IGVM files can contain an initial VMSA that should be applied to each
vcpu as part of the initial guest state. The sev_features flags are
provided as part of the VMSA structure. However, KVM only allows
sev_features to be set during initialization and not as the guest is
being prepared for launch.
This patch queries KVM for the supported set of sev_features flags and
processes the VP context entries in the IGVM file during kvm_init to
determine any sev_features flags set in the IGVM file. These are then
provided in the call to KVM_SEV_INIT2 to ensure the guest state
matches that specified in the IGVM file.
The igvm process() function is modified to allow a partial processing
of the file during initialization, with only the IGVM_VHT_VP_CONTEXT
fields being processed. This means the function is called twice,
firstly to extract the sev_features then secondly to actually
configure the guest.
Signed-off-by: Roy Hopkins <roy.hopkins@randomman.co.uk>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Stefano Garzarella <sgarzare@redhat.com>
Acked-by: Gerd Hoffman <kraxel@redhat.com>
Tested-by: Stefano Garzarella <sgarzare@redhat.com>
Reviewed-by: Liam Merwick <liam.merwick@oracle.com>
Reviewed-by: Ani Sinha <anisinha@redhat.com>
Link: https://lore.kernel.org/r/b2f986aae04e1da2aee530c9be22a54c0c59a560.1751554099.git.roy.hopkins@randomman.co.uk
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The new cgs_set_guest_policy() function is provided to receive the guest
policy flags, SNP ID block and SNP ID authentication from guest
configuration such as an IGVM file and apply it to the platform prior to
launching the guest.
The policy is used to populate values for the existing 'policy',
'id_block' and 'id_auth' parameters. When provided, the guest policy is
applied and the ID block configuration is used to verify the launch
measurement and signatures. The guest is only successfully started if
the expected launch measurements match the actual measurements and the
signatures are valid.
Signed-off-by: Roy Hopkins <roy.hopkins@randomman.co.uk>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Stefano Garzarella <sgarzare@redhat.com>
Acked-by: Gerd Hoffman <kraxel@redhat.com>
Reviewed-by: Ani Sinha <anisinha@redhat.com>
Link: https://lore.kernel.org/r/99e82ddec4ad2970c790db8bea16ea3f57eb0e53.1751554099.git.roy.hopkins@randomman.co.uk
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Adds a handler for the guest policy initialization IGVM section and
builds an SEV policy based on this information and the ID block
directive if present. The policy is applied using by calling
'set_guest_policy()' on the ConfidentialGuestSupport object.
Signed-off-by: Roy Hopkins <roy.hopkins@randomman.co.uk>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Stefano Garzarella <sgarzare@redhat.com>
Acked-by: Gerd Hoffman <kraxel@redhat.com>
Reviewed-by: Ani Sinha <anisinha@redhat.com>
Link: https://lore.kernel.org/r/57707230bef331b53e9366ce6a23ed25cd6f1293.1751554099.git.roy.hopkins@randomman.co.uk
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The initialization sections in IGVM files contain configuration that
should be applied to the guest platform before it is started. This
includes guest policy and other information that can affect the security
level and the startup measurement of a guest.
This commit introduces handling of the initialization sections during
processing of the IGVM file.
Signed-off-by: Roy Hopkins <roy.hopkins@randomman.co.uk>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Gerd Hoffman <kraxel@redhat.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Link: https://lore.kernel.org/r/9de24fb5df402024b40cbe02de0b13faa7cb4d84.1751554099.git.roy.hopkins@randomman.co.uk
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
For confidential guests a policy can be provided that defines the
security level, debug status, expected launch measurement and other
parameters that define the configuration of the confidential platform.
This commit adds a new function named set_guest_policy() that can be
implemented by each confidential platform, such as AMD SEV to set the
policy. This will allow configuration of the policy from a
multi-platform resource such as an IGVM file without the IGVM processor
requiring specific implementation details for each platform.
Signed-off-by: Roy Hopkins <roy.hopkins@randomman.co.uk>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Reviewed-by: Ani Sinha <anisinha@redhat.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Gerd Hoffman <kraxel@redhat.com>
Link: https://lore.kernel.org/r/d3888a2eb170c8d8c85a1c4b7e99accf3a15589c.1751554099.git.roy.hopkins@randomman.co.uk
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Create an enum entry within FirmwareDevice for 'igvm' to describe that
an IGVM file can be used to map firmware into memory as an alternative
to pre-existing firmware devices.
