Introduce a new API for thread-local blk_io_plug() that does not
traverse the block graph. The goal is to make blk_io_plug() multi-queue
friendly.
Instead of having block drivers track whether or not we're in a plugged
section, provide an API that allows them to defer a function call until
we're unplugged: blk_io_plug_call(fn, opaque). If blk_io_plug_call() is
called multiple times with the same fn/opaque pair, then fn() is only
called once at the end of the function - resulting in batching.
This patch introduces the API and changes blk_io_plug()/blk_io_unplug().
blk_io_plug()/blk_io_unplug() no longer require a BlockBackend argument
because the plug state is now thread-local.
Later patches convert block drivers to blk_io_plug_call() and then we
can finally remove .bdrv_co_io_plug() once all block drivers have been
converted.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Acked-by: Kevin Wolf <kwolf@redhat.com>
Message-id: 20230530180959.1108766-2-stefanha@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
The virtio-scsi Host Bus Adapter provides access to devices on a SCSI
bus. Those SCSI devices typically have a BlockBackend. When the
BlockBackend enters a drained section, the SCSI device must temporarily
stop submitting new I/O requests.
Implement this behavior by temporarily stopping virtio-scsi virtqueue
processing when one of the SCSI devices enters a drained section. The
new scsi_device_drained_begin() API allows scsi-disk to message the
virtio-scsi HBA.
scsi_device_drained_begin() uses a drain counter so that multiple SCSI
devices can have overlapping drained sections. The HBA only sees one
pair of .drained_begin/end() calls.
After this commit, virtio-scsi no longer depends on hw/virtio's
ioeventfd aio_set_event_notifier(is_external=true). This commit is a
step towards removing the aio_disable_external() API.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20230516190238.8401-19-stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This patch is part of an effort to remove the aio_disable_external()
API because it does not fit in a multi-queue block layer world where
many AioContexts may be submitting requests to the same disk.
The SCSI emulation code is already in good shape to stop using
aio_disable_external(). It was only used by commit 9c5aad84da
("virtio-scsi: fixed virtio_scsi_ctx_check failed when detaching scsi
disk") to ensure that virtio_scsi_hotunplug() works while the guest
driver is submitting I/O.
Ensure virtio_scsi_hotunplug() is safe as follows:
1. qdev_simple_device_unplug_cb() -> qdev_unrealize() ->
device_set_realized() calls qatomic_set(&dev->realized, false) so
that future scsi_device_get() calls return NULL because they exclude
SCSIDevices with realized=false.
That means virtio-scsi will reject new I/O requests to this
SCSIDevice with VIRTIO_SCSI_S_BAD_TARGET even while
virtio_scsi_hotunplug() is still executing. We are protected against
new requests!
2. scsi_qdev_unrealize() already contains a call to
scsi_device_purge_requests() so that in-flight requests are cancelled
synchronously. This ensures that no in-flight requests remain once
qdev_simple_device_unplug_cb() returns.
Thanks to these two conditions we don't need aio_disable_external()
anymore.
Cc: Zhengui Li <lizhengui@huawei.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Daniil Tatianin <d-tatianin@yandex-team.ru>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20230516190238.8401-5-stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Only report a transport reset event to the guest after the SCSIDevice
has been unrealized by qdev_simple_device_unplug_cb().
qdev_simple_device_unplug_cb() sets the SCSIDevice's qdev.realized field
to false so that scsi_device_find/get() no longer see it.
scsi_target_emulate_report_luns() also needs to be updated to filter out
SCSIDevices that are unrealized.
Change virtio_scsi_push_event() to take event information as an argument
instead of the SCSIDevice. This allows virtio_scsi_hotunplug() to emit a
VIRTIO_SCSI_T_TRANSPORT_RESET event after the SCSIDevice has already
been unrealized.
These changes ensure that the guest driver does not see the SCSIDevice
that's being unplugged if it responds very quickly to the transport
reset event.
