Linux Kernel Caller ID

Kernel Caller ID

Linux kernel boasts of a useful debug mechanism for printing the caller of a function.

printk(“Caller is %pS\n”, __builtin_return_address(0));

Without a JTAG debugger this is my primary tool to figure out the call stack. (Any better ideas?). Took couple of hours to dissect this mechanism but it was worth it.

There are two main parts to this:
1. Get the caller address.
&
2. Map the caller address to the caller name.

Step 1 : __builtin_return_address(0)
ARM assembly clearly shows that this is an assembler directive which simply fetches the caller address from the stack (or the return address register perhaps?)

Step 2: How do we map this address into a symbol string?
In simple words, when kernel is linked it created a compressed version of symbol table and parsing this provides the mapping from address to string. A one to one mapping of address and the ASCII string would lead to a gargantuan sized binary so we need some form of compression.

So the algorithm tends to encodes the symbol by exploiting the repeating character patterns, for example a string “__exit” might be represented by an arbitrary code 0x34, so the idea is to identify the these patterns and generates a custom representation of a string. Elegant and effective!! This works just fine on loadable kernel modules also, because the dynamic loading process takes care of this. More details might need a look into insmod!

Please look into the C file linux/kernel/kallsyms.c for discovering more.

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s