We observed some performance degradation on s390x with dynamic
halt polling. Until we can provide a proper fix, let's enable
halt_poll_ns as default only for supported architectures.
Architectures are now free to set their own halt_poll_ns
default value.
Signed-off-by: David Hildenbrand <dahi@linux.vnet.ibm.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
This new statistic can help diagnosing VCPUs that, for any reason,
trigger bad behavior of halt_poll_ns autotuning.
For example, say halt_poll_ns = 480000, and wakeups are spaced exactly
like 479us, 481us, 479us, 481us. Then KVM always fails polling and wastes
10+20+40+80+160+320+480 = 1110 microseconds out of every
479+481+479+481+479+481+479 = 3359 microseconds. The VCPU then
is consuming about 30% more CPU than it would use without
polling. This would show as an abnormally high number of
attempted polling compared to the successful polls.
Acked-by: Christian Borntraeger <borntraeger@de.ibm.com<
Reviewed-by: David Matlack <dmatlack@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
We have abstracted the data struct and functions which are used to check
reserved bit on guest page tables, now we extend the logic to check
zero bits on shadow page tables
The zero bits on sptes include not only reserved bits on hardware but also
the bits that SPTEs willnever use. For example, shadow pages will never
use GB pages unless the guest uses them too.
Signed-off-by: Xiao Guangrong <guangrong.xiao@linux.intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
These two fields, rsvd_bits_mask and bad_mt_xwr, in "struct kvm_mmu" are
used to check if reserved bits set on guest ptes, move them to a data
struct so that the approach can be applied to check host shadow page
table entries as well
Signed-off-by: Xiao Guangrong <guangrong.xiao@linux.intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
This patch introduce Hyper-V related source code file - hyperv.c and
per vm and per vcpu hyperv context structures.
All Hyper-V MSR's and hypercall code moved into hyperv.c.
All Hyper-V kvm/vcpu fields moved into appropriate hyperv context
structures. Copyrights and authors information copied from x86.c
to hyperv.c.
Signed-off-by: Andrey Smetanin <asmetanin@virtuozzo.com>
Signed-off-by: Denis V. Lunev <den@openvz.org>
Reviewed-by: Peter Hornyack <peterhornyack@google.com>
CC: Paolo Bonzini <pbonzini@redhat.com>
CC: Gleb Natapov <gleb@kernel.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
If there are no assigned devices, the guest PAT are not providing
any useful information and can be overridden to writeback; VMX
always does this because it has the "IPAT" bit in its extended
page table entries, but SVM does not have anything similar.
Hook into VFIO and legacy device assignment so that they
provide this information to KVM.
Reviewed-by: Alex Williamson <alex.williamson@redhat.com>
Tested-by: Joerg Roedel <jroedel@suse.de>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Writes were a bit racy, but hard to turn into a bug at the same time.
(Particularly because modern Linux doesn't use this feature anymore.)
Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
[Actually the next patch makes it much, much easier to trigger the race
so I'm including this one for stable@ as well. - Paolo]
Cc: stable@vger.kernel.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
for silicon that no one owns: these are really new features for
everyone.
* ARM: several features are in progress but missed the 4.2 deadline.
So here is just a smattering of bug fixes, plus enabling the VFIO
integration.
* s390: Some fixes/refactorings/optimizations, plus support for
2GB pages.
* x86: 1) host and guest support for marking kvmclock as a stable
scheduler clock. 2) support for write combining. 3) support for
system management mode, needed for secure boot in guests. 4) a bunch
of cleanups required for 2+3. 5) support for virtualized performance
counters on AMD; 6) legacy PCI device assignment is deprecated and
defaults to "n" in Kconfig; VFIO replaces it. On top of this there are
also bug fixes and eager FPU context loading for FPU-heavy guests.
* Common code: Support for multiple address spaces; for now it is
used only for x86 SMM but the s390 folks also have plans.
There are some x86 conflicts, one with the rc8 pull request and
the rest with Ingo's FPU rework.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.22 (GNU/Linux)
iQEcBAABAgAGBQJViYzhAAoJEL/70l94x66Dda0H/1IepMbfEy+o849d5G71fNTs
F8Y8qUP2GZuL7T53FyFUGSBw+AX7kimu9ia4gR/PmDK+QYsdosYeEjwlsolZfTBf
sHuzNtPoJhi5o1o/ur4NGameo0WjGK8f1xyzr+U8z74QDQyQv/QYCdK/4isp4BJL
ugHNHkuROX6Zng4i7jc9rfaSRg29I3GBxQUYpMkEnD3eMYMUBWGm6Rs8pHgGAMvL
vqzntgW00WNxehTqcAkmD/Wv+txxhkvIadZnjgaxH49e9JeXeBKTIR5vtb7Hns3s
SuapZUyw+c95DIipXq4EznxxaOrjbebOeFgLCJo8+XMXZum8RZf/ob24KroYad0=
=YsAR
-----END PGP SIGNATURE-----
Merge tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm
Pull first batch of KVM updates from Paolo Bonzini:
"The bulk of the changes here is for x86. And for once it's not for
silicon that no one owns: these are really new features for everyone.
