From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from smtp01-ext2.udag.de (smtp01-ext2.udag.de [62.146.106.18]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by smtp.subspace.kernel.org (Postfix) with ESMTPS id F0CF856472; Thu, 26 Mar 2026 18:01:51 +0000 (UTC) Authentication-Results: smtp.subspace.kernel.org; arc=none smtp.client-ip=62.146.106.18 ARC-Seal:i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1774548116; cv=none; b=FLYDZIULTEZYr0vvHHWMz3DAZQb5aWOwFo9PrJzUkGe2ZW7J4iNUK4eUsUQ3BkllDyN+JT99CUI0ic6H2GbakGJCJ/46id9+oAtLzRgjz5K5VW5hiv/Ba31kskkDg8Dmk67b/Z9DqAjeu1NLEa0S3udtjB+4lZTMk+E/cFvHANY= ARC-Message-Signature:i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1774548116; c=relaxed/simple; bh=hK+86MFgRbvMlqUBpIQbALyzfb8yb+CoCBa5h/BZ7Pc=; h=Date:From:To:Cc:Subject:Message-ID:References:MIME-Version: Content-Type:Content-Disposition:In-Reply-To; b=fwu8N8ojbXfNYmqg9w3pExrQg/zkgezhR5gqoxERUCEoNXayzdACJntHXgkmPqmbMzzWlm29KtXlIg3kaMqGpBihdbRwuG3hFk7929T2fJfUuH9ymsaN5ff9OlYpm5Bl8L2m5mH2uOwX0FmooqAtAvKguC+pZXuKrh4EzTyO4JQ= ARC-Authentication-Results:i=1; smtp.subspace.kernel.org; dmarc=pass (p=none dis=none) header.from=birthelmer.de; spf=pass smtp.mailfrom=birthelmer.de; arc=none smtp.client-ip=62.146.106.18 Authentication-Results: smtp.subspace.kernel.org; dmarc=pass (p=none dis=none) header.from=birthelmer.de Authentication-Results: smtp.subspace.kernel.org; spf=pass smtp.mailfrom=birthelmer.de Received: from localhost (108-141-067-156.ip-addr.inexio.net [156.67.141.108]) by smtp01-ext2.udag.de (Postfix) with ESMTPA id A7754E0068; Thu, 26 Mar 2026 18:54:20 +0100 (CET) Authentication-Results: smtp01-ext2.udag.de; auth=pass smtp.auth=birthelmercom-0001 smtp.mailfrom=horst@birthelmer.de Date: Thu, 26 Mar 2026 18:54:19 +0100 From: Horst Birthelmer To: Joanne Koong Cc: Miklos Szeredi , Bernd Schubert , Christian Brauner , Horst Birthelmer , linux-fsdevel@vger.kernel.org, linux-kernel@vger.kernel.org, Horst Birthelmer Subject: Re: Re: Re: [PATCH] fuse: fix inode initialization race Message-ID: References: <20260318-fix-inode-init-race-v1-1-a7e58b2ddb9a@ddn.com> <3a7d36c3-0ce0-4f1d-9649-1742f752c5f1@bsbernd.com> <20260326-reorganisation-bemessen-c6643edcf629@brauner> Precedence: bulk X-Mailing-List: linux-fsdevel@vger.kernel.org List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Disposition: inline Content-Transfer-Encoding: 8bit In-Reply-To: On Thu, Mar 26, 2026 at 09:43:00AM -0700, Joanne Koong wrote: > On Thu, Mar 26, 2026 at 8:48 AM Horst Birthelmer wrote: > > > > On Thu, Mar 26, 2026 at 04:19:24PM +0100, Miklos Szeredi wrote: > > > On Thu, 26 Mar 2026 at 16:13, Bernd Schubert wrote: > > > > > > > > > > > > > > > > On 3/26/26 15:26, Christian Brauner wrote: > > > > > On Wed, Mar 25, 2026 at 08:54:57AM +0100, Bernd Schubert wrote: > > > > >> > > > > >> > > > > >> On 3/18/26 14:43, Horst Birthelmer wrote: > > > > >>> From: Horst Birthelmer > > > > > > > >>> fi->attr_version = atomic64_inc_return(&fc->attr_version); > > > > >>> + wake_up_all(&fc->attr_version_waitq); > > > > >>> fi->i_time = attr_valid; > > > > > > > > > > > > While I'm looking at this again, wouldn't it make sense to make this > > > > conditional? Because we wake this queue on every attr change for every > > > > inode. And the conditional in fuse_iget() based on I_NEW? > > > > > > Right, should only wake if fi->attr_version old value was zero. > > > > > > BTW I have a hunch that there are better solutions, but it's simple > > > enough as a stopgap measure. > > > > OK, I'll send a new version. > > > > Just out of curiosity, what would be a better solution? > > I'm probably missing something here but why can't we just call the > > fi = get_fuse_inode(inode); > spin_lock(&fi->lock); > fi->nlookup++; > spin_unlock(&fi->lock); > fuse_change_attributes_i(inode, attr, NULL, attr_valid, attr_version, > evict_ctr); > > logic before releasing the inode lock (the unlock_new_inode() call) in > fuse_iget() to address the race? unlock_new_inode() clears I_NEW so > fuse_reverse_inval_inode()'s fuse_ilookup() would only get the inode > after the attributes initialization has finished. > > As I understand it, fuse_change_attributes_i() would be pretty > straightforward / fast for I_NEW inodes, as it doesn't send any > synchronous requests and for the I_NEW case the > invalidate_inode_pages2() and truncate_pagecache() calls would get > skipped. (truncate_pagecache() getting skipped because inode->i_size > is already attr->size from fuse_init_inode(), so "oldsize != > attr->size" is never true; and invalidate_inode_pages2() getting > skipped because "oldsize != attr->size" is never true and "if > (!timespec64_equal(&old_mtime, &new_mtime))" is never true because > fuse_init_inode() initialized the inode's mtime to attr->mtime). You understand the pretty well, I think. The problem I have there is that fuse_change_attributes_i() takes its own lock. That would be a pretty big operation to split that function. Is that required for this small (as Miklos put it, rare) case? If you guys want me to do that, I will. Thanks, Horst