On Fri, 2014-03-14 at 23:10 +0100, Daniel Vetter wrote: > > I've asked you on private irc whether this range matches/overlaps with > stolen - we know of things blowing up at least on earlier generations > in combination with dmar. Please boot with drm.debug=0xe and scan for > the stolen mem reporting:
Surely it *has* to be stolen? That's the whole *point* in the RMRR that
the BIOS provides, telling us that the gfx unit is expecting to do DMA
to this range of memory.
If it isn't stolen, it's just being wantonly "borrowed".
Have you *ever* known an RMRR point at memory other than the stolen
range?
I think I have found this problem on the IOMMU side. We usually assume
that RMRRs are for boot-time only, such as USB controllers for the
legacy keyboard/mouse emulation. And a patch sneaked in which
effectively *unmaps* the RMRR regions when you do the first "real"
mapping for the driver. Having fixed that, I think I should no longer
see these DMA faults.
Can't tell though, because my machine is still dying in an endless
stream of
[ 199.647850] [drm:intel_set_cpu_fifo_underrun_reporting] *ERROR* Interrupt
arrived before CRTCs were setup up
--
Sent with MeeGo's ActiveSync support.
David Woodhouse Open Source Technology Centre
[email protected] Intel Corporation
smime.p7s
Description: S/MIME cryptographic signature
_______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
