Reduce repeated work in HTTP header parsing by moonchen · Pull Request #13376 · apache/trafficserver

moonchen · 2026-07-13T18:24:25Z

Reduce repeated work in HTTP header parsing.

Replace per-byte libc calls in URL compliance validation with a vectorizable, locale-independent ASCII range check.
Simplify request-target validation using the stored URL fields.
Combine field-name colon scanning, hashing, and character validation into one pass, reusing the existing well-known-string table.
Skip duplicate lookup for well-known fields when their presence bit is clear.
Append adjacent duplicate fields in constant time instead of searching the duplicate chain.

Includes a header parsing benchmark harness and regression tests for validation, field-name scanning, duplicate attachment, and parser reuse.

Copilot

Copilot encountered an error and was unable to review this pull request. You can try again by re-requesting a review.

Copilot

Pull request overview

Copilot reviewed 9 out of 9 changed files in this pull request and generated 1 comment.

Suppressed comments (3)

tools/benchmark/benchmark_HdrParse.cc:797

The --profile unknown-target error message omits the supported "wks-lower"/"wkslower" target even though parse_target() accepts it, which can mislead users.

    Target t = parse_target(profile_target);
    if (t == Target::Unknown) {
      std::fprintf(stderr, "unknown target '%s' (want: request|response|mime|url|wks)\n", profile_target.c_str());
      return 2;

src/proxy/hdrs/HdrToken.cc:606

hdrtoken_tokenize_prehashed()’s comment says it does an “exact ASCII-case-insensitive byte compare”, but the implementation currently only checks (hash, length) and returns the bucket token without validating the bytes. Either adjust the comment or add the byte-compare; adding the compare also prevents accidental or crafted hash/length collisions from being misclassified as a WKS token.

// WKS lookup for a name whose FNV-1a hash the caller has already computed
// (e.g. fused into the field-name scan). Does the slot/length narrowing plus
// the exact ASCII-case-insensitive byte compare, but no hashing and no
// interned-pointer test, so it is only valid for a non-interned `string`.
int
hdrtoken_tokenize_prehashed(const char *string, int string_len, uint32_t hash, const char **wks_string_out)
{
  uint32_t            slot   = hash_to_slot(hash);
  HdrTokenHashBucket *bucket = &(hdrtoken_hash_table[slot]);

  if ((bucket->wks != nullptr) && (bucket->hash == hash) && (hdrtoken_wks_to_length(bucket->wks) == string_len)) {
    int wks_idx = hdrtoken_wks_to_index(bucket->wks);
    if (wks_string_out) {
      *wks_string_out = bucket->wks;
    }
    return wks_idx;
  }

src/proxy/hdrs/URL.cc:1210

url_is_mostly_compliant() no longer emits the debug message that previously identified the first offending byte (whitespace/non-printable). That makes troubleshooting strict_uri_parsing=2 failures harder. You can keep the vectorized scan and only do a second scalar scan when an invalid byte was detected, to log the first bad value.

  // Mode 2 accepts exactly the printable, non-space ASCII range 0x21..0x7E --
  // equivalent to the previous isspace()/isprint() pair, but locale-independent
  // and call-free. This runs on every request target under the default
  // strict_uri_parsing=2. OR-reducing an out-of-range flag over the whole target
  // (no early exit, no data-dependent branch) lets the compiler auto-vectorize
  // the scan to the build's SIMD; ATS builds -O3, where clang and GCC both do.
  unsigned char bad = 0;
  for (const char *i = start; i < end; ++i) {
    unsigned char const c  = static_cast<unsigned char>(*i);
    bad                   |= static_cast<unsigned char>((c < 0x21) | (c > 0x7E));
  }
  return bad == 0;

Copilot

Warning

Copilot couldn't run its full agentic review because it didn't start before the timeout. Make sure your repository has a runner available, or add a copilot-code-review.yml file specifying one with the runs-on attribute. See the docs for more details.

