qemu_co_queue_next does not need to release and re-acquire the mutex,
because the queued coroutine does not run immediately. However, this
does not hold for qemu_co_enter_next. Now that qemu_co_queue_wait
can synchronize (via QemuLockable) with code that is not running in
coroutine context, it's important that code using qemu_co_enter_next
can easily use a standardized locking idiom.
First of all, qemu_co_enter_next must use aio_co_wake to restart the
coroutine. Second, the function gains a second argument, a QemuLockable*,
and the comments of qemu_co_queue_next and qemu_co_queue_restart_all
are adjusted to clarify the difference.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20180203153935.8056-5-pbonzini@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Fam Zheng <famz@redhat.com>
There are cases in which a queued coroutine must be restarted from
non-coroutine context (with qemu_co_enter_next). In this cases,
qemu_co_enter_next also needs to be thread-safe, but it cannot use
a CoMutex and so cannot qemu_co_queue_wait. Use QemuLockable so
that the CoQueue can interchangeably use CoMutex or QemuMutex.
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20180203153935.8056-4-pbonzini@redhat.com>
Reviewed-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Fam Zheng <famz@redhat.com>
QemuLockable is a polymorphic lock type that takes an object and
knows which function to use for locking and unlocking. The
implementation could use C11 _Generic, but since the support is
not very widespread I am instead using __builtin_choose_expr and
__builtin_types_compatible_p, which are already used by
include/qemu/atomic.h.
QemuLockable can be used to implement lock guards, or to pass around
a lock in such a way that a function can release it and re-acquire it.
The next patch will do this for CoQueue.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20180203153935.8056-3-pbonzini@redhat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Fam Zheng <famz@redhat.com>
In preparation for adding a similar test using QemuLockable, add a very
simple testcase that has two interleaved calls to lock and unlock.
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20180203153935.8056-2-pbonzini@redhat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Fam Zheng <famz@redhat.com>
Using "fedora:latest" makes behavior different depending on when you
actually pulled the image from the docker repository. In my case,
the supposedly "latest" image was a Fedora 25 download from 8 months
ago, and the new "test-debug" test was failing.
Use "27" to improve reproducibility and make it clear when the image
is obsolete.
Cc: Fam Zheng <famz@redhat.com>
Cc: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <1515755504-21341-1-git-send-email-pbonzini@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Fam Zheng <famz@redhat.com>
The .count of HBitmap is forgot to set in function
hbitmap_deserialize_finish, let's set it to the right value.
Cc: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Cc: Fam Zheng <famz@redhat.com>
Cc: Max Reitz <mreitz@redhat.com>
Cc: John Snow <jsnow@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Signed-off-by: Weiping Zhang <zhangweiping@didichuxing.com>
Signed-off-by: Liang Li <liliangleo@didichuxing.com>
Reviewed-by: John Snow <jsnow@redhat.com>
Message-id: 20180118131308.GA2181@liangdeMacBook-Pro.local
Signed-off-by: John Snow <jsnow@redhat.com>
Implements the WHPX accelerator cpu enlightenments to actually use the whpx-all
accelerator on Windows platforms.
Signed-off-by: Justin Terry (VM) <juterry@microsoft.com>
Message-Id: <1516655269-1785-5-git-send-email-juterry@microsoft.com>
[Register/unregister VCPU thread with RCU. - Paolo]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Implements the Windows Hypervisor Platform accelerator (WHPX) target. Which
acts as a hypervisor accelerator for QEMU on the Windows platform. This enables
QEMU much greater speed over the emulated x86_64 path's that are taken on
Windows today.
1. Adds support for vPartition management.
2. Adds support for vCPU management.
3. Adds support for MMIO/PortIO.
4. Registers the WHPX ACCEL_CLASS.
Signed-off-by: Justin Terry (VM) <juterry@microsoft.com>
Message-Id: <1516655269-1785-4-git-send-email-juterry@microsoft.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Adds support for the Windows Hypervisor Platform accelerator (WHPX) stubs and
introduces the whpx.h sysemu API for managing the vcpu scheduling and
management.
