/*
GoCheese -- Python private package repository and caching proxy
-Copyright (C) 2019 Sergey Matveev <stargrave@stargrave.org>
+Copyright (C) 2019-2021 Sergey Matveev <stargrave@stargrave.org>
+ 2019-2021 Elena Balakhonova <balakhonova_e@riseup.net>
This program is free software: you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
import (
"bytes"
+ "context"
"crypto/sha256"
+ "crypto/tls"
"encoding/hex"
"errors"
"flag"
"fmt"
- "io"
"io/ioutil"
"log"
+ "net"
"net/http"
"net/url"
"os"
"runtime"
"strings"
"syscall"
+ "time"
+
+ "golang.org/x/net/netutil"
)
const (
- HTMLBegin = "<!DOCTYPE html><html><head><title>Links for %s</title></head><body><h1>Links for %s</h1>\n"
- HTMLEnd = "</body></html>"
- HTMLElement = "<a href='%s'>%s</a><br/>\n"
- SHA256Prefix = "sha256="
- SHA256Ext = ".sha256"
+ Version = "2.6.0"
+ HTMLBegin = `<!DOCTYPE html>
+<html>
+ <head>
+ <title>Links for %s</title>
+ </head>
+ <body>
+`
+ HTMLEnd = " </body>\n</html>\n"
+ HTMLElement = " <a href=\"%s\"%s>%s</a><br/>\n"
InternalFlag = ".internal"
+ GPGSigExt = ".asc"
Warranty = `This program is free software: you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
along with this program. If not, see <http://www.gnu.org/licenses/>.`
)
+const (
+ HashAlgoSHA256 = "sha256"
+ HashAlgoBLAKE2b256 = "blake2_256"
+ HashAlgoSHA512 = "sha512"
+ HashAlgoMD5 = "md5"
+)
+
var (
+ pkgPyPI = regexp.MustCompile(`^.*<a href="([^"]+)"[^>]*>(.+)</a><br/>.*$`)
+ normalizationRe = regexp.MustCompilePOSIX("[-_.]+")
+
+ knownHashAlgos []string = []string{
+ HashAlgoSHA256,
+ HashAlgoBLAKE2b256,
+ HashAlgoSHA512,
+ HashAlgoMD5,
+ }
+
root = flag.String("root", "./packages", "Path to packages directory")
bind = flag.String("bind", "[::]:8080", "Address to bind to")
+ tlsCert = flag.String("tls-cert", "", "Path to TLS X.509 certificate")
+ tlsKey = flag.String("tls-key", "", "Path to TLS X.509 private key")
norefreshURLPath = flag.String("norefresh", "/norefresh/", "Non-refreshing URL path")
refreshURLPath = flag.String("refresh", "/simple/", "Auto-refreshing URL path")
+ gpgUpdateURLPath = flag.String("gpgupdate", "/gpgupdate/", "GPG forceful refreshing URL path")
pypiURL = flag.String("pypi", "https://pypi.org/simple/", "Upstream PyPI URL")
+ pypiCertHash = flag.String("pypi-cert-hash", "", "Authenticate PyPI by its X.509 certificate's SHA256 hash")
passwdPath = flag.String("passwd", "passwd", "Path to file with authenticators")
+ logTimestamped = flag.Bool("log-timestamped", false, "Prepend timestmap to log messages")
passwdCheck = flag.Bool("passwd-check", false, "Test the -passwd file for syntax errors and exit")
- fsck = flag.Bool("fsck", false, "Check integrity of all packages")
+ fsck = flag.Bool("fsck", false, "Check integrity of all packages (errors are in stderr)")
+ maxClients = flag.Int("maxclients", 128, "Maximal amount of simultaneous clients")
version = flag.Bool("version", false, "Print version information")
warranty = flag.Bool("warranty", false, "Print warranty information")
- pkgPyPI = regexp.MustCompile(`^.*<a href="([^"]+)"[^>]*>(.+)</a><br/>.*$`)
