Merge branch 'bpf-get-stackid'
Alexei Starovoitov says:
====================
bpf_get_stackid() and stack_trace map
This patch set introduces new map type to store stack traces and
corresponding bpf_get_stackid() helper.
BPF programs already can walk the stack via unrolled loop
of bpf_probe_read()s which is ok for simple analysis, but it's
not efficient and limited to <30 frames after that the programs
don't fit into MAX_BPF_STACK. With bpf_get_stackid() helper
the programs can collect up to PERF_MAX_STACK_DEPTH both
user and kernel frames.
Using stack traces as a key in a map turned out to be very useful
for generating flame graphs, off-cpu graphs, waker and chain graphs.
Patch 3 is a simplified version of 'offwaketime' tool which is
described in detail here:
http://brendangregg.com/blog/2016-02-01/linux-wakeup-offwake-profiling.html
Earlier version of this patch were using save_stack_trace() helper,
but 'unreliable' frames add to much noise and two equiavlent
stack traces produce different 'stackid's.
Using lockdep style of storing frames with MAX_STACK_TRACE_ENTRIES is
great for lockdep, but not acceptable for bpf, since the stack_trace
map needs to be freed when user Ctrl-C the tool.
The ftrace style with per_cpu(struct ftrace_stack) is great, but it's
tightly coupled with ftrace ring buffer and has the same 'unreliable'
noise. perf_event's perf_callchain() mechanism is also very efficient
and it only needed minor generalization which is done in patch 1
to be used by bpf stack_trace maps.
Peter, please take a look at patch 1.
If you're ok with it, I'd like to take the whole set via net-next.
Patch 1 - generalization of perf_callchain()
Patch 2 - stack_trace map done as lock-less hashtable without link list
to avoid spinlock on insertion which is critical path when
bpf_get_stackid() helper is called for every task switch event
Patch 3 - offwaketime example
After the patch the 'perf report' for artificial 'sched_bench'
benchmark that doing pthread_cond_wait/signal and 'offwaketime'
example is running in the background:
16.35% swapper [kernel.vmlinux] [k] intel_idle
2.18% sched_bench [kernel.vmlinux] [k] __switch_to
2.18% sched_bench libpthread-2.12.so [.] pthread_cond_signal@@GLIBC_2.3.2
1.72% sched_bench libpthread-2.12.so [.] pthread_mutex_unlock
1.53% sched_bench [kernel.vmlinux] [k] bpf_get_stackid
1.44% sched_bench [kernel.vmlinux] [k] entry_SYSCALL_64
1.39% sched_bench [kernel.vmlinux] [k] __call_rcu.constprop.73
1.13% sched_bench libpthread-2.12.so [.] pthread_mutex_lock
1.07% sched_bench libpthread-2.12.so [.] pthread_cond_wait@@GLIBC_2.3.2
1.07% sched_bench [kernel.vmlinux] [k] hash_futex
1.05% sched_bench [kernel.vmlinux] [k] do_futex
1.05% sched_bench [kernel.vmlinux] [k] get_futex_key_refs.isra.13
The hotest part of bpf_get_stackid() is inlined jhash2, so we may consider
using some faster hash in the future, but it's good enough for now.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
diff --git a/arch/x86/include/asm/stacktrace.h b/arch/x86/include/asm/stacktrace.h
index 70bbe39..7c247e7 100644
--- a/arch/x86/include/asm/stacktrace.h
+++ b/arch/x86/include/asm/stacktrace.h
@@ -37,7 +37,7 @@
/* Generic stack tracer with callbacks */
struct stacktrace_ops {
- void (*address)(void *data, unsigned long address, int reliable);
+ int (*address)(void *data, unsigned long address, int reliable);
/* On negative return stop dumping */
int (*stack)(void *data, char *name);
walk_stack_t walk_stack;
diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c
index 1b443db..d276b31 100644
--- a/arch/x86/kernel/cpu/perf_event.c
+++ b/arch/x86/kernel/cpu/perf_event.c
@@ -2180,11 +2180,11 @@
return 0;
}
-static void backtrace_address(void *data, unsigned long addr, int reliable)
+static int backtrace_address(void *data, unsigned long addr, int reliable)
{
struct perf_callchain_entry *entry = data;
- perf_callchain_store(entry, addr);
+ return perf_callchain_store(entry, addr);
}
static const struct stacktrace_ops backtrace_ops = {
diff --git a/arch/x86/kernel/dumpstack.c b/arch/x86/kernel/dumpstack.c
index 9c30acf..0d1ff4b 100644
--- a/arch/x86/kernel/dumpstack.c
+++ b/arch/x86/kernel/dumpstack.c
@@ -135,7 +135,8 @@
if (!__kernel_text_address(addr))
break;
- ops->address(data, addr, 1);
+ if (ops->address(data, addr, 1))
+ break;
frame = frame->next_frame;
ret_addr = &frame->return_address;
print_ftrace_graph_addr(addr, data, ops, tinfo, graph);
@@ -154,10 +155,11 @@
/*
* Print one address/symbol entries per line.
