1 ================================================
2 HiSilicon PCIe Performance Monitoring Unit (PMU)
3 ================================================
5 On Hip09, HiSilicon PCIe Performance Monitoring Unit (PMU) could monitor
6 bandwidth, latency, bus utilization and buffer occupancy data of PCIe.
8 Each PCIe Core has a PMU to monitor multi Root Ports of this PCIe Core and
9 all Endpoints downstream these Root Ports.
12 HiSilicon PCIe PMU driver
13 =========================
15 The PCIe PMU driver registers a perf PMU with the name of its sicl-id and PCIe
18 /sys/bus/event_source/hisi_pcie<sicl>_core<core>
20 PMU driver provides description of available events and filter options in sysfs,
21 see /sys/bus/event_source/devices/hisi_pcie<sicl>_core<core>.
23 The "format" directory describes all formats of the config (events) and config1
24 (filter options) fields of the perf_event_attr structure. The "events" directory
25 describes all documented events shown in perf list.
27 The "identifier" sysfs file allows users to identify the version of the
30 The "bus" sysfs file allows users to get the bus number of Root Ports
31 monitored by PMU. Furthermore users can get the Root Ports range in
32 [bdf_min, bdf_max] from "bdf_min" and "bdf_max" sysfs attributes
35 Example usage of perf::
38 hisi_pcie0_core0/rx_mwr_latency/ [kernel PMU event]
39 hisi_pcie0_core0/rx_mwr_cnt/ [kernel PMU event]
40 ------------------------------------------
42 $# perf stat -e hisi_pcie0_core0/rx_mwr_latency,port=0xffff/
43 $# perf stat -e hisi_pcie0_core0/rx_mwr_cnt,port=0xffff/
45 The related events usually used to calculate the bandwidth, latency or others.
46 They need to start and end counting at the same time, therefore related events
47 are best used in the same event group to get the expected value. There are two
48 ways to know if they are related events:
50 a) By event name, such as the latency events "xxx_latency, xxx_cnt" or
51 bandwidth events "xxx_flux, xxx_time".
52 b) By event type, such as "event=0xXXXX, event=0x1XXXX".
54 Example usage of perf group::
56 $# perf stat -e "{hisi_pcie0_core0/rx_mwr_latency,port=0xffff/,hisi_pcie0_core0/rx_mwr_cnt,port=0xffff/}"
58 The current driver does not support sampling. So "perf record" is unsupported.
59 Also attach to a task is unsupported for PCIe PMU.
66 PMU could only monitor the performance of traffic downstream target Root
67 Ports or downstream target Endpoint. PCIe PMU driver support "port" and
68 "bdf" interfaces for users.
69 Please notice that, one of these two interfaces must be set, and these two
70 interfaces aren't supported at the same time. If they are both set, only
71 "port" filter is valid.
72 If "port" filter not being set or is set explicitly to zero (default), the
73 "bdf" filter will be in effect, because "bdf=0" meaning 0000:000:00.0.
77 "port" filter can be used in all PCIe PMU events, target Root Port can be
78 selected by configuring the 16-bits-bitmap "port". Multi ports can be
79 selected for AP-layer-events, and only one port can be selected for
82 For example, if target Root Port is 0000:00:00.0 (x8 lanes), bit0 of
83 bitmap should be set, port=0x1; if target Root Port is 0000:00:04.0 (x4
84 lanes), bit8 is set, port=0x100; if these two Root Ports are both
85 monitored, port=0x101.
87 Example usage of perf::
89 $# perf stat -e hisi_pcie0_core0/rx_mwr_latency,port=0x1/ sleep 5
93 "bdf" filter can only be used in bandwidth events, target Endpoint is
94 selected by configuring BDF to "bdf". Counter only counts the bandwidth of
95 message requested by target Endpoint.
97 For example, "bdf=0x3900" means BDF of target Endpoint is 0000:39:00.0.
99 Example usage of perf::
101 $# perf stat -e hisi_pcie0_core0/rx_mrd_flux,bdf=0x3900/ sleep 5
105 Event statistics start when the first time TLP length is greater/smaller
106 than trigger condition. You can set the trigger condition by writing
107 "trig_len", and set the trigger mode by writing "trig_mode". This filter can
108 only be used in bandwidth events.
110 For example, "trig_len=4" means trigger condition is 2^4 DW, "trig_mode=0"
111 means statistics start when TLP length > trigger condition, "trig_mode=1"
112 means start when TLP length < condition.
114 Example usage of perf::
116 $# perf stat -e hisi_pcie0_core0/rx_mrd_flux,port=0xffff,trig_len=0x4,trig_mode=1/ sleep 5
120 Counter counts when TLP length within the specified range. You can set the
121 threshold by writing "thr_len", and set the threshold mode by writing
122 "thr_mode". This filter can only be used in bandwidth events.
124 For example, "thr_len=4" means threshold is 2^4 DW, "thr_mode=0" means
125 counter counts when TLP length >= threshold, and "thr_mode=1" means counts
126 when TLP length < threshold.
128 Example usage of perf::
130 $# perf stat -e hisi_pcie0_core0/rx_mrd_flux,port=0xffff,thr_len=0x4,thr_mode=1/ sleep 5
134 When counting bandwidth, the data can be composed of certain parts of TLP
135 packets. You can specify it through "len_mode":
137 - 2'b00: Reserved (Do not use this since the behaviour is undefined)
138 - 2'b01: Bandwidth of TLP payloads
139 - 2'b10: Bandwidth of TLP headers
140 - 2'b11: Bandwidth of both TLP payloads and headers
142 For example, "len_mode=2" means only counting the bandwidth of TLP headers
143 and "len_mode=3" means the final bandwidth data is composed of both TLP
144 headers and payloads. Default value if not specified is 2'b11.
146 Example usage of perf::
148 $# perf stat -e hisi_pcie0_core0/rx_mrd_flux,port=0xffff,len_mode=0x1/ sleep 5