OSDN Git Service

qmiga/qemu.git
2 years agopython/aqmp-tui: Add syntax highlighting
G S Niteesh Babu [Mon, 23 Aug 2021 22:07:46 +0000 (03:37 +0530)]
python/aqmp-tui: Add syntax highlighting

Add syntax highlighting for the incoming and outgoing QMP messages.
This is achieved using the pygments module which was added in a
previous commit.

The current implementation is a really simple one which doesn't
allow for any configuration. In future this has to be improved
to allow for easier theme config using an external config of
some sort.

Signed-off-by: G S Niteesh Babu <niteesh.gs@gmail.com>
Message-Id: <20210823220746.28295-6-niteesh.gs@gmail.com>
Signed-off-by: John Snow <jsnow@redhat.com>
2 years agopython: add optional pygments dependency
G S Niteesh Babu [Mon, 23 Aug 2021 22:07:45 +0000 (03:37 +0530)]
python: add optional pygments dependency

Added pygments as optional dependency for AQMP TUI.
This is required for the upcoming syntax highlighting feature
in AQMP TUI.
The dependency has also been added in the devel optional group.

Added mypy 'ignore_missing_imports' for pygments since it does
not have any type stubs.

Signed-off-by: G S Niteesh Babu <niteesh.gs@gmail.com>
Message-Id: <20210823220746.28295-5-niteesh.gs@gmail.com>
Signed-off-by: John Snow <jsnow@redhat.com>
2 years agopython: Add entry point for aqmp-tui
G S Niteesh Babu [Mon, 23 Aug 2021 22:07:44 +0000 (03:37 +0530)]
python: Add entry point for aqmp-tui

Add an entry point for aqmp-tui. This will allow it to be run from
the command line using "aqmp-tui localhost:1234"
More options available in the TUI can be found using "aqmp-tui -h"

Signed-off-by: G S Niteesh Babu <niteesh.gs@gmail.com>
Message-Id: <20210823220746.28295-4-niteesh.gs@gmail.com>
Signed-off-by: John Snow <jsnow@redhat.com>
2 years agopython/aqmp-tui: Add AQMP TUI
G S Niteesh Babu [Mon, 23 Aug 2021 22:07:43 +0000 (03:37 +0530)]
python/aqmp-tui: Add AQMP TUI

Added AQMP TUI.

Implements the follwing basic features:
1) Command transmission/reception.
2) Shows events asynchronously.
3) Shows server status in the bottom status bar.
4) Automatic retries on disconnects and error conditions.

Also added type annotations and necessary pylint/mypy configurations.

Signed-off-by: G S Niteesh Babu <niteesh.gs@gmail.com>
Message-Id: <20210823220746.28295-3-niteesh.gs@gmail.com>
Signed-off-by: John Snow <jsnow@redhat.com>
2 years agopython: Add dependencies for AQMP TUI
G S Niteesh Babu [Mon, 23 Aug 2021 22:07:42 +0000 (03:37 +0530)]
python: Add dependencies for AQMP TUI

Added dependencies for the upcoming AQMP TUI under the optional
'tui' group.

The same dependencies have also been added under the devel group
since no work around has been found for optional groups to imply
other optional groups.

Signed-off-by: G S Niteesh Babu <niteesh.gs@gmail.com>
Message-Id: <20210823220746.28295-2-niteesh.gs@gmail.com>
Signed-off-by: John Snow <jsnow@redhat.com>
2 years agopython/aqmp: Add Coverage.py support
John Snow [Wed, 15 Sep 2021 16:29:55 +0000 (12:29 -0400)]
python/aqmp: Add Coverage.py support

I'm not exposing this via the Makefile help, it's not likely to be
useful to passersby. Switch the avocado runner to the 'legacy' runner
for now, as the new runner seems to obscure coverage reports, again.

Usage is to enter your venv of choice and then:
`make check-coverage && xdg-open htmlcov/index.html`.

Signed-off-by: John Snow <jsnow@redhat.com>
Message-id: 20210915162955.333025-28-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
2 years agopython/aqmp: add LineProtocol tests
John Snow [Wed, 15 Sep 2021 16:29:54 +0000 (12:29 -0400)]
python/aqmp: add LineProtocol tests

Tests a real connect, a real accept, and really sending and receiving a
message over a UNIX socket.

Brings coverage of protocol.py up to ~93%.

Signed-off-by: John Snow <jsnow@redhat.com>
Message-id: 20210915162955.333025-27-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
2 years agopython/aqmp: add AsyncProtocol unit tests
John Snow [Wed, 15 Sep 2021 16:29:53 +0000 (12:29 -0400)]
python/aqmp: add AsyncProtocol unit tests

This tests most of protocol.py -- From a hacked up Coverage.py run, it's
at about 86%. There's a few error cases that aren't very well tested
yet, they're hard to induce artificially so far. I'm working on it.

Signed-off-by: John Snow <jsnow@redhat.com>
Message-id: 20210915162955.333025-26-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
2 years agopython: bump avocado to v90.0
John Snow [Wed, 15 Sep 2021 16:29:52 +0000 (12:29 -0400)]
python: bump avocado to v90.0

Avocado v90 includes improved support for running async unit tests. The
workaround that existed prior to v90 causes the unit tests to fail
afterwards, however, so upgrade our minimum version pin to the very
latest and greatest.

Signed-off-by: John Snow <jsnow@redhat.com>
Message-id: 20210915162955.333025-25-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
2 years agopython/aqmp: add scary message
John Snow [Wed, 15 Sep 2021 16:29:51 +0000 (12:29 -0400)]
python/aqmp: add scary message

Add a warning whenever AQMP is used to steer people gently away from
using it for the time-being.

Signed-off-by: John Snow <jsnow@redhat.com>
Message-id: 20210915162955.333025-24-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
2 years agopython/aqmp: add asyncio_run compatibility wrapper
John Snow [Wed, 15 Sep 2021 16:29:50 +0000 (12:29 -0400)]
python/aqmp: add asyncio_run compatibility wrapper

As a convenience. It isn't used by the library itself, but it is used by
the test suite. It will also come in handy for users of the library
still on Python 3.6.

Signed-off-by: John Snow <jsnow@redhat.com>
Message-id: 20210915162955.333025-23-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
2 years agopython/aqmp: add _raw() execution interface
John Snow [Wed, 15 Sep 2021 16:29:49 +0000 (12:29 -0400)]
python/aqmp: add _raw() execution interface

This is added in anticipation of wanting it for a synchronous wrapper
for the iotest interface. Normally, execute() and execute_msg() both
raise QMP errors in the form of Python exceptions.

Many iotests expect the entire reply as-is. To reduce churn there, add a
private execution interface that will ease transition churn. However, I
do not wish to encourage its use, so it will remain a private interface.

Signed-off-by: John Snow <jsnow@redhat.com>
Message-id: 20210915162955.333025-22-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
2 years agopython/aqmp: add execute() interfaces
John Snow [Wed, 15 Sep 2021 16:29:48 +0000 (12:29 -0400)]
python/aqmp: add execute() interfaces

Add execute() and execute_msg().

_execute() is split into _issue() and _reply() halves so that
hypothetical subclasses of QMP that want to support different execution
paradigms can do so.

I anticipate a synchronous interface may have need of separating the
send/reply phases. However, I do not wish to expose that interface here
and want to actively discourage it, so they remain private interfaces.

Signed-off-by: John Snow <jsnow@redhat.com>
Message-id: 20210915162955.333025-21-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
2 years agopython/aqmp: Add message routing to QMP protocol
John Snow [Wed, 15 Sep 2021 16:29:47 +0000 (12:29 -0400)]
python/aqmp: Add message routing to QMP protocol

Add the ability to handle and route messages in qmp_protocol.py. The
interface for actually sending anything still isn't added until next
commit.

