vhost-vdpa: introduce vhost-vdpa net client

This patch set introduces a new net client type: vhost-vdpa.
vhost-vdpa net client will set up a vDPA device which is specified
by a "vhostdev" parameter.

Signed-off-by: Lingshan Zhu <lingshan.zhu@intel.com>
Signed-off-by: Tiwei Bie <tiwei.bie@intel.com>
Signed-off-by: Cindy Lu <lulu@redhat.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
Message-Id: <20200701145538.22333-15-lulu@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>

vhost-vdpa: introduce vhost-vdpa backend

Currently we have 2 types of vhost backends in QEMU: vhost kernel and
vhost-user. The above patch provides a generic device for vDPA purpose,
this vDPA device exposes to user space a non-vendor-specific configuration
interface for setting up a vhost HW accelerator, this patch set introduces
a third vhost backend called vhost-vdpa based on the vDPA interface.

Vhost-vdpa usage:

qemu-system-x86_64 -cpu host -enable-kvm \
    ......
    -netdev type=vhost-vdpa,vhostdev=/dev/vhost-vdpa-id,id=vhost-vdpa0 \
    -device virtio-net-pci,netdev=vhost-vdpa0,page-per-vq=on \

Signed-off-by: Lingshan zhu <lingshan.zhu@intel.com>
Signed-off-by: Tiwei Bie <tiwei.bie@intel.com>
Signed-off-by: Cindy Lu <lulu@redhat.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
Message-Id: <20200701145538.22333-14-lulu@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>

vhost_net: introduce set_config & get_config

This patch introduces set_config & get_config  method which allows
vhost_net set/get the config to backend

Signed-off-by: Cindy Lu <lulu@redhat.com>
Message-Id: <20200701145538.22333-13-lulu@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>

vhost: implement vhost_force_iommu method

use the vhost_force_iommu callback to force enable feature bit VIRTIO_F_IOMMU_PLATFORM

Signed-off-by: Cindy Lu <lulu@redhat.com>
Message-Id: <20200701145538.22333-12-lulu@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>

vhost: introduce new VhostOps vhost_force_iommu

This patch introduces new VhostOps vhost_force_iommu callback
to force enable features bit VIRTIO_F_IOMMU_PLATFORM.

Signed-off-by: Cindy Lu <lulu@redhat.com>
Message-Id: <20200701145538.22333-11-lulu@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>

vhost: implement vhost_vq_get_addr method

use vhost_vq_get_addr callback to get the vq address from backend

Signed-off-by: Cindy Lu <lulu@redhat.com>
Message-Id: <20200701145538.22333-10-lulu@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>

vhost: introduce new VhostOps vhost_vq_get_addr

This patch introduces new VhostOps vhost_vq_get_addr_op callback to get
the vring addr from the backend

Signed-off-by: Cindy Lu <lulu@redhat.com>
Message-Id: <20200701145538.22333-9-lulu@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>

vhost: implement vhost_dev_start method

use the vhost_dev_start callback to send the status to backend

Signed-off-by: Cindy Lu <lulu@redhat.com>
Message-Id: <20200701145538.22333-8-lulu@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>

vhost: introduce new VhostOps vhost_dev_start

This patch introduces new VhostOps vhost_dev_start callback which allows the
vhost_net set the start/stop status to backend

Signed-off-by: Cindy Lu <lulu@redhat.com>
Message-Id: <20200701145538.22333-7-lulu@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>

vhost: check the existence of vhost_set_iotlb_callback

Add the check of vhost_set_iotlb_callback
before calling

Signed-off-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Cindy Lu <lulu@redhat.com>
Message-Id: <20200701145538.22333-6-lulu@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>

virtio-pci: implement queue_enabled method

With version 1, we can detect whether a queue is enabled via
queue_enabled.

Signed-off-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Cindy Lu <lulu@redhat.com>
Message-Id: <20200701145538.22333-5-lulu@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>

virtio-bus: introduce queue_enabled method

This patch introduces queue_enabled() method which allows the
transport to implement its own way to report whether or not a queue is
enabled.

Signed-off-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Cindy Lu <lulu@redhat.com>
Message-Id: <20200701145538.22333-4-lulu@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>

vhost_net: use the function qemu_get_peer

user the qemu_get_peer to replace the old process

Signed-off-by: Cindy Lu <lulu@redhat.com>
Reviewed-by: Laurent Vivier <lvivier@redhat.com>
Message-Id: <20200701145538.22333-3-lulu@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>

net: introduce qemu_get_peer

This is a small function that can get the peer
from given NetClientState and queue_index

Signed-off-by: Cindy Lu <lulu@redhat.com>
Message-Id: <20200701145538.22333-2-lulu@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>

MAINTAINERS: add VT-d entry

Add this entry as suggested by Jason and Michael.

CC: Jason Wang <jasowang@redhat.com>
CC: Michael S. Tsirkin <mst@redhat.com>
CC: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Peter Xu <peterx@redhat.com>
Message-Id: <20200701124418.63060-1-peterx@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

docs: vhost-user: add Virtio status protocol feature

This patch specifies the VHOST_USER_SET_STATUS and
VHOST_USER_GET_STATUS requests, which are sent by
the master to update and query the Virtio status
in the backend.

Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>
Message-Id: <20200618134501.145747-1-maxime.coquelin@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

tests/acpi: remove stale allowed tables

Fixes: 93dd625f8bf7 ("tests/acpi: update expected data files")
Signed-off-by: Andrew Jones <drjones@redhat.com>
Message-Id: <20200629140938.17566-2-drjones@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

numa: Auto-enable NUMA when any memory devices are possible

Let's auto-enable it also when maxmem is specified but no slots are
defined. This will result in us properly creating ACPI srat tables,
indicating the maximum possible PFN to the guest OS. Based on this, e.g.,
Linux will enable the swiotlb properly.

This avoids having to manually force the switolb on (swiotlb=force) in
Linux in case we're booting only using DMA memory (e.g., 2GB on x86-64),
and virtio-mem adds memory later on that really needs the swiotlb to be
used for DMA.

Let's take care of backwards compatibility if somebody has a setup that
specifies "maxram" without "slots".

Reported-by: Alex Shi <alex.shi@linux.alibaba.com>
Cc: Peter Maydell <peter.maydell@linaro.org>
Cc: Eduardo Habkost <ehabkost@redhat.com>
Cc: Marcel Apfelbaum <marcel.apfelbaum@gmail.com>
Cc: Sergio Lopez <slp@redhat.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Richard Henderson <rth@twiddle.net>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Igor Mammedov <imammedo@redhat.com>
Cc: qemu-arm@nongnu.org <qemu-arm@nongnu.org>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20200626072248.78761-22-david@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

virtio-mem: Exclude unplugged memory during migration

The content of unplugged memory is undefined and should not be migrated,
ever. Exclude all unplugged memory during precopy using the precopy notifier
infrastructure introduced for free page hinting in virtio-balloon.

Unplugged memory is marked as "not dirty", meaning it won't be
considered for migration.

Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: "Dr. David Alan Gilbert" <dgilbert@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20200626072248.78761-21-david@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

virtio-mem: Add trace events

Let's add some trace events that might come in handy later.

Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: "Dr. David Alan Gilbert" <dgilbert@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20200626072248.78761-20-david@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

virtio-mem: Migration sanity checks

We want to make sure that certain properties don't change during
migration, especially to catch user errors in a nice way. Let's migrate
a temporary structure and validate that the properties didn't change.

Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: "Dr. David Alan Gilbert" <dgilbert@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20200626072248.78761-19-david@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

virtio-pci: Send qapi events when the virtio-mem size changes

Let's register the notifier and trigger the qapi event with the right
device id.

