[IA64] Backup/restore ACPI tables

We modify some of the ACPI tables for dom0 (limiting available CPUs,
modifying id/eid, and hiding SLIT/SRAT tables).  This causes problems
when we try to kexec with different dom0 CPU counts or from Xen to
Linux.  This introduces a mechanism to save ACPI tables before
modification and restoring them before kexec.

Signed-off-by: Alex Williamson <alex.williamson@hp.com>
Acked-by: Simon Horman <horms@verge.net.au>

[IA64] Cleanup: remove unused definitions.

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

[IA64] Declare field x of M28 (fc vs fc.i).

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

[IA64] Don't warn for EOI-ing an edge triggered interrupt.

Edge triggered interrupts don't need an EOI but EOI-ing them is harmless
according to the IOSAPIC documentation.  Do not emit a warning in such case.

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

[IA64] Cleanup of hypercall leave path.

p9 is set, but not used.

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

[IA64] Support console_timestamps on IA64

This patch intends to support console_timestamps on IA64.
At this moment, EFIRTC is used for start time.

If we support synchronize HV time to Dom0 system time for long period,
Dom0 sys_settimeofday and sys_adjtimex functions should implement
Dom0-HV time sync functionality.

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

[IA64] Correctly decode imm fields in mmio.c

Immediate field of ia64 instructions are 2's complement coded.

Instructions whose imm field was negative were not
correctly emulated in mmio.c

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

[IA64] Fix MCA error handler problems

Fixing MCA issues related to changes from kexec patch series...

[From "Kexec: Fix ia64_do_tlb_purge so that it works with XEN"]

> 2. Use the per_cpu variable to derive CURRENT_STACK_OFFSET rather
>    than reading it from a kernel register. See 1) for explanation
>    of why.

I added the same code in Reload DTR for stack part and also added a
code to avoid overlapping with kernel TR.

> 3. In the VHPT pruning code, don't use r25 as ia64_jump_to_sal,
>    which branches to ia64_do_tlb_purge expects r25 to be preserved.
>    There seems no reason not to use r2 as per the other purges
>    done in ia64_do_tlb_purge.  Furthermore use r16 and r18 instead
>    of r20 and r24 for consistency reasons.

The r25 kept the value of __va_ul(vcpu_vhpt_maddr(v)), and it was
referred to by the following lines.

468     // r25 = __va_ul(vcpu_vhpt_maddr(v));
469     dep r20=0,r25,0,IA64_GRANULE_SHIFT
470     movl r26=PAGE_KERNEL
471     ;;
472     mov r21=IA64_TR_VHPT
473     dep r22=0,r20,60,4              // physical address of

I defined GET_VA_VCPU_VHPT_MADDR() macro to re-calculate the value of
__va_ul(vcpu_vhpt_maddr(v)) in each part.
And I renamed the register names for same reasons.

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

[IA64] Fix TLB insertion for subpaging

Without this patch, Longhorn is sure to hang up.  .NET application
might hit this bug.  itc.i instruction is repeated forever, because
TLB entry with smaller page size is volatile.

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

[IA64] vti domain save/restore: make vmx_vcpu_set_rr() accept non-current

make vmx_vcpu_set_rr accept non-current vcpu for setting vcpu context.

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

[IA64] vti domain save/restore: add unwind directive to break fault handler

add unwind directive to fast_hypercall path.
While fast_hypercall path calls function (hypercall, do_softirq()) and
might be blocked, it doesn't have unwind infomation.
So stack unwinding fails. Add necessary unwind directive.

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

[IA64] vti domain save/restore: fix stack unwinder

fix stack unwinder.
- fix find_save_locs() and unw_unwind().
  instruction pointer check should be suite for xen.
- fix unw_unwind_to_user()
  VTi domain fault handler doesn't always updatevcpu->on_stack so that
  the pUStk check fails. Add more checking to stop winding.

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

[IA64] vti domain save/restore: add show_stack() prototype

add show_stack() prototype which will be used later.

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

[IA64] hvm domain builder clean up. use xc_vcpu_setcontext().

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

[IA64] vmx_setup_platform() may fail. make it return error value.

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

[IA64] Fix vmx_emul_mov_from_dbr/ibr()

dbr[] and ibr[] are confused.

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

[IA64] Clean up: vmx_vcpu_set_rr()

vmx_vcpu_set_rr(): remove the obsolete variable, oldrr.

