On Sat 29-10-16 13:09:25, Linus Torvalds wrote: > On Sat, Oct 29, 2016 at 12:17 PM, Al Viro <viro@xxxxxxxxxxxxxxxxxx> wrote: > > > > AFAICS, the possibility of dropping the last reference to struct file > > before ->write_iter() has returned is fundamentally broken. I might be > > missing something subtle here, but... > > Ok, let's add a get_file(); fput(); around that whole iter call sequence. > > And that's a separate issue from "we should hold the fs freezer lock > around the whole operation". So I think we need both. Yup, these are two separate issues. I'm fine with adding get_file(); fput() around the iter call sequence. Then, when we have struct file available for the whole time ->write_iter runs, I'd prefer to keep the call to fool lockdep where original file_end_write() call was - that gives us proper lockdep coverage for all the code behind iter_op(). The downside is we cannot keep helpers as elegant as you suggested in your patch but I believe it's bearable and worth the additional lockdep coverage. I'll send patches shortly... Honza -- Jan Kara <jack@xxxxxxxx> SUSE Labs, CR -- To unsubscribe from this list: send the line "unsubscribe stable" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html