From patchwork Thu Nov 17 21:23:10 2016 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Jens Axboe X-Patchwork-Id: 9435343 Return-Path: Received: from mail.wl.linuxfoundation.org (pdx-wl-mail.web.codeaurora.org [172.30.200.125]) by pdx-korg-patchwork.web.codeaurora.org (Postfix) with ESMTP id 4F99560471 for ; Thu, 17 Nov 2016 21:23:16 +0000 (UTC) Received: from mail.wl.linuxfoundation.org (localhost [127.0.0.1]) by mail.wl.linuxfoundation.org (Postfix) with ESMTP id 3F751296DB for ; Thu, 17 Nov 2016 21:23:16 +0000 (UTC) Received: by mail.wl.linuxfoundation.org (Postfix, from userid 486) id 343B3296DD; Thu, 17 Nov 2016 21:23:16 +0000 (UTC) X-Spam-Checker-Version: SpamAssassin 3.3.1 (2010-03-16) on pdx-wl-mail.web.codeaurora.org X-Spam-Level: X-Spam-Status: No, score=-6.3 required=2.0 tests=BAYES_00,DKIM_SIGNED, RCVD_IN_DNSWL_HI, RCVD_IN_SORBS_SPAM, T_DKIM_INVALID autolearn=ham version=3.3.1 Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by mail.wl.linuxfoundation.org (Postfix) with ESMTP id 5BCE5296DB for ; Thu, 17 Nov 2016 21:23:15 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1752463AbcKQVXO (ORCPT ); Thu, 17 Nov 2016 16:23:14 -0500 Received: from mail-it0-f43.google.com ([209.85.214.43]:36293 "EHLO mail-it0-f43.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751091AbcKQVXM (ORCPT ); Thu, 17 Nov 2016 16:23:12 -0500 Received: by mail-it0-f43.google.com with SMTP id l8so164071857iti.1 for ; Thu, 17 Nov 2016 13:23:12 -0800 (PST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=kernel-dk.20150623.gappssmtp.com; s=20150623; h=to:cc:from:subject:message-id:date:user-agent:mime-version :content-transfer-encoding; bh=P/T7iqdUVcrecvzu9I9FC0iuh/yxm+7B1wFkr1+KNY0=; b=0o0TSE4Ck9FG+SxdbgsG8hKtZNejfckjj++ikjULz1oTjf+DPV7Y5YV+CRA/BNTNw+ IKD1fZnf3KV2XAIrHctftH+JD9jdCJRvDid1T/A3azxsdzYWMjnt/IW48za9pb4YB5SI DxJFBh3f+/32jO9yDqJRGcWUuSfylgk1RUNeeKFByjBpKJI3B53zcxjE9rD+I3BfgCEd Cnz1e6ltT5cgPtA7k/K8Flb7xUeR29PUxPcOiy944eReTY6v8iCJXfOpqI0OtaAlRjyl ILQhx51EF2YB12cMZJWJ1h6QIYTnDHPCoXdr3ZLQE9OopSbbYS8K30tTpp8vsq8jzp5x kUFg== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20130820; h=x-gm-message-state:to:cc:from:subject:message-id:date:user-agent :mime-version:content-transfer-encoding; bh=P/T7iqdUVcrecvzu9I9FC0iuh/yxm+7B1wFkr1+KNY0=; b=ae1cX7YqY2EhEz4E3QNrdy0zdSK9UlLnzlrNO4ANlY36LoGzJhD2iN91Cf+BzGMIyT xuFpXKf3GbqHO/y4jU54c8I1+RyG1A12QupOiABRJCx/lNVU5QXraienbQhoeHBkl2pC eYXQuKwwqa3Y5My2K57602V1tl53jbJS413cAMxU73sOd7b9ySTRouzHXoawXXHkYhEO bBywcydsObd3u+Gp2uKjvUMAEFWgSOXZYaoqbdnrpZJHh8JByLulBsBMrfsTnJUCHh18 VgGBXZeNx3vsePwnNd8xEA1fmD1qdk3oG0jC56PXPjjENcVnKmXELBq3aHlgYApZC6ha sjWw== X-Gm-Message-State: AKaTC01s7DbegeaSAFgK6bsS9VRQLKF2Uvdx6SHlAfG8E5xkwqEJPuuzO8tM+JtZfTSnsQ== X-Received: by 10.107.129.150 with SMTP id l22mr5436175ioi.149.1479417791832; Thu, 17 Nov 2016 13:23:11 -0800 (PST) Received: from [192.168.1.129] ([216.160.245.98]) by smtp.gmail.com with ESMTPSA id q19sm1913915ioi.36.2016.11.17.13.23.10 (version=TLS1_2 cipher=ECDHE-RSA-AES128-GCM-SHA256 bits=128/128); Thu, 17 Nov 2016 13:23:11 -0800 (PST) To: Andrew Morton , Johannes Weiner , linux-mm@kvack.org Cc: "linux-kernel@vger.kernel.org" , "linux-block@vger.kernel.org" , Linus Torvalds From: Jens Axboe Subject: [PATCH v4] mm: don't cap request size based on read-ahead setting Message-ID: Date: Thu, 17 Nov 2016 14:23:10 -0700 User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:45.0) Gecko/20100101 