[PATCH net-next 0/5] Misc BPF updates

[PATCH net-next 0/5] Misc BPF updates

[Daniel Borkmann]

This set cleans up ldimm64 leftovers from early eBPF days and
adds couple of test cases related to this to the verifier test
suite. It also cleans up the kallsyms spinlock (had same patch
also in queue) by relaxing it through switching to _bh variant.
It fixes up test_progs in relation to htons/ntohs and adds
accessor macros for the percpu tests in test_maps.

Thanks!

Daniel Borkmann (4):
  bpf, x86_64/arm64: remove old ldimm64 artifacts from jits
  bpf: add various test cases to verifier selftests
  bpf: fix _htons occurences in test_progs
  bpf: provide a generic macro for percpu values for selftests

Hannes Frederic Sowa (1):
  bpf: bpf_lock on kallsysms doesn't need to be irqsave

 arch/arm64/net/bpf_jit_comp.c                 |   9 --
 arch/x86/net/bpf_jit_comp.c                   |   7 --
 kernel/bpf/core.c                             |  12 +--
 tools/testing/selftests/bpf/bpf_util.h        |  26 +++++
 tools/testing/selftests/bpf/test_l4lb.c       |  11 +--
 tools/testing/selftests/bpf/test_maps.c       |  37 +++----
 tools/testing/selftests/bpf/test_pkt_access.c |   6 +-
 tools/testing/selftests/bpf/test_progs.c      |  10 +-
 tools/testing/selftests/bpf/test_verifier.c   | 137 ++++++++++++++++++++++++++
 9 files changed, 199 insertions(+), 56 deletions(-)

-- 
1.9.3

[PATCH net-next 1/5] bpf, x86_64/arm64: remove old ldimm64 artifacts from jits

[Daniel Borkmann]

For both cases, the verifier is already rejecting such invalid
formed instructions. Thus, remove these artifacts from old times
and align it with ppc64, sparc64 and s390x JITs that don't have
them in the first place.

Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
---
 arch/arm64/net/bpf_jit_comp.c | 9 ---------
 arch/x86/net/bpf_jit_comp.c   | 7 -------
 2 files changed, 16 deletions(-)

diff --git a/arch/arm64/net/bpf_jit_comp.c b/arch/arm64/net/bpf_jit_comp.c
index a785554..3047368 100644
--- a/arch/arm64/net/bpf_jit_comp.c
+++ b/arch/arm64/net/bpf_jit_comp.c
@@ -604,15 +604,6 @@ static int build_insn(const struct bpf_insn *insn, struct jit_ctx *ctx)
 		const struct bpf_insn insn1 = insn[1];
 		u64 imm64;
 
-		if (insn1.code != 0 || insn1.src_reg != 0 ||
-		    insn1.dst_reg != 0 || insn1.off != 0) {
-			/* Note: verifier in BPF core must catch invalid
-			 * instructions.
-			 */
-			pr_err_once("Invalid BPF_LD_IMM64 instruction\n");
-			return -EINVAL;
-		}
-
 		imm64 = (u64)insn1.imm << 32 | (u32)imm;
 		emit_a64_mov_i64(dst, imm64, ctx);
 
diff --git a/arch/x86/net/bpf_jit_comp.c b/arch/x86/net/bpf_jit_comp.c
index 32322ce..14f840d 100644
--- a/arch/x86/net/bpf_jit_comp.c
+++ b/arch/x86/net/bpf_jit_comp.c
@@ -490,13 +490,6 @@ static int do_jit(struct bpf_prog *bpf_prog, int *addrs, u8 *image,
 			break;
 
 		case BPF_LD | BPF_IMM | BPF_DW:
-			if (insn[1].code != 0 || insn[1].src_reg != 0 ||
-			    insn[1].dst_reg != 0 || insn[1].off != 0) {
-				/* verifier must catch invalid insns */
-				pr_err("invalid BPF_LD_IMM64 insn\n");
-				return -EINVAL;
-			}
-
 			/* optimization: if imm64 is zero, use 'xor <dst>,<dst>'
 			 * to save 7 bytes.
 			 */
-- 
1.9.3

[PATCH net-next 2/5] bpf: add various test cases to verifier selftests

[Daniel Borkmann]

Add several test cases around ldimm64, fp arithmetic and direct
packet access.

Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
---
 tools/testing/selftests/bpf/test_verifier.c | 137 ++++++++++++++++++++++++++++
 1 file changed, 137 insertions(+)

diff --git a/tools/testing/selftests/bpf/test_verifier.c b/tools/testing/selftests/bpf/test_verifier.c
index 95a8d5f..d3395c1 100644
--- a/tools/testing/selftests/bpf/test_verifier.c
+++ b/tools/testing/selftests/bpf/test_verifier.c
@@ -191,6 +191,86 @@ struct test_val {
 		.result = REJECT,
 	},
 	{
+		"test6 ld_imm64",
+		.insns = {
+			BPF_RAW_INSN(BPF_LD | BPF_IMM | BPF_DW, 0, 0, 0, 0),
+			BPF_RAW_INSN(0, 0, 0, 0, 0),
+			BPF_EXIT_INSN(),
+		},
+		.result = ACCEPT,
+	},
+	{
+		"test7 ld_imm64",
+		.insns = {
+			BPF_RAW_INSN(BPF_LD | BPF_IMM | BPF_DW, 0, 0, 0, 1),
+			BPF_RAW_INSN(0, 0, 0, 0, 1),
+			BPF_EXIT_INSN(),
+		},
+		.result = ACCEPT,
+	},
+	{
+		"test8 ld_imm64",
+		.insns = {
+			BPF_RAW_INSN(BPF_LD | BPF_IMM | BPF_DW, 0, 0, 1, 1),
+			BPF_RAW_INSN(0, 0, 0, 0, 1),
+			BPF_EXIT_INSN(),
+		},
+		.errstr = "uses reserved fields",
+		.result = REJECT,
+	},
+	{
+		"test9 ld_imm64",
+		.insns = {
+			BPF_RAW_INSN(BPF_LD | BPF_IMM | BPF_DW, 0, 0, 0, 1),
+			BPF_RAW_INSN(0, 0, 0, 1, 1),
+			BPF_EXIT_INSN(),
+		},
+		.errstr = "invalid bpf_ld_imm64 insn",
+		.result = REJECT,
+	},
+	{
+		"test10 ld_imm64",
+		.insns = {
+			BPF_RAW_INSN(BPF_LD | BPF_IMM | BPF_DW, 0, 0, 0, 1),
+			BPF_RAW_INSN(0, BPF_REG_1, 0, 0, 1),
+			BPF_EXIT_INSN(),
+		},
+		.errstr = "invalid bpf_ld_imm64 insn",
+		.result = REJECT,
+	},
+	{
+		"test11 ld_imm64",
+		.insns = {
+			BPF_RAW_INSN(BPF_LD | BPF_IMM | BPF_DW, 0, 0, 0, 1),
+			BPF_RAW_INSN(0, 0, BPF_REG_1, 0, 1),
+			BPF_EXIT_INSN(),
+		},
+		.errstr = "invalid bpf_ld_imm64 insn",
+		.result = REJECT,
+	},
+	{
+		"test12 ld_imm64",
+		.insns = {
+			BPF_MOV64_IMM(BPF_REG_1, 0),
+			BPF_RAW_INSN(BPF_LD | BPF_IMM | BPF_DW, 0, BPF_REG_1, 0, 1),
+			BPF_RAW_INSN(0, 0, 0, 0, 1),
+			BPF_EXIT_INSN(),
+		},
+		.errstr = "not pointing to valid bpf_map",
+		.result = REJECT,
+	},
+	{
+		"test13 ld_imm64",
+		.insns = {
+			BPF_MOV64_IMM(BPF_REG_1, 0),
+			BPF_RAW_INSN(BPF_LD | BPF_IMM | BPF_DW, 0, BPF_REG_1, 0, 1),
+			BPF_RAW_INSN(0, 0, BPF_REG_1, 0, 1),
+			BPF_EXIT_INSN(),
+		},
+		.errstr = "invalid bpf_ld_imm64 insn",
+		.result = REJECT,
+	},
+	{
