[IA64] fix a bug in accelerating mov to rr

Signed-off-by: Anthony Xu <anthony.xu@intel.com>

[IA64] Correctly not handle VHPT long format.

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

[IA64] Remove unused code (vmx_uaccess).

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

[IA64] merge pal emulator

Merge VTi pal emulator with Xen pal emulator.
Xen/ia64 now has only one pal and one sal emulator.

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

[IA64] backport toolchain fixes from linux-ia64

This backports the ia64 specific parts of this upstream linux
changeset:

http://www.kernel.org/hg/linux-2.6/?cs=dfbee33b0693

Change suggested by Akio Takebe.

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

[IA64] add arch/ia64/kernel/Makefile to sparse tree

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

[IA64] code cleanups

A few more code cleanups

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

[IA64] deopfuscate vcpu.c and priop.c and related headers

De-opfuscate vcpu.c, priovop.c and related header files to make them
easier to debug and read. In addition eliminate bogus UINT64 and UINT
data types which only exist in the ia64 tree and are totally
unnecessary as we have u64 already.

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

[IA64] Add not-SMP-safe comment about PV-on-HVM

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

[IA64] warning fix for HVM PV driver

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

[IA64] Fix panic when destroying DomVTI

Signed-off-by: Kouya SHIMURA <kouya@jp.fujitsu.com>
Signed-off-by: Anthony Xu <anthony.xu@intel.com>

[IA64] micro optimize __domain_flush_vtlb_track_entry.

try to use local purge instead of global purge when possible.

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

[IA64] tlbflush_clock

This patch introduces xen compile time option, xen_ia64_tlbflush_clock=y.

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

[IA64] Remove dom0_align which isn't used anymore.

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

[IA64] warning fix

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

[IA64] Remove VHPT_ADDR

Remove VHPT_ADDR by mapping vhpt to xen identity mapping area.
and some clean ups.

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

[IA64] TLB tracking

Add tlb insert tracking to flush finer grained virtual address
range when a page is unmapped from a domain.
This functionality is enabled with a compile time option,
xen_ia64_tlb_track(default is y) and xen_ia64_tlb_track_cnt(default is n).

This patch forcuses on grant table mapping.
When page is unmapped, full vTLB flush is necessary.
By tracking tlb insert on grant mapped page, full vTLB flush
can be avoided.
Especially when vbd backend does only DMA, so dom0 doesn't insert tlb entry
on the grant mapped page. In such case any vTLB flush isn't needed.

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

[IA64] linux/include/linux/hash.h.

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

[IA64] Accelerate mov to rr

Signed-off-by: Anthony Xu <anthony.xu@intel.com>

[IA64] Xencomm fixes for HVM PV-drivers

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

[IA64] Xencomm fixes for HVM PV-drivers and driver domain.

Create xcom_mini.c from xcom_hcall.c

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

merge with xen-unstable.hg

[NET] front: Do not read features from backend until it enters InitWait.
Signed-off-by: Keir Fraser <keir@xensource.com>

[NET] back: Simplify probing of new interfaces. There's no need
to watch for creation of the handle node in xenstore, as it is
written by xend when creating the backend directory as part of
the same transaction.
Signed-off-by: Keir Fraser <keir@xensource.com>

[NET] front: No need to read handle field from xenstore. It's not used.
Signed-off-by: Keir Fraser <keir@xensource.com>

[IOEMU] Fix argument to parse_host().
Signed-off-by: Keir Fraser <keir@xensource.com>

[HVM][SVM] Use proper name for the K8 VM_CR MSR.
Signed-off-by: Keir Fraser <keir@xensource.com>

[HVM][SVM] Check if SVM is disabled by the BIOS before enabling it.

Newer BIOS implementations will be able to disable the SVM feature,
although an additional test of an MSR  (VMCR 0xC0010114 bit 4) is
necessary (set equals disabled).  Bit 4 of MSR 0xc0010114 returns 0
(SVM enabled) on machines with older BIOS' without the SVM disable
feature support.

Signed-off-by: Wei Huang <wei.huang2@amd.com>=20
Signed-off-by: Tom Woller <thomas.woller@amd.com>=20

[NET] back: Copy tx_ring data before verification

We need to make a copy of data from tx_ring before
verifying them as otherwise what we end up using may
be different from what was verified.

