RE: [PATCH v9 12/28] net/rnp: add support link update operations

["11" <caowenbo@mucse.com>]

Hi Stephen,

Thanks for your explain abort the memcpy,  on Fedora41  I had I understood
why
rte_memcpy will cause compile fault. I will be carefull to use memcpy,
when a address of src or dst is not alignment.

Regards Wenbo

> -----Original Message-----
> From: Stephen Hemminger <stephen@networkplumber.org>
> Sent: 2025年2月12日 23:38
> To: 11 <caowenbo@mucse.com>
> Cc: thomas@monjalon.net; dev@dpdk.org; ferruh.yigit@amd.com;
> andrew.rybchenko@oktetlabs.ru; yaojun@mucse.com
> Subject: Re: [PATCH v9 12/28] net/rnp: add support link update operations
> 
> On Wed, 12 Feb 2025 11:21:49 +0800
> "11" <caowenbo@mucse.com> wrote:
> 
> > Hi Stephen,
> > For memcpy what ever base code or other code all used memcpy not
> > rte_memcpy ?
> > Even the memory is malloc from rte_malloc/zmalloc ?
> >
> > Regards Wenbo.
> 
> 
> The documentation doesn't make it clear but rte_memcpy is the same as
> memcpy.
> It only exists because for some cases (like older versions of Gcc and
differences in
> alignment assumptions) when running micro benchmarks rte_memcpy was
> faster.
> 
> All new code should only use memcpy() unless there is a benchmark test
that
> shows a noticable difference. The reason is that tools like Coverity, Gcc,
Clang,
> and PVS studio all treat memcpy specially and can do checks for access
outside of
> range.