Add libFuzzer harnesses for the HTTP header and cache-index parsers (#508)

P3-5 fuzz Tier-2. Two more harnesses over hostile-input parsers that read
structured bytes into fixed buffers: fuzz-header drives treatfirstline plus
treathead on each response-header line (the Content-Type/-Encoding path hardened
in #506, and the cookie/Location/Content-Range fields); fuzz-cachendx drives the
hts-cache/*.ndx length-prefixed scan that cache_readex_new loads on --update.

fuzz-cachendx found the over-advance fixed in the parent commit; its seed corpus
carries the two crash reproducers as replay regressions. The cache harness stops
at the scan rather than the trailing coucal insert, whose murmur hash trips a
separate pointer-overflow the .ndx parser does not own.

Signed-off-by: Xavier Roche <roche@httrack.com>
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>

.gitignore

@@ -34,8 +34,9 @@ Makefile
 *.so.*
 *.a
 
-# make dist output.
+# make dist output; dist/ holds httrack-gh-release staging artifacts.
 /httrack-*.tar.gz
+/dist/
 
 # Editor / autotools backup files.
 *~

fuzz/Makefile.am

@@ -1,7 +1,7 @@
 # libFuzzer harnesses; built only with --enable-fuzzers (requires clang).
 if FUZZERS
 noinst_PROGRAMS = fuzz-charset fuzz-meta fuzz-idna fuzz-entities \
-	fuzz-unescape fuzz-filters fuzz-url
+	fuzz-unescape fuzz-filters fuzz-url fuzz-header fuzz-cachendx
 endif
 
 AM_CPPFLAGS = \
@@ -23,6 +23,8 @@ fuzz_entities_SOURCES = fuzz-entities.c fuzz.h
 fuzz_unescape_SOURCES = fuzz-unescape.c fuzz.h
 fuzz_filters_SOURCES = fuzz-filters.c fuzz.h
 fuzz_url_SOURCES = fuzz-url.c fuzz.h
+fuzz_header_SOURCES = fuzz-header.c fuzz.h
+fuzz_cachendx_SOURCES = fuzz-cachendx.c fuzz.h
 
 # List corpus files explicitly: automake does not expand EXTRA_DIST globs.
 EXTRA_DIST = README.md run-fuzzers.sh \
@@ -35,4 +37,8 @@ EXTRA_DIST = README.md run-fuzzers.sh \
 	corpus/filters/filter.bin corpus/filters/filter-size.bin \
 	corpus/filters/regress-empty-subject-unique.bin \
 	corpus/url/http-url.txt corpus/url/relative-path.txt \
-	corpus/url/regress-file-empty-path.txt corpus/url/regress-long-path-abort.txt
+	corpus/url/regress-file-empty-path.txt corpus/url/regress-long-path-abort.txt \
+	corpus/header/full-response.txt corpus/header/redirect.txt \
+	corpus/cachendx/new-format.txt corpus/cachendx/old-format.txt \
+	corpus/cachendx/regress-overadvance.bin \
+	corpus/cachendx/regress-truncated-entry.bin

fuzz/corpus/cachendx/new-format.txt

@@ -0,0 +1,7 @@
+8
+CACHE-1.1
+28
+Mon, 01 Jan 2024 00:00:00 GMT
+www.example.com
+/index.html
+123

fuzz/corpus/cachendx/old-format.txt

@@ -0,0 +1,5 @@
+3
+1.0
+www.example.com
+/page
+5

fuzz/corpus/cachendx/regress-overadvance.bin

@@ -0,0 +1,2 @@
+32768
+CACHE-1.1

fuzz/corpus/cachendx/regress-truncated-entry.bin

@@ -0,0 +1,5 @@
+8
+CACHE-1.1
+1
+x
+www.example.com

fuzz/corpus/header/full-response.txt

@@ -0,0 +1,10 @@
+HTTP/1.1 200 OK
+Content-Type: text/html; charset=utf-8
+Content-Length: 1234
+Content-Encoding: gzip
+Last-Modified: Mon, 01 Jan 2024 00:00:00 GMT
+Etag: "abc"
+Location: http://example.com/x
+Set-Cookie: ID=42; path=/; domain=.example.com
+Content-Range: bytes 0-99/100
+Transfer-Encoding: chunked

