QEMU must be careful when loading device state off migration streams to
prevent a malicious source from exploiting the emulator. Overdoing these
checks has the side effect of allowing a guest to "pin itself" in cloud
environments by messing with state which is entirely in its control.
Similarly to what f3081539 achieved in usb_device_post_load(), this
commit removes such a check from virtio_load(). Worth noting, the result
of a load without this check is the same as if a guest enables a VQ with
invalid indexes to begin with. That is, the virtual device is set in a
broken state (by the datapath handler) and must be reset.
Signed-off-by: Felipe Franciosi <felipe@nutanix.com>
Message-Id: <20201028134643.110698-1-felipe@nutanix.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Current the 'virtio_set_features' only update the 'MemorRegionCaches'
when the 'virtio_set_features_nocheck' return '0' which means it is
not bad features. However the guest can still trigger the access of the
used vring after set bad features. In this situation it will cause assert
failure in 'ADDRESS_SPACE_ST_CACHED'.
Buglink: https://bugs.launchpad.net/qemu/+bug/1890333
Fixes: db812c4073 ("virtio: update MemoryRegionCaches when guest negotiates features")
Reported-by: Alexander Bulekov <alxndr@bu.edu>
Signed-off-by: Li Qiang <liq3ea@163.com>
Message-Id: <20200919082706.6703-1-liq3ea@163.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Commit 9b3a35ec82 ("virtio: verify that legacy support is not accidentally
on") added a check that returns an error if legacy support is on, but the
device does not support legacy.
Unfortunately some devices were wrongly declared legacy capable even if
they were not (e.g vhost-vsock).
To avoid migration issues, we add a virtio-device property
(x-disable-legacy-check) to skip the legacy error, printing a warning
instead, for machine types < 5.1.
Cc: qemu-stable@nongnu.org
Fixes: 9b3a35ec82 ("virtio: verify that legacy support is not accidentally on")
Suggested-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Suggested-by: Cornelia Huck <cohuck@redhat.com>
Reviewed-by: Cornelia Huck <cohuck@redhat.com>
Signed-off-by: Stefano Garzarella <sgarzare@redhat.com>
Message-Id: <20200921122506.82515-2-sgarzare@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
clang's C11 atomic_fetch_*() functions only take a C11 atomic type
pointer argument. QEMU uses direct types (int, etc) and this causes a
compiler error when a QEMU code calls these functions in a source file
that also included <stdatomic.h> via a system header file:
$ CC=clang CXX=clang++ ./configure ... && make
../util/async.c:79:17: error: address argument to atomic operation must be a pointer to _Atomic type ('unsigned int *' invalid)
Avoid using atomic_*() names in QEMU's atomic.h since that namespace is
used by <stdatomic.h>. Prefix QEMU's APIs with 'q' so that atomic.h
and <stdatomic.h> can co-exist. I checked /usr/include on my machine and
searched GitHub for existing "qatomic_" users but there seem to be none.
This patch was generated using:
$ git grep -h -o '\<atomic\(64\)\?_[a-z0-9_]\+' include/qemu/atomic.h | \
sort -u >/tmp/changed_identifiers
$ for identifier in $(</tmp/changed_identifiers); do
sed -i "s%\<$identifier\>%q$identifier%g" \
$(git grep -I -l "\<$identifier\>")
done
I manually fixed line-wrap issues and misaligned rST tables.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20200923105646.47864-1-stefanha@redhat.com>
In legacy mode, virtio_pci_queue_enabled() falls back to
virtio_queue_enabled() to know if the queue is enabled.
But virtio_queue_enabled() calls again virtio_pci_queue_enabled()
if k->queue_enabled is set. This ends in a crash after a stack
overflow.
The problem can be reproduced with
"-device virtio-net-pci,disable-legacy=off,disable-modern=true
-net tap,vhost=on"
And a look to the backtrace is very explicit:
...
#4 0x000000010029a438 in virtio_queue_enabled ()
#5 0x0000000100497a9c in virtio_pci_queue_enabled ()
...
#130902 0x000000010029a460 in virtio_queue_enabled ()
#130903 0x0000000100497a9c in virtio_pci_queue_enabled ()
#130904 0x000000010029a460 in virtio_queue_enabled ()
#130905 0x0000000100454a20 in vhost_net_start ()
...
This patch fixes the problem by introducing a new function
for the legacy case and calls it from virtio_pci_queue_enabled().
It also calls it from virtio_queue_enabled() to avoid code duplication.
Fixes: f19bcdfedd ("virtio-pci: implement queue_enabled method")
Cc: Jason Wang <jasowang@redhat.com>
Cc: Cindy Lu <lulu@redhat.com>
CC: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Laurent Vivier <lvivier@redhat.com>
Message-Id: <20200727153319.43716-1-lvivier@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Several types of virtio devices had already been around before the
virtio standard was specified. These devices support virtio in legacy
(and transitional) mode.
Devices that have been added in the virtio standard are considered
non-transitional (i.e. with no support for legacy virtio).
Provide a helper function so virtio transports can figure that out
easily.
