linux-kernel - Re: [RFC] iomap: fix race between readahead and direct write

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <64d617cc-e7fe-6848-03bb-aab3498c9a07@huawei.com>
Date:   Sun, 19 Jan 2020 10:51:37 +0800
From:   "yukuai (C)" <yukuai3@...wei.com>
To:     Matthew Wilcox <willy@...radead.org>
CC:     <hch@...radead.org>, <darrick.wong@...cle.com>,
        <linux-xfs@...r.kernel.org>, <linux-fsdevel@...r.kernel.org>,
        <linux-kernel@...r.kernel.org>, <houtao1@...wei.com>,
        <zhengbin13@...wei.com>, <yi.zhang@...wei.com>
Subject: Re: [RFC] iomap: fix race between readahead and direct write



On 2020/1/19 9:42, Matthew Wilcox wrote:
> Did you read my patch series?  The current code allocates pages,
> but does not put them in the page cache until after iomap is called.
> My patch series changes that to put the pages in the page cache as soon
> as they're allocated, and before iomap is called.

I just read you patch series again.

At first, if you try to add all pages to pagecache and lock them before
iomap_begin. I thought aboult it before, but I throw away the idea
becacuse all other operation that will lock the page will need to wait
for readahead to finish. And it might cause problem for performance
overhead. And if you try to add each page to page cache and call iomap
before adding the next page. Then, we are facing the same CPU overhead
issure.

Then, there might be a problem in your implementation.
if 'use_list' is set to true here:
+	bool use_list = mapping->a_ops->readpages;

Your code do not call add_to_page_cache_lru for the page.
+		if (use_list) {
+			page->index = page_offset;
+			list_add(&page->lru, &page_pool);
+		} else if (!add_to_page_cache_lru(page, mapping, page_offset,
+					gfp_mask)) {
+			if (nr_pages)
+				read_pages(mapping, filp, &page_pool,
+						page_offset - nr_pages,
+						nr_pages);
+			nr_pages = 0;
+			continue;
+		}

And later, you replace 'iomap_next_page' with 'readahead_page'
+static inline
+struct page *readahead_page(struct address_space *mapping, loff_t pos)
+{
+	struct page *page = xa_load(&mapping->i_pages, pos / PAGE_SIZE);
+	VM_BUG_ON_PAGE(!PageLocked(page), page);
+
+	return page;
+}
+

It seems that the page will never add to page cache.

Thanks!
Yu Kuai