[HVM][VNC] Work around a bug in the newest version of xvncviewer which
could lead to the display locking up after it's been idle for a while.

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

[HVM][VNC] Make sure that qemu doesn't go into an infinite loop when
it receives certain invalid requests from the viewer.

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

[HVM] Make qemu_advance_timer do something sensible with timers which
aren't currently scheduled.

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

[HVM] Fix HLT with interrupts disabled. Only take down
the domain if no other VCPUs are online.
Signed-off-by: Keir Fraser <keir@xensource.com>

Adds support for the keyword 'all' to the vcpu-pin operation.
Using 'all' in place of a specific vcpu will apply the cpumask
to all vcpus in the domain.
Signed-off-by: Ryan Harper <ryanh@us.ibm.com>

[VMXASSIST] Small fix to virt-to-phys translation.
CR4.PSE is ignored in PAE mode, so remove test of it.
Signed-off-by: Keir Fraser <keir@xensource.com>

Remove static forward declaration in gdbstub code.
Signed-off-by: Keir Fraser <keir@xensource.com>

[VMXASSIST] Fix virt-to-phys translation routine.
Signed-off-by: Keir Fraser <keir@xensource.com>

[XEN] Simplify VMCS allocation (never bigger than a page).
Dynamically determine reserved bits in execution-control
fields.
Signed-off-by: Keir Fraser <keir@xensource.com>

[XEN] Small clean up.
Signed-off-by: Keir Fraser <keir@xensource.com>

[HVM] Rate limit guest accesses to the qemu virtual serial port.  This stops
grub's boot menu from hammering dom0.

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

Make dom0_vcpus_pin a platform independant option.
By default, it's off. When enabled on the command line, it
causes dom0's VCPUs to be pinned to their respective physical
CPUs at creation time. Once pinned, they can never be moved.
ALso cleanup VCPU initialization of scheduler parameters.
Signed-off-by: Emmanuel Ackaouy <ack@xensource.com>

[XEN] Fix 11607 to avoid bignum aritmetic in assembly file.
Signed-off-by: Keir Fraser <keir@xensource.com>

[HVM] Use AMD's isntruction-length decoder for VMX as well as SVM
MMIO decode. The VMX-defined instruction-length info field is not
valid for use during most page faults. Hence we have to obtain the
instruction length the slow-and-stupid way.

This *will* go away when we throw away the wretched MMIO emulator.

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

Consolidate the preferable use of a pragma with the feature detection
in the make script.
Signed-off-by: Jan Beulich <jbeulich@novell.com>

[XEN] Limit size of Xen heap to 2GB.

Avoids stub code created on the stack being unable to
reach targets in main Xen code via direct branches.

Signed-off-by: Jan Beulich <jbeulich@novell.com>

[XEN] Use symbolic constants to determine segment limits.
Signed-off-by: Jan Beulich <jbeulich@novell.com>

Merge.

[XEN][HVM] Initialise the ELCR to level triggered mode from pic_init1.
This should really be done from the BIOS, but since it isn't, work around
it from here.

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

[TOOLS] Wait for backend devices to set up before resuming execution
of a restored guest.
Signed-off-by: Keir Fraser <keir@xensource.com>

[XEN] Support unconnected IRQ lines in pirq_acktype().
Signed-off-by: Keir Fraser <keir@xensource.com>

[XEN] Add local-APIC-edge to list of PIC types supported by pirq_acktype()
function.
Signed-off-by: Keir Fraser <keir@xensource.com>

[QEMU] A better build fix for qemu-dm.
This extends 11592:5c58df8c78851d47bf7130308895c4e0ac5fa5ad
Signed-off-by: Keir Fraser <keir@xensource.com>

Ensure CONFIG_SND_FM801_TEA575X is unconfigured on all arches.
Signed-off-by: Al Stone <ahs3@fc.hp.com>

Fix tools build with latest glibc, which has an
extended PTRACE enumeration.
Original patch from Stephen Tweedie <sct@redhat.com>
Signed-off-by: Keir Fraser <keir@xensource.com>

Use GDB 'O' packets for console output if the serial line is shared
and GDB is attached.

It may be necessary for gdb and the console to share a serial port.
This patch utilises the GDB protocol to encode console output.

Based on a patch from Tony Breeds <tony@bakeyournoodle.com>
Signed-off-by: Keir Fraser <keir@xensource.com>

Move initialize_keytable declaration to keyhandler.h

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

Clean up and enhance "make tags"

- delete correct file in _tags rule
- don't prune nonexistent files/dirs
- call exuberant-ctags with additional flags from Linux:
    -I to ignore some misleading hits
    --extra=+f to include an entry for the basename of every file
    --c-kinds=+px to add prototypes and external variable declarations

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

[HVM] Fix timer interrupt underflow bug.

