Pull scheduler updates from Ingo Molnar:
"The main changes in this cycle are:
- Make schedstats a runtime tunable (disabled by default) and
optimize it via static keys.
As most distributions enable CONFIG_SCHEDSTATS=y due to its
instrumentation value, this is a nice performance enhancement.
(Mel Gorman)
- Implement 'simple waitqueues' (swait): these are just pure
waitqueues without any of the more complex features of full-blown
waitqueues (callbacks, wake flags, wake keys, etc.). Simple
waitqueues have less memory overhead and are faster.
Use simple waitqueues in the RCU code (in 4 different places) and
for handling KVM vCPU wakeups.
(Peter Zijlstra, Daniel Wagner, Thomas Gleixner, Paul Gortmaker,
Marcelo Tosatti)
- sched/numa enhancements (Rik van Riel)
- NOHZ performance enhancements (Rik van Riel)
- Various sched/deadline enhancements (Steven Rostedt)
- Various fixes (Peter Zijlstra)
- ... and a number of other fixes, cleanups and smaller enhancements"
* 'sched-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (29 commits)
sched/cputime: Fix steal_account_process_tick() to always return jiffies
sched/deadline: Remove dl_new from struct sched_dl_entity
Revert "kbuild: Add option to turn incompatible pointer check into error"
sched/deadline: Remove superfluous call to switched_to_dl()
sched/debug: Fix preempt_disable_ip recording for preempt_disable()
sched, time: Switch VIRT_CPU_ACCOUNTING_GEN to jiffy granularity
time, acct: Drop irq save & restore from __acct_update_integrals()
acct, time: Change indentation in __acct_update_integrals()
sched, time: Remove non-power-of-two divides from __acct_update_integrals()
sched/rt: Kick RT bandwidth timer immediately on start up
sched/debug: Add deadline scheduler bandwidth ratio to /proc/sched_debug
sched/debug: Move sched_domain_sysctl to debug.c
sched/debug: Move the /sys/kernel/debug/sched_features file setup into debug.c
sched/rt: Fix PI handling vs. sched_setscheduler()
sched/core: Remove duplicated sched_group_set_shares() prototype
sched/fair: Consolidate nohz CPU load update code
sched/fair: Avoid using decay_load_missed() with a negative value
sched/deadline: Always calculate end of period on sched_yield()
sched/cgroup: Fix cgroup entity load tracking tear-down
rcu: Use simple wait queues where possible in rcutree
...
Returning directly whatever copy_to_user(...) or copy_from_user(...)
returns may not do the right thing if there's a pagefault:
copy_to_user/copy_from_user return the number of bytes not copied in
this case, but ioctls need to return -EFAULT instead.
Fix up kvm on mips to do
return copy_to_user(...)) ? -EFAULT : 0;
and
return copy_from_user(...)) ? -EFAULT : 0;
everywhere.
Cc: stable@vger.kernel.org
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The problem:
On -rt, an emulated LAPIC timer instances has the following path:
1) hard interrupt
2) ksoftirqd is scheduled
3) ksoftirqd wakes up vcpu thread
4) vcpu thread is scheduled
This extra context switch introduces unnecessary latency in the
LAPIC path for a KVM guest.
The solution:
Allow waking up vcpu thread from hardirq context,
thus avoiding the need for ksoftirqd to be scheduled.
Normal waitqueues make use of spinlocks, which on -RT
are sleepable locks. Therefore, waking up a waitqueue
waiter involves locking a sleeping lock, which
is not allowed from hard interrupt context.
cyclictest command line:
This patch reduces the average latency in my tests from 14us to 11us.
Daniel writes:
Paolo asked for numbers from kvm-unit-tests/tscdeadline_latency
benchmark on mainline. The test was run 1000 times on
tip/sched/core 4.4.0-rc8-01134-g0905f04:
./x86-run x86/tscdeadline_latency.flat -cpu host
with idle=poll.
The test seems not to deliver really stable numbers though most of
them are smaller. Paolo write:
"Anything above ~10000 cycles means that the host went to C1 or
lower---the number means more or less nothing in that case.
The mean shows an improvement indeed."
