OSDN Git Service

qmiga/qemu.git
2 years agotests/qapi-schema: Demonstrate broken C code for 'if'
Markus Armbruster [Tue, 31 Aug 2021 12:38:01 +0000 (14:38 +0200)]
tests/qapi-schema: Demonstrate broken C code for 'if'

The C code generated for 'if' conditionals is incorrectly
parenthesized.  For instance,

    'if': { 'not': { 'any': [ { 'not': 'TEST_IF_EVT' },
      { 'not': 'TEST_IF_STRUCT' } ] } } }

generates

    #if !(!defined(TEST_IF_EVT)) || (!defined(TEST_IF_STRUCT))

This is wrong.  Correct would be:

    #if !(!defined(TEST_IF_EVT) || !defined(TEST_IF_STRUCT))

Cover the issue in qapi-schema-test.json.  This generates bad #if in
tests/test-qapi-events.h and other files.

Add a similar condition to doc-good.json.  The generated documentation
is fine.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20210831123809.1107782-5-armbru@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
2 years agotests/qapi-schema: Correct two 'if' conditionals
Markus Armbruster [Tue, 31 Aug 2021 12:38:00 +0000 (14:38 +0200)]
tests/qapi-schema: Correct two 'if' conditionals

A definition's conditional should imply the conditionals of types it
uses.  If it doesn't, some configurations won't compile.

Example (from tests/qapi-schema/qapi-schema-test.json):

    { 'union': 'TestIfUnion', 'data':
      { 'foo': 'TestStruct',
'bar': { 'type': 'str', 'if': 'TEST_IF_UNION_BAR'} },
      'if': { 'all': ['TEST_IF_UNION', 'TEST_IF_STRUCT'] } }

    { 'command': 'test-if-union-cmd',
      'data': { 'union-cmd-arg': 'TestIfUnion' },
      'if': 'TEST_IF_UNION' }

generates

    #if (defined(TEST_IF_UNION)) && (defined(TEST_IF_STRUCT))
    typedef struct TestIfUnion TestIfUnion;
    #endif /* (defined(TEST_IF_UNION)) && (defined(TEST_IF_STRUCT)) */

and

    #if defined(TEST_IF_UNION)
    void qmp_test_if_union_cmd(TestIfUnion *union_cmd_arg, Error **errp);
    void qmp_marshal_test_if_union_cmd(QDict *args, QObject **ret, Error **errp);
    #endif /* defined(TEST_IF_UNION) */

which doesn't compile when !defined(TEST_IF_STRUCT).

Messed up in f8c4fdd6ae "tests/qapi: Cover commands with 'if' and
union / alternate 'data'", v4.0.0.  Harmless, as we don't actually use
this configuration.  Correct it anyway, along with another instance.

This loses coverage for 'not'.  The next commit will bring it back.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20210831123809.1107782-4-armbru@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
2 years agoqapi: Simplify how QAPISchemaIfCond represents "no condition"
Markus Armbruster [Tue, 31 Aug 2021 12:37:59 +0000 (14:37 +0200)]
qapi: Simplify how QAPISchemaIfCond represents "no condition"

None works fine, there is no need to replace it by {} in .__init__().

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20210831123809.1107782-3-armbru@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
2 years agoqapi: Simplify QAPISchemaIfCond's interface for generating C
Markus Armbruster [Tue, 31 Aug 2021 12:37:58 +0000 (14:37 +0200)]
qapi: Simplify QAPISchemaIfCond's interface for generating C

QAPISchemaIfCond.cgen() is only ever used like

    gen_if(ifcond.cgen())

and

    gen_endif(ifcond.cgen())

Simplify to

    ifcond.gen_if()

and

    ifcond.gen_endif()

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20210831123809.1107782-2-armbru@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
[Import statements tidied up with isort]

2 years agoqapi: Set boolean value correctly in examples
Guoyi Tu [Fri, 27 Aug 2021 09:06:27 +0000 (17:06 +0800)]
qapi: Set boolean value correctly in examples

Signed-off-by: Guoyi Tu <tugy@chinatelecom.cn>
Message-Id: <a21a2b61-2653-a2c9-4478-715e5fb19120@chinatelecom.cn>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
2 years agoMerge remote-tracking branch 'remotes/cschoenebeck/tags/pull-9p-20210902' into staging
Peter Maydell [Fri, 3 Sep 2021 07:27:38 +0000 (08:27 +0100)]
Merge remote-tracking branch 'remotes/cschoenebeck/tags/pull-9p-20210902' into staging

9pfs: misc patches

* Fix an occasional crash when handling 'Twalk' requests.

* Two code cleanup patches.

# gpg: Signature made Thu 02 Sep 2021 12:42:32 BST
# gpg:                using RSA key 96D8D110CF7AF8084F88590134C2B58765A47395
# gpg:                issuer "qemu_oss@crudebyte.com"
# gpg: Good signature from "Christian Schoenebeck <qemu_oss@crudebyte.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: ECAB 1A45 4014 1413 BA38  4926 30DB 47C3 A012 D5F4
#      Subkey fingerprint: 96D8 D110 CF7A F808 4F88  5901 34C2 B587 65A4 7395

* remotes/cschoenebeck/tags/pull-9p-20210902:
  9pfs: fix crash in v9fs_walk()
  hw/9pfs: use g_autofree in v9fs_walk() where possible
  hw/9pfs: avoid 'path' copy in v9fs_walk()

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
2 years agoMerge remote-tracking branch 'remotes/a1xndr/tags/fuzz-pull-2021-09-01' into staging
Peter Maydell [Thu, 2 Sep 2021 13:59:05 +0000 (14:59 +0100)]
Merge remote-tracking branch 'remotes/a1xndr/tags/fuzz-pull-2021-09-01' into staging

Fuzzing Patches for 2021-09-01

# gpg: Signature made Wed 01 Sep 2021 12:42:00 BST
# gpg:                using RSA key FAD4E2BF871375D6340517C44E661DDE583A964E
# gpg: Good signature from "Alexander Bulekov <alxndr@bu.edu>" [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: FAD4 E2BF 8713 75D6 3405  17C4 4E66 1DDE 583A 964E

* remotes/a1xndr/tags/fuzz-pull-2021-09-01:
  MAINTAINERS: add fuzzing reviewer
  MAINTAINERS: Add myself as a reviewer for Device Fuzzing
  fuzz: unblock SIGALRM so the timeout works
  fuzz: use ITIMER_REAL for timeouts
  fuzz: add an instrumentation filter
  fuzz: make object-name matching case-insensitive
  fuzz: adjust timeout to allow for longer inputs
  fuzz: fix sparse memory access in the DMA callback

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
2 years agoMerge remote-tracking branch 'remotes/hreitz/tags/pull-block-2021-09-01' into staging
Peter Maydell [Thu, 2 Sep 2021 12:00:52 +0000 (13:00 +0100)]
Merge remote-tracking branch 'remotes/hreitz/tags/pull-block-2021-09-01' into staging

Block patches:
- Make the backup-top filter driver available for user-created block
  nodes (i.e. via blockdev-add)
- Allow running iotests with gdb or valgrind being attached to qemu
  instances
- Fix the raw format driver's permissions: There is no metadata, so we
  only need WRITE or RESIZE when the parent needs it
- Basic reopen implementation for win32 files (file-win32.c) so that
  qemu-img commit can work
- uclibc/musl build fix for the FUSE export code
- Some iotests delinting
- block-hmp-cmds.c refactoring

# gpg: Signature made Wed 01 Sep 2021 16:01:54 BST
# gpg:                using RSA key CB62D7A0EE3829E45F004D34A1FA40D098019CDF
# gpg:                issuer "hreitz@redhat.com"
# gpg: Good signature from "Hanna Reitz <hreitz@redhat.com>" [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: CB62 D7A0 EE38 29E4 5F00  4D34 A1FA 40D0 9801 9CDF

* remotes/hreitz/tags/pull-block-2021-09-01: (56 commits)
  block/file-win32: add reopen handlers
  block/export/fuse.c: fix fuse-lseek on uclibc or musl
  block/block-copy: block_copy_state_new(): drop extra arguments
  iotests/image-fleecing: add test-case for copy-before-write filter
  iotests/image-fleecing: prepare for adding new test-case
  iotests/image-fleecing: rename tgt_node
  iotests/image-fleecing: proper source device
  iotests.py: hmp_qemu_io: support qdev
  iotests: move 222 to tests/image-fleecing
  iotests/222: constantly use single quotes for strings
  iotests/222: fix pylint and mypy complains
  python:QEMUMachine: template typing for self returning methods
  python/qemu/machine: QEMUMachine: improve qmp() method
  python/qemu/machine.py: refactor _qemu_args()
  qapi: publish copy-before-write filter
  block/copy-before-write: make public block driver
  block/block-copy: make setting progress optional
  block/copy-before-write: initialize block-copy bitmap
  block/copy-before-write: cbw_init(): use options
  block/copy-before-write: bdrv_cbw_append(): drop unused compress arg
  ...

