diff --git a/tools/testing/selftests/bpf/prog_tests/bpf_verif_scale.c b/tools/testing/selftests/bpf/prog_tests/bpf_verif_scale.c
index e9f2f12ba06b54c31d4cddf62a38417ca4bfddab..e698ee6bb6c228919f7fdd1291cedc91834b146d 100644
--- a/tools/testing/selftests/bpf/prog_tests/bpf_verif_scale.c
+++ b/tools/testing/selftests/bpf/prog_tests/bpf_verif_scale.c
@@ -49,6 +49,7 @@ void test_bpf_verif_scale(void)
 		{ "test_verif_scale3.o", BPF_PROG_TYPE_SCHED_CLS },
 
 		{ "pyperf_global.o", BPF_PROG_TYPE_RAW_TRACEPOINT },
+		{ "pyperf_subprogs.o", BPF_PROG_TYPE_RAW_TRACEPOINT },
 
 		/* full unroll by llvm */
 		{ "pyperf50.o", BPF_PROG_TYPE_RAW_TRACEPOINT },
@@ -86,6 +87,9 @@ void test_bpf_verif_scale(void)
 		{ "strobemeta_nounroll1.o", BPF_PROG_TYPE_RAW_TRACEPOINT },
 		{ "strobemeta_nounroll2.o", BPF_PROG_TYPE_RAW_TRACEPOINT },
 
+		/* non-inlined subprogs */
+		{ "strobemeta_subprogs.o", BPF_PROG_TYPE_RAW_TRACEPOINT },
+
 		{ "test_sysctl_loop1.o", BPF_PROG_TYPE_CGROUP_SYSCTL },
 		{ "test_sysctl_loop2.o", BPF_PROG_TYPE_CGROUP_SYSCTL },
 
diff --git a/tools/testing/selftests/bpf/prog_tests/l4lb_all.c b/tools/testing/selftests/bpf/prog_tests/l4lb_all.c
index c2d373e294bbf963aa6df836266110737e866875..8073105548ff44380d157296c9035bd1555d883c 100644
--- a/tools/testing/selftests/bpf/prog_tests/l4lb_all.c
+++ b/tools/testing/selftests/bpf/prog_tests/l4lb_all.c
@@ -80,9 +80,8 @@ static void test_l4lb(const char *file)
 
 void test_l4lb_all(void)
 {
-	const char *file1 = "./test_l4lb.o";
-	const char *file2 = "./test_l4lb_noinline.o";
-
-	test_l4lb(file1);
-	test_l4lb(file2);
+	if (test__start_subtest("l4lb_inline"))
+		test_l4lb("test_l4lb.o");
+	if (test__start_subtest("l4lb_noinline"))
+		test_l4lb("test_l4lb_noinline.o");
 }
diff --git a/tools/testing/selftests/bpf/progs/pyperf.h b/tools/testing/selftests/bpf/progs/pyperf.h
index cc615b82b56e046081a8856ed4deb81b24b30597..2fb7adafb6b6437cf1c42de02df1901052ad5c39 100644
--- a/tools/testing/selftests/bpf/progs/pyperf.h
+++ b/tools/testing/selftests/bpf/progs/pyperf.h
@@ -67,7 +67,12 @@ typedef struct {
 	void* co_name; // PyCodeObject.co_name
 } FrameData;
 
-static __always_inline void *get_thread_state(void *tls_base, PidData *pidData)
+#ifdef SUBPROGS
+__noinline
+#else
+__always_inline
+#endif
+static void *get_thread_state(void *tls_base, PidData *pidData)
 {
 	void* thread_state;
 	int key;
@@ -155,7 +160,9 @@ struct {
 } stackmap SEC(".maps");
 
 #ifdef GLOBAL_FUNC
-__attribute__((noinline))
+__noinline
+#elif defined(SUBPROGS)
+static __noinline
 #else
 static __always_inline
 #endif
diff --git a/tools/testing/selftests/bpf/progs/pyperf_subprogs.c b/tools/testing/selftests/bpf/progs/pyperf_subprogs.c
new file mode 100644
index 0000000000000000000000000000000000000000..60e27a7f0cca9fcb694a9069561d1c7b63af51c8
--- /dev/null
+++ b/tools/testing/selftests/bpf/progs/pyperf_subprogs.c
@@ -0,0 +1,5 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (c) 2020 Facebook */
+#define STACK_MAX_LEN 50
+#define SUBPROGS
+#include "pyperf.h"
diff --git a/tools/testing/selftests/bpf/progs/strobemeta.h b/tools/testing/selftests/bpf/progs/strobemeta.h
index ad61b722a9ded0c7e705dd90dc6407f477115660..7de534f38c3f1f245671731c475ab7cafed30bdb 100644
--- a/tools/testing/selftests/bpf/progs/strobemeta.h
+++ b/tools/testing/selftests/bpf/progs/strobemeta.h
@@ -266,8 +266,12 @@ struct tls_index {
 	uint64_t offset;
 };
 
