With REPLY_NEEDED, libvhost-user sends both the acutal result and an
additional ACK reply for VHOST_USER_ADD_MEM_REG. This is incorrect, the
spec mandates that it behave the same with and without REPLY_NEEDED
because it always sends a reply.
Fixes: ec94c8e621
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20220627134500.94842-4-kwolf@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
With REPLY_NEEDED, libvhost-user sends both the acutal result and an
additional ACK reply for VHOST_USER_GET_MAX_MEM_SLOTS. This is
incorrect, the spec mandates that it behave the same with and without
REPLY_NEEDED because it always sends a reply.
Fixes: 6fb2e173d2
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20220627134500.94842-3-kwolf@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
The specification for VHOST_USER_ADD/REM_MEM_REG messages is unclear
in several points, which has led to clients having incompatible
implementations. This changes the specification to be more explicit
about them:
* VHOST_USER_ADD_MEM_REG is not specified as receiving a file
descriptor, though it obviously does need to do so. All
implementations agree on this one, fix the specification.
* VHOST_USER_REM_MEM_REG is not specified as receiving a file
descriptor either, and it also has no reason to do so. rust-vmm does
not send file descriptors for removing a memory region (in agreement
with the specification), libvhost-user and QEMU do (which is a bug),
though libvhost-user doesn't actually make any use of it.
Change the specification so that for compatibility QEMU's behaviour
becomes legal, even if discouraged, but rust-vmm's behaviour becomes
the explicitly recommended mode of operation.
* VHOST_USER_ADD_MEM_REG doesn't have a documented return value, which
is the desired behaviour in the non-postcopy case. It also implemented
like this in QEMU and rust-vmm, though libvhost-user is buggy and
sometimes sends an unexpected reply. This will be fixed in a separate
patch.
However, in postcopy mode it does reply like VHOST_USER_SET_MEM_TABLE.
This behaviour is shared between libvhost-user and QEMU; rust-vmm
doesn't implement postcopy mode yet. Mention it explicitly in the
spec.
* The specification doesn't mention how VHOST_USER_REM_MEM_REG
identifies the memory region to be removed. Change it to describe the
existing behaviour of libvhost-user (guest address, user address and
size must match).
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20220407133657.155281-2-kwolf@redhat.com>
Reviewed-by: Raphael Norwitz <raphael.norwitz@nutanix.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Commit 4c41c69e changed the way the coroutine pool is sized because for
virtio-blk devices with a large queue size and heavy I/O, it was just
too small and caused coroutines to be deleted and reallocated soon
afterwards. The change made the size dynamic based on the number of
queues and the queue size of virtio-blk devices.
There are two important numbers here: Slightly simplified, when a
coroutine terminates, it is generally stored in the global release pool
up to a certain pool size, and if the pool is full, it is freed.
Conversely, when allocating a new coroutine, the coroutines in the
release pool are reused if the pool already has reached a certain
minimum size (the batch size), otherwise we allocate new coroutines.
The problem after commit 4c41c69e is that it not only increases the
maximum pool size (which is the intended effect), but also the batch
size for reusing coroutines (which is a bug). It means that in cases
with many devices and/or a large queue size (which defaults to the
number of vcpus for virtio-blk-pci), many thousand coroutines could be
sitting in the release pool without being reused.
This is not only a waste of memory and allocations, but it actually
makes the QEMU process likely to hit the vm.max_map_count limit on Linux
because each coroutine requires two mappings (its stack and the guard
page for the stack), causing it to abort() in qemu_alloc_stack() because
when the limit is hit, mprotect() starts to fail with ENOMEM.
In order to fix the problem, change the batch size back to 64 to avoid
uselessly accumulating coroutines in the release pool, but keep the
dynamic maximum pool size so that coroutines aren't freed too early
in heavy I/O scenarios.
Note that this fix doesn't strictly make it impossible to hit the limit,
but this would only happen if most of the coroutines are actually in use
at the same time, not just sitting in a pool. This is the same behaviour
as we already had before commit 4c41c69e. Fully preventing this would
require allowing qemu_coroutine_create() to return an error, but it
doesn't seem to be a scenario that people hit in practice.
Cc: qemu-stable@nongnu.org
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=2079938
Fixes: 4c41c69e05
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20220510151020.105528-3-kwolf@redhat.com>
Tested-by: Hiroki Narukawa <hnarukaw@yahoo-corp.jp>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
It's true that these functions currently affect the batch size in which
coroutines are reused (i.e. moved from the global release pool to the
allocation pool of a specific thread), but this is a bug and will be
fixed in a separate patch.
In fact, the comment in the header file already just promises that it
influences the pool size, so reflect this in the name of the functions.
As a nice side effect, the shorter function name makes some line
wrapping unnecessary.
Cc: qemu-stable@nongnu.org
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20220510151020.105528-2-kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The spec clarifies now that QEMU should not send a file descriptor in a
request to remove a memory region. Change it accordingly.