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
2 years ago9pfs: fix crash in v9fs_walk()
Christian Schoenebeck [Wed, 1 Sep 2021 16:15:10 +0000 (18:15 +0200)]
9pfs: fix crash in v9fs_walk()

v9fs_walk() utilizes the v9fs_co_run_in_worker({...}) macro to run the
supplied fs driver code block on a background worker thread.

When either the 'Twalk' client request was interrupted or if the client
requested fid for that 'Twalk' request caused a stat error then that
fs driver code block was left by 'break' keyword, with the intention to
return from worker thread back to main thread as well:

    v9fs_co_run_in_worker({
        if (v9fs_request_cancelled(pdu)) {
            err = -EINTR;
            break;
        }
        err = s->ops->lstat(&s->ctx, &dpath, &fidst);
        if (err < 0) {
            err = -errno;
            break;
        }
        ...
    });

However that 'break;' statement also skipped the v9fs_co_run_in_worker()
macro's final and mandatory

    /* re-enter back to qemu thread */
    qemu_coroutine_yield();

call and thus caused the rest of v9fs_walk() to be continued being
executed on the worker thread instead of main thread, eventually
leading to a crash in the transport virtio transport driver.

To fix this issue and to prevent the same error from happening again by
other users of v9fs_co_run_in_worker() in future, auto wrap the supplied
code block into its own

    do { } while (0);

loop inside the 'v9fs_co_run_in_worker' macro definition.

Full discussion and backtrace:
https://lists.gnu.org/archive/html/qemu-devel/2021-08/msg05209.html
https://lists.gnu.org/archive/html/qemu-devel/2021-09/msg00174.html

Fixes: 8d6cb100731c4d28535adbf2a3c2d1f29be3fef4
Signed-off-by: Christian Schoenebeck <qemu_oss@crudebyte.com>
Cc: qemu-stable@nongnu.org
Reviewed-by: Greg Kurz <groug@kaod.org>
Message-Id: <E1mLTBg-0002Bh-2D@lizzy.crudebyte.com>

2 years agohw/9pfs: use g_autofree in v9fs_walk() where possible
Christian Schoenebeck [Tue, 17 Aug 2021 13:46:50 +0000 (15:46 +0200)]
hw/9pfs: use g_autofree in v9fs_walk() where possible

Suggested-by: Greg Kurz <groug@kaod.org>
Signed-off-by: Christian Schoenebeck <qemu_oss@crudebyte.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Greg Kurz <groug@kaod.org>
Message-Id: <b51670d2a39399535a035f6bc77c3cbeed85edae.1629208359.git.qemu_oss@crudebyte.com>

2 years agohw/9pfs: avoid 'path' copy in v9fs_walk()
Christian Schoenebeck [Tue, 17 Aug 2021 12:38:24 +0000 (14:38 +0200)]
hw/9pfs: avoid 'path' copy in v9fs_walk()

The v9fs_walk() function resolves all client submitted path nodes to the
local 'pathes' array. Using a separate string scalar variable 'path'
inside the background worker thread loop and copying that local 'path'
string scalar variable subsequently to the 'pathes' array (at the end of
each loop iteration) is not necessary.

Instead simply resolve each path directly to the 'pathes' array and
don't use the string scalar variable 'path' inside the fs worker thread
loop at all.

The only advantage of the 'path' scalar was that in case of an error
the respective 'pathes' element would not be filled. Right now this is
not an issue as the v9fs_walk() function returns as soon as any error
occurs.

Suggested-by: Greg Kurz <groug@kaod.org>
Signed-off-by: Christian Schoenebeck <qemu_oss@crudebyte.com>
Reviewed-by: Greg Kurz <groug@kaod.org>
Message-Id: <7dacbecf25b2c9b4a0ce12d689a8a535f09a31e3.1629208359.git.qemu_oss@crudebyte.com>

2 years agoMerge remote-tracking branch 'remotes/stefanberger/tags/pull-tpm-2021-09-01-1' into...
Peter Maydell [Thu, 2 Sep 2021 07:51:31 +0000 (08:51 +0100)]
Merge remote-tracking branch 'remotes/stefanberger/tags/pull-tpm-2021-09-01-1' into staging

Merge tpm 2021/09/01 v1

# gpg: Signature made Wed 01 Sep 2021 13:13:27 BST
# gpg:                using RSA key B818B9CADF9089C2D5CEC66B75AD65802A0B4211
# gpg: Good signature from "Stefan Berger <stefanb@linux.vnet.ibm.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: B818 B9CA DF90 89C2 D5CE  C66B 75AD 6580 2A0B 4211

* remotes/stefanberger/tags/pull-tpm-2021-09-01-1:
  tests: acpi: tpm1.2: Add expected TPM 1.2 ACPI blobs
  tests: acpi: Add test cases for TPM 1.2 with TCPA table
  tests: Use QMP to check whether a TPM device model is available
  tests: acpi: prepare for new TPM 1.2 related tables
  tests: tpm: Create TPM 1.2 response in TPM emulator
  tests: acpi: tpm2: Add the renamed ACPI files and drop old ones
  tests: Add suffix 'tpm2' or 'tpm12' to ACPI table files
  tests: acpi: Prepare for renaming of TPM2 related ACPI files
  tests: Add tpm_version field to TPMTestState and fill it
  tests: Rename TestState to TPMTestState

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
2 years agoMerge remote-tracking branch 'remotes/pmaydell/tags/pull-target-arm-20210901' into...
Peter Maydell [Wed, 1 Sep 2021 16:45:38 +0000 (17:45 +0100)]
Merge remote-tracking branch 'remotes/pmaydell/tags/pull-target-arm-20210901' into staging

 * Refactor M-profile systick to use Clocks instead of system_clock_scale global
 * clock: Provide builtin multiplier/divider
 * Add A64FX processor model
 * Enable MVE emulation in Cortex-M55
 * hw: Add compat machines for 6.2
 * hw/intc/arm_gicv3: Replace mis-used MEMTX_* constants by booleans
 * hw/arm/raspi: Remove deprecated raspi2/raspi3 aliases

# gpg: Signature made Wed 01 Sep 2021 11:35:57 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-20210901: (51 commits)
  arm: Remove system_clock_scale global
  hw/timer/stellaris-gptm: Use Clock input instead of system_clock_scale
  hw/arm/stellaris: Split stellaris-gptm into its own file
  hw/arm/stellaris: Fix code style issues in GPTM code
  hw/timer/armv7m_systick: Use clock inputs instead of system_clock_scale
  hw/arm/msf2-soc: Wire up refclk
  hw/arm/msf2: Use Clock input to MSF2_SOC instead of m3clk property
  hw/arm/msf2_soc: Don't allocate separate MemoryRegions
  hw/arm/stellaris: Wire sysclk up to armv7m
  hw/arm/stellaris: split stellaris_sys_init()
  hw/arm/nrf51: Wire up sysclk
  hw/arm/stm32vldiscovery: Delete trailing blank line
  hw/arm/stm32f405: Wire up sysclk and refclk
  hw/arm/stm32f205: Wire up sysclk and refclk
  hw/arm/stm32f100: Wire up sysclk and refclk
  hw/arm: Don't allocate separate MemoryRegions in stm32 SoC realize
  clock: Provide builtin multiplier/divider
  hw/arm/mps2.c: Connect up armv7m clocks
  armsse: Wire up systick cpuclk clock
  hw/arm/armv7m: Create input clocks
  ...

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
2 years agoMerge remote-tracking branch 'remotes/kraxel/tags/usb-20210901-pull-request' into...
Peter Maydell [Wed, 1 Sep 2021 14:13:07 +0000 (15:13 +0100)]
Merge remote-tracking branch 'remotes/kraxel/tags/usb-20210901-pull-request' into staging

usb: bugfixes.

# gpg: Signature made Wed 01 Sep 2021 07:53:33 BST
# gpg:                using RSA key A0328CFFB93A17A79901FE7D4CB6D8EED3E87138
# gpg: Good signature from "Gerd Hoffmann (work) <kraxel@redhat.com>" [full]
# gpg:                 aka "Gerd Hoffmann <gerd@kraxel.org>" [full]
# gpg:                 aka "Gerd Hoffmann (private) <kraxel@gmail.com>" [full]
# Primary key fingerprint: A032 8CFF B93A 17A7 9901  FE7D 4CB6 D8EE D3E8 7138

* remotes/kraxel/tags/usb-20210901-pull-request:
  hw/usb: Fix typo in comments and print
  uas: add stream number sanity checks.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
2 years agoblock/file-win32: add reopen handlers
Viktor Prutyanov [Wed, 25 Aug 2021 17:36:25 +0000 (20:36 +0300)]
block/file-win32: add reopen handlers

Make 'qemu-img commit' work on Windows.

Command 'commit' requires reopening backing file in RW mode. So,
add reopen prepare/commit/abort handlers and change dwShareMode
for CreateFile call in order to allow further read/write reopening.

