Thomas Gleixner | b886d83c | 2019-06-01 10:08:55 +0200 | [diff] [blame] | 1 | // SPDX-License-Identifier: GPL-2.0-only |
Gregory Haskins | 6e0534f | 2008-05-12 21:21:01 +0200 | [diff] [blame] | 2 | /* |
Peter Zijlstra | 391e43d | 2011-11-15 17:14:39 +0100 | [diff] [blame] | 3 | * kernel/sched/cpupri.c |
Gregory Haskins | 6e0534f | 2008-05-12 21:21:01 +0200 | [diff] [blame] | 4 | * |
| 5 | * CPU priority management |
| 6 | * |
| 7 | * Copyright (C) 2007-2008 Novell |
| 8 | * |
| 9 | * Author: Gregory Haskins <ghaskins@novell.com> |
| 10 | * |
| 11 | * This code tracks the priority of each CPU so that global migration |
| 12 | * decisions are easy to calculate. Each CPU can be in a state as follows: |
| 13 | * |
| 14 | * (INVALID), IDLE, NORMAL, RT1, ... RT99 |
| 15 | * |
| 16 | * going from the lowest priority to the highest. CPUs in the INVALID state |
| 17 | * are not eligible for routing. The system maintains this state with |
Ingo Molnar | 97fb7a0 | 2018-03-03 14:01:12 +0100 | [diff] [blame] | 18 | * a 2 dimensional bitmap (the first for priority class, the second for CPUs |
Gregory Haskins | 6e0534f | 2008-05-12 21:21:01 +0200 | [diff] [blame] | 19 | * in that class). Therefore a typical application without affinity |
| 20 | * restrictions can find a suitable CPU with O(1) complexity (e.g. two bit |
| 21 | * searches). For tasks with affinity restrictions, the algorithm has a |
| 22 | * worst case complexity of O(min(102, nr_domcpus)), though the scenario that |
| 23 | * yields the worst case search is fairly contrived. |
Gregory Haskins | 6e0534f | 2008-05-12 21:21:01 +0200 | [diff] [blame] | 24 | */ |
Ingo Molnar | 325ea10 | 2018-03-03 12:20:47 +0100 | [diff] [blame] | 25 | #include "sched.h" |
Gregory Haskins | 6e0534f | 2008-05-12 21:21:01 +0200 | [diff] [blame] | 26 | |
| 27 | /* Convert between a 140 based task->prio, and our 102 based cpupri */ |
| 28 | static int convert_prio(int prio) |
| 29 | { |
| 30 | int cpupri; |
| 31 | |
| 32 | if (prio == CPUPRI_INVALID) |
| 33 | cpupri = CPUPRI_INVALID; |
| 34 | else if (prio == MAX_PRIO) |
| 35 | cpupri = CPUPRI_IDLE; |
| 36 | else if (prio >= MAX_RT_PRIO) |
| 37 | cpupri = CPUPRI_NORMAL; |
| 38 | else |
| 39 | cpupri = MAX_RT_PRIO - prio + 1; |
| 40 | |
| 41 | return cpupri; |
| 42 | } |
| 43 | |
Gregory Haskins | 6e0534f | 2008-05-12 21:21:01 +0200 | [diff] [blame] | 44 | /** |
| 45 | * cpupri_find - find the best (lowest-pri) CPU in the system |
| 46 | * @cp: The cpupri context |
| 47 | * @p: The task |
Rusty Russell | 13b8bd0 | 2009-03-25 15:01:22 +1030 | [diff] [blame] | 48 | * @lowest_mask: A mask to fill in with selected CPUs (or NULL) |
Gregory Haskins | 6e0534f | 2008-05-12 21:21:01 +0200 | [diff] [blame] | 49 | * |
| 50 | * Note: This function returns the recommended CPUs as calculated during the |