Signed-off-by: Justin Terry (VM) <juterry@microsoft.com>
Message-Id: <1516655269-1785-3-git-send-email-juterry@microsoft.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Introduces the configure support for the new Windows Hypervisor Platform that
allows for hypervisor acceleration from usermode components on the Windows
platform.
Signed-off-by: Justin Terry (VM) <juterry@microsoft.com>
Message-Id: <1516655269-1785-2-git-send-email-juterry@microsoft.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Since we have separate handler on POLLHUP, which drops data
after closing the connection we need to fix this test, because
it sends data and instantly close the socket creating race condition.
In some cases on other end of socket client closes it faster than
reads data. To prevent it I suggest to close socket after recieving.
Signed-off-by: Klim Kireev <klim.kireev@virtuozzo.com>
Message-Id: <20180201134831.17709-1-klim.kireev@virtuozzo.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
This will exercise the memfd memory backend and should generally be
better for testing than memory-backend-file (thanks to anonymous files
and sealing).
If memfd is available, it is preferred.
However, in order to check that file & memfd backends both work
correctly, the read-guest-mem test is checked explicitly for each.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20180201132757.23063-8-marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Let's protect the failing tests under a QTEST_VHOST_USER_FIXME
environment variable, so we keep compiling the tests and we can easily
run them.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20180201132757.23063-6-marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Add a new memory backend, similar to hostmem-file, except that it
doesn't need to create files. It also enforces memory sealing.
This backend is mainly useful for sharing the memory with other
processes.
Note that Linux supports transparent huge-pages of shmem/memfd memory
since 4.8. It is relatively easier to set up THP than a dedicate
hugepage mount point by using "madvise" in
/sys/kernel/mm/transparent_hugepage/shmem_enabled.
Since 4.14, memfd allows to set hugetlb requirement explicitly.
Pending for merge in 4.16 is memfd sealing support for hugetlb backed
memory.
Usage:
-object memory-backend-memfd,id=mem1,size=1G
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20180201132757.23063-5-marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Linux commit 749df87bd7bee5a79cef073f5d032ddb2b211de8 (v4.14-rc1)
added a new flag MFD_HUGETLB to memfd_create() that specify the file
to be created resides in the hugetlbfs filesystem. This is the
generic hugetlbfs filesystem not associated with any specific mount
point.
hugetlbfs does not support sealing operations in v4.14, therefore
specifying MFD_ALLOW_SEALING with MFD_HUGETLB will result in EINVAL.
However, I added sealing support in "[PATCH v3 0/9] memfd: add sealing
to hugetlb-backed memory" series, queued in -mm tree for v4.16.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20180201132757.23063-3-marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
This will allow callers to silence error report when the call is
allowed to failed.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20180201132757.23063-2-marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
If no one joins the thread, its associated memory is leaked.
Reported-by: CheneyLin <linzc@zju.edu.cn>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The effects of ivshmem_enable_irqfd() was not undone on device reset.
This manifested as:
ivshmem_add_kvm_msi_virq: Assertion `!s->msi_vectors[vector].pdev' failed.
when irqfd was enabled before reset and then enabled again after reset, making
ivshmem_enable_irqfd() run for the second time.
To reproduce, run:
ivshmem-server
and QEMU with:
-device ivshmem-doorbell,chardev=iv
-chardev socket,path=/tmp/ivshmem_socket,id=iv
then install the Windows driver, at the time of writing available at:
https://github.com/virtio-win/kvm-guest-drivers-windows/tree/master/ivshmem
and crash-reboot the guest by inducing a BSOD.
Signed-off-by: Ladi Prosek <lprosek@redhat.com>
Message-Id: <20171211072110.9058-5-lprosek@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Adds a rollback path to ivshmem_enable_irqfd() and fixes
ivshmem_disable_irqfd() to bail if irqfd has not been enabled.
