]> git.kernelconcepts.de Git - karo-tx-linux.git/commitdiff
net: filter: BPF testsuite
authorAlexei Starovoitov <ast@plumgrid.com>
Thu, 8 May 2014 21:10:52 +0000 (14:10 -0700)
committerDavid S. Miller <davem@davemloft.net>
Mon, 12 May 2014 04:23:55 +0000 (00:23 -0400)
The testsuite covers classic and internal BPF instructions.
It is particularly useful for JIT compiler developers.
Adds to "net" selftest target.

The testsuite can be used as a set of micro-benchmarks.
It measures execution time of each BPF program in nsec.

This patch adds core framework.

Signed-off-by: Alexei Starovoitov <ast@plumgrid.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
lib/Kconfig.debug
lib/Makefile
lib/test_bpf.c [new file with mode: 0644]
tools/testing/selftests/net/Makefile

index 819ac51202c01006e105f91355d492db6bdd6eb6..423ca319a5f8f099355ee5999063fac7efb9c3fd 100644 (file)
@@ -1620,6 +1620,19 @@ config TEST_USER_COPY
 
          If unsure, say N.
 
+config TEST_BPF
+       tristate "Test BPF filter functionality"
+       default n
+       depends on m
+       help
+         This builds the "test_bpf" module that runs various test vectors
+         against the BPF interpreter or BPF JIT compiler depending on the
+         current setting. This is in particular useful for BPF JIT compiler
+         development, but also to run regression tests against changes in
+         the interpreter code.
+
+         If unsure, say N.
+
 source "samples/Kconfig"
 
 source "lib/Kconfig.kgdb"
index 0cd7b68e1382dee93301898da70a6ca51c3d764e..b2be1ef1e8ece2939550b3ffb029d5bbba24dac4 100644 (file)
@@ -33,6 +33,7 @@ obj-y += kstrtox.o
 obj-$(CONFIG_TEST_KSTRTOX) += test-kstrtox.o
 obj-$(CONFIG_TEST_MODULE) += test_module.o
 obj-$(CONFIG_TEST_USER_COPY) += test_user_copy.o
+obj-$(CONFIG_TEST_BPF) += test_bpf.o
 
