xen.git
18 years agoxen: Fix build with recent GCC (4.1+).
Keir Fraser [Sat, 28 Apr 2007 08:28:00 +0000 (09:28 +0100)]
xen: Fix build with recent GCC (4.1+).
Signed-off-by: Keir Fraser <keir@xensource.com>
18 years agoBump all API version numbers to 1.0.0.
ewan@localhost.localdomain [Sat, 28 Apr 2007 01:53:23 +0000 (18:53 -0700)]
Bump all API version numbers to 1.0.0.

Signed-off-by: Ewan Mellor <ewan@xensource.com>
18 years agoEnsure blk{tap/back}'s vm_area is not freed twice on error case.
Jake Wires [Sat, 28 Apr 2007 00:54:06 +0000 (17:54 -0700)]
Ensure blk{tap/back}'s vm_area is not freed twice on error case.
Signed-off-by: Jake Wires <jwires@xensource.com>
18 years agovnet: Fix build of the vnet module.
kfraser@localhost.localdomain [Fri, 27 Apr 2007 15:27:07 +0000 (16:27 +0100)]
vnet: Fix build of the vnet module.
- correcting KERNEL_SRC ( ./build-linux... )
- typing errors in Makefile.ver
- module-params
Signed-off-by: Robert Valentan <R.Valentan@solid-soft.at>
18 years agoAdd INIT INFO block to example init.d/xend file.
kfraser@localhost.localdomain [Fri, 27 Apr 2007 15:21:20 +0000 (16:21 +0100)]
Add INIT INFO block to example init.d/xend file.
Without it SuSE (and others?) will start xend before the
network.
Signed-off-by: Robert Valentan <R.Valentan@solid-soft.at>
18 years agoHandle EOF on xenstored socket.
Steven Hand [Fri, 27 Apr 2007 15:17:44 +0000 (16:17 +0100)]
Handle EOF on xenstored socket.
Suggested by Pavel Kankovsky <peak@argo.troja.mff.cuni.cz>.

Signed-off-by: Steven Hand <steven@xensource.com>
18 years agoxend: Auto-balloon a couple megabytes before creating an HVM domain as
kfraser@localhost.localdomain [Fri, 27 Apr 2007 14:22:34 +0000 (15:22 +0100)]
xend: Auto-balloon a couple megabytes before creating an HVM domain as
Xen will allocate 1MB for shadow memory immediately.
Suggested by Daniel Berrange.
Signed-off-by: Keir Fraser <keir@xensource.com>
18 years agoxend: Don't use -q flag to bootloader if activating text console.
kfraser@localhost.localdomain [Fri, 27 Apr 2007 14:12:48 +0000 (15:12 +0100)]
xend: Don't use -q flag to bootloader if activating text console.

In Xen 3.0.3 series release, if one did 'xm create -c someguest' then
you would be shown the pygrub boot menu, allowing one to pick a
kernel. If you didn't use the -c arg, then it'd just pick the default
kernel.

In Xen 3.0.5 testing, xm is forcing the addition of the '-q' flag
whenever the 'bootargs' parameter is not given in the config file. So
if one does not have 'bootargs' specified, and one uses 'xm create -c
someguest' you never get to see the pygrub boot menu.

This patch fixes 'xm create' so that it does not add the '-q' flag
to the bootloader args, if the '-c' flag is given to xm create. This
makes the default user experiance operate closer to user
expectations.

Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
18 years ago[XEND] make *.get_all_records return dict of ref->record
Tom Wilkie [Fri, 27 Apr 2007 14:09:39 +0000 (15:09 +0100)]
[XEND] make *.get_all_records return dict of ref->record

Also clean up some handling of uuids and refs in xm

signed-of-by: Tom Wilkie <tom.wilkie@gmail.com>

18 years agoxen: More 'IS_COMPAT' cleanups.
kfraser@localhost.localdomain [Fri, 27 Apr 2007 14:06:55 +0000 (15:06 +0100)]
xen: More 'IS_COMPAT' cleanups.
Signed-off-by: Keir Fraser <keir@xensource.com>
18 years agoAllow dynamic_min/dynamic_max to (temporarily) be out of sync; in practice
Steven Hand [Fri, 27 Apr 2007 13:45:06 +0000 (14:45 +0100)]
Allow dynamic_min/dynamic_max to (temporarily) be out of sync; in practice
we keep them identical but non-atomically.

Signed-off-by: Steven Hand <steven@xensource.com>
18 years agoSanity check changes to memory dynamic/static min/max and ensure we cannot
Steven Hand [Fri, 27 Apr 2007 13:16:22 +0000 (14:16 +0100)]
Sanity check changes to memory dynamic/static min/max and ensure we cannot
get ourselves into an inconsistent state.

Signed-off-by: Steven Hand <steven@xensource.com>
18 years agoMerge.
Steven Smith [Fri, 27 Apr 2007 12:31:09 +0000 (13:31 +0100)]
Merge.

