Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S932352Ab3GCLBy (ORCPT ); Wed, 3 Jul 2013 07:01:54 -0400 Received: from relay.parallels.com ([195.214.232.42]:46434 "EHLO relay.parallels.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754786Ab3GCLBw (ORCPT ); Wed, 3 Jul 2013 07:01:52 -0400 Message-ID: <51D4047F.2010700@parallels.com> Date: Wed, 3 Jul 2013 15:01:19 +0400 From: Maxim Patlasov User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:17.0) Gecko/20130620 Thunderbird/17.0.7 MIME-Version: 1.0 To: Andrew Morton CC: , , , , , , , , , , , , , Subject: Re: [PATCH] mm: strictlimit feature -v2 References: <20130629174706.20175.78184.stgit@maximpc.sw.ru> <20130702174316.15075.84993.stgit@maximpc.sw.ru> <20130702123804.9f252487f86c12b0f4edee57@linux-foundation.org> In-Reply-To: <20130702123804.9f252487f86c12b0f4edee57@linux-foundation.org> Content-Type: text/plain; charset="UTF-8"; format=flowed Content-Transfer-Encoding: 8bit X-Originating-IP: [10.30.17.2] Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 4352 Lines: 126 07/02/2013 11:38 PM, Andrew Morton пишет: > On Tue, 02 Jul 2013 21:44:47 +0400 Maxim Patlasov wrote: > >> From: Miklos Szeredi >> >> The feature prevents mistrusted filesystems to grow a large number of dirty >> pages before throttling. For such filesystems balance_dirty_pages always >> check bdi counters against bdi limits. I.e. even if global "nr_dirty" is under >> "freerun", it's not allowed to skip bdi checks. The only use case for now is >> fuse: it sets bdi max_ratio to 1% by default and system administrators are >> supposed to expect that this limit won't be exceeded. >> >> The feature is on if address space is marked by AS_STRICTLIMIT flag. >> A filesystem may set the flag when it initializes a new inode. >> >> Changed in v2 (thanks to Andrew Morton): >> - added a few explanatory comments >> - cleaned up the mess in backing_dev_info foo_stamp fields: now it's clearly >> stated that bw_time_stamp is measured in jiffies; renamed other foo_stamp >> fields to reflect that they are in units of number-of-pages. >> > Better, thanks. > > The writeback arithemtic makes my head spin - I'd really like Fengguang > to go over this, please. > > A quick visit from the spelling police: Great! Thank you, Andrew. I'll wait for Fengguang' feedback for a while before respin. > >> ... >> >> @@ -41,8 +43,15 @@ typedef int (congested_fn)(void *, int); >> enum bdi_stat_item { >> BDI_RECLAIMABLE, >> BDI_WRITEBACK, >> - BDI_DIRTIED, >> - BDI_WRITTEN, >> + >> + /* >> + * The three counters below reflects number of events of specific type >> + * happened since bdi_init(). The type is defined in comments below: > "The three counters below reflect the number of events of specific > types since bdi_init()" > >> + */ >> + BDI_DIRTIED, /* a page was dirtied */ >> + BDI_WRITTEN, /* writeout completed for a page */ >> + BDI_WRITTEN_BACK, /* a page went to writeback */ >> + >> NR_BDI_STAT_ITEMS >> }; >> >> >> ... >> >> @@ -680,28 +712,55 @@ static unsigned long bdi_position_ratio(struct backing_dev_info *bdi, >> return 0; >> >> /* >> - * global setpoint >> + * The strictlimit feature is a tool preventing mistrusted filesystems >> + * to grow a large number of dirty pages before throttling. For such > "from growing" > >> + * filesystems balance_dirty_pages always checks bdi counters against >> + * bdi limits. Even if global "nr_dirty" is under "freerun". This is >> + * especially important for fuse who sets bdi->max_ratio to 1% by > s/who/which/ > >> + * default. Without strictlimit feature, fuse writeback may consume >> + * arbitrary amount of RAM because it is accounted in >> + * NR_WRITEBACK_TEMP which is not involved in calculating "nr_dirty". >> >> ... >> >> @@ -994,6 +1054,26 @@ static void bdi_update_dirty_ratelimit(struct backing_dev_info *bdi, >> * keep that period small to reduce time lags). >> */ >> step = 0; >> + >> + /* >> + * For strictlimit case, balanced_dirty_ratelimit was calculated > balance_dirty_ratelimit? > >> + * above based on bdi counters and limits (see bdi_position_ratio()). >> + * Hence, to calculate "step" properly, we have to use bdi_dirty as >> + * "dirty" and bdi_setpoint as "setpoint". >> + * >> + * We rampup dirty_ratelimit forcibly if bdi_dirty is low because >> + * it's possible that bdi_thresh is close to zero due to inactivity >> + * of backing device (see the implementation of bdi_dirty_limit()). >> + */ >> + if (unlikely(strictlimit)) { >> + dirty = bdi_dirty; >> + if (bdi_dirty < 8) >> + setpoint = bdi_dirty + 1; >> + else >> >> ... >> >> @@ -1057,18 +1140,32 @@ void __bdi_update_bandwidth(struct backing_dev_info *bdi, >> if (elapsed > HZ && time_before(bdi->bw_time_stamp, start_time)) >> goto snapshot; >> >> + /* >> + * Skip periods when backing dev was idle due to abscence of pages > "absence" > >> + * under writeback (when over_bground_thresh() returns false) >> + */ >> + if (test_bit(BDI_idle, &bdi->state) && >> + bdi->writeback_nr_stamp == writeback) >> >> ... >> > > -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/