roms/openbios: update OpenBIOS images to af97fd7a built from submodule

Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>

gdbstub: move update guest debug to accel ops

Continuing the refactor of a48e7d9e52 (gdbstub: move guest debug support
check to ops) by removing hardcoded kvm_enabled() from generic cpu.c
code, and replace it with a property of AccelOpsClass.

Signed-off-by: Mads Ynddal <m.ynddal@samsung.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-Id: <20230207131721.49233-1-mads@ynddal.dk>
[AJB: add ifdef around update_guest_debug_ops, fix brace]
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20230302190846.2593720-27-alex.bennee@linaro.org>
Message-Id: <20230303025805.625589-30-richard.henderson@linaro.org>

gdbstub: Build syscall.c once

There is no longer anything target specific.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20230303025805.625589-29-richard.henderson@linaro.org>

stubs: split semihosting_get_target from system only stubs

As we are about to have a common syscalls.c for gdbstub we need to
stub out one semihosting helper function for all targets.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20230307170405.796822-1-alex.bennee@linaro.org>

gdbstub: Adjust gdb_do_syscall to only use uint32_t and uint64_t

Pass %x as uint32_t and %lx as uint64_t; pass the address
of %s as uint64_t and the length as uint32_t.

Add casts in semihosting/syscalls.c from target_ulong to
uint64_t; add casts from int to uint32_t for clarity.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20230303025805.625589-28-richard.henderson@linaro.org>

gdbstub: Remove gdb_do_syscallv

This function is unused, except to implement gdb_do_syscall.
Fold the implementations together.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20230303025805.625589-27-richard.henderson@linaro.org>

gdbstub: split out softmmu/user specifics for syscall handling

Most of the syscall code is config agnostic aside from the size of
target_ulong. In preparation for the next patch move the final bits
of specialisation into the appropriate user and softmmu helpers.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20230302190846.2593720-26-alex.bennee@linaro.org>
Message-Id: <20230303025805.625589-26-richard.henderson@linaro.org>

include: split target_long definition from cpu-defs

While we will continue to include this via cpu-defs it is useful to be
able to define this separately for 32 and 64 bit versions of an
otherwise target independent compilation unit.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20230302190846.2593720-25-alex.bennee@linaro.org>
Message-Id: <20230303025805.625589-25-richard.henderson@linaro.org>

testing: probe gdb for supported architectures ahead of time

Currently when we encounter a gdb that is old or not built with
multiarch in mind we fail rather messily. Try and improve the
situation by probing ahead of time and setting
HOST_GDB_SUPPORTS_ARCH=y in the relevant tcg configs. We can then skip
and give a more meaningful message if we don't run the test.

Cc: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20230302190846.2593720-24-alex.bennee@linaro.org>
Message-Id: <20230303025805.625589-24-richard.henderson@linaro.org>

gdbstub: only compile gdbstub twice for whole build

Now we have removed any target specific bits from the core gdbstub
code we only need to build it twice. We have to jump a few meson hoops
to manually define the CONFIG_USER_ONLY symbol but it seems to work.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20230302190846.2593720-23-alex.bennee@linaro.org>
Message-Id: <20230303025805.625589-23-richard.henderson@linaro.org>

gdbstub: move syscall handling to new file

Our GDB syscall support is the last chunk of code that needs target
specific support so move it to a new file. We take the opportunity to
move the syscall state into its own singleton instance and add in a
few helpers for the main gdbstub to interact with the module.

I also moved the gdb_exit() declaration into syscalls.h as it feels
pretty related and most of the callers of it treat it as such.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20230302190846.2593720-22-alex.bennee@linaro.org>
Message-Id: <20230303025805.625589-22-richard.henderson@linaro.org>

gdbstub: move register helpers into standalone include

These inline helpers are all used by target specific code so move them
out of the general header so we don't needlessly pollute the rest of
the API with target specific stuff.

Note we have to include cpu.h in semihosting as it was relying on a
side effect before.

Reviewed-by: Taylor Simpson <tsimpson@quicinc.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20230302190846.2593720-21-alex.bennee@linaro.org>
Message-Id: <20230303025805.625589-21-richard.henderson@linaro.org>

gdbstub: don't use target_ulong while handling registers

This is a hangover from the original code. addr is misleading as it is
only really a register id. While len will never exceed
MAX_PACKET_LENGTH I've used size_t as that is what strlen returns.

Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20230302190846.2593720-20-alex.bennee@linaro.org>
Message-Id: <20230303025805.625589-20-richard.henderson@linaro.org>

gdbstub: fix address type of gdb_set_cpu_pc

The underlying call uses vaddr and the comms API uses unsigned long
long which will always fit. We don't need to deal in target_ulong
here.

Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20230302190846.2593720-19-alex.bennee@linaro.org>
Message-Id: <20230303025805.625589-19-richard.henderson@linaro.org>

gdbstub: specialise stub_can_reverse

Currently we only support replay for softmmu mode so it is a constant
false for user-mode.

Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20230302190846.2593720-18-alex.bennee@linaro.org>
Message-Id: <20230303025805.625589-18-richard.henderson@linaro.org>

gdbstub: introduce gdb_get_max_cpus

This is needed for handling vcont packets as the way of calculating
max cpus vhanges between user and softmmu mode.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20230302190846.2593720-17-alex.bennee@linaro.org>
Message-Id: <20230303025805.625589-17-richard.henderson@linaro.org>

gdbstub: specialise target_memory_rw_debug

The two implementations are different enough to encourage having a
specialisation and we can move some of the softmmu only stuff out of
gdbstub.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20230302190846.2593720-16-alex.bennee@linaro.org>
Message-Id: <20230303025805.625589-16-richard.henderson@linaro.org>

gdbstub: specialise handle_query_attached

In both user and softmmu cases we are just replying with a constant.

Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20230302190846.2593720-15-alex.bennee@linaro.org>
Message-Id: <20230303025805.625589-15-richard.henderson@linaro.org>

gdbstub: abstract target specific details from gdb_put_packet_binary

We unfortunately handle the checking of packet acknowledgement
differently for user and softmmu modes. Abstract the user mode stuff
behind gdb_got_immediate_ack with a stub for softmmu.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20230302190846.2593720-14-alex.bennee@linaro.org>
Message-Id: <20230303025805.625589-14-richard.henderson@linaro.org>

gdbstub: rationalise signal mapping in softmmu

We don't really need a table for mapping two symbols.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20230302190846.2593720-13-alex.bennee@linaro.org>
Message-Id: <20230303025805.625589-13-richard.henderson@linaro.org>

gdbstub: move chunks of user code into own files

The process was pretty similar to the softmmu move except we take the
time to split stuff between user.c and user-target.c to avoid as much
target specific compilation as possible. We also start to make use of
our shiny new header scheme so the user-only helpers can be included
without the rest of the exec/gsbstub.h cruft.