 ifeq ($(CONFIG_DEBUG_KOBJECT),y)
 CFLAGS_kobject.o += -DDEBUG
diff --git a/lib/test_bpf.c b/lib/test_bpf.c
new file mode 100644 (file)
index 0000000..9f25dc1
--- /dev/null
@@ -0,0 +1,322 @@
+/*
+ * Testsuite for BPF interpreter and BPF JIT compiler
+ *
+ * Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of version 2 of the GNU General Public
+ * License as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * General Public License for more details.
+ */
+
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+
+#include <linux/init.h>
+#include <linux/module.h>
+#include <linux/filter.h>
+#include <linux/skbuff.h>
+#include <linux/netdevice.h>
+#include <linux/if_vlan.h>
+
+#define MAX_SUBTESTS   3
+#define MAX_DATA       128
+#define MAX_INSNS      512
+#define MAX_K          0xffffFFFF
+
+/* define few constants used to init test 'skb' */
+#define SKB_TYPE       3
+#define SKB_MARK       0x1234aaaa
+#define SKB_HASH       0x1234aaab
+#define SKB_QUEUE_MAP  123
+#define SKB_VLAN_TCI   0xffff
+#define SKB_DEV_IFINDEX        577
+#define SKB_DEV_TYPE   588
+
+/* redefine REGs to make tests less verbose */
+#define R0 BPF_REG_0
+#define R1 BPF_REG_1
+#define R2 BPF_REG_2
+#define R3 BPF_REG_3
+#define R4 BPF_REG_4
+#define R5 BPF_REG_5
+#define R6 BPF_REG_6
+#define R7 BPF_REG_7
+#define R8 BPF_REG_8
+#define R9 BPF_REG_9
+#define R10 BPF_REG_10
+
+struct bpf_test {
+       const char *descr;
+       union {
+               struct sock_filter insns[MAX_INSNS];
+               struct sock_filter_int insns_int[MAX_INSNS];
+       };
+       enum {
+               NO_DATA,
+               EXPECTED_FAIL,
+               SKB,
+               SKB_INT
+       } data_type;
+       __u8 data[MAX_DATA];
+       struct {
+               int data_size;
+               __u32 result;
+       } test[MAX_SUBTESTS];
+};
+
+static struct bpf_test tests[] = {
+       {
+               "TAX",
+               .insns = {
+                       BPF_STMT(BPF_LD | BPF_IMM, 1),
+                       BPF_STMT(BPF_MISC | BPF_TAX, 0),
+                       BPF_STMT(BPF_LD | BPF_IMM, 2),
+                       BPF_STMT(BPF_ALU | BPF_ADD | BPF_X, 0),
+                       BPF_STMT(BPF_ALU | BPF_NEG, 0), /* A == -3 */
+                       BPF_STMT(BPF_MISC | BPF_TAX, 0),
+                       BPF_STMT(BPF_LD | BPF_LEN, 0),
+                       BPF_STMT(BPF_ALU | BPF_ADD | BPF_X, 0),
+                       BPF_STMT(BPF_MISC | BPF_TAX, 0), /* X == len - 3 */
+                       BPF_STMT(BPF_LD | BPF_B | BPF_IND, 1),
+                       BPF_STMT(BPF_RET | BPF_A, 0)
+               },
+               SKB,
+               { 10, 20, 30, 40, 50 },
+               { { 2, 10 }, { 3, 20 }, { 4, 30 } },
+       },
+       {
+               "tcpdump port 22",
+               .insns = {
+                       { 0x28,  0,  0, 0x0000000c },
+                       { 0x15,  0,  8, 0x000086dd },
+                       { 0x30,  0,  0, 0x00000014 },
+                       { 0x15,  2,  0, 0x00000084 },
+                       { 0x15,  1,  0, 0x00000006 },
+                       { 0x15,  0, 17, 0x00000011 },
+                       { 0x28,  0,  0, 0x00000036 },
+                       { 0x15, 14,  0, 0x00000016 },
+                       { 0x28,  0,  0, 0x00000038 },
+                       { 0x15, 12, 13, 0x00000016 },
+                       { 0x15,  0, 12, 0x00000800 },
+                       { 0x30,  0,  0, 0x00000017 },
+                       { 0x15,  2,  0, 0x00000084 },
+                       { 0x15,  1,  0, 0x00000006 },
+                       { 0x15,  0,  8, 0x00000011 },
+                       { 0x28,  0,  0, 0x00000014 },
+                       { 0x45,  6,  0, 0x00001fff },
+                       { 0xb1,  0,  0, 0x0000000e },
+                       { 0x48,  0,  0, 0x0000000e },
+                       { 0x15,  2,  0, 0x00000016 },
+                       { 0x48,  0,  0, 0x00000010 },
+                       { 0x15,  0,  1, 0x00000016 },
+                       { 0x06,  0,  0, 0x0000ffff },
+                       { 0x06,  0,  0, 0x00000000 },
+               },
+               SKB,
+               /* 3c:07:54:43:e5:76 > 10:bf:48:d6:43:d6, ethertype IPv4(0x0800)
+                * length 114: 10.1.1.149.49700 > 10.1.2.10.22: Flags [P.],
+                * seq 1305692979:1305693027, ack 3650467037, win 65535,
+                * options [nop,nop,TS val 2502645400 ecr 3971138], length 48
+                */
+               { 0x10, 0xbf, 0x48, 0xd6, 0x43, 0xd6,
+                 0x3c, 0x07, 0x54, 0x43, 0xe5, 0x76,
+                 0x08, 0x00,
+                 0x45, 0x10, 0x00, 0x64, 0x75, 0xb5,
+                 0x40, 0x00, 0x40, 0x06, 0xad, 0x2e, /* IP header */
+                 0x0a, 0x01, 0x01, 0x95, /* ip src */
+                 0x0a, 0x01, 0x02, 0x0a, /* ip dst */
+                 0xc2, 0x24,
+                 0x00, 0x16 /* dst port */ },
+               { { 10, 0 }, { 30, 0 }, { 100, 65535 } },
+       },
+       {
+               "INT: DIV + ABS",
+               .insns_int = {
+                       BPF_ALU64_REG(BPF_MOV, R6, R1),
+                       BPF_LD_ABS(BPF_B, 3),
+                       BPF_ALU64_IMM(BPF_MOV, R2, 2),
+                       BPF_ALU32_REG(BPF_DIV, R0, R2),
+                       BPF_ALU64_REG(BPF_MOV, R8, R0),
+                       BPF_LD_ABS(BPF_B, 4),
+                       BPF_ALU64_REG(BPF_ADD, R8, R0),
+                       BPF_LD_IND(BPF_B, R8, -70),
+                       BPF_EXIT_INSN(),
+               },
+               SKB_INT,
+               { 10, 20, 30, 40, 50 },
+               { { 4, 0 }, { 5, 10 } }
+       },
+       {