*/
-static void print_trace_address(void *data, unsigned long addr, int reliable)
+static int print_trace_address(void *data, unsigned long addr, int reliable)
{
touch_nmi_watchdog();
printk_stack_address(addr, reliable, data);
+ return 0;
}
static const struct stacktrace_ops print_trace_ops = {
diff --git a/arch/x86/kernel/stacktrace.c b/arch/x86/kernel/stacktrace.c
index fdd0c64..9ee98ee 100644
--- a/arch/x86/kernel/stacktrace.c
+++ b/arch/x86/kernel/stacktrace.c
@@ -14,30 +14,34 @@
return 0;
}
-static void
+static int
__save_stack_address(void *data, unsigned long addr, bool reliable, bool nosched)
{
struct stack_trace *trace = data;
#ifdef CONFIG_FRAME_POINTER
if (!reliable)
- return;
+ return 0;
#endif
if (nosched && in_sched_functions(addr))
- return;
+ return 0;
if (trace->skip > 0) {
trace->skip--;
- return;
+ return 0;
}
- if (trace->nr_entries < trace->max_entries)
+ if (trace->nr_entries < trace->max_entries) {
trace->entries[trace->nr_entries++] = addr;
+ return 0;
+ } else {
+ return -1; /* no more room, stop walking the stack */
+ }
}
-static void save_stack_address(void *data, unsigned long addr, int reliable)
+static int save_stack_address(void *data, unsigned long addr, int reliable)
{
return __save_stack_address(data, addr, reliable, false);
}
-static void
+static int
save_stack_address_nosched(void *data, unsigned long addr, int reliable)
{
return __save_stack_address(data, addr, reliable, true);
diff --git a/arch/x86/oprofile/backtrace.c b/arch/x86/oprofile/backtrace.c
index 4e664bd..cb31a44 100644
--- a/arch/x86/oprofile/backtrace.c
+++ b/arch/x86/oprofile/backtrace.c
@@ -23,12 +23,13 @@
return 0;
}
-static void backtrace_address(void *data, unsigned long addr, int reliable)
+static int backtrace_address(void *data, unsigned long addr, int reliable)
{
unsigned int *depth = data;
if ((*depth)--)
oprofile_add_trace(addr);
+ return 0;
}
static struct stacktrace_ops backtrace_ops = {
diff --git a/include/linux/bpf.h b/include/linux/bpf.h
index 90ee6ab..0cadbb7 100644
--- a/include/linux/bpf.h
+++ b/include/linux/bpf.h
@@ -237,6 +237,7 @@
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/linux/perf_event.h b/include/linux/perf_event.h
index b35a61a..7da3c25 100644
--- a/include/linux/perf_event.h
+++ b/include/linux/perf_event.h
@@ -964,11 +964,20 @@
extern void perf_callchain_user(struct perf_callchain_entry *entry, struct pt_regs *regs);
extern void perf_callchain_kernel(struct perf_callchain_entry *entry, struct pt_regs *regs);
+extern struct perf_callchain_entry *
+get_perf_callchain(struct pt_regs *regs, u32 init_nr, bool kernel, bool user,
