Xiaoyao Li [Tue, 12 May 2026 08:21:08 +0000 (16:21 +0800)]
i386/tdx: Add CPUID_24_0_EBX_AVX10_VL_MASK as supported
AVX10 depends on CPUID_24_0_EBX_AVX10_VL_MASK as defined in
feature_dependencies[]. Currently CPUID_24_0_EBX_AVX10_VL_MASK is always
not supported for TDX, so AVX10 cannot be exposed to TD guest.
The TDX virtualization type of these bits is "XFAM & CPUID_Enabled &
Native": their value is determined by XFAM[5:7], the AVX10 CPUID bit,
and the native hardware value.
For simplicity, add CPUID_24_0_EBX_AVX10_VL_MASK to tdx_xfam_deps[]
under the AVX512/XFAM dependency, without separately checking the AVX10
bit. It's safe because any invalid combination supplied by the user will
be caught by tdx_check_features().
Bit[3:0] of CPUID(0x1e,1).EAX alias the AMX CPUID bits from leaf 7.
Their TDX virtualization type is "CPUID_Enabled & Native": the value is
determined by the leaf-7 AMX bit they are aliased to and the native
hardware value.
These bits must be added to the TDX supported bits list so that they can
be enabled without triggering the forced-set warning. For simplicity,
mark them as supported whenever the corresponding AMX XFAM bit is
supported, rather than checking each aliased leaf-7 bit individually.
This reduces code complexity. Any platform that supports the AMX XFAM bit
but not these alias bits will still be handled correctly, since the TDX
module provides the real value via tdx_check_features().
Xiaoyao Li [Tue, 12 May 2026 08:21:06 +0000 (16:21 +0800)]
i386/tdx: Use .has_gpa field to check if the gpa is valid
When translating the QAPI type GuestPanicInformationTdx into its C
struct, the generated code provides a .has_gpa boolean field to indicate
whether the optional gpa field is present.
Replace the magic sentinel value -1ULL, previously used to signal "no
valid GPA", with the idiomatic .has_gpa field. This removes the
implicit sentinel coupling and makes the validity check self-documenting.
Signed-off-by: Xiaoyao Li <xiaoyao.li@intel.com> Reviewed-by: Daniel P. Berrangé <berrange@redhat.com> Reviewed-by: Zhenzhong Duan <zhenzhong.duan@intel.com> Link: https://lore.kernel.org/r/20260512082108.621596-2-xiaoyao.li@intel.com Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Xiaoyao Li [Tue, 12 May 2026 08:44:55 +0000 (16:44 +0800)]
qemu-options: Change memory-encryption to confidential-guest-support in the example
"confidential-guest-support" is the recommended property instead of
"memory-encryption". Switch to "confidential-guest-support" in the
example of sev-guest.
Xiaoyao Li [Tue, 12 May 2026 08:44:54 +0000 (16:44 +0800)]
i386/sev: Remove the example that references memory-encryption
"confidential-guest-support" is the recommended property to configure
sev for the machine, and amd-memory-encryption.rst has already switched
to use "confidential-guest-support" in the example.
Instead of changing "memory-encryption" to "confidential-guest-support"
in the comment of struct SevGuestState, just drop the example for
simplicity.
target/i386/mshv: use the register page to get registers
Change the mshv_load_regs to use the register page when it is mmapped
and is valid.
Otherwise use the existing logic that uses ioctls to fetch registers.
When retrieving the special registers, there are some registers that are
not present in the register page: TR, LDTR, GDTR, IDTR, CR2, APIC_BASE.
For this ones we still need to use ioctls to correctly fetch.
Magnus Kulke [Fri, 17 Apr 2026 10:56:07 +0000 (12:56 +0200)]
target/i386/mshv: migrate CET/SS MSRs
This change migrates the MSRs required for CET shadow stack and indirect
branch tracking. They are gated behind cet_ss_support || cet_ibt_support
mshv processor feature flags.
Magnus Kulke [Fri, 17 Apr 2026 10:56:05 +0000 (12:56 +0200)]
target/i386/mshv: migrate MTRR MSRs
This change roundtrips memory access/caching MSRs. The mapping scheme
is a bit more elaborate on these, so we have added a special handling
instead of individual entries in the MSR mapping table.