 		"no bpf_exit",
 		.insns = {
 			BPF_ALU64_REG(BPF_MOV, BPF_REG_0, BPF_REG_2),
@@ -331,6 +411,30 @@ struct test_val {
 		.result = REJECT,
 	},
 	{
+		"invalid fp arithmetic",
+		/* If this gets ever changed, make sure JITs can deal with it. */
+		.insns = {
+			BPF_MOV64_IMM(BPF_REG_0, 0),
+			BPF_MOV64_REG(BPF_REG_1, BPF_REG_10),
+			BPF_ALU64_IMM(BPF_SUB, BPF_REG_1, 8),
+			BPF_STX_MEM(BPF_DW, BPF_REG_1, BPF_REG_0, 0),
+			BPF_EXIT_INSN(),
+		},
+		.errstr_unpriv = "R1 pointer arithmetic",
+		.result_unpriv = REJECT,
+		.errstr = "R1 invalid mem access",
+		.result = REJECT,
+	},
+	{
+		"non-invalid fp arithmetic",
+		.insns = {
+			BPF_MOV64_IMM(BPF_REG_0, 0),
+			BPF_STX_MEM(BPF_DW, BPF_REG_10, BPF_REG_0, -8),
+			BPF_EXIT_INSN(),
+		},
+		.result = ACCEPT,
+	},
+	{
 		"invalid argument register",
 		.insns = {
 			BPF_RAW_INSN(BPF_JMP | BPF_CALL, 0, 0, 0,
@@ -1801,6 +1905,20 @@ struct test_val {
 		.result = ACCEPT,
 	},
 	{
+		"unpriv: adding of fp",
+		.insns = {
+			BPF_MOV64_IMM(BPF_REG_0, 0),
+			BPF_MOV64_IMM(BPF_REG_1, 0),
+			BPF_ALU64_REG(BPF_ADD, BPF_REG_1, BPF_REG_10),
+			BPF_STX_MEM(BPF_DW, BPF_REG_1, BPF_REG_0, -8),
+			BPF_EXIT_INSN(),
+		},
+		.errstr_unpriv = "pointer arithmetic prohibited",
+		.result_unpriv = REJECT,
+		.errstr = "R1 invalid mem access",
+		.result = REJECT,
+	},
+	{
 		"unpriv: cmp of stack pointer",
 		.insns = {
 			BPF_MOV64_REG(BPF_REG_2, BPF_REG_10),
@@ -2472,6 +2590,25 @@ struct test_val {
 		.prog_type = BPF_PROG_TYPE_SCHED_CLS,
 	},
 	{
+		"direct packet access: test16 (arith on data_end)",
+		.insns = {
+			BPF_LDX_MEM(BPF_W, BPF_REG_2, BPF_REG_1,
+				    offsetof(struct __sk_buff, data)),
+			BPF_LDX_MEM(BPF_W, BPF_REG_3, BPF_REG_1,
+				    offsetof(struct __sk_buff, data_end)),
+			BPF_MOV64_REG(BPF_REG_0, BPF_REG_2),
+			BPF_ALU64_IMM(BPF_ADD, BPF_REG_0, 8),
+			BPF_ALU64_IMM(BPF_ADD, BPF_REG_3, 16),
+			BPF_JMP_REG(BPF_JGT, BPF_REG_0, BPF_REG_3, 1),
+			BPF_STX_MEM(BPF_B, BPF_REG_2, BPF_REG_2, 0),
+			BPF_MOV64_IMM(BPF_REG_0, 0),
+			BPF_EXIT_INSN(),
+		},
+		.errstr = "invalid access to packet",
+		.result = REJECT,
+		.prog_type = BPF_PROG_TYPE_SCHED_CLS,
+	},
+	{
 		"helper access to packet: test1, valid packet_ptr range",
 		.insns = {
 			BPF_LDX_MEM(BPF_W, BPF_REG_2, BPF_REG_1,
-- 
1.9.3