Before:
min max mean std
count 1000.000000 1000.000000 1000.000000 1000.000000
mean 5162.596000 2019270.084000 5824.491541 20681.645558
std 75.431231 622607.723969 89.575700 6492.272062
min 4466.000000 23928.000000 5537.926500 585.864966
25% 5163.000000 1613252.750000 5790.132275 16683.745433
50% 5175.000000 2281919.000000 5834.654000 23151.990026
75% 5190.000000 2382865.750000 5861.412950 24148.206168
max 5228.000000 4175158.000000 6254.827300 46481.048691
After
min max mean std
count 1000.000000 1000.00000 1000.000000 1000.000000
mean 5143.511000 2076886.10300 5813.312474 21207.357565
std 77.668322 610413.09583 86.541500 6331.915127
min 4427.000000 25103.00000 5529.756600 559.187707
25% 5148.000000 1691272.75000 5784.889825 17473.518244
50% 5160.000000 2308328.50000 5832.025000 23464.837068
75% 5172.000000 2393037.75000 5853.177675 24223.969976
max 5222.000000 3922458.00000 6186.720500 42520.379830
[Patch was originaly based on the swait implementation found in the -rt
tree. Daniel ported it to mainline's version and gathered the
benchmark numbers for tscdeadline_latency test.]
Signed-off-by: Daniel Wagner <daniel.wagner@bmw-carit.de>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: linux-rt-users@vger.kernel.org
Cc: Boqun Feng <boqun.feng@gmail.com>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Paul Gortmaker <paul.gortmaker@windriver.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
Link: http://lkml.kernel.org/r/1455871601-27484-4-git-send-email-wagi@monom.org
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Pull MIPS updates from Ralf Baechle:
"This is the main pull request for MIPS for 4.5 plus some 4.4 fixes.
The executive summary:
- ATH79 platform improvments, use DT bindings for the ATH79 USB PHY.
- Avoid useless rebuilds for zboot.
- jz4780: Add NEMC, BCH and NAND device tree nodes
- Initial support for the MicroChip's DT platform. As all the device
drivers are missing this is still of limited use.
- Some Loongson3 cleanups.
- The unavoidable whitespace polishing.
- Reduce clock skew when synchronizing the CPU cycle counters on CPU
startup.
- Add MIPS R6 fixes.
- Lots of cleanups across arch/mips as fallout from KVM.
- Lots of minor fixes and changes for IEEE 754-2008 support to the
FPU emulator / fp-assist software.
- Minor Ralink, BCM47xx and bcm963xx platform support improvments.
- Support SMP on BCM63168"
* 'upstream' of git://git.linux-mips.org/pub/scm/ralf/upstream-linus: (84 commits)
MIPS: zboot: Add support for serial debug using the PROM
MIPS: zboot: Avoid useless rebuilds
MIPS: BMIPS: Enable ARCH_WANT_OPTIONAL_GPIOLIB
MIPS: bcm63xx: nvram: Remove unused bcm63xx_nvram_get_psi_size() function
MIPS: bcm963xx: Update bcm_tag field image_sequence
MIPS: bcm963xx: Move extended flash address to bcm_tag header file
MIPS: bcm963xx: Move Broadcom BCM963xx image tag data structure
MIPS: bcm63xx: nvram: Use nvram structure definition from header file
MIPS: bcm963xx: Add Broadcom BCM963xx board nvram data structure
MAINTAINERS: Add KVM for MIPS entry
MIPS: KVM: Add missing newline to kvm_err()
MIPS: Move KVM specific opcodes into asm/inst.h
MIPS: KVM: Use cacheops.h definitions
MIPS: Break down cacheops.h definitions
MIPS: Use EXCCODE_ constants with set_except_vector()
MIPS: Update trap codes
MIPS: Move Cause.ExcCode trap codes to mipsregs.h
MIPS: KVM: Make kvm_mips_{init,exit}() static
MIPS: KVM: Refactor added offsetof()s
MIPS: KVM: Convert EXPORT_SYMBOL to _GPL
...
The header arch/mips/kvm/opcode.h defines a few extra opcodes which
aren't in arch/mips/include/uapi/asm/inst.h. There's nothing KVM
specific about them, so lets move them into inst.h where they belong and
delete the header.
