xen.git
20 years agoMove setting of configuration variables to Config.mk.
kaf24@firebug.cl.cam.ac.uk [Wed, 5 Oct 2005 12:32:00 +0000 (13:32 +0100)]
Move setting of configuration variables to Config.mk.
Allows for partial-tree builds to be invoked from a sub-directory.
Allow for users to include a ".config" with additional custom settings.

Signed-off-by: Michal Ostrowski <mostrows@watson.ibm.com>
20 years agoRemoved the EventServer, replacing the events fired with simple logging
emellor@ewan [Wed, 5 Oct 2005 10:43:23 +0000 (11:43 +0100)]
Removed the EventServer, replacing the events fired with simple logging
messages.  This closes bug #281.  Removed references to XendRoot where these
have become unnecessary either through the work above or just through time.

Renamed some parameters in event.py, to mark them as unused.  Fix a call to a
base class constructor there too.

Signed-off-by: Ewan Mellor <ewan@xensource.com>
20 years agoimported patch /home/emellor/block_detach.patch
emellor@ewan [Tue, 4 Oct 2005 23:11:30 +0000 (00:11 +0100)]
imported patch /home/emellor/block_detach.patch

20 years agoExplicitly close files on save and restore. Rename all local variables named id...
emellor@ewan [Tue, 4 Oct 2005 23:00:42 +0000 (00:00 +0100)]
Explicitly close files on save and restore.  Rename all local variables named id or slice so as not to clash with the builtins.

Signed-off-by: Ewan Mellor <ewan@xensource.com>
20 years agoExplicitly close files.
emellor@ewan [Tue, 4 Oct 2005 22:56:42 +0000 (23:56 +0100)]
Explicitly close files.

Signed-off-by: Ewan Mellor <ewan@xensource.com>
20 years agoClean up domains if creation/restoration fails.
emellor@ewan [Tue, 4 Oct 2005 17:23:58 +0000 (18:23 +0100)]
Clean up domains if creation/restoration fails.

Signed-off-by: Ewan Mellor <ewan@xensource.com>
20 years agoFix typo.
emellor@ewan [Tue, 4 Oct 2005 16:54:33 +0000 (17:54 +0100)]
Fix typo.

Signed-off-by: Ewan Mellor <ewan@xensource.com>
20 years agoMerge the child-process handling of the save and restore functions into one
emellor@ewan [Tue, 4 Oct 2005 16:49:52 +0000 (17:49 +0100)]
Merge the child-process handling of the save and restore functions into one
forkHelper function.  Change the handling of file descriptor closure to wait
until both stdout and stderr descriptors have closed.  This may fix the
intermittent bug seen whereby xm restore; xend restart leaves the domain in a
misconfigured state, presumably because IntroduceDomain is not being called by
XendCheckpoint.restore.

Signed-off-by: Ewan Mellor <ewan@xensource.com>
20 years agoRemoved unused xen_domain method, and unused imports. Comment those methods
emellor@ewan [Tue, 4 Oct 2005 15:38:04 +0000 (16:38 +0100)]
Removed unused xen_domain method, and unused imports.  Comment those methods
expecting domains_lock protection.  Remove notifications through eserver for
_add_domain and _delete_domain, and remove the notify parameters therein.

Tidy the exception handling in domain_restore_fd.

Signed-off-by: Ewan Mellor <ewan@xensource.com>
20 years agoFix call to state_set inside preserve (was erroneously calling set_state).
emellor@ewan [Tue, 4 Oct 2005 15:27:16 +0000 (16:27 +0100)]
Fix call to state_set inside preserve (was erroneously calling set_state).
Tidy exception handling, and rename local 'channel' variable which was clashing
with the identical global name.

Signed-off-by: Ewan Mellor <ewan@xensource.com>
20 years agoFix xm restore broken by missing return statement.
emellor@ewan [Tue, 4 Oct 2005 14:57:12 +0000 (15:57 +0100)]
Fix xm restore broken by missing return statement.

Signed-off-by: Ewan Mellor <ewan@xensource.com>
20 years agoClean up sched_op() hypercall interface. One particular
kaf24@firebug.cl.cam.ac.uk [Tue, 4 Oct 2005 14:34:02 +0000 (15:34 +0100)]
Clean up sched_op() hypercall interface. One particular
change to watch out for is that the suspend record, on
save/restore, is now passed via register %edx not %esi
(both i386 and x86/64).

