A CXL 2.0 component is any entity in the CXL topology. All components
have a analogous function in PCIe. Except for the CXL host bridge, all
have a PCIe config space that is accessible via the common PCIe
mechanisms. CXL components are enumerated via DVSEC fields in the
extended PCIe header space. CXL components will minimally implement some
subset of CXL.mem and CXL.cache registers defined in 8.2.5 of the CXL
2.0 specification. Two headers and a utility library are introduced to
support the minimum functionality needed to enumerate components.
The cxl_pci header manages bits associated with PCI, specifically the
DVSEC and related fields. The cxl_component.h variant has data
structures and APIs that are useful for drivers implementing any of the
CXL 2.0 components. The library takes care of making use of the DVSEC
bits and the CXL.[mem|cache] registers. Per spec, the registers are
little endian.
None of the mechanisms required to enumerate a CXL capable hostbridge
are introduced at this point.
Note that the CXL.mem and CXL.cache registers used are always 4B wide.
It's possible in the future that this constraint will not hold.
Signed-off-by: Ben Widawsky <ben.widawsky@intel.com>
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Adam Manzanares <a.manzanares@samsung.com>
Message-Id: <20220429144110.25167-3-Jonathan.Cameron@huawei.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
A CXL component is a hardware entity that implements CXL component
registers from the CXL 2.0 spec (8.2.3). Currently these represent 3
general types.
1. Host Bridge
2. Ports (root, upstream, downstream)
3. Devices (memory, other)
A CXL component can be conceptually thought of as a PCIe device with
extra functionality when enumerated and enabled. For this reason, CXL
does here, and will continue to add on to existing PCI code paths.
Host bridges will typically need to be handled specially and so they can
implement this newly introduced interface or not. All other components
should implement this interface. Implementing this interface allows the
core PCI code to treat these devices as special where appropriate.
Signed-off-by: Ben Widawsky <ben.widawsky@intel.com>
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Adam Manzanares <a.manzanares@samsung.com>
Message-Id: <20220429144110.25167-2-Jonathan.Cameron@huawei.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
error_setg_errno() expects a normal errno value, not a negated
one, so we should use ENOTSUP instead of -ENOSUP.
Fixes: Coverity CID 1487174
Fixes: ("intel_iommu: support snoop control")
Signed-off-by: Jason Wang <jasowang@redhat.com>
Message-Id: <20220401022824.9337-1-jasowang@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Peter Xu <peterx@redhat.com>
Unlike most virtio features ACCESS_PLATFORM is considered mandatory by
QEMU, i.e. the driver must accept it if offered by the device. The
virtio specification says that the driver SHOULD accept the
ACCESS_PLATFORM feature if offered, and that the device MAY fail to
operate if ACCESS_PLATFORM was offered but not negotiated.
While a SHOULD ain't exactly a MUST, we are certainly allowed to fail
the device when the driver fences ACCESS_PLATFORM. With commit
2943b53f68 ("virtio: force VIRTIO_F_IOMMU_PLATFORM") we already made the
decision to do so whenever the get_dma_as() callback is implemented (by
the bus), which in practice means for the entirety of virtio-pci.
That means, if the device needs to translate I/O addresses, then
ACCESS_PLATFORM is mandatory. The aforementioned commit tells us in the
commit message that this is for security reasons. More precisely if we
were to allow a less then trusted driver (e.g. an user-space driver, or
a nested guest) to make the device bypass the IOMMU by not negotiating
ACCESS_PLATFORM, then the guest kernel would have no ability to
control/police (by programming the IOMMU) what pieces of guest memory
the driver may manipulate using the device. Which would break security
assumptions within the guest.