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

[HVM] Fix resource leak in error path of AP bringup.
Based on a patch from Kouya Shimura.
Signed-off-by: Keir Fraser <keir@xensource.com>

[XEN] Fix booting with dom0_mem specifying 4GB or more.
Signed-off-by: Charles Coffing <ccoffing@novell.com>

Fix privcmd mmap() on 64b architectures for regions larger than 2GB.
Signed-off-by: Xin Li <xin.b.li@intel.com>
Signed-off-by: Keir Fraser <keir@xensource.com>

Include resources for PV xen-platform device in ACPI DSDT.
From: Qing He <qing.he@intel.com>
Signed-off-by: Keir Fraser <keir@xensource.com>

[HVMLOADER] Remove an unused variable from an SMBIOS function.
Signed-off-by: Keir Fraser <keir@xensource.com>

[IA64] Add support for enforcing singleshot mmap() semantics at the
privcmd interface.
From: Isaku Yamahata <yamahata@valinux.co.jp>
Signed-off-by: Keir Fraser <keir@xensource.com>

[PRIVCMD] Fix ia64 domain creation by abstracting the singleshot mapping check.
Signed-off-by: Keir Fraser <keir@xensource.com>

[xenstore] Don't create a transaction for singleton read/write operations.

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

[xenstore] Don't limit the number of concurrent transactions for domain0 connections
since both xend and the xenstore-* utilities used in the hotplug
scripts can easily reach the limit and we trust them anyway.

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

[XEN] Fix race in shadow invlpg
This fixes a crash under mmstress in SMP linux guests, where one vcpu
could remove shadow entries when another was reading them for invlpg.
Signed-off-by: Tim Deegan <Tim.Deegan@xensource.com>

[NET] back: Fix inter-domain networking when using rx-copy path.
Signed-off-by: Keir Fraser <keir@xensource.com>

[NET] front: Fix error handling to correctly skip bogus frags.
Signed-off-by: Keir Fraser <keir@xensource.com>

Revert 11728:30f13007be. Breaks Solaris guests.
Will need a new method for handling padded gzip files.
Signed-off-by: Keir Fraser <keir@xensource.com>

[XEND] Add missing line from changeset 11653:3b7e11cbeb940
Signed-off-by: Keir Fraser <keir@xensource.com>

[xend/hotplug] Remove happy gun logging.

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

[NET] back: Fix netif rate limiting.
Signed-off-by: Keir Fraser <keir@xensource.com>

Correct check for extended-cr3 support in __xen_guest fallback code.

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

[TOOLS] Fixes to miniterm serial terminal utility.
Signed-off-by: Keir Fraser <keir@xensource.com>

[NET] back: Refcount fixes to vif rate-limiting code.
Signed-off-by: Keir Fraser <keir@xensource.com>

[XEN] Zero PAE shadow l3es when destroying subshadows.

Normally, we can leave the contents of shadows in place when we destroy
them, but this is not the case for l3 subshadows, since they do not get
zeroed on reallocation unless the entire l3 page is unshadowed.
This fixes several crashes seen with SMP PAE HVM RHEL4.2 guests.
Signed-off-by: Tim Deegan <Tim.Deegan@xensource.com>

[XEND] No need to decompress the initrd when building a domain.
The guest OS should be capable of parsing an image file in
whatever format was produced by that OS's tools. Furthermore,
we have seen initrd images with trailing garbage. This causes
us to calculate the wrong size for the uncompressed data and hence
fail to build the guest.
Signed-off-by: Keir Fraser <keir@xensource.com>

[PRIVCMD] Fix trivial bug introduced in IOCTL_PRIVCMD_MMAP.
Signed-off-by: Keir Fraser <keir@xensource.com>

[LINUX] Various fixes for mmapping I/O and foreign memory pages.

First, auto-translate guests can use remap_pfn_range() rather than
direct_remap_pfn_range(). This actually works better because
remap_pfn_range() can legitimately assert VM_PFNMAP (this patch
removes this flag for direct_remap_pfn_range().

There are various cleanups and fixes to the privcmd interface:
 1. VMAs should be searched and used under the mmap semaphore
 2. Mapping should be single shot (since cirect_remap_pfn_range()
    expects the PTEs to be empty when it is called).
 3. Demand-fault population of the privcmd vma should be disallowed.
 4. Various others, including a more thorough check of input args.

