vhdx: Require GRAPH_RDLOCK for accessing a node's parent list

This adds GRAPH_RDLOCK annotations to declare that functions accessing
the parent list of a node need to hold a reader lock for the graph. As
it happens, they already do.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20230504115750.54437-12-kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

nbd: Mark nbd_co_do_establish_connection() and callers GRAPH_RDLOCK

This adds GRAPH_RDLOCK annotations to declare that callers of
nbd_co_do_establish_connection() need to hold a reader lock for the
graph.

Signed-off-by: Emanuele Giuseppe Esposito <eesposit@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20230504115750.54437-11-kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

nbd: Remove nbd_co_flush() wrapper function

The only thing nbd_co_flush() does is call nbd_client_co_flush(). Just
use that function directly in the BlockDriver definitions and remove the
wrapper.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20230504115750.54437-10-kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

block: .bdrv_open is non-coroutine and unlocked

Drivers were a bit confused about whether .bdrv_open can run in a
coroutine and whether or not it holds a graph lock.

It cannot keep a graph lock from the caller across the whole function
because it both changes the graph (requires a writer lock) and does I/O
(requires a reader lock). Therefore, it should take these locks
internally as needed.

The functions used to be called in coroutine context during image
creation. This was buggy for other reasons, and as of commit 32192301,
all block drivers go through no_co_wrappers. So it is not called in
coroutine context any more.

Fix qcow2 and qed to work with the correct assumptions: The graph lock
needs to be taken internally instead of just assuming it's already
there, and the coroutine path is dead code that can be removed.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20230504115750.54437-9-kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

graph-lock: Fix GRAPH_RDLOCK_GUARD*() to be reader lock

GRAPH_RDLOCK_GUARD() and GRAPH_RDLOCK_GUARD_MAINLOOP() only take a
reader lock for the graph, so the correct annotation for them to use is
TSA_ASSERT_SHARED rather than TSA_ASSERT.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20230504115750.54437-8-kwolf@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

graph-lock: Add GRAPH_UNLOCKED(_PTR)

For some functions, it is part of their interface to be called without
holding the graph lock. Add a new macro to document this.

The macro expands to TSA_EXCLUDES(), which is a relatively weak check
because it passes in cases where the compiler just doesn't know if the
lock is held. Function pointers can't be checked at all. Therefore, its
primary purpose is documentation.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20230504115750.54437-7-kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

test-bdrv-drain: Don't modify the graph in coroutines

test-bdrv-drain contains a few test cases that are run both in coroutine
and non-coroutine context. Running the entire code including the setup
and shutdown in coroutines is incorrect because graph modifications can
generally not happen in coroutines.

Change the test so that creating and destroying the test nodes and
BlockBackends always happens outside of coroutine context.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20230504115750.54437-6-kwolf@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

iotests: Test resizing image attached to an iothread

This tests that trying to resize an image with QMP block_resize doesn't
hang or otherwise fail when the image is attached to a device running in
an iothread.

This is a regression test for the recent fix that changed
qmp_block_resize, which is a coroutine based QMP handler, to avoid
calling no_coroutine_fns directly.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20230509134133.373408-1-kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

block: Don't call no_coroutine_fns in qmp_block_resize()

This QMP handler runs in a coroutine, so it must use the corresponding
no_co_wrappers instead.

Fixes: https://bugzilla.redhat.com/show_bug.cgi?id=2185688
Cc: qemu-stable@nongnu.org
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20230504115750.54437-5-kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

block: bdrv/blk_co_unref() for calls in coroutine context

These functions must not be called in coroutine context, because they
need write access to the graph.

Cc: qemu-stable@nongnu.org
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20230504115750.54437-4-kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

block: Consistently call bdrv_activate() outside coroutine

Migration code can call bdrv_activate() in coroutine context, whereas
other callers call it outside of coroutines. As it calls other code that
is not supposed to run in coroutines, standardise on running outside of
coroutines.

This adds a no_co_wrapper to switch to the main loop before calling
bdrv_activate().

Cc: qemu-stable@nongnu.org
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20230504115750.54437-3-kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

qcow2: Don't call bdrv_getlength() in coroutine_fns

There is a bdrv_co_getlength() now, which should be used in coroutine
context.

This requires adding GRAPH_RDLOCK to some functions so that this still
compiles with TSA because bdrv_co_getlength() is GRAPH_RDLOCK.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20230504115750.54437-2-kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

migration: Attempt disk reactivation in more failure scenarios

Commit fe904ea824 added a fail_inactivate label, which tries to
reactivate disks on the source after a failure while s->state ==
MIGRATION_STATUS_ACTIVE, but didn't actually use the label if
qemu_savevm_state_complete_precopy() failed.  This failure to
reactivate is also present in commit 6039dd5b1c (also covering the new
s->state == MIGRATION_STATUS_DEVICE state) and 403d18ae (ensuring
s->block_inactive is set more reliably).

Consolidate the two labels back into one - no matter HOW migration is
failed, if there is any chance we can reach vm_start() after having
attempted inactivation, it is essential that we have tried to restart
disks before then.  This also makes the cleanup more like
migrate_fd_cancel().

Suggested-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <20230502205212.134680-1-eblake@redhat.com>
Acked-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

iotests/nbd-reconnect-on-open: Fix NBD socket path

Socket paths need to be short to avoid failures. This is why there is a
iotests.sock_dir (defaulting to /tmp) separate from the disk image base
directory.

Make use of it to fix failures in too deeply nested test directories.

Fixes: ab7f7e67a7e7b49964109501dfcde4ec29bae60e
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20230503165019.8867-1-kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@yandex-team.ru>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

block: Fix use after free in blockdev_mark_auto_del()

job_cancel_locked() drops the job list lock temporarily and it may call
aio_poll(). We must assume that the list has changed after this call.
Also, with unlucky timing, it can end up freeing the job during
job_completed_txn_abort_locked(), making the job pointer invalid, too.

For both reasons, we can't just continue at block_job_next_locked(job).
Instead, start at the head of the list again after job_cancel_locked()
and skip those jobs that we already cancelled (or that are completing
anyway).

