linux-fsdevel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Andi Kleen <andi@firstfloor.org>
To: Waiman Long <Waiman.Long@hp.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>,
	Jeff Layton <jlayton@redhat.com>,
	Miklos Szeredi <mszeredi@suse.cz>, Ingo Molnar <mingo@redhat.com>,
	linux-fsdevel@vger.kernel.org, linux-kernel@vger.kernel.org,
	Linus Torvalds <torvalds@linux-foundation.org>,
	Benjamin Herrenschmidt <benh@kernel.crashing.org>,
	Andi Kleen <andi@firstfloor.org>,
	"Chandramouleeswaran, Aswin" <aswin@hp.com>,
	"Norton, Scott J" <scott.norton@hp.com>
Subject: Re: [PATCH v2 1/2] spinlock: New spinlock_refcount.h for lockless update of refcount
Date: Wed, 26 Jun 2013 22:17:13 +0200	[thread overview]
Message-ID: <20130626201713.GH6123@two.firstfloor.org> (raw)
In-Reply-To: <1372268603-46748-2-git-send-email-Waiman.Long@hp.com>

> + * The combined data structure is 8-byte aligned. So proper placement of this
> + * structure in the larger embedding data structure is needed to ensure that
> + * there is no hole in it.

On i386 u64 is only 4 bytes aligned. So you need to explicitely align
it to 8 bytes. Otherwise you risk the two members crossing a cache line, which
would be really expensive with atomics.

> +	/*
> +	 * Code doesn't work if raw spinlock is larger than 4 bytes
> +	 * or is empty.
> +	 */
> +	BUG_ON((sizeof(arch_spinlock_t) > 4) || (sizeof(arch_spinlock_t) == 0));

BUILD_BUG_ON 

> +
> +	spin_unlock_wait(plock);	/* Wait until lock is released */
> +	old.__lock_count = ACCESS_ONCE(*plockcnt);
> +	get_lock = ((threshold >= 0) && (old.count == threshold));
> +	if (likely(!get_lock && spin_can_lock(&old.lock))) {

What is that for? Why can't you do the CMPXCHG unconditially ?

If it's really needed, it is most likely a race?

The duplicated code should be likely an inline.

> +/*
> + * The presence of either one of the CONFIG_DEBUG_SPINLOCK or
> + * CONFIG_DEBUG_LOCK_ALLOC configuration parameter will force the
> + * spinlock_t structure to be 8-byte aligned.
> + *
> + * To support the spinlock/reference count combo data type for 64-bit SMP
> + * environment with spinlock debugging turned on, the reference count has
> + * to be integrated into the spinlock_t data structure in this special case.
> + * The spinlock_t data type will be 8 bytes larger if CONFIG_GENERIC_LOCKBREAK
> + * is also defined.

I would rather just disable the optimization when these CONFIGs are set

-Andi
-- 
ak@linux.intel.com -- Speaking for myself only.

  reply	other threads:[~2013-06-26 20:17 UTC|newest]

Thread overview: 28+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2013-06-26 17:43 [PATCH v2 0/2] Lockless update of reference count protected by spinlock Waiman Long
2013-06-26 17:43 ` [PATCH v2 1/2] spinlock: New spinlock_refcount.h for lockless update of refcount Waiman Long
2013-06-26 20:17   ` Andi Kleen [this message]
2013-06-26 21:07     ` Waiman Long
2013-06-26 21:22       ` Andi Kleen
2013-06-26 23:26         ` Waiman Long
2013-06-27  1:06           ` Andi Kleen
2013-06-27  1:15             ` Waiman Long
2013-06-27  1:24               ` Waiman Long
2013-06-27  1:37                 ` Andi Kleen
2013-06-27 14:56                   ` Waiman Long
2013-06-28 13:46                     ` Thomas Gleixner
2013-06-29 20:30                       ` Waiman Long
2013-06-26 23:27         ` Thomas Gleixner
2013-06-26 23:06   ` Thomas Gleixner
2013-06-27  0:16     ` Waiman Long
2013-06-27 14:44       ` Thomas Gleixner
2013-06-29 21:03         ` Waiman Long
2013-06-27  0:26     ` Waiman Long
2013-06-29 17:45   ` Linus Torvalds
2013-06-29 20:23     ` Waiman Long
2013-06-29 21:34       ` Waiman Long
2013-06-29 22:11         ` Linus Torvalds
2013-06-29 22:34           ` Waiman Long
2013-06-29 21:58       ` Linus Torvalds
2013-06-29 22:47         ` Linus Torvalds
2013-07-01 13:40           ` Waiman Long
2013-06-26 17:43 ` [PATCH v2 2/2] dcache: Locklessly update d_count whenever possible Waiman Long

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=20130626201713.GH6123@two.firstfloor.org \
    --to=andi@firstfloor.org \
    --cc=Waiman.Long@hp.com \
    --cc=aswin@hp.com \
    --cc=benh@kernel.crashing.org \
    --cc=jlayton@redhat.com \
    --cc=linux-fsdevel@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=mingo@redhat.com \
    --cc=mszeredi@suse.cz \
    --cc=scott.norton@hp.com \
    --cc=torvalds@linux-foundation.org \
    --cc=viro@zeniv.linux.org.uk \
    /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).