[PATCH net-next 3/5] bpf: bpf_lock on kallsysms doesn't need to be irqsave

[Daniel Borkmann]

From: Hannes Frederic Sowa <hannes@stressinduktion.org>

Hannes rightfully spotted that the bpf_lock doesn't need to be
irqsave variant. We never perform any such updates where this
would be necessary (neither right now nor in future), therefore
relax this further.

Signed-off-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
---
 kernel/bpf/core.c | 12 ++++--------
 1 file changed, 4 insertions(+), 8 deletions(-)

diff --git a/kernel/bpf/core.c b/kernel/bpf/core.c
index b4f1cb0..6f81e0f 100644
--- a/kernel/bpf/core.c
+++ b/kernel/bpf/core.c
@@ -394,27 +394,23 @@ static bool bpf_prog_kallsyms_verify_off(const struct bpf_prog *fp)
 
 void bpf_prog_kallsyms_add(struct bpf_prog *fp)
 {
-	unsigned long flags;
-
 	if (!bpf_prog_kallsyms_candidate(fp) ||
 	    !capable(CAP_SYS_ADMIN))
 		return;
 
-	spin_lock_irqsave(&bpf_lock, flags);
+	spin_lock_bh(&bpf_lock);
 	bpf_prog_ksym_node_add(fp->aux);
-	spin_unlock_irqrestore(&bpf_lock, flags);
+	spin_unlock_bh(&bpf_lock);
 }
 
 void bpf_prog_kallsyms_del(struct bpf_prog *fp)
 {
-	unsigned long flags;
-
 	if (!bpf_prog_kallsyms_candidate(fp))
 		return;
 
-	spin_lock_irqsave(&bpf_lock, flags);
+	spin_lock_bh(&bpf_lock);
 	bpf_prog_ksym_node_del(fp->aux);
-	spin_unlock_irqrestore(&bpf_lock, flags);
+	spin_unlock_bh(&bpf_lock);
 }
 
 static struct bpf_prog *bpf_prog_kallsyms_find(unsigned long addr)
-- 
1.9.3

[PATCH net-next 4/5] bpf: fix _htons occurences in test_progs

[Daniel Borkmann]

Dave reported that on sparc test_progs generates buggy swapped
eth->h_proto protocol comparisons:

  10: (15) if r3 == 0xdd86 goto pc+9
   R0=imm2,min_value=2,max_value=2 R1=pkt(id=0,off=0,r=14) R2=pkt_end R3=inv
   R4=pkt(id=0,off=14,r=14) R5=inv56 R10=fp

This is due to the unconditional ...

  #define htons __builtin_bswap16
  #define ntohs __builtin_bswap16

... in test_progs that causes this. Make use of asm/byteorder.h
and use __constant_htons() where possible and only perform the
bswap16 when on little endian in non-constant case.

Fixes: 6882804c916b ("selftests/bpf: add a test for overlapping packet range checks")
Fixes: 37821613626e ("selftests/bpf: add l4 load balancer test based on sched_cls")
Reported-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
---
 tools/testing/selftests/bpf/bpf_util.h        | 19 +++++++++++++++++++
 tools/testing/selftests/bpf/test_l4lb.c       | 11 +++++------
 tools/testing/selftests/bpf/test_pkt_access.c |  6 +++---
 tools/testing/selftests/bpf/test_progs.c      | 10 ++++------
 4 files changed, 31 insertions(+), 15 deletions(-)

diff --git a/tools/testing/selftests/bpf/bpf_util.h b/tools/testing/selftests/bpf/bpf_util.h
index 84a5d18..7de2796 100644
--- a/tools/testing/selftests/bpf/bpf_util.h
+++ b/tools/testing/selftests/bpf/bpf_util.h
@@ -6,6 +6,25 @@
 #include <string.h>
 #include <errno.h>
 
