Source-Changes-HG archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

[src/thorpej-futex2]: src/sys/kern At the end of futex_wait(), when sleepq_bl...



details:   https://anonhg.NetBSD.org/src/rev/7239dfe38a85
branches:  thorpej-futex2
changeset: 1022746:7239dfe38a85
user:      thorpej <thorpej%NetBSD.org@localhost>
date:      Thu Aug 05 23:23:50 2021 +0000

description:
At the end of futex_wait(), when sleepq_block() returns 0, we would like
to assert that l->l_futex == NULL, because all of the code paths that awaken
a blocked thread in sys_futex.c itself clear l->l_futex.  Unfortunately,
there are certain received-a-signal situations (e.g. SIGKILL) where
sleepq_block() will not return an error after being awakened by the signal,
rendering this assertion too strong.

So, rather than going down the rabbit hole of reasoning out and altering
long-standing behavior of the signals code, just don't assert there and
treat a zero-return from sleepq_block() as an aborted futex wait if
l->l_futex != NULL.

(Thanks chs@ for helping chase this one down.)

diffstat:

 sys/kern/sys_futex.c |  23 +++++++++++++++++++----
 1 files changed, 19 insertions(+), 4 deletions(-)

diffs (44 lines):

diff -r 112aba9b9e0b -r 7239dfe38a85 sys/kern/sys_futex.c
--- a/sys/kern/sys_futex.c      Thu Aug 05 23:14:21 2021 +0000
+++ b/sys/kern/sys_futex.c      Thu Aug 05 23:23:50 2021 +0000
@@ -1,4 +1,4 @@
-/*     $NetBSD: sys_futex.c,v 1.12.4.1 2021/08/05 23:14:21 thorpej Exp $       */
+/*     $NetBSD: sys_futex.c,v 1.12.4.2 2021/08/05 23:23:50 thorpej Exp $       */
 
 /*-
  * Copyright (c) 2018, 2019, 2020 The NetBSD Foundation, Inc.
@@ -30,7 +30,7 @@
  */
 
 #include <sys/cdefs.h>
-__KERNEL_RCSID(0, "$NetBSD: sys_futex.c,v 1.12.4.1 2021/08/05 23:14:21 thorpej Exp $");
+__KERNEL_RCSID(0, "$NetBSD: sys_futex.c,v 1.12.4.2 2021/08/05 23:23:50 thorpej Exp $");
 
 /*
  * Futexes
@@ -1122,8 +1122,23 @@
                else if (error != ETIMEDOUT)
                        error = EINTR;
        } else {
-               KASSERT(l->l_futex == NULL);
-               SDT_PROBE0(futex, wait, finish, normally);
+               f = l->l_futex;
+               if (__predict_false(f != NULL)) {
+                       /*
+                        * There are certain situations that can cause
+                        * sleepq_block() to return 0 even if we were
+                        * signalled (by e.g. SIGKILL).  In this case,
+                        * we will need to release our futex hold and
+                        * return EINTR (we're probably about to die
+                        * anyway).
+                        */
+                       SDT_PROBE0(futex, wait, finish, aborted);
+                       l->l_futex = NULL;
+                       futex_rele(f);
+                       error = EINTR;
+               } else {
+                       SDT_PROBE0(futex, wait, finish, normally);
+               }
        }
 
        return error;



Home | Main Index | Thread Index | Old Index