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
Message-Id: <20200707105446.677966-2-cohuck@redhat.com>
Cc: qemu-stable@nongnu.org
Acked-by: Halil Pasic <pasic@linux.ibm.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
This patch introduces queue_enabled() method which allows the
transport to implement its own way to report whether or not a queue is
enabled.
Signed-off-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Cindy Lu <lulu@redhat.com>
Message-Id: <20200701145538.22333-4-lulu@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Devices may have component devices and buses.
Device realization may fail. Realization is recursive: a device's
realize() method realizes its components, and device_set_realized()
realizes its buses (which should in turn realize the devices on that
bus, except bus_set_realized() doesn't implement that, yet).
When realization of a component or bus fails, we need to roll back:
unrealize everything we realized so far. If any of these unrealizes
failed, the device would be left in an inconsistent state. Must not
happen.
device_set_realized() lets it happen: it ignores errors in the roll
back code starting at label child_realize_fail.
Since realization is recursive, unrealization must be recursive, too.
But how could a partly failed unrealize be rolled back? We'd have to
re-realize, which can fail. This design is fundamentally broken.
device_set_realized() does not roll back at all. Instead, it keeps
unrealizing, ignoring further errors.
It can screw up even for a device with no buses: if the lone
dc->unrealize() fails, it still unregisters vmstate, and calls
listeners' unrealize() callback.
bus_set_realized() does not roll back either. Instead, it stops
unrealizing.
Fortunately, no unrealize method can fail, as we'll see below.
To fix the design error, drop parameter @errp from all the unrealize
methods.
Any unrealize method that uses @errp now needs an update. This leads
us to unrealize() methods that can fail. Merely passing it to another
unrealize method cannot cause failure, though. Here are the ones that
do other things with @errp:
* virtio_serial_device_unrealize()
Fails when qbus_set_hotplug_handler() fails, but still does all the
other work. On failure, the device would stay realized with its
resources completely gone. Oops. Can't happen, because
qbus_set_hotplug_handler() can't actually fail here. Pass
&error_abort to qbus_set_hotplug_handler() instead.
* hw/ppc/spapr_drc.c's unrealize()
Fails when object_property_del() fails, but all the other work is
already done. On failure, the device would stay realized with its
vmstate registration gone. Oops. Can't happen, because
object_property_del() can't actually fail here. Pass &error_abort
to object_property_del() instead.
* spapr_phb_unrealize()
Fails and bails out when remove_drcs() fails, but other work is
already done. On failure, the device would stay realized with some
of its resources gone. Oops. remove_drcs() fails only when
chassis_from_bus()'s object_property_get_uint() fails, and it can't
here. Pass &error_abort to remove_drcs() instead.
Therefore, no unrealize method can fail before this patch.
device_set_realized()'s recursive unrealization via bus uses
object_property_set_bool(). Can't drop @errp there, so pass
&error_abort.
We similarly unrealize with object_property_set_bool() elsewhere,
always ignoring errors. Pass &error_abort instead.
Several unrealize methods no longer handle errors from other unrealize
methods: virtio_9p_device_unrealize(),
virtio_input_device_unrealize(), scsi_qdev_unrealize(), ...
Much of the deleted error handling looks wrong anyway.
One unrealize methods no longer ignore such errors:
usb_ehci_pci_exit().
Several realize methods no longer ignore errors when rolling back:
v9fs_device_realize_common(), pci_qdev_unrealize(),
spapr_phb_realize(), usb_qdev_realize(), vfio_ccw_realize(),
virtio_device_realize().
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20200505152926.18877-17-armbru@redhat.com>
Description copied from Linux kernel commit from Gustavo A. R. Silva
(see [3]):
--v-- description start --v--
The current codebase makes use of the zero-length array language
extension to the C90 standard, but the preferred mechanism to
declare variable-length types such as these ones is a flexible
array member [1], introduced in C99:
struct foo {
int stuff;
struct boo array[];
};
By making use of the mechanism above, we will get a compiler
warning in case the flexible array does not occur last in the
structure, which will help us prevent some kind of undefined
behavior bugs from being unadvertenly introduced [2] to the
Linux codebase from now on.
--^-- description end --^--
Do the similar housekeeping in the QEMU codebase (which uses
C99 since commit 7be41675f7).
All these instances of code were found with the help of the
following Coccinelle script:
@@
identifier s, m, a;
type t, T;
@@
struct s {
...
t m;
- T a[0];
+ T a[];
};
@@
identifier s, m, a;
type t, T;
@@
struct s {
...
t m;
- T a[0];
+ T a[];
} QEMU_PACKED;
[1] https://gcc.gnu.org/onlinedocs/gcc/Zero-Length.html
[2] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=76497732932f
[3] https://git.kernel.org/pub/scm/linux/kernel/git/gustavoars/linux.git/commit/?id=17642a2fbd2c1
Inspired-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The virtqueue code sets up MemoryRegionCaches to access the virtqueue
guest RAM data structures. The code currently assumes that
VRingMemoryRegionCaches is initialized before device emulation code
accesses the virtqueue. An assertion will fail in
vring_get_region_caches() when this is not true. Device fuzzing found a
case where this assumption is false (see below).
