From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from userp2130.oracle.com ([156.151.31.86]:48260 "EHLO userp2130.oracle.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751024AbeCZTX3 (ORCPT ); Mon, 26 Mar 2018 15:23:29 -0400 Date: Mon, 26 Mar 2018 12:22:59 -0700 From: "Darrick J. Wong" Subject: Re: [PATCH] xfs: use hardlimit as sub-fs size if both hard/soft limits are set Message-ID: <20180326192259.GU4818@magnolia> References: <1521954996-203628-1-git-send-email-cgxu519@gmx.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <1521954996-203628-1-git-send-email-cgxu519@gmx.com> Sender: linux-xfs-owner@vger.kernel.org List-ID: List-Id: xfs To: Chengguang Xu Cc: linux-xfs@vger.kernel.org, bfoster@redhat.com, sandeen@sandeen.net, Dave Chinner , Christoph Hellwig On Sun, Mar 25, 2018 at 01:16:36PM +0800, Chengguang Xu wrote: > In current implementation, we size the fs(sub-fs via project quota) at > the soft limit and simply call it 100% used if the limit is exceeded. > It is reasonable when only a soft limit is set, but we should use the > hard limit if both hard/soft limits are set, so that quota-df reflects > the usage information more accurately. This is the followup patch to "xfs: adjust size/used/avail information for quota-df", correct? I also wonder, statvfs is a weird interface since there's no way to send back usage information, just blocks/free/avail. Isn't it more appropriate to use xfs_quota to find out the usage, hard limit, and soft limit of a directory? > Signed-off-by: Chengguang Xu > --- > fs/xfs/xfs_qm_bhv.c | 12 ++++++------ > 1 file changed, 6 insertions(+), 6 deletions(-) > > diff --git a/fs/xfs/xfs_qm_bhv.c b/fs/xfs/xfs_qm_bhv.c > index 2be6d27..43b0fe8 100644 > --- a/fs/xfs/xfs_qm_bhv.c > +++ b/fs/xfs/xfs_qm_bhv.c > @@ -35,9 +35,9 @@ > { > uint64_t limit; > > - limit = dqp->q_core.d_blk_softlimit ? > - be64_to_cpu(dqp->q_core.d_blk_softlimit) : > - be64_to_cpu(dqp->q_core.d_blk_hardlimit); > + limit = dqp->q_core.d_blk_hardlimit ? > + be64_to_cpu(dqp->q_core.d_blk_hardlimit) : > + be64_to_cpu(dqp->q_core.d_blk_softlimit); Ok, so now we report hard limit for f_blocks over the soft limit. So if this is the state of the filesystem: # xfs_quota -xc 'report -ahp' Project quota on /opt (/dev/sdf) Blocks Project ID Used Soft Hard Warn/Grace ---------- --------------------------------- #0 0 0 0 00 [------] vms 3M 2M 3M 00 [7 days] Then the df output goes from: # df /opt/b Filesystem Size Used Avail Use% Mounted on /dev/sdf 2.0M 2.0M 0 100% /opt to this: # df /opt/b Filesystem Size Used Avail Use% Mounted on /dev/sdf 3.0M 3.0M 0 100% /opt That makes to me, but as it /does/ change the behavior of an existing user-visible interface, I would like to know more about the current behavior. Dave/Christoph, do you recall why df reports the project quota soft limit? ---- Just for fun let's try the same on ext4... $ dd if=/dev/zero >> /opt/b/a ^C4129977+0 records in 4129977+0 records out 2114548224 bytes (2.1 GB, 2.0 GiB) copied, 23.9179 s, 88.4 MB/s $ sudo xfs_quota -fxc 'report -ahp' /opt Project quota on /opt (/dev/sdf) Blocks Project ID Used Soft Hard Warn/Grace ---------- --------------------------------- #0 20K 0 0 00 [------] vms 2.0G 2M 3M 00 [-none-] Only 1000x over soft quota... $ df /opt/b Filesystem Size Used Avail Use% Mounted on /dev/sdf 13G 2.8G 8.9G 24% /opt I guess we're going to need a couple more tests, then? One to check that we enforce project quotas, and another to check what we're reporting via df? --D > if (limit && statp->f_blocks > limit) { > statp->f_blocks = limit; > statp->f_bfree = statp->f_bavail = > @@ -45,9 +45,9 @@ > (statp->f_blocks - dqp->q_res_bcount) : 0; > } > > - limit = dqp->q_core.d_ino_softlimit ? > - be64_to_cpu(dqp->q_core.d_ino_softlimit) : > - be64_to_cpu(dqp->q_core.d_ino_hardlimit); > + limit = dqp->q_core.d_ino_hardlimit ? > + be64_to_cpu(dqp->q_core.d_ino_hardlimit) : > + be64_to_cpu(dqp->q_core.d_ino_softlimit); > if (limit && statp->f_files > limit) { > statp->f_files = limit; > statp->f_ffree = > -- > 1.8.3.1 > > -- > To unsubscribe from this list: send the line "unsubscribe linux-xfs" in > the body of a message to majordomo@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html