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

[IA64] Remove dead code

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

[IA64] Avoid set cr.ivt when context switch if possible

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

[IA64] Minor clean up of sync_vcpu_execstate()

vmx_save_state() is called by context_switch()

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

[IA64] Remove duplicate xc_get/set_hvm_param() definitions

Remove duplicated xc_get/set_hvm_param definition in xc_ia64_hvm_build.c
They are also defined in xc_domain.c.

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

svm: last branch recording MSR emulation
Signed-off-by: Jan Beulich <jbeulich@novell.com>
Signed-off-by: Keir Fraser <keir@xensource.com>

x86: Fix time-scale calculation to avoid #DE fault.
Signed-off-by: Keir Fraser <keir@xensource.com>

x86: Fix emulation of DIV instruction. Never actually used currently,
but this will avoid a #DE in Xen.
Signed-off-by: Keir Fraser <keir@xensource.com>

hvm: Fix CR0.TS handling.
Signed-off-by: Keir Fraser <keir@xensource.com>

Clean up SMP macros and always have IRQs disabled when executing an
'smp_call_function' callback function.
Signed-off-by: Keir Fraser <keir@xensource.com>

x86: disable AMD's C1E mode.

In C1E the APIC timer stops ticking, which Xen cannot tolerate.

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

x86: AMD Fam10/11 adjustments
Signed-off-by: Jan Beulich <jbeulich@novell.com>
Sync up affected files with 2.6.23-rc9.
Signed-off-by: Keir Fraser <keir@xensource.com>

Add local header files to HDRS.

Otherwise, changes to such headers do not trigger
re-compiles. (Powerpc already had this.)

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

More static shadow functions.
Signed-off-by: Jan Beulich <jbeulich@novell.com>

Shadow code printing adjustments.

Messages prior to BUG() should always be displayed, not just when
DEBUG_TRACE_DUMP is defined (which by default it isn't). Convert them
from SHADOW_PRINTK() to SHADOW_ERROR().

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

Do not include compile.h directly -- use version.h indirection functions.
Signed-off-by: Keir Fraser <keir@xensource.com>

[vTPM] Add set/get_other_config to Xen-API.

Add get_- and set_other_config methods to the vTPM class. Write the
parameters into the xenstore while the domain is running prefixing it
with 'oc_'. Also I add the methods to the C library as well as the
documentation.

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

hvm: Fix migration of 64-bit guests by permitting EFER_LMA to be set
in restored VCPU context.
Signed-off-by: David Lively <dlively@virtualiron.com>
Signed-off-by: Robert Phillips <rphillips@virtualiron.com>

ia64: Fix help message of xenitp.
Signed-off-by: Akio Takebe <takebe_akio@jp.fujitsu.com>

Fix printing month in printk prefix (1-12 rather than 0-11).
Signed-off-by: Atsushi SAKAI <sakaia@jp.fujitsu.com>

vt-d: Clean up iommu code.
Signed-off-by: Weidong Han <weidong.han@intel.com>

MiniOS: no-stack-protector added to the cflags if needed,
and the default domain config file fixed.

From: Grzegorz Milos <gm281@cam.ac.uk>
Signed-off-by: Keir Fraser <keir@xensource.com>

[IA64] Update for cset 16067:9f9f9b68cd08
Signed-off-by: Alex Williamson <alex.williamson@hp.com>

xend, xsm: Lock domain access while modifying policy.
Signed-off-by: Stefan Berger <stefanb@us.ibm.com>

xend: Remove unused xoptions.netback_type.
Signed-off-by: Keir Fraser <keir@xensource.com>

xenconsoled: portability fixes:
 - Use openpty(), which does the same as the sequence
   of open(), grantpt(), unlockpt(), ptsname(), tcgetattr()
   simplifies code
 - Check return code from tcsetattr()
 - sprintf() -> snprintf()
 - OpenBSD lacks POSIX grantpt() and unlockpt()
   requires use of openpty()
 - Solaris lacks POSIX openpty() via feedback from SUN (John Levon)
   implement openpty() for Solaris, tested and ok'd by SUN (John
   Levon)

Signed-off-by: Christoph Egger <Christoph.Egger@amd.com>

Remove tabs from python scripts xenmon and xentrace_format.
Signed-off-by: Masaki Kanno <kanno.masaki@jp.fujitsu.com>

Fix domU reboot failure when using "pci=" option

The failure only happens on the second or later auto reboot.
See the following thread for a long discussion on this issue where
most problems were fixed.

http://lists.xensource.com/archives/html/xen-users/2006-05/msg00154.html

Signed-off-by: Charles Arnold <carnold@novell.com>

[XEN] Shadow: fix locking in error path.
Signed-off-by: TIm Deegan <Tim.Deegan@xensource.com>

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

xenstore: sprintf->snprintf, and fix for OpenBSD.
Signed-off-by: Christoph Egger <Christoph.Egger@amd.com>

entrace: Remove assertions to avoid system crashes on debug build xen.
Signed-off-by: Yosuke Iwamatsu <y-iwamatsu@ab.jp.nec.com>

[NW scripts] Choose alternative interface if NFSroot is used.