-static __always_inline void *calc_location(struct strobe_value_loc *loc,
-					   void *tls_base)
+#ifdef SUBPROGS
+__noinline
+#else
+__always_inline
+#endif
+static void *calc_location(struct strobe_value_loc *loc, void *tls_base)
 {
 	/*
 	 * tls_mode value is:
@@ -327,10 +331,15 @@ static __always_inline void *calc_location(struct strobe_value_loc *loc,
 		: NULL;
 }
 
-static __always_inline void read_int_var(struct strobemeta_cfg *cfg,
-					 size_t idx, void *tls_base,
-					 struct strobe_value_generic *value,
-					 struct strobemeta_payload *data)
+#ifdef SUBPROGS
+__noinline
+#else
+__always_inline
+#endif
+static void read_int_var(struct strobemeta_cfg *cfg,
+			 size_t idx, void *tls_base,
+			 struct strobe_value_generic *value,
+			 struct strobemeta_payload *data)
 {
 	void *location = calc_location(&cfg->int_locs[idx], tls_base);
 	if (!location)
@@ -440,8 +449,13 @@ static __always_inline void *read_map_var(struct strobemeta_cfg *cfg,
  * read_strobe_meta returns NULL, if no metadata was read; otherwise returns
  * pointer to *right after* payload ends
  */
-static __always_inline void *read_strobe_meta(struct task_struct *task,
-					      struct strobemeta_payload *data)
+#ifdef SUBPROGS
+__noinline
+#else
+__always_inline
+#endif
+static void *read_strobe_meta(struct task_struct *task,
+			      struct strobemeta_payload *data)
 {
 	pid_t pid = bpf_get_current_pid_tgid() >> 32;
 	struct strobe_value_generic value = {0};
diff --git a/tools/testing/selftests/bpf/progs/strobemeta_subprogs.c b/tools/testing/selftests/bpf/progs/strobemeta_subprogs.c
new file mode 100644
index 0000000000000000000000000000000000000000..b6c01f8fc5590788f81f626577204a4f44aac208
--- /dev/null
+++ b/tools/testing/selftests/bpf/progs/strobemeta_subprogs.c
@@ -0,0 +1,10 @@
+// SPDX-License-Identifier: (LGPL-2.1 OR BSD-2-Clause)
+// Copyright (c) 2019 Facebook
+
+#define STROBE_MAX_INTS 2
+#define STROBE_MAX_STRS 25
+#define STROBE_MAX_MAPS 13
+#define STROBE_MAX_MAP_ENTRIES 20
+#define NO_UNROLL
+#define SUBPROGS
+#include "strobemeta.h"
diff --git a/tools/testing/selftests/bpf/progs/test_l4lb_noinline.c b/tools/testing/selftests/bpf/progs/test_l4lb_noinline.c
index 28351936a438edef17bec29cfa7da5b21aef8811..b9e2753f4f91473251243a3b489b85b6c139bb34 100644
--- a/tools/testing/selftests/bpf/progs/test_l4lb_noinline.c
+++ b/tools/testing/selftests/bpf/progs/test_l4lb_noinline.c
@@ -17,9 +17,7 @@
 #include "test_iptunnel_common.h"
 #include <bpf/bpf_endian.h>
 
-int _version SEC("version") = 1;
-
-static __u32 rol32(__u32 word, unsigned int shift)
+static __always_inline __u32 rol32(__u32 word, unsigned int shift)
 {
 	return (word << shift) | (word >> ((-shift) & 31));
 }
@@ -52,7 +50,7 @@ static __u32 rol32(__u32 word, unsigned int shift)
 
 typedef unsigned int u32;
 