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Daniil Tatianin <d-tatianin@yandex-team.ru>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20230516190238.8401-4-stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
When an IOThread is configured, the ctrl virtqueue is processed in the
IOThread. TMFs that reset SCSI devices are currently called directly
from the IOThread and trigger an assertion failure in blk_drain() from
the following call stack:
virtio_scsi_handle_ctrl_req -> virtio_scsi_do_tmf -> device_code_reset
-> scsi_disk_reset -> scsi_device_purge_requests -> blk_drain
../block/block-backend.c:1780: void blk_drain(BlockBackend *): Assertion `qemu_in_main_thread()' failed.
The blk_drain() function is not designed to be called from an IOThread
because it needs the Big QEMU Lock (BQL).
This patch defers TMFs that reset SCSI devices to a Bottom Half (BH)
that runs in the main loop thread under the BQL. This way it's safe to
call blk_drain() and the assertion failure is avoided.
Introduce s->tmf_bh_list for tracking TMF requests that have been
deferred to the BH. When the BH runs it will grab the entire list and
process all requests. Care must be taken to clear the list when the
virtio-scsi device is reset or unrealized. Otherwise deferred TMF
requests could execute later and lead to use-after-free or other
undefined behavior.
The s->resetting counter that's used by TMFs that reset SCSI devices is
accessed from multiple threads. This patch makes that explicit by using
atomic accessor functions. With this patch applied the counter is only
modified by the main loop thread under the BQL but can be read by any
thread.
Reported-by: Qing Wang <qinwang@redhat.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20230221212218.1378734-4-stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
hw/pci/pci_bridge.h and hw/cxl/cxl.h include each other.
Fortunately, breaking the loop is merely a matter of deleting
unnecessary includes from headers, and adding them back in places
where they are now missing.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20221222100330.380143-2-armbru@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
In the SCSI subsystem we currently use the legacy functions
qdev_reset_all() and qbus_reset_all(). These perform a recursive
reset, starting from either a qbus or a qdev. However they do not
permit any of the devices in the tree to use three-phase reset,
because device reset goes through the device_legacy_reset() function
that only calls the single DeviceClass::reset method.
Switch to using the device_cold_reset() and bus_cold_reset()
functions. These also perform a recursive reset, where first the
children are reset and then finally the parent, but they use the new
(...in 2020...) Resettable mechanism, which supports both the old
style single-reset method and also the new 3-phase reset handling.
Since no devices attached to SCSI buses currently try to use 3-phase
reset, this should be a no-behaviour-change commit which just reduces
the use of a deprecated API.
Commit created with:
sed -i -e 's/qdev_reset_all/device_cold_reset/g;s/qbus_reset_all/bus_cold_reset/g' hw/scsi/*.c
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Message-Id: <20221013160623.1296109-2-peter.maydell@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Section 5.6.6.3 of VirtIO specification states, "Events will also
be reported via sense codes..." However, no sense data is sent when
VIRTIO_SCSI_EVT_RESET_RESCAN or VIRTIO_SCSI_EVT_RESET_REMOVED events
are reported (when disk hotplug/hotunplug events occur). SCSI layer
on Solaris depends on this sense data, and hence does not handle disk
hotplug/hotunplug events.
When the disk inventory changes, use the bus unit attention mechanism
to return a CHECK_CONDITION status with sense data of 0x06/0x3F/0x0E
(sense code REPORTED_LUNS_CHANGED). The first device on the bus to
execute a command successfully will report and consume the unit
attention status.
Signed-off-by: Venu Busireddy <venu.busireddy@oracle.com>
Message-Id: <20221006194946.24134-1-venu.busireddy@oracle.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
When a SCSI command is received from the guest, the CDB length implied
by the first byte might exceed the number of bytes the guest sent. In
this case scsi_req_new() will read uninitialized data, causing
unpredictable behavior.
Adds the buf_len parameter to scsi_req_new() and plumbs it through the
call stack.
Signed-off-by: John Millikin <john@john-millikin.com>
Resolves: https://gitlab.com/qemu-project/qemu/-/issues/1127
Message-Id: <20220817053458.698416-1-john@john-millikin.com>
[Fill in correct length for adapters other than ESP. - Paolo]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
This patch drops the name parameter for the virtio_init function.
The pair between the numeric device ID and the string device ID
(name) of a virtio device already exists, but not in a way that
lets us map between them.
This patch lets us do this and removes the need for the name
parameter in the virtio_init function.
