[dpdk-dev] [PATCH v3] drivers: advertise kmod dependencies in pmdinfo

[Bruce Richardson]

On Fri, Dec 16, 2016 at 07:37:38AM -0500, Neil Horman wrote:
> On Fri, Dec 16, 2016 at 10:22:08AM +0100, Olivier Matz wrote:
> > Hi,
> > 
> > On Thu, 15 Dec 2016 09:22:07 -0800, Stephen Hemminger
> > <stephen at networkplumber.org> wrote:
> > > On Thu, 15 Dec 2016 11:09:12 -0500
> > > Neil Horman <nhorman at tuxdriver.com> wrote:
> > > 
> > > > On Thu, Dec 15, 2016 at 02:46:39PM +0100, Olivier Matz wrote:  
> > > > > Add a new macro RTE_PMD_REGISTER_KMOD_DEP() that allows a driver
> > > > > to declare the list of kernel modules required to run properly.
> > > > > 
> > > > > Today, most PCI drivers require uio/vfio.
> > > > > 
> > > > > Signed-off-by: Olivier Matz <olivier.matz at 6wind.com>
> > > > > Acked-by: Fiona Trahe <fiona.trahe at intel.com>
> > > > > ---
> > > > > 
> > > > > v2 -> v3:
> > > > > - fix kmods deps advertised by mellanox drivers as pointed out
> > > > >   by Adrien
> > > > > 
> > > > > v1 ->
> > > > > v2:                                                                                                
> > > > > - do not advertise uio_pci_generic for vf drivers
> > > > > - rebase on top of head: use new driver names and prefix
> > > > >   macro with
> > > > > RTE_                                                                                       
> > > > > 
> > > > > rfc -> v1:
> > > > > - the kmod information can be per-device using a modalias-like
> > > > >   pattern
> > > > > - change syntax to use '&' and '|' instead of ',' and ':'
> > > > > - remove useless prerequisites in kmod lis: no need to
> > > > >   specify both uio and uio_pci_generic, only the latter is
> > > > >   required
> > > > > - update kmod list in szedata2 driver
> > > > > - remove kmod list in qat driver: it requires more than just
> > > > > loading a kmod, which is described in documentation
> > > > > 
> > > > >  buildtools/pmdinfogen/pmdinfogen.c      |  1 +
> > > > >  buildtools/pmdinfogen/pmdinfogen.h      |  1 +
> > > > >  drivers/net/bnx2x/bnx2x_ethdev.c        |  2 ++
> > > > >  drivers/net/bnxt/bnxt_ethdev.c          |  1 +
> > > > >  drivers/net/cxgbe/cxgbe_ethdev.c        |  1 +
> > > > >  drivers/net/e1000/em_ethdev.c           |  1 +
> > > > >  drivers/net/e1000/igb_ethdev.c          |  2 ++
> > > > >  drivers/net/ena/ena_ethdev.c            |  1 +
> > > > >  drivers/net/enic/enic_ethdev.c          |  1 +
> > > > >  drivers/net/fm10k/fm10k_ethdev.c        |  1 +
> > > > >  drivers/net/i40e/i40e_ethdev.c          |  1 +
> > > > >  drivers/net/i40e/i40e_ethdev_vf.c       |  1 +
> > > > >  drivers/net/ixgbe/ixgbe_ethdev.c        |  2 ++
> > > > >  drivers/net/mlx4/mlx4.c                 |  2 ++
> > > > >  drivers/net/mlx5/mlx5.c                 |  1 +
> > > > >  drivers/net/nfp/nfp_net.c               |  1 +
> > > > >  drivers/net/qede/qede_ethdev.c          |  2 ++
> > > > >  drivers/net/szedata2/rte_eth_szedata2.c |  2 ++
> > > > >  drivers/net/thunderx/nicvf_ethdev.c     |  1 +
> > > > >  drivers/net/virtio/virtio_ethdev.c      |  1 +
> > > > >  drivers/net/vmxnet3/vmxnet3_ethdev.c    |  1 +
> > > > >  lib/librte_eal/common/include/rte_dev.h | 25
> > > > > +++++++++++++++++++++++++ tools/dpdk-pmdinfo.py
> > > > > |  5 ++++- 23 files changed, 56 insertions(+), 1 deletion(-)
> > > > >     
> > > > Its odd that all devices, regardless of vendor should depend on the
> > > > igb_uio module.  It seems to me that depending on uio_pci_generic
> > > > or vfio is sufficient.
> > 
> > igb_uio is the historical uio module of dpdk. Although it is called
> > igb_uio, it is not specific to Intel drivers.
> > 
> > Most drivers declare "igb_uio | uio_pci_generic | vfio", which means
> > that any of the 3 kernel modules can be used.
> > 
> > I think there are some cases where people will prefer using igb_uio,
> > for instance to use a vf pmd in a vm where there is no iommu,
> > and where the kernel vfio module does not support the no-iommu mode.
> > 
> > 
> > > 
> > > Yes it seems just a special case extension for Mellanox drivers.
> > 
> > Kmod deps are different whether it's a vf driver or not.
> > Mellanox drivers are not the only drivers that do not require uio,
> > there is also szedata2.
> > 
> > Is it an argument for not including this patch?
> > 
> Speaking only for myself, I'm not suggesting the patch not be included, only
> questioning the need to list igb_uio as an optional dependency.  From what I
> understand uio_pci_generic is equaly capable of being used in a vf as igb_uio,
> and so it seems like its sufficient to list in the deps alone, or am I missing
> something?
> 

Unfortunately uio_pci_generic cannot be used with VFs either inside or
outside a VM, as VFs use MSI/MSI-X rather than legacy interrupts. For
newer kernels with vfio-noiommu, that is an option, but for any other
case, igb_uio is needed.

/Bruce