Currently, the mapping handlers of the IOMMU backends, VFIO IOMMU Type
1 aka. legacy and IOMMUFD, return an errno and also report an error.
This can lead to excessive log messages at runtime for recurring DMA
mapping errors. Since these errors are already reported by the callers
in the vfio_container_dma_un/map() routines, simply remove them and
allow the callers to handle the reporting.
The mapping handler of the IOMMUFD backend has a comment suggesting
MMIO region mapping failures return EFAULT. I am not sure this is
entirely true, so keep the EFAULT case until the conditions are
clarified.
Reviewed-by: Alex Williamson <alex.williamson@redhat.com>
Link: https://lore.kernel.org/qemu-devel/20250206131438.1505542-7-clg@redhat.com
Signed-off-by: Cédric Le Goater <clg@redhat.com>
When the IOMMU address space width is smaller than the physical
address width, a MMIO region of a device can fail to map because the
region is outside the supported IOVA ranges of the VM. In this case,
PCI peer-to-peer transactions on BARs are not supported. This can
occur with the 39-bit IOMMU address space width, as can be the case on
some Intel consumer processors, or when using a vIOMMU device with
default settings.
The current error message is unclear, improve it and also change the
error report to a warning because it is a non fatal condition for the
VM. To prevent excessive log messages, restrict these recurring DMA
mapping errors to a single warning at runtime.
Reviewed-by: Alex Williamson <alex.williamson@redhat.com>
Link: https://lore.kernel.org/qemu-devel/20250206131438.1505542-6-clg@redhat.com
Signed-off-by: Cédric Le Goater <clg@redhat.com>
Depending on the configuration of the host and VM, a passthrough
device may generate recurring DMA mapping errors at runtime. In such
cases, reporting the issue once is sufficient.
We have already the warn/error_report_once() routines taking a format
and arguments. Using the same design pattern, add a new warning
variant taking an 'Error *' parameter.
Cc: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Alex Williamson <alex.williamson@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Link: https://lore.kernel.org/qemu-devel/20250206131438.1505542-2-clg@redhat.com
Signed-off-by: Cédric Le Goater <clg@redhat.com>
The risk is mainly theoretical since the applied bit mask will keep
the 'ggms' shift value below 3. Nevertheless, let's use a 64 bit
integer type and resolve the coverity issue.
Resolves: Coverity CID 1585908
Fixes: 1e1eac5f3d ("vfio/igd: canonicalize memory size calculations")
Reviewed-by: Alex Williamson <alex.williamson@redhat.com>
Link: https://lore.kernel.org/r/20250107130604.669697-1-clg@redhat.com
Signed-off-by: Cédric Le Goater <clg@redhat.com>
In do_cvttq() we set env->error_code with what is supposed to be a
set of FPCR exception bit values. However, if the set of float
exception flags we get back from softfloat for the conversion
includes a flag which is not one of the three we expect here
(invalid_cvti, invalid, inexact) then we will fall through the
if-ladder and set env->error_code to the unconverted softfloat
exception_flag value. This will then cause us to take a spurious
exception.
This is harmless now, but when we add new floating point exception
flags to softfloat it will cause problems. Add an else clause to the
if-ladder to make it ignore any float exception flags it doesn't care
about.
Specifically, without this fix, 'make check-tcg' will fail for Alpha
when the commit adding float_flag_input_denormal_used lands.
Fixes: aa3bad5b59 ("target/alpha: Use float64_to_int64_modulo for CVTTQ")
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Rather than a giant sed command with a hardcoded list of env var name,
we can now use the new(ish) cirrus-vars command that libvirt has added
to the 'cirrus-run' container.
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Message-ID: <20241204194807.1472261-3-berrange@redhat.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
In the last week of the month we have often run out of credits on
Cirrus CI, which causes the jobs to fail, in turn causing the
overall pipeline to fail.
The cirrus-run tool can now detect the "out of credits" scenario
and exits with a code of '3'. We can tell gitlab to treat this
exit code as special and mark the job as "warning" instead of
"failed". This allows the pipeline status overall to remain
green, when we have non-technical issues with Cirrus CI.
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Message-ID: <20241204194807.1472261-2-berrange@redhat.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
The test sequence boots from kernel a sam460ex machine with a
virtio-net device to check PCI.
The buildroot is built with config :
BR2_powerpc=y
BR2_powerpc_440fp=y
and the kernel with the '44x/canyonlands' deconfig and virtio support.
