Merge tag 'migration-20240423-pull-request' of https://gitlab.com/peterx/qemu into staging

Migration pull for 9.1

- Het's new test cases for "channels"
- Het's fix for a typo for vsock parsing
- Cedric's VFIO error report series
- Cedric's one more patch for dirty-bitmap error reports
- Zhijian's rdma deprecation patch
- Yuan's zeropage optimization to fix double faults on anon mem
- Zhijian's COLO fix on a crash

# -----BEGIN PGP SIGNATURE-----
#
# iIgEABYKADAWIQS5GE3CDMRX2s990ak7X8zN86vXBgUCZig4HxIccGV0ZXJ4QHJl
# ZGhhdC5jb20ACgkQO1/MzfOr1wbQiwD/V5nSJzSuAG4Ra1Fjo+LRG2TT6qk8eNCi
# fIytehSw6cYA/0wqarxOF0tr7ikeyhtG3w4xFf44kk6KcPkoVSl1tqoL
# =pJmQ
# -----END PGP SIGNATURE-----
# gpg: Signature made Tue 23 Apr 2024 03:37:19 PM PDT
# gpg:                using EDDSA key B9184DC20CC457DACF7DD1A93B5FCCCDF3ABD706
# gpg:                issuer "peterx@redhat.com"
# gpg: Good signature from "Peter Xu <xzpeter@gmail.com>" [unknown]
# gpg:                 aka "Peter Xu <peterx@redhat.com>" [unknown]
# 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: B918 4DC2 0CC4 57DA CF7D  D1A9 3B5F CCCD F3AB D706

* tag 'migration-20240423-pull-request' of https://gitlab.com/peterx/qemu: (26 commits)
  migration/colo: Fix bdrv_graph_rdlock_main_loop: Assertion `!qemu_in_coroutine()' failed.
  migration/multifd: solve zero page causing multiple page faults
  migration: Add Error** argument to add_bitmaps_to_list()
  migration: Modify ram_init_bitmaps() to report dirty tracking errors
  migration: Add Error** argument to xbzrle_init()
  migration: Add Error** argument to ram_state_init()
  memory: Add Error** argument to the global_dirty_log routines
  migration: Introduce ram_bitmaps_destroy()
  memory: Add Error** argument to .log_global_start() handler
  migration: Add Error** argument to .load_setup() handler
  migration: Add Error** argument to .save_setup() handler
  migration: Add Error** argument to qemu_savevm_state_setup()
  migration: Add Error** argument to vmstate_save()
  migration: Always report an error in ram_save_setup()
  migration: Always report an error in block_save_setup()
  vfio: Always report an error in vfio_save_setup()
  s390/stattrib: Add Error** argument to set_migrationmode() handler
  tests/qtest/migration: Fix typo for vsock in SocketAddress_to_str
  tests/qtest/migration: Add negative tests to validate migration QAPIs
  tests/qtest/migration: Add multifd_tcp_plain test using list of channels instead of uri
  ...

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

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

* cleanups for stubs
* do not link pixman automatically into all targets
* optimize computation of VGA dirty memory region
* kvm: use configs/ definition to conditionalize debug support
* hw: Add compat machines for 9.1
* target/i386: add guest-phys-bits cpu property
* target/i386: Introduce Icelake-Server-v7 and SierraForest models
* target/i386: Export RFDS bit to guests
* q35: SMM ranges cleanups
* target/i386: basic support for confidential guests
* linux-headers: update headers
* target/i386: SEV: use KVM_SEV_INIT2 if possible
* kvm: Introduce support for memory_attributes
* RAMBlock: Add support of KVM private guest memfd
* Consolidate use of warn_report_once()
* pythondeps.toml: warn about updates needed to docs/requirements.txt
* target/i386: always write 32-bits for SGDT and SIDT

# -----BEGIN PGP SIGNATURE-----
#
# iQFIBAABCAAyFiEE8TM4V0tmI4mGbHaCv/vSX3jHroMFAmYn1UkUHHBib256aW5p
# QHJlZGhhdC5jb20ACgkQv/vSX3jHroO1nwgAhRQhkYcdtFc649WJWTNvJCNzmek0
# Sg7trH2NKlwA75zG8Qv4TR3E71UrXoY9oItwYstc4Erz+tdf73WyaHMF3cEk1p82
# xx3LcBYhP7jGSjabxTkZsFU8+MM1raOjRN/tHvfcjYLaJOqJZplnkaVhMbNPsVuM
# IPJ5bVQohxpmHKPbeFNpF4QJ9wGyZAYOfJOFCk09xQtHnA8CtFjS9to33QPAR/Se
# OVZwRCigVjf0KNmCnHC8tJHoW8pG/cdQAr3qqd397XbM1vVELv9fiXiMoGF78UsY
# trO4K2yg6N5Sly4Qv/++zZ0OZNkL3BREGp3wf4eTSvLXxqSGvfi8iLpFGA==
# =lwSL
# -----END PGP SIGNATURE-----
# gpg: Signature made Tue 23 Apr 2024 08:35:37 AM PDT
# gpg:                using RSA key F13338574B662389866C7682BFFBD25F78C7AE83
# gpg:                issuer "pbonzini@redhat.com"
# gpg: Good signature from "Paolo Bonzini <bonzini@gnu.org>" [undefined]
# gpg:                 aka "Paolo Bonzini <pbonzini@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: 46F5 9FBD 57D6 12E7 BFD4  E2F7 7E15 100C CD36 69B1
#      Subkey fingerprint: F133 3857 4B66 2389 866C  7682 BFFB D25F 78C7 AE83

* tag 'for-upstream' of https://gitlab.com/bonzini/qemu: (63 commits)
  target/i386/translate.c: always write 32-bits for SGDT and SIDT
  pythondeps.toml: warn about updates needed to docs/requirements.txt
  accel/tcg/icount-common: Consolidate the use of warn_report_once()
  target/i386/cpu: Merge the warning and error messages for AMD HT check
  target/i386/cpu: Consolidate the use of warn_report_once()
  target/i386/host-cpu: Consolidate the use of warn_report_once()
  kvm/tdx: Ignore memory conversion to shared of unassigned region
  kvm/tdx: Don't complain when converting vMMIO region to shared
  kvm: handle KVM_EXIT_MEMORY_FAULT
  physmem: Introduce ram_block_discard_guest_memfd_range()
  RAMBlock: make guest_memfd require uncoordinated discard
  HostMem: Add mechanism to opt in kvm guest memfd via MachineState
  kvm/memory: Make memory type private by default if it has guest memfd backend
  kvm: Enable KVM_SET_USER_MEMORY_REGION2 for memslot
  RAMBlock: Add support of KVM private guest memfd
  kvm: Introduce support for memory_attributes
  trace/kvm: Split address space and slot id in trace_kvm_set_user_memory()
  hw/i386/sev: Use legacy SEV VM types for older machine types
  i386/sev: Add 'legacy-vm-type' parameter for SEV guest objects
  target/i386: SEV: use KVM_SEV_INIT2 if possible
  ...

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

Open 9.1 development tree

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

migration/colo: Fix bdrv_graph_rdlock_main_loop: Assertion `!qemu_in_coroutine()' failed.

bdrv_activate_all() should not be called from the coroutine context, move
it to the QEMU thread colo_process_incoming_thread() with the bql_lock
protected.

The backtrace is as follows:
 #4  0x0000561af7948362 in bdrv_graph_rdlock_main_loop () at ../block/graph-lock.c:260
 #5  0x0000561af7907a68 in graph_lockable_auto_lock_mainloop (x=0x7fd29810be7b) at /patch/to/qemu/include/block/graph-lock.h:259
 #6  0x0000561af79167d1 in bdrv_activate_all (errp=0x7fd29810bed0) at ../block.c:6906
 #7  0x0000561af762b4af in colo_incoming_co () at ../migration/colo.c:935
 #8  0x0000561af7607e57 in process_incoming_migration_co (opaque=0x0) at ../migration/migration.c:793
 #9  0x0000561af7adbeeb in coroutine_trampoline (i0=-106876144, i1=22042) at ../util/coroutine-ucontext.c:175
 #10 0x00007fd2a5cf21c0 in  () at /lib64/libc.so.6

Cc: qemu-stable@nongnu.org
Cc: Fabiano Rosas <farosas@suse.de>
Closes: https://gitlab.com/qemu-project/qemu/-/issues/2277
Fixes: 2b3912f135 ("block: Mark bdrv_first_blk() and bdrv_is_root_node() GRAPH_RDLOCK")
Signed-off-by: Li Zhijian <lizhijian@fujitsu.com>
Reviewed-by: Zhang Chen <chen.zhang@intel.com>
Tested-by: Zhang Chen <chen.zhang@intel.com>
Reviewed-by: Fabiano Rosas <farosas@suse.de>
Link: https://lore.kernel.org/r/20240417025634.1014582-1-lizhijian@fujitsu.com
Signed-off-by: Peter Xu <peterx@redhat.com>

migration/multifd: solve zero page causing multiple page faults

Implemented recvbitmap tracking of received pages in multifd.

