~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/tools/sched_ext/scx_central.bpf.c

Version: ~ [ linux-6.12-rc7 ] ~ [ linux-6.11.7 ] ~ [ linux-6.10.14 ] ~ [ linux-6.9.12 ] ~ [ linux-6.8.12 ] ~ [ linux-6.7.12 ] ~ [ linux-6.6.60 ] ~ [ linux-6.5.13 ] ~ [ linux-6.4.16 ] ~ [ linux-6.3.13 ] ~ [ linux-6.2.16 ] ~ [ linux-6.1.116 ] ~ [ linux-6.0.19 ] ~ [ linux-5.19.17 ] ~ [ linux-5.18.19 ] ~ [ linux-5.17.15 ] ~ [ linux-5.16.20 ] ~ [ linux-5.15.171 ] ~ [ linux-5.14.21 ] ~ [ linux-5.13.19 ] ~ [ linux-5.12.19 ] ~ [ linux-5.11.22 ] ~ [ linux-5.10.229 ] ~ [ linux-5.9.16 ] ~ [ linux-5.8.18 ] ~ [ linux-5.7.19 ] ~ [ linux-5.6.19 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.285 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.323 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.336 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.337 ] ~ [ linux-4.4.302 ] ~ [ linux-3.10.108 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.12 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

Diff markup

Differences between /tools/sched_ext/scx_central.bpf.c (Architecture sparc) and /tools/sched_ext/scx_central.bpf.c (Architecture i386)


  1 /* SPDX-License-Identifier: GPL-2.0 */              1 /* SPDX-License-Identifier: GPL-2.0 */
  2 /*                                                  2 /*
  3  * A central FIFO sched_ext scheduler which de      3  * A central FIFO sched_ext scheduler which demonstrates the followings:
  4  *                                                  4  *
  5  * a. Making all scheduling decisions from one      5  * a. Making all scheduling decisions from one CPU:
  6  *                                                  6  *
  7  *    The central CPU is the only one making s      7  *    The central CPU is the only one making scheduling decisions. All other
  8  *    CPUs kick the central CPU when they run       8  *    CPUs kick the central CPU when they run out of tasks to run.
  9  *                                                  9  *
 10  *    There is one global BPF queue and the ce     10  *    There is one global BPF queue and the central CPU schedules all CPUs by
 11  *    dispatching from the global queue to eac     11  *    dispatching from the global queue to each CPU's local dsq from dispatch().
 12  *    This isn't the most straightforward. e.g     12  *    This isn't the most straightforward. e.g. It'd be easier to bounce
 13  *    through per-CPU BPF queues. The current      13  *    through per-CPU BPF queues. The current design is chosen to maximally
 14  *    utilize and verify various SCX mechanism     14  *    utilize and verify various SCX mechanisms such as LOCAL_ON dispatching.
 15  *                                                 15  *
 16  * b. Tickless operation                           16  * b. Tickless operation
 17  *                                                 17  *
 18  *    All tasks are dispatched with the infini     18  *    All tasks are dispatched with the infinite slice which allows stopping the
 19  *    ticks on CONFIG_NO_HZ_FULL kernels runni     19  *    ticks on CONFIG_NO_HZ_FULL kernels running with the proper nohz_full
 20  *    parameter. The tickless operation can be     20  *    parameter. The tickless operation can be observed through
 21  *    /proc/interrupts.                            21  *    /proc/interrupts.
 22  *                                                 22  *
 23  *    Periodic switching is enforced by a peri     23  *    Periodic switching is enforced by a periodic timer checking all CPUs and
 24  *    preempting them as necessary. Unfortunat     24  *    preempting them as necessary. Unfortunately, BPF timer currently doesn't
 25  *    have a way to pin to a specific CPU, so      25  *    have a way to pin to a specific CPU, so the periodic timer isn't pinned to
 26  *    the central CPU.                             26  *    the central CPU.
