Re: [PATCH RFC] kprobes: Remove MODULES dependency

[Masami Hiramatsu <mhiramat@kernel.org>]

On Mon, 13 Jul 2020 08:49:55 +0300
Jarkko Sakkinen <jarkko.sakkinen@linux.intel.com> wrote:

> On Fri, Jul 10, 2020 at 01:32:38PM +0200, Peter Zijlstra wrote:
> > On Fri, Jul 10, 2020 at 07:32:57PM +0900, Masami Hiramatsu wrote:
> > > > -	page = module_alloc(PAGE_SIZE);
> > > > +	page = vmalloc(PAGE_SIZE);
> > > 
> > > No, you can not use vmalloc here. The reason why we use module_alloc()
> > > is to allocate the executable memory for trampoline code.
> > > So, you need to use vmalloc_exec() instead.
> > 
> > vmalloc_exec() would be broken too, also hch recently got rid of that
> > thing.
> > 
> > module_alloc() really is the only sane choice here.
> > 
> > We should make module_alloc() unconditionally available, and maybe even
> > rename it to text_alloc().
> 
> Thanks for the remarks.
> 
> The current module_alloc looks like this:
> 
> void * __weak module_alloc(unsigned long size)
> {
> 	return __vmalloc_node_range(size, 1, VMALLOC_START, VMALLOC_END,
> 			GFP_KERNEL, PAGE_KERNEL_EXEC, VM_FLUSH_RESET_PERMS,
> 			NUMA_NO_NODE, __builtin_return_address(0));
> }
> 
> What if I create inline functions for text_alloc() and text_memfree() and
> convert this function as:
> 
> void * __weak module_alloc(unsigned long size)
> {
> 	return text_alloc(size);
> }

Yeah, that'll be good.

Thank you,


-- 
Masami Hiramatsu <mhiramat@kernel.org>