Signed-off-by: John Snow <jsnow@redhat.com>
Message-id: 20210915162955.333025-20-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
2 years agopython/pylint: disable no-member check
John Snow [Wed, 15 Sep 2021 16:29:46 +0000 (12:29 -0400)]
python/pylint: disable no-member check

mypy handles this better -- but we only need the workaround because
pylint under Python 3.6 does not understand that a MutableMapping really
does have a .get() method attached.

We could remove this again once 3.7 is our minimum.

Signed-off-by: John Snow <jsnow@redhat.com>
Message-id: 20210915162955.333025-19-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
2 years agopython/aqmp: add QMP protocol support
John Snow [Wed, 15 Sep 2021 16:29:45 +0000 (12:29 -0400)]
python/aqmp: add QMP protocol support

The star of our show!

Add most of the QMP protocol, sans support for actually executing
commands. No problem, that happens in the next several commits.

Signed-off-by: John Snow <jsnow@redhat.com>
Message-id: 20210915162955.333025-18-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
2 years agopython/pylint: disable too-many-function-args
John Snow [Wed, 15 Sep 2021 16:29:44 +0000 (12:29 -0400)]
python/pylint: disable too-many-function-args

too-many-function-args seems prone to failure when considering
things like Method Resolution Order, which mypy gets correct. When
dealing with multiple inheritance, pylint doesn't seem to understand
which method will actually get called, while mypy does.

Remove the less powerful, redundant check.

Signed-off-by: John Snow <jsnow@redhat.com>
Message-id: 20210915162955.333025-17-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
2 years agopython/aqmp: add QMP event support
John Snow [Wed, 15 Sep 2021 16:29:43 +0000 (12:29 -0400)]
python/aqmp: add QMP event support

This class was designed as a "mix-in" primarily so that the feature
could be given its own treatment in its own python module.

It gets quite a bit too long otherwise.

Signed-off-by: John Snow <jsnow@redhat.com>
Message-id: 20210915162955.333025-16-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
2 years agopython/aqmp: add well-known QMP object models
John Snow [Wed, 15 Sep 2021 16:29:42 +0000 (12:29 -0400)]
python/aqmp: add well-known QMP object models

The QMP spec doesn't define very many objects that are iron-clad in
their format, but there are a few. This module makes it trivial to
validate them without relying on an external third-party library.

Signed-off-by: John Snow <jsnow@redhat.com>
Message-id: 20210915162955.333025-15-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
2 years agopython/aqmp: add QMP Message format
John Snow [Wed, 15 Sep 2021 16:29:41 +0000 (12:29 -0400)]
python/aqmp: add QMP Message format

The Message class is here primarily to serve as a solid type to use for
mypy static typing for unambiguous annotation and documentation.

We can also stuff JSON serialization and deserialization into this class
itself so it can be re-used even outside this infrastructure.

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 20210915162955.333025-14-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
2 years agopython/aqmp: add AsyncProtocol._readline() method
John Snow [Wed, 15 Sep 2021 16:29:40 +0000 (12:29 -0400)]
python/aqmp: add AsyncProtocol._readline() method

This is added as a courtesy: many protocols are line-based, including
QMP. Putting it in AsyncProtocol lets us keep the QMP class
implementation just a pinch more abstract.

(And, if we decide to add a QTEST implementation later, it will need
this, too. (Yes, I have a QTEST implementation.))

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 20210915162955.333025-13-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
2 years agopython/aqmp: add _cb_inbound and _cb_outbound logging hooks
John Snow [Wed, 15 Sep 2021 16:29:39 +0000 (12:29 -0400)]
python/aqmp: add _cb_inbound and _cb_outbound logging hooks

Add hooks designed to log/filter incoming/outgoing messages. The primary
intent for these is to be able to support iotests which may want to log
messages with specific filters for reproducible output.

Another use is for plugging into Urwid frameworks; all messages in/out
can be automatically added to a rendering list for the purposes of a
qmp-shell like tool.

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 20210915162955.333025-12-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
2 years agopython/aqmp: add configurable read buffer limit
John Snow [Wed, 15 Sep 2021 16:29:38 +0000 (12:29 -0400)]
python/aqmp: add configurable read buffer limit

QMP can transmit some pretty big messages, and the default limit of 64KB
isn't sufficient. Make sure that we can configure it.

Reported-by: G S Niteesh Babu <niteesh.gs@gmail.com>
Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 20210915162955.333025-11-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
2 years agopython/aqmp: add AsyncProtocol.accept() method
John Snow [Wed, 15 Sep 2021 16:29:37 +0000 (12:29 -0400)]
python/aqmp: add AsyncProtocol.accept() method

It's a little messier than connect, because it wasn't designed to accept
*precisely one* connection. Such is life.

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 20210915162955.333025-10-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
2 years agopython/aqmp: add logging to AsyncProtocol
John Snow [Wed, 15 Sep 2021 16:29:36 +0000 (12:29 -0400)]
python/aqmp: add logging to AsyncProtocol

Give the connection and the reader/writer tasks nicknames, and add
logging statements throughout.

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 20210915162955.333025-9-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
2 years agopython/aqmp: Add logging utility helpers
John Snow [Wed, 15 Sep 2021 16:29:35 +0000 (12:29 -0400)]
python/aqmp: Add logging utility helpers

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 20210915162955.333025-8-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
2 years agopython/aqmp: add runstate state machine to AsyncProtocol
John Snow [Wed, 15 Sep 2021 16:29:34 +0000 (12:29 -0400)]
python/aqmp: add runstate state machine to AsyncProtocol

This serves a few purposes:

1. Protect interfaces when it's not safe to call them (via @require)

2. Add an interface by which an async client can determine if the state
has changed, for the purposes of connection management.

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 20210915162955.333025-7-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
2 years agopython/aqmp: add generic async message-based protocol support
John Snow [Wed, 15 Sep 2021 16:29:33 +0000 (12:29 -0400)]
python/aqmp: add generic async message-based protocol support

This is the bare minimum that you need to establish a full-duplex async
message-based protocol with Python's asyncio.

The features to be added in forthcoming commits are:

- Runstate tracking
- Logging
- Support for incoming connections via accept()
- _cb_outbound, _cb_inbound message hooks
- _readline() method

Signed-off-by: John Snow <jsnow@redhat.com>
Message-id: 20210915162955.333025-6-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
2 years agopython/aqmp: add asyncio compatibility wrappers
John Snow [Wed, 15 Sep 2021 16:29:32 +0000 (12:29 -0400)]
python/aqmp: add asyncio compatibility wrappers

Python 3.6 does not have all of the goodies that Python 3.7 does, and we
need to support both. Add some compatibility wrappers needed for this
purpose.

(Note: Python 3.6 is EOL December 2021.)

Signed-off-by: John Snow <jsnow@redhat.com>
Message-id: 20210915162955.333025-5-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
2 years agopython/pylint: Add exception for TypeVar names ('T')
John Snow [Wed, 15 Sep 2021 16:29:31 +0000 (12:29 -0400)]
python/pylint: Add exception for TypeVar names ('T')

'T' is a common TypeVar name, allow its use.

See also https://github.com/PyCQA/pylint/issues/3401 -- In the future,
we might be able to have a separate list of acceptable names for
TypeVars exclusively.

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 20210915162955.333025-4-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
2 years agopython/aqmp: add error classes
John Snow [Wed, 15 Sep 2021 16:29:30 +0000 (12:29 -0400)]
python/aqmp: add error classes

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 20210915162955.333025-3-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
2 years agopython/aqmp: add asynchronous QMP (AQMP) subpackage
John Snow [Wed, 15 Sep 2021 16:29:29 +0000 (12:29 -0400)]
python/aqmp: add asynchronous QMP (AQMP) subpackage

For now, it's empty! Soon, it won't be.