Signed-off-by: Jonah Palmer <jonah.palmer@oracle.com>
Message-Id: <1648819405-25696-2-git-send-email-jonah.palmer@oracle.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
There is no longer a need to expose the request and related APIs in
virtio-scsi.h since there are no callers outside virtio-scsi.c.
Note the block comment in VirtIOSCSIReq has been adjusted to meet the
coding style.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Message-id: 20220427143541.119567-7-stefanha@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
virtio_scsi_handle_cmd_vq() is only called from hw/scsi/virtio-scsi.c
now and its return value is no longer used. Remove the function
prototype from virtio-scsi.h and drop the return value.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Message-id: 20220427143541.119567-6-stefanha@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
virtio_scsi_handle_ctrl_vq() is only called from hw/scsi/virtio-scsi.c
now and its return value is no longer used. Remove the function
prototype from virtio-scsi.h and drop the return value.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Message-id: 20220427143541.119567-5-stefanha@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
virtio_scsi_handle_event_vq() is only called from hw/scsi/virtio-scsi.c
now and its return value is no longer used. Remove the function
prototype from virtio-scsi.h and drop the return value.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Message-id: 20220427143541.119567-4-stefanha@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Commit f34e8d8b8d ("virtio-scsi: prepare
virtio_scsi_handle_cmd for dataplane") prepared the virtio-scsi cmd
virtqueue handler function to be used in both the dataplane and
non-datpalane code paths.
It failed to convert the ctrl and event virtqueue handler functions,
which are not designed to be called from the dataplane code path but
will be since the ioeventfd is set up for those virtqueues when
dataplane starts.
Convert the ctrl and event virtqueue handler functions now so they
operate correctly when called from the dataplane code path. Avoid code
duplication by extracting this code into a helper function.
Fixes: f34e8d8b8d ("virtio-scsi: prepare virtio_scsi_handle_cmd for dataplane")
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Message-id: 20220427143541.119567-2-stefanha@redhat.com
[Fixed s/by used/be used/ typo pointed out by Michael Tokarev
<mjt@tls.msk.ru>.
--Stefan]
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Prepare virtio_scsi_handle_cmd() to be used by both dataplane and
non-dataplane by making the condition for starting ioeventfd more
specific. This way it won't trigger when dataplane has already been
started.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Message-id: 20211207132336.36627-5-stefanha@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
The function scsi_bus_new() creates a new SCSI bus; callers can
either pass in a name argument to specify the name of the new bus, or
they can pass in NULL to allow the bus to be given an automatically
generated unique name. Almost all callers want to use the
autogenerated name; the only exception is the virtio-scsi device.
Taking a name argument that should almost always be NULL is an
easy-to-misuse API design -- it encourages callers to think perhaps
they should pass in some standard name like "scsi" or "scsi-bus". We
don't do this anywhere for SCSI, but we do (incorrectly) do it for
other bus types such as i2c.
The function name also implies that it will return a newly allocated
object, when it in fact does in-place allocation. We more commonly
name such functions foo_init(), with foo_new() being the
allocate-and-return variant.
Replace all the scsi_bus_new() callsites with either:
* scsi_bus_init() for the usual case where the caller wants
an autogenerated bus name
* scsi_bus_init_named() for the rare case where the caller
needs to specify the bus name
and document that for the _named() version it's then the caller's
responsibility to think about uniqueness of bus names.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Message-id: 20210923121153.23754-2-peter.maydell@linaro.org
Some SCSI drivers like virtio have an internal mapping for the
host_status. This patch moves the host_status translation into
the SCSI drivers to allow those drivers to set up the correct
values.
Signed-off-by: Hannes Reinecke <hare@suse.de>.
[Added default handling to avoid touching all drivers. - Paolo]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The command complete callback has a SCSIRequest as the first argument,
and the status field of that structure is identical to the 'status'
argument. So drop the argument from the callback.
Signed-off-by: Hannes Reinecke <hare@suse.de>
Message-Id: <20201116184041.60465-3-hare@suse.de>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The documentation for bdrv_set_aio_context_ignore() states this:
* The caller must own the AioContext lock for the old AioContext of bs, but it
* must not own the AioContext lock for new_context (unless new_context is the
* same as the current context of bs).
