Contemplate a memory space slack for PV domains,
since they do ballooning (or flipping network rx)
and need some extra room in their pfn space.
Note that this does not allocate any extra memory
to the domain, it simply extends the physmap with
some extra room for "bounce bufffering back" pfn's
that are yielded to dom0.
The default slack is set at 8MB.
Signed-off-by: Andres Lagar-Cavilla <andres@lagarcavilla.com>
Acked-by: Vincent Hanquez <vincent.hanquez@eu.citrix.com>
Added libxenlight implementation for resume domain.
This brings back a cooperative pv domain from the
shutdown state after save, enabling checkpointing.
Signed-off-by: Andres Lagar-Cavilla <andres@lagarcavilla.com>
Prevent segmentation fault caused by empty values
in key-value pairs for the /vm/ subdirectory
when restoring a pv domain.
Signed-off-by: Andres Lagar-Cavilla <andres@lagarcavilla.com>
Expose disk and nic device destroy calls
Also removes the obsolete device shutdown calls.
Signed-off-by: Andres Lagar-Cavilla <andres@lagarcavilla.com>
Refactor libxl device destroy code. Abstract function
waiting for the watch on the state node to fire.
Create a generic device delete function.
Only a single LIBXL_DESTROY_TIMEOUT elapses when
waiting for destruction of all the devices of a
domain.
Signed-off-by: Andres Lagar-Cavilla <andres@lagarcavilla.com>
Provide a function to clone a context. This is necessary
because simply copying the structs will eventually
corrup the GC: maxsize is updated in the cloned context
but not in the originating, yet they have the same array
of referenced pointers alloc_ptrs.
Signed-off-by: Andres Lagar-Cavilla <andres@lagarcavilla.com>
...even if there are active grants.
This triggers when checkpoint a guest which essentially resumes
without actually having gone through the suspend so the domain is
already latched to v2 inside Xen.
Also return the current actual version on success and failure. Not
terribly useful with only 2 options but is more robust to future
developments.
Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
A guest PFN may have been marked dirty and switched to p2m_ram_rw by
another CPU between the VMEXIT and lookup in this handler, so
we can't just check for p2m_ram_logdirty. Also, handle_mmio
doesn't handle passthrough MMIO.
Signed-off-by: Tim Deegan <Tim.Deegan@citrix.com>
Add a '-f' option to xentop to allow the full domain name to be
displayed. This is the original behavior which can cause the display
to be unaligned. Customers have requested this because only the
trailing characters of their domain names are unique and therefore
cannot be distinguished when the display is limited to a 10 character
width.
Signed-off-by: Charles Arnold <carnold@novell.com>
this patch fixes the multiple xenstore watches problem in libxenlight
opening a new xenstore connection to set and read temporary watches on
the device state nodes. This way they don't interfere with other long
running watches.
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
This patch reimplements libxl_wait_for_device_model using a xenstore
watch and a select loop.
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
The function dm_xenstore_record_pid is executed by a child of the main
process and therefore shouldn't use the same xenstore connection:
currently it opens a new connection but still uses the old one.
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
In above c/s I introduced dom0 statistics for case we use network
bonding. The indentation was not good for xenstat C codebase and also
some modifications were done to the logic, mainly not using the parsed
variables we don't care about (as we care only about
{tx|rx}{bytes,packets,errs,drops} and no other variable from
/proc/net/dev) by passing NULLs to variables we don't care about. Also
dom0 statistics alteration was fixed to include {tx|rx}{drop,errs} for
dom0 (previous version of my patch was not having this code applied).
Signed-off-by: Michal Novotny <minovotn@redhat.com>
The main idea of this patch is:
1) The admin sets aside some memory below 128G for 32-bit paravirtual
domain creation (via dom0_mem=-<value> in kernel comand line).
2) The admin also explicitly states to the tools (i..e xend) how much
memory is supposed to be left untouched by 64-bit domains
3) If a 32-bit pv DomU gets created, no ballooning ought to be
necessary (since if it is, no guarantee can be made about the address
range of the memory ballooned out), and memory gets allocated from the
reserved range.
4) Upon 64-bit (or 32-bit HVM or HVM) DomU creation, the tools
determine the amount of memory to be ballooned out of Dom0 by adding
the amount needed for the new guest and the amount still in the
reserved pool (and then of course subtracting the total amount of
memory the hypervisor has available for guest use).