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 20210915162955.333025-2-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
2 years agoMerge remote-tracking branch 'remotes/armbru/tags/pull-qapi-2021-09-25-v2' into staging
Peter Maydell [Mon, 27 Sep 2021 14:03:42 +0000 (15:03 +0100)]
Merge remote-tracking branch 'remotes/armbru/tags/pull-qapi-2021-09-25-v2' into staging

QAPI patches patches for 2021-09-25

# gpg: Signature made Mon 27 Sep 2021 13:44:23 BST
# gpg:                using RSA key 354BC8B3D7EB2A6B68674E5F3870B400EB918653
# gpg:                issuer "armbru@redhat.com"
# gpg: Good signature from "Markus Armbruster <armbru@redhat.com>" [full]
# gpg:                 aka "Markus Armbruster <armbru@pond.sub.org>" [full]
# Primary key fingerprint: 354B C8B3 D7EB 2A6B 6867  4E5F 3870 B400 EB91 8653

* remotes/armbru/tags/pull-qapi-2021-09-25-v2: (25 commits)
  tests/qapi-schema: Make test-qapi.py -u work when files are absent
  tests/qapi-schema: Use Python OSError instead of outmoded IOError
  test-clone-visitor: Correct an accidental rename
  tests/qapi-schema: Rename flat-union-* test cases to union-*
  qapi: Drop simple unions
  tests/qapi-schema: Purge simple unions from tests
  tests/qapi-schema: Drop simple union __org.qemu_x-Union1
  test-clone-visitor: Wean off __org.qemu_x-Union1
  tests/qapi-schema: Rewrite simple union TestIfUnion to be flat
  tests/qapi-schema: Simple union UserDefListUnion is now unused, drop
  tests/qapi-schema: Wean off UserDefListUnion
  test-clone-visitor: Wean off UserDefListUnion
  test-qobject-output-visitor: Wean off UserDefListUnion
  test-qobject-input-visitor: Wean off UserDefListUnion
  tests/qapi-schema: Prepare for simple union UserDefListUnion removal
  qapi: Convert simple union TransactionAction to flat one
  qapi: Convert simple union ImageInfoSpecific to flat one
  qapi: Convert simple union SocketAddressLegacy to flat one
  qapi: Convert simple union ChardevBackend to flat one
  qapi: Convert simple union MemoryDeviceInfo to flat one
  ...

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
2 years agoMerge remote-tracking branch 'remotes/vivier2/tags/trivial-branch-for-6.2-pull-reques...
Peter Maydell [Mon, 27 Sep 2021 10:08:36 +0000 (11:08 +0100)]
Merge remote-tracking branch 'remotes/vivier2/tags/trivial-branch-for-6.2-pull-request' into staging

Trivial patches pull request 20210927

# gpg: Signature made Mon 27 Sep 2021 10:52:10 BST
# gpg:                using RSA key CD2F75DDC8E3A4DC2E4F5173F30C38BD3F2FBE3C
# gpg:                issuer "laurent@vivier.eu"
# gpg: Good signature from "Laurent Vivier <lvivier@redhat.com>" [full]
# gpg:                 aka "Laurent Vivier <laurent@vivier.eu>" [full]
# gpg:                 aka "Laurent Vivier (Red Hat) <lvivier@redhat.com>" [full]
# Primary key fingerprint: CD2F 75DD C8E3 A4DC 2E4F  5173 F30C 38BD 3F2F BE3C

* remotes/vivier2/tags/trivial-branch-for-6.2-pull-request:
  multi-process: fix usage information
  hmp: Drop a bogus sentence from set_password's documentation
  hmp: Unbreak "change vnc"
  hw/loader: Restrict PC_ROM_* definitions to hw/i386/pc
  docs/nvdimm: Update nvdimm option value in machine example

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
2 years agomulti-process: fix usage information
Dongli Zhang [Tue, 13 Jul 2021 00:47:18 +0000 (17:47 -0700)]
multi-process: fix usage information

From source code, the 'devid' of x-remote-object should be one of devices
in remote QEMU process.

Signed-off-by: Dongli Zhang <dongli.zhang@oracle.com>
Reviewed-by: Jagannathan Raman <jag.raman@oracle.com>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20210713004718.20381-1-dongli.zhang@oracle.com>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
2 years agohmp: Drop a bogus sentence from set_password's documentation
Markus Armbruster [Thu, 9 Sep 2021 08:12:19 +0000 (10:12 +0200)]
hmp: Drop a bogus sentence from set_password's documentation

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Gerd Hoffmann <kraxel@redhat.com>
Message-Id: <20210909081219.308065-3-armbru@redhat.com>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
2 years agohmp: Unbreak "change vnc"
Markus Armbruster [Thu, 9 Sep 2021 08:12:18 +0000 (10:12 +0200)]
hmp: Unbreak "change vnc"

HMP command "change vnc" can take the password as argument, or prompt
for it:

    (qemu) change vnc password 123
    (qemu) change vnc password
    Password: ***
    (qemu)

This regressed in commit cfb5387a1d "hmp: remove "change vnc TARGET"
command", v6.0.0.

    (qemu) change vnc passwd 123
    Password: ***
    (qemu) change vnc passwd
    (qemu)

The latter passes NULL to qmp_change_vnc_password(), which is a no-no.
Looks like it puts the display into "password required, but none set"
state.

The logic error is easy to miss in review, but testing should've
caught it.

Fix the obvious way.

Fixes: cfb5387a1de2acda23fb5c97d2378b9e7ddf8025
Cc: qemu-stable@nongnu.org
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Gerd Hoffmann <kraxel@redhat.com>
Message-Id: <20210909081219.308065-2-armbru@redhat.com>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
2 years agohw/loader: Restrict PC_ROM_* definitions to hw/i386/pc
Philippe Mathieu-Daudé [Fri, 17 Sep 2021 18:59:49 +0000 (20:59 +0200)]
hw/loader: Restrict PC_ROM_* definitions to hw/i386/pc

The PC_ROM_* definitions are only used by the PC machine,
and are irrelevant to the other architectures / machines.
Reduce their scope by moving them to hw/i386/pc.c.

Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20210917185949.2244956-1-philmd@redhat.com>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
2 years agodocs/nvdimm: Update nvdimm option value in machine example
Pankaj Gupta [Thu, 23 Sep 2021 10:30:15 +0000 (12:30 +0200)]
docs/nvdimm: Update nvdimm option value in machine example

Update nvdimm option value in example command from "-machine pc,nvdimm"
to "-machine pc,nvdimm=on" as former complains with the below error:

"qemu-system-x86_64: -machine pc,nvdimm: Expected '=' after parameter 'nvdimm'"

Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Signed-off-by: Pankaj Gupta <pankaj.gupta.linux@gmail.com>
Message-Id: <20210923103015.135262-1-pankaj.gupta.linux@gmail.com>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
2 years agotests/qapi-schema: Make test-qapi.py -u work when files are absent
Markus Armbruster [Wed, 22 Sep 2021 12:56:19 +0000 (14:56 +0200)]
tests/qapi-schema: Make test-qapi.py -u work when files are absent

test-qapi.py -u updates the expected files.  Since it fails when they
are absent, users have to create them manually before they can use
test-qapi.py to fill in the contents, say for a new test.  Silly.
Improve -u to create them.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20210922125619.670673-3-armbru@redhat.com>
Reviewed-by: John Snow <jsnow@redhat.com>
2 years agotests/qapi-schema: Use Python OSError instead of outmoded IOError
Markus Armbruster [Wed, 22 Sep 2021 12:56:18 +0000 (14:56 +0200)]
tests/qapi-schema: Use Python OSError instead of outmoded IOError

https://docs.python.org/3.6/library/exceptions.html has

    Changed in version 3.3: EnvironmentError, IOError, WindowsError,
    socket.error, select.error and mmap.error have been merged into
    OSError, and the constructor may return a subclass.

and

    The following exceptions are kept for compatibility with previous
    versions; starting from Python 3.3, they are aliases of OSError.

    exception EnvironmentError

    exception IOError

    exception WindowsError

        Only available on Windows.

