The ladder governor has long pro/demotion delay shortcome while
applying to tickless mode, because it needs to count usage. Menu
governor chooses the next state simply via break-event prediction
including the factors of next timer event & last residency time etc,
so it would have faster response speed.
Signed-off-by: Gang Wei <gang.wei@intel.com>
LAPIC will stop during C3, and resume to work after exit from
C3. Considering below case:
The LAPIC timer was programmed to expire after 1000us, but CPU enter
C3 after 100us and exit C3 at 9xxus.
0us: reprogram_timer(1000us)
100us: entry C3, LAPIC timer stop
9xxus: exit C3 due to unexpected event, LAPIC timer continue running
10xxus: reprogram_timer(1000us), fail due to the past expiring time.
......: no timer softirq raised, no change to LAPIC timer.
......: if entry C3 again, HPET will be forced reprogramed to
now+small_slop.
......: if entry C2, no change to LAPIC.
18xxus: LAPIC timer expires unexpectedly if no C3 entries after
10xxus.
Signed-off-by: Wei Gang <gang.wei@intel.com>
Exit from C3 is mainly caused by HPET intr if force enable
hpetbroadcast. But it is still probably caused by other unexpected
events. In this case the HPET timer may still be alive while there is
no CPU in C3. Avoid those remnant HPET intr can save cpu handling time
and increase idle time.
Signed-off-by: Wei Gang <gang.wei@intel.com>
Taking security into accout, it's not suitable to bypass VT-d
translation for Dom0 by default when the pass-through field in
extended capability register is set. This feature is for people/usages
who are not overly worried about security/isolation, but want better
performance.
This patch adds a command line param that controls if it's enabled or
disabled.
Signed-off-by: Weidong Han <weidong.han@intel.com>
There is an python indentation issue keeping the full range of syntax
for the cpuid config file parameter from working correctly. This
patch fixes that. It also fixes some misspelling and a missing 'x' in
two of the example config files (must have 32 bits represented for
cpuid registers).
Signed-off-by: Bruce Rogers <brogers@novell.com>
This small patch fixes an issue leading to a crash (segfault, although
with earlier changesets I was seeing sigbus - not sure what changed)
in qemu-dm when the following conditions occur:
1. A valid mapping for a bucket on a low address exists
2. Immediately after accessing memory mapped in this bucket, an access
occurs to a high (beyond assigned ram) address beyond the 1GB limit
for 32bit map cache wrapping around to the previous bucket's entry
number.
3. The next call to map cache again accesses the low address.
In this scenario, the guest mem for the low bucket has been unmapped
by the remap_bucket caused by 2., but because the valid_mapping
bit-test fails, map_cache returns before last_address_index has been
updated. The subsequent call to map_cache therefore never remaps the
low, valid bucket and instead returns a vaddr pointing to memory that
has failed to get mapped.
Signed-off-by: Trolle Selander <trolle.selander@eu.citrix.com>
- fix ioemu segv with old firmware
Without notifying ioemu of address, ioemu will segv.
- fix qemu-dm segv with malicous firmware
If notifying ioemu more than once, ioemu will segv.
Usually such cases don't happen, but malicious guest can
do it intentionally.
Signed-off-by: Isaku Yamahata <yamahata@valinux.co.jp>
If pass-through field in extended capability register is set, set
pass-through translation type for Dom0, that means DMA requests with
Untranslated addresses are processed as pass-through in Dom0, needn't
translate DMA requests through a multi-level page-table.
Signed-off-by: Anthony Xu <anthony.xu@intel.com>
Signed-off-by: Weidong Han <weidong.han@intel.com>
This information is now redundant, as it can be determined using
runstate changes and lost record changes. This reduces log size for
long-running traces by 10-30%.
Signed-off-by: George Dunlap <george.dunlap@eu.citrix.com>
Signed-off-by: Trolle Selander <trolle.selander@eu.citrix.com>
and tsc of first lost record.
Including the current vcpu helps us identify who's currently
running on each processor, even if we've lost records. The tsc of the
first lost record helps us identify how much time we've spent lost,
and also when a pcpu started experiencing lost records.
Signed-off-by: George Dunlap <george.dunlap@eu.citrix.com>
Signed-off-by: Trolle Selander <trolle.selander@eu.citrix.com>
Also, introduce SCHED_MIN and SCHED_VERBOSE, so that we can trace
just runstate changes if we want to.
Signed-off-by: George Dunlap <george.dunlap@eu.citrix.com>
Signed-off-by: Trolle Selander <trolle.selander@eu.citrix.com>
cpufreq_dom_mask must not be confined to the number of CPUs in the
system - there's nothing disallowing e.g. 1-based domain identifiers
to be used in _PSD. It still remains risky to assume that these
numbers are 'small' and/or not allocated sparsely.
Signed-off-by: Jan Beulich <jbeulich@novell.com>
I've noticed that the suspend event channel becomes pending as soon as
it is bound. I'm not sure why or whether this is intentional, but it
means that the suspend function will return before the domain has
completed suspending unless the first notification is cleared. Without
this patch, xc_domain_save may find that the guest has not suspended
and sleep in 10ms chunks until it does. Typically this is several
milliseconds of wasted time.
From: Brendan Cully <brendan@cs.ubc.ca>
Signed-off-by: Keir Fraser <keir.fraser@citrix.com>
Current code does not set dev->msix->table_off variable.
