target/i386: document incorrect semantics of watchpoint following MOV/POP SS

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: fix TF/RF handling for HLT

HLT uses DISAS_NORETURN because the corresponding helper calls
cpu_loop_exit().  However, while gen_eob() clears HF_RF_MASK and
synthesizes a #DB exception if single-step is active, none of this is
done by HLT.  Note that the single-step trap is generated after the halt
is finished.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: fix INHIBIT_IRQ/TF/RF handling for PAUSE

PAUSE uses DISAS_NORETURN because the corresponding helper
calls cpu_loop_exit().  However, while HLT clear HF_INHIBIT_IRQ_MASK
to correctly handle "STI; HLT", the same is missing from PAUSE.
And also gen_eob() clears HF_RF_MASK and synthesizes a #DB exception
if single-step is active; none of this is done by HLT and PAUSE.
Start fixing PAUSE, HLT will follow.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: fix INHIBIT_IRQ/TF/RF handling for VMRUN

From vm entry to exit, VMRUN is handled as a single instruction.  It
uses DISAS_NORETURN in order to avoid processing TF or RF before
the first instruction executes in the guest.  However, the corresponding
handling is missing in vmexit.  Add it, and at the same time reorganize
the comments with quotes from the manual about the tasks performed
by a #VMEXIT.

Another gen_eob() task that is missing in VMRUN is preparing the
HF_INHIBIT_IRQ flag for the next instruction, in this case by loading
it from the VMCB control state.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: disable/enable breakpoints on vmentry/vmexit

If the required DR7 (either from the VMCB or from the host save
area) disables a breakpoint that was enabled prior to vmentry
or vmexit, it is left enabled and will trigger EXCP_DEBUG.
This causes a spurious #DB on the next crossing of the breakpoint.

To disable it, vmentry/vmexit must use cpu_x86_update_dr7
to load DR7.

Because cpu_x86_update_dr7 takes a 32-bit argument, check
reserved bits prior to calling cpu_x86_update_dr7, and do the
same for DR6 as well for consistency.

This scenario is tested by the "host_rflags" test in kvm-unit-tests.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: implement DR7.GD

DR7.GD triggers a #DB exception on any access to debug registers.
The GD bit is cleared so that the #DB handler itself can access
the debug registers.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: cleanup PAUSE helpers

Use decode.c's support for intercepts, doing the check in TCG-generated
code rather than the helper.  This is cleaner because it allows removing
the eip_addend argument to helper_pause(), even though it adds a bit of
bloat for opcode 0x90's new decoding function.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: cleanup HLT helpers

Use decode.c's support for intercepts, doing the check in TCG-generated
code rather than the helper.  This is cleaner because it allows removing
the eip_addend argument to helper_hlt().

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: fix implementation of ICEBP

ICEBP generates a trap-like exception, while gen_exception() produces
a fault.  Resurrect gen_update_eip_next() to implement the desired
semantics.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: fix pushed value of EFLAGS.RF

When preparing an exception stack frame for a fault exception, the value
pushed for RF is 1.  Take that into account.  The same should be true
of interrupts for repeated string instructions, but the situation there
is complicated.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

Merge tag 'for-upstream' of https://gitlab.com/bonzini/qemu into staging

* virtio-blk: remove SCSI passthrough functionality
* require x86-64-v2 baseline ISA
* SEV-SNP host support
* fix xsave.flat with TCG
* fixes for CPUID checks done by TCG

# -----BEGIN PGP SIGNATURE-----
#
# iQFIBAABCAAyFiEE8TM4V0tmI4mGbHaCv/vSX3jHroMFAmZgKVYUHHBib256aW5p
# QHJlZGhhdC5jb20ACgkQv/vSX3jHroPKYgf/QkWrNXdjjD3yAsv5LbJFVTVyCYW3
# b4Iax29kEDy8k9wbzfLxOfIk9jXIjmbOMO5ZN9LFiHK6VJxbXslsMh6hm50M3xKe
# 49X1Rvf9YuVA7KZX+dWkEuqLYI6Tlgj3HaCilYWfXrjyo6hY3CxzkPV/ChmaeYlV
# Ad4Y8biifoUuuEK8OTeTlcDWLhOHlFXylG3AXqULsUsXp0XhWJ9juXQ60eATv/W4
# eCEH7CSmRhYFu2/rV+IrWFYMnskLRTk1OC1/m6yXGPKOzgnOcthuvQfiUgPkbR/d
# llY6Ni5Aaf7+XX3S7Avcyvoq8jXzaaMzOrzL98rxYGDR1sYBYO+4h4ZToA==
# =qQeP
# -----END PGP SIGNATURE-----
# gpg: Signature made Wed 05 Jun 2024 02:01:10 AM PDT
# gpg:                using RSA key F13338574B662389866C7682BFFBD25F78C7AE83
# gpg:                issuer "pbonzini@redhat.com"
# gpg: Good signature from "Paolo Bonzini <bonzini@gnu.org>" [full]
# gpg:                 aka "Paolo Bonzini <pbonzini@redhat.com>" [full]

* tag 'for-upstream' of https://gitlab.com/bonzini/qemu: (46 commits)
  hw/i386: Add support for loading BIOS using guest_memfd
  hw/i386/sev: Use guest_memfd for legacy ROMs
  memory: Introduce memory_region_init_ram_guest_memfd()
  i386/sev: Allow measured direct kernel boot on SNP
  i386/sev: Reorder struct declarations
  i386/sev: Extract build_kernel_loader_hashes
  i386/sev: Enable KVM_HC_MAP_GPA_RANGE hcall for SNP guests
  i386/kvm: Add KVM_EXIT_HYPERCALL handling for KVM_HC_MAP_GPA_RANGE
  i386/sev: Invoke launch_updata_data() for SNP class
  i386/sev: Invoke launch_updata_data() for SEV class
  hw/i386/sev: Add support to encrypt BIOS when SEV-SNP is enabled
  i386/sev: Add support for SNP CPUID validation
  i386/sev: Add support for populating OVMF metadata pages
  hw/i386/sev: Add function to get SEV metadata from OVMF header
  i386/sev: Set CPU state to protected once SNP guest payload is finalized
  i386/sev: Add handling to encrypt/finalize guest launch data
  i386/sev: Add the SNP launch start context
  i386/sev: Update query-sev QAPI format to handle SEV-SNP
  i386/sev: Add a class method to determine KVM VM type for SNP guests
  i386/sev: Don't return launch measurements for SEV-SNP guests
  ...

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

hw/i386: Add support for loading BIOS using guest_memfd

When guest_memfd is enabled, the BIOS is generally part of the initial
encrypted guest image and will be accessed as private guest memory. Add
the necessary changes to set up the associated RAM region with a
guest_memfd backend to allow for this.

