bpf: introduce BPF_MAP_TYPE_STACK_TRACE

add new map type to store stack traces and corresponding helper
bpf_get_stackid(ctx, map, flags) - walk user or kernel stack and return id
@ctx: struct pt_regs*
@map: pointer to stack_trace map
@flags: bits 0-7 - numer of stack frames to skip
        bit 8 - collect user stack instead of kernel
        bit 9 - compare stacks by hash only
        bit 10 - if two different stacks hash into the same stackid
                 discard old
        other bits - reserved
Return: >= 0 stackid on success or negative error

stackid is a 32-bit integer handle that can be further combined with
other data (including other stackid) and used as a key into maps.

Userspace will access stackmap using standard lookup/delete syscall commands to
retrieve full stack trace for given stackid.

Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

diff --git a/include/linux/bpf.h b/include/linux/bpf.h
index 90ee6ab24bc53badebf5ab0f6362de1d4ad2062d..0cadbb7456c0b9ff98dd2ce7fa950166aa12eee4 100644 (file)
--- a/include/linux/bpf.h
+++ b/include/linux/bpf.h
@@ -237,6 +237,7 @@ extern const struct bpf_func_proto bpf_get_current_uid_gid_proto;
 extern const struct bpf_func_proto bpf_get_current_comm_proto;
 extern const struct bpf_func_proto bpf_skb_vlan_push_proto;
 extern const struct bpf_func_proto bpf_skb_vlan_pop_proto;
+extern const struct bpf_func_proto bpf_get_stackid_proto;
 
 /* Shared helpers among cBPF and eBPF. */
 void bpf_user_rnd_init_once(void);

diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
index 2ee0fde1bf9649739e8663dc480742fdf0ede72f..d3e77da8e9e8f355693dc39b71f7c59e8c77c7b6 100644 (file)
--- a/include/uapi/linux/bpf.h
+++ b/include/uapi/linux/bpf.h
@@ -83,6 +83,7 @@ enum bpf_map_type {
        BPF_MAP_TYPE_PERF_EVENT_ARRAY,
        BPF_MAP_TYPE_PERCPU_HASH,
        BPF_MAP_TYPE_PERCPU_ARRAY,
+       BPF_MAP_TYPE_STACK_TRACE,
 };
 
 enum bpf_prog_type {
@@ -272,6 +273,20 @@ enum bpf_func_id {
         */
        BPF_FUNC_perf_event_output,
        BPF_FUNC_skb_load_bytes,
+
+       /**
+        * bpf_get_stackid(ctx, map, flags) - walk user or kernel stack and return id
+        * @ctx: struct pt_regs*
+        * @map: pointer to stack_trace map
+        * @flags: bits 0-7 - numer of stack frames to skip
+        *         bit 8 - collect user stack instead of kernel
+        *         bit 9 - compare stacks by hash only
+        *         bit 10 - if two different stacks hash into the same stackid
+        *                  discard old
+        *         other bits - reserved
+        * Return: >= 0 stackid on success or negative error
+        */
+       BPF_FUNC_get_stackid,
        __BPF_FUNC_MAX_ID,
 };
 
@@ -294,6 +309,12 @@ enum bpf_func_id {
 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
 #define BPF_F_TUNINFO_IPV6             (1ULL << 0)
 
+/* BPF_FUNC_get_stackid flags. */
+#define BPF_F_SKIP_FIELD_MASK          0xffULL
+#define BPF_F_USER_STACK               (1ULL << 8)
+#define BPF_F_FAST_STACK_CMP           (1ULL << 9)
+#define BPF_F_REUSE_STACKID            (1ULL << 10)
+
 /* user accessible mirror of in-kernel sk_buff.
  * new fields can only be added to the end of this structure
  */

diff --git a/kernel/bpf/Makefile b/kernel/bpf/Makefile
index 13272582eee00099a45179d82b47556ea0eb3cf9..8a932d079c24dbdccb87022fa3976d32dd505130 100644 (file)
--- a/kernel/bpf/Makefile
+++ b/kernel/bpf/Makefile
@@ -2,3 +2,6 @@ obj-y := core.o
 
