optimize pthread termination in the non-detached case

we can avoid blocking signals by simply using a flag to mark that the
thread has exited and prevent it from getting counted in the rsyscall
signal-pingpong. this restores the original pthread create/join
throughput from before the sigprocmask call was added.

diff --git a/src/internal/pthread_impl.h b/src/internal/pthread_impl.h
index adb80f2f6783ae737a4d3ca44df16b8bf2943f99..c17401113107079c6d0c44b402055f7914cabb54 100644 (file)
--- a/src/internal/pthread_impl.h
+++ b/src/internal/pthread_impl.h
@@ -36,6 +36,7 @@ struct pthread {
        struct __ptcb *cancelbuf;
        void **tsd;
        pthread_attr_t attr;
+       volatile int dead;
 };
 
 #define __SU (sizeof(size_t)/sizeof(int))

diff --git a/src/thread/pthread_create.c b/src/thread/pthread_create.c
index 9c24b844183993cf9b1ea2c18138db6e1ae7dc55..d829fa26950c30d841cd7f3a946c6feda997d09b 100644 (file)
--- a/src/thread/pthread_create.c
+++ b/src/thread/pthread_create.c
@@ -25,13 +25,16 @@ void __pthread_unwind_next(struct __ptcb *cb)
                }
        }
 
-       syscall4(__NR_rt_sigprocmask, SIG_BLOCK, (long)(uint64_t[1]){-1},0,8);
+       /* Mark this thread dead before decrementing count */
+       self->dead = 1;
 
        if (!a_fetch_add(&libc.threads_minus_1, -1))
                exit(0);
 
-       if (self->detached && self->map_base)
+       if (self->detached && self->map_base) {
+               syscall4(__NR_rt_sigprocmask, SIG_BLOCK, (long)(uint64_t[1]){-1},0,8);
                __unmapself(self->map_base, self->map_size);
+       }
 
        __syscall_exit(0);
 }
@@ -76,9 +79,17 @@ static struct {
 
 static void rsyscall_handler(int sig, siginfo_t *si, void *ctx)
 {
-       if (si->si_code > 0 || si->si_pid != __pthread_self()->pid) return;
+       struct pthread *self = __pthread_self();
+
+       if (si->si_code > 0 || si->si_pid != self->pid ||
+               rs.cnt == libc.threads_minus_1) return;
 
-       if (rs.cnt == libc.threads_minus_1) return;
+       /* Threads which have already decremented themselves from the
+        * thread count must not increment rs.cnt or otherwise act. */
+       if (self->dead) {
+               __wait(&rs.hold, 0, 1, 1);
+               return;
+       }
 
        if (syscall6(rs.nr, rs.arg[0], rs.arg[1], rs.arg[2],
                rs.arg[3], rs.arg[4], rs.arg[5]) < 0 && !rs.err) rs.err=errno;