Use reader when scanning for package versions over reading entire binary into memory (#3558)

* use streaming readers

Signed-off-by: Alex Goodman <wagoodman@users.noreply.github.com>

* replace redis search patterns

Signed-off-by: Alex Goodman <wagoodman@users.noreply.github.com>

* address PR feedback

Signed-off-by: Alex Goodman <wagoodman@users.noreply.github.com>

---------

Signed-off-by: Alex Goodman <wagoodman@users.noreply.github.com>

internal/regex_helpers.go

@@ -1,6 +1,11 @@
 package internal
 
-import "regexp"
+import (
+	"io"
+	"regexp"
+)
+
+const readerChunkSize = 1024 * 1024
 
 // MatchNamedCaptureGroups takes a regular expression and string and returns all of the named capture group results in a map.
 // This is only for the first match in the regex. Callers shouldn't be providing regexes with multiple capture groups with the same name.
@@ -32,6 +37,89 @@ func MatchNamedCaptureGroups(regEx *regexp.Regexp, content string) map[string]st
 	return results
 }
 
+// MatchNamedCaptureGroupsFromReader matches named capture groups from a reader, assuming the pattern fits within
+// 1.5x the reader chunk size (1MB * 1.5).
+func MatchNamedCaptureGroupsFromReader(re *regexp.Regexp, r io.Reader) (map[string]string, error) {
+	results := make(map[string]string)
+	_, err := processReaderInChunks(r, readerChunkSize, matchNamedCaptureGroupsHandler(re, results))
+	if err != nil {
+		return nil, err
+	}
+	if len(results) == 0 {
+		return nil, nil
+	}
+	return results, nil
+}
+
+// MatchAnyFromReader matches any of the provided regular expressions from a reader, assuming the pattern fits within
+// 1.5x the reader chunk size (1MB * 1.5).
+func MatchAnyFromReader(r io.Reader, res ...*regexp.Regexp) (bool, error) {
+	return processReaderInChunks(r, readerChunkSize, matchAnyHandler(res))
+}
+
+func matchNamedCaptureGroupsHandler(re *regexp.Regexp, results map[string]string) func(data []byte) (bool, error) {
+	return func(data []byte) (bool, error) {
+		if match := re.FindSubmatch(data); match != nil {
+			groupNames := re.SubexpNames()
+			for i, name := range groupNames {
+				if i > 0 && name != "" {
+					results[name] = string(match[i])
+				}
+			}
+			return true, nil
+		}
+		return false, nil
+	}
+}
+
+func matchAnyHandler(res []*regexp.Regexp) func(data []byte) (bool, error) {
+	return func(data []byte) (bool, error) {
+		for _, re := range res {
+			if re.Match(data) {
+				return true, nil
+			}
+		}
+		return false, nil
+	}
+}
+
+// processReaderInChunks reads from the provided reader in chunks and calls the provided handler with each chunk + portion of the previous neighboring chunk.
+// Note that we only overlap the last half of the previous chunk with the current chunk to avoid missing matches that span chunk boundaries.
+func processReaderInChunks(rdr io.Reader, chunkSize int, handler func(data []byte) (bool, error)) (bool, error) {
+	half := chunkSize / 2
+	bufSize := chunkSize + half
+	buf := make([]byte, bufSize)
+	lastRead := 0
+
+	for {
+		offset := half
+		if lastRead < half {
+			offset = lastRead
+		}
+		start := half - offset
+		if lastRead > 0 {
+			copy(buf[start:], buf[half+offset:half+lastRead])
+		}
+		n, err := rdr.Read(buf[half:])
+		if err != nil {
+			break
+		}
+
+		// process the combined data with the handler
+		matched, handlerErr := handler(buf[start : half+n])
+		if handlerErr != nil {
+			return false, handlerErr
+		}
+		if matched {
+			return true, nil
+		}
+
+		lastRead = n
+	}
+
+	return false, nil
+}
+
 func isEmptyMap(m map[string]string) bool {
 	if len(m) == 0 {
 		return true

internal/regex_helpers_test.go

@@ -2,9 +2,11 @@ package internal
 
 import (
 	"regexp"
+	"strings"
 	"testing"
 
