Fuzzing showed that a guest could bind an interdomain port to itself, by
guessing the next port to be allocated and putting that as the 'remote'
port number. By chance, that works because the newly-allocated port has
type EVTCHNSTAT_unbound. It shouldn't.
Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Reviewed-by: Paul Durrant <paul@xen.org>
Message-Id: <20230801175747.145906-4-dwmw2@infradead.org>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Coverity points out (CID 1508128) a bounds checking error. We need to check
for gsi >= IOAPIC_NUM_PINS, not just greater-than.
Also fix up an assert() that has the same problem, that Coverity didn't see.
Fixes: 4f81baa33e ("hw/xen: Support GSI mapping to PIRQ")
Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-Id: <20230801175747.145906-2-dwmw2@infradead.org>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
The IDE unplug function needs to reset the entire PCI device, to make
sure all state is initialized to defaults. This is done by calling
pci_device_reset, which resets not only the chip specific registers, but
also all PCI state. This fixes "unplug" in a Xen HVM domU with the
modular legacy xenlinux PV drivers.
Commit ee358e919e ("hw/ide/piix: Convert reset handler to
DeviceReset") changed the way how the the disks are unplugged. Prior
this commit the PCI device remained unchanged. After this change,
piix_ide_reset is exercised after the "unplug" command, which was not
the case prior that commit. This function resets the command register.
As a result the ata_piix driver inside the domU will see a disabled PCI
device. The generic PCI code will reenable the PCI device. On the qemu
side, this runs pci_default_write_config/pci_update_mappings. Here a
changed address is returned by pci_bar_address, this is the address
which was truncated in piix_ide_reset. In case of a Xen HVM domU, the
address changes from 0xc120 to 0xc100. This truncation was a bug in
piix_ide_reset, which was fixed in commit 230dfd9257 ("hw/ide/piix:
properly initialize the BMIBA register"). If pci_xen_ide_unplug had used
pci_device_reset, the PCI registers would have been properly reset, and
commit ee358e919e would have not introduced a regression for this
specific domU environment.
While the unplug is supposed to hide the IDE disks, the changed BMIBA
address broke the UHCI device. In case the domU has an USB tablet
configured, to recive absolute pointer coordinates for the GUI, it will
cause a hang during device discovery of the partly discovered USB hid
device. Reading the USBSTS word size register will fail. The access ends
up in the QEMU piix-bmdma device, instead of the expected uhci device.
Here a byte size request is expected, and a value of ~0 is returned. As
a result the UCHI driver sees an error state in the register, and turns
off the UHCI controller.
Signed-off-by: Olaf Hering <olaf@aepfle.de>
Reviewed-by: Paul Durrant <paul@xen.org>
Message-Id: <20230720072950.20198-1-olaf@aepfle.de>
Signed-off-by: Anthony PERARD <anthony.perard@citrix.com>
Coverity points out (CID 1513106, 1513107) that MemoryListener is a
192 byte struct which we are passing around by value. Switch to
passing a const pointer into xen_register_ioreq() and then to
xen_do_ioreq_register(). We can also make the file-scope
MemoryListener variables const, since nothing changes them.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Acked-by: Anthony PERARD <anthony.perard@citrix.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-Id: <20230718101057.1110979-1-peter.maydell@linaro.org>
Signed-off-by: Anthony PERARD <anthony.perard@citrix.com>
Commit 1898293990 ("xen-block: Use specific blockdev driver")
introduced a new error path, without taking care of allocated
resources.
So only allocate the qdicts after the error check, and free both
`filename` and `driver` when we are about to return and thus taking
care of both success and error path.
Coverity only spotted the leak of qdicts (*_layer variables).
Reported-by: Peter Maydell <peter.maydell@linaro.org>
Fixes: Coverity CID 1508722, 1398649
Fixes: 1898293990 ("xen-block: Use specific blockdev driver")
Signed-off-by: Anthony PERARD <anthony.perard@citrix.com>
Reviewed-by: Paul Durrant <paul@xen.org>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Message-Id: <20230704171819.42564-1-anthony.perard@citrix.com>
Signed-off-by: Anthony PERARD <anthony.perard@citrix.com>
Coverity was unhappy (CID 1508359) because we didn't check the return of
init_walk_op() in transaction_commit(), despite doing so at every other
call site.
Strictly speaking, this is a false positive since it can never fail. It
only fails for invalid user input (transaction ID or path), and both of
those are hard-coded to known sane values in this invocation.
But Coverity doesn't know that, and neither does the casual reader of the
code.
