pthread_mutexattr_setrobust.3: On Linux execve(2) also causes EOWNERDIED notification

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

diff --git a/man3/pthread_mutexattr_setrobust.3 b/man3/pthread_mutexattr_setrobust.3
index 823b043265bb3006410eb8941ff5af2f5bdf7bbc..13550e1a40c9ade0cf8d5ac93b4ce9f39246ac3a 100644 (file)
--- a/man3/pthread_mutexattr_setrobust.3
+++ b/man3/pthread_mutexattr_setrobust.3
@@ -88,8 +88,6 @@ on the mutex will block indefinitely.
 If a mutex is initialized with the
 .BR PTHREAD_MUTEX_ROBUST ,
 attribute and its owner dies without unlocking it,
-.\" FIXME Yubin claims execve(2) without first unlocking yields
-.\" the same behavior. To be verified...
 any future attempts to call
 .BR pthread_mutex_lock (3)
 on this mutex will succeed and return
@@ -152,6 +150,17 @@ were added to glibc in version 2.12.
 .SH CONFORMING TO
 POSIX.1-2008.
 .SH NOTES
+In the Linux implementation,
+when using process-shared robust mutexes, a waiting thread also receives the
+.B EOWNERDEAD
+notification if the owner of a mutex performs an
+.BR execve (2)
+without first unlocking the mutex.
+POSIX.1 does not specify this detail,
+but the same behavior also occurs in at least some
+.\" E.g., Solaris, according to its manual page
+other implementations.
+.PP
 Before the addition of
 .BR pthread_mutexattr_getrobust ()
 and