Cc: qemu-stable@nongnu.org
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20230503140142.474404-1-kwolf@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

aio-wait: avoid AioContext lock in aio_wait_bh_oneshot()

There is no need for the AioContext lock in aio_wait_bh_oneshot().
It's easy to remove the lock from existing callers and then switch from
AIO_WAIT_WHILE() to AIO_WAIT_WHILE_UNLOCKED() in aio_wait_bh_oneshot().

Document that the AioContext lock should not be held across
aio_wait_bh_oneshot(). Holding a lock across aio_poll() can cause
deadlock so we don't want callers to do that.

This is a step towards getting rid of the AioContext lock.

Cc: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20230404153307.458883-1-stefanha@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Emanuele Giuseppe Esposito <eesposit@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

block: add missing coroutine_fn annotations

After the recent introduction of many new coroutine callbacks,
a couple calls from non-coroutine_fn to coroutine_fn have sneaked
in; fix them.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20230406101752.242125-1-pbonzini@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

block: add configure options for excluding vmdk, vhdx and vpc

Let's add --enable / --disable configure options for these formats,
so that those who don't need them may not build them.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@yandex-team.ru>
Message-Id: <20230421092758.814122-1-vsementsov@yandex-team.ru>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

Merge tag 'vfio-updates-20230509.0' of https://gitlab.com/alex.williamson/qemu into staging

VFIO updates 2023-05-09

 * Add vf-token device option allowing QEMU to assign VFs where the PF
   is managed by a userspace driver. (Minwoo Im)

 * Skip log_sync during migration setup as a potential source of failure
   and likely source of redundancy. (Avihai Horon)

 * Virtualize PCIe Resizable BAR capability rather than hiding it,
   exposing only the current size as available. (Alex Williamson)

# -----BEGIN PGP SIGNATURE-----
#
# iQJPBAABCAA5FiEEQvbATlQL0amee4qQI5ubbjuwiyIFAmRaqfobHGFsZXgud2ls
# bGlhbXNvbkByZWRoYXQuY29tAAoJECObm247sIsiwNYP/2KtCbKqylnGPuwLbRMP
# HC4Id4mme7jUribmhM7FP57nQrb0tgnQoGvalkmB6M3833e3p4ivH2ezTyPxIawx
# UH4mAEBtR03rxh54eVBbOvDVf+XHd6qll/rFw5dBI0C5s7JQyMOourNRLTZLvqzD
# 2bwI7dfQzWbXWPj8QGPmDti9wbeATZ3RjqC7onoWq6A6Cw4aRGj1gHBQH9v81iA+
# m8hnZh+e5eFkQRc4mPXxFjm1Kw6ZYXWGoEEZrYPXvQn9+3MDCLcNb++KIrLsGujP
# qOnZG534vs+EZtUsGI8F02CBBXMAQFuBZhxCtuuG8iI9OQSE6R3E29iIc0Lpz5aO
# s8rN5OW4m7wXPdGkU1/7/N7kdeZvg+R8Jc4ozx3Mez3eSFbVkABSSX9vyvdHAezi
# 02Np1+ZBldZWBbBhYbWfqhvcg4iYNnHknSkS2CYY8jdsGttbrNY2f7Xllf3KC/Iv
# 6Un5WpU//0LuJjmH6onzswUUEmulchzR7OpBj68jFsB8rnTaZWM4Sqb/Jx+KXlRB
# BnNck0PCPoblpT8lgjAD3H9NaXx3mdVsml8i/7YIZjx8Zc4eanRGlsH9DmnHbB7U
# i4orDvL3SR3ZKVy6Zssti5jt8GwrEnqg97uTbS/jiTai1tOCP9n6U4T/wslHIUR4
# rIxvyJnmqrPAiWtVF+0cvGmT
# =VTJU
# -----END PGP SIGNATURE-----
# gpg: Signature made Tue 09 May 2023 09:15:54 PM BST
# gpg:                using RSA key 42F6C04E540BD1A99E7B8A90239B9B6E3BB08B22
# gpg:                issuer "alex.williamson@redhat.com"
# gpg: Good signature from "Alex Williamson <alex.williamson@redhat.com>" [undefined]
# gpg:                 aka "Alex Williamson <alex@shazbot.org>" [undefined]
# gpg:                 aka "Alex Williamson <alwillia@redhat.com>" [undefined]
# gpg:                 aka "Alex Williamson <alex.l.williamson@gmail.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: 42F6 C04E 540B D1A9 9E7B  8A90 239B 9B6E 3BB0 8B22

* tag 'vfio-updates-20230509.0' of https://gitlab.com/alex.williamson/qemu:
  vfio/pci: Static Resizable BAR capability
  vfio/migration: Skip log_sync during migration SETUP state
  vfio/pci: add support for VF token

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

target/loongarch: Terminate vmstate subsections list

This list requires a NULL terminator.

Fixes: 16f5396cec23 ("target/loongarch: Add LSX data type VReg")
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20230510062405.127260-1-richard.henderson@linaro.org>

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

* target/i386: improved EPYC models
* more removal of mb_read/mb_set
* bump _WIN32_WINNT to the Windows 8 API
* fix for modular builds with --disable-system

# -----BEGIN PGP SIGNATURE-----
#
# iQFIBAABCAAyFiEE8TM4V0tmI4mGbHaCv/vSX3jHroMFAmRZK7wUHHBib256aW5p
# QHJlZGhhdC5jb20ACgkQv/vSX3jHroObngf8D6A5l1QQAnImRrZAny6HZV/9xseD
# 9QhkUW3fxXlUhb8tXomv2BlT8h9GzLIN6aWvcCotT+xK3kAX7mRcYKgPMr9CYL7y
# vev/hh+B6RY1CJ/xPT09/BMVjkj50AL0O/OuWMhcQ5nCO7F2sdMjMrsYqqeZcjYf
# zx9RTX7gVGt+wWFHxgCgdfL0kfgzexK55YuZU0vLzcA+pYsZWoEfW+fKBIf4rzDV
# r9M6mDBUkHBQ0rIVC3QFloAXnYb1JrpeqqL2i2qwhAkLz8LyGqk3lZF20hE/04im
# XZcZjWO5pxAxIEPeTken+2x1n8tn2BLkMtvwJdV5TpvICCFRtPZlbH79qw==
# =rXLN
# -----END PGP SIGNATURE-----
# gpg: Signature made Mon 08 May 2023 06:05:00 PM BST
# 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:
  meson: leave unnecessary modules out of the build
  docs: clarify --without-default-devices
  target/i386: Add EPYC-Genoa model to support Zen 4 processor series
  target/i386: Add VNMI and automatic IBRS feature bits
  target/i386: Add missing feature bits in EPYC-Milan model
  target/i386: Add feature bits for CPUID_Fn80000021_EAX
  target/i386: Add a couple of feature bits in 8000_0008_EBX
  target/i386: Add new EPYC CPU versions with updated cache_info
  target/i386: allow versioned CPUs to specify new cache_info
  include/qemu/osdep.h: Bump _WIN32_WINNT to the Windows 8 API
  MAINTAINERS: add stanza for Kconfig files
  tb-maint: do not use mb_read/mb_set
  call_rcu: stop using mb_set/mb_read
  test-aio-multithread: simplify test_multi_co_schedule
  test-aio-multithread: do not use mb_read/mb_set for simple flags
  rcu: remove qatomic_mb_set, expand comments