 	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
 )
 
 func TestMatchCaptureGroups(t *testing.T) {
@@ -68,3 +70,149 @@ func TestMatchCaptureGroups(t *testing.T) {
 		})
 	}
 }
+
+func TestMatchNamedCaptureGroupsFromReader(t *testing.T) {
+	tests := []struct {
+		name    string
+		pattern string
+		input   string
+		want    map[string]string
+		wantErr require.ErrorAssertionFunc
+	}{
+		{
+			name:    "match single group",
+			pattern: `(?P<key>[^1-9]+)`,
+			input:   "key",
+			want:    map[string]string{"key": "key"},
+			wantErr: require.NoError,
+		},
+		{
+			name:    "match multiple groups",
+			pattern: `(?P<key>[^1-9]+):(?P<value>\w+)`,
+			input:   "key:value",
+			want:    map[string]string{"key": "key", "value": "value"},
+			wantErr: require.NoError,
+		},
+		{
+			name:    "no match",
+			pattern: `(?P<key>[^1-9]+)`,
+			input:   "2345",
+			want:    nil,
+			wantErr: require.NoError,
+		},
+		{
+			name:    "error empty reader",
+			pattern: `(?P<key>\w+)`,
+			input:   "",
+			want:    nil,
+			wantErr: require.NoError,
+		},
+	}
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			re := regexp.MustCompile(tt.pattern)
+			r := strings.NewReader(tt.input)
+			got, err := MatchNamedCaptureGroupsFromReader(re, r)
+			tt.wantErr(t, err)
+			assert.Equal(t, tt.want, got)
+		})
+	}
+}
+
+func TestMatchAnyFromReader(t *testing.T) {
+	tests := []struct {
+		name     string
+		input    string
+		patterns []*regexp.Regexp
+		want     bool
+		wantErr  require.ErrorAssertionFunc
+	}{
+		{
+			name:     "match single pattern",
+			input:    "hello world",
+			patterns: []*regexp.Regexp{regexp.MustCompile(`hello`)},
+			want:     true,
+			wantErr:  require.NoError,
+		},
+		{
+			name:     "match multiple patterns",
+			input:    "test case",
+			patterns: []*regexp.Regexp{regexp.MustCompile(`case`), regexp.MustCompile(`test`)},
+			want:     true,
+			wantErr:  require.NoError,
+		},
+		{
+			name:     "no match",
+			input:    "nothing here",
+			patterns: []*regexp.Regexp{regexp.MustCompile(`absent`)},
+			want:     false,
+			wantErr:  require.NoError,
+		},
+		{
+			name:     "error empty reader",
+			input:    "",
+			patterns: []*regexp.Regexp{regexp.MustCompile(`match`)},
+			want:     false,
+			wantErr:  require.NoError,
+		},
+	}
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			r := strings.NewReader(tt.input)
+			got, err := MatchAnyFromReader(r, tt.patterns...)
+			tt.wantErr(t, err)
+			assert.Equal(t, tt.want, got)
+		})
+	}
+}
+
+func TestProcessReaderInChunks_ChunkBoundaries(t *testing.T) {
+	tests := []struct {
+		name          string
+		input         string
+		chunkSize     int
+		expectedCalls []string
+		returnOnChunk int
+		wantErr       require.ErrorAssertionFunc
+	}{
+		{
+			name:          "go case",
+			input:         "123456789012345",
+			chunkSize:     4,
+			returnOnChunk: 2,
+			expectedCalls: []string{"1234", "345678", "789012"},
+			wantErr:       require.NoError,
+		},
+		{
+			name:          "no match",
+			input:         "123456789012345",
+			chunkSize:     4,
+			returnOnChunk: -1,
+			expectedCalls: []string{"1234", "345678", "789012", "12345"},
+			wantErr:       require.NoError,
+		},
+	}
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			var actualCalls []string
+			var current int
+			handler := func(data []byte) (bool, error) {
+				actualCalls = append(actualCalls, string(data))
+				if current == tt.returnOnChunk {
+					return true, nil
+				}
+				current++
+				return false, nil
+			}
+			r := strings.NewReader(tt.input)
+			got, err := processReaderInChunks(r, tt.chunkSize, handler)
+			tt.wantErr(t, err)
+			if tt.returnOnChunk == -1 {
+				assert.False(t, got)
+			} else {
+				assert.True(t, got)
+			}
+			assert.Equal(t, tt.expectedCalls, actualCalls)
+		})
+	}
+}