Switch to the preferred name.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20210922125619.670673-2-armbru@redhat.com>
Reviewed-by: John Snow <jsnow@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
[Details added to commit message]

2 years agotest-clone-visitor: Correct an accidental rename
Markus Armbruster [Fri, 17 Sep 2021 14:31:34 +0000 (16:31 +0200)]
test-clone-visitor: Correct an accidental rename

Commit b359f4b203 "tests: Rename UserDefNativeListUnion to
UserDefListUnion" renamed test_clone_native_list() to
test_clone_list_union().  The function has nothing to do with unions.
Rename it to test_clone_list().

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20210917143134.412106-24-armbru@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
2 years agotests/qapi-schema: Rename flat-union-* test cases to union-*
Markus Armbruster [Fri, 17 Sep 2021 14:31:33 +0000 (16:31 +0200)]
tests/qapi-schema: Rename flat-union-* test cases to union-*

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20210917143134.412106-23-armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com
2 years agoqapi: Drop simple unions
Markus Armbruster [Fri, 17 Sep 2021 14:31:32 +0000 (16:31 +0200)]
qapi: Drop simple unions

Simple unions predate flat unions.  Having both complicates the QAPI
schema language and the QAPI generator.  We haven't been using simple
unions in new code for a long time, because they are less flexible and
somewhat awkward on the wire.

The previous commits eliminated simple union from the tree.  Now drop
them from the QAPI schema language entirely, and update mentions of
"flat union" to just "union".

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20210917143134.412106-22-armbru@redhat.com>

2 years agotests/qapi-schema: Purge simple unions from tests
Markus Armbruster [Fri, 17 Sep 2021 14:31:31 +0000 (16:31 +0200)]
tests/qapi-schema: Purge simple unions from tests

Drop tests that are specifically about simple unions:

* SugaredUnion in doc-good: flat unions are covered by @Object.

* union-branch-case and union-clash-branches: branch naming for flat
  unions is enforced for the tag enum instead, which is covered by
  enum-member-case and enum-clash-member.

* union-empty: empty flat unions are covered by flat-union-empty.

Rewrite the remainder to use flat unions: args-union, bad-base,
flat-union-base-union, union-branch-invalid-dict, union-unknown.

Except drop union-optional-branch. because converting this one is not
worth the trouble; we don't explicitly check names beginning with '*'
in other places, either.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20210917143134.412106-21-armbru@redhat.com>

2 years agotests/qapi-schema: Drop simple union __org.qemu_x-Union1
Markus Armbruster [Fri, 17 Sep 2021 14:31:30 +0000 (16:31 +0200)]
tests/qapi-schema: Drop simple union __org.qemu_x-Union1

Replace simple union __org.qemu_x-Union1 with flat union
__org.qemu_x-Union2, except drop it from __org.qemu_x-command, because
there it's only used to pull it into QMP.  Now drop the unused
-Union1, and rename -Union2 to -Union.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20210917143134.412106-20-armbru@redhat.com>

2 years agotest-clone-visitor: Wean off __org.qemu_x-Union1
Markus Armbruster [Fri, 17 Sep 2021 14:31:29 +0000 (16:31 +0200)]
test-clone-visitor: Wean off __org.qemu_x-Union1

test_clone_complex3() uses simple union __org.qemu_x-Union1 to cover
arrays.  Use UserDefOneList instead.  Unions are still covered by
test_clone_complex1().

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20210917143134.412106-19-armbru@redhat.com>

2 years agotests/qapi-schema: Rewrite simple union TestIfUnion to be flat
Markus Armbruster [Fri, 17 Sep 2021 14:31:28 +0000 (16:31 +0200)]
tests/qapi-schema: Rewrite simple union TestIfUnion to be flat

Simple unions predate flat unions.  Having both complicates the QAPI
schema language and the QAPI generator.  We haven't been using simple
unions in new code for a long time, because they are less flexible and
somewhat awkward on the wire.

To prepare for their removal, rewrite TestIfUnion to be flat.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20210917143134.412106-18-armbru@redhat.com>

2 years agotests/qapi-schema: Simple union UserDefListUnion is now unused, drop
Markus Armbruster [Fri, 17 Sep 2021 14:31:27 +0000 (16:31 +0200)]
tests/qapi-schema: Simple union UserDefListUnion is now unused, drop

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20210917143134.412106-17-armbru@redhat.com>

2 years agotests/qapi-schema: Wean off UserDefListUnion
Markus Armbruster [Fri, 17 Sep 2021 14:31:26 +0000 (16:31 +0200)]
tests/qapi-schema: Wean off UserDefListUnion

Command boxed-union uses simple union UserDefListUnion to cover
unions.  Use UserDefFlatUnion instead.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20210917143134.412106-16-armbru@redhat.com>

2 years agotest-clone-visitor: Wean off UserDefListUnion
Markus Armbruster [Fri, 17 Sep 2021 14:31:25 +0000 (16:31 +0200)]
test-clone-visitor: Wean off UserDefListUnion

test_clone_complex1() uses simple union UserDefListUnion to cover
unions.  Use UserDefFlatUnion instead.  Arrays are still covered by
test_clone_complex3().

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20210917143134.412106-15-armbru@redhat.com>

2 years agotest-qobject-output-visitor: Wean off UserDefListUnion
Markus Armbruster [Fri, 17 Sep 2021 14:31:24 +0000 (16:31 +0200)]
test-qobject-output-visitor: Wean off UserDefListUnion

The test_visitor_out_list_union_FOO() use simple union
UserDefListUnion to cover lists of builtin types.  Rewrite as
test_visitor_out_list_struct(), using struct ArrayStruct and a lot
less code.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20210917143134.412106-14-armbru@redhat.com>

2 years agotest-qobject-input-visitor: Wean off UserDefListUnion
Markus Armbruster [Fri, 17 Sep 2021 14:31:23 +0000 (16:31 +0200)]
test-qobject-input-visitor: Wean off UserDefListUnion

The test_visitor_in_list_union_FOO() use simple union UserDefListUnion
to cover lists of builtin types.  Rewrite as
test_visitor_in_list_struct(), using struct ArrayStruct and a lot less
code.

test_visitor_in_fail_union_list() uses UserDefListUnion to cover
"variant members don't match the discriminator value".  Cover that in
test_visitor_in_fail_union_flat() instead, and drop
test_visitor_in_fail_union_list().  Appropriating the former for this
purpose is okay, because it actually failed due to missing
discriminator, which is still covered by
test_visitor_in_fail_union_flat_no_discrim().

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20210917143134.412106-13-armbru@redhat.com>

2 years agotests/qapi-schema: Prepare for simple union UserDefListUnion removal
Markus Armbruster [Fri, 17 Sep 2021 14:31:22 +0000 (16:31 +0200)]
tests/qapi-schema: Prepare for simple union UserDefListUnion removal

Simple unions predate flat unions.  Having both complicates the QAPI
schema language and the QAPI generator.  We haven't been using simple
unions in new code for a long time, because they are less flexible and
somewhat awkward on the wire.

To prepare for their removal, simple union UserDefListUnion has to go.
It is used to cover arrays.  The next few commits will eliminate its
uses, and then it gets deleted.  As a first step, provide struct
ArrayStruct for the tests to be rewritten.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20210917143134.412106-12-armbru@redhat.com>

2 years agoqapi: Convert simple union TransactionAction to flat one
Markus Armbruster [Fri, 17 Sep 2021 14:31:21 +0000 (16:31 +0200)]
qapi: Convert simple union TransactionAction to flat one

Simple unions predate flat unions.  Having both complicates the QAPI
schema language and the QAPI generator.  We haven't been using simple
unions in new code for a long time, because they are less flexible and
somewhat awkward on the wire.

To prepare for their removal, convert simple union TransactionAction
to an equivalent flat one.  Adds some boilerplate to the schema, which
is a bit ugly, but a lot easier to maintain than the simple union
feature.