Virtqueue guest RAM addresses can also be changed from a vCPU thread
while an IOThread is accessing the virtqueue. This breaks the same
assumption but this time the caches could become invalid partway through
the virtqueue code. The code fetches the caches RCU pointer multiple
times so we will need to validate the pointer every time it is fetched.
Add checks each time we call vring_get_region_caches() and treat invalid
caches as a nop: memory stores are ignored and memory reads return 0.
The fuzz test failure is as follows:
$ qemu -M pc -device virtio-blk-pci,id=drv0,drive=drive0,addr=4.0 \
-drive if=none,id=drive0,file=null-co://,format=raw,auto-read-only=off \
-drive if=none,id=drive1,file=null-co://,file.read-zeroes=on,format=raw \
-display none \
-qtest stdio
endianness
outl 0xcf8 0x80002020
outl 0xcfc 0xe0000000
outl 0xcf8 0x80002004
outw 0xcfc 0x7
write 0xe0000000 0x24 0x00ffffffabffffffabffffffabffffffabffffffabffffffabffffffabffffffabffffffabffffffabffffffabffffffabffffffabffffffab5cffffffabffffffabffffffabffffffabffffffabffffffabffffffabffffffabffffffabffffffabffffffabffffffabffffffabffffffabffffffab0000000001
inb 0x4
writew 0xe000001c 0x1
write 0xe0000014 0x1 0x0d
The following error message is produced:
qemu-system-x86_64: /home/stefanha/qemu/hw/virtio/virtio.c:286: vring_get_region_caches: Assertion `caches != NULL' failed.
The backtrace looks like this:
#0 0x00007ffff5520625 in raise () at /lib64/libc.so.6
#1 0x00007ffff55098d9 in abort () at /lib64/libc.so.6
#2 0x00007ffff55097a9 in _nl_load_domain.cold () at /lib64/libc.so.6
#3 0x00007ffff5518a66 in annobin_assert.c_end () at /lib64/libc.so.6
#4 0x00005555559073da in vring_get_region_caches (vq=<optimized out>) at qemu/hw/virtio/virtio.c:286
#5 vring_get_region_caches (vq=<optimized out>) at qemu/hw/virtio/virtio.c:283
#6 0x000055555590818d in vring_used_flags_set_bit (mask=1, vq=0x5555575ceea0) at qemu/hw/virtio/virtio.c:398
#7 virtio_queue_split_set_notification (enable=0, vq=0x5555575ceea0) at qemu/hw/virtio/virtio.c:398
#8 virtio_queue_set_notification (vq=vq@entry=0x5555575ceea0, enable=enable@entry=0) at qemu/hw/virtio/virtio.c:451
#9 0x0000555555908512 in virtio_queue_set_notification (vq=vq@entry=0x5555575ceea0, enable=enable@entry=0) at qemu/hw/virtio/virtio.c:444
#10 0x00005555558c697a in virtio_blk_handle_vq (s=0x5555575c57e0, vq=0x5555575ceea0) at qemu/hw/block/virtio-blk.c:775
#11 0x0000555555907836 in virtio_queue_notify_aio_vq (vq=0x5555575ceea0) at qemu/hw/virtio/virtio.c:2244
#12 0x0000555555cb5dd7 in aio_dispatch_handlers (ctx=ctx@entry=0x55555671a420) at util/aio-posix.c:429
#13 0x0000555555cb67a8 in aio_dispatch (ctx=0x55555671a420) at util/aio-posix.c:460
#14 0x0000555555cb307e in aio_ctx_dispatch (source=<optimized out>, callback=<optimized out>, user_data=<optimized out>) at util/async.c:260
#15 0x00007ffff7bbc510 in g_main_context_dispatch () at /lib64/libglib-2.0.so.0
#16 0x0000555555cb5848 in glib_pollfds_poll () at util/main-loop.c:219
#17 os_host_main_loop_wait (timeout=<optimized out>) at util/main-loop.c:242
#18 main_loop_wait (nonblocking=<optimized out>) at util/main-loop.c:518
#19 0x00005555559b20c9 in main_loop () at vl.c:1683
#20 0x0000555555838115 in main (argc=<optimized out>, argv=<optimized out>, envp=<optimized out>) at vl.c:4441
Reported-by: Alexander Bulekov <alxndr@bu.edu>
Cc: Michael Tsirkin <mst@redhat.com>
Cc: Cornelia Huck <cohuck@redhat.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: qemu-stable@nongnu.org
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20200207104619.164892-1-stefanha@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
https://bugzilla.redhat.com/show_bug.cgi?id=1708480
Fix leak of region reference that prevents complete
device deletion on hot unplug.
Cc: qemu-stable@nongnu.org
Signed-off-by: Yuri Benditovich <yuri.benditovich@daynix.com>
Message-Id: <20191226043649.14481-2-yuri.benditovich@daynix.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Virtqueue notifications are not necessary during polling, so we disable
them. This allows the guest driver to avoid MMIO vmexits.