syft/pkg/cataloger/binary/classifier.go

@@ -18,6 +18,7 @@ import (
 	"github.com/anchore/syft/internal/log"
 	"github.com/anchore/syft/syft/cpe"
 	"github.com/anchore/syft/syft/file"
+	"github.com/anchore/syft/syft/internal/unionreader"
 	"github.com/anchore/syft/syft/pkg"
 )
 
@@ -74,9 +75,9 @@ func (cfg Classifier) MarshalJSON() ([]byte, error) {
 type EvidenceMatcher func(classifier Classifier, context matcherContext) ([]pkg.Package, error)
 
 type matcherContext struct {
-	resolver    file.Resolver
-	location    file.Location
-	getContents func(resolver matcherContext) ([]byte, error)
+	resolver  file.Resolver
+	location  file.Location
+	getReader func(resolver matcherContext) (unionreader.UnionReader, error)
 }
 
 func evidenceMatchers(matchers ...EvidenceMatcher) EvidenceMatcher {
@@ -124,12 +125,15 @@ func fileNameTemplateVersionMatcher(fileNamePattern string, contentTemplate stri
 			return nil, fmt.Errorf("unable to compile rendered regex=%q: %w", patternBuf.String(), err)
 		}
 
-		contents, err := getContents(context)
+		contents, err := getReader(context)
 		if err != nil {
 			return nil, fmt.Errorf("unable to get read contents for file: %w", err)
 		}
 
-		matchMetadata := internal.MatchNamedCaptureGroups(tmplPattern, string(contents))
+		matchMetadata, err := internal.MatchNamedCaptureGroupsFromReader(tmplPattern, contents)
+		if err != nil {
+			return nil, fmt.Errorf("unable to match version: %w", err)
+		}
 
 		p := newClassifierPackage(classifier, context.location, matchMetadata)
 		if p == nil {
@@ -143,12 +147,15 @@ func fileNameTemplateVersionMatcher(fileNamePattern string, contentTemplate stri
 func FileContentsVersionMatcher(pattern string) EvidenceMatcher {
 	pat := regexp.MustCompile(pattern)
 	return func(classifier Classifier, context matcherContext) ([]pkg.Package, error) {
-		contents, err := getContents(context)
+		contents, err := getReader(context)
 		if err != nil {
 			return nil, fmt.Errorf("unable to get read contents for file: %w", err)
 		}
 
-		matchMetadata := internal.MatchNamedCaptureGroups(pat, string(contents))
+		matchMetadata, err := internal.MatchNamedCaptureGroupsFromReader(pat, contents)
+		if err != nil {
+			return nil, fmt.Errorf("unable to match version: %w", err)
+		}
 
