When we statically assign a pci device (the pci=3D['xx:xx.x'] string
in guest config file) to guest, we make many checkings (for instance,
if the device is specified in 'pciback.hide', if it has
non-page-aligned MMIO BARs, if it has a proper FLR capability, if the
related devices should be co-assigned). However, with respect to the
guest hotplug, we only check if the device exists and not assigned yet
-- this is not enough, for instance, now xend allows us to assign an
in-use device (being used by Dom0) to an HVM guest (because
xc.test_assigned() returns OK) -- this will cause disaster... The
patch adds some necessary checkings.
Signed-off-by: Dexuan Cui <dexuan.cui@intel.com>
After changeset 18934 (VT-d: Fix PCI-X device assignment), my assigned
PCI E1000 NIC doesn't work in guest.
The NIC is 03:00.0. Its parent bridge is: 00:1e.0.
In domain_context_mapping():
case DEV_TYPE_PCI:
After we domain_context_mapping_one() 03:00.0 and 00:1e.0, the
'secbus' is 3 and 'bus' is 0, so we domain_context_mapping_one()
03:00.0 again -- this redundant invocation returns -EINVAL because
we have created the mapping but haven't changed pdev->domain from
Dom0 to a new domain at this time and eventually the
XEN_DOMCTL_assign_device hypercall returns a failure.
The attached patch detects this case and avoids the redundant
invocation.
Signed-off-by: Dexuan Cui <dexuan.cui@intel.com>
Add support to xenctx for guests that have a different word size
to the tools (x86 only, but shouldn't break ia64). Again, only 32-bit
HVM guests are supported until EFER.LMA is easier to get at.
Signed-off-by: Tim Deegan <Tim.Deegan@citrix.com>
(32-bit) HVM guests. 64-bit HVM guests are still not supported for
now, pending a sensible way of getting at the guest's EFER.LMA.
Signed-off-by: Tim Deegan <Tim.Deegan@citrix.com>
Xend integration for PoD functionality.
* Add python bindings for xc_hvm_domain_build() and
xc_domain_memory_set_pod_target()
* Always call xc_hvm_domain_build(), with memsize = memory_static_max
and target=memory_dynamic_max
* When setting a new memory target:
+ First make sure we actually have enough free memory for the target
setting to succeed
+ Call set_pod_target() with the new target, to Xen can do the Right
Thing.
Signed-off-by: George Dunlap <george.dunlap@eu.citrix.com>
Implement libxc interface to PoD functionality:
* Add xc_hvm_build_target_mem(), which takes both memsize and target.
Memsize is the total memory, allocated in PoD pages and reported in
the e820; target is the size of the cache. If these are the same, the
normal funcitonality is called. (So you can use the same function to
build always, and it will decide whether to use PoD or not.)
* Add xc_domain_memory_[gs]et_pod_target(), which sets and/or returns
information about the PoD cache and p2m entries.
Signed-off-by: George Dunlap <george.dunlap@eu.citrix.com>
Implement Xen interface to PoD functionality.
* Increase the number of MEMOP bits from 4 to 6 (increasing the number
of available memory operations from 16 to 64).
* Introduce XENMEMF_populate_on_demand, which will cause
populate_physmap() to fill a range with PoD entries rather than
backing it with ram
* Introduce XENMEM_[sg]et_pod_target operation to the memory
hypercall, to get and set PoD cache size. set_pod_target() should be
called during domain creation, as well as after modifying the memory
target of any domain which may have outstanding PoD entries.
Signed-off-by: George Dunlap <george.dunlap@eu.citrix.com>
Deal with splintering superpages in the PoD cache, and with
splintering superpage PoD entries in the p2m table.
Signed-off-by: George Dunlap <george.dunlap@eu.citrix.com>
Implement "emergency scan" for zero pages, to deal with start-of-day
page scrubbers.
If the cache is running out, scan through memory looking for "zero
pages" that we can reclaim for the cache. This is necessary for
operating systems which have a start-of-day page scrubber which runs
before the balloon driver can balloon down to the target.
Signed-off-by: George Dunlap <george.dunlap@eu.citrix.com>
Handle balloon driver's calls to decrease_reservation properly.
* Replace PoD entries with p2m_none
* Steal memory for the cache instead of freeing, if need be
Signed-off-by: George Dunlap <george.dunlap@eu.citrix.com>
Shadow code, and other important places, call gfn_to_mfn_query(). In
particular, any place that holds the shadow lock must make a query
call.
Signed-off-by: George Dunlap <george.dunlap@eu.citrix.com>
In DSDT, reserve the IO port ranges: [0x10c0, 0x10c2] and [0xb044,
0xb047] that are used by the virtual PCI hotplug.
Or else, for a hotplugged-in device, the port IO BAR assigned by guest
OS may conflict with the ranges here.