Magnus Kulke [Fri, 17 Apr 2026 10:56:04 +0000 (12:56 +0200)]
target/i386/mshv: migrate MSRs
In this change the we rewrite the existing MSR logic to make MSRs
migratable:
- we map them on existing QEMU fields in the CPU. A table and a macro
MSHV_ENV_FIELD is used to associate a HV register name to the their msr
index and their offset in the cpu state struct. The list is not
exhaustive and will be extended in follow-up commits.
- mshv_set/get_msrs() fns are called in the arch_load/store_vcpu_state()
fns. they use use generic registers ioctl's and map the input/output
via load/store_to/from_env() from/to the hv register content to the
cpu state representation.
- init_msrs() has been moved from mshv-vcpu to the msr source file
- we need to perform some filtering of MSR because before writing and
reading, because the hvcalls will fail if the partition doesn't
support a given MSRs.
- Some MSRs are partition-wide and so we will only write the to on the
BSP.
Magnus Kulke [Fri, 17 Apr 2026 10:56:16 +0000 (12:56 +0200)]
target/i386/mshv: reconstruct hflags after load
hflags is a cached bitmap derived from standard and special regs. We
want to reconstruct this state after regs and sregs have been read from
the hypervisor, similar to how it's one in other accelerators.
Magnus Kulke [Fri, 17 Apr 2026 10:56:15 +0000 (12:56 +0200)]
target/i386/mshv: migrate XSAVE state
We implement fn's that roundtrip XSAVE state in migration. We are using
the xsave_helper routines to move individual components from CPUX86State
to an xsave_buf and then we have to compact the buffer to XSAVEC format,
which is what the hypervisor expects.
And the same applies in the other direction for restoring state from the
hypervisor.
gitlab: greatly expand captured info about CI runner environment
To aid in debugging wierd CI failures we need greater information
about the CI runner environment. It is usually container based
and can have some unexpected characteristics that significantly
differ from a developer's local environment.
This captures the mount list, CPU model, memory information,
device node lists, kernel info, user identity and all environment
variables.
Since this information can get quite large it is not emitted
directly in the logs, rather it is exposed in job artifacts
under the "ci-runner-env" directory.
gitlab: pull before_script logic into .base_meson_job_template
This ensures that all jobs have the $JOBS env set and capture the
packages.txt content (where available) in their logs, and all use
the job section headers.
Stefan Hajnoczi [Thu, 25 Jun 2026 19:33:20 +0000 (15:33 -0400)]
tests/docker: bump emsdk-wasm64-cross zlib version to 1.3.2
Container image builds have started failing because zlib.net no longer
hosts the 1.3.1 tarball. Move to the 1.3.2 release from February 17,
2026.
$ make docker-image-emsdk-wasm64-cross
changing dir to build for make "docker-image-emsdk-wasm64-cross"...
make[1]: Entering directory 'qemu/build'
BUILD emsdk-wasm64-cross
...
