[IA64] Speedup ptc.e emulation

This patch makes ptc.e emulation faster.
It defers the initialization of collision chain area of VHPT.
Also removes an unused field from struct thash_cb.

Signed-off-by: Kouya Shimura <kouya@jp.fujitsu.com>

[IA64] Fix a memory allocation bug in MCA

Signed-off-by: Kazuhiro Suzuki <kaz@jp.fujitsu.com>

[IA64] Fix VTi domain shutdown

Event channel destruction is now done earlier, this patch does
the corresponding modification in IA64 side. It fixes dom0 crash
when VTI destroyed.

Signed-off-by: Zhang Xin <xing.z.zhang@intel.com>

[IA64] Fixup physinfo

Use max cpus per node to guess at sockets per node.  This avoids
averaging problems with offline cpus and nodes without cpus.  Also
fill in the cpu_to_node array.

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

merge with xen-unstable.hg

xend: Fix xm pause/unpause for inactive managed domains.
Signed-off-by: Masaki Kanno <kanno.masaki@jp.fujitsu.com>

xend: Count CPUs rather than relying on a static
nodes/sockets/cores/threads relationship.
Signed-off-by: Alex Williamson <alex.williamson@hp.com>

xm: Improve help message for vif creation.
Signed-off-by: Masaki Kanno <kanno.masaki@jp.fujitsu.com>

Simplify Unisys ES7000 code in light of the fact we do not support
legacy boxes with very non-standard APIC setup.
From: Raj Subrahmanian <raj.subrahmanian@unisys.com>
Signed-off-by: Keir Fraser <keir@xensource.com>

xend: Missing last line from Xen_API security patch.
Signed-off-by: Stefan Berger <stefanb@us.ibm.com>

Add accel option to vif xend config.
Signed-off-by: Kieran Mansley <kmansley@solarflare.com>

hvm: Clean ups and fix MSR access functions.
Signed-off-by: Xin Li <xin.b.li@intel.com>
Signed-off-by: Keir Fraser <keir@xensource.com>

[IA64] Support special guest optimizations in the hypervisor

For PV add a new hypercall __HYPERVISOR_opt_feature, which lets dom0/domU
switch on/off special optimization features in the hypervisor.
For HVM this is done in set_os_type().

Signed-off-by: Dietmar Hahn <dietmar.hahn@fujitsu-siemens.com>

[IA64] Rename NUM_MEM_DESC to NUM_EXTRA_MEM_DESC and reduce it.

Signed-off-by: Isaku Yamahata <yamahata@valinux.co.jp>

[Xen-API] Extension of the Xen-API for managing Xen Security Policies

This patch implements extensions for managing security policies in
xend. The XSPolicy and ACMPolicy classes provide the interface for the
Xen-API and implement functionality for setting, updating and
activating of a Xen security policy as well as labeling of virtual
machines and resources such as block devices. Labeling of network
devices will follow.

The acmpolicy class implements a compiler for translating an XML
policy into their binary format and provides functionality for
comparing a current policy against a new one when changing/updating a
policy.

The xspolicyadmin class administers the policy of the system.

Some of the xend-internal code deals with transforming the labeling
information from the S-Expression format into the new Xen-API
format. This is similar to much of the other code that is already
there.

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

Introduce 'bootscrub' boolean cmdline option. Default on.

no-bootscrub will disable boot-time RAM scrubbing, speeding up boot
time significantly on big-memory systems.

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

Also allow boolean cmdline params to be inverted in two other ways.

Now, a standard boolean (e.g., watchdog) can be inverted in three
ways:
 1. no-watchdog
 2. watchdog=no
 3. watchdog=off

Stacked inversions cancel each other: no-watchdog=no turns on the
watchdog.

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

Allow inversion of boolean cmdline parameters with 'no-' prefix.
Signed-off-by: Keir Fraser <keir@xensource.com>

Remove no-pv-compat cmdline option.
Signed-off-by: Keir Fraser <keir@xensource.com>

Add support for building -mm, -tip and -rc upstream trees.

