We need this to protect dataplane thread from race conditions with block
jobs until the latter is made dataplane-safe.
Signed-off-by: NFam Zheng <famz@redhat.com>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

blockdev_init() always creates a DriveInfo, but only drive_new() fills
it in.  qmp_blockdev_add() leaves it blank.  This results in a drive
with type = IF_IDE, bus = 0, unit = 0.  Screwed up in commit ee13ed1c.

Board initialization code looking for IDE drive (0,0) can pick up one
of these bogus drives.  The QMP command has to execute really early to
be visible.  Not sure how likely that is in practice.

Fix by creating DriveInfo in drive_new().  Block backends created by
blockdev-add don't get one.

Breaks the test for "has been created by qmp_blockdev_add()" in
blockdev_mark_auto_del() and do_drive_del(), because it changes the
value of dinfo && !dinfo->enable_auto_del from true to false.  Simply
test !dinfo instead.

Leaves DriveInfo member enable_auto_del unused.  Drop it.

A few places assume a block backend always has a DriveInfo.  Fix them
up.
Signed-off-by: NMarkus Armbruster <armbru@redhat.com>
Reviewed-by: NMax Reitz <mreitz@redhat.com>
Signed-off-by: NKevin Wolf <kwolf@redhat.com>

Device models should access their block backends only through the
block-backend.h API.  Convert them, and drop direct includes of
inappropriate headers.

Just four uses of BlockDriverState are left:

* The Xen paravirtual block device backend (xen_disk.c) opens images
  itself when set up via xenbus, bypassing blockdev.c.  I figure it
  should go through qmp_blockdev_add() instead.

* Device model "usb-storage" prompts for keys.  No other device model
  does, and this one probably shouldn't do it, either.

* ide_issue_trim_cb() uses bdrv_aio_discard() instead of
  blk_aio_discard() because it fishes its backend out of a BlockAIOCB,
  which has only the BlockDriverState.

* PC87312State has an unused BlockDriverState[] member.

The next two commits take care of the latter two.
Signed-off-by: NMarkus Armbruster <armbru@redhat.com>
Reviewed-by: NMax Reitz <mreitz@redhat.com>
Signed-off-by: NKevin Wolf <kwolf@redhat.com>

I'll use it with block backends shortly, and the name is going to fit
badly there.  It's a block layer thing anyway, not just a block driver
thing.
Signed-off-by: NMarkus Armbruster <armbru@redhat.com>
Reviewed-by: NMax Reitz <mreitz@redhat.com>
Signed-off-by: NKevin Wolf <kwolf@redhat.com>

The patch is big, but all it really does is replacing

    dinfo->bdrv

by

    blk_bs(blk_by_legacy_dinfo(dinfo))

The replacement is repetitive, but the conversion of device models to
BlockBackend is imminent, and will shorten it to just
blk_legacy_dinfo(dinfo).

Line wrapping muddies the waters a bit.  I also omit tests whether
dinfo->bdrv is null, because it never is.
Signed-off-by: NMarkus Armbruster <armbru@redhat.com>
Reviewed-by: NBenoît Canet <benoit.canet@nodalink.com>
Reviewed-by: NMax Reitz <mreitz@redhat.com>
Reviewed-by: NKevin Wolf <kwolf@redhat.com>
Signed-off-by: NKevin Wolf <kwolf@redhat.com>

On this way, we can assure the new bootindex take effect
during vm rebooting.
Signed-off-by: NGonglei <arei.gonglei@huawei.com>
Reviewed-by: NGerd Hoffmann <kraxel@redhat.com>
Signed-off-by: NGerd Hoffmann <kraxel@redhat.com>

Remove bootindex form qdev property to qom, things will
continue to work just fine, and we can use qom features
which are not supported by qdev property.
Signed-off-by: NGonglei <arei.gonglei@huawei.com>
Reviewed-by: NGerd Hoffmann <kraxel@redhat.com>
Signed-off-by: NGerd Hoffmann <kraxel@redhat.com>

Add a qom property with the same name 'bootindex',
when we remove it form qdev property, things will
continue to work just fine, and we can use qom features
which are not supported by qdev property.
Signed-off-by: NGonglei <arei.gonglei@huawei.com>
Reviewed-by: NGerd Hoffmann <kraxel@redhat.com>
Signed-off-by: NGerd Hoffmann <kraxel@redhat.com>

SCSI subsytem was converted to hotplug handler API and
doesn't use SCSIBusInfo{hotplug, hot_unplug} fields and
related callbacks anymore.
Signed-off-by: NIgor Mammedov <imammedo@redhat.com>
Reviewed-by: NPaolo Bonzini <pbonzini@redhat.com>
Signed-off-by: NAndreas Färber <afaerber@suse.de>

