since dirty ring has been introduced, there are two methods
to track dirty pages of vm. it seems that "logging" has
a hint on the method, so rename the global_dirty_log to
global_dirty_tracking would make description more accurate.
dirty rate measurement may start or stop dirty tracking during
calculation. this conflict with migration because stop dirty
tracking make migration leave dirty pages out then that'll be
a problem.
make global_dirty_tracking a bitmask can let both migration and
dirty rate measurement work fine. introduce GLOBAL_DIRTY_MIGRATION
and GLOBAL_DIRTY_DIRTY_RATE to distinguish what current dirty
tracking aims for, migration or dirty rate.
Signed-off-by: Hyman Huang(黄勇) <huangy81@chinatelecom.cn>
Message-Id: <9c9388657cfa0301bd2c1cfa36e7cf6da4aeca19.1624040308.git.huangy81@chinatelecom.cn>
Reviewed-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
dirty_pages is used to calculate dirtyrate via dirty ring, when
enabled, kvm-reaper will increase the dirty pages after gfns
being dirtied.
kvm_dirty_ring_enabled shows if kvm-reaper is working. dirtyrate
thread could use it to check if measurement can base on dirty
ring feature.
Signed-off-by: Hyman Huang(黄勇) <huangy81@chinatelecom.cn>
Message-Id: <fee5fb2ab17ec2159405fc54a3cff8e02322f816.1624040308.git.huangy81@chinatelecom.cn>
Reviewed-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
Now that we check sysbus device types during device creation, we
can remove the check in the machine init done notifier.
This was the only thing done by this notifier, so we remove the
whole sysbus_notifier structure of the MachineState.
Note: This notifier was checking all /peripheral and /peripheral-anon
sysbus devices. Now we only check those added by -device cli option or
device_add qmp command when handling the command/option. So if there
are some devices added in one of these containers manually (eg in
machine C code), these will not be checked anymore.
This use case does not seem to appear apart from
hw/xen/xen-legacy-backend.c (it uses qdev_set_id() and in this case,
not for a sysbus device, so it's ok).
Signed-off-by: Damien Hedde <damien.hedde@greensocs.com>
Acked-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Acked-by: Eduardo Habkost <ehabkost@redhat.com>
Message-Id: <20211029142258.484907-4-damien.hedde@greensocs.com>
Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Add an early check to test if the requested sysbus device type
is allowed by the current machine before creating the device. This
impacts both -device cli option and device_add qmp command.
Before this patch, the check was done well after the device has
been created (in a machine init done notifier). We can now report
the error right away.
Signed-off-by: Damien Hedde <damien.hedde@greensocs.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Acked-by: Eduardo Habkost <ehabkost@redhat.com>
Message-Id: <20211029142258.484907-3-damien.hedde@greensocs.com>
Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Right now the allowance check for adding a sysbus device using
-device cli option (or device_add qmp command) is done well after
the device has been created. It is done during the machine init done
notifier: machine_init_notify() in hw/core/machine.c
This new function will allow us to do the check at the right time and
issue an error if it fails.
Also make device_is_dynamic_sysbus() use the new function.
Signed-off-by: Damien Hedde <damien.hedde@greensocs.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Acked-by: Eduardo Habkost <ehabkost@redhat.com>
Message-Id: <20211029142258.484907-2-damien.hedde@greensocs.com>
Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Now that we have a generic parser smp_parse(), let's add an unit
test for the code. All possible valid/invalid SMP configurations
that the user can specify are covered.
Signed-off-by: Yanan Wang <wangyanan55@huawei.com>
Reviewed-by: Andrew Jones <drjones@redhat.com>
Tested-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20211026034659.22040-3-wangyanan55@huawei.com>
Acked-by: Eduardo Habkost <ehabkost@redhat.com>
Message-Id: <bfed7144-af86-7098-e7a6-731ff13c2cf7@huawei.com>
[PMD: Squashed format string fixup from Yanan Wang]
Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
We are going to introduce an unit test for the parser smp_parse()
in hw/core/machine.c, but now machine.c is only built in softmmu.
