wrong calculation of kprobe.addr?

[Hien Nguyen]

Guang Lei Li wrote:

>Hi,
> 
>I tried the following stap script:
>
>probe kernel.function("elv_next_request")
>{
>    log("elv_next_request")
>}
>
>On my Power5(2.6.9-22) it runs as desired, lots of "elv_next_request"
>is printed. But on my pc(2.6.14.2) it will print nothing, the probe 
>seems never to be hit at all. I think it may be architecture related.
>
>But I abandoned this idea after I modified the kernel src file
>elevator.c on my pc to add a printk, to make sure if elv_next_request
>really will never be called on my pc:
>
>struct request *elv_next_request(request_queue_t *q)
>{
>        struct request *rq;
>        int ret;
>        printk(KERN_INFO "elv_next_req from printk");
>        while ((rq = __elv_next_request(q)) != NULL) {
>
>
>I recompiled my kernel(name it 2.6.14.2-custom), reboot, then 
>/var/log/message shows that lots of "elv_next_req from printk". So 
>you can see that elv_next_request is really called on my pc. 
>
>Then I retry my script, this time it worked, lot of "elv_next_request"
>is printed out.
>
>So I am trying to figure out why my script doesn't work for me before.
>The only differences between the working and not-working of my script
>is that I added a printk inside elv_next_request.
>
>The following is the address from /proc/kallsyms and the address
>systemTap used to register a kprobe:
>
>------2.6.14.2 kernel (not modified)------
>from /proc/kallsyms:
>c02402e0 T elv_next_request
>
>snippet from the .c file got from: stap -p3 a.stp
>
>static struct kprobe dwarf_kprobe_0[1]= {
>  {.addr= (void *) 0xc02402e8}
>};
>
>
>------2.6.14.2-custom (modified kernel)------
>
>from /proc/kallsyms:
>c02402e0 T elv_next_request
>
>snippet from the .c file got from: stap -p3 a.stp
>
>static struct kprobe dwarf_kprobe_0[1]= {
>  {.addr= (void *) 0xc02402f0}
>};
>
>
>So you can see, the address elv_next_request doesn't change between
>the two kernels(c02402e0), but the address kprobe used changed(from
>0xc02402e8 to 0xc02402f0). Is this as designed, or there is something
>wrong with the calculation of kprobe address in systemTap?
>
>Thanks in advance for your help
>
>Best Regards,
>
>Li Guanglei
>
>- Linux Performance, China Systems & Technology Lab
>China Development Labs, Beijing Tel: 86-10-82782244 Ext.3516  Email: 
>liguangl@cn.ibm.com
>
>  
>
I tried the above script on my i386 box running kernel v 2.6.14-r3, it 
works fine
It could be a gcc issue.  Post your gcc version,  my gcc version is 4.0.1.
Hien.