To enable sharing of the arm_pmu code with arm64, this patch factors it
out to drivers/perf/. A new drivers/perf directory is added for
performance monitor drivers to live under.
MAINTAINERS is updated accordingly. Files added previously without a
corresponsing MAINTAINERS update (perf_regs.c, perf_callchain.c, and
perf_event.h) are also added.
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Linus Walleij <linus.walleij@linaro.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Russell King <linux@arm.linux.org.uk>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
[will: augmented Kconfig help slightly]
Signed-off-by: Will Deacon <will.deacon@arm.com>
arch_find_n_match_cpu_physical_id parses the device tree to get the
device node for a given logical cpu index. However, since ARM PMUs get
probed after the CPU device nodes are stashed while registering the
cpus, we can use of_cpu_device_node_get to avoid another DT parse.
This patch replaces arch_find_n_match_cpu_physical_id with
of_cpu_device_node_get to reuse the stashed value directly instead.
Cc: Will Deacon <will.deacon@arm.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
On systems containing multiple, heterogeneous clusters we need a way to
associate a PMU "device" with the CPU(s) on which it exists. For PMUs
that signal overflow with SPIs, this relationship is determined via the
"interrupt-affinity" property, which contains a list of phandles to CPU
nodes for the PMU. For PMUs using PPIs, the per-cpu nature of the
interrupt isn't enough to determine the set of CPUs which actually
contain the device.
This patch allows the interrupt-affinity property to be specified on a
PMU node irrespective of the interrupt type. For PPIs, it identifies
the set of CPUs signalling the PPI in question.
Tested-by: Stephen Boyd <sboyd@codeaurora.org> # Krait PMU
Signed-off-by: Will Deacon <will.deacon@arm.com>
For PPI based PMUs, we bail out early in of_pmu_irq_cfg() without
setting the PMU's supported_cpus bitmap. This causes the
smp_call_function_any() in armv7_probe_num_events() to fail. Set
the bitmap to be all CPUs so that we properly probe PMUs that use
PPIs.
Fixes: cc88116da0 ("arm: perf: treat PMUs as CPU affine")
Cc: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Will Deacon <will.deacon@arm.com>
It's possible, albeit unlikely, that using the of_node here will
reference freed memory. Call of_node_put() after printing the
name to be safe.
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Now that the arm_pmu framework is only used for CPU PMUs, there's no
reason to keep the pseudo-generic and CPU-specific framework portions
separate.
This patch folds the two into perf_event.c.
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
[will: fixed up irq cfg to match upstream]
Signed-off-by: Will Deacon <will.deacon@arm.com>
Currently the arm perf code has platdata callbacks for runtime PM and
irq handling, but no platform implements the hooks for the former. Kill
these off.
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Different CPU microarchitectures implement different PMU events, and
thus events which can be scheduled on one microarchitecture cannot be
scheduled on another, and vice-versa. Some archicted events behave
differently across microarchitectures, and thus cannot be meaningfully
summed. Due to this, we reject the scheduling of an event on a CPU of a
different microarchitecture to that the event targets.
When the core perf code is scheduling events and encounters an event
which cannot be scheduled, it stops attempting to schedule events. As
the perf core periodically rotates the list of events, for some
proportion of the time events which are unschedulable will block events
which are schedulable, resulting in low utilisation of the hardware
counters.
This patch implements a pmu::filter_match callback such that we can
detect and skip such events while scheduling early, before they can
block the schedulable events. This prevents the low HW counter
utilisation issue.
Acked-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
In multi-cluster systems, the PMUs can be different across clusters, and
so our logical PMU may not be able to schedule events on all CPUs.
This patch adds a cpumask to encode which CPUs a PMU driver supports
controlling events for, and limits the driver to scheduling events on
those CPUs, and enabling and disabling the physical PMUs on those CPUs.
The cpumask is built based on the interrupt-affinity property, and in
the absence of such a property a homogenous system is assumed.