fuzz/corpus/header/redirect.txt

@@ -0,0 +1,3 @@
+HTTP/1.0 301 Moved
+Location: /elsewhere
+Content-Disposition: attachment; filename="a.pdf"

fuzz/fuzz-cachendx.c

@@ -0,0 +1,66 @@
+/* ------------------------------------------------------------ */
+/*
+HTTrack Website Copier, Offline Browser for Windows and Unix
+Copyright (C) 1998 Xavier Roche and other contributors
+
+SPDX-License-Identifier: GPL-3.0-or-later
+
+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
+the Free Software Foundation, either version 3 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License
+along with this program. If not, see <http://www.gnu.org/licenses/>.
+
+Ethical use: we kindly ask that you NOT use this software to harvest email
+addresses or to collect any other private information about people. Doing so
+would dishonor our work and waste the many hours we have spent on it.
+
+Please visit our Website: http://www.httrack.com
+*/
+
+/* Fuzz the cache-index (.ndx) parser: a corrupt or truncated index must not
+   walk the length-prefixed cache_brstr/cache_binput scan past the buffer.
+   Mirrors the loader in cache_readex_new (htscache.c). */
+#include "fuzz.h"
+#include "htscache.h"
+#include "htslib.h"
+
+int LLVMFuzzerTestOneInput(const uint8_t *data, size_t size) {
+  char *buf = fuzz_strdup(data, size);
+  const char *const end = buf + size;
+  char firstline[256];
+  char *a = buf;
+
+  /* header: two length-prefixed fields (version, last-modified) */
+  a += cache_brstr(a, firstline, sizeof(firstline));
+  a += cache_brstr(a, firstline, sizeof(firstline));
+
+  /* body: newline-delimited host/file/position triples. The coucal insert
+     the real loader ends with is a separate library's concern; this harness
+     targets the length-prefixed scan that must stay inside the buffer. */
+  while (a != NULL && a < end) {
+    char BIGSTK line[HTS_URLMAXSIZE * 2];
+    char linepos[256];
+    int pos;
+
+    a = strchr(a + 1, '\n');
+    if (a == NULL)
+      break;
+    a++;
+    a += cache_binput(a, end, line, HTS_URLMAXSIZE);
+    a += cache_binput(a, end, line + strlen(line), HTS_URLMAXSIZE);
+    a += cache_binput(a, end, linepos, 200);
+    sscanf(linepos, "%d", &pos);
+    (void) pos;
+  }
+
+  freet(buf);
+  return 0;
+}