18 years ago32 on 64: Don't lock up Xen if you need to do lots of grant operations
Steven Smith [Fri, 27 Apr 2007 12:25:58 +0000 (13:25 +0100)]
32 on 64: Don't lock up Xen if you need to do lots of grant operations
in a single batch for a 32-on-64 domain.

Signed-off-by: Steven Smith <sos22@cam.ac.uk>
18 years ago[XM] Differentiate between VLAN PIFs and normal PIFs in network-show
Tom Wilkie [Fri, 27 Apr 2007 11:51:14 +0000 (12:51 +0100)]
[XM] Differentiate between VLAN PIFs and normal PIFs in network-show

signed-off-by: Tom Wilkie <tom.wilkie@gmail.com>

18 years ago[XEND] Fix up PIF class, make it work
Tom Wilkie [Fri, 27 Apr 2007 11:49:17 +0000 (12:49 +0100)]
[XEND] Fix up PIF class, make it work

signed-off-by: Tom Wilkie <tom.wilkie@gmail.com>

18 years ago[XEND] Used shutil.rename to cope with /tmp and /var on different file systems
Tom Wilkie [Fri, 27 Apr 2007 10:03:47 +0000 (11:03 +0100)]
[XEND] Used shutil.rename to cope with /tmp and /var on different file systems

signed-off-by: Tom Wilkie <tom.wilkie@gmail.com>

18 years ago[XM] Fix typo in XenAPI create code
Tom Wilkie [Fri, 27 Apr 2007 10:02:47 +0000 (11:02 +0100)]
[XM] Fix typo in XenAPI create code

signed-off-by: Tom Wilkie <tom.wilkie@gmail.com>

18 years agoMerge
Tim Deegan [Fri, 27 Apr 2007 09:54:37 +0000 (10:54 +0100)]
Merge

18 years ago[HVM] Save/restore: pad save structs up to a multiple of eight bytes.
Tim Deegan [Fri, 27 Apr 2007 09:54:00 +0000 (10:54 +0100)]
[HVM] Save/restore: pad save structs up to a multiple of eight bytes.
Signed-off-by: Tim Deegan <Tim.Deegan@xensource.com>
18 years ago[XEND] Start cleanly even with bad xml state files
Tom Wilkie [Fri, 27 Apr 2007 09:44:11 +0000 (10:44 +0100)]
[XEND] Start cleanly even with bad xml state files

signed-off-by: Tom Wilkie <tom.wilkie@gmail.com>

18 years ago[XEN] Replace IS_COMPAT() in the shadow code with pv_32on64_*()
Tim Deegan [Fri, 27 Apr 2007 09:24:09 +0000 (10:24 +0100)]
[XEN] Replace IS_COMPAT() in the shadow code with pv_32on64_*()
which test for the actual case we're interested in (PV PAE
pagetables disguised as 64bit ones by the compat mm code).
Signed-off-by: Tim Deegan <Tim.Deegan@xensource.com>
18 years agohvm: Take care when checking ->is_compat flag now that HVM as well as
kfraser@localhost.localdomain [Thu, 26 Apr 2007 17:33:34 +0000 (18:33 +0100)]
hvm: Take care when checking ->is_compat flag now that HVM as well as
PV guests can assert it.
Signed-off-by: Keir Fraser <keir@xensource.com>
18 years agoAdd support for creating networks to xm create XML file
Tom Wilkie [Thu, 26 Apr 2007 16:30:56 +0000 (17:30 +0100)]
Add support for creating networks to xm create XML file

signed-off-by: Tom Wilkie <tom.wilkie@gmail.com>

18 years agohvmloader: Place SMBIOS tables at 0xE9000. Previous location was lower
kfraser@localhost.localdomain [Thu, 26 Apr 2007 15:45:06 +0000 (16:45 +0100)]
hvmloader: Place SMBIOS tables at 0xE9000. Previous location was lower
than nay native system would place them.
Signed-off-by: Keir Fraser <keir@xensource.com>
18 years ago[XM-TEST] Add test for new network functions
Tom Wilkie [Thu, 26 Apr 2007 15:15:08 +0000 (16:15 +0100)]
[XM-TEST] Add test for new network functions

signed-off-by: Tom Wilkie <tom.wilkie@gmail.com>

18 years ago[XM-TEST] Move network_utils.py to lib directory
Tom Wilkie [Thu, 26 Apr 2007 14:48:09 +0000 (15:48 +0100)]
[XM-TEST] Move network_utils.py to lib directory

signed-off-by: Tom Wilkie <tom.wilkie@gmail.com>

18 years agoAdd network-del and network-show commands to xm
Tom Wilkie [Thu, 26 Apr 2007 14:40:35 +0000 (15:40 +0100)]
Add network-del and network-show commands to xm

signed-off-by: Tom Wilkie <tom.wilkie@gmail.com>

18 years agoAdd network.destroy() to XenAPI
Tom Wilkie [Thu, 26 Apr 2007 14:39:30 +0000 (15:39 +0100)]
Add network.destroy() to XenAPI