Cc: Kevin Wolf <kwolf@redhat.com>
Cc: Hanna Reitz <hreitz@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Acked-by: Hanna Reitz <hreitz@redhat.com>
Message-Id: <20210917143134.412106-11-armbru@redhat.com>

2 years agoqapi: Convert simple union ImageInfoSpecific to flat one
Markus Armbruster [Fri, 17 Sep 2021 14:31:20 +0000 (16:31 +0200)]
qapi: Convert simple union ImageInfoSpecific to flat one

Simple unions predate flat unions.  Having both complicates the QAPI
schema language and the QAPI generator.  We haven't been using simple
unions in new code for a long time, because they are less flexible and
somewhat awkward on the wire.

To prepare for their removal, convert simple union ImageInfoSpecific
to an equivalent flat one.  Adds some boilerplate to the schema, which
is a bit ugly, but a lot easier to maintain than the simple union
feature.

Implicit enum ImageInfoSpecificKind becomes explicit.  It duplicates
part of enum BlockdevDriver.  We could reuse BlockdevDriver instead.

Cc: Kevin Wolf <kwolf@redhat.com>
Cc: Hanna Reitz <hreitz@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Acked-by: Hanna Reitz <hreitz@redhat.com>
Message-Id: <20210917143134.412106-10-armbru@redhat.com>

2 years agoqapi: Convert simple union SocketAddressLegacy to flat one
Markus Armbruster [Fri, 17 Sep 2021 14:31:19 +0000 (16:31 +0200)]
qapi: Convert simple union SocketAddressLegacy to flat one

Simple unions predate flat unions.  Having both complicates the QAPI
schema language and the QAPI generator.  We haven't been using simple
unions in new code for a long time, because they are less flexible and
somewhat awkward on the wire.

To prepare for their removal, convert simple union SocketAddressLegacy
to an equivalent flat one, with existing enum SocketAddressType
replacing implicit enum type SocketAddressLegacyKind.  Adds some
boilerplate to the schema, which is a bit ugly, but a lot easier to
maintain than the simple union feature.

Cc: "Daniel P. Berrangé" <berrange@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20210917143134.412106-9-armbru@redhat.com>

2 years agoqapi: Convert simple union ChardevBackend to flat one
Markus Armbruster [Fri, 17 Sep 2021 14:31:18 +0000 (16:31 +0200)]
qapi: Convert simple union ChardevBackend to flat one

Simple unions predate flat unions.  Having both complicates the QAPI
schema language and the QAPI generator.  We haven't been using simple
unions in new code for a long time, because they are less flexible and
somewhat awkward on the wire.

To prepare for their removal, convert simple union ChardevBackend to
an equivalent flat one.  Adds some boilerplate to the schema, which is
a bit ugly, but a lot easier to maintain than the simple union
feature.

Cc: "Marc-André Lureau" <marcandre.lureau@redhat.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20210917143134.412106-8-armbru@redhat.com>
[Missing conditionals added]

2 years agoqapi: Convert simple union MemoryDeviceInfo to flat one
Markus Armbruster [Fri, 17 Sep 2021 14:31:17 +0000 (16:31 +0200)]
qapi: Convert simple union MemoryDeviceInfo to flat one

Simple unions predate flat unions.  Having both complicates the QAPI
schema language and the QAPI generator.  We haven't been using simple
unions in new code for a long time, because they are less flexible and
somewhat awkward on the wire.

To prepare for their removal, convert simple union MemoryDeviceInfo to
an equivalent flat one.  Adds some boilerplate to the schema, which is
a bit ugly, but a lot easier to maintain than the simple union
feature.

Cc: Eduardo Habkost <ehabkost@redhat.com>
Cc: Marcel Apfelbaum <marcel.apfelbaum@gmail.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20210917143134.412106-7-armbru@redhat.com>

2 years agoqapi: Convert simple union TpmTypeOptions to flat one
Markus Armbruster [Fri, 17 Sep 2021 14:31:16 +0000 (16:31 +0200)]
qapi: Convert simple union TpmTypeOptions to flat one

Simple unions predate flat unions.  Having both complicates the QAPI
schema language and the QAPI generator.  We haven't been using simple
unions in new code for a long time, because they are less flexible and
somewhat awkward on the wire.

To prepare for their removal, convert simple union TpmTypeOptions to
an equivalent flat one, with existing enum TpmType replacing implicit
enum TpmTypeOptionsKind.  Adds some boilerplate to the schema, which
is a bit ugly, but a lot easier to maintain than the simple union
feature.

Cc: Stefan Berger <stefanb@linux.vnet.ibm.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Acked-by: Stefan Berger <stefanb@linux.ibm.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20210917143134.412106-6-armbru@redhat.com>
[Indentation tidied up]

2 years agoqapi: Convert simple union InputEvent to flat one
Markus Armbruster [Fri, 17 Sep 2021 14:31:15 +0000 (16:31 +0200)]
qapi: Convert simple union InputEvent to flat one

Simple unions predate flat unions.  Having both complicates the QAPI
schema language and the QAPI generator.  We haven't been using simple
unions in new code for a long time, because they are less flexible and
somewhat awkward on the wire.

To prepare for their removal, convert simple union InputEvent to an
equivalent flat one.  Adds some boilerplate to the schema, which is a
bit ugly, but a lot easier to maintain than the simple union feature.

Cc: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20210917143134.412106-5-armbru@redhat.com>

2 years agoqapi: Convert simple union KeyValue to flat one
Markus Armbruster [Fri, 17 Sep 2021 14:31:14 +0000 (16:31 +0200)]
qapi: Convert simple union KeyValue to flat one

Simple unions predate flat unions.  Having both complicates the QAPI
schema language and the QAPI generator.  We haven't been using simple
unions in new code for a long time, because they are less flexible and
somewhat awkward on the wire.

To prepare for their removal, convert simple union KeyValue to an
equivalent flat one.  Adds some boilerplate to the schema, which is a
bit ugly, but a lot easier to maintain than the simple union feature.

Cc: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20210917143134.412106-4-armbru@redhat.com>

2 years agoqapi: Stop enforcing "type name should not end in 'Kind'
Markus Armbruster [Fri, 17 Sep 2021 14:31:13 +0000 (16:31 +0200)]
qapi: Stop enforcing "type name should not end in 'Kind'

I'm about to convert simple unions to flat unions, then drop simple
union support.  The conversion involves making the implict enum types
explicit.  To reduce churn, I'd like to name them exactly like the
implicit types they replace.  However, these names are reserved for
the generator's use.  They won't be once simple unions are gone.  Stop
enforcing this naming rule now rather than then.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20210917143134.412106-3-armbru@redhat.com>

2 years agoqapi: Tidy up unusual line breaks
Markus Armbruster [Fri, 17 Sep 2021 14:31:12 +0000 (16:31 +0200)]
qapi: Tidy up unusual line breaks

Break lines between members instead of within members.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20210917143134.412106-2-armbru@redhat.com>

2 years agoMerge remote-tracking branch 'remotes/rth/tags/pull-tcg-20210921' into staging
Richard Henderson [Fri, 24 Sep 2021 17:21:18 +0000 (13:21 -0400)]
Merge remote-tracking branch 'remotes/rth/tags/pull-tcg-20210921' into staging

Move cpu_signal_handler declaration.
Restrict cpu_handle_halt to sysemu.
Make do_unaligned_access noreturn.
Misc tcg/mips cleanup
Misc tcg/sparc cleanup
Misc tcg/riscv cleanup

# gpg: Signature made Tue 21 Sep 2021 10:47:29 PM EDT
# gpg:                using RSA key 7A481E78868B4DB6A85A05C064DF38E8AF7E215F
# gpg:                issuer "richard.henderson@linaro.org"
# gpg: Good signature from "Richard Henderson <richard.henderson@linaro.org>" [ultimate]