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

Since shadow_update_paging_modes() will modify v->arch.hvm_vcpu.hw_cr3,
so we should always do shadow_update_paging_modes() before we writes to
GUEST_CR3 with v->arch.hvm_vcpu.hw_cr3.

Signed-off-by: Xiaohui Xin xiaohui.xin@intel.com
Signed-off-by: Xin Li <xin.b.li@intel.com>

[IA64] per vcpu vhpt

Implement per vcpu vhpt option. allocate VHPT per vcpu.
added compile time option, xen_ia64_pervcpu_vhpt=y, to enable it.
Its default is on.
added xen boot time option, pervcpu_vhpt=0, to disable it.

This patch focuses on vcpu migration between physical cpus
becaseu vcpu is heavily migrated with credit scheduler.
This patch tries to reduce vTLB flush when vcpu is migrated

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

[IA64] Remove test of dead CONFIG_XEN_IA64_DOM0_NON_VP

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

[IA64] added __read_mostly to some variables

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

[IA64] __read_mostly

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

[IA64] make efi_print()'s output more readable

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

[IA64] add a warning message when ioremap hypercall fails

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

[BLKTAP] Remove bogus extra public definition of sig_handler().
Also do not build the daemon -fPIC.
Signed-off-by: Keir Fraser <keir@xensource.com>

[BLKTAP] Fix trivial indentation snafu in previous changeset.
Signed-off-by: Keir Fraser <keir@xensource.com>

[BLKTAP] Fix crash at start-of-day in blktapctrld
if /local/domain does not exist.
Signed-off-by: Keir Fraser <keir@xensource.com>

[BLKTAP] Formatting and style cleanups to blktapctrl.
Signed-off-by: Keir Fraser <keir@xensource.com>

[BALLOON] Replace alloc_empty_page_range with new helper alloc_empty_pages_and_pagevec.
This is a better fit with all the callers, who now allocate discontiguous empty pages.
Signed-off-by: Keir Fraser <keir@xensource.com>

[BLK] tap: Allocate pages for foreign mappings individually rather than contiguously.
Signed-off-by: Keir Fraser <keir@xensource.com>

[HVM][SVM] Revert changeset 11679:a949bd6ceb85

Although in the correct spirit, it seems there is some other
dependency in the code that assumes ExtInts are queued until
thay can be immediately delivered (EFLAGS.IF==1). This patch
therefore caused instability in Linux guests.

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

[TPM] back: Allocate pages for foreign mappings individually rather than contiguously.
Signed-off-by: Keir Fraser <keir@xensource.com>

[NET] back: Allocate pages for foreign mappings individually rather than contiguously.
Signed-off-by: Keir Fraser <keir@xensource.com>

[NET] back: Remove unused rx_mmap area. Copying happens inside Xen, so
there is no need for a mapping area inside the driver for copying
receive path.
Signed-off-by: Keir Fraser <keir@xensource.com>

[BLK] back: Allocate pages for foreign mappings individually rather
than in one large contiguous region.
Signed-off-by: Keir Fraser <keir@xensource.com>

merge with xen-unstable.hg

Merge

[XEN] Fix memtype accessor domctls.
From: Jan Beulich <jbeulich@novell.com>
Signed-off-by: Keir Fraser <keir@xensource.com>

[XEN] Change microcode_update function interface inside Xen.

Confine the knowledge that the buffer gets accessed without use of
proper guest handle accessors to a single file.

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

[XEN] Remove unused MULTICALL_arg6, and unnecessary use of STR() macro.
Signed-off-by: Jan Beulich <jbeulich@novell.com>

Make sure that if one of the intermediate steps fails, a subsequent
make invocation doesn't think there's nothing to do.
This should be viewed as an intermediate step towards splitting up
the rule into three independent ones, passing intermediate files
from one to the next.

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

[VMX] Remove vcpu->arch.hvm_vmx.cpu_state

The flags in hvm_vmx.cpu_state were shadows of shadows of bits in
guest CR0_READ_SHADOW, CR4_READ_SHADOW and EFER, and were not being
kept in sync with the other shadows if changes were made while
running under vmxassist.  Just use the full shadows of those control
regs instead, and replace cpu_state with a single vmxassist flag.

Signed-off-by: Tim Deegan <Tim.Deegan@xensource.com>

[XM] Fix 'xm help' when called on a command-name prefix.
Signed-off-by: Masaki Kanno <kanno.masaki@jp.fujitsu.com>

[XEND] Fix pygrub when run server-side on a domain with blktap disks.

