ioatcontrol(8) could exercise 8k-aligned copy with page-break, crc and crc-copy modes

[tychon]

Summary

In D19725 I touched a few area of the ioat(4) driver for which there didn't exist any existing unit test code. This patch aims to address that.

[cem]

arc-copy modes

typo: "crc-copy" (in the summary)

Looks mostly good to me. A few remarks:

sys/dev/ioat/ioat_test.c

319–322 ↗

(On Diff #55693)

I'm not sure these point to valid indices in buf at the limit. Why not just swap pages with the non-pagebreak case, i.e.,

src2 = ...buf[2*i+1] + PAGE_SIZE;
dst2 = ...buf[2*i] + PAGE_SIZE;

381–382 ↗

(On Diff #55693)

Oh, I see this is how you enforce src/dst+PAGESIZE*2 is valid. Still, it seems excessive when we only need 2*PAGE_SIZE in buffer, just a different access pattern to avoid the contiguous pages.

[tychon]

In D19780#424304, @cem wrote:

arc-copy modes

typo: "crc-copy" (in the summary)

Looks mostly good to me. A few remarks:

In D19780#424304, @cem wrote:

arc-copy modes

typo: "crc-copy" (in the summary)

Looks mostly good to me. A few remarks:

sys/dev/ioat/ioat_test.c

319–322 ↗

(On Diff #55693)

Why not? Sometimes when you start one way you don't see the obvious but cleaner alternative. Like here.

Good suggestion I've fixed it.

[cem]

sys/dev/ioat/ioat_test.c

372 ↗

(On Diff #55709)

Need || TEST_DMA_8K_PB now, I think :-)

[tychon]

Add || TEST_DMA_8K_PB to pre-condition verification.