When doing a snapshot/restore, Vcpus and Devices are made sure
to be asleep. The device controller gets the initial Vcpu state so that
Vcpus return to their prior state after the snapshot/restore
operations. Devices currently do not have the functionality of remaining
in the same state and will always go back to `running` at the end of the
operation. Adding a state save to devices.
BUG=N/A
TEST=presubmit, e2e_tests
Change-Id: Ibfb6c4055cb1f852ec6a85043b6464c621a83d59
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4641774
Reviewed-by: Frederick Mayle <fmayle@google.com>
Commit-Queue: Elie Kheirallah <khei@google.com>
Reviewed-by: Noah Gold <nkgold@google.com>
the BalloonWssConfigFfi struct was missing a #[repr(C)]
annotation, causing the generated header file to not include the struct
definition. this patch also removes a couple unneeded constants.
BUG=None
TEST=built crosvm_control, checked generated header
Change-Id: Ic02c2145d16143257f5096d4811a4012bb827d69
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4607224
Reviewed-by: T.J. Alumbaugh <talumbau@google.com>
Commit-Queue: Daniel Verkamp <dverkamp@chromium.org>
Auto-Submit: maciek swiech <drmasquatch@google.com>
Reviewed-by: Daniel Verkamp <dverkamp@chromium.org>
The VM restore path needs to make blocking `VmMemoryRequest`s from the
run_control thread (when it activates PCI devices). However, that's the
same thread responsible for processing the `VmMemoryRequest`s, so we end
up in a deadlock.
The IRQ chip had already been moved into new thread and this CL attempts
to closely copy that approach.
BUG=b:266514327
Change-Id: I914554f0d7318016102db1fe6f965d98a0274a94
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4598285
Reviewed-by: Elie Kheirallah <khei@google.com>
Commit-Queue: Frederick Mayle <fmayle@google.com>
Reviewed-by: Noah Gold <nkgold@google.com>
add a new event from balloon device to the hypervisor to indicate when
the requested balloon size has been reached. this will tell the
hypervisor that balloon is not going to change the page states anymore.
Bug: b:267051826
Signed-off-by: Vaibhav Nagarnaik <vnagarnaik@google.com>
Change-Id: Ie0be0b6293b7e4f4e1591050307f55501bd8c80b
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4602918
Reviewed-by: Noah Gold <nkgold@google.com>
remove `handle_inflate` and `handle_deflate` methods from the trait Vm
and replace with a single `handle_balloon_event` method. this will allow
more balloon events to be handled through the `enum BalloonEvent`.
the change is refactoring and a no-op.
Bug: b:267051826
Signed-off-by: Vaibhav Nagarnaik <vnagarnaik@google.com>
Change-Id: Ia5436eb3b72d6081ac3c05ae461eeb426e5f783c
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4601859
Reviewed-by: Vikram Auradkar <auradkar@google.com>
Reviewed-by: Noah Gold <nkgold@google.com>
for ease of initial implementation the number of WSS bins was hard-coded
to 4, however the WSS spec and kernel driver permit any number of bins
between 2-16. this patchset introduces a crosvm run flag which allows to
override the default 4 bins.
the protobuf-encoded WSS report will scale the number of bins correctly
in the return value, however the C crosvm_control_* API will always use
an array of 16 buckets, with a `num_bins` parameter that indicates how
many of them are actually valid. this has been done to preserve safety
across the ffi boundary.
this patchset also removes the sending of an initial WSS config. the
coordinating program (e.g. concierge et al.) is responsible for sending
the config to the guest after boot via the crosvm_control_* API. this is
largely due to the fact there may be various desirable configs for
varying scenarios and it does not make sense to have a baked-in default
config anymore.
BUG=b/278118696
TEST=https://x20.corp.google.com/users/dr/drmasquatch/socket-pinger-variable-bins
Change-Id: Ifaef2933a13a59d8d2580b5a9e283568fe6e86cf
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4577421
Commit-Queue: maciek swiech <drmasquatch@google.com>
Reviewed-by: Daniel Verkamp <dverkamp@chromium.org>
Bring some type safety to the wild west of vm_control Tubes.
This patch only converts the VmMemoryRequest/VmMemoryResponse instances,
but we should ideally apply this pattern to all of the Tube types
available in vm_control.
Change-Id: Id0d47b176af6feb0966aec5552f6cbff16c08881
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4504550
Commit-Queue: Daniel Verkamp <dverkamp@chromium.org>
Reviewed-by: Noah Gold <nkgold@google.com>
Move the snapshot logic to a function similar to do_restore.
BUG=N/A
Test=presubmit
Change-Id: I73787a89e8ec6051fd10c6e5a3ea64eabc461017
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4550001
Commit-Queue: Elie Kheirallah <khei@google.com>
Reviewed-by: Noah Gold <nkgold@google.com>
Reviewed-by: Frederick Mayle <fmayle@google.com>
Reviewed-by: Daniel Verkamp <dverkamp@chromium.org>
With the server support in place, we can connect to the control socket
just like unix does (albeit a named pipe instead of a UDS).
BUG=b:276361599
TEST=tested snapshot commands on Windows.
Change-Id: I874463026e3b4504009c6a19d7ba9ffb8fa169fa
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4558308
Commit-Queue: Noah Gold <nkgold@google.com>
Reviewed-by: Richard Zhang <rizhang@google.com>
If --socket is specified on a non-prod build, we will run a named pipe
server on that socket which will soon service requests like pressing the
power button, etc.
BUG=b:276361599
TEST=builds + later CLs in the chain.
Change-Id: I0be227f5e915feb29ab2883ab3bceb85a50cfc92
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4558307
Reviewed-by: Richard Zhang <rizhang@google.com>
Commit-Queue: Noah Gold <nkgold@google.com>
Add Irq_chip snapshot/restore to the existing operation of
snapshot/restore in the VM.
BUG=b:276357277
TEST=crosvm snapshot/restore && presubmit
Change-Id: I7bfb7eba11668e507d8a7d8b7fa75878c3b1a943
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4550006
Reviewed-by: Frederick Mayle <fmayle@google.com>
Reviewed-by: Steven Moreland <smoreland@google.com>
Reviewed-by: Daniel Verkamp <dverkamp@chromium.org>
Commit-Queue: Elie Kheirallah <khei@google.com>
since the RegisteredEvent model is effectively an external API, switch
to using protobuf as a more formal/stable means of communicating
messages. also introduces exporting the registered_events.proto file as
part of crosvm_control build, alongside the currently existing header
file.
this patch also introduces feature-gating for registered_events and
protos so as not to bring in too many third party dependencies for a
base build.
BUG=b/278117550
TEST=run bzImage locally
TEST=sidecar program available at https://x20.corp.google.com/users/dr/drmasquatch/socket-pinger-proto
Change-Id: I5d91d87f7807effc125352caf5c75eee2593f70d
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4521604
Reviewed-by: Daniel Verkamp <dverkamp@chromium.org>
Commit-Queue: maciek swiech <drmasquatch@google.com>
SwapStatus, SwapState, SwapStateTransition, SwapMetrics are directly
contained into crosvm_control.h
SwapState is translated as C++ enum.
SwapStateTransition::time_ms is now u64 because u128 does not support
repr(C).
Changed usize in the structs to u64 because usize is translated as
uintptr_t.
BUG=b:265386761
TEST=cargo build -p crosvm_control
Change-Id: Ia9df56d40c1884067a712620e0fc797e07af77a2
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4486549
Reviewed-by: Daniel Verkamp <dverkamp@chromium.org>
Reviewed-by: David Stevens <stevensd@chromium.org>
Commit-Queue: Shin Kawamura <kawasin@google.com>
Reviewed-by: Dennis Kempin <denniskempin@google.com>
Remove the architecture-specific requirements from the gdb config
checks. This enables gdb across every supported target_arch without
having to manually add new architectures to each cfg check.