Signed-off-by: Keir Fraser <keir@xensource.com>
20 years agoAllow EVTCHNOP_bind_{ipi,virq} to specify the vcpu to
kaf24@firebug.cl.cam.ac.uk [Tue, 4 Oct 2005 14:02:51 +0000 (15:02 +0100)]
Allow EVTCHNOP_bind_{ipi,virq} to specify the vcpu to
bind to. Previously the alloacted port was implicitly
bound to the calling vcpu.

Signed-off-by: Keir Fraser <keir@xensource.com>
20 years agoRename VCPUOP_craete to VCPUOP_initialise. It is likely that
kaf24@firebug.cl.cam.ac.uk [Tue, 4 Oct 2005 14:02:35 +0000 (15:02 +0100)]
Rename VCPUOP_craete to VCPUOP_initialise. It is likely that
VCPU creation will actually be moved to domain-build time.

Signed-off-by: Keir Fraser <keir@xensource.com>
20 years agoFix tools to pull public header files from standard
kaf24@firebug.cl.cam.ac.uk [Tue, 4 Oct 2005 14:02:18 +0000 (15:02 +0100)]
Fix tools to pull public header files from standard
location in libxc directory. Not direct from xen and
xenlinux build trees! Also fix a type name in
x86_emulator test harness.

Signed-off-by: Keir Fraser <keir@xensource.com>
20 years agoMerge to restore reverted changes.
jrb44@swoop.cl.cam.ac.uk [Tue, 4 Oct 2005 13:38:56 +0000 (14:38 +0100)]
Merge to restore reverted changes.

20 years agoThrow away bogus merge.
jrb44@swoop.cl.cam.ac.uk [Tue, 4 Oct 2005 13:37:59 +0000 (14:37 +0100)]
Throw away bogus merge.

20 years agoMerge.
jrb44@swoop.cl.cam.ac.uk [Tue, 4 Oct 2005 13:22:30 +0000 (14:22 +0100)]
Merge.

20 years agoCatch xen.xend.XendProtocol.XendError as well as xen.xend.XendError.XendError.
emellor@ewan [Tue, 4 Oct 2005 10:43:38 +0000 (11:43 +0100)]
Catch xen.xend.XendProtocol.XendError as well as xen.xend.XendError.XendError.
This horror needs cleaning up -- this is just a stopgap.

Signed-off-by: Ewan Mellor <ewan@xensource.com>
20 years agoMerged.
emellor@ewan [Tue, 4 Oct 2005 10:42:06 +0000 (11:42 +0100)]
Merged.

20 years agoMerged.
emellor@ewan [Tue, 4 Oct 2005 10:41:07 +0000 (11:41 +0100)]
Merged.

20 years agoMerged.
emellor@ewan [Tue, 4 Oct 2005 10:40:04 +0000 (11:40 +0100)]
Merged.

20 years agoMerged.
emellor@ewan [Tue, 4 Oct 2005 10:36:53 +0000 (11:36 +0100)]
Merged.

20 years agoMerged.
emellor@ewan [Tue, 4 Oct 2005 10:36:26 +0000 (11:36 +0100)]
Merged.

20 years agoMerged.
emellor@ewan [Tue, 4 Oct 2005 10:35:21 +0000 (11:35 +0100)]
Merged.

20 years agoMerged.
emellor@ewan [Tue, 4 Oct 2005 10:33:17 +0000 (11:33 +0100)]
Merged.

20 years agoMerged.
emellor@ewan [Tue, 4 Oct 2005 10:31:31 +0000 (11:31 +0100)]
Merged.

20 years agoMerged.
emellor@ewan [Tue, 4 Oct 2005 10:30:51 +0000 (11:30 +0100)]
Merged.

Signed-off-by: Ewan Mellor <ewan@xensource.com>
20 years agoMerged.
emellor@ewan [Tue, 4 Oct 2005 10:29:36 +0000 (11:29 +0100)]
Merged.

Signed-off-by: Ewan Mellor <ewan@xensource.com>
20 years agoFix debugging statement that was referring to non-existent field.
emellor@ewan [Tue, 4 Oct 2005 10:28:25 +0000 (11:28 +0100)]
Fix debugging statement that was referring to non-existent field.

Signed-off-by: Ewan Mellor <ewan@xensource.com>
20 years agoAdded some DPRINTK statements, for debugging.
emellor@ewan [Tue, 4 Oct 2005 10:27:37 +0000 (11:27 +0100)]
Added some DPRINTK statements, for debugging.

20 years agoMerged.
emellor@ewan [Tue, 4 Oct 2005 10:15:34 +0000 (11:15 +0100)]
Merged.

