sdk/emulator/qemu.git
12 years agoMerge remote-tracking branch 'sstabellini/xen-fixes-20120801' into staging
Anthony Liguori [Fri, 3 Aug 2012 18:54:35 +0000 (13:54 -0500)]
Merge remote-tracking branch 'sstabellini/xen-fixes-20120801' into staging

* sstabellini/xen-fixes-20120801:
  fix Xen compilation
  configure: Fix xen probe with Xen 4.2 and later

12 years agoMerge remote-tracking branch 'stefanha/net' into staging
Anthony Liguori [Fri, 3 Aug 2012 18:54:05 +0000 (13:54 -0500)]
Merge remote-tracking branch 'stefanha/net' into staging

* stefanha/net:
  net: add the support for -netdev socket, listen
  net: fix the coding style
  hub: add the support for hub own flow control
  net: determine if packets can be sent before net queue deliver packets
  net: cleanup deliver/deliver_iov func pointers
  net: Make "info network" output more readable info
  net: Rename qemu_del_vlan_client() to qemu_del_net_client()
  net: Rename vc local variables to nc
  net: Rename VLANClientState to NetClientState
  net: Rename non_vlan_clients to net_clients
  net: Remove VLANState
  net: Remove vlan code from net.c
  net: Convert qdev_prop_vlan to peer with hub
  net: Drop vlan argument to qemu_new_net_client()
  hub: Check that hubs are configured correctly
  net: Look up 'vlan' net clients using hubs
  net: Use hubs for the vlan feature
  net: Add a hub net client
  net: Add interface to bridge when SIOCBRADDIF isn't available

12 years agoMerge remote-tracking branch 'stefanha/trivial-patches' into staging
Anthony Liguori [Fri, 3 Aug 2012 18:51:19 +0000 (13:51 -0500)]
Merge remote-tracking branch 'stefanha/trivial-patches' into staging

* stefanha/trivial-patches:
  exec.c: Remove out of date comment
  exec.c: Use subpages for large unaligned mappings
  exec.c: Fix off-by-one error in register_subpage
  socket: clean up redundant assignment
  qom: Clean libuser object and dependency files
  usb: Clean common object and dependency files

12 years agoMerge remote-tracking branch 'aneesh/for-upstream' into staging
Anthony Liguori [Fri, 3 Aug 2012 18:50:41 +0000 (13:50 -0500)]
Merge remote-tracking branch 'aneesh/for-upstream' into staging

* aneesh/for-upstream:
  hw/9pfs: Fix assert when disabling migration
  configure: Fix build with capabilities

12 years agoexec.c: Remove out of date comment
Peter Maydell [Wed, 1 Aug 2012 13:35:47 +0000 (14:35 +0100)]
exec.c: Remove out of date comment

Remove an out of date comment: this comment used to be attached to
cpu_register_physical_memory_log(), before commit 0f0cb164 accidentally
inserted a couple of other functions between the comment and its function.
It is in any case obsolete since (a) the function arguments it refers
to have been replaced with a single MemoryRegionSection* argument and
(b) the inability to handle regions whose offset_within_address_space
and offset_within_region aren't equally aligned was fixed as part of
the rewrite of this code.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
12 years agoexec.c: Use subpages for large unaligned mappings
Tyler Hall [Wed, 25 Jul 2012 22:45:04 +0000 (18:45 -0400)]
exec.c: Use subpages for large unaligned mappings

Registering a multi-page memory region that is non-page-aligned results
in a subpage from the start to the page boundary, some number of full
pages, and possibly another subpage from the last page boundary to the
end. The full pages will have a value for offset_within_region that is
not a multiple of TARGET_PAGE_SIZE. Accesses through softmmu are unable
to handle this and will segfault.

Handling full pages through subpages is not optimal, but only
non-page-aligned mappings take the penalty.

Signed-off-by: Tyler Hall <tylerwhall@gmail.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Avi Kivity <avi@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
12 years agoexec.c: Fix off-by-one error in register_subpage
Tyler Hall [Wed, 25 Jul 2012 22:45:03 +0000 (18:45 -0400)]
exec.c: Fix off-by-one error in register_subpage

subpage_register() expects "end" to be the last byte in the mapping.
Registering a non-page-aligned memory region that extends up to or
beyond a page boundary causes subpage_register() to silently fail
through the (end >= PAGE_SIZE) check.

This bug does not cause noticeable problems for mappings that do not
extend to a page boundary, though they do register an extra byte.

Signed-off-by: Tyler Hall <tylerwhall@gmail.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Avi Kivity <avi@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
12 years agosocket: clean up redundant assignment
Amos Kong [Fri, 3 Aug 2012 03:06:22 +0000 (11:06 +0800)]
socket: clean up redundant assignment

Signed-off-by: Amos Kong <akong@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
12 years agoqom: Clean libuser object and dependency files
Jan Kiszka [Mon, 23 Jul 2012 11:45:01 +0000 (13:45 +0200)]
qom: Clean libuser object and dependency files

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
12 years agousb: Clean common object and dependency files
Jan Kiszka [Mon, 23 Jul 2012 11:44:35 +0000 (13:44 +0200)]
usb: Clean common object and dependency files

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
12 years agoSupport 'help' as a synonym for '?' in command line options
Peter Maydell [Thu, 2 Aug 2012 12:45:54 +0000 (13:45 +0100)]
Support 'help' as a synonym for '?' in command line options

For command line options which permit '?' meaning 'please list the
permitted values', add support for 'help' as a synonym, by abstracting
the check out into a helper function.

This change means that in some cases where we were being lazy in
our string parsing, "?junk" will now be rejected as an invalid option
rather than being (undocumentedly) treated the same way as "?".

Update the documentation to use 'help' rather than '?', since '?'
is a shell metacharacter and thus prone to fail confusingly if there
is a single character filename in the current working directory and
the '?' has not been escaped. It's therefore better to steer users
towards 'help', though '?' is retained for backwards compatibility.

