At present ARMv8 event counters are limited to 32-bits, though by
using the CHAIN event it's possible to combine adjacent counters to
achieve 64-bits. The perf config1:0 bit can be set to use such a
configuration.
With the introduction of ARMv8.5-PMU support, all event counters can
now be used as 64-bit counters.
Let's enable 64-bit event counters where support exists. Unless the
user sets config1:0 we will adjust the counter value such that it
overflows upon 32-bit overflow. This follows the same behaviour as
the cycle counter which has always been (and remains) 64-bits.
Signed-off-by: Andrew Murray <andrew.murray@arm.com>
Reviewed-by: Suzuki K Poulose <suzuki.poulose@arm.com>
[Mark: fix ID field names, compare with 8.5 value]
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Will Deacon <will@kernel.org>
Reading this code bordered on painful, what with all the repetition and
pointless return values. More fundamentally, dribbling the hardware
enables and disables in one bit at a time incurs needless system
register overhead for chained events and on reset. We already use
bitmask values for the KVM hooks, so consolidate all the register
accesses to match, and make a reasonable saving in both source and
object code.
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Signed-off-by: Will Deacon <will@kernel.org>
Add support for matching the new PMUs. For now, this just wires them up
as generic PMUv3 such that people writing DTs for new SoCs can do the
right thing, and at least have architectural and raw events be usable.
We can come back and fill in event maps for sysfs and/or perf tools at
a later date.
Acked-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Signed-off-by: Will Deacon <will@kernel.org>
The PMU init callbacks are already drowning in boilerplate, so before
doubling the number of supported PMU models, give it a sensible refactor
to significantly reduce the bloat, both in source and object code.
Although nobody uses non-default sysfs attributes today, there's minimal
impact to preserving the notion that maybe, some day, somebody might, so
we may as well keep up appearances.
Acked-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Signed-off-by: Will Deacon <will@kernel.org>
For each PMU event, there is a ARMV8_EVENT_ATTR(xx, XX) and
&armv8_event_attr_xx.attr.attr. Let's redefine the ARMV8_EVENT_ATTR
to simplify the armv8_pmuv3_event_attrs.
Cc: Will Deacon <will@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Shaokun Zhang <zhangshaokun@hisilicon.com>
[will: Dropped unnecessary array syntax]
Signed-off-by: Will Deacon <will@kernel.org>
In perf_event.c we use smp_processor_id(), but we haven't included
<linux/smp.h> where it is defined, and rely on this being pulled in
via a transitive include. Let's make this more robust by including
<linux.smp.h> explicitly.
Acked-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Raphael Gault <raphael.gault@arm.com>
Signed-off-by: Will Deacon <will@kernel.org>
ARMV8_EVENT_ATTR_RESOLVE became unused after commit <4b1a9e6934ec>
("arm64/perf: Filter common events based on PMCEIDn_EL0").
Remove it.
Cc: Will Deacon <will@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Shaokun Zhang <zhangshaokun@hisilicon.com>
Signed-off-by: Will Deacon <will@kernel.org>
Based on 1 normalized pattern(s):
this program is free software you can redistribute it and or modify
it under the terms of the gnu general public license version 2 as
published by the free software foundation this program is
distributed in the hope that it will be useful but without any
warranty without even the implied warranty of merchantability or
fitness for a particular purpose see the gnu general public license
for more details you should have received a copy of the gnu general
public license along with this program if not see http www gnu org
licenses
extracted by the scancode license scanner the SPDX license identifier
GPL-2.0-only
has been chosen to replace the boilerplate/reference in 503 file(s).
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Alexios Zavras <alexios.zavras@intel.com>
Reviewed-by: Allison Randal <allison@lohutok.net>
Reviewed-by: Enrico Weigelt <info@metux.net>
Cc: linux-spdx@vger.kernel.org
Link: https://lkml.kernel.org/r/20190602204653.811534538@linutronix.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
* POWER: support for direct access to the POWER9 XIVE interrupt controller,
memory and performance optimizations.
