lkml.org 
[lkml]   [2019]   [Mar]   [20]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [RFC v2 1/2] vfio/pci: export common symbols in vfio-pci
On Wed, 20 Mar 2019 11:49:37 +0000
"Liu, Yi L" <yi.l.liu@intel.com> wrote:

> > From: Alex Williamson [mailto:alex.williamson@redhat.com]
> > Sent: Wednesday, March 20, 2019 2:14 AM
> > To: Liu, Yi L <yi.l.liu@intel.com>
> > Subject: Re: [RFC v2 1/2] vfio/pci: export common symbols in vfio-pci
> >
> > On Tue, 12 Mar 2019 16:18:22 +0800
> > "Liu, Yi L" <yi.l.liu@intel.com> wrote:
> >
> > > This patch exports the following symbols from vfio-pci driver
> > > for vfio-pci alike driver. e.g. vfio-pci-mdev driver
> > >
> > > *) vfio_pci_set_vga_decode();
> > > *) vfio_pci_release();
> > > *) vfio_pci_open();
> > > *) vfio_pci_register_dev_region();
> > > *) vfio_pci_ioctl();
> > > *) vfio_pci_rw();
> > > *) vfio_pci_mmap();
> > > *) vfio_pci_request();
> > > *) vfio_pci_probe_misc();
> > > *) vfio_pci_remove_misc();
> > > *) vfio_err_handlers;
> > > *) vfio_pci_reflck_attach();
> > > *) vfio_pci_reflck_put();
> >
> > Exporting all these symbols scares me a bit. They're GPL and we don't
> > guarantee a kernel ABI, but I don't really want to support arbitrary
> > use cases either. What if we re-factored the shared bits into a common
> > file and just linked them together? Thanks,
>
> Hi, Alex,
>
> Before refactor the code, I'd like to check with you on the module
> parameters for the two modules. The existing vfio-pci driver has
> some module parameters. e.g. ids, nointxmask, disable_idle_d3.
> For future usage and maintain, I think it is better to let the two
> drivers have same parameters. However, I'm not 100% on whether
> we want to allow user load vfio-pci.ko and vfio-pci-mdev.ko with
> different parameter value? e.g. load vfio-pci.ko with nointxmask=false
> while load vfio-pci-mdev.ko with nointxmask=true. How about your
> opinion on it?

Hi Yi,

I agree that it makes sense to retain the module options for the mdev
wrapped version, but I expect we should also allow dissimilar user
settings. If those lived in the common code that gets linked separately
with each module, that should work fine, I think. We can worry about
refactoring for future driver that might not want those options later.
Thanks,

Alex

\
 
 \ /
  Last update: 2019-03-20 20:22    [W:0.071 / U:0.416 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site