signed-off-by: Tom Wilkie <tom.wilkie@gmail.com>

18 years agodomain builder: Fix sanity check for gzip size.
kfraser@localhost.localdomain [Thu, 26 Apr 2007 14:42:49 +0000 (15:42 +0100)]
domain builder: Fix sanity check for gzip size.
Signed-off-by: Keir Fraser <keir@xensource.com>
18 years agoacm: Remove unecessary compat shim from do_acm_op().
kfraser@localhost.localdomain [Thu, 26 Apr 2007 14:36:09 +0000 (15:36 +0100)]
acm: Remove unecessary compat shim from do_acm_op().
Signed-off-by: Keir Fraser <keir@xensource.com>
18 years agoacm: Fix 32-bit build with acm enabled.
kfraser@localhost.localdomain [Thu, 26 Apr 2007 14:25:09 +0000 (15:25 +0100)]
acm: Fix 32-bit build with acm enabled.
Signed-off-by: Stefan Berger <stefanb@us.ibm.com>
18 years agoblktap: Segments can span multiple clusters with tap:qcow.
kfraser@localhost.localdomain [Thu, 26 Apr 2007 10:29:41 +0000 (11:29 +0100)]
blktap: Segments can span multiple clusters with tap:qcow.

In blktap's qcow we need split up read/write requests if the requests
span multiple clusters. However, with our MAX_AIO_REQUESTS define we
assume that there is only ever a single aio request per tapdisk
request and under heavy i/o we can run out of room causing us to
cancel requests.

Dynamically allocate (based on cluster_bits) the various io request
queues the driver maintains.

Signed-off-by: Mark McLoughlin <markmc@redhat.com>
18 years agohvm ioemu: Fix SCSI initialisation after c/s 14942.
kfraser@localhost.localdomain [Thu, 26 Apr 2007 10:18:42 +0000 (11:18 +0100)]
hvm ioemu: Fix SCSI initialisation after c/s 14942.
Signed-off-by: Keir Fraser <keir@xensource.com>
18 years agoxen x86/64 compat: Fix size of accesses to trap_bounce.cs in asm.
kfraser@localhost.localdomain [Thu, 26 Apr 2007 09:51:49 +0000 (10:51 +0100)]
xen x86/64 compat: Fix size of accesses to trap_bounce.cs in asm.
Signed-off-by: Keir Fraser <keir@xensource.com>
18 years agoRemove non-ASCII chars in man page.
kfraser@localhost.localdomain [Thu, 26 Apr 2007 08:52:34 +0000 (09:52 +0100)]
Remove non-ASCII chars in man page.
Signed-off-by: Charles Coffing <ccoffing@novell.com>
18 years agoxend: Tidy bootloader device constant.
kfraser@localhost.localdomain [Thu, 26 Apr 2007 08:51:39 +0000 (09:51 +0100)]
xend: Tidy bootloader device constant.

In XendDomainInfo, BOOTLOADER_LOOPBACK_DEVICE is defined to be
'/dev/xvdp'.  However, there is still one place in this file that has
'/dev/xvdp' hard-coded.

Signed-off-by: Charles Coffing <ccoffing@novell.com>
18 years agohvm vmx: Mark bootstrap-initialised variables as read-mostly.
Keir Fraser [Wed, 25 Apr 2007 21:56:09 +0000 (22:56 +0100)]
hvm vmx: Mark bootstrap-initialised variables as read-mostly.
Signed-off-by: Keir Fraser <keir@xensource.com>
18 years agohvm vmx: Another small fix for CPUs with no TPR_SHADOW support.
Keir Fraser [Wed, 25 Apr 2007 21:29:39 +0000 (22:29 +0100)]
hvm vmx: Another small fix for CPUs with no TPR_SHADOW support.
Signed-off-by: Keir Fraser <keir@xensource.com>
18 years agoxen: Fix maximum_gpfn() hypercall to always return max_gpfn not nr_gpfns.
Keir Fraser [Wed, 25 Apr 2007 21:22:31 +0000 (22:22 +0100)]
xen: Fix maximum_gpfn() hypercall to always return max_gpfn not nr_gpfns.
Fix callers to convert this to nr_gpfns (aka p2m_size) if that's what
they actually need.
Signed-off-by: Keir Fraser <keir@xensource.com>
18 years agohvm vmx: Only use TPR_SHADOW feature if supported.
Keir Fraser [Wed, 25 Apr 2007 21:06:13 +0000 (22:06 +0100)]
hvm vmx: Only use TPR_SHADOW feature if supported.
Signed-off-by: Keir Fraser <keir@xensource.com>
18 years agohvm ioemu: Do not initialise the LSI SCSI interface if no SCSI discs
kfraser@localhost.localdomain [Wed, 25 Apr 2007 17:12:03 +0000 (18:12 +0100)]
hvm ioemu: Do not initialise the LSI SCSI interface if no SCSI discs
are attached to the guest.
Signed-off-by: Keir Fraser <keir@xensource.com>
18 years ago[XEND] No need to refresh bridges under new model...
Tom Wilkie [Wed, 25 Apr 2007 15:18:34 +0000 (16:18 +0100)]
[XEND] No need to refresh bridges under new model...

signed-off-by: Tom Wilkie <tom.wilkie@gmail.com>

18 years ago[XEND] Fully implement XenAPI network and PIF classes.
Tom Wilkie [Wed, 25 Apr 2007 15:17:51 +0000 (16:17 +0100)]
[XEND] Fully implement XenAPI network and PIF classes.