 27  *                                                 27  *
 28  * c. Preemption                                   28  * c. Preemption
 29  *                                                 29  *
 30  *    Kthreads are unconditionally queued to t     30  *    Kthreads are unconditionally queued to the head of a matching local dsq
 31  *    and dispatched with SCX_DSQ_PREEMPT. Thi     31  *    and dispatched with SCX_DSQ_PREEMPT. This ensures that a kthread is always
 32  *    prioritized over user threads, which is      32  *    prioritized over user threads, which is required for ensuring forward
 33  *    progress as e.g. the periodic timer may      33  *    progress as e.g. the periodic timer may run on a ksoftirqd and if the
 34  *    ksoftirqd gets starved by a user thread,     34  *    ksoftirqd gets starved by a user thread, there may not be anything else to
 35  *    vacate that user thread.                     35  *    vacate that user thread.
 36  *                                                 36  *
 37  *    SCX_KICK_PREEMPT is used to trigger sche     37  *    SCX_KICK_PREEMPT is used to trigger scheduling and CPUs to move to the
 38  *    next tasks.                                  38  *    next tasks.
 39  *                                                 39  *
 40  * This scheduler is designed to maximize usag     40  * This scheduler is designed to maximize usage of various SCX mechanisms. A
 41  * more practical implementation would likely      41  * more practical implementation would likely put the scheduling loop outside
 42  * the central CPU's dispatch() path and add s     42  * the central CPU's dispatch() path and add some form of priority mechanism.
 43  *                                                 43  *
 44  * Copyright (c) 2022 Meta Platforms, Inc. and     44  * Copyright (c) 2022 Meta Platforms, Inc. and affiliates.
 45  * Copyright (c) 2022 Tejun Heo <tj@kernel.org     45  * Copyright (c) 2022 Tejun Heo <tj@kernel.org>
 46  * Copyright (c) 2022 David Vernet <dvernet@me     46  * Copyright (c) 2022 David Vernet <dvernet@meta.com>
 47  */                                                47  */
 48 #include <scx/common.bpf.h>                        48 #include <scx/common.bpf.h>
 49                                                    49 
 50 char _license[] SEC("license") = "GPL";            50 char _license[] SEC("license") = "GPL";
 51                                                    51 
 52 enum {                                             52 enum {
 53         FALLBACK_DSQ_ID         = 0,               53         FALLBACK_DSQ_ID         = 0,
 54         MS_TO_NS                = 1000LLU * 10     54         MS_TO_NS                = 1000LLU * 1000,
 55         TIMER_INTERVAL_NS       = 1 * MS_TO_NS     55         TIMER_INTERVAL_NS       = 1 * MS_TO_NS,
 56 };                                                 56 };
 57                                                    57 
 58 const volatile s32 central_cpu;                    58 const volatile s32 central_cpu;
 59 const volatile u32 nr_cpu_ids = 1;      /* !0      59 const volatile u32 nr_cpu_ids = 1;      /* !0 for veristat, set during init */
 60 const volatile u64 slice_ns = SCX_SLICE_DFL;       60 const volatile u64 slice_ns = SCX_SLICE_DFL;
 61                                                    61 
 62 bool timer_pinned = true;                          62 bool timer_pinned = true;
 63 u64 nr_total, nr_locals, nr_queued, nr_lost_pi     63 u64 nr_total, nr_locals, nr_queued, nr_lost_pids;
 64 u64 nr_timers, nr_dispatches, nr_mismatches, n     64 u64 nr_timers, nr_dispatches, nr_mismatches, nr_retries;
 65 u64 nr_overflows;                                  65 u64 nr_overflows;
 66                                                    66 
 67 UEI_DEFINE(uei);                                   67 UEI_DEFINE(uei);
 68                                                    68 
 69 struct {                                           69 struct {
 70         __uint(type, BPF_MAP_TYPE_QUEUE);          70         __uint(type, BPF_MAP_TYPE_QUEUE);
 71         __uint(max_entries, 4096);                 71         __uint(max_entries, 4096);
 72         __type(value, s32);                        72         __type(value, s32);
 73 } central_q SEC(".maps");                          73 } central_q SEC(".maps");
 74                                                    74 
 75 /* can't use percpu map due to bad lookups */      75 /* can't use percpu map due to bad lookups */
 76 bool RESIZABLE_ARRAY(data, cpu_gimme_task);        76 bool RESIZABLE_ARRAY(data, cpu_gimme_task);
 77 u64 RESIZABLE_ARRAY(data, cpu_started_at);         77 u64 RESIZABLE_ARRAY(data, cpu_started_at);
 78                                                    78 
 79 struct central_timer {                             79 struct central_timer {
 80         struct bpf_timer timer;                    80         struct bpf_timer timer;