* x86: support for accessing memory not backed by struct page, fixes and refactoring
* Generic: dirty page tracking improvements
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.22 (GNU/Linux)
iQEcBAABAgAGBQJc3qV/AAoJEL/70l94x66Dn3QH/jX1Bn0P/RZAIt4w0SySklSg
PqxUKDyBQqB9vN9Qeb9jWXAKPH2CtM3+up/rz7oRnBWp7qA6vXcC/R/QJYAvzdXE
nklsR/oYCsflR1KdlVYuDvvPCPP2fLBU5zfN83OsaBQ8fNRkm3gN+N5XQ2SbXbLy
Mo9tybS4otY201UAC96e8N0ipwwyCRpDneQpLcl+F5nH3RBt63cVbs04O+70MXn7
eT4I+8K3+Go7LATzT8hglD21D/7uvE31qQb6yr5L33IfhU4GB51RZzBXTNaAdY8n
hT1rMrRkAMAFWYZPQDfoMadjWU3i5DIfstKjDxOr9oTfuOEp5Z+GvJwvVnUDg1I=
=D0+p
-----END PGP SIGNATURE-----
Merge tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm
Pull KVM updates from Paolo Bonzini:
"ARM:
- support for SVE and Pointer Authentication in guests
- PMU improvements
POWER:
- support for direct access to the POWER9 XIVE interrupt controller
- memory and performance optimizations
x86:
- support for accessing memory not backed by struct page
- fixes and refactoring
Generic:
- dirty page tracking improvements"
* tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm: (155 commits)
kvm: fix compilation on aarch64
Revert "KVM: nVMX: Expose RDPMC-exiting only when guest supports PMU"
kvm: x86: Fix L1TF mitigation for shadow MMU
KVM: nVMX: Disable intercept for FS/GS base MSRs in vmcs02 when possible
KVM: PPC: Book3S: Remove useless checks in 'release' method of KVM device
KVM: PPC: Book3S HV: XIVE: Fix spelling mistake "acessing" -> "accessing"
KVM: PPC: Book3S HV: Make sure to load LPID for radix VCPUs
kvm: nVMX: Set nested_run_pending in vmx_set_nested_state after checks complete
tests: kvm: Add tests for KVM_SET_NESTED_STATE
KVM: nVMX: KVM_SET_NESTED_STATE - Tear down old EVMCS state before setting new state
tests: kvm: Add tests for KVM_CAP_MAX_VCPUS and KVM_CAP_MAX_CPU_ID
tests: kvm: Add tests to .gitignore
KVM: Introduce KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2
KVM: Fix kvm_clear_dirty_log_protect off-by-(minus-)one
KVM: Fix the bitmap range to copy during clear dirty
KVM: arm64: Fix ptrauth ID register masking logic
KVM: x86: use direct accessors for RIP and RSP
KVM: VMX: Use accessors for GPRs outside of dedicated caching logic
KVM: x86: Omit caching logic for always-available GPRs
kvm, x86: Properly check whether a pfn is an MMIO or not
...
With VHE different exception levels are used between the host (EL2) and
guest (EL1) with a shared exception level for userpace (EL0). We can take
advantage of this and use the PMU's exception level filtering to avoid
enabling/disabling counters in the world-switch code. Instead we just
modify the counter type to include or exclude EL0 at vcpu_{load,put} time.
We also ensure that trapped PMU system register writes do not re-enable
EL0 when reconfiguring the backing perf events.
This approach completely avoids blackout windows seen with !VHE.
Suggested-by: Christoffer Dall <christoffer.dall@arm.com>
Signed-off-by: Andrew Murray <andrew.murray@arm.com>
Acked-by: Will Deacon <will.deacon@arm.com>
Reviewed-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Add support for the :G and :H attributes in perf by handling the
exclude_host/exclude_guest event attributes.
We notify KVM of counters that we wish to be enabled or disabled on
guest entry/exit and thus defer from starting or stopping events based
on their event attributes.
With !VHE we switch the counters between host/guest at EL2. We are able
to eliminate counters counting host events on the boundaries of guest
entry/exit when using :G by filtering out EL2 for exclude_host. When
using !exclude_hv there is a small blackout window at the guest
entry/exit where host events are not captured.
Signed-off-by: Andrew Murray <andrew.murray@arm.com>
Acked-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
The armv8pmu_enable_event_counter function issues an isb instruction
after enabling a pair of counters - this doesn't provide any value
and is inconsistent with the armv8pmu_disable_event_counter.
