[PATCH v8 00/22] vfio-ap: guest dedicated crypto adapters

From: Tony Krowiak
Date: Wed Aug 08 2018 - 10:44:54 EST


From: Tony Krowiak <akrowiak@xxxxxxxxxxxxx>

Several major objections were raised to design changes introduced in the v7
patch series, so in order to avoid an extended discussion around these
objections and to expedite acceptance of the series, the following changes
have been made for v8:

1. Removed the AP bus's ability to designate queues as 'used by host' or as
'used by alternate driver(s)'. The bind/unbind sysfs interfaces will be
used for managing the connection of AP queue devices to a zcrypt driver
or the VFIO AP driver.

2. Removed the 'activate' sysfs interfaces which allowed for over
provisioning of the mediated device as well as creation of mdevs with
overlapping matrixes. It was pointed out that both of these enhancements
break the mdev model. Consistency checking of the mdev matrix has
therefore been returned to the mediated matrix device's sysfs interfaces
for assigning adapters and domains:

* Verify that APQNs assigned to the mediated device are bound to the
VFIO AP device driver

* Verify that no APQN assigned to the mediated matrix device is assigned
to any other mediated matrix device.

4. Reworked the handling of the CRYCB in vSIE based upon patches introduced
by David in the mainline.

Notes:
=====

Patches 1-4 (by Harald) posted with this series are forthcoming via
Martins tree and are based on changes in the ap driver/bus that we use as a
foundation. They have been included here because some of the functions
in this patch series are dependent upon them.

Patches 5-6 (by David) are posted with this series because they are not
currently in our master branch. Patches 19 and 20 of this series are
dependent upon them. I believe David's patches are available in the
mainline now.

This patch series works with the v6 QEMU patches. There is no new QEMU
patchset version yet because there have been no review comments worthy of
creating a new series; only a couple of extremely minor nits.

Abstract:
========

On s390, we have cryptographic coprocessor cards, which are modeled on
Linux as devices on the AP bus. Each card can be partitioned into domains
which can be thought of as a set of hardware registers for processing
crypto commands. Crypto commands are sent to a specific domain within a
card is via a queue which is identified as a (card,domain) tuple. We model
this something like the following (assuming we have access to cards 3 and
4 and domains 1 and 2):

AP -> card3 -> queue (3,1)
-> queue (3,2)
-> card4 -> queue (4,1)
-> queue (4,2)

If we want to virtualize this, we can use a feature provided by the
hardware. We basically attach a satellite control block to our main
hardware virtualization control block and the hardware takes care of
most of the rest.

For this control block, we don't specify explicit tuples, but a list of
cards and a list of domains. The guest will get access to the cross
product.

Because of this, we need to take care that the lists provided to
different guests don't overlap; i.e., we need to enforce sane
configurations. Otherwise, one guest may get access to things like
secret keys for another guest.

The idea of this patch set is to introduce a new device, the matrix
device. This matrix device hangs off a different root and acts as the
parent node for mdev devices.

If you now want to give the tuples (4,1) and (4,2), you need to do the
following:

- Make sure the queues (4,1) and (4,2) belong to vfio_ap (see patches
#5 and #6)
- Create the mediated device.
- Assign card 4 and domains 1 and 2 to the mediated device
- Optionally activate the mediated device.

QEMU will now simply consume the mediated device and things should work.

For a complete description of the architecture and concepts underlying
the design, see the Documentation/s390/vfio-ap.txt file included with this
patch set.

v7 => v8 Change log:
===================
* Removed the AP bus's ability to designate queues as 'used by host' or as
'used by alternate driver(s)' and relying on bind/unbind for connecting
AP queue devices to a zcrypt driver or the VFIO AP driver.
* Removed 'activate' attribute from mediated device.
* Do consistency checking during device assignment:
1. Verify that APQNs assigned to the mediated device are bound to the
VFIO AP device driver
2. Verify that no APQN assigned to the mediated matrix device is assigned
to any other mediated matrix device.
* The attributes of a mediated matrix device that is in use by a guest can
not be changed - i.e., no device assignment/unassignment allowed
* A mediated matrix device that is in use by a guest can not be removed.
* Removed all printk logging from VFIO AP driver; allowing return codes
from interfaces to describe the error.
* Reworked the handling of the CRYCB in vSIE based upon patches introduced
by David in the mainline.

