Igor Mammedov [Fri, 26 Sep 2014 09:28:36 +0000 (09:28 +0000)]
usb-bot: Drop not needed "allow_hotplug = 0"
Drop useless hack that disables hotplug on bus by setting
"allow_hotplug = 0". Even if bus is hotpluggable, It won't
be possible to add another SCSI device to bus since its
realization will fail early with error "no free target"
in scsi_qdev_realize() method.
Igor Mammedov [Fri, 26 Sep 2014 09:28:35 +0000 (09:28 +0000)]
usb-bot: Mark device as non hotpluggable
usb-bot creates SCSI bus and immediately makes it
non hotpluggable which was making not possible to
hotplug usb-bot since QEMU would abort at
bus_add_child(scsi-hd) time when usb-bot is
realized.
Mark usb-bot as not hotpluggable so that attempt
to hotplug it would error out even before it gets
to device initialization point.
Igor Mammedov [Fri, 26 Sep 2014 09:28:28 +0000 (09:28 +0000)]
s390x: Drop not used allow_hotplug in event-facility
s390-sclp-event-facility creates s390-sclp-events-bus
and immediately sets its allow_hotplug field to 0,
which is NOP since it's already 0 by default.
Also since BUS is not hotpluggable, it's not possible
to call SCLP_EVENT{ DeviceClass::unplug } callback
from qdev_unplug() making this unreachable code,
so drop it as well.
Igor Mammedov [Fri, 26 Sep 2014 09:28:27 +0000 (09:28 +0000)]
virtio-mmio: Drop useless bus->allow_hotplug = 0
Bus by default is not hotpluggable.
virtio-mmio-bus and its parent types do not set allow_hotplug
anywhere explicitly, so remove not needed field access
and wrapper along with it.
Igor Mammedov [Fri, 26 Sep 2014 09:28:25 +0000 (09:28 +0000)]
virtio-pci: Drop BusState::allow_hotplug
virtio-pci-bus is an internal object of composite
virtio-pci device and it doesn't participate in
-device/device_add hotplug flow, and since it's
not required by bus_add_child() that BUS must
be hotpluggable to be able to add child at runtime,
it's possible to drop not needed 'allow_hotplug'
field.
Igor Mammedov [Fri, 26 Sep 2014 09:28:23 +0000 (09:28 +0000)]
qdev: Drop hotplug check from bus_add_child()
Check is too restrictive and does not allow
to add children to just created bus during hotplug
when the bus is part of composite device.
Removing check from bus_add_child() doesn't affect
devices creatable with device_add/del commands since
they have a similar builtin check and patch will
allow to create complex composite devices during
hotplug.
Igor Mammedov [Fri, 26 Sep 2014 09:28:22 +0000 (09:28 +0000)]
qdev: Add wrapper to set BUS as HotplugHandler
To be used for conversion of SCSI and USB devices,
and would allow to make every HBA/USB host switch
to HotplugHandler API without touching each controller
explicitly.
Igor Mammedov [Fri, 26 Sep 2014 09:28:21 +0000 (09:28 +0000)]
qdev: Add simple/generic unplug callback for HotplugHandler
It will be used in shallow conversion from legacy hotplug
mechanism and eventually replace all the uses of old mechanism
DeviceClass::unplug = qdev_simple_unplug_cb()
Igor Mammedov [Fri, 26 Sep 2014 09:28:20 +0000 (09:28 +0000)]
qdev: HotplugHandler: Provide unplug callback
It is to be called for actual device removal and
will allow to separate request and removal handling
phases of x86-CPU devices and also it's a handler
to be called for synchronously removable devices.
Igor Mammedov [Fri, 26 Sep 2014 09:28:19 +0000 (09:28 +0000)]
qdev: HotplugHandler: Rename unplug callback to unplug_request
'HotplugHandler.unplug' callback is currently used as async
call to issue unplug request for device that implements it.
Renaming 'unplug' callback to 'unplug_request' should help to
avoid confusion about what callback does and would allow to
introduce 'unplug' callback that would perform actual device
removal when guest is ready for it.
Igor Mammedov [Fri, 26 Sep 2014 09:28:18 +0000 (09:28 +0000)]
qdev: do not allow to instantiate non hotpluggable device with device_add
It will allow explicitly mark device as not hotpluggable and
avoid its creation with following error at realize time
and destroying it afterwards anyway. Instead of it will
error out even before instance of device is created.