To reproduce, run:
ivshmem-server -n 0
and QEMU with:
-device ivshmem-doorbell,chardev=iv
-chardev socket,path=/tmp/ivshmem_socket,id=iv
then load, unload, and load again the Windows driver, at the time of writing
available at:
https://github.com/virtio-win/kvm-guest-drivers-windows/tree/master/ivshmem
The issue is believed to have been masked by other guest drivers, notably
Linux ones, not enabling MSI-X on the device.
Signed-off-by: Ladi Prosek <lprosek@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20171211072110.9058-4-lprosek@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
As of commit 660c97eef6 ("ivshmem: use kvm irqfd for msi notifications"),
QEMU crashes with:
ivshmem: msix_set_vector_notifiers failed
msix_unset_vector_notifiers: Assertion `dev->msix_vector_use_notifier && dev->msix_vector_release_notifier' failed.
if MSI-X is repeatedly enabled and disabled on the ivshmem device, for example
by loading and unloading the Windows ivshmem driver. This is because
msix_unset_vector_notifiers() doesn't call any of the release notifier callbacks
since MSI-X is already disabled at that point (msix_enabled() returning false
is how this transition is detected in the first place). Thus ivshmem_vector_mask()
doesn't run and when MSI-X is subsequently enabled again ivshmem_vector_unmask()
fails.
This is fixed by keeping track of unmasked vectors and making sure that
ivshmem_vector_mask() always runs on MSI-X disable.
Fixes: 660c97eef6 ("ivshmem: use kvm irqfd for msi notifications")
Signed-off-by: Ladi Prosek <lprosek@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20171211072110.9058-3-lprosek@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
As of commit 660c97eef6 ("ivshmem: use kvm irqfd for msi notifications"),
QEMU crashes with:
kvm_irqchip_commit_routes: Assertion `ret == 0' failed.
if the ivshmem device is configured with more vectors than what the server
supports. This is caused by the ivshmem_vector_unmask() being called on
vectors that have not been initialized by ivshmem_add_kvm_msi_virq().
This commit fixes it by adding a simple check to the mask and unmask
callbacks.
Note that the opposite mismatch, if the server supplies more vectors than
what the device is configured for, is already handled and leads to output
like:
Too many eventfd received, device has 1 vectors
To reproduce the assert, run:
ivshmem-server -n 0
and QEMU with:
-device ivshmem-doorbell,chardev=iv
-chardev socket,path=/tmp/ivshmem_socket,id=iv
then load the Windows driver, at the time of writing available at:
https://github.com/virtio-win/kvm-guest-drivers-windows/tree/master/ivshmem
The issue is believed to have been masked by other guest drivers, notably
Linux ones, not enabling MSI-X on the device.
Fixes: 660c97eef6 ("ivshmem: use kvm irqfd for msi notifications")
Signed-off-by: Ladi Prosek <lprosek@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20171211072110.9058-2-lprosek@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The following behavior was observed for QEMU configured by libvirt
to use guest agent as usual for the guests without virtio-serial
driver (Windows or the guest remaining in BIOS stage).
In QEMU on first connect to listen character device socket
the listen socket is removed from poll just after the accept().
virtio_serial_guest_ready() returns 0 and the descriptor
of the connected Unix socket is removed from poll and it will
not be present in poll() until the guest will initialize the driver
and change the state of the serial to "guest connected".
In libvirt connect() to guest agent is performed on restart and
is run under VM state lock. Connect() is blocking and can
wait forever.
In this case libvirt can not perform ANY operation on that VM.
The bug can be easily reproduced this way:
Terminal 1:
qemu-system-x86_64 -m 512 -device pci-serial,chardev=serial1 -chardev socket,id=serial1,path=/tmp/console.sock,server,nowait
(virtio-serial and isa-serial also fit)
Terminal 2:
minicom -D unix\#/tmp/console.sock
(type something and press enter)
C-a x (to exit)
Do 3 times:
minicom -D unix\#/tmp/console.sock
C-a x
It needs 4 connections, because the first one is accepted by QEMU, then two are queued by
the kernel, and the 4th blocks.