As blk_set_aio_context() makes use of this function, this rule also
applies to it.
Fix all occurrences where this rule wasn't honored.
Suggested-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Sergio Lopez <slp@redhat.com>
Message-Id: <20201214170519.223781-2-slp@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
Add trace events for virtio command and response tracing.
Signed-off-by: Hannes Reinecke <hare@suse.de>
Message-Id: <20201116183114.55703-2-hare@suse.de>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
This will help us to avoid the scsi device disappearing
after we took a reference to it.
It doesn't by itself forbid case when we try to access
an unrealized device
Suggested-by: Stefan Hajnoczi <stefanha@gmail.com>
Signed-off-by: Maxim Levitsky <mlevitsk@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20200913160259.32145-9-mlevitsk@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20201006123904.610658-13-mlevitsk@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
This fixes the race between device emulation code that tries to find
a child device to dispatch the request to (e.g a scsi disk),
and hotplug of a new device to that bus.
Note that this doesn't convert all the readers of the list
but only these that might go over that list without BQL held.
This is a very small first step to make this code thread safe.
Suggested-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Maxim Levitsky <mlevitsk@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20200913160259.32145-5-mlevitsk@redhat.com>
[Use RCU_READ_LOCK_GUARD in more places, adjust testcase now that
the delay in DEVICE_DELETED due to RCU is more consistent. - Paolo]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20201006123904.610658-9-mlevitsk@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Automatically size the number of virtio-scsi-pci, vhost-scsi-pci, and
vhost-user-scsi-pci request virtqueues to match the number of vCPUs.
Other transports continue to default to 1 request virtqueue.
A 1:1 virtqueue:vCPU mapping ensures that completion interrupts are
handled on the same vCPU that submitted the request. No IPI is
necessary to complete an I/O request and performance is improved. The
maximum number of MSI-X vectors and virtqueues limit are respected.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20200818143348.310613-6-stefanha@redhat.com>
Reviewed-by: Cornelia Huck <cohuck@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>
The event and control virtqueues are always present, regardless of the
multi-queue configuration. Define a constant so that virtqueue number
calculations are easier to read.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Cornelia Huck <cohuck@redhat.com>
Reviewed-by: Pankaj Gupta <pankaj.gupta.linux@gmail.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Raphael Norwitz <raphael.norwitz@nutanix.com>
Message-Id: <20200818143348.310613-5-stefanha@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
qbus_set_hotplug_handler() is a simple wrapper around
object_property_set_link().
object_property_set_link() fails when the property doesn't exist, is
not settable, or its .check() method fails. These are all programming
errors here, so passing &error_abort to qbus_set_hotplug_handler() is
appropriate.
Most of its callers do. Exceptions:
* pcie_cap_slot_init(), shpc_init(), spapr_phb_realize() pass NULL,
i.e. they ignore errors.
* spapr_machine_init() passes &error_fatal.
* s390_pcihost_realize(), virtio_serial_device_realize(),
s390_pcihost_plug() pass the error to their callers. The latter two
keep going after the error, which looks wrong.
Drop the @errp parameter, and instead pass &error_abort to
object_property_set_link().
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>
Message-Id: <20200630090351.1247703-15-armbru@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>
The goal is to reduce the amount of requests issued by a guest on
1M reads/writes. This rises the performance up to 4% on that kind of
disk access pattern.
The maximum chunk size to be used for the guest disk accessing is
limited with seg_max parameter, which represents the max amount of
pices in the scatter-geather list in one guest disk request.
Since seg_max is virqueue_size dependent, increasing the virtqueue
size increases seg_max, which, in turn, increases the maximum size
of data to be read/write from a guest disk.
More details in the original problem statment:
https://lists.gnu.org/archive/html/qemu-devel/2017-12/msg03721.html
Suggested-by: Denis V. Lunev <den@openvz.org>
Signed-off-by: Denis Plotnikov <dplotnikov@virtuozzo.com>
Message-id: 20200214074648.958-1-dplotnikov@virtuozzo.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Use virtio_delete_queue to make it more clear.