If ACCESS_PLATFORM is offered not because we want the device to utilize
an IOMMU and do address translation, but because the device does not
have access to the entire guest RAM, and needs the driver to grant
access to the bits it needs access to (e.g. confidential guest support),
we still require the guest to have the corresponding logic and to accept
ACCESS_PLATFORM. If the driver does not accept ACCESS_PLATFORM, then
things are bound to go wrong, and we may see failures much less graceful
than failing the device because the driver didn't negotiate
ACCESS_PLATFORM.
So let us make ACCESS_PLATFORM mandatory for the driver regardless
of whether the get_dma_as() callback is implemented or not.
Signed-off-by: Halil Pasic <pasic@linux.ibm.com>
Fixes: 2943b53f68 ("virtio: force VIRTIO_F_IOMMU_PLATFORM")
Message-Id: <20220307112939.2780117-1-pasic@linux.ibm.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Cornelia Huck <cohuck@redhat.com>
common.rc has some complicated logic to find the common.config that
dates back to xfstests and is completely unnecessary now. Just include
the contents of the file.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20220505094723.732116-1-pbonzini@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
error_report should generally be followed by a failure; if we can proceed
anyway, that is just a warning and should be communicated properly to
the user with warn_report.
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20220511175043.27327-1-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
According to the NBD spec, a server that advertises
NBD_FLAG_CAN_MULTI_CONN promises that multiple client connections will
not see any cache inconsistencies: when properly separated by a single
flush, actions performed by one client will be visible to another
client, regardless of which client did the flush.
We always satisfy these conditions in qemu - even when we support
multiple clients, ALL clients go through a single point of reference
into the block layer, with no local caching. The effect of one client
is instantly visible to the next client. Even if our backend were a
network device, we argue that any multi-path caching effects that
would cause inconsistencies in back-to-back actions not seeing the
effect of previous actions would be a bug in that backend, and not the
fault of caching in qemu. As such, it is safe to unconditionally
advertise CAN_MULTI_CONN for any qemu NBD server situation that
supports parallel clients.
Note, however, that we don't want to advertise CAN_MULTI_CONN when we
know that a second client cannot connect (for historical reasons,
qemu-nbd defaults to a single connection while nbd-server-add and QMP
commands default to unlimited connections; but we already have
existing means to let either style of NBD server creation alter those
defaults). This is visible by no longer advertising MULTI_CONN for
'qemu-nbd -r' without -e, as in the iotest nbd-qemu-allocation.
The harder part of this patch is setting up an iotest to demonstrate
behavior of multiple NBD clients to a single server. It might be
possible with parallel qemu-io processes, but I found it easier to do
in python with the help of libnbd, and help from Nir and Vladimir in
writing the test.
Signed-off-by: Eric Blake <eblake@redhat.com>
Suggested-by: Nir Soffer <nsoffer@redhat.com>
Suggested-by: Vladimir Sementsov-Ogievskiy <v.sementsov-og@mail.ru>
Message-Id: <20220512004924.417153-3-eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The next patch wants to adjust whether the NBD server code advertises
MULTI_CONN based on whether it is known if the server limits to
exactly one client. For a server started by QMP, this information is
obtained through nbd_server_start (which can support more than one
export); but for qemu-nbd (which supports exactly one export), it is
controlled only by the command-line option -e/--shared. Since we
already have a hook function used by qemu-nbd, it's easiest to just
alter its signature to fit our needs.
Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <20220512004924.417153-2-eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Add the reproducer from https://gitlab.com/qemu-project/qemu/-/issues/339
Without the previous commit, when running 'make check-qtest-i386'
with QEMU configured with '--enable-sanitizers' we get:
==4028352==ERROR: AddressSanitizer: heap-buffer-overflow on address 0x619000062a00 at pc 0x5626d03c491a bp 0x7ffdb4199410 sp 0x7ffdb4198bc0
READ of size 786432 at 0x619000062a00 thread T0
#0 0x5626d03c4919 in __asan_memcpy (qemu-system-i386+0x1e65919)
#1 0x5626d1c023cc in flatview_write_continue softmmu/physmem.c:2787:13
#2 0x5626d1bf0c0f in flatview_write softmmu/physmem.c:2822:14
#3 0x5626d1bf0798 in address_space_write softmmu/physmem.c:2914:18
#4 0x5626d1bf0f37 in address_space_rw softmmu/physmem.c:2924:16
#5 0x5626d1bf14c8 in cpu_physical_memory_rw softmmu/physmem.c:2933:5
#6 0x5626d0bd5649 in cpu_physical_memory_write include/exec/cpu-common.h:82:5
#7 0x5626d0bd0a07 in i8257_dma_write_memory hw/dma/i8257.c:452:9
#8 0x5626d09f825d in fdctrl_transfer_handler hw/block/fdc.c:1616:13
#9 0x5626d0a048b4 in fdctrl_start_transfer hw/block/fdc.c:1539:13
#10 0x5626d09f4c3e in fdctrl_write_data hw/block/fdc.c:2266:13
#11 0x5626d09f22f7 in fdctrl_write hw/block/fdc.c:829:9
#12 0x5626d1c20bc5 in portio_write softmmu/ioport.c:207:17
0x619000062a00 is located 0 bytes to the right of 512-byte region [0x619000062800,0x619000062a00)
allocated by thread T0 here:
#0 0x5626d03c66ec in posix_memalign (qemu-system-i386+0x1e676ec)
#1 0x5626d2b988d4 in qemu_try_memalign util/oslib-posix.c:210:11
#2 0x5626d2b98b0c in qemu_memalign util/oslib-posix.c:226:27
#3 0x5626d09fbaf0 in fdctrl_realize_common hw/block/fdc.c:2341:20
#4 0x5626d0a150ed in isabus_fdc_realize hw/block/fdc-isa.c:113:5
#5 0x5626d2367935 in device_set_realized hw/core/qdev.c:531:13
SUMMARY: AddressSanitizer: heap-buffer-overflow (qemu-system-i386+0x1e65919) in __asan_memcpy
Shadow bytes around the buggy address:
0x0c32800044f0: fa fa fa fa fa fa fa fa fa fa fa fa fa fa fa fa
0x0c3280004500: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
0x0c3280004510: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
0x0c3280004520: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
0x0c3280004530: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
=>0x0c3280004540:[fa]fa fa fa fa fa fa fa fa fa fa fa fa fa fa fa
0x0c3280004550: fa fa fa fa fa fa fa fa fa fa fa fa fa fa fa fa
0x0c3280004560: fa fa fa fa fa fa fa fa fa fa fa fa fa fa fa fa
0x0c3280004570: fa fa fa fa fa fa fa fa fa fa fa fa fa fa fa fa
0x0c3280004580: fa fa fa fa fa fa fa fa fa fa fa fa fa fa fa fa
0x0c3280004590: fd fd fd fd fd fd fd fd fd fd fd fd fd fd fd fd
Shadow byte legend (one shadow byte represents 8 application bytes):
Addressable: 00
Heap left redzone: fa
Freed heap region: fd
==4028352==ABORTING
[ kwolf: Added snapshot=on to prevent write file lock failure ]
Reported-by: Alexander Bulekov <alxndr@bu.edu>
Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Alexander Bulekov <alxndr@bu.edu>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Per the 82078 datasheet, if the end-of-track (EOT byte in
the FIFO) is more than the number of sectors per side, the
command is terminated unsuccessfully:
* 5.2.5 DATA TRANSFER TERMINATION
The 82078 supports terminal count explicitly through
the TC pin and implicitly through the underrun/over-
run and end-of-track (EOT) functions. For full sector
transfers, the EOT parameter can define the last
sector to be transferred in a single or multisector
transfer. If the last sector to be transferred is a par-
tial sector, the host can stop transferring the data in
mid-sector, and the 82078 will continue to complete
the sector as if a hardware TC was received. The
only difference between these implicit functions and
TC is that they return "abnormal termination" result
status. Such status indications can be ignored if they
were expected.