Current support centers around using -bios to load the BIOS data.
Support for loading the BIOS via pflash requires additional enablement
since those interfaces rely on the use of ROM memory regions which make
use of the KVM_MEM_READONLY memslot flag, which is not supported for
guest_memfd-backed memslots.

Signed-off-by: Michael Roth <michael.roth@amd.com>
Signed-off-by: Pankaj Gupta <pankaj.gupta@amd.com>
Message-ID: <20240530111643.1091816-29-pankaj.gupta@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

hw/i386/sev: Use guest_memfd for legacy ROMs

Current SNP guest kernels will attempt to access these regions with
with C-bit set, so guest_memfd is needed to handle that. Otherwise,
kvm_convert_memory() will fail when the guest kernel tries to access it
and QEMU attempts to call KVM_SET_MEMORY_ATTRIBUTES to set these ranges
to private.

Whether guests should actually try to access ROM regions in this way (or
need to deal with legacy ROM regions at all), is a separate issue to be
addressed on kernel side, but current SNP guest kernels will exhibit
this behavior and so this handling is needed to allow QEMU to continue
running existing SNP guest kernels.

Signed-off-by: Michael Roth <michael.roth@amd.com>
[pankaj: Added sev_snp_enabled() check]
Signed-off-by: Pankaj Gupta <pankaj.gupta@amd.com>
Message-ID: <20240530111643.1091816-28-pankaj.gupta@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

memory: Introduce memory_region_init_ram_guest_memfd()

Introduce memory_region_init_ram_guest_memfd() to allocate private
guset memfd on the MemoryRegion initialization. It's for the use case of
TDVF, which must be private on TDX case.

Signed-off-by: Xiaoyao Li <xiaoyao.li@intel.com>
Signed-off-by: Michael Roth <michael.roth@amd.com>
Signed-off-by: Pankaj Gupta <pankaj.gupta@amd.com>
Message-ID: <20240530111643.1091816-4-pankaj.gupta@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

i386/sev: Allow measured direct kernel boot on SNP

In SNP, the hashes page designated with a specific metadata entry
published in AmdSev OVMF.

Therefore, if the user enabled kernel hashes (for measured direct boot),
QEMU should prepare the content of hashes table, and during the
processing of the metadata entry it copy the content into the designated
page and encrypt it.

Note that in SNP (unlike SEV and SEV-ES) the measurements is done in
whole 4KB pages.  Therefore QEMU zeros the whole page that includes the
hashes table, and fills in the kernel hashes area in that page, and then
encrypts the whole page.  The rest of the page is reserved for SEV
launch secrets which are not usable anyway on SNP.

If the user disabled kernel hashes, QEMU pre-validates the kernel hashes
page as a zero page.

Signed-off-by: Dov Murik <dovmurik@linux.ibm.com>
Signed-off-by: Michael Roth <michael.roth@amd.com>
Signed-off-by: Pankaj Gupta <pankaj.gupta@amd.com>
Message-ID: <20240530111643.1091816-24-pankaj.gupta@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

i386/sev: Reorder struct declarations

Move the declaration of PaddedSevHashTable before SevSnpGuest so
we can add a new such field to the latter.

No functional change intended.

Signed-off-by: Dov Murik <dovmurik@linux.ibm.com>
Signed-off-by: Michael Roth <michael.roth@amd.com>
Signed-off-by: Pankaj Gupta <pankaj.gupta@amd.com>
Message-ID: <20240530111643.1091816-23-pankaj.gupta@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

i386/sev: Extract build_kernel_loader_hashes

Extract the building of the kernel hashes table out from
sev_add_kernel_loader_hashes() to allow building it in
other memory areas (for SNP support).

No functional change intended.

Signed-off-by: Dov Murik <dovmurik@linux.ibm.com>
Signed-off-by: Michael Roth <michael.roth@amd.com>
Signed-off-by: Pankaj Gupta <pankaj.gupta@amd.com>
Message-ID: <20240530111643.1091816-22-pankaj.gupta@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

i386/sev: Enable KVM_HC_MAP_GPA_RANGE hcall for SNP guests

KVM will forward GHCB page-state change requests to userspace in the
form of KVM_HC_MAP_GPA_RANGE, so make sure the hypercall handling is
enabled for SNP guests.

Signed-off-by: Michael Roth <michael.roth@amd.com>
Signed-off-by: Pankaj Gupta <pankaj.gupta@amd.com>
Message-ID: <20240530111643.1091816-32-pankaj.gupta@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

i386/kvm: Add KVM_EXIT_HYPERCALL handling for KVM_HC_MAP_GPA_RANGE

KVM_HC_MAP_GPA_RANGE will be used to send requests to userspace for
private/shared memory attribute updates requested by the guest.
Implement handling for that use-case along with some basic
infrastructure for enabling specific hypercall events.

Signed-off-by: Michael Roth <michael.roth@amd.com>
Signed-off-by: Pankaj Gupta <pankaj.gupta@amd.com>
Message-ID: <20240530111643.1091816-31-pankaj.gupta@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

i386/sev: Invoke launch_updata_data() for SNP class

Invoke as sev_snp_launch_update_data() for SNP object.

Signed-off-by: Pankaj Gupta <pankaj.gupta@amd.com>
Message-ID: <20240530111643.1091816-27-pankaj.gupta@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

i386/sev: Invoke launch_updata_data() for SEV class

Add launch_update_data() in SevCommonStateClass and
invoke as sev_launch_update_data() for SEV object.

Signed-off-by: Pankaj Gupta <pankaj.gupta@amd.com>
Message-ID: <20240530111643.1091816-26-pankaj.gupta@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

hw/i386/sev: Add support to encrypt BIOS when SEV-SNP is enabled

As with SEV, an SNP guest requires that the BIOS be part of the initial
encrypted/measured guest payload. Extend sev_encrypt_flash() to handle
the SNP case and plumb through the GPA of the BIOS location since this
is needed for SNP.

Signed-off-by: Brijesh Singh <brijesh.singh@amd.com>
Signed-off-by: Michael Roth <michael.roth@amd.com>
Signed-off-by: Pankaj Gupta <pankaj.gupta@amd.com>
Message-ID: <20240530111643.1091816-25-pankaj.gupta@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

i386/sev: Add support for SNP CPUID validation

SEV-SNP firmware allows a special guest page to be populated with a
table of guest CPUID values so that they can be validated through
firmware before being loaded into encrypted guest memory where they can
be used in place of hypervisor-provided values[1].

As part of SEV-SNP guest initialization, use this interface to validate
the CPUID entries reported by KVM_GET_CPUID2 prior to initial guest
start and populate the CPUID page reserved by OVMF with the resulting
encrypted data.

