From mboxrd@z Thu Jan 1 00:00:00 1970 From: Christoph Hellwig Subject: Re: [PATCH 1/3] udf: Replace bkl with the inode->i_alloc_sem for protect udf_inode_info struct Date: Tue, 16 Nov 2010 13:05:44 -0500 Message-ID: <20101116180544.GA29896@infradead.org> References: <1289929249-4863-1-git-send-email-abogani@texware.it> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: Jan Kara , Arnd Bergmann , Christoph Hellwig , Tim Bird , LKML , linux-fsdevel@vger.kernel.org To: Alessio Igor Bogani Return-path: Received: from bombadil.infradead.org ([18.85.46.34]:51324 "EHLO bombadil.infradead.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1755585Ab0KPSFu (ORCPT ); Tue, 16 Nov 2010 13:05:50 -0500 Content-Disposition: inline In-Reply-To: <1289929249-4863-1-git-send-email-abogani@texware.it> Sender: linux-fsdevel-owner@vger.kernel.org List-ID: On Tue, Nov 16, 2010 at 06:40:47PM +0100, Alessio Igor Bogani wrote: > Replace bkl with the inode->i_alloc_sem rw semaphore in udf_release_file(), > udf_symlink(), udf_symlink_filler(), udf_get_block() and udf_block_map(). > Add protection in udf_evict_inode() using the same i_alloc_sem rw semaphore. I'd rather prefer not to introduce new users of i_alloc_sem. It's a quite nasty beast: the only rw_semaphore that is not released by the thread acquiring it. Thomas asked me if there's a way to get rid of it, and I've come up with some schemes that I need to prototype. Adding more uses that are unrelated to the original direct I/O use case are not very helpful in doing that.