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

vfio/pci: Static Resizable BAR capability

The PCI Resizable BAR (ReBAR) capability is currently hidden from the
VM because the protocol for interacting with the capability does not
support a mechanism for the device to reject an advertised supported
BAR size.  However, when assigned to a VM, the act of resizing the
BAR requires adjustment of host resources for the device, which
absolutely can fail.  Linux does not currently allow us to reserve
resources for the device independent of the current usage.

The only writable field within the ReBAR capability is the BAR Size
register.  The PCIe spec indicates that when written, the device
should immediately begin to operate with the provided BAR size.  The
spec however also notes that software must only write values
corresponding to supported sizes as indicated in the capability and
control registers.  Writing unsupported sizes produces undefined
results.  Therefore, if the hypervisor were to virtualize the
capability and control registers such that the current size is the
only indicated available size, then a write of anything other than
the current size falls into the category of undefined behavior,
where we can essentially expose the modified ReBAR capability as
read-only.

This may seem pointless, but users have reported that virtualizing
the capability in this way not only allows guest software to expose
related features as available (even if only cosmetic), but in some
scenarios can resolve guest driver issues.  Additionally, no
regressions in behavior have been reported for this change.

A caveat here is that the PCIe spec requires for compatibility that
devices report support for a size in the range of 1MB to 512GB,
therefore if the current BAR size falls outside that range we revert
to hiding the capability.

Reviewed-by: Cédric Le Goater <clg@redhat.com>
Link: https://lore.kernel.org/r/20230505232308.2869912-1-alex.williamson@redhat.com
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>

vfio/migration: Skip log_sync during migration SETUP state

Currently, VFIO log_sync can be issued while migration is in SETUP
state. However, doing this log_sync is at best redundant and at worst
can fail.

Redundant -- all RAM is marked dirty in migration SETUP state and is
transferred only after migration is set to ACTIVE state, so doing
log_sync during migration SETUP is pointless.

Can fail -- there is a time window, between setting migration state to
SETUP and starting dirty tracking by RAM save_live_setup handler, during
which dirty tracking is still not started. Any VFIO log_sync call that
is issued during this time window will fail. For example, this error can
be triggered by migrating a VM when a GUI is active, which constantly
calls log_sync.

Fix it by skipping VFIO log_sync while migration is in SETUP state.

Fixes: 758b96b61d5c ("vfio/migrate: Move switch of dirty tracking into vfio_memory_listener")
Signed-off-by: Avihai Horon <avihaih@nvidia.com>
Link: https://lore.kernel.org/r/20230403130000.6422-1-avihaih@nvidia.com
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>

vfio/pci: add support for VF token

VF token was introduced [1] to kernel vfio-pci along with SR-IOV
support [2].  This patch adds support VF token among PF and VF(s). To
passthu PCIe VF to a VM, kernel >= v5.7 needs this.

It can be configured with UUID like:

  -device vfio-pci,host=DDDD:BB:DD:F,vf-token=<uuid>,...

[1] https://lore.kernel.org/linux-pci/158396393244.5601.10297430724964025753.stgit@gimli.home/
[2] https://lore.kernel.org/linux-pci/158396044753.5601.14804870681174789709.stgit@gimli.home/

Cc: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Minwoo Im <minwoo.im@samsung.com>
Reviewed-by: Klaus Jensen <k.jensen@samsung.com>
Link: https://lore.kernel.org/r/20230320073522epcms2p48f682ecdb73e0ae1a4850ad0712fd780@epcms2p4
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>

Merge tag 'compression-code-pull-request' of https://gitlab.com/juan.quintela/qemu into staging

Migration PULL request (20230508 edition, take 2)

Hi

This is just the compression bits of the Migration PULL request for
20230428.  Only change is that we don't run the compression tests by
default.

The problem already exist with compression code.  The test just show
that it don't work.

- Add migration tests for (old) compress migration code (lukas)
- Make compression code independent of ram.c (lukas)
- Move compression code into ram-compress.c (lukas)

Please apply, Juan.

# -----BEGIN PGP SIGNATURE-----
#
# iQIzBAABCAAdFiEEGJn/jt6/WMzuA0uC9IfvGFhy1yMFAmRZRMwACgkQ9IfvGFhy
# 1yOdixAA1fOLanaYMUJZGLZ9sVTt7rDc4AEPRGkQOYYZNGK3LHaG2Dx9ob2/CEkS
# /YPp9Oth9QAYHZgiI2Xx8GSg98PRVr9b/GlQPseoCOFXnUL89rTpQtxQq4CV41E6
# AA5Dr8Z07hsr47ERQERFfDGD4zsvpn+NWM1ZBy+CCilf/o8UU4eIyfRF34YgSScv
# FVdWM4czUKei9fe2Go1KnMCz1GnT/6epl47Hs8zn9WAEeUfLILp7dbkbNq26F65G
# 8YC8YnrikxU+2j+NIyIbRxbIdjR+JUbR14AyezwWZ2zGbirwWN1DP2WQx0QIZOqM
# ZuCqIDj5HpNSlHmShI0gNDfPvs+iM+sFSwQ7JE8Q03hlES9HF5c+MOr3Pl3J91hH
# EEmkk5gBJ2v2tvBuHgwVAQ2UH1+XT+a7RXeoMU1iizc2sXRGDK12ZsyaAg4D0oaF
# eohzJk2j1QXcx/DNK2G5uhzwgKvKv1/+rHyYQFtg+XuWVVipSNwqRjDJkDANAYZP
# VwKOOqDd5lHLOIzE1j61Yu06DJhkSoMvz74RQlqnk+r1EKJcTUZL52uhQor//DaL
# ULpBsgYzoMUMrtw7myHxq4t0t6mmOtOkb0CvO8dTzkIV0YgIFTtPFB0ySXOFUFf5
# UoFoMFKlfbPpDsvTNEVErxpaG4FBwZNVt67V2KXQ53xRPShyBiQ=
# =SG8L
# -----END PGP SIGNATURE-----
# gpg: Signature made Mon 08 May 2023 07:51:56 PM BST
# gpg:                using RSA key 1899FF8EDEBF58CCEE034B82F487EF185872D723
# gpg: Good signature from "Juan Quintela <quintela@redhat.com>" [undefined]
# gpg:                 aka "Juan Quintela <quintela@trasno.org>" [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: 1899 FF8E DEBF 58CC EE03  4B82 F487 EF18 5872 D723

