[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <20190528155428.0f149994@x1.home>
Date: Tue, 28 May 2019 15:54:28 -0600
From: Alex Williamson <alex.williamson@...hat.com>
To: Cornelia Huck <cohuck@...hat.com>
Cc: Pierre Morel <pmorel@...ux.ibm.com>, sebott@...ux.vnet.ibm.com,
gerald.schaefer@...ibm.com, pasic@...ux.vnet.ibm.com,
borntraeger@...ibm.com, walling@...ux.ibm.com,
linux-s390@...r.kernel.org, iommu@...ts.linux-foundation.org,
joro@...tes.org, linux-kernel@...r.kernel.org, kvm@...r.kernel.org,
schwidefsky@...ibm.com, heiko.carstens@...ibm.com,
robin.murphy@....com
Subject: Re: [PATCH v3 2/3] vfio: zpci: defining the VFIO headers
On Thu, 23 May 2019 18:24:33 +0200
Cornelia Huck <cohuck@...hat.com> wrote:
> On Thu, 23 May 2019 14:25:25 +0200
> Pierre Morel <pmorel@...ux.ibm.com> wrote:
>
> > We define a new device region in vfio.h to be able to
> > get the ZPCI CLP information by reading this region from
> > userland.
> >
> > We create a new file, vfio_zdev.h to define the structure
> > of the new region we defined in vfio.h
> >
> > Signed-off-by: Pierre Morel <pmorel@...ux.ibm.com>
> > ---
> > include/uapi/linux/vfio.h | 4 ++++
> > include/uapi/linux/vfio_zdev.h | 34 ++++++++++++++++++++++++++++++++++
> > 2 files changed, 38 insertions(+)
> > create mode 100644 include/uapi/linux/vfio_zdev.h
> >
> > diff --git a/include/uapi/linux/vfio.h b/include/uapi/linux/vfio.h
> > index 8f10748..56595b8 100644
> > --- a/include/uapi/linux/vfio.h
> > +++ b/include/uapi/linux/vfio.h
> > @@ -306,6 +306,10 @@ struct vfio_region_info_cap_type {
> > #define VFIO_REGION_TYPE_GFX (1)
> > #define VFIO_REGION_SUBTYPE_GFX_EDID (1)
> >
> > +/* IBM Subtypes */
> > +#define VFIO_REGION_TYPE_IBM_ZDEV (1)
> > +#define VFIO_REGION_SUBTYPE_ZDEV_CLP (1)
>
> I'm afraid that confuses me a bit. You want to add the region to every
> vfio-pci device when we're running under s390, right? So this does not
> depend on the device type of the actual device (which may or may not be
> from IBM), but only on the architecture?
FWIW, I don't really have a strong opinion here but I welcome the
discussion. It seems fair to me that a PCI vendor type could be used
for either a device with that vendor ID or by the vendor of the
platform. We've got a lot of address space if want to use
VFIO_REGION_TYPE_IBM_ZDEV rather than a PCI vendor type (so long as
it's updated not to conflict with the GFX type). Thanks,
Alex
> (Generally speaking, I think using regions for this makes sense,
> though.)
>
> > +
> > /**
> > * struct vfio_region_gfx_edid - EDID region layout.
> > *
> > diff --git a/include/uapi/linux/vfio_zdev.h
> > b/include/uapi/linux/vfio_zdev.h new file mode 100644
> > index 0000000..84b1a82
> > --- /dev/null
> > +++ b/include/uapi/linux/vfio_zdev.h
> > @@ -0,0 +1,34 @@
> > +/* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
> > +/*
> > + * Region definition for ZPCI devices
> > + *
> > + * Copyright IBM Corp. 2019
> > + *
> > + * Author(s): Pierre Morel <pmorel@...ux.ibm.com>
> > + */
> > +
> > +#ifndef _VFIO_ZDEV_H_
> > +#define _VFIO_ZDEV_H_
> > +
> > +#include <linux/types.h>
> > +
> > +/**
> > + * struct vfio_region_zpci_info - ZPCI information.
> > + *
> > + */
> > +struct vfio_region_zpci_info {
> > + __u64 dasm;
> > + __u64 start_dma;
> > + __u64 end_dma;
> > + __u64 msi_addr;
> > + __u64 flags;
> > + __u16 pchid;
> > + __u16 mui;
> > + __u16 noi;
> > + __u8 gid;
> > + __u8 version;
> > +#define VFIO_PCI_ZDEV_FLAGS_REFRESH 1
> > + __u8 util_str[CLP_UTIL_STR_LEN];
> > +} __packed;
> > +
> > +#endif
>
Powered by blists - more mailing lists