Note that mfmcz_op is renamed to mfmc0_op to match the instruction set
manual, and wait_op was already added to inst.h in commit b0a3eae2b9
("MIPS: inst.h: define COP0 wait op"), merged in v3.16-rc1.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Gleb Natapov <gleb@kernel.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Patchwork: https://patchwork.linux-mips.org/patch/11895/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Drop the custom cache operation code definitions used by KVM for
emulating guest CACHE instructions, and switch to use the existing
definitions in <asm/cacheops.h>.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Gleb Natapov <gleb@kernel.org>
Cc: kvm@vger.kernel.org
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/11893/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Move the Cause.ExcCode trap code definitions from kvm_host.h to
mipsregs.h, since they describe architectural bits rather than KVM
specific constants, and change the prefix from T_ to EXCCODE_.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Gleb Natapov <gleb@kernel.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Patchwork: https://patchwork.linux-mips.org/patch/11891/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
The module init and exit functions have no need to be global, so make
them static.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Gleb Natapov <gleb@kernel.org>
Cc: kvm@vger.kernel.org
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/11889/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
When calculating the offsets into the commpage for dynamically
translated mtc0/mfc0 guest instructions, multiple offsetof()s are added
together to find the offset of the specific register in the mips_coproc,
within the commpage.
Simplify each of these cases to a single offsetof() to find the offset
of the specific register within the commpage.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Gleb Natapov <gleb@kernel.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Patchwork: https://patchwork.linux-mips.org/patch/11888/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Export symbols only to GPL modules to match other KVM symbols in
virt/kvm/ and arch/*/kvm/.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Gleb Natapov <gleb@kernel.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Patchwork: https://patchwork.linux-mips.org/patch/11887/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
The function kvm_mips_host_tlb_inv_index() is unused, so drop it
completely.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Gleb Natapov <gleb@kernel.org>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Patchwork: https://patchwork.linux-mips.org/patch/11886/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
To date, we have implemented two I/O usage models for persistent memory,
PMEM (a persistent "ram disk") and DAX (mmap persistent memory into
userspace). This series adds a third, DAX-GUP, that allows DAX mappings
to be the target of direct-i/o. It allows userspace to coordinate
DMA/RDMA from/to persistent memory.
The implementation leverages the ZONE_DEVICE mm-zone that went into
4.3-rc1 (also discussed at kernel summit) to flag pages that are owned
and dynamically mapped by a device driver. The pmem driver, after
mapping a persistent memory range into the system memmap via
devm_memremap_pages(), arranges for DAX to distinguish pfn-only versus
page-backed pmem-pfns via flags in the new pfn_t type.
The DAX code, upon seeing a PFN_DEV+PFN_MAP flagged pfn, flags the
resulting pte(s) inserted into the process page tables with a new
_PAGE_DEVMAP flag. Later, when get_user_pages() is walking ptes it keys
off _PAGE_DEVMAP to pin the device hosting the page range active.
Finally, get_page() and put_page() are modified to take references
against the device driver established page mapping.
Finally, this need for "struct page" for persistent memory requires
memory capacity to store the memmap array. Given the memmap array for a
large pool of persistent may exhaust available DRAM introduce a
mechanism to allocate the memmap from persistent memory. The new
"struct vmem_altmap *" parameter to devm_memremap_pages() enables
arch_add_memory() to use reserved pmem capacity rather than the page
allocator.
This patch (of 18):
The core has developed a need for a "pfn_t" type [1]. Move the existing
pfn_t in KVM to kvm_pfn_t [2].
[1]: https://lists.01.org/pipermail/linux-nvdimm/2015-September/002199.html
[2]: https://lists.01.org/pipermail/linux-nvdimm/2015-September/002218.html
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Acked-by: Christoffer Dall <christoffer.dall@linaro.org>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
If either of the memory allocations in kvm_arch_vcpu_create() fail, the
vcpu which has been allocated and kvm_vcpu_init'd doesn't get uninit'd
in the error handling path. Add a call to kvm_vcpu_uninit() to fix this.
