qapi/gen.py: delint with pylint

'fp' and 'fd' are self-evident in context, add them to the list of OK
names.

_top and _bottom also need to stay standard methods because some users
override the method and need to use `self`. Tell pylint to shush.

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Cleber Rosa <crosa@redhat.com>
Message-Id: <20201009161558.107041-32-jsnow@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>

qapi/gen.py: update write() to be more idiomatic

Make the file handling here just a tiny bit more idiomatic.
(I realize this is heavily subjective.)

Use exist_ok=True for os.makedirs and remove the exception,
use fdopen() to wrap the file descriptor in a File-like object,
and use a context manager for managing the file pointer.

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Cleber Rosa <crosa@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20201009161558.107041-31-jsnow@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>

qapi/gen.py: Remove unused parameter

_module_dirname doesn't use the 'what' argument, so remove it.

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Cleber Rosa <crosa@redhat.com>
Message-Id: <20201009161558.107041-30-jsnow@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>

qapi/gen.py: add type hint annotations

Annotations do not change runtime behavior.
This commit *only* adds annotations.

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Cleber Rosa <crosa@redhat.com>
Message-Id: <20201009161558.107041-28-jsnow@redhat.com>
Message-Id: <20201009161558.107041-29-jsnow@redhat.com>
[mypy.ini update squashed in]
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>

qapi/gen: Make _is_user_module() return bool

_is_user_module() returns thruth values.  The next commit wants it to
return bool.  Make it so.

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Cleber Rosa <crosa@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20201009161558.107041-27-jsnow@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
[Commit message rewritten]
Signed-off-by: Markus Armbruster <armbru@redhat.com>

qapi/source.py: delint with pylint

Shush an error and leave a hint for future cleanups when we're allowed
to use Python 3.7+.

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Cleber Rosa <crosa@redhat.com>
Tested-by: Cleber Rosa <crosa@redhat.com>
Message-Id: <20201009161558.107041-26-jsnow@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>

qapi/source.py: add type hint annotations

Annotations do not change runtime behavior.
This commit *only* adds annotations.

A note on typing of __init__: mypy requires init functions with no
parameters to document a return type of None to be considered fully
typed. In the case when there are input parameters, None may be omitted.

Since __init__ may never return any value, it is preferred to omit the
return annotation whenever possible.

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Cleber Rosa <crosa@redhat.com>
Tested-by: Cleber Rosa <crosa@redhat.com>
Message-Id: <20201009161558.107041-25-jsnow@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>

qapi/commands.py: add type hint annotations

Annotations do not change runtime behavior.
This commit *only* adds annotations.

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Cleber Rosa <crosa@redhat.com>
Message-Id: <20201009161558.107041-23-jsnow@redhat.com>
Message-Id: <20201009161558.107041-24-jsnow@redhat.com>
[mypy.ini update squashed in]
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>

qapi/commands.py: Don't re-bind to variable of different type

Mypy isn't a fan of rebinding a variable with a new data type.
It's easy enough to avoid.

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Cleber Rosa <crosa@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20201009161558.107041-22-jsnow@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>

qapi/events.py: Move comments into docstrings

Clarify them while we're here.

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Cleber Rosa <crosa@redhat.com>
Message-Id: <20201009161558.107041-21-jsnow@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>

qapi/events.py: add type hint annotations

Annotations do not change runtime behavior.
This commit *only* adds annotations.

Note: __init__ does not need its return type annotated, as it is special.
https://mypy.readthedocs.io/en/stable/class_basics.html#annotating-init-methods

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Cleber Rosa <crosa@redhat.com>
Message-Id: <20201009161558.107041-20-jsnow@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>

qapi: establish mypy type-checking baseline

Fix a minor typing issue, and then establish a mypy type-checking
baseline.

Like pylint, this should be run from the folder above:

 > mypy --config-file=qapi/mypy.ini qapi/

This is designed and tested for mypy 0.770 or greater.

Signed-off-by: John Snow <jsnow@redhat.com>
Tested-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Cleber Rosa <crosa@redhat.com>
Tested-by: Cleber Rosa <crosa@redhat.com>
Message-Id: <20201009161558.107041-19-jsnow@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>

qapi/common.py: move build_params into gen.py

Including it in common.py creates a circular import dependency; schema
relies on common, but common.build_params requires a type annotation
from schema. To type this properly, it needs to be moved outside the
cycle.

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Cleber Rosa <crosa@redhat.com>
Message-Id: <20201009161558.107041-18-jsnow@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>

qapi/common.py: Convert comments into docstrings, and elaborate

As docstrings, they'll show up in documentation and IDE help.

The docstring style being targeted is the Sphinx documentation
style. Sphinx uses an extension of ReST with "domains". We use the
(implicit) Python domain, which supports a number of custom "info
fields". Those info fields are documented here:
https://www.sphinx-doc.org/en/master/usage/restructuredtext/domains.html#info-field-lists

Primarily, we use `:param X: descr`, `:return[s]: descr`, and `:raise[s]
Z: when`. Everything else is the Sphinx dialect of ReST.

(No, nothing checks or enforces this style that I am aware of. Sphinx
either chokes or succeeds, but does not enforce a standard of what is
otherwise inside the docstring. Pycharm does highlight when your param
fields are not aligned with the actual fields present. It does not
highlight missing return or exception statements. There is no existing
style guide I am aware of that covers a standard for a minimally
acceptable docstring. I am debating writing one.)

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Cleber Rosa <crosa@redhat.com>
Message-Id: <20201009161558.107041-17-jsnow@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>

qapi/common.py: add type hint annotations

Annotations do not change runtime behavior.
This commit *only* adds annotations.

Note that build_params() cannot be fully annotated due to import
dependency issues.  The commit after next will take care of it.

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Cleber Rosa <crosa@redhat.com>
Message-Id: <20201009161558.107041-16-jsnow@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>

qapi/common.py: check with pylint

Remove qapi/common.py from the pylintrc ignore list.

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Cleber Rosa <crosa@redhat.com>
Tested-by: Cleber Rosa <crosa@redhat.com>
Tested-by: Eduardo Habkost <ehabkost@redhat.com>
Message-Id: <20201009161558.107041-15-jsnow@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>

qapi/common.py: Replace one-letter 'c' variable

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Cleber Rosa <crosa@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20201009161558.107041-14-jsnow@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>

qapi/common.py: delint with pylint

At this point, that just means using a consistent strategy for constant names.
constants get UPPER_CASE and names not used externally get a leading underscore.