* 6.1.3 READ TRACK
This command terminates when the EOT specified
number of sectors have been read. If the 82078
does not find an I D Address Mark on the diskette
after the second· occurrence of a pulse on the
INDX# pin, then it sets the IC code in Status Regis-
ter 0 to "01" (Abnormal termination), sets the MA bit
in Status Register 1 to "1", and terminates the com-
mand.
* 6.1.6 VERIFY
Refer to Table 6-6 and Table 6-7 for information
concerning the values of MT and EC versus SC and
EOT value.
* Table 6·6. Result Phase Table
* Table 6-7. Verify Command Result Phase Table
Fix by aborting the transfer when EOT > # Sectors Per Side.
Cc: qemu-stable@nongnu.org
Cc: Hervé Poussineau <hpoussin@reactos.org>
Fixes: baca51faff ("floppy driver: disk geometry auto detect")
Reported-by: Alexander Bulekov <alxndr@bu.edu>
Resolves: https://gitlab.com/qemu-project/qemu/-/issues/339
Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20211118115733.4038610-2-philmd@redhat.com>
Reviewed-by: Hanna Reitz <hreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Currently png support is dependent on vnc for linking object file to
libpng. This commit makes the parameter independent of vnc as it breaks
system emulator with --disable-vnc unless --disable-png is added.
Fixes: 9a0a119a38 ("Added parameter to take screenshot with screendump as PNG", 2022-04-27)
Signed-off-by: Kshitij Suri <kshitij.suri@nutanix.com>
Message-Id: <20220510161932.228481-1-kshitij.suri@nutanix.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The vsock callbacks .vhost_vsock_set_guest_cid and
.vhost_vsock_set_running are the only ones to be conditional
on #ifdef CONFIG_VHOST_VSOCK. This is different from any other
device-dependent callbacks like .vhost_scsi_set_endpoint, and it
also broke when CONFIG_VHOST_VSOCK was changed to a per-target
symbol.
It would be possible to also use the CONFIG_DEVICES include, but
really there is no reason for most virtio files to be per-target
so just remove the #ifdef to fix the issue.
Reported-by: Dov Murik <dovmurik@linux.ibm.com>
Fixes: 9972ae314f ("build: move vhost-vsock configuration to Kconfig")
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
qemu_co_queue_restart_all is basically the same as qemu_co_enter_all
but without a QemuLockable argument. That's perfectly fine, but only as
long as the function is marked coroutine_fn. If used outside coroutine
context, qemu_co_queue_wait will attempt to take the lock and that
is just broken: if you are calling qemu_co_queue_restart_all outside
coroutine context, the lock is going to be a QemuMutex which cannot be
taken twice by the same thread.
The patch adds the marker to qemu_co_queue_restart_all and to its sole
non-coroutine_fn caller; it then reimplements the function in terms of
qemu_co_enter_all_impl, to remove duplicated code and to clarify that the
latter also works in coroutine context.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20220427130830.150180-4-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Because qemu_co_queue_restart_all does not release the lock, it should
be used only in coroutine context. Introduce a new function that,
like qemu_co_enter_next, does release the lock, and use it whenever
qemu_co_queue_restart_all was used outside coroutine context.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20220427130830.150180-3-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
qemu_co_queue_next is basically the same as qemu_co_enter_next but
without a QemuLockable argument. That's perfectly fine, but only
as long as the function is marked coroutine_fn. If used outside
coroutine context, qemu_co_queue_wait will attempt to take the lock
and that is just broken: if you are calling qemu_co_queue_next outside
coroutine context, the lock is going to be a QemuMutex which cannot be
taken twice by the same thread.
The patch adds the marker and reimplements qemu_co_queue_next in terms of
qemu_co_enter_next_impl, to remove duplicated code and to clarify that the
latter also works in coroutine context.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20220427130830.150180-2-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
slirp 4.7 introduces a new CFI-friendly timer callback that does
not pass function pointers within libslirp as callbacks for timers.