Unfortunately the virtio-blk and virtio-scsi handler functions re-enable
notifications, defeating this optimization.
Fix virtio-blk and virtio-scsi emulation so they leave notifications
disabled. The key thing to remember for correctness is that polling
always checks one last time after ending its loop, therefore it's safe
to lose the race when re-enabling notifications at the end of polling.
There is a measurable performance improvement of 5-10% with the null-co
block driver. Real-life storage configurations will see a smaller
improvement because the MMIO vmexit overhead contributes less to
latency.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20191209210957.65087-1-stefanha@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Currently the SLOF firmware for pseries guests will disable/re-enable
a PCI device multiple times via IO/MEM/MASTER bits of PCI_COMMAND
register after the initial probe/feature negotiation, as it tends to
work with a single device at a time at various stages like probing
and running block/network bootloaders without doing a full reset
in-between.
In QEMU, when PCI_COMMAND_MASTER is disabled we disable the
corresponding IOMMU memory region, so DMA accesses (including to vring
fields like idx/flags) will no longer undergo the necessary
translation. Normally we wouldn't expect this to happen since it would
be misbehavior on the driver side to continue driving DMA requests.
However, in the case of pseries, with iommu_platform=on, we trigger the
following sequence when tearing down the virtio-blk dataplane ioeventfd
in response to the guest unsetting PCI_COMMAND_MASTER:
#2 0x0000555555922651 in virtqueue_map_desc (vdev=vdev@entry=0x555556dbcfb0, p_num_sg=p_num_sg@entry=0x7fffe657e1a8, addr=addr@entry=0x7fffe657e240, iov=iov@entry=0x7fffe6580240, max_num_sg=max_num_sg@entry=1024, is_write=is_write@entry=false, pa=0, sz=0)
at /home/mdroth/w/qemu.git/hw/virtio/virtio.c:757
#3 0x0000555555922a89 in virtqueue_pop (vq=vq@entry=0x555556dc8660, sz=sz@entry=184)
at /home/mdroth/w/qemu.git/hw/virtio/virtio.c:950
#4 0x00005555558d3eca in virtio_blk_get_request (vq=0x555556dc8660, s=0x555556dbcfb0)
at /home/mdroth/w/qemu.git/hw/block/virtio-blk.c:255
#5 0x00005555558d3eca in virtio_blk_handle_vq (s=0x555556dbcfb0, vq=0x555556dc8660)
at /home/mdroth/w/qemu.git/hw/block/virtio-blk.c:776
#6 0x000055555591dd66 in virtio_queue_notify_aio_vq (vq=vq@entry=0x555556dc8660)
at /home/mdroth/w/qemu.git/hw/virtio/virtio.c:1550
#7 0x000055555591ecef in virtio_queue_notify_aio_vq (vq=0x555556dc8660)
at /home/mdroth/w/qemu.git/hw/virtio/virtio.c:1546
#8 0x000055555591ecef in virtio_queue_host_notifier_aio_poll (opaque=0x555556dc86c8)
at /home/mdroth/w/qemu.git/hw/virtio/virtio.c:2527
#9 0x0000555555d02164 in run_poll_handlers_once (ctx=ctx@entry=0x55555688bfc0, timeout=timeout@entry=0x7fffe65844a8)
at /home/mdroth/w/qemu.git/util/aio-posix.c:520
#10 0x0000555555d02d1b in try_poll_mode (timeout=0x7fffe65844a8, ctx=0x55555688bfc0)
at /home/mdroth/w/qemu.git/util/aio-posix.c:607
#11 0x0000555555d02d1b in aio_poll (ctx=ctx@entry=0x55555688bfc0, blocking=blocking@entry=true)
at /home/mdroth/w/qemu.git/util/aio-posix.c:639
#12 0x0000555555d0004d in aio_wait_bh_oneshot (ctx=0x55555688bfc0, cb=cb@entry=0x5555558d5130 <virtio_blk_data_plane_stop_bh>, opaque=opaque@entry=0x555556de86f0)
at /home/mdroth/w/qemu.git/util/aio-wait.c:71
#13 0x00005555558d59bf in virtio_blk_data_plane_stop (vdev=<optimized out>)
at /home/mdroth/w/qemu.git/hw/block/dataplane/virtio-blk.c:288
#14 0x0000555555b906a1 in virtio_bus_stop_ioeventfd (bus=bus@entry=0x555556dbcf38)
at /home/mdroth/w/qemu.git/hw/virtio/virtio-bus.c:245
#15 0x0000555555b90dbb in virtio_bus_stop_ioeventfd (bus=bus@entry=0x555556dbcf38)
at /home/mdroth/w/qemu.git/hw/virtio/virtio-bus.c:237
#16 0x0000555555b92a8e in virtio_pci_stop_ioeventfd (proxy=0x555556db4e40)
at /home/mdroth/w/qemu.git/hw/virtio/virtio-pci.c:292
#17 0x0000555555b92a8e in virtio_write_config (pci_dev=0x555556db4e40, address=<optimized out>, val=1048832, len=<optimized out>)
at /home/mdroth/w/qemu.git/hw/virtio/virtio-pci.c:613
I.e. the calling code is only scheduling a one-shot BH for
virtio_blk_data_plane_stop_bh, but somehow we end up trying to process
an additional virtqueue entry before we get there. This is likely due
to the following check in virtio_queue_host_notifier_aio_poll:
static bool virtio_queue_host_notifier_aio_poll(void *opaque)
{
EventNotifier *n = opaque;
VirtQueue *vq = container_of(n, VirtQueue, host_notifier);
bool progress;
if (!vq->vring.desc || virtio_queue_empty(vq)) {
return false;
}
progress = virtio_queue_notify_aio_vq(vq);
namely the call to virtio_queue_empty(). In this case, since no new
requests have actually been issued, shadow_avail_idx == last_avail_idx,
so we actually try to access the vring via vring_avail_idx() to get
the latest non-shadowed idx:
int virtio_queue_empty(VirtQueue *vq)
{
bool empty;
...