-static u32 jhash(const void *key, u32 length, u32 initval)
+static __noinline u32 jhash(const void *key, u32 length, u32 initval)
 {
 	u32 a, b, c;
 	const unsigned char *k = key;
@@ -88,7 +86,7 @@ static u32 jhash(const void *key, u32 length, u32 initval)
 	return c;
 }
 
-static u32 __jhash_nwords(u32 a, u32 b, u32 c, u32 initval)
+static __noinline u32 __jhash_nwords(u32 a, u32 b, u32 c, u32 initval)
 {
 	a += initval;
 	b += initval;
@@ -97,7 +95,7 @@ static u32 __jhash_nwords(u32 a, u32 b, u32 c, u32 initval)
 	return c;
 }
 
-static u32 jhash_2words(u32 a, u32 b, u32 initval)
+static __noinline u32 jhash_2words(u32 a, u32 b, u32 initval)
 {
 	return __jhash_nwords(a, b, 0, initval + JHASH_INITVAL + (2 << 2));
 }
@@ -200,8 +198,7 @@ struct {
 	__type(value, struct ctl_value);
 } ctl_array SEC(".maps");
 
-static __u32 get_packet_hash(struct packet_description *pckt,
-			     bool ipv6)
+static __noinline __u32 get_packet_hash(struct packet_description *pckt, bool ipv6)
 {
 	if (ipv6)
 		return jhash_2words(jhash(pckt->srcv6, 16, MAX_VIPS),
@@ -210,10 +207,10 @@ static __u32 get_packet_hash(struct packet_description *pckt,
 		return jhash_2words(pckt->src, pckt->ports, CH_RINGS_SIZE);
 }
 
-static bool get_packet_dst(struct real_definition **real,
-			   struct packet_description *pckt,
-			   struct vip_meta *vip_info,
-			   bool is_ipv6)
+static __noinline bool get_packet_dst(struct real_definition **real,
+				      struct packet_description *pckt,
+				      struct vip_meta *vip_info,
+				      bool is_ipv6)
 {
 	__u32 hash = get_packet_hash(pckt, is_ipv6);
 	__u32 key = RING_SIZE * vip_info->vip_num + hash % RING_SIZE;
@@ -233,8 +230,8 @@ static bool get_packet_dst(struct real_definition **real,
 	return true;
 }
 
-static int parse_icmpv6(void *data, void *data_end, __u64 off,
-			struct packet_description *pckt)
+static __noinline int parse_icmpv6(void *data, void *data_end, __u64 off,
+				   struct packet_description *pckt)
 {
 	struct icmp6hdr *icmp_hdr;
 	struct ipv6hdr *ip6h;
@@ -255,8 +252,8 @@ static int parse_icmpv6(void *data, void *data_end, __u64 off,
 	return TC_ACT_UNSPEC;
 }
 
-static int parse_icmp(void *data, void *data_end, __u64 off,
-		      struct packet_description *pckt)
+static __noinline int parse_icmp(void *data, void *data_end, __u64 off,
+				 struct packet_description *pckt)
 {
 	struct icmphdr *icmp_hdr;
 	struct iphdr *iph;
@@ -280,8 +277,8 @@ static int parse_icmp(void *data, void *data_end, __u64 off,
 	return TC_ACT_UNSPEC;
 }
 
-static bool parse_udp(void *data, __u64 off, void *data_end,
-		      struct packet_description *pckt)
+static __noinline bool parse_udp(void *data, __u64 off, void *data_end,
+				 struct packet_description *pckt)
 {
 	struct udphdr *udp;
 	udp = data + off;
@@ -299,8 +296,8 @@ static bool parse_udp(void *data, __u64 off, void *data_end,
 	return true;
 }
 
-static bool parse_tcp(void *data, __u64 off, void *data_end,
-		      struct packet_description *pckt)
+static __noinline bool parse_tcp(void *data, __u64 off, void *data_end,
+				 struct packet_description *pckt)
 {
 	struct tcphdr *tcp;
 
@@ -321,8 +318,8 @@ static bool parse_tcp(void *data, __u64 off, void *data_end,
 	return true;
 }
 
-static int process_packet(void *data, __u64 off, void *data_end,
-			  bool is_ipv6, struct __sk_buff *skb)
+static __noinline int process_packet(void *data, __u64 off, void *data_end,
+				     bool is_ipv6, struct __sk_buff *skb)
 {
 	void *pkt_start = (void *)(long)skb->data;
 	struct packet_description pckt = {};