Allow the network bridging setup script to choose an alternative
network interface if it is determined that the root is mounted using
NFS, which is assumed to be done over the NIC serving the default
gw. In this case the prefix of the device is used to find other
devices, i.e., if eth0 is the default device an alternative device may
be an other ones starting with 'eth'.

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

xen: Changes to printk handling:
 1. Command-line option 'console_timestamps' adds a timestamp prefix
 to each line of Xen console output (x86 only, after CMOS has been
 interrogated).
 2. Clean up prefix handling and vanity banner info.
Signed-off-by: Keir Fraser <keir@xensource.com>

Revert 16065:671dcd08e806 and try again to fix the issue.

Changing LINUX_VER causes us to clone the wrong repository so instead
use the kernel's "kernelrelease" target to figure out the correct
Linux version.

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

xsview: a xenstore viewer written in PyQT.
From: Mark Williamson <mark.williamson@cl.cam.ac.uk>
Signed-off-by: Keir Fraser <keir@xensource.com>

Clean up grant-table code after recent changes.
Signed-off-by: Keir Fraser <keir@xensource.com>

Allow iomem permissions to be set up through grant table ops.
Signed-off-by: Kieran Mansley <kmansley@solarflare.com>

Fix TLB flush on grant unmap.
Signed-off-by: Kieran Mansley <kmansley@solarflare.com>

Fix kernel version after linux-2.6.18-xen.hg 219:3e8752eb69dc which
applied 2.6.18.8,

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

hvm: Fix the flush-all-tlbs hypercall.
From: Peter Johnston <pjohnston@xensource.com>
Signed-off-by: Keir Fraser <keir@xensource.com>

This patch ensures xen vmm has exclusive access of vt-d hw by zapping
vt-d "DMAR" ACPI table signature after xen has finished using it.

Signed-off-by: Allen Kay <allen.m.kay@intel.com>

README: Remove incorrect information about PAE builds.
Signed-off-by: Keir Fraser <keir@xensource.com>

xenbaked: Fix for new trace format.
Signed-off-by: Atsushi SAKAI <sakaia@jp.fujitsu.com>

[hotplug scripts] Unlock the 'block' lock before bailing.
Signed-off-by: Stefan Berger <stefanb@us.ibm.com>

svm: Enable nested paging only on x86/64.
Signed-off-by: Keir Fraser <keir@xensource.com>

hvm: More changes to reduce size of domain structure.
It is now smaller than a page (4kB) on x86/32 and x86/64.
Signed-off-by: Keir Fraser <keir@xensource.com>

hvm: Do not include full hvm_hw_hpet in 'struct domain'. It is very
large due to the number of (unused) reserved registers. This
unnecessarily blots the size of the domain structure.
Signed-off-by: Keir Fraser <keir@xensource.com>

xm: Add accel option to xm network-attach.
Signed-off-by: Kieran Mansley <kmansley@solarflare.com>

svm: Actually remove disabling of CR4-read-intercept when using NPT.
Should have been in the previosu changeset. Oops.
Signed-off-by: Keir Fraser <keir@xensource.com>

svm: Can only avoid CR3 reads when NPT is enabled.
CR0 and CR4 reads must still be intercepted.
Signed-off-by: Keir Fraser <keir@xensource.com>

Merge with ia64 tree.

xend: Check access to the privcmd interface before doing the call to
fetch the currently enforced policy. Assign 'INACCESSIBLE' to the
policy if it cannot be retrieved due to the user not being
privileged.

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

Extend 'xm dumppolicy' to support Xen-API

I am extending 'xm dumppolicy' to be used via the Xen-API. For this
there are two new functions in the ACM policy class:
 - get the currently enforced policy including statistical data from
   the hypervisor
- get the ACM 'ssidref' of a Domain. Since this may be a ACM-specific
  variable or type (int) I put it into the ACM class.

I extended the Xen-API documentation with the two new functions.

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

merge with xen-unstable.hg (staging)

[IA64] micro-optimize and comments in vmx_ivt.S

Comments added + micro-optimizations.
In PHY_D mode, alt tlb misses can call vmx_hpw_miss.

