Implement Nested-on-Nested.
This allows the guest to run nested guest with hap enabled.

Signed-off-by: Christoph Egger <Christoph.Egger@amd.com>
Acked-by: Tim Deegan <Tim.Deegan@citrix.com>
Committed-by: Tim Deegan <Tim.Deegan@citrix.com>

Implement generic piece to finally enable nested virtualization

Signed-off-by: Christoph Egger <Christoph.Egger@amd.com>
Acked-by: Tim Deegan <Tim.Deegan@citrix.com>
Committed-by: Tim Deegan <Tim.Deegan@citrix.com>

Implement SVM specific interrupt handling

Signed-off-by: Christoph Egger <Christoph.Egger@amd.com>
Acked-by: Tim Deegan <Tim.Deegan@citrix.com>
Committed-by: Tim Deegan <Tim.Deegan@citrix.com>

Implement SVM specific part for Nested Virtualization

Signed-off-by: Christoph Egger <Christoph.Egger@amd.com>
Acked-by: Tim Deegan <Tim.Deegan@citrix.com>
Committed-by: Tim Deegan <Tim.Deegan@citrix.com>

Handle interrupts (generic part)

Signed-off-by: Christoph Egger <Christoph.Egger@amd.com>
Acked-by: Eddie Dong <eddie.dong@intel.com>
Acked-by: Tim Deegan <Tim.Deegan@citrix.com>
Committed-by: Tim Deegan <Tim.Deegan@citrix.com>

Allow guest to enable SVM in EFER

Signed-off-by: Christoph Egger <Christoph.Egger@amd.com>
Acked-by: Eddie Dong <eddie.dong@intel.com>
Acked-by: Tim Deegan <Tim.Deegan@citrix.com>
Committed-by: Tim Deegan <Tim.Deegan@citrix.com>

When injecting an exception into L2 guest,
inject a #VMEXIT if L1 guest intercepts the exception

Signed-off-by: Christoph Egger <Christoph.Egger@amd.com>
Acked-by: Eddie Dong <eddie.dong@intel.com>
Acked-by: Tim Deegan <Tim.Deegan@citrix.com>
Committed-by: Tim Deegan <Tim.Deegan@citrix.com>

Allow paged real mode during vmrun emulation.
Emulate cr0 and cr4 when guest does not intercept them.

Signed-off-by: Christoph Egger <Christoph.Egger@amd.com>
Acked-by: Eddie Dong <eddie.dong@intel.com>
Acked-by: Tim Deegan <Tim.Deegan@citrix.com>
Committed-by: Tim Deegan <Tim.Deegan@citrix.com>

Nested Virtualization core implementation

Signed-off-by: Christoph Egger <Christoph.Egger@amd.com>
Acked-by: Eddie Dong <eddie.dong@intel.com>
Acked-by: Tim Deegan <Tim.Deegan@citrix.com>
Committed-by: Tim Deegan <Tim.Deegan@citrix.com>

add nestedhvm function hooks for svm/vmx specific code

Signed-off-by: Christoph Egger <Christoph.Egger@amd.com>
Acked-by: Eddie Dong <eddie.dong@intel.com>
Acked-by: Tim Deegan <Tim.Deegan@citrix.com>
Committed-by: Tim Deegan <Tim.Deegan@citrix.com>

Data structures for Nested Virtualization

Signed-off-by: Christoph Egger <Christoph.Egger@amd.com>
Acked-by: Eddie Dong <eddie.dong@intel.com>
Acked-by: Tim Deegan <Tim.Deegan@citrix.com>
Committed-by: Tim Deegan <Tim.Deegan@citrix.com>

tools: Add nestedhvm guest config option

Signed-off-by: Christoph Egger <Christoph.Egger@amd.com>
Acked-by: Eddie Dong <eddie.dong@intel.com>
Acked-by: Tim Deegan <Tim.Deegan@citrix.com>
Committed-by: Tim Deegan <Tim.Deegan@citrix.com>

Remove __init specifier from function declarations in header files.

The specifier only needs to be added to the function's definition.

At the same time, fix init_cpu_to_node() to be __init rather than
__devinit (it is only called at boot time).

Signed-off-by: Keir Fraser <keir@xen.org>

[VTD] Fixes to ACPI DMAR flag checks.