Resolves: https://gitlab.com/qemu-project/qemu/-/issues/418

Suggested-by: Hanna Reitz <hreitz@redhat.com>
Signed-off-by: Viktor Prutyanov <viktor.prutyanov@phystech.edu>
Tested-by: Helge Konetzka <hk@zapateado.de>
Message-Id: <20210825173625.19415-1-viktor.prutyanov@phystech.edu>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agoblock/export/fuse.c: fix fuse-lseek on uclibc or musl
Fabrice Fontaine [Fri, 27 Aug 2021 22:03:01 +0000 (00:03 +0200)]
block/export/fuse.c: fix fuse-lseek on uclibc or musl

Include linux/fs.h to avoid the following build failure on uclibc or
musl raised since version 6.0.0:

../block/export/fuse.c: In function 'fuse_lseek':
../block/export/fuse.c:641:19: error: 'SEEK_HOLE' undeclared (first use in this function)
  641 |     if (whence != SEEK_HOLE && whence != SEEK_DATA) {
      |                   ^~~~~~~~~
../block/export/fuse.c:641:19: note: each undeclared identifier is reported only once for each function it appears in
../block/export/fuse.c:641:42: error: 'SEEK_DATA' undeclared (first use in this function); did you mean 'SEEK_SET'?
  641 |     if (whence != SEEK_HOLE && whence != SEEK_DATA) {
      |                                          ^~~~~~~~~
      |                                          SEEK_SET

Fixes:
 - http://autobuild.buildroot.org/results/33c90ebf04997f4d3557cfa66abc9cf9a3076137

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Message-Id: <20210827220301.272887-1-fontaine.fabrice@gmail.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agoblock/block-copy: block_copy_state_new(): drop extra arguments
Vladimir Sementsov-Ogievskiy [Tue, 24 Aug 2021 08:38:56 +0000 (11:38 +0300)]
block/block-copy: block_copy_state_new(): drop extra arguments

The only caller pass copy_range and compress both false. Let's just
drop these arguments.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20210824083856.17408-35-vsementsov@virtuozzo.com>
Reviewed-by: Hanna Reitz <hreitz@redhat.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agoiotests/image-fleecing: add test-case for copy-before-write filter
Vladimir Sementsov-Ogievskiy [Tue, 24 Aug 2021 08:38:55 +0000 (11:38 +0300)]
iotests/image-fleecing: add test-case for copy-before-write filter

New fleecing method becomes available: copy-before-write filter.

Actually we don't need backup job to setup image fleecing. Add test
for new recommended way of image fleecing.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20210824083856.17408-34-vsementsov@virtuozzo.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agoiotests/image-fleecing: prepare for adding new test-case
Vladimir Sementsov-Ogievskiy [Tue, 24 Aug 2021 08:38:54 +0000 (11:38 +0300)]
iotests/image-fleecing: prepare for adding new test-case

We are going to add a test-case with some behavior modifications. So,
let's prepare a function to be reused.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20210824083856.17408-33-vsementsov@virtuozzo.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agoiotests/image-fleecing: rename tgt_node
Vladimir Sementsov-Ogievskiy [Tue, 24 Aug 2021 08:38:53 +0000 (11:38 +0300)]
iotests/image-fleecing: rename tgt_node

Actually target of backup(sync=None) is not a final backup target:
image fleecing is intended to be used with external tool, which will
copy data from fleecing node to some real backup target.

Also, we are going to add a test case for "push backup with fleecing",
where instead of exporting fleecing node by NBD, we'll start a backup
job from fleecing node to real backup target.

To avoid confusion, let's rename temporary fleecing node now.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20210824083856.17408-32-vsementsov@virtuozzo.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agoiotests/image-fleecing: proper source device
Vladimir Sementsov-Ogievskiy [Tue, 24 Aug 2021 08:38:52 +0000 (11:38 +0300)]
iotests/image-fleecing: proper source device

Define scsi device to operate with it by qom-set in further patch.

Give a new node-name to source block node, to not look like device
name.

Job now don't want to work without giving explicit id, so, let's call
it "fleecing".

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20210824083856.17408-31-vsementsov@virtuozzo.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agoiotests.py: hmp_qemu_io: support qdev
Vladimir Sementsov-Ogievskiy [Tue, 24 Aug 2021 08:38:51 +0000 (11:38 +0300)]
iotests.py: hmp_qemu_io: support qdev

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: John Snow <jsnow@redhat.com>
Message-Id: <20210824083856.17408-30-vsementsov@virtuozzo.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agoiotests: move 222 to tests/image-fleecing
Vladimir Sementsov-Ogievskiy [Tue, 24 Aug 2021 08:38:50 +0000 (11:38 +0300)]
iotests: move 222 to tests/image-fleecing

Give a good name to test file.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20210824083856.17408-29-vsementsov@virtuozzo.com>
[hreitz: Adjust .gitlab-ci.d/buildtest.yml]
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agoiotests/222: constantly use single quotes for strings
Vladimir Sementsov-Ogievskiy [Tue, 24 Aug 2021 08:38:49 +0000 (11:38 +0300)]
iotests/222: constantly use single quotes for strings

The file use both single and double quotes for strings. Let's be
consistent.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20210824083856.17408-28-vsementsov@virtuozzo.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agoiotests/222: fix pylint and mypy complains
Vladimir Sementsov-Ogievskiy [Tue, 24 Aug 2021 08:38:48 +0000 (11:38 +0300)]
iotests/222: fix pylint and mypy complains

Here:
 - long line
 - move to new interface of vm.qmp() (direct passing dict), to avoid
   mypy false-positive, as it thinks that unpacked dict is a positional
   argument.
 - extra parenthesis
 - handle event_wait possible None value

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20210824083856.17408-27-vsementsov@virtuozzo.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agopython:QEMUMachine: template typing for self returning methods
Vladimir Sementsov-Ogievskiy [Tue, 24 Aug 2021 08:38:47 +0000 (11:38 +0300)]
python:QEMUMachine: template typing for self returning methods

mypy thinks that return value of these methods in subclusses is
QEMUMachine, which is wrong. So, make typing smarter.

Suggested-by: John Snow <jsnow@redhat.com>
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20210824083856.17408-26-vsementsov@virtuozzo.com>
Reviewed-by: Hanna Reitz <hreitz@redhat.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agopython/qemu/machine: QEMUMachine: improve qmp() method
Vladimir Sementsov-Ogievskiy [Tue, 24 Aug 2021 08:38:46 +0000 (11:38 +0300)]
python/qemu/machine: QEMUMachine: improve qmp() method

We often call qmp() with unpacking dict, like qmp('foo', **{...}).
mypy don't really like it, it thinks that passed unpacked dict is a
positional argument and complains that it type should be bool (because
second argument of qmp() is conv_keys: bool).

Allow passing dict directly, simplifying interface, and giving a way to
satisfy mypy.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: John Snow <jsnow@redhat.com>
Message-Id: <20210824083856.17408-25-vsementsov@virtuozzo.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agopython/qemu/machine.py: refactor _qemu_args()
Vladimir Sementsov-Ogievskiy [Tue, 24 Aug 2021 08:38:45 +0000 (11:38 +0300)]
python/qemu/machine.py: refactor _qemu_args()

 - use shorter construction
 - don't create new dict if not needed
 - drop extra unpacking key-val arguments
 - drop extra default values

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: John Snow <jsnow@redhat.com>
Message-Id: <20210824083856.17408-24-vsementsov@virtuozzo.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agoqapi: publish copy-before-write filter
Vladimir Sementsov-Ogievskiy [Tue, 24 Aug 2021 08:38:44 +0000 (11:38 +0300)]
qapi: publish copy-before-write filter

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Acked-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20210824083856.17408-23-vsementsov@virtuozzo.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agoblock/copy-before-write: make public block driver
Vladimir Sementsov-Ogievskiy [Tue, 24 Aug 2021 08:38:43 +0000 (11:38 +0300)]
block/copy-before-write: make public block driver

Finally, copy-before-write gets own .bdrv_open and .bdrv_close
handlers, block_init() call and becomes available through bdrv_open().

To achieve this:

 - cbw_init gets unused flags argument and becomes cbw_open
 - block_copy_state_free() call moved to new cbw_close()
 - in bdrv_cbw_append:
   - options are completed with driver and node-name, and we can simply
     use bdrv_insert_node() to do both open and drained replacing
 - in bdrv_cbw_drop:
   - cbw_close() is now responsible for freeing s->bcs, so don't do it
     here

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20210824083856.17408-22-vsementsov@virtuozzo.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agoblock/block-copy: make setting progress optional
Vladimir Sementsov-Ogievskiy [Tue, 24 Aug 2021 08:38:42 +0000 (11:38 +0300)]
block/block-copy: make setting progress optional

Now block-copy will crash if user don't set progress meter by
block_copy_set_progress_meter(). copy-before-write filter will be used
in separate of backup job, and it doesn't want any progress meter (for
now). So, allow not setting it.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20210824083856.17408-21-vsementsov@virtuozzo.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agoblock/copy-before-write: initialize block-copy bitmap
Vladimir Sementsov-Ogievskiy [Tue, 24 Aug 2021 08:38:41 +0000 (11:38 +0300)]
block/copy-before-write: initialize block-copy bitmap

We are going to publish copy-before-write filter to be used in separate
of backup. Future step would support bitmap for the filter. But let's
start from full set bitmap.