For libvhost-user, this is a bug fix that makes it compatible with
rust-vmm's implementation that doesn't send a file descriptor. Keep
accepting, but ignoring a file descriptor for compatibility with older
QEMU versions.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20220407133657.155281-4-kwolf@redhat.com>
Reviewed-by: Raphael Norwitz <raphael.norwitz@nutanix.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Outside of postcopy mode, neither VHOST_USER_ADD_MEM_REG nor
VHOST_USER_REM_MEM_REG are supposed to send a reply unless explicitly
requested with the need_reply flag. Their current implementation always
sends a reply, even if it isn't requested. This confuses the master
because it will interpret the reply as a reply for the next message for
which it actually expects a reply.
need_reply is already handled correctly by vu_dispatch(), so just don't
send a reply in the non-postcopy part of the message handler for these
two commands.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20220407133657.155281-3-kwolf@redhat.com>
Reviewed-by: Raphael Norwitz <raphael.norwitz@nutanix.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The specification for VHOST_USER_ADD/REM_MEM_REG messages is unclear
in several points, which has led to clients having incompatible
implementations. This changes the specification to be more explicit
about them:
* VHOST_USER_ADD_MEM_REG is not specified as receiving a file
descriptor, though it obviously does need to do so. All
implementations agree on this one, fix the specification.
* VHOST_USER_REM_MEM_REG is not specified as receiving a file
descriptor either, and it also has no reason to do so. rust-vmm does
not send file descriptors for removing a memory region (in agreement
with the specification), libvhost-user and QEMU do (which is a bug),
though libvhost-user doesn't actually make any use of it.
Change the specification so that for compatibility QEMU's behaviour
becomes legal, even if discouraged, but rust-vmm's behaviour becomes
the explicitly recommended mode of operation.
* VHOST_USER_ADD_MEM_REG doesn't have a documented return value, which
is the desired behaviour in the non-postcopy case. It also implemented
like this in QEMU and rust-vmm, though libvhost-user is buggy and
sometimes sends an unexpected reply. This will be fixed in a separate
patch.
However, in postcopy mode it does reply like VHOST_USER_SET_MEM_TABLE.
This behaviour is shared between libvhost-user and QEMU; rust-vmm
doesn't implement postcopy mode yet. Mention it explicitly in the
spec.
* The specification doesn't mention how VHOST_USER_REM_MEM_REG
identifies the memory region to be removed. Change it to describe the
existing behaviour of libvhost-user (guest address, user address and
size must match).
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20220407133657.155281-2-kwolf@redhat.com>
Reviewed-by: Raphael Norwitz <raphael.norwitz@nutanix.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The 'throttle' block driver implements .bdrv_co_drain_end, so
blockdev-reopen will have to wait for it to complete in the polling
loop at the end of qmp_blockdev_reopen(). This makes AIO_WAIT_WHILE()
release the AioContext lock, which causes a crash if the lock hasn't
correctly been taken.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20220203140534.36522-3-kwolf@redhat.com>
Reviewed-by: Hanna Reitz <hreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The vhost-user-blk export runs requests asynchronously in their own
coroutine. When the vhost connection goes away and we want to stop the
vhost-user server, we need to wait for these coroutines to stop before
we can unmap the shared memory. Otherwise, they would still access the
unmapped memory and crash.
This introduces a refcount to VuServer which is increased when spawning
a new request coroutine and decreased before the coroutine exits. The
memory is only unmapped when the refcount reaches zero.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20220125151435.48792-1-kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This demonstrates what happens when the block status changes in
sub-min_sparse granularity, but all of the parts are zeroed out. The
alignment logic in is_allocated_sectors() prevents that the target image
remains fully sparse as expected, but turns it into a data cluster of
explicit zeros.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20211217164654.1184218-2-vsementsov@virtuozzo.com>
Tested-by: Peter Lieven <pl@kamp.de>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The calculation in sector2cluster() is done relative to the offset of
the root directory. Any writes to blocks before the start of the root
directory (in particular, writes to the FAT) result in negative values,
which are not handled correctly in vvfat_write().
This changes sector2cluster() to return a signed value, and makes sure
that vvfat_write() doesn't try to find mappings for negative cluster
number. It clarifies the code in vvfat_write() to make it more obvious
that the cluster numbers can be negative.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20211209152231.23756-1-kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The size of the qcow size was calculated so that only the FAT partition
would fit on it, but not the whole disk. However, offsets relative to
the whole disk are used to access it, so increase its size to be large
enough for that.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20211209151815.23495-1-kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
At the end of a reopen, we already call bdrv_refresh_limits(), which
should update bs->request_alignment according to the new file
descriptor. However, raw_probe_alignment() relies on s->needs_alignment
and just uses 1 if it isn't set. We neglected to update this field, so
starting with cache=writeback and then reopening with cache=none means
that we get an incorrect bs->request_alignment == 1 and unaligned
requests fail instead of being automatically aligned.