- Version string = "UNKNOWN"
-
- passwords map[string]Auther = make(map[string]Auther)
+ killed bool
+ pypiURLParsed *url.URL
)
-type Auther interface {
- Auth(password string) bool
-}
-
-func mkdirForPkg(w http.ResponseWriter, r *http.Request, dir string) bool {
- path := filepath.Join(*root, dir)
+func mkdirForPkg(w http.ResponseWriter, r *http.Request, pkgName string) bool {
+ path := filepath.Join(*root, pkgName)
if _, err := os.Stat(path); os.IsNotExist(err) {
- if err = os.Mkdir(path, 0700); err != nil {
- http.Error(w, err.Error(), http.StatusInternalServerError)
- return false
- }
- log.Println(r.RemoteAddr, "mkdir", dir)
- }
- return true
-}
-
-func refreshDir(w http.ResponseWriter, r *http.Request, dir, filenameGet string) bool {
- if _, err := os.Stat(filepath.Join(*root, dir, InternalFlag)); err == nil {
- log.Println(r.RemoteAddr, "pypi refresh skip, internal package", dir)
- return true
- }
- log.Println(r.RemoteAddr, "pypi refresh", dir)
- resp, err := http.Get(*pypiURL + dir + "/")
- if err != nil {
- http.Error(w, err.Error(), http.StatusBadGateway)
- return false
- }
- defer resp.Body.Close()
- body, err := ioutil.ReadAll(resp.Body)
- if err != nil {
- http.Error(w, err.Error(), http.StatusBadGateway)
- return false
- }
- if !mkdirForPkg(w, r, dir) {
- return false
- }
- var submatches []string
- var uri string
- var filename string
- var path string
- var pkgURL *url.URL
- var digest []byte
- for _, lineRaw := range bytes.Split(body, []byte("\n")) {
- submatches = pkgPyPI.FindStringSubmatch(string(lineRaw))
- if len(submatches) == 0 {
- continue
- }
- uri = submatches[1]
- filename = submatches[2]
- if pkgURL, err = url.Parse(uri); err != nil {
- http.Error(w, err.Error(), http.StatusInternalServerError)
- return false
- }
- digest, err = hex.DecodeString(strings.TrimPrefix(pkgURL.Fragment, SHA256Prefix))
- if err != nil {
- http.Error(w, err.Error(), http.StatusBadGateway)
- return false
- }
- if filename == filenameGet {
- log.Println(r.RemoteAddr, "pypi download", filename)
- path = filepath.Join(*root, dir, filename)
- resp, err = http.Get(uri)
- if err != nil {
- http.Error(w, err.Error(), http.StatusBadGateway)
- return false
- }
- defer resp.Body.Close()
- hasher := sha256.New()
- dst, err := ioutil.TempFile(filepath.Join(*root, dir), "")
- if err != nil {
- http.Error(w, err.Error(), http.StatusInternalServerError)
- return false
- }
- wr := io.MultiWriter(hasher, dst)
- if _, err = io.Copy(wr, resp.Body); err != nil {
- os.Remove(dst.Name())
- dst.Close()
- http.Error(w, err.Error(), http.StatusInternalServerError)
- return false
- }
- if bytes.Compare(hasher.Sum(nil), digest) != 0 {
- log.Println(r.RemoteAddr, "pypi", filename, "digest mismatch")
- os.Remove(dst.Name())
- dst.Close()
- http.Error(w, err.Error(), http.StatusBadGateway)
- return false
- }
- if err = dst.Sync(); err != nil {
- os.Remove(dst.Name())
- dst.Close()
- http.Error(w, err.Error(), http.StatusInternalServerError)
- return false
- }
- dst.Close()
- if err = os.Rename(dst.Name(), path); err != nil {
