From: Jaya Prakash <jayaprakash@ibm.com>
Date: Thu, 3 Jul 2025 11:25:39 +0000 (+0000)
Subject: common: add admin socket control and CMake flag for cputrace
X-Git-Url: http://git-server-git.apps.pok.os.sepia.ceph.com/?a=commitdiff_plain;h=905d2a0714ed48e5e6757a5a3839fe09e79b7c5c;p=ceph.git

common: add admin socket control and CMake flag for cputrace

- Introduced a separate CMake flag to enable cputrace:
    cmake -DWITH_CPUTRACE=1 ...

- Added admin socket interface for runtime control:
    - `ceph tell osd.X cputrace start`   → Start profiling
    - `ceph tell osd.X cputrace stop`    → Stop profiling
    - `ceph tell osd.X cputrace dump`    → Dump collected metrics
    - `ceph tell osd.X cputrace reset`   → Reset profiling state

This enables developers to dynamically control profiling during runtime
without rebuilding or restarting the OSD, and ensures `cputrace` remains
disabled in production unless explicitly enabled at build time.

Signed-off-by: Jaya Prakash <jayaprakash@ibm.com>
---

diff --git a/CMakeLists.txt b/CMakeLists.txt
index bc255b4df40e..a9a01b1db7d8 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -431,6 +431,8 @@ endif(WITH_LZ4)
 CMAKE_DEPENDENT_OPTION(WITH_CEPH_DEBUG_MUTEX "Use debug ceph::mutex with lockdep" ON
   "CMAKE_BUILD_TYPE STREQUAL Debug" OFF)
 
+option(WITH_CPUTRACE "Enable Cputrace support via the admin socket" OFF)
+
 #if allocator is set on command line make sure it matches below strings
 set(ALLOCATOR "" CACHE STRING
   "specify memory allocator to use. currently tcmalloc, tcmalloc_minimal, \
diff --git a/src/CMakeLists.txt b/src/CMakeLists.txt
index 19dc380d62c7..f82166752ce3 100644
--- a/src/CMakeLists.txt
+++ b/src/CMakeLists.txt
@@ -199,6 +199,10 @@ if(WITH_CEPH_DEBUG_MUTEX)
   add_compile_options($<$<COMPILE_LANGUAGE:CXX>:-DCEPH_DEBUG_MUTEX>)
 endif()
 
+if(WITH_CPUTRACE)
+  add_compile_options($<$<COMPILE_LANGUAGE:CXX>:-DWITH_CPUTRACE>)
+endif()
+
 include(CheckCCompilerFlag)
 if(CMAKE_CXX_COMPILER_ID STREQUAL GNU)
     CHECK_C_COMPILER_FLAG(-fstack-protector-strong HAS_STACK_PROTECT)
diff --git a/src/common/CMakeLists.txt b/src/common/CMakeLists.txt
index 1f21ad4c0c58..276c62aa3f8c 100644
--- a/src/common/CMakeLists.txt
+++ b/src/common/CMakeLists.txt
@@ -112,7 +112,13 @@ set(common_srcs
   mclock_common.cc
   tcp_info.cc)
 
+if(WITH_CPUTRACE)
+  list(APPEND common_srcs
+    cputrace.cc)
+endif()
+
 include_directories(${CMAKE_SOURCE_DIR}/src/dmclock/support/src)
+
 if(WITH_SYSTEMD)
   list(APPEND common_srcs
     Journald.cc)
diff --git a/src/common/ceph_context.cc b/src/common/ceph_context.cc
index db4cae2db0eb..e7179e20142d 100644
--- a/src/common/ceph_context.cc
+++ b/src/common/ceph_context.cc
@@ -67,6 +67,10 @@
 // for CINIT_FLAGS
 #include "common/common_init.h"
 
+#ifdef WITH_CPUTRACE
+#include "common/cputrace.h"
+#endif
+
 #include <iostream>
 #include <pthread.h>
 
@@ -685,6 +689,24 @@ int CephContext::_do_command(
     else if (command == "log reopen") {
       _log->reopen_log_file();
     }
+#ifdef WITH_CPUTRACE
+    else if (command == "cputrace start") {
+      cputrace_start(f);
+    }
+    else if (command == "cputrace stop") {
+      cputrace_stop(f);
+    }
+    else if (command == "cputrace dump") {
+      std::string logger;
+      std::string counter;
+      cmd_getval(cmdmap, "logger", logger);
+      cmd_getval(cmdmap, "counter", counter);
+      cputrace_dump(f, logger, counter);
+    }
+    else if (command == "cputrace reset") {
+      cputrace_reset(f);
+    }
+#endif
     else {
       ceph_abort_msg("registered under wrong command?");    
     }
@@ -781,6 +803,12 @@ CephContext::CephContext(uint32_t module_type_,
   _admin_socket->register_command("log dump", _admin_hook, "dump recent log entries to log file");
   _admin_socket->register_command("log reopen", _admin_hook, "reopen log file");
 
+#ifdef WITH_CPUTRACE
+  _admin_socket->register_command("cputrace start", _admin_hook, "start cpu profiling");
+  _admin_socket->register_command("cputrace stop", _admin_hook, "stop cpu profiling");
+  _admin_socket->register_command("cputrace reset", _admin_hook, "reset cpu profiling");
+  _admin_socket->register_command("cputrace dump name=logger,type=CephString,req=false name=counter,type=CephString,req=false", _admin_hook, "dump cpu profiling results");
+#endif
   _crypto_none = CryptoHandler::create(CEPH_CRYPTO_NONE);
   _crypto_aes = CryptoHandler::create(CEPH_CRYPTO_AES);
   _crypto_random.reset(new CryptoRandom());