drivers: perf: Add LLC-TAD perf counter support

This driver adds support for Last-level cache tag-and-data unit
(LLC-TAD) PMU that is featured in some of the Marvell's CN10K
infrastructure silicons.

The LLC is divided into 2N slices distributed across N Mesh tiles
in a single-socket configuration. The driver always configures the
same counter for all of the TADs. The user would end up effectively
reserving one of eight counters in every TAD to look across all TADs.
The occurrences of events are aggregated and presented to the user
at the end of an application run. The driver does not provide a way
for the user to partition TADs so that different TADs are used for
different applications.

The event counters are zeroed to start event counting to avoid any
rollover issues. TAD perf counters are 64-bit, so it's not currently
possible to overflow event counters at current mesh and core
frequencies.

To measure tad pmu events use perf tool stat command. For instance:

perf stat -e tad_dat_msh_in_dss,tad_req_msh_out_any <application>
perf stat -e tad_alloc_any,tad_hit_any,tad_tag_rd <application>

Signed-off-by: Bhaskara Budiredla <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Will Deacon <[email protected]>

Author: Bhaskara Budiredla

drivers/perf/Kconfig

@@ -139,6 +139,13 @@ config ARM_DMC620_PMU
 	  Support for PMU events monitoring on the ARM DMC-620 memory
 	  controller.
 
+config MARVELL_CN10K_TAD_PMU
+	tristate "Marvell CN10K LLC-TAD PMU"
+	depends on ARM64 || (COMPILE_TEST && 64BIT)
+	help
+	  Provides support for Last-Level cache Tag-and-data Units (LLC-TAD)
+	  performance monitors on CN10K family silicons.
+
 source "drivers/perf/hisilicon/Kconfig"
 
 endmenu

drivers/perf/Makefile

@@ -14,3 +14,4 @@ obj-$(CONFIG_THUNDERX2_PMU) += thunderx2_pmu.o
 obj-$(CONFIG_XGENE_PMU) += xgene_pmu.o
 obj-$(CONFIG_ARM_SPE_PMU) += arm_spe_pmu.o
 obj-$(CONFIG_ARM_DMC620_PMU) += arm_dmc620_pmu.o
+obj-$(CONFIG_MARVELL_CN10K_TAD_PMU) += marvell_cn10k_tad_pmu.o