- Move PAGE_SIZE and STACK_SIZE into __PAGE_SIZE and __STACK_SIZE in
arch_limits.h so as to permit getting them from there without
pulling all the internal Mini-OS defines.
- Setup a xen-elf cross-compilation environment in stubdom/cross-root
- Add a POSIX layer on top of Mini-OS by linking against the newlib C
library and lwIP, and implementing the Unixish part in mini-os/lib/sys.c
- Cross-compile zlib and libpci too.
- Add an xs.h-compatible layer on top of Mini-OS' xenbus.
- Cross-compile libxc with an additional xc_minios.c and a few things
disabled.
- Cross-compile ioemu with an additional block-vbd, but without sound,
tpm and other details. A few hacks are needed:
- Align ide and scsi buffers at least on sector size to permit
direct transmission to the block backend. While we are at it, just
page-align it to possibly save a segment. Also, limit the scsi
buffer size because of limitations of the block paravirtualization
protocol.
- Allocate big tables dynamically rather that letting them go to
bss: when Mini-OS gets installed in memory, bss is not lazily
allocated, and doing so during Mini-OS is unnecessarily trick while
we can simply use malloc.
- Had to change the Mini-OS compilation somehow, so as to export
Mini-OS compilation flags to the Makefiles of libxc and ioemu.
Signed-off-by: Samuel Thibault <samuel.thibault@eu.citrix.com>
mb/rmb/wmb. Instead add a xen_ prefix. Modify Xen's public headers to
expect the prefixed names instead of bare mb/rmb/wmb, but gate this
expectation on a bump of __XEN_INTERFACE_VERSION__.
Signed-off-by: Keir Fraser <keir.fraser@citrix.com>
Make qemu unblock SIGHUP and make sure the default handler is in
place. Have the domain killer send SIGHUP to the device-model script,
allow the script 10s to clean up, and if still not dead, send
SIGKILL.
Signed-off-by: Samuel Thibault <samuel.thibault@eu.citrix.com>
Currently, a wrong value of maximum memory size is checked by
_safe_set_memory() and _memory_sanity_check().
Signed-off-by: Masaki Kanno <kanno.masaki@jp.fujitsu.com>
The qemu vnc server changes its internal colour depth based on the
client request. This way just one colour conversion is done: the one
in vga_template.h, from the guest colour depth and the vnc server
internal colour depth.
This patch is meant to remove this colour conversion to improve
performances. It accomplishes the goal making the qemu internal colour
depth always the same as the guest colour depth.
The basic idea is that the vnc client is the one that should do the
colour conversion, if necessary. In general it should accept the pixel
format suggested by the server during the initial negotiation. This
behaviour can be set in most vnc clients (vncviewer included).
If the guest changes colour depth, the qemu vnc server changes colour
depth too and notifies the client. The problem is that the vnc
protocol doesn't provide a message from the server to the client to
ask for a colour depth change. So what I am doing is either:
1) quietly starting to do the conversion on vnc server (not gaining
any performance here);
2) closing the vnc connection with the client, so the client can
reconnect and choose the new pixel format.
By default I am doing 1), however the second choice can be enabled
passing the -vnc-switch-bpp command line option.
In order to do the colour conversion on the vnc server I had to
improve the colour conversion code already in place because it only
supported conversions from 32 bpp. The patch adds colour conversion
code that support conversions from any resolution to any resolution.
A last note: to get most out of this patch it is best to set Windows
to 16 bit colour depth, because the 24 bit mode is 24 bit depth and 24
bpp, meaning no alpha channel. The vnc protocol doesn't support 24
bpp, only 32 bpp, so this conversion is unavoidable.
From: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Signed-off-by: Keir Fraser <keir.fraser@citrix.com>
A new platform variable 'hpet' is added, which defaults to 0 for new
guests (that is, hpet disabled). Default is off (no hpet) because
hpet is currently less accurate in keeping time than PIT (because no
timer_mode adjustments).
Signed-off-by: Dan Magenheimer <dan.magenheimer@oracle.com>
Signed-off-by: Keir Fraser <keir.fraser@citrix.com>
via the 'kernel' config option:
1. Look for any string containing 'hvmloader'.
2. The 'kernel' option must be scrubbed to avoid taking
PV-kernel-loading paths during later guest setup.
Signed-off-by: Keir Fraser <keir.fraser@citrix.com>
When setting
vfb = [ 'type=sdl' ]
in a domain config file, qemu seems to be still using vnc. Make it use
sdl as expected.
Signed-off-by: Samuel Thibault <samuel.thibault@eu.citrix.com>
qemu doesn't enqueue mouse events, just records the latest mouse
state. This can cause some lost mouse double clicks if the events are
not processed fast enought. This patch implements a simple queue for
left mouse click events.
