uaddr and friends

[Mark Johnston]

On Tue, Apr 15, 2014 at 10:54 PM, carlos antonio neira bustos
<cneirabustos at gmail.com> wrote:
> Thanks Mark.
>
> You are right, I did not look at dt_subr.c, uaddr is a short version of
> ustack, I'll update to current now and try to pick on something else from
> the list.

Thanks! I've updated the TODOs a bit. I find that a good way to find
things to do is to just try random DTrace scripts until something
fails reproducibly. Large processes like firefox and Xorg have been a
fertile source of bugs. :)

And yes, please update to CURRENT; there have been quite a few bugs
fixed since 10.0. FreeBSD 10.1 will have much better userland DTrace
support.

-Mark

>
> Bests
>
>
> On Tue, Apr 15, 2014 at 11:36 PM, Mark Johnston <markj at freebsd.org> wrote:
>>
>> On Tue, Apr 15, 2014 at 10:15 PM, carlos antonio neira bustos
>> <cneirabustos at gmail.com> wrote:
>> > Hi Mark,
>> >
>> > I was looking at this document
>> >
>> > https://wikis.oracle.com/display/DTrace/Actions+and+Subroutines#ActionsandSubroutines-%7B%7Busym%7D%7D
>> > currently looking at the code uaddr is the same as usym, but according
>> > to
>> > this document
>> >
>> > usym will print the symbol for a specified address. This is analogous to
>> > how
>> > uaddr works, but without the hexadecimal offsets.
>> >
>> >         uaddr:          date`clock_val+0x1
>> >         usym:           date`clock_val
>> >
>> >
>> > here are my outputs from usym and uaddr :
>> >
>> > root at bsd:/home/cneira # dtrace -n 'pid$target::main:{usym(uregs[R_PC])}'
>> > -c
>> > ./test
>> > dtrace: description 'pid$target::main:' matched 2 probes
>> > abCPU     ID                    FUNCTION:NAME
>> >   0  56282                       main:entry   test`main
>> >   0  56281                      main:return   test`main
>> > dtrace: pid 10588 has exited
>> >
>> > root at bsd:/home/cneira # dtrace -n
>> > 'pid$target::main:{uaddr(uregs[R_PC])}' -c
>> > ./test
>> > dtrace: description 'pid$target::main:' matched 2 probes
>> > abCPU     ID                    FUNCTION:NAME
>> >   0  56282                       main:entry
>> >               test`main+0x1
>> >
>> >   0  56281                      main:return
>> >               test`main+0x23
>> >
>> > dtrace: pid 10591 has exited
>> >
>> > As current is only using dt_print_usym for uaddr the output should not
>> > have
>> > the  hexadecimal offset specified in that documentation.
>> > So I'm somewhat lost about which is the correct behavior , Do you have a
>> > uaddr and usym  output  example in current to check this?, all my
>> > changes
>> > were done in 10 prod release.
>>
>> You're right, dt_print_usym handles both usym and uaddr. In the usym
>> case, there's some special handling which looks up the symbol name
>> corresponding to the given pc, and then sets the pc variable to the
>> address of the symbol. Then dtrace_uaddr2str only prints the offset
>> from the symbol if that offset is non-zero, so in the uaddr case you
>> get "+0x1" and "+0x23" above.
>>
>> >
>> > Bests
>> >
>> >
>> >
>> >
>> >
>> >
>> >
>> > On Tue, Apr 15, 2014 at 10:09 PM, Mark Johnston <markj at freebsd.org>
>> > wrote:
>> >>
>> >> On Tue, Apr 15, 2014 at 8:35 PM, carlos antonio neira bustos
>> >> <cneirabustos at gmail.com> wrote:
>> >> > Hi All,
>> >> >
>> >> > I was looking at Dtrace todo list from
>> >> > https://wiki.freebsd.org/DTraceTODO
>> >> > and started to work on : Get uaddr(), ufunc(), umod(), and usym()
>> >> > action
>> >> > funtions working.
>> >> > I have  the  data recording action (uaddr)  working  now, here is an
>> >> > example:
>> >>
>> >> Hi!
>> >>
>> >> Hm, these should have been (at least mostly) working since r258902:
>> >> http://svnweb.freebsd.org/base?view=revision&revision=258902
>> >>
>> >> >
>> >> > root at bsd:/home/cneira # dtrace -n
>> >> > 'pid$target::main:{uaddr(uregs[R_PC])}'
>> >> > -c ./test
>> >> > dtrace: description 'pid$target::main:' matched 2 probes
>> >> > abCPU     ID                    FUNCTION:NAME
>> >> >   0  56282                       main:entry
>> >> >               test`main+0x1
>> >> >
>> >> >   0  56281                      main:return
>> >> >               test`main+0x23
>> >> >
>> >> > dtrace: pid 9687 has exited
>> >> >
>> >> > Here is the output of nm from the test binary I have used to check
>> >> > this
>> >> > change.
>> >> >
>> >> >         U _init_tls@@FBSD_1.0
>> >> > 08048380 T _start
>> >> > 080483a0 t _start1
>> >> > 0804814c r abitag
>> >> >          U atexit@@FBSD_1.0
>> >> > 08048164 r crt_noinit_tag
>> >> > 080485c0 T dosomething
>> >> > 08049764 B environ
>> >> >          U exit@@FBSD_1.0
>> >> > 080484f0 t finalizer
>> >> > 08048560 t frame_dummy
>> >> > 08048590 T main
>> >> >          U putchar@@FBSD_1.0
>> >> >
>> >> >
>> >> > root at bsd:/home/cneira # dtrace -c ./test -n 'pid$target::main:entry{
>> >> > uaddr(0x080485c0); }'
>> >> > dtrace: description 'pid$target::main:entry' matched 1 probe
>> >> > abCPU     ID                    FUNCTION:NAME
>> >> >   0  56281                       main:entry
>> >> >               test`dosomething
>> >> >
>> >> > dtrace: pid 9736 has exited
>> >> >
>> >> > I'll continue working on the rest, how do I submit  a patch with
>> >> > these
>> >> > changes ?
>> >>
>> >> Pasting it inline is generally fine if it's not too large. It's also
>> >> ok to put it in a public directory somewhere and post a link to it.
>> >>
>> >> -Mark
>> >
>> >
>
>