As a preference, while renaming constants to be UPPERCASE, move them to
the head of the file. Generally, it's nice to be able to audit the code
that runs on import in one central place.

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Cleber Rosa <crosa@redhat.com>
Message-Id: <20201009161558.107041-13-jsnow@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>

qapi/common.py: Add indent manager

Code style tools really dislike the use of global keywords, because it
generally involves re-binding the name at runtime which can have strange
effects depending on when and how that global name is referenced in
other modules.

Make a little indent level manager instead.

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Cleber Rosa <crosa@redhat.com>
Message-Id: <20201009161558.107041-12-jsnow@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>

qapi/common.py: Remove python compatibility workaround

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Cleber Rosa <crosa@redhat.com>
Message-Id: <20201009161558.107041-11-jsnow@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>

qapi: add pylintrc

Using `pylint --generate-rcfile > pylintrc`, generate a skeleton
pylintrc file. Sections that are not presently relevant (by the end of
this series) are removed leaving just the empty section as a search
engine / documentation hint to future authors.

I am targeting pylint 2.6.0. In the future (and hopefully before 5.2 is
released), I aim to have gitlab CI running the specific targeted
versions of pylint, mypy, flake8, etc in a job.

2.5.x will work if you additionally pass --disable=bad-whitespace.
This warning was removed from 2.6.x, for lack of consistent support.

Right now, quite a few modules are ignored as they are known to fail as
of this commit. modules will be removed from the known-bad list
throughout this and following series as they are repaired.

Note: Normally, pylintrc would go in the folder above the module, but as
that folder is shared by many things, it is going inside the module
folder (for now). Due to a bug in pylint 2.5+, pylint does not
correctly recognize when it is being run from "inside" a package, and
must be run *outside* of the package.

Therefore, to run it, you must:

 > pylint scripts/qapi/ --rcfile=scripts/qapi/pylintrc

Signed-off-by: John Snow <jsnow@redhat.com>
Tested-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Cleber Rosa <crosa@redhat.com>
Tested-by: Cleber Rosa <crosa@redhat.com>
Message-Id: <20201009161558.107041-10-jsnow@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>

qapi: delint using flake8

Petty style guide fixes and line length enforcement.  Not a big win, not
a big loss, but flake8 passes 100% on the qapi module, which gives us an
easy baseline to enforce hereafter.

A note on the flake8 exception: flake8 will warn on *any* bare except,
but pylint's is context-aware and will suppress the warning if you
re-raise the exception.

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Cleber Rosa <crosa@redhat.com>
Message-Id: <20201009161558.107041-9-jsnow@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>

qapi: enforce import order/styling with isort

While we're mucking around with imports, we might as well formalize the
style we use. Let's use isort to do it for us.

lines_after_imports=2: Use two lines after imports, to match PEP8's
desire to have "two lines before and after" class definitions, which are
likely to start immediately after imports.

force_sort_within_sections: Intermingles "from x" and "import x" style
statements, such that sorting is always performed strictly on the module
name itself.

force_grid_wrap=4: Four or more imports from a single module will force
the one-per-line style that's more git-friendly. This will generally
happen for 'typing' imports.

multi_line_output=3: Uses the one-per-line indented style for long
imports.

include_trailing_comma: Adds a comma to the last import in a group,
which makes git conflicts nicer to deal with, generally.

line_length: 72 is chosen to match PEP8's "docstrings and comments" line
length limit. If you have a single line import that exceeds 72
characters, your names are too long!

Suggested-by: Cleber Rosa <crosa@redhat.com>
Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Cleber Rosa <crosa@redhat.com>
Tested-by: Cleber Rosa <crosa@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20201009161558.107041-8-jsnow@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>

qapi: Remove wildcard includes

Wildcard includes become hard to manage when refactoring and dealing
with circular dependencies with strictly typed mypy.

flake8 also flags each one as a warning, as it is not smart enough to
know which names exist in the imported file.

Remove them and include things explicitly by name instead.

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Cleber Rosa <crosa@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20201009161558.107041-7-jsnow@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>

qapi: Prefer explicit relative imports

All of the QAPI include statements are changed to be package-aware, as
explicit relative imports.

A quirk of Python packages is that the name of the package exists only
*outside* of the package. This means that to a module inside of the qapi
folder, there is inherently no such thing as the "qapi" package. The
reason these imports work is because the "qapi" package exists in the
context of the caller -- the execution shim, where sys.path includes a
directory that has a 'qapi' folder in it.

When we write "from qapi import sibling", we are NOT referencing the folder
'qapi', but rather "any package named qapi in sys.path". If you should
so happen to have a 'qapi' package in your path, it will use *that*
package.

When we write "from .sibling import foo", we always reference explicitly
our sibling module; guaranteeing consistency in *where* we are importing
these modules from.

This can be useful when working with virtual environments and packages
in development mode. In development mode, a package is installed as a
series of symlinks that forwards to your same source files. The problem
arises because code quality checkers will follow "import qapi.x" to the
"installed" version instead of the sibling file and -- even though they
are the same file -- they have different module paths, and this causes
cyclic import problems, false positive type mismatch errors, and more.

It can also be useful when dealing with hierarchical packages, e.g. if
we allow qemu.core.qmp, qemu.qapi.parser, etc.

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Cleber Rosa <crosa@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20201009161558.107041-6-jsnow@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>

qapi: move generator entrypoint into package

As part of delinting and adding type hints to the QAPI generator, it's
helpful for the entrypoint to be part of the package, only leaving a
very tiny entrypoint shim outside of the package.

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Cleber Rosa <crosa@redhat.com>
Tested-by: Cleber Rosa <crosa@redhat.com>
Message-Id: <20201009161558.107041-5-jsnow@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
[invalid_char() renamed to invalid_prefix_char()]
Signed-off-by: Markus Armbruster <armbru@redhat.com>

qapi-gen: Separate arg-parsing from generation

This is a minor re-work of the entrypoint script. It isolates a
generate() method from the actual command-line mechanism.

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Cleber Rosa <crosa@redhat.com>
Tested-by: Cleber Rosa <crosa@redhat.com>
Message-Id: <20201009161558.107041-4-jsnow@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
[invalid_char() renamed to invalid_prefix_char()]
Signed-off-by: Markus Armbruster <armbru@redhat.com>

qapi: modify docstrings to be sphinx-compatible

A precise style guide and a package-wide overhaul is forthcoming pending
further discussion and consensus. For now, merely avoid obvious errors
that cause Sphinx documentation build problems, using a style loosely
based on PEP 257 and Sphinx Autodoc. It is chosen for interoperability
with our existing Sphinx framework, and because it has loose recognition
in the Pycharm IDE.