+#include <asm/byteorder.h>
+
+#if __BYTE_ORDER == __LITTLE_ENDIAN
+# define __bpf_ntohs(x)		__builtin_bswap16(x)
+# define __bpf_htons(x)		__builtin_bswap16(x)
+#elif __BYTE_ORDER == __BIG_ENDIAN
+# define __bpf_ntohs(x)		(x)
+# define __bpf_htons(x)		(x)
+#else
+# error "Fix your __BYTE_ORDER?!"
+#endif
+
+#define bpf_htons(x)				\
+	(__builtin_constant_p(x) ?		\
+	 __constant_htons(x) : __bpf_htons(x))
+#define bpf_ntohs(x)				\
+	(__builtin_constant_p(x) ?		\
+	 __constant_ntohs(x) : __bpf_ntohs(x))
+
 static inline unsigned int bpf_num_possible_cpus(void)
 {
 	static const char *fcpu = "/sys/devices/system/cpu/possible";
diff --git a/tools/testing/selftests/bpf/test_l4lb.c b/tools/testing/selftests/bpf/test_l4lb.c
index 368bfe8..b68b212 100644
--- a/tools/testing/selftests/bpf/test_l4lb.c
+++ b/tools/testing/selftests/bpf/test_l4lb.c
@@ -19,9 +19,8 @@
 #include <linux/udp.h>
 #include "bpf_helpers.h"
 #include "test_iptunnel_common.h"
+#include "bpf_util.h"
 
-#define htons __builtin_bswap16
-#define ntohs __builtin_bswap16
 int _version SEC("version") = 1;
 
 static inline __u32 rol32(__u32 word, unsigned int shift)
@@ -355,7 +354,7 @@ static __always_inline int process_packet(void *data, __u64 off, void *data_end,
 		iph_len = sizeof(struct ipv6hdr);
 		protocol = ip6h->nexthdr;
 		pckt.proto = protocol;
-		pkt_bytes = ntohs(ip6h->payload_len);
+		pkt_bytes = bpf_ntohs(ip6h->payload_len);
 		off += iph_len;
 		if (protocol == IPPROTO_FRAGMENT) {
 			return TC_ACT_SHOT;
@@ -377,7 +376,7 @@ static __always_inline int process_packet(void *data, __u64 off, void *data_end,
 
 		protocol = iph->protocol;
 		pckt.proto = protocol;
-		pkt_bytes = ntohs(iph->tot_len);
+		pkt_bytes = bpf_ntohs(iph->tot_len);
 		off += IPV4_HDR_LEN_NO_OPT;
 
 		if (iph->frag_off & PCKT_FRAGMENTED)
@@ -464,9 +463,9 @@ int balancer_ingress(struct __sk_buff *ctx)
 	if (data + nh_off > data_end)
 		return TC_ACT_SHOT;
 	eth_proto = eth->eth_proto;
-	if (eth_proto == htons(ETH_P_IP))
+	if (eth_proto == bpf_htons(ETH_P_IP))
 		return process_packet(data, nh_off, data_end, false, ctx);
-	else if (eth_proto == htons(ETH_P_IPV6))
+	else if (eth_proto == bpf_htons(ETH_P_IPV6))
 		return process_packet(data, nh_off, data_end, true, ctx);
 	else
 		return TC_ACT_SHOT;
diff --git a/tools/testing/selftests/bpf/test_pkt_access.c b/tools/testing/selftests/bpf/test_pkt_access.c
index fd1e083..7113005 100644
--- a/tools/testing/selftests/bpf/test_pkt_access.c
+++ b/tools/testing/selftests/bpf/test_pkt_access.c
@@ -14,8 +14,8 @@
 #include <linux/tcp.h>
 #include <linux/pkt_cls.h>
 #include "bpf_helpers.h"
+#include "bpf_util.h"
 
