Currently, sysctl kernel.bpf_stats_enabled controls BPF runtime stats.
Typical userspace tools use kernel.bpf_stats_enabled as follows:
1. Enable kernel.bpf_stats_enabled;
2. Check program run_time_ns;
3. Sleep for the monitoring period;
4. Check program run_time_ns again, calculate the difference;
5. Disable kernel.bpf_stats_enabled.
The problem with this approach is that only one userspace tool can toggle
this sysctl. If multiple tools toggle the sysctl at the same time, the
measurement may be inaccurate.
To fix this problem while keep backward compatibility, introduce a new
bpf command BPF_ENABLE_STATS. On success, this command enables stats and
returns a valid fd. BPF_ENABLE_STATS takes argument "type". Currently,
only one type, BPF_STATS_RUNTIME_CNT, is supported. We can extend the
command to support other types of stats in the future.
With BPF_ENABLE_STATS, user space tool would have the following flow:
1. Get a fd with BPF_ENABLE_STATS, and make sure it is valid;
2. Check program run_time_ns;
3. Sleep for the monitoring period;
4. Check program run_time_ns again, calculate the difference;
5. Close the fd.
Signed-off-by: Song Liu <songliubraving@xxxxxx>
---
Changes v2 => v3:
1. Rename the command to BPF_ENABLE_STATS, and make it extendible.
2. fix commit log;
3. remove unnecessary headers.
Changes RFC => v2:
1. Add a new bpf command instead of /dev/bpf_stats;
2. Remove the jump_label patch, which is no longer needed;
3. Add a static variable to save previous value of the sysctl.
---
include/linux/bpf.h | 1 +
include/uapi/linux/bpf.h | 17 ++++++++++--
kernel/bpf/syscall.c | 50 ++++++++++++++++++++++++++++++++++
kernel/sysctl.c | 36 +++++++++++++++++++++++-
tools/include/uapi/linux/bpf.h | 17 ++++++++++--
5 files changed, 114 insertions(+), 7 deletions(-)
diff --git a/include/linux/bpf.h b/include/linux/bpf.h
index fd2b2322412d..415be9d71f42 100644
--- a/include/linux/bpf.h
+++ b/include/linux/bpf.h
@@ -987,6 +987,7 @@ _out: \
#ifdef CONFIG_BPF_SYSCALL
DECLARE_PER_CPU(int, bpf_prog_active);
+extern struct mutex bpf_stats_enabled_mutex;
/*
* Block execution of BPF programs attached to instrumentation (perf,
diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
index 2e29a671d67e..7ad3f0f4b461 100644
--- a/include/uapi/linux/bpf.h
+++ b/include/uapi/linux/bpf.h
@@ -113,6 +113,7 @@ enum bpf_cmd {
BPF_MAP_DELETE_BATCH,
BPF_LINK_CREATE,
BPF_LINK_UPDATE,
+ BPF_ENABLE_STATS,
};
enum bpf_map_type {
@@ -379,6 +380,12 @@ enum {
*/
#define BPF_F_QUERY_EFFECTIVE (1U << 0)
+/* type for BPF_ENABLE_STATS */
+enum {
+ /* enabled run_time_ns and run_cnt */
+ BPF_STATS_RUNTIME_CNT = 0,
+};
+
enum bpf_stack_build_id_status {
/* user space need an empty entry to identify end of a trace */
BPF_STACK_BUILD_ID_EMPTY = 0,
@@ -589,6 +596,10 @@ union bpf_attr {
__u32 old_prog_fd;
} link_update;
+ struct { /* struct used by BPF_ENABLE_STATS command */
+ __u32 type;
+ } enable_stats;
+
} __attribute__((aligned(8)));
/* The description below is an attempt at providing documentation to eBPF
@@ -971,14 +982,14 @@ union bpf_attr {
*
* int ret;
* struct bpf_tunnel_key key = {};
- *
+ *
* ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
* if (ret < 0)
* return TC_ACT_SHOT; // drop packet
- *
+ *
* if (key.remote_ipv4 != 0x0a000001)
* return TC_ACT_SHOT; // drop packet
- *
+ *
* return TC_ACT_OK; // accept packet
*
* This interface can also be used with all encapsulation devices
diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
index d85f37239540..766778368c5f 100644
--- a/kernel/bpf/syscall.c
+++ b/kernel/bpf/syscall.c
@@ -3656,6 +3656,53 @@ static int link_update(union bpf_attr *attr)
return ret;
}
+DEFINE_MUTEX(bpf_stats_enabled_mutex);
+
+static int bpf_stats_release(struct inode *inode, struct file *file)
+{
+ mutex_lock(&bpf_stats_enabled_mutex);
+ static_key_slow_dec(&bpf_stats_enabled_key.key);
+ mutex_unlock(&bpf_stats_enabled_mutex);