See also:

https://www.python.org/dev/peps/pep-0257/
https://www.sphinx-doc.org/en/master/usage/restructuredtext/domains.html#info-field-lists
Signed-off-by: John Snow <jsnow@redhat.com>
Message-Id: <20201009161558.107041-3-jsnow@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>

docs: repair broken references

In two different places, we are not making a cross-reference to some
resource correctly.

Signed-off-by: John Snow <jsnow@redhat.com>
Message-Id: <20201009161558.107041-2-jsnow@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>

Merge remote-tracking branch 'remotes/dgibson/tags/ppc-for-5.2-20201009' into staging

ppc patch queue 2020-10-09

Here's the next set of ppc related patches for qemu-5.2.  There are
two main things here:

* Cleanups to error handling in spapr from Greg Kurz
* Improvements to NUMA handling for spapr from Daniel Barboza

There are also a handful of other bugfixes.

# gpg: Signature made Fri 09 Oct 2020 07:02:29 BST
# gpg:                using RSA key 75F46586AE61A66CC44E87DC6C38CACA20D9B392
# gpg: Good signature from "David Gibson <david@gibson.dropbear.id.au>" [full]
# gpg:                 aka "David Gibson (Red Hat) <dgibson@redhat.com>" [full]
# gpg:                 aka "David Gibson (ozlabs.org) <dgibson@ozlabs.org>" [full]
# gpg:                 aka "David Gibson (kernel.org) <dwg@kernel.org>" [unknown]
# Primary key fingerprint: 75F4 6586 AE61 A66C C44E  87DC 6C38 CACA 20D9 B392

* remotes/dgibson/tags/ppc-for-5.2-20201009:
  specs/ppc-spapr-numa: update with new NUMA support
  spapr_numa: consider user input when defining associativity
  spapr_numa: change reference-points and maxdomain settings
  spapr_numa: forbid asymmetrical NUMA setups
  spapr: add spapr_machine_using_legacy_numa() helper
  ppc/pnv: Increase max firmware size
  spapr: Add a return value to spapr_check_pagesize()
  spapr: Add a return value to spapr_nvdimm_validate()
  spapr: Simplify error handling in spapr_cpu_core_realize()
  spapr: Add a return value to spapr_set_vcpu_id()
  spapr: Simplify error handling in prop_get_fdt()
  spapr: Add a return value to spapr_drc_attach()
  spapr: Simplify error handling in spapr_vio_busdev_realize()
  spapr: Simplify error handling in do_client_architecture_support()
  spapr: Get rid of cas_check_pvr() error reporting
  spapr: Simplify error handling in callers of ppc_set_compat()
  ppc: Fix return value in cpu_post_load() error path
  ppc: Add a return value to ppc_set_compat() and ppc_set_compat_all()
  spapr: Fix error leak in spapr_realize_vcpu()
  spapr: Handle HPT allocation failure in nested guest

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

Merge remote-tracking branch 'remotes/armbru/tags/pull-error-2020-10-09' into staging

Error reporting patches for 2020-10-09

# gpg: Signature made Fri 09 Oct 2020 07:45:17 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-error-2020-10-09:
  error: Use error_fatal to simplify obvious fatal errors (again)
  error: Remove NULL checks on error_propagate() calls (again)

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

Merge remote-tracking branch 'remotes/armbru/tags/pull-monitor-2020-10-09' into staging

Monitor patches for 2020-10-09

# gpg: Signature made Fri 09 Oct 2020 06:16:51 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-monitor-2020-10-09:
  block: Convert 'block_resize' to coroutine
  block: Add bdrv_lock()/unlock()
  block: Add bdrv_co_enter()/leave()
  util/async: Add aio_co_reschedule_self()
  hmp: Add support for coroutine command handlers
  qmp: Move dispatcher to a coroutine
  qapi: Add a 'coroutine' flag for commands
  monitor: Make current monitor a per-coroutine property
  qmp: Call monitor_set_cur() only in qmp_dispatch()
  qmp: Assert that no other monitor is active
  hmp: Update current monitor only in handle_hmp_command()
  monitor: Use getter/setter functions for cur_mon
  monitor: Add Monitor parameter to monitor_get_cpu_index()
  monitor: Add Monitor parameter to monitor_set_cpu()

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

error: Use error_fatal to simplify obvious fatal errors (again)

Patch created mechanically by rerunning:

    $ spatch --in-place --sp-file scripts/coccinelle/use-error_fatal.cocci \
     --macro-file scripts/cocci-macro-file.h --use-gitgrep .

Variables now unused dropped manually.

Cc: Eric Auger <eric.auger@redhat.com>
Cc: David Hildenbrand <david@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20200722084048.1726105-5-armbru@redhat.com>
Reviewed-by: David Hildenbrand <david@redhat.com>

error: Remove NULL checks on error_propagate() calls (again)

Patch created mechanically by rerunning:

    $ spatch --sp-file scripts/coccinelle/error_propagate_null.cocci \
             --macro-file scripts/cocci-macro-file.h \
             --use-gitgrep .

Cc: Jens Freimann <jfreimann@redhat.com>
Cc: Hailiang Zhang <zhang.zhanghailiang@huawei.com>
Cc: Juan Quintela <quintela@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20200722084048.1726105-4-armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>

block: Convert 'block_resize' to coroutine

block_resize performs some I/O that could potentially take quite some
time, so use it as an example for the new 'coroutine': true annotation
in the QAPI schema.

bdrv_truncate() requires that we're already in the right AioContext for
the BlockDriverState if called in coroutine context. So instead of just
taking the AioContext lock, move the QMP handler coroutine to the
context.

Call blk_unref() only after switching back because blk_unref() may only
be called in the main thread.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20201005155855.256490-15-kwolf@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>

block: Add bdrv_lock()/unlock()

Inside of coroutine context, we can't directly use aio_context_acquire()
for the AioContext of a block node because we already own the lock of
the current AioContext and we need to avoid double locking to prevent
deadlocks.

This provides helper functions to lock the AioContext of a node only if
it's not the same as the current AioContext.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20201005155855.256490-14-kwolf@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>

block: Add bdrv_co_enter()/leave()

Add a pair of functions to temporarily move the current coroutine to the
AioContext of a given BlockDriverState.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20201005155855.256490-13-kwolf@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>

util/async: Add aio_co_reschedule_self()

Add a function that can be used to move the currently running coroutine
to a different AioContext (and therefore potentially a different
thread).