20 years agoFix localhost live migration. We were overvigorously wiping out the store
emellor@ewan [Tue, 4 Oct 2005 10:14:50 +0000 (11:14 +0100)]
Fix localhost live migration.  We were overvigorously wiping out the store
entries when a domain closed and on save, which meant that the /vm entries
disappeared when a localhost migration occurred.  XendCheckpoint has had extra
exception handling and logging added.  It also now calls back through
XendDomain.restore_,which has the correct locking semantics to prevent race
conditions during migration.

Added assertions to XendCheckpoint to ensure that the channels are set after
XendDomainInfo.restore.  I don't see why they would not be, and the old code
meant that in the case that they were not, IntroduceDomain would not be called
on the new domain, breaking Xend restart.

relocate calls through XendDomain.domain_restore_fd rather than directly to
XendCheckpoint to isolate XendCheckpoint from the rest of the world, and to
allow XendDomain to pass itself into XendCheckpoint for a callback.

Simplify the XendCheckpoint / XendDomainInfo interlock, giving only two
states, OK and TERMINATED.  If XendCheckpoint asks for a suspend, but sees a
shutdown, it is valid for it to proceed -- either way the domain has stopped.
Higher level tools may wish to disallow this, but at the very least, there is
no sense in waiting for a suspend that will never come.

Signed-off-by: Ewan Mellor <ewan@xensource.com>
20 years agoUse the domain ID to refer to backends, not the UUID. This means that a new
emellor@ewan [Tue, 4 Oct 2005 10:01:38 +0000 (11:01 +0100)]
Use the domain ID to refer to backends, not the UUID.  This means that a new
backend is used on a localhost live migrate; migration was failing otherwise.

Signed-off-by: Ewan Mellor <ewan@xensource.com>
20 years agoExport some more symbols.
kaf24@firebug.cl.cam.ac.uk [Tue, 4 Oct 2005 08:20:48 +0000 (09:20 +0100)]
Export some more symbols.

20 years agoCope with missing configuration information inside the configuration method.
emellor@ewan [Tue, 4 Oct 2005 01:28:41 +0000 (02:28 +0100)]
Cope with missing configuration information inside the configuration method.
This allows us to better recover if the store has been deleted.

Signed-off-by: Ewan Mellor <ewan@xensource.com>
20 years agoMove the import for XendError, and use a bare import, not a from...import.
emellor@ewan [Tue, 4 Oct 2005 01:25:04 +0000 (02:25 +0100)]
Move the import for XendError, and use a bare import, not a from...import.
This should silence the error message when the program is terminated during
startup.

Signed-off-by: Ewan Mellor <ewan@xensource.com>
20 years agoMove XendDomainInfo.{create,recreate,parseConfig} to the top level of the
emellor@ewan [Tue, 4 Oct 2005 01:21:28 +0000 (02:21 +0100)]
Move XendDomainInfo.{create,recreate,parseConfig} to the top level of the
domain.  This allows us to refer to them using an import statement, rather than
a from .. import.  This is a step towards getting rid of the xroot hack.  All
other references to XendDomainInfo methods need to be doubly qualified (once
for the module, once for the class).

Remove XendDomainDict, replacing it with a simple dictionary, folding the
get_by_name method into XendDomain.

Replace XendDomain.refresh_lock with a domains_lock which goes around any
reference to XendDomain.domains or anything that will create or destroy a
domain.  This serialises most accesses through XendDomain, ensuring that we will
not return stale state when racing against the watches fired in separate
threads.  This should have fixed bugs #270 and #234.

Added a number of domain_get_xyz methods.  Those named xyz_nr are to allow
components internal to XendDomain (XendDomainInfo, XendCheckpoint) to call back
into XendDomain without triggering further calls to XendDomain.refresh.  The
other methods simply make it clear which fallback behaviour is expected.

Replace XendDomainInfo.domain_exists with XendDomainInfo.domain_by_name; the
internals of this method needed to change to match those changes above, and it
has been a misnomer for some time.

Signed-off-by: Ewan Mellor <ewan@xensource.com>
20 years agoRemove the reason code from the destroy action -- the reason is only of value
emellor@ewan [Mon, 3 Oct 2005 23:57:34 +0000 (00:57 +0100)]
Remove the reason code from the destroy action -- the reason is only of value
on a graceful shutdown.  Clarify the semantic difference between poweroff and
halt (we use halt to mean 'shutdown and stop' and poweroff to mean 'shutdown
and do whatever the configured behaviour is').