Returning an error here would be weird, since the transaction *is*
committed by this point; all the walk_op is doing is firing watches on
the newly-committed changed nodes. So make it a g_assert(!ret), since
it really should never happen.
Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Reviewed-by: Paul Durrant <paul@xen.org>
Message-Id: <20076888f6bdf06a65aafc5cf954260965d45b97.camel@infradead.org>
Signed-off-by: Anthony PERARD <anthony.perard@citrix.com>
Use the stl/ldl pci dma api for writing/reading doorbells. This removes
the explicit endian conversions.
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Cédric Le Goater <clg@redhat.com>
Tested-by: Cédric Le Goater <clg@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Klaus Jensen <k.jensen@samsung.com>
This change is cosmetic. A comment is added explaining why we need to check for
the availability of function 0 when we hotplug a device.
CC: mst@redhat.com
CC: mjt@tls.msk.ru
Signed-off-by: Ani Sinha <anisinha@redhat.com>
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
The firmware of the m68k next-cube machine uses the loopback mode
for self-testing the hardware and currently fails during this step.
By implementing the loopback mode, we can make the firmware pass
to the next step.
Signed-off-by: Thomas Huth <huth@tuxfamily.org>
Reviewed-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-Id: <20230716153519.31722-1-huth@tuxfamily.org>
It's possible to compile QEMU without the USB devices (e.g. when using
"--without-default-devices" as option for the "configure" script).
To be still able to run the loongson3-virt machine in default mode with
such a QEMU binary, we have to check here for the availability of the
OHCI controller first before instantiating the USB devices.
Signed-off-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-Id: <20230714104903.284845-1-thuth@redhat.com>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Since commit c0a55a0c9d "hw/sd/sdhci: Support big endian SD host controller
interfaces" sdhci_common_realize() forces all SD card controllers to use either
sdhci_mmio_le_ops or sdhci_mmio_be_ops, depending on the "endianness" property.
However, there are device models which use different MMIO ops: TYPE_IMX_USDHC
uses usdhc_mmio_ops and TYPE_S3C_SDHCI uses sdhci_s3c_mmio_ops.
Forcing sdhci_mmio_le_ops breaks SD card handling on the "sabrelite" board, for
example. Fix this by defaulting the io_ops to little endian and switch to big
endian in sdhci_common_realize() only if there is a matchig big endian variant
available.
Fixes: c0a55a0c9d ("hw/sd/sdhci: Support big endian SD host controller
interfaces")
Signed-off-by: Bernhard Beschow <shentey@gmail.com>
Tested-by: Guenter Roeck <linux@roeck-us.net>
Message-Id: <20230709080950.92489-1-shentey@gmail.com>
The implementation of the SMMUv3 has multiple places where it reads a
data structure from the guest and directly operates on it without
doing a guest-to-host endianness conversion. Since all SMMU data
structures are little-endian, this means that the SMMU doesn't work
on a big-endian host. In particular, this causes the Avocado test
machine_aarch64_virt.py:Aarch64VirtMachine.test_alpine_virt_tcg_gic_max
to fail on an s390x host.
Add appropriate byte-swapping on reads and writes of guest in-memory
data structures so that the device works correctly on big-endian
hosts.
As part of this we constrain queue_read() to operate only on Cmd
structs and queue_write() on Evt structs, because in practice these
are the only data structures the two functions are used with, and we
need to know what the data structure is to be able to byte-swap its
parts correctly.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Tested-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Eric Auger <eric.auger@redhat.com>
Message-id: 20230717132641.764660-1-peter.maydell@linaro.org
Cc: qemu-stable@nongnu.org
Fourth RISC-V PR for 8.1
* Fix LMUL check to use VLEN
* Fix typo field in NUMA error_report
* check priv_ver before auto-enable zca/zcd/zcf
* Fix disas output of upper immediates
* tidy CPU firmware section
