Cross-compilation fixes.
Signed-off-by: Michal Ostrowski <mostrows@watson.ibm.com>

Make mbootpack work on big-endian systems.
Signed-off-by: Michal Ostrowski <mostrows@watson.ibm.com>

Replace dom_mem_op hypercall with memory_op hypercall with
better interface. Fix "mem=" parameter on x86/64 boot.

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

This patch export symbols in xenbus that is used by netfront
and blkfront so that vbd and vnif can be made to modules.

Signed-off-by: Xiaofeng Ling <xiaofeng.ling@intel.com>

Add CONFIG_BLK_DEV_UB as unconfigured item to x86/64 dom0
defconfig file. Spotted by Yongkang You at Intel.

Fix save/restore (for SMP guests in particularl);  the blkfront suspend code
was failing to unbind its irq handler correctly which (on SMP) caused the
guest to BUG() on resume when trying to request its callfunc_irq. Plus a
few small cosmetic changes.

Signed-off-by: Steven Hand <steven@xensource.com>
/bin/bash: :x: command not found

Simplify and fix SMP irq and ipi teardown/setup on suspend/resume.
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Update suspend/resume functions for xenbus.
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Only hotplug cpus and save/restore vcpu context on SMP builds.
The for_each_cpu_mask macro has a slightly odd behaviour in the
non-CONFIG_SMP case, making the code unexpectedly incorrect.
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Fix page_is_ram() start-of-day checks, and the watch routine
inside balloon driver.

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

Fix "bogus" check by moving it inside the length check.
Signed-off-by: Ryan Harper <ryanh@us.ibm.com>
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Allocate a request-id space per vbd in blkfront.
This should fix the assertions people have been seeing
in domU blkfront where we run out of IDs (because they
were being shared by multiple vbds).

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

Minor (partial) build fix for 2.4.

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

Fix nasty tools race between barking xu_autoreap() and xpopen3.wait() -
save/restore now seems more robust from a tools pov at least.

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

blkfront cleanup and retab.
Signed-off-by: Keir Fraser <keir@xensource.com>

CONFIG_X86_PM_TIMER is a no-op option when building Linux
for Xen, but make sure it doesn't break the build.

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

Don't build pcbios.c for xen/i386. No good can come of it.
Signed-off-by: Keir Fraser <keir@xensource.com>

Refactor sync_lazy_execstate_cpu() into the more sensible
sync_vcpu_execstate().

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

Clean up and fix hypercall macros in XenLinux: we must assume that
all register parameters are clobbered on return. Furthermore,
multicall arrays are clobbered on return from a multicall (only the
result field can be trusted!).

These restrictions are enforced by debug builds of Xen in the
following ways:
 1. On completion of a multicall, the call array is overwritten
    with garbage before writing in the result fields.
 2. On completion of a hypercall, all parameter registers are
    overwritten with garbage.

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

Remove the unused ia64 patch directory.
Signed-off-by: Keir Fraser <keir@xensource.com>

The patched attached enables x86_64 xenlinux with "late pin, early
unpin", which is already implemented for x86_32. Since we now only pin
the root rather than any of the other levels, the overall performance
became better especially with workloads that require heavy memory
management operations.

On 8-way x86_64 xenlinux (dom0) the kernel build was improved by about
10% (using make -j32). Even a small setup like a UP HT system, I see
about 3% performance gain with kernel build (make -j4).

Lmbench also shows improvements in fork/exec/sh:
Processor, Processes - times in microseconds - smaller is better
--------------------------------------------------------------------
Host    OS  Mhz   null null      open slct sig  sig  fork exec sh =20
                  call  I/O stat clos TCP  inst hndl proc proc proc
--------- ------------- ---- ---- ---- ---- ---- ---- ---- ---- ----=20
Linux 2.6.12- 3786 1.13 1.36 3.93 6.04 10.5 1.43 4.33 536. 1446 3614
Linux 2.6.12- 3786 1.13 1.36 3.91 6.03 10.4 1.44 4.38 346. 1050 2831

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

Fix GDT allocation for SMP guests.

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

Vnet update.
Fix the tools/vnet makefiles so that vnets will compile
and includes updates to the vnet implementation.
Vnet ids are increased to 128 bits. The vnet module will
insmod whether xen-br0 exists or not.
Signed-off-by: Mike Wray <mike.wray@hp.com>
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Fix vnets in xm.
The removal of 'xm call' to call a xend interface function directly
made the xend vnet functions inaccessible. This patch adds those functions
to xm and fixes the persistence in XmVnet to use xenstore.
Signed-off-by: Mike Wray <mike.wray@hp.com>
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

The patch removes broken, and very complicated malloc in
favour of much simpler (and working) Xen's allocator
(xmalloc by Rusty).

Signed-off-by: Grzegorz Milos <gm281@cam.ac.uk>

Clean up and fix domain_pause and friends. Synchronous
pause should not only wait for the running flag to clear,
but also for the scheduler lock to be released. Also get
rid of some unused sync_lazy_execstate functions.

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

merge?

Update xentop to use new name for Xen version function.
Signed-off-by: Josh Triplett <josht@us.ibm.com>
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Changed xentop.c to include version in the summary.
Signed-off-by: Judy Fischbach <jfisch@us.ibm.com>
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Bugfix: xenstat_collect_networks now returns the correct error code.
Signed-off-by: Josh Triplett <josht@us.ibm.com>
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Cache the Xen version in the xenstat_handle, since it should never change.
Use a collector and new flag for Xen version information, to avoid the
extra hypercall if the caller does not want the information.
Reference the xenstat_handle in the xenstat_node for ease of access
later, and update collectors accordingly.
Signed-off-by: Josh Triplett <josht@us.ibm.com>
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Formatting fixes for xenstat and xentop.
Signed-off-by: Josh Triplett <josht@us.ibm.com>
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Add file missed in c51562192ef6b1d94f6de7aba1daafe8353064a8 changeset.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Merge with -unstable

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

merge?

