pkg/vcs: pave way for multi-vcs support

Wrap current git interface in abstract interface.
Provide constructor that create repo interface for the given os/vm.

6 files changed, 348 insertions, 270 deletions

diff --git a/pkg/bisect/bisect.go b/pkg/bisect/bisect.go
index 3dab05290..3e136276e 100644
--- a/pkg/bisect/bisect.go
+++ b/pkg/bisect/bisect.go
@@ -51,6 +51,7 @@ type ReproConfig struct {
 
 type env struct {
 	cfg       *Config
+	repo      vcs.Repo
 	head      *vcs.Commit
 	inst      *instance.Env
 	numTests  int
@@ -63,8 +64,13 @@ type buildEnv struct {
 }
 
 func Run(cfg *Config) (*vcs.Commit, error) {
+	repo, err := vcs.NewRepo(cfg.Manager.TargetOS, cfg.Manager.Type, cfg.Manager.KernelSrc)
+	if err != nil {
+		return nil, err
+	}
 	env := &env{
-		cfg: cfg,
+		cfg:  cfg,
+		repo: repo,
 	}
 	if cfg.Fix {
 		env.log("searching for fixing commit since %v", cfg.Kernel.Commit)
@@ -98,7 +104,7 @@ func (env *env) bisect() (*vcs.Commit, error) {
 	if env.inst, err = instance.NewEnv(&cfg.Manager); err != nil {
 		return nil, err
 	}
-	if env.head, err = vcs.Poll(cfg.Manager.KernelSrc, cfg.Kernel.Repo, cfg.Kernel.Branch); err != nil {
+	if env.head, err = env.repo.Poll(cfg.Kernel.Repo, cfg.Kernel.Branch); err != nil {
 		return nil, err
 	}
 	if err := build.Clean(cfg.Manager.TargetOS, cfg.Manager.TargetArch,
@@ -109,7 +115,7 @@ func (env *env) bisect() (*vcs.Commit, error) {
 	if err := env.inst.BuildSyzkaller(cfg.Syzkaller.Repo, cfg.Syzkaller.Commit); err != nil {
 		return nil, err
 	}
-	if _, err := vcs.SwitchCommit(cfg.Manager.KernelSrc, cfg.Kernel.Commit); err != nil {
+	if _, err := env.repo.SwitchCommit(cfg.Kernel.Commit); err != nil {
 		return nil, err
 	}
 	if res, err := env.test(); err != nil {
@@ -127,7 +133,7 @@ func (env *env) bisect() (*vcs.Commit, error) {
 	if good == "" {
 		return nil, nil // still not fixed
 	}
-	return vcs.Bisect(cfg.Manager.KernelSrc, bad, good, cfg.Trace, func() (vcs.BisectResult, error) {
+	return env.repo.Bisect(bad, good, cfg.Trace, func() (vcs.BisectResult, error) {
 		res, err := env.test()
 		if cfg.Fix {
 			if res == vcs.BisectBad {
@@ -149,7 +155,7 @@ func (env *env) commitRange() (*vcs.Commit, string, string, error) {
 
 func (env *env) commitRangeForFix() (*vcs.Commit, string, string, error) {
 	env.log("testing current HEAD %v", env.head.Hash)
-	if _, err := vcs.SwitchCommit(env.cfg.Manager.KernelSrc, env.head.Hash); err != nil {
+	if _, err := env.repo.SwitchCommit(env.head.Hash); err != nil {
 		return nil, "", "", err
 	}
 	res, err := env.test()
@@ -164,7 +170,7 @@ func (env *env) commitRangeForFix() (*vcs.Commit, string, string, error) {
 