Signed-off-by: NIgor Mammedov <imammedo@redhat.com>
Reviewed-by: NPaolo Bonzini <pbonzini@redhat.com>
Reviewed-by: NMichael S. Tsirkin <mst@redhat.com>
Signed-off-by: NAndreas Färber <afaerber@suse.de>

Signed-off-by: NIgor Mammedov <imammedo@redhat.com>
Reviewed-by: NPaolo Bonzini <pbonzini@redhat.com>
Signed-off-by: NAndreas Färber <afaerber@suse.de>

That would allow to handle SCSI device unplug
on HBAs without dedicated hot(un)plug handlers
and avoid making such HBAs explicitly hotpluggable.
Signed-off-by: NIgor Mammedov <imammedo@redhat.com>
Reviewed-by: NPaolo Bonzini <pbonzini@redhat.com>
Signed-off-by: NAndreas Färber <afaerber@suse.de>

scsi_req_continue can complete the request and cause the VirtIOSCSIReq
to be freed.  Fetch req->sreq just once to avoid the bug.
Reported-by: NRichard Jones <rjones@redhat.com>
Tested-by: NRichard Jones <rjones@redhat.com>
Reviewed-by: NFam Zheng <famz@redhat.com>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

For VIRTIO_SCSI_T_TMF_ABORT_TASK and VIRTIO_SCSI_T_TMF_ABORT_TASK_SET,
use scsi_req_cancel_async to start the cancellation.

Because each tmf command may cancel multiple requests, we need to use a
counter to track the number of remaining requests we still need to wait
for.
Signed-off-by: NFam Zheng <famz@redhat.com>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

Devices will call this function to start an asynchronous cancellation. The
bus->info->cancel will be called after the request is canceled.

Devices will probably need to track a separate TMF request that triggers this
cancellation, and wait until the cancellation is done before completing it. So
we store a notifier list in SCSIRequest and in scsi_req_cancel_complete we
notify them.
Signed-off-by: NFam Zheng <famz@redhat.com>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

Let the aio cb do the clean up and notification job after scsi_req_cancel, in
preparation for asynchronous cancellation.
Signed-off-by: NFam Zheng <famz@redhat.com>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

The only two implementations are identical to each other, with nothing specific
to device: they only call bdrv_aio_cancel with the SCSIRequest.aiocb.

Let's move it to scsi-bus.
Signed-off-by: NFam Zheng <famz@redhat.com>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

Before, scsi_req_cancel will take ownership of the canceled request and unref
it. We did this because we didn't know whether AIO CB will be called or not
during the cancelling, so we set the io_canceled flag before calling it, and
skip unref in the potentially called callbacks, which is not very nice.

Now, bdrv_aio_cancel has a stricter contract that the completion callbacks are
always called, so we can remove the checks of req->io_canceled and just unref
it in callbacks.

It will also make implementing asynchronous cancellation easier.
Signed-off-by: NFam Zheng <famz@redhat.com>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

Now that we always called the cb in bdrv_aio_cancel, let's make scsi-generic
callbacks check io_canceled flag similarly to scsi-disk.
Signed-off-by: NFam Zheng <famz@redhat.com>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

The only user of this function is spapr_vscsi.c. We can convert to
scsi_req_cancel plus adding a check in vscsi_request_cancelled.
Suggested-by: NPaolo Bonzini <pbonzini@redhat.com>
Signed-off-by: NFam Zheng <famz@redhat.com>
[Drop prototype. - Paolo]
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

We are ready, now let's effectively enable dataplane.
Signed-off-by: NFam Zheng <famz@redhat.com>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

Signed-off-by: NFam Zheng <famz@redhat.com>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

Queue the popped requests while calling
virtio_scsi_handle_cmd_req_prepare(), then submit them after all
prepared.
Signed-off-by: NFam Zheng <famz@redhat.com>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

Mechanical change, in preparation for bdrv_io_plug/bdrv_io_unplug.
Signed-off-by: NFam Zheng <famz@redhat.com>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

Similar to virtio-blk-dataplane, we stop the iothread while migration
starts and restart it when migration finishes.
Signed-off-by: NFam Zheng <famz@redhat.com>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

This enables the virtio-scsi-dataplane code by setting the iothread
in virtio-scsi device, and makes any function that is called by
back from dataplane to cooperate with the caller: they need to be
vring/iothread aware when handling the requests and using scsi devices
on the bus.
Signed-off-by: NFam Zheng <famz@redhat.com>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

This implements the core part of dataplane feature of virtio-scsi.

A few fields are added in VirtIOSCSICommon to maintain the dataplane
status. These fields are managed by a new source file:
virtio-scsi-dataplane.c.

Most code in this file will run on an iothread, unless otherwise
commented as in a global mutex context, such as those functions to
start, stop and setting the iothread property.

