dtrace discuss - Nov 2005 - Virtual address of the pagefaults when the application is cold start

Hi,
   How to get the virtual address of the pagefaults when the application is cold
started. I mean that which prob should i use to get the virtual address due to
which the page fault occur?

Thanks,
Dummy
This message posted from opensolaris.org

G''Day,

On Sat, 5 Nov 2005, Dummy wrote:
> Hi,
>    How to get the virtual address of the pagefaults when the
> application is cold started. I mean that which prob should i use to get
> the virtual address due to which the page fault occur?
>From your description I''m guessing you are interested in executable
page
faults that cause a page in?

I''m not sure why the virtual address is most interesting, rather than
say
the pathname of the file it''s paging in from (which the io provider
does
nicely), but not to worry...

I''ve just punched out the following tool,

 # epiva.d
    PID CMD                 VADDR   SIZE PATH
   8766 atq              d27ca000   4096 /lib/libbsm.so.1
   8766 atq              d27c7000   4096 /lib/libbsm.so.1
   8766 atq              d2794000   4096 /lib/libsecdb.so.1
   8766 atq              d25b3000   4096 /lib/libcmd.so.1
   8766 atq              d27c8000   4096 /lib/libbsm.so.1
   8766 atq              d27b5000   4096 /lib/libbsm.so.1
   2065 nscd              8051000   4096 /usr/sbin/nscd
   8770 expr             d2744000   4096 /usr/lib/libc/libc_hwcap1.so.1

And the code behind it is below (which I bet I''ll regret posting since
I
haven''t thought too long and hard about this yet, and so there is
probably a bad assumption somewhere!).

cheers,

Brendan


-----
#!/usr/sbin/dtrace -s

#pragma D option quiet

dtrace:::BEGIN
{
        /* print header */
        printf("%6s %-16s %16s %6s %s\n", "PID",
"CMD", "VADDR",
            "SIZE", "PATH");
}

fbt::segvn_fault:entry
{
        /* save segvn fault details */
        self->vaddr = arg2;
        self->size = arg3;
        self->vp = (struct vnode *)((struct segvn_data
*)args[1]->s_data)->vp;
}

vminfo:::execpgin
{
        /* match on executable page ins */
        self->ok = 1;
}

fbt::segvn_fault:return
/self->ok/
{
        /* print details */
        printf("%6d %-16s %16x %6d %s\n",
            pid, execname, self->vaddr, self->size,
            self->vp == 0 ? "?" :
stringof(self->vp->v_path));
}

fbt::segvn_fault:return
{
        /* clear variables */
        self->vaddr = 0;
        self->size = 0;
        self->vp = 0;
        self->ok = 0;
}

Ok (regret already) I''ve just added a line to this function,

On Sun, 6 Nov 2005, Brendan Gregg wrote:
[...]
> fbt::segvn_fault:entry
> {
>         /* save segvn fault details */
>         self->vaddr = arg2;
>         self->size = arg3;
>         self->vp = (struct vnode *)((struct segvn_data
*)args[1]->s_data)->vp;
	self->ok = 0;
> }
It makes little difference, it was just lazy not to.

Brendan

G''Day folks,

Anyone notice the mailing list misbehaving by adding auto-quoting?

On Sun, 6 Nov 2005, Brendan Gregg wrote:
[...]
> >From your description I''m guessing you are interested in
executable page
> faults that cause a page in?
I don''t normally make those typos. This is a cat -vet from my sent
mail,

$
>From your description I''m guessing you are interested in executable
page$
faults that cause a page in?$
$

And now from the returned mail,

$
>From your description I''m guessing you are interested in executable
page$
faults that cause a page in?$
$

I wonder how the ">" got there. This could be pine misbehaving
(yeah, I
should be using mutt)...

(I think it''s another job for DTrace ;)

Brendan

On 11/5/05, Brendan Gregg <brendan.gregg at tpg.com.au>
wrote:
> G''Day folks,
>
> Anyone notice the mailing list misbehaving by adding auto-quoting?
...
> $
> >From your description I''m guessing you are interested in
executable page$
> faults that cause a page in?$
> $
>
> I wonder how the ">" got there. This could be pine misbehaving
(yeah, I
> should be using mutt)...
Standard Unix mailer thing;  the Unix mbox format uses ''From ''
at the
beginning of the line as a message deliminator, so any instance of it
in an outgoing piece of mail gets ''> '' tacked on to the
front.

Cheers,
- jonathan

Sorry, just realised this isn''t auto-quoting at all - it''s
that blastard
prehistoric sendmail feature to escape "From" lines.

grr, curse you sendmail!

I feel a "DTracing sendmail" website coming along... :)

Brendan


On Sun, 6 Nov 2005, Brendan Gregg wrote:
> G''Day folks,
>
> Anyone notice the mailing list misbehaving by adding auto-quoting?
>
> On Sun, 6 Nov 2005, Brendan Gregg wrote:
> [...]
> > >From your description I''m guessing you are interested in
executable page
> > faults that cause a page in?
>
> I don''t normally make those typos. This is a cat -vet from my sent
mail,
>
> $
> >From your description I''m guessing you are interested in
executable page$
> faults that cause a page in?$
> $
>
> And now from the returned mail,
>
> $
> >From your description I''m guessing you are interested in
executable page$
> faults that cause a page in?$
> $
>
> I wonder how the ">" got there. This could be pine misbehaving
(yeah, I
> should be using mutt)...
>
> (I think it''s another job for DTrace ;)
>
> Brendan
>
> _______________________________________________
> dtrace-discuss mailing list
> dtrace-discuss at opensolaris.org
>
>

G''Day Jonathan,

ahh, you''re right - it is mbox, can''t blame sendmail for this
one!

Brendan


On Sat, 5 Nov 2005, Jonathan Adams wrote:
> On 11/5/05, Brendan Gregg <brendan.gregg at tpg.com.au> wrote:
> > G''Day folks,
> >
> > Anyone notice the mailing list misbehaving by adding auto-quoting?
> ...
> > $
> > >From your description I''m guessing you are interested in
executable page$
> > faults that cause a page in?$
> > $
> >
> > I wonder how the ">" got there. This could be pine
misbehaving (yeah, I
> > should be using mutt)...
>
> Standard Unix mailer thing;  the Unix mbox format uses ''From
'' at the
> beginning of the line as a message deliminator, so any instance of it
> in an outgoing piece of mail gets ''> '' tacked on to
the front.
>
> Cheers,
> - jonathan
>
>

Brendan Gregg wrote:
> $
>>From your description I''m guessing you are interested in
executable page$
> faults that cause a page in?$
> $
> 
> I wonder how the ">" got there. This could be pine misbehaving
(yeah, I
> should be using mutt)...
It''s been a long time since I did MTA configuration, but sendmail used
to
have an option to quote From lines when doing final delivery for compatibility
with BSD mailbox format in which a line starting with "From " was the
message
delimiter.   The intermediate MTA''s like those forwarding the mailing
list
messages shouldn''t be doing that though.

-- 
	-Alan Coopersmith-           alan.coopersmith at sun.com
	 Sun Microsystems, Inc. - X Window System Engineering