* tag 'compression-code-pull-request' of https://gitlab.com/juan.quintela/qemu:
  migration: Initialize and cleanup decompression in migration.c
  ram-compress.c: Make target independent
  ram compress: Assert that the file buffer matches the result
  ram.c: Move core decompression code into its own file
  ram.c: Move core compression code into its own file
  ram.c: Remove last ram.c dependency from the core compress code
  ram.c: Call update_compress_thread_counts from compress_send_queued_data
  ram.c: Do not call save_page_header() from compress threads
  ram.c: Reset result after sending queued data
  ram.c: Dont change param->block in the compress thread
  ram.c: Let the compress threads return a CompressResult enum
  qtest/migration-test.c: Add postcopy tests with compress enabled
  qtest/migration-test.c: Add tests with compress enabled

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

meson: leave unnecessary modules out of the build

meson.build files choose whether to build modules based on foo.found()
expressions.  If a feature is enabled (e.g. --enable-gtk), these expressions
are true even if the code is not used by any emulator, and this results
in an unexpected difference between modular and non-modular builds.

For non-modular builds, the files are not included in any binary, and
therefore the source files are never processed.  For modular builds,
however, all .so files are unconditionally built by default, and therefore
a normal "make" tries to build them.  However, the corresponding trace-*.h
files are absent due to this conditional:

if have_system
  trace_events_subdirs += [
    ...
    'ui',
    ...
  ]
endif

which was added to avoid wasting time running tracetool on unused trace-events
files.  This causes a compilation failure; fix it by skipping module builds
entirely if (depending on the module directory) have_block or have_system
are false.

Reported-by: Michael Tokarev <mjt@tls.msk.ru>
Cc: qemu-stable@nongnu.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

docs: clarify --without-default-devices

--without-default-devices is a specialized option that should only be used
when configs/devices/ is changed manually.

Explain the model towards which we should tend, with respect to failures
to start guests and to run "make check".

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

target/i386: Add EPYC-Genoa model to support Zen 4 processor series

Adds the support for AMD EPYC Genoa generation processors. The model
display for the new processor will be EPYC-Genoa.

Adds the following new feature bits on top of the feature bits from
the previous generation EPYC models.

avx512f         : AVX-512 Foundation instruction
avx512dq        : AVX-512 Doubleword & Quadword Instruction
avx512ifma      : AVX-512 Integer Fused Multiply Add instruction
avx512cd        : AVX-512 Conflict Detection instruction
avx512bw        : AVX-512 Byte and Word Instructions
avx512vl        : AVX-512 Vector Length Extension Instructions
avx512vbmi      : AVX-512 Vector Byte Manipulation Instruction
avx512_vbmi2    : AVX-512 Additional Vector Byte Manipulation Instruction
gfni            : AVX-512 Galois Field New Instructions
avx512_vnni     : AVX-512 Vector Neural Network Instructions
avx512_bitalg   : AVX-512 Bit Algorithms, add bit algorithms Instructions
avx512_vpopcntdq: AVX-512 AVX-512 Vector Population Count Doubleword and
                  Quadword Instructions
avx512_bf16 : AVX-512 BFLOAT16 instructions
la57            : 57-bit virtual address support (5-level Page Tables)
vnmi            : Virtual NMI (VNMI) allows the hypervisor to inject the NMI
                  into the guest without using Event Injection mechanism
                  meaning not required to track the guest NMI and intercepting
                  the IRET.
auto-ibrs       : The AMD Zen4 core supports a new feature called Automatic IBRS.
                  It is a "set-and-forget" feature that means that, unlike e.g.,
                  s/w-toggled SPEC_CTRL.IBRS, h/w manages its IBRS mitigation
                  resources automatically across CPL transitions.

Signed-off-by: Babu Moger <babu.moger@amd.com>
Message-Id: <20230504205313.225073-8-babu.moger@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: Add VNMI and automatic IBRS feature bits

Add the following featute bits.

vnmi: Virtual NMI (VNMI) allows the hypervisor to inject the NMI into the
      guest without using Event Injection mechanism meaning not required to
      track the guest NMI and intercepting the IRET.
      The presence of this feature is indicated via the CPUID function
      0x8000000A_EDX[25].

automatic-ibrs :
      The AMD Zen4 core supports a new feature called Automatic IBRS.
      It is a "set-and-forget" feature that means that, unlike e.g.,
      s/w-toggled SPEC_CTRL.IBRS, h/w manages its IBRS mitigation
      resources automatically across CPL transitions.
      The presence of this feature is indicated via the CPUID function
      0x80000021_EAX[8].

