Add builtin numaid, example:

NUMA node information:

    $ numactl --hardware
    available: 2 nodes (0-1)
    node 0 cpus: 0 1 2 3 4 5 6 7 ...
    [...]
    node 1 cpus: 20 21 22 23 24 25 ...
    [...]
    node distances:
    node   0   1
    0:  10  21
    1:  21  10

bpftrace script code net_action.bt:
    kprobe:net_tx_action,
    kprobe:net_rx_action
    {
        @[numaid, comm] = count();
    }

Example1:
    ------------------------------------------------------------
    iperf3 server:
    $ taskset -c 1 iperf3 -s
    [...]
    [  5]   0.00-1.00   sec  6.04 GBytes  51.9 Gbits/sec
    [...]

    iperf3 client:
    $ taskset -c 2 iperf3 -c 0
    [...]
    [  5]   0.00-1.00   sec  6.04 GBytes  51.9 Gbits/sec    0   1.19 MBytes
    [...]

    net_action.bt output:
    @[0, iperf3]: 2430857

    Because both cpu1 and cpu2 are on the numa node 0.

Example2:
    ------------------------------------------------------------
    iperf3 server:
    $ taskset -c 1 iperf3 -s
    [...]
    [  5]   0.00-1.00   sec  2.79 GBytes  24.0 Gbits/sec
    [...]

    iperf3 client:
    $ taskset -c 21 iperf3 -c 0
    [...]
    [  5]   0.00-1.00   sec  2.79 GBytes  24.0 Gbits/sec    0   1023 KBytes
    [...]

    net_action.bt output:
    @[0, iperf3]: 456193
    @[1, iperf3]: 684853

Because cpu1 on numa node0, cpu21 on numa node1, Processes that
need to communicate running on different nodes cause poor performance,
It is very easy to get numa with `numaid`.

CHANGELOG.md

@@ -9,6 +9,8 @@ and this project adheres to
 ## Unreleased
 
 #### Added
+- Add builtin: `numaid`
+  - [#2177](https://github.com/iovisor/bpftrace/pull/2177)
 - Add builtin function: `cgroup_path`
   - [#2055](https://github.com/iovisor/bpftrace/pull/2055)
 - Limit number of generated BPF programs

docs/reference_guide.md

@@ -1831,6 +1831,7 @@ NetworkManager:1155 /var/lib/sss/mc/passwd (deleted)
 - `gid` - Group ID
 - `nsecs` - Nanosecond timestamp
 - `elapsed` - Nanoseconds since bpftrace initialization
+- `numaid` - NUMA Node ID
 - `cpu` - Processor ID
 - `comm` - Process name
 - `kstack` - Kernel stack trace

man/adoc/bpftrace.adoc

@@ -936,6 +936,12 @@ For string arguments use the `str()` call to retrieve the value.
 | n/a
 | PID of the child process
 
+| numaid
+| uint32
+| 5.8
+| numa_node_id
+| ID of the NUMA node executing the BPF program
+
 | cpu
 | uint32
 | 4.1

src/ast/irbuilderbpf.cpp

@@ -906,6 +906,19 @@ CallInst *IRBuilderBPF::CreateGetUidGid()
   return createCall(getuidgid_func, {}, "get_uid_gid");
 }
 
+CallInst *IRBuilderBPF::CreateGetNumaId()
+{
+  // u32 bpf_get_numa_node_id(void)
+  // Return: NUMA Node ID
+  FunctionType *getnumaid_func_type = FunctionType::get(getInt64Ty(), false);
+  PointerType *getnumaid_func_ptr_type = PointerType::get(getnumaid_func_type, 0);
+  Constant *getnumaid_func = ConstantExpr::getCast(
+      Instruction::IntToPtr,
+      getInt64(libbpf::BPF_FUNC_get_numa_node_id),
+      getnumaid_func_ptr_type);
+  return createCall(getnumaid_func, {}, "get_numa_id");
+}
+
 CallInst *IRBuilderBPF::CreateGetCpuId()
 {
   // u32 bpf_raw_smp_processor_id(void)

