The request restart mechanism is generic and could be reused for
scsi-generic. In the meanwhile, pushing it to SCSIDevice avoids
that scsi_dma_restart_bh looks at SCSIGenericReqs when working on
a scsi-block device.
The code is the same that is already in hw/scsi-disk.c, with
the type flags replaced by req->cmd.mode and a more generic way to
requeue SCSI_XFER_NONE commands.
I also added a missing call to qemu_del_vm_change_state_handler.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Otherwise, if cancellation is "faked" by the AIO layer and goes
through qemu_aio_flush, the whole request is completed synchronously
during scsi_req_cancel.
Using the enqueued flag would work here, but not in the next patches,
so I'm introducing a new io_canceled flag. That's because scsi_req_data
is a synchronous callback and the enqueued flag might be reset by the
time it returns. scsi-disk cannot unref the request until after calling
scsi_req_data.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This will let scsi-block choose between passthrough and emulation.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Also delete a stale occurrence of SCSIReqOps inside SCSIDeviceInfo.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The field is only in scsi-disk for now. Moving it up to SCSIDevice makes
it easier to reuse the scsi-generic reqops elsewhere.
At the same time, make scsi-generic get max_lba from snooped READ CAPACITY
commands as well.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This also requires little more than adding the new argument to
scsi_device_find, and the qdev property. All devices by default
end up on channel 0.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This only requires changes in two places: in SCSIBus, we need to look
for a free LUN if somebody creates a device with a pre-existing scsi-id
but the default LUN (-1, meaning "search for a free spot"); in vSCSI,
we need to actually parse the LUN according to the SCSI spec.
For vSCSI, max_target/max_lun are set according to the logical unit
addressing format in SAM.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Change the devs array into a linked list, and add a scsi_device_find
function to navigate the children list instead. This lets the SCSI
bus use more complex addressing, and HBAs can talk to the correct device
when there are multiple LUNs per target.
scsi_device_find may return another LUN on the same target if none is
found that matches exactly.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Building on the previous patch, this one adds a media change callback
to scsi-disk.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reporting media change events via unit attention sense codes requires
a small state machine: first report "NO MEDIUM", then report "MEDIUM MAY
HAVE CHANGED". Unfortunately there is no good hooking point for the
device to notice that its pending unit attention condition has been
reported. This patch reworks the generic machinery to add one.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
It's convenience stuff for block device models, so block.h isn't the
ideal home either, but better than block_int.h.
Permits moving some #include "block_int.h" from device model .h into
.c.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Unit attention conditions override any sense data the device already
has. Their signaling and clearing is handled entirely by the SCSIBus
code, and they are completely transparent to the SCSIDevices.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Also introduce the first occurrence of "independent" SCSIReqOps,
to handle invalid commands in common code.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
This will let SCSIBus detect requests sent to an invalid LUN, and
handle them itself. However, there will be still support for only one
LUN per target
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
This struct is currently unnamed. Give it a name and use it
explicitly to decouple (some parts of) CDB parsing from
SCSIRequest.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Right now the CDB is not passed to the SCSIBus until scsi_req_enqueue.
Passing it to scsi_req_new will let scsi_req_new dispatch common requests
through different reqops.
Moving the memcpy to scsi_req_new is a hack that will go away as
soon as scsi_req_new will also take care of the parsing.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
This will let allow requests to be dispatched through different callbacks,
either common or per-device.
This patch adjusts the API, the next one will move members to SCSIReqOps.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
With this patch, sense data is stored in the generic data structures
for SCSI devices and requests. The SCSI layer takes care of storing
sense data in the SCSIDevice for the subsequent REQUEST SENSE command.
At the same time, get_sense is removed and scsi_req_get_sense can use
an entirely generic implementation.
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
A small improvement in the SCSI request API. Pass the status
at the time the request is completed, so that we can assert that
no request is completed twice. This would have detected the
problem fixed in the previous patch.
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
'tag' is just an abstraction to identify the command
from the driver. So we should make that explicit by
replacing 'tag' with a driver-defined pointer 'hba_private'.
This saves the lookup for driver handling several commands
in parallel.
'tag' is still being kept for tracing purposes.
Signed-off-by: Hannes Reinecke <hare@suse.de>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The get_sense callback copies existing sense information into
the provided buffer. This is required if sense information
should be transferred together with the command response.
Signed-off-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
... and remove some SCSIDevice variables or fields that now become unused.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Move the common part of scsi-disk.c and scsi-generic.c to the SCSI layer.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
The SCSI spec has a quite detailed list of sense codes available.
It even mandates the use of specific ones for some failure cases.
The current implementation just has one type of generic error
which is actually a violation of the spec in certain cases.
This patch introduces various predefined sense codes to have the
sense code reporting more in line with the spec.
On top of Hannes's patch I fixed the reply to REQUEST SENSE commands
with DESC=0 and a small (<18) length.
Signed-off-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
This is for when the request must be dropped in the void,
but still memory should be freed. To this end, the devices
register a second callback in SCSIBusOps.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
This covers the case of canceling a request's I/O and still
completing it.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
The code for canceling requests upon reset is already the same. Clean
it up and move it to scsi-bus.c.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Currently the SCSIRequest structure is abstracted away and cannot accessed
directly from the driver. This requires the handler to do a lookup on
an abstract 'tag' which identifies the SCSIRequest structure.