The documention for the features are available in the links below.
a. Processor Programming Reference (PPR) for AMD Family 19h Model 01h,
   Revision B1 Processors
b. AMD64 Architecture Programmer’s Manual Volumes 1–5 Publication No. Revision
   40332 4.05 Date October 2022

Signed-off-by: Santosh Shukla <santosh.shukla@amd.com>
Signed-off-by: Kim Phillips <kim.phillips@amd.com>
Signed-off-by: Babu Moger <babu.moger@amd.com>
Link: https://www.amd.com/system/files/TechDocs/55898_B1_pub_0.50.zip
Link: https://www.amd.com/system/files/TechDocs/40332_4.05.pdf
Message-Id: <20230504205313.225073-7-babu.moger@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: Add missing feature bits in EPYC-Milan model

Add the following feature bits for EPYC-Milan model and bump the version.
vaes            : Vector VAES(ENC|DEC), VAES(ENC|DEC)LAST instruction support
vpclmulqdq : Vector VPCLMULQDQ instruction support
stibp-always-on : Single Thread Indirect Branch Prediction Mode has enhanced
                  performance and may be left Always on
amd-psfd : Predictive Store Forward Disable
no-nested-data-bp         : Processor ignores nested data breakpoints
lfence-always-serializing : LFENCE instruction is always serializing
null-sel-clr-base         : Null Selector Clears Base. When this bit is
                            set, a null segment load clears the segment base

These new features will be added in EPYC-Milan-v2. The "-cpu help" output
after the change will be.

    x86 EPYC-Milan             (alias configured by machine type)
    x86 EPYC-Milan-v1          AMD EPYC-Milan Processor
    x86 EPYC-Milan-v2          AMD EPYC-Milan Processor

The documentation for the features are available in the links below.
a. Processor Programming Reference (PPR) for AMD Family 19h Model 01h,
   Revision B1 Processors
b. SECURITY ANALYSIS OF AMD PREDICTIVE STORE FORWARDING
c. AMD64 Architecture Programmer’s Manual Volumes 1–5 Publication No. Revision
    40332 4.05 Date October 2022

Signed-off-by: Babu Moger <babu.moger@amd.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Link: https://www.amd.com/system/files/TechDocs/55898_B1_pub_0.50.zip
Link: https://www.amd.com/system/files/documents/security-analysis-predictive-store-forwarding.pdf
Link: https://www.amd.com/system/files/TechDocs/40332_4.05.pdf
Message-Id: <20230504205313.225073-6-babu.moger@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: Add feature bits for CPUID_Fn80000021_EAX

Add the following feature bits.
no-nested-data-bp   : Processor ignores nested data breakpoints.
lfence-always-serializing : LFENCE instruction is always serializing.
null-sel-cls-base   : Null Selector Clears Base. When this bit is
    set, a null segment load clears the segment base.

The documentation for the features are available in the links below.
a. Processor Programming Reference (PPR) for AMD Family 19h Model 01h,
   Revision B1 Processors
b. AMD64 Architecture Programmer’s Manual Volumes 1–5 Publication No. Revision
    40332 4.05 Date October 2022

Signed-off-by: Babu Moger <babu.moger@amd.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Link: https://www.amd.com/system/files/TechDocs/55898_B1_pub_0.50.zip
Link: https://www.amd.com/system/files/TechDocs/40332_4.05.pdf
Message-Id: <20230504205313.225073-5-babu.moger@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: Add a couple of feature bits in 8000_0008_EBX

Add the following feature bits.

amd-psfd : Predictive Store Forwarding Disable:
           PSF is a hardware-based micro-architectural optimization
           designed to improve the performance of code execution by
           predicting address dependencies between loads and stores.
           While SSBD (Speculative Store Bypass Disable) disables both
           PSF and speculative store bypass, PSFD only disables PSF.
           PSFD may be desirable for the software which is concerned
           with the speculative behavior of PSF but desires a smaller
           performance impact than setting SSBD.
   Depends on the following kernel commit:
           b73a54321ad8 ("KVM: x86: Expose Predictive Store Forwarding Disable")

stibp-always-on :
           Single Thread Indirect Branch Prediction mode has enhanced
           performance and may be left always on.

The documentation for the features are available in the links below.
a. Processor Programming Reference (PPR) for AMD Family 19h Model 01h,
   Revision B1 Processors
b. SECURITY ANALYSIS OF AMD PREDICTIVE STORE FORWARDING

Signed-off-by: Babu Moger <babu.moger@amd.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Link: https://www.amd.com/system/files/documents/security-analysis-predictive-store-forwarding.pdf
Link: https://www.amd.com/system/files/TechDocs/55898_B1_pub_0.50.zip
Message-Id: <20230504205313.225073-4-babu.moger@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: Add new EPYC CPU versions with updated cache_info

Introduce new EPYC cpu versions: EPYC-v4 and EPYC-Rome-v3.
The only difference vs. older models is an updated cache_info with
the 'complex_indexing' bit unset, since this bit is not currently
defined for AMD and may cause problems should it be used for
something else in the future. Setting this bit will also cause
CPUID validation failures when running SEV-SNP guests.

Signed-off-by: Michael Roth <michael.roth@amd.com>
Signed-off-by: Babu Moger <babu.moger@amd.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Message-Id: <20230504205313.225073-3-babu.moger@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: allow versioned CPUs to specify new cache_info

New EPYC CPUs versions require small changes to their cache_info's.
Because current QEMU x86 CPU definition does not support versioned
cach_info, we would have to declare a new CPU type for each such case.
To avoid the dup work, add "cache_info" in X86CPUVersionDefinition",
to allow new cache_info pointers to be specified for a new CPU version.

Co-developed-by: Wei Huang <wei.huang2@amd.com>
Signed-off-by: Wei Huang <wei.huang2@amd.com>
Signed-off-by: Michael Roth <michael.roth@amd.com>
Signed-off-by: Babu Moger <babu.moger@amd.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Message-Id: <20230504205313.225073-2-babu.moger@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

include/qemu/osdep.h: Bump _WIN32_WINNT to the Windows 8 API

Commit cf60ccc330 ("cutils: Introduce bundle mechanism") abandoned
compatibility with Windows older than 8 - we should reflect this
in our _WIN32_WINNT and set it to the value that corresponds to
Windows 8.

Signed-off-by: Thomas Huth <thuth@redhat.com>
Message-Id: <20230504081351.125140-1-thuth@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

migration: Initialize and cleanup decompression in migration.c

This fixes compress with colo.

Signed-off-by: Lukas Straub <lukasstraub2@web.de>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>

ram-compress.c: Make target independent

Make ram-compress.c target independent.

Signed-off-by: Lukas Straub <lukasstraub2@web.de>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>

ram compress: Assert that the file buffer matches the result

Before this series, "nothing to send" was handled by the file buffer
being empty. Now it is tracked via param->result.

Assert that the file buffer state matches the result.