Adam Buchbinder | 2a61aa4 | 2009-12-11 16:35:40 -0500 | [diff] [blame] | 51 | * current invocation. By the time the call returns, the CPUs may have in |
Gregory Haskins | 6e0534f | 2008-05-12 21:21:01 +0200 | [diff] [blame] | 52 | * fact changed priorities any number of times. While not ideal, it is not |
| 53 | * an issue of correctness since the normal rebalancer logic will correct |
| 54 | * any discrepancies created by racing against the uncertainty of the current |
| 55 | * priority configuration. |
| 56 | * |
Yacine Belkadi | e69f618 | 2013-07-12 20:45:47 +0200 | [diff] [blame] | 57 | * Return: (int)bool - CPUs were found |
Gregory Haskins | 6e0534f | 2008-05-12 21:21:01 +0200 | [diff] [blame] | 58 | */ |
| 59 | int cpupri_find(struct cpupri *cp, struct task_struct *p, |
Rusty Russell | 68e7456 | 2008-11-25 02:35:13 +1030 | [diff] [blame] | 60 | struct cpumask *lowest_mask) |
Gregory Haskins | 6e0534f | 2008-05-12 21:21:01 +0200 | [diff] [blame] | 61 | { |
Ying Xue | 014acbf | 2012-07-12 15:03:42 +0800 | [diff] [blame] | 62 | int idx = 0; |
| 63 | int task_pri = convert_prio(p->prio); |
Gregory Haskins | 6e0534f | 2008-05-12 21:21:01 +0200 | [diff] [blame] | 64 | |
Steven Rostedt (Red Hat) | 6227cb0 | 2014-04-13 09:34:53 -0400 | [diff] [blame] | 65 | BUG_ON(task_pri >= CPUPRI_NR_PRIORITIES); |
Steven Rostedt | c92211d | 2011-08-02 16:36:12 -0400 | [diff] [blame] | 66 | |
| 67 | for (idx = 0; idx < task_pri; idx++) { |
Gregory Haskins | 6e0534f | 2008-05-12 21:21:01 +0200 | [diff] [blame] | 68 | struct cpupri_vec *vec = &cp->pri_to_cpu[idx]; |
Steven Rostedt | d473750 | 2011-08-05 08:27:49 -0400 | [diff] [blame] | 69 | int skip = 0; |
Gregory Haskins | 6e0534f | 2008-05-12 21:21:01 +0200 | [diff] [blame] | 70 | |
Steven Rostedt | c92211d | 2011-08-02 16:36:12 -0400 | [diff] [blame] | 71 | if (!atomic_read(&(vec)->count)) |
Steven Rostedt | d473750 | 2011-08-05 08:27:49 -0400 | [diff] [blame] | 72 | skip = 1; |
Steven Rostedt | c92211d | 2011-08-02 16:36:12 -0400 | [diff] [blame] | 73 | /* |
| 74 | * When looking at the vector, we need to read the counter, |
| 75 | * do a memory barrier, then read the mask. |
| 76 | * |
| 77 | * Note: This is still all racey, but we can deal with it. |
| 78 | * Ideally, we only want to look at masks that are set. |
| 79 | * |
| 80 | * If a mask is not set, then the only thing wrong is that we |
| 81 | * did a little more work than necessary. |
| 82 | * |
| 83 | * If we read a zero count but the mask is set, because of the |
| 84 | * memory barriers, that can only happen when the highest prio |
| 85 | * task for a run queue has left the run queue, in which case, |
| 86 | * it will be followed by a pull. If the task we are processing |
| 87 | * fails to find a proper place to go, that pull request will |
| 88 | * pull this task if the run queue is running at a lower |
| 89 | * priority. |
| 90 | */ |