Without the patch, vlapic->intr_pending_count could become a very
small negative, causing some local APIC timer interrupts lost.

Signed-off-by: Dexuan Cui <dexuan.cui@intel.com>
Signed-off-by: Edwin Zhai <edwin.zhai@intel.com>

Update vnet kernel module to handle architectures without
socketcall(), for example Itanium, and some small changes
to be compatible with changed kernel interfaces.

Signed-off-by: Mike Wray <mike.wray@hp.com>

[MINIOS] Added a new file arc/x86/setup.c and moved some x86 specific
initialization stuff from kernel.c there. Two new functions are added
to handle this.

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

[QEMU] Fix build on some distros (e.g., FC6T3).
Signed-off-by: Keir Fraser <keir@xensource.com>

[XM] Fix argument parsing for dump-core subcommand

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

[HVM] A few cleanups to i8259.c -- in particular change
ELCR semantics (should not be modified by PIC reset).

This effectively removes Steve Smith's changeset
11586:08a582a98471da30fb86bca620495f3d91d55800
but the real bug here is not that the ELCR was reset to
zero, but that it was touched at all.

Our BIOS code should set the ELCR correctly (level-triggered)
for PCI IRQs.

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

[HVM][VNC] Make sure that the vnc viewer gets the right resolution when it
connects.

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

Merge.

[XM][ACM] Add try,catch back to main() for dry-run

Signed-off-by: Alastair Tse <atse@xensource.com>
Signed-off-by: Bryan D. Payne <bryan@thepaynes.cc>

Merge.

[XEN][HVM] Initialise the ELCR so that all interrupts are in level triggered
mode after pic_reset, since that seems to be what Linux expects.

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

[XM] Move 'xm sysrq' into main.py

* There's not need for sysrq to be in its own file.
* Bonus advantage: can use arg_check rather than gopts for argument
  validation.

Signed-off-by: Alastair Tse <atse@xensource.com>

[XEND] Do not attempt core dump to directories.

- Abort core dump if destination is a directory.
- Throw appropriate XendError expection on failure.
Reported by Masaki Kanno <kanno.masaki@jp.fujitsu.com>.

Signed-off-by: Alastair Tse <atse@xensource.com>

[XEND] Remove hard tabs from destroyDevice()
Reported by Hollis Blanchard.

Signed-off-by: Alastair Tse <atse@xensource.com>

[XEND] Check for invalid memory sizes.

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

[XM] Text wrapping fix, xm create --help_config fix.

* Fix text wrap so it doesn't chop off last word in help message for
  certain cases.
* Fix handling of xm create --help_config
* Remove redundant gopts.usage() call.

Signed-off-by: Alastair Tse <atse@xensource.com>

Add mmio emulation of opcodes 0x3A and 0x85 for SMP W2K3 installation.

Signed-off-by: Qing He <qing.he@intel.com>
Signed-off-by: Steven Hand <steven@xensource.com>

Revert default debug on cset 11576:4a3d9fa6ba24.

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

merge

merge

[XEN][ Remove unused declarations.
Signed-off-by: Keir Fraser <keir@xensource.com>

This patch fixes an error booting 32bit vista on VMX.

The shadow code uses hvm_get_guest_ctrl_reg(v, 4) to test whether
PAE is enabled or not.  But it is not always right if the hypervisor
calls hvm_get_guest_ctrl_reg(v, 4) between vmxassist_invoke and
vmxassist_restore

The patch uses the d->arch.hvm_vmx.cpu_state to test if the PAE is
enabled.

Also update SVM code to use the new 'pae_enabled' hvm func.

Signed-off-by: Xiaohui Xin <xiaohui.xin@intel.com>
Signed-off-by: Steven Hand <steven@xensource.com>

Fix vmxassist to boot Vista.

Currently, guest_linear_to_real to_real() in vmxassist assumes guest HVM
use 2 level page table when entering protect mode with PG enabled. Vista
uses 3 level (PAE enabled) page table, so memory addressing is wrong.
This patch fixes it, by considering all 4 possible cases.

Signed-off-by: Xiaowei Yang <xiaowei.yang@intel.com>

[XM] Error handling cleanup

Introducing an OptionError exception to be used by all xm subcommands
to signal an error in the arguments. "xm" will catch these and output
the appropriate error and usage message.