 obj-$(CONFIG_BPF_SYSCALL) += syscall.o verifier.o inode.o helpers.o
 obj-$(CONFIG_BPF_SYSCALL) += hashtab.o arraymap.o
+ifeq ($(CONFIG_PERF_EVENTS),y)
+obj-$(CONFIG_BPF_SYSCALL) += stackmap.o
+endif

diff --git a/kernel/bpf/stackmap.c b/kernel/bpf/stackmap.c
new file mode 100644 (file)
index 0000000..8a60ee1
--- /dev/null
+++ b/kernel/bpf/stackmap.c
@@ -0,0 +1,237 @@
+/* Copyright (c) 2016 Facebook
+ *
+ * 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.
+ */
+#include <linux/bpf.h>
+#include <linux/jhash.h>
+#include <linux/filter.h>
+#include <linux/vmalloc.h>
+#include <linux/stacktrace.h>
+#include <linux/perf_event.h>
+
+struct stack_map_bucket {
+       struct rcu_head rcu;
+       u32 hash;
+       u32 nr;
+       u64 ip[];
+};
+
+struct bpf_stack_map {
+       struct bpf_map map;
+       u32 n_buckets;
+       struct stack_map_bucket __rcu *buckets[];
+};
+
+/* Called from syscall */
+static struct bpf_map *stack_map_alloc(union bpf_attr *attr)
+{
+       u32 value_size = attr->value_size;
+       struct bpf_stack_map *smap;
+       u64 cost, n_buckets;
+       int err;
+
+       if (!capable(CAP_SYS_ADMIN))
+               return ERR_PTR(-EPERM);
+
+       /* check sanity of attributes */
+       if (attr->max_entries == 0 || attr->key_size != 4 ||
+           value_size < 8 || value_size % 8 ||
+           value_size / 8 > PERF_MAX_STACK_DEPTH)
+               return ERR_PTR(-EINVAL);
+
+       /* hash table size must be power of 2 */
+       n_buckets = roundup_pow_of_two(attr->max_entries);
+
+       cost = n_buckets * sizeof(struct stack_map_bucket *) + sizeof(*smap);
+       if (cost >= U32_MAX - PAGE_SIZE)
+               return ERR_PTR(-E2BIG);
+
+       smap = kzalloc(cost, GFP_USER | __GFP_NOWARN);
+       if (!smap) {
+               smap = vzalloc(cost);
+               if (!smap)
+                       return ERR_PTR(-ENOMEM);
+       }
+
+       err = -E2BIG;
+       cost += n_buckets * (value_size + sizeof(struct stack_map_bucket));
+       if (cost >= U32_MAX - PAGE_SIZE)
+               goto free_smap;
+
+       smap->map.map_type = attr->map_type;
+       smap->map.key_size = attr->key_size;
+       smap->map.value_size = value_size;
+       smap->map.max_entries = attr->max_entries;
+       smap->n_buckets = n_buckets;
+       smap->map.pages = round_up(cost, PAGE_SIZE) >> PAGE_SHIFT;
+
+       err = get_callchain_buffers();
+       if (err)
+               goto free_smap;
+
+       return &smap->map;
+
+free_smap:
+       kvfree(smap);
+       return ERR_PTR(err);
+}
+
+static u64 bpf_get_stackid(u64 r1, u64 r2, u64 flags, u64 r4, u64 r5)
+{
+       struct pt_regs *regs = (struct pt_regs *) (long) r1;
+       struct bpf_map *map = (struct bpf_map *) (long) r2;
+       struct bpf_stack_map *smap = container_of(map, struct bpf_stack_map, map);
+       struct perf_callchain_entry *trace;
+       struct stack_map_bucket *bucket, *new_bucket, *old_bucket;
+       u32 max_depth = map->value_size / 8;
+       /* stack_map_alloc() checks that max_depth <= PERF_MAX_STACK_DEPTH */
+       u32 init_nr = PERF_MAX_STACK_DEPTH - max_depth;
+       u32 skip = flags & BPF_F_SKIP_FIELD_MASK;
+       u32 hash, id, trace_nr, trace_len;
+       bool user = flags & BPF_F_USER_STACK;
+       bool kernel = !user;
+       u64 *ips;
+
+       if (unlikely(flags & ~(BPF_F_SKIP_FIELD_MASK | BPF_F_USER_STACK |
+                              BPF_F_FAST_STACK_CMP | BPF_F_REUSE_STACKID)))
+               return -EINVAL;
+
+       trace = get_perf_callchain(regs, init_nr, kernel, user, false, false);
+
+       if (unlikely(!trace))
+               /* couldn't fetch the stack trace */
+               return -EFAULT;
+
+       /* get_perf_callchain() guarantees that trace->nr >= init_nr
+        * and trace-nr <= PERF_MAX_STACK_DEPTH, so trace_nr <= max_depth
+        */
+       trace_nr = trace->nr - init_nr;
+
+       if (trace_nr <= skip)
+               /* skipping more than usable stack trace */