* remotes/rth/tags/pull-tcg-20210921:
  tcg/riscv: Remove add with zero on user-only memory access
  hw/core: Make do_unaligned_access noreturn
  tcg/sparc: Introduce tcg_out_mov_delay
  tcg/sparc: Drop inline markers
  tcg/mips: Drop special alignment for code_gen_buffer
  tcg/mips: Unset TCG_TARGET_HAS_direct_jump
  tcg/mips: Allow JAL to be out of range in tcg_out_bswap_subr
  tcg/mips: Drop inline markers
  accel/tcg: Restrict cpu_handle_halt() to sysemu
  include/exec: Move cpu_signal_handler declaration

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
2 years agoMerge remote-tracking branch 'remotes/vivier2/tags/linux-user-for-6.2-pull-request...
Peter Maydell [Fri, 24 Sep 2021 14:05:12 +0000 (15:05 +0100)]
Merge remote-tracking branch 'remotes/vivier2/tags/linux-user-for-6.2-pull-request' into staging

Pull request linux-user 20210924

Clean up siginfo_t handling for arm, aarch64

# gpg: Signature made Fri 24 Sep 2021 14:56:12 BST
# gpg:                using RSA key CD2F75DDC8E3A4DC2E4F5173F30C38BD3F2FBE3C
# gpg:                issuer "laurent@vivier.eu"
# gpg: Good signature from "Laurent Vivier <lvivier@redhat.com>" [full]
# gpg:                 aka "Laurent Vivier <laurent@vivier.eu>" [full]
# gpg:                 aka "Laurent Vivier (Red Hat) <lvivier@redhat.com>" [full]
# Primary key fingerprint: CD2F 75DD C8E3 A4DC 2E4F  5173 F30C 38BD 3F2F BE3C

* remotes/vivier2/tags/linux-user-for-6.2-pull-request:
  linux-user/aarch64: Use force_sig_fault()
  linux-user/arm: Use force_sig_fault()
  linux-user: Provide new force_sig_fault() function
  linux-user: Zero out target_siginfo_t in force_sig()
  linux-user/arm: Use force_sig() to deliver fpa11 emulation SIGFPE
  linux-user/arm: Set siginfo_t addr field for SIGTRAP signals
  linux-user/aarch64: Set siginfo_t addr field for SIGTRAP signals

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
2 years agoMerge remote-tracking branch 'remotes/nvme/tags/nvme-next-pull-request' into staging
Peter Maydell [Fri, 24 Sep 2021 09:28:13 +0000 (10:28 +0100)]
Merge remote-tracking branch 'remotes/nvme/tags/nvme-next-pull-request' into staging

hw/nvme updates

# gpg: Signature made Fri 24 Sep 2021 07:44:23 BST
# gpg:                using RSA key 522833AA75E2DCE6A24766C04DE1AF316D4F0DE9
# gpg: Good signature from "Klaus Jensen <its@irrelevant.dk>" [unknown]
# gpg:                 aka "Klaus Jensen <k.jensen@samsung.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: DDCA 4D9C 9EF9 31CC 3468  4272 63D5 6FC5 E55D A838
#      Subkey fingerprint: 5228 33AA 75E2 DCE6 A247  66C0 4DE1 AF31 6D4F 0DE9

* remotes/nvme/tags/nvme-next-pull-request:
  hw/nvme: Return error for fused operations
  hw/nvme: fix verification of select field in namespace attachment
  hw/nvme: fix validation of ASQ and ACQ

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
2 years agohw/nvme: Return error for fused operations
Pankaj Raghav [Wed, 15 Sep 2021 15:43:30 +0000 (17:43 +0200)]
hw/nvme: Return error for fused operations

Currently, FUSED operations are not supported by QEMU. As per the 1.4 SPEC,
controller should abort the command that requested a fused operation with
an INVALID FIELD error code if they are not supported.

Changes from v1:
Added FUSE flag check also to the admin cmd processing as the FUSED
operations are mentioned in the general SQE section in the SPEC.

Signed-off-by: Pankaj Raghav <p.raghav@samsung.com>
Reviewed-by: Keith Busch <kbusch@kernel.org>
Signed-off-by: Klaus Jensen <k.jensen@samsung.com>
2 years agohw/nvme: fix verification of select field in namespace attachment
Naveen Nagar [Mon, 23 Aug 2021 11:03:33 +0000 (16:33 +0530)]
hw/nvme: fix verification of select field in namespace attachment

Fix is added to check for reserved value in select field for
namespace attachment

CC: Minwoo Im <minwoo.im.dev@gmail.com>
Signed-off-by: Naveen Nagar <naveen.n1@samsung.com>
Signed-off-by: Klaus Jensen <k.jensen@samsung.com>
2 years agohw/nvme: fix validation of ASQ and ACQ
Klaus Jensen [Mon, 23 Aug 2021 12:11:33 +0000 (14:11 +0200)]
hw/nvme: fix validation of ASQ and ACQ

Address 0x0 is a valid address. Fix the admin submission and completion
queue address validation to not error out on this.

Signed-off-by: Klaus Jensen <k.jensen@samsung.com>
Reviewed-by: Keith Busch <kbusch@kernel.org>
2 years agolinux-user/aarch64: Use force_sig_fault()
Peter Maydell [Fri, 13 Aug 2021 13:18:09 +0000 (14:18 +0100)]
linux-user/aarch64: Use force_sig_fault()

Use the new force_sig_fault() function instead of setting up
a target_siginfo_t and calling queue_signal().

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20210813131809.28655-8-peter.maydell@linaro.org>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
2 years agolinux-user/arm: Use force_sig_fault()
Peter Maydell [Fri, 13 Aug 2021 13:18:08 +0000 (14:18 +0100)]
linux-user/arm: Use force_sig_fault()

Use the new force_sig_fault() function instead of setting up
a target_siginfo_t and calling queue_signal().

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20210813131809.28655-7-peter.maydell@linaro.org>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
2 years agolinux-user: Provide new force_sig_fault() function
Peter Maydell [Fri, 13 Aug 2021 13:18:07 +0000 (14:18 +0100)]
linux-user: Provide new force_sig_fault() function

In many places in the linux-user code we need to queue a signal for
the guest using the QEMU_SI_FAULT si_type.  This requires that the
caller sets up and passes us a target_siginfo, including setting the
appropriate part of the _sifields union for the si_type. In a number
of places the code forgets to set the _sifields union field.

Provide a new force_sig_fault() function, which does the same thing
as the Linux kernel function of that name -- it takes the signal
number, the si_code value and the address to use in
_sifields._sigfault, and assembles the target_siginfo itself.  This
makes the callsites simpler and means it's harder to forget to pass
in an address value.

We follow force_sig() and the kernel's force_sig_fault() in not
requiring the caller to pass in the CPU pointer but always acting
on the CPU of the current thread.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20210813131809.28655-6-peter.maydell@linaro.org>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
2 years agolinux-user: Zero out target_siginfo_t in force_sig()
Peter Maydell [Fri, 13 Aug 2021 13:18:06 +0000 (14:18 +0100)]
linux-user: Zero out target_siginfo_t in force_sig()

The target_siginfo_t we populate in force_sig() will eventually
get copied onto the target's stack. Zero it out so that any extra
padding in the sifields union is consistently zero when the guest
sees it.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20210813131809.28655-5-peter.maydell@linaro.org>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
2 years agolinux-user/arm: Use force_sig() to deliver fpa11 emulation SIGFPE
Peter Maydell [Fri, 13 Aug 2021 13:18:05 +0000 (14:18 +0100)]
linux-user/arm: Use force_sig() to deliver fpa11 emulation SIGFPE

In the Arm target code, when the fpa11 emulation code tells us we
need to send the guest a SIGFPE, we do this with queue_signal(), but
we are using the wrong si_type, and we aren't setting the _sifields
union members corresponding to either the si_type we are using or the
si_type we should be using.

