how to identify "Unknown Symbol"

I may come to regret my choice to write a webcam driver ;)

Anyway, now I am having this problem:
insmod: error inserting './usbgetinfo.ko': -1 Unknown symbol in module

I understand that

Quote:

---

Originally Posted by http://www.linuxchix.org/content/courses/kernel_hacking/lesson8

What this is saying is that the kernel is not allowing modules to see that variable. When the module loads, it has to resolve all it's external references, like function names or variable names. If it can't find all of it's unresolved names in the list of symbols that the kernel exports, then the module can't write to that variable or call that function. The variable my_variable has space allocated for it somewhere in the kernel, but the module can't figure out where.

To fix this, we're going to add my_variable to the list of symbols that the kernel exports...

---

However, what if I don't know which variable is causing the problem (the code is simple and straightforward, <70 lines, etc)...is there a way to identify the offending symbol?? From the sound of it, this could almost be just a missing header...

Try 'objdump -T mymodule.ko' to see all the symbols it's trying to bring in... Process of elimination?

Quote:

---

Originally Posted by brewbuck

Try 'objdump -T mymodule.ko' to see all the symbols it's trying to bring in... Process of elimination?

---

Not familiar with objdump but I would guess you meant -t, not -T. Is it safe to assume that I should confine my search to *UND* symbols?

Also, this does not list all the variables referenced, which in that quote from the OP, the error is produced by inserting a variable into the printk source, recompiling the kernel, and trying to reference it in the module code. In which case I imagine "printk" would be in the symbol list, but nothing about the variable. So this list doesn't seem any better than just looking at the code itself...

Quote:

---

Originally Posted by MK27

Not familiar with objdump but I would guess you meant -t, not -T. Is it safe to assume that I should confine my search to *UND* symbols?

---

It was a typo, I actually meant "-tT" to get both (is a ".ko" like a shared library or like an object file? I can't remember -- "-T" doesn't apply to object files).

Yes, you're concerned with the symbols marked *UND*. The symbol sections might be informative too. I have not looked at a kernel module in a while.

Quote:

---

Originally Posted by brewbuck

Yes, you're concerned with the symbols marked *UND*. The symbol sections might be informative too. I have not looked at a kernel module in a while.

---

No it's "not a dynamic object".

Not sure if this will help other than to round out my knowledge of what a symbol is. Which is helpful.

Problem is, this is already barebones. There's not much extraneous stuff I can take out one piece at a time and still expect it to compile or load. Sigh.

So what do you mean by "symbol sections" (--full-contents?)?

You can find currently defined symbols in the kernel by reading /proc/kallsyms. You might be able to match up the two (the contents of the objdump output) with a little perl or the shell.

Quote:

---

Originally Posted by tjb

You can find currently defined symbols in the kernel by reading /proc/kallsyms. You might be able to match up the two (the contents of the objdump output) with a little perl or the shell.

---

Tee hee. The problem turned out to be that I forgot to include

Code:

---

MODULE_LICENSE("GPL");

---

which meant the kernel was not willing to export anything at all...

But thanks for this tip, it may eventually come in handy.