Forw: RE: systemtap with USB

Forw: RE: systemtap with USB

[Frank Ch. Eigler]

From: <jaseen.shamsuddin@wipro.com>

Hi

Finally the USB script worked. 
created one soft link like
/boot/vmlinux-2.6.24 -> /root/linux/linux-2.6.24/vmlinux

for this to work.

Now the usb_register_driver probe point is getting hit. For this to
happen we need to insert the flash drive after the staprun is executed
during Pass 5 of stap execution.

Thanks a lot for your valuable inputs.

I would like to experiment more on this to see what all can be tapped
from the USB subsystem. Please suggest how to proceed.

Regards,
Jaseen 

Re: Forw: RE: systemtap with USB

[tsjaseen]

Hi,

I would like to experiment more on this to see what all can be tapped from
the USB subsystem. Please suggest how to proceed so that we can make some
contribution on this side.

Regards,
Jaseen


Frank Ch. Eigler wrote:
> 
> From: <jaseen.shamsuddin@wipro.com>
> 
> Hi
> 
> Finally the USB script worked. 
> created one soft link like
> /boot/vmlinux-2.6.24 -> /root/linux/linux-2.6.24/vmlinux
> 
> for this to work.
> 
> Now the usb_register_driver probe point is getting hit. For this to
> happen we need to insert the flash drive after the staprun is executed
> during Pass 5 of stap execution.
> 
> Thanks a lot for your valuable inputs.
> 
> I would like to experiment more on this to see what all can be tapped
> from the USB subsystem. Please suggest how to proceed.
> 
> Regards,
> Jaseen 
> 
> 

-- 
View this message in context: http://www.nabble.com/Forw%3A-RE%3A-systemtap-with-USB-tp16497588p17638677.html
Sent from the Sourceware - systemtap mailing list archive at Nabble.com.

Re: Forw: RE: systemtap with USB

[Felipe Balbi]

Hello,

On Tue, 3 Jun 2008 20:58:01 -0700 (PDT), tsjaseen
<jaseen.shamsuddin@wipro.com> wrote:
> I would like to experiment more on this to see what all can be tapped
from
> the USB subsystem. Please suggest how to proceed so that we can make some
> contribution on this side.

The simplest thing to do would be to track down all the function calls
starting form usb_register_driver. If you look at the implementation
of usb_register_driver [1] you can start thinking on how to do it.

Just follow the function call and print each one using thread_indent()
so it's easier to read. After that you could start trying to identify
hot paths in the middle and check if they could be optimized or not.

Also, checking usb throughput each 5 secs, or something, and printing
the exact function been called now would, maybe, help identifying wether
we should or shouldn't improve the code.

USB 2.0 devices attached to USB 2.0 hosts (ehci) could transfer
at a rate of 480Mbps, something around 300Mbps still quite acceptable,
but 60Mbps we could improve :-)

Tomorrow I'll start digging on it as well, maybe we might join work in
the near future ;-)

Good luck! :-D

[1] http://lxr.linux.no/linux/drivers/usb/core/driver.c#L736

-- 
Best Regards,

Felipe Balbi
http://felipebalbi.com
me@felipebalbi.com