Skip site navigation (1)Skip section navigation (2)
Date:      Fri, 3 Jan 2003 23:38:29 +1100 (EST)
From:      Bruce Evans <bde@zeta.org.au>
To:        phk@freebsd.org
Cc:        cvs-committers@freebsd.org, <cvs-all@freebsd.org>
Subject:   Re: cvs commit: src/sys/sys _mutex.h 
Message-ID:  <20030103230420.W3339-100000@gamplex.bde.org>
In-Reply-To: <10907.1041590410@critter.freebsd.dk>

next in thread | previous in thread | raw e-mail | index | archive | help
On Fri, 3 Jan 2003 phk@freebsd.org wrote:

> In message <20030103212617.K2888-100000@gamplex.bde.org>, Bruce Evans writes:
> >On Thu, 2 Jan 2003, Poul-Henning Kamp wrote:
> >>   MFC: Conditionalize 16 bytes of struct mutex on MUTEX_PROFILING
> >
> >Please back this out.  It is contentious even in -current.  Even its
> >misleading comment hasn't been fixed.
>
> I'll await re@'s decision.  It is necessary to prevent -current and
> 5.0-RELENG machines from running out of kmem_map space and as I
> read John, that takes precedence.

Really?  It should make more than a few percent difference.  E.g., it
bloats the size of a vnode from 284 to 300 on i386's (6%).  This isn't
all that small for a struct that has so many instances, but if we cared
about the size of a vnode then we wouldn't be using so many mutexes
and locks in it.  Mutexes without profiling have size 52 on i386's,
so using a mutex instead of a simplelock for v_interlock bloats the
size of a vnode from 236 to 284 (20%).

> I agree that we should probably fix the comment and remove
> MUTEX_PROFILING from conf/options to prevent unintended use.

That would make it a non-option.

Several other options have similar problems.  A quick grep in <sys>
shows:
- the non-option USE_BUFHASH modifies struct buf.
- ...
- the option DEBUG_LOCKS modifies struct lockmgr and struct vnode.

Bruce


To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe cvs-all" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20030103230420.W3339-100000>