", 2},
> > > > +     {"u32", 4},
> > > > +     {"s32", 4},
> > > > +     {"x32", 4},
> > > > +     {"u64", 8},
> > > > +     {"s64", 8},
> > > > +     {"x64", 8},
> > >
> > > Hmm, as far as I can see, you don't distinguish the prefix 'u','s','x'.
> > > If so, please support only 'x' at this moment. kprobe events supports
> > > those types, and it distinguishes the types when printing the logged
> > > data. E.g. 's16' shows '-1' for 0xffff, but 'x16' shows '0xffff'.
> > > You can add another patch to support such different types afterwards.
> >
> > I feel to let the objtrace trigger to distinguish the prefix 'u', 's',
> > 'x', It seems a very challenging work ;-)
> > I spent a lot of time thinking, I would like to add a callback
> > function(print function) in the struct trace_object_entry  for each
> > data type.
> > Not sure if this is possible or allowed, as I haven't seen any example
> > like this to add function in the  struct *_entry  ;-)
>
> Hmm, I don't recommend this, becuase this event record can be exposed
> to user space as binary data. So please do not put such a function
> pointer which will be used in the ftrace directly.
> Instead, add a new event type of the object-trace for each type-prefix,
> since each of them has different print-fmt.

This place is still a difficult point for me ;-)

> Anyway I would like to ask you is to share the next version of the
> series without that improvement. You can improve it after merging the
> basic feature. No need to stop the series until all possible feature
> set are implemented (unless it will change the user-exposed interface
> much.)

Thanks for your suggestion, I will submit the next version.

>
> >
> > The following is part of the code I have prepared. I don't know if you
> > can give any suggestions or wait until I submit the next version to
> > discuss.
>
> But thanks for sharing the code. This helps me to understand what you
> are trying :)
>
> Thank you,
>
>
>
> >
> > <snip>
> > diff --git a/kernel/trace/trace_entries.h b/kernel/trace/trace_entries.h
> > index 2407c45a568c..5f8289e26f91 100644
> > --- a/kernel/trace/trace_entries.h
> > +++ b/kernel/trace/trace_entries.h
> > @@ -414,6 +414,7 @@ FTRACE_ENTRY(object, trace_object_entry,
> >                 __field(        unsigned long,          parent_ip       )
> >                 __field(        unsigned long,          object          )
> >                 __field(        unsigned long,          value           )
> > +               __field(        unsigned long,          print           )
> >         ),
> >
> > +/* get the type size for the special object */
> > +struct objtrace_fetch_type {
> > +       char *name;
> > +       int type_size;
> > +       int is_signed;
> > +       print_type_func_t       print;
> > +};
> > +
> >
> >  static const struct objtrace_fetch_type objtrace_fetch_types[] = {
> > -       {"x8", 1},
> > -       {"x16", 2},
> > -       {"x32", 4},
> > -       {"x64", 8},
> > -       {NULL, 0}
> > +       {"u8", 1, 0, PRINT_TYPE_FUNC_NAME(u8)},
> > +       {"s8", 1, 1, PRINT_TYPE_FUNC_NAME(s8)},
> > +       {"x8", 1, 0, PRINT_TYPE_FUNC_NAME(x8)},
> > +       {"u16", 2, 0, PRINT_TYPE_FUNC_NAME(u16)},
> > +       {"s16", 2, 1, PRINT_TYPE_FUNC_NAME(s16)},
> > +       {"x16", 2, 0, PRINT_TYPE_FUNC_NAME(x16)},
> > +       {"u32", 4, 0, PRINT_TYPE_FUNC_NAME(u32)},
> > +       {"s32", 4, 1, PRINT_TYPE_FUNC_NAME(s32)},
> > +       {"x32", 4, 0, PRINT_TYPE_FUNC_NAME(x32)},
> > +       {"u64", 8, 0, PRINT_TYPE_FUNC_NAME(u64)},
> > +       {"s64", 8, 1, PRINT_TYPE_FUNC_NAME(s64)},
> > +       {"x64", 8, 1, PRINT_TYPE_FUNC_NAME(x64)},
> > +       {NULL, 0, 0, NULL}
> >  };
> > </snip>
> >
> > > > +     {}
> > >
> > > If this array is null terminated, please explictly do that, like
> > >
> > >         {NULL, 0},
> > >
> > > for readability.
> > >
> > > Thank you,
> > >
> > > --
> > > Masami Hiramatsu (Google) <mhiramat@kernel.org>
> >
> > Thanks,
> > JeffXie
>
>
> --
> Masami Hiramatsu (Google) <mhiramat@kernel.org>

Thanks,
JeffXie