lkvm by default sets up a virtio-pci transport for network, if none is
specified. This can be a problem on archs (e.g ARM64), where virtio-pci is
not supported yet and cause the following warning at exit.
# KVM compatibility warning.
virtio-net device was not detected.
This patch changes it to make use of the default transport method for the
architecture when none is specified. This will ensure that on every arch
we get the network up by default in the VM.
Signed-off-by: Suzuki K. Poulose <suzuki.poulose@arm.com>
Acked-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Add a utility function that transfers the endianness sampled
at device reset time to a queue being set up.
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Pekka Enberg <penberg@kernel.org>
Save the CPU endianness when the device is reset. It is widely
assumed that the guest won't change its endianness after, or at
least not without reseting the device first.
A default implementation of the endianness sampling just returns
the default "host endianness" value so that unsuspecting architectures
are not affected.
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Pekka Enberg <penberg@kernel.org>
Define a simple infrastructure to configure a virt_queue
depending on the guest endianness, as reported by the feature
flags. At this stage, the endianness is always the host's.
Wrap all accesses to virt_queue data structures shared between
host and guest with byte swapping helpers.
Should the architecture only support one endianness, these helpers
are reduced to the identity function.
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Pekka Enberg <penberg@kernel.org>
Commit "kvm tools: virtio: remove hardcoded assumptions
about guest page size" has introduced a bug that prevented
guests with more than 4gb of ram from booting.
The issue is that 'pfn' is a 32bit integer, so when multiplying
it by page size to get the actual page will cause an overflow if
the pfn referred to a memory area above 4gb.
Acked-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
Signed-off-by: Pekka Enberg <penberg@kernel.org>
Some devices want to know their status, use this hook to allow them to
get that notification.
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
Signed-off-by: Pekka Enberg <penberg@kernel.org>
virtio-based PCI devices deal only with 4k memory granules, making
direct use of the VIRTIO_PCI_VRING_ALIGN and VIRTIO_PCI_QUEUE_ADDR_SHIFT
constants when initialising the virtqueues for a device.
For MMIO-based devices, the guest page size is arbitrary and may differ
from that of the host (this is the case on AArch64, where both 4k and
64k pages are supported).
This patch fixes the virtio drivers to honour the guest page size passed
when configuring the virtio device and align the virtqueues accordingly.
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Pekka Enberg <penberg@kernel.org>
Instead of a get/set for config values, just request the address of the
config region, and handle that by simply reading directly from that region.
Signed-off-by: Sasha Levin <levinsasha928@gmail.com>
Signed-off-by: Pekka Enberg <penberg@kernel.org>
If vhost is enabled for a virtio device, vhost will poll the ioeventfd
in kernel side and there is no need to poll it in userspace. Otherwise,
both vhost kernel and userspace will race to poll.
Signed-off-by: Asias He <asias.hejun@gmail.com>
Signed-off-by: Pekka Enberg <penberg@kernel.org>
This patch introduces a helper virtio_compat_add_message() to simplify
adding compat message for virtio device.
Signed-off-by: Asias He <asias.hejun@gmail.com>
Signed-off-by: Pekka Enberg <penberg@kernel.org>
This patch reworks on virtio transport abstraction.
* Move virtio transport operations to virtio operations and drop
virtio/trans.c This makes the abstraction much cleaner.
* Rename struct virtio_trans to struct virtio_device
struct virtio_trans {
void *virtio;
enum virtio_trans_type type;
struct virtio_trans_ops *trans_ops;
struct virtio_ops *virtio_ops;
};
struct virtio_device {
void *virtio;
struct virtio_ops *ops;
};
The virtio_trans struct is bit confusing since it also includes virtio
operations.
* Introduce virtio_init()
To init device, e.g.
Before:
virtio_trans_init()
ndev->vtrans.trans_ops->init()
ndev->vtrans.virtio_ops = &net_dev_virtio_ops
After:
virtio_init()
Signed-off-by: Asias He <asias.hejun@gmail.com>
Signed-off-by: Pekka Enberg <penberg@kernel.org>
Rusty has just removed it out of the spec. Since we probably the only ones
who implemented support for it, we should remove it out of our code as well.
There is no issue with breaking anything since nothing else worked with it,
so it's fully backwards compatible.
Cc: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Sasha Levin <levinsasha928@gmail.com>
Signed-off-by: Pekka Enberg <penberg@kernel.org>
There are at most bdev->reqs[VIRTIO_BLK_QUEUE_SIZE] outstanding requests
at any time. We can simply use the head of each request to fetch the
right 'struct blk_dev_req' in bdev->reqs[].