We do not, however, update the output of the system emulator's -help
(or any documentation autogenerated from the qemu-options.hx which
is the source of the -help text) because libvirt parses our -help
output and will break. At a later date when QEMU provides a better
interface so libvirt can avoid having to do this, we can update the
-help text too.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
12 years agoReplace 'struct siginfo' with 'siginfo_t'.
Richard W.M. Jones [Thu, 5 Jul 2012 03:32:44 +0000 (03:32 +0000)]
Replace 'struct siginfo' with 'siginfo_t'.

glibc 2.16 will remove the undocumented definition of 'struct siginfo'
from <bits/siginfo.h>.

This change is already present in glibc 2.15.90, so qemu compilation
of certain targets (eg. cris-user) breaks.

This struct was always typedef'd to be the same as 'siginfo_t' which
is what POSIX documents, so use that instead.

Signed-off-by: Richard W.M. Jones <rjones@redhat.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
12 years agotarget-i386: move cpu_reset and reset callback to cpu.c
Igor Mammedov [Mon, 23 Jul 2012 13:22:28 +0000 (15:22 +0200)]
target-i386: move cpu_reset and reset callback to cpu.c

Moving reset callback into cpu object from board level and
resetting cpu at the end of x86_cpu_realize() will allow properly
create cpu object during run-time (hotplug) without calling reset externaly.

When reset over QOM hierarchy is implemented, reset callback
should be removed.

v2:
  - leave cpu_reset in pc_new_cpu() for now, it's to be cleaned up when APIC
    init is moved in cpu.c

Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
12 years agotarget-i386: move cpu halted decision into x86_cpu_reset
Igor Mammedov [Mon, 23 Jul 2012 13:22:27 +0000 (15:22 +0200)]
target-i386: move cpu halted decision into x86_cpu_reset

MP initialization protocol differs between cpu families, and for P6 and
onward models it is up to CPU to decide if it will be BSP using this
protocol, so try to model this. However there is no point in implementing
MP initialization protocol in qemu. Thus first CPU is always marked as BSP.

This patch:
 - moves decision to designate BSP from board into cpu, making cpu
self-sufficient in this regard. Later it will allow to cleanup hw/pc.c
and remove cpu_reset and wrappers from there.
 - stores flag that CPU is BSP in IA32_APIC_BASE to model behavior
described in Inted SDM vol 3a part 1 chapter 8.4.1
 - uses MSR_IA32_APICBASE_BSP flag in apic_base for checking if cpu is BSP

patch is based on Jan Kiszka's proposal:
    http://thread.gmane.org/gmane.comp.emulators.qemu/100806

Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
12 years agofix Xen compilation
Stefano Stabellini [Wed, 1 Aug 2012 13:44:11 +0000 (13:44 +0000)]
fix Xen compilation

xen_pt_unregister_device is used as PCIUnregisterFunc, so it should
match the type.

Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Tested-by: Andreas Färber <afaerber@suse.de>
12 years agofix Xen compilation
Stefano Stabellini [Wed, 1 Aug 2012 10:19:09 +0000 (11:19 +0100)]
fix Xen compilation

xen_pt_unregister_device is used as PCIUnregisterFunc, so it should
match the type.

Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
12 years agoconfigure: Fix xen probe with Xen 4.2 and later
Anthony PERARD [Wed, 1 Aug 2012 13:42:44 +0000 (13:42 +0000)]
configure: Fix xen probe with Xen 4.2 and later

The xs.h header is now deprecated and produces a warning. This prevents the
configure script from enabling Xen with xen-unstable whom will become 4.2. As
this header is not anymore common to every version of Xen, we just remove it
from the early probe for Xen.

Signed-off-by: Anthony PERARD <anthony.perard@citrix.com>
Acked-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
12 years agonet: add the support for -netdev socket, listen
Zhi Yong Wu [Fri, 20 Jul 2012 13:25:53 +0000 (14:25 +0100)]
net: add the support for -netdev socket, listen

The -net socket,listen option does not work with the newer -netdev
syntax:
 http://lists.gnu.org/archive/html/qemu-devel/2011-11/msg01508.html

This patch makes it work now.

For the case where one vlan has multiple listenning sockets,
the patch will also provide the support.

Supported syntax:
 1.) -net socket,listen=127.0.0.1:1234,vlan=0
 2.) -net socket,listen=127.0.0.1:1234,vlan=0 -net socket,listen=127.0.0.1:1235,vlan=0
 3.) -netdev socket,listen=127.0.0.1:1234,id=socket0

 Drop the NetSocketListenState struct and add a listen_fd field
to NetSocketState.  When a -netdev socket,listen= instance is created
there will be a NetSocketState with fd=-1 and a valid listen_fd.  The
net_socket_accept() handler waits for listen_fd to become readable and
then accepts the connection.  When this state transition happens, we no
longer monitor listen_fd for incoming connections...until the client
disconnects again.

Signed-off-by: Zhi Yong Wu <wuzhy@linux.vnet.ibm.com>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
12 years agonet: fix the coding style
Zhi Yong Wu [Fri, 20 Jul 2012 13:25:52 +0000 (14:25 +0100)]
net: fix the coding style

Signed-off-by: Zhi Yong Wu <wuzhy@linux.vnet.ibm.com>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
12 years agohub: add the support for hub own flow control
Zhi Yong Wu [Tue, 24 Jul 2012 15:35:19 +0000 (16:35 +0100)]
hub: add the support for hub own flow control

Only when all other hub port's *peer* .can_receive() all return 1,
the source hub port .can_receive() return 1.

Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Zhi Yong Wu <wuzhy@linux.vnet.ibm.com>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
12 years agonet: determine if packets can be sent before net queue deliver packets
Zhi Yong Wu [Tue, 24 Jul 2012 15:35:18 +0000 (16:35 +0100)]
net: determine if packets can be sent before net queue deliver packets

Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Zhi Yong Wu <wuzhy@linux.vnet.ibm.com>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
12 years agonet: cleanup deliver/deliver_iov func pointers
Zhi Yong Wu [Tue, 24 Jul 2012 15:35:17 +0000 (16:35 +0100)]
net: cleanup deliver/deliver_iov func pointers

Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Zhi Yong Wu <wuzhy@linux.vnet.ibm.com>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
12 years agonet: Make "info network" output more readable info
Zhi Yong Wu [Tue, 24 Jul 2012 15:35:16 +0000 (16:35 +0100)]
net: Make "info network" output more readable info

Reviewed-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Zhi Yong Wu <wuzhy@linux.vnet.ibm.com>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
12 years agonet: Rename qemu_del_vlan_client() to qemu_del_net_client()
Stefan Hajnoczi [Tue, 24 Jul 2012 15:35:15 +0000 (16:35 +0100)]
net: Rename qemu_del_vlan_client() to qemu_del_net_client()

Another step in moving the vlan feature out of net core.  Users only
deal with NetClientState and therefore qemu_del_vlan_client() should be
named qemu_del_net_client().

Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Zhi Yong Wu <wuzhy@linux.vnet.ibm.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
12 years agonet: Rename vc local variables to nc
Stefan Hajnoczi [Tue, 24 Jul 2012 15:35:14 +0000 (16:35 +0100)]
net: Rename vc local variables to nc

Now that VLANClientState has been renamed to NetClientState all 'vc'
local variables should be 'nc'.  Much of the code already used 'nc' but
there are places where 'vc' needs to be renamed.

Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Zhi Yong Wu <wuzhy@linux.vnet.ibm.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
12 years agonet: Rename VLANClientState to NetClientState
Stefan Hajnoczi [Tue, 24 Jul 2012 15:35:13 +0000 (16:35 +0100)]
net: Rename VLANClientState to NetClientState

The vlan feature is no longer part of net core.  Rename VLANClientState
to NetClientState because net clients are not explicitly associated with
a vlan at all, instead they have a peer net client to which they are
connected.

This patch is a mechanical search-and-replace except for a few
whitespace fixups where changing VLANClientState to NetClientState
misaligned whitespace.

Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Zhi Yong Wu <wuzhy@linux.vnet.ibm.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
12 years agonet: Rename non_vlan_clients to net_clients
Stefan Hajnoczi [Tue, 24 Jul 2012 15:35:12 +0000 (16:35 +0100)]
net: Rename non_vlan_clients to net_clients

There is no longer a distinction between vlan clients and non-vlan
clients in the net core.  The net core only knows about point-to-point
clients which are connected to a peer.  It's time to rename the global
list of net clients since it no longer refers to vlans at all.

Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Zhi Yong Wu <wuzhy@linux.vnet.ibm.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
12 years agonet: Remove VLANState
Stefan Hajnoczi [Tue, 24 Jul 2012 15:35:11 +0000 (16:35 +0100)]
net: Remove VLANState

VLANState is no longer used and can be removed.

Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Zhi Yong Wu <wuzhy@linux.vnet.ibm.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
12 years agonet: Remove vlan code from net.c
Stefan Hajnoczi [Tue, 24 Jul 2012 15:35:10 +0000 (16:35 +0100)]
net: Remove vlan code from net.c

The vlan implementation in net.c has been replaced by hubs so we can
remove the code.

Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Zhi Yong Wu <wuzhy@linux.vnet.ibm.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
12 years agonet: Convert qdev_prop_vlan to peer with hub
Zhi Yong Wu [Tue, 24 Jul 2012 15:35:09 +0000 (16:35 +0100)]
net: Convert qdev_prop_vlan to peer with hub

Instead of using VLANState use net/hub.h to support the vlan qdev
property.  The vlan qdev property becomes an alias for the peer qdev
property but is represented as a VLAN ID number.  When a VLAN ID is
selected the device will really peer with a hub port.

Signed-off-by: Zhi Yong Wu <wuzhy@linux.vnet.ibm.com>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
12 years agonet: Drop vlan argument to qemu_new_net_client()
Stefan Hajnoczi [Tue, 24 Jul 2012 15:35:08 +0000 (16:35 +0100)]
net: Drop vlan argument to qemu_new_net_client()

Since hubs are now used to implement the 'vlan' feature and the vlan
argument is always NULL, remove the argument entirely and update all net
clients that use qemu_new_net_client().

Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Zhi Yong Wu <wuzhy@linux.vnet.ibm.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
12 years agohub: Check that hubs are configured correctly
Stefan Hajnoczi [Tue, 24 Jul 2012 15:35:07 +0000 (16:35 +0100)]
hub: Check that hubs are configured correctly

Checks can be performed to make sure that hubs have at least one NIC and
one host device, warning the user if this is not the case.
Configurations which do not meet this rule tend to be broken but just
emit a warning.  This patch preserves compatibility with the checks
performed by net core on vlans.

Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Zhi Yong Wu <wuzhy@linux.vnet.ibm.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
12 years agonet: Look up 'vlan' net clients using hubs
Stefan Hajnoczi [Tue, 24 Jul 2012 15:35:06 +0000 (16:35 +0100)]
net: Look up 'vlan' net clients using hubs

Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Zhi Yong Wu <wuzhy@linux.vnet.ibm.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
12 years agonet: Use hubs for the vlan feature
Stefan Hajnoczi [Tue, 24 Jul 2012 15:35:05 +0000 (16:35 +0100)]
net: Use hubs for the vlan feature

Stop using the special-case vlan code in net.c.  Instead use the hub net
client to implement the vlan feature.  The next patch will remove vlan
code from net.c completely.

Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Zhi Yong Wu <wuzhy@linux.vnet.ibm.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
12 years agonet: Add a hub net client
Stefan Hajnoczi [Tue, 24 Jul 2012 15:35:04 +0000 (16:35 +0100)]
net: Add a hub net client

The vlan feature can be implemented in terms of hubs.  By introducing a
hub net client it becomes possible to remove the special case vlan code
from net.c and push the vlan feature out of generic networking code.

Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Zhi Yong Wu <wuzhy@linux.vnet.ibm.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
12 years agonet: Add interface to bridge when SIOCBRADDIF isn't available
Corey Bryant [Thu, 12 Jul 2012 13:24:31 +0000 (09:24 -0400)]
net: Add interface to bridge when SIOCBRADDIF isn't available

The bridge helper uses the SIOCBRADDIF ioctl to add an inteface to
a bridge.  SIOCBRADDIF is not available on old Linux versions.  This
patch adds support to use the SIOCDEVPRIVATE ioctl with BRCTL_ADD_IF
if SIOCBRADDIF is not available.

Reported-by: Fabien Chouteau <chouteau@adacore.com>
Signed-off-by: Corey Bryant <coreyb@linux.vnet.ibm.com>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
12 years agoconfigure: -I\$(SRC_PATH) goes in QEMU_INCLUDES not QEMU_CFLAGS
Peter Maydell [Wed, 18 Jul 2012 14:10:27 +0000 (15:10 +0100)]
configure: -I\$(SRC_PATH) goes in QEMU_INCLUDES not QEMU_CFLAGS

If the smartcard configure check passes, add '-I\$(SRC_PATH)/libcacard'
to QEMU_INCLUDES, not QEMU_CFLAGS. Otherwise the unexpanded SRC_PATH
will cause a warning in every following configure test.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
12 years agoconfigure: Fix compile warning in utimensat/futimens test
Peter Maydell [Wed, 18 Jul 2012 14:10:26 +0000 (15:10 +0100)]
configure: Fix compile warning in utimensat/futimens test

Fix compile warning in the utimensat/futimens test ("implicit
declaration of function 'utimensat'", ditto futimens) by
adding a missing include.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Stefan Weil <sw@weilnetz.de>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
12 years agoconfigure: Fix warnings in VDE library probe
Peter Maydell [Wed, 18 Jul 2012 14:10:25 +0000 (15:10 +0100)]
configure: Fix warnings in VDE library probe

Fix compile warnings in the VDE library probe ("passing argument 1 of
'vde_open_real' discards 'const' qualifier from pointer target type",
ditto argument 2).

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
12 years agoconfigure: Fix compile warning in PNG test
Peter Maydell [Wed, 18 Jul 2012 14:10:24 +0000 (15:10 +0100)]
configure: Fix compile warning in PNG test

Fix compile warning (variable 'png_ptr' set but not used) in the
PNG detection test code.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Stefan Weil <sw@weilnetz.de>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
12 years agoconfigure: Fix errors in test for__sync_fetch_and_and
Stefan Weil [Wed, 18 Jul 2012 14:10:23 +0000 (15:10 +0100)]
configure: Fix errors in test for__sync_fetch_and_and

The old test code raises two compiler warnings which are errors since
commit 417c9d72d48275d19c60861896efd4962d21aca2.

These errors could result in compilations with compiler flag
-march486 (so all nice features of newer processors got lost).

Signed-off-by: Stefan Weil <sw@weilnetz.de>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
12 years agoconfigure: -march=i486 belongs in QEMU_CFLAGS, not CFLAGS
Peter Maydell [Wed, 18 Jul 2012 14:10:22 +0000 (15:10 +0100)]
configure: -march=i486 belongs in QEMU_CFLAGS, not CFLAGS

The distinction between QEMU_CFLAGS and CFLAGS is that the
former is for flags without which QEMU can't compile, whereas
the latter is for flags like "-g -O2" which the user can
safely override. "-march=i486" is in the former category, and
so belongs in QEMU_CFLAGS.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
12 years agoconfigure: Replace bash code by standard shell code
Stefan Weil [Wed, 18 Jul 2012 14:10:21 +0000 (15:10 +0100)]
configure: Replace bash code by standard shell code

"+=" does not work with dash and other simple /bin/sh implementations.

The new code prepends the flag while the old code either did not work
(it continued after an error message which typically was not read) or
appended the flag. That difference should not matter here.

Reported-by: Olaf Hering <olaf@aepfle.de>
Signed-off-by: Stefan Weil <sw@weilnetz.de>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
12 years agoconfigure: Fix build with capabilities
Stefan Weil [Wed, 18 Jul 2012 14:10:20 +0000 (15:10 +0100)]
configure: Fix build with capabilities

Since commit 417c9d72d48275d19c60861896efd4962d21aca2 all configure tests
normally run with -Werror. Some of these tests now fail because they
raised a compiler warning.

This patch fixes support for capabilities.

Signed-off-by: Stefan Weil <sw@weilnetz.de>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
12 years agoconfigure: Fix build with ALSA audio driver
Stefan Weil [Wed, 18 Jul 2012 14:10:19 +0000 (15:10 +0100)]
configure: Fix build with ALSA audio driver

Since commit 417c9d72d48275d19c60861896efd4962d21aca2,
all configure tests normally run with -Werror.

Some of these tests now fail because they raised a compiler warning.

Here a build breakage for ALSA (configure --audio-drv-list=alsa) is fixed.

Signed-off-by: Stefan Weil <sw@weilnetz.de>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
12 years agoconfigure: Don't run configure tests with -Werror enabled
Peter Maydell [Wed, 18 Jul 2012 14:10:18 +0000 (15:10 +0100)]
configure: Don't run configure tests with -Werror enabled

Don't run configure tests with -Werror in the compiler flags. The idea
of -Werror is that it makes problems very obvious to developers, so
they get fixed quickly. However, when running configure tests, failures
due to -Werror are far from obvious -- they simply result in the test
quietly failing when it should have passed. Not using -Werror is in
line with recommended practice in the Autoconf world.

This commit is essentially backing out the changes in commit 417c9d72.
Instead we fix the problem that commit was trying to address in a
different way: we add -Werror only for the test of the nss headers,
with a comment that this is specifically intended to detect a bug
in some releases of nss.