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

x86/64: Fix build.
Signed-off-by: Keir Fraser <keir@xensource.com>

[IA64] cleanup in vmx headers

Reduce size of arch_vmx_struct (flags is now 1 byte).
Remove unused macro.
Add VMX_MMU_MODE macro.

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

[IA64] Cleanup: remove useless #ifndef and indentation.

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

vtd: Dynamically allocate IRQ-tracking structures, only for those
domains that actually have PCI-passthru devices. Greatly reduces size
of 'struct domain'.
Signed-off-by: Keir Fraser <keir@xensource.com>

[IA64] Remove warnings.

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

[IA64] xenitp improvements

Remove all the casts by using char * instead of unsigned char *.
The go command now accept a number.
The disass command can now accept a range.
Number of TRs is not hard-coded.
'-' (minus) can now be used in expressions.
'$iip' and '$b0' can be used in expressions.

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

x86/32: Re-factor mapcache data structure into per-domain and per-vcpu
components. Greatly reduces size of 'struct domain'.
Signed-off-by: Keir Fraser <keir@xensource.com>

[IA64] xenitp: Display symbolic names of cr.

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

[IA64] Fix + more cr regs for vcpu_get_context

Do not hard code number of TRs in vcpucontext.
Correctly fill rr7 and add more cr registers for get_vcpu_context.

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

xentop: Flush batch output on SIGTERM or SIGINT.
Signed-off-by: INAKOSHI Hiroya <inakoshi.hiroya@jp.fujitsu.com>

hvm svm: Fix #MC intercept routine for AMD SVM.
Signed-off-by: Christoph Egger <Christoph.Egger@amd.com>

Add XENPF_getidletime.
Signed-off-by: Rik van Riel <riel@redhat.com>
Signed-off-by: Keir Fraser <keir@xensource.com>

Clean up and fix XEN_SYSCTL_getcpuinfo error path.
Signed-off-by: Keir Fraser <keir@xensource.com>

xend: Fix name/uuid uniqueness checks.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>

Revert 15642:207582c8d88b532783da5c6f5839336187556f0a.

This patch is now bogus since 15168:a717cb2fac90 is reverted
by 15973:8817a53c030f.

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

Fix external-device-migrate infinite loop.

For some (yet) unknown reason extern-device-migrate is called
twice, once with a missing domname argument.

The following invocation loops forever:
 external-device-migrate -domname
because "shift 2" never shifts ($# < 2) and $# is always 1.

From: Pascal Bouchareine <pascal@gandi.net>
Signed-off-by: Keir Fraser <keir@xensource.com>

xenctx: Print EFLAGS for i386 and x86/64.
From: Peter Johnston <pjohnston@xensource.com>
Signed-off-by: Keir Fraser <keir@xensource.com>

hvm: Fix HVMOP_flush_tlbs to flush paging-mode state.
From: Peter Johnston <pjohnston@xensource.com>
Signed-off-by: Keir Fraser <keir@xensource.com>

xentrace: Fix xentrace_format for new file format.
Signed-off-by: Atsushi SAKAI <sakaia@jp.fujitsu.com>

xentrace: Fix man page (remove xentrace_cpuinfo(1)).
Signed-off-by: Atsushi SAKAI <sakaia@jp.fujitsu.com>

Fix ioemu compile on Solaris
Solaris doesn't have RLIMIT_RSS at least; don't try to set rlimits.
Signed-off-by: John Levon <john.levon@sun.com>

Fix non-portabilities in libflask.
Signed-off-by: John Levon <john.levon@sun.com>

xend: Fix refactoring done in CS 15938 in security.py.
Signed-off-by: Stefan Berger <stefanb@us.ibm.com>

xm-test: Fix bugzilla #1051.

This is a patch to fix xm-test's network errors
http://bugzilla.xensource.com/bugzilla/show_bug.cgi?id=1051

From: Lukasz Szymanski <Lukasz.Szymanski@unisys.com>
Signed-off-by: Keir Fraser <keir@xensource.com>

merge with xen-unstable.hg

[IA64] xenitp: correctly size vcpu_ctx array

vcpu_ctx[] array size should be MAX_VIRT_CPUS instead of 1.

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

[IA64] Fix wrong insertion of TLB entry in region 0

On PV domain with metaphysical mode, emulation of itc.d in region 0
doesn't work well and inserts an wrong TC entry.
Because set_one_rr() doesn't set the machine region register.
i.e. metaphyisical_rr0 is used instead of guest's rr[0].

This bug causes Dom0/U crash when an application uses region 0.
Actually I met the crash when I was building open GFW (java uses
region 0).

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

[IA64] xenctx: fix rse arithmetic

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