[1] SEV SNP Firmware ABI Specification, Rev. 0.8, 8.13.2.6

Signed-off-by: Michael Roth <michael.roth@amd.com>
Signed-off-by: Pankaj Gupta <pankaj.gupta@amd.com>
Message-ID: <20240530111643.1091816-21-pankaj.gupta@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

i386/sev: Add support for populating OVMF metadata pages

OVMF reserves various pages so they can be pre-initialized/validated
prior to launching the guest. Add support for populating these pages
with the expected content.

Signed-off-by: Brijesh Singh <brijesh.singh@amd.com>
Signed-off-by: Michael Roth <michael.roth@amd.com>
Co-developed-by: Pankaj Gupta <pankaj.gupta@amd.com>
Signed-off-by: Pankaj Gupta <pankaj.gupta@amd.com>
Message-ID: <20240530111643.1091816-20-pankaj.gupta@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

hw/i386/sev: Add function to get SEV metadata from OVMF header

A recent version of OVMF expanded the reset vector GUID list to add
SEV-specific metadata GUID. The SEV metadata describes the reserved
memory regions such as the secrets and CPUID page used during the SEV-SNP
guest launch.

The pc_system_get_ovmf_sev_metadata_ptr() is used to retieve the SEV
metadata pointer from the OVMF GUID list.

Signed-off-by: Brijesh Singh <brijesh.singh@amd.com>
Signed-off-by: Michael Roth <michael.roth@amd.com>
Signed-off-by: Pankaj Gupta <pankaj.gupta@amd.com>
Message-ID: <20240530111643.1091816-19-pankaj.gupta@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

i386/sev: Set CPU state to protected once SNP guest payload is finalized

Once KVM_SNP_LAUNCH_FINISH is called the vCPU state is copied into the
vCPU's VMSA page and measured/encrypted. Any attempt to read/write CPU
state afterward will only be acting on the initial data and so are
effectively no-ops.

Set the vCPU state to protected at this point so that QEMU don't
continue trying to re-sync vCPU data during guest runtime.

Signed-off-by: Michael Roth <michael.roth@amd.com>
Signed-off-by: Pankaj Gupta <pankaj.gupta@amd.com>
Message-ID: <20240530111643.1091816-18-pankaj.gupta@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

i386/sev: Add handling to encrypt/finalize guest launch data

Process any queued up launch data and encrypt/measure it into the SNP
guest instance prior to initial guest launch.

This also updates the KVM_SEV_SNP_LAUNCH_UPDATE call to handle partial
update responses.

Signed-off-by: Brijesh Singh <brijesh.singh@amd.com>
Co-developed-by: Michael Roth <michael.roth@amd.com>
Signed-off-by: Michael Roth <michael.roth@amd.com>
Co-developed-by: Pankaj Gupta <pankaj.gupta@amd.com>
Signed-off-by: Pankaj Gupta <pankaj.gupta@amd.com>
Message-ID: <20240530111643.1091816-17-pankaj.gupta@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

i386/sev: Add the SNP launch start context

The SNP_LAUNCH_START is called first to create a cryptographic launch
context within the firmware.

Signed-off-by: Brijesh Singh <brijesh.singh@amd.com>
Signed-off-by: Michael Roth <michael.roth@amd.com>
Co-developed-by: Pankaj Gupta <pankaj.gupta@amd.com>
Signed-off-by: Pankaj Gupta <pankaj.gupta@amd.com>
Message-ID: <20240530111643.1091816-16-pankaj.gupta@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

i386/sev: Update query-sev QAPI format to handle SEV-SNP

Most of the current 'query-sev' command is relevant to both legacy
SEV/SEV-ES guests and SEV-SNP guests, with 2 exceptions:

  - 'policy' is a 64-bit field for SEV-SNP, not 32-bit, and
    the meaning of the bit positions has changed
  - 'handle' is not relevant to SEV-SNP

To address this, this patch adds a new 'sev-type' field that can be
used as a discriminator to select between SEV and SEV-SNP-specific
fields/formats without breaking compatibility for existing management
tools (so long as management tools that add support for launching
SEV-SNP guest update their handling of query-sev appropriately).

The corresponding HMP command has also been fixed up similarly.

Signed-off-by: Michael Roth <michael.roth@amd.com>
Co-developed-by:Pankaj Gupta <pankaj.gupta@amd.com>
Signed-off-by: Pankaj Gupta <pankaj.gupta@amd.com>
Message-ID: <20240530111643.1091816-15-pankaj.gupta@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

i386/sev: Add a class method to determine KVM VM type for SNP guests

SEV guests can use either KVM_X86_DEFAULT_VM, KVM_X86_SEV_VM,
or KVM_X86_SEV_ES_VM depending on the configuration and what
the host kernel supports. SNP guests on the other hand can only
ever use KVM_X86_SNP_VM, so split determination of VM type out
into a separate class method that can be set accordingly for
sev-guest vs. sev-snp-guest objects and add handling for SNP.

Signed-off-by: Pankaj Gupta <pankaj.gupta@amd.com>
Message-ID: <20240530111643.1091816-14-pankaj.gupta@amd.com>
[Remove unnecessary function pointer declaration. - Paolo]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

i386/sev: Don't return launch measurements for SEV-SNP guests

For SEV-SNP guests, launch measurement is queried from within the guest
during attestation, so don't attempt to return it as part of
query-sev-launch-measure.

Signed-off-by: Michael Roth <michael.roth@amd.com>
Signed-off-by: Pankaj Gupta <pankaj.gupta@amd.com>
Message-ID: <20240530111643.1091816-13-pankaj.gupta@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

i386/cpu: Set SEV-SNP CPUID bit when SNP enabled

SNP guests will rely on this bit to determine certain feature support.

Signed-off-by: Michael Roth <michael.roth@amd.com>
Signed-off-by: Pankaj Gupta <pankaj.gupta@amd.com>
Message-ID: <20240530111643.1091816-12-pankaj.gupta@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

i386/sev: Add snp_kvm_init() override for SNP class

SNP does not support SMM and requires guest_memfd for
private guest memory, so add SNP specific kvm_init()
functionality in snp_kvm_init() class method.

Signed-off-by: Michael Roth <michael.roth@amd.com>
Co-developed-by: Pankaj Gupta <pankaj.gupta@amd.com>
Signed-off-by: Pankaj Gupta <pankaj.gupta@amd.com>
Message-ID: <20240530111643.1091816-11-pankaj.gupta@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

i386/sev: Add sev_kvm_init() override for SEV class

Some aspects of the init routine SEV are specific to SEV and not
applicable for SNP guests, so move the SEV-specific bits into
separate class method and retain only the common functionality.

Co-developed-by: Michael Roth <michael.roth@amd.com>
Signed-off-by: Michael Roth <michael.roth@amd.com>
Signed-off-by: Pankaj Gupta <pankaj.gupta@amd.com>
Message-ID: <20240530111643.1091816-10-pankaj.gupta@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

i386/sev: Add a sev_snp_enabled() helper

Add a simple helper to check if the current guest type is SNP. Also have
SNP-enabled imply that SEV-ES is enabled as well, and fix up any places
where the sev_es_enabled() check is expecting a pure/non-SNP guest.