Acked-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
The perf core implicitly rejects events spanning multiple HW PMUs, as in
these cases the event->ctx will differ. However this validation is
performed after pmu::event_init() is called in perf_init_event(), and
thus pmu::event_init() may be called with a group leader from a
different HW PMU.
The ARM PMU driver does not take this fact into account, and when
validating groups assumes that it can call to_arm_pmu(event->pmu) for
any HW event. When the event in question is from another HW PMU this is
wrong, and results in dereferencing garbage.
This patch updates the ARM PMU driver to first test for and reject
events from other PMUs, moving the to_arm_pmu and related logic after
this test. Fixes a crash triggered by perf_fuzzer on Linux-4.0-rc2, with
a CCI PMU present:
---
CPU: 0 PID: 1527 Comm: perf_fuzzer Not tainted 4.0.0-rc2 #57
Hardware name: ARM-Versatile Express
task: bd8484c0 ti: be676000 task.ti: be676000
PC is at 0xbf1bbc90
LR is at validate_event+0x34/0x5c
pc : [<bf1bbc90>] lr : [<80016060>] psr: 00000013
...
[<80016060>] (validate_event) from [<80016198>] (validate_group+0x28/0x90)
[<80016198>] (validate_group) from [<80016398>] (armpmu_event_init+0x150/0x218)
[<80016398>] (armpmu_event_init) from [<800882e4>] (perf_try_init_event+0x30/0x48)
[<800882e4>] (perf_try_init_event) from [<8008f544>] (perf_init_event+0x5c/0xf4)
[<8008f544>] (perf_init_event) from [<8008f8a8>] (perf_event_alloc+0x2cc/0x35c)
[<8008f8a8>] (perf_event_alloc) from [<8009015c>] (SyS_perf_event_open+0x498/0xa70)
[<8009015c>] (SyS_perf_event_open) from [<8000e420>] (ret_fast_syscall+0x0/0x34)
Code: bf1be000 bf1bb380 802a2664 00000000 (00000002)
---[ end trace 01aff0ff00926a0a ]---
Also cleans up the code to use the arm_pmu only when we know that
we are dealing with an arm pmu event.
Cc: Will Deacon <will.deacon@arm.com>
Acked-by: Mark Rutland <mark.rutland@arm.com>
Acked-by: Peter Ziljstra (Intel) <peterz@infradead.org>
Signed-off-by: Suzuki K. Poulose <suzuki.poulose@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
If the overflow threshold for a counter is set above or near the
0xffffffff boundary then the kernel may lose track of the overflow
causing only events that occur *after* the overflow to be recorded.
Specifically the problem occurs when the value of the performance counter
overtakes its original programmed value due to wrap around.
Typical solutions to this problem are either to avoid programming in
values likely to be overtaken or to treat the overflow bit as the 33rd
bit of the counter.
Its somewhat fiddly to refactor the code to correctly handle the 33rd bit
during irqsave sections (context switches for example) so instead we take
the simpler approach of avoiding values likely to be overtaken.
We set the limit to half of max_period because this matches the limit
imposed in __hw_perf_event_init(). This causes a doubling of the interrupt
rate for large threshold values, however even with a very fast counter
ticking at 4GHz the interrupt rate would only be ~1Hz.
Signed-off-by: Daniel Thompson <daniel.thompson@linaro.org>
Acked-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
After commit b2b49ccbdd (PM: Kconfig: Set PM_RUNTIME if PM_SLEEP is
selected) PM_RUNTIME is always set if PM is set, so #ifdef blocks
depending on CONFIG_PM_RUNTIME may now be changed to depend on
CONFIG_PM.
Replace CONFIG_PM_RUNTIME with CONFIG_PM everywhere in the code under
arch/arm/ (the defconfig files will be modified later).
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Acked-by: Nishanth Menon <nm@ti.com>
Acked-by: Sekhar Nori <nsekhar@ti.com>
Acked-by: Santosh Shilimkar <ssantosh@kernel.org>
Currently the percpu_pmu pointers used as percpu_irq dev_id values are
defined separately from the other per-cpu accounting data, which make
dynamically allocating the data (as will be required for systems with
heterogeneous CPUs) difficult.
