From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from verein.lst.de ([213.95.11.211]:36021 "EHLO newverein.lst.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751367AbdAYJNx (ORCPT ); Wed, 25 Jan 2017 04:13:53 -0500 Date: Wed, 25 Jan 2017 10:13:51 +0100 From: Christoph Hellwig Subject: Re: [PATCH 2/2] xfs: use per-AG reservations for the finobt Message-ID: <20170125091351.GA16307@lst.de> References: <1485288980-17075-1-git-send-email-hch@lst.de> <1485288980-17075-3-git-send-email-hch@lst.de> <20170124204257.GB5726@birch.djwong.org> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20170124204257.GB5726@birch.djwong.org> Sender: linux-xfs-owner@vger.kernel.org List-ID: List-Id: xfs To: "Darrick J. Wong" Cc: Christoph Hellwig , bfoster@redhat.com, linux-xfs@vger.kernel.org On Tue, Jan 24, 2017 at 12:42:57PM -0800, Darrick J. Wong wrote: > > + if (error) { > > + /* > > + * Because we didn't have per-AG reservations when the > > + * finobt feature was added we might not ne able to > > + * reservere all needed blocks. Warn and fall back to > > "...might not be able to reserve all..." Fixed. > So if __xfs_ag_resv_init() returns ENOSPC, we don't actually halt the > mount attempt, which means that we can start operating with > mp->m_inotbt_nores = false if there's not even enough space for the > refcountbt-only reservation. In theory this is impossible, but I think > we'd be in a better defensive position by moving this to before the > __xfs_ag_resv_init call. Ok. > > xfs_agnumber_t agno, struct xfs_buf **bpp); > > > > - > > #endif /* __XFS_IALLOC_H__ */ > > Unnecessary change? Fixed.