In various performance profiles of kernels with BPF programs attached,
bpf_local_storage_lookup() appears as a significant portion of CPU
cycles spent. To enable the compiler generate more optimal code, turn
bpf_local_storage_lookup() into a static inline function, where only the
cache insertion code path is outlined
Notably, outlining cache insertion helps avoid bloating callers by
duplicating setting up calls to raw_spin_{lock,unlock}_irqsave() (on
architectures which do not inline spin_lock/unlock, such as x86), which
would cause the compiler produce worse code by deciding to outline
otherwise inlinable functions. The call overhead is neutral, because we
make 2 calls either way: either calling raw_spin_lock_irqsave() and
raw_spin_unlock_irqsave(); or call __bpf_local_storage_insert_cache(),
which calls raw_spin_lock_irqsave(), followed by a tail-call to
raw_spin_unlock_irqsave() where the compiler can perform TCO and (in
optimized uninstrumented builds) turns it into a plain jump. The call to
__bpf_local_storage_insert_cache() can be elided entirely if
cacheit_lockit is a false constant expression.
Based on results from './benchs/run_bench_local_storage.sh' (21 trials,
reboot between each trial; x86 defconfig + BPF, clang 16) this produces
improvements in throughput and latency in the majority of cases, with an
average (geomean) improvement of 8%:
+---- Hashmap Control --------------------
|
| + num keys: 10
| : <before> | <after>
| +-+ hashmap (control) sequential get +----------------------+----------------------
| +- hits throughput | 14.789 M ops/s | 14.745 M ops/s ( ~ )
| +- hits latency | 67.679 ns/op | 67.879 ns/op ( ~ )
| +- important_hits throughput | 14.789 M ops/s | 14.745 M ops/s ( ~ )
|
| + num keys: 1000
| : <before> | <after>
| +-+ hashmap (control) sequential get +----------------------+----------------------
| +- hits throughput | 12.233 M ops/s | 12.170 M ops/s ( ~ )
| +- hits latency | 81.754 ns/op | 82.185 ns/op ( ~ )
| +- important_hits throughput | 12.233 M ops/s | 12.170 M ops/s ( ~ )
|
| + num keys: 10000
| : <before> | <after>
| +-+ hashmap (control) sequential get +----------------------+----------------------
| +- hits throughput | 7.220 M ops/s | 7.204 M ops/s ( ~ )
| +- hits latency | 138.522 ns/op | 138.842 ns/op ( ~ )
| +- important_hits throughput | 7.220 M ops/s | 7.204 M ops/s ( ~ )
|
| + num keys: 100000
| : <before> | <after>
| +-+ hashmap (control) sequential get +----------------------+----------------------
| +- hits throughput | 5.061 M ops/s | 5.165 M ops/s (+2.1%)
| +- hits latency | 198.483 ns/op | 194.270 ns/op (-2.1%)
| +- important_hits throughput | 5.061 M ops/s | 5.165 M ops/s (+2.1%)
|
| + num keys: 4194304
| : <before> | <after>
| +-+ hashmap (control) sequential get +----------------------+----------------------
| +- hits throughput | 2.864 M ops/s | 2.882 M ops/s ( ~ )
| +- hits latency | 365.220 ns/op | 361.418 ns/op (-1.0%)
| +- important_hits throughput | 2.864 M ops/s | 2.882 M ops/s ( ~ )
|
+---- Local Storage ----------------------
|
| + num_maps: 1
| : <before> | <after>
| +-+ local_storage cache sequential get +----------------------+----------------------
| +- hits throughput | 33.005 M ops/s | 39.068 M ops/s (+18.4%)
| +- hits latency | 30.300 ns/op | 25.598 ns/op (-15.5%)
| +- important_hits throughput | 33.005 M ops/s | 39.068 M ops/s (+18.4%)
| :
| : <before> | <after>
| +-+ local_storage cache interleaved get +----------------------+----------------------
| +- hits throughput | 37.151 M ops/s | 44.926 M ops/s (+20.9%)
| +- hits latency | 26.919 ns/op | 22.259 ns/op (-17.3%)
| +- important_hits throughput | 37.151 M ops/s | 44.926 M ops/s (+20.9%)
|
| + num_maps: 10
| : <before> | <after>
| +-+ local_storage cache sequential get +----------------------+----------------------
| +- hits throughput | 32.288 M ops/s | 38.099 M ops/s (+18.0%)
| +- hits latency | 30.972 ns/op | 26.248 ns/op (-15.3%)
| +- important_hits throughput | 3.229 M ops/s | 3.810 M ops/s (+18.0%)
| :
| : <before> | <after>
| +-+ local_storage cache interleaved get +----------------------+----------------------
| +- hits throughput | 34.473 M ops/s | 41.145 M ops/s (+19.4%)
| +- hits latency | 29.010 ns/op | 24.307 ns/op (-16.2%)
| +- important_hits throughput | 12.312 M ops/s | 14.695 M ops/s (+19.4%)
|
| + num_maps: 16
| : <before> | <after>
| +-+ local_storage cache sequential get +----------------------+----------------------
| +- hits throughput | 32.524 M ops/s | 38.341 M ops/s (+17.9%)
| +- hits latency | 30.748 ns/op | 26.083 ns/op (-15.2%)
| +- important_hits throughput | 2.033 M ops/s | 2.396 M ops/s (+17.9%)
| :
| : <before> | <after>
| +-+ local_storage cache interleaved get +----------------------+----------------------
| +- hits throughput | 34.575 M ops/s | 41.338 M ops/s (+19.6%)
| +- hits latency | 28.925 ns/op | 24.193 ns/op (-16.4%)
| +- important_hits throughput | 11.001 M ops/s | 13.153 M ops/s (+19.6%)
|
| + num_maps: 17
| : <before> | <after>
| +-+ local_storage cache sequential get +----------------------+----------------------
| +- hits throughput | 28.861 M ops/s | 32.756 M ops/s (+13.5%)
| +- hits latency | 34.649 ns/op | 30.530 ns/op (-11.9%)
| +- important_hits throughput | 1.700 M ops/s | 1.929 M ops/s (+13.5%)
| :
| : <before> | <after>
| +-+ local_storage cache interleaved get +----------------------+----------------------
| +- hits throughput | 31.529 M ops/s | 36.110 M ops/s (+14.5%)
| +- hits latency | 31.719 ns/op | 27.697 ns/op (-12.7%)
| +- important_hits throughput | 9.598 M ops/s | 10.993 M ops/s (+14.5%)
|
| + num_maps: 24
| : <before> | <after>
| +-+ local_storage cache sequential get +----------------------+----------------------
| +- hits throughput | 18.602 M ops/s | 19.937 M ops/s (+7.2%)
| +- hits latency | 53.767 ns/op | 50.166 ns/op (-6.7%)
| +- important_hits throughput | 0.776 M ops/s | 0.831 M ops/s (+7.2%)
| :
| : <before> | <after>
| +-+ local_storage cache interleaved get +----------------------+----------------------
| +- hits throughput | 21.718 M ops/s | 23.332 M ops/s (+7.4%)
| +- hits latency | 46.047 ns/op | 42.865 ns/op (-6.9%)
| +- important_hits throughput | 6.110 M ops/s | 6.564 M ops/s (+7.4%)
|
| + num_maps: 32
| : <before> | <after>
| +-+ local_storage cache sequential get +----------------------+----------------------
| +- hits throughput | 14.118 M ops/s | 14.626 M ops/s (+3.6%)
| +- hits latency | 70.856 ns/op | 68.381 ns/op (-3.5%)
| +- important_hits throughput | 0.442 M ops/s | 0.458 M ops/s (+3.6%)
| :
| : <before> | <after>
| +-+ local_storage cache interleaved get +----------------------+----------------------
| +- hits throughput | 17.111 M ops/s | 17.906 M ops/s (+4.6%)
| +- hits latency | 58.451 ns/op | 55.865 ns/op (-4.4%)
| +- important_hits throughput | 4.776 M ops/s | 4.998 M ops/s (+4.6%)
|
| + num_maps: 100
| : <before> | <after>
| +-+ local_storage cache sequential get +----------------------+----------------------
| +- hits throughput | 5.281 M ops/s | 5.528 M ops/s (+4.7%)
| +- hits latency | 192.398 ns/op | 183.059 ns/op (-4.9%)
| +- important_hits throughput | 0.053 M ops/s | 0.055 M ops/s (+4.9%)
| :
| : <before> | <after>
| +-+ local_storage cache interleaved get +----------------------+----------------------
| +- hits throughput | 6.265 M ops/s | 6.498 M ops/s (+3.7%)
| +- hits latency | 161.436 ns/op | 152.877 ns/op (-5.3%)
| +- important_hits throughput | 1.636 M ops/s | 1.697 M ops/s (+3.7%)
|
| + num_maps: 1000
| : <before> | <after>
| +-+ local_storage cache sequential get +----------------------+----------------------
| +- hits throughput | 0.355 M ops/s | 0.354 M ops/s ( ~ )
| +- hits latency | 2826.538 ns/op | 2827.139 ns/op ( ~ )
| +- important_hits throughput | 0.000 M ops/s | 0.000 M ops/s ( ~ )
| :
| : <before> | <after>
| +-+ local_storage cache interleaved get +----------------------+----------------------
| +- hits throughput | 0.404 M ops/s | 0.403 M ops/s ( ~ )
| +- hits latency | 2481.190 ns/op | 2487.555 ns/op ( ~ )
| +- important_hits throughput | 0.102 M ops/s | 0.101 M ops/s ( ~ )
Signed-off-by: Marco Elver <[email protected]>
Cc: Martin KaFai Lau <[email protected]>
---
v2:
* Inline most of bpf_local_storage_lookup(), which produces greater
speedup and avoids regressing the cases with large map arrays.
