Linux kernel mirror (for testing) git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel os linux

docs: perf: Add description for Synopsys DesignWare PCIe PMU driver

Alibaba's T-Head Yitan 710 SoC includes Synopsys' DesignWare Core PCIe
controller which implements PMU for performance and functional debugging to
facilitate system maintenance.

Document it to provide guidance on how to use it.

Signed-off-by: Shuai Xue <xueshuai@linux.alibaba.com>
Reviewed-by: Baolin Wang <baolin.wang@linux.alibaba.com>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Reviewed-by: Yicong Yang <yangyicong@hisilicon.com>
Tested-by: Ilkka Koskinen <ilkka@os.amperecomputing.com>
Link: https://lore.kernel.org/r/20231208025652.87192-2-xueshuai@linux.alibaba.com
Signed-off-by: Will Deacon <will@kernel.org>

authored by

Shuai Xue and committed by
Will Deacon
cae40614 eb183b2c

+95
+94
Documentation/admin-guide/perf/dwc_pcie_pmu.rst
··· 1 + ====================================================================== 2 + Synopsys DesignWare Cores (DWC) PCIe Performance Monitoring Unit (PMU) 3 + ====================================================================== 4 + 5 + DesignWare Cores (DWC) PCIe PMU 6 + =============================== 7 + 8 + The PMU is a PCIe configuration space register block provided by each PCIe Root 9 + Port in a Vendor-Specific Extended Capability named RAS D.E.S (Debug, Error 10 + injection, and Statistics). 11 + 12 + As the name indicates, the RAS DES capability supports system level 13 + debugging, AER error injection, and collection of statistics. To facilitate 14 + collection of statistics, Synopsys DesignWare Cores PCIe controller 15 + provides the following two features: 16 + 17 + - one 64-bit counter for Time Based Analysis (RX/TX data throughput and 18 + time spent in each low-power LTSSM state) and 19 + - one 32-bit counter for Event Counting (error and non-error events for 20 + a specified lane) 21 + 22 + Note: There is no interrupt for counter overflow. 23 + 24 + Time Based Analysis 25 + ------------------- 26 + 27 + Using this feature you can obtain information regarding RX/TX data 28 + throughput and time spent in each low-power LTSSM state by the controller. 29 + The PMU measures data in two categories: 30 + 31 + - Group#0: Percentage of time the controller stays in LTSSM states. 32 + - Group#1: Amount of data processed (Units of 16 bytes). 33 + 34 + Lane Event counters 35 + ------------------- 36 + 37 + Using this feature you can obtain Error and Non-Error information in 38 + specific lane by the controller. The PMU event is selected by all of: 39 + 40 + - Group i 41 + - Event j within the Group i 42 + - Lane k 43 + 44 + Some of the events only exist for specific configurations. 45 + 46 + DesignWare Cores (DWC) PCIe PMU Driver 47 + ======================================= 48 + 49 + This driver adds PMU devices for each PCIe Root Port named based on the BDF of 50 + the Root Port. For example, 51 + 52 + 30:03.0 PCI bridge: Device 1ded:8000 (rev 01) 53 + 54 + the PMU device name for this Root Port is dwc_rootport_3018. 55 + 56 + The DWC PCIe PMU driver registers a perf PMU driver, which provides 57 + description of available events and configuration options in sysfs, see 58 + /sys/bus/event_source/devices/dwc_rootport_{bdf}. 59 + 60 + The "format" directory describes format of the config fields of the 61 + perf_event_attr structure. The "events" directory provides configuration 62 + templates for all documented events. For example, 63 + "Rx_PCIe_TLP_Data_Payload" is an equivalent of "eventid=0x22,type=0x1". 64 + 65 + The "perf list" command shall list the available events from sysfs, e.g.:: 66 + 67 + $# perf list | grep dwc_rootport 68 + <...> 69 + dwc_rootport_3018/Rx_PCIe_TLP_Data_Payload/ [Kernel PMU event] 70 + <...> 71 + dwc_rootport_3018/rx_memory_read,lane=?/ [Kernel PMU event] 72 + 73 + Time Based Analysis Event Usage 74 + ------------------------------- 75 + 76 + Example usage of counting PCIe RX TLP data payload (Units of bytes):: 77 + 78 + $# perf stat -a -e dwc_rootport_3018/Rx_PCIe_TLP_Data_Payload/ 79 + 80 + The average RX/TX bandwidth can be calculated using the following formula: 81 + 82 + PCIe RX Bandwidth = Rx_PCIe_TLP_Data_Payload / Measure_Time_Window 83 + PCIe TX Bandwidth = Tx_PCIe_TLP_Data_Payload / Measure_Time_Window 84 + 85 + Lane Event Usage 86 + ------------------------------- 87 + 88 + Each lane has the same event set and to avoid generating a list of hundreds 89 + of events, the user need to specify the lane ID explicitly, e.g.:: 90 + 91 + $# perf stat -a -e dwc_rootport_3018/rx_memory_read,lane=4/ 92 + 93 + The driver does not support sampling, therefore "perf record" will not 94 + work. Per-task (without "-a") perf sessions are not supported.
+1
Documentation/admin-guide/perf/index.rst
··· 19 19 arm_dsu_pmu 20 20 thunderx2-pmu 21 21 alibaba_pmu 22 + dwc_pcie_pmu 22 23 nvidia-pmu 23 24 meson-ddr-pmu 24 25 cxl