So, we can eliminate the list and lock operations which introduced by
virtio_blk_req_{pop, push}.
Signed-off-by: Asias He <asias.hejun@gmail.com>
Signed-off-by: Pekka Enberg <penberg@kernel.org>
This patch is the base for enabling support for event index feature in the virtio spec.
We do so by updating and evaluating the used/avail event idx in the virtio ring functions.
Actual usage of this flag is in the following patches.
The results are less notifications between the guest and host, and in
result faster operation of the virt queues.
Signed-off-by: Sasha Levin <levinsasha928@gmail.com>
Signed-off-by: Pekka Enberg <penberg@kernel.org>
This function is not used anymore. Instead, We are using
virtio_pci__signal_vq() to trigger interrupt right now.
Signed-off-by: Asias He <asias.hejun@gmail.com>
Signed-off-by: Pekka Enberg <penberg@kernel.org>
This patch adds a helper used to retrieve the type of field used when guest
is writing or reading from virtio config space.
Since the config space is dynamic, it may change during runtime - so we must
calculate it before every read/write.
Signed-off-by: Sasha Levin <levinsasha928@gmail.com>
Signed-off-by: Pekka Enberg <penberg@kernel.org>
Instead of redefining virtio pci constants (or not using them at all), use
constants from kernel header.
Acked-and-tested-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Sasha Levin <levinsasha928@gmail.com>
Signed-off-by: Pekka Enberg <penberg@kernel.org>
queue->pfn may be used to point at addresses larger
than 32 bit.
Prevent a wraparound when shifting it left.
Acked-and-tested-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Sasha Levin <levinsasha928@gmail.com>
Signed-off-by: Pekka Enberg <penberg@kernel.org>
Inject IRQ to guest only when ISR status is low which means
guest has read ISR status and device has cleared this bit as
the side effect of this reading.
This reduces a lot of unnecessary IRQ inject from device to
guest.
Netpef test shows this patch changes:
the host to guest bandwidth
from 2866.27 Mbps (cpu 33.96%) to 5548.87 Mbps (cpu 53.87%),
the guest to host bandwitdth
form 1408.86 Mbps (cpu 99.9%) to 1301.29 Mbps (cpu 99.9%).
The bottleneck of the guest to host bandwidth is guest cpu power.
Signed-off-by: Asias He <asias.hejun@gmail.com>
Signed-off-by: Pekka Enberg <penberg@kernel.org>
virtio_console__inject_interrupt tries to use virt queues before guest
tell us to initialize them.
(gdb) r run -i linux-0.2.img -k ./vmlinuz-2.6.38-rc6+ -r ./initrd.img-2.6.38-rc6+ -p=init=1 -m 500 -c
Starting program: /project/rh/kvm-tools/tools/kvm/kvm run -i linux-0.2.img -k ./vmlinuz-2.6.38-rc6+ -r ./initrd.img-2.6.38-rc6+ -p=init=1 -m 500 -c
[Thread debugging using libthread_db enabled]
[New Thread 0x7fffd6e2d700 (LWP 19280)]
Warning: request type 8
Program received signal SIGSEGV, Segmentation fault.
0x00000000004026ca in virt_queue__available (vq=0x60d3c8) at include/kvm/virtio.h:31
31 return vq->vring.avail->idx != vq->last_avail_idx;
(gdb)
(gdb) bt
(gdb) p *vq
$2 = {vring = {num = 0, desc = 0x0, avail = 0x0, used = 0x0}, pfn = 0, last_avail_idx = 0}
include/kvm/virtio-console.h:
59 void virtio_console__inject_interrupt(struct kvm *self)
....
71 if (term_readable(CONSOLE_VIRTIO) && virt_queue__available(vq)) {
72 head = virt_queue__get_iov(vq, iov, &out, &in, self);
^^^^ then this block will not be executed if
virtio_queue is unavaiable.
Changes from v1:
- move the check of virt_queue out of virt_queue__get_iov()
Reported-by: Amos Kong <akong@redhat.com>
Acked-by: Cyrill Gorcunov <gorcunov@gmail.com>
Signed-off-by: Asias He <asias.hejun@gmail.com>
Signed-off-by: Amos Kong <akong@redhat.com>
Signed-off-by: Pekka Enberg <penberg@kernel.org>
This patch moves common virtio code to virtio.c and virtio.h.
Signed-off-by: Asias He <asias.hejun@gmail.com>
Signed-off-by: Pekka Enberg <penberg@kernel.org>
Suzuki K. Poulose