I have no issue with your proposal, but I would rather get this fixed and adjust the API with an additional fix. I suspect we probably want to mark this fix stable.
I agree, we should get it in and fix the API incrementally.
I did audit the kernel ULPs for FMR vs. FRMR usage and the only one that looked at max_fast_reg_page_list_len unconditionally was iser.
Correct, but it doesn't mean that others get it right. nfs uses a hard coded value without even knowing what the device is capable of, and srp calculates it from max_mr_size which is a user-space mr capability and has nothing to do with fmrs. So no one got it right, which is not surprising as we don't have a cap for it. -- To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