This patch moves the percpu_pmu pointers into pmu_hw_events (which is
itself allocated per cpu), which will allow for easier dynamic
allocation. Both percpu and regular irqs are requested using percpu_pmu
pointers as tokens, freeing us from having to know whether an irq is
percpu within the handler, and thus avoiding a radix tree lookup on the
handler path.
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Reviewed-by: Will Deacon <will.deacon@arm.com>
Reviewed-by: Stephen Boyd <sboyd@codeaurora.org>
Tested-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Now that the arm pmu code is limited to CPU PMUs the get_hw_events()
function is superfluous, as we'll always have a set of per-cpu
pmu_hw_events structures.
This patch removes the get_hw_events() function, replacing it with
a percpu hw_events pointer. Uses of get_hw_events are updated to use
this_cpu_ptr.
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Reviewed-by: Will Deacon <will.deacon@arm.com>
Reviewed-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Commit 3fc2c83087 (ARM: perf: remove event limit from pmu_hw_events) got
rid of the upper limit on the number of events an arm_pmu could handle,
but introduced additional complexity and places a burden on each PMU
driver to allocate accounting data somehow. So far this has not
generally been useful as the only users of arm_pmu are the CPU backend
and the CCI driver.
Now that the CCI driver plugs into the perf subsystem directly, we can
remove some of the complexities that get in the way of supporting
heterogeneous CPU PMUs.
This patch restores the original limits on pmu_hw_events fields such
that the pmu_hw_events data can be allocated as a contiguous block. This
will simplify dynamic pmu_hw_events allocation in later patches.
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Reviewed-by: Will Deacon <will.deacon@arm.com>
Reviewed-by: Stephen Boyd <sboyd@codeaurora.org>
Tested-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Will Deacon <will.deacon@arm.com>
For systems with heterogeneous CPUs (e.g. big.LITTLE systems) the PMUs
can be different in each cluster, and not all events can be migrated
between clusters. To allow userspace to deal with this, it must be
possible to address each PMU independently.
This patch changes PMUs to be registered with dynamic (IDR) types,
allowing them to be targeted individually. Each PMU's type can be found
in ${SYSFS_ROOT}/bus/event_source/devices/${PMU_NAME}/type.
From userspace, raw events can be targeted at a specific PMU:
$ perf stat -e ${PMU_NAME}/config=V,config1=V1,.../
Doing this does not break existing tools which use existing perf types:
when perf core can't find a PMU of matching type (in perf_init_event)
it'll iterate over the set of all PMUs. If a compatible PMU exists,
it'll be found eventually. If more than one compatible PMU exists, the
event will be handled by whichever PMU happens to be earlier in the pmus
list (which currently will be the last compatible PMU registered).
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Reviewed-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
The ARM callchain handling code is currently bundled with the ARM PMU
management code, despite the two having no dependency on each other.
This bundling has the unfortunate property of making callchain handling
depend on CONFIG_HW_PERF_EVENTS, even though the callchain handling
could be applied to software events in the absence of PMU hardware
support.
This patch separates the two, placing the callchain handling in
perf_callchain.c and making it depend on CONFIG_PERF_EVENTS rather than
CONFIG_HW_PERF_EVENTS, enabling callchain recording on kernels built
without hardware perf event support.
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Reviewed-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Make the perf backend use the API so that it correctly references the FP
when in THUMB2 mode
Signed-off-by: Nikolay Borisov <Nikolay.Borisov@arm.com>
Acked-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Under perf, the fp unwinding scheme requires access to user space memory
and can provoke a pagefault via call to __copy_from_user_inatomic from
user_backtrace. This unwinding can take place in response to an interrupt
(__perf_event_overflow). This is undesirable as we may already have
mmap_sem held for write. One example being a process that calls mprotect
just as a the PMU counters overflow.