* Drop "likely()" hint, it didn't produce much benefit.
* Re-run benchmark and collect 21 trials of results.
* Remove the on_lookup tests, which no longer work because
e.g. bpf_task_storage_delete() no longer does calls to
bpf_local_storage_*() helpers.
---
include/linux/bpf_local_storage.h | 30 ++++++++++-
kernel/bpf/bpf_local_storage.c | 52 +++++--------------
.../bpf/prog_tests/task_local_storage.c | 6 ---
.../selftests/bpf/progs/cgrp_ls_recursion.c | 26 ----------
.../selftests/bpf/progs/task_ls_recursion.c | 17 ------
5 files changed, 41 insertions(+), 90 deletions(-)
diff --git a/include/linux/bpf_local_storage.h b/include/linux/bpf_local_storage.h
index 173ec7f43ed1..dcddb0aef7d8 100644
--- a/include/linux/bpf_local_storage.h
+++ b/include/linux/bpf_local_storage.h
@@ -129,10 +129,36 @@ bpf_local_storage_map_alloc(union bpf_attr *attr,
struct bpf_local_storage_cache *cache,
bool bpf_ma);
-struct bpf_local_storage_data *
+void __bpf_local_storage_insert_cache(struct bpf_local_storage *local_storage,
+ struct bpf_local_storage_map *smap,
+ struct bpf_local_storage_elem *selem);
+/* If cacheit_lockit is false, this lookup function is lockless */
+static inline struct bpf_local_storage_data *
bpf_local_storage_lookup(struct bpf_local_storage *local_storage,
struct bpf_local_storage_map *smap,
- bool cacheit_lockit);
+ bool cacheit_lockit)
+{
+ struct bpf_local_storage_data *sdata;
+ struct bpf_local_storage_elem *selem;
+
+ /* Fast path (cache hit) */
+ sdata = rcu_dereference_check(local_storage->cache[smap->cache_idx],
+ bpf_rcu_lock_held());
+ if (sdata && rcu_access_pointer(sdata->smap) == smap)
+ return sdata;
+
+ /* Slow path (cache miss) */
+ hlist_for_each_entry_rcu(selem, &local_storage->list, snode,
+ rcu_read_lock_trace_held())
+ if (rcu_access_pointer(SDATA(selem)->smap) == smap)
+ break;
+
+ if (!selem)
+ return NULL;
+ if (cacheit_lockit)
+ __bpf_local_storage_insert_cache(local_storage, smap, selem);
+ return SDATA(selem);
+}
void bpf_local_storage_destroy(struct bpf_local_storage *local_storage);
diff --git a/kernel/bpf/bpf_local_storage.c b/kernel/bpf/bpf_local_storage.c
index 146824cc9689..bdea1a459153 100644
--- a/kernel/bpf/bpf_local_storage.c
+++ b/kernel/bpf/bpf_local_storage.c
@@ -414,47 +414,21 @@ void bpf_selem_unlink(struct bpf_local_storage_elem *selem, bool reuse_now)
bpf_selem_unlink_storage(selem, reuse_now);
}
-/* If cacheit_lockit is false, this lookup function is lockless */
-struct bpf_local_storage_data *
-bpf_local_storage_lookup(struct bpf_local_storage *local_storage,
- struct bpf_local_storage_map *smap,
- bool cacheit_lockit)
+void __bpf_local_storage_insert_cache(struct bpf_local_storage *local_storage,
+ struct bpf_local_storage_map *smap,
+ struct bpf_local_storage_elem *selem)
{
- struct bpf_local_storage_data *sdata;
- struct bpf_local_storage_elem *selem;
-
- /* Fast path (cache hit) */
- sdata = rcu_dereference_check(local_storage->cache[smap->cache_idx],
- bpf_rcu_lock_held());
- if (sdata && rcu_access_pointer(sdata->smap) == smap)
- return sdata;
-
- /* Slow path (cache miss) */
- hlist_for_each_entry_rcu(selem, &local_storage->list, snode,
- rcu_read_lock_trace_held())
- if (rcu_access_pointer(SDATA(selem)->smap) == smap)
- break;
-
- if (!selem)
- return NULL;
-
- sdata = SDATA(selem);
- if (cacheit_lockit) {
- unsigned long flags;
-
- /* spinlock is needed to avoid racing with the
- * parallel delete. Otherwise, publishing an already
- * deleted sdata to the cache will become a use-after-free
- * problem in the next bpf_local_storage_lookup().
- */
- raw_spin_lock_irqsave(&local_storage->lock, flags);
- if (selem_linked_to_storage(selem))
- rcu_assign_pointer(local_storage->cache[smap->cache_idx],
- sdata);
- raw_spin_unlock_irqrestore(&local_storage->lock, flags);
- }
+ unsigned long flags;
- return sdata;
+ /* spinlock is needed to avoid racing with the
+ * parallel delete. Otherwise, publishing an already
+ * deleted sdata to the cache will become a use-after-free
+ * problem in the next bpf_local_storage_lookup().