Signed-off-by: Ewan Mellor <ewan@xensource.com>
20 years agoRemove the reason code from the destroy action -- the reason is only of value
emellor@ewan [Mon, 3 Oct 2005 23:56:24 +0000 (00:56 +0100)]
Remove the reason code from the destroy action -- the reason is only of value
on a graceful shutdown.  Clarify the semantic difference between poweroff and
halt (we use halt to mean 'shutdown and stop' and poweroff to mean 'shutdown
and do whatever the configured behaviour is').

Signed-off-by: Ewan Mellor <ewan@xensource.com>
20 years agoRemove the reason parameter from XendDomain.destroy -- it is useless. Accept
emellor@ewan [Mon, 3 Oct 2005 23:30:49 +0000 (00:30 +0100)]
Remove the reason parameter from XendDomain.destroy -- it is useless.  Accept
the halt shutdown reason inside XendDomainInfo.  This indicates that the
domain should be shutdown just like a poweroff, except that the domain is not
restarted, regardless of the on_poweroff configuration setting.

Signed-off-by: Ewan Mellor <ewan@xensource.com>
20 years agoAdded method XendDomain.privilegedDomain and use that inside DevController to
emellor@ewan [Mon, 3 Oct 2005 23:21:49 +0000 (00:21 +0100)]
Added method XendDomain.privilegedDomain and use that inside DevController to
avoid hard-coding the domain ID 0.

Signed-off-by: Ewan Mellor <ewan@xensource.com>
20 years agoPossible fix for the SEDF scheduler assertion problem.
jrb44@plym.cl.cam.ac.uk [Mon, 3 Oct 2005 22:28:24 +0000 (23:28 +0100)]
Possible fix for the SEDF scheduler assertion problem.

Signed-off-by: Stephan Diestelhorst <Stephan.Diestelhorst@{cl.cam.ac.uk, inf.tu-dresden.de}>
20 years agoCreate new vcpu_op() hypercall. Replaces old boot_vcpu()
kaf24@firebug.cl.cam.ac.uk [Mon, 3 Oct 2005 18:14:02 +0000 (19:14 +0100)]
Create new vcpu_op() hypercall. Replaces old boot_vcpu()
hypercall and vcpu-related schedop commands.

Signed-off-by: Keir Fraser <keir@xensource.com>
20 years agoCope with missing configuration information inside the configuration method.
emellor@ewan [Mon, 3 Oct 2005 15:40:27 +0000 (16:40 +0100)]
Cope with missing configuration information inside the configuration method.
This allows us to better recover if the store has been deleted.

Signed-off-by: Ewan Mellor <ewan@xensource.com>
20 years agoRemove the reason code from the destroy action -- the reason is only of value
emellor@ewan [Mon, 3 Oct 2005 15:39:19 +0000 (16:39 +0100)]
Remove the reason code from the destroy action -- the reason is only of value
on a graceful shutdown.  Clarify the semantic difference between poweroff and
halt (we use halt to mean 'shutdown and stop' and poweroff to mean 'shutdown
and do whatever the configured behaviour is').

Signed-off-by: Ewan Mellor <ewan@xensource.com>
20 years agoMove the import for XendError, and use a bare import, not a from...import.
emellor@ewan [Mon, 3 Oct 2005 15:33:09 +0000 (16:33 +0100)]
Move the import for XendError, and use a bare import, not a from...import.
This should silence the error message when the program is terminated during
startup.

Signed-off-by: Ewan Mellor <ewan@xensource.com>
20 years agoAllow adjustment of the size of TPM transfer buffers
kaf24@firebug.cl.cam.ac.uk [Mon, 3 Oct 2005 14:05:37 +0000 (15:05 +0100)]
Allow adjustment of the size of TPM transfer buffers
to the size that a lower-layer driver supports.

Signed-off-by: Stefan Berger <stefanb@us.ibm.com>
20 years agox86/64 was using the plain i386 version of the microcode
kaf24@firebug.cl.cam.ac.uk [Mon, 3 Oct 2005 14:04:27 +0000 (15:04 +0100)]
x86/64 was using the plain i386 version of the microcode
driver instead of the Xen version.

Signed-Off-By: Nitin A Kamble <nitin.a.kamble@intel.com>
20 years agoDo not pickle vcpu state on save/restore. We can construct
kaf24@firebug.cl.cam.ac.uk [Mon, 3 Oct 2005 14:02:23 +0000 (15:02 +0100)]
Do not pickle vcpu state on save/restore. We can construct
suitable initial state for each vcpu during restore, without
need for pickle/unpickle.