Fix this by recalculating s->needs_alignment in raw_refresh_limits()
before calling raw_probe_alignment().
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20211104113109.56336-1-kwolf@redhat.com>
Reviewed-by: Hanna Reitz <hreitz@redhat.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20211115145409.176785-13-kwolf@redhat.com>
[hreitz: Fix iotest 142 for block sizes greater than 512 by operating on
a file with a size of 1 MB]
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
Message-Id: <20211116101431.105252-1-hreitz@redhat.com>
While introducing a non-QemuOpts code path for device creation for JSON
-device, we noticed that QMP device_add doesn't check its input
correctly (accepting arguments that should have been rejected), and that
users may be relying on this behaviour (libvirt did until it was fixed
recently).
Let's use a deprecation period before we fix this bug in QEMU to avoid
nasty surprises for users.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20211111143530.18985-1-kwolf@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20211115145409.176785-12-kwolf@redhat.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
Like we already do for -object, introduce support for JSON syntax in
-device, which can be kept stable in the long term and guarantees that a
single code path with identical behaviour is used for both QMP and the
command line. Compared to the QemuOpts based code, the parser contains
less surprises and has support for non-scalar options (lists and
structs). Switching management tools to JSON means that we can more
easily change the "human" CLI syntax from QemuOpts to the keyval parser
later.
In the QAPI schema, a feature flag is added to the device-add command to
allow management tools to detect support for this.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20211008133442.141332-16-kwolf@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Tested-by: Peter Krempa <pkrempa@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
QDicts are both what QMP natively uses and what the keyval parser
produces. Going through QemuOpts isn't useful for either one, so switch
the main device creation function to QDicts. By sharing more code with
the -object/object-add code path, we can even reduce the code size a
bit.
This commit doesn't remove the detour through QemuOpts from any code
path yet, but it allows the following commits to do so.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20211008133442.141332-15-kwolf@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Tested-by: Peter Krempa <pkrempa@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Don't go through the global QemuOptsList, it is state of the legacy
command line parser and we will create devices that are not contained
in it. It is also just the command line configuration and not
necessarily the current runtime state.
Instead, look at the qdev device tree which has the current state of all
existing devices.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20211008133442.141332-14-kwolf@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Tested-by: Peter Krempa <pkrempa@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Instead of accessing the global QemuOptsList, which really belong to the
command line parser and shouldn't be accessed from devices, store a
pointer to the QemuOpts in a new VirtIONet field.
This is not the final state, but just an intermediate step to get rid of
QemuOpts in devices. It will later be replaced with an options QDict.
Before this patch, two "primary" devices could be hidden for the same
standby device, but only one of them would actually be enabled and the
other one would be kept hidden forever, so this doesn't make sense.
After this patch, configuring a second primary device is an error.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20211008133442.141332-13-kwolf@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Tested-by: Peter Krempa <pkrempa@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
hide_device() is used for virtio-net failover, where the standby virtio
device delays creation of the primary device. It only makes sense to
have a single primary device for each standby device. Adding a second
one should result in an error instead of hiding it and never using it
afterwards.
Prepare for this by adding an Error parameter to the hide_device()
callback where virtio-net is informed about adding a primary device.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20211008133442.141332-12-kwolf@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Tested-by: Peter Krempa <pkrempa@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The only thing the string visitor adds compared to a keyval visitor is
list support. git grep for 'visit_start_list' and 'visit.*List' shows
that devices don't make use of this.
In a world with a QAPIfied command line interface, the keyval visitor is
used to parse the command line. In order to make sure that no devices
start using this feature that would make backwards compatibility harder,
just switch away from object_property_parse(), which internally uses the
string visitor, to a keyval visitor and object_property_set().
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20211008133442.141332-8-kwolf@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Tested-by: Peter Krempa <pkrempa@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The iothread isn't called 'iothread0', but 'thread0'. Depending on the
order that properties are parsed, the error message may change from the
expected one to another one saying that the iothread doesn't exist.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20211008133442.141332-7-kwolf@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Tested-by: Peter Krempa <pkrempa@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
vhost-vdpa works only with specific devices. At startup, it second
guesses what the command line option handling will do and error out if
it thinks a non-virtio device will attach to them.
This second guessing is not only ugly, it can lead to wrong error
messages ('-device floppy,netdev=foo' should complain about an unknown
property, not about the wrong kind of network device being attached) and
completely ignores hotplugging.
Drop the old checks and implement .check_peer_type() instead to fix
this. As a nice side effect, it also removes one more dependency on the
legacy QemuOpts infrastructure and even reduces the code size.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20211008133442.141332-4-kwolf@redhat.com>
Reviewed-by: Damien Hedde <damien.hedde@greensocs.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Tested-by: Peter Krempa <pkrempa@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
vhost-user works only with specific devices. At startup, it second
guesses what the command line option handling will do and error out if
it thinks a non-virtio device will attach to them.