| 91 | smp_rmb(); |
Gregory Haskins | 6e0534f | 2008-05-12 21:21:01 +0200 | [diff] [blame] | 92 | |
Steven Rostedt | d473750 | 2011-08-05 08:27:49 -0400 | [diff] [blame] | 93 | /* Need to do the rmb for every iteration */ |
| 94 | if (skip) |
| 95 | continue; |
| 96 | |
Sebastian Andrzej Siewior | 3bd3706 | 2019-04-23 16:26:36 +0200 | [diff] [blame] | 97 | if (cpumask_any_and(p->cpus_ptr, vec->mask) >= nr_cpu_ids) |
Gregory Haskins | 6e0534f | 2008-05-12 21:21:01 +0200 | [diff] [blame] | 98 | continue; |
| 99 | |
Gregory Haskins | 07903af | 2009-07-30 10:57:28 -0400 | [diff] [blame] | 100 | if (lowest_mask) { |
Sebastian Andrzej Siewior | 3bd3706 | 2019-04-23 16:26:36 +0200 | [diff] [blame] | 101 | cpumask_and(lowest_mask, p->cpus_ptr, vec->mask); |
Gregory Haskins | 07903af | 2009-07-30 10:57:28 -0400 | [diff] [blame] | 102 | |
| 103 | /* |
| 104 | * We have to ensure that we have at least one bit |
| 105 | * still set in the array, since the map could have |
| 106 | * been concurrently emptied between the first and |
| 107 | * second reads of vec->mask. If we hit this |
| 108 | * condition, simply act as though we never hit this |
| 109 | * priority level and continue on. |
| 110 | */ |
| 111 | if (cpumask_any(lowest_mask) >= nr_cpu_ids) |
| 112 | continue; |
| 113 | } |
| 114 | |
Gregory Haskins | 6e0534f | 2008-05-12 21:21:01 +0200 | [diff] [blame] | 115 | return 1; |
| 116 | } |
| 117 | |
| 118 | return 0; |
| 119 | } |
| 120 | |
| 121 | /** |
Ingo Molnar | 97fb7a0 | 2018-03-03 14:01:12 +0100 | [diff] [blame] | 122 | * cpupri_set - update the CPU priority setting |
Gregory Haskins | 6e0534f | 2008-05-12 21:21:01 +0200 | [diff] [blame] | 123 | * @cp: The cpupri context |
Ingo Molnar | 97fb7a0 | 2018-03-03 14:01:12 +0100 | [diff] [blame] | 124 | * @cpu: The target CPU |
Randy Dunlap | fa75728 | 2012-01-21 11:03:13 -0800 | [diff] [blame] | 125 | * @newpri: The priority (INVALID-RT99) to assign to this CPU |
Gregory Haskins | 6e0534f | 2008-05-12 21:21:01 +0200 | [diff] [blame] | 126 | * |
| 127 | * Note: Assumes cpu_rq(cpu)->lock is locked |
| 128 | * |
| 129 | * Returns: (void) |
| 130 | */ |
| 131 | void cpupri_set(struct cpupri *cp, int cpu, int newpri) |
| 132 | { |
Ying Xue | 014acbf | 2012-07-12 15:03:42 +0800 | [diff] [blame] | 133 | int *currpri = &cp->cpu_to_pri[cpu]; |
| 134 | int oldpri = *currpri; |
| 135 | int do_mb = 0; |
Gregory Haskins | 6e0534f | 2008-05-12 21:21:01 +0200 | [diff] [blame] | 136 | |
| 137 | newpri = convert_prio(newpri); |
| 138 | |
| 139 | BUG_ON(newpri >= CPUPRI_NR_PRIORITIES); |
| 140 | |
| 141 | if (newpri == oldpri) |
| 142 | return; |
| 143 | |
| 144 | /* |
Ingo Molnar | 97fb7a0 | 2018-03-03 14:01:12 +0100 | [diff] [blame] | 145 | * If the CPU was currently mapped to a different value, we |