Signed-off-by: Keir Fraser <keir@xensource.com>
20 years agoDynamically allocate mmu_update array in
kaf24@firebug.cl.cam.ac.uk [Mon, 3 Oct 2005 09:26:29 +0000 (10:26 +0100)]
Dynamically allocate mmu_update array in
direct_remap_pfn_range(). This allows larger batches and
avoids stack overflow.

Signed-off-by: Keir Fraser <keir@xensource.com>
20 years agoThe stopgap in changeset 7152:a73ab0ddf990 was broken -- revert the change to
emellor@ewan [Sun, 2 Oct 2005 22:56:13 +0000 (23:56 +0100)]
The stopgap in changeset 7152:a73ab0ddf990 was broken -- revert the change to
domain_lookup_by_name, as it causes an infinite loop on startup.  Leave the
domain_lookup change.

Signed-off-by: Ewan Mellor <ewan@xensource.com>
20 years agoremove useless ptep_establish_new function (called set_pte_at now)
vh249@kneesaa.uk.xensource.com [Sun, 2 Oct 2005 21:45:32 +0000 (17:45 -0400)]
remove useless ptep_establish_new function (called set_pte_at now)

Signed-off-by: Vincent Hanquez <vincent@xensource.com>
20 years agoMerged.
emellor@ewan [Sun, 2 Oct 2005 17:24:41 +0000 (18:24 +0100)]
Merged.

20 years agoMove XendVnet from using DBMap etc to using xstransact. This is the last
emellor@ewan [Sun, 2 Oct 2005 17:22:22 +0000 (18:22 +0100)]
Move XendVnet from using DBMap etc to using xstransact.  This is the last
dependency upon xsnode, xsobj, and xsresource, so these can all go.

Signed-off-by: Ewan Mellor <ewan@xensource.com>
20 years agoCall refresh in the domain_lookup and domain_lookup_by_name methods. This
emellor@ewan [Sun, 2 Oct 2005 17:21:01 +0000 (18:21 +0100)]
Call refresh in the domain_lookup and domain_lookup_by_name methods.  This
should reduce the incidence of stale domain information being used by xend, but
is only a stop-gap.

Signed-off-by: Ewan Mellor <ewan@xensource.com>
20 years agoFix call to XendCheckpoint.restore, since that method no longer needs a
emellor@ewan [Sun, 2 Oct 2005 17:17:26 +0000 (18:17 +0100)]
Fix call to XendCheckpoint.restore, since that method no longer needs a
XendDomain instance passed in.

Remove meaningless indirection of imports through web/reactor.py.  Tidy import
statements.

Signed-off-by: Ewan Mellor <ewan@xensource.com>
20 years agoAlso send xenbus notifications using evtchn as handle. It's
kaf24@firebug.cl.cam.ac.uk [Fri, 30 Sep 2005 16:37:52 +0000 (17:37 +0100)]
Also send xenbus notifications using evtchn as handle. It's
always up-to-date as it's reinitialised for us by external
tools during save/restore and migration.

20 years agoxencons must notify via evtchn. It can't use the irq as the
kaf24@firebug.cl.cam.ac.uk [Fri, 30 Sep 2005 16:34:01 +0000 (17:34 +0100)]
xencons must notify via evtchn. It can't use the irq as the
notification handle because console is used early, before the
irq is set up.
Signed-off-by: Keir Fraser <keir@xensource.com>
20 years agoMerge.
emellor@ewan [Fri, 30 Sep 2005 16:10:52 +0000 (17:10 +0100)]
Merge.

20 years agoAdd the config file parsing for the on_{poweroff,reboot,crash} options, so that
emellor@ewan [Fri, 30 Sep 2005 16:10:20 +0000 (17:10 +0100)]
Add the config file parsing for the on_{poweroff,reboot,crash} options, so that
they actually take effect.  Added behaviour "rename-restart" for debugging
purposes, that renames the domain out of the way, preserving it for debugging,
but starts a new domain too.

Add explicit remove of old domain paths when creating a new domain, to avoid
stale information affecting us (by shutting the domain down, for example).

Signed-off-by: Ewan Mellor <ewan@xensource.com>
20 years agoMinor tidy.
emellor@ewan [Fri, 30 Sep 2005 16:05:08 +0000 (17:05 +0100)]
Minor tidy.

Signed-off-by: Ewan Mellor <ewan@xensource.com>
20 years agoFix earlier merge with Ewan's xenstore changes.
kaf24@firebug.cl.cam.ac.uk [Fri, 30 Sep 2005 14:45:20 +0000 (15:45 +0100)]
Fix earlier merge with Ewan's xenstore changes.

