From: Jens Gustedt <Jens.Gustedt@inria.fr>
To: musl@lists.openwall.com
Subject: [PATCH 6/7] implement __unlock_requeue
Date: Wed, 3 Jan 2018 14:17:12 +0100 [thread overview]
Message-ID: <498d773729dc31d4e9cb0e58308254f2c4332d20.1514985618.git.Jens.Gustedt@inria.fr> (raw)
In-Reply-To: <cover.1514985618.git.Jens.Gustedt@inria.fr>
We need a special case of unlock when we might take advantage of the
futex requeue feature to move a thread from a condition queue to the one
of the correponding mutex
---
src/internal/__lock.h | 8 ++++++++
src/internal/libc.h | 2 ++
src/thread/__lock.c | 17 +++++++++++++++++
3 files changed, 27 insertions(+)
diff --git a/src/internal/__lock.h b/src/internal/__lock.h
index f16d6176..0c2e9bfb 100644
--- a/src/internal/__lock.h
+++ b/src/internal/__lock.h
@@ -19,3 +19,11 @@ static inline void __unlock_fast(volatile int *l)
}
}
}
+
+static inline void __unlock_requeue_fast(volatile int *l, volatile int *r, int shared)
+{
+ extern void __unlock_requeue_slow(volatile int*, volatile int*, int);
+ /* We have to check if l[0] had been touched at all. */
+ if (l[0] < 0 && (a_fetch_add(l, -(INT_MIN + 1)) != (INT_MIN + 1)))
+ __unlock_requeue_slow(l, r, shared);
+}
diff --git a/src/internal/libc.h b/src/internal/libc.h
index a594d0c5..999b4cba 100644
--- a/src/internal/libc.h
+++ b/src/internal/libc.h
@@ -50,6 +50,8 @@ extern char *__progname, *__progname_full;
void __lock_slow(volatile int *, int) ATTR_LIBC_VISIBILITY;
void __lock(volatile int *) ATTR_LIBC_VISIBILITY;
void __unlock(volatile int *) ATTR_LIBC_VISIBILITY;
+void __unlock_requeue(volatile int *, volatile int *, int) ATTR_LIBC_VISIBILITY;
+void __unlock_requeue_slow(volatile int *, volatile int *, int) ATTR_LIBC_VISIBILITY;
int __lockfile(FILE *) ATTR_LIBC_VISIBILITY;
void __unlockfile(FILE *) ATTR_LIBC_VISIBILITY;
#define LOCK(x) __lock(x)
diff --git a/src/thread/__lock.c b/src/thread/__lock.c
index a3d8d4d0..b0b0c67c 100644
--- a/src/thread/__lock.c
+++ b/src/thread/__lock.c
@@ -19,6 +19,23 @@
weak_alias(__lock_fast, __lock);
weak_alias(__unlock_fast, __unlock);
+weak_alias(__unlock_requeue_fast, __unlock_requeue);
+
+/* __unlock_requeue handles the case where we have a second futex
+ * queue to which the a waiting thread should be scheduled if shared
+ * is false. This is e.g the case when l is the futex of a condition
+ * variable and r is the one of a private mutex: if the mutex is
+ * private, it is more efficient to just insert the thread in the
+ * queue of that mutex, instead of waking him up and put it back to
+ * sleep immediately when trying to lock the mutex. */
+
+void __unlock_requeue_slow(volatile int *l, volatile int *r, int shared)
+{
+ if (shared) __wake(l, 1, 1);
+ else __syscall(SYS_futex, l, FUTEX_REQUEUE|FUTEX_PRIVATE, 0, 1, r) != -ENOSYS
+ || __syscall(SYS_futex, l, FUTEX_REQUEUE, 0, 1, r);
+}
+
void __lock_slow(volatile int *l, int current)
{
--
2.15.1
next prev parent reply other threads:[~2018-01-03 13:17 UTC|newest]
Thread overview: 13+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-01-03 13:20 [PATCH 0/7] V3 of the new lock algorithm Jens Gustedt
2018-01-03 13:17 ` Jens Gustedt [this message]
2018-01-03 13:17 ` [PATCH 1/7] a new lock algorithm with lock value and congestion count in the same atomic int Jens Gustedt
2018-01-03 13:17 ` [PATCH 3/7] revise the definition of multiple basic locks in the code Jens Gustedt
2018-01-03 13:17 ` [PATCH 2/7] consistently use the LOCK an UNLOCK macros Jens Gustedt
2018-01-03 13:17 ` [PATCH 7/7] implement the local lock for condition variables with the new lock feature Jens Gustedt
2018-01-03 13:17 ` [PATCH 5/7] use the new lock algorithm for malloc Jens Gustedt
2018-01-09 17:42 ` Rich Felker
2018-01-09 18:58 ` Jens Gustedt
2018-01-09 19:26 ` Rich Felker
2018-09-18 19:23 ` Rich Felker
2018-01-03 13:17 ` [PATCH 4/7] separate the fast parts of __lock and __unlock into a .h file that may be used by other TU Jens Gustedt
2018-01-09 17:41 ` Rich Felker
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=498d773729dc31d4e9cb0e58308254f2c4332d20.1514985618.git.Jens.Gustedt@inria.fr \
--to=jens.gustedt@inria.fr \
--cc=musl@lists.openwall.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
Code repositories for project(s) associated with this public inbox
https://git.vuxu.org/mirror/musl/
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).