Fix libxc warnings with gcc-4.1 and glibc-2.3.90.

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

Many fixes and cleanups for lomount:
- Fixed several overflows, off-by-one, and uninitialized variables.
- Added well-defined exit codes.
- Proper handling of system()'s return value.
- Errors parsing partition table cause it to stop now.
- etcetera...

Tested on 32 and 64 bit, with valgrind, with physical disks and disk
images.

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

Change debug level of vmx_io_instruction to DBG_LEVEL_IO.

From: Himanshu Raj

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

Add empty files need to fix ia64 build.

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

Clean up use of wait_event_interruptible().

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

linux-x86-64: eliminate duplicate exports
Symbols are exported from files shared with i386.

From: Jan Beulich <JBeulich@novell.com>
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Remove unnecessary include statements.

Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

linux: clean up setup_arch_post.h
- machine_specific_memory_setup() was unused on x86-64
- machine_specific_modify_cpu_capabilities() must not be __init (for CPU hotplug)
- machine_specific_modify_cpu_capabilities() should be called in similar ways on 32- and 64-bits

From: Jan Beulich <JBeulich@novell.com>
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

While native code range-checks the requested address, possibly
truncates the range, and then reads/writes page-wise until possibly
encountering a -EFAULT issue, Xen code accessed the whole range in a
single step, thus not allowing partially succeeding accesses.

From: Jan Beulich

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

A few changes to the new 'guest handle' interface:

 DEFINE_GUEST_HANDLE -> define_guest_handle

 GUEST_HANDLE        -> guest_handle

 New __define_guest_handle allows handle name different from
 encapsulated type's name (useful for awkward typenames).

 Got rid of xen_ulong, now use guest_handle(ulong).

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

Make sure that installed files have sensible permissions and are owned by the
user running install. Also clean up the install script a little.

Signed-off-by: Horms <horms@verge.net.au>
Signed-off-by: Keir Fraser <keir@xensource.com>

Fix problems when creating HVM guest on PAE host with >4G physical memory.

Signed-off-by: Yunhong Jiang yunhong.jiang@intel.com
Signed-off-by: Jun Nakajima jun.nakajima@intel.com

merge

Fix printing of u64 value 'msr_content' to use PRIx64 format.

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

Merge xen-ia64-unstable.hg

Merged.

Fix violation of C90 mixed-code-and-declarations restriction.

Signed-off-by: Ewan Mellor <ewan@xensource.com>

Removing out-of-date unused file.
current code available in io/xs_wire.h

Signed-off-by: Vincent Hanquez <vincent@xensource.com>

Fix printf.

Signed-off-by: Ewan Mellor <ewan@xensource.com>

Added a basic integrity checker, and some basic ability to recover from store
corruption, rather than just spewing error messages and exiting.

Added a xenstore-control executable, which sends commands to xenstored.
Currently, the only command is 'check', which triggers an integrity check.
(The integrity check is also triggered whenever a corrupted store is detected).

Signed-off-by: Ewan Mellor <ewan@xensource.com>

Silence the messages that are emitted when removing nodes that already have
been removed.  This is fine, in a cleanup script.

Signed-off-by: Ewan Mellor <ewan@xensource.com>

Avoid a race between Xend removing the vm entries corresponding to a domain,
and the hotplug scripts checking whether that domain is still alive.  This
triggers the "xenstore-read: couldn't read path /local/domain/107/vm" message
in /var/log/xen-hotplug.log, followed by "Device XYZ is mounted in a guest
domain and so cannot be mounted now".

Closes bug #514.

Signed-off-by: Ewan Mellor <ewan@xensource.com>

Fix problems with HVM PAE guests with >4GB memory.

Signed-off-by: Xiaohui Xin <xiaohui.xin@intel.com>
Signed-off-by: Jun Nakajima <jun.nakajima@intel.com>

Add SVM 32bit msr support (combined both 32bit with 64bit functions)
and enable vlapic for 64bit.

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

Clean up SVM 32-bit exits.S file (remove all tsc/fpu).

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

Revert two uses of CONFIG_PAGING_LEVELS to CPP rather than C predicate.
Fixes PAE and 64-bit builds.

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

[IA64] sparse update for 2.6.16-rc5

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

[IA64] default to SMP

UP builds are not maintained on xen-unstable, we seem to be
running SMP well enough to make the transition.

