OSDN Git Service

qmiga/qemu.git
2 years agotarget/alpha: Reorg integer memory operations
Richard Henderson [Sat, 31 Jul 2021 06:15:31 +0000 (20:15 -1000)]
target/alpha: Reorg integer memory operations

Pass in the MemOp instead of a callback.
Drop the fp argument; add a locked argument.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
2 years agotarget/alpha: Reorg fp memory operations
Richard Henderson [Sat, 31 Jul 2021 05:56:06 +0000 (19:56 -1000)]
target/alpha: Reorg fp memory operations

Pass in the context to each mini-helper, instead of an
incorrectly named "flags".  Separate gen_load_fp and
gen_store_fp, away from the integer helpers.

Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
2 years agoMerge remote-tracking branch 'remotes/jsnow/tags/python-pull-request' into staging
Richard Henderson [Tue, 12 Oct 2021 23:08:33 +0000 (16:08 -0700)]
Merge remote-tracking branch 'remotes/jsnow/tags/python-pull-request' into staging

Pull request

# gpg: Signature made Tue 12 Oct 2021 02:36:07 PM PDT
# gpg:                using RSA key F9B7ABDBBCACDF95BE76CBD07DEF8106AAFC390E
# gpg: Good signature from "John Snow (John Huston) <jsnow@redhat.com>" [full]

* remotes/jsnow/tags/python-pull-request:
  python, iotests: remove socket_scm_helper
  python/qmp: add send_fd_scm directly to QEMUMonitorProtocol
  python/qmp: clear events on get_events() call
  python/aqmp: Disable logging messages by default
  python/aqmp: Reduce severity of EOFError-caused loop terminations
  python/aqmp: Add dict conversion method to Greeting object
  python/aqmp: add send_fd_scm
  python/aqmp: Return cleared events from EventListener.clear()
  python/aqmp: add .empty() method to EventListener
  python/aqmp: add greeting property to QMPClient

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
2 years agoMerge remote-tracking branch 'remotes/rth/tags/pull-hppa-20211012' into staging
Richard Henderson [Tue, 12 Oct 2021 16:37:19 +0000 (09:37 -0700)]
Merge remote-tracking branch 'remotes/rth/tags/pull-hppa-20211012' into staging

seabios-hppa update

# gpg: Signature made Tue 12 Oct 2021 09:28:12 AM PDT
# 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-hppa-20211012:
  pc-bios: Update hppa-firmware.img

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
2 years agopc-bios: Update hppa-firmware.img
Helge Deller [Fri, 24 Sep 2021 19:53:27 +0000 (21:53 +0200)]
pc-bios: Update hppa-firmware.img

Update SeaBIOS to seabios-hppa-v2

Changes in seabios-hppa:
* Include all latest upstream SeaBIOS patches
* add support for the qemu "bootindex" parameter
* add support for the qemu "-boot order=g-m" parameter to choose
  SCSI ID

Signed-off-by: Helge Deller <deller@gmx.de>
Message-Id: <YU4st/zcLcg6RKNn@ls3530>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
2 years agopython, iotests: remove socket_scm_helper
John Snow [Thu, 23 Sep 2021 00:49:31 +0000 (20:49 -0400)]
python, iotests: remove socket_scm_helper

It's not used anymore, now.

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Hanna Reitz <hreitz@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Message-id: 20210923004938.3999963-11-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
2 years agopython/qmp: add send_fd_scm directly to QEMUMonitorProtocol
John Snow [Thu, 23 Sep 2021 00:49:30 +0000 (20:49 -0400)]
python/qmp: add send_fd_scm directly to QEMUMonitorProtocol

It turns out you can do this directly from Python ... and because of
this, you don't need to worry about setting the inheritability of the
fds or spawning another process.

Doing this is helpful because it allows QEMUMonitorProtocol to keep its
file descriptor and socket object as private implementation
details. /that/ is helpful in turn because it allows me to write a
compatible, alternative implementation.

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Hanna Reitz <hreitz@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Message-id: 20210923004938.3999963-10-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
2 years agopython/qmp: clear events on get_events() call
John Snow [Thu, 23 Sep 2021 00:49:29 +0000 (20:49 -0400)]
python/qmp: clear events on get_events() call

All callers in the tree *already* clear the events after a call to
get_events(). Do it automatically instead and update callsites to remove
the manual clear call.

These semantics are quite a bit easier to emulate with async QMP, and
nobody appears to be abusing some emergent properties of what happens if
you decide not to clear them, so let's dial down to the dumber, simpler
thing.

Specifically: callers of clear() right after a call to get_events() are
more likely expressing their desire to not see any events they just
retrieved, whereas callers of clear_events() not in relation to a recent
call to pull_event/get_events are likely expressing their desire to
simply drop *all* pending events straight onto the floor. In the sync
world, this is safe enough; in the async world it's nearly impossible to
promise that nothing happens between getting and clearing the
events.

Making the retrieval also clear the queue is vastly simpler.

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Hanna Reitz <hreitz@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Message-id: 20210923004938.3999963-9-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
2 years agopython/aqmp: Disable logging messages by default
John Snow [Thu, 23 Sep 2021 00:49:28 +0000 (20:49 -0400)]
python/aqmp: Disable logging messages by default

AQMP is a library, and ideally it should not print error diagnostics
unless a user opts into seeing them. By default, Python will print all
WARNING, ERROR or CRITICAL messages to screen if no logging
configuration has been created by a client application.

In AQMP's case, ERROR logging statements are used to report additional
detail about runtime failures that will also eventually be reported to the
client library via an Exception, so these messages should not be
rendered by default.

(Why bother to have them at all, then? In async contexts, there may be
multiple Exceptions and we are only able to report one of them back to
the client application. It is not reasonably easy to predict ahead of
time if one or more of these Exceptions will be squelched. Therefore,
it's useful to log intermediate failures to help make sense of the
ultimate, resulting failure.)

Add a NullHandler that will suppress these messages until a client
application opts into logging via logging.basicConfig or similar. Note
that upon calling basicConfig(), this handler will *not* suppress these
messages from being displayed by the client's configuration.

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 20210923004938.3999963-8-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
2 years agopython/aqmp: Reduce severity of EOFError-caused loop terminations
John Snow [Thu, 23 Sep 2021 00:49:27 +0000 (20:49 -0400)]
python/aqmp: Reduce severity of EOFError-caused loop terminations

When we encounter an EOFError, we don't know if it's an "error" in the
perspective of the user of the library yet. Therefore, we should not log
it as an error. Reduce the severity of this logging message to "INFO" to
indicate that it's something that we expect to occur during the normal
operation of the library.

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 20210923004938.3999963-7-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
2 years agopython/aqmp: Add dict conversion method to Greeting object
John Snow [Thu, 23 Sep 2021 00:49:26 +0000 (20:49 -0400)]
python/aqmp: Add dict conversion method to Greeting object

The iotests interface expects to return the greeting as a dict; AQMP
offers it as a rich object.

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 20210923004938.3999963-6-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
2 years agopython/aqmp: add send_fd_scm
John Snow [Thu, 23 Sep 2021 00:49:25 +0000 (20:49 -0400)]
python/aqmp: add send_fd_scm

Add an implementation for send_fd_scm to the async QMP implementation.
Like socket_scm_helper mentions, a non-empty payload is required for
QEMU to process the ancillary data. A space is most useful because it
does not disturb the parsing of subsequent JSON objects.

A note on "voiding the warranty":

Python 3.11 removes support for calling sendmsg directly from a
transport's socket. There is no other interface for doing this, our use
case is, I suspect, "quite unique".

As far as I can tell, this is safe to do -- send_fd_scm is a synchronous
function and we can be guaranteed that the async coroutines will *not* be
running when it is invoked. In testing, it works correctly.

I investigated quite thoroughly the possibility of creating my own
asyncio Transport (The class that ultimately manages the raw socket
object) so that I could manage the socket myself, but this is so wildly
invasive and unportable I scrapped the idea. It would involve a lot of
copy-pasting of various python utilities and classes just to re-create
the same infrastructure, and for extremely little benefit. Nah.

Just boldly void the warranty instead, while I try to follow up on
https://bugs.python.org/issue43232

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 20210923004938.3999963-5-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
2 years agopython/aqmp: Return cleared events from EventListener.clear()
John Snow [Thu, 23 Sep 2021 00:49:24 +0000 (20:49 -0400)]
python/aqmp: Return cleared events from EventListener.clear()

This serves two purposes:

(1) It is now possible to discern whether or not clear() removed any
event(s) from the queue with absolute certainty, and

(2) It is now very easy to get a List of all pending events in one
chunk, which is useful for the sync bridge.

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Hanna Reitz <hreitz@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Message-id: 20210923004938.3999963-4-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
2 years agopython/aqmp: add .empty() method to EventListener
John Snow [Thu, 23 Sep 2021 00:49:23 +0000 (20:49 -0400)]
python/aqmp: add .empty() method to EventListener

Synchronous clients may want to know if they're about to block waiting
for an event or not. A method such as this is necessary to implement a
compatible interface for the old QEMUMonitorProtocol using the new async
internals.

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Hanna Reitz <hreitz@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Message-id: 20210923004938.3999963-3-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
2 years agopython/aqmp: add greeting property to QMPClient
John Snow [Thu, 23 Sep 2021 00:49:22 +0000 (20:49 -0400)]
python/aqmp: add greeting property to QMPClient

Expose the greeting as a read-only property of QMPClient so it can be
retrieved at-will.

