xen.git
19 years agoAdded VM_BAD_POWER_STATE error code to VM.hard_shutdown.
Ewan Mellor [Thu, 8 Mar 2007 23:22:56 +0000 (23:22 +0000)]
Added VM_BAD_POWER_STATE error code to VM.hard_shutdown.

Signed-off-by: Ewan Mellor <ewan@xensource.com>
19 years agoRemove crashdump.create -- this doesn't make sense.
Ewan Mellor [Thu, 8 Mar 2007 23:16:34 +0000 (23:16 +0000)]
Remove crashdump.create -- this doesn't make sense.

Signed-off-by: Ewan Mellor <ewan@xensource.com>
19 years agoFix duplicate consts around xc_error declarations.
Ewan Mellor [Thu, 8 Mar 2007 22:33:47 +0000 (22:33 +0000)]
Fix duplicate consts around xc_error declarations.

Signed-off-by: Ewan Mellor <ewan@xensource.com>
19 years agoMerge
Tim Deegan [Thu, 8 Mar 2007 16:41:32 +0000 (16:41 +0000)]
Merge

19 years ago[HVM] Save/restore: tidy PFN handling in the tools.
Tim Deegan [Thu, 8 Mar 2007 16:40:35 +0000 (16:40 +0000)]
[HVM] Save/restore: tidy PFN handling in the tools.
Make the differences clearer between the number of frames, the max PFN
and the size of the pfn array, and handle te framebuffer more cleanly.
Signed-off-by: Tim Deegan <Tim.Deegan@xensource.com>
19 years ago[HVM] Save/restore: make sure all ioreqs are gone when a domain is saved
Tim Deegan [Thu, 8 Mar 2007 16:38:15 +0000 (16:38 +0000)]
[HVM] Save/restore: make sure all ioreqs are gone when a domain is saved
by pausing the domain and pulling them through.
Signed-off-by: Tim Deegan <Tim.Deegan@xensource.com>
19 years agolinux: Remove an unused data structure from TPM frontend.
kfraser@localhost.localdomain [Thu, 8 Mar 2007 16:26:17 +0000 (16:26 +0000)]
linux: Remove an unused data structure from TPM frontend.
Some more code style changes.

Signed-off-by: Stefan Berger <stefanb@us.ibm.com>
19 years ago[XEN] Shadow: fix mark_dirty invocation to use gfn, not mfn
Tim Deegan [Thu, 8 Mar 2007 16:20:01 +0000 (16:20 +0000)]
[XEN] Shadow: fix mark_dirty invocation to use gfn, not mfn
Signed-off-by: Tim Deegan <Tim.Deegan@xensource.com>
19 years ago[XEN] Shadow: tidy up page_info before handing back to the allocator
Tim Deegan [Thu, 8 Mar 2007 16:18:50 +0000 (16:18 +0000)]
[XEN] Shadow: tidy up page_info before handing back to the allocator
which expects count_info fields to be zero.
Signed-off-by: Tim Deegan <Tim.Deegan@xensource.com>
19 years ago[xen] Update the hvm rtc clock state if the domain's time offset is changed.
Christian Limpach [Thu, 8 Mar 2007 15:51:32 +0000 (15:51 +0000)]
[xen] Update the hvm rtc clock state if the domain's time offset is changed.
The settimeoffset domctl had no effect on hvm domains since the offset
was only read in rtc_init which is done when the domain is created.

Signed-off-by: Christian Limpach <Christian.Limpach@xensource.com>
19 years ago[BUILDER]: Initialize code segment selectors in virtual IDT to 0.
Ian Campbell [Thu, 8 Mar 2007 15:35:28 +0000 (15:35 +0000)]
[BUILDER]: Initialize code segment selectors in virtual IDT to 0.
These should not be set until the guest kernel kernel configures an
entry point otherwise the null_trap_bounce() check gets confused.

This change was made to the old domain builder in 12455:3fa6635d04b9
but was lost in the transition to the new builder.

Signed-off-by: Ian Campbell <ian.campbell@xensource.com>
19 years agoxen: Provide notification of console updates via VIRQ.
kfraser@localhost.localdomain [Thu, 8 Mar 2007 15:21:10 +0000 (15:21 +0000)]
xen: Provide notification of console updates via VIRQ.

The readconsolering capabilities provide the opportunity to
provide console output to other clients (remote systems,
logging systems, etc). This patchs adds the ability to generate
a notification of a change in the console buffer.

