Note about when device_find is fixed
Signed-off-by: Rusty Russel <rusty@rustcorp.com.au>
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Modify to assume transactions
Signed-off-by: Rusty Russel <rusty@rustcorp.com.au>
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Do not use device_find: crashes for some reason
Signed-off-by: Rusty Russel <rusty@rustcorp.com.au>
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Fix missing argument rename in xenbus_exists.
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Grab lock around registering watch.
Signed-off-by: Rusty Russel <rusty@rustcorp.com.au>
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Add watch for dynamic add/remove of devices.
Signed-off-by: Rusty Russel <rusty@rustcorp.com.au>
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Fix return from xenbus_scanf: returns number scanned, not 0!
Fix xenbus_write args in xenbus_printf
Signed-off-by: Rusty Russel <rusty@rustcorp.com.au>
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Major xenbus cleanup
1) Make xenbus_* functions first-class citizens, evict xs_ prefix.
2) Add xenbus_scanf and xenbus_printf instead of multiple functions
3) Remove unused code from xenbus.c
4) Use standard debugging macros.
Signed-off-by: Rusty Russel <rusty@rustcorp.com.au>
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Update python bindings - xs_watch doesn't support a priority anymore.
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Remove ill-conceived concept of watches blocking reply on
connection which did write/mkdir/rm/setperm etc.
This causes deadlocks in real life, and I can't see a sane way
of avoiding them: it is reasonable for someone to ignore watch
notifications while doing other actions, and that means that
we can do other writes.  These writes can block pending other
watchers; if one of these is the process blocked awaiting our
ack, we deadlock.
Signed-off-by: Rusty Russel <rusty@rustcorp.com.au>
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Fix bug where transaction fires multiple events.
Add general trace() call to help future debugging.
Signed-off-by: Rusty Russel <rusty@rustcorp.com.au>
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Fix bug where watch messages doesn't go out.  Also change xs_test
to more closely represent xenbus behaviour so we can see the bug.
Signed-off-by: Rusty Russel <rusty@rustcorp.com.au>
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Allow test.sh to take arg: name of test to start with
Signed-off-by: Rusty Russel <rusty@rustcorp.com.au>
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Make testing more reliable: wait for event after async so we know
child has done command
Signed-off-by: Rusty Russel <rusty@rustcorp.com.au>
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Change watches: operations block until everyone has acked.
Watch events are no longer sent to self
Watches no longer take a priority
async and asyncwait commands for xs_test, now we need to continue
despite blocking ops.
Print test name at end of verbose run on failure.
Use --trace-file arg to xenstored when testing
Signed-off-by: Rusty Russel <rusty@rustcorp.com.au>
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Change from bool indicating blocked to an enum: when watches go
synchronous this makes it easier (because then we have two reasons
to block)
Instead of using return value, use explicit state member inside
struct connection.
Signed-off-by: Rusty Russel <rusty@rustcorp.com.au>
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Test and fix acknowedge_watch from returning EINVAL.
Also ensure that daemon re-xmits event if they ack wrong thing
(otherwise confused clients get stuck, as we will send no more
data while awaiting ack).
Signed-off-by: Rusty Russel <rusty@rustcorp.com.au>
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Fix testsuite to ignore tool/ dir.
Signed-off-by: Rusty Russel <rusty@rustcorp.com.au>
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Start of code to persistent store connections when xenstored restarts
Signed-off-by: Rusty Russel <rusty@rustcorp.com.au>
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

# HG changeset patch
# User Rusty Russell <rusty@rustcorp.com.au>
# Node ID a92163adedcfcff0d05c965d09da747f3c8aa13e
# Parent  63ab20781afa311300f3a8e832744292014ea7f6

Remove ill-conceived concept of watches blocking reply on connection which did write/mkdir/rm/setperm etc.
This causes deadlocks in real life, and I can't see a sane way of avoiding them: it is reasonable for someone to ignore watch notifications while doing other actions, and that means that we can do other writes.  These writes can block pending other watchers; if one of these is the process blocked awaiting our ack, we deadlock.

diff -r 63ab20781afa -r a92163adedcf tools/xenstore/xenstored_core.c

Fix restart/poweroff properly. From aq.