An example that can provoke this behaviour:
perf record -e event:tocapture --call-graph fp ./application_to_test
This patch addresses this issue by disabling pagefaults briefly in
user_backtrace (as is done in the other architectures: ARM64, x86, Sparc etc.).
Without the patch a deadlock occurs when __perf_event_overflow is called
while reading the data from the user space:
[ INFO: possible recursive locking detected ]
3.16.0-rc2-00038-g0ed7ff6 #46 Not tainted
---------------------------------------------
stress/1634 is trying to acquire lock:
(&mm->mmap_sem){++++++}, at: [<c001dc04>] do_page_fault+0xa8/0x428
but task is already holding lock:
(&mm->mmap_sem){++++++}, at: [<c00f4098>] SyS_mprotect+0xa8/0x1c8
other info that might help us debug this:
Possible unsafe locking scenario:
CPU0
----
lock(&mm->mmap_sem);
lock(&mm->mmap_sem);
*** DEADLOCK ***
May be due to missing lock nesting notation
2 locks held by stress/1634:
#0: (&mm->mmap_sem){++++++}, at: [<c00f4098>] SyS_mprotect+0xa8/0x1c8
#1: (rcu_read_lock){......}, at: [<c00c29dc>] __perf_event_overflow+0x120/0x294
stack backtrace:
CPU: 1 PID: 1634 Comm: stress Not tainted 3.16.0-rc2-00038-g0ed7ff6 #46
[<c0017c8c>] (unwind_backtrace) from [<c0012eec>] (show_stack+0x20/0x24)
[<c0012eec>] (show_stack) from [<c04de914>] (dump_stack+0x7c/0x98)
[<c04de914>] (dump_stack) from [<c006a360>] (__lock_acquire+0x1484/0x1cf0)
[<c006a360>] (__lock_acquire) from [<c006b14c>] (lock_acquire+0xa4/0x11c)
[<c006b14c>] (lock_acquire) from [<c04e3880>] (down_read+0x40/0x7c)
[<c04e3880>] (down_read) from [<c001dc04>] (do_page_fault+0xa8/0x428)
[<c001dc04>] (do_page_fault) from [<c00084ec>] (do_DataAbort+0x44/0xa8)
[<c00084ec>] (do_DataAbort) from [<c0013a1c>] (__dabt_svc+0x3c/0x60)
Exception stack(0xed7c5ae0 to 0xed7c5b28)
5ae0: ed7c5b5c b6dadff4 ffffffec 00000000 b6dadff4 ebc08000 00000000 ebc08000
5b00: 0000007e 00000000 ed7c4000 ed7c5b94 00000014 ed7c5b2c c001a438 c0236c60
5b20: 00000013 ffffffff
[<c0013a1c>] (__dabt_svc) from [<c0236c60>] (__copy_from_user+0xa4/0x3a4)
Acked-by: Steve Capper <steve.capper@linaro.org>
Signed-off-by: Jean Pihet <jean.pihet@linaro.org>
Signed-off-by: Will Deacon <will.deacon@arm.com>
An event may occur when an mm is already released.
As per commit 20afc60f89
'x86, perf: Check that current->mm is alive before getting user callchain'
Acked-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Jean Pihet <jean.pihet@linaro.org>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Make the ARM perf code use the new common PMU interrupt disabled code.
This allows perf to work on ARM machines without a working PMU
interrupt (for example, raspberry pi).
Acked-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Vince Weaver <vincent.weaver@maine.edu>
[peterz: applied changes suggested by Will]
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Grant Likely <grant.likely@linaro.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Rob Herring <robh+dt@kernel.org>
Cc: Russell King <linux@arm.linux.org.uk>
Cc: Will Deacon <will.deacon@arm.com>
Cc: devicetree@vger.kernel.org
Cc: linux-arm-kernel@lists.infradead.org
Cc: linux-kernel@vger.kernel.org
Link: http://lkml.kernel.org/r/alpine.DEB.2.10.1405161712190.11099@vincent-weaver-1.umelst.maine.edu
[ Small readability tweaks to the code. ]
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Since we indirect all of our PMU IRQ handling through a dispatcher, it's
trivial to hook up perf_sample_event_took to prevent applications such
as oprofile from generating interrupt storms due to an unrealisticly
low sample period.
