[XEN] Define -fvisibility=hidden in CFLAGS to reduce -fpic cost
when compiling for x86/64.
Signed-off-by: Keir Fraser <keir@xensource.com>

merge with xen-unstable.hg

[IA64] swiotlb clash fix

Do not define dma_supported to platform_dma_supported when building
the Xen kernel. This happend to work for DIG builds because the
swiotlb.c version of dma_supported compiles to identical code as
arch/i386/kernel/pci-dma-xen.c so the linker didn't complain.

Signed-off-by: Jes Sorensen <jes@sgi.com>

[IA64] machvec cleanups

Move the DIG specific platform_dma_foo defines to where they belong.

Signed-off-by: Jes Sorensen <jes@sgi.com>

[IA64] Add xen boot option "dom0_vcpus_pin"

This patch adds the xen boot option to pin Domain-0 VCPUs.

Signed-off-by: Kouya SHIMURA <kouya@jp.fujitsu.com>
Signed-off-by: Masaki Kanno <kanno.masaki@jp.fujitsu.com>

[XEND] Return an Internal Server Error via the Xend HTTP server.

This allows domain creation failures can be detected when done via HTTP.

Signed-off-by: Daniel P. Berrange <berrange@redhat.com>

This patch fixes return codes for the acm-related Xen management scripts
(error conditions) and addresses minor issues that 'pycheck' complains
about. It is tested with xm-test and by manually running the xm
commands. It also handles the return case (reported by Masaki Kanno)
that was missed in the originally patch.

Signed-off by: Reiner Sailer <sailer@us.ibm.com>

[XEN] Revert changeset 11438. Needs fixing for PAE.
Signed-off-by: Keir Fraser <keir@xensource.com>

[IA64][XEND] Fix getRequiredShadowMemory parameters

Parameter count changed in xen-unstable.hg 11417:65a41e3206ac

Signed-off-by: Alex Williamson <alex.williamson@hp.com>

merge with xen-unstable.hg

[IA64] Add VTI related perfc

This patch intends to add VTI-related and fw_hypercall counters.

Signed-off-by: Atsushi SAKAI <sakaia@jp.fujitsu.com>
Signed-off-by: Hiroya INAKOSHI <inakoshi.hiroya@jp.fujitsu.com>

[IA64] Cleanup for PV-on-HVM for IPF

Signed-off-by: Tsunehisa Doi <Doi.Tsunehisa@jp.fujitsu.com>
Signed-off-by: Tomonari Horikoshi <t.horikoshi@jp.fujitsu.com>

[LINUX] Move .bss declaration after .data.* in x86_64 linker script.

Older binutils (prior to 2.16) have a problem with the linker script resulting
from the change introducing explicit segment maps. Namely does the
respective linker not properly handle @nobits sections (i.e. .bss) sitting
between @progbits ones (i.e. .data.*). The .bss section must therefore
be moved past all initialized sections (as is already the case on i386).
Replacement patch attached.

From: Jan Beulich <jbeulich@novell.com>
Signed-off-by: Ian Campbell <ian.campbell@xensource.com>

[XEN] Remove backpointer info from page type info. No longer used.
Replace with a flag for PAE final L2, needed for detecting PAE L2's
which contain Xen mappings.
Signed-off-by: Keir Fraser <keir@xensource.com>

Build fix.

[LINUX] console: ensure virtual console is disabled if mfn/evtchn
not provided.
Signed-off-by: Keir Fraser <keir@xensource.com>

[HVM] pv drivers: Allocate hypercall area as an executable region.
We have to use __vmalloc() and __PAGE_KERNEL because vmalloc_exec()
and PAGE_KERNEL_EXEC are not exported to modules.
Signed-off-by: Keir Fraser <keir@xensource.com>

[XEND] balloon enough memory to enable shadow mode before live migration.a
Signed-off-by: Tim Deegan <Tim.Deegan@xensource.com>

[XEN] gnttab: Initialise maptrack->flags

__gnttab_unmap_grant_ref verifies the validity of the handle by
checking it against maptrack_limit and making sure that the flags
field at the given handle is non-zero.

This means that the flags of an unused handle should be zero in
order to prevent unused handles from getting past this check.

However, we don't seem to set it to zero to begin with when we
allocate the maptrack table.  This could potentially lead to the
corruption of the other domain's internal state, and/or the hypervisor's
internal state.

This patch makes sure that it is zeroed.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

[BLKTAP] Fix in xend to properly destroy blktap devices when deviceClass=='tap'
Submitted by Andres Lagar Cavilla

[XEN] gnttab: Read domid/flags atomically.

This more neatly matches how the two adjacent fields in
the grant-table entry are accessed by the cmpxchg()
function.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

[XEN] Fix bitops inline asm to specify that memory is clobbered.
Necessary because the modified word may not be the one directly
addressed by the memory parameter (since the parameter actually
points at an array, not a scalar value).

