virtio_pmem_flush() treats a NULL return from virtqueue_pop() as a fatal
error and calls virtio_error(), which puts the device into NEEDS_RESET.
However, virtqueue handlers can be invoked when no element is available,
so an empty queue should be handled as a benign no-op.
With a Linux guest this avoids spurious NEEDS_RESET and the resulting
-EIO propagation (e.g. EXT4 journal abort and remount-ro).
Signed-off-by: Li Chen <me@linux.beauty>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Message-Id: <20260106083859.380338-1-me@linux.beauty>
When `owner` == `mr`, `object_unparent` will crash:
object_unparent(mr) ->
object_property_del_child(mr, mr) ->
object_finalize_child_property(mr, name, mr) ->
object_unref(mr) ->
object_finalize(mr) ->
object_property_del_all(mr) ->
object_finalize_child_property(mr, name, mr) ->
object_unref(mr) ->
fail on g_assert(obj->ref > 0)
However, passing a different `owner` to `memory_region_init` does not
work. `memory_region_ref` has an optimization where it takes a ref
only on the owner. That means when flatviews are created, it does not
take a ref on the region and you can get a UAF from `flatview_destroy`
called from RCU.
The correct fix therefore is to use `NULL` as the name which will set
the `owner` but not the `parent` (which is still NULL). This allows us
to use `memory_region_ref` on itself while not having to rely on unparent
for cleanup.
Signed-off-by: Joelle van Dyne <j@getutm.app>
Reviewed-by: Akihiko Odaki <odaki@rsg.ci.i.u-tokyo.ac.jp>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Message-Id: <20260103214400.71694-1-j@getutm.app>
Synced virtio_spi.h from upstream Linux kernelusing
scripts/update-linux-headers.sh.
Source:
- Linux commit dcb6fa37fd7bc9c3d2b066329b0d27dedf8becaa (v6.18-rc3)
- Author: Linus Torvalds <torvalds@linux-foundation.org>
- Date: Sun Oct 26 15:59:49 2025 -0700
This update ensures QEMU's standard headers are consistent with the latest
virtio SPI definitions from Linux v6.18-rc3.
Signed-off-by: Haixu Cui <quic_haixcui@quicinc.com>
Reviewed-by: Manos Pitsidianakis <manos.pitsidianakis@linaro.org>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Message-Id: <20251216032122.1316684-2-quic_haixcui@quicinc.com>
When migrating, dst QEMU by default has SMRAM unlocked,
and since wmask is not migrated, the migrated value of
MCH_HOST_BRIDGE_F_SMBASE in config space fall to prey of
mch_update_smbase_smram()
...
if (pd->wmask[MCH_HOST_BRIDGE_F_SMBASE] == 0xff) {
*reg = 0x00;
and is getting cleared and leads to unlocked smram
on dst even if on source it's been locked.
As Andrey has pointed out [1], we should derive wmask
from config and not other way around.
Drop offending chunk and resync wmask based on MCH_HOST_BRIDGE_F_SMBASE
register value. That would preserve the register during
migration and set smram regions into corresponding state.
What that changes is:
that it would let guest write junk values in register
(with no apparent effect) until it's stumbles upon
reserved 0x1 [|] 0x2 values, at which point it
would be only possible to lock register and trigger
switch to SMRAM blackhole in CPU AS.
While at it, fix up test by removing junk discard before negotiation hunk.
PS2:
Instead of adding a dedicated post_load handler for it,
reuse mch_update->mch_update_smbase_smram call chain
that is called on write/reset/post_load to be consistent
with how we handle mch registers.
PS3:
for prosterity here is erro message Andrey got due to this bug:
qemu: vfio_container_dma_map(0x..., 0x0, 0xa0000, 0x....) = -22 (Invalid argument)
qemu: hardware error: vfio: DMA mapping failed, unable to continue
1) https://patchew.org/QEMU/20251203180851.6390-1-arbn@yandex-team.com/
Fixes: f404220e27 ("q35: implement 128K SMRAM at default SMBASE address")
Reported-by: Andrey Ryabinin <arbn@yandex-team.com>
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Andrey Ryabinin <arbn@yandex-team.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Message-Id: <20251211165454.288476-1-imammedo@redhat.com>
PCI_SRIOV_* are offsets into the SR-IOV capability, not into the PCI
config space. pcie_sriov_pf_exit() erroneously takes them as the latter,
which makes it read PCI_HEADER_TYPE and PCI_BIST when it tries to read
PCI_SRIOV_TOTAL_VF.