If the zero page appears for the first time in the recvbitmap, this
page is not checked and set.

If the zero page has already appeared in the recvbitmap, there is no
need to check the data but directly set the data to 0, because it is
unlikely that the zero page will be migrated multiple times.

Signed-off-by: Yuan Liu <yuan1.liu@intel.com>
Reviewed-by: Fabiano Rosas <farosas@suse.de>
Link: https://lore.kernel.org/r/20240401154110.2028453-2-yuan1.liu@intel.com
[peterx: touch up the comment, as the bitmap is used outside postcopy now]
Signed-off-by: Peter Xu <peterx@redhat.com>

migration: Add Error** argument to add_bitmaps_to_list()

This allows to report more precise errors in the migration handler
dirty_bitmap_save_setup().

Suggested-by: Vladimir Sementsov-Ogievskiy <vsementsov@yandex-team.ru>
Signed-off-by: Cédric Le Goater <clg@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@yandex-team.ru>
Link: https://lore.kernel.org/r/20240329105627.311227-1-clg@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>

migration: Modify ram_init_bitmaps() to report dirty tracking errors

The .save_setup() handler has now an Error** argument that we can use
to propagate errors reported by the .log_global_start() handler. Do
that for the RAM. The caller qemu_savevm_state_setup() will store the
error under the migration stream for later detection in the migration
sequence.

Signed-off-by: Cédric Le Goater <clg@redhat.com>
Reviewed-by: Fabiano Rosas <farosas@suse.de>
Link: https://lore.kernel.org/r/20240320064911.545001-15-clg@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>

migration: Add Error** argument to xbzrle_init()

Since the return value (-ENOMEM) is not exploited, follow the
recommendations of qapi/error.h and change it to a bool

Signed-off-by: Cédric Le Goater <clg@redhat.com>
Reviewed-by: Fabiano Rosas <farosas@suse.de>
Link: https://lore.kernel.org/r/20240320064911.545001-14-clg@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>

migration: Add Error** argument to ram_state_init()

Since the return value not exploited, follow the recommendations of
qapi/error.h and change it to a bool

Signed-off-by: Cédric Le Goater <clg@redhat.com>
Reviewed-by: Fabiano Rosas <farosas@suse.de>
Link: https://lore.kernel.org/r/20240320064911.545001-13-clg@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>

memory: Add Error** argument to the global_dirty_log routines

Now that the log_global*() handlers take an Error** parameter and
return a bool, do the same for memory_global_dirty_log_start() and
memory_global_dirty_log_stop(). The error is reported in the callers
for now and it will be propagated in the call stack in the next
changes.

To be noted a functional change in ram_init_bitmaps(), if the dirty
pages logger fails to start, there is no need to synchronize the dirty
pages bitmaps. colo_incoming_start_dirty_log() could be modified in a
similar way.

Cc: Stefano Stabellini <sstabellini@kernel.org>
Cc: Anthony Perard <anthony.perard@citrix.com>
Cc: Paul Durrant <paul@xen.org>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Hyman Huang <yong.huang@smartx.com>
Signed-off-by: Cédric Le Goater <clg@redhat.com>
Reviewed-by: Fabiano Rosas <farosas@suse.de>
Acked-by: Peter Xu <peterx@redhat.com>
Link: https://lore.kernel.org/r/20240320064911.545001-12-clg@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>

migration: Introduce ram_bitmaps_destroy()

We will use it in ram_init_bitmaps() to clear the allocated bitmaps when
support for error reporting is added to memory_global_dirty_log_start().

Signed-off-by: Cédric Le Goater <clg@redhat.com>
Reviewed-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Fabiano Rosas <farosas@suse.de>
Link: https://lore.kernel.org/r/20240320064911.545001-11-clg@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>

memory: Add Error** argument to .log_global_start() handler

Modify all .log_global_start() handlers to take an Error** parameter
and return a bool. Adapt memory_global_dirty_log_start() to interrupt
on the first error the loop on handlers. In such case, a rollback is
performed to stop dirty logging on all listeners where it was
previously enabled.

Cc: Stefano Stabellini <sstabellini@kernel.org>
Cc: Anthony Perard <anthony.perard@citrix.com>
Cc: Paul Durrant <paul@xen.org>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: David Hildenbrand <david@redhat.com>
Signed-off-by: Cédric Le Goater <clg@redhat.com>
Reviewed-by: Peter Xu <peterx@redhat.com>
Link: https://lore.kernel.org/r/20240320064911.545001-10-clg@redhat.com
[peterx: modify & enrich the comment for listener_add_address_space() ]
Signed-off-by: Peter Xu <peterx@redhat.com>

migration: Add Error** argument to .load_setup() handler

This will be useful to report errors at a higher level, mostly in VFIO
today.

Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Cédric Le Goater <clg@redhat.com>
Link: https://lore.kernel.org/r/20240320064911.545001-9-clg@redhat.com
[peterx: drop comment for ERRP_GUARD, per Markus]
Signed-off-by: Peter Xu <peterx@redhat.com>

migration: Add Error** argument to .save_setup() handler

The purpose is to record a potential error in the migration stream if
qemu_savevm_state_setup() fails. Most of the current .save_setup()
handlers can be modified to use the Error argument instead of managing
their own and calling locally error_report().

Cc: Nicholas Piggin <npiggin@gmail.com>
Cc: Harsh Prateek Bora <harshpb@linux.ibm.com>
Cc: Halil Pasic <pasic@linux.ibm.com>
Cc: Thomas Huth <thuth@redhat.com>
Cc: Eric Blake <eblake@redhat.com>
Cc: Vladimir Sementsov-Ogievskiy <vsementsov@yandex-team.ru>
Cc: John Snow <jsnow@redhat.com>
Cc: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@yandex-team.ru>
Signed-off-by: Cédric Le Goater <clg@redhat.com>
Link: https://lore.kernel.org/r/20240320064911.545001-8-clg@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>

migration: Add Error** argument to qemu_savevm_state_setup()

This prepares ground for the changes coming next which add an Error**
argument to the .save_setup() handler. Callers of qemu_savevm_state_setup()
now handle the error and fail earlier setting the migration state from
MIGRATION_STATUS_SETUP to MIGRATION_STATUS_FAILED.

In qemu_savevm_state(), move the cleanup to preserve the error
reported by .save_setup() handlers.

Since the previous behavior was to ignore errors at this step of
migration, this change should be examined closely to check that
cleanups are still correctly done.

Signed-off-by: Cédric Le Goater <clg@redhat.com>
Reviewed-by: Peter Xu <peterx@redhat.com>
Link: https://lore.kernel.org/r/20240320064911.545001-7-clg@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>

migration: Add Error** argument to vmstate_save()

This will prepare ground for future changes adding an Error** argument
to qemu_savevm_state_setup().

Reviewed-by: Prasad Pandit <pjp@fedoraproject.org>
Signed-off-by: Cédric Le Goater <clg@redhat.com>
Link: https://lore.kernel.org/r/20240320064911.545001-6-clg@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>

migration: Always report an error in ram_save_setup()

This will prepare ground for future changes adding an Error** argument
to the save_setup() handler. We need to make sure that on failure,
ram_save_setup() sets a new error.

Reviewed-by: Fabiano Rosas <farosas@suse.de>
Signed-off-by: Cédric Le Goater <clg@redhat.com>
Link: https://lore.kernel.org/r/20240320064911.545001-5-clg@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>

migration: Always report an error in block_save_setup()

This will prepare ground for future changes adding an Error** argument
to the save_setup() handler. We need to make sure that on failure,
block_save_setup() always sets a new error.

Cc: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Fabiano Rosas <farosas@suse.de>
Signed-off-by: Cédric Le Goater <clg@redhat.com>
Link: https://lore.kernel.org/r/20240320064911.545001-4-clg@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>

vfio: Always report an error in vfio_save_setup()

This will prepare ground for future changes adding an Error** argument
to the save_setup() handler. We need to make sure that on failure,
vfio_save_setup() always sets a new error.

Reviewed-by: Fabiano Rosas <farosas@suse.de>
Reviewed-by: Eric Auger <eric.auger@redhat.com>
Signed-off-by: Cédric Le Goater <clg@redhat.com>
Link: https://lore.kernel.org/r/20240320064911.545001-3-clg@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>

s390/stattrib: Add Error** argument to set_migrationmode() handler

This will prepare ground for future changes adding an Error** argument
to the save_setup() handler. We need to make sure that on failure,
set_migrationmode() always sets a new error. See the Rules section in
qapi/error.h.