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20201005155855.256490-12-kwolf@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>

hmp: Add support for coroutine command handlers

Often, QMP command handlers are not only called to handle QMP commands,
but also from a corresponding HMP command handler. In order to give them
a consistent environment, optionally run HMP command handlers in a
coroutine, too.

The implementation is a lot simpler than in QMP because for HMP, we
still block the VM while the coroutine is running.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Message-Id: <20201005155855.256490-11-kwolf@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>

qmp: Move dispatcher to a coroutine

This moves the QMP dispatcher to a coroutine and runs all QMP command
handlers that declare 'coroutine': true in coroutine context so they
can avoid blocking the main loop while doing I/O or waiting for other
events.

For commands that are not declared safe to run in a coroutine, the
dispatcher drops out of coroutine context by calling the QMP command
handler from a bottom half.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20201005155855.256490-10-kwolf@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>

qapi: Add a 'coroutine' flag for commands

This patch adds a new 'coroutine' flag to QMP command definitions that
tells the QMP dispatcher that the command handler is safe to be run in a
coroutine.

The documentation of the new flag pretends that this flag is already
used as intended, which it isn't yet after this patch. We'll implement
this in another patch in this series.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20201005155855.256490-9-kwolf@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>

monitor: Make current monitor a per-coroutine property

This way, a monitor command handler will still be able to access the
current monitor, but when it yields, all other code code will correctly
get NULL from monitor_cur().

This uses a hash table to map the coroutine pointer to the current
monitor of that coroutine.  Outside of coroutine context, we associate
the current monitor with the leader coroutine of the current thread.

Approaches to implement some form of coroutine local storage directly in
the coroutine core code have been considered and discarded because they
didn't end up being much more generic than the hash table and their
performance impact on coroutines not using coroutine local storage was
unclear. As the block layer uses a coroutine per I/O request, this is a
fast path and we have to be careful. It's safest to just stay out of
this path with code only used by the monitor.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20201005155855.256490-8-kwolf@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>

qmp: Call monitor_set_cur() only in qmp_dispatch()

The correct way to set the current monitor for a coroutine handler will
be different than for a blocking handler, so monitor_set_cur() needs to
be called in qmp_dispatch().

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20201005155855.256490-7-kwolf@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>

qmp: Assert that no other monitor is active

monitor_qmp_dispatch() is never supposed to be called in the context of
another monitor, so assert that monitor_cur() is NULL instead of saving
and restoring it.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20201005155855.256490-6-kwolf@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>

hmp: Update current monitor only in handle_hmp_command()

The current monitor is updated relatively early in the command handling
code even though only the command handler actually needs it.

The current monitor will become coroutine-local later, so we can only
update it when we know in which coroutine the command will be exectued.
Move it to handle_hmp_command() where this information will be
available.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20201005155855.256490-5-kwolf@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>

monitor: Use getter/setter functions for cur_mon

cur_mon really needs to be coroutine-local as soon as we move monitor
command handlers to coroutines and let them yield. As a first step, just
remove all direct accesses to cur_mon so that we can implement this in
the getter function later.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20201005155855.256490-4-kwolf@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>

monitor: Add Monitor parameter to monitor_get_cpu_index()

Most callers actually don't have to rely on cur_mon, but already know
for which monitor they call monitor_get_cpu_index().

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20201005155855.256490-3-kwolf@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>

monitor: Add Monitor parameter to monitor_set_cpu()

Most callers actually don't have to rely on cur_mon, but already know
for which monitor they call monitor_set_cpu().

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20201005155855.256490-2-kwolf@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>

specs/ppc-spapr-numa: update with new NUMA support

This update provides more in depth information about the
choices and drawbacks of the new NUMA support for the
spapr machine.

Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Message-Id: <20201007172849.302240-6-danielhb413@gmail.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

spapr_numa: consider user input when defining associativity

A new function called spapr_numa_define_associativity_domains()
is created to calculate the associativity domains and change
the associativity arrays considering user input. This is how
the associativity domain between two NUMA nodes A and B is
calculated:

- get the distance D between them

- get the correspondent NUMA level 'n_level' for D. This is done
via a helper called spapr_numa_get_numa_level()

- all associativity arrays were initialized with their own
numa_ids, and we're calculating the distance in node_id ascending
order, starting from node id 0 (the first node retrieved by
numa_state). This will have a cascade effect in the algorithm because
the associativity domains that node 0 defines will be carried over to
other nodes, and node 1 associativities will be carried over after
taking node 0 associativities into account, and so on. This
happens because we'll assign assoc_src as the associativity domain
of dst as well, for all NUMA levels beyond and including n_level.

The PPC kernel expects the associativity domains of the first node
(node id 0) to be always 0 [1], and this algorithm will grant that
by default.

Ultimately, all of this results in a best effort approximation for
the actual NUMA distances the user input in the command line. Given
the nature of how PAPR itself interprets NUMA distances versus the
expectations risen by how ACPI SLIT works, there might be better
algorithms but, in the end, it'll also result in another way to
approximate what the user really wanted.

To keep this commit message no longer than it already is, the next
patch will update the existing documentation in ppc-spapr-numa.rst
with more in depth details and design considerations/drawbacks.

[1] https://lore.kernel.org/linuxppc-dev/5e8fbea3-8faf-0951-172a-b41a2138fbcf@gmail.com/

Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Message-Id: <20201007172849.302240-5-danielhb413@gmail.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

spapr_numa: change reference-points and maxdomain settings

This is the first guest visible change introduced in
spapr_numa.c. The previous settings of both reference-points
and maxdomains were too restrictive, but enough for the
existing associativity we're setting in the resources.

We'll change that in the following patches, populating the
associativity arrays based on user input. For those changes
to be effective, reference-points and maxdomains must be
more flexible. After this patch, we'll have 4 distinct
levels of NUMA (0x4, 0x3, 0x2, 0x1) and maxdomains will
allow for any type of configuration the user intends to
do - under the scope and limitations of PAPR itself, of
course.

Reviewed-by: Greg Kurz <groug@kaod.org>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Message-Id: <20201007172849.302240-4-danielhb413@gmail.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

spapr_numa: forbid asymmetrical NUMA setups

The pSeries machine does not support asymmetrical NUMA
configurations. This doesn't make much of a different
since we're not using user input for pSeries NUMA setup,
but this will change in the next patches.