xz: (stdin): File format not recognized
tar: Child returned status 1
tar: Error is not recoverable: exiting now
Error: building at STEP "RUN curl -Ls https://zlib.net/zlib-$ZLIB_VERSION.tar.xz | tar xJC /zlib --strip-components=1": while running runtime: exit status 2
make[1]: *** [qemu/tests/docker/Makefile.include:43: docker-image-emsdk-wasm64-cross] Error 2
make[1]: Leaving directory 'qemu/build'
make: *** [GNUmakefile:6: build] Error 2
Cc: Kohei Tokunaga <ktokunaga.mail@gmail.com> Cc: Brian Cain <brian.cain@oss.qualcomm.com> Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> Reviewed-by: Pierrick Bouvier <pierrick.bouvier@oss.qualcomm.com>
Message-id: 20260625193320.260312-1-stefanha@redhat.com Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Stefan Hajnoczi [Thu, 25 Jun 2026 19:34:25 +0000 (15:34 -0400)]
Merge tag 'next-pull-request' of https://gitlab.com/peterx/qemu into staging
Migration and mem pull request
- Maciej's patch to fix rare crash in VFIO multifd thread pool mgmt
- Peter's cleanup of @cpr-exec-command doc in migration.json
- Akihiko's patch to fix a TSAN warning on ram_list operations
- Bibo's migration-test coverage for loongarch
- Peter's update on a-b-boot image
- Marc-André's virtio-mem fix for CoCo
* tag 'next-pull-request' of https://gitlab.com/peterx/qemu:
system/physmem: make ram_block_discard_range() handle guest_memfd
tests: add unit tests for RamDiscardManager multi-source aggregation
system/memory: add RamDiscardManager reference counting and cleanup
system/physmem: destroy ram block attributes before RCU-deferred reclaim
system/memory: implement RamDiscardManager multi-source aggregation
system/ram-discard-manager: drop replay from source interface
virtio-mem: remove replay_populated/replay_discarded implementation
system/ram-discard-manager: implement replay via is_populated iteration
system/memory: constify section arguments
system/memory: move RamDiscardManager to separate compilation unit
system/memory: split RamDiscardManager into source and manager
migration/tests: Update a-b-boot images for all archs
tests/qtest/migration: Add migration test on loongarch
migration: Use OBJECT_DECLARE_SIMPLE_TYPE
system/memory: Remove MAX_PHYS_ADDR
system/physmem: Synchronize ram_list accesses
qapi/migration: Remove @cpr-exec-command doc in MigrationParameter
thread-pool: Allow at least 1 thread in thread_pool_adjust_max_threads_to_work()
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Magnus Kulke [Fri, 17 Apr 2026 10:56:18 +0000 (12:56 +0200)]
accel/mshv: enable dirty page tracking
This change introduces the functions required to perform dirty page
tracking to speed up migrations. We are using the sync, global_start,
and global_stop hooks.
The sync is implemented in batches.
Before we can disable the dirty page tracking we have to set all dirty bits.
Magnus Kulke [Fri, 17 Apr 2026 10:55:52 +0000 (12:55 +0200)]
accel/mshv: update s->irq_routes in add_msi_route
The irq_routes field of the state is populated with native mshv irq
route entries. The allocation logic is modelled after the KVM
implementation: we will always allocate a minumum of 64 entries and use
a bitmask to find/set/clear GSIs.
The old implementation of add_msi_routes will be removed in a subsequent
commit.
Magnus Kulke [Fri, 17 Apr 2026 10:55:50 +0000 (12:55 +0200)]
accel/accel-irq: add generic commit_route_changes
A generic accel_irqchip_commit_route_changes() fn has been introduced for
usage in the MSHV accelerator. The respective kvm_ fn can be removed
since we handle the commit op in a generic way.
Magnus Kulke [Fri, 17 Apr 2026 10:55:48 +0000 (12:55 +0200)]
accel/accel-irq: add AccelRouteChange abstraction
The accelerated irqchip routines use a record of changes to batch
changes when programming routes.
Currently this mechanism is coupled to the KVM accelerator, this change
introduces an abstraction that replaces KVMRouteChange and keeps a
pointer to an abstract AccelState instead of the concrete type,
converting the state where necessary.
This is done to further align the irqchip programming in the MSHV
accelerator with the existing KVM code in QEMU. Subsequent commits will
introduce AccelRouteChange to the MSHV accelerator code.
Magnus Kulke [Fri, 17 Apr 2026 10:55:47 +0000 (12:55 +0200)]
target/i386/mshv: impl init/load/store_vcpu_state
In migration we will handle more than registers, so we rework the
routines that were used to load & store CPU registers from/to the
hypervisor into more explicit init/load/store_vcpu_state() functions
that can be called from the appropriate hooks.
load/store_regs() still exists for the purpose of MMIO emulation, but it
will only address standard and special x86 registers.
Functions to retrieve FPU and XCR0 state from the hypervsisor have been
introduced.
MSR and APIC state covered are covered only as part of init_vcpu(). They
are not yet part of the load/store routines.
