From: Ingo Molnar <mingo@elte.hu>
To: Linus Torvalds <torvalds@osdl.org>
Cc: Peter Chubb <peterc@gelato.unsw.edu.au>,
Chris Wedgwood <cw@f00f.org>, Andrew Morton <akpm@osdl.org>,
paulus@samba.org, linux-kernel@vger.kernel.org,
tony.luck@intel.com, dsw@gelato.unsw.edu.au,
benh@kernel.crashing.org, linux-ia64@vger.kernel.org,
hch@infradead.org, wli@holomorphy.com, jbarnes@sgi.com
Subject: [patch 1/3] spinlock fix #1, *_can_lock() primitives
Date: Thu, 20 Jan 2005 17:08:39 +0100 [thread overview]
Message-ID: <20050120160839.GA13067@elte.hu> (raw)
In-Reply-To: <Pine.LNX.4.58.0501200747230.8178@ppc970.osdl.org>
* Linus Torvalds <torvalds@osdl.org> wrote:
> We have a sense problem with the "trylock()" cases - some return "it
> was locked" (semaphores), and some return "I succeeded" (spinlocks),
> so not only is the sense not immediately obvious from the usage, it's
> actually _different_ for semaphores and for spinlocks.
well, this is primarily a problem of the semaphore primitives.
anyway, here's my first patch again, with s/trylock_test/can_lock/.
Ingo
--
it fixes the BUILD_LOCK_OPS() bug by introducing the following 3 new
locking primitives:
spin_can_lock(lock)
read_can_lock(lock)
write_can_lock(lock)
this is what is needed by BUILD_LOCK_OPS(): a nonintrusive test to check
whether the real (intrusive) trylock op would succeed or not. Semantics
and naming is completely symmetric to the trylock counterpart. No
changes to exit.c.
build/boot-tested on x86. Architectures that want to support PREEMPT
need to add these definitions.
Signed-off-by: Ingo Molnar <mingo@elte.hu>
--- linux/kernel/spinlock.c.orig
+++ linux/kernel/spinlock.c
@@ -173,8 +173,8 @@ EXPORT_SYMBOL(_write_lock);
* (We do this in a function because inlining it would be excessive.)
*/
-#define BUILD_LOCK_OPS(op, locktype, is_locked_fn) \
-void __lockfunc _##op##_lock(locktype *lock) \
+#define BUILD_LOCK_OPS(op, locktype) \
+void __lockfunc _##op##_lock(locktype##_t *lock) \
{ \
preempt_disable(); \
for (;;) { \
@@ -183,7 +183,7 @@ void __lockfunc _##op##_lock(locktype *l
preempt_enable(); \
if (!(lock)->break_lock) \
(lock)->break_lock = 1; \
- while (is_locked_fn(lock) && (lock)->break_lock) \
+ while (!op##_can_lock(lock) && (lock)->break_lock) \
cpu_relax(); \
preempt_disable(); \
} \
@@ -191,7 +191,7 @@ void __lockfunc _##op##_lock(locktype *l
\
EXPORT_SYMBOL(_##op##_lock); \
\
-unsigned long __lockfunc _##op##_lock_irqsave(locktype *lock) \
+unsigned long __lockfunc _##op##_lock_irqsave(locktype##_t *lock) \
{ \
unsigned long flags; \
\
@@ -205,7 +205,7 @@ unsigned long __lockfunc _##op##_lock_ir
preempt_enable(); \
if (!(lock)->break_lock) \
(lock)->break_lock = 1; \
- while (is_locked_fn(lock) && (lock)->break_lock) \
+ while (!op##_can_lock(lock) && (lock)->break_lock) \
cpu_relax(); \
preempt_disable(); \
} \
@@ -214,14 +214,14 @@ unsigned long __lockfunc _##op##_lock_ir
\
EXPORT_SYMBOL(_##op##_lock_irqsave); \
\
-void __lockfunc _##op##_lock_irq(locktype *lock) \
+void __lockfunc _##op##_lock_irq(locktype##_t *lock) \
{ \
_##op##_lock_irqsave(lock); \
} \
\
EXPORT_SYMBOL(_##op##_lock_irq); \
\
-void __lockfunc _##op##_lock_bh(locktype *lock) \
+void __lockfunc _##op##_lock_bh(locktype##_t *lock) \
{ \
unsigned long flags; \
\
@@ -246,9 +246,9 @@ EXPORT_SYMBOL(_##op##_lock_bh)
* _[spin|read|write]_lock_irqsave()
* _[spin|read|write]_lock_bh()
*/
-BUILD_LOCK_OPS(spin, spinlock_t, spin_is_locked);
-BUILD_LOCK_OPS(read, rwlock_t, rwlock_is_locked);
-BUILD_LOCK_OPS(write, rwlock_t, spin_is_locked);
+BUILD_LOCK_OPS(spin, spinlock);
+BUILD_LOCK_OPS(read, rwlock);
+BUILD_LOCK_OPS(write, rwlock);
#endif /* CONFIG_PREEMPT */
--- linux/include/linux/spinlock.h.orig
+++ linux/include/linux/spinlock.h
@@ -584,4 +584,10 @@ static inline int bit_spin_is_locked(int
#define DEFINE_SPINLOCK(x) spinlock_t x = SPIN_LOCK_UNLOCKED
#define DEFINE_RWLOCK(x) rwlock_t x = RW_LOCK_UNLOCKED
+/**
+ * spin_can_lock - would spin_trylock() succeed?