We have to modify backup, as bitmap is first initialized by
copy-before-write filter, and then backup modifies it.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20210824083856.17408-20-vsementsov@virtuozzo.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agoblock/copy-before-write: cbw_init(): use options
Vladimir Sementsov-Ogievskiy [Tue, 24 Aug 2021 08:38:40 +0000 (11:38 +0300)]
block/copy-before-write: cbw_init(): use options

One more step closer to .bdrv_open(): use options instead of plain
arguments. Move to bdrv_open_child() calls, native for drive open
handlers.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Hanna Reitz <hreitz@redhat.com>
Message-Id: <20210824083856.17408-19-vsementsov@virtuozzo.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agoblock/copy-before-write: bdrv_cbw_append(): drop unused compress arg
Vladimir Sementsov-Ogievskiy [Tue, 24 Aug 2021 08:38:39 +0000 (11:38 +0300)]
block/copy-before-write: bdrv_cbw_append(): drop unused compress arg

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Hanna Reitz <hreitz@redhat.com>
Message-Id: <20210824083856.17408-18-vsementsov@virtuozzo.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agoblock/copy-before-write: cbw_init(): use file child after attaching
Vladimir Sementsov-Ogievskiy [Tue, 24 Aug 2021 08:38:38 +0000 (11:38 +0300)]
block/copy-before-write: cbw_init(): use file child after attaching

In the next commit we'll get rid of source argument of cbw_init().
Prepare to it now, to make next commit simpler: move the code block
that uses source below attaching the child and use bs->file->bs instead
of source variable.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20210824083856.17408-17-vsementsov@virtuozzo.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agoblock/copy-before-write: cbw_init(): rename variables
Vladimir Sementsov-Ogievskiy [Tue, 24 Aug 2021 08:38:37 +0000 (11:38 +0300)]
block/copy-before-write: cbw_init(): rename variables

One more step closer to real .bdrv_open() handler: use more usual names
for bs being initialized and its state.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20210824083856.17408-16-vsementsov@virtuozzo.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agoblock/copy-before-write: introduce cbw_init()
Vladimir Sementsov-Ogievskiy [Tue, 24 Aug 2021 08:38:36 +0000 (11:38 +0300)]
block/copy-before-write: introduce cbw_init()

Move part of bdrv_cbw_append() to new function cbw_open(). It's an
intermediate step for adding normal .bdrv_open() handler to the
filter. With this commit no logic is changed, but we have a function
which will be turned into .bdrv_open() handler in future commit.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20210824083856.17408-15-vsementsov@virtuozzo.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agoblock/copy-before-write: bdrv_cbw_append(): replace child at last
Vladimir Sementsov-Ogievskiy [Tue, 24 Aug 2021 08:38:35 +0000 (11:38 +0300)]
block/copy-before-write: bdrv_cbw_append(): replace child at last

Refactor the function to replace child at last. Thus we don't need to
revert it and code is simplified.

block-copy state initialization being done before replacing the child
doesn't need any drained section.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20210824083856.17408-14-vsementsov@virtuozzo.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agoblock/copy-before-write: use file child instead of backing
Vladimir Sementsov-Ogievskiy [Tue, 24 Aug 2021 08:38:34 +0000 (11:38 +0300)]
block/copy-before-write: use file child instead of backing

We are going to publish copy-before-write filter, and there no public
backing-child-based filter in Qemu. No reason to create a precedent, so
let's refactor copy-before-write filter instead.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20210824083856.17408-13-vsementsov@virtuozzo.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agoblock/copy-before-write: drop extra bdrv_unref on failure path
Vladimir Sementsov-Ogievskiy [Tue, 24 Aug 2021 08:38:33 +0000 (11:38 +0300)]
block/copy-before-write: drop extra bdrv_unref on failure path

bdrv_attach_child() do bdrv_unref() on failure, so we shouldn't do it
by hand here.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20210824083856.17408-12-vsementsov@virtuozzo.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agoblock/copy-before-write: relax permission requirements when no parents
Vladimir Sementsov-Ogievskiy [Tue, 24 Aug 2021 08:38:32 +0000 (11:38 +0300)]
block/copy-before-write: relax permission requirements when no parents

We are going to publish copy-before-write filter. So, user should be
able to create it with blockdev-add first, specifying both filtered and
target children. And then do blockdev-reopen, to actually insert the
filter where needed.

Currently, filter unshares write permission unconditionally on source
node. It's good, but it will not allow to do blockdev-add. So, let's
relax restrictions when filter doesn't have any parent.

Test output is modified, as now permission conflict happens only when
job creates a blk parent for filter node.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20210824083856.17408-11-vsementsov@virtuozzo.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agoblock/backup: move cluster size calculation to block-copy
Vladimir Sementsov-Ogievskiy [Tue, 24 Aug 2021 08:38:31 +0000 (11:38 +0300)]
block/backup: move cluster size calculation to block-copy

The main consumer of cluster-size is block-copy. Let's calculate it
here instead of passing through backup-top.

We are going to publish copy-before-write filter soon, so it will be
created through options. But we don't want for now to make explicit
option for cluster-size, let's continue to calculate it automatically.
So, now is the time to get rid of cluster_size argument for
bdrv_cbw_append().

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20210824083856.17408-10-vsementsov@virtuozzo.com>
[hreitz: Add qemu/error-report.h include to block/block-copy.c]
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agoMerge remote-tracking branch 'remotes/kraxel/tags/audio-20210901-pull-request' into...
Peter Maydell [Wed, 1 Sep 2021 11:38:14 +0000 (12:38 +0100)]
Merge remote-tracking branch 'remotes/kraxel/tags/audio-20210901-pull-request' into staging

audio: split audio entry in MAINTAINERS file.

# gpg: Signature made Wed 01 Sep 2021 05:42:36 BST
# gpg:                using RSA key A0328CFFB93A17A79901FE7D4CB6D8EED3E87138
# gpg: Good signature from "Gerd Hoffmann (work) <kraxel@redhat.com>" [full]
# gpg:                 aka "Gerd Hoffmann <gerd@kraxel.org>" [full]
# gpg:                 aka "Gerd Hoffmann (private) <kraxel@gmail.com>" [full]
# Primary key fingerprint: A032 8CFF B93A 17A7 9901  FE7D 4CB6 D8EE D3E8 7138

* remotes/kraxel/tags/audio-20210901-pull-request:
  MAINTAINERS: Split Audio backends sections
  MAINTAINERS: Remove SPICE from Audio backends section
  MAINTAINERS: Split Audio backends VS frontends

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
2 years agoMAINTAINERS: add fuzzing reviewer
Qiuhao Li [Tue, 24 Aug 2021 06:26:39 +0000 (14:26 +0800)]
MAINTAINERS: add fuzzing reviewer

To keep me cc-ed when something changes. Suggested by Alexander.

https://lists.gnu.org/archive/html/qemu-devel/2021-08/msg03631.html

Signed-off-by: Qiuhao Li <Qiuhao.Li@outlook.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Alexander Bulekov <alxndr@bu.edu>
Reviewed-by: Darren Kenny <darren.kenny@oracle.com>
2 years agoMAINTAINERS: Add myself as a reviewer for Device Fuzzing
Darren Kenny [Tue, 24 Aug 2021 14:04:01 +0000 (14:04 +0000)]
MAINTAINERS: Add myself as a reviewer for Device Fuzzing

Should have done this much sooner given the amount of reviewing I'm
already doing in this area.

Signed-off-by: Darren Kenny <darren.kenny@oracle.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Alexander Bulekov <alxndr@bu.edu>
2 years agofuzz: unblock SIGALRM so the timeout works
Alexander Bulekov [Wed, 4 Aug 2021 13:56:21 +0000 (09:56 -0400)]
fuzz: unblock SIGALRM so the timeout works

The timeout mechanism won't work if SIGALRM is blocked. This changes
unmasks SIGALRM when the timer is installed. This doesn't completely
solve the problem, as the fuzzer could trigger some device activity that
re-masks SIGALRM. However, there are currently no inputs on OSS-Fuzz
that re-mask SIGALRM and timeout. If that turns out to be a real issue,
we could try to hook sigmask-type calls, or use a separate timer thread.

Based-on: <20210713150037.9297-1-alxndr@bu.edu>
Signed-off-by: Alexander Bulekov <alxndr@bu.edu>
Reviewed-by: Darren Kenny <darren.kenny@oracle.com>
2 years agofuzz: use ITIMER_REAL for timeouts
Alexander Bulekov [Wed, 4 Aug 2021 13:56:20 +0000 (09:56 -0400)]
fuzz: use ITIMER_REAL for timeouts

Using ITIMER_VIRTUAL is a bad idea, if the fuzzer hits a blocking
syscall - e.g. ppoll with a NULL timespec. This causes timeout issues
while fuzzing some block-device code. Fix that by using wall-clock time.
This might cause inputs to timeout sometimes due to scheduling
effects/ambient load, but it is better than bringing the entire fuzzing
process to a halt.

