Naga Chumbalkar | 0f1d683 | 2009-12-17 20:18:27 +0000 | [diff] [blame] | 1 | /* |
| 2 | * pcc-cpufreq.txt - PCC interface documentation |
| 3 | * |
| 4 | * Copyright (C) 2009 Red Hat, Matthew Garrett <mjg@redhat.com> |
| 5 | * Copyright (C) 2009 Hewlett-Packard Development Company, L.P. |
| 6 | * Nagananda Chumbalkar <nagananda.chumbalkar@hp.com> |
| 7 | * |
| 8 | * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
| 9 | * |
| 10 | * This program is free software; you can redistribute it and/or modify |
| 11 | * it under the terms of the GNU General Public License as published by |
| 12 | * the Free Software Foundation; version 2 of the License. |
| 13 | * |
| 14 | * This program is distributed in the hope that it will be useful, but |
| 15 | * WITHOUT ANY WARRANTY; without even the implied warranty of |
| 16 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE, GOOD TITLE or NON |
| 17 | * INFRINGEMENT. See the GNU General Public License for more details. |
| 18 | * |
| 19 | * You should have received a copy of the GNU General Public License along |
| 20 | * with this program; if not, write to the Free Software Foundation, Inc., |
| 21 | * 675 Mass Ave, Cambridge, MA 02139, USA. |
| 22 | * |
| 23 | * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
| 24 | */ |
| 25 | |
| 26 | |
| 27 | Processor Clocking Control Driver |
| 28 | --------------------------------- |
| 29 | |
| 30 | Contents: |
| 31 | --------- |
| 32 | 1. Introduction |
| 33 | 1.1 PCC interface |
| 34 | 1.1.1 Get Average Frequency |
| 35 | 1.1.2 Set Desired Frequency |
| 36 | 1.2 Platforms affected |
| 37 | 2. Driver and /sys details |
| 38 | 2.1 scaling_available_frequencies |
| 39 | 2.2 cpuinfo_transition_latency |
| 40 | 2.3 cpuinfo_cur_freq |
| 41 | 2.4 related_cpus |
| 42 | 3. Caveats |
| 43 | |
| 44 | 1. Introduction: |
| 45 | ---------------- |
| 46 | Processor Clocking Control (PCC) is an interface between the platform |
| 47 | firmware and OSPM. It is a mechanism for coordinating processor |
| 48 | performance (ie: frequency) between the platform firmware and the OS. |
| 49 | |
| 50 | The PCC driver (pcc-cpufreq) allows OSPM to take advantage of the PCC |
| 51 | interface. |
| 52 | |
| 53 | OS utilizes the PCC interface to inform platform firmware what frequency the |
| 54 | OS wants for a logical processor. The platform firmware attempts to achieve |
| 55 | the requested frequency. If the request for the target frequency could not be |
| 56 | satisfied by platform firmware, then it usually means that power budget |
| 57 | conditions are in place, and "power capping" is taking place. |
| 58 | |
| 59 | 1.1 PCC interface: |
| 60 | ------------------ |
| 61 | The complete PCC specification is available here: |
| 62 | http://www.acpica.org/download/Processor-Clocking-Control-v1p0.pdf |
| 63 | |
| 64 | PCC relies on a shared memory region that provides a channel for communication |
| 65 | between the OS and platform firmware. PCC also implements a "doorbell" that |
| 66 | is used by the OS to inform the platform firmware that a command has been |
| 67 | sent. |
| 68 | |
| 69 | The ACPI PCCH() method is used to discover the location of the PCC shared |
| 70 | memory region. The shared memory region header contains the "command" and |
| 71 | "status" interface. PCCH() also contains details on how to access the platform |
| 72 | doorbell. |
| 73 | |
| 74 | The following commands are supported by the PCC interface: |
| 75 | * Get Average Frequency |
| 76 | * Set Desired Frequency |
| 77 | |
| 78 | The ACPI PCCP() method is implemented for each logical processor and is |
| 79 | used to discover the offsets for the input and output buffers in the shared |
| 80 | memory region. |
| 81 | |
| 82 | When PCC mode is enabled, the platform will not expose processor performance |
| 83 | or throttle states (_PSS, _TSS and related ACPI objects) to OSPM. Therefore, |
| 84 | the native P-state driver (such as acpi-cpufreq for Intel, powernow-k8 for |
| 85 | AMD) will not load. |
| 86 | |
| 87 | However, OSPM remains in control of policy. The governor (eg: "ondemand") |
| 88 | computes the required performance for each processor based on server workload. |
| 89 | The PCC driver fills in the command interface, and the input buffer and |
| 90 | communicates the request to the platform firmware. The platform firmware is |
| 91 | responsible for delivering the requested performance. |
| 92 | |
| 93 | Each PCC command is "global" in scope and can affect all the logical CPUs in |
| 94 | the system. Therefore, PCC is capable of performing "group" updates. With PCC |
| 95 | the OS is capable of getting/setting the frequency of all the logical CPUs in |
| 96 | the system with a single call to the BIOS. |
| 97 | |
| 98 | 1.1.1 Get Average Frequency: |
| 99 | ---------------------------- |
| 100 | This command is used by the OSPM to query the running frequency of the |
| 101 | processor since the last time this command was completed. The output buffer |
| 102 | indicates the average unhalted frequency of the logical processor expressed as |
| 103 | a percentage of the nominal (ie: maximum) CPU frequency. The output buffer |