The change to set_bit/__set_bit is reported to fix an issue
on x460 hardware, as reported (and this fix suggested) by
Ryan Harper <ryanh@us.ibm.com>

Signed-off-by: Keir Fraser <keir@xensource.com>

[XEN] Clean up grant-table pinning logic.
Signed-off-by: Keir Fraser <keir@xensource.com>

Fix xm create -F.  This can't be very widely used, because it looks like it's
been broken for ages!

Signed-off-by: Ewan Mellor <ewan@xensource.com>

Added vmunused flag to unit test results to match recent change.

Signed-off-by: Ewan Mellor <ewan@xensource.com>

Add missing import.

Signed-off-by: Ewan Mellor <ewan@xensource.com>

This patch adds the argument check of "--ms_per_sample=" option. For
"--ms_per_sample=", a negative value should be invalid.

e.g.
% xenmon.py -n --ms_per_sample=-1
usage: xenmon.py [options]

xenmon.py: error: option --ms_per_sample: invalid negative value: '-1'

Signed-off-by: KUWAMURA Shin'ya <kuwa@jp.fujitsu.com>

This patch improves error message of xm console command.

# xm console Domain-0
Can't specify Domain-0

Signed-off-by: Masaki Kanno <kanno.masaki@jp.fujitsu.com>

This patch intends for updating "man xm" about xm info description.
For example item:system is removed from xm info.

Signed-off-by: shaocyou <shaocyou@jp.fujitsu.com>

Fix inverted sense of getRequiredAvailableMemory and
getRequiredInitialReservation on x86 HVM.

Signed-off-by: Ewan Mellor <ewan@xensource.com>

[blktap] Fix block device getsize bug in qcow when backing_filename is a block device.

Merge.

[XEN] Add more checking when reusing an active grant reference.

Signed-off-by: Steven Smith <sos22@cam.ac.uk>
Based-on-a-patch-by: Herbert Xu <herbert@gondor.apana.org.au>

Fix the memory reservation calculations.  Introduce a new architecture-specific
call to determine the initial reservation separately from the amount of
memory to be freed up by ballooning -- when using QEMU, we lose 8 MiB to video
RAM, which must be accounted for when ballooning, but _not_ accounted for
within the domain's memory allocation.

Tidy this code up using the new architecture-specific interface arrangement.

Signed-off-by: Ewan Mellor <ewan@xensource.com>

The attached patch prevents a domain with an attached vTPM from doing
local migration since this does not seem to work correctly (see tests in
xm test suite). If no vTPM has been attached, the local migration
proceeds as usual.

Signed-off-by: Stefan Berger <stefanb@us.ibm.com>

Fix spellcheck-avoiding typo: s/not/now.
(Thanks Steven S. and Muli)

Signed-off-by: Andrew Warfield <andy@xensource.com>

Fix up cset 11228.

Signed-off-by: Steven Smith <sos22@cam.ac.uk>

Update user manual to use blktap for file-based VBDs.

Signed-off-by: Andrew Warfield <andrew.warfield@xensource.com>

[IA64] Bug fix: do not crash if no FPSWA.

Signed-off-by: Tristan Gingold <tristan.gingold@bull.net>

[IA64] hwclock support

This patch supports the hwclock on Domain-0.
It adds the following EFI runtime service emulations for the
hwclock support.
 1. SetTime
 2. Get/SetWakeupTime

Signed-off-by: Masaki Kanno <kanno.masaki@jp.fujitsu.com>

Add support for netif frontends which don't support checksum offload.

Signed-off-by: Steven Smith <sos22@cam.ac.uk>

[XEN] Fix mapping of p2m into monitor tables for external-mode guests.
The previous code could map frame zero into the P2M table on PAE hosts.
Signed-off-by: Tim Deegan <Tim.Deegan@xensource.com>

Tapdisk cleanup.

Sign-off-by: Jun Koi <junkoi2004@gmail.com>

merge with xen-unstable.hg

[IA64] change arch_memory_op #def to a prototype

Signed-off-by: Alex Williamson <alex.williamson@hp.com>

[IA64] update buildconfigs

This fixes the sparse tree buildconfigs to allow xenU to build
automatically.  There's no need to force netdev front/back on from
the arch Kconfig file anymore.  Based on patch from Akio Takebe &
Yoshiro Fujita.

Signed-off-by: Alex Williamson <alex.williamson@hp.com>

[LINUX] xenoprof: use alloc_vm_area instead of get_vm_area. The former
is exported to modules.
Signed-off-by: Keir Fraser <keir@xensource.com>

[IA64] Expand memory_op for PV-on-HVM on IPF

Signed-off-by: Tsunehisa Doi <Doi.Tsunehisa@jp.fujitsu.com>
Signed-off-by: Tomonari Horikoshi <t.horikoshi@jp.fujitsu.com>

More hvm common-code cleanups.