Based-on: <20210713150037.9297-1-alxndr@bu.edu>
Signed-off-by: Alexander Bulekov <alxndr@bu.edu>
Reviewed-by: Darren Kenny <darren.kenny@oracle.com>
2 years agofuzz: add an instrumentation filter
Alexander Bulekov [Tue, 13 Jul 2021 15:00:37 +0000 (11:00 -0400)]
fuzz: add an instrumentation filter

By default, -fsanitize=fuzzer instruments all code with coverage
information. However, this means that libfuzzer will track coverage over
hundreds of source files that are unrelated to virtual-devices. This
means that libfuzzer will optimize inputs for coverage observed in timer
code, memory APIs etc. This slows down the fuzzer and stores many inputs
that are not relevant to the actual virtual-devices.

With this change, clang versions that support the
"-fsanitize-coverage-allowlist" will only instrument a subset of the
compiled code, that is directly related to virtual-devices.

Signed-off-by: Alexander Bulekov <alxndr@bu.edu>
Reviewed-by: Darren Kenny <darren.kenny@oracle.com>
2 years agofuzz: make object-name matching case-insensitive
Alexander Bulekov [Tue, 13 Jul 2021 15:00:36 +0000 (11:00 -0400)]
fuzz: make object-name matching case-insensitive

We have some configs for devices such as the AC97 and ES1370 that were
not matching memory-regions correctly, because the configs provided
lowercase names. To resolve these problems and prevent them from
occurring again in the future, convert both the pattern and names to
lower-case, prior to checking for a match.

Suggested-by: Darren Kenny <darren.kenny@oracle.com>
Reviewed-by: Darren Kenny <darren.kenny@oracle.com>
Signed-off-by: Alexander Bulekov <alxndr@bu.edu>
2 years agofuzz: adjust timeout to allow for longer inputs
Alexander Bulekov [Tue, 13 Jul 2021 15:00:35 +0000 (11:00 -0400)]
fuzz: adjust timeout to allow for longer inputs

Using a custom timeout is useful to continue fuzzing complex devices,
even after we run into some slow code-path. However, simply adding a
fixed timeout to each input effectively caps the maximum input
length/number of operations at some artificial value. There are two
major problems with this:
1. Some code might only be reachable through long IO sequences.
2. Longer inputs can actually be _better_ for performance. While the
   raw number of fuzzer executions decreases with larger inputs, the
   number of MMIO/PIO/DMA operation/second actually increases, since
   were are speding proportionately less time fork()ing.

With this change, we keep the custom-timeout, but we renew it, prior to
each MMIO/PIO/DMA operation. Thus, we time-out only when a specific
operation takes a long time.

Reviewed-by: Darren Kenny <darren.kenny@oracle.com>
Signed-off-by: Alexander Bulekov <alxndr@bu.edu>
2 years agofuzz: fix sparse memory access in the DMA callback
Alexander Bulekov [Tue, 13 Jul 2021 15:00:34 +0000 (11:00 -0400)]
fuzz: fix sparse memory access in the DMA callback

The code mistakenly relied on address_space_translate to store the
length remaining until the next memory-region. We care about this
because when there is RAM or sparse-memory neighboring on an MMIO
region, we should only write up to the border, to prevent inadvertently
invoking MMIO handlers within the DMA callback.

However address_space_translate_internal only stores the length until
the end of the MemoryRegion if memory_region_is_ram(mr). Otherwise
the *len is left unmodified. This caused some false-positive issues,
where the fuzzer found a way to perform a nested MMIO write through a
DMA callback on an [address, length] that started within sparse memory
and spanned some device MMIO regions.

To fix this, write to sparse memory in small chunks of
memory_access_size (similar to the underlying address_space_write code),
which will prevent accidentally hitting MMIO handlers through large
writes.

Signed-off-by: Alexander Bulekov <alxndr@bu.edu>
Reviewed-by: Darren Kenny <darren.kenny@oracle.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
2 years agoblock/backup: set copy_range and compress after filter insertion
Vladimir Sementsov-Ogievskiy [Tue, 24 Aug 2021 08:38:30 +0000 (11:38 +0300)]
block/backup: set copy_range and compress after filter insertion

We are going to publish copy-before-write filter, so it would be
initialized through options. Still we don't want to publish compress
and copy-range options, as

1. Modern way to enable compression is to use compress filter.

2. For copy-range it's unclean how to make proper interface:
 - it's has experimental prefix for backup job anyway
 - the whole BackupPerf structure doesn't make sense for the filter
 So, let's just add copy-range possibility to the filter later if
 needed.

Still, we are going to continue support for compression and
experimental copy-range in backup job. So, set these options after
filter creation.

Note, that we can drop "compress" argument of bdrv_cbw_append() now, as
well as "perf". The only reason not doing so is that now, when I
prepare this patch the big series around it is already reviewed and I
want to avoid extra rebase conflicts to simplify review of the
following version.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Hanna Reitz <hreitz@redhat.com>
Message-Id: <20210824083856.17408-9-vsementsov@virtuozzo.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agoblock/block-copy: introduce block_copy_set_copy_opts()
Vladimir Sementsov-Ogievskiy [Tue, 24 Aug 2021 08:38:29 +0000 (11:38 +0300)]
block/block-copy: introduce block_copy_set_copy_opts()

We'll need a possibility to set compress and use_copy_range options
after initialization of the state. So make corresponding part of
block_copy_state_new() separate and public.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20210824083856.17408-8-vsementsov@virtuozzo.com>
Reviewed-by: Hanna Reitz <hreitz@redhat.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agoblock-copy: move detecting fleecing scheme to block-copy
Vladimir Sementsov-Ogievskiy [Tue, 24 Aug 2021 08:38:28 +0000 (11:38 +0300)]
block-copy: move detecting fleecing scheme to block-copy

We want to simplify initialization interface of copy-before-write
filter as we are going to make it public. So, let's detect fleecing
scheme exactly in block-copy code, to not pass this information through
extra levels.

Why not just set BDRV_REQ_SERIALISING unconditionally: because we are
going to implement new more efficient fleecing scheme which will not
rely on backing feature.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Hanna Reitz <hreitz@redhat.com>
Message-Id: <20210824083856.17408-7-vsementsov@virtuozzo.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agoblock: rename backup-top to copy-before-write
Vladimir Sementsov-Ogievskiy [Tue, 24 Aug 2021 08:38:27 +0000 (11:38 +0300)]
block: rename backup-top to copy-before-write

We are going to convert backup_top to full featured public filter,
which can be used in separate of backup job. Start from renaming from
"how it used" to "what it does".

While updating comments in 283 iotest, drop and rephrase also things
about ".active", as this field is now dropped, and filter doesn't have
"inactive" mode.

Note that this change may be considered as incompatible interface
change, as backup-top filter format name was visible through
query-block and query-named-block-nodes.

Still, consider the following reasoning:

1. backup-top was never documented, so if someone depends on format
   name (for driver that can't be used other than it is automatically
   inserted on backup job start), it's a kind of "undocumented feature
   use". So I think we are free to change it.

2. There is a hope, that there is no such users: it's a lot more native
   to give a good node-name to backup-top filter if need to operate
   with it somehow, and don't touch format name.

3. Another "incompatible" change in further commit would be moving
   copy-before-write filter from using backing child to file child. And
   this is even more reasonable than renaming: for now all public
   filters are file-child based.

So, it's a risky change, but risk seems small and good interface worth
it.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20210824083856.17408-6-vsementsov@virtuozzo.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agoqdev: allow setting drive property for realized device
Vladimir Sementsov-Ogievskiy [Tue, 24 Aug 2021 08:38:26 +0000 (11:38 +0300)]
qdev: allow setting drive property for realized device

We need an ability to insert filters above top block node, attached to
block device. It can't be achieved with blockdev-reopen command. So, we
want do it with help of qom-set.

Intended usage:

Assume there is a node A that is attached to some guest device.

1. blockdev-add to create a filter node B that has A as its child.

2. qom-set to change the node attached to the guest device’s
   BlockBackend from A to B.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20210824083856.17408-5-vsementsov@virtuozzo.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agoqdev-properties: PropertyInfo: add realized_set_allowed field
Vladimir Sementsov-Ogievskiy [Tue, 24 Aug 2021 08:38:25 +0000 (11:38 +0300)]
qdev-properties: PropertyInfo: add realized_set_allowed field

Add field, so property can declare support for setting the property
when device is realized. To be used in the following commit.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20210824083856.17408-4-vsementsov@virtuozzo.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agoblock: introduce blk_replace_bs
Vladimir Sementsov-Ogievskiy [Tue, 24 Aug 2021 08:38:24 +0000 (11:38 +0300)]
block: introduce blk_replace_bs

Add function to change bs inside blk.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20210824083856.17408-3-vsementsov@virtuozzo.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agoblock: introduce bdrv_replace_child_bs()
Vladimir Sementsov-Ogievskiy [Tue, 24 Aug 2021 08:38:23 +0000 (11:38 +0300)]
block: introduce bdrv_replace_child_bs()

Add function to transactionally replace bs inside BdrvChild.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20210824083856.17408-2-vsementsov@virtuozzo.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agoiotests: use subprocess.DEVNULL instead of open("/dev/null")
John Snow [Tue, 20 Jul 2021 17:33:21 +0000 (13:33 -0400)]
iotests: use subprocess.DEVNULL instead of open("/dev/null")

Avoids a warning from pylint not to use open() outside of a
with-statement, and is ... probably more portable anyway. Not that I
think we care too much about running tests *on* Windows, but... eh.

