Pull x86 spinlock changes from Ingo Molnar:
"The biggest change here are paravirtualized ticket spinlocks (PV
spinlocks), which bring a nice speedup on various benchmarks.
The KVM host side will come to you via the KVM tree"
* 'x86-spinlocks-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
x86/kvm/guest: Fix sparse warning: "symbol 'klock_waiting' was not declared as static"
kvm: Paravirtual ticketlocks support for linux guests running on KVM hypervisor
kvm guest: Add configuration support to enable debug information for KVM Guests
kvm uapi: Add KICK_CPU and PV_UNHALT definition to uapi
xen, pvticketlock: Allow interrupts to be enabled while blocking
x86, ticketlock: Add slowpath logic
jump_label: Split jumplabel ratelimit
x86, pvticketlock: When paravirtualizing ticket locks, increment by 2
x86, pvticketlock: Use callee-save for lock_spinning
xen, pvticketlocks: Add xen_nopvspin parameter to disable xen pv ticketlocks
xen, pvticketlock: Xen implementation for PV ticket locks
xen: Defer spinlock setup until boot CPU setup
x86, ticketlock: Collapse a layer of functions
x86, ticketlock: Don't inline _spin_unlock when using paravirt spinlocks
x86, spinlock: Replace pv spinlocks with pv ticketlocks
Although the lock_spinning calls in the spinlock code are on the
uncommon path, their presence can cause the compiler to generate many
more register save/restores in the function pre/postamble, which is in
the fast path. To avoid this, convert it to using the pvops callee-save
calling convention, which defers all the save/restores until the actual
function is called, keeping the fastpath clean.
Signed-off-by: Jeremy Fitzhardinge <jeremy@goop.org>
Link: http://lkml.kernel.org/r/1376058122-8248-8-git-send-email-raghavendra.kt@linux.vnet.ibm.com
Reviewed-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Tested-by: Attilio Rao <attilio.rao@citrix.com>
Signed-off-by: Raghavendra K T <raghavendra.kt@linux.vnet.ibm.com>
Acked-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
Rather than outright replacing the entire spinlock implementation in
order to paravirtualize it, keep the ticket lock implementation but add
a couple of pvops hooks on the slow patch (long spin on lock, unlocking
a contended lock).
Ticket locks have a number of nice properties, but they also have some
surprising behaviours in virtual environments. They enforce a strict
FIFO ordering on cpus trying to take a lock; however, if the hypervisor
scheduler does not schedule the cpus in the correct order, the system can
waste a huge amount of time spinning until the next cpu can take the lock.
(See Thomas Friebel's talk "Prevent Guests from Spinning Around"
http://www.xen.org/files/xensummitboston08/LHP.pdf for more details.)
To address this, we add two hooks:
- __ticket_spin_lock which is called after the cpu has been
spinning on the lock for a significant number of iterations but has
failed to take the lock (presumably because the cpu holding the lock
has been descheduled). The lock_spinning pvop is expected to block
the cpu until it has been kicked by the current lock holder.
- __ticket_spin_unlock, which on releasing a contended lock
(there are more cpus with tail tickets), it looks to see if the next
cpu is blocked and wakes it if so.
When compiled with CONFIG_PARAVIRT_SPINLOCKS disabled, a set of stub
functions causes all the extra code to go away.
Results:
=======
setup: 32 core machine with 32 vcpu KVM guest (HT off) with 8GB RAM
base = 3.11-rc
patched = base + pvspinlock V12
+-----------------+----------------+--------+
dbench (Throughput in MB/sec. Higher is better)
+-----------------+----------------+--------+
| base (stdev %)|patched(stdev%) | %gain |
+-----------------+----------------+--------+
| 15035.3 (0.3) |15150.0 (0.6) | 0.8 |
| 1470.0 (2.2) | 1713.7 (1.9) | 16.6 |
| 848.6 (4.3) | 967.8 (4.3) | 14.0 |
| 652.9 (3.5) | 685.3 (3.7) | 5.0 |
+-----------------+----------------+--------+
pvspinlock shows benefits for overcommit ratio > 1 for PLE enabled cases,
and undercommits results are flat
Signed-off-by: Jeremy Fitzhardinge <jeremy@goop.org>
Link: http://lkml.kernel.org/r/1376058122-8248-2-git-send-email-raghavendra.kt@linux.vnet.ibm.com
Reviewed-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Tested-by: Attilio Rao <attilio.rao@citrix.com>
[ Raghavendra: Changed SPIN_THRESHOLD, fixed redefinition of arch_spinlock_t]
Signed-off-by: Raghavendra K T <raghavendra.kt@linux.vnet.ibm.com>
Acked-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
Pull x86 paravirt update from Ingo Molnar:
"Various paravirtualization related changes - the biggest one makes
guest support optional via CONFIG_HYPERVISOR_GUEST"
* 'x86-paravirt-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
x86, wakeup, sleep: Use pvops functions for changing GDT entries
x86, xen, gdt: Remove the pvops variant of store_gdt.
