Unwrap N-level chained proxy URLs to the innermost upstream.

Recursively peel embedded root/http(s) hops in parseChainedFileUrl so
multi_origin, allow_origin, and HTTP ext see the final endpoint.

Assisted-by: Cursor:Composer-2.5 CursorAI

Author: apeters1971

src/XrdApps/XrdClJournalCachePlugin/README.md

@@ -481,14 +481,22 @@ Clients can name a dynamic upstream inside a proxy URL:
 root://proxy.example:1095//root://origin.cern.ch:1094//store/file.dat
 ```
 
-PSS forwarding (`pss.origin =root,http,https`) accepts path-embedded upstreams such as `/root://origin.cern.ch:1094//store/file.dat` as well.
+Longer chains with **N proxies** are supported; JournalCache unwraps through every embedded hop to the innermost upstream:
+
+```
+root://proxy1//root://proxy2//root://proxy3//root://origin.cern.ch:1094//store/file.dat
+```
+
+Use the usual XRootD `//` separator before each path segment (including each embedded URL).
+
+PSS forwarding (`pss.origin =root,http,https`) accepts path-embedded upstreams such as `/root://origin.cern.ch:1094//store/file.dat` as well. Each PSS hop forwards one layer; with `multi_origin = 1`, JournalCache on a proxy unwraps the full chain in one step for open, allowlist checks, and cache keys.
 
 **Plugin options** (client config for the xrootd/PSS process):
 
 | Key | Meaning |
 |-----|---------|
-| `multi_origin = 1` | Unwrap chained URLs to the inner upstream for open + journal cache key |
-| `allow_origin = <regex>` | Allowed upstream patterns (comma-separated or repeated key); matched against full URL, location, or host |
+| `multi_origin = 1` | Unwrap chained URLs to the innermost upstream for open + journal cache key |
+| `allow_origin = <regex>` | Allowed upstream patterns (comma-separated or repeated key); matched against the fully unwrapped URL, location, or host |
 
 Environment overrides: `XRD_JOURNALCACHE_MULTI_ORIGIN`, `XRD_JOURNALCACHE_ALLOW_ORIGIN`.
 

src/XrdApps/XrdClJournalCachePlugin/html/index.html

@@ -530,11 +530,13 @@ <h3>HTTP forwarding proxy (dynamic origins)</h3>
       <h3>Chained multi-origin URLs (<code>root://</code> and friends)</h3>
       <p>Clients can name a dynamic upstream inside a proxy URL:</p>
       <pre><code>root://proxy.example:1095//root://origin.cern.ch:1094//store/file.dat</code></pre>
-      <p>PSS forwarding (<code>pss.origin =root,http,https</code>) also accepts path-embedded upstreams such as <code>/root://origin.cern.ch:1094//store/file.dat</code>.</p>
+      <p>Longer chains with <strong>N proxies</strong> are supported; JournalCache unwraps through every embedded hop to the innermost upstream:</p>
+      <pre><code>root://proxy1//root://proxy2//root://proxy3//root://origin.cern.ch:1094//store/file.dat</code></pre>
+      <p>Use the usual XRootD <code>//</code> separator before each path segment. PSS forwarding (<code>pss.origin =root,http,https</code>) also accepts path-embedded upstreams such as <code>/root://origin.cern.ch:1094//store/file.dat</code>. Each PSS hop forwards one layer; with <code>multi_origin = 1</code>, JournalCache unwraps the full chain in one step for open, allowlist checks, and cache keys.</p>
       <table>
         <tr><th>Key</th><th>Meaning</th></tr>
-        <tr><td><code>multi_origin = 1</code></td><td>Unwrap chained URLs to the inner upstream for open + journal cache key</td></tr>
-        <tr><td><code>allow_origin = &lt;regex&gt;</code></td><td>Allowed upstream patterns (comma-separated or repeated)</td></tr>
+        <tr><td><code>multi_origin = 1</code></td><td>Unwrap chained URLs to the innermost upstream for open + journal cache key</td></tr>
+        <tr><td><code>allow_origin = &lt;regex&gt;</code></td><td>Allowed upstream patterns; matched against the fully unwrapped URL</td></tr>
       </table>
       <p>Environment overrides: <code>XRD_JOURNALCACHE_MULTI_ORIGIN</code>, <code>XRD_JOURNALCACHE_ALLOW_ORIGIN</code>. The HTTP ext handler accepts the same <code>allow_origin</code> lines and rejects disallowed upstreams with <strong>403</strong>.</p>
       <pre><code>multi_origin = 1

src/XrdApps/XrdClJournalCachePlugin/html/index.md

@@ -315,12 +315,20 @@ Clients can name a dynamic upstream inside a proxy URL:
 root://proxy.example:1095//root://origin.cern.ch:1094//store/file.dat
 ```
 
-PSS forwarding (`pss.origin =root,http,https`) also accepts path-embedded upstreams such as `/root://origin.cern.ch:1094//store/file.dat`.
+Longer chains with **N proxies** are supported; JournalCache unwraps through every embedded hop to the innermost upstream:
+
+```
+root://proxy1//root://proxy2//root://proxy3//root://origin.cern.ch:1094//store/file.dat
+```
+
+Use the usual XRootD `//` separator before each path segment (including each embedded URL).
+
+PSS forwarding (`pss.origin =root,http,https`) also accepts path-embedded upstreams such as `/root://origin.cern.ch:1094//store/file.dat`. Each PSS hop forwards one layer; with `multi_origin = 1`, JournalCache on a proxy unwraps the full chain in one step for open, allowlist checks, and cache keys.
 