Magnus Kulke [Fri, 17 Apr 2026 10:55:45 +0000 (12:55 +0200)]
target/i386/mshv: use arch_load/store_reg fns
Improved consistency around the naming of load/store register fn's. this
is required since we want to roundtrip more registers in a migration
than what's currently required for MMIO emulation.
Magnus Kulke [Thu, 16 Apr 2026 12:11:16 +0000 (14:11 +0200)]
accel/mshv: disable la57 (5lvl paging)
This change disable la57 paging on the mshv hypervisor on both the
mshv processor feature bitmap and mask the cpuid feature leaf to the
guest.
Since the removal of hypervisor-assisted gva=>gpa translation in 1c85a4a3d7 we have seen MMIO errors in guests on la57-enabled hw. We
will have to investigate and test this further.
Signed-off-by: Magnus Kulke <magnuskulke@linux.microsoft.com> Reviewed-by: Anirudh Rayabharam (Microsoft) <anirudh@anirudhrb.com> Reviewed-by: Wei Liu <wei.liu@kernel.org> Reviewed-by: Doru Blânzeanu <dblanzeanu@linux.microsoft.com> Link: https://lore.kernel.org/r/20260416121116.527927-10-magnuskulke@linux.microsoft.com Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Magnus Kulke [Thu, 16 Apr 2026 12:11:15 +0000 (14:11 +0200)]
target/i386/mshv: use hv-provided [0xD,1+2].EBX
We cannot statically set the responses for CPUID[0xD,{1,2}].EBX, b/c
those are dynamic, dependent on which features the guest enables.
Hence we mask EBX when registering answers for those subleaves at the
hypervisor, which will result in the hypervisor providing us answers,
considering XCR0 and XSS.
The reported size now reflects the field masks properly (without the
mask they were 576 and 10728, which is wrong):
$ cpuid -l 0xd -s 0
CPU 0:
XSAVE features (0xd/0):
XCR0 valid bit field mask = 0x00000000000600e7
...
bytes required by fields in XCR0 = 0x00002b00 (11008)
bytes required by XSAVE/XRSTOR area = 0x00002b00 (11008)
$ cpuid -l 0xd -s 1
CPU 0:
XSAVE features (0xd/1):
...
SAVE area size in bytes = 0x000029c0 (10688)
IA32_XSS lower 32 bits valid bit field mask = 0x00001800
IA32_XSS upper 32 bits valid bit field mask = 0x00000000
Magnus Kulke [Thu, 16 Apr 2026 12:11:13 +0000 (14:11 +0200)]
target/i386: query mshv accel for supported cpuids
We implement mshv_get_supported_cpuid() and invoke it in
x86_cpu_get_supported_feature_word() retrieve the cpu features that the
host is supporting. Initially we mask the virtualization capabilitities
potentially we might need to mask more in the future.
Magnus Kulke [Thu, 16 Apr 2026 12:11:11 +0000 (14:11 +0200)]
target/i386/mshv: change cpuid mask to UINT32_MAX
The current implementation would only override the feature bits that are
enabled, however we also want to consider disabled features, hence all
bits are set on the masks in the hypercall argument.
Magnus Kulke [Thu, 16 Apr 2026 12:11:10 +0000 (14:11 +0200)]
target/i386/mshv: fix various cpuid traversal bugs
- Hardcoded max_leaf was not accurate. We query leaf 0x0 and 0x80000000 to
get the actual max leaves
- On all 0 zeroes on leaf 0x0d, we register 0-63 subleaves with zeros
indicating XSAVE is disabled
- Subleaf 0 was hardcoded, so the Hypervisor returned defaults for other
subleaves
- Subleaf 0 was hardcoded, so we were passing 0 instead of actual subleaf
when adding entries. We now pass the correct subleaf value to
add_cpuid_entry()
- Leaves 0x04,0x07,0d,0f,10 weren't marked as subleaf-specific
Magnus Kulke [Thu, 16 Apr 2026 12:11:08 +0000 (14:11 +0200)]
accel/mshv: use mshv_create_partition_v2 payload
When using the extended request format we can instruct the hypervisor to
provision enough space for requested XSAVE features. This is required
for supporting QEMU models provided via the -cpu flag properly.
