Mark Lord wrote:
> Heh. Okay, apparently it now needs a .pmp_hardreset function,
> which was never there before. So I've now added one that does this:
Ah... right. That's something changed while changing libata EH to favor
hardreset over softreset. With the change, resuming link on PMP fan out
ports became hardreset's responsibility, which BTW is necessary anyway
for certain PMPs (marvell ones). So, one way or another, you need
> mv_pmp_select(link->ap, sata_srst_pmp(link));
> return sata_std_hardreset(link, class, deadline);
> And it seems to be working now with that function supplied, at least for
> Anything else new that it should have? I'll scan through libata-pmp.c
> and see.
The hardreset thing was the only behavior change if I didn't screw up.
It wasn't supposed to cause any actual behavior change tho as all PMP
supporting PMPs are supposed to have SCR access on fan-out ports and
thus the capability to hardreset them.
> Time to clean out the debug cruft and test a few things a bit more.