As before we split some functions into user and softmmu versions

Reviewed-by: Fabiano Rosas <farosas@suse.de>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20230302190846.2593720-12-alex.bennee@linaro.org>
Message-Id: <20230303025805.625589-12-richard.henderson@linaro.org>

vfio: Fix vfio_get_dev_region() trace event

Simply transpose 'x8' to fix the typo and remove the ending '8'

Fixes: e61a424f05 ("vfio: Create device specific region info helper")
Resolves: https://gitlab.com/qemu-project/qemu/-/issues/1526
Signed-off-by: Cédric Le Goater <clg@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Link: https://lore.kernel.org/r/20230303074330.2609377-1-clg@kaod.org
[aw: commit log s/revert/transpose/]
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>

vfio/migration: Rename entry points

Pick names that align with the section drivers should use them from,
avoiding the confusion of calling a _finalize() function from _exit()
and generalizing the actual _finalize() to handle removing the viommu
blocker.

Reviewed-by: Cédric Le Goater <clg@redhat.com>
Reviewed-by: Joao Martins <joao.m.martins@oracle.com>
Link: https://lore.kernel.org/r/167820912978.606734.12740287349119694623.stgit@omen
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>

hw/mem/cxl_type3: Add CXL RAS Error Injection Support.

CXL uses PCI AER Internal errors to signal to the host that an error has
occurred. The host can then read more detailed status from the CXL RAS
capability.

For uncorrectable errors: support multiple injection in one operation
as this is needed to reliably test multiple header logging support in an
OS. The equivalent feature doesn't exist for correctable errors, so only
one error need be injected at a time.

Note:
 - Header content needs to be manually specified in a fashion that
   matches the specification for what can be in the header for each
   error type.

Injection via QMP:
{ "execute": "qmp_capabilities" }
...
{ "execute": "cxl-inject-uncorrectable-errors",
  "arguments": {
    "path": "/machine/peripheral/cxl-pmem0",
    "errors": [
        {
            "type": "cache-address-parity",
            "header": [ 3, 4]
        },
        {
            "type": "cache-data-parity",
            "header": [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31]
        },
        {
            "type": "internal",
            "header": [ 1, 2, 4]
        }
        ]
  }}
...
{ "execute": "cxl-inject-correctable-error",
    "arguments": {
        "path": "/machine/peripheral/cxl-pmem0",
        "type": "physical"
    } }

Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Message-Id: <20230302133709.30373-9-Jonathan.Cameron@huawei.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

hw/pci/aer: Make PCIE AER error injection facility available for other emulation to use.

This infrastructure will be reused for CXL RAS error injection
in patches that follow.

Reviewed-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Message-Id: <20230302133709.30373-8-Jonathan.Cameron@huawei.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Fan Ni <fan.ni@samsung.com>

hw/cxl: Fix endian issues in CXL RAS capability defaults / masks

As these are about to be modified, fix the endian handle for
this set of registers rather than making it worse.

Note that CXL is currently only supported in QEMU on
x86 (arm64 patches out of tree) so we aren't going to yet hit
an problems with big endian. However it is good to avoid making
things worse for that support in the future.

Reviewed-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Message-Id: <20230302133709.30373-7-Jonathan.Cameron@huawei.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Fan Ni <fan.ni@samsung.com>

hw/mem/cxl-type3: Add AER extended capability

This enables AER error injection to function as expected.
It is intended as a building block in enabling CXL RAS error injection
in the following patches.

Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Reviewed-by: Dave Jiang <dave.jiang@intel.com>
Message-Id: <20230302133709.30373-6-Jonathan.Cameron@huawei.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Fan Ni <fan.ni@samsung.com>

hw/pci-bridge/cxl_root_port: Wire up MSI

Done to avoid fixing ACPI route description of traditional PCI interrupts on q35
and because we should probably move with the times anyway.

Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Reviewed-by: Dave Jiang <dave.jiang@intel.com>
Message-Id: <20230302133709.30373-5-Jonathan.Cameron@huawei.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Fan Ni <fan.ni@samsung.com>

hw/pci-bridge/cxl_root_port: Wire up AER

We are missing necessary config write handling for AER emulation in
the CXL root port. Add it based on pcie_root_port.c

Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Reviewed-by: Dave Jiang <dave.jiang@intel.com>
Message-Id: <20230302133709.30373-4-Jonathan.Cameron@huawei.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Fan Ni <fan.ni@samsung.com>

hw/pci/aer: Add missing routing for AER errors

PCIe r6.0 Figure 6-3 "Pseudo Logic Diagram for Selected Error Message Control
and Status Bits" includes a right hand branch under "All PCI Express devices"
that allows for messages to be generated or sent onwards without SERR#
being set as long as the appropriate per error class bit in the PCIe
Device Control Register is set.

Implement that branch thus enabling routing of ERR_COR, ERR_NONFATAL
and ERR_FATAL under OSes that set these bits appropriately (e.g. Linux)

Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Reviewed-by: Dave Jiang <dave.jiang@intel.com>
Message-Id: <20230302133709.30373-3-Jonathan.Cameron@huawei.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Fan Ni <fan.ni@samsung.com>

hw/pci/aer: Implement PCI_ERR_UNCOR_MASK register

This register in AER should be both writeable and should
have a default value with a couple of the errors masked
including the Uncorrectable Internal Error used by CXL for
it's error reporting.

Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Reviewed-by: Dave Jiang <dave.jiang@intel.com>
Message-Id: <20230302133709.30373-2-Jonathan.Cameron@huawei.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Fan Ni <fan.ni@samsung.com>

pcihp: add ACPI PCI hotplug specific is_hotpluggable_bus() callback

Provide pcihp specific callback to check if bus is hotpluggable
and consolidate its scattered hotplug criteria there.
While at it clean up no longer needed
   qbus_set_hotplug_handler(BUS(bus), NULL)
workarounds since callback makes qbus_is_hotpluggable() return
correct answer even if hotplug_handler is set on bus.

PS:
see ("pci: fix 'hotplugglable' property behavior") for details
why callback was introduced.

Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-35-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

pcihp: move fields enabling hotplug into AcpiPciHpState

... instead of duplicating them in piix4 and lpc and then
trying to pass them to pcihp routines as arguments.
it simplifies call sites and places pcihp specific in
its own structure.

Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-34-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

acpi: pci: move out ACPI PCI hotplug generator from generic slot generator build_append_pci_bus_devices()

Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-33-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

acpi: pci: move BSEL into build_append_pcihp_slots()

Generic PCI enumeration code doesn't really need access to
BSEL value, it is only used as means to decide if hotplug
enumerator should be called.

Use stateless object_property_find() to do that, and move
the rest of BSEL handling into build_append_pcihp_slots()
where it belongs.

This cleans up generic code a bit from hotplug stuff
and follow up patch will remove remaining call to
build_append_pcihp_slots() from generic code, making
it possible to use without ACPI PCI hotplug dependencies.

Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-32-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

acpi: pci: drop BSEL usage when deciding that device isn't hotpluggable

previous commit ("pci: fix 'hotplugglable' property behavior") fixed
pcie root port's 'hotpluggable' property to behave consistently.

So we don't need a BSEL crutch anymore to see of device is not
hotpluggable, drop it from 'generic' PCI slots description handling.

BSEL is still used to decide if hotplug part should be called
but that will be moved out of generic code to hotplug one by
followup patches.

Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-31-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

pci: move acpi-index uniqueness check to generic PCI device code

acpi-index is now working with non-hotpluggable buses
(pci/q35 machine hostbridge), it can be used even if
ACPI PCI hotplug is disabled and as result acpi-index
uniqueness check will be omitted (since the check is
done by ACPI PCI hotplug handler, which isn't wired
when ACPI PCI hotplug is disabled).
Move check and related code to generic PCIDevice so it
would be independent of ACPI PCI hotplug.

Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-30-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

tests: acpi: update expected blobs

an extra devices at non-zero function address with static
_DSM method get exposed, ex:

  +            Device (S15)
  +            {
  +                Name (_ADR, 0x00020005)  // _ADR: Address
  +                Method (_DSM, 4, Serialized)  // _DSM: Device-Specific Method
  +                {
  +                    Local0 = Package (0x01)
  +                        {
  +                            0x66
  +                        }
  +                    Return (EDSM (Arg0, Arg1, Arg2, Arg3, Local0))
  +                }
  +            }

Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-29-imammedo@redhat.com>

tests: acpi: add non zero function device with acpi-index on non-hotpluggble bus

Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-28-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

tests: acpi: whitelist DSDT before adding non-0 function device with acpi-index to testcases

Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-27-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

tests: acpi: update expected blobs

in PC machine case piix3-ide and PIIX4_PM get exposed

  +            Device (S09)
  +            {
  +                Name (_ADR, 0x00010001)  // _ADR: Address
  +            }
  +
  +            Device (S0B)
  +            {
  +                Name (_ADR, 0x00010003)  // _ADR: Address
  +            }

in q35 machine case ich9-ahci gets exposed
  +            Device (SFA)
  +            {
  +                Name (_ADR, 0x001F0002)  // _ADR: Address
  +            }

and addtional pci-testdev, virtio-balloon exposed in q35 multi-bridge test case
  +            Device (S14)
  +            {
  +                Name (_ADR, 0x00020004)  // _ADR: Address
  +            }
  +
  ...
  +            Device (S22)
  +            {
  +                Name (_ADR, 0x00040002)  // _ADR: Address
  +            }

Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-26-imammedo@redhat.com>

acpi: pci: describe all functions on populated slots

describing all present devices on functions other than
0 was complicated when non hotplug and hotplug code
was intermixed. So QEMU has been excluding non zero
functions since they are not supported by hotplug code,
then a condition to whitelist coldplugged bridges was
added and later whitelisting of devices that advertise
presence of their own AML description.

With non hotplug and hotplug code separated, it is
possible to relax rules and allow describing all
non-hotpluggble functions and hence simplify
conditions whether PCI device should be enumerated by
generic (non-hotplug) code.

Price of that simplification is an extra few Device()
descriptors in DSDT exposing built-in chipset functions,
which has no functional effect on guest side.

Apart from that, the enumeration of non zero functions,
allows to attach more NICs with acpi-index enabled
directly on hostbridge (if hotplug is not required).

Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-25-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

tests: acpi: whitelist DSDT before exposing non zero functions

Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-24-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

tests: acpi: update expected blobs

the only chenge is addition of _DSM- > EDSM method
on non-hotpluggable devices with configured acpi-index.
Something like:

  +                Method (_DSM, 4, Serialized)  // _DSM: Device-Specific Method
  +                {
  +                    Local0 = Package (0x01)
  +                        {
  +                            0x65
  +                        }
  +                    Return (EDSM (Arg0, Arg1, Arg2, Arg3, Local0))
  +                }

Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-23-imammedo@redhat.com>

acpi: pci: support acpi-index for non-hotpluggable devices

Inject static _DSM (EDSM) if non-hotpluggable device has
acpi-index configured on it.
It lets use acpi-index non-hotpluggable devices / devices
attached to non-hotpluggable bus.

Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-22-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

tests: acpi: add device with acpi-index on non-hotpluggble bus

Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-21-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

tests: acpi: whitelist DSDT before adding device with acpi-index to testcases

Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-20-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

tests: acpi: update expected blobs

Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-19-imammedo@redhat.com>

acpi: pci: add EDSM method to DSDT

it's a helper method for acpi-index support on PCI buses
that do no support or have disabled ACPI PCI hotplug
or for non-hotpluggble endpoint devices.
(like non-hotpluggble NICs, integrated endpoints and
later for machines that do not support ACPI PCI hotplug)

no functional change, commit adds only EDSM method in DSDT
without any users. (the follow up patches will use it)

Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-18-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

tests: acpi: whitelist DSDT before adding EDSM method

Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-17-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

tests: acpi: update expected blobs

only following context change:
 -  Local1 = Zero
    If ((Arg0 != ToUUID ("e5c937d0-3553-4d7a-9117-ea4d19c3434d") /* Device Labeling Interface */))
    {
        Return (Local0)
 ...
        Return (Local0)
    }

 +  Local1 = Zero
    Local2 = AIDX (DerefOf (Arg4 [Zero]), DerefOf (Arg4 [One]

Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-16-imammedo@redhat.com>

pcihp: move PCI _DSM function 0 prolog into separate function

it will be reused by follow up patches that will implement
static _DSM for non-hotpluggable devices.

no functional AML change, only context one, where 'cap' (Local1)
initialization is moved after UUID/revision checks.

Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-15-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

tests: acpi: whitelist DSDT blobs before isolating PCI _DSM func 0 prolog

Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-14-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

pci: fix 'hotplugglable' property behavior

Currently the property may flip its state
during VM bring up or just doesn't work as
the name implies.

In particular with PCIE root port that has
'hotplug={on|off}' property, and when it's
turned off, one would expect
  'hotpluggable' == false
for any devices attached to it.
Which is not the case since qbus_is_hotpluggable()
used by the property just checks for presence
of any hotplug_handler set on bus.

The problem is that name BusState::hotplug_handler
from its inception is misnomer, as it handles
not only hotplug but also in many cases coldplug
as well (i.e. generic wiring interface), and
it's fine to have hotplug_handler set on bus
while it doesn't support hotplug (ex. pcie-slot
with hotplug=off).

Another case of root port flipping 'hotpluggable'
state when ACPI PCI hotplug is enabled in this
case root port with 'hotplug=off' starts as
hotpluggable and then later on, pcihp
hotplug_handler clears hotplug_handler
explicitly after checking root port's 'hotplug'
property.

So root-port hotpluggablity check sort of works
if pcihp is enabled but is broken if pcihp is
disabled.

One way to deal with the issue is to ask
hotplug_handler if bus it controls is hotpluggable
or not. To do that add is_hotpluggable_bus()
hook to HotplugHandler interface and use it in
'hotpluggable' property + teach pcie-slot to
actually look into 'hotplug' property state
before deciding if bus is hotpluggable.

Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-13-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

pcihp: piix4: do not redirect hotplug controller to piix4 when ACPI hotplug is disabled

commit [1] added ability to disable ACPI PCI hotplug
on hostbridge but forgot to take into account that it
should disable all ACPI hotplug machinery in case both
hostbridge and bridge hotplug are disabled.

Commit [2] tried to fix that, however it forgot to
remove hotplug_handler override which hands hotplug
control over to piix4 hotplug controller
(uninitialized after [2]).

As result at the time bridge is plugged in, its default
(SHPC) hotplug handler is replaced by piix4 one in
  acpi_pcihp_device_plug_cb()
    ...
    if (!s->legacy_piix &&
       ...
       qbus_set_hotplug_handler(BUS(sec), OBJECT(hotplug_dev));

which is acting on uninitialized s->legacy_piix value
(0 by default) that was supposed to be initialized by
acpi_pcihp_init(), that is no longer called due to
following condition being false:

  piix4_acpi_system_hot_add_init()
    if (s->use_acpi_hotplug_bridge || s->use_acpi_root_pci_hotplug) {

and the bridge ends up with piix4 as hotplug handler
instead of shpc one.

Followup hotplug on that bridge as result yields
piix4 specific error:

  Error: Unsupported bus. Bus doesn't have property 'acpi-pcihp-bsel' set

1) 3d7e78aa777 (Introduce a new flag for i440fx to disable PCI hotplug on the root bus)
2) df4008c9c59 (piix4: don't reserve hw resources when hotplug is off globally)

Fixes: df4008c9c59 (piix4: don't reserve hw resources when hotplug is off globally)
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-12-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

tests: acpi: update expected blobs

BNUM numbering changes across DSDT due to addition of new bridges.

Fixed missing PCI tree brunch (q35/DSDT.multi-bridge case):

  //  -device pcie-root-port,id=rpnohp,chassis=8,addr=0xA.0,hotplug=off
  +            Device (S50)
  +            {
  +                Name (_ADR, 0x000A0000)  // _ADR: Address
  //  -device pcie-root-port,id=rp3,chassis=9,bus=rpnohp
  +                Device (S00)
  +                {
  +                    Name (_ADR, Zero)  // _ADR: Address
  +                    Name (BSEL, Zero)
  +                    Device (S00)
  +                    {
  +                        Name (_ADR, Zero)  // _ADR: Address
  +                        Name (ASUN, Zero)
  +                        Method (_DSM, 4, Serialized)  // _DSM: Device-Specific Method
  +                        {
  +                            Local0 = Package (0x02)
  +                                {
  +                                    BSEL,
  +                                    ASUN
  +                                }
  +                            Return (PDSM (Arg0, Arg1, Arg2, Arg3, Local0))
  +                        }
  +
  +                        Name (_SUN, Zero)  // _SUN: Slot User Number
  +                        Method (_EJ0, 1, NotSerialized)  // _EJx: Eject Device
  +                        {
  +                            PCEJ (BSEL, _SUN)
  +                        }
  +                    }
  +
  +                    Method (DVNT, 2, NotSerialized)
  +                    {
  +                        If ((Arg0 & One))
  +                        {
  +                            Notify (S00, Arg1)
  +                        }
  +                    }
  +                }
  +            }

Fixed hotplug notification for leaf root port (hotplug=on) attached to
intermediate root port (hotplug=off) (q35/DSDT.multi-bridge case)

  //  -device pcie-root-port,id=rpnohp,chassis=8,addr=0xA.0,hotplug=off
  +        Scope (S50)
  +        {
  //  -device pcie-root-port,id=rp3,chassis=9,bus=rpnohp
  +            Scope (S00)
  +            {
  +                Method (PCNT, 0, NotSerialized)
  +                {
  +                    BNUM = Zero
  +                    DVNT (PCIU, One)
  +                    DVNT (PCID, 0x03)
  +                }
  +            }
  +
  +            Method (PCNT, 0, NotSerialized)
  +            {
  +                ^S00.PCNT ()
  +            }
  +        }
  ...
           Method (PCNT, 0, NotSerialized)
           {
  +            ^S50.PCNT ()
               ^S13.PCNT ()

Populated slots being described on coldplugged bridges even if
ACPI bridge hotplug is disabled.
(pc/DSDT.hpbridge and pc/DSDT.hpbrroot)
  ...
               Device (S18)
               {
                   Name (_ADR, 0x00030000)  // _ADR: Address
  +                Device (S08)
  +                {
  +                    Name (_ADR, 0x00010000)  // _ADR: Address
  +                }
  +
  +                Device (S10)
  +                {
  +                    Name (_ADR, 0x00020000)  // _ADR: Address
  +                }
               }
  ...
               Device (S18)
               {
                   Name (_ADR, 0x00030000)  // _ADR: Address
  +                Device (S00)
  +                {
  +                    Name (_ADR, Zero)  // _ADR: Address
  +                }
               }

Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-11-imammedo@redhat.com>

x86: pcihp: fix missing bridge AML when intermediate root-port has 'hotplug=off' set

(I practice [1] hasn't broke anything since on hardware side we unset
hotplug_handler on such intermediate port => hotplug behind it has
never worked)

When deciding if bridge should be described, the original
condition was

  cold_plugged_bridge && pcihp_bridge_en

which was replaced [1] by

  bridge has ACPI_PCIHP_PROP_BSEL

the later however is not the same thing as the original
and flips to false if intermediate bridge has hotplug
turned off (root-port with 'hotplug=off' option).

