Leaving the aiocb to a non-NULL value leads to an assertion failure when
rerror/werror are set to stop or enospc, and the operation is retried.
scsi-disk checks that the aiocb member is NULL before filling it.

This patch correctly resets the aiocb to NULL values everywhere,
and adds the dual assertion that the aiocb was non-NULL before
calling bdrv_acct_done.
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

Reported-by: NGerhard Wiesinger <lists@wiesinger.com>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

The START STOP UNIT command will only eject/load media if
power condition is zero.

If power condition is !0 then LOEJ and START will be ignored.

From MMC (sbc contains similar wordings too)
  The Power Conditions field requests the block device to be placed
  in the power condition defined in
  Table 558. If this field has a value other than 0h then the Start
  and LoEj bits shall be ignored.
Signed-off-by: NRonnie Sahlberg <ronniesahlberg@gmail.com>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

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

Linux will not use these, but a very similar mechanism will be used to
report the condition via virtio-scsi events.
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

When a device is resized, we will report a unit attention condition
for CAPACITY DATA HAS CHANGED.  However, we should ensure that this
condition does not override a more important unit attention condition.
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

Support for the LOEJ bit of the START/STOP UNIT command right now is
limited to CD-ROMs.  This is wrong, since removable hard disks (in the
real world: SD card readers) also support it in pretty much the same way.

Without the LOEJ bit, START/STOP UNIT does nothing for all devices.
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

Add sense code for DATA_PROTECT/WRITE_PROTECTED and return this error
for any WRITE*/WRITE_VERIFY* calls if the device is readonly=on,
i.e. write-protected
Signed-off-by: NRonnie Sahlberg <ronniesahlberg@gmail.com>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

Improve the tests for the LBA to cover more cases.

For the 16 byte opcodes, the lba is a uint64, so we need to check is to
make sure that we do not wrap.  For example if an opcode would specify
the LBA:0xffffffffffffffff and LEN:2 then lba+len would wrap to 1.

Also verify that ALL requested blocks are available, not just the first one.
Signed-off-by: NRonnie Sahlberg <ronniesahlberg@gmail.com>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

The QEMU SCSI emulation does not support protection information,
so any READ/WRITE/VERIFY commands that has the protect bits set to
non-zero should fail with ILLEGAL_REQUEST/INVALID_FIELD_IN_CDB

From SCSI SBC :
If the logical unit does not support protection information,
then the device server should terminate the command with CHECK CONDITION
status with the sense key set to ILLEGAL REQUEST and the additional sense
code set to INVALID FIELD IN CDB.
Signed-off-by: NRonnie Sahlberg <ronniesahlberg@gmail.com>
[ Rebase after scsi_dma_reqops introduction - Paolo ]
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

Finally, this uses the "plumbing" in the previous patch to
add support for toggling the WCE bit of the caching mode page.
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

This adds the bulk of the parsing code for MODE SELECT, including
breaking out changes to different mode pages, and checking that only
changeable values are modified.

In order to report errors correctly two passes are made through the
parameters; the first only looks for errors, the second actually
applies the changes to the mode page.
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

The changeable values were not all-zeros for this mode page, fix it.
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

This will make offsets the same when implementing MODE SELECT.  This is
because MODE SELECT has to deal with both 2-byte and 4-byte headers.
Unfortunately, this means that the offsets are now off by two compared
to the descriptions in the SCSI specs, which include the header.
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

This adds the implementation of write_data for the emulated
command case.  The first time through it asks for more data,
the second time it finishes the processing of the command.

MODE SELECT and MODE SELECT(10) can now be re-enabled, but they
will not do much.
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

The previous patch only separated the send_command callback.
Use different implementations also for read_data and write_data.
The latter is still unreachable, so it aborts for now.

read_data passes the data buffer that was prepared and completes
the command.
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

Only checks for present medium were still done in scsi_send_command
for emulated commands.  So move those to scsi_disk_emulate_command
and return different SCSIReqOps depending on the kind of command.

Checks for present medium can be done unconditionally for the
scsi_disk_dma_reqops case.
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

We want to use separate SCSIReqOps for emulated commands needing an
allocated buffer vs. those that are zerocopy when the HBA supports
S/G lists.  Ensure that all of the former are in scsi_disk_emulate_command.

Commands that do not have any parameters are more similar to emulated
commands, so also move them, even if they do I/O.

Finally, MODE SELECT and MODE SELECT(10) are broken because we do not
yet support passing parameter data _to_ emulated commands, so disable
them.
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

By making discard asynchronous, we can reuse all the error handling
code that is used for other commands.
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

This patch adds two new properties vendor and product to SCSI disks.
These options let the user customize the inquiry data returned by the
disk.
Signed-off-by: NYan Vugenfirer <yan@ravellosystems.com>
Signed-off-by: NDmitry Fleytman <dmitry.fleytman@ravellosystems.com>
[ Use vendor and product property names, avoid "if" statements. - PB ]
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

scsi-block is a passthrough device and does not allow customization
of vendor, product, removable, DPOFUA, block size or any other piece of
information.  Thus, drop DEFINE_SCSI_DISK_PROPERTIES() from the
list of qdev properties.
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

Signed-off-by: NMarkus Armbruster <armbru@redhat.com>
Signed-off-by: NKevin Wolf <kwolf@redhat.com>

Signed-off-by: NMarkus Armbruster <armbru@redhat.com>
Signed-off-by: NKevin Wolf <kwolf@redhat.com>

Geometry needs to be qdev properties, because it belongs to the
disk's guest part.