+ */
+ raw_spin_lock_irqsave(&local_storage->lock, flags);
+ if (selem_linked_to_storage(selem))
+ rcu_assign_pointer(local_storage->cache[smap->cache_idx], SDATA(selem));
+ raw_spin_unlock_irqrestore(&local_storage->lock, flags);
}
static int check_flags(const struct bpf_local_storage_data *old_sdata,
diff --git a/tools/testing/selftests/bpf/prog_tests/task_local_storage.c b/tools/testing/selftests/bpf/prog_tests/task_local_storage.c
index ea8537c54413..c33c05161a9e 100644
--- a/tools/testing/selftests/bpf/prog_tests/task_local_storage.c
+++ b/tools/testing/selftests/bpf/prog_tests/task_local_storage.c
@@ -117,12 +117,6 @@ static void test_recursion(void)
ASSERT_OK(err, "lookup map_b");
ASSERT_EQ(value, 100, "map_b value");
- prog_fd = bpf_program__fd(skel->progs.on_lookup);
- memset(&info, 0, sizeof(info));
- err = bpf_prog_get_info_by_fd(prog_fd, &info, &info_len);
- ASSERT_OK(err, "get prog info");
- ASSERT_GT(info.recursion_misses, 0, "on_lookup prog recursion");
-
prog_fd = bpf_program__fd(skel->progs.on_update);
memset(&info, 0, sizeof(info));
err = bpf_prog_get_info_by_fd(prog_fd, &info, &info_len);
diff --git a/tools/testing/selftests/bpf/progs/cgrp_ls_recursion.c b/tools/testing/selftests/bpf/progs/cgrp_ls_recursion.c
index 610c2427fd93..3500e4b69ebe 100644
--- a/tools/testing/selftests/bpf/progs/cgrp_ls_recursion.c
+++ b/tools/testing/selftests/bpf/progs/cgrp_ls_recursion.c
@@ -27,32 +27,6 @@ bool is_cgroup1 = 0;
struct cgroup *bpf_task_get_cgroup1(struct task_struct *task, int hierarchy_id) __ksym;
void bpf_cgroup_release(struct cgroup *cgrp) __ksym;
-static void __on_lookup(struct cgroup *cgrp)
-{
- bpf_cgrp_storage_delete(&map_a, cgrp);
- bpf_cgrp_storage_delete(&map_b, cgrp);
-}
-
-SEC("fentry/bpf_local_storage_lookup")
-int BPF_PROG(on_lookup)
-{
- struct task_struct *task = bpf_get_current_task_btf();
- struct cgroup *cgrp;
-
- if (is_cgroup1) {
- cgrp = bpf_task_get_cgroup1(task, target_hid);
- if (!cgrp)
- return 0;
-
- __on_lookup(cgrp);
- bpf_cgroup_release(cgrp);
- return 0;
- }
-
- __on_lookup(task->cgroups->dfl_cgrp);
- return 0;
-}
-
static void __on_update(struct cgroup *cgrp)
{
long *ptr;
diff --git a/tools/testing/selftests/bpf/progs/task_ls_recursion.c b/tools/testing/selftests/bpf/progs/task_ls_recursion.c
index 4542dc683b44..f1853c38aada 100644
--- a/tools/testing/selftests/bpf/progs/task_ls_recursion.c
+++ b/tools/testing/selftests/bpf/progs/task_ls_recursion.c
@@ -27,23 +27,6 @@ struct {
__type(value, long);
} map_b SEC(".maps");
-SEC("fentry/bpf_local_storage_lookup")
-int BPF_PROG(on_lookup)
-{
- struct task_struct *task = bpf_get_current_task_btf();
-
- if (!test_pid || task->pid != test_pid)
- return 0;
-
- /* The bpf_task_storage_delete will call
- * bpf_local_storage_lookup. The prog->active will
- * stop the recursion.
- */
- bpf_task_storage_delete(&map_a, task);
- bpf_task_storage_delete(&map_b, task);
- return 0;
-}
-
SEC("fentry/bpf_local_storage_update")
int BPF_PROG(on_update)
{
--
2.43.0.594.gd9cf4e227d-goog
On 2/7/24 4:26 AM, Marco Elver wrote:
> In various performance profiles of kernels with BPF programs attached,
> bpf_local_storage_lookup() appears as a significant portion of CPU
> cycles spent. To enable the compiler generate more optimal code, turn
> bpf_local_storage_lookup() into a static inline function, where only the
> cache insertion code path is outlined
>
> Notably, outlining cache insertion helps avoid bloating callers by
> duplicating setting up calls to raw_spin_{lock,unlock}_irqsave() (on
> architectures which do not inline spin_lock/unlock, such as x86), which
> would cause the compiler produce worse code by deciding to outline
> otherwise inlinable functions. The call overhead is neutral, because we
> make 2 calls either way: either calling raw_spin_lock_irqsave() and
> raw_spin_unlock_irqsave(); or call __bpf_local_storage_insert_cache(),
> which calls raw_spin_lock_irqsave(), followed by a tail-call to
> raw_spin_unlock_irqsave() where the compiler can perform TCO and (in
> optimized uninstrumented builds) turns it into a plain jump. The call to
> __bpf_local_storage_insert_cache() can be elided entirely if
> cacheit_lockit is a false constant expression.
>
> Based on results from './benchs/run_bench_local_storage.sh' (21 trials,
> reboot between each trial; x86 defconfig + BPF, clang 16) this produces
> improvements in throughput and latency in the majority of cases, with an
> average (geomean) improvement of 8%:
>
> +---- Hashmap Control --------------------
> |
> | + num keys: 10
> | : <before> | <after>
> | +-+ hashmap (control) sequential get +----------------------+----------------------
> | +- hits throughput | 14.789 M ops/s | 14.745 M ops/s ( ~ )
> | +- hits latency | 67.679 ns/op | 67.879 ns/op ( ~ )
> | +- important_hits throughput | 14.789 M ops/s | 14.745 M ops/s ( ~ )
> |
> | + num keys: 1000
> | : <before> | <after>
> | +-+ hashmap (control) sequential get +----------------------+----------------------
> | +- hits throughput | 12.233 M ops/s | 12.170 M ops/s ( ~ )
> | +- hits latency | 81.754 ns/op | 82.185 ns/op ( ~ )
> | +- important_hits throughput | 12.233 M ops/s | 12.170 M ops/s ( ~ )
> |
> | + num keys: 10000
> | : <before> | <after>
> | +-+ hashmap (control) sequential get +----------------------+----------------------
> | +- hits throughput | 7.220 M ops/s | 7.204 M ops/s ( ~ )
> | +- hits latency | 138.522 ns/op | 138.842 ns/op ( ~ )
> | +- important_hits throughput | 7.220 M ops/s | 7.204 M ops/s ( ~ )
> |
> | + num keys: 100000
> | : <before> | <after>
> | +-+ hashmap (control) sequential get +----------------------+----------------------
> | +- hits throughput | 5.061 M ops/s | 5.165 M ops/s (+2.1%)
> | +- hits latency | 198.483 ns/op | 194.270 ns/op (-2.1%)
> | +- important_hits throughput | 5.061 M ops/s | 5.165 M ops/s (+2.1%)
> |
> | + num keys: 4194304
> | : <before> | <after>
> | +-+ hashmap (control) sequential get +----------------------+----------------------
> | +- hits throughput | 2.864 M ops/s | 2.882 M ops/s ( ~ )
> | +- hits latency | 365.220 ns/op | 361.418 ns/op (-1.0%)
> | +- important_hits throughput | 2.864 M ops/s | 2.882 M ops/s ( ~ )
> |
> +---- Local Storage ----------------------
> |
> | + num_maps: 1
> | : <before> | <after>
> | +-+ local_storage cache sequential get +----------------------+----------------------
> | +- hits throughput | 33.005 M ops/s | 39.068 M ops/s (+18.4%)
> | +- hits latency | 30.300 ns/op | 25.598 ns/op (-15.5%)
> | +- important_hits throughput | 33.005 M ops/s | 39.068 M ops/s (+18.4%)
> | :
> | : <before> | <after>
> | +-+ local_storage cache interleaved get +----------------------+----------------------
> | +- hits throughput | 37.151 M ops/s | 44.926 M ops/s (+20.9%)
> | +- hits latency | 26.919 ns/op | 22.259 ns/op (-17.3%)
> | +- important_hits throughput | 37.151 M ops/s | 44.926 M ops/s (+20.9%)
> |
> | + num_maps: 10
> | : <before> | <after>
> | +-+ local_storage cache sequential get +----------------------+----------------------
> | +- hits throughput | 32.288 M ops/s | 38.099 M ops/s (+18.0%)
> | +- hits latency | 30.972 ns/op | 26.248 ns/op (-15.3%)
> | +- important_hits throughput | 3.229 M ops/s | 3.810 M ops/s (+18.0%)
> | :
> | : <before> | <after>
> | +-+ local_storage cache interleaved get +----------------------+----------------------
> | +- hits throughput | 34.473 M ops/s | 41.145 M ops/s (+19.4%)
> | +- hits latency | 29.010 ns/op | 24.307 ns/op (-16.2%)
> | +- important_hits throughput | 12.312 M ops/s | 14.695 M ops/s (+19.4%)
