* [PATCH v4 1/6] Documentation: perf: hisi: Documentation for HiSilicon SoC PMU driver
2017-07-25 12:10 [PATCH v4 0/6] Add HiSilicon SoC uncore Performance Monitoring Unit driver Shaokun Zhang
@ 2017-07-25 12:10 ` Shaokun Zhang
2017-08-15 9:50 ` Mark Rutland
2017-07-25 12:10 ` [PATCH v4 2/6] perf: hisi: Add support for HiSilicon SoC uncore " Shaokun Zhang
` (5 subsequent siblings)
6 siblings, 1 reply; 19+ messages in thread
From: Shaokun Zhang @ 2017-07-25 12:10 UTC (permalink / raw)
To: mark.rutland, will.deacon
Cc: jonathan.cameron, linux-arm-kernel, linux-kernel, linux-doc, linuxarm
This patch adds documentation for the uncore PMUs on HiSilicon SoC.
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Signed-off-by: Shaokun Zhang <zhangshaokun@hisilicon.com>
Signed-off-by: Anurup M <anurup.m@huawei.com>
---
Documentation/perf/hisi-pmu.txt | 52 +++++++++++++++++++++++++++++++++++++++++
1 file changed, 52 insertions(+)
create mode 100644 Documentation/perf/hisi-pmu.txt
diff --git a/Documentation/perf/hisi-pmu.txt b/Documentation/perf/hisi-pmu.txt
new file mode 100644
index 0000000..f45a03d
--- /dev/null
+++ b/Documentation/perf/hisi-pmu.txt
@@ -0,0 +1,52 @@
+HiSilicon SoC uncore Performance Monitoring Unit (PMU)
+======================================================
+The HiSilicon SoC chip comprehends various independent system device PMUs
+such as L3 cache (L3C), Hydra Home Agent (HHA) and DDRC. These PMUs are
+independent and have hardware logic to gather statistics and performance
+information.
+
+HiSilicon SoC encapsulates multiple CPU and IO dies. Each CPU cluster
+(CCL) is made up of 4 cpu cores sharing one L3 cache; Each CPU die is
+called Super CPU cluster (SCCL) and is made up of 6 CCLs. Each SCCL has
+two HHAs (0 - 1) and four DDRCs (0 - 3), respectively.
+
+HiSilicon SoC uncore PMU driver
+---------------------------------------
+Each device PMU has separate registers for event counting, control and
+interrupt, and the PMU driver shall register perf PMU drivers like L3C,
+HHA and DDRC etc. The available events and configuration options shall
+be described in the sysfs, see /sys/devices/hisi_* or /sys/bus/
+event_source/devices/hisi_*.
+The "perf list" command shall list the available events from sysfs.
+
+Each L3C, HHA and DDRC in one SCCL are registered as an separate PMU with perf.
+The PMU name will appear in event listing as hisi_module <index-id>_<sccl-id>.
+where "index-id" is the index of module and "sccl-id" is the identifier of
+the SCCL.
+e.g. hisi_l3c0_1/rd_hit_cpipe is READ_HIT_CPIPE event of L3C index #0 and SCCL
+ID #1.
+e.g. hisi_hha0_1/rx_operations is RX_OPERATIONS event of HHA index #0 and SCCL
+ID #1.
+
+The driver also provides a "cpumask" sysfs attribute, which shows the CPU core
+ID used to count the uncore PMU event.
+
+Example usage of perf:
+$# perf list
+hisi_l3c0_3/rd_hit_cpipe/ [kernel PMU event]
+------------------------------------------
+hisi_l3c0_3/wr_hit_cpipe/ [kernel PMU event]
+------------------------------------------
+hisi_l3c0_1/rd_hit_cpipe/ [kernel PMU event]
+------------------------------------------
+hisi_l3c0_1/wr_hit_cpipe/ [kernel PMU event]
+------------------------------------------
+
+$# perf stat -a -e hisi_l3c0_1/rd_hit_cpipe/ sleep 5
+$# perf stat -a -e hisi_l3c0_1/config=0x02/ sleep 5
+
+The current driver does not support sampling. So "perf record" is unsupported.
+Also attach to a task is unsupported as the events are all uncore.
+
+Note: Please contact the maintainer for a complete list of events supported for
+the PMU devices in the SoC and its information if needed.
--
1.9.1
^ permalink raw reply related [flat|nested] 19+ messages in thread* Re: [PATCH v4 1/6] Documentation: perf: hisi: Documentation for HiSilicon SoC PMU driver
2017-07-25 12:10 ` [PATCH v4 1/6] Documentation: perf: hisi: Documentation for HiSilicon SoC PMU driver Shaokun Zhang
@ 2017-08-15 9:50 ` Mark Rutland
2017-08-17 2:30 ` Zhangshaokun
0 siblings, 1 reply; 19+ messages in thread
From: Mark Rutland @ 2017-08-15 9:50 UTC (permalink / raw)
To: Shaokun Zhang
Cc: will.deacon, jonathan.cameron, linux-arm-kernel, linux-kernel,
linux-doc, linuxarm
Hi,
On Tue, Jul 25, 2017 at 08:10:37PM +0800, Shaokun Zhang wrote:
> This patch adds documentation for the uncore PMUs on HiSilicon SoC.
>
> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
> Signed-off-by: Shaokun Zhang <zhangshaokun@hisilicon.com>
> Signed-off-by: Anurup M <anurup.m@huawei.com>
> ---
> Documentation/perf/hisi-pmu.txt | 52 +++++++++++++++++++++++++++++++++++++++++
> 1 file changed, 52 insertions(+)
> create mode 100644 Documentation/perf/hisi-pmu.txt
>
> diff --git a/Documentation/perf/hisi-pmu.txt b/Documentation/perf/hisi-pmu.txt
> new file mode 100644
> index 0000000..f45a03d
> --- /dev/null
> +++ b/Documentation/perf/hisi-pmu.txt
> @@ -0,0 +1,52 @@
> +HiSilicon SoC uncore Performance Monitoring Unit (PMU)
> +======================================================
> +The HiSilicon SoC chip comprehends various independent system device PMUs
Nit: s/comprehends/comprises/ would be easier to read.
> +such as L3 cache (L3C), Hydra Home Agent (HHA) and DDRC. These PMUs are
> +independent and have hardware logic to gather statistics and performance
> +information.
> +
> +HiSilicon SoC encapsulates multiple CPU and IO dies. Each CPU cluster
Nit: The Hisilicon SoC
> +(CCL) is made up of 4 cpu cores sharing one L3 cache; Each CPU die is
Nit: s/Each/each/
> +called Super CPU cluster (SCCL) and is made up of 6 CCLs. Each SCCL has
> +two HHAs (0 - 1) and four DDRCs (0 - 3), respectively.
> +
> +HiSilicon SoC uncore PMU driver
> +---------------------------------------
> +Each device PMU has separate registers for event counting, control and
> +interrupt, and the PMU driver shall register perf PMU drivers like L3C,
> +HHA and DDRC etc. The available events and configuration options shall
> +be described in the sysfs, see /sys/devices/hisi_*
What exactly its exposed under /sys/devices/hisi_* ?
> or /sys/bus/
> +event_source/devices/hisi_*.
Please don't wrap paths; keep this on one line.
> +The "perf list" command shall list the available events from sysfs.
> +
> +Each L3C, HHA and DDRC in one SCCL are registered as an separate PMU with perf.
> +The PMU name will appear in event listing as hisi_module <index-id>_<sccl-id>.
> +where "index-id" is the index of module and "sccl-id" is the identifier of
> +the SCCL.
> +e.g. hisi_l3c0_1/rd_hit_cpipe is READ_HIT_CPIPE event of L3C index #0 and SCCL
> +ID #1.
> +e.g. hisi_hha0_1/rx_operations is RX_OPERATIONS event of HHA index #0 and SCCL
> +ID #1.
It would make more sense for this to be hierarichal, e.g. hisi_sccl{X}_l3c{Y}.
Other than the above nits, this documentation is very useful. Thanks for
putting this together.
Thanks,
Mark.
^ permalink raw reply [flat|nested] 19+ messages in thread* Re: [PATCH v4 1/6] Documentation: perf: hisi: Documentation for HiSilicon SoC PMU driver
2017-08-15 9:50 ` Mark Rutland
@ 2017-08-17 2:30 ` Zhangshaokun
2017-08-17 3:56 ` Jonathan Cameron
0 siblings, 1 reply; 19+ messages in thread
From: Zhangshaokun @ 2017-08-17 2:30 UTC (permalink / raw)
To: Mark Rutland
Cc: linux-doc, will.deacon, linuxarm, linux-kernel, linux-arm-kernel
Hi Mark,
Thanks for your comments.
On 2017/8/15 17:50, Mark Rutland wrote:
> Hi,
>
> On Tue, Jul 25, 2017 at 08:10:37PM +0800, Shaokun Zhang wrote:
>> This patch adds documentation for the uncore PMUs on HiSilicon SoC.
>>
>> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
>> Signed-off-by: Shaokun Zhang <zhangshaokun@hisilicon.com>
>> Signed-off-by: Anurup M <anurup.m@huawei.com>
>> ---
>> Documentation/perf/hisi-pmu.txt | 52 +++++++++++++++++++++++++++++++++++++++++
>> 1 file changed, 52 insertions(+)
>> create mode 100644 Documentation/perf/hisi-pmu.txt
>>
>> diff --git a/Documentation/perf/hisi-pmu.txt b/Documentation/perf/hisi-pmu.txt
>> new file mode 100644
>> index 0000000..f45a03d
>> --- /dev/null
>> +++ b/Documentation/perf/hisi-pmu.txt
>> @@ -0,0 +1,52 @@
>> +HiSilicon SoC uncore Performance Monitoring Unit (PMU)
>> +======================================================
>> +The HiSilicon SoC chip comprehends various independent system device PMUs
>
> Nit: s/comprehends/comprises/ would be easier to read.
>
Ok.
>> +such as L3 cache (L3C), Hydra Home Agent (HHA) and DDRC. These PMUs are
>> +independent and have hardware logic to gather statistics and performance
>> +information.
>> +
>> +HiSilicon SoC encapsulates multiple CPU and IO dies. Each CPU cluster
>
> Nit: The Hisilicon SoC
>
Ok.
>> +(CCL) is made up of 4 cpu cores sharing one L3 cache; Each CPU die is
>
> Nit: s/Each/each/
>
Ok.
>> +called Super CPU cluster (SCCL) and is made up of 6 CCLs. Each SCCL has
>> +two HHAs (0 - 1) and four DDRCs (0 - 3), respectively.
>> +
>> +HiSilicon SoC uncore PMU driver
>> +---------------------------------------
>> +Each device PMU has separate registers for event counting, control and
>> +interrupt, and the PMU driver shall register perf PMU drivers like L3C,
>> +HHA and DDRC etc. The available events and configuration options shall
>> +be described in the sysfs, see /sys/devices/hisi_*
>
> What exactly its exposed under /sys/devices/hisi_* ?
>
Apologies that i shall list /sys/devices/hisi_sccl{X}_<l3c{Y}/hha{Y}/ddrc{Y}>/ and
will change it in next version.
>> or /sys/bus/
>> +event_source/devices/hisi_*.
>
> Please don't wrap paths; keep this on one line.
>
Ok.
>> +The "perf list" command shall list the available events from sysfs.
>> +
>> +Each L3C, HHA and DDRC in one SCCL are registered as an separate PMU with perf.
>> +The PMU name will appear in event listing as hisi_module <index-id>_<sccl-id>.
>> +where "index-id" is the index of module and "sccl-id" is the identifier of
>> +the SCCL.
>> +e.g. hisi_l3c0_1/rd_hit_cpipe is READ_HIT_CPIPE event of L3C index #0 and SCCL
>> +ID #1.
>> +e.g. hisi_hha0_1/rx_operations is RX_OPERATIONS event of HHA index #0 and SCCL
>> +ID #1.
>
> It would make more sense for this to be hierarichal, e.g. hisi_sccl{X}_l3c{Y}.
>
Surely, it is nicer.
Thanks.
Shaokun
> Other than the above nits, this documentation is very useful. Thanks for
> putting this together.
>
> Thanks,
> Mark.
> _______________________________________________
> linuxarm mailing list
> linuxarm@huawei.com
> http://rnd-openeuler.huawei.com/mailman/listinfo/linuxarm
>
> .
>
^ permalink raw reply [flat|nested] 19+ messages in thread* Re: [PATCH v4 1/6] Documentation: perf: hisi: Documentation for HiSilicon SoC PMU driver
2017-08-17 2:30 ` Zhangshaokun
@ 2017-08-17 3:56 ` Jonathan Cameron
0 siblings, 0 replies; 19+ messages in thread
From: Jonathan Cameron @ 2017-08-17 3:56 UTC (permalink / raw)
To: Zhangshaokun
Cc: Mark Rutland, linux-kernel, will.deacon, linuxarm,
linux-arm-kernel, linux-doc
On Thu, 17 Aug 2017 10:30:23 +0800
Zhangshaokun <zhangshaokun@hisilicon.com> wrote:
> Hi Mark,
>
> Thanks for your comments.
>
> On 2017/8/15 17:50, Mark Rutland wrote:
> > Hi,
> >
> > On Tue, Jul 25, 2017 at 08:10:37PM +0800, Shaokun Zhang wrote:
> >> This patch adds documentation for the uncore PMUs on HiSilicon SoC.
> >>
> >> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
> >> Signed-off-by: Shaokun Zhang <zhangshaokun@hisilicon.com>
> >> Signed-off-by: Anurup M <anurup.m@huawei.com>
> >> ---
> >> Documentation/perf/hisi-pmu.txt | 52 +++++++++++++++++++++++++++++++++++++++++
> >> 1 file changed, 52 insertions(+)
> >> create mode 100644 Documentation/perf/hisi-pmu.txt
> >>
> >> diff --git a/Documentation/perf/hisi-pmu.txt b/Documentation/perf/hisi-pmu.txt
> >> new file mode 100644
> >> index 0000000..f45a03d
> >> --- /dev/null
> >> +++ b/Documentation/perf/hisi-pmu.txt
> >> @@ -0,0 +1,52 @@
> >> +HiSilicon SoC uncore Performance Monitoring Unit (PMU)
> >> +======================================================
> >> +The HiSilicon SoC chip comprehends various independent system device PMUs
> >
> > Nit: s/comprehends/comprises/ would be easier to read.
> >
>
> Ok.
s/comprises/includes/ would perhaps be even better. There are a few other
things in the SoC beyond independent system device PMUs :)
(good spot though - I completely missed comprehends when doing the internal
review!)
>
> >> +such as L3 cache (L3C), Hydra Home Agent (HHA) and DDRC. These PMUs are
> >> +independent and have hardware logic to gather statistics and performance
> >> +information.
> >> +
> >> +HiSilicon SoC encapsulates multiple CPU and IO dies. Each CPU cluster
> >
> > Nit: The Hisilicon SoC
> >
>
> Ok.
I disagree. It is odd but the company name is HiSilicon with the capital S.
>
> >> +(CCL) is made up of 4 cpu cores sharing one L3 cache; Each CPU die is
> >
> > Nit: s/Each/each/
> >
>
> Ok.
>
> >> +called Super CPU cluster (SCCL) and is made up of 6 CCLs. Each SCCL has
> >> +two HHAs (0 - 1) and four DDRCs (0 - 3), respectively.
> >> +
> >> +HiSilicon SoC uncore PMU driver
> >> +---------------------------------------
> >> +Each device PMU has separate registers for event counting, control and
> >> +interrupt, and the PMU driver shall register perf PMU drivers like L3C,
> >> +HHA and DDRC etc. The available events and configuration options shall
> >> +be described in the sysfs, see /sys/devices/hisi_*
> >
> > What exactly its exposed under /sys/devices/hisi_* ?
> >
>
> Apologies that i shall list /sys/devices/hisi_sccl{X}_<l3c{Y}/hha{Y}/ddrc{Y}>/ and
> will change it in next version.
>
> >> or /sys/bus/
> >> +event_source/devices/hisi_*.
> >
> > Please don't wrap paths; keep this on one line.
> >
>
> Ok.
>
> >> +The "perf list" command shall list the available events from sysfs.
> >> +
> >> +Each L3C, HHA and DDRC in one SCCL are registered as an separate PMU with perf.
> >> +The PMU name will appear in event listing as hisi_module <index-id>_<sccl-id>.
> >> +where "index-id" is the index of module and "sccl-id" is the identifier of
> >> +the SCCL.
> >> +e.g. hisi_l3c0_1/rd_hit_cpipe is READ_HIT_CPIPE event of L3C index #0 and SCCL
> >> +ID #1.
> >> +e.g. hisi_hha0_1/rx_operations is RX_OPERATIONS event of HHA index #0 and SCCL
> >> +ID #1.
> >
> > It would make more sense for this to be hierarichal, e.g. hisi_sccl{X}_l3c{Y}.
> >
>
> Surely, it is nicer.
>
> Thanks.
> Shaokun
>
> > Other than the above nits, this documentation is very useful. Thanks for
> > putting this together.
> >
> > Thanks,
> > Mark.
> > _______________________________________________
> > linuxarm mailing list
> > linuxarm@huawei.com
> > http://rnd-openeuler.huawei.com/mailman/listinfo/linuxarm
> >
> > .
> >
>
> _______________________________________________
> linuxarm mailing list
> linuxarm@huawei.com
> http://rnd-openeuler.huawei.com/mailman/listinfo/linuxarm
^ permalink raw reply [flat|nested] 19+ messages in thread
* [PATCH v4 2/6] perf: hisi: Add support for HiSilicon SoC uncore PMU driver
2017-07-25 12:10 [PATCH v4 0/6] Add HiSilicon SoC uncore Performance Monitoring Unit driver Shaokun Zhang
2017-07-25 12:10 ` [PATCH v4 1/6] Documentation: perf: hisi: Documentation for HiSilicon SoC PMU driver Shaokun Zhang
@ 2017-07-25 12:10 ` Shaokun Zhang
2017-08-15 10:16 ` Mark Rutland
2017-07-25 12:10 ` [PATCH v4 3/6] perf: hisi: Add support for HiSilicon SoC L3C " Shaokun Zhang
` (4 subsequent siblings)
6 siblings, 1 reply; 19+ messages in thread
From: Shaokun Zhang @ 2017-07-25 12:10 UTC (permalink / raw)
To: mark.rutland, will.deacon
Cc: jonathan.cameron, linux-arm-kernel, linux-kernel, linux-doc, linuxarm
This patch adds support HiSilicon SoC uncore PMU driver framework and
interfaces.
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Signed-off-by: Shaokun Zhang <zhangshaokun@hisilicon.com>
Signed-off-by: Anurup M <anurup.m@huawei.com>
---
drivers/perf/Kconfig | 7 +
drivers/perf/Makefile | 1 +
drivers/perf/hisilicon/Makefile | 1 +
drivers/perf/hisilicon/hisi_uncore_pmu.c | 398 +++++++++++++++++++++++++++++++
drivers/perf/hisilicon/hisi_uncore_pmu.h | 103 ++++++++
5 files changed, 510 insertions(+)
create mode 100644 drivers/perf/hisilicon/Makefile
create mode 100644 drivers/perf/hisilicon/hisi_uncore_pmu.c
create mode 100644 drivers/perf/hisilicon/hisi_uncore_pmu.h
diff --git a/drivers/perf/Kconfig b/drivers/perf/Kconfig
index e5197ff..78fc4bc 100644
--- a/drivers/perf/Kconfig
+++ b/drivers/perf/Kconfig
@@ -17,6 +17,13 @@ config ARM_PMU_ACPI
depends on ARM_PMU && ACPI
def_bool y
+config HISI_PMU
+ bool "HiSilicon SoC PMU"
+ depends on ARM64 && ACPI
+ help
+ Support for HiSilicon SoC uncore performance monitoring
+ unit (PMU), such as: L3C, HHA and DDRC.
+
config QCOM_L2_PMU
bool "Qualcomm Technologies L2-cache PMU"
depends on ARCH_QCOM && ARM64 && ACPI
diff --git a/drivers/perf/Makefile b/drivers/perf/Makefile
index 6420bd4..41d3342 100644
--- a/drivers/perf/Makefile
+++ b/drivers/perf/Makefile
@@ -1,5 +1,6 @@
obj-$(CONFIG_ARM_PMU) += arm_pmu.o arm_pmu_platform.o
obj-$(CONFIG_ARM_PMU_ACPI) += arm_pmu_acpi.o
+obj-$(CONFIG_HISI_PMU) += hisilicon/
obj-$(CONFIG_QCOM_L2_PMU) += qcom_l2_pmu.o
obj-$(CONFIG_QCOM_L3_PMU) += qcom_l3_pmu.o
obj-$(CONFIG_XGENE_PMU) += xgene_pmu.o
diff --git a/drivers/perf/hisilicon/Makefile b/drivers/perf/hisilicon/Makefile
new file mode 100644
index 0000000..2783bb3
--- /dev/null
+++ b/drivers/perf/hisilicon/Makefile
@@ -0,0 +1 @@
+obj-$(CONFIG_HISI_PMU) += hisi_uncore_pmu.o
diff --git a/drivers/perf/hisilicon/hisi_uncore_pmu.c b/drivers/perf/hisilicon/hisi_uncore_pmu.c
new file mode 100644
index 0000000..d868447
--- /dev/null
+++ b/drivers/perf/hisilicon/hisi_uncore_pmu.c
@@ -0,0 +1,398 @@
+/*
+ * HiSilicon SoC Hardware event counters support
+ *
+ * Copyright (C) 2017 Hisilicon Limited
+ * Author: Anurup M <anurup.m@huawei.com>
+ * Shaokun Zhang <zhangshaokun@hisilicon.com>
+ *
+ * This code is based on the uncore PMUs like arm-cci and arm-ccn.
+ *
+ * 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.