In many cases we're lucky enough that the PCI config space will be 0
there, so we just skip the whole for loop, but this isn't guaranteed.
For example, setting the multifunction bit on the PF and then doing a
'device_del' on it will get a larger number and cause a segfault.
Fix this and access the real PCI_SRIOV_* fields in the capability.
Cc: qemu-stable@nongnu.org
Fixes: 19e55471d4 ('pcie_sriov: Allow user to create SR-IOV device')
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Message-Id: <20251205145718.55136-1-kwolf@redhat.com>
Setting the sriov-pf property on devices that aren't PCI Express causes
an assertion failure:
$ qemu-system-x86_64 \
-blockdev null-co,node-name=null \
-blockdev null-co,node-name=null2 \
-device virtio-blk,drive=null,id=pf \
-device virtio-blk,sriov-pf=pf,drive=null2
qemu-system-x86_64: ../hw/pci/pcie.c:1062: void pcie_add_capability(PCIDevice *, uint16_t, uint8_t, uint16_t, uint16_t): Assertion `offset >= PCI_CONFIG_SPACE_SIZE' failed.
This is because proxy->last_pcie_cap_offset is only initialised to a
non-zero value in virtio_pci_realize() if it's a PCI Express device, and
then virtio_pci_device_plugged() still tries to use it.
To fix this, just skip the SR-IOV code for !pci_is_express(). Then the
next thing pci_qdev_realize() does is call pcie_sriov_register_device(),
which returns the appropriate error.
Cc: qemu-stable@nongnu.org
Fixes: d0c280d3fa ('pcie_sriov: Make a PCI device with user-created VF ARI-capable')
Signed-off-by: Kevin Wolf <kwolf@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>
Message-Id: <20251204172657.174391-1-kwolf@redhat.com>
In `virtio_add_resource` function, the UUID used as a key for
`g_hash_table_insert` was temporary, which could lead to
invalid lookups when accessed later. This patch ensures that
the UUID remains valid by duplicating it into a newly allocated
memory space. The value is then inserted into the hash table
with this persistent UUID key to ensure that the key stored in
the hash table remains valid as long as the hash table entry
exists.
Fixes: faefdba847 ("hw/display: introduce virtio-dmabuf")
Signed-off-by: Dorinda Bassey <dbassey@redhat.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Reviewed-by: Albert Esteve <aesteve@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Jim MacArthur <jim.macarthur@linaro.org>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Message-Id: <20251204162129.262745-1-dbassey@redhat.com>
The device is not able to generate indirect descriptors by itself, only
the driver can do it. And SVQ is already able to handle them by reusing
VirtQueue struct.
Shadow VirtQueue just transform the indirect descriptors to chained
buffers. This way we don't need to map a new buffer for the indirect
table and handle its exhaustion etc. As they are only used in control
plane and HW devices don't like indirect, we should not see significant
downsides with these.
Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
--
Tested functionally by forwarding the CVQ descriptors.
Acked-by: Jason Wang <jasowang@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Message-Id: <20251201125149.2151026-1-eperezma@redhat.com>
In ghes_record_cper_errors(), get_ghes_source_offsets() can return
a error initialized by error_setg(). Without bailing on this error,
it can call into the second error_setg() due to the unexpected value
returned from the read acknowledgement register. The second error_setg()
can trigger assert(*errp == NULL) in its callee error_setv(), which
isn't expected.
Bail early in ghes_record_cper_errors() when error is received from
get_ghes_source_offsets() to avoid the unexpected behavior.
Signed-off-by: Gavin Shan <gshan@redhat.com>
Reviewed-by: Jonathan Cameron <jonathan.cameron@huawei.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Message-Id: <20251201141803.2386129-5-gshan@redhat.com>
When stopping a vhost-vdpa device, only the first queue pair is marked as suspended,
while the remaining queues are not updated to the suspended state.