Igor Mammedov [Fri, 26 Sep 2014 09:28:17 +0000 (09:28 +0000)]
Access BusState::allow_hotplug using wraper qbus_is_hotpluggable()
It would allow to transparently switch detection whether Bus
is hotpluggable from allow_hotplug field to hotplug_handler
link and to drop allow_hotplug field once all users are
converted to hotplug handler API.
Igor Mammedov [Fri, 26 Sep 2014 09:28:16 +0000 (09:28 +0000)]
tests: usb: usb-uas hotplug test
checks that it's possible to hotplug usb-uas HBA and
then if it's possible to hot(un)plug scsi-disk to it.
Thest basically covers hot(un)plug on dummy HBAs
without means of hot(un)plug notification of the guest.
Igor Mammedov [Fri, 26 Sep 2014 09:28:15 +0000 (09:28 +0000)]
tests: usb: usb-storage hotplug test
usb-storage is different from usual usb devices
in that it uses a child SCSI bus for underlying storage.
This commit verifies that the SCSI bus is hotpluggable, as
hotplug operation wouldn't succeed without it.
object_property_add_alias() is called at some
places at present. And its parameter errp may not NULL,
such as
object_property_add_alias(obj, "iothread", OBJECT(&dev->vdev),"iothread",
&error_abort);
This patch add error handler for security.
qdev: gpio: Don't allow name share between I and O
Only allow a GPIO name to be one or the other. Inputs and outputs are
functionally different and should be in different namespaces. Prepares
support for the QOMification of IRQs as Links or Child objects.
The alternative is to munge names .e.g. with "-in" or "-out" suffixes
when giving QOM names. But that reduces clarity and if there are cases
out there where users want I and O with same name they can manually add
their own suffixes.
Peter Maydell [Fri, 10 Oct 2014 13:55:29 +0000 (14:55 +0100)]
Merge remote-tracking branch 'remotes/cohuck/tags/s390x-20141010' into staging
various s390x updates:
- cpu state handling in qemu and migration
- vhost-scsi-ccw bugfix
# gpg: Signature made Fri 10 Oct 2014 14:01:34 BST using RSA key ID C6F02FAF
# gpg: Can't check signature: public key not found
* remotes/cohuck/tags/s390x-20141010:
s390x/virtio-ccw: fix vhost-scsi intialization
s390x/migration: migrate CPU state
s390x/kvm: synchronize the cpu state after SIGP (INITIAL) CPU RESET
s390x/kvm: reuse kvm_s390_reset_vcpu() to get rid of ifdefs
s390x/kvm: propagate s390 cpu state to kvm
s390x/kvm: proper use of the cpu states OPERATING and STOPPED
s390x/kvm: introduce proper states for s390 cpus
linux-headers: update to 3.17-rc7
Paolo Bonzini [Fri, 10 Oct 2014 10:23:35 +0000 (12:23 +0200)]
kvm fix compilation with GCC 4.3.4
As usual, SLES11's GCC complained about double typedefs:
/home/cohuck/git/qemu/kvm-all.c:110: error: redefinition of typedef ‘KVMState’
/home/cohuck/git/qemu/include/sysemu/kvm.h:161: error: previous declaration of ‘KVMState’ was here
Thomas Huth [Tue, 30 Sep 2014 08:58:42 +0000 (10:58 +0200)]
s390x/migration: migrate CPU state
This patch provides the cpu save information for dumps and later life
migration and enables migration of the CPU state. The code is based on
earlier work from Christian Borntraeger and Jason Herne.
s390x/kvm: synchronize the cpu state after SIGP (INITIAL) CPU RESET
We need to synchronize registers after a reset has been performed. The
current code does that in qemu_system_reset(), load_normal_reset() and
modified_clear_reset() for all vcpus. After SIGP (INITIAL) CPU RESET,
this needs to be done for the targeted vcpu as well, so let's call
cpu_synchronize_post_reset() in the respective handlers.
s390x/kvm: reuse kvm_s390_reset_vcpu() to get rid of ifdefs
This patch reuses kvm_s390_reset_vcpu() to get rid of some CONFIG_KVM and
CONFIG_USER_ONLY ifdefs in cpu.c.