Signed-off-by: Pan Nengyuan <pannengyuan@huawei.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20200117075547.60864-3-pannengyuan@huawei.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Before the patch, seg_max parameter was immutable and hardcoded
to 126 (128 - 2) without respect to queue size. This has two negative effects:
1. when queue size is < 128, we have Virtio 1.1 specfication violation:
(2.6.5.3.1 Driver Requirements) seq_max must be <= queue_size.
This violation affects the old Linux guests (ver < 4.14). These guests
crash on these queue_size setups.
2. when queue_size > 128, as was pointed out by Denis Lunev <den@virtuozzo.com>,
seg_max restrics guest's block request length which affects guests'
performance making them issues more block request than needed.
https://lists.gnu.org/archive/html/qemu-devel/2017-12/msg03721.html
To mitigate this two effects, the patch adds the property adjusting seg_max
to queue size automaticaly. Since seg_max is a guest visible parameter,
the property is machine type managable and allows to choose between
old (seg_max = 126 always) and new (seg_max = queue_size - 2) behaviors.
Not to change the behavior of the older VMs, prevent setting the default
seg_max_adjust value for older machine types.
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Denis Plotnikov <dplotnikov@virtuozzo.com>
Message-Id: <20191220140905.1718-2-dplotnikov@virtuozzo.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>
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>
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 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>
commit a6f230c move blockbackend back to main AioContext on unplug. It set the AioContext of
SCSIDevice to the main AioContex, but s->ctx is still the iothread AioContex（if the scsi controller
is configure with iothread）. So if there are having in-flight requests during unplug, a failing assertion
happend. The bt is below:
(gdb) bt
#0 0x0000ffff86aacbd0 in raise () from /lib64/libc.so.6
#1 0x0000ffff86aadf7c in abort () from /lib64/libc.so.6
#2 0x0000ffff86aa6124 in __assert_fail_base () from /lib64/libc.so.6
#3 0x0000ffff86aa61a4 in __assert_fail () from /lib64/libc.so.6
#4 0x0000000000529118 in virtio_scsi_ctx_check (d=<optimized out>, s=<optimized out>, s=<optimized out>) at /home/qemu-4.0.0/hw/scsi/virtio-scsi.c:246
#5 0x0000000000529ec4 in virtio_scsi_handle_cmd_req_prepare (s=0x2779ec00, req=0xffff740397d0) at /home/qemu-4.0.0/hw/scsi/virtio-scsi.c:559
#6 0x000000000052a228 in virtio_scsi_handle_cmd_vq (s=0x2779ec00, vq=0xffff7c6d7110) at /home/qemu-4.0.0/hw/scsi/virtio-scsi.c:603
#7 0x000000000052afa8 in virtio_scsi_data_plane_handle_cmd (vdev=<optimized out>, vq=0xffff7c6d7110) at /home/qemu-4.0.0/hw/scsi/virtio-scsi-dataplane.c:59
#8 0x000000000054d94c in virtio_queue_host_notifier_aio_poll (opaque=<optimized out>) at /home/qemu-4.0.0/hw/virtio/virtio.c:2452
assert(blk_get_aio_context(d->conf.blk) == s->ctx) failed.
To avoid assertion failed, moving the "if" after qdev_simple_device_unplug_cb.
In addition, to avoid another qemu crash below, add aio_disable_external before
qdev_simple_device_unplug_cb, which disable the further processing of external clients
when doing qdev_simple_device_unplug_cb.