Details:
- ARM:
several features are in progress but missed the 4.2 deadline.
So here is just a smattering of bug fixes, plus enabling the
VFIO integration.
- s390:
Some fixes/refactorings/optimizations, plus support for 2GB
pages.
- x86:
* host and guest support for marking kvmclock as a stable
scheduler clock.
* support for write combining.
* support for system management mode, needed for secure boot in
guests.
* a bunch of cleanups required for the above
* support for virtualized performance counters on AMD
* legacy PCI device assignment is deprecated and defaults to "n"
in Kconfig; VFIO replaces it
On top of this there are also bug fixes and eager FPU context
loading for FPU-heavy guests.
- Common code:
Support for multiple address spaces; for now it is used only for
x86 SMM but the s390 folks also have plans"
* tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm: (124 commits)
KVM: s390: clear floating interrupt bitmap and parameters
KVM: x86/vPMU: Enable PMU handling for AMD PERFCTRn and EVNTSELn MSRs
KVM: x86/vPMU: Implement AMD vPMU code for KVM
KVM: x86/vPMU: Define kvm_pmu_ops to support vPMU function dispatch
KVM: x86/vPMU: introduce kvm_pmu_msr_idx_to_pmc
KVM: x86/vPMU: reorder PMU functions
KVM: x86/vPMU: whitespace and stylistic adjustments in PMU code
KVM: x86/vPMU: use the new macros to go between PMC, PMU and VCPU
KVM: x86/vPMU: introduce pmu.h header
KVM: x86/vPMU: rename a few PMU functions
KVM: MTRR: do not map huge page for non-consistent range
KVM: MTRR: simplify kvm_mtrr_get_guest_memory_type
KVM: MTRR: introduce mtrr_for_each_mem_type
KVM: MTRR: introduce fixed_mtrr_addr_* functions
KVM: MTRR: sort variable MTRRs
KVM: MTRR: introduce var_mtrr_range
KVM: MTRR: introduce fixed_mtrr_segment table
KVM: MTRR: improve kvm_mtrr_get_guest_memory_type
KVM: MTRR: do not split 64 bits MSR content
KVM: MTRR: clean up mtrr default type
...
This patch defines a new function pointer struct (kvm_pmu_ops) to
support vPMU for both Intel and AMD. The functions pointers defined in
this new struct will be linked with Intel and AMD functions later. In the
meanwhile the struct that maps from event_sel bits to PERF_TYPE_HARDWARE
events is renamed and moved from Intel specific code to kvm_host.h as a
common struct.
Reviewed-by: Joerg Roedel <jroedel@suse.de>
Tested-by: Joerg Roedel <jroedel@suse.de>
Signed-off-by: Wei Huang <wei@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
This will be used for private function used by AMD- and Intel-specific
PMU implementations.
Signed-off-by: Wei Huang <wei@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Sort all valid variable MTRRs based on its base address, it will help us to
check a range to see if it's fully contained in variable MTRRs
Signed-off-by: Xiao Guangrong <guangrong.xiao@linux.intel.com>
[Fix list insertion sort, simplify var_mtrr_range_is_valid to just
test the V bit. - Paolo]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Variable MTRR MSRs are 64 bits which are directly accessed with full length,
no reason to split them to two 32 bits
Signed-off-by: Xiao Guangrong <guangrong.xiao@linux.intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Drop kvm_mtrr->enable, omit the decode/code workload and get rid of
all the hard code
Signed-off-by: Xiao Guangrong <guangrong.xiao@linux.intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Only KVM_NR_VAR_MTRR variable MTRRs are available in KVM guest
Signed-off-by: Xiao Guangrong <guangrong.xiao@linux.intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
vMTRR does not depend on any host MTRR feature and fixed MTRRs have always
been implemented, so drop this field
Signed-off-by: Xiao Guangrong <guangrong.xiao@linux.intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
MTRR code locates in x86.c and mmu.c so that move them to a separate file to
make the organization more clearer and it will be the place where we fully
implement vMTRR
Signed-off-by: Xiao Guangrong <guangrong.xiao@linux.intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
... and we're done. :)
Because SMBASE is usually relocated above 1M on modern chipsets, and
SMM handlers might indeed rely on 4G segment limits, we only expose it
if KVM is able to run the guest in big real mode. This includes any
of VMX+emulate_invalid_guest_state, VMX+unrestricted_guest, or SVM.
