> > I did not explain myself well. > > > > This should be enough IMO: > > > > @@ -483,7 +483,7 @@ static int ovl_copy_up_inode(struct > > ovl_copy_up_ctx *c, struct dentry *temp) > > } > > > > inode_lock(temp->d_inode); > > - if (c->metacopy) > > + if (S_ISREG(c->stat.mode)) > > err = ovl_set_size(temp, &c->stat); > > if (!err) > > err = ovl_set_attr(temp, &c->stat); > > > > There is no special reason IMO to try to spare an unneeded ovl_set_size > > if it simplifies the code a bit. > > We can try this but I'm afraid that someone could complain > we do unnecessary ovl_set_size() in the case of full copy-up > or data-end file's copy-up. > > There is no one to complain. The cost of ovl_set_size() is insignificant compared to the cost of copying data (unless I am missing something). Please post a version as above and if Miklos finds it a problem, we can add a boolean c->should_set_size to the copy up context, initialize it: c->should_set_size = (S_ISREG(c->stat.mode) && c->stat.size) and set it to false in case all data was copied. I think that won't be necessary though. Thanks, Amir.