+               return -EFAULT;
+
+       trace_nr -= skip;
+       trace_len = trace_nr * sizeof(u64);
+       ips = trace->ip + skip + init_nr;
+       hash = jhash2((u32 *)ips, trace_len / sizeof(u32), 0);
+       id = hash & (smap->n_buckets - 1);
+       bucket = rcu_dereference(smap->buckets[id]);
+
+       if (bucket && bucket->hash == hash) {
+               if (flags & BPF_F_FAST_STACK_CMP)
+                       return id;
+               if (bucket->nr == trace_nr &&
+                   memcmp(bucket->ip, ips, trace_len) == 0)
+                       return id;
+       }
+
+       /* this call stack is not in the map, try to add it */
+       if (bucket && !(flags & BPF_F_REUSE_STACKID))
+               return -EEXIST;
+
+       new_bucket = kmalloc(sizeof(struct stack_map_bucket) + map->value_size,
+                            GFP_ATOMIC | __GFP_NOWARN);
+       if (unlikely(!new_bucket))
+               return -ENOMEM;
+
+       memcpy(new_bucket->ip, ips, trace_len);
+       memset(new_bucket->ip + trace_len / 8, 0, map->value_size - trace_len);
+       new_bucket->hash = hash;
+       new_bucket->nr = trace_nr;
+
+       old_bucket = xchg(&smap->buckets[id], new_bucket);
+       if (old_bucket)
+               kfree_rcu(old_bucket, rcu);
+       return id;
+}
+
+const struct bpf_func_proto bpf_get_stackid_proto = {
+       .func           = bpf_get_stackid,
+       .gpl_only       = true,
+       .ret_type       = RET_INTEGER,
+       .arg1_type      = ARG_PTR_TO_CTX,
+       .arg2_type      = ARG_CONST_MAP_PTR,
+       .arg3_type      = ARG_ANYTHING,
+};
+
+/* Called from syscall or from eBPF program */
+static void *stack_map_lookup_elem(struct bpf_map *map, void *key)
+{
+       struct bpf_stack_map *smap = container_of(map, struct bpf_stack_map, map);
+       struct stack_map_bucket *bucket;
+       u32 id = *(u32 *)key;
+
+       if (unlikely(id >= smap->n_buckets))
+               return NULL;
+       bucket = rcu_dereference(smap->buckets[id]);
+       return bucket ? bucket->ip : NULL;
+}
+
+static int stack_map_get_next_key(struct bpf_map *map, void *key, void *next_key)
+{
+       return -EINVAL;
+}
+
+static int stack_map_update_elem(struct bpf_map *map, void *key, void *value,
+                                u64 map_flags)
+{
+       return -EINVAL;
+}
+
+/* Called from syscall or from eBPF program */
+static int stack_map_delete_elem(struct bpf_map *map, void *key)
+{
+       struct bpf_stack_map *smap = container_of(map, struct bpf_stack_map, map);
+       struct stack_map_bucket *old_bucket;
+       u32 id = *(u32 *)key;
+
+       if (unlikely(id >= smap->n_buckets))
+               return -E2BIG;
+
+       old_bucket = xchg(&smap->buckets[id], NULL);
+       if (old_bucket) {
+               kfree_rcu(old_bucket, rcu);
+               return 0;
+       } else {
+               return -ENOENT;
+       }
+}
+
+/* Called when map->refcnt goes to zero, either from workqueue or from syscall */
+static void stack_map_free(struct bpf_map *map)
+{
+       struct bpf_stack_map *smap = container_of(map, struct bpf_stack_map, map);
+       int i;
+
+       synchronize_rcu();
+
+       for (i = 0; i < smap->n_buckets; i++)
+               if (smap->buckets[i])
+                       kfree_rcu(smap->buckets[i], rcu);
+       kvfree(smap);
+       put_callchain_buffers();
+}
+
+static const struct bpf_map_ops stack_map_ops = {
+       .map_alloc = stack_map_alloc,
+       .map_free = stack_map_free,
+       .map_get_next_key = stack_map_get_next_key,
+       .map_lookup_elem = stack_map_lookup_elem,
+       .map_update_elem = stack_map_update_elem,
+       .map_delete_elem = stack_map_delete_elem,
+};
+
+static struct bpf_map_type_list stack_map_type __read_mostly = {
+       .ops = &stack_map_ops,
+       .type = BPF_MAP_TYPE_STACK_TRACE,
+};
+
+static int __init register_stack_map(void)
+{
+       bpf_register_map_type(&stack_map_type);
+       return 0;
+}
+late_initcall(register_stack_map);

diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
index d1d3e8f57de907764fe3080632062485f5639443..42ba4ccc020b5e9a0bcd5cac4ac7216b3a68ff6a 100644 (file)
--- a/kernel/bpf/verifier.c
+++ b/kernel/bpf/verifier.c
@@ -246,6 +246,7 @@ static const struct {
        {BPF_MAP_TYPE_PROG_ARRAY, BPF_FUNC_tail_call},
        {BPF_MAP_TYPE_PERF_EVENT_ARRAY, BPF_FUNC_perf_event_read},
        {BPF_MAP_TYPE_PERF_EVENT_ARRAY, BPF_FUNC_perf_event_output},
+       {BPF_MAP_TYPE_STACK_TRACE, BPF_FUNC_get_stackid},
 };
 
 static void print_verifier_state(struct verifier_env *env)
@@ -911,8 +912,11 @@ static int check_map_func_compatibility(struct bpf_map *map, int func_id)
                 * don't allow any other map type to be passed into
                 * the special func;
                 */
-               if (bool_func && bool_map != bool_func)
+               if (bool_func && bool_map != bool_func) {
+                       verbose("cannot pass map_type %d into func %d\n",
+                               map->map_type, func_id);
                        return -EINVAL;
+               }
        }
 
        return 0;

diff --git a/kernel/trace/bpf_trace.c b/kernel/trace/bpf_trace.c
index 326a75e884dbf3f46513538861a028c7be459f09..4b8caa392b8644d92de4d2482b3b376c8db5d8af 100644 (file)
--- a/kernel/trace/bpf_trace.c
+++ b/kernel/trace/bpf_trace.c
@@ -299,6 +299,8 @@ static const struct bpf_func_proto *kprobe_prog_func_proto(enum bpf_func_id func
                return &bpf_perf_event_read_proto;
        case BPF_FUNC_perf_event_output:
                return &bpf_perf_event_output_proto;
+       case BPF_FUNC_get_stackid:
+               return &bpf_get_stackid_proto;
        default:
                return NULL;
        }