For the specific case of target_arch = "arm", this patch will newly
enable gdb support. The gdbstub protocol will still send aarch64 state,
matching the guest.
This stubs out enough riscv64 gdb functions and types to make
`cargo build --features=gdb` compile, but gdb support will not be
functional on that architecture without additional work.
Change-Id: I63b079b7a3dca4aec2c13c775c0ccb8850625884
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4506285
Commit-Queue: Daniel Verkamp <dverkamp@chromium.org>
Reviewed-by: Dennis Kempin <denniskempin@google.com>
Reviewed-by: Dylan Reid <dgreid@rivosinc.com>
Status does not depend on any swap related feature. Exposing Status to
all variants is easy to implement swap status FFI at crosvm_control.
The new "swap/enable" feature switches the actual vmm-swap functionality
to be compiled. The feature is enabled by default on "swap" crate but
disabled on the root package for test dependency.
Vmm-swap feature is enabled by `--features=swap` flag on cargo build as
before.
swap/src/controller.rs is copied from swap/src/lib.rs.
BUG=b:265386761
TEST=tools/dev_container tools/run_tests2
Change-Id: Ifc2539a62d0f594fd5bbb41623c735ea2621f7b6
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4486546
Reviewed-by: Dennis Kempin <denniskempin@google.com>
Commit-Queue: Shin Kawamura <kawasin@google.com>
Reviewed-by: David Stevens <stevensd@chromium.org>
When we restore a snapshot, we need to refresh the set of Events that
the IRQ handler thread is listening on, otherwise it won't forward IRQs
from the restored/re-created events.
BUG=b:277651574
TEST=tools/dev-container e2e_tests/run
Change-Id: Icd61112c23f063af0f6f217efe2b3359f884e351
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4416558
Reviewed-by: Steven Moreland <smoreland@google.com>
Reviewed-by: Elie Kheirallah <khei@google.com>
Commit-Queue: Noah Gold <nkgold@google.com>
Reviewed-by: Richard Zhang <rizhang@google.com>
since run_control is single threaded, we should be able to always
discard stale wss/stats messages in the handlers in vm_control,
obviating the need to keep a separate tube for wss crosvm_commands.
BUG=None
TEST=run bzImage
Change-Id: I7de192a08ae7a42641cbabe183dbad1c6ef627a4
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4428055
Reviewed-by: Daniel Verkamp <dverkamp@chromium.org>
Commit-Queue: Maciek Swiech <drmasquatch@google.com>
balloon size is useful for balloon policies using working set size, so
we return the balloon-device cached balloon size. unless there is some
error in adjustement of the balloon, this should reflect that real
balloon size and saves having to query the guest an additional time.
BUG=b/278119213
TEST=run bzImage
Change-Id: I7c6b558358b8db099a737430e8431997bb110b2b
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4428532
Commit-Queue: Daniel Verkamp <dverkamp@chromium.org>
Reviewed-by: T.J. Alumbaugh <talumbau@google.com>
Reviewed-by: Daniel Verkamp <dverkamp@chromium.org>
Auto-Submit: Maciek Swiech <drmasquatch@google.com>
This is a reland of commit 22c212d54f
This reland avoids using the new pre-mapped memory regions on arm
devices, since there is insufficient address space. The new path is
still used on aarch64.
Original change's description:
> vm_control: reduce user memory region modifications
>
> Reduce how often KVM_SET_USER_MEMORY_REGION is called when the tdp mmu
> is enabled. With the tdp mmu, there is no memory overhead from creating
> large memory regions (at least until a nested VM is started). Simply
> mmap'ing/munmap'ing fds within a pre-created memory region is more
> efficient. It also addresses audio jank caused by removing a memory
> region.
>
> Adding this support to VmMemoryRequest will allow FsMappingRequest to be
> removed in a later change.
>
> BUG=b:274037632
> TEST=tast run arc.Boot.vm
> TEST=manually launch gedit in crostini
>
> Change-Id: I2ac02454ecb734c9707b6d67546135134b887527
> Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4402068
> Reviewed-by: Dennis Kempin <denniskempin@google.com>
> Reviewed-by: Daniel Verkamp <dverkamp@chromium.org>
> Commit-Queue: David Stevens <stevensd@chromium.org>
Bug: b:274037632
Change-Id: I5deedfd3a030640f9af950cee675fac0d9a411a0
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4421352
Reviewed-by: Daniel Verkamp <dverkamp@chromium.org>
Commit-Queue: Dennis Kempin <denniskempin@google.com>
Reviewed-by: Dennis Kempin <denniskempin@google.com>
This reverts commit 22c212d54f.
Reason for revert: Suspected to break crostini http://go/bbid/8783938922329906785
Original change's description:
> vm_control: reduce user memory region modifications
>
> Reduce how often KVM_SET_USER_MEMORY_REGION is called when the tdp mmu
> is enabled. With the tdp mmu, there is no memory overhead from creating
> large memory regions (at least until a nested VM is started). Simply
> mmap'ing/munmap'ing fds within a pre-created memory region is more
> efficient. It also addresses audio jank caused by removing a memory
> region.
>
> Adding this support to VmMemoryRequest will allow FsMappingRequest to be
> removed in a later change.
>
> BUG=b:274037632
> TEST=tast run arc.Boot.vm
> TEST=manually launch gedit in crostini
>
> Change-Id: I2ac02454ecb734c9707b6d67546135134b887527
> Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4402068
> Reviewed-by: Dennis Kempin <denniskempin@google.com>
> Reviewed-by: Daniel Verkamp <dverkamp@chromium.org>
> Commit-Queue: David Stevens <stevensd@chromium.org>
Bug: b:274037632
Change-Id: Id5caebd0bd33f4984567ff604f117ae4d95a71ab
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4422453
Commit-Queue: Dennis Kempin <denniskempin@google.com>
Reviewed-by: Daniel Verkamp <dverkamp@chromium.org>
Bot-Commit: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
this patch adds the crosvm_control command (and internal implementation)
to send a new WSS config command to a guest balloon driver. included is
a refactor of how wss op notifications are dispatched in the balloon
device.
BUG=b/271891148
TEST=run bzImage
TEST=socket-pinger program available at https://x20.corp.google.com/users/dr/drmasquatch/socket-pinger
Change-Id: Id9696b0e82b3ba54d87cd0928070c55da9a77d1f
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4415596
Reviewed-by: Daniel Verkamp <dverkamp@chromium.org>
Commit-Queue: Maciek Swiech <drmasquatch@google.com>
Reduce how often KVM_SET_USER_MEMORY_REGION is called when the tdp mmu
is enabled. With the tdp mmu, there is no memory overhead from creating
large memory regions (at least until a nested VM is started). Simply
mmap'ing/munmap'ing fds within a pre-created memory region is more
efficient. It also addresses audio jank caused by removing a memory
region.
Adding this support to VmMemoryRequest will allow FsMappingRequest to be
removed in a later change.
BUG=b:274037632
TEST=tast run arc.Boot.vm
TEST=manually launch gedit in crostini
Change-Id: I2ac02454ecb734c9707b6d67546135134b887527
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4402068
Reviewed-by: Dennis Kempin <denniskempin@google.com>
Reviewed-by: Daniel Verkamp <dverkamp@chromium.org>
Commit-Queue: David Stevens <stevensd@chromium.org>
this patch implements the virtio-balloon extensions proposed in
go/roziere - namely adding the Working Set Size vqueues and listener on
the wss vqueue. in addition we add a crosvm_control command which allows
a user to request that the balloon device send a request to the driver
to produce a WSS report, similar to asking for balloon stats.
BUG=b:271757427,b:273212204
TEST=sidecar process, available at https://x20.corp.google.com/users/dr/drmasquatch/socket-pinger
TEST=deploy to DUT, restart termina and arcvm
Change-Id: I45753f523aae97d14bfc5dcbad5ddfcb70d10beb
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4260158
Reviewed-by: T.J. Alumbaugh <talumbau@google.com>
Reviewed-by: Daniel Verkamp <dverkamp@chromium.org>
Commit-Queue: Maciek Swiech <drmasquatch@google.com>
This is supported since Rust 1.62.
