From b8a3baffddb0f0082a9b250693d26d934eaf2650 Mon Sep 17 00:00:00 2001
From: Matt Benjamin <mbenjamin@redhat.com>
Date: Thu, 24 May 2018 16:09:01 -0400
Subject: [PATCH] rgw: add configurable AWS-compat invalid range get behavior

If rgw_ignore_get_invalid_range is set, treat invalid range
restrictions as a request for the full object.  By default, retain
the RGW behavior to fail with ERANGE.

Fixes: http://tracker.ceph.com/issues/24317

Signed-off-by: Matt Benjamin <mbenjamin@redhat.com>
---
 src/common/legacy_config_opts.h |  1 +
 src/common/options.cc           |  6 ++++
 src/rgw/rgw_op.cc               | 61 +++++++++++++++++++--------------
 src/rgw/rgw_op.h                |  3 ++
 4 files changed, 45 insertions(+), 26 deletions(-)

diff --git a/src/common/legacy_config_opts.h b/src/common/legacy_config_opts.h
index 5597b62548495..cdaf6f1f8818a 100644
--- a/src/common/legacy_config_opts.h
+++ b/src/common/legacy_config_opts.h
@@ -1299,6 +1299,7 @@ OPTION(rgw_lc_max_objs, OPT_INT)
 OPTION(rgw_lc_debug_interval, OPT_INT)  // Debug run interval, in seconds
 OPTION(rgw_script_uri, OPT_STR) // alternative value for SCRIPT_URI if not set in request
 OPTION(rgw_request_uri, OPT_STR) // alternative value for REQUEST_URI if not set in request
+OPTION(rgw_ignore_get_invalid_range, OPT_BOOL) // treat invalid (e.g., negative) range requests as full
 OPTION(rgw_swift_url, OPT_STR)             // the swift url, being published by the internal swift auth
 OPTION(rgw_swift_url_prefix, OPT_STR) // entry point for which a url is considered a swift url
 OPTION(rgw_swift_auth_url, OPT_STR)        // default URL to go and verify tokens for v1 auth (if not using internal swift auth)
diff --git a/src/common/options.cc b/src/common/options.cc
index 5768f2755909e..14865f9a0988f 100644
--- a/src/common/options.cc
+++ b/src/common/options.cc
@@ -5080,6 +5080,12 @@ std::vector<Option> get_rgw_options() {
     .set_default("")
     .set_description(""),
 
+    Option("rgw_ignore_get_invalid_range", Option::TYPE_BOOL, Option::LEVEL_ADVANCED)
+    .set_default(false)
+    .set_description("Treat invalid (e.g., negative) range request as full")
+    .set_long_description("Treat invalid (e.g., negative) range request "
+			  "as request for the full object (AWS compatibility)"),
+
     Option("rgw_swift_url", Option::TYPE_STR, Option::LEVEL_ADVANCED)
     .set_default("")
     .set_description("Swift-auth storage URL")
diff --git a/src/rgw/rgw_op.cc b/src/rgw/rgw_op.cc
index ca56cbf1b88d8..d539af4a4b96c 100644
--- a/src/rgw/rgw_op.cc
+++ b/src/rgw/rgw_op.cc
@@ -86,41 +86,42 @@ static MultipartMetaFilter mp_filter;
 // at some point
 static constexpr auto S3_EXISTING_OBJTAG = "s3:ExistingObjectTag";
 
-static int parse_range(const char *range, off_t& ofs, off_t& end, bool *partial_content)
+int RGWGetObj::parse_range(void)
 {
   int r = -ERANGE;
-  string s(range);
+  string rs(range_str);
   string ofs_str;
   string end_str;
 
-  *partial_content = false;
+  ignore_invalid_range = s->cct->_conf->rgw_ignore_get_invalid_range;
+  partial_content = false;
 