 		// Convert {major: 1, minor: 2, patch: 3} to "1.2.3"
 		_, versionOk := matchMetadata["version"]
@@ -183,14 +190,16 @@ func matchExcluding(matcher EvidenceMatcher, contentPatternsToExclude ...string)
 		nonMatchPatterns = append(nonMatchPatterns, regexp.MustCompile(p))
 	}
 	return func(classifier Classifier, context matcherContext) ([]pkg.Package, error) {
-		contents, err := getContents(context)
+		contents, err := getReader(context)
 		if err != nil {
 			return nil, fmt.Errorf("unable to get read contents for file: %w", err)
 		}
-		for _, nonMatch := range nonMatchPatterns {
-			if nonMatch.Match(contents) {
-				return nil, nil
-			}
+		matches, err := internal.MatchAnyFromReader(contents, nonMatchPatterns...)
+		if err != nil {
+			return nil, fmt.Errorf("unable to match content: %w", err)
+		}
+		if matches {
+			return nil, nil
 		}
 		return matcher(classifier, context)
 	}
@@ -214,9 +223,9 @@ func sharedLibraryLookup(sharedLibraryPattern string, sharedLibraryMatcher Evide
 			}
 			for _, libraryLocation := range locations {
 				newResolver := matcherContext{
-					resolver:    context.resolver,
-					location:    libraryLocation,
-					getContents: context.getContents,
+					resolver:  context.resolver,
+					location:  libraryLocation,
+					getReader: context.getReader,
 				}
 				newResolver.location = libraryLocation
 				pkgs, err := sharedLibraryMatcher(classifier, newResolver)
@@ -253,23 +262,16 @@ func mustPURL(purl string) packageurl.PackageURL {
 	return p
 }
 
-func getContents(context matcherContext) ([]byte, error) {
-	if context.getContents != nil {
-		return context.getContents(context)
+func getReader(context matcherContext) (unionreader.UnionReader, error) {
+	if context.getReader != nil {
+		return context.getReader(context)
 	}
-	reader, err := context.resolver.FileContentsByLocation(context.location)
+	reader, err := context.resolver.FileContentsByLocation(context.location) //nolint:gocritic
 	if err != nil {
 		return nil, err
 	}
-	defer internal.CloseAndLogError(reader, context.location.AccessPath)
 
-	// TODO: there may be room for improvement here, as this may use an excessive amount of memory. Alternate approach is to leverage a RuneReader.
-	contents, err := io.ReadAll(reader)
-	if err != nil {
-		return nil, fmt.Errorf("unable to get contents for file: %w", err)
-	}
-
-	return contents, nil
+	return unionreader.GetUnionReader(reader)
 }
 
 // singleCPE returns a []cpe.CPE with Source: Generated based on the cpe string or panics if the
@@ -287,14 +289,13 @@ func singleCPE(cpeString string, source ...cpe.Source) []cpe.CPE {
 // sharedLibraries returns a list of all shared libraries found within a binary, currently
 // supporting: elf, macho, and windows pe
 func sharedLibraries(context matcherContext) ([]string, error) {
-	contents, err := getContents(context)
+	contents, err := getReader(context)
 	if err != nil {
 		return nil, err
 	}
+	defer internal.CloseAndLogError(contents, context.location.RealPath)
 
-	r := bytes.NewReader(contents)
-
-	e, _ := elf.NewFile(r)
+	e, _ := elf.NewFile(contents)
 	if e != nil {
 		symbols, err := e.ImportedLibraries()
 		if err != nil {
@@ -302,8 +303,11 @@ func sharedLibraries(context matcherContext) ([]string, error) {
 		}
 		return symbols, nil
 	}
+	if _, err := contents.Seek(0, io.SeekStart); err != nil {
+		return nil, fmt.Errorf("unable to seek to beginning of file: %w", err)
+	}
 
-	m, _ := macho.NewFile(r)
+	m, _ := macho.NewFile(contents)
 	if m != nil {
 		symbols, err := m.ImportedLibraries()
 		if err != nil {
@@ -311,8 +315,11 @@ func sharedLibraries(context matcherContext) ([]string, error) {
 		}
 		return symbols, nil
 	}
+	if _, err := contents.Seek(0, io.SeekStart); err != nil {
+		return nil, fmt.Errorf("unable to seek to beginning of file: %w", err)
+	}
 