Check the version number and, if it is new enough, allow using CFI
even with a system libslirp.
Reviewed-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
Reviewed-by: Marc-André Lureau <malureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
libslirp 4.7 introduces a CFI-friendly version of the .timer_new callback.
The new callback replaces the function pointer with an enum; invoking the
callback is done with a new function slirp_handle_timer.
Support the new API so that CFI can be made compatible with using a system
libslirp.
Reviewed-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
Reviewed-by: Marc-André Lureau <malureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Replace slirp_init with slirp_new, so that a more recent cfg.version
can be specified. The function appeared in version 4.1.0.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
This struct will be extended in the next few patches to support the
new slirp_handle_timer() call. For that we need to store an additional
"int" for each SLIRP timer, in addition to the cb_opaque.
Reviewed-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
Reviewed-by: Marc-André Lureau <malureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Version 4.7 of slirp provides a new timer API that works better with CFI,
together with several other improvements:
* Allow disabling the internal DHCP server !22
* Support Unix sockets in hostfwd !103
* IPv6 DNS proxying support !110
* bootp: add support for UEFI HTTP boot !111
and bugfixes.
The submodule update also includes 2 commits to fix warnings in the
Win32 build.
Reviewed-by: Marc-André Lureau <malureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
This allows setting memory properties without going through
vl.c, and have them validated just the same.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20220414165300.555321-6-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Handle HostMemoryBackend creation and setting of ms->ram entirely in
machine_run_board_init.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20220414165300.555321-5-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Make -m syntactic sugar for a compound property "-machine
mem.{size,max-size,slots}". The new property does not have
the magic conversion to megabytes of unsuffixed arguments,
and also does not understand that "0" means the default size
(you have to leave it out to get the default). This means
that we need to convert the QemuOpts by hand to a QDict.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20220414165300.555321-4-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Make -boot syntactic sugar for a compound property "-machine boot.{order,menu,...}".
machine_boot_parse is replaced by the setter for the property.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20220414165300.555321-3-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
As part of converting -boot to a property with a QAPI type, define
the struct and use it throughout QEMU to access boot configuration.
machine_boot_parse takes care of doing the QemuOpts->QAPI conversion by
hand, for now.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20220414165300.555321-2-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
When running 'make check' we only get a summary of progress on the
console. Fortunately meson/ninja have saved the raw test output to a
logfile. Exposing this log will make it easier to debug failures that
happen in CI.
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20220509124134.867431-3-berrange@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
When running I/O tests using TAP output mode, we get a single TAP test
with a sub-test reported for each I/O test that is run. The output looks
something like this:
1..123
ok qcow2 011
ok qcow2 012
ok qcow2 013
ok qcow2 217
...
If everything runs or fails normally this is fine, but periodically we
have been seeing the test harness abort early before all 123 tests have
been run, just leaving a fairly useless message like
TAP parsing error: Too few tests run (expected 123, got 107)
we have no idea which tests were running at the time the test harness
abruptly exited. This change causes us to print a message about our
intent to run each test, so we have a record of what is active at the
time the harness exits abnormally.
1..123
# running qcow2 011
ok qcow2 011
# running qcow2 012
ok qcow2 012
# running qcow2 013
ok qcow2 013
# running qcow2 217
ok qcow2 217
...
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20220509124134.867431-2-berrange@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
When stdout is not a terminal, the buffer may not be flushed at each end
of line, so we should flush after each test is done. This is especially
apparent when run by check-block, in two ways:
First, when running make check-block -jX with X > 1, progress indication
was missing, even though testrunner.py does theoretically print each
test's status once it has been run, even in multi-processing mode.
Flushing after each test restores this progress indication.
