On Tue, May 30, 2023 at 10:07:54AM +0800, Yu Kuai wrote: >> If we don't care about blktrace for passthrough io after del_gendisk(), >> and /dev/sg* has separate handling, I think it's better just to check >> QUEUE_FLAG_REGISTERED in blk_trace_setup(), and don't enable blktrace >> in the first place. > > Any suggestions about this problem? Should we use separate handling for > /dev/sd? Or just free blktrace in blk_free_queue(). I'd be fine with trying to either remove the /dev/sg blktrace handling and / or splitting it up so that it doesn't interact with the main disk based one. I can look into this if you want, or leave it to you.