> |
> | + num_maps: 16
> | : <before> | <after>
> | +-+ local_storage cache sequential get +----------------------+----------------------
> | +- hits throughput | 32.524 M ops/s | 38.341 M ops/s (+17.9%)
> | +- hits latency | 30.748 ns/op | 26.083 ns/op (-15.2%)
> | +- important_hits throughput | 2.033 M ops/s | 2.396 M ops/s (+17.9%)
> | :
> | : <before> | <after>
> | +-+ local_storage cache interleaved get +----------------------+----------------------
> | +- hits throughput | 34.575 M ops/s | 41.338 M ops/s (+19.6%)
> | +- hits latency | 28.925 ns/op | 24.193 ns/op (-16.4%)
> | +- important_hits throughput | 11.001 M ops/s | 13.153 M ops/s (+19.6%)
> |
> | + num_maps: 17
> | : <before> | <after>
> | +-+ local_storage cache sequential get +----------------------+----------------------
> | +- hits throughput | 28.861 M ops/s | 32.756 M ops/s (+13.5%)
> | +- hits latency | 34.649 ns/op | 30.530 ns/op (-11.9%)
> | +- important_hits throughput | 1.700 M ops/s | 1.929 M ops/s (+13.5%)
> | :
> | : <before> | <after>
> | +-+ local_storage cache interleaved get +----------------------+----------------------
> | +- hits throughput | 31.529 M ops/s | 36.110 M ops/s (+14.5%)
> | +- hits latency | 31.719 ns/op | 27.697 ns/op (-12.7%)
> | +- important_hits throughput | 9.598 M ops/s | 10.993 M ops/s (+14.5%)
> |
> | + num_maps: 24
> | : <before> | <after>
> | +-+ local_storage cache sequential get +----------------------+----------------------
> | +- hits throughput | 18.602 M ops/s | 19.937 M ops/s (+7.2%)
> | +- hits latency | 53.767 ns/op | 50.166 ns/op (-6.7%)
> | +- important_hits throughput | 0.776 M ops/s | 0.831 M ops/s (+7.2%)
> | :
> | : <before> | <after>
> | +-+ local_storage cache interleaved get +----------------------+----------------------
> | +- hits throughput | 21.718 M ops/s | 23.332 M ops/s (+7.4%)
> | +- hits latency | 46.047 ns/op | 42.865 ns/op (-6.9%)
> | +- important_hits throughput | 6.110 M ops/s | 6.564 M ops/s (+7.4%)
> |
> | + num_maps: 32
> | : <before> | <after>
> | +-+ local_storage cache sequential get +----------------------+----------------------
> | +- hits throughput | 14.118 M ops/s | 14.626 M ops/s (+3.6%)
> | +- hits latency | 70.856 ns/op | 68.381 ns/op (-3.5%)
> | +- important_hits throughput | 0.442 M ops/s | 0.458 M ops/s (+3.6%)
> | :
> | : <before> | <after>
> | +-+ local_storage cache interleaved get +----------------------+----------------------
> | +- hits throughput | 17.111 M ops/s | 17.906 M ops/s (+4.6%)
> | +- hits latency | 58.451 ns/op | 55.865 ns/op (-4.4%)
> | +- important_hits throughput | 4.776 M ops/s | 4.998 M ops/s (+4.6%)
> |
> | + num_maps: 100
> | : <before> | <after>
> | +-+ local_storage cache sequential get +----------------------+----------------------
> | +- hits throughput | 5.281 M ops/s | 5.528 M ops/s (+4.7%)
> | +- hits latency | 192.398 ns/op | 183.059 ns/op (-4.9%)
> | +- important_hits throughput | 0.053 M ops/s | 0.055 M ops/s (+4.9%)
> | :
> | : <before> | <after>
> | +-+ local_storage cache interleaved get +----------------------+----------------------
> | +- hits throughput | 6.265 M ops/s | 6.498 M ops/s (+3.7%)
> | +- hits latency | 161.436 ns/op | 152.877 ns/op (-5.3%)
> | +- important_hits throughput | 1.636 M ops/s | 1.697 M ops/s (+3.7%)
> |
> | + num_maps: 1000
> | : <before> | <after>
> | +-+ local_storage cache sequential get +----------------------+----------------------
> | +- hits throughput | 0.355 M ops/s | 0.354 M ops/s ( ~ )
> | +- hits latency | 2826.538 ns/op | 2827.139 ns/op ( ~ )
> | +- important_hits throughput | 0.000 M ops/s | 0.000 M ops/s ( ~ )
> | :
> | : <before> | <after>
> | +-+ local_storage cache interleaved get +----------------------+----------------------
> | +- hits throughput | 0.404 M ops/s | 0.403 M ops/s ( ~ )
> | +- hits latency | 2481.190 ns/op | 2487.555 ns/op ( ~ )
> | +- important_hits throughput | 0.102 M ops/s | 0.101 M ops/s ( ~ )
>
> Signed-off-by: Marco Elver <[email protected]>
> Cc: Martin KaFai Lau <[email protected]>
> ---
> v2:
> * Inline most of bpf_local_storage_lookup(), which produces greater
> speedup and avoids regressing the cases with large map arrays.
> * Drop "likely()" hint, it didn't produce much benefit.
> * Re-run benchmark and collect 21 trials of results.
> * Remove the on_lookup tests, which no longer work because
> e.g. bpf_task_storage_delete() no longer does calls to
> bpf_local_storage_*() helpers.
> ---
> include/linux/bpf_local_storage.h | 30 ++++++++++-
> kernel/bpf/bpf_local_storage.c | 52 +++++--------------
> .../bpf/prog_tests/task_local_storage.c | 6 ---
> .../selftests/bpf/progs/cgrp_ls_recursion.c | 26 ----------
> .../selftests/bpf/progs/task_ls_recursion.c | 17 ------
> 5 files changed, 41 insertions(+), 90 deletions(-)
>
> diff --git a/include/linux/bpf_local_storage.h b/include/linux/bpf_local_storage.h
> index 173ec7f43ed1..dcddb0aef7d8 100644
> --- a/include/linux/bpf_local_storage.h
> +++ b/include/linux/bpf_local_storage.h
> @@ -129,10 +129,36 @@ bpf_local_storage_map_alloc(union bpf_attr *attr,
> struct bpf_local_storage_cache *cache,
> bool bpf_ma);
>
> -struct bpf_local_storage_data *
> +void __bpf_local_storage_insert_cache(struct bpf_local_storage *local_storage,
> + struct bpf_local_storage_map *smap,
> + struct bpf_local_storage_elem *selem);
> +/* If cacheit_lockit is false, this lookup function is lockless */
> +static inline struct bpf_local_storage_data *
> bpf_local_storage_lookup(struct bpf_local_storage *local_storage,
> struct bpf_local_storage_map *smap,
> - bool cacheit_lockit);
> + bool cacheit_lockit)
> +{
> + struct bpf_local_storage_data *sdata;
> + struct bpf_local_storage_elem *selem;
> +
> + /* Fast path (cache hit) */
> + sdata = rcu_dereference_check(local_storage->cache[smap->cache_idx],
> + bpf_rcu_lock_held());
> + if (sdata && rcu_access_pointer(sdata->smap) == smap)
> + return sdata;
I think we should focus on fast path (your v1 patch)
and I suppose most production environments
want to hit fast path in most times. In your production environment did
you see more than 16 local storage maps per entity (task/sk/inode)?
In the fast path, the memory accesses are
two from local_storage->cache[smap->cache_idx] and
one from sdata->smap
> +
> + /* Slow path (cache miss) */
> + hlist_for_each_entry_rcu(selem, &local_storage->list, snode,
> + rcu_read_lock_trace_held())
> + if (rcu_access_pointer(SDATA(selem)->smap) == smap)
> + break;
But if we reach slow path here which means we have more than 16 local
storage maps, then traversing the list and getting SDATA(selem)->smap
will be very expensive, in addition to memory accesses in fast path.
I suppose here we mostly care about socket local storage since it is
totally possible for a production workload to have millions of sockets.
To improve performance, fast path should hit in most cases.
If there are too many sk local storage maps, some kind of sharing
can be done so multiple applications might be using a single sk
local storage.
Your above inlining/outlining analysis also show how tricky it is
for compilation optimization. Without profiling, it is totally
possible that compiler might do optimization differently in
the future. So here is my suggestion, let us do inlining
for fast path and focus on performance of fast path.
> + if (!selem)
> + return NULL;
> + if (cacheit_lockit)
> + __bpf_local_storage_insert_cache(local_storage, smap, selem);
> + return SDATA(selem);
> +}
>
> void bpf_local_storage_destroy(struct bpf_local_storage *local_storage);
>
[...]