 func (env *env) commitRangeForBug() (*vcs.Commit, string, string, error) {
 	cfg := env.cfg
-	tags, err := vcs.PreviousReleaseTags(cfg.Manager.KernelSrc, cfg.Kernel.Commit)
+	tags, err := env.repo.PreviousReleaseTags(cfg.Kernel.Commit)
 	if err != nil {
 		return nil, "", "", err
 	}
@@ -183,7 +189,7 @@ func (env *env) commitRangeForBug() (*vcs.Commit, string, string, error) {
 	lastBad := cfg.Kernel.Commit
 	for i, tag := range tags {
 		env.log("testing release %v", tag)
-		commit, err := vcs.SwitchCommit(cfg.Manager.KernelSrc, tag)
+		commit, err := env.repo.SwitchCommit(tag)
 		if err != nil {
 			return nil, "", "", err
 		}
@@ -207,7 +213,7 @@ func (env *env) commitRangeForBug() (*vcs.Commit, string, string, error) {
 func (env *env) test() (vcs.BisectResult, error) {
 	cfg := env.cfg
 	env.numTests++
-	current, err := vcs.HeadCommit(cfg.Manager.KernelSrc)
+	current, err := env.repo.HeadCommit()
 	if err != nil {
 		return 0, err
 	}
@@ -303,7 +309,7 @@ func (env *env) processResults(current *vcs.Commit, results []error) (bad, good
 // Note: linux-specific.
 func (env *env) buildEnvForCommit(commit string) (*buildEnv, error) {
 	cfg := env.cfg
-	tags, err := vcs.PreviousReleaseTags(cfg.Manager.KernelSrc, commit)
+	tags, err := env.repo.PreviousReleaseTags(commit)
 	if err != nil {
 		return nil, err
 	}
diff --git a/pkg/instance/instance.go b/pkg/instance/instance.go
index a41a4cbc2..79616f3fa 100644
--- a/pkg/instance/instance.go
+++ b/pkg/instance/instance.go
@@ -59,7 +59,7 @@ func (env *Env) BuildSyzkaller(repo, commit string) error {
 	if srcIndex == -1 {
 		return fmt.Errorf("syzkaller path %q is not in GOPATH", cfg.Syzkaller)
 	}
-	if _, err := vcs.CheckoutCommit(cfg.Syzkaller, repo, commit); err != nil {
+	if _, err := vcs.NewSyzkallerRepo(cfg.Syzkaller).CheckoutCommit(repo, commit); err != nil {
 		return fmt.Errorf("failed to checkout syzkaller repo: %v", err)
 	}
 	cmd := osutil.Command("make", "target")
diff --git a/pkg/vcs/git.go b/pkg/vcs/git.go
index e942b3c39..d1595aaa8 100644
--- a/pkg/vcs/git.go
+++ b/pkg/vcs/git.go
@@ -12,7 +12,6 @@ import (
 	"net/mail"
 	"os"
 	"os/exec"
-	"regexp"
 	"sort"
 	"strconv"
 	"strings"
@@ -21,21 +20,28 @@ import (
 	"github.com/google/syzkaller/pkg/osutil"
 )
 
-const (
-	DateFormat = "Mon Jan 2 15:04:05 2006 -0700"
-	timeout    = time.Hour // timeout for all git invocations
-)
+type git struct {
+	os  string
+	vm  string
+	dir string
+}
+
+func newGit(os, vm, dir string) *git {
+	return &git{
+		os:  os,
+		vm:  vm,
+		dir: dir,
+	}
+}
 
-// Poll checkouts the specified repository/branch in dir.
-// This involves fetching/resetting/cloning as necessary to recover from all possible problems.
-// Returns hash of the HEAD commit in the specified branch.
-func Poll(dir, repo, branch string) (*Commit, error) {
+func (git *git) Poll(repo, branch string) (*Commit, error) {
+	dir := git.dir
 	runSandboxed(dir, "git", "bisect", "reset")
 	runSandboxed(dir, "git", "reset", "--hard")
 	origin, err := runSandboxed(dir, "git", "remote", "get-url", "origin")
 	if err != nil || strings.TrimSpace(string(origin)) != repo {
 		// The repo is here, but it has wrong origin (e.g. repo in config has changed), re-clone.
-		if err := clone(dir, repo, branch); err != nil {
+		if err := git.clone(repo, branch); err != nil {
 			return nil, err
 		}
 	}
@@ -44,27 +50,27 @@ func Poll(dir, repo, branch string) (*Commit, error) {
 	// stuck with the local version forever (git checkout won't fail).
 	if _, err := runSandboxed(dir, "git", "checkout", "origin/"+branch); err != nil {
 		// No such branch (e.g. branch in config has changed), re-clone.
-		if err := clone(dir, repo, branch); err != nil {
+		if err := git.clone(repo, branch); err != nil {
 			return nil, err
 		}
 	}
 	if _, err := runSandboxed(dir, "git", "fetch", "--no-tags"); err != nil {
 		// Something else is wrong, re-clone.
-		if err := clone(dir, repo, branch); err != nil {
+		if err := git.clone(repo, branch); err != nil {
 			return nil, err
 		}
 	}
 	if _, err := runSandboxed(dir, "git", "checkout", "origin/"+branch); err != nil {
 		return nil, err
 	}
-	return HeadCommit(dir)
+	return git.HeadCommit()
 }
 
-// CheckoutBranch checkouts the specified repository/branch in dir.
-func CheckoutBranch(dir, repo, branch string) (*Commit, error) {
+func (git *git) CheckoutBranch(repo, branch string) (*Commit, error) {
+	dir := git.dir
 	runSandboxed(dir, "git", "bisect", "reset")
 	if _, err := runSandboxed(dir, "git", "reset", "--hard"); err != nil {
-		if err := initRepo(dir); err != nil {
+		if err := git.initRepo(); err != nil {
 			return nil, err
 		}
 	}
@@ -75,14 +81,14 @@ func CheckoutBranch(dir, repo, branch string) (*Commit, error) {
 	if _, err := runSandboxed(dir, "git", "checkout", "FETCH_HEAD"); err != nil {
 		return nil, err
 	}
-	return HeadCommit(dir)
+	return git.HeadCommit()
 }
 
