merge gen_kobject_list.py and gen_priv_stacks.py

[andrewboie]

Generation of privilege mode stack buffers can take place in the main gen_kobject_list.py script, there's no actual need for them to be separate. Any data/bss for objects declared in the gperf C file already goes in an area of memory which can expand without limit.

We just need the main gperf table; for any thread's kobject, the data field will point to a struct containing the thread ID and a pointer to its stack buffer.

Should help with build times as we only need to scan the DWARF data once instead of twice.

[andrewboie]

Further detail:

/* Table generated by gperf, these objects are retrieved via
 * z_object_find() */
struct _k_object {
        char *name;
        u8_t perms[CONFIG_MAX_THREAD_BYTES];
        u8_t type;
        u8_t flags;
        u32_t data;
} __packed __aligned(4);

Right now each script generates a gperf hash table and instantiates some objects

gen_kobject_list.py:

• Scan the elf binary for instances of any type of kernel object
• Generate a table of struct _k_object for every kernel object found in the build.
• The hash table map kernel object memory addresses to kernel object metadata (struct _k_object), where the _k_object gets instantiated by the script.
  • • For any instances of k_futex, instantiate a wait_q and assign it to the futex's associated _k_object's data member
  • • For any instance of sys_mutex, instantiate a k_mutex and assign it to the sys_mutex's associated k_object's data member
  • • For threads, assign a numeric ID and place in the data member
  • • For thread stacks, set the data member to be the size of the stack object

gen_priv_stacks.py:

• Scan the elf binary for instances of k_thread_stack element arrays (which is what thread stacks are)
• Generate a table mapping stack object memory addresses to pointers to privilege mode stacks
• Generate a privilege mode stack for every stack object found, which can be looked up.

These do not have to be separate. These scripts do the same thing: perform a time-consuming analysis of the DWARF debug info to find kernel objects
We can generate the privilege mode stacks in gen_kobject_list directly just like we already do with futex waitqs, etc.
The data member for a thread stack object currently just has the stack size. Instead, create an instance of:

struct z_stack_metadata {
   char *privilege_stack;
   size_t size;
}

Populate these appropriately for every stack object found, and set the data member of the stack's struct k_object to point to it.

[andrewboie]

Note that elf_helper.py script was split out (in a rather rough fashion) to support this extra script, so really I'd like to see gen_kobject_list.py, gen_priv_stacks.py, and elf_helper.py all combined into a single script, that creates a single gperf table.