We also have to clean up a bug in the smartcard test where it was
trying to include smartcard_cflags in the test compile flags: this
would always result in a failure with -Werror, because they include
an escaped "$(SRC_PATH)" which is only valid when used in the final
makefile.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Stefan Weil <sw@weilnetz.de>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
12 years agoconfigure: Split valgrind test into pragma test and valgrind.h test
Peter Maydell [Mon, 30 Jul 2012 15:13:07 +0000 (16:13 +0100)]
configure: Split valgrind test into pragma test and valgrind.h test

Split the configure test that checks for valgrind into two, one
part checking whether we have the gcc pragma to disable unused-but-set
variables, and the other part checking for the existence of valgrind.h.
The first of these has to be compiled with -Werror and the second
does not and shouldn't generate any warnings.

This (a) allows us to enable "make errors in configure tests be
build failures" and (b) enables use of valgrind on systems with
a gcc which doesn't know about -Wunused-but-set-varibale, like
Debian squeeze.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
12 years agohw/9pfs: Fix assert when disabling migration
Aneesh Kumar K.V [Tue, 31 Jul 2012 07:23:18 +0000 (12:53 +0530)]
hw/9pfs: Fix assert when disabling migration

For 9p we can get the attach request multiple times for the
same export. So don't adding migration blocker for every
attach request.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
12 years agoconfigure: Fix build with capabilities
Stefan Weil [Wed, 18 Jul 2012 14:10:20 +0000 (15:10 +0100)]
configure: Fix build with capabilities

Since commit 417c9d72d48275d19c60861896efd4962d21aca2 all configure tests
normally run with -Werror. Some of these tests now fail because they
raised a compiler warning.

This patch fixes support for capabilities.

Signed-off-by: Stefan Weil <sw@weilnetz.de>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
12 years agoMerge remote-tracking branch 'mst/tags/for_anthony' into staging
Anthony Liguori [Mon, 30 Jul 2012 15:00:48 +0000 (10:00 -0500)]
Merge remote-tracking branch 'mst/tags/for_anthony' into staging

* mst/tags/for_anthony:
  msi/msix: added API to set MSI message address and data
  pci: Add INTx routing notifier
  pci: Add pci_device_route_intx_to_irq
  pci: Unregister BARs before device exit
  pci: convert PCIUnregisterFunc to void
  msix: Switch msix_uninit to return void
  msix: Allow full specification of MSIX layout
  msix: Split PBA into it's own MemoryRegion
  msix: Note endian TODO item
  msix: Move msix_mmio_read
  virtio: Convert to msix_init_exclusive_bar() interface
  ivshmem: Convert to msix_init_exclusive_bar() interface
  msix: Add simple BAR allocation MSIX setup functions
  msix: fix PCIDevice naming inconsistency
  msix: drop unused msix_bar_size, require valid bar_size

12 years agoMerge remote-tracking branch 'bonzini/nbd-next' into staging
Anthony Liguori [Mon, 30 Jul 2012 14:59:38 +0000 (09:59 -0500)]
Merge remote-tracking branch 'bonzini/nbd-next' into staging

* bonzini/nbd-next:
  qemu-nbd: add --cache and --aio options
  qemu-nbd: reorganize help message

12 years agoMerge remote-tracking branch 'bonzini/scsi-next' into staging
Anthony Liguori [Mon, 30 Jul 2012 14:59:23 +0000 (09:59 -0500)]
Merge remote-tracking branch 'bonzini/scsi-next' into staging

* bonzini/scsi-next: (32 commits)
  virtio-scsi: enable MSI-X support
  virtio-scsi: add ioeventfd support
  virtio-scsi: report parameter change events
  virtio-scsi: do not report dropped events after reset
  virtio-scsi: Report missed events
  virtio-scsi: Implement hotplug support for virtio-scsi
  scsi: report parameter changes to HBA drivers
  scsi-disk: report resized disk via sense codes
  scsi: establish precedence levels for unit attention
  scsi: introduce hotplug() and hot_unplug() interfaces for SCSI bus
  scsi: add tracepoint for scsi_req_cancel
  scsi-disk: removable hard disks support load/eject
  scsi-disk: Fail medium writes with proper sense for readonly LUNs
  scsi-disk: improve the lba-out-of-range tests for read/write/verify
  scsi-disk: rd/wr/vr-protect !=0 is an error
  scsi-disk: support toggling the write cache
  scsi-disk: parse MODE SELECT commands and parameters
  scsi-disk: fix changeable values for MODE_PAGE_R_W_ERROR
  scsi-disk: adjust offsets in MODE SENSE by 2
  scsi-disk: support emulated TO_DEV requests
  ...

12 years agoMerge commit 'quintela/migration-next-v5' into staging
Anthony Liguori [Mon, 30 Jul 2012 14:58:41 +0000 (09:58 -0500)]
Merge commit 'quintela/migration-next-v5' into staging

* commit '6c779f22a93cc6e4565b940ef616e3efc5b50ba5':
  Change ram_save_block to return -1 if there are no more changes
  ram: save_live_setup() we don't need to synchronize the dirty bitmap.
  ram: iterate phase
  ram: save_live_complete() only do one loop
  ram: save_live_setup() don't need to sent pages
  savevm: split save_live into stage2 and stage3
  savevm: split save_live_setup from save_live_state
  savevm: introduce is_active method
  savevm: Refactor cancel operation in its own operation
  savevm: remove SaveLiveStateHandler
  savevm: remove SaveSetParamsHandler
  savevm: Live migration handlers register the struct directly
  savevm: Use a struct to pass all handlers

12 years agoMerge branch pci into master
Michael S. Tsirkin [Sun, 29 Jul 2012 14:03:19 +0000 (17:03 +0300)]
Merge branch pci into master

Merge master and pci branch, resolve build breakage in hw/esp.c
introduced by f90c2bcd.