-// CheckoutCommit checkouts the specified repository on the specified commit in dir.
-func CheckoutCommit(dir, repo, commit string) (*Commit, error) {
+func (git *git) CheckoutCommit(repo, commit string) (*Commit, error) {
+	dir := git.dir
 	runSandboxed(dir, "git", "bisect", "reset")
 	if _, err := runSandboxed(dir, "git", "reset", "--hard"); err != nil {
-		if err := initRepo(dir); err != nil {
+		if err := git.initRepo(); err != nil {
 			return nil, err
 		}
 	}
@@ -90,73 +96,65 @@ func CheckoutCommit(dir, repo, commit string) (*Commit, error) {
 	if err != nil {
 		return nil, err
 	}
-	return SwitchCommit(dir, commit)
+	return git.SwitchCommit(commit)
 }
 
-// SwitchCommit checkouts the specified commit without fetching.
-func SwitchCommit(dir, commit string) (*Commit, error) {
+func (git *git) SwitchCommit(commit string) (*Commit, error) {
+	dir := git.dir
 	if _, err := runSandboxed(dir, "git", "checkout", commit); err != nil {
 		return nil, err
 	}
-	return HeadCommit(dir)
+	return git.HeadCommit()
 }
 
-func clone(dir, repo, branch string) error {
-	if err := initRepo(dir); err != nil {
+func (git *git) clone(repo, branch string) error {
+	if err := git.initRepo(); err != nil {
 		return err
 	}
-	if _, err := runSandboxed(dir, "git", "remote", "add", "origin", repo); err != nil {
+	if _, err := runSandboxed(git.dir, "git", "remote", "add", "origin", repo); err != nil {
 		return err
 	}
-	if _, err := runSandboxed(dir, "git", "fetch", "origin", branch); err != nil {
+	if _, err := runSandboxed(git.dir, "git", "fetch", "origin", branch); err != nil {
 		return err
 	}
 	return nil
 }
 
-func initRepo(dir string) error {
-	if err := os.RemoveAll(dir); err != nil {
+func (git *git) initRepo() error {
+	if err := os.RemoveAll(git.dir); err != nil {
 		return fmt.Errorf("failed to remove repo dir: %v", err)
 	}
-	if err := osutil.MkdirAll(dir); err != nil {
+	if err := osutil.MkdirAll(git.dir); err != nil {
 		return fmt.Errorf("failed to create repo dir: %v", err)
 	}
-	if err := osutil.SandboxChown(dir); err != nil {
+	if err := osutil.SandboxChown(git.dir); err != nil {
 		return err
 	}
-	if _, err := runSandboxed(dir, "git", "init"); err != nil {
+	if _, err := runSandboxed(git.dir, "git", "init"); err != nil {
 		return err
 	}
 	return nil
 }
 
-type Commit struct {
-	Hash   string
-	Title  string
-	Author string
-	CC     []string
-	Date   time.Time
+func (git *git) HeadCommit() (*Commit, error) {
+	return git.GetCommit("HEAD")
 }
 
-// HeadCommit returns info about the HEAD commit of the current branch of git repository in dir.
-func HeadCommit(dir string) (*Commit, error) {
-	return GetCommit(dir, "HEAD")
-}
-
-func GetCommit(dir, commit string) (*Commit, error) {
-	output, err := runSandboxed(dir, "git", "log", "--format=%H%n%s%n%ae%n%ad%n%b", "-n", "1", commit)
+func (git *git) GetCommit(commit string) (*Commit, error) {
+	output, err := runSandboxed(git.dir, "git", "log", "--format=%H%n%s%n%ae%n%ad%n%b", "-n", "1", commit)
 	if err != nil {
 		return nil, err
 	}
-	return parseCommit(output)
+	return gitParseCommit(output)
 }
 
-func parseCommit(output []byte) (*Commit, error) {
+func gitParseCommit(output []byte) (*Commit, error) {
 	lines := bytes.Split(output, []byte{'\n'})
 	if len(lines) < 4 || len(lines[0]) != 40 {
 		return nil, fmt.Errorf("unexpected git log output: %q", output)
 	}
-	date, err := time.Parse(DateFormat, string(lines[3]))
+	const dateFormat = "Mon Jan 2 15:04:05 2006 -0700"
+	date, err := time.Parse(dateFormat, string(lines[3]))
 	if err != nil {
 		return nil, fmt.Errorf("failed to parse date in git log output: %v\n%q", err, output)
 	}
@@ -191,12 +189,11 @@ func parseCommit(output []byte) (*Commit, error) {
 	return com, nil
 }
 
-// ListRecentCommits returns list of recent commit titles starting from baseCommit.
-func ListRecentCommits(dir, baseCommit string) ([]string, error) {
+func (git *git) ListRecentCommits(baseCommit string) ([]string, error) {
 	// On upstream kernel this produces ~11MB of output.
 	// Somewhat inefficient to collect whole output in a slice
 	// and then convert to string, but should be bearable.
-	output, err := runSandboxed(dir, "git", "log",
+	output, err := runSandboxed(git.dir, "git", "log",
 		"--pretty=format:%s", "--no-merges", "-n", "200000", baseCommit)
 	if err != nil {
 		return nil, err
@@ -204,18 +201,10 @@ func ListRecentCommits(dir, baseCommit string) ([]string, error) {
 	return strings.Split(string(output), "\n"), nil
 }
 