On Thu, 8 Feb 2024 at 00:58, Yonghong Song <[email protected]> wrote:
> On 2/7/24 4:26 AM, Marco Elver wrote:
> > In various performance profiles of kernels with BPF programs attached,
> > bpf_local_storage_lookup() appears as a significant portion of CPU
> > cycles spent. To enable the compiler generate more optimal code, turn
> > bpf_local_storage_lookup() into a static inline function, where only the
> > cache insertion code path is outlined
> >
> > Notably, outlining cache insertion helps avoid bloating callers by
> > duplicating setting up calls to raw_spin_{lock,unlock}_irqsave() (on
> > architectures which do not inline spin_lock/unlock, such as x86), which
> > would cause the compiler produce worse code by deciding to outline
> > otherwise inlinable functions. The call overhead is neutral, because we
> > make 2 calls either way: either calling raw_spin_lock_irqsave() and
> > raw_spin_unlock_irqsave(); or call __bpf_local_storage_insert_cache(),
> > which calls raw_spin_lock_irqsave(), followed by a tail-call to
> > raw_spin_unlock_irqsave() where the compiler can perform TCO and (in
> > optimized uninstrumented builds) turns it into a plain jump. The call to
> > __bpf_local_storage_insert_cache() can be elided entirely if
> > cacheit_lockit is a false constant expression.
> >
> > Based on results from './benchs/run_bench_local_storage.sh' (21 trials,
> > reboot between each trial; x86 defconfig + BPF, clang 16) this produces
> > improvements in throughput and latency in the majority of cases, with an
> > average (geomean) improvement of 8%:
[...]
> > include/linux/bpf_local_storage.h | 30 ++++++++++-
> > kernel/bpf/bpf_local_storage.c | 52 +++++--------------
> > .../bpf/prog_tests/task_local_storage.c | 6 ---
> > .../selftests/bpf/progs/cgrp_ls_recursion.c | 26 ----------
> > .../selftests/bpf/progs/task_ls_recursion.c | 17 ------
> > 5 files changed, 41 insertions(+), 90 deletions(-)
> >
> > diff --git a/include/linux/bpf_local_storage.h b/include/linux/bpf_local_storage.h
> > index 173ec7f43ed1..dcddb0aef7d8 100644
> > --- a/include/linux/bpf_local_storage.h
> > +++ b/include/linux/bpf_local_storage.h
> > @@ -129,10 +129,36 @@ bpf_local_storage_map_alloc(union bpf_attr *attr,
> > struct bpf_local_storage_cache *cache,
> > bool bpf_ma);
> >
> > -struct bpf_local_storage_data *
> > +void __bpf_local_storage_insert_cache(struct bpf_local_storage *local_storage,
> > + struct bpf_local_storage_map *smap,
> > + struct bpf_local_storage_elem *selem);
> > +/* If cacheit_lockit is false, this lookup function is lockless */
> > +static inline struct bpf_local_storage_data *
> > bpf_local_storage_lookup(struct bpf_local_storage *local_storage,
> > struct bpf_local_storage_map *smap,
> > - bool cacheit_lockit);
> > + bool cacheit_lockit)
> > +{
> > + struct bpf_local_storage_data *sdata;
> > + struct bpf_local_storage_elem *selem;
> > +
> > + /* Fast path (cache hit) */
> > + sdata = rcu_dereference_check(local_storage->cache[smap->cache_idx],
> > + bpf_rcu_lock_held());
> > + if (sdata && rcu_access_pointer(sdata->smap) == smap)
> > + return sdata;
>
> I think we should focus on fast path (your v1 patch)
> and I suppose most production environments
> want to hit fast path in most times. In your production environment did
> you see more than 16 local storage maps per entity (task/sk/inode)?
I think having more than 16 local storage maps isn't entirely unlikely
as eBPF usage grows. But at the moment, it should be rare.
> In the fast path, the memory accesses are
> two from local_storage->cache[smap->cache_idx] and
> one from sdata->smap
>
>
> > +
> > + /* Slow path (cache miss) */
> > + hlist_for_each_entry_rcu(selem, &local_storage->list, snode,
> > + rcu_read_lock_trace_held())
> > + if (rcu_access_pointer(SDATA(selem)->smap) == smap)
> > + break;
>
> But if we reach slow path here which means we have more than 16 local
> storage maps, then traversing the list and getting SDATA(selem)->smap
> will be very expensive, in addition to memory accesses in fast path.
>
> I suppose here we mostly care about socket local storage since it is
> totally possible for a production workload to have millions of sockets.
> To improve performance, fast path should hit in most cases.
> If there are too many sk local storage maps, some kind of sharing
> can be done so multiple applications might be using a single sk
> local storage.
>
> Your above inlining/outlining analysis also show how tricky it is
> for compilation optimization. Without profiling, it is totally
> possible that compiler might do optimization differently in
> the future.
Sure, but it's usually the case that we have to help the compiler a
little to produce more optimal code - if the compiler becomes stupid
in future, we need either fix the compiler or help it some more.
> So here is my suggestion, let us do inlining
> for fast path and focus on performance of fast path.
The slow-path (iterate list w/o cache insertion) is still relatively
small (it's a pointer-chasing loop and a compare), and I decided that
it can be justified inlining it. Martin asked in v1 why there were
slowdowns above 16 local maps, and I analyzed, and concluded that
inlining most is needed to fix and does not hurt performance: in fact,
the current version is better than v1 in all cases (even for 16 maps
or below).
Let me know which version you prefer, and I'll change it. However,
based on the results, I would prefer the current version.
Thanks,
-- Marco
On Thu, Feb 08, 2024 at 08:37AM +0100, Marco Elver wrote:
> On Thu, 8 Feb 2024 at 00:58, Yonghong Song <[email protected]> wrote:
> > On 2/7/24 4:26 AM, Marco Elver wrote:
> > > In various performance profiles of kernels with BPF programs attached,
> > > bpf_local_storage_lookup() appears as a significant portion of CPU
> > > cycles spent. To enable the compiler generate more optimal code, turn
> > > bpf_local_storage_lookup() into a static inline function, where only the
> > > cache insertion code path is outlined
> > >
> > > Notably, outlining cache insertion helps avoid bloating callers by
> > > duplicating setting up calls to raw_spin_{lock,unlock}_irqsave() (on
> > > architectures which do not inline spin_lock/unlock, such as x86), which
> > > would cause the compiler produce worse code by deciding to outline
> > > otherwise inlinable functions. The call overhead is neutral, because we
> > > make 2 calls either way: either calling raw_spin_lock_irqsave() and
> > > raw_spin_unlock_irqsave(); or call __bpf_local_storage_insert_cache(),
> > > which calls raw_spin_lock_irqsave(), followed by a tail-call to
> > > raw_spin_unlock_irqsave() where the compiler can perform TCO and (in
> > > optimized uninstrumented builds) turns it into a plain jump. The call to
> > > __bpf_local_storage_insert_cache() can be elided entirely if
> > > cacheit_lockit is a false constant expression.
> > >
> > > Based on results from './benchs/run_bench_local_storage.sh' (21 trials,
> > > reboot between each trial; x86 defconfig + BPF, clang 16) this produces
> > > improvements in throughput and latency in the majority of cases, with an
> > > average (geomean) improvement of 8%:
> [...]
