of a data transfer operation immediately after the data transfer
was finished, instead of waiting for the chip to interrupt us and
tell us that it was finished and had the result for us. This worked
okay for read and write since the operation would be finished very
shortly after the data transfer completed. However, with formatting,
the chip still had most of the rest of the track to do, so we ended
up timing out before the operation was finished.
enable the rdmsr call in msr_write_ipi(), so that when it's not
defined we don't read it before writing; disabled in powernow_k8
and enabled in the others.
so avoid making them.
- When parking an LWP on a condition variable, point the hint argument at
the mutex's waiters queue. Chances are we will be awoken from that later.
powernow_k8 driver (much better than undeffing and write it again).
* Fix the WRITE_FIDVID macro, I changed it to use the third argument
for the bitmask, but it's not correct.
Last change should fix the problem reported by FUKUMOTO Atsushi.
memory and scrolling trough it. If the hardware is not capable it will
disable hardware scrolling.
For 640x480 at 8 bpp the speedup is around 4 times, at 1280x1280 at 32 bpp
the speedup is around 9.4 times.
Checked and OK'd by Jared McNeill. Thanks go to Jared and Michael Lorenz
for their tips and vcons knowledge!
Basically, when we switch from WSDISPLAYIO_MODE_EMUL to anything else, we
release any bus_space resources. When we return to WSDISPLAYIO_MODE_EMUL,
reclaim these resources and reinitialize the display.
Idea from macallan@
too, but that's not critical): real hubs send port status change
notifications until acknowledged, ehci/ohci send only once so there is
a chance that this gets lost if notifications are disabled during a scan.
The temporary fix is to force a rescan; the real fix will involve a
change in interrupt handling in ehci/ohci which needs more work and
testing.
ehci is critical because a port status change during scan regularely
happens if a full/lowspeed device is connected, due to the handover
process.
Fixes a problem where full/lowspeed devices were not reattached after
a disconnect, reported by Kouichirou Hiratsuka and Juan RP
on current-users.