Detailed Changes:

main.py:
 *  Cleaned up imports and moved warning filter outside of
    import block.
 *  Converted usage parameters and description to a python
    dict rather than strings to enable better usage help
    formatting.
 *  Removed unused list_label domain command.
 *  Added cmdHelp() prints out usage message for any command
 *  Added shortHelp() prints out the default help message when
    xm is invoked with no arguments.
 *  Added longHelp() prints out long help message when invoked
    with xm --help or xm help.
 *  Added extra optional paramter to getDomains() so we can
    tell Xend not to poll devices. This will speed up xm list.
    (PENDING changes to Xend itself.)
 *  Changed all references where 'dom' actually means 'domid'
    to use the correct name.
 *  Changed 'xm list' header format to use printf formatting style.
 *  Renamed xm_subcommand to xm_importcommand so it is more
    clear what it is doing (all xm commands are subcommands!)
 *  Moved cpu_make_map() inside xm_vcpu_pin as an anonymous func.
 *  Use OptionError whenever an invalid option is detected in xm.
 *  Added proper catch and error printing for XendError and
    OptionErrors in main().

addlabel.py:
cfgbootpolicy.py:
dry-run.py:
dump-policy.py:
get-label.py:
labels.py:
loadpolicy.py:
makepolicy.py:
rmlabel.py:
resources.py:
 *  Replaced usage() with help() that is called from main.py
 *  Replaced usage() invokation with raising OptionError

opts.py:
 *  Added very simple wrap() function that behaves differently
    to textwrap.wrap()
 *  Added OptionError()
 *  Replaced the string representation of Opt, Opts to output
    a nicely formatted usage message.
 *  Changed class Opts itself will throw approriate OptionErrors.
 *  Changed set_bool to recognise 'y' and 'n' as valid input

create.py:
 *  Some whitespace and column width cleanup.
 *  throws OptionError if encounters option error.

migrate.py:
 *  Replace usage() message with the string representation of
    gopts.

sysrq.py:
 * Replace usage message with throwing OptionErrors

Signed-off-by: Alastair Tse <atse@xensource.com>

[HVM][VMX] Clean up and audit VMX uses of instruction-length
info field. Todo: use by mmio decoder needs to be removed.
Signed-off-by: Keir Fraser <keir@xensource.com>

[HVM][VNC] Fix bug in the VNC copyrect implementation which crept in
with the update backoff patch.

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

[XEN] Do not emulate user-mode updates to page tables.
Instead use as a hint that the page is no longer a page table.
Signed-off-by: Keir Fraser <keir@xensource.com>

Merge?

[XEN] Fix domctl for changing VCPU affinity.

Now works for any VCPU, including the caller's VCPU.

By not synchronously pausing the affected VCPU we avoid
any risk of deadlock.

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

[qemu patches] Refresh.

Signed-off-by: Christian Limpach <Christian.Limpach@xensource.com>

[qemu patches] Update patches for changeset 11562:449dcaff2551.

Signed-off-by: Christian Limpach <Christian.Limpach@xensource.com>

[qemu patches] Update patches for changeset 11561:c742b2ae920c.

Signed-off-by: Christian Limpach <Christian.Limpach@xensource.com>

[qemu patches] Update patches for changeset 11554:39aaa54756db.

Signed-off-by: Christian Limpach <Christian.Limpach@xensource.com>

[HVM][VNC] Add a backoff feature to the vnc server, so that if it detects
that the display hasn't changed for a while it starts scanning more slowly.

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

[HVM] Reduce VNC overhead, by (a) only scanning framebuffer when a client
is connected, and (b) fixing an overflow bug in the scanning code which
prevented the dirty bit from ever getting cleared.

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

Merge

[XEND] Don't prepend '/dev/' to any absolute path.
Check for '/' rather than '/dev/' when translating e.g. 'hda'->'/dev/hda'
Signed-off-by: Tim Deegan <tim.deegan@xensource.com>

[XEND] split command-line arguments to the bootloader script properly.
Signed-off-by: Tim Deegan <tim.deegan@xensource.com>

Pygrub: add /var/lib/xen to the install tree.
(Pygrub uses /var/lib/xen to store fifos and image files.)
Signed-off-by: Tim Deegan <tim.deegan@xensource.com>

Pygrub: stop the FS identification code from leaking open fds.
Signed-off-by: Tim Deegan <tim.deegn@xensource.com>

This patch adds value checking of sched-credit xm options.
Signed-off-by: Masaki Kanno <kanno.masaki@jp.fujitsu.com>

[tools] Increases some of the timeouts
since e.g. stealing the hotplug lock after five seconds is completely insane.