Signed-off-by: Magnus Kulke <magnuskulke@linux.microsoft.com> Acked-by: Wei Liu <wei.liu@kernel.org> Reviewed-by: Anirudh Rayabharam (Microsoft) <anirudh@anirudhrb.com> Reviewed-by: Doru Blânzeanu <dblanzeanu@linux.microsoft.com> Link: https://lore.kernel.org/r/20260416121116.527927-2-magnuskulke@linux.microsoft.com Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
When bit 3 of the VGA Attribute Mode Control register is set, attribute
bit 7 switches from selecting bright background colors to enabling
character blink.
Implement this by tracking a separate blink phase timer that toggles
every 32 frames (matching real VGA hardware frame counter bit 5 @60hz),
and rendering blinking characters by replacing their foreground with
background during the off phase.
As with cursor, no VMState migration of the fields, as they are
transient display-side states.
Related to: https://gitlab.com/qemu-project/qemu/-/work_items/1585 Acked-by: Gerd Hoffmann <kraxel@redhat.com> Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-ID: <20260525081944.1494798-1-marcandre.lureau@redhat.com>
Add a qtest that verifies display consoles are dynamically added and
removed over D-Bus when a bochs-display device is hotplugged and
unplugged on a q35 machine.
The test plugs device_add a bochs-display, waits for the DEVICE_ADDED
QMP event, and checks that the D-Bus VM interface reports a second
console. It then device_del it, forces a system reset
(q35 removal is ACPI-based and needs guest cooperation qtest cannot
provide), waits for DEVICE_DELETED, and checks the console count again.
Subscribe to QemuConsoleEvent notifications to dynamically add and
remove D-Bus display consoles. This mirrors the GTK backend's
handling added in the previous commits.
ui/console: unregister console from QOM tree on close
Call object_unparent() in graphic_console_close() to remove the
console from /backend/console[N]. This drops the QOM tree
reference, while the initial object_new() reference keeps the
console alive for potential reuse.
When graphic_console_init() reuses a closed console, re-register
it in the QOM tree.
ui/console: register console in QOM tree dynamically
Consoles created after init_displaystate() (e.g. hotplugged
display devices) were never added to the /backend/console[N]
QOM tree. Extract qemu_console_add_to_qom() and call it from
qemu_console_register() when the display is already initialized.
Register a console notifier so the GTK display dynamically creates and
destroys VirtualConsole tabs when graphic consoles are added or removed
at runtime (e.g. vfio-pci with display=on hotplug).
Add skips consoles that already have a VC binding, remove skips unknown
consoles.
ui/gtk: centralize console menu and shortcut management
Replace the per-console gd_vc_menu_init() with gd_rebuild_vc_menu()
that tears down and rebuilds all console radio menu items and
Ctrl+Alt+N accelerators at once. This is called from initialization
and whenever consoles are detached or reattached.
Shortcuts now skip detached (windowed) consoles, so they always map
to reachable tabs. Rename gd_vc_gfx_init() to add_gfx_console()
and simplify the init function signatures now that menu creation is
decoupled.
ui/gtk: fix tab re-insertion order on window close
Add gd_vc_notebook_pos() which computes the correct notebook position
for a console by counting only non-detached (non-windowed) tabs before
it. Use it in gd_tab_window_close() so a re-attached tab is inserted
at its logical position rather than appended at the end.
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com> Reviewed-by: Akihiko Odaki <odaki@rsg.ci.i.u-tokyo.ac.jp> Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-ID: <20260623-b4-ui-v4-28-4656aec3398d@redhat.com>
ui/gtk: move global display settings out of per-console init
Move zoom_to_fit, keep_aspect_ratio and touch_slots initialisation
from gd_vc_gfx_init() to gd_create_menu_view(). These are global
display settings that should be set once after all consoles are
created, not repeated on every per-console init (where the last
iteration's values silently win).