Now support creation of Networks and PIFs.  Auto discover current network settings.

signed-off-by: Tom Wilkie <tom.wilkie@gmail.com>

18 years ago[XEND] Remove autoplug class signatures, have it all handled in the class now
Tom Wilkie [Wed, 25 Apr 2007 15:16:21 +0000 (16:16 +0100)]
[XEND] Remove autoplug class signatures, have it all handled in the class now

signed-off-by: Tom Wilkie <tom.wilkie@gmail.com>

18 years ago[XEND] Make existing autoplug classes implement new XendBase
Tom Wilkie [Wed, 25 Apr 2007 15:14:20 +0000 (16:14 +0100)]
[XEND] Make existing autoplug classes implement new XendBase

signed-off-by: Tom Wilkie <tom.wilkie@gmail.com>

18 years ago[XM] Fix network-attach when a bridge is specified, add network-new command to create...
Tom Wilkie [Wed, 25 Apr 2007 15:12:41 +0000 (16:12 +0100)]
[XM] Fix network-attach when a bridge is specified, add network-new command to create new XenAPI network

signed-off-by: Tom Wilkie <tom.wilkie@gmail.com>

18 years ago[XEND] Add more exceptions and error messages
Tom Wilkie [Wed, 25 Apr 2007 15:11:20 +0000 (16:11 +0100)]
[XEND] Add more exceptions and error messages

signed-off-by: Tom Wilkie <tom.wilkie@gmail.com>

18 years agoUse new instance store to resolve network uuid to bridge name
Tom Wilkie [Wed, 25 Apr 2007 14:42:34 +0000 (15:42 +0100)]
Use new instance store to resolve network uuid to bridge name

signed-off-by: Tom Wilkie <tom.wilkie@gmail.com>

18 years agoChange how XendStateStore store XML for persistance.
Tom Wilkie [Wed, 25 Apr 2007 14:39:48 +0000 (15:39 +0100)]
Change how XendStateStore store XML for persistance.

Make it explicity store empty dicts and list.

signed-off-by: Tom Wilkie <tom.wilkie@gmail.com>

18 years agoAdd new base class and instance store for new autoplugged api classes
Tom Wilkie [Wed, 25 Apr 2007 14:38:04 +0000 (15:38 +0100)]
Add new base class and instance store for new autoplugged api classes

signed-off-by: Tom Wilkie <tom.wilkie@gmail.com>

18 years agoIf bridge name hasn't been recorded in XendConfig then try and work it out using...
Tom Wilkie [Wed, 25 Apr 2007 12:55:18 +0000 (13:55 +0100)]
If bridge name hasn't been recorded in XendConfig then try and work it out using Brctl.

This is needed when a VIF is created without specifying a bridge.

signed-off-by: Tom Wilkie <tom.wilkie@gmail.com>

18 years agoTranslate network uuid to bridge name and store it in XendConfig when createing a...
Tom Wilkie [Wed, 25 Apr 2007 12:44:37 +0000 (13:44 +0100)]
Translate network uuid to bridge name and store it in XendConfig when createing a new vif.

Fixes a bug where a new vif gets added to the wrong bridge.

signed-off-by: Tom Wilkie <tom.wilkie@gmail.com>

18 years agoxen: Fix up use of trap_bounce structure.
kfraser@localhost.localdomain [Wed, 25 Apr 2007 12:50:20 +0000 (13:50 +0100)]
xen: Fix up use of trap_bounce structure.
Fixes suggested by Jan Beulich.
Signed-off-by: Keir Fraser <keir@xensource.com>
18 years agohvm save/restore: Use a GUEST_HANDLE_64 rather than an ordinary GUEST_HANDLE
Steven Smith [Wed, 25 Apr 2007 11:04:55 +0000 (12:04 +0100)]
hvm save/restore: Use a GUEST_HANDLE_64 rather than an ordinary GUEST_HANDLE
when getting the HVM context so as 32 bit dom0s on a 64 bit hypervisor
do the right thing.