For some slightly-crazy reason the blktap patches changed the name
of the SEXPR block containing the disk definition 'vbd' to 'tap', even
though merely changing the 'uname' fragement was sufficient. eg

 (vbd (backend 0) (dev hda:disk) (uname file:/xen/rhel4x86_64.img)
 (mode w) )

changes to

 (tap (backend 0) (dev hda:disk) (uname tap:aio:/xen/rhel4x86_64.img)
 (mode w) )

If you create a domain with xm, then xm runs pygrub client side which
works fine. If you create a domain by feeding a block of SEXPR to Xend
over its network port, then pygrub is run server side. The code in
XenD which decides which disk to run pygrub against only looks for
'device/vbd' in the SEXPR so fails to see any blktap disks.

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

[NET] front: Fix size of grant_rx_ref[]. It should use the RX ring
size, rather than the TX ring size.
Signed-off-by: Jose Renato Santos <jsantos@hpl.hp.com>

[HVM/vncserver] Implement a 'vnclisten' option to limit the interface
that the VNC server from qemu listens on.

Defaults to only listen on 127.0.0.1

The old behaviour (listen on all interfaces) can be restored, by
- changing the system-wide default in /etc/xen/xend-config.sxp by adding:
(vnc-listen '0.0.0.0')
- changing individual domain config files by adding:
vnclisten="0.0.0.0"

Also allows specifying the hostname associated with an interface to limit
to that interface.

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

[NET] front: Include linux/moduleparam.h since we use module_param_*

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

[MBOOTPACK] Fix building mbootpack on x86/64.
Signed-off-by: Keir Fraser <keir@xensource.com>

[BLKTAP] Blktapctrl: Add asynchronous watch on dom0 name entry for blktapctrl on startup. This patch cleans up the synchronisation issues between blktapctrl and xend at initialisation if the Dom0 entries have not yet been populated.

Signed-off-by: Julian Chesterfield <julian@xensource.com>

[NET] front: If the kernel does not include GSO then look for TSO.

This allows us to take advantage of TSO when running as PV-on-HVM on
older kernels.

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

[NET] front: Invert the test for a kernel with GSO support.

This simplifies a subsequent patch.

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

[LIBXC] Fix ptrace memory-mapping code to map pagetables read-only.
Signed-off-by: Keir Fraser <keir@xensource.com>

[IA64] update buildconfigs for p2m exposure

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

[IA64] fix sparse tree build with p2m exposure disabled

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

[IA64] p2m exposure test module

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

[IA64] p2m exposure. linux side part.

This patch introduce compile time option XEN_IA64_EXPOSE_P2M
to enable this feature and boot option xen_ia64_p2m_expose to
disable the functionality.
This patch also introduce XEN_IA64_EXPOSE_P2M_USE_DTR to map p2m table
with dtr and boot option xen_ia64_p2m_expose_use_dtr to disable it.

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

[IA64] expose p2m table. xen side part.

This patch introduces xen compile time option, xen_ia64_expose_p2m=y.
default is enabled.

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

[IA64] Clean up extern declarations in arch/ia64/xen/domain.c

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

[IA64] remove getmeminfo.nr_pages trick from xencomm

getmeminfo nr_page trick is not anymore necessary.

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

[IA64] xencomm_privcmd_sched_op

Necessary for reboot on VTi domains.

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

[IA64] revert xen-ia64-unstable.hg cset 11039

No longer necessary with xencomm

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

Remove redundant '&' before array name in __RING_SIZE macro.
Signed-off-by: Richard Miller <9xen@hamnavoe.com>

[LINUX] Don't bother to write a 5-bit value to a 2-bit field.
From: Jan Beulich <jbeulich@novell.com>
Signed-off-by: Keir Fraser <keir@xensource.com>

[NET] back: Disable packet queuing when a client has no receive buffers.

This turns out to be dangerous as packets can be queued indefinitely,
introducing resource dependencies between guests (if the queued
packets originate from another virtual machine). For example, this
can prevent interfaces from being torn down and hence defunct
virtual machines from being destroyed.

The queuing functionality is retained as a module option, but a
warning is printed if it is enabled. It is not intended for general use!

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

[IOEMU] When vncunused is specified, allocate a port from 5900
rather than 5898.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>