-----Original Message----- > > - Is it possible to use the new logic when blk-mq has sbitmap, even if > > a kernel has rq_dispatched and rq_completed? > > > > > > > > If we would like to replace the rq_dispatched/rq_completed with the sbitmap > > to calculate the disk I/O statistics, that might be extra work. > > > > But anyway, if it's worth doing, we might consider doing it at the next stage. > > What do you think? > > sorry, cannot get it. I just mean this flow: > > if blk-mq uses sbitmap then > get_mq_diskio_from_hw_queues(q, &tmp); > return; > else if rq_dispatched/rq_completed exists then > ... > get_one_mctx_diskio(mctx_addr, &tmp); > > > > It means that crash will preferentially use the sbitmap to calculate the disk I/O statistics > since kernel started to introduce the sbitmap for the blk-mq feature. Yes, right. > > > What is the extra work? > > > > The extra work is that: crash has to support parsing sbitmap on old kernel, and need to > follow up the relevant kernel structure changes, it will make the implementation of the > get_mq_diskio_from_hw_queues() become more complex. That is my concern. Got it, thanks. I thought that we wanted RHEL8 crash also fixed with this first series as the bz is for RHEL8. But ok, we can fix it later one by one. Thanks, Kazu -- Crash-utility mailing list Crash-utility@xxxxxxxxxx https://listman.redhat.com/mailman/listinfo/crash-utility Contribution Guidelines: https://github.com/crash-utility/crash/wiki