+ * @lock: the spinlock in question.
+ */
+#define spin_can_lock(lock) (!spin_is_locked(lock))
+
#endif /* __LINUX_SPINLOCK_H */
--- linux/include/asm-i386/spinlock.h.orig
+++ linux/include/asm-i386/spinlock.h
@@ -188,6 +188,18 @@ typedef struct {
#define rwlock_is_locked(x) ((x)->lock != RW_LOCK_BIAS)
+/**
+ * read_can_lock - would read_trylock() succeed?
+ * @lock: the rwlock in question.
+ */
+#define read_can_lock(x) (atomic_read((atomic_t *)&(x)->lock) > 0)
+
+/**
+ * write_can_lock - would write_trylock() succeed?
+ * @lock: the rwlock in question.
+ */
+#define write_can_lock(x) ((x)->lock == RW_LOCK_BIAS)
+
/*
* On x86, we implement read-write locks as a 32-bit counter
* with the high bit (sign) being the "contended" bit.
next prev parent reply other threads:[~2005-01-20 16:13 UTC|newest]
Thread overview: 48+ messages / expand[flat|nested] mbox.gz Atom feed top
2005-01-17 5:50 Horrible regression with -CURRENT from "Don't busy-lock-loop in preemptable spinlocks" patch Chris Wedgwood
2005-01-17 7:09 ` Andrew Morton
2005-01-17 7:33 ` Chris Wedgwood
2005-01-17 7:50 ` Paul Mackerras
2005-01-17 8:00 ` Chris Wedgwood
2005-01-17 14:33 ` Ingo Molnar
2005-01-18 1:47 ` Darren Williams
2005-01-18 4:28 ` Darren Williams
2005-01-18 7:08 ` Chris Wedgwood
2005-01-19 0:14 ` Peter Chubb
2005-01-19 8:04 ` Ingo Molnar
2005-01-19 9:18 ` Peter Chubb
2005-01-19 9:20 ` Ingo Molnar
2005-01-19 21:43 ` Paul Mackerras
2005-01-20 2:34 ` [PATCH RFC] 'spinlock/rwlock fixes' V3 [1/1] Chris Wedgwood
2005-01-20 3:01 ` Andrew Morton
2005-01-20 3:18 ` Chris Wedgwood
2005-01-20 3:33 ` Andrew Morton
2005-01-20 8:59 ` Peter Chubb
2005-01-20 13:04 ` Ingo Molnar
2005-01-20 15:51 ` Linus Torvalds
2005-01-20 16:08 ` Ingo Molnar [this message]
2005-01-20 16:11 ` [patch 2/3] spinlock fix #2: generalize [spin|rw]lock yielding Ingo Molnar
2005-01-20 16:12 ` [patch 3/3] spinlock fix #3: type-checking spinlock primitives, x86 Ingo Molnar
2005-01-20 16:14 ` [patch] stricter type-checking rwlock " Ingo Molnar
2005-01-20 16:16 ` [patch] minor spinlock cleanups Ingo Molnar
2005-01-20 16:31 ` [patch 1/3] spinlock fix #1, *_can_lock() primitives Linus Torvalds
2005-01-20 16:40 ` Ingo Molnar
2005-01-20 17:48 ` Linus Torvalds
2005-01-20 17:53 ` Ingo Molnar
2005-01-20 18:22 ` [patch, BK-curr] nonintrusive spin-polling loop in kernel/spinlock.c Ingo Molnar
2005-01-20 18:25 ` [patch, BK-curr] rename 'lock' to 'slock' in asm-i386/spinlock.h Ingo Molnar
2005-01-20 23:45 ` [patch, BK-curr] nonintrusive spin-polling loop in kernel/spinlock.c Linus Torvalds
2005-01-20 16:44 ` [patch 1/3] spinlock fix #1, *_can_lock() primitives Ingo Molnar
2005-01-20 16:59 ` Ingo Molnar
2005-01-20 16:47 ` Ingo Molnar
2005-01-20 16:57 ` Ingo Molnar
2005-01-20 16:05 ` [PATCH RFC] 'spinlock/rwlock fixes' V3 [1/1] Linus Torvalds
2005-01-20 16:20 ` Ingo Molnar
2005-01-20 16:18 ` Linus Torvalds
2005-01-20 16:23 ` Ingo Molnar
2005-01-20 17:30 ` Linus Torvalds
2005-01-20 17:38 ` Ingo Molnar
2005-01-20 16:28 ` Ingo Molnar
2005-01-20 5:49 ` Horrible regression with -CURRENT from "Don't busy-lock-loop in preemptable spinlocks" patch Grant Grundler
2005-01-17 7:38 ` [PATCH] __get_cpu_var should use __smp_processor_id() not smp_processor_id() Chris Wedgwood
2005-01-17 14:40 ` Ingo Molnar
2005-01-17 18:53 ` Chris Wedgwood
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=20050120160839.GA13067@elte.hu \
--to=mingo@elte.hu \
--cc=akpm@osdl.org \
--cc=benh@kernel.crashing.org \
--cc=cw@f00f.org \
--cc=dsw@gelato.unsw.edu.au \
--cc=hch@infradead.org \
--cc=jbarnes@sgi.com \
--cc=linux-ia64@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=paulus@samba.org \
--cc=peterc@gelato.unsw.edu.au \
--cc=tony.luck@intel.com \
--cc=torvalds@osdl.org \
--cc=wli@holomorphy.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.
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).