-#define _htons __builtin_bswap16
 #define barrier() __asm__ __volatile__("": : :"memory")
 int _version SEC("version") = 1;
 
@@ -32,7 +32,7 @@ int process(struct __sk_buff *skb)
 	if (eth + 1 > data_end)
 		return TC_ACT_SHOT;
 
-	if (eth->h_proto == _htons(ETH_P_IP)) {
+	if (eth->h_proto == bpf_htons(ETH_P_IP)) {
 		struct iphdr *iph = (struct iphdr *)(eth + 1);
 
 		if (iph + 1 > data_end)
@@ -40,7 +40,7 @@ int process(struct __sk_buff *skb)
 		ihl_len = iph->ihl * 4;
 		proto = iph->protocol;
 		tcp = (struct tcphdr *)((void *)(iph) + ihl_len);
-	} else if (eth->h_proto == _htons(ETH_P_IPV6)) {
+	} else if (eth->h_proto == bpf_htons(ETH_P_IPV6)) {
 		struct ipv6hdr *ip6h = (struct ipv6hdr *)(eth + 1);
 
 		if (ip6h + 1 > data_end)
diff --git a/tools/testing/selftests/bpf/test_progs.c b/tools/testing/selftests/bpf/test_progs.c
index 5275d4a..7c2d899 100644
--- a/tools/testing/selftests/bpf/test_progs.c
+++ b/tools/testing/selftests/bpf/test_progs.c
@@ -30,8 +30,6 @@
 #include "test_iptunnel_common.h"
 #include "bpf_util.h"
 
-#define _htons __builtin_bswap16
-
 static int error_cnt, pass_cnt;
 
 #define MAGIC_BYTES 123
@@ -42,10 +40,10 @@
 	struct iphdr iph;
 	struct tcphdr tcp;
 } __packed pkt_v4 = {
-	.eth.h_proto = _htons(ETH_P_IP),
+	.eth.h_proto = bpf_htons(ETH_P_IP),
 	.iph.ihl = 5,
 	.iph.protocol = 6,
-	.iph.tot_len = _htons(MAGIC_BYTES),
+	.iph.tot_len = bpf_htons(MAGIC_BYTES),
 	.tcp.urg_ptr = 123,
 };
 
@@ -55,9 +53,9 @@
 	struct ipv6hdr iph;
 	struct tcphdr tcp;
 } __packed pkt_v6 = {
-	.eth.h_proto = _htons(ETH_P_IPV6),
+	.eth.h_proto = bpf_htons(ETH_P_IPV6),
 	.iph.nexthdr = 6,
-	.iph.payload_len = _htons(MAGIC_BYTES),
+	.iph.payload_len = bpf_htons(MAGIC_BYTES),
 	.tcp.urg_ptr = 123,
 };
 
-- 
1.9.3

[PATCH net-next 5/5] bpf: provide a generic macro for percpu values for selftests

[Daniel Borkmann]

To overcome bugs as described and fixed in 89087c456fb5 ("bpf: Fix
values type used in test_maps"), provide a generic BPF_DECLARE_PERCPU()
and bpf_percpu() accessor macro for all percpu map values used in
tests.

Declaring variables works as follows (also works for structs):

  BPF_DECLARE_PERCPU(uint32_t, my_value);

They can then be accessed normally as uint32_t type through:

  bpf_percpu(my_value, <cpu_nr>)

For example:

  bpf_percpu(my_value, 0)++;

Implicitly, we make sure that the passed type is allocated and aligned
by gcc at least on a 8-byte boundary, so that it works together with
the map lookup/update syscall for percpu maps. We use it as a usage
example in test_maps, so that others are free to adapt this into their
code when necessary.

Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
---
 tools/testing/selftests/bpf/bpf_util.h  |  7 +++++++
 tools/testing/selftests/bpf/test_maps.c | 37 ++++++++++++++++++---------------
 2 files changed, 27 insertions(+), 17 deletions(-)

diff --git a/tools/testing/selftests/bpf/bpf_util.h b/tools/testing/selftests/bpf/bpf_util.h
index 7de2796..369e7d7 100644
--- a/tools/testing/selftests/bpf/bpf_util.h
+++ b/tools/testing/selftests/bpf/bpf_util.h
@@ -54,4 +54,11 @@ static inline unsigned int bpf_num_possible_cpus(void)
 	return possible_cpus;
 }
 
+#define __bpf_percpu_val_align	__attribute__((__aligned__(8)))
+
+#define BPF_DECLARE_PERCPU(type, name)				\
+	struct { type v; /* padding */ } __bpf_percpu_val_align	\
+		name[bpf_num_possible_cpus()]
+#define bpf_percpu(name, cpu) name[(cpu)].v
+
 #endif /* __BPF_UTIL__ */
diff --git a/tools/testing/selftests/bpf/test_maps.c b/tools/testing/selftests/bpf/test_maps.c
index a977c4f..9331452 100644
--- a/tools/testing/selftests/bpf/test_maps.c
+++ b/tools/testing/selftests/bpf/test_maps.c
@@ -137,20 +137,20 @@ static void test_hashmap_sizes(int task, void *data)
 static void test_hashmap_percpu(int task, void *data)
 {
 	unsigned int nr_cpus = bpf_num_possible_cpus();
-	long long value[nr_cpus];
+	BPF_DECLARE_PERCPU(long, value);
 	long long key, next_key, first_key;
 	int expected_key_mask = 0;
 	int fd, i;
 
 	fd = bpf_create_map(BPF_MAP_TYPE_PERCPU_HASH, sizeof(key),
-			    sizeof(value[0]), 2, map_flags);
+			    sizeof(bpf_percpu(value, 0)), 2, map_flags);
 	if (fd < 0) {
 		printf("Failed to create hashmap '%s'!\n", strerror(errno));
 		exit(1);
 	}
 
 	for (i = 0; i < nr_cpus; i++)
-		value[i] = i + 100;
+		bpf_percpu(value, i) = i + 100;
 
 	key = 1;
 	/* Insert key=1 element. */
@@ -170,8 +170,9 @@ static void test_hashmap_percpu(int task, void *data)
 	/* Check that key=1 can be found. Value could be 0 if the lookup
 	 * was run from a different CPU.
 	 */
-	value[0] = 1;
-	assert(bpf_map_lookup_elem(fd, &key, value) == 0 && value[0] == 100);
+	bpf_percpu(value, 0) = 1;
+	assert(bpf_map_lookup_elem(fd, &key, value) == 0 &&
+	       bpf_percpu(value, 0) == 100);
 
 	key = 2;
 	/* Check that key=2 is not found. */
@@ -211,7 +212,7 @@ static void test_hashmap_percpu(int task, void *data)
 		assert(bpf_map_lookup_elem(fd, &next_key, value) == 0);
 
 		for (i = 0; i < nr_cpus; i++)
-			assert(value[i] == i + 100);
+			assert(bpf_percpu(value, i) == i + 100);
 
 		key = next_key;
 	}
@@ -296,34 +297,36 @@ static void test_arraymap(int task, void *data)
 static void test_arraymap_percpu(int task, void *data)
 {
 	unsigned int nr_cpus = bpf_num_possible_cpus();
+	BPF_DECLARE_PERCPU(long, values);
 	int key, next_key, fd, i;
-	long long values[nr_cpus];
 
 	fd = bpf_create_map(BPF_MAP_TYPE_PERCPU_ARRAY, sizeof(key),
-			    sizeof(values[0]), 2, 0);
+			    sizeof(bpf_percpu(values, 0)), 2, 0);
 	if (fd < 0) {
 		printf("Failed to create arraymap '%s'!\n", strerror(errno));
 		exit(1);
 	}
 
 	for (i = 0; i < nr_cpus; i++)
-		values[i] = i + 100;
+		bpf_percpu(values, i) = i + 100;
 
 	key = 1;
 	/* Insert key=1 element. */
 	assert(bpf_map_update_elem(fd, &key, values, BPF_ANY) == 0);
 