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Hanna Reitz <hreitz@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Message-id: 20210923004938.3999963-2-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
2 years agoMerge remote-tracking branch 'remotes/stsquad/tags/pull-for-6.2-121021-2' into staging
Richard Henderson [Tue, 12 Oct 2021 13:16:25 +0000 (06:16 -0700)]
Merge remote-tracking branch 'remotes/stsquad/tags/pull-for-6.2-121021-2' into staging

Some testing and plugin updates:

  - don't override the test compiler when specified
  - split some multiarch tests by guest OS
  - add riscv64 docker image and cross-compile tests
  - drop release tarball test from Travis
  - skip check-patch on master repo
  - fix passing of TEST_TARGETS to cirrus
  - fix missing symbols in plugins
  - ensure s390x insn start ops precede plugin instrumentation
  - refactor plugin instruction boundary detection
  - update github repo lockdown
  - add a debian-native test image for multi-arch builds

# gpg: Signature made Tue 12 Oct 2021 02:35:00 AM PDT
# gpg:                using RSA key 6685AE99E75167BCAFC8DF35FBD0DB095A9E2A44
# gpg: Good signature from "Alex Bennée (Master Work Key) <alex.bennee@linaro.org>" [full]

* remotes/stsquad/tags/pull-for-6.2-121021-2:
  tests/docker: add a debian-native image and make available
  .github: move repo lockdown to the v2 configuration
  accel/tcg: re-factor plugin_inject_cb so we can assert insn_idx is valid
  target/s390x: move tcg_gen_insn_start to s390x_tr_insn_start
  plugins/: Add missing functions to symbol list
  gitlab: fix passing of TEST_TARGETS env to cirrus
  gitlab: skip the check-patch job on the upstream repo
  travis.yml: Remove the "Release tarball" job
  gitlab: Add cross-riscv64-system, cross-riscv64-user
  tests/docker: promote debian-riscv64-cross to a full image
  tests/tcg: move some multiarch files and make conditional
  tests/tcg/sha1: remove endian include
  configure: don't override the selected host test compiler if defined

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
2 years agoMerge remote-tracking branch 'remotes/clg/tags/pull-aspeed-20211012' into staging
Richard Henderson [Tue, 12 Oct 2021 11:41:59 +0000 (04:41 -0700)]
Merge remote-tracking branch 'remotes/clg/tags/pull-aspeed-20211012' into staging

Aspeed patches :

* I2C QOMify (Cedric)
* SMC model cleanup and QOMify (Cedric)
* ADC model (Peter and Andrew)
* GPIO fixes (Peter)

# gpg: Signature made Tue 12 Oct 2021 12:36:22 AM PDT
# gpg:                using RSA key A0F66548F04895EBFE6B0B6051A343C7CFFBECA1
# gpg: Good signature from "Cédric Le Goater <clg@kaod.org>" [marginal]
# gpg: WARNING: This key is not certified with sufficiently trusted signatures!
# gpg:          It is not certain that the signature belongs to the owner.
# Primary key fingerprint: A0F6 6548 F048 95EB FE6B  0B60 51A3 43C7 CFFB ECA1

* remotes/clg/tags/pull-aspeed-20211012:
  aspeed/smc: Dump address offset in trace events
  aspeed/wdt: Add trace events
  hw/arm: Integrate ADC model into Aspeed SoC
  hw/adc: Add basic Aspeed ADC model
  hw: aspeed_gpio: Fix GPIO array indexing
  hw: aspeed_gpio: Fix pin I/O type declarations
  aspeed/i2c: QOMify AspeedI2CBus
  aspeed/smc: Remove unused attribute 'irqline'
  aspeed/smc: Introduce a new addr_width() class handler
  aspeed/smc: Add default reset values
  aspeed/smc: QOMify AspeedSMCFlash
  aspeed/smc: Rename AspeedSMCFlash 'id' to 'cs'
  aspeed/smc: Remove the 'size' attribute from AspeedSMCFlash
  aspeed/smc: Remove the 'flash' attribute from AspeedSMCFlash
  aspeed/smc: Drop AspeedSMCController structure
  aspeed/smc: Stop using the model name for the memory regions
  aspeed/smc: Introduce aspeed_smc_error() helper
  aspeed/smc: Add watchdog Control/Status Registers

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
2 years agotests/docker: add a debian-native image and make available
Alex Bennée [Wed, 22 Sep 2021 15:15:27 +0000 (16:15 +0100)]
tests/docker: add a debian-native image and make available

This image is intended for building whatever the native versions of
QEMU are for the host architecture. This will hopefully be an aid for
3rd parties who want to be able to build QEMU themselves without
redoing all the dependencies themselves.

We disable the registry because we currently don't have multi-arch
support there.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Tested-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Tested-by: Anders Roxell <anders.roxell@linaro.org>
Acked-by: Willian Rampazzo <willianr@redhat.com>
Message-Id: <20210922151528.2192966-1-alex.bennee@linaro.org>

2 years ago.github: move repo lockdown to the v2 configuration
Alex Bennée [Mon, 4 Oct 2021 15:43:08 +0000 (16:43 +0100)]
.github: move repo lockdown to the v2 configuration

I was getting prompted by GitHub for new permissions but it turns out
per https://github.com/dessant/repo-lockdown/issues/6:

  Repo Lockdown has been rewritten for GitHub Actions, offering new
  features and better control over your automation presets. The legacy
  GitHub App has been deprecated, and the public instance of the app
  has been shut down.

So this is what I've done. As the issues tab is disabled I've removed
the handling for issues from the new version.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Willian Rampazzo <willianr@redhat.com>
Message-Id: <20211004154308.2114870-1-alex.bennee@linaro.org>

2 years agoaccel/tcg: re-factor plugin_inject_cb so we can assert insn_idx is valid
Alex Bennée [Fri, 17 Sep 2021 16:23:32 +0000 (17:23 +0100)]
accel/tcg: re-factor plugin_inject_cb so we can assert insn_idx is valid

Coverity doesn't know enough about how we have arranged our plugin TCG
ops to know we will always have incremented insn_idx before injecting
the callback. Let us assert it for the benefit of Coverity and protect
ourselves from accidentally breaking the assumption and triggering
harder to grok errors deeper in the code if we attempt a negative
indexed array lookup.

However to get to this point we re-factor the code and remove the
second hand instruction boundary detection in favour of scanning the
full set of ops and using the existing INDEX_op_insn_start to cleanly
detect when the instruction has started. As we no longer need the
plugin specific list of ops we delete that.

My initial benchmarks shows no discernible impact of dropping the
plugin specific ops list.

Fixes: Coverity 1459509
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Cc: Peter Maydell <peter.maydell@linaro.org>
Message-Id: <20210917162332.3511179-12-alex.bennee@linaro.org>

2 years agotarget/s390x: move tcg_gen_insn_start to s390x_tr_insn_start
Richard Henderson [Mon, 11 Oct 2021 18:53:32 +0000 (11:53 -0700)]
target/s390x: move tcg_gen_insn_start to s390x_tr_insn_start

We use INDEX_op_insn_start to make the start of instruction boundaries.
If we don't do it in the .insn_start hook things get confused especially
now plugins want to use that marking to identify the start of instructions
and will bomb out if it sees instrumented ops before the first instruction
boundary.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20211011185332.166763-1-richard.henderson@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
2 years agoplugins/: Add missing functions to symbol list
Lukas Jünger [Fri, 17 Sep 2021 16:23:31 +0000 (17:23 +0100)]
plugins/: Add missing functions to symbol list

Some functions of the plugin API were missing in
the symbol list. However, they are all used by
the contributed example plugins. QEMU fails to
load the plugin if the function symbol is not
exported.

Signed-off-by: Lukas Jünger <lukas.junger@greensocs.com>
Message-Id: <20210905140939.638928-2-lukas.junger@greensocs.com>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20210917162332.3511179-11-alex.bennee@linaro.org>

2 years agogitlab: fix passing of TEST_TARGETS env to cirrus
Daniel P. Berrangé [Fri, 17 Sep 2021 16:23:30 +0000 (17:23 +0100)]
gitlab: fix passing of TEST_TARGETS env to cirrus

A typo meant the substitution would not work, and the placeholder in the
target file didn't even exist.

The result was that tests were never run on the FreeBSD and macOS jobs,
only a basic build.

Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Acked-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Willian Rampazzo <willianr@redhat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20210915125452.1704899-3-berrange@redhat.com>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20210917162332.3511179-10-alex.bennee@linaro.org>

2 years agogitlab: skip the check-patch job on the upstream repo
Daniel P. Berrangé [Fri, 17 Sep 2021 16:23:29 +0000 (17:23 +0100)]
gitlab: skip the check-patch job on the upstream repo

The check-patch job is intended to be used by contributors or
subsystem maintainers to see if there are style mistakes. The
false positive rate is too high to be used in a gating scenario
so should not run it on the upstream repo ever.

Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Acked-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20210915125452.1704899-2-berrange@redhat.com>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20210917162332.3511179-9-alex.bennee@linaro.org>

2 years agotravis.yml: Remove the "Release tarball" job
Thomas Huth [Fri, 17 Sep 2021 16:23:28 +0000 (17:23 +0100)]
travis.yml: Remove the "Release tarball" job

This is a leftover from the days when we were using Travis excessively,
but since x86 jobs are not really usable there anymore, this job has
likely never been used since many months. Let's simply remove it now.

Signed-off-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20210917094826.466047-1-thuth@redhat.com>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20210917162332.3511179-8-alex.bennee@linaro.org>

2 years agogitlab: Add cross-riscv64-system, cross-riscv64-user
Richard Henderson [Fri, 17 Sep 2021 16:23:26 +0000 (17:23 +0100)]
gitlab: Add cross-riscv64-system, cross-riscv64-user

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20210914185830.1378771-3-richard.henderson@linaro.org>
[AJB: add allow_failure]
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Willian Rampazzo <willianr@redhat.com>
Message-Id: <20210917162332.3511179-6-alex.bennee@linaro.org>

2 years agotests/docker: promote debian-riscv64-cross to a full image
Alex Bennée [Fri, 17 Sep 2021 16:23:25 +0000 (17:23 +0100)]
tests/docker: promote debian-riscv64-cross to a full image

To be able to cross build QEMU itself we need to include a few more
libraries. These are only available in Debian's unstable ports repo
for now so we need to base the riscv64 image on sid with the the
minimal libs needed to build QEMU (glib/pixman).