In order to get rid of CONFIG_USER_ONLY, kvm_s390_reset_vcpu() has to provide a
dummy implementation - the two definitions are moved to the proper section in
cpu.h.
Let QEMU propagate the cpu state to kvm. If kvm doesn't yet support it, it is
silently ignored as kvm will still handle the cpu state itself in that case.
The state is not synced back, thus kvm won't have a chance to actively modify
the cpu state. To do so, control has to be given back to QEMU (which is already
done so in all relevant cases).
Setting of the cpu state can fail either because kvm doesn't support the
interface yet, or because the state is invalid/not supported. Failed attempts
will be traced
s390x/kvm: proper use of the cpu states OPERATING and STOPPED
This patch makes sure that halting a cpu and stopping a cpu are two different
things. Stopping a cpu will also set the cpu halted - this is needed for common
infrastructure to work (note that the stop and stopped flag cannot be used for
our purpose because they are already used by other mechanisms).
A cpu can be halted ("waiting") when it is operating. If interrupts are
disabled, this is called a "disabled wait", as it can't be woken up anymore. A
stopped cpu is treated like a "disabled wait" cpu, but in order to prepare for a
proper cpu state synchronization with the kvm part, we need to track the real
logical state of a cpu.
Until now, when a s390 cpu was stopped or halted, the number of running
CPUs was tracked in a global variable. This was problematic for migration,
so Jason came up with a per-cpu running state.
As it turns out, we want to track the full logical state of a target vcpu,
so we need real s390 cpu states.
This patch is based on an initial patch by Jason Herne, but was heavily
rewritten when adding the cpu states STOPPED and OPERATING. On the way we
move add_del_running to cpu.c (the declaration is already in cpu.h) and
modify the users where appropriate.
Please note that the cpu is still set to be stopped when it is
halted, which is wrong. This will be fixed in the next patch. The LOAD and
CHECK-STOP state will not be used in the first step.
Peter Maydell [Thu, 9 Oct 2014 14:09:05 +0000 (15:09 +0100)]
Merge remote-tracking branch 'remotes/bonzini/tags/for-upstream' into staging
Four changes here. Polling for reconnection of character devices,
the QOMification of accelerators, a fix for -kernel support on x86, and one
for a recently-introduced virtio-scsi optimization.
# gpg: Signature made Thu 09 Oct 2014 14:36:50 BST using RSA key ID 4E6B09D7
# gpg: Good signature from "Paolo Bonzini <[email protected]>"
# gpg: aka "Paolo Bonzini <[email protected]>"
* remotes/bonzini/tags/for-upstream: (28 commits)
qemu-char: Fix reconnect socket error reporting
qemu-sockets: Add error to non-blocking connect handler
qemu-error: Add error_vreport()
virtio-scsi: fix use-after-free of VirtIOSCSIReq
linuxboot: compute initrd loading address
kvm: Make KVMState be the TYPE_KVM_ACCEL instance struct
accel: Create accel object when initializing machine
accel: Pass MachineState object to accel init functions
accel: Rename 'init' method to 'init_machine'
accel: Move accel init/allowed code to separate function
accel: Remove tcg_available() function
accel: Move qtest accel registration to qtest.c
accel: Move Xen registration code to xen-common.c
accel: Move KVM accel registration to kvm-all.c
accel: Report unknown accelerator as "not found" instead of "does not exist"
accel: Make AccelClass.available() optional
accel: Use QOM classes for accel types
accel: Move accel name lookup to separate function
accel: Simplify configure_accelerator() using AccelType *acc variable
accel: Create AccelType typedef
...
Corey Minyard [Wed, 8 Oct 2014 12:11:55 +0000 (07:11 -0500)]
qemu-char: Fix reconnect socket error reporting
If reconnect was set, errors wouldn't always be reported.
Fix that and also only report a connect error once until a
connection has been made.
The primary purpose of this is to tell the user that a
connection failed so they can know they need to figure out
what went wrong. So we don't want to spew too much
out here, just enough so they know.
Corey Minyard [Wed, 8 Oct 2014 12:11:56 +0000 (07:11 -0500)]
qemu-sockets: Add error to non-blocking connect handler
An error value here would be quite handy and more consistent
with the rest of the code.