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
user-specified polling interval in order to determine when to empty
the trace buffers. Removed the old and unused/unimplemented
new_data_threshold logic.
Signed-off-by: Michael A Fetterman <Michael.Fetterman@cl.cam.ac.uk>
It was previously assert'ing when it hit 4G.
Also, because the trace buffer is not a power of 2 in size,
using modulo arithmetic to address the buffer does not work
when the index wraps around 2^32.
This patch fixes both issues, and as a side effect, removes all
integer division from the hypervisor side of the trace mechanism.
Signed-off-by: Michael A Fetterman <Michael.Fetterman@cl.cam.ac.uk>
Xen-API already specifies these actions for the 'on_crash' domain exit
event. This patch makes them available for use in traditional domU
config files and through the xm tool as well.
Signed-off-by: Jim Fehlig <jfehlig@novell.com>
When I rebooted a domain by xm reboot command with wait option,
I saw the following message. But, rebooting the domain succeeded.
Domain vm1 destroyed for failed in rebooting
The cause why the message was shown is the domain is destroyed
temporarily by processing of xm reboot command. The domain
information is not gotten from Xend by server.xend.domains()
function till recreating the domain is completed.
This patch fixes processing of xm reboot command in Xm side.
It waits just a bit till recreating the domain is completed,
then it measures the success or failure of the reboot of the
domain.
Signed-off-by: Masaki Kanno <kanno.masaki@jp.fujitsu.com>
Here is what I wrote in my submission to qemu upstream:
qemu's audio subdirectory contains a copy of BSD's sys-queue.h, which
defines a bunch of LIST_ macros. This makes it difficult to build a
program made partly out of qemu and partly out of the Linux
kernel[1], since Linux has a different set of LIST_ macros. It might
also cause trouble when mixing with BSD-derived code.
Under the circumstances it's probably best to rename the versions in
qemu. The attached patch does this.
[1] You might well ask why anyone would want to do this. In Xen we
are moving our emulation of IO devices from processes which run on
the host into a dedicated VM (one per actual VM) which we call a
`stub domain'. This dedicated VM runs a very cut-down `operating
system' which uses some code from Linux.
Signed-off-by: Ian Jackson <ian.jackson@eu.citrix.com>
Here is what I wrote in my submission to qemu upstream:
The patch below makes it possible to disable AF_UNIX (unix-domain)
sockets in host environments which do not define _WIN32, by adding
-DNO_UNIX_SOCKETS to the compiler flags. This is useful in the
effectively-embedded qemu host which are going to be using for device
emulation in Xen.
Signed-off-by: Ian Jackson <ian.jackson@eu.citrix.com>
When guest domains are restarted, previous values of /vm/uuid/xend/*
in xenstore are lost. (e.g. previous_restart_time,
last_shutdown_reason). This patch restores them to restarting domains.
And we should update /vm/uuid/xend/restart_count of restarting
domains, not previous domains.
Signed-off-by: Masaki Kanno <kanno.masaki@jp.fujitsu.com>
The stub domain device model needs to trap the termination signal so
as to actually destroy the stub domain. SIGKILL can't be trapped,
SIGTERM is caught by SDL and so may be unnoticed. SIGHUP can be
trapped and is not caught by SDL (and by default causes a process
termination without core).
Signed-off-by: Samuel Thibault <samuel.thibault@eu.citrix.com>
Continue to SIGKILL qemu-dm so that 'xm destroy' should always work
rather than hang if qemu-dm is in a bad state.
Signed-off-by: Keir Fraser <keir.fraser@citrix.com>
tools/firmware/etherboot by default (the dependency of eb_roms.h on
Config does not seem to always prevent the build from going ahead, for
some reason).
Signed-off-by: Keir Fraser <keir.fraser@citrix.com>
Xen recently (changeset 16873:86c32269ba60) started looking for a
DHCP defaults file in /etc/sysconfig/dhcpd or /etc/defaults/dhcp.
Debian (lenny) puts this file in a different place though:
/etc/default/dhcp3-server. I added this location to Xen's search
path.
From: Michael Abd-El-Malek <mabdelmalek@cmu.edu>
Signed-off-by: Keir Fraser <keir.fraser@citrix.com>
The save failure cleanup introduced in 13543:207523704fb1 is
incorrect: if we didn't get as far as actually suspending the domain,
then the guest domain will not be expecting the devices to be removed
(seen on both Linux and Solaris, which don't expect a 'Closing' state
when they hold the device open). Only re-jig devices if we definitely
shut the domain down.