| 104 | also signifies if the CPU frequency is limited by a power budget condition. |
| 105 | |
| 106 | 1.1.2 Set Desired Frequency: |
| 107 | ---------------------------- |
| 108 | This command is used by the OSPM to communicate to the platform firmware the |
| 109 | desired frequency for a logical processor. The output buffer is currently |
| 110 | ignored by OSPM. The next invocation of "Get Average Frequency" will inform |
| 111 | OSPM if the desired frequency was achieved or not. |
| 112 | |
| 113 | 1.2 Platforms affected: |
| 114 | ----------------------- |
| 115 | The PCC driver will load on any system where the platform firmware: |
| 116 | * supports the PCC interface, and the associated PCCH() and PCCP() methods |
| 117 | * assumes responsibility for managing the hardware clocking controls in order |
| 118 | to deliver the requested processor performance |
| 119 | |
| 120 | Currently, certain HP ProLiant platforms implement the PCC interface. On those |
| 121 | platforms PCC is the "default" choice. |
| 122 | |
| 123 | However, it is possible to disable this interface via a BIOS setting. In |
| 124 | such an instance, as is also the case on platforms where the PCC interface |
| 125 | is not implemented, the PCC driver will fail to load silently. |
| 126 | |
| 127 | 2. Driver and /sys details: |
| 128 | --------------------------- |
| 129 | When the driver loads, it merely prints the lowest and the highest CPU |
| 130 | frequencies supported by the platform firmware. |
| 131 | |
| 132 | The PCC driver loads with a message such as: |
| 133 | pcc-cpufreq: (v1.00.00) driver loaded with frequency limits: 1600 MHz, 2933 |
| 134 | MHz |
| 135 | |
| 136 | This means that the OPSM can request the CPU to run at any frequency in |
| 137 | between the limits (1600 MHz, and 2933 MHz) specified in the message. |
| 138 | |
| 139 | Internally, there is no need for the driver to convert the "target" frequency |
| 140 | to a corresponding P-state. |
| 141 | |
| 142 | The VERSION number for the driver will be of the format v.xy.ab. |
| 143 | eg: 1.00.02 |
| 144 | ----- -- |
| 145 | | | |
| 146 | | -- this will increase with bug fixes/enhancements to the driver |
| 147 | |-- this is the version of the PCC specification the driver adheres to |
| 148 | |
| 149 | |
| 150 | The following is a brief discussion on some of the fields exported via the |
| 151 | /sys filesystem and how their values are affected by the PCC driver: |
| 152 | |
| 153 | 2.1 scaling_available_frequencies: |
| 154 | ---------------------------------- |
| 155 | scaling_available_frequencies is not created in /sys. No intermediate |
| 156 | frequencies need to be listed because the BIOS will try to achieve any |
| 157 | frequency, within limits, requested by the governor. A frequency does not have |
| 158 | to be strictly associated with a P-state. |
| 159 | |
| 160 | 2.2 cpuinfo_transition_latency: |
| 161 | ------------------------------- |
| 162 | The cpuinfo_transition_latency field is 0. The PCC specification does |
| 163 | not include a field to expose this value currently. |
| 164 | |
| 165 | 2.3 cpuinfo_cur_freq: |
| 166 | --------------------- |
| 167 | A) Often cpuinfo_cur_freq will show a value different than what is declared |
| 168 | in the scaling_available_frequencies or scaling_cur_freq, or scaling_max_freq. |
| 169 | This is due to "turbo boost" available on recent Intel processors. If certain |
| 170 | conditions are met the BIOS can achieve a slightly higher speed than requested |
| 171 | by OSPM. An example: |
| 172 | |
| 173 | scaling_cur_freq : 2933000 |
| 174 | cpuinfo_cur_freq : 3196000 |
| 175 | |
| 176 | B) There is a round-off error associated with the cpuinfo_cur_freq value. |
| 177 | Since the driver obtains the current frequency as a "percentage" (%) of the |
| 178 | nominal frequency from the BIOS, sometimes, the values displayed by |
| 179 | scaling_cur_freq and cpuinfo_cur_freq may not match. An example: |
| 180 | |
| 181 | scaling_cur_freq : 1600000 |
| 182 | cpuinfo_cur_freq : 1583000 |
| 183 | |
| 184 | In this example, the nominal frequency is 2933 MHz. The driver obtains the |
| 185 | current frequency, cpuinfo_cur_freq, as 54% of the nominal frequency: |
| 186 | |
| 187 | 54% of 2933 MHz = 1583 MHz |
| 188 | |
| 189 | Nominal frequency is the maximum frequency of the processor, and it usually |
| 190 | corresponds to the frequency of the P0 P-state. |
| 191 | |
| 192 | 2.4 related_cpus: |
| 193 | ----------------- |
| 194 | The related_cpus field is identical to affected_cpus. |
| 195 | |
| 196 | affected_cpus : 4 |
| 197 | related_cpus : 4 |
| 198 | |
| 199 | Currently, the PCC driver does not evaluate _PSD. The platforms that support |
| 200 | PCC do not implement SW_ALL. So OSPM doesn't need to perform any coordination |
| 201 | to ensure that the same frequency is requested of all dependent CPUs. |
| 202 | |
| 203 | 3. Caveats: |
| 204 | ----------- |
| 205 | The "cpufreq_stats" module in its present form cannot be loaded and |
| 206 | expected to work with the PCC driver. Since the "cpufreq_stats" module |
| 207 | provides information wrt each P-state, it is not applicable to the PCC driver. |