Fixes the new derivable_impls clippy lint, added in Rust 1.67.
BUG=b:276487055
TEST=tools/clippy # with rust 1.68
Change-Id: I3696dd3cf4acbf9b5a2cfa8c2046e7656e13dbbd
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4390735
Reviewed-by: Alexandre Courbot <acourbot@chromium.org>
Commit-Queue: Daniel Verkamp <dverkamp@chromium.org>
The deadlock could happen on this scenario:
1. User enables vmm-swap. The main process sends Command::Enable to the
monitor process.
2. User request the current status at the same time. The main process
sends Command::Status to the monitor process and wait for the
response from the monitor process.
3. The monitor process start enabling vmm-swap. Sends
VmSwapCommand::Suspend to the main process and wait for
VmSwapResponse::SuspendCompleted.
4. The main process is blocked by the step 2 and the request from step 3
is never consumed.
The root issue issue is that Command::Status can be inserted between
Command::Enable and VmSwapCommand::Suspend. This CL simplifies the
communication between the main and monitor processes on enabling
vmm-swap and resolve the deadlock.
VmSwapCommand and corresponding Tube was added by
https://crrev.com/c/4293656, but is now removed by this CL.
BUG=b:275671628
TEST=manual test
Change-Id: Ia838b1feddb4a3f41bd729e9147adc5e2df866aa
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4387662
Reviewed-by: David Stevens <stevensd@chromium.org>
Commit-Queue: Shin Kawamura <kawasin@google.com>
We can require the ID field to be present in all the arch specific
structs, then we don't need an extra layer.
Change-Id: Icbbf0ec5b99fd6816863a611a538e67def09d87f
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4385864
Reviewed-by: Daniel Verkamp <dverkamp@chromium.org>
Reviewed-by: Elie Kheirallah <khei@google.com>
This CL adds BAR reprogramming support to a VirtioPciDevice. A tube for
handling ioevent register requests is added. The requests are handled
similarly to existing VmMemoryRequests.
Test: Presubmit tests. Hotplugged virtio device without VM memory
allocation functions after guest OS gives the memory.
Bug: 271204121
Change-Id: Ia0d421d8cd461eeee8f402256d0c7620c507f0e9
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4311103
Reviewed-by: Ningyuan Wang <ningyuan@google.com>
Commit-Queue: Ningyuan Wang <ningyuan@google.com>
Reviewed-by: Zihan Chen <zihanchen@google.com>
Reviewed-by: Daniel Verkamp <dverkamp@chromium.org>
"crosvm swap trim <socket_path>" command for trimming 2 types of pages
in the staging memory before swapping out to reduce the disk I/O.
* clean pages
* zero pages
The doc comment of TrimContext explains the page types.
BUG=b:265592787
TEST=cargo test -p swap
Change-Id: I5c33149f7d0bfd712f07fd11eb9aa07c1a8b0e7a
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4357224
Reviewed-by: Daniel Verkamp <dverkamp@chromium.org>
Commit-Queue: Shin Kawamura <kawasin@google.com>
Reviewed-by: David Stevens <stevensd@chromium.org>
The base crate has some nifty abstractions but also
pulls in things like minijail, audio_streams and
various random items.
This prevents true cross-platform interoperatibility,
which requires the ability to upload to crates.io
with relative ease.
This change removes the base crate. The replacement
strategy is two fold:
- For things like SafeDescriptor which don't have
adequate solutions on crates.io, just copy it
internally. I think SafeDescriptor in particular --
since it works on bot Windows/Linux should probably
be made into a separate crate long-term since it is
generically valuable. Similarly for the Event and
WaitContext APIs. For now, live with the duplication.
- Otherwise, the nix crate does most of the Linux
specific items.
The cross-domain context type has a lot of dependencies
on Linux sockets / epoll. This is refactored to look more
like the Fuchsia solution, which uses std::sync::mspc
channels to connect to an utility Wayland FIDL library.
The rest of Rutabaga mostly relies just on SafeDescriptor
and the ability to clone it. rutabaga_gralloc allows
provides shared memory used via virtwl.
Another thing to call out is zerocopy is now strongly
versioned to the latest stable release on crates.io
I also got rid of #[cfg(feature = "virgl_renderer_next")]
on the renderer server fd, since it was complicating the
refactoring. SafeDescriptor works everywhere and adding
the #ifdef should be reserved for the lowest-levels of
platform specific code when possible.
Overall, this makes compiling on Fuchsia hosts much
simpler.
BUG=b:273555494
TEST= - run cross_domain context type, see Wayland apps
- note there is a previous bug in cross domain
when resizing windows, it occurs before this CL.
The reason seems to be raciness and can likely be fixed
gfxstream-like seqno.
- run gfxstream + virgl on Linux, verify everything works.
Change-Id: I7519f2b8ed0cd57c47980d4e76403057498cf064
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4353700
Reviewed-by: Dennis Kempin <denniskempin@google.com>
Reviewed-by: Daniel Verkamp <dverkamp@chromium.org>
Commit-Queue: Gurchetan Singh <gurchetansingh@chromium.org>
This is a reland of commit 42a437de5e
This change fixes a crash that occurs when no-rtc is set by properly
checking the Option that can contain the cmos's control tube, instead of
just calling unwrap. It also replaces an unwrap during cmos
initialization with proper error propagation.
Original change's description:
> devices: cmos: implement rtc alarm
>
> Implement CMOS alarm functionality. This allows guests to set rtc
> alarms, which is useful when putting a Linux guest into s2idle.
>
> This change allocates a dedicated interrupt for the goldfish battery,
> instead of reusing the sci irq. Sharing an irq line between a wakeup
> interrupt and a non-wakeup interrupt doesn't work well because the
> non-wakeup interrupt won't be delivered to its driver while the driver
> is suspended, which results in the irq being permanently asserted.
> Simply using a dedicated interrupt for the goldfish battery is much
> easier than adding suspend/resume support to the Linux driver.
>
> BUG=None
> TEST=cargo test -p devices cmos, manual tests w/rtcwake
>
> Change-Id: I757acc64b61e414d5d9df5a1fb4770943ef985bf
> Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4314184
> Reviewed-by: Elie Kheirallah <khei@google.com>
> Reviewed-by: Daniel Verkamp <dverkamp@chromium.org>
> Commit-Queue: David Stevens <stevensd@chromium.org>
BUG=None
TEST=cargo test -p devices cmos, manual tests w/rtcwake and no-rtc
Change-Id: Iec73c0aa08b4f2081c3b7a824254723ea9da5365
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4364558
Reviewed-by: Elie Kheirallah <khei@google.com>
Reviewed-by: Daniel Verkamp <dverkamp@chromium.org>
Commit-Queue: David Stevens <stevensd@chromium.org>
Auto-Submit: David Stevens <stevensd@chromium.org>
Add restore to Vcpu to restore VCPU state from snapshot.
Add kick_vcpu to kick a specific vcpu.
BUG=b:266515147
BUG=b:232437513
Test=crosvm run && crosvm snapshot take img_name socket
Change-Id: Ie09419fd50dcecd0226b2b716236d2c8196b58ec
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4260157
Reviewed-by: Daniel Verkamp <dverkamp@chromium.org>
Commit-Queue: Elie Kheirallah <khei@google.com>
Auto-Submit: Elie Kheirallah <khei@google.com>
Add snapshot to vCPU for x86 and x86_64 arch.