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

[IA64] Turn on CONFIG_XEN_BLKDEV_BACKEND

We relied on the Makefile for building this before, now we
need it set in our .config file.

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

merge

map_domain_page() now handles running on idle page tables.

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

Disallow cmpxchg8b ptwr emulation for non-pae. Also a few cleanups.

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

[IA64] Every processor call xen_early_setup (required on SMP-guest to set ivt).

Only BP set running_on_xen.
xensetup.S slightly compacted.

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

[IA64] warning cleanup

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

Remove unused #include <sys/ioctl.h>.

Signed-off-by: Ewan Mellor <ewan@xensource.com>

[IA64] VTI: updated vtlb, support_non_contiguous memory on vtidomain

Previously VTI-domain only supported contiguous memory,
this patch is intended to make VTI-domain support non-contiguous memory.

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

Clean wrpt state when emulating CR3 write.

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

SVM patch to add 64bit hv support.
This patch only modifies svm files.
Tested with c/s 9015 with 32bit hv using UP Dom0, with UP linux and
winxpsp1 unmodified guests.
Tested with c/s 9015 with 64bit hv using UP Dom0, with 32bit and 64bit
UP linux and 32bit winxpsp1.

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

[IA64] Unused fields in mm_struct commented out.

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

[IA64] Misc clean-up and warnings removal.

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

[IA64] xen_timer_interval removed (was not used).

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

[IA64] delivery_mask removed (was unused).

The field was renamed in arch-ia64.h for backward compatibility with xen linux.

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

[IA64] Found the lost memory

This patch can work around virtual memmap/discontig memory issues,
but memory gain may not be realized.

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

Allow tpmfront/tpmback to be built as modules.

From: Jan Beulich

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

Allow netback to be built as a module.

From: Jan Beulich

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

Allow blkback to be built as a module.

From: Jan Beulich

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

[IA64] cleanup warning in xen/ia64 (arch/ia64/xen)

These patches fix many issue (ex. panic dom0, oops domU/dom0...).
we will fix compile warnnings one step at a time.

Signed-off-by: Tsunehisa Doi <doi.tsunehisa@jp.fujitsu.com>
Signed-off-by: Kouya SHIMURA <kouya@jp.fujitsu.com>
Signed-off-by: Masaki Kanno <kanno.masaki@jp.fujitsu.com>
Signed-off-by: Akio Takebe <takebe_akio@jp.fujitsu.com>

[IA64] cleanup warning in xen/ia64 (arch/ia64/linux-xen)

These patches fix many issue (ex. panic dom0, oops domU/dom0...).
we will fix compile warnnings one step at a time.

Signed-off-by: Tsunehisa Doi <doi.tsunehisa@jp.fujitsu.com>
Signed-off-by: Kouya SHIMURA <kouya@jp.fujitsu.com>
Signed-off-by: Masaki Kanno <kanno.masaki@jp.fujitsu.com>
Signed-off-by: Akio Takebe <takebe_akio@jp.fujitsu.com>

[IA64] cleanup warning in xen/ia64 (arch/ia64/linux)

These patches fix many issue (ex. panic dom0, oops domU/dom0...).
we will fix compile warnnings one step at a time.

Signed-off-by: Tsunehisa Doi <doi.tsunehisa@jp.fujitsu.com>
Signed-off-by: Kouya SHIMURA <kouya@jp.fujitsu.com>
Signed-off-by: Masaki Kanno <kanno.masaki@jp.fujitsu.com>
Signed-off-by: Akio Takebe <takebe_akio@jp.fujitsu.com>

[IA64] cleanup warning in xen/ia64 (include/asm-ia64)

These patches fix many issue (ex. panic dom0, oops domU/dom0...).
we will fix compile warnnings one step at a time.

Signed-off-by: Tsunehisa Doi <doi.tsunehisa@jp.fujitsu.com>
Signed-off-by: Kouya SHIMURA <kouya@jp.fujitsu.com>
Signed-off-by: Masaki Kanno <kanno.masaki@jp.fujitsu.com>
Signed-off-by: Akio Takebe <takebe_akio@jp.fujitsu.com>

Avoid using unexported add_timer_on() function in netback driver.

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

[IA64] Cleanup hypercall.h

Cleanup to hypercall.h of ia64 xenlinux, and remove used stuff.

