Hi Jonathan,
I am running snapshot 96. The boot drive is using the mpt driver.
And the same thing happens -- no output, although the driver has been
modloaded via running the dtrace command (and I did wait a bit before
typing <ctrl><c>).
dell690pr# modinfo | grep mpt
46 fffffffff7883000 3c768 169 1 mpt (MPT HBA Driver v1.144)
dell690pr# modinfo | grep mega
dell690pr# dtrace -n fbt:mega_sas:megasas_attach:entry'{
@[probefunc]=count(); }'
dtrace: description 'fbt:mega_sas:megasas_attach:entry' matched 1 probe
^C
dell690pr# modinfo | grep mega
241 fffffffff8379000 7c10 211 1 mega_sas (LSIv1.26)
dell690pr#
Thanks,
-Susan
Jonathan Adams wrote:
> On Tue, Sep 02, 2008 at 03:47:25PM -0700, Susan Scheufele wrote:
>> Hello dtrace gurus,
>>
>> Is it possible to put DTRACE_PROBEs in a driver attach routine, and
>> expect to see them show up when running a dtrace script to trigger them
>> (which incidentally causes the driver to be modload-ed)?
>>
>> I'm trying to change all of the funky con_log() calls in the mega_sas
>> driver to use DTRACE_PROBEs, but I'm wondering if this will really work
>> for attach. I do see the probes from other non-attach mega_sas function
>> calls after the driver attaches, but not the probes from during
>> megasas_attach. Maybe I need to retain the logging code for the
>> attach/detach routines?
>
> What version of Solaris is this?
>
> I believe they should show up; dtrace gets notification very early on (before
> _init is run)[1], which should cause the sdt probes to show up.
>
> What happens if you enable fbt::megasas_attach:entry?
>
> Cheers,
> - jonathan
>
> [1] Unfortunately, an old linker bug makes instrumenting _init and _fini
> impossible at the moment.
_______________________________________________
dtrace-discuss mailing list
[email protected]