BUG=b:266515147
BUG=b:232437513
Test=crosvm run && crosvm snapshot take img_name socket
Change-Id: I65d059a7537c713e9ef6c1456a7aeae2b6bc4728
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4237759
Auto-Submit: Elie Kheirallah <khei@google.com>
Reviewed-by: Daniel Verkamp <dverkamp@chromium.org>
Reviewed-by: Frederick Mayle <fmayle@google.com>
Commit-Queue: Elie Kheirallah <khei@google.com>
MsixConfig is now serializable. After deserialization, there is a new function
`restore_msix_all`, which will re-create the MSI-X vectors based on the
data that was deserialized. We could try to weave this into the actual
serde deserialization process, but that will likely be more hassle than
it is worth (it will also stop us from fixing up the msi control Tube on
restore, which we'll have to do in a future CL).
BUG=b:272137957
TEST=presubmit
Change-Id: I36fbc35a13fe11ae2f7c899431d2ef239ac00f90
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4337964
Reviewed-by: Elie Kheirallah <khei@google.com>
Reviewed-by: Frederick Mayle <fmayle@google.com>
Commit-Queue: Noah Gold <nkgold@google.com>
This CL renames some functions and variables introduced for PCI BAR
reprogramming to avoid confusion.
Test: presubmit tests.
Bug: b:271204121
Change-Id: I77aaca123a677b56da0b45ce78a54e84bec64a59
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4332623
Commit-Queue: Ningyuan Wang <ningyuan@google.com>
Reviewed-by: Hugo Benichi <hugobenichi@google.com>
Reviewed-by: Daniel Verkamp <dverkamp@chromium.org>
Instead of wiring up a channel ahead of time, set one up on demand. This
significantly reduces the amount of plumbing required. We'll need more
VcpuControl cases in the near future that require responses like this
(e.g. vcpu snapshotting and restoring) and the extra plumbing for each
in unpalatable.
It isn't obvious what the performance difference is and we don't have a
relevant benchmark. None of the use cases are particularly performance
sensitive yet.
BUG=b:232437513
Change-Id: I96e720065275894ac0eb2965d5f707ec57b19c64
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4326945
Commit-Queue: Frederick Mayle <fmayle@google.com>
Reviewed-by: Daniel Verkamp <dverkamp@chromium.org>
Reviewed-by: Noah Gold <nkgold@google.com>
Reviewed-by: Elie Kheirallah <khei@google.com>
Upstreams blocking overlapped read/write commmands to base.
These funtions will be used by the broker to forward messages exchanged
between main and service.
Bug: 269191436
Test: none
Change-Id: I029e8c440bb6b8234fe2d1ca3b1dd5ddb408d814
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4329913
Auto-Submit: Vikram Auradkar <auradkar@google.com>
Commit-Queue: Vikram Auradkar <auradkar@google.com>
Reviewed-by: Noah Gold <nkgold@google.com>
crrev.com/v/4237140 was reverted in crrev.com/c/4310064 due to the
generated header file not defining the RegisteredEvent type. along with
re-reverting the original cl (and corresponding bugfix cl) this patch
introduces a RegisteredEventFfi type which mirrors the original
RegisteredEvent type but should be exposed in the generated header file.
BUG=b:269609274
TEST=sidecar process (patch: crrev.com/c/4276642)
TEST=cargo build, header file inspection
Change-Id: I4a17163e36eb466a540a72d2bb4356419739c974
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4311831
Reviewed-by: Daniel Verkamp <dverkamp@chromium.org>
Commit-Queue: Maciek Swiech <drmasquatch@google.com>
This reverts commit 42a437de5e.
Reason for revert: This is crashing if no_rtc is set (http://go/bbid/8787052669966742145/test-results)
Original change's description:
> devices: cmos: implement rtc alarm
>
> Implement CMOS alarm functionality. This allows guests to set rtc
> alarms, which is useful when putting a Linux guest into s2idle.
>
> This change allocates a dedicated interrupt for the goldfish battery,
> instead of reusing the sci irq. Sharing an irq line between a wakeup
> interrupt and a non-wakeup interrupt doesn't work well because the
> non-wakeup interrupt won't be delivered to its driver while the driver
> is suspended, which results in the irq being permanently asserted.
> Simply using a dedicated interrupt for the goldfish battery is much
> easier than adding suspend/resume support to the Linux driver.
>
> BUG=None
> TEST=cargo test -p devices cmos, manual tests w/rtcwake
>
> Change-Id: I757acc64b61e414d5d9df5a1fb4770943ef985bf
> Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4314184
> Reviewed-by: Elie Kheirallah <khei@google.com>
> Reviewed-by: Daniel Verkamp <dverkamp@chromium.org>
> Commit-Queue: David Stevens <stevensd@chromium.org>
Bug: None
Change-Id: I25737095f84ca0b56d94924a9d8ccb5fd42ff51c
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4329909
Commit-Queue: Dennis Kempin <denniskempin@google.com>
Bot-Commit: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
This CL adds trait implementation for PciDevice to handle BAR
reprogramming. When BAR reprogramming is detected, the PciDevice checks
all ioevents needed, and requests ioevent registeration for events that
are in the changed domain. The requests are handled by instances of
PciDevice via the trait function ioevent_register_request.
Before the CL, ioevent re-registering is unhandled, and PciDevice is
broken without notification. After CL, an error message will appear, but
device still would not function as the added trait is not implemented.
Test: Presubmit Tests. Hotplugged virtio device triggers error
IoEventNotSupported as expected.
Bug: 271204121
Change-Id: Ibf4bc00cfdce24e508cd50b5009f148c5a49db30
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4310071
Reviewed-by: Daniel Verkamp <dverkamp@chromium.org>
Commit-Queue: Ningyuan Wang <ningyuan@google.com>
Implement CMOS alarm functionality. This allows guests to set rtc
alarms, which is useful when putting a Linux guest into s2idle.
This change allocates a dedicated interrupt for the goldfish battery,
instead of reusing the sci irq. Sharing an irq line between a wakeup
interrupt and a non-wakeup interrupt doesn't work well because the
non-wakeup interrupt won't be delivered to its driver while the driver
is suspended, which results in the irq being permanently asserted.
Simply using a dedicated interrupt for the goldfish battery is much
easier than adding suspend/resume support to the Linux driver.
BUG=None
TEST=cargo test -p devices cmos, manual tests w/rtcwake
Change-Id: I757acc64b61e414d5d9df5a1fb4770943ef985bf
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4314184
Reviewed-by: Elie Kheirallah <khei@google.com>
Reviewed-by: Daniel Verkamp <dverkamp@chromium.org>
Commit-Queue: David Stevens <stevensd@chromium.org>
This reverts commit 2404c5ed44.
Reason for revert: This fails to build crosvm C interface and blocking ChromeOS CQ.
Original change's description:
> balloon: add event registration mechanism for sidecar processes.
>
> this patch introduces a mechanism for so-called 'sidecar processes' to
> register as event listeners, consuming events that are generated in
> various parts of crosvm. in this patch we focus on events that enable
> roziere cooperative ballooning, but lay the groundwork for a more
> general framework.
>
> the general idea is that a sidecar process would open a listening socket
> and register the path with crosvm for a specific event. crosvm will pass
> along a handle to a registered event tube to the appropriate device (or
> other internal component) which would then be responsible for passing
> along events. once events arrive back in the crosvm control loop, an
> attempt is made to dispatch the events to any registered sockets.
>
> BUG=b:269609274
> TEST=sidecar program that performs registration and receives events
>
> Change-Id: Iaff41aad8f862ed99a104c75623caaabc53e9e88
> Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4237140
> Commit-Queue: Maciek Swiech <drmasquatch@google.com>
> Reviewed-by: Daniel Verkamp <dverkamp@chromium.org>
Bug: b:269609274
Change-Id: I135e0d55bb296cf591548b7bd4feffc7ff9f3447
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4310064
Commit-Queue: Shin Kawamura <kawasin@google.com>
Reviewed-by: Junichi Uekawa <uekawa@chromium.org>
To safely flush IRQs, we need devices to be asleep. This CL splits our
existing snapshot device control command into three: sleep, snapshot,
and wake. This way we can ensure the devices are sleeping while we flush
IRQs & snapshot the VCPUs (in the future).