In order to solve the build dependency on the smp parsing code and
avoid building unrelated stuff for the unit tests, move the tested
code from machine.c into a separate file, i.e., machine-smp.c and
build it in common field.
Signed-off-by: Yanan Wang <wangyanan55@huawei.com>
Reviewed-by: Andrew Jones <drjones@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Tested-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20211026034659.22040-2-wangyanan55@huawei.com>
Acked-by: Eduardo Habkost <ehabkost@redhat.com>
Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Restrict hotplug to system emulation, add stubs for the other uses.
Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Yanan Wang <wangyanan55@huawei.com>
Acked-by: Eduardo Habkost <ehabkost@redhat.com>
Message-Id: <20211028150521.1973821-5-philmd@redhat.com>
As we want to be able to conditionally add files to the hw/core
file list, use a source set.
Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Yanan Wang <wangyanan55@huawei.com>
Tested-by: Yanan Wang <wangyanan55@huawei.com>
Acked-by: Eduardo Habkost <ehabkost@redhat.com>
Message-Id: <20211028150521.1973821-3-philmd@redhat.com>
All these files don't make sense for tools and user emulation,
restrict them to system emulation.
Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Yanan Wang <wangyanan55@huawei.com>
Tested-by: Yanan Wang <wangyanan55@huawei.com>
Acked-by: Eduardo Habkost <ehabkost@redhat.com>
Message-Id: <20211028150521.1973821-2-philmd@redhat.com>
Only softmmu code uses gpio, so move gpio code from qdev.c to
gpio.c and compile it only on softmmu mode.
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Tested-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20190425200051.19906-2-ehabkost@redhat.com>
[PMD: Rebased]
Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
The MSI-X structures of some devices and other non-MSI-X structures
may be in the same BAR. They may share one host page, especially in
the case of large page granularity, such as 64K.
For example, MSIX-Table size of 82599 NIC is 0x30 and the offset in
Bar 3(size 64KB) is 0x0. vfio_listener_region_add() will be called
to map the remaining range (0x30-0xffff). If host page size is 64KB,
it will return early at 'int128_ge((int128_make64(iova), llend))'
without any message. Let's add a trace point to inform users like commit
5c08600547 ("vfio: Use a trace point when a RAM section cannot be DMA mapped")
did.
Signed-off-by: Kunkun Jiang <jiangkunkun@huawei.com>
Link: https://lore.kernel.org/r/20211027090406.761-3-jiangkunkun@huawei.com
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
We can expand MemoryRegions of sub-page MMIO BARs in
vfio_pci_write_config() to improve IO performance for some
devices. However, the MemoryRegions of destination VM are
not expanded any more after live migration. Because their
addresses have been updated in vmstate_load_state()
(vfio_pci_load_config) and vfio_sub_page_bar_update_mapping()
will not be called.
This may result in poor performance after live migration.
So iterate BARs in vfio_pci_load_config() and try to update
sub-page BARs.
Reported-by: Nianyao Tang <tangnianyao@huawei.com>
Reported-by: Qixin Gan <ganqixin@huawei.com>
Signed-off-by: Kunkun Jiang <jiangkunkun@huawei.com>
Link: https://lore.kernel.org/r/20211027090406.761-2-jiangkunkun@huawei.com
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Swap out the synchronous QEMUMonitorProtocol from qemu.qmp with the sync
wrapper from qemu.aqmp instead.
Add an escape hatch in the form of the environment variable
QEMU_PYTHON_LEGACY_QMP which allows you to cajole QEMUMachine into using
the old implementation, proving that both implementations work
concurrently.
Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Hanna Reitz <hreitz@redhat.com>
Message-id: 20211026175612.4127598-9-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
This is a wrapper around the async QMPClient that mimics the old,
synchronous QEMUMonitorProtocol class. It is designed to be
interchangeable with the old implementation.