if (vq->shadow_avail_idx != vq->last_avail_idx) {
return 0;
}
rcu_read_lock();
empty = vring_avail_idx(vq) == vq->last_avail_idx;
rcu_read_unlock();
return empty;
but since the IOMMU region has been disabled we get a bogus value (0
usually), which causes virtio_queue_empty() to falsely report that
there are entries to be processed, which causes errors such as:
"virtio: zero sized buffers are not allowed"
or
"virtio-blk missing headers"
and puts the device in an error state.
This patch works around the issue by introducing virtio_set_disabled(),
which sets a 'disabled' flag to bypass checks like virtio_queue_empty()
when bus-mastering is disabled. Since we'd check this flag at all the
same sites as vdev->broken, we replace those checks with an inline
function which checks for either vdev->broken or vdev->disabled.
The 'disabled' flag is only migrated when set, which should be fairly
rare, but to maintain migration compatibility we disable it's use for
older machine types. Users requiring the use of the flag in conjunction
with older machine types can set it explicitly as a virtio-device
option.
NOTES:
- This leaves some other oddities in play, like the fact that
DRIVER_OK also gets unset in response to bus-mastering being
disabled, but not restored (however the device seems to continue
working)
- Similarly, we disable the host notifier via
virtio_bus_stop_ioeventfd(), which seems to move the handling out
of virtio-blk dataplane and back into the main IO thread, and it
ends up staying there till a reset (but otherwise continues working
normally)
Cc: David Gibson <david@gibson.dropbear.id.au>,
Cc: Alexey Kardashevskiy <aik@ozlabs.ru>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Message-Id: <20191120005003.27035-1-mdroth@linux.vnet.ibm.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Let's make sure calling this twice is harmless -
no known instances, but seems safer.
Suggested-by: Pan Nengyuan <pannengyuan@huawei.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Devices tend to maintain vq pointers, allow deleting them trough a vq pointer.
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Host notifiers are used in several cases:
1. Traditional ioeventfd where virtqueue notifications are handled in
the main loop thread.
2. IOThreads (aio_handle_output) where virtqueue notifications are
handled in an IOThread AioContext.
3. vhost where virtqueue notifications are handled by kernel vhost or
a vhost-user device backend.
Most virtqueue notifications from the guest use the ioeventfd mechanism,
but there are corner cases where QEMU code calls virtio_queue_notify().
This currently honors the host notifier for the IOThreads
aio_handle_output case, but not for the vhost case. The result is that
vhost does not receive virtqueue notifications from QEMU when
virtio_queue_notify() is called.
This patch extends virtio_queue_notify() to set the host notifier
whenever it is enabled instead of calling the vq->(aio_)handle_output()
function directly. We track the host notifier state for each virtqueue
separately since some devices may use it only for certain virtqueues.
This fixes the vhost case although it does add a trip through the
eventfd for the traditional ioeventfd case. I don't think it's worth
adding a fast path for the traditional ioeventfd case because calling
virtio_queue_notify() is rare when ioeventfd is enabled.
Reported-by: Felipe Franciosi <felipe@nutanix.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20191105140946.165584-1-stefanha@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Use RCU_READ_LOCK_GUARD and WITH_RCU_READ_LOCK_GUARD
to replace the manual rcu_read_(un)lock calls.