Signed-off-by: Cédric Le Goater <clg@redhat.com>
Message-ID: <20250203092606.491933-1-clg@redhat.com>
[thuth: sort meson.build alphabetically]
Signed-off-by: Thomas Huth <thuth@redhat.com>
testing and gdbstub updates:
- add a check-rust test to docker builds
- re-factor the qtest logic to be cleaner
- fix tests to not clock_step when no timers enabled
- roll-up log prefix into qtest_send
- cleaner error reporting when qtest_clock_set fails
- revert old deadlock fix now tests are updated
- only run full set of migration tests under HW acceleration
- support late attachment to user-mode gdbstubs
# -----BEGIN PGP SIGNATURE-----
#
# iQEzBAABCgAdFiEEZoWumedRZ7yvyN81+9DbCVqeKkQFAmeqBSsACgkQ+9DbCVqe
# KkQS/Af+K0hpdGc1msiuMsqmuESBvhoQniYZFLN1/pwe2KpG8i/+fq2fsCuxJhJ1
# 2TzPH7aj54p9MGCZf2k9JLhO22XldN+oezZMc1crhoWK0AtrWhnLs58I2oEPIsUo
# NmGO6Zfm98ge89o2y8GCvd0QXAtUf+jduDKnW0mfnOnw+w/mky5KzWS7/1091VGW
# 42LSY4KnqgdLSqLyuLBOrgADEjB1ChWS4/bSC+kEYSGrmNQB+n1KeIzzlJBGpOr0
# Z9yzmhMCm7TWdkFNPmnVfYH/7ZUNcpv6PtQSpkku4f6b/gybyvJBknHpM4i+Gpb5
# 87wSjljrCpdNm/9KFRjiJuUWdS/jCg==
# =UF0n
# -----END PGP SIGNATURE-----
# gpg: Signature made Mon 10 Feb 2025 08:54:51 EST
# gpg: using RSA key 6685AE99E75167BCAFC8DF35FBD0DB095A9E2A44
# gpg: Good signature from "Alex Bennée (Master Work Key) <alex.bennee@linaro.org>" [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: 6685 AE99 E751 67BC AFC8 DF35 FBD0 DB09 5A9E 2A44
* tag 'pull-10.0-testing-and-gdstub-updates-100225-1' of https://gitlab.com/stsquad/qemu:
tests/tcg: Add late gdbstub attach test
docs/user: Document the %d placeholder and suspend=n QEMU_GDB features
gdbstub: Allow late attachment
osdep: Introduce qemu_kill_thread()
user: Introduce host_interrupt_signal
user: Introduce user/signal.h
gdbstub: Try unlinking the unix socket before binding
gdbstub: Allow the %d placeholder in the socket path
tests/qtest/migration: Pick smoke tests
tests/qtest/migration: Add --full option
Revert "util/timer: avoid deadlock when shutting down"
tests/qtest: tighten up the checks on clock_step
tests/qtest: rename qtest_send_prefix and roll-up into qtest_send
tests/qtest: simplify qtest_process_inbuf
tests/qtest: don't step clock at start of npcm7xx periodic IRQ test
tests/qtest: don't attempt to clock_step while waiting for virtio ISR
tests/docker: replicate the check-rust-tools-nightly CI job
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
This replaces use of the constants from the QapiSpecialFeatures
enum, with constants from the auto-generate QapiFeatures enum
in qapi-features.h
The 'deprecated' and 'unstable' features still have a little bit of
special handling, being force defined to be the 1st + 2nd features
in the enum, regardless of whether they're used in the schema. This
retains compatibility with common code that references the features
via the QapiSpecialFeatures constants.
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Message-ID: <20250205123550.2754387-5-berrange@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
[Imports tidied up with isort]
Signed-off-by: Markus Armbruster <armbru@redhat.com>
This updates the QAPI code generation to refer to 'features' instead
of 'special_features', in preparation for generalizing their exposure.
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Message-ID: <20250205123550.2754387-4-berrange@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
[Imports tidied up with isort]
Signed-off-by: Markus Armbruster <armbru@redhat.com>
The "special_features" field / parameter holds the subset of schema
features that are for internal code use. Specifically 'DEPRECATED'
and 'UNSTABLE'.
This special casing of internal features is going to be removed, so
prepare for that by renaming to 'features'. Using a fixed size type
is also best practice for bit fields.
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Message-ID: <20250205123550.2754387-3-berrange@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
The general expectation is that header files should follow the same
file/path naming scheme as the corresponding source file. There are
various historical exceptions to this practice in QEMU, with one of
the most notable being the include/qapi/qmp/ directory.
include/qapi/qmp/dispatch.h corresponds mostly to qapi/qmp-registry.c.
Move and rename it to include/qapi/qmp-registry.h.
Now just qerror.h is left in include/qapi/qmp/. Since it's deprecated
& (slowly) getting eliminated anyway, it isn't worth moving.
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Zhao Liu <zhao1.liu@intel.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-ID: <20241118151235.2665921-3-armbru@redhat.com>
The general expectation is that header files should follow the same
file/path naming scheme as the corresponding source file. There are
various historical exceptions to this practice in QEMU, with one of
the most notable being the include/qapi/qmp/ directory. Most of the
headers there correspond to source files in qobject/.
This patch corrects most of that inconsistency by creating
include/qobject/ and moving the headers for qobject/ there.
This also fixes MAINTAINERS for include/qapi/qmp/dispatch.h:
scripts/get_maintainer.pl now reports "QAPI" instead of "No
maintainers found".