+ bool crosstask, bool add_mark);
+extern int get_callchain_buffers(void);
+extern void put_callchain_buffers(void);
-static inline void perf_callchain_store(struct perf_callchain_entry *entry, u64 ip)
+static inline int perf_callchain_store(struct perf_callchain_entry *entry, u64 ip)
{
- if (entry->nr < PERF_MAX_STACK_DEPTH)
+ if (entry->nr < PERF_MAX_STACK_DEPTH) {
entry->ip[entry->nr++] = ip;
+ return 0;
+ } else {
+ return -1; /* no more room, stop walking the stack */
+ }
}
extern int sysctl_perf_event_paranoid;
diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
index 2ee0fde..d3e77da 100644
--- a/include/uapi/linux/bpf.h
+++ b/include/uapi/linux/bpf.h
@@ -83,6 +83,7 @@
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 @@
*/
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 @@
/* 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 1327258..8a932d0 100644
--- a/kernel/bpf/Makefile
+++ b/kernel/bpf/Makefile
@@ -2,3 +2,6 @@
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
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 d1d3e8f..42ba4cc 100644
--- a/kernel/bpf/verifier.c
+++ b/kernel/bpf/verifier.c
@@ -246,6 +246,7 @@
{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 @@
* 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/events/callchain.c b/kernel/events/callchain.c
index 9c41800..343c22f 100644
--- a/kernel/events/callchain.c
+++ b/kernel/events/callchain.c
@@ -159,15 +159,24 @@
struct perf_callchain_entry *
perf_callchain(struct perf_event *event, struct pt_regs *regs)
{
- int rctx;
- struct perf_callchain_entry *entry;
-
- int kernel = !event->attr.exclude_callchain_kernel;
- int user = !event->attr.exclude_callchain_user;
+ bool kernel = !event->attr.exclude_callchain_kernel;
+ bool user = !event->attr.exclude_callchain_user;
+ /* Disallow cross-task user callchains. */
+ bool crosstask = event->ctx->task && event->ctx->task != current;
if (!kernel && !user)
return NULL;
+ return get_perf_callchain(regs, 0, kernel, user, crosstask, true);
+}
+
+struct perf_callchain_entry *
+get_perf_callchain(struct pt_regs *regs, u32 init_nr, bool kernel, bool user,
+ bool crosstask, bool add_mark)
+{
+ struct perf_callchain_entry *entry;
+ int rctx;
+
entry = get_callchain_entry(&rctx);
if (rctx == -1)
return NULL;
@@ -175,10 +184,11 @@
if (!entry)
goto exit_put;
- entry->nr = 0;
+ entry->nr = init_nr;
if (kernel && !user_mode(regs)) {
- perf_callchain_store(entry, PERF_CONTEXT_KERNEL);
+ if (add_mark)
+ perf_callchain_store(entry, PERF_CONTEXT_KERNEL);
perf_callchain_kernel(entry, regs);
}
@@ -191,13 +201,11 @@
}
if (regs) {
- /*
- * Disallow cross-task user callchains.