v6 => v7 Change log:
===================
* The AP bus gained the ability to designate queues as 'used by host'
or as 'used by alternate driver(s)'. This allows us to authorise access
(via the CRYCB) to queues that are not currently bound to the vfio_ap
driver. If a vfio_ap owned queue diss- and reapears it's guaranteed
to get bound back to the vfio_ap driver.
* The mediated device gained an 'activate' attribute. Sharing conflicts are
checked on activation now. If the device was not activated, the mdev
open still implies activation. An active ap_matrix_mdev device claims
it's resources -- an inactive does not.
* An active ap_matrix_mdev device can not be removed. An ap_matrix_mdev
that is hooked up with a guest can not be deactivated.
* An active ap_matrix_mdev device rejects assign_* and deassign_*
operations. Thus changing the CRYCB masks of a guest in order to
accomplys certain hotplug scenarios is planned, but not supported yet. In
previous versions it was possible to do those operations on a ap_matrix_mdev
that is hooked up to a guest, but the changes would take effect on the next
mdev_open.
* Synchronisation was reworked.
* The sysfs path of the parent device changed from /sys/devices/vfio_ap/matrix/
to /sys/devices/virtual/misc/vfio_ap/. The parent device is a misc
device now.
* The severity for most of the messages were reduced form error to
warning.
* We are not as thick headed about the zapq as we used to be in v6.

v5 => v6 Change log:
===================
* Added VSIE support - thanks to Pierre Morel
* Added VFIO_DEVICE_RESET ioctl
* Zeroizing AP queues when mediated device released and when
VFIO_DEVICE_RESET ioctl is invoked
* Removed /arch/s390/kvm/kvm-ap.c and arch/s390/include/asm/kvm-ap.h and
moved guest matrix configuration into vfio driver
* Removed temporary interfaces to be supplied by AP bus
* Made the variable that keeps track of mdev instance count an atomic_t
type
* Removed code iterating through vm_list to determine if another guest has
a queue .... not keep a list of matrix_mdev devices and verify against
that list. Removes the need for the kvm_lock.
* Added a sysfs attribute for the mediated matrix device to display the
matrix contained in the guest's CRYCB if a guest is using the mdev.


v4 => v5 Change log:
===================
* Verify AP queues bound to driver in mediated device open callback, prior
to configuring the matrix in the CRYCB
* Implement VFIO_DEVICE_RESET ioctl
* Zeroize queues on guest entry and exit
* Removed vnet from all email IBM email addresses referenced
* Add synchronization in mdev create/remove and open/release.

v4 => v5 Change log:
===================
* Added code to mdev open callback to ensure not more than one vfio-ap
device can be opened by a guest.
* Interpret AP instructions by default
* Removed patch implementing interface to enable/disable AP interpretation,
since that will now be done by default
* Removed patch to reset crypto attributes for ALL vcpus. That will be
submitted as a single patch since it will not be needed in this series -
i.e., it was called from the interface to enable/disable AP instructions
* All code for initializing crypto for a guest has been moved back to
kvm-s390.c, kvm_s390_crypto_init(kvm) function
* Maintaining a module reference count for the vfio_ap module so it is not
removed while a guest with AP devices is running.



v3 => v4 Change log:
===================
* Resolved issue with enabling ZCRYPT when KVM is enabled by using
#ifdef ZCRYPT in relevant functions
* Added patch with a new function for resetting the crypto attributes
for all vcpus to resolve the issue raised with running vcpus getting out
of sync.
* Removed KVM_S390_VM_CRYPTO_INTERPRET_AP: Setting interpretive exec mode
from vfio_ap driver when mdev device is opened.


v2 => v3 Change log:
===================
* Set APIE in VCPU setup function
* Renamed patch 13/15:
KVM: s390: Configure the guest's CRYCB
KVM: s390: Configure the guest's AP devices
* Fixed problem with building arch/s390/kvm/kvm-ap.c when CONFIG_ZCRYPT
not selected
* Removed patch introducing VSIE support for AP pending further
investigation
* Initialized AP maximum mask sizes - i.e., APM, AQM and ADM - from info
returned from PQAP(QCI) function
* Introduced a new device attribute to the KVM_S390_VM_CRYPTO attribute
group for setting a flag via the KVM_SET_DEVICE_ATTR ioctl to indicate
whether ECA_APIE should be set or not. The flag is used in the
kvm_s390_vcpu_crypto_setup() function to set ECA_APIE in the SIE block.


