From: Willy Tarreau <w@1wt.eu>
Date: Wed, 25 Oct 2023 13:42:27 +0000 (+0200)
Subject: DEBUG: pools: detect that malloc_trim() is in progress
X-Git-Tag: v2.9-dev9~28
X-Git-Url: http://git.ipfire.org/cgi-bin/gitweb.cgi?a=commitdiff_plain;h=96bb99a87da409c21364d38d2d429c9239ef08df;p=thirdparty%2Fhaproxy.git

DEBUG: pools: detect that malloc_trim() is in progress

Now when calling ha_panic() with a thread still under malloc_trim(),
we'll set a new tainted flag to easily report it, and the output
trace will report that this condition happened and will suggest to
use no-memory-trimming to avoid it in the future.
---

diff --git a/include/haproxy/bug.h b/include/haproxy/bug.h
index fbed1203d9..c980727e7a 100644
--- a/include/haproxy/bug.h
+++ b/include/haproxy/bug.h
@@ -251,6 +251,7 @@ enum tainted_flags {
 	TAINTED_PANIC                  = 0x00000200, /* a panic dump has started */
 	TAINTED_LUA_STUCK              = 0x00000400, /* stuck in a Lua context */
 	TAINTED_LUA_STUCK_SHARED       = 0x00000800, /* stuck in a shared Lua context */
+	TAINTED_MEM_TRIMMING_STUCK     = 0x00001000, /* stuck while trimming memory */
 };
 
 /* this is a bit field made of TAINTED_*, and is declared in haproxy.c */
diff --git a/include/haproxy/pool.h b/include/haproxy/pool.h
index 133e954fc9..5375b35c91 100644
--- a/include/haproxy/pool.h
+++ b/include/haproxy/pool.h
@@ -100,6 +100,9 @@
 /* poison each newly allocated area with this byte if >= 0 */
 extern int mem_poison_byte;
 
+/* trim() in progress */
+extern int pool_trim_in_progress;
+
 /* set of POOL_DBG_* flags */
 extern uint pool_debugging;
 
diff --git a/src/debug.c b/src/debug.c
index fbab47e2e2..9417a9613c 100644
--- a/src/debug.c
+++ b/src/debug.c
@@ -240,6 +240,10 @@ void ha_thread_dump_one(int thr, int from_signal)
 			}
 		}
 #endif
+
+		if (HA_ATOMIC_LOAD(&pool_trim_in_progress))
+			mark_tainted(TAINTED_MEM_TRIMMING_STUCK);
+
 		/* We only emit the backtrace for stuck threads in order not to
 		 * waste precious output buffer space with non-interesting data.
 		 * Please leave this as the last instruction in this function
@@ -468,6 +472,14 @@ void ha_panic()
 		DISGUISE(write(2, trash.area, trash.data));
 	}
 #endif
+	if (get_tainted() & TAINTED_MEM_TRIMMING_STUCK) {
+		chunk_printf(&trash,
+			     "### Note: one thread was found stuck under malloc_trim(), which can run for a\n"
+			     "          very long time on large memory systems. You way want to disable this\n"
+			     "          memory reclaiming feature by setting 'no-memory-trimming' in the\n"
+			     "          'global' section of your configuration to avoid this in the future.\n");
+		DISGUISE(write(2, trash.area, trash.data));
+	}
 
 	for (;;)
 		abort();
diff --git a/src/pool.c b/src/pool.c
index 6af1af48a3..02b5c6d250 100644
--- a/src/pool.c
+++ b/src/pool.c
@@ -38,6 +38,7 @@ THREAD_LOCAL size_t pool_cache_count = 0;                /* #cache objects   */
 
 static struct list pools __read_mostly = LIST_HEAD_INIT(pools);
 int mem_poison_byte __read_mostly = 'P';
+int pool_trim_in_progress = 0;
 uint pool_debugging __read_mostly =               /* set of POOL_DBG_* flags */
 #ifdef DEBUG_FAIL_ALLOC
 	POOL_DBG_FAIL_ALLOC |
@@ -218,6 +219,8 @@ int malloc_trim(size_t pad)
 	if (disable_trim)
 		return ret;
 
+	HA_ATOMIC_INC(&pool_trim_in_progress);
+
 	if (my_mallctl) {
 		/* here we're on jemalloc and malloc_trim() is called either
 		 * by haproxy or another dependency (the worst case that
@@ -263,6 +266,8 @@ int malloc_trim(size_t pad)
 		}
 	}
 #endif
+	HA_ATOMIC_DEC(&pool_trim_in_progress);
+
 	/* here we have ret=0 if nothing was release, or 1 if some were */
 	return ret;
 }