malc [Sat, 21 Feb 2009 05:48:19 +0000 (05:48 +0000)]
Properly handle pthread_cond_timedwait timing out
pthread_cond_timedwait is allowed to both consume the signal and
return with the value indicating the timeout, hence predicate should
always be (re)checked before taking an action
malc [Sat, 21 Feb 2009 05:48:15 +0000 (05:48 +0000)]
Avoid thundering herd problem
Broadcast was used so that the I/O threads would wakeup, reset their
ts values and all but one go to sleep, in other words an optimization
to prevent threads from exiting in presence of continuing I/O
activity. Spurious wakeups make the looping around cond_timedwait with
ever reinitialized ts potentially unsafe and as such ts in no longer
reinitilized inside the loop, hence switch to signal is warranted and
this benefits of this particlaur optimization are lost.
(It's worth noting that timed variants of pthread calls use realtime
clock by default, and therefore can hang "forever" should the host
time be changed. Unfortunatelly not all host systems QEMU runs on
support CLOCK_MONOTONIC and/or pthread_condattr_setclock with this
value)
aurel32 [Wed, 18 Feb 2009 21:37:17 +0000 (21:37 +0000)]
Fix cpu_physical_memory_rw() for 64-bit I/O accesses
KVM uses cpu_physical_memory_rw() to access the I/O devices. When a
read or write with a length of 8-byte is requested, it is split into 2
4-byte accesses.
This has been broken in revision 5849. After this revision, only the
first 4 bytes are actually read/write to the device, as the target
address is changed, so on the next iteration of the loop the next 4
bytes are actually read/written elsewhere (in the RAM for the graphic
card).
This patch fixes screen corruption (and most probably data corruption)
with FreeBSD/amd64. Bug #2556746 in KVM bugzilla.
aliguori [Mon, 16 Feb 2009 14:59:35 +0000 (14:59 +0000)]
Fix hardware accelerated video to video copy on Cirrus VGA (Brian Kress)
cirrus_do_copy() in hw/cirrus_vga.c seems to make some incorrect assumptions
about video memory layout. It tries to convert addresses to coordinates
assuming that one row of data is (width * depth) bytes long. The correct way
seems to be to use the pitch fields in the CirrusVGAState structure instead.
Without this patch, I get lots of screen corruption when I try to drag a window
under X as it's passing the wrong coordinates to the display surface for the
copy. With this patch I can drag a window with no screen corruption.
Signed-off-by: Brian Kress <kressb@moose.net> Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@6622 c046a42c-6fe2-441c-8c8c-71466251a162
aliguori [Mon, 16 Feb 2009 14:59:30 +0000 (14:59 +0000)]
Support multiple VNC clients (Brian Kress)
Change structure associated with a display from VncState to a new structure
VncDisplay. Remove client specific fields from VncDisplay. Remove display
specific fields from VncState. Maintain a linked list of VncStates per
VncDisplay structure, update as necessary. When updates/resizes/copies come in
from the hardware, dispatch to all clients.
Signed-off-by: Brian Kress <kressb@moose.net> Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@6621 c046a42c-6fe2-441c-8c8c-71466251a162
aliguori [Wed, 11 Feb 2009 21:00:43 +0000 (21:00 +0000)]
KVM: cpuid function 2: store all values (Amit Shah)
Incrementing the array index was missed in the previous series which causes
us to not store all the values.
Signed-off-by: Amit Shah <amit.shah@redhat.com> Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@6617 c046a42c-6fe2-441c-8c8c-71466251a162
aliguori [Wed, 11 Feb 2009 21:00:32 +0000 (21:00 +0000)]
Fix qemu_realloc() (Kevin Wolf)
For qemu_realloc with size == 0 a result of NULL is perfectly fine
Signed-off-by: Kevin Wolf <kwolf@suse.de> Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@6615 c046a42c-6fe2-441c-8c8c-71466251a162
It returns the domain, bus and slot for the newly added device on success.
It is possible to attach a disk to a device after PCI initialization via
the drive_add command. If so, a manual scan of the SCSI bus on the guest
is necessary.
Save QEMUMachine necessary for drive_init.
Add monitor command to hot-remove devices, remove device data on _EJ0 notification.
aurel32 [Mon, 9 Feb 2009 16:50:00 +0000 (16:50 +0000)]
target-ppc: Model e500v{1,2} CPUs more accurately
The e500v1 chips only have single-precision floating point; don't say we
support the double-precision floating-point instructions on such chips.