v1 => v2 Change log:
===================
* Added documentation vfio-ap.txt
* Renamed vfio_ap_matrix module and device driver to vfio_ap
* Use device core device list instead of maintaining list of matrix
devices in driver
* Added VSIE support for AP
* Create matrix device before registering VFIO AP device driver with the
AP bus
* Renamed the following files in drivers/s390/crypto:
* vfio_ap_matrix.drv -> vfio_ap_drv
* vfio_ap_matrix_private.h -> vfio_ap_private.h
* vfio_ap_matrix_ops.c -> vfio_ap_ops.c
* arch/s390/include/asm/kvm/ap-matrix-config.h
* Renamed to kvm-ap.h
* Changed the data type of the bit mask fields for the matrix structure
to unsigned long and create them with DECLARE_BITMAP
* Changed #define prefixes from AP_MATRIX to KVM_AP
* Changed function and structure prefixes from ap_matrix to kvm_ap
* Added function interface to check if AP Extended Addressing (APXA)
facility is installedCRYCB_FORMAT_MASK
* Added function interface to get the maximum ID for AP mask type
* Added function interface to set the AP execution mode
* arch/s390/kvm/ap-matrix-config.c
* Renamed to kvm-ap.c
* Changed function prefixes from ap_matrix to kvm_ap
* Added function to check if AP Extended Addressing (APXA) facility is
installed
* Added function to get the maximum ID for AP mask type
* Added function to set the AP execution mode
* Added a boolean parameter to the functions that retrieve the APM, AQM
and ADM bit mask fields from the CRYCB. If true, then the function
will clear the bits in the mask before returning a reference to it
* Added validation to verify that APM, AQM and ADM bits that are set do
not exceed the maximum ID value allowed
*
* arch/s390/include/asm/kvm_host.h
* Changed defined for ECA_AP to ECA_APIE - interpretive execution mode
* Added a flag to struct kvm_s390_crypto to indicate whether the
KVM_S390_VM_CPU_FEAT_AP CPU model feature for AP facilities is set
* Added two CPU facilities features to set STFLE.12 and STFLE.15
* arch/s390/kvm/kvm-s390.c
* Added initialization for new KVM_S390_VM_CPU_FEAT_AP CPU model feature
* Removed kvm_s390_apxa_installed() function
* Changed call to kvm_s390_apxa_installed() which has been removed to a
call to new kvm_ap_apxa_installed() function.
* Added code to kvm_s390_vcpu_crypto_setup() to set the new CPU model
feature flag in the kvm_s390_crypto structure
* Added CRYCB_FORMAT_MASK to mask CRYCBD
* arch/s390/tools/gen_facilities.c
* Added STFLE.12 and STFLE.15 to struct facility _def
* drivers/s390/crypto/vfio_ap_matrix_private.h
* Changed name of file to vfio_ap.private.h
* Changed #define prefixes from VFIO_AP_MATRIX to VFIO_AP
* struct ap_matrix: removed list fields and locks
* struct vfio_ap_queue: removed list field
* Renamed functions ap_matrix_mdev_register and ap_matrix_mdev_unregister
to vfio_ap_mdev_register and vfio_ap_mdev_unregister respectively
* drivers/s390/crypto/vfio_ap_matrix_drv.c
* Renamed file to drivers/s390/crypto/vfio_ap_drv.c
* Changed all #define, structure and function prefixes to vfio_ap
* probe function
* Changed root device name for the matrix device to vfio_ap:
i.e., /sys/devices/vfio_ap/matrix
* No longer storing the AP queue device in a list, it is retrievable via
the device core
* Removed unnecessary check whether matrix device exists
* Store the vfio_ap_queue structure in the private field of the ap_queue
structure rather than using list interface
* remove function
* Retrieve vfio_ap_queue structure from the struct ap_queue private
data rather than from a list
* Removed unnecesary check
* drivers/s390/crypto/vfio_ap_matrix_ops.c
* Renamed file to vfio_ap_ops.c
* Changed #define prefixes from AP_MATRIX to VFIO_AP
* Changed function name prefixes from ap_matrix to vfio_ap
* Removed ioctl to configure the CRYCB
* create function
* Removed ap_matrix_mdev_find_by_uuid() function - function is provided
by mdev core
* Removed available_instances verification, provided by mdev core
* Removed check to see if mediated device exists, handled by mdev core
* notifier function
* Configuring matrix here instead of via ioctl
* Set interpretive execution mode for all VCPUs
* Removed R/O attributes to display adapters and domains
* Added an R/O attribute to display the matrix
* assign_control_domain mdev attribute:
* Removed check to see if the domain is installed on the linux host
* Added check to verify the control domain ID does not exceed the max
value
* assign_adapter mdev attribute:
* Added check to verify the adapter ID does not exceed the max
value
* If any APQNs configured for the mediated matrix device that
have an APID matching the adapter ID being assigned are not
bound to the vfio_ap device driver then it is assumed that the APQN is
bound to another driver and assignment will fail
* assign_domain mdev attribute:
* Added check to verify the domain ID does not exceed the max
value
* If any APQNs configured for the mediated matrix device that
have an APQI matching the domain ID being assigned are not
bound to the vfio_ap device driver then it is assumed that the APQN is
bound to another driver and assignment will fail
* tools/arch/s390/include/uapi/asm/kvm.h
* removed KVM_S390_VM_CPU_FEAT_AP feature definition