Signed-off-by: Michael Roth <michael.roth@amd.com>
Signed-off-by: Pankaj Gupta <pankaj.gupta@amd.com>
Message-ID: <20240530111643.1091816-9-pankaj.gupta@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

i386/sev: Introduce 'sev-snp-guest' object

SEV-SNP support relies on a different set of properties/state than the
existing 'sev-guest' object. This patch introduces the 'sev-snp-guest'
object, which can be used to configure an SEV-SNP guest. For example,
a default-configured SEV-SNP guest with no additional information
passed in for use with attestation:

  -object sev-snp-guest,id=sev0

or a fully-specified SEV-SNP guest where all spec-defined binary
blobs are passed in as base64-encoded strings:

  -object sev-snp-guest,id=sev0, \
    policy=0x30000, \
    init-flags=0, \
    id-block=YWFhYWFhYWFhYWFhYWFhCg==, \
    id-auth=CxHK/OKLkXGn/KpAC7Wl1FSiisWDbGTEKz..., \
    author-key-enabled=on, \
    host-data=LNkCWBRC5CcdGXirbNUV1OrsR28s..., \
    guest-visible-workarounds=AA==, \

See the QAPI schema updates included in this patch for more usage
details.

In some cases these blobs may be up to 4096 characters, but this is
generally well below the default limit for linux hosts where
command-line sizes are defined by the sysconf-configurable ARG_MAX
value, which defaults to 2097152 characters for Ubuntu hosts, for
example.

Signed-off-by: Brijesh Singh <brijesh.singh@amd.com>
Co-developed-by: Michael Roth <michael.roth@amd.com>
Acked-by: Markus Armbruster <armbru@redhat.com> (for QAPI schema)
Signed-off-by: Michael Roth <michael.roth@amd.com>
Co-developed-by: Pankaj Gupta <pankaj.gupta@amd.com>
Signed-off-by: Pankaj Gupta <pankaj.gupta@amd.com>
Message-ID: <20240530111643.1091816-8-pankaj.gupta@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

i386/sev: Move sev_launch_finish to separate class method

When sev-snp-guest objects are introduced there will be a number of
differences in how the launch finish is handled compared to the existing
sev-guest object. Move sev_launch_finish() to a class method to make it
easier to implement SNP-specific launch update functionality later.

Signed-off-by: Pankaj Gupta <pankaj.gupta@amd.com>
Message-ID: <20240530111643.1091816-7-pankaj.gupta@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

i386/sev: Move sev_launch_update to separate class method

When sev-snp-guest objects are introduced there will be a number of
differences in how the launch data is handled compared to the existing
sev-guest object. Move sev_launch_start() to a class method to make it
easier to implement SNP-specific launch update functionality later.

Signed-off-by: Pankaj Gupta <pankaj.gupta@amd.com>
Message-ID: <20240530111643.1091816-6-pankaj.gupta@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

i386/sev: Introduce "sev-common" type to encapsulate common SEV state

Currently all SEV/SEV-ES functionality is managed through a single
'sev-guest' QOM type. With upcoming support for SEV-SNP, taking this
same approach won't work well since some of the properties/state
managed by 'sev-guest' is not applicable to SEV-SNP, which will instead
rely on a new QOM type with its own set of properties/state.

To prepare for this, this patch moves common state into an abstract
'sev-common' parent type to encapsulate properties/state that are
common to both SEV/SEV-ES and SEV-SNP, leaving only SEV/SEV-ES-specific
properties/state in the current 'sev-guest' type. This should not
affect current behavior or command-line options.

As part of this patch, some related changes are also made:

  - a static 'sev_guest' variable is currently used to keep track of
    the 'sev-guest' instance. SEV-SNP would similarly introduce an
    'sev_snp_guest' static variable. But these instances are now
    available via qdev_get_machine()->cgs, so switch to using that
    instead and drop the static variable.

  - 'sev_guest' is currently used as the name for the static variable
    holding a pointer to the 'sev-guest' instance. Re-purpose the name
    as a local variable referring the 'sev-guest' instance, and use
    that consistently throughout the code so it can be easily
    distinguished from sev-common/sev-snp-guest instances.

  - 'sev' is generally used as the name for local variables holding a
    pointer to the 'sev-guest' instance. In cases where that now points
    to common state, use the name 'sev_common'; in cases where that now
    points to state specific to 'sev-guest' instance, use the name
    'sev_guest'

In order to enable kernel-hashes for SNP, pull it from
SevGuestProperties to its parent SevCommonProperties so
it will be available for both SEV and SNP.

Signed-off-by: Michael Roth <michael.roth@amd.com>
Co-developed-by: Dov Murik <dovmurik@linux.ibm.com>
Signed-off-by: Dov Murik <dovmurik@linux.ibm.com>
Acked-by: Markus Armbruster <armbru@redhat.com> (QAPI schema)
Co-developed-by: Pankaj Gupta <pankaj.gupta@amd.com>
Signed-off-by: Pankaj Gupta <pankaj.gupta@amd.com>
Message-ID: <20240530111643.1091816-5-pankaj.gupta@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

i386/sev: Replace error_report with error_setg

Signed-off-by: Pankaj Gupta <pankaj.gupta@amd.com>
Message-ID: <20240530111643.1091816-2-pankaj.gupta@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

machine: allow early use of machine_require_guest_memfd

Ask the ConfidentialGuestSupport object whether to use guest_memfd
for KVM-backend private memory.  This bool can be set in instance_init
(or user_complete) so that it is available when the machine is created.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

update-linux-headers: import linux/kvm_para.h header

Right now QEMU is importing arch/x86/include/uapi/asm/kvm_para.h
because it includes definitions for kvmclock and for KVM CPUID
bits.  However, other definitions for KVM hypercall values and return
codes are included in include/uapi/linux/kvm_para.h and they will be
used by SEV-SNP.

To ensure that it is possible to include both <linux/kvm_para.h> and
"standard-headers/asm-x86/kvm_para.h" without conflicts, provide
linux/kvm_para.h as a portable header too, and forward linux-headers/
files to those in include/standard-headers.  Note that <linux/kvm_para.h>
will include architecture-specific definitions as well, but
"standard-headers/linux/kvm_para.h" will not because it can be used in
architecture-independent files.

This could easily be extended to other architectures, but right now
they do not need any symbol in their specific kvm_para.h files.

Reviewed-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

linux-headers: Update to current kvm/next

This updates kernel headers to commit 6f627b425378 ("KVM: SVM: Add module
parameter to enable SEV-SNP", 2024-05-12).  The SNP host patches will
be included in Linux 6.11, to be released next July.

