tools/profile: add support for PID-namespacing (#4709)

This adds translation logic to perform PID translation across
PID-namespaces.
It is now possible to profile a target process from within a
nested PID-namespace (e.g. in a container).
Also add a note in profile_example.txt file.

2 files changed, 27 insertions, 4 deletions

diff --git a/tools/profile.py b/tools/profile.py
index 796e2bf5..7961550a 100755
--- a/tools/profile.py
+++ b/tools/profile.py
@@ -162,9 +162,18 @@ BPF_STACK_TRACE(stack_traces, STACK_STORAGE_SIZE);
 // This code gets a bit complex. Probably not suitable for casual hacking.
 
 int do_perf_event(struct bpf_perf_event_data *ctx) {
-    u64 id = bpf_get_current_pid_tgid();
-    u32 tgid = id >> 32;
-    u32 pid = id;
+    u32 tgid = 0;
+    u32 pid = 0;
+
+    struct bpf_pidns_info ns = {};
+    if (USE_PIDNS && !bpf_get_ns_current_pid_tgid(PIDNS_DEV, PIDNS_INO, &ns, sizeof(struct bpf_pidns_info))) {
+        tgid = ns.tgid;
+        pid = ns.pid;
+    } else {
+        u64 id = bpf_get_current_pid_tgid();
+        tgid = id >> 32;
+        pid = id;
+    }
 
     if (IDLE_FILTER)
         return 0;
@@ -216,6 +225,17 @@ int do_perf_event(struct bpf_perf_event_data *ctx) {
 }
 """
 
+# pid-namespace translation
+try:
+    devinfo = os.stat("/proc/self/ns/pid")
+    bpf_text = bpf_text.replace('USE_PIDNS', "1")
+    bpf_text = bpf_text.replace('PIDNS_DEV', str(devinfo.st_dev))
+    bpf_text = bpf_text.replace('PIDNS_INO', str(devinfo.st_ino))
+except:
+    bpf_text = bpf_text.replace('USE_PIDNS', "0")
+    bpf_text = bpf_text.replace('PIDNS_DEV', "0")
+    bpf_text = bpf_text.replace('PIDNS_INO', "0")
+
 # set idle filter
 idle_filter = "pid == 0"
 if args.include_idle:
diff --git a/tools/profile_example.txt b/tools/profile_example.txt
index 65a10048..7dbc33af 100644
--- a/tools/profile_example.txt
+++ b/tools/profile_example.txt
@@ -154,7 +154,10 @@ Sampling at 49 Hertz of PID 25036 by user + kernel stack... Hit Ctrl-C to end.
         7
 
 Again, I've truncated some lines. Now we're just analyzing the dd process.
-The filtering is performed in kernel context, for efficiency.
+The filter is configured by specifying the target PID (from the current PID
+namespace where we are profiling) via the "-p" flag.
+Filtering is performed in kernel context, for efficiency, with automatic
+PID translation to the top-level namespace (if required).
 
 This output has some "[unknown]" frames that probably have valid addresses,
 but we're lacking the symbol translation. This is a common for all profilers