Weird Probe points on PPC

Weird Probe points on PPC

[Breno Leitao]

Hi Guys, 

I found something that I can't explain in SystemTap.
When probing module("jsm").function("*) it is probing more
function that those inside the jsm module (ie, out_8, in_8, writeb, etc)

At the same time, if I probe module("jsm").function("*").return
it works properly. Did anyone see anything like that ?


kernel: 2.6.31-rc8-ppc64
stap version: version 0.9.9/0.142.90 commit release-0.9.9-319-ge08217d + changes


----- jsm.stp -----
probe module("jsm").function("*"){
	printf("-> %s\n",  probefunc())
}


probe module("jsm").function("*").return {
	printf("<- %s\n",  probefunc())
}


---- output -----

-> jsm_tty_open
-> neo_flush_uart_write
-> writeb
-> out_8
-> eeh_readb
-> readb
-> in_8
<- neo_flush_uart_write
-> neo_flush_uart_read
-> writeb
-> out_8
-> eeh_readb
-> readb
-> in_8
<- neo_flush_uart_read
-> neo_uart_init
-> writeb
-> out_8
-> writeb
-> out_8
-> writeb
-> out_8
-> eeh_readb
-> readb
-> in_8
-> writeb
-> out_8
-> eeh_readb
-> readb
-> in_8
-> eeh_readb
-> readb
-> in_8
-> writeb
-> out_8
<- neo_uart_init
-> neo_param
-> eeh_readb
-> readb
-> in_8
-> eeh_readb
-> readb
-> in_8
-> writeb
-> out_8
-> writeb
-> out_8
-> writeb
-> out_8
-> writeb
-> out_8
-> writeb
-> out_8
-> neo_set_new_start_stop_chars
-> writeb
-> out_8
-> writeb
-> out_8
-> writeb
-> out_8
-> writeb
-> out_8

Thanks, 
Breno

Re: Weird Probe points on PPC

[Breno Leitao]

Breno Leitao wrote:
> Hi Guys, 
> 
> I found something that I can't explain in SystemTap.
> When probing module("jsm").function("*) it is probing more
> function that those inside the jsm module (ie, out_8, in_8, writeb, etc)
I just found that if I add .call, then it works properly. 
So, what is the difference between module("").function("*") and module("").function("*").call ?

Thanks

Re: Weird Probe points on PPC

[Eugeniy Meshcheryakov]

[-- Attachment #1: Type: text/plain, Size: 521 bytes --]

29 вересня 2009 о 11:17 -0300 Breno Leitao написав(-ла):
> Breno Leitao wrote:
> > Hi Guys, 
> > 
> > I found something that I can't explain in SystemTap.
> > When probing module("jsm").function("*) it is probing more
> > function that those inside the jsm module (ie, out_8, in_8, writeb, etc)
> I just found that if I add .call, then it works properly. 
> So, what is the difference between module("").function("*") and module("").function("*").call ?
First one probes also inline functions?

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

Re: Weird Probe points on PPC

[Breno Leitao]

Eugeniy Meshcheryakov wrote:
>> I just found that if I add .call, then it works properly. 
>> So, what is the difference between module("").function("*") and module("").function("*").call ?
> First one probes also inline functions?
Well, in_8() for example is an inline fuction, but it's not related to jsm driver at all. 
It's part of the arch/powerpc/boot/io.h, and it is defined as follow: 

static inline int in_8(const volatile unsigned char *addr)

So, If I understood correctly, those inline function becomes part of the module, and then
the probe considers them ?

Re: Weird Probe points on PPC

[Eugeniy Meshcheryakov]

[-- Attachment #1: Type: text/plain, Size: 907 bytes --]

29 вересня 2009 о 11:28 -0300 Breno Leitao написав(-ла):
> Eugeniy Meshcheryakov wrote:
> >> I just found that if I add .call, then it works properly. 
> >> So, what is the difference between module("").function("*") and module("").function("*").call ?
> > First one probes also inline functions?
> Well, in_8() for example is an inline fuction, but it's not related to jsm driver at all. 
> It's part of the arch/powerpc/boot/io.h, and it is defined as follow: 
> 
> static inline int in_8(const volatile unsigned char *addr)
> 
> So, If I understood correctly, those inline function becomes part of the module, and then
> the probe considers them ?
> 
Inline functions are inlined in places were they are called, so any
module that uses them contains one or more copies of such functions. It does
not matter where they are defined (but you may try function@file syntax). 

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 198 bytes --]