Re: [PATCH v6 bpf-next 0/5] bpf_prog_pack followup

[Song Liu <songliubraving@fb.com>]

> On Jul 11, 2022, at 9:18 PM, Luis Chamberlain <mcgrof@kernel.org> wrote:
> 
> On Sat, Jul 09, 2022 at 01:14:23AM +0000, Song Liu wrote:
>>> On Jul 8, 2022, at 3:24 PM, Luis Chamberlain <mcgrof@kernel.org> wrote:
>>> 
>>>> 1) Rename module_alloc_huge as module_alloc_text_huge();
>>> 
>>> module_alloc_text_huge() is too long, but I've suggested names before
>>> which are short and generic, and also suggested that if modules are
>>> not the only users this needs to go outside of modules and so
>>> vmalloc_text_huge() or whatever.
>>> 
>>> To do this right it begs the question why we don't do that for the
>>> existing module_alloc(), as the users of this code is well outside of
>>> modules now. Last time a similar generic name was used all the special
>>> arch stuff was left to be done by the module code still, but still
>>> non-modules were still using that allocator. From my perspective the
>>> right thing to do is to deal with all the arch stuff as well in the
>>> generic handler, and have the module code *and* the other users which
>>> use module_alloc() to use that new caller as well.
>> 
>> The key difference between module_alloc() and the new API is that the 
>> API will return RO+X memory, and the user need text-poke like API to
>> modify this buffer. Archs that do not support text-poke will not be
>> able to use the new API. Does this sound like a reasonable design?

[...]

> I believe you are mentioning requiring text_poke() because the way
> eBPF code uses the module_alloc() is different. Correct me if I'm
> wrong, but from what I gather is you use the text_poke_copy() as the data
> is already RO+X, contrary module_alloc() use cases. You do this since your
> bpf_prog_pack_alloc() calls set_memory_ro() and set_memory_x() after
> module_alloc() and before you can use this memory. This is a different type
> of allocator. And, again please correct me if I'm wrong but now you want to
> share *one* 2 MiB huge-page for multiple BPF programs to help with the
> impact of TLB misses.

Yes, sharing 1x 2MiB huge page is the main reason to require text_poke. 
OTOH, 2MiB huge pages without sharing is not really useful. Both kprobe
and ftrace only uses a fraction of a 4kB page. Most BPF programs and 
modules cannot use 2MiB either. Therefore, vmalloc_rw_exec() doesn't add
much value on top of current module_alloc(). 

> A vmalloc_ro_exec() by definition would imply a text_poke().
> 
> Can kprobes, ftrace and modules use it too? It would be nice
> so to not have to deal with the loose semantics on the user to
> have to use set_vm_flush_reset_perms() on ro+x later, but
> I think this can be addressed separately on a case by case basis.

I am pretty confident that kprobe and ftrace can share huge pages with 
BPF programs. I haven't looked into all the details with modules, but 
given CONFIG_ARCH_WANTS_MODULES_DATA_IN_VMALLOC, I think it is also 
possible. 

Once this is done, a regular system (without huge BPF program or huge
modules) will just use 1x 2MB page for text from module, ftrace, kprobe, 
and bpf programs. 

> 
> But a vmalloc_ro_exec() with a respective free can remove the
> requirement to do set_vm_flush_reset_perms().

Removing the requirement to set_vm_flush_reset_perms() is the other
reason to go directly to vmalloc_ro_exec(). 

My current version looks like this:

void *vmalloc_exec(unsigned long size);
void vfree_exec(void *ptr, unsigned int size);

ro is eliminated as there is no rw version of the API. 

The ugly part is @size for vfree_exec(). We need it to share huge 
pages. 

Under the hood, it looks similar to current bpf_prog_pack_alloc
and bpf_prog_pack_free. 

Thanks,
Song