Add hotplug script to run vif-bridge up script when a backend vif starts up.
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Fix MPT mapping for x86/64. Also i386 no longer needs
4MB mapping for PAE, as xenlinux can handle faults when
accessing unmapped sections of the table.

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

merge?

merge?

Memory management fixes. Page tables are created, buddy allocator now
has ability to free memory (as well as allocate it).

Signed-off-by: Grzegorz Milos <gm281@cam.ac.uk>

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

This patch fixes Bugzilla #169. (It should fix #187 too -- Keir).

The root cause of bug 169 is, machine_to_phys_mapping, starting from
0xffff800000000000, is mapped using 2M pages.  When the system has RAM
no more than 2G, only one 2M page is allocated and only one PDE entry is
created correspondingly, so calling mfn_to_pfn with mfn > 0x80000 will
overflow this 2M page and cause a unable handled kernel paging request.
The mfn > 0x80000 comes from PCI device I/O memory, here from AGP
display card when booting X server.  Jun suggested to use something like
get_user() when accessing machine_to_phys_mapping.

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

The mpt entry is 4 bytes even on x86_64, so we use 4 when
calculating mpt size.

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

1. Remove workaround patch
2. Update scripts/vif-bridge to not call 'brctl delif'

When a domU is shutdown/destroyed and the netif is destroyed, the
notify_call_chain triggered from unregister_netdevice() will trigger the
bridge event handler and which will call the proper code to remove the
device from the bridge.

I can't see any reason why brtcl delif should be called when taking out
a domain if the call chain will delete the interface from the bridge
when the vif is destroyed automatically.

Signed-off-by: Ryan Harper <ryanh@us.ibm.com>

Remove broken patch preventing net device unregister.
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Cleanup vif entries in store when domain goes away.
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

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

Store correct rx ring handle in netback driver.
Signed-off-by: Keir Fraser <keir@xensource.com>

merge?

g/c Linux 2.4 netback support.
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

More cleanups and g/c of unused code.
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Fix grant-ref allocation in netfront.

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

merge?

g/c unused code.
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Fix SMP booting: x86/64 startup initialisation fixes and so on.
Signed-off-by: Keir Fraser <keir@xensource.com>

Fix install of xenctrl.h.

merge?

Fix the exception thrown when doing, for example, "xm domid" on a non-existent domain.
Signed-off-by: Dan Smith <danms@us.ibm.com>
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

merge?

Updated to reflect new tools added here.
Signed-off-by: Judy Fischbach <jfisch@us.ibm.com>
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Ensure explicit asm statememt goes in the correct section.

merge?

Break the building/save/restore code out into a separate library libxenguest.
Also update the tools accordingly.
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Rename libxc => libxenctrl and xc.h => xen/xenctrl.h
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>
--HG--
rename : tools/libxc/xc.h => tools/libxc/xenctrl.h

merge?

Switch network setup over to xenbus.
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

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

Rename grant-id to ring-ref.
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Watch may fire before we've allocated the interface structure.
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Unregister watch in error path.
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Avoid calling device_unregister from device_register callback.
Also simplify code.
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Disable debug output again.
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Print status of Flush Filter on AMD64 processors during
Xen boot. Also provide a new boot parameter:
 amd_flush_filter=on  : Forcibly enable the Filter
 amd_flush_filter=off : Forcibly disable the Filter

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

merge?

Always free the lowest 1MB of pseudo-physical memory when booting
a guest.

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

Revert changeset a3822cb3b2b3c24e6b46c1fed3e6c366817f52ad.
With this changeset the error message is incorrect.
Revert: Fix the exception thrown when doing, for example,
        "xm domid" on a non-existent domain.
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Fix x86/64 pagetable initialisation so that only things that
need to be are mapped read-only.

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

Add EXT3 as built-in filesystem for x86/64 domU.

Copy correct number of bytes for xen_start_info in x86_64
head.S.

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

Initialise syscall32 vsyscall page early, as it may be needed
for some initcalls (e.g., if initrd /bin/hotplug is an i386
binary).

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

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

Don't use the grant pre-allocation system for the ring grants.
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Add some additional error nodes when block device setup fails.
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Fix the exception thrown when doing, for example, "xm domid" on a non-existent domain.
Signed-off-by: Dan Smith <danms@us.ibm.com>
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Ensure percpu data area not used before the TR is set.

Xenstore testsuite robustness: save output rather than rerun on failure.
"make check" reruns a test which fails with more verbosity.  If the test
fails intermittently, that doesn't work well: save the output and simply
dump it if the test fails.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Fix xenstore testsuite for valgrind 3.0 and latest Debian glibc.
Debian upgrade revealed two problems: glibc spits out a warning on every
binary, and valgrind 3.0 gratuitously changed --logfile-fd to --log-fd,
so don't use it.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Make directories implicitly for mkdir and write.
Change directory code: make then move (safer than delete-if-fail).
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

g/c extra test.
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Put the call to notifier_call_chain() back in do_xenbus_probe().
This fixes some dom0 activities, such as ballooning, sysrqing,
and cpu hotplugging.
Signed-off-by: Dan Smith <danms@us.ibm.com>
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

merge?

xenstat/xentop merge.

merge?

g/c vbd active.
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

Fix non-hotplug i386 build. Spotted by Rik van Riel.

Remove blkback 2.4 support since we don't support 2.4 dom0 anymore.
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>

merge?

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

Build topology.c into x86/64 xenlinux kernel. This omission
spotted by Ryan Harper at IBM.