fuzz/fuzz-header.c

@@ -0,0 +1,70 @@
+/* ------------------------------------------------------------ */
+/*
+HTTrack Website Copier, Offline Browser for Windows and Unix
+Copyright (C) 1998 Xavier Roche and other contributors
+
+SPDX-License-Identifier: GPL-3.0-or-later
+
+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
+the Free Software Foundation, either version 3 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License
+along with this program. If not, see <http://www.gnu.org/licenses/>.
+
+Ethical use: we kindly ask that you NOT use this software to harvest email
+addresses or to collect any other private information about people. Doing so
+would dishonor our work and waste the many hours we have spent on it.
+
+Please visit our Website: http://www.httrack.com
+*/
+
+/* Fuzz the HTTP response-header parser (htslib.c): treatfirstline on the
+   status line, treathead on each following header. Both consume raw bytes
+   off the wire and copy fields into fixed htsblk buffers; treathead also
+   mutates its line in place and drives cookie parsing. */
+#include "fuzz.h"
+#include "htslib.h"
+#include "htsbauth.h"
+
+int LLVMFuzzerTestOneInput(const uint8_t *data, size_t size) {
+  char *buf = fuzz_strdup(data, size);
+  htsblk r;
+  t_cookie *cookie = calloct(1, sizeof(*cookie));
+  char *line = malloct(size + 1);
+  char *p = buf;
+  int first = 1;
+
+  memset(&r, 0, sizeof(r));
+  cookie->max_len = (int) sizeof(cookie->data);
+  r.location = malloct(HTS_URLMAXSIZE * 2);
+  r.location[0] = '\0';
+
+  /* feed one header line at a time, as the receive loop does */
+  while (p != NULL && *p != '\0') {
+    char *nl = strchr(p, '\n');
+    size_t n = (nl != NULL) ? (size_t) (nl - p) : strlen(p);
+
+    memcpy(line, p, n);
+    line[n] = '\0';
+    if (first) {
+      treatfirstline(&r, line);
+      first = 0;
+    } else {
+      treathead(cookie, "www.example.com", "/", &r, line);
+    }
+    p = (nl != NULL) ? nl + 1 : NULL;
+  }
+
+  freet(r.location);
+  freet(line);
+  freet(cookie);
+  freet(buf);
+  return 0;
+}

src/htscache.c

@@ -1938,15 +1938,17 @@ void cache_init(cache_back * cache, httrackp * opt) {
             char linepos[256];
             int pos;
 
-            while((a != NULL) && (a < (cache->use + buffl))) {
+            const char *const end = cache->use + buffl;
+
+            while ((a != NULL) && (a < end)) {
               a = strchr(a + 1, '\n');  /* start of line */
               if (a) {
                 a++;
                 /* read "host/file" */
-                a += binput(a, line, HTS_URLMAXSIZE);
-                a += binput(a, line + strlen(line), HTS_URLMAXSIZE);
+                a += cache_binput(a, end, line, HTS_URLMAXSIZE);
+                a += cache_binput(a, end, line + strlen(line), HTS_URLMAXSIZE);
                 /* read position */
-                a += binput(a, linepos, 200);
+                a += cache_binput(a, end, linepos, 200);
                 sscanf(linepos, "%d", &pos);
                 coucal_add(cache->hashtable, line, pos);
               }
@@ -2291,23 +2293,40 @@ int cache_brstr(char *adr, char *s, size_t s_size) {
   char buff[256 + 4];
 
   off = binput(adr, buff, 256);
-  adr += off;
-  sscanf(buff, "%d", &i);
-  if (i < 0 || i > 32768) /* guard a corrupt length */
+  /* binput stops at the buffer's terminating NUL; a value can only follow a
+     real line terminator, so never step past end-of-buffer. */
+  if (adr[off - 1] == '\0') {
+    s[0] = '\0';
+    return off - 1;
+  }
+  /* an empty/non-numeric field leaves i unset: treat as length 0 */
+  if (sscanf(buff, "%d", &i) != 1 || i < 0 || i > 32768)
     i = 0;
   if (i > 0) {
-    /* copy at most s_size-1 bytes; advance past the full field regardless */
-    const size_t store = (size_t) i < s_size ? (size_t) i : s_size - 1;
+    /* A corrupt/truncated cache may declare a length past the buffer end;
+       bound both the copy and the advance to the bytes actually present. */
+    const size_t avail = strnlen(adr + off, (size_t) i);
+    const size_t store = avail < s_size ? avail : s_size - 1;
 
-    strncpy(s, adr, store);
+    memcpy(s, adr + off, store);
     s[store] = '\0';
+    off += (int) avail;
   } else {
     s[0] = '\0';
   }
-  off += i;
   return off;
 }
 