From: Steven Smith <sos22@cam.ac.uk>
Signed-off-by: Christian Limpach <Christian.Limpach@xensource.com>

[qemu patches] Refresh qemu-pci patch without -p3.

Signed-off-by: Christian Limpach <Christian.Limpach@xensource.com>

[qemu patches] Update patches for changeset 11551:f872300b672f.

Signed-off-by: Christian Limpach <Christian.Limpach@xensource.com>

[HVM] Use the right, PCI SIG assigned, vendor ID for xensource rather than
just making one up.

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

[XEN] Shadow mode no longer obtains page type references.
This allows the shadow destructor hook in free_page_type()
to work properly.

Also, move mark_dirty() back to alloc/free_page_type(). It
doesn't matter that this happens before the type count is
modified -- bitmap is extracted by the tools with the domain
paused, so these non-blocking paths are atomic from p.o.v of
the tools.

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

[LINUX] Update pci-mmconfig patch to refer to machine_e820 map on Xen.
Signed-off-by: Keir Fraser <keir@xensource.com>

[LINUX] Provide permanent global version of machine_e820 map for
general use.
Signed-off-by: Keir Fraser <keir@xensource.com>

[XEN] Move mark_dirty() from alloc_page_type() to get_page_type(),
for symmetry with put_page_type(). Limit calls to mark_dirty() in
put_page_type() so that they happen only for page-table pages.
Signed-off-by: Keir Fraser <keir@xensource.com>

merge

fix -xen x86_32 linux build
Signed-off-by: Emmanuel Ackaouy <ack@xensource.com>

[XEN] Trivial simplification.
Signed-off-by: Keir Fraser <keir@xensource.com>

[XEN] Clean up mark_dirty() usage in mm.c some more.
Add clarifying comments. Move mark_dirty() in alloc_page_type()
to end of the function (more correct).
Signed-off-by: Keir Fraser <keir@xensource.com>

Simplify + fix previous cset (remove_all_shadows() is indeed required :-)

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

Ensure page-table pin changes are noted in shadow log-dirty mode.
Also mark dirty in put_page_type() instead of free_page_type()

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

[LINUX] Backport PCI MMCONFIG patch from 2.6.17 -- only use MMCONFIG
access method if the memory area is reserved in the E820 map.
Signed-off-by: Keir Fraser <keir@xensource.com>

[XEN] Fix arg clobber test for debug version of hypercall code.
Signed-off-by: Jan Beulich <jbeulich@novell.com>
Signed-off-by: Keir Fraser <keir@xensource.com>

[XEN][X86_64] Remove hardcoded addresses in x86/64 boot assembly.
Signed-off-by: Jan Beulich <jbeulich@novell.com>

[XEN] Upgrade some platform files to Linux 2.6.17.13.
Signed-off-by: Keir Fraser <keir@xensource.com>

[linux patches] Update patches for linux-2.6.16.29.