Also brings in an linux-headers/linux/vhost.h fix from v6.9-rc4.

Co-developed-by: Michael Roth <michael.roth@amd.com>
Signed-off-by: Michael Roth <michael.roth@amd.com>
Signed-off-by: Pankaj Gupta <pankaj.gupta@amd.com>
Message-ID: <20240530111643.1091816-3-pankaj.gupta@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

update-linux-headers: move pvpanic.h to correct directory

Linux has <misc/pvpanic.h>, not <linux/pvpanic.h>.  Use the same
directory for QEMU's include/standard-headers/ copy.

Reviewed-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

update-linux-headers: fix forwarding to asm-generic headers

Afer commit 3efc75ad9d9 ("scripts/update-linux-headers.sh: Remove
temporary directory inbetween", 2024-05-29), updating linux-headers/
results in errors such as

   cp: cannot stat '/tmp/tmp.1A1Eejh1UE/headers/include/asm/bitsperlong.h': No such file or directory

because Loongarch does not have an asm/bitsperlong.h file and uses the
generic version.  Before commit 3efc75ad9d9, the missing file would
incorrectly cause stale files to be included in linux-headers/.  The files
were never committed to qemu.git, but were wrong nevertheless. The build
would just use the system version of the files, which is opposite to
the idea of importing Linux header files into QEMU's tree.

Create forwarding headers, resembling the ones that are generated during a
kernel build by scripts/Makefile.asm-generic, if a file is only installed
under include/asm-generic/.

Reviewed-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: fix xsave.flat from kvm-unit-tests

xsave.flat checks that "executing the XSETBV instruction causes a general-
protection fault (#GP) if ECX = 0 and EAX[2:1] has the value 10b".  QEMU allows
that option, so the test fails.  Add the condition.

Cc: qemu-stable@nongnu.org
Fixes: 892544317fe ("target/i386: implement XSAVE and XRSTOR of AVX registers", 2022-10-18)
Reported-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386/tcg: Fix RDPID feature check

DisasContext.cpuid_ext_features indicates CPUID.01H.ECX.

Use DisasContext.cpuid_7_0_ecx_features field to check RDPID feature bit
(CPUID_7_0_ECX_RDPID).

Fixes: 6750485bf42a ("target/i386: implement RDPID in TCG")
Inspired-by: Xinyu Li <lixinyu20s@ict.ac.cn>
Signed-off-by: Zhao Liu <zhao1.liu@intel.com>
Message-ID: <20240603080723.1256662-1-zhao1.liu@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: fix memory opsize for Mov to/from Seg

This commit fixes an issue with MOV instructions (0x8C and 0x8E)
involving segment registers; MOV to segment register's source is
16-bit, while MOV from segment register has to explicitly set the
memory operand size to 16 bits.  Introduce a new flag
X86_SPECIAL_Op0_Mw to handle this specification correctly.

Signed-off-by: Xinyu Li <lixinyu20s@ict.ac.cn>
Message-ID: <20240602100528.2135717-1-lixinyu20s@ict.ac.cn>
Fixes: 5e9e21bcc4d ("target/i386: move 60-BF opcodes to new decoder", 2024-05-07)
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: fix SSE and SSE2 feature check

Features check of CPUID_SSE and CPUID_SSE2 should use cpuid_features,
rather than cpuid_ext_features.

Signed-off-by: Xinyu Li <lixinyu20s@ict.ac.cn>
Reviewed-by: Zhao Liu <zhao1.liu@intel.com>
Message-ID: <20240602100904.2137939-1-lixinyu20s@ict.ac.cn>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

host/i386: assume presence of POPCNT

QEMU now requires an x86-64-v2 host, which has the POPCNT instruction.
Use it freely in TCG-generated code.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

host/i386: assume presence of SSSE3

QEMU now requires an x86-64-v2 host, which has SSSE3 instructions
(notably, PSHUFB which is used by QEMU's AES implementation).
Do not bother checking it.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

host/i386: assume presence of SSE2

QEMU now requires an x86-64-v2 host, which has SSE2.
Use it freely in buffer_is_zero.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

host/i386: assume presence of CMOV

QEMU now requires an x86-64-v2 host, which always has CMOV.
Use it freely in TCG generated code.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

meson: assume x86-64-v2 baseline ISA

x86-64-v2 processors were released in 2008, assume that we have one.
Unfortunately there is no GCC flag to enable all the features
without disabling what came after; so enable them one by one.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

host/i386: nothing looks at CPUINFO_SSE4

The only user was the SSE4.1 variant of buffer_is_zero, which has
been removed; code to compute CPUINFO_SSE4 is dead.

Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

virtio-blk: remove SCSI passthrough functionality

The legacy SCSI passthrough functionality has never been enabled for
VIRTIO 1.0 and was deprecated more than four years ago.

Get rid of it---almost, because QEMU is advertising it unconditionally
for legacy virtio-blk devices.  Just parse the header and return a
nonzero status.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

docs, tests: do not specify scsi=off

This has been the default forever.

Acked-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

Merge tag 'hw-misc-accel-20240604' of https://github.com/philmd/qemu into staging

Misc HW & accelerators patch queue

- Use async exit in debugexit model (Thomas)
- Fixed bug reading xlnx_dpdma descriptor (Peter)
- Initialise plugin state before vCPU/thread creation (Alex)
- Few sprintf() calls removed (Richard & Philippe)
- Few deprecated QMP events removed (Philippe)
- Housekeeping in Xen (Edgar & Philippe)
- Split USB HID/HUB & update MAINTAINERS (Gerd)

# -----BEGIN PGP SIGNATURE-----
#
# iQIzBAABCAAdFiEE+qvnXhKRciHc/Wuy4+MsLN6twN4FAmZe5FMACgkQ4+MsLN6t
# wN5QghAAl1j5++wKcJaWoribnhxYctcBPhg3eEAwg2qVRbx5tz7HwYw8jYp+qB47
# abAtE3WOUg+Z+NaBi0bTj8Ns0wrBTZy0toJrQb+k5M9CzJJIFSURcFV2J40ORld8
# VHNKmzNN+ZscxAW3fSFruJlGLaskwB0bk9VF0cYSip2Y+xjbPlJZ8dlffX/HAvwB
# Ul4gJW04FT3k6KoeJASNiAMm7mPW9iP4kY7rKz+j+NWydAfl8yPzxFwW6J6U2npi
# Hvmy1519W74cftZSqxkBETtbw1YVKyT6QEA9HjJOjc05E6UYnxoR13taEzLNaDHA
# B7Xhp6IUYZGRD1bt85WZCP5t5yUINLgmzJU2OQnv9dfU1HdSkk0FQiBr0Iif3tCI
# /2cQumkGI4etXX77oVIwkioo0ZTvEggDNsN1eKHIY7yIE5etqXduPwnTfwHFI3lv
# Nj18MGF0cftbYDEbUwNUnN5U6+msW3Si6hQaUxpUWXy7Idkl9pBiWjv1OpBg9PqX
# YIuw9/USs7i8yZkh0/q8UcZxdzB3LF+LxjVVVoTVD6t+KbRSkJSOJm7hfqCKuXbC
# lEmtLt+aH/pdZ2Banug1Ayen2L7s0czRHQWC++DfNQve1xEUkSFp5VE0mxsr+6Rp
# 18/dSGo++Nh52jUt2QFEv9LyAAQCKxg72I9DW9sB7uC004+ub84=
# =xcID
# -----END PGP SIGNATURE-----
# gpg: Signature made Tue 04 Jun 2024 04:54:27 AM CDT
# gpg:                using RSA key FAABE75E12917221DCFD6BB2E3E32C2CDEADC0DE
# gpg: Good signature from "Philippe Mathieu-Daudé (F4BUG) <f4bug@amsat.org>" [full]