Reported-by: Robert Richter <rric@kernel.org>
Signed-off-by: Will Deacon <will.deacon@arm.com>
On Krait processors we have a many-to-one relationship between
raw CPU events and the event programmed into the PMNx counter.
Two raw CPU events could map to the same value programmed in the
PMNx counter. To avoid this problem, we check for collisions
during the get_event_idx() callback by setting a bit in a bitmap
whenever a certain event is used in a PMNx counter (see the next
patch). Unfortunately, we don't have a hook to clear this bit in
the bitmap when the event is deleted so let's add an optional
clear_event_idx() callback for this purpose.
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Some CPU PMUs are wired up with one PPI for all the CPUs instead
of with a different SPI for each CPU. Add support for these
devices.
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Will Deacon <will.deacon@arm.com>
This reverts commit 3581fe0ef3.
Fixes to the handling of PERF_EVENT_IOC_PERIOD in the core code mean
we no longer have to play this horrible game.
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1385560479-11014-2-git-send-email-will.deacon@arm.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Since software events can always be scheduled, perf allows software and
hardware events to be mixed together in the same event group. There are
two ways in which this can come about:
(1) A SW event is added to a HW group. This validates using the HW PMU
of the group leader.
(2) A HW event is added to a SW group. This inserts the SW events and
the new HW event into a HW context, but the SW event remains the
group leader.
When validating the latter case, we would ideally compare the PMU of
each event in the group with the relevant HW PMU. The problem is, in the
face of potentially multiple HW PMUs, we don't have a handle on the
relevant structure. Commit 7b9f72c62e ("ARM: perf: clean up event
group validation") attempting to resolve this issue, but actually made
things *worse* by comparing with the leader PMU. If the leader is a SW
event, then we automatically `pass' all the HW events during validation!
This patch removes the check against the leader PMU. Whilst this will
allow events from multiple HW PMUs to be grouped together, that should
probably be dealt with in perf core as the result of a later patch.
Acked-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
It is possible to construct an event group with a software event as a
group leader and then subsequently add a hardware event to the group.
This results in the event group being validated by adding all members
of the group to a fake PMU and attempting to allocate each event on
their respective PMU.
Unfortunately, for software events wthout a corresponding arm_pmu, this
results in a kernel crash attempting to dereference the ->get_event_idx
function pointer.
This patch fixes the problem by checking explicitly for software events
and ignoring those in event validation (since they can always be
scheduled). We will probably want to revisit this for 3.12, since the
validation checks don't appear to work correctly when dealing with
multiple hardware PMUs anyway.
Cc: <stable@vger.kernel.org>
Reported-by: Vince Weaver <vincent.weaver@maine.edu>
Tested-by: Vince Weaver <vincent.weaver@maine.edu>
Tested-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
With this change, we no longer lose the innermost entry in the user-mode
part of the call chain. See also the x86 port, which includes the ip.
It's possible to partially work around this problem by post-processing
the data to use the PERF_SAMPLE_IP value, but this works only if the CPU
wasn't in the kernel when the sample was taken.
Cc: <stable@vger.kernel.org>
Signed-off-by: Jed Davis <jld@mozilla.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Events may be created with attr->disabled == 1 and attr->enable_on_exec
== 1, which confuses the group validation code because events with the
PERF_EVENT_STATE_OFF are not considered candidates for scheduling, which
may lead to failure at group scheduling time.
This patch fixes the validation check for ARM, so that events in the
OFF state are still considered when enable_on_exec is true.
Cc: stable@vger.kernel.org
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
Cc: Jiri Olsa <jolsa@redhat.com>
Reported-by: Sudeep KarkadaNagesha <Sudeep.KarkadaNagesha@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
WARNING: vmlinux.o(.text+0xfb80): Section mismatch in reference
from the function armpmu_register() to the function
.init.text:armpmu_init()
The function armpmu_register() references
the function __init armpmu_init().