Fire CONSOLE_REMOVED in char_vc_finalize() before dropping the console
reference, so the console is still in a valid state when listeners
handle the event. Also fixes a console object leak by adding the
missing object_unref().
ui/console: fire console ADDED/REMOVED notifications
Fire CONSOLE_ADDED at the end of graphic_console_init() and
CONSOLE_REMOVED at the start of graphic_console_close(), so display
backends can react to console hotplug/unplug events.
REMOVED fires before the device link is cleared and before the
placeholder surface swap, so handlers can unregister their DCL while
the console is still in a known state.
Add gtk_display_cleanup() to properly tear down GTK display state:
remove console and mouse notifiers, unregister clipboard peer,
destroy the main window and virtual consoles.
Move cbowner release from QemuCocoaAppController -dealloc to
cocoa_display_cleanup(), since cbowner is allocated in
cocoa_display_init() and cleanup is the symmetric teardown path.
Add egl_cleanup() to tear down the EGL render node context, GBM device,
and EGL display. Add egl_headless_cleanup() to unregister listeners,
destroy framebuffers, and free per-console state tracked via a new
GPtrArray.
Replace the atexit() handler with a proper cleanup callback. Extend
sdl_cleanup() to unregister display listeners, free keyboard state,
destroy windows, and clean up all cursor resources.
Replace the atexit() handler with a proper cleanup callback. The new
curses_cleanup() unregisters the display listener, destroy & free the
allocated resources.
Add a cleanup callback to QemuDisplay and a qemu_display_cleanup()
function that iterates all registered display types and calls their
cleanup handler. Wire it into qemu_cleanup() in runstate.c, replacing
the ad-hoc vnc_cleanup() call.
This provides a structured alternative to atexit() handlers, giving
deterministic teardown ordering and making resource leaks visible to
sanitizers.
The cleanup should happen before user_creatable_cleanup(), since some
display have weak user-creatable references to cleanup before.
ui/console: init gl_unblock_timer in qemu_console_init
Move gl_unblock_timer allocation from graphic_console_init() to
qemu_console_init(), similar to what was done in commit cfde05d15b
("ui/console: allocate ui_timer in QemuConsole").
This fixes leaking timers on console recycling.
Fixes: a9b1e471e17 ("ui: add a gl-unblock warning timer") Reviewed-by: Daniel P. Berrangé <berrange@redhat.com> Reviewed-by: Akihiko Odaki <odaki@rsg.ci.i.u-tokyo.ac.jp> Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-ID: <20260623-b4-ui-v4-12-4656aec3398d@redhat.com>
vfio/pci: close display console during unrealize, not finalize
The QemuGraphicConsole holds a strong QOM link back to the device via
its "device" property (OBJ_PROP_LINK_STRONG). When graphic_console_close()
is only called from vfio_display_finalize() during object finalize, this
creates a ref-cycle deadlock: the device can't reach refcount 0 because
the console holds a strong ref, but the console's ref is only dropped by
graphic_console_close() which runs inside finalize.
Split the display teardown into two phases:
- vfio_display_exit(): called during unrealize (vfio_exitfn), closes
the graphic console to break the ref cycle, and removes display
region subregions while the parent memory regions are still alive.
- vfio_display_finalize(): remains in finalize (vfio_pci_put_device),
frees display region memory, dmabuf, and edid resources. The region
memory contains QOM child objects (MemoryRegions) that must stay
alive until QOM finalization has processed them.
irq: add per-IRQ observer to fix qemu_irq_intercept_in leak
qemu_irq_intercept_in() saves original IRQ handlers by allocating
new QOM objects, which are never freed. On a PC machine, this leaks
IRQ objects (one per IOAPIC pin) on every qtest run.
Rather than tracking allocations to free later, avoid them: add an
"observer" field to IRQState, called by qemu_set_irq() after the
real handler. Interception sets the observer instead of rewriting
handlers, so there's nothing to save and nothing to leak.
Fix qemu_notirq() to route through qemu_set_irq() so inverted IRQs
trigger observers too. Drop the LSan suppression.
The GitLab macOS runners are only available in Premium and Ultimate
tier projects. This permits their use in QEMU upstream via the OSS
Program membership perk, but this does not extend to forks of QEMU.