* tag 'hw-misc-accel-20240604' of https://github.com/philmd/qemu: (32 commits)
  usb: add config options for the hub and hid devices
  qga: Remove deprecated 'blacklist' argument / config key
  trace: Remove deprecated 'vcpu' field from QMP trace events
  hw/acpi: Remove the deprecated QAPI MEM_UNPLUG_ERROR event
  hw/dma/xlnx_dpdma: Read descriptor into buffer, not into pointer-to-buffer
  hw/misc/debugexit: use runstate API instead of plain exit()
  hw/xen: Register framebuffer backend via xen_backend_init()
  hw/xen: Make XenDevOps structures const
  hw/xen: Constify xenstore_be::XenDevOps
  hw/xen: Constify XenLegacyDevice::XenDevOps
  physmem: Replace check for RAMBlock offset 0 with xen_mr_is_memory
  physmem: Always pass offset + addr to xen_map_cache
  xen: Add xen_mr_is_memory()
  core/cpu-common: initialise plugin state before thread creation
  plugins: remove special casing for cpu->realized
  cpu-target: don't set cpu->thread_id to bogus value
  cpu: move Qemu[Thread|Cond] setup into common code
  hw/core: expand on the alignment of CPUState
  accel/kvm: Fix two lines with hard-coded tabs
  accel/tcg: Move common declarations to 'internal-common.h'
  ...

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

Merge tag 'net-pull-request' of https://github.com/jasowang/qemu into staging

# -----BEGIN PGP SIGNATURE-----
#
# iQEzBAABCAAdFiEEIV1G9IJGaJ7HfzVi7wSWWzmNYhEFAmZewo4ACgkQ7wSWWzmN
# YhHhxgf/ZaECxru4fP8wi34XdSG/PR+BF+W5M9gZIRGrHg3vIf3/LRTpZTDccbRN
# Qpwtypr9O6/AWG9Os80rn7alsmMDxN8PDDNLa9T3wf5pJUQSyQ87Yy0MiuTNPSKD
# HKYUIfIlbFCM5WUW4huMmg98gKTgnzZMqOoRyMFZitbkR59qCm+Exws4HtXvCH68
# 3k4lgvnFccmzO9iIzaOUIPs+Yf04Kw/FrY0Q/6nypvqbF2W80Md6w02JMQuTLwdF
# Guxeg/n6g0NLvCBbkjiM2VWfTaWJYbwFSwRTAMxM/geqh7qAgGsmD0N5lPlgqRDy
# uAy2GvFyrwzcD0lYqf0/fRK0Go0HPA==
# =J70K
# -----END PGP SIGNATURE-----
# gpg: Signature made Tue 04 Jun 2024 02:30:22 AM CDT
# gpg:                using RSA key 215D46F48246689EC77F3562EF04965B398D6211
# gpg: Good signature from "Jason Wang (Jason Wang on RedHat) <jasowang@redhat.com>" [undefined]
# gpg: WARNING: This key is not certified with a trusted signature!
# gpg:          There is no indication that the signature belongs to the owner.
# Primary key fingerprint: 215D 46F4 8246 689E C77F  3562 EF04 965B 398D 6211

* tag 'net-pull-request' of https://github.com/jasowang/qemu:
  ebpf: Added traces back. Changed source set for eBPF to 'system'.
  virtio-net: drop too short packets early
  ebpf: Add a separate target for skeleton
  ebpf: Refactor tun_rss_steering_prog()
  ebpf: Return 0 when configuration fails
  ebpf: Fix RSS error handling
  virtio-net: Do not write hashes to peer buffer
  virtio-net: Always set populate_hash
  virtio-net: Unify the logic to update NIC state for RSS
  virtio-net: Disable RSS on reset
  virtio-net: Shrink header byte swapping buffer
  virtio-net: Copy header only when necessary
  virtio-net: Add only one queue pair when realizing
  virtio-net: Do not propagate ebpf-rss-fds errors
  tap: Shrink zeroed virtio-net header
  tap: Call tap_receive_iov() from tap_receive()
  net: Remove receive_raw()
  net: Move virtio-net header length assertion
  tap: Remove qemu_using_vnet_hdr()
  tap: Remove tap_probe_vnet_hdr_len()

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

Merge tag 'pull-nbd-2024-05-30-v2' of https://repo.or.cz/qemu/ericb into staging

NBD patches for 2024-05-30

- Fix AioContext assertion with NBD+TLS

# -----BEGIN PGP SIGNATURE-----
#
# iQEzBAABCAAdFiEEccLMIrHEYCkn0vOqp6FrSiUnQ2oFAmZd0MYACgkQp6FrSiUn
# Q2rrxwf8C+nhCy5d/d0DfBvDANEWB5zFGffC89t9P2ULJ6ehGlZuAyF8bCqnptPM
# 76qhZfCz1XCUM0B+H+JXzcOsphTY5Y+hQ6Aq95VbuJgzq4CbAz4D30RNS0P5Q/ei
# v1+goMTEegH1sSS6b0NueTj4chesMui3bJpiUnb3qPptZ+L0OWhIsk6Pe0N418UM
# kmWC3kPWUPkmpxqXprcV/OunS6ipkfSVORfjxFNbFQcceuP45lt+f0YxhMHU8NIU
# PwFiUaPPCy/mDFTHgqCmpqoivFroUyhQIFl5ucHmds2wTOTx9VpxunYNcrVCq6vY
# UUfKc/p5lRArTqJXEth8cHlo1sFOEw==
# =7ZGy
# -----END PGP SIGNATURE-----
# gpg: Signature made Mon 03 Jun 2024 09:18:46 AM CDT
# gpg:                using RSA key 71C2CC22B1C4602927D2F3AAA7A16B4A2527436A
# gpg: Good signature from "Eric Blake <eblake@redhat.com>" [full]
# gpg:                 aka "Eric Blake (Free Software Programmer) <ebb9@byu.net>" [full]
# gpg:                 aka "[jpeg image of size 6874]" [full]