-	p, _ := pe.NewFile(r)
+	p, _ := pe.NewFile(contents)
 	if p != nil {
 		symbols, err := p.ImportedLibraries()
 		if err != nil {
@@ -320,6 +327,9 @@ func sharedLibraries(context matcherContext) ([]string, error) {
 		}
 		return symbols, nil
 	}
+	if _, err := contents.Seek(0, io.SeekStart); err != nil {
+		return nil, fmt.Errorf("unable to seek to beginning of file: %w", err)
+	}
 
 	return nil, nil
 }

syft/pkg/cataloger/binary/classifier_test.go

@@ -1,6 +1,8 @@
 package binary
 
 import (
+	"bytes"
+	"io"
 	"testing"
 
 	"github.com/stretchr/testify/assert"
@@ -9,6 +11,7 @@ import (
 	"github.com/anchore/packageurl-go"
 	"github.com/anchore/syft/syft/cpe"
 	"github.com/anchore/syft/syft/file"
+	"github.com/anchore/syft/syft/internal/unionreader"
 )
 
 func Test_ClassifierCPEs(t *testing.T) {
@@ -162,12 +165,12 @@ func TestFileContentsVersionMatcher(t *testing.T) {
 	}
 	for _, tt := range tests {
 		t.Run(tt.name, func(t *testing.T) {
-			mockGetContent := func(context matcherContext) ([]byte, error) {
-				return []byte(tt.data), nil
+			mockGetContent := func(context matcherContext) (unionreader.UnionReader, error) {
+				return unionreader.GetUnionReader(io.NopCloser(bytes.NewBufferString(tt.data)))
 			}
 			fn := FileContentsVersionMatcher(tt.pattern)
 			p, err := fn(Classifier{}, matcherContext{
-				getContents: mockGetContent,
+				getReader: mockGetContent,
 			})
 
 			if err != nil {

syft/pkg/cataloger/binary/classifiers.go

@@ -77,8 +77,12 @@ func DefaultClassifiers() []Classifier {
 			Class:    "redis-binary",
 			FileGlob: "**/redis-server",
 			EvidenceMatcher: evidenceMatchers(
-				FileContentsVersionMatcher(`(?s)payload %5.*?(?P<version>\d.\d\.\d\d*)[a-z0-9]{12,15}-[0-9]{19}`),
-				FileContentsVersionMatcher(`(?s)\x00(?P<version>\d.\d\.\d\d*)[a-z0-9]{12,15}-[0-9]{19}\x00.*?payload %5`),
+				// matches most recent versions of redis (~v7), e.g. "7.0.14buildkitsandbox-1702957741000000000"
+				FileContentsVersionMatcher(`[^\d](?P<version>\d+.\d+\.\d+)buildkitsandbox-\d+`),
+				// matches against older versions of redis (~v3 - v6), e.g. "4.0.11841ce7054bd9-1542359302000000000"
+				FileContentsVersionMatcher(`[^\d](?P<version>[0-9]+\.[0-9]+\.[0-9]+)\w{12}-\d+`),
+				// matches against older versions of redis (~v2), e.g. "Server started, Redis version 2.8.23"
+				FileContentsVersionMatcher(`Redis version (?P<version>[0-9]+\.[0-9]+\.[0-9]+)`),
 			),
 			Package: "redis",
 			PURL:    mustPURL("pkg:generic/redis@version"),

syft/pkg/cataloger/golang/parse_go_binary.go

@@ -262,12 +262,11 @@ func (c *goBinaryCataloger) findMainModuleVersion(metadata *pkg.GolangBinaryBuil
 }
 
 func extractVersionFromContents(reader io.Reader) string {
-	contents, err := io.ReadAll(reader)
+	matchMetadata, err := internal.MatchNamedCaptureGroupsFromReader(semverPattern, reader)
 	if err != nil {
-		log.WithFields("error", err).Trace("unable to read from go binary reader")
+		log.WithFields("error", err).Trace("unable to extract version from go binary reader")
 		return ""
 	}
-	matchMetadata := internal.MatchNamedCaptureGroups(semverPattern, string(contents))
 
 	version, ok := matchMetadata["version"]
 	if ok {