To avoid breaking existing setups, gate this change by
checking for legacy NUMA support.

Reviewed-by: Greg Kurz <groug@kaod.org>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Message-Id: <20201007172849.302240-3-danielhb413@gmail.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

spapr: add spapr_machine_using_legacy_numa() helper

The changes to come to NUMA support are all guest visible. In
theory we could just create a new 5_1 class option flag to
avoid the changes to cascade to 5.1 and under. The reality is that
these changes are only relevant if the machine has more than one
NUMA node. There is no need to change guest behavior that has
been around for years needlesly.

This new helper will be used by the next patches to determine
whether we should retain the (soon to be) legacy NUMA behavior
in the pSeries machine. The new behavior will only be exposed
if:

- machine is pseries-5.2 and newer;
- more than one NUMA node is declared in NUMA state.

Reviewed-by: Greg Kurz <groug@kaod.org>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Message-Id: <20201007172849.302240-2-danielhb413@gmail.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

ppc/pnv: Increase max firmware size

Builds enabling GCOV can be bigger than 4MB and the limit on FSP
systems is 16MB.

Signed-off-by: Cédric Le Goater <clg@kaod.org>
Message-Id: <20201002091440.1349326-1-clg@kaod.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

spapr: Add a return value to spapr_check_pagesize()

As recommended in "qapi/error.h", return true on success and false on
failure. This allows to reduce error propagation overhead in the callers.

Signed-off-by: Greg Kurz <groug@kaod.org>
Message-Id: <20200914123505.612812-14-groug@kaod.org>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

spapr: Add a return value to spapr_nvdimm_validate()

As recommended in "qapi/error.h", return true on success and false on
failure. This allows to reduce error propagation overhead in the callers.

Signed-off-by: Greg Kurz <groug@kaod.org>
Message-Id: <20200914123505.612812-13-groug@kaod.org>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

spapr: Simplify error handling in spapr_cpu_core_realize()

As recommended in "qapi/error.h", add a bool return value to
spapr_realize_vcpu() and use it in spapr_cpu_core_realize()
in order to get rid of the error propagation overhead.

Signed-off-by: Greg Kurz <groug@kaod.org>
Message-Id: <20200914123505.612812-12-groug@kaod.org>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

spapr: Add a return value to spapr_set_vcpu_id()

As recommended in "qapi/error.h", return true on success and false on
failure. This allows to reduce error propagation overhead in the callers.

Signed-off-by: Greg Kurz <groug@kaod.org>
Message-Id: <20200914123505.612812-11-groug@kaod.org>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

spapr: Simplify error handling in prop_get_fdt()

Use the return value of visit_check_struct() and visit_check_list()
for error checking instead of local_err. This allows to get rid of
the error propagation overhead.

Signed-off-by: Greg Kurz <groug@kaod.org>
Message-Id: <20200914123505.612812-10-groug@kaod.org>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

spapr: Add a return value to spapr_drc_attach()

As recommended in "qapi/error.h", return true on success and false on
failure. This allows to reduce error propagation overhead in the callers.

Signed-off-by: Greg Kurz <groug@kaod.org>
Message-Id: <20200914123505.612812-9-groug@kaod.org>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

spapr: Simplify error handling in spapr_vio_busdev_realize()

Use the return value of spapr_irq_findone() and spapr_irq_claim()
to detect failures. This allows to reduce the error propagation
overhead.

Signed-off-by: Greg Kurz <groug@kaod.org>
Message-Id: <20200914123505.612812-8-groug@kaod.org>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

spapr: Simplify error handling in do_client_architecture_support()

Use the return value of ppc_set_compat_all() to check failures,
which is preferred over hijacking local_err.

Signed-off-by: Greg Kurz <groug@kaod.org>
Message-Id: <20200914123505.612812-7-groug@kaod.org>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

spapr: Get rid of cas_check_pvr() error reporting

The cas_check_pvr() function has two purposes:
- finding the "best" logical PVR, ie. the most recent one supported by
  the guest for this CPU type
- checking if the guest supports the real PVR of this CPU type, which
  is just an optional extra information to workaround the lack of
  support for "compat" mode in PR KVM

This logic doesn't need error reporting, really. If we don't find a
suitable logical PVR, we return the special value 0 which is definitely
not a valid PVR. Let the caller decide on whether it should error out
or not.

This doesn't change the behavior.

Signed-off-by: Greg Kurz <groug@kaod.org>
Message-Id: <20200914123505.612812-6-groug@kaod.org>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

spapr: Simplify error handling in callers of ppc_set_compat()

Now that ppc_set_compat() indicates success/failure with a return
value, use it and reduce error propagation overhead.

Signed-off-by: Greg Kurz <groug@kaod.org>
Message-Id: <20200914123505.612812-5-groug@kaod.org>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

ppc: Fix return value in cpu_post_load() error path

VMState handlers are supposed to return negative errno values on failure.

Signed-off-by: Greg Kurz <groug@kaod.org>
Message-Id: <20200914123505.612812-4-groug@kaod.org>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

ppc: Add a return value to ppc_set_compat() and ppc_set_compat_all()

As recommended in "qapi/error.h", indicate success / failure with a
return value. Since ppc_set_compat() is called from a VMState handler,
let's make it an int so that it propagates any negative errno returned
by kvmppc_set_compat(). Do the same for ppc_set_compat_all() for
consistency, even if it isn't called in a context where a negative errno
is required on failure.

This will allow to simplify error handling in the callers.

Signed-off-by: Greg Kurz <groug@kaod.org>
Message-Id: <20200914123505.612812-3-groug@kaod.org>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

spapr: Fix error leak in spapr_realize_vcpu()

If spapr_irq_cpu_intc_create() fails, local_err isn't propagated and
thus leaked.

Fixes: 992861fb1e4c ("error: Eliminate error_propagate() manually")
Cc: armbru@redhat.com
Signed-off-by: Greg Kurz <groug@kaod.org>
Message-Id: <20200914123505.612812-2-groug@kaod.org>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

spapr: Handle HPT allocation failure in nested guest

The nested KVM code does not yet support HPT guests. Calling the
KVM_CAP_PPC_ALLOC_HTAB ioctl currently leads to KVM setting the guest
as HPT and erroneously executing code in L1 that should only run in
hypervisor mode, leading to an exception in the L1 vcpu thread when it
enters the nested guest.

This can be reproduced with -machine max-cpu-compat=power8 in the L2
guest command line.