In any case armv8pmu_enable_event_counter is always called with the
PMU stopped. Starting the PMU with armv8pmu_start results in an isb
instruction being issued prior to writing to PMCR_EL0.
Let's remove the unnecessary isb instruction.
Signed-off-by: Andrew Murray <andrew.murray@arm.com>
Reviewed-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Acked-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
The functions armv8pmu_read_counter() and armv8pmu_write_counter()
are `static inline` while they are only referenced when assigned
to a function pointer field in a `struct arm_pmu` instance.
The inline keyword is thus counter intuitive and shouldn't be used.
Acked-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Raphael Gault <raphael.gault@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
The comment for the armv8pmu_set_event_filter function suggests that
it only works for PMUv2 PMUs - this is incorrect.
Let's remove the incorrect comment.
Acked-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Andrew Murray <andrew.murray@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Fix up one typos: Onl -> Only
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Shaokun Zhang <zhangshaokun@hisilicon.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
There have been some additional events added to the PMU architecture
since Armv8.0, so expose them via our sysfs infrastructure.
Signed-off-by: Will Deacon <will.deacon@arm.com>
The PMU event numbers are split between perf_event.h and perf_event.c,
which makes it difficult to spot any gaps in the numbers which may be
allocated in the future.
This patch sorts the events numerically, adds some missing events and
moves the definitions into perf_event.h.
Signed-off-by: Will Deacon <will.deacon@arm.com>
We cannot distinguish reads from writes in our generic cache events, so
drop the WRITE entries and leave the READ entries pointing to the combined
read/write events, as is done by other CPUs and architectures.
Reported-by: Ganapatrao Kulkarni <Ganapatrao.Kulkarni@cavium.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Armv8.1 allocated the upper 32-bits of the PMCEID registers to describe
the common architectural and microarchitecture events beginning at 0x4000.
Add support for these registers to our probing code, so that we can
advertise the SPE events when they are supported by the CPU.
Signed-off-by: Will Deacon <will.deacon@arm.com>
As a hangover from when this code used a designated initialiser, we've
been using commas to terminate the arm_pmu field assignments. Whilst
harmless, it's also weird, so replace them with semicolons instead.
Signed-off-by: Will Deacon <will.deacon@arm.com>
It doesn't make sense for a perf event to be configured as a CHAIN event
in isolation, so extend the arm_pmu structure with a ->filter_match()
function to allow the backend PMU implementation to reject CHAIN events
early.
Cc: <stable@vger.kernel.org>
Reviewed-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
It is useful to get the running time of a thread. Doing so in an
efficient manner can be important for performance of user applications.
Avoiding system calls in `clock_gettime` when handling
CLOCK_THREAD_CPUTIME_ID is important. Other clocks are handled in the
VDSO, but CLOCK_THREAD_CPUTIME_ID falls back on the system call.
CLOCK_THREAD_CPUTIME_ID is not handled in the VDSO since it would have
costs associated with maintaining updated user space accessible time
offsets. These offsets have to be updated everytime the a thread is
scheduled/descheduled. However, for programs regularly checking the
running time of a thread, this is a performance improvement.
This patch takes a middle ground, and adds support for cap_user_time an
optional feature of the perf_event API. This way costs are only
incurred when the perf_event api is enabled. This is done the same way
as it is in x86.
Ultimately this allows calculating the thread running time in userspace
on aarch64 as follows (adapted from perf_event_open manpage):
u32 seq, time_mult, time_shift;
u64 running, count, time_offset, quot, rem, delta;
struct perf_event_mmap_page *pc;
pc = buf; // buf is the perf event mmaped page as documented in the API.
if (pc->cap_usr_time) {
do {
seq = pc->lock;
barrier();
running = pc->time_running;
count = readCNTVCT_EL0(); // Read ARM hardware clock.
time_offset = pc->time_offset;
time_mult = pc->time_mult;
time_shift = pc->time_shift;
barrier();
} while (pc->lock != seq);
quot = (count >> time_shift);
rem = count & (((u64)1 << time_shift) - 1);
delta = time_offset + quot * time_mult +
((rem * time_mult) >> time_shift);
running += delta;
// running now has the current nanosecond level thread time.