Cc: Halil Pasic <pasic@linux.ibm.com>
Cc: Christian Borntraeger <borntraeger@linux.ibm.com>
Cc: Thomas Huth <thuth@redhat.com>
Reviewed-by: Fabiano Rosas <farosas@suse.de>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Cédric Le Goater <clg@redhat.com>
Link: https://lore.kernel.org/r/20240320064911.545001-2-clg@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>

tests/qtest/migration: Fix typo for vsock in SocketAddress_to_str

Signed-off-by: Het Gala <het.gala@nutanix.com>
Link: https://lore.kernel.org/r/20240319204840.211632-2-het.gala@nutanix.com
Signed-off-by: Peter Xu <peterx@redhat.com>

tests/qtest/migration: Add negative tests to validate migration QAPIs

Migration QAPI arguments - uri and channels are mutually exhaustive.
Add negative validation tests, one with both arguments present and
one with none present.

Signed-off-by: Het Gala <het.gala@nutanix.com>
Suggested-by: Fabiano Rosas <farosas@suse.de>
Reviewed-by: Fabiano Rosas <farosas@suse.de>
Link: https://lore.kernel.org/r/20240312202634.63349-9-het.gala@nutanix.com
Signed-off-by: Peter Xu <peterx@redhat.com>

tests/qtest/migration: Add multifd_tcp_plain test using list of channels instead of uri

Add a positive test to check multifd live migration but this time
using list of channels (restricted to 1) as the starting point
instead of simple uri string.

Signed-off-by: Het Gala <het.gala@nutanix.com>
Suggested-by: Fabiano Rosas <farosas@suse.de>
Reviewed-by: Fabiano Rosas <farosas@suse.de>
Link: https://lore.kernel.org/r/20240312202634.63349-8-het.gala@nutanix.com
Signed-off-by: Peter Xu <peterx@redhat.com>

tests/qtest/migration: Add channels parameter in migrate_qmp

Alter migrate_qmp() to allow use of channels parameter, but only
fill the uri with correct port number if there are no channels.
Here we don't want to allow the wrong cases of having both or
none (ex: migrate_qmp_fail).

Signed-off-by: Het Gala <het.gala@nutanix.com>
Suggested-by: Fabiano Rosas <farosas@suse.de>
Reviewed-by: Fabiano Rosas <farosas@suse.de>
Link: https://lore.kernel.org/r/20240312202634.63349-7-het.gala@nutanix.com
Signed-off-by: Peter Xu <peterx@redhat.com>

tests/qtest/migration: Add migrate_set_ports into migrate_qmp to update migration port value

migrate_get_connect_qdict gets qdict with the dst QEMU parameters.

migrate_set_ports() from list of channels reads each QDict for port,
and fills the port with correct value in case it was 0 in the test.

Signed-off-by: Het Gala <het.gala@nutanix.com>
Suggested-by: Fabiano Rosas <farosas@suse.de>
Reviewed-by: Fabiano Rosas <farosas@suse.de>
Link: https://lore.kernel.org/r/20240312202634.63349-6-het.gala@nutanix.com
Signed-off-by: Peter Xu <peterx@redhat.com>

tests/qtest/migration: Add channels parameter in migrate_qmp_fail

Alter migrate_qmp_fail() to allow both uri and channels
independently. For channels, convert string to a Dict.
No dealing with migrate_get_socket_address() here because
we will fail before starting the migration anyway.

Signed-off-by: Het Gala <het.gala@nutanix.com>
Suggested-by: Fabiano Rosas <farosas@suse.de>
Reviewed-by: Fabiano Rosas <farosas@suse.de>
Link: https://lore.kernel.org/r/20240312202634.63349-5-het.gala@nutanix.com
Signed-off-by: Peter Xu <peterx@redhat.com>

tests/qtest/migration: Replace migrate_get_connect_uri inplace of migrate_get_socket_address

Refactor migrate_get_socket_address to internally utilize 'socket-address'
parameter, reducing redundancy in the function definition.

migrate_get_socket_address implicitly converts SocketAddress into str.
Move migrate_get_socket_address inside migrate_get_connect_uri which
should return the uri string instead.

Signed-off-by: Het Gala <het.gala@nutanix.com>
Suggested-by: Fabiano Rosas <farosas@suse.de>
Reviewed-by: Fabiano Rosas <farosas@suse.de>
Link: https://lore.kernel.org/r/20240312202634.63349-4-het.gala@nutanix.com
Signed-off-by: Peter Xu <peterx@redhat.com>

tests/qtest/migration: Replace connect_uri and move migrate_get_socket_address inside migrate_qmp

Move the calls to migrate_get_socket_address() into migrate_qmp().
Get rid of connect_uri and replace it with args->connect_uri only
because 'to' object will help to generate connect_uri with the
correct port number.

Signed-off-by: Het Gala <het.gala@nutanix.com>
Suggested-by: Fabiano Rosas <farosas@suse.de>
Reviewed-by: Fabiano Rosas <farosas@suse.de>
Link: https://lore.kernel.org/r/20240312202634.63349-3-het.gala@nutanix.com
Signed-off-by: Peter Xu <peterx@redhat.com>

tests/qtest/migration: Add 'to' object into migrate_qmp()

Add the 'to' object into migrate_qmp(), so we can use
migrate_get_socket_address() inside migrate_qmp() to get
the port value. This is not applied to other migrate_qmp*
because they don't need the port.

Signed-off-by: Het Gala <het.gala@nutanix.com>
Suggested-by: Fabiano Rosas <farosas@suse.de>
Reviewed-by: Fabiano Rosas <farosas@suse.de>
Link: https://lore.kernel.org/r/20240312202634.63349-2-het.gala@nutanix.com
Signed-off-by: Peter Xu <peterx@redhat.com>

target/i386/translate.c: always write 32-bits for SGDT and SIDT

The various Intel CPU manuals claim that SGDT and SIDT can write either 24-bits
or 32-bits depending upon the operand size, but this is incorrect. Not only do
the Intel CPU manuals give contradictory information between processor
revisions, but this information doesn't even match real-life behaviour.

In fact, tests on real hardware show that the CPU always writes 32-bits for SGDT
and SIDT, and this behaviour is required for at least OS/2 Warp and WFW 3.11 with
Win32s to function correctly. Remove the masking applied due to the operand size
for SGDT and SIDT so that the TCG behaviour matches the behaviour on real
hardware.

Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Resolves: https://gitlab.com/qemu-project/qemu/-/issues/2198

--
MCA: Whilst I don't have a copy of OS/2 Warp handy, I've confirmed that this
patch fixes the issue in WFW 3.11 with Win32s. For more technical information I
highly recommend the excellent write-up at
https://www.os2museum.com/wp/sgdtsidt-fiction-and-reality/.
Message-ID: <20240419195147.434894-1-mark.cave-ayland@ilande.co.uk>

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

pythondeps.toml: warn about updates needed to docs/requirements.txt

docs/requirements.txt is expected by readthedocs and should be in sync
with pythondeps.toml.  Add a comment to both.

Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

accel/tcg/icount-common: Consolidate the use of warn_report_once()

Use warn_report_once() to get rid of the static local variable "notified".

Signed-off-by: Zhao Liu <zhao1.liu@intel.com>
Message-ID: <20240418100716.1085491-1-zhao1.liu@linux.intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386/cpu: Merge the warning and error messages for AMD HT check

Currently, the difference between warn_report_once() and
error_report_once() is the former has the "warning:" prefix, while the
latter does not have a similar level prefix.

At the meantime, considering that there is no error handling logic here,
and the purpose of error_report_once() is only to prompt the user with
an abnormal message, there is no need to use an error-level message here,
and instead we can just use a warning.

Therefore, downgrade the message in error_report_once() to warning, and
merge it into the previous warn_report_once().

Signed-off-by: Zhao Liu <zhao1.liu@intel.com>
Message-ID: <20240327103951.3853425-4-zhao1.liu@linux.intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386/cpu: Consolidate the use of warn_report_once()

The difference between error_printf() and error_report() is the latter
may contain more information, such as the name of the program
("qemu-system-x86_64").

Thus its variant error_report_once() and warn_report()'s variant
warn_report_once() can be used here to print the information only once
without a static local variable "ht_warned".

Signed-off-by: Zhao Liu <zhao1.liu@intel.com>
Message-ID: <20240327103951.3853425-3-zhao1.liu@linux.intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386/host-cpu: Consolidate the use of warn_report_once()

Use warn_report_once() to get rid of the static local variable "warned".

Signed-off-by: Zhao Liu <zhao1.liu@intel.com>
Message-ID: <20240327103951.3853425-2-zhao1.liu@linux.intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

kvm/tdx: Ignore memory conversion to shared of unassigned region

TDX requires vMMIO region to be shared.  For KVM, MMIO region is the region
which kvm memslot isn't assigned to (except in-kernel emulation).
qemu has the memory region for vMMIO at each device level.