The result works but is not as clean as using build-dep to bring in
more dependencies. However sid is by definition a shifting pile of
sand and by keeping the list of libs minimal we reduce the chance of
having an image we can't build. It's good enough for a basic cross
build testing of TCG.

Cc: "Daniel P. Berrangé" <berrange@redhat.com>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20210914185830.1378771-2-richard.henderson@linaro.org>
[AJB: tweak allow_failure]
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20210917162332.3511179-5-alex.bennee@linaro.org>

2 years agotests/tcg: move some multiarch files and make conditional
Alex Bennée [Fri, 17 Sep 2021 16:23:24 +0000 (17:23 +0100)]
tests/tcg: move some multiarch files and make conditional

We had some messy code to filter out stuff we can't build. Lets junk
that and simplify the logic by pushing some stuff into subdirs. In
particular we move:

  float_helpers into libs - not a standalone test
  linux-test into linux - so we only build on Linux hosts

This allows for at least some of the tests to be nominally usable
by *BSD user builds.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Cc: Warner Losh <imp@bsdimp.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Warner Losh <imp@bsdimp.com>
Message-Id: <20210917162332.3511179-4-alex.bennee@linaro.org>

2 years agotests/tcg/sha1: remove endian include
Alex Bennée [Fri, 17 Sep 2021 16:23:23 +0000 (17:23 +0100)]
tests/tcg/sha1: remove endian include

This doesn't exist in BSD world and doesn't seem to be needed by
either.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Warner Losh <imp@bsdimp.com>
Message-Id: <20210917162332.3511179-3-alex.bennee@linaro.org>

2 years agoconfigure: don't override the selected host test compiler if defined
Alex Bennée [Fri, 17 Sep 2021 16:23:22 +0000 (17:23 +0100)]
configure: don't override the selected host test compiler if defined

There are not many cases you would want to do this but one is if you
want to use a test friendly compiler like gcc instead of a system
compiler like clang. Either way we should honour the users choice if
they have made it.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Cc: Warner Losh <imp@bsdimp.com>
Reviewed-by: Warner Losh <imp@bsdimp.com>
Message-Id: <20210917162332.3511179-2-alex.bennee@linaro.org>

2 years agoaspeed/smc: Dump address offset in trace events
Cédric Le Goater [Tue, 12 Oct 2021 06:20:08 +0000 (08:20 +0200)]
aspeed/smc: Dump address offset in trace events

The register index is currently printed and this is confusing.

Reviewed-by: Francisco Iglesias <frasse.iglesias@gmail.com>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
2 years agoaspeed/wdt: Add trace events
Cédric Le Goater [Tue, 12 Oct 2021 06:20:08 +0000 (08:20 +0200)]
aspeed/wdt: Add trace events

Reviewed-by: Francisco Iglesias <frasse.iglesias@gmail.com>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
2 years agohw/arm: Integrate ADC model into Aspeed SoC
Andrew Jeffery [Tue, 12 Oct 2021 06:20:08 +0000 (08:20 +0200)]
hw/arm: Integrate ADC model into Aspeed SoC

Signed-off-by: Andrew Jeffery <andrew@aj.id.au>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
Signed-off-by: Peter Delevoryas <pdel@fb.com>
Message-Id: <20211005052604.1674891-3-pdel@fb.com>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
2 years agohw/adc: Add basic Aspeed ADC model
Andrew Jeffery [Tue, 12 Oct 2021 06:20:08 +0000 (08:20 +0200)]
hw/adc: Add basic Aspeed ADC model

This model implements enough behaviour to do basic functionality tests
such as device initialisation and read out of dummy sample values. The
sample value generation strategy is similar to the STM ADC already in
the tree.

Signed-off-by: Andrew Jeffery <andrew@aj.id.au>
[clg : support for multiple engines (AST2600) ]
Signed-off-by: Cédric Le Goater <clg@kaod.org>
[pdel : refactored engine register struct fields to regs[] array field]
[pdel : added guest-error checking for upper-8 channel regs in AST2600]
[pdel : allow 16-bit reads of the channel data registers]
Signed-off-by: Peter Delevoryas <pdel@fb.com>
Message-Id: <20211005052604.1674891-2-pdel@fb.com>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
2 years agohw: aspeed_gpio: Fix GPIO array indexing
Peter Delevoryas [Tue, 12 Oct 2021 06:20:08 +0000 (08:20 +0200)]
hw: aspeed_gpio: Fix GPIO array indexing

The gpio array is declared as a dense array:

  qemu_irq gpios[ASPEED_GPIO_NR_PINS];

(AST2500 has 228, AST2400 has 216, AST2600 has 208)

However, this array is used like a matrix of GPIO sets
(e.g. gpio[NR_SETS][NR_PINS_PER_SET] = gpio[8][32])

  size_t offset = set * GPIOS_PER_SET + gpio;
  qemu_set_irq(s->gpios[offset], !!(new & mask));

This can result in an out-of-bounds access to "s->gpios" because the
gpio sets do _not_ have the same length. Some of the groups (e.g.
GPIOAB) only have 4 pins. 228 != 8 * 32 == 256.

To fix this, I converted the gpio array from dense to sparse, to that
match both the hardware layout and this existing indexing code.

Fixes: 4b7f956862dc2db4c5c ("hw/gpio: Add basic Aspeed GPIO model for AST2400 and AST2500")
Signed-off-by: Peter Delevoryas <pdel@fb.com>
Message-Id: <20211008033501.934729-2-pdel@fb.com>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
2 years agohw: aspeed_gpio: Fix pin I/O type declarations
Peter Delevoryas [Tue, 12 Oct 2021 06:20:08 +0000 (08:20 +0200)]
hw: aspeed_gpio: Fix pin I/O type declarations

Some of the pin declarations in the Aspeed GPIO module were incorrect,
probably because of confusion over which bits in the input and output
uint32_t's correspond to which groups in the label array. Since the
uint32_t literals are in big endian, it's sort of the opposite of what
would be intuitive. The least significant bit in ast2500_set_props[6]
corresponds to GPIOY0, not GPIOAB7.

GPIOxx indicates input and output capabilities, GPIxx indicates only
input, GPOxx indicates only output.

AST2500:
- Previously had GPIW0..GPIW7 and GPIX0..GPIX7, that's correct.
- Previously had GPIOY0..GPIOY3, should have been GPIOY0..GPIOY7.
- Previously had GPIOAB0..GPIOAB3 and GPIAB4..GPIAB7, should only have
  been GPIOAB0..GPIOAB3.

AST2600:
- GPIOT0..GPIOT7 should have been GPIT0..GPIT7.
- GPIOU0..GPIOU7 should have been GPIU0..GPIU7.
- GPIW0..GPIW7 should have been GPIOW0..GPIOW7.
- GPIOY0..GPIOY7 and GPIOZ0...GPIOZ7 were disabled.

Fixes: 4b7f956862dc2db4c5c ("hw/gpio: Add basic Aspeed GPIO model for AST2400 and AST2500")
Fixes: 36d737ee82b2972167e ("hw/gpio: Add in AST2600 specific implementation")
Signed-off-by: Peter Delevoryas <pdel@fb.com>
Reviewed-by: Damien Hedde <damien.hedde@greensocs.com>
Reviewed-by: Rashmica Gupta <rashmica.g@gmail.com>
Message-Id: <20210928032456.3192603-2-pdel@fb.com>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
2 years agoaspeed/i2c: QOMify AspeedI2CBus
Cédric Le Goater [Tue, 12 Oct 2021 06:20:08 +0000 (08:20 +0200)]
aspeed/i2c: QOMify AspeedI2CBus

Introduce an AspeedI2CBus SysBusDevice model and attach the associated
memory region and IRQ to the newly instantiated objects.

Before this change, the I2C bus IRQs were all attached to the
SysBusDevice model of the I2C controller. Adapt the AST2600 SoC
realize routine to take into account this change.

Signed-off-by: Cédric Le Goater <clg@kaod.org>
2 years agoaspeed/smc: Remove unused attribute 'irqline'
Cédric Le Goater [Tue, 12 Oct 2021 06:20:08 +0000 (08:20 +0200)]
aspeed/smc: Remove unused attribute 'irqline'

Signed-off-by: Cédric Le Goater <clg@kaod.org>
2 years agoaspeed/smc: Introduce a new addr_width() class handler
Cédric Le Goater [Tue, 12 Oct 2021 06:20:08 +0000 (08:20 +0200)]
aspeed/smc: Introduce a new addr_width() class handler

The AST2400 SPI controller has a transitional HW interface and it
stores the address width currently in use in a different register than
all the other SMC controllers. It needs special handling when working
in 4B mode.

Make it clear through a class handler. This also removes another use
of the segments array.

Signed-off-by: Cédric Le Goater <clg@kaod.org>
2 years agoaspeed/smc: Add default reset values
Cédric Le Goater [Tue, 12 Oct 2021 06:20:08 +0000 (08:20 +0200)]
aspeed/smc: Add default reset values

This simplifies the reset handler and has the benefit to remove some
"bad" use of the segments array as an identifier of the controller model.

Signed-off-by: Cédric Le Goater <clg@kaod.org>
2 years agoaspeed/smc: QOMify AspeedSMCFlash
Cédric Le Goater [Tue, 12 Oct 2021 06:20:08 +0000 (08:20 +0200)]
aspeed/smc: QOMify AspeedSMCFlash

AspeedSMCFlash is a small structure representing the AHB memory window
through which the contents of a flash device can be accessed with MMIOs.

Introduce an AspeedSMCFlash SysBusDevice model and attach the associated
memory region to the newly instantiated objects.

Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
2 years agoaspeed/smc: Rename AspeedSMCFlash 'id' to 'cs'
Cédric Le Goater [Tue, 12 Oct 2021 06:20:08 +0000 (08:20 +0200)]
aspeed/smc: Rename AspeedSMCFlash 'id' to 'cs'

'cs' is a more appropriate name to index SPI flash devices.

Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
2 years agoaspeed/smc: Remove the 'size' attribute from AspeedSMCFlash
Cédric Le Goater [Tue, 12 Oct 2021 06:20:08 +0000 (08:20 +0200)]
aspeed/smc: Remove the 'size' attribute from AspeedSMCFlash

AspeedSMCFlash::size is only used to compute the initial size of the
boot_rom region. Not very useful, so directly call memory_region_size()
instead.

Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
2 years agoaspeed/smc: Remove the 'flash' attribute from AspeedSMCFlash
Cédric Le Goater [Tue, 12 Oct 2021 06:20:08 +0000 (08:20 +0200)]
aspeed/smc: Remove the 'flash' attribute from AspeedSMCFlash

There is no need to keep a reference of the flash qdev in the AspeedSMCFlash
state: the SPI bus takes ownership and will release its resources. Remove
AspeedSMCFlash::flash.

Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
2 years agoaspeed/smc: Drop AspeedSMCController structure
Cédric Le Goater [Tue, 12 Oct 2021 06:20:08 +0000 (08:20 +0200)]
aspeed/smc: Drop AspeedSMCController structure

The characteristics of the Aspeed controllers are described in a
AspeedSMCController structure which is redundant with the
AspeedSMCClass. Move all attributes under the class and adapt the code
to use class attributes instead.

This is a large change but it is functionally equivalent.

Signed-off-by: Cédric Le Goater <clg@kaod.org>
2 years agoaspeed/smc: Stop using the model name for the memory regions
Cédric Le Goater [Tue, 12 Oct 2021 06:20:08 +0000 (08:20 +0200)]
aspeed/smc: Stop using the model name for the memory regions

There is no real reason to use this name. It's simply nice to have in
the monitor output but it's a burden for the following patch which
removes the AspeedSMCController structure describing the controller.

Signed-off-by: Cédric Le Goater <clg@kaod.org>
2 years agoaspeed/smc: Introduce aspeed_smc_error() helper
Cédric Le Goater [Tue, 12 Oct 2021 06:20:07 +0000 (08:20 +0200)]
aspeed/smc: Introduce aspeed_smc_error() helper

It unifies the errors reported by the Aspeed SMC model and also
removes some use of ctrl->name which will help us for the next
patches.

Signed-off-by: Cédric Le Goater <clg@kaod.org>
2 years agoaspeed/smc: Add watchdog Control/Status Registers
Cédric Le Goater [Tue, 12 Oct 2021 06:20:07 +0000 (08:20 +0200)]
aspeed/smc: Add watchdog Control/Status Registers

The Aspeed SoCs have a dual boot function for firmware fail-over
recovery. The system auto-reboots from the second flash if the main
flash does not boot successfully within a certain amount of time. This
function is called alternate boot (ABR) in the FMC controllers.

On AST2400/AST2500, ABR is enabled by hardware strapping in SCU70 to
enable the 2nd watchdog timer, on AST2600, through register SCU510.
If the boot on the the main flash succeeds, the firmware should
disable the 2nd watchdog timer. If not, the BMC is reset and the CE0
and CE1 mappings are swapped to restart the BMC from the 2nd flash.

On the AST2600, the ABR registers controlling the 2nd watchdog timer
were moved from the watchdog register to the FMC controller and the
FMC model should be able to control WDT2 through its own register set.
This requires more work. For now, add dummy read/write handlers to let
the FW disable the 2nd watchdog without error.

Reviewed-by: Peter Delevoryas <pdel@fb.com>
Reported-by: Peter Delevoryas <pdel@fb.com>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
2 years agoMerge remote-tracking branch 'remotes/stefanha/tags/block-pull-request' into staging
Richard Henderson [Mon, 11 Oct 2021 15:15:32 +0000 (08:15 -0700)]
Merge remote-tracking branch 'remotes/stefanha/tags/block-pull-request' into staging

Pull request

# gpg: Signature made Mon 11 Oct 2021 05:40:35 AM PDT
# gpg:                using RSA key 8695A8BFD3F97CDAAC35775A9CA4ABB381AB73C8
# gpg: Good signature from "Stefan Hajnoczi <stefanha@redhat.com>" [full]
# gpg:                 aka "Stefan Hajnoczi <stefanha@gmail.com>" [full]

* remotes/stefanha/tags/block-pull-request:
  iothread: use IOThreadParamInfo in iothread_[set|get]_param()
  iothread: rename PollParamInfo to IOThreadParamInfo

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
2 years agoMerge remote-tracking branch 'remotes/vivier-m68k/tags/m68k-next-pull-request' into...
Richard Henderson [Fri, 8 Oct 2021 12:02:19 +0000 (05:02 -0700)]
Merge remote-tracking branch 'remotes/vivier-m68k/tags/m68k-next-pull-request' into staging

Pull request q800 20211008

macfb: fixes for booting MacOS

# gpg: Signature made Fri 08 Oct 2021 04:44:44 AM PDT
# 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]

* remotes/vivier-m68k/tags/m68k-next-pull-request:
  q800: wire macfb IRQ to separate video interrupt on VIA2
  macfb: add vertical blank interrupt
  macfb: fix 24-bit RGB pixel encoding
  macfb: fix up 1-bit pixel encoding
  macfb: add common monitor modes supported by the MacOS toolbox ROM
  macfb: add qdev property to specify display type
  macfb: implement mode sense to allow display type to be detected
  macfb: add trace events for reading and writing the control registers
  macfb: use memory_region_init_ram() in macfb_common_realize() for the framebuffer
  macfb: fix overflow of color_palette array
  macfb: fix invalid object reference in macfb_common_realize()
  macfb: update macfb.c to use the Error API best practices
  macfb: handle errors that occur during realize

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
2 years agoq800: wire macfb IRQ to separate video interrupt on VIA2
Mark Cave-Ayland [Thu, 7 Oct 2021 22:12:53 +0000 (23:12 +0100)]
q800: wire macfb IRQ to separate video interrupt on VIA2

Whilst the in-built Quadra 800 framebuffer exists within the Nubus address
space for slot 9, it has its own dedicated interrupt on VIA2. Force the
macfb device to occupy slot 9 in the q800 machine and wire its IRQ to the
separate video interrupt since this is what is expected by the MacOS
interrupt handler.

Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20211007221253.29024-14-mark.cave-ayland@ilande.co.uk>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
2 years agomacfb: add vertical blank interrupt
Mark Cave-Ayland [Thu, 7 Oct 2021 22:12:52 +0000 (23:12 +0100)]
macfb: add vertical blank interrupt

The MacOS driver expects a 60.15Hz vertical blank interrupt to be generated by
the framebuffer which in turn schedules the mouse driver via the Vertical Retrace
Manager.

Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20211007221253.29024-13-mark.cave-ayland@ilande.co.uk>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
2 years agomacfb: fix 24-bit RGB pixel encoding
Mark Cave-Ayland [Thu, 7 Oct 2021 22:12:51 +0000 (23:12 +0100)]
macfb: fix 24-bit RGB pixel encoding

According to Apple Technical Note HW26: "Macintosh Quadra Built-In Video" the
in-built framebuffer encodes each 24-bit pixel into 4 bytes. Adjust the 24-bit
RGB pixel encoding accordingly which agrees with the encoding expected by MacOS
when changing into 24-bit colour mode.

Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20211007221253.29024-12-mark.cave-ayland@ilande.co.uk>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
2 years agomacfb: fix up 1-bit pixel encoding
Mark Cave-Ayland [Thu, 7 Oct 2021 22:12:50 +0000 (23:12 +0100)]
macfb: fix up 1-bit pixel encoding

The MacOS driver expects the RGB values for the pixel to be in entries 0 and 1
of the colour palette.

Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20211007221253.29024-11-mark.cave-ayland@ilande.co.uk>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
2 years agomacfb: add common monitor modes supported by the MacOS toolbox ROM
Mark Cave-Ayland [Thu, 7 Oct 2021 22:12:49 +0000 (23:12 +0100)]
macfb: add common monitor modes supported by the MacOS toolbox ROM

The monitor modes table is found by experimenting with the Monitors Control
Panel in MacOS and analysing the reads/writes. From this it can be found that
the mode is controlled by writes to the DAFB_MODE_CTRL1 and DAFB_MODE_CTRL2
registers.

Implement the first block of DAFB registers as a register array including the
existing sense register, the newly discovered control registers above, and also
the DAFB_MODE_VADDR1 and DAFB_MODE_VADDR2 registers which are used by NetBSD to
determine the current video mode.

These experiments also show that the offset of the start of video RAM and the
stride can change depending upon the monitor mode, so update macfb_draw_graphic()
and both the BI_MAC_VADDR and BI_MAC_VROW bootinfo for the q800 machine
accordingly.

Finally update macfb_common_realize() so that only the resolution and depth
supported by the display type can be specified on the command line, and add an
error hint showing the list of supported resolutions and depths if the user tries
to specify an invalid display mode.

Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20211007221253.29024-10-mark.cave-ayland@ilande.co.uk>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
2 years agomacfb: add qdev property to specify display type
Mark Cave-Ayland [Thu, 7 Oct 2021 22:12:48 +0000 (23:12 +0100)]
macfb: add qdev property to specify display type

Since the available resolutions and colour depths are determined by the attached
display type, add a qdev property to allow the display type to be specified.

The main resolutions of interest are high resolution 1152x870 with 8-bit colour
and SVGA resolution up to 800x600 with 24-bit colour so update the q800 machine
to allow high resolution mode if specified and otherwise fall back to SVGA.

Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20211007221253.29024-9-mark.cave-ayland@ilande.co.uk>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
2 years agomacfb: implement mode sense to allow display type to be detected
Mark Cave-Ayland [Thu, 7 Oct 2021 22:12:47 +0000 (23:12 +0100)]
macfb: implement mode sense to allow display type to be detected

The MacOS toolbox ROM uses the monitor sense to detect the display type and then
offer a fixed set of resolutions and colour depths accordingly. Implement the
monitor sense using information found in Apple Technical Note HW26: "Macintosh
Quadra Built-In Video" along with some local experiments.