Pull request overview

Copilot reviewed 9 out of 9 changed files in this pull request and generated 5 comments.

moonchen · 2026-09-09T15:45:41Z

+int
+hdrtoken_field_name_scan(const char *string, int maxlen, uint32_t *hash_out, bool *all_valid_out)
+{
+  uint32_t hval      = HDRTOKEN_HASH_SEED; // same FNV-1a name hash as hdrtoken_hash
+  bool     all_valid = true;
+  int      i         = 0;
+
+  for (; i < maxlen; ++i) {
+    unsigned char const uc = static_cast<unsigned char>(string[i]);
+    if (uc == ':') {
+      break;
+    }
+    hval       = hdrtoken_hash_step(hval, hdrtoken_ascii_toupper(uc));
+    all_valid &= (ParseRules::is_http_field_name(static_cast<char>(uc)) != 0);
+  }
+
+  *hash_out      = hval;
+  *all_valid_out = all_valid;
+  return (i < maxlen) ? i : -1;


hash_out is gone from hdrtoken_tokenize — it was dead, its only caller wrote it and never read it back.

For hdrtoken_field_name_scan both out-params are now documented as required rather than made optional. It is a hot-path helper with two in-tree callers; a conditional store would hide a caller bug instead of surfacing it.

moonchen · 2026-09-09T15:45:45Z

+  if (!profile_target.empty()) {
+    Target t = parse_target(profile_target);
+    if (t == Target::Unknown) {
+      std::fprintf(stderr, "unknown target '%s' (want: request|response|mime|url|wks)\n", profile_target.c_str());


Added wks-lower to the list.

Header parsing needs repeatable measurements before its hot paths can be optimized. Cover zero-copy parsing under the default strict URI mode, copied inputs, modern browser headers, duplicate-heavy responses, and canonical and lowercase WKS names in one harness. Include a profiling loop and file-loaded corpora for investigating representative workloads.

Default URI validation pays for libc character classification on every byte. Use a branchless ASCII range reduction to enable vectorization and make acceptance locale-independent; rejection no longer logs the offending byte. Exhaustive differential tests cover every byte value across vector boundaries and scalar tails.

Request-target validation calls three URL getters and follows branches that reduce to a direct test of the stored host and scheme fields. Express that condition directly while preserving the existing treatment of origin, asterisk, absolute, and authority forms.

Field names were walked separately to find the colon, hash the name, and validate its characters. Fuse those passes and reuse the WKS lookup through a prehashed entry point, preserving whitespace normalization and the current table representation. Parity tests check the delimiter position, character validation, and token lookup.

A clear presence bit already proves that a well-known field has no duplicate in the header. Use that result when attaching parsed fields to avoid redundant lookup work. Well-known names without a presence mask and non-well-known names retain the normal duplicate search.

Consecutive fields such as Set-Cookie repeatedly search an existing duplicate chain even though their predecessor is already its tail. Derive that predecessor from the current header block and append directly, avoiding parser-held pointers that could outlive the header. Tests compare duplicate chains with normal attachment and cover parser reuse after a header is destroyed.

Copilot

Warning

Copilot couldn't run its full agentic review because it didn't start before the timeout. Make sure your repository has a runner available, or add a copilot-code-review.yml file specifying one with the runs-on attribute. See the docs for more details.

Pull request overview

Copilot reviewed 9 out of 9 changed files in this pull request and generated 4 comments.

+extern void hdrtoken_init();
+extern int  hdrtoken_tokenize(const char *string, int string_len, const char **wks_string_out = nullptr);
+extern int  hdrtoken_tokenize_prehashed(const char *string, int string_len, uint32_t hash, const char **wks_string_out = nullptr);
+extern int  hdrtoken_field_name_scan(const char *string, int maxlen, uint32_t *hash_out, bool *all_valid_out);
+extern int  hdrtoken_method_tokenize(const char *string, int string_len);