Also add an e500v1 -cpu argument for a generic e500v1.
aurel32 [Mon, 9 Feb 2009 16:49:50 +0000 (16:49 +0000)]
target-ppc: Model SPE floating-point instructions more accurately
Single-precision and double-precision floating-point instructions should
be separated into their own categories, since some chips only support
single-precision instructions.
aliguori [Mon, 9 Feb 2009 15:50:36 +0000 (15:50 +0000)]
KVM: Get all cpuid values from function 2 (Amit Shah)
cpuid function 2 can have multiple values to describe cache behaviour.
Loop till we have fetched all the values.
Signed-off-by: Amit Shah <amit.shah@redhat.com> Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@6567 c046a42c-6fe2-441c-8c8c-71466251a162
CPUID functions 4, 0xb and 0xd have sub-leaf values which depend on the
input value of ECX. Store these values as well.
Signed-off-by: Amit Shah <amit.shah@redhat.com> Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@6566 c046a42c-6fe2-441c-8c8c-71466251a162
aliguori [Mon, 9 Feb 2009 15:50:08 +0000 (15:50 +0000)]
KVM: CPUID takes ecx as input value for some functions (Amit Shah)
The CPUID instruction takes the value of ECX as an input parameter
in addition to the value of EAX as the count for functions 4, 0xb
and 0xd. Make sure we pass the value to the instruction.
Also convert to the qemu-style whitespace for the surrounding code.
Signed-off-by: Amit Shah <amit.shah@redhat.com> Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@6565 c046a42c-6fe2-441c-8c8c-71466251a162
aliguori [Fri, 6 Feb 2009 00:15:19 +0000 (00:15 +0000)]
Revert changes to fmopl.c
fmopl.c was taken from MAME and doesn't include QEMU header files so we cannot
use qemu_malloc in it. It happens to build because C is a silly language.
Unfortunately, it doesn't play nicely with the QEMU headers so lets just revert
the changes that were made to it.
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@6541 c046a42c-6fe2-441c-8c8c-71466251a162
Use the control virtqueue to allow the guest to enable and manipulate
a VLAN filter table. This allows us to drop more packets the guest
doesn't want to see. We define a new VLAN class for the control
virtqueue with commands ADD and DEL with usage defined in virtio-net.h.
Signed-off-by: Alex Williamson <alex.williamson@hp.com> Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@6540 c046a42c-6fe2-441c-8c8c-71466251a162
aliguori [Thu, 5 Feb 2009 22:36:28 +0000 (22:36 +0000)]
qemu:virtio-net: Add additional MACs via a filter table (Alex Williamson)
Create a filter table and allow the guest to populate it with the
MAC class control commands. We manage the size and usage of the
filter table including enabling promiscuous and all-multi modes
as necessary. The guest should therefore assume the table is
infinite. Eventually this might allow us to bind directly to a
hardware NIC and manipulate a physical MAC filter.
The specifics of the TABLE_SET command are documented in
virtio-net.h. Separate buffers in the same command are used
for unicaste and multicast addresses for priority and
sychronization. With this we can export the VIRTIO_NET_F_CTRL_RX
feature bit.
Signed-off-by: Alex Williamson <alex.williamson@hp.com> Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@6539 c046a42c-6fe2-441c-8c8c-71466251a162
aliguori [Thu, 5 Feb 2009 22:36:20 +0000 (22:36 +0000)]
qemu:virtio-net: Add promiscuous and all-multicast mode bits (Alex Williamson)
Add a new RX_MODE control virtqueue class with commands PROMISC and
ALLMULTI and usage documented in virtio-net.h allowing the guest to
manipulate packet receiving options. We don't export a feature for
this until we also add the MAC filter table.
Note, for compatibility with older guest drivers we need to default
to promiscuous.
Signed-off-by: Alex Williamson <alex.williamson@hp.com> Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@6537 c046a42c-6fe2-441c-8c8c-71466251a162
aliguori [Thu, 5 Feb 2009 22:36:16 +0000 (22:36 +0000)]
qemu:virtio-net: Add a virtqueue for control commands from the guest (Alex Williamson)
This will be used for RX mode, MAC table, VLAN table control, etc...
The control transaction consists of one or more "out" sg entries and
one or more "in" sg entries. The first out entry contains a header
defining the class and command. Additional out entries may provide
data for the command. A response via the ack entry is required
and the guest will typically be waiting for it.
Signed-off-by: Alex Williamson <alex.williamson@hp.com> Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@6536 c046a42c-6fe2-441c-8c8c-71466251a162
projects / xenclient / ioemu.git / log