On Mon, Aug 20, 2012 at 9:48 AM, Jessica Allison <jessica.allison.2012@xxxxxxxxxxx> wrote: > >> Date: Wed, 15 Aug 2012 16:20:12 +0100 >> From: marc.zyngier@xxxxxxx >> To: jessica.allison.2012@xxxxxxxxxxx >> CC: peter.maydell@xxxxxxxxxx; kvmarm@xxxxxxxxxxxxxxxxxxxxx > >> Subject: Re: Boot fails with Unexpected/invalid >> #address-cells/#size-cells in device tree >> >> On 15/08/12 16:13, Jessica Allison wrote: >> > >> > >> >> Date: Wed, 15 Aug 2012 16:06:49 +0100 >> >> From: marc.zyngier@xxxxxxx >> >> To: peter.maydell@xxxxxxxxxx >> >> CC: jessica.allison.2012@xxxxxxxxxxx; kvmarm@xxxxxxxxxxxxxxxxxxxxx >> >> Subject: Re: Boot fails with Unexpected/invalid >> >> #address-cells/#size-cells in device tree >> >> >> >> On 15/08/12 15:50, Peter Maydell wrote: >> >>> On 15 August 2012 15:45, Jessica Allison >> >>> <jessica.allison.2012@xxxxxxxxxxx> wrote: >> >>>> Yes, when I quickly remove that check in the code, it boots up just >> >>>> fine. >> >>> >> >>> Good (although we should probably do a proper fix and make sure >> >>> we handle 64 bit sizes correctly in the following code). >> >>> >> >>>> Is there already a way to boot up a full graphical user environment >> >>>> on the >> >>>> FastModel simulator with the KVM on ARM kernel? >> >>>> I use >> >>>> >> >>>> http://git.kernel.org/?p=linux/kernel/git/maz/arm-platforms.git;a=summary >> >>>> for my kernel and the system boots fine into a root shell. >> >>>> Is there a kernel driver for the LCD by now, and included in the >> >>>> device >> >>>> tree? >> >>> >> >>> Graphics via the pl11x should work fine, but I don't think anybody has >> >>> actually tested because the effect of stacking KVM on top of the Fast >> >>> Model means it would run pretty slowly. Try getting it running with >> >>> plain QEMU on x86 first, then the same kernel config should behave >> >>> the same on KVM. >> >> >> >> Only the A9 tile code has some support for its private pl111. The new >> >> DT >> >> code has no knowledge of the VE baseboard pl111. Not to mention the >> >> HDLCD on TC1/TC2, which doesn't even have a driver. >> >> >> >> As mentioned in an email to Jessica a while ago, this bit is waiting >> >> for >> >> a kind soul who'd actually cares about video output to pick it up... >> > >> > When my kernel boots with the rtsm_ve-cortex_a15x2.dts device tree I see >> > the following error which, I assume, is related to the pl111 not working? >> > >> > clcd-pl11x: probe of 1c1f0000.clcd failed with error -22 >> >> That's one of the many problems. I have a really ugly workaround in one >> of my branches that allows the frame buffer to be used, but that patch >> will not go anywhere near mainline: >> >> >> http://git.us.kernel.org/?p=linux/kernel/git/maz/arm-platforms.git;a=commitdiff;h=de81b299148b84548fe6c563d8d70dcfddf3145c;hp=edf2c53a89824cb5752d42b89997ae8bc96dcff8 >> > > With the current kernel on the Fast Model simulator, which is the fastest > configuration for booting a host+guest? From MMC or NFS or initframfs? Which > would be the best practice configuration? > initramfs is by far the fastest. best practice depends on what you want to do I guess. Both MMC and NFS are useful, but it becomes slow on the fast model, which can be helped somewhat by using virtio, but it's still highly experimental (some patches are floating around there somewhere). _______________________________________________ kvmarm mailing list kvmarm@xxxxxxxxxxxxxxxxxxxxx https://lists.cs.columbia.edu/cucslists/listinfo/kvmarm