BUG=b:269652620
TEST=./tools/dev_container e2e_tests/run --tests suspend_resume
Change-Id: I82a3f280c24378ea5d2a28a5cd4e3db49c9b78f2
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4304243
Commit-Queue: Noah Gold <nkgold@google.com>
Reviewed-by: Elie Kheirallah <khei@google.com>
Reviewed-by: Frederick Mayle <fmayle@google.com>
https://crrev.com/c/4304741 broke these because the expected responses
are not `VmResponse::Ok`. There isn't good reason for them to have
custom types though, so I've moved them to `VmResponse::Ok` and added a
generic `VmResponse::ErrString` case to support stringy errors. Maybe
more stuff could be moved to `VmResponse::ErrString`, but I'm holding
off for now.
Change-Id: Ib5760217037ceb8718e36a77aeb300a4b0eb2a9f
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4304683
Reviewed-by: Elie Kheirallah <khei@google.com>
Reviewed-by: Noah Gold <nkgold@google.com>
Commit-Queue: Frederick Mayle <fmayle@google.com>
The tube response should be deserialized as a `SnapshotControlResult`
instead of `()`.
Also, we want to flush IRQs before snapshotting the IRQ chip state, so
it should be done before sending the snapshot command. However, this
part is still not right because the devices aren't asleep.
Change-Id: I94245b23a5166ec231dd36020e55930a80e007cd
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4304682
Reviewed-by: Noah Gold <nkgold@google.com>
Commit-Queue: Frederick Mayle <fmayle@google.com>
A command like `crosvm stop` could fail, but report an exit code of 0
with no error message.
Change-Id: I6491f8fcb755717999816d34d4a4eaf3964e1270
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4304741
Commit-Queue: Frederick Mayle <fmayle@google.com>
Reviewed-by: Daniel Verkamp <dverkamp@chromium.org>
this patch introduces a mechanism for so-called 'sidecar processes' to
register as event listeners, consuming events that are generated in
various parts of crosvm. in this patch we focus on events that enable
roziere cooperative ballooning, but lay the groundwork for a more
general framework.
the general idea is that a sidecar process would open a listening socket
and register the path with crosvm for a specific event. crosvm will pass
along a handle to a registered event tube to the appropriate device (or
other internal component) which would then be responsible for passing
along events. once events arrive back in the crosvm control loop, an
attempt is made to dispatch the events to any registered sockets.
BUG=b:269609274
TEST=sidecar program that performs registration and receives events
Change-Id: Iaff41aad8f862ed99a104c75623caaabc53e9e88
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4237140
Commit-Queue: Maciek Swiech <drmasquatch@google.com>
Reviewed-by: Daniel Verkamp <dverkamp@chromium.org>
The design sending SIGSTOP to the main process and the device processes
from the monitor process does not work with sandboxing the monitor
process because user namespace without pid namespace is not supported by
minijail.
This changes the design so that monitor process sends a message to the
main process, which is then responsible for guaranteeing that nothing
except the monitor process will access guest memory when vmm-swap is
being enabled. The main process does this by sending SIGSTOP to device
processes and suspending the vCPU threads. Although there are other
threads in the main process, none of them currently access guest memory.
The vmm-swap feature now does not support `--disable-sandbox`. It would
be possible to support it by sending `SIGSTOP` from the monitor process
to the main process. However there is no clear use case for vmm-swap
with `--disable-sandbox`, so the extra complexity is not worth it.
BUG=b:270248453
TEST=manual test
Change-Id: Ie24d5a5b5f8e6999d133beb9b4c3562e26427838
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4293656
Reviewed-by: David Stevens <stevensd@chromium.org>
Commit-Queue: Shin Kawamura <kawasin@google.com>
The constant was for a max, not an expected loop value.
BUG=none
TEST=tools/clippy
Change-Id: I71a69612951b0f12f32e7d800d7a892e7f472371
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4299921
Reviewed-by: Frederick Mayle <fmayle@google.com>
Commit-Queue: Noah Gold <nkgold@google.com>
Just prior to taking a snapshot, we want to ensure that all pending IRQs
are flushed to LAPICs so that they can be safely saved & restored with a
snapshot. This CL wires the snapshot handler to the IRQ handler so that
the IRQs can be confirmed to be flushed.
BUG=b:269652620
TEST=./tools/dev_container e2e_tests/run --tests suspend_resume exercises this
code path, but does not directly test the changed behavior.
TEST=this fixes a race condition. Writing a test to reproduce that race
condition would just introduce a flakey test. When we have a cold boot
test (which we can't write yet because it depends future features), it
will flake on restoring LAPIC state *if* this CL has not fixed the race
condition properly. Thus, the testing strategy for this CL is to make
sure our cold boot test is run with *many* iterations to detect such
flakes (see b/270778349).
Change-Id: I2f326a7fdee9f21edda8402920661ef984943761
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4292623
Reviewed-by: Elie Kheirallah <khei@google.com>
Reviewed-by: Frederick Mayle <fmayle@google.com>
Commit-Queue: Noah Gold <nkgold@google.com>
Add unit test cases to test disk resize and device reset.
BUG=b:270280130
TEST=`cargo test -p devices`
Change-Id: Idac5e5e18ff520b6ae7d85ca9a6d5d103b114e78
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4272806
Reviewed-by: Keiichi Watanabe <keiichiw@chromium.org>
Reviewed-by: Noah Gold <nkgold@google.com>
Reviewed-by: Daniel Verkamp <dverkamp@chromium.org>
Commit-Queue: Takaya Saeki <takayas@chromium.org>
do_while_vcpus_suspended() deeply depends on device snapshot feature.
It should be more generic.
The vmm-swap feature will need to execute enabling logic while vCPUs are
suspended.
BUG=b:270248453
TEST=cargo build
Change-Id: I4f1acb2cf9ad61ee0887c6c7bfa7f68ba1dfb3ed
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4279545
Reviewed-by: Daniel Verkamp <dverkamp@chromium.org>
Reviewed-by: Elie Kheirallah <khei@google.com>
Commit-Queue: Shin Kawamura <kawasin@google.com>
page fault logger is only for debugging purpose. It has a security
concern about exposing the trigger to enable the logger. The logger will
make the page fault handling slower and consumes disk resources.
Instead, the page fault logger is only enabled by compiler feature flag
like:
```
cargo build --features=swap,swap/log_page_fault
```
Opening a file on the monitor process is blocking the upcoming minijail
migration. Open the log file on the main process and pass the file
descriptor to the monitor process.
BUG=b:269372016
TEST=manual test
Change-Id: I5df2a32da7fb149680fa24617ffd905b2483108b
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4253287
Reviewed-by: David Stevens <stevensd@chromium.org>
Commit-Queue: Shin Kawamura <kawasin@google.com>
Allow devices to notice wakeup events supported by
VFIO_DEVICE_FEATURE_LOW_POWER_ENTRY_WITH_WAKEUP.
BUG=b:241526471
TEST=Trigger a wakeup event and verify the output from `lspci`
Change-Id: Ie93780606fbfaa7efff853b6ff12314b3f54f276
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4048547
Reviewed-by: Daniel Verkamp <dverkamp@chromium.org>
Auto-Submit: Victor Ding <victording@chromium.org>
Commit-Queue: Victor Ding <victording@chromium.org>
Behavior difference: will now call `irq_chip.kick_halted_vcpus()` (which
is a noop for many irq chip impls).
Change-Id: I165393fe22e0c90b9d751ab2f04246e3e93fb414
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4149735
Reviewed-by: Daniel Verkamp <dverkamp@chromium.org>
Reviewed-by: Elie Kheirallah <khei@google.com>
Commit-Queue: Frederick Mayle <fmayle@google.com>
This splits the `crosvm swap enable` command, which pauses the whole
crosvm processes while it write all the active pages on the guest memory
into the swap file, into 2 commands.