Signed-off-by: Ben Thomas <ben@virtualiron.com>
Signed-off-by: Keir Fraser <keir@xensource.com>
19 years agoxend: Platform feature 'multiprocessor suspend' is x86-only for now.
kfraser@localhost.localdomain [Thu, 8 Mar 2007 14:57:33 +0000 (14:57 +0000)]
xend: Platform feature 'multiprocessor suspend' is x86-only for now.
Signed-off-by: Keir Fraser <keir@xensource.com>
19 years ago[HVM] Save/restore: back-port IDE controller save routine from current qemu.
Tim Deegan [Thu, 8 Mar 2007 14:37:21 +0000 (14:37 +0000)]
[HVM] Save/restore: back-port IDE controller save routine from current qemu.
Also, send an interrupt and an error code to the guest on restore if we
dropped an active request when we saved.
Signed-off-by: Tim Deegan <Tim.Deegan@xensource.com>
19 years agoxen: Clean up heap allocator.
kfraser@localhost.localdomain [Thu, 8 Mar 2007 14:29:09 +0000 (14:29 +0000)]
xen: Clean up heap allocator.
Move some common free/alloc code into core heap-allocator
functions. Make it clear that alloc/free can only ever be done outside
hardirq context (previously it was unsafe: could deadlock on heap_lock).
Signed-off-by: Keir Fraser <keir@xensource.com>
19 years ago[HVM] Add support for hardware-assisted paging
Tim Deegan [Thu, 8 Mar 2007 10:54:56 +0000 (10:54 +0000)]
[HVM] Add support for hardware-assisted paging
as the second implementation of the generic paging-assistance interface.

Signed-off-by: Wei Huang <wei.huang2@amd.com>
19 years agolinux: Restore time on APs if they are not hotplugged on save/restore.
kfraser@localhost.localdomain [Wed, 7 Mar 2007 16:21:21 +0000 (16:21 +0000)]
linux: Restore time on APs if they are not hotplugged on save/restore.
Signed-off-by: Keir Fraser <keir@xensource.com>
19 years agonetback: Disable debug interrupt (accidentally left enabled in
kfraser@localhost.localdomain [Wed, 7 Mar 2007 16:16:09 +0000 (16:16 +0000)]
netback: Disable debug interrupt (accidentally left enabled in
previous checkin).
Signed-off-by: Keir Fraser <keir@xensource.com>
19 years agolinux: Use fake carrier flag for netfront/netback rather than the real
kfraser@localhost.localdomain [Wed, 7 Mar 2007 16:00:47 +0000 (16:00 +0000)]
linux: Use fake carrier flag for netfront/netback rather than the real
netif_carrier_XXX() functions. This makes network bringup much faster.
Signed-off-by: Keir Fraser <keir@xensource.com>
19 years ago[HVM] Disallow soft resets in the BIOS.
Tim Deegan [Wed, 7 Mar 2007 13:07:12 +0000 (13:07 +0000)]
[HVM] Disallow soft resets in the BIOS.
They aren't safe if the guest has brought up PV device drivers.
Signed-off-by: Tim Deegan <Tim.Deegan@xensource.com>
19 years agoxen: Replace stupid page_alloc fix.
kfraser@localhost.localdomain [Wed, 7 Mar 2007 11:17:03 +0000 (11:17 +0000)]
xen: Replace stupid page_alloc fix.

I broke the 'correct' fix when I copied it out of an email. The
actual correct version, with an extra +1 in the for-loop header is
rather abusive of for loops, so I've changed it now to a do-while loop
and an extra comment so I don't screw up this backwards loop ever
again.

This version does actually boot. :-)

Signed-off-by: Keir Fraser <keir@xensource.com>
19 years agodump-core: always use elfclass64 format to support x86_32pae.
kfraser@localhost.localdomain [Wed, 7 Mar 2007 10:53:34 +0000 (10:53 +0000)]
dump-core: always use elfclass64 format to support x86_32pae.
Without this patch xen dump-core format can't handle over 4GB case.

Signed-off-by: Isaku Yamahata <yamahata@valinux.co.jp>
19 years agotools: Clean up use of 'install' command.
kfraser@localhost.localdomain [Wed, 7 Mar 2007 10:49:20 +0000 (10:49 +0000)]
tools: Clean up use of 'install' command.

- convert raw "install" command to use $(INSTALL)
- convert some $(INSTALL) to $(INSTALL_DATA) as appropriate
- modify the specific $(INSTALL) definitions to use -p.

Original patch by Ben Thomas <ben@virtualiron.com>
Signed-off-by: Keir Fraser <keir@xensource.com>
19 years agoioemu: Remove use of 'echo -c' in script.
kfraser@localhost.localdomain [Wed, 7 Mar 2007 10:37:05 +0000 (10:37 +0000)]
ioemu: Remove use of 'echo -c' in script.

I've no idea what it does, and it doesn't appear to be valid on some
systems.  Use a more generic approach.

