From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from bombadil.infradead.org ([198.137.202.9]:50624 "EHLO bombadil.infradead.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S966328AbcLVUA0 (ORCPT ); Thu, 22 Dec 2016 15:00:26 -0500 From: Christoph Hellwig Subject: minleft fixes V2 Date: Thu, 22 Dec 2016 21:00:17 +0100 Message-Id: <1482436822-31546-1-git-send-email-hch@lst.de> Sender: linux-xfs-owner@vger.kernel.org List-ID: List-Id: xfs To: linux-xfs@vger.kernel.org Cc: eguan@redhat.com, darrick.wong@oracle.com his is my attempt to fix the problems with rmap and reflink file system running out of space during delayed extent conversions. It turns out the way minleft has been handled has always been bogus, but the rmap enablement made it even worse.. This has survived xfstests xfs/109 on a 2k fs for a couple hours now but Eryu was much better than me at reproducing the problems, so it could use some further testing. Changes since V1: - new patch to fix the noalign case in xfs_bmap_btalloc - new patch to fix xfs_alloc_set_aside - dropped the patch to increase the minleft value for bmap allocations, we only need the current minimal values as we only allocate one extent per call - minor style fixes per review of the last round