Fixes: 669e846e6c ("KVM/MIPS32: MIPS arch specific APIs for KVM")
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Gleb Natapov <gleb@kernel.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Cc: <stable@vger.kernel.org> # 3.10.x-
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The immediate field of the CACHE instruction is signed, so ensure that
it gets sign extended by casting it to an int16_t rather than just
masking the low 16 bits.
Fixes: e685c689f3 ("KVM/MIPS32: Privileged instruction/target branch emulation.")
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Gleb Natapov <gleb@kernel.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Cc: <stable@vger.kernel.org> # 3.10.x-
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
ASID restoration on guest resume should determine the guest execution
mode based on the guest Status register rather than bit 30 of the guest
PC.
Fix the two places in locore.S that do this, loading the guest status
from the cop0 area. Note, this assembly is specific to the trap &
emulate implementation of KVM, so it doesn't need to check the
supervisor bit as that mode is not implemented in the guest.
Fixes: b680f70fc1 ("KVM/MIPS32: Entry point for trampolining to...")
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Gleb Natapov <gleb@kernel.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Cc: <stable@vger.kernel.org> # 3.10.x-
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Remove the definition in locore.S and move a few of the other similar
definitions in asm/mipsregs.h too. CP0_INTCTL, CP0_SRSCTL, & CP0_SRSMAP
are unused so they're just dropped instead. CP0_DDATA_LO is left where
it is as I have patches to eliminate its use in locore.S and it
otherwise is unlikely to need to be used from assembly code.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Paul Burton <paul.burton@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Gleb Natapov <gleb@kernel.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Patchwork: https://patchwork.linux-mips.org/patch/11461/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
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>
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
...
Fix possible unintended sign extension in unsigned MMIO loads by casting
to uint16_t in the case of mmio_needed != 2.
Signed-off-by: Nicholas Mc Guire <hofrat@osadl.org>
Reviewed-by: James Hogan <james.hogan@imgtec.com>
Tested-by: James Hogan <james.hogan@imgtec.com>
Cc: Gleb Natapov <gleb@kernel.org>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: kvm@vger.kernel.org
Cc: linux-mips@linux-mips.org
Cc: linux-kernel@vger.kernel.org
Patchwork: https://patchwork.linux-mips.org/patch/9985/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
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>
Architecture-specific helpers are not supposed to muck with
struct kvm_userspace_memory_region contents. Add const to
enforce this.
In order to eliminate the only write in __kvm_set_memory_region,
the cleaning of deleted slots is pulled up from update_memslots
to __kvm_set_memory_region.
Reviewed-by: Takuya Yoshikawa <yoshikawa_takuya_b1@lab.ntt.co.jp>
Reviewed-by: Radim Krcmar <rkrcmar@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
kvm_memslots provides lockdep checking. Use it consistently instead of
explicit dereferencing of kvm->memslots.
Reviewed-by: Radim Krcmar <rkrcmar@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The argument to KVM_GET_DIRTY_LOG is a memslot id; it may not match the
position in the memslots array, which is sorted by gfn.
Cc: stable@vger.kernel.org
Reviewed-by: James Hogan <james.hogan@imgtec.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Use __kvm_guest_{enter|exit} instead of kvm_guest_{enter|exit}
where interrupts are disabled.
Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Now that the code is in place for KVM to support MIPS SIMD Architecutre
(MSA) in MIPS guests, wire up the new KVM_CAP_MIPS_MSA capability.
For backwards compatibility, the capability must be explicitly enabled
in order to detect or make use of MSA from the guest.
The capability is not supported if the hardware supports MSA vector
partitioning, since the extra support cannot be tested yet and it
extends the state that the userland program would have to save.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Gleb Natapov <gleb@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Cc: linux-api@vger.kernel.org
Cc: linux-doc@vger.kernel.org
Add KVM register numbers for the MIPS SIMD Architecture (MSA) registers,
and implement access to them with the KVM_GET_ONE_REG / KVM_SET_ONE_REG
ioctls when the MSA capability is enabled (exposed in a later patch) and
present in the guest according to its Config3.MSAP bit.