MEMORY_DEVICE_SIZE_CHANGE is similar to BALLOON_CHANGE, however on a
memory device level.

Don't unregister the notifier (we neither have finalize() nor unrealize()
for VirtIOPCIProxy, so it's not that simple to do it) - both devices are
expected to vanish at the same time.

Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Markus Armbruster <armbru@redhat.com>
Cc: "Dr. David Alan Gilbert" <dgilbert@redhat.com>
Cc: Eric Blake <eblake@redhat.com>
Cc: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20200626072248.78761-18-david@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

virtio-mem: Allow notifiers for size changes

We want to send qapi events in case the size of a virtio-mem device
changes. This allows upper layers to always know how much memory is
actually currently consumed via a virtio-mem device.

Unfortuantely, we have to report the id of our proxy device. Let's provide
an easy way for our proxy device to register, so it can send the qapi
events. Piggy-backing on the notifier infrastructure (although we'll
only ever have one notifier registered) seems to be an easy way.

Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: "Dr. David Alan Gilbert" <dgilbert@redhat.com>
Cc: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20200626072248.78761-17-david@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

pc: Support for virtio-mem-pci

Let's wire it up similar to virtio-pmem. Also disallow unplug, so it's
harder for users to shoot themselves into the foot.

Reviewed-by: Pankaj Gupta <pankaj.gupta.linux@gmail.com>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Marcel Apfelbaum <marcel.apfelbaum@gmail.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Eduardo Habkost <ehabkost@redhat.com>
Cc: Eric Blake <eblake@redhat.com>
Cc: Markus Armbruster <armbru@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20200626072248.78761-16-david@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

numa: Handle virtio-mem in NUMA stats

Account the memory to the configured nid.

Reviewed-by: Pankaj Gupta <pankaj.gupta.linux@gmail.com>
Cc: Eduardo Habkost <ehabkost@redhat.com>
Cc: Marcel Apfelbaum <marcel.apfelbaum@gmail.com>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20200626072248.78761-15-david@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

hmp: Handle virtio-mem when printing memory device info

Print the memory device info just like for other memory devices.

Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Cc: "Dr. David Alan Gilbert" <dgilbert@redhat.com>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20200626072248.78761-14-david@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

MAINTAINERS: Add myself as virtio-mem maintainer

Let's make sure patches/bug reports find the right person.

Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Peter Maydell <peter.maydell@linaro.org>
Cc: Markus Armbruster <armbru@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20200626072248.78761-13-david@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

virtio-pci: Proxy for virtio-mem

Let's add a proxy for virtio-mem, make it a memory device, and
pass-through the properties.

Reviewed-by: Pankaj Gupta <pankaj.gupta.linux@gmail.com>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Marcel Apfelbaum <marcel.apfelbaum@gmail.com>
Cc: "Dr. David Alan Gilbert" <dgilbert@redhat.com>
Cc: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20200626072248.78761-12-david@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

virtio-mem: Paravirtualized memory hot(un)plug

This is the very basic/initial version of virtio-mem. An introduction to
virtio-mem can be found in the Linux kernel driver [1]. While it can be
used in the current state for hotplug of a smaller amount of memory, it
will heavily benefit from resizeable memory regions in the future.

Each virtio-mem device manages a memory region (provided via a memory
backend). After requested by the hypervisor ("requested-size"), the
guest can try to plug/unplug blocks of memory within that region, in order
to reach the requested size. Initially, and after a reboot, all memory is
unplugged (except in special cases - reboot during postcopy).

The guest may only try to plug/unplug blocks of memory within the usable
region size. The usable region size is a little bigger than the
requested size, to give the device driver some flexibility. The usable
region size will only grow, except on reboots or when all memory is
requested to get unplugged. The guest can never plug more memory than
requested. Unplugged memory will get zapped/discarded, similar to in a
balloon device.

The block size is variable, however, it is always chosen in a way such that
THP splits are avoided (e.g., 2MB). The state of each block
(plugged/unplugged) is tracked in a bitmap.

As virtio-mem devices (e.g., virtio-mem-pci) will be memory devices, we now
expose "VirtioMEMDeviceInfo" via "query-memory-devices".

--------------------------------------------------------------------------

There are two important follow-up items that are in the works:
1. Resizeable memory regions: Use resizeable allocations/RAM blocks to
   grow/shrink along with the usable region size. This avoids creating
   initially very big VMAs, RAM blocks, and KVM slots.
2. Protection of unplugged memory: Make sure the gust cannot actually
   make use of unplugged memory.

Other follow-up items that are in the works:
1. Exclude unplugged memory during migration (via precopy notifier).
2. Handle remapping of memory.
3. Support for other architectures.

--------------------------------------------------------------------------

Example usage (virtio-mem-pci is introduced in follow-up patches):

Start QEMU with two virtio-mem devices (one per NUMA node):
 $ qemu-system-x86_64 -m 4G,maxmem=20G \
  -smp sockets=2,cores=2 \
  -numa node,nodeid=0,cpus=0-1 -numa node,nodeid=1,cpus=2-3 \
  [...]
  -object memory-backend-ram,id=mem0,size=8G \
  -device virtio-mem-pci,id=vm0,memdev=mem0,node=0,requested-size=0M \
  -object memory-backend-ram,id=mem1,size=8G \
  -device virtio-mem-pci,id=vm1,memdev=mem1,node=1,requested-size=1G

Query the configuration:
 (qemu) info memory-devices
 Memory device [virtio-mem]: "vm0"
   memaddr: 0x140000000
   node: 0
   requested-size: 0
   size: 0
   max-size: 8589934592
   block-size: 2097152
   memdev: /objects/mem0
 Memory device [virtio-mem]: "vm1"
   memaddr: 0x340000000
   node: 1
   requested-size: 1073741824
   size: 1073741824
   max-size: 8589934592
   block-size: 2097152
   memdev: /objects/mem1

Add some memory to node 0:
 (qemu) qom-set vm0 requested-size 500M

Remove some memory from node 1:
 (qemu) qom-set vm1 requested-size 200M

Query the configuration again:
 (qemu) info memory-devices
 Memory device [virtio-mem]: "vm0"
   memaddr: 0x140000000
   node: 0
   requested-size: 524288000
   size: 524288000
   max-size: 8589934592
   block-size: 2097152
   memdev: /objects/mem0
 Memory device [virtio-mem]: "vm1"
   memaddr: 0x340000000
   node: 1
   requested-size: 209715200
   size: 209715200
   max-size: 8589934592
   block-size: 2097152
   memdev: /objects/mem1

[1] https://lkml.kernel.org/r/20200311171422.10484-1-david@redhat.com

Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Eric Blake <eblake@redhat.com>
Cc: Markus Armbruster <armbru@redhat.com>
Cc: "Dr. David Alan Gilbert" <dgilbert@redhat.com>
Cc: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20200626072248.78761-11-david@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

migration/colo: Use ram_block_discard_disable()

COLO will copy all memory in a RAM block, disable discarding of RAM.

Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Tested-by: Lukas Straub <lukasstraub2@web.de>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Hailiang Zhang <zhang.zhanghailiang@huawei.com>
Cc: Juan Quintela <quintela@redhat.com>
Cc: "Dr. David Alan Gilbert" <dgilbert@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20200626072248.78761-10-david@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

migration/rdma: Use ram_block_discard_disable()

RDMA will pin all guest memory (as documented in docs/rdma.txt). We want
to disable RAM block discards - however, to keep it simple use
ram_block_discard_is_required() instead of inhibiting.

Note: It is not sufficient to limit disabling to pin_all. Even when only
conditionally pinning 1 MB chunks, as soon as one page within such a
chunk was discarded and one page not, the discarded pages will be pinned
as well.

Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Juan Quintela <quintela@redhat.com>
Cc: "Dr. David Alan Gilbert" <dgilbert@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20200626072248.78761-9-david@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

target/i386: sev: Use ram_block_discard_disable()

AMD SEV will pin all guest memory, mark discarding of RAM broken. At the
time this is called, we cannot have anyone active that relies on discards
to work properly - let's still implement error handling.

Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Eduardo Habkost <ehabkost@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20200626072248.78761-8-david@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

virtio-balloon: Rip out qemu_balloon_inhibit()

The only remaining special case is postcopy. It cannot handle
concurrent discards yet, which would result in requesting already sent
pages from the source. Special-case it in virtio-balloon instead.

Introduce migration_in_incoming_postcopy(), to find out if incoming
postcopy is active.

Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Juan Quintela <quintela@redhat.com>
Cc: "Dr. David Alan Gilbert" <dgilbert@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20200626072248.78761-7-david@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

s390x/pv: Convert to ram_block_discard_disable()

Discarding RAM does not work as expected with protected VMs. Let's
switch to ram_block_discard_disable() for now, as we want to get rid
of qemu_balloon_inhibit(). Note that it will currently never fail, but
might fail in the future with new technologies (e.g., virtio-mem).

Acked-by: Cornelia Huck <cohuck@redhat.com>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Cornelia Huck <cohuck@redhat.com>
Cc: Halil Pasic <pasic@linux.ibm.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: Janosch Frank <frankja@linux.ibm.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20200626072248.78761-6-david@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

accel/kvm: Convert to ram_block_discard_disable()

Discarding memory does not work as expected. At the time this is called,
we cannot have anyone active that relies on discards to work properly.

Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20200626072248.78761-5-david@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

vfio: Convert to ram_block_discard_disable()

VFIO is (except devices without a physical IOMMU or some mediated devices)
incompatible with discarding of RAM. The kernel will pin basically all VM
memory. Let's convert to ram_block_discard_disable(), which can now
fail, in contrast to qemu_balloon_inhibit().

Leave "x-balloon-allowed" named as it is for now.

Reviewed-by: Tony Krowiak <akrowiak@linux.ibm.com>
Acked-by: Cornelia Huck <cohuck@redhat.com>
Cc: Cornelia Huck <cohuck@redhat.com>
Cc: Alex Williamson <alex.williamson@redhat.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: Tony Krowiak <akrowiak@linux.ibm.com>
Cc: Halil Pasic <pasic@linux.ibm.com>
Cc: Pierre Morel <pmorel@linux.ibm.com>
Cc: Eric Farman <farman@linux.ibm.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20200626072248.78761-4-david@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

exec: Introduce ram_block_discard_(disable|require)()

We want to replace qemu_balloon_inhibit() by something more generic.
Especially, we want to make sure that technologies that really rely on
RAM block discards to work reliably to run mutual exclusive with
technologies that effectively break it.

E.g., vfio will usually pin all guest memory, turning the virtio-balloon
basically useless and make the VM consume more memory than reported via
the balloon. While the balloon is special already (=> no guarantees, same
behavior possible afer reboots and with huge pages), this will be
different, especially, with virtio-mem.

Let's implement a way such that we can make both types of technology run
mutually exclusive. We'll convert existing balloon inhibitors in successive
patches and add some new ones. Add the check to
qemu_balloon_is_inhibited() for now. We might want to make
virtio-balloon an acutal inhibitor in the future - however, that
requires more thought to not break existing setups.

Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20200626072248.78761-3-david@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

pc: Support coldplugging of virtio-pmem-pci devices on all buses

E.g., with "pc-q35-4.2", trying to coldplug a virtio-pmem-pci devices
results in
    "virtio-pmem-pci not supported on this bus"

Reasons is, that the bus does not support hotplug and, therefore, does
not have a hotplug handler. Let's allow coldplugging virtio-pmem devices
on such buses. The hotplug order is only relevant for virtio-pmem-pci
when the guest is already alive and the device is visible before
memory_device_plug() wired up the memory device bits.

Hotplug attempts will still fail with:
    "Error: Bus 'pcie.0' does not support hotplugging"

Hotunplug attempts will still fail with:
    "Error: Bus 'pcie.0' does not support hotplugging"

Reported-by: Vivek Goyal <vgoyal@redhat.com>
Reviewed-by: Pankaj Gupta <pankaj.gupta.linux@gmail.com>
Cc: Pankaj Gupta <pankaj.gupta.linux@gmail.com>
Cc: Igor Mammedov <imammedo@redhat.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Eduardo Habkost <ehabkost@redhat.com>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Marcel Apfelbaum <marcel.apfelbaum@gmail.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20200626072248.78761-2-david@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

virtio-balloon: always indicate S_DONE when migration fails

If something goes wrong during precopy, before stopping the VM, we will
never send a S_DONE indication to the VM, resulting in the hinted pages
not getting released to be used by the guest OS (e.g., Linux).

Easy to reproduce:
1. Start migration (e.g., HMP "migrate -d 'exec:gzip -c > STATEFILE.gz'")
2. Cancel migration (e.g., HMP "migrate_cancel")
3. Oberve in the guest (e.g., cat /proc/meminfo) that there is basically
   no free memory left.

While at it, add similar locking to virtio_balloon_free_page_done() as
done in virtio_balloon_free_page_stop. Locking is still weird, but that
has to be sorted out separately.

There is nothing to do in the PRECOPY_NOTIFY_COMPLETE case. Add some
comments regarding S_DONE handling.

Fixes: c13c4153f76d ("virtio-balloon: VIRTIO_BALLOON_F_FREE_PAGE_HINT")
Reviewed-by: Alexander Duyck <alexander.h.duyck@linux.intel.com>
Cc: Wei Wang <wei.w.wang@intel.com>
Cc: Alexander Duyck <alexander.duyck@gmail.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20200629080615.26022-1-david@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

Revert "tests/migration: Reduce autoconverge initial bandwidth"

This reverts commit 6d1da867e65f ("tests/migration: Reduce autoconverge initial bandwidth")
since that change makes unit tests much slower for all developers, while it's not
a robust way to fix migration tests. Migration tests need to find
a more robust way to discover a reasonable bandwidth without slowing
things down for everyone.

Fixes: 6d1da867e65f ("tests/migration: Reduce autoconverge initial bandwidth")
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Acked-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Acked-by: Thomas Huth <thuth@redhat.com>

tests: disassemble-aml.sh: generate AML in readable format

On systems where the IASL tool exists, we can convert
extected ACPI tables to ASL format, which is useful
for debugging and documentation purposes.
This script does this for all ACPI tables under tests/data/acpi/.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

hw/misc/pca9552: Add missing TypeInfo::class_size field

When adding the generic PCA955xClass in commit 736132e455, we
forgot to set the class_size field. Fill it now to avoid:

  (gdb) run -machine mcimx6ul-evk -m 128M -display none -serial stdio -kernel ./OS.elf
  Starting program: ../../qemu/qemu/arm-softmmu/qemu-system-arm -machine mcimx6ul-evk -m 128M -display none -serial stdio -kernel ./OS.elf
  double free or corruption (!prev)
  Thread 1 "qemu-system-arm" received signal SIGABRT, Aborted.
  __GI_raise (sig=sig@entry=6) at ../sysdeps/unix/sysv/linux/raise.c:50
  (gdb) where
  #0  __GI_raise (sig=sig@entry=6) at ../sysdeps/unix/sysv/linux/raise.c:50
  #1  0x00007ffff75d8859 in __GI_abort () at abort.c:79
  #2  0x00007ffff76433ee in __libc_message
      (action=action@entry=do_abort, fmt=fmt@entry=0x7ffff776d285 "%s\n")
      at ../sysdeps/posix/libc_fatal.c:155
  #3  0x00007ffff764b47c in malloc_printerr
      (str=str@entry=0x7ffff776f690 "double free or corruption (!prev)")
      at malloc.c:5347
  #4  0x00007ffff764d12c in _int_free
      (av=0x7ffff779eb80 <main_arena>, p=0x5555567a3990, have_lock=<optimized out>) at malloc.c:4317
  #5  0x0000555555c906c3 in type_initialize_interface
      (ti=ti@entry=0x5555565b8f40, interface_type=0x555556597ad0, parent_type=0x55555662ca10) at qom/object.c:259
  #6  0x0000555555c902da in type_initialize (ti=ti@entry=0x5555565b8f40)
      at qom/object.c:323
  #7  0x0000555555c90d20 in type_initialize (ti=0x5555565b8f40)
      at qom/object.c:1028

  $ valgrind --track-origins=yes qemu-system-arm -M mcimx6ul-evk -m 128M -display none -serial stdio -kernel ./OS.elf
  ==77479== Memcheck, a memory error detector
  ==77479== Copyright (C) 2002-2017, and GNU GPL'd, by Julian Seward et al.
  ==77479== Using Valgrind-3.15.0 and LibVEX; rerun with -h for copyright info
  ==77479== Command: qemu-system-arm -M mcimx6ul-evk -m 128M -display none -serial stdio -kernel ./OS.elf
  ==77479==
  ==77479== Invalid write of size 2
  ==77479==    at 0x6D8322: pca9552_class_init (pca9552.c:424)
  ==77479==    by 0x844D1F: type_initialize (object.c:1029)
  ==77479==    by 0x844D1F: object_class_foreach_tramp (object.c:1016)
  ==77479==    by 0x4AE1057: g_hash_table_foreach (in /usr/lib/x86_64-linux-gnu/libglib-2.0.so.0.6400.2)
  ==77479==    by 0x8453A4: object_class_foreach (object.c:1038)
  ==77479==    by 0x8453A4: object_class_get_list (object.c:1095)
  ==77479==    by 0x556194: select_machine (vl.c:2416)
  ==77479==    by 0x556194: qemu_init (vl.c:3828)
  ==77479==    by 0x40AF9C: main (main.c:48)
  ==77479==  Address 0x583f108 is 0 bytes after a block of size 200 alloc'd
  ==77479==    at 0x483DD99: calloc (in /usr/lib/x86_64-linux-gnu/valgrind/vgpreload_memcheck-amd64-linux.so)
  ==77479==    by 0x4AF8D30: g_malloc0 (in /usr/lib/x86_64-linux-gnu/libglib-2.0.so.0.6400.2)
  ==77479==    by 0x844258: type_initialize.part.0 (object.c:306)
  ==77479==    by 0x844D1F: type_initialize (object.c:1029)
  ==77479==    by 0x844D1F: object_class_foreach_tramp (object.c:1016)
  ==77479==    by 0x4AE1057: g_hash_table_foreach (in /usr/lib/x86_64-linux-gnu/libglib-2.0.so.0.6400.2)
  ==77479==    by 0x8453A4: object_class_foreach (object.c:1038)
  ==77479==    by 0x8453A4: object_class_get_list (object.c:1095)
  ==77479==    by 0x556194: select_machine (vl.c:2416)
  ==77479==    by 0x556194: qemu_init (vl.c:3828)
  ==77479==    by 0x40AF9C: main (main.c:48)

Fixes: 736132e455 ("hw/misc/pca9552: Add generic PCA955xClass")
Reported-by: Jean-Christophe DUBOIS <jcd@tribudubois.net>
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Tested-by: Jean-Christophe DUBOIS <jcd@tribudubois.net>
Message-id: 20200629074704.23028-1-f4bug@amsat.org
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

Merge remote-tracking branch 'remotes/amarkovic/tags/mips-gsoc-queue-jun-27-2020' into staging

MIPS + TCG Continuous Benchmarking queue for June 27th, 2020

# gpg: Signature made Sat 27 Jun 2020 20:30:48 BST
# gpg:                using RSA key D4972A8967F75A65
# gpg: Good signature from "Aleksandar Markovic <amarkovic@wavecomp.com>" [full]
# Primary key fingerprint: 8526 FBF1 5DA3 811F 4A01  DD75 D497 2A89 67F7 5A65

* remotes/amarkovic/tags/mips-gsoc-queue-jun-27-2020:
  MAINTAINERS: Add 'Performance Tools and Tests' subsection
  scripts/performance: Add topN_callgrind.py script
  scripts/performance: Add topN_perf.py script
  MAINTAINERS: Add Loongson-3 maintainer and reviewer
  hw/intc: Add Loongson LIOINTC support
  hw/mips: Implement the kvm_type() hook in MachineClass

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

MAINTAINERS: Add 'Performance Tools and Tests' subsection

This commit creates a new 'Miscellaneous' section which hosts a new
'Performance Tools and Tests' subsection. This subsection will contain
the the performance scripts and benchmarks written as a part of the
'TCG Continuous Benchmarking' project. Also, it will be a placeholder
for follow-ups to this project, if any.

Signed-off-by: Ahmed Karaman <ahmedkhaledkaraman@gmail.com>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Aleksandar Markovic <aleksandar.qemu.devel@gmail.com>
Signed-off-by: Aleksandar Markovic <aleksandar.qemu.devel@gmail.com>
Message-Id: <20200626164546.22102-4-ahmedkhaledkaraman@gmail.com>

scripts/performance: Add topN_callgrind.py script

Python script that prints the top N most executed functions in QEMU
using callgrind.

Syntax:
topN_callgrind.py [-h] [-n] <number of displayed top functions>  -- \
                      <qemu executable> [<qemu executable options>] \
                      <target executable> [<target execurable options>]

[-h] - Print the script arguments help message.
[-n] - Specify the number of top functions to print.
     - If this flag is not specified, the tool defaults to 25.

Example of usage:
topN_callgrind.py -n 20 -- qemu-arm coulomb_double-arm

Example Output:
No.  Percentage Function Name         Source File
----  --------- ------------------    ------------------------------
   1    24.577% 0x00000000082db000    ???
   2    20.467% float64_mul           <qemu>/fpu/softfloat.c
   3    14.720% float64_sub           <qemu>/fpu/softfloat.c
   4    13.864% float64_add           <qemu>/fpu/softfloat.c
   5     4.876% helper_mulsd          <qemu>/target/i386/ops_sse.h
   6     3.767% helper_subsd          <qemu>/target/i386/ops_sse.h
   7     3.549% helper_addsd          <qemu>/target/i386/ops_sse.h
   8     2.185% helper_ucomisd        <qemu>/target/i386/ops_sse.h
   9     1.667% helper_lookup_tb_ptr  <qemu>/include/exec/tb-lookup.h
  10     1.662% f64_compare           <qemu>/fpu/softfloat.c
  11     1.509% helper_lookup_tb_ptr  <qemu>/accel/tcg/tcg-runtime.c
  12     0.635% helper_lookup_tb_ptr  <qemu>/include/exec/exec-all.h
  13     0.616% float64_div           <qemu>/fpu/softfloat.c
  14     0.502% helper_pand_xmm       <qemu>/target/i386/ops_sse.h
  15     0.502% float64_mul           <qemu>/include/fpu/softfloat.h
  16     0.476% helper_lookup_tb_ptr  <qemu>/target/i386/cpu.h
  17     0.437% float64_compare_quiet <qemu>/fpu/softfloat.c
  18     0.414% helper_pxor_xmm       <qemu>/target/i386/ops_sse.h
  19     0.353% round_to_int          <qemu>/fpu/softfloat.c
  20     0.347% helper_cc_compute_all <qemu>/target/i386/cc_helper.c

Signed-off-by: Ahmed Karaman <ahmedkhaledkaraman@gmail.com>
Signed-off-by: Aleksandar Markovic <aleksandar.qemu.devel@gmail.com>
Reviewed-by: Aleksandar Markovic <aleksandar.qemu.devel@gmail.com>
Message-Id: <20200626164546.22102-3-ahmedkhaledkaraman@gmail.com>

scripts/performance: Add topN_perf.py script

Syntax:
topN_perf.py [-h] [-n] <number of displayed top functions>  -- \
                 <qemu executable> [<qemu executable options>] \
                 <target executable> [<target execurable options>]

[-h] - Print the script arguments help message.
[-n] - Specify the number of top functions to print.
     - If this flag is not specified, the tool defaults to 25.