As the existing comment notes, the kernel code for this calls the old
send_sig() function to deliver the signal.  This eventually results
in the kernel's signal handling code fabricating a siginfo_t with a
SI_KERNEL code and a zero pid and uid.  For QEMU this means we need
to use QEMU_SI_KILL.  We already have a function for that:
force_sig() sets up the whole target_siginfo_t the way we need it.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20210813131809.28655-4-peter.maydell@linaro.org>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
2 years agolinux-user/arm: Set siginfo_t addr field for SIGTRAP signals
Peter Maydell [Fri, 13 Aug 2021 13:18:04 +0000 (14:18 +0100)]
linux-user/arm: Set siginfo_t addr field for SIGTRAP signals

When generating a TRAP_BRKPT SIGTRAP, set the siginfo_t addr field
to the PC where the breakpoint/singlestep trap occurred; this is
what the kernel does for this signal for this architecture.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20210813131809.28655-3-peter.maydell@linaro.org>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
2 years agolinux-user/aarch64: Set siginfo_t addr field for SIGTRAP signals
Peter Maydell [Fri, 13 Aug 2021 13:18:03 +0000 (14:18 +0100)]
linux-user/aarch64: Set siginfo_t addr field for SIGTRAP signals

When generating a TRAP_BRKPT SIGTRAP, set the siginfo_t addr field
to the PC where the breakpoint/singlestep trap occurred; this is
what the kernel does for this signal for this architecture.

Fixes: Coverity 1459154
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20210813131809.28655-2-peter.maydell@linaro.org>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
2 years agotcg/riscv: Remove add with zero on user-only memory access
Richard Henderson [Thu, 5 Aug 2021 17:38:14 +0000 (07:38 -1000)]
tcg/riscv: Remove add with zero on user-only memory access

Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
2 years agohw/core: Make do_unaligned_access noreturn
Richard Henderson [Thu, 29 Jul 2021 20:45:10 +0000 (10:45 -1000)]
hw/core: Make do_unaligned_access noreturn

While we may have had some thought of allowing system-mode
to return from this hook, we have no guests that require this.

Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
2 years agotcg/sparc: Introduce tcg_out_mov_delay
Richard Henderson [Wed, 4 Aug 2021 22:15:53 +0000 (01:15 +0300)]
tcg/sparc: Introduce tcg_out_mov_delay

This version of tcg_out_mov is emits a nop to fill the
delay slot if the move is not required.

The only current use, for INDEX_op_goto_ptr, will always
require the move but properly documents the delay slot.

Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
2 years agotcg/sparc: Drop inline markers
Richard Henderson [Wed, 4 Aug 2021 22:09:41 +0000 (01:09 +0300)]
tcg/sparc: Drop inline markers

Let the compiler decide about inlining.

Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
2 years agotcg/mips: Drop special alignment for code_gen_buffer
Richard Henderson [Fri, 6 Aug 2021 18:20:16 +0000 (08:20 -1000)]
tcg/mips: Drop special alignment for code_gen_buffer

Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
2 years agotcg/mips: Unset TCG_TARGET_HAS_direct_jump
Richard Henderson [Fri, 6 Aug 2021 17:49:03 +0000 (07:49 -1000)]
tcg/mips: Unset TCG_TARGET_HAS_direct_jump

Only use indirect jumps.  Finish weaning away from the
unique alignment requirements for code_gen_buffer.

Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
2 years agotcg/mips: Allow JAL to be out of range in tcg_out_bswap_subr
Richard Henderson [Fri, 6 Aug 2021 18:03:45 +0000 (08:03 -1000)]
tcg/mips: Allow JAL to be out of range in tcg_out_bswap_subr

Weaning off of unique alignment requirements, so allow JAL
to not reach the target.  TCG_TMP1 is always available for
use as a scratch because it is clobbered by the subroutine
being called.

Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
2 years agotcg/mips: Drop inline markers
Richard Henderson [Fri, 6 Aug 2021 17:04:00 +0000 (07:04 -1000)]
tcg/mips: Drop inline markers

Let the compiler decide about inlining.
Remove tcg_out_ext8s and tcg_out_ext16s as unused.

Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
2 years agoaccel/tcg: Restrict cpu_handle_halt() to sysemu
Philippe Mathieu-Daudé [Sun, 12 Sep 2021 17:27:02 +0000 (19:27 +0200)]
accel/tcg: Restrict cpu_handle_halt() to sysemu

Commit 372579427a5 ("tcg: enable thread-per-vCPU") added the following
comment describing EXCP_HALTED in qemu_tcg_cpu_thread_fn():

    case EXCP_HALTED:
         /* during start-up the vCPU is reset and the thread is
          * kicked several times. If we don't ensure we go back
          * to sleep in the halted state we won't cleanly
          * start-up when the vCPU is enabled.
          *
          * cpu->halted should ensure we sleep in wait_io_event
          */
         g_assert(cpu->halted);
         break;

qemu_wait_io_event() is sysemu-specific, so we can restrict the
cpu_handle_halt() call in cpu_exec() to system emulation.

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20210912172731.789788-2-f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
2 years agoinclude/exec: Move cpu_signal_handler declaration
Richard Henderson [Tue, 3 Aug 2021 15:31:43 +0000 (05:31 -1000)]
include/exec: Move cpu_signal_handler declaration

There is nothing target specific about this.  The implementation
is host specific, but the declaration is 100% common.

Reviewed-By: Warner Losh <imp@bsdimp.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
2 years agoMerge remote-tracking branch 'remotes/alistair23/tags/pull-riscv-to-apply-20210921...
Richard Henderson [Tue, 21 Sep 2021 17:57:48 +0000 (10:57 -0700)]
Merge remote-tracking branch 'remotes/alistair23/tags/pull-riscv-to-apply-20210921' into staging

Second RISC-V PR for QEMU 6.2

 - ePMP CSR address updates
 - Convert internal interrupts to use QEMU GPIO lines
 - SiFive PWM support
 - Support for RISC-V ACLINT
 - SiFive PDMA fixes
 - Update to u-boot instructions for sifive_u
 - mstatus.SD bug fix for hypervisor extensions
 - OpenTitan fix for USB dev address

# gpg: Signature made Mon 20 Sep 2021 11:52:26 PM PDT
# gpg:                using RSA key F6C4AC46D4934868D3B8CE8F21E10D29DF977054
# gpg: Good signature from "Alistair Francis <alistair@alistair23.me>" [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: F6C4 AC46 D493 4868 D3B8  CE8F 21E1 0D29 DF97 7054

* remotes/alistair23/tags/pull-riscv-to-apply-20210921: (21 commits)
  hw/riscv: opentitan: Correct the USB Dev address
  target/riscv: csr: Rename HCOUNTEREN_CY and friends
  target/riscv: Backup/restore mstatus.SD bit when virtual register swapped
  docs/system/riscv: sifive_u: Update U-Boot instructions
  hw/dma: sifive_pdma: don't set Control.error if 0 bytes to transfer
  hw/dma: sifive_pdma: allow non-multiple transaction size transactions
  hw/dma: sifive_pdma: claim bit must be set before DMA transactions
  hw/dma: sifive_pdma: reset Next* registers when Control.claim is set
  hw/riscv: virt: Add optional ACLINT support to virt machine
  hw/riscv: virt: Re-factor FDT generation
  hw/intc: Upgrade the SiFive CLINT implementation to RISC-V ACLINT
  hw/intc: Rename sifive_clint sources to riscv_aclint sources
  sifive_u: Connect the SiFive PWM device
  hw/timer: Add SiFive PWM support
  hw/intc: ibex_timer: Convert the timer to use RISC-V CPU GPIO lines
  hw/intc: sifive_plic: Convert the PLIC to use RISC-V CPU GPIO lines
  hw/intc: ibex_plic: Convert the PLIC to use RISC-V CPU GPIO lines
  hw/intc: sifive_clint: Use RISC-V CPU GPIO lines
  target/riscv: Expose interrupt pending bits as GPIO lines
  target/riscv: Fix satp write
  ...