(gdb) bt
#0 scsi_req_unref (req=0xffff6802c6f0) at hw/scsi/scsi-bus.c:1283
#1 0x00000000005294a4 in virtio_scsi_handle_cmd_req_submit (req=<optimized out>,
s=<optimized out>) at /home/qemu-4.0.0/hw/scsi/virtio-scsi.c:589
#2 0x000000000052a2a8 in virtio_scsi_handle_cmd_vq (s=s@entry=0x9c90e90,
vq=vq@entry=0xffff7c05f110) at /home/qemu-4.0.0/hw/scsi/virtio-scsi.c:625
#3 0x000000000052afd8 in virtio_scsi_data_plane_handle_cmd (vdev=<optimized out>,
vq=0xffff7c05f110) at /home/qemu-4.0.0/hw/scsi/virtio-scsi-dataplane.c:60
#4 0x000000000054d97c in virtio_queue_host_notifier_aio_poll (opaque=<optimized out>)
at /home/qemu-4.0.0/hw/virtio/virtio.c:2447
#5 0x00000000009b204c in run_poll_handlers_once (ctx=ctx@entry=0x6efea40,
timeout=timeout@entry=0xffff7d7f7308) at util/aio-posix.c:521
#6 0x00000000009b2b64 in run_poll_handlers (ctx=ctx@entry=0x6efea40,
max_ns=max_ns@entry=4000, timeout=timeout@entry=0xffff7d7f7308) at util/aio-posix.c:559
#7 0x00000000009b2ca0 in try_poll_mode (ctx=ctx@entry=0x6efea40, timeout=0xffff7d7f7308,
timeout@entry=0xffff7d7f7348) at util/aio-posix.c:594
#8 0x00000000009b31b8 in aio_poll (ctx=0x6efea40, blocking=blocking@entry=true)
at util/aio-posix.c:636
#9 0x00000000006973cc in iothread_run (opaque=0x6ebd800) at iothread.c:75
#10 0x00000000009b592c in qemu_thread_start (args=0x6efef60) at util/qemu-thread-posix.c:502
#11 0x0000ffff8057f8bc in start_thread () from /lib64/libpthread.so.0
#12 0x0000ffff804e5f8c in thread_start () from /lib64/libc.so.6
(gdb) p bus
$1 = (SCSIBus *) 0x0
Signed-off-by: Zhengui li <lizhengui@huawei.com>
Message-Id: <1563696502-7972-1-git-send-email-lizhengui@huawei.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <1563829520-17525-1-git-send-email-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The argument is not used and passing it clutters error propagation in the
callers. So, get rid of it.
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Rather than looking inside the definition of a BusState with "s->bus.qbus",
use the QOM prefered style: "BUS(&s->bus)".
This patch was generated using the following Coccinelle script:
// Use BUS() macros to access BusState.qbus
@use_bus_macro_to_access_qbus@
expression obj;
identifier bus;
@@
-&obj->bus.qbus
+BUS(&obj->bus)
Suggested-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20190528164020.32250-3-philmd@redhat.com>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
This makes use of qdev_prop_drive_iothread for scsi-disk so that the
disk can be attached to a node that is already in the target AioContext.
We need to check that the HBA actually supports iothreads, otherwise
scsi-disk must make sure that the node is already in the main
AioContext.
This changes the error message for conflicting iothread settings.
Previously, virtio-scsi produced the error message, now it comes from
blk_set_aio_context(). Update a test case accordingly.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Add an Error parameter to blk_set_aio_context() and use
bdrv_child_try_set_aio_context() internally to check whether all
involved nodes can actually support the AioContext switch.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Build fails with gcc 9:
CC ppc64-softmmu/hw/scsi/virtio-scsi.o
hw/scsi/virtio-scsi.c: In function ‘virtio_scsi_do_tmf’:
hw/scsi/virtio-scsi.c:265:39: error: taking address of packed member of ‘struct virtio_scsi_ctrl_tmf_req’ may result in an unaligned pointer value [-Werror=address-of-packed-member]
265 | virtio_tswap32s(VIRTIO_DEVICE(s), &req->req.tmf.subtype);
| ^~~~~~~~~~~~~~~~~~~~~
cc1: all warnings being treated as errors
All the fields in struct virtio_scsi_ctrl_tmf_req are naturally aligned,
so we could in theory drop QEMU_PACKED. Unfortunately, the header file
is imported from linux which already has the packed attribute. Trying to
fix that in the update-linux-headers.sh script is likely to produce
ugliness. Turn the call to virtio_tswap32s() into an assignment instead.
Signed-off-by: Greg Kurz <groug@kaod.org>
Message-Id: <155137678223.44753.5438092367451176318.stgit@bahia.lan>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Certain devices types, like memory/CPU, are now being handled using a
hotplug interface provided by a top-level MachineClass. Hotpluggable
host bridges are another such device where it makes sense to use a
machine-level hotplug handler. However, unlike those devices,
host-bridges have a parent bus (the main system bus), and devices with
a parent bus use a different mechanism for registering their hotplug
handlers: qbus_set_hotplug_handler(). This interface currently expects
a handler to be a subclass of DeviceClass, but this is not the case
for MachineClass, which derives directly from ObjectClass.
