dtrace discuss - Mar 2006 - Missing probes in libc.so.1

Hi,

I did "dtrace -l" on libc.so.1 with "_c??????" 
matching pattern and yield only two probes but
there are 6 of them matches that pattern when
invoke "nm" on libc.  Why?

See below for details:

%dtrace -w -qn ''BEGIN {system("dtrace -l -mlibc.so.1 >
libc_dtrace");} pid$target:libc:_c??????:entry{}'' -c
''sleep 1''

% cat libc_dtrace
   ID   PROVIDER            MODULE                          FUNCTION NAME
42288   pid22010         libc.so.1                          _cleanup entry
42289   pid22010         libc.so.1                          _creat64 entry
% /usr/ccs/bin/nm -n /lib/libc.so.1 | grep ''\|_c......$'' |
grep FUNC
[8747]  |    260828|     500|FUNC |GLOB |0    |9      |_catgets
[8845]  |    261328|     148|FUNC |GLOB |0    |9      |_catopen
[6100]  |    525792|      16|FUNC |WEAK |0    |9      |_cleanup
[7789]  |    651320|      24|FUNC |GLOB |0    |9      |_creat64
[8799]  |    515852|      56|FUNC |GLOB |0    |9      |_ctermid
[8814]  |    271940|      72|FUNC |GLOB |0    |9      |_ctime_r
[8071]  |    515964|     144|FUNC |GLOB |0    |9      |_cuserid
% 

/CP Lin

Upon further investigation, I found that a lot of
libc functions with leading ''_'' actually have a twin
without leading ''_".  For example, _ctermid and
ctermid are both the same function with different
names.  An elfdump of libc reveal this:

# /usr/ccs/bin/elfdump /lib/libc.so.1 | grep ''ctermid$'' | grep
FUNC
     [2321]  0x0007df0c 0x00000038  FUNC WEAK  D   35 .text       ctermid
     [2802]  0x0007df0c 0x00000038  FUNC GLOB  D   35 .text       _ctermid
     [8318]  0x0007df0c 0x00000038  FUNC WEAK  D    0 .text       ctermid
     [8799]  0x0007df0c 0x00000038  FUNC GLOB  D    0 .text       _ctermid
#

So, how do I know Dtrace is going to pick ctermid
not _ctermid?

THX.

Chian-Phon Lin wrote On 03/30/06 20:38,:
> Hi,
> 
> I did "dtrace -l" on libc.so.1 with "_c??????" 
> matching pattern and yield only two probes but
> there are 6 of them matches that pattern when
> invoke "nm" on libc.  Why?
> 
> See below for details:
> 
> %dtrace -w -qn ''BEGIN {system("dtrace -l -mlibc.so.1 >
libc_dtrace");} pid$target:libc:_c??????:entry{}'' -c
''sleep 1''
> 
> % cat libc_dtrace
>    ID   PROVIDER            MODULE                          FUNCTION NAME
> 42288   pid22010         libc.so.1                          _cleanup entry
> 42289   pid22010         libc.so.1                          _creat64 entry
> % /usr/ccs/bin/nm -n /lib/libc.so.1 | grep ''\|_c......$''
| grep FUNC
> [8747]  |    260828|     500|FUNC |GLOB |0    |9      |_catgets
> [8845]  |    261328|     148|FUNC |GLOB |0    |9      |_catopen
> [6100]  |    525792|      16|FUNC |WEAK |0    |9      |_cleanup
> [7789]  |    651320|      24|FUNC |GLOB |0    |9      |_creat64
> [8799]  |    515852|      56|FUNC |GLOB |0    |9      |_ctermid
> [8814]  |    271940|      72|FUNC |GLOB |0    |9      |_ctime_r
> [8071]  |    515964|     144|FUNC |GLOB |0    |9      |_cuserid
> % 
> 
> /CP Lin
> 
> _______________________________________________
> dtrace-discuss mailing list
> dtrace-discuss at opensolaris.org
-- 
************************************************
* C P Lin, Common Technology Project Lead.     *
* Sun Microsystems Inc.                        *
* E-Mail:  c.lin at sun.com                       *
* Address: 4150 Network Circle, M/S UMPK12-330 *
* Santa Clara, CA 95054                        *
* Phone:   650/352-4967	   Fax: 650/786-7816   *
************************************************

Hi,

On 3/31/06, CHIAN-PHON LIN <C.Lin at sun.com>
wrote:
> Upon further investigation, I found that a lot of
> libc functions with leading ''_'' actually have a twin
> without leading ''_".  For example, _ctermid and
> ctermid are both the same function with different
> names.  An elfdump of libc reveal this:
>
> # /usr/ccs/bin/elfdump /lib/libc.so.1 | grep ''ctermid$'' |
grep FUNC
>      [2321]  0x0007df0c 0x00000038  FUNC WEAK  D   35 .text       ctermid
>      [2802]  0x0007df0c 0x00000038  FUNC GLOB  D   35 .text       _ctermid
>      [8318]  0x0007df0c 0x00000038  FUNC WEAK  D    0 .text       ctermid
>      [8799]  0x0007df0c 0x00000038  FUNC GLOB  D    0 .text       _ctermid
> #
>
> So, how do I know Dtrace is going to pick ctermid
> not _ctermid?
When the functions _X and X are equivalent, observed from their
starting address, only 1 probe is available.  In this case, X is
probably chosen since it is the outward facing function.


--
Just me,
Wire ...