Virtio v1.2 says that if CONFIG_WCE is negotiated
but FLUSH is not, the device must initialize writeback to 0.
It also says that if CONFIG_WCE was not negotiated but FLUSH
was, the driver should assume presence of a writeback cache.
Introduce a pure is_writeback_enabled helper and a
set_writeback_mode helper. This makes the two call flows
explicit:
* write_config resolves the guest requested mode against the
negotiated features before storing it back
* activate starts from the default writeback preference and then
resolves it against the negotiated features
* reset restores the initial writeback state
This keeps the config space value and the runtime writeback flag
in sync and makes the spec driven fallback easier to follow.
Signed-off-by: Anatol Belski <anbelski@linux.microsoft.com>
Refactor PciDevice::allocate_bars trait and all implementations
to take &mut SystemAllocator instead of &Arc<Mutex<SystemAllocator>>,
removing double indirection.
The caller in device_manager.rs now acquires the lock before
calling allocate_bars.
Signed-off-by: Chinmoy <daschinmoyy21@gmail.com>
The control queue handler passed the total length of all
descriptors (header + data + status) as used_len to add_used.
Per virtio spec section 2.6.8, used_len must only count bytes
written to device-writable descriptors. The device only writes
the 1-byte status/ack field.
Windows NetKVM >= 0.1.285 strictly checks this value and calls
NdisMRemoveMiniport when len != sizeof(virtio_net_ctrl_ack),
removing the network adapter immediately after activation.
Signed-off-by: CMGS <ilskdw@gmail.com>
Windows NetKVM driver (>= 0.1.271) issues unsupported command classes
even when they are not even advertised.
According to the Virtio 1.2 specification:
RX, VLAN, and ANNOUNCE control paths are only meaningful when their
corresponding features are negotiated in sections 5.1.3.1, 5.1.6.5.1.2,
5.1.6.5.2.2, and 5.1.6.5.4.1.
RX and VLAN are explicitly described as best-effort in sections
5.1.6.5.1 and 5.1.6.5.3.
Instead of returning an error to the guest, return success to the guest.
Fixes: #7925
Signed-off-by: Wei Liu <liuwe@microsoft.com>
Change the return type of RawFileAsyncAio::new() from
std::io::Result<Self> to BlockResult<Self>, wrapping
internal errors from EventFd::new() and IoContext::new()
in BlockError with DiskFileError::NewAsyncIo.
This simplifies the caller in AsyncDiskFile::new_async_io()
which no longer needs its own error mapping.
Signed-off-by: Muminul Islam <muislam@microsoft.com>
Remove the old async_io::DiskFile trait implementation from
RawFileDiskAio, now that the new disk_file trait hierarchy
is fully implemented.
Clean up unused imports: DiskFile and DiskFileResult from
crate::async_io.
Signed-off-by: Muminul Islam <muislam@microsoft.com>
Add disk_file::AsyncDiskFile trait implementation for
RawFileDiskAio with try_clone() and new_async_io() methods.
try_clone() duplicates the underlying file descriptor and
wraps it in a new RawFileDiskAio. new_async_io() creates a
RawFileAsyncAio (Linux AIO) backend, wrapping errors in
BlockError instead of DiskFileError.
Signed-off-by: Muminul Islam <muislam@microsoft.com>
Add empty disk_file::DiskFile impl for RawFileDiskAio.
This marker supertrait requires DiskSize + Geometry + Sync,
all of which are now satisfied.
Signed-off-by: Muminul Islam <muislam@microsoft.com>
Add disk_file::Resizable trait implementation for
RawFileDiskAio. Calls file.set_len(size) and wraps the
I/O error in BlockError on failure.
Signed-off-by: Muminul Islam <muislam@microsoft.com>
Add disk_file::SparseCapable trait implementation for
RawFileDiskAio. Delegates to probe_sparse_support() to
detect whether the underlying file supports hole-punching.
