From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1752740AbaEBVZJ (ORCPT ); Fri, 2 May 2014 17:25:09 -0400 Received: from mga02.intel.com ([134.134.136.20]:53419 "EHLO mga02.intel.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751501AbaEBVZI (ORCPT ); Fri, 2 May 2014 17:25:08 -0400 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="4.97,974,1389772800"; d="scan'208";a="505094135" Subject: Re: [PATCH] rwsem: Comments to explain the meaning of the rwsem's count field From: Tim Chen To: Peter Hurley Cc: Ingo Molnar , Peter Zijlstra , Andrew Morton , Davidlohr Bueso , Alex Shi , Andi Kleen , Michel Lespinasse , Rik van Riel , Thomas Gleixner , "Paul E.McKenney" , Jason Low , linux-kernel@vger.kernel.org In-Reply-To: <5363F969.5060601@hurleysoftware.com> References: <1398966637.2970.129.camel@schen9-DESK> <5362AC24.9050508@hurleysoftware.com> <1398985513.2970.141.camel@schen9-DESK> <5363F969.5060601@hurleysoftware.com> Content-Type: text/plain; charset="UTF-8" Date: Fri, 02 May 2014 14:25:00 -0700 Message-ID: <1399065900.2970.182.camel@schen9-DESK> Mime-Version: 1.0 X-Mailer: Evolution 2.32.3 (2.32.3-1.fc14) Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org > > /* > > + * Guide to the rw_semaphore's count field for common values. > > + * (32 bit case illustrated, similar for 64 bit) > > The values below are x86-specific; the actual defines are arch-dependent. > Do other archs use different values? This is also the value used also for generic case. I don't see other arch specific values defined. > > > + * > > + * 0x0000000X (1) X readers active or attempting lock, no writer waiting > > + * X = #active_readers + #readers attempting to lock > > + * (X*ACTIVE_BIAS) > > Not sure it matters, but maybe you want to note that it's possible for 0 readers > to be active with this value, and all of the other readers may have initially > failed to claim the lock but may be successful if one can claim the wait_lock while > the count is still > 0. I'll add the explanation for the down_read_failed scenario in the note section below. > > > + * > > + * 0x00000000 rwsem is unlocked, and no one is waiting for the lock or > > + * attempting to read lock or write lock. > > + * > > + * 0xffff000X (1) X readers active or attempt lock, there are waiters for lock > > + * X = #active readers + # readers attempting lock > > + * (X*ACTIVE_BIAS + WAITING_BIAS) > > + * (2) 1 writer attempting lock, no waiters for lock > > + * X-1 = #active readers + #readers attempting lock > > + * ((X-1)*ACTIVE_BIAS + ACTIVE_WRITE_BIAS) > > + * (3) 1 writer active, no waiters for lock > > + * X-1 = #active readers + #readers attempting lock > > + * ((X-1)*ACTIVE_BIAS + ACTIVE_WRITE_BIAS) > > + * > > + * 0xffff0001 (1) 1 reader active or attempting lock, waiters for lock > > + * (WAITING_BIAS + ACTIVE_BIAS) > > + * (2) 1 writer active or attempt lock, no waiters for lock > > + * (ACTIVE_BIAS + ACTIVE_WRITE_BIAS) > > + * > > + * 0xffff0000 (1) There are writers or readers queued but none active > > + * or in the process of attempting lock. > > + * (WAITING_BIAS) > > + * Note: writer can attempt to steal lock for this count by adding > > + * ACTIVE_WRITE_BIAS in cmpxchg and checking the old count > > + * > > + * 0xfffe0001 (1) 1 writer active, or attempting lock. Waiters on queue. > > + * (ACTIVE_WRITE_BIAS + WAITING_BIAS) > > The count can have more values than just 0xfffe0001 because multiple > failed write lock attempts plus failed read lock attempts can produce other > values than those listed. You're correct. The values are not comprehensive. I tried to show the common ones and how they arose. How about I replace the 0xfffe0001 case with count < WAITING_BIAS (1) X writer active, Y writers attempting lock, Z readers attempting lock, no waiters where X = 0 or 1, (X+Y) >= 2, Z >= 0 (X+Y) * ACTIVE_WRITE_BIAS + Z * ACTIVE_BIAS (2) X writer active, Y writers attempting lock, Z readers attempting lock, with waiters where X = 0 or 1, (X+Y) >= 1, Z >= 0 (X+Y) * ACTIVE_WRITE_BIAS + Z * ACTIVE_BIAS + WAITING_BIAS Thanks. Tim