 * platform_supports_{intremap,x2apic} should not be marked __init as
   they are used during S3 resume.
 * DMAR flags should be taken from the table passed to
   acpi_parse_dmar() -- this is the trusted copy of the DMAR, when
   running in TXT mode.

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

x86, amd, MTRR: remove k8_enable_fixed_iorrs()

AMD64 defines two special bits (bit 3 and 4) RdMem and WrMem in fixed
MTRR type. Their values are supposed to be 0 after BIOS hands the
control to OS according to AMD BKDG. Unless OS specificially turn them
on, they are kept 0 all the time. As a result, k8_enable_fixed_iorrs()
is unnecessary and removed from upstream kernel (see
https://patchwork.kernel.org/patch/11425/). This patch does the same
thing.

Signed-off-by: Wei Huang <wei.huang2@amd.com>

x86, amd, MTRR: correct DramModEn bit of SYS_CFG MSR

Some buggy BIOS might set SYS_CFG DramModEn bit to 1, which can cause
unexpected behavior on AMD platforms. This patch clears DramModEn bit
if it is 1.

Signed-off-by: Wei Huang <wei.huang2@amd.com>

libxl: Drop unnecessary \n from log message

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>

libxl: specific explicit disk image format to new qemu

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Signed-off-by: Ian Jackson <ian.jackson@eu.citrix.com>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>

libxl: specify disks using supported command line syntax for new qemu

The -hdX syntax is only retained for compatibility reasons and the
-sdX syntax doesn't even exist.

Additionally convert the first four non-SCSI disks to hd[a-d] and
ignore any further non-SCSI disks (since qemu only supports 4 IDE
devices).

SCSI disks are passed through as is. qemu-xen was limited to 7 SCSI
devices but upstream qemu supports 256, therefore do not limit the
number of disks on the libxl side.

qemu-xen did all this itself internally.

Fixes "qemu: -xvda: invalid option" and allows PVHVM to work with
upstream qemu.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>

libxl: pass list of disks to libxl__build_device_model_args

Given that we have the information available this is preferable to
picking it out of xenstore instead. We already do this for VIFs.

Only the qemu upstream version makes use of it since old qemu-xen
actually parses xenstore itself.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>

libxl: return raw disk and partition number from libxl__device_disk_dev_number

Optional parameters, caller to follow.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>

libxl: explicitly set disk format in libxl__append_disk_list_of_type

Ideally we should be able to infer the format from something stashed
in xenstore but this is better than letting users see garbage values.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>

passthrough: use domain pirq as index of struct hvm_irq_dpci's hvm_timer array

Since d->nr_pirqs is guaranteed to be not larger than nr_irqs,
indexing arrays by the former ought to be preferred. In the case
given, the indices so far had to be computed specially in a number of
cases, whereas the indexes use now are all readily available.

This opens the possibility to fold the ->mirq[] and ->hvm_timer[]
members of struct hvm_irq_dpci into a single array, possibly with some
members overlayed in a union to reduce size (see
http://lists.xensource.com/archives/html/xen-devel/2011-03/msg02006.html).
Such space saving wouldn't, however, suffice to generally get the
respective allocation sizes here to below PAGE_SIZE, not even when
converting the array of structures into an array of pointers to
structures. Whether a multi-level lookup mechanism would make sense
here is questionable, as it can be expected that for other than Dom0
(which isn't hvm, and hence shouldn't use these data structures - see
http://lists.xensource.com/archives/html/xen-devel/2011-03/msg02004.html)
only very few entries would commonly be used here. An obvious
alternative would be to use rb or radix trees (both currently only
used in tmem).

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

x86: introduce alloc_vcpu_guest_context()

This is necessary because on x86-64 struct vcpu_guest_context is
larger than PAGE_SIZE, and hence not suitable for a general purpose
runtime allocation. On x86-32, FIX_PAE_HIGHMEM_* fixmap entries are
being re-used, whiule on x86-64 new per-CPU fixmap entries get
introduced.  The implication of using per-CPU fixmaps is that these
allocations have to happen from non-preemptable hypercall context
(which they all do).