- http.Error(w, err.Error(), http.StatusInternalServerError)
- return false
- }
- }
- path = filepath.Join(*root, dir, filename+SHA256Ext)
- _, err = os.Stat(path)
- if err == nil {
- continue
- } else {
- if !os.IsNotExist(err) {
- http.Error(w, err.Error(), http.StatusInternalServerError)
- return false
- }
- }
- log.Println(r.RemoteAddr, "pypi touch", filename)
- if err = ioutil.WriteFile(path, digest, os.FileMode(0600)); err != nil {
+ if err = os.Mkdir(path, os.FileMode(0777)); err != nil {
+ log.Println("error", r.RemoteAddr, "mkdir", pkgName, err)
http.Error(w, err.Error(), http.StatusInternalServerError)
return false
}
+ log.Println(r.RemoteAddr, "mkdir", pkgName)
}
return true
}
func listRoot(w http.ResponseWriter, r *http.Request) {
- log.Println(r.RemoteAddr, "root")
files, err := ioutil.ReadDir(*root)
if err != nil {
+ log.Println("error", r.RemoteAddr, "root", err)
http.Error(w, err.Error(), http.StatusInternalServerError)
return
}
- w.Write([]byte(fmt.Sprintf(HTMLBegin, "root", "root")))
+ var result bytes.Buffer
+ result.WriteString(fmt.Sprintf(HTMLBegin, "root"))
for _, file := range files {
if file.Mode().IsDir() {
- w.Write([]byte(fmt.Sprintf(
+ result.WriteString(fmt.Sprintf(
HTMLElement,
*refreshURLPath+file.Name()+"/",
- file.Name(),
- )))
+ "", file.Name(),
+ ))
}
}
- w.Write([]byte(HTMLEnd))
+ result.WriteString(HTMLEnd)
+ w.Write(result.Bytes())
}
-func listDir(w http.ResponseWriter, r *http.Request, dir string, autorefresh bool) {
- log.Println(r.RemoteAddr, "dir", dir)
- dirPath := filepath.Join(*root, dir)
+func listDir(
+ w http.ResponseWriter,
+ r *http.Request,
+ pkgName string,
+ autorefresh, gpgUpdate bool,
+) {
+ dirPath := filepath.Join(*root, pkgName)
if autorefresh {
- if !refreshDir(w, r, dir, "") {
+ if !refreshDir(w, r, pkgName, "", gpgUpdate) {
return
}
- } else if _, err := os.Stat(dirPath); os.IsNotExist(err) && !refreshDir(w, r, dir, "") {
+ } else if _, err := os.Stat(dirPath); os.IsNotExist(err) && !refreshDir(w, r, pkgName, "", false) {
return
}
- files, err := ioutil.ReadDir(dirPath)
+ fis, err := ioutil.ReadDir(dirPath)
if err != nil {
+ log.Println("error", r.RemoteAddr, "list", pkgName, err)
http.Error(w, err.Error(), http.StatusInternalServerError)
return
}
- w.Write([]byte(fmt.Sprintf(HTMLBegin, dir, dir)))
- var data []byte
- var filenameClean string
- for _, file := range files {
- if !strings.HasSuffix(file.Name(), SHA256Ext) {
- continue
- }
- data, err = ioutil.ReadFile(filepath.Join(dirPath, file.Name()))
- if err != nil {
- http.Error(w, err.Error(), http.StatusInternalServerError)
- return
+ files := make(map[string]struct{}, len(fis)/2)
+ for _, fi := range fis {
+ files[fi.Name()] = struct{}{}
+ }
+ var result bytes.Buffer
+ result.WriteString(fmt.Sprintf(HTMLBegin, pkgName))
+ for _, algo := range knownHashAlgos {
+ for fn := range files {
+ if killed {
+ // Skip expensive I/O when shutting down
+ http.Error(w, "shutting down", http.StatusInternalServerError)
+ return
+ }
+ if !strings.HasSuffix(fn, "."+algo) {