I think the only change is virtio_load which was missing unlocks
in error paths; those end up being fatal errors so it's not
that important anyway.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Message-Id: <20191028161109.60205-1-dgilbert@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Version: GnuPG v1
iQEcBAABAgAGBQJdt6UAAAoJEO8Ells5jWIRdaAH/3+dufJuFijZY44VYbob92ud
lZR1dYah1fBL1bq0F2siFUb+/wgF1IXHJl9tuUJe8Kp0+hnsXji2s4Iuq5lNQoJj
wwMGziL1TPkhxwgy4jObIC+/bqZVrzAO4Cd+PARrSGAAbAqjxLPizOaf72/t4kdn
C2n87ZlR5k0EOPmUY6Y2DtHtrJ20usSS6EThGhdW7iPSzfQSGiOdRzfZrSiEV2XT
cuKbSzQxk7pbPcz4jIgLzaoA7FIXwm99dBosUkjPszNNFbO4+OPDNdUBanYuqmn/
0ZPe/9YZpEMV64ps/Ab7lx7YB04wZ+A9Etln2JULhBWXg/oyri9gsqgOc6bfCXg=
=uE5S
-----END PGP SIGNATURE-----
Merge remote-tracking branch 'remotes/jasowang/tags/net-pull-request' into staging
# gpg: Signature made Tue 29 Oct 2019 02:33:36 GMT
# gpg: using RSA key EF04965B398D6211
# gpg: Good signature from "Jason Wang (Jason Wang on RedHat) <jasowang@redhat.com>" [marginal]
# gpg: WARNING: This key is not certified with sufficiently trusted signatures!
# gpg: It is not certain that the signature belongs to the owner.
# Primary key fingerprint: 215D 46F4 8246 689E C77F 3562 EF04 965B 398D 6211
* remotes/jasowang/tags/net-pull-request:
COLO-compare: Fix incorrect `if` logic
virtio-net: prevent offloads reset on migration
virtio: new post_load hook
net: add tulip (dec21143) driver
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Post load hook in virtio vmsd is called early while device is processed,
and when VirtIODevice core isn't fully initialized. Most device
specific code isn't ready to deal with a device in such state, and
behaves weirdly.
Add a new post_load hook in a device class instead. Devices should use
this unless they specifically want to verify the migration stream as
it's processed, e.g. for bounds checking.
Cc: qemu-stable@nongnu.org
Suggested-by: "Dr. David Alan Gilbert" <dgilbert@redhat.com>
Cc: Mikhail Sennikovsky <mikhail.sennikovskii@cloud.ionos.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
virtio_device_stop_ioeventfd() has not been used since commit
310837de6c ("virtio: introduce
grab/release_ioeventfd to fix vhost") in 2016.
Nowadays ioeventfd is stopped implicitly by the virtio transport when
lifecycle events such as the VM pausing or device unplug occur.
Cc: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20191021150343.30742-1-stefanha@redhat.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
This patch implements event suppression through device/driver
area. Please refer virtio specification for more information.
Signed-off-by: Wei Xu <wexu@redhat.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
Message-Id: <20191025083527.30803-7-eperezma@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
This patch implements basic support for the packed virtqueue. Compare
the split virtqueue which has three rings, packed virtqueue only have
one which is supposed to have better cache utilization and more
hardware friendly.
Please refer virtio specification for more information.
Signed-off-by: Wei Xu <wexu@redhat.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
Message-Id: <20191025083527.30803-6-eperezma@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
There is slight size difference between split/packed rings.
This is the refactor of split ring as well as a helper to expanding
device and driver area size calculation for packed ring.
Signed-off-by: Wei Xu <wexu@redhat.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
Reviewed-by: Jens Freimann <jfreimann@redhat.com>
Message-Id: <20191025083527.30803-3-eperezma@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Define packed ring structure according to Qemu nomenclature,
field data(wrap counter, etc) are also included.
Signed-off-by: Wei Xu <wexu@redhat.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
Reviewed-by: Jens Freimann <jfreimann@redhat.com>
Message-Id: <20191025083527.30803-2-eperezma@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
sysemu/sysemu.h is a rather unfocused dumping ground for stuff related
to the system-emulator. Evidence:
* It's included widely: in my "build everything" tree, changing
sysemu/sysemu.h still triggers a recompile of some 1100 out of 6600
objects (not counting tests and objects that don't depend on
qemu/osdep.h, down from 5400 due to the previous two commits).
* It pulls in more than a dozen additional headers.
Split stuff related to run state management into its own header
sysemu/runstate.h.
Touching sysemu/sysemu.h now recompiles some 850 objects. qemu/uuid.h
also drops from 1100 to 850, and qapi/qapi-types-run-state.h from 4400
to 4200. Touching new sysemu/runstate.h recompiles some 500 objects.
Since I'm touching MAINTAINERS to add sysemu/runstate.h anyway, also
add qemu/main-loop.h.
Suggested-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20190812052359.30071-30-armbru@redhat.com>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
[Unbreak OS-X build]
In my "build everything" tree, changing sysemu/sysemu.h triggers a
recompile of some 1800 out of 6600 objects (not counting tests and
objects that don't depend on qemu/osdep.h, down from 5400 due to the
previous commit).
Several headers include sysemu/sysemu.h just to get typedef
VMChangeStateEntry. Move it from sysemu/sysemu.h to qemu/typedefs.h.
Spell its structure tag the same while there. Drop the now
superfluous includes of sysemu/sysemu.h from headers.
Touching sysemu/sysemu.h now recompiles some 1100 objects.
qemu/uuid.h also drops from 1800 to 1100, and
qapi/qapi-types-run-state.h from 5000 to 4400.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20190812052359.30071-29-armbru@redhat.com>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Tested-by: Philippe Mathieu-Daudé <philmd@redhat.com>
In my "build everything" tree, changing hw/qdev-properties.h triggers
a recompile of some 2700 out of 6600 objects (not counting tests and
objects that don't depend on qemu/osdep.h).