Signed-off-by: Christian Limpach <Christian.Limpach@xensource.com>
--HG--
rename : patches/linux-2.6.16.13/blktap-aio-16_03_06.patch => patches/linux-2.6.16.29/blktap-aio-16_03_06.patch
rename : patches/linux-2.6.16.13/device_bind.patch => patches/linux-2.6.16.29/device_bind.patch
rename : patches/linux-2.6.16.13/fix-hz-suspend.patch => patches/linux-2.6.16.29/fix-hz-suspend.patch
rename : patches/linux-2.6.16.13/fix-ide-cd-pio-mode.patch => patches/linux-2.6.16.29/fix-ide-cd-pio-mode.patch
rename : patches/linux-2.6.16.13/i386-mach-io-check-nmi.patch => patches/linux-2.6.16.29/i386-mach-io-check-nmi.patch
rename : patches/linux-2.6.16.13/ipv6-no-autoconf.patch => patches/linux-2.6.16.29/ipv6-no-autoconf.patch
rename : patches/linux-2.6.16.13/net-csum.patch => patches/linux-2.6.16.29/net-csum.patch
rename : patches/linux-2.6.16.13/net-gso-0-base.patch => patches/linux-2.6.16.29/net-gso-0-base.patch
rename : patches/linux-2.6.16.13/net-gso-1-check-dodgy.patch => patches/linux-2.6.16.29/net-gso-1-check-dodgy.patch
rename : patches/linux-2.6.16.13/net-gso-2-checksum-fix.patch => patches/linux-2.6.16.29/net-gso-2-checksum-fix.patch
rename : patches/linux-2.6.16.13/net-gso-3-fix-errorcheck.patch => patches/linux-2.6.16.29/net-gso-3-fix-errorcheck.patch
rename : patches/linux-2.6.16.13/net-gso-4-kill-warnon.patch => patches/linux-2.6.16.29/net-gso-4-kill-warnon.patch
rename : patches/linux-2.6.16.13/pmd-shared.patch => patches/linux-2.6.16.29/pmd-shared.patch
rename : patches/linux-2.6.16.13/rcu_needs_cpu.patch => patches/linux-2.6.16.29/rcu_needs_cpu.patch
rename : patches/linux-2.6.16.13/rename-TSS_sysenter_esp0-SYSENTER_stack_esp0.patch => patches/linux-2.6.16.29/rename-TSS_sysenter_esp0-SYSENTER_stack_esp0.patch
rename : patches/linux-2.6.16.13/series => patches/linux-2.6.16.29/series
rename : patches/linux-2.6.16.13/smp-alts.patch => patches/linux-2.6.16.29/smp-alts.patch
rename : patches/linux-2.6.16.13/tpm_plugin_2.6.17.patch => patches/linux-2.6.16.29/tpm_plugin_2.6.17.patch
rename : patches/linux-2.6.16.13/x86-elfnote-as-preprocessor-macro.patch => patches/linux-2.6.16.29/x86-elfnote-as-preprocessor-macro.patch
rename : patches/linux-2.6.16.13/x86-increase-interrupt-vector-range.patch => patches/linux-2.6.16.29/x86-increase-interrupt-vector-range.patch
rename : patches/linux-2.6.16.13/x86-put-note-sections-into-a-pt_note-segment-in-vmlinux.patch => patches/linux-2.6.16.29/x86-put-note-sections-into-a-pt_note-segment-in-vmlinux.patch
rename : patches/linux-2.6.16.13/x86_64-put-note-sections-into-a-pt_note-segment-in-vmlinux.patch => patches/linux-2.6.16.29/x86_64-put-note-sections-into-a-pt_note-segment-in-vmlinux.patch
rename : patches/linux-2.6.16.13/xen-hotplug.patch => patches/linux-2.6.16.29/xen-hotplug.patch
rename : patches/linux-2.6.16.13/xenoprof-generic.patch => patches/linux-2.6.16.29/xenoprof-generic.patch

[linux] Update to linux-2.6.16.29.

Signed-off-by: Christian Limpach <Christian.Limpach@xensource.com>

[XEN] Fix the emulation of instructions in vm86 mode. It fetches
them using cs and eip instead of only eip. This makes it at least
possible to use the i945GM vesa bios from the running system.

Signed-off-by: Bastian Blank <waldi@debian.org>

[XENMON] Add argument check of "--ms_per_sample" option.

If a value of "--ms_per_sample" option is larger than a value of
"--time" option, xenmon may generate no log file. So, too large
"--ms_per_sample" should be treated as an error.

e.g.
  % xenmon.py --ms_per_sample=2000 -t 1 -n
  usage: xenmon.py [options]

  xenmon.py: error: option --ms_per_sample: too large (> 1000 ms)

# Notice that a unit of "--time" is a second.

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

[HVM][SVM] Add cr8 intercept to/from code for SVM.
Add proper vlapic unmapping and freeing in the svm relinquish
resources.

Signed-off-by: Tom Woller <thomas.woller@amd.com>=20

[HVM][SVM] This patch cleans up the SVM CPUID logic a bit, adds proper clearing
of MCA/MCE bits, and also ties clearing the PAE bit with the NX bit. We
have seen issues with Solaris if the NX bit is set but the PAE bit is
not.

Signed-off-by: Tom Woller <thomas.woller@amd.com>=20

[HVM] getmemlist hypercall returns pages in strict
physmap order.
From: Ke Yu <ke.yu@intel.com>
Signed-off-by: Keir Fraser <keir@xensource.com>

[XEN][X86_64] USe GLOBAL bit to build user mappings.
Avoids need to flush user mappings when switching between
user and kernel contexts.
Signed-off-by: Jun Nakajima <jun.nakajima@intel.com>
Signed-off-by: Keir Fraser <keir@xensource.com>

Merge with xenppc-unstable-merge.hg

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

Clean up dom0_vga_console_info structure for v3.0.3.
Signed-off-by: Keir Fraser <keir@xensource.com>

Revert 11475:11645dda144c3c8365dd2a6a64cb5a7d7da01170
Signed-off-by: Keir Fraser <keir@xensource.com>

[POWERPC][XEN] Track the Hard CPUID as configured by the FW

This patch correctly implements and supports hard_smp_processor_id().

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