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Zhao Liu <zhao1.liu@intel.com>
Acked-by: Halil Pasic <pasic@linux.ibm.com> #s390x
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-ID: <20241118151235.2665921-2-armbru@redhat.com>
[Rebased]
Allow debugging individual processes in multi-process applications by
starting them with export QEMU_GDB=/tmp/qemu-%d.sock,suspend=n.
Currently one would have to attach to every process to ensure the app
makes progress.
In case suspend=n is not specified, the flow remains unchanged. If it
is specified, then accepting the client connection is delegated to a
thread. In the future this machinery may be reused for handling
reconnections and interruptions.
On accepting a connection, the thread schedules gdb_handlesig() on the
first CPU and wakes it up with host_interrupt_signal. Note that the
result of this gdb_handlesig() invocation is handled, as opposed to
many other existing call sites. These other call sites probably need to
be fixed separately.
Signed-off-by: Ilya Leoshkevich <iii@linux.ibm.com>
Message-Id: <20250117001542.8290-7-iii@linux.ibm.com>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20250207153112.3939799-16-alex.bennee@linaro.org>
Attaching to the gdbstub of a running process requires stopping its
threads. For threads that run on a CPU, cpu_exit() is enough, but the
only way to grab attention of a thread that is stuck in a long-running
syscall is to interrupt it with a signal.
Reserve a host realtime signal for this, just like it's already done
for TARGET_SIGABRT on Linux. This may reduce the number of available
guest realtime signals by one, but this is acceptable, since there are
quite a lot of them, and it's unlikely that there are apps that need
them all.
Set signal_pending for the safe_sycall machinery to prevent invoking
the syscall. This is a lie, since we don't queue a guest signal, but
process_pending_signals() can handle the absence of pending signals.
The syscall returns with QEMU_ERESTARTSYS errno, which arranges for
the automatic restart. This is important, because it helps avoiding
disturbing poorly written guests.
Reviewed-by: Warner Losh <imp@bsdimp.com>
Signed-off-by: Ilya Leoshkevich <iii@linux.ibm.com>
Message-Id: <20250117001542.8290-5-iii@linux.ibm.com>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20250207153112.3939799-14-alex.bennee@linaro.org>
In case an emulated process execve()s another emulated process, bind()
will fail, because the socket already exists. So try deleting it. Use
the existing unix_listen() function which does this. Link qemu-user
with qemu-sockets.c and add the monitor_get_fd() stub.
Note that it is not possible to handle this in do_execv(): deleting
gdbserver_user_state.socket_path before safe_execve() is not correct,
because the latter may fail, and afterwards we may lose control.
Signed-off-by: Ilya Leoshkevich <iii@linux.ibm.com>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20250117001542.8290-3-iii@linux.ibm.com>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20250207153112.3939799-12-alex.bennee@linaro.org>
Add a new command line option to allow selecting between running the
full set of tests or a smaller set of tests. The default will be to
run the small set (i.e. no comand line option provided) so we can
reduce the amount of tests run by default. Only hosts which support
KVM for the target architecture being tested will run the complete set
of tests.
Adjust the meson.build file to pass in the --full option when
appropriate.
(for now, set the option unconditionally until the next patch actually
creates the small set)
Use cases:
configure --target-list=aarch64-softmmu,ppc64-softmmu,s390x-softmmu,x86_64-softmmu
| before - 615s/244 tests | after - 244s/100 tests
------------------------+--------------------------+-----------------------------
make check | full set for all archs | full set for the KVM arch,
make check-qtest | | small set for the rest
| |
qemu-system-$ARCH | full set for $ARCH | small set for $ARCH, KVM or
./migration-test | | TCG automatically chosen
| |
qemu-system-$ARCH | N/A | full set for $ARCH, KVM or
./migration-test --full | | TCG automatically chosen
| |
migration-compat-x86_64 | full set for x86_64 | small set for x86_64
CI job | |
------------------------+--------------------------+-----------------------------
Signed-off-by: Fabiano Rosas <farosas@suse.de>
Reviewed-by: Peter Xu <peterx@redhat.com>
Message-Id: <20250130184012.5711-2-farosas@suse.de>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20250207153112.3939799-9-alex.bennee@linaro.org>
This reverts commit bc02be4508.
Now we catch attempts to clock_step to the next timer when none are
enabled we can revert the previous attempt to prevent deadlock. As
long as a new target time is given we will move time forward even if
no timers will fire. This is desirable for tests which are checking
that nothing changes when things are disabled.
Previously most tests got away with it because --enable-slirp always
has a timer running while the test is active.
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20250207153112.3939799-8-alex.bennee@linaro.org>
qtest_send_prefix never actually sent something over the chardev, all
it does is print the timestamp to the QTEST_LOG when enabled. So
rename the function, make it static, remove the unused CharDev and
simplify all the call sites by handling that directly with
qtest_send (and qtest_log_send).
Reviewed-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20250207153112.3939799-6-alex.bennee@linaro.org>
Cédric Le Goater