Signed-off-by: Ben Thomas <ben@virtualiron.com>
19 years agolinux: Don't use set_fixmap/clear_fixmap after boot-time. It's not
kfraser@localhost.localdomain [Wed, 7 Mar 2007 10:13:39 +0000 (10:13 +0000)]
linux: Don't use set_fixmap/clear_fixmap after boot-time. It's not
allowed and the functions are marked __init on x86/64.
Use HYPERVISOR_update_va_mapping() directly instead.
Signed-off-by: Keir Fraser <keir@xensource.com>
19 years agoA correct fix for page_alloc crash when Xen heap empty.
kfraser@localhost.localdomain [Wed, 7 Mar 2007 09:43:36 +0000 (09:43 +0000)]
A correct fix for page_alloc crash when Xen heap empty.
Signed-off-by: Keir Fraser <keir@xensource.com>
19 years agoioemu: Close file descriptors when forking processes.
kfraser@localhost.localdomain [Tue, 6 Mar 2007 16:51:42 +0000 (16:51 +0000)]
ioemu: Close file descriptors when forking processes.
Just leave stdin/stdout/stderr (and possibly tap fd) open.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
19 years agoMake Python tools (i.e., xend and pygrub) conditionally compiled.
kfraser@localhost.localdomain [Tue, 6 Mar 2007 16:46:20 +0000 (16:46 +0000)]
Make Python tools (i.e., xend and pygrub) conditionally compiled.

Config option is PYTHON_TOOLS, default is 'y'.

Signed-off-by: Ben Thomas <ben@virtualiron.com>
19 years ago[XEN] Shadow: ignore shadow operations on dying domains
Tim Deegan [Tue, 6 Mar 2007 15:02:20 +0000 (15:02 +0000)]
[XEN] Shadow: ignore shadow operations on dying domains
so that they don't get confused by the way that shadow_teardown()
leaves some of the mode bits set after releasing the shadow resources.
Signed-off-by: Tim Deegan <Tim.Deegan@xensource.com>
19 years agolinux x86/64: No need to pre-allocate all fixmap pagetables:
kfraser@localhost.localdomain [Tue, 6 Mar 2007 13:41:05 +0000 (13:41 +0000)]
linux x86/64: No need to pre-allocate all fixmap pagetables:
just those for bt_ioremap().
Signed-off-by: Keir Fraser <keir@xensource.com>
19 years agoxend: more flexibility in gathering data from /proc/cpuinfo
kfraser@localhost.localdomain [Tue, 6 Mar 2007 11:53:37 +0000 (11:53 +0000)]
xend: more flexibility in gathering data from /proc/cpuinfo

Add ia64 mapping for cpuinfo data and make the case where #vcpu !=
#pcpu more flexible (don't rely on cpu0 existing).

Signed-off-by: Alex Williamson <alex.williamson@hp.com>
19 years ago[SVM] Clear VMCB's EFER.LME when guest disables paging
Travis Betak [Mon, 5 Mar 2007 22:22:21 +0000 (16:22 -0600)]
[SVM] Clear VMCB's EFER.LME when guest disables paging

Since the guest's CR0.PG is always set (in shadow paging), EFER.LME must
be cleared along with EFER.LMA when the guest is disabling paging.

Signed-off-by: Travis Betak <travis.betak@amd.com>
19 years ago[SVM] Use svm_long_mode_enabled() when checking if guest is in long mode
Travis Betak [Mon, 5 Mar 2007 22:21:50 +0000 (16:21 -0600)]
[SVM] Use svm_long_mode_enabled() when checking if guest is in long mode

This patch fixes the inconsistent method of checking if the guest is in
long mode (EFER.LMA == 1).

Signed-off-by: Travis Betak <travis.betak@amd.com>
19 years ago[SVM] Cleanup guest writes to EFER
Travis Betak [Mon, 5 Mar 2007 22:21:11 +0000 (16:21 -0600)]
[SVM] Cleanup guest writes to EFER

19 years ago[SVM] Return the shadow EFER value on guest EFER reads.
Travis Betak [Mon, 5 Mar 2007 22:18:10 +0000 (16:18 -0600)]
[SVM] Return the shadow EFER value on guest EFER reads.

Return the shadow of the guest's EFER instead of the VMCB.  This helps
with things such as the unlikely event the guest wants to check EFER.LME
immediately after setting it (The VMCB's EFER.LME remains clear until
the guest enables paging).

Signed-off-by: Travis Betak <travis.betak@amd.com>
19 years ago[SVM] Remove SVM cpu_state and add EFER shadow to svm structure.
Travis Betak [Mon, 5 Mar 2007 22:17:25 +0000 (16:17 -0600)]
[SVM] Remove SVM cpu_state and add EFER shadow to svm structure.

The SVM cpu state was being used inconsistently and, besides, it makes
more sense to instead rely on the guest's shadow registers to determine
the current guest state.