Signed-off-by Kevin Tian <kevin.tian@intel.com>

Sketch a new interface for transferring hypercall arguments in memory.
Instead of manipulating guest virtual addresses, 'guest handles' are
passed across the hypercall interface, which may only be manipulated
via the interface exported by guest_access.h.

So far this has only been applied to the memory_op hypercall. The
interfaces are still subject to change. Other hypercalls can be updated
after the interface is agreed upon.

Also cleaned up the hypercall_create_continuation() interface to take
a format string and different-typed varargs (they do not all need to
be castable to longs).

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

Clean up management of 'initial images' temporary buffers.
Avoid double-freeing boundary pages.

Signed-off-by: Anthony Xu <anthony.xu@intel.com>
Signed-off-by: Keir Fraser <keir@xensource.com>

Fix tools Makefiles to respect CFLAGS. Disable stack-smash
protection when building vmxassist/hvmloader.

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

Add 'm' debug key to print free-memory info.

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

Fix a multi HVM domain issue (bugzilla bug #542):
        This issue can be reproduced on SMP platform, while the domain 0
is UP. The reason is, after finishing a dma request, the dma thread will
trigger the interrupt and then clear the call back function. When guest
get the interrupt , it will try to check the call back function, if it
is set, then it will trigger a dma request again. So if the checking for
the callbackfunction happens between trigger interrupt and clear
callback function on dma thread, it will cause NULL function calling.
After with domain 0 UP and smp platform, this situation can be
reproduced easily.

Signed-off-by: Yunhong Jiang <yunhong.jiang@intel.com>

Fix SVM cr0 handling.

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

Fix SVM PAE support.

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

Fix SVM guest pin timer migration logic.

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

Fix SVM vlapic logic.

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

Removing hvm_safe_block() from the SVM hlt function avoids SVM guest problems.
This is a temporary solution.

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

[IA64] fix SMP bug for vhpt

vhpt_paddr & vhpt_pend are now per-cpu variables, since VHPT is allocated
per cpu.  This should really improve stability.

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

[IA64] warning removal: NR_CPUS

linux/threads nullified (moved to linux-null).

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

[IA64] SMP_HOST: Alloc vhpt from domheap

It's more reasonable to alloc vhpt from domheap, instead of xenheap.

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

Merged.

Add support for configuring feature flags when starting guests.

Guest configuration files may now contain a features= declaration
listing the features which should be enabled in a | separated
list.

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

Update to Linux 2.6.16-rc5.

Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Check feature flag supervisor_mode_kernel in dom0 builder.

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

Reconstitute the ip parameter correctly.  Closes bug #491.

Signed-off-by: Ewan Mellor <ewan@xensource.com>

Allow a guest to switch pagetables 'atomically'. That is, the new pagetables
do not need to be read-only in the old pagetables, and the old pagetables do
not need to be read-only in the new pagetables.

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

Use cleverer sed pattern to avoid use of head -1.

Signed-off-by: Ewan Mellor <ewan@xensource.com>

Break out the hotplug cleanup code into xen-hotplug-cleanup.  Claim the block
script's lock before doing the cleanup, to avoid a race that causes the message
"xenstore-list: could not list path backend/vbd/40" in /var/log/hotplug.log,
with a subsequent refusal to create the domain, after a domain reboot.

Signed-off-by: Ewan Mellor <ewan@xensource.com>

Merged.

Move files that have appeared in the wrong place due to incorrect patch -p setting.

Signed-off-by: Ewan Mellor <ewan@xensource.com>

xen/ia64 uaccess is still linux style with different declaration as
common xen. It's meaningless to patch original file with many #ifdef
XEN, and so instead copy that file to create a new xen specific
uaccess file, with unused lines removed. May need render later.

Or else it breaks compilation on xen-unstable.

Signed-off-by Kevin Tian <kevin.tian@intel.com>

Fix a missed unmap in __entry().
On x86_32p, after several creation/destroy of VMX domain, xen crashes
because we forget to unmap phys_table.

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

While testing TCP & UDP tests (coming soon), I tracked down a bug in the
IP addressing scheme
used by xm-test: each machine running network tests is using the same IP
addresses, so two machines
in the same network can cross-pollinate/interfere with each other's tests.

