Steffen,
As I said, I don't have a problem with having module parameters. > There's one more important thing that has performance impact: We need to > pack payload and protection data into the same queue of limited > length. So for the worst case with DIX, we have to use half the size for > sg_tablesize to get the other half for sg_prot_tablesize. Interesting. With the exception of NVMe over PCIe, it's kind of unusual for modern controllers to have real limits in this area. > This limits the maximum I/O request size and thus throughput. Using > read_verify and write_generate does not change the tablesizes, as zfcp > would still announce support for DIF and DIX. With the new zfcp.dif=1 > and zfcp.dix=0, we can use the full sg_tablesize for payload data and > sg_prot_tablesize=0. (The DIF "overhead" on the fibre still exists of > course.) > > Are there other ways for accomplishing this which I'm not aware of? You can set your shost->sg_prot_tablesize. There are pathological corner cases like dd to the raw block device where you'll suffer if there is not a 1:1 mapping between data and protection segments. But for regular I/O where the protection is generated over a whole bio at a time, you can get away with something smaller. Anyway. I don't have any problems with you making DIX experimental for zfcp. Just want to make sure it's done for the right reasons (i.e. not problems in SCSI or the block layer). -- Martin K. Petersen Oracle Linux Engineering