> > > include/linux/bpf_local_storage.h | 30 ++++++++++-
> > > kernel/bpf/bpf_local_storage.c | 52 +++++--------------
> > > .../bpf/prog_tests/task_local_storage.c | 6 ---
> > > .../selftests/bpf/progs/cgrp_ls_recursion.c | 26 ----------
> > > .../selftests/bpf/progs/task_ls_recursion.c | 17 ------
> > > 5 files changed, 41 insertions(+), 90 deletions(-)
> > >
> > > diff --git a/include/linux/bpf_local_storage.h b/include/linux/bpf_local_storage.h
> > > index 173ec7f43ed1..dcddb0aef7d8 100644
> > > --- a/include/linux/bpf_local_storage.h
> > > +++ b/include/linux/bpf_local_storage.h
> > > @@ -129,10 +129,36 @@ bpf_local_storage_map_alloc(union bpf_attr *attr,
> > > struct bpf_local_storage_cache *cache,
> > > bool bpf_ma);
> > >
> > > -struct bpf_local_storage_data *
> > > +void __bpf_local_storage_insert_cache(struct bpf_local_storage *local_storage,
> > > + struct bpf_local_storage_map *smap,
> > > + struct bpf_local_storage_elem *selem);
> > > +/* If cacheit_lockit is false, this lookup function is lockless */
> > > +static inline struct bpf_local_storage_data *
> > > bpf_local_storage_lookup(struct bpf_local_storage *local_storage,
> > > struct bpf_local_storage_map *smap,
> > > - bool cacheit_lockit);
> > > + bool cacheit_lockit)
> > > +{
> > > + struct bpf_local_storage_data *sdata;
> > > + struct bpf_local_storage_elem *selem;
> > > +
> > > + /* Fast path (cache hit) */
> > > + sdata = rcu_dereference_check(local_storage->cache[smap->cache_idx],
> > > + bpf_rcu_lock_held());
> > > + if (sdata && rcu_access_pointer(sdata->smap) == smap)
> > > + return sdata;
> >
> > I think we should focus on fast path (your v1 patch)
> > and I suppose most production environments
> > want to hit fast path in most times. In your production environment did
> > you see more than 16 local storage maps per entity (task/sk/inode)?
>
> I think having more than 16 local storage maps isn't entirely unlikely
> as eBPF usage grows. But at the moment, it should be rare.
>
> > In the fast path, the memory accesses are
> > two from local_storage->cache[smap->cache_idx] and
> > one from sdata->smap
> >
> >
> > > +
> > > + /* Slow path (cache miss) */
> > > + hlist_for_each_entry_rcu(selem, &local_storage->list, snode,
> > > + rcu_read_lock_trace_held())
> > > + if (rcu_access_pointer(SDATA(selem)->smap) == smap)
> > > + break;
> >
> > But if we reach slow path here which means we have more than 16 local
> > storage maps, then traversing the list and getting SDATA(selem)->smap
> > will be very expensive, in addition to memory accesses in fast path.
> >
> > I suppose here we mostly care about socket local storage since it is
> > totally possible for a production workload to have millions of sockets.
> > To improve performance, fast path should hit in most cases.
> > If there are too many sk local storage maps, some kind of sharing
> > can be done so multiple applications might be using a single sk
> > local storage.
> >
> > Your above inlining/outlining analysis also show how tricky it is
> > for compilation optimization. Without profiling, it is totally
> > possible that compiler might do optimization differently in
> > the future.
>
> Sure, but it's usually the case that we have to help the compiler a
> little to produce more optimal code - if the compiler becomes stupid
> in future, we need either fix the compiler or help it some more.
>
> > So here is my suggestion, let us do inlining
> > for fast path and focus on performance of fast path.
>
> The slow-path (iterate list w/o cache insertion) is still relatively
> small (it's a pointer-chasing loop and a compare), and I decided that
> it can be justified inlining it. Martin asked in v1 why there were
> slowdowns above 16 local maps, and I analyzed, and concluded that
> inlining most is needed to fix and does not hurt performance: in fact,
> the current version is better than v1 in all cases (even for 16 maps
> or below).
>
> Let me know which version you prefer, and I'll change it. However,
> based on the results, I would prefer the current version.
FTR, these were the results going from v1 (before) -> v2 (after):
+---- Local Storage ----------------------
|
| + num_maps: 1
| : <before> | <after>
| +-+ local_storage cache sequential get +----------------------+----------------------
| +- hits throughput | 38.593 M ops/s | 39.068 M ops/s (+1.2%)
| +- hits latency | 25.913 ns/op | 25.598 ns/op (-1.2%)
| +- important_hits throughput | 38.593 M ops/s | 39.068 M ops/s (+1.2%)
| :
| : <before> | <after>
| +-+ local_storage cache interleaved get +----------------------+----------------------
| +- hits throughput | 44.406 M ops/s | 44.926 M ops/s (+1.2%)
| +- hits latency | 22.521 ns/op | 22.259 ns/op (-1.2%)
| +- important_hits throughput | 44.406 M ops/s | 44.926 M ops/s (+1.2%)
|
| + num_maps: 10
| : <before> | <after>
| +-+ local_storage cache sequential get +----------------------+----------------------
| +- hits throughput | 37.583 M ops/s | 38.099 M ops/s (+1.4%)
| +- hits latency | 26.609 ns/op | 26.248 ns/op (-1.4%)
| +- important_hits throughput | 3.758 M ops/s | 3.810 M ops/s (+1.4%)
| :
| : <before> | <after>
| +-+ local_storage cache interleaved get +----------------------+----------------------
| +- hits throughput | 40.698 M ops/s | 41.145 M ops/s (+1.1%)
| +- hits latency | 24.573 ns/op | 24.307 ns/op (-1.1%)
| +- important_hits throughput | 14.535 M ops/s | 14.695 M ops/s (+1.1%)
|
| + num_maps: 16
| : <before> | <after>
| +-+ local_storage cache sequential get +----------------------+----------------------
| +- hits throughput | 38.061 M ops/s | 38.341 M ops/s ( ~ )
| +- hits latency | 26.275 ns/op | 26.083 ns/op ( ~ )
| +- important_hits throughput | 2.379 M ops/s | 2.396 M ops/s ( ~ )
| :
| : <before> | <after>
| +-+ local_storage cache interleaved get +----------------------+----------------------
| +- hits throughput | 40.890 M ops/s | 41.338 M ops/s (+1.1%)
| +- hits latency | 24.458 ns/op | 24.193 ns/op (-1.1%)
| +- important_hits throughput | 13.010 M ops/s | 13.153 M ops/s (+1.1%)
|
| + num_maps: 17
| : <before> | <after>
| +-+ local_storage cache sequential get +----------------------+----------------------
| +- hits throughput | 31.799 M ops/s | 32.756 M ops/s (+3.0%)
| +- hits latency | 31.448 ns/op | 30.530 ns/op (-2.9%)
| +- important_hits throughput | 1.873 M ops/s | 1.929 M ops/s (+3.0%)
| :
| : <before> | <after>
| +-+ local_storage cache interleaved get +----------------------+----------------------
| +- hits throughput | 35.284 M ops/s | 36.110 M ops/s (+2.3%)
| +- hits latency | 28.343 ns/op | 27.697 ns/op (-2.3%)
| +- important_hits throughput | 10.742 M ops/s | 10.993 M ops/s (+2.3%)
|
| + num_maps: 24
| : <before> | <after>
| +-+ local_storage cache sequential get +----------------------+----------------------
| +- hits throughput | 17.947 M ops/s | 19.937 M ops/s (+11.1%)
| +- hits latency | 55.725 ns/op | 50.166 ns/op (-10.0%)
| +- important_hits throughput | 0.748 M ops/s | 0.831 M ops/s (+11.1%)
| :
| : <before> | <after>
| +-+ local_storage cache interleaved get +----------------------+----------------------
| +- hits throughput | 21.379 M ops/s | 23.332 M ops/s (+9.1%)
| +- hits latency | 46.775 ns/op | 42.865 ns/op (-8.4%)
| +- important_hits throughput | 6.014 M ops/s | 6.564 M ops/s (+9.1%)
|
| + num_maps: 32
| : <before> | <after>
| +-+ local_storage cache sequential get +----------------------+----------------------
| +- hits throughput | 13.279 M ops/s | 14.626 M ops/s (+10.1%)
| +- hits latency | 75.317 ns/op | 68.381 ns/op (-9.2%)
| +- important_hits throughput | 0.416 M ops/s | 0.458 M ops/s (+10.2%)
| :
| : <before> | <after>