+ */
+#include <linux/bitmap.h>
+#include <linux/bitops.h>
+#include <linux/bug.h>
+#include <linux/err.h>
+#include <linux/errno.h>
+#include <asm/local64.h>
+#include "hisi_uncore_pmu.h"
+
+#define HISI_GET_EVENTID(ev) (ev->hw.config_base & 0xff)
+#define HISI_MAX_PERIOD(nr) (BIT_ULL(nr) - 1)
+
+/*
+ * PMU format attributes
+ */
+ssize_t hisi_format_sysfs_show(struct device *dev,
+ struct device_attribute *attr, char *buf)
+{
+ struct dev_ext_attribute *eattr;
+
+ eattr = container_of(attr, struct dev_ext_attribute, attr);
+
+ return sprintf(buf, "%s\n", (char *)eattr->var);
+}
+
+/*
+ * PMU event attributes
+ */
+ssize_t hisi_event_sysfs_show(struct device *dev,
+ struct device_attribute *attr, char *page)
+{
+ struct dev_ext_attribute *eattr;
+
+ eattr = container_of(attr, struct dev_ext_attribute, attr);
+
+ return sprintf(page, "config=0x%lx\n", (unsigned long)eattr->var);
+}
+
+/*
+ * sysfs cpumask attributes
+ */
+ssize_t hisi_cpumask_sysfs_show(struct device *dev,
+ struct device_attribute *attr, char *buf)
+{
+ struct hisi_pmu *hisi_pmu = to_hisi_pmu(dev_get_drvdata(dev));
+
+ return cpumap_print_to_pagebuf(true, buf, &hisi_pmu->cpus);
+}
+
+/* Read Super CPU cluster and CPU cluster ID from MPIDR_EL1 */
+void hisi_read_sccl_and_ccl_id(u32 *sccl_id, u32 *ccl_id)
+{
+ u64 mpidr;
+
+ mpidr = read_cpuid_mpidr();
+ if (mpidr & MPIDR_MT_BITMASK) {
+ if (sccl_id)
+ *sccl_id = MPIDR_AFFINITY_LEVEL(mpidr, 3);
+ if (ccl_id)
+ *ccl_id = MPIDR_AFFINITY_LEVEL(mpidr, 2);
+ } else {
+ if (sccl_id)
+ *sccl_id = MPIDR_AFFINITY_LEVEL(mpidr, 2);
+ if (ccl_id)
+ *ccl_id = MPIDR_AFFINITY_LEVEL(mpidr, 1);
+ }
+}
+
+static bool hisi_validate_event_group(struct perf_event *event)
+{
+ struct perf_event *sibling, *leader = event->group_leader;
+ struct hisi_pmu *hisi_pmu = to_hisi_pmu(event->pmu);
+ /* Include count for the event */
+ int counters = 1;
+
+ /*
+ * We must NOT create groups containing mixed PMUs, although
+ * software events are acceptable
+ */
+ if (leader->pmu != event->pmu && !is_software_event(leader))
+ return false;
+
+ /* Increment counter for the leader */
+ counters++;
+
+ list_for_each_entry(sibling, &event->group_leader->sibling_list,
+ group_entry) {
+ if (is_software_event(sibling))
+ continue;
+ if (sibling->pmu != event->pmu)
+ return false;
+ /* Increment counter for each sibling */
+ counters++;
+ }
+
+ /* The group can not count events more than the counters in the HW */
+ return counters <= hisi_pmu->num_counters;
+}
+
+int hisi_uncore_pmu_counter_valid(struct hisi_pmu *hisi_pmu, int idx)
+{
+ return (idx >= 0 && idx < hisi_pmu->num_counters);
+}
+
+int hisi_uncore_pmu_get_event_idx(struct perf_event *event)
+{
+ struct hisi_pmu *hisi_pmu = to_hisi_pmu(event->pmu);
+ unsigned long *used_mask = hisi_pmu->pmu_events.used_mask;
+ u32 num_counters = hisi_pmu->num_counters;
+ int idx;
+
+ idx = find_first_zero_bit(used_mask, num_counters);
+ if (idx == num_counters)
+ return -EAGAIN;
+
+ set_bit(idx, used_mask);
+
+ return idx;
+}
+
+static void hisi_uncore_pmu_clear_event_idx(struct hisi_pmu *hisi_pmu, int idx)
+{
+ if (!hisi_uncore_pmu_counter_valid(hisi_pmu, idx)) {
+ dev_err(hisi_pmu->dev, "Unsupported event index:%d!\n", idx);
+ return;
+ }
+
+ clear_bit(idx, hisi_pmu->pmu_events.used_mask);
+}
+
+int hisi_uncore_pmu_event_init(struct perf_event *event)
+{
+ struct hw_perf_event *hwc = &event->hw;
+ struct hisi_pmu *hisi_pmu;
+ int cpu;
+
+ if (event->attr.type != event->pmu->type)
+ return -ENOENT;
+
+ /*
+ * We do not support sampling as the counters are all
+ * shared by all CPU cores in a CPU die(SCCL). Also we
+ * do not support attach to a task(per-process mode)
+ */
+ if (is_sampling_event(event) || event->attach_state & PERF_ATTACH_TASK)
+ return -EOPNOTSUPP;
+
+ /* counters do not have these bits */
+ if (event->attr.exclude_user ||
+ event->attr.exclude_kernel ||
+ event->attr.exclude_host ||
+ event->attr.exclude_guest ||
+ event->attr.exclude_hv ||
+ event->attr.exclude_idle)
+ return -EINVAL;
+
+ /*
+ * The uncore counters not specific to any CPU, so cannot
+ * support per-task
+ */
+ if (event->cpu < 0)
+ return -EINVAL;
+
+ /*
+ * Validate if the events in group does not exceed the
+ * available counters in hardware.
+ */
+ if (!hisi_validate_event_group(event))
+ return -EINVAL;
+
+ /*
+ * We don't assign an index until we actually place the event onto
+ * hardware. Use -1 to signify that we haven't decided where to put it
+ * yet.
+ */
+ hwc->idx = -1;
+ hwc->config_base = event->attr.config;
+
+ /* Select an available CPU to monitor events in this PMU */
+ hisi_pmu = to_hisi_pmu(event->pmu);
+ cpu = cpumask_first(&hisi_pmu->cpus);
+ if (cpu >= nr_cpu_ids)
+ return -EINVAL;
+
+ /* Enforce to use the same CPU for all events in this PMU */
+ event->cpu = cpu;
+
+ return 0;
+}
+
+/*
+ * Set the counter to count the event that we're interested in,
+ * and enable counter and interrupt.
+ */
+static void hisi_uncore_pmu_enable_event(struct perf_event *event)
+{
+ struct hisi_pmu *hisi_pmu = to_hisi_pmu(event->pmu);
+ struct hw_perf_event *hwc = &event->hw;
+
+ /*
+ * Write event code in event select registers(for DDRC PMU,
+ * event has been mapped to fixed-purpose counter, there is
+ * no need to write event type).
+ */
+ if (hisi_pmu->ops->write_evtype)
+ hisi_pmu->ops->write_evtype(hisi_pmu, hwc->idx,
+ HISI_GET_EVENTID(event));
+
+ /* Enable the hardware event interrupt and counting */
+ hisi_pmu->ops->enable_counter_int(hisi_pmu, hwc);
+ hisi_pmu->ops->enable_counter(hisi_pmu, hwc);
+}
+
+/*
+ * Disable counting and interrupt.
+ */
+static void hisi_uncore_pmu_disable_event(struct perf_event *event)
+{
+ struct hisi_pmu *hisi_pmu = to_hisi_pmu(event->pmu);
+ struct hw_perf_event *hwc = &event->hw;
+
+ hisi_pmu->ops->disable_counter(hisi_pmu, hwc);
+ hisi_pmu->ops->disable_counter_int(hisi_pmu, hwc);
+}
+
+void hisi_uncore_pmu_set_event_period(struct perf_event *event)
+{
+ struct hisi_pmu *hisi_pmu = to_hisi_pmu(event->pmu);
+ struct hw_perf_event *hwc = &event->hw;
+
+ /*
+ * The HiSilicon PMU counters support 32 bits or 48 bits, depending on
+ * the PMU. We reduce it to 2^(counter_bits - 1) to account for the
+ * extreme interrupt latency. So we could hopefully handle the overflow
+ * interrupt before another 2^(counter_bits - 1) events occur and the
+ * counter overtakes its previous value.
+ */
+ u64 val = BIT_ULL(hisi_pmu->counter_bits - 1);
+
+ local64_set(&hwc->prev_count, val);
+ /* Write start value to the hardware event counter */
+ hisi_pmu->ops->write_counter(hisi_pmu, hwc, val);
+}
+
+void hisi_uncore_pmu_event_update(struct perf_event *event)
+{
+ struct hisi_pmu *hisi_pmu = to_hisi_pmu(event->pmu);
+ struct hw_perf_event *hwc = &event->hw;
+ u64 delta, prev_raw_count, new_raw_count;
+
+ do {
+ /* Read the count from the counter register */
+ new_raw_count = hisi_pmu->ops->read_counter(hisi_pmu, hwc);
+ prev_raw_count = local64_read(&hwc->prev_count);
+ } while (local64_cmpxchg(&hwc->prev_count, prev_raw_count,
+ new_raw_count) != prev_raw_count);
+ /*
+ * compute the delta
+ */
+ delta = (new_raw_count - prev_raw_count) &
+ HISI_MAX_PERIOD(hisi_pmu->counter_bits);
+ local64_add(delta, &event->count);
+}
+
+void hisi_uncore_pmu_start(struct perf_event *event, int flags)
+{
+ struct hisi_pmu *hisi_pmu = to_hisi_pmu(event->pmu);
+ struct hw_perf_event *hwc = &event->hw;
+
+ if (WARN_ON_ONCE(!(hwc->state & PERF_HES_STOPPED)))
+ return;
+
+ WARN_ON_ONCE(!(hwc->state & PERF_HES_UPTODATE));
+ hwc->state = 0;
+ hisi_uncore_pmu_set_event_period(event);
+
+ if (flags & PERF_EF_RELOAD) {
+ u64 prev_raw_count = local64_read(&hwc->prev_count);
+
+ hisi_pmu->ops->write_counter(hisi_pmu, hwc, prev_raw_count);
+ }
+
+ hisi_uncore_pmu_enable_event(event);
+ perf_event_update_userpage(event);
+}
+
+void hisi_uncore_pmu_stop(struct perf_event *event, int flags)
+{
+ struct hw_perf_event *hwc = &event->hw;
+
+ hisi_uncore_pmu_disable_event(event);
+ WARN_ON_ONCE(hwc->state & PERF_HES_STOPPED);
+ hwc->state |= PERF_HES_STOPPED;
+
+ if (hwc->state & PERF_HES_UPTODATE)
+ return;
+
+ /* Read hardware counter and update the Perf counter statistics */
+ hisi_uncore_pmu_event_update(event);
+ hwc->state |= PERF_HES_UPTODATE;
+}
+
+int hisi_uncore_pmu_add(struct perf_event *event, int flags)
+{
+ struct hisi_pmu *hisi_pmu = to_hisi_pmu(event->pmu);
+ struct hw_perf_event *hwc = &event->hw;
+ int idx;
+
+ hwc->state = PERF_HES_STOPPED | PERF_HES_UPTODATE;
+
+ /* Get an available counter index for counting */
+ idx = hisi_pmu->ops->get_event_idx(event);
+ if (idx < 0)
+ return idx;
+
+ event->hw.idx = idx;
+ hisi_pmu->pmu_events.hw_events[idx] = event;
+
+ if (flags & PERF_EF_START)
+ hisi_uncore_pmu_start(event, PERF_EF_RELOAD);
+
+ return 0;
+}
+
+void hisi_uncore_pmu_del(struct perf_event *event, int flags)
+{
+ struct hisi_pmu *hisi_pmu = to_hisi_pmu(event->pmu);
+ struct hw_perf_event *hwc = &event->hw;
+
+ hisi_uncore_pmu_stop(event, PERF_EF_UPDATE);
+ hisi_uncore_pmu_clear_event_idx(hisi_pmu, hwc->idx);
+ perf_event_update_userpage(event);
+ hisi_pmu->pmu_events.hw_events[hwc->idx] = NULL;
+}
+
+struct hisi_pmu *hisi_pmu_alloc(struct device *dev, u32 num_cntrs)
+{
+ struct hisi_pmu *hisi_pmu;
+ struct hisi_pmu_hwevents *pmu_events;
+
+ hisi_pmu = devm_kzalloc(dev, sizeof(*hisi_pmu), GFP_KERNEL);
+ if (!hisi_pmu)
+ return ERR_PTR(-ENOMEM);
+
+ pmu_events = &hisi_pmu->pmu_events;
+ pmu_events->hw_events = devm_kcalloc(dev,
+ num_cntrs,
+ sizeof(*pmu_events->hw_events),
+ GFP_KERNEL);
+ if (!pmu_events->hw_events)
+ return ERR_PTR(-ENOMEM);
+
+ pmu_events->used_mask = devm_kcalloc(dev,
+ BITS_TO_LONGS(num_cntrs),
+ sizeof(*pmu_events->used_mask),
+ GFP_KERNEL);
+ if (!pmu_events->used_mask)
+ return ERR_PTR(-ENOMEM);
+
+ return hisi_pmu;
+}
+
+void hisi_uncore_pmu_read(struct perf_event *event)
+{
+ /* Read hardware counter and update the perf counter statistics */
+ hisi_uncore_pmu_event_update(event);
+}
+
+void hisi_uncore_pmu_enable(struct pmu *pmu)
+{
+ struct hisi_pmu *hisi_pmu = to_hisi_pmu(pmu);
+ int enabled = bitmap_weight(hisi_pmu->pmu_events.used_mask,
+ hisi_pmu->num_counters);
+
+ if (!enabled)
+ return;
+
+ hisi_pmu->ops->start_counters(hisi_pmu);
+}
+
+void hisi_uncore_pmu_disable(struct pmu *pmu)
+{
+ struct hisi_pmu *hisi_pmu = to_hisi_pmu(pmu);
+
+ hisi_pmu->ops->stop_counters(hisi_pmu);
+}
diff --git a/drivers/perf/hisilicon/hisi_uncore_pmu.h b/drivers/perf/hisilicon/hisi_uncore_pmu.h
new file mode 100644
index 0000000..5bb099f
--- /dev/null
+++ b/drivers/perf/hisilicon/hisi_uncore_pmu.h
@@ -0,0 +1,103 @@
+/*
+ * HiSilicon SoC Hardware event counters support
+ *
+ * Copyright (C) 2017 Hisilicon Limited
+ * Author: Anurup M <anurup.m@huawei.com>
+ * Shaokun Zhang <zhangshaokun@hisilicon.com>
+ *
+ * This code is based on the uncore PMUs like arm-cci and arm-ccn.
+ *
+ * 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.
+ */
+#ifndef __HISI_UNCORE_PMU_H__
+#define __HISI_UNCORE_PMU_H__
+
+#include <linux/cpumask.h>
+#include <linux/device.h>
+#include <linux/kernel.h>
+#include <linux/perf_event.h>
+#include <linux/types.h>
+
+#undef pr_fmt
+#define pr_fmt(fmt) "hisi_pmu: " fmt
+
+#define to_hisi_pmu(p) (container_of(p, struct hisi_pmu, pmu))
+
+#define HISI_PMU_ATTR(_name, _func, _config) \
+ (&((struct dev_ext_attribute[]) { \
+ { __ATTR(_name, 0444, _func, NULL), (void *)_config } \
+ })[0].attr.attr)
+
+#define HISI_PMU_FORMAT_ATTR(_name, _config) \
+ HISI_PMU_ATTR(_name, hisi_format_sysfs_show, (void *)_config)
+#define HISI_PMU_EVENT_ATTR(_name, _config) \
+ HISI_PMU_ATTR(_name, hisi_event_sysfs_show, (unsigned long)_config)
+
+struct hisi_pmu;
+
+struct hisi_uncore_ops {
+ void (*write_evtype)(struct hisi_pmu *, int, u32);
+ int (*get_event_idx)(struct perf_event *);
+ u64 (*read_counter)(struct hisi_pmu *, struct hw_perf_event *);
+ void (*write_counter)(struct hisi_pmu *, struct hw_perf_event *, u64);
+ void (*enable_counter)(struct hisi_pmu *, struct hw_perf_event *);
+ void (*disable_counter)(struct hisi_pmu *, struct hw_perf_event *);
+ void (*enable_counter_int)(struct hisi_pmu *, struct hw_perf_event *);
+ void (*disable_counter_int)(struct hisi_pmu *, struct hw_perf_event *);
+ void (*start_counters)(struct hisi_pmu *);
+ void (*stop_counters)(struct hisi_pmu *);
+};
+
+struct hisi_pmu_hwevents {
+ struct perf_event **hw_events;
+ unsigned long *used_mask;
+};
+
+/* Generic pmu struct for different pmu types */
+struct hisi_pmu {
+ const char *name;
+ struct pmu pmu;
+ const struct hisi_uncore_ops *ops;
+ struct hisi_pmu_hwevents pmu_events;
+ cpumask_t cpus;
+ int irq;
+ struct device *dev;
+ struct hlist_node node;
+ u32 sccl_id;
+ u32 ccl_id;
+ /* Hardware information for different pmu types */
+ void __iomem *base;
+ union {
+ u32 ddrc_chn_id;
+ u32 l3c_tag_id;
+ u32 hha_uid;
+ };
+ int num_counters;
+ int num_events;
+ int counter_bits;
+};
+
+int hisi_uncore_pmu_counter_valid(struct hisi_pmu *hisi_pmu, int idx);
+int hisi_uncore_pmu_get_event_idx(struct perf_event *event);
+void hisi_uncore_pmu_read(struct perf_event *event);
+int hisi_uncore_pmu_add(struct perf_event *event, int flags);
+void hisi_uncore_pmu_del(struct perf_event *event, int flags);
+void hisi_uncore_pmu_start(struct perf_event *event, int flags);
+void hisi_uncore_pmu_stop(struct perf_event *event, int flags);
+void hisi_uncore_pmu_set_event_period(struct perf_event *event);
+void hisi_uncore_pmu_event_update(struct perf_event *event);
+int hisi_uncore_pmu_event_init(struct perf_event *event);
+int hisi_uncore_pmu_setup(struct hisi_pmu *hisi_pmu, const char *pmu_name);
+void hisi_uncore_pmu_enable(struct pmu *pmu);
+void hisi_uncore_pmu_disable(struct pmu *pmu);
+struct hisi_pmu *hisi_pmu_alloc(struct device *dev, u32 num_cntrs);
+ssize_t hisi_event_sysfs_show(struct device *dev,
+ struct device_attribute *attr, char *buf);
+ssize_t hisi_format_sysfs_show(struct device *dev,
+ struct device_attribute *attr, char *buf);
+ssize_t hisi_cpumask_sysfs_show(struct device *dev,
+ struct device_attribute *attr, char *buf);
+void hisi_read_sccl_and_ccl_id(u32 *sccl_id, u32 *ccl_id);
+#endif /* __HISI_UNCORE_PMU_H__ */
--
1.9.1
^ permalink raw reply related [flat|nested] 19+ messages in thread* Re: [PATCH v4 2/6] perf: hisi: Add support for HiSilicon SoC uncore PMU driver
2017-07-25 12:10 ` [PATCH v4 2/6] perf: hisi: Add support for HiSilicon SoC uncore " Shaokun Zhang
@ 2017-08-15 10:16 ` Mark Rutland
2017-08-17 3:08 ` Zhangshaokun
0 siblings, 1 reply; 19+ messages in thread
From: Mark Rutland @ 2017-08-15 10:16 UTC (permalink / raw)
To: Shaokun Zhang
Cc: will.deacon, jonathan.cameron, linux-arm-kernel, linux-kernel,
linux-doc, linuxarm
Hi,
On Tue, Jul 25, 2017 at 08:10:38PM +0800, Shaokun Zhang wrote:
> +/* Read Super CPU cluster and CPU cluster ID from MPIDR_EL1 */
> +void hisi_read_sccl_and_ccl_id(u32 *sccl_id, u32 *ccl_id)
> +{
> + u64 mpidr;
> +
> + mpidr = read_cpuid_mpidr();
> + if (mpidr & MPIDR_MT_BITMASK) {
> + if (sccl_id)
> + *sccl_id = MPIDR_AFFINITY_LEVEL(mpidr, 3);
> + if (ccl_id)
> + *ccl_id = MPIDR_AFFINITY_LEVEL(mpidr, 2);
> + } else {
> + if (sccl_id)
> + *sccl_id = MPIDR_AFFINITY_LEVEL(mpidr, 2);
> + if (ccl_id)
> + *ccl_id = MPIDR_AFFINITY_LEVEL(mpidr, 1);
> + }
> +}
How exactly are SCCLs organised w.r.t. MPIDRS?
Is this guaranteed to be correct for future SoCs?
It would be nicer if this were described explicitly by FW rather than
guessed at based on the MPIDR.
> +static bool hisi_validate_event_group(struct perf_event *event)
> +{
> + struct perf_event *sibling, *leader = event->group_leader;
> + struct hisi_pmu *hisi_pmu = to_hisi_pmu(event->pmu);
> + /* Include count for the event */
> + int counters = 1;
> +
> + /*
> + * We must NOT create groups containing mixed PMUs, although
> + * software events are acceptable
> + */
> + if (leader->pmu != event->pmu && !is_software_event(leader))
> + return false;
> +
> + /* Increment counter for the leader */
> + counters++;
If this event is the leader, you account for it twice.
I guess you get away with that assuming you have at least two counters,
but it's less than ideal.
> +
> + list_for_each_entry(sibling, &event->group_leader->sibling_list,
> + group_entry) {
> + if (is_software_event(sibling))
> + continue;
> + if (sibling->pmu != event->pmu)
> + return false;
> + /* Increment counter for each sibling */
> + counters++;
> + }
> +
> + /* The group can not count events more than the counters in the HW */
> + return counters <= hisi_pmu->num_counters;
> +}
[...]
> +/*
> + * Set the counter to count the event that we're interested in,
> + * and enable counter and interrupt.
> + */
> +static void hisi_uncore_pmu_enable_event(struct perf_event *event)
> +{
> + struct hisi_pmu *hisi_pmu = to_hisi_pmu(event->pmu);
> + struct hw_perf_event *hwc = &event->hw;
> +
> + /*
> + * Write event code in event select registers(for DDRC PMU,
> + * event has been mapped to fixed-purpose counter, there is
> + * no need to write event type).
> + */
> + if (hisi_pmu->ops->write_evtype)
> + hisi_pmu->ops->write_evtype(hisi_pmu, hwc->idx,
> + HISI_GET_EVENTID(event));
It looks like this is the only op which might be NULL. It would be
cleaner for the DDRC PMU code to provide an empty callback.
[...]
> +struct hisi_pmu *hisi_pmu_alloc(struct device *dev, u32 num_cntrs)
> +{
> + struct hisi_pmu *hisi_pmu;
> + struct hisi_pmu_hwevents *pmu_events;
> +
> + hisi_pmu = devm_kzalloc(dev, sizeof(*hisi_pmu), GFP_KERNEL);
> + if (!hisi_pmu)
> + return ERR_PTR(-ENOMEM);
> +
> + pmu_events = &hisi_pmu->pmu_events;
> + pmu_events->hw_events = devm_kcalloc(dev,
> + num_cntrs,
> + sizeof(*pmu_events->hw_events),
> + GFP_KERNEL);
> + if (!pmu_events->hw_events)
> + return ERR_PTR(-ENOMEM);
> +
> + pmu_events->used_mask = devm_kcalloc(dev,
> + BITS_TO_LONGS(num_cntrs),
> + sizeof(*pmu_events->used_mask),
> + GFP_KERNEL);
How big can num_counters be?
Assuming it's not too big, it would be nicer to embed these within the
hisi_pmu_hwevents.
[...]
> +
> +/* Generic pmu struct for different pmu types */
> +struct hisi_pmu {
> + const char *name;
> + struct pmu pmu;
struct pmu has a name field. Why do we need another?
> + union {
> + u32 ddrc_chn_id;
> + u32 l3c_tag_id;
> + u32 hha_uid;
> + };
This would be simpler as a `u32 id` rather than a union.
> + int num_counters;
> + int num_events;
Subsequent patches intialise num_events, but it is never used. Was it
supposed to be checked at event_init time? Or is it unnnecessary?
Thanks,
Mark.