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
2 years agoMerge remote-tracking branch 'remotes/pmaydell/tags/pull-target-arm-20210921' into...
Peter Maydell [Tue, 21 Sep 2021 15:32:19 +0000 (16:32 +0100)]
Merge remote-tracking branch 'remotes/pmaydell/tags/pull-target-arm-20210921' into staging

target-arm queue:
 * Optimize codegen for MVE when predication not active
 * hvf: Add Apple Silicon support
 * hw/intc: Set GIC maintenance interrupt level to only 0 or 1
 * Fix mishandling of MVE FPSCR.LTPSIZE reset for usermode emulator
 * elf2dmp: Fix coverity nits

# gpg: Signature made Tue 21 Sep 2021 16:31:17 BST
# gpg:                using RSA key E1A5C593CD419DE28E8315CF3C2525ED14360CDE
# gpg:                issuer "peter.maydell@linaro.org"
# gpg: Good signature from "Peter Maydell <peter.maydell@linaro.org>" [ultimate]
# gpg:                 aka "Peter Maydell <pmaydell@gmail.com>" [ultimate]
# gpg:                 aka "Peter Maydell <pmaydell@chiark.greenend.org.uk>" [ultimate]
# Primary key fingerprint: E1A5 C593 CD41 9DE2 8E83  15CF 3C25 25ED 1436 0CDE

* remotes/pmaydell/tags/pull-target-arm-20210921: (27 commits)
  target/arm: Optimize MVE 1op-immediate insns
  target/arm: Optimize MVE VSLI and VSRI
  target/arm: Optimize MVE VSHLL and VMOVL
  target/arm: Optimize MVE VSHL, VSHR immediate forms
  target/arm: Optimize MVE VMVN
  target/arm: Optimize MVE VDUP
  target/arm: Optimize MVE VNEG, VABS
  target/arm: Optimize MVE arithmetic ops
  target/arm: Optimize MVE logic ops
  target/arm: Add TB flag for "MVE insns not predicated"
  target/arm: Enforce that FPDSCR.LTPSIZE is 4 on inbound migration
  target/arm: Avoid goto_tb if we're trying to exit to the main loop
  hvf: arm: Add rudimentary PMC support
  arm: Add Hypervisor.framework build target
  hvf: arm: Implement PSCI handling
  hvf: arm: Implement -cpu host
  arm/hvf: Add a WFI handler
  hvf: Add Apple Silicon support
  hvf: Introduce hvf_arch_init() callback
  hvf: Add execute to dirty log permission bitmap
  ...

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
2 years agotarget/arm: Optimize MVE 1op-immediate insns
Peter Maydell [Mon, 13 Sep 2021 09:54:40 +0000 (10:54 +0100)]
target/arm: Optimize MVE 1op-immediate insns

Optimize the MVE 1op-immediate insns (VORR, VBIC, VMOV) to
use TCG vector ops when possible.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20210913095440.13462-13-peter.maydell@linaro.org

2 years agotarget/arm: Optimize MVE VSLI and VSRI
Peter Maydell [Mon, 13 Sep 2021 09:54:39 +0000 (10:54 +0100)]
target/arm: Optimize MVE VSLI and VSRI

Optimize the MVE shift-and-insert insns by using TCG
vector ops when possible.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20210913095440.13462-12-peter.maydell@linaro.org

2 years agotarget/arm: Optimize MVE VSHLL and VMOVL
Peter Maydell [Mon, 13 Sep 2021 09:54:38 +0000 (10:54 +0100)]
target/arm: Optimize MVE VSHLL and VMOVL

Optimize the MVE VSHLL insns by using TCG vector ops when possible.
This includes the VMOVL insn, which we handle in mve.decode as "VSHLL
with zero shift count".

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20210913095440.13462-11-peter.maydell@linaro.org

2 years agotarget/arm: Optimize MVE VSHL, VSHR immediate forms
Peter Maydell [Mon, 13 Sep 2021 09:54:37 +0000 (10:54 +0100)]
target/arm: Optimize MVE VSHL, VSHR immediate forms

Optimize the MVE VSHL and VSHR immediate forms by using TCG vector
ops when possible.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20210913095440.13462-10-peter.maydell@linaro.org

2 years agotarget/arm: Optimize MVE VMVN
Peter Maydell [Mon, 13 Sep 2021 09:54:36 +0000 (10:54 +0100)]
target/arm: Optimize MVE VMVN

Optimize the MVE VMVN insn by using TCG vector ops when possible.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20210913095440.13462-9-peter.maydell@linaro.org

2 years agotarget/arm: Optimize MVE VDUP
Peter Maydell [Mon, 13 Sep 2021 09:54:35 +0000 (10:54 +0100)]
target/arm: Optimize MVE VDUP

Optimize the MVE VDUP insns by using TCG vector ops when possible.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20210913095440.13462-8-peter.maydell@linaro.org

2 years agotarget/arm: Optimize MVE VNEG, VABS
Peter Maydell [Mon, 13 Sep 2021 09:54:34 +0000 (10:54 +0100)]
target/arm: Optimize MVE VNEG, VABS

Optimize the MVE VNEG and VABS insns by using TCG
vector ops when possible.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20210913095440.13462-7-peter.maydell@linaro.org

2 years agotarget/arm: Optimize MVE arithmetic ops
Peter Maydell [Mon, 13 Sep 2021 09:54:33 +0000 (10:54 +0100)]
target/arm: Optimize MVE arithmetic ops

Optimize MVE arithmetic ops when we have a TCG
vector operation we can use.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20210913095440.13462-6-peter.maydell@linaro.org

2 years agotarget/arm: Optimize MVE logic ops
Peter Maydell [Mon, 13 Sep 2021 09:54:32 +0000 (10:54 +0100)]
target/arm: Optimize MVE logic ops

When not predicating, implement the MVE bitwise logical insns
directly using TCG vector operations.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20210913095440.13462-5-peter.maydell@linaro.org

2 years agotarget/arm: Add TB flag for "MVE insns not predicated"
Peter Maydell [Mon, 13 Sep 2021 09:54:31 +0000 (10:54 +0100)]
target/arm: Add TB flag for "MVE insns not predicated"

Our current codegen for MVE always calls out to helper functions,
because some byte lanes might be predicated.  The common case is that
in fact there is no predication active and all lanes should be
updated together, so we can produce better code by detecting that and
using the TCG generic vector infrastructure.

Add a TB flag that is set when we can guarantee that there is no
active MVE predication, and a bool in the DisasContext.  Subsequent
patches will use this flag to generate improved code for some
instructions.

In most cases when the predication state changes we simply end the TB
after that instruction.  For the code called from vfp_access_check()
that handles lazy state preservation and creating a new FP context,
we can usually avoid having to try to end the TB because luckily the
new value of the flag following the register changes in those
sequences doesn't depend on any runtime decisions.  We do have to end
the TB if the guest has enabled lazy FP state preservation but not
automatic state preservation, but this is an odd corner case that is
not going to be common in real-world code.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20210913095440.13462-4-peter.maydell@linaro.org

2 years agotarget/arm: Enforce that FPDSCR.LTPSIZE is 4 on inbound migration
Peter Maydell [Mon, 13 Sep 2021 09:54:30 +0000 (10:54 +0100)]
target/arm: Enforce that FPDSCR.LTPSIZE is 4 on inbound migration

Architecturally, for an M-profile CPU with the LOB feature the
LTPSIZE field in FPDSCR is always constant 4.  QEMU's implementation
enforces this everywhere, except that we don't check that it is true
in incoming migration data.

We're going to add come in gen_update_fp_context() which relies on
the "always 4" property.  Since this is TCG-only, we don't actually
need to be robust to bogus incoming migration data, and the effect of
it being wrong would be wrong code generation rather than a QEMU
crash; but if it did ever happen somehow it would be very difficult
to track down the cause.  Add a check so that we fail the inbound
migration if the FPDSCR.LTPSIZE value is incorrect.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20210913095440.13462-3-peter.maydell@linaro.org