On `crosvm swap enable` command, the monitor process just moves the
guest memroy to the staging memory while blocking the whole crosvm
processes with SIGSTOP.
On `crosvm swap out` command, the monitor process writes the active
pages in the staging memory to the swap file.
The page faults between `swap enable` and `swap out` are swapped in from
the staging memory.
By waiting some time before sending swap out request after enabling
vmm-swap, hot pages on the guest memory, which are frequently accessed,
are swapped in from the staging memory and we can reduce the TBW of the
swap file and reduce the latency of page faults for hot pages.
BUG=b:263830401
TEST=cargo test -p swap -p base
Change-Id: Ia83f5052cc0bd4d25089c6a6680e40a1fa1661c0
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4134149
Reviewed-by: David Stevens <stevensd@chromium.org>
Commit-Queue: Shin Kawamura <kawasin@google.com>
These should be written as ChromeOS and ChromiumOS (without the space)
to match the updated branding. The copyright headers were already
migrated to the new style (https://crrev.com/c/3894243), but there were
some more instances left over.
BUG=None
TEST=tools/cargo-doc
Change-Id: I8c76aea2eb33b2e370ab71ee9b5cc0a4cfd00585
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4129934
Reviewed-by: Dennis Kempin <denniskempin@google.com>
Commit-Queue: Daniel Verkamp <dverkamp@chromium.org>
Adding restore to crosvm
Adding restore in vm_control
Cleaning snapshot and restore to be in DeviceCommand (single point of
entry for modifying devices)
Bug=b🅱️232437513
Test=WIP
Change-Id: Id7d51cb5689bc1872aa0dbcbc9a8778e0d26130e
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/3936662
Reviewed-by: Frederick Mayle <fmayle@google.com>
Reviewed-by: Steven Moreland <smoreland@google.com>
Auto-Submit: Elie Kheirallah <khei@google.com>
Commit-Queue: Elie Kheirallah <khei@google.com>
Reviewed-by: Daniel Verkamp <dverkamp@chromium.org>
Add snapshot functionality in Bus.
Add thread to handle communication between VM cmdline and devices.
Add DeviceCommandControl in vm_control to handle device-specific
actions.
Bug=b:232437513
Test=cargo build && cargo run ...
Change-Id: If3795c9e1f12ab4cd34ab36cad8eae43d469a50a
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/3930627
Auto-Submit: Elie Kheirallah <khei@google.com>
Commit-Queue: Elie Kheirallah <khei@google.com>
Reviewed-by: Frederick Mayle <fmayle@google.com>
Reviewed-by: Daniel Verkamp <dverkamp@chromium.org>
Reviewed-by: Steven Moreland <smoreland@google.com>
The current syntax is:
--gpu-display horizontal-dpi=100,vertical-dpi=200
We would like to change that to:
--gpu-display dpi=[100,200]
We will keep the backward compatibility for now, but the caller
should not use both of syntax at the same time.
FixedGpuDisplayParameters is introduced for backward compatibility
support, and it will be removed once the old syntax is deprecated.
BUG=b:260101753
TEST=presubmit
Change-Id: I66a31c448ecc1eba60f6d1cbcbc753315c6c1df6
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4049340
Reviewed-by: Alexandre Courbot <acourbot@chromium.org>
Reviewed-by: Jason Macnak <natsu@google.com>
Commit-Queue: Pujun Lun <lunpujun@google.com>
Handling page faults using userfaultfd for all the time after vmm-swap
is enabled originally has overhead. By disabling swapping, we can use
crosvm without the overhead when the crosvm become used heavily again.
This also refactor PageHandler and extract register logic to lib.rs.
BUG=b:215093219
TEST=cargo test -p swap
Change-Id: Id81d84ebe40067b4f19b212f6b81cbaf249c0c3c
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4005474
Reviewed-by: David Stevens <stevensd@chromium.org>
Commit-Queue: Shin Kawamura <kawasin@google.com>
The new command "crosvm swap log_pagefault <socket>" starts logging
page fault events. The log file "page_fault.log" will be located under
the swap directory.
Pagefault event log help up debugging the vmm-swap feature and
understand the memory usage of the guest os.
design document: go/tanooki-phase1-dd
BUG=b:253555760
TEST=manually tested. this is debug feature.
Change-Id: I13e36cef873cd83179032fe8a289c8df71384644
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/3954806
Reviewed-by: Daniel Verkamp <dverkamp@chromium.org>
Reviewed-by: David Stevens <stevensd@chromium.org>
Commit-Queue: Shin Kawamura <kawasin@google.com>
Added new crosvm commands "crosvm swap enable" and "crosvm swap status".
swap crate exposes SwapController to control the vmm-swap feature.
When vmm-swap is enabled, all the pages on the GuestMemory are swapped
out to a file and are freed to give a lot of memory back to OS.
design document: go/tanooki-phase1-dd
BUG=b:215093219
TEST=manually tested
Change-Id: Ideb10b0119edd9f47b66c8ac61add21ac08181d4
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/3871758
Reviewed-by: David Stevens <stevensd@chromium.org>
Commit-Queue: Shin Kawamura <kawasin@google.com>
As part of commit cadc84b32a the crosvm
suspend command has been extended to generating sleep button injection
in order to put VM in suspend state. Additionally the crosvm suspend
command completed after the guest entered proper power state.
Nevertheless in some circumstances above implementation caused issues.
E.g. when the guest with pass-through devices entered run-time device
suspension, triggering a subsequent system suspend request fails and
makes VM inaccessible. It is because before entering system suspension,
devices which entered device run-time suspension needs to be resumed.
Therefore in problematic scenario:
- VmRequest::Suspend was triggered and it resulted with blocking till
guest entered proper suspend state
- during VM suspension process some pci config write accesses are made,
which in some cases triggers write_msi_reg-> write_msi_capability->
add_msi_route, which underneath sends VmIrqRequest::AddMsiRoute
request
- above VmIrqRequest::AddMsiRoute request can't be satisfied since
VmRequest::Suspend made the run_control wait loop busy waiting and
preventing further requests to be processed
In order to fix the above issue, instead of busy waiting directly in the
run_control wait loop for suspend notification, spawn a new thread, which
will generate the acpi power button and start busy waiting for a
notification that the VM actually entered suspend state. This allows to
postpone sending response over the control tube indicating that the
'crosvm suspend ...' command completed and in the same time process
other Vm*Requests.
Mentioned postponement of 'crosvm suspend ..' completion is crucial
since the full system suspension needs to be blocked till VM actually
enters proper suspend state.
BUG=b:248462336
TEST=With a setup where some device is pass-through to the VM and
--force-s2idle flag is used: wait till pass-through device enters
run-time suspend state and trigger "crosvm suspend
/run/vm/vm.<hash>/crosvm.sock" followed by "crosvm resume ..". Also make
sure that full system suspension/resumption doesn't make VM inaccessible
anymore.
Change-Id: Ic23461a78a62d2116cf5674c71d89f4f86ad96c3
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/3944915
Reviewed-by: Daniel Verkamp <dverkamp@chromium.org>
Commit-Queue: Grzegorz Jaszczyk <jaszczyk@google.com>
Reviewed-by: Noah Gold <nkgold@google.com>
This CL contains a few fixes to needless_return which need approval from
someone familiar with the code. In two cases, the return was removed
from a function that returns a value, so it is considered safe because
if code is added later into the function, it will be clear that this is
removing an implicit return from the function. In the last case
(iommu.rs), the function has several error! macros that do not trigger
an early return, so it seems that it is okay to remove the last early
return.
BUG=b:157245930, chromium:908640
TEST=cargo clippy (no logical code changes)
Change-Id: I00685c9033499a92713eb591eadb5f90c29e761d
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/4014835
Reviewed-by: Daniel Verkamp <dverkamp@chromium.org>
Commit-Queue: Clarissa Garvey <clarissagarvey@chromium.org>
Reviewed-by: David Stevens <stevensd@chromium.org>
Trying to reconcile the difference between the linux and windows
implementations.