20 years agoChange rebooting specification so that we can have domains crash and stay
emellor@ewan [Fri, 30 Sep 2005 12:59:28 +0000 (13:59 +0100)]
Change rebooting specification so that we can have domains crash and stay
around to be debugged, for example.

Fix problem in XendDomainInfo.eventChannel caused by a mis-merge.

Signed-off-by: Ewan Mellor <ewan@xensource.com>
20 years agoUpdate comment/variable names to reflect the fact that the backend paths now
emellor@ewan [Fri, 30 Sep 2005 12:58:08 +0000 (13:58 +0100)]
Update comment/variable names to reflect the fact that the backend paths now
contain the frontend domid, not the uuid.  The actual functionality is the same,
because the ID is referred to by path level, so the change in semantics is not
a problem.

Signed-off-by: Ewan Mellor <ewan@xensource.com>
20 years agoWithin the store, split the persistent information regarding a VM from the
emellor@ewan [Fri, 30 Sep 2005 12:41:10 +0000 (13:41 +0100)]
Within the store, split the persistent information regarding a VM from the
transient information regarding a domain.  This allows live localhost migration,
which is important for testing migration, and also allows migration when we
have a distributed store.  The backend paths in the store now refer to the
frontend domain ID, not its UUID, and blktap has changed to match.

To support this split in the information, the cleanup procedure has been split
to match.

Change the save-restore interface between XendDomain, XendDomainInfo, and
XendCheckpoint, to remove some intermingling, in particular taking XendDomain
out of the loop for the restore procedure.

Improved the recovery procedure to avoid trying to destroy dom0 when recovering.

Added a lock around XendDomain.refresh and one around
XendDomainInfo.refreshShutdown, to improve the behaviour when recreating
domains at startup.  There are still races here, so more thought needs to be
given to the locking.

Live reconfiguration of an existing domain is temporarily broken.

Signed-off-by: Ewan Mellor <ewan@xensource.com>
20 years agoAdded cache-control headers to pages returned by HTTP server so that pages
emellor@ewan [Fri, 30 Sep 2005 10:55:49 +0000 (11:55 +0100)]
Added cache-control headers to pages returned by HTTP server so that pages
expire immediately.  This prevents navigation to stale pages.

Signed-off-by: Ewan Mellor <ewan@xensource.com>
20 years agoRemove the bizarre arrangement whereby EventChannel inherits from dict. Move
emellor@ewan [Fri, 30 Sep 2005 10:40:59 +0000 (11:40 +0100)]
Remove the bizarre arrangement whereby EventChannel inherits from dict.  Move
the channel-creation logic into eventChannel, making the EventChannel class
simpler.  Remove the closeEventChannel method -- it was simply doing a check
for None, and that isn't enough value to justify the extra method when there
is an EventChannel.close method there too.

Signed-off-by: Ewan Mellor <ewan@xensource.com>
20 years agoMerge.
emellor@ewan [Fri, 30 Sep 2005 09:50:19 +0000 (10:50 +0100)]
Merge.

20 years agoAdded DevController.configuration and configurations methods. These parse the
emellor@ewan [Fri, 30 Sep 2005 09:48:49 +0000 (10:48 +0100)]
Added DevController.configuration and configurations methods.  These parse the
store for the current device configuration, and convert it back into the
s-expression used in the config files and the save files.  This means that we
can bring the devices up properly on restore.

Signed-off-by: Ewan Mellor <ewan@xensource.com>
20 years agoFix typos I introduced into tpmfront driver.
kaf24@firebug.cl.cam.ac.uk [Fri, 30 Sep 2005 08:17:07 +0000 (09:17 +0100)]
Fix typos I introduced into tpmfront driver.

20 years agoFix compiler warnings when shadow debugging is enabled.
kaf24@firebug.cl.cam.ac.uk [Thu, 29 Sep 2005 22:44:34 +0000 (23:44 +0100)]
Fix compiler warnings when shadow debugging is enabled.

Fix two classes of warnings when debugging of shadow code is enabled:
  - The type_info field in pfn_info is a long, so use a %lx format
    specifier when printing it.
  - Fix a warning about uninitialized use of eff_guest_pte.

