> Don't tell me this fixes SATA on the xserve G5...
It does.
> Wonder if a "posting read" is necessary here. If that removes the
> need to increase the delay, it probably is a better solution.
I've tried this diff and I have not been able to reproduce the bogus
SStatus value in a few reboots yet:
Index: pciide.c
===================================================================
RCS file: /OpenBSD/src/sys/dev/pci/pciide.c,v
retrieving revision 1.355
diff -u -p -r1.355 pciide.c
--- pciide.c 18 Oct 2015 20:24:10 -0000 1.355
+++ pciide.c 15 Nov 2015 10:21:36 -0000
@@ -7924,10 +7924,14 @@ svwsata_drv_probe(struct channel_softc *
bus_space_write_4(ss->ba5_st, ss->ba5_sh,
(channel << 8) + SVWSATA_SCONTROL, scontrol);
+ (void)bus_space_read_4(ss->ba5_st, ss->ba5_sh,
+ (channel << 8) + SVWSATA_SCONTROL);
delay(50 * 1000);
scontrol &= ~SControl_DET_INIT;
bus_space_write_4(ss->ba5_st, ss->ba5_sh,
(channel << 8) + SVWSATA_SCONTROL, scontrol);
+ (void)bus_space_read_4(ss->ba5_st, ss->ba5_sh,
+ (channel << 8) + SVWSATA_SCONTROL);
delay(50 * 1000);
sstatus = bus_space_read_4(ss->ba5_st, ss->ba5_sh,