Signed-off-by: Travis Betak <travis.betak@amd.com>
19 years ago[tools] Import ptsname from its new location (after cset 14259:7f0d8e3c538e).
Christian Limpach [Mon, 5 Mar 2007 22:42:13 +0000 (22:42 +0000)]
[tools] Import ptsname from its new location (after cset 14259:7f0d8e3c538e).

Signed-off-by: Christian Limpach <Christian.Limpach@xensource.com>
19 years ago[tools] Use auxbin.py to determine libpath at build time.
Christian Limpach [Mon, 5 Mar 2007 19:22:21 +0000 (19:22 +0000)]
[tools] Use auxbin.py to determine libpath at build time.

Signed-off-by: Christian Limpach <Christian.Limpach@xensource.com>
19 years agoInstall pygrub according to auxbin rules.
kfraser@localhost.localdomain [Mon, 5 Mar 2007 19:01:19 +0000 (19:01 +0000)]
Install pygrub according to auxbin rules.
Signed-off-by: Keir Fraser <keir@xensource.com>
19 years agoMove ptsname module under tools/python.
kfraser@localhost.localdomain [Mon, 5 Mar 2007 18:54:26 +0000 (18:54 +0000)]
Move ptsname module under tools/python.
Signed-off-by: Keir Fraser <keir@xensource.com>
19 years agolinux x86/64: Clean up mm initialisation a little.
kfraser@localhost.localdomain [Mon, 5 Mar 2007 18:10:41 +0000 (18:10 +0000)]
linux x86/64: Clean up mm initialisation a little.
Signed-off-by: Keir Fraser <keir@xensource.com>
19 years ago[tools] Use auxbin.py to determine libpath at build time.
Christian Limpach [Mon, 5 Mar 2007 16:43:12 +0000 (16:43 +0000)]
[tools] Use auxbin.py to determine libpath at build time.

Signed-off-by: Christian Limpach <Christian.Limpach@xensource.com>
19 years agolinux: no pc speaker device in domUs
kfraser@localhost.localdomain [Mon, 5 Mar 2007 15:53:18 +0000 (15:53 +0000)]
linux: no pc speaker device in domUs
Signed-off-by: Jan Beulich <jbeulich@novell.com>
19 years agolinux: scan DMI early
kfraser@localhost.localdomain [Mon, 5 Mar 2007 15:46:33 +0000 (15:46 +0000)]
linux: scan DMI early

While shuffling quite a few things around, this gets us closer to
native, which clearly had a reason to do the DMI scan early.

Signed-off-by: Jan Beulich <jbeulich@novell.com>
19 years agolinux/x86-64 2.6.18: reserve bootmem differently
kfraser@localhost.localdomain [Mon, 5 Mar 2007 15:08:17 +0000 (15:08 +0000)]
linux/x86-64 2.6.18: reserve bootmem differently

Do init memory reservation in a single step rather then piece-wise.
This is not really addressing a problem, but simplifies things a
little.

Signed-off-by: Jan Beulich <jbeulich@novell.com>
19 years agolinux/x86-64 2.6.18: do away with pud_offset_k()
kfraser@localhost.localdomain [Mon, 5 Mar 2007 15:06:13 +0000 (15:06 +0000)]
linux/x86-64 2.6.18: do away with pud_offset_k()
Signed-off-by: Jan Beulich <jbeulich@novell.com>
19 years agolinux/x86_64 2.6.18: remove orig_ist
kfraser@localhost.localdomain [Mon, 5 Mar 2007 15:04:28 +0000 (15:04 +0000)]
linux/x86_64 2.6.18: remove orig_ist

orig_ist is bound to presence of a TSS, hence is meaningless on Xen
kernels.

Signed-off-by: Jan Beulich <jbeulich@novell.com>
19 years agoFix crash in heap allocator when Xen 'zone' is empty.
kfraser@localhost.localdomain [Mon, 5 Mar 2007 14:39:03 +0000 (14:39 +0000)]
Fix crash in heap allocator when Xen 'zone' is empty.
Signed-off-by: Keir Fraser <keir@xensource.com>
19 years agolinux: Prefer testing io_bitmap_ptr to the TIF_IO_BITMAP.
kfraser@localhost.localdomain [Mon, 5 Mar 2007 14:20:26 +0000 (14:20 +0000)]
linux: Prefer testing io_bitmap_ptr to the TIF_IO_BITMAP.
Former is safer (in case TIF_IO_BITMAP is out-of-sync with actual
existence of an I/O bitmap for this process) and more readable, and
probably results in faster code too.
Signed-off-by: Keir Fraser <keir@xensource.com>
19 years agolinux: Remove unnecessary locking from pcifront free_root_bus_devs().
kfraser@localhost.localdomain [Mon, 5 Mar 2007 14:09:19 +0000 (14:09 +0000)]
linux: Remove unnecessary locking from pcifront free_root_bus_devs().
Signed-off-by: Keir Fraser <keir@xensource.com>
19 years ago[LINUX] CONFIG_X86_TSC is defined upstream in arch/i386/Kconfig.cpu so use that
Ian Campbell [Mon, 5 Mar 2007 14:04:45 +0000 (14:04 +0000)]
[LINUX] CONFIG_X86_TSC is defined upstream in arch/i386/Kconfig.cpu so use that
instead of duplicating it in arch/i386/Kconfig.