Signed-off-by: Dexuan Cui <dexuan.cui@intel.com>
Set userspace governor as default, which stays same effect
as when cpufreq in xen is not enabled. As a result, enable cpufreq
in xen by default to avoid reboot to activate cpufreq. Now it's
always on but w/o performance impact if user doesn't attempt
to change governor.
Add governor option at cmdline, add some warning info for debug.
Signed-off-by: Liu, Jinsong <jinsong.liu@intel.com>
The machine_check_poll() is called with mi which is set by
x86_mcinfo_getptr() everytime. But, I think it should not be called
when there is no error, because error_idx and fetch_idx cannot work
together.
Signed-off-by: Kazuhiro Suzuki <kaz@jp.fujitsu.com>
Since cmci might happened when cpu is taking down (cpu hotplug) before
setting new cmci owner while old owner is down. We need to clear the
corrected error counter field to make sure CMCI could be triggered on
the new owner.
Signed-off-by: Yunhong Jiang <yunhong.jiang@intel.com>
Signed-off-by: Liping Ke <liping.ke@intel.com>
As pcidevs_lock is changed from protecting only the alldevs_list to
more than that, it doesn't benifit too much from the rw_lock. Also the
previous patch 18906:2941b1a97c60 is wrong to use read_lock to protect some
sensitive data (thanks Espen pointed out that).
Also two minor fix in this patch:
a) deassign_device will deadlock when try to get the pcidevs_lock if
called by pci_release_devices, remove the lock to the caller.
b) The iommu_domain_teardown should not ASSERT for the pcidevs_lock
because it just update the domain's vt-d mapping.
Signed-off-by: Yunhong Jiang <yunhong.jiang@intel.com>
When assign PCI device, current code just map its bridge and its
secondary bus number and devfn 0. It doesn't work for PCI-x device
assignment, because the request may be the source-id in the original
PCI-X transaction or the source-id provided by the bridge. It needs to
map the device itself, and its upstream bridges till PCIe-to-PCI/PCI-x
bridge.
In addition, add description for DEV_TYPE_PCIe_BRIDGE and
DEV_TYPE_PCI_BRIDGE for understandability.
Signed-off-by: Weidong Han <weidong.han@intel.com>
this means that in some cases it cannot be owner of its own xenstore
nodes.
This bug was pointed out by Daniel Berrange at Red Hat. This patch is
my own more generic fix that automatically covers a range of callers
(albeit the patch is arguably a bit of a hack ;-).
Signed-off-by: Keir Fraser <keir.fraser@citrix.com>
- Avoid nested function to avoid a trampoline.
- Do not link mini-os_app.o when it is empty.
Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
external-mode guest if the fault came from Xen; it would be making
changes to the wrong pagetables, potentially causing a pagefault loop
in Xen.
Signed-off-by: Tim Deegan <Tim.Deegan@citrix.com>
tune the parameters manually.
Now, xenpm can be invoked with the following options:
Usage:
xenpm get-cpuidle-states [cpuid]: list cpu idle information on
CPU cpuid or all CPUs.
xenpm get-cpufreq-states [cpuid]: list cpu frequency
information on CPU cpuid or all CPUs.
xenpm get-cpufreq-para [cpuid]: list cpu frequency information
on CPU cpuid or all CPUs.
xenpm set-scaling-maxfreq <cpuid> <HZ>: set max cpu frequency
<HZ> on CPU <cpuid>.
xenpm set-scaling-minfreq <cpuid> <HZ>: set min cpu frequency
<HZ> on CPU <cpuid>.
xenpm set-scaling-governor <cpuid> <name>: set scaling governor
on CPU <cpuid>.
xenpm set-scaling-speed <cpuid> <num>: set scaling speed on CPU
<cpuid>.
xenpm set-sampling-rate <cpuid> <num>: set sampling rate on CPU
<cpuid>.
xenpm set-up-threshold <cpuid> <num>: set up threshold on CPU
<cpuid>.
To ease the use of this tool, the shortcut option is supported,
i.e. `xenpm get-cpui' is equal to `xenpm get-cpuidle-states'.
Signed-off-by: Guanqun Lu <guanqun.lu@intel.com>
We must not disable EFER.NXE in host mode since shadow code relies on
accessing shadow mappings with NX set.
We do not want to write EFER on every vmentry/vmexit if we can avoid
it, since it will be somewhat slow.
Finally, we don't believe that any guest relies on NX really being
disabled when EFER.NXE is cleared.
This given, it makes sense to ignore the guest's setting of EFER.NXE.
Signed-off-by: Keir Fraser <keir.fraser@citrix.com>
Extended Bios Data Area (EBDA) can be relocated by the initialization
of PCI option ROM. The IPL boot table is also.
EBDA must be accessed via 0x40E after the initialization.
Signed-off-by: Kouya Shimura <kouya@jp.fujitsu.com>
Keir Fraser