We currently have two invocations, which is unnecessary. Fetch it only once and use the smp_processor_id() variant, so we also get preemption checks along with it when DEBUG_PREEMPT is set.
Unified · Split
| @@ -233,6 +233,7 @@ static u64 bpf_perf_event_output(u64 r1, u64 r2, u64 flags, u64 r4, u64 size) · 232 unmodified lines | |||
| 233 | struct pt_regs *regs = (struct pt_regs *) (long) r1; | 233 | struct pt_regs *regs = (struct pt_regs *) (long) r1; |
| 234 | struct bpf_map *map = (struct bpf_map *) (long) r2; | 234 | struct bpf_map *map = (struct bpf_map *) (long) r2; |
| 235 | struct bpf_array *array = container_of(map, struct bpf_array, map); | 235 | struct bpf_array *array = container_of(map, struct bpf_array, map); |
| 236 | unsigned int cpu = smp_processor_id(); | ||
| 236 | u64 index = flags & BPF_F_INDEX_MASK; | 237 | u64 index = flags & BPF_F_INDEX_MASK; |
| 237 | void *data = (void *) (long) r4; | 238 | void *data = (void *) (long) r4; |
| 238 | struct perf_sample_data sample_data; | 239 | struct perf_sample_data sample_data; |
| @@ -246,7 +247,7 @@ static u64 bpf_perf_event_output(u64 r1, u64 r2, u64 flags, u64 r4, u64 size) · 7 unmodified lines | |||
| 246 | if (unlikely(flags & ~(BPF_F_INDEX_MASK))) | 247 | if (unlikely(flags & ~(BPF_F_INDEX_MASK))) |
| 247 | return -EINVAL; | 248 | return -EINVAL; |
| 248 | if (index == BPF_F_CURRENT_CPU) | 249 | if (index == BPF_F_CURRENT_CPU) |
| 249 | index = raw_smp_processor_id(); | 250 | index = cpu; |
| 250 | if (unlikely(index >= array->map.max_entries)) | 251 | if (unlikely(index >= array->map.max_entries)) |
| 251 | return -E2BIG; | 252 | return -E2BIG; |
| 252 | 253 | ||
| @@ -259,7 +260,7 @@ static u64 bpf_perf_event_output(u64 r1, u64 r2, u64 flags, u64 r4, u64 size) · 6 unmodified lines | |||
| 259 | event->attr.config != PERF_COUNT_SW_BPF_OUTPUT)) | 260 | event->attr.config != PERF_COUNT_SW_BPF_OUTPUT)) |
| 260 | return -EINVAL; | 261 | return -EINVAL; |
| 261 | 262 | ||
| 262 | if (unlikely(event->oncpu != smp_processor_id())) | 263 | if (unlikely(event->oncpu != cpu)) |
| 263 | return -EOPNOTSUPP; | 264 | return -EOPNOTSUPP; |
| 264 | 265 | ||
| 265 | perf_sample_data_init(&sample_data, 0, 0); | 266 | perf_sample_data_init(&sample_data, 0, 0); |