The offset of MSI-X memory mapped table is treated as 0.
The wrong region is unmapped from guest physical memory space.
As a result, guest device driver can't access memory mapped resource.
Signed-off-by: Yuji Shimada <shimada-yxb@necst.nec.co.jp>
- The patch includes a policy for xen that can be booted into
enforcing mode and supports creation and management of
paravirtualized guests. The policy follows the dom0/domU usage
model, extension to other models or the addition of management or IO
permissions should be much more straightforward now. The option
flask_enforcing=1 can be passed on the xen line in grub to boot
into enforcing mode.
- The policy provides a basic policy for booting the platform and
creating a domU with the label system_u:object_r:domU_t. The policy
can be easily extended to support new types by modifying the xen.te
source file.
- The policy includes some basic macros which may be helpful in
extending the policy.
- The policy is compatible with and requires the most recent XSM
patch, xsm-flask-io-sysctl-hooks-090308.diff.
- The policy is not built as part of the make all as it requires the
SELinux policy compiler which may/may not be installed on all
systems. Users must go into the tools/flask/policy directory and
explicitly compile the policy.
Signed-off-by: George Coker <gscoker@alpha.ncsc.mil>
- The patch refactors the IO resource checks into the rangeset add/del
code. This produces a much more architecture friendly implementation and
ensures that HVM and paravirtualized guests are checked consistently.
- The patch removes the following hooks in support of the refactoring
of the IO resource checks:
- xsm_irq_permission
- xsm_iomem_permission
- xsm_ioport_permission
- The patch adds the following hooks in support of the refactoring of
the IO resource checks:
- xsm_add_range
- xsm_remove_range
- These IO refactoring changes are transparent to any pre-existing
Flask policies.
- The patch adds also adds hooks for sysctl functionality that was
added since the last major XSM patch. The following hooks were added:
- xsm_set_target
- xsm_debug_keys
- xsm_getcpuinfo
- xsm_availheap
- xsm_firmware_info
- xsm_acpi_sleep
- xsm_change_freq
- xsm_getidletime
- xsm_sendtrigger
- xsm_test_assign_device
- xsm_assign_device
- xsm_deassign_device
- xsm_bind_pt_irq
- xsm_pin_mem_cacheattr
- xsm_ext_vcpucontext
Signed-off-by: George Coker <gscoker@alpha.ncsc.mil>
- The patch does away with the autogenerated xsm.py file and
introduces a config parameter in xend-config.sxp to determine the
security module. The parameter is (xsm_module_name {acm, dummy,
flask}). The default setting/option is dummy. .hgignore is also
updated to stop ignoring xsm.py on commits.
- The patch has created an xsconstant for XS_POLICY_FLASK and updated
the toolchain to check the instance of XS_POLICY_USE. XS_POLICY_USE
evalauates to XS_POLICY_FLASK or XS_POLICY_ACM or XS_POLICY_DUMMY
depending on configuration.
- Flask relies on the current value of ssidref returned by dominfo to
ensure that the label to sid mapping is consistent. ssidref had
been pop'ed from the dominfo object. The patch addresses this
issue.
- Flask python module style cleanups.
Signed-off-by: George Coker <gscoker@alpha.ncsc.mil>
By protecting entry/exit with a spinlock we can safely determine
precisely when it is required that we assert/deassert ARB_DISABLE.
Signed-off-by: Wei Gang <gang.wei@intel.com>
Signed-off-by: Keir Fraser <keir.fraser@citrix.com>
The guest handle checks should use paging_* predicates, not shadow_*.
Also tidy up a few places where p2m definitions were being imported
via asm/guest_access.h -> asm/shadow.h -> asm/p2m.h
Signed-off-by: Tim Deegan <Tim.Deegan@citrix.com>
Idle vcpu periodic timer is useless. It increased the lapic timer
interrupt number, which decreased the cpu idle residency. This patch
disables idle vcpu periodic timer via keeping v->periodic_period = 0.
The vcpu periodic timer may be expired 50us before expected time
because there is a 50us TIMER_SLOP used for soft timer (xen/common/timer.c,
timer_softirq_action()). This will cause vcpu_periodic_timer_work() be
continuously called tens of times in a single call of
timer_softirq_action() until (now > periodic_next_event). It brings
unnecessary overhead. This patch adds a similar time slop in vcpu
periodic timer to eliminate this overhead.
Signed-off-by: Wei Gang <gang.wei@intel.com>
This patch moves the call to XendDevices.destroy_device_state() to the
function _remove_domain. Previously only Xen-API calls were calling
the destroy_device_state() function.
Signed-off-by: Stefan Berger <stefanb@us.ibm.com>
Linux 2.6.27 adds code to enable extended config space accesses in the
Northbridge Configuration MSR; Xen should allow Dom0 to control the
respective bit.
Likewise, 2.6.26 added support to enable the MMIO config space access
method for certain Sun systems, so similarly Xen should allow Dom0 to
control the respective fields of the MMIO Configuration Base Address
Register.
Signed-off-by: Jan Beulich <jbeulich@novell.com>
This is only a first step - more call sites need to be hooked up.
Most of this is really Keir's work, I just took what he handed me and
fixed a few remaining issues.
Signed-off-by: Jan Beulich <jbeulich@novell.com>
Signed-off-by: Keir Fraser <keir.fraser@citrix.com>
Keir Fraser