- allows to optionally boot the system with a policy already being
  active at startup; this works by adding a module line into the grub
  configuration file and placing the binary policy generated by the
  policy tool into the boot directory; This assumes that a maximum of
  one module line is used for the initrd in the grub configuration
  file - Question: do users pass more than one module to the kernel?
- enables the policy hypervisor call on x86/64
- some function prototypes moved to include files
- moves the version number in the java tool up to the current
  version (a better way of doing this will be submitted soon)

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

Improve guest time keeping

Even though we accounted for lost ticks, the accounting was not precise.
Specifically, we didn't account for the delay in calling the ac_timer
handler.

Signed-off-by: Eddie Dong <eddie.dong@intel.com>
Signed-off-by: Arun Sharma <arun.sharma@intel.com>

Really enable gdbserver for real mode code.

Missed out one instance where we need to check if we're running on
1:1 page tables with guest paging disabled.

Signed-off-by: Arun Sharma <arun.sharma@intel.com>

Use hg root instead of bk root.

Signed-off-by: Arun Sharma <arun.sharma@intel.com>

Restore command line parsing code from qemu-0.6.1

This enables cdrom ejection (both file based and physical i.e. /dev/cdrom)

Signed-off-by: Edwin Zhai <edwin.zhai@intel.com>
Signed-off-by: Arun Sharma <arun.sharma@intel.com>

Allow multiple disk images per guest.

Signed-off-by: Yunfeng Zhao <yunfeng.zhao@intel.com>
Signed-off-by: Arun Sharma <arun.sharma@intel.com>

Fix the path to qemu-dm

Implement the I/O bitmap vm execution control.

With this patch, we should not bee seeing any vmexits for debug port
accesses.

Signed-off-by: Edwin Zhai <edwin.zhai@intel.com>
Signed-off-by: Arun Sharma <arun.sharma@intel.com>

Don't leave behind zombie device model processes

Signed-off-by: Edwin Zhai <edwin.zhai@intel.com>
Signed-off-by: Arun Sharma <arun.sharma@intel.com>

Xen full virtualization does not handle unaligned IO with page crossing.
Disable 32-bit PIO as a workaround.

Signed-off-by: Asit Mallick <asit.k.mallick@intel.com>
Signed-off-by: Arun Sharma <arun.sharma@intel.com>

Translate some control characters in the qemu monitor.

This removes odd control characters displayed in the qemu monitor with vnc

Signed-off-by: Edwin Zhain <edwin.zhai@intel.com>
Signed-off-by: Arun Sharma <arun.sharma@intel.com>

Remove extra definition of DBG in pci/irq.c in XenLinux.

Fix cross-compile builds by using NM macro instead of nm
binary directly.

XenLinux restart/poweroff calls 'reboot' or 'poweroff'
as appropriate.
Signed-off-by: Keir Fraser <keir@xensource.com>

split deallocation from net_tx_action

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

add grant table capabilities to netback and netfront.

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

regenerate smp-alt.patch for 2.6.12

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

Local merge
Stephan.Diestelhorst@{cl.cam.ac.uk, inf.tu-dresden.de}

Port kallsyms to Xen, as 'symbols'.
Signed-off-by: Keir Fraser <keir@xensource.com>

Fixed extratime scheduling bug, which caused a prioritised dom0 to overtake the whole cpu.
Signed-off by: Stephan.Diestelhorst@{cl.cam.ac.uk, inf.tu-dresden.de}

In 2.6.12-land, do_debug and do_int3 return void.  Make sure we don't try
to use their return values anywhere.

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

Here is a patch to enable Xen to run on a Unisys ES7000 x86_64 system.
Signed-off-by: Aravindh Puthiyaparambil <aravindh.puthiyaparambil@unisys.com>

Fix problem of can not create two or more vmx guest.

The original gdb server patch will cause the new vmx guest break the=20
old one. This patch make the arch_set_guest_info modify the correct
vmcs and also skip modifying during  creating.

