I didn't modified the attach function,I modified the rte_eal_config_create.
primay process first mmap's address, use a lower address, to avoid
secondary process mmap the fixed address fail. the secondary still
mmap the same address as the primary.
dpdk do the same thing in rte_fbarray_init() function.
------------------origin------------------
from：DavidMarchand <david.march...@redhat.com>
to：Burakov, Anatoly <anatoly.bura...@intel.com>;
copy：hanli....@zte.com.cn;Thomas Monjalon <tho...@monjalon.net>;dev
<dev@dpdk.org>;
date ：2019年10月24日 15:38
title ：Re: [dpdk-dev] [PATCH v3] lib/librte_eal/linuxapp: fix runtime
configmmap issue
On Wed, Oct 23, 2019 at 7:47 PM Burakov, Anatoly
<anatoly.bura...@intel.com> wrote:
>
> On 23-Oct-19 11:42 AM, Li Han wrote:
> > In rte_eal_config_reattach(),the secondary mmap may fail
> > due to the rte_mem_cfg_addr already be used by others.do
> > the change just as the rte_fbarray_init() do.if no
> > base_virtaddr,use the default 0x100000000.
> >
> > v2/v3:
> > -fix code style issues
> >
> > Signed-off-by: Li Han <han....@zte.com.cn>
> > ---
>
> Nack. There's a reason why we map it at the same address, and it's to
> have all pointers working across processes. Remapping it at a different
> address has potential to break things.
Marked as rejected.
Thanks.
--
David Marchand