(note that not all of those contains Xen support just yet and that a
native kernel will be silently built in that case.)

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

Add support to the tarball source type for using the ketchup tool.

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

Add hooks for a script to update kernel configuration after tree has been
prepared. This is to support upstream sources which do not enable Xen in their
default configurations.

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

Remove .hgskip. This was an ancient mercurial patch which was never
taken upstream.

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

VMX: Fix function return code when a GP fault is generated on CR accesses.
Signed-off-by: Xin Li <xin.b.li@intel.com>

Fix xm save command for a inactive managed domain.
Signed-off-by: Masaki Kanno <kanno.masaki@jp.fujitsu.com>

VMX: No VMExits on SYSENTER MSR accesses since VMX already has VMCS
fields for both host and guest state.
Signed-off-by: Xin Li <xin.b.li@intel.com>

Better error message for 'xm sysrq' on inactive domain.
Signed-off-by: Masaki Kanno <kanno.masaki@jp.fujitsu.com>

SVM: Sync with VMX code changes.
Signed-off-by: Keir Fraser <keir@xensource.com>

vms: A cleanup to vmx.c:
1) move some functions from vmx.h to vmx.c, since they are private.
2) adjust vmx_set_cr0 handling.
3) majorly coding style tidying.
Signed-off-by: Xin Li <xin.b.li@intel.com>
Signed-off-by: Keir Fraser <keir@xensource.com>

Specify -fno-strict-aliasing in root build config file.

Fixes correctness issues with xenstored and gcc-4.2.

Original patch by Charles Coffing <ccoffing@novell.com>

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

Merge with PPC tree.

[POWERPC][XEN] Stop sharing the MPIC between dom0 and Xen.
Xen is now the sole owner of the MPIC. Requires Xen Linux 2.6.18.
Signed-off-by: Hollis Blanchard <hollisb@us.ibm.com>

[POWERPC][XEN] Increase foreign mapping size to accommodate Linux 2.6.18 usage.
Signed-off-by: Christian Ehrhardt <ehrhardt@linux.vnet.ibm.com>
Signed-off-by: Hollis Blanchard <hollisb@us.ibm.com>

[POWERPC][XEN] Quiet time debug messages.
Signed-off-by: Hollis Blanchard <hollisb@us.ibm.com>

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

[PPC] Add missing '{' in conditional statement.
Signed-off-by: Jerone Young <jyoung5@us.ibm.com>

hvm: Add global default keymap setting

Add global default keymap setting to be used when a VM does not have
a keymap configuration set.

Signed-off-by: Pat Campbell <plc@novell.com>

Extended the physinfo sysctl to export NUMA cpu_to_node topology info.
Print this in 'xm info'.
Signed-off-by: Ryan Harper <ryanh@us.ibm.com>
Signed-off-by: Keir Fraser <keir@xensource.com>

Revert 15471:7ac7f147241405af83e7a9d748cf7b01279734fc

Block-device specifiers in ioemu can contain colons, so skipping
always past the first colon is not a good idea. Better solutions are
in the pipeline to solve the blktap issues.

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

docs: update xm man page

- Fixed description of "Mem" column in "xm list" output.
- Added a bit of text for the credit scheduler.
- Described the --force option to block-detach.
- Made formatting and spelling more consistent.
- etc...

Signed-off-by: Charles Coffing <ccoffing@novell.com>

[ACM] Support for initial unicode handling of strings in policies.
Signed-off-by: Stefan Berger <stefanb@us.ibm.com>

