Commit 52dbd67d authored by Jose E. Marchesi's avatar Jose E. Marchesi Committed by Andrii Nakryiko
Browse files

bpf: Abstract loop unrolling pragmas in BPF selftests 



[Changes from V1:
- Avoid conflict by rebasing with latest master.]

Some BPF tests use loop unrolling compiler pragmas that are clang
specific and not supported by GCC.  These pragmas, along with their
GCC equivalences are:

  #pragma clang loop unroll_count(N)
  #pragma GCC unroll N

  #pragma clang loop unroll(full)
  #pragma GCC unroll 65534

  #pragma clang loop unroll(disable)
  #pragma GCC unroll 1

  #pragma unroll [aka #pragma clang loop unroll(enable)]
  There is no GCC equivalence to this pragma.  It enables unrolling on
  loops that the compiler would not ordinarily unroll even with
  -O2|-funroll-loops, but it is not equivalent to full unrolling
  either.

This patch adds a new header progs/bpf_compiler.h that defines the
following macros, which correspond to each pair of compiler-specific
pragmas above:

  __pragma_loop_unroll_count(N)
  __pragma_loop_unroll_full
  __pragma_loop_no_unroll
  __pragma_loop_unroll

The selftests using loop unrolling pragmas are then changed to include
the header and use these macros in place of the explicit pragmas.

Tested in bpf-next master.
No regressions.

Signed-off-by: default avatarJose E. Marchesi <jose.marchesi@oracle.com>
Signed-off-by: default avatarAndrii Nakryiko <andrii@kernel.org>
Acked-by: default avatarYonghong Song <yonghong.song@linux.dev>
Link: https://lore.kernel.org/bpf/20240208203612.29611-1-jose.marchesi@oracle.com
parent fc1c9e40

tools/testing/selftests/bpf/progs/bpf_compiler.h

0 → 100644
+33 −0
Original line number Diff line number Diff line 
/* SPDX-License-Identifier: GPL-2.0 */

#ifndef __BPF_COMPILER_H__

#define __BPF_COMPILER_H__



#define DO_PRAGMA_(X) _Pragma(#X)



#if __clang__

#define __pragma_loop_unroll DO_PRAGMA_(clang loop unroll(enable))

#else

/* In GCC -funroll-loops, which is enabled with -O2, should have the

   same impact than the loop-unroll-enable pragma above.  */

#define __pragma_loop_unroll

#endif



#if __clang__

#define __pragma_loop_unroll_count(N) DO_PRAGMA_(clang loop unroll_count(N))

#else

#define __pragma_loop_unroll_count(N) DO_PRAGMA_(GCC unroll N)

#endif



#if __clang__

#define __pragma_loop_unroll_full DO_PRAGMA_(clang loop unroll(full))

#else

#define __pragma_loop_unroll_full DO_PRAGMA_(GCC unroll 65534)

#endif



#if __clang__

#define __pragma_loop_no_unroll DO_PRAGMA_(clang loop unroll(disable))

#else

#define __pragma_loop_no_unroll DO_PRAGMA_(GCC unroll 1)

#endif



#endif

tools/testing/selftests/bpf/progs/iters.c

+3 −2
Original line number Diff line number Diff line
@@ -5,6 +5,7 @@ 
#include <linux/bpf.h>

#include <bpf/bpf_helpers.h>

#include "bpf_misc.h"

#include "bpf_compiler.h"



#define ARRAY_SIZE(x) (int)(sizeof(x) / sizeof((x)[0]))
@@ -183,7 +184,7 @@ int iter_pragma_unroll_loop(const void *ctx) 
	MY_PID_GUARD();



	bpf_iter_num_new(&it, 0, 2);

#pragma nounroll

	__pragma_loop_no_unroll

	for (i = 0; i < 3; i++) {

		v = bpf_iter_num_next(&it);

		bpf_printk("ITER_BASIC: E3 VAL: i=%d v=%d", i, v ? *v : -1);
@@ -238,7 +239,7 @@ int iter_multiple_sequential_loops(const void *ctx) 
	bpf_iter_num_destroy(&it);



	bpf_iter_num_new(&it, 0, 2);

#pragma nounroll

	__pragma_loop_no_unroll

	for (i = 0; i < 3; i++) {

		v = bpf_iter_num_next(&it);

		bpf_printk("ITER_BASIC: E3 VAL: i=%d v=%d", i, v ? *v : -1);

tools/testing/selftests/bpf/progs/loop4.c

+3 −1
Original line number Diff line number Diff line
@@ -3,6 +3,8 @@ 
#include <linux/bpf.h>

#include <bpf/bpf_helpers.h>



#include "bpf_compiler.h"



char _license[] SEC("license") = "GPL";



SEC("socket")
@@ -10,7 +12,7 @@ int combinations(volatile struct __sk_buff* skb) 
{

	int ret = 0, i;