-  size_t pos = s.find("bytes=");
+  size_t pos = rs.find("bytes=");
   if (pos == string::npos) {
     pos = 0;
-    while (isspace(s[pos]))
+    while (isspace(rs[pos]))
       pos++;
     int end = pos;
-    while (isalpha(s[end]))
+    while (isalpha(rs[end]))
       end++;
-    if (strncasecmp(s.c_str(), "bytes", end - pos) != 0)
+    if (strncasecmp(rs.c_str(), "bytes", end - pos) != 0)
       return 0;
-    while (isspace(s[end]))
+    while (isspace(rs[end]))
       end++;
-    if (s[end] != '=')
+    if (rs[end] != '=')
       return 0;
-    s = s.substr(end + 1);
+    rs = rs.substr(end + 1);
   } else {
-    s = s.substr(pos + 6); /* size of("bytes=")  */
+    rs = rs.substr(pos + 6); /* size of("bytes=")  */
   }
-  pos = s.find('-');
+  pos = rs.find('-');
   if (pos == string::npos)
     goto done;
 
-  *partial_content = true;
+  partial_content = true;
 
-  ofs_str = s.substr(0, pos);
-  end_str = s.substr(pos + 1);
+  ofs_str = rs.substr(0, pos);
+  end_str = rs.substr(pos + 1);
   if (end_str.length()) {
     end = atoll(end_str.c_str());
     if (end < 0)
@@ -137,8 +138,18 @@ static int parse_range(const char *range, off_t& ofs, off_t& end, bool *partial_
   if (end >= 0 && end < ofs)
     goto done;
 
-  r = 0;
+  range_parsed = true;
+  return 0;
+
 done:
+  if (ignore_invalid_range) {
+    partial_content = false;
+    ofs = 0;
+    end = -1;
+    range_parsed = false; // allow retry
+    r = 0;
+  }
+
   return r;
 }
 
@@ -1673,16 +1684,13 @@ bool RGWGetObj::prefetch_data()
   bool prefetch_first_chunk = true;
   range_str = s->info.env->get("HTTP_RANGE");
 
-  if(range_str) {
-    int r = parse_range(range_str, ofs, end, &partial_content);
-    /* error on parsing the range, stop prefetch and will fail in execte() */
+  if (range_str) {
+    int r = parse_range();
+    /* error on parsing the range, stop prefetch and will fail in execute() */
     if (r < 0) {
-      range_parsed = false;
-      return false;
-    } else {
-      range_parsed = true;
+      return false; /* range_parsed==false */
     }
-    /* range get goes to shadown objects, stop prefetch */
+    /* range get goes to shadow objects, stop prefetch */
     if (ofs >= s->cct->_conf->rgw_max_chunk_size) {
       prefetch_first_chunk = false;
     }
@@ -1690,6 +1698,7 @@ bool RGWGetObj::prefetch_data()
 
   return get_data && prefetch_first_chunk;
 }
+
 void RGWGetObj::pre_exec()
 {
   rgw_bucket_object_pre_exec(s);
@@ -1889,9 +1898,9 @@ done_err:
 int RGWGetObj::init_common()
 {
   if (range_str) {
-    /* range parsed error when prefetch*/
+    /* range parsed error when prefetch */
     if (!range_parsed) {
-      int r = parse_range(range_str, ofs, end, &partial_content);
+      int r = parse_range();
       if (r < 0)
         return r;
     }
diff --git a/src/rgw/rgw_op.h b/src/rgw/rgw_op.h
index 23306b9ba77f7..0219c67091074 100644
--- a/src/rgw/rgw_op.h
+++ b/src/rgw/rgw_op.h
@@ -247,6 +247,7 @@ protected:
   map<string, bufferlist> attrs;
   bool get_data;
   bool partial_content;
+  bool ignore_invalid_range;
   bool range_parsed;
   bool skip_manifest;
   bool skip_decrypt{false};
@@ -300,9 +301,11 @@ public:
   void set_get_data(bool get_data) {
     this->get_data = get_data;
   }
+
   int verify_permission() override;
   void pre_exec() override;
   void execute() override;
+  int parse_range();
   int read_user_manifest_part(
     rgw_bucket& bucket,
     const rgw_bucket_dir_entry& ent,
-- 
2.39.5