[patch] lockdep: fix seqlock_init()

[patch] lockdep: fix seqlock_init()

[Ingo Molnar]

Subject: [patch] lockdep: fix seqlock_init()
From: Ingo Molnar <mingo@elte.hu>

seqlock_init() needs to use spin_lock_init() for dynamic locks, so that 
lockdep is notified about the presence of a new lock.

(this is a fallout of the recent networking merge, which started using 
the so-far unused seqlock_init() API.)

This fix solves the following lockdep-internal warning on current -git:

 INFO: trying to register non-static key.
 the code is fine but needs lockdep annotation.
 turning off the locking correctness validator.
  [<c0104fd9>] dump_trace+0x63/0x1e8
  [<c0105177>] show_trace_log_lvl+0x19/0x2e
  [<c010557e>] show_trace+0x12/0x14
  [<c0105594>] dump_stack+0x14/0x16
  [<c0142992>] __lock_acquire+0x10c/0x9f9
  [<c0143564>] lock_acquire+0x56/0x72
  [<c03c514f>] _spin_lock+0x35/0x42
  [<c0369875>] neigh_destroy+0x9d/0x12e
  [<c036a1d5>] neigh_periodic_timer+0x10a/0x15c
  [<c01302a5>] run_timer_softirq+0x126/0x18e
  [<c012c530>] __do_softirq+0x6b/0xe6
  [<c0106404>] do_softirq+0x64/0xd2
  [<c012c249>] ksoftirqd+0x82/0x138
  [<c01398f1>] kthread+0xb2/0xd7
  [<c0104c1b>] kernel_thread_helper+0x7/0x10

Signed-off-by: Ingo Molnar <mingo@elte.hu>
---
 include/linux/seqlock.h |    7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

Index: linux-hres-timers.q/include/linux/seqlock.h
===================================================================
--- linux-hres-timers.q.orig/include/linux/seqlock.h
+++ linux-hres-timers.q/include/linux/seqlock.h
@@ -44,8 +44,11 @@ typedef struct {
 #define SEQLOCK_UNLOCKED \
 		 __SEQLOCK_UNLOCKED(old_style_seqlock_init)
 
-#define seqlock_init(x) \
-		do { *(x) = (seqlock_t) __SEQLOCK_UNLOCKED(x); } while (0)
+#define seqlock_init(x)					\
+	do {						\
+		(x)->sequence = 0;			\
+		spin_lock_init(&(x)->lock);		\
+	} while (0)
 
 #define DEFINE_SEQLOCK(x) \
 		seqlock_t x = __SEQLOCK_UNLOCKED(x)

Re: [patch] lockdep: fix seqlock_init()

[Andrew Morton]

On Tue, 12 Dec 2006 12:10:28 +0100
Ingo Molnar <mingo@elte.hu> wrote:

> +#define seqlock_init(x)					\
> +	do {						\
> +		(x)->sequence = 0;			\
> +		spin_lock_init(&(x)->lock);		\
> +	} while (0)

This does not have to be a macro, does it?

Re: [patch] lockdep: fix seqlock_init()

[Ingo Molnar]

* Andrew Morton <akpm@osdl.org> wrote:

> On Tue, 12 Dec 2006 12:10:28 +0100
> Ingo Molnar <mingo@elte.hu> wrote:
> 
> > +#define seqlock_init(x)					\
> > +	do {						\
> > +		(x)->sequence = 0;			\
> > +		spin_lock_init(&(x)->lock);		\
> > +	} while (0)
> 
> This does not have to be a macro, does it?

Maybe it could be an __always_inline inline function (it has to be 
inlined to get the callsite based lock class key right) - but i'm not 
sure about the include file dependencies. Will probably work out fine as 
seqlock.h is supposed to be a late one in the order of inclusion - but i 
didnt want to make a blind bet.

	Ingo

Re: [patch] lockdep: fix seqlock_init()

[Andrew Morton]

On Tue, 12 Dec 2006 21:50:01 +0100
Ingo Molnar <mingo@elte.hu> wrote:

> 
> * Andrew Morton <akpm@osdl.org> wrote:
> 
> > On Tue, 12 Dec 2006 12:10:28 +0100
> > Ingo Molnar <mingo@elte.hu> wrote:
> > 
> > > +#define seqlock_init(x)					\
> > > +	do {						\
> > > +		(x)->sequence = 0;			\
> > > +		spin_lock_init(&(x)->lock);		\
> > > +	} while (0)
> > 
> > This does not have to be a macro, does it?
> 
> Maybe it could be an __always_inline inline function (it has to be 
> inlined to get the callsite based lock class key right)

the compiler darn better inline it, else we'll have an out-of-line copy of
everything in everywhere.

> - but i'm not 
> sure about the include file dependencies. Will probably work out fine as 
> seqlock.h is supposed to be a late one in the order of inclusion - but i 
> didnt want to make a blind bet.

seqlock.h already includes spinlock.h.

Re: [patch] lockdep: fix seqlock_init()

[Ingo Molnar]

* Andrew Morton <akpm@osdl.org> wrote:

> > > > +#define seqlock_init(x)					\
> > > > +	do {						\
> > > > +		(x)->sequence = 0;			\
> > > > +		spin_lock_init(&(x)->lock);		\
> > > > +	} while (0)
> > > 
> > > This does not have to be a macro, does it?
> > 
> > Maybe it could be an __always_inline inline function (it has to be 
> > inlined to get the callsite based lock class key right)
> 
> the compiler darn better inline it, else we'll have an out-of-line 
> copy of everything in everywhere.

the compiler will do the uninlining happily if it sees a size advantage 
(when a single .c module calls the function several times), and creates 
a private per-object-file uninlined function. So an __always_inline 
would definitely be needed.

> > - but i'm not 
> > sure about the include file dependencies. Will probably work out fine as 
> > seqlock.h is supposed to be a late one in the order of inclusion - but i 
> > didnt want to make a blind bet.
> 
> seqlock.h already includes spinlock.h.

yes ... i just preserved the status quo.

	Ingo