Signed-off-by: John Snow <jsnow@redhat.com>
Message-Id: <20210720173336.1876937-3-jsnow@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agoiotests: use with-statement for open() calls
John Snow [Tue, 20 Jul 2021 17:33:20 +0000 (13:33 -0400)]
iotests: use with-statement for open() calls

Silences a new pylint warning. The dangers of *not* doing this are
somewhat unclear; I believe the file object gets garbage collected
eventually, but possibly the way in which it happens is
non-deterministic. Maybe this is a valid warning, but if there are
consequences of not doing it, I am not aware of them at present.

Signed-off-by: John Snow <jsnow@redhat.com>
Message-Id: <20210720173336.1876937-2-jsnow@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agoraw-format: drop WRITE and RESIZE child perms when possible
Stefan Hajnoczi [Mon, 26 Jul 2021 12:28:39 +0000 (13:28 +0100)]
raw-format: drop WRITE and RESIZE child perms when possible

The following command-line fails due to a permissions conflict:

  $ qemu-storage-daemon \
      --blockdev driver=nvme,node-name=nvme0,device=0000:08:00.0,namespace=1 \
      --blockdev driver=raw,node-name=l1-1,file=nvme0,offset=0,size=1073741824 \
      --blockdev driver=raw,node-name=l1-2,file=nvme0,offset=1073741824,size=1073741824 \
      --nbd-server addr.type=unix,addr.path=/tmp/nbd.sock,max-connections=2 \
      --export type=nbd,id=nbd-l1-1,node-name=l1-1,name=l1-1,writable=on \
      --export type=nbd,id=nbd-l1-2,node-name=l1-2,name=l1-2,writable=on

  qemu-storage-daemon: --export type=nbd,id=nbd-l1-1,node-name=l1-1,name=l1-1,writable=on: Permission conflict on node 'nvme0': permissions 'resize' are both required by node 'l1-1' (uses node 'nvme0' as 'file' child) and unshared by node 'l1-2' (uses node 'nvme0' as 'file' child).

The problem is that block/raw-format.c relies on bdrv_default_perms() to
set permissions on the nvme node. The default permissions add RESIZE in
anticipation of a format driver like qcow2 that needs to grow the image
file. This fails because RESIZE is unshared, so we cannot get the RESIZE
permission.

Max Reitz pointed out that block/crypto.c already handles this case by
implementing a custom ->bdrv_child_perm() function that adjusts the
result of bdrv_default_perms().

This patch takes the same approach in block/raw-format.c so that RESIZE
is only required if it's actually necessary (e.g. the parent is qcow2).

Cc: Max Reitz <mreitz@redhat.com>
Cc: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20210726122839.822900-1-stefanha@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agoblock/monitor: Consolidate hmp_handle_error calls to reduce redundant code
Mao Zhongyi [Mon, 2 Aug 2021 06:25:07 +0000 (14:25 +0800)]
block/monitor: Consolidate hmp_handle_error calls to reduce redundant code

Signed-off-by: Mao Zhongyi <maozhongyi@cmss.chinamobile.com>
Message-Id: <20210802062507.347555-1-maozhongyi@cmss.chinamobile.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agodocs/devel/testing: add -p option to the debug section of QEMU iotests
Emanuele Giuseppe Esposito [Mon, 9 Aug 2021 09:01:14 +0000 (11:01 +0200)]
docs/devel/testing: add -p option to the debug section of QEMU iotests

Signed-off-by: Emanuele Giuseppe Esposito <eesposit@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20210809090114.64834-17-eesposit@redhat.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agoqemu-iotests: add option to show qemu binary logs on stdout
Emanuele Giuseppe Esposito [Mon, 9 Aug 2021 09:01:13 +0000 (11:01 +0200)]
qemu-iotests: add option to show qemu binary logs on stdout

Using the flag -p, allow the qemu binary to print to stdout.

Also create the common function _close_qemu_log_file() to
avoid accessing machine.py private fields directly and have
duplicate code.

Signed-off-by: Emanuele Giuseppe Esposito <eesposit@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20210809090114.64834-16-eesposit@redhat.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agodocs/devel/testing: add -valgrind option to the debug section of QEMU iotests
Emanuele Giuseppe Esposito [Mon, 9 Aug 2021 09:01:12 +0000 (11:01 +0200)]
docs/devel/testing: add -valgrind option to the debug section of QEMU iotests

Signed-off-by: Emanuele Giuseppe Esposito <eesposit@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20210809090114.64834-15-eesposit@redhat.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agoqemu-iotests: insert valgrind command line as wrapper for qemu binary
Emanuele Giuseppe Esposito [Mon, 9 Aug 2021 09:01:11 +0000 (11:01 +0200)]
qemu-iotests: insert valgrind command line as wrapper for qemu binary

If -gdb and -valgrind are both defined, return an error.

Signed-off-by: Emanuele Giuseppe Esposito <eesposit@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20210809090114.64834-14-eesposit@redhat.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agoqemu-iotests: allow valgrind to read/delete the generated log file
Emanuele Giuseppe Esposito [Mon, 9 Aug 2021 09:01:10 +0000 (11:01 +0200)]
qemu-iotests: allow valgrind to read/delete the generated log file

When using -valgrind on the script tests, it generates a log file
in $TEST_DIR that is either read (if valgrind finds problems) or
otherwise deleted. Provide the same exact behavior when using
-valgrind on the python tests.

Signed-off-by: Emanuele Giuseppe Esposito <eesposit@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20210809090114.64834-13-eesposit@redhat.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agoqemu-iotests: extend QMP socket timeout when using valgrind
Emanuele Giuseppe Esposito [Mon, 9 Aug 2021 09:01:09 +0000 (11:01 +0200)]
qemu-iotests: extend QMP socket timeout when using valgrind

As with gdbserver, valgrind delays the test execution, so
the default QMP socket timeout and the generic class
Timeout in iotests.py timeouts too soon.

Signed-off-by: Emanuele Giuseppe Esposito <eesposit@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20210809090114.64834-12-eesposit@redhat.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agoqemu-iotests: extend the check script to prepare supporting valgrind for python tests
Emanuele Giuseppe Esposito [Mon, 9 Aug 2021 09:01:08 +0000 (11:01 +0200)]
qemu-iotests: extend the check script to prepare supporting valgrind for python tests

Currently, the check script only parses the option and sets the
VALGRIND_QEMU environmental variable to "y".
Add another local python variable that prepares the command line,
identical to the one provided in the test scripts.

Because the python script does not know in advance the valgrind
PID to assign to the log file name, use the "%p" flag in valgrind
log file name that automatically puts the process PID at runtime.

Signed-off-by: Emanuele Giuseppe Esposito <eesposit@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20210809090114.64834-11-eesposit@redhat.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agodocs/devel/testing: add -gdb option to the debugging section of QEMU iotests
Emanuele Giuseppe Esposito [Mon, 9 Aug 2021 09:01:07 +0000 (11:01 +0200)]
docs/devel/testing: add -gdb option to the debugging section of QEMU iotests

Signed-off-by: Emanuele Giuseppe Esposito <eesposit@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20210809090114.64834-10-eesposit@redhat.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agoqemu-iotests: add gdbserver option to script tests too
Emanuele Giuseppe Esposito [Mon, 9 Aug 2021 09:01:06 +0000 (11:01 +0200)]
qemu-iotests: add gdbserver option to script tests too

Remove read timer in test script when GDB_OPTIONS are set,
so that the bash tests won't timeout while running gdb.

The only limitation here is that running a script with gdbserver
will make the test output mismatch with the expected
results, making the test fail.

Signed-off-by: Emanuele Giuseppe Esposito <eesposit@redhat.com>
Message-Id: <20210809090114.64834-9-eesposit@redhat.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agoqemu_iotests: insert gdbserver command line as wrapper for qemu binary
Emanuele Giuseppe Esposito [Mon, 9 Aug 2021 09:01:05 +0000 (11:01 +0200)]
qemu_iotests: insert gdbserver command line as wrapper for qemu binary

Signed-off-by: Emanuele Giuseppe Esposito <eesposit@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20210809090114.64834-8-eesposit@redhat.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agoqemu-iotests: delay QMP socket timers
Emanuele Giuseppe Esposito [Mon, 9 Aug 2021 09:01:04 +0000 (11:01 +0200)]
qemu-iotests: delay QMP socket timers

Attaching gdbserver implies that the qmp socket
should wait indefinitely for an answer from QEMU.

Signed-off-by: Emanuele Giuseppe Esposito <eesposit@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20210809090114.64834-7-eesposit@redhat.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agoqemu-iotests: add option to attach gdbserver
Emanuele Giuseppe Esposito [Mon, 9 Aug 2021 09:01:03 +0000 (11:01 +0200)]
qemu-iotests: add option to attach gdbserver

Define -gdb flag and GDB_OPTIONS environment variable
to python tests to attach a gdbserver to each qemu instance.
This patch only adds and parses this flag, it does not yet add
the implementation for it.

if -gdb is not provided but $GDB_OPTIONS is set, ignore the
environment variable.