Reviewed-by: Radim Krčmář <rkrcmar@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
This is now very simple to do. The only interesting part is a simple
trick to find the right memslot in gfn_to_rmap, retrieving the address
space from the spte role word. The same trick is used in the auditing
code.
The comment on top of union kvm_mmu_page_role has been stale forever,
so remove it. Speaking of stale code, remove pad_for_nice_hex_output
too: it was splitting the "access" bitfield across two bytes and thus
had effectively turned into pad_for_ugly_hex_output.
Reviewed-by: Radim Krčmář <rkrcmar@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
This patch has no semantic change, but it prepares for the introduction
of a second address space for system management mode.
A new function x86_set_memory_region (and the "slots_lock taken"
counterpart __x86_set_memory_region) is introduced in order to
operate on all address spaces when adding or deleting private
memory slots.
Reviewed-by: Radim Krčmář <rkrcmar@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
We need to hide SMRAM from guests not running in SMM. Therefore,
all uses of kvm_read_guest* and kvm_write_guest* must be changed to
check whether the VCPU is in system management mode and use a
different set of memslots. Switch from kvm_* to the newly-introduced
kvm_vcpu_*, which call into kvm_arch_vcpu_memslots_id.
Reviewed-by: Radim Krčmář <rkrcmar@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
This patch adds the interface between x86.c and the emulator: the
SMBASE register, a new emulator flag, the RSM instruction. It also
adds a new request bit that will be used by the KVM_SMI ioctl.
Reviewed-by: Radim Krčmář <rkrcmar@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
This patch includes changes to the external API for SMM support.
Userspace can predicate the availability of the new fields and
ioctls on a new capability, KVM_CAP_X86_SMM, which is added at the end
of the patch series.
Reviewed-by: Radim Krčmář <rkrcmar@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
SMBASE is only readable from SMM for the VCPU, but it must be always
accessible if userspace is accessing it. Thus, all functions that
read MSRs are changed to accept a struct msr_data; the host_initiated
and index fields are pre-initialized, while the data field is filled
on return.
Reviewed-by: Radim Krčmář <rkrcmar@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
This lets the function access the new memory slot without going through
kvm_memslots and id_to_memslot. It will simplify the code when more
than one address space will be supported.
Unfortunately, the "const"ness of the new argument must be casted
away in two places. Fixing KVM to accept const struct kvm_memory_slot
pointers would require modifications in pretty much all architectures,
and is left for later.
Reviewed-by: Radim Krcmar <rkrcmar@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The MPX feature requires eager KVM FPU restore support. We have verified
that MPX cannot work correctly with the current lazy KVM FPU restore
mechanism. Eager KVM FPU restore should be enabled if the MPX feature is
exposed to VM.
Signed-off-by: Yang Zhang <yang.z.zhang@intel.com>
Signed-off-by: Liang Li <liang.z.li@intel.com>
[Also activate the FPU on AMD processors. - Paolo]
Cc: stable@vger.kernel.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
KVM may turn a user page to a kernel page when kernel writes a readonly
user page if CR0.WP = 1. This shadow page entry will be reused after
SMAP is enabled so that kernel is allowed to access this user page
Fix it by setting SMAP && !CR0.WP into shadow page's role and reset mmu
once CR4.SMAP is updated
Signed-off-by: Xiao Guangrong <guangrong.xiao@linux.intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Now that fpstate_init() cannot fail the error return of fx_init()
has lost its purpose. Eliminate the error return and propagate this
change to all callers.
Reviewed-by: Borislav Petkov <bp@alien8.de>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Fenghua Yu <fenghua.yu@intel.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
KVM may turn a user page to a kernel page when kernel writes a readonly
user page if CR0.WP = 1. This shadow page entry will be reused after
SMAP is enabled so that kernel is allowed to access this user page
Fix it by setting SMAP && !CR0.WP into shadow page's role and reset mmu
once CR4.SMAP is updated
Signed-off-by: Xiao Guangrong <guangrong.xiao@linux.intel.com>
Cc: stable@vger.kernel.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Extended struct kvm_lapic_irq with bool msi_redir_hint, which will
be used to determine if the delivery of the MSI should target only
the lowest priority CPU in the logical group specified for delivery.