The MSA vector registers use the same register numbers as the FPU
registers except with a different size (128bits). Since MSA depends on
Status.FR=1, these registers are inaccessible when Status.FR=0. These
registers are returned as a single native endian 128bit value, rather
than least significant half first with each 64-bit half native endian as
the kernel uses internally.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Paul Burton <paul.burton@imgtec.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Gleb Natapov <gleb@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Cc: linux-api@vger.kernel.org
Cc: linux-doc@vger.kernel.org
Add guest exception handling for MIPS SIMD Architecture (MSA) floating
point exceptions and MSA disabled exceptions.
MSA floating point exceptions from the guest need passing to the guest
kernel, so for these a guest MSAFPE is emulated.
MSA disabled exceptions are normally handled by passing a reserved
instruction exception to the guest (because no guest MSA was supported),
but the hypervisor can now handle them if the guest has MSA by passing
an MSA disabled exception to the guest, or if the guest has MSA enabled
by transparently restoring the guest MSA context and enabling MSA and
the FPU.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Paul Burton <paul.burton@imgtec.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Gleb Natapov <gleb@kernel.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Emulate MSA related parts of COP0 interface so that the guest will be
able to enable/disable MSA (Config5.MSAEn) once the MSA capability has
been wired up.
As with the FPU (Status.CU1) setting Config5.MSAEn has no immediate
effect if the MSA state isn't live, as MSA state is restored lazily on
first use. Changes after the MSA state has been restored take immediate
effect, so that the guest can start getting MSA disabled exceptions
right away for guest MSA operations. The MSA state is saved lazily too,
as MSA may get re-enabled in the near future anyway.
A special case is also added for when Status.CU1 is set while FR=0 and
the MSA state is live. In this case we are at risk of getting reserved
instruction exceptions if we try and save the MSA state, so we lose the
MSA state sooner while MSA is still usable.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Paul Burton <paul.burton@imgtec.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Gleb Natapov <gleb@kernel.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Add base code for supporting the MIPS SIMD Architecture (MSA) in MIPS
KVM guests. MSA cannot yet be enabled in the guest, we're just laying
the groundwork.
As with the FPU, whether the guest's MSA context is loaded is stored in
another bit in the fpu_inuse vcpu member. This allows MSA to be disabled
when the guest disables it, but keeping the MSA context loaded so it
doesn't have to be reloaded if the guest re-enables it.
New assembly code is added for saving and restoring the MSA context,
restoring only the upper half of the MSA context (for if the FPU context
is already loaded) and for saving/clearing and restoring MSACSR (which
can itself cause an MSA FP exception depending on the value). The MSACSR
is restored before returning to the guest if MSA is already enabled, and
the existing FP exception die notifier is extended to catch the possible
MSA FP exception and step over the ctcmsa instruction.
The helper function kvm_own_msa() is added to enable MSA and restore
the MSA context if it isn't already loaded, which will be used in a
later patch when the guest attempts to use MSA for the first time and
triggers an MSA disabled exception.
The existing FPU helpers are extended to handle MSA. kvm_lose_fpu()
saves the full MSA context if it is loaded (which includes the FPU
context) and both kvm_lose_fpu() and kvm_drop_fpu() disable MSA.
kvm_own_fpu() also needs to lose any MSA context if FR=0, since there
would be a risk of getting reserved instruction exceptions if CU1 is
enabled and we later try and save the MSA context. We shouldn't usually
hit this case since it will be handled when emulating CU1 changes,
however there's nothing to stop the guest modifying the Status register
directly via the comm page, which will cause this case to get hit.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Paul Burton <paul.burton@imgtec.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Gleb Natapov <gleb@kernel.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Now that the code is in place for KVM to support FPU in MIPS KVM guests,
wire up the new KVM_CAP_MIPS_FPU capability.
For backwards compatibility, the capability must be explicitly enabled
in order to detect or make use of the FPU from the guest.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Gleb Natapov <gleb@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Cc: linux-api@vger.kernel.org
Cc: linux-doc@vger.kernel.org
Add KVM register numbers for the MIPS FPU registers, and implement
access to them with the KVM_GET_ONE_REG / KVM_SET_ONE_REG ioctls when
the FPU capability is enabled (exposed in a later patch) and present in
the guest according to its Config1.FP bit.