Maintain backward compatibility exactly like for serial: fall back to
DriveInfo's geometry, set with -drive cyls=...

Do this only for scsi-hd.  scsi-disk is legacy.  scsi-cd doesn't have
a geometry.  scsi-block should get geometry from the host disk.

Bonus: info qtree now shows the geometry.
Signed-off-by: NMarkus Armbruster <armbru@redhat.com>
Signed-off-by: NKevin Wolf <kwolf@redhat.com>

Best to use the same type, to avoid unwanted truncation or sign
extension.

BlockConf can't use plain int for cyls, heads and secs, because
integer properties require an exact width.
Signed-off-by: NMarkus Armbruster <armbru@redhat.com>
Signed-off-by: NKevin Wolf <kwolf@redhat.com>

hd_geometry_guess() picks geometry and translation.  Callers can get
the geometry directly, via parameters, but for translation they need
to go through the block layer.

Add a parameter for translation, so it can optionally be gotten just
like geometry.  In preparation of purging translation from the block
layer, which will happen later in this series.
Signed-off-by: NMarkus Armbruster <armbru@redhat.com>
Signed-off-by: NKevin Wolf <kwolf@redhat.com>

Commit f3d54fc4 factored it out of hw/ide.c for reuse.  Sensible,
except it was put into block.c.  Device-specific functionality should
be kept in device code, not the block layer.  Move it to
hw/hd-geometry.c, and make stylistic changes required to keep
checkpatch.pl happy.
Signed-off-by: NMarkus Armbruster <armbru@redhat.com>
Signed-off-by: NKevin Wolf <kwolf@redhat.com>

This command is not necessary for CD-ROM and DVD-ROM, but some versions of
udev trip on its absence.

Cc: Kevin Wolf <kwolf@redhat.com>
Cc: Markus Armbruster <armbru@redhat.com>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

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

The last four bytes of the thin provisioning page were cut out.
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

QEMU exposes its version to the guest's hardware and in some cases that is wrong
(e.g. Windows prints messages about driver updates when you switch
the QEMU version).
There is a new field now on the struct QEmuMachine, hw_version, which may
contain the version that the specific machine should report. If that field is
set, then that machine will report that version to the guest.
Signed-off-by: NCrístian Viana <vianac@linux.vnet.ibm.com>
Signed-off-by: NAnthony Liguori <aliguori@us.ibm.com>

Use the appropriate interface instead.
Signed-off-by: NMarkus Armbruster <armbru@redhat.com>
Acked-by: NPaolo Bonzini <pbonzini@redhat.com>
Signed-off-by: NKevin Wolf <kwolf@redhat.com>

usb-storage can't handle requests in one go as the data transfer can be
splitted into lots of usb packets.  Because of that there can be
normal in-flight requests at savevm time and we need to handle that.
With other scsi hba's this happens only in case i/o is stopped due to
errors and there are pending requests which need to be restarted
(req->retry = true).

So, first we need to save req->retry and then handle the req->retry =
false case.  Write requests are handled fine already.  For read requests
we have to save the buffer as we will not restart the request (and thus
not refill the buffer) on the target host.

Cc: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: NGerd Hoffmann <kraxel@redhat.com>

Optional inquiry information is declared obsolete in the latest versions
of the standard; invalid CDBs or unsupported VPD pages are supported
can be diagnosed with trace_scsi_inquiry.
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

The requirements on the INQUIRY buffer size are not in my copy of SPC
(SPC-4 r27) and not observed by LIO.  Rip them out.
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

Linux expects REQ_FUA to be advertised only if WRITE+FUA is faster than
WRITE+SYNCHRONIZE CACHE, so we should not set the DPOFUA bit.  However,
it is useful to have it for testing purposes, so add a qdev property to
set it.
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

It is pointless to add a uint32_t field for every new feature.
Since we will need a new feature soon, convert accesses to "removable"
to look at bit 0 only.
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

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

Recently introduced FUA support also gave us a use-after-free
of the BlockAcctCookie within a SCSIDiskReq, due to unbalanced
reference counting.

The patch fixes this by making scsi_do_read look like a combination
of scsi_*_complete + scsi_*_data.  It does both a ref (like
scsi_read_data) and an unref (like scsi_flush_complete).
Reported-by: NDavid Gibson <david@gibson.dropbear.id.au>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

Signed-off-by: NRonnie Sahlberg <ronniesahlberg@gmail.com>
[Actually, we should report it only if discard_granularity is nonzero.
 Older SBC drafts assigned 0 to thin provisioning and 1 to thick
 (resource-provisioned, they call it).  Newer drafts assign respectively
 1 and 2 - Paolo]
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>