On Fri, Jul 01, 2016 at 07:40:41PM +0200, Jan Kara wrote: > > So we are waiting for transaction commit to finish with unsubmitted pages > that already have PageWriteback set (and also potentially other pages that > are locked and we didn't prepare them for writing because the block mapping > we got was too short). Now JBD2 goes on trying to do the transaction > commit: Ah, I see, so this is only an issue in those cases where the handle is synchronous. Is this the only case where there is a concern? (e.g., could we test handle->h_sync and stop the handle early if h_sync is not set?) This would put the uninit->init conversion into potentially a separate transaction, but that should be OK. The reason why I'm pushing so hard here is that long running handles is a major contributor to ext4 haveing poor CPU scalability numbers, since we can end up having lots of threads waiting on the last transaction to complete. So keeping transactions small and fast is a big deal. - Ted -- 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