It does not, however, attempt to mimic Exception compatibility.
Signed-off-by: John Snow <jsnow@redhat.com>
Acked-by: Hanna Reitz <hreitz@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Hanna Reitz <hreitz@redhat.com>
Message-id: 20211026175612.4127598-8-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
Wait for the destination VM to close itself instead of racing to shut it
down first, which produces different error log messages from AQMP
depending on precisely when we tried to shut it down.
(For example: We may try to issue 'quit' immediately prior to the target
VM closing its QMP socket, which will cause an ECONNRESET error to be
logged. Waiting for the VM to exit itself avoids the race on shutdown
behavior.)
Reported-by: Hanna Reitz <hreitz@redhat.com>
Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Hanna Reitz <hreitz@redhat.com>
Message-id: 20211026175612.4127598-7-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
AQMP likes to be very chatty about errors it encounters. In general,
this is good because it allows us to get good diagnostic information for
otherwise complex async failures.
For example, during a failed QMP connection attempt, we might see:
+ERROR:qemu.aqmp.qmp_client.qemub-2536319:Negotiation failed: EOFError
+ERROR:qemu.aqmp.qmp_client.qemub-2536319:Failed to establish session: EOFError
This might be nice in iotests output, because failure scenarios
involving the new QMP library will be spelled out plainly in the output
diffs.
For tests that are intentionally causing this scenario though, filtering
that log output could be a hassle. For now, add a context manager that
simply lets us toggle this output off during a critical region.
(Additionally, a forthcoming patch allows the use of either legacy or
async QMP to be toggled with an environment variable. In this
circumstance, we can't amend the iotest output to just always expect the
error message, either. Just suppress it for now. More rigorous log
filtering can be investigated later if/when it is deemed safe to
permanently replace the legacy QMP library.)
Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Hanna Reitz <hreitz@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Message-id: 20211026175612.4127598-6-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
(But continue to support the old ones for now, too.)
There are very few cases of any user of QEMUMachine or a subclass
thereof relying on a QMP Exception type. If you'd like to check for
yourself, you want to grep for all of the derivatives of QMPError,
excluding 'AQMPError' and its derivatives. That'd be these:
- QMPError
- QMPConnectError
- QMPCapabilitiesError
- QMPTimeoutError
- QMPProtocolError
- QMPResponseError
- QMPBadPortError
Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Hanna Reitz <hreitz@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Message-id: 20211026175612.4127598-5-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
The scary message interferes with the iotests output. Coincidentally, if
iotests works by removing this, then it's good evidence that we don't
really need to scare people away from using it.
Signed-off-by: John Snow <jsnow@redhat.com>
Acked-by: Hanna Reitz <hreitz@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Hanna Reitz <hreitz@redhat.com>
Message-id: 20211026175612.4127598-4-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
To use the AQMP backend, Machine just needs to be a little more diligent
about what happens when closing a QMP connection. The operation is no
longer a freebie in the async world; it may return errors encountered in
the async bottom half on incoming message receipt, etc.
(AQMP's disconnect, ultimately, serves as the quiescence point where all
async contexts are gathered together, and any final errors reported at
that point.)
Because async QMP continues to check for messages asynchronously, it's
almost certainly likely that the loop will have exited due to EOF after
issuing the last 'quit' command. That error will ultimately be bubbled
up when attempting to close the QMP connection. The manager class here
then is free to discard it -- if it was expected.
Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Hanna Reitz <hreitz@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Message-id: 20211026175612.4127598-3-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
If we spy on the QMP commands instead, we don't need callers to remember
to pass it. Seems like a fair trade-off.
The one slightly weird bit is overloading this instance variable for
wait(), where we use it to mean "don't issue the qmp 'quit'
command". This means that wait() will "fail" if the QEMU process does
not terminate of its own accord.