Many places including hw/qdev-properties.h (directly or via hw/qdev.h)
actually need only hw/qdev-core.h. Include hw/qdev-core.h there
instead.
hw/qdev.h is actually pointless: all it does is include hw/qdev-core.h
and hw/qdev-properties.h, which in turn includes hw/qdev-core.h.
Replace the remaining uses of hw/qdev.h by hw/qdev-properties.h.
While there, delete a few superfluous inclusions of hw/qdev-core.h.
Touching hw/qdev-properties.h now recompiles some 1200 objects.
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Daniel P. Berrangé" <berrange@redhat.com>
Cc: Eduardo Habkost <ehabkost@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Message-Id: <20190812052359.30071-22-armbru@redhat.com>
In my "build everything" tree, changing qemu/main-loop.h triggers a
recompile of some 5600 out of 6600 objects (not counting tests and
objects that don't depend on qemu/osdep.h). It includes block/aio.h,
which in turn includes qemu/event_notifier.h, qemu/notify.h,
qemu/processor.h, qemu/qsp.h, qemu/queue.h, qemu/thread-posix.h,
qemu/thread.h, qemu/timer.h, and a few more.
Include qemu/main-loop.h only where it's needed. Touching it now
recompiles only some 1700 objects. For block/aio.h and
qemu/event_notifier.h, these numbers drop from 5600 to 2800. For the
others, they shrink only slightly.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20190812052359.30071-21-armbru@redhat.com>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Tested-by: Philippe Mathieu-Daudé <philmd@redhat.com>
In my "build everything" tree, changing migration/qemu-file-types.h
triggers a recompile of some 2600 out of 6600 objects (not counting
tests and objects that don't depend on qemu/osdep.h).
The culprit is again hw/hw.h, which supposedly includes it for
convenience.
Include migration/qemu-file-types.h only where it's needed. Touching
it now recompiles less than 200 objects.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20190812052359.30071-10-armbru@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Tested-by: Philippe Mathieu-Daudé <philmd@redhat.com>
When the 'cont' command resumes guest execution the vm change state
handlers are invoked. Unfortunately there is no explicit ordering
between classic qemu_add_vm_change_state_handler() callbacks. When two
layers of code both use vm change state handlers, we don't control which
handler runs first.
virtio-scsi with iothreads hits a deadlock when a failed SCSI command is
restarted and completes before the iothread is re-initialized.
This patch uses the new qdev_add_vm_change_state_handler() API to
guarantee that virtio-scsi's virtio change state handler executes before
the SCSI bus children. This way DMA is restarted after the iothread has
re-initialized.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
We will call virtio_set_status() on virtio_vmstate_change().
The "started" flag should not be changed in this case. Otherwise,
we may get an incorrect value when we set "started" flag but
not set DRIVER_OK in source VM.
Signed-off-by: Xie Yongji <xieyongji@baidu.com>
Message-Id: <20190626023130.31315-6-xieyongji@baidu.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
We should set the flags: "start_on_kick" and "started" after we call
the kick functions (handle_aio_output() and handle_output()).
Signed-off-by: Xie Yongji <xieyongji@baidu.com>
Message-Id: <20190626023130.31315-5-xieyongji@baidu.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
The guest feature is not set correctly on virtio_reset() and
virtio_init(). So we should not use it to set "start_on_kick" at that
point. This patch set "start_on_kick" on virtio_set_features() instead.
Fixes: badaf79cfd ("virtio: Introduce started flag to VirtioDevice")
Signed-off-by: Xie Yongji <xieyongji@baidu.com>
Reviewed-by: Greg Kurz <groug@kaod.org>
Message-Id: <20190626023130.31315-4-xieyongji@baidu.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Besides virtio 1.0 transitional devices, we should also
set "start_on_kick" flag for legacy devices (virtio 0.9).
Signed-off-by: Xie Yongji <xieyongji@baidu.com>
Reviewed-by: Greg Kurz <groug@kaod.org>
Message-Id: <20190626023130.31315-3-xieyongji@baidu.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
In order to avoid migration issues, we introduce a "use-started"
property to the base virtio device to indicate whether use
"started" flag or not. This property will be true by default and
set to false when machine type <= 4.0.
Suggested-by: Greg Kurz <groug@kaod.org>
Signed-off-by: Xie Yongji <xieyongji@baidu.com>
Message-Id: <20190626023130.31315-2-xieyongji@baidu.com>
Reviewed-by: Greg Kurz <groug@kaod.org>
Tested-by: Greg Kurz <groug@kaod.org>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
As explained in commit aff39be0ed97:
Both functions, object_initialize() and object_property_add_child()
increase the reference counter of the new object, so one of the
references has to be dropped afterwards to get the reference
counting right. Otherwise the child object will not be properly
cleaned up when the parent gets destroyed.
Thus let's use now object_initialize_child() instead to get the
reference counting here right.