* tag 'pull-nbd-2024-05-30-v2' of https://repo.or.cz/qemu/ericb:
  iotests: test NBD+TLS+iothread
  qio: Inherit follow_coroutine_ctx across TLS

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

usb: add config options for the hub and hid devices

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-ID: <20240530112718.1752905-3-kraxel@redhat.com>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>

qga: Remove deprecated 'blacklist' argument / config key

The 'blacklist' argument / config key are deprecated since commit
582a098e6c ("qga: Replace 'blacklist' command line and config file
options by 'block-rpcs'"), time to remove them.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Konstantin Kostiuk <kkostiuk@redhat.com>
Message-Id: <20240530070413.19181-1-philmd@linaro.org>

trace: Remove deprecated 'vcpu' field from QMP trace events

'vcpu' fields are deprecated since commit 5485e52a33
("qapi: make the vcpu parameters deprecated for 8.1"),
time to remove them.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20240530071548.20074-3-philmd@linaro.org>

hw/acpi: Remove the deprecated QAPI MEM_UNPLUG_ERROR event

The MEM_UNPLUG_ERROR event is deprecated since commit d43f1670c7
("qapi/qdev.json: add DEVICE_UNPLUG_GUEST_ERROR QAPI event"),
time to remove it.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Harsh Prateek Bora <harshpb@linux.ibm.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20240530071548.20074-2-philmd@linaro.org>

hw/dma/xlnx_dpdma: Read descriptor into buffer, not into pointer-to-buffer

In fdf029762f501 we factored out the handling of reading and writing
DMA descriptors from guest memory.  Unfortunately we accidentally
made the descriptor-read read the descriptor into the address of the
buffer rather than into the buffer, because we didn't notice we
needed to update the arguments to the dma_memory_read() call. Before
the refactoring, "&desc" is the address of a local struct DPDMADescriptor
variable in xlnx_dpdma_start_operation(), which is the correct target
for the guest-memory-read. But after the refactoring 'desc' is the
"DPDMADescriptor *desc" argument to the new function, and so it is
already an address.

This bug is an overrun of a stack variable, since a pointer is at
most 8 bytes long and we try to read 64 bytes, as well as being
incorrect behaviour.

Pass 'desc' rather than '&desc' as the dma_memory_read() argument
to fix this.

(The same bug is not present in xlnx_dpdma_write_descriptor(),
because there we are writing the descriptor from a local struct
variable "DPDMADescriptor tmp_desc" and so passing &tmp_desc to
dma_memory_write() is correct.)

Spotted by Coverity: CID 1546649

Fixes: fdf029762f50101 ("xlnx_dpdma: fix descriptor endianness bug")
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Edgar E. Iglesias <edgar.iglesias@amd.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-ID: <20240531124628.476938-1-peter.maydell@linaro.org>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>

hw/misc/debugexit: use runstate API instead of plain exit()

Directly calling exit() prevents any kind of management or handling.
Instead use the corresponding runstate API.
The default behavior of the runstate API is the same as exit().

Signed-off-by: Thomas Weißschuh <thomas@t-8ch.de>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-ID: <20240523-debugexit-v1-1-d52fcaf7bf8b@t-8ch.de>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>

hw/xen: Register framebuffer backend via xen_backend_init()

Align the framebuffer backend with the other legacy ones,
register it via xen_backend_init() when '-vga xenfb' is
used. It is safe because MODULE_INIT_XEN_BACKEND is called
in xen_bus_realize(), long after CLI processing initialized
the vga_interface_type variable.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Paul Durrant <paul@xen.org>
Message-Id: <20240510104908.76908-8-philmd@linaro.org>

hw/xen: Make XenDevOps structures const

Keep XenDevOps structures in .rodata.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Paul Durrant <paul@xen.org>
Message-Id: <20240510104908.76908-5-philmd@linaro.org>

hw/xen: Constify xenstore_be::XenDevOps

XenDevOps @ops is not updated, mark it const.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Paul Durrant <paul@xen.org>
Message-Id: <20240510104908.76908-4-philmd@linaro.org>

hw/xen: Constify XenLegacyDevice::XenDevOps

XenDevOps @ops is not updated, mark it const.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Paul Durrant <paul@xen.org>
Message-Id: <20240510104908.76908-3-philmd@linaro.org>

physmem: Replace check for RAMBlock offset 0 with xen_mr_is_memory

For xen, when checking for the first RAM (xen_memory), use
xen_mr_is_memory() rather than checking for a RAMBlock with
offset 0.

All Xen machines create xen_memory first so this has no
functional change for existing machines.

Signed-off-by: Edgar E. Iglesias <edgar.iglesias@amd.com>
Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>
Reviewed-by: David Hildenbrand <david@redhat.com>
Message-ID: <20240529140739.1387692-6-edgar.iglesias@gmail.com>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>

physmem: Always pass offset + addr to xen_map_cache

Always pass address with offset to xen_map_cache().
This is in preparation for support for grant mappings.

Since this is within a block that checks for offset == 0,
this has no functional changes.

Signed-off-by: Edgar E. Iglesias <edgar.iglesias@amd.com>
Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>
Reviewed-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-ID: <20240529140739.1387692-5-edgar.iglesias@gmail.com>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>

xen: Add xen_mr_is_memory()

Add xen_mr_is_memory() to abstract away tests for the
xen_memory MR.

No functional changes.

Signed-off-by: Edgar E. Iglesias <edgar.iglesias@amd.com>
Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>
Acked-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-ID: <20240529140739.1387692-4-edgar.iglesias@gmail.com>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>

core/cpu-common: initialise plugin state before thread creation

Originally I tried to move where vCPU thread initialisation to later
in realize. However pulling that thread (sic) got gnarly really
quickly. It turns out some steps of CPU realization need values that
can only be determined from the running vCPU thread.

However having moved enough out of the thread creation we can now
queue work before the thread starts (at least for TCG guests) and
avoid the race between vcpu_init and other vcpu states a plugin might
subscribe to.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Message-ID: <20240530194250.1801701-6-alex.bennee@linaro.org>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>

plugins: remove special casing for cpu->realized

Now the condition variable is initialised early on we don't need to go
through hoops to avoid calling async_run_on_cpu.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Message-ID: <20240530194250.1801701-5-alex.bennee@linaro.org>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>

cpu-target: don't set cpu->thread_id to bogus value

The thread_id isn't valid until the threads are created. There is no
point setting it here. The only thing that cares about the thread_id
is qmp_query_cpus_fast.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Message-ID: <20240530194250.1801701-4-alex.bennee@linaro.org>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>

cpu: move Qemu[Thread|Cond] setup into common code

Aside from the round robin threads this is all common code. By
moving the halt_cond setup we also no longer need hacks to work around
the race between QOM object creation and thread creation.