This second guessing is not only ugly, it can lead to wrong error
messages ('-device floppy,netdev=foo' should complain about an unknown
property, not about the wrong kind of network device being attached) and
completely ignores hotplugging.
Drop the old checks and implement .check_peer_type() instead to fix
this. As a nice side effect, it also removes one more dependency on the
legacy QemuOpts infrastructure and even reduces the code size.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20211008133442.141332-3-kwolf@redhat.com>
Reviewed-by: Damien Hedde <damien.hedde@greensocs.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Tested-by: Peter Krempa <pkrempa@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Some network backends (vhost-user and vhost-vdpa) work only with
specific devices. At startup, they second guess what the command line
option handling will do and error out if they think a non-virtio device
will attach to them.
This second guessing is not only ugly, it can lead to wrong error
messages ('-device floppy,netdev=foo' should complain about an unknown
property, not about the wrong kind of network device being attached) and
completely ignores hotplugging.
Add a callback where backends can check compatibility with a device when
it actually tries to attach, even on hotplug.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20211008133442.141332-2-kwolf@redhat.com>
Reviewed-by: Damien Hedde <damien.hedde@greensocs.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Tested-by: Peter Krempa <pkrempa@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
When bdrv_pad_request() fails in bdrv_co_preadv_part(), bs->in_flight
has been increased, but is never decreased again. This leads to a hang
when trying to drain the block node.
This bug was observed with Windows guests which issue a request that
fully uses IOV_MAX during installation, so that when padding is
necessary (O_DIRECT with a 4k sector size block device on the host),
adding another entry causes failure.
Call bdrv_dec_in_flight() to fix this. There is a larger problem to
solve here because this request shouldn't even fail, but Windows doesn't
seem to care and with this minimal fix the installation succeeds. So
given that we're already in freeze, let's take this minimal fix for 6.1.
Fixes: 98ca45494f
Fixes: https://bugzilla.redhat.com/show_bug.cgi?id=1972079
Reported-by: Qing Wang <qinwang@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20210727154923.91067-1-kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Currently, the block driver whitelists are only applied for the system
emulator. All other binaries still give unrestricted access to all block
drivers. There are use cases where this made sense because the main
concern was avoiding customers running VMs on less optimised block
drivers and getting bad performance. Allowing the same image format e.g.
as a target for 'qemu-img convert' is not a problem then.
However, if the concern is the supportability of the driver in general,
either in full or when used read-write, not applying the list driver
whitelist in tools doesn't help - especially since qemu-nbd and
qemu-storage-daemon now give access to more or less the same operations
in block drivers as running a system emulator.
In order to address this, introduce a new configure option that enforces
the driver whitelist in all binaries.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20210709164141.254097-1-kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
As the BlockReopenQueue can contain nodes in multiple AioContexts, only
one of which may be locked when AIO_WAIT_WHILE() can be called, we can't
let the caller lock the right contexts. Instead, individually lock the
AioContext of a single node when iterating the queue.
Reintroduce bdrv_reopen() as a wrapper for reopening a single node that
drains the node and temporarily drops the AioContext lock for
bdrv_reopen_multiple().
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20210708114709.206487-4-kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Without an external data file, s->data_file is a second pointer with the
same value as bs->file. When changing bs->file to a different BdrvChild
and freeing the old BdrvChild, s->data_file must also be updated,
otherwise it points to freed memory and causes crashes.
This problem was caught by iotests case 245.
Fixes: df2b7086f169239ebad5d150efa29c9bb6d4f820
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20210708114709.206487-2-kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
dev->max_queues was never initialised for backends that don't support
VHOST_USER_PROTOCOL_F_MQ, so it would use 0 as the maximum number of
queues to check against and consequently fail for any such backend.
Set it to 1 if the backend doesn't have multiqueue support.
Fixes: c90bd505a3
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20210705171429.29286-1-kwolf@redhat.com>
Reviewed-by: Cornelia Huck <cohuck@redhat.com>
Reviewed-by: Raphael Norwitz <raphael.norwitz@nutanix.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Commit dabefdd6 removed code that was supposed to try reconnecting
during .realize(), but actually just crashed and had several design
problems.
This adds the feature back without the crash in simple cases while also
fixing some design problems: Reconnection is now only tried if there was
a problem with the connection and not an error related to the content
(which would fail again the same way in the next attempt). Reconnection
is limited to three attempts (four with the initial attempt) so that we
won't end up in an infinite loop if a problem is permanent. If the
backend restarts three times in the very short time window of device
initialisation, we have bigger problems and erroring out is the right
course of action.
In the case that a connection error occurs and we reconnect, the error
message is printed using error_report_err(), but otherwise ignored.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20210609154658.350308-8-kwolf@redhat.com>
Reviewed-by: Raphael Norwitz <raphael.norwitz@nutanix.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This function is the part that we will want to retry if the connection
is lost during initialisation, so factor it out to keep the following
patch simpler.