Signed-off-by: Lukas Straub <lukasstraub2@web.de>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>

ram.c: Move core decompression code into its own file

No functional changes intended.

Signed-off-by: Lukas Straub <lukasstraub2@web.de>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>

ram.c: Move core compression code into its own file

No functional changes intended.

Signed-off-by: Lukas Straub <lukasstraub2@web.de>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>

ram.c: Remove last ram.c dependency from the core compress code

Make compression interfaces take send_queued_data() as an argument.
Remove save_page_use_compression() from flush_compressed_data().

This removes the last ram.c dependency from the core compress code.

Signed-off-by: Lukas Straub <lukasstraub2@web.de>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>

ram.c: Call update_compress_thread_counts from compress_send_queued_data

This makes the core compress code more independend from ram.c.

Signed-off-by: Lukas Straub <lukasstraub2@web.de>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>

ram.c: Do not call save_page_header() from compress threads

save_page_header() accesses several global variables, so calling it
from multiple threads is pretty ugly.

Instead, call save_page_header() before writing out the compressed
data from the compress buffer to the migration stream.

This also makes the core compress code more independend from ram.c.

Signed-off-by: Lukas Straub <lukasstraub2@web.de>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>

ram.c: Reset result after sending queued data

And take the param->mutex lock for the whole section to ensure
thread-safety.
Now, it is explicitly clear if there is no queued data to send.
Before, this was handled by param->file stream being empty and thus
qemu_put_qemu_file() not sending anything.

This will be used in the next commits to move save_page_header()
out of compress code.

Signed-off-by: Lukas Straub <lukasstraub2@web.de>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>

ram.c: Dont change param->block in the compress thread

Instead introduce a extra parameter to trigger the compress thread.
Now, when the compress thread is done, we know what RAMBlock and
offset it did compress.

This will be used in the next commits to move save_page_header()
out of compress code.

Signed-off-by: Lukas Straub <lukasstraub2@web.de>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>

ram.c: Let the compress threads return a CompressResult enum

This will be used in the next commits to move save_page_header()
out of compress code.

Signed-off-by: Lukas Straub <lukasstraub2@web.de>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>

qtest/migration-test.c: Add postcopy tests with compress enabled

Add postcopy tests with compress enabled to ensure nothing breaks
with the refactoring in the next commits.

preempt+compress is blocked, so no test needed for that case.

Signed-off-by: Lukas Straub <lukasstraub2@web.de>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>

qtest/migration-test.c: Add tests with compress enabled

There has never been tests for migration with compress enabled.

Add suitable tests, testing with compress-wait-thread = false
too.

Signed-off-by: Lukas Straub <lukasstraub2@web.de>
Acked-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>

MAINTAINERS: add stanza for Kconfig files

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

tb-maint: do not use mb_read/mb_set

The load side can use a relaxed load, which will surely happen before
the work item is run by async_safe_run_on_cpu() or before double-checking
under mmap_lock.  The store side can use an atomic RMW operation.

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

call_rcu: stop using mb_set/mb_read

Use a store-release when enqueuing a new call_rcu, and a load-acquire
when dequeuing; and read the tail after checking that node->next is
consistent, which is the standard message passing pattern and it is
clearer than mb_read/mb_set.

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

test-aio-multithread: simplify test_multi_co_schedule

Instead of using qatomic_mb_{read,set} mindlessly, just use a per-coroutine
flag that requires no synchronization.

Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

test-aio-multithread: do not use mb_read/mb_set for simple flags

The remaining use of mb_read/mb_set is just to force a thread to exit
eventually.  It does not order two memory accesses and therefore can be
just read/set.

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

rcu: remove qatomic_mb_set, expand comments

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

Merge tag 'pull-loongarch-20230506' of https://gitlab.com/gaosong/qemu into staging

Add LoongArch LSX instructions.

# -----BEGIN PGP SIGNATURE-----
#
# iLMEAAEIAB0WIQS4/x2g0v3LLaCcbCxAov/yOSY+3wUCZFXxGwAKCRBAov/yOSY+
# 39EoA/0Uy2DPz6g7J5+9tcIRk9jLrp36aYQJ9J8zRJd226YFvHSfiBWSIteMFOEX
# Z0Jx1bL6N97KK/HA74Nx++x0kVuplEGp1s5cO/odL3gYy8RaJm23p9iaDa0D/UaB
# ygLvXtuzN4unDFP5EF/wa9zRkDb7qX2iBBvc8OIal7eT4dDX+g==
# =gyVU
# -----END PGP SIGNATURE-----
# gpg: Signature made Sat 06 May 2023 07:18:03 AM BST
# gpg:                using RSA key B8FF1DA0D2FDCB2DA09C6C2C40A2FFF239263EDF
# gpg: Good signature from "Song Gao <m17746591750@163.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: B8FF 1DA0 D2FD CB2D A09C  6C2C 40A2 FFF2 3926 3EDF

* tag 'pull-loongarch-20230506' of https://gitlab.com/gaosong/qemu: (45 commits)
  hw/intc: don't use target_ulong for LoongArch ipi
  target/loongarch: CPUCFG support LSX
  target/loongarch: Use {set/get}_gpr replace to cpu_fpr
  target/loongarch: Implement vldi
  target/loongarch: Implement vld vst
  target/loongarch: Implement vilvl vilvh vextrins vshuf
  target/loongarch: Implement vreplve vpack vpick
  target/loongarch: Implement vinsgr2vr vpickve2gr vreplgr2vr
  target/loongarch: Implement vbitsel vset
  target/loongarch: Implement vfcmp
  target/loongarch: Implement vseq vsle vslt
  target/loongarch: Implement LSX fpu fcvt instructions
  target/loongarch: Implement LSX fpu arith instructions
  target/loongarch: Implement vfrstp
  target/loongarch: Implement vbitclr vbitset vbitrev
  target/loongarch: Implement vpcnt
  target/loongarch: Implement vclo vclz
  target/loongarch: Implement vssrlrn vssrarn
  target/loongarch: Implement vssrln vssran
  target/loongarch: Implement vsrlrn vsrarn
  ...