- */
- if (event->ctx->task && event->ctx->task != current)
+ if (crosstask)
goto exit_put;
- perf_callchain_store(entry, PERF_CONTEXT_USER);
+ if (add_mark)
+ perf_callchain_store(entry, PERF_CONTEXT_USER);
perf_callchain_user(entry, regs);
}
}
diff --git a/kernel/events/internal.h b/kernel/events/internal.h
index 2bbad9c..4199b6d 100644
--- a/kernel/events/internal.h
+++ b/kernel/events/internal.h
@@ -182,8 +182,6 @@
/* Callchain handling */
extern struct perf_callchain_entry *
perf_callchain(struct perf_event *event, struct pt_regs *regs);
-extern int get_callchain_buffers(void);
-extern void put_callchain_buffers(void);
static inline int get_recursion_context(int *recursion)
{
diff --git a/kernel/trace/bpf_trace.c b/kernel/trace/bpf_trace.c
index 326a75e..4b8caa3 100644
--- a/kernel/trace/bpf_trace.c
+++ b/kernel/trace/bpf_trace.c
@@ -299,6 +299,8 @@
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;
}
diff --git a/samples/bpf/Makefile b/samples/bpf/Makefile
index edd638b..c4f8ae0 100644
--- a/samples/bpf/Makefile
+++ b/samples/bpf/Makefile
@@ -16,6 +16,7 @@
hostprogs-y += tracex6
hostprogs-y += trace_output
hostprogs-y += lathist
+hostprogs-y += offwaketime
test_verifier-objs := test_verifier.o libbpf.o
test_maps-objs := test_maps.o libbpf.o
@@ -32,6 +33,7 @@
tracex6-objs := bpf_load.o libbpf.o tracex6_user.o
trace_output-objs := bpf_load.o libbpf.o trace_output_user.o
lathist-objs := bpf_load.o libbpf.o lathist_user.o
+offwaketime-objs := bpf_load.o libbpf.o offwaketime_user.o
# Tell kbuild to always build the programs
always := $(hostprogs-y)
@@ -47,6 +49,7 @@
always += trace_output_kern.o
always += tcbpf1_kern.o
always += lathist_kern.o
+always += offwaketime_kern.o
HOSTCFLAGS += -I$(objtree)/usr/include
@@ -63,6 +66,7 @@
HOSTLOADLIBES_tracex6 += -lelf
HOSTLOADLIBES_trace_output += -lelf -lrt
HOSTLOADLIBES_lathist += -lelf
+HOSTLOADLIBES_offwaketime += -lelf
# point this to your LLVM backend with bpf support
LLC=$(srctree)/tools/bpf/llvm/bld/Debug+Asserts/bin/llc
diff --git a/samples/bpf/bpf_helpers.h b/samples/bpf/bpf_helpers.h
index 7ad19e1..811bcca 100644
--- a/samples/bpf/bpf_helpers.h
+++ b/samples/bpf/bpf_helpers.h
@@ -39,6 +39,8 @@
(void *) BPF_FUNC_redirect;
static int (*bpf_perf_event_output)(void *ctx, void *map, int index, void *data, int size) =
(void *) BPF_FUNC_perf_event_output;
+static int (*bpf_get_stackid)(void *ctx, void *map, int flags) =
+ (void *) BPF_FUNC_get_stackid;
/* llvm builtin functions that eBPF C program may use to
* emit BPF_LD_ABS and BPF_LD_IND instructions
diff --git a/samples/bpf/offwaketime_kern.c b/samples/bpf/offwaketime_kern.c
new file mode 100644
index 0000000..c0aa5a9
--- /dev/null
+++ b/samples/bpf/offwaketime_kern.c
@@ -0,0 +1,131 @@
+/* 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 <uapi/linux/bpf.h>
+#include "bpf_helpers.h"
+#include <uapi/linux/ptrace.h>
+#include <uapi/linux/perf_event.h>
+#include <linux/version.h>
+#include <linux/sched.h>
+
+#define _(P) ({typeof(P) val = 0; bpf_probe_read(&val, sizeof(val), &P); val;})
+
+#define MINBLOCK_US 1
+
+struct key_t {
+ char waker[TASK_COMM_LEN];
+ char target[TASK_COMM_LEN];
+ u32 wret;
+ u32 tret;
+};
+
