Re: [PATCH 4/4] vfio-pci/zdev: Introduce the zPCI I/O vfio region

From: Matthew Rosato
Date: Thu Jan 21 2021 - 11:01:06 EST


On 1/21/21 5:01 AM, Niklas Schnelle wrote:


On 1/19/21 9:02 PM, Matthew Rosato wrote:
Some s390 PCI devices (e.g. ISM) perform I/O operations that have very
specific requirements in terms of alignment as well as the patterns in
which the data is read/written. Allowing these to proceed through the
typical vfio_pci_bar_rw path will cause them to be broken in up in such a
way that these requirements can't be guaranteed. In addition, ISM devices
do not support the MIO codepaths that might be triggered on vfio I/O coming
from userspace; we must be able to ensure that these devices use the
non-MIO instructions. To facilitate this, provide a new vfio region by
which non-MIO instructions can be passed directly to the host kernel s390
PCI layer, to be reliably issued as non-MIO instructions.

This patch introduces the new vfio VFIO_REGION_SUBTYPE_IBM_ZPCI_IO region
and implements the ability to pass PCISTB and PCILG instructions over it,
as these are what is required for ISM devices.

Signed-off-by: Matthew Rosato <mjrosato@xxxxxxxxxxxxx>
---
drivers/vfio/pci/vfio_pci.c | 8 ++
drivers/vfio/pci/vfio_pci_private.h | 6 ++
drivers/vfio/pci/vfio_pci_zdev.c | 158 ++++++++++++++++++++++++++++++++++++
include/uapi/linux/vfio.h | 4 +
include/uapi/linux/vfio_zdev.h | 33 ++++++++
5 files changed, 209 insertions(+)

Related to the discussion on the QEMU side, if we have a check
to make sure this is only used for ISM, then this patch should
make that clear in its wording and also in the paths
(drivers/vfio/pci/vfio_pci_ism.c instead of vfio_pci_zdev.c.)
This also has precedent with the region for IGD in
drivers/vfio/pci/vfio_pci_igd.c.


This is a fair point, but just to tie up threads here -- the QEMU discussion has since moved towards making the use-case less restrictive rather than ISM-only (though getting ISM working is still the motivating factor here). So as such I think I'd prefer to keep this in vfio_pci_zdev.c as other hardware could use the region if they meet the necessary criteria.