Second, sometimes make check-block failed altogether, with an error
message that "too few tests [were] run". I presume that's because one
worker process in the job pool did not get to flush its stdout before
the main process exited, and so meson did not get to see that worker's
test results. In any case, by flushing at the end of run_test(), the
problem has disappeared for me.
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
Message-Id: <20220506134215.10086-1-hreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Commit 4c41c69e changed the way the coroutine pool is sized because for
virtio-blk devices with a large queue size and heavy I/O, it was just
too small and caused coroutines to be deleted and reallocated soon
afterwards. The change made the size dynamic based on the number of
queues and the queue size of virtio-blk devices.
There are two important numbers here: Slightly simplified, when a
coroutine terminates, it is generally stored in the global release pool
up to a certain pool size, and if the pool is full, it is freed.
Conversely, when allocating a new coroutine, the coroutines in the
release pool are reused if the pool already has reached a certain
minimum size (the batch size), otherwise we allocate new coroutines.
The problem after commit 4c41c69e is that it not only increases the
maximum pool size (which is the intended effect), but also the batch
size for reusing coroutines (which is a bug). It means that in cases
with many devices and/or a large queue size (which defaults to the
number of vcpus for virtio-blk-pci), many thousand coroutines could be
sitting in the release pool without being reused.
This is not only a waste of memory and allocations, but it actually
makes the QEMU process likely to hit the vm.max_map_count limit on Linux
because each coroutine requires two mappings (its stack and the guard
page for the stack), causing it to abort() in qemu_alloc_stack() because
when the limit is hit, mprotect() starts to fail with ENOMEM.
In order to fix the problem, change the batch size back to 64 to avoid
uselessly accumulating coroutines in the release pool, but keep the
dynamic maximum pool size so that coroutines aren't freed too early
in heavy I/O scenarios.
Note that this fix doesn't strictly make it impossible to hit the limit,
but this would only happen if most of the coroutines are actually in use
at the same time, not just sitting in a pool. This is the same behaviour
as we already had before commit 4c41c69e. Fully preventing this would
require allowing qemu_coroutine_create() to return an error, but it
doesn't seem to be a scenario that people hit in practice.
Cc: qemu-stable@nongnu.org
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=2079938
Fixes: 4c41c69e05
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20220510151020.105528-3-kwolf@redhat.com>
Tested-by: Hiroki Narukawa <hnarukaw@yahoo-corp.jp>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
It's true that these functions currently affect the batch size in which
coroutines are reused (i.e. moved from the global release pool to the
allocation pool of a specific thread), but this is a bug and will be
fixed in a separate patch.
In fact, the comment in the header file already just promises that it
influences the pool size, so reflect this in the name of the functions.
As a nice side effect, the shorter function name makes some line
wrapping unnecessary.
Cc: qemu-stable@nongnu.org
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20220510151020.105528-2-kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Up to now the virt-machine node contains a virtio-mmio node.
However no driver produces any PCI interface node. Hence, PCI
tests cannot be run with aarch64 binary.
Add a GPEX driver node that produces a pci interface node. This latter
then can be consumed by all the pci tests. One of the first motivation
was to be able to run the virtio-iommu-pci tests.
We still face an issue with pci hotplug tests as hotplug cannot happen
on the pcie root bus and require a generic root port. This will be
addressed later on.
We force cpu=max along with aarch64/virt machine as some PCI tests
require high MMIO regions to be available.
Signed-off-by: Eric Auger <eric.auger@redhat.com>
Message-Id: <20220504152025.1785704-4-eric.auger@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
ARM does not not support hotplug on pcie.0. Add a flag on the bus
which tells if devices can be hotplugged and skip hotplug tests
if the bus cannot be hotplugged. This is a temporary solution to
enable the other pci tests on aarch64.
Signed-off-by: Eric Auger <eric.auger@redhat.com>
Acked-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20220504152025.1785704-3-eric.auger@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
At the moment the IO space limit is hardcoded to
QPCI_PIO_LIMIT = 0x10000. When accesses are performed to a bar,
the base address of this latter is compared against the limit
to decide whether we perform an IO or a memory access.