^ permalink raw reply [flat|nested] 19+ messages in thread* Re: [PATCH v4 2/6] perf: hisi: Add support for HiSilicon SoC uncore PMU driver
2017-08-15 10:16 ` Mark Rutland
@ 2017-08-17 3:08 ` Zhangshaokun
0 siblings, 0 replies; 19+ messages in thread
From: Zhangshaokun @ 2017-08-17 3:08 UTC (permalink / raw)
To: Mark Rutland
Cc: linux-doc, will.deacon, linuxarm, linux-kernel, linux-arm-kernel
Hi Mark,
On 2017/8/15 18:16, Mark Rutland wrote:
> Hi,
>
> On Tue, Jul 25, 2017 at 08:10:38PM +0800, Shaokun Zhang wrote:
>> +/* Read Super CPU cluster and CPU cluster ID from MPIDR_EL1 */
>> +void hisi_read_sccl_and_ccl_id(u32 *sccl_id, u32 *ccl_id)
>> +{
>> + u64 mpidr;
>> +
>> + mpidr = read_cpuid_mpidr();
>> + if (mpidr & MPIDR_MT_BITMASK) {
>> + if (sccl_id)
>> + *sccl_id = MPIDR_AFFINITY_LEVEL(mpidr, 3);
>> + if (ccl_id)
>> + *ccl_id = MPIDR_AFFINITY_LEVEL(mpidr, 2);
>> + } else {
>> + if (sccl_id)
>> + *sccl_id = MPIDR_AFFINITY_LEVEL(mpidr, 2);
>> + if (ccl_id)
>> + *ccl_id = MPIDR_AFFINITY_LEVEL(mpidr, 1);
>> + }
>> +}
>
> How exactly are SCCLs organised w.r.t. MPIDRS?
>
For single-thread core, sccl_id is in MPIDR[aff2] and ccl_id is MPIDR[aff1];
For MT core, sccl_id is in MPIDR[aff3] and ccl_id in MPIDR[aff2].
I shall add comments here.
> Is this guaranteed to be correct for future SoCs?
>
Sorry that it is uncertain.
> It would be nicer if this were described explicitly by FW rather than
> guessed at based on the MPIDR.
>
Whilst I agree, we assume this isn't going to happen now and the logic
can be updated to support this if it we have more complex topology in
the future.
>> +static bool hisi_validate_event_group(struct perf_event *event)
>> +{
>> + struct perf_event *sibling, *leader = event->group_leader;
>> + struct hisi_pmu *hisi_pmu = to_hisi_pmu(event->pmu);
>> + /* Include count for the event */
>> + int counters = 1;
>> +
>> + /*
>> + * We must NOT create groups containing mixed PMUs, although
>> + * software events are acceptable
>> + */
>> + if (leader->pmu != event->pmu && !is_software_event(leader))
>> + return false;
>> +
>> + /* Increment counter for the leader */
>> + counters++;
>
> If this event is the leader, you account for it twice.
>
> I guess you get away with that assuming you have at least two counters,
> but it's less than ideal.
>
We update this as per
https://marc.info/?l=linux-arm-kernel&m=149096885106554&w=2
Any thoughts to avoid this issue?
>> +
>> + list_for_each_entry(sibling, &event->group_leader->sibling_list,
>> + group_entry) {
>> + if (is_software_event(sibling))
>> + continue;
>> + if (sibling->pmu != event->pmu)
>> + return false;
>> + /* Increment counter for each sibling */
>> + counters++;
>> + }
>> +
>> + /* The group can not count events more than the counters in the HW */
>> + return counters <= hisi_pmu->num_counters;
>> +}
>
> [...]
>
>> +/*
>> + * Set the counter to count the event that we're interested in,
>> + * and enable counter and interrupt.
>> + */
>> +static void hisi_uncore_pmu_enable_event(struct perf_event *event)
>> +{
>> + struct hisi_pmu *hisi_pmu = to_hisi_pmu(event->pmu);
>> + struct hw_perf_event *hwc = &event->hw;
>> +
>> + /*
>> + * Write event code in event select registers(for DDRC PMU,
>> + * event has been mapped to fixed-purpose counter, there is
>> + * no need to write event type).
>> + */
>> + if (hisi_pmu->ops->write_evtype)
>> + hisi_pmu->ops->write_evtype(hisi_pmu, hwc->idx,
>> + HISI_GET_EVENTID(event));
>
> It looks like this is the only op which might be NULL. It would be
> cleaner for the DDRC PMU code to provide an empty callback.
>
Ok.
> [...]
>
>> +struct hisi_pmu *hisi_pmu_alloc(struct device *dev, u32 num_cntrs)
>> +{
>> + struct hisi_pmu *hisi_pmu;
>> + struct hisi_pmu_hwevents *pmu_events;
>> +
>> + hisi_pmu = devm_kzalloc(dev, sizeof(*hisi_pmu), GFP_KERNEL);
>> + if (!hisi_pmu)
>> + return ERR_PTR(-ENOMEM);
>> +
>> + pmu_events = &hisi_pmu->pmu_events;
>> + pmu_events->hw_events = devm_kcalloc(dev,
>> + num_cntrs,
>> + sizeof(*pmu_events->hw_events),
>> + GFP_KERNEL);
>> + if (!pmu_events->hw_events)
>> + return ERR_PTR(-ENOMEM);
>> +
>> + pmu_events->used_mask = devm_kcalloc(dev,
>> + BITS_TO_LONGS(num_cntrs),
>> + sizeof(*pmu_events->used_mask),
>> + GFP_KERNEL);
>
> How big can num_counters be?
>
At the moment, the max num_counters is 0x10 for HHA PMU.
> Assuming it's not too big, it would be nicer to embed these within the
> hisi_pmu_hwevents.
>
Ok, shall refactor hisi_pmu_hwevents, will use the max num_counters and
remove num_cntrs for hisi_pmu_alloc function.
> [...]
>
>> +
>> +/* Generic pmu struct for different pmu types */
>> +struct hisi_pmu {
>> + const char *name;
>> + struct pmu pmu;
>
> struct pmu has a name field. Why do we need another?
>
It is redundant and shall remove it.
>> + union {
>> + u32 ddrc_chn_id;
>> + u32 l3c_tag_id;
>> + u32 hha_uid;
>> + };
>
> This would be simpler as a `u32 id` rather than a union.
>
Ok.
>> + int num_counters;
>> + int num_events;
>
> Subsequent patches intialise num_events, but it is never used. Was it
> supposed to be checked at event_init time? Or is it unnnecessary?
>
Yes, it is unnecessary and shall remove it.
Thanks.
Shaokun
> Thanks,
> Mark.
> _______________________________________________
> linuxarm mailing list
> linuxarm@huawei.com
> http://rnd-openeuler.huawei.com/mailman/listinfo/linuxarm
>
> .
>
^ permalink raw reply [flat|nested] 19+ messages in thread
* [PATCH v4 3/6] perf: hisi: Add support for HiSilicon SoC L3C PMU driver
2017-07-25 12:10 [PATCH v4 0/6] Add HiSilicon SoC uncore Performance Monitoring Unit driver Shaokun Zhang
2017-07-25 12:10 ` [PATCH v4 1/6] Documentation: perf: hisi: Documentation for HiSilicon SoC PMU driver Shaokun Zhang
2017-07-25 12:10 ` [PATCH v4 2/6] perf: hisi: Add support for HiSilicon SoC uncore " Shaokun Zhang
@ 2017-07-25 12:10 ` Shaokun Zhang
2017-08-15 10:41 ` Mark Rutland
2017-07-25 12:10 ` [PATCH v4 4/6] perf: hisi: Add support for HiSilicon SoC HHA " Shaokun Zhang
` (3 subsequent siblings)
6 siblings, 1 reply; 19+ messages in thread
From: Shaokun Zhang @ 2017-07-25 12:10 UTC (permalink / raw)
To: mark.rutland, will.deacon
Cc: jonathan.cameron, linux-arm-kernel, linux-kernel, linux-doc, linuxarm
This patch adds support for L3C PMU driver in HiSilicon SoC chip, Each
L3C has own control, counter and interrupt registers and is an separate
PMU. For each L3C PMU, it has 8-programable counters and supports 0x60
events, event code is 8-bits and every counter is free-running.
Interrupt is supported to handle counter (48-bits) overflow.
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Signed-off-by: Shaokun Zhang <zhangshaokun@hisilicon.com>
Signed-off-by: Anurup M <anurup.m@huawei.com>
---
drivers/perf/hisilicon/Makefile | 2 +-
drivers/perf/hisilicon/hisi_uncore_l3c_pmu.c | 538 +++++++++++++++++++++++++++
include/linux/cpuhotplug.h | 1 +
3 files changed, 540 insertions(+), 1 deletion(-)
create mode 100644 drivers/perf/hisilicon/hisi_uncore_l3c_pmu.c
diff --git a/drivers/perf/hisilicon/Makefile b/drivers/perf/hisilicon/Makefile
index 2783bb3..4a3d3e6 100644
--- a/drivers/perf/hisilicon/Makefile
+++ b/drivers/perf/hisilicon/Makefile
@@ -1 +1 @@
-obj-$(CONFIG_HISI_PMU) += hisi_uncore_pmu.o
+obj-$(CONFIG_HISI_PMU) += hisi_uncore_pmu.o hisi_uncore_l3c_pmu.o
diff --git a/drivers/perf/hisilicon/hisi_uncore_l3c_pmu.c b/drivers/perf/hisilicon/hisi_uncore_l3c_pmu.c
new file mode 100644
index 0000000..33146bb
--- /dev/null
+++ b/drivers/perf/hisilicon/hisi_uncore_l3c_pmu.c
@@ -0,0 +1,538 @@
+/*
+ * HiSilicon SoC L3C uncore Hardware event counters support
+ *
+ * Copyright (C) 2017 Hisilicon Limited
+ * Author: Anurup M <anurup.m@huawei.com>
+ * Shaokun Zhang <zhangshaokun@hisilicon.com>
+ *
+ * This code is based on the uncore PMUs like arm-cci and arm-ccn.
+ *
+ * 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.
+ */
+#include <linux/acpi.h>
+#include <linux/bug.h>
+#include <linux/cpuhotplug.h>
+#include <linux/interrupt.h>
+#include <linux/irq.h>
+#include <linux/list.h>
+#include <linux/platform_device.h>
+#include <linux/topology.h>
+#include "hisi_uncore_pmu.h"
+
+/* L3C register definition */
+#define L3C_PERF_CTRL 0x0408
+#define L3C_INT_MASK 0x0800
+#define L3C_INT_STATUS 0x0808
+#define L3C_INT_CLEAR 0x080c
+#define L3C_EVENT_CTRL 0x1c00
+#define L3C_EVENT_TYPE0 0x1d00
+#define L3C_CNTR0_LOWER 0x1e00
+
+/* L3C has 8-counters and supports 0x60 events */
+#define L3C_NR_COUNTERS 0x8
+#define L3C_NR_EVENTS 0x60
+
+#define L3C_PERF_CTRL_EN 0x20000
+#define L3C_EVTYPE_NONE 0xff
+
+/*
+ * Select the counter register offset using the counter index
+ * every counter is 48-bits and [48:63] is reserved.
+ */
+static u32 get_counter_reg_off(int cntr_idx)
+{
+ return (L3C_CNTR0_LOWER + (cntr_idx * 8));
+}
+
+static u64 hisi_l3c_pmu_read_counter(struct hisi_pmu *l3c_pmu,
+ struct hw_perf_event *hwc)
+{
+ u32 idx = hwc->idx;
+ u32 reg;
+
+ if (!hisi_uncore_pmu_counter_valid(l3c_pmu, idx)) {
+ dev_err(l3c_pmu->dev, "Unsupported event index:%d!\n", idx);
+ return 0;
+ }
+
+ reg = get_counter_reg_off(idx);
+
+ /* Read 64-bits and the upper 16 bits are Read-As-Zero */
+ return readq(l3c_pmu->base + reg);
+}
+
+static void hisi_l3c_pmu_write_counter(struct hisi_pmu *l3c_pmu,
+ struct hw_perf_event *hwc, u64 val)
+{
+ u32 idx = hwc->idx;
+ u32 reg;
+
+ if (!hisi_uncore_pmu_counter_valid(l3c_pmu, idx)) {
+ dev_err(l3c_pmu->dev, "Unsupported event index:%d!\n", idx);
+ return;
+ }
+
+ reg = get_counter_reg_off(idx);
+ /* Write 64-bits and the upper 16 bits are Writes-Ignored */
+ writeq(val, l3c_pmu->base + reg);
+}
+
+static void hisi_l3c_pmu_write_evtype(struct hisi_pmu *l3c_pmu, int idx,
+ u32 type)
+{
+ u32 reg, reg_idx, shift, val;
+
+ /*
+ * Select the appropriate event select register(L3C_EVENT_TYPE0/1).
+ * There are 2 event select registers for the 8 hardware counters.
+ * Event code is 8-bits and for the former 4 hardware counters,
+ * L3C_EVENT_TYPE0 is chosen. For the latter 4 hardware counters,
+ * L3C_EVENT_TYPE1 is chosen.
+ */
+ reg = L3C_EVENT_TYPE0 + (idx / 4) * 4;
+ reg_idx = idx % 4;
+ shift = 8 * reg_idx;
+
+ /* Write event code to L3C_EVENT_TYPEx Register */
+ val = readl(l3c_pmu->base + reg);
+ val &= ~(L3C_EVTYPE_NONE << shift);
+ val |= (type << shift);
+ writel(val, l3c_pmu->base + reg);
+}
+
+static void hisi_l3c_pmu_start_counters(struct hisi_pmu *l3c_pmu)
+{
+ u32 val;
+
+ /*
+ * Set perf_enable bit in L3C_PERF_CTRL register to start counting
+ * for all enabled counters.
+ */
+ val = readl(l3c_pmu->base + L3C_PERF_CTRL);
+ val |= L3C_PERF_CTRL_EN;
+ writel(val, l3c_pmu->base + L3C_PERF_CTRL);
+}
+
+static void hisi_l3c_pmu_stop_counters(struct hisi_pmu *l3c_pmu)
+{
+ u32 val;
+
+ /*
+ * Clear perf_enable bit in L3C_PERF_CTRL register to stop counting
+ * for all enabled counters.
+ */
+ val = readl(l3c_pmu->base + L3C_PERF_CTRL);
+ val &= ~(L3C_PERF_CTRL_EN);
+ writel(val, l3c_pmu->base + L3C_PERF_CTRL);
+}
+
+static void hisi_l3c_pmu_enable_counter(struct hisi_pmu *l3c_pmu,
+ struct hw_perf_event *hwc)
+{
+ u32 val;
+
+ /* Enable counter index in L3C_EVENT_CTRL register */
+ val = readl(l3c_pmu->base + L3C_EVENT_CTRL);
+ val |= (1 << hwc->idx);
+ writel(val, l3c_pmu->base + L3C_EVENT_CTRL);
+}
+
+static void hisi_l3c_pmu_disable_counter(struct hisi_pmu *l3c_pmu,
+ struct hw_perf_event *hwc)
+{
+ u32 val;
+
+ /* Clear counter index in L3C_EVENT_CTRL register */
+ val = readl(l3c_pmu->base + L3C_EVENT_CTRL);
+ val &= ~(1 << hwc->idx);
+ writel(val, l3c_pmu->base + L3C_EVENT_CTRL);
+}
+
+static void hisi_l3c_pmu_enable_counter_int(struct hisi_pmu *l3c_pmu,
+ struct hw_perf_event *hwc)
+{
+ u32 val;
+
+ val = readl(l3c_pmu->base + L3C_INT_MASK);
+ /* Write 0 to enable interrupt */
+ val &= ~(1 << hwc->idx);
+ writel(val, l3c_pmu->base + L3C_INT_MASK);
+}
+
+static void hisi_l3c_pmu_disable_counter_int(struct hisi_pmu *l3c_pmu,
+ struct hw_perf_event *hwc)
+{
+ u32 val;
+
+ val = readl(l3c_pmu->base + L3C_INT_MASK);
+ /* Write 1 to mask interrupt */
+ val |= (1 << hwc->idx);
+ writel(val, l3c_pmu->base + L3C_INT_MASK);
+}
+
+static irqreturn_t hisi_l3c_pmu_isr(int irq, void *dev_id)
+{
+ struct hisi_pmu *l3c_pmu = dev_id;
+ struct perf_event *event;
+ unsigned long overflown;
+ u32 status;
+ int idx;
+
+ /* Read L3C_INT_STATUS register */
+ status = readl(l3c_pmu->base + L3C_INT_STATUS);
+ if (!status)
+ return IRQ_NONE;
+ overflown = status;
+
+ /*
+ * Find the counter index which overflowed if the bit was set
+ * and handle it.
+ */
+ for_each_set_bit(idx, &overflown, L3C_NR_COUNTERS) {
+ /* Write 1 to clear the IRQ status flag */
+ writel((1 << idx), l3c_pmu->base + L3C_INT_CLEAR);
+
+ /* Get the corresponding event struct */
+ event = l3c_pmu->pmu_events.hw_events[idx];
+ if (!event)
+ continue;
+
+ hisi_uncore_pmu_event_update(event);
+ hisi_uncore_pmu_set_event_period(event);
+ }
+
+ return IRQ_HANDLED;
+}
+
+static int hisi_l3c_pmu_init_irq(struct hisi_pmu *l3c_pmu,
+ struct platform_device *pdev)
+{
+ struct device *dev = &pdev->dev;
+ int irq, ret;
+
+ /* Read and init IRQ */
+ irq = platform_get_irq(pdev, 0);
+ if (irq < 0) {
+ dev_err(dev, "irq init: fail L3C overflow interrupt\n");
+ return irq;
+ }
+
+ ret = devm_request_irq(dev, irq, hisi_l3c_pmu_isr,
+ IRQF_NOBALANCING | IRQF_NO_THREAD,
+ dev_name(dev), l3c_pmu);
+ if (ret < 0) {
+ dev_err(dev, "Fail to request IRQ:%d ret:%d\n", irq, ret);
+ return ret;
+ }
+
+ l3c_pmu->irq = irq;
+
+ return 0;
+}
+
+/* Check if the CPU belongs to the SCCL and CCL of PMU */
+static bool hisi_l3c_is_cpu_in_ccl(struct hisi_pmu *l3c_pmu)
+{
+ u32 ccl_id, sccl_id;
+
+ hisi_read_sccl_and_ccl_id(&sccl_id, &ccl_id);
+
+ if (sccl_id != l3c_pmu->sccl_id)
+ return false;
+
+ if (ccl_id != l3c_pmu->ccl_id)
+ return false;
+
+ /* Return true if matched */
+ return true;
+}
+
+static int hisi_l3c_pmu_online_cpu(unsigned int cpu, struct hlist_node *node)
+{
+ struct hisi_pmu *l3c_pmu;
+
+ l3c_pmu = hlist_entry_safe(node, struct hisi_pmu, node);
+
+ /* Proceed only when CPU belongs to the same SCCL and CCL */
+ if (!hisi_l3c_is_cpu_in_ccl(l3c_pmu))
+ return 0;
+
+ /* If another CPU is already managing the CPU cluster, simply return */
+ if (!cpumask_empty(&l3c_pmu->cpus))
+ return 0;
+
+ /* Use this CPU for event counting in the CCL */
+ cpumask_set_cpu(cpu, &l3c_pmu->cpus);
+
+ /* Overflow interrupt also should use the same CPU */
+ WARN_ON(irq_set_affinity(l3c_pmu->irq, &l3c_pmu->cpus));
+
+ return 0;
+}
+
+static int hisi_l3c_pmu_offline_cpu(unsigned int cpu, struct hlist_node *node)
+{
+ struct hisi_pmu *l3c_pmu;
+ cpumask_t ccl_online_cpus;
+ unsigned int target;
+
+ l3c_pmu = hlist_entry_safe(node, struct hisi_pmu, node);
+
+ /* Proceed only when CPU belongs to the same SCCL and CCL */
+ if (!hisi_l3c_is_cpu_in_ccl(l3c_pmu))
+ return 0;
+
+ /* Proceed if this CPU is used for event counting */
+ if (!cpumask_test_cpu(cpu, &l3c_pmu->cpus))
+ return 0;
+
+ /* Give up ownership of CCL */
+ cpumask_test_and_clear_cpu(cpu, &l3c_pmu->cpus);
+
+ /* Any other CPU for this CCL which is still online */
+ cpumask_and(&ccl_online_cpus, cpu_coregroup_mask(cpu),
+ cpu_online_mask);
+ target = cpumask_any_but(&ccl_online_cpus, cpu);
+ if (target >= nr_cpu_ids)
+ return 0;
+
+ perf_pmu_migrate_context(&l3c_pmu->pmu, cpu, target);
+ /* Use this CPU for event counting in the CCL */
+ cpumask_set_cpu(target, &l3c_pmu->cpus);
+
+ return 0;
+}
+
+static const struct acpi_device_id hisi_l3c_pmu_acpi_match[] = {
+ { "HISI0213", },
+ {},
+};
+MODULE_DEVICE_TABLE(acpi, hisi_l3c_pmu_acpi_match);
+
+static int hisi_l3c_pmu_init_data(struct platform_device *pdev,
+ struct hisi_pmu *l3c_pmu)
+{
+ struct device *dev = &pdev->dev;
+ struct resource *res;
+ unsigned long long id;
+ acpi_status status;
+
+ status = acpi_evaluate_integer(ACPI_HANDLE(dev), "_UID", NULL, &id);
+ if (ACPI_FAILURE(status))
+ return false;
+
+ l3c_pmu->l3c_tag_id = id;
+
+ /* Get the L3C SCCL ID */
+ if (device_property_read_u32(dev, "hisilicon,scl-id",
+ &l3c_pmu->sccl_id)) {
+ dev_err(dev, "Can not read l3c sccl-id!\n");
+ return -EINVAL;
+ }
+
+ /* Get the L3C CPU cluster(CCL) ID */
+ if (device_property_read_u32(dev, "hisilicon,ccl-id",
+ &l3c_pmu->ccl_id)) {
+ dev_err(dev, "Can not read l3c ccl-id!\n");
+ return -EINVAL;
+ }
+
+ res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+ l3c_pmu->base = devm_ioremap_resource(dev, res);
+ if (IS_ERR(l3c_pmu->base)) {
+ dev_err(dev, "ioremap failed for l3c_pmu resource\n");
+ return PTR_ERR(l3c_pmu->base);
+ }
+
+ return 0;
+}
+
+static struct attribute *hisi_l3c_pmu_format_attr[] = {
+ HISI_PMU_FORMAT_ATTR(event, "config:0-7"),
+ NULL,
+};
+
+static const struct attribute_group hisi_l3c_pmu_format_group = {
+ .name = "format",
+ .attrs = hisi_l3c_pmu_format_attr,
+};
+
+static struct attribute *hisi_l3c_pmu_events_attr[] = {
+ HISI_PMU_EVENT_ATTR(rd_cpipe, 0x00),
+ HISI_PMU_EVENT_ATTR(wr_cpipe, 0x01),
+ HISI_PMU_EVENT_ATTR(rd_hit_cpipe, 0x02),
+ HISI_PMU_EVENT_ATTR(wr_hit_cpipe, 0x03),
+ HISI_PMU_EVENT_ATTR(victim_num, 0x04),
+ HISI_PMU_EVENT_ATTR(rd_spipe, 0x20),
+ HISI_PMU_EVENT_ATTR(wr_spipe, 0x21),