# -----BEGIN PGP SIGNATURE-----
#
# iQIzBAABCAAdFiEEaukCtqfKh31tZZKWr3yVEwxTgBMFAmS3akMACgkQr3yVEwxT
# gBPQ/BAArrieEkrRco3tIQJFZqTLfII28M0cYdwN+gjMAkL6RlauCh5yKkc+gsGy
# bhhpr0AE+EzrjKfJgdyMQe2ZH08WEpoAfJHAmLTSm2ktgIlnDAjyJtVksZ3FSwfG
# MRK3v0CChyOav3EfDZzK9jcaXeaSSfjCIG8JW3enoZxf2TnpoXlsCIQdRTnMw7Um
# C73BWoOGOfixFehywHBnkkAPo/nkQPofELrRKNTlefAIsH1RcgYw+s3IgCIuYxJN
# zCjM1y6ye1aiaQhKcNJiLoiP4Eq2R6vUuL8RKWkXqTP3QBZUqKMPnRVgI+W0qRAj
# 9DS+l37zMdxytovQ4gmIqnENT8ty9bholOtWM8nI54subJBplQhkRednG3RBFYjH
# hqbsakcHfE1lyyNI7WoBpO8UMtnOad6eBNmMOM48VduSdNuBZN3ksoRVomnJTlCY
# nq1ZdteywHEZ3uBqk3k/4yzKH+jLj0McPz5FswxsMIGScVjd6H8rMYmM95r1He4k
# YTJ8GwnOTBs1tFxOz5DaM3BVfq5hrzB0SbpDHMOdQHNXnqkyfvSd/QWeXfnY09Ux
# kbNvSpzjn7wWRSP7s4KMcTmas4oGtPS2dheREB/gmoC1ubrfuhbzduDNXJt+omuC
# GDcn9cpouyE/Vp/358PuEe1gW9GFMH0CbYBJ66P0hI/76iPfwLY=
# =MOsI
# -----END PGP SIGNATURE-----
# gpg: Signature made Wed 19 Jul 2023 05:44:51 BST
# gpg: using RSA key 6AE902B6A7CA877D6D659296AF7C95130C538013
# gpg: Good signature from "Alistair Francis <alistair@alistair23.me>" [unknown]
# gpg: WARNING: This key is not certified with a trusted signature!
# gpg: There is no indication that the signature belongs to the owner.
# Primary key fingerprint: 6AE9 02B6 A7CA 877D 6D65 9296 AF7C 9513 0C53 8013
* tag 'pull-riscv-to-apply-20230719-1' of https://github.com/alistair23/qemu:
target/riscv: Fix LMUL check to use VLEN
hw/riscv: Fix typo field in error_report
target/riscv/cpu.c: check priv_ver before auto-enable zca/zcd/zcf
riscv/disas: Fix disas output of upper immediates
docs/system/target-riscv.rst: tidy CPU firmware section
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
"smp.cpus" means the number of online CPUs and "smp.max_cpus" means the
total number of CPUs.
riscv_numa_get_default_cpu_node_id() checks "smp.cpus" and the
"available CPUs" description in the next error message also indicates
online CPUs.
So report "smp.cpus" in error_report() instand of "smp.max_cpus".
Since "smp.cpus" is "unsigned int", use "%u".
Signed-off-by: Zhao Liu <zhao1.liu@intel.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-Id: <20230718080712.503333-1-zhao1.liu@linux.intel.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
If QEMU is built with --without-default-devices, the s390-flic-kvm
device is missing and QEMU aborts when started with the KVM accelerator.
Make sure it's available by selecting S390_FLIC_KVM in Kconfig.
Consequently, this also fixes an abort in tests/qtest/migration-test.
Signed-off-by: Cédric Le Goater <clg@redhat.com>
Message-Id: <20230711151440.716822-1-clg@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Thomas Huth <thuth@redhat.com>
The primary guest scanout shows the booting screen right after reboot
but additional guest displays (i.e. max_ouptuts > 1) will keep displaying
the old frames until the guest virtio gpu driver gets initialized, which
could cause some confusion. A better way is to to replace the surface with
a place holder that tells the display is not active during the reset of
virtio-gpu device.
And to immediately update the surface with the place holder image after
the switch, displaychangelistener_gfx_switch needs to be called with
'update == TRUE' in dpy_gfx_replace_surface when the new surface is NULL.
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Marc-André Lureau <marcandre.lureau@redhat.com>
Cc: Vivek Kasireddy <vivek.kasireddy@intel.com>
Signed-off-by: Dongwon Kim <dongwon.kim@intel.com>
Acked-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-ID: <20230627224451.11739-1-dongwon.kim@intel.com>
Add a check in the bit-set operation to write the backstore
only if the affected bit is 0 before.
With this in place, there will be no need for callers to
do the checking in order to avoid unnecessary writes.
Signed-off-by: Tong Ho <tong.ho@amd.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Francisco Iglesias <frasse.iglesias@gmail.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
The ppi command line option for the TIS device on sysbus never worked
and caused an immediate segfault. Remove support for it since it also
needs support in the firmware and needs testing inside the VM.