The problem is that QEMU doesn't add a read watcher after succesful read
until the guest device wants to acquire recieved data, so
I propose to install a separate pullhup watcher regardless of
whether the device waits for data or not.
Signed-off-by: Klim Kireev <klim.kireev@virtuozzo.com>
Message-Id: <20180125135129.9305-1-klim.kireev@virtuozzo.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
When unregister memory listeners, we should call, e.g.,
region_del() (and possibly other undo operations) on every existing
memory region sections there, otherwise we may leak resources that are
held during the region_add(). This patch undo the stuff for the
listeners, which emulates the case when the address space is set from
current to an empty state.
I found this problem when debugging a refcount leak issue that leads to
a device unplug event lost (please see the "Bug:" line below). In that
case, the leakage of resource is the PCI BAR memory region refcount.
And since memory regions are not keeping their own refcount but onto
their owners, so the vfio-pci device's (who is the owner of the PCI BAR
memory regions) refcount is leaked, and event missing.
We had encountered similar issues before and fixed in other
way (ee4c112846, "vhost: Release memory references on cleanup"). This
patch can be seen as a more high-level fix of similar problems that are
caused by the resource leaks from memory listeners. So now we can remove
the explicit unref of memory regions since that'll be done altogether
during unregistering of listeners now.
Bug: https://bugzilla.redhat.com/show_bug.cgi?id=1531393
Signed-off-by: Peter Xu <peterx@redhat.com>
Message-Id: <20180122060244.29368-5-peterx@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
After next patch, listener unregister will need the container to be
alive. Let's move this unregister phase to be before unset container,
since that operation will free the backend container in kernel,
otherwise we'll get these after next patch:
qemu-system-x86_64: VFIO_UNMAP_DMA: -22
qemu-system-x86_64: vfio_dma_unmap(0x559bf53a4590, 0x0, 0xa0000) = -22 (Invalid argument)
Signed-off-by: Peter Xu <peterx@redhat.com>
Message-Id: <20180122060244.29368-4-peterx@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Acked-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
It's a preparation for follow-up patch to call region_del() in
memory_listener_unregister(), otherwise all device addr attached with
kvm_devices_head will be reset before calling kvm_arm_set_device_addr.
Signed-off-by: Peter Xu <peterx@redhat.com>
Message-Id: <20180122060244.29368-3-peterx@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Trace these operations on two memory listeners. It helps to verify the
new memory listener fix, and good to keep them there.
Signed-off-by: Peter Xu <peterx@redhat.com>
Message-Id: <20180122060244.29368-2-peterx@redhat.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
It helps ASAN to detect more leaks on coroutine stacks, and to get rid
of some extra warnings.
Before:
tests/test-coroutine -p
/basic/lifecycle
/basic/lifecycle: ==20781==WARNING: ASan doesn't fully support
makecontext/swapcontext functions and may produce false positives in
some cases!
==20781==WARNING: ASan is ignoring requested __asan_handle_no_return:
stack top: 0x7ffcb184d000; bottom 0x7ff6c4cfd000; size: 0x0005ecb50000
(25446121472)
False positive error reports may follow
For details see https://github.com/google/sanitizers/issues/189
OK
After:
tests/test-coroutine -p /basic/lifecycle
/basic/lifecycle: ==21110==WARNING: ASan doesn't fully support
makecontext/swapcontext functions and may produce false positives in
some cases!
OK
A similar work would need to be done for sigaltstack & windows fibers
to have similar coverage. Since ucontext is preferred, I didn't bother
checking the other coroutine implementations for now.
Update travis to fix the build with ASAN annotations.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20180116151152.4040-4-marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Typical slowdown introduced by AddressSanitizer is 2x.
