Merge branch 'generate-enums' (#92)

Author: fuhry

attach_type.gen.go

@@ -0,0 +1,70 @@
+// @generated by bpf_enum_gen. DO NOT EDIT.
+// Source: /usr/include/linux/bpf.h
+
+package goebpf
+
+// AttachType is generated from enum bpf_attach_type in /usr/include/linux/bpf.h
+type AttachType int
+
+const (
+	AttachTypeCgroupInetIngress AttachType = iota
+	AttachTypeCgroupInetEgress
+	AttachTypeCgroupInetSockCreate
+	AttachTypeCgroupSockOps
+	AttachTypeSkSkbStreamParser
+	AttachTypeSkSkbStreamVerdict
+	AttachTypeCgroupDevice
+	AttachTypeSkMsgVerdict
+	AttachTypeCgroupInet4Bind
+	AttachTypeCgroupInet6Bind
+	AttachTypeCgroupInet4Connect
+	AttachTypeCgroupInet6Connect
+	AttachTypeCgroupInet4PostBind
+	AttachTypeCgroupInet6PostBind
+	AttachTypeCgroupUdp4Sendmsg
+	AttachTypeCgroupUdp6Sendmsg
+	AttachTypeLircMode2
+	AttachTypeFlowDissector
+	AttachTypeCgroupSysctl
+	AttachTypeCgroupUdp4Recvmsg
+	AttachTypeCgroupUdp6Recvmsg
+	AttachTypeCgroupGetsockopt
+	AttachTypeCgroupSetsockopt
+	AttachTypeTraceRawTp
+	AttachTypeTraceFentry
+	AttachTypeTraceFexit
+	AttachTypeModifyReturn
+	AttachTypeLsmMac
+	AttachTypeTraceIter
+	AttachTypeCgroupInet4Getpeername
+	AttachTypeCgroupInet6Getpeername
+	AttachTypeCgroupInet4Getsockname
+	AttachTypeCgroupInet6Getsockname
+	AttachTypeXdpDevmap
+	AttachTypeCgroupInetSockRelease
+	AttachTypeXdpCpumap
+	AttachTypeSkLookup
+	AttachTypeXdp
+	AttachTypeSkSkbVerdict
+	AttachTypeSkReuseportSelect
+	AttachTypeSkReuseportSelectOrMigrate
+	AttachTypePerfEvent
+	AttachTypeTraceKprobeMulti
+	AttachTypeLsmCgroup
+	AttachTypeStructOps
+	AttachTypeNetfilter
+	AttachTypeTcxIngress
+	AttachTypeTcxEgress
+	AttachTypeTraceUprobeMulti
+	AttachTypeCgroupUnixConnect
+	AttachTypeCgroupUnixSendmsg
+	AttachTypeCgroupUnixRecvmsg
+	AttachTypeCgroupUnixGetpeername
+	AttachTypeCgroupUnixGetsockname
+	AttachTypeNetkitPrimary
+	AttachTypeNetkitPeer
+	AttachTypeTraceKprobeSession
+	AttachTypeTraceUprobeSession
+	AttachTypeTraceFsession
+	maxBpfAttachType
+)