Signed-off-by: Xiaofeng Ling <xiaofeng.ling@intel.com>^
Signed-off-by: Arun Sharma <arun.sharma@intel.com>

Save per cpu IDT information into the VMCS host area.

Without this, we may be saving the IDT info from the wrong cpu into the
vmcs.
This makes domU work in the presence of VMX domains on SMP/HT enabled
machines.

Signed-off-by: Chengyuan Li <chengyuan.li@intel.com>
Signed-off-by: Xiaofeng Ling <xiaofeng.ling@intel.com>=20
Signed-off-by: Arun Sharma <arun.sharma@intel.com>

Propagate guest MSR writes to machine MSRs immediately

Right now, we have an exposure between the time the MSR is written
and used by an instruction such as syscall. If there is a context
switch and we do vmx_do_restore_msrs(), everything goes fine. But
if we don't, then we execute the syscall with the wrong MSR.

Signed-off-by: Yunhong Jiang <yunhong.jiang@intel.com>
Signed-off-by: Arun Sharma <arun.sharma@intel.com>

Fix bug.

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

No need to define io_remap_pfn_range in linux/mm.h.

Merged by hand.

Shadow mode's writable-PTs should only allow guest kernels access to PTs

Fix wallclock timebase extraction in XenLinux.
Signed-off-by: Keir Fraser <keir@xensource.com>

Simplify and fix the error-correction factor during
local timer calibration.
Signed-off-by: Keir Fraser <keir@xensource.com>

First cut of new time interfaces and synchronisation mechanisms.
Based on an initial patch from Don Fry at IBM.
Still TODO:
 1. Testing
 2. NTP synchronisation
 3. Fix wallclock interface a bit
 4. Support for platform timers other than PIT (e.g., HPET, IBM Cyclone)
 5. Scale 64-bit TSC diffs instead of 32-bit, just for sanity
 6. Error-correcting scale factor is still slightly wrong
 6. More testing
Signed-off-by: Keir Fraser <keir@xensource.com>

pdb: read/write registers for process target

Fix 'make uninstall' target.

Fix 'make mrproper' to remove pristine and ref trees.

Delete resurrected files (from ia64 merge).

Simple grant-table fix (check_unmap).
By Chris Clark and Stefan Berger.

Trivial fixes for numeric arg to head/tail.

Merge with ia64-unstable.hg

Relax the bounce-buffer copying constraints in dma_sync_single_for_*
functions. This fixes certain drivers (e.g., ivtv).
Signed-off-by: Keir Fraser <keir@xensource.com>

Avoid some unnecessary TLB flushes.  This will probably make no real
difference on any sensible guest operating system.

Signed-off-by: Steven Smith, sos22@cl.cam.ac.uk.

Fix exit path from DOM0_CREATEDOMAIN, which I broke earlier
today.

Enable VMX domains on a SMP dom0.

Not all vectors go through do_IRQ(). It's possible to look up the handler
in the IDT and call that one, but it'll result in two trap frames.

Signed-off-by: Xiaofeng Ling <xiaofeng.ling@intel.com>
Signed-off-by: Arun Sharma <arun.sharma@intel.com>

Recently some people complain that they cannot set the maxmem for
domU. The problem is that some code wrongly treat dom.max_memkb as in
byte unit, instead of KB unit. This patch fixs the problem.

Signed-off-by: Nguyen Anh Quynh <aquynh@gmail.com>

Temporary fix for netif ring overflow, until we move
netfront/netback to grant tables.

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

Update x86_64 -xen config for Linux 2.6.12
Signed-off-by: ian@xensource.com

Update the -xen default config to Linux 2.6.12

Signed-off-by: ian@xensource.com

Incorporate lomount from QEMU into tools/misc. This tool makes it easy to mount
individual partitions held within whole disk images.

Original source from http://www.dad-answers.com/qemu/utilities/QEMU-HD-Mounter/l
omount/lomount.c

 * Copyright (c) 2004 Jim Brown
 * Copyright (c) 2004 Brad Watson
 * Copyright (c) 2004 Mulyadi Santosa

Signed-off-by: ian@xensource.com

Fix for new ns16550 serial code in xen-unstable.hg

Missing quirks.c file.