src/ast/irbuilderbpf.h

@@ -120,6 +120,7 @@ class IRBuilderBPF : public IRBuilder<>
   CallInst   *CreateGetPidTgid();
   CallInst   *CreateGetCurrentCgroupId();
   CallInst   *CreateGetUidGid();
+  CallInst   *CreateGetNumaId();
   CallInst   *CreateGetCpuId();
   CallInst   *CreateGetCurrentTask();
   CallInst   *CreateGetRandom();

src/ast/passes/codegen_llvm.cpp

@@ -169,6 +169,10 @@ void CodegenLLVM::visit(Builtin &builtin)
       expr_ = b_.CreateLShr(uidgid, 32);
     }
   }
+  else if (builtin.ident == "numaid")
+  {
+    expr_ = b_.CreateGetNumaId();
+  }
   else if (builtin.ident == "cpu")
   {
     expr_ = b_.CreateGetCpuId();

src/ast/passes/semantic_analyser.cpp

@@ -285,7 +285,7 @@ void SemanticAnalyser::visit(Builtin &builtin)
            builtin.ident == "pid" || builtin.ident == "tid" ||
            builtin.ident == "cgroup" || builtin.ident == "uid" ||
            builtin.ident == "gid" || builtin.ident == "cpu" ||
-           builtin.ident == "rand")
+           builtin.ident == "rand" || builtin.ident == "numaid")
   {
     builtin.type = CreateUInt64();
     if (builtin.ident == "cgroup" &&

src/lexer.l

@@ -37,7 +37,7 @@ hspace   [ \t]
 vspace   [\n\r]
 space    {hspace}|{vspace}
 path     :(\\.|[_\-\./a-zA-Z0-9#\*])+
-builtin  arg[0-9]|args|cgroup|comm|cpid|cpu|ctx|curtask|elapsed|func|gid|nsecs|pid|probe|rand|retval|sarg[0-9]|tid|uid|username
+builtin  arg[0-9]|args|cgroup|comm|cpid|numaid|cpu|ctx|curtask|elapsed|func|gid|nsecs|pid|probe|rand|retval|sarg[0-9]|tid|uid|username
 call     avg|buf|cat|cgroupid|clear|count|delete|exit|hist|join|kaddr|kptr|ksym|lhist|macaddr|max|min|ntop|override|print|printf|cgroup_path|reg|signal|sizeof|stats|str|strftime|strncmp|sum|system|time|uaddr|uptr|usym|zero|path|unwatch
 
 /* Don't add to this! Use builtin OR call not both */

tests/codegen/builtin_numaid.cpp

@@ -0,0 +1,16 @@
+#include "common.h"
+
+namespace bpftrace {
+namespace test {
+namespace codegen {
+
+TEST(codegen, builtin_numaid)
+{
+  test("kprobe:f { @x = numaid }",
+
+       NAME);
+}
+
+} // namespace codegen
+} // namespace test
+} // namespace bpftrace

tests/parser.cpp

@@ -53,6 +53,7 @@ TEST(Parser, builtin_variables)
   test("kprobe:f { gid }", "Program\n kprobe:f\n  builtin: gid\n");
   test("kprobe:f { nsecs }", "Program\n kprobe:f\n  builtin: nsecs\n");
   test("kprobe:f { elapsed }", "Program\n kprobe:f\n  builtin: elapsed\n");
+  test("kprobe:f { numaid }", "Program\n kprobe:f\n  builtin: numaid\n");
   test("kprobe:f { cpu }", "Program\n kprobe:f\n  builtin: cpu\n");
   test("kprobe:f { curtask }", "Program\n kprobe:f\n  builtin: curtask\n");
   test("kprobe:f { rand }", "Program\n kprobe:f\n  builtin: rand\n");

tests/semantic_analyser.cpp

@@ -146,6 +146,7 @@ TEST(semantic_analyser, builtin_variables)
   test("kprobe:f { gid }", 0);
   test("kprobe:f { nsecs }", 0);
   test("kprobe:f { elapsed }", 0);
+  test("kprobe:f { numaid }", 0);
   test("kprobe:f { cpu }", 0);
   test("kprobe:f { curtask }", 0);
   test("kprobe:f { rand }", 0);