The registers are accessible in the current mode of the guest, with each
sized access showing what the guest would see with an equivalent access,
and like the architecture they may become UNPREDICTABLE if the FR mode
is changed. When FR=0, odd doubles are inaccessible as they do not exist
in that mode.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Cc: Paul Burton <paul.burton@imgtec.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Gleb Natapov <gleb@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Cc: linux-api@vger.kernel.org
Cc: linux-doc@vger.kernel.org
Add guest exception handling for floating point exceptions and
coprocessor 1 unusable exceptions.
Floating point exceptions from the guest need passing to the guest
kernel, so for these a guest FPE is emulated.
Also, coprocessor 1 unusable exceptions are normally passed straight
through to the guest (because no guest FPU was supported), but the
hypervisor can now handle them if the guest has its FPU enabled by
restoring the guest FPU context and enabling the FPU.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Paul Burton <paul.burton@imgtec.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Gleb Natapov <gleb@kernel.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Emulate FPU related parts of COP0 interface so that the guest will be
able to enable/disable the following once the FPU capability has been
wired up:
- The FPU (Status.CU1)
- 64-bit FP register mode (Status.FR)
- Hybrid FP register mode (Config5.FRE)
Changing Status.CU1 has no immediate effect if the FPU state isn't live,
as the FPU state is restored lazily on first use. After that, changes
take place immediately in the host Status.CU1, so that the guest can
start getting coprocessor unusable exceptions right away for guest FPU
operations if it is disabled. The FPU state is saved lazily too, as the
FPU may get re-enabled in the near future anyway.
Any change to Status.FR causes the FPU state to be discarded and FPU
disabled, as the register state is architecturally UNPREDICTABLE after
such a change. This should also ensure that the FPU state is fully
initialised (with stale state, but that's fine) when it is next used in
the new FP mode.
Any change to the Config5.FRE bit is immediately updated in the host
state so that the guest can get the relevant exceptions right away for
single-precision FPU operations.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Paul Burton <paul.burton@imgtec.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Gleb Natapov <gleb@kernel.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Add base code for supporting FPU in MIPS KVM guests. The FPU cannot yet
be enabled in the guest, we're just laying the groundwork.
Whether the guest's FPU context is loaded is stored in a bit in the
fpu_inuse vcpu member. This allows the FPU to be disabled when the guest
disables it, but keeping the FPU context loaded so it doesn't have to be
reloaded if the guest re-enables it.
An fpu_enabled vcpu member stores whether userland has enabled the FPU
capability (which will be wired up in a later patch).
New assembly code is added for saving and restoring the FPU context, and
for saving/clearing and restoring FCSR (which can itself cause an FP
exception depending on the value). The FCSR is restored before returning
to the guest if the FPU is already enabled, and a die notifier is
registered to catch the possible FP exception and step over the ctc1
instruction.
The helper function kvm_lose_fpu() is added to save FPU context and
disable the FPU, which is used when saving hardware state before a
context switch or KVM exit (the vcpu_get_regs() callback).
The helper function kvm_own_fpu() is added to enable the FPU and restore
the FPU context if it isn't already loaded, which will be used in a
later patch when the guest attempts to use the FPU for the first time
and triggers a co-processor unusable exception.
The helper function kvm_drop_fpu() is added to discard the FPU context
and disable the FPU, which will be used in a later patch when the FPU
state will become architecturally UNPREDICTABLE (change of FR mode) to
force a reload of [stale] context in the new FR mode.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Paul Burton <paul.burton@imgtec.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Gleb Natapov <gleb@kernel.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Add a vcpu_get_regs() and vcpu_set_regs() callbacks for loading and
restoring context which may be in hardware registers. This may include
floating point and MIPS SIMD Architecture (MSA) state which may be
accessed directly by the guest (but restored lazily by the hypervisor),
and also dedicated guest registers as provided by the VZ ASE.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Gleb Natapov <gleb@kernel.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Add Config4 and Config5 co-processor 0 registers, and add capability to
write the Config1, Config3, Config4, and Config5 registers using the KVM
API.