The error path for vhost_dev_get_config() forgot disconnecting the
chardev, add this while touching the code.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20210609154658.350308-7-kwolf@redhat.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Reviewed-by: Raphael Norwitz <raphael.norwitz@nutanix.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Instead of just returning 0/-1 and letting the caller make up a
meaningless error message, add an Error parameter to allow reporting the
real error and switch to 0/-errno so that different kind of errors can
be distinguished in the caller.
config_len in vhost_user_get_config() is defined by the device, so if
it's larger than VHOST_USER_MAX_CONFIG_SIZE, this is a programming
error. Turn the corresponding check into an assertion.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20210609154658.350308-6-kwolf@redhat.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Reviewed-by: Raphael Norwitz <raphael.norwitz@nutanix.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Instead of just returning 0/-1 and letting the caller make up a
meaningless error message, switch to 0/-errno so that different kinds of
errors can be distinguished in the caller.
This involves changing a few more callbacks in VhostOps to return
0/-errno: .vhost_set_owner(), .vhost_get_features() and
.vhost_virtqueue_set_busyloop_timeout(). The implementations of these
functions are trivial as they generally just send a message to the
backend.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20210609154658.350308-4-kwolf@redhat.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Reviewed-by: Raphael Norwitz <raphael.norwitz@nutanix.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Instead of just returning 0/-1 and letting the caller make up a
meaningless error message, add an Error parameter to allow reporting the
real error and switch to 0/-errno so that different kind of errors can
be distinguished in the caller.
Specifically, in vhost-user, EPROTO is used for all errors that relate
to the connection itself, whereas other error codes are used for errors
relating to the content of the connection. This will allow us later to
automatically reconnect when the connection goes away, without ending up
in an endless loop if it's a permanent error in the configuration.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20210609154658.350308-3-kwolf@redhat.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Reviewed-by: Raphael Norwitz <raphael.norwitz@nutanix.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This allows callers to return better error messages instead of making
one up while the real error ends up on stderr. Most callers can
immediately make use of this because they already have an Error
parameter themselves. The others just keep printing the error with
error_report_err().
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20210609154658.350308-2-kwolf@redhat.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Reviewed-by: Raphael Norwitz <raphael.norwitz@nutanix.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Creating a device with a number of queues that isn't supported by the
backend is pointless, the device won't work properly and the error
messages are rather confusing.
Just fail to create the device if num-queues is higher than what the
backend supports.
Since the relationship between num-queues and the number of virtqueues
depends on the specific device, this is an additional value that needs
to be initialised by the device. For convenience, allow leaving it 0 if
the check should be skipped. This makes sense for vhost-user-net where
separate vhost devices are used for the queues and custom initialisation
code is needed to perform the check.
Fixes: https://bugzilla.redhat.com/show_bug.cgi?id=1935031
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Raphael Norwitz <raphael.norwitz@nutanix.com>
Message-Id: <20210429171316.162022-7-kwolf@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Commit 2943b53f6 (' virtio: force VIRTIO_F_IOMMU_PLATFORM') made sure
that vhost can't just reject VIRTIO_F_IOMMU_PLATFORM when it was
requested. However, just adding it back to the negotiated flags isn't
right either because it promises support to the guest that the device
actually doesn't support. One example of a vhost-user device that
doesn't have support for the flag is the vhost-user-blk export of QEMU.
Instead of successfully creating a device that doesn't work, just fail
to plug the device when it doesn't support the feature, but it was
requested. This results in much clearer error messages.
Fixes: https://bugzilla.redhat.com/show_bug.cgi?id=1935019
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Raphael Norwitz <raphael.norwitz@nutanix.com>
Message-Id: <20210429171316.162022-6-kwolf@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Now that vhost_user_blk_connect() is not called from an event handler
any more, but directly from vhost_user_blk_device_realize(), we can
actually make use of Error again instead of calling error_report() in
the inner function and setting a more generic and therefore less useful
error message in realize() itself.
With Error, the callers are responsible for adding context if necessary
(such as the "-device" option the error refers to). Additional prefixes
are redundant and better omitted.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Acked-by: Raphael Norwitz <raphael.norwitz@nutanix.com>
Message-Id: <20210429171316.162022-4-kwolf@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This is a partial revert of commits 77542d4314 and bc79c87bcd.
Usually, an error during initialisation means that the configuration was
wrong. Reconnecting won't make the error go away, but just turn the
error condition into an endless loop. Avoid this and return errors
again.
Additionally, calling vhost_user_blk_disconnect() from the chardev event
handler could result in use-after-free because none of the
initialisation code expects that the device could just go away in the
middle. So removing the call fixes crashes in several places.