Code relying on the eventfd count must now use the linux specific
`EventExt` interface.
BUG=b:231344063
TEST=presubmits
Change-Id: I14eb50f7a02d766a00f27aca388823309633e193
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/3864030
Commit-Queue: Frederick Mayle <fmayle@google.com>
Reviewed-by: Daniel Verkamp <dverkamp@chromium.org>
Reviewed-by: Noah Gold <nkgold@google.com>
This replaces the handwritten arg parser with the serde_keyvalue
based parser. Eventually we will unify the arg parsing with Unix.
BUG=b:233676779
TEST=cargo b --features all-msvc64,gpu --no-default-features
TEST=cargo t -p vm_control --features all-msvc64,gpu
--no-default-features
TEST=cargo t -p crosvm sys::windows::config::
--features all-msvc64,gpu --no-default-features
Change-Id: I36a563be9767c7e5cbd3ab44f6a9ba23cd64cdb6
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/3939033
Commit-Queue: Pujun Lun <lunpujun@google.com>
Reviewed-by: Alexandre Courbot <acourbot@chromium.org>
... and include this information in the EDID blob.
BUG=b:163864461
TEST=cargo build
TEST=Android build
TEST=launch Cuttlefish with various DPIs and inspect
`adb shell dumpsys SurfaceFlinger`
Change-Id: I3f9ca0283f335d51063bf8e555ff2ffa7ead104b
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/3954473
Reviewed-by: Pujun Lun <lunpujun@google.com>
Reviewed-by: Daniel Verkamp <dverkamp@chromium.org>
Commit-Queue: Jason Macnak <natsu@google.com>
Adds `crosvm gpu <>` commands which allows for interactively
adding/listing/removing virtio-gpu displays/scanouts.
Also updates the virtio gpu device to use a Map<u32, Scanout>
instead of a Vec<Scanout> and always report the maximum number
of displays (16) with enabled/disabled. This helps handle the
case of removing a display that is not the last display number.
BUG=b:163864461
TEST=cargo build
TEST=cargo build (with gpu feature disabled)
TEST=Android build
TEST=launch Cuttlefish
TEST=crosvm gpu list-displays <socket>
TEST=crosvm gpu add-displays
--gpu-display=width=900,height=600
--gpu-display=width=600,height=480
<socket>
TEST=crosvm gpu list-displays <socket>
TEST=crosvm gpu remove-displays
--display-id=1
<socket>
Change-Id: I7530498897eb4e84199eaeb3d66df006df62817f
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/3911102
Commit-Queue: Jason Macnak <natsu@google.com>
Reviewed-by: Daniel Verkamp <dverkamp@chromium.org>
Add a new identifier for a device, the "device id". This is a
combination of the device UUID + driver UUID. Reason for not only using
device UUID is that the spec calls out both must match:
https://registry.khronos.org/vulkan/specs/1.3-extensions/html/vkspec.html#external-memory-handle-types-compatibility
We have also seen some less-ideal implementations of device UUID on
Windows that return mostly zero with a few bits set, so adding the
driver UUID should increase confidence.
physical_device_idx is removed.
BUG=b:244622199
TEST=presubmit & downstream
Change-Id: I8174227e3d4bcadf778b7a73c9f84bb475169326
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/3926105
Reviewed-by: Daniel Verkamp <dverkamp@chromium.org>
Commit-Queue: Idan Raiter <idanr@google.com>
Reviewed-by: Dennis Kempin <denniskempin@google.com>
Reviewed-by: Gurchetan Singh <gurchetansingh@chromium.org>
Rework how to specify which VmMemoryRequests are exposed to passthrough
devices via virtio-iommu. Instead of specifying this on a per-request
basis, specify it at at a property of the VirtioDevice. This allows the
feature to work even for vhost-user devices.
BUG=b:243061269
TEST=presubmit
Change-Id: I3c381705f10ae0822896a4b9be760202cf79d925
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/3865353
Reviewed-by: Dennis Kempin <denniskempin@google.com>
Reviewed-by: Gurchetan Singh <gurchetansingh@chromium.org>
Commit-Queue: David Stevens <stevensd@chromium.org>
Add the infrastructre for GDB 'p'/'P' packets, used by the client to get
or set individual registers.
Implement a back-end using KVM_{G,S}ET_ONE_REG on AArch64, enabling the
client to access the architecture's numerous system registers.
Add stubs for x86 to please the compiler; leave them unimplemented as,
although the ioctl is available, the architecture doesn't seem to make
use of it.
BUG=b:222222882
BUG=b:240711627
BUG=chromium:1141812
TEST=tools/dev_container ./tools/run_tests
Change-Id: I01968275e88e9d0098de40b3f78e2cd3a61b75da
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/3785468
Commit-Queue: Keiichi Watanabe <keiichiw@chromium.org>
Reviewed-by: Keiichi Watanabe <keiichiw@chromium.org>
Virglrenderer must not unmap from any thread other than the virtio-gpu
thread because virglrenderer is not thread-safe and EGL requires that
only one thread own a context concurrently.
For the VirglRenderer RutabagaComponent, we can unmap directly from the
virtio-gpu thread. In fact, we can greatly simplify the use of
ExternalMapping if we eliminate the associated map/unmap callbacks for
all RutabagaComponents.
Now VirtioGpu is responsible for ensuring that the map/unmap occurs and
is registered/unregistered. Also, the shared `map_request:
Arc<Mutex<ExternalMapping>>` spanning the KVM and virtio-gpu threads is
no longer necessary.
BUG=b:244626679
TEST=OpenGL and Vulkan gameplay with sandboxing disabled
TEST=OpenGL and Vulkan gameplay on ChromeOS with sandboxing
Change-Id: I4acaf4aabff3891536ca0439221cfbd021dba123
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/3888602
Reviewed-by: Daniel Verkamp <dverkamp@chromium.org>
Commit-Queue: Ryan Neph <ryanneph@google.com>
This search/replace updates all copyright notices to drop the
"All rights reserved", Use "ChromiumOS" instead of "Chromium OS"
and drops the trailing dots.
This fulfills the request from legal and unifies our notices.
./tools/health-check has been updated to only accept this style.
BUG=b:246579983
TEST=./tools/health-check
Change-Id: I87a80701dc651f1baf4820e5cc42469d7c5f5bf7
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/3894243
Reviewed-by: Daniel Verkamp <dverkamp@chromium.org>
Commit-Queue: Dennis Kempin <denniskempin@google.com>
Add an in-process control server-like mechanism which runs in the
devices command to enable controlling vhost-user devices with control
commands such as DiskControlCommands. Like the in-process version,
the vhost-user control server waits for connection on the specified
socket, processes the recevied command, and forwards it to the
corresponding device via established tubes. Currently, DiskCommands are
the only commands supported by the control server, as block devices are
the only devices which need this mechanism.
This patch also adds the mechanism to process forwarded commands in
the vhost-user block backend device, similar to the command processing
mechanism in the worker provided in the in-process block device.
BUG=b:191845881
TEST=cargo run devices -s /path/to/socket.sock --block ..., then run
cargo run disk resize 0 $SIZE /path/to/socket.sock
Change-Id: I7b96e6c4bb7371424ca220da5f95be88e82c1fc0
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/3841001
Reviewed-by: Alexandre Courbot <acourbot@chromium.org>
Reviewed-by: Keiichi Watanabe <keiichiw@chromium.org>
Tested-by: Keita Suzuki <suzukikeita@google.com>
Reviewed-by: Morg <morg@chromium.org>
Commit-Queue: Keita Suzuki <suzukikeita@google.com>
Bump the versions of these crates as we're about to require support for
unavailable register values and AArch64. Do it in a separate commit to
ease future bisects.