Only supported bits can be written, to minimise the chances of the guest
being given a configuration from e.g. QEMU that is inconsistent with
that being emulated, and as such the handling is in trap_emul.c as it
may need to be different for VZ. Currently the only modification
permitted is to make Config4 and Config5 exist via the M bits, but other
bits will be added for FPU and MSA support in future patches.
Care should be taken by userland not to change bits without fully
handling the possible extra state that may then exist and which the
guest may begin to use and depend on.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Gleb Natapov <gleb@kernel.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Various semi-used definitions exist in kvm_host.h for the default guest
config registers. Remove them and use the appropriate values directly
when initialising the Config registers.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Gleb Natapov <gleb@kernel.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
The information messages when the KVM module is loaded and unloaded are
a bit pointless and out of line with other architectures, so lets drop
them.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Gleb Natapov <gleb@kernel.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Sort the registers in the kvm_mips_get_reg() switch by register number,
which puts ERROREPC after the CONFIG registers.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Gleb Natapov <gleb@kernel.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Implement access to the guest Processor Identification CP0 register
using the KVM_GET_ONE_REG and KVM_SET_ONE_REG ioctls. This allows the
owning process to modify and read back the value that is exposed to the
guest in this register.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Gleb Natapov <gleb@kernel.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Trap instructions are used by Linux to implement BUG_ON(), however KVM
doesn't pass trap exceptions on to the guest if they occur in guest
kernel mode, instead triggering an internal error "Exception Code: 13,
not yet handled". The guest kernel then doesn't get a chance to print
the usual BUG message and stack trace.
Implement handling of the trap exception so that it gets passed to the
guest and the user is left with a more useful log message.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Gleb Natapov <gleb@kernel.org>
Cc: kvm@vger.kernel.org
Cc: linux-mips@linux-mips.org
Guest user mode can generate a guest MSA Disabled exception on an MSA
capable core by simply trying to execute an MSA instruction. Since this
exception is unknown to KVM it will be passed on to the guest kernel.
However guest Linux kernels prior to v3.15 do not set up an exception
handler for the MSA Disabled exception as they don't support any MSA
capable cores. This results in a guest OS panic.
Since an older processor ID may be being emulated, and MSA support is
not advertised to the guest, the correct behaviour is to generate a
Reserved Instruction exception in the guest kernel so it can send the
guest process an illegal instruction signal (SIGILL), as would happen
with a non-MSA-capable core.
Fix this as minimally as reasonably possible by preventing
kvm_mips_check_privilege() from relaying MSA Disabled exceptions from
guest user mode to the guest kernel, and handling the MSA Disabled
exception by emulating a Reserved Instruction exception in the guest,
via a new handle_msa_disabled() KVM callback.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Paul Burton <paul.burton@imgtec.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Gleb Natapov <gleb@kernel.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Cc: <stable@vger.kernel.org> # v3.15+
Enable disabled interrupt, on unsuccessful operation.
Found by Coccinelle.
Signed-off-by: Tapasweni Pathak <tapaswenipathak@gmail.com>
Acked-by: Julia Lawall <julia.lawall@lip6.fr>
Reviewed-by: James Hogan <james.hogan@imgtec.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Currently the guest exit trace event saves the VCPU pointer to the
structure, and the guest PC is retrieved by dereferencing it when the
event is printed rather than directly from the trace record. This isn't
safe as the printing may occur long afterwards, after the PC has changed
and potentially after the VCPU has been freed. Usually this results in
the same (wrong) PC being printed for multiple trace events. It also
isn't portable as userland has no way to access the VCPU data structure
when interpreting the trace record itself.
Lets save the actual PC in the structure so that the correct value is
accessible later.
Fixes: 669e846e6c ("KVM/MIPS32: MIPS arch specific APIs for KVM")
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: Gleb Natapov <gleb@kernel.org>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Cc: <stable@vger.kernel.org> # v3.10+
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Common: Optional support for adding a small amount of polling on each HLT
instruction executed in the guest (or equivalent for other architectures).
This can improve latency up to 50% on some scenarios (e.g. O_DSYNC writes
or TCP_RR netperf tests). This also has to be enabled manually for now,
but the plan is to auto-tune this in the future.