With this patch the SCSIRequest structure is exposed to the driver. This
allows use to use it directly as an argument to the SCSIDeviceInfo
callback functions and remove the lookup.
A new callback function 'alloc_req' is introduced matching 'free
req'; unref'ing to free up resources after use is moved into the
scsi_command_complete callbacks.
This temporarily introduces a leak of requests that are cancelled,
when they are removed from the queue and not from the driver. This
is fixed later by introducing scsi_req_cancel. That patch in turn
depends on this one, because the argument to scsi_req_cancel is a
SCSIRequest.
Signed-off-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
With the next patch, a device may hold SCSIRequest for an indefinite
time. Split a rather big patch, and protect against access errors,
by reference counting them.
There is some ugliness in scsi_send_command implementation due to
the need to unref the request when it fails. This will go away
with the next patches, which move the unref'ing to the devices.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
There are more operations than a SCSI bus can handle, besides completing
commands. One example, which this series will introduce, is cleaning up
after a request is cancelled.
More long term, a "SCSI bus" can represent the LUNs attached to a
target; in this case, while all commands will ultimately reach a logical
unit, it is the target who is in charge of answering REPORT LUNs.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
This abstracts calling the command_complete callback, reducing churn
in the following patches.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Before commit 622b520f, index=12 meant bus=1,unit=5.
Since the commit, it means bus=0,unit=12. The drive is created, but
not the guest device. That's because the controllers we use with
if=scsi drives (lsi53c895a and esp) support only 7 units, and
scsi_bus_legacy_handle_cmdline() ignores drives with unit numbers
exceeding that limit.
Changing the mapping of index to bus, unit is a regression. Breaking
-drive invocations that used to work just makes it worse.
Revert the part of commit 622b520f that causes this, and clean up
some.
Note that the fix only affects if=scsi. You can still put more than 7
units on a SCSI bus with -device & friends.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
scsi-disk devices may wish to override the removable bit. Add support
for a qdev property on SCSI devices. This is will be used by usb-msd.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The current sense handling in scsi-bus is only used by the
scsi-disk driver; the scsi-generic driver is using its own.
So we should move the current sense handling into the
scsi-disk driver.
Signed-off-by: Hannes Reinecke <hare@suse.de>
Acked-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The SCSI parallel interface has a limit of 8 devices, but
not the SCSI stack in general. So we should be removing the
hard-coded limit and use MAX_SCSI_DEVS instead.
And we only need to scan those devices which are allocated
by the bus.
Signed-off-by: Hannes Reinecke <hare@suse.de>
Acked-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The request completion callback of the LSI controller may start the next
request that can use the same tag as the completed one. As the latter is
still enqueued at that point, scsi_send_command will complain about the
tag reuse and cancel the completed request. That will cause a double
free later on when the completion path cleans up as well.
Fix this by dequeuing the request before invoking the callback.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
Make the property point to BlockDriverState, cutting out the DriveInfo
middleman. This prepares the ground for block devices that don't have
a DriveInfo.
Currently all user-defined ones have a DriveInfo, because the only way
to define one is -drive & friends (they go through drive_init()).
DriveInfo is closely tied to -drive, and like -drive, it mixes
information about host and guest part of the block device. I'm
working towards a new way to define block devices, with clean
host/guest separation, and I need to get DriveInfo out of the way for
that.
Fortunately, the device models are perfectly happy with
BlockDriverState, except for two places: ide_drive_initfn() and
scsi_disk_initfn() need to check the DriveInfo for a serial number set
with legacy -drive serial=... Use drive_get_by_blockdev() there.
Device model code should now use DriveInfo only when explicitly
dealing with drives defined the old way, i.e. without -device.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
None of its callers checks for failure. scsi_hot_add() can crash
because of that:
(qemu) drive_add 4 if=scsi,format=host_device,file=/dev/sg1
scsi-generic: scsi generic interface too old
Segmentation fault (core dumped)
Fix all callers, not just scsi_hot_add().
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Add three new qdev properties to export block topology information to
the guest. This is needed to get optimal I/O alignment for RAID arrays
or SSDs.
The options are:
- physical_block_size to specify the physical block size of the device,
this is going to increase from 512 bytes to 4096 kilobytes for many
modern storage devices
- min_io_size to specify the minimal I/O size without performance impact,
this is typically set to the RAID chunk size for arrays.
- opt_io_size to specify the optimal sustained I/O size, this is
typically the RAID stripe width for arrays.
I decided to not auto-probe these values from blkid which might easily
be possible as I don't know how to deal with these issues on migration.
Note that we specificly only set the physical_block_size, and not the
logial one which is the unit all I/O is described in. The reason for
that is that IDE does not support increasing the logical block size and
at last for now I want to stick to one meachnisms in queue and allow
for easy switching of transports for a given backing image which would
not be possible if scsi and virtio use real 4k sectors, while ide only
uses the physical block exponent.
To make this more common for the different block drivers introduce a
new BlockConf structure holding all common block properties and a
DEFINE_BLOCK_PROPERTIES macro to add them all together, mirroring
what is done for network drivers. Also switch over all block drivers
to use it, except for the floppy driver which has weird driveA/driveB
properties and probably won't require any advanced block options ever.
Example usage for a virtio device with 4k physical block size and
8k optimal I/O size:
-drive file=scratch.img,media=disk,cache=none,id=scratch \
-device virtio-blk-pci,drive=scratch,physical_block_size=4096,opt_io_size=8192
aliguori: updated patch to take into account BLOCK events
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