Signed-off-by: Corey Minyard <[email protected]>
[Make sure SO_ERROR value is passed to error_setg_errno. - Paolo] Signed-off-by: Paolo Bonzini <[email protected]>
Paolo Bonzini [Mon, 6 Oct 2014 14:49:57 +0000 (16:49 +0200)]
linuxboot: compute initrd loading address
Even though hw/i386/pc.c tries to compute a valid loading address for the
initrd, close to the top of RAM, this does not take into account other
data that is malloced into that memory by SeaBIOS.
Luckily we can easily look at the memory map to find out how much memory is
used up there. This patch places the initrd in the first four gigabytes,
below the first hole (as returned by INT 15h, AX=e801h).
Without this patch:
[ 0.000000] init_memory_mapping: [mem 0x07000000-0x07fdffff]
[ 0.000000] RAMDISK: [mem 0x0710a000-0x07fd7fff]
With this patch:
[ 0.000000] init_memory_mapping: [mem 0x07000000-0x07fdffff]
[ 0.000000] RAMDISK: [mem 0x07112000-0x07fdffff]
So linuxboot is able to use the 64k that were added as padding for
QEMU <= 2.1.
kvm: Make KVMState be the TYPE_KVM_ACCEL instance struct
Now that we create an accel object before calling machine_init, we can
simply use the accel object to save all KVMState data, instead of
allocationg KVMState manually.
accel: Create accel object when initializing machine
Create an actual TYPE_ACCEL object when initializing a machine. This
will allow accelerator classes to implement some initialization on
instance_init, and to save state on the TYPE_ACCEL object.
accel: Pass MachineState object to accel init functions
Most of the machine options and machine state information is in the
MachineState object, not on the MachineClass. This will allow init
functions to use the MachineState object directly instead of
qemu_get_machine_opts() or the current_machine global.
Peter Maydell [Tue, 7 Oct 2014 09:41:48 +0000 (10:41 +0100)]
Merge remote-tracking branch 'remotes/riku/tags/pull-linux-user-20141006-2' into staging
linux-user pull for 2.2
Clearest linux-user patches sent to the list since august,
Apart from Mikhails patch, the rest are quite trivial.
v2: check for CONFIG_TIMERFD only after it has been defined
# gpg: Signature made Mon 06 Oct 2014 20:08:10 BST using RSA key ID DE3C9BC0
# gpg: Good signature from "Riku Voipio <[email protected]>"
# gpg: aka "Riku Voipio <[email protected]>"
* remotes/riku/tags/pull-linux-user-20141006-2:
translate-all.c: memory walker initial address miscalculation
linux-user: don't include timerfd if not needed
linux-user: Simplify timerid checks on g_posix_timers range
linux-user: Convert blkpg to use a special subop handler
linux-user: Enable epoll_pwait syscall for ARM
The initial base address is miscalculated in walk_memory_regions().
It has to be shifted TARGET_PAGE_BITS more. Holder variables are
extended to target_ulong size otherwise they don't fit for MIPS N32
(a 32-bit ABI with a 64-bit address space) and qemu won't compile.
The issue led to incorrect debug output of memory maps and a
mis-formed coredumped file.
Alexander Graf [Fri, 22 Aug 2014 11:56:18 +0000 (13:56 +0200)]
linux-user: Simplify timerid checks on g_posix_timers range
We check whether the passed in timer id is negative on all calls
that involve g_posix_timers.
However, these checks are bogus. First off we limit the timer_id to
16 bits which is not what Linux does. Then we check whether it's negative
which it can't be because we masked it.
We can safely remove the masking. For the negativity check we can just
treat the timerid as unsigned and only check for upper boundaries.
Alexander Graf [Fri, 22 Aug 2014 11:15:50 +0000 (13:15 +0200)]
linux-user: Convert blkpg to use a special subop handler
The blkpg ioctl can take different payloads depending on the opcode in
its payload structure. Create a new special ioctl handler that can only
deal with partition style ones for now.
Peter Maydell [Fri, 19 Sep 2014 01:35:20 +0000 (02:35 +0100)]
linux-user: Enable epoll_pwait syscall for ARM
We have support for the epoll_pwait syscall, but it wasn't enabled for
ARM guests because we hadn't defined the syscall number; correct this
deficiency.