Signed-off-by: Steven Smith <sos22@cam.ac.uk>
18 years agolibxc: Fix xc_translate_foreign_address() to parse non-pae pagetables
kfraser@localhost.localdomain [Wed, 25 Apr 2007 08:49:18 +0000 (09:49 +0100)]
libxc: Fix xc_translate_foreign_address() to parse non-pae pagetables
correctly.
From: Chris Morrow <cmorrow@yottayotta.com>
Signed-off-by: Keir Fraser <keir@xensource.com>
18 years agolinux x86/64: Fix supervisor_mode_kernel test to read a single byte.
kfraser@localhost.localdomain [Wed, 25 Apr 2007 08:46:43 +0000 (09:46 +0100)]
linux x86/64: Fix supervisor_mode_kernel test to read a single byte.
Spotted by Jan Beulich.
Signed-off-by: Keir Fraser <keir@xensource.com>
18 years agoxend: Fix race in vfb/vkbd device setup.
kfraser@localhost.localdomain [Wed, 25 Apr 2007 08:44:20 +0000 (09:44 +0100)]
xend: Fix race in vfb/vkbd device setup.

 1. XendDomainInfo._createDevices() gets a list of devices to be
    created from XendConfig.ordered_device_refs().
    On a simple guest, this has 4 devices, vfb, vbd, vif, vkbd - in
    that order.

 2. It iterates over those devices, creating the appropriate
    DevController subclass instance, and then calling createDevice()
    on that object.

 3. When createDevice() is called on the vfb  device, it spawns
    xen-vncfb daemon.

 4. During startup xen-vncfb writes into the backend paths
            /local/domain/0/backend/vfb/0
     And
            /local/domain/0/backend/vkbd/0

 5. When createDevice() is called on the vkbd device in XenD, if the
    2nd xenstore path write from step 4 has occurred, then you'll hit
    the 'Device 0 (vkbd) is already connected' error. If the 2nd path
    write didn't complete yet then everything is fine.

I think the reason it often works once after boot is that loading
xen-vncfb from disk the first time around is just enough of a slow
down to ensure step 5 occurs before the 2nd xenstore write in step 4
has occurred.

The key seems to be to ensure the vkbd device is initialized in
xenstore before the vfb device - this ensures all the xenstored setup
from XenD is complete before the xen-vncfb daemon starts. I'm now able
to create & destroy a domain many times over with this patch & never
hit the error message any more.

Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
18 years agoacm: Updating a policy on a running system.
kfraser@localhost.localdomain [Wed, 25 Apr 2007 08:40:02 +0000 (09:40 +0100)]
acm: Updating a policy on a running system.

Allow a policy to be updated on a running system and domains to be
relabeled. The updating of a policy is happening in several steps:
relabeling the domains, testing whether the system would be in a valid
state after the relabeling (according to the policy), committing the
changes if state is determined to be valid.

I have followed Keir's suggestion of building a 2nd linked list
parallel to the domain list. That 2nd list holds security information
related to the running domains. Each entry is pointed to by its domain
structure. The list is protected by its own read/write-lock. I have
moved nearly all ACM-related code that was traversing the domain list
previously to traverse this list instead and not hold onto the domain
list lock.

Signed-off-by: Stefan Berger <stefanb@us.ibm.com>
18 years agoacm: Changes to XML schema of the policy
kfraser@localhost.localdomain [Wed, 25 Apr 2007 08:31:52 +0000 (09:31 +0100)]
acm: Changes to XML schema of the policy

This patch changes the XML schema of the ACM policy to require a
version and that every conflict set have a name. Every VM label must
have one Chinese Wall Type and every resource label one Simple Type
Enforcement Type. As a consequence of this some example policies
needed to be changed. Also not offering that many configuration
options for compiling xen anymore to make things simpler.

Signed-off-by: Stefan Berger <stefanb@us.ibm.com>
18 years agoxend: Settle on 0666 pre-umask permissions for xend-debug.log.
kfraser@localhost.localdomain [Tue, 24 Apr 2007 22:40:40 +0000 (23:40 +0100)]
xend: Settle on 0666 pre-umask permissions for xend-debug.log.
Signed-off-by: Keir Fraser <keir@xensource.com>
18 years agoxend: Suppress logging of "Listing domains in state foo" in xend.log.
Keir Fraser [Tue, 24 Apr 2007 22:12:10 +0000 (23:12 +0100)]
xend: Suppress logging of "Listing domains in state foo" in xend.log.
Signed-off-by: Jim Fehlig <jfehlig@novell.com>
18 years agoLINUX: Fixes for XENFEAT_supervisor_mode_kernel and XENFEAT_writable_page_tables.
Ian Campbell [Tue, 24 Apr 2007 21:35:16 +0000 (22:35 +0100)]
LINUX: Fixes for XENFEAT_supervisor_mode_kernel and XENFEAT_writable_page_tables.

