Date: Thu, 10 Apr 2014 23:29:20 +0400 From: Gleb Smirnoff <glebius@FreeBSD.org> To: John Baldwin <jhb@FreeBSD.org>, Julien Charbon <jcharbon@verisign.com> Cc: adrian@FreeBSD.org, src-committers@freebsd.org, svn-src-all@freebsd.org, rrs@FreeBSD.org, rwatson@FreeBSD.org, svn-src-head@freebsd.org Subject: Re: svn commit: r264321 - head/sys/netinet Message-ID: <20140410192920.GT44326@FreeBSD.org> In-Reply-To: <201404101815.s3AIFZx3065541@svn.freebsd.org> References: <201404101815.s3AIFZx3065541@svn.freebsd.org>
next in thread | previous in thread | raw e-mail | index | archive | help
Hi,
one comment:
On Thu, Apr 10, 2014 at 06:15:35PM +0000, John Baldwin wrote:
J> +/*
J> + * Drop a refcount on an tw elevated using tw_pcbref(). Return
J> + * the tw lock released.
J> + */
J> +static int
J> +tw_pcbrele(struct tcptw *tw)
J> +{
J> +
J> + TW_WLOCK_ASSERT(V_tw_lock);
J> + KASSERT(tw->tw_refcount > 0, ("%s: refcount 0", __func__));
J> +
J> + if (!refcount_release(&tw->tw_refcount)) {
J> + TW_WUNLOCK(V_tw_lock);
J> + return (0);
J> + }
J> +
J> + uma_zfree(V_tcptw_zone, tw);
J> + TW_WUNLOCK(V_tw_lock);
J> + return (1);
J> +}
We can unlock before uma_zfree(), that would reduce contention.
Why do we need the lock in this function at all? We don't manipulate
the TAILQ.
--
Totus tuus, Glebius.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20140410192920.GT44326>