Internally, the interface only requires an ObjectClass, so expose that
in qbus_set_hotplug_handler().
Cc: Michael S. Tsirkin <mst@redhat.com>
Cc: Eduardo Habkost <ehabkost@redhat.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Signed-off-by: Greg Kurz <groug@kaod.org>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Cornelia Huck <cohuck@redhat.com>
Acked-by: Halil Pasic <pasic@linux.ibm.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Message-Id: <154999589921.690774.3640149277362188566.stgit@bahia.lan>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
This patch forbids attaching a disk to a SCSI device if its using a
different AioContext. Test case included.
Signed-off-by: Alberto Garcia <berto@igalia.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This fixes a crash when attaching a disk to a SCSI device using
iothreads, then detaching it and reattaching it again. Test case
included.
Signed-off-by: Alberto Garcia <berto@igalia.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
When [2] was fixed it was agreed that adding and calling post_plug()
callback after device_reset() was low risk approach to hotfix issue
right before release. So it was merged instead of moving already
existing plug() callback after device_reset() is called which would
be more risky and require all plug() callbacks audit.
Looking at the current plug() callbacks, it doesn't seem that moving
plug() callback after device_reset() is breaking anything, so here
goes agreed upon [3] proper fix which essentially reverts [1][2]
and moves plug() callback after device_reset().
This way devices always comes to plug() stage, after it's been fully
initialized (including being reset), which fixes race condition [2]
without need for an extra post_plug() callback.
1. (25e897881 "qdev: add HotplugHandler->post_plug() callback")
2. (8449bcf94 "virtio-scsi: fix hotplug ->reset() vs event race")
3. https://www.mail-archive.com/qemu-devel@nongnu.org/msg549915.html
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <1539696820-273275-1-git-send-email-imammedo@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Tested-by: Pierre Morel<pmorel@linux.ibm.com>
Acked-by: Pierre Morel<pmorel@linux.ibm.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
There is a race condition during hotplug when iothread is used. It
occurs because virtio-scsi may be processing command queues in the
iothread while the monitor performs SCSI device hotplug.
When a SCSI device is hotplugged the HotplugHandler->plug() callback is
invoked and virtio-scsi emits a rescan event to the guest.
If the guest submits a SCSI command at this point then it may be
cancelled before hotplug completes. This happens because ->reset() is
called by hw/core/qdev.c:device_set_realized() after
HotplugHandler->plug() has been called and
hw/scsi/scsi-disk.c:scsi_disk_reset() purges all requests.
This patch uses the new HotplugHandler->post_plug() callback to emit the
rescan event after ->reset(). This eliminates the race conditions where
requests could be cancelled.
Reported-by: l00284672 <lizhengui@huawei.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Fam Zheng <famz@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20180716083732.3347-3-stefanha@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Complete the transition by renaming this header, which was
shared by block/iscsi.c and the SCSI emulation code.
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Since Linux switched to blk-mq as the default in Linux commit
5c279bd9e406 ("scsi: default to scsi-mq"), virtio-scsi LUNs consume
about 10x as much guest kernel memory.
This commit allows you to choose the virtqueue size for each
virtio-scsi-pci controller like this:
-device virtio-scsi-pci,id=scsi,virtqueue_size=16
The default is still 128 as before. Using smaller virtqueue_size
allows many more disks to be added to small memory virtual machines.
For a 1 vCPU, 500 MB, no swap VM I observed:
With scsi-mq enabled (upstream kernel): 175 disks
-"- ditto -"- virtqueue_size=64: 318 disks
-"- ditto -"- virtqueue_size=16: 775 disks
With scsi-mq disabled (kernel before 5c279bd9e406): 1755 disks
Note that to have any effect, this requires a kernel patch:
https://lkml.org/lkml/2017/8/10/689
Signed-off-by: Richard W.M. Jones <rjones@redhat.com>
Message-Id: <20170810165255.20865-1-rjones@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