 81 };                                                 81 };
 82                                                    82 
 83 struct {                                           83 struct {
 84         __uint(type, BPF_MAP_TYPE_ARRAY);          84         __uint(type, BPF_MAP_TYPE_ARRAY);
 85         __uint(max_entries, 1);                    85         __uint(max_entries, 1);
 86         __type(key, u32);                          86         __type(key, u32);
 87         __type(value, struct central_timer);       87         __type(value, struct central_timer);
 88 } central_timer SEC(".maps");                      88 } central_timer SEC(".maps");
 89                                                    89 
 90 static bool vtime_before(u64 a, u64 b)             90 static bool vtime_before(u64 a, u64 b)
 91 {                                                  91 {
 92         return (s64)(a - b) < 0;                   92         return (s64)(a - b) < 0;
 93 }                                                  93 }
 94                                                    94 
 95 s32 BPF_STRUCT_OPS(central_select_cpu, struct      95 s32 BPF_STRUCT_OPS(central_select_cpu, struct task_struct *p,
 96                    s32 prev_cpu, u64 wake_flag     96                    s32 prev_cpu, u64 wake_flags)
 97 {                                                  97 {
 98         /*                                         98         /*
 99          * Steer wakeups to the central CPU as     99          * Steer wakeups to the central CPU as much as possible to avoid
100          * disturbing other CPUs. It's safe to    100          * disturbing other CPUs. It's safe to blindly return the central cpu as
101          * select_cpu() is a hint and if @p ca    101          * select_cpu() is a hint and if @p can't be on it, the kernel will
102          * automatically pick a fallback CPU.     102          * automatically pick a fallback CPU.
103          */                                       103          */
104         return central_cpu;                       104         return central_cpu;
105 }                                                 105 }
106                                                   106 
107 void BPF_STRUCT_OPS(central_enqueue, struct ta    107 void BPF_STRUCT_OPS(central_enqueue, struct task_struct *p, u64 enq_flags)
108 {                                                 108 {
109         s32 pid = p->pid;                         109         s32 pid = p->pid;
110                                                   110 
111         __sync_fetch_and_add(&nr_total, 1);       111         __sync_fetch_and_add(&nr_total, 1);
112                                                   112 
113         /*                                        113         /*
114          * Push per-cpu kthreads at the head o    114          * Push per-cpu kthreads at the head of local dsq's and preempt the
115          * corresponding CPU. This ensures tha    115          * corresponding CPU. This ensures that e.g. ksoftirqd isn't blocked
116          * behind other threads which is neces    116          * behind other threads which is necessary for forward progress
117          * guarantee as we depend on the BPF t    117          * guarantee as we depend on the BPF timer which may run from ksoftirqd.
118          */                                       118          */
119         if ((p->flags & PF_KTHREAD) && p->nr_c    119         if ((p->flags & PF_KTHREAD) && p->nr_cpus_allowed == 1) {
120                 __sync_fetch_and_add(&nr_local    120                 __sync_fetch_and_add(&nr_locals, 1);
121                 scx_bpf_dispatch(p, SCX_DSQ_LO    121                 scx_bpf_dispatch(p, SCX_DSQ_LOCAL, SCX_SLICE_INF,
122                                  enq_flags | S    122                                  enq_flags | SCX_ENQ_PREEMPT);
123                 return;                           123                 return;
124         }                                         124         }
125                                                   125 
126         if (bpf_map_push_elem(&central_q, &pid    126         if (bpf_map_push_elem(&central_q, &pid, 0)) {
127                 __sync_fetch_and_add(&nr_overf    127                 __sync_fetch_and_add(&nr_overflows, 1);
128                 scx_bpf_dispatch(p, FALLBACK_D    128                 scx_bpf_dispatch(p, FALLBACK_DSQ_ID, SCX_SLICE_INF, enq_flags);
129                 return;                           129                 return;
130         }                                         130         }
131                                                   131 
132         __sync_fetch_and_add(&nr_queued, 1);      132         __sync_fetch_and_add(&nr_queued, 1);
133                                                   133 
134         if (!scx_bpf_task_running(p))             134         if (!scx_bpf_task_running(p))