This patch was generated using the following Coccinelle script:
@use_object_initialize_child@
expression parent_obj;
expression child_ptr;
expression child_name;
expression child_type;
expression child_size;
expression errp;
@@
(
- object_initialize(child_ptr, child_size, child_type);
+ object_initialize_child(parent_obj, child_name, child_ptr, child_size,
+ child_type, &error_abort, NULL);
... when != parent_obj
- object_property_add_child(parent_obj, child_name, OBJECT(child_ptr), NULL);
...
?- object_unref(OBJECT(child_ptr));
|
- object_initialize(child_ptr, child_size, child_type);
+ object_initialize_child(parent_obj, child_name, child_ptr, child_size,
+ child_type, errp, NULL);
... when != parent_obj
- object_property_add_child(parent_obj, child_name, OBJECT(child_ptr), errp);
...
?- object_unref(OBJECT(child_ptr));
)
While the object_initialize() function doesn't take an
'Error *errp' argument, the object_initialize_child() does.
Since this code is used when a machine is created (and is not
yet running), we deliberately choose to use the &error_abort
argument instead of ignoring errors if an object creation failed.
Suggested-by: Eduardo Habkost <ehabkost@redhat.com>
Inspired-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20190507163416.24647-4-philmd@redhat.com>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Currently, we use DRIVER_OK status bit to check whether guest
driver has started the device in virtio_vmstate_change(). But it's
not the case for virtio 1.0 transitional devices. If migration completes
between kicking virtqueue and setting VIRTIO_CONFIG_S_DRIVER_OK, guest
may be hung. So here we use started flag to check guest state instead.
Signed-off-by: Xie Yongji <xieyongji@baidu.com>
Signed-off-by: Zhang Yu <zhangyu31@baidu.com>
Message-Id: <20190320112646.3712-3-xieyongji@baidu.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
The virtio 1.0 transitional devices support driver uses the device
before setting the DRIVER_OK status bit. So we introduce a started
flag to indicate whether driver has started the device or not.
Signed-off-by: Xie Yongji <xieyongji@baidu.com>
Signed-off-by: Zhang Yu <zhangyu31@baidu.com>
Message-Id: <20190320112646.3712-2-xieyongji@baidu.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
We used to set backend unconditionally, this won't work for some
guests (e.g windows driver) who may not initialize all virtqueues. For
kernel backend, this will fail since it may try to validate the rings
during setting backend.
Fixing this by simply skipping the backend set when we find desc is
not ready.
Reviewed-by: Michael S. Tsirkin<mst@redhat.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
In order to use VirtIOFeature also in other virtio devices, we move
its declaration and the endof() macro (renamed in virtio_endof())
in virtio.h.
We add virtio_feature_get_config_size() function to iterate the array
of VirtIOFeature and to return the config size depending on the
features enabled. (as virtio_net_set_config_size() did)
Suggested-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Stefano Garzarella <sgarzare@redhat.com>
Message-id: 20190221103314.58500-5-sgarzare@redhat.com
Message-Id: <20190221103314.58500-5-sgarzare@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
The virtqueue_pop() and virtqueue_get_avail_bytes() routines can use the
INDIRECT table to get the data. It is possible to create a packet which
will lead to the assert message like:
include/exec/memory.h:1995: void
address_space_read_cached(MemoryRegionCache *, hwaddr, void *, int):
Assertion `addr < cache->len && len <= cache->len - addr' failed.
Aborted
To do it the first descriptor should have a link to the INDIRECT table
and set the size of it to 0. It doesn't look good that the guest should
be able to trigger the assert in qemu. Add additional check for the size
of the INDIRECT table, which should not be 0.
Signed-off-by: Dima Stepanov <dimastep@yandex-team.ru>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Cornelia Huck <cohuck@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Pass num_sg by value instead of by pointer, as num_sg is never modified in
virtqueue_map_iovec().
Signed-off-by: Dongli Zhang <dongli.zhang@oracle.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <1541139396-4727-1-git-send-email-dongli.zhang@oracle.com>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
Because they are supposed to remain const.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20181114132931.22624-1-marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Before, we did not clear callback like handle_output when delete
the virtqueue which may result be segmentfault.
The scene is as follows:
1. Start a vm with multiqueue vhost-net,
2. then we write VIRTIO_PCI_GUEST_FEATURES in PCI configuration to
triger multiqueue disable in this vm which will delete the virtqueue.
In this step, the tx_bh is deleted but the callback virtio_net_handle_tx_bh
still exist.
3. Finally, we write VIRTIO_PCI_QUEUE_NOTIFY in PCI configuration to
notify the deleted virtqueue. In this way, virtio_net_handle_tx_bh
will be called and qemu will be crashed.
Although the way described above is uncommon, we had better reinforce it.
CC: qemu-stable@nongnu.org
Signed-off-by: liujunjie <liujunjie23@huawei.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
We've got three places already that provide a prototype for this
function in a .c file - that's ugly. Let's provide a proper prototype
in a header instead, with a proper description why this function should
not be used in most cases.
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