Export debugger attached state to xc API.
Signed-off-by: Tristan Gingold <tgingold@free.fr>

xend: Fix xm block/network-attach command for inactive managed domain
Signed-off-by: Masaki Kanno <kanno.masaki@jp.fujitsu.com>

xend: Fix op_domain_sched_credit_set() in SrvDomain.py
This patch adds "cap" into op_domain_sched_credit_set().
Signed-off-by: Masaki Kanno <kanno.masaki@jp.fujitsu.com>

vmx: Fix a typo.
Signed-off-by: Xin Li <xin.b.li@intel.com>

hvm: Fix vnc keymap for Japanese keyboards.
Signed-off-by: Takanori Kasai <kasai.takanori@jp.fujitsu.com>
Signed-off-by: Junko Ichino <ichino.junko@jp.fujitsu.com>

Use short name format when reference to vcpu vmx union member.
Signed-off-by: Xin Li <xin.b.li@intel.com>

Improve xm sched-credit command for inactive managed domains
Allows to change the scheduler parameters of inactive managed domains.
Signed-off-by: Masaki Kanno <kanno.masaki@jp.fujitsu.com>

hvm: Allow blktap to be able to be booted as system volume for PV-on-HVM

When blktap was booting system volume, PV-on-HVM domain was not able
to be  started.

The configuration file is specified as follows.
disk = [ 'tap:aio:/xen/test/rhel5ga_full.img,hda,w' ]

The error occurred by the initialization of system volume in qemu-dm.
qemu: could not open hard disk image 'aio:/xen/test/rhel5ga_full.img'

Signed-off-by: Takanori Kasai <kasai.takanori@jp.fujitsu.com>
Signed-off-by: Tomonari Horikoshi <t.horikoshi@jp.fujitsu.com>
Signed-off-by: Tsunehisa Doi <Doi.Tsunehisa@jp.fujitsu.com>
Signed-off-by: Keir Fraser <keir@xensource.com>

[HVM] ROMBIOS: reboot when we see Ctrl+Alt+Del
Signed-off-by: Tim Deegan <Tim.Deegan@xensource.com>

[IA64] Removed unused __get_free_pages() and __free_pages()

Signed-off-by: Isaku Yamahata <yamahata@valinux.co.jp>

[IA64] cleanup: remove dead code.

Signed-off-by: Tristan Gingold <tgingold@free.fr>

[IA64] Minor reformating and comments for vmx_fault.c

Signed-off-by: Tristan Gingold <tgingold@free.fr>

[IA64] Renames vmx_process.c into vmx_fault.c

Signed-off-by: Tristan Gingold <tgingold@free.fr>
--HG--
rename : xen/arch/ia64/vmx/vmx_process.c => xen/arch/ia64/vmx/vmx_fault.c

[IA64] Makefile fix: do not always redo .S targets.

Signed-off-by: Tristan Gingold <tgingold@free.fr>

[IA64] Clear the key part of cr.itir

Clear the key part of cr.itir before itc.X in alt_dtlb_miss(), alt_itlb_miss()
and frametable_miss().  Preparation for using protection keys.

Signed-off-by: Dietmar Hahn <dietmar.hahn@fujitsu-siemens.com>

[IA64] Fix soft lock up caused by xen_timer_interrupt()

This patch intends to fix softlockup caused by xen_timer_interrupt().
This is caused by local_cpu_data->itm_next and stime_irq, itc_at_irq
inconsistency at CPU0 of hypervisor.  This patch sets stime_irq and
itc_at_irq every time in xen_timer_interrupt() to avoid this soft
lock up.

In other words, it is caused by competition of local_cpu_data->itm_next
and domain_itm in xen_timer_interrupt() and reprogram_timer() (more
specific vcpu_set_next_timer()).

For example:
 1) reprogram_timer() runs and set local_cpu_data->itm_next and set
    domain_itm as next itm.
 2) xen_timer_interrupt() called but following condition is not satisfied:
    while(time_after(ia64_get_itc(), local_cpu_data->itm_next)
    This skips stime_irq and itc_at_irq setting.
 3) goto 1)
 4) sometimes local_cpu_data->itm_next is rollback because
    ns_to_cycle()/IA64 is representing almost 32bit.
    (This occured at reprogram_timer())
 5) It causes soft lock up.
 6) Hypervisor returns to work(not hang).