The KVM code has since been modified to fail the ioctl when running in
a nested environment so QEMU needs to be able to handle that. This
patch provides an error message informing the user about the lack of
support for HPT in nested guests.

Reported-by: Satheesh Rajendran <sathnaga@linux.vnet.ibm.com>
Signed-off-by: Fabiano Rosas <farosas@linux.ibm.com>
Message-Id: <20200911043123.204162-1-farosas@linux.ibm.com>
Reviewed-by: Greg Kurz <groug@kaod.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

Merge remote-tracking branch 'remotes/pmaydell/tags/pull-target-arm-20201008-1' into staging

target-arm queue:
 * hw/ssi/npcm7xx_fiu: Fix handling of unsigned integer
 * hw/arm/fsl-imx25: Fix a typo
 * hw/arm/sbsa-ref : Fix SMMUv3 Initialisation
 * hw/arm/sbsa-ref : allocate IRQs for SMMUv3
 * hw/char/bcm2835_aux: Allow less than 32-bit accesses
 * hw/arm/virt: Implement kvm-steal-time
 * target/arm: Make '-cpu max' have a 48-bit PA

# gpg: Signature made Thu 08 Oct 2020 21:40:31 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-20201008-1:
  target/arm: Make '-cpu max' have a 48-bit PA
  hw/arm/virt: Implement kvm-steal-time
  tests/qtest: Restore aarch64 arm-cpu-features test
  hw/arm/virt: Move kvm pmu setup to virt_cpu_post_init
  hw/arm/virt: Move post cpu realize check into its own function
  target/arm/kvm: Make uncalled stubs explicitly unreachable
  linux headers: sync to 5.9-rc7
  hw/char/bcm2835_aux: Allow less than 32-bit accesses
  hw/arm/sbsa-ref : allocate IRQs for SMMUv3
  hw/arm/sbsa-ref : Fix SMMUv3 Initialisation
  hw/arm/fsl-imx25: Fix a typo
  hw/ssi/npcm7xx_fiu: Fix handling of unsigned integer

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

target/arm: Make '-cpu max' have a 48-bit PA

QEMU supports a 48-bit physical address range, but we don't currently
expose it in the '-cpu max' ID registers (you get the same range as
Cortex-A57, which is 44 bits).

Set the ID_AA64MMFR0.PARange field to indicate 48 bits.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20201001160116.18095-1-peter.maydell@linaro.org

Merge remote-tracking branch 'remotes/rth/tags/pull-tcg-20201008' into staging

Extend maximum gvec vector size
Fix i386 avx2 dupi
Fix mips host user-only write detection
Misc cleanups.

# gpg: Signature made Thu 08 Oct 2020 13:55:22 BST
# gpg:                using RSA key 7A481E78868B4DB6A85A05C064DF38E8AF7E215F
# gpg:                issuer "richard.henderson@linaro.org"
# gpg: Good signature from "Richard Henderson <richard.henderson@linaro.org>" [full]
# Primary key fingerprint: 7A48 1E78 868B 4DB6 A85A  05C0 64DF 38E8 AF7E 215F

* remotes/rth/tags/pull-tcg-20201008:
  accel/tcg: Fix computing of is_write for MIPS
  tcg: Remove TCG_TARGET_HAS_cmp_vec
  tcg/optimize: Fold dup2_vec
  tcg: Fix generation of dupi_vec for 32-bit host
  tcg/i386: Fix dupi for avx2 32-bit hosts
  tcg: Remove TCGOpDef.used
  tcg: Move some TCG_CT_* bits to TCGArgConstraint bitfields
  tcg: Remove TCG_CT_REG
  tcg: Move sorted_args into TCGArgConstraint.sort_index
  tcg: Drop union from TCGArgConstraint
  tcg: Adjust simd_desc size encoding

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

hw/arm/virt: Implement kvm-steal-time

We add the kvm-steal-time CPU property and implement it for machvirt.
A tiny bit of refactoring was also done to allow pmu and pvtime to
use the same vcpu device helper functions.

Reviewed-by: Eric Auger <eric.auger@redhat.com>
Signed-off-by: Andrew Jones <drjones@redhat.com>
Message-id: 20201001061718.101915-7-drjones@redhat.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

tests/qtest: Restore aarch64 arm-cpu-features test

arm-cpu-features got dropped from the AArch64 tests during the meson
conversion shuffle.

Signed-off-by: Andrew Jones <drjones@redhat.com>
Message-id: 20201001061718.101915-6-drjones@redhat.com
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

hw/arm/virt: Move kvm pmu setup to virt_cpu_post_init

Move the KVM PMU setup part of fdt_add_pmu_nodes() to
virt_cpu_post_init(), which is a more appropriate location. Now
fdt_add_pmu_nodes() is also named more appropriately, because it
no longer does anything but fdt node creation.

No functional change intended.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Eric Auger <eric.auger@redhat.com>
Signed-off-by: Andrew Jones <drjones@redhat.com>
Message-id: 20201001061718.101915-5-drjones@redhat.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

hw/arm/virt: Move post cpu realize check into its own function

We'll add more to this new function in coming patches so we also
state the gic must be created and call it below create_gic().

No functional change intended.

Reviewed-by: Eric Auger <eric.auger@redhat.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Andrew Jones <drjones@redhat.com>
Message-id: 20201001061718.101915-4-drjones@redhat.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm/kvm: Make uncalled stubs explicitly unreachable

When we compile without KVM support !defined(CONFIG_KVM) we generate
stubs for functions that the linker will still encounter. Sometimes
these stubs can be executed safely and are placed in paths where they
get executed with or without KVM. Other functions should never be
called without KVM. Those functions should be guarded by kvm_enabled(),
but should also be robust to refactoring mistakes. Putting a
g_assert_not_reached() in the function should help. Additionally,
the g_assert_not_reached() calls may actually help the linker remove
some code.

We remove the stubs for kvm_arm_get/put_virtual_time(), as they aren't
necessary at all - the only caller is in kvm.c

Reviewed-by: Eric Auger <eric.auger@redhat.com>
Signed-off-by: Andrew Jones <drjones@redhat.com>
Message-id: 20201001061718.101915-3-drjones@redhat.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

linux headers: sync to 5.9-rc7

Update against Linux 5.9-rc7.

