On 1/21/22 11:17, John Pittman wrote:
Thanks for looking Steffen. I failed to notice the no_hash_pointers option. I do think it would be a reasonable convenience for debug scsi logging to print the actual address without having to go through the trouble of rebooting enabling the no_hash_pointers parameter then rebooting again to disable afterward. Perhaps the system is a production box that can't be rebooted. Maybe an additional sysctl to enable no_hash_pointers would be reasonable? That way it could be changed online? Thanks again.
Are SCSI command pointers really useful to anyone? How about modifying the debug statements such that the SCSI command tags are reported instead of the SCSI command pointers? There is a 1:1 correspondence between the values returned by blk_mq_unique_tag() and the SCSI command pointers.
Bart.