ARM/ARM64: the highlights are support for GICv3 emulation and dirty page
tracking
s390: several optimizations and bugfixes. Also a first: a feature
exposed by KVM (UUID and long guest name in /proc/sysinfo) before
it is available in IBM's hypervisor! :)
MIPS: Bugfixes.
x86: Support for PML (page modification logging, a new feature in
Broadwell Xeons that speeds up dirty page tracking), nested virtualization
improvements (nested APICv---a nice optimization), usual round of emulation
fixes. There is also a new option to reduce latency of the TSC deadline
timer in the guest; this needs to be tuned manually.
Some commits are common between this pull and Catalin's; I see you
have already included his tree.
ARM has other conflicts where functions are added in the same place
by 3.19-rc and 3.20 patches. These are not large though, and entirely
within KVM.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.22 (GNU/Linux)
iQEcBAABAgAGBQJU28rkAAoJEL/70l94x66DXqQH/1TDOfJIjW7P2kb0Sw7Fy1wi
cEX1KO/VFxAqc8R0E/0Wb55CXyPjQJM6xBXuFr5cUDaIjQ8ULSktL4pEwXyyv/s5
DBDkN65mriry2w5VuEaRLVcuX9Wy+tqLQXWNkEySfyb4uhZChWWHvKEcgw5SqCyg
NlpeHurYESIoNyov3jWqvBjr4OmaQENyv7t2c6q5ErIgG02V+iCux5QGbphM2IC9
LFtPKxoqhfeB2xFxTOIt8HJiXrZNwflsTejIlCl/NSEiDVLLxxHCxK2tWK/tUXMn
JfLD9ytXBWtNMwInvtFm4fPmDouv2VDyR0xnK2db+/axsJZnbxqjGu1um4Dqbak=
=7gdx
-----END PGP SIGNATURE-----
Merge tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm
Pull KVM update from Paolo Bonzini:
"Fairly small update, but there are some interesting new features.
Common:
Optional support for adding a small amount of polling on each HLT
instruction executed in the guest (or equivalent for other
architectures). This can improve latency up to 50% on some
scenarios (e.g. O_DSYNC writes or TCP_RR netperf tests). This
also has to be enabled manually for now, but the plan is to
auto-tune this in the future.
ARM/ARM64:
The highlights are support for GICv3 emulation and dirty page
tracking
s390:
Several optimizations and bugfixes. Also a first: a feature
exposed by KVM (UUID and long guest name in /proc/sysinfo) before
it is available in IBM's hypervisor! :)
MIPS:
Bugfixes.
x86:
Support for PML (page modification logging, a new feature in
Broadwell Xeons that speeds up dirty page tracking), nested
virtualization improvements (nested APICv---a nice optimization),
usual round of emulation fixes.
There is also a new option to reduce latency of the TSC deadline
timer in the guest; this needs to be tuned manually.
Some commits are common between this pull and Catalin's; I see you
have already included his tree.
Powerpc:
Nothing yet.
The KVM/PPC changes will come in through the PPC maintainers,
because I haven't received them yet and I might end up being
offline for some part of next week"
* tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm: (130 commits)
KVM: ia64: drop kvm.h from installed user headers
KVM: x86: fix build with !CONFIG_SMP
KVM: x86: emulate: correct page fault error code for NoWrite instructions
KVM: Disable compat ioctl for s390
KVM: s390: add cpu model support
KVM: s390: use facilities and cpu_id per KVM
KVM: s390/CPACF: Choose crypto control block format
s390/kernel: Update /proc/sysinfo file with Extended Name and UUID
KVM: s390: reenable LPP facility
KVM: s390: floating irqs: fix user triggerable endless loop
kvm: add halt_poll_ns module parameter
kvm: remove KVM_MMIO_SIZE
KVM: MIPS: Don't leak FPU/DSP to guest
KVM: MIPS: Disable HTW while in guest
KVM: nVMX: Enable nested posted interrupt processing
KVM: nVMX: Enable nested virtual interrupt delivery
KVM: nVMX: Enable nested apic register virtualization
KVM: nVMX: Make nested control MSRs per-cpu
KVM: nVMX: Enable nested virtualize x2apic mode
KVM: nVMX: Prepare for using hardware MSR bitmap
...
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>