-type FixCommit struct {
-	Tag   string
-	Title string
-}
-
-// ExtractFixTagsFromCommits extracts fixing tags for bugs from git log.
-// Given email = "user@domain.com", it searches for tags of the form "user+tag@domain.com"
-// and return pairs {tag, commit title}.
-func ExtractFixTagsFromCommits(dir, baseCommit, email string) ([]FixCommit, error) {
+func (git *git) ExtractFixTagsFromCommits(baseCommit, email string) ([]FixCommit, error) {
 	since := time.Now().Add(-time.Hour * 24 * 365).Format("01-02-2006")
 	cmd := exec.Command("git", "log", "--no-merges", "--since", since, baseCommit)
-	cmd.Dir = dir
+	cmd.Dir = git.dir
 	stdout, err := cmd.StdoutPipe()
 	if err != nil {
 		return nil, err
@@ -225,10 +214,10 @@ func ExtractFixTagsFromCommits(dir, baseCommit, email string) ([]FixCommit, erro
 	}
 	defer cmd.Wait()
 	defer cmd.Process.Kill()
-	return extractFixTags(stdout, email)
+	return gitExtractFixTags(stdout, email)
 }
 
-func extractFixTags(r io.Reader, email string) ([]FixCommit, error) {
+func gitExtractFixTags(r io.Reader, email string) ([]FixCommit, error) {
 	user, domain, err := splitEmail(email)
 	if err != nil {
 		return nil, fmt.Errorf("failed to parse email %q: %v", email, err)
@@ -292,81 +281,11 @@ func splitEmail(email string) (user, domain string, err error) {
 	return
 }
 
-// CanonicalizeCommit returns commit title that can be used when checking
-// if a particular commit is present in a git tree.
-// Some trees add prefixes to commit titles during backporting,
-// so we want e.g. commit "foo bar" match "BACKPORT: foo bar".
-func CanonicalizeCommit(title string) string {
-	for _, prefix := range commitPrefixes {
-		if strings.HasPrefix(title, prefix) {
-			title = title[len(prefix):]
-			break
-		}
-	}
-	return strings.TrimSpace(title)
-}
-
-var commitPrefixes = []string{
-	"UPSTREAM:",
-	"CHROMIUM:",
-	"FROMLIST:",
-	"BACKPORT:",
-	"FROMGIT:",
-	"net-backports:",
-}
-
-func Patch(dir string, patch []byte) error {
-	// Do --dry-run first to not mess with partially consistent state.
-	cmd := osutil.Command("patch", "-p1", "--force", "--ignore-whitespace", "--dry-run")
-	if err := osutil.Sandbox(cmd, true, true); err != nil {
-		return err
-	}
-	cmd.Stdin = bytes.NewReader(patch)
-	cmd.Dir = dir
-	if output, err := cmd.CombinedOutput(); err != nil {
-		// If it reverses clean, then it's already applied
-		// (seems to be the easiest way to detect it).
-		cmd = osutil.Command("patch", "-p1", "--force", "--ignore-whitespace", "--reverse", "--dry-run")
-		if err := osutil.Sandbox(cmd, true, true); err != nil {
-			return err
-		}
-		cmd.Stdin = bytes.NewReader(patch)
-		cmd.Dir = dir
-		if _, err := cmd.CombinedOutput(); err == nil {
-			return fmt.Errorf("patch is already applied")
-		}
-		return fmt.Errorf("failed to apply patch:\n%s", output)
-	}
-	// Now apply for real.
-	cmd = osutil.Command("patch", "-p1", "--force", "--ignore-whitespace")
-	if err := osutil.Sandbox(cmd, true, true); err != nil {
-		return err
-	}
-	cmd.Stdin = bytes.NewReader(patch)
-	cmd.Dir = dir
-	if output, err := cmd.CombinedOutput(); err != nil {
-		return fmt.Errorf("failed to apply patch after dry run:\n%s", output)
-	}
-	return nil
-}
-
-type BisectResult int
-
-const (
-	BisectBad BisectResult = iota
-	BisectGood
-	BisectSkip
-)
-
-// Bisect bisects good..bad commit range against the provided predicate (wrapper around git bisect).
-// The predicate should return an error only if there is no way to proceed
-// (it will abort the process), if possible it should prefer to return BisectSkip.
-// Progress of the process is streamed to the provided trace.
-// Returns the first commit on which the predicate returns BisectBad.
-func Bisect(dir, bad, good string, trace io.Writer, pred func() (BisectResult, error)) (*Commit, error) {
+func (git *git) Bisect(bad, good string, trace io.Writer, pred func() (BisectResult, error)) (*Commit, error) {
+	dir := git.dir
 	runSandboxed(dir, "git", "bisect", "reset")
 	runSandboxed(dir, "git", "reset", "--hard")
-	firstBad, err := GetCommit(dir, bad)
+	firstBad, err := git.GetCommit(bad)
 	if err != nil {
 		return nil, err
 	}
@@ -376,7 +295,7 @@ func Bisect(dir, bad, good string, trace io.Writer, pred func() (BisectResult, e
 	}
 	defer runSandboxed(dir, "git", "bisect", "reset")
 	fmt.Fprintf(trace, "# git bisect start %v %v\n%s", bad, good, output)
-	current, err := HeadCommit(dir)
+	current, err := git.HeadCommit()
 	if err != nil {
 		return nil, err
 	}
@@ -398,7 +317,7 @@ func Bisect(dir, bad, good string, trace io.Writer, pred func() (BisectResult, e
 			return nil, err
 		}
 		fmt.Fprintf(trace, "# git bisect %v %v\n%s", bisectTerms[res], current.Hash, output)
-		next, err := HeadCommit(dir)
+		next, err := git.HeadCommit()
 		if err != nil {
 			return nil, err
 		}
@@ -409,30 +328,29 @@ func Bisect(dir, bad, good string, trace io.Writer, pred func() (BisectResult, e
 	}
 }
 
