From: Jan Kara Subject: Re: [PATCH 1/3] jbd2 : Make jbd2 transaction handle allocation to return errors and handle them gracefully. Date: Fri, 4 Feb 2011 16:53:58 +0100 Message-ID: <20110204155358.GF4104@quack.suse.cz> References: <20110123054049.GC3237@thunk.org> <20110123062900.GA7436@noexit> <20110124133143.GA5058@quack.suse.cz> <20110125114656.GB4088@quack.suse.cz> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: Jan Kara , Andreas Dilger , Joel Becker , Ted Ts'o , ext4 To: Manish Katiyar Return-path: Received: from cantor.suse.de ([195.135.220.2]:33077 "EHLO mx1.suse.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751082Ab1BDPyG (ORCPT ); Fri, 4 Feb 2011 10:54:06 -0500 Content-Disposition: inline In-Reply-To: Sender: linux-ext4-owner@vger.kernel.org List-ID: On Sat 29-01-11 21:40:03, Manish Katiyar wrote: Hi Manish, > Have we reached on any conclusion yet on the function name which I can > use to send my updated patch ? My preference from the above list is to > use ext4_journal_start_nofs() as that seems the closest match, but I > would like hear the conclusion from experts. How about "ext4_journal_start_tryhard()"? I don't like "nofs" because the fact whether we use GFP_NOFS is separate from the fact whether we are able to handle memory allocation failure. In fact, your patch set should keep calling kzalloc() with GFP_NOFS in all cases to avoid new recursions into the filesystem (which could cause deadlocks - changing that would need a separate audit if we care enough). Just the retrying logic should be influenced by your work. Honza -- Jan Kara SUSE Labs, CR