From: Willy Tarreau <w@1wt.eu>
Date: Wed, 12 Mar 2025 14:54:36 +0000 (+0100)
Subject: BUG/MEDIUM: thread: use pthread_self() not ha_pthread[tid] in set_affinity
X-Git-Tag: v3.2-dev8~107
X-Git-Url: http://git.ipfire.org/cgi-bin/gitweb.cgi?a=commitdiff_plain;h=12383fd9f5b3614dbffec6260b82659c3c5fd0df;p=thirdparty%2Fhaproxy.git

BUG/MEDIUM: thread: use pthread_self() not ha_pthread[tid] in set_affinity

A bug was uncovered by the work on NUMA. It only triggers in the CI
with libmusl due to a race condition. What happens is that the call
to set_thread_cpu_affinity() is done very early in the polling loop,
and that it relies on ha_pthread[tid] instead of pthread_self(). The
problem is that ha_pthread[tid] is only set by the return from
pthread_create(), which might happen later depending on the number of
CPUs available to run the starting thread.

Let's just use pthread_self() here. ha_pthread[] is only used to send
signals between threads, there's no point in using it here.

This can be backported to 2.6.
---

diff --git a/src/thread.c b/src/thread.c
index fa47cd67d..e54a32ef9 100644
--- a/src/thread.c
+++ b/src/thread.c
@@ -287,14 +287,14 @@ void set_thread_cpu_affinity()
 			thread_affinity_policy_data_t cpu_set = { j - 1 };
 			thread_port_t mthread;
 
-			mthread = pthread_mach_thread_np(ha_pthread[tid]);
+			mthread = pthread_mach_thread_np(pthread_self());
 			thread_policy_set(mthread, THREAD_AFFINITY_POLICY, (thread_policy_t)&cpu_set, 1);
 			set &= ~(1UL << (j - 1));
 		}
 #  else
 		struct hap_cpuset *set = &cpu_map[tgid - 1].thread[ti->ltid];
 
-		pthread_setaffinity_np(ha_pthread[tid], sizeof(set->cpuset), &set->cpuset);
+		pthread_setaffinity_np(pthread_self(), sizeof(set->cpuset), &set->cpuset);
 #  endif
 	}
 #endif /* USE_CPU_AFFINITY */