e1000: always enable PCSD when RSS hashing is used

[kbowling]

Summary

To enable RSS hashing in the NIC, the PCSD bit must be set.

By default, this is never set when RXCSUM is disabled - which
causes problems higher up in the stack.

While here improve the RXCSUM flag assignments when enabling or disabling
IFCAP_RXCSUM.

See also: https://lists.freebsd.org/pipermail/freebsd-current/2020-May/076148.html

Co-authored-by: Stephan de Wit <[e-mail]>
Co-authored-by: Franco Fichtner <[e-mail]>

[kbowling]

Taking a look at this.. I suspect the 82574 MQ case will be mishandled let me check the various datasheets.

[kbowling]

Taking this so I can show you my changes, feel free to re-take after looking

[kbowling]

PCSD is supposed to be set when using RSS features so I moved setting it into there.

I noticed some other issues and tried to clean this up based on your review, let me know what you think.

Here is my datasheet stash: http://iommu.com/datasheets/e1000-datasheets/

The main ones I check are for lem, em, and igb:

• http://iommu.com/datasheets/e1000-datasheets/pci-pci-x-family-gbe-controllers-software-dev-manual.pdf
• http://iommu.com/datasheets/e1000-datasheets/pcie-gbe-controllers-open-source-manual.pdf
• http://iommu.com/datasheets/e1000-datasheets/OpenSDM_82576-DS-2p0.pdf

I also checked NetBSD's if_wm for this which doesn't directly use the Intel code and is a good cross-reference for errata.

[franco_opnsense.org]

Much nicer, thanks! Added Stephan as reviewer who originally worked on this to give it a go.

One small thing while we edit defines: can E1000_MRQC_ENABLE_RSS_4Q either be removed or defined as E1000_MRQC_ENABLE_RSS_8Q since it's the same value and practically unused?

[kbowling]

Address review feedback

[stephan.dewt_yahoo.co.uk]

Looks good to me.