Re: [PATCH 2/2] module: fix bne2 "gave up waiting for init of module libcrc32c"

[Rusty Russell <rusty@rustcorp.com.au>]

On Fri, 4 Jun 2010 01:54:19 am Linus Torvalds wrote:
> 
> On Thu, 3 Jun 2010, Rusty Russell wrote:
> > 
> > However, you're right that it has potential.  I'll rename module_info to
> > load_info if you don't mind tho: contains more semantic punch IMHO.
> 
> Umm. One problem is that you will almost certainly eventually want to 
> expose that to the architecture "fixup" routines (ie things like 
> module_frob_arch_sections(), arch_mod_section_prepend()), and at that 
> point "load_info" is a horribly bad structure name, since it would show 
> up in <linux/module.h> and thus be exported all over.
> 
> At least call it "struct module_load_info". But yes, I do agree that the 
> "load" part is important.

Looking at the arch code, it has the advantage that it's self-contained.
They've been pleasantly undemanding from the core over the years; I think
archs doing tricky things with elf prefer to parse the object themselves
anyway.  And I'm not sure they want to revisit it, either.

So I don't think we'd win much from changing them.  I'm wrong later, I'll
prepend "module_" to the struct name as an internal change then hit them
all.

> I looked at that particularly when doing that whole
> 
> 	mod = setup_module_info(&info);
> 	if (IS_ERR(mod)) {
> 		err = PTR_ERR(mod);
> 		goto free_hdr;
> 	}
> 
> thing, because that made "mod" have _three_ totally different values 
> (error, before, after) when jumping out to the failure paths. 

Yep, it now is back to sanity.  Let's see if today's linux-next is
happy.

If so, do you want just the fixes or the whole refactoring too, while
it's nice and fresh?

Thanks!
Rusty.