Peter Maydell [Fri, 12 Sep 2014 18:04:17 +0000 (19:04 +0100)]
gdbstub: Allow target CPUs to specify watchpoint STOP_BEFORE_ACCESS flag
GDB assumes that watchpoint set via the gdbstub remote protocol will
behave in the same way as hardware watchpoints for the target. In
particular, whether the CPU stops with the PC before or after the insn
which triggers the watchpoint is target dependent. Allow guest CPU
code to specify which behaviour to use. This fixes a bug where with
guest CPUs which stop before the accessing insn GDB would manually
step forward over what it thought was the insn and end up one insn
further forward than it should be.
We set this flag for the CPU architectures which set
gdbarch_have_nonsteppable_watchpoint in gdb 7.7:
ARM, CRIS, LM32, MIPS and Xtensa.
Peter Maydell [Mon, 6 Oct 2014 09:59:55 +0000 (10:59 +0100)]
Merge remote-tracking branch 'remotes/stefanha/tags/block-pull-request' into staging
# gpg: Signature made Sat 04 Oct 2014 21:24:46 BST using RSA key ID 81AB73C8
# gpg: Good signature from "Stefan Hajnoczi <[email protected]>"
# gpg: aka "Stefan Hajnoczi <[email protected]>"
* remotes/stefanha/tags/block-pull-request: (23 commits)
blockdev-test: Test device_del after drive_del
blockdev-test: Factor out some common code into helpers
blockdev-test: Simplify by using g_assert_cmpstr()
blockdev-test: Clean up bogus drive_add argument
blockdev-test: Use single rather than double quotes in QMP
drive_del-test: Merge of qdev-monitor-test, blockdev-test
iotests: qemu-img info output for corrupt image
qapi: Add corrupt field to ImageInfoSpecificQCow2
iotests: Use _img_info
util: Emancipate id_wellformed() from QemuOpts
q35/ahci: Pick up -cdrom and -hda options
qtest/bios-tables: Correct Q35 command line
ide: Update ide_drive_get to be HBA agnostic
pc/vl: Add units-per-default-bus property
blockdev: Allow overriding if_max_dev property
blockdev: Orphaned drive search
qemu-iotests: Fix supported cache modes for 052
make check-block: Use default cache modes
Modify qemu_opt_rename to realize renaming all items in opts
vmdk: Fix integer overflow in offset calculation
...
The first argument should be a PCI address, which pci-addr=auto isn't.
Doesn't really matter, as drive_add ignores its first argument when
its second argument has if=none. Clean it up anyway.
blockdev-test: Use single rather than double quotes in QMP
QMP accepts both single and double quotes. This is the only test
using double quotes. They need to be quoted in C strings. Replace
them by single quotes.
Max Reitz [Tue, 30 Sep 2014 19:31:27 +0000 (21:31 +0200)]
iotests: Use _img_info
qemu-img info should only be used directly if the format-specific
information or the name of the format is relevant (some tests explicitly
test format-specific information; test 082 uses qcow2-specific settings
to test the qemu-img interface); otherwise, tests should always use
_img_info instead.
Test 082 was touched only partially. It does test the qemu-img
interface; however, its invocations of qemu-img info are not real tests
but rather verifications, so if format-specific information is not
important for the test, there is no reason not to use _img_info. In
contrast to directly invoking qemu-img info, "qcow2" is replaced by
"IMGFMT"; but as "qcow2" is only mentioned once in test 082 (in
_supported_fmt), I consider this an improvement.
Today, all accelerator init functions affect some global state:
* tcg_init() calls tcg_exec_init() and affects globals such as tcg_tcx,
page size globals, and possibly others;
* kvm_init() changes the kvm_state global, cpu_interrupt_handler, and possibly
others;
* xen_init() changes the xen_xc global, and registers a change state handler.
With the new accelerator QOM classes, initialization may now be split in two
steps:
* instance_init() will do basic initialization that doesn't affect any global
state and don't need MachineState or MachineClass data. This will allow
probing code to safely create multiple accelerator objects on the fly just
for reporting host/accelerator capabilities, for example.