UBSan shouldn't have much impact on runtime cost.
Enable it by default when --enable-debug, unless --disable-sanitizers.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20180116151152.4040-3-marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
This is based off Juan's last pull with a few extras, but
also removing:
Add migration xbzrle test
Add migration precopy test
As well as my normal test boxes, I also gave it a test
on a 32 bit ARM box and it seems happy (a Calxeda highbank)
and a big-endian power box.
Dave
-----BEGIN PGP SIGNATURE-----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=ntXb
-----END PGP SIGNATURE-----
Merge remote-tracking branch 'remotes/dgilbert/tags/pull-migration-20180206a' into staging
Migration pull 2018-02-06
This is based off Juan's last pull with a few extras, but
also removing:
Add migration xbzrle test
Add migration precopy test
As well as my normal test boxes, I also gave it a test
on a 32 bit ARM box and it seems happy (a Calxeda highbank)
and a big-endian power box.
Dave
# gpg: Signature made Tue 06 Feb 2018 15:33:31 GMT
# gpg: using RSA key 0516331EBC5BFDE7
# gpg: Good signature from "Dr. David Alan Gilbert (RH2) <dgilbert@redhat.com>"
# Primary key fingerprint: 45F5 C71B 4A0C B7FB 977A 9FA9 0516 331E BC5B FDE7
* remotes/dgilbert/tags/pull-migration-20180206a:
migration: incoming postcopy advise sanity checks
migration: Don't leak IO channels
migration: Recover block devices if failure in device state
tests: Adjust sleeps for migration test
tests: Create migrate-start-postcopy command
tests: Add deprecated commands migration test
tests: Use consistent names for migration
tests: Consolidate accelerators declaration
tests: Remove deprecated migration tests commands
migration: Drop current address parameter from save_zero_page()
migration: use s->threshold_size inside migration_update_counters
migration/savevm.c: set MAX_VM_CMD_PACKAGED_SIZE to 1ul << 32
migration: Route errors down through migration_channel_connect
migration: Allow migrate_fd_connect to take an Error *
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
-----BEGIN PGP SIGNATURE-----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=wKYf
-----END PGP SIGNATURE-----
Merge remote-tracking branch 'remotes/ehabkost/tags/python-next-pull-request' into staging
Python queue, 2018-02-05
# gpg: Signature made Mon 05 Feb 2018 23:07:57 GMT
# gpg: using RSA key 2807936F984DC5A6
# gpg: Good signature from "Eduardo Habkost <ehabkost@redhat.com>"
# Primary key fingerprint: 5A32 2FD5 ABC4 D3DB ACCF D1AA 2807 936F 984D C5A6
* remotes/ehabkost/tags/python-next-pull-request: (21 commits)
docker: change Fedora images to run with python3
travis: improve python version test coverage
ui: update keycodemapdb to get py3 fixes
input: add missing JIS keys to virtio input
qemu.py: don't launch again before shutdown()
qemu.py: cleanup redundant calls in launch()
qemu.py: use poll() instead of 'returncode'
qemu.py: always cleanup on shutdown()
qemu.py: refactor launch()
qemu.py: better control of created files
qemu.py: remove unused import
configure: allow use of python 3
scripts: ensure signrom treats data as bytes
qapi: force a UTF-8 locale for running Python
qapi: ensure stable sort ordering when checking QAPI entities
qapi: remove '-q' arg to diff when comparing QAPI output
qapi: Adapt to moved location of 'maketrans' function in py3
qapi: adapt to moved location of StringIO module in py3
qapi: Use OrderedDict from standard library if available
qapi: use items()/values() intead of iteritems()/itervalues()
...
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
These quirks are necessary for GeForce, but not for Quadro/GRID/Tesla
assignment. Leaving them enabled is fully functional and provides the
most compatibility, but due to the unique NVIDIA MSI ACK behavior[1],
it also introduces latency in re-triggering the MSI interrupt. This
overhead is typically negligible, but has been shown to adversely
affect some (very) high interrupt rate applications. This adds the
vfio-pci device option "x-no-geforce-quirks=" which can be set to
"on" to disable this additional overhead.