Signed-off-by: Emanuele Giuseppe Esposito <eesposit@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20210809090114.64834-6-eesposit@redhat.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agodocs/devel/testing: add debug section to the QEMU iotests chapter
Emanuele Giuseppe Esposito [Mon, 9 Aug 2021 09:01:02 +0000 (11:01 +0200)]
docs/devel/testing: add debug section to the QEMU iotests chapter

Introduce the "Debugging a test case" section, in preparation
to the additional flags that will be added in the next patches.

Signed-off-by: Emanuele Giuseppe Esposito <eesposit@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20210809090114.64834-5-eesposit@redhat.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agopython: qemu: pass the wrapper field from QEMUQtestmachine to QEMUMachine
Emanuele Giuseppe Esposito [Mon, 9 Aug 2021 09:01:01 +0000 (11:01 +0200)]
python: qemu: pass the wrapper field from QEMUQtestmachine to QEMUMachine

Signed-off-by: Emanuele Giuseppe Esposito <eesposit@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: John Snow <jsnow@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Acked-by: John Snow <jsnow@redhat.com>
Message-Id: <20210809090114.64834-4-eesposit@redhat.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agopython: Reduce strictness of pylint's duplicate-code check
John Snow [Mon, 9 Aug 2021 09:01:00 +0000 (11:01 +0200)]
python: Reduce strictness of pylint's duplicate-code check

Pylint prior to 2.8.3 (We pin at >= 2.8.0) includes function and method
signatures as part of its duplicate checking algorithm. This check does
not listen to pragmas, so the only way to disable it is to turn it off
completely or increase the minimum duplicate lines so that it doesn't
trigger for functions with long, multi-line signatures.

When we decide to upgrade to pylint 2.8.3 or greater, we will be able to
use 'ignore-signatures = true' to the config instead.

I'd prefer not to keep us on the very bleeding edge of pylint if I can
help it -- 2.8.3 came out only three days ago at time of writing.

See: https://github.com/PyCQA/pylint/pull/4474
Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Acked-by: John Snow <jsnow@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20210809090114.64834-3-eesposit@redhat.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agopython: qemu: add timer parameter for qmp.accept socket
Emanuele Giuseppe Esposito [Mon, 9 Aug 2021 09:00:59 +0000 (11:00 +0200)]
python: qemu: add timer parameter for qmp.accept socket

Also add a new _qmp_timer field to the QEMUMachine class.

Let's change the default socket timeout to None, so that if
a subclass needs to add a timer, it can be done by modifying
this private field.

At the same time, restore the timer to be 15 seconds in iotests.py, to
give an upper bound to the QMP monitor test command execution.

Signed-off-by: Emanuele Giuseppe Esposito <eesposit@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: John Snow <jsnow@redhat.com>
Acked-by: John Snow <jsnow@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20210809090114.64834-2-eesposit@redhat.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
2 years agoarm: Remove system_clock_scale global
Peter Maydell [Thu, 12 Aug 2021 09:33:56 +0000 (10:33 +0100)]
arm: Remove system_clock_scale global

All the devices that used to use system_clock_scale have now been
converted to use Clock inputs instead, so the global is no longer
needed; remove it and all the code that sets it.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-id: 20210812093356.1946-26-peter.maydell@linaro.org

2 years agohw/timer/stellaris-gptm: Use Clock input instead of system_clock_scale
Peter Maydell [Thu, 12 Aug 2021 09:33:55 +0000 (10:33 +0100)]
hw/timer/stellaris-gptm: Use Clock input instead of system_clock_scale

The stellaris-gptm timer currently uses system_clock_scale for one of
its timer modes where the timer runs at the CPU clock rate.  Make it
use a Clock input instead.

We don't try to make the timer handle changes in the clock frequency
while the downcounter is running.  This is not a change in behaviour
from the previous system_clock_scale implementation -- we will pick
up the new frequency only when the downcounter hits zero.  Handling
dynamic clock changes when the counter is running would require state
that the current gptm implementation doesn't have.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Damien Hedde <damien.hedde@greensocs.com>
Message-id: 20210812093356.1946-25-peter.maydell@linaro.org

2 years agohw/arm/stellaris: Split stellaris-gptm into its own file
Peter Maydell [Thu, 12 Aug 2021 09:33:54 +0000 (10:33 +0100)]
hw/arm/stellaris: Split stellaris-gptm into its own file

The implementation of the Stellaris general purpose timer module
device stellaris-gptm is currently in the same source file as the
board model.  Split it out into its own source file in hw/timer.

Apart from the new file comment headers and the Kconfig and
meson.build changes, this is just code movement.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Damien Hedde <damien.hedde@greensocs.com>
Message-id: 20210812093356.1946-24-peter.maydell@linaro.org

2 years agohw/arm/stellaris: Fix code style issues in GPTM code
Peter Maydell [Thu, 12 Aug 2021 09:33:53 +0000 (10:33 +0100)]
hw/arm/stellaris: Fix code style issues in GPTM code

Fix the code style issues in the Stellaris general purpose timer
module code, so that when we move it to a different file in a
following patch checkpatch doesn't complain.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Alexandre Iooss <erdnaxe@crans.org>
Message-id: 20210812093356.1946-23-peter.maydell@linaro.org

2 years agohw/timer/armv7m_systick: Use clock inputs instead of system_clock_scale
Peter Maydell [Thu, 12 Aug 2021 09:33:52 +0000 (10:33 +0100)]
hw/timer/armv7m_systick: Use clock inputs instead of system_clock_scale

Now that all users of the systick devices wire up the clock inputs,
use those instead of the system_clock_scale and the hardwired 1MHz
value for the reference clock.

This will fix various board models where we were incorrectly
providing a 1MHz reference clock instead of some other value or
instead of providing no reference clock at all.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Damien Hedde <damien.hedde@greensocs.com>
Message-id: 20210812093356.1946-22-peter.maydell@linaro.org

2 years agohw/arm/msf2-soc: Wire up refclk
Peter Maydell [Thu, 12 Aug 2021 09:33:51 +0000 (10:33 +0100)]
hw/arm/msf2-soc: Wire up refclk

Wire up the refclk for the msf2 SoC.  This SoC runs the refclk at a
frequency which is programmably either /4, /8, /16 or /32 of the main
CPU clock.  We don't currently model the register which allows the
guest to set the divisor, so implement the refclk as a fixed /32 of
the CPU clock (which is the value of the divisor at reset).

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Damien Hedde <damien.hedde@greensocs.com>
Message-id: 20210812093356.1946-21-peter.maydell@linaro.org

2 years agohw/arm/msf2: Use Clock input to MSF2_SOC instead of m3clk property
Peter Maydell [Thu, 12 Aug 2021 09:33:50 +0000 (10:33 +0100)]
hw/arm/msf2: Use Clock input to MSF2_SOC instead of m3clk property

Instead of passing the MSF2 SoC an integer property specifying the
CPU clock rate, pass it a Clock instead.  This lets us wire that
clock up to the armv7m object.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Alexandre Iooss <erdnaxe@crans.org>
Message-id: 20210812093356.1946-20-peter.maydell@linaro.org

2 years agohw/arm/msf2_soc: Don't allocate separate MemoryRegions
Peter Maydell [Thu, 12 Aug 2021 09:33:49 +0000 (10:33 +0100)]
hw/arm/msf2_soc: Don't allocate separate MemoryRegions

In the realize method of the msf2-soc SoC object, we call g_new() to
create new MemoryRegion objects for the nvm, nvm_alias, and sram.
This is unnecessary; make these MemoryRegions member fields of the
device state struct instead.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Alexandre Iooss <erdnaxe@crans.org>
Message-id: 20210812093356.1946-19-peter.maydell@linaro.org

2 years agohw/arm/stellaris: Wire sysclk up to armv7m
Peter Maydell [Thu, 12 Aug 2021 09:33:48 +0000 (10:33 +0100)]
hw/arm/stellaris: Wire sysclk up to armv7m

Connect the sysclk to the armv7m object.  This board's SoC does not
connect up the systick reference clock, so we don't need to connect a
refclk.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Alexandre Iooss <erdnaxe@crans.org>
Message-id: 20210812093356.1946-18-peter.maydell@linaro.org

2 years agohw/arm/stellaris: split stellaris_sys_init()
Peter Maydell [Thu, 12 Aug 2021 09:33:47 +0000 (10:33 +0100)]
hw/arm/stellaris: split stellaris_sys_init()

Currently the stellaris_sys_init() function creates the
TYPE_STELLARIS_SYS object, sets its properties, realizes it, maps its
MMIO region and connects its IRQ.  In order to support wiring the
sysclk up to the armv7m object, we need to split this function apart,
because to connect the clock output of the STELLARIS_SYS object to
the armv7m object we need to create the STELLARIS_SYS object before
the armv7m object, but we can't wire up the IRQ until after we've
created the armv7m object.