Signed-off-by: Roy Hopkins <roy.hopkins@randomman.co.uk>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Gerd Hoffman <kraxel@redhat.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Reviewed-by: Ani Sinha <anisinha@redhat.com>
Link: https://lore.kernel.org/r/2eca2611d372facbffa65ee8244cf2d321eb9d17.1751554099.git.roy.hopkins@randomman.co.uk
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
IGVM support has been implemented for Confidential Guests that support
AMD SEV and AMD SEV-ES. Add some documentation that gives some
background on the IGVM format and how to use it to configure a
confidential guest.
Signed-off-by: Roy Hopkins <roy.hopkins@randomman.co.uk>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Reviewed-by: Pankaj Gupta <pankaj.gupta@amd.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Gerd Hoffman <kraxel@redhat.com>
Reviewed-by: Ani Sinha <anisinha@redhat.com>
Link: https://lore.kernel.org/r/b4dc920a30717e19cd79bbbe2cc769f3b9ff3d37.1751554099.git.roy.hopkins@randomman.co.uk
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The ConfidentialGuestSupport object defines a number of virtual
functions that are called during processing of IGVM directives to query
or configure initial guest state. In order to support processing of IGVM
files, these functions need to be implemented by relevant isolation
hardware support code such as SEV.
This commit implements the required functions for SEV-ES and adds
support for processing IGVM files for configuring the guest.
Signed-off-by: Roy Hopkins <roy.hopkins@randomman.co.uk>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Stefano Garzarella <sgarzare@redhat.com>
Acked-by: Gerd Hoffman <kraxel@redhat.com>
Reviewed-by: Ani Sinha <anisinha@redhat.com>
Link: https://lore.kernel.org/r/7145835f729e6195f2fbda308aa90e089a96ae6e.1751554099.git.roy.hopkins@randomman.co.uk
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
When an SEV guest is started, the reset vector and state are
extracted from metadata that is contained in the firmware volume.
In preparation for using IGVM to setup the initial CPU state,
the code has been refactored to populate vmcb_save_area for each
CPU which is then applied during guest startup and CPU reset.
Signed-off-by: Roy Hopkins <roy.hopkins@randomman.co.uk>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Stefano Garzarella <sgarzare@redhat.com>
Acked-by: Gerd Hoffman <kraxel@redhat.com>
Reviewed-by: Pankaj Gupta <pankaj.gupta@amd.com>
Link: https://lore.kernel.org/r/d3c2debca496c4366a278b135f951908f3b9c341.1751554099.git.roy.hopkins@randomman.co.uk
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The x86 segment registers are identified by the X86Seg enumeration which
includes LDTR and TR as well as the normal segment registers. The
function 'cpu_x86_load_seg_cache()' uses the enum to determine which
segment to set. However, specifying R_LDTR or R_TR results in an
out-of-bounds access of the segment array.
Possibly by coincidence, the function does correctly set LDTR or TR in
this case as the structures for these registers immediately follow the
array which is accessed out of bounds.
This patch adds correct handling for R_LDTR and R_TR in the function.
Signed-off-by: Roy Hopkins <roy.hopkins@randomman.co.uk>
Acked-by: Gerd Hoffman <kraxel@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Reviewed-by: Ani Sinha <anisinha@redhat.com>
Link: https://lore.kernel.org/r/95c69253ea4f91107625872d5e3f0c586376771d.1751554099.git.roy.hopkins@randomman.co.uk
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The class function and implementations for updating launch data return
a code in case of error. In some cases an error message is generated and
in other cases, just the error return value is used.
This small refactor adds an 'Error **errp' parameter to all functions
which consistently set an error condition if a non-zero value is
returned.
Signed-off-by: Roy Hopkins <roy.hopkins@randomman.co.uk>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Gerd Hoffman <kraxel@redhat.com>
Reviewed-by: Pankaj Gupta <pankaj.gupta@amd.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Reviewed-by: Ani Sinha <anisinha@redhat.com>
Link: https://lore.kernel.org/r/d59721f7b99cfc87aab71f8f551937e98e983615.1751554099.git.roy.hopkins@randomman.co.uk
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
When using an IGVM file the configuration of the system firmware is
defined by IGVM directives contained in the file. In this case the user
should not configure any pflash devices.
This commit skips initialization of the ROM mode when pflash0 is not set
then checks to ensure no pflash devices have been configured when using
IGVM, exiting with an error message if this is not the case.