| +-+ local_storage cache interleaved get +----------------------+----------------------
| +- hits throughput | 16.444 M ops/s | 17.906 M ops/s (+8.9%)
| +- hits latency | 60.816 ns/op | 55.865 ns/op (-8.1%)
| +- important_hits throughput | 4.590 M ops/s | 4.998 M ops/s (+8.9%)
|
| + num_maps: 100
| : <before> | <after>
| +-+ local_storage cache sequential get +----------------------+----------------------
| +- hits throughput | 4.912 M ops/s | 5.528 M ops/s (+12.5%)
| +- hits latency | 207.291 ns/op | 183.059 ns/op (-11.7%)
| +- important_hits throughput | 0.049 M ops/s | 0.055 M ops/s (+12.7%)
| :
| : <before> | <after>
| +-+ local_storage cache interleaved get +----------------------+----------------------
| +- hits throughput | 6.039 M ops/s | 6.498 M ops/s (+7.6%)
| +- hits latency | 167.325 ns/op | 152.877 ns/op (-8.6%)
| +- important_hits throughput | 1.577 M ops/s | 1.697 M ops/s (+7.6%)
|
| + num_maps: 1000
| : <before> | <after>
| +-+ local_storage cache sequential get +----------------------+----------------------
| +- hits throughput | 0.342 M ops/s | 0.354 M ops/s (+3.6%)
| +- hits latency | 2930.550 ns/op | 2827.139 ns/op (-3.5%)
| +- important_hits throughput | 0.000 M ops/s | 0.000 M ops/s ( ~ )
| :
| : <before> | <after>
| +-+ local_storage cache interleaved get +----------------------+----------------------
| +- hits throughput | 0.413 M ops/s | 0.403 M ops/s (-2.5%)
| +- hits latency | 2427.830 ns/op | 2487.555 ns/op (+2.5%)
| +- important_hits throughput | 0.104 M ops/s | 0.101 M ops/s (-2.6%)
|
| Geomean:
| hits throughput: 102.93%
| hits latency: 97.11%
| important_hits throughput: 102.77%
On 2/8/24 2:54 AM, Marco Elver wrote:
> On Thu, Feb 08, 2024 at 08:37AM +0100, Marco Elver wrote:
>> On Thu, 8 Feb 2024 at 00:58, Yonghong Song <[email protected]> wrote:
>>> On 2/7/24 4:26 AM, Marco Elver wrote:
>>>> In various performance profiles of kernels with BPF programs attached,
>>>> bpf_local_storage_lookup() appears as a significant portion of CPU
>>>> cycles spent. To enable the compiler generate more optimal code, turn
>>>> bpf_local_storage_lookup() into a static inline function, where only the
>>>> cache insertion code path is outlined
>>>>
>>>> Notably, outlining cache insertion helps avoid bloating callers by
>>>> duplicating setting up calls to raw_spin_{lock,unlock}_irqsave() (on
>>>> architectures which do not inline spin_lock/unlock, such as x86), which
>>>> would cause the compiler produce worse code by deciding to outline
>>>> otherwise inlinable functions. The call overhead is neutral, because we
>>>> make 2 calls either way: either calling raw_spin_lock_irqsave() and
>>>> raw_spin_unlock_irqsave(); or call __bpf_local_storage_insert_cache(),
>>>> which calls raw_spin_lock_irqsave(), followed by a tail-call to
>>>> raw_spin_unlock_irqsave() where the compiler can perform TCO and (in
>>>> optimized uninstrumented builds) turns it into a plain jump. The call to
>>>> __bpf_local_storage_insert_cache() can be elided entirely if
>>>> cacheit_lockit is a false constant expression.
>>>>
>>>> Based on results from './benchs/run_bench_local_storage.sh' (21 trials,
>>>> reboot between each trial; x86 defconfig + BPF, clang 16) this produces
>>>> improvements in throughput and latency in the majority of cases, with an
>>>> average (geomean) improvement of 8%:
>> [...]
>>>> include/linux/bpf_local_storage.h | 30 ++++++++++-
>>>> kernel/bpf/bpf_local_storage.c | 52 +++++--------------
>>>> .../bpf/prog_tests/task_local_storage.c | 6 ---
>>>> .../selftests/bpf/progs/cgrp_ls_recursion.c | 26 ----------
>>>> .../selftests/bpf/progs/task_ls_recursion.c | 17 ------
>>>> 5 files changed, 41 insertions(+), 90 deletions(-)
>>>>
>>>> diff --git a/include/linux/bpf_local_storage.h b/include/linux/bpf_local_storage.h
>>>> index 173ec7f43ed1..dcddb0aef7d8 100644
>>>> --- a/include/linux/bpf_local_storage.h
>>>> +++ b/include/linux/bpf_local_storage.h
>>>> @@ -129,10 +129,36 @@ bpf_local_storage_map_alloc(union bpf_attr *attr,
>>>> struct bpf_local_storage_cache *cache,
>>>> bool bpf_ma);
>>>>
>>>> -struct bpf_local_storage_data *
>>>> +void __bpf_local_storage_insert_cache(struct bpf_local_storage *local_storage,
>>>> + struct bpf_local_storage_map *smap,
>>>> + struct bpf_local_storage_elem *selem);
>>>> +/* If cacheit_lockit is false, this lookup function is lockless */
>>>> +static inline struct bpf_local_storage_data *
>>>> bpf_local_storage_lookup(struct bpf_local_storage *local_storage,
>>>> struct bpf_local_storage_map *smap,
>>>> - bool cacheit_lockit);
>>>> + bool cacheit_lockit)
>>>> +{
>>>> + struct bpf_local_storage_data *sdata;
>>>> + struct bpf_local_storage_elem *selem;
>>>> +
>>>> + /* Fast path (cache hit) */
>>>> + sdata = rcu_dereference_check(local_storage->cache[smap->cache_idx],
>>>> + bpf_rcu_lock_held());
>>>> + if (sdata && rcu_access_pointer(sdata->smap) == smap)
>>>> + return sdata;
>>> I think we should focus on fast path (your v1 patch)
>>> and I suppose most production environments
>>> want to hit fast path in most times. In your production environment did
>>> you see more than 16 local storage maps per entity (task/sk/inode)?
>> I think having more than 16 local storage maps isn't entirely unlikely
>> as eBPF usage grows. But at the moment, it should be rare.
>>
>>> In the fast path, the memory accesses are
>>> two from local_storage->cache[smap->cache_idx] and
>>> one from sdata->smap
>>>
>>>
>>>> +
>>>> + /* Slow path (cache miss) */
>>>> + hlist_for_each_entry_rcu(selem, &local_storage->list, snode,
>>>> + rcu_read_lock_trace_held())
>>>> + if (rcu_access_pointer(SDATA(selem)->smap) == smap)
>>>> + break;
>>> But if we reach slow path here which means we have more than 16 local
>>> storage maps, then traversing the list and getting SDATA(selem)->smap
>>> will be very expensive, in addition to memory accesses in fast path.
>>>
>>> I suppose here we mostly care about socket local storage since it is
>>> totally possible for a production workload to have millions of sockets.
>>> To improve performance, fast path should hit in most cases.
>>> If there are too many sk local storage maps, some kind of sharing
>>> can be done so multiple applications might be using a single sk
>>> local storage.
>>>
>>> Your above inlining/outlining analysis also show how tricky it is
>>> for compilation optimization. Without profiling, it is totally
>>> possible that compiler might do optimization differently in
>>> the future.
>> Sure, but it's usually the case that we have to help the compiler a
>> little to produce more optimal code - if the compiler becomes stupid
>> in future, we need either fix the compiler or help it some more.
>>
>>> So here is my suggestion, let us do inlining
>>> for fast path and focus on performance of fast path.
>> The slow-path (iterate list w/o cache insertion) is still relatively
>> small (it's a pointer-chasing loop and a compare), and I decided that
>> it can be justified inlining it. Martin asked in v1 why there were
>> slowdowns above 16 local maps, and I analyzed, and concluded that
>> inlining most is needed to fix and does not hurt performance: in fact,
>> the current version is better than v1 in all cases (even for 16 maps
>> or below).
>>
>> Let me know which version you prefer, and I'll change it. However,
>> based on the results, I would prefer the current version.
I compared asm code between v1 and v2. In bpf_local_storage_update(),
bpf_local_storage_lookup() is called twice with cacheit_lockit == false.
Fully inlining bpf_local_storage_lookup() in bpf_local_storage_update()
should generate better code compared to v1.
So let us do v2.