Signed-of-by: Ian Campbell <ian.campbell@xensource.com>
18 years agohvm: Choose correct shared_info format for 32-bit PV drivers on 64-bit Xen.
Keir Fraser [Tue, 24 Apr 2007 21:00:45 +0000 (22:00 +0100)]
hvm: Choose correct shared_info format for 32-bit PV drivers on 64-bit Xen.
Signed-off-by: K. Y. Srinivasan <ksrinivasan@novell.com>
Signed-off-by: Keir Fraser <keir@xensource.com>
18 years agoacm: Changes to files missed in previous cleanup patch.
Keir Fraser [Tue, 24 Apr 2007 20:50:00 +0000 (21:50 +0100)]
acm: Changes to files missed in previous cleanup patch.
Signed-off-by: Stefan Berger <stefanb@us.ibm.com>
18 years agoRemove execute permission from xend-debug.log
Keir Fraser [Tue, 24 Apr 2007 20:48:09 +0000 (21:48 +0100)]
Remove execute permission from xend-debug.log

The file /var/log/xen/xend-debug.log is currently being created with
executable permission bits set. This is because the os.open() method
defaults to using a mode of 0777 if no third parameter is provided.
The attached patch changes the mode to 0600 to ensure that the file
permissions come out as -rw-------  instead of -rwxr-xr-x

Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
18 years agoacm: Move acm_domain_destroy() hook to complete_domain_destroy().
Keir Fraser [Tue, 24 Apr 2007 20:44:41 +0000 (21:44 +0100)]
acm: Move acm_domain_destroy() hook to complete_domain_destroy().
Signed-off-by: Keir Fraser <keir@xensource.com>
18 years agoxen ia64: Fix build breakage after ACM cleanup.
Keir Fraser [Tue, 24 Apr 2007 20:42:08 +0000 (21:42 +0100)]
xen ia64: Fix build breakage after ACM cleanup.
Signed-off-by: Alex Williamson <alex.williamson@hp.com>
18 years agoMerge with xen-ia64-unstable.hg
Keir Fraser [Tue, 24 Apr 2007 20:39:23 +0000 (21:39 +0100)]
Merge with xen-ia64-unstable.hg

18 years ago[qemu-dm] Check for null pointer in dma handler function pointer.
Christian Limpach [Tue, 24 Apr 2007 16:02:10 +0000 (17:02 +0100)]
[qemu-dm] Check for null pointer in dma handler function pointer.

Signed-off-by: Christian Limpach <Christian.Limpach@xensource.com>
18 years ago[qemu-dm] Add bounds checks to block device reads/writes.
Christian Limpach [Tue, 24 Apr 2007 16:02:09 +0000 (17:02 +0100)]
[qemu-dm] Add bounds checks to block device reads/writes.

Signed-off-by: Christian Limpach <Christian.Limpach@xensource.com>
18 years ago[qemu-dm] Add bounds checks for cirrus bitblit memory accesses.
Christian Limpach [Tue, 24 Apr 2007 16:02:07 +0000 (17:02 +0100)]
[qemu-dm] Add bounds checks for cirrus bitblit memory accesses.

Signed-off-by: Christian Limpach <Christian.Limpach@xensource.com>
18 years agoacm: Code restructuring on domain create/destroy.
kfraser@localhost.localdomain [Tue, 24 Apr 2007 15:52:15 +0000 (16:52 +0100)]
acm: Code restructuring on domain create/destroy.

When a domain is created, the function acm_domain_create() in
domain_create() is called that does what previously the pre- and
post_domain_create functions were doing. Similarly there's a function
acm_domain_destroy() in domain_kill() that reverts changes to state
when destroying a domain. There's no more separate initialization
necessary for domain-0, but domain_create takes one more parameter,
the ssidref. It is usually passed through the hypercall when a domain
is created.

Signed-off-by: Stefan Berger <stefanb@us.ibm.com>
18 years agohvm: Fix ioreq structure size to be 32/64-bit invariant.
kfraser@localhost.localdomain [Tue, 24 Apr 2007 15:28:37 +0000 (16:28 +0100)]
hvm: Fix ioreq structure size to be 32/64-bit invariant.
This fixes booting HVM guests with 32-bit dom0 on 64-bit Xen.
Signed-off-by: Keir Fraser <keir@xensource.com>
18 years agomerge with xen-unstable.hg
Alex Williamson [Tue, 24 Apr 2007 15:26:32 +0000 (09:26 -0600)]
merge with xen-unstable.hg

18 years ago[IA64] Fix ptc.g race
Alex Williamson [Tue, 24 Apr 2007 15:07:24 +0000 (09:07 -0600)]
[IA64] Fix ptc.g race

If one vcpu is executing ptc.g, while the other vcpu is executing itc,
the VHPT(VTLB) entry which should be purged may unexpectedly survive.
Then issue happens.

Signed-off-by: Anthony Xu <anthony.xu@intel.com>
18 years agoxm-test: Scale memory parameter for tests using XenAPIDomain.
kfraser@localhost.localdomain [Tue, 24 Apr 2007 14:01:13 +0000 (15:01 +0100)]
xm-test: Scale memory parameter for tests using XenAPIDomain.