#pragma nounroll

	__pragma_loop_no_unroll

	for (i = 0; i < 20; i++)

		if (skb->len)

			ret |= 1 << i;

tools/testing/selftests/bpf/progs/profiler.inc.h

+9 −8
Original line number Diff line number Diff line
@@ -8,6 +8,7 @@ 
#include "profiler.h"

#include "err.h"

#include "bpf_experimental.h"

#include "bpf_compiler.h"



#ifndef NULL

#define NULL 0
@@ -169,7 +170,7 @@ static INLINE int get_var_spid_index(struct var_kill_data_arr_t* arr_struct, 
				     int spid)

{

#ifdef UNROLL

#pragma unroll

	__pragma_loop_unroll

#endif

	for (int i = 0; i < ARRAY_SIZE(arr_struct->array); i++)

		if (arr_struct->array[i].meta.pid == spid)
@@ -185,7 +186,7 @@ static INLINE void populate_ancestors(struct task_struct* task, 


	ancestors_data->num_ancestors = 0;

#ifdef UNROLL

#pragma unroll

	__pragma_loop_unroll

#endif

	for (num_ancestors = 0; num_ancestors < MAX_ANCESTORS; num_ancestors++) {

		parent = BPF_CORE_READ(parent, real_parent);
@@ -212,7 +213,7 @@ static INLINE void* read_full_cgroup_path(struct kernfs_node* cgroup_node, 
	size_t filepart_length;



#ifdef UNROLL

#pragma unroll

	__pragma_loop_unroll

#endif

	for (int i = 0; i < MAX_CGROUPS_PATH_DEPTH; i++) {

		filepart_length =
@@ -261,7 +262,7 @@ static INLINE void* populate_cgroup_info(struct cgroup_data_t* cgroup_data, 
		int cgrp_id = bpf_core_enum_value(enum cgroup_subsys_id___local,

						  pids_cgrp_id___local);

#ifdef UNROLL

#pragma unroll

		__pragma_loop_unroll

#endif

		for (int i = 0; i < CGROUP_SUBSYS_COUNT; i++) {

			struct cgroup_subsys_state* subsys =
@@ -402,7 +403,7 @@ static INLINE int trace_var_sys_kill(void* ctx, int tpid, int sig) 
			if (kill_data == NULL)

				return 0;

#ifdef UNROLL

#pragma unroll

			__pragma_loop_unroll

#endif

			for (int i = 0; i < ARRAY_SIZE(arr_struct->array); i++)

				if (arr_struct->array[i].meta.pid == 0) {
@@ -482,7 +483,7 @@ read_absolute_file_path_from_dentry(struct dentry* filp_dentry, void* payload) 
	struct dentry* parent_dentry;



#ifdef UNROLL

#pragma unroll

	__pragma_loop_unroll

#endif

	for (int i = 0; i < MAX_PATH_DEPTH; i++) {

		filepart_length =
@@ -508,7 +509,7 @@ is_ancestor_in_allowed_inodes(struct dentry* filp_dentry) 
{

	struct dentry* parent_dentry;

#ifdef UNROLL

#pragma unroll

	__pragma_loop_unroll

#endif

	for (int i = 0; i < MAX_PATH_DEPTH; i++) {

		u64 dir_ino = BPF_CORE_READ(filp_dentry, d_inode, i_ino);
@@ -629,7 +630,7 @@ int raw_tracepoint__sched_process_exit(void* ctx) 
	struct kernfs_node* proc_kernfs = BPF_CORE_READ(task, cgroups, dfl_cgrp, kn);



#ifdef UNROLL

#pragma unroll

	__pragma_loop_unroll

#endif

	for (int i = 0; i < ARRAY_SIZE(arr_struct->array); i++) {

		struct var_kill_data_t* past_kill_data = &arr_struct->array[i];

tools/testing/selftests/bpf/progs/pyperf.h

+4 −3
Original line number Diff line number Diff line
@@ -8,6 +8,7 @@ 
#include <linux/bpf.h>

#include <bpf/bpf_helpers.h>

#include "bpf_misc.h"

#include "bpf_compiler.h"



#define FUNCTION_NAME_LEN 64

#define FILE_NAME_LEN 128
@@ -298,11 +299,11 @@ int __on_event(struct bpf_raw_tracepoint_args *ctx) 
#if defined(USE_ITER)

/* no for loop, no unrolling */

#elif defined(NO_UNROLL)

#pragma clang loop unroll(disable)

	__pragma_loop_no_unroll

#elif defined(UNROLL_COUNT)

#pragma clang loop unroll_count(UNROLL_COUNT)

	__pragma_loop_unroll_count(UNROLL_COUNT)

#else

#pragma clang loop unroll(full)

	__pragma_loop_unroll_full

#endif /* NO_UNROLL */

		/* Unwind python stack */

#ifdef USE_ITER