For example, using a num-queues setting that is incompatible with the
backend would result in a crash like this (dereferencing dev->opaque,
which is already NULL):
#0 0x0000555555d0a4bd in vhost_user_read_cb (source=0x5555568f4690, condition=(G_IO_IN | G_IO_HUP), opaque=0x7fffffffcbf0) at ../hw/virtio/vhost-user.c:313
#1 0x0000555555d950d3 in qio_channel_fd_source_dispatch (source=0x555557c3f750, callback=0x555555d0a478 <vhost_user_read_cb>, user_data=0x7fffffffcbf0) at ../io/channel-watch.c:84
#2 0x00007ffff7b32a9f in g_main_context_dispatch () at /lib64/libglib-2.0.so.0
#3 0x00007ffff7b84a98 in g_main_context_iterate.constprop () at /lib64/libglib-2.0.so.0
#4 0x00007ffff7b32163 in g_main_loop_run () at /lib64/libglib-2.0.so.0
#5 0x0000555555d0a724 in vhost_user_read (dev=0x555557bc62f8, msg=0x7fffffffcc50) at ../hw/virtio/vhost-user.c:402
#6 0x0000555555d0ee6b in vhost_user_get_config (dev=0x555557bc62f8, config=0x555557bc62ac "", config_len=60) at ../hw/virtio/vhost-user.c:2133
#7 0x0000555555d56d46 in vhost_dev_get_config (hdev=0x555557bc62f8, config=0x555557bc62ac "", config_len=60) at ../hw/virtio/vhost.c:1566
#8 0x0000555555cdd150 in vhost_user_blk_device_realize (dev=0x555557bc60b0, errp=0x7fffffffcf90) at ../hw/block/vhost-user-blk.c:510
#9 0x0000555555d08f6d in virtio_device_realize (dev=0x555557bc60b0, errp=0x7fffffffcff0) at ../hw/virtio/virtio.c:3660
Note that this removes the ability to reconnect during initialisation
(but not during operation) when there is no permanent error, but the
backend restarts, as the implementation was buggy. This feature can be
added back in a follow-up series after changing error paths to
distinguish cases where retrying could help from cases with permanent
errors.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20210429171316.162022-3-kwolf@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
For a successful conversion of an image, we must make sure that its
content doesn't change during the conversion.
A special case of this is using the same image file both as the source
and as the destination. If both input and output format are raw, the
operation would just be useless work, with other formats it is a sure
way to destroy the image. This will now fail because the image file
can't be opened a second time for the output when opening it for the
input has already acquired file locks to unshare BLK_PERM_WRITE.
Nevertheless, if there is some reason in a special case why it is
actually okay to allow writes to the image while it is being converted,
-U can still be used to force sharing all permissions.
Note that for most image formats, BLK_PERM_WRITE would already be
unshared by the format driver, so this only really makes a difference
for raw source images (but any output format).
Reported-by: Xueqiang Wei <xuwei@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20210422164344.283389-3-kwolf@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Normally, blk_new_open() just shares all permissions. This was fine
originally when permissions only protected against uses in the same
process because no other part of the code would actually get to access
the block nodes opened with blk_new_open(). However, since we use it for
file locking now, unsharing permissions becomes desirable.
Add a new BDRV_O_NO_SHARE flag that is used in blk_new_open() to unshare
any permissions that can be unshared.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20210422164344.283389-2-kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Both callers use callbacks that don't do anything when they are called
for CLI aliases. Instead of passing the cli_alias parameter, just don't
call the callbacks for aliases in the first place.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20210311164253.338723-4-kwolf@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
QAPI doesn't know the aliases 'tty' and 'parport' and there is no
reason to prefer them to the real names of the backends 'serial' and
'parallel'.
Since warnings are not allowed in 'make check' output, we can't test
the deprecated alias any more. Remove it from test-char.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20210311164253.338723-3-kwolf@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The aliases "tty" and "parport" are only valid on the command line, QMP
commands like chardev-add don't know them. query-chardev-backends should
describe QMP and therefore not include them in the list of available
backends.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20210311164253.338723-2-kwolf@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The system emulator has a more complicated way of handling command line
options in that it reorders options before it processes them. This means
that parsing object options and creating the object happen at two
different points. Split the parsing part into a separate function that
can be reused by the system emulator command line.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Acked-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
This switches the HMP command object_add from a QemuOpts-based parser to
user_creatable_add_from_str() which uses a keyval parser and enforces
the QAPI schema.
Apart from being a cleanup, this makes non-scalar properties and help
accessible. In order for help to be printed to the monitor instead of
stdout, the printf() calls in the help functions are changed to
qemu_printf().
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Acked-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
This switches qemu-img from a QemuOpts-based parser for --object to
user_creatable_process_cmdline() which uses a keyval parser and enforces
the QAPI schema.
Apart from being a cleanup, this makes non-scalar properties accessible.
As a side effect, fix wrong exit codes in the object parsing error path
of 'qemu-img compare'. This was broken in commit 334c43e2c3 because
&error_fatal exits with an exit code of 1, while it should have been 2.