+ continue
+ }
+ delete(files, fn)
+ digest, err := ioutil.ReadFile(filepath.Join(dirPath, fn))
+ if err != nil {
+ log.Println("error", r.RemoteAddr, "list", fn, err)
+ http.Error(w, err.Error(), http.StatusInternalServerError)
+ return
+ }
+ fnClean := strings.TrimSuffix(fn, "."+algo)
+ delete(files, fnClean)
+ gpgSigAttr := ""
+ if _, err = os.Stat(filepath.Join(dirPath, fnClean+GPGSigExt)); err == nil {
+ gpgSigAttr = " data-gpg-sig=true"
+ delete(files, fnClean+GPGSigExt)
+ }
+ result.WriteString(fmt.Sprintf(
+ HTMLElement,
+ strings.Join([]string{
+ *refreshURLPath, pkgName, "/", fnClean,
+ "#", algo, "=", hex.EncodeToString(digest),
+ }, ""),
+ gpgSigAttr,
+ fnClean,
+ ))
}
- filenameClean = strings.TrimSuffix(file.Name(), SHA256Ext)
- w.Write([]byte(fmt.Sprintf(
- HTMLElement,
- strings.Join([]string{
- *refreshURLPath, dir, "/",
- filenameClean, "#", SHA256Prefix, string(data),
- }, ""),
- filenameClean,
- )))
}
- w.Write([]byte(HTMLEnd))
+ result.WriteString(HTMLEnd)
+ w.Write(result.Bytes())
}
-func servePkg(w http.ResponseWriter, r *http.Request, dir, filename string) {
- log.Println(r.RemoteAddr, "pkg", filename)
- path := filepath.Join(*root, dir, filename)
+func servePkg(w http.ResponseWriter, r *http.Request, pkgName, filename string) {
+ log.Println(r.RemoteAddr, "get", filename)
+ path := filepath.Join(*root, pkgName, filename)
if _, err := os.Stat(path); os.IsNotExist(err) {
- if !refreshDir(w, r, dir, filename) {
+ if !refreshDir(w, r, pkgName, filename, false) {
return
}
}
http.ServeFile(w, r, path)
}
-func strToAuther(verifier string) (string, Auther, error) {
- st := strings.SplitN(verifier, "$", 3)
- if len(st) != 3 || st[0] != "" {
- return "", nil, errors.New("invalid verifier structure")
- }
- algorithm := st[1]
- var auther Auther
- var err error
- switch algorithm {
- case "argon2i":
- auther, err = parseArgon2i(st[2])
- case "sha256":
- auther, err = parseSHA256(st[2])
- default:
- err = errors.New("unknown hashing algorithm")
- }
- return algorithm, auther, err
-}
-
-func serveUpload(w http.ResponseWriter, r *http.Request) {
- username, password, ok := r.BasicAuth()
- if !ok {
- log.Println(r.RemoteAddr, "unauthenticated", username)
- http.Error(w, "unauthenticated", http.StatusUnauthorized)
- return
- }
- auther, ok := passwords[username]
- if !ok || !auther.Auth(password) {
- log.Println(r.RemoteAddr, "unauthenticated", username)
- http.Error(w, "unauthenticated", http.StatusUnauthorized)
- return
- }
- var err error
- if err = r.ParseMultipartForm(1 << 20); err != nil {
- http.Error(w, err.Error(), http.StatusBadRequest)
- return
- }
- for _, file := range r.MultipartForm.File["content"] {
- filename := file.Filename
- log.Println(r.RemoteAddr, "upload", filename, "by", username)
- dir := filename[:strings.LastIndex(filename, "-")]
- dirPath := filepath.Join(*root, dir)
- path := filepath.Join(dirPath, filename)
- if _, err = os.Stat(path); err == nil {
- log.Println(r.RemoteAddr, "already exists", filename)
- http.Error(w, "Already exists", http.StatusBadRequest)