+ return 0;
+}
+
+static const struct file_operations bpf_stats_fops = {
+ .release = bpf_stats_release,
+};
+
+static int bpf_enable_runtime_stats(void)
+{
+ int fd;
+
+ if (!capable(CAP_SYS_ADMIN))
+ return -EPERM;
+
+ mutex_lock(&bpf_stats_enabled_mutex);
+ /* Set a very high limit to avoid overflow */
+ if (static_key_count(&bpf_stats_enabled_key.key) > INT_MAX / 2) {
+ mutex_unlock(&bpf_stats_enabled_mutex);
+ return -EBUSY;
+ }
+
+ fd = anon_inode_getfd("bpf-stats", &bpf_stats_fops, NULL, 0);
+ if (fd >= 0)
+ static_key_slow_inc(&bpf_stats_enabled_key.key);
+
+ mutex_unlock(&bpf_stats_enabled_mutex);
+ return fd;
+}
+
+static int bpf_enable_stats(union bpf_attr *attr)
+{
+ switch (attr->enable_stats.type) {
+ case BPF_STATS_RUNTIME_CNT:
+ return bpf_enable_runtime_stats();
+ default:
+ break;
+ }
+ return -EINVAL;
+}
+
SYSCALL_DEFINE3(bpf, int, cmd, union bpf_attr __user *, uattr, unsigned int, size)
{
union bpf_attr attr;
@@ -3773,6 +3820,9 @@ SYSCALL_DEFINE3(bpf, int, cmd, union bpf_attr __user *, uattr, unsigned int, siz
case BPF_LINK_UPDATE:
err = link_update(&attr);
break;
+ case BPF_ENABLE_STATS:
+ err = bpf_enable_stats(&attr);
+ break;
default:
err = -EINVAL;
break;
diff --git a/kernel/sysctl.c b/kernel/sysctl.c
index 8a176d8727a3..2d72f4ad3063 100644
--- a/kernel/sysctl.c
+++ b/kernel/sysctl.c
@@ -304,6 +304,40 @@ static int min_extfrag_threshold;
static int max_extfrag_threshold = 1000;
#endif
+#ifdef CONFIG_BPF_SYSCALL
+static int bpf_stats_handler(struct ctl_table *table, int write,
+ void __user *buffer, size_t *lenp,
+ loff_t *ppos)
+{
+ struct static_key *key = (struct static_key *)table->data;
+ static int saved_val;
+ int val, ret;
+ struct ctl_table tmp = {
+ .data = &val,
+ .maxlen = sizeof(val),
+ .mode = table->mode,
+ .extra1 = SYSCTL_ZERO,
+ .extra2 = SYSCTL_ONE,
+ };
+
+ if (write && !capable(CAP_SYS_ADMIN))
+ return -EPERM;
+
+ mutex_lock(&bpf_stats_enabled_mutex);
+ val = saved_val;
+ ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
+ if (write && !ret && val != saved_val) {
+ if (val)
+ static_key_slow_inc(key);
+ else
+ static_key_slow_dec(key);
+ saved_val = val;
+ }
+ mutex_unlock(&bpf_stats_enabled_mutex);
+ return ret;
+}
+#endif
+
static struct ctl_table kern_table[] = {
{
.procname = "sched_child_runs_first",
@@ -1244,7 +1278,7 @@ static struct ctl_table kern_table[] = {
.data = &bpf_stats_enabled_key.key,
.maxlen = sizeof(bpf_stats_enabled_key),
.mode = 0644,
- .proc_handler = proc_do_static_key,
+ .proc_handler = bpf_stats_handler,
},
#endif
#if defined(CONFIG_TREE_RCU)
diff --git a/tools/include/uapi/linux/bpf.h b/tools/include/uapi/linux/bpf.h
index 2e29a671d67e..7ad3f0f4b461 100644
--- a/tools/include/uapi/linux/bpf.h
+++ b/tools/include/uapi/linux/bpf.h
@@ -113,6 +113,7 @@ enum bpf_cmd {
BPF_MAP_DELETE_BATCH,
BPF_LINK_CREATE,
BPF_LINK_UPDATE,
+ BPF_ENABLE_STATS,
};
enum bpf_map_type {
@@ -379,6 +380,12 @@ enum {
*/
#define BPF_F_QUERY_EFFECTIVE (1U << 0)
+/* type for BPF_ENABLE_STATS */
+enum {
+ /* enabled run_time_ns and run_cnt */
+ BPF_STATS_RUNTIME_CNT = 0,
+};
+
enum bpf_stack_build_id_status {
/* user space need an empty entry to identify end of a trace */
BPF_STACK_BUILD_ID_EMPTY = 0,
@@ -589,6 +596,10 @@ union bpf_attr {
__u32 old_prog_fd;
} link_update;
+ struct { /* struct used by BPF_ENABLE_STATS command */
+ __u32 type;
+ } enable_stats;
+
} __attribute__((aligned(8)));
/* The description below is an attempt at providing documentation to eBPF
@@ -971,14 +982,14 @@ union bpf_attr {
*
* int ret;
* struct bpf_tunnel_key key = {};
- *
+ *
* ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
* if (ret < 0)
* return TC_ACT_SHOT; // drop packet
- *
+ *
* if (key.remote_ipv4 != 0x0a000001)
* return TC_ACT_SHOT; // drop packet
- *
+ *
* return TC_ACT_OK; // accept packet
*
* This interface can also be used with all encapsulation devices
--
2.24.1
