jbk: it's an interrupt for one of the chelsio NICs, and I initially kind of looked past it because it didn't seem interesting, but now I realized it's an error interrupt, which I don't think we often see, so that's fun

so maybe I had some transient hardware failure, or maybe IRM did something screwy

i'm wondering if there's a timing bit there -- apix_wait_till_seen() appears to spin until there's no active or pending interrupts on the cpu

so maybe that interrupt hit at the wrong time

but the fact that cpu hadn't switched for so long does seem problematic