Example of usage:
topN_perf.py -n 20 -- qemu-arm coulomb_double-arm

Example Output:
 No.  Percentage  Name                       Invoked by
----  ----------  -------------------------  -------------------------
   1      16.25%  float64_mul                qemu-x86_64
   2      12.01%  float64_sub                qemu-x86_64
   3      11.99%  float64_add                qemu-x86_64
   4       5.69%  helper_mulsd               qemu-x86_64
   5       4.68%  helper_addsd               qemu-x86_64
   6       4.43%  helper_lookup_tb_ptr       qemu-x86_64
   7       4.28%  helper_subsd               qemu-x86_64
   8       2.71%  f64_compare                qemu-x86_64
   9       2.71%  helper_ucomisd             qemu-x86_64
  10       1.04%  helper_pand_xmm            qemu-x86_64
  11       0.71%  float64_div                qemu-x86_64
  12       0.63%  helper_pxor_xmm            qemu-x86_64
  13       0.50%  0x00007f7b7004ef95         [JIT] tid 491
  14       0.50%  0x00007f7b70044e83         [JIT] tid 491
  15       0.36%  helper_por_xmm             qemu-x86_64
  16       0.32%  helper_cc_compute_all      qemu-x86_64
  17       0.30%  0x00007f7b700433f0         [JIT] tid 491
  18       0.30%  float64_compare_quiet      qemu-x86_64
  19       0.27%  soft_f64_addsub            qemu-x86_64
  20       0.26%  round_to_int               qemu-x86_64

Signed-off-by: Ahmed Karaman <ahmedkhaledkaraman@gmail.com>
Signed-off-by: Aleksandar Markovic <aleksandar.qemu.devel@gmail.com>
Reviewed-by: Aleksandar Markovic <aleksandar.qemu.devel@gmail.com>
Message-Id: <20200626164546.22102-2-ahmedkhaledkaraman@gmail.com>

MAINTAINERS: Add Loongson-3 maintainer and reviewer

Add myself as the maintainer for Loongson-3 virtual platforms, and
also add Jiaxun Yang as the reviewer.

Signed-off-by: Huacai Chen <chenhc@lemote.com>
Co-developed-by: Jiaxun Yang <jiaxun.yang@flygoat.com>
Reviewed-by: Aleksandar Markovic <aleksandar.qemu.devel@gmail.com>
Signed-off-by: Aleksandar Markovic <aleksandar.qemu.devel@gmail.com>
Message-Id: <1592995531-32600-5-git-send-email-chenhc@lemote.com>

hw/intc: Add Loongson LIOINTC support

Loongson-3 has an integrated liointc (Local I/O Interrupt Controller).
It is similar to Goldfish interrupt controller, but more powerful (e.g.,
it can route external interrupt to multi-cores).

Documents about Loongson-3's liointc:
1, https://wiki.godson.ac.cn/ip_block:liointc;
2, The "I/O中断" section of Loongson-3's user mannual, part 1.

Signed-off-by: Huacai Chen <chenhc@lemote.com>
Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com>
Signed-off-by: Aleksandar Markovic <aleksandar.qemu.devel@gmail.com>
Reviewed-by: Aleksandar Markovic <aleksandar.qemu.devel@gmail.com>
Message-Id: <1592995531-32600-3-git-send-email-chenhc@lemote.com>

hw/mips: Implement the kvm_type() hook in MachineClass

MIPS has two types of KVM: TE & VZ, and TE is the default type. Now we
can't create a VZ guest in QEMU because it lacks the kvm_type() hook in
MachineClass. This patch add the the kvm_type() hook to support both of
the two types.

[AM: Added "if defined" guards.]

Reviewed-by: Aleksandar Markovic <aleksandar.qemu.devel@gmail.com>
Signed-off-by: Aleksandar Markovic <aleksandar.qemu.devel@gmail.com>
Signed-off-by: Huacai Chen <chenhc@lemote.com>
Co-developed-by: Jiaxun Yang <jiaxun.yang@flygoat.com>
Message-Id: <1592995531-32600-2-git-send-email-chenhc@lemote.com>

Merge remote-tracking branch 'remotes/pmaydell/tags/pull-target-arm-20200626' into staging

target-arm queue:
 * hw/arm/aspeed: improve QOM usage
 * hw/misc/pca9552: trace GPIO change events
 * target/arm: Implement ARMv8.5-MemTag for system emulation

# gpg: Signature made Fri 26 Jun 2020 16:13:27 BST
# gpg:                using RSA key E1A5C593CD419DE28E8315CF3C2525ED14360CDE
# gpg:                issuer "peter.maydell@linaro.org"
# gpg: Good signature from "Peter Maydell <peter.maydell@linaro.org>" [ultimate]
# gpg:                 aka "Peter Maydell <pmaydell@gmail.com>" [ultimate]
# gpg:                 aka "Peter Maydell <pmaydell@chiark.greenend.org.uk>" [ultimate]
# Primary key fingerprint: E1A5 C593 CD41 9DE2 8E83  15CF 3C25 25ED 1436 0CDE

* remotes/pmaydell/tags/pull-target-arm-20200626: (57 commits)
  target/arm: Enable MTE
  target/arm: Add allocation tag storage for system mode
  target/arm: Create tagged ram when MTE is enabled
  target/arm: Cache the Tagged bit for a page in MemTxAttrs
  target/arm: Always pass cacheattr to get_phys_addr
  target/arm: Set PSTATE.TCO on exception entry
  target/arm: Implement data cache set allocation tags
  target/arm: Complete TBI clearing for user-only for SVE
  target/arm: Add mte helpers for sve scatter/gather memory ops
  target/arm: Handle TBI for sve scalar + int memory ops
  target/arm: Add mte helpers for sve scalar + int ff/nf loads
  target/arm: Add mte helpers for sve scalar + int stores
  target/arm: Add mte helpers for sve scalar + int loads
  target/arm: Add arm_tlb_bti_gp
  target/arm: Tidy trans_LD1R_zpri
  target/arm: Use mte_check1 for sve LD1R
  target/arm: Use mte_checkN for sve unpredicated stores
  target/arm: Use mte_checkN for sve unpredicated loads
  target/arm: Add helper_mte_check_zva
  target/arm: Implement helper_mte_checkN
  ...

