[PATCH] Revert lockdep check in raw_seqcount_begin

[PATCH] Revert lockdep check in raw_seqcount_begin

[Trond Myklebust]

This commit reverts the addition of lockdep checking to raw_seqcount_begin
for the following reasons:

1) It violates the naming convention that raw_* functions should not
   do lockdep checks (a convention that is also followed by the other
   raw_*_seqcount_begin functions).
2) raw_seqcount_begin does not spin, so it can only be part of an ABBA
   deadlock in very special circumstances (for instance if a lock
   is held across the entire raw_seqcount_begin()+read_seqcount_retry()
   loop while also being taken inside the write_seqcount protected area).
3) It is causing false positives with some existing callers, and there
   is no non-lockdep alternative for those callers to use.

None of the three existing callers (__d_lookup_rcu, netdev_get_name, and
the NFS state code) appear to use the function in a manner that is ABBA
deadlock prone.

Fixes: 1ca7d67cf5d5: seqcount: Add lockdep functionality to seqcount/seqlock
Cc: John Stultz <john.stultz@linaro.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Link: http://lkml.kernel.org/r/CAHQdGtRR6SvEhXiqWo24hoUh9AU9cL82Z8Z-d8-7u951F_d+5g@mail.gmail.com
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
---
 include/linux/seqlock.h | 2 --
 1 file changed, 2 deletions(-)

diff --git a/include/linux/seqlock.h b/include/linux/seqlock.h
index 535f158977b9..8cf350325dc6 100644
--- a/include/linux/seqlock.h
+++ b/include/linux/seqlock.h
@@ -164,8 +164,6 @@ static inline unsigned read_seqcount_begin(const seqcount_t *s)
 static inline unsigned raw_seqcount_begin(const seqcount_t *s)
 {
 	unsigned ret = ACCESS_ONCE(s->sequence);
-
-	seqcount_lockdep_reader_access(s);
 	smp_rmb();
 	return ret & ~1;
 }
-- 
1.9.3

Re: [PATCH] Revert lockdep check in raw_seqcount_begin

[Peter Zijlstra]

On Thu, Jun 05, 2014 at 11:31:01AM -0400, Trond Myklebust wrote:
> This commit reverts the addition of lockdep checking to raw_seqcount_begin
> for the following reasons:
> 
> 1) It violates the naming convention that raw_* functions should not
>    do lockdep checks (a convention that is also followed by the other
>    raw_*_seqcount_begin functions).
> 2) raw_seqcount_begin does not spin, so it can only be part of an ABBA
>    deadlock in very special circumstances (for instance if a lock
>    is held across the entire raw_seqcount_begin()+read_seqcount_retry()
>    loop while also being taken inside the write_seqcount protected area).
> 3) It is causing false positives with some existing callers, and there
>    is no non-lockdep alternative for those callers to use.
> 
> None of the three existing callers (__d_lookup_rcu, netdev_get_name, and
> the NFS state code) appear to use the function in a manner that is ABBA
> deadlock prone.
> 
> Fixes: 1ca7d67cf5d5: seqcount: Add lockdep functionality to seqcount/seqlock
> Cc: John Stultz <john.stultz@linaro.org>
> Cc: Peter Zijlstra <peterz@infradead.org>
> Cc: Ingo Molnar <mingo@kernel.org>
> Cc: "David S. Miller" <davem@davemloft.net>
> Cc: Al Viro <viro@zeniv.linux.org.uk>
> Link: http://lkml.kernel.org/r/CAHQdGtRR6SvEhXiqWo24hoUh9AU9cL82Z8Z-d8-7u951F_d+5g@mail.gmail.com
> Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

Thanks Trond!

Re: [PATCH] Revert lockdep check in raw_seqcount_begin

[John Stultz]

On Thu, Jun 5, 2014 at 8:31 AM, Trond Myklebust
<trond.myklebust@primarydata.com> wrote:
> This commit reverts the addition of lockdep checking to raw_seqcount_begin
> for the following reasons:
>
> 1) It violates the naming convention that raw_* functions should not
>    do lockdep checks (a convention that is also followed by the other
>    raw_*_seqcount_begin functions).
> 2) raw_seqcount_begin does not spin, so it can only be part of an ABBA
>    deadlock in very special circumstances (for instance if a lock
>    is held across the entire raw_seqcount_begin()+read_seqcount_retry()
>    loop while also being taken inside the write_seqcount protected area).
> 3) It is causing false positives with some existing callers, and there
>    is no non-lockdep alternative for those callers to use.
>
> None of the three existing callers (__d_lookup_rcu, netdev_get_name, and
> the NFS state code) appear to use the function in a manner that is ABBA
> deadlock prone.
>
> Fixes: 1ca7d67cf5d5: seqcount: Add lockdep functionality to seqcount/seqlock
> Cc: John Stultz <john.stultz@linaro.org>
> Cc: Peter Zijlstra <peterz@infradead.org>
> Cc: Ingo Molnar <mingo@kernel.org>
> Cc: "David S. Miller" <davem@davemloft.net>
> Cc: Al Viro <viro@zeniv.linux.org.uk>
> Link: http://lkml.kernel.org/r/CAHQdGtRR6SvEhXiqWo24hoUh9AU9cL82Z8Z-d8-7u951F_d+5g@mail.gmail.com
> Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
> ---
>  include/linux/seqlock.h | 2 --
>  1 file changed, 2 deletions(-)
>
> diff --git a/include/linux/seqlock.h b/include/linux/seqlock.h
> index 535f158977b9..8cf350325dc6 100644
> --- a/include/linux/seqlock.h
> +++ b/include/linux/seqlock.h
> @@ -164,8 +164,6 @@ static inline unsigned read_seqcount_begin(const seqcount_t *s)
>  static inline unsigned raw_seqcount_begin(const seqcount_t *s)
>  {
>         unsigned ret = ACCESS_ONCE(s->sequence);
> -
> -       seqcount_lockdep_reader_access(s);
>         smp_rmb();
>         return ret & ~1;
>  }

Just back from vacation, sorry for the slow response. Thanks for
submitting this! Also sorry for my confusion mixing
raw_seqcount_begin() and raw_seqlock_begin() in our earlier discussion
(I blame the hectic vacation prep ;).

Acked-by: John Stultz <john.stultz@linaro.org>

thanks
-john

Re: [PATCH] Revert lockdep check in raw_seqcount_begin

[John Stultz]

On Wed, Jun 11, 2014 at 11:39 AM, John Stultz <john.stultz@linaro.org> wrote:
> Just back from vacation, sorry for the slow response. Thanks for
> submitting this! Also sorry for my confusion mixing
> raw_seqcount_begin() and raw_seqlock_begin() in our earlier discussion

Bah. Not raw_seqlock_begin.. I meant raw_read_seqcount_begin().
-john