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

x86: split struct domain

This is accomplished by converting a couple of embedded arrays (in one
case a structure containing an array) into separately allocated
pointers, and (just as for struct arch_vcpu in a prior patch)
overlaying some PV-only fields with HVM-only ones.

One particularly noteworthy change in the opposite direction is that
of PITState - this field so far lived in the HVM-only portion, but is
being used by PV guests too, and hence needed to be moved out of
struct hvm_domain.

The change to XENMEM_set_memory_map (and hence libxl__build_pre() and
the movement of the E820 related pieces to struct pv_domain) are
subject to a positive response to a query sent to xen-devel regarding
the need for this to happen for HVM guests (see
http://lists.xensource.com/archives/html/xen-devel/2011-03/msg01848.html).

The protection of arch.hvm_domain.irq.dpci accesses by is_hvm_domain()
is subject to confirmation that the field is used for HVM guests only
(see
http://lists.xensource.com/archives/html/xen-devel/2011-03/msg02004.html).

In the absence of any reply to these queries, and given the early
state of 4.2 development, I think it should be acceptable to take the
risk of having to later undo/redo some of this.

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

x86: move pv-only members of struct vcpu to struct pv_vcpu

... thus further shrinking overall size of struct arch_vcpu.

This has a minor effect on XEN_DOMCTL_{get,set}_ext_vcpucontext - for
HVM guests, some meaningless fields will no longer get stored or
retrieved: reads will now return zero, and writes are required to be
(mostly) zero (the same as was already done on x86-32).

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

x86: split struct vcpu

This is accomplished by splitting the guest_context member, which by
itself is larger than a page on x86-64. Quite a number of fields of
this structure is completely meaningless for HVM guests, and thus a
new struct pv_vcpu gets introduced, which is being overlaid with
struct hvm_vcpu in struct arch_vcpu. The one member that is mostly
responsible for the large size is trap_ctxt, which now gets allocated
separately (unless fitting on the same page as struct arch_vcpu, as is
currently the case for x86-32), and only for non-hvm, non-idle
domains.

This change pointed out a latent problem in arch_set_info_guest(),
which is permitted to be called on already initialized vCPU-s, but
so far copied the new state into struct arch_vcpu without (in this
case) actually going through all the necessary accounting/validation
steps. The logic gets changed so that the pieces that bypass
accounting
will at least be verified to be no different from the currently active
bits, and the whole change will fail in case they are. The logic does
*not* get adjusted here to do full error recovery, that is, partially
modified state continues to not get unrolled in case of failure.

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

Remove direct cpumask_t members from struct vcpu and struct domain

The CPU masks embedded in these structures prevent NR_CPUS-independent
sizing of these structures.

Basic concept (in xen/include/cpumask.h) taken from recent Linux.

For scalability purposes, many other uses of cpumask_t should be
replaced by cpumask_var_t, particularly local variables of functions.
This implies that no functions should have by-value cpumask_t
parameters, and that the whole old cpumask interface (cpus_...())
should go away in favor of the new (cpumask_...()) one.

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

xsm: Error code consistency

Signed-off-by: Machon Gregory <mbgrego@tycho.ncsc.mil>

libxl: add CODING_STYLE

libxenlight and xl grew enough to need a CODING_STYLE, that I blatantly
copied from qemu and linux, just adding few specific modifications.
The result should be as less controversial as possible, mostly
documenting what we are already doing.

[ Message and document originally posted to xen-devel on 2010-09-01 ]

Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Signed-off-by: Ian Jackson <ian.jackson@eu.citrix.com>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>

x86/mm: fix PSE alignment checks in pagetable walker.

The PSE handling doesn't check bits 1...8 (or 1...9 for non-PAE guests)
being zero, thus allowing bad (not 2Mb/4Mb aligned) large pages to be
handled.

Signed-off-by: Jan Beulich <jbeulich@novell.com>
Acked-by: Tim Deegan <Tim.Deegan@citrix.com>

x86: cleanup bogus CONFIG_ACPI_PCI uses

We're building for one case (CONFIG_ACPI_PCI defined) only, yet still
had the other case's code in there. Additionally there was quite a bit
of pseudo-duplication between disabled(!) DMI scan and ACPI boot code.

acpi_pci_disabled had only a single reader, which is off by default
(i.e. must be enable on the command line), so it seems pointless to
keep it.