Acked-by: Yonghong Song <[email protected]>
> FTR, these were the results going from v1 (before) -> v2 (after):
>
> +---- Local Storage ----------------------
> |
> | + num_maps: 1
> | : <before> | <after>
> | +-+ local_storage cache sequential get +----------------------+----------------------
> | +- hits throughput | 38.593 M ops/s | 39.068 M ops/s (+1.2%)
> | +- hits latency | 25.913 ns/op | 25.598 ns/op (-1.2%)
> | +- important_hits throughput | 38.593 M ops/s | 39.068 M ops/s (+1.2%)
> | :
> | : <before> | <after>
> | +-+ local_storage cache interleaved get +----------------------+----------------------
> | +- hits throughput | 44.406 M ops/s | 44.926 M ops/s (+1.2%)
> | +- hits latency | 22.521 ns/op | 22.259 ns/op (-1.2%)
> | +- important_hits throughput | 44.406 M ops/s | 44.926 M ops/s (+1.2%)
> |
> | + num_maps: 10
> | : <before> | <after>
> | +-+ local_storage cache sequential get +----------------------+----------------------
> | +- hits throughput | 37.583 M ops/s | 38.099 M ops/s (+1.4%)
> | +- hits latency | 26.609 ns/op | 26.248 ns/op (-1.4%)
> | +- important_hits throughput | 3.758 M ops/s | 3.810 M ops/s (+1.4%)
> | :
> | : <before> | <after>
> | +-+ local_storage cache interleaved get +----------------------+----------------------
> | +- hits throughput | 40.698 M ops/s | 41.145 M ops/s (+1.1%)
> | +- hits latency | 24.573 ns/op | 24.307 ns/op (-1.1%)
> | +- important_hits throughput | 14.535 M ops/s | 14.695 M ops/s (+1.1%)
> |
> | + num_maps: 16
> | : <before> | <after>
> | +-+ local_storage cache sequential get +----------------------+----------------------
> | +- hits throughput | 38.061 M ops/s | 38.341 M ops/s ( ~ )
> | +- hits latency | 26.275 ns/op | 26.083 ns/op ( ~ )
> | +- important_hits throughput | 2.379 M ops/s | 2.396 M ops/s ( ~ )
> | :
> | : <before> | <after>
> | +-+ local_storage cache interleaved get +----------------------+----------------------
> | +- hits throughput | 40.890 M ops/s | 41.338 M ops/s (+1.1%)
> | +- hits latency | 24.458 ns/op | 24.193 ns/op (-1.1%)
> | +- important_hits throughput | 13.010 M ops/s | 13.153 M ops/s (+1.1%)
> |
> | + num_maps: 17
> | : <before> | <after>
> | +-+ local_storage cache sequential get +----------------------+----------------------
> | +- hits throughput | 31.799 M ops/s | 32.756 M ops/s (+3.0%)
> | +- hits latency | 31.448 ns/op | 30.530 ns/op (-2.9%)
> | +- important_hits throughput | 1.873 M ops/s | 1.929 M ops/s (+3.0%)
> | :
> | : <before> | <after>
> | +-+ local_storage cache interleaved get +----------------------+----------------------
> | +- hits throughput | 35.284 M ops/s | 36.110 M ops/s (+2.3%)
> | +- hits latency | 28.343 ns/op | 27.697 ns/op (-2.3%)
> | +- important_hits throughput | 10.742 M ops/s | 10.993 M ops/s (+2.3%)
> |
> | + num_maps: 24
> | : <before> | <after>
> | +-+ local_storage cache sequential get +----------------------+----------------------
> | +- hits throughput | 17.947 M ops/s | 19.937 M ops/s (+11.1%)
> | +- hits latency | 55.725 ns/op | 50.166 ns/op (-10.0%)
> | +- important_hits throughput | 0.748 M ops/s | 0.831 M ops/s (+11.1%)
> | :
> | : <before> | <after>
> | +-+ local_storage cache interleaved get +----------------------+----------------------
> | +- hits throughput | 21.379 M ops/s | 23.332 M ops/s (+9.1%)
> | +- hits latency | 46.775 ns/op | 42.865 ns/op (-8.4%)
> | +- important_hits throughput | 6.014 M ops/s | 6.564 M ops/s (+9.1%)
> |
> | + num_maps: 32
> | : <before> | <after>
> | +-+ local_storage cache sequential get +----------------------+----------------------
> | +- hits throughput | 13.279 M ops/s | 14.626 M ops/s (+10.1%)
> | +- hits latency | 75.317 ns/op | 68.381 ns/op (-9.2%)
> | +- important_hits throughput | 0.416 M ops/s | 0.458 M ops/s (+10.2%)
> | :
> | : <before> | <after>
> | +-+ local_storage cache interleaved get +----------------------+----------------------
> | +- hits throughput | 16.444 M ops/s | 17.906 M ops/s (+8.9%)
> | +- hits latency | 60.816 ns/op | 55.865 ns/op (-8.1%)
> | +- important_hits throughput | 4.590 M ops/s | 4.998 M ops/s (+8.9%)
> |
> | + num_maps: 100
> | : <before> | <after>
> | +-+ local_storage cache sequential get +----------------------+----------------------
> | +- hits throughput | 4.912 M ops/s | 5.528 M ops/s (+12.5%)
> | +- hits latency | 207.291 ns/op | 183.059 ns/op (-11.7%)
> | +- important_hits throughput | 0.049 M ops/s | 0.055 M ops/s (+12.7%)
> | :
> | : <before> | <after>
> | +-+ local_storage cache interleaved get +----------------------+----------------------
> | +- hits throughput | 6.039 M ops/s | 6.498 M ops/s (+7.6%)
> | +- hits latency | 167.325 ns/op | 152.877 ns/op (-8.6%)
> | +- important_hits throughput | 1.577 M ops/s | 1.697 M ops/s (+7.6%)
> |
> | + num_maps: 1000
> | : <before> | <after>
> | +-+ local_storage cache sequential get +----------------------+----------------------
> | +- hits throughput | 0.342 M ops/s | 0.354 M ops/s (+3.6%)
> | +- hits latency | 2930.550 ns/op | 2827.139 ns/op (-3.5%)
> | +- important_hits throughput | 0.000 M ops/s | 0.000 M ops/s ( ~ )
> | :
> | : <before> | <after>
> | +-+ local_storage cache interleaved get +----------------------+----------------------
> | +- hits throughput | 0.413 M ops/s | 0.403 M ops/s (-2.5%)
> | +- hits latency | 2427.830 ns/op | 2487.555 ns/op (+2.5%)
> | +- important_hits throughput | 0.104 M ops/s | 0.101 M ops/s (-2.6%)
> |
> | Geomean:
> | hits throughput: 102.93%
> | hits latency: 97.11%
> | important_hits throughput: 102.77%
Hello:
This patch was applied to bpf/bpf-next.git (master)
by Martin KaFai Lau <[email protected]>:
On Wed, 7 Feb 2024 13:26:17 +0100 you wrote:
> In various performance profiles of kernels with BPF programs attached,
> bpf_local_storage_lookup() appears as a significant portion of CPU
> cycles spent. To enable the compiler generate more optimal code, turn
> bpf_local_storage_lookup() into a static inline function, where only the
> cache insertion code path is outlined
>
> Notably, outlining cache insertion helps avoid bloating callers by
> duplicating setting up calls to raw_spin_{lock,unlock}_irqsave() (on
> architectures which do not inline spin_lock/unlock, such as x86), which
> would cause the compiler produce worse code by deciding to outline
> otherwise inlinable functions. The call overhead is neutral, because we
> make 2 calls either way: either calling raw_spin_lock_irqsave() and
> raw_spin_unlock_irqsave(); or call __bpf_local_storage_insert_cache(),
> which calls raw_spin_lock_irqsave(), followed by a tail-call to
> raw_spin_unlock_irqsave() where the compiler can perform TCO and (in
> optimized uninstrumented builds) turns it into a plain jump. The call to
> __bpf_local_storage_insert_cache() can be elided entirely if
> cacheit_lockit is a false constant expression.
>
> [...]
Here is the summary with links:
- [bpf-next,v2] bpf: Allow compiler to inline most of bpf_local_storage_lookup()
https://git.kernel.org/bpf/bpf-next/c/68bc61c26cac
You are awesome, thank you!
--
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html