+struct bpf_map_def SEC("maps") counts = {
+ .type = BPF_MAP_TYPE_HASH,
+ .key_size = sizeof(struct key_t),
+ .value_size = sizeof(u64),
+ .max_entries = 10000,
+};
+
+struct bpf_map_def SEC("maps") start = {
+ .type = BPF_MAP_TYPE_HASH,
+ .key_size = sizeof(u32),
+ .value_size = sizeof(u64),
+ .max_entries = 10000,
+};
+
+struct wokeby_t {
+ char name[TASK_COMM_LEN];
+ u32 ret;
+};
+
+struct bpf_map_def SEC("maps") wokeby = {
+ .type = BPF_MAP_TYPE_HASH,
+ .key_size = sizeof(u32),
+ .value_size = sizeof(struct wokeby_t),
+ .max_entries = 10000,
+};
+
+struct bpf_map_def SEC("maps") stackmap = {
+ .type = BPF_MAP_TYPE_STACK_TRACE,
+ .key_size = sizeof(u32),
+ .value_size = PERF_MAX_STACK_DEPTH * sizeof(u64),
+ .max_entries = 10000,
+};
+
+#define STACKID_FLAGS (0 | BPF_F_FAST_STACK_CMP)
+
+SEC("kprobe/try_to_wake_up")
+int waker(struct pt_regs *ctx)
+{
+ struct task_struct *p = (void *) PT_REGS_PARM1(ctx);
+ struct wokeby_t woke = {};
+ u32 pid;
+
+ pid = _(p->pid);
+
+ bpf_get_current_comm(&woke.name, sizeof(woke.name));
+ woke.ret = bpf_get_stackid(ctx, &stackmap, STACKID_FLAGS);
+
+ bpf_map_update_elem(&wokeby, &pid, &woke, BPF_ANY);
+ return 0;
+}
+
+static inline int update_counts(struct pt_regs *ctx, u32 pid, u64 delta)
+{
+ struct key_t key = {};
+ struct wokeby_t *woke;
+ u64 zero = 0, *val;
+
+ bpf_get_current_comm(&key.target, sizeof(key.target));
+ key.tret = bpf_get_stackid(ctx, &stackmap, STACKID_FLAGS);
+
+ woke = bpf_map_lookup_elem(&wokeby, &pid);
+ if (woke) {
+ key.wret = woke->ret;
+ __builtin_memcpy(&key.waker, woke->name, TASK_COMM_LEN);
+ bpf_map_delete_elem(&wokeby, &pid);
+ }
+
+ val = bpf_map_lookup_elem(&counts, &key);
+ if (!val) {
+ bpf_map_update_elem(&counts, &key, &zero, BPF_NOEXIST);
+ val = bpf_map_lookup_elem(&counts, &key);
+ if (!val)
+ return 0;
+ }
+ (*val) += delta;
+ return 0;
+}
+
+SEC("kprobe/finish_task_switch")
+int oncpu(struct pt_regs *ctx)
+{
+ struct task_struct *p = (void *) PT_REGS_PARM1(ctx);
+ u64 delta, ts, *tsp;
+ u32 pid;
+
+ /* record previous thread sleep time */
+ pid = _(p->pid);
+ ts = bpf_ktime_get_ns();
+ bpf_map_update_elem(&start, &pid, &ts, BPF_ANY);
+
+ /* calculate current thread's delta time */
+ pid = bpf_get_current_pid_tgid();
+ tsp = bpf_map_lookup_elem(&start, &pid);
+ if (!tsp)
+ /* missed start or filtered */
+ return 0;
+
+ delta = bpf_ktime_get_ns() - *tsp;
+ bpf_map_delete_elem(&start, &pid);
+ delta = delta / 1000;
+ if (delta < MINBLOCK_US)
+ return 0;
+
+ return update_counts(ctx, pid, delta);
+}
+char _license[] SEC("license") = "GPL";
+u32 _version SEC("version") = LINUX_VERSION_CODE;
diff --git a/samples/bpf/offwaketime_user.c b/samples/bpf/offwaketime_user.c
new file mode 100644
index 0000000..17cf302
--- /dev/null
+++ b/samples/bpf/offwaketime_user.c
@@ -0,0 +1,185 @@
+/* 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 <stdio.h>
+#include <unistd.h>
+#include <stdlib.h>
+#include <signal.h>
+#include <linux/bpf.h>
+#include <string.h>
+#include <linux/perf_event.h>
+#include <errno.h>
+#include <assert.h>
+#include <stdbool.h>
+#include <sys/resource.h>
+#include "libbpf.h"
+#include "bpf_load.h"
+
+#define MAX_SYMS 300000
+#define PRINT_RAW_ADDR 0
+