David Hildenbrand (2):
KVM: s390: vsie: simulate VCPU SIE entry/exit
KVM: s390: introduce and use KVM_REQ_VSIE_RESTART

Harald Freudenberger (4):
s390/zcrypt: Add ZAPQ inline function.
s390/zcrypt: Review inline assembler constraints.
s390/zcrypt: Show load of cards and queues in sysfs
s390/zcrypt: Integrate ap_asm.h into include/asm/ap.h.

Pierre Morel (2):
KVM: s390: Clear Crypto Control Block when using vSIE
KVM: s390: Handling of Cypto control block in VSIE

Tony Krowiak (14):
KVM: s390: refactor crypto initialization
s390: vfio-ap: base implementation of VFIO AP device driver
s390: vfio-ap: register matrix device with VFIO mdev framework
s390: vfio-ap: sysfs interfaces to configure adapters
s390: vfio-ap: sysfs interfaces to configure domains
s390: vfio-ap: sysfs interfaces to configure control domains
s390: vfio-ap: sysfs interface to view matrix mdev matrix
KVM: s390: interfaces to clear CRYCB masks
s390: vfio-ap: implement mediated device open callback
s390: vfio-ap: implement VFIO_DEVICE_GET_INFO ioctl
s390: vfio-ap: zeroize the AP queues.
s390: vfio-ap: implement VFIO_DEVICE_RESET ioctl
KVM: s390: CPU model support for AP virtualization
s390: doc: detailed specifications for AP virtualization

Documentation/s390/vfio-ap.txt | 616 +++++++++++++++++++++++
MAINTAINERS | 12 +
arch/s390/Kconfig | 11 +
arch/s390/include/asm/ap.h | 284 ++++++++++-
arch/s390/include/asm/kvm_host.h | 7 +
arch/s390/include/uapi/asm/kvm.h | 1 +
arch/s390/kvm/kvm-s390.c | 131 ++++--
arch/s390/kvm/kvm-s390.h | 1 +
arch/s390/kvm/vsie.c | 244 ++++++++-
arch/s390/tools/gen_facilities.c | 2 +
drivers/s390/crypto/Makefile | 4 +
drivers/s390/crypto/ap_asm.h | 236 ---------
drivers/s390/crypto/ap_bus.c | 21 +-
drivers/s390/crypto/ap_bus.h | 1 +
drivers/s390/crypto/ap_card.c | 1 -
drivers/s390/crypto/ap_queue.c | 1 -
drivers/s390/crypto/vfio_ap_drv.c | 137 +++++
drivers/s390/crypto/vfio_ap_ops.c | 885 +++++++++++++++++++++++++++++++++
drivers/s390/crypto/vfio_ap_private.h | 106 ++++
drivers/s390/crypto/zcrypt_card.c | 12 +
drivers/s390/crypto/zcrypt_queue.c | 12 +
include/uapi/linux/vfio.h | 2 +
22 files changed, 2373 insertions(+), 354 deletions(-)
create mode 100644 Documentation/s390/vfio-ap.txt
delete mode 100644 drivers/s390/crypto/ap_asm.h
create mode 100644 drivers/s390/crypto/vfio_ap_drv.c
create mode 100644 drivers/s390/crypto/vfio_ap_ops.c
create mode 100644 drivers/s390/crypto/vfio_ap_private.h