A follow-on optimization for GeForce might be to make use of an
ioeventfd to allow KVM to trigger an irqfd in the kernel vfio-pci
driver, avoiding the bounce through userspace to handle this device
write.
[1] Background: the NVIDIA driver has been observed to issue a write
to the MMIO mirror of PCI config space in BAR0 in order to allow the
MSI interrupt for the device to retrigger. Older reports indicated a
write of 0xff to the (read-only) MSI capability ID register, while
more recently a write of 0x0 is observed at config space offset 0x704,
non-architected, extended config space of the device (BAR0 offset
0x88704). Virtualization of this range is only required for GeForce.
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
There is already @hostwin in vfio_listener_region_add() so there is no
point in having the other one.
Fixes: 2e4109de8e ("vfio/spapr: Create DMA window dynamically (SPAPR IOMMU v2)")
Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Add the initialization of the mutex protecting the interrupt list.
Signed-off-by: Eric Auger <eric.auger@redhat.com>
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Recently proposed vfio-pci kernel changes (v4.16) remove the
restriction preventing userspace from mmap'ing PCI BARs in areas
overlapping the MSI-X vector table. This change is primarily intended
to benefit host platforms which make use of system page sizes larger
than the PCI spec recommendation for alignment of MSI-X data
structures (ie. not x86_64). In the case of POWER systems, the SPAPR
spec requires the VM to program MSI-X using hypercalls, rendering the
MSI-X vector table unused in the VM view of the device. However,
ARM64 platforms also support 64KB pages and rely on QEMU emulation of
MSI-X. Regardless of the kernel driver allowing mmaps overlapping
the MSI-X vector table, emulation of the MSI-X vector table also
prevents direct mapping of device MMIO spaces overlapping this page.
Thanks to the fact that PCI devices have a standard self discovery
mechanism, we can try to resolve this by relocating the MSI-X data
structures, either by creating a new PCI BAR or extending an existing
BAR and updating the MSI-X capability for the new location. There's
even a very slim chance that this could benefit devices which do not
adhere to the PCI spec alignment guidelines on x86_64 systems.
This new x-msix-relocation option accepts the following choices:
off: Disable MSI-X relocation, use native device config (default)
auto: Use a known good combination for the platform/device (none yet)
bar0..bar5: Specify the target BAR for MSI-X data structures
If compatible, the target BAR will either be created or extended and
the new portion will be used for MSI-X emulation.
The first obvious user question with this option is how to determine
whether a given platform and device might benefit from this option.
In most cases, the answer is that it won't, especially on x86_64.
Devices often dedicate an entire BAR to MSI-X and therefore no
performance sensitive registers overlap the MSI-X area. Take for
example:
# lspci -vvvs 0a:00.0
0a:00.0 Ethernet controller: Intel Corporation I350 Gigabit Network Connection
...
Region 0: Memory at db680000 (32-bit, non-prefetchable) [size=512K]
Region 3: Memory at db7f8000 (32-bit, non-prefetchable) [size=16K]
...
Capabilities: [70] MSI-X: Enable+ Count=10 Masked-
Vector table: BAR=3 offset=00000000
PBA: BAR=3 offset=00002000
This device uses the 16K bar3 for MSI-X with the vector table at
offset zero and the pending bits arrary at offset 8K, fully honoring
the PCI spec alignment guidance. The data sheet specifically refers
to this as an MSI-X BAR. This device would not see a benefit from
MSI-X relocation regardless of the platform, regardless of the page
size.
However, here's another example:
# lspci -vvvs 02:00.0
02:00.0 Serial Attached SCSI controller: xxxxxxxx
...