Steven Rostedt | c3a2ae3 | 2009-07-29 00:21:23 -0400 | [diff] [blame] | 146 | * need to map it to the new value then remove the old value. |
| 147 | * Note, we must add the new value first, otherwise we risk the |
Yong Zhang | 5710f15 | 2011-08-06 08:10:04 +0800 | [diff] [blame] | 148 | * cpu being missed by the priority loop in cpupri_find. |
Gregory Haskins | 6e0534f | 2008-05-12 21:21:01 +0200 | [diff] [blame] | 149 | */ |
Gregory Haskins | 6e0534f | 2008-05-12 21:21:01 +0200 | [diff] [blame] | 150 | if (likely(newpri != CPUPRI_INVALID)) { |
| 151 | struct cpupri_vec *vec = &cp->pri_to_cpu[newpri]; |
| 152 | |
Rusty Russell | 68e7456 | 2008-11-25 02:35:13 +1030 | [diff] [blame] | 153 | cpumask_set_cpu(cpu, vec->mask); |
Steven Rostedt | c92211d | 2011-08-02 16:36:12 -0400 | [diff] [blame] | 154 | /* |
| 155 | * When adding a new vector, we update the mask first, |
| 156 | * do a write memory barrier, and then update the count, to |
| 157 | * make sure the vector is visible when count is set. |
| 158 | */ |
Peter Zijlstra | 4e857c5 | 2014-03-17 18:06:10 +0100 | [diff] [blame] | 159 | smp_mb__before_atomic(); |
Steven Rostedt | c92211d | 2011-08-02 16:36:12 -0400 | [diff] [blame] | 160 | atomic_inc(&(vec)->count); |
Steven Rostedt | d473750 | 2011-08-05 08:27:49 -0400 | [diff] [blame] | 161 | do_mb = 1; |
Gregory Haskins | 6e0534f | 2008-05-12 21:21:01 +0200 | [diff] [blame] | 162 | } |
Steven Rostedt | c3a2ae3 | 2009-07-29 00:21:23 -0400 | [diff] [blame] | 163 | if (likely(oldpri != CPUPRI_INVALID)) { |
| 164 | struct cpupri_vec *vec = &cp->pri_to_cpu[oldpri]; |
| 165 | |
Steven Rostedt | c92211d | 2011-08-02 16:36:12 -0400 | [diff] [blame] | 166 | /* |
Steven Rostedt | d473750 | 2011-08-05 08:27:49 -0400 | [diff] [blame] | 167 | * Because the order of modification of the vec->count |
| 168 | * is important, we must make sure that the update |
| 169 | * of the new prio is seen before we decrement the |
| 170 | * old prio. This makes sure that the loop sees |
| 171 | * one or the other when we raise the priority of |
| 172 | * the run queue. We don't care about when we lower the |
| 173 | * priority, as that will trigger an rt pull anyway. |
| 174 | * |
| 175 | * We only need to do a memory barrier if we updated |
| 176 | * the new priority vec. |
| 177 | */ |
| 178 | if (do_mb) |
Peter Zijlstra | 4e857c5 | 2014-03-17 18:06:10 +0100 | [diff] [blame] | 179 | smp_mb__after_atomic(); |
Steven Rostedt | d473750 | 2011-08-05 08:27:49 -0400 | [diff] [blame] | 180 | |
| 181 | /* |
Steven Rostedt | c92211d | 2011-08-02 16:36:12 -0400 | [diff] [blame] | 182 | * When removing from the vector, we decrement the counter first |
| 183 | * do a memory barrier and then clear the mask. |
| 184 | */ |
| 185 | atomic_dec(&(vec)->count); |
Peter Zijlstra | 4e857c5 | 2014-03-17 18:06:10 +0100 | [diff] [blame] | 186 | smp_mb__after_atomic(); |