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

x86/ACPI: __init-annotate

xen/arch/x86/acpi/boot.c consists of almost only code/data in .init.*,
so move the few bits that aren't into a new file and then use the
recently introduced .init.o mechanism to move all the literal strings
into .init.rodata.

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

amd-iommu: __init-annotate

Besides marking a few more items __init/__initdata, use the recently
introduced .init.o mechanism to move all the literal strings into
.init.rodata in those files that consist of only contributions to
.init.*.

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

move register_cpu_notifier() into .init.text

With no modular drivers, all CPU notifier setup is supposed to happen
during boot. There also is a respective comment in the function.=20

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

move setup_irq() into .init.text

With no modular drivers, all interrupt setup is supposed to happen
during boot.

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

move request_irq() into .init.text

With no modular drivers, all interrupt setup is supposed to happen
during boot.

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

x86: eliminate old-CPU support bits

Some pieces can be removed altogether, others are only of possible use
in 32-bit builds.

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

xen: fix reloc.S generation for non-gnu toolchain

The output of AT&T UNIX and GNU od(1) are different.

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

xentrace: remove unneeded debug printk

The pointer value in case of an allocation failure is rather
uninteresting.

Signed-off-by: Olaf Hering <olaf@aepfle.de>
Acked-by: George Dunlap <george.dunlap@eu.citrix.com>

xentrace: correct formula to calculate t_info_pages

The current formula to calculate t_info_pages, based on the initial
code, is slightly incorrect. It may allocate more than needed.
Each cpu has some pages/mfns stored as uint32_t.
That list is stored with an offset at tinfo.

Signed-off-by: Olaf Hering <olaf@aepfle.de>
Acked-by: George Dunlap <george.dunlap@eu.citrix.com>

[VTD] check BIOS settings before enabling interrupt remapping or x2apic

Check flags field in ACPI DMAR structure before enabling interrupt
remapping or x2apic.  This allows platform vendors to disable
interrupt remapping or x2apic features if on board BIOS does not
support them.

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

libxl: typedef libxl_action_on_shutdown

This is the libxl convention. By using the "typedef enum foo { } foo;"
syntax we can retain backwards compatibility with existing users.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>

tools: remove some .o and binary files on clean

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>

tools: remove pattern matched linking rules

Most subdirs only build a single tool to start with and those which
build multiple tools often have different linkage requirements.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>

tools: Remove $(CFLAGS) from links lines.

The relevant variable in these circumstances is called $(LDFLAGS).

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>

tools: ocaml: drop redefinition of xen_mb

This is always available to the in-tree build.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>

tools: ocaml: drop some commented out build rules.

These are covered by tools/ocaml/Makefile.rules now

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>

tools: ocaml: push CFLAGS usage down into the specific bindings

So each binding only builds against what it requires.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>

tools: xenstored: make usage info of xenstored accurate

The usage printed for xenstored was not in sync with the parameters
defined in the code.

Signed-off-by: juergen.gross@ts.fujitsu.com
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>

libxl: fix memory management in "xl network-attach"

The libxl_device_nic struct has strings which are initially strdup()ed
and then free()ed in libxl_device_nic_destroy(). In the
"network-attach" parser we need to free() the existing string and
strdup((*argv) + N), rather than just copying the pointer.

Signed-off-by: David Scott <dave.scott@eu.citrix.com>
Acked-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>

tools: ocaml: do not try to link dynamic objects into .a libraries

lib<foo>_stubs.a was linking against $(LIBS_<foo>) which only includes
dynamic linkage information which is not useful for a static library.

$(LIBS_<foo>) is still used when linking <foo>.cmxa

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>

libxl: use system include for libxl_uuid.h

We install libxl_uuid.h into the system include directory and refer to
it from libxl.h, hence we should include with <> not "".

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>

tools: ocaml: Add an ocaml binding for libxl_get_topologyinfo

Add an ocaml binding for libxl_get_topologyinfo in the style of
libxl_get_physinfo.

The function returns an array of optional records, one per possible pCPU.

Signed-off-by: David Scott <dave.scott@eu.citrix.com>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>

tools: blktap2: copy xenstore/hashtable.h into blktap2 (really)

hg add the missing files from 23112:e15ca5c3effc

libxl: export libxl_utils.h

libxl_utils.h is the header for libxlu.{so,a} and should have been
exported for the benefit of libxl callers.