+               "check: missing ret",
+               .insns = {
+                       BPF_STMT(BPF_LD | BPF_IMM, 1),
+               },
+               EXPECTED_FAIL,
+               { },
+               { }
+       },
+};
+
+static int get_length(struct sock_filter *fp)
+{
+       int len = 0;
+
+       while (fp->code != 0 || fp->k != 0) {
+               fp++;
+               len++;
+       }
+
+       return len;
+}
+
+struct net_device dev;
+struct sk_buff *populate_skb(char *buf, int size)
+{
+       struct sk_buff *skb;
+
+       if (size >= MAX_DATA)
+               return NULL;
+
+       skb = alloc_skb(MAX_DATA, GFP_KERNEL);
+       if (!skb)
+               return NULL;
+
+       memcpy(__skb_put(skb, size), buf, size);
+       skb_reset_mac_header(skb);
+       skb->protocol = htons(ETH_P_IP);
+       skb->pkt_type = SKB_TYPE;
+       skb->mark = SKB_MARK;
+       skb->hash = SKB_HASH;
+       skb->queue_mapping = SKB_QUEUE_MAP;
+       skb->vlan_tci = SKB_VLAN_TCI;
+       skb->dev = &dev;
+       skb->dev->ifindex = SKB_DEV_IFINDEX;
+       skb->dev->type = SKB_DEV_TYPE;
+       skb_set_network_header(skb, min(size, ETH_HLEN));
+
+       return skb;
+}
+
+static int run_one(struct sk_filter *fp, struct bpf_test *t)
+{
+       u64 start, finish, res, cnt = 100000;
+       int err_cnt = 0, err, i, j;
+       u32 ret = 0;
+       void *data;
+
+       for (i = 0; i < MAX_SUBTESTS; i++) {
+               if (t->test[i].data_size == 0 &&
+                   t->test[i].result == 0)
+                       break;
+               if (t->data_type == SKB ||
+                   t->data_type == SKB_INT) {
+                       data = populate_skb(t->data, t->test[i].data_size);
+                       if (!data)
+                               return -ENOMEM;
+               } else {
+                       data = NULL;
+               }
+
+               start = ktime_to_us(ktime_get());
+               for (j = 0; j < cnt; j++)
+                       ret = SK_RUN_FILTER(fp, data);
+               finish = ktime_to_us(ktime_get());
+
+               res = (finish - start) * 1000;
+               do_div(res, cnt);
+
+               err = ret != t->test[i].result;
+               if (!err)
+                       pr_cont("%lld ", res);
+
+               if (t->data_type == SKB || t->data_type == SKB_INT)
+                       kfree_skb(data);
+
+               if (err) {
+                       pr_cont("ret %d != %d ", ret, t->test[i].result);
+                       err_cnt++;
+               }
+       }
+
+       return err_cnt;
+}
+
+static __init int test_bpf(void)
+{
+       struct sk_filter *fp, *fp_ext = NULL;
+       struct sock_fprog fprog;
+       int err, i, err_cnt = 0;
+
+       for (i = 0; i < ARRAY_SIZE(tests); i++) {
+               pr_info("#%d %s ", i, tests[i].descr);
+
+               fprog.filter = tests[i].insns;
+               fprog.len = get_length(fprog.filter);
+
+               if (tests[i].data_type == SKB_INT) {
+                       fp_ext = kzalloc(4096, GFP_KERNEL);
+                       if (!fp_ext)
+                               return -ENOMEM;
+                       fp = fp_ext;
+                       memcpy(fp_ext->insns, tests[i].insns_int,
+                              fprog.len * 8);
+                       fp->len = fprog.len;
+                       fp->bpf_func = sk_run_filter_int_skb;
+               } else {
+                       err = sk_unattached_filter_create(&fp, &fprog);
+                       if (tests[i].data_type == EXPECTED_FAIL) {
+                               if (err == -EINVAL) {
+                                       pr_cont("PASS\n");
+                                       continue;
+                               } else {
+                                       pr_cont("UNEXPECTED_PASS\n");
+                                       /* verifier didn't reject the test
+                                        * that's bad enough, just return
+                                        */
+                                       return -EINVAL;
+                               }
+                       }
+                       if (err) {
+                               pr_cont("FAIL to attach err=%d len=%d\n",
+                                       err, fprog.len);
+                               return err;
+                       }
+               }
+
+               err = run_one(fp, &tests[i]);
+
+               if (tests[i].data_type != SKB_INT)
+                       sk_unattached_filter_destroy(fp);
+               else
+                       kfree(fp);
+
+               if (err) {
+                       pr_cont("FAIL %d\n", err);
+                       err_cnt++;
+               } else {
+                       pr_cont("PASS\n");
+               }
+       }
+
+       if (err_cnt)
+               return -EINVAL;
+       else
+               return 0;
+}
+
+static int __init test_bpf_init(void)
+{
+       return test_bpf();
+}
+
+static void __exit test_bpf_exit(void)
+{
+}
+
+module_init(test_bpf_init);
+module_exit(test_bpf_exit);
+MODULE_LICENSE("GPL");
index 750512ba2c8846d2d1275abb796b425fdd4a987d..c7493b8f9b0ee5b951d73d64e7502e809be2d5ed 100644 (file)
@@ -14,6 +14,12 @@ all: $(NET_PROGS)
 run_tests: all
        @/bin/sh ./run_netsocktests || echo "sockettests: [FAIL]"
        @/bin/sh ./run_afpackettests || echo "afpackettests: [FAIL]"
-
+       @if /sbin/modprobe test_bpf ; then \
+               /sbin/rmmod test_bpf; \
+               echo "test_bpf: ok"; \
+       else \
+               echo "test_bpf: [FAIL]"; \
+               exit 1; \
+       fi
 clean:
        $(RM) $(NET_PROGS)