+ HISI_PMU_EVENT_ATTR(rd_hit_spipe, 0x22),
+ HISI_PMU_EVENT_ATTR(wr_hit_spipe, 0x23),
+ HISI_PMU_EVENT_ATTR(back_invalid, 0x29),
+ HISI_PMU_EVENT_ATTR(retry_cpu, 0x40),
+ HISI_PMU_EVENT_ATTR(retry_ring, 0x41),
+ HISI_PMU_EVENT_ATTR(prefetch_drop, 0x42),
+ NULL,
+};
+
+static const struct attribute_group hisi_l3c_pmu_events_group = {
+ .name = "events",
+ .attrs = hisi_l3c_pmu_events_attr,
+};
+
+static DEVICE_ATTR(cpumask, 0444, hisi_cpumask_sysfs_show, NULL);
+
+static struct attribute *hisi_l3c_pmu_cpumask_attrs[] = {
+ &dev_attr_cpumask.attr,
+ NULL,
+};
+
+static const struct attribute_group hisi_l3c_pmu_cpumask_attr_group = {
+ .attrs = hisi_l3c_pmu_cpumask_attrs,
+};
+
+static const struct attribute_group *hisi_l3c_pmu_attr_groups[] = {
+ &hisi_l3c_pmu_format_group,
+ &hisi_l3c_pmu_events_group,
+ &hisi_l3c_pmu_cpumask_attr_group,
+ NULL,
+};
+
+static const struct hisi_uncore_ops hisi_uncore_l3c_ops = {
+ .write_evtype = hisi_l3c_pmu_write_evtype,
+ .get_event_idx = hisi_uncore_pmu_get_event_idx,
+ .start_counters = hisi_l3c_pmu_start_counters,
+ .stop_counters = hisi_l3c_pmu_stop_counters,
+ .enable_counter = hisi_l3c_pmu_enable_counter,
+ .disable_counter = hisi_l3c_pmu_disable_counter,
+ .enable_counter_int = hisi_l3c_pmu_enable_counter_int,
+ .disable_counter_int = hisi_l3c_pmu_disable_counter_int,
+ .write_counter = hisi_l3c_pmu_write_counter,
+ .read_counter = hisi_l3c_pmu_read_counter,
+};
+
+static int hisi_l3c_pmu_dev_probe(struct platform_device *pdev,
+ struct hisi_pmu *l3c_pmu)
+{
+ struct device *dev = &pdev->dev;
+ int ret;
+
+ ret = hisi_l3c_pmu_init_data(pdev, l3c_pmu);
+ if (ret)
+ return ret;
+
+ ret = hisi_l3c_pmu_init_irq(l3c_pmu, pdev);
+ if (ret)
+ return ret;
+
+ l3c_pmu->name = devm_kasprintf(dev, GFP_KERNEL, "hisi_l3c%u_%u",
+ l3c_pmu->l3c_tag_id, l3c_pmu->sccl_id);
+ l3c_pmu->num_events = L3C_NR_EVENTS;
+ l3c_pmu->num_counters = L3C_NR_COUNTERS;
+ l3c_pmu->counter_bits = 48;
+ l3c_pmu->ops = &hisi_uncore_l3c_ops;
+ l3c_pmu->dev = dev;
+
+ return 0;
+}
+
+static int hisi_l3c_pmu_probe(struct platform_device *pdev)
+{
+ struct device *dev = &pdev->dev;
+ struct hisi_pmu *l3c_pmu;
+ int ret;
+
+ l3c_pmu = hisi_pmu_alloc(dev, L3C_NR_COUNTERS);
+ if (!l3c_pmu)
+ return -ENOMEM;
+
+ platform_set_drvdata(pdev, l3c_pmu);
+
+ ret = hisi_l3c_pmu_dev_probe(pdev, l3c_pmu);
+ if (ret)
+ return ret;
+
+ ret = cpuhp_state_add_instance(CPUHP_AP_PERF_ARM_HISI_L3_ONLINE,
+ &l3c_pmu->node);
+ if (ret) {
+ dev_err(dev, "Error %d registering hotplug\n", ret);
+ return ret;
+ }
+
+ l3c_pmu->pmu = (struct pmu) {
+ .name = l3c_pmu->name,
+ .task_ctx_nr = perf_invalid_context,
+ .event_init = hisi_uncore_pmu_event_init,
+ .pmu_enable = hisi_uncore_pmu_enable,
+ .pmu_disable = hisi_uncore_pmu_disable,
+ .add = hisi_uncore_pmu_add,
+ .del = hisi_uncore_pmu_del,
+ .start = hisi_uncore_pmu_start,
+ .stop = hisi_uncore_pmu_stop,
+ .read = hisi_uncore_pmu_read,
+ .attr_groups = hisi_l3c_pmu_attr_groups,
+ };
+
+ ret = perf_pmu_register(&l3c_pmu->pmu, l3c_pmu->name, -1);
+ if (ret) {
+ dev_err(l3c_pmu->dev, "l3c_pmu register failed!\n");
+ cpuhp_state_remove_instance(CPUHP_AP_PERF_ARM_HISI_L3_ONLINE,
+ &l3c_pmu->node);
+ }
+
+ return ret;
+}
+
+static int hisi_l3c_pmu_remove(struct platform_device *pdev)
+{
+ struct hisi_pmu *l3c_pmu = platform_get_drvdata(pdev);
+
+ perf_pmu_unregister(&l3c_pmu->pmu);
+ cpuhp_state_remove_instance(CPUHP_AP_PERF_ARM_HISI_L3_ONLINE,
+ &l3c_pmu->node);
+
+ return 0;
+}
+
+static struct platform_driver hisi_l3c_pmu_driver = {
+ .driver = {
+ .name = "hisi_l3c_pmu",
+ .acpi_match_table = ACPI_PTR(hisi_l3c_pmu_acpi_match),
+ },
+ .probe = hisi_l3c_pmu_probe,
+ .remove = hisi_l3c_pmu_remove,
+};
+
+static int __init hisi_l3c_pmu_module_init(void)
+{
+ int ret;
+
+ ret = cpuhp_setup_state_multi(CPUHP_AP_PERF_ARM_HISI_L3_ONLINE,
+ "AP_PERF_ARM_HISI_L3_ONLINE",
+ hisi_l3c_pmu_online_cpu,
+ hisi_l3c_pmu_offline_cpu);
+ if (ret) {
+ pr_err("l3c_pmu_module_init: Error setup hotplug, ret=%d", ret);
+ return ret;
+ }
+
+ ret = platform_driver_register(&hisi_l3c_pmu_driver);
+ if (ret)
+ cpuhp_remove_multi_state(CPUHP_AP_PERF_ARM_HISI_L3_ONLINE);
+
+ return ret;
+}
+module_init(hisi_l3c_pmu_module_init);
+
+static void __exit hisi_l3c_pmu_module_exit(void)
+{
+ platform_driver_unregister(&hisi_l3c_pmu_driver);
+ cpuhp_remove_multi_state(CPUHP_AP_PERF_ARM_HISI_L3_ONLINE);
+}
+module_exit(hisi_l3c_pmu_module_exit);
+
+MODULE_DESCRIPTION("HiSilicon SoC L3C uncore PMU driver");
+MODULE_LICENSE("GPL v2");
+MODULE_AUTHOR("Anurup M <anurup.m@huawei.com>");
+MODULE_AUTHOR("Shaokun Zhang <zhangshaokun@hisilicon.com>");
diff --git a/include/linux/cpuhotplug.h b/include/linux/cpuhotplug.h
index b56573b..2eab157 100644
--- a/include/linux/cpuhotplug.h
+++ b/include/linux/cpuhotplug.h
@@ -136,6 +136,7 @@ enum cpuhp_state {
CPUHP_AP_PERF_S390_SF_ONLINE,
CPUHP_AP_PERF_ARM_CCI_ONLINE,
CPUHP_AP_PERF_ARM_CCN_ONLINE,
+ CPUHP_AP_PERF_ARM_HISI_L3_ONLINE,
CPUHP_AP_PERF_ARM_L2X0_ONLINE,
CPUHP_AP_PERF_ARM_QCOM_L2_ONLINE,
CPUHP_AP_PERF_ARM_QCOM_L3_ONLINE,
--
1.9.1
^ permalink raw reply related [flat|nested] 19+ messages in thread* Re: [PATCH v4 3/6] perf: hisi: Add support for HiSilicon SoC L3C PMU driver
2017-07-25 12:10 ` [PATCH v4 3/6] perf: hisi: Add support for HiSilicon SoC L3C " Shaokun Zhang
@ 2017-08-15 10:41 ` Mark Rutland
2017-08-17 3:31 ` Zhangshaokun
0 siblings, 1 reply; 19+ messages in thread
From: Mark Rutland @ 2017-08-15 10:41 UTC (permalink / raw)
To: Shaokun Zhang
Cc: will.deacon, jonathan.cameron, linux-arm-kernel, linux-kernel,
linux-doc, linuxarm
On Tue, Jul 25, 2017 at 08:10:39PM +0800, Shaokun Zhang wrote:
> This patch adds support for L3C PMU driver in HiSilicon SoC chip, Each
> L3C has own control, counter and interrupt registers and is an separate
> PMU. For each L3C PMU, it has 8-programable counters and supports 0x60
> events, event code is 8-bits and every counter is free-running.
> Interrupt is supported to handle counter (48-bits) overflow.
[...]
> +/* L3C register definition */
> +#define L3C_PERF_CTRL 0x0408
> +#define L3C_INT_MASK 0x0800
> +#define L3C_INT_STATUS 0x0808
> +#define L3C_INT_CLEAR 0x080c
> +#define L3C_EVENT_CTRL 0x1c00
> +#define L3C_EVENT_TYPE0 0x1d00
> +#define L3C_CNTR0_LOWER 0x1e00
Why does this have a _LOWER suffix?
How big is this regsiter? is it part of a pair of registers?
> +
> +/* L3C has 8-counters and supports 0x60 events */
> +#define L3C_NR_COUNTERS 0x8
> +#define L3C_NR_EVENTS 0x60
What exactly is meant by "supports 0x60 events"?
e.g. does tha mean event IDs 0-0x5f are valid?
> +static irqreturn_t hisi_l3c_pmu_isr(int irq, void *dev_id)
> +{
> + struct hisi_pmu *l3c_pmu = dev_id;
> + struct perf_event *event;
> + unsigned long overflown;
> + u32 status;
> + int idx;
> +
> + /* Read L3C_INT_STATUS register */
> + status = readl(l3c_pmu->base + L3C_INT_STATUS);
> + if (!status)
> + return IRQ_NONE;
> + overflown = status;
You don't need the temporary u32 value here; you can assign directly to
an unsigned lnog and do all the manipulation on that.
[...]
> +/* Check if the CPU belongs to the SCCL and CCL of PMU */
> +static bool hisi_l3c_is_cpu_in_ccl(struct hisi_pmu *l3c_pmu)
> +{
> + u32 ccl_id, sccl_id;
> +
> + hisi_read_sccl_and_ccl_id(&sccl_id, &ccl_id);
> +
> + if (sccl_id != l3c_pmu->sccl_id)
> + return false;
> +
> + if (ccl_id != l3c_pmu->ccl_id)
> + return false;
> +
> + /* Return true if matched */
> + return true;
> +}
The conditionals would be simpler as:
return (sccl_id == l3c_pmu->sccl_id &&
ccl_id == l3c_pmu->ccl_id);
> +
> +static int hisi_l3c_pmu_online_cpu(unsigned int cpu, struct hlist_node *node)
> +{
> + struct hisi_pmu *l3c_pmu;
> +
> + l3c_pmu = hlist_entry_safe(node, struct hisi_pmu, node);
> +
> + /* Proceed only when CPU belongs to the same SCCL and CCL */
> + if (!hisi_l3c_is_cpu_in_ccl(l3c_pmu))
> + return 0;
Surely you have a mask of CPUs that you can check instead? You'll need
that for the offline path.
[...]
> +static int hisi_l3c_pmu_offline_cpu(unsigned int cpu, struct hlist_node *node)
> +{
> + struct hisi_pmu *l3c_pmu;
> + cpumask_t ccl_online_cpus;
> + unsigned int target;
> +
> + l3c_pmu = hlist_entry_safe(node, struct hisi_pmu, node);
> +
> + /* Proceed only when CPU belongs to the same SCCL and CCL */
> + if (!hisi_l3c_is_cpu_in_ccl(l3c_pmu))
> + return 0;
Again, surely you can check a pre-computed mask?
> +
> + /* Proceed if this CPU is used for event counting */
> + if (!cpumask_test_cpu(cpu, &l3c_pmu->cpus))
> + return 0;
You need to set up the CPU state regardless of whether there are active
events currently. Otherwise the cpumask can be stale, pointing at an
offline CPU, leaving the PMU unusable.
> +
> + /* Give up ownership of CCL */
> + cpumask_test_and_clear_cpu(cpu, &l3c_pmu->cpus);
> +
> + /* Any other CPU for this CCL which is still online */
> + cpumask_and(&ccl_online_cpus, cpu_coregroup_mask(cpu),
> + cpu_online_mask);
What is cpu_coregroup_mask? I do not think you can rely on that
happening to align with the physical CCL mask.
Instead, please:
* Keep track of which CPU(s) this PMU can be used from, with a cpumask.
Either initialise that at probe time, or add CPUs to that in the
hotplug callback.
* Use only that mask to determine which CPU to move the PMU context to.
* Use an int to hold the current CPU; there's no need to use a mask to
hold what shoule be a single CPU ID.
[...]
> + /* Get the L3C SCCL ID */
> + if (device_property_read_u32(dev, "hisilicon,scl-id",
> + &l3c_pmu->sccl_id)) {
> + dev_err(dev, "Can not read l3c sccl-id!\n");
> + return -EINVAL;
> + }
> +
> + /* Get the L3C CPU cluster(CCL) ID */
> + if (device_property_read_u32(dev, "hisilicon,ccl-id",
> + &l3c_pmu->ccl_id)) {
> + dev_err(dev, "Can not read l3c ccl-id!\n");
> + return -EINVAL;
> + }
Previously, you expect that these happen to match particular bits of the
MPIDR, which vary for multi-threaded cores. Please document this.
> +static int hisi_l3c_pmu_dev_probe(struct platform_device *pdev,
> + struct hisi_pmu *l3c_pmu)
> +{
> + struct device *dev = &pdev->dev;
> + int ret;
> +
> + ret = hisi_l3c_pmu_init_data(pdev, l3c_pmu);
> + if (ret)
> + return ret;
> +
> + ret = hisi_l3c_pmu_init_irq(l3c_pmu, pdev);
> + if (ret)
> + return ret;
> +
> + l3c_pmu->name = devm_kasprintf(dev, GFP_KERNEL, "hisi_l3c%u_%u",
> + l3c_pmu->l3c_tag_id, l3c_pmu->sccl_id);
As mentioned on the documentation patch, it would be nicer for the name
to be hierarchical, i.e. placing the SCCL ID first.
Thanks,
Mark.
^ permalink raw reply [flat|nested] 19+ messages in thread* Re: [PATCH v4 3/6] perf: hisi: Add support for HiSilicon SoC L3C PMU driver
2017-08-15 10:41 ` Mark Rutland
@ 2017-08-17 3:31 ` Zhangshaokun
0 siblings, 0 replies; 19+ messages in thread
From: Zhangshaokun @ 2017-08-17 3:31 UTC (permalink / raw)
To: Mark Rutland
Cc: linux-doc, will.deacon, linuxarm, linux-kernel, linux-arm-kernel
Hi Mark,
On 2017/8/15 18:41, Mark Rutland wrote:
> On Tue, Jul 25, 2017 at 08:10:39PM +0800, Shaokun Zhang wrote:
>> This patch adds support for L3C PMU driver in HiSilicon SoC chip, Each
>> L3C has own control, counter and interrupt registers and is an separate
>> PMU. For each L3C PMU, it has 8-programable counters and supports 0x60
>> events, event code is 8-bits and every counter is free-running.
>> Interrupt is supported to handle counter (48-bits) overflow.
>
> [...]
>
>> +/* L3C register definition */
>> +#define L3C_PERF_CTRL 0x0408
>> +#define L3C_INT_MASK 0x0800
>> +#define L3C_INT_STATUS 0x0808
>> +#define L3C_INT_CLEAR 0x080c
>> +#define L3C_EVENT_CTRL 0x1c00
>> +#define L3C_EVENT_TYPE0 0x1d00
>> +#define L3C_CNTR0_LOWER 0x1e00
>
> Why does this have a _LOWER suffix?
>
> How big is this regsiter? is it part of a pair of registers?
>
Each counter is 48-bits, for counter0, the register offset of the lower
32-bits is 0x1e00 and the higher 16-bits is in 0x1e04 (while the upper
16-bits is reserved for 0x1e04, RAZ and WI), other counters are the same.
>> +
>> +/* L3C has 8-counters and supports 0x60 events */
>> +#define L3C_NR_COUNTERS 0x8
>> +#define L3C_NR_EVENTS 0x60
>
> What exactly is meant by "supports 0x60 events"?
>
> e.g. does tha mean event IDs 0-0x5f are valid?
>
It is event IDs, my apologies to describe it vaguely.
>> +static irqreturn_t hisi_l3c_pmu_isr(int irq, void *dev_id)
>> +{
>> + struct hisi_pmu *l3c_pmu = dev_id;
>> + struct perf_event *event;
>> + unsigned long overflown;
>> + u32 status;
>> + int idx;
>> +
>> + /* Read L3C_INT_STATUS register */
>> + status = readl(l3c_pmu->base + L3C_INT_STATUS);
>> + if (!status)
>> + return IRQ_NONE;
>> + overflown = status;
>
> You don't need the temporary u32 value here; you can assign directly to
> an unsigned lnog and do all the manipulation on that.
>
Ok.
> [...]
>
>> +/* Check if the CPU belongs to the SCCL and CCL of PMU */
>> +static bool hisi_l3c_is_cpu_in_ccl(struct hisi_pmu *l3c_pmu)
>> +{
>> + u32 ccl_id, sccl_id;
>> +
>> + hisi_read_sccl_and_ccl_id(&sccl_id, &ccl_id);
>> +
>> + if (sccl_id != l3c_pmu->sccl_id)
>> + return false;
>> +
>> + if (ccl_id != l3c_pmu->ccl_id)
>> + return false;
>> +
>> + /* Return true if matched */
>> + return true;
>> +}
>
> The conditionals would be simpler as:
>
> return (sccl_id == l3c_pmu->sccl_id &&
> ccl_id == l3c_pmu->ccl_id);
>
Ok.
>> +
>> +static int hisi_l3c_pmu_online_cpu(unsigned int cpu, struct hlist_node *node)
>> +{
>> + struct hisi_pmu *l3c_pmu;
>> +
>> + l3c_pmu = hlist_entry_safe(node, struct hisi_pmu, node);
>> +
>> + /* Proceed only when CPU belongs to the same SCCL and CCL */
>> + if (!hisi_l3c_is_cpu_in_ccl(l3c_pmu))
>> + return 0;
>
> Surely you have a mask of CPUs that you can check instead? You'll need
> that for the offline path.
>
Ok, Shall create the cpumask and update it during CPU hotplug callback.
> [...]
>
>> +static int hisi_l3c_pmu_offline_cpu(unsigned int cpu, struct hlist_node *node)
>> +{
>> + struct hisi_pmu *l3c_pmu;
>> + cpumask_t ccl_online_cpus;
>> + unsigned int target;
>> +
>> + l3c_pmu = hlist_entry_safe(node, struct hisi_pmu, node);
>> +
>> + /* Proceed only when CPU belongs to the same SCCL and CCL */
>> + if (!hisi_l3c_is_cpu_in_ccl(l3c_pmu))
>> + return 0;
>
> Again, surely you can check a pre-computed mask?
>
Ok.
>> +
>> + /* Proceed if this CPU is used for event counting */
>> + if (!cpumask_test_cpu(cpu, &l3c_pmu->cpus))
>> + return 0;
>
> You need to set up the CPU state regardless of whether there are active
> events currently. Otherwise the cpumask can be stale, pointing at an
> offline CPU, leaving the PMU unusable.
>
Ok. Shall update the cpumask and also hisi_pmu::init cpu to hold the next
online CPU
>> +
>> + /* Give up ownership of CCL */
>> + cpumask_test_and_clear_cpu(cpu, &l3c_pmu->cpus);
>> +
>> + /* Any other CPU for this CCL which is still online */
>> + cpumask_and(&ccl_online_cpus, cpu_coregroup_mask(cpu),
>> + cpu_online_mask);
>
> What is cpu_coregroup_mask? I do not think you can rely on that
> happening to align with the physical CCL mask.
>
The cpu_coregroup_mask return the CPU cores within the cluster. So we used this.
> Instead, please:
>
> * Keep track of which CPU(s) this PMU can be used from, with a cpumask.
> Either initialise that at probe time, or add CPUs to that in the
> hotplug callback.
>
> * Use only that mask to determine which CPU to move the PMU context to.
>
> * Use an int to hold the current CPU; there's no need to use a mask to
> hold what shoule be a single CPU ID.
>
Shall modify as suggested.
> [...]
>
>> + /* Get the L3C SCCL ID */
>> + if (device_property_read_u32(dev, "hisilicon,scl-id",
>> + &l3c_pmu->sccl_id)) {
>> + dev_err(dev, "Can not read l3c sccl-id!\n");
>> + return -EINVAL;
>> + }
>> +
>> + /* Get the L3C CPU cluster(CCL) ID */
>> + if (device_property_read_u32(dev, "hisilicon,ccl-id",
>> + &l3c_pmu->ccl_id)) {
>> + dev_err(dev, "Can not read l3c ccl-id!\n");
>> + return -EINVAL;
>> + }
>
> Previously, you expect that these happen to match particular bits of the
> MPIDR, which vary for multi-threaded cores. Please document this.
>
Ok.
>> +static int hisi_l3c_pmu_dev_probe(struct platform_device *pdev,
>> + struct hisi_pmu *l3c_pmu)
>> +{
>> + struct device *dev = &pdev->dev;
>> + int ret;
>> +
>> + ret = hisi_l3c_pmu_init_data(pdev, l3c_pmu);
>> + if (ret)
>> + return ret;
>> +
>> + ret = hisi_l3c_pmu_init_irq(l3c_pmu, pdev);
>> + if (ret)
>> + return ret;
>> +
>> + l3c_pmu->name = devm_kasprintf(dev, GFP_KERNEL, "hisi_l3c%u_%u",
>> + l3c_pmu->l3c_tag_id, l3c_pmu->sccl_id);
>
> As mentioned on the documentation patch, it would be nicer for the name
> to be hierarchical, i.e. placing the SCCL ID first.
>
Surely.
Thanks.
Shaokun
> Thanks,
> Mark.
> _______________________________________________
> linuxarm mailing list
> linuxarm@huawei.com
> http://rnd-openeuler.huawei.com/mailman/listinfo/linuxarm
>
> .
>
^ permalink raw reply [flat|nested] 19+ messages in thread
* [PATCH v4 4/6] perf: hisi: Add support for HiSilicon SoC HHA PMU driver
2017-07-25 12:10 [PATCH v4 0/6] Add HiSilicon SoC uncore Performance Monitoring Unit driver Shaokun Zhang
` (2 preceding siblings ...)