Since the default configuration is 640 x 480 with 8-bit colour then hardcode
the sense register to return MACFB_DISPLAY_VGA for now.

Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20211007221253.29024-8-mark.cave-ayland@ilande.co.uk>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
2 years agomacfb: add trace events for reading and writing the control registers
Mark Cave-Ayland [Thu, 7 Oct 2021 22:12:46 +0000 (23:12 +0100)]
macfb: add trace events for reading and writing the control registers

Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20211007221253.29024-7-mark.cave-ayland@ilande.co.uk>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
2 years agomacfb: use memory_region_init_ram() in macfb_common_realize() for the framebuffer
Mark Cave-Ayland [Thu, 7 Oct 2021 22:12:45 +0000 (23:12 +0100)]
macfb: use memory_region_init_ram() in macfb_common_realize() for the framebuffer

Currently macfb_common_realize() defines the framebuffer RAM memory region as
being non-migrateable but then immediately registers it for migration. Replace
memory_region_init_ram_nomigrate() with memory_region_init_ram() which is clearer
and does exactly the same thing.

Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: BALATON Zoltan <balaton@eik.bme.hu>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20211007221253.29024-6-mark.cave-ayland@ilande.co.uk>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
2 years agomacfb: fix overflow of color_palette array
Mark Cave-Ayland [Thu, 7 Oct 2021 22:12:44 +0000 (23:12 +0100)]
macfb: fix overflow of color_palette array

The palette_current index counter has a maximum size of 256 * 3 to cover a full
color palette of 256 RGB entries. Linux assumes that the palette_current index
wraps back around to zero after writing 256 RGB entries so ensure that
palette_current is reset at this point to prevent data corruption within
MacfbState.

Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20211007221253.29024-5-mark.cave-ayland@ilande.co.uk>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
2 years agomacfb: fix invalid object reference in macfb_common_realize()
Mark Cave-Ayland [Thu, 7 Oct 2021 22:12:43 +0000 (23:12 +0100)]
macfb: fix invalid object reference in macfb_common_realize()

During realize memory_region_init_ram_nomigrate() is used to initialise the RAM
memory region used for the framebuffer but the owner object reference is
incorrect since MacFbState is a typedef and not a QOM type.

Change the memory region owner to be the corresponding DeviceState to fix the
issue and prevent random crashes during macfb_common_realize().

Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Fixes: 8ac919a0654 ("hw/m68k: add Nubus macfb video card")
Reviewed-by: BALATON Zoltan <balaton@eik.bme.hu>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20211007221253.29024-4-mark.cave-ayland@ilande.co.uk>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
2 years agomacfb: update macfb.c to use the Error API best practices
Mark Cave-Ayland [Thu, 7 Oct 2021 22:12:42 +0000 (23:12 +0100)]
macfb: update macfb.c to use the Error API best practices

As per the current Error API best practices, change macfb_commom_realize() to return
a boolean indicating success to reduce errp boiler-plate handling code. Note that
memory_region_init_ram_nomigrate() is also updated to use &error_abort to indicate
a non-recoverable error, matching the behaviour recommended after similar
discussions on memory API failures for the recent nubus changes.

Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20211007221253.29024-3-mark.cave-ayland@ilande.co.uk>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
2 years agomacfb: handle errors that occur during realize
Mark Cave-Ayland [Thu, 7 Oct 2021 22:12:41 +0000 (23:12 +0100)]
macfb: handle errors that occur during realize

Make sure any errors that occur within the macfb realize chain are detected
and handled correctly to prevent crashes and to ensure that error messages are
reported back to the user.

Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: BALATON Zoltan <balaton@eik.bme.hu>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20211007221253.29024-2-mark.cave-ayland@ilande.co.uk>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
2 years agoMerge remote-tracking branch 'remotes/vsementsov/tags/pull-jobs-2021-10-07-v2' into...
Richard Henderson [Thu, 7 Oct 2021 17:26:35 +0000 (10:26 -0700)]
Merge remote-tracking branch 'remotes/vsementsov/tags/pull-jobs-2021-10-07-v2' into staging

mirror: Handle errors after READY cancel
v2: add small fix by Stefano, Hanna's series fixed

# gpg: Signature made Thu 07 Oct 2021 08:25:07 AM PDT
# gpg:                using RSA key 8B9C26CDB2FD147C880E86A1561F24C1F19F79FB
# gpg: Good signature from "Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.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: 8B9C 26CD B2FD 147C 880E  86A1 561F 24C1 F19F 79FB

* remotes/vsementsov/tags/pull-jobs-2021-10-07-v2:
  iotests: Add mirror-ready-cancel-error test
  mirror: Do not clear .cancelled
  mirror: Stop active mirroring after force-cancel
  mirror: Check job_is_cancelled() earlier
  mirror: Use job_is_cancelled()
  job: Add job_cancel_requested()
  job: Do not soft-cancel after a job is done
  jobs: Give Job.force_cancel more meaning
  job: @force parameter for job_cancel_sync()
  job: Force-cancel jobs in a failed transaction
  mirror: Drop s->synced
  mirror: Keep s->synced on error
  job: Context changes in job_completed_txn_abort()
  block/aio_task: assert `max_busy_tasks` is greater than 0
  block/backup: avoid integer overflow of `max-workers`

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
2 years agoiothread: use IOThreadParamInfo in iothread_[set|get]_param()
Stefano Garzarella [Tue, 27 Jul 2021 14:59:36 +0000 (16:59 +0200)]
iothread: use IOThreadParamInfo in iothread_[set|get]_param()

Commit 0445409d74 ("iothread: generalize
iothread_set_param/iothread_get_param") moved common code to set and
get IOThread parameters in two new functions.

These functions are called inside callbacks, so we don't need to use an
opaque pointer. Let's replace `void *opaque` parameter with
`IOThreadParamInfo *info`.

Suggested-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Stefano Garzarella <sgarzare@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-id: 20210727145936.147032-3-sgarzare@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
2 years agoiothread: rename PollParamInfo to IOThreadParamInfo
Stefano Garzarella [Tue, 27 Jul 2021 14:59:35 +0000 (16:59 +0200)]
iothread: rename PollParamInfo to IOThreadParamInfo

Commit 1793ad0247 ("iothread: add aio-max-batch parameter") added
a new parameter (aio-max-batch) to IOThread and used PollParamInfo
structure to handle it.

Since it is not a parameter of the polling mechanism, we rename the
structure to a more generic IOThreadParamInfo.

Suggested-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Stefano Garzarella <sgarzare@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-id: 20210727145936.147032-2-sgarzare@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
2 years agoMerge remote-tracking branch 'remotes/alistair23/tags/pull-riscv-to-apply-20211007...
Richard Henderson [Thu, 7 Oct 2021 14:18:39 +0000 (07:18 -0700)]
Merge remote-tracking branch 'remotes/alistair23/tags/pull-riscv-to-apply-20211007' into staging

Third RISC-V PR for QEMU 6.2

 - Add Zb[abcs] instruction support
 - Remove RVB support
 - Bug fix of setting mstatus_hs.[SD|FS] bits
 - Mark some UART devices as 'input'
 - QOMify PolarFire MMUART
 - Fixes for sifive PDMA
 - Mark shakti_c as not user creatable

# gpg: Signature made Wed 06 Oct 2021 11:42:53 PM PDT
# gpg:                using RSA key F6C4AC46D4934868D3B8CE8F21E10D29DF977054
# gpg: Good signature from "Alistair Francis <alistair@alistair23.me>" [full]

* remotes/alistair23/tags/pull-riscv-to-apply-20211007: (26 commits)
  hw/riscv: shakti_c: Mark as not user creatable
  hw/dma: sifive_pdma: Don't run DMA when channel is disclaimed
  hw/dma: sifive_pdma: Fix Control.claim bit detection
  hw/char/mchp_pfsoc_mmuart: QOM'ify PolarFire MMUART
  hw/char/mchp_pfsoc_mmuart: Use a MemoryRegion container
  hw/char/mchp_pfsoc_mmuart: Simplify MCHP_PFSOC_MMUART_REG definition
  hw/char: sifive_uart: Register device in 'input' category
  hw/char: shakti_uart: Register device in 'input' category
  hw/char: ibex_uart: Register device in 'input' category
  target/riscv: Set mstatus_hs.[SD|FS] bits if Clean and V=1 in mark_fs_dirty()
  disas/riscv: Add Zb[abcs] instructions
  target/riscv: Remove RVB (replaced by Zb[abcs])
  target/riscv: Add zext.h instructions to Zbb, removing pack/packu/packh
  target/riscv: Add rev8 instruction, removing grev/grevi
  target/riscv: Add a REQUIRE_32BIT macro
  target/riscv: Add orc.b instruction for Zbb, removing gorc/gorci
  target/riscv: Reassign instructions to the Zbb-extension
  target/riscv: Add instructions of the Zbc-extension
  target/riscv: Reassign instructions to the Zbs-extension
  target/riscv: Remove shift-one instructions (proposed Zbo in pre-0.93 draft-B)
  ...

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
2 years agoiotests: Add mirror-ready-cancel-error test
Hanna Reitz [Wed, 6 Oct 2021 15:19:40 +0000 (17:19 +0200)]
iotests: Add mirror-ready-cancel-error test

Test what happens when there is an I/O error after a mirror job in the
READY phase has been cancelled.

Signed-off-by: Hanna Reitz <hreitz@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Tested-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20211006151940.214590-14-hreitz@redhat.com>
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
2 years agomirror: Do not clear .cancelled
Hanna Reitz [Wed, 6 Oct 2021 15:19:39 +0000 (17:19 +0200)]
mirror: Do not clear .cancelled

Clearing .cancelled before leaving the main loop when the job has been
soft-cancelled is no longer necessary since job_is_cancelled() only
returns true for jobs that have been force-cancelled.