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

hw/intc: don't use target_ulong for LoongArch ipi

The calling function is already working with hwaddr and uint64_t so
lets avoid bringing target_ulong in if we don't need to.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20230404132711.2563638-1-alex.bennee@linaro.org>
Signed-off-by: Song Gao <gaosong@loongson.cn>

target/loongarch: CPUCFG support LSX

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20230504122810.4094787-45-gaosong@loongson.cn>

target/loongarch: Use {set/get}_gpr replace to cpu_fpr

Introduce set_fpr() and get_fpr() and remove cpu_fpr.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20230504122810.4094787-44-gaosong@loongson.cn>

target/loongarch: Implement vldi

This patch includes:
- VLDI.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20230504122810.4094787-43-gaosong@loongson.cn>

target/loongarch: Implement vld vst

This patch includes:
- VLD[X], VST[X];
- VLDREPL.{B/H/W/D};
- VSTELM.{B/H/W/D}.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20230504122810.4094787-42-gaosong@loongson.cn>

target/loongarch: Implement vilvl vilvh vextrins vshuf

This patch includes:
- VILV{L/H}.{B/H/W/D};
- VSHUF.{B/H/W/D};
- VSHUF4I.{B/H/W/D};
- VPERMI.W;
- VEXTRINS.{B/H/W/D}.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20230504122810.4094787-41-gaosong@loongson.cn>

target/loongarch: Implement vreplve vpack vpick

This patch includes:
- VREPLVE[I].{B/H/W/D};
- VBSLL.V, VBSRL.V;
- VPACK{EV/OD}.{B/H/W/D};
- VPICK{EV/OD}.{B/H/W/D}.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20230504122810.4094787-40-gaosong@loongson.cn>

target/loongarch: Implement vinsgr2vr vpickve2gr vreplgr2vr

This patch includes:
- VINSGR2VR.{B/H/W/D};
- VPICKVE2GR.{B/H/W/D}[U];
- VREPLGR2VR.{B/H/W/D}.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20230504122810.4094787-39-gaosong@loongson.cn>

target/loongarch: Implement vbitsel vset

This patch includes:
- VBITSEL.V;
- VBITSELI.B;
- VSET{EQZ/NEZ}.V;
- VSETANYEQZ.{B/H/W/D};
- VSETALLNEZ.{B/H/W/D}.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20230504122810.4094787-38-gaosong@loongson.cn>

target/loongarch: Implement vfcmp

This patch includes:
- VFCMP.cond.{S/D}.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20230504122810.4094787-37-gaosong@loongson.cn>

target/loongarch: Implement vseq vsle vslt

This patch includes:
- VSEQ[I].{B/H/W/D};
- VSLE[I].{B/H/W/D}[U];
- VSLT[I].{B/H/W/D/}[U].

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20230504122810.4094787-36-gaosong@loongson.cn>

target/loongarch: Implement LSX fpu fcvt instructions

This patch includes:
- VFCVT{L/H}.{S.H/D.S};
- VFCVT.{H.S/S.D};
- VFRINT[{RNE/RZ/RP/RM}].{S/D};
- VFTINT[{RNE/RZ/RP/RM}].{W.S/L.D};
- VFTINT[RZ].{WU.S/LU.D};
- VFTINT[{RNE/RZ/RP/RM}].W.D;
- VFTINT[{RNE/RZ/RP/RM}]{L/H}.L.S;
- VFFINT.{S.W/D.L}[U];
- VFFINT.S.L, VFFINT{L/H}.D.W.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20230504122810.4094787-35-gaosong@loongson.cn>

target/loongarch: Implement LSX fpu arith instructions

This patch includes:
- VF{ADD/SUB/MUL/DIV}.{S/D};
- VF{MADD/MSUB/NMADD/NMSUB}.{S/D};
- VF{MAX/MIN}.{S/D};
- VF{MAXA/MINA}.{S/D};
- VFLOGB.{S/D};
- VFCLASS.{S/D};
- VF{SQRT/RECIP/RSQRT}.{S/D}.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20230504122810.4094787-34-gaosong@loongson.cn>

target/loongarch: Implement vfrstp

This patch includes:
- VFRSTP[I].{B/H}.

Acked-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20230504122810.4094787-33-gaosong@loongson.cn>

target/loongarch: Implement vbitclr vbitset vbitrev

This patch includes:
- VBITCLR[I].{B/H/W/D};
- VBITSET[I].{B/H/W/D};
- VBITREV[I].{B/H/W/D}.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20230504122810.4094787-32-gaosong@loongson.cn>

target/loongarch: Implement vpcnt

This patch includes:
- VPCNT.{B/H/W/D}.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20230504122810.4094787-31-gaosong@loongson.cn>

target/loongarch: Implement vclo vclz

This patch includes:
- VCLO.{B/H/W/D};
- VCLZ.{B/H/W/D}.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20230504122810.4094787-30-gaosong@loongson.cn>

target/loongarch: Implement vssrlrn vssrarn

This patch includes:
- VSSRLRN.{B.H/H.W/W.D};
- VSSRARN.{B.H/H.W/W.D};
- VSSRLRN.{BU.H/HU.W/WU.D};
- VSSRARN.{BU.H/HU.W/WU.D};
- VSSRLRNI.{B.H/H.W/W.D/D.Q};
- VSSRARNI.{B.H/H.W/W.D/D.Q};
- VSSRLRNI.{BU.H/HU.W/WU.D/DU.Q};
- VSSRARNI.{BU.H/HU.W/WU.D/DU.Q}.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20230504122810.4094787-29-gaosong@loongson.cn>

target/loongarch: Implement vssrln vssran

This patch includes:
- VSSRLN.{B.H/H.W/W.D};
- VSSRAN.{B.H/H.W/W.D};
- VSSRLN.{BU.H/HU.W/WU.D};
- VSSRAN.{BU.H/HU.W/WU.D};
- VSSRLNI.{B.H/H.W/W.D/D.Q};
- VSSRANI.{B.H/H.W/W.D/D.Q};
- VSSRLNI.{BU.H/HU.W/WU.D/DU.Q};
- VSSRANI.{BU.H/HU.W/WU.D/DU.Q}.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20230504122810.4094787-28-gaosong@loongson.cn>

target/loongarch: Implement vsrlrn vsrarn

This patch includes:
- VSRLRN.{B.H/H.W/W.D};
- VSRARN.{B.H/H.W/W.D};
- VSRLRNI.{B.H/H.W/W.D/D.Q};
- VSRARNI.{B.H/H.W/W.D/D.Q}.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20230504122810.4094787-27-gaosong@loongson.cn>