x86-32, gdt: Store/load GDT for ACPI S3 or hibernation/resume path is not needed
x86-64, gdt: Store/load GDT for ACPI S3 or hibernate/resume path is not needed.
x86: Make Linux guest support optional
x86, Kconfig: Move PARAVIRT_DEBUG into the paravirt menu
The two use-cases where we needed to store the GDT were during ACPI S3 suspend
and resume. As the patches:
x86/gdt/i386: store/load GDT for ACPI S3 or hibernation/resume path is not needed
x86/gdt/64-bit: store/load GDT for ACPI S3 or hibernate/resume path is not needed.
have demonstrated - there are other mechanism by which the GDT is
saved and reloaded during early resume path.
Hence we do not need to worry about the pvops call-chain for saving the
GDT and can and can eliminate it. The other areas where the store_gdt is
used are never going to be hit when running under the pvops platforms.
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Link: http://lkml.kernel.org/r/1365194544-14648-4-git-send-email-konrad.wilk@oracle.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
Invoking arch_flush_lazy_mmu_mode() results in calls to
preempt_enable()/disable() which may have performance impact.
Since lazy MMU is not used on bare metal we can patch away
arch_flush_lazy_mmu_mode() so that it is never called in such
environment.
[ hpa: the previous patch "Fix vmalloc_fault oops during lazy MMU
updates" may cause a minor performance regression on
bare metal. This patch resolves that performance regression. It is
somewhat unclear to me if this is a good -stable candidate. ]
Signed-off-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Link: http://lkml.kernel.org/r/1364045796-10720-2-git-send-email-konrad.wilk@oracle.com
Tested-by: Josh Boyer <jwboyer@redhat.com>
Tested-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Acked-by: Borislav Petkov <bp@suse.de>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
Cc: <stable@vger.kernel.org> SEE NOTE ABOVE
Pull x86/mm changes from Peter Anvin:
"The big change here is the patchset by Alex Shi to use INVLPG to flush
only the affected pages when we only need to flush a small page range.
It also removes the special INVALIDATE_TLB_VECTOR interrupts (32
vectors!) and replace it with an ordinary IPI function call."
Fix up trivial conflicts in arch/x86/include/asm/apic.h (added code next
to changed line)
* 'x86-mm-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
x86/tlb: Fix build warning and crash when building for !SMP
x86/tlb: do flush_tlb_kernel_range by 'invlpg'
x86/tlb: replace INVALIDATE_TLB_VECTOR by CALL_FUNCTION_VECTOR
x86/tlb: enable tlb flush range support for x86
mm/mmu_gather: enable tlb flush range in generic mmu_gather
x86/tlb: add tlb_flushall_shift knob into debugfs
x86/tlb: add tlb_flushall_shift for specific CPU
x86/tlb: fall back to flush all when meet a THP large page
x86/flush_tlb: try flush_tlb_single one by one in flush_tlb_range
x86/tlb_info: get last level TLB entry number of CPU
x86: Add read_mostly declaration/definition to variables from smp.h
x86: Define early read-mostly per-cpu macros
x86 has no flush_tlb_range support in instruction level. Currently the
flush_tlb_range just implemented by flushing all page table. That is not
the best solution for all scenarios. In fact, if we just use 'invlpg' to
flush few lines from TLB, we can get the performance gain from later
remain TLB lines accessing.
But the 'invlpg' instruction costs much of time. Its execution time can
compete with cr3 rewriting, and even a bit more on SNB CPU.
So, on a 512 4KB TLB entries CPU, the balance points is at:
(512 - X) * 100ns(assumed TLB refill cost) =
X(TLB flush entries) * 100ns(assumed invlpg cost)
Here, X is 256, that is 1/2 of 512 entries.
But with the mysterious CPU pre-fetcher and page miss handler Unit, the
assumed TLB refill cost is far lower then 100ns in sequential access. And
2 HT siblings in one core makes the memory access more faster if they are
accessing the same memory. So, in the patch, I just do the change when
the target entries is less than 1/16 of whole active tlb entries.
Actually, I have no data support for the percentage '1/16', so any
suggestions are welcomed.
As to hugetlb, guess due to smaller page table, and smaller active TLB
entries, I didn't see benefit via my benchmark, so no optimizing now.
