cci_pmu_sync_counters and pmu_event_set_period are internal functions
to the CCI PMU driver, so make them static to avoid polluting the kernel
namespace.
Signed-off-by: Will Deacon <will.deacon@arm.com>
Add ARM CoreLink CCI-550 cache coherent interconnect PMU
driver support. The CCI-550 PMU shares all the attributes of CCI-500
PMU, except for an additional master interface (MI-6 - 0xe).
CCI-550 requires the same work around as for CCI-500 to
write to the PMU counter.
Acked-by: Olof Johansson <olof@lixom.net>
Acked-by: Punit Agrawal <punit.agrawal@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>
CCI-550 PMU shares most of the CCI-500 PMU attributes including the
event format, PMU event codes. The only difference is an additional
master interface (MI6 - 0xe). Hence we share the driver code for both,
except for a model specific event validate method.
This patch renames the common CCI500 symbols to CCI5xx, including the
Kconfig symbol.
No functional changes to the PMU driver.
Acked-by: Olof Johansson <olof@lixom.net>
Acked-by: Punit Agrawal <punit.agrawal@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 CCI PMU driver sets the event counter to the half of the maximum
value(2^31) it can count before we start the counters via
pmu_event_set_period(). This is done to give us the best chance to
handle the overflow interrupt, taking care of extreme interrupt latencies.
However, CCI-500 comes with advanced power saving schemes, which
disables the clock to the event counters unless the counters are enabled to
count (PMCR.CEN). This prevents the driver from writing the period to the
counters before starting them. Also, there is no way we can reset the
individual event counter to 0 (PMCR.RST resets all the counters, losing
their current readings). However the value of the counter is preserved and
could be read back, when the counters are not enabled.
So we cannot reliably use the counters and compute the number of events
generated during the sampling period since we don't have the value of the
counter at start.
This patch works around this issue by changing writes to the counter
with the following steps.
1) Disable all the counters (remembering any counters which were enabled)
2) Enable the PMU, now that all the counters are disabled.
For each counter to be programmed, repeat steps 3-7
3) Save the current event and program the target counter to count an
invalid event, which by spec is guaranteed to not-generate any events.
4) Enable the target counter.
5) Write to the target counter.
6) Disable the target counter
7) Restore the event back on the target counter.
8) Disable the PMU
9) Restore the status of the all the counters
Cc: Punit Agrawal <punit.agrawal@arm.com>
Acked-by: Olof Johansson <olof@lixom.net>
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>
Add a hook for writing to CCI PMU counters. This callback
can be used for CCI models which requires some extra work
to program the PMU counter values. To accommodate group writes
and single counter writes, the call back accepts a bitmask
of the counter indices which need to be programmed with the
given value.
Cc: Punit Agrawal <punit.agrawal@arm.com>
Acked-by: Olof Johansson <olof@lixom.net>
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>
On CCI-500 writing to a counter requires turning the PMU on. So,
synchronising the counter state should not be performed for such special cases,
while turning the PMU on. This patch adds a helper, __cci_pmu_enable_nosync(),
without flushing the counter states.
Cc: Punit Agrawal <punit.agrawal@arm.com>
Acked-by: Olof Johansson <olof@lixom.net>
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>
Adds helper routines to disable the counter controls for
all the counters on the CCI PMU and restore it back, by
preserving the original state in caller provided mask.
Cc: Punit Agrawal <punit.agrawal@arm.com>
Acked-by: Olof Johansson <olof@lixom.net>
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>
pmu_write_counter() is now only called from pmu_write_counters(),
which does so for each set index in the given mask, bounded by
cci_pmu->num_cntrs. So, there is no need for an extra check to
make sure the given counter is valid inside pmu_write_counter.
This patch gets rid of that.
Cc: Punit Agrawal <punit.agrawal@arm.com>
Acked-by: Olof Johansson <olof@lixom.net>
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>
CCI PMU driver always reprograms the counters to a safe value (half of the
counter max, = 2^31) before starting the profiling to account for extreme
interrupt latencies. Also, the cost of writing to a PMU counter could be
very costly on some PMUs(e.g, CCI-500). In order to ammortise the cost of
programming the counters, this patch delays the counter writes to pmu::pmu_enable().
We use the PER_HES_ARCH flag to keep track of the counters which need to
be programmed. Before turning on the PMU, we go through the counters that
were marked for write, and perform the operation in a batch.
To unify all the counter writes to pmu_enable(), this patch also makes sure that
we disable-and-enable the PMU in the irq handler to program any counters that
overflowed.
Cc: Punit Agrawal <punit.agrawal@arm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Acked-by: Olof Johansson <olof@lixom.net>
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>
This patch refactors the CCI PMU driver code a little bit to
make it easier share the code for enabling/disabling the CCI
PMU. This will be used by the hooks to work around the special cases
where writing to a counter is not always that easy(e.g, CCI-500)
No functional changes.