Upon start, we set up guest/host event notifiers, in a same way as
virtio-blk does. The handlers then pop request from vring and call into
virtio-scsi.c functions to process it. So we need to make sure make all
those called functions work with iothread, too.
Signed-off-by: NFam Zheng <famz@redhat.com>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

Similar to this property in virtio-blk for dataplane, add it as a QOM
link in virtio-scsi and an alias in virtio-scsi-pci and virtio-scsi-ccw,
in order to assign an iothread to the device.
Signed-off-by: NFam Zheng <famz@redhat.com>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

Later this will be called by dataplane code.
Signed-off-by: NFam Zheng <famz@redhat.com>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

To share with dataplane code later.
Signed-off-by: NFam Zheng <famz@redhat.com>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

To share with datplane code later.
Signed-off-by: NFam Zheng <famz@redhat.com>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

To share with dataplane code.
Signed-off-by: NFam Zheng <famz@redhat.com>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

This is the "common part" to handle one cmd request. Refactor out for
later usage of dataplane iothread code.
Signed-off-by: NFam Zheng <famz@redhat.com>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

The command direction according to the guest-passed buffers
is already stored in the VirtIOSCSIReq.  We can use it instead
of computing it again from req->elem.

Cc: Laszlo Ersek <lersek@redhat.com>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

This helps for cross-endian configurations.

Cc: qemu-stable@nongnu.org
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

The VirtQueueElement is a very big structure (>48k!), since it will be
initialzed by virtqueue_pop, we can save the expensive zeroing here.

This saves a few microseconds per request in my test:

[fio-test]      rw         bs         iodepth    jobs       bw         iops       latency
--------------------------------------------------------------------------------------------
Before          read       4k         1          1          110        28269      34
After           read       4k         1          1          131        33745      28

Whereas,

virtio-blk      read       4k         1          1          217        55673      16
Signed-off-by: NFam Zheng <famz@redhat.com>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

Zeroing sense buffer for each scsi request is not efficient, we can just
leave it uninitialized because sense_len is set to 0.

Move the implicitly zeroed fields to the end of the structure and use a
partial memset.

The explicitly initialized fields (by scsi_req_alloc or scsi_req_new)
are moved to the beginning of the structure, before sense buffer, to
skip the memset.

Also change g_malloc0 to g_slice_alloc.
Signed-off-by: NFam Zheng <famz@redhat.com>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

This is the next step for decoupling block accounting functions from
BlockDriverState.
In a future commit the BlockAcctStats structure will be moved from
BlockDriverState to the device models structures.

Note that bdrv_get_stats was introduced so device models can retrieve the
BlockAcctStats structure of a BlockDriverState without being aware of it's
layout.
This function should go away when BlockAcctStats will be embedded in the device
models structures.

CC: Kevin Wolf <kwolf@redhat.com>
CC: Stefan Hajnoczi <stefanha@redhat.com>
CC: Keith Busch <keith.busch@intel.com>
CC: Anthony Liguori <aliguori@amazon.com>
CC: "Michael S. Tsirkin" <mst@redhat.com>
CC: Paolo Bonzini <pbonzini@redhat.com>
CC: Eric Blake <eblake@redhat.com>
CC: Peter Maydell <peter.maydell@linaro.org>
CC: Michael Tokarev <mjt@tls.msk.ru>
CC: John Snow <jsnow@redhat.com>
CC: Markus Armbruster <armbru@redhat.com>
CC: Alexander Graf <agraf@suse.de>
CC: Max Reitz <mreitz@redhat.com>
Signed-off-by: NBenoît Canet <benoit.canet@nodalink.com>
Signed-off-by: NKevin Wolf <kwolf@redhat.com>

The middle term goal is to move the BlockAcctStats structure in the device models.
(Capturing I/O accounting statistics in the device models is good for billing)
This patch make a small step in this direction by removing a reference to BDRV.

CC: Kevin Wolf <kwolf@redhat.com>
CC: Stefan Hajnoczi <stefanha@redhat.com>
CC: Keith Busch <keith.busch@intel.com>
CC: Anthony Liguori <aliguori@amazon.com>
CC: "Michael S. Tsirkin" <mst@redhat.com>
CC: Paolo Bonzini <pbonzini@redhat.com>
CC: John Snow <jsnow@redhat.com>
CC: Richard Henderson <rth@twiddle.net>
CC: Markus Armbruster <armbru@redhat.com>
CC: Alexander Graf <agraf@suse.de>i
Signed-off-by: NBenoît Canet <benoit.canet@nodalink.com>
Signed-off-by: NKevin Wolf <kwolf@redhat.com>

As vhost core can use backend_features during init, clear it earlier to
avoid using uninitialized memory.
This use would be harmless since vhost scsi ignores the result
anyway, but initializing earlier will help prevent valgrind errors,
and make scsi and net behave similarly.

Cc: qemu-stable@nongnu.org
Acked-by: NPaolo Bonzini <pbonzini@redhat.com>
Acked-by: NJason Wang <jasowang@redhat.com>
Signed-off-by: NMichael S. Tsirkin <mst@redhat.com>