 | Key | Meaning |
 |-----|---------|
-| `multi_origin = 1` | Unwrap chained URLs to the inner upstream for open + journal cache key |
-| `allow_origin = <regex>` | Allowed upstream patterns (comma-separated or repeated); matched against full URL, location, or host |
+| `multi_origin = 1` | Unwrap chained URLs to the innermost upstream for open + journal cache key |
+| `allow_origin = <regex>` | Allowed upstream patterns (comma-separated or repeated); matched against the fully unwrapped URL, location, or host |
 
 Environment overrides: `XRD_JOURNALCACHE_MULTI_ORIGIN`, `XRD_JOURNALCACHE_ALLOW_ORIGIN`.
 

src/XrdApps/XrdClJournalCachePlugin/http/ForwardingUrl.cc

@@ -69,6 +69,26 @@ EmbeddedFileUrl parseEmbeddedFromRest(std::string rest) {
   return canonicalizeFileUrl(rest);
 }
 
+EmbeddedFileUrl unwrapFullyChained(const EmbeddedFileUrl &first) {
+  EmbeddedFileUrl result = first;
+  if (!result.valid) {
+    return result;
+  }
+
+  while (true) {
+    XrdCl::URL current(result.fileUrl);
+    if (!current.IsValid()) {
+      break;
+    }
+    const EmbeddedFileUrl inner = parseEmbeddedFileUrl(current.GetPath());
+    if (!inner.valid) {
+      break;
+    }
+    result = inner;
+  }
+  return result;
+}
+
 } // namespace
 
 EmbeddedFileUrl parseEmbeddedFileUrl(const std::string &path) {
@@ -80,7 +100,7 @@ EmbeddedFileUrl parseChainedFileUrl(const std::string &url) {
   if (!url.empty() && url.front() == '/') {
     EmbeddedFileUrl embedded = parseEmbeddedFileUrl(url);
     if (embedded.valid) {
-      return embedded;
+      return unwrapFullyChained(embedded);
     }
   }
 
@@ -91,7 +111,7 @@ EmbeddedFileUrl parseChainedFileUrl(const std::string &url) {
 
   EmbeddedFileUrl embedded = parseEmbeddedFileUrl(outer.GetPath());
   if (embedded.valid) {
-    return embedded;
+    return unwrapFullyChained(embedded);
   }
 
   return canonicalizeFileUrl(url);

src/XrdApps/XrdClJournalCachePlugin/http/ForwardingUrl.hh

@@ -13,7 +13,7 @@ struct EmbeddedFileUrl {
 //! Parse `/https://host/path`, `/root://host//path`, etc.
 EmbeddedFileUrl parseEmbeddedFileUrl(const std::string &path);
 
-//! Parse `root://proxy//root://origin//path` and path-embedded upstream URLs.
+//! Parse chained URLs and unwrap through N embedded proxies to the innermost upstream.
 EmbeddedFileUrl parseChainedFileUrl(const std::string &url);
 
 std::string resolveJournalDirWithSettings(const std::string &cacheRoot,

tests/XrdCl/XrdClJournalCache.cc

@@ -728,6 +728,35 @@ TEST(ForwardingUrlTest, ParseChainedRootUrl) {
   EXPECT_EQ(chained.fileUrl.find("proxy.cern.ch"), std::string::npos);
 }
 
+TEST(ForwardingUrlTest, ParseChainedTripleRootUrl) {
+  const auto chained = JournalCache::parseChainedFileUrl(
+      "root://proxy1:1095//root://proxy2:1096//root://origin.cern.ch:1094//"
+      "store/file.dat");
+  ASSERT_TRUE(chained.valid);
+  EXPECT_NE(chained.fileUrl.find("origin.cern.ch"), std::string::npos);
+  EXPECT_NE(chained.fileUrl.find("/store/file.dat"), std::string::npos);
+  EXPECT_EQ(chained.fileUrl.find("proxy1"), std::string::npos);
+  EXPECT_EQ(chained.fileUrl.find("proxy2"), std::string::npos);
+}
+
+TEST(ForwardingUrlTest, ParseChainedTripleEmbeddedPath) {
+  const auto chained = JournalCache::parseChainedFileUrl(
+      "/root://proxy2:1096//root://origin.cern.ch:1094//store/file.dat");
+  ASSERT_TRUE(chained.valid);
+  EXPECT_NE(chained.fileUrl.find("origin.cern.ch"), std::string::npos);
+  EXPECT_EQ(chained.fileUrl.find("proxy2"), std::string::npos);
+}
+
+TEST(ForwardingUrlTest, ParseChainedMixedProtocols) {
+  const auto chained = JournalCache::parseChainedFileUrl(
+      "root://proxy:1095//root://relay:1096//https://cdn.example.org/store/"
+      "file.dat");
+  ASSERT_TRUE(chained.valid);
+  EXPECT_NE(chained.fileUrl.find("cdn.example.org"), std::string::npos);
+  EXPECT_EQ(chained.fileUrl.find("proxy"), std::string::npos);
+  EXPECT_EQ(chained.fileUrl.find("relay"), std::string::npos);
+}
+
 TEST(OriginAllowlistTest, AllowsMatchingHostOrUrl) {
   JournalCache::OriginAllowlist allowlist;
   allowlist.addPattern(R"(^root://([a-z0-9.-]+\.)?cern\.ch(:1094)?/)");