Conflicts:
hw/esp.c

12 years agox86: Fixed incorrect segment base address addition in 64-bits mode
Vitaly Chipounov [Mon, 2 Jul 2012 22:20:49 +0000 (00:20 +0200)]
x86: Fixed incorrect segment base address addition in 64-bits mode

According to the Intel manual
"Intel® 64 and IA-32 Architectures Software Developer’s Manual
Volume 3", "3.4.4 Segment Loading Instructions in IA-32e Mode":

"When in compatibility mode, FS and GS overrides operate as defined by
32-bit mode behavior regardless of the value loaded into the upper 32
linear-address bits of the hidden descriptor register base field.
Compatibility mode ignores the upper 32 bits when calculating an effective address."

However, the code misses the 64-bit mode case, where an instruction with
address and segment size override would be translated incorrectly. For example,
inc dword ptr gs:260h[ebx*4] gets incorrectly translated to:

(uint32_t)(gs.base + ebx * 4 + 0x260)
instead of
gs.base + (uint32_t)(ebx * 4 + 0x260)

Signed-off-by: Vitaly Chipounov <vitaly.chipounov@epfl.ch>
Reviewed-by: Max Filippov <jcmvbkbc@gmail.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
12 years agoplace qemu-sockets.c contributions since 2012-01-13 under GPLv2+
Paolo Bonzini [Mon, 9 Jul 2012 09:08:39 +0000 (11:08 +0200)]
place qemu-sockets.c contributions since 2012-01-13 under GPLv2+

Past contributions since 2012-01-13 were only made by Red Hat people,
so they are already available under GPLv2+.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Stefan Weil <sw@weilnetz.de>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
12 years agobuild: Fix linking failure for qemu-ga
Dunrong Huang [Sat, 28 Jul 2012 19:11:24 +0000 (03:11 +0800)]
build: Fix linking failure for qemu-ga

This patch will fix the following linking failed:

  LINK  qemu-ga
gcc: error: qga/../qapi-generated/qga-qapi-types.o: No such file or directory
gcc: error: qga/../qapi-generated/qga-qapi-visit.o: No such file or directory
gcc: error: qga/../qapi-generated/qga-qmp-marshal.o: No such file or directory
make: *** [qemu-ga] Error 1

Commit cdc976b040f2760103ce2fb49f4d504093a7163f changes the
dependencies of qemu-ga to depend "../qapi-generated/qga-qapi-types.o",
which will be expanded to "qga/../qapi-generated/qga-qapi-types.o" when
building qemu-ga.

In top-level Makefile, we defined a target "qapi-generated/qga-qapi-types.o"
which was not equal to "qga/../qapi-generated/qga-qapi-types" in the
Makefile world. So "No such file" error happened when qemu-ga was linking.

The easy approach to fix is to change the target name to
"qga/../qapi-generated/qga-qapi-types.o", but it is weird.

So, in order to solve it more graciously, I move those temporary
files(qga-qapi-*.{c,h}) qemu-ga depends on to qemu-ga/qapi-generated,
this makes dependencies more clearer.

Signed-off-by: Dunrong Huang <riegamaths@gmail.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
12 years agoAvoid redefining inline
Blue Swirl [Sun, 8 Jul 2012 07:00:50 +0000 (07:00 +0000)]
Avoid redefining inline

System headers (e.g. gutils.h) may define 'inline', #undefine it
before #defining.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
12 years agoAvoid returning void
Blue Swirl [Sun, 8 Jul 2012 06:56:53 +0000 (06:56 +0000)]
Avoid returning void

It's silly and non-conforming to standards to return void,
don't do it.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
12 years agohw/escc: Drop duplicate definition of 'disabled' property
Peter Maydell [Tue, 24 Jul 2012 14:10:27 +0000 (15:10 +0100)]
hw/escc: Drop duplicate definition of 'disabled' property

Drop a duplicate definition of the 'disabled' property from
the escc qdev property list: this redefinition is currently
effectively ignored but will become an error. (The duplication
was inadvertently introduced in 2009 in commit ec02f7dec2.)

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Acked-by: Andreas Färber <afaerber@suse.de>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
12 years agobuild: get dependency file directories from object file names
Paolo Bonzini [Wed, 11 Jul 2012 14:40:21 +0000 (16:40 +0200)]
build: get dependency file directories from object file names

After commit dcff25f2cd8c11a9368cc2369aeb0319c32d9e26, Dependency file
are taken from the directories that have a Makefile.objs file.  This is
not enough, since files can be included from other directories.
So, pick them from directories that have an object file in them.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
12 years agobuild: include qapi-generated/ files in qga/Makefile.objs
Paolo Bonzini [Wed, 11 Jul 2012 14:40:20 +0000 (16:40 +0200)]
build: include qapi-generated/ files in qga/Makefile.objs

No reason to leave them out, and it will ensure that the dependencies
are picked up.  Later we can perhaps move the files to another
directory to avoid ../ usage.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
12 years agotests: Makefile: include dependency files
Eduardo Habkost [Wed, 18 Jul 2012 17:22:27 +0000 (14:22 -0300)]
tests: Makefile: include dependency files

Otherwise 'make check' won't recompile files that need to be recompiled
because of header changes.

To reproduce the bug, run:

 $ make check  # succeeds
 $ echo B0RKED > hw/mc146818rtc_regs.h
 $ make check  # is supposed to try to rebuild tests/rtc-test.o and fail

Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
12 years agovl.c: Don't print errno after failed qemu_chr_new()
Peter Maydell [Mon, 9 Jul 2012 04:28:30 +0000 (04:28 +0000)]
vl.c: Don't print errno after failed qemu_chr_new()

The qemu_chr_new() function doesn't set errno on failure, so
don't print strerror(errno) on the error handling path when
dealing with the -serial, -parallel and -virtioconsole arguments.
This avoids nonsensical error messages like:
  $ ./arm-softmmu/qemu-system-arm -serial wombat
  qemu: could not open serial device 'wombat': Success

We also rephrase the message slightly to make it a little clearer
that we're expecting the name of a QEMU chr backend rather than
a host or guest serial/parallel/etc device.