* accel_init_machine()/init_machine() will save the accelerator object in
MachineState, and do initialization steps which still affect global state,
machine state, or that need data from MachineClass or MachineState.
To clarify the difference between those two steps, rename init() to
init_machine().
As qtest_availble() returns 1 only when CONFIG_POSIX is set, keep
setting AccelClass.available to keep current behavior (this is different
from what we did for KVM and Xen).
This also allows us to make qtest_init_accel() static.
Note that this has an user-visible side-effect: instead of reporting
"Xen is not supported for this target", QEMU binaries not supporting Xen
will report "xen accelerator does not exist".
As xen_available() always return 1 when CONFIG_XEN is enabled, we don't
need to set AccelClass.available anymore. xen_enabled() is not being
removed yet, but only because vl.c is still using it.
Note that this has an user-visible side-effect: instead of reporting
"KVM is not supported for this target", QEMU binaries not supporting KVM
will report "kvm accelerator does not exist".
As kvm_availble() always return 1 when CONFIG_KVM is enabled, we don't
need to set AccelClass.available anymore. kvm_enabled() is not being
completely removed yet only because qmp_query_kvm() still uses it.
accel: Report unknown accelerator as "not found" instead of "does not exist"
As the accelerator classes won't be registered anymore if they are not
enabled at compile time, saying "does not exist" may be misleading, as
the accelerator may be simply disabled. Change the wording to just say
"not found".
When we move accel classes outside accel.c, the available() function
won't be necessary anymore, because the classes will be registered only
if the accelerator code is really enabled at build time.
Corey Minyard [Thu, 2 Oct 2014 16:17:37 +0000 (11:17 -0500)]
qemu-char: Add reconnecting to client sockets
Adds a "reconnect" option to socket backends that gives a reconnect
timeout. This only applies to client sockets. If the other end
of a socket closes the connection, qemu will attempt to reconnect
after the given number of seconds.
Corey Minyard [Thu, 2 Oct 2014 16:17:36 +0000 (11:17 -0500)]
qemu-char: set socket filename to disconnected when not connected
This way we can tell if the socket is connected or not. It also splits
the string conversions out into separate functions to make this more
convenient.
Corey Minyard [Thu, 2 Oct 2014 16:17:34 +0000 (11:17 -0500)]
qemu-char: Rework qemu_chr_open_socket() for reconnect
Move all socket configuration to qmp_chardev_open_socket().
qemu_chr_open_socket_fd() just opens the socket. This is getting ready
for the reconnect code, which will call open_sock_fd() on a reconnect
attempt.
IDs have long spread beyond QemuOpts: not everything with an ID
necessarily goes through QemuOpts. Commit 9aebf3b is about such a
case: block layer names are meant to be well-formed IDs, but some of
them don't go through QemuOpts, and thus weren't checked. The commit
fixed that the straightforward way: rename the internal QemuOpts
helper id_wellformed() to qemu_opts_id_wellformed() and give it
external linkage.
Instead of using it directly in block.c, the commit adds wrapper
bdrv_is_valid_name(), probably to hide the connection to QemuOpts.
Go one logical step further: emancipate IDs from QemuOpts. Rename the
function back to id_wellformed(), and put it in another file. While
there, clean up its value to bool. Peel off the bdrv_is_valid_name()
wrapper.
[Replaced stray return 0 with return false to match bool returns used
elsewhere in id_wellformed().
--Stefan]
John Snow [Wed, 1 Oct 2014 18:19:29 +0000 (14:19 -0400)]
q35/ahci: Pick up -cdrom and -hda options
This patch implements the backend for the Q35 board
for us to be able to pick up and use drives defined
by the -cdrom, -hda, or -drive if=ide shorthand options.
John Snow [Wed, 1 Oct 2014 18:19:28 +0000 (14:19 -0400)]
qtest/bios-tables: Correct Q35 command line
If the Q35 board types are to begin recognizing
and decoding syntactic sugar for drive/device
declarations, then workarounds found within
the qtests suite need to be adjusted to prevent
any test failures after the fix.
bios-tables-test improperly uses this cli:
-drive file=etc,id=hd -device ide-hd,drive=hd
Which will create a drive and device due to
the lack of specifying if=none. Then, it will
attempt to create a second device and fail.