-// PreviousReleaseTags returns list of preceding release tags that are reachable from the given commit.
 // Note: linux-specific.
-func PreviousReleaseTags(dir, commit string) ([]string, error) {
-	output, err := runSandboxed(dir, "git", "tag", "--no-contains", commit, "--merged", commit, "v*.*")
+func (git *git) PreviousReleaseTags(commit string) ([]string, error) {
+	output, err := runSandboxed(git.dir, "git", "tag", "--no-contains", commit, "--merged", commit, "v*.*")
 	if err != nil {
 		return nil, err
 	}
-	return parseReleaseTags(output)
+	return gitParseReleaseTags(output)
 }
 
-func parseReleaseTags(output []byte) ([]string, error) {
+func gitParseReleaseTags(output []byte) ([]string, error) {
 	var tags []string
 	for _, tag := range bytes.Split(output, []byte{'\n'}) {
-		if releaseTagRe.Match(tag) && releaseTagToInt(string(tag)) != 0 {
+		if releaseTagRe.Match(tag) && gitReleaseTagToInt(string(tag)) != 0 {
 			tags = append(tags, string(tag))
 		}
 	}
 	sort.Slice(tags, func(i, j int) bool {
-		return releaseTagToInt(tags[i]) > releaseTagToInt(tags[j])
+		return gitReleaseTagToInt(tags[i]) > gitReleaseTagToInt(tags[j])
 	})
 	return tags, nil
 }
 
-func releaseTagToInt(tag string) uint64 {
+func gitReleaseTagToInt(tag string) uint64 {
 	matches := releaseTagRe.FindStringSubmatchIndex(tag)
 	v1, err := strconv.ParseUint(tag[matches[2]:matches[3]], 10, 64)
 	if err != nil {
@@ -460,37 +378,3 @@ func runSandboxed(dir, command string, args ...string) ([]byte, error) {
 	}
 	return osutil.Run(timeout, cmd)
 }
-
-// CheckRepoAddress does a best-effort approximate check of a git repo address.
-func CheckRepoAddress(repo string) bool {
-	return gitRepoRe.MatchString(repo)
-}
-
-// CheckBranch does a best-effort approximate check of a git branch name.
-func CheckBranch(branch string) bool {
-	return gitBranchRe.MatchString(branch)
-}
-
-func CheckCommitHash(hash string) bool {
-	if !gitHashRe.MatchString(hash) {
-		return false
-	}
-	ln := len(hash)
-	return ln == 8 || ln == 10 || ln == 12 || ln == 16 || ln == 20 || ln == 40
-}
-
-var (
-	// nolint: lll
-	gitRepoRe    = regexp.MustCompile(`^(git|ssh|http|https|ftp|ftps)://[a-zA-Z0-9-_]+(\.[a-zA-Z0-9-_]+)+(:[0-9]+)?/[a-zA-Z0-9-_./]+\.git(/)?$`)
-	gitBranchRe  = regexp.MustCompile("^[a-zA-Z0-9-_/.]{2,200}$")
-	gitHashRe    = regexp.MustCompile("^[a-f0-9]+$")
-	releaseTagRe = regexp.MustCompile(`^v([0-9]+).([0-9]+)(?:\.([0-9]+))?$`)
-	ccRes        = []*regexp.Regexp{
-		regexp.MustCompile(`^Reviewed\-.*: (.*)$`),
-		regexp.MustCompile(`^[A-Za-z-]+\-and\-[Rr]eviewed\-.*: (.*)$`),
-		regexp.MustCompile(`^Acked\-.*: (.*)$`),
-		regexp.MustCompile(`^[A-Za-z-]+\-and\-[Aa]cked\-.*: (.*)$`),
-		regexp.MustCompile(`^Tested\-.*: (.*)$`),
-		regexp.MustCompile(`^[A-Za-z-]+\-and\-[Tt]ested\-.*: (.*)$`),
-	}
-)
diff --git a/pkg/vcs/git_test.go b/pkg/vcs/git_test.go
index 6b7cf183e..095aca29a 100644
--- a/pkg/vcs/git_test.go
+++ b/pkg/vcs/git_test.go
@@ -10,7 +10,7 @@ import (
 	"time"
 )
 