Signed-off-by: Ian Campbell <ian.campbell@xensource.com>
19 years agolinux 2.6.18: constify instances of 'struct file_operations'
kfraser@localhost.localdomain [Mon, 5 Mar 2007 14:00:05 +0000 (14:00 +0000)]
linux 2.6.18: constify instances of 'struct file_operations'
Signed-off-by: Jan Beulich <jbeulich@novell.com>
19 years agolinux 2.6.18: miscellaneous
kfraser@localhost.localdomain [Mon, 5 Mar 2007 13:57:11 +0000 (13:57 +0000)]
linux 2.6.18: miscellaneous

Various minor changes not fitting elsewhere and not warranting
separate patches.

Signed-off-by: Jan Beulich <jbeulich@novell.com>
19 years agolinux 2.6.18: assembly annotations
kfraser@localhost.localdomain [Mon, 5 Mar 2007 13:49:11 +0000 (13:49 +0000)]
linux 2.6.18: assembly annotations

This adds (or moves to the proper place) annotations paralleling those
added in native code in 2.6.18. To make this look less ugly, it
required moving around a few things.

Signed-off-by: Jan Beulich <jbeulich@novell.com>
19 years agolinux 2.6.18: eliminate Xen special: x8664_ksyms.c
kfraser@localhost.localdomain [Mon, 5 Mar 2007 13:33:42 +0000 (13:33 +0000)]
linux 2.6.18: eliminate Xen special: x8664_ksyms.c

The differences of arch/x86_64/kernel/x8664_ksyms-xen.c over its
native original are benign - the Xen special can therefore be
removed.

Signed-off-by: Jan Beulich <jbeulich@novell.com>
19 years agolinux 2.6.18: eliminate Xen special: alternative.c
kfraser@localhost.localdomain [Mon, 5 Mar 2007 13:32:50 +0000 (13:32 +0000)]
linux 2.6.18: eliminate Xen special: alternative.c

The sole (x86-64 specific) difference of
arch/i386/kernel/alternative-xen.c
over its native original can be eliminated by slightly adjusting
x86-64's make_readonly(), and the Xen special can then be removed.

Signed-off-by: Jan Beulich <jbeulich@novell.com>
19 years agolinux 2.6.18: eliminate Xen special: intel_cacheinfo.c
kfraser@localhost.localdomain [Mon, 5 Mar 2007 13:30:52 +0000 (13:30 +0000)]
linux 2.6.18: eliminate Xen special: intel_cacheinfo.c

arch/i386/kernel/cpu/intel_cacheinfo-xen.c is identical to
arch/i386/kernel/cpu/intel_cacheinfo.c, hence the Xen special can be
removed.

Signed-off-by: Jan Beulich <jbeulich@novell.com>
19 years agolinux/x86-64: allow kernel init memory to be freed.
kfraser@localhost.localdomain [Mon, 5 Mar 2007 13:27:51 +0000 (13:27 +0000)]
linux/x86-64: allow kernel init memory to be freed.
Signed-off-by: Jan Beulich <jbeulich@novell.com>
19 years agoVM.actions_after_[shutdown|reboot|suspend|crash] were not set properly
kfraser@localhost.localdomain [Mon, 5 Mar 2007 13:18:28 +0000 (13:18 +0000)]
VM.actions_after_[shutdown|reboot|suspend|crash] were not set properly
in VM record returned via Xen API.
This patch fixes typos in key name for these fields.

Signed-off-by: Jim Fehlig <jfehlig@novell.com>
19 years ago[XEN] Shadow: fix predicate for when to use up-pointers
Tim Deegan [Mon, 5 Mar 2007 11:54:55 +0000 (11:54 +0000)]
[XEN] Shadow: fix predicate for when to use up-pointers
which was broken by the introduction of "pinnable" types
Signed-off-by: Tim Deegan <Tim.Deegan@xensource.com>
19 years agoxen: Remove bogus file.
kfraser@localhost.localdomain [Mon, 5 Mar 2007 11:23:34 +0000 (11:23 +0000)]
xen: Remove bogus file.
Signed-off-by: Keir Fraser <keir@xensource.com>
19 years agoMerge with xenppc-unstable.hg
kfraser@localhost.localdomain [Mon, 5 Mar 2007 11:18:55 +0000 (11:18 +0000)]
Merge with xenppc-unstable.hg

19 years ago[xen] Fix deferred mmu op logic on invalidate shadow ldt.
Christian Limpach [Mon, 5 Mar 2007 11:17:52 +0000 (11:17 +0000)]
[xen] Fix deferred mmu op logic on invalidate shadow ldt.
On domain destroy, put_page_from_l1e is called on a domain
which is not current.