It is a little ugly to free stuff up for the round robin thread but
better it deal with its own specialises than making the other
accelerators jump through hoops.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Message-ID: <20240530194250.1801701-3-alex.bennee@linaro.org>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>

hw/core: expand on the alignment of CPUState

Make the relationship between CPUState, ArchCPU and cpu_env a bit
clearer in the kdoc comments.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Message-ID: <20240530194250.1801701-2-alex.bennee@linaro.org>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>

accel/kvm: Fix two lines with hard-coded tabs

In kvm-all.c, two lines have been accidentally indented with
hard-coded tabs rather than spaces. Normalise to match the rest
of the file.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Zhao Liu <zhao1.liu@intel.com>
Message-ID: <20240531170952.505323-1-peter.maydell@linaro.org>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>

accel/tcg: Move common declarations to 'internal-common.h'

'internal-target.h' is meant for target-specific declarations,
while 'internal-common.h' for common ones. Move common declarations
to it.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Message-Id: <20240528145953.65398-3-philmd@linaro.org>

system/runstate: Remove unused 'qemu/plugin.h' header

system/runstate.c never required "qemu/plugin.h".

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Message-Id: <20240528145953.65398-2-philmd@linaro.org>

MAINTAINERS: drop usb maintainership

Remove myself from usb entries.
Flip status to "Orphan" for entries which have nobody else listed.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Manos Pitsidianakis <manos.pitsidianakis@linaro.org>
Message-ID: <20240528083858.836262-3-kraxel@redhat.com>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>

util/hexdump: Remove ascii parameter from qemu_hexdump_line

Split out asciidump_line as a separate function, local to hexdump.c,
for use by qemu_hexdump.  Use "%-*s" to generate the alignment
between the hex and the ascii, rather than explicit spaces.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-ID: <20240412073346.458116-3-richard.henderson@linaro.org>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>

util/hexdump: Remove b parameter from qemu_hexdump_line

Require that the caller output the offset and increment bufptr.
Use QEMU_HEXDUMP_LINE_BYTES in vhost_vdpa_dump_config instead
of raw integer.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-ID: <20240412073346.458116-2-richard.henderson@linaro.org>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>

disas/microblaze: Replace sprintf() by snprintf()

sprintf() is deprecated on Darwin since macOS 13.0 / XCode 14.1,
resulting in painful developper experience. Use snprintf() instead.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Edgar E. Iglesias <edgar.iglesias@amd.com>
Message-Id: <20240411104340.6617-3-philmd@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

disas/m68k: Replace sprintf() by snprintf()

sprintf() is deprecated on Darwin since macOS 13.0 / XCode 14.1,
resulting in painful developper experience. Use snprintf() instead.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Message-Id: <20240411104340.6617-2-philmd@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

target/i386/kvm: Improve KVM_EXIT_NOTIFY warnings

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Zhao Liu <zhao1.liu@intel.com>
Message-ID: <20240412073346.458116-28-richard.henderson@linaro.org>
[PMD: Fixed typo reported by Peter Maydell]
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>

target/arm: Replace sprintf() by snprintf()

sprintf() is deprecated on Darwin since macOS 13.0 / XCode 14.1,
resulting in painful developper experience. Use snprintf() instead.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Message-Id: <20240411104340.6617-9-philmd@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

target/mips: Remove unused 'hw/misc/mips_itu.h' header

Since commit e1152f8166 ("target/mips: Remove helpers accessing
SAAR registers") this header is not needed.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Jiaxun Yang <jiaxun.yang@flygoat.com>
Message-Id: <20240529155216.5574-1-philmd@linaro.org>

target/riscv: Restrict riscv_cpu_do_interrupt() to sysemu

riscv_cpu_do_interrupt() is not reachable on user emulation.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-Id: <20230626232007.8933-7-philmd@linaro.org>

target/riscv: Restrict 'rv128' machine to TCG accelerator

We only build for 32/64-bit hosts, so TCG is required for
128-bit targets.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Acked-by: Alistair Francis <alistair.francis@wdc.com>
Message-Id: <20230626232007.8933-5-philmd@linaro.org>

target/riscv: Remove unused 'instmap.h' header in translate.c

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Acked-by: Alistair Francis <alistair.francis@wdc.com>
Message-Id: <20230626232007.8933-2-philmd@linaro.org>

ebpf: Added traces back. Changed source set for eBPF to 'system'.

There was an issue with Qemu build with "--disable-system".
The traces could be generated and the build fails.
The traces were 'cut out' for previous patches, and overall,
the 'system' source set should be used like in pre-'eBPF blob' patches.

Signed-off-by: Andrew Melnychenko <andrew@daynix.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>

virtio-net: drop too short packets early

Reproducer from https://gitlab.com/qemu-project/qemu/-/issues/1451
creates small packet (1 segment, len = 10 == n->guest_hdr_len),
then destroys queue.

"if (n->host_hdr_len != n->guest_hdr_len)" is triggered, if body creates
zero length/zero segment packet as there is nothing after guest header.

qemu_sendv_packet_async() tries to send it.

slirp discards it because it is smaller than Ethernet header,
but returns 0 because tx hooks are supposed to return total length of data.

0 is propagated upwards and is interpreted as "packet has been sent"
which is terrible because queue is being destroyed, nobody is waiting for TX
to complete and assert it triggered.

Fix is discard such empty packets instead of sending them.

Length 1 packets will go via different codepath:

virtqueue_push(q->tx_vq, elem, 0);
virtio_notify(vdev, q->tx_vq);
g_free(elem);

and aren't problematic.

Signed-off-by: Alexey Dobriyan <adobriyan@yandex-team.ru>
Signed-off-by: Jason Wang <jasowang@redhat.com>

ebpf: Add a separate target for skeleton

This generalizes the rule to generate the skeleton and allows to add
another.

Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>

ebpf: Refactor tun_rss_steering_prog()

This saves branches and makes later BPF program changes easier.

Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>

ebpf: Return 0 when configuration fails

The kernel interprets the returned value as an unsigned 32-bit so -1
will mean queue 4294967295, which is awkward. Return 0 instead.

Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>

ebpf: Fix RSS error handling

calculate_rss_hash() was using hash value 0 to tell if it calculated
a hash, but the hash value may be 0 on a rare occasion. Have a
distinct bool value for correctness.

Fixes: f3fa412de2 ("ebpf: Added eBPF RSS program.")
Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>

virtio-net: Do not write hashes to peer buffer

The peer buffer is qualified with const and not meant to be modified.

Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>

virtio-net: Always set populate_hash

The member is not cleared during reset so may have a stale value.

Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>