}
Summary of changes in the patch:
For aarch64 systems, make arch_perf_update_userpage update the timing
information stored in the perf_event page. Requiring the following
calculations:
- Calculate the appropriate time_mult, and time_shift factors to convert
ticks to nano seconds for the current clock frequency.
- Adjust the mult and shift factors to avoid shift factors of 32 bits.
(possibly unnecessary)
- The time_offset userspace should apply when doing calculations:
negative the current sched time (now), because time_running and
time_enabled fields of the perf_event page have just been updated.
Toggle bits to appropriate values:
- Enable cap_user_time
Signed-off-by: Michael O'Farrell <micpof@gmail.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Add support for 64bit event by using chained event counters
and 64bit cycle counters.
PMUv3 allows chaining a pair of adjacent 32-bit counters, effectively
forming a 64-bit counter. The low/even counter is programmed to count
the event of interest, and the high/odd counter is programmed to count
the CHAIN event, taken when the low/even counter overflows.
For CPU cycles, when 64bit mode is requested, the cycle counter
is used in 64bit mode. If the cycle counter is not available,
falls back to chaining.
Cc: Will Deacon <will.deacon@arm.com>
Acked-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
The arm64 PMU updates the event counters and reprograms the
counters in the overflow IRQ handler without disabling the
PMU. This could potentially cause skews in for group counters,
where the overflowed counters may potentially loose some event
counts, while they are reprogrammed. To prevent this, disable
the PMU while we process the counter overflows and enable it
right back when we are done.
This patch also moves the PMU stop/start routines to avoid a
forward declaration.
Suggested-by: Mark Rutland <mark.rutland@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Acked-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
armv8pmu_select_counter always returns the passed idx. So
let us make that void and get rid of the pointless checks.
Suggested-by: Mark Rutland <mark.rutland@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Acked-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
The armpmu uses get_event_idx callback to allocate an event
counter for a given event, which marks the selected counter
as "used". Now, when we delete the counter, the arm_pmu goes
ahead and clears the "used" bit and then invokes the "clear_event_idx"
call back, which kind of splits the job between the core code
and the backend. To keep things tidy, mandate the implementation
of clear_event_idx() and add it for exisiting backends.
This will be useful for adding the chained event support, where
we leave the event idx maintenance to the backend.
Also, when an event is removed from the PMU, reset the hw.idx
to indicate that a counter is not allocated for this event,
to help the backends do better checks. This will be also used
for the chain counter support.
Cc: Will Deacon <will.deacon@arm.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Reviewed-by: Julien Thierry <julien.thierry@arm.com>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Convert the {read/write}_counter APIs to handle 64bit values
to enable supporting chained event counters. The backends still
use 32bit values and we pass them 32bit values only. So in effect
there are no functional changes.
Cc: Will Deacon <will.deacon@arm.com>
Acked-by: Mark Rutland <mark.rutland@arm.com>
Reviewed-by: Julien Thierry <julien.thierry@arm.com>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Each PMU defines their max_period of the counter as the maximum
value that can be counted. Since all the PMU backends support
32bit counters by default, let us remove the redundant field.
No functional changes.
Cc: Will Deacon <will.deacon@arm.com>
Acked-by: Mark Rutland <mark.rutland@arm.com>
Reviewed-by: Julien Thierry <julien.thierry@arm.com>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
The arm_pmu::handle_irq() callback has the same prototype as a generic
IRQ handler, taking the IRQ number and a void pointer argument which it
must convert to an arm_pmu pointer.
This means that all arm_pmu::handle_irq() take an IRQ number they never
use, and all must explicitly cast the void pointer to an arm_pmu
pointer.
Instead, let's change arm_pmu::handle_irq to take an arm_pmu pointer,
allowing these casts to be removed. The redundant IRQ number parameter
is also removed.
Suggested-by: Hoeun Ryu <hoeun.ryu@lge.com>
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
The ID_AA64DFR0_EL1.PMUVer field doesn't follow the usual ID registers
scheme. While value 0xf indicates a non-architected PMU is implemented,
values 0x1 to 0xe indicate an increasingly featureful architected PMU,
as if the field were unsigned.
For more details, see ARM DDI 0487C.a, D10.1.4, "Alternative ID scheme
used for the Performance Monitors Extension version".
Currently, we treat the field as signed, and erroneously bail out for
values 0x8 to 0xe. Let's correct that.