Cc: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Andrew Jones <drjones@redhat.com>
Message-id: 20201001061718.101915-2-drjones@redhat.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

hw/char/bcm2835_aux: Allow less than 32-bit accesses

The "BCM2835 ARM Peripherals" datasheet [*] chapter 2
("Auxiliaries: UART1 & SPI1, SPI2"), list the register
sizes as 3/8/16/32 bits. We assume this means this
peripheral allows 8-bit accesses.

This was not an issue until commit 5d971f9e67 which reverted
("memory: accept mismatching sizes in memory_region_access_valid").

The model is implemented as 32-bit accesses (see commit 97398d900c,
all registers are 32-bit) so replace MemoryRegionOps.valid as
MemoryRegionOps.impl, and re-introduce MemoryRegionOps.valid
with a 8/32-bit range.

[*] https://www.raspberrypi.org/app/uploads/2012/02/BCM2835-ARM-Peripherals.pdf

Fixes: 97398d900c ("bcm2835_aux: add emulation of BCM2835 AUX (aka UART1) block")
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-id: 20201002181032.1899463-1-f4bug@amsat.org
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

hw/arm/sbsa-ref : allocate IRQs for SMMUv3

Original commit did not allocate IRQs for the SMMUv3 in the irqmap
effectively using irq 0->3 (shared with other devices). Assuming
original intent was to allocate unique IRQs then add an allocation
to the irqmap.

Fixes: e9fdf453240 ("hw/arm: Add arm SBSA reference machine, devices part")
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Graeme Gregory <graeme@nuviainc.com>
Reviewed-by: Eric Auger <eric.auger@redhat.com>
Message-id: 20201007100732.4103790-3-graeme@nuviainc.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

hw/arm/sbsa-ref : Fix SMMUv3 Initialisation

SMMUv3 has an error in a previous patch where an i was transposed to a 1
meaning interrupts would not have been correctly assigned to the SMMUv3
instance.

Fixes: 48ba18e6d3f3 ("hw/arm/sbsa-ref: Simplify by moving the gic in the machine state")
Signed-off-by: Graeme Gregory <graeme@nuviainc.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Eric Auger <eric.auger@redhat.com>
Message-id: 20201007100732.4103790-2-graeme@nuviainc.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

hw/arm/fsl-imx25: Fix a typo

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-id: 20201002080935.1660005-1-f4bug@amsat.org
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

hw/ssi/npcm7xx_fiu: Fix handling of unsigned integer

Fix integer handling issues handling issue reported by Coverity:

  hw/ssi/npcm7xx_fiu.c: 162 in npcm7xx_fiu_flash_read()
  >>>     CID 1432730:  Integer handling issues  (NEGATIVE_RETURNS)
  >>>     "npcm7xx_fiu_cs_index(fiu, f)" is passed to a parameter that cannot be negative.
  162         npcm7xx_fiu_select(fiu, npcm7xx_fiu_cs_index(fiu, f));

  hw/ssi/npcm7xx_fiu.c: 221 in npcm7xx_fiu_flash_write()
  218         cs_id = npcm7xx_fiu_cs_index(fiu, f);
  219         trace_npcm7xx_fiu_flash_write(DEVICE(fiu)->canonical_path, cs_id, addr,
  220                                       size, v);
  >>>     CID 1432729:  Integer handling issues  (NEGATIVE_RETURNS)
  >>>     "cs_id" is passed to a parameter that cannot be negative.
  221         npcm7xx_fiu_select(fiu, cs_id);

Since the index of the flash can not be negative, return an
unsigned type.

Reported-by: Coverity (CID 1432729 & 1432730: NEGATIVE_RETURNS)
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Havard Skinnemoen <hskinnemoen@google.com>
Message-id: 20200919132435.310527-1-f4bug@amsat.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

accel/tcg: Fix computing of is_write for MIPS

Detect all MIPS store instructions in cpu_signal_handler for all available
MIPS versions, and set is_write if encountering such store instructions.

This fixed the error while dealing with self-modified code for MIPS.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Kele Huang <kele.hwang@gmail.com>
Signed-off-by: Xu Zou <iwatchnima@gmail.com>
Message-Id: <20201002081420.10814-1-kele.hwang@gmail.com>
[rth: Use uintptr_t for pc to fix n32 build error.]
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

tcg: Remove TCG_TARGET_HAS_cmp_vec

The cmp_vec opcode is mandatory; this symbol is unused.

Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

tcg/optimize: Fold dup2_vec

When the two arguments are identical, this can be reduced to
dup_vec or to mov_vec from a tcg_constant_vec.

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

tcg: Fix generation of dupi_vec for 32-bit host

The definition of INDEX_op_dupi_vec is that it operates on
units of tcg_target_ulong -- in this case 32 bits.  It does
not work to use this for a uint64_t value that happens to be
small enough to fit in tcg_target_ulong.

Fixes: d2fd745fe8b
Fixes: db432672dc5
Cc: qemu-stable@nongnu.org
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

tcg/i386: Fix dupi for avx2 32-bit hosts

The previous change wrongly stated that 32-bit avx2 should have
used VPBROADCASTW.  But that's a 16-bit broadcast and we want a
32-bit broadcast.

Fixes: 7b60ef3264e
Cc: qemu-stable@nongnu.org
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

tcg: Remove TCGOpDef.used

The last user of this field disappeared in f69d277ece4.

Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

tcg: Move some TCG_CT_* bits to TCGArgConstraint bitfields

These are easier to set and test when they have their own fields.
Reduce the size of alias_index and sort_index to 4 bits, which is
sufficient for TCG_MAX_OP_ARGS.  This leaves only the bits indicating
constants within the ct field.

Move all initialization to allocation time, rather than init
individual fields in process_op_defs.

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

tcg: Remove TCG_CT_REG

This wasn't actually used for anything, really.  All variable
operands must accept registers, and which are indicated by the
set in TCGArgConstraint.regs.

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

tcg: Move sorted_args into TCGArgConstraint.sort_index

This uses an existing hole in the TCGArgConstraint structure
and will be convenient for keeping the data in one place.

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

tcg: Drop union from TCGArgConstraint

The union is unused; let "regs" appear in the main structure
without the "u.regs" wrapping.

Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

tcg: Adjust simd_desc size encoding

With larger vector sizes, it turns out oprsz == maxsz, and we only
need to represent mismatch for oprsz <= 32.  We do, however, need
to represent larger oprsz and do so without reducing SIMD_DATA_BITS.

Reduce the size of the oprsz field and increase the maxsz field.
Steal the oprsz value of 24 to indicate equality with maxsz.