Document that exit code 0 is also returned when just requested help was
printed instead of comparing images. This is preexisting behaviour that
isn't changed by this patch, though another instance of it is added with
'--object help'.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Acked-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
This is a version of user_creatable_process_cmdline() with an Error
parameter that never calls exit() and is therefore usable in HMP.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Acked-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
This switches qemu-nbd from a QemuOpts-based parser for --object to
user_creatable_process_cmdline() which uses a keyval parser and enforces
the QAPI schema.
Apart from being a cleanup, this makes non-scalar properties accessible.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Acked-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
This switches qemu-io from a QemuOpts-based parser for --object to
user_creatable_process_cmdline() which uses a keyval parser and enforces
the QAPI schema.
Apart from being a cleanup, this makes non-scalar properties accessible.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Acked-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
The implementation for --object can be shared between
qemu-storage-daemon and other binaries, so move it into a function in
qom/object_interfaces.c that is accessible from everywhere.
This also requires moving the implementation of qmp_object_add() into a
new user_creatable_add_qapi(), because qom/qom-qmp-cmds.c is not linked
for tools.
user_creatable_print_help_from_qdict() can become static now.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Acked-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
This code is going away anyway, but for a few more commits, we'll be in
a state where some binaries still use QemuOpts and others don't. If the
"object" QemuOptsList doesn't even exist, we don't have to remove (or
fail to remove, and therefore abort) a user creatable object from it.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Acked-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
This converts object-add from 'gen': false to the ObjectOptions QAPI
type. As an immediate benefit, clients can now use QAPI schema
introspection for user creatable QOM objects.
It is also the first step towards making the QAPI schema the only
external interface for the creation of user creatable objects. Once all
other places (HMP and command lines of the system emulator and all
tools) go through QAPI, too, some object implementations can be
simplified because some checks (e.g. that mandatory options are set) are
already performed by QAPI, and in another step, QOM boilerplate code
could be generated from the schema.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Acked-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
This adds a QAPI schema for the properties of the input-* objects.
ui.json cannot be included in qom.json because the storage daemon can't
use it, so move GrabToggleKeys to common.json.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Acked-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
This adds a QAPI schema for the properties of the objects implementing
the confidential-guest-support interface.
pef-guest and s390x-pv-guest don't have any properties, so they only
need to be added to the ObjectType enum without adding a new branch to
ObjectOptions.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Acked-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
This adds a QAPI schema for the properties of the filter-* objects.
Some parts of the interface (in particular NetfilterProperties.position)
are very unusual for QAPI, but for now just describe the existing
interface.
net.json can't be included in qom.json because the storage daemon
doesn't have it. NetFilterDirection is still required in the new object
property definitions in qom.json, so move this enum to common.json.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Acked-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
This adds a QAPI schema for the properties of the can-* objects.
can-bus doesn't have any properties, so it only needs to be added to the
ObjectType enum without adding a new branch to ObjectOptions.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Acked-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
This adds a QAPI schema for the properties of the tls-* objects.
The 'loaded' property doesn't seem to make sense as an external
interface: It is automatically set to true in ucc->complete, and
explicitly setting it to true earlier just means that additional options
will be silently ignored.
In other words, the 'loaded' property is useless. Mark it as deprecated
in the schema from the start.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Acked-by: Peter Krempa <pkrempa@redhat.com>
Acked-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
This adds a QAPI schema for the properties of the secret* objects.
The 'loaded' property doesn't seem to make sense as an external
interface: It is automatically set to true in ucc->complete, and
explicitly setting it to true earlier just means that additional options
will be silently ignored.
In other words, the 'loaded' property is useless. Mark it as deprecated
in the schema from the start.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Acked-by: Peter Krempa <pkrempa@redhat.com>
Acked-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
This adds a QAPI schema for the properties of the throttle-group object.
The only purpose of the x-* properties is to make the nested options in
'limits' available for a command line parser that doesn't support
structs. Any parser that will use the QAPI schema will supports structs,
though, so they will not be needed in the schema in the future.
To keep the conversion straightforward, add them to the schema anyway.
We can then remove the options and adjust documentation, test cases etc.
in a separate patch.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Acked-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
This adds a QAPI schema for the properties of the rng-* objects.
The 'opened' property doesn't seem to make sense as an external
interface: It is automatically set to true in ucc->complete, and
explicitly setting it to true earlier just means that trying to set
additional options will result in an error. After the property has once
been set to true (i.e. when the object construction has completed), it
can never be reset to false. In other words, the 'opened' property is
useless. Mark it as deprecated in the schema from the start.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Acked-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
This adds a QAPI schema for the properties of the memory-backend-*
objects.
HostMemPolicy has to be moved to an include file that can be used by the
storage daemon, too, because ObjectOptions must be the same in all
binaries if we don't want to compile the whole code multiple times.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Acked-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
This adds a QAPI schema for the properties of the dbus-vmstate object.
