walking

Thank you very much! I am sending packets based on the modified module. This module
is like UDP/IP. Sometimes, I can send packets smoothly, sometimes the sending makes the
computer crashed (only the source node). Yesterday I found sometimes it made the system
"segmentation fault" instead of hanging up, that is much better for me. Here is the error:


However, I don't think I access an NULL pointer in my module. Is there any kernel function
I used wrong? Can I get some information from the above errors? I read the file under the directory /var/log/ksysmoops trying to find the above xxxxxxx code's meaning, but I didn't get them.

What's more, when I tried to remove the module, it said " Device or Resource Busy". Why this happened? Thanks a lot!

Thanks again!

Mod edit - added code tags.

Dolda2000

How can you be so sure? Is it really impossible that you may have stored 0 or NULL in a pointer anywhere, or alternatively not initialized some structure insufficiently, and then derefenced it?

The first thing you should do is recompile the kernel with debugging support. That way, the panic handler will be able to print symbolic function names instead of addresses in the backtrace. That'll make it very much easier to debug.
That could just be because the addresses were inside your module, in which case they won't be found by ksymoops. Anyhow, like I said, recompiling the kernel with debug support will solve that part of the problem.

Probably because some refcount mismatch. Are you releasing all resources properly? Either way, if you're segfaulting inside the kernel, there's really no telling what may happen.

walking

Thank you so much for your reply! I really appreciate it.

I found the error which makes the kenel panic. As I tried to print "skb->mac.ethernet->h_proto " , probably sometimes it is a NULL pointer. I am not very
sure. I haven't read how the mac header works. However, after I removed the printk
statement. No segmentation fault any more.


However, I returned to the original question. Sometimes, I sent traffic based on the module,
everything works. However, sometimes, it made the whole machine HANG up. No any
kernel panic information in log file or console. Because I tried to debug the code to see
where is problem, I set some printk statements in every function. However, when I enabled
the printk statements, the machine doesn't hangs up ( at least, very lower possibility than disabling the printk statements.). Why could it be that?

I don't want to enable printk statments, because it really affected my measurements. such as
throughput, etc.

Any advice? Thanks a lot!