2017-07-25 12:10 ` [PATCH v4 3/6] perf: hisi: Add support for HiSilicon SoC L3C " Shaokun Zhang
@ 2017-07-25 12:10 ` Shaokun Zhang
2017-08-15 11:05 ` Mark Rutland
2017-07-25 12:10 ` [PATCH v4 5/6] perf: hisi: Add support for HiSilicon SoC DDRC " Shaokun Zhang
` (2 subsequent siblings)
6 siblings, 1 reply; 19+ messages in thread
From: Shaokun Zhang @ 2017-07-25 12:10 UTC (permalink / raw)
To: mark.rutland, will.deacon
Cc: jonathan.cameron, linux-arm-kernel, linux-kernel, linux-doc, linuxarm
L3 cache coherence is maintained by Hydra Home Agent (HHA) in HiSilicon
SoC. This patch adds support for HHA PMU driver, Each HHA has own
control, counter and interrupt registers and is an separate PMU. For
each HHA PMU, it has 16-programable counters and supports 0x50 events,
event code is 8-bits and every counter is free-running. Interrupt is
supported to handle counter (48-bits) overflow.
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Signed-off-by: Shaokun Zhang <zhangshaokun@hisilicon.com>
Signed-off-by: Anurup M <anurup.m@huawei.com>
---
drivers/perf/hisilicon/Makefile | 2 +-
drivers/perf/hisilicon/hisi_uncore_hha_pmu.c | 436 +++++++++++++++++++++++++++
2 files changed, 437 insertions(+), 1 deletion(-)
create mode 100644 drivers/perf/hisilicon/hisi_uncore_hha_pmu.c
diff --git a/drivers/perf/hisilicon/Makefile b/drivers/perf/hisilicon/Makefile
index 4a3d3e6..a72afe8 100644
--- a/drivers/perf/hisilicon/Makefile
+++ b/drivers/perf/hisilicon/Makefile
@@ -1 +1 @@
-obj-$(CONFIG_HISI_PMU) += hisi_uncore_pmu.o hisi_uncore_l3c_pmu.o
+obj-$(CONFIG_HISI_PMU) += hisi_uncore_pmu.o hisi_uncore_l3c_pmu.o hisi_uncore_hha_pmu.o
diff --git a/drivers/perf/hisilicon/hisi_uncore_hha_pmu.c b/drivers/perf/hisilicon/hisi_uncore_hha_pmu.c
new file mode 100644
index 0000000..6798d5f
--- /dev/null
+++ b/drivers/perf/hisilicon/hisi_uncore_hha_pmu.c
@@ -0,0 +1,436 @@
+/*
+ * HiSilicon SoC HHA uncore Hardware event counters support
+ *
+ * Copyright (C) 2017 Hisilicon Limited
+ * Author: Shaokun Zhang <zhangshaokun@hisilicon.com>
+ * Anurup M <anurup.m@huawei.com>
+ *
+ * This code is based on the uncore PMUs like arm-cci and arm-ccn.
+ *
+ * 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.
+ */
+#include <linux/acpi.h>
+#include <linux/bug.h>
+#include <linux/interrupt.h>
+#include <linux/irq.h>
+#include <linux/list.h>
+#include <linux/platform_device.h>
+#include "hisi_uncore_pmu.h"
+
+/* HHA register definition */
+#define HHA_INT_MASK 0x0804
+#define HHA_INT_STATUS 0x0808
+#define HHA_INT_CLEAR 0x080C
+#define HHA_PERF_CTRL 0x1E00
+#define HHA_EVENT_CTRL 0x1E04
+#define HHA_EVENT_TYPE0 0x1E80
+#define HHA_CNT0_LOWER 0x1F00
+
+/* HHA has 16-counters and supports 0x50 events */
+#define HHA_NR_COUNTERS 0x10
+#define HHA_NR_EVENTS 0x50
+
+#define HHA_PERF_CTRL_EN 0x1
+#define HHA_EVTYPE_NONE 0xff
+
+#define HHA_EVTYPE_REG(idx) (HHA_EVENT_TYPE0 + 4 * ((idx) / 4))
+
+/*
+ * Select the counter register offset using the counter index
+ * every counter is 48-bits and [48:63] is reserved.
+ */
+static u32 get_counter_reg_off(int cntr_idx)
+{
+ return (HHA_CNT0_LOWER + (cntr_idx * 8));
+}
+
+static u64 hisi_hha_pmu_read_counter(struct hisi_pmu *hha_pmu,
+ struct hw_perf_event *hwc)
+{
+ u32 idx = hwc->idx;
+ u32 reg;
+
+ if (!hisi_uncore_pmu_counter_valid(hha_pmu, idx)) {
+ dev_err(hha_pmu->dev, "Unsupported event index:%d!\n", idx);
+ return 0;
+ }
+
+ reg = get_counter_reg_off(idx);
+
+ /* Read 64 bits and like L3C, top 16 bits are RAZ */
+ return readq(hha_pmu->base + reg);
+}
+
+static void hisi_hha_pmu_write_counter(struct hisi_pmu *hha_pmu,
+ struct hw_perf_event *hwc, u64 val)
+{
+ u32 idx = hwc->idx;
+ u32 reg;
+
+ if (!hisi_uncore_pmu_counter_valid(hha_pmu, idx)) {
+ dev_err(hha_pmu->dev, "Unsupported event index:%d!\n", idx);
+ return;
+ }
+
+ reg = get_counter_reg_off(idx);
+ /* Write 64 bits and like L3C, top 16 bits are WI */
+ writeq(val, hha_pmu->base + reg);
+}
+
+static void hisi_hha_pmu_write_evtype(struct hisi_pmu *hha_pmu, int idx,
+ u32 type)
+{
+ u32 reg, reg_idx, shift, val;
+
+ /*
+ * Select the appropriate event select register(HHA_EVENT_TYPEx).
+ * There are 4 event select registers for the 16 hardware counters.
+ * Event code is 8-bits and for the first 4 hardware counters,
+ * HHA_EVENT_TYPE0 is chosen. For the next 4 hardware counters,
+ * HHA_EVENT_TYPE1 is chosen and so on.
+ */
+ reg = HHA_EVTYPE_REG(idx);
+ reg_idx = idx % 4;
+ shift = 8 * reg_idx;
+
+ /* Write event code to HHA_EVENT_TYPEx register */
+ val = readl(hha_pmu->base + reg);
+ val &= ~(HHA_EVTYPE_NONE << shift);
+ val |= (type << shift);
+ writel(val, hha_pmu->base + reg);
+}
+
+static void hisi_hha_pmu_start_counters(struct hisi_pmu *hha_pmu)
+{
+ u32 val;
+
+ /*
+ * Set perf_enable bit in HHA_PERF_CTRL to start event
+ * counting for all enabled counters.
+ */
+ val = readl(hha_pmu->base + HHA_PERF_CTRL);
+ val |= HHA_PERF_CTRL_EN;
+ writel(val, hha_pmu->base + HHA_PERF_CTRL);
+}
+
+static void hisi_hha_pmu_stop_counters(struct hisi_pmu *hha_pmu)
+{
+ u32 val;
+
+ /*
+ * Clear perf_enable bit in HHA_PERF_CTRL to stop event
+ * counting for all enabled counters.
+ */
+ val = readl(hha_pmu->base + HHA_PERF_CTRL);
+ val &= ~(HHA_PERF_CTRL_EN);
+ writel(val, hha_pmu->base + HHA_PERF_CTRL);
+}
+
+static void hisi_hha_pmu_enable_counter(struct hisi_pmu *hha_pmu,
+ struct hw_perf_event *hwc)
+{
+ u32 val;
+
+ /* Enable counter index in HHA_EVENT_CTRL register */
+ val = readl(hha_pmu->base + HHA_EVENT_CTRL);
+ val |= (1 << hwc->idx);
+ writel(val, hha_pmu->base + HHA_EVENT_CTRL);
+}
+
+static void hisi_hha_pmu_disable_counter(struct hisi_pmu *hha_pmu,
+ struct hw_perf_event *hwc)
+{
+ u32 val;
+
+ /* Clear counter index in HHA_EVENT_CTRL register */
+ val = readl(hha_pmu->base + HHA_EVENT_CTRL);
+ val &= ~(1 << hwc->idx);
+ writel(val, hha_pmu->base + HHA_EVENT_CTRL);
+}
+
+static void hisi_hha_pmu_enable_counter_int(struct hisi_pmu *hha_pmu,
+ struct hw_perf_event *hwc)
+{
+ u32 val;
+
+ /* Write 0 to enable interrupt */
+ val = readl(hha_pmu->base + HHA_INT_MASK);
+ val &= ~(1 << hwc->idx);
+ writel(val, hha_pmu->base + HHA_INT_MASK);
+}
+
+static void hisi_hha_pmu_disable_counter_int(struct hisi_pmu *hha_pmu,
+ struct hw_perf_event *hwc)
+{
+ u32 val;
+
+ /* Write 1 to mask interrupt */
+ val = readl(hha_pmu->base + HHA_INT_MASK);
+ val |= (1 << hwc->idx);
+ writel(val, hha_pmu->base + HHA_INT_MASK);
+}
+
+static irqreturn_t hisi_hha_pmu_isr(int irq, void *dev_id)
+{
+ struct hisi_pmu *hha_pmu = dev_id;
+ struct perf_event *event;
+ unsigned long overflown;
+ u32 status;
+ int idx;
+
+ /* Read HHA_INT_STATUS register */
+ status = readl(hha_pmu->base + HHA_INT_STATUS);
+ if (!status)
+ return IRQ_NONE;
+ overflown = status;
+
+ /*
+ * Find the counter index which overflowed if the bit was set
+ * and handle it
+ */
+ for_each_set_bit(idx, &overflown, HHA_NR_COUNTERS) {
+ /* Write 1 to clear the IRQ status flag */
+ writel((1 << idx), hha_pmu->base + HHA_INT_CLEAR);
+
+ /* Get the corresponding event struct */
+ event = hha_pmu->pmu_events.hw_events[idx];
+ if (!event)
+ continue;
+
+ hisi_uncore_pmu_event_update(event);
+ hisi_uncore_pmu_set_event_period(event);
+ }
+
+ return IRQ_HANDLED;
+}
+
+static int hisi_hha_pmu_init_irq(struct hisi_pmu *hha_pmu,
+ struct platform_device *pdev)
+{
+ struct device *dev = &pdev->dev;
+ int irq, ret;
+
+ /* Read and init IRQ */
+ irq = platform_get_irq(pdev, 0);
+ if (irq < 0) {
+ dev_err(dev, "irq init: fail map HHA overflow interrupt\n");
+ return irq;
+ }
+
+ ret = devm_request_irq(dev, irq, hisi_hha_pmu_isr,
+ IRQF_NOBALANCING | IRQF_NO_THREAD,
+ dev_name(dev), hha_pmu);
+ if (ret < 0) {
+ dev_err(dev, "Fail to request IRQ:%d ret:%d\n", irq, ret);
+ return ret;
+ }
+
+ /* Overflow interrupt also should use the same CPU */
+ WARN_ON(irq_set_affinity(irq, &hha_pmu->cpus));
+
+ return 0;
+}
+
+static const struct acpi_device_id hisi_hha_pmu_acpi_match[] = {
+ { "HISI0243", },
+ {},
+};
+MODULE_DEVICE_TABLE(acpi, hisi_hha_pmu_acpi_match);
+
+static int hisi_hha_pmu_init_data(struct platform_device *pdev,
+ struct hisi_pmu *hha_pmu)
+{
+ struct device *dev = &pdev->dev;
+ struct resource *res;
+ unsigned long long uid;
+ acpi_status status;
+
+ status = acpi_evaluate_integer(ACPI_HANDLE(dev), "_UID", NULL, &uid);
+ if (ACPI_FAILURE(status))
+ return false;
+
+ hha_pmu->hha_uid = uid;
+
+ /* Get the HHA SCCL ID */
+ if (device_property_read_u32(dev, "hisilicon,scl-id",
+ &hha_pmu->sccl_id)) {
+ dev_err(dev, "Can not read hha sccl-id!\n");
+ return -EINVAL;
+ }
+
+ res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+ hha_pmu->base = devm_ioremap_resource(dev, res);
+ if (IS_ERR(hha_pmu->base)) {
+ dev_err(dev, "ioremap failed for hha_pmu resource\n");
+ return PTR_ERR(hha_pmu->base);
+ }
+
+ return 0;
+}
+
+static struct attribute *hisi_hha_pmu_format_attr[] = {
+ HISI_PMU_FORMAT_ATTR(event, "config:0-7"),
+ NULL,
+};
+
+static const struct attribute_group hisi_hha_pmu_format_group = {
+ .name = "format",
+ .attrs = hisi_hha_pmu_format_attr,
+};
+
+static struct attribute *hisi_hha_pmu_events_attr[] = {
+ HISI_PMU_EVENT_ATTR(rx_ops_num, 0x00),
+ HISI_PMU_EVENT_ATTR(rx_outer, 0x01),
+ HISI_PMU_EVENT_ATTR(rx_sccl, 0x02),
+ HISI_PMU_EVENT_ATTR(rx_ccix, 0x03),
+ HISI_PMU_EVENT_ATTR(rx_wbi, 0x04),
+ HISI_PMU_EVENT_ATTR(rx_wbip, 0x05),
+ HISI_PMU_EVENT_ATTR(rx_wtistash, 0x11),
+ HISI_PMU_EVENT_ATTR(rd_ddr_64b, 0x1c),
+ HISI_PMU_EVENT_ATTR(wr_dr_64b, 0x1d),
+ HISI_PMU_EVENT_ATTR(rd_ddr_128b, 0x1e),
+ HISI_PMU_EVENT_ATTR(wr_ddr_128b, 0x1f),
+ HISI_PMU_EVENT_ATTR(spill_num, 0x20),
+ HISI_PMU_EVENT_ATTR(spill_success, 0x21),
+ HISI_PMU_EVENT_ATTR(bi_num, 0x23),
+ HISI_PMU_EVENT_ATTR(mediated_num, 0x32),
+ HISI_PMU_EVENT_ATTR(tx_snp_num, 0x33),
+ HISI_PMU_EVENT_ATTR(tx_snp_outer, 0x34),
+ HISI_PMU_EVENT_ATTR(tx_snp_ccix, 0x35),
+ HISI_PMU_EVENT_ATTR(rx_snprspdata, 0x38),
+ HISI_PMU_EVENT_ATTR(rx_snprsp_outer, 0x3c),
+ HISI_PMU_EVENT_ATTR(sdir-lookup, 0x40),
+ HISI_PMU_EVENT_ATTR(edir-lookup, 0x41),
+ HISI_PMU_EVENT_ATTR(sdir-hit, 0x42),
+ HISI_PMU_EVENT_ATTR(edir-hit, 0x43),
+ HISI_PMU_EVENT_ATTR(sdir-home-migrate, 0x4c),
+ HISI_PMU_EVENT_ATTR(edir-home-migrate, 0x4d),
+ NULL,
+};
+
+static const struct attribute_group hisi_hha_pmu_events_group = {
+ .name = "events",
+ .attrs = hisi_hha_pmu_events_attr,
+};
+
+static DEVICE_ATTR(cpumask, 0444, hisi_cpumask_sysfs_show, NULL);
+
+static struct attribute *hisi_hha_pmu_cpumask_attrs[] = {
+ &dev_attr_cpumask.attr,
+ NULL,
+};
+
+static const struct attribute_group hisi_hha_pmu_cpumask_attr_group = {
+ .attrs = hisi_hha_pmu_cpumask_attrs,
+};
+
+static const struct attribute_group *hisi_hha_pmu_attr_groups[] = {
+ &hisi_hha_pmu_format_group,
+ &hisi_hha_pmu_events_group,
+ &hisi_hha_pmu_cpumask_attr_group,
+ NULL,
+};
+
+static const struct hisi_uncore_ops hisi_uncore_hha_ops = {
+ .write_evtype = hisi_hha_pmu_write_evtype,
+ .get_event_idx = hisi_uncore_pmu_get_event_idx,
+ .start_counters = hisi_hha_pmu_start_counters,
+ .stop_counters = hisi_hha_pmu_stop_counters,
+ .enable_counter = hisi_hha_pmu_enable_counter,
+ .disable_counter = hisi_hha_pmu_disable_counter,
+ .enable_counter_int = hisi_hha_pmu_enable_counter_int,
+ .disable_counter_int = hisi_hha_pmu_disable_counter_int,
+ .write_counter = hisi_hha_pmu_write_counter,
+ .read_counter = hisi_hha_pmu_read_counter,
+};
+
+static int hisi_hha_pmu_dev_probe(struct platform_device *pdev,
+ struct hisi_pmu *hha_pmu)
+{
+ struct device *dev = &pdev->dev;
+ int ret;
+
+ ret = hisi_hha_pmu_init_data(pdev, hha_pmu);
+ if (ret)
+ return ret;
+
+ /* Pick one core to use for cpumask attributes */
+ cpumask_set_cpu(smp_processor_id(), &hha_pmu->cpus);
+
+ ret = hisi_hha_pmu_init_irq(hha_pmu, pdev);
+ if (ret)
+ return ret;
+
+ hha_pmu->name = devm_kasprintf(dev, GFP_KERNEL, "hisi_hha%u_%u",
+ hha_pmu->hha_uid, hha_pmu->sccl_id);
+ hha_pmu->num_events = HHA_NR_EVENTS;
+ hha_pmu->num_counters = HHA_NR_COUNTERS;
+ hha_pmu->counter_bits = 48;
+ hha_pmu->ops = &hisi_uncore_hha_ops;
+ hha_pmu->dev = dev;
+
+ return 0;
+}
+
+static int hisi_hha_pmu_probe(struct platform_device *pdev)
+{
+ struct device *dev = &pdev->dev;
+ struct hisi_pmu *hha_pmu;
+ int ret;
+
+ hha_pmu = hisi_pmu_alloc(dev, HHA_NR_COUNTERS);
+ if (!hha_pmu)
+ return -ENOMEM;
+
+ platform_set_drvdata(pdev, hha_pmu);
+
+ ret = hisi_hha_pmu_dev_probe(pdev, hha_pmu);
+ if (ret)
+ return ret;
+
+ hha_pmu->pmu = (struct pmu) {
+ .name = hha_pmu->name,
+ .task_ctx_nr = perf_invalid_context,
+ .event_init = hisi_uncore_pmu_event_init,
+ .pmu_enable = hisi_uncore_pmu_enable,
+ .pmu_disable = hisi_uncore_pmu_disable,
+ .add = hisi_uncore_pmu_add,
+ .del = hisi_uncore_pmu_del,
+ .start = hisi_uncore_pmu_start,
+ .stop = hisi_uncore_pmu_stop,
+ .read = hisi_uncore_pmu_read,
+ .attr_groups = hisi_hha_pmu_attr_groups,
+ };
+
+ ret = perf_pmu_register(&hha_pmu->pmu, hha_pmu->name, -1);
+ if (ret)
+ dev_err(hha_pmu->dev, "hha_pmu register failed!\n");
+
+ return ret;
+}
+
+static int hisi_hha_pmu_remove(struct platform_device *pdev)
+{
+ struct hisi_pmu *hha_pmu = platform_get_drvdata(pdev);
+
+ perf_pmu_unregister(&hha_pmu->pmu);
+
+ return 0;
+}
+
+static struct platform_driver hha_pmu_driver = {
+ .driver = {
+ .name = "hisi_hha_pmu",
+ .acpi_match_table = ACPI_PTR(hisi_hha_pmu_acpi_match),
+ },
+ .probe = hisi_hha_pmu_probe,
+ .remove = hisi_hha_pmu_remove,
+};
+module_platform_driver(hha_pmu_driver);
+
+MODULE_DESCRIPTION("HiSilicon SoC HHA uncore PMU driver");
+MODULE_LICENSE("GPL v2");
+MODULE_AUTHOR("Shaokun Zhang <zhangshaokun@hisilicon.com>");
+MODULE_AUTHOR("Anurup M <anurup.m@huawei.com>");
--
1.9.1
^ permalink raw reply related [flat|nested] 19+ messages in thread* Re: [PATCH v4 4/6] perf: hisi: Add support for HiSilicon SoC HHA PMU driver
2017-07-25 12:10 ` [PATCH v4 4/6] perf: hisi: Add support for HiSilicon SoC HHA " Shaokun Zhang
@ 2017-08-15 11:05 ` Mark Rutland
2017-08-17 3:38 ` Zhangshaokun
0 siblings, 1 reply; 19+ messages in thread
From: Mark Rutland @ 2017-08-15 11:05 UTC (permalink / raw)
To: Shaokun Zhang
Cc: will.deacon, jonathan.cameron, linux-arm-kernel, linux-kernel,
linux-doc, linuxarm
On Tue, Jul 25, 2017 at 08:10:40PM +0800, Shaokun Zhang wrote:
> +/* HHA register definition */
> +#define HHA_INT_MASK 0x0804
> +#define HHA_INT_STATUS 0x0808
> +#define HHA_INT_CLEAR 0x080C
> +#define HHA_PERF_CTRL 0x1E00
> +#define HHA_EVENT_CTRL 0x1E04
> +#define HHA_EVENT_TYPE0 0x1E80
> +#define HHA_CNT0_LOWER 0x1F00
> +
> +/* HHA has 16-counters and supports 0x50 events */
As with the L3C PMU, what exactly does this mean?
Does this mean event IDs 0-0x4f are valid?
[...]
> +static irqreturn_t hisi_hha_pmu_isr(int irq, void *dev_id)
> +{
> + struct hisi_pmu *hha_pmu = dev_id;
> + struct perf_event *event;
> + unsigned long overflown;
> + u32 status;
> + int idx;
> +
> + /* Read HHA_INT_STATUS register */
> + status = readl(hha_pmu->base + HHA_INT_STATUS);
> + if (!status)
> + return IRQ_NONE;
> + overflown = status;
No need for the u32 temporary here.
[....]
> +static int hisi_hha_pmu_dev_probe(struct platform_device *pdev,
> + struct hisi_pmu *hha_pmu)
> +{
> + struct device *dev = &pdev->dev;
> + int ret;
> +
> + ret = hisi_hha_pmu_init_data(pdev, hha_pmu);
> + if (ret)
> + return ret;
> +
> + /* Pick one core to use for cpumask attributes */
> + cpumask_set_cpu(smp_processor_id(), &hha_pmu->cpus);
> +
Why does this not have the usual event migration callbacks, across CPUs
in the same SCCL?
> + ret = hisi_hha_pmu_init_irq(hha_pmu, pdev);
> + if (ret)
> + return ret;
> +
> + hha_pmu->name = devm_kasprintf(dev, GFP_KERNEL, "hisi_hha%u_%u",
> + hha_pmu->hha_uid, hha_pmu->sccl_id);
As on the doc patch, this should be hierarchical.
Thanks,
Mark
^ permalink raw reply [flat|nested] 19+ messages in thread* Re: [PATCH v4 4/6] perf: hisi: Add support for HiSilicon SoC HHA PMU driver
2017-08-15 11:05 ` Mark Rutland
@ 2017-08-17 3:38 ` Zhangshaokun
0 siblings, 0 replies; 19+ messages in thread
From: Zhangshaokun @ 2017-08-17 3:38 UTC (permalink / raw)
To: Mark Rutland
Cc: linux-doc, will.deacon, linuxarm, linux-kernel, jonathan.cameron,
linux-arm-kernel
Hi Mark,
On 2017/8/15 19:05, Mark Rutland wrote:
> On Tue, Jul 25, 2017 at 08:10:40PM +0800, Shaokun Zhang wrote:
>> +/* HHA register definition */
>> +#define HHA_INT_MASK 0x0804
>> +#define HHA_INT_STATUS 0x0808
>> +#define HHA_INT_CLEAR 0x080C
>> +#define HHA_PERF_CTRL 0x1E00
>> +#define HHA_EVENT_CTRL 0x1E04
>> +#define HHA_EVENT_TYPE0 0x1E80
>> +#define HHA_CNT0_LOWER 0x1F00
>> +
>> +/* HHA has 16-counters and supports 0x50 events */
>
> As with the L3C PMU, what exactly does this mean?
