On Mon 25-01-21 20:42:49, Christoph Hellwig wrote: > On Mon, Jan 25, 2021 at 04:45:07PM +0100, Jan Kara wrote: > > > What do you mean by "take"? Take the superblock as an argument to > > > quotactl_sb() or take a reference to the superblock? > > > Sorry, I don't really get where you aiming at. > > > > I think Christoph was pointing at the fact it is suboptimal to search for > > superblock by device number when you already have a pointer to it. And I > > guess he was suggesting we could pass 'sb' pointer to quotactl_sb() when we > > already have it. Although to be honest, I'm not sure how Christoph imagines > > the refactoring of user_get_super() he mentions - when we have a path > > looked up through user_path(), that pins the superblock the path is on so > > it cannot be unmounted. So perhaps quotactl_sb() can done like: > > I don't think we need a quotactl_sb at all, do_quotactl is in fact a > pretty good abstraction as-is. > > For the path based one we just need to factor out a little helper > to set excl and thaw and then call it like: > > sb = path.dentry->d_inode->i_sb; > > if (excl) > down_write(&sb->s_umount); > else > down_read(&sb->s_umount); > if (thawed && sb->s_writers.frozen != SB_UNFROZEN) > ret = -EBUSY; > else > ret = do_quotactl(sb, type, cmds, id, addr, &path); > if (excl) > up_write(&sb->s_umount); > else > up_read(&sb->s_umount); > > as there is no good reason to bring over the somewhat strange wait until > unfrozen semantics to a new syscall. Well, I don't think that "wait until unfrozen" is that strange e.g. for Q_SETQUOTA - it behaves like setxattr() or any other filesystem modification operation. And IMO it is desirable that filesystem freezing is transparent for operations like these. For stuff like Q_QUOTAON, I agree that returning EBUSY makes sense but then I'm not convinced it's really simpler or more useful behavior... Honza -- Jan Kara <jack@xxxxxxxx> SUSE Labs, CR