While OVMF issues MapGPA(to-shared) conservatively on 32bit PCI MMIO
region, qemu doesn't find corresponding vMMIO region because it's before
PCI device allocation and memory_region_find() finds the device region, not
PCI bus region.  It's safe to ignore MapGPA(to-shared) because when guest
accesses those region they use GPA with shared bit set for vMMIO.  Ignore
memory conversion request of non-assigned region to shared and return
success.  Otherwise OVMF is confused and panics there.

Signed-off-by: Isaku Yamahata <isaku.yamahata@intel.com>
Signed-off-by: Xiaoyao Li <xiaoyao.li@intel.com>
Message-ID: <20240229063726.610065-35-xiaoyao.li@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

kvm/tdx: Don't complain when converting vMMIO region to shared

Because vMMIO region needs to be shared region, guest TD may explicitly
convert such region from private to shared.  Don't complain such
conversion.

Signed-off-by: Isaku Yamahata <isaku.yamahata@intel.com>
Signed-off-by: Xiaoyao Li <xiaoyao.li@intel.com>
Message-ID: <20240229063726.610065-34-xiaoyao.li@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

kvm: handle KVM_EXIT_MEMORY_FAULT

Upon an KVM_EXIT_MEMORY_FAULT exit, userspace needs to do the memory
conversion on the RAMBlock to turn the memory into desired attribute,
switching between private and shared.

Currently only KVM_MEMORY_EXIT_FLAG_PRIVATE in flags is valid when
KVM_EXIT_MEMORY_FAULT happens.

Note, KVM_EXIT_MEMORY_FAULT makes sense only when the RAMBlock has
guest_memfd memory backend.

Note, KVM_EXIT_MEMORY_FAULT returns with -EFAULT, so special handling is
added.

When page is converted from shared to private, the original shared
memory can be discarded via ram_block_discard_range(). Note, shared
memory can be discarded only when it's not back'ed by hugetlb because
hugetlb is supposed to be pre-allocated and no need for discarding.

Signed-off-by: Chao Peng <chao.p.peng@linux.intel.com>
Co-developed-by: Xiaoyao Li <xiaoyao.li@intel.com>
Signed-off-by: Xiaoyao Li <xiaoyao.li@intel.com>
Message-ID: <20240320083945.991426-13-michael.roth@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

physmem: Introduce ram_block_discard_guest_memfd_range()

When memory page is converted from private to shared, the original
private memory is back'ed by guest_memfd. Introduce
ram_block_discard_guest_memfd_range() for discarding memory in
guest_memfd.

Based on a patch by Isaku Yamahata <isaku.yamahata@intel.com>.

Signed-off-by: Xiaoyao Li <xiaoyao.li@intel.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Michael Roth <michael.roth@amd.com>
Message-ID: <20240320083945.991426-12-michael.roth@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

RAMBlock: make guest_memfd require uncoordinated discard

Some subsystems like VFIO might disable ram block discard, but guest_memfd
uses discard operations to implement conversions between private and
shared memory.  Because of this, sequences like the following can result
in stale IOMMU mappings:

1. allocate shared page
2. convert page shared->private
3. discard shared page
4. convert page private->shared
5. allocate shared page
6. issue DMA operations against that shared page

This is not a use-after-free, because after step 3 VFIO is still pinning
the page.  However, DMA operations in step 6 will hit the old mapping
that was allocated in step 1.

Address this by taking ram_block_discard_is_enabled() into account when
deciding whether or not to discard pages.

Since kvm_convert_memory()/guest_memfd doesn't implement a
RamDiscardManager handler to convey and replay discard operations,
this is a case of uncoordinated discard, which is blocked/released
by ram_block_discard_require().  Interestingly, this function had
no use so far.

Alternative approaches would be to block discard of shared pages, but
this would cause guests to consume twice the memory if they use VFIO;
or to implement a RamDiscardManager and only block uncoordinated
discard, i.e. use ram_block_coordinated_discard_require().

[Commit message mostly by Michael Roth <michael.roth@amd.com>]

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

HostMem: Add mechanism to opt in kvm guest memfd via MachineState

Add a new member "guest_memfd" to memory backends. When it's set
to true, it enables RAM_GUEST_MEMFD in ram_flags, thus private kvm
guest_memfd will be allocated during RAMBlock allocation.

Memory backend's @guest_memfd is wired with @require_guest_memfd
field of MachineState. It avoid looking up the machine in phymem.c.

MachineState::require_guest_memfd is supposed to be set by any VMs
that requires KVM guest memfd as private memory, e.g., TDX VM.

Signed-off-by: Xiaoyao Li <xiaoyao.li@intel.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Message-ID: <20240320083945.991426-8-michael.roth@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

kvm/memory: Make memory type private by default if it has guest memfd backend

KVM side leaves the memory to shared by default, which may incur the
overhead of paging conversion on the first visit of each page. Because
the expectation is that page is likely to private for the VMs that
require private memory (has guest memfd).

Explicitly set the memory to private when memory region has valid
guest memfd backend.

Signed-off-by: Xiaoyao Li <xiaoyao.li@intel.com>
Signed-off-by: Michael Roth <michael.roth@amd.com>
Message-ID: <20240320083945.991426-16-michael.roth@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

kvm: Enable KVM_SET_USER_MEMORY_REGION2 for memslot

Switch to KVM_SET_USER_MEMORY_REGION2 when supported by KVM.

With KVM_SET_USER_MEMORY_REGION2, QEMU can set up memory region that
backend'ed both by hva-based shared memory and guest memfd based private
memory.

Signed-off-by: Chao Peng <chao.p.peng@linux.intel.com>
Co-developed-by: Xiaoyao Li <xiaoyao.li@intel.com>
Signed-off-by: Xiaoyao Li <xiaoyao.li@intel.com>
Message-ID: <20240320083945.991426-10-michael.roth@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

RAMBlock: Add support of KVM private guest memfd

Add KVM guest_memfd support to RAMBlock so both normal hva based memory
and kvm guest memfd based private memory can be associated in one RAMBlock.

Introduce new flag RAM_GUEST_MEMFD. When it's set, it calls KVM ioctl to
create private guest_memfd during RAMBlock setup.

Allocating a new RAM_GUEST_MEMFD flag to instruct the setup of guest memfd
is more flexible and extensible than simply relying on the VM type because
in the future we may have the case that not all the memory of a VM need
guest memfd. As a benefit, it also avoid getting MachineState in memory
subsystem.

Note, RAM_GUEST_MEMFD is supposed to be set for memory backends of
confidential guests, such as TDX VM. How and when to set it for memory
backends will be implemented in the following patches.

Introduce memory_region_has_guest_memfd() to query if the MemoryRegion has
KVM guest_memfd allocated.

Signed-off-by: Xiaoyao Li <xiaoyao.li@intel.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Message-ID: <20240320083945.991426-7-michael.roth@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

kvm: Introduce support for memory_attributes

Introduce the helper functions to set the attributes of a range of
memory to private or shared.

This is necessary to notify KVM the private/shared attribute of each gpa
range. KVM needs the information to decide the GPA needs to be mapped at
hva-based shared memory or guest_memfd based private memory.

Signed-off-by: Xiaoyao Li <xiaoyao.li@intel.com>
Message-ID: <20240320083945.991426-11-michael.roth@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

trace/kvm: Split address space and slot id in trace_kvm_set_user_memory()

The upper 16 bits of kvm_userspace_memory_region::slot are
address space id. Parse it separately in trace_kvm_set_user_memory().

Signed-off-by: Xiaoyao Li <xiaoyao.li@intel.com>
Message-ID: <20240229063726.610065-5-xiaoyao.li@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

hw/i386/sev: Use legacy SEV VM types for older machine types

Newer 9.1 machine types will default to using the KVM_SEV_INIT2 API for
creating SEV/SEV-ES going forward. However, this API results in guest
measurement changes which are generally not expected for users of these
older guest types and can cause disruption if they switch to a newer
QEMU/kernel version. Avoid this by continuing to use the older
KVM_SEV_INIT/KVM_SEV_ES_INIT APIs for older machine types.

Signed-off-by: Michael Roth <michael.roth@amd.com>
Message-ID: <20240409230743.962513-4-michael.roth@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

i386/sev: Add 'legacy-vm-type' parameter for SEV guest objects

QEMU will currently automatically make use of the KVM_SEV_INIT2 API for
initializing SEV and SEV-ES guests verses the older
KVM_SEV_INIT/KVM_SEV_ES_INIT interfaces.

However, the older interfaces will silently avoid sync'ing FPU/XSAVE
state to the VMSA prior to encryption, thus relying on behavior and
measurements that assume the related fields to be allow zero.

With KVM_SEV_INIT2, this state is now synced into the VMSA, resulting in
measurements changes and, theoretically, behaviorial changes, though the
latter are unlikely to be seen in practice.

To allow a smooth transition to the newer interface, while still
providing a mechanism to maintain backward compatibility with VMs
created using the older interfaces, provide a new command-line
parameter:

  -object sev-guest,legacy-vm-type=true,...

and have it default to false.