Signed-off-by: Ian Jackson <ian.jackson@eu.citrix.com>
Acked-by: Jim Fehlig <jfehlig@novell.com>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>

tools: blktap2: copy xenstore/hashtable.h into blktap2

hashtable.c has already been copied, forked and modified, there
doesn't seem much point in avoiding the same for the header until
someone feels motivated to properly refactor.

Add comments to the various duplicated files cross-referencing each
other for future reference and as a barrier to forking again...

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>

tools: ocaml: xl: propagate simplfied libxl interface to PCI BDFs to bindings.

22166:251694a87f1d changed the libxl interface to remove the need for
users to understand BDF encoding but did not propagate the change to
the ocaml bindings. Do that now.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>

libxl: remove "reg" and "enable" fields from PCI device.

The structure of the BDF argument used with PCI passthrough related
hypercalls was taken from the structure of the PCI config_address
register (I/O port 0xCF8) which allows I/O mapped access to PCI
configuration space but these fields have no meaning in the context of
PCI passthrough configuration and hence do not need to be exposed via
libxl.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>

remus: fix check for installed qdiscs on ifb

current check includes ingress and pfifo_fast.
Add mq to the list of allowed qdiscs already installed
on ifb. This patch fixes cases where remus fails to start,
due to an mq qdisc already present on the vif.

Signed-off-by: Shriram Rajagopalan <rshriram@cs.ubc.ca>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>

spinlock: Define check_barrier() for non-debug build, fixes build.

Signed-off-by: Keir Fraser <keir@xen.org>

x86: Remove _PAGE_NX defintiion (with implicit use of cpu_has_nx).

Most users can use _PAGE_NX_BIT directly.

The few genuine users in mm.c can do the cpu_has_nx check more clearly
in other ways.

Signed-off-by: Keir Fraser <keir@xen.org>

x86: cleanup after tboot fix (c/s 23101:dd386a4b6595)

When submitting the patch that became 23101:dd386a4b6595 I forgot that
I actually intended to remove the pointless IRQ disabling and
restoring (and the then pointless comment). Pointless because the
code, other than its pre-23013:65d26504e843 original, now runs before
interrupts get enabled for the first time.

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

xend: Fix startup after removal of ACM support.

Signed-off-by: Keir Fraser <keir@xen.org>

x86_32: Fix _raw_read_trylock() build on some gcc versions.

Was broken by 23099:612171ff82ea.

A bool_t is a single byte, and needs a 'q' register constraint. Avoid
the whole issue by changing the variable to an int, and explicitly
specify the operand suffix as 'l' for good measure.

Signed-off-by: Keir Fraser <keir@xen.org>

x86: __pirq_guest_eoi() must check it is called for a fully
guest-bound irq before accessing desc->action.

Signed-off-by: Keir Fraser <keir@xen.org>

x86/ATS: remove unreferenced lock member from struct pci_ats_dev

Otherwise this may give the false impression that some explicit
locking is done in this code.

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

x86: fix tboot after c/s 23013:65d26504e843 (ACPI cleanup)

TXT code calls acpi_parse_dmar() with a transient copy of the DMAR
table, and hence storing the pointer for later reference was wrong.
Obtain the pointer in acpi_dmar_init() instead.

Signed-off-by: Jan Beulich <jbeulich@novell.com>
Tested-by: Gang Wei <gang.wei@intel.com>

Remove spin_barrier_irq(). It is pointless.

Add a barrier-appropriate consistency check to spinlock.c, and add
code comments to explain why barrier operations are more relaxed than
lock-acquisition operations.

Signed-off-by: Keir Fraser <keir@xen.org>

rwlock: Allow to scale to 2^31-1 readers on x86.

Also rework to match the 'trylock' style of raw function used for
spinlocks.

Inspired by Jan Beulich's patch to do similar improved scaling.