As a result, when stopping a multi-queue vhost-vdpa device,
the following error message will be printed.
qemu-system-x86_64:vhost VQ 2 ring restore failed: -1: Operation not permitted (1)
qemu-system-x86_64:vhost VQ 3 ring restore failed: -1: Operation not permitted (1)
So move v->suspended to v->shared, and then all the vhost_vdpa devices cannot
have different suspended states.
Fixes: 0bb302a996 ("vdpa: add vhost_vdpa_suspend")
Suggested-by: Eugenio Pérez <eperezma@redhat.com>
Acked-by: Eugenio Pérez <eperezma@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Wafer Xie <wafer@jaguarmicro.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>
Message-Id: <20251119132452.3117-1-wafer@jaguarmicro.com>
Just having a file descriptor is not enough to consider a memory region
public. If QEMU didn't map it as MAP_SHARED (in case of share=off), guest
writes to this region won't be visible to the vhost-user backend, thus
causing it to read all zeroes or garbage. Make sure we don't pass such
regions and include that to our definition of what a private region is.
Signed-off-by: Daniil Tatianin <d-tatianin@yandex-team.ru>
Fixes: 552b25229c ("vhost: Rework memslot filtering and fix "used_memslot" tracking")
Acked-by: Stefano Garzarella <sgarzare@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Message-Id: <20251113085842.323745-1-d-tatianin@yandex-team.ru>
Children are automatically unparented so manually unparenting is
unnecessary.
Worse, automatic unparenting happens before the instance_finalize()
callback of the parent gets called, so object_unparent() calls in
the callback will refer to objects that are already unparented, which
is semantically incorrect.
Signed-off-by: Akihiko Odaki <odaki@rsg.ci.i.u-tokyo.ac.jp>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Message-ID: <20251027-shpc-v1-1-00e9b20a355d@rsg.ci.i.u-tokyo.ac.jp>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Message-Id: <20251027-shpc-v1-1-00e9b20a355d@rsg.ci.i.u-tokyo.ac.jp>
Misc HW & memory API patches
- Add unit test for qemu_hexdump()
- Remove legacy native endianness API uses on the Alpha target
- Remove unused memory_region_init_rom_device_nomigrate()
- Fix use-after-free in NvmeNamespace "bootindex" suffix
- Correct documentation of SCSI Rotation Rate field
- Make iotlb_to_section() work with non-CPU AddressSpaces
- Reduce few monitor target-specific methods
# -----BEGIN PGP SIGNATURE-----
#
# iQIzBAABCAAdFiEE+qvnXhKRciHc/Wuy4+MsLN6twN4FAmmBFLoACgkQ4+MsLN6t
# wN4jrhAAjSb6hqXTTGWqVdyovdjCWmUOa3b8CZvTVycI2xTRA2HDqawnZSZrYwJS
# Wz20vyfylaoVI9OUibQOYBrpuZw1TBI4G7IeLBgdtzVW6oygIiHO1IlhtEENJtm7
# IXOMoU59ToQ85JWUZN6dwd6mbhBkR5roPJzJseyh/D5gf3t7/BE+jMXyAPNgls2A
# jnrgxw/1f1/DnIfAwbKaQfPuTVGerFmuicLhZsIrE9KFW2z8+acEPckAQKugsUqO
# vRrNz4m5ekNIijQfQwgU5XMhLDmiOfh0lRruecWOmCham3IKUaQ4YngGiArM8Q/p
# I9zcejrISWqlEYteu65OvB0ZOStEDGuCCu+sCDdvvFLTTfzkAEyyAKf4+rLdYS3L
# AWk1pm5YRQUb62VdNnFYXNVatDODixOSj2icoj4ojud2J9BsdHgvf7qPJ9IMgp2a
# eGnLj/F/38OUKRkj6kzzIUf5t/g9ZmC+JsdLcW1EleQLGiv4+yXqiH7c7swalXK5
# dZCjIrr92/iiemx9/kIi6oxkjlibzhDm9fF0NLoHhMOs9YGarbJx+J1YjRfQSnvh
# MehsqLV/D8KQvhy4VLsrEIPEx53g38SnXeHKBmyBXvZKeD2aqvzfaSilgHbrmFpu
# F6fD8FASkS+QVK0Fo+zOU7Zq3wuMebiDD7ioiE1JO8yg3KFS8jQ=
# =djFg
# -----END PGP SIGNATURE-----
# gpg: Signature made Tue 03 Feb 2026 07:18:50 AM AEST
# gpg: using RSA key FAABE75E12917221DCFD6BB2E3E32C2CDEADC0DE
# gpg: Good signature from "Philippe Mathieu-Daudé (F4BUG) <f4bug@amsat.org>" [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: FAAB E75E 1291 7221 DCFD 6BB2 E3E3 2C2C DEAD C0DE
* tag 'hw-misc-20260202' of https://github.com/philmd/qemu:
monitor: Reduce target-specific methods
monitor: Add hmp_cmds_for_target() helper
monitor: Reduce target-specific declarations
target/i386: Include missing 'svm.h' header in 'sev.h'
system/physmem: Remove the assertion of page-aligned section number
accel/tcg: Fix iotlb_to_section() for different AddressSpace
accel/tcg: Send the CPUTLBEntryFull struct into io_prepare()
hw/ide, scsi-disk: Fix typo on the rotation_rate documentation
hw/nvme: Fix bootindex suffix use-after-free
memory: Add internal memory_region_set_ops helper function
memory: Remove memory_region_init_rom_device_nomigrate()
target/alpha: Replace legacy ld_phys() -> address_space_ld()
configs/targets: Forbid Alpha to use legacy native endianness APIs
target/alpha: Inline translator_ldl()
target/alpha: Use explicit little-endian LD/ST API
tests/unit: add unit test for qemu_hexdump()
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
The following methods don't use target-specific code anymore:
- hmp_compare_cmd()
- monitor_register_hmp()
- monitor_register_hmp_info_hrt()
Move them to hmp.c which is target-agnostic, being built once.
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Dr. David Alan Gilbert <dave@treblig.org>
Message-Id: <20260129164039.58472-5-philmd@linaro.org>
HMPCommand arrays are filled with target-specific
commands, so defined in a target-specific unit.
Introduce the hmp_cmds_for_target() to allow
target-agnostic code to access the arrays.
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Dr. David Alan Gilbert <dave@treblig.org>
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Message-Id: <20260129164039.58472-4-philmd@linaro.org>
Some declarations do not depend on target-specific types,
move them out of "monitor/hmp-target.h" to "monitor/hmp.h".
Commit 409e9f7131 ("mos6522: add "info via" HMP command
for debugging") declared hmp_info_via() is declared twice.
Remove the one in "hw/misc/mos6522.h" otherwise we get:
In file included from ../hw/misc/mos6522.c:33:
include/monitor/hmp.h:43:6: error: redundant redeclaration of 'hmp_info_via' [-Werror=redundant-decls]
43 | void hmp_info_via(Monitor *mon, const QDict *qdict);
| ^~~~~~~~~~~~
In file included from ../hw/misc/mos6522.c:29:
include/hw/misc/mos6522.h:175:6: note: previous declaration of 'hmp_info_via' with type 'void(Monitor *, const QDict *)'
175 | void hmp_info_via(Monitor *mon, const QDict *qdict);
| ^~~~~~~~~~~~
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-Id: <20260129164039.58472-3-philmd@linaro.org>
"target/i386/sev.h" uses the vmcb_seg structure type, which
is defined in "target/i386/svm.h". Current builds succeed
because the files including "target/i386/sev.h" also include
"monitor/hmp-target.h", itself including "cpu.h" and finally
"target/i386/svm.h".
Include the latter, otherwise removing "cpu.h" from
"monitor/hmp-target.h" triggers:
../target/i386/sev.h:62:21: error: field has incomplete type 'struct vmcb_seg'
62 | struct vmcb_seg es;
| ^
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Zhao Liu <zhao1.liu@intel.com>
Message-Id: <20260129164039.58472-2-philmd@linaro.org>
'CPUTLBEntryFull.xlat_section' stores section_index in last 12 bits to
find the correct section when CPU access the IO region over the IOTLB.