target/loongarch: Implement vsrln vsran

This patch includes:
- VSRLN.{B.H/H.W/W.D};
- VSRAN.{B.H/H.W/W.D};
- VSRLNI.{B.H/H.W/W.D/D.Q};
- VSRANI.{B.H/H.W/W.D/D.Q}.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20230504122810.4094787-26-gaosong@loongson.cn>

target/loongarch: Implement vsrlr vsrar

This patch includes:
- VSRLR[I].{B/H/W/D};
- VSRAR[I].{B/H/W/D}.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20230504122810.4094787-25-gaosong@loongson.cn>

target/loongarch: Implement vsllwil vextl

This patch includes:
- VSLLWIL.{H.B/W.H/D.W};
- VSLLWIL.{HU.BU/WU.HU/DU.WU};
- VEXTL.Q.D, VEXTL.QU.DU.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20230504122810.4094787-24-gaosong@loongson.cn>

target/loongarch: Implement vsll vsrl vsra vrotr

This patch includes:
- VSLL[I].{B/H/W/D};
- VSRL[I].{B/H/W/D};
- VSRA[I].{B/H/W/D};
- VROTR[I].{B/H/W/D}.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20230504122810.4094787-23-gaosong@loongson.cn>

target/loongarch: Implement LSX logic instructions

This patch includes:
- V{AND/OR/XOR/NOR/ANDN/ORN}.V;
- V{AND/OR/XOR/NOR}I.B.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20230504122810.4094787-22-gaosong@loongson.cn>

target/loongarch: Implement vmskltz/vmskgez/vmsknz

This patch includes:
- VMSKLTZ.{B/H/W/D};
- VMSKGEZ.B;
- VMSKNZ.B.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20230504122810.4094787-21-gaosong@loongson.cn>

target/loongarch: Implement vsigncov

This patch includes:
- VSIGNCOV.{B/H/W/D}.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20230504122810.4094787-20-gaosong@loongson.cn>

target/loongarch: Implement vexth

This patch includes:
- VEXTH.{H.B/W.H/D.W/Q.D};
- VEXTH.{HU.BU/WU.HU/DU.WU/QU.DU}.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20230504122810.4094787-19-gaosong@loongson.cn>

target/loongarch: Implement vsat

This patch includes:
- VSAT.{B/H/W/D}[U].

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20230504122810.4094787-18-gaosong@loongson.cn>

target/loongarch: Implement vdiv/vmod

This patch includes:
- VDIV.{B/H/W/D}[U];
- VMOD.{B/H/W/D}[U].

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20230504122810.4094787-17-gaosong@loongson.cn>

target/loongarch: Implement vmadd/vmsub/vmaddw{ev/od}

This patch includes:
- VMADD.{B/H/W/D};
- VMSUB.{B/H/W/D};
- VMADDW{EV/OD}.{H.B/W.H/D.W/Q.D}[U];
- VMADDW{EV/OD}.{H.BU.B/W.HU.H/D.WU.W/Q.DU.D}.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20230504122810.4094787-16-gaosong@loongson.cn>

target/loongarch: Implement vmul/vmuh/vmulw{ev/od}

This patch includes:
- VMUL.{B/H/W/D};
- VMUH.{B/H/W/D}[U];
- VMULW{EV/OD}.{H.B/W.H/D.W/Q.D}[U];
- VMULW{EV/OD}.{H.BU.B/W.HU.H/D.WU.W/Q.DU.D}.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20230504122810.4094787-15-gaosong@loongson.cn>

target/loongarch: Implement vmax/vmin

This patch includes:
- VMAX[I].{B/H/W/D}[U];
- VMIN[I].{B/H/W/D}[U].

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20230504122810.4094787-14-gaosong@loongson.cn>

target/loongarch: Implement vadda

This patch includes:
- VADDA.{B/H/W/D}.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20230504122810.4094787-13-gaosong@loongson.cn>

target/loongarch: Implement vabsd

This patch includes:
- VABSD.{B/H/W/D}[U].

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20230504122810.4094787-12-gaosong@loongson.cn>

target/loongarch: Implement vavg/vavgr

This patch includes:
- VAVG.{B/H/W/D}[U];
- VAVGR.{B/H/W/D}[U].

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20230504122810.4094787-11-gaosong@loongson.cn>

target/loongarch: Implement vaddw/vsubw

This patch includes:
- VADDW{EV/OD}.{H.B/W.H/D.W/Q.D}[U];
- VSUBW{EV/OD}.{H.B/W.H/D.W/Q.D}[U];
- VADDW{EV/OD}.{H.BU.B/W.HU.H/D.WU.W/Q.DU.D}.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20230504122810.4094787-10-gaosong@loongson.cn>

target/loongarch: Implement vhaddw/vhsubw

This patch includes:
- VHADDW.{H.B/W.H/D.W/Q.D/HU.BU/WU.HU/DU.WU/QU.DU};
- VHSUBW.{H.B/W.H/D.W/Q.D/HU.BU/WU.HU/DU.WU/QU.DU}.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20230504122810.4094787-9-gaosong@loongson.cn>

target/loongarch: Implement vsadd/vssub

This patch includes:
- VSADD.{B/H/W/D}[U];
- VSSUB.{B/H/W/D}[U].

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20230504122810.4094787-8-gaosong@loongson.cn>

target/loongarch: Implement vneg

This patch includes;
- VNEG.{B/H/W/D}.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20230504122810.4094787-7-gaosong@loongson.cn>

target/loongarch: Implement vaddi/vsubi

This patch includes:
- VADDI.{B/H/W/D}U;
- VSUBI.{B/H/W/D}U.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20230504122810.4094787-6-gaosong@loongson.cn>

target/loongarch: Implement vadd/vsub

This patch includes:
- VADD.{B/H/W/D/Q};
- VSUB.{B/H/W/D/Q}.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20230504122810.4094787-5-gaosong@loongson.cn>

target/loongarch: Add CHECK_SXE maccro for check LSX enable

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20230504122810.4094787-4-gaosong@loongson.cn>

target/loongarch: meson.build support build LSX

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20230504122810.4094787-3-gaosong@loongson.cn>

target/loongarch: Add LSX data type VReg

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20230504122810.4094787-2-gaosong@loongson.cn>