On 03/28/2012 11:59 AM, Benjamin Herrenschmidt wrote: > On Wed, 2012-03-28 at 11:37 +0200, Avi Kivity wrote: > > > > Now I see that x86 just seems to flush everything, which is quite heavy > > > handed considering how often cirrus does it, but maybe it doesn't have a > > > choice (lack of reverse mapping from GPA ?). > > > > We do have a reverse mapping, so we could easily flush just a single > > slot. The reason it hasn't been done is that slot changes are very are > > on x86. They're usually only done by 16-bit software; 32-bit software > > just maps the entire framebuffer BAR and accesses it directly. It's > > also usually done in a tight loop, so flushing everything doesn't have a > > large impact (and with a 20-bit address space, you couldn't cause a > > large impact if you wanted to - memory is all of 256 pages). > > Right ... except that it definitely seems to be happening here with > cirrusfb in the guest kernel :-) > > Every time it does an imageblit it switches the BAR to emulation and > back to direct mapping when the "upload" of the image is complete. That's strange, the cirrus BAR allows the framebuffer and bitblt region to coexist: 0000000000000000-7ffffffffffffffe (prio 0, RW): pci 00000000000a0000-00000000000bffff (prio 1, RW): cirrus-lowmem-container 00000000000a0000-00000000000a7fff (prio 1, RW): alias vga.bank0 @vga.vram 0000000000000000-0000000000007fff 00000000000a0000-00000000000bffff (prio 0, RW): cirrus-low-memory 00000000000a8000-00000000000affff (prio 1, RW): alias vga.bank1 @vga.vram ^ those are continuously flipped when running 16-bit software 0000000000008000-000000000000ffff 00000000000c0000-00000000000dffff (prio 1, RW): pc.rom 00000000000e0000-00000000000fffff (prio 1, R-): isa-bios 00000000fc000000-00000000fdffffff (prio 1, RW): cirrus-pci-bar0 00000000fc000000-00000000fc7fffff (prio 1, RW): vga.vram 00000000fc000000-00000000fc7fffff (prio 0, RW): cirrus-linear-io 00000000fd000000-00000000fd3fffff (prio 0, RW): cirrus-bitblt-mmio ^ the cirrus BAR, write to 0xfc000000 and you hit vga.vram, write to 0xfd000000 and you trigger a bitblt. 00000000feba0000-00000000febbffff (prio 1, RW): e1000-mmio 00000000febf0000-00000000febf0fff (prio 1, RW): cirrus-mmio A guest driver problem perhaps? > The X cirrus driver doesn't seem to trigger that (at least didn't from > my limited testing so far) so it may not be using host data blits ... > I'll have to compare what they do in more details. Or maybe it understands the BAR layout better. -- error compiling committee.c: too many arguments to function -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html