Signed-off-by: Michael Vrable <mvrable@cs.ucsd.edu>
20 years agoAllow kernel built with privileged operation support (e.g. -xen/-xen0) to be used...
iap10@firebug.cl.cam.ac.uk [Thu, 29 Sep 2005 18:59:18 +0000 (19:59 +0100)]
Allow kernel built with privileged operation support (e.g. -xen/-xen0) to be used as a guest kernel on x86_64
Signed-off-by: ian@xensource.com
20 years agoMore save/restore code in interrupts-enabled contest, where
kaf24@firebug.cl.cam.ac.uk [Thu, 29 Sep 2005 17:28:28 +0000 (18:28 +0100)]
More save/restore code in interrupts-enabled contest, where
it properly belongs.
Signed-off-by: Keir Fraser <keir@xensource.com>
20 years agoRemove unused debug code.
kaf24@firebug.cl.cam.ac.uk [Thu, 29 Sep 2005 17:15:49 +0000 (18:15 +0100)]
Remove unused debug code.
Signed-off-by: Keir Fraser <keir@xensource.com>
20 years agoMove xenbus_suspend outside critical region with
kaf24@firebug.cl.cam.ac.uk [Thu, 29 Sep 2005 16:39:33 +0000 (17:39 +0100)]
Move xenbus_suspend outside critical region with
interrupts disabled.
Signed-off-by: Keir Fraser <keir@xensource.com>
20 years agoUpdate GDB server documentation for crashed guests.
kaf24@firebug.cl.cam.ac.uk [Thu, 29 Sep 2005 16:12:48 +0000 (17:12 +0100)]
Update GDB server documentation for crashed guests.
Signed-off-by: Keir Fraser <keir@xensource.com>
20 years agoAvoid verbosity when connection xenbus -> xenstore.
kaf24@firebug.cl.cam.ac.uk [Thu, 29 Sep 2005 16:12:20 +0000 (17:12 +0100)]
Avoid verbosity when connection xenbus -> xenstore.
Signed-off-by: Keir Fraser <keir@xensource.com>
20 years agoDon't define vectors for IPIs we don't use in Xen
iap10@firebug.cl.cam.ac.uk [Thu, 29 Sep 2005 15:57:23 +0000 (16:57 +0100)]
Don't define vectors for IPIs we don't use in Xen
Signed-off-by: ian@xensource.com
20 years agoSlight rejig avoids grant entry warning message on xm save.
kaf24@firebug.cl.cam.ac.uk [Thu, 29 Sep 2005 15:39:32 +0000 (16:39 +0100)]
Slight rejig avoids grant entry warning message on xm save.
Signed-off-by: Keir Fraser <keir@xensource.com>
20 years agofull devices expect to be able to read the partition table during xlvbd_add
vh249@kneesaa.uk.xensource.com [Thu, 29 Sep 2005 15:25:50 +0000 (11:25 -0400)]
full devices expect to be able to read the partition table during xlvbd_add
but couldn't because the device was not marked as BLKIF_CONNECTED and thus
blkif_queue_request discarded the read request. fix bug #220

Signed-off-by: Vincent Hanquez <vincent@xensource.com>
20 years agoDefine new notify_remote_via_irq() interface which is a
kaf24@firebug.cl.cam.ac.uk [Thu, 29 Sep 2005 14:49:54 +0000 (15:49 +0100)]
Define new notify_remote_via_irq() interface which is a
safe version for drivers to use that is aware of
save/restore. Post-restore notifications are silently
dropped, in the expectation that xenbus will eventually
get round to telling the driver it needs to reconnect.

Signed-off-by: Keir Fraser <keir@xensource.com>
20 years agoChange semantics of bind_evtchn_to_xxx and
kaf24@firebug.cl.cam.ac.uk [Thu, 29 Sep 2005 14:14:03 +0000 (15:14 +0100)]
Change semantics of bind_evtchn_to_xxx and
unbind_evtchn_from_xxx. The bind now returns the IRQ
number on success. The unbind takes this as a parameter
instead of the event-channel port. Also, unbind closes
down the underlying event-channel port if it is still
live.

Signed-off-by: Keir Fraser <keir@xensource.com>
20 years agoVIRQs and IPIs on VCPU#0 are automatically re-bound on save/restore.
kaf24@firebug.cl.cam.ac.uk [Thu, 29 Sep 2005 12:05:43 +0000 (13:05 +0100)]
VIRQs and IPIs on VCPU#0 are automatically re-bound on save/restore.
Signed-off-by: Keir Fraser <keir@xensource.com>
20 years agoReindent more xenlinux files. Remove defunct header file.
kaf24@firebug.cl.cam.ac.uk [Thu, 29 Sep 2005 10:10:27 +0000 (11:10 +0100)]
Reindent more xenlinux files. Remove defunct header file.
Signed-off-by: Keir Fraser <keir@xensource.com>
20 years agoSet guest selectors to be null if it's not valid when switch to
kaf24@firebug.cl.cam.ac.uk [Thu, 29 Sep 2005 08:59:46 +0000 (09:59 +0100)]
Set guest selectors to be null if it's not valid when switch to
protected-mode in the guest firmware.