Since we already in build_pci_bridge_aml(), the question
if it's bridge is answered. Use DeviceState::hotplugged
to make decision if bridge should describe its slots.

What's left out is pcihp_bridge_en, which tells us if
ACPI bridge hotplug is enabled.

With hotplug and non hotplug part now being mostly
separated, omitting this check will only lead to
colplugged bridges describe occupied slots in case
when ACPI bridge hotplug is disabled.
Which makes behavior consistent with occupied slots
on hostbridge.

Ex (pc/DSDT.hpbrroot diff):
  ...
               Device (S20)
               {
                   Name (_ADR, 0x00040000)  // _ADR: Address
  +                Device (S08)
  +                {
  +                    Name (_ADR, 0x00010000)  // _ADR: Address
  +                }
  +
  +                Device (S10)
  +                {
  +                    Name (_ADR, 0x00020000)  // _ADR: Address
  +                }
               }
  ...

PS:
testing shows that above doesn't affect adversely guest OS
behavior: i.e. if ACPI bridge hotplug is enabled it's
expected behaviour, and with ACPI bridge hotplug is disabled
(a.k. native hotplug), it doesn't break slot enumeration
nor native hotplug. (tested with RHEL9.0 and WS2022).

1)
Fixes: 6c36ec46b0d ("pcihp: make bridge describe itself using AcpiDevAmlIfClass:build_dev_aml")
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-10-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

tests: acpi: whitelist pc/DSDT.hpbrroot and pc/DSDT.hpbridge tests

follow up fix for missing root-port AML will affect these tests
by adding non-hotpluggable Device descriptors of colplugged
bridges when bridge hotplug is disabled.

Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-9-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

x86: pcihp: fix missing PCNT callchain when intermediate root-port has 'hotplug=off' set

Beside BSEL numbers change (due to 2 extra root-ports in q35/miltibridge test),
following change is expected:

       Scope (\_SB.PCI0)
       {
  ...
  +        Scope (S50)
  +        {
  +            Scope (S00)
  +            {
  +                Method (PCNT, 0, NotSerialized)
  +                {
  +                    BNUM = Zero
  +                    DVNT (PCIU, One)
  +                    DVNT (PCID, 0x03)
  +                }
  +            }
  +
  +            Method (PCNT, 0, NotSerialized)
  +            {
  +                ^S00.PCNT
  +            }
  +        }
  ...
           Method (PCNT, 0, NotSerialized)
           {
  +            ^S50.PCNT ()
               ^S13.PCNT ()
               ^S12.PCNT ()
               ^S11.PCNT ()

I practice [1] hasn't broke anything since on hardware side we unset
hotplug_handler on such intermediate port => hotplug behind it has
not been properly wired and as result not worked.

1)
Fixes: ddab4d3fae4e8 ("pcihp: compose PCNT callchain right before its user _GPE._E01")
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-8-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

tests: acpi: extend multi-bridge case with case 'root-port,id=HOHP,hotplug=off root-port,bus=NOHP'

Following corner case wasn't covered:

  -device pcie-root-port,id=NO_HOTPLUG,hotplug=off
  -device pcie-root-port,bus=NO_HOTPLUG

when intermediate root-port has explicitly disabled hotplug,
all hierarchy below it is not described anymore (used to be
described in 7.2)

So as result we see only NO_HOTPLUG root-port described

  +            Device (S50)
  +            {
  +                Name (_ADR, 0x000A0000)  // _ADR: Address
  +            }

and no children nor notification chain for them are being composed.
Follow up patches will fix missing leaf root-port descriptor
and notification chain that should accompany it.

Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-7-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

tests: acpi: whitelist q35/DSDT.multi-bridge before extending testcase

Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-6-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

tests: acpi: update expected blobs

expected changes:
Basically adds devices present on root bus in form:
  Device (SXX)
  {
     Name (_ADR, 0xYYYYYYYY)  // _ADR: Address
  }

On top of that For q35.noacpihp, all ACPI PCI hotplug
AML is removed and _OSC get native hotplug enabled:

                       CreateDWordField (Arg3, 0x04, CDW2)
                       CreateDWordField (Arg3, 0x08, CDW3)
                       Local0 = CDW3 /* \_SB_.PCI0._OSC.CDW3 */
  -                    Local0 &= 0x1E
  +                    Local0 &= 0x1F
                       If ((Arg1 != One))
                       {
                           CDW1 |= 0x08
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-5-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

tests: acpi: add test_acpi_q35_tcg_no_acpi_hotplug test and extend test_acpi_piix4_no_acpi_pci_hotplug

test bridge AML generator with ACPI PCI hotplug disabled
(i.e. with native hotplug enabled/disabled per bridge/root port)

PS:
while at make sure that devices on pci-bridge are starting
from addr=1.0 as slot 0 is not available there and test
passes only because of a bug in ACPI hotplug that will be
fixed by follow up patch

Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-4-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

tests: acpi: whitelist new q35.noacpihp test and pc.hpbrroot

for q35.noacpihp use plain default Q35 DSDT table as a starting point.

Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-3-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

Revert "tests/qtest: Check for devices in bios-tables-test"

This reverts commit c471eb4f40445908c1be7bb11a37ac676a0edae7.

which broke acpi tables test and rebuild due to skipping some tests
even thought none of devices tests depend on weren't disabled.

As result it leads to some expected tables not being updated,
merge conflicts and tests failure.

Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-2-imammedo@redhat.com>
Acked-by: Fabiano Rosas <farosas@suse.de>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

vdpa: return VHOST_F_LOG_ALL in vhost-vdpa devices

vhost-vdpa devices can return this feature now that blockers have been
set in case some features are not met.

Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Message-Id: <20230303172445.1089785-15-eperezma@redhat.com>
Tested-by: Lei Yang <leiyang@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

vdpa net: allow VHOST_F_LOG_ALL

Since some actions move to the start function instead of init, the
device features may not be the parent vdpa device's, but the one
returned by vhost backend.  If transition to SVQ is supported, the vhost
backend will return _F_LOG_ALL to signal the device is migratable.

Add VHOST_F_LOG_ALL.  HW dirty page tracking can be added on top of this
change if the device supports it in the future.

Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Message-Id: <20230303172445.1089785-14-eperezma@redhat.com>
Tested-by: Lei Yang <leiyang@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

vdpa: block migration if SVQ does not admit a feature

Next patches enable devices to be migrated even if vdpa netdev has not
been started with x-svq. However, not all devices are migratable, so we
need to block migration if we detect that.

Block migration if we detect the device expose a feature SVQ does not
know how to work with.

Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
Message-Id: <20230303172445.1089785-13-eperezma@redhat.com>
Tested-by: Lei Yang <leiyang@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

vdpa: block migration if device has unsupported features

A vdpa net device must initialize with SVQ in order to be migratable at
this moment, and initialization code verifies some conditions.  If the
device is not initialized with the x-svq parameter, it will not expose
_F_LOG so the vhost subsystem will block VM migration from its
initialization.

Next patches change this, so we need to verify migration conditions
differently.

QEMU only supports a subset of net features in SVQ, and it cannot
migrate state that cannot track or restore in the destination.  Add a
migration blocker if the device offers an unsupported feature.

Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
Message-Id: <20230303172445.1089785-12-eperezma@redhat.com>
Tested-by: Lei Yang <leiyang@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

vdpa net: block migration if the device has CVQ

Devices with CVQ need to migrate state beyond vq state.  Leaving this to
future series.

Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
Message-Id: <20230303172445.1089785-11-eperezma@redhat.com>
Tested-by: Lei Yang <leiyang@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

vdpa: disable RAM block discard only for the first device

Although it does not make a big difference, its more correct and
simplifies the cleanup path in subsequent patches.

Move ram_block_discard_disable(false) call to the top of
vhost_vdpa_cleanup because:
* We cannot use vhost_vdpa_first_dev after dev->opaque = NULL
  assignment.
* Improve the stack order in cleanup: since it is the last action taken
  in init, it should be the first at cleanup.

Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
Message-Id: <20230303172445.1089785-10-eperezma@redhat.com>
Tested-by: Lei Yang <leiyang@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

vdpa: add vdpa net migration state notifier

This allows net to restart the device backend to configure SVQ on it.

Ideally, these changes should not be net specific and they could be done
in:
* vhost_vdpa_set_features (with VHOST_F_LOG_ALL)
* vhost_vdpa_set_vring_addr (with .enable_log)
* vhost_vdpa_set_log_base.

However, the vdpa net backend is the one with enough knowledge to
configure everything because of some reasons:
* Queues might need to be shadowed or not depending on its kind (control
  vs data).
* Queues need to share the same map translations (iova tree).

Also, there are other problems that may have solutions but complicates
the implementation at this stage:
* We're basically duplicating vhost_dev_start and vhost_dev_stop, and
  they could go out of sync.  If we want to reuse them, we need a way to
  skip some function calls to avoid recursiveness (either vhost_ops ->
  vhost_set_features, vhost_set_vring_addr, ...).
* We need to traverse all vhost_dev of a given net device twice: one to
  stop and get the vq state and another one after the reset to
  configure properties like address, fd, etc.

Because of that it is cleaner to restart the whole net backend and
configure again as expected, similar to how vhost-kernel moves between
userspace and passthrough.

If more kinds of devices need dynamic switching to SVQ we can:
* Create a callback struct like VhostOps and move most of the code
  there.  VhostOps cannot be reused since all vdpa backend share them,
  and to personalize just for networking would be too heavy.
* Add a parent struct or link all the vhost_vdpa or vhost_dev structs so
  we can traverse them.

Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
Message-Id: <20230303172445.1089785-9-eperezma@redhat.com>
Tested-by: Lei Yang <leiyang@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

vdpa: move vhost reset after get vring base

The function vhost.c:vhost_dev_stop calls vhost operation
vhost_dev_start(false). In the case of vdpa it totally reset and wipes
the device, making the fetching of the vring base (virtqueue state) totally
useless.

The kernel backend does not use vhost_dev_start vhost op callback, but
vhost-user do. A patch to make vhost_user_dev_start more similar to vdpa
is desirable, but it can be added on top.

Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
Message-Id: <20230303172445.1089785-8-eperezma@redhat.com>
Tested-by: Lei Yang <leiyang@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

vdpa: add vhost_vdpa_suspend

The function vhost.c:vhost_dev_stop fetches the vring base so the vq
state can be migrated to other devices.  However, this is unreliable in
vdpa, since we didn't signal the device to suspend the queues, making
the value fetched useless.

Suspend the device if possible before fetching first and subsequent
vring bases.

Moreover, vdpa totally reset and wipes the device at the last device
before fetch its vrings base, making that operation useless in the last
device. This will be fixed in later patches of this series.

Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
Message-Id: <20230303172445.1089785-7-eperezma@redhat.com>
Tested-by: Lei Yang <leiyang@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

vdpa: add vhost_vdpa->suspended parameter

This allows vhost_vdpa to track if it is safe to get the vring base from
the device or not.  If it is not, vhost can fall back to fetch idx from
the guest buffer again.

No functional change intended in this patch, later patches will use this
field.

Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
Message-Id: <20230303172445.1089785-6-eperezma@redhat.com>
Tested-by: Lei Yang <leiyang@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

vdpa: rewind at get_base, not set_base

At this moment it is only possible to migrate to a vdpa device running
with x-svq=on. As a protective measure, the rewind of the inflight
descriptors was done at the destination. That way if the source sent a
virtqueue with inuse descriptors they are always discarded.

Since this series allows to migrate also to passthrough devices with no
SVQ, the right thing to do is to rewind at the source so the base of
vrings are correct.

Support for inflight descriptors may be added in the future.

Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Message-Id: <20230303172445.1089785-5-eperezma@redhat.com>
Tested-by: Lei Yang <leiyang@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

vdpa: Negotiate _F_SUSPEND feature

This is needed for qemu to know it can suspend the device to retrieve
its status and enable SVQ with it, so all the process is transparent to
the guest.

Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Message-Id: <20230303172445.1089785-4-eperezma@redhat.com>
Tested-by: Lei Yang <leiyang@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

vdpa: Remember last call fd set

As SVQ can be enabled dynamically at any time, it needs to store call fd
always.

Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
Message-Id: <20230303172445.1089785-3-eperezma@redhat.com>
Tested-by: Lei Yang <leiyang@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

vdpa net: move iova tree creation from init to start

Only create iova_tree if and when it is needed.

The cleanup keeps being responsible for the last VQ but this change
allows it to merge both cleanup functions.

Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Message-Id: <20230303172445.1089785-2-eperezma@redhat.com>
Tested-by: Lei Yang <leiyang@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

MAINTAINERS: add myself as the maintainer for cryptodev

I developed the akcipher service, QoS setting, QMP/HMP commands and
statistics accounting for crypto device. Making myself as the
maintainer for QEMU's cryptodev.

Cc: Gonglei <arei.gonglei@huawei.com>
Signed-off-by: zhenwei pi <pizhenwei@bytedance.com>
Message-Id: <20230301105847.253084-13-pizhenwei@bytedance.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

cryptodev: Support query-stats QMP command

Now we can use "query-stats" QMP command to query statistics of
crypto devices. (Originally this was designed to show statistics
by '{"execute": "query-cryptodev"}'. Daniel Berrangé suggested that
querying configuration info by "query-cryptodev", and querying
runtime performance info by "query-stats". This makes sense!)

Example:
~# virsh qemu-monitor-command vm '{"execute": "query-stats", \
   "arguments": {"target": "cryptodev"} }' | jq
{
  "return": [
    {
      "provider": "cryptodev",
      "stats": [
        {
          "name": "asym-verify-bytes",
          "value": 7680
        },
        ...
        {
          "name": "asym-decrypt-ops",
          "value": 32
        },
        {
          "name": "asym-encrypt-ops",
          "value": 48
        }
      ],
      "qom-path": "/objects/cryptodev0" # support asym only
    },
    {
      "provider": "cryptodev",
      "stats": [
        {
          "name": "asym-verify-bytes",
          "value": 0
        },
        ...
        {
          "name": "sym-decrypt-bytes",
          "value": 5376
        },
        ...
      ],
      "qom-path": "/objects/cryptodev1" # support asym/sym
    }
  ],
  "id": "libvirt-422"
}

Suggested-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Signed-off-by: zhenwei pi <pizhenwei@bytedance.com>
Message-Id: <20230301105847.253084-12-pizhenwei@bytedance.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

cryptodev: support QoS

Add 'throttle-bps' and 'throttle-ops' limitation to set QoS. The
two arguments work with both QEMU command line and QMP command.

Example of QEMU command line:
-object cryptodev-backend-builtin,id=cryptodev1,throttle-bps=1600,\
throttle-ops=100

Example of QMP command:
virsh qemu-monitor-command buster --hmp qom-set /objects/cryptodev1 \
throttle-ops 100

or cancel limitation:
virsh qemu-monitor-command buster --hmp qom-set /objects/cryptodev1 \
throttle-ops 0

Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Signed-off-by: zhenwei pi <pizhenwei@bytedance.com>
Message-Id: <20230301105847.253084-11-pizhenwei@bytedance.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

cryptodev: Account statistics

Account OPS/BPS for crypto device, this will be used for 'query-stats'
QEMU monitor command and QoS in the next step.

Note that a crypto device may support symmetric mode, asymmetric mode,
both symmetric and asymmetric mode. So we use two structure to
describe the statistics of a crypto device.

Signed-off-by: zhenwei pi <pizhenwei@bytedance.com>
Message-Id: <20230301105847.253084-10-pizhenwei@bytedance.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>

cryptodev: Use CryptoDevBackendOpInfo for operation

Move queue_index, CryptoDevCompletionFunc and opaque into struct
CryptoDevBackendOpInfo, then cryptodev_backend_crypto_operation()
needs an argument CryptoDevBackendOpInfo *op_info only. And remove
VirtIOCryptoReq from cryptodev. It's also possible to hide
VirtIOCryptoReq into virtio-crypto.c in the next step. (In theory,
VirtIOCryptoReq is a private structure used by virtio-crypto only)

Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Signed-off-by: zhenwei pi <pizhenwei@bytedance.com>
Message-Id: <20230301105847.253084-9-pizhenwei@bytedance.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

hmp: add cryptodev info command

Example of this command:
 # virsh qemu-monitor-command vm --hmp info cryptodev
cryptodev1: service=[akcipher|mac|hash|cipher]
    queue 0: type=builtin
cryptodev0: service=[akcipher]
    queue 0: type=lkcf

Acked-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Signed-off-by: zhenwei pi <pizhenwei@bytedance.com>
Message-Id: <20230301105847.253084-8-pizhenwei@bytedance.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

cryptodev-builtin: Detect akcipher capability

Rather than exposing akcipher service/RSA algorithm to virtio crypto
device unconditionally, detect akcipher capability from akcipher
crypto framework. This avoids unsuccessful requests.

Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Signed-off-by: zhenwei pi <pizhenwei@bytedance.com>
Message-Id: <20230301105847.253084-7-pizhenwei@bytedance.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

cryptodev: Introduce 'query-cryptodev' QMP command

Now we have a QMP command to query crypto devices:
virsh qemu-monitor-command vm '{"execute": "query-cryptodev"}' | jq
{
  "return": [
    {
      "service": [
        "akcipher",
        "mac",
        "hash",
        "cipher"
      ],
      "id": "cryptodev1",
      "client": [
        {
          "queue": 0,
          "type": "builtin"
        }
      ]
    },
    {
      "service": [
        "akcipher"
      ],
      "id": "cryptodev0",
      "client": [
        {
          "queue": 0,
          "type": "lkcf"
        }
      ]
    }
  ],
  "id": "libvirt-417"
}

Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Signed-off-by: zhenwei pi <pizhenwei@bytedance.com>
Message-Id: <20230301105847.253084-6-pizhenwei@bytedance.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

cryptodev: Introduce server type in QAPI

Introduce cryptodev service type in cryptodev.json, then apply this
to related codes. Now we can remove VIRTIO_CRYPTO_SERVICE_xxx
dependence from QEMU cryptodev.

Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Signed-off-by: zhenwei pi <pizhenwei@bytedance.com>
Message-Id: <20230301105847.253084-5-pizhenwei@bytedance.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

cryptodev: Introduce cryptodev alg type in QAPI

Introduce cryptodev alg type in cryptodev.json, then apply this to
related codes, and drop 'enum CryptoDevBackendAlgType'.

