Re: [dpdk-dev] [PATCH] vfio: correct system call error checking

["Richardson, Bruce" <bruce.richardson@intel.com>]

> -----Original Message-----
> From: Neil Horman [mailto:nhorman@tuxdriver.com]
> Sent: Tuesday, June 17, 2014 1:40 PM
> To: Richardson, Bruce
> Cc: dev@dpdk.org
> Subject: Re: [dpdk-dev] [PATCH] vfio: correct system call error checking
> 
> On Tue, Jun 17, 2014 at 08:21:29PM +0000, Richardson, Bruce wrote:
> > > -----Original Message-----
> > > From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Neil Horman
> > > Sent: Tuesday, June 17, 2014 12:04 PM
> > > To: dev@dpdk.org
> > > Subject: [dpdk-dev] [PATCH] vfio: correct system call error checking
> > >
> > > Noticed today that ioctl error code return checking was incorrect in some of
> the
> > > vfio code.  ioctl can return a negative value if the system detects an error
> > > before the target device/driver can produce a return code.  The dpdk vfio
> code
> > > only checks specfically for the values that it expects, which leaves it open to
> > > accepting unexpected error codes as success.  For instance, if the vfio layer
> > > noted that the iommu driver hadn't finished registering yet, it would return
> an
> > > -EINVAL error code, but the dpdk would accept that as success, becuase it
> > > wasn't
> > > 0.
> > >
> > > Fix this to specifically check for < 0 error codes
> > >
> > > Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
> > > CC: Thomas Monjalon <thomas.monjalon@6wind.com>
> > > ---
> > >  lib/librte_eal/linuxapp/eal/eal_pci_vfio.c | 8 ++++----
> > >  1 file changed, 4 insertions(+), 4 deletions(-)
> > >
> > > diff --git a/lib/librte_eal/linuxapp/eal/eal_pci_vfio.c
> > > b/lib/librte_eal/linuxapp/eal/eal_pci_vfio.c
> > > index 4de6061..65aa8ad 100644
> > > --- a/lib/librte_eal/linuxapp/eal/eal_pci_vfio.c
> > > +++ b/lib/librte_eal/linuxapp/eal/eal_pci_vfio.c
> > > @@ -319,16 +319,16 @@ pci_vfio_get_container_fd(void)
> > >
> > >  		/* check VFIO API version */
> > >  		ret = ioctl(vfio_container_fd, VFIO_GET_API_VERSION);
> > > -		if (ret != VFIO_API_VERSION) {
> > > -			RTE_LOG(ERR, EAL, "  unknown VFIO API version!\n");
> > > +		if ((ret < 0) || (ret != VFIO_API_VERSION)) {
> > > +			RTE_LOG(ERR, EAL, "  unknown VFIO API version! errno
> > > = %d\n", errno);
> > >  			close(vfio_container_fd);
> > >  			return -1;
> > >  		}
> >
> > Not sure how this change improves things, since the existing check will already
> trigger an error on all values <0. Can you please clarify why you think this needs
> to be changed?
> Ah, my bad, the ret < 0 is superfulous, as the != already catches it, but the
> log message change is valuable in that it differentiates bad API version
> detection from other system errors.  I can respin that if you like.
> Neil

Perhaps a respin separating out ioctl errors vs version errors might be good, giving different error messages for each case.

/Bruce