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Reviewed-by: Robin Murphy <robin.murphy@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
with bitmap_{from,to}_arr32 over the kernel. Additionally to it:
* __check_eq_bitmap() now takes single nbits argument.
* __check_eq_u32_array is not used in new test but may be used in
future. So I don't remove it here, but annotate as __used.
Tested on arm64 and 32-bit BE mips.
[arnd@arndb.de: perf: arm_dsu_pmu: convert to bitmap_from_arr32]
Link: http://lkml.kernel.org/r/20180201172508.5739-2-ynorov@caviumnetworks.com
[ynorov@caviumnetworks.com: fix net/core/ethtool.c]
Link: http://lkml.kernel.org/r/20180205071747.4ekxtsbgxkj5b2fz@yury-thinkpad
Link: http://lkml.kernel.org/r/20171228150019.27953-2-ynorov@caviumnetworks.com
Signed-off-by: Yury Norov <ynorov@caviumnetworks.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Cc: Ben Hutchings <ben@decadent.org.uk>
Cc: David Decotigny <decot@googlers.com>,
Cc: David S. Miller <davem@davemloft.net>,
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Matthew Wilcox <mawilcox@microsoft.com>
Cc: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Cc: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
bus access read/write events are not supported in A73, based on the
Cortex-A73 TRM r0p2, section 11.9 Events (pages 11-457 to 11-460).
Fixes: 5561b6c5e9 "arm64: perf: add support for Cortex-A73"
Acked-by: Julien Thierry <julien.thierry@arm.com>
Signed-off-by: Xu YiPing <xuyiping@hisilicon.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
The Cortex-A35 uses some implementation defined perf events.
The Cortex-A35 derives from the Cortex-A53 core, using the same event mapings
based on Cortex-A35 TRM r0p2, section C2.3 - Performance monitoring events
(pages C2-562 to C2-565).
Signed-off-by: Julien Thierry <julien.thierry@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
The Cortex-A73 uses some implementation defined perf events.
This patch sets up the necessary mapping for Cortex-A73.
Mappings are based on Cortex-A73 TRM r0p2, section 11.9 Events
(pages 11-457 to 11-460).
Signed-off-by: Julien Thierry <julien.thierry@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Now that the event mapping code always looks into the PMUv3 events
before any extended mappings, the extended mappings can be reduced to
only those events that are not discoverable through the PMCEID registers.
Signed-off-by: Will Deacon <will.deacon@arm.com>
Last level caches and node events were almost never connected in current
supported cores.
We connect last level caches to the actual last level within the core and
node events are connected to bus accesses.
Signed-off-by: Julien Thierry <julien.thierry@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Rather than continue adding CPU-specific event maps, instead look up by
default in the PMUv3 event map and only fallback to the CPU-specific maps
if either the event isn't described by PMUv3, or it is described but
the PMCEID registers say that it is unsupported by the current CPU.
Signed-off-by: Will Deacon <will.deacon@arm.com>
Currently:
$ perf stat -e cycles:u -e cycles:k true
Performance counter stats for 'true':
2,24,699 cycles:u
<not counted> cycles:k (0.00%)
0.000788087 seconds time elapsed
We can not count more than one cycle counter in one instance,because we
allow to map cycle counter into PMCCNTR_EL0 only. However, if I did not
miss anything then specification do not prohibit to use PMEVCNTR<n>_EL0
for cycle count as well.
Modify the code so that it still prefers to use PMCCNTR_EL0 for cycle
counter, however allow to use PMEVCNTR<n>_EL0 if PMCCNTR_EL0 is already
in use.
After this patch:
$ perf stat -e cycles:u -e cycles:k true
Performance counter stats for 'true':
2,17,310 cycles:u
7,40,009 cycles:k
0.000764149 seconds time elapsed
Signed-off-by: Pratyush Anand <panand@redhat.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Perf has supported ARMv8.1 feature with 16-bit evtCount filed [see c210ae8
arm64: perf: Extend event mask for ARMv8.1], event config should be
extended to 16-bit too, otherwise, if use -e event_name whose event_code
is more than 0x3ff, pmu_config_term will return -EINVAL because function
pmu_format_max_value depends on event config.
This patch extends event config to 16-bit.