Signed-off-by: Muminul Islam <muislam@microsoft.com>
Add disk_file::Geometry trait implementation for
RawFileDiskAio. Probes disk topology from the file,
falling back to defaults on failure. Takes &self instead
of &mut self and uses unwrap_or_else for cleaner error
handling.
Signed-off-by: Muminul Islam <muislam@microsoft.com>
Add disk_file::DiskFd trait implementation for
RawFileDiskAio. Delegates to file.as_raw_fd() via
BorrowedDiskFd, taking &self instead of &mut self.
Signed-off-by: Muminul Islam <muislam@microsoft.com>
Add disk_file::PhysicalSize trait implementation for
RawFileDiskAio. Returns the physical size from
query_device_size wrapped in BlockError on failure,
consistent with the DiskSize impl.
Signed-off-by: Muminul Islam <muislam@microsoft.com>
Add disk_file::DiskSize trait implementation for
RawFileDiskAio using BlockError and BlockResult. Takes
&self instead of &mut self.
Add BlockError, BlockErrorKind, BlockResult, and disk_file
imports needed by this and subsequent trait impls.
Signed-off-by: Muminul Islam <muislam@microsoft.com>
Add #[derive(Debug)] to RawFileDiskAio. This is required
by the new disk_file traits which have Send + Debug bounds.
Signed-off-by: Muminul Islam <muislam@microsoft.com>
Add a ramp time before measuring the block rate limiter tests so
both the single device and group workloads are measured after
warm up to make the measurements less sensitive to startup
transients.
Signed-off-by: Anatol Belski <anbelski@linux.microsoft.com>
Increase the group block refill time from 100 ms to 1000 ms and
scale the shared bucket sizes to preserve the target rate.
Set the one time burst to 0 to avoid transient overshoot and use
the shared block runtime constant directly in the group path.
Signed-off-by: Anatol Belski <anbelski@linux.microsoft.com>
The rate limiter token bucket has a fixed 100 ms cool down time
that pauses I/O whenever the bucket empties. With a 100 ms refill
time, the actual throughput drops to roughly half of the target
rate and causes the tests to miss their target.
Increase the net and single block refill time from 100 ms to
1000 ms and scale the bucket sizes by 10x to preserve the target
rate. Set the one time burst to 0 to avoid overshooting the upper
bound, and raise the runtime constants from 10 s to 20 s for
steadier measurements.
Signed-off-by: Anatol Belski <anbelski@linux.microsoft.com>
Introduce named constants for the net and single block rate limiter
test runtimes and use them directly at the call sites.
Signed-off-by: Anatol Belski <anbelski@linux.microsoft.com>
Unfortunately with a single ARM64 machine this has now become a
bottleneck for landing PRs. Copy the methodology we use for existing
jobs that we only run on the MQ by creating dummy jobs that run on the
GH hosted runner (ubuntu-latest) allowing the PR to transition into
the MQ by passing the required checks.
Signed-off-by: Rob Bradford <rbradford@meta.com>
Now that the VhostUserCommon::shutdown implementation has been filled
out to support migration it can also be used for the drop
implementations in the vhost-user devices.
It's worth noting that the call to wait_for_epoll_threads() was a no-op
as those threads are only configured on conventional virtio devices.
Signed-off-by: Rob Bradford <rbradford@meta.com>
Advertising support for this virtio feature is required to enable
support for migration. (Along with the LOG_SHMFD protocol feature.)
Signed-off-by: Rob Bradford <rbradford@meta.com>
If the epoll thread is paused, which would be expected as a part of live
migration/snapshot-restore unpause the thread so that it can receive the
kill event. This mirrors the reset() behaviour of virtio devices. It is
important here so as to close the connection with the vhost-user-backend
to allow same host and --local migration and since after getting the
device state the vhost-user backend should no longer be used.
As a result of this change we can do --local and same-host migration
with virtio-fs.
Signed-off-by: Rob Bradford <rbradford@meta.com>
This is used by all devices so it can be part of the common state.
Moving it simplifies the code and simplifies some future improvements
around shutdown for migration.
