[dpdk-dev] about memzone name size issue

Xu, HuilongX huilongx.xu at intel.com
Wed Jun 1 02:34:34 CEST 2016


Hi ,
I will tester-by this patch, thanks  a lot.

> -----Original Message-----
> From: Iremonger, Bernard
> Sent: Tuesday, May 31, 2016 9:59 PM
> To: Wiles, Keith; Xu, HuilongX; dev at dpdk.org
> Subject: RE: [dpdk-dev] about memzone name size issue
> 
> Hi Huilong,
> 
> > -----Original Message-----
> > From: dev [mailto:dev-bounces at dpdk.org] On Behalf Of Wiles, Keith
> > Sent: Tuesday, May 31, 2016 2:18 PM
> > To: Xu, HuilongX <huilongx.xu at intel.com>; dev at dpdk.org
> > Subject: Re: [dpdk-dev] about memzone name size issue
> >
> > >Hi all,
> > >I find a issue on link_bonding unit test case.
> > >
> > >When I run model6 test case, will generate core dump error.
> > >I debug it, find the error code in function:
> > >rte_mempool_create_empty(const char *name, unsigned n, unsigned
> > elt_size,
> > >                unsigned cache_size, unsigned private_data_size,
> > >                int socket_id, unsigned flags) {
> > >          .......................
> > >                ret = snprintf(mz_name, sizeof(mz_name),
> > RTE_MEMPOOL_MZ_FORMAT, name);
> > >                if (ret < 0 || ret >= (int)sizeof(mz_name)) {
> > >                                rte_errno = ENAMETOOLONG;
> > >                                goto exit_unlock;
> > >                }
> > >           .........................
> > >}
> > >The memzone name size only 32 bytes, but the mz_name in link_bonding
> is
> > bigger 32 bytes. Could we set memzone name size to 64 bytes ?
> > >Thanks  a lot
> >
> > Having a name of 32 bytes is pretty big IMO, what would be a better
> reason
> > for changing a internal structure as it will take two releases to make
> that
> > change?
> > How big is the name string you are passing into the routine?
> > >
> >
> >
> The following patches were submitted last Friday to fix this issue with
> the link bonding tests.
> 
> http://dpdk.org/dev/patchwork/patch/13052/
> http://dpdk.org/dev/patchwork/patch/13053/
> 
> Regards,
> 
> Bernard.
> 



More information about the dev mailing list