Williams, Dan J wrote:

+	cookie = iop_adma_tx_submit(tx);
+	iop_adma_issue_pending(dma_chan);
+	msleep(8);

Why sleeping exactly 8 ms?
Has this value been somehow experiomentally obtained?
How about introducing wait_for_completion, just like it is in ioatdma now?

General remark:
Both iop_adma_pq_zero_sum_self_test() and iop_adma_xor_zero_sum_self_test()
are quite long and they consist of series of subsequent similar tests.
I guess the code would be easier to read if every test was a separate function
called from iop_adma_XX_zero_sum_self_test().

Regards,
Maciej