[RFC] Get values from per-cpu maps #2
Conversation
Useful for per-cpu maps lookup implementation.
Convert tabs to spaces.
|
Hi! As you see there's no per-cpu support yet, I think the way to go about it would be:
the rationale for this is that right now |
| @@ -315,6 +316,11 @@ Napi::Value BpfObjGet(const CallbackInfo& info) { | |||
| return ToStatus(env, bpf_obj_get(path.c_str())); | |||
| } | |||
|
|
|||
| Napi::Value GetNumPossibleCPUs(const CallbackInfo& info) { | |||
| Napi::Env env = info.Env(); | |||
| return ToStatus(env, libbpf_num_possible_cpus()); | |||
There was a problem hiding this comment.
nit: we can't use ToStatus here, the function returns the error code directly (not through errno)
|
Thanks, that sounds like a great plan! Unfortunately, I have to focus on some other stuff, so let me just put this on my weekend project list. |
The node_bpf packages has no support for per-cpu bpf maps yet. As a quick workaround we stick to non percpu-maps. See mildsunrise/node_bpf#2 for details.
The node_bpf packages has no support for per-cpu bpf maps yet. As a quick workaround we stick to non percpu-maps. See mildsunrise/node_bpf#2 for details.
Per-cpu maps are special, they store multiple values to a single key; a value for each CPU core. The
bpf_map_lookup_elem()returns all these values. But when I call aget()on a RawMap, it returns a single value only since it allocates a buffer which is as long as the map's value size. To handle per-cpu maps, I propose a newgetPerCPU()function which returns a buffer containing all values. wdyt?