Signed-off-by: Roy Hopkins <roy.hopkins@randomman.co.uk>
Acked-by: Gerd Hoffman <kraxel@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Reviewed-by: Pankaj Gupta <pankaj.gupta@amd.com>
Reviewed-by: Ani Sinha <anisinha@redhat.com>
Link: https://lore.kernel.org/r/c6166cfe128933b04003a9288566b7affe170dfe.1751554099.git.roy.hopkins@randomman.co.uk
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
An IGVM file contains configuration of guest state that should be
applied during configuration of the guest, before the guest is started.
This patch allows the user to add an igvm-cfg object to an X86 machine
configuration that allows an IGVM file to be configured that will be
applied to the guest before it is started.
If an IGVM configuration is provided then the IGVM file is processed at
the end of the board initialization, before the state transition to
PHASE_MACHINE_INITIALIZED.
Signed-off-by: Roy Hopkins <roy.hopkins@randomman.co.uk>
Acked-by: Gerd Hoffman <kraxel@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Reviewed-by: Ani Sinha <anisinha@redhat.com>
Link: https://lore.kernel.org/r/23bc66ae4504ba5cf2134826e055b25df3fc9cd9.1751554099.git.roy.hopkins@randomman.co.uk
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Adds an IGVM loader to QEMU which processes a given IGVM file and
applies the directives within the file to the current guest
configuration.
The IGVM loader can be used to configure both confidential and
non-confidential guests. For confidential guests, the
ConfidentialGuestSupport object for the system is used to encrypt
memory, apply the initial CPU state and perform other confidential guest
operations.
The loader is configured via a new IgvmCfg QOM object which allows the
user to provide a path to the IGVM file to process.
Signed-off-by: Roy Hopkins <roy.hopkins@randomman.co.uk>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Gerd Hoffman <kraxel@redhat.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Link: https://lore.kernel.org/r/ae3a07d8f514d93845a9c16bb155c847cb567b0d.1751554099.git.roy.hopkins@randomman.co.uk
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
In preparation for supporting the processing of IGVM files to configure
guests, this adds a set of functions to ConfidentialGuestSupport
allowing configuration of secure virtual machines that can be
implemented for each supported isolation platform type such as Intel TDX
or AMD SEV-SNP. These functions will be called by IGVM processing code
in subsequent patches.
This commit provides a default implementation of the functions that
either perform no action or generate an error when they are called.
Targets that support ConfidentalGuestSupport should override these
implementations.
Signed-off-by: Roy Hopkins <roy.hopkins@randomman.co.uk>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Gerd Hoffman <kraxel@redhat.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Reviewed-by: Ani Sinha <anisinha@redhat.com>
Link: https://lore.kernel.org/r/23e34a106da87427899f93178102e4a6ef50c966.1751554099.git.roy.hopkins@randomman.co.uk
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The IGVM library allows Independent Guest Virtual Machine files to be
parsed and processed. IGVM files are used to configure guest memory
layout, initial processor state and other configuration pertaining to
secure virtual machines.
This adds the --enable-igvm configure option, enabled by default, which
attempts to locate and link against the IGVM library via pkgconfig and
sets CONFIG_IGVM if found.
The library is added to the system_ss target in backends/meson.build
where the IGVM parsing will be performed by the ConfidentialGuestSupport
object.
Signed-off-by: Roy Hopkins <roy.hopkins@randomman.co.uk>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Gerd Hoffman <kraxel@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Reviewed-by: Ani Sinha <anisinha@redhat.com>
Link: https://lore.kernel.org/r/45945a83a638c3f08e68c025f378e7b7f4f6d593.1751554099.git.roy.hopkins@randomman.co.uk
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Gpa is defined in QAPI but never reported to monitor because has_gpa is
never set to ture.
Fix it by setting has_gpa to ture when TDX_REPORT_FATAL_ERROR_GPA_VALID
is set in error_code.
Fixes: 6e250463b0 ("i386/tdx: Wire TDX_REPORT_FATAL_ERROR with GuestPanic facility")
Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Link: https://lore.kernel.org/r/20250710035538.303136-1-zhenzhong.duan@intel.com
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Record the interrupt vector and the apic id of the vcpu that calls
TDVMCALL_SETUP_EVENT_NOTIFY_INTERRUPT.
Inject the interrupt to TD guest to notify the completion of <GetQuote>
when notify interrupt vector is valid.
Signed-off-by: Xiaoyao Li <xiaoyao.li@intel.com>
Link: https://lore.kernel.org/r/20250703024021.3559286-5-xiaoyao.li@intel.com
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
KVM reports the supported TDVMCALL sub leafs in TDX capabilities.
one for kernel-supported
TDVMCALLs (userspace can set those blindly) and one for user-supported
TDVMCALLs (userspace can set those if it knows how to handle them)
Signed-off-by: Xiaoyao Li <xiaoyao.li@intel.com>
Link: https://lore.kernel.org/r/20250703024021.3559286-4-xiaoyao.li@intel.com
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
GHCI is finalized with the <GetQuote> being one of the base VMCALLs, and
not enuemrated via <GetTdVmCallInfo>.
