From mboxrd@z Thu Jan 1 00:00:00 1970 From: Vivek Goyal Subject: Re: [PATCH 3/7] ovl: Copy up only metadata during copy up where it makes sense Date: Tue, 3 Oct 2017 10:39:49 -0400 Message-ID: <20171003143949.GC14308@redhat.com> References: <1506951605-31440-1-git-send-email-vgoyal@redhat.com> <1506951605-31440-4-git-send-email-vgoyal@redhat.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Received: from mx1.redhat.com ([209.132.183.28]:54984 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751460AbdJCOju (ORCPT ); Tue, 3 Oct 2017 10:39:50 -0400 Content-Disposition: inline In-Reply-To: Sender: linux-unionfs-owner@vger.kernel.org List-Id: linux-unionfs@vger.kernel.org To: Amir Goldstein Cc: overlayfs , Miklos Szeredi On Mon, Oct 02, 2017 at 10:21:05PM +0300, Amir Goldstein wrote: > On Mon, Oct 2, 2017 at 4:40 PM, Vivek Goyal wrote: > > If it makes sense to copy up only metadata during copy up, do it. This > > is done for regular files which are not opened for WRITE and have origin > > being saved. > > > > Signed-off-by: Vivek Goyal > > --- > > fs/overlayfs/copy_up.c | 27 +++++++++++++++++++-------- > > 1 file changed, 19 insertions(+), 8 deletions(-) > > > > diff --git a/fs/overlayfs/copy_up.c b/fs/overlayfs/copy_up.c > > index cdc4d79a1249..b2d9ed81e9ff 100644 > > --- a/fs/overlayfs/copy_up.c > > +++ b/fs/overlayfs/copy_up.c > > @@ -306,11 +306,8 @@ static int ovl_set_origin(struct dentry *dentry, struct dentry *lower, > > return PTR_ERR(fh); > > } > > > > - /* > > - * Do not fail when upper doesn't support xattrs. > > - */ > > err = ovl_check_setxattr(dentry, upper, OVL_XATTR_ORIGIN, fh, > > - fh ? fh->len : 0, 0); > > + fh ? fh->len : 0, -EOPNOTSUPP); > > kfree(fh); > > > > return err; > > @@ -328,6 +325,7 @@ struct ovl_copy_up_ctx { > > struct dentry *workdir; > > bool tmpfile; > > bool origin; > > + bool metadata_only; > > }; > > > > static int ovl_link_up(struct ovl_copy_up_ctx *c) > > @@ -464,11 +462,15 @@ static int ovl_copy_up_inode(struct ovl_copy_up_ctx *c, struct dentry *temp) > > */ > > if (c->origin) { > > err = ovl_set_origin(c->dentry, c->lowerpath.dentry, temp); > > - if (err) > > - return err; > > + if (err) { > > + if (err != -EOPNOTSUPP) > > + return err; > > + /* Upper does not support xattr. Copy up data as well */ > > + c->metadata_only = false; > > + } > > } > > > > - if (S_ISREG(c->stat.mode)) { > > + if (S_ISREG(c->stat.mode) && !c->metadata_only) { > > struct path upperpath; > > > > I would set the meta inode size here. see: > https://github.com/amir73il/linux/commits/ovl-rocopyup > Hi Amir, Ok. So instead of having a zero size file in upper, set the size to same as lower file. Just that there will not be any data blocks in upper file yet? Will do. Vivek > > ovl_path_upper(c->dentry, &upperpath); > > @@ -596,6 +598,7 @@ static int ovl_copy_up_one(struct dentry *parent, struct dentry *dentry, > > .dentry = dentry, > > .workdir = ovl_workdir(dentry), > > .origin = true, > > + .metadata_only = true, > > }; > > > > if (WARN_ON(!ctx.workdir)) > > @@ -616,9 +619,17 @@ static int ovl_copy_up_one(struct dentry *parent, struct dentry *dentry, > > if (err) > > return err; > > > > + if (!S_ISREG(ctx.stat.mode)) > > + ctx.metadata_only = false; > > + > > /* maybe truncate regular file. this has no effect on dirs */ > > - if (flags & O_TRUNC) > > + if (flags & O_TRUNC) { > > ctx.stat.size = 0; > > + ctx.metadata_only = false; > > + } > > + > > + if (flags & (OPEN_FMODE(flags) & FMODE_WRITE)) > > + ctx.metadata_only = false; > > > > if (S_ISLNK(ctx.stat.mode)) { > > ctx.link = vfs_get_link(ctx.lowerpath.dentry, &done); > > -- > > 2.13.5 > >