Signed-off-by: james song (wei) <jsong@novell.com>
Currently iommu_flush_cache_entry uses a fixed size 8 bytes to flush
cache. But it also needs to flush caches with different sizes,
e.g. struct root_entry is 16 bytes. This patch fixes the hardcode by
using a parameter "size" to flush caches with different sizes.
Signed-off-by: Weidong Han <weidong.han@intel.com>
BaseException.message has been deprecated since Python 2.6. To
prevent DeprecationWarning from popping up over this pre-existing
attribute, use a new property that takes lookup precedence.
Signed-off-by: Wei Kong <weikong.cn@gmail.com>
It turns out that there is a better way to write stubdoms logs to file
than using libxl_console_attach: qemu is the one that provides the
console backend for stubdoms and qemu is able to redirect a serial to
file, so we can use this feature to make sure the first stubdom
console is always redirected to a logfile.
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
- set the domid of the guest and not the one of the stubdom in the
libxl_device_model_starting returned to the user;
- check that the length of the two strings matches in
libxl_name_to_domid, otherwise we can get a match for two different
domains that have the same initial part of the name.
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Without this patch, loading a bzimage dom0 kernel while also
requesting a dynamically-allocated crashkernel area is broken.
Signed-off-by: Keir Fraser <keir.fraser@citrix.com>
this patch makes libxl_create_stubdom wait for pv qemu to be properly
initialized before unpausing the stubdom.
A new libxl_device_model_starting pointer is used to wait for pv qemu
initialization while the libxl_device_model_starting pointer given by
the user is initialized to a new structure with an empty for_spawn
member, because nothing that was spawn has to be waited for anymore.
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
This attempts to address all the concerns raised in
http://lists.xensource.com/archives/html/xen-devel/2009-11/msg01195.html,
but I'm nevertheless still not convinced that all aspects of the
injection handling really work reliably. In particular, while the
patch here on top of the fixes for the problems menioned in the
referenced mail also adds code to keep send_guest_trap() from
injecting multiple events at a time, I don't think the is the right
mechanism - it should be possible to handle NMI/MCE nested within
each other.
Another fix on top of the ones for the earlier described problems is
that the vCPU affinity restore logic didn't account for software
injected NMIs - these never set cpu_affinity_tmp, but due to it most
likely being different from cpu_affinity it would have got restored
(to a potentially random value) nevertheless.
Signed-off-by: Jan Beulich <jbeulich@novell.com>
I did some benchmark runs (lmbench & kernel compile) with a number of
guests running in parallel to compare the performance of numa=on vs.
numa=off. As soon as one starts to load the machine, the performance
goes down in the numa=off case. The tests were done on an 8-node
machine (4 cores each). lmbench (actually copying large amounts of
memory) shows a dramatic dropdown, but I even noticed significant
performance decrease for a tmpfs based Linux kernel compile. Here a
summary of the data:
lmbench's rd benchmark (normalized to native Linux (=100)):
guests numa=off numa=on avg increase
min avg max min avg max
1 78.0 102.3
7 37.4 45.6 62.0 90.6 102.3 110.9 124.4%
15 21.0 25.8 31.7 41.7 48.7 54.1 88.2%
23 13.4 17.5 23.2 25.0 28.0 30.1 60.2%
kernel compile in tmpfs, 1 VCPU, 2GB RAM, average of elapsed time:
guests numa=off numa=on increase
1 480.610 464.320 3.4%
7 482.109 461.721 4.2%
15 515.297 477.669 7.3%
23 548.427 495.180 9.7%
again with 2 VCPUs and make -j2:
1 264.580 261.690 1.1%
7 279.763 258.907 7.7%
15 330.385 272.762 17.4%
23 463.510 390.547 15.7% (46 VCPUs on 32pCPUs)
Selected tests on a 4-node machine showed similar behavior (7.9 %
increase with 6 parallel guests on the 2 VCPU kernel compile
benchmark).
Note that this does not affect non-NUMA machines at all, since NUMA
will be turned off again by the code if no NUMA topology is detected.
Signed-off-by: Andre Przywara <andre.przywara@amd.com>
Enable turning on a different amount of vcpus than
the maximum during domain creation/restore.
Signed-off-by: Andres Lagar-Cavilla <andres@lagarcavilla.com>
Allow domain to publish its suspend event channel.
Otherwise, the fast event-channel-based suspend
path is disabled.
Signed-off-by: Andres Lagar-Cavilla <andres@lagarcavilla.com>
Keir Fraser