Signed-off-by: Chengyuan Li <chengyuan.li@intel.com>
Signed-off-by: Asit Mallick <asit.k.mallick@intel.com>
20 years agoFix fixed-point long multiplication used in time delta
kaf24@firebug.cl.cam.ac.uk [Wed, 28 Sep 2005 21:18:30 +0000 (22:18 +0100)]
Fix fixed-point long multiplication used in time delta
calculations. The carry flag was clobbered before it was
added to the result.

Signed-off-by: Keir Fraser <keir@xensource.com>
20 years agomerge
kaf24@firebug.cl.cam.ac.uk [Wed, 28 Sep 2005 20:06:49 +0000 (21:06 +0100)]
merge

20 years agoFix gnttab setup, which I broke 2 changesets ago.
kaf24@firebug.cl.cam.ac.uk [Wed, 28 Sep 2005 17:21:09 +0000 (18:21 +0100)]
Fix gnttab setup, which I broke 2 changesets ago.
Signed-off-by: Keir Fraser <keir@xensource.com>
20 years agoThis patch fixes a race between when the disable cpu is marked online
kaf24@firebug.cl.cam.ac.uk [Wed, 28 Sep 2005 17:03:20 +0000 (18:03 +0100)]
This patch fixes a race between when the disable cpu is marked online
and binding IPIs back to the CPU.  In some cases, an IPI would be sent
to CPU1 before it had allocated a new evtchn.  Moving smp_resume() call
before setting the cpu online fixes this race.  This fixes bug #228
(http://bugzilla.xensource.com/bugzilla/show_bug.cgi?id=228).

Signed-off-by: Ryan Harper <ryanh@us.ibm.com>
20 years agoReindent grant-table kernel code and reserve first few
kaf24@firebug.cl.cam.ac.uk [Wed, 28 Sep 2005 17:01:40 +0000 (18:01 +0100)]
Reindent grant-table kernel code and reserve first few
entries for tools use.

Signed-off-by: Keir Fraser <keir@xensource.com>
20 years agoFix netif save/restore.
kaf24@firebug.cl.cam.ac.uk [Wed, 28 Sep 2005 16:51:24 +0000 (17:51 +0100)]
Fix netif save/restore.

Signed-off-by: Keir Fraser <keir@xensource.com>
20 years agoFixes for blkif save/restore. A 'dd if=/dev/sda1 of=/dev/null'
kaf24@firebug.cl.cam.ac.uk [Wed, 28 Sep 2005 16:25:08 +0000 (17:25 +0100)]
Fixes for blkif save/restore. A 'dd if=/dev/sda1 of=/dev/null'
seems to lock up the domU kernel on restore still (can be
pinged but not ssh'ed). This still needs investigation.

Signed-off-by: Keir Fraser <keir@xensource.com>
20 years agoMerge.
emellor@ewan [Wed, 28 Sep 2005 16:02:17 +0000 (17:02 +0100)]
Merge.

20 years agoRemove unused domain_devtype_get.
emellor@ewan [Wed, 28 Sep 2005 16:00:22 +0000 (17:00 +0100)]
Remove unused domain_devtype_get.

Signed-off-by: Ewan Mellor <ewan@xensource.com>
20 years agoRemove unused xend_domain_device.
emellor@ewan [Wed, 28 Sep 2005 15:59:34 +0000 (16:59 +0100)]
Remove unused xend_domain_device.

Signed-off-by: Ewan Mellor <ewan@xensource.com>
20 years agoChange device IDs to be integers, as required now by DevController. Remove
emellor@ewan [Wed, 28 Sep 2005 15:59:13 +0000 (16:59 +0100)]
Change device IDs to be integers, as required now by DevController.  Remove
unused op_device.  Remove broken code to show the domain configuration using
the HTTP server.

Signed-off-by: Ewan Mellor <ewan@xensource.com>
20 years agoRemove broken assertion in 64-bit shadow pagetable code.
kaf24@firebug.cl.cam.ac.uk [Wed, 28 Sep 2005 15:15:20 +0000 (16:15 +0100)]
Remove broken assertion in 64-bit shadow pagetable code.

20 years agoNothing.
emellor@ewan [Wed, 28 Sep 2005 14:09:45 +0000 (15:09 +0100)]
Nothing.

20 years agoMinor tidy up.
emellor@ewan [Wed, 28 Sep 2005 14:09:24 +0000 (15:09 +0100)]
Minor tidy up.

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