cmd/bpf_enum_gen/main.go

@@ -0,0 +1,303 @@
+// bpf_enum_gen generates Go constants from a C enum in a header file.
+// Usage: bpf_enum_gen -header <file> -enum <enum_name> -type <GoType> -pkg <package> -out <file>
+//
+// The C enum members are converted to Go identifiers by stripping the
+// longest common prefix (e.g. "BPF_PROG_TYPE_"), then title-casing each
+// underscore-separated word.  The __MAX_* sentinel becomes a private
+// constant named maxBpf<suffix>.
+package main
+
+import (
+	"bufio"
+	"flag"
+	"fmt"
+	"os"
+	"strings"
+	"text/template"
+	"unicode"
+)
+
+const outputTemplate = `// @generated by bpf_enum_gen. DO NOT EDIT.
+// Source: {{.Header}}
+
+package {{.Package}}
+
+// {{.GoType}} is generated from enum {{.EnumName}} in {{.Header}}
+type {{.GoType}} int
+
+const (
+{{- range .Members}}
+{{- if .First}}
+	{{.Name}} {{$.GoType}} = iota
+{{- else}}
+	{{.Name}}
+{{- end}}
+{{- end}}
+)
+`
+
+type member struct {
+	Name  string
+	First bool
+}
+
+type templateData struct {
+	Header   string
+	Package  string
+	GoType   string
+	EnumName string
+	Members  []member
+}
+
+func main() {
+	header := flag.String("header", "", "path to C header file")
+	enumName := flag.String("enum", "", "C enum name (e.g. bpf_prog_type)")
+	goType := flag.String("type", "", "Go type name (e.g. ProgramType)")
+	pkg := flag.String("pkg", "goebpf", "Go package name")
+	out := flag.String("out", "", "output file (default: stdout)")
+	flag.Parse()
+
+	if *header == "" || *enumName == "" || *goType == "" {
+		fmt.Fprintln(os.Stderr, "bpf_enum_gen: -header, -enum, and -type are required")
+		flag.Usage()
+		os.Exit(1)
+	}
+
+	members, err := parseEnum(*header, *enumName)
+	if err != nil {
+		fmt.Fprintf(os.Stderr, "bpf_enum_gen: %v\n", err)
+		os.Exit(1)
+	}
+
+	goNames := toGoNames(members, *goType)
+
+	data := templateData{
+		Header:   *header,
+		Package:  *pkg,
+		GoType:   *goType,
+		EnumName: *enumName,
+		Members:  goNames,
+	}
+
+	w := os.Stdout
+	if *out != "" {
+		f, err := os.Create(*out)
+		if err != nil {
+			fmt.Fprintf(os.Stderr, "bpf_enum_gen: %v\n", err)
+			os.Exit(1)
+		}
+		defer f.Close()
+		w = f
+	}
+
+	tmpl := template.Must(template.New("output").Parse(outputTemplate))
+	if err := tmpl.Execute(w, data); err != nil {
+		fmt.Fprintf(os.Stderr, "bpf_enum_gen: %v\n", err)
+		os.Exit(1)
+	}
+}
+
+// parseEnum extracts the raw C member names from the named enum in the file.
+func parseEnum(path, enumName string) ([]string, error) {
+	f, err := os.Open(path)
+	if err != nil {
+		return nil, err
+	}
+	defer f.Close()
+
+	target := "enum " + enumName + " {"
+	var members []string
+	inEnum := false
+	depth := 0
+
+	scanner := bufio.NewScanner(f)
+	for scanner.Scan() {
+		line := strings.TrimSpace(scanner.Text())
+
+		// Strip line comments
+		if idx := strings.Index(line, "//"); idx >= 0 {
+			line = strings.TrimSpace(line[:idx])
+		}
+		if idx := strings.Index(line, "/*"); idx >= 0 {
+			line = strings.TrimSpace(line[:idx])
+		}
+
+		if !inEnum {
+			if strings.Contains(line, target) {
+				inEnum = true
+				depth = strings.Count(line, "{") - strings.Count(line, "}")
+				// Grab any member on the same line as the opening brace
+				after := line[strings.Index(line, "{")+1:]
+				collectMembers(after, &members)
+			}
+			continue
+		}
+
+		depth += strings.Count(line, "{") - strings.Count(line, "}")
+		if depth <= 0 {
+			// Last line of enum body before closing brace
+			before := line
+			if idx := strings.LastIndex(line, "}"); idx >= 0 {
+				before = line[:idx]
+			}
+			collectMembers(before, &members)
+			break
+		}
+		collectMembers(line, &members)
+	}
+
+	if err := scanner.Err(); err != nil {
+		return nil, err
+	}
+	if !inEnum {
+		return nil, fmt.Errorf("enum %q not found in %s", enumName, path)
+	}
+	return members, nil
+}
+