Reproducer with the ppi=on option passed:
qemu-system-aarch64 \
-machine virt,gic-version=3 \
-m 4G \
-nographic -no-acpi \
-chardev socket,id=chrtpm,path=/tmp/mytpm1/swtpm-sock \
-tpmdev emulator,id=tpm0,chardev=chrtpm \
-device tpm-tis-device,tpmdev=tpm0,ppi=on
[...]
Segmentation fault (core dumped)
Signed-off-by: Stefan Berger <stefanb@linux.ibm.com>
Reviewed-by: Eric Auger <eric.auger@redhat.com>
Message-id: 20230713171955.149236-1-stefanb@linux.ibm.com
scsi_clear_unit_attention() now only handles REPORTED LUNS DATA HAS
CHANGED.
This only happens when we handle REPORT LUNS commands, so let's rename
the function in scsi_clear_reported_luns_changed() and call it only in
scsi_target_emulate_report_luns().
Suggested-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Stefano Garzarella <sgarzare@redhat.com>
Message-ID: <20230712134352.118655-4-sgarzare@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The previous commit moved the unit attention clearing when we create
the request. So now we can clean scsi_clear_unit_attention() to handle
only the case of the REPORT LUNS command: this is the only case in
which a UNIT ATTENTION is cleared without having been reported.
Suggested-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Stefano Garzarella <sgarzare@redhat.com>
Message-ID: <20230712134352.118655-3-sgarzare@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Commit 1880ad4f4e ("virtio-scsi: Batched prepare for cmd reqs") split
calls to scsi_req_new() and scsi_req_enqueue() in the virtio-scsi device.
No ill effects were observed until commit 8cc5583abe ("virtio-scsi: Send
"REPORTED LUNS CHANGED" sense data upon disk hotplug events") added a
unit attention that was easy to trigger with device hotplug and
hot-unplug.
Because the two calls were separated, all requests in the batch were
prepared calling scsi_req_new() to report a sense. The first one
submitted would report the right sense and reset it to NO_SENSE, while
the others reported CHECK_CONDITION with no sense data. This caused
SCSI errors in Linux.
To solve this issue, let's fetch the unit attention as early as possible
when we prepare the request, so that only the first request in the batch
will use the unit attention SCSIReqOps and the others will not report
CHECK CONDITION.
Fixes: 1880ad4f4e ("virtio-scsi: Batched prepare for cmd reqs")
Fixes: 8cc5583abe ("virtio-scsi: Send "REPORTED LUNS CHANGED" sense data upon disk hotplug events")
Reported-by: Thomas Huth <thuth@redhat.com>
Buglink: https://bugzilla.redhat.com/show_bug.cgi?id=2176702
Co-developed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Stefano Garzarella <sgarzare@redhat.com>
Message-ID: <20230712134352.118655-2-sgarzare@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
It is useful to extend the number of available PCIe devices to KVM guests
for passthrough scenarios and also to expose these models to a different
(big endian) architecture. Introduce a new config PCIE_DEVICES to select
models, Intel Ethernet adapters and one USB controller. These devices all
support MSI-X which is a requirement on s390x as legacy INTx are not
supported.
Cc: Matthew Rosato <mjrosato@linux.ibm.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Thomas Huth <thuth@redhat.com>
Signed-off-by: Cédric Le Goater <clg@redhat.com>
Message-ID: <20230712080146.839113-1-clg@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
According to the 82371FB documentation (82371FB.pdf, 2.3.9. BMIBA-BUS
MASTER INTERFACE BASE ADDRESS REGISTER, April 1997), the register is
32bit wide. To properly reset it to default values, all 32bit need to be
cleared. Bit #0 "Resource Type Indicator (RTE)" needs to be enabled.
The initial change wrote just the lower 8 bit, leaving parts of the "Bus
Master Interface Base Address" address at bit 15:4 unchanged.
Fixes: e6a71ae327 ("Add support for 82371FB (Step A1) and Improved support for 82371SB (Function 1)")
Signed-off-by: Olaf Hering <olaf@aepfle.de>
Reviewed-by: Bernhard Beschow <shentey@gmail.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-ID: <20230712074721.14728-1-olaf@aepfle.de>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The main loop thread can consume 100% CPU when using --device
virtio-blk-pci,iothread=<iothread>. ppoll() constantly returns but
reading virtqueue host notifiers fails with EAGAIN. The file descriptors
are stale and remain registered with the AioContext because of bugs in
the virtio-blk dataplane start/stop code.
The problem is that the dataplane start/stop code involves drain
operations, which call virtio_blk_drained_begin() and
virtio_blk_drained_end() at points where the host notifier is not
operational:
- In virtio_blk_data_plane_start(), blk_set_aio_context() drains after
vblk->dataplane_started has been set to true but the host notifier has
not been attached yet.