This patch changes the IP addressing scheme to add some randomness. It's
gone through a couple of dozen
passes of xm-test, some on -bridge and some on -route, while xm-test was
running on another box on the same network.

Signed-off-by: Jim Dykman <dykman@us.ibm.com>

The attached patch does the following:

- introduces two basic tests for the virtual TPM (skipped on hardware
where a TPM is not available or vtpm manager has not been started
manually before)
- adapts the Makefiles to make the test cases available
- splits up the hotplug files into two files where the vtpm-common.sh
can be sourced by 'other' scripts as well
- uses echo "$var" to return variables from functions
- gets rid of the 'set +e" work-around
- introduces a script (vtpm-delete) to delete entries from the virtual
TPM directory in /etc/xen/vtpm.db

The xm-tests will skip over the 2nd test on most systems and the patch
has otherwise no effect on existing tests.

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

The attached patch breaks out

- log() function from xen-hotplug-common.sh -> logging.sh
- locking-related functions from xen-hotplug-common.sh  -> locking.sh

- xen-hotplug-common.sh 'sources' both those files now to not loose
these functions (which should make the break-out atomic)

- adapts the Makefile to install locking.sh and logging.sh at the right
place.

I ran the xm-tests over it and no (additional) errors appear due to this
change.

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

Add a compile time option to enable domain 0 running in ring 0.

In this mode only a single guest kernel is supported.

This mode only works for x86/32 (not x86/64).

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

merge

List pae_pgdir_above_4gb in features for x86/32 Linux kernel.

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

merge.

Support XENFEAT_supervisor_mode_kernel in x86/64 HYPERVISOR_IRET macro

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

Only panic() if the dom0 kernel requires an unknown feature flag.

It is acceptable for a guest kernel to support (but not require)
a feature which the hypervisor does not understand.

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

Merged.

Fix non-xen build for: Make the fixmap area moveable in kernel address space.

From: Gerd Hoffmann <kraxel@suse.de>
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Add arch/i386/mm/pgtable.c to sparse tree.

Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Fix non-xen build for: Move the gate page (vsyscall) out of the fixmap area
into user address space, just below PAGE_OFFSET.

From: Gerd Hoffmann <kraxel@suse.de>
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Add include/asm-i386/{fixmap,page}.h to sparse tree.

Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Define strlcpy() in include/xen/string.h

Signed-off-by: Tsunehisa Doi <doi.tsunehisa@jp.fujitsu.com>
Signed-off-by: Kouya SHIMURA <kouya@jp.fujitsu.com>
Signed-off-by: Masaki Kanno <kanno.masaki@jp.fujitsu.com>
Signed-off-by: Akio Takebe <takebe_akio@jp.fujitsu.com>

Periodically sytnchronize Xen's wallclock time with NTP-synchronized time in domain0.

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

The disk isn't guaranteed to be the first device in the config sxp --
make sure we can restart domains when using a boot loader even if not.

Signed-off-by: Jeremy Katz <katzj@redhat.com>

pygrub currently takes a file on the command line to mean a config file
if it can't figure out partitions and filesystems.  This then makes it
balloon to absurd sizes.  Make this only happen if you pass a debugging
flag

Signed-off-by: Jeremy Katz <katzj@redhat.com>

Instead of using a static filename in the guest bootloader, use a random
filename so that multiple guests can start up simultaneously.

Signed-off-by: Jeremy Katz <katzj@redhat.com>

Add 'console=xxx' to the Xen serial console section to make it obvious
that one might need to specify 'console=xxx' and not just 'comX=xxx'
if one isn't using the default com1,vga.

Signed-Off-By: Muli Ben-Yehuda <mulix@mulix.org>

Cross-compilation fixes (x86/64 on i386).

Use strtoull to parse the final_exec_addr value from the command
line. Whe cross compiling on i386 for x86-64 unsigned long is 32 bit,
which leads to final_exec_addr being truncated to 32 bits by strtoul,
which leads to grub refusing to load the resulting image. Please
apply.

Use the target nm(1) when cross compiling to pick the _end address
from the Xen binary.

Signed-Off-By: Muli Ben-Yehuda <mulix@mulix.org>

Fix assertion.

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

New VCPUOP_register_runstate_memory_area hypercall. Avoids
need for a hypercall in the guest timer interrupt handler.

Cleaned up stolen/blocked tick handling in Linux.

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