In most cases, we probably did already actually issue quit -- some
iotests do this -- but in some others, we may be waiting for QEMU to
terminate for some other reason, such as a test wherein we tell the
guest (directly) to shut down.
Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Hanna Reitz <hreitz@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Message-id: 20211026175612.4127598-2-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
Run mypy and pylint on the iotests files directly from the Python CI
test infrastructure. This ensures that any accidental breakages to the
qemu.[qmp|aqmp|machine|utils] packages will be caught by that test
suite.
It also ensures that these linters are run with well-known versions and
test against a wide variety of python versions, which helps to find
accidental cross-version python compatibility issues.
Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Hanna Reitz <hreitz@redhat.com>
Message-id: 20211019144918.3159078-15-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
This one is insidious: if you write an import as "from {namespace}
import {subpackage}" as mirror-top-perms (now) does, mypy will fail on
every-other invocation *if* the package being imported is a typed,
installed, namespace-scoped package.
Upsettingly, that's exactly what 'qemu.[aqmp|qmp|machine]' et al are in
the context of Python CI tests.
Now, I could just edit mirror-top-perms to avoid this invocation, but
since I tripped on a landmine, I might as well head it off at the pass
and make sure nobody else trips on that same landmine.
It seems to have something to do with the order in which files are
checked as well, meaning the random order in which set(os.listdir())
produces the list of files to test will cause problems intermittently
and not just strictly "every other run".
This will be fixed in mypy >= 0.920, which is not released yet. The
workaround for now is to disable incremental checking, which avoids the
issue.
Note: This workaround is not applied when running iotest 297 directly,
because the bug does not surface there! Given the nature of CI jobs not
starting with any stale cache to begin with, this really only has a
half-second impact on manual runs of the Python test suite when executed
directly by a developer on their local machine. The workaround may be
removed when the Python package requirements can stipulate mypy 0.920 or
higher, which can happen as soon as it is released. (Barring any
unforseen compatibility issues that 0.920 may bring with it.)
See also:
https://github.com/python/mypy/issues/11010https://github.com/python/mypy/issues/9852
Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Hanna Reitz <hreitz@redhat.com>
Message-id: 20211019144918.3159078-14-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
We need at least a tiny little shim here to join test file discovery
with test invocation. This logic could conceivably be hosted somewhere
in python/, but I felt it was strictly the least-rude thing to keep the
test logic here in iotests/, even if this small function isn't itself an
iotest.
Note that we don't actually even need the executable bit here, we'll be
relying on the ability to run this module as a script using Python CLI
arguments. No chance it gets misunderstood as an actual iotest that way.
(It's named, not in tests/, doesn't have the execute bit, and doesn't
have an execution shebang.)
Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Hanna Reitz <hreitz@redhat.com>
Message-id: 20211019144918.3159078-13-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
Now, 297 is just the iotests-specific incantations and linters.py is as
minimal as I can think to make it. The only remaining element in here
that ought to be configuration and not code is the list of skip files,
but they're still numerous enough that repeating them for mypy and
pylint configurations both would be ... a hassle.
Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Hanna Reitz <hreitz@redhat.com>
Message-id: 20211019144918.3159078-12-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
Take iotest 297's main() test function and split it into two sub-cases
that can be skipped individually. We can also drop custom environment
setup from the pylint test as it isn't needed.
Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Hanna Reitz <hreitz@redhat.com>
Message-id: 20211019144918.3159078-11-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
As mentioned in 'iotests/297: Don't rely on distro-specific linter
binaries', these checks are overly strict. Update them to be in-line
with how we actually invoke the linters themselves.
Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Hanna Reitz <hreitz@redhat.com>
Message-id: 20211019144918.3159078-10-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
Instead of using a process return code as the python function return
value (or just not returning anything at all), allow run_linter() to
raise an exception instead.