- return
- }
- if !mkdirForPkg(w, r, dir) {
- return
- }
- internalPath := filepath.Join(dirPath, InternalFlag)
- var dst *os.File
- if _, err = os.Stat(internalPath); os.IsNotExist(err) {
- if dst, err = os.Create(internalPath); err != nil {
- http.Error(w, err.Error(), http.StatusInternalServerError)
- return
- }
- dst.Close()
- }
- src, err := file.Open()
- defer src.Close()
- if err != nil {
- http.Error(w, err.Error(), http.StatusInternalServerError)
- return
- }
- dst, err = ioutil.TempFile(dirPath, "")
- if err != nil {
- http.Error(w, err.Error(), http.StatusInternalServerError)
- return
- }
- hasher := sha256.New()
- wr := io.MultiWriter(hasher, dst)
- if _, err = io.Copy(wr, src); err != nil {
- os.Remove(dst.Name())
- dst.Close()
- http.Error(w, err.Error(), http.StatusInternalServerError)
- return
- }
- if err = dst.Sync(); err != nil {
- os.Remove(dst.Name())
- dst.Close()
- http.Error(w, err.Error(), http.StatusInternalServerError)
- return
- }
- dst.Close()
- if err = os.Rename(dst.Name(), path); err != nil {
- http.Error(w, err.Error(), http.StatusInternalServerError)
- return
- }
- if err = ioutil.WriteFile(
- path+SHA256Ext,
- hasher.Sum(nil),
- os.FileMode(0600),
- ); err != nil {
- http.Error(w, err.Error(), http.StatusInternalServerError)
- return
- }
- }
-}
-
func handler(w http.ResponseWriter, r *http.Request) {
- if r.Method == "GET" {
+ switch r.Method {
+ case "GET":
var path string
var autorefresh bool
+ var gpgUpdate bool
if strings.HasPrefix(r.URL.Path, *norefreshURLPath) {
path = strings.TrimPrefix(r.URL.Path, *norefreshURLPath)
- autorefresh = false
- } else {
+ } else if strings.HasPrefix(r.URL.Path, *refreshURLPath) {
path = strings.TrimPrefix(r.URL.Path, *refreshURLPath)
autorefresh = true
+ } else if strings.HasPrefix(r.URL.Path, *gpgUpdateURLPath) {
+ path = strings.TrimPrefix(r.URL.Path, *gpgUpdateURLPath)
+ autorefresh = true
+ gpgUpdate = true
+ } else {
+ http.Error(w, "unknown action", http.StatusBadRequest)
+ return
}
parts := strings.Split(strings.TrimSuffix(path, "/"), "/")
if len(parts) > 2 {
if parts[0] == "" {
listRoot(w, r)
} else {
- listDir(w, r, parts[0], autorefresh)
+ listDir(w, r, parts[0], autorefresh, gpgUpdate)
}
} else {
servePkg(w, r, parts[0], parts[1])
}
- } else if r.Method == "POST" {
+ case "POST":
serveUpload(w, r)
+ default:
+ http.Error(w, "unknown action", http.StatusBadRequest)
}
}
-func goodIntegrity() bool {
- dirs, err := ioutil.ReadDir(*root)
- if err != nil {
- log.Fatal(err)
- }
- hasher := sha256.New()
- digest := make([]byte, sha256.Size)
- isGood := true
- var data []byte
- var pkgName string
- for _, dir := range dirs {
- files, err := ioutil.ReadDir(filepath.Join(*root, dir.Name()))
- if err != nil {
- log.Fatal(err)
- }
- for _, file := range files {
- if !strings.HasSuffix(file.Name(), SHA256Ext) {
- continue
- }
- pkgName = strings.TrimSuffix(file.Name(), SHA256Ext)
- data, err = ioutil.ReadFile(filepath.Join(*root, dir.Name(), pkgName))
- if err != nil {
- if os.IsNotExist(err) {
- continue
- }
- log.Fatal(err)