Cc: Punit Agrawal <punit.agrawal@arm.com>
Acked-by: Olof Johansson <olof@lixom.net>
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>
There's no need to dynamically initialise attribute pointers when we can
get the compiler to do it for us. We also don't need a dev_ext_attribute
for the cpumask, as the drvdata for a PMU device is a pointer to struct
pmu.
Cc: Punit Agrawal <punit.agrawal@arm.com>
Acked-by: Olof Johansson <olof@lixom.net>
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Tested-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Reviewed-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Each CCI model have different event/source codes and formats. This
patch exports this information via the sysfs, which includes the
aliases for the events. The aliases are listed by 'perf list', helping
the users to specify the name of the event instead of the binary
config values.
Each event alias must accompany the 'source' code except for the
following cases :
1) CCI-400 - cycles event, doesn't relate to an interface.
2) CCI-500 - Global events to the CCI. (Fixed source code = 0xf)
Each CCI model provides two sets of attributes(format and event),
which are dynamically populated before registering the PMU, to
allow for the appropriate information.
Cc: Punit Agrawal <punit.agrawal@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Pawel Moll <pawel.moll@arm.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Suzuki K. Poulose <suzuki.poulose@arm.com>
Acked-by: Punit Agrawal <punit.agrawal@arm.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
CCI-500 provides 8 event counters which can count any of the
supported events independently. The PMU event id is a 9-bit
value made of two parts.
bits [8:5] - Source port
0x0-0x6 Slave Ports
0x8-0xD Master Ports
0xf Global Events to CCI
0x7,0xe Reserved
bits [0:4] - Event code (specific to each type of port)
The generic CCI-500 controlling interface remains the same with CCI-400.
However there are some differences in the PMU event counters.
- No cycle counter
- Upto 8 counters(4 in CCI-400)
- Each counter area is 64K(4K in CCI400)
- The counter0 starts at offset 0x10000 from the base of CCI
Cc: Punit Agrawal <punit.agrawal@arm.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: devicetree@vger.kernel.org
Signed-off-by: Suzuki K. Poulose <suzuki.poulose@arm.com>
Acked-by: Punit Agrawal <punit.agrawal@arm.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Given that each CCI has different set of interfaces and
its associated events, it is good to abstract the validation of the
event codes to make it easier to add support for a new CCI model.
This patch also abstracts the mapping of a given event to a counter,
as there are some special counters for certain specific events.
We assume that the fixed hardware counters are always at the beginning,
so that we can use cci_model->fixed_hw_events as an upper bound to given
idx to check if we need to program the counter for an event.
Cc: Punit Agrawal <punit.agrawal@arm.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Suzuki K. Poulose <suzuki.poulose@arm.com>
Acked-by: Punit Agrawal <punit.agrawal@arm.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Adds the PMU model specific counters to the PMU model
abstraction to make it easier to add a new PMU.
The patch cleans up the naming convention used all over
the code.
e.g, CCI_PMU_MAX_HW_EVENTS => maximum number of events that
can be counted at any time, which is in fact the maximum
number of counters available.
Change all such namings to use 'counters' instead of events.
This patch also abstracts the following:
1) Size of a PMU event counter area.
2) Maximum number of programmable counters supported by the PMU model
3) Number of counters which counts fixed events (e.g, cycle
counter on CCI-400).
Also changes some of the static allocation of the data
structures to dynamic, to accommodate the number of events
supported by a PMU.
Gets rid ofthe CCI_PMU_* defines for the model. All such
data should be accessed via the model abstraction.
Limits the number of counters to the maximum supported
by the 'model'.
Cc: Punit Agrawal <punit.agrawal@arm.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Suzuki K. Poulose <suzuki.poulose@arm.com>
Acked-by: Punit Agrawal <punit.agrawal@arm.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Currently in validate_group(), there is a static initializer
for fake_pmu.used_mask which is based on CPU_BITS_NONE but
the used_mask array size is based on CCI_PMU_MAX_HW_EVENTS.
CCI_PMU_MAX_HW_EVENTS is not based on NR_CPUS, so CPU_BITS_NONE
is not correct and will cause a build failure if NR_CPUS
is set high enough to make CPU_BITS_NONE larger than used_mask.
Reviewed-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Mark Salter <msalter@redhat.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
We mask the event with the CCI_PMU_EVENT_MASK, before passing
the config to pmu_validate_hw_event(), which causes extra bits
to be ignored and qualifies an invalid event code as valid.
e.g,
$ perf stat -a -C 0 -e CCI_400/config=0x1ff,name=cycles/ sleep 1
Performance counter stats for 'system wide':
506951142 cycles
1.013879626 seconds time elapsed
where, cycles has an event coding of 0xff. This patch also removes
the unnecessary 'event' mask in pmu_write_register, since the config_base
is set by the pmu code after the event is validated.
Acked-by: Punit Agrawal <punit.agrawal@arm.com>
Reviewed-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Suzuki K. Poulose <suzuki.poulose@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Will Deacon
Suzuki K Poulose
Andrzej Hajda
Mark Rutland
Mark Salter