Signed-off-by: John Levon <john.levon@sun.com>
When saving a guest in early ChangeSet (Before 16750), Xend will
send "SIGKILL" signal to kill qemu process. But in ChangeSet 16750,
the signal is modified to "SIGTERM". However, "SIGTERM" handler has
been registered by SDL in qemu. So when qemu process receives a
"SIGTERM", SDL will catch it and it will push out a "SDL_QUIT" even to
qemu. If qemu can receive that event, it will be successfully
terminated. While we do "xm save", qemu can't go out from the dead
loop "while (suspend_requested) {...}" in main_loop() to handle the
SQL_QUIT event. So it is unable to poll that "SDL_QUIT" event. This
makes xm save command hang. This fix enables qemu to poll the
"SDL_QUIT" event while doing save.
Signed-off-by: Xu Dongxiao <dongxiao.xu@intel.com>
This patch provides the tools support for direct kernel boot of
HVM guests. Currently the config files in /etc/xen support the
args 'kernel', 'ramdisk' and 'extra'. For PV guests these have
the obvious meaning.
Unfortunately HVM guest configs hijacked the 'kernel' parameter
and use it to refer to the path of the HVM firmware. So, this
patch adds a new config file parameter called 'loader' which is
used to refer to the HVM firmware instead.
The conventions for loading the initrd image say that it should
live at the end of memory. This requires QEMU to know the size
of the guest's initial RAM allocation, so image.py is changed
to pass the '-m' flag to QEMU.
The HVMImageHandler class in image.py is changed so that if the
'kernel', 'ramdisk' or 'extra' params were given in the config
these are passed to QEMU with the '-kernel', '-initrd' and
'-append' flags respectively.
Finally, the 'loader' param is used as the arg to 'xc_hvm_build'
instead of the old 'kernel' param.
For the sake of compatability with old HVM guest config files,
if the config file has a 'kernel' param whose path matches that
of the HVM firmware, then we automatically convert this 'kernel'
param into the 'loader' param. This ensures existing HVM guests
work without changes required.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
This patch introduces a hack to make non-relocatable kernels
bootable too. Non-relocatable kernels absolutely want to run
at 0x100000 and are not at all happy about being at 0x200000.
Fortunately, thanks to crazy programs like LOADLIN, Linux has
a couple of hooks in its boot process which can be used to
play games. The 'code32_switch' hook is executed immediately
following the switch to protected mode.
So, this patch installs a hook at 0x200000+kernel_size. The hook
is hand crafted assembly which sets up all the segments as needed,
then essentially does memmove(0x100000,0x200000,kernel_size) and
finally does an unconditional jmp to 0x100000.
Amazingly this actually really does work. It has been successfully
tested with RHEL-2.1 and Fedora Core 6 install kernels on i386, and
Fedora Core 6 and 7 kernels on x86_64.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
This patch introduces the basic infrastructure for direct kernel
boot in the ioemu copy of QEMU. The current #ifdef disabled
code is actually obsolete wrt to upstream QEMU code. So this
is removed entirely. In its place I have imported the latest
upstream QEMU code. The QEMU code assumes that the guest RAM
is directly mapped into the QEMU process, so there were some
changes neccessary. Instead of strcpy/memcpy'ing the args
and kernel header into guest RAM, cpu_physical_memory_rw is
used. Intead of fread() the initrd and kernel into guest RAM
a helper function is used fread2guest which reads into a small
buffer and then uses cpu_physical_memory_rw.
NB in reading the following, Documentation/i386/boot.txt is
a useful reference for what's going on.
Next, instead of loading the kernel at 0x100000, this code
loads it at 0x200000. This is far enough away that there's
no risk of it overlapping with the HVM firmware image. If the
Linux kernel boot protocol is 0x205 or later, and the flag
at offset 0x234 in the kernel header is 1, then the guest
kernel was built with CONFIG_RELOCATABLE=y.
In this scenario we merely need to tell the kernel what address
it has been relocated to by writing 0x200000 into the kernel
header at offset 0x214. When switching from real mode into
protected mode the kernel will immediately start executing at
0x200000 and be happy with life. This should work for 2.6.20 or
later on i386, and 2.6.22 or later on x86_64.
This has been verified with Fedora 7 and Fedora 8 bare metal kernels
on i386 and x86_64 from the $TREE/images/pxeboot of the install trees.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
xc.domain_destroy_hook() is called twice when we execute
xm shutdown/reboot command to an HVM domain without PV drivers.
The first calling is from shutdown() in XendDomainInfo.py.
The second calling is from destroyDomain() in XendDomainInfo.py.
The first calling is not necessary, so this patch removes it.
A discussion about this patch is as follows.
http://lists.xensource.com/archives/html/xen-ia64-devel/2008-01/msg00232.html
Signed-off-by: Masaki Kanno <kanno.masaki@jp.fujitsu.com>
Keir Fraser