Adjust tdx_handle_get_tdvmcall_info() to match with GHCI.
Opportunistically fix the wrong indentation and explicitly set the
ret to TDG_VP_VMCALL_SUCCESS (in case KVM leaves unexpected value).
Signed-off-by: Xiaoyao Li <xiaoyao.li@intel.com>
Link: https://lore.kernel.org/r/20250703024021.3559286-2-xiaoyao.li@intel.com
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Simplify the accelerators' cpu_instance_init callbacks by doing all
host-cpu setup in a single function.
Based-on: <20250711000603.438312-1-pbonzini@redhat.com>
Cc: Xiaoyao Li <xiaoyao.li@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
With the reordering of instance_post_init callbacks that is new in 10.1
accel_cpu_instance_init must execute in .instance_init as is already
the case for RISC-V. Otherwise, for example, setting the vendor
property is broken when using KVM or Hypervisor.framework, because
KVM sets it *after* the user's value is set by DeviceState's
intance_post_init callback.
Reported-by: Like Xu <like.xu.linux@gmail.com>
Reported-by: Dongli Zhang <dongli.zhang@oracle.com>
Reviewed-by: Xiaoyao Li <xiaoyao.li@intel.com>
Reviewed-by: Zhao Liu <zhao1.liu@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The PMU feature is only supported by KVM, so move it there. And since
all accelerators other than TCG overwrite the vendor, set it in
max_x86_cpu_initfn only if it has not been initialized by the
superclass. This makes it possible to run max_x86_cpu_initfn
after accelerator init.
Reviewed-by: Xiaoyao Li <xiaoyao.li@intel.com>
Reviewed-by: Zhao Liu <zhao1.liu@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
NVMM and WHPX are virtualizers, and therefore they need to use
(at least by default) the host vendor for the guest CPUID.
Add a cpu_instance_init implementation to these accelerators.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
In the linux-user do_fork() function we try to set the FD_CLOEXEC
flag on a pidfd like this:
fcntl(pid_fd, F_SETFD, fcntl(pid_fd, F_GETFL) | FD_CLOEXEC);
This has two problems:
(1) it doesn't check errors, which Coverity complains about
(2) we use F_GETFL when we mean F_GETFD
Deal with both of these problems by using qemu_set_cloexec() instead.
That function will assert() if the fcntls fail, which is fine (we are
inside fork_start()/fork_end() so we know nothing can mess around
with our file descriptors here, and we just got this one from
pidfd_open()).
(As we are touching the if() statement here, we correct the
indentation.)
Coverity: CID 1508111
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-ID: <20250711141217.1429412-1-peter.maydell@linaro.org>
Avoid ubsan failure with clang-20,
tcg.h:715:19: runtime error: applying non-zero offset 64 to null pointer
by not using pointers.
Acked-by: Ilya Leoshkevich <iii@linux.ibm.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Recent patch [1] renames the save_live_complete_precopy handler to
save_complete, as the machine is not live in most cases when this
handler is executed. The same is true also for
save_live_complete_precopy_thread, therefore this patch removes the
"live" keyword from the handler itself and related types to keep the
naming unified.
In contrast to save_complete, this handler is only executed at the end
of precopy, therefore the "precopy" keyword is retained.
[1]: https://lore.kernel.org/all/20250613140801.474264-7-peterx@redhat.com/
Cc: Alex Williamson <alex.williamson@redhat.com>
Cc: Cédric Le Goater <clg@redhat.com>
Signed-off-by: Juraj Marcin <jmarcin@redhat.com>
Link: https://lore.kernel.org/r/20250626085235.294690-1-jmarcin@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Fabiano Rosas <farosas@suse.de>
Add the latency distribution too for blocktime, using order-of-two buckets.
It accounts for all the faults, from either vCPU or non-vCPU threads. With
prior rework, it's very easy to achieve by adding an array to account for
faults in each buckets.
