The ioeventfd mechanism is used by vhost, dataplane, and virtio-pci to
turn guest MMIO/PIO writes into eventfd file descriptor events. This
allows arbitrary threads to be notified when the guest writes to a
specific MMIO/PIO address.
qtest and TCG do not support ioeventfd because memory writes are not
checked against registered ioeventfds in QEMU. This patch implements
this in memory_region_dispatch_write() so qtest can use ioeventfd.
Also this patch fixes vhost aborting on some misconfigured old kernels
like 3.18.0 on ARM. It is possible to explicitly enable CONFIG_EVENTFD
in expert settings, while MMIO binding support in KVM will still be
missing.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Pavel Fedin <p.fedin@samsung.com>
Message-Id: <006e01d12377$0b9c2d40$22d487c0$@samsung.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Currently hostmem backend fails if CONFIG_NUMA is enabled in QEMU
(the default) but NUMA is not supported by the kernel. This makes
it impossible to use ivshmem in such configurations.
This patch fixes the problem by ignoring ENOSYS error if policy is set to
MPOL_DEFAULT. This way the code behaves in the same way as if CONFIG_NUMA
was not defined. qemu will still fail if the user specifies some other
policy, so that the user knows it.
Signed-off-by: Pavel Fedin <p.fedin@samsung.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
ARM GICv3 systems with large number of CPUs create lots of IRQ pins. Since
every pin is represented as a property, number of these properties becomes
very large. Every property add first makes sure there's no duplicates.
Traversing the list becomes very slow, therefore QEMU initialization takes
significant time (several seconds for e. g. 16 CPUs).
This patch replaces list with GHashTable, making lookup very fast. The only
drawback is that object_child_foreach() and object_child_foreach_recursive()
cannot add or remove properties during traversal, since GHashTableIter does
not have modify-safe version. However, the code seems not to modify objects
via these functions.
Signed-off-by: Pavel Fedin <p.fedin@samsung.com>
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Tested-by: Pavel Fedin <p.fedin@samsung.com>
[AF: Fixed object_property_del_{all,child}() issues;
g_hash_table_contains() -> g_hash_table_lookup(), suggested by Daniel]
Reviewed-by: Daniel P. Berrange <berrange@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
This allows to explicitly specify file name to use with the backend. This
is important when using it together with ivshmem in order to make it backed
by hugetlbfs. By default filename is autogenerated using mkstemp(), and the
file is unlink()ed after creation, effectively making it anonymous. This is
not very useful with ivshmem because it ends up in a memory which cannot be
accessed by something else.
Distinction between directory and file name is done by stat() check. If an
existing directory is given, the code keeps old behavior. Otherwise it
creates or opens a file with the given pathname.
Signed-off-by: Pavel Fedin <p.fedin@samsung.com>
Tested-by: Igor Skalkin <i.skalkin@samsung.com>
Message-Id: <004301d11166$9672fe30$c358fa90$@samsung.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Currently, if the kernel does not have live migration API, the migration
will still be attempted, but vGIC save/restore functions will just not do
anything. This will result in a broken machine state.
This patch fixes the problem by adding migration blocker if kernel API is
not supported.
Signed-off-by: Pavel Fedin <p.fedin@samsung.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
In-kernel ITS emulation on ARM64 will require to supply requester IDs.
These IDs can now be retrieved from the device pointer using new
pci_requester_id() function.
This patch adds pci_dev pointer to KVM GSI routing functions and makes
callers passing it.
x86 architecture does not use requester IDs, but hw/i386/kvm/pci-assign.c
also made passing PCI device pointer instead of NULL for consistency with
the rest of the code.
Signed-off-by: Pavel Fedin <p.fedin@samsung.com>
Message-Id: <ce081423ba2394a4efc30f30708fca07656bc500.1444916432.git.p.fedin@samsung.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
For GICv3 ITS implementation we are going to use requester IDs in KVM IRQ
routing code. This patch introduces reusable convenient way to obtain this
ID from the device pointer. The new function is now used in some places,
where the same calculation was used.
MemTxAttrs.stream_id also renamed to requester_id in order to better
reflect semantics of the field.