Scale the memory parameter for domains using the XenAPIDomain to
create a domain. Those domains use the normal configuration where the
memory parameter is given in 'Mb', so it needs to be adapted to
'bytes'.

Signed-off-by: Stefan Berger <stefanb@us.ibm.com>
18 years ago[HVM][HAP] Don't kill domains if dom0 requests an invalid shadow op.
Tim Deegan [Tue, 24 Apr 2007 13:08:57 +0000 (14:08 +0100)]
[HVM][HAP] Don't kill domains if dom0 requests an invalid shadow op.
Signed-off-by: Tim Deegan <Tim.Deegan@xensource.com>
18 years agohvm rombios: No need for SS prefix when EA base is %bp.
kfraser@localhost.localdomain [Tue, 24 Apr 2007 12:49:11 +0000 (13:49 +0100)]
hvm rombios: No need for SS prefix when EA base is %bp.
Signed-off-by: Keir Fraser <keir@xensource.com>
18 years agohvm rombios: Do not screw around with the PIT in the BIOS 32-bit
kfraser@localhost.localdomain [Tue, 24 Apr 2007 11:14:57 +0000 (12:14 +0100)]
hvm rombios: Do not screw around with the PIT in the BIOS 32-bit
extensions. mdelay() can more easily be implemented by polling the
DRAM refresh bit in port 0x61. And this does not mess with state that
a guest may be relying on.
Signed-off-by: Keir Fraser <keir@xensource.com>
18 years agohvm rombios: Fix int15, func 0x86 (udelay).
kfraser@localhost.localdomain [Tue, 24 Apr 2007 11:13:58 +0000 (12:13 +0100)]
hvm rombios: Fix int15, func 0x86 (udelay).
Was reading argument from wrong stack location with recent versions of
bcc. Doing the access from [bp] rather than [sp] is safer as the
former does not change all the time.
Signed-off-by: Keir Fraser <keir@xensource.com>
18 years agoqemu: Do not kill ioemu if undefined LSI-SCI registers are read. Just
kfraser@localhost.localdomain [Tue, 24 Apr 2007 11:12:48 +0000 (12:12 +0100)]
qemu: Do not kill ioemu if undefined LSI-SCI registers are read. Just
return zeroes. This path is triggered by ASPI8DOS.SYS in DRDOS.
Signed-off-by: Keir Fraser <keir@xensource.com>
18 years ago[XEN] Truncate arithmetic for PERFSTATUS counters to the width of perfc_t.
Tim Deegan [Tue, 24 Apr 2007 10:39:13 +0000 (11:39 +0100)]
[XEN] Truncate arithmetic for PERFSTATUS counters to the width of perfc_t.
Since these are all calculated per-cpu now, a status counter that is
incremented on one cpu and decremented on another will underflow on
the decrement.  When we add the per-cpu values to get the total, if
we don't truncate the sum it will be max-perfc + 1 instead of zero.
Signed-off-by: Tim Deegan <Tim.Deegan@xensource.com>
18 years agohvmloader: Fix e820_collapse() to not lose E820 entries.
kfraser@localhost.localdomain [Tue, 24 Apr 2007 09:34:47 +0000 (10:34 +0100)]
hvmloader: Fix e820_collapse() to not lose E820 entries.
Fixes bugzilla bug #971.
Signed-off-by: Keir Fraser <keir@xensource.com>
18 years agovtpm: Decrease amount of pages on shared ring.
kfraser@localhost.localdomain [Tue, 24 Apr 2007 08:50:56 +0000 (09:50 +0100)]
vtpm: Decrease amount of pages on shared ring.

Decrease the amount of shared pages on the ring to only one page -
more is not needed.

Signed-off-by: Stefan Berger <stefanb@us.ibm.com>
18 years agoDon't bypass the spin_unlock at the end of do_domctl().
kfraser@localhost.localdomain [Tue, 24 Apr 2007 08:50:12 +0000 (09:50 +0100)]
Don't bypass the spin_unlock at the end of do_domctl().
Signed-off-by: Stefan Berger <stefanb@us.ibm.com>
18 years agokexec: Avoid unaligned writes when producing crash info.
kfraser@localhost.localdomain [Tue, 24 Apr 2007 08:48:08 +0000 (09:48 +0100)]
kexec: Avoid unaligned writes when producing crash info.

The alignment of info is 32bits in line with the elf specification.
This means that on 64 bit architectures accessing it directly may
result unaligned access and a panic. I have been experiencing
this on IA64.

It seems that a simple approach of having an crash_xen_info_t on
the stack and simply memcopying it into info at the end alleviates
the problem.

