nvme: complete requests when payload DMA mapping fails

[seuros]

Summary

bus_dmamap_load_mem() reports most mapping failures, including EFBIG,
only through its callback and then returns zero. nvme_payload_map()
logged the error without telling the submission path, so the tracker
stayed on the outstanding list with no command submitted and no
timeout armed, stalling all later I/O on the queue behind it.

MFC after: 2 weeks

[seuros]

rebase

[imp]

sys/dev/nvme/nvme_qpair.c

1181

Should we set this to EINPROGRESS here? Are there paths out which don't do the DMA that aren't a mapping failure?

1200

I'd think that we don't want to wait here.
If it's just to detect this error, can't we detect it in the callback instead?

[ngie]

sys/dev/nvme/nvme_qpair.c

1200

+1. Doing this in the callback (if possible) seems much cleaner so you don't have to deal with the truth table of err x tr->map_err conditional combinations and the associated complexity in all of the downstream callers.