-	values[0] = 0;
+	bpf_percpu(values, 0) = 0;
 	assert(bpf_map_update_elem(fd, &key, values, BPF_NOEXIST) == -1 &&
 	       errno == EEXIST);
 
 	/* Check that key=1 can be found. */
-	assert(bpf_map_lookup_elem(fd, &key, values) == 0 && values[0] == 100);
+	assert(bpf_map_lookup_elem(fd, &key, values) == 0 &&
+	       bpf_percpu(values, 0) == 100);
 
 	key = 0;
 	/* Check that key=0 is also found and zero initialized. */
 	assert(bpf_map_lookup_elem(fd, &key, values) == 0 &&
-	       values[0] == 0 && values[nr_cpus - 1] == 0);
+	       bpf_percpu(values, 0) == 0 &&
+	       bpf_percpu(values, nr_cpus - 1) == 0);
 
 	/* Check that key=2 cannot be inserted due to max_entries limit. */
 	key = 2;
@@ -353,15 +356,15 @@ static void test_arraymap_percpu(int task, void *data)
 static void test_arraymap_percpu_many_keys(void)
 {
 	unsigned int nr_cpus = bpf_num_possible_cpus();
+	BPF_DECLARE_PERCPU(long, values);
 	/* nr_keys is not too large otherwise the test stresses percpu
 	 * allocator more than anything else
 	 */
 	unsigned int nr_keys = 2000;
-	long long values[nr_cpus];
 	int key, fd, i;
 
 	fd = bpf_create_map(BPF_MAP_TYPE_PERCPU_ARRAY, sizeof(key),
-			    sizeof(values[0]), nr_keys, 0);
+			    sizeof(bpf_percpu(values, 0)), nr_keys, 0);
 	if (fd < 0) {
 		printf("Failed to create per-cpu arraymap '%s'!\n",
 		       strerror(errno));
@@ -369,19 +372,19 @@ static void test_arraymap_percpu_many_keys(void)
 	}
 
 	for (i = 0; i < nr_cpus; i++)
-		values[i] = i + 10;
+		bpf_percpu(values, i) = i + 10;
 
 	for (key = 0; key < nr_keys; key++)
 		assert(bpf_map_update_elem(fd, &key, values, BPF_ANY) == 0);
 
 	for (key = 0; key < nr_keys; key++) {
 		for (i = 0; i < nr_cpus; i++)
-			values[i] = 0;
+			bpf_percpu(values, i) = 0;
 
 		assert(bpf_map_lookup_elem(fd, &key, values) == 0);
 
 		for (i = 0; i < nr_cpus; i++)
-			assert(values[i] == i + 10);
+			assert(bpf_percpu(values, i) == i + 10);
 	}
 
 	close(fd);
-- 
1.9.3

Re: [PATCH net-next 0/5] Misc BPF updates

[David Miller]

From: Daniel Borkmann <daniel@iogearbox.net>
Date: Thu, 27 Apr 2017 01:39:30 +0200

> This set cleans up ldimm64 leftovers from early eBPF days and
> adds couple of test cases related to this to the verifier test
> suite. It also cleans up the kallsyms spinlock (had same patch
> also in queue) by relaxing it through switching to _bh variant.
> It fixes up test_progs in relation to htons/ntohs and adds
> accessor macros for the percpu tests in test_maps.

Series applied, thanks for fixing all of this stuff.