Signed-off-by: Christian Limpach <Christian.Limpach@xensource.com>
19 years agoMerge with xenppc-unstable.hg
kfraser@localhost.localdomain [Mon, 5 Mar 2007 11:14:15 +0000 (11:14 +0000)]
Merge with xenppc-unstable.hg

19 years agoxen: Make dom0 domain builder behaviour match that of domU.
kfraser@localhost.localdomain [Mon, 5 Mar 2007 10:52:54 +0000 (10:52 +0000)]
xen: Make dom0 domain builder behaviour match that of domU.
Signed-off-by: Keir Fraser <keir@xensource.com>
19 years agoxen: Fix ACPI SLIT/SRAT parsing to correctly map highmem.
kfraser@localhost.localdomain [Mon, 5 Mar 2007 10:37:01 +0000 (10:37 +0000)]
xen: Fix ACPI SLIT/SRAT parsing to correctly map highmem.
Signed-off-by: Keir Fraser <keir@xensource.com>
19 years agolinux: Clean up mm_pin/unpin.
kfraser@localhost.localdomain [Mon, 5 Mar 2007 10:26:45 +0000 (10:26 +0000)]
linux: Clean up mm_pin/unpin.
Signed-off-by: Keir Fraser <keir@xensource.com>
19 years ago[domain builder] Remove unnecessary alignment of virtual address base.
Christian Limpach [Sat, 3 Mar 2007 20:42:38 +0000 (20:42 +0000)]
[domain builder] Remove unnecessary alignment of virtual address base.
Unbreaks loading NetBSD kernels with the new ^^ domain builder.

Signed-off-by: Christian Limpach <Christian.Limpach@xensource.com>
19 years agolinux: Some save/restore simplifications and cleanups.
Keir Fraser [Fri, 2 Mar 2007 23:01:36 +0000 (23:01 +0000)]
linux: Some save/restore simplifications and cleanups.
Signed-off-by: Keir Fraser <keir@xensource.com>
19 years agox86/64: Save/restore user pagetable pointer for x86/64 PV guests.
Keir Fraser [Fri, 2 Mar 2007 22:57:27 +0000 (22:57 +0000)]
x86/64: Save/restore user pagetable pointer for x86/64 PV guests.
Signed-off-by: Keir Fraser <keir@xensource.com>
19 years agoFix -xen build.
kfraser@localhost.localdomain [Fri, 2 Mar 2007 16:57:24 +0000 (16:57 +0000)]
Fix -xen build.
Signed-off-by: Keir Fraser <keir@xensource.com>
19 years agolinux: xenbus dev write function resets packet data on *all* error conditions.
kfraser@localhost.localdomain [Fri, 2 Mar 2007 16:29:48 +0000 (16:29 +0000)]
linux: xenbus dev write function resets packet data on *all* error conditions.
Signed-off-by: Keir Fraser <keir@xensource.com>
19 years agolinux: Fix xenbus device write function.
kfraser@localhost.localdomain [Fri, 2 Mar 2007 16:26:56 +0000 (16:26 +0000)]
linux: Fix xenbus device write function.
Signed-off-by: Keir Fraser <keir@xensource.com>
19 years agotpm: Code style cleanups.
kfraser@localhost.localdomain [Fri, 2 Mar 2007 16:15:28 +0000 (16:15 +0000)]
tpm: Code style cleanups.
Signed-off-by: Keir Fraser <keir@xensource.com>
19 years ago[Linux] [TPM driver] Resume frontend after checkpointing
kfraser@localhost.localdomain [Fri, 2 Mar 2007 16:03:21 +0000 (16:03 +0000)]
[Linux] [TPM driver] Resume frontend after checkpointing

This patch implements TPM frontend driver resume after checkpointing a
domain. I also now let the front-end wait for the response from the
TPM until the response is there. A response must come due to the
strict request/response nature of the protocol, but the time until the
responses arrives depends on the type of operation that is executed by
the vTPM. Resending a request is not possible with this type of device.