Sample output for HMP (while for QMP it's simply an array):
Postcopy Latency Distribution:
[ 1 us - 2 us ]: 0
[ 2 us - 4 us ]: 0
[ 4 us - 8 us ]: 1
[ 8 us - 16 us ]: 2
[ 16 us - 32 us ]: 2
[ 32 us - 64 us ]: 3
[ 64 us - 128 us ]: 10169
[ 128 us - 256 us ]: 50151
[ 256 us - 512 us ]: 12876
[ 512 us - 1 ms ]: 97
[ 1 ms - 2 ms ]: 42
[ 2 ms - 4 ms ]: 44
[ 4 ms - 8 ms ]: 93
[ 8 ms - 16 ms ]: 138
[ 16 ms - 32 ms ]: 0
[ 32 ms - 65 ms ]: 0
[ 65 ms - 131 ms ]: 0
[ 131 ms - 262 ms ]: 0
[ 262 ms - 524 ms ]: 0
[ 524 ms - 1 sec ]: 0
[ 1 sec - 2 sec ]: 0
[ 2 sec - 4 sec ]: 0
[ 4 sec - 8 sec ]: 0
[ 8 sec - 16 sec ]: 0
Cc: Markus Armbruster <armbru@redhat.com>
Acked-by: Dr. David Alan Gilbert <dave@treblig.org>
Reviewed-by: Fabiano Rosas <farosas@suse.de>
Link: https://lore.kernel.org/r/20250613141217.474825-15-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Fabiano Rosas <farosas@suse.de>
When used to report page fault latencies, the blocktime feature can be
almost useless when KVM async page fault is enabled, because in most cases
such remote fault will kickoff async page faults, then it's not trackable
from blocktime layer.
After all these recent rewrites to blocktime layer, it's finally so easy to
also support tracking non-vCPU faults. It'll be even faster if we could
always index fault records with TIDs, unfortunately we need to maintain the
blocktime API which report things in vCPU indexes.
Of course this can work not only for kworkers, but also any guest accesses
that may reach a missing page, for example, very likely when in the QEMU
main thread too (and all other threads whenever applicable).
In this case, we don't care about "how long the threads are blocked", but
we only care about "how long the fault will be resolved".
Cc: Markus Armbruster <armbru@redhat.com>
Cc: Dr. David Alan Gilbert <dave@treblig.org>
Reviewed-by: Fabiano Rosas <farosas@suse.de>
Tested-by: Mario Casquero <mcasquer@redhat.com>
Link: https://lore.kernel.org/r/20250613141217.474825-14-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Fabiano Rosas <farosas@suse.de>
Currently, the postcopy blocktime feature maintains vCPU fault information
using an array (vcpu_addr[]). It has two issues.
Issue 1: Performance Concern
============================
The old algorithm was almost OK and fast on inserts, except that the lookup
is slow and won't scale if there are a lot of vCPUs: when a page is copied
during postcopy, mark_postcopy_blocktime_end() will walk the whole array
trying to find which vCPUs are blocked by the address. So it needs
constant O(N) walk for each page resolution.
Alexey (the author of postcopy blocktime) mentioned the perf issue and how
to optimize it in a piece of comment in the page resolution path. The
comment was (interestingly..) not complete, but it's relatively clear what
he wanted to say about this perf issue.
Issue 2: Wrong Accounting on re-entrancies
==========================================
People might think that each vCPU should only and always get one fault at a
time, so that when the blocktime layer captured one fault on one vCPU, we
should never see another fault message on this vCPU.
It's almost correct, except in some extreme rare cases.
Case 1: it's possible the fault thread processes the userfaultfd messages
too fast so it can see >1 messages on one vCPU before the previous one was
resolved.
Case 2: it's theoretically also possible one vCPU can get even more than
one message on the same fault address if a fault is retried by the
kernel (e.g., handle_userfault() got interrupted before page resolution).
As this info might be important, instead of using commit message, I put
more details into the code as comment, when introducing an array
maintaining concurrent faults on one vCPU. Please refer to the comments
for details on both cases, especially case 1 which can be tricky.
Case 1 sounds rare, but it can be easily reproduced locally for me when we
run blocktime together with the migration-test on the vanilla postcopy.
New Design
==========
This patch should do almost what Alexey mentioned, but slightly
differently: instead of having an array to maintain vCPU fault addresses,
for each of the fault message we push a message into a hash, indexed by the
fault address.
With the hash, it can replace the old two structs: both the vcpu_addr[]
array, and also the array to store the start time of the fault. However
due to above we need one more counter array to account concurrent faults on
the same vCPU - that should even be needed in the old code, it's just that
the old code was buggy and it will blindly overwrite an existing
entry.. now we'll start to really track everything.
The hash structure might be more efficient than tree to maintain such
addr->(cpu, fault_time) information, so that the insert() and lookup()
paths should ideally both be ~O(1). After all, we do not need to sort.