This is often because armpmu_register lacks a __init
annotation or the annotation of armpmu_init is wrong.
Just drop the __init marking on armpmu_init() because
armpmu_register() no longer has an __init marking.
Acked-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Commit 9dcbf46655 ("ARM: perf: simplify __hw_perf_event_init err
handling") tidied up the error handling code for perf event
initialisation on ARM, but a copy-and-paste error left a dangling
semicolon at the end of an if statement.
This patch removes the broken semicolon, restoring the old group
validation semantics.
Cc: Mark Rutland <mark.rutland@arm.com>
Acked-by: Dirk Behme <dirk.behme@gmail.com>
Signed-off-by: Chen Gang <gang.chen@asianux.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Currently __hw_perf_event_init has an err variable that's ignored right
until the end, where it's initialised, conditionally set, and then used
as a boolean flag deciding whether to return another error code.
This patch removes the err variable and simplifies the associated error
handling logic.
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
We currently check for hwx->idx < 0 in armpmu_read and armpmu_del
unnecessarily. The only case where hwc->idx < 0 is when armpmu_add
fails, in which case the event's state is set to
PERF_EVENT_STATE_INACTIVE.
The perf core will not attempt to read from an event in
PERF_EVENT_STATE_INACTIVE, and so the check in armpmu_read is
unnecessary. Similarly, if perf core cannot add an event it will not
attempt to delete it, so the WARN_ON in armpmu_del is unnecessary.
This patch removes these two redundant checks.
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Commit 7be2958 (ARM: PMU: Add runtime PM Support) updated the ARM PMU code to
use runtime PM which was prototyped and validated on the OMAP devices. In this
commit, there is no call pm_runtime_enable() and for OMAP devices
pm_runtime_enable() is currently being called from the OMAP PMU code when the
PMU device is created. However, there are two problems with this:
1. For any other ARM device wishing to use runtime PM for PMU they will need
to call pm_runtime_enable() for runtime PM to work.
2. When booting with device-tree and using device-tree to create the PMU
device, pm_runtime_enable() needs to be called from within the ARM PERF
driver as we are no longer calling any device specific code to create the
device. Hence, PMU does not work on OMAP devices that use the runtime PM
callbacks when using device-tree to create the PMU device.
Therefore, call pm_runtime_enable() directly from the ARM PMU driver when
registering the device. For platforms that do not use runtime PM,
pm_runtime_enable() does nothing and for platforms that do use runtime PM but
may not require it specifically for PMU, this will just add a little overhead
when initialising and uninitialising the PMU device.
Tested with PERF on OMAP2420, OMAP3430 and OMAP4460.
Acked-by: Kevin Hilman <khilman@ti.com>
Acked-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: Jon Hunter <jon-hunter@ti.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Perf has three ways to name a PMU: either by passing an explicit char *,
reading arm_pmu->name or accessing arm_pmu->pmu.name.
Just use arm_pmu->name consistently in the ARM backend.
Signed-off-by: Will Deacon <will.deacon@arm.com>
The arm_pmu functions have wildly varied parameters which can often be
derived from struct perf_event.
This patch changes the arm_pmu function prototypes so that struct
perf_event pointers are passed in preference to fields that can be
derived from the event.
Signed-off-by: Sudeep KarkadaNagesha <Sudeep.KarkadaNagesha@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Add minimal guest support to perf, so it can distinguish whether
the PMU interrupt was in the host or the guest, as well as collecting
some very basic information (guest PC, user vs kernel mode).
This is not feature complete though, as it doesn't support backtracing
in the guest.
Based on the x86 implementation, tested with KVM/ARM.
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
The PERF_EVENT_IOC_PERIOD ioctl command can be used to change the
sample period of a running perf_event. Consequently, when calculating
the next event period, the new period will only be considered after the
previous one has overflowed.
