From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1757037AbcASXij (ORCPT ); Tue, 19 Jan 2016 18:38:39 -0500 Received: from gum.cmpxchg.org ([85.214.110.215]:53652 "EHLO gum.cmpxchg.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1757156AbcASXi3 (ORCPT ); Tue, 19 Jan 2016 18:38:29 -0500 Date: Tue, 19 Jan 2016 18:38:22 -0500 From: Johannes Weiner To: Andrew Morton Cc: Shaohua Li , Siddhesh Poyarekar , linux-mm@kvack.org, linux-kernel@vger.kernel.org, kernel-team@fb.com Subject: Re: [PATCH] proc: revert /proc//maps [stack:TID] annotation Message-ID: <20160119233822.GA10788@cmpxchg.org> References: <1453226559-17322-1-git-send-email-hannes@cmpxchg.org> <20160119141430.8ff9c464.akpm@linux-foundation.org> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20160119141430.8ff9c464.akpm@linux-foundation.org> User-Agent: Mutt/1.5.24 (2015-08-30) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Tue, Jan 19, 2016 at 02:14:30PM -0800, Andrew Morton wrote: > On Tue, 19 Jan 2016 13:02:39 -0500 Johannes Weiner wrote: > > > b764375 ("procfs: mark thread stack correctly in proc//maps") > > added [stack:TID] annotation to /proc//maps. Finding the task of > > a stack VMA requires walking the entire thread list, turning this into > > quadratic behavior: a thousand threads means a thousand stacks, so the > > rendering of /proc//maps needs to look at a million threads. The > > cost is not in proportion to the usefulness as described in the patch. > > > > Drop the [stack:TID] annotation to make /proc//maps (and > > /proc//numa_maps) usable again for higher thread counts. > > > > The [stack] annotation inside /proc//task//maps is retained, > > as identifying the stack VMA there is an O(1) operation. > > Four years ago, ouch. > > Any thoughts on the obvious back-compatibility concerns? ie, why did > Siddhesh implement this in the first place? My bad for not ensuring > that the changelog told us this. I thought about storing the TID of the thread using the VMA as the stack directly inside vm_area_struct; maybe using vm_private_data? However, that's a bit of work and ugliness that I wouldn't want to commit to until we know that people ended up using this in practice. > I note that this patch is a partial revert - the smaps and numa_maps > parts of b764375 remain in place. What's up with that? I left the stack annotations in the thread-specific files because that sounds useful and is cheap enough - we only have to test the vma range against that thread's stack pointer. The last changelog paragraph says that for maps, I'll update it to include smaps and numa_maps.