>
> Does this mean event IDs 0-0x4f are valid?
>
It is the same as L3C PMU and it is event IDs.
> [...]
>
>> +static irqreturn_t hisi_hha_pmu_isr(int irq, void *dev_id)
>> +{
>> + struct hisi_pmu *hha_pmu = dev_id;
>> + struct perf_event *event;
>> + unsigned long overflown;
>> + u32 status;
>> + int idx;
>> +
>> + /* Read HHA_INT_STATUS register */
>> + status = readl(hha_pmu->base + HHA_INT_STATUS);
>> + if (!status)
>> + return IRQ_NONE;
>> + overflown = status;
>
> No need for the u32 temporary here.
>
Ok.
> [....]
>
>> +static int hisi_hha_pmu_dev_probe(struct platform_device *pdev,
>> + struct hisi_pmu *hha_pmu)
>> +{
>> + struct device *dev = &pdev->dev;
>> + int ret;
>> +
>> + ret = hisi_hha_pmu_init_data(pdev, hha_pmu);
>> + if (ret)
>> + return ret;
>> +
>> + /* Pick one core to use for cpumask attributes */
>> + cpumask_set_cpu(smp_processor_id(), &hha_pmu->cpus);
>> +
>
> Why does this not have the usual event migration callbacks, across CPUs
> in the same SCCL?
>
Sorry we missed this. Shall add CPU hotplug callbacks and handle the cpumask
update and event migration.
>> + ret = hisi_hha_pmu_init_irq(hha_pmu, pdev);
>> + if (ret)
>> + return ret;
>> +
>> + hha_pmu->name = devm_kasprintf(dev, GFP_KERNEL, "hisi_hha%u_%u",
>> + hha_pmu->hha_uid, hha_pmu->sccl_id);
>
> As on the doc patch, this should be hierarchical.
>
Surely.
Thanks,
Shaokun
> Thanks,
> Mark
>
> _______________________________________________
> linux-arm-kernel mailing list
> linux-arm-kernel@lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
>
>
^ permalink raw reply [flat|nested] 19+ messages in thread
* [PATCH v4 5/6] perf: hisi: Add support for HiSilicon SoC DDRC PMU driver
2017-07-25 12:10 [PATCH v4 0/6] Add HiSilicon SoC uncore Performance Monitoring Unit driver Shaokun Zhang
` (3 preceding siblings ...)
2017-07-25 12:10 ` [PATCH v4 4/6] perf: hisi: Add support for HiSilicon SoC HHA " Shaokun Zhang
@ 2017-07-25 12:10 ` Shaokun Zhang
2017-08-15 13:02 ` Mark Rutland
2017-07-25 12:10 ` [PATCH v4 6/6] arm64: MAINTAINERS: hisi: Add HiSilicon SoC PMU support Shaokun Zhang
2017-08-07 8:57 ` [PATCH v4 0/6] Add HiSilicon SoC uncore Performance Monitoring Unit driver Zhangshaokun
6 siblings, 1 reply; 19+ messages in thread
From: Shaokun Zhang @ 2017-07-25 12:10 UTC (permalink / raw)
To: mark.rutland, will.deacon
Cc: jonathan.cameron, linux-arm-kernel, linux-kernel, linux-doc, linuxarm
This patch adds support for DDRC PMU driver in HiSilicon SoC chip, Each
DDRC has own control, counter and interrupt registers and is an separate
PMU. For each DDRC PMU, it has 8-fixed-purpose counters which have been
mapped to 8-events by hardware, it assumes that counter index is equal
to event code (0 - 7) in DDRC PMU driver. Interrupt is supported to
handle counter (32-bits) overflow.
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Signed-off-by: Shaokun Zhang <zhangshaokun@hisilicon.com>
Signed-off-by: Anurup M <anurup.m@huawei.com>
---
drivers/perf/hisilicon/Makefile | 2 +-
drivers/perf/hisilicon/hisi_uncore_ddrc_pmu.c | 420 ++++++++++++++++++++++++++
2 files changed, 421 insertions(+), 1 deletion(-)
create mode 100644 drivers/perf/hisilicon/hisi_uncore_ddrc_pmu.c
diff --git a/drivers/perf/hisilicon/Makefile b/drivers/perf/hisilicon/Makefile
index a72afe8..2621d51 100644
--- a/drivers/perf/hisilicon/Makefile
+++ b/drivers/perf/hisilicon/Makefile
@@ -1 +1 @@
-obj-$(CONFIG_HISI_PMU) += hisi_uncore_pmu.o hisi_uncore_l3c_pmu.o hisi_uncore_hha_pmu.o
+obj-$(CONFIG_HISI_PMU) += hisi_uncore_pmu.o hisi_uncore_l3c_pmu.o hisi_uncore_hha_pmu.o hisi_uncore_ddrc_pmu.o
diff --git a/drivers/perf/hisilicon/hisi_uncore_ddrc_pmu.c b/drivers/perf/hisilicon/hisi_uncore_ddrc_pmu.c
new file mode 100644
index 0000000..e178a09
--- /dev/null
+++ b/drivers/perf/hisilicon/hisi_uncore_ddrc_pmu.c
@@ -0,0 +1,420 @@
+/*
+ * HiSilicon SoC DDRC uncore Hardware event counters support
+ *
+ * Copyright (C) 2017 Hisilicon Limited
+ * Author: Shaokun Zhang <zhangshaokun@hisilicon.com>
+ * Anurup M <anurup.m@huawei.com>
+ *
+ * This code is based on the uncore PMUs like arm-cci and arm-ccn.
+ *
+ * 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.
+ */
+#include <linux/acpi.h>
+#include <linux/bug.h>
+#include <linux/interrupt.h>
+#include <linux/irq.h>
+#include <linux/list.h>
+#include <linux/platform_device.h>
+#include "hisi_uncore_pmu.h"
+
+/* DDRC register definition */
+#define DDRC_PERF_CTRL 0x010
+#define DDRC_FLUX_WR 0x380
+#define DDRC_FLUX_RD 0x384
+#define DDRC_FLUX_WCMD 0x388
+#define DDRC_FLUX_RCMD 0x38c
+#define DDRC_PRE_CMD 0x3c0
+#define DDRC_ACT_CMD 0x3c4
+#define DDRC_BNK_CHG 0x3c8
+#define DDRC_RNK_CHG 0x3cc
+#define DDRC_EVENT_CTRL 0x6C0
+#define DDRC_INT_MASK 0x6c8
+#define DDRC_INT_STATUS 0x6cc
+#define DDRC_INT_CLEAR 0x6d0
+
+/* DDRC supports 8-events and counter is fixed-purpose */
+#define DDRC_NR_COUNTERS 0x8
+#define DDRC_NR_EVENTS DDRC_NR_COUNTERS
+
+#define DDRC_PERF_CTRL_EN 0x2
+
+/*
+ * For DDRC PMU, there are eight-events and every event has been mapped
+ * to fixed-purpose counters which register offset is not consistent.
+ * Therefore there is no write event type and we assume that event
+ * code (0 to 7) is equal to counter index in PMU driver.
+ */
+#define GET_DDRC_EVENTID(hwc) (hwc->config_base & 0x7)
+
+static const u32 ddrc_reg_off[] = {
+ DDRC_FLUX_WR, DDRC_FLUX_RD, DDRC_FLUX_WCMD, DDRC_FLUX_RCMD,
+ DDRC_PRE_CMD, DDRC_ACT_CMD, DDRC_BNK_CHG, DDRC_RNK_CHG
+};
+
+/*
+ * Select the counter register offset using the counter index.
+ * In DDRC there are no programmable counter, the count
+ * is readed form the statistics counter register itself.
+ */
+static u32 get_counter_reg_off(int cntr_idx)
+{
+ return ddrc_reg_off[cntr_idx];
+}
+
+static u64 hisi_ddrc_pmu_read_counter(struct hisi_pmu *ddrc_pmu,
+ struct hw_perf_event *hwc)
+{
+ /* Use event code as counter index */
+ u32 idx = GET_DDRC_EVENTID(hwc);
+ u32 reg;
+
+ if (!hisi_uncore_pmu_counter_valid(ddrc_pmu, idx)) {
+ dev_err(ddrc_pmu->dev, "Unsupported event index:%d!\n", idx);
+ return 0;
+ }
+
+ reg = get_counter_reg_off(idx);
+
+ return readl(ddrc_pmu->base + reg);
+}
+
+static void hisi_ddrc_pmu_write_counter(struct hisi_pmu *ddrc_pmu,
+ struct hw_perf_event *hwc, u64 val)
+{
+ u32 idx = GET_DDRC_EVENTID(hwc);
+ u32 reg;
+
+ if (!hisi_uncore_pmu_counter_valid(ddrc_pmu, idx)) {
+ dev_err(ddrc_pmu->dev, "Unsupported event index:%d!\n", idx);
+ return;
+ }
+
+ reg = get_counter_reg_off(idx);
+ writel((u32)val, ddrc_pmu->base + reg);
+}
+
+static void hisi_ddrc_pmu_start_counters(struct hisi_pmu *ddrc_pmu)
+{
+ u32 val;
+
+ /* Set perf_enable in DDRC_PERF_CTRL to start event counting */
+ val = readl(ddrc_pmu->base + DDRC_PERF_CTRL);
+ val |= DDRC_PERF_CTRL_EN;
+ writel(val, ddrc_pmu->base + DDRC_PERF_CTRL);
+}
+
+static void hisi_ddrc_pmu_stop_counters(struct hisi_pmu *ddrc_pmu)
+{
+ u32 val;
+
+ /* Clear perf_enable in DDRC_PERF_CTRL to stop event counting */
+ val = readl(ddrc_pmu->base + DDRC_PERF_CTRL);
+ val &= ~DDRC_PERF_CTRL_EN;
+ writel(val, ddrc_pmu->base + DDRC_PERF_CTRL);
+}
+
+static void hisi_ddrc_pmu_enable_counter(struct hisi_pmu *ddrc_pmu,
+ struct hw_perf_event *hwc)
+{
+ u32 val;
+
+ /* Set counter index(event code) in DDRC_EVENT_CTRL register */
+ val = readl(ddrc_pmu->base + DDRC_EVENT_CTRL);
+ val |= (1 << GET_DDRC_EVENTID(hwc));
+ writel(val, ddrc_pmu->base + DDRC_EVENT_CTRL);
+}
+
+static void hisi_ddrc_pmu_disable_counter(struct hisi_pmu *ddrc_pmu,
+ struct hw_perf_event *hwc)
+{
+ u32 val;
+
+ /* Clear counter index(event code) in DDRC_EVENT_CTRL register */
+ val = readl(ddrc_pmu->base + DDRC_EVENT_CTRL);
+ val &= ~(1 << GET_DDRC_EVENTID(hwc));
+ writel(val, ddrc_pmu->base + DDRC_EVENT_CTRL);
+}
+
+static int hisi_ddrc_pmu_get_event_idx(struct perf_event *event)
+{
+ struct hisi_pmu *ddrc_pmu = to_hisi_pmu(event->pmu);
+ unsigned long *used_mask = ddrc_pmu->pmu_events.used_mask;
+ struct hw_perf_event *hwc = &event->hw;
+ /* For DDRC PMU, we use event code as counter index */
+ int idx = GET_DDRC_EVENTID(hwc);
+
+ if (test_bit(idx, used_mask))
+ return -EAGAIN;
+
+ set_bit(idx, used_mask);
+
+ return idx;
+}
+
+static void hisi_ddrc_pmu_enable_counter_int(struct hisi_pmu *ddrc_pmu,
+ struct hw_perf_event *hwc)
+{
+ u32 val;
+
+ /* Write 0 to enable interrupt */
+ val = readl(ddrc_pmu->base + DDRC_INT_MASK);
+ val &= ~(1 << GET_DDRC_EVENTID(hwc));
+ writel(val, ddrc_pmu->base + DDRC_INT_MASK);
+}
+
+static void hisi_ddrc_pmu_disable_counter_int(struct hisi_pmu *ddrc_pmu,
+ struct hw_perf_event *hwc)
+{
+ u32 val;
+
+ /* Write 1 to mask interrupt */
+ val = readl(ddrc_pmu->base + DDRC_INT_MASK);
+ val |= (1 << GET_DDRC_EVENTID(hwc));
+ writel(val, ddrc_pmu->base + DDRC_INT_MASK);
+}
+
+static irqreturn_t hisi_ddrc_pmu_isr(int irq, void *dev_id)
+{
+ struct hisi_pmu *ddrc_pmu = dev_id;
+ struct perf_event *event;
+ unsigned long overflown;
+ u32 status;
+ int idx;
+
+ /* Read the DDRC_INT_STATUS register */
+ status = readl(ddrc_pmu->base + DDRC_INT_STATUS);
+ if (!status)
+ return IRQ_NONE;
+ overflown = status;
+
+ /*
+ * Find the counter index which overflowed if the bit was set
+ * and handle it
+ */
+ for_each_set_bit(idx, &overflown, DDRC_NR_COUNTERS) {
+ /* Write 1 to clear the IRQ status flag */
+ writel((1 << idx), ddrc_pmu->base + DDRC_INT_CLEAR);
+
+ /* Get the corresponding event struct */
+ event = ddrc_pmu->pmu_events.hw_events[idx];
+ if (!event)
+ continue;
+
+ hisi_uncore_pmu_event_update(event);
+ hisi_uncore_pmu_set_event_period(event);
+ }
+
+ return IRQ_HANDLED;
+}
+
+static int hisi_ddrc_pmu_init_irq(struct hisi_pmu *ddrc_pmu,
+ struct platform_device *pdev)
+{
+ struct device *dev = &pdev->dev;
+ int irq, ret;
+
+ /* Read and init IRQ */
+ irq = platform_get_irq(pdev, 0);
+ if (irq < 0) {
+ dev_err(dev, "irq init: fail map DDRC overflow interrupt\n");
+ return irq;
+ }
+
+ ret = devm_request_irq(dev, irq, hisi_ddrc_pmu_isr,
+ IRQF_NOBALANCING | IRQF_NO_THREAD,
+ dev_name(dev), ddrc_pmu);
+ if (ret < 0) {
+ dev_err(dev, "Fail to request IRQ:%d ret:%d\n", irq, ret);
+ return ret;
+ }
+
+ /* Overflow interrupt also should use the same CPU */
+ WARN_ON(irq_set_affinity(irq, &ddrc_pmu->cpus));
+
+ return 0;
+}
+
+static const struct acpi_device_id hisi_ddrc_pmu_acpi_match[] = {
+ { "HISI0233", },
+ {},
+};
+MODULE_DEVICE_TABLE(acpi, hisi_ddrc_pmu_acpi_match);
+
+static int hisi_ddrc_pmu_init_data(struct platform_device *pdev,
+ struct hisi_pmu *ddrc_pmu)
+{
+ struct device *dev = &pdev->dev;
+ struct resource *res;
+
+ /* Get the DDRC Channel ID */
+ if (device_property_read_u32(dev, "hisilicon,ch-id",
+ &ddrc_pmu->ddrc_chn_id)) {
+ dev_err(dev, "Can not read ddrc ch-id!\n");
+ return -EINVAL;
+ }
+
+ /* Get the DDRC SCCL ID */
+ if (device_property_read_u32(dev, "hisilicon,scl-id",
+ &ddrc_pmu->sccl_id)) {
+ dev_err(dev, "Can not read ddrc sccl-id!\n");
+ return -EINVAL;
+ }
+
+ res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+ ddrc_pmu->base = devm_ioremap_resource(dev, res);
+ if (IS_ERR(ddrc_pmu->base)) {
+ dev_err(dev, "ioremap failed for ddrc_pmu resource\n");
+ return PTR_ERR(ddrc_pmu->base);
+ }
+
+ return 0;
+}
+
+static struct attribute *hisi_ddrc_pmu_format_attr[] = {
+ HISI_PMU_FORMAT_ATTR(event, "config:0-4"),
+ NULL,
+};
+
+static const struct attribute_group hisi_ddrc_pmu_format_group = {
+ .name = "format",
+ .attrs = hisi_ddrc_pmu_format_attr,
+};
+
+static struct attribute *hisi_ddrc_pmu_events_attr[] = {
+ HISI_PMU_EVENT_ATTR(flux_wr, 0x00),
+ HISI_PMU_EVENT_ATTR(flux_rd, 0x01),
+ HISI_PMU_EVENT_ATTR(flux_wcmd, 0x02),
+ HISI_PMU_EVENT_ATTR(flux_rcmd, 0x03),
+ HISI_PMU_EVENT_ATTR(pre_cmd, 0x04),
+ HISI_PMU_EVENT_ATTR(act_cmd, 0x05),
+ HISI_PMU_EVENT_ATTR(rnk_chg, 0x06),
+ HISI_PMU_EVENT_ATTR(rw_chg, 0x07),
+ NULL,
+};
+
+static const struct attribute_group hisi_ddrc_pmu_events_group = {
+ .name = "events",
+ .attrs = hisi_ddrc_pmu_events_attr,
+};
+
+static DEVICE_ATTR(cpumask, 0444, hisi_cpumask_sysfs_show, NULL);
+
+static struct attribute *hisi_ddrc_pmu_cpumask_attrs[] = {
+ &dev_attr_cpumask.attr,
+ NULL,
+};
+
+static const struct attribute_group hisi_ddrc_pmu_cpumask_attr_group = {
+ .attrs = hisi_ddrc_pmu_cpumask_attrs,
+};
+
+static const struct attribute_group *hisi_ddrc_pmu_attr_groups[] = {
+ &hisi_ddrc_pmu_format_group,
+ &hisi_ddrc_pmu_events_group,
+ &hisi_ddrc_pmu_cpumask_attr_group,
+ NULL,
+};
+
+static const struct hisi_uncore_ops hisi_uncore_ddrc_ops = {
+ .get_event_idx = hisi_ddrc_pmu_get_event_idx,
+ .start_counters = hisi_ddrc_pmu_start_counters,
+ .stop_counters = hisi_ddrc_pmu_stop_counters,
+ .enable_counter = hisi_ddrc_pmu_enable_counter,
+ .disable_counter = hisi_ddrc_pmu_disable_counter,
+ .enable_counter_int = hisi_ddrc_pmu_enable_counter_int,
+ .disable_counter_int = hisi_ddrc_pmu_disable_counter_int,
+ .write_counter = hisi_ddrc_pmu_write_counter,
+ .read_counter = hisi_ddrc_pmu_read_counter,
+};
+
+static int hisi_ddrc_pmu_dev_probe(struct platform_device *pdev,
+ struct hisi_pmu *ddrc_pmu)
+{
+ struct device *dev = &pdev->dev;
+ int ret;
+
+ ret = hisi_ddrc_pmu_init_data(pdev, ddrc_pmu);
+ if (ret)
+ return ret;
+
+ /* Pick one core to use for cpumask attributes */
+ cpumask_set_cpu(smp_processor_id(), &ddrc_pmu->cpus);
+
+ ret = hisi_ddrc_pmu_init_irq(ddrc_pmu, pdev);
+ if (ret)
+ return ret;
+
+ ddrc_pmu->name = devm_kasprintf(dev, GFP_KERNEL, "hisi_ddrc%u_%u",
+ ddrc_pmu->ddrc_chn_id,
+ ddrc_pmu->sccl_id);
+ ddrc_pmu->num_events = DDRC_NR_EVENTS;
+ ddrc_pmu->num_counters = DDRC_NR_COUNTERS;
+ ddrc_pmu->counter_bits = 32;
+ ddrc_pmu->ops = &hisi_uncore_ddrc_ops;
+ ddrc_pmu->dev = dev;
+
+ return 0;
+}
+
+static int hisi_ddrc_pmu_probe(struct platform_device *pdev)
+{
+ struct device *dev = &pdev->dev;
+ struct hisi_pmu *ddrc_pmu;
+ int ret;
+
+ ddrc_pmu = hisi_pmu_alloc(dev, DDRC_NR_COUNTERS);
+ if (!ddrc_pmu)
+ return -ENOMEM;
+
+ platform_set_drvdata(pdev, ddrc_pmu);
+
+ ret = hisi_ddrc_pmu_dev_probe(pdev, ddrc_pmu);
+ if (ret)
+ return ret;
+
+ ddrc_pmu->pmu = (struct pmu) {
+ .name = ddrc_pmu->name,
+ .task_ctx_nr = perf_invalid_context,
+ .event_init = hisi_uncore_pmu_event_init,
+ .pmu_enable = hisi_uncore_pmu_enable,
+ .pmu_disable = hisi_uncore_pmu_disable,
+ .add = hisi_uncore_pmu_add,
+ .del = hisi_uncore_pmu_del,
+ .start = hisi_uncore_pmu_start,
+ .stop = hisi_uncore_pmu_stop,
+ .read = hisi_uncore_pmu_read,
+ .attr_groups = hisi_ddrc_pmu_attr_groups,
+ };
+
+ ret = perf_pmu_register(&ddrc_pmu->pmu, ddrc_pmu->name, -1);
+ if (ret)
+ dev_err(ddrc_pmu->dev, "ddrc_pmu register failed!\n");
+
+ return ret;
+}
+
+static int hisi_ddrc_pmu_remove(struct platform_device *pdev)
+{
+ struct hisi_pmu *ddrc_pmu = platform_get_drvdata(pdev);
+
+ perf_pmu_unregister(&ddrc_pmu->pmu);
+
+ return 0;
+}
+
+static struct platform_driver hisi_ddrc_pmu_driver = {
+ .driver = {
+ .name = "hisi_ddrc_pmu",
+ .acpi_match_table = ACPI_PTR(hisi_ddrc_pmu_acpi_match),
+ },
+ .probe = hisi_ddrc_pmu_probe,
+ .remove = hisi_ddrc_pmu_remove,
+};
+module_platform_driver(hisi_ddrc_pmu_driver);
+
+MODULE_DESCRIPTION("HiSilicon SoC DDRC uncore PMU driver");
+MODULE_LICENSE("GPL v2");
+MODULE_AUTHOR("Shaokun Zhang <zhangshaokun@hisilicon.com>");
+MODULE_AUTHOR("Anurup M <anurup.m@huawei.com>");
--
1.9.1
^ permalink raw reply related [flat|nested] 19+ messages in thread* Re: [PATCH v4 5/6] perf: hisi: Add support for HiSilicon SoC DDRC PMU driver
2017-07-25 12:10 ` [PATCH v4 5/6] perf: hisi: Add support for HiSilicon SoC DDRC " Shaokun Zhang
@ 2017-08-15 13:02 ` Mark Rutland
2017-08-17 3:40 ` Zhangshaokun
0 siblings, 1 reply; 19+ messages in thread
From: Mark Rutland @ 2017-08-15 13:02 UTC (permalink / raw)
To: Shaokun Zhang
Cc: will.deacon, jonathan.cameron, linux-arm-kernel, linux-kernel,
linux-doc, linuxarm
On Tue, Jul 25, 2017 at 08:10:41PM +0800, Shaokun Zhang wrote:
> This patch adds support for DDRC PMU driver in HiSilicon SoC chip, Each
> DDRC has own control, counter and interrupt registers and is an separate
> PMU. For each DDRC PMU, it has 8-fixed-purpose counters which have been
> mapped to 8-events by hardware, it assumes that counter index is equal
> to event code (0 - 7) in DDRC PMU driver. Interrupt is supported to
> handle counter (32-bits) overflow.