The macOS jobs thus need to be disabled in forks, otherwise all
pipelines will immediately report a failure due to inability to
access a runner.
gitlab: ensure "check-XXX' jobs capture functional test logs
A small subset of functional tests are run by default with
'make check', and so run in the context of 'check-XXX' CI
jobs, rather than 'functional-XXX' CI jobs. Thus we need
to capture the functional test logs unconditionally for all
test jobs.
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com> Reviewed-by: Pierrick Bouvier <pierrick.bouvier@oss.qualcomm.com>
Message-ID: <20260610121254.2870259-1-berrange@redhat.com>
Message-ID: <20260619155657.944220-6-alex.bennee@linaro.org> Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Alex Bennée [Fri, 19 Jun 2026 15:56:53 +0000 (16:56 +0100)]
tests/functional: tell pylint not to check c-modules
To fix:
qemu-test.test_pylint "/home/alex/lsrc/qemu.git/tests/functional/arm/test_integratorcp.py:83: I1101: Module 'cv2' has no 'imread' member, but source is unavailable. Consider adding this module to extension-pkg-allow-list if you want to perform analysis based on run-time introspection of living objects. (c-extension-no-member)"
Manually running python3 showed I could indeed import cv2 and call
those functions. Rather than allowing pylint to introspect lets just
tell it to skip c modules.
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com> Reviewed-by: Thomas Huth <thuth@redhat.com>
Message-ID: <20260619155657.944220-5-alex.bennee@linaro.org> Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Alex Bennée [Fri, 19 Jun 2026 15:56:52 +0000 (16:56 +0100)]
python/qemu: dump a developer friendly version of cmdline to logs
Now we have the arguments nicely split up we can make _console_args a
function call and present a slightly different version to the logs to
save developers manually hacking the command line up.
Reviewed-by: Philippe Mathieu-Daudé <philmd@oss.qualcomm.com> Reviewed-by: Thomas Huth <thuth@redhat.com>
Message-ID: <20260619155657.944220-4-alex.bennee@linaro.org> Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
system/physmem: make ram_block_discard_range() handle guest_memfd
Most callers of ram_block_discard_range() want to discard both the
shared and guest_memfd backing. Only kvm_convert_memory() intentionally
discards a single plane during private/shared conversions.
Rename the current implementation to ram_block_discard_shared_range()
and make ram_block_discard_range() a composite that also discards
guest_memfd when present (rb->guest_memfd >= 0). This ensures callers
like virtio-mem, virtio-balloon, hv-balloon, migration.. reclaim
private pages on discard.
Update kvm_convert_memory() to use the plane-specific
ram_block_discard_shared_range() since it only needs to discard
the shared backing when converting to private.
Likewise, after TDVF image copy, use ram_block_discard_shared_range().
system/memory: add RamDiscardManager reference counting and cleanup
Listeners now hold a reference to the RamDiscardManager, ensuring it
stays alive while listeners are registered. The RDM is eagerly freed
when the last source and listener are removed, and also unreffed during
MemoryRegion finalization as a safety net.
This completes the TODO left in the previous commit and prevents both
use-after-free and memory leaks of the RamDiscardManager.
system/physmem: destroy ram block attributes before RCU-deferred reclaim
ram_block_attributes_destroy() was called from reclaim_ramblock(), which
runs as an RCU callback deferred by call_rcu().
However,when the RamDiscardManager is finalized, it will assert that its
source_list is empty in the next commit. Since the RCU callback hasn't
run yet, the source added by ram_block_attributes_create() is still
attached.
Move ram_block_attributes_destroy() into qemu_ram_free() so the source
is removed synchronously. This is safe because qemu_ram_free() during
shutdown runs after pause_all_vcpus(), so no vCPU thread can
concurrently access the attributes via kvm_convert_memory().
Reviewed-by: Peter Xu <peterx@redhat.com> Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com> Link: https://lore.kernel.org/r/20260604-rdm5-v5-8-5768e6a0943d@redhat.com
[peterx: rebase on top of qemu_ram_free() change] Signed-off-by: Peter Xu <peterx@redhat.com>
projects / thirdparty / qemu.git / log