+/* binput bounded to a NUL-terminated buffer: refuse to start a read at or
+   past `end`, so a prior over-advance can't walk a cache-index parse OOB. */
+int cache_binput(char *adr, const char *end, char *s, int max) {
+  if (adr >= end) {
+    s[0] = '\0';
+    return 0;
+  }
+  return binput(adr, s, max);
+}
+
 /* idem, mais en int */
 int cache_brint(char *adr, int *i) {
   char s[256];

src/htscache.h

@@ -97,6 +97,8 @@ int cache_readdata(cache_back * cache, const char *str1, const char *str2,
 void cache_rstr(FILE *fp, char *s, size_t s_size);
 char *cache_rstr_addr(FILE * fp);
 int cache_brstr(char *adr, char *s, size_t s_size);
+/* binput over a NUL-terminated buffer, bounded: no read starts at/past end. */
+int cache_binput(char *adr, const char *end, char *s, int max);
 int cache_brint(char *adr, int *i);
 void cache_rint(FILE * fp, int *i);
 void cache_rLLint(FILE * fp, LLint * i);

src/htscoremain.c

@@ -1953,18 +1953,20 @@ static int hts_main_internal(int argc, char **argv, httrackp * opt) {
                     char BIGSTK url[HTS_URLMAXSIZE * 2];
                     char linepos[256];
                     int pos;
-                    char *cacheNdx =
-                      readfile(fconcat
-                               (OPT_GET_BUFF(opt), OPT_GET_BUFF_SIZE(opt), StringBuff(opt->path_log),
-                                "hts-cache/new.ndx"));
+                    LLint cacheNdxLen = 0;
+                    char *cacheNdx = readfile2(
+                        fconcat(OPT_GET_BUFF(opt), OPT_GET_BUFF_SIZE(opt),
+                                StringBuff(opt->path_log), "hts-cache/new.ndx"),
+                        &cacheNdxLen);
                     cache_init(&cache, opt);    /* load cache */
                     if (cacheNdx != NULL) {
                       char firstline[256];
                       char *a = cacheNdx;
+                      const char *const end = cacheNdx + cacheNdxLen;
 
                       a += cache_brstr(a, firstline, sizeof(firstline));
                       a += cache_brstr(a, firstline, sizeof(firstline));
-                      while(a != NULL) {
+                      while (a != NULL && a < end) {
                         a = strchr(a + 1, '\n');        /* start of line */
                         if (a) {
                           htsblk r;
@@ -1972,15 +1974,15 @@ static int hts_main_internal(int argc, char **argv, httrackp * opt) {
                           /* */
                           a++;
                           /* read "host/file" */
-                          a += binput(a, afs.af.adr, HTS_URLMAXSIZE);
-                          a += binput(a, afs.af.fil, HTS_URLMAXSIZE);
+                          a += cache_binput(a, end, afs.af.adr, HTS_URLMAXSIZE);
+                          a += cache_binput(a, end, afs.af.fil, HTS_URLMAXSIZE);
                           url[0] = '\0';
                           if (!link_has_authority(afs.af.adr))
                             strcatbuff(url, "http://");
                           strcatbuff(url, afs.af.adr);
                           strcatbuff(url, afs.af.fil);
                           /* read position */
-                          a += binput(a, linepos, 200);
+                          a += cache_binput(a, end, linepos, 200);
                           sscanf(linepos, "%d", &pos);
                           if (!hasFilter
                               || (strjoker(url, filter, NULL, NULL) != NULL)

src/htsselftest.c

@@ -46,6 +46,7 @@ Please visit our Website: http://www.httrack.com
 #include "htsdefines.h"
 #include "htslib.h"
 #include "htsparse.h"
+#include "htscache.h"
 #include "htscache_selftest.h"
 #include "htsdns_selftest.h"
 #include "htscharset.h"
@@ -1390,6 +1391,78 @@ static int st_cache(httrackp *opt, int argc, char **argv) {
   return err;
 }
 