Remove the stellaris_sys_init() function, and instead put the
create/configure/realize parts before we create the armv7m object and
the mmio/irq connection parts afterwards.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Alexandre Iooss <erdnaxe@crans.org>
Message-id: 20210812093356.1946-17-peter.maydell@linaro.org

2 years agohw/arm/nrf51: Wire up sysclk
Peter Maydell [Thu, 12 Aug 2021 09:33:46 +0000 (10:33 +0100)]
hw/arm/nrf51: Wire up sysclk

Wire up the sysclk input to the armv7m object.

Strictly this SoC should not have a systick device at all, but our
armv7m container object doesn't currently support disabling the
systick device.  For the moment, add a TODO comment, but note that
this is why we aren't wiring up a refclk (no need for one).

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Alexandre Iooss <erdnaxe@crans.org>
Message-id: 20210812093356.1946-16-peter.maydell@linaro.org

2 years agohw/arm/stm32vldiscovery: Delete trailing blank line
Peter Maydell [Thu, 12 Aug 2021 09:33:45 +0000 (10:33 +0100)]
hw/arm/stm32vldiscovery: Delete trailing blank line

Delete the trailing blank line at the end of the source file.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Alexandre Iooss <erdnaxe@crans.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Luc Michel <luc@lmichel.fr>
Message-id: 20210812093356.1946-15-peter.maydell@linaro.org

2 years agohw/arm/stm32f405: Wire up sysclk and refclk
Peter Maydell [Thu, 12 Aug 2021 09:33:44 +0000 (10:33 +0100)]
hw/arm/stm32f405: Wire up sysclk and refclk

Wire up the sysclk and refclk for the stm32f405 SoC.  This SoC always
runs the systick refclk at 1/8 the frequency of the main CPU clock,
so the board code only needs to provide a single sysclk clock.

Because there is only one board using this SoC, we convert the SoC
and the board together, rather than splitting it into "add clock to
SoC; connect clock in board; add error check in SoC code that clock
is wired up".

When the systick device starts honouring its clock inputs, this will
fix an emulation inaccuracy in the netduinoplus2 board where the
systick reference clock was running at 1MHz rather than 21MHz.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Alexandre Iooss <erdnaxe@crans.org>
Reviewed-by: Luc Michel <luc@lmichel.fr>
Message-id: 20210812093356.1946-14-peter.maydell@linaro.org

2 years agohw/arm/stm32f205: Wire up sysclk and refclk
Peter Maydell [Thu, 12 Aug 2021 09:33:43 +0000 (10:33 +0100)]
hw/arm/stm32f205: Wire up sysclk and refclk

Wire up the sysclk and refclk for the stm32f205 SoC.  This SoC always
runs the systick refclk at 1/8 the frequency of the main CPU clock,
so the board code only needs to provide a single sysclk clock.

Because there is only one board using this SoC, we convert the SoC
and the board together, rather than splitting it into "add clock to
SoC; connect clock in board; add error check in SoC code that clock
is wired up".

When the systick device starts honouring its clock inputs, this will
fix an emulation inaccuracy in the netduino2 board where the systick
reference clock was running at 1MHz rather than 15MHz.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Alexandre Iooss <erdnaxe@crans.org>
Reviewed-by: Luc Michel <luc@lmichel.fr>
Message-id: 20210812093356.1946-13-peter.maydell@linaro.org

2 years agohw/arm/stm32f100: Wire up sysclk and refclk
Peter Maydell [Thu, 12 Aug 2021 09:33:42 +0000 (10:33 +0100)]
hw/arm/stm32f100: Wire up sysclk and refclk

Wire up the sysclk and refclk for the stm32f100 SoC.  This SoC always
runs the systick refclk at 1/8 the frequency of the main CPU clock,
so the board code only needs to provide a single sysclk clock.

Because there is only one board using this SoC, we convert the SoC
and the board together, rather than splitting it into "add clock to
SoC; connect clock in board; add error check in SoC code that clock
is wired up".

When the systick device starts honouring its clock inputs, this will
fix an emulation inaccuracy in the stm32vldiscovery board where the
systick reference clock was running at 1MHz rather than 3MHz.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Alexandre Iooss <erdnaxe@crans.org>
Reviewed-by: Luc Michel <luc@lmichel.fr>
Message-id: 20210812093356.1946-12-peter.maydell@linaro.org

2 years agohw/arm: Don't allocate separate MemoryRegions in stm32 SoC realize
Peter Maydell [Thu, 12 Aug 2021 09:33:41 +0000 (10:33 +0100)]
hw/arm: Don't allocate separate MemoryRegions in stm32 SoC realize

In the realize methods of the stm32f100 and stm32f205 SoC objects, we
call g_new() to create new MemoryRegion objects for the sram, flash,
and flash_alias.  This is unnecessary (and leaves open the
possibility of leaking the allocations if we exit from realize with
an error).  Make these MemoryRegions member fields of the device
state struct instead, as stm32f405 already does.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Alexandre Iooss <erdnaxe@crans.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Luc Michel <luc@lmichel.fr>
Message-id: 20210812093356.1946-11-peter.maydell@linaro.org

2 years agoclock: Provide builtin multiplier/divider
Peter Maydell [Thu, 12 Aug 2021 09:33:40 +0000 (10:33 +0100)]
clock: Provide builtin multiplier/divider

It is quite common for a clock tree to involve possibly programmable
clock multipliers or dividers, where the frequency of a clock is for
instance divided by 8 to produce a slower clock to feed to a
particular device.

Currently we provide no convenient mechanism for modelling this.  You
can implement it by having an input Clock and an output Clock, and
manually setting the period of the output clock in the period-changed
callback of the input clock, but that's quite clunky.

This patch adds support in the Clock objects themselves for setting a
multiplier or divider.  The effect of setting this on a clock is that
when the clock's period is changed, all the children of the clock are
set to period * multiplier / divider, rather than being set to the
same period as the parent clock.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Alexandre Iooss <erdnaxe@crans.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Luc Michel <luc@lmichel.fr>
Reviewed-by: Damien Hedde <damien.hedde@greensocs.com>
Message-id: 20210812093356.1946-10-peter.maydell@linaro.org

2 years agohw/arm/mps2.c: Connect up armv7m clocks
Peter Maydell [Thu, 12 Aug 2021 09:33:39 +0000 (10:33 +0100)]
hw/arm/mps2.c: Connect up armv7m clocks

Connect up the armv7m clocks on the mps2-an385/386/500/511.

Connect up the armv7m object's clocks on the MPS boards defined in
mps2.c.  The documentation for these FPGA images doesn't specify what
systick reference clock is used (if any), so for the moment we
provide a 1MHz refclock, which will result in no behavioural change
from the current hardwired 1MHz clock implemented in
armv7m_systick.c:systick_scale().

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Luc Michel <luc@lmichel.fr>
Message-id: 20210812093356.1946-9-peter.maydell@linaro.org

2 years agoarmsse: Wire up systick cpuclk clock
Peter Maydell [Thu, 12 Aug 2021 09:33:38 +0000 (10:33 +0100)]
armsse: Wire up systick cpuclk clock

Wire up the cpuclk for the systick devices to the SSE object's
existing mainclk clock.

We do not wire up the refclk because the SSE subsystems do not
provide a refclk.  (This is documented in the IoTKit and SSE-200
TRMs; the SSE-300 TRM doesn't mention it but we assume it follows the
same approach.) When we update the systick device later to honour "no
refclk connected" this will fix a minor emulation inaccuracy for the
SSE-based boards.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Acked-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Luc Michel <luc@lmichel.fr>
Message-id: 20210812093356.1946-8-peter.maydell@linaro.org

2 years agohw/arm/armv7m: Create input clocks
Peter Maydell [Thu, 12 Aug 2021 09:33:37 +0000 (10:33 +0100)]
hw/arm/armv7m: Create input clocks

Create input clocks on the armv7m container object which pass through
to the systick timers, so that users of the armv7m object can specify
the clocks being used.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Luc Michel <luc@lmichel.fr>
Message-id: 20210812093356.1946-7-peter.maydell@linaro.org

2 years agohw/timer/armv7m_systick: Add input clocks
Peter Maydell [Thu, 12 Aug 2021 09:33:36 +0000 (10:33 +0100)]
hw/timer/armv7m_systick: Add input clocks

The v7M systick timer can be programmed to run from either of
two clocks:
 * an "external reference clock" (when SYST_CSR.CLKSOURCE == 0)
 * the main CPU clock (when SYST_CSR.CLKSOURCE == 1)

Our implementation currently hardwires the external reference clock
to be 1MHz, and allows boards to set the main CPU clock frequency via
the global 'system_clock_scale'.  (Most boards set that to a constant
value; the Stellaris boards allow the guest to reprogram it via the
board-specific RCC registers).

As the first step in converting this to use the Clock infrastructure,
add input clocks to the systick device for the reference clock and
the CPU clock.  The device implementation ignores them; once we have
made all the users of the device correctly wire up the new Clocks we
will switch the implementation to use them and ignore the old
system_clock_scale.

This is a migration compat break for all M-profile boards, because of
the addition of the new clock objects to the vmstate struct.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Luc Michel <luc@lmichel.fr>
Message-id: 20210812093356.1946-6-peter.maydell@linaro.org