Therefore, this only makes a differences in places that call
job_cancel_requested().  In block/mirror.c, this is done only before
.cancelled was cleared.

In job.c, there are two callers:
- job_completed_txn_abort() asserts that .cancelled is true, so keeping
  it true will not affect this place.

- job_complete() refuses to let a job complete that has .cancelled set.
  It is correct to refuse to let the user invoke job-complete on mirror
  jobs that have already been soft-cancelled.

With this change, there are no places that reset .cancelled to false and
so we can be sure that .force_cancel can only be true if .cancelled is
true as well.  Assert this in job_is_cancelled().

Signed-off-by: Hanna Reitz <hreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20211006151940.214590-13-hreitz@redhat.com>
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
2 years agomirror: Stop active mirroring after force-cancel
Hanna Reitz [Wed, 6 Oct 2021 15:19:38 +0000 (17:19 +0200)]
mirror: Stop active mirroring after force-cancel

Once the mirror job is force-cancelled (job_is_cancelled() is true), we
should not generate new I/O requests.  This applies to active mirroring,
too, so stop it once the job is cancelled.

(We must still forward all I/O requests to the source, though, of
course, but those are not really I/O requests generated by the job, so
this is fine.)

Signed-off-by: Hanna Reitz <hreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20211006151940.214590-12-hreitz@redhat.com>
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
2 years agomirror: Check job_is_cancelled() earlier
Hanna Reitz [Wed, 6 Oct 2021 15:19:37 +0000 (17:19 +0200)]
mirror: Check job_is_cancelled() earlier

We must check whether the job is force-cancelled early in our main loop,
most importantly before any `continue` statement.  For example, we used
to have `continue`s before our current checking location that are
triggered by `mirror_flush()` failing.  So, if `mirror_flush()` kept
failing, force-cancelling the job would not terminate it.

Jobs can be cancelled while they yield, and once they are
(force-cancelled), they should not generate new I/O requests.
Therefore, we should put the check after the last yield before
mirror_iteration() is invoked.

Buglink: https://gitlab.com/qemu-project/qemu/-/issues/462
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20211006151940.214590-11-hreitz@redhat.com>
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
2 years agomirror: Use job_is_cancelled()
Hanna Reitz [Wed, 6 Oct 2021 15:19:36 +0000 (17:19 +0200)]
mirror: Use job_is_cancelled()

mirror_drained_poll() returns true whenever the job is cancelled,
because "we [can] be sure that it won't issue more requests".  However,
this is only true for force-cancelled jobs, so use job_is_cancelled().

Signed-off-by: Hanna Reitz <hreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20211006151940.214590-10-hreitz@redhat.com>
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
2 years agojob: Add job_cancel_requested()
Hanna Reitz [Wed, 6 Oct 2021 15:19:35 +0000 (17:19 +0200)]
job: Add job_cancel_requested()

Most callers of job_is_cancelled() actually want to know whether the job
is on its way to immediate termination.  For example, we refuse to pause
jobs that are cancelled; but this only makes sense for jobs that are
really actually cancelled.

A mirror job that is cancelled during READY with force=false should
absolutely be allowed to pause.  This "cancellation" (which is actually
a kind of completion) may take an indefinite amount of time, and so
should behave like any job during normal operation.  For example, with
on-target-error=stop, the job should stop on write errors.  (In
contrast, force-cancelled jobs should not get write errors, as they
should just terminate and not do further I/O.)

Therefore, redefine job_is_cancelled() to only return true for jobs that
are force-cancelled (which as of HEAD^ means any job that interprets the
cancellation request as a request for immediate termination), and add
job_cancel_requested() as the general variant, which returns true for
any jobs which have been requested to be cancelled, whether it be
immediately or after an arbitrarily long completion phase.

Finally, here is a justification for how different job_is_cancelled()
invocations are treated by this patch:

- block/mirror.c (mirror_run()):
  - The first invocation is a while loop that should loop until the job
    has been cancelled or scheduled for completion.  What kind of cancel
    does not matter, only the fact that the job is supposed to end.

  - The second invocation wants to know whether the job has been
    soft-cancelled.  Calling job_cancel_requested() is a bit too broad,
    but if the job were force-cancelled, we should leave the main loop
    as soon as possible anyway, so this should not matter here.

  - The last two invocations already check force_cancel, so they should
    continue to use job_is_cancelled().

- block/backup.c, block/commit.c, block/stream.c, anything in tests/:
  These jobs know only force-cancel, so there is no difference between
  job_is_cancelled() and job_cancel_requested().  We can continue using
  job_is_cancelled().

- job.c:
  - job_pause_point(), job_yield(), job_sleep_ns(): Only force-cancelled
    jobs should be prevented from being paused.  Continue using job_is_cancelled().

  - job_update_rc(), job_finalize_single(), job_finish_sync(): These
    functions are all called after the job has left its main loop.  The
    mirror job (the only job that can be soft-cancelled) will clear
    .cancelled before leaving the main loop if it has been
    soft-cancelled.  Therefore, these functions will observe .cancelled
    to be true only if the job has been force-cancelled.  We can
    continue to use job_is_cancelled().
    (Furthermore, conceptually, a soft-cancelled mirror job should not
    report to have been cancelled.  It should report completion (see
    also the block-job-cancel QAPI documentation).  Therefore, it makes
    sense for these functions not to distinguish between a
    soft-cancelled mirror job and a job that has completed as normal.)

  - job_completed_txn_abort(): All jobs other than @job have been
    force-cancelled.  job_is_cancelled() must be true for them.
    Regarding @job itself: job_completed_txn_abort() is mostly called
    when the job's return value is not 0.  A soft-cancelled mirror has a
    return value of 0, and so will not end up here then.
    However, job_cancel() invokes job_completed_txn_abort() if the job
    has been deferred to the main loop, which is mostly the case for
    completed jobs (which skip the assertion), but not for sure.
    To be safe, use job_cancel_requested() in this assertion.

  - job_complete(): This is function eventually invoked by the user
    (through qmp_block_job_complete() or qmp_job_complete(), or
    job_complete_sync(), which comes from qemu-img).  The intention here
    is to prevent a user from invoking job-complete after the job has
    been cancelled.  This should also apply to soft cancelling: After a
    mirror job has been soft-cancelled, the user should not be able to
    decide otherwise and have it complete as normal (i.e. pivoting to
    the target).

  - job_cancel(): Both functions are equivalent (see comment there), but
    we want to use job_is_cancelled(), because this shows that we call
    job_completed_txn_abort() only for force-cancelled jobs.  (As
    explained for job_update_rc(), soft-cancelled jobs should be treated
    as if they have completed as normal.)

Buglink: https://gitlab.com/qemu-project/qemu/-/issues/462
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20211006151940.214590-9-hreitz@redhat.com>
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
2 years agojob: Do not soft-cancel after a job is done
Hanna Reitz [Wed, 6 Oct 2021 15:19:34 +0000 (17:19 +0200)]
job: Do not soft-cancel after a job is done

The only job that supports a soft cancel mode is the mirror job, and in
such a case it resets its .cancelled field before it leaves its .run()
function, so it does not really count as cancelled.

However, it is possible to cancel the job after .run() returns and
before job_exit() (which is run in the main loop) is executed.  Then,
.cancelled would still be true and the job would count as cancelled.
This does not seem to be in the interest of the mirror job, so adjust
job_cancel_async() to not set .cancelled in such a case, and
job_cancel() to not invoke job_completed_txn_abort().

Signed-off-by: Hanna Reitz <hreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20211006151940.214590-8-hreitz@redhat.com>
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
2 years agojobs: Give Job.force_cancel more meaning
Hanna Reitz [Wed, 6 Oct 2021 15:19:33 +0000 (17:19 +0200)]
jobs: Give Job.force_cancel more meaning

We largely have two cancel modes for jobs:

First, there is actual cancelling.  The job is terminated as soon as
possible, without trying to reach a consistent result.

Second, we have mirror in the READY state.  Technically, the job is not
really cancelled, but it just is a different completion mode.  The job
can still run for an indefinite amount of time while it tries to reach a
consistent result.

We want to be able to clearly distinguish which cancel mode a job is in
(when it has been cancelled).  We can use Job.force_cancel for this, but
right now it only reflects cancel requests from the user with
force=true, but clearly, jobs that do not even distinguish between
force=false and force=true are effectively always force-cancelled.

So this patch has Job.force_cancel signify whether the job will
terminate as soon as possible (force_cancel=true) or whether it will
effectively remain running despite being "cancelled"
(force_cancel=false).

To this end, we let jobs that provide JobDriver.cancel() tell the
generic job code whether they will terminate as soon as possible or not,
and for jobs that do not provide that method we assume they will.

Signed-off-by: Hanna Reitz <hreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20211006151940.214590-7-hreitz@redhat.com>
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
2 years agojob: @force parameter for job_cancel_sync()
Hanna Reitz [Wed, 6 Oct 2021 15:19:32 +0000 (17:19 +0200)]
job: @force parameter for job_cancel_sync()

Callers should be able to specify whether they want job_cancel_sync() to
force-cancel the job or not.

In fact, almost all invocations do not care about consistency of the
result and just want the job to terminate as soon as possible, so they
should pass force=true.  The replication block driver is the exception,
specifically the active commit job it runs.

As for job_cancel_sync_all(), all callers want it to force-cancel all
jobs, because that is the point of it: To cancel all remaining jobs as
quickly as possible (generally on process termination).  So make it
invoke job_cancel_sync() with force=true.

This changes some iotest outputs, because quitting qemu while a mirror
job is active will now lead to it being cancelled instead of completed,
which is what we want.  (Cancelling a READY mirror job with force=false
may take an indefinite amount of time, which we do not want when
quitting.  If users want consistent results, they must have all jobs be
done before they quit qemu.)