There are two options:
1, { 'enum': 'QCryptodevBackendAlgType',
  'prefix': 'CRYPTODEV_BACKEND_ALG',
  'data': ['sym', 'asym']}
Then we can keep 'CRYPTODEV_BACKEND_ALG_SYM' and avoid lots of
changes.
2, changes in this patch(with prefix 'QCRYPTODEV_BACKEND_ALG').

To avoid breaking the rule of QAPI, use 2 here.

Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Signed-off-by: zhenwei pi <pizhenwei@bytedance.com>
Message-Id: <20230301105847.253084-4-pizhenwei@bytedance.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

cryptodev: Remove 'name' & 'model' fields

We have already used qapi to generate crypto device types, this allows
to convert type to a string 'model', so the 'model' field is not
needed.

And the 'name' field is not used by any backend driver, drop it.

Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Signed-off-by: zhenwei pi <pizhenwei@bytedance.com>
Message-Id: <20230301105847.253084-3-pizhenwei@bytedance.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

cryptodev: Introduce cryptodev.json

Introduce QCryptodevBackendType in cryptodev.json, also apply this to
related codes. Then we can drop 'enum CryptoDevBackendOptionsType'.

Note that `CRYPTODEV_BACKEND_TYPE_NONE` is *NOT* used by anywhere, so
drop it(no 'none' enum in QCryptodevBackendType).

Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Signed-off-by: zhenwei pi <pizhenwei@bytedance.com>
Message-Id: <20230301105847.253084-2-pizhenwei@bytedance.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

docs/devel: Document VFIO device dirty page tracking

Adjust the VFIO dirty page tracking documentation and add a section to
describe device dirty page tracking.

Signed-off-by: Avihai Horon <avihaih@nvidia.com>
Signed-off-by: Joao Martins <joao.m.martins@oracle.com>
Reviewed-by: Cédric Le Goater <clg@redhat.com>
Link: https://lore.kernel.org/r/20230307125450.62409-16-joao.m.martins@oracle.com
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>

vfio/migration: Query device dirty page tracking support

Now that everything has been set up for device dirty page tracking,
query the device for device dirty page tracking support.

Signed-off-by: Avihai Horon <avihaih@nvidia.com>
Signed-off-by: Joao Martins <joao.m.martins@oracle.com>
Reviewed-by: Cédric Le Goater <clg@redhat.com>
Link: https://lore.kernel.org/r/20230307125450.62409-15-joao.m.martins@oracle.com
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>

vfio/migration: Block migration with vIOMMU

Migrating with vIOMMU will require either tracking maximum
IOMMU supported address space (e.g. 39/48 address width on Intel)
or range-track current mappings and dirty track the new ones
post starting dirty tracking. This will be done as a separate
series, so add a live migration blocker until that is fixed.

Signed-off-by: Joao Martins <joao.m.martins@oracle.com>
Reviewed-by: Cédric Le Goater <clg@redhat.com>
Link: https://lore.kernel.org/r/20230307125450.62409-14-joao.m.martins@oracle.com
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>

vfio/common: Add device dirty page bitmap sync

Add device dirty page bitmap sync functionality. This uses the device
DMA logging uAPI to sync dirty page bitmap from the device.

Device dirty page bitmap sync is used only if all devices within a
container support device dirty page tracking.

Signed-off-by: Avihai Horon <avihaih@nvidia.com>
Signed-off-by: Joao Martins <joao.m.martins@oracle.com>
Reviewed-by: Cédric Le Goater <clg@redhat.com>
Link: https://lore.kernel.org/r/20230307125450.62409-13-joao.m.martins@oracle.com
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>

vfio/common: Extract code from vfio_get_dirty_bitmap() to new function

Extract the VFIO_IOMMU_DIRTY_PAGES ioctl code in vfio_get_dirty_bitmap()
to its own function.

This will help the code to be more readable after next patch will add
device dirty page bitmap sync functionality.

Signed-off-by: Avihai Horon <avihaih@nvidia.com>
Signed-off-by: Joao Martins <joao.m.martins@oracle.com>
Reviewed-by: Cédric Le Goater <clg@redhat.com>
Link: https://lore.kernel.org/r/20230307125450.62409-12-joao.m.martins@oracle.com
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>

vfio/common: Add device dirty page tracking start/stop

Add device dirty page tracking start/stop functionality. This uses the
device DMA logging uAPI to start and stop dirty page tracking by device.

Device dirty page tracking is used only if all devices within a
container support device dirty page tracking.

Signed-off-by: Avihai Horon <avihaih@nvidia.com>
Signed-off-by: Joao Martins <joao.m.martins@oracle.com>
Link: https://lore.kernel.org/r/20230307125450.62409-11-joao.m.martins@oracle.com
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>

target/mips: Fix JALS32/J32 instruction handling for microMIPS

microMIPS J & JAL instructions perform a jump in a 128MB region and 5
top bits of the address need to be preserved. This is different behavior
compared to standard mips systems, where the jump is executed within a
256MB region.
Note that microMIPS32 instruction set documentation appears to have
inconsistent information regarding JALX32 instruction - it is written in
the doc that:

"To execute a procedure call within the current 256 MB-aligned region
(...)
The low 26 bits of the target address is the target field shifted left
2 bits."

But the target address is already 26 bits. Moreover, the operation
description indicates that 28 bits are copied, so the statement about
use of 26 bits is _most likely_ incorrect and the corresponding code
remains the same as for standard mips instruction set.

Signed-off-by: Marcin Nowakowski <marcin.nowakowski@fungible.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20230216051717.3911212-2-marcin.nowakowski@fungible.com>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>

target/mips: Replace [g_]assert(0) -> g_assert_not_reached()

In order to avoid warnings such commit c0a6665c3c ("target/i386:
Remove compilation errors when -Werror=maybe-uninitialized"),
replace all assert(0) and g_assert(0) by g_assert_not_reached().

Remove any code following g_assert_not_reached().

See previous commit for rationale.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20230221232520.14480-4-philmd@linaro.org>

docs/system: Remove "mips" board from target-mips.rst

This board had been removed long ago in commit f169413c27
("hw/mips: Remove the 'r4k' machine")

Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-Id: <20230202132138.30945-2-jiaxun.yang@flygoat.com>
[PMD: Mention commit f169413c27]
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>