A list represented as a comma separated string is clearly not very
QAPI-like, but for now just describe the existing interface.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Acked-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
This adds a QAPI schema for the properties of the cryptodev-* objects.
These interfaces have some questionable aspects (cryptodev-backend is
really an abstract base class without function, and the queues option
only makes sense for cryptodev-vhost-user), but as the goal is to
represent the existing interface in QAPI, leave these things in place.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Acked-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Add an ObjectOptions union that will eventually describe the options of
all user creatable object types. As unions can't exist without any
branches, also add the first object type.
This adds a QAPI schema for the properties of the iothread object.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Acked-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
The 'props' option has been deprecated in 5.0 in favour of a flattened
object-add command. Time to change our test cases to drop the deprecated
option.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Acked-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
The image streaming block job restricts shared permissions of the nodes
it accesses. This can obviously fail when other users already got these
permissions. &error_abort is therefore wrong and can crash. Handle these
errors gracefully and just fail starting the block job.
Reported-by: Nini Gu <ngu@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20210309173451.45152-1-kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Alberto Garcia <berto@igalia.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Most code paths for creating a user creatable object go through
QemuOpts, which ensures that the provided 'id' option is actually a
valid identifier.
However, there are some code paths that don't go through QemuOpts:
qemu-storage-daemon --object (since commit 8db1efd3) and QMP object-add
(since it was first introduced in commit cff8b2c6). We need to have the
same validity check for those, too.
This adds the check and makes it print the same error message as
QemuOpts on failure.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20210302171623.49709-1-kwolf@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Before this patch, monitor_qmp_dispatcher_co() used to check whether
shutdown is requested only when it would have to wait for new requests.
If there were still some queued requests, it would try to execute all of
them before shutting down.
This can be surprising when the queued QMP commands take long or hang
because Ctrl-C may not actually exit QEMU as soon as possible.
Change monitor_qmp_dispatcher_co() so that it additionally checks
whether shutdown is request before it gets a new request from the queue.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20210212172028.288825-3-kwolf@redhat.com>
Tested-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Commit 357bda95 already tried to fix the order in monitor_cleanup() by
moving shutdown of the dispatcher coroutine further to the start.
However, it didn't go far enough:
iothread_stop() makes sure that all pending work (bottom halves) in the
AioContext of the monitor iothread is completed. iothread_destroy()
depends on this and fails an assertion if there is still a pending BH.
While the dispatcher coroutine is running, it will try to resume the
monitor after taking a request out of the queue, which involves a BH.
The dispatcher is run until it terminates in the AIO_WAIT_WHILE() loop.
However, adding new BHs between iothread_stop() and iothread_destroy()
is forbidden.
Fix this by stopping the dispatcher first before shutting down the other
parts of the monitor. This means we can now receive requests that aren't
handled any more when QEMU is shutting down, but this is unlikely to be
a problem for QMP clients.
Fixes: 357bda9590
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20210212172028.288825-2-kwolf@redhat.com>
Tested-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
As we don't have a fully QAPIfied version of object-add yet and it still
has 'gen': false in the schema, it needs to be registered explicitly in
init_qmp_commands() to be available for users.
Fixes: 2af282ec51
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20210204072137.19663-1-kwolf@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
size_to_str() can return a size like "4.24 MiB", with a single digit
integer part and two fractional digits. This is eight characters, but
commit b39847a5 changed the format string to only reserve seven
characters for the column.
This can result in unaligned columns, which in turn changes the output of
iotests case 267 because exceeding the column size defeats the attempt
to filter the size out of the output (observed with the ppc64 emulator).
The resulting change is only a whitespace change, but since commit
f203080b this is enough for iotests to consider the test failed.
Taking a character away from the tag name column and adding it to the VM
size column doesn't change anything in the common case (the tag name is
left justified, the VM size is right justified), but fixes this case.
Fixes: b39847a505
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20210202155911.179865-1-kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
For -makecheck, the old 'check' implementation skipped the output when
starting a test. It only had the condensed output at the end of a test.
testrunner.py prints the normal output when starting a test even for
-makecheck. This output contains '\r' at the end so that it can be
overwritten with the result at the end of the test. However, for
-makecheck this is shorter output in a different format, so effectively
we end up with garbled output that mixes both output forms.
Revert to the old behaviour of only printing a message after the test
had completed in -makecheck mode.
Fixes: d74c754c92
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20210201161024.127921-1-kwolf@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
If the qemu-system-{arch} binary for the host architecture can't be
found, the old 'check' implementation selected the alphabetically first
system emulator binary that it could find. The new Python implementation
just uses the first result of glob.iglob(), which has an undefined
order.
This is a problem that breaks CI because the iotests aren't actually
prepared to run on any emulator. They should be, so this is really a bug
in the failing test cases that should be fixed there, but as a quick
fix, let's revert to the old behaviour to let CI runs succeed again.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20210202142802.119999-1-kwolf@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Kevin Wolf