However, section_index is only unique inside single AddressSpace. If
address space translation is over IOMMUMemoryRegion, it could return
section from other AddressSpace. 'iotlb_to_section()' API only finds the
sections from CPU's AddressSpace so that it couldn't find section in
other AddressSpace. Thus, using 'iotlb_to_section()' API will find the
wrong section and QEMU will have wrong load/store access.
To fix this bug of iotlb_to_section(), store complete MemoryRegionSection
pointer in CPUTLBEntryFull to replace the section_index in xlat_section.
Rename 'xlat_section' to 'xlat' as we remove last 12 bits section_index
inside. Also, since we directly use section pointer in the
CPUTLBEntryFull (full->section), we can remove the unused functions:
iotlb_to_section(), memory_region_section_get_iotlb().
This bug occurs only when
(1) IOMMUMemoryRegion is in the path of CPU access.
(2) IOMMUMemoryRegion returns different target_as and the section is in
the IO region.
Common IOMMU devices don't have this issue since they are only in the
path of DMA access. Currently, the bug only occurs when ARM MPC device
(hw/misc/tz-mpc.c) returns 'blocked_io_as' to emulate blocked access
handling. Upcoming RISC-V wgChecker [1] and IOPMP [2] devices are also
affected by this bug.
[1] RISC-V WG:
https://patchew.org/QEMU/20251021155548.584543-1-jim.shu@sifive.com/
[2] RISC-V IOPMP:
https://patchew.org/QEMU/20250312093735.1517740-1-ethan84@andestech.com/
Signed-off-by: Jim Shu <jim.shu@sifive.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Tested-by: Mark Burton <mburton@qti.qualcomm.com>
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Message-ID: <20260128152348.2095427-3-jim.shu@sifive.com>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Prefer the address_space_ld/st API over the legacy ld_phys()
because it allow checking for bus access fault.
Since we removed the last legacy uses of the legacy ldst_phys()
API, set the TARGET_NOT_USING_LEGACY_LDST_PHYS_API variable to
hide the legacy API to alpha binaries, avoiding further API uses
to creep in.
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20251224160040.88612-7-philmd@linaro.org>
All Alpha-related binaries are buildable without a single use
of the legacy "native endian" API. Unset the transitional
TARGET_USE_LEGACY_NATIVE_ENDIAN_API definition to forbid
further uses of the legacy API.
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20251224160040.88612-6-philmd@linaro.org>
The Alpha architecture uses little endianness. Directly
use the little-endian LD/ST API.
Mechanical change running:
$ for a in uw w l q; do \
sed -i -e "s/ld${a}_p(/ld${a}_le_p(/" \
$(git grep -wlE '(ld|st)u?[wlq]_p' target/alpha/);
done
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Manos Pitsidianakis <manos.pitsidianakis@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20251224160040.88612-2-philmd@linaro.org>
Test that the fix in commit 20aa05edc2 ("util/hexdump: fix
QEMU_HEXDUMP_LINE_WIDTH logic") make sense.
To not break compilation when we build without 'block', move
hexdump.c out of "if have_block" in meson.build.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@yandex-team.ru>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-Id: <20260202112826.38018-1-philmd@linaro.org>
Added a test that sends 32 NOP Out commands asynchronously. Since the CQ
has 31 entries by default, this tests the scenario where CQ processing
needs to wait for space to become available.
Additionally, added two minor fixes to existing tests:
* advance CQ head after reading from CQ
* initialize command descriptor slots bitmap in ufs_init()
Signed-off-by: Ilia Levi <ilia.levi@intel.com>
Acked-by: Fabiano Rosas <farosas@suse.de>
Reviewed-by: Jeuk Kim <jeuk20.kim@samsung.com>
Signed-off-by: Jeuk Kim <jeuk20.kim@samsung.com>
Currently, ufs_mcq_process_cq() writes to the CQ without checking whether
there is available space. This can cause CQ entries to be discarded and
overwritten. The solution is to stop writing when CQ is full and exert
backpressure on the affected SQs. This is similar to how NVMe CQs operate.
Signed-off-by: Ilia Levi <ilia.levi@intel.com>
Reviewed-by: Jeuk Kim <jeuk20.kim@samsung.com>
Signed-off-by: Jeuk Kim <jeuk20.kim@samsung.com>