Region 0: I/O ports at c000 [size=256]
Region 1: Memory at ef640000 (64-bit, non-prefetchable) [size=64K]
Region 3: Memory at ef600000 (64-bit, non-prefetchable) [size=256K]
...
Capabilities: [c0] MSI-X: Enable+ Count=16 Masked-
Vector table: BAR=1 offset=0000e000
PBA: BAR=1 offset=0000f000
Here the MSI-X data structures are placed on separate 4K pages at the
end of a 64KB BAR. If our host page size is 4K, we're likely fine,
but at 64KB page size, MSI-X emulation at that location prevents the
entire BAR from being directly mapped into the VM address space.
Overlapping performance sensitive registers then starts to be a very
likely scenario on such a platform. At this point, the user could
enable tracing on vfio_region_read and vfio_region_write to determine
more conclusively if device accesses are being trapped through QEMU.
Upon finding a device and platform in need of MSI-X relocation, the
next problem is how to choose target PCI BAR to host the MSI-X data
structures. A few key rules to keep in mind for this selection
include:
* There are only 6 BAR slots, bar0..bar5
* 64-bit BARs occupy two BAR slots, 'lspci -vvv' lists the first slot
* PCI BARs are always a power of 2 in size, extending == doubling
* The maximum size of a 32-bit BAR is 2GB
* MSI-X data structures must reside in an MMIO BAR
Using these rules, we can evaluate each BAR of the second example
device above as follows:
bar0: I/O port BAR, incompatible with MSI-X tables
bar1: BAR could be extended, incurring another 64KB of MMIO
bar2: Unavailable, bar1 is 64-bit, this register is used by bar1
bar3: BAR could be extended, incurring another 256KB of MMIO
bar4: Unavailable, bar3 is 64bit, this register is used by bar3
bar5: Available, empty BAR, minimum additional MMIO
A secondary optimization we might wish to make in relocating MSI-X
is to minimize the additional MMIO required for the device, therefore
we might test the available choices in order of preference as bar5,
bar1, and finally bar3. The original proposal for this feature
included an 'auto' option which would choose bar5 in this case, but
various drivers have been found that make assumptions about the
properties of the "first" BAR or the size of BARs such that there
appears to be no foolproof automatic selection available, requiring
known good combinations to be sourced from users. This patch is
pre-enabled for an 'auto' selection making use of a validated lookup
table, but no entries are yet identified.
Tested-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Reviewed-by: Eric Auger <eric.auger@redhat.com>
Tested-by: Eric Auger <eric.auger@redhat.com>
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Add an option which allows the user to specify a PCI BAR number,
including an 'off' and 'auto' selection.
Cc: Markus Armbruster <armbru@redhat.com>
Cc: Eric Blake <eblake@redhat.com>
Tested-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Reviewed-by: Eric Auger <eric.auger@redhat.com>
Tested-by: Eric Auger <eric.auger@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
The kernel provides similar emulation of PCI BAR register access to
QEMU, so up until now we've used that for things like BAR sizing and
storing the BAR address. However, if we intend to resize BARs or add
BARs that don't exist on the physical device, we need to switch to the
pure QEMU emulation of the BAR.
Tested-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Reviewed-by: Eric Auger <eric.auger@redhat.com>
Tested-by: Eric Auger <eric.auger@redhat.com>
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Add one more layer to our stack of MemoryRegions, this base region
allows us to register BARs independently of the vfio region or to
extend the size of BARs which do map to a region. This will be
useful when we want hypervisor defined BARs or sections of BARs,
for purposes such as relocating MSI-X emulation. We therefore call
msix_init() based on this new base MemoryRegion, while the quirks,
which only modify regions still operate on those sub-MemoryRegions.
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
The fields were removed in the referenced commit, but the comment
still mentions them.
Fixes: 2fb9636ebf ("vfio-pci: Remove unused fields from VFIOMSIXInfo")
Tested-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Reviewed-by: Eric Auger <eric.auger@redhat.com>
Tested-by: Eric Auger <eric.auger@redhat.com>
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>