On ARM, we cannot keep this PIO limit as the arm-virt machine
uses [0x3eff0000, 0x3f000000 ] for the IO space map and we
are mandated to allocate at 0x0.
Add a new flag in QPCIBar indicating whether it is an IO bar
or a memory bar. This flag is set on QPCIBar allocation and
provisionned based on the BAR configuration. Then the new flag
is used in access functions and in iomap() function.
Signed-off-by: Eric Auger <eric.auger@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20220504152025.1785704-2-eric.auger@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Now that igd_passthrough_isa_bridge_create() is implemented within the
xen context it may use Xen* data types directly and become
xen_igd_passthrough_isa_bridge_create(). This resolves an indirection.
Signed-off-by: Bernhard Beschow <shentey@gmail.com>
Acked-by: Anthony PERARD <anthony.perard@citrix.com>
Message-Id: <20220326165825.30794-3-shentey@gmail.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
igd-passthrough-isa-bridge is only requested in xen_pt but was
implemented in pc_piix.c. This caused xen_pt to dependend on i386/pc
which is hereby resolved.
Signed-off-by: Bernhard Beschow <shentey@gmail.com>
Acked-by: Anthony PERARD <anthony.perard@citrix.com>
Message-Id: <20220326165825.30794-2-shentey@gmail.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Without this (at least in Fedora 35) it don't detect mremap()
correctly.
Signed-off-by: Juan Quintela <quintela@redhat.com>
Message-Id: <20220502131119.2345-1-quintela@redhat.com>
[Also switch the LEGACY_RDMA_REG_MR test to cc.links, otherwise
Debian fails to build. - Paolo]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Use the string equality operator "eq", and ensure that $1 is defined by
using "(try|)" instead of "(try)?". The alternative "((?:try)?)" is
longer and less readable.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
When cache_info_passthrough is requested, QEMU passes the host values
of the cache information CPUID leaves down to the guest. However,
it blindly assumes that the CPUID leaf exists on the host, and this
cannot be guaranteed: for example, KVM has recently started to
synthesize AMD leaves up to 0x80000021 in order to provide accurate
CPU bug information to guests.
Querying a nonexistent host leaf fills the output arguments of
host_cpuid with data that (albeit deterministic) is nonsensical
as cache information, namely the data in the highest Intel CPUID
leaf. If said highest leaf is not ECX-dependent, this can even
cause an infinite loop when kvm_arch_init_vcpu prepares the input
to KVM_SET_CPUID2. The infinite loop is only terminated by an
abort() when the array gets full.
Reported-by: Maxim Levitsky <mlevitsk@redhat.com>
Reviewed-by: Maxim Levitsky <mlevitsk@redhat.com>
Cc: qemu-stable@nongnu.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Do not rely on the detection that was done in the configure script,
since in the future we may want to cross-compile this file.
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Cleaned up with scripts/clean-header-guards.pl.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20220506134911.2856099-5-armbru@redhat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
We commonly define the header guard symbol without an explicit value.
Normalize the exceptions.
Done with scripts/clean-header-guards.pl.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20220506134911.2856099-4-armbru@redhat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Leading underscores are ill-advised because such identifiers are
reserved. Trailing underscores are merely ugly. Strip both.
Our header guards commonly end in _H. Normalize the exceptions.
Macros should be ALL_CAPS. Normalize the exception.
Done with scripts/clean-header-guards.pl.
include/hw/xen/interface/ and tools/virtiofsd/ left alone, because
these were imported from Xen and libfuse respectively.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20220506134911.2856099-3-armbru@redhat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Header guard symbols should match their file name to make guard
collisions less likely.
