From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1752361AbbCLFL2 (ORCPT ); Thu, 12 Mar 2015 01:11:28 -0400 Received: from mail-pd0-f181.google.com ([209.85.192.181]:46747 "EHLO mail-pd0-f181.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751565AbbCLFLY (ORCPT ); Thu, 12 Mar 2015 01:11:24 -0400 Date: Thu, 12 Mar 2015 14:11:15 +0900 From: Minchan Kim To: Sergey Senozhatsky Cc: Sergey Senozhatsky , Andrew Morton , Nitin Gupta , linux-kernel@vger.kernel.org Subject: Re: [PATCH 0/6] new zram statistics reporting scheme Message-ID: <20150312051106.GA20144@blaptop> References: <1426000114-4471-1-git-send-email-sergey.senozhatsky@gmail.com> <20150312015558.GG10820@blaptop> <20150312021609.GA3508@swordfish> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20150312021609.GA3508@swordfish> User-Agent: Mutt/1.5.23 (2014-03-12) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Thu, Mar 12, 2015 at 11:16:09AM +0900, Sergey Senozhatsky wrote: > On (03/12/15 10:55), Minchan Kim wrote: > > I really appreciate you enhance stat functions, esp, working with iostat! > > thanks! my pleasure. > > > One thing I want to discuss is sometime we could remove RO fields > > in /sys/block/zram/ but we couldn't remove RW fields because > > io_stat/mm_stat doesn't have any writable option now so users will > > have two options to read stat. For exmaple, > > I played with CONFIG_ZRAM_OLD_STATS option, which turns RW attrs into > WO attrs (where possible/needed). but it turned out to be a rather > ugly patch and I eventually decided that I don't want to have these > #ifdef-s in zram code for the next two years. so providing both RW/RO > old stats (with a warning in the logs) and RO [mm|io]_stat sound like > a better plan. I think we don't need CONFIG_ZRAM_OLD_STATS. For example, for mem_used_max, we could add pr_warn_once in *mem_used_max_show* but not *mem_used_max_store*. so, old users will see deprecated message when they try to *read* the vaule old stat while they will write new value. One or two year later, we could remove mem_used_max_show then everyone should read the vaule /sys/block/zram0/mm_stat. > > > cat /sys/block/zram/mem_used_max > > cat /sys/block/zram/mm_stat | awk friend > > > > How about changing only writeable, not readable for duplicated stats > > in /sys/block/zram? So, user will have writeable stat to set some > > options in /sys/block/zram and readable stat to get some data in > > /sys/block/zram/[io|mm]_stat if the stat is duplicated in both. > > Sorry, I probably didn't drink enough coffee today, can you please > rephrase or give a trivial example? Sorry, I was vague. I meant users can read the vaule for mem_used_max and mem_limit in two places(ie, /sys/block/zram/ and /sys/block/zram/mm_stat). If we feel it's handy for user, i am not against that. But if there is no reason, I want to make /sys/block/zram/ stat write-only if it is possible stat like mem_used_max. I guess It's easier rule for user. > > -ss -- Kind regards, Minchan Kim