+static struct ksym {
+ long addr;
+ char *name;
+} syms[MAX_SYMS];
+static int sym_cnt;
+
+static int ksym_cmp(const void *p1, const void *p2)
+{
+ return ((struct ksym *)p1)->addr - ((struct ksym *)p2)->addr;
+}
+
+static int load_kallsyms(void)
+{
+ FILE *f = fopen("/proc/kallsyms", "r");
+ char func[256], buf[256];
+ char symbol;
+ void *addr;
+ int i = 0;
+
+ if (!f)
+ return -ENOENT;
+
+ while (!feof(f)) {
+ if (!fgets(buf, sizeof(buf), f))
+ break;
+ if (sscanf(buf, "%p %c %s", &addr, &symbol, func) != 3)
+ break;
+ if (!addr)
+ continue;
+ syms[i].addr = (long) addr;
+ syms[i].name = strdup(func);
+ i++;
+ }
+ sym_cnt = i;
+ qsort(syms, sym_cnt, sizeof(struct ksym), ksym_cmp);
+ return 0;
+}
+
+static void *search(long key)
+{
+ int start = 0, end = sym_cnt;
+ int result;
+
+ while (start < end) {
+ size_t mid = start + (end - start) / 2;
+
+ result = key - syms[mid].addr;
+ if (result < 0)
+ end = mid;
+ else if (result > 0)
+ start = mid + 1;
+ else
+ return &syms[mid];
+ }
+
+ if (start >= 1 && syms[start - 1].addr < key &&
+ key < syms[start].addr)
+ /* valid ksym */
+ return &syms[start - 1];
+
+ /* out of range. return _stext */
+ return &syms[0];
+}
+
+static void print_ksym(__u64 addr)
+{
+ struct ksym *sym;
+
+ if (!addr)
+ return;
+ sym = search(addr);
+ if (PRINT_RAW_ADDR)
+ printf("%s/%llx;", sym->name, addr);
+ else
+ printf("%s;", sym->name);
+}
+
+#define TASK_COMM_LEN 16
+
+struct key_t {
+ char waker[TASK_COMM_LEN];
+ char target[TASK_COMM_LEN];
+ __u32 wret;
+ __u32 tret;
+};
+
+static void print_stack(struct key_t *key, __u64 count)
+{
+ __u64 ip[PERF_MAX_STACK_DEPTH] = {};
+ static bool warned;
+ int i;
+
+ printf("%s;", key->target);
+ if (bpf_lookup_elem(map_fd[3], &key->tret, ip) != 0) {
+ printf("---;");
+ } else {
+ for (i = PERF_MAX_STACK_DEPTH - 1; i >= 0; i--)
+ print_ksym(ip[i]);
+ }
+ printf("-;");
+ if (bpf_lookup_elem(map_fd[3], &key->wret, ip) != 0) {
+ printf("---;");
+ } else {
+ for (i = 0; i < PERF_MAX_STACK_DEPTH; i++)
+ print_ksym(ip[i]);
+ }
+ printf(";%s %lld\n", key->waker, count);
+
+ if ((key->tret == -EEXIST || key->wret == -EEXIST) && !warned) {
+ printf("stackmap collisions seen. Consider increasing size\n");
+ warned = true;
+ } else if (((int)(key->tret) < 0 || (int)(key->wret) < 0)) {
+ printf("err stackid %d %d\n", key->tret, key->wret);
+ }
+}
+
+static void print_stacks(int fd)
+{
+ struct key_t key = {}, next_key;
+ __u64 value;
+
+ while (bpf_get_next_key(fd, &key, &next_key) == 0) {
+ bpf_lookup_elem(fd, &next_key, &value);
+ print_stack(&next_key, value);
+ key = next_key;
+ }
+}
+
+static void int_exit(int sig)
+{
+ print_stacks(map_fd[0]);
+ exit(0);
+}
+
+int main(int argc, char **argv)
+{
+ struct rlimit r = {RLIM_INFINITY, RLIM_INFINITY};
+ char filename[256];
+ int delay = 1;
+
+ snprintf(filename, sizeof(filename), "%s_kern.o", argv[0]);
+ setrlimit(RLIMIT_MEMLOCK, &r);
+
+ signal(SIGINT, int_exit);
+
+ if (load_kallsyms()) {
+ printf("failed to process /proc/kallsyms\n");
+ return 2;
+ }
+
+ if (load_bpf_file(filename)) {
+ printf("%s", bpf_log_buf);
+ return 1;
+ }
+
+ if (argc > 1)
+ delay = atoi(argv[1]);
+ sleep(delay);
+ print_stacks(map_fd[0]);
+
+ return 0;
+}