Buglink: https://gitlab.com/qemu-project/qemu/-/issues/462
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20211006151940.214590-6-hreitz@redhat.com>
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
2 years agojob: Force-cancel jobs in a failed transaction
Hanna Reitz [Wed, 6 Oct 2021 15:19:31 +0000 (17:19 +0200)]
job: Force-cancel jobs in a failed transaction

When a transaction is aborted, no result matters, and so all jobs within
should be force-cancelled.

Signed-off-by: Hanna Reitz <hreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20211006151940.214590-5-hreitz@redhat.com>
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
2 years agomirror: Drop s->synced
Hanna Reitz [Wed, 6 Oct 2021 15:19:30 +0000 (17:19 +0200)]
mirror: Drop s->synced

As of HEAD^, there is no meaning to s->synced other than whether the job
is READY or not.  job_is_ready() gives us that information, too.

Suggested-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20211006151940.214590-4-hreitz@redhat.com>
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
2 years agomirror: Keep s->synced on error
Hanna Reitz [Wed, 6 Oct 2021 15:19:29 +0000 (17:19 +0200)]
mirror: Keep s->synced on error

An error does not take us out of the READY phase, which is what
s->synced signifies.  It does of course mean that source and target are
no longer in sync, but that is what s->actively_sync is for -- s->synced
never meant that source and target are in sync, only that they were at
some point (and at that point we transitioned into the READY phase).

The tangible problem is that we transition to READY once we are in sync
and s->synced is false.  By resetting s->synced here, we will transition
from READY to READY once the error is resolved (if the job keeps
running), and that transition is not allowed.

Signed-off-by: Hanna Reitz <hreitz@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20211006151940.214590-3-hreitz@redhat.com>
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
2 years agojob: Context changes in job_completed_txn_abort()
Hanna Reitz [Wed, 6 Oct 2021 15:19:28 +0000 (17:19 +0200)]
job: Context changes in job_completed_txn_abort()

Finalizing the job may cause its AioContext to change.  This is noted by
job_exit(), which points at job_txn_apply() to take this fact into
account.

However, job_completed() does not necessarily invoke job_txn_apply()
(through job_completed_txn_success()), but potentially also
job_completed_txn_abort().  The latter stores the context in a local
variable, and so always acquires the same context at its end that it has
released in the beginning -- which may be a different context from the
one that job_exit() releases at its end.  If it is different, qemu
aborts ("qemu_mutex_unlock_impl: Operation not permitted").

Drop the local @outer_ctx variable from job_completed_txn_abort(), and
instead re-acquire the actual job's context at the end of the function,
so job_exit() will release the same.

Signed-off-by: Hanna Reitz <hreitz@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20211006151940.214590-2-hreitz@redhat.com>
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
2 years agohw/riscv: shakti_c: Mark as not user creatable
Alistair Francis [Tue, 28 Sep 2021 23:29:55 +0000 (09:29 +1000)]
hw/riscv: shakti_c: Mark as not user creatable

Mark the shakti_c machine as not user creatable.

Resolves: https://gitlab.com/qemu-project/qemu/-/issues/639
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Tested-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <c617a04d4e3dd041a3427b47a1b1d5ab475a2edd.1632871759.git.alistair.francis@wdc.com>

2 years agohw/dma: sifive_pdma: Don't run DMA when channel is disclaimed
Bin Meng [Mon, 27 Sep 2021 07:21:24 +0000 (15:21 +0800)]
hw/dma: sifive_pdma: Don't run DMA when channel is disclaimed

If Control.run bit is set while not preserving the Control.claim
bit, the DMA transfer shall not be started.

The following result is PDMA tested in U-Boot on Unleashed board:

=> mw.l 0x3000000 0x0                      <= Disclaim channel 0
=> mw.l 0x3000000 0x1                      <= Claim channel 0
=> mw.l 0x3000004 0x55000000               <= wsize = rsize = 5 (2^5 = 32 bytes)
=> mw.q 0x3000008 0x2                      <= NextBytes = 2
=> mw.q 0x3000010 0x84000000               <= NextDestination = 0x84000000
=> mw.q 0x3000018 0x84001000               <= NextSource = 0x84001000
=> mw.l 0x84000000 0x87654321              <= Fill test data to dst
=> mw.l 0x84001000 0x12345678              <= Fill test data to src
=> md.l 0x84000000 1; md.l 0x84001000 1    <= Dump src/dst memory contents
8400000087654321                               !Ce.
8400100012345678                               xV4.
=> md.l 0x3000000 8                        <= Dump PDMA status
0300000000000001 55000000 00000002 00000000    .......U........
0300001084000000 00000000 84001000 00000000    ................
=> mw.l 0x3000000 0x2                      <= Set channel 0 run bit only
=> md.l 0x3000000 8                        <= Dump PDMA status
0300000000000000 55000000 00000002 00000000    .......U........
0300001084000000 00000000 84001000 00000000    ................
=> md.l 0x84000000 1; md.l 0x84001000 1    <= Dump src/dst memory contents
8400000087654321                               !Ce.
8400100012345678                               xV4.

Signed-off-by: Bin Meng <bmeng.cn@gmail.com>
Acked-by: Alistair Francis <alistair.francis@wdc.com>
Message-id: 20210927072124.1564129-2-bmeng.cn@gmail.com
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
2 years agohw/dma: sifive_pdma: Fix Control.claim bit detection
Bin Meng [Mon, 27 Sep 2021 07:21:23 +0000 (15:21 +0800)]
hw/dma: sifive_pdma: Fix Control.claim bit detection

At present the codes detect whether the DMA channel is claimed by:

  claimed = !!s->chan[ch].control & CONTROL_CLAIM;

As ! has higher precedence over & (bitwise and), this is essentially

  claimed = (!!s->chan[ch].control) & CONTROL_CLAIM;

which is wrong, as any non-zero bit set in the control register will
produce a result of a claimed channel.

Fixes: de7c7988d25d ("hw/dma: sifive_pdma: reset Next* registers when Control.claim is set")
Signed-off-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-id: 20210927072124.1564129-1-bmeng.cn@gmail.com
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
2 years agohw/char/mchp_pfsoc_mmuart: QOM'ify PolarFire MMUART
Philippe Mathieu-Daudé [Sat, 25 Sep 2021 13:34:07 +0000 (15:34 +0200)]
hw/char/mchp_pfsoc_mmuart: QOM'ify PolarFire MMUART

- Embed SerialMM in MchpPfSoCMMUartState and QOM-initialize it
- Alias SERIAL_MM 'chardev' property on MCHP_PFSOC_UART
- Forward SerialMM sysbus IRQ in mchp_pfsoc_mmuart_realize()
- Add DeviceReset() method
- Add vmstate structure for migration
- Register device in 'input' category
- Keep mchp_pfsoc_mmuart_create() behavior

Note, serial_mm_init() calls qdev_set_legacy_instance_id().
This call is only needed for backwards-compatibility of incoming
migration data with old versions of QEMU which implemented migration
of devices with hand-rolled code. Since this device didn't previously
handle migration at all, then it doesn't need to set the legacy
instance ID.

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Bin Meng <bin.meng@windriver.com>
Tested-by: Bin Meng <bin.meng@windriver.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-id: 20210925133407.1259392-4-f4bug@amsat.org
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
2 years agohw/char/mchp_pfsoc_mmuart: Use a MemoryRegion container
Philippe Mathieu-Daudé [Sat, 25 Sep 2021 13:34:06 +0000 (15:34 +0200)]
hw/char/mchp_pfsoc_mmuart: Use a MemoryRegion container

Our device have 2 different I/O regions:
- a 16550 UART mapped for 32-bit accesses
- 13 extra registers

Instead of mapping each region on the main bus, introduce
a container, map the 2 devices regions on the container,
and map the container on the main bus.

Before:

  (qemu) info mtree
    ...
    0000000020100000-000000002010001f (prio 0, i/o): serial
    0000000020100020-000000002010101f (prio 0, i/o): mchp.pfsoc.mmuart
    0000000020102000-000000002010201f (prio 0, i/o): serial
    0000000020102020-000000002010301f (prio 0, i/o): mchp.pfsoc.mmuart
    0000000020104000-000000002010401f (prio 0, i/o): serial
    0000000020104020-000000002010501f (prio 0, i/o): mchp.pfsoc.mmuart
    0000000020106000-000000002010601f (prio 0, i/o): serial
    0000000020106020-000000002010701f (prio 0, i/o): mchp.pfsoc.mmuart

After:

  (qemu) info mtree
    ...
    0000000020100000-0000000020100fff (prio 0, i/o): mchp.pfsoc.mmuart
      0000000020100000-000000002010001f (prio 0, i/o): serial
      0000000020100020-0000000020100fff (prio 0, i/o): mchp.pfsoc.mmuart.regs
    0000000020102000-0000000020102fff (prio 0, i/o): mchp.pfsoc.mmuart
      0000000020102000-000000002010201f (prio 0, i/o): serial
      0000000020102020-0000000020102fff (prio 0, i/o): mchp.pfsoc.mmuart.regs
    0000000020104000-0000000020104fff (prio 0, i/o): mchp.pfsoc.mmuart
      0000000020104000-000000002010401f (prio 0, i/o): serial
      0000000020104020-0000000020104fff (prio 0, i/o): mchp.pfsoc.mmuart.regs
    0000000020106000-0000000020106fff (prio 0, i/o): mchp.pfsoc.mmuart
      0000000020106000-000000002010601f (prio 0, i/o): serial
      0000000020106020-0000000020106fff (prio 0, i/o): mchp.pfsoc.mmuart.regs

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Bin Meng <bin.meng@windriver.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Tested-by: Bin Meng <bin.meng@windriver.com>
Message-id: 20210925133407.1259392-3-f4bug@amsat.org
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
2 years agohw/char/mchp_pfsoc_mmuart: Simplify MCHP_PFSOC_MMUART_REG definition
Philippe Mathieu-Daudé [Sat, 25 Sep 2021 13:34:05 +0000 (15:34 +0200)]
hw/char/mchp_pfsoc_mmuart: Simplify MCHP_PFSOC_MMUART_REG definition

The current MCHP_PFSOC_MMUART_REG_SIZE definition represent the
size occupied by all the registers. However all registers are
32-bit wide, and the MemoryRegionOps handlers are restricted to
32-bit:

  static const MemoryRegionOps mchp_pfsoc_mmuart_ops = {
      .read = mchp_pfsoc_mmuart_read,
      .write = mchp_pfsoc_mmuart_write,
      .impl = {
          .min_access_size = 4,
          .max_access_size = 4,
      },

Avoid being triskaidekaphobic, simplify by using the number of
registers.