This patch changes the calculation of the remaining event ticks so that
they are offset if the period has changed.
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Reported-by: Andreas Sandberg <andreas.sandberg@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
This patch moves the CPU-specific IRQ registration and parsing code into
the CPU PMU backend. This is required because a PMU may have more than
one interrupt, which in turn can be either PPI (per-cpu) or SPI
(requiring strict affinity setting at the interrupt distributor).
Signed-off-by: Sudeep KarkadaNagesha <Sudeep.KarkadaNagesha@arm.com>
[will: cosmetic edits and reworked interrupt dispatching]
Signed-off-by: Will Deacon <will.deacon@arm.com>
The CPU PMU code is tightly coupled with generic ARM PMU handling code.
This makes it cumbersome when trying to add support for other ARM PMUs
(e.g. interconnect, L2 cache controller, bus) as the generic parts of
the code are not readily reusable.
This patch cleans up perf_event.c so that reusable code is exposed via
header files to other potential PMU drivers. The CPU code is
consistently named to identify it as such and also to prepare for moving
it into a separate file.
Signed-off-by: Will Deacon <will.deacon@arm.com>
The CPU PMU is probed using the current cpuid information as part of the
early_initcall initialising the architecture perf backend. For
architectures without NMI (such as ARM), this does not need to be
performed early and can be deferred to the driver probe callback. This
also allows us to probe the devicetree in preference to parsing the
current cpuid, which may be invalid on a big.LITTLE multi-cluster
system.
This patch defers the PMU probing and uses the devicetree information
when available.
Signed-off-by: Will Deacon <will.deacon@arm.com>
There's a rather strange compiler barrier in the PMU disabling code
which was presumably placed there by aliens. There's no valid reason for
the barrier and one can only suspect that it's up to no good.
This patch removes it before it has a chance to spread.
Signed-off-by: Will Deacon <will.deacon@arm.com>
The PMU reservation mechanism was originally intended to allow OProfile
and perf-events to co-ordinate over access to the CPU PMU. Since then,
OProfile for ARM has moved to using perf as its backend, so the
reservation code is no longer used.
This patch removes the reservation code for the CPU PMU on ARM.
Signed-off-by: Will Deacon <will.deacon@arm.com>
This patch adds separate devicetree bindings for 11MPcore and
Cortex-{A5,A7,A15} PMUs in preparation for improved devicetree parsing
in the ARM perf-event CPU PMU driver.
Cc: Grant Likely <grant.likely@secretlab.ca>
Acked-by: Rob Herring <rob.herring@calxeda.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Add runtime PM support to the ARM PMU driver so that devices such as OMAP
supporting dynamic PM can use the platform->runtime_* hooks to initialise
hardware at runtime. Without having these runtime PM hooks in place any
configuration of the PMU hardware would be lost when low power states are
entered and hence would prevent PMU from working.
This change also replaces the PMU platform functions enable_irq and disable_irq
added by Ming Lei with runtime_resume and runtime_suspend funtions. Ming had
added the enable_irq and disable_irq functions as a method to configure the
cross trigger interface on OMAP4 for routing the PMU interrupts. By adding
runtime PM support, we can move the code called by enable_irq and disable_irq
into the runtime PM callbacks runtime_resume and runtime_suspend.
Cc: Ming Lei <ming.lei@canonical.com>
Cc: Benoit Cousson <b-cousson@ti.com>
Cc: Paul Walmsley <paul@pwsan.com>
Cc: Kevin Hilman <khilman@ti.com>
Signed-off-by: Jon Hunter <jon-hunter@ti.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
In order to provide PMU name strings compatible with the OProfile
user ABI, an enumeration of all PMUs is currently used by perf to
identify each PMU uniquely. Unfortunately, this does not scale well
in the presence of multiple PMUs and creates a single, global namespace
across all PMUs in the system.
This patch removes the enumeration and instead uses the name string
for the PMU to map onto the OProfile variant. perf_pmu_name is
implemented for CPU PMUs, which is all that OProfile cares about anyway.
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