+hdrtoken_field_name_scan(const char *string, int maxlen, uint32_t *hash_out, bool *all_valid_out)
+{
+  uint32_t hval      = HDRTOKEN_HASH_SEED; // same FNV-1a name hash as hdrtoken_hash
+  bool     all_valid = true;
+  int      i         = 0;
+
+  for (; i < maxlen; ++i) {
+    unsigned char const uc = static_cast<unsigned char>(string[i]);
+    if (uc == ':') {
+      break;
+    }
+    hval       = hdrtoken_hash_step(hval, hdrtoken_ascii_toupper(uc));
+    all_valid &= (ParseRules::is_http_field_name(static_cast<char>(uc)) != 0);
+  }
+
+  *hash_out      = hval;
+  *all_valid_out = all_valid;
+  return (i < maxlen) ? i : -1;
+}


+    bad                   |= static_cast<unsigned char>((c < 0x21) | (c > 0x7E));
  }
-  return true;
+  return bad == 0;


+  auto   t1     = std::chrono::steady_clock::now();
+  double ns     = std::chrono::duration<double, std::nano>(t1 - t0).count();
+  double ns_per = ns / static_cast<double>(count);
+


moonchen self-assigned this Jul 13, 2026

moonchen added the Performance label Jul 13, 2026

moonchen force-pushed the header-parse-optimization branch from dd3ee23 to 0442cf1 Compare July 13, 2026 20:38

moonchen added this to the 11.0.0 milestone Jul 13, 2026

moonchen force-pushed the header-parse-optimization branch from 0442cf1 to 656735c Compare July 13, 2026 22:18

moonchen marked this pull request as ready for review August 17, 2026 16:07

Copilot AI lite review requested due to automatic review settings August 17, 2026 16:07

Copilot started reviewing on behalf of moonchen August 17, 2026 16:08 View session

Copilot AI reviewed Aug 17, 2026

moonchen marked this pull request as draft August 17, 2026 20:54

moonchen force-pushed the header-parse-optimization branch 3 times, most recently from 52512ce to 25227f3 Compare August 20, 2026 00:01

moonchen marked this pull request as ready for review August 20, 2026 00:01

Copilot AI review requested due to automatic review settings August 20, 2026 00:01

Copilot started reviewing on behalf of moonchen August 20, 2026 00:04 View session

Copilot AI reviewed Aug 20, 2026

View reviewed changes

Comment thread tools/benchmark/benchmark_HdrParse.cc

moonchen marked this pull request as draft August 20, 2026 18:47

moonchen force-pushed the header-parse-optimization branch from 25227f3 to c0d1671 Compare September 9, 2026 14:50

moonchen changed the title ~~Header Parsing Optimizations~~ Reduce repeated work in HTTP header parsing Sep 9, 2026

moonchen marked this pull request as ready for review September 9, 2026 14:50

Copilot AI review requested due to automatic review settings September 9, 2026 14:50

moonchen added this to ATS v11.0.x Sep 9, 2026

Copilot AI reviewed Sep 9, 2026

View reviewed changes

moonchen added 5 commits September 9, 2026 10:41

Copilot AI review requested due to automatic review settings September 9, 2026 15:45

moonchen force-pushed the header-parse-optimization branch from c0d1671 to 238fb0c Compare September 9, 2026 15:45

Copilot AI reviewed Sep 9, 2026

View reviewed changes

Copilot started reviewing on behalf of moonchen September 9, 2026 20:00 View session

Copilot started reviewing on behalf of moonchen September 9, 2026 20:16 View session

Navigation Menu

Reduce repeated work in HTTP header parsing - #13376

Reduce repeated work in HTTP header parsing#13376
moonchen wants to merge 6 commits into
apache:masterfrom
moonchen:header-parse-optimization

moonchen commented Jul 13, 2026 •

edited

Loading

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI left a comment

Uh oh!

Uh oh!

Copilot AI left a comment

Uh oh!

Uh oh!

moonchen Sep 9, 2026

Uh oh!

Uh oh!

Uh oh!

moonchen Sep 9, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

moonchen commented Jul 13, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

moonchen Sep 9, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

moonchen Sep 9, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

moonchen commented Jul 13, 2026 •

edited

Loading