Re: [dpdk-dev] [PATCH] eal: fix errno on service cores init failure

[Olivier Matz <olivier.matz@6wind.com>]

Hi Harry,

On Thu, Nov 26, 2020 at 02:46:30PM +0000, Van Haaren, Harry wrote:
> > -----Original Message-----
> > From: Olivier Matz <olivier.matz@6wind.com>
> > Sent: Thursday, November 26, 2020 2:25 PM
> > To: dev@dpdk.org
> > Cc: Richardson, Bruce <bruce.richardson@intel.com>; Jerin Jacob
> > <jerin.jacob@caviumnetworks.com>; Van Haaren, Harry
> > <harry.van.haaren@intel.com>; stable@dpdk.org
> > Subject: [PATCH] eal: fix errno on service cores init failure
> > 
> > Currently, when rte_service_init() fails at initialization, we
> > see the following message:
> > 
> >   Cannot init EAL: Exec format error
> > 
> > This error code does describe the real issue. Instead, use the error
> > code returned by the function.
> 
> Should the above read as "does NOT describe" .. ?
> 
> > Fixes: e39824500825 ("service: initialize with EAL")
> > Cc: stable@dpdk.org
> > 
> > Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
> 
> Few comments below, assuming agree on those, add my Ack on v2?
> 
> Checked, -ENOMEM and -EALREADY are returned today, which seem
> better descriptive terms. Thanks for fixing,
> 
> Acked-by: Harry van Haaren <harry.van.haaren@intel.com>
> 
> 
> > ---
> >  lib/librte_eal/freebsd/eal.c | 4 ++--
> >  lib/librte_eal/linux/eal.c   | 4 ++--
> >  2 files changed, 4 insertions(+), 4 deletions(-)
> > 
> > diff --git a/lib/librte_eal/freebsd/eal.c b/lib/librte_eal/freebsd/eal.c
> > index d6ea023750..51478358c7 100644
> > --- a/lib/librte_eal/freebsd/eal.c
> > +++ b/lib/librte_eal/freebsd/eal.c
> > @@ -906,7 +906,7 @@ rte_eal_init(int argc, char **argv)
> >  	ret = rte_service_init();
> >  	if (ret) {
> >  		rte_eal_init_alert("rte_service_init() failed");
> > -		rte_errno = ENOEXEC;
> > +		rte_errno = -ret;
> >  		return -1;
> >  	}
> 
> Here we set   rte_errno   as   -ret,  as in rte_service_init() we return the negative, e.g.   -ENOMEM.
> Perhaps it is cleaner to to return ENOMEM from rte_service_init(), and avoid the duplicate negation?
> 
> rte_service_init() is not exported publicly in the .map file, so is internal only, and hence not an ABI break.

I think returning -errno is common in dpdk, so I'll keep it like
this. Or it can eventually return -1 and set rte_errno.

> 
> 
> > @@ -922,7 +922,7 @@ rte_eal_init(int argc, char **argv)
> >  	 */
> >  	ret = rte_service_start_with_defaults();
> >  	if (ret < 0 && ret != -ENOTSUP) {
> > -		rte_errno = ENOEXEC;
> > +		rte_errno = -ret;
> >  		return -1;
> >  	}
> > 
> > diff --git a/lib/librte_eal/linux/eal.c b/lib/librte_eal/linux/eal.c
> > index a4161be630..32b48c3de9 100644
> > --- a/lib/librte_eal/linux/eal.c
> > +++ b/lib/librte_eal/linux/eal.c
> > @@ -1273,7 +1273,7 @@ rte_eal_init(int argc, char **argv)
> >  	ret = rte_service_init();
> >  	if (ret) {
> >  		rte_eal_init_alert("rte_service_init() failed");
> > -		rte_errno = ENOEXEC;
> > +		rte_errno = -ret;
> >  		return -1;
> >  	}
> > 
> > @@ -1295,7 +1295,7 @@ rte_eal_init(int argc, char **argv)
> >  	 */
> >  	ret = rte_service_start_with_defaults();
> >  	if (ret < 0 && ret != -ENOTSUP) {
> > -		rte_errno = ENOEXEC;
> > +		rte_errno = -ret;
> >  		return -1;
> >  	}
> > 
> > --
> > 2.25.1
>