Signed-off-by: Rob Bradford <rbradford@meta.com>
Since vhost-user devices are always virtio devices it makes sense to
structure this struct inside the VhostUserCommon struct. This then also
makes some of the methods on VhostUserCommon cleaner since they can now
act directly on the common virtio bits (e.g. for kill_evt)
Signed-off-by: Rob Bradford <rbradford@meta.com>
OVMF can reprogram PCI BARs while memory space decoding is disabled.
Cloud Hypervisor defers the corresponding BAR move in
`pending_bar_reprogram` until the PCI command register enables Memory
Space again.
That deferred state was not part of `PciConfigurationState`. A
snapshot taken in that window restored the new BAR values in PCI
config space, but lost the pending BAR relocation needed to update the
VMM-side BAR mapping.
The restore logs show guest MMIO accesses to the reprogrammed BAR
addresses `0xc0000000`, `0x100000000`, and `0x100080000` hitting
unregistered addresses. The firmware serial output shows OVMF
assigning those same BAR addresses during PCI resource allocation,
then reaching BDS, finding the mass-storage device, and failing to
boot from it.
Serialize and restore `pending_bar_reprogram` so deferred BAR moves
survive snapshot and restore.
Co-authored-by: Thomas Prescher <thomas.prescher@cyberus-technology.de>
Co-authored-by: Julian Schindel <julian.schindel@cyberus-technology.de>
On-behalf-of: SAP leander.kohler@sap.com
Signed-off-by: Leander Kohler <leander.kohler@cyberus-technology.de>
Previously, GSIs were eagerly allocated for all MSI-X vectors a device
advertises (i.e. the maximum the device can support). This can easily
exhaust KVM_MAX_IRQ_ROUTES (4096) with modern NVMe devices that support
up to 2048 MSI-X vectors.
Defer GSI allocation to the first time an interrupt vector is
unmasked. The EventFd is still created eagerly since external
components (e.g. VFIO) need it at device init time.
Signed-off-by: Bo Chen <bchen@crusoe.ai>
Introduce ExecuteError::UnsupportedFlags to carry both the
request type and the rejected flags value, replacing the
generic ExecuteError::Unsupported at discard and write zeroes
flag validation sites. This provides structured context for
debugging without changing the returned VIRTIO_BLK_S_UNSUPP
status.
Signed-off-by: Anatol Belski <anbelski@linux.microsoft.com>
PR #7852 fixed the missing VirtioBlockConfig fields but did not
change the feature advertisement logic. The condition
`sparse || disk_image.supports_zero_flag()` causes qcow2 to
advertise DISCARD even with sparse=false, because qcow2 can
mark clusters as zero (supports_zero_flag() returns true).
Windows viostor BSODs (DRIVER_IRQL_NOT_LESS_OR_EQUAL) when
DISCARD is advertised on qcow2 backends, making sparse=off
ineffective as a workaround for qcow2 images.
Restrict DISCARD to explicit sparse=true only. WRITE_ZEROES
remains available for all sparse-capable backends.
Fixes#7849
Signed-off-by: CMGS <ilskdw@gmail.com>
The virtio spec v1.2 in 5.2.6.2 requires that the device
MUST return VIRTIO_BLK_S_UNSUPP for write zeroes commands
if any unknown flag is set.
Add an early check that rejects requests with reserved flag
bits set by returning VIRTIO_BLK_S_UNSUPP via the existing
ExecuteError::Unsupported variant.
Signed-off-by: Anatol Belski <anbelski@linux.microsoft.com>
The virtio spec v1.2 in 5.2.6.2 requires that the device
MUST return VIRTIO_BLK_S_UNSUPP for discard commands if the
unmap flag is set or if any unknown flag is set.
The discard handler was not reading the flags field at all,
silently accepting requests with arbitrary flags. Read and
validate the flags, rejecting any non-zero value with
VIRTIO_BLK_S_UNSUPP.