(In physical dest mode, the RH bit is not relevant). Initialized the value
of msi_redir_hint to true when RH=1 in kvm_set_msi_irq(), and initialized
to false in all other cases.
Added value of msi_redir_hint to a debug message dump of an IRQ in
apic_send_ipi().
Signed-off-by: James Sullivan <sullivan.james.f@gmail.com>
Reviewed-by: Radim Krčmář <rkrcmar@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Change to u16 if they only contain data in the low 16 bits.
Change the level field to bool, since we assign 1 sometimes, but
just mask icr_low with APIC_INT_ASSERT in apic_send_ipi.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
x86 architecture defines differences between the reset and INIT sequences.
INIT does not initialize the FPU (including MMX, XMM, YMM, etc.), TSC, PMU,
MSRs (in general), MTRRs machine-check, APIC ID, APIC arbitration ID and BSP.
References (from Intel SDM):
"If the MP protocol has completed and a BSP is chosen, subsequent INITs (either
to a specific processor or system wide) do not cause the MP protocol to be
repeated." [8.4.2: MP Initialization Protocol Requirements and Restrictions]
[Table 9-1. IA-32 Processor States Following Power-up, Reset, or INIT]
"If the processor is reset by asserting the INIT# pin, the x87 FPU state is not
changed." [9.2: X87 FPU INITIALIZATION]
"The state of the local APIC following an INIT reset is the same as it is after
a power-up or hardware reset, except that the APIC ID and arbitration ID
registers are not affected." [10.4.7.3: Local APIC State After an INIT Reset
("Wait-for-SIPI" State)]
Signed-off-by: Nadav Amit <namit@cs.technion.ac.il>
Message-Id: <1428924848-28212-1-git-send-email-namit@cs.technion.ac.il>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Introducing KVM_CAP_DISABLE_QUIRKS for disabling x86 quirks that were previous
created in order to overcome QEMU issues. Those issue were mostly result of
invalid VM BIOS. Currently there are two quirks that can be disabled:
1. KVM_QUIRK_LINT0_REENABLED - LINT0 was enabled after boot
2. KVM_QUIRK_CD_NW_CLEARED - CD and NW are cleared after boot
These two issues are already resolved in recent releases of QEMU, and would
therefore be disabled by QEMU.
Signed-off-by: Nadav Amit <namit@cs.technion.ac.il>
Message-Id: <1428879221-29996-1-git-send-email-namit@cs.technion.ac.il>
[Report capability from KVM_CHECK_EXTENSION too. - Paolo]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Dirty logging tracks sptes in 4k granularity, meaning that large sptes
have to be split. If live migration is successful, the guest in the
source machine will be destroyed and large sptes will be created in the
destination. However, the guest continues to run in the source machine
(for example if live migration fails), small sptes will remain around
and cause bad performance.
This patch introduce lazy collapsing of small sptes into large sptes.
The rmap will be scanned in ioctl context when dirty logging is stopped,
dropping those sptes which can be collapsed into a single large-page spte.
Later page faults will create the large-page sptes.
Reviewed-by: Xiao Guangrong <guangrong.xiao@linux.intel.com>
Signed-off-by: Wanpeng Li <wanpeng.li@linux.intel.com>
Message-Id: <1428046825-6905-1-git-send-email-wanpeng.li@linux.intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
DR0-DR3 are not cleared as they should during reset and when they are set from
userspace. It appears to be caused by c77fb5fe6f ("KVM: x86: Allow the guest
to run with dirty debug registers").
Force their reload on these situations.
Signed-off-by: Nadav Amit <namit@cs.technion.ac.il>
Message-Id: <1427933438-12782-4-git-send-email-namit@cs.technion.ac.il>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
recalculate_apic_map() uses two passes over all VCPUs. This is a relic
from time when we selected a global mode in the first pass and set up
the optimized table in the second pass (to have a consistent mode).
Recent changes made mixed mode unoptimized and we can do it in one pass.
Format of logical MDA is a function of the mode, so we encode it in
apic_logical_id() and drop obsoleted variables from the struct.
Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
Message-Id: <1423766494-26150-5-git-send-email-rkrcmar@redhat.com>
[Add lid_bits temporary in apic_logical_id. - Paolo]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
We want to support mixed modes and the easiest solution is to avoid
optimizing those weird and unlikely scenarios.
Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
Message-Id: <1423766494-26150-4-git-send-email-rkrcmar@redhat.com>
[Add comment above KVM_APIC_MODE_* defines. - Paolo]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Broadcast allowed only one global APIC mode, but mixed modes are
theoretically possible. x2APIC IPI doesn't mean 0xff as broadcast,
the rest does.
x2APIC broadcasts are accepted by xAPIC. If we take SDM to be logical,
even addreses beginning with 0xff should be accepted, but real hardware
disagrees. This patch aims for simple code by considering most of real
behavior as undefined.
Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
Message-Id: <1423766494-26150-3-git-send-email-rkrcmar@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
cpuid_maxphyaddr(), which performs lot of memory accesses is called
extensively across KVM, especially in nVMX code.
This patch adds a cached value of maxphyaddr to vcpu.arch to reduce the
pressure onto CPU cache and simplify the code of cpuid_maxphyaddr()
callers. The cached value is initialized in kvm_arch_vcpu_init() and
reloaded every time CPUID is updated by usermode. It is obvious that
these reloads occur infrequently.
Signed-off-by: Eugene Korenevsky <ekorenevsky@gmail.com>
Message-Id: <20150329205612.GA1223@gnote>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Some constants are redfined in emulate.c. Avoid it.
s/SELECTOR_RPL_MASK/SEGMENT_RPL_MASK
s/SELECTOR_TI_MASK/SEGMENT_TI_MASK
No functional change.
Signed-off-by: Nadav Amit <namit@cs.technion.ac.il>
Message-Id: <1427635984-8113-3-git-send-email-namit@cs.technion.ac.il>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The eflags are redefined (using other defines) in emulate.c.
Use the definition from processor-flags.h as some mess already started.
No functional change.
Signed-off-by: Nadav Amit <namit@cs.technion.ac.il>
Message-Id: <1427635984-8113-2-git-send-email-namit@cs.technion.ac.il>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Currently kvm_emulate() skips the instruction but kvm_emulate_* sometimes
don't. The end reult is the caller ends up doing the skip themselves.
Let's make them consistant.
Signed-off-by: Joel Schopp <joel.schopp@amd.com>
Reviewed-by: Radim Krčmář <rkrcmar@redhat.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
This patch introduces a new module parameter for the KVM module; when it
is present, KVM attempts a bit of polling on every HLT before scheduling
itself out via kvm_vcpu_block.
This parameter helps a lot for latency-bound workloads---in particular
I tested it with O_DSYNC writes with a battery-backed disk in the host.
In this case, writes are fast (because the data doesn't have to go all
the way to the platters) but they cannot be merged by either the host or
the guest. KVM's performance here is usually around 30% of bare metal,
or 50% if you use cache=directsync or cache=writethrough (these
parameters avoid that the guest sends pointless flush requests, and
at the same time they are not slow because of the battery-backed cache).
The bad performance happens because on every halt the host CPU decides
to halt itself too. When the interrupt comes, the vCPU thread is then
migrated to a new physical CPU, and in general the latency is horrible
because the vCPU thread has to be scheduled back in.
With this patch performance reaches 60-65% of bare metal and, more
important, 99% of what you get if you use idle=poll in the guest. This
means that the tunable gets rid of this particular bottleneck, and more
work can be done to improve performance in the kernel or QEMU.
Of course there is some price to pay; every time an otherwise idle vCPUs
is interrupted by an interrupt, it will poll unnecessarily and thus
impose a little load on the host. The above results were obtained with
a mostly random value of the parameter (500000), and the load was around
1.5-2.5% CPU usage on one of the host's core for each idle guest vCPU.
The patch also adds a new stat, /sys/kernel/debug/kvm/halt_successful_poll,
that can be used to tune the parameter. It counts how many HLT
instructions received an interrupt during the polling period; each
successful poll avoids that Linux schedules the VCPU thread out and back
in, and may also avoid a likely trip to C1 and back for the physical CPU.
While the VM is idle, a Linux 4 VCPU VM halts around 10 times per second.
Of these halts, almost all are failed polls. During the benchmark,
instead, basically all halts end within the polling period, except a more
or less constant stream of 50 per second coming from vCPUs that are not
running the benchmark. The wasted time is thus very low. Things may
be slightly different for Windows VMs, which have a ~10 ms timer tick.
The effect is also visible on Marcelo's recently-introduced latency
test for the TSC deadline timer. Though of course a non-RT kernel has
awful latency bounds, the latency of the timer is around 8000-10000 clock
cycles compared to 20000-120000 without setting halt_poll_ns. For the TSC
deadline timer, thus, the effect is both a smaller average latency and
a smaller variance.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Revert 7c6a98dfa1, given
that testing PIR is not necessary anymore.
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
David Hildenbrand