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

Merge remote-tracking branch 'remotes/bonzini/tags/for-upstream' into staging

* Various fixes
* libdaxctl support to correctly align devdax character devices (Jingqi)
* initial-all-set support for live migration (Jay)
* forbid '-numa node, mem' for 5.1 and newer machine types (Igor)
* x87 fixes (Joseph)
* Tighten memory_region_access_valid (Michael) and fix fallout (myself)
* Replay fixes (Pavel)

# gpg: Signature made Fri 26 Jun 2020 14:42:17 BST
# gpg:                using RSA key F13338574B662389866C7682BFFBD25F78C7AE83
# gpg:                issuer "pbonzini@redhat.com"
# gpg: Good signature from "Paolo Bonzini <bonzini@gnu.org>" [full]
# gpg:                 aka "Paolo Bonzini <pbonzini@redhat.com>" [full]
# Primary key fingerprint: 46F5 9FBD 57D6 12E7 BFD4  E2F7 7E15 100C CD36 69B1
#      Subkey fingerprint: F133 3857 4B66 2389 866C  7682 BFFB D25F 78C7 AE83

* remotes/bonzini/tags/for-upstream: (31 commits)
  i386: Mask SVM features if nested SVM is disabled
  ibex_uart: fix XOR-as-pow
  vmport: move compat properties to hw_compat_5_0
  hyperv: vmbus: Remove the 2nd IRQ
  kvm: i386: allow TSC to differ by NTP correction bounds without TSC scaling
  numa: forbid '-numa node, mem' for 5.1 and newer machine types
  osdep: Make MIN/MAX evaluate arguments only once
  target/i386: Add notes for versioned CPU models
  target/i386: reimplement fpatan using floatx80 operations
  target/i386: reimplement fyl2x using floatx80 operations
  target/i386: reimplement fyl2xp1 using floatx80 operations
  target/i386: reimplement fprem, fprem1 using floatx80 operations
  softfloat: return low bits of quotient from floatx80_modrem
  softfloat: do not set denominator high bit for floatx80 remainder
  softfloat: do not return pseudo-denormal from floatx80 remainder
  softfloat: fix floatx80 remainder pseudo-denormal check for zero
  softfloat: merge floatx80_mod and floatx80_rem
  target/i386: reimplement f2xm1 using floatx80 operations
  xen: Actually fix build without passthrough
  Makefile: Install qemu-[qmp/ga]-ref.* into the directory "interop"
  ...

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

i386: Mask SVM features if nested SVM is disabled

QEMU incorrectly validates FEAT_SVM feature flags against
GET_SUPPORTED_CPUID even if SVM features are being masked out by
cpu_x86_cpuid().  This can make QEMU print warnings on most AMD
CPU models, even when SVM nesting is disabled (which is the
default).

This bug was never detected before because of a Linux KVM bug:
until Linux v5.6, KVM was not filtering out SVM features in
GET_SUPPORTED_CPUID when nested was disabled.  This KVM bug was
fixed in Linux v5.7-rc1, on Linux commit a50718cc3f43 ("KVM:
nSVM: Expose SVM features to L1 iff nested is enabled").

Fix the problem by adding a CPUID_EXT3_SVM dependency to all
FEAT_SVM feature flags in the feature_dependencies table.

Reported-by: Yanan Fu <yfu@redhat.com>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Message-Id: <20200623230116.277409-1-ehabkost@redhat.com>
[Fix testcase. - Paolo]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

ibex_uart: fix XOR-as-pow

The xor-as-pow warning in clang actually detected a genuine bug.
Fix it.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

vmport: move compat properties to hw_compat_5_0

The patches that introduced the properties were submitted when QEMU 5.0
had not been released yet, so they got merged under the wrong heading.
Move them to hw_compat_5_0 so that 5.0 machine types get the pre-patch
behavior.

Fixes: b889212973da ("hw/i386/vmport: Propagate IOPort read to vCPU EAX register")
Fixes: 0342ee761ef2 ("hw/i386/vmport: Set EAX to -1 on failed and unsupported commands")
Fixes: f8bdc550370f ("hw/i386/vmport: Report vmware-vmx-type in CMD_GETVERSION")
Fixes: aaacf1c15a22 ("hw/i386/vmport: Add support for CMD_GETBIOSUUID")
Reported-by: Laurent Vivier <lvivier@redhat.com>
Cc: Liran Alon <liran.alon@oracle.com>
Reviewed-by: Laurent Vivier <lvivier@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

hyperv: vmbus: Remove the 2nd IRQ

It seems like Windows does not really require 2 IRQs to have a
functioning VMBus.

Signed-off-by: Jon Doron <arilou@gmail.com>
Message-Id: <20200617160904.681845-2-arilou@gmail.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

kvm: i386: allow TSC to differ by NTP correction bounds without TSC scaling

The Linux TSC calibration procedure is subject to small variations
(its common to see +-1 kHz difference between reboots on a given CPU, for example).

So migrating a guest between two hosts with identical processor can fail, in case
of a small variation in calibrated TSC between them.

Allow a conservative 250ppm error between host TSC and VM TSC frequencies,
rather than requiring an exact match. NTP daemon in the guest can
correct this difference.

Also change migration to accept this bound.

KVM_SET_TSC_KHZ depends on a kernel interface change. Without this change,
the behaviour remains the same: in case of a different frequency
between host and VM, KVM_SET_TSC_KHZ will fail and QEMU will exit.

Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Message-Id: <20200616165805.GA324612@fuller.cnet>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

numa: forbid '-numa node, mem' for 5.1 and newer machine types

Deprecation period is run out and it's a time to flip the switch
introduced by cd5ff8333a.  Disable legacy option for new machine
types (since 5.1) and amend documentation.

'-numa node,memdev' shall be used instead of disabled option
with new machine types.

Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Greg Kurz <groug@kaod.org>
Message-Id: <20200609135635.761587-1-imammedo@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

osdep: Make MIN/MAX evaluate arguments only once

I'm not aware of any immediate bugs in qemu where a second runtime
evaluation of the arguments to MIN() or MAX() causes a problem, but
proactively preventing such abuse is easier than falling prey to an
unintended case down the road.  At any rate, here's the conversation
that sparked the current patch:
https://lists.gnu.org/archive/html/qemu-devel/2018-12/msg05718.html

Update the MIN/MAX macros to only evaluate their argument once at
runtime; this uses typeof(1 ? (a) : (b)) to ensure that we are
promoting the temporaries to the same type as the final comparison (we
have to trigger type promotion, as typeof(bitfield) won't compile; and
we can't use typeof((a) + (b)) or even typeof((a) + 0), as some of our
uses of MAX are on void* pointers where such addition is undefined).

However, we are unable to work around gcc refusing to compile ({}) in
a constant context (such as the array length of a static variable),
even when only used in the dead branch of a __builtin_choose_expr(),
so we have to provide a second macro pair MIN_CONST and MAX_CONST for
use when both arguments are known to be compile-time constants and
where the result must also be usable as a constant; this second form
evaluates arguments multiple times but that doesn't matter for
constants.  By using a void expression as the expansion if a
non-constant is presented to this second form, we can enlist the
compiler to ensure the double evaluation is not attempted on
non-constants.

Alas, as both macros now rely on compiler intrinsics, they are no
longer usable in preprocessor #if conditions; those will just have to
be open-coded or the logic rewritten into #define or runtime 'if'
conditions (but where the compiler dead-code-elimination will probably
still apply).