Signed-off-by: Simon Horman <horms@verge.net.au>
18 years agoxend: Incluide processing of 'monitor' config option in qemu-dm
kfraser@localhost.localdomain [Tue, 24 Apr 2007 08:44:51 +0000 (09:44 +0100)]
xend: Incluide processing of 'monitor' config option in qemu-dm
cmdline processing.
Signed-off-by: Yongkang You <yongkang.you@intel.com>
18 years ago[TAPDISK] Move tapdisk control files from /dev/xen to /var/run/tap
Jake Wires [Tue, 24 Apr 2007 00:15:38 +0000 (17:15 -0700)]
[TAPDISK] Move tapdisk control files from /dev/xen to /var/run/tap
Signed-off-by: Jake Wires <jwires@xensource.com>
18 years ago[TAPDISK] Use incremental cookies and fix blktapctrl::test_path parameter typo.
Jake Wires [Mon, 23 Apr 2007 22:37:53 +0000 (15:37 -0700)]
[TAPDISK] Use incremental cookies and fix blktapctrl::test_path parameter typo.
Signed-off-by: Jake Wires <jwires@xensource.com>
18 years agonetback: Copy skbuffs that are presented to the start_xmit() function
kfraser@localhost.localdomain [Mon, 23 Apr 2007 15:37:03 +0000 (16:37 +0100)]
netback: Copy skbuffs that are presented to the start_xmit() function
with a multi-page head buffer. The new skbuff will not have a
multi-page head buffer, which is a necessary restriction until this is
supported by the internals of the netback driver.
Signed-off-by: Steven Smith <ssmith@xensource.com>
Signed-off-by: Keir Fraser <keir@xensource.com>
18 years agoAPI version to 0.9.1.
ewan@localhost.localdomain [Sat, 21 Apr 2007 21:12:09 +0000 (14:12 -0700)]
API version to 0.9.1.

Signed-off-by: Ewan Mellor <ewan@xensource.com>
18 years agoIgnore the libxenapi tarball temp directory.
ewan@localhost.localdomain [Sat, 21 Apr 2007 19:43:25 +0000 (12:43 -0700)]
Ignore the libxenapi tarball temp directory.

Signed-off-by: Ewan Mellor <ewan@xensource.com>
18 years agoAdded Makefile.dist, which can make standalone tarballs.
ewan@localhost.localdomain [Sat, 21 Apr 2007 19:42:58 +0000 (12:42 -0700)]
Added Makefile.dist, which can make standalone tarballs.

Signed-off-by: Ewan Mellor <ewan@xensource.com>
18 years agoMove all internal header files to a separate directory, so that it's clear what
ewan@localhost.localdomain [Sat, 21 Apr 2007 19:21:06 +0000 (12:21 -0700)]
Move all internal header files to a separate directory, so that it's clear what
forms the public API.  Tidy up a couple of macros in xen_internal.h.  Remove
test_hvm_bindings, which no longer builds, and which has had its functionality
subsumed into test_bindings.  Bump the library version to 0.9.

Signed-off-by: Ewan Mellor <ewan@xensource.com>
18 years agoxenstored: Only dom0 can do XS_DEBUG in non-test environments.
kfraser@localhost.localdomain [Sat, 21 Apr 2007 10:37:34 +0000 (11:37 +0100)]
xenstored: Only dom0 can do XS_DEBUG in non-test environments.
Signed-off-by: Keir Fraser <keir@xensource.com>
18 years agoDeprecate XendDomainInfo.state - now use _stateGet() to get the live
Tom Wilkie [Fri, 20 Apr 2007 16:56:28 +0000 (17:56 +0100)]
Deprecate XendDomainInfo.state - now use _stateGet() to get the live
state of a domain.  Should still call _stateSet() to notify others
when you expect the state has changed.

Also some changes to locking in save/suspend, restore/resume and migrate

Passes xm-test against XenAPI...

signed-off-by: Tom Wilkie <tom.wilkie@gmail.com>

18 years agoRevert 14883:6568c1882af5 and 14862:98efd2e410ae.
kfraser@localhost.localdomain [Fri, 20 Apr 2007 16:51:10 +0000 (17:51 +0100)]
Revert 14883:6568c1882af5 and 14862:98efd2e410ae.

This reverts the preparation for removing the sparse tree, since these
modified scripts are not yet ready:
 1. XENGUEST suffix was lost yet depended on still by sparse tree
 Makefiles (hacked around by 14883; also reverted).
 2. XEN_TARGET_X86_PAE default value seems to be ignored
 (out-of-the-box 'make dist' will build non-PAE kernels).

Signed-off-by: Keir Fraser <keir@xensource.com>
18 years ago[XEND] Fix test for when to record HVM VNC console as a VFB.
Tim Deegan [Fri, 20 Apr 2007 16:32:30 +0000 (17:32 +0100)]
[XEND] Fix test for when to record HVM VNC console as a VFB.
This code may still need more work to properly integrate HVM SDL consoles.
Signed-off-by: Tim Deegan <Tim.Deegan@xensource.com>
18 years agoFix dist target to keep previous symlink format.
Steven Hand [Fri, 20 Apr 2007 14:29:11 +0000 (15:29 +0100)]
Fix dist target to keep previous symlink format.
May be a cleaner fix which I'll let IanC do.

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