The responsibility for printing output on error shifts from the function
itself to the caller, who will know best how to present/format that
information. (Also, "suppress_output" is now a lot more accurate of a
parameter name.)
Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Hanna Reitz <hreitz@redhat.com>
Message-id: 20211019144918.3159078-9-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
There's virtually nothing special here anymore; we can combine these
into a single, rather generic function.
Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Hanna Reitz <hreitz@redhat.com>
Message-id: 20211019144918.3159078-8-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
Move environment setup into main(), and split the actual linter
execution into run_pylint and run_mypy, respectively.
Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Hanna Reitz <hreitz@redhat.com>
Message-id: 20211019144918.3159078-7-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
'pylint-3' is another Fedora-ism. Use "python3 -m pylint" or "python3 -m
mypy" to access these scripts instead. This style of invocation will
prefer the "correct" tool when run in a virtual environment.
Note that we still check for "pylint-3" before the test begins -- this
check is now "overly strict", but shouldn't cause anything that was
already running correctly to start failing. This is addressed by a
commit later in this series;
'iotests/297: update tool availability checks'.
Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Hanna Reitz <hreitz@redhat.com>
Message-id: 20211019144918.3159078-6-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
Instead of running "run_linters" directly, create a main() function that
will be responsible for environment setup, leaving run_linters()
responsible only for execution of the linters.
(That environment setup will be moved over in forthcoming commits.)
Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Hanna Reitz <hreitz@redhat.com>
Message-id: 20211019144918.3159078-5-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
Split out file discovery into its own method to begin separating out
configuration/setup and test execution.
Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Hanna Reitz <hreitz@redhat.com>
Message-id: 20211019144918.3159078-4-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
More separation of code and configuration.
Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Hanna Reitz <hreitz@redhat.com>
Message-id: 20211019144918.3159078-3-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
Move --score=n and --notes=XXX,FIXME into pylintrc. This pulls
configuration out of code, which I think is probably a good thing in
general.
Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Hanna Reitz <hreitz@redhat.com>
Message-id: 20211019144918.3159078-2-jsnow@redhat.com
Signed-off-by: John Snow <jsnow@redhat.com>
destination:
../qemu/build/qemu-system-x86_64 -enable-kvm -netdev tap,id=hn0,script=/etc/qemu-ifup,downscript=/etc/qemu-ifdown -device e1000,netdev=hn0,mac=50:52:54:00:11:22 -boot c -drive if=none,file=./Fedora-rdma-server-migration.qcow2,id=drive-virtio-disk0 -device virtio-blk-pci,bus=pci.0,addr=0x4,drive=drive-virtio-disk0,id=virtio-disk0 -m 2048 -smp 2 -device piix3-usb-uhci -device usb-tablet -monitor stdio -vga qxl -spice streaming-video=filter,port=5902,disable-ticketing -incoming rdma:192.168.22.23:8888