To reproduce this issue, I do following configuration.

 1) boot Xen with pcpu=4 and Dom0 with vcpu=4
 2) boot domU1 with vcpu with vcpu-pin 0-1
 3) boot domU2 with vcpu with vcpu-pin 0-1
 4) run yes > /dev/null  2 process on domU1
 5) run nothing on domU2(to check softlock up occured or not)
 6) run kernel compile with -j4 on Dom0 continuously
 7) wait 4 or 8 hours to occur softlockup.

Signed-off-by: Atsushi SAKAI <sakaia@jp.fujitsu.com>

[HVM] Shadow: avoid xen crash if guest uses special memory for pagetables
(just crash the guest and don't do any more PTE propagations).
Signed-off-by: Tim Deegan <Tim.Deegan@xensource.com>

[HVM] update Etherboot ROM: new version, stricter DHCP compliance.
Signed-off-by: Tim Deegan <Tim.Deegan@xensource.com>

vmx: Write back modified interruptibility-info in enable_intr_window().
Signed-off-by: Keir Fraser <keir@xensource.com>

hvm vmx: Support 'virtual NMI' feature of VMX.
Signed-off-by: Haitao Shan <haitao.shan@intel.com>
Signed-off-by: Eddie Dong <eddie.dong@intel.com>
Signed-off-by: Dexuan Cui <dexuan.cui@intel.com>
Signed-off-by: Keir Fraser <keir@xensource.com>

xen: Fix event-channel destruction.

Lots of churn to fix a couple of bugs:

 1. If we tried to close an interdomain channel, and the remote domain
    is being destroyed, we immediately return success. But the remote
    domain will place our port in state 'unbound', not 'free'. Hence
    the port is effectively leaked.

 2. If two domains are being destroyed at the same time, and share an
    interdomain channel, the second to attempt the close may
    dereference an invalid domain pointer. Aiee!

Doing som ework to be able to destroy event-channel context much
earlier in domain death was the civilised thing to do.

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

[x86/32] Support two byte CMOVcc opcodes in 4G segment fixup.

These are present in Debian Etch's libc6-i686 package.

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

x86: Clean up do_iret() hypercall.
Signed-off-by: Keir Fraser <keir@xensource.com>

xenstore: Remove broken and unmaintained test code.
Signed-off-by: Keir Fraser <keir@xensource.com>

Remove over-zealous assertion.

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

[XEN] Fix p2m->shadow callback to pass the mfn being written to
as well as the pointer and contents.  This was being calculated but
got disconnected from its use when the p2m and shadow functions were
separated.
Signed-off-by: Tim Deegan <Tim.Deegan@xensource.com>

xen: Fix my changes to I/O-region-finding routine.
Signed-off-by: Keir Fraser <keir@xensource.com>

xen: Remove useless check from dom0 builder.
Signed-off-by: Keir Fraser <keir@xensource.com>

x86: Improve e820 scanning for all I/O regions.
Signed-off-by: Keir Fraser <keir@xensource.com>

hvm vlapic: Fix one_shot argument passed to create_periodic_time().
Signed-off-by: Yosuke Iwamatsu <y-iwamatsu@ab.jp.nec.com>

x86-64: clear DF for kernel when forwarding syscall.
Signed-off-by: Jan Beulich <jbeulich@novell.com>

x86-64: bump STACK_SIZE to 32 so that trampoline and IST stacks fit
without undue squeezing.
Signed-off-by: Jan Beulich <jbeulich@novell.com>

xen: Fix kexec_crash_area reservation/allocation.
Signed-off-by: Keir Fraser <keir@xensource.com>

merge

[32on64 kexec] Add an explicit local branch after re-enabling paging
in 32 bit mode.

While it doesn't appear to be strictly required the documentation
calls for it. Thanks to Edwin Zhai for bringing the omission to my
attention.