Steven Rostedt | c3a2ae3 | 2009-07-29 00:21:23 -0400 | [diff] [blame] | 187 | cpumask_clear_cpu(cpu, vec->mask); |
Steven Rostedt | c3a2ae3 | 2009-07-29 00:21:23 -0400 | [diff] [blame] | 188 | } |
Gregory Haskins | 6e0534f | 2008-05-12 21:21:01 +0200 | [diff] [blame] | 189 | |
| 190 | *currpri = newpri; |
| 191 | } |
| 192 | |
| 193 | /** |
| 194 | * cpupri_init - initialize the cpupri structure |
| 195 | * @cp: The cpupri context |
| 196 | * |
Yacine Belkadi | e69f618 | 2013-07-12 20:45:47 +0200 | [diff] [blame] | 197 | * Return: -ENOMEM on memory allocation failure. |
Gregory Haskins | 6e0534f | 2008-05-12 21:21:01 +0200 | [diff] [blame] | 198 | */ |
Pekka Enberg | 68c38fc | 2010-07-15 23:18:22 +0300 | [diff] [blame] | 199 | int cpupri_init(struct cpupri *cp) |
Gregory Haskins | 6e0534f | 2008-05-12 21:21:01 +0200 | [diff] [blame] | 200 | { |
| 201 | int i; |
| 202 | |
Gregory Haskins | 6e0534f | 2008-05-12 21:21:01 +0200 | [diff] [blame] | 203 | for (i = 0; i < CPUPRI_NR_PRIORITIES; i++) { |
| 204 | struct cpupri_vec *vec = &cp->pri_to_cpu[i]; |
| 205 | |
Steven Rostedt | c92211d | 2011-08-02 16:36:12 -0400 | [diff] [blame] | 206 | atomic_set(&vec->count, 0); |
Pekka Enberg | 68c38fc | 2010-07-15 23:18:22 +0300 | [diff] [blame] | 207 | if (!zalloc_cpumask_var(&vec->mask, GFP_KERNEL)) |
Rusty Russell | 68e7456 | 2008-11-25 02:35:13 +1030 | [diff] [blame] | 208 | goto cleanup; |
Gregory Haskins | 6e0534f | 2008-05-12 21:21:01 +0200 | [diff] [blame] | 209 | } |
| 210 | |
Peter Zijlstra | 4dac0b6 | 2014-05-14 16:04:26 +0200 | [diff] [blame] | 211 | cp->cpu_to_pri = kcalloc(nr_cpu_ids, sizeof(int), GFP_KERNEL); |
| 212 | if (!cp->cpu_to_pri) |
| 213 | goto cleanup; |
| 214 | |
Gregory Haskins | 6e0534f | 2008-05-12 21:21:01 +0200 | [diff] [blame] | 215 | for_each_possible_cpu(i) |
| 216 | cp->cpu_to_pri[i] = CPUPRI_INVALID; |
Peter Zijlstra | 4dac0b6 | 2014-05-14 16:04:26 +0200 | [diff] [blame] | 217 | |
Rusty Russell | 68e7456 | 2008-11-25 02:35:13 +1030 | [diff] [blame] | 218 | return 0; |
| 219 | |
| 220 | cleanup: |
| 221 | for (i--; i >= 0; i--) |
| 222 | free_cpumask_var(cp->pri_to_cpu[i].mask); |
| 223 | return -ENOMEM; |
Gregory Haskins | 6e0534f | 2008-05-12 21:21:01 +0200 | [diff] [blame] | 224 | } |
| 225 | |
Rusty Russell | 68e7456 | 2008-11-25 02:35:13 +1030 | [diff] [blame] | 226 | /** |
| 227 | * cpupri_cleanup - clean up the cpupri structure |
| 228 | * @cp: The cpupri context |
| 229 | */ |
| 230 | void cpupri_cleanup(struct cpupri *cp) |
| 231 | { |
| 232 | int i; |
Gregory Haskins | 6e0534f | 2008-05-12 21:21:01 +0200 | [diff] [blame] | 233 | |
Peter Zijlstra | 4dac0b6 | 2014-05-14 16:04:26 +0200 | [diff] [blame] | 234 | kfree(cp->cpu_to_pri); |
Rusty Russell | 68e7456 | 2008-11-25 02:35:13 +1030 | [diff] [blame] | 235 | for (i = 0; i < CPUPRI_NR_PRIORITIES; i++) |
| 236 | free_cpumask_var(cp->pri_to_cpu[i].mask); |
| 237 | } |