qemu-system-x86_64: -spice streaming-video=filter,port=5902,disable-ticketing: warning: short-form boolean option 'disable-ticketing' deprecated
Please use disable-ticketing=on instead
QEMU 6.0.50 monitor - type 'help' for more information
(qemu) trace-event qemu_rdma_block_for_wrid_miss on
(qemu) dest_init RDMA Device opened: kernel name rxe_eth0 uverbs device name uverbs2, infiniband_verbs class device path /sys/class/infiniband_verbs/uverbs2, infiniband class device path /sys/class/infiniband/rxe_eth0, transport: (2) Ethernet
qemu_rdma_block_for_wrid_miss A Wanted wrid CONTROL SEND (2000) but got CONTROL RECV (4000)
source:
../qemu/build/qemu-system-x86_64 -enable-kvm -netdev tap,id=hn0,script=/etc/qemu-ifup,downscript=/etc/qemu-ifdown -device e1000,netdev=hn0,mac=50:52:54:00:11:22 -boot c -drive if=none,file=./Fedora-rdma-server.qcow2,id=drive-virtio-disk0 -device virtio-blk-pci,bus=pci.0,addr=0x4,drive=drive-virtio-disk0,id=virtio-disk0 -m 2048 -smp 2 -device piix3-usb-uhci -device usb-tablet -monitor stdio -vga qxl -spice streaming-video=filter,port=5901,disable-ticketing -S
qemu-system-x86_64: -spice streaming-video=filter,port=5901,disable-ticketing: warning: short-form boolean option 'disable-ticketing' deprecated
Please use disable-ticketing=on instead
QEMU 6.0.50 monitor - type 'help' for more information
(qemu)
(qemu) trace-event qemu_rdma_block_for_wrid_miss on
(qemu) migrate -d rdma:192.168.22.23:8888
source_resolve_host RDMA Device opened: kernel name rxe_eth0 uverbs device name uverbs2, infiniband_verbs class device path /sys/class/infiniband_verbs/uverbs2, infiniband class device path /sys/class/infiniband/rxe_eth0, transport: (2) Ethernet
(qemu) qemu_rdma_block_for_wrid_miss A Wanted wrid WRITE RDMA (1) but got CONTROL RECV (4000)
NOTE: we use soft RoCE as the rdma device.
[root@iaas-rpma images]# rdma link show rxe_eth0/1
link rxe_eth0/1 state ACTIVE physical_state LINK_UP netdev eth0
This migration could not be completed when out of order(OOO) CQ event occurs.
The send queue and receive queue shared a same completion queue, and
qemu_rdma_block_for_wrid() will drop the CQs it's not interested in. But
the dropped CQs by qemu_rdma_block_for_wrid() could be later CQs it wants.
So in this case, qemu_rdma_block_for_wrid() will block forever.
OOO cases will occur in both source side and destination side. And a
forever blocking happens on only SEND and RECV are out of order. OOO between
'WRITE RDMA' and 'RECV' doesn't matter.
below the OOO sequence:
source destination
rdma_write_one() qemu_rdma_registration_handle()
1. S1: post_recv X D1: post_recv Y
2. wait for recv CQ event X
3. D2: post_send X ---------------+
4. wait for send CQ send event X (D2) |
5. recv CQ event X reaches (D2) |
6. +-S2: post_send Y |
7. | wait for send CQ event Y |
8. | recv CQ event Y (S2) (drop it) |
9. +-send CQ event Y reaches (S2) |
10. send CQ event X reaches (D2) -----+
11. wait recv CQ event Y (dropped by (8))
Although a hardware IB works fine in my a hundred of runs, the IB specification
doesn't guaratee the CQ order in such case.
Here we introduce a independent send completion queue to distinguish
ibv_post_send completion queue from the original mixed completion queue.
It helps us to poll the specific CQE we are really interested in.