Here we need to do one remove() though after the lookup(). It could be
slow but only if many vCPUs faulted exactly on the same address (so when
the list of cpu entries is long), which should be unlikely. Even with that,
it's still a worst case O(N) (consider 400 vCPUs faulted on the same
address and how likely is it..) rather than a constant O(N) complexity.
When at it, touch up the tracepoints to make them slightly more useful.
One tracepoint is added when walking all the fault entries.
Reviewed-by: Fabiano Rosas <farosas@suse.de>
Link: https://lore.kernel.org/r/20250613141217.474825-13-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Fabiano Rosas <farosas@suse.de>
The variable vcpu_total_blocktime isn't easy to follow. In reality, it
wants to capture the case where all vCPUs are stopped, and now there will
be some vCPUs starts running.
The name now starts to conflict with vcpu_blocktime_total[], meanwhile it's
actually not necessary to have the variable at all: since nobody is
touching smp_cpus_down except ourselves, we can safely do the calculation
at the end before decrementing smp_cpus_down.
Hopefully this makes the logic easier to read, side benefit is we drop one
temp var.
Reviewed-by: Fabiano Rosas <farosas@suse.de>
Link: https://lore.kernel.org/r/20250613141217.474825-12-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Fabiano Rosas <farosas@suse.de>
Looking up the vCPU index for each fault can be expensive when there're
hundreds of vCPUs. Provide a cache for tid->vcpu instead with a hash
table, then lookup from there.
When at it, add another counter to record how many non-vCPU faults it gets.
For example, the main thread can also access a guest page that was missing.
These kind of faults are not accounted by blocktime so far.
Reviewed-by: Fabiano Rosas <farosas@suse.de>
Link: https://lore.kernel.org/r/20250613141217.474825-11-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Fabiano Rosas <farosas@suse.de>
Before this patch, the blocktime context can be created very early, because
postcopy_ram_supported_by_host() <- migrate_caps_check() can happen during
migration object init.
The trick here is the blocktime context needs system vCPU information,
which seems to be possible to change after that point. I didn't verify it,
but it doesn't sound right.
Now move it out and initialize the context only when postcopy listen
starts. That is already during a migration so it should be guaranteed the
vCPU topology can never change on both sides.
While at it, assert that the ctx isn't created instead this time; the old
"if" trick isn't needed when we're sure it will only happen once now.
Reviewed-by: Fabiano Rosas <farosas@suse.de>
Link: https://lore.kernel.org/r/20250613141217.474825-10-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Fabiano Rosas <farosas@suse.de>
Blocktime so far only cares about the time one vcpu (or the whole system)
got blocked. It would be also be helpful if it can also report the latency
of page requests, which could be very sensitive during postcopy.
Blocktime itself is sometimes not very important, especially when one
thinks about KVM async PF support, which means vCPUs are literally almost
not blocked at all because the guest OS is smart enough to switch to
another task when a remote fault is needed.
However, latency is still sensitive and important because even if the guest
vCPU is running on threads that do not need a remote fault, the workload
that accesses some missing page is still affected.
Add two entries to the report, showing how long it takes to resolve a
remote fault. Mention in the QAPI doc that this is not the real average
fault latency, but only the ones that was requested for a remote fault.
Unwrap get_vcpu_blocktime_list() so we don't need to walk the list twice,
meanwhile add the entry checks in qtests for all postcopy tests.
Cc: Markus Armbruster <armbru@redhat.com>
Cc: Dr. David Alan Gilbert <dave@treblig.org>
Reviewed-by: Fabiano Rosas <farosas@suse.de>
Tested-by: Mario Casquero <mcasquer@redhat.com>
Link: https://lore.kernel.org/r/20250613141217.474825-9-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Fabiano Rosas <farosas@suse.de>
Add a field to count how many remote faults one vCPU has taken. So far
it's still not used, but will be soon.
Reviewed-by: Fabiano Rosas <farosas@suse.de>
Link: https://lore.kernel.org/r/20250613141217.474825-8-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Fabiano Rosas <farosas@suse.de>
With 64-bit fields, it is trivial. The caution is when exposing any values
in QMP, it was still declared with milliseconds (ms). Hence it's needed to
do the convertion when exporting the values to existing QMP queries.
Reviewed-by: Fabiano Rosas <farosas@suse.de>
Link: https://lore.kernel.org/r/20250613141217.474825-7-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Fabiano Rosas <farosas@suse.de>
Now with 64bits, the offseting using start_time is not needed anymore,
because the array can always remember the whole timestamp.