Reported-by: Christian Müller <christian.mueller@heig-vd.ch>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
12 years agotarget-xtensa: fix big-endian BBS/BBC implementation
Max Filippov [Tue, 17 Jul 2012 19:45:23 +0000 (23:45 +0400)]
target-xtensa: fix big-endian BBS/BBC implementation

Quote from ISA, 2.1:

For most Xtensa instructions, bit numbering is irrelevant; only the BBC
and BBS instructions assign bit numbers to values on which the processor
operates. The BBC/BBS instructions use big-endian bit ordering (0 is the
most-significant bit) on a big-endian processor configuration.

Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
12 years agotarget-or32: Add testcases
Jia Liu [Fri, 20 Jul 2012 07:50:53 +0000 (15:50 +0800)]
target-or32: Add testcases

Add testcases for OpenRISC.

Signed-off-by: Jia Liu <proljc@gmail.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
12 years agotarget-or32: Add linux user support
Jia Liu [Fri, 20 Jul 2012 07:50:52 +0000 (15:50 +0800)]
target-or32: Add linux user support

Add QEMU OpenRISC linux user support.

Signed-off-by: Jia Liu <proljc@gmail.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
12 years agotarget-or32: Add linux syscall, signal and termbits
Jia Liu [Fri, 20 Jul 2012 07:50:51 +0000 (15:50 +0800)]
target-or32: Add linux syscall, signal and termbits

Add OpenRISC linux syscall, signal and termbits.

Signed-off-by: Jia Liu <proljc@gmail.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
12 years agotarget-or32: Add gdb stub support
Jia Liu [Fri, 20 Jul 2012 07:50:50 +0000 (15:50 +0800)]
target-or32: Add gdb stub support

Add OpenRISC gdb stub support.

Signed-off-by: Jia Liu <proljc@gmail.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
12 years agotarget-or32: Add system instructions
Jia Liu [Fri, 20 Jul 2012 07:50:49 +0000 (15:50 +0800)]
target-or32: Add system instructions

Add OpenRISC system instructions.

Signed-off-by: Jia Liu <proljc@gmail.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
12 years agotarget-or32: Add a IIS dummy board
Jia Liu [Fri, 20 Jul 2012 07:50:48 +0000 (15:50 +0800)]
target-or32: Add a IIS dummy board

Add a IIS dummy board.

Signed-off-by: Jia Liu <proljc@gmail.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
12 years agotarget-or32: Add timer support
Jia Liu [Fri, 20 Jul 2012 07:50:47 +0000 (15:50 +0800)]
target-or32: Add timer support

Add OpenRISC timer support.

Signed-off-by: Jia Liu <proljc@gmail.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
12 years agotarget-or32: Add PIC support
Jia Liu [Fri, 20 Jul 2012 07:50:46 +0000 (15:50 +0800)]
target-or32: Add PIC support

Add OpenRISC Programmable Interrupt Controller support.

Signed-off-by: Jia Liu <proljc@gmail.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
12 years agotarget-or32: Add instruction translation
Jia Liu [Fri, 20 Jul 2012 07:50:45 +0000 (15:50 +0800)]
target-or32: Add instruction translation

Add OpenRISC instruction tanslation routines.

Signed-off-by: Jia Liu <proljc@gmail.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
12 years agotarget-or32: Add float instruction helpers
Jia Liu [Fri, 20 Jul 2012 07:50:44 +0000 (15:50 +0800)]
target-or32: Add float instruction helpers

Add OpenRISC float instruction helpers.

Signed-off-by: Jia Liu <proljc@gmail.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
12 years agotarget-or32: Add int instruction helpers
Jia Liu [Fri, 20 Jul 2012 07:50:43 +0000 (15:50 +0800)]
target-or32: Add int instruction helpers

Add OpenRISC int instruction helpers.

Signed-off-by: Jia Liu <proljc@gmail.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
12 years agotarget-or32: Add exception support
Jia Liu [Fri, 20 Jul 2012 07:50:42 +0000 (15:50 +0800)]
target-or32: Add exception support

Add OpenRISC exception support.

Signed-off-by: Jia Liu <proljc@gmail.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
12 years agotarget-or32: Add interrupt support
Jia Liu [Fri, 20 Jul 2012 07:50:41 +0000 (15:50 +0800)]
target-or32: Add interrupt support

Add OpenRISC interrupt support.

Signed-off-by: Jia Liu <proljc@gmail.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
12 years agotarget-or32: Add MMU support
Jia Liu [Fri, 20 Jul 2012 07:50:40 +0000 (15:50 +0800)]
target-or32: Add MMU support

Add OpenRISC MMU support.

Signed-off-by: Jia Liu <proljc@gmail.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
12 years agotarget-or32: Add target stubs and QOM cpu
Jia Liu [Fri, 20 Jul 2012 07:50:39 +0000 (15:50 +0800)]
target-or32: Add target stubs and QOM cpu

Add OpenRISC target stubs, QOM cpu and basic machine.

Signed-off-by: Jia Liu <proljc@gmail.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
12 years agovirtio-scsi: enable MSI-X support
Paolo Bonzini [Fri, 27 Jul 2012 12:38:03 +0000 (14:38 +0200)]
virtio-scsi: enable MSI-X support

While virtio-scsi does support multiqueue, the default number of
interrupt vectors is not enough to actually enable usage of
multiple queues in the driver; this is because with only 2
vectors the driver will not be able to use a separate
interrupt for each request queue.  Derive the desired number
of vectors from the number of request queues.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
12 years agovirtio-scsi: add ioeventfd support
Paolo Bonzini [Fri, 27 Jul 2012 12:36:07 +0000 (14:36 +0200)]
virtio-scsi: add ioeventfd support

Probably due to bad merge months ago, virtio-scsi-pci did not have
ioeventfd support.  Fix this and enable it by default, as is the
case for other virtio-pci devices.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
12 years agovirtio-scsi: report parameter change events
Paolo Bonzini [Mon, 16 Jul 2012 12:22:52 +0000 (14:22 +0200)]
virtio-scsi: report parameter change events