Signed-off-by: Steven Hand <steven@xensource.com>

Fix typo.

With this patch, a checked build of Windows Server 2003 can be installed
and run on top of a VMX domain.

Signed-off-by: Eddie Dong <eddie.dong@intel.com>
Signed-off-by: Qing He <qing.he@intel.com>

Fix build w/ perfc after cset 11394:96a4379be74a

Signed-off-by: Steven Hand <steven@xensource.com>

[XEN] Improve Xen information in register dumps.
Signed-off-by: Keir Fraser <keir@xensource.com>

[XEN] Include CR2 and CR4 in register dumps.
Signed-off-by: Jan Beulich <jbeulich@novell.com>

[XENBUS] Make frontend drivers shutdown cleanly.

This patch makes frontend drivers being shutdown when devices_shutdown()
is called and walks down the device tree.  Most of it is handled by the
xenbus core, which got a new function for the bus->shutdown callback.

Signed-off-by: Gerd Hoffmann <kraxel@suse.de>

[XENBUS] Introduce new "online" node for backend drivers.

Driver changes:  Make backend drivers check it when entering Closed
state, only unregister devices when the "online" node is either "0" or
not present.  The later maintains backward compatibility.

Tools changes:  Update "online" node when creating and hot-unplugging
devices.

Background: When kexec'ing a new kernel, the old kernel's frontend
driver will shutdown, but the backend device should *NOT* disappear
due to that, so the new kernel instance can reconnect.

Signed-off-by: Gerd Hoffmann <kraxel@suse.de>

Add xenbus_strstate() helper function for more readable debug output.

From: Gerd Hoffmann <kraxel@suse.de>
Signed-off-by: Keir Fraser <keir@xensource.com>

[HVM] Add a mmio decode for 0F BA /4 for HVM guest.
Signed-off-by: Xiaohui Xin <xiaohui.xin@intel.com>

[HVM][VMX] Add CR8 virtualization. Needed for x86/64 Windows boot.
Signed-off-by: Xiaohui Xin <xiaohui.xin@intel.com>

[HVM][VMX] Report MCE/MCA features in CPUID.
Required for x86/64 Windows guest.
Signed-off-by: Xiaohui Xin <xiaohui.xin@intel.com>

[HVM][SVM] Remove redundant do_softirq in the VMEXIT_INTR handler.

Signed-off-by: Tom Woller <thomas.woller@amd.com>

[XEN] Clean up BASE_DISALLOW_MASK and fix for x86/64.
Original patch from Xiaohui Xin <xiaohui.xin@intel.com>
Signed-off-by: Keir Fraser <keir@xensource.com>

[XEN] Use vcpu_info pointer in vcpu structure.
Signed-off-by: Jan Beulich <jbeulich@novell.com>

[XEN] Remove unused IS_INVALID_M2P_ENTRY macro.
Signed-off-by: Jan Beulich <jbeulich@novell.com>

[XEN] interfaces: Remove explicit alignment/padding from domctl/sysctl hypercalls.

Generation of 32-on-64 compat shims is being semi-automated for all
other hypercalls. It seems sensible to teat all hypercalls uniformly
and do the same for the control-stack management interfaces too.

Signed-off-by: Keir Fraser <keir@xensource.com>

[TOOLS] xenctrl.h defines __XEN_TOOLS__ if it's not already.
Signed-off-by: Keir Fraser <keir@xensource.com>

[LINUX] Remove vm_map_xen_pages(), clean up __direct_remap_page_range.
Signed-off-by: Keir Fraser <keir@xensource.com>

[XEN] Fix shadow linear-mapping recopy code for SMP PAE HVM on 64bit Xen.
Signed-off-by: Tim Deegan <Tim.Deegan@xensource.com>

[MINIOS] Declare a __XEN_INTERFACE_VERSION__ to fix the build

[LINUX] /proc/iomem should not be initialised from pseudophysical e820 map.
This is because it should only contain real I/O memory address ranges,
otherwise we can get false conflicts between a real I/O range and a
'pseudophysical' RAM range, causing drivers to fail to work in a domU.
Also prevent kernel code/data being requested as iomem resources --
they are not contiguous in 'io memory' space so this would not make
sense.

Signed-off-by: Keir Fraser <keir@xensource.com>

Merge with xenppc-unstable-merge.hg.

Merge with xen-ia64-unstable.hg

[POWERPC] merge with xen-unstable.hg

[NET] front: move register_netdev() + sysfs attribute registering from create_netdev() into the new open_netdev() function.
open_netdev is called as very last thing in netdev_probe.