-func TestParseCommit(t *testing.T) {
+func TestGitParseCommit(t *testing.T) {
 	tests := map[string]*Commit{
 		`2075b16e32c26e4031b9fd3cbe26c54676a8fcb5
 rbtree: include rcu.h
@@ -46,7 +46,7 @@ Signed-off-by: Linux Master <linux@linux-foundation.org>
 		},
 	}
 	for input, com := range tests {
-		res, err := parseCommit([]byte(input))
+		res, err := gitParseCommit([]byte(input))
 		if err != nil && com != nil {
 			t.Fatalf("want %+v, got error: %v", com, err)
 		}
@@ -74,80 +74,7 @@ Signed-off-by: Linux Master <linux@linux-foundation.org>
 	}
 }
 
-func TestCanonicalizeCommit(t *testing.T) {
-	tests := map[string]string{
-		"foo bar":                     "foo bar",
-		" foo ":                       "foo",
-		"UPSTREAM: foo bar":           "foo bar",
-		"BACKPORT: UPSTREAM: foo bar": "UPSTREAM: foo bar",
-	}
-	for in, want := range tests {
-		got := CanonicalizeCommit(in)
-		if got != want {
-			t.Errorf("input %q: got %q, want %q", in, got, want)
-		}
-	}
-}
-
-func TestCheckRepoAddress(t *testing.T) {
-	testPredicate(t, CheckRepoAddress, map[string]bool{
-		"git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git":      true,
-		"https://github.com/torvalds/linux.git":                                 true,
-		"git://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable.git": true,
-		"git://git.cmpxchg.org/linux-mmots.git":                                 true,
-		"https://anonscm.debian.org/git/kernel/linux.git":                       true,
-		"git://kernel.ubuntu.com/ubuntu/ubuntu-zesty.git":                       true,
-		"http://host.xz:123/path/to/repo.git/":                                  true,
-		"":           false,
-		"foobar":     false,
-		"linux-next": false,
-		"foo://kernel.ubuntu.com/ubuntu/ubuntu-zesty.git":    false,
-		"git://kernel/ubuntu.git":                            false,
-		"git://kernel.com/ubuntu":                            false,
-		"gitgit://kernel.ubuntu.com/ubuntu/ubuntu-zesty.git": false,
-	})
-}
-
-func TestCheckBranch(t *testing.T) {
-	testPredicate(t, CheckBranch, map[string]bool{
-		"master":                  true,
-		"core/core":               true,
-		"irq-irqdomain-for-linus": true,
-		"timers/2038":             true,
-		"ubuntu-zesty/v4.9.4":     true,
-		"WIP.locking/atomics":     true,
-		"linux-4.9.y":             true,
-		"abi_spec":                true,
-		"@":                       false,
-		"":                        false,
-	})
-}
-
-func TestCheckCommitHash(t *testing.T) {
-	testPredicate(t, CheckCommitHash, map[string]bool{
-		"ff12bea91c22bba93d3ffc3034d813d686bc7eeb": true, // 40
-		"eae05cb0aaeae05cb0aa":                     true, // 20
-		"449dd6984d0eaabb":                         true, // 16
-		"449dd6984d0e":                             true, // 12
-		"eae05cb0aa":                               true, // 10
-		"eae05cb0":                                 true, // 8
-		"":                                         false,
-		"aa":                                       false,
-		"eae05cb0aab":                              false,
-		"xxxxxxxx":                                 false,
-	})
-}
-
-func testPredicate(t *testing.T, fn func(string) bool, tests map[string]bool) {
-	for input, want := range tests {
-		res := fn(input)
-		if res != want {
-			t.Errorf("%v: got %v, want %v", input, res, want)
-		}
-	}
-}
-
-func TestParseReleaseTags(t *testing.T) {
+func TestGitParseReleaseTags(t *testing.T) {
 	input := `
 v3.1
 v2.6.12
@@ -183,7 +110,7 @@ v1.
 		"v2.6.13",
 		"v2.6.12",
 	}
-	got, err := parseReleaseTags([]byte(input))
+	got, err := gitParseReleaseTags([]byte(input))
 	if err != nil {
 		t.Fatal(err)
 	}
@@ -192,8 +119,8 @@ v1.
 	}
 }
 