Thunderbird/45.4.0 MIME-Version: 1.0 Sender: linux-block-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-block@vger.kernel.org X-Virus-Scanned: ClamAV using ClamSMTP We ran into a funky issue, where someone doing 256K buffered reads saw 128K requests at the device level. Turns out it is read-ahead capping the request size, since we use 128K as the default setting. This doesn't make a lot of sense - if someone is issuing 256K reads, they should see 256K reads, regardless of the read-ahead setting, if the underlying device can support a 256K read in a single command. To make matters more confusing, there's an odd interaction with the fadvise hint setting. If we tell the kernel we're doing sequential IO on this file descriptor, we can get twice the read-ahead size. But if we tell the kernel that we are doing random IO, hence disabling read-ahead, we do get nice 256K requests at the lower level. This is because ondemand and forced read-ahead behave differently, with the latter doing the right thing. An application developer will be, rightfully, scratching his head at this point, wondering wtf is going on. A good one will dive into the kernel source, and silently weep. This patch introduces a bdi hint, io_pages. This is the soft max IO size for the lower level, I've hooked it up to the bdev settings here. Read-ahead is modified to issue the maximum of the user request size, and the read-ahead max size, but capped to the max request size on the device side. The latter is done to avoid reading ahead too much, if the application asks for a huge read. With this patch, the kernel behaves like the application expects. Signed-off-by: Jens Axboe Acked-by: Johannes Weiner --- Changes since v3: - Went over it with Johannes, cleaned up the the logic as a result Changes since v2: - Fix up the last minute typo on io_pages (Johannes/Hillf) - Apply the same limit to force_page_cache_readahead(). int err; @@ -369,10 +374,18 @@ ondemand_readahead(struct address_space *mapping, bool hit_readahead_marker, pgoff_t offset, unsigned long req_size) { - unsigned long max = ra->ra_pages; + struct backing_dev_info *bdi = inode_to_bdi(mapping->host); + unsigned long max_pages = ra->ra_pages; pgoff_t prev_offset; /* + * If the request exceeds the readahead window, allow the read to + * be up to the optimal hardware IO size + */ + if (req_size > max_pages && bdi->io_pages > max_pages) + max_pages = min(req_size, bdi->io_pages); + + /* * start of file */ if (!offset) @@ -385,7 +398,7 @@ ondemand_readahead(struct address_space *mapping, if ((offset == (ra->start + ra->size - ra->async_size) || offset == (ra->start + ra->size))) { ra->start += ra->size; - ra->size = get_next_ra_size(ra, max); + ra->size = get_next_ra_size(ra, max_pages); ra->async_size = ra->size; goto readit; } @@ -400,16 +413,16 @@ ondemand_readahead(struct address_space *mapping, pgoff_t start; rcu_read_lock(); - start = page_cache_next_hole(mapping, offset + 1, max); + start = page_cache_next_hole(mapping, offset + 1, max_pages); rcu_read_unlock(); - if (!start || start - offset > max) + if (!start || start - offset > max_pages) return 0; ra->start = start; ra->size = start - offset; /* old async_size */ ra->size += req_size; - ra->size = get_next_ra_size(ra, max); + ra->size = get_next_ra_size(ra, max_pages); ra->async_size = ra->size; goto readit; } @@ -417,7 +430,7 @@ ondemand_readahead(struct address_space *mapping, /* * oversize read */ - if (req_size > max) + if (req_size > max_pages) goto initial_readahead; /* @@ -433,7 +446,7 @@ ondemand_readahead(struct address_space *mapping, * Query the page cache and look for the traces(cached history pages) * that a sequential stream would leave behind. */ - if (try_context_readahead(mapping, ra, offset, req_size, max)) + if (try_context_readahead(mapping, ra, offset, req_size, max_pages)) goto readit; /* @@ -444,7 +457,7 @@ ondemand_readahead(struct address_space *mapping, initial_readahead: ra->start = offset; - ra->size = get_init_ra_size(req_size, max); + ra->size = get_init_ra_size(req_size, max_pages); ra->async_size = ra->size > req_size ? ra->size - req_size : ra->size; readit: @@ -454,7 +467,7 @@ ondemand_readahead(struct address_space *mapping, * the resulted next readahead window into the current one. */ if (offset == ra->start && ra->size == ra->async_size) { - ra->async_size = get_next_ra_size(ra, max); + ra->async_size = get_next_ra_size(ra, max_pages); ra->size += ra->async_size; } diff --git a/block/blk-settings.c b/block/blk-settings.c index f679ae1..65f16cf 100644 --- a/block/blk-settings.c +++ b/block/blk-settings.c @@ -249,6 +249,7 @@ void blk_queue_max_hw_sectors(struct request_queue *q, unsigned int max_hw_secto max_sectors = min_not_zero(max_hw_sectors, limits->max_dev_sectors); max_sectors = min_t(unsigned int, max_sectors, BLK_DEF_MAX_SECTORS); limits->max_sectors = max_sectors; + q->backing_dev_info.io_pages = max_sectors >> (PAGE_SHIFT - 9); } EXPORT_SYMBOL(blk_queue_max_hw_sectors); diff --git a/block/blk-sysfs.c b/block/blk-sysfs.c index 9cc8d7c..ea374e8 100644 --- a/block/blk-sysfs.c +++ b/block/blk-sysfs.c @@ -212,6 +212,7 @@ queue_max_sectors_store(struct request_queue *q, const char *page, size_t count) spin_lock_irq(q->queue_lock); q->limits.max_sectors = max_sectors_kb << 1; + q->backing_dev_info.io_pages = max_sectors_kb >> (PAGE_SHIFT - 10); spin_unlock_irq(q->queue_lock); return ret; diff --git a/include/linux/backing-dev-defs.h b/include/linux/backing-dev-defs.h index c357f27..b8144b2 100644 --- a/include/linux/backing-dev-defs.h +++ b/include/linux/backing-dev-defs.h @@ -136,6 +136,7 @@ struct bdi_writeback { struct backing_dev_info { struct list_head bdi_list; unsigned long ra_pages; /* max readahead in PAGE_SIZE units */ + unsigned long io_pages; /* max allowed IO size */ unsigned int capabilities; /* Device capabilities */ congested_fn *congested_fn; /* Function pointer if device is md/dm */ void *congested_data; /* Pointer to aux data for congested func */ diff --git a/mm/readahead.c b/mm/readahead.c index c8a955b..344c1da 100644 --- a/mm/readahead.c +++ b/mm/readahead.c @@ -207,12 +207,17 @@ int __do_page_cache_readahead(struct address_space *mapping, struct file *filp, * memory at once. */ int force_page_cache_readahead(struct address_space *mapping, struct file *filp, - pgoff_t offset, unsigned long nr_to_read) + pgoff_t offset, unsigned long nr_to_read) { + struct backing_dev_info *bdi = inode_to_bdi(mapping->host); + struct file_ra_state *ra = &filp->f_ra; + unsigned long max_pages; + if (unlikely(!mapping->a_ops->readpage && !mapping->a_ops->readpages)) return -EINVAL; - nr_to_read = min(nr_to_read, inode_to_bdi(mapping->host)->ra_pages); + max_pages = max_t(unsigned long, bdi->io_pages, ra->ra_pages); + nr_to_read = min(nr_to_read, max_pages); while (nr_to_read) {