Signed-off-by: Keir Fraser <keir@xen.org>

tools: vnet: Remove

Build has been broken since at least 18969:d6889b3b6423 (early
2009) and it has been unhooked from the top level build since forever
AFAICT. The last actual development (as opposed to tree wide
cleanups and build fixes) appears to have been 11594:6d7bba6443ef in
2006. The functionality of vnet has apparently been superceded by
VLANs, ebtables, Ethernet-over-IP etc all of which are well integrated
with upstream kernels and distros.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Signed-off-by: Keir Fraser <keir@xen.org>

Remove unmaintained Access Control Module (ACM) from hypervisor.

Signed-off-by: Keir Fraser <keir@xen.org>

x86/hpet: eliminate cpumask_lock

According to the (now getting removed) comment in struct
hpet_event_channel, this was to prevent accessing a CPU's
timer_deadline after it got cleared from cpumask. This can be done
without a lock altogether - hpet_broadcast_exit() can simply clear
the bit, and handle_hpet_broadcast() can read timer_deadline before
looking at the mask a second time (the cpumask bit was already
found set by the surrounding loop).

Signed-off-by: Jan Beulich <jbeulich@novell.com>
Acked-by: Gang Wei <gang.wei@intel.com>

xentrace: use consistent printk prefix

Signed-off-by: Olaf Hering <olaf@aepfle.de>

xentrace: update comments

Fix a typo, remove redundant comment.

Signed-off-by: Olaf Hering <olaf@aepfle.de>

xentrace: remove gdprintk usage since they are not in guest context

Signed-off-by: Olaf Hering <olaf@aepfle.de>

xentrace: print calculated numbers in calculate_tbuf_size()

Print number of pages to allocate for per-cpu tracebuffer and metadata
to ease debugging when allocation fails.

Signed-off-by: Olaf Hering <olaf@aepfle.de>

xentrace: fix t_info_pages calculation.

Signed-off-by: Olaf Hering <olaf@aepfle.de>

tools/xenstat: 23075:b4351d57464b added libxenstat.so but didn't build fPIC

This broke on x86_64.

Reported-by: Olaf Hering
Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>

COPYING: Add discussion about GPL compatibility of Modified BSD licenses.

Signed-off-by: Keir Fraser <keir@xen.org>

tools: add missing libxl_uuid.c

This should have been added in 23087:d9faf0aa9755.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>

tools/libxl: move uuid wrapper functions out of line.

This isolates users of libxenlight from the need to know about the
different OS schemes for UUIDs, in particular the linkage
requirements.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>

tools: drop further uses of -Wp,-MD,.$(@F).d to generate dependencies.

19025:bd78714b8594 switched to "-MMD -MF" in preference, adding the
correct runes to tools/Rules.mk but appears to have missed some other
uses of -Wp,...

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>

libxl: remove XS transaction from public API.

All external users pass 0 anyway so make the version of
libxl_domain_rename which takes a transaction internal and provide an
external facing function which does not expose a transaction.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>

libxl: drop protype for libxl_ctx_set_log

Nothing defines or uses it.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>

tools: provide generic rules for compiling .S files

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>

tools: flask: remove $(LOADLIBES)

Probably a typo and never defined, I can't tell what it was supposed
to be.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>

tools: flask: Remove BASECFLAGS, just use CFLAGS

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>

tools: users of libxl currently need to see libxc and libxenstore
headers

So add them to CFLAGS_libxenlight

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>

tools: ocaml: link evtchn bindings against libxenctrl

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>

tools: ocaml: link xl bindings against libxl

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>

tools: allow Makefiles to define CFLAGS_foo.o

To provide per-file flags.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>

tools: Drop $(X11_LDPATH) from build.

Nothing uses it.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>

tools: xenstat: install and use shared library

Remove .so files from the .a file.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>

Define new <pfn.h> header for PFN_{DOWN,UP} macros.

Signed-off-by: Keir Fraser <keir@xen.org>

tools: consistently use $({CFLAGS,LDLIBS}_libxenlight) instead of open coding.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>

tools: consistently use $({CFLAGS,LDLIBS}_libxenstore) instead of open coding.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>

tools: consistently use $({CFLAGS,LDLIBS}_libxenctrl) instead of open coding.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>

tools: consistently use $(CFLAGS_xeninclude) instead of open coding.

Renamed from the slightly ambiguous CFLAGS_include.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>

tools: Drop XEN_XC variable

There is nothing to include in the python bindings source directory
and likely never was/will be ...

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>