- In virtio_blk_data_plane_stop(), blk_drain() and blk_set_aio_context()
drain after the host notifier has already been detached but with
vblk->dataplane_started still set to true.
I would like to simplify ->ioeventfd_start/stop() to avoid interactions
with drain entirely, but couldn't find a way to do that. Instead, this
patch accepts the fragile nature of the code and reorders it so that
vblk->dataplane_started is false during drain operations. This way the
virtio_blk_drained_begin() and virtio_blk_drained_end() calls don't
touch the host notifier. The result is that
virtio_blk_data_plane_start() and virtio_blk_data_plane_stop() have
complete control over the host notifier and stale file descriptors are
no longer left in the AioContext.
This patch fixes the 100% CPU consumption in the main loop thread and
correctly moves host notifier processing to the IOThread.
Fixes: 1665d9326f ("virtio-blk: implement BlockDevOps->drained_begin()")
Reported-by: Lukáš Doktor <ldoktor@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Tested-by: Lukas Doktor <ldoktor@redhat.com>
Message-id: 20230704151527.193586-1-stefanha@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Let's support device unplug by forwarding the unplug_request_check()
callback to the virtio-mem device.
Further, disallow changing the requested-size once an unplug request is
pending.
Disallowing requested-size changes handles corner cases such as
(1) pausing the VM (2) requesting device unplug and (3) adjusting the
requested size. If the VM would plug memory (due to the requested size
change) before processing the unplug request, we would be in trouble.
Message-ID: <20230711153445.514112-8-david@redhat.com>
Tested-by: Mario Casquero <mcasquer@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
In many cases, blindly unplugging a virtio-mem device is problematic. We
can only safely remove a device once:
* The guest is not expecting to be able to read unplugged memory
(unplugged-inaccessible == on)
* The virtio-mem device does not have memory plugged (size == 0)
* The virtio-mem device does not have outstanding requests to the VM to
plug memory (requested-size == 0)
So let's add a callback to the virtio-mem device class to check for that.
We'll wire-up virtio-mem-pci next.
Message-ID: <20230711153445.514112-7-david@redhat.com>
Tested-by: Mario Casquero <mcasquer@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
While we fence unplug requests from the outside, the VM can still
trigger unplug of virtio based memory devices, for example, in Linux
doing on a virtio-mem-pci device:
# echo 0 > /sys/bus/pci/slots/3/power
While doing that is not really expected to work without harming the
guest OS (e.g., removing a virtio-mem device while it still provides
memory), let's make sure that we properly handle it on the QEMU side.
We'll add support for unplugging of virtio-mem devices in some
configurations next.
Message-ID: <20230711153445.514112-5-david@redhat.com>
Tested-by: Mario Casquero <mcasquer@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Let's factor out (un)plug handling, to be reused from arm/virt code.
Provide stubs for the case that CONFIG_VIRTIO_MD is not selected because
neither virtio-mem nor virtio-pmem is enabled. While this cannot
currently happen for x86, it will be possible for arm/virt.
Message-ID: <20230711153445.514112-3-david@redhat.com>
Tested-by: Mario Casquero <mcasquer@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
To achieve desired "x-ignore-shared" functionality, we should not
discard all RAM when realizing the device and not mess with
preallocation/postcopy when loading device state. In essence, we should
not touch RAM content.
As "x-ignore-shared" gets set after realizing the device, we cannot
rely on that. Let's simply skip discarding of RAM on incoming migration.
Note that virtio_mem_post_load() will call
virtio_mem_restore_unplugged() -- unless "x-ignore-shared" is set. So
once migration finished we'll have a consistent state.
The initial system reset will also not discard any RAM, because
virtio_mem_unplug_all() will not call virtio_mem_unplug_all() when no
memory is plugged (which is the case before loading the device state).
Note that something like VM templating -- see commit b17fbbe55c
("migration: allow private destination ram with x-ignore-shared") -- is
currently incompatible with virtio-mem and ram_block_discard_range() will
warn in case a private file mapping is supplied by virtio-mem.
For VM templating with virtio-mem, it makes more sense to either
(a) Create the template without the virtio-mem device and hotplug a
virtio-mem device to the new VM instances using proper own memory
backend.
(b) Use a virtio-mem device that doesn't provide any memory in the
template (requested-size=0) and use private anonymous memory.
Message-ID: <20230706075612.67404-5-david@redhat.com>
Tested-by: Mario Casquero <mcasquer@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