BUG=b:222222882
BUG=chromium:1141812
TEST=tools/dev_container ./tools/run_tests
Change-Id: I0bfa3559d172faf2df6bcffdc77714830f442051
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/3785466
Reviewed-by: Keiichi Watanabe <keiichiw@chromium.org>
Auto-Submit: Pierre-Clément Tosi <ptosi@google.com>
Tested-by: Pierre-Clément Tosi <ptosi@google.com>
Commit-Queue: Pierre-Clément Tosi <ptosi@google.com>
Reviewed-by: Takaya Saeki <takayas@chromium.org>
Make GdbStub query the number of available hardware breakpoints from the
VM as some architectures (e.g. AArch64) might permit a flexible number.
BUG=b:222222882
BUG=chromium:1141812
TEST=tools/dev_container ./tools/run_tests
Change-Id: I9220f642fc01939305bd17461eaf50c424d998bc
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/3785465
Auto-Submit: Pierre-Clément Tosi <ptosi@google.com>
Commit-Queue: Pierre-Clément Tosi <ptosi@google.com>
Tested-by: Pierre-Clément Tosi <ptosi@google.com>
Reviewed-by: Keiichi Watanabe <keiichiw@chromium.org>
Reviewed-by: Takaya Saeki <takayas@chromium.org>
Remove target_arch-specific #[cfg]s where the underlying code is
architecture-agnostic.
Introduce a GdbOps trait for architectures to implement.
Make use of the generic gdbstub::arch::Arch trait where relevant.
Import base::Tube unconditionally in arch/src/sys/unix.rs.
Expand crate::gdb::* in vm_control/src/lib.rs for clarity.
Keep target_arch checks in x86-specific code to exclude 32-bit builds as
those don't seem to provide GDB support.
BUG=b:222222882
BUG=chromium:1141812
TEST=tools/dev_container ./tools/run_tests
Change-Id: I3f5ceeeb9031bee222ecd388dddb815e256748e8
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/3785464
Tested-by: Pierre-Clément Tosi <ptosi@google.com>
Reviewed-by: Keiichi Watanabe <keiichiw@chromium.org>
Auto-Submit: Pierre-Clément Tosi <ptosi@google.com>
Commit-Queue: Pierre-Clément Tosi <ptosi@google.com>
Instead of proxying doorbell MMIO writes via the proxy device in crosvm,
register the vring call eventfds with KVM so that the kernel signals the
sibling directly.
BUG=b:240374485
TEST=boot crostini on brya-manatee
Change-Id: I4e33143a38eb01c50fe900a09e8715164792614f
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/3804519
Reviewed-by: Keiichi Watanabe <keiichiw@chromium.org>
Tested-by: David Stevens <stevensd@chromium.org>
Commit-Queue: David Stevens <stevensd@chromium.org>
This CL migrates Tube from directly using a UnixSeqpacket to instead use
a StreamChannel backed by UnixSeqpackets. This CL also migrates users of
the old Tube::new(UnixSeqpacket) function to its renamed counterpart.
Tests are added for new functionality.
This CL enforces that the StreamChannel used by the Tube is backed by a
UnixSeqpacket because the reading and writing behavior of a UnixStream
is different, and it cannot guarantee the same multi-process
concurrent-writers behavior that a UnixSeqpacket can. This is because
the UnixStream does not send data as packets and the size cannot be
queried, so two underlyig writes are required for each Tube write: one
to send a header indicating message size, and one for the actual content
of the sent data. These can be interleaved during concurrent writes,
causing corruption of the Tube's data.
This CL makes the Windows and Unix API closer to each other, with some
differences remaining or added, including Tube::new_from_unix_seqpacket
existing in Unix but necessarily not Windows.
This migration (Tube from UnixSeqpacket to Stream) is intended to move
the Unix Tube API closer to the Windows API, with the eventual goal of
creating a platform-agnostic API for Tube.
Bug: b:231641496
Change-Id: I2422e76b7efa5a20bc40a8d194f5f8390dc71438
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/3818253
Reviewed-by: Alexandre Courbot <acourbot@chromium.org>
Reviewed-by: Noah Gold <nkgold@google.com>
Tested-by: Clarissa Garvey <clarissagarvey@chromium.org>
Commit-Queue: Clarissa Garvey <clarissagarvey@chromium.org>
Add tracking to VmMemoryRequest.execute so that only requests which were
mapped with VfioDmabufMap get unmapped with VfioDmabufUnmap. This fixes
a deadlock that occurs when unregistering memory before the viommu
device is started, which is triggered by VfioPciDevice on ManaTEE.
Note that although a deadlock with registering memory is still possible,
this doesn't seem to be a problem in practice. Registration of such
buffers is driven by the virtio-gpu driver, and thus tends to occur
later during guest boot.
BUG=b:235508487
TEST=boot brya-manatee
Change-Id: I080d27b43d01a01b7d962b62c986d837be72b288
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/3830133
Tested-by: David Stevens <stevensd@chromium.org>
Commit-Queue: Keiichi Watanabe <keiichiw@chromium.org>
Auto-Submit: David Stevens <stevensd@chromium.org>
Commit-Queue: David Stevens <stevensd@chromium.org>
Reviewed-by: Keiichi Watanabe <keiichiw@chromium.org>
Host dma-buf refers to the buffers allocated in host memory and shared
with guest passthrough devices or vCPU through dma-buf interface. To let
guest passthrough devices access the host dma-buf, host needs to pin
those buffers in memory and set up the mapping in vt-d table before
those buffers are exposed to guest and do the unpin thing right after
those buffers get returned from guest.
BUG=b:235508487
TEST=test with a crostini VM equiped with a passthrough GPU, e.g.
--vfio /sys/bus/pci/devices/0000:04:00.0,guest-address="00:1f.0",iommu=viommu
Change-Id: Ie7b3db3a7f7aa8afff03425be16c952db4a5f67b
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/3764931
Commit-Queue: David Stevens <stevensd@chromium.org>
Reviewed-by: David Stevens <stevensd@chromium.org>
Tested-by: David Stevens <stevensd@chromium.org>
Currently we use VfioCommand to hotplug only vfio devices, but
pcie switch will also be hotplugged into the system and other type
of devices could also be added later. So we change VfioCommand to
HotplugCommand and introduce HotplugDeviceInfo structure to track
details of hotplug devices and send it in HotplugCommand.
BUG=b:199986018
TEST=tools/presubmit
Change-Id: I1a7fb316178ffff002a65b4f57f6e857a0993caa
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/3769816
Tested-by: Daniel Verkamp <dverkamp@chromium.org>
Commit-Queue: Daniel Verkamp <dverkamp@chromium.org>
Reviewed-by: Daniel Verkamp <dverkamp@chromium.org>
We don't need to log that a request succeeded.
BUG=b:240470388
TEST=tast run dut multivm.LifecycleShifting.arc_host
verify those lines are gone in messages
Change-Id: I375ca717328323ba4ebaebe87e90673f1e4776e8
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/3805091
Commit-Queue: Charles William Dick <cwd@google.com>
Tested-by: Charles William Dick <cwd@google.com>
Auto-Submit: Charles William Dick <cwd@google.com>
Reviewed-by: Kameron Lutes <kalutes@chromium.org>
Reviewed-by: Junichi Uekawa <uekawa@chromium.org>
This requires the introduction of a BatteryConfig structure to parse,
which is probably a good idea anyway as it reserves some space to add
more battery-related options.
BUG=b:218223240
TEST=cargo test parse_battery
Change-Id: I82db7fe7ce2daeca004e734aaafa542583998561
Reviewed-on: https://chromium-review.googlesource.com/c/crosvm/crosvm/+/3784661
Tested-by: Alexandre Courbot <acourbot@chromium.org>
Reviewed-by: Daniel Verkamp <dverkamp@chromium.org>
Commit-Queue: Alexandre Courbot <acourbot@chromium.org>