Signed-off-by: Stefan Berger <stefanb@us.ibm.com>
19 years agolinux: Use old-mode of save/restore for uniproc guests (it's faster).
kfraser@localhost.localdomain [Fri, 2 Mar 2007 16:00:22 +0000 (16:00 +0000)]
linux: Use old-mode of save/restore for uniproc guests (it's faster).
Signed-off-by: Keir Fraser <keir@xensource.com>
19 years agolinux: allow use of split page table locks
kfraser@localhost.localdomain [Fri, 2 Mar 2007 15:59:12 +0000 (15:59 +0000)]
linux: allow use of split page table locks

This changes netback to use struct page's mapping field instead of the
index one, and the PageForeign logic the other way around. That way,
CONFIG_SPLIT_PTLOCK_CPUS doesn't have to be suppressed anymore for x86.

Signed-off-by: Jan Beulich <jbeulich@novell.com>
19 years ago[PVFB][TOOLS] Terminate backend when frontend vanishes early
kfraser@localhost.localdomain [Fri, 2 Mar 2007 15:32:42 +0000 (15:32 +0000)]
[PVFB][TOOLS] Terminate backend when frontend vanishes early

Ensure the backend terminates when the frontend vanishes before the
connection is established.  This happens when the guest doesn't have
the devices.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
19 years agolinux/x86: avoid a re-definition warning on x86-64 and change i386
kfraser@localhost.localdomain [Fri, 2 Mar 2007 15:12:54 +0000 (15:12 +0000)]
linux/x86: avoid a re-definition warning on x86-64 and change i386
similarly to be consistent.

Signed-off-by: Jan Beulich <jbeulich@novell.com>
19 years ago.S->.o generation should depend on asm-offsets.h
kfraser@localhost.localdomain [Fri, 2 Mar 2007 15:10:54 +0000 (15:10 +0000)]
.S->.o generation should depend on asm-offsets.h

While object files generated from C sources should generally not
depend on asm-offsets.h, those generated from assembly sources
should.

Signed-off-by: Jan Beulich <jbeulich@novell.com>
19 years agoxend: Fix startup when dom0 does not see all physical cpus.
kfraser@localhost.localdomain [Fri, 2 Mar 2007 15:08:26 +0000 (15:08 +0000)]
xend: Fix startup when dom0 does not see all physical cpus.
Signed-off-by: Keir Fraser <keir@xensource.com>
19 years agoxend: Catch VmError exception when creating a domain and do
kfraser@localhost.localdomain [Fri, 2 Mar 2007 14:57:18 +0000 (14:57 +0000)]
xend: Catch VmError exception when creating a domain and do
appropriate cleanup.
Signed-off-by: Masaki Kanno <kanno.masaki@jp.fujitsu.com>
19 years agoacm: Fix bogus domain_rcu_lock() invocations to rcu_unlock_domain().
kfraser@localhost.localdomain [Fri, 2 Mar 2007 14:42:25 +0000 (14:42 +0000)]
acm: Fix bogus domain_rcu_lock() invocations to rcu_unlock_domain().
Signed-off-by: Stefan Berger <stefanb@us.ibm.com>
19 years ago[XEN] Remove a race condition in shadow mark_dirty code.
Tim Deegan [Fri, 2 Mar 2007 13:46:39 +0000 (13:46 +0000)]
[XEN] Remove a race condition in shadow mark_dirty code.
Signed-off-by: Tim Deegan <Tim.Deegan@xensource.com>
19 years agolinux: Support new 'fast suspend' mode which does not require us to
kfraser@localhost.localdomain [Fri, 2 Mar 2007 12:11:52 +0000 (12:11 +0000)]
linux: Support new 'fast suspend' mode which does not require us to
hotplug all auxiliary CPUs.
Signed-off-by: Keir Fraser <keir@xensource.com>
19 years agotools: Advertise save/restore multiple VCPU contexts capability.
kfraser@localhost.localdomain [Fri, 2 Mar 2007 12:11:10 +0000 (12:11 +0000)]
tools: Advertise save/restore multiple VCPU contexts capability.
This capability means that save/restore will remember contexts of all
online VCPUs at the time of final guest suspend. Any initialised but
offline VCPUs will not be saved!
Signed-off-by: Keir Fraser <keir@xensource.com>
19 years ago[xenstored] Fix error path when xc_map_foreign_range fails.
Christian Limpach [Fri, 2 Mar 2007 11:25:12 +0000 (11:25 +0000)]
[xenstored] Fix error path when xc_map_foreign_range fails.
If xc_map_foreign_range fails, the new domain object is freed,
but it is not removed from the connection list.