Signed-off-by: Michael Roth <michael.roth@amd.com>
Message-ID: <20240409230743.962513-2-michael.roth@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: SEV: use KVM_SEV_INIT2 if possible

Implement support for the KVM_X86_SEV_VM and KVM_X86_SEV_ES_VM virtual
machine types, and the KVM_SEV_INIT2 function of KVM_MEMORY_ENCRYPT_OP.

These replace the KVM_SEV_INIT and KVM_SEV_ES_INIT functions, and have
several advantages:

- sharing the initialization sequence with SEV-SNP and TDX

- allowing arguments including the set of desired VMSA features

- protection against invalid use of KVM_GET/SET_* ioctls for guests
  with encrypted state

If the KVM_X86_SEV_VM and KVM_X86_SEV_ES_VM types are not supported,
fall back to KVM_SEV_INIT and KVM_SEV_ES_INIT (which use the
default x86 VM type).

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

target/i386: Implement mc->kvm_type() to get VM type

KVM is introducing a new API to create confidential guests, which
will be used by TDX and SEV-SNP but is also available for SEV and
SEV-ES.  The API uses the VM type argument to KVM_CREATE_VM to
identify which confidential computing technology to use.

Since there are no other expected uses of VM types, delegate
mc->kvm_type() for x86 boards to the confidential-guest-support
object pointed to by ms->cgs.

For example, if a sev-guest object is specified to confidential-guest-support,
like,

  qemu -machine ...,confidential-guest-support=sev0 \
       -object sev-guest,id=sev0,...

it will check if a VM type KVM_X86_SEV_VM or KVM_X86_SEV_ES_VM
is supported, and if so use them together with the KVM_SEV_INIT2
function of the KVM_MEMORY_ENCRYPT_OP ioctl. If not, it will fall back to
KVM_SEV_INIT and KVM_SEV_ES_INIT.

This is a preparatory work towards TDX and SEV-SNP support, but it
will also enable support for VMSA features such as DebugSwap, which
are only available via KVM_SEV_INIT2.

Co-developed-by: Xiaoyao Li <xiaoyao.li@intel.com>
Signed-off-by: Xiaoyao Li <xiaoyao.li@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: introduce x86-confidential-guest

Introduce a common superclass for x86 confidential guest implementations.
It will extend ConfidentialGuestSupportClass with a method that provides
the VM type to be passed to KVM_CREATE_VM.

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

KVM: remove kvm_arch_cpu_check_are_resettable

Board reset requires writing a fresh CPU state.  As far as KVM is
concerned, the only thing that blocks reset is that CPU state is
encrypted; therefore, kvm_cpus_are_resettable() can simply check
if that is the case.

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

KVM: track whether guest state is encrypted

So far, KVM has allowed KVM_GET/SET_* ioctls to execute even if the
guest state is encrypted, in which case they do nothing.  For the new
API using VM types, instead, the ioctls will fail which is a safer and
more robust approach.

The new API will be the only one available for SEV-SNP and TDX, but it
is also usable for SEV and SEV-ES.  In preparation for that, require
architecture-specific KVM code to communicate the point at which guest
state is protected (which must be after kvm_cpu_synchronize_post_init(),
though that might change in the future in order to suppor migration).
From that point, skip reading registers so that cpu->vcpu_dirty is
never true: if it ever becomes true, kvm_arch_put_registers() will
fail miserably.

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

runstate: skip initial CPU reset if reset is not actually possible

Right now, the system reset is concluded by a call to
cpu_synchronize_all_post_reset() in order to sync any changes
that the machine reset callback applied to the CPU state.

However, for VMs with encrypted state such as SEV-ES guests (currently
the only case of guests with non-resettable CPUs) this cannot be done,
because guest state has already been finalized by machine-init-done notifiers.
cpu_synchronize_all_post_reset() does nothing on these guests, and actually
we would like to make it fail if called once guest has been encrypted.
So, assume that boards that support non-resettable CPUs do not touch
CPU state and that all such setup is done before, at the time of
cpu_synchronize_all_post_init().

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

linux-headers: update to current kvm/next

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

scripts/update-linux-headers: Add bits.h to file imports

Signed-off-by: Michael Roth <michael.roth@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

scripts/update-linux-headers: Add setup_data.h to import list

Data structures like struct setup_data have been moved to a separate
setup_data.h header which bootparam.h relies on. Add setup_data.h to
the cp_portable() list and sync it along with the other header files.

Note that currently struct setup_data is stripped away as part of
generating bootparam.h, but that handling is no currently needed for
setup_data.h since it doesn't pull in many external
headers/dependencies. However, QEMU currently redefines struct
setup_data in hw/i386/x86.c, so that will need to be removed as part of
any header update that pulls in the new setup_data.h to avoid build
bisect breakage.

Because <asm/setup_data.h> is the first architecture specific #include
in include/standard-headers/, add a new sed substitution to rewrite
asm/ include to the standard-headers/asm-* subdirectory for the current
architecture.

And while at it, remove asm-generic/kvm_para.h from the list of
allowed includes: it does not have a matching substitution, and therefore
it would not be possible to use it on non-Linux systems where there is
no /usr/include/asm-generic/ directory.

Signed-off-by: Michael Roth <michael.roth@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

s390: Switch to use confidential_guest_kvm_init()

Use unified confidential_guest_kvm_init() for consistency with
other architectures.

Signed-off-by: Xiaoyao Li <xiaoyao.li@intel.com>
Message-Id: <20240229060038.606591-1-xiaoyao.li@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

ppc/pef: switch to use confidential_guest_kvm_init/reset()

Use the unified interface to call confidential guest related kvm_init()
and kvm_reset(), to avoid exposing pef specific functions.

As a bonus, pef.h goes away since there is no direct call from sPAPR
board code to PEF code anymore.

Signed-off-by: Xiaoyao Li <xiaoyao.li@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

i386/sev: Switch to use confidential_guest_kvm_init()

Use confidential_guest_kvm_init() instead of calling SEV
specific sev_kvm_init(). This allows the introduction of multiple
confidential-guest-support subclasses for different x86 vendors.

As a bonus, stubs are not needed anymore since there is no
direct call from target/i386/kvm/kvm.c to SEV code.

Signed-off-by: Xiaoyao Li <xiaoyao.li@intel.com>
Message-Id: <20240229060038.606591-1-xiaoyao.li@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

confidential guest support: Add kvm_init() and kvm_reset() in class

Different confidential VMs in different architectures all have the same
needs to do their specific initialization (and maybe resetting) stuffs
with KVM. Currently each of them exposes individual *_kvm_init()
functions and let machine code or kvm code to call it.

To facilitate the introduction of confidential guest technology from
different x86 vendors, add two virtual functions, kvm_init() and kvm_reset()
in ConfidentialGuestSupportClass, and expose two helpers functions for
invodking them.

Signed-off-by: Xiaoyao Li <xiaoyao.li@intel.com>
Message-Id: <20240229060038.606591-1-xiaoyao.li@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

hw/i386/acpi: Set PCAT_COMPAT bit only when pic is not disabled

A value 1 of PCAT_COMPAT (bit 0) of MADT.Flags indicates that the system
also has a PC-AT-compatible dual-8259 setup, i.e., the PIC.  When PIC
is not enabled (pic=off) for x86 machine, the PCAT_COMPAT bit needs to
be cleared.  The PIC probe should then print:

   [    0.155970] Using NULL legacy PIC

However, no such log printed in guest kernel unless PCAT_COMPAT is
cleared.

Signed-off-by: Xiaoyao Li <xiaoyao.li@intel.com>
Message-ID: <20240403145953.3082491-1-xiaoyao.li@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

q35: Introduce smm_ranges property for q35-pci-host

Add a q35 property to check whether or not SMM ranges, e.g. SMRAM, TSEG,
etc... exist for the target platform.  TDX doesn't support SMM and doesn't
play nice with QEMU modifying related guest memory ranges.

Signed-off-by: Isaku Yamahata <isaku.yamahata@linux.intel.com>
Co-developed-by: Sean Christopherson <sean.j.christopherson@intel.com>
Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>
Signed-off-by: Xiaoyao Li <xiaoyao.li@intel.com>
Signed-off-by: Michael Roth <michael.roth@amd.com>
Message-ID: <20240320083945.991426-19-michael.roth@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

pci-host/q35: Move PAM initialization above SMRAM initialization

In mch_realize(), process PAM initialization before SMRAM initialization so
that later patch can skill all the SMRAM related with a single check.