My micro benchmark show in ideal scenarios, the performance improves 70
percent in reading. And in worst scenario, the reading/writing
performance is similar with unpatched 3.4-rc4 kernel.
Here is the reading data on my 2P * 4cores *HT NHM EP machine, with THP
'always':
multi thread testing, '-t' paramter is thread number:
with patch unpatched 3.4-rc4
./mprotect -t 1 14ns 24ns
./mprotect -t 2 13ns 22ns
./mprotect -t 4 12ns 19ns
./mprotect -t 8 14ns 16ns
./mprotect -t 16 28ns 26ns
./mprotect -t 32 54ns 51ns
./mprotect -t 128 200ns 199ns
Single process with sequencial flushing and memory accessing:
with patch unpatched 3.4-rc4
./mprotect 7ns 11ns
./mprotect -p 4096 -l 8 -n 10240
21ns 21ns
[ hpa: http://lkml.kernel.org/r/1B4B44D9196EFF41AE41FDA404FC0A100BFF94@SHSMSX101.ccr.corp.intel.com
has additional performance numbers. ]
Signed-off-by: Alex Shi <alex.shi@intel.com>
Link: http://lkml.kernel.org/r/1340845344-27557-3-git-send-email-alex.shi@intel.com
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
There were paravirt_ops hooks for the full register set variant of
{rd,wr}msr_safe which are actually not used by anyone anymore. Remove
them to make the code cleaner and avoid silent breakages when the pvops
members were uninitialized. This has been boot-tested natively and under
Xen with PVOPS enabled and disabled on one machine.
Signed-off-by: Andre Przywara <andre.przywara@amd.com>
Link: http://lkml.kernel.org/r/1338562358-28182-2-git-send-email-bp@amd64.org
Acked-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
* 'x86-vdso-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-tip:
x86-64: Rework vsyscall emulation and add vsyscall= parameter
x86-64: Wire up getcpu syscall
x86: Remove unnecessary compile flag tweaks for vsyscall code
x86-64: Add vsyscall:emulate_vsyscall trace event
x86-64: Add user_64bit_mode paravirt op
x86-64, xen: Enable the vvar mapping
x86-64: Work around gold bug 13023
x86-64: Move the "user" vsyscall segment out of the data segment.
x86-64: Pad vDSO to a page boundary
Three places in the kernel assume that the only long mode CPL 3
selector is __USER_CS. This is not true on Xen -- Xen's sysretq
changes cs to the magic value 0xe033.
Two of the places are corner cases, but as of "x86-64: Improve
vsyscall emulation CS and RIP handling"
(c9712944b2), vsyscalls will segfault
if called with Xen's extra CS selector. This causes a panic when
older init builds die.
It seems impossible to make Xen use __USER_CS reliably without
taking a performance hit on every system call, so this fixes the
tests instead with a new paravirt op. It's a little ugly because
ptrace.h can't include paravirt.h.
Signed-off-by: Andy Lutomirski <luto@mit.edu>
Link: http://lkml.kernel.org/r/f4fcb3947340d9e96ce1054a432f183f9da9db83.1312378163.git.luto@mit.edu
Reported-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
This patch adds a function pointer in one of the many paravirt_ops
structs, to allow guests to register a steal time function. Besides
a steal time function, we also declare two jump_labels. They will be
used to allow the steal time code to be easily bypassed when not
in use.
Signed-off-by: Glauber Costa <glommer@redhat.com>
Acked-by: Rik van Riel <riel@redhat.com>
Tested-by: Eric B Munson <emunson@mgebm.net>
CC: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>
CC: Peter Zijlstra <peterz@infradead.org>
CC: Anthony Liguori <aliguori@us.ibm.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
Paravirt ops pmd_update/pmd_update_defer/pmd_set_at. Not all might be
necessary (vmware needs pmd_update, Xen needs set_pmd_at, nobody needs
pmd_update_defer), but this is to keep full simmetry with pte paravirt
ops, which looks cleaner and simpler from a common code POV.
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Acked-by: Rik van Riel <riel@redhat.com>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
VMI was the only user of the alloc_pmd_clone hook, given that VMI
is now removed we can also remove this hook.
Signed-off-by: Alok N Kataria <akataria@vmware.com>
LKML-Reference: <1282608357.19396.36.camel@ank32.eng.vmware.com>
Cc: Jeremy Fitzhardinge <jeremy@xensource.com>
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
Now that both Xen and VMI disable allocations of PTE pages from high
memory this paravirt op serves no further purpose.
This effectively reverts ce6234b5 "add kmap_atomic_pte for mapping
highpte pages".
Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
LKML-Reference: <1267204562-11844-3-git-send-email-ian.campbell@citrix.com>
Acked-by: Alok Kataria <akataria@vmware.com>
Cc: Jeremy Fitzhardinge <jeremy@goop.org>
Cc: Ingo Molnar <mingo@elte.hu>
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
The raw_spin* namespace was taken by lockdep for the architecture
specific implementations. raw_spin_* would be the ideal name space for
the spinlocks which are not converted to sleeping locks in preempt-rt.
Linus suggested to convert the raw_ to arch_ locks and cleanup the
name space instead of using an artifical name like core_spin,
atomic_spin or whatever
No functional change.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Acked-by: David S. Miller <davem@davemloft.net>
Acked-by: Ingo Molnar <mingo@elte.hu>
Cc: linux-arch@vger.kernel.org
Bastian Blank reported a boot crash with stackprotector enabled,
and debugged it back to edx register corruption.
For historical reasons irq enable/disable/save/restore had special
calling sequences to make them more efficient. With the more
recent introduction of higher-level and more general optimisations
this is no longer necessary so we can just use the normal PVOP_
macros.
This fixes some residual bugs in the old implementations which left
edx liable to inadvertent clobbering. Also, fix some bugs in
__PVOP_VCALLEESAVE which were revealed by actual use.
Reported-by: Bastian Blank <bastian@waldi.eu.org>
Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>
Cc: Stable Kernel <stable@kernel.org>
Cc: Xen-devel <xen-devel@lists.xensource.com>
LKML-Reference: <4AD3BC9B.7040501@goop.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
* 'x86-platform-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip: (38 commits)
x86: Move get/set_wallclock to x86_platform_ops
x86: platform: Fix section annotations
x86: apic namespace cleanup
x86: Distangle ioapic and i8259
x86: Add Moorestown early detection
x86: Add hardware_subarch ID for Moorestown
x86: Add early platform detection
x86: Move tsc_init to late_time_init
x86: Move tsc_calibration to x86_init_ops
x86: Replace the now identical time_32/64.c by time.c
x86: time_32/64.c unify profile_pc
x86: Move calibrate_cpu to tsc.c
x86: Make timer setup and global variables the same in time_32/64.c
x86: Remove mca bus ifdef from timer interrupt
x86: Simplify timer_ack magic in time_32.c
x86: Prepare unification of time_32/64.c
x86: Remove do_timer hook
x86: Add timer_init to x86_init_ops
x86: Move percpu clockevents setup to x86_init_ops
x86: Move xen_post_allocator_init into xen_pagetable_setup_done
...
Fix up conflicts in arch/x86/include/asm/io_apic.h
get/set_wallclock() have already a set of platform dependent
implementations (default, EFI, paravirt). MRST will add another
variant.
Moving them to platform ops simplifies the existing code and minimizes
the effort to integrate new variants.
Signed-off-by: Feng Tang <feng.tang@intel.com>
LKML-Reference: <new-submission>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
The timer init code is convoluted with several quirks and the paravirt
timer chooser. Figuring out which code path is actually taken is not
for the faint hearted.
Move the numaq TSC quirk to tsc_pre_init x86_init_ops function and
replace the paravirt time chooser and the remaining x86 quirk with a
simple x86_init_ops function.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
paravirt overrides the setup of the default apic timers as per cpu
timers. Moorestown needs to override that as well.
Move it to x86_init_ops setup and create a separate x86_cpuinit struct
which holds the function for the secondary evtl. hotplugabble CPUs.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
We really do not need two paravirt/x86_init_ops functions which are
called in two consecutive source lines. Move the only user of
post_allocator_init into the already existing pagetable_setup_done
function.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Replace another obscure paravirt magic and move it to
x86_init_ops. Such a hook is also useful for embedded and special
hardware.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
ARCH_SETUP is a horrible leftover from the old arch/i386 mach support
code. It still has a lonely user in xen. Move it to x86_init_ops.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
irq_init is overridden by x86_quirks and by paravirts. Unify the whole
mess and make it an unconditional x86_init_ops function which defaults
to the standard function and can be overridden by the early platform
code.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
memory_setup is overridden by x86_quirks and by paravirts with weak
functions and quirks. Unify the whole mess and make it an
unconditional x86_init_ops function which defaults to the standard
function and can be overridden by the early platform code.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Split the monolithic asm/paravirt.h into separate paravirt.h (inlines and other
"active" definitions), and paravirt_types.h (types, constants and other "passive"
definitions). This makes it easier to use the type/constant definitions without
pulling in everything else and causing circular dependency problems.
[ Impact: cleanup ]
Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>
Linus Torvalds