Auto-disable IRQ balancing/affinity on buggy chipsets.
Signed-off-by: Keir Fraser <keir@xensource.com>

Allow IRQ balancing to be disabled with 'noirqbalance'
boot parameter.
Signed-off-by: Keir Fraser <keir@xensource.com>

Add debug-key for printing IRQ/IOAPIC info.
Signed-off-by: Keir Fraser <keir@xensource.com>

Misc fixes for new sparse tree names: xen-clone script; -xen kernel build (the one with all the modules); Backout changes to legacy targets.

Signed-off-by: ian.pratt@cl.cam.ac.uk

Allow arch-specific defaults to be specified for ns16550
uart configuration. Based on a patch from Hollis Blanchard
at IBM.
Signed-off-by: Keir Fraser <keir@xensource.com>

The VMCS control bits in the current tree are not optimal.  Attached is
a patch to improve the VMCS control bits setting. It also adds checks to
detect any non-optimal or incompatible settings for the VMCS control
bits based on MSRs, and fixes a bug associated with vmcs region freeing.

Signed-off-by: Nitin A Kamble <nitin.a.kamble@intel.com>
Signed-off-by: Jun Nakajima <jun.nakajima@intel.com>
Signed-off-by: Xin Li <xin.b.li@intel.com>

Cleanup mov to CR4 handling.

Signed-off-by: Arun Sharma <arun.sharma@intel.com>

Enable CR4.VME to improve the performance of vm86 emulation.

Setting this bit eliminates the calls to the vm86 monitor to handle
interrupts and exceptions that occur while executing programs in vm86
mode.

Signed-off-by: Asit Mallick <asit.k.mallick@intel.com>
Signed-off-by: Arun Sharma <arun.sharma@intel.com>

repair legacy Makefile targets linux26 and linux24

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

Oops remove debug output from last changeset

Fix PAL_PTCE_INFO and cpuid (pfmon now works)

pdb: reconcile register manipulation between processes and domains

Initial tools support for 32-bit x86 pae.
Signed-off-by: Gerd Knorr <kraxel@suse.de>

Force PAE build of Xen by specifying 'pae=y' to make, or
by setting XEN_TARGET_X86_PAE=y (e.g., in Config.mk).
Signed-off-by: Keir Fraser <keir@xensource.com>

Remove CONFIG_X86_PAE hack from config.h.
Signed-off-by: Keir Fraser <keir@xensource.com>

Readd pristine-% rules that went MIA.

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

Add easy way to turn off all fast paths to rule them out when debugging

Enable NX/XD feature for 32-bit PAE Xen.
Signed-off-by: Keir Fraser <keir@xensource.com>

Build hotplug by default in dom0.

The patch attached restores NX/XD support in x86_64 xenlinux. When the
port was being done, Xen did not want to see the bit 63 in ptes, and it
now allows the guests to set it. Also include cleanups (warning when
compiling domU).

Signed-off-by: Jun Nakajima <jun.nakajima@intel.com>

Need to save and restore MSRs for VMX domains across context switches.

Signed-off-by: Yunhong Jiang <yunhong.jiang@intel.com>
Signed-off-by: Jun Nakajima <jun.nakajima@intel.com>
Signed-off-by: Arun Sharma <arun.sharma@intel.com>

Fix NX/XD enable on secondary CPUs.
Signed-off-by: Keir Fraser <keir@xensource.com>

pdb: get rid of absolute path in module makefile

Let Xen have 256 PIRQs to guests.

pdb: execution_domain -> vcpu

Native x86_64 linux uses paranoidentry for exceptions with ist !=3D 0,
while x86_64 xenlinux doesn't have special stacks for these exceptions.
Actually we don't need paranoidentry for x86_64 xenlinux, and these
exceptions except #DB should be handled in hypervisor.  With this patch,
#DB is forwarded to guest, and let it use zeroentry should be OK.
-Xin

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

Small fix for PAE in 2.6.12.
Signed-off-by: Gerd Knorr <kraxel@suse.de>

Some more new files for Linux PAE.
Signed-off-by: Gerd Knorr <kraxel@suse.de>