Signed-off-by: Isaku Yamahata <isaku.yamahata@intel.com>
Signed-off-by: Xiaoyao Li <xiaoyao.li@intel.com>
Signed-off-by: Michael Roth <michael.roth@amd.com>
Message-ID: <20240320083945.991426-18-michael.roth@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: Export RFDS bit to guests

Register File Data Sampling (RFDS) is a CPU side-channel vulnerability
that may expose stale register value. CPUs that set RFDS_NO bit in MSR
IA32_ARCH_CAPABILITIES indicate that they are not vulnerable to RFDS.
Similarly, RFDS_CLEAR indicates that CPU is affected by RFDS, and has
the microcode to help mitigate RFDS.

Make RFDS_CLEAR and RFDS_NO bits available to guests.

Signed-off-by: Pawan Gupta <pawan.kumar.gupta@linux.intel.com>
Reviewed-by: Xiaoyao Li <xiaoyao.li@intel.com>
Reviewed-by: Zhao Liu <zhao1.liu@intel.com>
Message-ID: <9a38877857392b5c2deae7e7db1b170d15510314.1710341348.git.pawan.kumar.gupta@linux.intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: Add new CPU model SierraForest

According to table 1-2 in Intel Architecture Instruction Set Extensions and
Future Features (rev 051) [1], SierraForest has the following new features
which have already been virtualized:

- CMPCCXADD CPUID.(EAX=7,ECX=1):EAX[bit 7]
- AVX-IFMA CPUID.(EAX=7,ECX=1):EAX[bit 23]
- AVX-VNNI-INT8 CPUID.(EAX=7,ECX=1):EDX[bit 4]
- AVX-NE-CONVERT CPUID.(EAX=7,ECX=1):EDX[bit 5]

Add above features to new CPU model SierraForest. Comparing with GraniteRapids
CPU model, SierraForest bare-metal removes the following features:

- HLE CPUID.(EAX=7,ECX=0):EBX[bit 4]
- RTM CPUID.(EAX=7,ECX=0):EBX[bit 11]
- AVX512F CPUID.(EAX=7,ECX=0):EBX[bit 16]
- AVX512DQ CPUID.(EAX=7,ECX=0):EBX[bit 17]
- AVX512_IFMA CPUID.(EAX=7,ECX=0):EBX[bit 21]
- AVX512CD CPUID.(EAX=7,ECX=0):EBX[bit 28]
- AVX512BW CPUID.(EAX=7,ECX=0):EBX[bit 30]
- AVX512VL CPUID.(EAX=7,ECX=0):EBX[bit 31]
- AVX512_VBMI CPUID.(EAX=7,ECX=0):ECX[bit 1]
- AVX512_VBMI2 CPUID.(EAX=7,ECX=0):ECX[bit 6]
- AVX512_VNNI CPUID.(EAX=7,ECX=0):ECX[bit 11]
- AVX512_BITALG CPUID.(EAX=7,ECX=0):ECX[bit 12]
- AVX512_VPOPCNTDQ CPUID.(EAX=7,ECX=0):ECX[bit 14]
- LA57 CPUID.(EAX=7,ECX=0):ECX[bit 16]
- TSXLDTRK CPUID.(EAX=7,ECX=0):EDX[bit 16]
- AMX-BF16 CPUID.(EAX=7,ECX=0):EDX[bit 22]
- AVX512_FP16 CPUID.(EAX=7,ECX=0):EDX[bit 23]
- AMX-TILE CPUID.(EAX=7,ECX=0):EDX[bit 24]
- AMX-INT8 CPUID.(EAX=7,ECX=0):EDX[bit 25]
- AVX512_BF16 CPUID.(EAX=7,ECX=1):EAX[bit 5]
- fast zero-length MOVSB CPUID.(EAX=7,ECX=1):EAX[bit 10]
- fast short CMPSB, SCASB CPUID.(EAX=7,ECX=1):EAX[bit 12]
- AMX-FP16 CPUID.(EAX=7,ECX=1):EAX[bit 21]
- PREFETCHI CPUID.(EAX=7,ECX=1):EDX[bit 14]
- XFD CPUID.(EAX=0xD,ECX=1):EAX[bit 4]
- EPT_PAGE_WALK_LENGTH_5 VMX_EPT_VPID_CAP(0x48c)[bit 7]

Add all features of GraniteRapids CPU model except above features to
SierraForest CPU model.

SierraForest doesn’t support TSX and RTM but supports TAA_NO. When RTM is
not enabled in host, KVM will not report TAA_NO. So, just don't include
TAA_NO in SierraForest CPU model.

[1] https://cdrdv2.intel.com/v1/dl/getContent/671368

Reviewed-by: Zhao Liu <zhao1.liu@intel.com>
Reviewed-by: Xiaoyao Li <xiaoyao.li@intel.com>
Signed-off-by: Tao Su <tao1.su@linux.intel.com>
Message-ID: <20240320021044.508263-1-tao1.su@linux.intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: Introduce Icelake-Server-v7 to enable TSX

When start L2 guest with both L1/L2 using Icelake-Server-v3 or above,
QEMU reports below warning:

"warning: host doesn't support requested feature: MSR(10AH).taa-no [bit 8]"

Reason is QEMU Icelake-Server-v3 has TSX feature disabled but enables taa-no
bit. It's meaningless that TSX isn't supported but still claim TSX is secure.
So L1 KVM doesn't expose taa-no to L2 if TSX is unsupported, then starting L2
triggers the warning.

Fix it by introducing a new version Icelake-Server-v7 which has both TSX
and taa-no features. Then guest can use TSX securely when it see taa-no.

This matches the production Icelake which supports TSX and isn't susceptible
to TSX Async Abort (TAA) vulnerabilities, a.k.a, taa-no.

Ideally, TSX should have being enabled together with taa-no since v3, but for
compatibility, we'd better to add v7 to enable it.

Fixes: d965dc35592d ("target/i386: Add ARCH_CAPABILITIES related bits into Icelake-Server CPU model")
Tested-by: Xiangfei Ma <xiangfeix.ma@intel.com>
Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
Message-ID: <20240320093138.80267-2-zhenzhong.duan@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

i386/kvm: Move architectural CPUID leaf generation to separate helper

Move the architectural (for lack of a better term) CPUID leaf generation
to a separate helper so that the generation code can be reused by TDX,
which needs to generate a canonical VM-scoped configuration.

For now this is just a cleanup, so keep the function static.

Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>
Signed-off-by: Xiaoyao Li <xiaoyao.li@intel.com>
Message-ID: <20240229063726.610065-23-xiaoyao.li@intel.com>
Reviewed-by: Xiaoyao Li <xiaoyao.li@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

Update version for 9.0.0 release

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

kvm: add support for guest physical bits

Query kvm for supported guest physical address bits, in cpuid
function 80000008, eax[23:16].  Usually this is identical to host
physical address bits.  With NPT or EPT being used this might be
restricted to 48 (max 4-level paging address space size) even if
the host cpu supports more physical address bits.

When set pass this to the guest, using cpuid too.  Guest firmware
can use this to figure how big the usable guest physical address
space is, so PCI bar mapping are actually reachable.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Xiaoyao Li <xiaoyao.li@intel.com>
Reviewed-by: Zhao Liu <zhao1.liu@intel.com>
Message-ID: <20240318155336.156197-2-kraxel@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: add guest-phys-bits cpu property

Allows to set guest-phys-bits (cpuid leaf 80000008, eax[23:16])
via -cpu $model,guest-phys-bits=$nr.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Message-ID: <20240318155336.156197-3-kraxel@redhat.com>
Reviewed-by: Zhao Liu <zhao1.liu@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

hw: Add compat machines for 9.1

Add 9.1 machine types for arm/i440fx/m68k/q35/s390x/spapr.

Reviewed-by: Cornelia Huck <cohuck@redhat.com>
Acked-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Harsh Prateek Bora <harshpb@linux.ibm.com>
Reviewed-by: Zhao Liu <zhao1.liu@intel.com>
Cc: Gavin Shan <gshan@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

kvm: use configs/ definition to conditionalize debug support

If an architecture adds support for KVM_CAP_SET_GUEST_DEBUG but QEMU does not
have the necessary code, QEMU will fail to build after updating kernel headers.
Avoid this by using a #define in config-target.h instead of KVM_CAP_SET_GUEST_DEBUG.

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

vga: move dirty memory region code together

Take into account split screen mode close to wrap around, which is the
other special case for dirty memory region computation.

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

vga: optimize computation of dirty memory region

The depth == 0 and depth == 15 have to be special cased because
width * depth / 8 does not provide the correct scanline length.
However, thanks to the recent reorganization of vga_draw_graphic()
the correct value of VRAM bits per pixel is available in "bits".

Use it (via the same "bwidth" computation that is used later in
the function), thus restricting the slow path to the wraparound case.

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

stubs: move monitor_fdsets_cleanup with other fdset stubs

Even though monitor_get_fd() has to remain separate because it is mocked by
tests/unit/test-util-sockets, monitor_fdsets_cleanup() is logically part
of the stubs for monitor/fds.c, so move it there.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-ID: <20240408155330.522792-19-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

stubs: include stubs only if needed

Currently it is not documented anywhere why some functions need to
be stubbed.