135                 scx_bpf_kick_cpu(central_cpu,     135                 scx_bpf_kick_cpu(central_cpu, SCX_KICK_PREEMPT);
136 }                                                 136 }
137                                                   137 
138 static bool dispatch_to_cpu(s32 cpu)              138 static bool dispatch_to_cpu(s32 cpu)
139 {                                                 139 {
140         struct task_struct *p;                    140         struct task_struct *p;
141         s32 pid;                                  141         s32 pid;
142                                                   142 
143         bpf_repeat(BPF_MAX_LOOPS) {               143         bpf_repeat(BPF_MAX_LOOPS) {
144                 if (bpf_map_pop_elem(&central_    144                 if (bpf_map_pop_elem(&central_q, &pid))
145                         break;                    145                         break;
146                                                   146 
147                 __sync_fetch_and_sub(&nr_queue    147                 __sync_fetch_and_sub(&nr_queued, 1);
148                                                   148 
149                 p = bpf_task_from_pid(pid);       149                 p = bpf_task_from_pid(pid);
150                 if (!p) {                         150                 if (!p) {
151                         __sync_fetch_and_add(&    151                         __sync_fetch_and_add(&nr_lost_pids, 1);
152                         continue;                 152                         continue;
153                 }                                 153                 }
154                                                   154 
155                 /*                                155                 /*
156                  * If we can't run the task at    156                  * If we can't run the task at the top, do the dumb thing and
157                  * bounce it to the fallback d    157                  * bounce it to the fallback dsq.
158                  */                               158                  */
159                 if (!bpf_cpumask_test_cpu(cpu,    159                 if (!bpf_cpumask_test_cpu(cpu, p->cpus_ptr)) {
160                         __sync_fetch_and_add(&    160                         __sync_fetch_and_add(&nr_mismatches, 1);
161                         scx_bpf_dispatch(p, FA    161                         scx_bpf_dispatch(p, FALLBACK_DSQ_ID, SCX_SLICE_INF, 0);
162                         bpf_task_release(p);      162                         bpf_task_release(p);
163                         /*                        163                         /*
164                          * We might run out of    164                          * We might run out of dispatch buffer slots if we continue dispatching
165                          * to the fallback DSQ    165                          * to the fallback DSQ, without dispatching to the local DSQ of the
166                          * target CPU. In such    166                          * target CPU. In such a case, break the loop now as will fail the
167                          * next dispatch opera    167                          * next dispatch operation.
168                          */                       168                          */
169                         if (!scx_bpf_dispatch_    169                         if (!scx_bpf_dispatch_nr_slots())
170                                 break;            170                                 break;
171                         continue;                 171                         continue;
172                 }                                 172                 }
173                                                   173 
174                 /* dispatch to local and mark     174                 /* dispatch to local and mark that @cpu doesn't need more */
175                 scx_bpf_dispatch(p, SCX_DSQ_LO    175                 scx_bpf_dispatch(p, SCX_DSQ_LOCAL_ON | cpu, SCX_SLICE_INF, 0);
176                                                   176 
177                 if (cpu != central_cpu)           177                 if (cpu != central_cpu)
178                         scx_bpf_kick_cpu(cpu,     178                         scx_bpf_kick_cpu(cpu, SCX_KICK_IDLE);
179                                                   179 
180                 bpf_task_release(p);              180                 bpf_task_release(p);
181                 return true;                      181                 return true;
182         }                                         182         }
183                                                   183 
184         return false;                             184         return false;
185 }                                                 185 }
186                                                   186 
