Re: mod->klp set on copy ok ?

[Josh Poimboeuf <jpoimboe@kernel.org>]

On Thu, Mar 16, 2023 at 02:50:55PM -0700, Luis Chamberlain wrote:
> The comment for "Update sh_addr to point to copy in image." seems pretty
> misleading to me, what we are doing there is actually ensuring that we update
> the copy's ELF section address to point to our newly allocated memory.
> Do folks agree?
> 
> And how about the size on the memcpy()? That's a shd->sh_size. No matter
> how much I increase my struct module in include/linux/module.h I see
> thes same sh_size. Do folks see same?
> 
> nm --print-size --size-sort fs/xfs/xfs.ko | grep __this_module
> 0000000000000000 0000000000000500 D __this_module
> 
> This is what is supposed to make the final part of layout_and_allocate() work:
> 
> 	mod = (void *)info->sechdrs[info->index.mod].sh_addr;
> 
> This works off of the copy of the module. Let's recall that
> setup_load_info() sets the copy mod to:
> 
> 	info->mod = (void *)info->hdr + info->sechdrs[info->index.mod].sh_offset;
> 
> The memcpy() in move_module() is what *should* be copying over the entire
> mod stuff properly over, that includes the mod->klp for live patching
> but also any new data we muck with in-kernel as the new mod->mem stuff
> in layout_sections(). In short, anything in struct module should be
> shoved into an ELF section. But I'm not quite sure this is all right.

I dug into that code years ago, and the above sounds right.

The .ko file has a .gnu.linkonce.this_module section whose data is just
the original "struct module __this_module" which is created by the
module build (from foo.mod.c).

At the beginning of the finit_module() syscall, the .ko file's ELF
sections get copied (and optionally decompressed) into kernel memory.
Then 'mod' just points to the copied __this_module struct.

Then mod->klp (and possibly mod->taint) get set.

Then in layout_and_allocate(), that 'mod' gets memcpy'd into the second
(and final) in-kernel copy of 'struct module':

 		if (shdr->sh_type != SHT_NOBITS)
 			memcpy(dest, (void *)shdr->sh_addr, shdr->sh_size);
 		/* Update sh_addr to point to copy in image. */
 		shdr->sh_addr = (unsigned long)dest;

I suspect you don't see the size changing when you add to 'struct
module' because it's ____cacheline_aligned.

It's all rather obtuse, but working as designed as far as I can tell.

-- 
Josh