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Bin Meng <bin.meng@windriver.com>
Tested-by: Bin Meng <bin.meng@windriver.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-id: 20210925133407.1259392-2-f4bug@amsat.org
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
2 years agohw/char: sifive_uart: Register device in 'input' category
Bin Meng [Sun, 26 Sep 2021 10:50:03 +0000 (18:50 +0800)]
hw/char: sifive_uart: Register device in 'input' category

The category of sifive_uart device is not set. Put it into the
'input' category.

Signed-off-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-id: 20210926105003.2716-3-bmeng.cn@gmail.com
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
2 years agohw/char: shakti_uart: Register device in 'input' category
Bin Meng [Sun, 26 Sep 2021 10:50:02 +0000 (18:50 +0800)]
hw/char: shakti_uart: Register device in 'input' category

The category of shakti_uart device is not set. Put it into the
'input' category.

Signed-off-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-id: 20210926105003.2716-2-bmeng.cn@gmail.com
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
2 years agohw/char: ibex_uart: Register device in 'input' category
Bin Meng [Sun, 26 Sep 2021 10:50:01 +0000 (18:50 +0800)]
hw/char: ibex_uart: Register device in 'input' category

The category of ibex_uart device is not set. Put it into the
'input' category.

Signed-off-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-id: 20210926105003.2716-1-bmeng.cn@gmail.com
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
2 years agotarget/riscv: Set mstatus_hs.[SD|FS] bits if Clean and V=1 in mark_fs_dirty()
Frank Chang [Tue, 21 Sep 2021 02:02:33 +0000 (10:02 +0800)]
target/riscv: Set mstatus_hs.[SD|FS] bits if Clean and V=1 in mark_fs_dirty()

When V=1, both vsstauts.FS and HS-level sstatus.FS are in effect.
Modifying the floating-point state when V=1 causes both fields to
be set to 3 (Dirty).

However, it's possible that HS-level sstatus.FS is Clean and VS-level
vsstatus.FS is Dirty at the time mark_fs_dirty() is called when V=1.
We can't early return for this case because we still need to set
sstatus.FS to Dirty according to spec.

Signed-off-by: Frank Chang <frank.chang@sifive.com>
Reviewed-by: Vincent Chen <vincent.chen@sifive.com>
Tested-by: Vincent Chen <vincent.chen@sifive.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-id: 20210921020234.123448-1-frank.chang@sifive.com
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
2 years agodisas/riscv: Add Zb[abcs] instructions
Philipp Tomsich [Sat, 11 Sep 2021 14:00:16 +0000 (16:00 +0200)]
disas/riscv: Add Zb[abcs] instructions

With the addition of Zb[abcs], we also need to add disassembler
support for these new instructions.

Signed-off-by: Philipp Tomsich <philipp.tomsich@vrull.eu>
Acked-by: Alistair Francis <alistair.francis@wdc.com>
Message-id: 20210911140016.834071-17-philipp.tomsich@vrull.eu
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
2 years agotarget/riscv: Remove RVB (replaced by Zb[abcs])
Philipp Tomsich [Sat, 11 Sep 2021 14:00:15 +0000 (16:00 +0200)]
target/riscv: Remove RVB (replaced by Zb[abcs])

With everything classified as Zb[abcs] and pre-0.93 draft-B
instructions that are not part of Zb[abcs] removed, we can remove the
remaining support code for RVB.

Note that RVB has been retired for good and misa.B will neither mean
'some' or 'all of' Zb*:
  https://lists.riscv.org/g/tech-bitmanip/message/532

Signed-off-by: Philipp Tomsich <philipp.tomsich@vrull.eu>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Message-id: 20210911140016.834071-16-philipp.tomsich@vrull.eu
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
2 years agotarget/riscv: Add zext.h instructions to Zbb, removing pack/packu/packh
Philipp Tomsich [Sat, 11 Sep 2021 14:00:14 +0000 (16:00 +0200)]
target/riscv: Add zext.h instructions to Zbb, removing pack/packu/packh

The 1.0.0 version of Zbb does not contain pack/packu/packh. However, a
zext.h instruction is provided (built on pack/packh from pre-0.93
draft-B) is available.

This commit adds zext.h and removes the pack* instructions.

Note that the encodings for zext.h are different between RV32 and
RV64, which is handled through REQUIRE_32BIT.

Signed-off-by: Philipp Tomsich <philipp.tomsich@vrull.eu>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-id: 20210911140016.834071-15-philipp.tomsich@vrull.eu
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
2 years agotarget/riscv: Add rev8 instruction, removing grev/grevi
Philipp Tomsich [Sat, 11 Sep 2021 14:00:13 +0000 (16:00 +0200)]
target/riscv: Add rev8 instruction, removing grev/grevi

The 1.0.0 version of Zbb does not contain grev/grevi.  Instead, a
rev8 instruction (equivalent to the rev8 pseudo-instruction built on
grevi from pre-0.93 draft-B) is available.

This commit adds the new rev8 instruction and removes grev/grevi.

Note that there is no W-form of this instruction (both a
sign-extending and zero-extending 32-bit version can easily be
synthesized by following rev8 with either a srai or srli instruction
on RV64) and that the opcode encodings for rev8 in RV32 and RV64 are
different.

Signed-off-by: Philipp Tomsich <philipp.tomsich@vrull.eu>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-id: 20210911140016.834071-14-philipp.tomsich@vrull.eu
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
2 years agotarget/riscv: Add a REQUIRE_32BIT macro
Philipp Tomsich [Sat, 11 Sep 2021 14:00:12 +0000 (16:00 +0200)]
target/riscv: Add a REQUIRE_32BIT macro

With the changes to Zb[abcs], there's some encodings that are
different in RV64 and RV32 (e.g., for rev8 and zext.h). For these,
we'll need a helper macro allowing us to select on RV32, as well.

Signed-off-by: Philipp Tomsich <philipp.tomsich@vrull.eu>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Message-id: 20210911140016.834071-13-philipp.tomsich@vrull.eu
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
2 years agotarget/riscv: Add orc.b instruction for Zbb, removing gorc/gorci
Philipp Tomsich [Sat, 11 Sep 2021 14:00:11 +0000 (16:00 +0200)]
target/riscv: Add orc.b instruction for Zbb, removing gorc/gorci

The 1.0.0 version of Zbb does not contain gorc/gorci.  Instead, a
orc.b instruction (equivalent to the orc.b pseudo-instruction built on
gorci from pre-0.93 draft-B) is available, mainly targeting
string-processing workloads.

This commit adds the new orc.b instruction and removed gorc/gorci.

Signed-off-by: Philipp Tomsich <philipp.tomsich@vrull.eu>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-id: 20210911140016.834071-12-philipp.tomsich@vrull.eu
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
2 years agotarget/riscv: Reassign instructions to the Zbb-extension
Philipp Tomsich [Sat, 11 Sep 2021 14:00:10 +0000 (16:00 +0200)]
target/riscv: Reassign instructions to the Zbb-extension

This reassigns the instructions that are part of Zbb into it, with the
notable exceptions of the instructions (rev8, zext.w and orc.b) that
changed due to gorci, grevi and pack not being part of Zb[abcs].

Signed-off-by: Philipp Tomsich <philipp.tomsich@vrull.eu>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Acked-by: Bin Meng <bmeng.cn@gmail.com>
Message-id: 20210911140016.834071-11-philipp.tomsich@vrull.eu
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
2 years agotarget/riscv: Add instructions of the Zbc-extension
Philipp Tomsich [Sat, 11 Sep 2021 14:00:09 +0000 (16:00 +0200)]
target/riscv: Add instructions of the Zbc-extension

The following instructions are part of Zbc:
 - clmul
 - clmulh
 - clmulr

Note that these instructions were already defined in the pre-0.93 and
the 0.93 draft-B proposals, but had not been omitted in the earlier
addition of draft-B to QEmu.

Signed-off-by: Philipp Tomsich <philipp.tomsich@vrull.eu>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-id: 20210911140016.834071-10-philipp.tomsich@vrull.eu
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
2 years agotarget/riscv: Reassign instructions to the Zbs-extension
Philipp Tomsich [Sat, 11 Sep 2021 14:00:08 +0000 (16:00 +0200)]
target/riscv: Reassign instructions to the Zbs-extension

The following instructions are part of Zbs:
 - b{set,clr,ext,inv}
 - b{set,clr,ext,inv}i

Signed-off-by: Philipp Tomsich <philipp.tomsich@vrull.eu>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Acked-by: Bin Meng <bmeng.cn@gmail.com>
Message-id: 20210911140016.834071-9-philipp.tomsich@vrull.eu
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
2 years agotarget/riscv: Remove shift-one instructions (proposed Zbo in pre-0.93 draft-B)
Philipp Tomsich [Sat, 11 Sep 2021 14:00:07 +0000 (16:00 +0200)]
target/riscv: Remove shift-one instructions (proposed Zbo in pre-0.93 draft-B)

The Zb[abcs] ratification package does not include the proposed
shift-one instructions. There currently is no clear plan to whether
these (or variants of them) will be ratified as Zbo (or a different
extension) or what the timeframe for such a decision could be.

Signed-off-by: Philipp Tomsich <philipp.tomsich@vrull.eu>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Acked-by: Bin Meng <bmeng.cn@gmail.com>
Message-id: 20210911140016.834071-8-philipp.tomsich@vrull.eu
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>