Hello, this is a second version of the patches to fix data exposure after a crash in data=ordered mode. Since previous revision I have implemented suggestion by Ted that it is enough to wait for outstanding writeback from jbd2 thread when IO was already submitted during delalloc writeback (see patches 3 and 4 for details). I have checked that indeed jbd2 did not submit any data buffers under appropriate load (I've used randwrite workload from fio) with the improvement but I didn't observe any measurable improvement in transaction commit times - waiting for already outstanding writeback was dominating the commit time and the additional buffers jbd2 was originally submitting apparently didn't make a big difference. At least for this workload. I have a hard time coming up with some other workload where the optimization would make a big difference since flusher thread ends up submitting overwritten blocks anyway which makes jbd2 thread wait for the IO. But still conceptually the change makes sense so it may be worth it. So Ted, feel free to either merge or drop the last two patches in this series... Honza -- To unsubscribe from this list: send the line "unsubscribe linux-ext4" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