+/* A corrupt cache index (.ndx) must not walk the length-prefixed scan past
+   the buffer. Checks the two primitives the loader is built from. */
+static int st_cacheindex(httrackp *opt, int argc, char **argv) {
+  int fail = 0;
+
+  (void) opt;
+  (void) argc;
+  (void) argv;
+
+  /* A length prefix that overstates the bytes present must bound the advance
+     to the buffer, not trust the declared length. */
+  {
+    static const char src[] = "32768\nCACHE-1.1";
+    const size_t len = sizeof(src) - 1;
+    char *buf = malloct(len + 1);
+    char s[256];
+    int off;
+
+    memcpy(buf, src, len + 1);
+    off = cache_brstr(buf, s, sizeof(s));
+    if (off > (int) len) {
+      printf("cacheindex: over-advance off=%d len=%d\n", off, (int) len);
+      fail = 1;
+    }
+    if (strcmp(s, "CACHE-1.1") != 0) {
+      printf("cacheindex: value=%s\n", s);
+      fail = 1;
+    }
+    freet(buf);
+  }
+
+  /* cache_binput reads a field while in bounds, but refuses one starting at
+     or past end-of-buffer. */
+  {
+    char buf[8] = "ab\ncd";
+    const char *const end = buf + 5;
+    char s[16];
+
+    if (cache_binput(buf, end, s, sizeof(s)) != 3 || strcmp(s, "ab") != 0)
+      fail = 1; /* normal read: "ab" then the '\n', 3 bytes consumed */
+    if (cache_binput(end, end, s, sizeof(s)) != 0 || s[0] != '\0')
+      fail = 1;
+  }
+
+  /* Drive the full loader scan over a truncated index: a declared length that
+     overshoots plus a half-written entry. ASan aborts here on the pre-fix
+     scan; the cursor must never leave the buffer. */
+  {
+    static const char src[] = "9\nCACHE-1.1\n99\nwww.example.com\n/a";
+    const size_t len = sizeof(src) - 1;
+    char *buf = malloct(len + 1);
+    const char *const end = buf + len;
+    char line[256];
+    char *a = buf;
+
+    memcpy(buf, src, len + 1);
+    a += cache_brstr(a, line, sizeof(line));
+    a += cache_brstr(a, line, sizeof(line));
+    while (a != NULL && a < end) {
+      a = strchr(a + 1, '\n');
+      if (a == NULL)
+        break;
+      a++;
+      a += cache_binput(a, end, line, sizeof(line));
+    }
+    freet(buf);
+  }
+
+  printf("cacheindex: %s\n", fail ? "FAIL" : "OK");
+  return fail;
+}
+
 static int st_cache_golden(httrackp *opt, int argc, char **argv) {
   int regen, err;
 
@@ -2213,6 +2286,8 @@ static const struct selftest_entry {
     {"sniff", "<content-type> <hex:..|text>", "MIME magic consistency",
      st_sniff},
     {"cache", "<dir>", "cache read/write round-trip self-test", st_cache},
+    {"cacheindex", "", "cache-index (.ndx) parse must stay in bounds",
+     st_cacheindex},
     {"cache-golden", "<dir> [regen]", "frozen cache-format read self-test",
      st_cache_golden},
     {"cache-writefail", "<dir>", "cache write-failure handling self-test",

tests/01_engine-cacheindex.test

@@ -0,0 +1,10 @@
+#!/bin/bash
+#
+
+set -euo pipefail
+
+# Cache-index (.ndx) parse must stay inside the buffer on a corrupt length
+# prefix (ASan aborts here on the pre-fix binary; the OK check gates the
+# non-sanitized build too).
+
+test "$(httrack -O /dev/null -#test=cacheindex)" == 'cacheindex: OK'

tests/Makefile.am

@@ -38,6 +38,7 @@ TESTS = \
 	01_engine-filterdual.test \
 	01_engine-ftp-line.test \
 	01_engine-ftp-userpass.test \
+	01_engine-cacheindex.test \
 	01_engine-hashtable.test \
 	01_engine-header.test \
 	01_engine-idna.test \