This patch corrects this test to always use
the full, non-sugared -device/-drive syntax
for both PC and Q35.
John Snow [Wed, 1 Oct 2014 18:19:27 +0000 (14:19 -0400)]
ide: Update ide_drive_get to be HBA agnostic
Instead of duplicating the logic for the if_ide
(bus,unit) mappings, rely on the blockdev layer
for managing those mappings for us, and use the
drive_get_by_index call instead.
This allows ide_drive_get to work for AHCI HBAs
as well, and can be used in the Q35 initialization.
Lastly, change the nature of the argument to
ide_drive_get so that represents the number of
total drives we can support, and not the total
number of buses. This will prevent array overflows
if the units-per-default-bus property ever needs
to be adjusted for compatibility reasons.
John Snow [Wed, 1 Oct 2014 18:19:26 +0000 (14:19 -0400)]
pc/vl: Add units-per-default-bus property
This patch adds the 'units_per_default_bus' property which
allows individual boards to declare their desired
index => (bus,unit) mapping for their default HBA, so that
boards such as Q35 can specify that its default if_ide HBA,
AHCI, only accepts one unit per bus.
This property only overrides the mapping for drives matching
the block_default_type interface.
This patch also adds this property to *all* past and present
Q35 machine types. This retroactive addition is justified
because the previous erroneous index=>(bus,unit) mappings
caused by lack of such a property were not utilized due to
lack of initialization code in the Q35 init routine.
Further, semantically, the Q35 board type has always had the
property that its default HBA, AHCI, only accepts one unit per
bus. The new code added to add devices to drives relies upon
the accuracy of this mapping. Thus, the property is applied
retroactively to reduce complexity of allowing IDE HBAs with
different units per bus.
Examples:
Prior to this patch, all IDE HBAs were assumed to use 2 units
per bus (Master, Slave). When using Q35 and AHCI, however, we
only allow one unit per bus.
-hdb foo.qcow2 would become index=1, or bus=0,unit=1.
-hdd foo.qcow2 would become index=3, or bus=1,unit=1.
-drive file=foo.qcow2,index=5 becomes bus=2,unit=1.
These are invalid for AHCI. They now become, under Q35 only:
The mapping is adjusted based on the fact that the default IF
for the Q35 machine type is IF_IDE, and units-per-default-bus
overrides the IDE mapping from its default of 2 units per bus
to just 1 unit per bus.
John Snow [Wed, 1 Oct 2014 18:19:25 +0000 (14:19 -0400)]
blockdev: Allow overriding if_max_dev property
The if_max_devs table as in the past been an immutable
default that controls the mapping of index => (bus,unit)
for all boards and all HBAs for each interface type.
Since adding this mapping information to the HBA device
itself is currently unwieldly from the perspective of
retrieving this information at option parsing time
(e.g, within drive_new), we consider the alternative
of marking the if_max_devs table mutable so that
later configuration and initialization can adjust the
mapping at will, but only up until a drive is added,
at which point the mapping is finalized.
John Snow [Wed, 1 Oct 2014 18:19:24 +0000 (14:19 -0400)]
blockdev: Orphaned drive search
When users use command line options like -hda, -cdrom,
or even -drive if=ide, it is up to the board initialization
routines to pick up these drives and create backing
devices for them.
Some boards, like Q35, have not been doing this.
However, there is no warning explaining why certain
drive specifications are just silently ignored,
so this function adds a check to print some warnings
to assist users in debugging these sorts of issues
in the future.
This patch will not warn about drives added with if_none,
for which it is not possible to tell in advance if
the omission of a backing device is an issue.
A warning in these cases is considered appropriate.
Kevin Wolf [Tue, 30 Sep 2014 11:27:10 +0000 (13:27 +0200)]
qemu-iotests: Fix supported cache modes for 052
The requirement for this test case is really "no O_DIRECT", because the
temporary snapshot for BDRV_O_SNAPSHOT is created in /tmp, which often
is a tmpfs.
Commit f210a83c ('qemu-iotests: Add _default_cache_mode and
_supported_cache_modes') turned the restriction into writethrough-only,
but that's not really necessary.
Allow to run the test for any non-O_DIRECT cache modes, and use the
global default of writeback if no cache mode is specified.
projects / qemu.git / log