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
12 years agovirtio-scsi: do not report dropped events after reset
Paolo Bonzini [Mon, 16 Jul 2012 12:50:27 +0000 (14:50 +0200)]
virtio-scsi: do not report dropped events after reset

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
12 years agovirtio-scsi: Report missed events
Paolo Bonzini [Mon, 2 Jul 2012 08:47:35 +0000 (10:47 +0200)]
virtio-scsi: Report missed events

When an event is reported but no buffers are present in the event vq,
we can set a flag and report a dummy event as soon as one is added.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
12 years agovirtio-scsi: Implement hotplug support for virtio-scsi
Cong Meng [Wed, 20 Jun 2012 06:47:11 +0000 (14:47 +0800)]
virtio-scsi: Implement hotplug support for virtio-scsi

Implement the hotplug() and hot_unplug() interfaces in virtio-scsi, by signal
the virtio_scsi.ko in guest kernel via event virtual queue.

The counterpart patch of virtio_scsi.ko will be sent soon in another thread.

Signed-off-by: Sen Wang <senwang@linux.vnet.ibm.com>
Signed-off-by: Cong Meng <mc@linux.vnet.ibm.com>
[ Add memset, fix LUN field, placate checkpatch - Paolo ]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
12 years agoscsi: report parameter changes to HBA drivers
Paolo Bonzini [Mon, 16 Jul 2012 12:22:36 +0000 (14:22 +0200)]
scsi: report parameter changes to HBA drivers

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
12 years agoscsi-disk: report resized disk via sense codes
Paolo Bonzini [Mon, 16 Jul 2012 12:07:55 +0000 (14:07 +0200)]
scsi-disk: report resized disk via sense codes

Linux will not use these, but a very similar mechanism will be used to
report the condition via virtio-scsi events.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
12 years agoscsi: establish precedence levels for unit attention
Paolo Bonzini [Mon, 16 Jul 2012 12:18:58 +0000 (14:18 +0200)]
scsi: establish precedence levels for unit attention

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: Paolo Bonzini <pbonzini@redhat.com>
12 years agoscsi: introduce hotplug() and hot_unplug() interfaces for SCSI bus
Cong Meng [Wed, 20 Jun 2012 06:47:10 +0000 (14:47 +0800)]
scsi: introduce hotplug() and hot_unplug() interfaces for SCSI bus

Add two interfaces hotplug() and hot_unplug() to scsi bus info.
The scsi bus can implement these two interfaces to signal the HBA driver
of guest kernel to add/remove the scsi device in question.

Signed-off-by: Sen Wang <senwang@linux.vnet.ibm.com>
Signed-off-by: Cong Meng <mc@linux.vnet.ibm.com>
[ Fixed braces and indentation - Paolo ]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
12 years agoscsi: add tracepoint for scsi_req_cancel
Paolo Bonzini [Fri, 27 Jul 2012 06:23:07 +0000 (08:23 +0200)]
scsi: add tracepoint for scsi_req_cancel

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
12 years agoscsi-disk: removable hard disks support load/eject
Paolo Bonzini [Mon, 16 Jul 2012 13:59:57 +0000 (15:59 +0200)]
scsi-disk: removable hard disks support load/eject

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: Paolo Bonzini <pbonzini@redhat.com>
12 years agoscsi-disk: Fail medium writes with proper sense for readonly LUNs
Ronnie Sahlberg [Mon, 16 Jul 2012 06:53:28 +0000 (08:53 +0200)]
scsi-disk: Fail medium writes with proper sense for readonly LUNs

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: Ronnie Sahlberg <ronniesahlberg@gmail.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
12 years agoscsi-disk: improve the lba-out-of-range tests for read/write/verify
Ronnie Sahlberg [Thu, 12 Jul 2012 06:52:47 +0000 (16:52 +1000)]
scsi-disk: improve the lba-out-of-range tests for read/write/verify

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: Ronnie Sahlberg <ronniesahlberg@gmail.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
12 years agoscsi-disk: rd/wr/vr-protect !=0 is an error
Ronnie Sahlberg [Thu, 12 Jul 2012 00:19:38 +0000 (10:19 +1000)]
scsi-disk: rd/wr/vr-protect !=0 is an error

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: Ronnie Sahlberg <ronniesahlberg@gmail.com>
[ Rebase after scsi_dma_reqops introduction - Paolo ]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
12 years agoscsi-disk: support toggling the write cache
Paolo Bonzini [Tue, 10 Jul 2012 13:03:17 +0000 (15:03 +0200)]
scsi-disk: support toggling the write cache

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: Paolo Bonzini <pbonzini@redhat.com>
12 years agoscsi-disk: parse MODE SELECT commands and parameters
Paolo Bonzini [Tue, 10 Jul 2012 13:02:55 +0000 (15:02 +0200)]
scsi-disk: parse MODE SELECT commands and parameters

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: Paolo Bonzini <pbonzini@redhat.com>
12 years agoscsi-disk: fix changeable values for MODE_PAGE_R_W_ERROR
Paolo Bonzini [Tue, 10 Jul 2012 12:04:51 +0000 (14:04 +0200)]
scsi-disk: fix changeable values for MODE_PAGE_R_W_ERROR

The changeable values were not all-zeros for this mode page, fix it.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
12 years agoscsi-disk: adjust offsets in MODE SENSE by 2
Paolo Bonzini [Thu, 12 Jul 2012 14:08:24 +0000 (16:08 +0200)]
scsi-disk: adjust offsets in MODE SENSE by 2

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: Paolo Bonzini <pbonzini@redhat.com>
12 years agoscsi-disk: support emulated TO_DEV requests
Paolo Bonzini [Wed, 8 Feb 2012 09:32:55 +0000 (10:32 +0100)]
scsi-disk: support emulated TO_DEV requests

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: Paolo Bonzini <pbonzini@redhat.com>