+// collectMembers extracts comma-separated C identifiers (member names) from a
+// partial enum body line, stripping optional = <value> assignments.
+func collectMembers(line string, out *[]string) {
+	// Strip inline comments again (already done at line level, but handle residue)
+	if idx := strings.Index(line, "/*"); idx >= 0 {
+		line = line[:idx]
+	}
+	for _, tok := range strings.Split(line, ",") {
+		tok = strings.TrimSpace(tok)
+		if tok == "" {
+			continue
+		}
+		// Strip = value
+		if idx := strings.Index(tok, "="); idx >= 0 {
+			tok = strings.TrimSpace(tok[:idx])
+		}
+		if tok == "" || tok == "{" || tok == "}" {
+			continue
+		}
+		// Must look like a C identifier
+		if isIdent(tok) {
+			*out = append(*out, tok)
+		}
+	}
+}
+
+func isIdent(s string) bool {
+	if s == "" {
+		return false
+	}
+	for i, r := range s {
+		if i == 0 && !unicode.IsLetter(r) && r != '_' {
+			return false
+		}
+		if i > 0 && !unicode.IsLetter(r) && !unicode.IsDigit(r) && r != '_' {
+			return false
+		}
+	}
+	return true
+}
+
+// toGoNames converts C enum member names to Go identifiers.
+//
+// Strategy:
+//  1. Find the longest common prefix that is a full underscore-delimited
+//     token boundary (shared by all non-sentinel members).
+//  2. Strip that prefix, title-case the remaining words, prepend goType.
+//  3. __MAX_* sentinel → maxBpf<TitleCaseSuffix>.
+func toGoNames(cNames []string, goType string) []member {
+	// Separate sentinel (__MAX_*) from real members
+	var real []string
+	var sentinels []string
+	for _, n := range cNames {
+		if strings.HasPrefix(n, "__MAX_") || (strings.HasPrefix(n, "__") && strings.HasSuffix(n, "_MAX")) {
+			sentinels = append(sentinels, n)
+		} else {
+			real = append(real, n)
+		}
+	}
+
+	prefix := longestCommonTokenPrefix(real)
+
+	var result []member
+	for i, n := range real {
+		rest := strings.TrimPrefix(n, prefix)
+		result = append(result, member{
+			Name:  goType + titleCase(rest),
+			First: i == 0,
+		})
+	}
+
+	for range sentinels {
+		// __MAX_BPF_PROG_TYPE → maxBpfProgramType (always derived from goType)
+		// __MAX_BPF_ATTACH_TYPE → maxBpfAttachType
+		result = append(result, member{Name: "maxBpf" + goType})
+	}
+
+	return result
+}
+
+// longestCommonTokenPrefix finds the longest prefix, measured in whole
+// underscore-delimited tokens, shared by all strings.
+func longestCommonTokenPrefix(ss []string) string {
+	if len(ss) == 0 {
+		return ""
+	}
+	parts := strings.Split(ss[0], "_")
+	// Try prefixes from longest to shortest
+	for length := len(parts); length > 0; length-- {
+		candidate := strings.Join(parts[:length], "_") + "_"
+		allMatch := true
+		for _, s := range ss[1:] {
+			if !strings.HasPrefix(s, candidate) {
+				allMatch = false
+				break
+			}
+		}
+		if allMatch {
+			return candidate
+		}
+	}
+	return ""
+}
+
+// titleCase converts SNAKE_CASE to TitleCase.
+// Special handling for well-known abbreviations to match existing names.
+func titleCase(s string) string {
+	words := strings.Split(s, "_")
+	var b strings.Builder
+	for _, w := range words {
+		if w == "" {
+			continue
+		}
+		b.WriteString(titleCaseWord(w))
+	}
+	return b.String()
+}
+
+// knownAbbreviations are substrings that should keep their capitalisation.
+// Checked case-insensitively; the value is the exact output spelling.
+var knownAbbreviations = map[string]string{
+	"xdp":  "Xdp",
+	"skb":  "Skb",
+	"cls":  "Cls",
+	"act":  "Act",
+	"ops":  "Ops",
+	"msg":  "Msg",
+	"sk":   "Sk",
+	"lsm":  "Lsm",
+	"lrc":  "Lrc",
+	"udp":  "Udp",
+	"tcp":  "Tcp",
+	"ipv4": "Ipv4",
+	"ipv6": "Ipv6",
+}
+
+func titleCaseWord(w string) string {
+	lower := strings.ToLower(w)
+	if v, ok := knownAbbreviations[lower]; ok {
+		return v
+	}
+	if len(w) == 0 {
+		return w
+	}
+	return strings.ToUpper(w[:1]) + strings.ToLower(w[1:])
+}