Then drop the unused parameter in get_low_time_offset() altogether.
Reviewed-by: Fabiano Rosas <farosas@suse.de>
Link: https://lore.kernel.org/r/20250613141217.474825-6-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Fabiano Rosas <farosas@suse.de>
I am guessing it was used to be 32bits because of the atomic ops. Now all
the atomic ops are gone and we're protected by a mutex instead, it's ok we
can switch to 64 bits.
Reasons to move over:
- Allow further patches to change the unit from ms to us: with postcopy
preempt mode, we're really into hundreds of microseconds level on
blocktime. We'd better be able to trap those.
- This also paves way for some other tricks that the original version
used to avoid overflows, e.g., start_time was almost only useful before
to make sure the sampled timestamp won't overflow a 32-bit field.
- This prepares further reports on top of existing data collected,
e.g. average page fault latencies. When average operation is taken into
account, milliseconds are simply too coarse grained.
When at it:
- Rename page_fault_vcpu_time to vcpu_blocktime_start.
- Rename vcpu_blocktime to vcpu_blocktime_total.
- Touch up the trace-events to not dump blocktime ctx pointer
Reviewed-by: Fabiano Rosas <farosas@suse.de>
Link: https://lore.kernel.org/r/20250613141217.474825-5-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Fabiano Rosas <farosas@suse.de>
The postcopy blocktime feature was tricky that it used quite some atomic
operations over quite a few arrays and vars, without explaining how that
would be thread safe. The thread safety here is about concurrency between
the fault thread and the fault resolution threads, possible to access the
same chunk of data. All these atomic ops can be expensive too before
knowing clearly how it works.
OTOH, postcopy has one page_request_mutex used to serialize the received
bitmap updates. So far it's ok - we don't yet have a lot of threads
contending the lock. It might change after multifd will be supported, but
that's a separate story. What is important is, with that mutex, it's
pretty lightweight to move all the blocktime maintenance into the mutex
critical section. It's because the blocktime layer is lightweighted:
almost "remember which vcpu faulted on which address", and "ok we get some
fault resolved, calculate how long it takes". It's also an optional
feature for now (but I have thought of changing that, maybe in the future).
Let's push the blocktime layer into the mutex, so that it's always
thread-safe even without any atomic ops.
To achieve that, I'll need to add a tid parameter on fault path so that
it'll start to pass the faulted thread ID into deeper the stack, but not
too deep. When at it, add a comment for the shared fault handler (for
example, vhost-user devices running with postcopy), to mention a TODO. One
reason it might not be trivial is that vhost-user's userfaultfds should be
opened by vhost-user process, so it's pretty hard to control making sure
the TID feature will be around. It wasn't supported before, so keep it
like that for now.
Now we should be as ease when everything is protected by a mutex that we
always take anyway.
One side effect: we can finally remove one ramblock_recv_bitmap_test() in
mark_postcopy_blocktime_begin(), which was pretty weird and which also
includes a weird (but maybe necessary.. but maybe not?) operation to inject
a blocktime entry then quickly erase it.. When we're with the mutex, and
when we make sure it's invoked after checking the receive bitmap, it's not
needed anymore. Instead, we assert.
As another side effect, this paves way for removing all atomic ops in all
the mem accesses in blocktime layer.
Note that we need a stub for mark_postcopy_blocktime_begin() for Windows
builds.
Reviewed-by: Fabiano Rosas <farosas@suse.de>
Link: https://lore.kernel.org/r/20250613141217.474825-3-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Fabiano Rosas <farosas@suse.de>
This is a follow up on the other commit "migration/ram: avoid to do log
clear in the last round" but for postcopy.
https://lore.kernel.org/r/20250514115827.3216082-1-yanfei.xu@bytedance.com
I can observe more than 10% reduction of average page fault latency during
postcopy phase with this optimization:
Before: 268.00us (+-1.87%)
After: 232.67us (+-2.01%)
The test was done with a 16GB VM with 80 vCPUs, running a workload that
busy random writes to 13GB memory.
Cc: Yanfei Xu <yanfei.xu@bytedance.com>
Reviewed-by: Fabiano Rosas <farosas@suse.de>
Link: https://lore.kernel.org/r/20250613140801.474264-12-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Fabiano Rosas <farosas@suse.de>
There're a few things off here in that logic, rewrite it. When at it, add
rich comment to explain each of the decisions.
Since this is very sensitive path for migration, below are the list of
things changed with their reasonings.