Signed-off-by: Anatol Belski <anbelski@linux.microsoft.com>
Previously, KVM_KVMCLOCK_CTRL was skipped when kvm_hyperv=on because
Windows does not use pvclock directly. However, KVM internally uses
pvclock data structures as the basis for computing the Hyper-V
Reference TSC page parameters. Not calling KVM_KVMCLOCK_CTRL means
there is no mechanism to signal time discontinuity to Windows guests
after pause/resume, contributing to multi-minute hangs.
Remove the kvm_hyperv guard so all guests receive the clock-paused
notification.
Signed-off-by: CMGS <ilskdw@gmail.com>
The queue_msix_vector register (offset 0x1a in virtio PCI common
config) was indexed into the msix_queues Vec using the guest-controlled
queue_select value without bounds checking. A malicious guest can set
queue_select to any u16 value via offset 0x16, then read or write
offset 0x1a to trigger an out-of-bounds panic, crashing the VMM.
Replace direct Vec indexing with .get()/.get_mut() for bounds-checked
access, returning VIRTQ_MSI_NO_VECTOR (0xFFFF) on OOB reads to match
the virtio "no vector" sentinel. Add a regression test that sets
queue_select to 0xFFFF and exercises both the read and write paths.
AI/LLM disclosure: this patch was co-authored with Claude Code.
Fixes#7917
Signed-off-by: Tobias Kässer <t.kaesser@gmail.com>
When booting an SEV-SNP guest VM using IGVM with -pvalidate_opt 1 (lazy
page acceptance), the guest kernel's #VC exception handler may issue
VMGEXIT with SVM_EXIT_CPUID (0x72) or SVM_EXIT_MSR (0x7c) exit codes
via the GHCB page protocol. The hypervisor had no handlers for these
exit codes, causing the guest's #VC handler to fail and trigger
sev_es_terminate(), which sends GHCB_MSR_TERM_REQ (0x100). The
hypervisor then panicked on the unhandled 0x100 operation.
Add the following handlers to the GHCB VMGEXIT processing:
- SVM_EXIT_CPUID (0x72): Read function/index/xcr0/xss from the GHCB
page and return CPUID results via get_cpuid_values().
- SVM_EXIT_MSR (0x7c): Handle MSR read (RDMSR) and write (WRMSR)
requests from the guest via the GHCB page protocol.
- GHCB_MSR_TERM_REQ (0x100): Decode reason_set and reason_val from the
GHCB MSR and return an error instead of panicking, allowing graceful
error propagation.
Testing:
Reproducer (on Azure DC16as_cc_v5, /dev/mshv):
cloud-hypervisor --cpus boot=1,nested=off --memory size=512M \
--disk path=osdisk.img path=cloudinit \
--net "tap=,mac=12:34:56:78:90:06,ip=192.168.6.1,mask=255.255.255.128" \
--serial null --console pty \
--api-socket /tmp/ch.sock \
--igvm /igvm_files/linux-ttyS0.bin \
--host-data <hash> --platform sev_snp=on -v
Before fix:
thread 'vcpu0' panicked at hypervisor/src/mshv/mod.rs:1207:30:
Unsupported VMGEXIT operation: 100
After fix: VM boots successfully to login prompt with no panics.
All virtio devices (console, rng, disks) activate normally.
No regression risk for non-SEV-SNP: all new code is within the
HVMSG_X64_SEV_VMGEXIT_INTERCEPT handler which is only reached for
SEV-SNP guests.
Signed-off-by: Souradeep Chakrabarti <schakrabarti@microsoft.com>
Reorder resume() to: set_clock, device_manager.resume,
cpu_manager.resume. This matches the inverse of pause()
which correctly saves the clock before pausing vCPUs.
Signed-off-by: CMGS <ilskdw@gmail.com>
This gives the community more time to react to possible security chain
compromises.
We have high confidence that rust-vmm crates are trusted, and the
community is fully capable of spotting any issues. There is no need to
delay that group.
Signed-off-by: Wei Liu <liuwe@microsoft.com>
Delegate try_clone() to FixedVhd::clone() and new_async_io() to
FixedVhdSync, preserving DiskFileError::NewAsyncIo.
Signed-off-by: Anatol Belski <anbelski@linux.microsoft.com>