Group the files in stubs/meson.build according to who needs them, both
to reduce the size of the compilation and to clarify the use of stubs.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-ID: <20240408155330.522792-18-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

stubs: split record/replay stubs further

replay.c symbols are only needed by user mode emulation, with the
exception of replay_mode that is needed by both user mode emulation
(by way of qemu_guest_getrandom) and block layer tools (by way of
util/qemu-timer.c).

Since it is needed by libqemuutil rather than specific files that
are part of the tools and emulators, split the replay_mode stub
into its own file.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-ID: <20240408155330.522792-17-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

colo: move stubs out of stubs/

Since the colo stubs are needed exactly when the build options are not
enabled, move them together with the code they stub.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-ID: <20240408155330.522792-16-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

memory-device: move stubs out of stubs/

Since the memory-device stubs are needed exactly when the Kconfig symbols are not
needed, move them to hw/mem/.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-ID: <20240408155330.522792-15-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

ramfb: move stubs out of stubs/

Since the ramfb stubs are needed exactly when the Kconfig symbols are not
needed, move them to hw/display/ and compile them when ramfb.c is absent.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-ID: <20240408155330.522792-14-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

semihosting: move stubs out of stubs/

Since the semihosting stubs are needed exactly when the Kconfig symbols
are not needed, move them to semihosting/ and conditionalize them
on CONFIG_SEMIHOSTING and/or CONFIG_SYSTEM_ONLY.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-ID: <20240408155330.522792-13-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

hw/virtio: move stubs out of stubs/

Since the virtio memory device stubs are needed exactly when the
Kconfig symbol is not enabled, they can be placed in hw/virtio/ and
conditionalized on CONFIG_VIRTIO_MD.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-ID: <20240408155330.522792-12-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

hw/usb: move stubs out of stubs/

Since the USB stubs are needed exactly when the Kconfig symbols are not
enabled, they can be placed in hw/usb/ and conditionalized on CONFIG_USB.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-ID: <20240408155330.522792-11-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

stubs: remove obsolete stubs

These file define functions are are not called from common code
anymore.  Delete those functions and, if applicable, the entire files.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Acked-by: Richard Henderson <richard.henderson@linaro.org>
Message-ID: <20240408155330.522792-10-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

hw: Include minimal source set in user emulation build

Only the files in hwcore_ss[] are required to link a user emulation
binary.

Have meson process the hw/ sub-directories if system emulation is
selected, otherwise directly process hw/core/ to get hwcore_ss[], which
is the only set required by user emulation.

This removes about 10% from the time needed to run
"../configure --disable-system --disable-tools --disable-guest-agent".

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-ID: <20240404194757.9343-8-philmd@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-ID: <20240408155330.522792-9-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

hw/core: Move system emulation files to system_ss

hotplug.c, qdev-hotplug.c and reset.c are not used by user emulation
and need not be included in hwcore_ss.  Move them to system_ss, where
they belong, by letting the linker pull in the stubs when needed.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-ID: <20240408155330.522792-8-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

util/qemu-config: Extract QMP commands to qemu-config-qmp.c

QMP is irrelevant for user emulation. Extract the code
related to QMP in a different source file, which won't
be build for user emulation binaries. This avoid pulling
pointless code.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-ID: <20240404194757.9343-5-philmd@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-ID: <20240408155330.522792-7-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

yank: only build if needed

The yank feature is not used in user emulation.

Suggested-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-ID: <20240408155330.522792-6-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

tests/unit: match some unit tests to corresponding feature switches

Try not to test code that is not used by user mode emulation, or by the
block layer, unless they are being compiled; and fix test-timed-average
which was not compiled with --disable-system --enable-tools.

This is by no means complete, it only touches the more blatantly
wrong cases.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-ID: <20240408155330.522792-5-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

ebpf: Restrict to system emulation

eBPF is not used in user emulation.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-ID: <20240404194757.9343-2-philmd@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-ID: <20240408155330.522792-4-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

tests: only build plugins if TCG is enabled

There is no way to use them for testing, if all the available
accelerators use hardware virtualization.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-ID: <20240408155330.522792-3-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

meson: do not link pixman automatically into all targets

The dependency on pixman is listed manually in all sourcesets that need it.
There is no need to bring into libqemuutil, since there is nothing in
util/ that needs pixman either.

Reported-by: Michael Tokarev <mjt@tls.msk.ru>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-ID: <20240408155330.522792-2-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

Update version for v9.0.0-rc4 release

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

usb-storage: Fix BlockConf defaults

Commit 30896374 started to pass the full BlockConf from usb-storage to
scsi-disk, while previously only a few select properties would be
forwarded. This enables the user to set more properties, e.g. the block
size, that are actually taking effect.

However, now the calls to blkconf_apply_backend_options() and
blkconf_blocksizes() in usb_msd_storage_realize() that modify some of
these properties take effect, too, instead of being silently ignored.
This means at least that the block sizes get an unconditional default of
512 bytes before the configuration is passed to scsi-disk.

Before commit 30896374, the property wouldn't be set for scsi-disk and
therefore the device dependent defaults would apply - 512 for scsi-hd,
but 2048 for scsi-cd. The latter default has now become 512, too, which
makes at least Windows 11 installation fail when installing from
usb-storage.

Fix this by simply not calling these functions any more in usb-storage
and passing BlockConf on unmodified (except for the BlockBackend). The
same functions are called by the SCSI code anyway and it sets the right
defaults for the actual media type.

Fixes: 308963746169 ('scsi: Don't ignore most usb-storage properties')
Resolves: https://gitlab.com/qemu-project/qemu/-/issues/2260
Reported-by: Jonas Svensson
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Hanna Czenczek <hreitz@redhat.com>
Message-id: 20240412144202.13786-1-kwolf@redhat.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

Merge tag 'hw-misc-20240415' of https://github.com/philmd/qemu into staging

Misc HW patch queue

Fixes for hardware used by machines running AmigaOS.

# -----BEGIN PGP SIGNATURE-----
#
# iQIzBAABCAAdFiEE+qvnXhKRciHc/Wuy4+MsLN6twN4FAmYdP44ACgkQ4+MsLN6t
# wN5nZw//enjS8WL5jRk77FogdJW8Xa4JIsfrsDgZrZJ+Pzj378ssq6oIJgELAgFg
# 6h8CCP9hbS1TML984w907Psl0KP1dG/ar1Egu1yMaJT4c6HULwZeyKdzYpWm2E1R
# e4CCip+Zh33wy8TGivyodSSCN0oQgOLv2h1IqFArZ4n3TKMirhTyK+otzXXbXmyw
# U9ZlYVUxS3zljcFz6ZVoYISc9cNqbZe1GI6R6KvXqX/frvZI1KF/GLZovJiKR25H
# IbF+wfCbD/4sTPX5AR/gY5XfKd3zthFtZlbBViRawmTs6BPlcV9p2BXa4V3eXMBP
# +WXTNz+vRtGBUunEMSBlFWdz4ka4Q65MU+q7DLPdaCIOFOn9w3VDINotpL8oV8Nm
# e4IsM2Du5sUf0QSRopPFsorFY70kW5mH+WkF1MTXfTqZTZy/I2meTD5s1OkZLJA5
# g9+o17bn25jtZvJnEmAilVdopBSBclmniAsR9A1sCGooyVjn3Byo5ylcLTNIQ+de
# nScnyR0cvKqBjKkmMOLbDHo/sszH7jAYqedv7Aoh2dS8/uk3KuHtgi6GeLhSYF5Y
# ZTCYbFnpuohQ8ueOL9oa3abYUCzQBu+UivgWdSVhgA7W97zihqIj2oWmorIwBpc9
# uuaZBOpTyzhGUafRS6/J6pSTxcWIL3HZqzAQMz9D1kGSGlXsS/s=
# =xMfK
# -----END PGP SIGNATURE-----
# gpg: Signature made Mon 15 Apr 2024 15:54:06 BST
# gpg:                using RSA key FAABE75E12917221DCFD6BB2E3E32C2CDEADC0DE
# gpg: Good signature from "Philippe Mathieu-Daudé (F4BUG) <f4bug@amsat.org>" [full]
# Primary key fingerprint: FAAB E75E 1291 7221 DCFD  6BB2 E3E3 2C2C DEAD C0DE

* tag 'hw-misc-20240415' of https://github.com/philmd/qemu:
  hw/pci-host/ppc440_pcix: Do not expose a bridge device on PCI bus
  hw/isa/vt82c686: Keep track of PIRQ/PINT pins separately