187 void BPF_STRUCT_OPS(central_dispatch, s32 cpu,    187 void BPF_STRUCT_OPS(central_dispatch, s32 cpu, struct task_struct *prev)
188 {                                                 188 {
189         if (cpu == central_cpu) {                 189         if (cpu == central_cpu) {
190                 /* dispatch for all other CPUs    190                 /* dispatch for all other CPUs first */
191                 __sync_fetch_and_add(&nr_dispa    191                 __sync_fetch_and_add(&nr_dispatches, 1);
192                                                   192 
193                 bpf_for(cpu, 0, nr_cpu_ids) {     193                 bpf_for(cpu, 0, nr_cpu_ids) {
194                         bool *gimme;              194                         bool *gimme;
195                                                   195 
196                         if (!scx_bpf_dispatch_    196                         if (!scx_bpf_dispatch_nr_slots())
197                                 break;            197                                 break;
198                                                   198 
199                         /* central's gimme is     199                         /* central's gimme is never set */
200                         gimme = ARRAY_ELEM_PTR    200                         gimme = ARRAY_ELEM_PTR(cpu_gimme_task, cpu, nr_cpu_ids);
201                         if (!gimme || !*gimme)    201                         if (!gimme || !*gimme)
202                                 continue;         202                                 continue;
203                                                   203 
204                         if (dispatch_to_cpu(cp    204                         if (dispatch_to_cpu(cpu))
205                                 *gimme = false    205                                 *gimme = false;
206                 }                                 206                 }
207                                                   207 
208                 /*                                208                 /*
209                  * Retry if we ran out of disp    209                  * Retry if we ran out of dispatch buffer slots as we might have
210                  * skipped some CPUs and also     210                  * skipped some CPUs and also need to dispatch for self. The ext
211                  * core automatically retries     211                  * core automatically retries if the local dsq is empty but we
212                  * can't rely on that as we're    212                  * can't rely on that as we're dispatching for other CPUs too.
213                  * Kick self explicitly to ret    213                  * Kick self explicitly to retry.
214                  */                               214                  */
215                 if (!scx_bpf_dispatch_nr_slots    215                 if (!scx_bpf_dispatch_nr_slots()) {
216                         __sync_fetch_and_add(&    216                         __sync_fetch_and_add(&nr_retries, 1);
217                         scx_bpf_kick_cpu(centr    217                         scx_bpf_kick_cpu(central_cpu, SCX_KICK_PREEMPT);
218                         return;                   218                         return;
219                 }                                 219                 }
220                                                   220 
221                 /* look for a task to run on t    221                 /* look for a task to run on the central CPU */
222                 if (scx_bpf_consume(FALLBACK_D    222                 if (scx_bpf_consume(FALLBACK_DSQ_ID))
223                         return;                   223                         return;
224                 dispatch_to_cpu(central_cpu);     224                 dispatch_to_cpu(central_cpu);
225         } else {                                  225         } else {
226                 bool *gimme;                      226                 bool *gimme;
227                                                   227 
228                 if (scx_bpf_consume(FALLBACK_D    228                 if (scx_bpf_consume(FALLBACK_DSQ_ID))
229                         return;                   229                         return;
230                                                   230 
231                 gimme = ARRAY_ELEM_PTR(cpu_gim    231                 gimme = ARRAY_ELEM_PTR(cpu_gimme_task, cpu, nr_cpu_ids);
232                 if (gimme)                        232                 if (gimme)
233                         *gimme = true;            233                         *gimme = true;
234                                                   234 
235                 /*                                235                 /*
236                  * Force dispatch on the sched    236                  * Force dispatch on the scheduling CPU so that it finds a task
237                  * to run for us.                 237                  * to run for us.