Signed-off-by: Shaokun Zhang <zhangshaokun@hisilicon.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
commit d98ecdaca2 ("arm64: perf: Count EL2 events if the kernel is
running in HYP") returns -EINVAL when perf system call perf_event_open is
called with exclude_hv != exclude_kernel. This change breaks applications
on VHE enabled ARMv8.1 platforms. The issue was observed with HHVM
application, which calls perf_event_open with exclude_hv = 1 and
exclude_kernel = 0.
There is no separate hypervisor privilege level when VHE is enabled, the
host kernel runs at EL2. So when VHE is enabled, we should ignore
exclude_hv from the application. This behaviour is consistent with PowerPC
where the exclude_hv is ignored when the hypervisor is not present and with
x86 where this flag is ignored.
Signed-off-by: Ganapatrao Kulkarni <ganapatrao.kulkarni@cavium.com>
[will: added comment to justify the behaviour of exclude_hv]
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Add missing L2 cache events: read/write accesses and misses, as well as
the DTLB refills.
Acked-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Commit f1b36dcb5c ("arm64: pmuv3: handle !PMUv3 when probing") is
a little too restrictive, and prevents the use of of backwards
compatible PMUv3 extenstions, which have a PMUver value other than 1.
For instance, ARMv8.1 PMU extensions (as implemented by ThunderX2) are
reported with PMUver value 4.
Per the usual ID register principles, at least 0x1-0x7 imply a
PMUv3-compatible PMU. It's not currently clear whether 0x8-0xe imply the
same.
For the time being, treat the value as signed, and with 0x1-0x7 treated
as meaning PMUv3 is implemented. This may be relaxed by future patches.
Reported-by: Jayachandran C <jnair@caviumnetworks.com>
Tested-by: Jayachandran C <jnair@caviumnetworks.com>
Acked-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Now that we have a framework to handle the ACPI bits, make the PMUv3
code use this. The framework is a little different to what was
originally envisaged, and we can drop some unused support code in the
process of moving over to it.
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Tested-by: Jeremy Linton <jeremy.linton@arm.com>
[will: make armv8_pmu_driver_init static]
Signed-off-by: Will Deacon <will.deacon@arm.com>
When probing via ACPI, we won't know up-front whether a CPU has a PMUv3
compatible PMU. Thus we need to consult ID registers during probe time.
This patch updates our PMUv3 probing code to test for the presence of
PMUv3 functionality before touching an PMUv3-specific registers, and
before updating the struct arm_pmu with PMUv3 data.
When a PMUv3-compatible PMU is not present, probing will return -ENODEV.
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
KVM calls kvm_pmu_set_counter_event_type() when PMCCFILTR is configured.
But this function can't deals with PMCCFILTR correctly because the evtCount
bits of PMCCFILTR, which is reserved 0, conflits with the SW_INCR event
type of other PMXEVTYPER<n> registers. To fix it, when eventsel == 0, this
function shouldn't return immediately; instead it needs to check further
if select_idx is ARMV8_PMU_CYCLE_IDX.
Another issue is that KVM shouldn't copy the eventsel bits of PMCCFILTER
blindly to attr.config. Instead it ought to convert the request to the
"cpu cycle" event type (i.e. 0x11).
To support this patch and to prevent duplicated definitions, a limited
set of ARMv8 perf event types were relocated from perf_event.c to
asm/perf_event.h.
Cc: stable@vger.kernel.org # 4.6+
Acked-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Wei Huang <wei@redhat.com>
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Move the PMU name into a common header file so it may
be referenced by other users.
Signed-off-by: Jeremy Linton <jeremy.linton@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
ARMv8 machines can identify the micro/arch defined counters
that are available on a machine. Add all these counters to the
default armv8 perf map. At run-time disable the counters which
are not available on the given PMU.
Signed-off-by: Jeremy Linton <jeremy.linton@arm.com>
Acked-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
In preparation for ACPI support, add a pmu_probe_info table to
the arm_pmu_device_probe() call. This table gets used when
probing in the absence of a devicetree node for PMU.
Signed-off-by: Mark Salter <msalter@redhat.com>
Signed-off-by: Jeremy Linton <jeremy.linton@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
By using a common attr_groups array, the common arm_pmu code can set up
common files (e.g. cpumask) for us in subsequent patches.
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