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

Merge tag 'for_upstream' of https://git./virt/kvm/mst/qemu into staging

virtio: bugfix

A last minute fix for a use of a vector after it's released.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
# -----BEGIN PGP SIGNATURE-----
#
# iQFDBAABCAAtFiEEXQn9CHHI+FuUyooNKB8NuNKNVGkFAmYdBssPHG1zdEByZWRo
# YXQuY29tAAoJECgfDbjSjVRpTHcH/Rtl2jNJ5myZOuEylw+T6/GSvyEne6CoreHK
# zUNPxmXY+uJzCskXkJXyd4uIaci5iIH1JC9Tc0FzFYaYrTsoA1dlQridqoajKyN5
# E6zjKqepi3sLnvDE1VbZ1kVcNEX2xSAFX++iv4Rbn4HHO49yKR0jNajusTOsq505
# NObgNQXK/Yj1q0IXYrWDETV7xywpQqiiAzwnmhi6ac72+trqmPrUXnUulhitWR3K
# iZBuGxAHn9c/ilW3J4FeSbqe6sC/AhqUz3RSM6dB+rkpvA0E675T526uVMWxND2H
# auE+ou0kzZ8HNit3AHBg8316seHXzWP+ndVEZlifX33HoR1pltY=
# =H3M5
# -----END PGP SIGNATURE-----
# gpg: Signature made Mon 15 Apr 2024 11:51:55 BST
# gpg:                using RSA key 5D09FD0871C8F85B94CA8A0D281F0DB8D28D5469
# gpg:                issuer "mst@redhat.com"
# gpg: Good signature from "Michael S. Tsirkin <mst@kernel.org>" [full]
# gpg:                 aka "Michael S. Tsirkin <mst@redhat.com>" [full]
# Primary key fingerprint: 0270 606B 6F3C DF3D 0B17  0970 C350 3912 AFBE 8E67
#      Subkey fingerprint: 5D09 FD08 71C8 F85B 94CA  8A0D 281F 0DB8 D28D 5469

* tag 'for_upstream' of https://git.kernel.org/pub/scm/virt/kvm/mst/qemu:
  virtio-pci: fix use of a released vector

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

hw/pci-host/ppc440_pcix: Do not expose a bridge device on PCI bus

Real 460EX SoC apparently does not expose a bridge device and having
it appear on PCI bus confuses an AmigaOS file system driver that uses
this to detect which machine it is running on.

Cc: qemu-stable@nongnu.org
Signed-off-by: BALATON Zoltan <balaton@eik.bme.hu>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-ID: <20240411192443.B4D644E6026@zero.eik.bme.hu>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>

hw/isa/vt82c686: Keep track of PIRQ/PINT pins separately

Move calculation of mask after the switch which sets the function
number for PIRQ/PINT pins to make sure the state of these pins are
kept track of separately and IRQ is raised if any of them is active.

Cc: qemu-stable@nongnu.org
Fixes: 7e01bd80c1 hw/isa/vt82c686: Bring back via_isa_set_irq()
Signed-off-by: BALATON Zoltan <balaton@eik.bme.hu>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-ID: <20240410222543.0EA534E6005@zero.eik.bme.hu>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>

virtio-pci: fix use of a released vector

During the booting process of the non-standard image, the behavior of the
called function in qemu is as follows:

1. vhost_net_stop() was triggered by guest image. This will call the function
virtio_pci_set_guest_notifiers() with assgin= false,
virtio_pci_set_guest_notifiers(） will release the irqfd for vector 0

2. virtio_reset() was triggered, this will set configure vector to VIRTIO_NO_VECTOR

3.vhost_net_start() was called (at this time, the configure vector is
still VIRTIO_NO_VECTOR) and then call virtio_pci_set_guest_notifiers() with
assgin=true, so the irqfd for vector 0 is still not "init" during this process

4. The system continues to boot and sets the vector back to 0. After that
msix_fire_vector_notifier() was triggered to unmask the vector 0 and  meet the crash

To fix the issue, we need to support changing the vector after VIRTIO_CONFIG_S_DRIVER_OK is set.

(gdb) bt
0  __pthread_kill_implementation (threadid=<optimized out>, signo=signo@entry=6, no_tid=no_tid@entry=0)
    at pthread_kill.c:44
1  0x00007fc87148ec53 in __pthread_kill_internal (signo=6, threadid=<optimized out>) at pthread_kill.c:78
2  0x00007fc87143e956 in __GI_raise (sig=sig@entry=6) at ../sysdeps/posix/raise.c:26
3  0x00007fc8714287f4 in __GI_abort () at abort.c:79
4  0x00007fc87142871b in __assert_fail_base
    (fmt=0x7fc8715bbde0 "%s%s%s:%u: %s%sAssertion `%s' failed.\n%n", assertion=0x5606413efd53 "ret == 0", file=0x5606413ef87d "../accel/kvm/kvm-all.c", line=1837, function=<optimized out>) at assert.c:92
5  0x00007fc871437536 in __GI___assert_fail
    (assertion=0x5606413efd53 "ret == 0", file=0x5606413ef87d "../accel/kvm/kvm-all.c", line=1837, function=0x5606413f06f0 <__PRETTY_FUNCTION__.19> "kvm_irqchip_commit_routes") at assert.c:101
6  0x0000560640f884b5 in kvm_irqchip_commit_routes (s=0x560642cae1f0) at ../accel/kvm/kvm-all.c:1837
7  0x0000560640c98f8e in virtio_pci_one_vector_unmask
    (proxy=0x560643c65f00, queue_no=4294967295, vector=0, msg=..., n=0x560643c6e4c8)
    at ../hw/virtio/virtio-pci.c:1005
8  0x0000560640c99201 in virtio_pci_vector_unmask (dev=0x560643c65f00, vector=0, msg=...)
    at ../hw/virtio/virtio-pci.c:1070
9  0x0000560640bc402e in msix_fire_vector_notifier (dev=0x560643c65f00, vector=0, is_masked=false)
    at ../hw/pci/msix.c:120
10 0x0000560640bc40f1 in msix_handle_mask_update (dev=0x560643c65f00, vector=0, was_masked=true)
    at ../hw/pci/msix.c:140
11 0x0000560640bc4503 in msix_table_mmio_write (opaque=0x560643c65f00, addr=12, val=0, size=4)
    at ../hw/pci/msix.c:231
12 0x0000560640f26d83 in memory_region_write_accessor
    (mr=0x560643c66540, addr=12, value=0x7fc86b7bc628, size=4, shift=0, mask=4294967295, attrs=...)
    at ../system/memory.c:497
13 0x0000560640f270a6 in access_with_adjusted_size

     (addr=12, value=0x7fc86b7bc628, size=4, access_size_min=1, access_size_max=4, access_fn=0x560640f26c8d <memory_region_write_accessor>, mr=0x560643c66540, attrs=...) at ../system/memory.c:573
14 0x0000560640f2a2b5 in memory_region_dispatch_write (mr=0x560643c66540, addr=12, data=0, op=MO_32, attrs=...)
    at ../system/memory.c:1521
15 0x0000560640f37bac in flatview_write_continue
    (fv=0x7fc65805e0b0, addr=4273803276, attrs=..., ptr=0x7fc871e9c028, len=4, addr1=12, l=4, mr=0x560643c66540)
    at ../system/physmem.c:2714
16 0x0000560640f37d0f in flatview_write
    (fv=0x7fc65805e0b0, addr=4273803276, attrs=..., buf=0x7fc871e9c028, len=4) at ../system/physmem.c:2756
17 0x0000560640f380bf in address_space_write
    (as=0x560642161ae0 <address_space_memory>, addr=4273803276, attrs=..., buf=0x7fc871e9c028, len=4)
    at ../system/physmem.c:2863
18 0x0000560640f3812c in address_space_rw
    (as=0x560642161ae0 <address_space_memory>, addr=4273803276, attrs=..., buf=0x7fc871e9c028, len=4, is_write=true) at ../system/physmem.c:2873
--Type <RET> for more, q to quit, c to continue without paging--
19 0x0000560640f8aa55 in kvm_cpu_exec (cpu=0x560642f205e0) at ../accel/kvm/kvm-all.c:2915
20 0x0000560640f8d731 in kvm_vcpu_thread_fn (arg=0x560642f205e0) at ../accel/kvm/kvm-accel-ops.c:51
21 0x00005606411949f4 in qemu_thread_start (args=0x560642f292b0) at ../util/qemu-thread-posix.c:541
22 0x00007fc87148cdcd in start_thread (arg=<optimized out>) at pthread_create.c:442
23 0x00007fc871512630 in clone3 () at ../sysdeps/unix/sysv/linux/x86_64/clone3.S:81
(gdb)

MST: coding style and typo fixups

Fixes: f9a09ca3ea ("vhost: add support for configure interrupt")
Cc: qemu-stable@nongnu.org
Signed-off-by: Cindy Lu <lulu@redhat.com>
Message-ID: <2321ade5f601367efe7380c04e3f61379c59b48f.1713173550.git.mst@redhat.com>
Cc: Lei Yang <leiyang@redhat.com>
Cc: Jason Wang <jasowang@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Tested-by: Cindy Lu <lulu@redhat.com>