Signed-off-by: Li Zhijian <lizhijian@cn.fujitsu.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
Patches from Zoltan:
- Various clean up to align the code style with the rest of the code base
- QOM'ify the SH_SERIAL device
- Modify few memory region size to better match the hardware manual
-----BEGIN PGP SIGNATURE-----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=ktI+
-----END PGP SIGNATURE-----
Merge remote-tracking branch 'remotes/philmd/tags/renesas-20211030' into staging
Renesas SH-4 patches queue
Patches from Zoltan:
- Various clean up to align the code style with the rest of the code base
- QOM'ify the SH_SERIAL device
- Modify few memory region size to better match the hardware manual
# gpg: Signature made Sat 30 Oct 2021 10:05:03 AM PDT
# gpg: using RSA key FAABE75E12917221DCFD6BB2E3E32C2CDEADC0DE
# gpg: Good signature from "Philippe Mathieu-Daudé (F4BUG) <f4bug@amsat.org>" [full]
* remotes/philmd/tags/renesas-20211030: (30 commits)
hw/timer/sh_timer: Remove use of hw_error
hw/timer/sh_timer: Fix timer memory region size
hw/timer/sh_timer: Do not wrap lines that are not too long
hw/timer/sh_timer: Rename sh_timer_state to SHTimerState
hw/intc/sh_intc: Remove unneeded local variable initialisers
hw/intc/sh_intc: Simplify allocating sources array
hw/intc/sh_intc: Avoid using continue in loops
hw/intc/sh_intc: Replace abort() with g_assert_not_reached()
hw/intc/sh_intc: Inline and drop sh_intc_source() function
hw/intc/sh_intc: Use array index instead of pointer arithmetics
hw/intc/sh_intc: Remove excessive parenthesis
hw/intc/sh_intc: Move sh_intc_register() closer to its only user
hw/intc/sh_intc: Drop another useless macro
hw/intc/sh_intc: Rename iomem region
hw/intc/sh_intc: Turn some defines into an enum
hw/intc/sh_intc: Use existing macro instead of local one
hw/char/sh_serial: Add device id to trace output
hw/char/sh_serial: QOM-ify
hw/char/sh_serial: Split off sh_serial_reset() from sh_serial_init()
hw/char/sh_serial: Embed QEMUTimer in state struct
...
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Add stub host-signal.h for all linux-user hosts.
Add new code replacing cpu_signal_handler.
Full migration will happen one host at a time.
Reviewed-by: Warner Losh <imp@bsdimp.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Acked-by: Alistair Francis <alistair.francis@wdc.com>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
The existing code for safe-syscall.inc.S will compile
without change for riscv32 and riscv64. We may also
drop the meson.build stanza that merges them for tcg/.
Reviewed-by: Warner Losh <imp@bsdimp.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Remove the comment about siglongjmp. We do use sigsetjmp
in the main cpu loop, but we do not save the signal mask
as most exits from the cpu loop do not require them.
Reviewed-by: Warner Losh <imp@bsdimp.com>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
This is the major portion of handle_cpu_signal which is specific
to tcg, handling the page protections for the translations.
Most of the rest will migrate to linux-user/ shortly.
Reviewed-by: Warner Losh <imp@bsdimp.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
v2: Pass guest address to handle_sigsegv_accerr_write.
Currently there are only two places that require we reset this
value before exiting to the main loop, but that will change.
Reviewed-by: Warner Losh <imp@bsdimp.com>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Split out a function to adjust the raw signal pc into a
value that could be passed to cpu_restore_state.
Reviewed-by: Warner Losh <imp@bsdimp.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
v2: Adjust pc in place; return MMUAccessType.
The hw_error function calls abort and is not meant to be used by
devices. Use qemu_log_mask instead to log and ignore invalid accesses.
Also fix format strings to allow dropping type casts of hwaddr and use
__func__ instead of hard coding function name in the message which
were wrong in two cases.
Signed-off-by: BALATON Zoltan <balaton@eik.bme.hu>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <f818dc3dd2ac8c3b3d53067f316a716d7f9683d8.1635541329.git.balaton@eik.bme.hu>
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
The timer unit only has registers that fit in a region 0x30 bytes
long. No need to have the timer region larger than that.
Signed-off-by: BALATON Zoltan <balaton@eik.bme.hu>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <b1cd196cf1395a602c7a08a4f858e69e50c446a1.1635550060.git.balaton@eik.bme.hu>
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
It's more readable to keep things on one line if it fits the length limit.
Signed-off-by: BALATON Zoltan <balaton@eik.bme.hu>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <97bc2a38991f33fd0c8cc2e4d0a3a29b20c47d1f.1635541329.git.balaton@eik.bme.hu>
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
According to coding style types should be camel case, also remove
unneded casts from void *.
Signed-off-by: BALATON Zoltan <balaton@eik.bme.hu>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <d9a9d160c1153a583397e366ab06477f5a31c507.1635541329.git.balaton@eik.bme.hu>
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Hyman Huang(黄勇)