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

vmx: Use per-cpu vmcs pointer to avoid unnecessary vmcs loading in
vmx_do_resume().
Signed-off-by: Eric Liu <eric.e.liu@intel.com>

xend: Save modified scheduler params of managed guests to config.sxp.
Signed-off-by: Masaki Kanno <kanno.masaki@jp.fujitsu.com>

xendomains: Better sanity checking of potetntial save records.
Signed-off-by: Gawain Lynch <gawain.lynch@gmail.com>

vmxassist: Decode mov instruction in protected to real mode in vmxassist.
Also some clean up.
Signed-off-by: Xin Li <xin.b.li@intel.com>

hvm: Fix keymap for Windows keys.
Signed-off-by: Takanori Kasai <kasai.takanori@jp.fujitsu.com>
Signed-off-by: Junko Ichino <ichino.junko@jp.fujitsu.com>

[IA64] Better warning fix

This fixes it for gcc-3.x too.

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

[IA64] Typo fix

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

merge with xen-unstable.hg

[IA64] Build warning cleanup

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

[IA64] Cleanup: Move is_platform_hp_ski() from xenmisc.c to xensetup.c

- only caller is start_kernel
- change to static __init
- also move running_on_sim to xensetup.c, and change it from unsigned
  long to int, since it's just a boolean
- declare running_on_sim in config.h near some other externs

Tested by building, booting, starting a PV guest on rx2620.

Signed-off-by: Aron Griffis <aron@hp.com>

[IA64] frametable_miss only exists when CONFIG_VIRTUAL_FRAME_TABLE

Tested by building without CONFIG_VIRTUAL_FRAME_TABLE

Signed-off-by: Aron Griffis <aron@hp.com>

[IA64] Add ia64 ignores

These files are created or modified during the build.

Signed-off-by: Aron Griffis <aron@hp.com>

[IA64] cleanup: remove unused code.

Signed-off-by: Tristan Gingold <tgingold@free.fr>

[IA64] Domain debugger for VTi: virtualize ibr and dbr.

Misc cleanup.

Signed-off-by: Tristan Gingold <tgingold@free.fr>

[IA64] Cleanup: remove cpu_user_regs from arch-ia64.h

Add a constant for isr.ni

Signed-off-by: Tristan Gingold <tgingold@free.fr>

[IA64] Fix NVRAM data cannot be saved when guest execute "reboot"

Also fix "xm reboot" instruction in Xend.

Signed-off-by: Zhang Xin <xing.z.zhang@intel.com>

[IA64] Don't use hard wired privilege level 2 for domain kernel

PV domains now use CPL 1 instead of 2 for CPL 0 emulation

Signed-off-by: Juergen Gross <juergen.gross@fujitsu-siemens.com>

[IA64] Fix incorrect NVRAM saving if domain is destroyed by config error

Nvram saving is always executed even if a domain is destroyed by a
configuration parameter error.  In this case, Nvram saving function
will get a bad address for the NVRAM data and save garbage into the
NVRAM file. Configuring a wrong vif parameter can expose this issue.

This patch fixes the issue by adding an address check function in
NVRAM saving path.

Signed-off-by: Zhang Xin <xing.z.zhang@intel.com>

[IA64] Virtualize ibr/dbr for PV domains.

Signed-off-by: Tristan Gingold <tgingold@free.fr>

[IA64] Remove IA64_GUEST_CONTEXT_RBS_OFFSET

Pass the offset in the vcpu_reg struct.
Fix displaying stack register in xenctx.

Signed-off-by: Tristan Gingold <tgingold@free.fr>

Fix 15435:ab95b9764b20.

The changeset did not consider about ":cdrom" and "ioemu:"
of a disk definition.

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

Use 32bit operand and address during VMXAssist protected to real.
Signed-off-by: Xin Li <xin.b.li@intel.com>

xm: Fix xm block-detach error message duplication.
Signed-off-by: Masaki Kanno <kanno.masaki@jp.fujitsu.com>

xm: Fix a typo for xm network-detach.
Signed-off-by: Masaki Kanno <kanno.masaki@jp.fujitsu.com>