I tested that both gcc 10.1.1 and clang 10.0.0 produce errors for all
forms of macro mis-use.  As the errors can sometimes be cryptic, I'm
demonstrating the gcc output:

Use of MIN when MIN_CONST is needed:

In file included from /home/eblake/qemu/qemu-img.c:25:
/home/eblake/qemu/include/qemu/osdep.h:249:5: error: braced-group within expression allowed only inside a function
  249 |     ({                                                  \
      |     ^
/home/eblake/qemu/qemu-img.c:92:12: note: in expansion of macro ‘MIN’
   92 | char array[MIN(1, 2)] = "";
      |            ^~~

Use of MIN_CONST when MIN is needed:

/home/eblake/qemu/qemu-img.c: In function ‘is_allocated_sectors’:
/home/eblake/qemu/qemu-img.c:1225:15: error: void value not ignored as it ought to be
 1225 |             i = MIN_CONST(i, n);
      |               ^

Use of MIN in the preprocessor:

In file included from /home/eblake/qemu/accel/tcg/translate-all.c:20:
/home/eblake/qemu/accel/tcg/translate-all.c: In function ‘page_check_range’:
/home/eblake/qemu/include/qemu/osdep.h:249:6: error: token "{" is not valid in preprocessor expressions
  249 |     ({                                                  \
      |      ^

Fix the resulting callsites that used #if or computed a compile-time
constant min or max to use the new macros.  cpu-defs.h is interesting,
as CPU_TLB_DYN_MAX_BITS is sometimes used as a constant and sometimes
dynamic.

It may be worth improving glib's MIN/MAX definitions to be saner, but
that is a task for another day.

Signed-off-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Tested-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20200625162602.700741-1-eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: Add notes for versioned CPU models

Add which features are added or removed in this version.

Signed-off-by: Tao Xu <tao3.xu@intel.com>
Message-Id: <20200324051034.30541-1-tao3.xu@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: reimplement fpatan using floatx80 operations

The x87 fpatan emulation is currently based around conversion to
double.  This is inherently unsuitable for a good emulation of any
floatx80 operation.  Reimplement using the soft-float operations, as
for other such instructions.

Signed-off-by: Joseph Myers <joseph@codesourcery.com>
Message-Id: <alpine.DEB.2.21.2006230000340.24721@digraph.polyomino.org.uk>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: reimplement fyl2x using floatx80 operations

The x87 fyl2x emulation is currently based around conversion to
double.  This is inherently unsuitable for a good emulation of any
floatx80 operation.  Reimplement using the soft-float operations,
building on top of the reimplementation of fyl2xp1 and factoring out
code to be shared between the two instructions.

The included test assumes that the result in round-to-nearest mode
should always be one of the two closest floating-point numbers to the
mathematically exact result (including that it should be exact, in the
exact cases which cover more cases than for fyl2xp1).

Signed-off-by: Joseph Myers <joseph@codesourcery.com>
Message-Id: <alpine.DEB.2.21.2006172321530.20587@digraph.polyomino.org.uk>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: reimplement fyl2xp1 using floatx80 operations

The x87 fyl2xp1 emulation is currently based around conversion to
double.  This is inherently unsuitable for a good emulation of any
floatx80 operation, even before considering that it is a particularly
naive implementation using double (adding 1 then using log rather than
attempting a better emulation using log1p).

Reimplement using the soft-float operations, as was done for f2xm1; as
in that case, m68k has related operations but not exactly this one and
it seemed safest to implement directly rather than reusing the m68k
code to avoid accumulation of errors.

A test is included with many randomly generated inputs.  The
assumption of the test is that the result in round-to-nearest mode
should always be one of the two closest floating-point numbers to the
mathematical value of y * log2(x + 1); the implementation aims to do
somewhat better than that (about 70 correct bits before rounding).  I
haven't investigated how accurate hardware is.

Intel manuals describe a narrower range of valid arguments to this
instruction than AMD manuals.  The implementation accepts the wider
range (it's needed anyway for the core code to be reusable in a
subsequent patch reimplementing fyl2x), but the test only has inputs
in the narrower range so that it's valid on hardware that may reject
or produce poor results for inputs outside that range.

Code in the previous implementation that sets C2 for some out-of-range
arguments is not carried forward to the new implementation; C2 is
undefined for this instruction and I suspect that code was just
cut-and-pasted from the trigonometric instructions (fcos, fptan, fsin,
fsincos) where C2 *is* defined to be set for out-of-range arguments.

Signed-off-by: Joseph Myers <joseph@codesourcery.com>
Message-Id: <alpine.DEB.2.21.2006172320190.20587@digraph.polyomino.org.uk>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: reimplement fprem, fprem1 using floatx80 operations

The x87 fprem and fprem1 emulation is currently based around
conversion to double, which is inherently unsuitable for a good
emulation of any floatx80 operation.  Reimplement using the soft-float
floatx80 remainder operations.

Signed-off-by: Joseph Myers <joseph@codesourcery.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <alpine.DEB.2.21.2006081657200.23637@digraph.polyomino.org.uk>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

softfloat: return low bits of quotient from floatx80_modrem

Both x87 and m68k need the low parts of the quotient for their
remainder operations.  Arrange for floatx80_modrem to track those bits
and return them via a pointer.

The architectures using float32_rem and float64_rem do not appear to
need this information, so the *_rem interface is left unchanged and
the information returned only from floatx80_modrem.  The logic used to
determine the low 7 bits of the quotient for m68k
(target/m68k/fpu_helper.c:make_quotient) appears completely bogus (it
looks at the result of converting the remainder to integer, the
quotient having been discarded by that point); this patch does not
change that, but the m68k maintainers may wish to do so.

Signed-off-by: Joseph Myers <joseph@codesourcery.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <alpine.DEB.2.21.2006081656500.23637@digraph.polyomino.org.uk>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

softfloat: do not set denominator high bit for floatx80 remainder

The floatx80 remainder implementation unnecessarily sets the high bit
of bSig explicitly.  By that point in the function, arguments that are
invalid, zero, infinity or NaN have already been handled and
subnormals have been through normalizeFloatx80Subnormal, so the high
bit will already be set.  Remove the unnecessary code.

Signed-off-by: Joseph Myers <joseph@codesourcery.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <alpine.DEB.2.21.2006081656220.23637@digraph.polyomino.org.uk>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

softfloat: do not return pseudo-denormal from floatx80 remainder

The floatx80 remainder implementation sometimes returns the numerator
unchanged when the denominator is sufficiently larger than the
numerator.  But if the value to be returned unchanged is a
pseudo-denormal, that is incorrect.  Fix it to normalize the numerator
in that case.

Signed-off-by: Joseph Myers <joseph@codesourcery.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <alpine.DEB.2.21.2006081655520.23637@digraph.polyomino.org.uk>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

softfloat: fix floatx80 remainder pseudo-denormal check for zero

The floatx80 remainder implementation ignores the high bit of the
significand when checking whether an operand (numerator) with zero
exponent is zero.  This means it mishandles a pseudo-denormal
representation of 0x1p-16382L by treating it as zero.  Fix this by
checking the whole significand instead.

Signed-off-by: Joseph Myers <joseph@codesourcery.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <alpine.DEB.2.21.2006081655180.23637@digraph.polyomino.org.uk>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

softfloat: merge floatx80_mod and floatx80_rem

The m68k-specific softfloat code includes a function floatx80_mod that
is extremely similar to floatx80_rem, but computing the remainder
based on truncating the quotient toward zero rather than rounding it
to nearest integer.  This is also useful for emulating the x87 fprem
and fprem1 instructions.  Change the floatx80_rem implementation into
floatx80_modrem that can perform either operation, with both
floatx80_rem and floatx80_mod as thin wrappers available for all
targets.

There does not appear to be any use for the _mod operation for other
floating-point formats in QEMU (the only other architectures using
_rem at all are linux-user/arm/nwfpe, for FPA emulation, and openrisc,
for instructions that have been removed in the latest version of the
architecture), so no change is made to the code for other formats.

Signed-off-by: Joseph Myers <joseph@codesourcery.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <alpine.DEB.2.21.2006081654280.23637@digraph.polyomino.org.uk>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: reimplement f2xm1 using floatx80 operations

The x87 f2xm1 emulation is currently based around conversion to
double.  This is inherently unsuitable for a good emulation of any
floatx80 operation, even before considering that it is a particularly
naive implementation using double (computing with pow and then
subtracting 1 rather than attempting a better emulation using expm1).