Signed-off-by: Christian Limpach <Christian.Limpach@xensource.com>
19 years agoSave/restore context of all online VCPUs in PV save/restore/migrate.
kfraser@localhost.localdomain [Thu, 1 Mar 2007 17:59:17 +0000 (17:59 +0000)]
Save/restore context of all online VCPUs in PV save/restore/migrate.
Signed-off-by: Keir Fraser <keir@xensource.com>
19 years ago[tools] Set guest address size to be the same as dom0 address size on restore.
Christian Limpach [Thu, 1 Mar 2007 17:28:31 +0000 (17:28 +0000)]
[tools] Set guest address size to be the same as dom0 address size on restore.

Signed-off-by: Christian Limpach <Christian.Limpach@xensource.com>
19 years ago[XEN] Check that the cr3 mfn is valid before using it.
Christian Limpach [Thu, 1 Mar 2007 17:27:31 +0000 (17:27 +0000)]
[XEN] Check that the cr3 mfn is valid before using it.

Signed-off-by: Christian Limpach <Christian.Limpach@xensource.com>
19 years agolinux: No need to kmem_cache_shrink(pgd_cache) on save/restore.
kfraser@localhost.localdomain [Thu, 1 Mar 2007 15:56:45 +0000 (15:56 +0000)]
linux: No need to kmem_cache_shrink(pgd_cache) on save/restore.
Signed-off-by: Keir Fraser <keir@xensource.com>
19 years agoFully reset the xenstore connection when a domain is (re)introduced to xenstored.
kfraser@localhost.localdomain [Thu, 1 Mar 2007 14:51:12 +0000 (14:51 +0000)]
Fully reset the xenstore connection when a domain is (re)introduced to xenstored.
Signed-off-by: Keir Fraser <keir@xensource.com>
19 years ago[HVM] Save/restore: remove bogus "valid state" check
Tim Deegan [Thu, 1 Mar 2007 14:00:05 +0000 (14:00 +0000)]
[HVM] Save/restore: remove bogus "valid state" check
Signed-off-by: Tim Deegan <Tim.Deegan@xensource.com>
19 years agohvm: Clean up console-information passing via xenstore.
kfraser@localhost.localdomain [Thu, 1 Mar 2007 13:57:25 +0000 (13:57 +0000)]
hvm: Clean up console-information passing via xenstore.

Each serial, parallel and monitor device in qemu that is connected to
a pty creates a xenstore node of the form:
 <domain-path>/monitor/tty
 <domain-path>/serial/<n>/tty
 <domain-path>/parallel/<n>/tty

In addition, serial/0 (com1) also registers its information at:
 <domain-path>/console/tty

Also fix a realloc() failure memory leak.

Signed-off-by: Ben Thomas <ben@virtualiron.com>
19 years agoFix Xend domain_lookup error report when looking up an invalid
kfraser@localhost.localdomain [Thu, 1 Mar 2007 13:48:31 +0000 (13:48 +0000)]
Fix Xend domain_lookup error report when looking up an invalid
domain: raise the correct error code (XendInvalidDomain instead of
XendError). Also remove redundance exception code that is handled by
domain_lookup.

Signed-off-by: Yung Giang <ygiang@gmail.com>
19 years agoacm: Fix the traversal of the event channel buckets and use the active
kfraser@localhost.localdomain [Thu, 1 Mar 2007 13:45:53 +0000 (13:45 +0000)]
acm: Fix the traversal of the event channel buckets and use the active
grant table entries instead of the shared ones. I had to move some
functions from grant_table.c into grant_table.h to make them usable by
the ACM module.

Signed-off-by: Stefan Berger <stefanb@us.ibm.com>
19 years agoCreate P2M map during Mini-OS boot (fixes 'xm dump-core').
kfraser@localhost.localdomain [Thu, 1 Mar 2007 12:23:44 +0000 (12:23 +0000)]
Create P2M map during Mini-OS boot (fixes 'xm dump-core').
Signed-off-by: Grzegorz Milos <gm281@cam.ac.uk>
19 years agoFix PAE Mini-OS build.
kfraser@localhost.localdomain [Thu, 1 Mar 2007 12:23:04 +0000 (12:23 +0000)]
Fix PAE Mini-OS build.
Signed-off-by: Grzegorz Milos <gm281@cam.ac.uk>
19 years agox86: No need to take full reference when doing mmu_update on a foreign
kfraser@localhost.localdomain [Thu, 1 Mar 2007 12:14:53 +0000 (12:14 +0000)]
x86: No need to take full reference when doing mmu_update on a foreign
domain. Also fix a pre-existing race setting PGT_pinned versus
clearing it during domain_kill().
Signed-off-by: Keir Fraser <keir@xensource.com>
19 years agoxen: Cleanups and bug fixes after the rcu_lock_domain patch.
kfraser@localhost.localdomain [Thu, 1 Mar 2007 11:38:55 +0000 (11:38 +0000)]
xen: Cleanups and bug fixes after the rcu_lock_domain patch.
Signed-off-by: Keir Fraser <keir@xensource.com>