238                  */                               238                  */
239                 scx_bpf_kick_cpu(central_cpu,     239                 scx_bpf_kick_cpu(central_cpu, SCX_KICK_PREEMPT);
240         }                                         240         }
241 }                                                 241 }
242                                                   242 
243 void BPF_STRUCT_OPS(central_running, struct ta    243 void BPF_STRUCT_OPS(central_running, struct task_struct *p)
244 {                                                 244 {
245         s32 cpu = scx_bpf_task_cpu(p);            245         s32 cpu = scx_bpf_task_cpu(p);
246         u64 *started_at = ARRAY_ELEM_PTR(cpu_s    246         u64 *started_at = ARRAY_ELEM_PTR(cpu_started_at, cpu, nr_cpu_ids);
247         if (started_at)                           247         if (started_at)
248                 *started_at = bpf_ktime_get_ns    248                 *started_at = bpf_ktime_get_ns() ?: 1;  /* 0 indicates idle */
249 }                                                 249 }
250                                                   250 
251 void BPF_STRUCT_OPS(central_stopping, struct t    251 void BPF_STRUCT_OPS(central_stopping, struct task_struct *p, bool runnable)
252 {                                                 252 {
253         s32 cpu = scx_bpf_task_cpu(p);            253         s32 cpu = scx_bpf_task_cpu(p);
254         u64 *started_at = ARRAY_ELEM_PTR(cpu_s    254         u64 *started_at = ARRAY_ELEM_PTR(cpu_started_at, cpu, nr_cpu_ids);
255         if (started_at)                           255         if (started_at)
256                 *started_at = 0;                  256                 *started_at = 0;
257 }                                                 257 }
258                                                   258 
259 static int central_timerfn(void *map, int *key    259 static int central_timerfn(void *map, int *key, struct bpf_timer *timer)
260 {                                                 260 {
261         u64 now = bpf_ktime_get_ns();             261         u64 now = bpf_ktime_get_ns();
262         u64 nr_to_kick = nr_queued;               262         u64 nr_to_kick = nr_queued;
263         s32 i, curr_cpu;                          263         s32 i, curr_cpu;
264                                                   264 
265         curr_cpu = bpf_get_smp_processor_id();    265         curr_cpu = bpf_get_smp_processor_id();
266         if (timer_pinned && (curr_cpu != centr    266         if (timer_pinned && (curr_cpu != central_cpu)) {
267                 scx_bpf_error("Central timer r    267                 scx_bpf_error("Central timer ran on CPU %d, not central CPU %d",
268                               curr_cpu, centra    268                               curr_cpu, central_cpu);
269                 return 0;                         269                 return 0;
270         }                                         270         }
271                                                   271 
272         bpf_for(i, 0, nr_cpu_ids) {               272         bpf_for(i, 0, nr_cpu_ids) {
273                 s32 cpu = (nr_timers + i) % nr    273                 s32 cpu = (nr_timers + i) % nr_cpu_ids;
274                 u64 *started_at;                  274                 u64 *started_at;
275                                                   275 
276                 if (cpu == central_cpu)           276                 if (cpu == central_cpu)
277                         continue;                 277                         continue;
278                                                   278 
279                 /* kick iff the current one ex    279                 /* kick iff the current one exhausted its slice */
280                 started_at = ARRAY_ELEM_PTR(cp    280                 started_at = ARRAY_ELEM_PTR(cpu_started_at, cpu, nr_cpu_ids);
281                 if (started_at && *started_at     281                 if (started_at && *started_at &&
282                     vtime_before(now, *started    282                     vtime_before(now, *started_at + slice_ns))
283                         continue;                 283                         continue;
284                                                   284 
285                 /* and there's something pendi    285                 /* and there's something pending */
286                 if (scx_bpf_dsq_nr_queued(FALL    286                 if (scx_bpf_dsq_nr_queued(FALLBACK_DSQ_ID) ||
287                     scx_bpf_dsq_nr_queued(SCX_    287                     scx_bpf_dsq_nr_queued(SCX_DSQ_LOCAL_ON | cpu))
288                         ;                         288                         ;
289                 else if (nr_to_kick)              289                 else if (nr_to_kick)
290                         nr_to_kick--;             290                         nr_to_kick--;
291                 else                              291                 else
292                         continue;                 292                         continue;
293                                                   293 
294                 scx_bpf_kick_cpu(cpu, SCX_KICK    294                 scx_bpf_kick_cpu(cpu, SCX_KICK_PREEMPT);
295         }                                         295         }
296                                                   296 
297         bpf_timer_start(timer, TIMER_INTERVAL_    297         bpf_timer_start(timer, TIMER_INTERVAL_NS, BPF_F_TIMER_CPU_PIN);
298         __sync_fetch_and_add(&nr_timers, 1);      298         __sync_fetch_and_add(&nr_timers, 1);
299         return 0;                                 299         return 0;
300 }                                                 300 }
301                                                   301 
302 int BPF_STRUCT_OPS_SLEEPABLE(central_init)        302 int BPF_STRUCT_OPS_SLEEPABLE(central_init)
