Allow positional parameters in attachpoint definitions

[danobi]

This commit supports using positional parameters as follows:

  # bpftrace -e 'kprobe:$1 { 1 }' -- f
  Attaching 1 probe...

  # bpftrace -e 'kprobe:$1asdf { 1 }' -- f
  stdin:1:1-14: ERROR: Found trailing text 'asdf' in positional parameter index. Try quoting the trailing text.
  kprobe:$1asdf { 1 }
  ~~~~~~~~~~~~~

  # bpftrace -e 'kprobe:$1"asdf" { 1 }' -- f
  Attaching 1 probe...
  cannot attach kprobe, probe entry may not exist
  Error attaching probe: 'kprobe:fasdf'

This closes #606 .

Checklist

• Language changes are updated in docs/reference_guide.md
• User-visible and non-trivial changes updated in CHANGELOG.md

[mmisono]

This is definitely useful; I found that the followings do not work.

1. Use as int

% ./src/bpftrace -e 'i:ms:$1 { printf("hi\n"); }' 100 
stdin:1:1-8: ERROR: Failed to parse '': stoi
i:ss:$1 { printf("hi\n"); }
~~~~~~~

2. Use as offset

sudo ./src/bpftrace -e 'k:vfs_open+$1 { printf("hi\n"); }' 5
stdin:1:1-14: ERROR: Invalid offset
k:vfs_open+$1 { printf("hi\n"); }
~~~~~~~~~~~~~

% sudo ./src/bpftrace -e 'BEG$1 { printf("hi\n"); }' IN
stdin:1:1-6: ERROR: Unrecognized probe type: BEG
BEG$1 { printf("hi\n"); }
~~~~~

% sudo ./src/bpftrace -e 'BEGIN$1 { printf("hi\n"); }' IN
stdin:1:1-8: ERROR: Unrecognized probe type: BEGININ
BEGIN$1 { printf("hi\n"); }
~~~~~~~

[danobi]

@mmisono I fixed 1-3 with:

commit d925f154416d28d5a5bf8acb9a9748702db1c796 (HEAD -> params_in_aps)
Author: Daniel Xu <dxu@dxuuu.xyz>
Date:   Mon May 18 14:27:39 2020 -0700

    Load positional parameters before driver runs

    Before, we were loading positional parameters into bpftrace after the
    driver ran and before the AST was processed. This was fine for
    positional parameters used inside the probe body but not ok for
    positional parameters parsed in the driver (ie attach point defns).

diff --git a/src/main.cpp b/src/main.cpp
index cafe0b4..d497585 100644
--- a/src/main.cpp
+++ b/src/main.cpp
@@ -346,6 +346,13 @@ int main(int argc, char *argv[])
   BPFtrace bpftrace(std::move(output));
   Driver driver(bpftrace);

+  // positional parameters
+  while (optind < argc)
+  {
+    bpftrace.add_param(argv[optind]);
+    optind++;
+  }
+
   bpftrace.usdt_file_activation_ = usdt_file_activation;
   bpftrace.safe_mode_ = safe_mode;
   bpftrace.force_btf_ = force_btf;
@@ -453,12 +460,6 @@ int main(int argc, char *argv[])

   cap_memory_limits();

-  // positional parameters
-  while (optind < argc) {
-    bpftrace.add_param(argv[optind]);
-    optind++;
-  }
-
   // defaults
   bpftrace.join_argnum_ = 16;
   bpftrace.join_argsize_ = 1024;

What is the issue with 4? That seems like reasonable behavior to me.

[mmisono]

@danobi
oops, there is no issue with 4.

[danobi]

Ah I figured out why my tests passed initially. driver.parse() runs twice. My tests only had string arguments and the AttachPointParser accepts an empty string as an argument, but cannot accept an empty string as an numeric parameter (stoi fails). But before second driver.parse() runs the positional params are added.

I think the fix I put above is correct (modulo a small change).