Signed-off-by: Pavel Fedin <p.fedin@samsung.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Message-Id: <5814bcb03a297f198e796b13ed9c35059c52f89b.1444916432.git.p.fedin@samsung.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Add gic_version to VirtMachineState, set it to value of the option
and pass it around where necessary. Instantiate devices and fdt
nodes according to the choice.
max_cpus for virt machine increased to 123 (calculated from redistributor
space available in the memory map). GICv2 compatibility check happens
inside arm_gic_common_realize().
ITS region is added to the memory map too, however currently it not used,
just reserved.
Signed-off-by: Pavel Fedin <p.fedin@samsung.com>
Tested-by: Ashok kumar <ashoks@broadcom.com>
[PMM: Added missing cpu_to_le* calls, thanks to Shannon Zhao]
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Some functions previously used only by vGICv2 are useful also for vGICv3
implementation. Untie them from GICState and make accessible from within
other modules:
- kvm_arm_gic_set_irq()
- kvm_gic_supports_attr() - moved to common code and renamed to
kvm_device_check_attr()
- kvm_gic_access() - turned into GIC-independent kvm_device_access().
Data pointer changed to void * because some GICv3 registers are
64-bit wide
Some of these changes are not used right now, but they will be helpful for
implementing live migration.
Actually kvm_dist_get() and kvm_dist_put() could also be made reusable, but
they would require two extra parameters (s->dev_fd and s->num_cpu) as well as
lots of typecasts of 's' to DeviceState * and back to GICState *. This makes
the code very ugly so i decided to stop at this point. I tried also an
approach with making a base class for all possible GICs, but it would contain
only three variables (dev_fd, cpu_num and irq_num), and accessing them through
the rest of the code would be again tedious (either ugly casts or qemu-style
separate object pointer). So i disliked it too.
Signed-off-by: Pavel Fedin <p.fedin@samsung.com>
Tested-by: Ashok kumar <ashoks@broadcom.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 2ef56d1dd64ffb75ed02a10dcdaf605e5b8ff4f8.1441784344.git.p.fedin@samsung.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Expansion of [*] suffix is very slow because index expansion is done using
trial and error strategy, starting every time from zero and retrying with
the next index until insertion succeeds. With large number of already added
properties this process takes huge amount of time (O(n^2) complexity).
Some architectures (like ARM) use very large amount of IRQ pins in interrupt
controller models. This flaw makes machine startup extremely slow
(~20 seconds for ARM64 with 32 CPUs). This patch decreases this time down to
~10 seconds.
Also in qdev_init_gpio_out_named() memset() is now called only once for the
whole array instead of per-cell cleaning
Signed-off-by: Pavel Fedin <p.fedin@samsung.com>
Reviewed-by: Daniel P. Berrange <berrange@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
This large region is necessary for some devices like ivshmem and video cards
32-bit kernels can be built without LPAE support. In this case such a kernel
will not be able to use PCI controller which has windows in high addresses.
In order to work around the problem, "highmem" option is introduced. It
defaults to on on, but can be manually set to off in order to be able to run
those old 32-bit guests.
Signed-off-by: Pavel Fedin <p.fedin@samsung.com>
Reviewed-by: Alexander Graf <agraf@suse.de>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Shannon Zhao <shannon.zhao@linaro.org>
[PMM: Added missing ULL suffixes and a comment to the a15memmap[] entry]
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
When we're using KVM, the kernel's internal idea of the MPIDR
affinity fields must match the values we tell it for the guest
vcpu cluster configuration in the device tree. Since at the moment
the kernel doesn't support letting userspace tell it the correct
affinity fields to use, we must read the kernel's view and
reflect that back in the device tree.
Signed-off-by: Shlomo Pongratz <shlomo.pongratz@huawei.com>
Signed-off-by: Pavel Fedin <p.fedin@samsung.com>
Message-id: 02f601d0a1e6$90c7d630$b2578290$@samsung.com
[PMM: Use a local #define rather than a global variable for
the TCG ARM_CPUS_PER_CLUSTER setting. Tweak a comment. Update the
commit message.]
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
GICv3 ITS distinguishes between devices by using hardwired device IDs passed on the bus.
This patch implements passing these IDs in qemu.
SMMU is also known to use stream IDs, therefore this addition can also be useful for
implementing platforms with SMMU.
Signed-off-by: Pavel Fedin <p.fedin@samsung.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Changes from v1:
- Added bus number to the stream ID
- Added stream ID not only to MSI-X, but also to plain MSI. Some common code was made into
msi_send_message() function.
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Pavel Fedin