>
> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
> Signed-off-by: Shaokun Zhang <zhangshaokun@hisilicon.com>
> Signed-off-by: Anurup M <anurup.m@huawei.com>
> ---
> drivers/perf/hisilicon/Makefile | 2 +-
> drivers/perf/hisilicon/hisi_uncore_ddrc_pmu.c | 420 ++++++++++++++++++++++++++
> 2 files changed, 421 insertions(+), 1 deletion(-)
> create mode 100644 drivers/perf/hisilicon/hisi_uncore_ddrc_pmu.c
My comments for the HHA PMU driver all apply here, too.
Thanks,
Mark.
> diff --git a/drivers/perf/hisilicon/Makefile b/drivers/perf/hisilicon/Makefile
> index a72afe8..2621d51 100644
> --- a/drivers/perf/hisilicon/Makefile
> +++ b/drivers/perf/hisilicon/Makefile
> @@ -1 +1 @@
> -obj-$(CONFIG_HISI_PMU) += hisi_uncore_pmu.o hisi_uncore_l3c_pmu.o hisi_uncore_hha_pmu.o
> +obj-$(CONFIG_HISI_PMU) += hisi_uncore_pmu.o hisi_uncore_l3c_pmu.o hisi_uncore_hha_pmu.o hisi_uncore_ddrc_pmu.o
> diff --git a/drivers/perf/hisilicon/hisi_uncore_ddrc_pmu.c b/drivers/perf/hisilicon/hisi_uncore_ddrc_pmu.c
> new file mode 100644
> index 0000000..e178a09
> --- /dev/null
> +++ b/drivers/perf/hisilicon/hisi_uncore_ddrc_pmu.c
> @@ -0,0 +1,420 @@
> +/*
> + * HiSilicon SoC DDRC uncore Hardware event counters support
> + *
> + * Copyright (C) 2017 Hisilicon Limited
> + * Author: Shaokun Zhang <zhangshaokun@hisilicon.com>
> + * Anurup M <anurup.m@huawei.com>
> + *
> + * This code is based on the uncore PMUs like arm-cci and arm-ccn.
> + *
> + * 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.
> + */
> +#include <linux/acpi.h>
> +#include <linux/bug.h>
> +#include <linux/interrupt.h>
> +#include <linux/irq.h>
> +#include <linux/list.h>
> +#include <linux/platform_device.h>
> +#include "hisi_uncore_pmu.h"
> +
> +/* DDRC register definition */
> +#define DDRC_PERF_CTRL 0x010
> +#define DDRC_FLUX_WR 0x380
> +#define DDRC_FLUX_RD 0x384
> +#define DDRC_FLUX_WCMD 0x388
> +#define DDRC_FLUX_RCMD 0x38c
> +#define DDRC_PRE_CMD 0x3c0
> +#define DDRC_ACT_CMD 0x3c4
> +#define DDRC_BNK_CHG 0x3c8
> +#define DDRC_RNK_CHG 0x3cc
> +#define DDRC_EVENT_CTRL 0x6C0
> +#define DDRC_INT_MASK 0x6c8
> +#define DDRC_INT_STATUS 0x6cc
> +#define DDRC_INT_CLEAR 0x6d0
> +
> +/* DDRC supports 8-events and counter is fixed-purpose */
> +#define DDRC_NR_COUNTERS 0x8
> +#define DDRC_NR_EVENTS DDRC_NR_COUNTERS
> +
> +#define DDRC_PERF_CTRL_EN 0x2
> +
> +/*
> + * For DDRC PMU, there are eight-events and every event has been mapped
> + * to fixed-purpose counters which register offset is not consistent.
> + * Therefore there is no write event type and we assume that event
> + * code (0 to 7) is equal to counter index in PMU driver.
> + */
> +#define GET_DDRC_EVENTID(hwc) (hwc->config_base & 0x7)
> +
> +static const u32 ddrc_reg_off[] = {
> + DDRC_FLUX_WR, DDRC_FLUX_RD, DDRC_FLUX_WCMD, DDRC_FLUX_RCMD,
> + DDRC_PRE_CMD, DDRC_ACT_CMD, DDRC_BNK_CHG, DDRC_RNK_CHG
> +};
> +
> +/*
> + * Select the counter register offset using the counter index.
> + * In DDRC there are no programmable counter, the count
> + * is readed form the statistics counter register itself.
> + */
> +static u32 get_counter_reg_off(int cntr_idx)
> +{
> + return ddrc_reg_off[cntr_idx];
> +}
> +
> +static u64 hisi_ddrc_pmu_read_counter(struct hisi_pmu *ddrc_pmu,
> + struct hw_perf_event *hwc)
> +{
> + /* Use event code as counter index */
> + u32 idx = GET_DDRC_EVENTID(hwc);
> + u32 reg;
> +
> + if (!hisi_uncore_pmu_counter_valid(ddrc_pmu, idx)) {
> + dev_err(ddrc_pmu->dev, "Unsupported event index:%d!\n", idx);
> + return 0;
> + }
> +
> + reg = get_counter_reg_off(idx);
> +
> + return readl(ddrc_pmu->base + reg);
> +}
> +
> +static void hisi_ddrc_pmu_write_counter(struct hisi_pmu *ddrc_pmu,
> + struct hw_perf_event *hwc, u64 val)
> +{
> + u32 idx = GET_DDRC_EVENTID(hwc);
> + u32 reg;
> +
> + if (!hisi_uncore_pmu_counter_valid(ddrc_pmu, idx)) {
> + dev_err(ddrc_pmu->dev, "Unsupported event index:%d!\n", idx);
> + return;
> + }
> +
> + reg = get_counter_reg_off(idx);
> + writel((u32)val, ddrc_pmu->base + reg);
> +}
> +
> +static void hisi_ddrc_pmu_start_counters(struct hisi_pmu *ddrc_pmu)
> +{
> + u32 val;
> +
> + /* Set perf_enable in DDRC_PERF_CTRL to start event counting */
> + val = readl(ddrc_pmu->base + DDRC_PERF_CTRL);
> + val |= DDRC_PERF_CTRL_EN;
> + writel(val, ddrc_pmu->base + DDRC_PERF_CTRL);
> +}
> +
> +static void hisi_ddrc_pmu_stop_counters(struct hisi_pmu *ddrc_pmu)
> +{
> + u32 val;
> +
> + /* Clear perf_enable in DDRC_PERF_CTRL to stop event counting */
> + val = readl(ddrc_pmu->base + DDRC_PERF_CTRL);
> + val &= ~DDRC_PERF_CTRL_EN;
> + writel(val, ddrc_pmu->base + DDRC_PERF_CTRL);
> +}
> +
> +static void hisi_ddrc_pmu_enable_counter(struct hisi_pmu *ddrc_pmu,
> + struct hw_perf_event *hwc)
> +{
> + u32 val;
> +
> + /* Set counter index(event code) in DDRC_EVENT_CTRL register */
> + val = readl(ddrc_pmu->base + DDRC_EVENT_CTRL);
> + val |= (1 << GET_DDRC_EVENTID(hwc));
> + writel(val, ddrc_pmu->base + DDRC_EVENT_CTRL);
> +}
> +
> +static void hisi_ddrc_pmu_disable_counter(struct hisi_pmu *ddrc_pmu,
> + struct hw_perf_event *hwc)
> +{
> + u32 val;
> +
> + /* Clear counter index(event code) in DDRC_EVENT_CTRL register */
> + val = readl(ddrc_pmu->base + DDRC_EVENT_CTRL);
> + val &= ~(1 << GET_DDRC_EVENTID(hwc));
> + writel(val, ddrc_pmu->base + DDRC_EVENT_CTRL);
> +}
> +
> +static int hisi_ddrc_pmu_get_event_idx(struct perf_event *event)
> +{
> + struct hisi_pmu *ddrc_pmu = to_hisi_pmu(event->pmu);
> + unsigned long *used_mask = ddrc_pmu->pmu_events.used_mask;
> + struct hw_perf_event *hwc = &event->hw;
> + /* For DDRC PMU, we use event code as counter index */
> + int idx = GET_DDRC_EVENTID(hwc);
> +
> + if (test_bit(idx, used_mask))
> + return -EAGAIN;
> +
> + set_bit(idx, used_mask);
> +
> + return idx;
> +}
> +
> +static void hisi_ddrc_pmu_enable_counter_int(struct hisi_pmu *ddrc_pmu,
> + struct hw_perf_event *hwc)
> +{
> + u32 val;
> +
> + /* Write 0 to enable interrupt */
> + val = readl(ddrc_pmu->base + DDRC_INT_MASK);
> + val &= ~(1 << GET_DDRC_EVENTID(hwc));
> + writel(val, ddrc_pmu->base + DDRC_INT_MASK);
> +}
> +
> +static void hisi_ddrc_pmu_disable_counter_int(struct hisi_pmu *ddrc_pmu,
> + struct hw_perf_event *hwc)
> +{
> + u32 val;
> +
> + /* Write 1 to mask interrupt */
> + val = readl(ddrc_pmu->base + DDRC_INT_MASK);
> + val |= (1 << GET_DDRC_EVENTID(hwc));
> + writel(val, ddrc_pmu->base + DDRC_INT_MASK);
> +}
> +
> +static irqreturn_t hisi_ddrc_pmu_isr(int irq, void *dev_id)
> +{
> + struct hisi_pmu *ddrc_pmu = dev_id;
> + struct perf_event *event;
> + unsigned long overflown;
> + u32 status;
> + int idx;
> +
> + /* Read the DDRC_INT_STATUS register */
> + status = readl(ddrc_pmu->base + DDRC_INT_STATUS);
> + if (!status)
> + return IRQ_NONE;
> + overflown = status;
> +
> + /*
> + * Find the counter index which overflowed if the bit was set
> + * and handle it
> + */
> + for_each_set_bit(idx, &overflown, DDRC_NR_COUNTERS) {
> + /* Write 1 to clear the IRQ status flag */
> + writel((1 << idx), ddrc_pmu->base + DDRC_INT_CLEAR);
> +
> + /* Get the corresponding event struct */
> + event = ddrc_pmu->pmu_events.hw_events[idx];
> + if (!event)
> + continue;
> +
> + hisi_uncore_pmu_event_update(event);
> + hisi_uncore_pmu_set_event_period(event);
> + }
> +
> + return IRQ_HANDLED;
> +}
> +
> +static int hisi_ddrc_pmu_init_irq(struct hisi_pmu *ddrc_pmu,
> + struct platform_device *pdev)
> +{
> + struct device *dev = &pdev->dev;
> + int irq, ret;
> +
> + /* Read and init IRQ */
> + irq = platform_get_irq(pdev, 0);
> + if (irq < 0) {
> + dev_err(dev, "irq init: fail map DDRC overflow interrupt\n");
> + return irq;
> + }
> +
> + ret = devm_request_irq(dev, irq, hisi_ddrc_pmu_isr,
> + IRQF_NOBALANCING | IRQF_NO_THREAD,
> + dev_name(dev), ddrc_pmu);
> + if (ret < 0) {
> + dev_err(dev, "Fail to request IRQ:%d ret:%d\n", irq, ret);
> + return ret;
> + }
> +
> + /* Overflow interrupt also should use the same CPU */
> + WARN_ON(irq_set_affinity(irq, &ddrc_pmu->cpus));
> +
> + return 0;
> +}
> +
> +static const struct acpi_device_id hisi_ddrc_pmu_acpi_match[] = {
> + { "HISI0233", },
> + {},
> +};
> +MODULE_DEVICE_TABLE(acpi, hisi_ddrc_pmu_acpi_match);
> +
> +static int hisi_ddrc_pmu_init_data(struct platform_device *pdev,
> + struct hisi_pmu *ddrc_pmu)
> +{
> + struct device *dev = &pdev->dev;
> + struct resource *res;
> +
> + /* Get the DDRC Channel ID */
> + if (device_property_read_u32(dev, "hisilicon,ch-id",
> + &ddrc_pmu->ddrc_chn_id)) {
> + dev_err(dev, "Can not read ddrc ch-id!\n");
> + return -EINVAL;
> + }
> +
> + /* Get the DDRC SCCL ID */
> + if (device_property_read_u32(dev, "hisilicon,scl-id",
> + &ddrc_pmu->sccl_id)) {
> + dev_err(dev, "Can not read ddrc sccl-id!\n");
> + return -EINVAL;
> + }
> +
> + res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
> + ddrc_pmu->base = devm_ioremap_resource(dev, res);
> + if (IS_ERR(ddrc_pmu->base)) {
> + dev_err(dev, "ioremap failed for ddrc_pmu resource\n");
> + return PTR_ERR(ddrc_pmu->base);
> + }
> +
> + return 0;
> +}
> +
> +static struct attribute *hisi_ddrc_pmu_format_attr[] = {
> + HISI_PMU_FORMAT_ATTR(event, "config:0-4"),
> + NULL,
> +};
> +
> +static const struct attribute_group hisi_ddrc_pmu_format_group = {
> + .name = "format",
> + .attrs = hisi_ddrc_pmu_format_attr,
> +};
> +
> +static struct attribute *hisi_ddrc_pmu_events_attr[] = {
> + HISI_PMU_EVENT_ATTR(flux_wr, 0x00),
> + HISI_PMU_EVENT_ATTR(flux_rd, 0x01),
> + HISI_PMU_EVENT_ATTR(flux_wcmd, 0x02),
> + HISI_PMU_EVENT_ATTR(flux_rcmd, 0x03),
> + HISI_PMU_EVENT_ATTR(pre_cmd, 0x04),
> + HISI_PMU_EVENT_ATTR(act_cmd, 0x05),
> + HISI_PMU_EVENT_ATTR(rnk_chg, 0x06),
> + HISI_PMU_EVENT_ATTR(rw_chg, 0x07),
> + NULL,
> +};
> +
> +static const struct attribute_group hisi_ddrc_pmu_events_group = {
> + .name = "events",
> + .attrs = hisi_ddrc_pmu_events_attr,
> +};
> +
> +static DEVICE_ATTR(cpumask, 0444, hisi_cpumask_sysfs_show, NULL);
> +
> +static struct attribute *hisi_ddrc_pmu_cpumask_attrs[] = {
> + &dev_attr_cpumask.attr,
> + NULL,
> +};
> +
> +static const struct attribute_group hisi_ddrc_pmu_cpumask_attr_group = {
> + .attrs = hisi_ddrc_pmu_cpumask_attrs,
> +};
> +
> +static const struct attribute_group *hisi_ddrc_pmu_attr_groups[] = {
> + &hisi_ddrc_pmu_format_group,
> + &hisi_ddrc_pmu_events_group,
> + &hisi_ddrc_pmu_cpumask_attr_group,
> + NULL,
> +};
> +
> +static const struct hisi_uncore_ops hisi_uncore_ddrc_ops = {
> + .get_event_idx = hisi_ddrc_pmu_get_event_idx,
> + .start_counters = hisi_ddrc_pmu_start_counters,
> + .stop_counters = hisi_ddrc_pmu_stop_counters,
> + .enable_counter = hisi_ddrc_pmu_enable_counter,
> + .disable_counter = hisi_ddrc_pmu_disable_counter,
> + .enable_counter_int = hisi_ddrc_pmu_enable_counter_int,
> + .disable_counter_int = hisi_ddrc_pmu_disable_counter_int,
> + .write_counter = hisi_ddrc_pmu_write_counter,
> + .read_counter = hisi_ddrc_pmu_read_counter,
> +};
> +
> +static int hisi_ddrc_pmu_dev_probe(struct platform_device *pdev,
> + struct hisi_pmu *ddrc_pmu)
> +{
> + struct device *dev = &pdev->dev;
> + int ret;
> +
> + ret = hisi_ddrc_pmu_init_data(pdev, ddrc_pmu);
> + if (ret)
> + return ret;
> +
> + /* Pick one core to use for cpumask attributes */
> + cpumask_set_cpu(smp_processor_id(), &ddrc_pmu->cpus);
> +
> + ret = hisi_ddrc_pmu_init_irq(ddrc_pmu, pdev);
> + if (ret)
> + return ret;
> +
> + ddrc_pmu->name = devm_kasprintf(dev, GFP_KERNEL, "hisi_ddrc%u_%u",
> + ddrc_pmu->ddrc_chn_id,
> + ddrc_pmu->sccl_id);
> + ddrc_pmu->num_events = DDRC_NR_EVENTS;
> + ddrc_pmu->num_counters = DDRC_NR_COUNTERS;
> + ddrc_pmu->counter_bits = 32;
> + ddrc_pmu->ops = &hisi_uncore_ddrc_ops;
> + ddrc_pmu->dev = dev;
> +
> + return 0;
> +}
> +
> +static int hisi_ddrc_pmu_probe(struct platform_device *pdev)
> +{
> + struct device *dev = &pdev->dev;
> + struct hisi_pmu *ddrc_pmu;
> + int ret;
> +
> + ddrc_pmu = hisi_pmu_alloc(dev, DDRC_NR_COUNTERS);
> + if (!ddrc_pmu)
> + return -ENOMEM;
> +
> + platform_set_drvdata(pdev, ddrc_pmu);
> +
> + ret = hisi_ddrc_pmu_dev_probe(pdev, ddrc_pmu);
> + if (ret)
> + return ret;
> +
> + ddrc_pmu->pmu = (struct pmu) {
> + .name = ddrc_pmu->name,
> + .task_ctx_nr = perf_invalid_context,
> + .event_init = hisi_uncore_pmu_event_init,
> + .pmu_enable = hisi_uncore_pmu_enable,
> + .pmu_disable = hisi_uncore_pmu_disable,
> + .add = hisi_uncore_pmu_add,
> + .del = hisi_uncore_pmu_del,
> + .start = hisi_uncore_pmu_start,
> + .stop = hisi_uncore_pmu_stop,
> + .read = hisi_uncore_pmu_read,
> + .attr_groups = hisi_ddrc_pmu_attr_groups,
> + };
> +
> + ret = perf_pmu_register(&ddrc_pmu->pmu, ddrc_pmu->name, -1);
> + if (ret)
> + dev_err(ddrc_pmu->dev, "ddrc_pmu register failed!\n");
> +
> + return ret;
> +}
> +
> +static int hisi_ddrc_pmu_remove(struct platform_device *pdev)
> +{
> + struct hisi_pmu *ddrc_pmu = platform_get_drvdata(pdev);
> +
> + perf_pmu_unregister(&ddrc_pmu->pmu);
> +
> + return 0;
> +}
> +
> +static struct platform_driver hisi_ddrc_pmu_driver = {
> + .driver = {
> + .name = "hisi_ddrc_pmu",
> + .acpi_match_table = ACPI_PTR(hisi_ddrc_pmu_acpi_match),
> + },
> + .probe = hisi_ddrc_pmu_probe,
> + .remove = hisi_ddrc_pmu_remove,
> +};
> +module_platform_driver(hisi_ddrc_pmu_driver);
> +
> +MODULE_DESCRIPTION("HiSilicon SoC DDRC uncore PMU driver");
> +MODULE_LICENSE("GPL v2");
> +MODULE_AUTHOR("Shaokun Zhang <zhangshaokun@hisilicon.com>");
> +MODULE_AUTHOR("Anurup M <anurup.m@huawei.com>");
> --
> 1.9.1
>
^ permalink raw reply [flat|nested] 19+ messages in thread* Re: [PATCH v4 5/6] perf: hisi: Add support for HiSilicon SoC DDRC PMU driver
2017-08-15 13:02 ` Mark Rutland
@ 2017-08-17 3:40 ` Zhangshaokun
0 siblings, 0 replies; 19+ messages in thread
From: Zhangshaokun @ 2017-08-17 3:40 UTC (permalink / raw)
To: Mark Rutland
Cc: linux-doc, will.deacon, linuxarm, linux-kernel, jonathan.cameron,
linux-arm-kernel
Hi Mark,
On 2017/8/15 21:02, Mark Rutland wrote:
> On Tue, Jul 25, 2017 at 08:10:41PM +0800, Shaokun Zhang wrote:
>> This patch adds support for DDRC PMU driver in HiSilicon SoC chip, Each
>> DDRC has own control, counter and interrupt registers and is an separate
>> PMU. For each DDRC PMU, it has 8-fixed-purpose counters which have been
>> mapped to 8-events by hardware, it assumes that counter index is equal
>> to event code (0 - 7) in DDRC PMU driver. Interrupt is supported to
>> handle counter (32-bits) overflow.
>>
>> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
>> Signed-off-by: Shaokun Zhang <zhangshaokun@hisilicon.com>
>> Signed-off-by: Anurup M <anurup.m@huawei.com>
>> ---
>> drivers/perf/hisilicon/Makefile | 2 +-
>> drivers/perf/hisilicon/hisi_uncore_ddrc_pmu.c | 420 ++++++++++++++++++++++++++
>> 2 files changed, 421 insertions(+), 1 deletion(-)
>> create mode 100644 drivers/perf/hisilicon/hisi_uncore_ddrc_pmu.c
>
> My comments for the HHA PMU driver all apply here, too.
>
Surely.
Thanks.
Shaokun
> Thanks,
> Mark.
>
>> diff --git a/drivers/perf/hisilicon/Makefile b/drivers/perf/hisilicon/Makefile
>> index a72afe8..2621d51 100644
>> --- a/drivers/perf/hisilicon/Makefile
>> +++ b/drivers/perf/hisilicon/Makefile
>> @@ -1 +1 @@
>> -obj-$(CONFIG_HISI_PMU) += hisi_uncore_pmu.o hisi_uncore_l3c_pmu.o hisi_uncore_hha_pmu.o
>> +obj-$(CONFIG_HISI_PMU) += hisi_uncore_pmu.o hisi_uncore_l3c_pmu.o hisi_uncore_hha_pmu.o hisi_uncore_ddrc_pmu.o
>> diff --git a/drivers/perf/hisilicon/hisi_uncore_ddrc_pmu.c b/drivers/perf/hisilicon/hisi_uncore_ddrc_pmu.c
>> new file mode 100644
>> index 0000000..e178a09
>> --- /dev/null
>> +++ b/drivers/perf/hisilicon/hisi_uncore_ddrc_pmu.c
>> @@ -0,0 +1,420 @@
>> +/*
>> + * HiSilicon SoC DDRC uncore Hardware event counters support
>> + *
>> + * Copyright (C) 2017 Hisilicon Limited
>> + * Author: Shaokun Zhang <zhangshaokun@hisilicon.com>
>> + * Anurup M <anurup.m@huawei.com>
>> + *
>> + * This code is based on the uncore PMUs like arm-cci and arm-ccn.
>> + *
>> + * 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.