Cleaned up with scripts/clean-header-guards.pl, followed by some
renaming of new guard symbols picked by the script to better ones.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20220506134911.2856099-2-armbru@redhat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
[Change to generated file ebpf/rss.bpf.skeleton.h backed out]
- Add new thread-pool-min/thread-pool-max parameters to control the thread pool
used for async I/O.
- Fix virtio-scsi IOThread 100% CPU consumption QEMU 7.0 regression.
-----BEGIN PGP SIGNATURE-----
iQEzBAABCAAdFiEEhpWov9P5fNqsNXdanKSrs4Grc8gFAmJ5DqgACgkQnKSrs4Gr
c8iAqAf/WEJzEso0Hu3UUYJi2lAXpLxWPjoNBlPdQlKIJ/I0zQIF0P7GeCifF+0l
iMjgBv0ofyAuV47gaTJlVrAR75+hJ/IXNDhnu3UuvNWfVOqvksgw6kuHkMo9A2hC
4tIHEU9J8jbQSSdQTaZR8Zj4FX1/zcxMBAXT3YO3De6zo78RatBTuNP4dsZzt8bI
Qs1a4A0p2ScNXK8EcF4QwAWfoxu9OPPzN52DBCNxcIcnn0SUab4NbDxzpRV4ZhDP
08WoafI5O+2Kb36QysJN01LqajHrClG/fozrPzBLq5aZUK3xewJGB1hEdGTLkkmz
NJNBg5Ldszwj4PDZ1dFU3/03aigb3g==
=t5eR
-----END PGP SIGNATURE-----
Merge tag 'block-pull-request' of https://gitlab.com/stefanha/qemu into staging
Pull request
- Add new thread-pool-min/thread-pool-max parameters to control the thread pool
used for async I/O.
- Fix virtio-scsi IOThread 100% CPU consumption QEMU 7.0 regression.
# -----BEGIN PGP SIGNATURE-----
#
# iQEzBAABCAAdFiEEhpWov9P5fNqsNXdanKSrs4Grc8gFAmJ5DqgACgkQnKSrs4Gr
# c8iAqAf/WEJzEso0Hu3UUYJi2lAXpLxWPjoNBlPdQlKIJ/I0zQIF0P7GeCifF+0l
# iMjgBv0ofyAuV47gaTJlVrAR75+hJ/IXNDhnu3UuvNWfVOqvksgw6kuHkMo9A2hC
# 4tIHEU9J8jbQSSdQTaZR8Zj4FX1/zcxMBAXT3YO3De6zo78RatBTuNP4dsZzt8bI
# Qs1a4A0p2ScNXK8EcF4QwAWfoxu9OPPzN52DBCNxcIcnn0SUab4NbDxzpRV4ZhDP
# 08WoafI5O+2Kb36QysJN01LqajHrClG/fozrPzBLq5aZUK3xewJGB1hEdGTLkkmz
# NJNBg5Ldszwj4PDZ1dFU3/03aigb3g==
# =t5eR
# -----END PGP SIGNATURE-----
# gpg: Signature made Mon 09 May 2022 05:52:56 AM PDT
# gpg: using RSA key 8695A8BFD3F97CDAAC35775A9CA4ABB381AB73C8
# gpg: Good signature from "Stefan Hajnoczi <stefanha@redhat.com>" [full]
# gpg: aka "Stefan Hajnoczi <stefanha@gmail.com>" [full]
* tag 'block-pull-request' of https://gitlab.com/stefanha/qemu:
virtio-scsi: move request-related items from .h to .c
virtio-scsi: clean up virtio_scsi_handle_cmd_vq()
virtio-scsi: clean up virtio_scsi_handle_ctrl_vq()
virtio-scsi: clean up virtio_scsi_handle_event_vq()
virtio-scsi: don't waste CPU polling the event virtqueue
virtio-scsi: fix ctrl and event handler functions in dataplane mode
util/event-loop-base: Introduce options to set the thread pool size
util/main-loop: Introduce the main loop into QOM
Introduce event-loop-base abstract class
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