Reimplement using the soft-float operations, including additions and
multiplications with higher precision where appropriate to limit
accumulation of errors.  I considered reusing some of the m68k code
for transcendental operations, but the instructions don't generally
correspond exactly to x87 operations (for example, m68k has 2^x and
e^x - 1, but not 2^x - 1); to avoid possible accumulation of errors
from applying multiple such operations each rounding to floatx80
precision, I wrote a direct implementation of 2^x - 1 instead.  It
would be possible in principle to make the implementation more
efficient by doing the intermediate operations directly with
significands, signs and exponents and not packing / unpacking floatx80
format for each operation, but that would make it significantly more
complicated and it's not clear that's worthwhile; the m68k emulation
doesn't try to do that.

A test is included with many randomly generated inputs.  The
assumption of the test is that the result in round-to-nearest mode
should always be one of the two closest floating-point numbers to the
mathematical value of 2^x - 1; the implementation aims to do somewhat
better than that (about 70 correct bits before rounding).  I haven't
investigated how accurate hardware is.

Signed-off-by: Joseph Myers <joseph@codesourcery.com>
Message-Id: <alpine.DEB.2.21.2006112341010.18393@digraph.polyomino.org.uk>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

xen: Actually fix build without passthrough

Fix typo.

Fixes: acd0c9416d48 ("xen: fix build without pci passthrough")
Signed-off-by: Anthony PERARD <anthony.perard@citrix.com>
Message-Id: <20200619103115.254127-1-anthony.perard@citrix.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

Makefile: Install qemu-[qmp/ga]-ref.* into the directory "interop"

We need install qemu-[qmp/ga]-ref.* files into the subdirectory of qemu docs: interop.

If we visit the following address and click the link to qemu-qmp-ref.html:
https://www.qemu.org/docs/master/interop/bitmaps.html#basic-qmp-usage

It will report following error:
"
Not Found
The requested URL /docs/master/interop/qemu-qmp-ref.html was not found on this server.
"

Signed-off-by: Liao Pingfang <liao.pingfang@zte.com.cn>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <1591663670-47712-1-git-send-email-wang.yi59@zte.com.cn>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

hw/scsi/megasas: Fix possible out-of-bounds array access in tracepoints

Some tracepoints in megasas.c use a guest-controlled value as an index
into the mfi_frame_desc[] array. Thus a malicious guest could cause an
out-of-bounds error here. Fortunately, the impact is very low since this
can only happen when the corresponding tracepoints have been enabled
before, but the problem should be fixed anyway with a proper check.

Buglink: https://bugs.launchpad.net/qemu/+bug/1882065
Signed-off-by: Thomas Huth <thuth@redhat.com>
Message-Id: <20200615072629.32321-1-thuth@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

docs/nvdimm: add description of alignment requirement of device dax

For device dax (e.g., /dev/dax0.0), the NUM of 'align=NUM' option
needs to match the alignment requirement of the device dax.
It must be larger than or equal to the 'align' of device dax.

Reviewed-by: Joao Martins <joao.m.martins@oracle.com>
Signed-off-by: Jingqi Liu <jingqi.liu@intel.com>
Message-Id: <20200429085011.63752-3-jingqi.liu@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/arm: Enable MTE

We now implement all of the components of MTE, without actually
supporting any tagged memory.  All MTE instructions will work,
trivially, so we can enable support.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20200626033144.790098-46-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Add allocation tag storage for system mode

Look up the physical address for the given virtual address,
convert that to a tag physical address, and finally return
the host address that backs it.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20200626033144.790098-45-richard.henderson@linaro.org
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Create tagged ram when MTE is enabled

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 20200626033144.790098-44-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Cache the Tagged bit for a page in MemTxAttrs

This "bit" is a particular value of the page's MemAttr.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20200626033144.790098-43-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Always pass cacheattr to get_phys_addr

We need to check the memattr of a page in order to determine
whether it is Tagged for MTE.  Between Stage1 and Stage2,
this becomes simpler if we always collect this data, instead
of occasionally being presented with NULL.

Use the nonnull attribute to allow the compiler to check that
all pointer arguments are non-null.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20200626033144.790098-42-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Set PSTATE.TCO on exception entry

D1.10 specifies that exception handlers begin with tag checks overridden.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20200626033144.790098-41-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Implement data cache set allocation tags

This is DC GVA and DC GZVA, and the tag check for DC ZVA.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20200626033144.790098-40-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Complete TBI clearing for user-only for SVE

There are a number of paths by which the TBI is still intact
for user-only in the SVE helpers.

Because we currently always set TBI for user-only, we do not
need to pass down the actual TBI setting from above, and we
can remove the top byte in the inner-most primitives, so that
none are forgotten.  Moreover, this keeps the "dirty" pointer
around at the higher levels, where we need it for any MTE checking.

Since the normal case, especially for user-only, goes through
RAM, this clearing merely adds two insns per page lookup, which
will be completely in the noise.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20200626033144.790098-39-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Add mte helpers for sve scatter/gather memory ops

Because the elements are non-sequential, we cannot eliminate many
tests straight away like we can for sequential operations.  But
we often have the PTE details handy, so we can test for Tagged.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20200626033144.790098-38-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Handle TBI for sve scalar + int memory ops

We still need to handle tbi for user-only when mte is inactive.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20200626033144.790098-37-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Add mte helpers for sve scalar + int ff/nf loads

Because the elements are sequential, we can eliminate many tests all
at once when the tag hits TCMA, or if the page(s) are not Tagged.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20200626033144.790098-36-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Add mte helpers for sve scalar + int stores

Because the elements are sequential, we can eliminate many tests all
at once when the tag hits TCMA, or if the page(s) are not Tagged.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20200626033144.790098-35-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Add mte helpers for sve scalar + int loads

Because the elements are sequential, we can eliminate many tests all
at once when the tag hits TCMA, or if the page(s) are not Tagged.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20200626033144.790098-34-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Add arm_tlb_bti_gp

Introduce an lvalue macro to wrap target_tlb_bit0.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20200626033144.790098-33-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Tidy trans_LD1R_zpri

Move the variable declarations to the top of the function,
but do not create a new label before sve_access_check.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20200626033144.790098-32-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Use mte_check1 for sve LD1R

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20200626033144.790098-31-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Use mte_checkN for sve unpredicated stores

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20200626033144.790098-30-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Use mte_checkN for sve unpredicated loads

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20200626033144.790098-29-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Add helper_mte_check_zva

Use a special helper for DC_ZVA, rather than the more
general mte_checkN.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20200626033144.790098-28-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Implement helper_mte_checkN

Fill out the stub that was added earlier.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20200626033144.790098-27-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Implement helper_mte_check1

Fill out the stub that was added earlier.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20200626033144.790098-26-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Add gen_mte_checkN

Replace existing uses of check_data_tbi in translate-a64.c that
perform multiple logical memory access.  Leave the helper blank
for now to reduce the patch size.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20200626033144.790098-25-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Add gen_mte_check1

Replace existing uses of check_data_tbi in translate-a64.c that
perform a single logical memory access.  Leave the helper blank
for now to reduce the patch size.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20200626033144.790098-24-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Move regime_tcr to internals.h

We will shortly need this in mte_helper.c as well.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20200626033144.790098-23-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Move regime_el to internals.h

We will shortly need this in mte_helper.c as well.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20200626033144.790098-22-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Implement the access tag cache flushes

Like the regular data cache flushes, these are nops within qemu.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20200626033144.790098-21-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Implement the LDGM, STGM, STZGM instructions

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20200626033144.790098-20-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>