>> + */
>> +#include <linux/acpi.h>
>> +#include <linux/bug.h>
>> +#include <linux/interrupt.h>
>> +#include <linux/irq.h>
>> +#include <linux/list.h>
>> +#include <linux/platform_device.h>
>> +#include "hisi_uncore_pmu.h"
>> +
>> +/* DDRC register definition */
>> +#define DDRC_PERF_CTRL 0x010
>> +#define DDRC_FLUX_WR 0x380
>> +#define DDRC_FLUX_RD 0x384
>> +#define DDRC_FLUX_WCMD 0x388
>> +#define DDRC_FLUX_RCMD 0x38c
>> +#define DDRC_PRE_CMD 0x3c0
>> +#define DDRC_ACT_CMD 0x3c4
>> +#define DDRC_BNK_CHG 0x3c8
>> +#define DDRC_RNK_CHG 0x3cc
>> +#define DDRC_EVENT_CTRL 0x6C0
>> +#define DDRC_INT_MASK 0x6c8
>> +#define DDRC_INT_STATUS 0x6cc
>> +#define DDRC_INT_CLEAR 0x6d0
>> +
>> +/* DDRC supports 8-events and counter is fixed-purpose */
>> +#define DDRC_NR_COUNTERS 0x8
>> +#define DDRC_NR_EVENTS DDRC_NR_COUNTERS
>> +
>> +#define DDRC_PERF_CTRL_EN 0x2
>> +
>> +/*
>> + * For DDRC PMU, there are eight-events and every event has been mapped
>> + * to fixed-purpose counters which register offset is not consistent.
>> + * Therefore there is no write event type and we assume that event
>> + * code (0 to 7) is equal to counter index in PMU driver.
>> + */
>> +#define GET_DDRC_EVENTID(hwc) (hwc->config_base & 0x7)
>> +
>> +static const u32 ddrc_reg_off[] = {
>> + DDRC_FLUX_WR, DDRC_FLUX_RD, DDRC_FLUX_WCMD, DDRC_FLUX_RCMD,
>> + DDRC_PRE_CMD, DDRC_ACT_CMD, DDRC_BNK_CHG, DDRC_RNK_CHG
>> +};
>> +
>> +/*
>> + * Select the counter register offset using the counter index.
>> + * In DDRC there are no programmable counter, the count
>> + * is readed form the statistics counter register itself.
>> + */
>> +static u32 get_counter_reg_off(int cntr_idx)
>> +{
>> + return ddrc_reg_off[cntr_idx];
>> +}
>> +
>> +static u64 hisi_ddrc_pmu_read_counter(struct hisi_pmu *ddrc_pmu,
>> + struct hw_perf_event *hwc)
>> +{
>> + /* Use event code as counter index */
>> + u32 idx = GET_DDRC_EVENTID(hwc);
>> + u32 reg;
>> +
>> + if (!hisi_uncore_pmu_counter_valid(ddrc_pmu, idx)) {
>> + dev_err(ddrc_pmu->dev, "Unsupported event index:%d!\n", idx);
>> + return 0;
>> + }
>> +
>> + reg = get_counter_reg_off(idx);
>> +
>> + return readl(ddrc_pmu->base + reg);
>> +}
>> +
>> +static void hisi_ddrc_pmu_write_counter(struct hisi_pmu *ddrc_pmu,
>> + struct hw_perf_event *hwc, u64 val)
>> +{
>> + u32 idx = GET_DDRC_EVENTID(hwc);
>> + u32 reg;
>> +
>> + if (!hisi_uncore_pmu_counter_valid(ddrc_pmu, idx)) {
>> + dev_err(ddrc_pmu->dev, "Unsupported event index:%d!\n", idx);
>> + return;
>> + }
>> +
>> + reg = get_counter_reg_off(idx);
>> + writel((u32)val, ddrc_pmu->base + reg);
>> +}
>> +
>> +static void hisi_ddrc_pmu_start_counters(struct hisi_pmu *ddrc_pmu)
>> +{
>> + u32 val;
>> +
>> + /* Set perf_enable in DDRC_PERF_CTRL to start event counting */
>> + val = readl(ddrc_pmu->base + DDRC_PERF_CTRL);
>> + val |= DDRC_PERF_CTRL_EN;
>> + writel(val, ddrc_pmu->base + DDRC_PERF_CTRL);
>> +}
>> +
>> +static void hisi_ddrc_pmu_stop_counters(struct hisi_pmu *ddrc_pmu)
>> +{
>> + u32 val;
>> +
>> + /* Clear perf_enable in DDRC_PERF_CTRL to stop event counting */
>> + val = readl(ddrc_pmu->base + DDRC_PERF_CTRL);
>> + val &= ~DDRC_PERF_CTRL_EN;
>> + writel(val, ddrc_pmu->base + DDRC_PERF_CTRL);
>> +}
>> +
>> +static void hisi_ddrc_pmu_enable_counter(struct hisi_pmu *ddrc_pmu,
>> + struct hw_perf_event *hwc)
>> +{
>> + u32 val;
>> +
>> + /* Set counter index(event code) in DDRC_EVENT_CTRL register */
>> + val = readl(ddrc_pmu->base + DDRC_EVENT_CTRL);
>> + val |= (1 << GET_DDRC_EVENTID(hwc));
>> + writel(val, ddrc_pmu->base + DDRC_EVENT_CTRL);
>> +}
>> +
>> +static void hisi_ddrc_pmu_disable_counter(struct hisi_pmu *ddrc_pmu,
>> + struct hw_perf_event *hwc)
>> +{
>> + u32 val;
>> +
>> + /* Clear counter index(event code) in DDRC_EVENT_CTRL register */
>> + val = readl(ddrc_pmu->base + DDRC_EVENT_CTRL);
>> + val &= ~(1 << GET_DDRC_EVENTID(hwc));
>> + writel(val, ddrc_pmu->base + DDRC_EVENT_CTRL);
>> +}
>> +
>> +static int hisi_ddrc_pmu_get_event_idx(struct perf_event *event)
>> +{
>> + struct hisi_pmu *ddrc_pmu = to_hisi_pmu(event->pmu);
>> + unsigned long *used_mask = ddrc_pmu->pmu_events.used_mask;
>> + struct hw_perf_event *hwc = &event->hw;
>> + /* For DDRC PMU, we use event code as counter index */
>> + int idx = GET_DDRC_EVENTID(hwc);
>> +
>> + if (test_bit(idx, used_mask))
>> + return -EAGAIN;
>> +
>> + set_bit(idx, used_mask);
>> +
>> + return idx;
>> +}
>> +
>> +static void hisi_ddrc_pmu_enable_counter_int(struct hisi_pmu *ddrc_pmu,
>> + struct hw_perf_event *hwc)
>> +{
>> + u32 val;
>> +
>> + /* Write 0 to enable interrupt */
>> + val = readl(ddrc_pmu->base + DDRC_INT_MASK);
>> + val &= ~(1 << GET_DDRC_EVENTID(hwc));
>> + writel(val, ddrc_pmu->base + DDRC_INT_MASK);
>> +}
>> +
>> +static void hisi_ddrc_pmu_disable_counter_int(struct hisi_pmu *ddrc_pmu,
>> + struct hw_perf_event *hwc)
>> +{
>> + u32 val;
>> +
>> + /* Write 1 to mask interrupt */
>> + val = readl(ddrc_pmu->base + DDRC_INT_MASK);
>> + val |= (1 << GET_DDRC_EVENTID(hwc));
>> + writel(val, ddrc_pmu->base + DDRC_INT_MASK);
>> +}
>> +
>> +static irqreturn_t hisi_ddrc_pmu_isr(int irq, void *dev_id)
>> +{
>> + struct hisi_pmu *ddrc_pmu = dev_id;
>> + struct perf_event *event;
>> + unsigned long overflown;
>> + u32 status;
>> + int idx;
>> +
>> + /* Read the DDRC_INT_STATUS register */
>> + status = readl(ddrc_pmu->base + DDRC_INT_STATUS);
>> + if (!status)
>> + return IRQ_NONE;
>> + overflown = status;
>> +
>> + /*
>> + * Find the counter index which overflowed if the bit was set
>> + * and handle it
>> + */
>> + for_each_set_bit(idx, &overflown, DDRC_NR_COUNTERS) {
>> + /* Write 1 to clear the IRQ status flag */
>> + writel((1 << idx), ddrc_pmu->base + DDRC_INT_CLEAR);
>> +
>> + /* Get the corresponding event struct */
>> + event = ddrc_pmu->pmu_events.hw_events[idx];
>> + if (!event)
>> + continue;
>> +
>> + hisi_uncore_pmu_event_update(event);
>> + hisi_uncore_pmu_set_event_period(event);
>> + }
>> +
>> + return IRQ_HANDLED;
>> +}
>> +
>> +static int hisi_ddrc_pmu_init_irq(struct hisi_pmu *ddrc_pmu,
>> + struct platform_device *pdev)
>> +{
>> + struct device *dev = &pdev->dev;
>> + int irq, ret;
>> +
>> + /* Read and init IRQ */
>> + irq = platform_get_irq(pdev, 0);
>> + if (irq < 0) {
>> + dev_err(dev, "irq init: fail map DDRC overflow interrupt\n");
>> + return irq;
>> + }
>> +
>> + ret = devm_request_irq(dev, irq, hisi_ddrc_pmu_isr,
>> + IRQF_NOBALANCING | IRQF_NO_THREAD,
>> + dev_name(dev), ddrc_pmu);
>> + if (ret < 0) {
>> + dev_err(dev, "Fail to request IRQ:%d ret:%d\n", irq, ret);
>> + return ret;
>> + }
>> +
>> + /* Overflow interrupt also should use the same CPU */
>> + WARN_ON(irq_set_affinity(irq, &ddrc_pmu->cpus));
>> +
>> + return 0;
>> +}
>> +
>> +static const struct acpi_device_id hisi_ddrc_pmu_acpi_match[] = {
>> + { "HISI0233", },
>> + {},
>> +};
>> +MODULE_DEVICE_TABLE(acpi, hisi_ddrc_pmu_acpi_match);
>> +
>> +static int hisi_ddrc_pmu_init_data(struct platform_device *pdev,
>> + struct hisi_pmu *ddrc_pmu)
>> +{
>> + struct device *dev = &pdev->dev;
>> + struct resource *res;
>> +
>> + /* Get the DDRC Channel ID */
>> + if (device_property_read_u32(dev, "hisilicon,ch-id",
>> + &ddrc_pmu->ddrc_chn_id)) {
>> + dev_err(dev, "Can not read ddrc ch-id!\n");
>> + return -EINVAL;
>> + }
>> +
>> + /* Get the DDRC SCCL ID */
>> + if (device_property_read_u32(dev, "hisilicon,scl-id",
>> + &ddrc_pmu->sccl_id)) {
>> + dev_err(dev, "Can not read ddrc sccl-id!\n");
>> + return -EINVAL;
>> + }
>> +
>> + res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
>> + ddrc_pmu->base = devm_ioremap_resource(dev, res);
>> + if (IS_ERR(ddrc_pmu->base)) {
>> + dev_err(dev, "ioremap failed for ddrc_pmu resource\n");
>> + return PTR_ERR(ddrc_pmu->base);
>> + }
>> +
>> + return 0;
>> +}
>> +
>> +static struct attribute *hisi_ddrc_pmu_format_attr[] = {
>> + HISI_PMU_FORMAT_ATTR(event, "config:0-4"),
>> + NULL,
>> +};
>> +
>> +static const struct attribute_group hisi_ddrc_pmu_format_group = {
>> + .name = "format",
>> + .attrs = hisi_ddrc_pmu_format_attr,
>> +};
>> +
>> +static struct attribute *hisi_ddrc_pmu_events_attr[] = {
>> + HISI_PMU_EVENT_ATTR(flux_wr, 0x00),
>> + HISI_PMU_EVENT_ATTR(flux_rd, 0x01),
>> + HISI_PMU_EVENT_ATTR(flux_wcmd, 0x02),
>> + HISI_PMU_EVENT_ATTR(flux_rcmd, 0x03),
>> + HISI_PMU_EVENT_ATTR(pre_cmd, 0x04),
>> + HISI_PMU_EVENT_ATTR(act_cmd, 0x05),
>> + HISI_PMU_EVENT_ATTR(rnk_chg, 0x06),
>> + HISI_PMU_EVENT_ATTR(rw_chg, 0x07),
>> + NULL,
>> +};
>> +
>> +static const struct attribute_group hisi_ddrc_pmu_events_group = {
>> + .name = "events",
>> + .attrs = hisi_ddrc_pmu_events_attr,
>> +};
>> +
>> +static DEVICE_ATTR(cpumask, 0444, hisi_cpumask_sysfs_show, NULL);
>> +
>> +static struct attribute *hisi_ddrc_pmu_cpumask_attrs[] = {
>> + &dev_attr_cpumask.attr,
>> + NULL,
>> +};
>> +
>> +static const struct attribute_group hisi_ddrc_pmu_cpumask_attr_group = {
>> + .attrs = hisi_ddrc_pmu_cpumask_attrs,
>> +};
>> +
>> +static const struct attribute_group *hisi_ddrc_pmu_attr_groups[] = {
>> + &hisi_ddrc_pmu_format_group,
>> + &hisi_ddrc_pmu_events_group,
>> + &hisi_ddrc_pmu_cpumask_attr_group,
>> + NULL,
>> +};
>> +
>> +static const struct hisi_uncore_ops hisi_uncore_ddrc_ops = {
>> + .get_event_idx = hisi_ddrc_pmu_get_event_idx,
>> + .start_counters = hisi_ddrc_pmu_start_counters,
>> + .stop_counters = hisi_ddrc_pmu_stop_counters,
>> + .enable_counter = hisi_ddrc_pmu_enable_counter,
>> + .disable_counter = hisi_ddrc_pmu_disable_counter,
>> + .enable_counter_int = hisi_ddrc_pmu_enable_counter_int,
>> + .disable_counter_int = hisi_ddrc_pmu_disable_counter_int,
>> + .write_counter = hisi_ddrc_pmu_write_counter,
>> + .read_counter = hisi_ddrc_pmu_read_counter,
>> +};
>> +
>> +static int hisi_ddrc_pmu_dev_probe(struct platform_device *pdev,
>> + struct hisi_pmu *ddrc_pmu)
>> +{
>> + struct device *dev = &pdev->dev;
>> + int ret;
>> +
>> + ret = hisi_ddrc_pmu_init_data(pdev, ddrc_pmu);
>> + if (ret)
>> + return ret;
>> +
>> + /* Pick one core to use for cpumask attributes */
>> + cpumask_set_cpu(smp_processor_id(), &ddrc_pmu->cpus);
>> +
>> + ret = hisi_ddrc_pmu_init_irq(ddrc_pmu, pdev);
>> + if (ret)
>> + return ret;
>> +
>> + ddrc_pmu->name = devm_kasprintf(dev, GFP_KERNEL, "hisi_ddrc%u_%u",
>> + ddrc_pmu->ddrc_chn_id,
>> + ddrc_pmu->sccl_id);
>> + ddrc_pmu->num_events = DDRC_NR_EVENTS;
>> + ddrc_pmu->num_counters = DDRC_NR_COUNTERS;
>> + ddrc_pmu->counter_bits = 32;
>> + ddrc_pmu->ops = &hisi_uncore_ddrc_ops;
>> + ddrc_pmu->dev = dev;
>> +
>> + return 0;
>> +}
>> +
>> +static int hisi_ddrc_pmu_probe(struct platform_device *pdev)
>> +{
>> + struct device *dev = &pdev->dev;
>> + struct hisi_pmu *ddrc_pmu;
>> + int ret;
>> +
>> + ddrc_pmu = hisi_pmu_alloc(dev, DDRC_NR_COUNTERS);
>> + if (!ddrc_pmu)
>> + return -ENOMEM;
>> +
>> + platform_set_drvdata(pdev, ddrc_pmu);
>> +
>> + ret = hisi_ddrc_pmu_dev_probe(pdev, ddrc_pmu);
>> + if (ret)
>> + return ret;
>> +
>> + ddrc_pmu->pmu = (struct pmu) {
>> + .name = ddrc_pmu->name,
>> + .task_ctx_nr = perf_invalid_context,
>> + .event_init = hisi_uncore_pmu_event_init,
>> + .pmu_enable = hisi_uncore_pmu_enable,
>> + .pmu_disable = hisi_uncore_pmu_disable,
>> + .add = hisi_uncore_pmu_add,
>> + .del = hisi_uncore_pmu_del,
>> + .start = hisi_uncore_pmu_start,
>> + .stop = hisi_uncore_pmu_stop,
>> + .read = hisi_uncore_pmu_read,
>> + .attr_groups = hisi_ddrc_pmu_attr_groups,
>> + };
>> +
>> + ret = perf_pmu_register(&ddrc_pmu->pmu, ddrc_pmu->name, -1);
>> + if (ret)
>> + dev_err(ddrc_pmu->dev, "ddrc_pmu register failed!\n");
>> +
>> + return ret;
>> +}
>> +
>> +static int hisi_ddrc_pmu_remove(struct platform_device *pdev)
>> +{
>> + struct hisi_pmu *ddrc_pmu = platform_get_drvdata(pdev);
>> +
>> + perf_pmu_unregister(&ddrc_pmu->pmu);
>> +
>> + return 0;
>> +}
>> +
>> +static struct platform_driver hisi_ddrc_pmu_driver = {
>> + .driver = {
>> + .name = "hisi_ddrc_pmu",
>> + .acpi_match_table = ACPI_PTR(hisi_ddrc_pmu_acpi_match),
>> + },
>> + .probe = hisi_ddrc_pmu_probe,
>> + .remove = hisi_ddrc_pmu_remove,
>> +};
>> +module_platform_driver(hisi_ddrc_pmu_driver);
>> +
>> +MODULE_DESCRIPTION("HiSilicon SoC DDRC uncore PMU driver");
>> +MODULE_LICENSE("GPL v2");
>> +MODULE_AUTHOR("Shaokun Zhang <zhangshaokun@hisilicon.com>");
>> +MODULE_AUTHOR("Anurup M <anurup.m@huawei.com>");
>> --
>> 1.9.1
>>
>
> _______________________________________________
> linux-arm-kernel mailing list
> linux-arm-kernel@lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
>
> .
>
^ permalink raw reply [flat|nested] 19+ messages in thread
* [PATCH v4 6/6] arm64: MAINTAINERS: hisi: Add HiSilicon SoC PMU support
2017-07-25 12:10 [PATCH v4 0/6] Add HiSilicon SoC uncore Performance Monitoring Unit driver Shaokun Zhang
` (4 preceding siblings ...)
2017-07-25 12:10 ` [PATCH v4 5/6] perf: hisi: Add support for HiSilicon SoC DDRC " Shaokun Zhang
@ 2017-07-25 12:10 ` Shaokun Zhang
2017-08-07 8:57 ` [PATCH v4 0/6] Add HiSilicon SoC uncore Performance Monitoring Unit driver Zhangshaokun
6 siblings, 0 replies; 19+ messages in thread
From: Shaokun Zhang @ 2017-07-25 12:10 UTC (permalink / raw)
To: mark.rutland, will.deacon
Cc: jonathan.cameron, linux-arm-kernel, linux-kernel, linux-doc, linuxarm
Add support HiSilicon SoC uncore PMU driver.
Signed-off-by: Shaokun Zhang <zhangshaokun@hisilicon.com>
---
MAINTAINERS | 7 +++++++
1 file changed, 7 insertions(+)
diff --git a/MAINTAINERS b/MAINTAINERS
index 205d397..649b144 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -6197,6 +6197,13 @@ S: Maintained
F: drivers/net/ethernet/hisilicon/
F: Documentation/devicetree/bindings/net/hisilicon*.txt
+HISILICON PMU DRIVER
+M: Shaokun Zhang <zhangshaokun@hisilicon.com>
+W: http://www.hisilicon.com
+S: Supported
+F: drivers/perf/hisilicon
+F: Documentation/perf/hisi-pmu.txt
+
HISILICON ROCE DRIVER
M: Lijun Ou <oulijun@huawei.com>
M: Wei Hu(Xavier) <xavier.huwei@huawei.com>
--
1.9.1
^ permalink raw reply related [flat|nested] 19+ messages in thread* Re: [PATCH v4 0/6] Add HiSilicon SoC uncore Performance Monitoring Unit driver
2017-07-25 12:10 [PATCH v4 0/6] Add HiSilicon SoC uncore Performance Monitoring Unit driver Shaokun Zhang
` (5 preceding siblings ...)
2017-07-25 12:10 ` [PATCH v4 6/6] arm64: MAINTAINERS: hisi: Add HiSilicon SoC PMU support Shaokun Zhang
@ 2017-08-07 8:57 ` Zhangshaokun
6 siblings, 0 replies; 19+ messages in thread
From: Zhangshaokun @ 2017-08-07 8:57 UTC (permalink / raw)
To: mark.rutland, will.deacon
Cc: linux-doc, linuxarm, linux-kernel, linux-arm-kernel, Jonathan Cameron
Hi Mark/Will,
Appreciate your comments.
Thanks.
Shaokun
On 2017/7/25 20:10, Shaokun Zhang wrote:
> This patchset adds support for HiSilicon SoC uncore PMUs driver. It
> includes L3C, Hydra Home Agent (HHA) and DDRC.
>
> Changes in v4:
> * remove redundant code and comments
> * reverse the functions order in exit function
> * remove some GPL information
> * revise including header file
> * fix Jonathan's other comments
>
> Changes in v3:
> * rebase to 4.13-rc1
> * add dev_err if ioremap fails for PMUs
>
> Changes in v2:
> * fix kbuild test robot error
> * make hisi_uncore_ops static
>
> Shaokun Zhang (6):
> Documentation: perf: hisi: Documentation for HiSilicon SoC PMU driver
> perf: hisi: Add support for HiSilicon SoC uncore PMU driver
> perf: hisi: Add support for HiSilicon SoC L3C PMU driver
> perf: hisi: Add support for HiSilicon SoC HHA PMU driver
> perf: hisi: Add support for HiSilicon SoC DDRC PMU driver
> arm64: MAINTAINERS: hisi: Add HiSilicon SoC PMU support
>
> Documentation/perf/hisi-pmu.txt | 52 +++
> MAINTAINERS | 7 +
> drivers/perf/Kconfig | 7 +
> drivers/perf/Makefile | 1 +
> drivers/perf/hisilicon/Makefile | 1 +
> drivers/perf/hisilicon/hisi_uncore_ddrc_pmu.c | 420 ++++++++++++++++++++
> drivers/perf/hisilicon/hisi_uncore_hha_pmu.c | 436 +++++++++++++++++++++
> drivers/perf/hisilicon/hisi_uncore_l3c_pmu.c | 538 ++++++++++++++++++++++++++
> drivers/perf/hisilicon/hisi_uncore_pmu.c | 398 +++++++++++++++++++
> drivers/perf/hisilicon/hisi_uncore_pmu.h | 103 +++++
> include/linux/cpuhotplug.h | 1 +
> 11 files changed, 1964 insertions(+)
> create mode 100644 Documentation/perf/hisi-pmu.txt
> create mode 100644 drivers/perf/hisilicon/Makefile
> create mode 100644 drivers/perf/hisilicon/hisi_uncore_ddrc_pmu.c
> create mode 100644 drivers/perf/hisilicon/hisi_uncore_hha_pmu.c
> create mode 100644 drivers/perf/hisilicon/hisi_uncore_l3c_pmu.c
> create mode 100644 drivers/perf/hisilicon/hisi_uncore_pmu.c
> create mode 100644 drivers/perf/hisilicon/hisi_uncore_pmu.h
>
^ permalink raw reply [flat|nested] 19+ messages in thread