When dataplane is stopping, the s->vdev->binding->set_host_notifier(...,
false) call can invoke the virtqueue handler if an ioeventfd
notification is pending. This causes hw/virtio-blk.c to invoke
virtio_blk_data_plane_start() before virtio_blk_data_plane_stop()
returns!
The result is that we try to restart dataplane while trying to stop it
and the following assertion is raised:
msix_set_mask_notifier: Assertion `!dev->msix_mask_notifier' failed.
Although the code was intended to prevent this scenario, the s->started
boolean isn't enough. Add s->stopping so that we can postpone clearing
s->started until we've completely stopped dataplane.
This way, virtqueue handler calls during virtio_blk_data_plane_stop()
are ignored. When dataplane is legitimately started again later we
already self-kick ourselves to resume processing.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
virtio_pci_set_guest_notifiers() now takes an additional argument to
specify the number of virtqueues to assign a guest notifier for. This
causes a build breakage for CONFIG_VIRTIO_BLK_DATA_PLANE builds:
/home/mdroth/w/qemu2.git/hw/dataplane/virtio-blk.c: In function
‘virtio_blk_data_plane_start’:
/home/mdroth/w/qemu2.git/hw/dataplane/virtio-blk.c:451:47: error: too
few arguments to function ‘s->vdev->binding->set_guest_notifiers’
/home/mdroth/w/qemu2.git/hw/dataplane/virtio-blk.c: In function
‘virtio_blk_data_plane_stop’:
/home/mdroth/w/qemu2.git/hw/dataplane/virtio-blk.c:511:5: error: too few
arguments to function ‘s->vdev->binding->set_guest_notifiers’
make[1]: *** [hw/dataplane/virtio-blk.o] Error 1
make[1]: *** Waiting for unfinished jobs....
make: *** [subdir-x86_64-softmmu] Error 2
Fix this by passing 1 as the number of virtqueues to assign notifiers
for.
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
O_DIRECT on Linux has alignment requirements on I/O buffers and
misaligned requests result in -EINVAL. The Linux virtio_blk guest
driver usually submits aligned requests so I forgot to handle misaligned
requests.
It turns out that virtio-win guest drivers submit misaligned requests.
Handle them using a bounce buffer that meets alignment requirements.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Extract code for read/write command processing into do_rdwr_cmd(). This
brings together pieces that are spread across process_request().
The real motivation is to set the stage for handling misaligned
requests, which the next patch tackles.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The hw/dataplane/vring.c code includes linux/virtio_ring.h. Ensure that
we use linux-headers/ instead of the system-wide headers, which may be
out-of-date on older distros.
This resolves the following build error on Debian 6:
CC hw/dataplane/vring.o
cc1: warnings being treated as errors
hw/dataplane/vring.c: In function 'vring_enable_notification':
hw/dataplane/vring.c:71: error: implicit declaration of function 'vring_avail_event'
hw/dataplane/vring.c:71: error: nested extern declaration of 'vring_avail_event'
hw/dataplane/vring.c:71: error: lvalue required as left operand of assignment
Note that we now build dataplane/ for each target instead of only once.
There is no way around this since linux-headers/ is only available for
per-target objects - and it's how virtio, vfio, kvm, and friends are
built.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
virtio-blk-data-plane is a subset implementation of virtio-blk. It only
handles read, write, and flush requests. It does this using a dedicated
thread that executes an epoll(2)-based event loop and processes I/O
using Linux AIO.
This approach performs very well but can be used for raw image files
only. The number of IOPS achieved has been reported to be several times
higher than the existing virtio-blk implementation.
Eventually it should be possible to unify virtio-blk-data-plane with the
main body of QEMU code once the block layer and hardware emulation is
able to run outside the global mutex.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
The IOQueue has a pool of iocb structs and a function to add new
read/write requests. Multiple requests can be added before calling the
submit function to actually tell the host kernel to begin I/O. This
allows callers to batch requests and submit them in one go.
The actual I/O is performed using Linux AIO.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Outside the safety of the global mutex we need to poll on file
descriptors. I found epoll(2) is a convenient way to do that, although
other options could replace this module in the future (such as an
AioContext-based loop or glib's GMainLoop).
One important feature of this small event loop implementation is that
the loop can be terminated in a thread-safe way. This allows QEMU to
stop the data plane thread cleanly.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
The virtio-blk-data-plane cannot access memory using the usual QEMU
functions since it executes outside the global mutex and the memory APIs
are this time are not thread-safe.
This patch introduces a virtqueue module based on the kernel's vhost
vring code. The trick is that we map guest memory ahead of time and
access it cheaply outside the global mutex.
Once the hardware emulation code can execute outside the global mutex it
will be possible to drop this code.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
The data plane thread needs to map guest physical addresses to host
pointers. Normally this is done with cpu_physical_memory_map() but the
function assumes the global mutex is held. The data plane thread does
not touch the global mutex and therefore needs a thread-safe memory
mapping mechanism.
Hostmem registers a MemoryListener similar to how vhost collects and
pushes memory region information into the kernel. There is a
fine-grained lock on the regions list which is held during lookup and
when installing a new regions list.
When the physical memory map changes the MemoryListener callbacks are
invoked. They build up a new list of memory regions which is finally
installed when the list has been completed.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