-func TestExtractFixTags(t *testing.T) {
-	commits, err := extractFixTags(strings.NewReader(extractFixTagsInput), extractFixTagsEmail)
+func TestGitExtractFixTags(t *testing.T) {
+	commits, err := gitExtractFixTags(strings.NewReader(extractFixTagsInput), extractFixTagsEmail)
 	if err != nil {
 		t.Fatal(err)
 	}
diff --git a/pkg/vcs/vcs.go b/pkg/vcs/vcs.go
new file mode 100644
index 000000000..bef279bdc
--- /dev/null
+++ b/pkg/vcs/vcs.go
@@ -0,0 +1,180 @@
+// Copyright 2018 syzkaller project authors. All rights reserved.
+// Use of this source code is governed by Apache 2 LICENSE that can be found in the LICENSE file.
+
+// Package vcs provides helper functions for working with various repositories (e.g. git).
+package vcs
+
+import (
+	"bytes"
+	"fmt"
+	"io"
+	"regexp"
+	"strings"
+	"time"
+
+	"github.com/google/syzkaller/pkg/osutil"
+)
+
+type Repo interface {
+	// Poll checkouts the specified repository/branch.
+	// This involves fetching/resetting/cloning as necessary to recover from all possible problems.
+	// Returns hash of the HEAD commit in the specified branch.
+	Poll(repo, branch string) (*Commit, error)
+
+	// CheckoutBranch checkouts the specified repository/branch.
+	CheckoutBranch(repo, branch string) (*Commit, error)
+
+	// CheckoutCommit checkouts the specified repository on the specified commit.
+	CheckoutCommit(repo, commit string) (*Commit, error)
+
+	// SwitchCommit checkouts the specified commit without fetching.
+	SwitchCommit(commit string) (*Commit, error)
+
+	// HeadCommit returns info about the HEAD commit of the current branch of git repository.
+	HeadCommit() (*Commit, error)
+
+	// ListRecentCommits returns list of recent commit titles starting from baseCommit.
+	ListRecentCommits(baseCommit string) ([]string, error)
+
+	// ExtractFixTagsFromCommits extracts fixing tags for bugs from git log.
+	// Given email = "user@domain.com", it searches for tags of the form "user+tag@domain.com"
+	// and return pairs {tag, commit title}.
+	ExtractFixTagsFromCommits(baseCommit, email string) ([]FixCommit, error)
+
+	// PreviousReleaseTags returns list of preceding release tags that are reachable from the given commit.
+	PreviousReleaseTags(commit string) ([]string, error)
+
+	// Bisect bisects good..bad commit range against the provided predicate (wrapper around git bisect).
+	// The predicate should return an error only if there is no way to proceed
+	// (it will abort the process), if possible it should prefer to return BisectSkip.
+	// Progress of the process is streamed to the provided trace.
+	// Returns the first commit on which the predicate returns BisectBad.
+	Bisect(bad, good string, trace io.Writer, pred func() (BisectResult, error)) (*Commit, error)
+}
+
+type Commit struct {
+	Hash   string
+	Title  string
+	Author string
+	CC     []string
+	Date   time.Time
+}
+
+type FixCommit struct {
+	Tag   string
+	Title string
+}
+
+type BisectResult int
+
+const (
+	BisectBad BisectResult = iota
+	BisectGood
+	BisectSkip
+)
+
+func NewRepo(os, vm, dir string) (Repo, error) {
+	switch os {
+	case "linux":
+		return newGit(os, vm, dir), nil
+	}
+	return nil, fmt.Errorf("vcs is unsupported for %v", os)
+}
+
+func NewSyzkallerRepo(dir string) Repo {
+	return newGit("syzkaller", "", dir)
+}
+
+func Patch(dir string, patch []byte) error {
+	// Do --dry-run first to not mess with partially consistent state.
+	cmd := osutil.Command("patch", "-p1", "--force", "--ignore-whitespace", "--dry-run")
+	if err := osutil.Sandbox(cmd, true, true); err != nil {
+		return err
+	}
+	cmd.Stdin = bytes.NewReader(patch)
+	cmd.Dir = dir
+	if output, err := cmd.CombinedOutput(); err != nil {
+		// If it reverses clean, then it's already applied
+		// (seems to be the easiest way to detect it).
+		cmd = osutil.Command("patch", "-p1", "--force", "--ignore-whitespace", "--reverse", "--dry-run")
+		if err := osutil.Sandbox(cmd, true, true); err != nil {
+			return err
+		}
+		cmd.Stdin = bytes.NewReader(patch)
+		cmd.Dir = dir
+		if _, err := cmd.CombinedOutput(); err == nil {
+			return fmt.Errorf("patch is already applied")
+		}
+		return fmt.Errorf("failed to apply patch:\n%s", output)
+	}
+	// Now apply for real.
+	cmd = osutil.Command("patch", "-p1", "--force", "--ignore-whitespace")
+	if err := osutil.Sandbox(cmd, true, true); err != nil {
+		return err
+	}
+	cmd.Stdin = bytes.NewReader(patch)
+	cmd.Dir = dir
+	if output, err := cmd.CombinedOutput(); err != nil {
+		return fmt.Errorf("failed to apply patch after dry run:\n%s", output)
+	}
+	return nil
+}
+
+// CheckRepoAddress does a best-effort approximate check of a git repo address.
+func CheckRepoAddress(repo string) bool {
+	return gitRepoRe.MatchString(repo)
+}
+
+// CheckBranch does a best-effort approximate check of a git branch name.
+func CheckBranch(branch string) bool {
