lkml.org 
[lkml]   [2010]   [Sep]   [27]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
SubjectRe: [PATCH 3/3] VFIO V4: VFIO driver: Non-privileged user level PCI drivers
From
Date
On Mon, 2010-09-27 at 13:54 +0200, Michael S. Tsirkin wrote:
> On Wed, Sep 22, 2010 at 02:18:24PM -0700, Tom Lyon wrote:
> >
> > Signed-off-by: Tom Lyon <pugs@cisco.com>
>
> Some comments on the pci bits:
>
> After going over them for the Nth time - something needs to be done
> with the rvirt/write tables. I doubt anyone besides me and you
> has gone over them:

/me bites tongue...

> > +static void vfio_bar_fixup(struct vfio_dev *vdev)
> > +{
>
> So you do this on each read?
> Why don't you mask the appropriate bits on write?
> This is what real hardware does, after all.
> Then you won't need the bardirty field.
>
>
> > + struct pci_dev *pdev = vdev->pdev;
> > + int bar;
> > + u32 *lp;
> > + u64 mask;
> > +
> > + for (bar = 0; bar <= 5; bar++) {
> > + if (pci_resource_start(pdev, bar))
> > + mask = ~(pci_resource_len(pdev, bar) - 1);
> > + else
> > + mask = 0;
> > + lp = (u32 *)vdev->vconfig + PCI_BASE_ADDRESS_0 + 4*bar;
> > + *lp &= (u32)mask;
> > +
> > + if (pci_resource_flags(pdev, bar) & IORESOURCE_IO)
> > + *lp |= PCI_BASE_ADDRESS_SPACE_IO;
> > + else if (pci_resource_flags(pdev, bar) & IORESOURCE_MEM) {
> > + *lp |= PCI_BASE_ADDRESS_SPACE_MEMORY;
> > + if (pci_resource_flags(pdev, bar) & IORESOURCE_PREFETCH)
> > + *lp |= PCI_BASE_ADDRESS_MEM_PREFETCH;
> > + if (pci_resource_flags(pdev, bar) & IORESOURCE_MEM_64) {
> > + *lp |= PCI_BASE_ADDRESS_MEM_TYPE_64;
> > + lp++;
> > + *lp &= (u32)(mask >> 32);
> > + bar++;
> > + }
> > + }
> > + }
> > +
> > + if (pci_resource_start(pdev, PCI_ROM_RESOURCE))
>
> Not if (pci_resource_len)?

Using this test covers both zero length/unimplemented BARs and BARs that
aren't mapped by the host. Unmapped BARs on the host are exposed as
unimplemented BARs to the vfio user. I wonder if this is still
necessary if we do the iomap/rom_map on open as I propose.

> > + mask = ~(pci_resource_len(pdev, PCI_ROM_RESOURCE) - 1);
> > + else
> > + mask = 0;
> > + lp = (u32 *)vdev->vconfig + PCI_ROM_ADDRESS;
> > + *lp &= (u32)mask;
> > +
> > + vdev->bardirty = 0;
>
> Aren't the pci values in little endian format?
> If so doing operations on them in native endianness is wrong.
> sparse generally is good at catching these, but you will
> have to avoid so many type casts and annotate endian-ness
> for it to be of any use.

Yep, I expect there are a number of endian issues here, good call
pointing them out.

Alex



\
 
 \ /
  Last update: 2010-09-27 23:01    [W:1.308 / U:1.112 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site