303 {                                                 303 {
304         u32 key = 0;                              304         u32 key = 0;
305         struct bpf_timer *timer;                  305         struct bpf_timer *timer;
306         int ret;                                  306         int ret;
307                                                   307 
308         ret = scx_bpf_create_dsq(FALLBACK_DSQ_    308         ret = scx_bpf_create_dsq(FALLBACK_DSQ_ID, -1);
309         if (ret)                                  309         if (ret)
310                 return ret;                       310                 return ret;
311                                                   311 
312         timer = bpf_map_lookup_elem(&central_t    312         timer = bpf_map_lookup_elem(&central_timer, &key);
313         if (!timer)                               313         if (!timer)
314                 return -ESRCH;                    314                 return -ESRCH;
315                                                   315 
316         if (bpf_get_smp_processor_id() != cent    316         if (bpf_get_smp_processor_id() != central_cpu) {
317                 scx_bpf_error("init from non-c    317                 scx_bpf_error("init from non-central CPU");
318                 return -EINVAL;                   318                 return -EINVAL;
319         }                                         319         }
320                                                   320 
321         bpf_timer_init(timer, &central_timer,     321         bpf_timer_init(timer, &central_timer, CLOCK_MONOTONIC);
322         bpf_timer_set_callback(timer, central_    322         bpf_timer_set_callback(timer, central_timerfn);
323                                                   323 
324         ret = bpf_timer_start(timer, TIMER_INT    324         ret = bpf_timer_start(timer, TIMER_INTERVAL_NS, BPF_F_TIMER_CPU_PIN);
325         /*                                        325         /*
326          * BPF_F_TIMER_CPU_PIN is pretty new (    326          * BPF_F_TIMER_CPU_PIN is pretty new (>=6.7). If we're running in a
327          * kernel which doesn't have it, bpf_t    327          * kernel which doesn't have it, bpf_timer_start() will return -EINVAL.
328          * Retry without the PIN. This would b    328          * Retry without the PIN. This would be the perfect use case for
329          * bpf_core_enum_value_exists() but th    329          * bpf_core_enum_value_exists() but the enum type doesn't have a name
330          * and can't be used with bpf_core_enu    330          * and can't be used with bpf_core_enum_value_exists(). Oh well...
331          */                                       331          */
332         if (ret == -EINVAL) {                     332         if (ret == -EINVAL) {
333                 timer_pinned = false;             333                 timer_pinned = false;
334                 ret = bpf_timer_start(timer, T    334                 ret = bpf_timer_start(timer, TIMER_INTERVAL_NS, 0);
335         }                                         335         }
336         if (ret)                                  336         if (ret)
337                 scx_bpf_error("bpf_timer_start    337                 scx_bpf_error("bpf_timer_start failed (%d)", ret);
338         return ret;                               338         return ret;
339 }                                                 339 }
340                                                   340 
341 void BPF_STRUCT_OPS(central_exit, struct scx_e    341 void BPF_STRUCT_OPS(central_exit, struct scx_exit_info *ei)
342 {                                                 342 {
343         UEI_RECORD(uei, ei);                      343         UEI_RECORD(uei, ei);
344 }                                                 344 }
345                                                   345 
346 SCX_OPS_DEFINE(central_ops,                       346 SCX_OPS_DEFINE(central_ops,
347                /*                                 347                /*
348                 * We are offloading all schedu    348                 * We are offloading all scheduling decisions to the central CPU
349                 * and thus being the last task    349                 * and thus being the last task on a given CPU doesn't mean
350                 * anything special. Enqueue th    350                 * anything special. Enqueue the last tasks like any other tasks.
351                 */                                351                 */
352                .flags                   = SCX_    352                .flags                   = SCX_OPS_ENQ_LAST,
353                                                   353 
354                .select_cpu              = (voi    354                .select_cpu              = (void *)central_select_cpu,
355                .enqueue                 = (voi    355                .enqueue                 = (void *)central_enqueue,
356                .dispatch                = (voi    356                .dispatch                = (void *)central_dispatch,
357                .running                 = (voi    357                .running                 = (void *)central_running,
358                .stopping                = (voi    358                .stopping                = (void *)central_stopping,
359                .init                    = (voi    359                .init                    = (void *)central_init,
360                .exit                    = (voi    360                .exit                    = (void *)central_exit,
361                .name                    = "cen    361                .name                    = "central");
362                                                   362 

~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

kernel.org | git.kernel.org | LWN.net | Project Home | SVN repository | Mail admin

Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.

sflogo.php