+	return gitBranchRe.MatchString(branch)
+}
+
+func CheckCommitHash(hash string) bool {
+	if !gitHashRe.MatchString(hash) {
+		return false
+	}
+	ln := len(hash)
+	return ln == 8 || ln == 10 || ln == 12 || ln == 16 || ln == 20 || ln == 40
+}
+
+const timeout = time.Hour // timeout for all git invocations
+
+var (
+	// nolint: lll
+	gitRepoRe    = regexp.MustCompile(`^(git|ssh|http|https|ftp|ftps)://[a-zA-Z0-9-_]+(\.[a-zA-Z0-9-_]+)+(:[0-9]+)?/[a-zA-Z0-9-_./]+\.git(/)?$`)
+	gitBranchRe  = regexp.MustCompile("^[a-zA-Z0-9-_/.]{2,200}$")
+	gitHashRe    = regexp.MustCompile("^[a-f0-9]+$")
+	releaseTagRe = regexp.MustCompile(`^v([0-9]+).([0-9]+)(?:\.([0-9]+))?$`)
+	ccRes        = []*regexp.Regexp{
+		regexp.MustCompile(`^Reviewed\-.*: (.*)$`),
+		regexp.MustCompile(`^[A-Za-z-]+\-and\-[Rr]eviewed\-.*: (.*)$`),
+		regexp.MustCompile(`^Acked\-.*: (.*)$`),
+		regexp.MustCompile(`^[A-Za-z-]+\-and\-[Aa]cked\-.*: (.*)$`),
+		regexp.MustCompile(`^Tested\-.*: (.*)$`),
+		regexp.MustCompile(`^[A-Za-z-]+\-and\-[Tt]ested\-.*: (.*)$`),
+	}
+)
+
+// CanonicalizeCommit returns commit title that can be used when checking
+// if a particular commit is present in a git tree.
+// Some trees add prefixes to commit titles during backporting,
+// so we want e.g. commit "foo bar" match "BACKPORT: foo bar".
+func CanonicalizeCommit(title string) string {
+	for _, prefix := range commitPrefixes {
+		if strings.HasPrefix(title, prefix) {
+			title = title[len(prefix):]
+			break
+		}
+	}
+	return strings.TrimSpace(title)
+}
+
+var commitPrefixes = []string{
+	"UPSTREAM:",
+	"CHROMIUM:",
+	"FROMLIST:",
+	"BACKPORT:",
+	"FROMGIT:",
+	"net-backports:",
+}
diff --git a/pkg/vcs/vcs_test.go b/pkg/vcs/vcs_test.go
new file mode 100644
index 000000000..2c0660849
--- /dev/null
+++ b/pkg/vcs/vcs_test.go
@@ -0,0 +1,81 @@
+// Copyright 2018 syzkaller project authors. All rights reserved.
+// Use of this source code is governed by Apache 2 LICENSE that can be found in the LICENSE file.
+
+package vcs
+
+import (
+	"testing"
+)
+
+func TestCanonicalizeCommit(t *testing.T) {
+	tests := map[string]string{
+		"foo bar":                     "foo bar",
+		" foo ":                       "foo",
+		"UPSTREAM: foo bar":           "foo bar",
+		"BACKPORT: UPSTREAM: foo bar": "UPSTREAM: foo bar",
+	}
+	for in, want := range tests {
+		got := CanonicalizeCommit(in)
+		if got != want {
+			t.Errorf("input %q: got %q, want %q", in, got, want)
+		}
+	}
+}
+
+func TestCheckRepoAddress(t *testing.T) {
+	testPredicate(t, CheckRepoAddress, map[string]bool{
+		"git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git":      true,
+		"https://github.com/torvalds/linux.git":                                 true,
+		"git://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable.git": true,
+		"git://git.cmpxchg.org/linux-mmots.git":                                 true,
+		"https://anonscm.debian.org/git/kernel/linux.git":                       true,
+		"git://kernel.ubuntu.com/ubuntu/ubuntu-zesty.git":                       true,
+		"http://host.xz:123/path/to/repo.git/":                                  true,
+		"":           false,
+		"foobar":     false,
+		"linux-next": false,
+		"foo://kernel.ubuntu.com/ubuntu/ubuntu-zesty.git":    false,
+		"git://kernel/ubuntu.git":                            false,
+		"git://kernel.com/ubuntu":                            false,
+		"gitgit://kernel.ubuntu.com/ubuntu/ubuntu-zesty.git": false,
+	})
+}
+
+func TestCheckBranch(t *testing.T) {
+	testPredicate(t, CheckBranch, map[string]bool{
+		"master":                  true,
+		"core/core":               true,
+		"irq-irqdomain-for-linus": true,
+		"timers/2038":             true,
+		"ubuntu-zesty/v4.9.4":     true,
+		"WIP.locking/atomics":     true,
+		"linux-4.9.y":             true,
+		"abi_spec":                true,
+		"@":                       false,
+		"":                        false,
+	})
+}
+
+func TestCheckCommitHash(t *testing.T) {
+	testPredicate(t, CheckCommitHash, map[string]bool{
+		"ff12bea91c22bba93d3ffc3034d813d686bc7eeb": true, // 40
+		"eae05cb0aaeae05cb0aa":                     true, // 20
+		"449dd6984d0eaabb":                         true, // 16
+		"449dd6984d0e":                             true, // 12
+		"eae05cb0aa":                               true, // 10
+		"eae05cb0":                                 true, // 8
+		"":                                         false,
+		"aa":                                       false,
+		"eae05cb0aab":                              false,
+		"xxxxxxxx":                                 false,
+	})
+}
+
+func testPredicate(t *testing.T, fn func(string) bool, tests map[string]bool) {
+	for input, want := range tests {
+		res := fn(input)
+		if res != want {
+			t.Errorf("%v: got %v, want %v", input, res, want)
+		}
+	}
+}