This closes at least one initialization race: uninitialized
mac address is visible for a short time in
/sys/class/net/${if}/address

Signed-off-by: Gerd Hoffmann <kraxel@suse.de>

[XEN] Add a start_pfn field in xen_domctl_getmemlist.
This is used by ia64 because it directly reads the P2M table.
Signed-off-by: Tristan Gingold <tristan.gingold@bull.net>

Revert powerpc build break after dom0_ops split.
Signed-off-by: Hollis Blanchard <hollisb@us.ibm.com>

[XEN] Reserve hypercall number for kexec operations.
Signed-off-by: Keir Fraser <keir@xensource.com>

[XEN] physdev: Check for negative irqs in alloc_irq_vector.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

[LINUX][X86_64] P2M map must be allocated on page aligned boundary.
Otherwise save/restore can break.
Signed-off-by: Kazuo Moriwaka <moriwaka@valinux.co.jp>

[TOOLS] Fix gdbserver build.
Signed-off-by: Travis Betak <travis.betak@amd.com>

Fix fix :-)

Signed-off-by: Steven Hand <steven@xensource.com>

Fix hard tabs.

Signed-off-by: Steven Hand <steven@xensource.com>

Block IO safety patch for live migration. Xend checks that VBDs are destroyed before unpausing the domain.

Further shadow_mem_control fixes -- we need to balloon the amount of memory
that we're going to request i.e. the rounded-up value.

Signed-off-by: Ewan Mellor <ewan@xensource.com>

Shadow memory should be in MiB, not KiB.  It also needs rounding up to be safe.

Signed-off-by: Ewan Mellor <ewan@xensource.com>

Fix reboot and reconfigure of HVM guests when they are configured with an
empty CD-ROM drive.

Signed-off-by: Ewan Mellor <ewan@xensource.com>

Added an entry to arch._types for x86_64.

Signed-off-by: Ewan Mellor <ewan@xensource.com>

Fix typo.

Signed-off-by: <andy@xensource.com>

merge

Small blktap cleanups.

Signed-off-by: Andrew Warfield <andy@xensource.com>

[XEN][POWERPC] catch up with xen-unstable changes
Signed-off-by: Hollis Blanchard <hollisb@us.ibm.com>

[XEND] abstract architecture-specific bits in image.py
- create arch.type (which evaluates to "x86", "ia64" or "powerpc")
- create subclasses for x86 and ia64 HVM loaders
- rework findImageHandlerClass()
Signed-off-by: Hollis Blanchard <hollisb@us.ibm.com>

This patch fixed 'xm reboot' to work as you expected with the wait
option. This patch adds a restart sequence counter to xenstore.
The completion of a VM reboot is detected by observing that
its restart sequence counter changes.

Signed-off-by: Masaki Kanno <kanno.masaki@jp.fujitsu.com>

Improve xm migrate error message by printing usage instead of "Invalid
arguments".

Signed-off-by: Masaki Kanno <kanno.masaki@jp.fujitsu.com>

Reduce AIO context allocations in blktap.

Signed-off-by: Andrew Warfield <andy@xensource.com>

[IA64] revert xen-ia64-unstable.hg cset 11301:a19dbbe4cff5

revert arch_domain_destory() for old PV-on-HVM on IPF

Signed-off-by: Tsunehisa Doi <Doi.Tsunehisa@jp.fujitsu.com>
Signed-off-by: Tomonari Horikoshi <t.horikoshi@jp.fujitsu.com>

[IA64] revert xen-ia64-unstable.hg 11271:c232365128cf

Delete functions introduced for old PV-on-HVM on IPF.

Signed-off-by: Tsunehisa Doi <Doi.Tsunehisa@jp.fujitsu.com>
Signed-off-by: Tomonari Horikoshi <t.horikoshi@jp.fujitsu.com>

merge with xen-unstable.hg

[IA64] init the hlt timer on a valid cpu and migrate

Based on a patch from Anthony Xu, v->processor is not initialized
in alloc_vcpu_struct() when we initialize the hlt_timer.  Init the
timer on the first cpu and migrate it in schedule_tail().

Signed-off-by: Alex Williamson <alex.williamson@hp.com>

Fix definition of LINUX_VER so that doesn't pickup LINUX_VER3
from mk.linux-2.6-xen. This makes "make mkpatches" work again.

Signed-off-by: Ian Campbell <ian.campbell@xensource.com>

[XEN][POWERPC] cleanup for (;;);

When the port first started, we didn't even have panic() working
properly so we used "for(;;);" loops to stop execution.  Now that output
is buffered these look like hangs, since you usually do not see the
printk() before.

Signed-off-by: Jimi Xenidis <jimix@watson.ibm.com>
Signed-off-by: Hollis Blanchard <hollisb@us.ibm.com>