- }
- hasher.Write(data)
- data, err = ioutil.ReadFile(filepath.Join(*root, dir.Name(), file.Name()))
- if err != nil {
- log.Fatal(err)
- }
- if bytes.Compare(hasher.Sum(digest[:0]), data) == 0 {
- log.Println(pkgName, "GOOD")
- } else {
- isGood = false
- log.Println(pkgName, "BAD")
- }
- hasher.Reset()
- }
- }
- return isGood
-}
-
func main() {
flag.Parse()
if *warranty {
return
}
if *version {
- fmt.Println("GoCheese version " + Version + " built with " + runtime.Version())
+ fmt.Println("GoCheese", Version, "built with", runtime.Version())
return
}
+
+ if *logTimestamped {
+ log.SetFlags(log.Ldate | log.Lmicroseconds | log.Lshortfile)
+ } else {
+ log.SetFlags(log.Lshortfile)
+ }
+ log.SetOutput(os.Stdout)
+
if *fsck {
if !goodIntegrity() {
os.Exit(1)
}
return
}
+
if *passwdCheck {
refreshPasswd()
return
}
+
+ if (*tlsCert != "" && *tlsKey == "") || (*tlsCert == "" && *tlsKey != "") {
+ log.Fatalln("Both -tls-cert and -tls-key are required")
+ }
+
+ var err error
+ pypiURLParsed, err = url.Parse(*pypiURL)
+ if err != nil {
+ log.Fatalln(err)
+ }
refreshPasswd()
- log.Println("root:", *root, "bind:", *bind)
+ if *pypiCertHash == "" {
+ pypiHTTPTransport = http.Transport{}
+ } else {
+ ourDgst, err := hex.DecodeString(*pypiCertHash)
+ if err != nil {
+ log.Fatalln(err)
+ }
+ pypiHTTPTransport = http.Transport{
+ TLSClientConfig: &tls.Config{
+ VerifyConnection: func(s tls.ConnectionState) error {
+ spki := s.VerifiedChains[0][0].RawSubjectPublicKeyInfo
+ theirDgst := sha256.Sum256(spki)
+ if bytes.Compare(ourDgst, theirDgst[:]) != 0 {
+ return errors.New("certificate's digest mismatch")
+ }
+ return nil
+ }},
+ }
+ }
+
+ ln, err := net.Listen("tcp", *bind)
+ if err != nil {
+ log.Fatal(err)
+ }
+ ln = netutil.LimitListener(ln, *maxClients)
+ server := &http.Server{
+ ReadTimeout: time.Minute,
+ WriteTimeout: time.Minute,
+ }
+ http.HandleFunc(*norefreshURLPath, handler)
+ http.HandleFunc(*refreshURLPath, handler)
+ if *gpgUpdateURLPath != "" {
+ http.HandleFunc(*gpgUpdateURLPath, handler)
+ }
+
needsRefreshPasswd := make(chan os.Signal, 0)
+ needsShutdown := make(chan os.Signal, 0)
+ exitErr := make(chan error, 0)
signal.Notify(needsRefreshPasswd, syscall.SIGHUP)
+ signal.Notify(needsShutdown, syscall.SIGTERM, syscall.SIGINT)
go func() {
for range needsRefreshPasswd {
+ log.Println("refreshing passwords")
refreshPasswd()
}
}()
- http.HandleFunc(*norefreshURLPath, handler)
- http.HandleFunc(*refreshURLPath, handler)
- log.Fatal(http.ListenAndServe(*bind, nil))
+ go func(s *http.Server) {
+ <-needsShutdown
+ killed = true
+ log.Println("shutting down")
+ ctx, cancel := context.WithTimeout(context.TODO(), time.Minute)
+ exitErr <- s.Shutdown(ctx)
+ cancel()
+ }(server)
+
+ log.Println(
+ "GoCheese", Version, "listens:",
+ "root:", *root,
+ "bind:", *bind,
+ "pypi:", *pypiURL,
+ )
+ if *tlsCert == "" {
+ err = server.Serve(ln)
+ } else {
+ err = server.ServeTLS(ln, *tlsCert, *tlsKey)
+ }
+ if err != http.ErrServerClosed {
+ log.Fatal(err)
+ }
+ if err := <-exitErr; err != nil {
+ log.Fatal(err)
+ }
}