(1) Exact pending size is only needed for precopy not postcopy
Fundamentally it's because "exact" version only does one more deep
sync to fetch the pending results, while in postcopy's case it's
never going to sync anything more than estimate as the VM on source
is stopped.
(2) Do _not_ rely on threshold_size anymore to decide whether postcopy
should complete
threshold_size was calculated from the expected downtime and
bandwidth only during precopy as an efficient way to decide when to
switchover. It's not sensible to rely on threshold_size in postcopy.
For precopy, if switchover is decided, the migration will complete
soon. It's not true for postcopy. Logically speaking, postcopy
should only complete the migration if all pending data is flushed.
Here it used to work because save_complete() used to implicitly
contain save_live_iterate() when there's pending size.
Even if that looks benign, having RAMs to be migrated in postcopy's
save_complete() has other bad side effects:
(a) Since save_complete() needs to be run once at a time, it means
when moving RAM there's no way moving other things (rather than
round-robin iterating the vmstate handlers like what we do with
ITERABLE phase). Not an immediate concern, but it may stop working
in the future when there're more than one iterables (e.g. vfio
postcopy).
(b) postcopy recovery, unfortunately, only works during ITERABLE
phase. IOW, if src QEMU moves RAM during postcopy's save_complete()
and network failed, then it'll crash both QEMUs... OTOH if it failed
during iteration it'll still be recoverable. IOW, this change should
further reduce the window QEMU split brain and crash in extreme cases.
If we enable the ram_save_complete() tracepoints, we'll see this
before this patch:
1267959@1748381938.294066:ram_save_complete dirty=9627, done=0
1267959@1748381938.308884:ram_save_complete dirty=0, done=1
It means in this migration there're 9627 pages migrated at complete()
of postcopy phase.
After this change, all the postcopy RAM should be migrated in iterable
phase, rather than save_complete():
1267959@1748381938.294066:ram_save_complete dirty=0, done=0
1267959@1748381938.308884:ram_save_complete dirty=0, done=1
(3) Adjust when to decide to switch to postcopy
This shouldn't be super important, the movement makes sure there's
only one in_postcopy check, then we are clear on what we do with the
two completely differnt use cases (precopy v.s. postcopy).
(4) Trivial touch up on threshold_size comparision
Which changes:
"(!pending_size || pending_size < s->threshold_size)"
into:
"(pending_size <= s->threshold_size)"
Reviewed-by: Juraj Marcin <jmarcin@redhat.com>
Reviewed-by: Fabiano Rosas <farosas@suse.de>
Link: https://lore.kernel.org/r/20250613140801.474264-11-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Fabiano Rosas <farosas@suse.de>
Take notes on start/end state of dirty pages for the whole system.
Reviewed-by: Juraj Marcin <jmarcin@redhat.com>
Link: https://lore.kernel.org/r/20250613140801.474264-10-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Fabiano Rosas <farosas@suse.de>
The check over PAGE_DIRTY_FOUND isn't necessary. We could indent one less
and assert that instead.
Reviewed-by: Juraj Marcin <jmarcin@redhat.com>
Link: https://lore.kernel.org/r/20250613140801.474264-9-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Fabiano Rosas <farosas@suse.de>
Since we use the same save_complete() hook for both precopy and postcopy,
add a set of helpers to invoke the hook() to dedup the code.
Reviewed-by: Juraj Marcin <jmarcin@redhat.com>
Reviewed-by: Fabiano Rosas <farosas@suse.de>
Link: https://lore.kernel.org/r/20250613140801.474264-8-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Fabiano Rosas <farosas@suse.de>
Now after merging the precopy and postcopy version of complete() hook,
rename the precopy version from save_live_complete_precopy() to
save_complete().
Dropping the "live" when at it, because it's in most cases not live when
happening (in precopy).
No functional change intended.
Reviewed-by: Juraj Marcin <jmarcin@redhat.com>
Reviewed-by: Fabiano Rosas <farosas@suse.de>
Link: https://lore.kernel.org/r/20250613140801.474264-7-peterx@redhat.com
[peterx: squash the fixup that covers a few more doc spots, per Juraj]
Signed-off-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Fabiano Rosas <farosas@suse.de>
The hook is only defined in two vmstate users ("ram" and "block dirty
bitmap"), meanwhile both of them define the hook exactly the same as the
precopy version. Hence, this postcopy version isn't needed.
No functional change intended.
Reviewed-by: Juraj Marcin <jmarcin@redhat.com>
Reviewed-by: Fabiano Rosas <farosas@suse.de>
Link: https://lore.kernel.org/r/20250613140801.474264-6-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Fabiano Rosas <farosas@suse.de>
