blob: a2d32829c321411eb26ca1ce5d8d3eb9bd779aa0 [file] [log] [blame]
// Copyright 2024 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
package openvex
import (
"crypto/sha256"
"encoding/json"
"fmt"
"io"
"slices"
"time"
"golang.org/x/vuln/internal/govulncheck"
"golang.org/x/vuln/internal/osv"
)
type findingLevel int
const (
invalid findingLevel = iota
required
imported
called
)
type handler struct {
w io.Writer
cfg *govulncheck.Config
sbom *govulncheck.SBOM
osvs map[string]*osv.Entry
// findings contains same-level findings for an
// OSV at the most precise level of granularity
// available. This means, for instance, that if
// an osv is indeed called, then all findings for
// the osv will have call stack info.
findings map[string][]*govulncheck.Finding
}
func NewHandler(w io.Writer) *handler {
return &handler{
w: w,
osvs: make(map[string]*osv.Entry),
findings: make(map[string][]*govulncheck.Finding),
}
}
func (h *handler) Config(cfg *govulncheck.Config) error {
h.cfg = cfg
return nil
}
func (h *handler) Progress(progress *govulncheck.Progress) error {
return nil
}
func (h *handler) SBOM(s *govulncheck.SBOM) error {
h.sbom = s
return nil
}
func (h *handler) OSV(e *osv.Entry) error {
h.osvs[e.ID] = e
return nil
}
// foundAtLevel returns the level at which a specific finding is present in the
// scanned product.
func foundAtLevel(f *govulncheck.Finding) findingLevel {
frame := f.Trace[0]
if frame.Function != "" {
return called
}
if frame.Package != "" {
return imported
}
return required
}
// moreSpecific favors a call finding over a non-call
// finding and a package finding over a module finding.
func moreSpecific(f1, f2 *govulncheck.Finding) int {
if len(f1.Trace) > 1 && len(f2.Trace) > 1 {
// Both are call stack findings.
return 0
}
if len(f1.Trace) > 1 {
return -1
}
if len(f2.Trace) > 1 {
return 1
}
fr1, fr2 := f1.Trace[0], f2.Trace[0]
if fr1.Function != "" && fr2.Function == "" {
return -1
}
if fr1.Function == "" && fr2.Function != "" {
return 1
}
if fr1.Package != "" && fr2.Package == "" {
return -1
}
if fr1.Package == "" && fr2.Package != "" {
return -1
}
return 0 // findings always have module info
}
func (h *handler) Finding(f *govulncheck.Finding) error {
fs := h.findings[f.OSV]
if len(fs) == 0 {
fs = []*govulncheck.Finding{f}
} else {
if ms := moreSpecific(f, fs[0]); ms == -1 {
// The new finding is more specific, so we need
// to erase existing findings and add the new one.
fs = []*govulncheck.Finding{f}
} else if ms == 0 {
// The new finding is at the same level of precision.
fs = append(fs, f)
}
// Otherwise, the new finding is at a less precise level.
}
h.findings[f.OSV] = fs
return nil
}
// Flush is used to print the vex json to w.
// This is needed as vex is not streamed.
func (h *handler) Flush() error {
doc := toVex(h)
out, err := json.MarshalIndent(doc, "", " ")
if err != nil {
return err
}
_, err = h.w.Write(out)
return err
}
func toVex(h *handler) Document {
doc := Document{
Context: ContextURI,
Author: DefaultAuthor,
Timestamp: time.Now().UTC(),
Version: 1,
Tooling: Tooling,
Statements: statements(h),
}
id := hashVex(doc)
doc.ID = "govulncheck/vex:" + id
return doc
}
// Given a slice of findings, returns those findings as a set of subcomponents
// that are unique per the vulnerable artifact's PURL.
func subcomponentSet(findings []*govulncheck.Finding) []Component {
var scs []Component
seen := make(map[string]bool)
for _, f := range findings {
purl := purlFromFinding(f)
if !seen[purl] {
scs = append(scs, Component{
ID: purlFromFinding(f),
})
seen[purl] = true
}
}
return scs
}
// statements combines all OSVs found by govulncheck and generates the list of
// vex statements with the proper affected level and justification to match the
// openVex specification.
func statements(h *handler) []Statement {
var scanLevel findingLevel
switch h.cfg.ScanLevel {
case govulncheck.ScanLevelModule:
scanLevel = required
case govulncheck.ScanLevelPackage:
scanLevel = imported
case govulncheck.ScanLevelSymbol:
scanLevel = called
}
var statements []Statement
for id, osv := range h.osvs {
// if there are no findings emitted for a given OSV that means that
// the vulnerable module is not required at a vulnerable version.
if len(h.findings[id]) == 0 {
continue
}
description := osv.Summary
if description == "" {
description = osv.Details
}
s := Statement{
Vulnerability: Vulnerability{
ID: fmt.Sprintf("https://pkg.go.dev/vuln/%s", id),
Name: id,
Description: description,
Aliases: osv.Aliases,
},
Products: []Product{
{
Component: Component{ID: DefaultPID},
Subcomponents: subcomponentSet(h.findings[id]),
},
},
}
// Findings are guaranteed to be at the same level, so we can just check the first element
fLevel := foundAtLevel(h.findings[id][0])
if fLevel >= scanLevel {
s.Status = StatusAffected
} else {
s.Status = StatusNotAffected
s.ImpactStatement = Impact
s.Justification = JustificationNotPresent
// We only reach this case if running in symbol mode
if fLevel == imported {
s.Justification = JustificationNotExecuted
}
}
statements = append(statements, s)
}
slices.SortFunc(statements, func(a, b Statement) int {
if a.Vulnerability.ID > b.Vulnerability.ID {
return 1
}
if a.Vulnerability.ID < b.Vulnerability.ID {
return -1
}
// this should never happen in practice, since statements are being
// populated from a map with the vulnerability IDs as keys
return 0
})
return statements
}
func hashVex(doc Document) string {
// json.Marshal should never error here (because of the structure of Document).
// If an error does occur, it won't be a jsonerror, but instead a panic
d := Document{
Context: doc.Context,
ID: doc.ID,
Author: doc.Author,
Version: doc.Version,
Tooling: doc.Tooling,
Statements: doc.Statements,
}
out, err := json.Marshal(d)
if err != nil {
panic(err)
}
return fmt.Sprintf("%x", sha256.Sum256(out))
}