Tested-by: Frank Chang <frank.chang@sifive.com>
Reviewed-by: Frank Chang <frank.chang@sifive.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

Merge remote-tracking branch 'remotes/ehabkost/tags/machine-next-pull-request' into staging

machine + QOM queue, 2020-10-06

* QOM documentation fixes and cleanups (Eduardo Habkost)
* user-mode: Prune build dependencies (Philippe Mathieu-Daudé)
* qom: Improve error message (Philippe Mathieu-Daudé)
* numa: hmat: require parent cache description before the next
  level one (Igor Mammedov)

# gpg: Signature made Tue 06 Oct 2020 23:09:03 BST
# gpg:                using RSA key 5A322FD5ABC4D3DBACCFD1AA2807936F984DC5A6
# gpg:                issuer "ehabkost@redhat.com"
# gpg: Good signature from "Eduardo Habkost <ehabkost@redhat.com>" [full]
# Primary key fingerprint: 5A32 2FD5 ABC4 D3DB ACCF  D1AA 2807 936F 984D C5A6

* remotes/ehabkost/tags/machine-next-pull-request: (21 commits)
  numa: hmat: require parent cache description before the next level one
  kernel-doc: Remove $decl_type='type name' hack
  memory: Explicitly tag doc comments for structs
  qom: Explicitly tag doc comments for typedefs and structs
  kernel-doc: Handle function typedefs without asterisks
  kernel-doc: Handle function typedefs that return pointers
  docs/devel/qom: Avoid long lines
  docs/devel/qom: Remove usage of <code>
  docs/devel/qom: Use *emphasis* for emphasis
  docs/devel/qom: Fix indentation of code blocks
  docs/devel/qom: Fix indentation of bulleted list
  qom: Fix DECLARE_*CHECKER documentation
  qom: Improve error message displayed with missing object properties
  hw/core/cpu: Add missing 'exec/cpu-common.h' include
  hw/core/qdev-properties: Extract system-mode specific properties
  hw/core/qdev-properties: Export some integer-related functions
  hw/core/qdev-properties: Export qdev_prop_enum
  hw/core/qdev-properties: Export enum-related functions
  hw/core/qdev-properties: Fix code style
  hw/core/qdev-properties: Use qemu_strtoul() in set_pci_host_devaddr()
  ...

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

Merge remote-tracking branch 'remotes/huth-gitlab/tags/pull-request-2020-10-06' into staging

* Don't stop at the first unbootable device, continue scanning
* Fix corner cases in booting from ECKD
* s390x-ccw bios cleanup part 2​

# gpg: Signature made Tue 06 Oct 2020 19:17:46 BST
# gpg:                using RSA key 27B88847EEE0250118F3EAB92ED9D774FE702DB5
# gpg:                issuer "thuth@redhat.com"
# gpg: Good signature from "Thomas Huth <th.huth@gmx.de>" [full]
# gpg:                 aka "Thomas Huth <thuth@redhat.com>" [full]
# gpg:                 aka "Thomas Huth <huth@tuxfamily.org>" [full]
# gpg:                 aka "Thomas Huth <th.huth@posteo.de>" [unknown]
# Primary key fingerprint: 27B8 8847 EEE0 2501 18F3  EAB9 2ED9 D774 FE70 2DB5

* remotes/huth-gitlab/tags/pull-request-2020-10-06:
  tests/qtest/cdrom: Add more s390x-related boot tests
  pc-bios/s390: Update the s390-ccw bios binaries
  pc-bios: s390x: Go into disabled wait when encountering a PGM exception
  pc-bios: s390x: Use reset PSW if avaliable
  pc-bios: s390x: Save PSW rework
  pc-bios: s390x: Fix bootmap.c zipl component entry data handling
  pc-bios/s390-ccw: break loop if a null block number is reached
  pc-bios/s390-ccw: fix off-by-one error
  pc-bios/s390-ccw/main: Remove superfluous call to enable_subchannel()
  pc-bios/s390-ccw: Allow booting in case the first virtio-blk disk is bad
  pc-bios/s390-ccw: Scan through all devices if no boot device specified
  pc-bios/s390-ccw: Do not bail out early if not finding a SCSI disk
  pc-bios/s390-ccw: Move the inner logic of find_subch() to a separate function
  pc-bios/s390-ccw: Introduce ENODEV define and remove guards of others
  pc-bios/s390-ccw: Move ipl-related code from main() into a separate function
  pc-bios/s390-ccw/Makefile: Compile with -std=gnu99, -fwrapv and -fno-common

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

tests/qtest/cdrom: Add more s390x-related boot tests

Let's add two new tests:

1) Booting with "bootindex" is the architected default behavior on the
s390x target, so we should have at least one test that is using the
"bootindex" property.

2) The s390-ccw bios used to fail when other unbootable devices have
been specified before the bootable device (without "bootindex"). Now
that the s390-ccw bios is a little bit smarter here, we should test
this scenario, too, to avoid regressions.

Message-Id: <20200806105349.632-10-thuth@redhat.com>
Reviewed-by: Cornelia Huck <cohuck@redhat.com>
Acked-by: Janosch Frank <frankja@linux.ibm.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>

pc-bios/s390: Update the s390-ccw bios binaries

Make sure that the binaries match the current state of the sources.

Signed-off-by: Thomas Huth <thuth@redhat.com>

pc-bios: s390x: Go into disabled wait when encountering a PGM exception

Let's setup a PGM PSW, so we won't load 0s when a program exception
happens. Instead we'll load a disabled wait PSW.

Signed-off-by: Janosch Frank <frankja@linux.ibm.com>
Reviewed-by: Christian Borntraeger <borntraeger@de.ibm.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Message-Id: <20201006094249.50640-5-frankja@linux.ibm.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>

pc-bios: s390x: Use reset PSW if avaliable

If a blob provides a reset PSW then we should use it instead of
branching to the PSW address and using our own mask.

Signed-off-by: Janosch Frank <frankja@linux.ibm.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Message-Id: <20201006094249.50640-4-frankja@linux.ibm.com>
[thuth: Use Elvis operator to shorten long line]
Signed-off-by: Thomas Huth <thuth@redhat.com>

pc-bios: s390x: Save PSW rework

We don't need to save the ipl_continue variable in lowcore and have it
limited to 32 bits because of the lowcore layout. Let's move it to a
new 64 bit variable and get rid of the reset info struct.

Signed-off-by: Janosch Frank <frankja@linux.ibm.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Message-Id: <20201006094249.50640-3-frankja@linux.ibm.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>